From 1897bdc4d33167e9036460631d1349e59d841f2d Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Thu, 13 Nov 2014 13:46:09 +1100 Subject: mmu_notifier: add mmu_notifier_invalidate_range() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This notifier closes an important gap in the current mmu_notifier implementation, the existing callbacks are called too early or too late to reliably manage a non-CPU TLB. Specifically, invalidate_range_start() is called when all pages are still mapped and invalidate_range_end() when all pages are unmapped and potentially freed. This is fine when the users of the mmu_notifiers manage their own SoftTLB, like KVM does. When the TLB is managed in software it is easy to wipe out entries for a given range and prevent new entries to be established until invalidate_range_end is called. But when the user of mmu_notifiers has to manage a hardware TLB it can still wipe out TLB entries in invalidate_range_start, but it can't make sure that no new TLB entries in the given range are established between invalidate_range_start and invalidate_range_end. To avoid silent data corruption the entries in the non-CPU TLB need to be flushed when the pages are unmapped (at this point in time no _new_ TLB entries can be established in the non-CPU TLB) but not yet freed (as the non-CPU TLB may still have _existing_ entries pointing to the pages about to be freed). To fix this problem we need to catch the moment when the Linux VMM flushes remote TLBs (as a non-CPU TLB is not very CPU TLB), as this is the point in time when the pages are unmapped but _not_ yet freed. The mmu_notifier_invalidate_range() function aims to catch that moment. IOMMU code will be one user of the notifier-callback. Currently this is only the AMD IOMMUv2 driver, but its code is about to be more generalized and converted to a generic IOMMU-API extension to fit the needs of similar functionality in other IOMMUs as well. The current attempt in the AMD IOMMUv2 driver to work around the invalidate_range_start/end() shortcoming is to assign an empty page table to the non-CPU TLB between any invalidata_range_start/end calls. With the empty page-table assigned, every page-table walk to re-fill the non-CPU TLB will cause a page-fault reported to the IOMMU driver via an interrupt, possibly causing interrupt storms. The page-fault handler in the AMD IOMMUv2 driver doesn't handle the fault if an invalidate_range_start/end pair is active, it just reports back SUCCESS to the device and let it refault the page. But existing hardware (newer Radeon GPUs) that makes use of this feature don't re-fault indefinitly, after a certain number of faults for the same address the device enters a failure state and needs to be resetted. To avoid the GPUs entering a failure state we need to get rid of the empty-page-table workaround and use the mmu_notifier_invalidate_range() function introduced with this patch. Signed-off-by: Joerg Roedel Reviewed-by: Andrea Arcangeli Reviewed-by: Jérôme Glisse Cc: Peter Zijlstra Cc: Rik van Riel Cc: Hugh Dickins Cc: Mel Gorman Cc: Johannes Weiner Cc: Jay Cornwall Cc: Oded Gabbay Cc: Suravee Suthikulpanit Cc: Jesse Barnes Cc: David Woodhouse Signed-off-by: Andrew Morton Signed-off-by: Oded Gabbay diff --git a/include/linux/mmu_notifier.h b/include/linux/mmu_notifier.h index 88787bb..1790790 100644 --- a/include/linux/mmu_notifier.h +++ b/include/linux/mmu_notifier.h @@ -242,6 +242,11 @@ static inline void mmu_notifier_invalidate_range_end(struct mm_struct *mm, __mmu_notifier_invalidate_range_end(mm, start, end); } +static inline void mmu_notifier_invalidate_range(struct mm_struct *mm, + unsigned long start, unsigned long end) +{ +} + static inline void mmu_notifier_mm_init(struct mm_struct *mm) { mm->mmu_notifier_mm = NULL; @@ -342,6 +347,11 @@ static inline void mmu_notifier_invalidate_range_end(struct mm_struct *mm, { } +static inline void mmu_notifier_invalidate_range(struct mm_struct *mm, + unsigned long start, unsigned long end) +{ +} + static inline void mmu_notifier_mm_init(struct mm_struct *mm) { } -- cgit v0.10.2 From 34ee645e83b60ae3d5955f70ab9ab9a159136673 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Thu, 13 Nov 2014 13:46:09 +1100 Subject: mmu_notifier: call mmu_notifier_invalidate_range() from VMM MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add calls to the new mmu_notifier_invalidate_range() function to all places in the VMM that need it. Signed-off-by: Joerg Roedel Reviewed-by: Andrea Arcangeli Reviewed-by: Jérôme Glisse Cc: Peter Zijlstra Cc: Rik van Riel Cc: Hugh Dickins Cc: Mel Gorman Cc: Johannes Weiner Cc: Jay Cornwall Cc: Oded Gabbay Cc: Suravee Suthikulpanit Cc: Jesse Barnes Cc: David Woodhouse Signed-off-by: Andrew Morton Signed-off-by: Oded Gabbay diff --git a/include/linux/mmu_notifier.h b/include/linux/mmu_notifier.h index 1790790..966da2b 100644 --- a/include/linux/mmu_notifier.h +++ b/include/linux/mmu_notifier.h @@ -284,6 +284,44 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm) __young; \ }) +#define ptep_clear_flush_notify(__vma, __address, __ptep) \ +({ \ + unsigned long ___addr = __address & PAGE_MASK; \ + struct mm_struct *___mm = (__vma)->vm_mm; \ + pte_t ___pte; \ + \ + ___pte = ptep_clear_flush(__vma, __address, __ptep); \ + mmu_notifier_invalidate_range(___mm, ___addr, \ + ___addr + PAGE_SIZE); \ + \ + ___pte; \ +}) + +#define pmdp_clear_flush_notify(__vma, __haddr, __pmd) \ +({ \ + unsigned long ___haddr = __haddr & HPAGE_PMD_MASK; \ + struct mm_struct *___mm = (__vma)->vm_mm; \ + pmd_t ___pmd; \ + \ + ___pmd = pmdp_clear_flush(__vma, __haddr, __pmd); \ + mmu_notifier_invalidate_range(___mm, ___haddr, \ + ___haddr + HPAGE_PMD_SIZE); \ + \ + ___pmd; \ +}) + +#define pmdp_get_and_clear_notify(__mm, __haddr, __pmd) \ +({ \ + unsigned long ___haddr = __haddr & HPAGE_PMD_MASK; \ + pmd_t ___pmd; \ + \ + ___pmd = pmdp_get_and_clear(__mm, __haddr, __pmd); \ + mmu_notifier_invalidate_range(__mm, ___haddr, \ + ___haddr + HPAGE_PMD_SIZE); \ + \ + ___pmd; \ +}) + /* * set_pte_at_notify() sets the pte _after_ running the notifier. * This is safe to start by updating the secondary MMUs, because the primary MMU @@ -362,6 +400,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm) #define ptep_clear_flush_young_notify ptep_clear_flush_young #define pmdp_clear_flush_young_notify pmdp_clear_flush_young +#define ptep_clear_flush_notify ptep_clear_flush +#define pmdp_clear_flush_notify pmdp_clear_flush +#define pmdp_get_and_clear_notify pmdp_get_and_clear #define set_pte_at_notify set_pte_at #endif /* CONFIG_MMU_NOTIFIER */ diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c index 1d0af8a..bc143cf 100644 --- a/kernel/events/uprobes.c +++ b/kernel/events/uprobes.c @@ -193,7 +193,7 @@ static int __replace_page(struct vm_area_struct *vma, unsigned long addr, } flush_cache_page(vma, addr, pte_pfn(*ptep)); - ptep_clear_flush(vma, addr, ptep); + ptep_clear_flush_notify(vma, addr, ptep); set_pte_at_notify(mm, addr, ptep, mk_pte(kpage, vma->vm_page_prot)); page_remove_rmap(page); diff --git a/mm/fremap.c b/mm/fremap.c index 72b8fa3..9129013 100644 --- a/mm/fremap.c +++ b/mm/fremap.c @@ -37,7 +37,7 @@ static void zap_pte(struct mm_struct *mm, struct vm_area_struct *vma, if (pte_present(pte)) { flush_cache_page(vma, addr, pte_pfn(pte)); - pte = ptep_clear_flush(vma, addr, ptep); + pte = ptep_clear_flush_notify(vma, addr, ptep); page = vm_normal_page(vma, addr, pte); if (page) { if (pte_dirty(pte)) diff --git a/mm/huge_memory.c b/mm/huge_memory.c index de98415..1d89526 100644 --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -1036,7 +1036,7 @@ static int do_huge_pmd_wp_page_fallback(struct mm_struct *mm, goto out_free_pages; VM_BUG_ON_PAGE(!PageHead(page), page); - pmdp_clear_flush(vma, haddr, pmd); + pmdp_clear_flush_notify(vma, haddr, pmd); /* leave pmd empty until pte is filled */ pgtable = pgtable_trans_huge_withdraw(mm, pmd); @@ -1179,7 +1179,7 @@ alloc: pmd_t entry; entry = mk_huge_pmd(new_page, vma->vm_page_prot); entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); - pmdp_clear_flush(vma, haddr, pmd); + pmdp_clear_flush_notify(vma, haddr, pmd); page_add_new_anon_rmap(new_page, vma, haddr); mem_cgroup_commit_charge(new_page, memcg, false); lru_cache_add_active_or_unevictable(new_page, vma); @@ -1512,7 +1512,7 @@ int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd, pmd_t entry; ret = 1; if (!prot_numa) { - entry = pmdp_get_and_clear(mm, addr, pmd); + entry = pmdp_get_and_clear_notify(mm, addr, pmd); if (pmd_numa(entry)) entry = pmd_mknonnuma(entry); entry = pmd_modify(entry, newprot); @@ -1644,6 +1644,7 @@ static int __split_huge_page_splitting(struct page *page, * serialize against split_huge_page*. */ pmdp_splitting_flush(vma, address, pmd); + ret = 1; spin_unlock(ptl); } @@ -2834,7 +2835,7 @@ static void __split_huge_zero_page_pmd(struct vm_area_struct *vma, pmd_t _pmd; int i; - pmdp_clear_flush(vma, haddr, pmd); + pmdp_clear_flush_notify(vma, haddr, pmd); /* leave pmd empty until pte is filled */ pgtable = pgtable_trans_huge_withdraw(mm, pmd); diff --git a/mm/hugetlb.c b/mm/hugetlb.c index 9fd7227..2e6add0 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c @@ -2598,8 +2598,11 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src, } set_huge_pte_at(dst, addr, dst_pte, entry); } else { - if (cow) + if (cow) { huge_ptep_set_wrprotect(src, addr, src_pte); + mmu_notifier_invalidate_range(src, mmun_start, + mmun_end); + } entry = huge_ptep_get(src_pte); ptepage = pte_page(entry); get_page(ptepage); @@ -2899,6 +2902,7 @@ retry_avoidcopy: /* Break COW */ huge_ptep_clear_flush(vma, address, ptep); + mmu_notifier_invalidate_range(mm, mmun_start, mmun_end); set_huge_pte_at(mm, address, ptep, make_huge_pte(vma, new_page, 1)); page_remove_rmap(old_page); @@ -3374,6 +3378,7 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma, * and that page table be reused and filled with junk. */ flush_tlb_range(vma, start, end); + mmu_notifier_invalidate_range(mm, start, end); mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex); mmu_notifier_invalidate_range_end(mm, start, end); diff --git a/mm/ksm.c b/mm/ksm.c index 6b2e337..d247efa 100644 --- a/mm/ksm.c +++ b/mm/ksm.c @@ -892,7 +892,7 @@ static int write_protect_page(struct vm_area_struct *vma, struct page *page, * this assure us that no O_DIRECT can happen after the check * or in the middle of the check. */ - entry = ptep_clear_flush(vma, addr, ptep); + entry = ptep_clear_flush_notify(vma, addr, ptep); /* * Check that no O_DIRECT or similar I/O is in progress on the * page @@ -960,7 +960,7 @@ static int replace_page(struct vm_area_struct *vma, struct page *page, page_add_anon_rmap(kpage, vma, addr); flush_cache_page(vma, addr, pte_pfn(*ptep)); - ptep_clear_flush(vma, addr, ptep); + ptep_clear_flush_notify(vma, addr, ptep); set_pte_at_notify(mm, addr, ptep, mk_pte(kpage, vma->vm_page_prot)); page_remove_rmap(page); diff --git a/mm/memory.c b/mm/memory.c index 3e50383..655fd3d 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -238,6 +238,7 @@ static void tlb_flush_mmu_tlbonly(struct mmu_gather *tlb) { tlb->need_flush = 0; tlb_flush(tlb); + mmu_notifier_invalidate_range(tlb->mm, tlb->start, tlb->end); #ifdef CONFIG_HAVE_RCU_TABLE_FREE tlb_table_flush(tlb); #endif @@ -2234,7 +2235,7 @@ gotten: * seen in the presence of one thread doing SMC and another * thread doing COW. */ - ptep_clear_flush(vma, address, page_table); + ptep_clear_flush_notify(vma, address, page_table); page_add_new_anon_rmap(new_page, vma, address); mem_cgroup_commit_charge(new_page, memcg, false); lru_cache_add_active_or_unevictable(new_page, vma); diff --git a/mm/migrate.c b/mm/migrate.c index 0143995..41945cb 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -1854,7 +1854,7 @@ fail_putback: */ flush_cache_range(vma, mmun_start, mmun_end); page_add_anon_rmap(new_page, vma, mmun_start); - pmdp_clear_flush(vma, mmun_start, pmd); + pmdp_clear_flush_notify(vma, mmun_start, pmd); set_pmd_at(mm, mmun_start, pmd, entry); flush_tlb_range(vma, mmun_start, mmun_end); update_mmu_cache_pmd(vma, address, &entry); @@ -1862,6 +1862,7 @@ fail_putback: if (page_count(page) != 2) { set_pmd_at(mm, mmun_start, pmd, orig_entry); flush_tlb_range(vma, mmun_start, mmun_end); + mmu_notifier_invalidate_range(mm, mmun_start, mmun_end); update_mmu_cache_pmd(vma, address, &entry); page_remove_rmap(new_page); goto fail_putback; diff --git a/mm/rmap.c b/mm/rmap.c index 19886fb..d3eb1e0 100644 --- a/mm/rmap.c +++ b/mm/rmap.c @@ -1378,7 +1378,7 @@ static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount, /* Nuke the page table entry. */ flush_cache_page(vma, address, pte_pfn(*pte)); - pteval = ptep_clear_flush(vma, address, pte); + pteval = ptep_clear_flush_notify(vma, address, pte); /* If nonlinear, store the file page offset in the pte. */ if (page->index != linear_page_index(vma, address)) { -- cgit v0.10.2 From 0f0a327fa12cd55de5e7f8c05a70ac3d047f405e Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Thu, 13 Nov 2014 13:46:09 +1100 Subject: mmu_notifier: add the callback for mmu_notifier_invalidate_range() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Now that the mmu_notifier_invalidate_range() calls are in place, add the callback to allow subsystems to register against it. Signed-off-by: Joerg Roedel Reviewed-by: Andrea Arcangeli Reviewed-by: Jérôme Glisse Cc: Peter Zijlstra Cc: Rik van Riel Cc: Hugh Dickins Cc: Mel Gorman Cc: Johannes Weiner Cc: Jay Cornwall Cc: Oded Gabbay Cc: Suravee Suthikulpanit Cc: Jesse Barnes Cc: David Woodhouse Signed-off-by: Andrew Morton Signed-off-by: Oded Gabbay diff --git a/include/linux/mmu_notifier.h b/include/linux/mmu_notifier.h index 966da2b..94d19f6 100644 --- a/include/linux/mmu_notifier.h +++ b/include/linux/mmu_notifier.h @@ -98,11 +98,11 @@ struct mmu_notifier_ops { /* * invalidate_range_start() and invalidate_range_end() must be * paired and are called only when the mmap_sem and/or the - * locks protecting the reverse maps are held. The subsystem - * must guarantee that no additional references are taken to - * the pages in the range established between the call to - * invalidate_range_start() and the matching call to - * invalidate_range_end(). + * locks protecting the reverse maps are held. If the subsystem + * can't guarantee that no additional references are taken to + * the pages in the range, it has to implement the + * invalidate_range() notifier to remove any references taken + * after invalidate_range_start(). * * Invalidation of multiple concurrent ranges may be * optionally permitted by the driver. Either way the @@ -144,6 +144,29 @@ struct mmu_notifier_ops { void (*invalidate_range_end)(struct mmu_notifier *mn, struct mm_struct *mm, unsigned long start, unsigned long end); + + /* + * invalidate_range() is either called between + * invalidate_range_start() and invalidate_range_end() when the + * VM has to free pages that where unmapped, but before the + * pages are actually freed, or outside of _start()/_end() when + * a (remote) TLB is necessary. + * + * If invalidate_range() is used to manage a non-CPU TLB with + * shared page-tables, it not necessary to implement the + * invalidate_range_start()/end() notifiers, as + * invalidate_range() alread catches the points in time when an + * external TLB range needs to be flushed. + * + * The invalidate_range() function is called under the ptl + * spin-lock and not allowed to sleep. + * + * Note that this function might be called with just a sub-range + * of what was passed to invalidate_range_start()/end(), if + * called between those functions. + */ + void (*invalidate_range)(struct mmu_notifier *mn, struct mm_struct *mm, + unsigned long start, unsigned long end); }; /* @@ -190,6 +213,8 @@ extern void __mmu_notifier_invalidate_range_start(struct mm_struct *mm, unsigned long start, unsigned long end); extern void __mmu_notifier_invalidate_range_end(struct mm_struct *mm, unsigned long start, unsigned long end); +extern void __mmu_notifier_invalidate_range(struct mm_struct *mm, + unsigned long start, unsigned long end); static inline void mmu_notifier_release(struct mm_struct *mm) { @@ -245,6 +270,8 @@ static inline void mmu_notifier_invalidate_range_end(struct mm_struct *mm, static inline void mmu_notifier_invalidate_range(struct mm_struct *mm, unsigned long start, unsigned long end) { + if (mm_has_notifiers(mm)) + __mmu_notifier_invalidate_range(mm, start, end); } static inline void mmu_notifier_mm_init(struct mm_struct *mm) diff --git a/mm/mmu_notifier.c b/mm/mmu_notifier.c index 2c8da98..3b9b3d0 100644 --- a/mm/mmu_notifier.c +++ b/mm/mmu_notifier.c @@ -193,6 +193,16 @@ void __mmu_notifier_invalidate_range_end(struct mm_struct *mm, id = srcu_read_lock(&srcu); hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) { + /* + * Call invalidate_range here too to avoid the need for the + * subsystem of having to register an invalidate_range_end + * call-back when there is invalidate_range already. Usually a + * subsystem registers either invalidate_range_start()/end() or + * invalidate_range(), so this will be no additional overhead + * (besides the pointer check). + */ + if (mn->ops->invalidate_range) + mn->ops->invalidate_range(mn, mm, start, end); if (mn->ops->invalidate_range_end) mn->ops->invalidate_range_end(mn, mm, start, end); } @@ -200,6 +210,21 @@ void __mmu_notifier_invalidate_range_end(struct mm_struct *mm, } EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range_end); +void __mmu_notifier_invalidate_range(struct mm_struct *mm, + unsigned long start, unsigned long end) +{ + struct mmu_notifier *mn; + int id; + + id = srcu_read_lock(&srcu); + hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) { + if (mn->ops->invalidate_range) + mn->ops->invalidate_range(mn, mm, start, end); + } + srcu_read_unlock(&srcu, id); +} +EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range); + static int do_mmu_notifier_register(struct mmu_notifier *mn, struct mm_struct *mm, int take_mmap_sem) -- cgit v0.10.2 From e7cc3dd48c7b0a4b2135a5a2145ad43a8192fa6a Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Thu, 13 Nov 2014 13:46:09 +1100 Subject: iommu/amd: use new invalidate_range mmu-notifier MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Make use of the new invalidate_range mmu_notifier call-back and remove the old logic of assigning an empty page-table between invalidate_range_start and invalidate_range_end. Signed-off-by: Joerg Roedel Tested-by: Oded Gabbay Reviewed-by: Andrea Arcangeli Reviewed-by: Jérôme Glisse Cc: Peter Zijlstra Cc: Rik van Riel Cc: Hugh Dickins Cc: Mel Gorman Cc: Johannes Weiner Cc: Jay Cornwall Cc: Suravee Suthikulpanit Cc: Jesse Barnes Cc: David Woodhouse Signed-off-by: Andrew Morton Signed-off-by: Oded Gabbay diff --git a/drivers/iommu/amd_iommu_v2.c b/drivers/iommu/amd_iommu_v2.c index 90d734b..f724bdc 100644 --- a/drivers/iommu/amd_iommu_v2.c +++ b/drivers/iommu/amd_iommu_v2.c @@ -92,13 +92,6 @@ static spinlock_t state_lock; static struct workqueue_struct *iommu_wq; -/* - * Empty page table - Used between - * mmu_notifier_invalidate_range_start and - * mmu_notifier_invalidate_range_end - */ -static u64 *empty_page_table; - static void free_pasid_states(struct device_state *dev_state); static u16 device_id(struct pci_dev *pdev) @@ -418,46 +411,21 @@ static void mn_invalidate_page(struct mmu_notifier *mn, __mn_flush_page(mn, address); } -static void mn_invalidate_range_start(struct mmu_notifier *mn, - struct mm_struct *mm, - unsigned long start, unsigned long end) -{ - struct pasid_state *pasid_state; - struct device_state *dev_state; - unsigned long flags; - - pasid_state = mn_to_state(mn); - dev_state = pasid_state->device_state; - - spin_lock_irqsave(&pasid_state->lock, flags); - if (pasid_state->mmu_notifier_count == 0) { - amd_iommu_domain_set_gcr3(dev_state->domain, - pasid_state->pasid, - __pa(empty_page_table)); - } - pasid_state->mmu_notifier_count += 1; - spin_unlock_irqrestore(&pasid_state->lock, flags); -} - -static void mn_invalidate_range_end(struct mmu_notifier *mn, - struct mm_struct *mm, - unsigned long start, unsigned long end) +static void mn_invalidate_range(struct mmu_notifier *mn, + struct mm_struct *mm, + unsigned long start, unsigned long end) { struct pasid_state *pasid_state; struct device_state *dev_state; - unsigned long flags; pasid_state = mn_to_state(mn); dev_state = pasid_state->device_state; - spin_lock_irqsave(&pasid_state->lock, flags); - pasid_state->mmu_notifier_count -= 1; - if (pasid_state->mmu_notifier_count == 0) { - amd_iommu_domain_set_gcr3(dev_state->domain, - pasid_state->pasid, - __pa(pasid_state->mm->pgd)); - } - spin_unlock_irqrestore(&pasid_state->lock, flags); + if ((start ^ (end - 1)) < PAGE_SIZE) + amd_iommu_flush_page(dev_state->domain, pasid_state->pasid, + start); + else + amd_iommu_flush_tlb(dev_state->domain, pasid_state->pasid); } static void mn_release(struct mmu_notifier *mn, struct mm_struct *mm) @@ -482,8 +450,7 @@ static struct mmu_notifier_ops iommu_mn = { .release = mn_release, .clear_flush_young = mn_clear_flush_young, .invalidate_page = mn_invalidate_page, - .invalidate_range_start = mn_invalidate_range_start, - .invalidate_range_end = mn_invalidate_range_end, + .invalidate_range = mn_invalidate_range, }; static void set_pri_tag_status(struct pasid_state *pasid_state, @@ -954,18 +921,10 @@ static int __init amd_iommu_v2_init(void) if (iommu_wq == NULL) goto out; - ret = -ENOMEM; - empty_page_table = (u64 *)get_zeroed_page(GFP_KERNEL); - if (empty_page_table == NULL) - goto out_destroy_wq; - amd_iommu_register_ppr_notifier(&ppr_nb); return 0; -out_destroy_wq: - destroy_workqueue(iommu_wq); - out: return ret; } @@ -999,8 +958,6 @@ static void __exit amd_iommu_v2_exit(void) } destroy_workqueue(iommu_wq); - - free_page((unsigned long)empty_page_table); } module_init(amd_iommu_v2_init); -- cgit v0.10.2 From a015c1e92639cd65ebb49350abdf5ad15bce4448 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Mon, 10 Nov 2014 10:57:36 +0200 Subject: iommu/amd: fix accounting of device_state This patch fixes a bug in the accounting of the device_state. In the current code, the device_state was put (decremented) too many times, which sometimes lead to the driver getting stuck permanently in put_device_state_wait(). That happen because the device_state->count would go below zero, which is never supposed to happen. The root cause is that the device_state was decremented in put_pasid_state() and put_pasid_state_wait() but also in all the functions that call those functions. Therefore, the device_state was decremented twice in each of these code paths. The fix is to decouple the device_state accounting from the pasid_state accounting - remove the call to put_device_state() from the put_pasid_state() and the put_pasid_state_wait()) Signed-off-by: Oded Gabbay diff --git a/drivers/iommu/amd_iommu_v2.c b/drivers/iommu/amd_iommu_v2.c index f724bdc..1e6360e 100644 --- a/drivers/iommu/amd_iommu_v2.c +++ b/drivers/iommu/amd_iommu_v2.c @@ -272,10 +272,8 @@ static void free_pasid_state(struct pasid_state *pasid_state) static void put_pasid_state(struct pasid_state *pasid_state) { - if (atomic_dec_and_test(&pasid_state->count)) { - put_device_state(pasid_state->device_state); + if (atomic_dec_and_test(&pasid_state->count)) wake_up(&pasid_state->wq); - } } static void put_pasid_state_wait(struct pasid_state *pasid_state) @@ -284,9 +282,7 @@ static void put_pasid_state_wait(struct pasid_state *pasid_state) prepare_to_wait(&pasid_state->wq, &wait, TASK_UNINTERRUPTIBLE); - if (atomic_dec_and_test(&pasid_state->count)) - put_device_state(pasid_state->device_state); - else + if (!atomic_dec_and_test(&pasid_state->count)) schedule(); finish_wait(&pasid_state->wq, &wait); -- cgit v0.10.2 From 62a7b7fbd08ef745bb51e8728e89125a0ba6327e Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 16 Jan 2014 17:35:44 +0200 Subject: drm/radeon: reduce number of free VMIDs and pipes in KV To support HSA on KV, we need to limit the number of vmids and pipes that are available for radeon's use with KV. This patch reserves VMIDs 8-15 for amdkfd (so radeon can only use VMIDs 0-7) and also makes radeon thinks that KV has only a single MEC with a single pipe in it v3: Use define for static vmid allocation in radeon Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 57a359d..03e160b 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -4684,12 +4684,11 @@ static int cik_mec_init(struct radeon_device *rdev) /* * KV: 2 MEC, 4 Pipes/MEC, 8 Queues/Pipe - 64 Queues total * CI/KB: 1 MEC, 4 Pipes/MEC, 8 Queues/Pipe - 32 Queues total + * Nonetheless, we assign only 1 pipe because all other pipes will + * be handled by KFD */ - if (rdev->family == CHIP_KAVERI) - rdev->mec.num_mec = 2; - else - rdev->mec.num_mec = 1; - rdev->mec.num_pipe = 4; + rdev->mec.num_mec = 1; + rdev->mec.num_pipe = 1; rdev->mec.num_queue = rdev->mec.num_mec * rdev->mec.num_pipe * 8; if (rdev->mec.hpd_eop_obj == NULL) { @@ -4831,28 +4830,24 @@ static int cik_cp_compute_resume(struct radeon_device *rdev) /* init the pipes */ mutex_lock(&rdev->srbm_mutex); - for (i = 0; i < (rdev->mec.num_pipe * rdev->mec.num_mec); i++) { - int me = (i < 4) ? 1 : 2; - int pipe = (i < 4) ? i : (i - 4); - eop_gpu_addr = rdev->mec.hpd_eop_gpu_addr + (i * MEC_HPD_SIZE * 2); + eop_gpu_addr = rdev->mec.hpd_eop_gpu_addr; - cik_srbm_select(rdev, me, pipe, 0, 0); + cik_srbm_select(rdev, 0, 0, 0, 0); - /* write the EOP addr */ - WREG32(CP_HPD_EOP_BASE_ADDR, eop_gpu_addr >> 8); - WREG32(CP_HPD_EOP_BASE_ADDR_HI, upper_32_bits(eop_gpu_addr) >> 8); + /* write the EOP addr */ + WREG32(CP_HPD_EOP_BASE_ADDR, eop_gpu_addr >> 8); + WREG32(CP_HPD_EOP_BASE_ADDR_HI, upper_32_bits(eop_gpu_addr) >> 8); - /* set the VMID assigned */ - WREG32(CP_HPD_EOP_VMID, 0); + /* set the VMID assigned */ + WREG32(CP_HPD_EOP_VMID, 0); + + /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ + tmp = RREG32(CP_HPD_EOP_CONTROL); + tmp &= ~EOP_SIZE_MASK; + tmp |= order_base_2(MEC_HPD_SIZE / 8); + WREG32(CP_HPD_EOP_CONTROL, tmp); - /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ - tmp = RREG32(CP_HPD_EOP_CONTROL); - tmp &= ~EOP_SIZE_MASK; - tmp |= order_base_2(MEC_HPD_SIZE / 8); - WREG32(CP_HPD_EOP_CONTROL, tmp); - } - cik_srbm_select(rdev, 0, 0, 0, 0); mutex_unlock(&rdev->srbm_mutex); /* init the queues. Just two for now. */ @@ -5906,8 +5901,13 @@ int cik_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib) */ int cik_vm_init(struct radeon_device *rdev) { - /* number of VMs */ - rdev->vm_manager.nvm = 16; + /* + * number of VMs + * VMID 0 is reserved for System + * radeon graphics/compute will use VMIDs 1-7 + * amdkfd will use VMIDs 8-15 + */ + rdev->vm_manager.nvm = RADEON_NUM_OF_VMIDS; /* base offset of vram pages */ if (rdev->flags & RADEON_IS_IGP) { u64 tmp = RREG32(MC_VM_FB_OFFSET); diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h index 0c6e1b5..fae4d0c 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -30,6 +30,8 @@ #define CIK_RB_BITMAP_WIDTH_PER_SH 2 #define HAWAII_RB_BITMAP_WIDTH_PER_SH 4 +#define RADEON_NUM_OF_VMIDS 8 + /* DIDT IND registers */ #define DIDT_SQ_CTRL0 0x0 # define DIDT_CTRL_EN (1 << 0) -- cgit v0.10.2 From 28b57b856b635ea0d44f1281e2efdc963c100ea3 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Tue, 11 Feb 2014 18:28:24 +0200 Subject: drm/radeon/cik: Don't touch int of pipes 1-7 amdkfd should set interrupts for pipes 1-7. Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 03e160b..dd8f50f 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -7297,8 +7297,7 @@ static int cik_irq_init(struct radeon_device *rdev) int cik_irq_set(struct radeon_device *rdev) { u32 cp_int_cntl; - u32 cp_m1p0, cp_m1p1, cp_m1p2, cp_m1p3; - u32 cp_m2p0, cp_m2p1, cp_m2p2, cp_m2p3; + u32 cp_m1p0; u32 crtc1 = 0, crtc2 = 0, crtc3 = 0, crtc4 = 0, crtc5 = 0, crtc6 = 0; u32 hpd1, hpd2, hpd3, hpd4, hpd5, hpd6; u32 grbm_int_cntl = 0; @@ -7332,13 +7331,6 @@ int cik_irq_set(struct radeon_device *rdev) dma_cntl1 = RREG32(SDMA0_CNTL + SDMA1_REGISTER_OFFSET) & ~TRAP_ENABLE; cp_m1p0 = RREG32(CP_ME1_PIPE0_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; - cp_m1p1 = RREG32(CP_ME1_PIPE1_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; - cp_m1p2 = RREG32(CP_ME1_PIPE2_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; - cp_m1p3 = RREG32(CP_ME1_PIPE3_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; - cp_m2p0 = RREG32(CP_ME2_PIPE0_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; - cp_m2p1 = RREG32(CP_ME2_PIPE1_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; - cp_m2p2 = RREG32(CP_ME2_PIPE2_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; - cp_m2p3 = RREG32(CP_ME2_PIPE3_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; if (rdev->flags & RADEON_IS_IGP) thermal_int = RREG32_SMC(CG_THERMAL_INT_CTRL) & @@ -7360,33 +7352,6 @@ int cik_irq_set(struct radeon_device *rdev) case 0: cp_m1p0 |= TIME_STAMP_INT_ENABLE; break; - case 1: - cp_m1p1 |= TIME_STAMP_INT_ENABLE; - break; - case 2: - cp_m1p2 |= TIME_STAMP_INT_ENABLE; - break; - case 3: - cp_m1p2 |= TIME_STAMP_INT_ENABLE; - break; - default: - DRM_DEBUG("si_irq_set: sw int cp1 invalid pipe %d\n", ring->pipe); - break; - } - } else if (ring->me == 2) { - switch (ring->pipe) { - case 0: - cp_m2p0 |= TIME_STAMP_INT_ENABLE; - break; - case 1: - cp_m2p1 |= TIME_STAMP_INT_ENABLE; - break; - case 2: - cp_m2p2 |= TIME_STAMP_INT_ENABLE; - break; - case 3: - cp_m2p2 |= TIME_STAMP_INT_ENABLE; - break; default: DRM_DEBUG("si_irq_set: sw int cp1 invalid pipe %d\n", ring->pipe); break; @@ -7403,33 +7368,6 @@ int cik_irq_set(struct radeon_device *rdev) case 0: cp_m1p0 |= TIME_STAMP_INT_ENABLE; break; - case 1: - cp_m1p1 |= TIME_STAMP_INT_ENABLE; - break; - case 2: - cp_m1p2 |= TIME_STAMP_INT_ENABLE; - break; - case 3: - cp_m1p2 |= TIME_STAMP_INT_ENABLE; - break; - default: - DRM_DEBUG("si_irq_set: sw int cp2 invalid pipe %d\n", ring->pipe); - break; - } - } else if (ring->me == 2) { - switch (ring->pipe) { - case 0: - cp_m2p0 |= TIME_STAMP_INT_ENABLE; - break; - case 1: - cp_m2p1 |= TIME_STAMP_INT_ENABLE; - break; - case 2: - cp_m2p2 |= TIME_STAMP_INT_ENABLE; - break; - case 3: - cp_m2p2 |= TIME_STAMP_INT_ENABLE; - break; default: DRM_DEBUG("si_irq_set: sw int cp2 invalid pipe %d\n", ring->pipe); break; @@ -7518,13 +7456,6 @@ int cik_irq_set(struct radeon_device *rdev) WREG32(SDMA0_CNTL + SDMA1_REGISTER_OFFSET, dma_cntl1); WREG32(CP_ME1_PIPE0_INT_CNTL, cp_m1p0); - WREG32(CP_ME1_PIPE1_INT_CNTL, cp_m1p1); - WREG32(CP_ME1_PIPE2_INT_CNTL, cp_m1p2); - WREG32(CP_ME1_PIPE3_INT_CNTL, cp_m1p3); - WREG32(CP_ME2_PIPE0_INT_CNTL, cp_m2p0); - WREG32(CP_ME2_PIPE1_INT_CNTL, cp_m2p1); - WREG32(CP_ME2_PIPE2_INT_CNTL, cp_m2p2); - WREG32(CP_ME2_PIPE3_INT_CNTL, cp_m2p3); WREG32(GRBM_INT_CNTL, grbm_int_cntl); -- cgit v0.10.2 From ebff8453d3a57a2405c4d96d9f9c4f4acc7d4d79 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Tue, 28 Jan 2014 14:43:19 +0200 Subject: drm/radeon: Report doorbell configuration to amdkfd radeon and amdkfd share the doorbell aperture. radeon sets it up, takes the doorbells required for its own rings and reports the setup to amdkfd. radeon reserved doorbells are at the start of the doorbell aperture. Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 908f349..f6cb42e 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -701,6 +701,10 @@ struct radeon_doorbell { int radeon_doorbell_get(struct radeon_device *rdev, u32 *page); void radeon_doorbell_free(struct radeon_device *rdev, u32 doorbell); +void radeon_doorbell_get_kfd_info(struct radeon_device *rdev, + phys_addr_t *aperture_base, + size_t *aperture_size, + size_t *start_offset); /* * IRQS. diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index ea26769..0b41f5e 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -377,6 +377,37 @@ void radeon_doorbell_free(struct radeon_device *rdev, u32 doorbell) __clear_bit(doorbell, rdev->doorbell.used); } +/** + * radeon_doorbell_get_kfd_info - Report doorbell configuration required to + * setup KFD + * + * @rdev: radeon_device pointer + * @aperture_base: output returning doorbell aperture base physical address + * @aperture_size: output returning doorbell aperture size in bytes + * @start_offset: output returning # of doorbell bytes reserved for radeon. + * + * Radeon and the KFD share the doorbell aperture. Radeon sets it up, + * takes doorbells required for its own rings and reports the setup to KFD. + * Radeon reserved doorbells are at the start of the doorbell aperture. + */ +void radeon_doorbell_get_kfd_info(struct radeon_device *rdev, + phys_addr_t *aperture_base, + size_t *aperture_size, + size_t *start_offset) +{ + /* The first num_doorbells are used by radeon. + * KFD takes whatever's left in the aperture. */ + if (rdev->doorbell.size > rdev->doorbell.num_doorbells * sizeof(u32)) { + *aperture_base = rdev->doorbell.base; + *aperture_size = rdev->doorbell.size; + *start_offset = rdev->doorbell.num_doorbells * sizeof(u32); + } else { + *aperture_base = 0; + *aperture_size = 0; + *start_offset = 0; + } +} + /* * radeon_wb_*() * Writeback is the the method by which the the GPU updates special pages -- cgit v0.10.2 From 1c0a46255f8d7daf5b601668836e185fd1294e94 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Mon, 14 Jul 2014 15:36:08 +0300 Subject: drm/radeon: adding synchronization for GRBM GFX Implementing a lock for selecting and accessing shader engines and arrays. This lock will make sure that radeon and amdkfd are not colliding when accessing shader engines and arrays with GRBM_GFX_INDEX register. Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index dd8f50f..2e742a5 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -1563,6 +1563,8 @@ static const u32 godavari_golden_registers[] = static void cik_init_golden_registers(struct radeon_device *rdev) { + /* Some of the registers might be dependent on GRBM_GFX_INDEX */ + mutex_lock(&rdev->grbm_idx_mutex); switch (rdev->family) { case CHIP_BONAIRE: radeon_program_register_sequence(rdev, @@ -1637,6 +1639,7 @@ static void cik_init_golden_registers(struct radeon_device *rdev) default: break; } + mutex_unlock(&rdev->grbm_idx_mutex); } /** @@ -3428,6 +3431,7 @@ static void cik_setup_rb(struct radeon_device *rdev, u32 disabled_rbs = 0; u32 enabled_rbs = 0; + mutex_lock(&rdev->grbm_idx_mutex); for (i = 0; i < se_num; i++) { for (j = 0; j < sh_per_se; j++) { cik_select_se_sh(rdev, i, j); @@ -3439,6 +3443,7 @@ static void cik_setup_rb(struct radeon_device *rdev, } } cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + mutex_unlock(&rdev->grbm_idx_mutex); mask = 1; for (i = 0; i < max_rb_num_per_se * se_num; i++) { @@ -3449,6 +3454,7 @@ static void cik_setup_rb(struct radeon_device *rdev, rdev->config.cik.backend_enable_mask = enabled_rbs; + mutex_lock(&rdev->grbm_idx_mutex); for (i = 0; i < se_num; i++) { cik_select_se_sh(rdev, i, 0xffffffff); data = 0; @@ -3476,6 +3482,7 @@ static void cik_setup_rb(struct radeon_device *rdev, WREG32(PA_SC_RASTER_CONFIG, data); } cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + mutex_unlock(&rdev->grbm_idx_mutex); } /** @@ -3693,6 +3700,12 @@ static void cik_gpu_init(struct radeon_device *rdev) /* set HW defaults for 3D engine */ WREG32(CP_MEQ_THRESHOLDS, MEQ1_START(0x30) | MEQ2_START(0x60)); + mutex_lock(&rdev->grbm_idx_mutex); + /* + * making sure that the following register writes will be broadcasted + * to all the shaders + */ + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); WREG32(SX_DEBUG_1, 0x20); WREG32(TA_CNTL_AUX, 0x00010000); @@ -3748,6 +3761,7 @@ static void cik_gpu_init(struct radeon_device *rdev) WREG32(PA_CL_ENHANCE, CLIP_VTX_REORDER_ENA | NUM_CLIP_SEQ(3)); WREG32(PA_SC_ENHANCE, ENABLE_PA_SC_OUT_OF_ORDER); + mutex_unlock(&rdev->grbm_idx_mutex); udelay(50); } @@ -6068,6 +6082,7 @@ static void cik_wait_for_rlc_serdes(struct radeon_device *rdev) u32 i, j, k; u32 mask; + mutex_lock(&rdev->grbm_idx_mutex); for (i = 0; i < rdev->config.cik.max_shader_engines; i++) { for (j = 0; j < rdev->config.cik.max_sh_per_se; j++) { cik_select_se_sh(rdev, i, j); @@ -6079,6 +6094,7 @@ static void cik_wait_for_rlc_serdes(struct radeon_device *rdev) } } cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + mutex_unlock(&rdev->grbm_idx_mutex); mask = SE_MASTER_BUSY_MASK | GC_MASTER_BUSY | TC0_MASTER_BUSY | TC1_MASTER_BUSY; for (k = 0; k < rdev->usec_timeout; k++) { @@ -6213,10 +6229,12 @@ static int cik_rlc_resume(struct radeon_device *rdev) WREG32(RLC_LB_CNTR_INIT, 0); WREG32(RLC_LB_CNTR_MAX, 0x00008000); + mutex_lock(&rdev->grbm_idx_mutex); cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); WREG32(RLC_LB_INIT_CU_MASK, 0xffffffff); WREG32(RLC_LB_PARAMS, 0x00600408); WREG32(RLC_LB_CNTL, 0x80000004); + mutex_unlock(&rdev->grbm_idx_mutex); WREG32(RLC_MC_CNTL, 0); WREG32(RLC_UCODE_CNTL, 0); @@ -6283,11 +6301,13 @@ static void cik_enable_cgcg(struct radeon_device *rdev, bool enable) tmp = cik_halt_rlc(rdev); + mutex_lock(&rdev->grbm_idx_mutex); cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); WREG32(RLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); WREG32(RLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); tmp2 = BPM_ADDR_MASK | CGCG_OVERRIDE_0 | CGLS_ENABLE; WREG32(RLC_SERDES_WR_CTRL, tmp2); + mutex_unlock(&rdev->grbm_idx_mutex); cik_update_rlc(rdev, tmp); @@ -6329,11 +6349,13 @@ static void cik_enable_mgcg(struct radeon_device *rdev, bool enable) tmp = cik_halt_rlc(rdev); + mutex_lock(&rdev->grbm_idx_mutex); cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); WREG32(RLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); WREG32(RLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); data = BPM_ADDR_MASK | MGCG_OVERRIDE_0; WREG32(RLC_SERDES_WR_CTRL, data); + mutex_unlock(&rdev->grbm_idx_mutex); cik_update_rlc(rdev, tmp); @@ -6377,11 +6399,13 @@ static void cik_enable_mgcg(struct radeon_device *rdev, bool enable) tmp = cik_halt_rlc(rdev); + mutex_lock(&rdev->grbm_idx_mutex); cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); WREG32(RLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); WREG32(RLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); data = BPM_ADDR_MASK | MGCG_OVERRIDE_1; WREG32(RLC_SERDES_WR_CTRL, data); + mutex_unlock(&rdev->grbm_idx_mutex); cik_update_rlc(rdev, tmp); } @@ -6810,10 +6834,12 @@ static u32 cik_get_cu_active_bitmap(struct radeon_device *rdev, u32 se, u32 sh) u32 mask = 0, tmp, tmp1; int i; + mutex_lock(&rdev->grbm_idx_mutex); cik_select_se_sh(rdev, se, sh); tmp = RREG32(CC_GC_SHADER_ARRAY_CONFIG); tmp1 = RREG32(GC_USER_SHADER_ARRAY_CONFIG); cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + mutex_unlock(&rdev->grbm_idx_mutex); tmp &= 0xffff0000; diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index f6cb42e..e95718f0 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -2397,6 +2397,8 @@ struct radeon_device { struct radeon_atcs atcs; /* srbm instance registers */ struct mutex srbm_mutex; + /* GRBM index mutex. Protects concurrents access to GRBM index */ + struct mutex grbm_idx_mutex; /* clock, powergating flags */ u32 cg_flags; u32 pg_flags; diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index 0b41f5e..ae87310 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -1303,6 +1303,7 @@ int radeon_device_init(struct radeon_device *rdev, mutex_init(&rdev->pm.mutex); mutex_init(&rdev->gpu_clock_mutex); mutex_init(&rdev->srbm_mutex); + mutex_init(&rdev->grbm_idx_mutex); init_rwsem(&rdev->pm.mclk_lock); init_rwsem(&rdev->exclusive_lock); init_waitqueue_head(&rdev->irq.vblank_queue); -- cgit v0.10.2 From e28740ece34d314002b1ddfa14e8fb7c7b909489 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Tue, 15 Jul 2014 13:53:32 +0300 Subject: drm/radeon: Add radeon <--> amdkfd interface This patch adds the interface between the radeon driver and the amdkfd driver. The interface implementation is contained in radeon_kfd.c and radeon_kfd.h. The interface itself is represented by a pointer to struct kfd_dev. The pointer is located inside radeon_device structure. All the register accesses that amdkfd need are done using this interface. This allows us to avoid direct register accesses in amdkfd proper, while also avoiding locking between amdkfd and radeon. The single exception is the doorbells that are used in both of the drivers. However, because they are located in separate pci bar pages, the danger of sharing registers between the drivers is minimal. Having said that, we are planning to move the doorbells as well to radeon. v3: Add interface for sa manager init and fini. The init function will allocate a buffer on system memory and pin it to the GART address space via the radeon sa manager. All mappings of buffers to GART address space are done via the radeon sa manager. The interface of allocate memory will use the radeon sa manager to sub allocate from the single buffer that was allocated during the init function. Change lower_32/upper_32 calls to use linux macros Add documentation for the interface v4: Change ptr field type in kgd_mem from uint32_t* to void* to match to type that is returned by radeon_sa_bo_cpu_addr v5: Change format of mqd structure to work with latest KV firmware Add support for AQL queues creation to enable working with open-source HSA runtime. Move generic kfd-->kgd interface and other generic kgd definitions to a generic header file that will be used by AMD's radeon and amdgpu drivers Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h new file mode 100644 index 0000000..9c729dd --- /dev/null +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -0,0 +1,185 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file defines the private interface between the + * AMD kernel graphics drivers and the AMD KFD. + */ + +#ifndef KGD_KFD_INTERFACE_H_INCLUDED +#define KGD_KFD_INTERFACE_H_INCLUDED + +#include + +struct pci_dev; + +#define KFD_INTERFACE_VERSION 1 + +struct kfd_dev; +struct kgd_dev; + +struct kgd_mem; + +enum kgd_memory_pool { + KGD_POOL_SYSTEM_CACHEABLE = 1, + KGD_POOL_SYSTEM_WRITECOMBINE = 2, + KGD_POOL_FRAMEBUFFER = 3, +}; + +struct kgd2kfd_shared_resources { + /* Bit n == 1 means VMID n is available for KFD. */ + unsigned int compute_vmid_bitmap; + + /* Compute pipes are counted starting from MEC0/pipe0 as 0. */ + unsigned int first_compute_pipe; + + /* Number of MEC pipes available for KFD. */ + unsigned int compute_pipe_count; + + /* Base address of doorbell aperture. */ + phys_addr_t doorbell_physical_address; + + /* Size in bytes of doorbell aperture. */ + size_t doorbell_aperture_size; + + /* Number of bytes at start of aperture reserved for KGD. */ + size_t doorbell_start_offset; +}; + +/** + * struct kgd2kfd_calls + * + * @exit: Notifies amdkfd that kgd module is unloaded + * + * @probe: Notifies amdkfd about a probe done on a device in the kgd driver. + * + * @device_init: Initialize the newly probed device (if it is a device that + * amdkfd supports) + * + * @device_exit: Notifies amdkfd about a removal of a kgd device + * + * @suspend: Notifies amdkfd about a suspend action done to a kgd device + * + * @resume: Notifies amdkfd about a resume action done to a kgd device + * + * This structure contains function callback pointers so the kgd driver + * will notify to the amdkfd about certain status changes. + * + */ +struct kgd2kfd_calls { + void (*exit)(void); + struct kfd_dev* (*probe)(struct kgd_dev *kgd, struct pci_dev *pdev); + bool (*device_init)(struct kfd_dev *kfd, + const struct kgd2kfd_shared_resources *gpu_resources); + void (*device_exit)(struct kfd_dev *kfd); + void (*interrupt)(struct kfd_dev *kfd, const void *ih_ring_entry); + void (*suspend)(struct kfd_dev *kfd); + int (*resume)(struct kfd_dev *kfd); +}; + +/** + * struct kfd2kgd_calls + * + * @init_sa_manager: Initialize an instance of the sa manager, used by + * amdkfd for all system memory allocations that are mapped to the GART + * address space + * + * @fini_sa_manager: Releases all memory allocations for amdkfd that are + * handled by kgd sa manager + * + * @allocate_mem: Allocate a buffer from amdkfd's sa manager. The buffer can + * be used for mqds, hpds, kernel queue, fence and runlists + * + * @free_mem: Frees a buffer that was allocated by amdkfd's sa manager + * + * @get_vmem_size: Retrieves (physical) size of VRAM + * + * @get_gpu_clock_counter: Retrieves GPU clock counter + * + * @get_max_engine_clock_in_mhz: Retrieves maximum GPU clock in MHz + * + * @program_sh_mem_settings: A function that should initiate the memory + * properties such as main aperture memory type (cache / non cached) and + * secondary aperture base address, size and memory type. + * This function is used only for no cp scheduling mode. + * + * @set_pasid_vmid_mapping: Exposes pasid/vmid pair to the H/W for no cp + * scheduling mode. Only used for no cp scheduling mode. + * + * @init_memory: Initializes memory apertures to fixed base/limit address + * and non cached memory types. + * + * @init_pipeline: Initialized the compute pipelines. + * + * @hqd_load: Loads the mqd structure to a H/W hqd slot. used only for no cp + * sceduling mode. + * + * @hqd_is_occupies: Checks if a hqd slot is occupied. + * + * @hqd_destroy: Destructs and preempts the queue assigned to that hqd slot. + * + * This structure contains function pointers to services that the kgd driver + * provides to amdkfd driver. + * + */ +struct kfd2kgd_calls { + /* Memory management. */ + int (*init_sa_manager)(struct kgd_dev *kgd, unsigned int size); + void (*fini_sa_manager)(struct kgd_dev *kgd); + int (*allocate_mem)(struct kgd_dev *kgd, size_t size, size_t alignment, + enum kgd_memory_pool pool, struct kgd_mem **mem); + + void (*free_mem)(struct kgd_dev *kgd, struct kgd_mem *mem); + + uint64_t (*get_vmem_size)(struct kgd_dev *kgd); + uint64_t (*get_gpu_clock_counter)(struct kgd_dev *kgd); + + uint32_t (*get_max_engine_clock_in_mhz)(struct kgd_dev *kgd); + + /* Register access functions */ + void (*program_sh_mem_settings)(struct kgd_dev *kgd, uint32_t vmid, + uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, + uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases); + + int (*set_pasid_vmid_mapping)(struct kgd_dev *kgd, unsigned int pasid, + unsigned int vmid); + + int (*init_memory)(struct kgd_dev *kgd); + int (*init_pipeline)(struct kgd_dev *kgd, uint32_t pipe_id, + uint32_t hpd_size, uint64_t hpd_gpu_addr); + + int (*hqd_load)(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, + uint32_t queue_id, uint32_t __user *wptr); + + bool (*hqd_is_occupies)(struct kgd_dev *kgd, uint64_t queue_address, + uint32_t pipe_id, uint32_t queue_id); + + int (*hqd_destroy)(struct kgd_dev *kgd, uint32_t reset_type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id); +}; + +bool kgd2kfd_init(unsigned interface_version, + const struct kfd2kgd_calls *f2g, + const struct kgd2kfd_calls **g2f); + +#endif /* KGD_KFD_INTERFACE_H_INCLUDED */ diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index d01b879..bad6caa 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -104,6 +104,7 @@ radeon-y += \ radeon_vce.o \ vce_v1_0.o \ vce_v2_0.o \ + radeon_kfd.o radeon-$(CONFIG_COMPAT) += radeon_ioc32.o radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 2e742a5..bce73b6 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -32,6 +32,7 @@ #include "cik_blit_shaders.h" #include "radeon_ucode.h" #include "clearstate_ci.h" +#include "radeon_kfd.h" MODULE_FIRMWARE("radeon/BONAIRE_pfp.bin"); MODULE_FIRMWARE("radeon/BONAIRE_me.bin"); @@ -7798,6 +7799,10 @@ restart_ih: while (rptr != wptr) { /* wptr/rptr are in bytes! */ ring_index = rptr / 4; + + radeon_kfd_interrupt(rdev, + (const void *) &rdev->ih.ring[ring_index]); + src_id = le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff; src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff; ring_id = le32_to_cpu(rdev->ih.ring[ring_index + 2]) & 0xff; @@ -8487,6 +8492,10 @@ static int cik_startup(struct radeon_device *rdev) if (r) return r; + r = radeon_kfd_resume(rdev); + if (r) + return r; + return 0; } @@ -8535,6 +8544,7 @@ int cik_resume(struct radeon_device *rdev) */ int cik_suspend(struct radeon_device *rdev) { + radeon_kfd_suspend(rdev); radeon_pm_suspend(rdev); dce6_audio_fini(rdev); radeon_vm_manager_fini(rdev); diff --git a/drivers/gpu/drm/radeon/cik_reg.h b/drivers/gpu/drm/radeon/cik_reg.h index ca1bb61..79c45e8 100644 --- a/drivers/gpu/drm/radeon/cik_reg.h +++ b/drivers/gpu/drm/radeon/cik_reg.h @@ -147,4 +147,140 @@ #define CIK_LB_DESKTOP_HEIGHT 0x6b0c +#define CP_HQD_IQ_RPTR 0xC970u +#define AQL_ENABLE (1U << 0) + +#define IDLE (1 << 2) + +struct cik_mqd { + uint32_t header; + uint32_t compute_dispatch_initiator; + uint32_t compute_dim_x; + uint32_t compute_dim_y; + uint32_t compute_dim_z; + uint32_t compute_start_x; + uint32_t compute_start_y; + uint32_t compute_start_z; + uint32_t compute_num_thread_x; + uint32_t compute_num_thread_y; + uint32_t compute_num_thread_z; + uint32_t compute_pipelinestat_enable; + uint32_t compute_perfcount_enable; + uint32_t compute_pgm_lo; + uint32_t compute_pgm_hi; + uint32_t compute_tba_lo; + uint32_t compute_tba_hi; + uint32_t compute_tma_lo; + uint32_t compute_tma_hi; + uint32_t compute_pgm_rsrc1; + uint32_t compute_pgm_rsrc2; + uint32_t compute_vmid; + uint32_t compute_resource_limits; + uint32_t compute_static_thread_mgmt_se0; + uint32_t compute_static_thread_mgmt_se1; + uint32_t compute_tmpring_size; + uint32_t compute_static_thread_mgmt_se2; + uint32_t compute_static_thread_mgmt_se3; + uint32_t compute_restart_x; + uint32_t compute_restart_y; + uint32_t compute_restart_z; + uint32_t compute_thread_trace_enable; + uint32_t compute_misc_reserved; + uint32_t compute_user_data_0; + uint32_t compute_user_data_1; + uint32_t compute_user_data_2; + uint32_t compute_user_data_3; + uint32_t compute_user_data_4; + uint32_t compute_user_data_5; + uint32_t compute_user_data_6; + uint32_t compute_user_data_7; + uint32_t compute_user_data_8; + uint32_t compute_user_data_9; + uint32_t compute_user_data_10; + uint32_t compute_user_data_11; + uint32_t compute_user_data_12; + uint32_t compute_user_data_13; + uint32_t compute_user_data_14; + uint32_t compute_user_data_15; + uint32_t cp_compute_csinvoc_count_lo; + uint32_t cp_compute_csinvoc_count_hi; + uint32_t cp_mqd_base_addr_lo; + uint32_t cp_mqd_base_addr_hi; + uint32_t cp_hqd_active; + uint32_t cp_hqd_vmid; + uint32_t cp_hqd_persistent_state; + uint32_t cp_hqd_pipe_priority; + uint32_t cp_hqd_queue_priority; + uint32_t cp_hqd_quantum; + uint32_t cp_hqd_pq_base_lo; + uint32_t cp_hqd_pq_base_hi; + uint32_t cp_hqd_pq_rptr; + uint32_t cp_hqd_pq_rptr_report_addr_lo; + uint32_t cp_hqd_pq_rptr_report_addr_hi; + uint32_t cp_hqd_pq_wptr_poll_addr_lo; + uint32_t cp_hqd_pq_wptr_poll_addr_hi; + uint32_t cp_hqd_pq_doorbell_control; + uint32_t cp_hqd_pq_wptr; + uint32_t cp_hqd_pq_control; + uint32_t cp_hqd_ib_base_addr_lo; + uint32_t cp_hqd_ib_base_addr_hi; + uint32_t cp_hqd_ib_rptr; + uint32_t cp_hqd_ib_control; + uint32_t cp_hqd_iq_timer; + uint32_t cp_hqd_iq_rptr; + uint32_t cp_hqd_dequeue_request; + uint32_t cp_hqd_dma_offload; + uint32_t cp_hqd_sema_cmd; + uint32_t cp_hqd_msg_type; + uint32_t cp_hqd_atomic0_preop_lo; + uint32_t cp_hqd_atomic0_preop_hi; + uint32_t cp_hqd_atomic1_preop_lo; + uint32_t cp_hqd_atomic1_preop_hi; + uint32_t cp_hqd_hq_status0; + uint32_t cp_hqd_hq_control0; + uint32_t cp_mqd_control; + uint32_t cp_mqd_query_time_lo; + uint32_t cp_mqd_query_time_hi; + uint32_t cp_mqd_connect_start_time_lo; + uint32_t cp_mqd_connect_start_time_hi; + uint32_t cp_mqd_connect_end_time_lo; + uint32_t cp_mqd_connect_end_time_hi; + uint32_t cp_mqd_connect_end_wf_count; + uint32_t cp_mqd_connect_end_pq_rptr; + uint32_t cp_mqd_connect_end_pq_wptr; + uint32_t cp_mqd_connect_end_ib_rptr; + uint32_t reserved_96; + uint32_t reserved_97; + uint32_t reserved_98; + uint32_t reserved_99; + uint32_t iqtimer_pkt_header; + uint32_t iqtimer_pkt_dw0; + uint32_t iqtimer_pkt_dw1; + uint32_t iqtimer_pkt_dw2; + uint32_t iqtimer_pkt_dw3; + uint32_t iqtimer_pkt_dw4; + uint32_t iqtimer_pkt_dw5; + uint32_t iqtimer_pkt_dw6; + uint32_t reserved_108; + uint32_t reserved_109; + uint32_t reserved_110; + uint32_t reserved_111; + uint32_t queue_doorbell_id0; + uint32_t queue_doorbell_id1; + uint32_t queue_doorbell_id2; + uint32_t queue_doorbell_id3; + uint32_t queue_doorbell_id4; + uint32_t queue_doorbell_id5; + uint32_t queue_doorbell_id6; + uint32_t queue_doorbell_id7; + uint32_t queue_doorbell_id8; + uint32_t queue_doorbell_id9; + uint32_t queue_doorbell_id10; + uint32_t queue_doorbell_id11; + uint32_t queue_doorbell_id12; + uint32_t queue_doorbell_id13; + uint32_t queue_doorbell_id14; + uint32_t queue_doorbell_id15; +}; + #endif diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h index fae4d0c..068cbb0 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -1139,6 +1139,9 @@ #define SH_MEM_ALIGNMENT_MODE_UNALIGNED 3 #define DEFAULT_MTYPE(x) ((x) << 4) #define APE1_MTYPE(x) ((x) << 7) +/* valid for both DEFAULT_MTYPE and APE1_MTYPE */ +#define MTYPE_CACHED 0 +#define MTYPE_NONCACHED 3 #define SX_DEBUG_1 0x9060 @@ -1449,6 +1452,16 @@ #define CP_HQD_ACTIVE 0xC91C #define CP_HQD_VMID 0xC920 +#define CP_HQD_PERSISTENT_STATE 0xC924u +#define DEFAULT_CP_HQD_PERSISTENT_STATE (0x33U << 8) + +#define CP_HQD_PIPE_PRIORITY 0xC928u +#define CP_HQD_QUEUE_PRIORITY 0xC92Cu +#define CP_HQD_QUANTUM 0xC930u +#define QUANTUM_EN 1U +#define QUANTUM_SCALE_1MS (1U << 4) +#define QUANTUM_DURATION(x) ((x) << 8) + #define CP_HQD_PQ_BASE 0xC934 #define CP_HQD_PQ_BASE_HI 0xC938 #define CP_HQD_PQ_RPTR 0xC93C @@ -1476,12 +1489,32 @@ #define PRIV_STATE (1 << 30) #define KMD_QUEUE (1 << 31) -#define CP_HQD_DEQUEUE_REQUEST 0xC974 +#define CP_HQD_IB_BASE_ADDR 0xC95Cu +#define CP_HQD_IB_BASE_ADDR_HI 0xC960u +#define CP_HQD_IB_RPTR 0xC964u +#define CP_HQD_IB_CONTROL 0xC968u +#define IB_ATC_EN (1U << 23) +#define DEFAULT_MIN_IB_AVAIL_SIZE (3U << 20) + +#define CP_HQD_DEQUEUE_REQUEST 0xC974 +#define DEQUEUE_REQUEST_DRAIN 1 +#define DEQUEUE_REQUEST_RESET 2 #define CP_MQD_CONTROL 0xC99C #define MQD_VMID(x) ((x) << 0) #define MQD_VMID_MASK (0xf << 0) +#define CP_HQD_SEMA_CMD 0xC97Cu +#define CP_HQD_MSG_TYPE 0xC980u +#define CP_HQD_ATOMIC0_PREOP_LO 0xC984u +#define CP_HQD_ATOMIC0_PREOP_HI 0xC988u +#define CP_HQD_ATOMIC1_PREOP_LO 0xC98Cu +#define CP_HQD_ATOMIC1_PREOP_HI 0xC990u +#define CP_HQD_HQ_SCHEDULER0 0xC994u +#define CP_HQD_HQ_SCHEDULER1 0xC998u + +#define SH_STATIC_MEM_CONFIG 0x9604u + #define DB_RENDER_CONTROL 0x28000 #define PA_SC_RASTER_CONFIG 0x28350 @@ -2071,4 +2104,20 @@ #define VCE_CMD_IB_AUTO 0x00000005 #define VCE_CMD_SEMAPHORE 0x00000006 +#define ATC_VMID0_PASID_MAPPING 0x339Cu +#define ATC_VMID_PASID_MAPPING_UPDATE_STATUS 0x3398u +#define ATC_VMID_PASID_MAPPING_VALID (1U << 31) + +#define ATC_VM_APERTURE0_CNTL 0x3310u +#define ATS_ACCESS_MODE_NEVER 0 +#define ATS_ACCESS_MODE_ALWAYS 1 + +#define ATC_VM_APERTURE0_CNTL2 0x3318u +#define ATC_VM_APERTURE0_HIGH_ADDR 0x3308u +#define ATC_VM_APERTURE0_LOW_ADDR 0x3300u +#define ATC_VM_APERTURE1_CNTL 0x3314u +#define ATC_VM_APERTURE1_CNTL2 0x331Cu +#define ATC_VM_APERTURE1_HIGH_ADDR 0x330Cu +#define ATC_VM_APERTURE1_LOW_ADDR 0x3304u + #endif diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index e95718f0..1f61ff0 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -2411,6 +2411,10 @@ struct radeon_device { u64 vram_pin_size; u64 gart_pin_size; + /* amdkfd interface */ + struct kfd_dev *kfd; + struct radeon_sa_manager kfd_bo; + struct mutex mn_lock; DECLARE_HASHTABLE(mn_hash, 7); }; diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c index dcffa30..4f50fb0 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.c +++ b/drivers/gpu/drm/radeon/radeon_drv.c @@ -41,6 +41,8 @@ #include #include "drm_crtc_helper.h" +#include "radeon_kfd.h" + /* * KMS wrapper. * - 2.0.0 - initial interface @@ -654,12 +656,15 @@ static int __init radeon_init(void) #endif } + radeon_kfd_init(); + /* let modprobe override vga console setting */ return drm_pci_init(driver, pdriver); } static void __exit radeon_exit(void) { + radeon_kfd_fini(); drm_pci_exit(driver, pdriver); radeon_unregister_atpx_handler(); } diff --git a/drivers/gpu/drm/radeon/radeon_kfd.c b/drivers/gpu/drm/radeon/radeon_kfd.c new file mode 100644 index 0000000..065d020 --- /dev/null +++ b/drivers/gpu/drm/radeon/radeon_kfd.c @@ -0,0 +1,563 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include +#include +#include +#include "radeon.h" +#include "cikd.h" +#include "cik_reg.h" +#include "radeon_kfd.h" + +#define CIK_PIPE_PER_MEC (4) + +struct kgd_mem { + struct radeon_sa_bo *sa_bo; + uint64_t gpu_addr; + void *ptr; +}; + +static int init_sa_manager(struct kgd_dev *kgd, unsigned int size); +static void fini_sa_manager(struct kgd_dev *kgd); + +static int allocate_mem(struct kgd_dev *kgd, size_t size, size_t alignment, + enum kgd_memory_pool pool, struct kgd_mem **mem); + +static void free_mem(struct kgd_dev *kgd, struct kgd_mem *mem); + +static uint64_t get_vmem_size(struct kgd_dev *kgd); +static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd); + +static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd); + +/* + * Register access functions + */ + +static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, + uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, + uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases); + +static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, + unsigned int vmid); + +static int kgd_init_memory(struct kgd_dev *kgd); + +static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id, + uint32_t hpd_size, uint64_t hpd_gpu_addr); + +static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, + uint32_t queue_id, uint32_t __user *wptr); + +static bool kgd_hqd_is_occupies(struct kgd_dev *kgd, uint64_t queue_address, + uint32_t pipe_id, uint32_t queue_id); + +static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id); + +static const struct kfd2kgd_calls kfd2kgd = { + .init_sa_manager = init_sa_manager, + .fini_sa_manager = fini_sa_manager, + .allocate_mem = allocate_mem, + .free_mem = free_mem, + .get_vmem_size = get_vmem_size, + .get_gpu_clock_counter = get_gpu_clock_counter, + .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, + .program_sh_mem_settings = kgd_program_sh_mem_settings, + .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, + .init_memory = kgd_init_memory, + .init_pipeline = kgd_init_pipeline, + .hqd_load = kgd_hqd_load, + .hqd_is_occupies = kgd_hqd_is_occupies, + .hqd_destroy = kgd_hqd_destroy, +}; + +static const struct kgd2kfd_calls *kgd2kfd; + +bool radeon_kfd_init(void) +{ + bool (*kgd2kfd_init_p)(unsigned, const struct kfd2kgd_calls*, + const struct kgd2kfd_calls**); + + kgd2kfd_init_p = symbol_request(kgd2kfd_init); + + if (kgd2kfd_init_p == NULL) + return false; + + if (!kgd2kfd_init_p(KFD_INTERFACE_VERSION, &kfd2kgd, &kgd2kfd)) { + symbol_put(kgd2kfd_init); + kgd2kfd = NULL; + + return false; + } + + return true; +} + +void radeon_kfd_fini(void) +{ + if (kgd2kfd) { + kgd2kfd->exit(); + symbol_put(kgd2kfd_init); + } +} + +void radeon_kfd_device_probe(struct radeon_device *rdev) +{ + if (kgd2kfd) + rdev->kfd = kgd2kfd->probe((struct kgd_dev *)rdev, rdev->pdev); +} + +void radeon_kfd_device_init(struct radeon_device *rdev) +{ + if (rdev->kfd) { + struct kgd2kfd_shared_resources gpu_resources = { + .compute_vmid_bitmap = 0xFF00, + + .first_compute_pipe = 1, + .compute_pipe_count = 8 - 1, + }; + + radeon_doorbell_get_kfd_info(rdev, + &gpu_resources.doorbell_physical_address, + &gpu_resources.doorbell_aperture_size, + &gpu_resources.doorbell_start_offset); + + kgd2kfd->device_init(rdev->kfd, &gpu_resources); + } +} + +void radeon_kfd_device_fini(struct radeon_device *rdev) +{ + if (rdev->kfd) { + kgd2kfd->device_exit(rdev->kfd); + rdev->kfd = NULL; + } +} + +void radeon_kfd_interrupt(struct radeon_device *rdev, const void *ih_ring_entry) +{ + if (rdev->kfd) + kgd2kfd->interrupt(rdev->kfd, ih_ring_entry); +} + +void radeon_kfd_suspend(struct radeon_device *rdev) +{ + if (rdev->kfd) + kgd2kfd->suspend(rdev->kfd); +} + +int radeon_kfd_resume(struct radeon_device *rdev) +{ + int r = 0; + + if (rdev->kfd) + r = kgd2kfd->resume(rdev->kfd); + + return r; +} + +static u32 pool_to_domain(enum kgd_memory_pool p) +{ + switch (p) { + case KGD_POOL_FRAMEBUFFER: return RADEON_GEM_DOMAIN_VRAM; + default: return RADEON_GEM_DOMAIN_GTT; + } +} + +static int init_sa_manager(struct kgd_dev *kgd, unsigned int size) +{ + struct radeon_device *rdev = (struct radeon_device *)kgd; + int r; + + BUG_ON(kgd == NULL); + + r = radeon_sa_bo_manager_init(rdev, &rdev->kfd_bo, + size, + RADEON_GPU_PAGE_SIZE, + RADEON_GEM_DOMAIN_GTT, + RADEON_GEM_GTT_WC); + + if (r) + return r; + + r = radeon_sa_bo_manager_start(rdev, &rdev->kfd_bo); + if (r) + radeon_sa_bo_manager_fini(rdev, &rdev->kfd_bo); + + return r; +} + +static void fini_sa_manager(struct kgd_dev *kgd) +{ + struct radeon_device *rdev = (struct radeon_device *)kgd; + + BUG_ON(kgd == NULL); + + radeon_sa_bo_manager_suspend(rdev, &rdev->kfd_bo); + radeon_sa_bo_manager_fini(rdev, &rdev->kfd_bo); +} + +static int allocate_mem(struct kgd_dev *kgd, size_t size, size_t alignment, + enum kgd_memory_pool pool, struct kgd_mem **mem) +{ + struct radeon_device *rdev = (struct radeon_device *)kgd; + u32 domain; + int r; + + BUG_ON(kgd == NULL); + + domain = pool_to_domain(pool); + if (domain != RADEON_GEM_DOMAIN_GTT) { + dev_err(rdev->dev, + "Only allowed to allocate gart memory for kfd\n"); + return -EINVAL; + } + + *mem = kmalloc(sizeof(struct kgd_mem), GFP_KERNEL); + if ((*mem) == NULL) + return -ENOMEM; + + r = radeon_sa_bo_new(rdev, &rdev->kfd_bo, &(*mem)->sa_bo, size, + alignment); + if (r) { + dev_err(rdev->dev, "failed to get memory for kfd (%d)\n", r); + return r; + } + + (*mem)->ptr = radeon_sa_bo_cpu_addr((*mem)->sa_bo); + (*mem)->gpu_addr = radeon_sa_bo_gpu_addr((*mem)->sa_bo); + + return 0; +} + +static void free_mem(struct kgd_dev *kgd, struct kgd_mem *mem) +{ + struct radeon_device *rdev = (struct radeon_device *)kgd; + + BUG_ON(kgd == NULL); + + radeon_sa_bo_free(rdev, &mem->sa_bo, NULL); + kfree(mem); +} + +static uint64_t get_vmem_size(struct kgd_dev *kgd) +{ + struct radeon_device *rdev = (struct radeon_device *)kgd; + + BUG_ON(kgd == NULL); + + return rdev->mc.real_vram_size; +} + +static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd) +{ + struct radeon_device *rdev = (struct radeon_device *)kgd; + + return rdev->asic->get_gpu_clock_counter(rdev); +} + +static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd) +{ + struct radeon_device *rdev = (struct radeon_device *)kgd; + + /* The sclk is in quantas of 10kHz */ + return rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk / 100; +} + +static inline struct radeon_device *get_radeon_device(struct kgd_dev *kgd) +{ + return (struct radeon_device *)kgd; +} + +static void write_register(struct kgd_dev *kgd, uint32_t offset, uint32_t value) +{ + struct radeon_device *rdev = get_radeon_device(kgd); + + writel(value, (void __iomem *)(rdev->rmmio + offset)); +} + +static uint32_t read_register(struct kgd_dev *kgd, uint32_t offset) +{ + struct radeon_device *rdev = get_radeon_device(kgd); + + return readl((void __iomem *)(rdev->rmmio + offset)); +} + +static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe, + uint32_t queue, uint32_t vmid) +{ + struct radeon_device *rdev = get_radeon_device(kgd); + uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue); + + mutex_lock(&rdev->srbm_mutex); + write_register(kgd, SRBM_GFX_CNTL, value); +} + +static void unlock_srbm(struct kgd_dev *kgd) +{ + struct radeon_device *rdev = get_radeon_device(kgd); + + write_register(kgd, SRBM_GFX_CNTL, 0); + mutex_unlock(&rdev->srbm_mutex); +} + +static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, + uint32_t queue_id) +{ + uint32_t mec = (++pipe_id / CIK_PIPE_PER_MEC) + 1; + uint32_t pipe = (pipe_id % CIK_PIPE_PER_MEC); + + lock_srbm(kgd, mec, pipe, queue_id, 0); +} + +static void release_queue(struct kgd_dev *kgd) +{ + unlock_srbm(kgd); +} + +static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, + uint32_t sh_mem_config, + uint32_t sh_mem_ape1_base, + uint32_t sh_mem_ape1_limit, + uint32_t sh_mem_bases) +{ + lock_srbm(kgd, 0, 0, 0, vmid); + + write_register(kgd, SH_MEM_CONFIG, sh_mem_config); + write_register(kgd, SH_MEM_APE1_BASE, sh_mem_ape1_base); + write_register(kgd, SH_MEM_APE1_LIMIT, sh_mem_ape1_limit); + write_register(kgd, SH_MEM_BASES, sh_mem_bases); + + unlock_srbm(kgd); +} + +static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, + unsigned int vmid) +{ + /* + * We have to assume that there is no outstanding mapping. + * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 + * because a mapping is in progress or because a mapping finished and + * the SW cleared it. + * So the protocol is to always wait & clear. + */ + uint32_t pasid_mapping = (pasid == 0) ? 0 : + (uint32_t)pasid | ATC_VMID_PASID_MAPPING_VALID; + + write_register(kgd, ATC_VMID0_PASID_MAPPING + vmid*sizeof(uint32_t), + pasid_mapping); + + while (!(read_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS) & + (1U << vmid))) + cpu_relax(); + write_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS, 1U << vmid); + + return 0; +} + +static int kgd_init_memory(struct kgd_dev *kgd) +{ + /* + * Configure apertures: + * LDS: 0x60000000'00000000 - 0x60000001'00000000 (4GB) + * Scratch: 0x60000001'00000000 - 0x60000002'00000000 (4GB) + * GPUVM: 0x60010000'00000000 - 0x60020000'00000000 (1TB) + */ + int i; + uint32_t sh_mem_bases = PRIVATE_BASE(0x6000) | SHARED_BASE(0x6000); + + for (i = 8; i < 16; i++) { + uint32_t sh_mem_config; + + lock_srbm(kgd, 0, 0, 0, i); + + sh_mem_config = ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED); + sh_mem_config |= DEFAULT_MTYPE(MTYPE_NONCACHED); + + write_register(kgd, SH_MEM_CONFIG, sh_mem_config); + + write_register(kgd, SH_MEM_BASES, sh_mem_bases); + + /* Scratch aperture is not supported for now. */ + write_register(kgd, SH_STATIC_MEM_CONFIG, 0); + + /* APE1 disabled for now. */ + write_register(kgd, SH_MEM_APE1_BASE, 1); + write_register(kgd, SH_MEM_APE1_LIMIT, 0); + + unlock_srbm(kgd); + } + + return 0; +} + +static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id, + uint32_t hpd_size, uint64_t hpd_gpu_addr) +{ + uint32_t mec = (++pipe_id / CIK_PIPE_PER_MEC) + 1; + uint32_t pipe = (pipe_id % CIK_PIPE_PER_MEC); + + lock_srbm(kgd, mec, pipe, 0, 0); + write_register(kgd, CP_HPD_EOP_BASE_ADDR, + lower_32_bits(hpd_gpu_addr >> 8)); + write_register(kgd, CP_HPD_EOP_BASE_ADDR_HI, + upper_32_bits(hpd_gpu_addr >> 8)); + write_register(kgd, CP_HPD_EOP_VMID, 0); + write_register(kgd, CP_HPD_EOP_CONTROL, hpd_size); + unlock_srbm(kgd); + + return 0; +} + +static inline struct cik_mqd *get_mqd(void *mqd) +{ + return (struct cik_mqd *)mqd; +} + +static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, + uint32_t queue_id, uint32_t __user *wptr) +{ + uint32_t wptr_shadow, is_wptr_shadow_valid; + struct cik_mqd *m; + + m = get_mqd(mqd); + + is_wptr_shadow_valid = !get_user(wptr_shadow, wptr); + + acquire_queue(kgd, pipe_id, queue_id); + write_register(kgd, CP_MQD_BASE_ADDR, m->cp_mqd_base_addr_lo); + write_register(kgd, CP_MQD_BASE_ADDR_HI, m->cp_mqd_base_addr_hi); + write_register(kgd, CP_MQD_CONTROL, m->cp_mqd_control); + + write_register(kgd, CP_HQD_PQ_BASE, m->cp_hqd_pq_base_lo); + write_register(kgd, CP_HQD_PQ_BASE_HI, m->cp_hqd_pq_base_hi); + write_register(kgd, CP_HQD_PQ_CONTROL, m->cp_hqd_pq_control); + + write_register(kgd, CP_HQD_IB_CONTROL, m->cp_hqd_ib_control); + write_register(kgd, CP_HQD_IB_BASE_ADDR, m->cp_hqd_ib_base_addr_lo); + write_register(kgd, CP_HQD_IB_BASE_ADDR_HI, m->cp_hqd_ib_base_addr_hi); + + write_register(kgd, CP_HQD_IB_RPTR, m->cp_hqd_ib_rptr); + + write_register(kgd, CP_HQD_PERSISTENT_STATE, + m->cp_hqd_persistent_state); + write_register(kgd, CP_HQD_SEMA_CMD, m->cp_hqd_sema_cmd); + write_register(kgd, CP_HQD_MSG_TYPE, m->cp_hqd_msg_type); + + write_register(kgd, CP_HQD_ATOMIC0_PREOP_LO, + m->cp_hqd_atomic0_preop_lo); + + write_register(kgd, CP_HQD_ATOMIC0_PREOP_HI, + m->cp_hqd_atomic0_preop_hi); + + write_register(kgd, CP_HQD_ATOMIC1_PREOP_LO, + m->cp_hqd_atomic1_preop_lo); + + write_register(kgd, CP_HQD_ATOMIC1_PREOP_HI, + m->cp_hqd_atomic1_preop_hi); + + write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR, + m->cp_hqd_pq_rptr_report_addr_lo); + + write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR_HI, + m->cp_hqd_pq_rptr_report_addr_hi); + + write_register(kgd, CP_HQD_PQ_RPTR, m->cp_hqd_pq_rptr); + + write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR, + m->cp_hqd_pq_wptr_poll_addr_lo); + + write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR_HI, + m->cp_hqd_pq_wptr_poll_addr_hi); + + write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL, + m->cp_hqd_pq_doorbell_control); + + write_register(kgd, CP_HQD_VMID, m->cp_hqd_vmid); + + write_register(kgd, CP_HQD_QUANTUM, m->cp_hqd_quantum); + + write_register(kgd, CP_HQD_PIPE_PRIORITY, m->cp_hqd_pipe_priority); + write_register(kgd, CP_HQD_QUEUE_PRIORITY, m->cp_hqd_queue_priority); + + write_register(kgd, CP_HQD_IQ_RPTR, m->cp_hqd_iq_rptr); + + if (is_wptr_shadow_valid) + write_register(kgd, CP_HQD_PQ_WPTR, wptr_shadow); + + write_register(kgd, CP_HQD_ACTIVE, m->cp_hqd_active); + release_queue(kgd); + + return 0; +} + +static bool kgd_hqd_is_occupies(struct kgd_dev *kgd, uint64_t queue_address, + uint32_t pipe_id, uint32_t queue_id) +{ + uint32_t act; + bool retval = false; + uint32_t low, high; + + acquire_queue(kgd, pipe_id, queue_id); + act = read_register(kgd, CP_HQD_ACTIVE); + if (act) { + low = lower_32_bits(queue_address >> 8); + high = upper_32_bits(queue_address >> 8); + + if (low == read_register(kgd, CP_HQD_PQ_BASE) && + high == read_register(kgd, CP_HQD_PQ_BASE_HI)) + retval = true; + } + release_queue(kgd); + return retval; +} + +static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id) +{ + uint32_t temp; + + acquire_queue(kgd, pipe_id, queue_id); + write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL, 0); + + write_register(kgd, CP_HQD_DEQUEUE_REQUEST, reset_type); + + while (true) { + temp = read_register(kgd, CP_HQD_ACTIVE); + if (temp & 0x1) + break; + if (timeout == 0) { + pr_err("kfd: cp queue preemption time out (%dms)\n", + temp); + return -ETIME; + } + msleep(20); + timeout -= 20; + } + + release_queue(kgd); + return 0; +} diff --git a/drivers/gpu/drm/radeon/radeon_kfd.h b/drivers/gpu/drm/radeon/radeon_kfd.h new file mode 100644 index 0000000..f90e161 --- /dev/null +++ b/drivers/gpu/drm/radeon/radeon_kfd.h @@ -0,0 +1,47 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * radeon_kfd.h defines the private interface between the + * AMD kernel graphics drivers and the AMD KFD. + */ + +#ifndef RADEON_KFD_H_INCLUDED +#define RADEON_KFD_H_INCLUDED + +#include +#include "../amd/include/kgd_kfd_interface.h" + +struct radeon_device; + +bool radeon_kfd_init(void); +void radeon_kfd_fini(void); + +void radeon_kfd_suspend(struct radeon_device *rdev); +int radeon_kfd_resume(struct radeon_device *rdev); +void radeon_kfd_interrupt(struct radeon_device *rdev, + const void *ih_ring_entry); +void radeon_kfd_device_probe(struct radeon_device *rdev); +void radeon_kfd_device_init(struct radeon_device *rdev); +void radeon_kfd_device_fini(struct radeon_device *rdev); + +#endif /* RADEON_KFD_H_INCLUDED */ diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 8309b11..6eb561d 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -34,6 +34,8 @@ #include #include +#include "radeon_kfd.h" + #if defined(CONFIG_VGA_SWITCHEROO) bool radeon_has_atpx(void); #else @@ -63,6 +65,8 @@ int radeon_driver_unload_kms(struct drm_device *dev) pm_runtime_get_sync(dev->dev); + radeon_kfd_device_fini(rdev); + radeon_acpi_fini(rdev); radeon_modeset_fini(rdev); @@ -142,6 +146,9 @@ int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags) "Error during ACPI methods call\n"); } + radeon_kfd_device_probe(rdev); + radeon_kfd_device_init(rdev); + if (radeon_is_px(dev)) { pm_runtime_use_autosuspend(dev->dev); pm_runtime_set_autosuspend_delay(dev->dev, 5000); -- cgit v0.10.2 From 16423d67936f87e320a7b11771675b982cc9de02 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Tue, 15 Jul 2014 13:08:36 +0300 Subject: Update MAINTAINERS and CREDITS files with amdkfd info v6: Update entries to reflect new name & location of driver Signed-off-by: Oded Gabbay diff --git a/CREDITS b/CREDITS index bb62788..c56d8aa 100644 --- a/CREDITS +++ b/CREDITS @@ -1197,6 +1197,13 @@ S: R. Tocantins, 89 - Cristo Rei S: 80050-430 - Curitiba - Paraná S: Brazil +N: Oded Gabbay +E: oded.gabbay@gmail.com +D: AMD KFD maintainer +S: 12 Shraga Raphaeli +S: Petah-Tikva, 4906418 +S: Israel + N: Kumar Gala E: galak@kernel.crashing.org D: Embedded PowerPC 6xx/7xx/74xx/82xx/83xx/85xx support diff --git a/MAINTAINERS b/MAINTAINERS index ea4d005..3c69a3c7 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -618,6 +618,16 @@ S: Maintained F: drivers/iommu/amd_iommu*.[ch] F: include/linux/amd-iommu.h +AMD KFD +M: Oded Gabbay +L: dri-devel@lists.freedesktop.org +T: git git://people.freedesktop.org/~gabbayo/linux.git +S: Supported +F: drivers/gpu/drm/amd/amdkfd/ +F: drivers/gpu/drm/radeon/radeon_kfd.c +F: drivers/gpu/drm/radeon/radeon_kfd.h +F: include/uapi/linux/kfd_ioctl.h + AMD MICROCODE UPDATE SUPPORT M: Andreas Herrmann L: amd64-microcode@amd64.org -- cgit v0.10.2 From b7facbaec75a20f34c2065121dc423971682f922 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Wed, 16 Jul 2014 15:55:29 +0300 Subject: amdkfd: Add IOCTL set definitions of amdkfd - KFD_IOC_GET_VERSION: Retrieves the interface version of amdkfd - KFD_IOC_CREATE_QUEUE: Creates a usermode queue that runs on a specific GPU device - KFD_IOC_DESTROY_QUEUE: Destroys an existing usermode queue - KFD_IOC_SET_MEMORY_POLICY: Sets the memory policy of the default and alternate aperture of the calling process - KFD_IOC_GET_CLOCK_COUNTERS: Retrieves counters (timestamps) of CPU and GPU - KFD_IOC_GET_PROCESS_APERTURES: Retrieves information about process apertures that were initialized during the open() call of the amdkfd device - KFD_IOC_UPDATE_QUEUE: Updates configuration of an existing usermode queue v3: Remove pragma pack and pmc ioctls. Added parameter for doorbell offset and a comment on counters v5: Add define for AQL queues. Fix arguments of Get Version IOCTL Make IOCTL's structures to be the same size on 32/64 bit v6: Change the version of the amdkfd-thunk interface Signed-off-by: Oded Gabbay diff --git a/include/uapi/linux/kfd_ioctl.h b/include/uapi/linux/kfd_ioctl.h new file mode 100644 index 0000000..7acef41 --- /dev/null +++ b/include/uapi/linux/kfd_ioctl.h @@ -0,0 +1,154 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef KFD_IOCTL_H_INCLUDED +#define KFD_IOCTL_H_INCLUDED + +#include +#include + +#define KFD_IOCTL_MAJOR_VERSION 1 +#define KFD_IOCTL_MINOR_VERSION 0 + +struct kfd_ioctl_get_version_args { + uint32_t major_version; /* from KFD */ + uint32_t minor_version; /* from KFD */ +}; + +/* For kfd_ioctl_create_queue_args.queue_type. */ +#define KFD_IOC_QUEUE_TYPE_COMPUTE 0 +#define KFD_IOC_QUEUE_TYPE_SDMA 1 +#define KFD_IOC_QUEUE_TYPE_COMPUTE_AQL 2 + +#define KFD_MAX_QUEUE_PERCENTAGE 100 +#define KFD_MAX_QUEUE_PRIORITY 15 + +struct kfd_ioctl_create_queue_args { + uint64_t ring_base_address; /* to KFD */ + uint64_t write_pointer_address; /* from KFD */ + uint64_t read_pointer_address; /* from KFD */ + uint64_t doorbell_offset; /* from KFD */ + + uint32_t ring_size; /* to KFD */ + uint32_t gpu_id; /* to KFD */ + uint32_t queue_type; /* to KFD */ + uint32_t queue_percentage; /* to KFD */ + uint32_t queue_priority; /* to KFD */ + uint32_t queue_id; /* from KFD */ + + uint64_t eop_buffer_address; /* to KFD */ + uint64_t eop_buffer_size; /* to KFD */ + uint64_t ctx_save_restore_address; /* to KFD */ + uint64_t ctx_save_restore_size; /* to KFD */ +}; + +struct kfd_ioctl_destroy_queue_args { + uint32_t queue_id; /* to KFD */ + uint32_t pad; +}; + +struct kfd_ioctl_update_queue_args { + uint64_t ring_base_address; /* to KFD */ + + uint32_t queue_id; /* to KFD */ + uint32_t ring_size; /* to KFD */ + uint32_t queue_percentage; /* to KFD */ + uint32_t queue_priority; /* to KFD */ +}; + +/* For kfd_ioctl_set_memory_policy_args.default_policy and alternate_policy */ +#define KFD_IOC_CACHE_POLICY_COHERENT 0 +#define KFD_IOC_CACHE_POLICY_NONCOHERENT 1 + +struct kfd_ioctl_set_memory_policy_args { + uint64_t alternate_aperture_base; /* to KFD */ + uint64_t alternate_aperture_size; /* to KFD */ + + uint32_t gpu_id; /* to KFD */ + uint32_t default_policy; /* to KFD */ + uint32_t alternate_policy; /* to KFD */ + uint32_t pad; +}; + +/* + * All counters are monotonic. They are used for profiling of compute jobs. + * The profiling is done by userspace. + * + * In case of GPU reset, the counter should not be affected. + */ + +struct kfd_ioctl_get_clock_counters_args { + uint64_t gpu_clock_counter; /* from KFD */ + uint64_t cpu_clock_counter; /* from KFD */ + uint64_t system_clock_counter; /* from KFD */ + uint64_t system_clock_freq; /* from KFD */ + + uint32_t gpu_id; /* to KFD */ + uint32_t pad; +}; + +#define NUM_OF_SUPPORTED_GPUS 7 + +struct kfd_process_device_apertures { + uint64_t lds_base; /* from KFD */ + uint64_t lds_limit; /* from KFD */ + uint64_t scratch_base; /* from KFD */ + uint64_t scratch_limit; /* from KFD */ + uint64_t gpuvm_base; /* from KFD */ + uint64_t gpuvm_limit; /* from KFD */ + uint32_t gpu_id; /* from KFD */ + uint32_t pad; +}; + +struct kfd_ioctl_get_process_apertures_args { + struct kfd_process_device_apertures + process_apertures[NUM_OF_SUPPORTED_GPUS];/* from KFD */ + + /* from KFD, should be in the range [1 - NUM_OF_SUPPORTED_GPUS] */ + uint32_t num_of_nodes; + uint32_t pad; +}; + +#define KFD_IOC_MAGIC 'K' + +#define KFD_IOC_GET_VERSION \ + _IOR(KFD_IOC_MAGIC, 1, struct kfd_ioctl_get_version_args) + +#define KFD_IOC_CREATE_QUEUE \ + _IOWR(KFD_IOC_MAGIC, 2, struct kfd_ioctl_create_queue_args) + +#define KFD_IOC_DESTROY_QUEUE \ + _IOWR(KFD_IOC_MAGIC, 3, struct kfd_ioctl_destroy_queue_args) + +#define KFD_IOC_SET_MEMORY_POLICY \ + _IOW(KFD_IOC_MAGIC, 4, struct kfd_ioctl_set_memory_policy_args) + +#define KFD_IOC_GET_CLOCK_COUNTERS \ + _IOWR(KFD_IOC_MAGIC, 5, struct kfd_ioctl_get_clock_counters_args) + +#define KFD_IOC_GET_PROCESS_APERTURES \ + _IOR(KFD_IOC_MAGIC, 6, struct kfd_ioctl_get_process_apertures_args) + +#define KFD_IOC_UPDATE_QUEUE \ + _IOW(KFD_IOC_MAGIC, 7, struct kfd_ioctl_update_queue_args) + +#endif -- cgit v0.10.2 From 4a488a7ad71401169cecee75dc94bcce642e2c53 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Wed, 16 Jul 2014 21:08:55 +0300 Subject: amdkfd: Add amdkfd skeleton driver This patch adds the amdkfd skeleton driver. The driver does nothing except define a /dev/kfd device. It returns -ENODEV on all amdkfd IOCTLs. v3: Move bool field to the end of structure, removed the pmc ioctls and added a meaningful error message for ioctl error. v5: Create a new folder drm/amd and move amdkfd from drm/radeon/ to drm/amd/ Remove scheduler_class from kfd_priv.h as it was never used Add skeleton implementation of the Get Version IOCTL v6: Update module version to the correct number and remove the "default m" from the Kconfig file Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig index e3b4b0f..37c5a6e 100644 --- a/drivers/gpu/drm/Kconfig +++ b/drivers/gpu/drm/Kconfig @@ -200,3 +200,5 @@ source "drivers/gpu/drm/tegra/Kconfig" source "drivers/gpu/drm/panel/Kconfig" source "drivers/gpu/drm/sti/Kconfig" + +source "drivers/gpu/drm/amd/amdkfd/Kconfig" diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile index c3cf64c..dd9d35b 100644 --- a/drivers/gpu/drm/Makefile +++ b/drivers/gpu/drm/Makefile @@ -65,3 +65,4 @@ obj-$(CONFIG_DRM_STI) += sti/ obj-y += i2c/ obj-y += panel/ obj-y += bridge/ +obj-$(CONFIG_HSA_AMD) += amd/amdkfd/ diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig new file mode 100644 index 0000000..e13c67c --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig @@ -0,0 +1,9 @@ +# +# Heterogenous system architecture configuration +# + +config HSA_AMD + tristate "HSA kernel driver for AMD GPU devices" + depends on (DRM_RADEON || DRM_AMDGPU) && AMD_IOMMU_V2 && X86_64 + help + Enable this if you want to use HSA features on AMD GPU devices. diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile new file mode 100644 index 0000000..d875ce4 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -0,0 +1,9 @@ +# +# Makefile for Heterogenous System Architecture support for AMD GPU devices +# + +ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/amd/include/ + +amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o + +obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c new file mode 100644 index 0000000..d7c32eb --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -0,0 +1,210 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "kfd_priv.h" + +static long kfd_ioctl(struct file *, unsigned int, unsigned long); +static int kfd_open(struct inode *, struct file *); + +static const char kfd_dev_name[] = "kfd"; + +static const struct file_operations kfd_fops = { + .owner = THIS_MODULE, + .unlocked_ioctl = kfd_ioctl, + .compat_ioctl = kfd_ioctl, + .open = kfd_open, +}; + +static int kfd_char_dev_major = -1; +static struct class *kfd_class; +struct device *kfd_device; + +int kfd_chardev_init(void) +{ + int err = 0; + + kfd_char_dev_major = register_chrdev(0, kfd_dev_name, &kfd_fops); + err = kfd_char_dev_major; + if (err < 0) + goto err_register_chrdev; + + kfd_class = class_create(THIS_MODULE, kfd_dev_name); + err = PTR_ERR(kfd_class); + if (IS_ERR(kfd_class)) + goto err_class_create; + + kfd_device = device_create(kfd_class, NULL, + MKDEV(kfd_char_dev_major, 0), + NULL, kfd_dev_name); + err = PTR_ERR(kfd_device); + if (IS_ERR(kfd_device)) + goto err_device_create; + + return 0; + +err_device_create: + class_destroy(kfd_class); +err_class_create: + unregister_chrdev(kfd_char_dev_major, kfd_dev_name); +err_register_chrdev: + return err; +} + +void kfd_chardev_exit(void) +{ + device_destroy(kfd_class, MKDEV(kfd_char_dev_major, 0)); + class_destroy(kfd_class); + unregister_chrdev(kfd_char_dev_major, kfd_dev_name); +} + +struct device *kfd_chardev(void) +{ + return kfd_device; +} + + +static int kfd_open(struct inode *inode, struct file *filep) +{ + if (iminor(inode) != 0) + return -ENODEV; + + return 0; +} + +static long kfd_ioctl_get_version(struct file *filep, struct kfd_process *p, + void __user *arg) +{ + return -ENODEV; +} + +static long kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p, + void __user *arg) +{ + return -ENODEV; +} + +static int kfd_ioctl_destroy_queue(struct file *filp, struct kfd_process *p, + void __user *arg) +{ + return -ENODEV; +} + +static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p, + void __user *arg) +{ + return -ENODEV; +} + +static long kfd_ioctl_set_memory_policy(struct file *filep, + struct kfd_process *p, void __user *arg) +{ + return -ENODEV; +} + +static long kfd_ioctl_get_clock_counters(struct file *filep, + struct kfd_process *p, void __user *arg) +{ + return -ENODEV; +} + + +static int kfd_ioctl_get_process_apertures(struct file *filp, + struct kfd_process *p, void __user *arg) +{ + return -ENODEV; +} + +static long kfd_ioctl(struct file *filep, unsigned int cmd, unsigned long arg) +{ + struct kfd_process *process; + long err = -EINVAL; + + dev_dbg(kfd_device, + "ioctl cmd 0x%x (#%d), arg 0x%lx\n", + cmd, _IOC_NR(cmd), arg); + + /* TODO: add function that retrieves process */ + process = NULL; + + switch (cmd) { + case KFD_IOC_GET_VERSION: + err = kfd_ioctl_get_version(filep, process, (void __user *)arg); + break; + case KFD_IOC_CREATE_QUEUE: + err = kfd_ioctl_create_queue(filep, process, + (void __user *)arg); + break; + + case KFD_IOC_DESTROY_QUEUE: + err = kfd_ioctl_destroy_queue(filep, process, + (void __user *)arg); + break; + + case KFD_IOC_SET_MEMORY_POLICY: + err = kfd_ioctl_set_memory_policy(filep, process, + (void __user *)arg); + break; + + case KFD_IOC_GET_CLOCK_COUNTERS: + err = kfd_ioctl_get_clock_counters(filep, process, + (void __user *)arg); + break; + + case KFD_IOC_GET_PROCESS_APERTURES: + err = kfd_ioctl_get_process_apertures(filep, process, + (void __user *)arg); + break; + + case KFD_IOC_UPDATE_QUEUE: + err = kfd_ioctl_update_queue(filep, process, + (void __user *)arg); + break; + + default: + dev_err(kfd_device, + "unknown ioctl cmd 0x%x, arg 0x%lx)\n", + cmd, arg); + err = -EINVAL; + break; + } + + if (err < 0) + dev_err(kfd_device, + "ioctl error %ld for ioctl cmd 0x%x (#%d)\n", + err, cmd, _IOC_NR(cmd)); + + return err; +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c new file mode 100644 index 0000000..d81a58e --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -0,0 +1,130 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include +#include +#include +#include "kfd_priv.h" + +static const struct kfd_device_info kaveri_device_info = { + .max_pasid_bits = 16, +}; + +struct kfd_deviceid { + unsigned short did; + const struct kfd_device_info *device_info; +}; + +/* Please keep this sorted by increasing device id. */ +static const struct kfd_deviceid supported_devices[] = { + { 0x1304, &kaveri_device_info }, /* Kaveri */ + { 0x1305, &kaveri_device_info }, /* Kaveri */ + { 0x1306, &kaveri_device_info }, /* Kaveri */ + { 0x1307, &kaveri_device_info }, /* Kaveri */ + { 0x1309, &kaveri_device_info }, /* Kaveri */ + { 0x130A, &kaveri_device_info }, /* Kaveri */ + { 0x130B, &kaveri_device_info }, /* Kaveri */ + { 0x130C, &kaveri_device_info }, /* Kaveri */ + { 0x130D, &kaveri_device_info }, /* Kaveri */ + { 0x130E, &kaveri_device_info }, /* Kaveri */ + { 0x130F, &kaveri_device_info }, /* Kaveri */ + { 0x1310, &kaveri_device_info }, /* Kaveri */ + { 0x1311, &kaveri_device_info }, /* Kaveri */ + { 0x1312, &kaveri_device_info }, /* Kaveri */ + { 0x1313, &kaveri_device_info }, /* Kaveri */ + { 0x1315, &kaveri_device_info }, /* Kaveri */ + { 0x1316, &kaveri_device_info }, /* Kaveri */ + { 0x1317, &kaveri_device_info }, /* Kaveri */ + { 0x1318, &kaveri_device_info }, /* Kaveri */ + { 0x131B, &kaveri_device_info }, /* Kaveri */ + { 0x131C, &kaveri_device_info }, /* Kaveri */ + { 0x131D, &kaveri_device_info }, /* Kaveri */ +}; + +static const struct kfd_device_info *lookup_device_info(unsigned short did) +{ + size_t i; + + for (i = 0; i < ARRAY_SIZE(supported_devices); i++) { + if (supported_devices[i].did == did) { + BUG_ON(supported_devices[i].device_info == NULL); + return supported_devices[i].device_info; + } + } + + return NULL; +} + +struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev) +{ + struct kfd_dev *kfd; + + const struct kfd_device_info *device_info = + lookup_device_info(pdev->device); + + if (!device_info) + return NULL; + + kfd = kzalloc(sizeof(*kfd), GFP_KERNEL); + if (!kfd) + return NULL; + + kfd->kgd = kgd; + kfd->device_info = device_info; + kfd->pdev = pdev; + + return kfd; +} + +bool kgd2kfd_device_init(struct kfd_dev *kfd, + const struct kgd2kfd_shared_resources *gpu_resources) +{ + kfd->shared_resources = *gpu_resources; + + kfd->init_complete = true; + dev_info(kfd_device, "added device (%x:%x)\n", kfd->pdev->vendor, + kfd->pdev->device); + + return true; +} + +void kgd2kfd_device_exit(struct kfd_dev *kfd) +{ + kfree(kfd); +} + +void kgd2kfd_suspend(struct kfd_dev *kfd) +{ + BUG_ON(kfd == NULL); +} + +int kgd2kfd_resume(struct kfd_dev *kfd) +{ + BUG_ON(kfd == NULL); + + return 0; +} + +void kgd2kfd_interrupt(struct kfd_dev *dev, const void *ih_ring_entry) +{ +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c new file mode 100644 index 0000000..9519f33 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -0,0 +1,101 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include +#include +#include +#include +#include "kfd_priv.h" + +#define KFD_DRIVER_AUTHOR "AMD Inc. and others" + +#define KFD_DRIVER_DESC "Standalone HSA driver for AMD's GPUs" +#define KFD_DRIVER_DATE "20141113" +#define KFD_DRIVER_MAJOR 0 +#define KFD_DRIVER_MINOR 7 +#define KFD_DRIVER_PATCHLEVEL 0 + +const struct kfd2kgd_calls *kfd2kgd; +static const struct kgd2kfd_calls kgd2kfd = { + .exit = kgd2kfd_exit, + .probe = kgd2kfd_probe, + .device_init = kgd2kfd_device_init, + .device_exit = kgd2kfd_device_exit, + .interrupt = kgd2kfd_interrupt, + .suspend = kgd2kfd_suspend, + .resume = kgd2kfd_resume, +}; + +bool kgd2kfd_init(unsigned interface_version, + const struct kfd2kgd_calls *f2g, + const struct kgd2kfd_calls **g2f) +{ + /* + * Only one interface version is supported, + * no kfd/kgd version skew allowed. + */ + if (interface_version != KFD_INTERFACE_VERSION) + return false; + + kfd2kgd = f2g; + *g2f = &kgd2kfd; + + return true; +} +EXPORT_SYMBOL(kgd2kfd_init); + +void kgd2kfd_exit(void) +{ +} + +static int __init kfd_module_init(void) +{ + int err; + + err = kfd_chardev_init(); + if (err < 0) + goto err_ioctl; + + dev_info(kfd_device, "Initialized module\n"); + + return 0; + +err_ioctl: + return err; +} + +static void __exit kfd_module_exit(void) +{ + kfd_chardev_exit(); + dev_info(kfd_device, "Removed module\n"); +} + +module_init(kfd_module_init); +module_exit(kfd_module_exit); + +MODULE_AUTHOR(KFD_DRIVER_AUTHOR); +MODULE_DESCRIPTION(KFD_DRIVER_DESC); +MODULE_LICENSE("GPL and additional rights"); +MODULE_VERSION(__stringify(KFD_DRIVER_MAJOR) "." + __stringify(KFD_DRIVER_MINOR) "." + __stringify(KFD_DRIVER_PATCHLEVEL)); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h new file mode 100644 index 0000000..219a0d3 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -0,0 +1,81 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef KFD_PRIV_H_INCLUDED +#define KFD_PRIV_H_INCLUDED + +#include +#include +#include +#include +#include +#include +#include +#include + +struct kfd_device_info { + unsigned int max_pasid_bits; + size_t ih_ring_entry_size; +}; + +struct kfd_dev { + struct kgd_dev *kgd; + + const struct kfd_device_info *device_info; + struct pci_dev *pdev; + + unsigned int id; /* topology stub index */ + + struct kgd2kfd_shared_resources shared_resources; + + bool init_complete; + +}; + +/* KGD2KFD callbacks */ +void kgd2kfd_exit(void); +struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev); +bool kgd2kfd_device_init(struct kfd_dev *kfd, + const struct kgd2kfd_shared_resources *gpu_resources); +void kgd2kfd_device_exit(struct kfd_dev *kfd); + +extern const struct kfd2kgd_calls *kfd2kgd; + +/* Character device interface */ +int kfd_chardev_init(void); +void kfd_chardev_exit(void); +struct device *kfd_chardev(void); + +/* Process data */ +struct kfd_process { +}; + +extern struct device *kfd_device; + +/* Interrupts */ +void kgd2kfd_interrupt(struct kfd_dev *dev, const void *ih_ring_entry); + +/* Power Management */ +void kgd2kfd_suspend(struct kfd_dev *dev); +int kgd2kfd_resume(struct kfd_dev *dev); + +#endif -- cgit v0.10.2 From 5b5c4e40a37e858e2bff8cd91be8e972256392c4 Mon Sep 17 00:00:00 2001 From: Evgeny Pinchuk Date: Wed, 16 Jul 2014 21:22:32 +0300 Subject: amdkfd: Add topology module to amdkfd This patch adds the topology module to the driver. The topology is exposed to userspace through the sysfs. The calls to add and remove a device to/from topology are done by the radeon driver. v3: The CPU information, that is provided in the topology section of the amdkfd driver, is extracted from the CRAT table. Unlike the CPU information located in /sys/devices/system/cpu/cpu*, which is extracted from the SRAT table. While the CPU information provided by the CRAT and the SRAT tables might be identical, the node topology might be different. The SRAT table contains the topology of CPU nodes only. The CRAT table contains the topology of CPU and GPU nodes together (and can be interleaved). For example CPU node 1 in SRAT can be CPU node 3 in CRAT. Furthermore it's worth to mention that the CRAT table contains only HSA compatible nodes (nodes which are compliant with the HSA spec). To recap, amdkfd exposes a different kind of topology than the one exposed by /sys/devices/system/cpu/cpu even though it may contain similar information. v4: The topology module doesn't support uevent handling and doesn't notify the userspace about runtime modifications. It is up to the userspace to acquire snapshots of the topology information created by the amdkfd and exposed in sysfs. The following is an example of how the topology looks on a Kaveri A10-7850K system with amdkfd installed: /sys/devices/virtual/kfd/kfd/ | --- topology/ | |--- generation_id |--- system_properties |--- nodes/ | |--- 0/ | |--- gpu_id |--- name |--- properties |--- caches/ | |--- 0/ | |--- properties |--- 1/ | |--- properties |--- 2/ | |--- properties |--- io_links/ | |--- mem_banks/ | |--- 0/ | |--- properties |--- 1/ | |--- properties |--- 2/ | |--- properties |--- 3/ | |--- properties v5: Move amdkfd from drm/radeon/ to drm/amd/ Add a check if dev->gpu pointer is null before accessing it in the node_show function in kfd_topology.c This situation may occur when amdkfd is loaded and there is a GPU with a CRAT table, but that GPU isn't supported by amdkfd Signed-off-by: Evgeny Pinchuk Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index d875ce4..978654f 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -4,6 +4,6 @@ ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/amd/include/ -amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o +amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.h b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h new file mode 100644 index 0000000..a374fa3 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h @@ -0,0 +1,294 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef KFD_CRAT_H_INCLUDED +#define KFD_CRAT_H_INCLUDED + +#include + +#pragma pack(1) + +/* + * 4CC signature values for the CRAT and CDIT ACPI tables + */ + +#define CRAT_SIGNATURE "CRAT" +#define CDIT_SIGNATURE "CDIT" + +/* + * Component Resource Association Table (CRAT) + */ + +#define CRAT_OEMID_LENGTH 6 +#define CRAT_OEMTABLEID_LENGTH 8 +#define CRAT_RESERVED_LENGTH 6 + +#define CRAT_OEMID_64BIT_MASK ((1ULL << (CRAT_OEMID_LENGTH * 8)) - 1) + +struct crat_header { + uint32_t signature; + uint32_t length; + uint8_t revision; + uint8_t checksum; + uint8_t oem_id[CRAT_OEMID_LENGTH]; + uint8_t oem_table_id[CRAT_OEMTABLEID_LENGTH]; + uint32_t oem_revision; + uint32_t creator_id; + uint32_t creator_revision; + uint32_t total_entries; + uint16_t num_domains; + uint8_t reserved[CRAT_RESERVED_LENGTH]; +}; + +/* + * The header structure is immediately followed by total_entries of the + * data definitions + */ + +/* + * The currently defined subtype entries in the CRAT + */ +#define CRAT_SUBTYPE_COMPUTEUNIT_AFFINITY 0 +#define CRAT_SUBTYPE_MEMORY_AFFINITY 1 +#define CRAT_SUBTYPE_CACHE_AFFINITY 2 +#define CRAT_SUBTYPE_TLB_AFFINITY 3 +#define CRAT_SUBTYPE_CCOMPUTE_AFFINITY 4 +#define CRAT_SUBTYPE_IOLINK_AFFINITY 5 +#define CRAT_SUBTYPE_MAX 6 + +#define CRAT_SIBLINGMAP_SIZE 32 + +/* + * ComputeUnit Affinity structure and definitions + */ +#define CRAT_CU_FLAGS_ENABLED 0x00000001 +#define CRAT_CU_FLAGS_HOT_PLUGGABLE 0x00000002 +#define CRAT_CU_FLAGS_CPU_PRESENT 0x00000004 +#define CRAT_CU_FLAGS_GPU_PRESENT 0x00000008 +#define CRAT_CU_FLAGS_IOMMU_PRESENT 0x00000010 +#define CRAT_CU_FLAGS_RESERVED 0xffffffe0 + +#define CRAT_COMPUTEUNIT_RESERVED_LENGTH 4 + +struct crat_subtype_computeunit { + uint8_t type; + uint8_t length; + uint16_t reserved; + uint32_t flags; + uint32_t proximity_domain; + uint32_t processor_id_low; + uint16_t num_cpu_cores; + uint16_t num_simd_cores; + uint16_t max_waves_simd; + uint16_t io_count; + uint16_t hsa_capability; + uint16_t lds_size_in_kb; + uint8_t wave_front_size; + uint8_t num_banks; + uint16_t micro_engine_id; + uint8_t num_arrays; + uint8_t num_cu_per_array; + uint8_t num_simd_per_cu; + uint8_t max_slots_scatch_cu; + uint8_t reserved2[CRAT_COMPUTEUNIT_RESERVED_LENGTH]; +}; + +/* + * HSA Memory Affinity structure and definitions + */ +#define CRAT_MEM_FLAGS_ENABLED 0x00000001 +#define CRAT_MEM_FLAGS_HOT_PLUGGABLE 0x00000002 +#define CRAT_MEM_FLAGS_NON_VOLATILE 0x00000004 +#define CRAT_MEM_FLAGS_RESERVED 0xfffffff8 + +#define CRAT_MEMORY_RESERVED_LENGTH 8 + +struct crat_subtype_memory { + uint8_t type; + uint8_t length; + uint16_t reserved; + uint32_t flags; + uint32_t promixity_domain; + uint32_t base_addr_low; + uint32_t base_addr_high; + uint32_t length_low; + uint32_t length_high; + uint32_t width; + uint8_t reserved2[CRAT_MEMORY_RESERVED_LENGTH]; +}; + +/* + * HSA Cache Affinity structure and definitions + */ +#define CRAT_CACHE_FLAGS_ENABLED 0x00000001 +#define CRAT_CACHE_FLAGS_DATA_CACHE 0x00000002 +#define CRAT_CACHE_FLAGS_INST_CACHE 0x00000004 +#define CRAT_CACHE_FLAGS_CPU_CACHE 0x00000008 +#define CRAT_CACHE_FLAGS_SIMD_CACHE 0x00000010 +#define CRAT_CACHE_FLAGS_RESERVED 0xffffffe0 + +#define CRAT_CACHE_RESERVED_LENGTH 8 + +struct crat_subtype_cache { + uint8_t type; + uint8_t length; + uint16_t reserved; + uint32_t flags; + uint32_t processor_id_low; + uint8_t sibling_map[CRAT_SIBLINGMAP_SIZE]; + uint32_t cache_size; + uint8_t cache_level; + uint8_t lines_per_tag; + uint16_t cache_line_size; + uint8_t associativity; + uint8_t cache_properties; + uint16_t cache_latency; + uint8_t reserved2[CRAT_CACHE_RESERVED_LENGTH]; +}; + +/* + * HSA TLB Affinity structure and definitions + */ +#define CRAT_TLB_FLAGS_ENABLED 0x00000001 +#define CRAT_TLB_FLAGS_DATA_TLB 0x00000002 +#define CRAT_TLB_FLAGS_INST_TLB 0x00000004 +#define CRAT_TLB_FLAGS_CPU_TLB 0x00000008 +#define CRAT_TLB_FLAGS_SIMD_TLB 0x00000010 +#define CRAT_TLB_FLAGS_RESERVED 0xffffffe0 + +#define CRAT_TLB_RESERVED_LENGTH 4 + +struct crat_subtype_tlb { + uint8_t type; + uint8_t length; + uint16_t reserved; + uint32_t flags; + uint32_t processor_id_low; + uint8_t sibling_map[CRAT_SIBLINGMAP_SIZE]; + uint32_t tlb_level; + uint8_t data_tlb_associativity_2mb; + uint8_t data_tlb_size_2mb; + uint8_t instruction_tlb_associativity_2mb; + uint8_t instruction_tlb_size_2mb; + uint8_t data_tlb_associativity_4k; + uint8_t data_tlb_size_4k; + uint8_t instruction_tlb_associativity_4k; + uint8_t instruction_tlb_size_4k; + uint8_t data_tlb_associativity_1gb; + uint8_t data_tlb_size_1gb; + uint8_t instruction_tlb_associativity_1gb; + uint8_t instruction_tlb_size_1gb; + uint8_t reserved2[CRAT_TLB_RESERVED_LENGTH]; +}; + +/* + * HSA CCompute/APU Affinity structure and definitions + */ +#define CRAT_CCOMPUTE_FLAGS_ENABLED 0x00000001 +#define CRAT_CCOMPUTE_FLAGS_RESERVED 0xfffffffe + +#define CRAT_CCOMPUTE_RESERVED_LENGTH 16 + +struct crat_subtype_ccompute { + uint8_t type; + uint8_t length; + uint16_t reserved; + uint32_t flags; + uint32_t processor_id_low; + uint8_t sibling_map[CRAT_SIBLINGMAP_SIZE]; + uint32_t apu_size; + uint8_t reserved2[CRAT_CCOMPUTE_RESERVED_LENGTH]; +}; + +/* + * HSA IO Link Affinity structure and definitions + */ +#define CRAT_IOLINK_FLAGS_ENABLED 0x00000001 +#define CRAT_IOLINK_FLAGS_COHERENCY 0x00000002 +#define CRAT_IOLINK_FLAGS_RESERVED 0xfffffffc + +/* + * IO interface types + */ +#define CRAT_IOLINK_TYPE_UNDEFINED 0 +#define CRAT_IOLINK_TYPE_HYPERTRANSPORT 1 +#define CRAT_IOLINK_TYPE_PCIEXPRESS 2 +#define CRAT_IOLINK_TYPE_OTHER 3 +#define CRAT_IOLINK_TYPE_MAX 255 + +#define CRAT_IOLINK_RESERVED_LENGTH 24 + +struct crat_subtype_iolink { + uint8_t type; + uint8_t length; + uint16_t reserved; + uint32_t flags; + uint32_t proximity_domain_from; + uint32_t proximity_domain_to; + uint8_t io_interface_type; + uint8_t version_major; + uint16_t version_minor; + uint32_t minimum_latency; + uint32_t maximum_latency; + uint32_t minimum_bandwidth_mbs; + uint32_t maximum_bandwidth_mbs; + uint32_t recommended_transfer_size; + uint8_t reserved2[CRAT_IOLINK_RESERVED_LENGTH]; +}; + +/* + * HSA generic sub-type header + */ + +#define CRAT_SUBTYPE_FLAGS_ENABLED 0x00000001 + +struct crat_subtype_generic { + uint8_t type; + uint8_t length; + uint16_t reserved; + uint32_t flags; +}; + +/* + * Component Locality Distance Information Table (CDIT) + */ +#define CDIT_OEMID_LENGTH 6 +#define CDIT_OEMTABLEID_LENGTH 8 + +struct cdit_header { + uint32_t signature; + uint32_t length; + uint8_t revision; + uint8_t checksum; + uint8_t oem_id[CDIT_OEMID_LENGTH]; + uint8_t oem_table_id[CDIT_OEMTABLEID_LENGTH]; + uint32_t oem_revision; + uint32_t creator_id; + uint32_t creator_revision; + uint32_t total_entries; + uint16_t num_domains; + uint8_t entry[1]; +}; + +#pragma pack() + +#endif /* KFD_CRAT_H_INCLUDED */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index d81a58e..53b2e19 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -101,6 +101,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, { kfd->shared_resources = *gpu_resources; + if (kfd_topology_add_device(kfd) != 0) + return false; + kfd->init_complete = true; dev_info(kfd_device, "added device (%x:%x)\n", kfd->pdev->vendor, kfd->pdev->device); @@ -110,6 +113,10 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, void kgd2kfd_device_exit(struct kfd_dev *kfd) { + int err = kfd_topology_remove_device(kfd); + + BUG_ON(err != 0); + kfree(kfd); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c index 9519f33..45654be 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -76,16 +76,23 @@ static int __init kfd_module_init(void) if (err < 0) goto err_ioctl; + err = kfd_topology_init(); + if (err < 0) + goto err_topology; + dev_info(kfd_device, "Initialized module\n"); return 0; +err_topology: + kfd_chardev_exit(); err_ioctl: return err; } static void __exit kfd_module_exit(void) { + kfd_topology_shutdown(); kfd_chardev_exit(); dev_info(kfd_device, "Removed module\n"); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 219a0d3..b58b86d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -32,6 +32,15 @@ #include #include +#define KFD_SYSFS_FILE_MODE 0444 + +/* GPU ID hash width in bits */ +#define KFD_GPU_ID_HASH_WIDTH 16 + +/* Macro for allocating structures */ +#define kfd_alloc_struct(ptr_to_struct) \ + ((typeof(ptr_to_struct)) kzalloc(sizeof(*ptr_to_struct), GFP_KERNEL)) + struct kfd_device_info { unsigned int max_pasid_bits; size_t ih_ring_entry_size; @@ -71,6 +80,15 @@ struct kfd_process { extern struct device *kfd_device; +/* Topology */ +int kfd_topology_init(void); +void kfd_topology_shutdown(void); +int kfd_topology_add_device(struct kfd_dev *gpu); +int kfd_topology_remove_device(struct kfd_dev *gpu); +struct kfd_dev *kfd_device_by_id(uint32_t gpu_id); +struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev); +struct kfd_dev *kfd_topology_enum_kfd_devices(uint8_t idx); + /* Interrupts */ void kgd2kfd_interrupt(struct kfd_dev *dev, const void *ih_ring_entry); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c new file mode 100644 index 0000000..77cd7d5 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -0,0 +1,1235 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include +#include +#include +#include +#include +#include + +#include "kfd_priv.h" +#include "kfd_crat.h" +#include "kfd_topology.h" + +static struct list_head topology_device_list; +static int topology_crat_parsed; +static struct kfd_system_properties sys_props; + +static DECLARE_RWSEM(topology_lock); + +struct kfd_dev *kfd_device_by_id(uint32_t gpu_id) +{ + struct kfd_topology_device *top_dev; + struct kfd_dev *device = NULL; + + down_read(&topology_lock); + + list_for_each_entry(top_dev, &topology_device_list, list) + if (top_dev->gpu_id == gpu_id) { + device = top_dev->gpu; + break; + } + + up_read(&topology_lock); + + return device; +} + +struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev) +{ + struct kfd_topology_device *top_dev; + struct kfd_dev *device = NULL; + + down_read(&topology_lock); + + list_for_each_entry(top_dev, &topology_device_list, list) + if (top_dev->gpu->pdev == pdev) { + device = top_dev->gpu; + break; + } + + up_read(&topology_lock); + + return device; +} + +static int kfd_topology_get_crat_acpi(void *crat_image, size_t *size) +{ + struct acpi_table_header *crat_table; + acpi_status status; + + if (!size) + return -EINVAL; + + /* + * Fetch the CRAT table from ACPI + */ + status = acpi_get_table(CRAT_SIGNATURE, 0, &crat_table); + if (status == AE_NOT_FOUND) { + pr_warn("CRAT table not found\n"); + return -ENODATA; + } else if (ACPI_FAILURE(status)) { + const char *err = acpi_format_exception(status); + + pr_err("CRAT table error: %s\n", err); + return -EINVAL; + } + + if (*size >= crat_table->length && crat_image != 0) + memcpy(crat_image, crat_table, crat_table->length); + + *size = crat_table->length; + + return 0; +} + +static void kfd_populated_cu_info_cpu(struct kfd_topology_device *dev, + struct crat_subtype_computeunit *cu) +{ + BUG_ON(!dev); + BUG_ON(!cu); + + dev->node_props.cpu_cores_count = cu->num_cpu_cores; + dev->node_props.cpu_core_id_base = cu->processor_id_low; + if (cu->hsa_capability & CRAT_CU_FLAGS_IOMMU_PRESENT) + dev->node_props.capability |= HSA_CAP_ATS_PRESENT; + + pr_info("CU CPU: cores=%d id_base=%d\n", cu->num_cpu_cores, + cu->processor_id_low); +} + +static void kfd_populated_cu_info_gpu(struct kfd_topology_device *dev, + struct crat_subtype_computeunit *cu) +{ + BUG_ON(!dev); + BUG_ON(!cu); + + dev->node_props.simd_id_base = cu->processor_id_low; + dev->node_props.simd_count = cu->num_simd_cores; + dev->node_props.lds_size_in_kb = cu->lds_size_in_kb; + dev->node_props.max_waves_per_simd = cu->max_waves_simd; + dev->node_props.wave_front_size = cu->wave_front_size; + dev->node_props.mem_banks_count = cu->num_banks; + dev->node_props.array_count = cu->num_arrays; + dev->node_props.cu_per_simd_array = cu->num_cu_per_array; + dev->node_props.simd_per_cu = cu->num_simd_per_cu; + dev->node_props.max_slots_scratch_cu = cu->max_slots_scatch_cu; + if (cu->hsa_capability & CRAT_CU_FLAGS_HOT_PLUGGABLE) + dev->node_props.capability |= HSA_CAP_HOT_PLUGGABLE; + pr_info("CU GPU: simds=%d id_base=%d\n", cu->num_simd_cores, + cu->processor_id_low); +} + +/* kfd_parse_subtype_cu is called when the topology mutex is already acquired */ +static int kfd_parse_subtype_cu(struct crat_subtype_computeunit *cu) +{ + struct kfd_topology_device *dev; + int i = 0; + + BUG_ON(!cu); + + pr_info("Found CU entry in CRAT table with proximity_domain=%d caps=%x\n", + cu->proximity_domain, cu->hsa_capability); + list_for_each_entry(dev, &topology_device_list, list) { + if (cu->proximity_domain == i) { + if (cu->flags & CRAT_CU_FLAGS_CPU_PRESENT) + kfd_populated_cu_info_cpu(dev, cu); + + if (cu->flags & CRAT_CU_FLAGS_GPU_PRESENT) + kfd_populated_cu_info_gpu(dev, cu); + break; + } + i++; + } + + return 0; +} + +/* + * kfd_parse_subtype_mem is called when the topology mutex is + * already acquired + */ +static int kfd_parse_subtype_mem(struct crat_subtype_memory *mem) +{ + struct kfd_mem_properties *props; + struct kfd_topology_device *dev; + int i = 0; + + BUG_ON(!mem); + + pr_info("Found memory entry in CRAT table with proximity_domain=%d\n", + mem->promixity_domain); + list_for_each_entry(dev, &topology_device_list, list) { + if (mem->promixity_domain == i) { + props = kfd_alloc_struct(props); + if (props == 0) + return -ENOMEM; + + if (dev->node_props.cpu_cores_count == 0) + props->heap_type = HSA_MEM_HEAP_TYPE_FB_PRIVATE; + else + props->heap_type = HSA_MEM_HEAP_TYPE_SYSTEM; + + if (mem->flags & CRAT_MEM_FLAGS_HOT_PLUGGABLE) + props->flags |= HSA_MEM_FLAGS_HOT_PLUGGABLE; + if (mem->flags & CRAT_MEM_FLAGS_NON_VOLATILE) + props->flags |= HSA_MEM_FLAGS_NON_VOLATILE; + + props->size_in_bytes = + ((uint64_t)mem->length_high << 32) + + mem->length_low; + props->width = mem->width; + + dev->mem_bank_count++; + list_add_tail(&props->list, &dev->mem_props); + + break; + } + i++; + } + + return 0; +} + +/* + * kfd_parse_subtype_cache is called when the topology mutex + * is already acquired + */ +static int kfd_parse_subtype_cache(struct crat_subtype_cache *cache) +{ + struct kfd_cache_properties *props; + struct kfd_topology_device *dev; + uint32_t id; + + BUG_ON(!cache); + + id = cache->processor_id_low; + + pr_info("Found cache entry in CRAT table with processor_id=%d\n", id); + list_for_each_entry(dev, &topology_device_list, list) + if (id == dev->node_props.cpu_core_id_base || + id == dev->node_props.simd_id_base) { + props = kfd_alloc_struct(props); + if (props == 0) + return -ENOMEM; + + props->processor_id_low = id; + props->cache_level = cache->cache_level; + props->cache_size = cache->cache_size; + props->cacheline_size = cache->cache_line_size; + props->cachelines_per_tag = cache->lines_per_tag; + props->cache_assoc = cache->associativity; + props->cache_latency = cache->cache_latency; + + if (cache->flags & CRAT_CACHE_FLAGS_DATA_CACHE) + props->cache_type |= HSA_CACHE_TYPE_DATA; + if (cache->flags & CRAT_CACHE_FLAGS_INST_CACHE) + props->cache_type |= HSA_CACHE_TYPE_INSTRUCTION; + if (cache->flags & CRAT_CACHE_FLAGS_CPU_CACHE) + props->cache_type |= HSA_CACHE_TYPE_CPU; + if (cache->flags & CRAT_CACHE_FLAGS_SIMD_CACHE) + props->cache_type |= HSA_CACHE_TYPE_HSACU; + + dev->cache_count++; + dev->node_props.caches_count++; + list_add_tail(&props->list, &dev->cache_props); + + break; + } + + return 0; +} + +/* + * kfd_parse_subtype_iolink is called when the topology mutex + * is already acquired + */ +static int kfd_parse_subtype_iolink(struct crat_subtype_iolink *iolink) +{ + struct kfd_iolink_properties *props; + struct kfd_topology_device *dev; + uint32_t i = 0; + uint32_t id_from; + uint32_t id_to; + + BUG_ON(!iolink); + + id_from = iolink->proximity_domain_from; + id_to = iolink->proximity_domain_to; + + pr_info("Found IO link entry in CRAT table with id_from=%d\n", id_from); + list_for_each_entry(dev, &topology_device_list, list) { + if (id_from == i) { + props = kfd_alloc_struct(props); + if (props == 0) + return -ENOMEM; + + props->node_from = id_from; + props->node_to = id_to; + props->ver_maj = iolink->version_major; + props->ver_min = iolink->version_minor; + + /* + * weight factor (derived from CDIR), currently always 1 + */ + props->weight = 1; + + props->min_latency = iolink->minimum_latency; + props->max_latency = iolink->maximum_latency; + props->min_bandwidth = iolink->minimum_bandwidth_mbs; + props->max_bandwidth = iolink->maximum_bandwidth_mbs; + props->rec_transfer_size = + iolink->recommended_transfer_size; + + dev->io_link_count++; + dev->node_props.io_links_count++; + list_add_tail(&props->list, &dev->io_link_props); + + break; + } + i++; + } + + return 0; +} + +static int kfd_parse_subtype(struct crat_subtype_generic *sub_type_hdr) +{ + struct crat_subtype_computeunit *cu; + struct crat_subtype_memory *mem; + struct crat_subtype_cache *cache; + struct crat_subtype_iolink *iolink; + int ret = 0; + + BUG_ON(!sub_type_hdr); + + switch (sub_type_hdr->type) { + case CRAT_SUBTYPE_COMPUTEUNIT_AFFINITY: + cu = (struct crat_subtype_computeunit *)sub_type_hdr; + ret = kfd_parse_subtype_cu(cu); + break; + case CRAT_SUBTYPE_MEMORY_AFFINITY: + mem = (struct crat_subtype_memory *)sub_type_hdr; + ret = kfd_parse_subtype_mem(mem); + break; + case CRAT_SUBTYPE_CACHE_AFFINITY: + cache = (struct crat_subtype_cache *)sub_type_hdr; + ret = kfd_parse_subtype_cache(cache); + break; + case CRAT_SUBTYPE_TLB_AFFINITY: + /* + * For now, nothing to do here + */ + pr_info("Found TLB entry in CRAT table (not processing)\n"); + break; + case CRAT_SUBTYPE_CCOMPUTE_AFFINITY: + /* + * For now, nothing to do here + */ + pr_info("Found CCOMPUTE entry in CRAT table (not processing)\n"); + break; + case CRAT_SUBTYPE_IOLINK_AFFINITY: + iolink = (struct crat_subtype_iolink *)sub_type_hdr; + ret = kfd_parse_subtype_iolink(iolink); + break; + default: + pr_warn("Unknown subtype (%d) in CRAT\n", + sub_type_hdr->type); + } + + return ret; +} + +static void kfd_release_topology_device(struct kfd_topology_device *dev) +{ + struct kfd_mem_properties *mem; + struct kfd_cache_properties *cache; + struct kfd_iolink_properties *iolink; + + BUG_ON(!dev); + + list_del(&dev->list); + + while (dev->mem_props.next != &dev->mem_props) { + mem = container_of(dev->mem_props.next, + struct kfd_mem_properties, list); + list_del(&mem->list); + kfree(mem); + } + + while (dev->cache_props.next != &dev->cache_props) { + cache = container_of(dev->cache_props.next, + struct kfd_cache_properties, list); + list_del(&cache->list); + kfree(cache); + } + + while (dev->io_link_props.next != &dev->io_link_props) { + iolink = container_of(dev->io_link_props.next, + struct kfd_iolink_properties, list); + list_del(&iolink->list); + kfree(iolink); + } + + kfree(dev); + + sys_props.num_devices--; +} + +static void kfd_release_live_view(void) +{ + struct kfd_topology_device *dev; + + while (topology_device_list.next != &topology_device_list) { + dev = container_of(topology_device_list.next, + struct kfd_topology_device, list); + kfd_release_topology_device(dev); +} + + memset(&sys_props, 0, sizeof(sys_props)); +} + +static struct kfd_topology_device *kfd_create_topology_device(void) +{ + struct kfd_topology_device *dev; + + dev = kfd_alloc_struct(dev); + if (dev == 0) { + pr_err("No memory to allocate a topology device"); + return 0; + } + + INIT_LIST_HEAD(&dev->mem_props); + INIT_LIST_HEAD(&dev->cache_props); + INIT_LIST_HEAD(&dev->io_link_props); + + list_add_tail(&dev->list, &topology_device_list); + sys_props.num_devices++; + + return dev; + } + +static int kfd_parse_crat_table(void *crat_image) +{ + struct kfd_topology_device *top_dev; + struct crat_subtype_generic *sub_type_hdr; + uint16_t node_id; + int ret; + struct crat_header *crat_table = (struct crat_header *)crat_image; + uint16_t num_nodes; + uint32_t image_len; + + if (!crat_image) + return -EINVAL; + + num_nodes = crat_table->num_domains; + image_len = crat_table->length; + + pr_info("Parsing CRAT table with %d nodes\n", num_nodes); + + for (node_id = 0; node_id < num_nodes; node_id++) { + top_dev = kfd_create_topology_device(); + if (!top_dev) { + kfd_release_live_view(); + return -ENOMEM; + } + } + + sys_props.platform_id = + (*((uint64_t *)crat_table->oem_id)) & CRAT_OEMID_64BIT_MASK; + sys_props.platform_oem = *((uint64_t *)crat_table->oem_table_id); + sys_props.platform_rev = crat_table->revision; + + sub_type_hdr = (struct crat_subtype_generic *)(crat_table+1); + while ((char *)sub_type_hdr + sizeof(struct crat_subtype_generic) < + ((char *)crat_image) + image_len) { + if (sub_type_hdr->flags & CRAT_SUBTYPE_FLAGS_ENABLED) { + ret = kfd_parse_subtype(sub_type_hdr); + if (ret != 0) { + kfd_release_live_view(); + return ret; + } + } + + sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr + + sub_type_hdr->length); + } + + sys_props.generation_count++; + topology_crat_parsed = 1; + + return 0; +} + + +#define sysfs_show_gen_prop(buffer, fmt, ...) \ + snprintf(buffer, PAGE_SIZE, "%s"fmt, buffer, __VA_ARGS__) +#define sysfs_show_32bit_prop(buffer, name, value) \ + sysfs_show_gen_prop(buffer, "%s %u\n", name, value) +#define sysfs_show_64bit_prop(buffer, name, value) \ + sysfs_show_gen_prop(buffer, "%s %llu\n", name, value) +#define sysfs_show_32bit_val(buffer, value) \ + sysfs_show_gen_prop(buffer, "%u\n", value) +#define sysfs_show_str_val(buffer, value) \ + sysfs_show_gen_prop(buffer, "%s\n", value) + +static ssize_t sysprops_show(struct kobject *kobj, struct attribute *attr, + char *buffer) +{ + ssize_t ret; + + /* Making sure that the buffer is an empty string */ + buffer[0] = 0; + + if (attr == &sys_props.attr_genid) { + ret = sysfs_show_32bit_val(buffer, sys_props.generation_count); + } else if (attr == &sys_props.attr_props) { + sysfs_show_64bit_prop(buffer, "platform_oem", + sys_props.platform_oem); + sysfs_show_64bit_prop(buffer, "platform_id", + sys_props.platform_id); + ret = sysfs_show_64bit_prop(buffer, "platform_rev", + sys_props.platform_rev); + } else { + ret = -EINVAL; + } + + return ret; +} + +static const struct sysfs_ops sysprops_ops = { + .show = sysprops_show, +}; + +static struct kobj_type sysprops_type = { + .sysfs_ops = &sysprops_ops, +}; + +static ssize_t iolink_show(struct kobject *kobj, struct attribute *attr, + char *buffer) +{ + ssize_t ret; + struct kfd_iolink_properties *iolink; + + /* Making sure that the buffer is an empty string */ + buffer[0] = 0; + + iolink = container_of(attr, struct kfd_iolink_properties, attr); + sysfs_show_32bit_prop(buffer, "type", iolink->iolink_type); + sysfs_show_32bit_prop(buffer, "version_major", iolink->ver_maj); + sysfs_show_32bit_prop(buffer, "version_minor", iolink->ver_min); + sysfs_show_32bit_prop(buffer, "node_from", iolink->node_from); + sysfs_show_32bit_prop(buffer, "node_to", iolink->node_to); + sysfs_show_32bit_prop(buffer, "weight", iolink->weight); + sysfs_show_32bit_prop(buffer, "min_latency", iolink->min_latency); + sysfs_show_32bit_prop(buffer, "max_latency", iolink->max_latency); + sysfs_show_32bit_prop(buffer, "min_bandwidth", iolink->min_bandwidth); + sysfs_show_32bit_prop(buffer, "max_bandwidth", iolink->max_bandwidth); + sysfs_show_32bit_prop(buffer, "recommended_transfer_size", + iolink->rec_transfer_size); + ret = sysfs_show_32bit_prop(buffer, "flags", iolink->flags); + + return ret; +} + +static const struct sysfs_ops iolink_ops = { + .show = iolink_show, +}; + +static struct kobj_type iolink_type = { + .sysfs_ops = &iolink_ops, +}; + +static ssize_t mem_show(struct kobject *kobj, struct attribute *attr, + char *buffer) +{ + ssize_t ret; + struct kfd_mem_properties *mem; + + /* Making sure that the buffer is an empty string */ + buffer[0] = 0; + + mem = container_of(attr, struct kfd_mem_properties, attr); + sysfs_show_32bit_prop(buffer, "heap_type", mem->heap_type); + sysfs_show_64bit_prop(buffer, "size_in_bytes", mem->size_in_bytes); + sysfs_show_32bit_prop(buffer, "flags", mem->flags); + sysfs_show_32bit_prop(buffer, "width", mem->width); + ret = sysfs_show_32bit_prop(buffer, "mem_clk_max", mem->mem_clk_max); + + return ret; +} + +static const struct sysfs_ops mem_ops = { + .show = mem_show, +}; + +static struct kobj_type mem_type = { + .sysfs_ops = &mem_ops, +}; + +static ssize_t kfd_cache_show(struct kobject *kobj, struct attribute *attr, + char *buffer) +{ + ssize_t ret; + uint32_t i; + struct kfd_cache_properties *cache; + + /* Making sure that the buffer is an empty string */ + buffer[0] = 0; + + cache = container_of(attr, struct kfd_cache_properties, attr); + sysfs_show_32bit_prop(buffer, "processor_id_low", + cache->processor_id_low); + sysfs_show_32bit_prop(buffer, "level", cache->cache_level); + sysfs_show_32bit_prop(buffer, "size", cache->cache_size); + sysfs_show_32bit_prop(buffer, "cache_line_size", cache->cacheline_size); + sysfs_show_32bit_prop(buffer, "cache_lines_per_tag", + cache->cachelines_per_tag); + sysfs_show_32bit_prop(buffer, "association", cache->cache_assoc); + sysfs_show_32bit_prop(buffer, "latency", cache->cache_latency); + sysfs_show_32bit_prop(buffer, "type", cache->cache_type); + snprintf(buffer, PAGE_SIZE, "%ssibling_map ", buffer); + for (i = 0; i < KFD_TOPOLOGY_CPU_SIBLINGS; i++) + ret = snprintf(buffer, PAGE_SIZE, "%s%d%s", + buffer, cache->sibling_map[i], + (i == KFD_TOPOLOGY_CPU_SIBLINGS-1) ? + "\n" : ","); + + return ret; +} + +static const struct sysfs_ops cache_ops = { + .show = kfd_cache_show, +}; + +static struct kobj_type cache_type = { + .sysfs_ops = &cache_ops, +}; + +static ssize_t node_show(struct kobject *kobj, struct attribute *attr, + char *buffer) +{ + ssize_t ret; + struct kfd_topology_device *dev; + char public_name[KFD_TOPOLOGY_PUBLIC_NAME_SIZE]; + uint32_t i; + + /* Making sure that the buffer is an empty string */ + buffer[0] = 0; + + if (strcmp(attr->name, "gpu_id") == 0) { + dev = container_of(attr, struct kfd_topology_device, + attr_gpuid); + ret = sysfs_show_32bit_val(buffer, dev->gpu_id); + } else if (strcmp(attr->name, "name") == 0) { + dev = container_of(attr, struct kfd_topology_device, + attr_name); + for (i = 0; i < KFD_TOPOLOGY_PUBLIC_NAME_SIZE; i++) { + public_name[i] = + (char)dev->node_props.marketing_name[i]; + if (dev->node_props.marketing_name[i] == 0) + break; + } + public_name[KFD_TOPOLOGY_PUBLIC_NAME_SIZE-1] = 0x0; + ret = sysfs_show_str_val(buffer, public_name); + } else { + dev = container_of(attr, struct kfd_topology_device, + attr_props); + sysfs_show_32bit_prop(buffer, "cpu_cores_count", + dev->node_props.cpu_cores_count); + sysfs_show_32bit_prop(buffer, "simd_count", + dev->node_props.simd_count); + + if (dev->mem_bank_count < dev->node_props.mem_banks_count) { + pr_warn("kfd: mem_banks_count truncated from %d to %d\n", + dev->node_props.mem_banks_count, + dev->mem_bank_count); + sysfs_show_32bit_prop(buffer, "mem_banks_count", + dev->mem_bank_count); + } else { + sysfs_show_32bit_prop(buffer, "mem_banks_count", + dev->node_props.mem_banks_count); + } + + sysfs_show_32bit_prop(buffer, "caches_count", + dev->node_props.caches_count); + sysfs_show_32bit_prop(buffer, "io_links_count", + dev->node_props.io_links_count); + sysfs_show_32bit_prop(buffer, "cpu_core_id_base", + dev->node_props.cpu_core_id_base); + sysfs_show_32bit_prop(buffer, "simd_id_base", + dev->node_props.simd_id_base); + sysfs_show_32bit_prop(buffer, "capability", + dev->node_props.capability); + sysfs_show_32bit_prop(buffer, "max_waves_per_simd", + dev->node_props.max_waves_per_simd); + sysfs_show_32bit_prop(buffer, "lds_size_in_kb", + dev->node_props.lds_size_in_kb); + sysfs_show_32bit_prop(buffer, "gds_size_in_kb", + dev->node_props.gds_size_in_kb); + sysfs_show_32bit_prop(buffer, "wave_front_size", + dev->node_props.wave_front_size); + sysfs_show_32bit_prop(buffer, "array_count", + dev->node_props.array_count); + sysfs_show_32bit_prop(buffer, "simd_arrays_per_engine", + dev->node_props.simd_arrays_per_engine); + sysfs_show_32bit_prop(buffer, "cu_per_simd_array", + dev->node_props.cu_per_simd_array); + sysfs_show_32bit_prop(buffer, "simd_per_cu", + dev->node_props.simd_per_cu); + sysfs_show_32bit_prop(buffer, "max_slots_scratch_cu", + dev->node_props.max_slots_scratch_cu); + sysfs_show_32bit_prop(buffer, "engine_id", + dev->node_props.engine_id); + sysfs_show_32bit_prop(buffer, "vendor_id", + dev->node_props.vendor_id); + sysfs_show_32bit_prop(buffer, "device_id", + dev->node_props.device_id); + sysfs_show_32bit_prop(buffer, "location_id", + dev->node_props.location_id); + + if (dev->gpu) { + sysfs_show_32bit_prop(buffer, "max_engine_clk_fcompute", + kfd2kgd->get_max_engine_clock_in_mhz( + dev->gpu->kgd)); + sysfs_show_64bit_prop(buffer, "local_mem_size", + kfd2kgd->get_vmem_size(dev->gpu->kgd)); + } + + ret = sysfs_show_32bit_prop(buffer, "max_engine_clk_ccompute", + cpufreq_quick_get_max(0)/1000); + } + + return ret; +} + +static const struct sysfs_ops node_ops = { + .show = node_show, +}; + +static struct kobj_type node_type = { + .sysfs_ops = &node_ops, +}; + +static void kfd_remove_sysfs_file(struct kobject *kobj, struct attribute *attr) +{ + sysfs_remove_file(kobj, attr); + kobject_del(kobj); + kobject_put(kobj); +} + +static void kfd_remove_sysfs_node_entry(struct kfd_topology_device *dev) +{ + struct kfd_iolink_properties *iolink; + struct kfd_cache_properties *cache; + struct kfd_mem_properties *mem; + + BUG_ON(!dev); + + if (dev->kobj_iolink) { + list_for_each_entry(iolink, &dev->io_link_props, list) + if (iolink->kobj) { + kfd_remove_sysfs_file(iolink->kobj, + &iolink->attr); + iolink->kobj = 0; + } + kobject_del(dev->kobj_iolink); + kobject_put(dev->kobj_iolink); + dev->kobj_iolink = 0; + } + + if (dev->kobj_cache) { + list_for_each_entry(cache, &dev->cache_props, list) + if (cache->kobj) { + kfd_remove_sysfs_file(cache->kobj, + &cache->attr); + cache->kobj = 0; + } + kobject_del(dev->kobj_cache); + kobject_put(dev->kobj_cache); + dev->kobj_cache = 0; + } + + if (dev->kobj_mem) { + list_for_each_entry(mem, &dev->mem_props, list) + if (mem->kobj) { + kfd_remove_sysfs_file(mem->kobj, &mem->attr); + mem->kobj = 0; + } + kobject_del(dev->kobj_mem); + kobject_put(dev->kobj_mem); + dev->kobj_mem = 0; + } + + if (dev->kobj_node) { + sysfs_remove_file(dev->kobj_node, &dev->attr_gpuid); + sysfs_remove_file(dev->kobj_node, &dev->attr_name); + sysfs_remove_file(dev->kobj_node, &dev->attr_props); + kobject_del(dev->kobj_node); + kobject_put(dev->kobj_node); + dev->kobj_node = 0; + } +} + +static int kfd_build_sysfs_node_entry(struct kfd_topology_device *dev, + uint32_t id) +{ + struct kfd_iolink_properties *iolink; + struct kfd_cache_properties *cache; + struct kfd_mem_properties *mem; + int ret; + uint32_t i; + + BUG_ON(!dev); + + /* + * Creating the sysfs folders + */ + BUG_ON(dev->kobj_node); + dev->kobj_node = kfd_alloc_struct(dev->kobj_node); + if (!dev->kobj_node) + return -ENOMEM; + + ret = kobject_init_and_add(dev->kobj_node, &node_type, + sys_props.kobj_nodes, "%d", id); + if (ret < 0) + return ret; + + dev->kobj_mem = kobject_create_and_add("mem_banks", dev->kobj_node); + if (!dev->kobj_mem) + return -ENOMEM; + + dev->kobj_cache = kobject_create_and_add("caches", dev->kobj_node); + if (!dev->kobj_cache) + return -ENOMEM; + + dev->kobj_iolink = kobject_create_and_add("io_links", dev->kobj_node); + if (!dev->kobj_iolink) + return -ENOMEM; + + /* + * Creating sysfs files for node properties + */ + dev->attr_gpuid.name = "gpu_id"; + dev->attr_gpuid.mode = KFD_SYSFS_FILE_MODE; + sysfs_attr_init(&dev->attr_gpuid); + dev->attr_name.name = "name"; + dev->attr_name.mode = KFD_SYSFS_FILE_MODE; + sysfs_attr_init(&dev->attr_name); + dev->attr_props.name = "properties"; + dev->attr_props.mode = KFD_SYSFS_FILE_MODE; + sysfs_attr_init(&dev->attr_props); + ret = sysfs_create_file(dev->kobj_node, &dev->attr_gpuid); + if (ret < 0) + return ret; + ret = sysfs_create_file(dev->kobj_node, &dev->attr_name); + if (ret < 0) + return ret; + ret = sysfs_create_file(dev->kobj_node, &dev->attr_props); + if (ret < 0) + return ret; + + i = 0; + list_for_each_entry(mem, &dev->mem_props, list) { + mem->kobj = kzalloc(sizeof(struct kobject), GFP_KERNEL); + if (!mem->kobj) + return -ENOMEM; + ret = kobject_init_and_add(mem->kobj, &mem_type, + dev->kobj_mem, "%d", i); + if (ret < 0) + return ret; + + mem->attr.name = "properties"; + mem->attr.mode = KFD_SYSFS_FILE_MODE; + sysfs_attr_init(&mem->attr); + ret = sysfs_create_file(mem->kobj, &mem->attr); + if (ret < 0) + return ret; + i++; + } + + i = 0; + list_for_each_entry(cache, &dev->cache_props, list) { + cache->kobj = kzalloc(sizeof(struct kobject), GFP_KERNEL); + if (!cache->kobj) + return -ENOMEM; + ret = kobject_init_and_add(cache->kobj, &cache_type, + dev->kobj_cache, "%d", i); + if (ret < 0) + return ret; + + cache->attr.name = "properties"; + cache->attr.mode = KFD_SYSFS_FILE_MODE; + sysfs_attr_init(&cache->attr); + ret = sysfs_create_file(cache->kobj, &cache->attr); + if (ret < 0) + return ret; + i++; + } + + i = 0; + list_for_each_entry(iolink, &dev->io_link_props, list) { + iolink->kobj = kzalloc(sizeof(struct kobject), GFP_KERNEL); + if (!iolink->kobj) + return -ENOMEM; + ret = kobject_init_and_add(iolink->kobj, &iolink_type, + dev->kobj_iolink, "%d", i); + if (ret < 0) + return ret; + + iolink->attr.name = "properties"; + iolink->attr.mode = KFD_SYSFS_FILE_MODE; + sysfs_attr_init(&iolink->attr); + ret = sysfs_create_file(iolink->kobj, &iolink->attr); + if (ret < 0) + return ret; + i++; +} + + return 0; +} + +static int kfd_build_sysfs_node_tree(void) +{ + struct kfd_topology_device *dev; + int ret; + uint32_t i = 0; + + list_for_each_entry(dev, &topology_device_list, list) { + ret = kfd_build_sysfs_node_entry(dev, 0); + if (ret < 0) + return ret; + i++; + } + + return 0; +} + +static void kfd_remove_sysfs_node_tree(void) +{ + struct kfd_topology_device *dev; + + list_for_each_entry(dev, &topology_device_list, list) + kfd_remove_sysfs_node_entry(dev); +} + +static int kfd_topology_update_sysfs(void) +{ + int ret; + + pr_info("Creating topology SYSFS entries\n"); + if (sys_props.kobj_topology == 0) { + sys_props.kobj_topology = + kfd_alloc_struct(sys_props.kobj_topology); + if (!sys_props.kobj_topology) + return -ENOMEM; + + ret = kobject_init_and_add(sys_props.kobj_topology, + &sysprops_type, &kfd_device->kobj, + "topology"); + if (ret < 0) + return ret; + + sys_props.kobj_nodes = kobject_create_and_add("nodes", + sys_props.kobj_topology); + if (!sys_props.kobj_nodes) + return -ENOMEM; + + sys_props.attr_genid.name = "generation_id"; + sys_props.attr_genid.mode = KFD_SYSFS_FILE_MODE; + sysfs_attr_init(&sys_props.attr_genid); + ret = sysfs_create_file(sys_props.kobj_topology, + &sys_props.attr_genid); + if (ret < 0) + return ret; + + sys_props.attr_props.name = "system_properties"; + sys_props.attr_props.mode = KFD_SYSFS_FILE_MODE; + sysfs_attr_init(&sys_props.attr_props); + ret = sysfs_create_file(sys_props.kobj_topology, + &sys_props.attr_props); + if (ret < 0) + return ret; + } + + kfd_remove_sysfs_node_tree(); + + return kfd_build_sysfs_node_tree(); +} + +static void kfd_topology_release_sysfs(void) +{ + kfd_remove_sysfs_node_tree(); + if (sys_props.kobj_topology) { + sysfs_remove_file(sys_props.kobj_topology, + &sys_props.attr_genid); + sysfs_remove_file(sys_props.kobj_topology, + &sys_props.attr_props); + if (sys_props.kobj_nodes) { + kobject_del(sys_props.kobj_nodes); + kobject_put(sys_props.kobj_nodes); + sys_props.kobj_nodes = 0; + } + kobject_del(sys_props.kobj_topology); + kobject_put(sys_props.kobj_topology); + sys_props.kobj_topology = 0; + } +} + +int kfd_topology_init(void) +{ + void *crat_image = 0; + size_t image_size = 0; + int ret; + + /* + * Initialize the head for the topology device list + */ + INIT_LIST_HEAD(&topology_device_list); + init_rwsem(&topology_lock); + topology_crat_parsed = 0; + + memset(&sys_props, 0, sizeof(sys_props)); + + /* + * Get the CRAT image from the ACPI + */ + ret = kfd_topology_get_crat_acpi(crat_image, &image_size); + if (ret == 0 && image_size > 0) { + pr_info("Found CRAT image with size=%zd\n", image_size); + crat_image = kmalloc(image_size, GFP_KERNEL); + if (!crat_image) { + ret = -ENOMEM; + pr_err("No memory for allocating CRAT image\n"); + goto err; + } + ret = kfd_topology_get_crat_acpi(crat_image, &image_size); + + if (ret == 0) { + down_write(&topology_lock); + ret = kfd_parse_crat_table(crat_image); + if (ret == 0) + ret = kfd_topology_update_sysfs(); + up_write(&topology_lock); + } else { + pr_err("Couldn't get CRAT table size from ACPI\n"); + } + kfree(crat_image); + } else if (ret == -ENODATA) { + ret = 0; + } else { + pr_err("Couldn't get CRAT table size from ACPI\n"); + } + +err: + pr_info("Finished initializing topology ret=%d\n", ret); + return ret; +} + +void kfd_topology_shutdown(void) +{ + kfd_topology_release_sysfs(); + kfd_release_live_view(); +} + +static void kfd_debug_print_topology(void) +{ + struct kfd_topology_device *dev; + uint32_t i = 0; + + pr_info("DEBUG PRINT OF TOPOLOGY:"); + list_for_each_entry(dev, &topology_device_list, list) { + pr_info("Node: %d\n", i); + pr_info("\tGPU assigned: %s\n", (dev->gpu ? "yes" : "no")); + pr_info("\tCPU count: %d\n", dev->node_props.cpu_cores_count); + pr_info("\tSIMD count: %d", dev->node_props.simd_count); + i++; + } +} + +static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu) +{ + uint32_t hashout; + uint32_t buf[7]; + int i; + + if (!gpu) + return 0; + + buf[0] = gpu->pdev->devfn; + buf[1] = gpu->pdev->subsystem_vendor; + buf[2] = gpu->pdev->subsystem_device; + buf[3] = gpu->pdev->device; + buf[4] = gpu->pdev->bus->number; + buf[5] = (uint32_t)(kfd2kgd->get_vmem_size(gpu->kgd) & 0xffffffff); + buf[6] = (uint32_t)(kfd2kgd->get_vmem_size(gpu->kgd) >> 32); + + for (i = 0, hashout = 0; i < 7; i++) + hashout ^= hash_32(buf[i], KFD_GPU_ID_HASH_WIDTH); + + return hashout; +} + +static struct kfd_topology_device *kfd_assign_gpu(struct kfd_dev *gpu) +{ + struct kfd_topology_device *dev; + struct kfd_topology_device *out_dev = 0; + + BUG_ON(!gpu); + + list_for_each_entry(dev, &topology_device_list, list) + if (dev->gpu == 0 && dev->node_props.simd_count > 0) { + dev->gpu = gpu; + out_dev = dev; + break; + } + + return out_dev; +} + +static void kfd_notify_gpu_change(uint32_t gpu_id, int arrival) +{ + /* + * TODO: Generate an event for thunk about the arrival/removal + * of the GPU + */ +} + +int kfd_topology_add_device(struct kfd_dev *gpu) +{ + uint32_t gpu_id; + struct kfd_topology_device *dev; + int res; + + BUG_ON(!gpu); + + gpu_id = kfd_generate_gpu_id(gpu); + + pr_debug("kfd: Adding new GPU (ID: 0x%x) to topology\n", gpu_id); + + down_write(&topology_lock); + /* + * Try to assign the GPU to existing topology device (generated from + * CRAT table + */ + dev = kfd_assign_gpu(gpu); + if (!dev) { + pr_info("GPU was not found in the current topology. Extending.\n"); + kfd_debug_print_topology(); + dev = kfd_create_topology_device(); + if (!dev) { + res = -ENOMEM; + goto err; + } + dev->gpu = gpu; + + /* + * TODO: Make a call to retrieve topology information from the + * GPU vBIOS + */ + + /* + * Update the SYSFS tree, since we added another topology device + */ + if (kfd_topology_update_sysfs() < 0) + kfd_topology_release_sysfs(); + + } + + dev->gpu_id = gpu_id; + gpu->id = gpu_id; + dev->node_props.vendor_id = gpu->pdev->vendor; + dev->node_props.device_id = gpu->pdev->device; + dev->node_props.location_id = (gpu->pdev->bus->number << 24) + + (gpu->pdev->devfn & 0xffffff); + /* + * TODO: Retrieve max engine clock values from KGD + */ + + res = 0; + +err: + up_write(&topology_lock); + + if (res == 0) + kfd_notify_gpu_change(gpu_id, 1); + + return res; +} + +int kfd_topology_remove_device(struct kfd_dev *gpu) +{ + struct kfd_topology_device *dev; + uint32_t gpu_id; + int res = -ENODEV; + + BUG_ON(!gpu); + + down_write(&topology_lock); + + list_for_each_entry(dev, &topology_device_list, list) + if (dev->gpu == gpu) { + gpu_id = dev->gpu_id; + kfd_remove_sysfs_node_entry(dev); + kfd_release_topology_device(dev); + res = 0; + if (kfd_topology_update_sysfs() < 0) + kfd_topology_release_sysfs(); + break; + } + + up_write(&topology_lock); + + if (res == 0) + kfd_notify_gpu_change(gpu_id, 0); + + return res; +} + +/* + * When idx is out of bounds, the function will return NULL + */ +struct kfd_dev *kfd_topology_enum_kfd_devices(uint8_t idx) +{ + + struct kfd_topology_device *top_dev; + struct kfd_dev *device = NULL; + uint8_t device_idx = 0; + + down_read(&topology_lock); + + list_for_each_entry(top_dev, &topology_device_list, list) { + if (device_idx == idx) { + device = top_dev->gpu; + break; + } + + device_idx++; + } + + up_read(&topology_lock); + + return device; + +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h new file mode 100644 index 0000000..989624b --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h @@ -0,0 +1,168 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef __KFD_TOPOLOGY_H__ +#define __KFD_TOPOLOGY_H__ + +#include +#include +#include "kfd_priv.h" + +#define KFD_TOPOLOGY_PUBLIC_NAME_SIZE 128 + +#define HSA_CAP_HOT_PLUGGABLE 0x00000001 +#define HSA_CAP_ATS_PRESENT 0x00000002 +#define HSA_CAP_SHARED_WITH_GRAPHICS 0x00000004 +#define HSA_CAP_QUEUE_SIZE_POW2 0x00000008 +#define HSA_CAP_QUEUE_SIZE_32BIT 0x00000010 +#define HSA_CAP_QUEUE_IDLE_EVENT 0x00000020 +#define HSA_CAP_VA_LIMIT 0x00000040 +#define HSA_CAP_WATCH_POINTS_SUPPORTED 0x00000080 +#define HSA_CAP_WATCH_POINTS_TOTALBITS_MASK 0x00000f00 +#define HSA_CAP_WATCH_POINTS_TOTALBITS_SHIFT 8 +#define HSA_CAP_RESERVED 0xfffff000 + +struct kfd_node_properties { + uint32_t cpu_cores_count; + uint32_t simd_count; + uint32_t mem_banks_count; + uint32_t caches_count; + uint32_t io_links_count; + uint32_t cpu_core_id_base; + uint32_t simd_id_base; + uint32_t capability; + uint32_t max_waves_per_simd; + uint32_t lds_size_in_kb; + uint32_t gds_size_in_kb; + uint32_t wave_front_size; + uint32_t array_count; + uint32_t simd_arrays_per_engine; + uint32_t cu_per_simd_array; + uint32_t simd_per_cu; + uint32_t max_slots_scratch_cu; + uint32_t engine_id; + uint32_t vendor_id; + uint32_t device_id; + uint32_t location_id; + uint32_t max_engine_clk_fcompute; + uint32_t max_engine_clk_ccompute; + uint16_t marketing_name[KFD_TOPOLOGY_PUBLIC_NAME_SIZE]; +}; + +#define HSA_MEM_HEAP_TYPE_SYSTEM 0 +#define HSA_MEM_HEAP_TYPE_FB_PUBLIC 1 +#define HSA_MEM_HEAP_TYPE_FB_PRIVATE 2 +#define HSA_MEM_HEAP_TYPE_GPU_GDS 3 +#define HSA_MEM_HEAP_TYPE_GPU_LDS 4 +#define HSA_MEM_HEAP_TYPE_GPU_SCRATCH 5 + +#define HSA_MEM_FLAGS_HOT_PLUGGABLE 0x00000001 +#define HSA_MEM_FLAGS_NON_VOLATILE 0x00000002 +#define HSA_MEM_FLAGS_RESERVED 0xfffffffc + +struct kfd_mem_properties { + struct list_head list; + uint32_t heap_type; + uint64_t size_in_bytes; + uint32_t flags; + uint32_t width; + uint32_t mem_clk_max; + struct kobject *kobj; + struct attribute attr; +}; + +#define KFD_TOPOLOGY_CPU_SIBLINGS 256 + +#define HSA_CACHE_TYPE_DATA 0x00000001 +#define HSA_CACHE_TYPE_INSTRUCTION 0x00000002 +#define HSA_CACHE_TYPE_CPU 0x00000004 +#define HSA_CACHE_TYPE_HSACU 0x00000008 +#define HSA_CACHE_TYPE_RESERVED 0xfffffff0 + +struct kfd_cache_properties { + struct list_head list; + uint32_t processor_id_low; + uint32_t cache_level; + uint32_t cache_size; + uint32_t cacheline_size; + uint32_t cachelines_per_tag; + uint32_t cache_assoc; + uint32_t cache_latency; + uint32_t cache_type; + uint8_t sibling_map[KFD_TOPOLOGY_CPU_SIBLINGS]; + struct kobject *kobj; + struct attribute attr; +}; + +struct kfd_iolink_properties { + struct list_head list; + uint32_t iolink_type; + uint32_t ver_maj; + uint32_t ver_min; + uint32_t node_from; + uint32_t node_to; + uint32_t weight; + uint32_t min_latency; + uint32_t max_latency; + uint32_t min_bandwidth; + uint32_t max_bandwidth; + uint32_t rec_transfer_size; + uint32_t flags; + struct kobject *kobj; + struct attribute attr; +}; + +struct kfd_topology_device { + struct list_head list; + uint32_t gpu_id; + struct kfd_node_properties node_props; + uint32_t mem_bank_count; + struct list_head mem_props; + uint32_t cache_count; + struct list_head cache_props; + uint32_t io_link_count; + struct list_head io_link_props; + struct kfd_dev *gpu; + struct kobject *kobj_node; + struct kobject *kobj_mem; + struct kobject *kobj_cache; + struct kobject *kobj_iolink; + struct attribute attr_gpuid; + struct attribute attr_name; + struct attribute attr_props; +}; + +struct kfd_system_properties { + uint32_t num_devices; /* Number of H-NUMA nodes */ + uint32_t generation_count; + uint64_t platform_oem; + uint64_t platform_id; + uint64_t platform_rev; + struct kobject *kobj_topology; + struct kobject *kobj_nodes; + struct attribute attr_genid; + struct attribute attr_props; +}; + + + +#endif /* __KFD_TOPOLOGY_H__ */ -- cgit v0.10.2 From 19f6d2a660340d01bcdb7a09557efeeee28d1517 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Wed, 16 Jul 2014 23:25:31 +0300 Subject: amdkfd: Add basic modules to amdkfd This patch adds the process module and three helper modules: - kfd_process, which handles process which open /dev/kfd - kfd_doorbell, which provides helper functions for doorbell allocation, release and mapping to userspace - kfd_pasid, which provides helper functions for pasid allocation and release - kfd_aperture, which provides helper functions for managing the LDS, Local GPU memory and Scratch memory apertures of the process This patch only contains the basic kfd_process module, which doesn't contain the reference to the queue scheduler. This was done to allow easier code review. Also, this patch doesn't contain the calls to the IOMMU driver for binding the pasid to the device. Again, this was done to allow easier code review The kfd_process object is created when a process opens /dev/kfd and is closed when the mm_struct of that process is teared-down. v3: Removed kfd_vidmem.c file Replaced direct mmput call to mmu_notifier release Removed typedefs Moved bool field to end of the structure Added new kernel params for gart usage limitation Added initialization of sa manager Fixed debug messages Remove support for LDS in 32 bit Changed code to support mmap of doorbell pages from userspace Added documentation for apertures v4: Replaced RCU by SRCU for kfd_process list management v5: Move amdkfd from drm/radeon/ to drm/amd/ Rename kfd_aperture.c to kfd_flat_memory.c Protect against multiple init calls MQD size is H/W dependent so moved it to device info structure Rename kfd_mem_obj structure's members Use delayed function for process tear-down Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 978654f..e829a3f 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -4,6 +4,8 @@ ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/amd/include/ -amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o +amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ + kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ + kfd_process.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index d7c32eb..58441cd 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -38,6 +38,7 @@ static long kfd_ioctl(struct file *, unsigned int, unsigned long); static int kfd_open(struct inode *, struct file *); +static int kfd_mmap(struct file *, struct vm_area_struct *); static const char kfd_dev_name[] = "kfd"; @@ -46,6 +47,7 @@ static const struct file_operations kfd_fops = { .unlocked_ioctl = kfd_ioctl, .compat_ioctl = kfd_ioctl, .open = kfd_open, + .mmap = kfd_mmap, }; static int kfd_char_dev_major = -1; @@ -98,9 +100,22 @@ struct device *kfd_chardev(void) static int kfd_open(struct inode *inode, struct file *filep) { + struct kfd_process *process; + if (iminor(inode) != 0) return -ENODEV; + process = kfd_create_process(current); + if (IS_ERR(process)) + return PTR_ERR(process); + + process->is_32bit_user_mode = is_compat_task(); + + dev_dbg(kfd_device, "process %d opened, compat mode (32 bit) - %d\n", + process->pasid, process->is_32bit_user_mode); + + kfd_init_apertures(process); + return 0; } @@ -156,8 +171,9 @@ static long kfd_ioctl(struct file *filep, unsigned int cmd, unsigned long arg) "ioctl cmd 0x%x (#%d), arg 0x%lx\n", cmd, _IOC_NR(cmd), arg); - /* TODO: add function that retrieves process */ - process = NULL; + process = kfd_get_process(current); + if (IS_ERR(process)) + return PTR_ERR(process); switch (cmd) { case KFD_IOC_GET_VERSION: @@ -208,3 +224,14 @@ static long kfd_ioctl(struct file *filep, unsigned int cmd, unsigned long arg) return err; } + +static int kfd_mmap(struct file *filp, struct vm_area_struct *vma) +{ + struct kfd_process *process; + + process = kfd_get_process(current); + if (IS_ERR(process)) + return PTR_ERR(process); + + return kfd_doorbell_mmap(process, vma); +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 53b2e19..71a03f7 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -26,8 +26,11 @@ #include #include "kfd_priv.h" +#define MQD_SIZE_ALIGNED 768 + static const struct kfd_device_info kaveri_device_info = { .max_pasid_bits = 16, + .mqd_size_aligned = MQD_SIZE_ALIGNED }; struct kfd_deviceid { @@ -92,6 +95,7 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev) kfd->kgd = kgd; kfd->device_info = device_info; kfd->pdev = pdev; + kfd->init_complete = false; return kfd; } @@ -99,23 +103,53 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev) bool kgd2kfd_device_init(struct kfd_dev *kfd, const struct kgd2kfd_shared_resources *gpu_resources) { + unsigned int size; + kfd->shared_resources = *gpu_resources; - if (kfd_topology_add_device(kfd) != 0) - return false; + /* calculate max size of mqds needed for queues */ + size = max_num_of_processes * + max_num_of_queues_per_process * + kfd->device_info->mqd_size_aligned; + + /* add another 512KB for all other allocations on gart */ + size += 512 * 1024; + + if (kfd2kgd->init_sa_manager(kfd->kgd, size)) { + dev_err(kfd_device, + "Error initializing sa manager for device (%x:%x)\n", + kfd->pdev->vendor, kfd->pdev->device); + goto out; + } + + kfd_doorbell_init(kfd); + + if (kfd_topology_add_device(kfd) != 0) { + dev_err(kfd_device, + "Error adding device (%x:%x) to topology\n", + kfd->pdev->vendor, kfd->pdev->device); + goto kfd_topology_add_device_error; + } + kfd->init_complete = true; dev_info(kfd_device, "added device (%x:%x)\n", kfd->pdev->vendor, kfd->pdev->device); - return true; + goto out; + +kfd_topology_add_device_error: + kfd2kgd->fini_sa_manager(kfd->kgd); + dev_err(kfd_device, + "device (%x:%x) NOT added due to errors\n", + kfd->pdev->vendor, kfd->pdev->device); +out: + return kfd->init_complete; } void kgd2kfd_device_exit(struct kfd_dev *kfd) { - int err = kfd_topology_remove_device(kfd); - - BUG_ON(err != 0); + kfd_topology_remove_device(kfd); kfree(kfd); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c new file mode 100644 index 0000000..0dcb787 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c @@ -0,0 +1,255 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "kfd_priv.h" +#include +#include +#include + +/* + * This extension supports a kernel level doorbells management for + * the kernel queues. + * Basically the last doorbells page is devoted to kernel queues + * and that's assures that any user process won't get access to the + * kernel doorbells page + */ +static DEFINE_MUTEX(doorbell_mutex); +static unsigned long doorbell_available_index[ + DIV_ROUND_UP(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS, BITS_PER_LONG)] = { 0 }; + +#define KERNEL_DOORBELL_PASID 1 +#define KFD_SIZE_OF_DOORBELL_IN_BYTES 4 + +/* + * Each device exposes a doorbell aperture, a PCI MMIO aperture that + * receives 32-bit writes that are passed to queues as wptr values. + * The doorbells are intended to be written by applications as part + * of queueing work on user-mode queues. + * We assign doorbells to applications in PAGE_SIZE-sized and aligned chunks. + * We map the doorbell address space into user-mode when a process creates + * its first queue on each device. + * Although the mapping is done by KFD, it is equivalent to an mmap of + * the /dev/kfd with the particular device encoded in the mmap offset. + * There will be other uses for mmap of /dev/kfd, so only a range of + * offsets (KFD_MMAP_DOORBELL_START-END) is used for doorbells. + */ + +/* # of doorbell bytes allocated for each process. */ +static inline size_t doorbell_process_allocation(void) +{ + return roundup(KFD_SIZE_OF_DOORBELL_IN_BYTES * + KFD_MAX_NUM_OF_QUEUES_PER_PROCESS, + PAGE_SIZE); +} + +/* Doorbell calculations for device init. */ +void kfd_doorbell_init(struct kfd_dev *kfd) +{ + size_t doorbell_start_offset; + size_t doorbell_aperture_size; + size_t doorbell_process_limit; + + /* + * We start with calculations in bytes because the input data might + * only be byte-aligned. + * Only after we have done the rounding can we assume any alignment. + */ + + doorbell_start_offset = + roundup(kfd->shared_resources.doorbell_start_offset, + doorbell_process_allocation()); + + doorbell_aperture_size = + rounddown(kfd->shared_resources.doorbell_aperture_size, + doorbell_process_allocation()); + + if (doorbell_aperture_size > doorbell_start_offset) + doorbell_process_limit = + (doorbell_aperture_size - doorbell_start_offset) / + doorbell_process_allocation(); + else + doorbell_process_limit = 0; + + kfd->doorbell_base = kfd->shared_resources.doorbell_physical_address + + doorbell_start_offset; + + kfd->doorbell_id_offset = doorbell_start_offset / sizeof(u32); + kfd->doorbell_process_limit = doorbell_process_limit - 1; + + kfd->doorbell_kernel_ptr = ioremap(kfd->doorbell_base, + doorbell_process_allocation()); + + BUG_ON(!kfd->doorbell_kernel_ptr); + + pr_debug("kfd: doorbell initialization:\n"); + pr_debug("kfd: doorbell base == 0x%08lX\n", + (uintptr_t)kfd->doorbell_base); + + pr_debug("kfd: doorbell_id_offset == 0x%08lX\n", + kfd->doorbell_id_offset); + + pr_debug("kfd: doorbell_process_limit == 0x%08lX\n", + doorbell_process_limit); + + pr_debug("kfd: doorbell_kernel_offset == 0x%08lX\n", + (uintptr_t)kfd->doorbell_base); + + pr_debug("kfd: doorbell aperture size == 0x%08lX\n", + kfd->shared_resources.doorbell_aperture_size); + + pr_debug("kfd: doorbell kernel address == 0x%08lX\n", + (uintptr_t)kfd->doorbell_kernel_ptr); +} + +int kfd_doorbell_mmap(struct kfd_process *process, struct vm_area_struct *vma) +{ + phys_addr_t address; + struct kfd_dev *dev; + + /* + * For simplicitly we only allow mapping of the entire doorbell + * allocation of a single device & process. + */ + if (vma->vm_end - vma->vm_start != doorbell_process_allocation()) + return -EINVAL; + + /* Find kfd device according to gpu id */ + dev = kfd_device_by_id(vma->vm_pgoff); + if (dev == NULL) + return -EINVAL; + + /* Find if pdd exists for combination of process and gpu id */ + if (!kfd_get_process_device_data(dev, process, 0)) + return -EINVAL; + + /* Calculate physical address of doorbell */ + address = kfd_get_process_doorbells(dev, process); + + vma->vm_flags |= VM_IO | VM_DONTCOPY | VM_DONTEXPAND | VM_NORESERVE | + VM_DONTDUMP | VM_PFNMAP; + + vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot); + + pr_debug("kfd: mapping doorbell page in kfd_doorbell_mmap\n" + " target user address == 0x%08llX\n" + " physical address == 0x%08llX\n" + " vm_flags == 0x%04lX\n" + " size == 0x%04lX\n", + (unsigned long long) vma->vm_start, address, vma->vm_flags, + doorbell_process_allocation()); + + + return io_remap_pfn_range(vma, + vma->vm_start, + address >> PAGE_SHIFT, + doorbell_process_allocation(), + vma->vm_page_prot); +} + + +/* get kernel iomem pointer for a doorbell */ +u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, + unsigned int *doorbell_off) +{ + u32 inx; + + BUG_ON(!kfd || !doorbell_off); + + mutex_lock(&doorbell_mutex); + inx = find_first_zero_bit(doorbell_available_index, + KFD_MAX_NUM_OF_QUEUES_PER_PROCESS); + + __set_bit(inx, doorbell_available_index); + mutex_unlock(&doorbell_mutex); + + if (inx >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) + return NULL; + + /* + * Calculating the kernel doorbell offset using "faked" kernel + * pasid that allocated for kernel queues only + */ + *doorbell_off = KERNEL_DOORBELL_PASID * (doorbell_process_allocation() / + sizeof(u32)) + inx; + + pr_debug("kfd: get kernel queue doorbell\n" + " doorbell offset == 0x%08d\n" + " kernel address == 0x%08lX\n", + *doorbell_off, (uintptr_t)(kfd->doorbell_kernel_ptr + inx)); + + return kfd->doorbell_kernel_ptr + inx; +} + +void kfd_release_kernel_doorbell(struct kfd_dev *kfd, u32 __iomem *db_addr) +{ + unsigned int inx; + + BUG_ON(!kfd || !db_addr); + + inx = (unsigned int)(db_addr - kfd->doorbell_kernel_ptr); + + mutex_lock(&doorbell_mutex); + __clear_bit(inx, doorbell_available_index); + mutex_unlock(&doorbell_mutex); +} + +inline void write_kernel_doorbell(u32 __iomem *db, u32 value) +{ + if (db) { + writel(value, db); + pr_debug("writing %d to doorbell address 0x%p\n", value, db); + } +} + +/* + * queue_ids are in the range [0,MAX_PROCESS_QUEUES) and are mapped 1:1 + * to doorbells with the process's doorbell page + */ +unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd, + struct kfd_process *process, + unsigned int queue_id) +{ + /* + * doorbell_id_offset accounts for doorbells taken by KGD. + * pasid * doorbell_process_allocation/sizeof(u32) adjusts + * to the process's doorbells + */ + return kfd->doorbell_id_offset + + process->pasid * (doorbell_process_allocation()/sizeof(u32)) + + queue_id; +} + +uint64_t kfd_get_number_elems(struct kfd_dev *kfd) +{ + uint64_t num_of_elems = (kfd->shared_resources.doorbell_aperture_size - + kfd->shared_resources.doorbell_start_offset) / + doorbell_process_allocation() + 1; + + return num_of_elems; + +} + +phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev, + struct kfd_process *process) +{ + return dev->doorbell_base + + process->pasid * doorbell_process_allocation(); +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c new file mode 100644 index 0000000..2dfc4c0 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c @@ -0,0 +1,355 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "kfd_priv.h" +#include +#include +#include + +/* + * The primary memory I/O features being added for revisions of gfxip + * beyond 7.0 (Kaveri) are: + * + * Access to ATC/IOMMU mapped memory w/ associated extension of VA to 48b + * + * “Flat” shader memory access – These are new shader vector memory + * operations that do not reference a T#/V# so a “pointer” is what is + * sourced from the vector gprs for direct access to memory. + * This pointer space has the Shared(LDS) and Private(Scratch) memory + * mapped into this pointer space as apertures. + * The hardware then determines how to direct the memory request + * based on what apertures the request falls in. + * + * Unaligned support and alignment check + * + * + * System Unified Address - SUA + * + * The standard usage for GPU virtual addresses are that they are mapped by + * a set of page tables we call GPUVM and these page tables are managed by + * a combination of vidMM/driver software components. The current virtual + * address (VA) range for GPUVM is 40b. + * + * As of gfxip7.1 and beyond we’re adding the ability for compute memory + * clients (CP/RLC, DMA, SHADER(ifetch, scalar, and vector ops)) to access + * the same page tables used by host x86 processors and that are managed by + * the operating system. This is via a technique and hardware called ATC/IOMMU. + * The GPU has the capability of accessing both the GPUVM and ATC address + * spaces for a given VMID (process) simultaneously and we call this feature + * system unified address (SUA). + * + * There are three fundamental address modes of operation for a given VMID + * (process) on the GPU: + * + * HSA64 – 64b pointers and the default address space is ATC + * HSA32 – 32b pointers and the default address space is ATC + * GPUVM – 64b pointers and the default address space is GPUVM (driver + * model mode) + * + * + * HSA64 - ATC/IOMMU 64b + * + * A 64b pointer in the AMD64/IA64 CPU architecture is not fully utilized + * by the CPU so an AMD CPU can only access the high area + * (VA[63:47] == 0x1FFFF) and low area (VA[63:47 == 0) of the address space + * so the actual VA carried to translation is 48b. There is a “hole” in + * the middle of the 64b VA space. + * + * The GPU not only has access to all of the CPU accessible address space via + * ATC/IOMMU, but it also has access to the GPUVM address space. The “system + * unified address” feature (SUA) is the mapping of GPUVM and ATC address + * spaces into a unified pointer space. The method we take for 64b mode is + * to map the full 40b GPUVM address space into the hole of the 64b address + * space. + + * The GPUVM_Base/GPUVM_Limit defines the aperture in the 64b space where we + * direct requests to be translated via GPUVM page tables instead of the + * IOMMU path. + * + * + * 64b to 49b Address conversion + * + * Note that there are still significant portions of unused regions (holes) + * in the 64b address space even for the GPU. There are several places in + * the pipeline (sw and hw), we wish to compress the 64b virtual address + * to a 49b address. This 49b address is constituted of an “ATC” bit + * plus a 48b virtual address. This 49b address is what is passed to the + * translation hardware. ATC==0 means the 48b address is a GPUVM address + * (max of 2^40 – 1) intended to be translated via GPUVM page tables. + * ATC==1 means the 48b address is intended to be translated via IOMMU + * page tables. + * + * A 64b pointer is compared to the apertures that are defined (Base/Limit), in + * this case the GPUVM aperture (red) is defined and if a pointer falls in this + * aperture, we subtract the GPUVM_Base address and set the ATC bit to zero + * as part of the 64b to 49b conversion. + * + * Where this 64b to 49b conversion is done is a function of the usage. + * Most GPU memory access is via memory objects where the driver builds + * a descriptor which consists of a base address and a memory access by + * the GPU usually consists of some kind of an offset or Cartesian coordinate + * that references this memory descriptor. This is the case for shader + * instructions that reference the T# or V# constants, or for specified + * locations of assets (ex. the shader program location). In these cases + * the driver is what handles the 64b to 49b conversion and the base + * address in the descriptor (ex. V# or T# or shader program location) + * is defined as a 48b address w/ an ATC bit. For this usage a given + * memory object cannot straddle multiple apertures in the 64b address + * space. For example a shader program cannot jump in/out between ATC + * and GPUVM space. + * + * In some cases we wish to pass a 64b pointer to the GPU hardware and + * the GPU hw does the 64b to 49b conversion before passing memory + * requests to the cache/memory system. This is the case for the + * S_LOAD and FLAT_* shader memory instructions where we have 64b pointers + * in scalar and vector GPRs respectively. + * + * In all cases (no matter where the 64b -> 49b conversion is done), the gfxip + * hardware sends a 48b address along w/ an ATC bit, to the memory controller + * on the memory request interfaces. + * + * _MC_rdreq_atc // read request ATC bit + * + * 0 : _MC_rdreq_addr is a GPUVM VA + * + * 1 : _MC_rdreq_addr is a ATC VA + * + * + * “Spare” aperture (APE1) + * + * We use the GPUVM aperture to differentiate ATC vs. GPUVM, but we also use + * apertures to set the Mtype field for S_LOAD/FLAT_* ops which is input to the + * config tables for setting cache policies. The “spare” (APE1) aperture is + * motivated by getting a different Mtype from the default. + * The default aperture isn’t an actual base/limit aperture; it is just the + * address space that doesn’t hit any defined base/limit apertures. + * The following diagram is a complete picture of the gfxip7.x SUA apertures. + * The APE1 can be placed either below or above + * the hole (cannot be in the hole). + * + * + * General Aperture definitions and rules + * + * An aperture register definition consists of a Base, Limit, Mtype, and + * usually an ATC bit indicating which translation tables that aperture uses. + * In all cases (for SUA and DUA apertures discussed later), aperture base + * and limit definitions are 64KB aligned. + * + * _Base[63:0] = { _Base_register[63:16], 0x0000 } + * + * _Limit[63:0] = { _Limit_register[63:16], 0xFFFF } + * + * The base and limit are considered inclusive to an aperture so being + * inside an aperture means (address >= Base) AND (address <= Limit). + * + * In no case is a payload that straddles multiple apertures expected to work. + * For example a load_dword_x4 that starts in one aperture and ends in another, + * does not work. For the vector FLAT_* ops we have detection capability in + * the shader for reporting a “memory violation” back to the + * SQ block for use in traps. + * A memory violation results when an op falls into the hole, + * or a payload straddles multiple apertures. The S_LOAD instruction + * does not have this detection. + * + * Apertures cannot overlap. + * + * + * + * HSA32 - ATC/IOMMU 32b + * + * For HSA32 mode, the pointers are interpreted as 32 bits and use a single GPR + * instead of two for the S_LOAD and FLAT_* ops. The entire GPUVM space of 40b + * will not fit so there is only partial visibility to the GPUVM + * space (defined by the aperture) for S_LOAD and FLAT_* ops. + * There is no spare (APE1) aperture for HSA32 mode. + * + * + * GPUVM 64b mode (driver model) + * + * This mode is related to HSA64 in that the difference really is that + * the default aperture is GPUVM (ATC==0) and not ATC space. + * We have gfxip7.x hardware that has FLAT_* and S_LOAD support for + * SUA GPUVM mode, but does not support HSA32/HSA64. + * + * + * Device Unified Address - DUA + * + * Device unified address (DUA) is the name of the feature that maps the + * Shared(LDS) memory and Private(Scratch) memory into the overall address + * space for use by the new FLAT_* vector memory ops. The Shared and + * Private memories are mapped as apertures into the address space, + * and the hardware detects when a FLAT_* memory request is to be redirected + * to the LDS or Scratch memory when it falls into one of these apertures. + * Like the SUA apertures, the Shared/Private apertures are 64KB aligned and + * the base/limit is “in” the aperture. For both HSA64 and GPUVM SUA modes, + * the Shared/Private apertures are always placed in a limited selection of + * options in the hole of the 64b address space. For HSA32 mode, the + * Shared/Private apertures can be placed anywhere in the 32b space + * except at 0. + * + * + * HSA64 Apertures for FLAT_* vector ops + * + * For HSA64 SUA mode, the Shared and Private apertures are always placed + * in the hole w/ a limited selection of possible locations. The requests + * that fall in the private aperture are expanded as a function of the + * work-item id (tid) and redirected to the location of the + * “hidden private memory”. The hidden private can be placed in either GPUVM + * or ATC space. The addresses that fall in the shared aperture are + * re-directed to the on-chip LDS memory hardware. + * + * + * HSA32 Apertures for FLAT_* vector ops + * + * In HSA32 mode, the Private and Shared apertures can be placed anywhere + * in the 32b space except at 0 (Private or Shared Base at zero disables + * the apertures). If the base address of the apertures are non-zero + * (ie apertures exists), the size is always 64KB. + * + * + * GPUVM Apertures for FLAT_* vector ops + * + * In GPUVM mode, the Shared/Private apertures are specified identically + * to HSA64 mode where they are always in the hole at a limited selection + * of locations. + * + * + * Aperture Definitions for SUA and DUA + * + * The interpretation of the aperture register definitions for a given + * VMID is a function of the “SUA Mode” which is one of HSA64, HSA32, or + * GPUVM64 discussed in previous sections. The mode is first decoded, and + * then the remaining register decode is a function of the mode. + * + * + * SUA Mode Decode + * + * For the S_LOAD and FLAT_* shader operations, the SUA mode is decoded from + * the COMPUTE_DISPATCH_INITIATOR:DATA_ATC bit and + * the SH_MEM_CONFIG:PTR32 bits. + * + * COMPUTE_DISPATCH_INITIATOR:DATA_ATC SH_MEM_CONFIG:PTR32 Mode + * + * 1 0 HSA64 + * + * 1 1 HSA32 + * + * 0 X GPUVM64 + * + * In general the hardware will ignore the PTR32 bit and treat + * as “0” whenever DATA_ATC = “0”, but sw should set PTR32=0 + * when DATA_ATC=0. + * + * The DATA_ATC bit is only set for compute dispatches. + * All “Draw” dispatches are hardcoded to GPUVM64 mode + * for FLAT_* / S_LOAD operations. + */ + +#define MAKE_GPUVM_APP_BASE(gpu_num) \ + (((uint64_t)(gpu_num) << 61) + 0x1000000000000) + +#define MAKE_GPUVM_APP_LIMIT(base) \ + (((uint64_t)(base) & 0xFFFFFF0000000000) | 0xFFFFFFFFFF) + +#define MAKE_SCRATCH_APP_BASE(gpu_num) \ + (((uint64_t)(gpu_num) << 61) + 0x100000000) + +#define MAKE_SCRATCH_APP_LIMIT(base) \ + (((uint64_t)base & 0xFFFFFFFF00000000) | 0xFFFFFFFF) + +#define MAKE_LDS_APP_BASE(gpu_num) \ + (((uint64_t)(gpu_num) << 61) + 0x0) +#define MAKE_LDS_APP_LIMIT(base) \ + (((uint64_t)(base) & 0xFFFFFFFF00000000) | 0xFFFFFFFF) + +int kfd_init_apertures(struct kfd_process *process) +{ + uint8_t id = 0; + struct kfd_dev *dev; + struct kfd_process_device *pdd; + + mutex_lock(&process->mutex); + + /*Iterating over all devices*/ + while ((dev = kfd_topology_enum_kfd_devices(id)) != NULL && + id < NUM_OF_SUPPORTED_GPUS) { + + pdd = kfd_get_process_device_data(dev, process, 1); + + /* + * For 64 bit process aperture will be statically reserved in + * the x86_64 non canonical process address space + * amdkfd doesn't currently support apertures for 32 bit process + */ + if (process->is_32bit_user_mode) { + pdd->lds_base = pdd->lds_limit = 0; + pdd->gpuvm_base = pdd->gpuvm_limit = 0; + pdd->scratch_base = pdd->scratch_limit = 0; + } else { + /* + * node id couldn't be 0 - the three MSB bits of + * aperture shoudn't be 0 + */ + pdd->lds_base = MAKE_LDS_APP_BASE(id + 1); + + pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base); + + pdd->gpuvm_base = MAKE_GPUVM_APP_BASE(id + 1); + + pdd->gpuvm_limit = + MAKE_GPUVM_APP_LIMIT(pdd->gpuvm_base); + + pdd->scratch_base = MAKE_SCRATCH_APP_BASE(id + 1); + + pdd->scratch_limit = + MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base); + } + + dev_dbg(kfd_device, "node id %u\n", id); + dev_dbg(kfd_device, "gpu id %u\n", pdd->dev->id); + dev_dbg(kfd_device, "lds_base %llX\n", pdd->lds_base); + dev_dbg(kfd_device, "lds_limit %llX\n", pdd->lds_limit); + dev_dbg(kfd_device, "gpuvm_base %llX\n", pdd->gpuvm_base); + dev_dbg(kfd_device, "gpuvm_limit %llX\n", pdd->gpuvm_limit); + dev_dbg(kfd_device, "scratch_base %llX\n", pdd->scratch_base); + dev_dbg(kfd_device, "scratch_limit %llX\n", pdd->scratch_limit); + + id++; + } + + mutex_unlock(&process->mutex); + + return 0; +} + + diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c index 45654be..a05116b 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -22,7 +22,6 @@ #include #include -#include #include #include #include "kfd_priv.h" @@ -46,6 +45,16 @@ static const struct kgd2kfd_calls kgd2kfd = { .resume = kgd2kfd_resume, }; +int max_num_of_processes = KFD_MAX_NUM_OF_PROCESSES_DEFAULT; +module_param(max_num_of_processes, int, 0444); +MODULE_PARM_DESC(max_num_of_processes, + "Kernel cmdline parameter that defines the amdkfd maximum number of supported processes"); + +int max_num_of_queues_per_process = KFD_MAX_NUM_OF_QUEUES_PER_PROCESS_DEFAULT; +module_param(max_num_of_queues_per_process, int, 0444); +MODULE_PARM_DESC(max_num_of_queues_per_process, + "Kernel cmdline parameter that defines the amdkfd maximum number of supported queues per process"); + bool kgd2kfd_init(unsigned interface_version, const struct kfd2kgd_calls *f2g, const struct kgd2kfd_calls **g2f) @@ -57,6 +66,10 @@ bool kgd2kfd_init(unsigned interface_version, if (interface_version != KFD_INTERFACE_VERSION) return false; + /* Protection against multiple amd kgd loads */ + if (kfd2kgd) + return true; + kfd2kgd = f2g; *g2f = &kgd2kfd; @@ -72,6 +85,26 @@ static int __init kfd_module_init(void) { int err; + kfd2kgd = NULL; + + /* Verify module parameters */ + if ((max_num_of_processes < 0) || + (max_num_of_processes > KFD_MAX_NUM_OF_PROCESSES)) { + pr_err("kfd: max_num_of_processes must be between 0 to KFD_MAX_NUM_OF_PROCESSES\n"); + return -1; + } + + if ((max_num_of_queues_per_process < 0) || + (max_num_of_queues_per_process > + KFD_MAX_NUM_OF_QUEUES_PER_PROCESS)) { + pr_err("kfd: max_num_of_queues_per_process must be between 0 to KFD_MAX_NUM_OF_QUEUES_PER_PROCESS\n"); + return -1; + } + + err = kfd_pasid_init(); + if (err < 0) + goto err_pasid; + err = kfd_chardev_init(); if (err < 0) goto err_ioctl; @@ -80,6 +113,8 @@ static int __init kfd_module_init(void) if (err < 0) goto err_topology; + kfd_process_create_wq(); + dev_info(kfd_device, "Initialized module\n"); return 0; @@ -87,13 +122,17 @@ static int __init kfd_module_init(void) err_topology: kfd_chardev_exit(); err_ioctl: + kfd_pasid_exit(); +err_pasid: return err; } static void __exit kfd_module_exit(void) { + kfd_process_destroy_wq(); kfd_topology_shutdown(); kfd_chardev_exit(); + kfd_pasid_exit(); dev_info(kfd_device, "Removed module\n"); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c b/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c new file mode 100644 index 0000000..2458ab7 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c @@ -0,0 +1,97 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include +#include "kfd_priv.h" + +static unsigned long *pasid_bitmap; +static unsigned int pasid_limit; +static DEFINE_MUTEX(pasid_mutex); + +int kfd_pasid_init(void) +{ + pasid_limit = max_num_of_processes; + + pasid_bitmap = kzalloc(DIV_ROUND_UP(pasid_limit, BITS_PER_BYTE), + GFP_KERNEL); + if (!pasid_bitmap) + return -ENOMEM; + + set_bit(0, pasid_bitmap); /* PASID 0 is reserved. */ + + return 0; +} + +void kfd_pasid_exit(void) +{ + kfree(pasid_bitmap); +} + +bool kfd_set_pasid_limit(unsigned int new_limit) +{ + if (new_limit < pasid_limit) { + bool ok; + + mutex_lock(&pasid_mutex); + + /* ensure that no pasids >= new_limit are in-use */ + ok = (find_next_bit(pasid_bitmap, pasid_limit, new_limit) == + pasid_limit); + if (ok) + pasid_limit = new_limit; + + mutex_unlock(&pasid_mutex); + + return ok; + } + + return true; +} + +inline unsigned int kfd_get_pasid_limit(void) +{ + return pasid_limit; +} + +unsigned int kfd_pasid_alloc(void) +{ + unsigned int found; + + mutex_lock(&pasid_mutex); + + found = find_first_zero_bit(pasid_bitmap, pasid_limit); + if (found == pasid_limit) + found = 0; + else + set_bit(found, pasid_bitmap); + + mutex_unlock(&pasid_mutex); + + return found; +} + +void kfd_pasid_free(unsigned int pasid) +{ + BUG_ON(pasid == 0 || pasid >= pasid_limit); + clear_bit(pasid, pasid_bitmap); +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index b58b86d..77d15db 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -30,6 +30,7 @@ #include #include #include +#include #include #define KFD_SYSFS_FILE_MODE 0444 @@ -41,9 +42,26 @@ #define kfd_alloc_struct(ptr_to_struct) \ ((typeof(ptr_to_struct)) kzalloc(sizeof(*ptr_to_struct), GFP_KERNEL)) +/* Kernel module parameter to specify maximum number of supported processes */ +extern int max_num_of_processes; + +#define KFD_MAX_NUM_OF_PROCESSES_DEFAULT 32 +#define KFD_MAX_NUM_OF_PROCESSES 512 + +/* + * Kernel module parameter to specify maximum number of supported queues + * per process + */ +extern int max_num_of_queues_per_process; + +#define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS_DEFAULT 128 +#define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS 1024 + + struct kfd_device_info { unsigned int max_pasid_bits; size_t ih_ring_entry_size; + uint16_t mqd_size_aligned; }; struct kfd_dev { @@ -54,6 +72,21 @@ struct kfd_dev { unsigned int id; /* topology stub index */ + phys_addr_t doorbell_base; /* Start of actual doorbells used by + * KFD. It is aligned for mapping + * into user mode + */ + size_t doorbell_id_offset; /* Doorbell offset (from KFD doorbell + * to HW doorbell, GFX reserved some + * at the start) + */ + size_t doorbell_process_limit; /* Number of processes we have doorbell + * space for. + */ + u32 __iomem *doorbell_kernel_ptr; /* This is a pointer for a doorbells + * page used by kernel queue + */ + struct kgd2kfd_shared_resources shared_resources; bool init_complete; @@ -69,15 +102,122 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd); extern const struct kfd2kgd_calls *kfd2kgd; +struct kfd_mem_obj { + void *bo; + uint64_t gpu_addr; + uint32_t *cpu_ptr; +}; + +enum kfd_mempool { + KFD_MEMPOOL_SYSTEM_CACHEABLE = 1, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE = 2, + KFD_MEMPOOL_FRAMEBUFFER = 3, +}; + /* Character device interface */ int kfd_chardev_init(void); void kfd_chardev_exit(void); struct device *kfd_chardev(void); + +/* Data that is per-process-per device. */ +struct kfd_process_device { + /* + * List of all per-device data for a process. + * Starts from kfd_process.per_device_data. + */ + struct list_head per_device_list; + + /* The device that owns this data. */ + struct kfd_dev *dev; + + + /*Apertures*/ + uint64_t lds_base; + uint64_t lds_limit; + uint64_t gpuvm_base; + uint64_t gpuvm_limit; + uint64_t scratch_base; + uint64_t scratch_limit; + + /* Is this process/pasid bound to this device? (amd_iommu_bind_pasid) */ + bool bound; +}; + /* Process data */ struct kfd_process { + /* + * kfd_process are stored in an mm_struct*->kfd_process* + * hash table (kfd_processes in kfd_process.c) + */ + struct hlist_node kfd_processes; + + struct mm_struct *mm; + + struct mutex mutex; + + /* + * In any process, the thread that started main() is the lead + * thread and outlives the rest. + * It is here because amd_iommu_bind_pasid wants a task_struct. + */ + struct task_struct *lead_thread; + + /* We want to receive a notification when the mm_struct is destroyed */ + struct mmu_notifier mmu_notifier; + + /* Use for delayed freeing of kfd_process structure */ + struct rcu_head rcu; + + unsigned int pasid; + + /* + * List of kfd_process_device structures, + * one for each device the process is using. + */ + struct list_head per_device_data; + + /* The process's queues. */ + size_t queue_array_size; + + /* Size is queue_array_size, up to MAX_PROCESS_QUEUES. */ + struct kfd_queue **queues; + + unsigned long allocated_queue_bitmap[DIV_ROUND_UP(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS, BITS_PER_LONG)]; + + /*Is the user space process 32 bit?*/ + bool is_32bit_user_mode; }; +void kfd_process_create_wq(void); +void kfd_process_destroy_wq(void); +struct kfd_process *kfd_create_process(const struct task_struct *); +struct kfd_process *kfd_get_process(const struct task_struct *); + +struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, + struct kfd_process *p, + int create_pdd); + +/* PASIDs */ +int kfd_pasid_init(void); +void kfd_pasid_exit(void); +bool kfd_set_pasid_limit(unsigned int new_limit); +unsigned int kfd_get_pasid_limit(void); +unsigned int kfd_pasid_alloc(void); +void kfd_pasid_free(unsigned int pasid); + +/* Doorbells */ +void kfd_doorbell_init(struct kfd_dev *kfd); +int kfd_doorbell_mmap(struct kfd_process *process, struct vm_area_struct *vma); +u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, + unsigned int *doorbell_off); +void kfd_release_kernel_doorbell(struct kfd_dev *kfd, u32 __iomem *db_addr); +u32 read_kernel_doorbell(u32 __iomem *db); +void write_kernel_doorbell(u32 __iomem *db, u32 value); +unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd, + struct kfd_process *process, + unsigned int queue_id); + extern struct device *kfd_device; /* Topology */ @@ -96,4 +236,11 @@ void kgd2kfd_interrupt(struct kfd_dev *dev, const void *ih_ring_entry); void kgd2kfd_suspend(struct kfd_dev *dev); int kgd2kfd_resume(struct kfd_dev *dev); +/* amdkfd Apertures */ +int kfd_init_apertures(struct kfd_process *process); + +uint64_t kfd_get_number_elems(struct kfd_dev *kfd); +phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev, + struct kfd_process *process); + #endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c new file mode 100644 index 0000000..5596f69 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -0,0 +1,383 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include +#include +#include +#include +struct mm_struct; + +#include "kfd_priv.h" + +/* + * Initial size for the array of queues. + * The allocated size is doubled each time + * it is exceeded up to MAX_PROCESS_QUEUES. + */ +#define INITIAL_QUEUE_ARRAY_SIZE 16 + +/* + * List of struct kfd_process (field kfd_process). + * Unique/indexed by mm_struct* + */ +#define KFD_PROCESS_TABLE_SIZE 5 /* bits: 32 entries */ +static DEFINE_HASHTABLE(kfd_processes_table, KFD_PROCESS_TABLE_SIZE); +static DEFINE_MUTEX(kfd_processes_mutex); + +DEFINE_STATIC_SRCU(kfd_processes_srcu); + +static struct workqueue_struct *kfd_process_wq; + +struct kfd_process_release_work { + struct work_struct kfd_work; + struct kfd_process *p; +}; + +static struct kfd_process *find_process(const struct task_struct *thread); +static struct kfd_process *create_process(const struct task_struct *thread); + +void kfd_process_create_wq(void) +{ + if (!kfd_process_wq) + kfd_process_wq = create_workqueue("kfd_process_wq"); +} + +void kfd_process_destroy_wq(void) +{ + if (kfd_process_wq) { + flush_workqueue(kfd_process_wq); + destroy_workqueue(kfd_process_wq); + kfd_process_wq = NULL; + } +} + +struct kfd_process *kfd_create_process(const struct task_struct *thread) +{ + struct kfd_process *process; + + BUG_ON(!kfd_process_wq); + + if (thread->mm == NULL) + return ERR_PTR(-EINVAL); + + /* Only the pthreads threading model is supported. */ + if (thread->group_leader->mm != thread->mm) + return ERR_PTR(-EINVAL); + + /* Take mmap_sem because we call __mmu_notifier_register inside */ + down_write(&thread->mm->mmap_sem); + + /* + * take kfd processes mutex before starting of process creation + * so there won't be a case where two threads of the same process + * create two kfd_process structures + */ + mutex_lock(&kfd_processes_mutex); + + /* A prior open of /dev/kfd could have already created the process. */ + process = find_process(thread); + if (process) + pr_debug("kfd: process already found\n"); + + if (!process) + process = create_process(thread); + + mutex_unlock(&kfd_processes_mutex); + + up_write(&thread->mm->mmap_sem); + + return process; +} + +struct kfd_process *kfd_get_process(const struct task_struct *thread) +{ + struct kfd_process *process; + + if (thread->mm == NULL) + return ERR_PTR(-EINVAL); + + /* Only the pthreads threading model is supported. */ + if (thread->group_leader->mm != thread->mm) + return ERR_PTR(-EINVAL); + + process = find_process(thread); + + return process; +} + +static struct kfd_process *find_process_by_mm(const struct mm_struct *mm) +{ + struct kfd_process *process; + + hash_for_each_possible_rcu(kfd_processes_table, process, + kfd_processes, (uintptr_t)mm) + if (process->mm == mm) + return process; + + return NULL; +} + +static struct kfd_process *find_process(const struct task_struct *thread) +{ + struct kfd_process *p; + int idx; + + idx = srcu_read_lock(&kfd_processes_srcu); + p = find_process_by_mm(thread->mm); + srcu_read_unlock(&kfd_processes_srcu, idx); + + return p; +} + +static void kfd_process_wq_release(struct work_struct *work) +{ + struct kfd_process_release_work *my_work; + struct kfd_process_device *pdd, *temp; + struct kfd_process *p; + + my_work = (struct kfd_process_release_work *) work; + + p = my_work->p; + + mutex_lock(&p->mutex); + + list_for_each_entry_safe(pdd, temp, &p->per_device_data, + per_device_list) { + list_del(&pdd->per_device_list); + + kfree(pdd); + } + + kfd_pasid_free(p->pasid); + + mutex_unlock(&p->mutex); + + mutex_destroy(&p->mutex); + + kfree(p->queues); + + kfree(p); + + kfree((void *)work); +} + +static void kfd_process_destroy_delayed(struct rcu_head *rcu) +{ + struct kfd_process_release_work *work; + struct kfd_process *p; + + BUG_ON(!kfd_process_wq); + + p = container_of(rcu, struct kfd_process, rcu); + BUG_ON(atomic_read(&p->mm->mm_count) <= 0); + + mmdrop(p->mm); + + work = (struct kfd_process_release_work *) + kmalloc(sizeof(struct kfd_process_release_work), GFP_KERNEL); + + if (work) { + INIT_WORK((struct work_struct *) work, kfd_process_wq_release); + work->p = p; + queue_work(kfd_process_wq, (struct work_struct *) work); + } +} + +static void kfd_process_notifier_release(struct mmu_notifier *mn, + struct mm_struct *mm) +{ + struct kfd_process *p; + + /* + * The kfd_process structure can not be free because the + * mmu_notifier srcu is read locked + */ + p = container_of(mn, struct kfd_process, mmu_notifier); + BUG_ON(p->mm != mm); + + mutex_lock(&kfd_processes_mutex); + hash_del_rcu(&p->kfd_processes); + mutex_unlock(&kfd_processes_mutex); + synchronize_srcu(&kfd_processes_srcu); + + /* + * Because we drop mm_count inside kfd_process_destroy_delayed + * and because the mmu_notifier_unregister function also drop + * mm_count we need to take an extra count here. + */ + atomic_inc(&p->mm->mm_count); + mmu_notifier_unregister_no_release(&p->mmu_notifier, p->mm); + mmu_notifier_call_srcu(&p->rcu, &kfd_process_destroy_delayed); +} + +static const struct mmu_notifier_ops kfd_process_mmu_notifier_ops = { + .release = kfd_process_notifier_release, +}; + +static struct kfd_process *create_process(const struct task_struct *thread) +{ + struct kfd_process *process; + int err = -ENOMEM; + + process = kzalloc(sizeof(*process), GFP_KERNEL); + + if (!process) + goto err_alloc_process; + + process->queues = kmalloc_array(INITIAL_QUEUE_ARRAY_SIZE, + sizeof(process->queues[0]), GFP_KERNEL); + if (!process->queues) + goto err_alloc_queues; + + process->pasid = kfd_pasid_alloc(); + if (process->pasid == 0) + goto err_alloc_pasid; + + mutex_init(&process->mutex); + + process->mm = thread->mm; + + /* register notifier */ + process->mmu_notifier.ops = &kfd_process_mmu_notifier_ops; + err = __mmu_notifier_register(&process->mmu_notifier, process->mm); + if (err) + goto err_mmu_notifier; + + hash_add_rcu(kfd_processes_table, &process->kfd_processes, + (uintptr_t)process->mm); + + process->lead_thread = thread->group_leader; + + process->queue_array_size = INITIAL_QUEUE_ARRAY_SIZE; + + INIT_LIST_HEAD(&process->per_device_data); + + return process; + +err_mmu_notifier: + kfd_pasid_free(process->pasid); +err_alloc_pasid: + kfree(process->queues); +err_alloc_queues: + kfree(process); +err_alloc_process: + return ERR_PTR(err); +} + +struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, + struct kfd_process *p, + int create_pdd) +{ + struct kfd_process_device *pdd = NULL; + + list_for_each_entry(pdd, &p->per_device_data, per_device_list) + if (pdd->dev == dev) + return pdd; + + if (create_pdd) { + pdd = kzalloc(sizeof(*pdd), GFP_KERNEL); + if (pdd != NULL) { + pdd->dev = dev; + list_add(&pdd->per_device_list, &p->per_device_data); + } + } + + return pdd; +} + +/* + * Direct the IOMMU to bind the process (specifically the pasid->mm) + * to the device. + * Unbinding occurs when the process dies or the device is removed. + * + * Assumes that the process lock is held. + */ +struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, + struct kfd_process *p) +{ + struct kfd_process_device *pdd = kfd_get_process_device_data(dev, p, 1); + + if (pdd == NULL) + return ERR_PTR(-ENOMEM); + + if (pdd->bound) + return pdd; + + pdd->bound = true; + + return pdd; +} + +void kfd_unbind_process_from_device(struct kfd_dev *dev, unsigned int pasid) +{ + struct kfd_process *p; + struct kfd_process_device *pdd; + int idx, i; + + BUG_ON(dev == NULL); + + idx = srcu_read_lock(&kfd_processes_srcu); + + hash_for_each_rcu(kfd_processes_table, i, p, kfd_processes) + if (p->pasid == pasid) + break; + + srcu_read_unlock(&kfd_processes_srcu, idx); + + BUG_ON(p->pasid != pasid); + + mutex_lock(&p->mutex); + + pdd = kfd_get_process_device_data(dev, p, 0); + + /* + * Just mark pdd as unbound, because we still need it to call + * amd_iommu_unbind_pasid() in when the process exits. + * We don't call amd_iommu_unbind_pasid() here + * because the IOMMU called us. + */ + if (pdd) + pdd->bound = false; + + mutex_unlock(&p->mutex); +} + +struct kfd_process_device *kfd_get_first_process_device_data(struct kfd_process *p) +{ + return list_first_entry(&p->per_device_data, + struct kfd_process_device, + per_device_list); +} + +struct kfd_process_device *kfd_get_next_process_device_data(struct kfd_process *p, + struct kfd_process_device *pdd) +{ + if (list_is_last(&pdd->per_device_list, &p->per_device_data)) + return NULL; + return list_next_entry(pdd, per_device_list); +} + +bool kfd_has_process_device_data(struct kfd_process *p) +{ + return !(list_empty(&p->per_device_data)); +} -- cgit v0.10.2 From b17f068a09fbe9b193a18080ff8ad5114a900124 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 17 Jul 2014 00:06:27 +0300 Subject: amdkfd: Add binding/unbinding calls to amd_iommu driver This patch adds the functions to bind and unbind pasid from a device through the amd_iommu driver. The unbind function is called when the mm_struct of the process is released. The bind function is not called here because it is called only in the IOCTLs which are not yet implemented at this stage of the patchset. Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 71a03f7..4b1cced 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -100,6 +100,65 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev) return kfd; } +static bool device_iommu_pasid_init(struct kfd_dev *kfd) +{ + const u32 required_iommu_flags = AMD_IOMMU_DEVICE_FLAG_ATS_SUP | + AMD_IOMMU_DEVICE_FLAG_PRI_SUP | + AMD_IOMMU_DEVICE_FLAG_PASID_SUP; + + struct amd_iommu_device_info iommu_info; + unsigned int pasid_limit; + int err; + + err = amd_iommu_device_info(kfd->pdev, &iommu_info); + if (err < 0) { + dev_err(kfd_device, + "error getting iommu info. is the iommu enabled?\n"); + return false; + } + + if ((iommu_info.flags & required_iommu_flags) != required_iommu_flags) { + dev_err(kfd_device, "error required iommu flags ats(%i), pri(%i), pasid(%i)\n", + (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_ATS_SUP) != 0, + (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PRI_SUP) != 0, + (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP) != 0); + return false; + } + + pasid_limit = min_t(unsigned int, + (unsigned int)1 << kfd->device_info->max_pasid_bits, + iommu_info.max_pasids); + /* + * last pasid is used for kernel queues doorbells + * in the future the last pasid might be used for a kernel thread. + */ + pasid_limit = min_t(unsigned int, + pasid_limit, + kfd->doorbell_process_limit - 1); + + err = amd_iommu_init_device(kfd->pdev, pasid_limit); + if (err < 0) { + dev_err(kfd_device, "error initializing iommu device\n"); + return false; + } + + if (!kfd_set_pasid_limit(pasid_limit)) { + dev_err(kfd_device, "error setting pasid limit\n"); + amd_iommu_free_device(kfd->pdev); + return false; + } + + return true; +} + +static void iommu_pasid_shutdown_callback(struct pci_dev *pdev, int pasid) +{ + struct kfd_dev *dev = kfd_device_by_pci_dev(pdev); + + if (dev) + kfd_unbind_process_from_device(dev, pasid); +} + bool kgd2kfd_device_init(struct kfd_dev *kfd, const struct kgd2kfd_shared_resources *gpu_resources) { @@ -131,6 +190,14 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, goto kfd_topology_add_device_error; } + if (!device_iommu_pasid_init(kfd)) { + dev_err(kfd_device, + "Error initializing iommuv2 for device (%x:%x)\n", + kfd->pdev->vendor, kfd->pdev->device); + goto device_iommu_pasid_error; + } + amd_iommu_set_invalidate_ctx_cb(kfd->pdev, + iommu_pasid_shutdown_callback); kfd->init_complete = true; dev_info(kfd_device, "added device (%x:%x)\n", kfd->pdev->vendor, @@ -138,6 +205,8 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, goto out; +device_iommu_pasid_error: + kfd_topology_remove_device(kfd); kfd_topology_add_device_error: kfd2kgd->fini_sa_manager(kfd->kgd); dev_err(kfd_device, @@ -149,7 +218,10 @@ out: void kgd2kfd_device_exit(struct kfd_dev *kfd) { - kfd_topology_remove_device(kfd); + if (kfd->init_complete) { + amd_iommu_free_device(kfd->pdev); + kfd_topology_remove_device(kfd); + } kfree(kfd); } @@ -157,12 +229,28 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd) void kgd2kfd_suspend(struct kfd_dev *kfd) { BUG_ON(kfd == NULL); + + if (kfd->init_complete) + amd_iommu_free_device(kfd->pdev); } int kgd2kfd_resume(struct kfd_dev *kfd) { + unsigned int pasid_limit; + int err; + BUG_ON(kfd == NULL); + pasid_limit = kfd_get_pasid_limit(); + + if (kfd->init_complete) { + err = amd_iommu_init_device(kfd->pdev, pasid_limit); + if (err < 0) + return -ENXIO; + amd_iommu_set_invalidate_ctx_cb(kfd->pdev, + iommu_pasid_shutdown_callback); + } + return 0; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 77d15db..431a09b 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -194,6 +194,7 @@ void kfd_process_destroy_wq(void); struct kfd_process *kfd_create_process(const struct task_struct *); struct kfd_process *kfd_get_process(const struct task_struct *); +void kfd_unbind_process_from_device(struct kfd_dev *dev, unsigned int pasid); struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, struct kfd_process *p, int create_pdd); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 5596f69..5084794 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -24,6 +24,7 @@ #include #include #include +#include #include struct mm_struct; @@ -163,6 +164,7 @@ static void kfd_process_wq_release(struct work_struct *work) list_for_each_entry_safe(pdd, temp, &p->per_device_data, per_device_list) { + amd_iommu_unbind_pasid(pdd->dev->pdev, p->pasid); list_del(&pdd->per_device_list); kfree(pdd); @@ -316,6 +318,7 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, struct kfd_process *p) { struct kfd_process_device *pdd = kfd_get_process_device_data(dev, p, 1); + int err; if (pdd == NULL) return ERR_PTR(-ENOMEM); @@ -323,6 +326,15 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, if (pdd->bound) return pdd; + err = amd_iommu_bind_pasid(dev->pdev, p->pasid, p->lead_thread); + if (err < 0) + return ERR_PTR(err); + + if (err < 0) { + amd_iommu_unbind_pasid(dev->pdev, p->pasid); + return ERR_PTR(err); + } + pdd->bound = true; return pdd; -- cgit v0.10.2 From ed8aab4594e0c5054b25b4a8810901ccf70efdcf Mon Sep 17 00:00:00 2001 From: Ben Goz Date: Thu, 17 Jul 2014 00:18:51 +0300 Subject: amdkfd: Add queue module The queue module enables allocating and initializing queues uniformly. v3: Removed typedef and redundant memset call. Broke long pr_debug print to one liners and Added documentation. v5: Move amdkfd from drm/radeon/ to drm/amd/ Signed-off-by: Ben Goz Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index e829a3f..42df022 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -6,6 +6,6 @@ ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/amd/include/ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ - kfd_process.o + kfd_process.o kfd_queue.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 431a09b..5f6f13d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -57,7 +57,6 @@ extern int max_num_of_queues_per_process; #define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS_DEFAULT 128 #define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS 1024 - struct kfd_device_info { unsigned int max_pasid_bits; size_t ih_ring_entry_size; @@ -120,6 +119,129 @@ void kfd_chardev_exit(void); struct device *kfd_chardev(void); +/** + * enum kfd_queue_type + * + * @KFD_QUEUE_TYPE_COMPUTE: Regular user mode queue type. + * + * @KFD_QUEUE_TYPE_SDMA: Sdma user mode queue type. + * + * @KFD_QUEUE_TYPE_HIQ: HIQ queue type. + * + * @KFD_QUEUE_TYPE_DIQ: DIQ queue type. + */ +enum kfd_queue_type { + KFD_QUEUE_TYPE_COMPUTE, + KFD_QUEUE_TYPE_SDMA, + KFD_QUEUE_TYPE_HIQ, + KFD_QUEUE_TYPE_DIQ +}; + +/** + * struct queue_properties + * + * @type: The queue type. + * + * @queue_id: Queue identifier. + * + * @queue_address: Queue ring buffer address. + * + * @queue_size: Queue ring buffer size. + * + * @priority: Defines the queue priority relative to other queues in the + * process. + * This is just an indication and HW scheduling may override the priority as + * necessary while keeping the relative prioritization. + * the priority granularity is from 0 to f which f is the highest priority. + * currently all queues are initialized with the highest priority. + * + * @queue_percent: This field is partially implemented and currently a zero in + * this field defines that the queue is non active. + * + * @read_ptr: User space address which points to the number of dwords the + * cp read from the ring buffer. This field updates automatically by the H/W. + * + * @write_ptr: Defines the number of dwords written to the ring buffer. + * + * @doorbell_ptr: This field aim is to notify the H/W of new packet written to + * the queue ring buffer. This field should be similar to write_ptr and the user + * should update this field after he updated the write_ptr. + * + * @doorbell_off: The doorbell offset in the doorbell pci-bar. + * + * @is_interop: Defines if this is a interop queue. Interop queue means that the + * queue can access both graphics and compute resources. + * + * @is_active: Defines if the queue is active or not. + * + * @vmid: If the scheduling mode is no cp scheduling the field defines the vmid + * of the queue. + * + * This structure represents the queue properties for each queue no matter if + * it's user mode or kernel mode queue. + * + */ +struct queue_properties { + enum kfd_queue_type type; + unsigned int queue_id; + uint64_t queue_address; + uint64_t queue_size; + uint32_t priority; + uint32_t queue_percent; + uint32_t *read_ptr; + uint32_t *write_ptr; + uint32_t *doorbell_ptr; + uint32_t doorbell_off; + bool is_interop; + bool is_active; + /* Not relevant for user mode queues in cp scheduling */ + unsigned int vmid; +}; + +/** + * struct queue + * + * @list: Queue linked list. + * + * @mqd: The queue MQD. + * + * @mqd_mem_obj: The MQD local gpu memory object. + * + * @gart_mqd_addr: The MQD gart mc address. + * + * @properties: The queue properties. + * + * @mec: Used only in no cp scheduling mode and identifies to micro engine id + * that the queue should be execute on. + * + * @pipe: Used only in no cp scheduling mode and identifies the queue's pipe id. + * + * @queue: Used only in no cp scheduliong mode and identifies the queue's slot. + * + * @process: The kfd process that created this queue. + * + * @device: The kfd device that created this queue. + * + * This structure represents user mode compute queues. + * It contains all the necessary data to handle such queues. + * + */ + +struct queue { + struct list_head list; + void *mqd; + struct kfd_mem_obj *mqd_mem_obj; + uint64_t gart_mqd_addr; + struct queue_properties properties; + + uint32_t mec; + uint32_t pipe; + uint32_t queue; + + struct kfd_process *process; + struct kfd_dev *device; +}; + /* Data that is per-process-per device. */ struct kfd_process_device { /* diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_queue.c new file mode 100644 index 0000000..9a0c90b --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_queue.c @@ -0,0 +1,85 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include "kfd_priv.h" + +void print_queue_properties(struct queue_properties *q) +{ + if (!q) + return; + + pr_debug("Printing queue properties:\n"); + pr_debug("Queue Type: %u\n", q->type); + pr_debug("Queue Size: %llu\n", q->queue_size); + pr_debug("Queue percent: %u\n", q->queue_percent); + pr_debug("Queue Address: 0x%llX\n", q->queue_address); + pr_debug("Queue Id: %u\n", q->queue_id); + pr_debug("Queue Process Vmid: %u\n", q->vmid); + pr_debug("Queue Read Pointer: 0x%p\n", q->read_ptr); + pr_debug("Queue Write Pointer: 0x%p\n", q->write_ptr); + pr_debug("Queue Doorbell Pointer: 0x%p\n", q->doorbell_ptr); + pr_debug("Queue Doorbell Offset: %u\n", q->doorbell_off); +} + +void print_queue(struct queue *q) +{ + if (!q) + return; + pr_debug("Printing queue:\n"); + pr_debug("Queue Type: %u\n", q->properties.type); + pr_debug("Queue Size: %llu\n", q->properties.queue_size); + pr_debug("Queue percent: %u\n", q->properties.queue_percent); + pr_debug("Queue Address: 0x%llX\n", q->properties.queue_address); + pr_debug("Queue Id: %u\n", q->properties.queue_id); + pr_debug("Queue Process Vmid: %u\n", q->properties.vmid); + pr_debug("Queue Read Pointer: 0x%p\n", q->properties.read_ptr); + pr_debug("Queue Write Pointer: 0x%p\n", q->properties.write_ptr); + pr_debug("Queue Doorbell Pointer: 0x%p\n", q->properties.doorbell_ptr); + pr_debug("Queue Doorbell Offset: %u\n", q->properties.doorbell_off); + pr_debug("Queue MQD Address: 0x%p\n", q->mqd); + pr_debug("Queue MQD Gart: 0x%llX\n", q->gart_mqd_addr); + pr_debug("Queue Process Address: 0x%p\n", q->process); + pr_debug("Queue Device Address: 0x%p\n", q->device); +} + +int init_queue(struct queue **q, struct queue_properties properties) +{ + struct queue *tmp; + + BUG_ON(!q); + + tmp = kzalloc(sizeof(struct queue), GFP_KERNEL); + if (!tmp) + return -ENOMEM; + + memcpy(&tmp->properties, &properties, sizeof(struct queue_properties)); + + *q = tmp; + return 0; +} + +void uninit_queue(struct queue *q) +{ + kfree(q); +} -- cgit v0.10.2 From 6e99df5741b3332dcceb86915b2818af640c51f9 Mon Sep 17 00:00:00 2001 From: Ben Goz Date: Thu, 17 Jul 2014 00:36:17 +0300 Subject: amdkfd: Add mqd_manager module The mqd_manager module handles MQD data structures. MQD stands for Memory Queue Descriptor, which is used by the H/W to keep the usermode queue state in memory. v3: Removed new typedefs Removed pragma pack 4 Remove cik_mqds.h file Changed lower_32/upper_32 calls to use linux macros Used new gart allocation functions Added documentation v4: Added missing initialization of the addr field in init_mqd() Setting the hqd persistent.preload_req bit ON so that when queues switches on/off, their context will kept and read from the mqd when the cp reassign them, and thus the dispatched workload context kept consistent without any interrupts. v5: Move amdkfd from drm/radeon/ to drm/amd/ Change format of mqd structure to match latest KV firmware Add support for AQL queues creation to enable working with open-source HSA runtime. Various fixes Signed-off-by: Ben Goz Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 42df022..301e847 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -6,6 +6,6 @@ ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/amd/include/ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ - kfd_process.o kfd_queue.o + kfd_process.o kfd_queue.o kfd_mqd_manager.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/cik_regs.h b/drivers/gpu/drm/amd/amdkfd/cik_regs.h new file mode 100644 index 0000000..607fc5c --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/cik_regs.h @@ -0,0 +1,221 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef CIK_REGS_H +#define CIK_REGS_H + +#define IH_VMID_0_LUT 0x3D40u + +#define BIF_DOORBELL_CNTL 0x530Cu + +#define SRBM_GFX_CNTL 0xE44 +#define PIPEID(x) ((x) << 0) +#define MEID(x) ((x) << 2) +#define VMID(x) ((x) << 4) +#define QUEUEID(x) ((x) << 8) + +#define SQ_CONFIG 0x8C00 + +#define SH_MEM_BASES 0x8C28 +/* if PTR32, these are the bases for scratch and lds */ +#define PRIVATE_BASE(x) ((x) << 0) /* scratch */ +#define SHARED_BASE(x) ((x) << 16) /* LDS */ +#define SH_MEM_APE1_BASE 0x8C2C +/* if PTR32, this is the base location of GPUVM */ +#define SH_MEM_APE1_LIMIT 0x8C30 +/* if PTR32, this is the upper limit of GPUVM */ +#define SH_MEM_CONFIG 0x8C34 +#define PTR32 (1 << 0) +#define PRIVATE_ATC (1 << 1) +#define ALIGNMENT_MODE(x) ((x) << 2) +#define SH_MEM_ALIGNMENT_MODE_DWORD 0 +#define SH_MEM_ALIGNMENT_MODE_DWORD_STRICT 1 +#define SH_MEM_ALIGNMENT_MODE_STRICT 2 +#define SH_MEM_ALIGNMENT_MODE_UNALIGNED 3 +#define DEFAULT_MTYPE(x) ((x) << 4) +#define APE1_MTYPE(x) ((x) << 7) + +/* valid for both DEFAULT_MTYPE and APE1_MTYPE */ +#define MTYPE_CACHED 0 +#define MTYPE_NONCACHED 3 + + +#define SH_STATIC_MEM_CONFIG 0x9604u + +#define TC_CFG_L1_LOAD_POLICY0 0xAC68 +#define TC_CFG_L1_LOAD_POLICY1 0xAC6C +#define TC_CFG_L1_STORE_POLICY 0xAC70 +#define TC_CFG_L2_LOAD_POLICY0 0xAC74 +#define TC_CFG_L2_LOAD_POLICY1 0xAC78 +#define TC_CFG_L2_STORE_POLICY0 0xAC7C +#define TC_CFG_L2_STORE_POLICY1 0xAC80 +#define TC_CFG_L2_ATOMIC_POLICY 0xAC84 +#define TC_CFG_L1_VOLATILE 0xAC88 +#define TC_CFG_L2_VOLATILE 0xAC8C + +#define CP_PQ_WPTR_POLL_CNTL 0xC20C +#define WPTR_POLL_EN (1 << 31) + +#define CPC_INT_CNTL 0xC2D0 +#define CP_ME1_PIPE0_INT_CNTL 0xC214 +#define CP_ME1_PIPE1_INT_CNTL 0xC218 +#define CP_ME1_PIPE2_INT_CNTL 0xC21C +#define CP_ME1_PIPE3_INT_CNTL 0xC220 +#define CP_ME2_PIPE0_INT_CNTL 0xC224 +#define CP_ME2_PIPE1_INT_CNTL 0xC228 +#define CP_ME2_PIPE2_INT_CNTL 0xC22C +#define CP_ME2_PIPE3_INT_CNTL 0xC230 +#define DEQUEUE_REQUEST_INT_ENABLE (1 << 13) +#define WRM_POLL_TIMEOUT_INT_ENABLE (1 << 17) +#define PRIV_REG_INT_ENABLE (1 << 23) +#define TIME_STAMP_INT_ENABLE (1 << 26) +#define GENERIC2_INT_ENABLE (1 << 29) +#define GENERIC1_INT_ENABLE (1 << 30) +#define GENERIC0_INT_ENABLE (1 << 31) +#define CP_ME1_PIPE0_INT_STATUS 0xC214 +#define CP_ME1_PIPE1_INT_STATUS 0xC218 +#define CP_ME1_PIPE2_INT_STATUS 0xC21C +#define CP_ME1_PIPE3_INT_STATUS 0xC220 +#define CP_ME2_PIPE0_INT_STATUS 0xC224 +#define CP_ME2_PIPE1_INT_STATUS 0xC228 +#define CP_ME2_PIPE2_INT_STATUS 0xC22C +#define CP_ME2_PIPE3_INT_STATUS 0xC230 +#define DEQUEUE_REQUEST_INT_STATUS (1 << 13) +#define WRM_POLL_TIMEOUT_INT_STATUS (1 << 17) +#define PRIV_REG_INT_STATUS (1 << 23) +#define TIME_STAMP_INT_STATUS (1 << 26) +#define GENERIC2_INT_STATUS (1 << 29) +#define GENERIC1_INT_STATUS (1 << 30) +#define GENERIC0_INT_STATUS (1 << 31) + +#define CP_HPD_EOP_BASE_ADDR 0xC904 +#define CP_HPD_EOP_BASE_ADDR_HI 0xC908 +#define CP_HPD_EOP_VMID 0xC90C +#define CP_HPD_EOP_CONTROL 0xC910 +#define EOP_SIZE(x) ((x) << 0) +#define EOP_SIZE_MASK (0x3f << 0) +#define CP_MQD_BASE_ADDR 0xC914 +#define CP_MQD_BASE_ADDR_HI 0xC918 +#define CP_HQD_ACTIVE 0xC91C +#define CP_HQD_VMID 0xC920 + +#define CP_HQD_PERSISTENT_STATE 0xC924u +#define DEFAULT_CP_HQD_PERSISTENT_STATE (0x33U << 8) +#define PRELOAD_REQ (1 << 0) + +#define CP_HQD_PIPE_PRIORITY 0xC928u +#define CP_HQD_QUEUE_PRIORITY 0xC92Cu +#define CP_HQD_QUANTUM 0xC930u +#define QUANTUM_EN 1U +#define QUANTUM_SCALE_1MS (1U << 4) +#define QUANTUM_DURATION(x) ((x) << 8) + +#define CP_HQD_PQ_BASE 0xC934 +#define CP_HQD_PQ_BASE_HI 0xC938 +#define CP_HQD_PQ_RPTR 0xC93C +#define CP_HQD_PQ_RPTR_REPORT_ADDR 0xC940 +#define CP_HQD_PQ_RPTR_REPORT_ADDR_HI 0xC944 +#define CP_HQD_PQ_WPTR_POLL_ADDR 0xC948 +#define CP_HQD_PQ_WPTR_POLL_ADDR_HI 0xC94C +#define CP_HQD_PQ_DOORBELL_CONTROL 0xC950 +#define DOORBELL_OFFSET(x) ((x) << 2) +#define DOORBELL_OFFSET_MASK (0x1fffff << 2) +#define DOORBELL_SOURCE (1 << 28) +#define DOORBELL_SCHD_HIT (1 << 29) +#define DOORBELL_EN (1 << 30) +#define DOORBELL_HIT (1 << 31) +#define CP_HQD_PQ_WPTR 0xC954 +#define CP_HQD_PQ_CONTROL 0xC958 +#define QUEUE_SIZE(x) ((x) << 0) +#define QUEUE_SIZE_MASK (0x3f << 0) +#define RPTR_BLOCK_SIZE(x) ((x) << 8) +#define RPTR_BLOCK_SIZE_MASK (0x3f << 8) +#define MIN_AVAIL_SIZE(x) ((x) << 20) +#define PQ_ATC_EN (1 << 23) +#define PQ_VOLATILE (1 << 26) +#define NO_UPDATE_RPTR (1 << 27) +#define UNORD_DISPATCH (1 << 28) +#define ROQ_PQ_IB_FLIP (1 << 29) +#define PRIV_STATE (1 << 30) +#define KMD_QUEUE (1 << 31) + +#define DEFAULT_RPTR_BLOCK_SIZE RPTR_BLOCK_SIZE(5) +#define DEFAULT_MIN_AVAIL_SIZE MIN_AVAIL_SIZE(3) + +#define CP_HQD_IB_BASE_ADDR 0xC95Cu +#define CP_HQD_IB_BASE_ADDR_HI 0xC960u +#define CP_HQD_IB_RPTR 0xC964u +#define CP_HQD_IB_CONTROL 0xC968u +#define IB_ATC_EN (1U << 23) +#define DEFAULT_MIN_IB_AVAIL_SIZE (3U << 20) + +#define CP_HQD_DEQUEUE_REQUEST 0xC974 +#define DEQUEUE_REQUEST_DRAIN 1 +#define DEQUEUE_REQUEST_RESET 2 +#define DEQUEUE_INT (1U << 8) + +#define CP_HQD_SEMA_CMD 0xC97Cu +#define CP_HQD_MSG_TYPE 0xC980u +#define CP_HQD_ATOMIC0_PREOP_LO 0xC984u +#define CP_HQD_ATOMIC0_PREOP_HI 0xC988u +#define CP_HQD_ATOMIC1_PREOP_LO 0xC98Cu +#define CP_HQD_ATOMIC1_PREOP_HI 0xC990u +#define CP_HQD_HQ_SCHEDULER0 0xC994u +#define CP_HQD_HQ_SCHEDULER1 0xC998u + + +#define CP_MQD_CONTROL 0xC99C +#define MQD_VMID(x) ((x) << 0) +#define MQD_VMID_MASK (0xf << 0) +#define MQD_CONTROL_PRIV_STATE_EN (1U << 8) + +#define GRBM_GFX_INDEX 0x30800 +#define INSTANCE_INDEX(x) ((x) << 0) +#define SH_INDEX(x) ((x) << 8) +#define SE_INDEX(x) ((x) << 16) +#define SH_BROADCAST_WRITES (1 << 29) +#define INSTANCE_BROADCAST_WRITES (1 << 30) +#define SE_BROADCAST_WRITES (1 << 31) + +#define SQC_CACHES 0x30d20 +#define SQC_POLICY 0x8C38u +#define SQC_VOLATILE 0x8C3Cu + +#define CP_PERFMON_CNTL 0x36020 + +#define ATC_VMID0_PASID_MAPPING 0x339Cu +#define ATC_VMID_PASID_MAPPING_UPDATE_STATUS 0x3398u +#define ATC_VMID_PASID_MAPPING_VALID (1U << 31) + +#define ATC_VM_APERTURE0_CNTL 0x3310u +#define ATS_ACCESS_MODE_NEVER 0 +#define ATS_ACCESS_MODE_ALWAYS 1 + +#define ATC_VM_APERTURE0_CNTL2 0x3318u +#define ATC_VM_APERTURE0_HIGH_ADDR 0x3308u +#define ATC_VM_APERTURE0_LOW_ADDR 0x3300u +#define ATC_VM_APERTURE1_CNTL 0x3314u +#define ATC_VM_APERTURE1_CNTL2 0x331Cu +#define ATC_VM_APERTURE1_HIGH_ADDR 0x330Cu +#define ATC_VM_APERTURE1_LOW_ADDR 0x3304u + +#endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c new file mode 100644 index 0000000..59d2407 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c @@ -0,0 +1,346 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include "kfd_priv.h" +#include "kfd_mqd_manager.h" +#include "cik_regs.h" +#include "../../radeon/cik_reg.h" + +inline void busy_wait(unsigned long ms) +{ + while (time_before(jiffies, ms)) + cpu_relax(); +} + +static inline struct cik_mqd *get_mqd(void *mqd) +{ + return (struct cik_mqd *)mqd; +} + +static int init_mqd(struct mqd_manager *mm, void **mqd, + struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, + struct queue_properties *q) +{ + uint64_t addr; + struct cik_mqd *m; + int retval; + + BUG_ON(!mm || !q || !mqd); + + pr_debug("kfd: In func %s\n", __func__); + + retval = kfd2kgd->allocate_mem(mm->dev->kgd, + sizeof(struct cik_mqd), + 256, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) mqd_mem_obj); + + if (retval != 0) + return -ENOMEM; + + m = (struct cik_mqd *) (*mqd_mem_obj)->cpu_ptr; + addr = (*mqd_mem_obj)->gpu_addr; + + memset(m, 0, ALIGN(sizeof(struct cik_mqd), 256)); + + m->header = 0xC0310800; + m->compute_pipelinestat_enable = 1; + m->compute_static_thread_mgmt_se0 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se1 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se2 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se3 = 0xFFFFFFFF; + + /* + * Make sure to use the last queue state saved on mqd when the cp + * reassigns the queue, so when queue is switched on/off (e.g over + * subscription or quantum timeout) the context will be consistent + */ + m->cp_hqd_persistent_state = + DEFAULT_CP_HQD_PERSISTENT_STATE | PRELOAD_REQ; + + m->cp_mqd_control = MQD_CONTROL_PRIV_STATE_EN; + m->cp_mqd_base_addr_lo = lower_32_bits(addr); + m->cp_mqd_base_addr_hi = upper_32_bits(addr); + + m->cp_hqd_ib_control = DEFAULT_MIN_IB_AVAIL_SIZE | IB_ATC_EN; + /* Although WinKFD writes this, I suspect it should not be necessary */ + m->cp_hqd_ib_control = IB_ATC_EN | DEFAULT_MIN_IB_AVAIL_SIZE; + + m->cp_hqd_quantum = QUANTUM_EN | QUANTUM_SCALE_1MS | + QUANTUM_DURATION(10); + + /* + * Pipe Priority + * Identifies the pipe relative priority when this queue is connected + * to the pipeline. The pipe priority is against the GFX pipe and HP3D. + * In KFD we are using a fixed pipe priority set to CS_MEDIUM. + * 0 = CS_LOW (typically below GFX) + * 1 = CS_MEDIUM (typically between HP3D and GFX + * 2 = CS_HIGH (typically above HP3D) + */ + m->cp_hqd_pipe_priority = 1; + m->cp_hqd_queue_priority = 15; + + *mqd = m; + if (gart_addr != NULL) + *gart_addr = addr; + retval = mm->update_mqd(mm, m, q); + + return retval; +} + +static void uninit_mqd(struct mqd_manager *mm, void *mqd, + struct kfd_mem_obj *mqd_mem_obj) +{ + BUG_ON(!mm || !mqd); + kfd2kgd->free_mem(mm->dev->kgd, (struct kgd_mem *) mqd_mem_obj); +} + +static int load_mqd(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, + uint32_t queue_id, uint32_t __user *wptr) +{ + return kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id, wptr); + +} + +static int update_mqd(struct mqd_manager *mm, void *mqd, + struct queue_properties *q) +{ + struct cik_mqd *m; + + BUG_ON(!mm || !q || !mqd); + + pr_debug("kfd: In func %s\n", __func__); + + m = get_mqd(mqd); + m->cp_hqd_pq_control = DEFAULT_RPTR_BLOCK_SIZE | + DEFAULT_MIN_AVAIL_SIZE | PQ_ATC_EN; + + /* + * Calculating queue size which is log base 2 of actual queue size -1 + * dwords and another -1 for ffs + */ + m->cp_hqd_pq_control |= ffs(q->queue_size / sizeof(unsigned int)) + - 1 - 1; + m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); + m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8); + m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr); + m->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits((uint64_t)q->read_ptr); + m->cp_hqd_pq_doorbell_control = DOORBELL_EN | + DOORBELL_OFFSET(q->doorbell_off); + + m->cp_hqd_vmid = q->vmid; + + if (q->format == KFD_QUEUE_FORMAT_AQL) { + m->cp_hqd_iq_rptr = AQL_ENABLE; + m->cp_hqd_pq_control |= NO_UPDATE_RPTR; + } + + m->cp_hqd_active = 0; + q->is_active = false; + if (q->queue_size > 0 && + q->queue_address != 0 && + q->queue_percent > 0) { + m->cp_hqd_active = 1; + q->is_active = true; + } + + return 0; +} + +static int destroy_mqd(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id) +{ + return kfd2kgd->hqd_destroy(mm->dev->kgd, type, timeout, + pipe_id, queue_id); +} + +bool is_occupied(struct mqd_manager *mm, void *mqd, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) +{ + + return kfd2kgd->hqd_is_occupies(mm->dev->kgd, queue_address, + pipe_id, queue_id); + +} + +/* + * HIQ MQD Implementation, concrete implementation for HIQ MQD implementation. + * The HIQ queue in Kaveri is using the same MQD structure as all the user mode + * queues but with different initial values. + */ + +static int init_mqd_hiq(struct mqd_manager *mm, void **mqd, + struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, + struct queue_properties *q) +{ + uint64_t addr; + struct cik_mqd *m; + int retval; + + BUG_ON(!mm || !q || !mqd || !mqd_mem_obj); + + pr_debug("kfd: In func %s\n", __func__); + + retval = kfd2kgd->allocate_mem(mm->dev->kgd, + sizeof(struct cik_mqd), + 256, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) mqd_mem_obj); + + if (retval != 0) + return -ENOMEM; + + m = (struct cik_mqd *) (*mqd_mem_obj)->cpu_ptr; + addr = (*mqd_mem_obj)->gpu_addr; + + memset(m, 0, ALIGN(sizeof(struct cik_mqd), 256)); + + m->header = 0xC0310800; + m->compute_pipelinestat_enable = 1; + m->compute_static_thread_mgmt_se0 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se1 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se2 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se3 = 0xFFFFFFFF; + + m->cp_hqd_persistent_state = DEFAULT_CP_HQD_PERSISTENT_STATE | + PRELOAD_REQ; + m->cp_hqd_quantum = QUANTUM_EN | QUANTUM_SCALE_1MS | + QUANTUM_DURATION(10); + + m->cp_mqd_control = MQD_CONTROL_PRIV_STATE_EN; + m->cp_mqd_base_addr_lo = lower_32_bits(addr); + m->cp_mqd_base_addr_hi = upper_32_bits(addr); + + m->cp_hqd_ib_control = DEFAULT_MIN_IB_AVAIL_SIZE; + + /* + * Pipe Priority + * Identifies the pipe relative priority when this queue is connected + * to the pipeline. The pipe priority is against the GFX pipe and HP3D. + * In KFD we are using a fixed pipe priority set to CS_MEDIUM. + * 0 = CS_LOW (typically below GFX) + * 1 = CS_MEDIUM (typically between HP3D and GFX + * 2 = CS_HIGH (typically above HP3D) + */ + m->cp_hqd_pipe_priority = 1; + m->cp_hqd_queue_priority = 15; + + *mqd = m; + if (gart_addr) + *gart_addr = addr; + retval = mm->update_mqd(mm, m, q); + + return retval; +} + +static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, + struct queue_properties *q) +{ + struct cik_mqd *m; + + BUG_ON(!mm || !q || !mqd); + + pr_debug("kfd: In func %s\n", __func__); + + m = get_mqd(mqd); + m->cp_hqd_pq_control = DEFAULT_RPTR_BLOCK_SIZE | + DEFAULT_MIN_AVAIL_SIZE | + PRIV_STATE | + KMD_QUEUE; + + /* + * Calculating queue size which is log base 2 of actual queue + * size -1 dwords + */ + m->cp_hqd_pq_control |= ffs(q->queue_size / sizeof(unsigned int)) + - 1 - 1; + m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); + m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8); + m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr); + m->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits((uint64_t)q->read_ptr); + m->cp_hqd_pq_doorbell_control = DOORBELL_EN | + DOORBELL_OFFSET(q->doorbell_off); + + m->cp_hqd_vmid = q->vmid; + + m->cp_hqd_active = 0; + q->is_active = false; + if (q->queue_size > 0 && + q->queue_address != 0 && + q->queue_percent > 0) { + m->cp_hqd_active = 1; + q->is_active = true; + } + + return 0; +} + +struct mqd_manager *mqd_manager_init(enum KFD_MQD_TYPE type, + struct kfd_dev *dev) +{ + struct mqd_manager *mqd; + + BUG_ON(!dev); + BUG_ON(type >= KFD_MQD_TYPE_MAX); + + pr_debug("kfd: In func %s\n", __func__); + + mqd = kzalloc(sizeof(struct mqd_manager), GFP_KERNEL); + if (!mqd) + return NULL; + + mqd->dev = dev; + + switch (type) { + case KFD_MQD_TYPE_CIK_CP: + case KFD_MQD_TYPE_CIK_COMPUTE: + mqd->init_mqd = init_mqd; + mqd->uninit_mqd = uninit_mqd; + mqd->load_mqd = load_mqd; + mqd->update_mqd = update_mqd; + mqd->destroy_mqd = destroy_mqd; + mqd->is_occupied = is_occupied; + break; + case KFD_MQD_TYPE_CIK_HIQ: + mqd->init_mqd = init_mqd_hiq; + mqd->uninit_mqd = uninit_mqd; + mqd->load_mqd = load_mqd; + mqd->update_mqd = update_mqd_hiq; + mqd->destroy_mqd = destroy_mqd; + mqd->is_occupied = is_occupied; + break; + default: + kfree(mqd); + return NULL; + } + + return mqd; +} + +/* SDMA queues should be implemented here when the cp will supports them */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h new file mode 100644 index 0000000..213a71e --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h @@ -0,0 +1,91 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef KFD_MQD_MANAGER_H_ +#define KFD_MQD_MANAGER_H_ + +#include "kfd_priv.h" + +/** + * struct mqd_manager + * + * @init_mqd: Allocates the mqd buffer on local gpu memory and initialize it. + * + * @load_mqd: Loads the mqd to a concrete hqd slot. Used only for no cp + * scheduling mode. + * + * @update_mqd: Handles a update call for the MQD + * + * @destroy_mqd: Destroys the HQD slot and by that preempt the relevant queue. + * Used only for no cp scheduling. + * + * @uninit_mqd: Releases the mqd buffer from local gpu memory. + * + * @is_occupied: Checks if the relevant HQD slot is occupied. + * + * @mqd_mutex: Mqd manager mutex. + * + * @dev: The kfd device structure coupled with this module. + * + * MQD stands for Memory Queue Descriptor which represents the current queue + * state in the memory and initiate the HQD (Hardware Queue Descriptor) state. + * This structure is actually a base class for the different types of MQDs + * structures for the variant ASICs that should be supported in the future. + * This base class is also contains all the MQD specific operations. + * Another important thing to mention is that each queue has a MQD that keeps + * his state (or context) after each preemption or reassignment. + * Basically there are a instances of the mqd manager class per MQD type per + * ASIC. Currently the kfd driver supports only Kaveri so there are instances + * per KFD_MQD_TYPE for each device. + * + */ + +struct mqd_manager { + int (*init_mqd)(struct mqd_manager *mm, void **mqd, + struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, + struct queue_properties *q); + + int (*load_mqd)(struct mqd_manager *mm, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t __user *wptr); + + int (*update_mqd)(struct mqd_manager *mm, void *mqd, + struct queue_properties *q); + + int (*destroy_mqd)(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id); + + void (*uninit_mqd)(struct mqd_manager *mm, void *mqd, + struct kfd_mem_obj *mqd_mem_obj); + + bool (*is_occupied)(struct mqd_manager *mm, void *mqd, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id); + + struct mutex mqd_mutex; + struct kfd_dev *dev; +}; + +#endif /* KFD_MQD_MANAGER_H_ */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 5f6f13d..0eb2032 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -119,6 +119,11 @@ void kfd_chardev_exit(void); struct device *kfd_chardev(void); +enum kfd_preempt_type { + KFD_PREEMPT_TYPE_WAVEFRONT, + KFD_PREEMPT_TYPE_WAVEFRONT_RESET +}; + /** * enum kfd_queue_type * @@ -137,6 +142,11 @@ enum kfd_queue_type { KFD_QUEUE_TYPE_DIQ }; +enum kfd_queue_format { + KFD_QUEUE_FORMAT_PM4, + KFD_QUEUE_FORMAT_AQL +}; + /** * struct queue_properties * @@ -183,6 +193,7 @@ enum kfd_queue_type { */ struct queue_properties { enum kfd_queue_type type; + enum kfd_queue_format format; unsigned int queue_id; uint64_t queue_address; uint64_t queue_size; @@ -242,6 +253,17 @@ struct queue { struct kfd_dev *device; }; +/* + * Please read the kfd_mqd_manager.h description. + */ +enum KFD_MQD_TYPE { + KFD_MQD_TYPE_CIK_COMPUTE = 0, /* for no cp scheduling */ + KFD_MQD_TYPE_CIK_HIQ, /* for hiq */ + KFD_MQD_TYPE_CIK_CP, /* for cp queues and diq */ + KFD_MQD_TYPE_CIK_SDMA, /* for sdma queues */ + KFD_MQD_TYPE_MAX +}; + /* Data that is per-process-per device. */ struct kfd_process_device { /* -- cgit v0.10.2 From ed6e6a3487bd736cbcfc74fe0f9d9220bae00c72 Mon Sep 17 00:00:00 2001 From: Ben Goz Date: Thu, 17 Jul 2014 00:45:35 +0300 Subject: amdkfd: Add kernel queue module The kernel queue module enables the amdkfd to establish kernel queues, not exposed to user space. The kernel queues are used for HIQ (HSA Interface Queue) and DIQ (Debug Interface Queue) operations v3: Removed use of internal typedefs and added use of the new gart allocation functions v4: Fixed a miscalculation in kernel queue wrapping v5: Move amdkfd from drm/radeon/ to drm/amd/ Change format of mqd structure to match latest KV firmware Add support for AQL queues creation to enable working with open-source HSA runtime Add define for kernel queue size Various fixes Signed-off-by: Ben Goz Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 301e847..047b834 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -6,6 +6,7 @@ ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/amd/include/ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ - kfd_process.o kfd_queue.o kfd_mqd_manager.o + kfd_process.o kfd_queue.o kfd_mqd_manager.o \ + kfd_kernel_queue.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h new file mode 100644 index 0000000..e495b38 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h @@ -0,0 +1,101 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef KFD_DEVICE_QUEUE_MANAGER_H_ +#define KFD_DEVICE_QUEUE_MANAGER_H_ + +#include +#include +#include "kfd_priv.h" +#include "kfd_mqd_manager.h" + +#define QUEUE_PREEMPT_DEFAULT_TIMEOUT_MS (500) +#define QUEUES_PER_PIPE (8) +#define PIPE_PER_ME_CP_SCHEDULING (3) +#define CIK_VMID_NUM (8) +#define KFD_VMID_START_OFFSET (8) +#define VMID_PER_DEVICE CIK_VMID_NUM +#define KFD_DQM_FIRST_PIPE (0) + +struct device_process_node { + struct qcm_process_device *qpd; + struct list_head list; +}; + +struct device_queue_manager { + int (*create_queue)(struct device_queue_manager *dqm, + struct queue *q, + struct qcm_process_device *qpd, + int *allocate_vmid); + int (*destroy_queue)(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + struct queue *q); + int (*update_queue)(struct device_queue_manager *dqm, + struct queue *q); + struct mqd_manager * (*get_mqd_manager)( + struct device_queue_manager *dqm, + enum KFD_MQD_TYPE type); + + int (*register_process)(struct device_queue_manager *dqm, + struct qcm_process_device *qpd); + int (*unregister_process)(struct device_queue_manager *dqm, + struct qcm_process_device *qpd); + int (*initialize)(struct device_queue_manager *dqm); + int (*start)(struct device_queue_manager *dqm); + int (*stop)(struct device_queue_manager *dqm); + void (*uninitialize)(struct device_queue_manager *dqm); + int (*create_kernel_queue)(struct device_queue_manager *dqm, + struct kernel_queue *kq, + struct qcm_process_device *qpd); + void (*destroy_kernel_queue)(struct device_queue_manager *dqm, + struct kernel_queue *kq, + struct qcm_process_device *qpd); + bool (*set_cache_memory_policy)(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + enum cache_policy default_policy, + enum cache_policy alternate_policy, + void __user *alternate_aperture_base, + uint64_t alternate_aperture_size); + + + struct mqd_manager *mqds[KFD_MQD_TYPE_MAX]; + struct packet_manager packets; + struct kfd_dev *dev; + struct mutex lock; + struct list_head queues; + unsigned int processes_count; + unsigned int queue_count; + unsigned int next_pipe_to_allocate; + unsigned int *allocated_queues; + unsigned int vmid_bitmap; + uint64_t pipelines_addr; + struct kfd_mem_obj *pipeline_mem; + uint64_t fence_gpu_addr; + unsigned int *fence_addr; + struct kfd_mem_obj *fence_mem; + bool active_runlist; +}; + + + +#endif /* KFD_DEVICE_QUEUE_MANAGER_H_ */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c new file mode 100644 index 0000000..555af45 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -0,0 +1,347 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include +#include +#include "kfd_kernel_queue.h" +#include "kfd_priv.h" +#include "kfd_device_queue_manager.h" +#include "kfd_pm4_headers.h" +#include "kfd_pm4_opcodes.h" + +#define PM4_COUNT_ZERO (((1 << 15) - 1) << 16) + +static bool initialize(struct kernel_queue *kq, struct kfd_dev *dev, + enum kfd_queue_type type, unsigned int queue_size) +{ + struct queue_properties prop; + int retval; + union PM4_MES_TYPE_3_HEADER nop; + + BUG_ON(!kq || !dev); + BUG_ON(type != KFD_QUEUE_TYPE_DIQ && type != KFD_QUEUE_TYPE_HIQ); + + pr_debug("kfd: In func %s initializing queue type %d size %d\n", + __func__, KFD_QUEUE_TYPE_HIQ, queue_size); + + nop.opcode = IT_NOP; + nop.type = PM4_TYPE_3; + nop.u32all |= PM4_COUNT_ZERO; + + kq->dev = dev; + kq->nop_packet = nop.u32all; + switch (type) { + case KFD_QUEUE_TYPE_DIQ: + case KFD_QUEUE_TYPE_HIQ: + kq->mqd = dev->dqm->get_mqd_manager(dev->dqm, + KFD_MQD_TYPE_CIK_HIQ); + break; + default: + BUG(); + break; + } + + if (kq->mqd == NULL) + return false; + + prop.doorbell_ptr = + (uint32_t *)kfd_get_kernel_doorbell(dev, &prop.doorbell_off); + + if (prop.doorbell_ptr == NULL) + goto err_get_kernel_doorbell; + + retval = kfd2kgd->allocate_mem(dev->kgd, + queue_size, + PAGE_SIZE, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) &kq->pq); + + if (retval != 0) + goto err_pq_allocate_vidmem; + + kq->pq_kernel_addr = kq->pq->cpu_ptr; + kq->pq_gpu_addr = kq->pq->gpu_addr; + + retval = kfd2kgd->allocate_mem(dev->kgd, + sizeof(*kq->rptr_kernel), + 32, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) &kq->rptr_mem); + + if (retval != 0) + goto err_rptr_allocate_vidmem; + + kq->rptr_kernel = kq->rptr_mem->cpu_ptr; + kq->rptr_gpu_addr = kq->rptr_mem->gpu_addr; + + retval = kfd2kgd->allocate_mem(dev->kgd, + sizeof(*kq->wptr_kernel), + 32, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) &kq->wptr_mem); + + if (retval != 0) + goto err_wptr_allocate_vidmem; + + kq->wptr_kernel = kq->wptr_mem->cpu_ptr; + kq->wptr_gpu_addr = kq->wptr_mem->gpu_addr; + + memset(kq->pq_kernel_addr, 0, queue_size); + memset(kq->rptr_kernel, 0, sizeof(*kq->rptr_kernel)); + memset(kq->wptr_kernel, 0, sizeof(*kq->wptr_kernel)); + + prop.queue_size = queue_size; + prop.is_interop = false; + prop.priority = 1; + prop.queue_percent = 100; + prop.type = type; + prop.vmid = 0; + prop.queue_address = kq->pq_gpu_addr; + prop.read_ptr = (uint32_t *) kq->rptr_gpu_addr; + prop.write_ptr = (uint32_t *) kq->wptr_gpu_addr; + + if (init_queue(&kq->queue, prop) != 0) + goto err_init_queue; + + kq->queue->device = dev; + kq->queue->process = kfd_get_process(current); + + retval = kq->mqd->init_mqd(kq->mqd, &kq->queue->mqd, + &kq->queue->mqd_mem_obj, + &kq->queue->gart_mqd_addr, + &kq->queue->properties); + if (retval != 0) + goto err_init_mqd; + + /* assign HIQ to HQD */ + if (type == KFD_QUEUE_TYPE_HIQ) { + pr_debug("assigning hiq to hqd\n"); + kq->queue->pipe = KFD_CIK_HIQ_PIPE; + kq->queue->queue = KFD_CIK_HIQ_QUEUE; + kq->mqd->load_mqd(kq->mqd, kq->queue->mqd, kq->queue->pipe, + kq->queue->queue, NULL); + } else { + /* allocate fence for DIQ */ + + retval = kfd2kgd->allocate_mem(dev->kgd, + sizeof(uint32_t), + 32, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) &kq->fence_mem_obj); + + if (retval != 0) + goto err_alloc_fence; + + kq->fence_kernel_address = kq->fence_mem_obj->cpu_ptr; + kq->fence_gpu_addr = kq->fence_mem_obj->gpu_addr; + } + + print_queue(kq->queue); + + return true; +err_alloc_fence: +err_init_mqd: + uninit_queue(kq->queue); +err_init_queue: + kfd2kgd->free_mem(dev->kgd, (struct kgd_mem *) kq->wptr_mem); +err_wptr_allocate_vidmem: + kfd2kgd->free_mem(dev->kgd, (struct kgd_mem *) kq->rptr_mem); +err_rptr_allocate_vidmem: + kfd2kgd->free_mem(dev->kgd, (struct kgd_mem *) kq->pq); +err_pq_allocate_vidmem: + pr_err("kfd: error init pq\n"); + kfd_release_kernel_doorbell(dev, (u32 *)prop.doorbell_ptr); +err_get_kernel_doorbell: + pr_err("kfd: error init doorbell"); + return false; + +} + +static void uninitialize(struct kernel_queue *kq) +{ + BUG_ON(!kq); + + if (kq->queue->properties.type == KFD_QUEUE_TYPE_HIQ) + kq->mqd->destroy_mqd(kq->mqd, + NULL, + false, + QUEUE_PREEMPT_DEFAULT_TIMEOUT_MS, + kq->queue->pipe, + kq->queue->queue); + + kfd2kgd->free_mem(kq->dev->kgd, (struct kgd_mem *) kq->rptr_mem); + kfd2kgd->free_mem(kq->dev->kgd, (struct kgd_mem *) kq->wptr_mem); + kfd2kgd->free_mem(kq->dev->kgd, (struct kgd_mem *) kq->pq); + kfd_release_kernel_doorbell(kq->dev, + (u32 *)kq->queue->properties.doorbell_ptr); + uninit_queue(kq->queue); +} + +static int acquire_packet_buffer(struct kernel_queue *kq, + size_t packet_size_in_dwords, unsigned int **buffer_ptr) +{ + size_t available_size; + size_t queue_size_dwords; + uint32_t wptr, rptr; + unsigned int *queue_address; + + BUG_ON(!kq || !buffer_ptr); + + rptr = *kq->rptr_kernel; + wptr = *kq->wptr_kernel; + queue_address = (unsigned int *)kq->pq_kernel_addr; + queue_size_dwords = kq->queue->properties.queue_size / sizeof(uint32_t); + + pr_debug("kfd: In func %s\nrptr: %d\nwptr: %d\nqueue_address 0x%p\n", + __func__, rptr, wptr, queue_address); + + available_size = (rptr - 1 - wptr + queue_size_dwords) % + queue_size_dwords; + + if (packet_size_in_dwords >= queue_size_dwords || + packet_size_in_dwords >= available_size) + return -ENOMEM; + + if (wptr + packet_size_in_dwords >= queue_size_dwords) { + while (wptr > 0) { + queue_address[wptr] = kq->nop_packet; + wptr = (wptr + 1) % queue_size_dwords; + } + } + + *buffer_ptr = &queue_address[wptr]; + kq->pending_wptr = wptr + packet_size_in_dwords; + + return 0; +} + +static void submit_packet(struct kernel_queue *kq) +{ +#ifdef DEBUG + int i; +#endif + + BUG_ON(!kq); + +#ifdef DEBUG + for (i = *kq->wptr_kernel; i < kq->pending_wptr; i++) { + pr_debug("0x%2X ", kq->pq_kernel_addr[i]); + if (i % 15 == 0) + pr_debug("\n"); + } + pr_debug("\n"); +#endif + + *kq->wptr_kernel = kq->pending_wptr; + write_kernel_doorbell((u32 *)kq->queue->properties.doorbell_ptr, + kq->pending_wptr); +} + +static int sync_with_hw(struct kernel_queue *kq, unsigned long timeout_ms) +{ + unsigned long org_timeout_ms; + + BUG_ON(!kq); + + org_timeout_ms = timeout_ms; + timeout_ms += jiffies * 1000 / HZ; + while (*kq->wptr_kernel != *kq->rptr_kernel) { + if (time_after(jiffies * 1000 / HZ, timeout_ms)) { + pr_err("kfd: kernel_queue %s timeout expired %lu\n", + __func__, org_timeout_ms); + pr_err("kfd: wptr: %d rptr: %d\n", + *kq->wptr_kernel, *kq->rptr_kernel); + return -ETIME; + } + cpu_relax(); + } + + return 0; +} + +static void rollback_packet(struct kernel_queue *kq) +{ + BUG_ON(!kq); + kq->pending_wptr = *kq->queue->properties.write_ptr; +} + +struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, + enum kfd_queue_type type) +{ + struct kernel_queue *kq; + + BUG_ON(!dev); + + kq = kzalloc(sizeof(struct kernel_queue), GFP_KERNEL); + if (!kq) + return NULL; + + kq->initialize = initialize; + kq->uninitialize = uninitialize; + kq->acquire_packet_buffer = acquire_packet_buffer; + kq->submit_packet = submit_packet; + kq->sync_with_hw = sync_with_hw; + kq->rollback_packet = rollback_packet; + + if (kq->initialize(kq, dev, type, KFD_KERNEL_QUEUE_SIZE) == false) { + pr_err("kfd: failed to init kernel queue\n"); + kfree(kq); + return NULL; + } + return kq; +} + +void kernel_queue_uninit(struct kernel_queue *kq) +{ + BUG_ON(!kq); + + kq->uninitialize(kq); + kfree(kq); +} + +void test_kq(struct kfd_dev *dev) +{ + struct kernel_queue *kq; + uint32_t *buffer, i; + int retval; + + BUG_ON(!dev); + + pr_debug("kfd: starting kernel queue test\n"); + + kq = kernel_queue_init(dev, KFD_QUEUE_TYPE_HIQ); + BUG_ON(!kq); + + retval = kq->acquire_packet_buffer(kq, 5, &buffer); + BUG_ON(retval != 0); + for (i = 0; i < 5; i++) + buffer[i] = kq->nop_packet; + kq->submit_packet(kq); + kq->sync_with_hw(kq, 1000); + + pr_debug("kfd: ending kernel queue test\n"); +} + + diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h new file mode 100644 index 0000000..dcd2bdb --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h @@ -0,0 +1,69 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef KFD_KERNEL_QUEUE_H_ +#define KFD_KERNEL_QUEUE_H_ + +#include +#include +#include "kfd_priv.h" + +struct kernel_queue { + /* interface */ + bool (*initialize)(struct kernel_queue *kq, struct kfd_dev *dev, + enum kfd_queue_type type, unsigned int queue_size); + void (*uninitialize)(struct kernel_queue *kq); + int (*acquire_packet_buffer)(struct kernel_queue *kq, + size_t packet_size_in_dwords, + unsigned int **buffer_ptr); + + void (*submit_packet)(struct kernel_queue *kq); + int (*sync_with_hw)(struct kernel_queue *kq, + unsigned long timeout_ms); + void (*rollback_packet)(struct kernel_queue *kq); + + /* data */ + struct kfd_dev *dev; + struct mqd_manager *mqd; + struct queue *queue; + uint32_t pending_wptr; + unsigned int nop_packet; + + struct kfd_mem_obj *rptr_mem; + uint32_t *rptr_kernel; + uint64_t rptr_gpu_addr; + struct kfd_mem_obj *wptr_mem; + uint32_t *wptr_kernel; + uint64_t wptr_gpu_addr; + struct kfd_mem_obj *pq; + uint64_t pq_gpu_addr; + uint32_t *pq_kernel_addr; + + struct kfd_mem_obj *fence_mem_obj; + uint64_t fence_gpu_addr; + void *fence_kernel_address; + + struct list_head list; +}; + +#endif /* KFD_KERNEL_QUEUE_H_ */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers.h new file mode 100644 index 0000000..071ad57 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers.h @@ -0,0 +1,405 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef KFD_PM4_HEADERS_H_ +#define KFD_PM4_HEADERS_H_ + +#ifndef PM4_MES_HEADER_DEFINED +#define PM4_MES_HEADER_DEFINED +union PM4_MES_TYPE_3_HEADER { + struct { + uint32_t reserved1:8; /* < reserved */ + uint32_t opcode:8; /* < IT opcode */ + uint32_t count:14; /* < number of DWORDs - 1 + * in the information body. + */ + uint32_t type:2; /* < packet identifier. + * It should be 3 for type 3 packets + */ + }; + uint32_t u32all; +}; +#endif /* PM4_MES_HEADER_DEFINED */ + +/* --------------------MES_SET_RESOURCES-------------------- */ + +#ifndef PM4_MES_SET_RESOURCES_DEFINED +#define PM4_MES_SET_RESOURCES_DEFINED +enum set_resources_queue_type_enum { + queue_type__mes_set_resources__kernel_interface_queue_kiq = 0, + queue_type__mes_set_resources__hsa_interface_queue_hiq = 1, + queue_type__mes_set_resources__hsa_debug_interface_queue = 4 +}; + +struct pm4_set_resources { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t vmid_mask:16; + uint32_t unmap_latency:8; + uint32_t reserved1:5; + enum set_resources_queue_type_enum queue_type:3; + } bitfields2; + uint32_t ordinal2; + }; + + uint32_t queue_mask_lo; + uint32_t queue_mask_hi; + uint32_t gws_mask_lo; + uint32_t gws_mask_hi; + + union { + struct { + uint32_t oac_mask:16; + uint32_t reserved2:16; + } bitfields7; + uint32_t ordinal7; + }; + + union { + struct { + uint32_t gds_heap_base:6; + uint32_t reserved3:5; + uint32_t gds_heap_size:6; + uint32_t reserved4:15; + } bitfields8; + uint32_t ordinal8; + }; + +}; +#endif + +/*--------------------MES_RUN_LIST-------------------- */ + +#ifndef PM4_MES_RUN_LIST_DEFINED +#define PM4_MES_RUN_LIST_DEFINED + +struct pm4_runlist { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t reserved1:2; + uint32_t ib_base_lo:30; + } bitfields2; + uint32_t ordinal2; + }; + + union { + struct { + uint32_t ib_base_hi:16; + uint32_t reserved2:16; + } bitfields3; + uint32_t ordinal3; + }; + + union { + struct { + uint32_t ib_size:20; + uint32_t chain:1; + uint32_t offload_polling:1; + uint32_t reserved3:1; + uint32_t valid:1; + uint32_t reserved4:8; + } bitfields4; + uint32_t ordinal4; + }; + +}; +#endif + +/*--------------------MES_MAP_PROCESS-------------------- */ + +#ifndef PM4_MES_MAP_PROCESS_DEFINED +#define PM4_MES_MAP_PROCESS_DEFINED + +struct pm4_map_process { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t pasid:16; + uint32_t reserved1:8; + uint32_t diq_enable:1; + uint32_t process_quantum:7; + } bitfields2; + uint32_t ordinal2; + }; + + union { + struct { + uint32_t page_table_base:28; + uint32_t reserved3:4; + } bitfields3; + uint32_t ordinal3; + }; + + uint32_t sh_mem_bases; + uint32_t sh_mem_ape1_base; + uint32_t sh_mem_ape1_limit; + uint32_t sh_mem_config; + uint32_t gds_addr_lo; + uint32_t gds_addr_hi; + + union { + struct { + uint32_t num_gws:6; + uint32_t reserved4:2; + uint32_t num_oac:4; + uint32_t reserved5:4; + uint32_t gds_size:6; + uint32_t num_queues:10; + } bitfields10; + uint32_t ordinal10; + }; + +}; +#endif + +/*--------------------MES_MAP_QUEUES--------------------*/ + +#ifndef PM4_MES_MAP_QUEUES_DEFINED +#define PM4_MES_MAP_QUEUES_DEFINED +enum map_queues_queue_sel_enum { + queue_sel__mes_map_queues__map_to_specified_queue_slots = 0, + queue_sel__mes_map_queues__map_to_hws_determined_queue_slots = 1, + queue_sel__mes_map_queues__enable_process_queues = 2 +}; + +enum map_queues_vidmem_enum { + vidmem__mes_map_queues__uses_no_video_memory = 0, + vidmem__mes_map_queues__uses_video_memory = 1 +}; + +enum map_queues_alloc_format_enum { + alloc_format__mes_map_queues__one_per_pipe = 0, + alloc_format__mes_map_queues__all_on_one_pipe = 1 +}; + +enum map_queues_engine_sel_enum { + engine_sel__mes_map_queues__compute = 0, + engine_sel__mes_map_queues__sdma0 = 2, + engine_sel__mes_map_queues__sdma1 = 3 +}; + +struct pm4_map_queues { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t reserved1:4; + enum map_queues_queue_sel_enum queue_sel:2; + uint32_t reserved2:2; + uint32_t vmid:4; + uint32_t reserved3:4; + enum map_queues_vidmem_enum vidmem:2; + uint32_t reserved4:6; + enum map_queues_alloc_format_enum alloc_format:2; + enum map_queues_engine_sel_enum engine_sel:3; + uint32_t num_queues:3; + } bitfields2; + uint32_t ordinal2; + }; + + struct { + union { + struct { + uint32_t reserved5:2; + uint32_t doorbell_offset:21; + uint32_t reserved6:3; + uint32_t queue:6; + } bitfields3; + uint32_t ordinal3; + }; + + uint32_t mqd_addr_lo; + uint32_t mqd_addr_hi; + uint32_t wptr_addr_lo; + uint32_t wptr_addr_hi; + + } mes_map_queues_ordinals[1]; /* 1..N of these ordinal groups */ + +}; +#endif + +/*--------------------MES_QUERY_STATUS--------------------*/ + +#ifndef PM4_MES_QUERY_STATUS_DEFINED +#define PM4_MES_QUERY_STATUS_DEFINED +enum query_status_interrupt_sel_enum { + interrupt_sel__mes_query_status__completion_status = 0, + interrupt_sel__mes_query_status__process_status = 1, + interrupt_sel__mes_query_status__queue_status = 2 +}; + +enum query_status_command_enum { + command__mes_query_status__interrupt_only = 0, + command__mes_query_status__fence_only_immediate = 1, + command__mes_query_status__fence_only_after_write_ack = 2, + command__mes_query_status__fence_wait_for_write_ack_send_interrupt = 3 +}; + +enum query_status_engine_sel_enum { + engine_sel__mes_query_status__compute = 0, + engine_sel__mes_query_status__sdma0_queue = 2, + engine_sel__mes_query_status__sdma1_queue = 3 +}; + +struct pm4_query_status { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t context_id:28; + enum query_status_interrupt_sel_enum interrupt_sel:2; + enum query_status_command_enum command:2; + } bitfields2; + uint32_t ordinal2; + }; + + union { + struct { + uint32_t pasid:16; + uint32_t reserved1:16; + } bitfields3a; + struct { + uint32_t reserved2:2; + uint32_t doorbell_offset:21; + uint32_t reserved3:3; + enum query_status_engine_sel_enum engine_sel:3; + uint32_t reserved4:3; + } bitfields3b; + uint32_t ordinal3; + }; + + uint32_t addr_lo; + uint32_t addr_hi; + uint32_t data_lo; + uint32_t data_hi; +}; +#endif + +/*--------------------MES_UNMAP_QUEUES--------------------*/ + +#ifndef PM4_MES_UNMAP_QUEUES_DEFINED +#define PM4_MES_UNMAP_QUEUES_DEFINED +enum unmap_queues_action_enum { + action__mes_unmap_queues__preempt_queues = 0, + action__mes_unmap_queues__reset_queues = 1, + action__mes_unmap_queues__disable_process_queues = 2 +}; + +enum unmap_queues_queue_sel_enum { + queue_sel__mes_unmap_queues__perform_request_on_specified_queues = 0, + queue_sel__mes_unmap_queues__perform_request_on_pasid_queues = 1, + queue_sel__mes_unmap_queues__perform_request_on_all_active_queues = 2 +}; + +enum unmap_queues_engine_sel_enum { + engine_sel__mes_unmap_queues__compute = 0, + engine_sel__mes_unmap_queues__sdma0 = 2, + engine_sel__mes_unmap_queues__sdma1 = 3 +}; + +struct pm4_unmap_queues { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + enum unmap_queues_action_enum action:2; + uint32_t reserved1:2; + enum unmap_queues_queue_sel_enum queue_sel:2; + uint32_t reserved2:20; + enum unmap_queues_engine_sel_enum engine_sel:3; + uint32_t num_queues:3; + } bitfields2; + uint32_t ordinal2; + }; + + union { + struct { + uint32_t pasid:16; + uint32_t reserved3:16; + } bitfields3a; + struct { + uint32_t reserved4:2; + uint32_t doorbell_offset0:21; + uint32_t reserved5:9; + } bitfields3b; + uint32_t ordinal3; + }; + + union { + struct { + uint32_t reserved6:2; + uint32_t doorbell_offset1:21; + uint32_t reserved7:9; + } bitfields4; + uint32_t ordinal4; + }; + + union { + struct { + uint32_t reserved8:2; + uint32_t doorbell_offset2:21; + uint32_t reserved9:9; + } bitfields5; + uint32_t ordinal5; + }; + + union { + struct { + uint32_t reserved10:2; + uint32_t doorbell_offset3:21; + uint32_t reserved11:9; + } bitfields6; + uint32_t ordinal6; + }; + +}; +#endif + +enum { + CACHE_FLUSH_AND_INV_TS_EVENT = 0x00000014 +}; + +#endif /* KFD_PM4_HEADERS_H_ */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_opcodes.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_opcodes.h new file mode 100644 index 0000000..b72fa3b --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_opcodes.h @@ -0,0 +1,107 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + + +#ifndef KFD_PM4_OPCODES_H +#define KFD_PM4_OPCODES_H + +enum it_opcode_type { + IT_NOP = 0x10, + IT_SET_BASE = 0x11, + IT_CLEAR_STATE = 0x12, + IT_INDEX_BUFFER_SIZE = 0x13, + IT_DISPATCH_DIRECT = 0x15, + IT_DISPATCH_INDIRECT = 0x16, + IT_ATOMIC_GDS = 0x1D, + IT_OCCLUSION_QUERY = 0x1F, + IT_SET_PREDICATION = 0x20, + IT_REG_RMW = 0x21, + IT_COND_EXEC = 0x22, + IT_PRED_EXEC = 0x23, + IT_DRAW_INDIRECT = 0x24, + IT_DRAW_INDEX_INDIRECT = 0x25, + IT_INDEX_BASE = 0x26, + IT_DRAW_INDEX_2 = 0x27, + IT_CONTEXT_CONTROL = 0x28, + IT_INDEX_TYPE = 0x2A, + IT_DRAW_INDIRECT_MULTI = 0x2C, + IT_DRAW_INDEX_AUTO = 0x2D, + IT_NUM_INSTANCES = 0x2F, + IT_DRAW_INDEX_MULTI_AUTO = 0x30, + IT_INDIRECT_BUFFER_CNST = 0x33, + IT_STRMOUT_BUFFER_UPDATE = 0x34, + IT_DRAW_INDEX_OFFSET_2 = 0x35, + IT_DRAW_PREAMBLE = 0x36, + IT_WRITE_DATA = 0x37, + IT_DRAW_INDEX_INDIRECT_MULTI = 0x38, + IT_MEM_SEMAPHORE = 0x39, + IT_COPY_DW = 0x3B, + IT_WAIT_REG_MEM = 0x3C, + IT_INDIRECT_BUFFER = 0x3F, + IT_COPY_DATA = 0x40, + IT_PFP_SYNC_ME = 0x42, + IT_SURFACE_SYNC = 0x43, + IT_COND_WRITE = 0x45, + IT_EVENT_WRITE = 0x46, + IT_EVENT_WRITE_EOP = 0x47, + IT_EVENT_WRITE_EOS = 0x48, + IT_RELEASE_MEM = 0x49, + IT_PREAMBLE_CNTL = 0x4A, + IT_DMA_DATA = 0x50, + IT_ACQUIRE_MEM = 0x58, + IT_REWIND = 0x59, + IT_LOAD_UCONFIG_REG = 0x5E, + IT_LOAD_SH_REG = 0x5F, + IT_LOAD_CONFIG_REG = 0x60, + IT_LOAD_CONTEXT_REG = 0x61, + IT_SET_CONFIG_REG = 0x68, + IT_SET_CONTEXT_REG = 0x69, + IT_SET_CONTEXT_REG_INDIRECT = 0x73, + IT_SET_SH_REG = 0x76, + IT_SET_SH_REG_OFFSET = 0x77, + IT_SET_QUEUE_REG = 0x78, + IT_SET_UCONFIG_REG = 0x79, + IT_SCRATCH_RAM_WRITE = 0x7D, + IT_SCRATCH_RAM_READ = 0x7E, + IT_LOAD_CONST_RAM = 0x80, + IT_WRITE_CONST_RAM = 0x81, + IT_DUMP_CONST_RAM = 0x83, + IT_INCREMENT_CE_COUNTER = 0x84, + IT_INCREMENT_DE_COUNTER = 0x85, + IT_WAIT_ON_CE_COUNTER = 0x86, + IT_WAIT_ON_DE_COUNTER_DIFF = 0x88, + IT_SWITCH_BUFFER = 0x8B, + IT_SET_RESOURCES = 0xA0, + IT_MAP_PROCESS = 0xA1, + IT_MAP_QUEUES = 0xA2, + IT_UNMAP_QUEUES = 0xA3, + IT_QUERY_STATUS = 0xA4, + IT_RUN_LIST = 0xA5, +}; + +#define PM4_TYPE_0 0 +#define PM4_TYPE_2 2 +#define PM4_TYPE_3 3 + +#endif /* KFD_PM4_OPCODES_H */ + diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 0eb2032..bba3c2a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -35,6 +35,16 @@ #define KFD_SYSFS_FILE_MODE 0444 +/* + * When working with cp scheduler we should assign the HIQ manually or via + * the radeon driver to a fixed hqd slot, here are the fixed HIQ hqd slot + * definitions for Kaveri. In Kaveri only the first ME queues participates + * in the cp scheduling taking that in mind we set the HIQ slot in the + * second ME. + */ +#define KFD_CIK_HIQ_PIPE 4 +#define KFD_CIK_HIQ_QUEUE 0 + /* GPU ID hash width in bits */ #define KFD_GPU_ID_HASH_WIDTH 16 @@ -57,6 +67,13 @@ extern int max_num_of_queues_per_process; #define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS_DEFAULT 128 #define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS 1024 +#define KFD_KERNEL_QUEUE_SIZE 2048 + +enum cache_policy { + cache_policy_coherent, + cache_policy_noncoherent +}; + struct kfd_device_info { unsigned int max_pasid_bits; size_t ih_ring_entry_size; @@ -88,8 +105,10 @@ struct kfd_dev { struct kgd2kfd_shared_resources shared_resources; - bool init_complete; + /* QCM Device instance */ + struct device_queue_manager *dqm; + bool init_complete; }; /* KGD2KFD callbacks */ @@ -384,6 +403,21 @@ int kgd2kfd_resume(struct kfd_dev *dev); /* amdkfd Apertures */ int kfd_init_apertures(struct kfd_process *process); +/* Queue Context Management */ +int init_queue(struct queue **q, struct queue_properties properties); +void uninit_queue(struct queue *q); +void print_queue(struct queue *q); + +/* Packet Manager */ + +struct packet_manager { + struct device_queue_manager *dqm; + struct kernel_queue *priv_queue; + struct mutex lock; + bool allocated; + struct kfd_mem_obj *ib_buffer_obj; +}; + uint64_t kfd_get_number_elems(struct kfd_dev *kfd); phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev, struct kfd_process *process); -- cgit v0.10.2 From 31c21fece7a5f279ebdd1ee260493450017183f4 Mon Sep 17 00:00:00 2001 From: Ben Goz Date: Thu, 17 Jul 2014 00:48:28 +0300 Subject: amdkfd: Add module parameter of scheduling policy This patch adds a new parameter to the amdkfd driver. This parameter enables the user to select the scheduling policy of the CP. The choices are: * CP Scheduling with support for over-subscription * CP Scheduling without support for over-subscription * Without CP Scheduling Note that the third option (Without CP scheduling) is only for debug purposes and bringup of new H/W. As such, it is _not_ guaranteed to work at all times on all H/W versions. v3: Fixed description of parameter, changed the permissions to read_only, added a verification of the value and added documentation v5: Set default sched_policy to HWS as it is now supported by firmware Signed-off-by: Ben Goz Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c index a05116b..95d5af1 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -45,6 +45,11 @@ static const struct kgd2kfd_calls kgd2kfd = { .resume = kgd2kfd_resume, }; +int sched_policy = KFD_SCHED_POLICY_HWS; +module_param(sched_policy, int, 0444); +MODULE_PARM_DESC(sched_policy, + "Kernel cmdline parameter that defines the amdkfd scheduling policy"); + int max_num_of_processes = KFD_MAX_NUM_OF_PROCESSES_DEFAULT; module_param(max_num_of_processes, int, 0444); MODULE_PARM_DESC(max_num_of_processes, @@ -88,6 +93,13 @@ static int __init kfd_module_init(void) kfd2kgd = NULL; /* Verify module parameters */ + if ((sched_policy < KFD_SCHED_POLICY_HWS) || + (sched_policy > KFD_SCHED_POLICY_NO_HWS)) { + pr_err("kfd: sched_policy has invalid value\n"); + return -1; + } + + /* Verify module parameters */ if ((max_num_of_processes < 0) || (max_num_of_processes > KFD_MAX_NUM_OF_PROCESSES)) { pr_err("kfd: max_num_of_processes must be between 0 to KFD_MAX_NUM_OF_PROCESSES\n"); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index bba3c2a..32bca8a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -69,6 +69,36 @@ extern int max_num_of_queues_per_process; #define KFD_KERNEL_QUEUE_SIZE 2048 +/* Kernel module parameter to specify the scheduling policy */ +extern int sched_policy; + +/** + * enum kfd_sched_policy + * + * @KFD_SCHED_POLICY_HWS: H/W scheduling policy known as command processor (cp) + * scheduling. In this scheduling mode we're using the firmware code to + * schedule the user mode queues and kernel queues such as HIQ and DIQ. + * the HIQ queue is used as a special queue that dispatches the configuration + * to the cp and the user mode queues list that are currently running. + * the DIQ queue is a debugging queue that dispatches debugging commands to the + * firmware. + * in this scheduling mode user mode queues over subscription feature is + * enabled. + * + * @KFD_SCHED_POLICY_HWS_NO_OVERSUBSCRIPTION: The same as above but the over + * subscription feature disabled. + * + * @KFD_SCHED_POLICY_NO_HWS: no H/W scheduling policy is a mode which directly + * set the command processor registers and sets the queues "manually". This + * mode is used *ONLY* for debugging proposes. + * + */ +enum kfd_sched_policy { + KFD_SCHED_POLICY_HWS = 0, + KFD_SCHED_POLICY_HWS_NO_OVERSUBSCRIPTION, + KFD_SCHED_POLICY_NO_HWS +}; + enum cache_policy { cache_policy_coherent, cache_policy_noncoherent -- cgit v0.10.2 From 241f24f823631ced1c6f162db1deebc35037f7d0 Mon Sep 17 00:00:00 2001 From: Ben Goz Date: Thu, 17 Jul 2014 00:55:28 +0300 Subject: amdkfd: Add packet manager module The packet manager module builds PM4 packets for the sole use of the CP scheduler. Those packets are used by the HIQ to submit runlists to the CP. v3: Removed include of cik_mqds.h Changed lower_32/upper_32 calls to use linux macros Used new gart allocation functions Added documentation v5: Move amdkfd from drm/radeon/ to drm/amd/ Change format of mqd structure to match latest KV firmware Add support for AQL queues creation to enable working with open-source HSA runtime Always chain runlist if you have more than 1 process or if you have over-subscription over the number of queues. Various fixes (typos, style) Signed-off-by: Ben Goz Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 047b834..1164a3a 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -7,6 +7,6 @@ ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/amd/include/ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ kfd_process.o kfd_queue.o kfd_mqd_manager.o \ - kfd_kernel_queue.o + kfd_kernel_queue.o kfd_packet_manager.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c new file mode 100644 index 0000000..5ce9233 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -0,0 +1,565 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include "kfd_device_queue_manager.h" +#include "kfd_kernel_queue.h" +#include "kfd_priv.h" +#include "kfd_pm4_headers.h" +#include "kfd_pm4_opcodes.h" + +static inline void inc_wptr(unsigned int *wptr, unsigned int increment_bytes, + unsigned int buffer_size_bytes) +{ + unsigned int temp = *wptr + increment_bytes / sizeof(uint32_t); + + BUG_ON((temp * sizeof(uint32_t)) > buffer_size_bytes); + *wptr = temp; +} + +static unsigned int build_pm4_header(unsigned int opcode, size_t packet_size) +{ + union PM4_MES_TYPE_3_HEADER header; + + header.u32all = 0; + header.opcode = opcode; + header.count = packet_size/sizeof(uint32_t) - 2; + header.type = PM4_TYPE_3; + + return header.u32all; +} + +static void pm_calc_rlib_size(struct packet_manager *pm, + unsigned int *rlib_size, + bool *over_subscription) +{ + unsigned int process_count, queue_count; + + BUG_ON(!pm || !rlib_size || !over_subscription); + + process_count = pm->dqm->processes_count; + queue_count = pm->dqm->queue_count; + + /* check if there is over subscription*/ + *over_subscription = false; + if ((process_count > 1) || + queue_count > PIPE_PER_ME_CP_SCHEDULING * QUEUES_PER_PIPE) { + *over_subscription = true; + pr_debug("kfd: over subscribed runlist\n"); + } + + /* calculate run list ib allocation size */ + *rlib_size = process_count * sizeof(struct pm4_map_process) + + queue_count * sizeof(struct pm4_map_queues); + + /* + * Increase the allocation size in case we need a chained run list + * when over subscription + */ + if (*over_subscription) + *rlib_size += sizeof(struct pm4_runlist); + + pr_debug("kfd: runlist ib size %d\n", *rlib_size); +} + +static int pm_allocate_runlist_ib(struct packet_manager *pm, + unsigned int **rl_buffer, + uint64_t *rl_gpu_buffer, + unsigned int *rl_buffer_size, + bool *is_over_subscription) +{ + int retval; + + BUG_ON(!pm); + BUG_ON(pm->allocated == true); + BUG_ON(is_over_subscription == NULL); + + pm_calc_rlib_size(pm, rl_buffer_size, is_over_subscription); + + retval = kfd2kgd->allocate_mem(pm->dqm->dev->kgd, + *rl_buffer_size, + PAGE_SIZE, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) &pm->ib_buffer_obj); + + if (retval != 0) { + pr_err("kfd: failed to allocate runlist IB\n"); + return retval; + } + + *(void **)rl_buffer = pm->ib_buffer_obj->cpu_ptr; + *rl_gpu_buffer = pm->ib_buffer_obj->gpu_addr; + + memset(*rl_buffer, 0, *rl_buffer_size); + pm->allocated = true; + return retval; +} + +static int pm_create_runlist(struct packet_manager *pm, uint32_t *buffer, + uint64_t ib, size_t ib_size_in_dwords, bool chain) +{ + struct pm4_runlist *packet; + + BUG_ON(!pm || !buffer || !ib); + + packet = (struct pm4_runlist *)buffer; + + memset(buffer, 0, sizeof(struct pm4_runlist)); + packet->header.u32all = build_pm4_header(IT_RUN_LIST, + sizeof(struct pm4_runlist)); + + packet->bitfields4.ib_size = ib_size_in_dwords; + packet->bitfields4.chain = chain ? 1 : 0; + packet->bitfields4.offload_polling = 0; + packet->bitfields4.valid = 1; + packet->ordinal2 = lower_32_bits(ib); + packet->bitfields3.ib_base_hi = upper_32_bits(ib); + + return 0; +} + +static int pm_create_map_process(struct packet_manager *pm, uint32_t *buffer, + struct qcm_process_device *qpd) +{ + struct pm4_map_process *packet; + struct queue *cur; + uint32_t num_queues; + + BUG_ON(!pm || !buffer || !qpd); + + packet = (struct pm4_map_process *)buffer; + + pr_debug("kfd: In func %s\n", __func__); + + memset(buffer, 0, sizeof(struct pm4_map_process)); + + packet->header.u32all = build_pm4_header(IT_MAP_PROCESS, + sizeof(struct pm4_map_process)); + packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0; + packet->bitfields2.process_quantum = 1; + packet->bitfields2.pasid = qpd->pqm->process->pasid; + packet->bitfields3.page_table_base = qpd->page_table_base; + packet->bitfields10.gds_size = qpd->gds_size; + packet->bitfields10.num_gws = qpd->num_gws; + packet->bitfields10.num_oac = qpd->num_oac; + num_queues = 0; + list_for_each_entry(cur, &qpd->queues_list, list) + num_queues++; + packet->bitfields10.num_queues = num_queues; + + packet->sh_mem_config = qpd->sh_mem_config; + packet->sh_mem_bases = qpd->sh_mem_bases; + packet->sh_mem_ape1_base = qpd->sh_mem_ape1_base; + packet->sh_mem_ape1_limit = qpd->sh_mem_ape1_limit; + + packet->gds_addr_lo = lower_32_bits(qpd->gds_context_area); + packet->gds_addr_hi = upper_32_bits(qpd->gds_context_area); + + return 0; +} + +static int pm_create_map_queue(struct packet_manager *pm, uint32_t *buffer, + struct queue *q) +{ + struct pm4_map_queues *packet; + + BUG_ON(!pm || !buffer || !q); + + pr_debug("kfd: In func %s\n", __func__); + + packet = (struct pm4_map_queues *)buffer; + memset(buffer, 0, sizeof(struct pm4_map_queues)); + + packet->header.u32all = build_pm4_header(IT_MAP_QUEUES, + sizeof(struct pm4_map_queues)); + packet->bitfields2.alloc_format = + alloc_format__mes_map_queues__one_per_pipe; + packet->bitfields2.num_queues = 1; + packet->bitfields2.queue_sel = + queue_sel__mes_map_queues__map_to_hws_determined_queue_slots; + + packet->bitfields2.vidmem = (q->properties.is_interop) ? + vidmem__mes_map_queues__uses_video_memory : + vidmem__mes_map_queues__uses_no_video_memory; + + switch (q->properties.type) { + case KFD_QUEUE_TYPE_COMPUTE: + case KFD_QUEUE_TYPE_DIQ: + packet->bitfields2.engine_sel = + engine_sel__mes_map_queues__compute; + break; + case KFD_QUEUE_TYPE_SDMA: + packet->bitfields2.engine_sel = + engine_sel__mes_map_queues__sdma0; + break; + default: + BUG(); + break; + } + + packet->mes_map_queues_ordinals[0].bitfields3.doorbell_offset = + q->properties.doorbell_off; + + packet->mes_map_queues_ordinals[0].mqd_addr_lo = + lower_32_bits(q->gart_mqd_addr); + + packet->mes_map_queues_ordinals[0].mqd_addr_hi = + upper_32_bits(q->gart_mqd_addr); + + packet->mes_map_queues_ordinals[0].wptr_addr_lo = + lower_32_bits((uint64_t)q->properties.write_ptr); + + packet->mes_map_queues_ordinals[0].wptr_addr_hi = + upper_32_bits((uint64_t)q->properties.write_ptr); + + return 0; +} + +static int pm_create_runlist_ib(struct packet_manager *pm, + struct list_head *queues, + uint64_t *rl_gpu_addr, + size_t *rl_size_bytes) +{ + unsigned int alloc_size_bytes; + unsigned int *rl_buffer, rl_wptr, i; + int retval, proccesses_mapped; + struct device_process_node *cur; + struct qcm_process_device *qpd; + struct queue *q; + struct kernel_queue *kq; + bool is_over_subscription; + + BUG_ON(!pm || !queues || !rl_size_bytes || !rl_gpu_addr); + + rl_wptr = retval = proccesses_mapped = 0; + + retval = pm_allocate_runlist_ib(pm, &rl_buffer, rl_gpu_addr, + &alloc_size_bytes, &is_over_subscription); + if (retval != 0) + return retval; + + *rl_size_bytes = alloc_size_bytes; + + pr_debug("kfd: In func %s\n", __func__); + pr_debug("kfd: building runlist ib process count: %d queues count %d\n", + pm->dqm->processes_count, pm->dqm->queue_count); + + /* build the run list ib packet */ + list_for_each_entry(cur, queues, list) { + qpd = cur->qpd; + /* build map process packet */ + if (proccesses_mapped >= pm->dqm->processes_count) { + pr_debug("kfd: not enough space left in runlist IB\n"); + pm_release_ib(pm); + return -ENOMEM; + } + retval = pm_create_map_process(pm, &rl_buffer[rl_wptr], qpd); + if (retval != 0) + return retval; + proccesses_mapped++; + inc_wptr(&rl_wptr, sizeof(struct pm4_map_process), + alloc_size_bytes); + + list_for_each_entry(kq, &qpd->priv_queue_list, list) { + if (kq->queue->properties.is_active != true) + continue; + retval = pm_create_map_queue(pm, &rl_buffer[rl_wptr], + kq->queue); + if (retval != 0) + return retval; + inc_wptr(&rl_wptr, sizeof(struct pm4_map_queues), + alloc_size_bytes); + } + + list_for_each_entry(q, &qpd->queues_list, list) { + if (q->properties.is_active != true) + continue; + retval = pm_create_map_queue(pm, + &rl_buffer[rl_wptr], q); + if (retval != 0) + return retval; + inc_wptr(&rl_wptr, sizeof(struct pm4_map_queues), + alloc_size_bytes); + } + } + + pr_debug("kfd: finished map process and queues to runlist\n"); + + if (is_over_subscription) + pm_create_runlist(pm, &rl_buffer[rl_wptr], *rl_gpu_addr, + alloc_size_bytes / sizeof(uint32_t), true); + + for (i = 0; i < alloc_size_bytes / sizeof(uint32_t); i++) + pr_debug("0x%2X ", rl_buffer[i]); + pr_debug("\n"); + + return 0; +} + +int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm) +{ + BUG_ON(!dqm); + + pm->dqm = dqm; + mutex_init(&pm->lock); + pm->priv_queue = kernel_queue_init(dqm->dev, KFD_QUEUE_TYPE_HIQ); + if (pm->priv_queue == NULL) { + mutex_destroy(&pm->lock); + return -ENOMEM; + } + pm->allocated = false; + + return 0; +} + +void pm_uninit(struct packet_manager *pm) +{ + BUG_ON(!pm); + + mutex_destroy(&pm->lock); + kernel_queue_uninit(pm->priv_queue); +} + +int pm_send_set_resources(struct packet_manager *pm, + struct scheduling_resources *res) +{ + struct pm4_set_resources *packet; + + BUG_ON(!pm || !res); + + pr_debug("kfd: In func %s\n", __func__); + + mutex_lock(&pm->lock); + pm->priv_queue->acquire_packet_buffer(pm->priv_queue, + sizeof(*packet) / sizeof(uint32_t), + (unsigned int **)&packet); + if (packet == NULL) { + mutex_unlock(&pm->lock); + pr_err("kfd: failed to allocate buffer on kernel queue\n"); + return -ENOMEM; + } + + memset(packet, 0, sizeof(struct pm4_set_resources)); + packet->header.u32all = build_pm4_header(IT_SET_RESOURCES, + sizeof(struct pm4_set_resources)); + + packet->bitfields2.queue_type = + queue_type__mes_set_resources__hsa_interface_queue_hiq; + packet->bitfields2.vmid_mask = res->vmid_mask; + packet->bitfields2.unmap_latency = KFD_UNMAP_LATENCY; + packet->bitfields7.oac_mask = res->oac_mask; + packet->bitfields8.gds_heap_base = res->gds_heap_base; + packet->bitfields8.gds_heap_size = res->gds_heap_size; + + packet->gws_mask_lo = lower_32_bits(res->gws_mask); + packet->gws_mask_hi = upper_32_bits(res->gws_mask); + + packet->queue_mask_lo = lower_32_bits(res->queue_mask); + packet->queue_mask_hi = upper_32_bits(res->queue_mask); + + pm->priv_queue->submit_packet(pm->priv_queue); + pm->priv_queue->sync_with_hw(pm->priv_queue, KFD_HIQ_TIMEOUT); + + mutex_unlock(&pm->lock); + + return 0; +} + +int pm_send_runlist(struct packet_manager *pm, struct list_head *dqm_queues) +{ + uint64_t rl_gpu_ib_addr; + uint32_t *rl_buffer; + size_t rl_ib_size, packet_size_dwords; + int retval; + + BUG_ON(!pm || !dqm_queues); + + retval = pm_create_runlist_ib(pm, dqm_queues, &rl_gpu_ib_addr, + &rl_ib_size); + if (retval != 0) + goto fail_create_runlist_ib; + + pr_debug("kfd: runlist IB address: 0x%llX\n", rl_gpu_ib_addr); + + packet_size_dwords = sizeof(struct pm4_runlist) / sizeof(uint32_t); + mutex_lock(&pm->lock); + + retval = pm->priv_queue->acquire_packet_buffer(pm->priv_queue, + packet_size_dwords, &rl_buffer); + if (retval != 0) + goto fail_acquire_packet_buffer; + + retval = pm_create_runlist(pm, rl_buffer, rl_gpu_ib_addr, + rl_ib_size / sizeof(uint32_t), false); + if (retval != 0) + goto fail_create_runlist; + + pm->priv_queue->submit_packet(pm->priv_queue); + pm->priv_queue->sync_with_hw(pm->priv_queue, KFD_HIQ_TIMEOUT); + + mutex_unlock(&pm->lock); + + return retval; + +fail_create_runlist: + pm->priv_queue->rollback_packet(pm->priv_queue); +fail_acquire_packet_buffer: + mutex_unlock(&pm->lock); +fail_create_runlist_ib: + if (pm->allocated == true) + pm_release_ib(pm); + return retval; +} + +int pm_send_query_status(struct packet_manager *pm, uint64_t fence_address, + uint32_t fence_value) +{ + int retval; + struct pm4_query_status *packet; + + BUG_ON(!pm || !fence_address); + + mutex_lock(&pm->lock); + retval = pm->priv_queue->acquire_packet_buffer( + pm->priv_queue, + sizeof(struct pm4_query_status) / sizeof(uint32_t), + (unsigned int **)&packet); + if (retval != 0) + goto fail_acquire_packet_buffer; + + packet->header.u32all = build_pm4_header(IT_QUERY_STATUS, + sizeof(struct pm4_query_status)); + + packet->bitfields2.context_id = 0; + packet->bitfields2.interrupt_sel = + interrupt_sel__mes_query_status__completion_status; + packet->bitfields2.command = + command__mes_query_status__fence_only_after_write_ack; + + packet->addr_hi = upper_32_bits((uint64_t)fence_address); + packet->addr_lo = lower_32_bits((uint64_t)fence_address); + packet->data_hi = upper_32_bits((uint64_t)fence_value); + packet->data_lo = lower_32_bits((uint64_t)fence_value); + + pm->priv_queue->submit_packet(pm->priv_queue); + pm->priv_queue->sync_with_hw(pm->priv_queue, KFD_HIQ_TIMEOUT); + mutex_unlock(&pm->lock); + + return 0; + +fail_acquire_packet_buffer: + mutex_unlock(&pm->lock); + return retval; +} + +int pm_send_unmap_queue(struct packet_manager *pm, enum kfd_queue_type type, + enum kfd_preempt_type_filter mode, + uint32_t filter_param, bool reset, + unsigned int sdma_engine) +{ + int retval; + uint32_t *buffer; + struct pm4_unmap_queues *packet; + + BUG_ON(!pm); + + mutex_lock(&pm->lock); + retval = pm->priv_queue->acquire_packet_buffer( + pm->priv_queue, + sizeof(struct pm4_unmap_queues) / sizeof(uint32_t), + &buffer); + if (retval != 0) + goto err_acquire_packet_buffer; + + packet = (struct pm4_unmap_queues *)buffer; + memset(buffer, 0, sizeof(struct pm4_unmap_queues)); + + packet->header.u32all = build_pm4_header(IT_UNMAP_QUEUES, + sizeof(struct pm4_unmap_queues)); + switch (type) { + case KFD_QUEUE_TYPE_COMPUTE: + case KFD_QUEUE_TYPE_DIQ: + packet->bitfields2.engine_sel = + engine_sel__mes_unmap_queues__compute; + break; + case KFD_QUEUE_TYPE_SDMA: + packet->bitfields2.engine_sel = + engine_sel__mes_unmap_queues__sdma0 + sdma_engine; + break; + default: + BUG(); + break; + } + + if (reset) + packet->bitfields2.action = + action__mes_unmap_queues__reset_queues; + else + packet->bitfields2.action = + action__mes_unmap_queues__preempt_queues; + + switch (mode) { + case KFD_PREEMPT_TYPE_FILTER_SINGLE_QUEUE: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__perform_request_on_specified_queues; + packet->bitfields2.num_queues = 1; + packet->bitfields3b.doorbell_offset0 = filter_param; + break; + case KFD_PREEMPT_TYPE_FILTER_BY_PASID: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__perform_request_on_pasid_queues; + packet->bitfields3a.pasid = filter_param; + break; + case KFD_PREEMPT_TYPE_FILTER_ALL_QUEUES: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__perform_request_on_all_active_queues; + break; + default: + BUG(); + break; + }; + + pm->priv_queue->submit_packet(pm->priv_queue); + pm->priv_queue->sync_with_hw(pm->priv_queue, KFD_HIQ_TIMEOUT); + + mutex_unlock(&pm->lock); + return 0; + +err_acquire_packet_buffer: + mutex_unlock(&pm->lock); + return retval; +} + +void pm_release_ib(struct packet_manager *pm) +{ + BUG_ON(!pm); + + mutex_lock(&pm->lock); + if (pm->allocated) { + kfd2kgd->free_mem(pm->dqm->dev->kgd, + (struct kgd_mem *) pm->ib_buffer_obj); + pm->allocated = false; + } + mutex_unlock(&pm->lock); +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 32bca8a..219c85a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -167,6 +167,23 @@ int kfd_chardev_init(void); void kfd_chardev_exit(void); struct device *kfd_chardev(void); +/** + * enum kfd_preempt_type_filter + * + * @KFD_PREEMPT_TYPE_FILTER_SINGLE_QUEUE: Preempts single queue. + * + * @KFD_PRERMPT_TYPE_FILTER_ALL_QUEUES: Preempts all queues in the + * running queues list. + * + * @KFD_PRERMPT_TYPE_FILTER_BY_PASID: Preempts queues that belongs to + * specific process. + * + */ +enum kfd_preempt_type_filter { + KFD_PREEMPT_TYPE_FILTER_SINGLE_QUEUE, + KFD_PREEMPT_TYPE_FILTER_ALL_QUEUES, + KFD_PREEMPT_TYPE_FILTER_BY_PASID +}; enum kfd_preempt_type { KFD_PREEMPT_TYPE_WAVEFRONT, @@ -313,6 +330,51 @@ enum KFD_MQD_TYPE { KFD_MQD_TYPE_MAX }; +struct scheduling_resources { + unsigned int vmid_mask; + enum kfd_queue_type type; + uint64_t queue_mask; + uint64_t gws_mask; + uint32_t oac_mask; + uint32_t gds_heap_base; + uint32_t gds_heap_size; +}; + +struct process_queue_manager { + /* data */ + struct kfd_process *process; + unsigned int num_concurrent_processes; + struct list_head queues; + unsigned long *queue_slot_bitmap; +}; + +struct qcm_process_device { + /* The Device Queue Manager that owns this data */ + struct device_queue_manager *dqm; + struct process_queue_manager *pqm; + /* Device Queue Manager lock */ + struct mutex *lock; + /* Queues list */ + struct list_head queues_list; + struct list_head priv_queue_list; + + unsigned int queue_count; + unsigned int vmid; + bool is_debug; + /* + * All the memory management data should be here too + */ + uint64_t gds_context_area; + uint32_t sh_mem_config; + uint32_t sh_mem_bases; + uint32_t sh_mem_ape1_base; + uint32_t sh_mem_ape1_limit; + uint32_t page_table_base; + uint32_t gds_size; + uint32_t num_gws; + uint32_t num_oac; +}; + /* Data that is per-process-per device. */ struct kfd_process_device { /* @@ -434,12 +496,23 @@ int kgd2kfd_resume(struct kfd_dev *dev); int kfd_init_apertures(struct kfd_process *process); /* Queue Context Management */ +inline uint32_t lower_32(uint64_t x); +inline uint32_t upper_32(uint64_t x); + int init_queue(struct queue **q, struct queue_properties properties); void uninit_queue(struct queue *q); void print_queue(struct queue *q); +struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, + enum kfd_queue_type type); +void kernel_queue_uninit(struct kernel_queue *kq); + /* Packet Manager */ +#define KFD_HIQ_TIMEOUT (500) + +#define KFD_UNMAP_LATENCY (150) + struct packet_manager { struct device_queue_manager *dqm; struct kernel_queue *priv_queue; @@ -448,6 +521,8 @@ struct packet_manager { struct kfd_mem_obj *ib_buffer_obj; }; +void pm_release_ib(struct packet_manager *pm); + uint64_t kfd_get_number_elems(struct kfd_dev *kfd); phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev, struct kfd_process *process); -- cgit v0.10.2 From 45102048f77ef72a57a877efad0f012b917d4df7 Mon Sep 17 00:00:00 2001 From: Ben Goz Date: Thu, 17 Jul 2014 01:04:10 +0300 Subject: amdkfd: Add process queue manager module The queue scheduler divides into two sections, one section is process bounded and the other section is device bounded. The process bounded section is handled by this module. The PQM handles usermode queue setup, updates and tear-down. v3: Used kernel parameter to limit queues per process instead of define Added use of doorbell address from user v4: Modified pqm_create_queue so that only when creating usermode queues the driver should return the queue properties to the userspace. Added an info message print when no more queues can be opened because of the queue per process limitation v5: Move amdkfd from drm/radeon/ to drm/amd/ Various fixes Signed-off-by: Ben Goz Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 1164a3a..104f279 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -7,6 +7,7 @@ ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/amd/include/ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ kfd_process.o kfd_queue.o kfd_mqd_manager.o \ - kfd_kernel_queue.o kfd_packet_manager.o + kfd_kernel_queue.o kfd_packet_manager.o \ + kfd_process_queue_manager.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 219c85a..72b2e27 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -387,6 +387,9 @@ struct kfd_process_device { struct kfd_dev *dev; + /* per-process-per device QCM data structure */ + struct qcm_process_device qpd; + /*Apertures*/ uint64_t lds_base; uint64_t lds_limit; @@ -432,6 +435,8 @@ struct kfd_process { */ struct list_head per_device_data; + struct process_queue_manager pqm; + /* The process's queues. */ size_t queue_array_size; @@ -501,12 +506,33 @@ inline uint32_t upper_32(uint64_t x); int init_queue(struct queue **q, struct queue_properties properties); void uninit_queue(struct queue *q); +void print_queue_properties(struct queue_properties *q); void print_queue(struct queue *q); struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, enum kfd_queue_type type); void kernel_queue_uninit(struct kernel_queue *kq); +/* Process Queue Manager */ +struct process_queue_node { + struct queue *q; + struct kernel_queue *kq; + struct list_head process_queue_list; +}; + +int pqm_init(struct process_queue_manager *pqm, struct kfd_process *p); +void pqm_uninit(struct process_queue_manager *pqm); +int pqm_create_queue(struct process_queue_manager *pqm, + struct kfd_dev *dev, + struct file *f, + struct queue_properties *properties, + unsigned int flags, + enum kfd_queue_type type, + unsigned int *qid); +int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid); +int pqm_update_queue(struct process_queue_manager *pqm, unsigned int qid, + struct queue_properties *p); + /* Packet Manager */ #define KFD_HIQ_TIMEOUT (500) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 5084794..a17e2dd 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -222,6 +222,13 @@ static void kfd_process_notifier_release(struct mmu_notifier *mn, mutex_unlock(&kfd_processes_mutex); synchronize_srcu(&kfd_processes_srcu); + mutex_lock(&p->mutex); + + /* In case our notifier is called before IOMMU notifier */ + pqm_uninit(&p->pqm); + + mutex_unlock(&p->mutex); + /* * Because we drop mm_count inside kfd_process_destroy_delayed * and because the mmu_notifier_unregister function also drop @@ -274,8 +281,16 @@ static struct kfd_process *create_process(const struct task_struct *thread) INIT_LIST_HEAD(&process->per_device_data); + err = pqm_init(&process->pqm, process); + if (err != 0) + goto err_process_pqm_init; + return process; +err_process_pqm_init: + hash_del_rcu(&process->kfd_processes); + synchronize_rcu(); + mmu_notifier_unregister_no_release(&process->mmu_notifier, process->mm); err_mmu_notifier: kfd_pasid_free(process->pasid); err_alloc_pasid: @@ -300,6 +315,9 @@ struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, pdd = kzalloc(sizeof(*pdd), GFP_KERNEL); if (pdd != NULL) { pdd->dev = dev; + INIT_LIST_HEAD(&pdd->qpd.queues_list); + INIT_LIST_HEAD(&pdd->qpd.priv_queue_list); + pdd->qpd.dqm = dev->dqm; list_add(&pdd->per_device_list, &p->per_device_data); } } @@ -360,6 +378,8 @@ void kfd_unbind_process_from_device(struct kfd_dev *dev, unsigned int pasid) mutex_lock(&p->mutex); + pqm_uninit(&p->pqm); + pdd = kfd_get_process_device_data(dev, p, 0); /* diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c new file mode 100644 index 0000000..c7859fc --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -0,0 +1,342 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include "kfd_device_queue_manager.h" +#include "kfd_priv.h" +#include "kfd_kernel_queue.h" + +static inline struct process_queue_node *get_queue_by_qid( + struct process_queue_manager *pqm, unsigned int qid) +{ + struct process_queue_node *pqn; + + BUG_ON(!pqm); + + list_for_each_entry(pqn, &pqm->queues, process_queue_list) { + if (pqn->q && pqn->q->properties.queue_id == qid) + return pqn; + if (pqn->kq && pqn->kq->queue->properties.queue_id == qid) + return pqn; + } + + return NULL; +} + +static int find_available_queue_slot(struct process_queue_manager *pqm, + unsigned int *qid) +{ + unsigned long found; + + BUG_ON(!pqm || !qid); + + pr_debug("kfd: in %s\n", __func__); + + found = find_first_zero_bit(pqm->queue_slot_bitmap, + max_num_of_queues_per_process); + + pr_debug("kfd: the new slot id %lu\n", found); + + if (found >= max_num_of_queues_per_process) { + pr_info("amdkfd: Can not open more queues for process with pasid %d\n", + pqm->process->pasid); + return -ENOMEM; + } + + set_bit(found, pqm->queue_slot_bitmap); + *qid = found; + + return 0; +} + +int pqm_init(struct process_queue_manager *pqm, struct kfd_process *p) +{ + BUG_ON(!pqm); + + INIT_LIST_HEAD(&pqm->queues); + pqm->queue_slot_bitmap = + kzalloc(DIV_ROUND_UP(max_num_of_queues_per_process, + BITS_PER_BYTE), GFP_KERNEL); + if (pqm->queue_slot_bitmap == NULL) + return -ENOMEM; + pqm->process = p; + + return 0; +} + +void pqm_uninit(struct process_queue_manager *pqm) +{ + int retval; + struct process_queue_node *pqn, *next; + + BUG_ON(!pqm); + + pr_debug("In func %s\n", __func__); + + list_for_each_entry_safe(pqn, next, &pqm->queues, process_queue_list) { + retval = pqm_destroy_queue( + pqm, + (pqn->q != NULL) ? + pqn->q->properties.queue_id : + pqn->kq->queue->properties.queue_id); + + if (retval != 0) { + pr_err("kfd: failed to destroy queue\n"); + return; + } + } + kfree(pqm->queue_slot_bitmap); + pqm->queue_slot_bitmap = NULL; +} + +static int create_cp_queue(struct process_queue_manager *pqm, + struct kfd_dev *dev, struct queue **q, + struct queue_properties *q_properties, + struct file *f, unsigned int qid) +{ + int retval; + + retval = 0; + + /* Doorbell initialized in user space*/ + q_properties->doorbell_ptr = NULL; + + q_properties->doorbell_off = + kfd_queue_id_to_doorbell(dev, pqm->process, qid); + + /* let DQM handle it*/ + q_properties->vmid = 0; + q_properties->queue_id = qid; + q_properties->type = KFD_QUEUE_TYPE_COMPUTE; + + retval = init_queue(q, *q_properties); + if (retval != 0) + goto err_init_queue; + + (*q)->device = dev; + (*q)->process = pqm->process; + + pr_debug("kfd: PQM After init queue"); + + return retval; + +err_init_queue: + return retval; +} + +int pqm_create_queue(struct process_queue_manager *pqm, + struct kfd_dev *dev, + struct file *f, + struct queue_properties *properties, + unsigned int flags, + enum kfd_queue_type type, + unsigned int *qid) +{ + int retval; + struct kfd_process_device *pdd; + struct queue_properties q_properties; + struct queue *q; + struct process_queue_node *pqn; + struct kernel_queue *kq; + + BUG_ON(!pqm || !dev || !properties || !qid); + + memset(&q_properties, 0, sizeof(struct queue_properties)); + memcpy(&q_properties, properties, sizeof(struct queue_properties)); + q = NULL; + kq = NULL; + + pdd = kfd_get_process_device_data(dev, pqm->process, 1); + BUG_ON(!pdd); + + retval = find_available_queue_slot(pqm, qid); + if (retval != 0) + return retval; + + if (list_empty(&pqm->queues)) { + pdd->qpd.pqm = pqm; + dev->dqm->register_process(dev->dqm, &pdd->qpd); + } + + pqn = kzalloc(sizeof(struct process_queue_node), GFP_KERNEL); + if (!pqn) { + retval = -ENOMEM; + goto err_allocate_pqn; + } + + switch (type) { + case KFD_QUEUE_TYPE_COMPUTE: + /* check if there is over subscription */ + if ((sched_policy == KFD_SCHED_POLICY_HWS_NO_OVERSUBSCRIPTION) && + ((dev->dqm->processes_count >= VMID_PER_DEVICE) || + (dev->dqm->queue_count >= PIPE_PER_ME_CP_SCHEDULING * QUEUES_PER_PIPE))) { + pr_err("kfd: over-subscription is not allowed in radeon_kfd.sched_policy == 1\n"); + retval = -EPERM; + goto err_create_queue; + } + + retval = create_cp_queue(pqm, dev, &q, &q_properties, f, *qid); + if (retval != 0) + goto err_create_queue; + pqn->q = q; + pqn->kq = NULL; + retval = dev->dqm->create_queue(dev->dqm, q, &pdd->qpd, + &q->properties.vmid); + print_queue(q); + break; + case KFD_QUEUE_TYPE_DIQ: + kq = kernel_queue_init(dev, KFD_QUEUE_TYPE_DIQ); + if (kq == NULL) { + kernel_queue_uninit(kq); + goto err_create_queue; + } + kq->queue->properties.queue_id = *qid; + pqn->kq = kq; + pqn->q = NULL; + retval = dev->dqm->create_kernel_queue(dev->dqm, kq, &pdd->qpd); + break; + default: + BUG(); + break; + } + + if (retval != 0) { + pr_err("kfd: error dqm create queue\n"); + goto err_create_queue; + } + + pr_debug("kfd: PQM After DQM create queue\n"); + + list_add(&pqn->process_queue_list, &pqm->queues); + + if (q) { + *properties = q->properties; + pr_debug("kfd: PQM done creating queue\n"); + print_queue_properties(properties); + } + + return retval; + +err_create_queue: + kfree(pqn); +err_allocate_pqn: + clear_bit(*qid, pqm->queue_slot_bitmap); + return retval; +} + +int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) +{ + struct process_queue_node *pqn; + struct kfd_process_device *pdd; + struct device_queue_manager *dqm; + struct kfd_dev *dev; + int retval; + + dqm = NULL; + + BUG_ON(!pqm); + retval = 0; + + pr_debug("kfd: In Func %s\n", __func__); + + pqn = get_queue_by_qid(pqm, qid); + if (pqn == NULL) { + pr_err("kfd: queue id does not match any known queue\n"); + return -EINVAL; + } + + dev = NULL; + if (pqn->kq) + dev = pqn->kq->dev; + if (pqn->q) + dev = pqn->q->device; + BUG_ON(!dev); + + pdd = kfd_get_process_device_data(dev, pqm->process, 1); + BUG_ON(!pdd); + + if (pqn->kq) { + /* destroy kernel queue (DIQ) */ + dqm = pqn->kq->dev->dqm; + dqm->destroy_kernel_queue(dqm, pqn->kq, &pdd->qpd); + kernel_queue_uninit(pqn->kq); + } + + if (pqn->q) { + dqm = pqn->q->device->dqm; + retval = dqm->destroy_queue(dqm, &pdd->qpd, pqn->q); + if (retval != 0) + return retval; + + uninit_queue(pqn->q); + } + + list_del(&pqn->process_queue_list); + kfree(pqn); + clear_bit(qid, pqm->queue_slot_bitmap); + + if (list_empty(&pqm->queues)) + dqm->unregister_process(dqm, &pdd->qpd); + + return retval; +} + +int pqm_update_queue(struct process_queue_manager *pqm, unsigned int qid, + struct queue_properties *p) +{ + int retval; + struct process_queue_node *pqn; + + BUG_ON(!pqm); + + pqn = get_queue_by_qid(pqm, qid); + BUG_ON(!pqn); + + pqn->q->properties.queue_address = p->queue_address; + pqn->q->properties.queue_size = p->queue_size; + pqn->q->properties.queue_percent = p->queue_percent; + pqn->q->properties.priority = p->priority; + + retval = pqn->q->device->dqm->update_queue(pqn->q->device->dqm, pqn->q); + if (retval != 0) + return retval; + + return 0; +} + +struct kernel_queue *pqm_get_kernel_queue(struct process_queue_manager *pqm, + unsigned int qid) +{ + struct process_queue_node *pqn; + + BUG_ON(!pqm); + + pqn = get_queue_by_qid(pqm, qid); + if (pqn && pqn->kq) + return pqn->kq; + + return NULL; +} + + -- cgit v0.10.2 From 64c7f8cf792776aaca036fb983006b6b21204934 Mon Sep 17 00:00:00 2001 From: Ben Goz Date: Thu, 17 Jul 2014 01:27:00 +0300 Subject: amdkfd: Add device queue manager module The queue scheduler divides into two sections, one section is process bounded and the other section is device bounded. The device bounded section is handled by this module. The DQM module handles queue setup, update and tear-down from the device side. It also supports suspend/resume operation. v3: Changed device_init, added the use of the new gart allocation functions an Added documentation. v4: Fixed a race in DQM queue scheduler where dqm->lock must be held when accessing dqm->queue_count and dqm->processes_count. This fixes runlist IB allocation failures when DQM is under load. Fixed race in DQM queue destruction where queues being destroyed must be removed from qpd->queues_list prior to preemption, or concurrent queue creation activity may reschedule them while their MQD is destroyed. Fixed EOP queue size setting in CP_HPD_EOP_CONTROL, because the size is specified as (log2(size_dwords)-1). The previous calculation assumed the size was specified in bytes, which caused interference between EOP queues when multiple MEC pipelines were active. v5: Move amdkfd from drm/radeon/ to drm/amd/ Change format of mqd structure to match latest KV firmware Add support for AQL queues creation to enable working with open-source HSA runtime Remove unused unmap_queue function Various fixes (Style, typos) Signed-off-by: Ben Goz Signed-off-by: Jay Cornwall Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 104f279..307a309 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -8,6 +8,6 @@ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ kfd_process.o kfd_queue.o kfd_mqd_manager.o \ kfd_kernel_queue.o kfd_packet_manager.o \ - kfd_process_queue_manager.o + kfd_process_queue_manager.o kfd_device_queue_manager.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 4b1cced..bdf3e94 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -25,6 +25,7 @@ #include #include #include "kfd_priv.h" +#include "kfd_device_queue_manager.h" #define MQD_SIZE_ALIGNED 768 @@ -199,12 +200,34 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, amd_iommu_set_invalidate_ctx_cb(kfd->pdev, iommu_pasid_shutdown_callback); + kfd->dqm = device_queue_manager_init(kfd); + if (!kfd->dqm) { + dev_err(kfd_device, + "Error initializing queue manager for device (%x:%x)\n", + kfd->pdev->vendor, kfd->pdev->device); + goto device_queue_manager_error; + } + + if (kfd->dqm->start(kfd->dqm) != 0) { + dev_err(kfd_device, + "Error starting queuen manager for device (%x:%x)\n", + kfd->pdev->vendor, kfd->pdev->device); + goto dqm_start_error; + } + kfd->init_complete = true; dev_info(kfd_device, "added device (%x:%x)\n", kfd->pdev->vendor, kfd->pdev->device); + pr_debug("kfd: Starting kfd with the following scheduling policy %d\n", + sched_policy); + goto out; +dqm_start_error: + device_queue_manager_uninit(kfd->dqm); +device_queue_manager_error: + amd_iommu_free_device(kfd->pdev); device_iommu_pasid_error: kfd_topology_remove_device(kfd); kfd_topology_add_device_error: @@ -219,6 +242,7 @@ out: void kgd2kfd_device_exit(struct kfd_dev *kfd) { if (kfd->init_complete) { + device_queue_manager_uninit(kfd->dqm); amd_iommu_free_device(kfd->pdev); kfd_topology_remove_device(kfd); } @@ -230,8 +254,10 @@ void kgd2kfd_suspend(struct kfd_dev *kfd) { BUG_ON(kfd == NULL); - if (kfd->init_complete) + if (kfd->init_complete) { + kfd->dqm->stop(kfd->dqm); amd_iommu_free_device(kfd->pdev); + } } int kgd2kfd_resume(struct kfd_dev *kfd) @@ -249,6 +275,7 @@ int kgd2kfd_resume(struct kfd_dev *kfd) return -ENXIO; amd_iommu_set_invalidate_ctx_cb(kfd->pdev, iommu_pasid_shutdown_callback); + kfd->dqm->start(kfd->dqm); } return 0; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c new file mode 100644 index 0000000..8c40d04 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -0,0 +1,1059 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include +#include +#include +#include "kfd_priv.h" +#include "kfd_device_queue_manager.h" +#include "kfd_mqd_manager.h" +#include "cik_regs.h" +#include "kfd_kernel_queue.h" +#include "../../radeon/cik_reg.h" + +/* Size of the per-pipe EOP queue */ +#define CIK_HPD_EOP_BYTES_LOG2 11 +#define CIK_HPD_EOP_BYTES (1U << CIK_HPD_EOP_BYTES_LOG2) + +static bool is_mem_initialized; + +static int init_memory(struct device_queue_manager *dqm); +static int set_pasid_vmid_mapping(struct device_queue_manager *dqm, + unsigned int pasid, unsigned int vmid); + +static int create_compute_queue_nocpsch(struct device_queue_manager *dqm, + struct queue *q, + struct qcm_process_device *qpd); +static int execute_queues_cpsch(struct device_queue_manager *dqm, bool lock); +static int destroy_queues_cpsch(struct device_queue_manager *dqm, bool lock); + + +static inline unsigned int get_pipes_num(struct device_queue_manager *dqm) +{ + BUG_ON(!dqm || !dqm->dev); + return dqm->dev->shared_resources.compute_pipe_count; +} + +static inline unsigned int get_first_pipe(struct device_queue_manager *dqm) +{ + BUG_ON(!dqm); + return dqm->dev->shared_resources.first_compute_pipe; +} + +static inline unsigned int get_pipes_num_cpsch(void) +{ + return PIPE_PER_ME_CP_SCHEDULING; +} + +static unsigned int get_sh_mem_bases_nybble_64(struct kfd_process *process, + struct kfd_dev *dev) +{ + struct kfd_process_device *pdd; + uint32_t nybble; + + pdd = kfd_get_process_device_data(dev, process, 1); + nybble = (pdd->lds_base >> 60) & 0x0E; + + return nybble; + +} + +static unsigned int get_sh_mem_bases_32(struct kfd_process *process, + struct kfd_dev *dev) +{ + struct kfd_process_device *pdd; + unsigned int shared_base; + + pdd = kfd_get_process_device_data(dev, process, 1); + shared_base = (pdd->lds_base >> 16) & 0xFF; + + return shared_base; +} + +static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble); +static void init_process_memory(struct device_queue_manager *dqm, + struct qcm_process_device *qpd) +{ + unsigned int temp; + + BUG_ON(!dqm || !qpd); + + /* check if sh_mem_config register already configured */ + if (qpd->sh_mem_config == 0) { + qpd->sh_mem_config = + ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED) | + DEFAULT_MTYPE(MTYPE_NONCACHED) | + APE1_MTYPE(MTYPE_NONCACHED); + qpd->sh_mem_ape1_limit = 0; + qpd->sh_mem_ape1_base = 0; + } + + if (qpd->pqm->process->is_32bit_user_mode) { + temp = get_sh_mem_bases_32(qpd->pqm->process, dqm->dev); + qpd->sh_mem_bases = SHARED_BASE(temp); + qpd->sh_mem_config |= PTR32; + } else { + temp = get_sh_mem_bases_nybble_64(qpd->pqm->process, dqm->dev); + qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp); + } + + pr_debug("kfd: is32bit process: %d sh_mem_bases nybble: 0x%X and register 0x%X\n", + qpd->pqm->process->is_32bit_user_mode, temp, qpd->sh_mem_bases); +} + +static void program_sh_mem_settings(struct device_queue_manager *dqm, + struct qcm_process_device *qpd) +{ + return kfd2kgd->program_sh_mem_settings(dqm->dev->kgd, qpd->vmid, + qpd->sh_mem_config, + qpd->sh_mem_ape1_base, + qpd->sh_mem_ape1_limit, + qpd->sh_mem_bases); +} + +static int allocate_vmid(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + struct queue *q) +{ + int bit, allocated_vmid; + + if (dqm->vmid_bitmap == 0) + return -ENOMEM; + + bit = find_first_bit((unsigned long *)&dqm->vmid_bitmap, CIK_VMID_NUM); + clear_bit(bit, (unsigned long *)&dqm->vmid_bitmap); + + /* Kaveri kfd vmid's starts from vmid 8 */ + allocated_vmid = bit + KFD_VMID_START_OFFSET; + pr_debug("kfd: vmid allocation %d\n", allocated_vmid); + qpd->vmid = allocated_vmid; + q->properties.vmid = allocated_vmid; + + set_pasid_vmid_mapping(dqm, q->process->pasid, q->properties.vmid); + program_sh_mem_settings(dqm, qpd); + + return 0; +} + +static void deallocate_vmid(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + struct queue *q) +{ + int bit = qpd->vmid - KFD_VMID_START_OFFSET; + + set_bit(bit, (unsigned long *)&dqm->vmid_bitmap); + qpd->vmid = 0; + q->properties.vmid = 0; +} + +static int create_queue_nocpsch(struct device_queue_manager *dqm, + struct queue *q, + struct qcm_process_device *qpd, + int *allocated_vmid) +{ + int retval; + + BUG_ON(!dqm || !q || !qpd || !allocated_vmid); + + pr_debug("kfd: In func %s\n", __func__); + print_queue(q); + + mutex_lock(&dqm->lock); + + if (list_empty(&qpd->queues_list)) { + retval = allocate_vmid(dqm, qpd, q); + if (retval != 0) { + mutex_unlock(&dqm->lock); + return retval; + } + } + *allocated_vmid = qpd->vmid; + q->properties.vmid = qpd->vmid; + + retval = create_compute_queue_nocpsch(dqm, q, qpd); + + if (retval != 0) { + if (list_empty(&qpd->queues_list)) { + deallocate_vmid(dqm, qpd, q); + *allocated_vmid = 0; + } + mutex_unlock(&dqm->lock); + return retval; + } + + list_add(&q->list, &qpd->queues_list); + dqm->queue_count++; + + mutex_unlock(&dqm->lock); + return 0; +} + +static int allocate_hqd(struct device_queue_manager *dqm, struct queue *q) +{ + bool set; + int pipe, bit; + + set = false; + + for (pipe = dqm->next_pipe_to_allocate; pipe < get_pipes_num(dqm); + pipe = (pipe + 1) % get_pipes_num(dqm)) { + if (dqm->allocated_queues[pipe] != 0) { + bit = find_first_bit( + (unsigned long *)&dqm->allocated_queues[pipe], + QUEUES_PER_PIPE); + + clear_bit(bit, + (unsigned long *)&dqm->allocated_queues[pipe]); + q->pipe = pipe; + q->queue = bit; + set = true; + break; + } + } + + if (set == false) + return -EBUSY; + + pr_debug("kfd: DQM %s hqd slot - pipe (%d) queue(%d)\n", + __func__, q->pipe, q->queue); + /* horizontal hqd allocation */ + dqm->next_pipe_to_allocate = (pipe + 1) % get_pipes_num(dqm); + + return 0; +} + +static inline void deallocate_hqd(struct device_queue_manager *dqm, + struct queue *q) +{ + set_bit(q->queue, (unsigned long *)&dqm->allocated_queues[q->pipe]); +} + +static int create_compute_queue_nocpsch(struct device_queue_manager *dqm, + struct queue *q, + struct qcm_process_device *qpd) +{ + int retval; + struct mqd_manager *mqd; + + BUG_ON(!dqm || !q || !qpd); + + mqd = dqm->get_mqd_manager(dqm, KFD_MQD_TYPE_CIK_COMPUTE); + if (mqd == NULL) + return -ENOMEM; + + retval = allocate_hqd(dqm, q); + if (retval != 0) + return retval; + + retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj, + &q->gart_mqd_addr, &q->properties); + if (retval != 0) { + deallocate_hqd(dqm, q); + return retval; + } + + return 0; +} + +static int destroy_queue_nocpsch(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + struct queue *q) +{ + int retval; + struct mqd_manager *mqd; + + BUG_ON(!dqm || !q || !q->mqd || !qpd); + + retval = 0; + + pr_debug("kfd: In Func %s\n", __func__); + + mutex_lock(&dqm->lock); + mqd = dqm->get_mqd_manager(dqm, KFD_MQD_TYPE_CIK_COMPUTE); + if (mqd == NULL) { + retval = -ENOMEM; + goto out; + } + + retval = mqd->destroy_mqd(mqd, q->mqd, + KFD_PREEMPT_TYPE_WAVEFRONT, + QUEUE_PREEMPT_DEFAULT_TIMEOUT_MS, + q->pipe, q->queue); + + if (retval != 0) + goto out; + + deallocate_hqd(dqm, q); + + mqd->uninit_mqd(mqd, q->mqd, q->mqd_mem_obj); + + list_del(&q->list); + if (list_empty(&qpd->queues_list)) + deallocate_vmid(dqm, qpd, q); + dqm->queue_count--; +out: + mutex_unlock(&dqm->lock); + return retval; +} + +static int update_queue(struct device_queue_manager *dqm, struct queue *q) +{ + int retval; + struct mqd_manager *mqd; + + BUG_ON(!dqm || !q || !q->mqd); + + mutex_lock(&dqm->lock); + mqd = dqm->get_mqd_manager(dqm, KFD_MQD_TYPE_CIK_COMPUTE); + if (mqd == NULL) { + mutex_unlock(&dqm->lock); + return -ENOMEM; + } + + retval = mqd->update_mqd(mqd, q->mqd, &q->properties); + if (q->properties.is_active == true) + dqm->queue_count++; + else + dqm->queue_count--; + + if (sched_policy != KFD_SCHED_POLICY_NO_HWS) + retval = execute_queues_cpsch(dqm, false); + + mutex_unlock(&dqm->lock); + return retval; +} + +static struct mqd_manager *get_mqd_manager_nocpsch( + struct device_queue_manager *dqm, enum KFD_MQD_TYPE type) +{ + struct mqd_manager *mqd; + + BUG_ON(!dqm || type >= KFD_MQD_TYPE_MAX); + + pr_debug("kfd: In func %s mqd type %d\n", __func__, type); + + mqd = dqm->mqds[type]; + if (!mqd) { + mqd = mqd_manager_init(type, dqm->dev); + if (mqd == NULL) + pr_err("kfd: mqd manager is NULL"); + dqm->mqds[type] = mqd; + } + + return mqd; +} + +static int register_process_nocpsch(struct device_queue_manager *dqm, + struct qcm_process_device *qpd) +{ + struct device_process_node *n; + + BUG_ON(!dqm || !qpd); + + pr_debug("kfd: In func %s\n", __func__); + + n = kzalloc(sizeof(struct device_process_node), GFP_KERNEL); + if (!n) + return -ENOMEM; + + n->qpd = qpd; + + mutex_lock(&dqm->lock); + list_add(&n->list, &dqm->queues); + + init_process_memory(dqm, qpd); + dqm->processes_count++; + + mutex_unlock(&dqm->lock); + + return 0; +} + +static int unregister_process_nocpsch(struct device_queue_manager *dqm, + struct qcm_process_device *qpd) +{ + int retval; + struct device_process_node *cur, *next; + + BUG_ON(!dqm || !qpd); + + BUG_ON(!list_empty(&qpd->queues_list)); + + pr_debug("kfd: In func %s\n", __func__); + + retval = 0; + mutex_lock(&dqm->lock); + + list_for_each_entry_safe(cur, next, &dqm->queues, list) { + if (qpd == cur->qpd) { + list_del(&cur->list); + dqm->processes_count--; + goto out; + } + } + /* qpd not found in dqm list */ + retval = 1; +out: + mutex_unlock(&dqm->lock); + return retval; +} + +static int +set_pasid_vmid_mapping(struct device_queue_manager *dqm, unsigned int pasid, + unsigned int vmid) +{ + uint32_t pasid_mapping; + + pasid_mapping = (pasid == 0) ? 0 : (uint32_t)pasid | + ATC_VMID_PASID_MAPPING_VALID; + return kfd2kgd->set_pasid_vmid_mapping(dqm->dev->kgd, pasid_mapping, + vmid); +} + +static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble) +{ + /* In 64-bit mode, we can only control the top 3 bits of the LDS, + * scratch and GPUVM apertures. + * The hardware fills in the remaining 59 bits according to the + * following pattern: + * LDS: X0000000'00000000 - X0000001'00000000 (4GB) + * Scratch: X0000001'00000000 - X0000002'00000000 (4GB) + * GPUVM: Y0010000'00000000 - Y0020000'00000000 (1TB) + * + * (where X/Y is the configurable nybble with the low-bit 0) + * + * LDS and scratch will have the same top nybble programmed in the + * top 3 bits of SH_MEM_BASES.PRIVATE_BASE. + * GPUVM can have a different top nybble programmed in the + * top 3 bits of SH_MEM_BASES.SHARED_BASE. + * We don't bother to support different top nybbles + * for LDS/Scratch and GPUVM. + */ + + BUG_ON((top_address_nybble & 1) || top_address_nybble > 0xE || + top_address_nybble == 0); + + return PRIVATE_BASE(top_address_nybble << 12) | + SHARED_BASE(top_address_nybble << 12); +} + +static int init_memory(struct device_queue_manager *dqm) +{ + int i, retval; + + for (i = 8; i < 16; i++) + set_pasid_vmid_mapping(dqm, 0, i); + + retval = kfd2kgd->init_memory(dqm->dev->kgd); + if (retval == 0) + is_mem_initialized = true; + return retval; +} + + +static int init_pipelines(struct device_queue_manager *dqm, + unsigned int pipes_num, unsigned int first_pipe) +{ + void *hpdptr; + struct mqd_manager *mqd; + unsigned int i, err, inx; + uint64_t pipe_hpd_addr; + + BUG_ON(!dqm || !dqm->dev); + + pr_debug("kfd: In func %s\n", __func__); + + /* + * Allocate memory for the HPDs. This is hardware-owned per-pipe data. + * The driver never accesses this memory after zeroing it. + * It doesn't even have to be saved/restored on suspend/resume + * because it contains no data when there are no active queues. + */ + + err = kfd2kgd->allocate_mem(dqm->dev->kgd, + CIK_HPD_EOP_BYTES * pipes_num, + PAGE_SIZE, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) &dqm->pipeline_mem); + + if (err) { + pr_err("kfd: error allocate vidmem num pipes: %d\n", + pipes_num); + return -ENOMEM; + } + + hpdptr = dqm->pipeline_mem->cpu_ptr; + dqm->pipelines_addr = dqm->pipeline_mem->gpu_addr; + + memset(hpdptr, 0, CIK_HPD_EOP_BYTES * pipes_num); + + mqd = dqm->get_mqd_manager(dqm, KFD_MQD_TYPE_CIK_COMPUTE); + if (mqd == NULL) { + kfd2kgd->free_mem(dqm->dev->kgd, + (struct kgd_mem *) dqm->pipeline_mem); + return -ENOMEM; + } + + for (i = 0; i < pipes_num; i++) { + inx = i + first_pipe; + pipe_hpd_addr = dqm->pipelines_addr + i * CIK_HPD_EOP_BYTES; + pr_debug("kfd: pipeline address %llX\n", pipe_hpd_addr); + /* = log2(bytes/4)-1 */ + kfd2kgd->init_pipeline(dqm->dev->kgd, i, + CIK_HPD_EOP_BYTES_LOG2 - 3, pipe_hpd_addr); + } + + return 0; +} + + +static int init_scheduler(struct device_queue_manager *dqm) +{ + int retval; + + BUG_ON(!dqm); + + pr_debug("kfd: In %s\n", __func__); + + retval = init_pipelines(dqm, get_pipes_num(dqm), KFD_DQM_FIRST_PIPE); + if (retval != 0) + return retval; + + retval = init_memory(dqm); + + return retval; +} + +static int initialize_nocpsch(struct device_queue_manager *dqm) +{ + int i; + + BUG_ON(!dqm); + + pr_debug("kfd: In func %s num of pipes: %d\n", + __func__, get_pipes_num(dqm)); + + mutex_init(&dqm->lock); + INIT_LIST_HEAD(&dqm->queues); + dqm->queue_count = dqm->next_pipe_to_allocate = 0; + dqm->allocated_queues = kcalloc(get_pipes_num(dqm), + sizeof(unsigned int), GFP_KERNEL); + if (!dqm->allocated_queues) { + mutex_destroy(&dqm->lock); + return -ENOMEM; + } + + for (i = 0; i < get_pipes_num(dqm); i++) + dqm->allocated_queues[i] = (1 << QUEUES_PER_PIPE) - 1; + + dqm->vmid_bitmap = (1 << VMID_PER_DEVICE) - 1; + + init_scheduler(dqm); + return 0; +} + +static void uninitialize_nocpsch(struct device_queue_manager *dqm) +{ + BUG_ON(!dqm); + + BUG_ON(dqm->queue_count > 0 || dqm->processes_count > 0); + + kfree(dqm->allocated_queues); + mutex_destroy(&dqm->lock); + kfd2kgd->free_mem(dqm->dev->kgd, + (struct kgd_mem *) dqm->pipeline_mem); +} + +static int start_nocpsch(struct device_queue_manager *dqm) +{ + return 0; +} + +static int stop_nocpsch(struct device_queue_manager *dqm) +{ + return 0; +} + +/* + * Device Queue Manager implementation for cp scheduler + */ + +static int set_sched_resources(struct device_queue_manager *dqm) +{ + struct scheduling_resources res; + unsigned int queue_num, queue_mask; + + BUG_ON(!dqm); + + pr_debug("kfd: In func %s\n", __func__); + + queue_num = get_pipes_num_cpsch() * QUEUES_PER_PIPE; + queue_mask = (1 << queue_num) - 1; + res.vmid_mask = (1 << VMID_PER_DEVICE) - 1; + res.vmid_mask <<= KFD_VMID_START_OFFSET; + res.queue_mask = queue_mask << (get_first_pipe(dqm) * QUEUES_PER_PIPE); + res.gws_mask = res.oac_mask = res.gds_heap_base = + res.gds_heap_size = 0; + + pr_debug("kfd: scheduling resources:\n" + " vmid mask: 0x%8X\n" + " queue mask: 0x%8llX\n", + res.vmid_mask, res.queue_mask); + + return pm_send_set_resources(&dqm->packets, &res); +} + +static int initialize_cpsch(struct device_queue_manager *dqm) +{ + int retval; + + BUG_ON(!dqm); + + pr_debug("kfd: In func %s num of pipes: %d\n", + __func__, get_pipes_num_cpsch()); + + mutex_init(&dqm->lock); + INIT_LIST_HEAD(&dqm->queues); + dqm->queue_count = dqm->processes_count = 0; + dqm->active_runlist = false; + retval = init_pipelines(dqm, get_pipes_num(dqm), 0); + if (retval != 0) + goto fail_init_pipelines; + + return 0; + +fail_init_pipelines: + mutex_destroy(&dqm->lock); + return retval; +} + +static int start_cpsch(struct device_queue_manager *dqm) +{ + struct device_process_node *node; + int retval; + + BUG_ON(!dqm); + + retval = 0; + + retval = pm_init(&dqm->packets, dqm); + if (retval != 0) + goto fail_packet_manager_init; + + retval = set_sched_resources(dqm); + if (retval != 0) + goto fail_set_sched_resources; + + pr_debug("kfd: allocating fence memory\n"); + + /* allocate fence memory on the gart */ + retval = kfd2kgd->allocate_mem(dqm->dev->kgd, + sizeof(*dqm->fence_addr), + 32, + KFD_MEMPOOL_SYSTEM_WRITECOMBINE, + (struct kgd_mem **) &dqm->fence_mem); + + if (retval != 0) + goto fail_allocate_vidmem; + + dqm->fence_addr = dqm->fence_mem->cpu_ptr; + dqm->fence_gpu_addr = dqm->fence_mem->gpu_addr; + + list_for_each_entry(node, &dqm->queues, list) + if (node->qpd->pqm->process && dqm->dev) + kfd_bind_process_to_device(dqm->dev, + node->qpd->pqm->process); + + execute_queues_cpsch(dqm, true); + + return 0; +fail_allocate_vidmem: +fail_set_sched_resources: + pm_uninit(&dqm->packets); +fail_packet_manager_init: + return retval; +} + +static int stop_cpsch(struct device_queue_manager *dqm) +{ + struct device_process_node *node; + struct kfd_process_device *pdd; + + BUG_ON(!dqm); + + destroy_queues_cpsch(dqm, true); + + list_for_each_entry(node, &dqm->queues, list) { + pdd = kfd_get_process_device_data(dqm->dev, + node->qpd->pqm->process, 1); + pdd->bound = false; + } + kfd2kgd->free_mem(dqm->dev->kgd, + (struct kgd_mem *) dqm->fence_mem); + pm_uninit(&dqm->packets); + + return 0; +} + +static int create_kernel_queue_cpsch(struct device_queue_manager *dqm, + struct kernel_queue *kq, + struct qcm_process_device *qpd) +{ + BUG_ON(!dqm || !kq || !qpd); + + pr_debug("kfd: In func %s\n", __func__); + + mutex_lock(&dqm->lock); + list_add(&kq->list, &qpd->priv_queue_list); + dqm->queue_count++; + qpd->is_debug = true; + execute_queues_cpsch(dqm, false); + mutex_unlock(&dqm->lock); + + return 0; +} + +static void destroy_kernel_queue_cpsch(struct device_queue_manager *dqm, + struct kernel_queue *kq, + struct qcm_process_device *qpd) +{ + BUG_ON(!dqm || !kq); + + pr_debug("kfd: In %s\n", __func__); + + mutex_lock(&dqm->lock); + destroy_queues_cpsch(dqm, false); + list_del(&kq->list); + dqm->queue_count--; + qpd->is_debug = false; + execute_queues_cpsch(dqm, false); + mutex_unlock(&dqm->lock); +} + +static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, + struct qcm_process_device *qpd, int *allocate_vmid) +{ + int retval; + struct mqd_manager *mqd; + + BUG_ON(!dqm || !q || !qpd); + + retval = 0; + + if (allocate_vmid) + *allocate_vmid = 0; + + mutex_lock(&dqm->lock); + + mqd = dqm->get_mqd_manager(dqm, KFD_MQD_TYPE_CIK_CP); + if (mqd == NULL) { + mutex_unlock(&dqm->lock); + return -ENOMEM; + } + + retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj, + &q->gart_mqd_addr, &q->properties); + if (retval != 0) + goto out; + + list_add(&q->list, &qpd->queues_list); + if (q->properties.is_active) { + dqm->queue_count++; + retval = execute_queues_cpsch(dqm, false); + } + +out: + mutex_unlock(&dqm->lock); + return retval; +} + +int fence_wait_timeout(unsigned int *fence_addr, unsigned int fence_value, + unsigned long timeout) +{ + BUG_ON(!fence_addr); + timeout += jiffies; + + while (*fence_addr != fence_value) { + if (time_after(jiffies, timeout)) { + pr_err("kfd: qcm fence wait loop timeout expired\n"); + return -ETIME; + } + cpu_relax(); + } + + return 0; +} + +static int destroy_queues_cpsch(struct device_queue_manager *dqm, bool lock) +{ + int retval; + + BUG_ON(!dqm); + + retval = 0; + + if (lock) + mutex_lock(&dqm->lock); + if (dqm->active_runlist == false) + goto out; + retval = pm_send_unmap_queue(&dqm->packets, KFD_QUEUE_TYPE_COMPUTE, + KFD_PREEMPT_TYPE_FILTER_ALL_QUEUES, 0, false, 0); + if (retval != 0) + goto out; + + *dqm->fence_addr = KFD_FENCE_INIT; + pm_send_query_status(&dqm->packets, dqm->fence_gpu_addr, + KFD_FENCE_COMPLETED); + /* should be timed out */ + fence_wait_timeout(dqm->fence_addr, KFD_FENCE_COMPLETED, + QUEUE_PREEMPT_DEFAULT_TIMEOUT_MS); + pm_release_ib(&dqm->packets); + dqm->active_runlist = false; + +out: + if (lock) + mutex_unlock(&dqm->lock); + return retval; +} + +static int execute_queues_cpsch(struct device_queue_manager *dqm, bool lock) +{ + int retval; + + BUG_ON(!dqm); + + if (lock) + mutex_lock(&dqm->lock); + + retval = destroy_queues_cpsch(dqm, false); + if (retval != 0) { + pr_err("kfd: the cp might be in an unrecoverable state due to an unsuccessful queues preemption"); + goto out; + } + + if (dqm->queue_count <= 0 || dqm->processes_count <= 0) { + retval = 0; + goto out; + } + + if (dqm->active_runlist) { + retval = 0; + goto out; + } + + retval = pm_send_runlist(&dqm->packets, &dqm->queues); + if (retval != 0) { + pr_err("kfd: failed to execute runlist"); + goto out; + } + dqm->active_runlist = true; + +out: + if (lock) + mutex_unlock(&dqm->lock); + return retval; +} + +static int destroy_queue_cpsch(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + struct queue *q) +{ + int retval; + struct mqd_manager *mqd; + + BUG_ON(!dqm || !qpd || !q); + + retval = 0; + + /* remove queue from list to prevent rescheduling after preemption */ + mutex_lock(&dqm->lock); + + mqd = dqm->get_mqd_manager(dqm, KFD_MQD_TYPE_CIK_CP); + if (!mqd) { + retval = -ENOMEM; + goto failed; + } + + list_del(&q->list); + dqm->queue_count--; + + execute_queues_cpsch(dqm, false); + + mqd->uninit_mqd(mqd, q->mqd, q->mqd_mem_obj); + + mutex_unlock(&dqm->lock); + + return 0; + +failed: + mutex_unlock(&dqm->lock); + return retval; +} + +/* + * Low bits must be 0000/FFFF as required by HW, high bits must be 0 to + * stay in user mode. + */ +#define APE1_FIXED_BITS_MASK 0xFFFF80000000FFFFULL +/* APE1 limit is inclusive and 64K aligned. */ +#define APE1_LIMIT_ALIGNMENT 0xFFFF + +static bool set_cache_memory_policy(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + enum cache_policy default_policy, + enum cache_policy alternate_policy, + void __user *alternate_aperture_base, + uint64_t alternate_aperture_size) +{ + uint32_t default_mtype; + uint32_t ape1_mtype; + + pr_debug("kfd: In func %s\n", __func__); + + mutex_lock(&dqm->lock); + + if (alternate_aperture_size == 0) { + /* base > limit disables APE1 */ + qpd->sh_mem_ape1_base = 1; + qpd->sh_mem_ape1_limit = 0; + } else { + /* + * In FSA64, APE1_Base[63:0] = { 16{SH_MEM_APE1_BASE[31]}, + * SH_MEM_APE1_BASE[31:0], 0x0000 } + * APE1_Limit[63:0] = { 16{SH_MEM_APE1_LIMIT[31]}, + * SH_MEM_APE1_LIMIT[31:0], 0xFFFF } + * Verify that the base and size parameters can be + * represented in this format and convert them. + * Additionally restrict APE1 to user-mode addresses. + */ + + uint64_t base = (uintptr_t)alternate_aperture_base; + uint64_t limit = base + alternate_aperture_size - 1; + + if (limit <= base) + goto out; + + if ((base & APE1_FIXED_BITS_MASK) != 0) + goto out; + + if ((limit & APE1_FIXED_BITS_MASK) != APE1_LIMIT_ALIGNMENT) + goto out; + + qpd->sh_mem_ape1_base = base >> 16; + qpd->sh_mem_ape1_limit = limit >> 16; + } + + default_mtype = (default_policy == cache_policy_coherent) ? + MTYPE_NONCACHED : + MTYPE_CACHED; + + ape1_mtype = (alternate_policy == cache_policy_coherent) ? + MTYPE_NONCACHED : + MTYPE_CACHED; + + qpd->sh_mem_config = (qpd->sh_mem_config & PTR32) + | ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED) + | DEFAULT_MTYPE(default_mtype) + | APE1_MTYPE(ape1_mtype); + + if ((sched_policy == KFD_SCHED_POLICY_NO_HWS) && (qpd->vmid != 0)) + program_sh_mem_settings(dqm, qpd); + + pr_debug("kfd: sh_mem_config: 0x%x, ape1_base: 0x%x, ape1_limit: 0x%x\n", + qpd->sh_mem_config, qpd->sh_mem_ape1_base, + qpd->sh_mem_ape1_limit); + + mutex_unlock(&dqm->lock); + return true; + +out: + mutex_unlock(&dqm->lock); + return false; +} + +struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev) +{ + struct device_queue_manager *dqm; + + BUG_ON(!dev); + + dqm = kzalloc(sizeof(struct device_queue_manager), GFP_KERNEL); + if (!dqm) + return NULL; + + dqm->dev = dev; + switch (sched_policy) { + case KFD_SCHED_POLICY_HWS: + case KFD_SCHED_POLICY_HWS_NO_OVERSUBSCRIPTION: + /* initialize dqm for cp scheduling */ + dqm->create_queue = create_queue_cpsch; + dqm->initialize = initialize_cpsch; + dqm->start = start_cpsch; + dqm->stop = stop_cpsch; + dqm->destroy_queue = destroy_queue_cpsch; + dqm->update_queue = update_queue; + dqm->get_mqd_manager = get_mqd_manager_nocpsch; + dqm->register_process = register_process_nocpsch; + dqm->unregister_process = unregister_process_nocpsch; + dqm->uninitialize = uninitialize_nocpsch; + dqm->create_kernel_queue = create_kernel_queue_cpsch; + dqm->destroy_kernel_queue = destroy_kernel_queue_cpsch; + dqm->set_cache_memory_policy = set_cache_memory_policy; + break; + case KFD_SCHED_POLICY_NO_HWS: + /* initialize dqm for no cp scheduling */ + dqm->start = start_nocpsch; + dqm->stop = stop_nocpsch; + dqm->create_queue = create_queue_nocpsch; + dqm->destroy_queue = destroy_queue_nocpsch; + dqm->update_queue = update_queue; + dqm->get_mqd_manager = get_mqd_manager_nocpsch; + dqm->register_process = register_process_nocpsch; + dqm->unregister_process = unregister_process_nocpsch; + dqm->initialize = initialize_nocpsch; + dqm->uninitialize = uninitialize_nocpsch; + dqm->set_cache_memory_policy = set_cache_memory_policy; + break; + default: + BUG(); + break; + } + + if (dqm->initialize(dqm) != 0) { + kfree(dqm); + return NULL; + } + + return dqm; +} + +void device_queue_manager_uninit(struct device_queue_manager *dqm) +{ + BUG_ON(!dqm); + + dqm->uninitialize(dqm); + kfree(dqm); +} + diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h index e495b38..c3f189e8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h @@ -42,6 +42,50 @@ struct device_process_node { struct list_head list; }; +/** + * struct device_queue_manager + * + * @create_queue: Queue creation routine. + * + * @destroy_queue: Queue destruction routine. + * + * @update_queue: Queue update routine. + * + * @get_mqd_manager: Returns the mqd manager according to the mqd type. + * + * @exeute_queues: Dispatches the queues list to the H/W. + * + * @register_process: This routine associates a specific process with device. + * + * @unregister_process: destroys the associations between process to device. + * + * @initialize: Initializes the pipelines and memory module for that device. + * + * @start: Initializes the resources/modules the the device needs for queues + * execution. This function is called on device initialization and after the + * system woke up after suspension. + * + * @stop: This routine stops execution of all the active queue running on the + * H/W and basically this function called on system suspend. + * + * @uninitialize: Destroys all the device queue manager resources allocated in + * initialize routine. + * + * @create_kernel_queue: Creates kernel queue. Used for debug queue. + * + * @destroy_kernel_queue: Destroys kernel queue. Used for debug queue. + * + * @set_cache_memory_policy: Sets memory policy (cached/ non cached) for the + * memory apertures. + * + * This struct is a base class for the kfd queues scheduler in the + * device level. The device base class should expose the basic operations + * for queue creation and queue destruction. This base class hides the + * scheduling mode of the driver and the specific implementation of the + * concrete device. This class is the only class in the queues scheduler + * that configures the H/W. + */ + struct device_queue_manager { int (*create_queue)(struct device_queue_manager *dqm, struct queue *q, @@ -52,8 +96,9 @@ struct device_queue_manager { struct queue *q); int (*update_queue)(struct device_queue_manager *dqm, struct queue *q); - struct mqd_manager * (*get_mqd_manager)( - struct device_queue_manager *dqm, + + struct mqd_manager * (*get_mqd_manager) + (struct device_queue_manager *dqm, enum KFD_MQD_TYPE type); int (*register_process)(struct device_queue_manager *dqm, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 72b2e27..7f3b4fd 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -454,6 +454,8 @@ void kfd_process_destroy_wq(void); struct kfd_process *kfd_create_process(const struct task_struct *); struct kfd_process *kfd_get_process(const struct task_struct *); +struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, + struct kfd_process *p); void kfd_unbind_process_from_device(struct kfd_dev *dev, unsigned int pasid); struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, struct kfd_process *p, @@ -509,6 +511,10 @@ void uninit_queue(struct queue *q); void print_queue_properties(struct queue_properties *q); void print_queue(struct queue *q); +struct mqd_manager *mqd_manager_init(enum KFD_MQD_TYPE type, + struct kfd_dev *dev); +struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev); +void device_queue_manager_uninit(struct device_queue_manager *dqm); struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, enum kfd_queue_type type); void kernel_queue_uninit(struct kernel_queue *kq); @@ -537,6 +543,8 @@ int pqm_update_queue(struct process_queue_manager *pqm, unsigned int qid, #define KFD_HIQ_TIMEOUT (500) +#define KFD_FENCE_COMPLETED (100) +#define KFD_FENCE_INIT (10) #define KFD_UNMAP_LATENCY (150) struct packet_manager { @@ -547,6 +555,19 @@ struct packet_manager { struct kfd_mem_obj *ib_buffer_obj; }; +int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm); +void pm_uninit(struct packet_manager *pm); +int pm_send_set_resources(struct packet_manager *pm, + struct scheduling_resources *res); +int pm_send_runlist(struct packet_manager *pm, struct list_head *dqm_queues); +int pm_send_query_status(struct packet_manager *pm, uint64_t fence_address, + uint32_t fence_value); + +int pm_send_unmap_queue(struct packet_manager *pm, enum kfd_queue_type type, + enum kfd_preempt_type_filter mode, + uint32_t filter_param, bool reset, + unsigned int sdma_engine); + void pm_release_ib(struct packet_manager *pm); uint64_t kfd_get_number_elems(struct kfd_dev *kfd); -- cgit v0.10.2 From b3f5e6b4412e69ebd7094c51aa345188a30a5cbe Mon Sep 17 00:00:00 2001 From: Andrew Lewycky Date: Thu, 17 Jul 2014 01:37:30 +0300 Subject: amdkfd: Add interrupt handling module This patch adds the interrupt handling module, in kfd_interrupt.c, and its related members in different data structures to the amdkfd driver. The amdkfd interrupt module maintains an internal interrupt ring per amdkfd device. The internal interrupt ring contains interrupts that needs further handling. The extra handling is deferred to a later time through a workqueue. There's no acknowledgment for the interrupts we use. The hardware simply queues a new interrupt each time without waiting. The fixed-size internal queue means that it's possible for us to lose interrupts because we have no back-pressure to the hardware. v3: Move amdkfd from drm/radeon/ to drm/amd/ Change device init Made sure spin lock is taken only if init is complete Moved bool field to the end of the structure Signed-off-by: Andrew Lewycky Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 307a309..be6246d 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -8,6 +8,7 @@ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ kfd_process.o kfd_queue.o kfd_mqd_manager.o \ kfd_kernel_queue.o kfd_packet_manager.o \ - kfd_process_queue_manager.o kfd_device_queue_manager.o + kfd_process_queue_manager.o kfd_device_queue_manager.o \ + kfd_interrupt.o obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index bdf3e94..9beb6f7 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -31,6 +31,7 @@ static const struct kfd_device_info kaveri_device_info = { .max_pasid_bits = 16, + .ih_ring_entry_size = 4 * sizeof(uint32_t), .mqd_size_aligned = MQD_SIZE_ALIGNED }; @@ -191,6 +192,13 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, goto kfd_topology_add_device_error; } + if (kfd_interrupt_init(kfd)) { + dev_err(kfd_device, + "Error initializing interrupts for device (%x:%x)\n", + kfd->pdev->vendor, kfd->pdev->device); + goto kfd_interrupt_error; + } + if (!device_iommu_pasid_init(kfd)) { dev_err(kfd_device, "Error initializing iommuv2 for device (%x:%x)\n", @@ -229,6 +237,8 @@ dqm_start_error: device_queue_manager_error: amd_iommu_free_device(kfd->pdev); device_iommu_pasid_error: + kfd_interrupt_exit(kfd); +kfd_interrupt_error: kfd_topology_remove_device(kfd); kfd_topology_add_device_error: kfd2kgd->fini_sa_manager(kfd->kgd); @@ -244,6 +254,7 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd) if (kfd->init_complete) { device_queue_manager_uninit(kfd->dqm); amd_iommu_free_device(kfd->pdev); + kfd_interrupt_exit(kfd); kfd_topology_remove_device(kfd); } @@ -281,6 +292,16 @@ int kgd2kfd_resume(struct kfd_dev *kfd) return 0; } -void kgd2kfd_interrupt(struct kfd_dev *dev, const void *ih_ring_entry) +/* This is called directly from KGD at ISR. */ +void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry) { + if (kfd->init_complete) { + spin_lock(&kfd->interrupt_lock); + + if (kfd->interrupts_active + && enqueue_ih_ring_entry(kfd, ih_ring_entry)) + schedule_work(&kfd->interrupt_work); + + spin_unlock(&kfd->interrupt_lock); + } } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c new file mode 100644 index 0000000..5b99909 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c @@ -0,0 +1,176 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * KFD Interrupts. + * + * AMD GPUs deliver interrupts by pushing an interrupt description onto the + * interrupt ring and then sending an interrupt. KGD receives the interrupt + * in ISR and sends us a pointer to each new entry on the interrupt ring. + * + * We generally can't process interrupt-signaled events from ISR, so we call + * out to each interrupt client module (currently only the scheduler) to ask if + * each interrupt is interesting. If they return true, then it requires further + * processing so we copy it to an internal interrupt ring and call each + * interrupt client again from a work-queue. + * + * There's no acknowledgment for the interrupts we use. The hardware simply + * queues a new interrupt each time without waiting. + * + * The fixed-size internal queue means that it's possible for us to lose + * interrupts because we have no back-pressure to the hardware. + */ + +#include +#include +#include "kfd_priv.h" + +#define KFD_INTERRUPT_RING_SIZE 256 + +static void interrupt_wq(struct work_struct *); + +int kfd_interrupt_init(struct kfd_dev *kfd) +{ + void *interrupt_ring = kmalloc_array(KFD_INTERRUPT_RING_SIZE, + kfd->device_info->ih_ring_entry_size, + GFP_KERNEL); + if (!interrupt_ring) + return -ENOMEM; + + kfd->interrupt_ring = interrupt_ring; + kfd->interrupt_ring_size = + KFD_INTERRUPT_RING_SIZE * kfd->device_info->ih_ring_entry_size; + atomic_set(&kfd->interrupt_ring_wptr, 0); + atomic_set(&kfd->interrupt_ring_rptr, 0); + + spin_lock_init(&kfd->interrupt_lock); + + INIT_WORK(&kfd->interrupt_work, interrupt_wq); + + kfd->interrupts_active = true; + + /* + * After this function returns, the interrupt will be enabled. This + * barrier ensures that the interrupt running on a different processor + * sees all the above writes. + */ + smp_wmb(); + + return 0; +} + +void kfd_interrupt_exit(struct kfd_dev *kfd) +{ + /* + * Stop the interrupt handler from writing to the ring and scheduling + * workqueue items. The spinlock ensures that any interrupt running + * after we have unlocked sees interrupts_active = false. + */ + unsigned long flags; + + spin_lock_irqsave(&kfd->interrupt_lock, flags); + kfd->interrupts_active = false; + spin_unlock_irqrestore(&kfd->interrupt_lock, flags); + + /* + * Flush_scheduled_work ensures that there are no outstanding + * work-queue items that will access interrupt_ring. New work items + * can't be created because we stopped interrupt handling above. + */ + flush_scheduled_work(); + + kfree(kfd->interrupt_ring); +} + +/* + * This assumes that it can't be called concurrently with itself + * but only with dequeue_ih_ring_entry. + */ +bool enqueue_ih_ring_entry(struct kfd_dev *kfd, const void *ih_ring_entry) +{ + unsigned int rptr = atomic_read(&kfd->interrupt_ring_rptr); + unsigned int wptr = atomic_read(&kfd->interrupt_ring_wptr); + + if ((rptr - wptr) % kfd->interrupt_ring_size == + kfd->device_info->ih_ring_entry_size) { + /* This is very bad, the system is likely to hang. */ + dev_err_ratelimited(kfd_chardev(), + "Interrupt ring overflow, dropping interrupt.\n"); + return false; + } + + memcpy(kfd->interrupt_ring + wptr, ih_ring_entry, + kfd->device_info->ih_ring_entry_size); + + wptr = (wptr + kfd->device_info->ih_ring_entry_size) % + kfd->interrupt_ring_size; + smp_wmb(); /* Ensure memcpy'd data is visible before wptr update. */ + atomic_set(&kfd->interrupt_ring_wptr, wptr); + + return true; +} + +/* + * This assumes that it can't be called concurrently with itself + * but only with enqueue_ih_ring_entry. + */ +static bool dequeue_ih_ring_entry(struct kfd_dev *kfd, void *ih_ring_entry) +{ + /* + * Assume that wait queues have an implicit barrier, i.e. anything that + * happened in the ISR before it queued work is visible. + */ + + unsigned int wptr = atomic_read(&kfd->interrupt_ring_wptr); + unsigned int rptr = atomic_read(&kfd->interrupt_ring_rptr); + + if (rptr == wptr) + return false; + + memcpy(ih_ring_entry, kfd->interrupt_ring + rptr, + kfd->device_info->ih_ring_entry_size); + + rptr = (rptr + kfd->device_info->ih_ring_entry_size) % + kfd->interrupt_ring_size; + + /* + * Ensure the rptr write update is not visible until + * memcpy has finished reading. + */ + smp_mb(); + atomic_set(&kfd->interrupt_ring_rptr, rptr); + + return true; +} + +static void interrupt_wq(struct work_struct *work) +{ + struct kfd_dev *dev = container_of(work, struct kfd_dev, + interrupt_work); + + uint32_t ih_ring_entry[DIV_ROUND_UP( + dev->device_info->ih_ring_entry_size, + sizeof(uint32_t))]; + + while (dequeue_ih_ring_entry(dev, ih_ring_entry)) + ; +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 7f3b4fd..2bc34aa1 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -135,10 +135,22 @@ struct kfd_dev { struct kgd2kfd_shared_resources shared_resources; + void *interrupt_ring; + size_t interrupt_ring_size; + atomic_t interrupt_ring_rptr; + atomic_t interrupt_ring_wptr; + struct work_struct interrupt_work; + spinlock_t interrupt_lock; + /* QCM Device instance */ struct device_queue_manager *dqm; bool init_complete; + /* + * Interrupts of interest to KFD are copied + * from the HW ring into a SW ring. + */ + bool interrupts_active; }; /* KGD2KFD callbacks */ @@ -493,11 +505,14 @@ struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev); struct kfd_dev *kfd_topology_enum_kfd_devices(uint8_t idx); /* Interrupts */ -void kgd2kfd_interrupt(struct kfd_dev *dev, const void *ih_ring_entry); +int kfd_interrupt_init(struct kfd_dev *dev); +void kfd_interrupt_exit(struct kfd_dev *dev); +void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry); +bool enqueue_ih_ring_entry(struct kfd_dev *kfd, const void *ih_ring_entry); /* Power Management */ -void kgd2kfd_suspend(struct kfd_dev *dev); -int kgd2kfd_resume(struct kfd_dev *dev); +void kgd2kfd_suspend(struct kfd_dev *kfd); +int kgd2kfd_resume(struct kfd_dev *kfd); /* amdkfd Apertures */ int kfd_init_apertures(struct kfd_process *process); -- cgit v0.10.2 From 39b027d957d4a9666a815ccf5bdce82e7796b2c6 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Sun, 19 Oct 2014 23:46:40 +0300 Subject: amdkfd: Implement the create/destroy/update queue IOCTLs v3: Removed the use of internal typedefs, fixed debug prints, added checks for parameters and moved to using doorbell address from user v4: Extracted some of the code in the create queue ioctl to a different function that may be also called from other ioctls in the future. Also fixed the check of the ring size argument. v5: Add support for AQL queues creation to enable working with open-source HSA runtime Signed-off-by: Ben Goz Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 58441cd..d342035 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -125,22 +125,221 @@ static long kfd_ioctl_get_version(struct file *filep, struct kfd_process *p, return -ENODEV; } +static int set_queue_properties_from_user(struct queue_properties *q_properties, + struct kfd_ioctl_create_queue_args *args) +{ + if (args->queue_percentage > KFD_MAX_QUEUE_PERCENTAGE) { + pr_err("kfd: queue percentage must be between 0 to KFD_MAX_QUEUE_PERCENTAGE\n"); + return -EINVAL; + } + + if (args->queue_priority > KFD_MAX_QUEUE_PRIORITY) { + pr_err("kfd: queue priority must be between 0 to KFD_MAX_QUEUE_PRIORITY\n"); + return -EINVAL; + } + + if ((args->ring_base_address) && + (!access_ok(VERIFY_WRITE, args->ring_base_address, sizeof(uint64_t)))) { + pr_err("kfd: can't access ring base address\n"); + return -EFAULT; + } + + if (!is_power_of_2(args->ring_size) && (args->ring_size != 0)) { + pr_err("kfd: ring size must be a power of 2 or 0\n"); + return -EINVAL; + } + + if (!access_ok(VERIFY_WRITE, args->read_pointer_address, sizeof(uint32_t))) { + pr_err("kfd: can't access read pointer\n"); + return -EFAULT; + } + + if (!access_ok(VERIFY_WRITE, args->write_pointer_address, sizeof(uint32_t))) { + pr_err("kfd: can't access write pointer\n"); + return -EFAULT; + } + + q_properties->is_interop = false; + q_properties->queue_percent = args->queue_percentage; + q_properties->priority = args->queue_priority; + q_properties->queue_address = args->ring_base_address; + q_properties->queue_size = args->ring_size; + q_properties->read_ptr = (uint32_t *) args->read_pointer_address; + q_properties->write_ptr = (uint32_t *) args->write_pointer_address; + if (args->queue_type == KFD_IOC_QUEUE_TYPE_COMPUTE || + args->queue_type == KFD_IOC_QUEUE_TYPE_COMPUTE_AQL) + q_properties->type = KFD_QUEUE_TYPE_COMPUTE; + else + return -ENOTSUPP; + + if (args->queue_type == KFD_IOC_QUEUE_TYPE_COMPUTE_AQL) + q_properties->format = KFD_QUEUE_FORMAT_AQL; + else + q_properties->format = KFD_QUEUE_FORMAT_PM4; + + pr_debug("Queue Percentage (%d, %d)\n", + q_properties->queue_percent, args->queue_percentage); + + pr_debug("Queue Priority (%d, %d)\n", + q_properties->priority, args->queue_priority); + + pr_debug("Queue Address (0x%llX, 0x%llX)\n", + q_properties->queue_address, args->ring_base_address); + + pr_debug("Queue Size (0x%llX, %u)\n", + q_properties->queue_size, args->ring_size); + + pr_debug("Queue r/w Pointers (0x%llX, 0x%llX)\n", + (uint64_t) q_properties->read_ptr, + (uint64_t) q_properties->write_ptr); + + pr_debug("Queue Format (%d)\n", q_properties->format); + + return 0; +} + static long kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p, void __user *arg) { - return -ENODEV; + struct kfd_ioctl_create_queue_args args; + struct kfd_dev *dev; + int err = 0; + unsigned int queue_id; + struct kfd_process_device *pdd; + struct queue_properties q_properties; + + memset(&q_properties, 0, sizeof(struct queue_properties)); + + if (copy_from_user(&args, arg, sizeof(args))) + return -EFAULT; + + pr_debug("kfd: creating queue ioctl\n"); + + err = set_queue_properties_from_user(&q_properties, &args); + if (err) + return err; + + dev = kfd_device_by_id(args.gpu_id); + if (dev == NULL) + return -EINVAL; + + mutex_lock(&p->mutex); + + pdd = kfd_bind_process_to_device(dev, p); + if (IS_ERR(pdd) < 0) { + err = PTR_ERR(pdd); + goto err_bind_process; + } + + pr_debug("kfd: creating queue for PASID %d on GPU 0x%x\n", + p->pasid, + dev->id); + + err = pqm_create_queue(&p->pqm, dev, filep, &q_properties, 0, + KFD_QUEUE_TYPE_COMPUTE, &queue_id); + if (err != 0) + goto err_create_queue; + + args.queue_id = queue_id; + + /* Return gpu_id as doorbell offset for mmap usage */ + args.doorbell_offset = args.gpu_id << PAGE_SHIFT; + + if (copy_to_user(arg, &args, sizeof(args))) { + err = -EFAULT; + goto err_copy_args_out; + } + + mutex_unlock(&p->mutex); + + pr_debug("kfd: queue id %d was created successfully\n", args.queue_id); + + pr_debug("ring buffer address == 0x%016llX\n", + args.ring_base_address); + + pr_debug("read ptr address == 0x%016llX\n", + args.read_pointer_address); + + pr_debug("write ptr address == 0x%016llX\n", + args.write_pointer_address); + + return 0; + +err_copy_args_out: + pqm_destroy_queue(&p->pqm, queue_id); +err_create_queue: +err_bind_process: + mutex_unlock(&p->mutex); + return err; } static int kfd_ioctl_destroy_queue(struct file *filp, struct kfd_process *p, void __user *arg) { - return -ENODEV; + int retval; + struct kfd_ioctl_destroy_queue_args args; + + if (copy_from_user(&args, arg, sizeof(args))) + return -EFAULT; + + pr_debug("kfd: destroying queue id %d for PASID %d\n", + args.queue_id, + p->pasid); + + mutex_lock(&p->mutex); + + retval = pqm_destroy_queue(&p->pqm, args.queue_id); + + mutex_unlock(&p->mutex); + return retval; } static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p, void __user *arg) { - return -ENODEV; + int retval; + struct kfd_ioctl_update_queue_args args; + struct queue_properties properties; + + if (copy_from_user(&args, arg, sizeof(args))) + return -EFAULT; + + if (args.queue_percentage > KFD_MAX_QUEUE_PERCENTAGE) { + pr_err("kfd: queue percentage must be between 0 to KFD_MAX_QUEUE_PERCENTAGE\n"); + return -EINVAL; + } + + if (args.queue_priority > KFD_MAX_QUEUE_PRIORITY) { + pr_err("kfd: queue priority must be between 0 to KFD_MAX_QUEUE_PRIORITY\n"); + return -EINVAL; + } + + if ((args.ring_base_address) && + (!access_ok(VERIFY_WRITE, args.ring_base_address, sizeof(uint64_t)))) { + pr_err("kfd: can't access ring base address\n"); + return -EFAULT; + } + + if (!is_power_of_2(args.ring_size) && (args.ring_size != 0)) { + pr_err("kfd: ring size must be a power of 2 or 0\n"); + return -EINVAL; + } + + properties.queue_address = args.ring_base_address; + properties.queue_size = args.ring_size; + properties.queue_percent = args.queue_percentage; + properties.priority = args.queue_priority; + + pr_debug("kfd: updating queue id %d for PASID %d\n", + args.queue_id, p->pasid); + + mutex_lock(&p->mutex); + + retval = pqm_update_queue(&p->pqm, args.queue_id, &properties); + + mutex_unlock(&p->mutex); + + return retval; } static long kfd_ioctl_set_memory_policy(struct file *filep, -- cgit v0.10.2 From 41a286fa54e850add4482146c511b088354f6f2d Mon Sep 17 00:00:00 2001 From: Andrew Lewycky Date: Thu, 17 Jul 2014 01:46:17 +0300 Subject: amdkfd: Implement the Set Memory Policy IOCTL Signed-off-by: Andrew Lewycky Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index d342035..1020faf 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -35,6 +35,7 @@ #include #include #include "kfd_priv.h" +#include "kfd_device_queue_manager.h" static long kfd_ioctl(struct file *, unsigned int, unsigned long); static int kfd_open(struct inode *, struct file *); @@ -345,7 +346,56 @@ static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p, static long kfd_ioctl_set_memory_policy(struct file *filep, struct kfd_process *p, void __user *arg) { - return -ENODEV; + struct kfd_ioctl_set_memory_policy_args args; + struct kfd_dev *dev; + int err = 0; + struct kfd_process_device *pdd; + enum cache_policy default_policy, alternate_policy; + + if (copy_from_user(&args, arg, sizeof(args))) + return -EFAULT; + + if (args.default_policy != KFD_IOC_CACHE_POLICY_COHERENT + && args.default_policy != KFD_IOC_CACHE_POLICY_NONCOHERENT) { + return -EINVAL; + } + + if (args.alternate_policy != KFD_IOC_CACHE_POLICY_COHERENT + && args.alternate_policy != KFD_IOC_CACHE_POLICY_NONCOHERENT) { + return -EINVAL; + } + + dev = kfd_device_by_id(args.gpu_id); + if (dev == NULL) + return -EINVAL; + + mutex_lock(&p->mutex); + + pdd = kfd_bind_process_to_device(dev, p); + if (IS_ERR(pdd) < 0) { + err = PTR_ERR(pdd); + goto out; + } + + default_policy = (args.default_policy == KFD_IOC_CACHE_POLICY_COHERENT) + ? cache_policy_coherent : cache_policy_noncoherent; + + alternate_policy = + (args.alternate_policy == KFD_IOC_CACHE_POLICY_COHERENT) + ? cache_policy_coherent : cache_policy_noncoherent; + + if (!dev->dqm->set_cache_memory_policy(dev->dqm, + &pdd->qpd, + default_policy, + alternate_policy, + (void __user *)args.alternate_aperture_base, + args.alternate_aperture_size)) + err = -EINVAL; + +out: + mutex_unlock(&p->mutex); + + return err; } static long kfd_ioctl_get_clock_counters(struct file *filep, -- cgit v0.10.2 From 4fac47c8206750128634622853c400570c44040d Mon Sep 17 00:00:00 2001 From: Evgeny Pinchuk Date: Thu, 17 Jul 2014 01:47:58 +0300 Subject: amdkfd: Implement the Get Clock Counters IOCTL Signed-off-by: Evgeny Pinchuk Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 1020faf..473f8e1 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -401,7 +401,34 @@ out: static long kfd_ioctl_get_clock_counters(struct file *filep, struct kfd_process *p, void __user *arg) { - return -ENODEV; + struct kfd_ioctl_get_clock_counters_args args; + struct kfd_dev *dev; + struct timespec time; + + if (copy_from_user(&args, arg, sizeof(args))) + return -EFAULT; + + dev = kfd_device_by_id(args.gpu_id); + if (dev == NULL) + return -EINVAL; + + /* Reading GPU clock counter from KGD */ + args.gpu_clock_counter = kfd2kgd->get_gpu_clock_counter(dev->kgd); + + /* No access to rdtsc. Using raw monotonic time */ + getrawmonotonic(&time); + args.cpu_clock_counter = (uint64_t)timespec_to_ns(&time); + + get_monotonic_boottime(&time); + args.system_clock_counter = (uint64_t)timespec_to_ns(&time); + + /* Since the counter is in nano-seconds we use 1GHz frequency */ + args.system_clock_freq = 1000000000; + + if (copy_to_user(arg, &args, sizeof(args))) + return -EFAULT; + + return 0; } -- cgit v0.10.2 From 775921edc1c10f6022ee6614ff2914704fc605bf Mon Sep 17 00:00:00 2001 From: Alexey Skidanov Date: Thu, 17 Jul 2014 01:49:36 +0300 Subject: amdkfd: Implement the Get Process Aperture IOCTL v3: Fixed debug messages Signed-off-by: Alexey Skidanov Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 473f8e1..3130265 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -435,7 +435,61 @@ static long kfd_ioctl_get_clock_counters(struct file *filep, static int kfd_ioctl_get_process_apertures(struct file *filp, struct kfd_process *p, void __user *arg) { - return -ENODEV; + struct kfd_ioctl_get_process_apertures_args args; + struct kfd_process_device_apertures *pAperture; + struct kfd_process_device *pdd; + + dev_dbg(kfd_device, "get apertures for PASID %d", p->pasid); + + if (copy_from_user(&args, arg, sizeof(args))) + return -EFAULT; + + args.num_of_nodes = 0; + + mutex_lock(&p->mutex); + + /*if the process-device list isn't empty*/ + if (kfd_has_process_device_data(p)) { + /* Run over all pdd of the process */ + pdd = kfd_get_first_process_device_data(p); + do { + pAperture = &args.process_apertures[args.num_of_nodes]; + pAperture->gpu_id = pdd->dev->id; + pAperture->lds_base = pdd->lds_base; + pAperture->lds_limit = pdd->lds_limit; + pAperture->gpuvm_base = pdd->gpuvm_base; + pAperture->gpuvm_limit = pdd->gpuvm_limit; + pAperture->scratch_base = pdd->scratch_base; + pAperture->scratch_limit = pdd->scratch_limit; + + dev_dbg(kfd_device, + "node id %u\n", args.num_of_nodes); + dev_dbg(kfd_device, + "gpu id %u\n", pdd->dev->id); + dev_dbg(kfd_device, + "lds_base %llX\n", pdd->lds_base); + dev_dbg(kfd_device, + "lds_limit %llX\n", pdd->lds_limit); + dev_dbg(kfd_device, + "gpuvm_base %llX\n", pdd->gpuvm_base); + dev_dbg(kfd_device, + "gpuvm_limit %llX\n", pdd->gpuvm_limit); + dev_dbg(kfd_device, + "scratch_base %llX\n", pdd->scratch_base); + dev_dbg(kfd_device, + "scratch_limit %llX\n", pdd->scratch_limit); + + args.num_of_nodes++; + } while ((pdd = kfd_get_next_process_device_data(p, pdd)) != NULL && + (args.num_of_nodes < NUM_OF_SUPPORTED_GPUS)); + } + + mutex_unlock(&p->mutex); + + if (copy_to_user(arg, &args, sizeof(args))) + return -EFAULT; + + return 0; } static long kfd_ioctl(struct file *filep, unsigned int cmd, unsigned long arg) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 2bc34aa1..41e608d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -473,6 +473,12 @@ struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, struct kfd_process *p, int create_pdd); +/* Process device data iterator */ +struct kfd_process_device *kfd_get_first_process_device_data(struct kfd_process *p); +struct kfd_process_device *kfd_get_next_process_device_data(struct kfd_process *p, + struct kfd_process_device *pdd); +bool kfd_has_process_device_data(struct kfd_process *p); + /* PASIDs */ int kfd_pasid_init(void); void kfd_pasid_exit(void); -- cgit v0.10.2 From 8c50f10d73b50139dcfe48bc22f2c8c7822c1983 Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Fri, 5 Sep 2014 14:13:16 +0100 Subject: drm/i915: Enable full PPGTT on gen7 Use full PPGTT as the default option in gen7. Note that aliasing PPGTT is the default option for gen8 (see HAS_PPGTT) since we're still fighting troubles around context switching and execlists. This may well come back to bite me later. Signed-off-by: Michel Thierry [danvet: Explain that gen8 full ppgtt is blocked on execlists for now.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 6f410cf..14f078c 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -59,7 +59,7 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) return 0; } - return HAS_ALIASING_PPGTT(dev) ? 1 : 0; + return HAS_PPGTT(dev) ? 2 : HAS_ALIASING_PPGTT(dev) ? 1 : 0; } -- cgit v0.10.2 From d518ce50e7c5bbdf0f1ff404b40cd073930af5a2 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 27 Aug 2014 10:43:37 +0200 Subject: drm/i915: WARN if interrupts aren't on in en/disable_pipestat Now that vlv has runtime pm we kinda should check for that like on the pch split platforms. Looks like this was simply lost in the vlv rpm enabling. Cc: Paulo Zanoni Cc: Imre Deak Cc: Jesse Barnes Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index c96ddc9..4847ed5 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -589,6 +589,7 @@ __i915_enable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe, u32 pipestat = I915_READ(reg) & PIPESTAT_INT_ENABLE_MASK; assert_spin_locked(&dev_priv->irq_lock); + WARN_ON(!intel_irqs_enabled(dev_priv)); if (WARN_ONCE(enable_mask & ~PIPESTAT_INT_ENABLE_MASK || status_mask & ~PIPESTAT_INT_STATUS_MASK, @@ -615,6 +616,7 @@ __i915_disable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe, u32 pipestat = I915_READ(reg) & PIPESTAT_INT_ENABLE_MASK; assert_spin_locked(&dev_priv->irq_lock); + WARN_ON(!intel_irqs_enabled(dev_priv)); if (WARN_ONCE(enable_mask & ~PIPESTAT_INT_ENABLE_MASK || status_mask & ~PIPESTAT_INT_STATUS_MASK, -- cgit v0.10.2 From eeca778a3ec71022630e50935cfb7edab6f6f9e5 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 5 Sep 2014 17:04:46 -0300 Subject: drm/i915: create struct intel_plane_state MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This new struct will be the storage of src and dst coordinates between the check and commit stages of a plane update. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 07ce046..dd5e0f1 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -34,6 +34,7 @@ #include #include #include +#include /** * _wait_for - magic (register) wait macro @@ -237,6 +238,17 @@ typedef struct dpll { int p; } intel_clock_t; +struct intel_plane_state { + struct drm_crtc *crtc; + struct drm_framebuffer *fb; + struct drm_rect src; + struct drm_rect dst; + struct drm_rect clip; + struct drm_rect orig_src; + struct drm_rect orig_dst; + bool visible; +}; + struct intel_plane_config { bool tiled; int size; -- cgit v0.10.2 From 96d61a7f267ff355a401ca23a732810027d10ba2 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 5 Sep 2014 17:04:47 -0300 Subject: drm/i915: split intel_update_plane into check() and commit() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Due to the upcoming atomic modesetting feature we need to separate some update functions into a check step that can fail and a commit step that should, ideally, never fail. This commit splits intel_update_plane() and its commit part can still fail due to the fb pinning procedure. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 07a74ef..a4306cf 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -845,57 +845,24 @@ static bool colorkey_enabled(struct intel_plane *intel_plane) } static int -intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, - struct drm_framebuffer *fb, int crtc_x, int crtc_y, - unsigned int crtc_w, unsigned int crtc_h, - uint32_t src_x, uint32_t src_y, - uint32_t src_w, uint32_t src_h) +intel_check_sprite_plane(struct drm_plane *plane, + struct intel_plane_state *state) { - struct drm_device *dev = plane->dev; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(state->crtc); struct intel_plane *intel_plane = to_intel_plane(plane); - enum pipe pipe = intel_crtc->pipe; + struct drm_framebuffer *fb = state->fb; struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); struct drm_i915_gem_object *obj = intel_fb->obj; - struct drm_i915_gem_object *old_obj = intel_plane->obj; - int ret; - bool primary_enabled; - bool visible; + int crtc_x, crtc_y; + unsigned int crtc_w, crtc_h; + uint32_t src_x, src_y, src_w, src_h; + struct drm_rect *src = &state->src; + struct drm_rect *dst = &state->dst; + struct drm_rect *orig_src = &state->orig_src; + const struct drm_rect *clip = &state->clip; int hscale, vscale; int max_scale, min_scale; int pixel_size = drm_format_plane_cpp(fb->pixel_format, 0); - struct drm_rect src = { - /* sample coordinates in 16.16 fixed point */ - .x1 = src_x, - .x2 = src_x + src_w, - .y1 = src_y, - .y2 = src_y + src_h, - }; - struct drm_rect dst = { - /* integer pixels */ - .x1 = crtc_x, - .x2 = crtc_x + crtc_w, - .y1 = crtc_y, - .y2 = crtc_y + crtc_h, - }; - const struct drm_rect clip = { - .x2 = intel_crtc->active ? intel_crtc->config.pipe_src_w : 0, - .y2 = intel_crtc->active ? intel_crtc->config.pipe_src_h : 0, - }; - const struct { - int crtc_x, crtc_y; - unsigned int crtc_w, crtc_h; - uint32_t src_x, src_y, src_w, src_h; - } orig = { - .crtc_x = crtc_x, - .crtc_y = crtc_y, - .crtc_w = crtc_w, - .crtc_h = crtc_h, - .src_x = src_x, - .src_y = src_y, - .src_w = src_w, - .src_h = src_h, - }; /* Don't modify another pipe's plane */ if (intel_plane->pipe != intel_crtc->pipe) { @@ -927,55 +894,55 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, max_scale = intel_plane->max_downscale << 16; min_scale = intel_plane->can_scale ? 1 : (1 << 16); - drm_rect_rotate(&src, fb->width << 16, fb->height << 16, + drm_rect_rotate(src, fb->width << 16, fb->height << 16, intel_plane->rotation); - hscale = drm_rect_calc_hscale_relaxed(&src, &dst, min_scale, max_scale); + hscale = drm_rect_calc_hscale_relaxed(src, dst, min_scale, max_scale); BUG_ON(hscale < 0); - vscale = drm_rect_calc_vscale_relaxed(&src, &dst, min_scale, max_scale); + vscale = drm_rect_calc_vscale_relaxed(src, dst, min_scale, max_scale); BUG_ON(vscale < 0); - visible = drm_rect_clip_scaled(&src, &dst, &clip, hscale, vscale); + state->visible = drm_rect_clip_scaled(src, dst, clip, hscale, vscale); - crtc_x = dst.x1; - crtc_y = dst.y1; - crtc_w = drm_rect_width(&dst); - crtc_h = drm_rect_height(&dst); + crtc_x = dst->x1; + crtc_y = dst->y1; + crtc_w = drm_rect_width(dst); + crtc_h = drm_rect_height(dst); - if (visible) { + if (state->visible) { /* check again in case clipping clamped the results */ - hscale = drm_rect_calc_hscale(&src, &dst, min_scale, max_scale); + hscale = drm_rect_calc_hscale(src, dst, min_scale, max_scale); if (hscale < 0) { DRM_DEBUG_KMS("Horizontal scaling factor out of limits\n"); - drm_rect_debug_print(&src, true); - drm_rect_debug_print(&dst, false); + drm_rect_debug_print(src, true); + drm_rect_debug_print(dst, false); return hscale; } - vscale = drm_rect_calc_vscale(&src, &dst, min_scale, max_scale); + vscale = drm_rect_calc_vscale(src, dst, min_scale, max_scale); if (vscale < 0) { DRM_DEBUG_KMS("Vertical scaling factor out of limits\n"); - drm_rect_debug_print(&src, true); - drm_rect_debug_print(&dst, false); + drm_rect_debug_print(src, true); + drm_rect_debug_print(dst, false); return vscale; } /* Make the source viewport size an exact multiple of the scaling factors. */ - drm_rect_adjust_size(&src, - drm_rect_width(&dst) * hscale - drm_rect_width(&src), - drm_rect_height(&dst) * vscale - drm_rect_height(&src)); + drm_rect_adjust_size(src, + drm_rect_width(dst) * hscale - drm_rect_width(src), + drm_rect_height(dst) * vscale - drm_rect_height(src)); - drm_rect_rotate_inv(&src, fb->width << 16, fb->height << 16, + drm_rect_rotate_inv(src, fb->width << 16, fb->height << 16, intel_plane->rotation); /* sanity check to make sure the src viewport wasn't enlarged */ - WARN_ON(src.x1 < (int) src_x || - src.y1 < (int) src_y || - src.x2 > (int) (src_x + src_w) || - src.y2 > (int) (src_y + src_h)); + WARN_ON(src->x1 < (int) orig_src->x1 || + src->y1 < (int) orig_src->y1 || + src->x2 > (int) orig_src->x2 || + src->y2 > (int) orig_src->y2); /* * Hardware doesn't handle subpixel coordinates. @@ -983,10 +950,10 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, * increase the source viewport size, because that could * push the downscaling factor out of bounds. */ - src_x = src.x1 >> 16; - src_w = drm_rect_width(&src) >> 16; - src_y = src.y1 >> 16; - src_h = drm_rect_height(&src) >> 16; + src_x = src->x1 >> 16; + src_w = drm_rect_width(src) >> 16; + src_y = src->y1 >> 16; + src_h = drm_rect_height(src) >> 16; if (format_is_yuv(fb->pixel_format)) { src_x &= ~1; @@ -1000,12 +967,12 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, crtc_w &= ~1; if (crtc_w == 0) - visible = false; + state->visible = false; } } /* Check size restrictions when scaling */ - if (visible && (src_w != crtc_w || src_h != crtc_h)) { + if (state->visible && (src_w != crtc_w || src_h != crtc_h)) { unsigned int width_bytes; WARN_ON(!intel_plane->can_scale); @@ -1013,12 +980,13 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, /* FIXME interlacing min height is 6 */ if (crtc_w < 3 || crtc_h < 3) - visible = false; + state->visible = false; if (src_w < 3 || src_h < 3) - visible = false; + state->visible = false; - width_bytes = ((src_x * pixel_size) & 63) + src_w * pixel_size; + width_bytes = ((src_x * pixel_size) & 63) + + src_w * pixel_size; if (src_w > 2048 || src_h > 2048 || width_bytes > 4096 || fb->pitches[0] > 4096) { @@ -1027,17 +995,48 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, } } - dst.x1 = crtc_x; - dst.x2 = crtc_x + crtc_w; - dst.y1 = crtc_y; - dst.y2 = crtc_y + crtc_h; + if (state->visible) { + src->x1 = src_x; + src->x2 = src_x + src_w; + src->y1 = src_y; + src->y2 = src_y + src_h; + } + + dst->x1 = crtc_x; + dst->x2 = crtc_x + crtc_w; + dst->y1 = crtc_y; + dst->y2 = crtc_y + crtc_h; + + return 0; +} + +static int +intel_commit_sprite_plane(struct drm_plane *plane, + struct intel_plane_state *state) +{ + struct drm_device *dev = plane->dev; + struct drm_crtc *crtc = state->crtc; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_plane *intel_plane = to_intel_plane(plane); + enum pipe pipe = intel_crtc->pipe; + struct drm_framebuffer *fb = state->fb; + struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); + struct drm_i915_gem_object *obj = intel_fb->obj; + struct drm_i915_gem_object *old_obj = intel_plane->obj; + int crtc_x, crtc_y; + unsigned int crtc_w, crtc_h; + uint32_t src_x, src_y, src_w, src_h; + struct drm_rect *dst = &state->dst; + const struct drm_rect *clip = &state->clip; + bool primary_enabled; + int ret; /* * If the sprite is completely covering the primary plane, * we can disable the primary and save power. */ - primary_enabled = !drm_rect_equals(&dst, &clip) || colorkey_enabled(intel_plane); - WARN_ON(!primary_enabled && !visible && intel_crtc->active); + primary_enabled = !drm_rect_equals(dst, clip) || colorkey_enabled(intel_plane); + WARN_ON(!primary_enabled && !state->visible && intel_crtc->active); mutex_lock(&dev->struct_mutex); @@ -1055,14 +1054,14 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, if (ret) return ret; - intel_plane->crtc_x = orig.crtc_x; - intel_plane->crtc_y = orig.crtc_y; - intel_plane->crtc_w = orig.crtc_w; - intel_plane->crtc_h = orig.crtc_h; - intel_plane->src_x = orig.src_x; - intel_plane->src_y = orig.src_y; - intel_plane->src_w = orig.src_w; - intel_plane->src_h = orig.src_h; + intel_plane->crtc_x = state->orig_dst.x1; + intel_plane->crtc_y = state->orig_dst.y1; + intel_plane->crtc_w = drm_rect_width(&state->orig_dst); + intel_plane->crtc_h = drm_rect_height(&state->orig_dst); + intel_plane->src_x = state->orig_src.x1; + intel_plane->src_y = state->orig_src.y1; + intel_plane->src_w = drm_rect_width(&state->orig_src); + intel_plane->src_h = drm_rect_height(&state->orig_src); intel_plane->obj = obj; if (intel_crtc->active) { @@ -1076,12 +1075,22 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, if (primary_was_enabled && !primary_enabled) intel_pre_disable_primary(crtc); - if (visible) + if (state->visible) { + crtc_x = state->dst.x1; + crtc_y = state->dst.x2; + crtc_w = drm_rect_width(&state->dst); + crtc_h = drm_rect_height(&state->dst); + src_x = state->src.x1; + src_y = state->src.y1; + src_w = drm_rect_width(&state->src); + src_h = drm_rect_height(&state->src); intel_plane->update_plane(plane, crtc, fb, obj, crtc_x, crtc_y, crtc_w, crtc_h, src_x, src_y, src_w, src_h); - else + } else { intel_plane->disable_plane(plane, crtc); + } + intel_frontbuffer_flip(dev, INTEL_FRONTBUFFER_SPRITE(pipe)); @@ -1109,6 +1118,46 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, } static int +intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, + struct drm_framebuffer *fb, int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t src_x, uint32_t src_y, + uint32_t src_w, uint32_t src_h) +{ + struct intel_plane_state state; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + int ret; + + state.crtc = crtc; + state.fb = fb; + + /* sample coordinates in 16.16 fixed point */ + state.src.x1 = src_x; + state.src.x2 = src_x + src_w; + state.src.y1 = src_y; + state.src.y2 = src_y + src_h; + + /* integer pixels */ + state.dst.x1 = crtc_x; + state.dst.x2 = crtc_x + crtc_w; + state.dst.y1 = crtc_y; + state.dst.y2 = crtc_y + crtc_h; + + state.clip.x1 = 0; + state.clip.y1 = 0; + state.clip.x2 = intel_crtc->active ? intel_crtc->config.pipe_src_w : 0; + state.clip.y2 = intel_crtc->active ? intel_crtc->config.pipe_src_h : 0; + state.orig_src = state.src; + state.orig_dst = state.dst; + + ret = intel_check_sprite_plane(plane, &state); + if (ret) + return ret; + + return intel_commit_sprite_plane(plane, &state); +} + +static int intel_disable_plane(struct drm_plane *plane) { struct drm_device *dev = plane->dev; -- cgit v0.10.2 From 852e787c4cb9fcc34cf81d2a764a1aee2d9f24bd Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 5 Sep 2014 17:22:31 -0300 Subject: drm/i915: split intel_cursor_plane_update() into check() and commit() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Due to the upcoming atomic modesetting feature we need to separate some update functions into a check step that can fail and a commit step that should, ideally, never fail. The commit part can still fail, but that should be solved in another upcoming patch. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index c5079f2..8de7d4c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -12040,51 +12040,41 @@ intel_cursor_plane_disable(struct drm_plane *plane) } static int -intel_cursor_plane_update(struct drm_plane *plane, struct drm_crtc *crtc, - struct drm_framebuffer *fb, int crtc_x, int crtc_y, - unsigned int crtc_w, unsigned int crtc_h, - uint32_t src_x, uint32_t src_y, - uint32_t src_w, uint32_t src_h) +intel_check_cursor_plane(struct drm_plane *plane, + struct intel_plane_state *state) { - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); - struct drm_i915_gem_object *obj = intel_fb->obj; - struct drm_rect dest = { - /* integer pixels */ - .x1 = crtc_x, - .y1 = crtc_y, - .x2 = crtc_x + crtc_w, - .y2 = crtc_y + crtc_h, - }; - struct drm_rect src = { - /* 16.16 fixed point */ - .x1 = src_x, - .y1 = src_y, - .x2 = src_x + src_w, - .y2 = src_y + src_h, - }; - const struct drm_rect clip = { - /* integer pixels */ - .x2 = intel_crtc->active ? intel_crtc->config.pipe_src_w : 0, - .y2 = intel_crtc->active ? intel_crtc->config.pipe_src_h : 0, - }; - bool visible; - int ret; + struct drm_crtc *crtc = state->crtc; + struct drm_framebuffer *fb = state->fb; + struct drm_rect *dest = &state->dst; + struct drm_rect *src = &state->src; + const struct drm_rect *clip = &state->clip; - ret = drm_plane_helper_check_update(plane, crtc, fb, - &src, &dest, &clip, + return drm_plane_helper_check_update(plane, crtc, fb, + src, dest, clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, - true, true, &visible); - if (ret) - return ret; + true, true, &state->visible); +} - crtc->cursor_x = crtc_x; - crtc->cursor_y = crtc_y; +static int +intel_commit_cursor_plane(struct drm_plane *plane, + struct intel_plane_state *state) +{ + struct drm_crtc *crtc = state->crtc; + struct drm_framebuffer *fb = state->fb; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); + struct drm_i915_gem_object *obj = intel_fb->obj; + int crtc_w, crtc_h; + + crtc->cursor_x = state->orig_dst.x1; + crtc->cursor_y = state->orig_dst.y1; if (fb != crtc->cursor->fb) { + crtc_w = drm_rect_width(&state->orig_dst); + crtc_h = drm_rect_height(&state->orig_dst); return intel_crtc_cursor_set_obj(crtc, obj, crtc_w, crtc_h); } else { - intel_crtc_update_cursor(crtc, visible); + intel_crtc_update_cursor(crtc, state->visible); intel_frontbuffer_flip(crtc->dev, INTEL_FRONTBUFFER_CURSOR(intel_crtc->pipe)); @@ -12092,6 +12082,48 @@ intel_cursor_plane_update(struct drm_plane *plane, struct drm_crtc *crtc, return 0; } } + +static int +intel_cursor_plane_update(struct drm_plane *plane, struct drm_crtc *crtc, + struct drm_framebuffer *fb, int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t src_x, uint32_t src_y, + uint32_t src_w, uint32_t src_h) +{ + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_plane_state state; + int ret; + + state.crtc = crtc; + state.fb = fb; + + /* sample coordinates in 16.16 fixed point */ + state.src.x1 = src_x; + state.src.x2 = src_x + src_w; + state.src.y1 = src_y; + state.src.y2 = src_y + src_h; + + /* integer pixels */ + state.dst.x1 = crtc_x; + state.dst.x2 = crtc_x + crtc_w; + state.dst.y1 = crtc_y; + state.dst.y2 = crtc_y + crtc_h; + + state.clip.x1 = 0; + state.clip.y1 = 0; + state.clip.x2 = intel_crtc->active ? intel_crtc->config.pipe_src_w : 0; + state.clip.y2 = intel_crtc->active ? intel_crtc->config.pipe_src_h : 0; + + state.orig_src = state.src; + state.orig_dst = state.dst; + + ret = intel_check_cursor_plane(plane, &state); + if (ret) + return ret; + + return intel_commit_cursor_plane(plane, &state); +} + static const struct drm_plane_funcs intel_cursor_plane_funcs = { .update_plane = intel_cursor_plane_update, .disable_plane = intel_cursor_plane_disable, -- cgit v0.10.2 From 3c692a4162b7bf9eff9ba677228254d2f633a785 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 5 Sep 2014 17:04:49 -0300 Subject: drm/i915: split intel_primary_plane_setplane() into check() and commit() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit As a preparation for atomic updates we need to split the code to check everything we are going to commit first. This patch starts the work to split intel_primary_plane_setplane() into check() and commit() parts. More work is expected on this to get a better split of the two steps. Ideally the commit() step should never fail. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 8de7d4c..fc1844d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11807,63 +11807,37 @@ disable_unpin: } static int -intel_primary_plane_setplane(struct drm_plane *plane, struct drm_crtc *crtc, - struct drm_framebuffer *fb, int crtc_x, int crtc_y, - unsigned int crtc_w, unsigned int crtc_h, - uint32_t src_x, uint32_t src_y, - uint32_t src_w, uint32_t src_h) +intel_check_primary_plane(struct drm_plane *plane, + struct intel_plane_state *state) { + struct drm_crtc *crtc = state->crtc; + struct drm_framebuffer *fb = state->fb; + struct drm_rect *dest = &state->dst; + struct drm_rect *src = &state->src; + const struct drm_rect *clip = &state->clip; + + return drm_plane_helper_check_update(plane, crtc, fb, + src, dest, clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + false, true, &state->visible); +} + +static int +intel_commit_primary_plane(struct drm_plane *plane, + struct intel_plane_state *state) +{ + struct drm_crtc *crtc = state->crtc; + struct drm_framebuffer *fb = state->fb; struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); struct drm_i915_gem_object *obj = intel_fb_obj(fb); struct drm_i915_gem_object *old_obj = intel_fb_obj(plane->fb); - struct drm_rect dest = { - /* integer pixels */ - .x1 = crtc_x, - .y1 = crtc_y, - .x2 = crtc_x + crtc_w, - .y2 = crtc_y + crtc_h, - }; - struct drm_rect src = { - /* 16.16 fixed point */ - .x1 = src_x, - .y1 = src_y, - .x2 = src_x + src_w, - .y2 = src_y + src_h, - }; - const struct drm_rect clip = { - /* integer pixels */ - .x2 = intel_crtc->active ? intel_crtc->config.pipe_src_w : 0, - .y2 = intel_crtc->active ? intel_crtc->config.pipe_src_h : 0, - }; - const struct { - int crtc_x, crtc_y; - unsigned int crtc_w, crtc_h; - uint32_t src_x, src_y, src_w, src_h; - } orig = { - .crtc_x = crtc_x, - .crtc_y = crtc_y, - .crtc_w = crtc_w, - .crtc_h = crtc_h, - .src_x = src_x, - .src_y = src_y, - .src_w = src_w, - .src_h = src_h, - }; struct intel_plane *intel_plane = to_intel_plane(plane); - bool visible; + struct drm_rect *src = &state->src; int ret; - ret = drm_plane_helper_check_update(plane, crtc, fb, - &src, &dest, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - false, true, &visible); - - if (ret) - return ret; - /* * If the CRTC isn't enabled, we're just pinning the framebuffer, * updating the fb pointer, and returning without touching the @@ -11898,7 +11872,7 @@ intel_primary_plane_setplane(struct drm_plane *plane, struct drm_crtc *crtc, * happens if userspace explicitly disables the plane by passing fb=0 * because plane->fb still gets set and pinned. */ - if (!visible) { + if (!state->visible) { mutex_lock(&dev->struct_mutex); /* @@ -11945,7 +11919,7 @@ intel_primary_plane_setplane(struct drm_plane *plane, struct drm_crtc *crtc, intel_disable_fbc(dev); } } - ret = intel_pipe_set_base(crtc, src.x1, src.y1, fb); + ret = intel_pipe_set_base(crtc, src->x1, src->y1, fb); if (ret) return ret; @@ -11953,19 +11927,62 @@ intel_primary_plane_setplane(struct drm_plane *plane, struct drm_crtc *crtc, intel_enable_primary_hw_plane(plane, crtc); } - intel_plane->crtc_x = orig.crtc_x; - intel_plane->crtc_y = orig.crtc_y; - intel_plane->crtc_w = orig.crtc_w; - intel_plane->crtc_h = orig.crtc_h; - intel_plane->src_x = orig.src_x; - intel_plane->src_y = orig.src_y; - intel_plane->src_w = orig.src_w; - intel_plane->src_h = orig.src_h; + intel_plane->crtc_x = state->orig_dst.x1; + intel_plane->crtc_y = state->orig_dst.y1; + intel_plane->crtc_w = drm_rect_width(&state->orig_dst); + intel_plane->crtc_h = drm_rect_height(&state->orig_dst); + intel_plane->src_x = state->orig_src.x1; + intel_plane->src_y = state->orig_src.y1; + intel_plane->src_w = drm_rect_width(&state->orig_src); + intel_plane->src_h = drm_rect_height(&state->orig_src); intel_plane->obj = obj; return 0; } +static int +intel_primary_plane_setplane(struct drm_plane *plane, struct drm_crtc *crtc, + struct drm_framebuffer *fb, int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t src_x, uint32_t src_y, + uint32_t src_w, uint32_t src_h) +{ + struct intel_plane_state state; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + int ret; + + state.crtc = crtc; + state.fb = fb; + + /* sample coordinates in 16.16 fixed point */ + state.src.x1 = src_x; + state.src.x2 = src_x + src_w; + state.src.y1 = src_y; + state.src.y2 = src_y + src_h; + + /* integer pixels */ + state.dst.x1 = crtc_x; + state.dst.x2 = crtc_x + crtc_w; + state.dst.y1 = crtc_y; + state.dst.y2 = crtc_y + crtc_h; + + state.clip.x1 = 0; + state.clip.y1 = 0; + state.clip.x2 = intel_crtc->active ? intel_crtc->config.pipe_src_w : 0; + state.clip.y2 = intel_crtc->active ? intel_crtc->config.pipe_src_h : 0; + + state.orig_src = state.src; + state.orig_dst = state.dst; + + ret = intel_check_primary_plane(plane, &state); + if (ret) + return ret; + + intel_commit_primary_plane(plane, &state); + + return 0; +} + /* Common destruction function for both primary and cursor planes */ static void intel_plane_destroy(struct drm_plane *plane) { -- cgit v0.10.2 From 2363d8c97f87738dba7af7d2103f4f26b3bf5916 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 8 Sep 2014 18:28:20 +0200 Subject: drm/i915: Restore resume irq ordering comment This was lost in commit e11aa362308f5de467ce355a2a2471321b15a35c Author: Jesse Barnes Date: Wed Jun 18 09:52:55 2014 -0700 drm/i915: use runtime irq suspend/resume in freeze/thaw which makes the second part of this commen a bit nonsense. Both were originally added in commit 15239099d7a7a9ecdc1ccb5b187ae4cda5488ff9 Author: Daniel Vetter Date: Tue Mar 5 09:50:58 2013 +0100 drm/i915: enable irqs earlier when resuming Cc: Jesse Barnes Cc: Imre Deak Cc: Chris Wilson Cc: Mika Kuoppala Signed-off-by: Daniel Vetter Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 055d5e7..b8bd008 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -680,6 +680,7 @@ static int __i915_drm_thaw(struct drm_device *dev, bool restore_gtt_mappings) } mutex_unlock(&dev->struct_mutex); + /* We need working interrupts for modeset enabling ... */ intel_runtime_pm_restore_interrupts(dev); intel_modeset_init_hw(dev); -- cgit v0.10.2 From 94f598f176600a857817e010f266cdc30efa585b Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Tue, 9 Sep 2014 11:43:19 -0300 Subject: drm/i915: remove !enabled handling from commit primary plane step MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The !crtc->enabled case will now be handled by the !visible code, since the handling is basically the same. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index fc1844d..870555b 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11838,32 +11838,6 @@ intel_commit_primary_plane(struct drm_plane *plane, struct drm_rect *src = &state->src; int ret; - /* - * If the CRTC isn't enabled, we're just pinning the framebuffer, - * updating the fb pointer, and returning without touching the - * hardware. This allows us to later do a drmModeSetCrtc with fb=-1 to - * turn on the display with all planes setup as desired. - */ - if (!crtc->enabled) { - mutex_lock(&dev->struct_mutex); - - /* - * If we already called setplane while the crtc was disabled, - * we may have an fb pinned; unpin it. - */ - if (plane->fb) - intel_unpin_fb_obj(old_obj); - - i915_gem_track_fb(old_obj, obj, - INTEL_FRONTBUFFER_PRIMARY(intel_crtc->pipe)); - - /* Pin and return without programming hardware */ - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); - mutex_unlock(&dev->struct_mutex); - - return ret; - } - intel_crtc_wait_for_pending_flips(crtc); /* -- cgit v0.10.2 From 25067bfc060d1a481584dcb51ef4b5680176ecb6 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Wed, 10 Sep 2014 12:03:17 -0300 Subject: drm/i915: pin sprite fb only if it changed MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Optimize code avoiding helding dev mutex if old fb and current fb are the same. v2: take Ville's comments - move comment along with the pin_and_fence call - check for error before calling i915_gem_track_fb - move old_obj != obj to an upper if condition Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index a4306cf..90bb45f 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -1038,21 +1038,24 @@ intel_commit_sprite_plane(struct drm_plane *plane, primary_enabled = !drm_rect_equals(dst, clip) || colorkey_enabled(intel_plane); WARN_ON(!primary_enabled && !state->visible && intel_crtc->active); - mutex_lock(&dev->struct_mutex); - - /* Note that this will apply the VT-d workaround for scanouts, - * which is more restrictive than required for sprites. (The - * primary plane requires 256KiB alignment with 64 PTE padding, - * the sprite planes only require 128KiB alignment and 32 PTE padding. - */ - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); - i915_gem_track_fb(old_obj, obj, - INTEL_FRONTBUFFER_SPRITE(pipe)); - mutex_unlock(&dev->struct_mutex); + if (old_obj != obj) { + mutex_lock(&dev->struct_mutex); - if (ret) - return ret; + /* Note that this will apply the VT-d workaround for scanouts, + * which is more restrictive than required for sprites. (The + * primary plane requires 256KiB alignment with 64 PTE padding, + * the sprite planes only require 128KiB alignment and 32 PTE + * padding. + */ + ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); + if (ret == 0) + i915_gem_track_fb(old_obj, obj, + INTEL_FRONTBUFFER_SPRITE(pipe)); + mutex_unlock(&dev->struct_mutex); + if (ret) + return ret; + } intel_plane->crtc_x = state->orig_dst.x1; intel_plane->crtc_y = state->orig_dst.y1; @@ -1099,14 +1102,15 @@ intel_commit_sprite_plane(struct drm_plane *plane, } /* Unpin old obj after new one is active to avoid ugliness */ - if (old_obj) { + if (old_obj && old_obj != obj) { + /* * It's fairly common to simply update the position of * an existing object. In that case, we don't need to * wait for vblank to avoid ugliness, we only need to * do the pin & ref bookkeeping. */ - if (old_obj != obj && intel_crtc->active) + if (intel_crtc->active) intel_wait_for_vblank(dev, intel_crtc->pipe); mutex_lock(&dev->struct_mutex); -- cgit v0.10.2 From e30e8f7536f14ce42841f0d30bb2ca3d14f120f3 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Wed, 10 Sep 2014 12:04:17 -0300 Subject: drm/i915: create intel_update_pipe_size() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Factor out a piece of code from intel_pipe_set_base() that updates the pipe size and adjust fitter. This will help refactor the update primary plane path. v2: use struct intel_crtc as argument to intel_update_pipe_size() v3: use 'crtc' as argument name Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 870555b..1556c70 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2779,6 +2779,45 @@ static bool intel_crtc_has_pending_flip(struct drm_crtc *crtc) return pending; } +static void intel_update_pipe_size(struct intel_crtc *crtc) +{ + struct drm_device *dev = crtc->base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + const struct drm_display_mode *adjusted_mode; + + if (!i915.fastboot) + return; + + /* + * Update pipe size and adjust fitter if needed: the reason for this is + * that in compute_mode_changes we check the native mode (not the pfit + * mode) to see if we can flip rather than do a full mode set. In the + * fastboot case, we'll flip, but if we don't update the pipesrc and + * pfit state, we'll end up with a big fb scanned out into the wrong + * sized surface. + * + * To fix this properly, we need to hoist the checks up into + * compute_mode_changes (or above), check the actual pfit state and + * whether the platform allows pfit disable with pipe active, and only + * then update the pipesrc and pfit state, even on the flip path. + */ + + adjusted_mode = &crtc->config.adjusted_mode; + + I915_WRITE(PIPESRC(crtc->pipe), + ((adjusted_mode->crtc_hdisplay - 1) << 16) | + (adjusted_mode->crtc_vdisplay - 1)); + if (!crtc->config.pch_pfit.enabled && + (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS) || + intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_EDP))) { + I915_WRITE(PF_CTL(crtc->pipe), 0); + I915_WRITE(PF_WIN_POS(crtc->pipe), 0); + I915_WRITE(PF_WIN_SZ(crtc->pipe), 0); + } + crtc->config.pipe_src_w = adjusted_mode->crtc_hdisplay; + crtc->config.pipe_src_h = adjusted_mode->crtc_vdisplay; +} + static int intel_pipe_set_base(struct drm_crtc *crtc, int x, int y, struct drm_framebuffer *fb) @@ -2821,36 +2860,7 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y, return ret; } - /* - * Update pipe size and adjust fitter if needed: the reason for this is - * that in compute_mode_changes we check the native mode (not the pfit - * mode) to see if we can flip rather than do a full mode set. In the - * fastboot case, we'll flip, but if we don't update the pipesrc and - * pfit state, we'll end up with a big fb scanned out into the wrong - * sized surface. - * - * To fix this properly, we need to hoist the checks up into - * compute_mode_changes (or above), check the actual pfit state and - * whether the platform allows pfit disable with pipe active, and only - * then update the pipesrc and pfit state, even on the flip path. - */ - if (i915.fastboot) { - const struct drm_display_mode *adjusted_mode = - &intel_crtc->config.adjusted_mode; - - I915_WRITE(PIPESRC(intel_crtc->pipe), - ((adjusted_mode->crtc_hdisplay - 1) << 16) | - (adjusted_mode->crtc_vdisplay - 1)); - if (!intel_crtc->config.pch_pfit.enabled && - (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) || - intel_pipe_has_type(crtc, INTEL_OUTPUT_EDP))) { - I915_WRITE(PF_CTL(intel_crtc->pipe), 0); - I915_WRITE(PF_WIN_POS(intel_crtc->pipe), 0); - I915_WRITE(PF_WIN_SZ(intel_crtc->pipe), 0); - } - intel_crtc->config.pipe_src_w = adjusted_mode->crtc_hdisplay; - intel_crtc->config.pipe_src_h = adjusted_mode->crtc_vdisplay; - } + intel_update_pipe_size(intel_crtc); dev_priv->display.update_primary_plane(crtc, fb, x, y); -- cgit v0.10.2 From be2d599b5da3936ca92e0187ff50b34b6b8ff997 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 10 Sep 2014 19:52:18 +0100 Subject: drm/i915: Remove dead code, i915_gem_verify_gtt The data structure it was supposed to be sanity checking has long gone. Signed-off-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 28f91df..2fb87cf 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3384,46 +3384,6 @@ static bool i915_gem_valid_gtt_space(struct i915_vma *vma, return true; } -static void i915_gem_verify_gtt(struct drm_device *dev) -{ -#if WATCH_GTT - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_gem_object *obj; - int err = 0; - - list_for_each_entry(obj, &dev_priv->mm.gtt_list, global_list) { - if (obj->gtt_space == NULL) { - printk(KERN_ERR "object found on GTT list with no space reserved\n"); - err++; - continue; - } - - if (obj->cache_level != obj->gtt_space->color) { - printk(KERN_ERR "object reserved space [%08lx, %08lx] with wrong color, cache_level=%x, color=%lx\n", - i915_gem_obj_ggtt_offset(obj), - i915_gem_obj_ggtt_offset(obj) + i915_gem_obj_ggtt_size(obj), - obj->cache_level, - obj->gtt_space->color); - err++; - continue; - } - - if (!i915_gem_valid_gtt_space(dev, - obj->gtt_space, - obj->cache_level)) { - printk(KERN_ERR "invalid GTT space found at [%08lx, %08lx] - color=%x\n", - i915_gem_obj_ggtt_offset(obj), - i915_gem_obj_ggtt_offset(obj) + i915_gem_obj_ggtt_size(obj), - obj->cache_level); - err++; - continue; - } - } - - WARN_ON(err); -#endif -} - /** * Finds free space in the GTT aperture and binds the object there. */ @@ -3532,7 +3492,6 @@ search_free: vma->bind_vma(vma, obj->cache_level, flags & (PIN_MAPPABLE | PIN_GLOBAL) ? GLOBAL_BIND : 0); - i915_gem_verify_gtt(dev); return vma; err_remove_node: @@ -3769,7 +3728,6 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj, old_write_domain); } - i915_gem_verify_gtt(dev); return 0; } -- cgit v0.10.2 From e259f172adf25b40c7a0c9baeb71b783f3ef9036 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Thu, 11 Sep 2014 17:42:15 -0300 Subject: drm/i915: Fix regression in the sprite plane update split 7e4bf45dbd99a965c7b5d5944c6dc4246f171eb5 introduced the regression. We fix it by doing the right assignment of crtc_y Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83747 Signed-off-by: Gustavo Padovan Reviewed-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 90bb45f..78044bb 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -1080,7 +1080,7 @@ intel_commit_sprite_plane(struct drm_plane *plane, if (state->visible) { crtc_x = state->dst.x1; - crtc_y = state->dst.x2; + crtc_y = state->dst.y1; crtc_w = drm_rect_width(&state->dst); crtc_h = drm_rect_height(&state->dst); src_x = state->src.x1; -- cgit v0.10.2 From c04d016124f2290762d9535690b817855dd80fc1 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 12 Sep 2014 15:18:13 +0200 Subject: drm/i915: Drop get/put_pages for scratch page While discussing/reviewing __GFP_MOVEABLE behaviour and interactions with our various page allocations on irc Chris brought up that the scratch page isn't allocated as moveable, but we still grab/put a reference to lock it in place. Which is unecessary. So drop that. Acked-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 14f078c..a234446 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -1752,7 +1752,6 @@ static int setup_scratch_page(struct drm_device *dev) page = alloc_page(GFP_KERNEL | GFP_DMA32 | __GFP_ZERO); if (page == NULL) return -ENOMEM; - get_page(page); set_pages_uc(page, 1); #ifdef CONFIG_INTEL_IOMMU @@ -1777,7 +1776,6 @@ static void teardown_scratch_page(struct drm_device *dev) set_pages_wb(page, 1); pci_unmap_page(dev->pdev, dev_priv->gtt.base.scratch.addr, PAGE_SIZE, PCI_DMA_BIDIRECTIONAL); - put_page(page); __free_page(page); } -- cgit v0.10.2 From f77c44b902a032718e2431a5fc9211ef0b0282d3 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 12 Sep 2014 15:20:24 +0200 Subject: agp/intel-gtt: Remove get/put_pages If a page isn't allocated as __GFP_MOVEABLE it won't move around, so no need to grab a reference to lock it into place. Discovered while reviewing page allocation handling in i915 gem. Acked-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/char/agp/intel-gtt.c b/drivers/char/agp/intel-gtt.c index 9a024f8..f333482 100644 --- a/drivers/char/agp/intel-gtt.c +++ b/drivers/char/agp/intel-gtt.c @@ -153,7 +153,6 @@ static struct page *i8xx_alloc_pages(void) __free_pages(page, 2); return NULL; } - get_page(page); atomic_inc(&agp_bridge->current_memory_agp); return page; } @@ -164,7 +163,6 @@ static void i8xx_destroy_pages(struct page *page) return; set_pages_wb(page, 4); - put_page(page); __free_pages(page, 2); atomic_dec(&agp_bridge->current_memory_agp); } @@ -300,7 +298,6 @@ static int intel_gtt_setup_scratch_page(void) page = alloc_page(GFP_KERNEL | GFP_DMA32 | __GFP_ZERO); if (page == NULL) return -ENOMEM; - get_page(page); set_pages_uc(page, 1); if (intel_private.needs_dmar) { @@ -560,7 +557,6 @@ static void intel_gtt_teardown_scratch_page(void) set_pages_wb(intel_private.scratch_page, 1); pci_unmap_page(intel_private.pcidev, intel_private.scratch_page_dma, PAGE_SIZE, PCI_DMA_BIDIRECTIONAL); - put_page(intel_private.scratch_page); __free_page(intel_private.scratch_page); } -- cgit v0.10.2 From 99d1f3878baa86d5ce9e275ae9568fd903616f54 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 12 Sep 2014 20:53:32 +0300 Subject: drm/i915: Move the cursor_base setup to i{845, 9xx}_update_cursor() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit To make the code a bit more undestandable move the intel_crtc->cursor_base assignment into the low level update cursor routines. That's were we compare the current value with the new one so immediately seeing that it gets assigned only afterwards helps one to understand that it gets assigned only after the comparison. Signed-off-by: Ville Syrjälä Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 1556c70..fbfdaf4 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8258,8 +8258,10 @@ static void i845_update_cursor(struct drm_crtc *crtc, u32 base) intel_crtc->cursor_cntl = 0; } - if (intel_crtc->cursor_base != base) + if (intel_crtc->cursor_base != base) { I915_WRITE(_CURABASE, base); + intel_crtc->cursor_base = base; + } if (intel_crtc->cursor_size != size) { I915_WRITE(CURSIZE, size); @@ -8312,6 +8314,8 @@ static void i9xx_update_cursor(struct drm_crtc *crtc, u32 base) /* and commit changes on next vblank */ I915_WRITE(CURBASE(pipe), base); POSTING_READ(CURBASE(pipe)); + + intel_crtc->cursor_base = base; } /* If no-part of the cursor is visible on the framebuffer, then the GPU may hang... */ @@ -8362,7 +8366,6 @@ static void intel_crtc_update_cursor(struct drm_crtc *crtc, i845_update_cursor(crtc, base); else i9xx_update_cursor(crtc, base); - intel_crtc->cursor_base = base; } static bool cursor_size_ok(struct drm_device *dev, -- cgit v0.10.2 From 47bf17a7d1676f5f0e5c8d45f0fd0c5050add78f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 12 Sep 2014 20:53:33 +0300 Subject: drm/i915: Only set CURSOR_PIPE_CSC_ENABLE when cursor is enabled MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It seems cleaner if we keep CURCNTR at 0 when the cursor is disabled, so don't set the CURSOR_PIPE_CSC_ENABLE bit unless the cursor is enabled. Signed-off-by: Ville Syrjälä Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index fbfdaf4..007f465 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8301,9 +8301,10 @@ static void i9xx_update_cursor(struct drm_crtc *crtc, u32 base) return; } cntl |= pipe << 28; /* Connect to correct pipe */ + + if (IS_HASWELL(dev) || IS_BROADWELL(dev)) + cntl |= CURSOR_PIPE_CSC_ENABLE; } - if (IS_HASWELL(dev) || IS_BROADWELL(dev)) - cntl |= CURSOR_PIPE_CSC_ENABLE; if (intel_crtc->cursor_cntl != cntl) { I915_WRITE(CURCNTR(pipe), cntl); -- cgit v0.10.2 From 950eabaf5a87257040e0c207be09487954113f54 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Mon, 8 Sep 2014 15:21:09 +0300 Subject: drm/i915: vlv: fix display IRQ enable/disable We want to enable/disable display IRQs only if global i915 IRQs are enabled. To check the latter it's not enough to consult the DRM dev->irq_enabled flag, since runtime PM can disable/enable IRQs and it won't adjust this flag only the i915 specific dev_priv->pm._irqs_disabled flag. Fix this by using the proper intel_irqs_enabled() helper instead. Fortunately this didn't cause an actual problem since even if we enabled display IRQs too early (before enabling global i915 IRQs) the VLV_MASTER_IER would still be clear masking all IRQs. This issue was caught by commit 920dd15a2b2fc60d054646a8a1ffd6aeb6090e05 Author: Daniel Vetter Date: Wed Aug 27 10:43:37 2014 +0200 drm/i915: WARN if interrupts aren't on in en/disable_pipestat Signed-off-by: Imre Deak Cc: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 4847ed5..d22f870 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3723,7 +3723,7 @@ void valleyview_enable_display_irqs(struct drm_i915_private *dev_priv) dev_priv->display_irqs_enabled = true; - if (dev_priv->dev->irq_enabled) + if (intel_irqs_enabled(dev_priv)) valleyview_display_irqs_install(dev_priv); } @@ -3736,7 +3736,7 @@ void valleyview_disable_display_irqs(struct drm_i915_private *dev_priv) dev_priv->display_irqs_enabled = false; - if (dev_priv->dev->irq_enabled) + if (intel_irqs_enabled(dev_priv)) valleyview_display_irqs_uninstall(dev_priv); } -- cgit v0.10.2 From 7cd512f1520f85bf8e45f75b82fece58f0265cec Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 11:38:57 +0200 Subject: drm/i915: Fix irq checks in ring->irq_get/put functions Yet another place that wasn't properly transformed when implementing SOix. While at it convert the checks to WARN_ON on gen5+ (since we don't have UMS potentially doing stupid things on those platforms). And also add the corresponding checks to the put functions (again with a WARN_ON) for gen5+. v2: Drop the WARNINGS in the irq_put functions (including the existing one for vebox), Chris convinced me that they're not that terribly useful. v3: Don't forget about execlist code. Cc: Imre Deak Cc: Jesse Barnes Cc: "Volkin, Bradley D" Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index bafd38b..803fc38 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1063,7 +1063,7 @@ static bool gen8_logical_ring_get_irq(struct intel_engine_cs *ring) struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; - if (!dev->irq_enabled) + if (WARN_ON(!intel_irqs_enabled(dev_priv))) return false; spin_lock_irqsave(&dev_priv->irq_lock, flags); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 25795f2..922d6bc 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1197,7 +1197,7 @@ gen5_ring_get_irq(struct intel_engine_cs *ring) struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; - if (!dev->irq_enabled) + if (WARN_ON(!intel_irqs_enabled(dev_priv))) return false; spin_lock_irqsave(&dev_priv->irq_lock, flags); @@ -1228,7 +1228,7 @@ i9xx_ring_get_irq(struct intel_engine_cs *ring) struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; - if (!dev->irq_enabled) + if (!intel_irqs_enabled(dev_priv)) return false; spin_lock_irqsave(&dev_priv->irq_lock, flags); @@ -1265,7 +1265,7 @@ i8xx_ring_get_irq(struct intel_engine_cs *ring) struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; - if (!dev->irq_enabled) + if (!intel_irqs_enabled(dev_priv)) return false; spin_lock_irqsave(&dev_priv->irq_lock, flags); @@ -1399,8 +1399,8 @@ gen6_ring_get_irq(struct intel_engine_cs *ring) struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; - if (!dev->irq_enabled) - return false; + if (WARN_ON(!intel_irqs_enabled(dev_priv))) + return false; spin_lock_irqsave(&dev_priv->irq_lock, flags); if (ring->irq_refcount++ == 0) { @@ -1442,7 +1442,7 @@ hsw_vebox_get_irq(struct intel_engine_cs *ring) struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; - if (!dev->irq_enabled) + if (WARN_ON(!intel_irqs_enabled(dev_priv))) return false; spin_lock_irqsave(&dev_priv->irq_lock, flags); @@ -1462,9 +1462,6 @@ hsw_vebox_put_irq(struct intel_engine_cs *ring) struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; - if (!dev->irq_enabled) - return; - spin_lock_irqsave(&dev_priv->irq_lock, flags); if (--ring->irq_refcount == 0) { I915_WRITE_IMR(ring, ~0); @@ -1480,7 +1477,7 @@ gen8_ring_get_irq(struct intel_engine_cs *ring) struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; - if (!dev->irq_enabled) + if (WARN_ON(!intel_irqs_enabled(dev_priv))) return false; spin_lock_irqsave(&dev_priv->irq_lock, flags); -- cgit v0.10.2 From 4dc49272bdfc11cdefb41175771ee811e363dfc7 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 15 Sep 2014 16:59:28 +0300 Subject: drm/i915/bios: add missing __packed to structs used for reading vbt This does not seem to make a difference for the structs in question, but document the intent. v2: also pack union child_device_config (Daniel) Signed-off-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_bios.h b/drivers/gpu/drm/i915/intel_bios.h index 905999b..7603765 100644 --- a/drivers/gpu/drm/i915/intel_bios.h +++ b/drivers/gpu/drm/i915/intel_bios.h @@ -46,7 +46,7 @@ struct bdb_header { u16 version; /**< decimal */ u16 header_size; /**< in bytes */ u16 bdb_size; /**< in bytes */ -}; +} __packed; /* strictly speaking, this is a "skip" block, but it has interesting info */ struct vbios_data { @@ -252,7 +252,7 @@ union child_device_config { /* This one should also be safe to use anywhere, even without version * checks. */ struct common_child_dev_config common; -}; +} __packed; struct bdb_general_definitions { /* DDC GPIO */ @@ -888,12 +888,12 @@ struct mipi_pps_data { u16 bl_disable_delay; u16 panel_off_delay; u16 panel_power_cycle_delay; -}; +} __packed; struct bdb_mipi_config { struct mipi_config config[MAX_MIPI_CONFIGURATIONS]; struct mipi_pps_data pps[MAX_MIPI_CONFIGURATIONS]; -}; +} __packed; /* Block 53 contains MIPI sequences as needed by the panel * for enabling it. This block can be variable in size and @@ -902,7 +902,7 @@ struct bdb_mipi_config { struct bdb_mipi_sequence { u8 version; u8 data[0]; -}; +} __packed; /* MIPI Sequnece Block definitions */ enum mipi_seq { -- cgit v0.10.2 From 07f11d49f120dfffb5b8603dd3863e903d50c456 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:35:09 +0200 Subject: drm/i915: Convert backlight_lock to a mutex Originally the irq safe spinlock was required because of asle interrupts. But since commit 91a60f20712179e56b7a6c3d332a5f6f9a54aa11 Author: Jani Nikula Date: Thu Oct 31 18:55:48 2013 +0200 drm/i915: move opregion asle request handling to a work queue there's no need for this any more. So switch to the simpler mutex. v2: Cite the right commit, spotted by Jani. Cc: Jani Nikula Reviewed-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 1403b01..0bc1583 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1614,7 +1614,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) spin_lock_init(&dev_priv->irq_lock); spin_lock_init(&dev_priv->gpu_error.lock); - spin_lock_init(&dev_priv->backlight_lock); + mutex_init(&dev_priv->backlight_lock); spin_lock_init(&dev_priv->uncore.lock); spin_lock_init(&dev_priv->mm.object_stat_lock); spin_lock_init(&dev_priv->mmio_flip_lock); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 17dfce0..07dafa2 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1528,7 +1528,7 @@ struct drm_i915_private { struct intel_overlay *overlay; /* backlight registers and fields in struct intel_panel */ - spinlock_t backlight_lock; + struct mutex backlight_lock; /* LVDS info */ bool no_aux_handshake; diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c index 1878447..f17ada3 100644 --- a/drivers/gpu/drm/i915/intel_panel.c +++ b/drivers/gpu/drm/i915/intel_panel.c @@ -538,14 +538,13 @@ static u32 intel_panel_get_backlight(struct intel_connector *connector) struct drm_device *dev = connector->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; u32 val; - unsigned long flags; - spin_lock_irqsave(&dev_priv->backlight_lock, flags); + mutex_lock(&dev_priv->backlight_lock); val = dev_priv->display.get_backlight(connector); val = intel_panel_compute_brightness(connector, val); - spin_unlock_irqrestore(&dev_priv->backlight_lock, flags); + mutex_unlock(&dev_priv->backlight_lock); DRM_DEBUG_DRIVER("get backlight PWM = %d\n", val); return val; @@ -629,12 +628,11 @@ static void intel_panel_set_backlight(struct intel_connector *connector, struct intel_panel *panel = &connector->panel; enum pipe pipe = intel_get_pipe_from_connector(connector); u32 hw_level; - unsigned long flags; if (!panel->backlight.present || pipe == INVALID_PIPE) return; - spin_lock_irqsave(&dev_priv->backlight_lock, flags); + mutex_lock(&dev_priv->backlight_lock); WARN_ON(panel->backlight.max == 0); @@ -644,7 +642,7 @@ static void intel_panel_set_backlight(struct intel_connector *connector, if (panel->backlight.enabled) intel_panel_actually_set_backlight(connector, hw_level); - spin_unlock_irqrestore(&dev_priv->backlight_lock, flags); + mutex_unlock(&dev_priv->backlight_lock); } /* set backlight brightness to level in range [0..max], assuming hw min is @@ -658,12 +656,11 @@ void intel_panel_set_backlight_acpi(struct intel_connector *connector, struct intel_panel *panel = &connector->panel; enum pipe pipe = intel_get_pipe_from_connector(connector); u32 hw_level; - unsigned long flags; if (!panel->backlight.present || pipe == INVALID_PIPE) return; - spin_lock_irqsave(&dev_priv->backlight_lock, flags); + mutex_lock(&dev_priv->backlight_lock); WARN_ON(panel->backlight.max == 0); @@ -679,7 +676,7 @@ void intel_panel_set_backlight_acpi(struct intel_connector *connector, if (panel->backlight.enabled) intel_panel_actually_set_backlight(connector, hw_level); - spin_unlock_irqrestore(&dev_priv->backlight_lock, flags); + mutex_unlock(&dev_priv->backlight_lock); } static void pch_disable_backlight(struct intel_connector *connector) @@ -733,7 +730,6 @@ void intel_panel_disable_backlight(struct intel_connector *connector) struct drm_i915_private *dev_priv = dev->dev_private; struct intel_panel *panel = &connector->panel; enum pipe pipe = intel_get_pipe_from_connector(connector); - unsigned long flags; if (!panel->backlight.present || pipe == INVALID_PIPE) return; @@ -749,14 +745,14 @@ void intel_panel_disable_backlight(struct intel_connector *connector) return; } - spin_lock_irqsave(&dev_priv->backlight_lock, flags); + mutex_lock(&dev_priv->backlight_lock); if (panel->backlight.device) panel->backlight.device->props.power = FB_BLANK_POWERDOWN; panel->backlight.enabled = false; dev_priv->display.disable_backlight(connector); - spin_unlock_irqrestore(&dev_priv->backlight_lock, flags); + mutex_unlock(&dev_priv->backlight_lock); } static void bdw_enable_backlight(struct intel_connector *connector) @@ -937,14 +933,13 @@ void intel_panel_enable_backlight(struct intel_connector *connector) struct drm_i915_private *dev_priv = dev->dev_private; struct intel_panel *panel = &connector->panel; enum pipe pipe = intel_get_pipe_from_connector(connector); - unsigned long flags; if (!panel->backlight.present || pipe == INVALID_PIPE) return; DRM_DEBUG_KMS("pipe %c\n", pipe_name(pipe)); - spin_lock_irqsave(&dev_priv->backlight_lock, flags); + mutex_lock(&dev_priv->backlight_lock); WARN_ON(panel->backlight.max == 0); @@ -962,7 +957,7 @@ void intel_panel_enable_backlight(struct intel_connector *connector) if (panel->backlight.device) panel->backlight.device->props.power = FB_BLANK_UNBLANK; - spin_unlock_irqrestore(&dev_priv->backlight_lock, flags); + mutex_unlock(&dev_priv->backlight_lock); } #if IS_ENABLED(CONFIG_BACKLIGHT_CLASS_DEVICE) @@ -1267,7 +1262,6 @@ int intel_panel_setup_backlight(struct drm_connector *connector) struct drm_i915_private *dev_priv = dev->dev_private; struct intel_connector *intel_connector = to_intel_connector(connector); struct intel_panel *panel = &intel_connector->panel; - unsigned long flags; int ret; if (!dev_priv->vbt.backlight.present) { @@ -1280,9 +1274,9 @@ int intel_panel_setup_backlight(struct drm_connector *connector) } /* set level and max in panel struct */ - spin_lock_irqsave(&dev_priv->backlight_lock, flags); + mutex_lock(&dev_priv->backlight_lock); ret = dev_priv->display.setup_backlight(intel_connector); - spin_unlock_irqrestore(&dev_priv->backlight_lock, flags); + mutex_unlock(&dev_priv->backlight_lock); if (ret) { DRM_DEBUG_KMS("failed to setup backlight for connector %s\n", -- cgit v0.10.2 From 51e31d49c89055299e34b8f44d13f70e19aaaad1 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 12:36:02 +0200 Subject: drm/i915: Use generic vblank wait MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This has the upside that it will no longer steal interrupts from the interrupt handler on pre-g4x. Furthermore this will now scream properly on all platforms if we don't have hw counters enabled. v2: Adjust to the new names. Cc: Ville Syrjälä Reviewed-by: Matt Roper Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 007f465..474e615 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -892,18 +892,6 @@ enum transcoder intel_pipe_to_cpu_transcoder(struct drm_i915_private *dev_priv, return intel_crtc->config.cpu_transcoder; } -static void g4x_wait_for_vblank(struct drm_device *dev, int pipe) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - u32 frame, frame_reg = PIPE_FRMCOUNT_GM45(pipe); - - frame = I915_READ(frame_reg); - - if (wait_for(I915_READ_NOTRACE(frame_reg) != frame, 50)) - WARN(1, "vblank wait on pipe %c timed out\n", - pipe_name(pipe)); -} - /** * intel_wait_for_vblank - wait for vblank on a given pipe * @dev: drm device @@ -914,36 +902,7 @@ static void g4x_wait_for_vblank(struct drm_device *dev, int pipe) */ void intel_wait_for_vblank(struct drm_device *dev, int pipe) { - struct drm_i915_private *dev_priv = dev->dev_private; - int pipestat_reg = PIPESTAT(pipe); - - if (IS_G4X(dev) || INTEL_INFO(dev)->gen >= 5) { - g4x_wait_for_vblank(dev, pipe); - return; - } - - /* Clear existing vblank status. Note this will clear any other - * sticky status fields as well. - * - * This races with i915_driver_irq_handler() with the result - * that either function could miss a vblank event. Here it is not - * fatal, as we will either wait upon the next vblank interrupt or - * timeout. Generally speaking intel_wait_for_vblank() is only - * called during modeset at which time the GPU should be idle and - * should *not* be performing page flips and thus not waiting on - * vblanks... - * Currently, the result of us stealing a vblank from the irq - * handler is that a single frame will be skipped during swapbuffers. - */ - I915_WRITE(pipestat_reg, - I915_READ(pipestat_reg) | PIPE_VBLANK_INTERRUPT_STATUS); - - /* Wait for vblank interrupt bit to set */ - if (wait_for(I915_READ(pipestat_reg) & - PIPE_VBLANK_INTERRUPT_STATUS, - 50)) - DRM_DEBUG_KMS("vblank wait on pipe %c timed out\n", - pipe_name(pipe)); + drm_wait_one_vblank(dev, pipe); } static bool pipe_dsl_stopped(struct drm_device *dev, enum pipe pipe) -- cgit v0.10.2 From 4f905cf97fb09b34193c7494bb27e49a56de934b Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:12:21 +0200 Subject: drm/i915: static inline for intel_wait_for_vblank Requested by Chris, and also requested to keep it since it's a more accurate name in his opinion. Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 474e615..ef1d8fb 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -892,19 +892,6 @@ enum transcoder intel_pipe_to_cpu_transcoder(struct drm_i915_private *dev_priv, return intel_crtc->config.cpu_transcoder; } -/** - * intel_wait_for_vblank - wait for vblank on a given pipe - * @dev: drm device - * @pipe: pipe to wait for - * - * Wait for vblank to occur on a given pipe. Needed for various bits of - * mode setting code. - */ -void intel_wait_for_vblank(struct drm_device *dev, int pipe) -{ - drm_wait_one_vblank(dev, pipe); -} - static bool pipe_dsl_stopped(struct drm_device *dev, enum pipe pipe) { struct drm_i915_private *dev_priv = dev->dev_private; diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index dd5e0f1..1b72c15 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -853,7 +853,11 @@ int intel_get_pipe_from_crtc_id(struct drm_device *dev, void *data, struct drm_file *file_priv); enum transcoder intel_pipe_to_cpu_transcoder(struct drm_i915_private *dev_priv, enum pipe pipe); -void intel_wait_for_vblank(struct drm_device *dev, int pipe); +static inline void +intel_wait_for_vblank(struct drm_device *dev, int pipe) +{ + drm_wait_one_vblank(dev, pipe); +} int ironlake_get_lanes_required(int target_clock, int link_bw, int bpp); void vlv_wait_port_ready(struct drm_i915_private *dev_priv, struct intel_digital_port *dport); -- cgit v0.10.2 From 4b3a9526fc3228e74011b88f58088336acd2c9e2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 14 Aug 2014 22:04:37 +0300 Subject: drm/i915: Move vblank enable earlier and disable later MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We changed to an interrupt based vblank wait (as opposed to polling) in: commit 44bd93a3d367913d883be6abba9a6e51a53c4e90 Author: Daniel Vetter Date: Fri Jul 25 23:36:44 2014 +0200 drm/i915: Use generic vblank wait However we already had vblank waits on the wrong side of drm_vblank_{on,off}() calls due to various workarounds, so now we get a warning more or less every time we do a modeset, and we fail to wait for the vblank like we should. Move the drm_vblank_{on,off}() calls back out from intel_crtc_{enable,disable}_planes() so that all of these vblank waits return to proper operation. Also move the cxsr wait a bit earlier so that we can keep the encoder disable after we've turned off vblanks. Moving stuff out from the plane enable/disable functions seems preferrable to moving the workaround stuff in since the workarounds are required only on specific platforms. While at it switch over to the drm_crtc_ variants of the vblank on/off functions. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82525 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82490 Cc: Daniel Vetter Signed-off-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ef1d8fb..365e04c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -3997,10 +3997,6 @@ static void intel_crtc_enable_planes(struct drm_crtc *crtc) struct intel_crtc *intel_crtc = to_intel_crtc(crtc); int pipe = intel_crtc->pipe; - assert_vblank_disabled(crtc); - - drm_vblank_on(dev, pipe); - intel_enable_primary_hw_plane(crtc->primary, crtc); intel_enable_planes(crtc); intel_crtc_update_cursor(crtc, true); @@ -4046,10 +4042,6 @@ static void intel_crtc_disable_planes(struct drm_crtc *crtc) * consider this a flip to a NULL plane. */ intel_frontbuffer_flip(dev, INTEL_FRONTBUFFER_ALL_MASK(pipe)); - - drm_vblank_off(dev, pipe); - - assert_vblank_disabled(crtc); } static void ironlake_crtc_enable(struct drm_crtc *crtc) @@ -4119,6 +4111,9 @@ static void ironlake_crtc_enable(struct drm_crtc *crtc) if (HAS_PCH_CPT(dev)) cpt_verify_modeset(dev, intel_crtc->pipe); + assert_vblank_disabled(crtc); + drm_crtc_vblank_on(crtc); + intel_crtc_enable_planes(crtc); } @@ -4226,6 +4221,9 @@ static void haswell_crtc_enable(struct drm_crtc *crtc) intel_opregion_notify_encoder(encoder, true); } + assert_vblank_disabled(crtc); + drm_crtc_vblank_on(crtc); + /* If we change the relative order between pipe/planes enabling, we need * to change the workaround. */ haswell_mode_set_planes_workaround(intel_crtc); @@ -4261,6 +4259,9 @@ static void ironlake_crtc_disable(struct drm_crtc *crtc) intel_crtc_disable_planes(crtc); + drm_crtc_vblank_off(crtc); + assert_vblank_disabled(crtc); + for_each_encoder_on_crtc(dev, crtc, encoder) encoder->disable(encoder); @@ -4323,6 +4324,9 @@ static void haswell_crtc_disable(struct drm_crtc *crtc) intel_crtc_disable_planes(crtc); + drm_crtc_vblank_off(crtc); + assert_vblank_disabled(crtc); + for_each_encoder_on_crtc(dev, crtc, encoder) { intel_opregion_notify_encoder(encoder, false); encoder->disable(encoder); @@ -4789,6 +4793,9 @@ static void valleyview_crtc_enable(struct drm_crtc *crtc) for_each_encoder_on_crtc(dev, crtc, encoder) encoder->enable(encoder); + assert_vblank_disabled(crtc); + drm_crtc_vblank_on(crtc); + intel_crtc_enable_planes(crtc); /* Underruns don't raise interrupts, so check manually. */ @@ -4846,6 +4853,9 @@ static void i9xx_crtc_enable(struct drm_crtc *crtc) for_each_encoder_on_crtc(dev, crtc, encoder) encoder->enable(encoder); + assert_vblank_disabled(crtc); + drm_crtc_vblank_on(crtc); + intel_crtc_enable_planes(crtc); /* @@ -4909,9 +4919,6 @@ static void i9xx_crtc_disable(struct drm_crtc *crtc) intel_set_memory_cxsr(dev_priv, false); intel_crtc_disable_planes(crtc); - for_each_encoder_on_crtc(dev, crtc, encoder) - encoder->disable(encoder); - /* * On gen2 planes are double buffered but the pipe isn't, so we must * wait for planes to fully turn off before disabling the pipe. @@ -4920,6 +4927,12 @@ static void i9xx_crtc_disable(struct drm_crtc *crtc) */ intel_wait_for_vblank(dev, pipe); + drm_crtc_vblank_off(crtc); + assert_vblank_disabled(crtc); + + for_each_encoder_on_crtc(dev, crtc, encoder) + encoder->disable(encoder); + intel_disable_pipe(intel_crtc); i9xx_pfit_disable(intel_crtc); -- cgit v0.10.2 From 5e2d7afcfefa7cda3f341ce78523bb9b0acdbacf Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:22 +0200 Subject: drm/i915: Clarify event_lock locking, process context It's good practice to use the more specific versions for irq save spinlocks both as executable documentation and to enforce saner design. The _irqsave version really should only be used if the calling context is unknown and there's a good reason to call a function from all kinds of places. This is the first step whice replaces all occurances of _irqsave in process context with the simpler irq disable/enable variants. We don't have any funky spinlock nesting going on, especially since the event_lock is the outermost of the irq/vblank related spinlocks. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 063b448..0ba5c71 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -516,7 +516,6 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data) struct drm_info_node *node = m->private; struct drm_device *dev = node->minor->dev; struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long flags; struct intel_crtc *crtc; int ret; @@ -529,7 +528,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data) const char plane = plane_name(crtc->plane); struct intel_unpin_work *work; - spin_lock_irqsave(&dev->event_lock, flags); + spin_lock_irq(&dev->event_lock); work = crtc->unpin_work; if (work == NULL) { seq_printf(m, "No flip due on pipe %c (plane %c)\n", @@ -575,7 +574,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data) seq_printf(m, "MMIO update completed? %d\n", addr == work->gtt_offset); } } - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock_irq(&dev->event_lock); } mutex_unlock(&dev->struct_mutex); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 365e04c..624d9bf 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2711,16 +2711,15 @@ static bool intel_crtc_has_pending_flip(struct drm_crtc *crtc) struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - unsigned long flags; bool pending; if (i915_reset_in_progress(&dev_priv->gpu_error) || intel_crtc->reset_counter != atomic_read(&dev_priv->gpu_error.reset_counter)) return false; - spin_lock_irqsave(&dev->event_lock, flags); + spin_lock_irq(&dev->event_lock); pending = to_intel_crtc(crtc)->unpin_work != NULL; - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock_irq(&dev->event_lock); return pending; } @@ -3431,14 +3430,13 @@ void intel_crtc_wait_for_pending_flips(struct drm_crtc *crtc) !intel_crtc_has_pending_flip(crtc), 60*HZ) == 0)) { struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - unsigned long flags; - spin_lock_irqsave(&dev->event_lock, flags); + spin_lock_irq(&dev->event_lock); if (intel_crtc->unpin_work) { WARN_ONCE(1, "Removing stuck page flip\n"); page_flip_completed(intel_crtc); } - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock_irq(&dev->event_lock); } if (crtc->primary->fb) { @@ -9280,12 +9278,11 @@ static void intel_crtc_destroy(struct drm_crtc *crtc) struct intel_crtc *intel_crtc = to_intel_crtc(crtc); struct drm_device *dev = crtc->dev; struct intel_unpin_work *work; - unsigned long flags; - spin_lock_irqsave(&dev->event_lock, flags); + spin_lock_irq(&dev->event_lock); work = intel_crtc->unpin_work; intel_crtc->unpin_work = NULL; - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock_irq(&dev->event_lock); if (work) { cancel_work_sync(&work->work); @@ -9896,7 +9893,6 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, enum pipe pipe = intel_crtc->pipe; struct intel_unpin_work *work; struct intel_engine_cs *ring; - unsigned long flags; int ret; //trigger software GT busyness calculation @@ -9940,7 +9936,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, goto free_work; /* We borrow the event spin lock for protecting unpin_work */ - spin_lock_irqsave(&dev->event_lock, flags); + spin_lock_irq(&dev->event_lock); if (intel_crtc->unpin_work) { /* Before declaring the flip queue wedged, check if * the hardware completed the operation behind our backs. @@ -9950,7 +9946,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, page_flip_completed(intel_crtc); } else { DRM_DEBUG_DRIVER("flip queue: crtc already busy\n"); - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock_irq(&dev->event_lock); drm_crtc_vblank_put(crtc); kfree(work); @@ -9958,7 +9954,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, } } intel_crtc->unpin_work = work; - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock_irq(&dev->event_lock); if (atomic_read(&intel_crtc->unpin_work_count) >= 2) flush_workqueue(dev_priv->wq); @@ -10045,9 +10041,9 @@ cleanup_pending: mutex_unlock(&dev->struct_mutex); cleanup: - spin_lock_irqsave(&dev->event_lock, flags); + spin_lock_irq(&dev->event_lock); intel_crtc->unpin_work = NULL; - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock_irq(&dev->event_lock); drm_crtc_vblank_put(crtc); free_work: @@ -10058,9 +10054,9 @@ out_hang: intel_crtc_wait_for_pending_flips(crtc); ret = intel_pipe_set_base(crtc, crtc->x, crtc->y, fb); if (ret == 0 && event) { - spin_lock_irqsave(&dev->event_lock, flags); + spin_lock_irq(&dev->event_lock); drm_send_vblank_event(dev, pipe, event); - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock_irq(&dev->event_lock); } } return ret; @@ -13769,9 +13765,8 @@ void intel_modeset_preclose(struct drm_device *dev, struct drm_file *file) for_each_intel_crtc(dev, crtc) { struct intel_unpin_work *work; - unsigned long irqflags; - spin_lock_irqsave(&dev->event_lock, irqflags); + spin_lock_irq(&dev->event_lock); work = crtc->unpin_work; @@ -13781,6 +13776,6 @@ void intel_modeset_preclose(struct drm_device *dev, struct drm_file *file) work->event = NULL; } - spin_unlock_irqrestore(&dev->event_lock, irqflags); + spin_unlock_irq(&dev->event_lock); } } -- cgit v0.10.2 From f326038a29092534b59626f736a3c6e599bda017 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:23 +0200 Subject: drm/i915: Clarify event_lock locking, irq&mixed context Now we tackle the functions also called from interrupt handlers. - intel_check_page_flip is exclusively called from irq handlers, so a plain spin_lock is all we need. In i915_irq.c we have the convention to give all such functions an _irq_handler postfix, but that would look strange and als be a bit a misleading name. I've opted for a WARN_ON(!in_irq()) instead. - The other two places left are called both from interrupt handlers and from our reset work, so need the full irqsave dance. Annotate them with a short comment. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 624d9bf..0c6350c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9328,6 +9328,10 @@ static void do_intel_finish_page_flip(struct drm_device *dev, if (intel_crtc == NULL) return; + /* + * This is called both by irq handlers and the reset code (to complete + * lost pageflips) so needs the full irqsave spinlocks. + */ spin_lock_irqsave(&dev->event_lock, flags); work = intel_crtc->unpin_work; @@ -9409,7 +9413,12 @@ void intel_prepare_page_flip(struct drm_device *dev, int plane) to_intel_crtc(dev_priv->plane_to_crtc_mapping[plane]); unsigned long flags; - /* NB: An MMIO update of the plane base pointer will also + + /* + * This is called both by irq handlers and the reset code (to complete + * lost pageflips) so needs the full irqsave spinlocks. + * + * NB: An MMIO update of the plane base pointer will also * generate a page-flip completion irq, i.e. every modeset * is also accompanied by a spurious intel_prepare_page_flip(). */ @@ -9866,18 +9875,19 @@ void intel_check_page_flip(struct drm_device *dev, int pipe) struct drm_i915_private *dev_priv = dev->dev_private; struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pipe]; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - unsigned long flags; + + WARN_ON(!in_irq()); if (crtc == NULL) return; - spin_lock_irqsave(&dev->event_lock, flags); + spin_lock(&dev->event_lock); if (intel_crtc->unpin_work && __intel_pageflip_stall_check(dev, crtc)) { WARN_ONCE(1, "Kicking stuck page flip: queued at %d, now %d\n", intel_crtc->unpin_work->flip_queued_vblank, drm_vblank_count(dev, pipe)); page_flip_completed(intel_crtc); } - spin_unlock_irqrestore(&dev->event_lock, flags); + spin_unlock(&dev->event_lock); } static int intel_crtc_page_flip(struct drm_crtc *crtc, -- cgit v0.10.2 From 5b254c597871fa638a1ba0621f545e303090e970 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:24 +0200 Subject: drm/i915: Clarify gpu_error.lock locking i915_capture_error_state can be called from all kinds of contexts, so needs the full irqsave dance. But the other two places to grab and release the error state are only called from process context. So simplify them to the plaine _irq spinlock versions to clarify the locking semantics. Cc: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 2c87a79..386e45db 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -1326,13 +1326,12 @@ void i915_error_state_get(struct drm_device *dev, struct i915_error_state_file_priv *error_priv) { struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long flags; - spin_lock_irqsave(&dev_priv->gpu_error.lock, flags); + spin_lock_irq(&dev_priv->gpu_error.lock); error_priv->error = dev_priv->gpu_error.first_error; if (error_priv->error) kref_get(&error_priv->error->ref); - spin_unlock_irqrestore(&dev_priv->gpu_error.lock, flags); + spin_unlock_irq(&dev_priv->gpu_error.lock); } @@ -1346,12 +1345,11 @@ void i915_destroy_error_state(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_error_state *error; - unsigned long flags; - spin_lock_irqsave(&dev_priv->gpu_error.lock, flags); + spin_lock_irq(&dev_priv->gpu_error.lock); error = dev_priv->gpu_error.first_error; dev_priv->gpu_error.first_error = NULL; - spin_unlock_irqrestore(&dev_priv->gpu_error.lock, flags); + spin_unlock_irq(&dev_priv->gpu_error.lock); if (error) kref_put(&error->ref, i915_error_state_free); -- cgit v0.10.2 From 2795aa4864c9391c3107ecd91d663e58ff43dbfc Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:25 +0200 Subject: drm/i915: Clarify irq_lock locking, intel_tv_detect ->detect callbacks are only ever called from process context, and there's no fancy nesting going on here. So plain _irq spinlock variants is what we want. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_tv.c b/drivers/gpu/drm/i915/intel_tv.c index c14341c..6f5f59b 100644 --- a/drivers/gpu/drm/i915/intel_tv.c +++ b/drivers/gpu/drm/i915/intel_tv.c @@ -1182,18 +1182,17 @@ intel_tv_detect_type(struct intel_tv *intel_tv, struct intel_crtc *intel_crtc = to_intel_crtc(crtc); struct drm_device *dev = encoder->dev; struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long irqflags; u32 tv_ctl, save_tv_ctl; u32 tv_dac, save_tv_dac; int type; /* Disable TV interrupts around load detect or we'll recurse */ if (connector->polled & DRM_CONNECTOR_POLL_HPD) { - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); i915_disable_pipestat(dev_priv, 0, PIPE_HOTPLUG_INTERRUPT_STATUS | PIPE_HOTPLUG_TV_INTERRUPT_STATUS); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); } save_tv_dac = tv_dac = I915_READ(TV_DAC); @@ -1266,11 +1265,11 @@ intel_tv_detect_type(struct intel_tv *intel_tv, /* Restore interrupt config */ if (connector->polled & DRM_CONNECTOR_POLL_HPD) { - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); i915_enable_pipestat(dev_priv, 0, PIPE_HOTPLUG_INTERRUPT_STATUS | PIPE_HOTPLUG_TV_INTERRUPT_STATUS); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); } return type; -- cgit v0.10.2 From 4cb21832280643f4ca96cad0edbef6f45baacf8c Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:26 +0200 Subject: drm/i915: Clarify irq_lock locking, work functions Work functions are in process context, so plain _irq spinlock variants is all we need. The hpd reenable work didn't follow the _work/_work_func postfix naming scheme, so adjust that while at it. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index d22f870..4906823 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1096,18 +1096,17 @@ static void i915_digport_work_func(struct work_struct *work) { struct drm_i915_private *dev_priv = container_of(work, struct drm_i915_private, dig_port_work); - unsigned long irqflags; u32 long_port_mask, short_port_mask; struct intel_digital_port *intel_dig_port; int i, ret; u32 old_bits = 0; - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); long_port_mask = dev_priv->long_hpd_port_mask; dev_priv->long_hpd_port_mask = 0; short_port_mask = dev_priv->short_hpd_port_mask; dev_priv->short_hpd_port_mask = 0; - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); for (i = 0; i < I915_MAX_PORTS; i++) { bool valid = false; @@ -1132,9 +1131,9 @@ static void i915_digport_work_func(struct work_struct *work) } if (old_bits) { - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); dev_priv->hpd_event_bits |= old_bits; - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); schedule_work(&dev_priv->hotplug_work); } } @@ -1153,7 +1152,6 @@ static void i915_hotplug_work_func(struct work_struct *work) struct intel_connector *intel_connector; struct intel_encoder *intel_encoder; struct drm_connector *connector; - unsigned long irqflags; bool hpd_disabled = false; bool changed = false; u32 hpd_event_bits; @@ -1161,7 +1159,7 @@ static void i915_hotplug_work_func(struct work_struct *work) mutex_lock(&mode_config->mutex); DRM_DEBUG_KMS("running encoder hotplug functions\n"); - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); hpd_event_bits = dev_priv->hpd_event_bits; dev_priv->hpd_event_bits = 0; @@ -1195,7 +1193,7 @@ static void i915_hotplug_work_func(struct work_struct *work) msecs_to_jiffies(I915_REENABLE_HOTPLUG_DELAY)); } - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); list_for_each_entry(connector, &mode_config->connector_list, head) { intel_connector = to_intel_connector(connector); @@ -1490,7 +1488,6 @@ static void ivybridge_parity_work(struct work_struct *work) u32 error_status, row, bank, subbank; char *parity_event[6]; uint32_t misccpctl; - unsigned long flags; uint8_t slice = 0; /* We must turn off DOP level clock gating to access the L3 registers. @@ -1549,9 +1546,9 @@ static void ivybridge_parity_work(struct work_struct *work) out: WARN_ON(dev_priv->l3_parity.which_slice); - spin_lock_irqsave(&dev_priv->irq_lock, flags); + spin_lock_irq(&dev_priv->irq_lock); gen5_enable_gt_irq(dev_priv, GT_PARITY_ERROR(dev_priv->dev)); - spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + spin_unlock_irq(&dev_priv->irq_lock); mutex_unlock(&dev_priv->dev->struct_mutex); } @@ -4606,19 +4603,18 @@ static void i965_irq_uninstall(struct drm_device * dev) I915_WRITE(IIR, I915_READ(IIR)); } -static void intel_hpd_irq_reenable(struct work_struct *work) +static void intel_hpd_irq_reenable_work(struct work_struct *work) { struct drm_i915_private *dev_priv = container_of(work, typeof(*dev_priv), hotplug_reenable_work.work); struct drm_device *dev = dev_priv->dev; struct drm_mode_config *mode_config = &dev->mode_config; - unsigned long irqflags; int i; intel_runtime_pm_get(dev_priv); - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); for (i = (HPD_NONE + 1); i < HPD_NUM_PINS; i++) { struct drm_connector *connector; @@ -4642,7 +4638,7 @@ static void intel_hpd_irq_reenable(struct work_struct *work) } if (dev_priv->display.hpd_irq_setup) dev_priv->display.hpd_irq_setup(dev); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); intel_runtime_pm_put(dev_priv); } @@ -4668,7 +4664,7 @@ void intel_irq_init(struct drm_device *dev) i915_hangcheck_elapsed, (unsigned long) dev); INIT_DELAYED_WORK(&dev_priv->hotplug_reenable_work, - intel_hpd_irq_reenable); + intel_hpd_irq_reenable_work); pm_qos_add_request(&dev_priv->pm_qos, PM_QOS_CPU_DMA_LATENCY, PM_QOS_DEFAULT_VALUE); -- cgit v0.10.2 From d62074358de148da197a90262551217f26937b5e Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:27 +0200 Subject: drm/i915: Clarify irq_lock locking, interrupt install/uninstall All the interrupt setup/teardown hooks are always run from plain process context. So again just the _irq variant is good enough. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 4906823..a829619 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3603,7 +3603,6 @@ static void gen5_gt_irq_postinstall(struct drm_device *dev) static int ironlake_irq_postinstall(struct drm_device *dev) { - unsigned long irqflags; struct drm_i915_private *dev_priv = dev->dev_private; u32 display_mask, extra_mask; @@ -3642,9 +3641,9 @@ static int ironlake_irq_postinstall(struct drm_device *dev) * spinlocking not required here for correctness since interrupt * setup is guaranteed to run in single-threaded context. But we * need it to make the assert_spin_locked happy. */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); ironlake_enable_display_irq(dev_priv, DE_PCU_EVENT); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); } return 0; @@ -3740,7 +3739,6 @@ void valleyview_disable_display_irqs(struct drm_i915_private *dev_priv) static int valleyview_irq_postinstall(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long irqflags; dev_priv->irq_mask = ~0; @@ -3754,10 +3752,10 @@ static int valleyview_irq_postinstall(struct drm_device *dev) /* Interrupt setup is already guaranteed to be single-threaded, this is * just to make the assert_spin_locked check happy. */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); if (dev_priv->display_irqs_enabled) valleyview_display_irqs_install(dev_priv); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); I915_WRITE(VLV_IIR, 0xffffffff); I915_WRITE(VLV_IIR, 0xffffffff); @@ -3848,7 +3846,6 @@ static int cherryview_irq_postinstall(struct drm_device *dev) I915_DISPLAY_PIPE_C_EVENT_INTERRUPT; u32 pipestat_enable = PLANE_FLIP_DONE_INT_STATUS_VLV | PIPE_CRC_DONE_INTERRUPT_STATUS; - unsigned long irqflags; int pipe; /* @@ -3860,11 +3857,11 @@ static int cherryview_irq_postinstall(struct drm_device *dev) for_each_pipe(dev_priv, pipe) I915_WRITE(PIPESTAT(pipe), 0xffff); - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); i915_enable_pipestat(dev_priv, PIPE_A, PIPE_GMBUS_INTERRUPT_STATUS); for_each_pipe(dev_priv, pipe) i915_enable_pipestat(dev_priv, pipe, pipestat_enable); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); I915_WRITE(VLV_IIR, 0xffffffff); I915_WRITE(VLV_IMR, dev_priv->irq_mask); @@ -3891,7 +3888,6 @@ static void gen8_irq_uninstall(struct drm_device *dev) static void valleyview_irq_uninstall(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long irqflags; int pipe; if (!dev_priv) @@ -3906,10 +3902,12 @@ static void valleyview_irq_uninstall(struct drm_device *dev) I915_WRITE(PORT_HOTPLUG_EN, 0); I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + /* Interrupt setup is already guaranteed to be single-threaded, this is + * just to make the assert_spin_locked check happy. */ + spin_lock_irq(&dev_priv->irq_lock); if (dev_priv->display_irqs_enabled) valleyview_display_irqs_uninstall(dev_priv); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); dev_priv->irq_mask = 0; @@ -3995,7 +3993,6 @@ static void i8xx_irq_preinstall(struct drm_device * dev) static int i8xx_irq_postinstall(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long irqflags; I915_WRITE16(EMR, ~(I915_ERROR_PAGE_TABLE | I915_ERROR_MEMORY_REFRESH)); @@ -4018,10 +4015,10 @@ static int i8xx_irq_postinstall(struct drm_device *dev) /* Interrupt setup is already guaranteed to be single-threaded, this is * just to make the assert_spin_locked check happy. */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); i915_enable_pipestat(dev_priv, PIPE_A, PIPE_CRC_DONE_INTERRUPT_STATUS); i915_enable_pipestat(dev_priv, PIPE_B, PIPE_CRC_DONE_INTERRUPT_STATUS); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); return 0; } @@ -4168,7 +4165,6 @@ static int i915_irq_postinstall(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; u32 enable_mask; - unsigned long irqflags; I915_WRITE(EMR, ~(I915_ERROR_PAGE_TABLE | I915_ERROR_MEMORY_REFRESH)); @@ -4206,10 +4202,10 @@ static int i915_irq_postinstall(struct drm_device *dev) /* Interrupt setup is already guaranteed to be single-threaded, this is * just to make the assert_spin_locked check happy. */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); i915_enable_pipestat(dev_priv, PIPE_A, PIPE_CRC_DONE_INTERRUPT_STATUS); i915_enable_pipestat(dev_priv, PIPE_B, PIPE_CRC_DONE_INTERRUPT_STATUS); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); return 0; } @@ -4391,7 +4387,6 @@ static int i965_irq_postinstall(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; u32 enable_mask; u32 error_mask; - unsigned long irqflags; /* Unmask the interrupts that we always want on. */ dev_priv->irq_mask = ~(I915_ASLE_INTERRUPT | @@ -4412,11 +4407,11 @@ static int i965_irq_postinstall(struct drm_device *dev) /* Interrupt setup is already guaranteed to be single-threaded, this is * just to make the assert_spin_locked check happy. */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); i915_enable_pipestat(dev_priv, PIPE_A, PIPE_GMBUS_INTERRUPT_STATUS); i915_enable_pipestat(dev_priv, PIPE_A, PIPE_CRC_DONE_INTERRUPT_STATUS); i915_enable_pipestat(dev_priv, PIPE_B, PIPE_CRC_DONE_INTERRUPT_STATUS); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); /* * Enable some error detection, note the instruction error mask @@ -4756,7 +4751,6 @@ void intel_hpd_init(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; struct drm_mode_config *mode_config = &dev->mode_config; struct drm_connector *connector; - unsigned long irqflags; int i; for (i = 1; i < HPD_NUM_PINS; i++) { @@ -4774,10 +4768,10 @@ void intel_hpd_init(struct drm_device *dev) /* Interrupt setup is already guaranteed to be single-threaded, this is * just to make the assert_spin_locked checks happy. */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); if (dev_priv->display.hpd_irq_setup) dev_priv->display.hpd_irq_setup(dev); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); } /* Disable interrupts so we can allow runtime PM. */ -- cgit v0.10.2 From 222c7f51b074cc8bfbda5821ee7f875e51f841b5 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:28 +0200 Subject: drm/i915: Clarify irq_lock locking, irq handlers irq handlers always run with interrupts locally disabled, so plain spinlocks is all we need. I've also reviewed again that they all follow the _irq_handler postfix convention. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index a829619..6a4f389 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -4063,7 +4063,6 @@ static irqreturn_t i8xx_irq_handler(int irq, void *arg) struct drm_i915_private *dev_priv = dev->dev_private; u16 iir, new_iir; u32 pipe_stats[2]; - unsigned long irqflags; int pipe; u16 flip_mask = I915_DISPLAY_PLANE_A_FLIP_PENDING_INTERRUPT | @@ -4079,7 +4078,7 @@ static irqreturn_t i8xx_irq_handler(int irq, void *arg) * It doesn't set the bit in iir again, but it still produces * interrupts (for non-MSI). */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock(&dev_priv->irq_lock); if (iir & I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT) i915_handle_error(dev, false, "Command parser error, iir 0x%08x", @@ -4095,7 +4094,7 @@ static irqreturn_t i8xx_irq_handler(int irq, void *arg) if (pipe_stats[pipe] & 0x8000ffff) I915_WRITE(reg, pipe_stats[pipe]); } - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock(&dev_priv->irq_lock); I915_WRITE16(IIR, iir & ~flip_mask); new_iir = I915_READ16(IIR); /* Flush posted writes */ @@ -4249,7 +4248,6 @@ static irqreturn_t i915_irq_handler(int irq, void *arg) struct drm_device *dev = arg; struct drm_i915_private *dev_priv = dev->dev_private; u32 iir, new_iir, pipe_stats[I915_MAX_PIPES]; - unsigned long irqflags; u32 flip_mask = I915_DISPLAY_PLANE_A_FLIP_PENDING_INTERRUPT | I915_DISPLAY_PLANE_B_FLIP_PENDING_INTERRUPT; @@ -4265,7 +4263,7 @@ static irqreturn_t i915_irq_handler(int irq, void *arg) * It doesn't set the bit in iir again, but it still produces * interrupts (for non-MSI). */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock(&dev_priv->irq_lock); if (iir & I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT) i915_handle_error(dev, false, "Command parser error, iir 0x%08x", @@ -4281,7 +4279,7 @@ static irqreturn_t i915_irq_handler(int irq, void *arg) irq_received = true; } } - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock(&dev_priv->irq_lock); if (!irq_received) break; @@ -4476,7 +4474,6 @@ static irqreturn_t i965_irq_handler(int irq, void *arg) struct drm_i915_private *dev_priv = dev->dev_private; u32 iir, new_iir; u32 pipe_stats[I915_MAX_PIPES]; - unsigned long irqflags; int ret = IRQ_NONE, pipe; u32 flip_mask = I915_DISPLAY_PLANE_A_FLIP_PENDING_INTERRUPT | @@ -4493,7 +4490,7 @@ static irqreturn_t i965_irq_handler(int irq, void *arg) * It doesn't set the bit in iir again, but it still produces * interrupts (for non-MSI). */ - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock(&dev_priv->irq_lock); if (iir & I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT) i915_handle_error(dev, false, "Command parser error, iir 0x%08x", @@ -4511,7 +4508,7 @@ static irqreturn_t i965_irq_handler(int irq, void *arg) irq_received = true; } } - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock(&dev_priv->irq_lock); if (!irq_received) break; -- cgit v0.10.2 From 133217869ea7f328acedab2ed2ec1ad41b6390bd Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:29 +0200 Subject: drm/i915: Clarify irq_lock locking, special cases Grab bag for all the special cases: - i9xx_check_fifo_underruns is only called from crtc_enable hooks, i.e. process context. - i915_enable_asle_pipestat is only called from interrupt postinstall hooks. So again process context. - gen8_irq_power_well_post_enable is called from the runtime pm code, which again means process context. - The open-coded hpd_irq_setup loop in _thaw is also running in process context. So for all of them the plain _irq variant is sufficient. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index b8bd008..8ce1b13 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -686,11 +686,10 @@ static int __i915_drm_thaw(struct drm_device *dev, bool restore_gtt_mappings) intel_modeset_init_hw(dev); { - unsigned long irqflags; - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); if (dev_priv->display.hpd_irq_setup) dev_priv->display.hpd_irq_setup(dev); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); } intel_dp_mst_resume(dev); diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 6a4f389..a08cdc6 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -310,9 +310,8 @@ void i9xx_check_fifo_underruns(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *crtc; - unsigned long flags; - spin_lock_irqsave(&dev_priv->irq_lock, flags); + spin_lock_irq(&dev_priv->irq_lock); for_each_intel_crtc(dev, crtc) { u32 reg = PIPESTAT(crtc->pipe); @@ -331,7 +330,7 @@ void i9xx_check_fifo_underruns(struct drm_device *dev) DRM_ERROR("pipe %c underrun\n", pipe_name(crtc->pipe)); } - spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + spin_unlock_irq(&dev_priv->irq_lock); } static void i9xx_set_fifo_underrun_reporting(struct drm_device *dev, @@ -696,19 +695,18 @@ i915_disable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe, static void i915_enable_asle_pipestat(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long irqflags; if (!dev_priv->opregion.asle || !IS_MOBILE(dev)) return; - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); i915_enable_pipestat(dev_priv, PIPE_B, PIPE_LEGACY_BLC_EVENT_STATUS); if (INTEL_INFO(dev)->gen >= 4) i915_enable_pipestat(dev_priv, PIPE_A, PIPE_LEGACY_BLC_EVENT_STATUS); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); } /** @@ -3477,14 +3475,12 @@ static void gen8_irq_reset(struct drm_device *dev) void gen8_irq_power_well_post_enable(struct drm_i915_private *dev_priv) { - unsigned long irqflags; - - spin_lock_irqsave(&dev_priv->irq_lock, irqflags); + spin_lock_irq(&dev_priv->irq_lock); GEN8_IRQ_INIT_NDX(DE_PIPE, PIPE_B, dev_priv->de_irq_mask[PIPE_B], ~dev_priv->de_irq_mask[PIPE_B]); GEN8_IRQ_INIT_NDX(DE_PIPE, PIPE_C, dev_priv->de_irq_mask[PIPE_C], ~dev_priv->de_irq_mask[PIPE_C]); - spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); + spin_unlock_irq(&dev_priv->irq_lock); } static void cherryview_irq_preinstall(struct drm_device *dev) -- cgit v0.10.2 From d2e40e2741e32e03ce060cc74ef6d47eaf833bf9 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:31 +0200 Subject: drm/i915: Clarify uncore.lock locking Only one place looked in need of a bit of polish: hsw_restore_lcpll. It's used by the runtime pm code and hence is always called from process context. No irq flag saving required. Another thing I've stumbled over is that we might need to add a raw forcewake_get/put helpers which don't grab a runtime pm reference but just check that the device isn't suspended - we have this duplicated in the execlist code, too. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 0c6350c..8b73d11 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7603,7 +7603,6 @@ static void hsw_disable_lcpll(struct drm_i915_private *dev_priv, static void hsw_restore_lcpll(struct drm_i915_private *dev_priv) { uint32_t val; - unsigned long irqflags; val = I915_READ(LCPLL_CTL); @@ -7623,10 +7622,10 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv) * to call special forcewake code that doesn't touch runtime PM and * doesn't enable the forcewake delayed work. */ - spin_lock_irqsave(&dev_priv->uncore.lock, irqflags); + spin_lock_irq(&dev_priv->uncore.lock); if (dev_priv->uncore.forcewake_count++ == 0) dev_priv->uncore.funcs.force_wake_get(dev_priv, FORCEWAKE_ALL); - spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); + spin_unlock_irq(&dev_priv->uncore.lock); if (val & LCPLL_POWER_DOWN_ALLOW) { val &= ~LCPLL_POWER_DOWN_ALLOW; @@ -7657,10 +7656,10 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv) } /* See the big comment above. */ - spin_lock_irqsave(&dev_priv->uncore.lock, irqflags); + spin_lock_irq(&dev_priv->uncore.lock); if (--dev_priv->uncore.forcewake_count == 0) dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL); - spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); + spin_unlock_irq(&dev_priv->uncore.lock); } /* -- cgit v0.10.2 From 24955f2412fa16e30bb678bad1450ebd9fb8e180 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 15 Sep 2014 14:55:32 +0200 Subject: drm/i915: Clarify mmio_flip_lock locking The ->queue_flip callback is always called from process context, so plain _irq spinlock variants are enough. Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 8b73d11..5b05ddb 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9792,7 +9792,6 @@ static int intel_queue_mmio_flip(struct drm_device *dev, { struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - unsigned long irq_flags; int ret; if (WARN_ON(intel_crtc->mmio_flip.seqno)) @@ -9806,10 +9805,10 @@ static int intel_queue_mmio_flip(struct drm_device *dev, return 0; } - spin_lock_irqsave(&dev_priv->mmio_flip_lock, irq_flags); + spin_lock_irq(&dev_priv->mmio_flip_lock); intel_crtc->mmio_flip.seqno = obj->last_write_seqno; intel_crtc->mmio_flip.ring_id = obj->ring->id; - spin_unlock_irqrestore(&dev_priv->mmio_flip_lock, irq_flags); + spin_unlock_irq(&dev_priv->mmio_flip_lock); /* * Double check to catch cases where irq fired before -- cgit v0.10.2 From 01d06e9f963ba6a83154ab81929b7f5e04bbe5dd Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 5 Sep 2014 16:57:20 -0400 Subject: drm/i915: Only flush fbc on sw when fbc is enabled. Avoid touching fbc register when fbc is disabled. Cc: Paulo Zanoni Signed-off-by: Rodrigo Vivi Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 675e8a2..6f3b94b 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -352,6 +352,9 @@ void gen8_fbc_sw_flush(struct drm_device *dev, u32 value) if (!IS_GEN8(dev)) return; + if (!intel_fbc_enabled(dev)) + return; + I915_WRITE(MSG_FBC_REND_STATE, value); } -- cgit v0.10.2 From 342e36c6b0cab1d7461287d553154ff35fee989d Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 17 Sep 2014 16:59:20 -0400 Subject: drm/i915: Avoid reading fbc registers in vain when fbc was never enabled. If it wasn't never enabled by kernel parameter or platform default we can avoid reading registers so many times in vain Cc: Paulo Zanoni Signed-off-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 6f3b94b..1ec3c8f 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -339,6 +339,12 @@ bool intel_fbc_enabled(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; + /* If it wasn't never enabled by kernel parameter or platform default + * we can avoid reading registers so many times in vain + */ + if (!i915.enable_fbc) + return false; + if (!dev_priv->display.fbc_enabled) return false; -- cgit v0.10.2 From 5cb13c07dae73380d8b3ddc792740487b8742938 Mon Sep 17 00:00:00 2001 From: Deepak S Date: Thu, 18 Sep 2014 18:51:50 +0530 Subject: drm/i915/vlv: Remove check for Old Ack during forcewake MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Based on the HW team inputs. We can should not wait for the old ack, Waiting for old ack might fail, when other forcewake came before the present one is desserted. for example, if forcewake bit 0 was set and before it could get cleared forcewake bit 1 got set, HW eventually clear bit 0, when the bit 1 is cleared. i.e, bit 1 is still sent then forcewake bit 0 will still be set. Signed-off-by: Deepak S Reviewed-by: Ville Syrjälä [danvet: Add comment Ville requested.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 918b761..0e99852 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -194,13 +194,15 @@ static void vlv_force_wake_reset(struct drm_i915_private *dev_priv) static void __vlv_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) { + /* + * WaRsDontPollForAckOnClearingFWBits:vlv + * Hardware clears ack bits lazily (only when all ack + * bits become 0) so don't poll for individiual ack + * bits to be clear here like on other platforms. + */ + /* Check for Render Engine */ if (FORCEWAKE_RENDER & fw_engine) { - if (wait_for_atomic((__raw_i915_read32(dev_priv, - FORCEWAKE_ACK_VLV) & - FORCEWAKE_KERNEL) == 0, - FORCEWAKE_ACK_TIMEOUT_MS)) - DRM_ERROR("Timed out: Render forcewake old ack to clear.\n"); __raw_i915_write32(dev_priv, FORCEWAKE_VLV, _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL)); @@ -214,11 +216,6 @@ static void __vlv_force_wake_get(struct drm_i915_private *dev_priv, /* Check for Media Engine */ if (FORCEWAKE_MEDIA & fw_engine) { - if (wait_for_atomic((__raw_i915_read32(dev_priv, - FORCEWAKE_ACK_MEDIA_VLV) & - FORCEWAKE_KERNEL) == 0, - FORCEWAKE_ACK_TIMEOUT_MS)) - DRM_ERROR("Timed out: Media forcewake old ack to clear.\n"); __raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_VLV, _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL)); -- cgit v0.10.2 From 9d533c5a5219ed7e3d52a05309723964be2307b3 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 19 Sep 2014 17:07:10 +0200 Subject: drm/i915: Update DRIVER_DATE to 20140919 Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 07dafa2..03f59ec 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -54,7 +54,7 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20140905" +#define DRIVER_DATE "20140919" enum pipe { INVALID_PIPE = -1, -- cgit v0.10.2 From 9beb0ccb66ec5d011ac13a02d3b1098035195251 Mon Sep 17 00:00:00 2001 From: Brad Volkin Date: Thu, 18 Sep 2014 16:26:26 -0700 Subject: drm/i915: Re-enable the command parser when using PPGTT In commit commit 896ab1a5d54269b463a24194c2e4a369103b46d8 Author: Daniel Vetter Date: Wed Aug 6 15:04:51 2014 +0200 drm/i915: Fix up checks for aliasing ppgtt it looks like we accidentally inverted the check that the command parser should only run when the driver enables some form of PPGTT. Testcase: igt/gem_exec_parse Cc: Daniel Vetter Signed-off-by: Brad Volkin [danvet: Also drop the comment right above, all production vlv now have hw ppgtt enabled.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c index c45856b..d40f264 100644 --- a/drivers/gpu/drm/i915/i915_cmd_parser.c +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c @@ -845,12 +845,7 @@ bool i915_needs_cmd_parser(struct intel_engine_cs *ring) if (!ring->needs_cmd_parser) return false; - /* - * XXX: VLV is Gen7 and therefore has cmd_tables, but has PPGTT - * disabled. That will cause all of the parser's PPGTT checks to - * fail. For now, disable parsing when PPGTT is off. - */ - if (USES_PPGTT(ring->dev)) + if (!USES_PPGTT(ring->dev)) return false; return (i915.enable_cmd_parser == 1); -- cgit v0.10.2 From 00caf0199f66871b0e2c28d7c2079de0ce1d646c Mon Sep 17 00:00:00 2001 From: Brad Volkin Date: Thu, 18 Sep 2014 16:26:27 -0700 Subject: drm/i915: Log a message when rejecting LRM to OACONTROL The other paths in the command parser that reject a batch all log a message indicating the reason. We simply missed this one. Signed-off-by: Brad Volkin Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c index d40f264..4c35e2a 100644 --- a/drivers/gpu/drm/i915/i915_cmd_parser.c +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c @@ -881,8 +881,10 @@ static bool check_cmd(const struct intel_engine_cs *ring, * OACONTROL writes to only MI_LOAD_REGISTER_IMM commands. */ if (reg_addr == OACONTROL) { - if (desc->cmd.value == MI_LOAD_REGISTER_MEM) + if (desc->cmd.value == MI_LOAD_REGISTER_MEM) { + DRM_DEBUG_DRIVER("CMD: Rejected LRM to OACONTROL\n"); return false; + } if (desc->cmd.value == MI_LOAD_REGISTER_IMM(1)) *oacontrol_set = (cmd[2] != 0); -- cgit v0.10.2 From 1893a71b1eba79a46ac5b356208e80f822543fa4 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 19 Sep 2014 11:56:27 +0100 Subject: drm/i915: Inline feature detection into sanitize_enable_ppgtt Rather than splitting and hiding away critical parts of sanitize_enable_ppgtt() into single use macros in the headers, inline them into the function for clarity. Signed-off-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 03f59ec..5fce16c 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2139,8 +2139,6 @@ struct drm_i915_cmd_table { #define HAS_HW_CONTEXTS(dev) (INTEL_INFO(dev)->gen >= 6) #define HAS_LOGICAL_RING_CONTEXTS(dev) (INTEL_INFO(dev)->gen >= 8) -#define HAS_ALIASING_PPGTT(dev) (INTEL_INFO(dev)->gen >= 6) -#define HAS_PPGTT(dev) (INTEL_INFO(dev)->gen >= 7 && !IS_GEN8(dev)) #define USES_PPGTT(dev) (i915.enable_ppgtt) #define USES_FULL_PPGTT(dev) (i915.enable_ppgtt == 2) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index a234446..90c9bf6 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -35,13 +35,21 @@ static void chv_setup_private_ppat(struct drm_i915_private *dev_priv); static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) { - if (enable_ppgtt == 0 || !HAS_ALIASING_PPGTT(dev)) + bool has_aliasing_ppgtt; + bool has_full_ppgtt; + + has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6; + has_full_ppgtt = INTEL_INFO(dev)->gen >= 7; + if (IS_GEN8(dev)) + has_full_ppgtt = false; /* XXX why? */ + + if (enable_ppgtt == 0 || !has_aliasing_ppgtt) return 0; if (enable_ppgtt == 1) return 1; - if (enable_ppgtt == 2 && HAS_PPGTT(dev)) + if (enable_ppgtt == 2 && has_full_ppgtt) return 2; #ifdef CONFIG_INTEL_IOMMU @@ -59,7 +67,7 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) return 0; } - return HAS_PPGTT(dev) ? 2 : HAS_ALIASING_PPGTT(dev) ? 1 : 0; + return has_full_ppgtt ? 2 : has_aliasing_ppgtt ? 1 : 0; } -- cgit v0.10.2 From b680c37a4d145cf4d8f2b24e46b1163e5ceb1d35 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 19 Sep 2014 18:27:27 +0200 Subject: drm/i915: DocBook integration for frontbuffer tracking I shouldn't ask everyone to do this and fail myself ... This extracts all the frontbuffer tracking functions into intel_frontbuffer.c, adds a DOC overview section and also adds the missing kerneldoc for i915_gem_track_fb and also pulls it into the same section for convenience. v2: Don't forget about the header files. v3: Oops, might check compilation next time around. To make my life easier drop the increase_pllclock from set_base_atomic since really, it doesn't matter if you see your Oops or kgdb with a tiny bit of lag. v4: Try to better explain how to actually use this, requested by Paulo on irc. v5: Explain invalidate/flush a bit clearer. v6: s/business/busyness/ Acked-by: Paulo Zanoni Cc: Paulo Zanoni Cc: Vandana Kannan Signed-off-by: Daniel Vetter diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index ca44d9f..7ad6128 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -3804,6 +3804,13 @@ int num_ioctls; + Frontbuffer Tracking +!Pdrivers/gpu/drm/i915/intel_frontbuffer.c frontbuffer tracking +!Idrivers/gpu/drm/i915/intel_frontbuffer.c +!Fdrivers/gpu/drm/i915/intel_drv.h intel_frontbuffer_flip +!Fdrivers/gpu/drm/i915/i915_gem.c i915_gem_track_fb + + Plane Configuration This section covers plane configuration and composition with the diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index c1dd485..2d8317d 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -43,6 +43,7 @@ i915-y += intel_renderstate_gen6.o \ # modesetting core code i915-y += intel_bios.o \ intel_display.o \ + intel_frontbuffer.o \ intel_modes.o \ intel_overlay.o \ intel_sideband.o \ diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 2fb87cf..55a2ebb5 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5077,6 +5077,15 @@ int i915_gem_open(struct drm_device *dev, struct drm_file *file) return ret; } +/** + * i915_gem_track_fb - update frontbuffer tracking + * old: current GEM buffer for the frontbuffer slots + * new: new GEM buffer for the frontbuffer slots + * frontbuffer_bits: bitmask of frontbuffer slots + * + * This updates the frontbuffer tracking bits @frontbuffer_bits by clearing them + * from @old and setting them in @new. Both @old and @new can be NULL. + */ void i915_gem_track_fb(struct drm_i915_gem_object *old, struct drm_i915_gem_object *new, unsigned frontbuffer_bits) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5b05ddb..858011d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -76,8 +76,6 @@ static const uint32_t intel_cursor_formats[] = { #define DIV_ROUND_CLOSEST_ULL(ll, d) \ ({ unsigned long long _tmp = (ll)+(d)/2; do_div(_tmp, d); _tmp; }) -static void intel_increase_pllclock(struct drm_device *dev, - enum pipe pipe); static void intel_crtc_update_cursor(struct drm_crtc *crtc, bool on); static void i9xx_crtc_clock_get(struct intel_crtc *crtc, @@ -1138,8 +1136,8 @@ void assert_fdi_rx_pll(struct drm_i915_private *dev_priv, state_string(state), state_string(cur_state)); } -static void assert_panel_unlocked(struct drm_i915_private *dev_priv, - enum pipe pipe) +void assert_panel_unlocked(struct drm_i915_private *dev_priv, + enum pipe pipe) { struct drm_device *dev = dev_priv->dev; int pp_reg; @@ -2631,7 +2629,6 @@ intel_pipe_set_base_atomic(struct drm_crtc *crtc, struct drm_framebuffer *fb, if (dev_priv->display.disable_fbc) dev_priv->display.disable_fbc(dev); - intel_increase_pllclock(dev, to_intel_crtc(crtc)->pipe); dev_priv->display.update_primary_plane(crtc, fb, x, y); @@ -8986,35 +8983,6 @@ struct drm_display_mode *intel_crtc_mode_get(struct drm_device *dev, return mode; } -static void intel_increase_pllclock(struct drm_device *dev, - enum pipe pipe) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int dpll_reg = DPLL(pipe); - int dpll; - - if (!HAS_GMCH_DISPLAY(dev)) - return; - - if (!dev_priv->lvds_downclock_avail) - return; - - dpll = I915_READ(dpll_reg); - if (!HAS_PIPE_CXSR(dev) && (dpll & DISPLAY_RATE_SELECT_FPA1)) { - DRM_DEBUG_DRIVER("upclocking LVDS\n"); - - assert_panel_unlocked(dev_priv, pipe); - - dpll &= ~DISPLAY_RATE_SELECT_FPA1; - I915_WRITE(dpll_reg, dpll); - intel_wait_for_vblank(dev, pipe); - - dpll = I915_READ(dpll_reg); - if (dpll & DISPLAY_RATE_SELECT_FPA1) - DRM_DEBUG_DRIVER("failed to upclock LVDS!\n"); - } -} - static void intel_decrease_pllclock(struct drm_crtc *crtc) { struct drm_device *dev = crtc->dev; @@ -9090,188 +9058,6 @@ out: intel_runtime_pm_put(dev_priv); } - -/** - * intel_mark_fb_busy - mark given planes as busy - * @dev: DRM device - * @frontbuffer_bits: bits for the affected planes - * @ring: optional ring for asynchronous commands - * - * This function gets called every time the screen contents change. It can be - * used to keep e.g. the update rate at the nominal refresh rate with DRRS. - */ -static void intel_mark_fb_busy(struct drm_device *dev, - unsigned frontbuffer_bits, - struct intel_engine_cs *ring) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - enum pipe pipe; - - if (!i915.powersave) - return; - - for_each_pipe(dev_priv, pipe) { - if (!(frontbuffer_bits & INTEL_FRONTBUFFER_ALL_MASK(pipe))) - continue; - - intel_increase_pllclock(dev, pipe); - if (ring && intel_fbc_enabled(dev)) - ring->fbc_dirty = true; - } -} - -/** - * intel_fb_obj_invalidate - invalidate frontbuffer object - * @obj: GEM object to invalidate - * @ring: set for asynchronous rendering - * - * This function gets called every time rendering on the given object starts and - * frontbuffer caching (fbc, low refresh rate for DRRS, panel self refresh) must - * be invalidated. If @ring is non-NULL any subsequent invalidation will be delayed - * until the rendering completes or a flip on this frontbuffer plane is - * scheduled. - */ -void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj, - struct intel_engine_cs *ring) -{ - struct drm_device *dev = obj->base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - - WARN_ON(!mutex_is_locked(&dev->struct_mutex)); - - if (!obj->frontbuffer_bits) - return; - - if (ring) { - mutex_lock(&dev_priv->fb_tracking.lock); - dev_priv->fb_tracking.busy_bits - |= obj->frontbuffer_bits; - dev_priv->fb_tracking.flip_bits - &= ~obj->frontbuffer_bits; - mutex_unlock(&dev_priv->fb_tracking.lock); - } - - intel_mark_fb_busy(dev, obj->frontbuffer_bits, ring); - - intel_edp_psr_invalidate(dev, obj->frontbuffer_bits); -} - -/** - * intel_frontbuffer_flush - flush frontbuffer - * @dev: DRM device - * @frontbuffer_bits: frontbuffer plane tracking bits - * - * This function gets called every time rendering on the given planes has - * completed and frontbuffer caching can be started again. Flushes will get - * delayed if they're blocked by some oustanding asynchronous rendering. - * - * Can be called without any locks held. - */ -void intel_frontbuffer_flush(struct drm_device *dev, - unsigned frontbuffer_bits) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - /* Delay flushing when rings are still busy.*/ - mutex_lock(&dev_priv->fb_tracking.lock); - frontbuffer_bits &= ~dev_priv->fb_tracking.busy_bits; - mutex_unlock(&dev_priv->fb_tracking.lock); - - intel_mark_fb_busy(dev, frontbuffer_bits, NULL); - - intel_edp_psr_flush(dev, frontbuffer_bits); - - /* - * FIXME: Unconditional fbc flushing here is a rather gross hack and - * needs to be reworked into a proper frontbuffer tracking scheme like - * psr employs. - */ - if (IS_BROADWELL(dev)) - gen8_fbc_sw_flush(dev, FBC_REND_CACHE_CLEAN); -} - -/** - * intel_fb_obj_flush - flush frontbuffer object - * @obj: GEM object to flush - * @retire: set when retiring asynchronous rendering - * - * This function gets called every time rendering on the given object has - * completed and frontbuffer caching can be started again. If @retire is true - * then any delayed flushes will be unblocked. - */ -void intel_fb_obj_flush(struct drm_i915_gem_object *obj, - bool retire) -{ - struct drm_device *dev = obj->base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - unsigned frontbuffer_bits; - - WARN_ON(!mutex_is_locked(&dev->struct_mutex)); - - if (!obj->frontbuffer_bits) - return; - - frontbuffer_bits = obj->frontbuffer_bits; - - if (retire) { - mutex_lock(&dev_priv->fb_tracking.lock); - /* Filter out new bits since rendering started. */ - frontbuffer_bits &= dev_priv->fb_tracking.busy_bits; - - dev_priv->fb_tracking.busy_bits &= ~frontbuffer_bits; - mutex_unlock(&dev_priv->fb_tracking.lock); - } - - intel_frontbuffer_flush(dev, frontbuffer_bits); -} - -/** - * intel_frontbuffer_flip_prepare - prepare asnychronous frontbuffer flip - * @dev: DRM device - * @frontbuffer_bits: frontbuffer plane tracking bits - * - * This function gets called after scheduling a flip on @obj. The actual - * frontbuffer flushing will be delayed until completion is signalled with - * intel_frontbuffer_flip_complete. If an invalidate happens in between this - * flush will be cancelled. - * - * Can be called without any locks held. - */ -void intel_frontbuffer_flip_prepare(struct drm_device *dev, - unsigned frontbuffer_bits) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - mutex_lock(&dev_priv->fb_tracking.lock); - dev_priv->fb_tracking.flip_bits - |= frontbuffer_bits; - mutex_unlock(&dev_priv->fb_tracking.lock); -} - -/** - * intel_frontbuffer_flip_complete - complete asynchronous frontbuffer flush - * @dev: DRM device - * @frontbuffer_bits: frontbuffer plane tracking bits - * - * This function gets called after the flip has been latched and will complete - * on the next vblank. It will execute the fush if it hasn't been cancalled yet. - * - * Can be called without any locks held. - */ -void intel_frontbuffer_flip_complete(struct drm_device *dev, - unsigned frontbuffer_bits) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - mutex_lock(&dev_priv->fb_tracking.lock); - /* Mask any cancelled flips. */ - frontbuffer_bits &= dev_priv->fb_tracking.flip_bits; - dev_priv->fb_tracking.flip_bits &= ~frontbuffer_bits; - mutex_unlock(&dev_priv->fb_tracking.lock); - - intel_frontbuffer_flush(dev, frontbuffer_bits); -} - static void intel_crtc_destroy(struct drm_crtc *crtc) { struct intel_crtc *intel_crtc = to_intel_crtc(crtc); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 1b72c15..6171267 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -801,11 +801,7 @@ void intel_ddi_clock_get(struct intel_encoder *encoder, struct intel_crtc_config *pipe_config); void intel_ddi_set_vc_payload_alloc(struct drm_crtc *crtc, bool state); -/* intel_display.c */ -const char *intel_output_name(int output); -bool intel_has_pending_fb_unpin(struct drm_device *dev); -int intel_pch_rawclk(struct drm_device *dev); -void intel_mark_busy(struct drm_device *dev); +/* intel_frontbuffer.c */ void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj, struct intel_engine_cs *ring); void intel_frontbuffer_flip_prepare(struct drm_device *dev, @@ -833,6 +829,13 @@ void intel_frontbuffer_flip(struct drm_device *dev, } void intel_fb_obj_flush(struct drm_i915_gem_object *obj, bool retire); + + +/* intel_display.c */ +const char *intel_output_name(int output); +bool intel_has_pending_fb_unpin(struct drm_device *dev); +int intel_pch_rawclk(struct drm_device *dev); +void intel_mark_busy(struct drm_device *dev); void intel_mark_idle(struct drm_device *dev); void intel_crtc_restore_mode(struct drm_crtc *crtc); void intel_crtc_control(struct drm_crtc *crtc, bool enable); @@ -891,6 +894,8 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc); void intel_put_shared_dpll(struct intel_crtc *crtc); /* modesetting asserts */ +void assert_panel_unlocked(struct drm_i915_private *dev_priv, + enum pipe pipe); void assert_pll(struct drm_i915_private *dev_priv, enum pipe pipe, bool state); #define assert_pll_enabled(d, p) assert_pll(d, p, true) diff --git a/drivers/gpu/drm/i915/intel_frontbuffer.c b/drivers/gpu/drm/i915/intel_frontbuffer.c new file mode 100644 index 0000000..f74744c --- /dev/null +++ b/drivers/gpu/drm/i915/intel_frontbuffer.c @@ -0,0 +1,277 @@ +/* + * Copyright © 2014 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + * Authors: + * Daniel Vetter + */ + +/** + * DOC: frontbuffer tracking + * + * Many features require us to track changes to the currently active + * frontbuffer, especially rendering targetted at the frontbuffer. + * + * To be able to do so GEM tracks frontbuffers using a bitmask for all possible + * frontbuffer slots through i915_gem_track_fb(). The function in this file are + * then called when the contents of the frontbuffer are invalidated, when + * frontbuffer rendering has stopped again to flush out all the changes and when + * the frontbuffer is exchanged with a flip. Subsystems interested in + * frontbuffer changes (e.g. PSR, FBC, DRRS) should directly put their callbacks + * into the relevant places and filter for the frontbuffer slots that they are + * interested int. + * + * On a high level there are two types of powersaving features. The first one + * work like a special cache (FBC and PSR) and are interested when they should + * stop caching and when to restart caching. This is done by placing callbacks + * into the invalidate and the flush functions: At invalidate the caching must + * be stopped and at flush time it can be restarted. And maybe they need to know + * when the frontbuffer changes (e.g. when the hw doesn't initiate an invalidate + * and flush on its own) which can be achieved with placing callbacks into the + * flip functions. + * + * The other type of display power saving feature only cares about busyness + * (e.g. DRRS). In that case all three (invalidate, flush and flip) indicate + * busyness. There is no direct way to detect idleness. Instead an idle timer + * work delayed work should be started from the flush and flip functions and + * cancelled as soon as busyness is detected. + * + * Note that there's also an older frontbuffer activity tracking scheme which + * just trackings general activity. This is done by the various mark_busy and + * mark_idle functions. For display power management features using these + * functions is deprecated and should be avoided. + */ + +#include + +#include "intel_drv.h" +#include "i915_drv.h" + +static void intel_increase_pllclock(struct drm_device *dev, + enum pipe pipe) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int dpll_reg = DPLL(pipe); + int dpll; + + if (!HAS_GMCH_DISPLAY(dev)) + return; + + if (!dev_priv->lvds_downclock_avail) + return; + + dpll = I915_READ(dpll_reg); + if (!HAS_PIPE_CXSR(dev) && (dpll & DISPLAY_RATE_SELECT_FPA1)) { + DRM_DEBUG_DRIVER("upclocking LVDS\n"); + + assert_panel_unlocked(dev_priv, pipe); + + dpll &= ~DISPLAY_RATE_SELECT_FPA1; + I915_WRITE(dpll_reg, dpll); + intel_wait_for_vblank(dev, pipe); + + dpll = I915_READ(dpll_reg); + if (dpll & DISPLAY_RATE_SELECT_FPA1) + DRM_DEBUG_DRIVER("failed to upclock LVDS!\n"); + } +} + +/** + * intel_mark_fb_busy - mark given planes as busy + * @dev: DRM device + * @frontbuffer_bits: bits for the affected planes + * @ring: optional ring for asynchronous commands + * + * This function gets called every time the screen contents change. It can be + * used to keep e.g. the update rate at the nominal refresh rate with DRRS. + */ +static void intel_mark_fb_busy(struct drm_device *dev, + unsigned frontbuffer_bits, + struct intel_engine_cs *ring) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + enum pipe pipe; + + if (!i915.powersave) + return; + + for_each_pipe(dev_priv, pipe) { + if (!(frontbuffer_bits & INTEL_FRONTBUFFER_ALL_MASK(pipe))) + continue; + + intel_increase_pllclock(dev, pipe); + if (ring && intel_fbc_enabled(dev)) + ring->fbc_dirty = true; + } +} + +/** + * intel_fb_obj_invalidate - invalidate frontbuffer object + * @obj: GEM object to invalidate + * @ring: set for asynchronous rendering + * + * This function gets called every time rendering on the given object starts and + * frontbuffer caching (fbc, low refresh rate for DRRS, panel self refresh) must + * be invalidated. If @ring is non-NULL any subsequent invalidation will be delayed + * until the rendering completes or a flip on this frontbuffer plane is + * scheduled. + */ +void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj, + struct intel_engine_cs *ring) +{ + struct drm_device *dev = obj->base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + + WARN_ON(!mutex_is_locked(&dev->struct_mutex)); + + if (!obj->frontbuffer_bits) + return; + + if (ring) { + mutex_lock(&dev_priv->fb_tracking.lock); + dev_priv->fb_tracking.busy_bits + |= obj->frontbuffer_bits; + dev_priv->fb_tracking.flip_bits + &= ~obj->frontbuffer_bits; + mutex_unlock(&dev_priv->fb_tracking.lock); + } + + intel_mark_fb_busy(dev, obj->frontbuffer_bits, ring); + + intel_edp_psr_invalidate(dev, obj->frontbuffer_bits); +} + +/** + * intel_frontbuffer_flush - flush frontbuffer + * @dev: DRM device + * @frontbuffer_bits: frontbuffer plane tracking bits + * + * This function gets called every time rendering on the given planes has + * completed and frontbuffer caching can be started again. Flushes will get + * delayed if they're blocked by some oustanding asynchronous rendering. + * + * Can be called without any locks held. + */ +void intel_frontbuffer_flush(struct drm_device *dev, + unsigned frontbuffer_bits) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + /* Delay flushing when rings are still busy.*/ + mutex_lock(&dev_priv->fb_tracking.lock); + frontbuffer_bits &= ~dev_priv->fb_tracking.busy_bits; + mutex_unlock(&dev_priv->fb_tracking.lock); + + intel_mark_fb_busy(dev, frontbuffer_bits, NULL); + + intel_edp_psr_flush(dev, frontbuffer_bits); + + /* + * FIXME: Unconditional fbc flushing here is a rather gross hack and + * needs to be reworked into a proper frontbuffer tracking scheme like + * psr employs. + */ + if (IS_BROADWELL(dev)) + gen8_fbc_sw_flush(dev, FBC_REND_CACHE_CLEAN); +} + +/** + * intel_fb_obj_flush - flush frontbuffer object + * @obj: GEM object to flush + * @retire: set when retiring asynchronous rendering + * + * This function gets called every time rendering on the given object has + * completed and frontbuffer caching can be started again. If @retire is true + * then any delayed flushes will be unblocked. + */ +void intel_fb_obj_flush(struct drm_i915_gem_object *obj, + bool retire) +{ + struct drm_device *dev = obj->base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned frontbuffer_bits; + + WARN_ON(!mutex_is_locked(&dev->struct_mutex)); + + if (!obj->frontbuffer_bits) + return; + + frontbuffer_bits = obj->frontbuffer_bits; + + if (retire) { + mutex_lock(&dev_priv->fb_tracking.lock); + /* Filter out new bits since rendering started. */ + frontbuffer_bits &= dev_priv->fb_tracking.busy_bits; + + dev_priv->fb_tracking.busy_bits &= ~frontbuffer_bits; + mutex_unlock(&dev_priv->fb_tracking.lock); + } + + intel_frontbuffer_flush(dev, frontbuffer_bits); +} + +/** + * intel_frontbuffer_flip_prepare - prepare asnychronous frontbuffer flip + * @dev: DRM device + * @frontbuffer_bits: frontbuffer plane tracking bits + * + * This function gets called after scheduling a flip on @obj. The actual + * frontbuffer flushing will be delayed until completion is signalled with + * intel_frontbuffer_flip_complete. If an invalidate happens in between this + * flush will be cancelled. + * + * Can be called without any locks held. + */ +void intel_frontbuffer_flip_prepare(struct drm_device *dev, + unsigned frontbuffer_bits) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + mutex_lock(&dev_priv->fb_tracking.lock); + dev_priv->fb_tracking.flip_bits + |= frontbuffer_bits; + mutex_unlock(&dev_priv->fb_tracking.lock); +} + +/** + * intel_frontbuffer_flip_complete - complete asynchronous frontbuffer flush + * @dev: DRM device + * @frontbuffer_bits: frontbuffer plane tracking bits + * + * This function gets called after the flip has been latched and will complete + * on the next vblank. It will execute the fush if it hasn't been cancalled yet. + * + * Can be called without any locks held. + */ +void intel_frontbuffer_flip_complete(struct drm_device *dev, + unsigned frontbuffer_bits) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + mutex_lock(&dev_priv->fb_tracking.lock); + /* Mask any cancelled flips. */ + frontbuffer_bits &= dev_priv->fb_tracking.flip_bits; + dev_priv->fb_tracking.flip_bits &= ~frontbuffer_bits; + mutex_unlock(&dev_priv->fb_tracking.lock); + + intel_frontbuffer_flush(dev, frontbuffer_bits); +} + -- cgit v0.10.2 From d2dee86cece9deee33923ee71be918f0452c8ebe Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Fri, 19 Sep 2014 16:04:54 -0300 Subject: drm/i915: extract intel_init_fbc() Because I plan to expand it a little bit. Signed-off-by: Paulo Zanoni Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 1ec3c8f..2ca9fdb 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7358,33 +7358,39 @@ void intel_fini_runtime_pm(struct drm_i915_private *dev_priv) pm_runtime_disable(device); } +static void intel_init_fbc(struct drm_i915_private *dev_priv) +{ + if (!HAS_FBC(dev_priv)) + return; + + if (INTEL_INFO(dev_priv)->gen >= 7) { + dev_priv->display.fbc_enabled = ironlake_fbc_enabled; + dev_priv->display.enable_fbc = gen7_enable_fbc; + dev_priv->display.disable_fbc = ironlake_disable_fbc; + } else if (INTEL_INFO(dev_priv)->gen >= 5) { + dev_priv->display.fbc_enabled = ironlake_fbc_enabled; + dev_priv->display.enable_fbc = ironlake_enable_fbc; + dev_priv->display.disable_fbc = ironlake_disable_fbc; + } else if (IS_GM45(dev_priv)) { + dev_priv->display.fbc_enabled = g4x_fbc_enabled; + dev_priv->display.enable_fbc = g4x_enable_fbc; + dev_priv->display.disable_fbc = g4x_disable_fbc; + } else { + dev_priv->display.fbc_enabled = i8xx_fbc_enabled; + dev_priv->display.enable_fbc = i8xx_enable_fbc; + dev_priv->display.disable_fbc = i8xx_disable_fbc; + + /* This value was pulled out of someone's hat */ + I915_WRITE(FBC_CONTROL, 500 << FBC_CTL_INTERVAL_SHIFT); + } +} + /* Set up chip specific power management-related functions */ void intel_init_pm(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - if (HAS_FBC(dev)) { - if (INTEL_INFO(dev)->gen >= 7) { - dev_priv->display.fbc_enabled = ironlake_fbc_enabled; - dev_priv->display.enable_fbc = gen7_enable_fbc; - dev_priv->display.disable_fbc = ironlake_disable_fbc; - } else if (INTEL_INFO(dev)->gen >= 5) { - dev_priv->display.fbc_enabled = ironlake_fbc_enabled; - dev_priv->display.enable_fbc = ironlake_enable_fbc; - dev_priv->display.disable_fbc = ironlake_disable_fbc; - } else if (IS_GM45(dev)) { - dev_priv->display.fbc_enabled = g4x_fbc_enabled; - dev_priv->display.enable_fbc = g4x_enable_fbc; - dev_priv->display.disable_fbc = g4x_disable_fbc; - } else { - dev_priv->display.fbc_enabled = i8xx_fbc_enabled; - dev_priv->display.enable_fbc = i8xx_enable_fbc; - dev_priv->display.disable_fbc = i8xx_disable_fbc; - - /* This value was pulled out of someone's hat */ - I915_WRITE(FBC_CONTROL, 500 << FBC_CTL_INTERVAL_SHIFT); - } - } + intel_init_fbc(dev_priv); /* For cxsr */ if (IS_PINEVIEW(dev)) -- cgit v0.10.2 From 9adccc6063d1cf6ba38a5a26b87001554105be18 Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Fri, 19 Sep 2014 16:04:55 -0300 Subject: drm/i915: add SW tracking to FBC enabling Currently, calling intel_fbc_enabled() will trigger a register read. And we call it a lot of times, even when FBC is disabled, so saving a few cycles would be a good thing. Another reason for this patch is because we currently call intel_fbc_enabled() while the HW is runtime suspended, so the read makes no sense and triggers a WARN. This happens even if FBC is disabled by default. Of course one could argue that we just shouldn't be calling intel_fbc_enabled() while the driver is runtime suspended, and I agree that's a good argument, but I still think that the reason explained in the first paragraph already justifies the patch. This problem can easily be reproduced with many subtests of igt/pm_rpm, and it is a regression introduced by: commit c5ad011d7d256ecbe173324029e992817194d2b0 Author: Rodrigo Vivi Date: Mon Aug 4 03:51:38 2014 -0700 drm/i915: FBC flush nuke for BDW Testcase: igt/pm_rpm/cursor (and others) Cc: Rodrigo Vivi Signed-off-by: Paulo Zanoni Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 5fce16c..999bd57 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -662,6 +662,10 @@ struct i915_fbc { bool false_color; + /* Tracks whether the HW is actually enabled, not whether the feature is + * possible. */ + bool enabled; + struct intel_fbc_work { struct delayed_work work; struct drm_crtc *crtc; diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 2ca9fdb..6b41620 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -71,6 +71,8 @@ static void i8xx_disable_fbc(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; u32 fbc_ctl; + dev_priv->fbc.enabled = false; + /* Disable compression */ fbc_ctl = I915_READ(FBC_CONTROL); if ((fbc_ctl & FBC_CTL_EN) == 0) @@ -99,6 +101,8 @@ static void i8xx_enable_fbc(struct drm_crtc *crtc) int i; u32 fbc_ctl; + dev_priv->fbc.enabled = true; + cfb_pitch = dev_priv->fbc.size / FBC_LL_SIZE; if (fb->pitches[0] < cfb_pitch) cfb_pitch = fb->pitches[0]; @@ -153,6 +157,8 @@ static void g4x_enable_fbc(struct drm_crtc *crtc) struct intel_crtc *intel_crtc = to_intel_crtc(crtc); u32 dpfc_ctl; + dev_priv->fbc.enabled = true; + dpfc_ctl = DPFC_CTL_PLANE(intel_crtc->plane) | DPFC_SR_EN; if (drm_format_plane_cpp(fb->pixel_format, 0) == 2) dpfc_ctl |= DPFC_CTL_LIMIT_2X; @@ -173,6 +179,8 @@ static void g4x_disable_fbc(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; u32 dpfc_ctl; + dev_priv->fbc.enabled = false; + /* Disable compression */ dpfc_ctl = I915_READ(DPFC_CONTROL); if (dpfc_ctl & DPFC_CTL_EN) { @@ -224,6 +232,8 @@ static void ironlake_enable_fbc(struct drm_crtc *crtc) struct intel_crtc *intel_crtc = to_intel_crtc(crtc); u32 dpfc_ctl; + dev_priv->fbc.enabled = true; + dpfc_ctl = DPFC_CTL_PLANE(intel_crtc->plane); if (drm_format_plane_cpp(fb->pixel_format, 0) == 2) dev_priv->fbc.threshold++; @@ -264,6 +274,8 @@ static void ironlake_disable_fbc(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; u32 dpfc_ctl; + dev_priv->fbc.enabled = false; + /* Disable compression */ dpfc_ctl = I915_READ(ILK_DPFC_CONTROL); if (dpfc_ctl & DPFC_CTL_EN) { @@ -290,6 +302,8 @@ static void gen7_enable_fbc(struct drm_crtc *crtc) struct intel_crtc *intel_crtc = to_intel_crtc(crtc); u32 dpfc_ctl; + dev_priv->fbc.enabled = true; + dpfc_ctl = IVB_DPFC_CTL_PLANE(intel_crtc->plane); if (drm_format_plane_cpp(fb->pixel_format, 0) == 2) dev_priv->fbc.threshold++; @@ -339,16 +353,7 @@ bool intel_fbc_enabled(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - /* If it wasn't never enabled by kernel parameter or platform default - * we can avoid reading registers so many times in vain - */ - if (!i915.enable_fbc) - return false; - - if (!dev_priv->display.fbc_enabled) - return false; - - return dev_priv->display.fbc_enabled(dev); + return dev_priv->fbc.enabled; } void gen8_fbc_sw_flush(struct drm_device *dev, u32 value) @@ -7360,8 +7365,10 @@ void intel_fini_runtime_pm(struct drm_i915_private *dev_priv) static void intel_init_fbc(struct drm_i915_private *dev_priv) { - if (!HAS_FBC(dev_priv)) + if (!HAS_FBC(dev_priv)) { + dev_priv->fbc.enabled = false; return; + } if (INTEL_INFO(dev_priv)->gen >= 7) { dev_priv->display.fbc_enabled = ironlake_fbc_enabled; @@ -7383,6 +7390,8 @@ static void intel_init_fbc(struct drm_i915_private *dev_priv) /* This value was pulled out of someone's hat */ I915_WRITE(FBC_CONTROL, 500 << FBC_CTL_INTERVAL_SHIFT); } + + dev_priv->fbc.enabled = dev_priv->display.fbc_enabled(dev_priv->dev); } /* Set up chip specific power management-related functions */ -- cgit v0.10.2 From ba80f4d4266b2c804a2dd6deec6e2df32cb033ce Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 16 Sep 2014 19:19:05 -0400 Subject: drm/i915: PSR: organize setup function. psr_enabled is already by itself a setup once so let's put the W/As there and rename old setup once to setup_vsc. Signed-off-by: Rodrigo Vivi Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 5975445..e0066c0 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1991,10 +1991,8 @@ static void intel_edp_psr_write_vsc(struct intel_dp *intel_dp, POSTING_READ(ctl_reg); } -static void intel_edp_psr_setup(struct intel_dp *intel_dp) +static void intel_edp_psr_setup_vsc(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = dev->dev_private; struct edp_vsc_psr psr_vsc; /* Prepare VSC packet as per EDP 1.3 spec, Table 3.10 */ @@ -2004,10 +2002,6 @@ static void intel_edp_psr_setup(struct intel_dp *intel_dp) psr_vsc.sdp_header.HB2 = 0x2; psr_vsc.sdp_header.HB3 = 0x8; intel_edp_psr_write_vsc(intel_dp, &psr_vsc); - - /* Avoid continuous PSR exit by masking memup and hpd */ - I915_WRITE(EDP_PSR_DEBUG_CTL(dev), EDP_PSR_DEBUG_MASK_MEMUP | - EDP_PSR_DEBUG_MASK_HPD | EDP_PSR_DEBUG_MASK_LPSP); } static void intel_edp_psr_enable_sink(struct intel_dp *intel_dp) @@ -2160,8 +2154,11 @@ void intel_edp_psr_enable(struct intel_dp *intel_dp) dev_priv->psr.busy_frontbuffer_bits = 0; - /* Setup PSR once */ - intel_edp_psr_setup(intel_dp); + intel_edp_psr_setup_vsc(intel_dp); + + /* Avoid continuous PSR exit by masking memup and hpd */ + I915_WRITE(EDP_PSR_DEBUG_CTL(dev), EDP_PSR_DEBUG_MASK_MEMUP | + EDP_PSR_DEBUG_MASK_HPD | EDP_PSR_DEBUG_MASK_LPSP); if (intel_edp_psr_match_conditions(intel_dp)) dev_priv->psr.enabled = intel_dp; -- cgit v0.10.2 From 0aa48783979d40a1de59eb5b04989e1398400c3c Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 16 Sep 2014 19:19:06 -0400 Subject: drm/i915: PSR: Organize PSR enable function We don't need to setup everything else if it doesn't match all conditions. Signed-off-by: Rodrigo Vivi Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index e0066c0..43b381a 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2148,10 +2148,12 @@ void intel_edp_psr_enable(struct intel_dp *intel_dp) mutex_lock(&dev_priv->psr.lock); if (dev_priv->psr.enabled) { DRM_DEBUG_KMS("PSR already in use\n"); - mutex_unlock(&dev_priv->psr.lock); - return; + goto unlock; } + if (!intel_edp_psr_match_conditions(intel_dp)) + goto unlock; + dev_priv->psr.busy_frontbuffer_bits = 0; intel_edp_psr_setup_vsc(intel_dp); @@ -2160,8 +2162,8 @@ void intel_edp_psr_enable(struct intel_dp *intel_dp) I915_WRITE(EDP_PSR_DEBUG_CTL(dev), EDP_PSR_DEBUG_MASK_MEMUP | EDP_PSR_DEBUG_MASK_HPD | EDP_PSR_DEBUG_MASK_LPSP); - if (intel_edp_psr_match_conditions(intel_dp)) - dev_priv->psr.enabled = intel_dp; + dev_priv->psr.enabled = intel_dp; +unlock: mutex_unlock(&dev_priv->psr.lock); } -- cgit v0.10.2 From 72bbf0af0c76cbefe9cecbd2ed670b7555e03625 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 13 Feb 2013 15:27:37 +0000 Subject: drm/i915/skl: Add the Skylake PCI ids v2: Squash in 2nd patch from Damien for more ids (Daniel) v3: info->has*ring -> info->ring_mask conversion. Also add VEBOX support. v4: Fold in update from Damien v5: Rebase and add GEN_DEFAULT_PIPEOFFSETS v6: Add more PCI ID (Vandana) v7: Rebase and add IVB_CURSOR_OFFSETS v8: Renamed the macro from _PCI_IDS to _IDS for consistency Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Vandana Kannan Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 3870c73..bec58e8 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -356,6 +356,17 @@ static const struct intel_device_info intel_cherryview_info = { CURSOR_OFFSETS, }; +static const struct intel_device_info intel_skylake_info = { + .is_preliminary = 1, + .gen = 9, .num_pipes = 3, + .need_gfx_hws = 1, .has_hotplug = 1, + .ring_mask = RENDER_RING | BSD_RING | BLT_RING | VEBOX_RING, + .has_llc = 1, + .has_ddi = 1, + GEN_DEFAULT_PIPEOFFSETS, + IVB_CURSOR_OFFSETS, +}; + /* * Make sure any device matches here are from most specific to most * general. For example, since the Quanta match is based on the subsystem @@ -392,7 +403,8 @@ static const struct intel_device_info intel_cherryview_info = { INTEL_BDW_GT12D_IDS(&intel_broadwell_d_info), \ INTEL_BDW_GT3M_IDS(&intel_broadwell_gt3m_info), \ INTEL_BDW_GT3D_IDS(&intel_broadwell_gt3d_info), \ - INTEL_CHV_IDS(&intel_cherryview_info) + INTEL_CHV_IDS(&intel_cherryview_info), \ + INTEL_SKL_IDS(&intel_skylake_info) static const struct pci_device_id pciidlist[] = { /* aka */ INTEL_PCI_IDS, diff --git a/include/drm/i915_pciids.h b/include/drm/i915_pciids.h index a70d456..180ad0e 100644 --- a/include/drm/i915_pciids.h +++ b/include/drm/i915_pciids.h @@ -259,4 +259,21 @@ INTEL_VGA_DEVICE(0x22b2, info), \ INTEL_VGA_DEVICE(0x22b3, info) +#define INTEL_SKL_IDS(info) \ + INTEL_VGA_DEVICE(0x1916, info), /* ULT GT2 */ \ + INTEL_VGA_DEVICE(0x1906, info), /* ULT GT1 */ \ + INTEL_VGA_DEVICE(0x1926, info), /* ULT GT3 */ \ + INTEL_VGA_DEVICE(0x1921, info), /* ULT GT2F */ \ + INTEL_VGA_DEVICE(0x190E, info), /* ULX GT1 */ \ + INTEL_VGA_DEVICE(0x191E, info), /* ULX GT2 */ \ + INTEL_VGA_DEVICE(0x1912, info), /* DT GT2 */ \ + INTEL_VGA_DEVICE(0x1902, info), /* DT GT1 */ \ + INTEL_VGA_DEVICE(0x191B, info), /* Halo GT2 */ \ + INTEL_VGA_DEVICE(0x192B, info), /* Halo GT3 */ \ + INTEL_VGA_DEVICE(0x190B, info), /* Halo GT1 */ \ + INTEL_VGA_DEVICE(0x191A, info), /* SRV GT2 */ \ + INTEL_VGA_DEVICE(0x192A, info), /* SRV GT3 */ \ + INTEL_VGA_DEVICE(0x190A, info), /* SRV GT1 */ \ + INTEL_VGA_DEVICE(0x191D, info) /* WKS GT2 */ + #endif /* _I915_PCIIDS_H */ -- cgit v0.10.2 From b71252dc48697157e114fc12801f89398018b9ef Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 13 Feb 2013 15:27:24 +0000 Subject: drm/i915/skl: Add an IS_GEN9() define Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 19c0dd8..7892428d 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2123,6 +2123,7 @@ struct drm_i915_cmd_table { #define IS_GEN6(dev) (INTEL_INFO(dev)->gen == 6) #define IS_GEN7(dev) (INTEL_INFO(dev)->gen == 7) #define IS_GEN8(dev) (INTEL_INFO(dev)->gen == 8) +#define IS_GEN9(dev) (INTEL_INFO(dev)->gen == 9) #define RENDER_RING (1< Date: Wed, 2 Apr 2014 11:24:50 +0530 Subject: drm/i915/skl: Add an IS_SKYLAKE macro Adding new macro IS_SKYLAKE for skylake specific implementation. Reviewed-by: Thomas Wood Signed-off-by: Satheeshakrishna M Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index bec58e8..a1ba155 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -358,6 +358,7 @@ static const struct intel_device_info intel_cherryview_info = { static const struct intel_device_info intel_skylake_info = { .is_preliminary = 1, + .is_skylake = 1, .gen = 9, .num_pipes = 3, .need_gfx_hws = 1, .has_hotplug = 1, .ring_mask = RENDER_RING | BSD_RING | BLT_RING | VEBOX_RING, diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 7892428d..2c652e8 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -551,6 +551,7 @@ struct intel_uncore { func(is_ivybridge) sep \ func(is_valleyview) sep \ func(is_haswell) sep \ + func(is_skylake) sep \ func(is_preliminary) sep \ func(has_fbc) sep \ func(has_pipe_cxsr) sep \ @@ -2093,6 +2094,7 @@ struct drm_i915_cmd_table { #define IS_CHERRYVIEW(dev) (INTEL_INFO(dev)->is_valleyview && IS_GEN8(dev)) #define IS_HASWELL(dev) (INTEL_INFO(dev)->is_haswell) #define IS_BROADWELL(dev) (!INTEL_INFO(dev)->is_valleyview && IS_GEN8(dev)) +#define IS_SKYLAKE(dev) (INTEL_INFO(dev)->is_skylake) #define IS_MOBILE(dev) (INTEL_INFO(dev)->is_mobile) #define IS_HSW_EARLY_SDV(dev) (IS_HASWELL(dev) && \ (INTEL_DEVID(dev) & 0xFF00) == 0x0C00) -- cgit v0.10.2 From 043efb1111694981c99613b8f944d6eabfb93273 Mon Sep 17 00:00:00 2001 From: Daisy Sun Date: Wed, 23 Apr 2014 17:13:09 -0700 Subject: drm/i915/skl: SKL FBC enablement Enable FBC feature on Skylake Issue: VIZ-3788 Reviewed-by: Thomas Wood Signed-off-by: Daisy Sun Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index a1ba155..7088bde 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -364,6 +364,7 @@ static const struct intel_device_info intel_skylake_info = { .ring_mask = RENDER_RING | BSD_RING | BLT_RING | VEBOX_RING, .has_llc = 1, .has_ddi = 1, + .has_fbc = 1, GEN_DEFAULT_PIPEOFFSETS, IVB_CURSOR_OFFSETS, }; -- cgit v0.10.2 From 01209dd56edf77b02290bf1db7ac0bed32381235 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 13 Feb 2013 15:27:25 +0000 Subject: drm/i915/skl: Fence registers on SKL are the same as SNB v2: Rebased on top of the i915_gpu_error.c extraction. Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 4ca3a6d..2a38ef2 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3151,6 +3151,7 @@ static void i915_gem_write_fence(struct drm_device *dev, int reg, obj->stride, obj->tiling_mode); switch (INTEL_INFO(dev)->gen) { + case 9: case 8: case 7: case 6: diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 2c87a79..68fd767 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -765,6 +765,7 @@ static void i915_gem_record_fences(struct drm_device *dev, /* Fences */ switch (INTEL_INFO(dev)->gen) { + case 9: case 8: case 7: case 6: -- cgit v0.10.2 From da2078cd001e958409b6cef248679ff1772a3ad9 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 13 Feb 2013 15:27:27 +0000 Subject: drm/i915/skl: Provide a placeholder for init_clock_gating() v2: Rebase on top of the broadwell_init_clock_gating() name change Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 45f71e6..d760e06 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -66,6 +66,10 @@ * i915.i915_enable_fbc parameter */ +static void gen9_init_clock_gating(struct drm_device *dev) +{ +} + static void i8xx_disable_fbc(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -7397,6 +7401,8 @@ void intel_init_pm(struct drm_device *dev) dev_priv->display.init_clock_gating = haswell_init_clock_gating; else if (INTEL_INFO(dev)->gen == 8) dev_priv->display.init_clock_gating = broadwell_init_clock_gating; + else if (INTEL_INFO(dev)->gen == 9) + dev_priv->display.init_clock_gating = gen9_init_clock_gating; } else if (IS_CHERRYVIEW(dev)) { dev_priv->display.update_wm = cherryview_update_wm; dev_priv->display.update_sprite_wm = valleyview_update_sprite_wm; -- cgit v0.10.2 From 8a8b009d133714c0856ee08f7c68b908103e8383 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Mon, 8 Apr 2013 15:42:02 +0200 Subject: drm/i915/skl: Skylake shares the interrupt logic with Broadwell Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index c96ddc9..1d2849e 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -4713,7 +4713,7 @@ void intel_irq_init(struct drm_device *dev) dev->driver->enable_vblank = valleyview_enable_vblank; dev->driver->disable_vblank = valleyview_disable_vblank; dev_priv->display.hpd_irq_setup = i915_hpd_irq_setup; - } else if (IS_GEN8(dev)) { + } else if (INTEL_INFO(dev)->gen >= 8) { dev->driver->irq_handler = gen8_irq_handler; dev->driver->irq_preinstall = gen8_irq_reset; dev->driver->irq_postinstall = gen8_irq_postinstall; -- cgit v0.10.2 From fbdcb06880bf414afafd4053d0d9906725f8b117 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 13 Feb 2013 15:27:34 +0000 Subject: drm/i915/skl: don't set the AsyncFlip performance mode for Gen9+ The following sets the AsyncFlip performance mode for everything above Gen6: commit 4790cb36b3eede8fb0cca529dc1d31b9936fa24b Author: Chris Wilson Date: Sun Jan 20 16:11:20 2013 +0000 drm/i915: Disable AsyncFlip performance optimisations Starting from Gen9 the MI_MODE register layout changes and doesn't include the above bit. Reviewed-by: Thomas Wood Signed-off-by: Imre Deak Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 109de2e..a6a64ae 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -823,7 +823,7 @@ static int init_render_ring(struct intel_engine_cs *ring) * * WaDisableAsyncFlipPerfMode:snb,ivb,hsw,vlv,bdw,chv */ - if (INTEL_INFO(dev)->gen >= 6) + if (INTEL_INFO(dev)->gen >= 6 && INTEL_INFO(dev)->gen < 9) I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(ASYNC_FLIP_PERF_DISABLE)); /* Required for the hardware to program scanline values for waiting */ -- cgit v0.10.2 From 1fada4cc6a49fa8fb4530bba8fb0cfb42708fc6a Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 3 Jul 2013 21:06:02 +0100 Subject: drm/i915/skl: Framebuffers need to be aligned to 256KB on Skylake v2: Also align X tiled fbs to 256KB (Thomas) Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 1386086..dc264de 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2199,7 +2199,9 @@ intel_pin_and_fence_fb_obj(struct drm_device *dev, switch (obj->tiling_mode) { case I915_TILING_NONE: - if (IS_BROADWATER(dev) || IS_CRESTLINE(dev)) + if (INTEL_INFO(dev)->gen >= 9) + alignment = 256 * 1024; + else if (IS_BROADWATER(dev) || IS_CRESTLINE(dev)) alignment = 128 * 1024; else if (INTEL_INFO(dev)->gen >= 4) alignment = 4 * 1024; @@ -2207,8 +2209,12 @@ intel_pin_and_fence_fb_obj(struct drm_device *dev, alignment = 64 * 1024; break; case I915_TILING_X: - /* pin() will align the object as required by fence */ - alignment = 0; + if (INTEL_INFO(dev)->gen >= 9) + alignment = 256 * 1024; + else { + /* pin() will align the object as required by fence */ + alignment = 0; + } break; case I915_TILING_Y: WARN(1, "Y tiled bo slipped through, driver bug!\n"); -- cgit v0.10.2 From 5907f5fb6a16ae70cde25bcf4be201a93dfa0b9c Mon Sep 17 00:00:00 2001 From: Robert Beckett Date: Thu, 23 Jan 2014 14:23:14 +0000 Subject: drm/i915/skl: i915_swizzle_info gen9 fix Fix ARB_MODE register read for gen >= 8 in i915_swizzle_info Reviewed-by: Thomas Wood Signed-off-by: Robert Beckett Signed-off-by: Damien Lespiau diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 2cbc85f..707eddd 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1986,7 +1986,7 @@ static int i915_swizzle_info(struct seq_file *m, void *data) I915_READ(MAD_DIMM_C2)); seq_printf(m, "TILECTL = 0x%08x\n", I915_READ(TILECTL)); - if (IS_GEN8(dev)) + if (INTEL_INFO(dev)->gen >= 8) seq_printf(m, "GAMTARBMODE = 0x%08x\n", I915_READ(GAMTARBMODE)); else -- cgit v0.10.2 From 70d21f0e914415c39bfdffdcaebb6bc8fb6ce79c Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 3 Jul 2013 21:06:04 +0100 Subject: drm/i915/skl: Implement the new update_plane() for primary planes Skylake makes primary planes the same as sprite planes and call the result "universal planes". This commit emulates a primary plane with plane 0, taking the opportunity to redefine primary and sprite registers to be identical now that the underlying hardware is. It also makes sense as plenty of fields have changed. v2: Rebase on top of the vma code. v3: Follow upstream evolution: - Drop return values. - Remove pipe checks since redudant and BUG instead. - Remove tiling checks and BUG instead. - Drop commented out DISP_MODIFY usage. v4: s/plane/primary_plane/ v5: Misc fixes: - Fix the fields we need to clear up - Disable trickle feed - Correctly use PLANE_OFFSET for the panning v6: (Jesse) Use pipe src size when programming plane size. This makes cloned configs work correctly w/o the use of a panel fitter. v7: Rebase on top of Ville's rmw elimination series v8: Remove clearing the trickle feed bit now that we don't do a RMW (Rodrigo, Damien) Add a comment about the stride unit (Rodrigo) Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau (v1,5,6,7) Signed-off-by: Daniel Vetter (v2,3) Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index b65bdfc..1eaa341 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -26,8 +26,8 @@ #define _I915_REG_H_ #define _PIPE(pipe, a, b) ((a) + (pipe)*((b)-(a))) +#define _PLANE(plane, a, b) _PIPE(plane, a, b) #define _TRANSCODER(tran, a, b) ((a) + (tran)*((b)-(a))) - #define _PORT(port, a, b) ((a) + (port)*((b)-(a))) #define _PIPE3(pipe, a, b, c) ((pipe) == PIPE_A ? (a) : \ (pipe) == PIPE_B ? (b) : (c)) @@ -4499,6 +4499,114 @@ enum punit_power_well { #define SPCONSTALPHA(pipe, plane) _PIPE(pipe * 2 + plane, _SPACONSTALPHA, _SPBCONSTALPHA) #define SPGAMC(pipe, plane) _PIPE(pipe * 2 + plane, _SPAGAMC, _SPBGAMC) +/* Skylake plane registers */ + +#define _PLANE_CTL_1_A 0x70180 +#define _PLANE_CTL_2_A 0x70280 +#define _PLANE_CTL_3_A 0x70380 +#define PLANE_CTL_ENABLE (1 << 31) +#define PLANE_CTL_PIPE_GAMMA_ENABLE (1 << 30) +#define PLANE_CTL_FORMAT_MASK (0xf << 24) +#define PLANE_CTL_FORMAT_YUV422 ( 0 << 24) +#define PLANE_CTL_FORMAT_NV12 ( 1 << 24) +#define PLANE_CTL_FORMAT_XRGB_2101010 ( 2 << 24) +#define PLANE_CTL_FORMAT_XRGB_8888 ( 4 << 24) +#define PLANE_CTL_FORMAT_XRGB_16161616F ( 6 << 24) +#define PLANE_CTL_FORMAT_AYUV ( 8 << 24) +#define PLANE_CTL_FORMAT_INDEXED ( 12 << 24) +#define PLANE_CTL_FORMAT_RGB_565 ( 14 << 24) +#define PLANE_CTL_PIPE_CSC_ENABLE (1 << 23) +#define PLANE_CTL_KEY_ENABLE (1 << 22) +#define PLANE_CTL_ORDER_BGRX (0 << 20) +#define PLANE_CTL_ORDER_RGBX (1 << 20) +#define PLANE_CTL_YUV422_ORDER_MASK (0x3 << 16) +#define PLANE_CTL_YUV422_YUYV ( 0 << 16) +#define PLANE_CTL_YUV422_UYVY ( 1 << 16) +#define PLANE_CTL_YUV422_YVYU ( 2 << 16) +#define PLANE_CTL_YUV422_VYUY ( 3 << 16) +#define PLANE_CTL_DECOMPRESSION_ENABLE (1 << 15) +#define PLANE_CTL_TRICKLE_FEED_DISABLE (1 << 14) +#define PLANE_CTL_PLANE_GAMMA_DISABLE (1 << 13) +#define PLANE_CTL_TILED_MASK (0x7 << 10) +#define PLANE_CTL_TILED_LINEAR ( 0 << 10) +#define PLANE_CTL_TILED_X ( 1 << 10) +#define PLANE_CTL_TILED_Y ( 4 << 10) +#define PLANE_CTL_TILED_YF ( 5 << 10) +#define PLANE_CTL_ALPHA_MASK (0x3 << 4) +#define PLANE_CTL_ALPHA_DISABLE ( 0 << 4) +#define PLANE_CTL_ALPHA_SW_PREMULTIPLY ( 2 << 4) +#define PLANE_CTL_ALPHA_HW_PREMULTIPLY ( 3 << 4) +#define _PLANE_STRIDE_1_A 0x70188 +#define _PLANE_STRIDE_2_A 0x70288 +#define _PLANE_STRIDE_3_A 0x70388 +#define _PLANE_POS_1_A 0x7018c +#define _PLANE_POS_2_A 0x7028c +#define _PLANE_POS_3_A 0x7038c +#define _PLANE_SIZE_1_A 0x70190 +#define _PLANE_SIZE_2_A 0x70290 +#define _PLANE_SIZE_3_A 0x70390 +#define _PLANE_SURF_1_A 0x7019c +#define _PLANE_SURF_2_A 0x7029c +#define _PLANE_SURF_3_A 0x7039c +#define _PLANE_OFFSET_1_A 0x701a4 +#define _PLANE_OFFSET_2_A 0x702a4 +#define _PLANE_OFFSET_3_A 0x703a4 + +#define _PLANE_CTL_1_B 0x71180 +#define _PLANE_CTL_2_B 0x71280 +#define _PLANE_CTL_3_B 0x71380 +#define _PLANE_CTL_1(pipe) _PIPE(pipe, _PLANE_CTL_1_A, _PLANE_CTL_1_B) +#define _PLANE_CTL_2(pipe) _PIPE(pipe, _PLANE_CTL_2_A, _PLANE_CTL_2_B) +#define _PLANE_CTL_3(pipe) _PIPE(pipe, _PLANE_CTL_3_A, _PLANE_CTL_3_B) +#define PLANE_CTL(pipe, plane) \ + _PLANE(plane, _PLANE_CTL_1(pipe), _PLANE_CTL_2(pipe)) + +#define _PLANE_STRIDE_1_B 0x71188 +#define _PLANE_STRIDE_2_B 0x71288 +#define _PLANE_STRIDE_3_B 0x71388 +#define _PLANE_STRIDE_1(pipe) \ + _PIPE(pipe, _PLANE_STRIDE_1_A, _PLANE_STRIDE_1_B) +#define _PLANE_STRIDE_2(pipe) \ + _PIPE(pipe, _PLANE_STRIDE_2_A, _PLANE_STRIDE_2_B) +#define _PLANE_STRIDE_3(pipe) \ + _PIPE(pipe, _PLANE_STRIDE_3_A, _PLANE_STRIDE_3_B) +#define PLANE_STRIDE(pipe, plane) \ + _PLANE(plane, _PLANE_STRIDE_1(pipe), _PLANE_STRIDE_2(pipe)) + +#define _PLANE_POS_1_B 0x7118c +#define _PLANE_POS_2_B 0x7128c +#define _PLANE_POS_3_B 0x7138c +#define _PLANE_POS_1(pipe) _PIPE(pipe, _PLANE_POS_1_A, _PLANE_POS_1_B) +#define _PLANE_POS_2(pipe) _PIPE(pipe, _PLANE_POS_2_A, _PLANE_POS_2_B) +#define _PLANE_POS_3(pipe) _PIPE(pipe, _PLANE_POS_3_A, _PLANE_POS_3_B) +#define PLANE_POS(pipe, plane) \ + _PLANE(plane, _PLANE_POS_1(pipe), _PLANE_POS_2(pipe)) + +#define _PLANE_SIZE_1_B 0x71190 +#define _PLANE_SIZE_2_B 0x71290 +#define _PLANE_SIZE_3_B 0x71390 +#define _PLANE_SIZE_1(pipe) _PIPE(pipe, _PLANE_SIZE_1_A, _PLANE_SIZE_1_B) +#define _PLANE_SIZE_2(pipe) _PIPE(pipe, _PLANE_SIZE_2_A, _PLANE_SIZE_2_B) +#define _PLANE_SIZE_3(pipe) _PIPE(pipe, _PLANE_SIZE_3_A, _PLANE_SIZE_3_B) +#define PLANE_SIZE(pipe, plane) \ + _PLANE(plane, _PLANE_SIZE_1(pipe), _PLANE_SIZE_2(pipe)) + +#define _PLANE_SURF_1_B 0x7119c +#define _PLANE_SURF_2_B 0x7129c +#define _PLANE_SURF_3_B 0x7139c +#define _PLANE_SURF_1(pipe) _PIPE(pipe, _PLANE_SURF_1_A, _PLANE_SURF_1_B) +#define _PLANE_SURF_2(pipe) _PIPE(pipe, _PLANE_SURF_2_A, _PLANE_SURF_2_B) +#define _PLANE_SURF_3(pipe) _PIPE(pipe, _PLANE_SURF_3_A, _PLANE_SURF_3_B) +#define PLANE_SURF(pipe, plane) \ + _PLANE(plane, _PLANE_SURF_1(pipe), _PLANE_SURF_2(pipe)) + +#define _PLANE_OFFSET_1_B 0x711a4 +#define _PLANE_OFFSET_2_B 0x712a4 +#define _PLANE_OFFSET_1(pipe) _PIPE(pipe, _PLANE_OFFSET_1_A, _PLANE_OFFSET_1_B) +#define _PLANE_OFFSET_2(pipe) _PIPE(pipe, _PLANE_OFFSET_2_A, _PLANE_OFFSET_2_B) +#define PLANE_OFFSET(pipe, plane) \ + _PLANE(plane, _PLANE_OFFSET_1(pipe), _PLANE_OFFSET_2(pipe)) + /* VBIOS regs */ #define VGACNTRL 0x71400 # define VGA_DISP_DISABLE (1 << 31) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index dc264de..65f4ff6 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2644,6 +2644,90 @@ static void ironlake_update_primary_plane(struct drm_crtc *crtc, POSTING_READ(reg); } +static void skylake_update_primary_plane(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + int x, int y) +{ + struct drm_device *dev = crtc->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_framebuffer *intel_fb; + struct drm_i915_gem_object *obj; + int pipe = intel_crtc->pipe; + u32 plane_ctl, stride; + + if (!intel_crtc->primary_enabled) { + I915_WRITE(PLANE_CTL(pipe, 0), 0); + I915_WRITE(PLANE_SURF(pipe, 0), 0); + POSTING_READ(PLANE_CTL(pipe, 0)); + return; + } + + plane_ctl = PLANE_CTL_ENABLE | + PLANE_CTL_PIPE_GAMMA_ENABLE | + PLANE_CTL_PIPE_CSC_ENABLE; + + switch (fb->pixel_format) { + case DRM_FORMAT_RGB565: + plane_ctl |= PLANE_CTL_FORMAT_RGB_565; + break; + case DRM_FORMAT_XRGB8888: + plane_ctl |= PLANE_CTL_FORMAT_XRGB_8888; + break; + case DRM_FORMAT_XBGR8888: + plane_ctl |= PLANE_CTL_ORDER_RGBX; + plane_ctl |= PLANE_CTL_FORMAT_XRGB_8888; + break; + case DRM_FORMAT_XRGB2101010: + plane_ctl |= PLANE_CTL_FORMAT_XRGB_2101010; + break; + case DRM_FORMAT_XBGR2101010: + plane_ctl |= PLANE_CTL_ORDER_RGBX; + plane_ctl |= PLANE_CTL_FORMAT_XRGB_2101010; + break; + default: + BUG(); + } + + intel_fb = to_intel_framebuffer(fb); + obj = intel_fb->obj; + + /* + * The stride is either expressed as a multiple of 64 bytes chunks for + * linear buffers or in number of tiles for tiled buffers. + */ + switch (obj->tiling_mode) { + case I915_TILING_NONE: + stride = fb->pitches[0] >> 6; + break; + case I915_TILING_X: + plane_ctl |= PLANE_CTL_TILED_X; + stride = fb->pitches[0] >> 9; + break; + default: + BUG(); + } + + plane_ctl |= PLANE_CTL_PLANE_GAMMA_DISABLE; + + I915_WRITE(PLANE_CTL(pipe, 0), plane_ctl); + + DRM_DEBUG_KMS("Writing base %08lX %d,%d,%d,%d pitch=%d\n", + i915_gem_obj_ggtt_offset(obj), + x, y, fb->width, fb->height, + fb->pitches[0]); + + I915_WRITE(PLANE_POS(pipe, 0), 0); + I915_WRITE(PLANE_OFFSET(pipe, 0), (y << 16) | x); + I915_WRITE(PLANE_SIZE(pipe, 0), + (intel_crtc->config.pipe_src_h - 1) << 16 | + (intel_crtc->config.pipe_src_w - 1)); + I915_WRITE(PLANE_STRIDE(pipe, 0), stride); + I915_WRITE(PLANE_SURF(pipe, 0), i915_gem_obj_ggtt_offset(obj)); + + POSTING_READ(PLANE_SURF(pipe, 0)); +} + /* Assume fb object is pinned & idle & fenced and just update base pointers */ static int intel_pipe_set_base_atomic(struct drm_crtc *crtc, struct drm_framebuffer *fb, @@ -12581,8 +12665,12 @@ static void intel_init_display(struct drm_device *dev) dev_priv->display.crtc_enable = haswell_crtc_enable; dev_priv->display.crtc_disable = haswell_crtc_disable; dev_priv->display.off = ironlake_crtc_off; - dev_priv->display.update_primary_plane = - ironlake_update_primary_plane; + if (INTEL_INFO(dev)->gen >= 9) + dev_priv->display.update_primary_plane = + skylake_update_primary_plane; + else + dev_priv->display.update_primary_plane = + ironlake_update_primary_plane; } else if (HAS_PCH_SPLIT(dev)) { dev_priv->display.get_pipe_config = ironlake_get_pipe_config; dev_priv->display.get_plane_config = ironlake_get_plane_config; -- cgit v0.10.2 From 884497edd9e982eaf5c7e1216029625cec3c5f5e Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 3 Dec 2013 13:56:23 +0000 Subject: drm/i915/skl: Don't create a VGA connector on Skylake v2: Rebase on top of the intel_crt_present() addition v3: Fix rebase error (we were patching the wrong function) Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 65f4ff6..3c3656cd 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -12322,6 +12322,9 @@ static bool intel_crt_present(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; + if (INTEL_INFO(dev)->gen >= 9) + return false; + if (IS_ULT(dev)) return false; -- cgit v0.10.2 From ca3704555c1360b98c645a0c5f882a3011a225f0 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 3 Dec 2013 13:56:24 +0000 Subject: drm/i915/skl: Don't try to read out the PCH transcoder state if not present When the platform doesn't have a FDI link, don't try to read out the state of a potential PCH transcoder. Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau [danvet: Don't open-code HAS_FDI if there's only one place that needs it. Acked by Damien on irc.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 3c3656cd..40f73fc 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7867,7 +7867,8 @@ static void haswell_get_ddi_port_state(struct intel_crtc *crtc, * DDI E. So just check whether this pipe is wired to DDI E and whether * the PCH transcoder is on. */ - if ((port == PORT_E) && I915_READ(LPT_TRANSCONF) & TRANS_ENABLE) { + if (INTEL_INFO(dev)->gen < 9 && + (port == PORT_E) && I915_READ(LPT_TRANSCONF) & TRANS_ENABLE) { pipe_config->has_pch_encoder = true; tmp = I915_READ(FDI_RX_CTL(PIPE_A)); -- cgit v0.10.2 From 7f88e3af2047c008d5bef5632854c883c7d577fb Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 3 Dec 2013 13:56:25 +0000 Subject: drm/i915/skl: Program the DDI buffer translation tables A couple of things have changed compared to Broadwell: - Entry 9 is used for eDP - No more FDI v2: Update the translation values to latest specs. v3: Rebase on top of the BDW HDMI translation patch v4: Remove the low voltage edp tables, Rebase on top of the patch not writing the HDMI entry on eDP/FDI DDIs (Satheesh, Paulo). v5: Apply the / 2 fix for the number of HDMI entries (Satheesh) v6: Rebase on top of Jani's clean up for the DDI_BUF_TRANS tables v7: Restore the commit message that was mangled by error Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index b63d4fa..ba1103f 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -127,6 +127,32 @@ static const struct ddi_buf_trans bdw_ddi_translations_hdmi[] = { { 0x80FFFFFF, 0x001B0002 }, /* 9: 1000 1000 0 */ }; +static const struct ddi_buf_trans skl_ddi_translations_dp[] = { + { 0x00000018, 0x000000a0 }, + { 0x00004014, 0x00000098 }, + { 0x00006012, 0x00000088 }, + { 0x00008010, 0x00000080 }, + { 0x00000018, 0x00000098 }, + { 0x00004014, 0x00000088 }, + { 0x00006012, 0x00000080 }, + { 0x00000018, 0x00000088 }, + { 0x00004014, 0x00000080 }, +}; + +static const struct ddi_buf_trans skl_ddi_translations_hdmi[] = { + /* Idx NT mV T mV db */ + { 0x00000018, 0x000000a0 }, /* 0: 400 400 0 */ + { 0x00004014, 0x00000098 }, /* 1: 400 600 3.5 */ + { 0x00006012, 0x00000088 }, /* 2: 400 800 6 */ + { 0x00000018, 0x0000003c }, /* 3: 450 450 0 */ + { 0x00000018, 0x00000098 }, /* 4: 600 600 0 */ + { 0x00003015, 0x00000088 }, /* 5: 600 800 2.5 */ + { 0x00005013, 0x00000080 }, /* 6: 600 1000 4.5 */ + { 0x00000018, 0x00000088 }, /* 7: 800 800 0 */ + { 0x00000096, 0x00000080 }, /* 8: 800 1000 2 */ + { 0x00000018, 0x00000080 }, /* 9: 1200 1200 0 */ +}; + enum port intel_ddi_get_encoder_port(struct intel_encoder *intel_encoder) { struct drm_encoder *encoder = &intel_encoder->base; @@ -169,7 +195,14 @@ static void intel_prepare_ddi_buffers(struct drm_device *dev, enum port port) const struct ddi_buf_trans *ddi_translations_hdmi; const struct ddi_buf_trans *ddi_translations; - if (IS_BROADWELL(dev)) { + if (IS_SKYLAKE(dev)) { + ddi_translations_fdi = NULL; + ddi_translations_dp = skl_ddi_translations_dp; + ddi_translations_edp = skl_ddi_translations_dp; + ddi_translations_hdmi = skl_ddi_translations_hdmi; + n_hdmi_entries = ARRAY_SIZE(skl_ddi_translations_hdmi); + hdmi_800mV_0dB = 7; + } else if (IS_BROADWELL(dev)) { ddi_translations_fdi = bdw_ddi_translations_fdi; ddi_translations_dp = bdw_ddi_translations_dp; ddi_translations_edp = bdw_ddi_translations_edp; @@ -208,7 +241,10 @@ static void intel_prepare_ddi_buffers(struct drm_device *dev, enum port port) ddi_translations = ddi_translations_dp; break; case PORT_E: - ddi_translations = ddi_translations_fdi; + if (ddi_translations_fdi) + ddi_translations = ddi_translations_fdi; + else + ddi_translations = ddi_translations_dp; break; default: BUG(); -- cgit v0.10.2 From 5a9d1f1a8f62108828641c9224eb5d080db598c5 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 3 Dec 2013 13:56:26 +0000 Subject: drm/i915/skl: Add support for DP voltage swings and pre-emphasis They are similar to Haswell. v2: Rebased on top of drm-intel-nightly v3: Rebased on top of Sonika's DP train defines renaming Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 2a26774..581ba1d 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2840,7 +2840,9 @@ intel_dp_voltage_max(struct intel_dp *intel_dp) struct drm_device *dev = intel_dp_to_dev(intel_dp); enum port port = dp_to_dig_port(intel_dp)->port; - if (IS_VALLEYVIEW(dev)) + if (INTEL_INFO(dev)->gen >= 9) + return DP_TRAIN_VOLTAGE_SWING_LEVEL_2; + else if (IS_VALLEYVIEW(dev)) return DP_TRAIN_VOLTAGE_SWING_LEVEL_3; else if (IS_GEN7(dev) && port == PORT_A) return DP_TRAIN_VOLTAGE_SWING_LEVEL_2; @@ -2856,7 +2858,18 @@ intel_dp_pre_emphasis_max(struct intel_dp *intel_dp, uint8_t voltage_swing) struct drm_device *dev = intel_dp_to_dev(intel_dp); enum port port = dp_to_dig_port(intel_dp)->port; - if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { + if (INTEL_INFO(dev)->gen >= 9) { + switch (voltage_swing & DP_TRAIN_VOLTAGE_SWING_MASK) { + case DP_TRAIN_VOLTAGE_SWING_LEVEL_0: + return DP_TRAIN_PRE_EMPH_LEVEL_3; + case DP_TRAIN_VOLTAGE_SWING_LEVEL_1: + return DP_TRAIN_PRE_EMPH_LEVEL_2; + case DP_TRAIN_VOLTAGE_SWING_LEVEL_2: + return DP_TRAIN_PRE_EMPH_LEVEL_1; + default: + return DP_TRAIN_PRE_EMPH_LEVEL_0; + } + } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { switch (voltage_swing & DP_TRAIN_VOLTAGE_SWING_MASK) { case DP_TRAIN_VOLTAGE_SWING_LEVEL_0: return DP_TRAIN_PRE_EMPH_LEVEL_3; @@ -3338,7 +3351,7 @@ intel_dp_set_signal_levels(struct intel_dp *intel_dp, uint32_t *DP) uint32_t signal_levels, mask; uint8_t train_set = intel_dp->train_set[0]; - if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { + if (IS_HASWELL(dev) || IS_BROADWELL(dev) || INTEL_INFO(dev)->gen >= 9) { signal_levels = intel_hsw_signal_levels(train_set); mask = DDI_BUF_EMP_MASK; } else if (IS_CHERRYVIEW(dev)) { -- cgit v0.10.2 From 1b1aad754cb26fd073b82eb860126cb812157ae9 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 3 Dec 2013 13:56:29 +0000 Subject: drm/i915/skl: Skylake moves AUX_CTL from PCH to CPU So we can apply the old aux_ctl = dp_ctl + 0x10 rule again. Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 581ba1d..17c6910 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -925,7 +925,16 @@ intel_dp_aux_init(struct intel_dp *intel_dp, struct intel_connector *connector) BUG(); } - if (!HAS_DDI(dev)) + /* + * The AUX_CTL register is usually DP_CTL + 0x10. + * + * On Haswell and Broadwell though: + * - Both port A DDI_BUF_CTL and DDI_AUX_CTL are on the CPU + * - Port B/C/D AUX channels are on the PCH, DDI_BUF_CTL on the CPU + * + * Skylake moves AUX_CTL back next to DDI_BUF_CTL, on the CPU. + */ + if (!IS_HASWELL(dev) && !IS_BROADWELL(dev)) intel_dp->aux_ch_ctl_reg = intel_dp->output_reg + 0x10; intel_dp->aux.name = name; -- cgit v0.10.2 From 663750141ed953c42b6b0e3cfb2450453bc38151 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Thu, 9 Jan 2014 18:02:46 +0000 Subject: drm/i915/skl: Add the additional graphics stolen sizes Skylake introduces new stolen memory sizes starting at 0xf0 (4MB) and growing by 4MB increments from there. v2: Rebase on top of the early-quirk changes from Ville. v3: Rebase on top of the PCI_IDS/IDS macro rename Reviewed-by: Thomas Wood Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/arch/x86/kernel/early-quirks.c b/arch/x86/kernel/early-quirks.c index 2e1a685..fe9f0b7 100644 --- a/arch/x86/kernel/early-quirks.c +++ b/arch/x86/kernel/early-quirks.c @@ -455,6 +455,23 @@ struct intel_stolen_funcs { u32 (*base)(int num, int slot, int func, size_t size); }; +static size_t __init gen9_stolen_size(int num, int slot, int func) +{ + u16 gmch_ctrl; + + gmch_ctrl = read_pci_config_16(num, slot, func, SNB_GMCH_CTRL); + gmch_ctrl >>= BDW_GMCH_GMS_SHIFT; + gmch_ctrl &= BDW_GMCH_GMS_MASK; + + if (gmch_ctrl < 0xf0) + return gmch_ctrl << 25; /* 32 MB units */ + else + /* 4MB increments starting at 0xf0 for 4MB */ + return (gmch_ctrl - 0xf0 + 1) << 22; +} + +typedef size_t (*stolen_size_fn)(int num, int slot, int func); + static const struct intel_stolen_funcs i830_stolen_funcs __initconst = { .base = i830_stolen_base, .size = i830_stolen_size, @@ -490,6 +507,11 @@ static const struct intel_stolen_funcs gen8_stolen_funcs __initconst = { .size = gen8_stolen_size, }; +static const struct intel_stolen_funcs gen9_stolen_funcs __initconst = { + .base = intel_stolen_base, + .size = gen9_stolen_size, +}; + static const struct intel_stolen_funcs chv_stolen_funcs __initconst = { .base = intel_stolen_base, .size = chv_stolen_size, @@ -523,6 +545,7 @@ static const struct pci_device_id intel_stolen_ids[] __initconst = { INTEL_BDW_M_IDS(&gen8_stolen_funcs), INTEL_BDW_D_IDS(&gen8_stolen_funcs), INTEL_CHV_IDS(&chv_stolen_funcs), + INTEL_SKL_IDS(&gen9_stolen_funcs), }; static void __init intel_graphics_stolen(int num, int slot, int func) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 6f410cf..d78695d 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -1847,6 +1847,18 @@ static size_t chv_get_stolen_size(u16 gmch_ctrl) return (gmch_ctrl - 0x17 + 9) << 22; } +static size_t gen9_get_stolen_size(u16 gen9_gmch_ctl) +{ + gen9_gmch_ctl >>= BDW_GMCH_GMS_SHIFT; + gen9_gmch_ctl &= BDW_GMCH_GMS_MASK; + + if (gen9_gmch_ctl < 0xf0) + return gen9_gmch_ctl << 25; /* 32 MB units */ + else + /* 4MB increments starting at 0xf0 for 4MB */ + return (gen9_gmch_ctl - 0xf0 + 1) << 22; +} + static int ggtt_probe_common(struct drm_device *dev, size_t gtt_size) { @@ -1943,7 +1955,10 @@ static int gen8_gmch_probe(struct drm_device *dev, pci_read_config_word(dev->pdev, SNB_GMCH_CTRL, &snb_gmch_ctl); - if (IS_CHERRYVIEW(dev)) { + if (INTEL_INFO(dev)->gen >= 9) { + *stolen = gen9_get_stolen_size(snb_gmch_ctl); + gtt_size = gen8_get_total_gtt_size(snb_gmch_ctl); + } else if (IS_CHERRYVIEW(dev)) { *stolen = chv_get_stolen_size(snb_gmch_ctl); gtt_size = chv_get_total_gtt_size(snb_gmch_ctl); } else { -- cgit v0.10.2 From fb8aad4becbfdf913d09a41e7240ef1f58420f76 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Thu, 16 Jan 2014 16:42:32 +0000 Subject: drm/i915/skl: gen9 uses the same bind_vma() vfuncs as gen6+ Temporary plug a BUG() while waiting for a better solution. See: http://lists.freedesktop.org/archives/intel-gfx/2014-January/038132.html However Chris was looking at cleaning-up this as well, so went for the easy intermediate solution instead. Reviewed-by: Ben Widawsky Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index d78695d..688dd00 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -2130,6 +2130,7 @@ static struct i915_vma *__i915_gem_vma_create(struct drm_i915_gem_object *obj, vma->obj = obj; switch (INTEL_INFO(vm->dev)->gen) { + case 9: case 8: case 7: case 6: -- cgit v0.10.2 From b6b5e38323b276cae53cf34970d968084de701bc Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Mon, 20 Jan 2014 16:00:59 +0000 Subject: drm/i915/skl: Implement the get_aux_clock_divider() DP vfunc We need to provide a vfunc that will make the code in intel_dp_aux_ch() loop once to start the AUX transaction. The return value (clock divider) is unused on SKL, so just return 1. Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 17c6910..4d2f704 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -661,6 +661,16 @@ static uint32_t vlv_get_aux_clock_divider(struct intel_dp *intel_dp, int index) return index ? 0 : 100; } +static uint32_t skl_get_aux_clock_divider(struct intel_dp *intel_dp, int index) +{ + /* + * SKL doesn't need us to program the AUX clock divider (Hardware will + * derive the clock from CDCLK automatically). We still implement the + * get_aux_clock_divider vfunc to plug-in into the existing code. + */ + return index ? 0 : 1; +} + static uint32_t i9xx_get_aux_send_ctl(struct intel_dp *intel_dp, bool has_aux_irq, int send_bytes, @@ -5083,7 +5093,9 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, intel_dp->pps_pipe = INVALID_PIPE; /* intel_dp vfuncs */ - if (IS_VALLEYVIEW(dev)) + if (INTEL_INFO(dev)->gen >= 9) + intel_dp->get_aux_clock_divider = skl_get_aux_clock_divider; + else if (IS_VALLEYVIEW(dev)) intel_dp->get_aux_clock_divider = vlv_get_aux_clock_divider; else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) intel_dp->get_aux_clock_divider = hsw_get_aux_clock_divider; -- cgit v0.10.2 From b9ca5fadb3842a7a90934f3ef1795e95b382def1 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Mon, 20 Jan 2014 16:01:00 +0000 Subject: drm/i915/skl: Provide a get_aux_send_ctl() vfunc for skylake Skylake doesn't use the pre-charge field now, but, instead, we need to specify the total number of SYNC pulses for the SYNC phase (pre-charge + SYNC pattern pules). Let's use the default value (32) for that. v3: increase DP AUX TX timeout as 400us is not to be used on SKL apparently (Jesse). Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 1eaa341..d4af99e 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -3634,6 +3634,7 @@ enum punit_power_well { #define DP_AUX_CH_CTL_PRECHARGE_TEST (1 << 11) #define DP_AUX_CH_CTL_BIT_CLOCK_2X_MASK (0x7ff) #define DP_AUX_CH_CTL_BIT_CLOCK_2X_SHIFT 0 +#define DP_AUX_CH_CTL_SYNC_PULSE_SKL(c) ((c) - 1) /* * Computing GMCH M and N values for the Display Port link diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 4d2f704..2e9a6ab 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -701,6 +701,21 @@ static uint32_t i9xx_get_aux_send_ctl(struct intel_dp *intel_dp, (aux_clock_divider << DP_AUX_CH_CTL_BIT_CLOCK_2X_SHIFT); } +static uint32_t skl_get_aux_send_ctl(struct intel_dp *intel_dp, + bool has_aux_irq, + int send_bytes, + uint32_t unused) +{ + return DP_AUX_CH_CTL_SEND_BUSY | + DP_AUX_CH_CTL_DONE | + (has_aux_irq ? DP_AUX_CH_CTL_INTERRUPT : 0) | + DP_AUX_CH_CTL_TIME_OUT_ERROR | + DP_AUX_CH_CTL_TIME_OUT_1600us | + DP_AUX_CH_CTL_RECEIVE_ERROR | + (send_bytes << DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT) | + DP_AUX_CH_CTL_SYNC_PULSE_SKL(32); +} + static int intel_dp_aux_ch(struct intel_dp *intel_dp, uint8_t *send, int send_bytes, @@ -5104,7 +5119,10 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, else intel_dp->get_aux_clock_divider = i9xx_get_aux_clock_divider; - intel_dp->get_aux_send_ctl = i9xx_get_aux_send_ctl; + if (INTEL_INFO(dev)->gen >= 9) + intel_dp->get_aux_send_ctl = skl_get_aux_send_ctl; + else + intel_dp->get_aux_send_ctl = i9xx_get_aux_send_ctl; /* Preserve the current hw state. */ intel_dp->DP = I915_READ(intel_dp->output_reg); -- cgit v0.10.2 From 3fdcf80f75814fe2f28db43771b50f9aa70d43b6 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Thu, 23 Jan 2014 13:59:49 +0000 Subject: drm/i915/skl: Initialize PPGTT like gen8 gen9 uses very similar memory management to what gen8 has. Just follow the flow. v2: Fix trivial conflict (Damien) Reviewed-by: Rodrigo Vivi Reviewed-by: Ben Widawsky Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 688dd00..66e4b2b 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -1092,7 +1092,7 @@ static int __hw_ppgtt_init(struct drm_device *dev, struct i915_hw_ppgtt *ppgtt) if (INTEL_INFO(dev)->gen < 8) return gen6_ppgtt_init(ppgtt); - else if (IS_GEN8(dev)) + else if (IS_GEN8(dev) || IS_GEN9(dev)) return gen8_ppgtt_init(ppgtt, dev_priv->gtt.base.total); else BUG(); -- cgit v0.10.2 From c3f59a67ad6ff5ac11fefc0c0d9d0f12ff7eb226 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Sun, 30 Mar 2014 16:28:23 +0100 Subject: drm/i915/skl: Allow the reg_read ioctl to return RCS_TIMESTAMP Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 918b761..3b27fb0 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -968,7 +968,7 @@ static const struct register_whitelist { /* supported gens, 0x10 for 4, 0x30 for 4 and 5, etc. */ uint32_t gen_bitmask; } whitelist[] = { - { RING_TIMESTAMP(RENDER_RING_BASE), 8, GEN_RANGE(4, 8) }, + { RING_TIMESTAMP(RENDER_RING_BASE), 8, GEN_RANGE(4, 9) }, }; int i915_reg_read_ioctl(struct drm_device *dev, -- cgit v0.10.2 From 2fcdcd8a2eaa70d110fb27bd297ef4aea3b291fa Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Thu, 3 Apr 2014 16:13:29 +0100 Subject: drm/i915/skl: report the same INSTDONE registers as gen8 Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 68fd767..1bf6e69 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -1390,6 +1390,7 @@ void i915_get_extra_instdone(struct drm_device *dev, uint32_t *instdone) WARN_ONCE(1, "Unsupported platform\n"); case 7: case 8: + case 9: instdone[0] = I915_READ(GEN7_INSTDONE_1); instdone[1] = I915_READ(GEN7_SC_INSTDONE); instdone[2] = I915_READ(GEN7_SAMPLER_INSTDONE); -- cgit v0.10.2 From 2a9b7539667bb347502d7c7e00cae30479c164fb Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Thu, 3 Apr 2014 16:17:51 +0100 Subject: drm/i915/skl: Report the PDP regs as in gen8 Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 1bf6e69..fe14980 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -924,6 +924,7 @@ static void i915_record_ring_state(struct drm_device *dev, ering->vm_info.gfx_mode = I915_READ(RING_MODE_GEN7(ring)); switch (INTEL_INFO(dev)->gen) { + case 9: case 8: for (i = 0; i < 4; i++) { ering->vm_info.pdp[i] = -- cgit v0.10.2 From a440ca6f2859bc64525dfbb437f031f43163d5ae Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 4 Jun 2014 19:25:23 +0100 Subject: drm/i915/skl: SKL shares the same underrun interrupt as BDW Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 1d2849e..a8e0bc4 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -503,7 +503,7 @@ static bool __intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, ironlake_set_fifo_underrun_reporting(dev, pipe, enable); else if (IS_GEN7(dev)) ivybridge_set_fifo_underrun_reporting(dev, pipe, enable, old); - else if (IS_GEN8(dev)) + else if (IS_GEN8(dev) || IS_GEN9(dev)) broadwell_set_fifo_underrun_reporting(dev, pipe, enable); return old; -- cgit v0.10.2 From 3cdf122ccfd150e022247c2709c0737b9fb235f1 Mon Sep 17 00:00:00 2001 From: Satheeshakrishna M Date: Tue, 8 Apr 2014 15:46:53 +0530 Subject: drm/i915/skl: SKL pipe misc programming Pipe misc programming in gen9 is similar to BDW. Extending the BDW implementation to gen 9. Reviewed-by: Rodrigo Vivi Signed-off-by: Satheeshakrishna M Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 40f73fc..901c035 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7064,7 +7064,7 @@ static void haswell_set_pipeconf(struct drm_crtc *crtc) I915_WRITE(GAMMA_MODE(intel_crtc->pipe), GAMMA_MODE_MODE_8BIT); POSTING_READ(GAMMA_MODE(intel_crtc->pipe)); - if (IS_BROADWELL(dev)) { + if (IS_BROADWELL(dev) || INTEL_INFO(dev)->gen >= 9) { val = 0; switch (intel_crtc->config.pipe_bpp) { -- cgit v0.10.2 From 02c292596af4785091a851df441f3c6c27099a29 Mon Sep 17 00:00:00 2001 From: Satheeshakrishna M Date: Tue, 8 Apr 2014 15:46:54 +0530 Subject: drm/i915/skl: vfuncs for skl eld and global resource Set gen 9 function pointers for eld write and global resource. Implementation remains same as HSW. v2: Rebase on top of Sonika's untangling of the if/else ladder (Damien) Reviewed-by: Rodrigo Vivi Signed-off-by: Satheeshakrishna M (v1) Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 901c035..2087e8b 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -12758,6 +12758,10 @@ static void intel_init_display(struct drm_device *dev) dev_priv->display.modeset_global_resources = valleyview_modeset_global_resources; dev_priv->display.write_eld = ironlake_write_eld; + } else if (INTEL_INFO(dev)->gen >= 9) { + dev_priv->display.write_eld = haswell_write_eld; + dev_priv->display.modeset_global_resources = + haswell_modeset_global_resources; } /* Default just returns -ENODEV to indicate unsupported */ -- cgit v0.10.2 From 7879a7ebff7b5e9313ef6c28019a0bd62046b10d Mon Sep 17 00:00:00 2001 From: Satheeshakrishna M Date: Tue, 8 Apr 2014 15:46:55 +0530 Subject: drm/i915/skl: SKL backlight enabling Extending the BDW backlight implementation to SKL. Reviewed-by: Rodrigo Vivi Signed-off-by: Satheeshakrishna M Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c index 1878447..d69a3cb6 100644 --- a/drivers/gpu/drm/i915/intel_panel.c +++ b/drivers/gpu/drm/i915/intel_panel.c @@ -1317,7 +1317,7 @@ void intel_panel_init_backlight_funcs(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - if (IS_BROADWELL(dev)) { + if (IS_BROADWELL(dev) || (INTEL_INFO(dev)->gen >= 9)) { dev_priv->display.setup_backlight = bdw_setup_backlight; dev_priv->display.enable_backlight = bdw_enable_backlight; dev_priv->display.disable_backlight = pch_disable_backlight; -- cgit v0.10.2 From 08524a9ffa396c56ff3fbec9cfd80edd3fa6a058 Mon Sep 17 00:00:00 2001 From: Satheeshakrishna M Date: Tue, 8 Apr 2014 15:46:56 +0530 Subject: drm/i915/skl: Restore pipe B/C interrupts Extending BDW implementation to gen9. Pipe B/C interrupt restoration after exiting LPSP. v2: Fix minor rebasing conflict. Reviewed-by: Rodrigo Vivi Signed-off-by: Satheeshakrishna M Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index d760e06..fa87f1e 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6277,7 +6277,7 @@ static void hsw_power_well_post_enable(struct drm_i915_private *dev_priv) outb(inb(VGA_MSR_READ), VGA_MSR_WRITE); vga_put(dev->pdev, VGA_RSRC_LEGACY_IO); - if (IS_BROADWELL(dev)) + if (IS_BROADWELL(dev) || (INTEL_INFO(dev)->gen >= 9)) gen8_irq_power_well_post_enable(dev_priv); } -- cgit v0.10.2 From 770de83dc0b4f13b3ec1a923b1769ed668221779 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Thu, 20 Mar 2014 20:45:01 +0000 Subject: drm/i915/skl: Adjust the display engine interrupts To accomodate the extra planes, the bit definitions were shuffled around a bit. v2: Rebase on top of the for_each_pipe() change adding dev_priv as first argument. v3: Rebase after yet another change int that area (done with wiggle) Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index a8e0bc4..ba5ba63 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -2587,7 +2587,7 @@ static irqreturn_t gen8_irq_handler(int irq, void *arg) } for_each_pipe(dev_priv, pipe) { - uint32_t pipe_iir; + uint32_t pipe_iir, flip_done = 0, fault_errors = 0; if (!(master_ctl & GEN8_DE_PIPE_IRQ(pipe))) continue; @@ -2596,11 +2596,17 @@ static irqreturn_t gen8_irq_handler(int irq, void *arg) if (pipe_iir) { ret = IRQ_HANDLED; I915_WRITE(GEN8_DE_PIPE_IIR(pipe), pipe_iir); + if (pipe_iir & GEN8_PIPE_VBLANK && intel_pipe_handle_vblank(dev, pipe)) intel_check_page_flip(dev, pipe); - if (pipe_iir & GEN8_PIPE_PRIMARY_FLIP_DONE) { + if (IS_GEN9(dev)) + flip_done = pipe_iir & GEN9_PIPE_PLANE1_FLIP_DONE; + else + flip_done = pipe_iir & GEN8_PIPE_PRIMARY_FLIP_DONE; + + if (flip_done) { intel_prepare_page_flip(dev, pipe); intel_finish_page_flip_plane(dev, pipe); } @@ -2615,11 +2621,16 @@ static irqreturn_t gen8_irq_handler(int irq, void *arg) pipe_name(pipe)); } - if (pipe_iir & GEN8_DE_PIPE_IRQ_FAULT_ERRORS) { + + if (IS_GEN9(dev)) + fault_errors = pipe_iir & GEN9_DE_PIPE_IRQ_FAULT_ERRORS; + else + fault_errors = pipe_iir & GEN8_DE_PIPE_IRQ_FAULT_ERRORS; + + if (fault_errors) DRM_ERROR("Fault errors on pipe %c\n: 0x%08x", pipe_name(pipe), pipe_iir & GEN8_DE_PIPE_IRQ_FAULT_ERRORS); - } } else DRM_ERROR("The master control interrupt lied (DE PIPE)!\n"); } @@ -3803,12 +3814,20 @@ static void gen8_gt_irq_postinstall(struct drm_i915_private *dev_priv) static void gen8_de_irq_postinstall(struct drm_i915_private *dev_priv) { - uint32_t de_pipe_masked = GEN8_PIPE_PRIMARY_FLIP_DONE | - GEN8_PIPE_CDCLK_CRC_DONE | - GEN8_DE_PIPE_IRQ_FAULT_ERRORS; - uint32_t de_pipe_enables = de_pipe_masked | GEN8_PIPE_VBLANK | - GEN8_PIPE_FIFO_UNDERRUN; + uint32_t de_pipe_masked = GEN8_PIPE_CDCLK_CRC_DONE; + uint32_t de_pipe_enables; int pipe; + + if (IS_GEN9(dev_priv)) + de_pipe_masked |= GEN9_PIPE_PLANE1_FLIP_DONE | + GEN9_DE_PIPE_IRQ_FAULT_ERRORS; + else + de_pipe_masked |= GEN8_PIPE_PRIMARY_FLIP_DONE | + GEN8_DE_PIPE_IRQ_FAULT_ERRORS; + + de_pipe_enables = de_pipe_masked | GEN8_PIPE_VBLANK | + GEN8_PIPE_FIFO_UNDERRUN; + dev_priv->de_irq_mask[PIPE_A] = ~de_pipe_masked; dev_priv->de_irq_mask[PIPE_B] = ~de_pipe_masked; dev_priv->de_irq_mask[PIPE_C] = ~de_pipe_masked; diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index d4af99e..5060b51 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4844,10 +4844,23 @@ enum punit_power_well { #define GEN8_PIPE_SCAN_LINE_EVENT (1 << 2) #define GEN8_PIPE_VSYNC (1 << 1) #define GEN8_PIPE_VBLANK (1 << 0) +#define GEN9_PIPE_CURSOR_FAULT (1 << 11) +#define GEN9_PIPE_PLANE3_FAULT (1 << 9) +#define GEN9_PIPE_PLANE2_FAULT (1 << 8) +#define GEN9_PIPE_PLANE1_FAULT (1 << 7) +#define GEN9_PIPE_PLANE3_FLIP_DONE (1 << 5) +#define GEN9_PIPE_PLANE2_FLIP_DONE (1 << 4) +#define GEN9_PIPE_PLANE1_FLIP_DONE (1 << 3) +#define GEN9_PIPE_PLANE_FLIP_DONE(p) (1 << (3 + p)) #define GEN8_DE_PIPE_IRQ_FAULT_ERRORS \ (GEN8_PIPE_CURSOR_FAULT | \ GEN8_PIPE_SPRITE_FAULT | \ GEN8_PIPE_PRIMARY_FAULT) +#define GEN9_DE_PIPE_IRQ_FAULT_ERRORS \ + (GEN9_PIPE_CURSOR_FAULT | \ + GEN9_PIPE_PLANE3_FAULT | \ + GEN9_PIPE_PLANE2_FAULT | \ + GEN9_PIPE_PLANE1_FAULT) #define GEN8_DE_PORT_ISR 0x44440 #define GEN8_DE_PORT_IMR 0x44444 -- cgit v0.10.2 From e7e7ea20297afae464bb7da69a2e4c67d1d2b354 Mon Sep 17 00:00:00 2001 From: Satheeshakrishna M Date: Wed, 9 Apr 2014 11:08:57 +0530 Subject: drm/i915/skl: Sunrise Point PCH detection This patch implements detection of SPT and SPT-LP PCH devices. v2: Added HAS_PCH_SPT macro Reviewed-by: Rodrigo Vivi Signed-off-by: Satheeshakrishna M Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 7088bde..51fbb34 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -475,6 +475,16 @@ void intel_detect_pch(struct drm_device *dev) DRM_DEBUG_KMS("Found LynxPoint LP PCH\n"); WARN_ON(!IS_HASWELL(dev)); WARN_ON(!IS_ULT(dev)); + } else if (id == INTEL_PCH_SPT_DEVICE_ID_TYPE) { + dev_priv->pch_type = PCH_SPT; + DRM_DEBUG_KMS("Found SunrisePoint PCH\n"); + WARN_ON(!IS_SKYLAKE(dev)); + WARN_ON(IS_ULT(dev)); + } else if (id == INTEL_PCH_SPT_LP_DEVICE_ID_TYPE) { + dev_priv->pch_type = PCH_SPT; + DRM_DEBUG_KMS("Found SunrisePoint LP PCH\n"); + WARN_ON(!IS_SKYLAKE(dev)); + WARN_ON(!IS_ULT(dev)); } else continue; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 2c652e8..a8d4e23 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -705,6 +705,7 @@ enum intel_pch { PCH_IBX, /* Ibexpeak PCH */ PCH_CPT, /* Cougarpoint PCH */ PCH_LPT, /* Lynxpoint PCH */ + PCH_SPT, /* Sunrisepoint PCH */ PCH_NOP, }; @@ -2191,8 +2192,11 @@ struct drm_i915_cmd_table { #define INTEL_PCH_PPT_DEVICE_ID_TYPE 0x1e00 #define INTEL_PCH_LPT_DEVICE_ID_TYPE 0x8c00 #define INTEL_PCH_LPT_LP_DEVICE_ID_TYPE 0x9c00 +#define INTEL_PCH_SPT_DEVICE_ID_TYPE 0xA100 +#define INTEL_PCH_SPT_LP_DEVICE_ID_TYPE 0x9D00 #define INTEL_PCH_TYPE(dev) (to_i915(dev)->pch_type) +#define HAS_PCH_SPT(dev) (INTEL_PCH_TYPE(dev) == PCH_SPT) #define HAS_PCH_LPT(dev) (INTEL_PCH_TYPE(dev) == PCH_LPT) #define HAS_PCH_CPT(dev) (INTEL_PCH_TYPE(dev) == PCH_CPT) #define HAS_PCH_IBX(dev) (INTEL_PCH_TYPE(dev) == PCH_IBX) -- cgit v0.10.2 From acd5c3464038a550359433362055849e83351d89 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 26 Mar 2014 16:55:46 +0000 Subject: drm/i915/skl: Implement WaDisableSDEUnitClockGating:skl Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index fa87f1e..81465e6 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -68,6 +68,14 @@ static void gen9_init_clock_gating(struct drm_device *dev) { + struct drm_i915_private *dev_priv = dev->dev_private; + + /* + * WaDisableSDEUnitClockGating:skl + * This seems to be a pre-production w/a. + */ + I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) | + GEN8_SDEUNIT_CLOCK_GATE_DISABLE); } static void i8xx_disable_fbc(struct drm_device *dev) -- cgit v0.10.2 From 91e41d1665c868a26a24580b83c321e04be43dfe Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 26 Mar 2014 17:42:50 +0000 Subject: drm/i915/skl: Implement Wa4x4STCOptimizationDisable:skl Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 81465e6..64eb3b8 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -76,6 +76,10 @@ static void gen9_init_clock_gating(struct drm_device *dev) */ I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) | GEN8_SDEUNIT_CLOCK_GATE_DISABLE); + + /* Wa4x4STCOptimizationDisable:skl */ + I915_WRITE(CACHE_MODE_1, + _MASKED_BIT_ENABLE(GEN8_4x4_STC_OPTIMIZATION_DISABLE)); } static void i8xx_disable_fbc(struct drm_device *dev) -- cgit v0.10.2 From 3ca5da43003a5bd0ef1d4c15d015c77411e0df0d Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 26 Mar 2014 18:18:01 +0000 Subject: drm/i915/skl: Implement WaDisableDgMirrorFixInHalfSliceChicken5:skl Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 5060b51..c5c3ee3 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5866,6 +5866,9 @@ enum punit_power_well { #define GEN7_SINGLE_SUBSCAN_DISPATCH_ENABLE (1<<10) #define GEN7_PSD_SINGLE_PORT_DISPATCH_ENABLE (1<<3) +#define GEN9_HALF_SLICE_CHICKEN5 0xe188 +#define GEN9_DG_MIRROR_FIX_ENABLE (1<<5) + #define GEN8_ROW_CHICKEN 0xe4f0 #define PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE (1<<8) #define STALL_DOP_GATING_DISABLE (1<<5) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 64eb3b8..4f5dcf5 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -77,6 +77,14 @@ static void gen9_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) | GEN8_SDEUNIT_CLOCK_GATE_DISABLE); + /* + * WaDisableDgMirrorFixInHalfSliceChicken5:skl + * This is a pre-production w/a. + */ + I915_WRITE(GEN9_HALF_SLICE_CHICKEN5, + I915_READ(GEN9_HALF_SLICE_CHICKEN5) & + ~GEN9_DG_MIRROR_FIX_ENABLE); + /* Wa4x4STCOptimizationDisable:skl */ I915_WRITE(CACHE_MODE_1, _MASKED_BIT_ENABLE(GEN8_4x4_STC_OPTIMIZATION_DISABLE)); -- cgit v0.10.2 From 1fc8ac3e5c355fb0f2864f5deebba3c7d77b4f21 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 12 Feb 2014 19:13:31 +0000 Subject: drm/i915/skl: Skylake has 2 "sprite" planes per pipe On Skylake, we use plane1 as primary plane and plane2/3 as sprite planes. v2: Rebase on top of the for_each_pipe() change adding dev_priv as first argument. Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 1403b01..f64050d 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1534,7 +1534,7 @@ static void intel_device_info_runtime_init(struct drm_device *dev) info = (struct intel_device_info *)&dev_priv->info; - if (IS_VALLEYVIEW(dev)) + if (IS_VALLEYVIEW(dev) || INTEL_INFO(dev)->gen == 9) for_each_pipe(dev_priv, pipe) info->num_sprites[pipe] = 2; else -- cgit v0.10.2 From dc2a41b4cd2b54422b788e4a6f35aacd4fa1fea3 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 4 Dec 2013 00:49:41 +0000 Subject: drm/i915/skl: Implement drm_plane vfuncs SKL Uses the same hardware for all planes now, so called "universal" planes. Ie both the primary planes and sprite planes share the same logic. This patch implements the drm_plane vfuncs for "sprites" ie planes that aren't the primary plane. v2: Couple of fixes: - Actually enabled the planes and fix the plane number Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index c5c3ee3..bc14aaa 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4517,7 +4517,9 @@ enum punit_power_well { #define PLANE_CTL_FORMAT_INDEXED ( 12 << 24) #define PLANE_CTL_FORMAT_RGB_565 ( 14 << 24) #define PLANE_CTL_PIPE_CSC_ENABLE (1 << 23) -#define PLANE_CTL_KEY_ENABLE (1 << 22) +#define PLANE_CTL_KEY_ENABLE_MASK (0x3 << 21) +#define PLANE_CTL_KEY_ENABLE_SOURCE ( 1 << 21) +#define PLANE_CTL_KEY_ENABLE_DESTINATION ( 2 << 21) #define PLANE_CTL_ORDER_BGRX (0 << 20) #define PLANE_CTL_ORDER_RGBX (1 << 20) #define PLANE_CTL_YUV422_ORDER_MASK (0x3 << 16) @@ -4552,6 +4554,12 @@ enum punit_power_well { #define _PLANE_OFFSET_1_A 0x701a4 #define _PLANE_OFFSET_2_A 0x702a4 #define _PLANE_OFFSET_3_A 0x703a4 +#define _PLANE_KEYVAL_1_A 0x70194 +#define _PLANE_KEYVAL_2_A 0x70294 +#define _PLANE_KEYMSK_1_A 0x70198 +#define _PLANE_KEYMSK_2_A 0x70298 +#define _PLANE_KEYMAX_1_A 0x701a0 +#define _PLANE_KEYMAX_2_A 0x702a0 #define _PLANE_CTL_1_B 0x71180 #define _PLANE_CTL_2_B 0x71280 @@ -4608,6 +4616,27 @@ enum punit_power_well { #define PLANE_OFFSET(pipe, plane) \ _PLANE(plane, _PLANE_OFFSET_1(pipe), _PLANE_OFFSET_2(pipe)) +#define _PLANE_KEYVAL_1_B 0x71194 +#define _PLANE_KEYVAL_2_B 0x71294 +#define _PLANE_KEYVAL_1(pipe) _PIPE(pipe, _PLANE_KEYVAL_1_A, _PLANE_KEYVAL_1_B) +#define _PLANE_KEYVAL_2(pipe) _PIPE(pipe, _PLANE_KEYVAL_2_A, _PLANE_KEYVAL_2_B) +#define PLANE_KEYVAL(pipe, plane) \ + _PLANE(plane, _PLANE_KEYVAL_1(pipe), _PLANE_KEYVAL_2(pipe)) + +#define _PLANE_KEYMSK_1_B 0x71198 +#define _PLANE_KEYMSK_2_B 0x71298 +#define _PLANE_KEYMSK_1(pipe) _PIPE(pipe, _PLANE_KEYMSK_1_A, _PLANE_KEYMSK_1_B) +#define _PLANE_KEYMSK_2(pipe) _PIPE(pipe, _PLANE_KEYMSK_2_A, _PLANE_KEYMSK_2_B) +#define PLANE_KEYMSK(pipe, plane) \ + _PLANE(plane, _PLANE_KEYMSK_1(pipe), _PLANE_KEYMSK_2(pipe)) + +#define _PLANE_KEYMAX_1_B 0x711a0 +#define _PLANE_KEYMAX_2_B 0x712a0 +#define _PLANE_KEYMAX_1(pipe) _PIPE(pipe, _PLANE_KEYMAX_1_A, _PLANE_KEYMAX_1_B) +#define _PLANE_KEYMAX_2(pipe) _PIPE(pipe, _PLANE_KEYMAX_2_A, _PLANE_KEYMAX_2_B) +#define PLANE_KEYMAX(pipe, plane) \ + _PLANE(plane, _PLANE_KEYMAX_1(pipe), _PLANE_KEYMAX_2(pipe)) + /* VBIOS regs */ #define VGACNTRL 0x71400 # define VGA_DISP_DISABLE (1 << 31) diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 07a74ef..57e7190 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -139,6 +139,184 @@ static void intel_update_primary_plane(struct intel_crtc *crtc) } static void +skl_update_plane(struct drm_plane *drm_plane, struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_i915_gem_object *obj, int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t x, uint32_t y, + uint32_t src_w, uint32_t src_h) +{ + struct drm_device *dev = drm_plane->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_plane *intel_plane = to_intel_plane(drm_plane); + const int pipe = intel_plane->pipe; + const int plane = intel_plane->plane + 1; + u32 plane_ctl, stride; + int pixel_size = drm_format_plane_cpp(fb->pixel_format, 0); + + plane_ctl = I915_READ(PLANE_CTL(pipe, plane)); + + /* Mask out pixel format bits in case we change it */ + plane_ctl &= ~PLANE_CTL_FORMAT_MASK; + plane_ctl &= ~PLANE_CTL_ORDER_RGBX; + plane_ctl &= ~PLANE_CTL_YUV422_ORDER_MASK; + plane_ctl &= ~PLANE_CTL_TILED_MASK; + plane_ctl &= ~PLANE_CTL_ALPHA_MASK; + + /* Trickle feed has to be enabled */ + plane_ctl &= ~PLANE_CTL_TRICKLE_FEED_DISABLE; + + switch (fb->pixel_format) { + case DRM_FORMAT_RGB565: + plane_ctl |= PLANE_CTL_FORMAT_RGB_565; + break; + case DRM_FORMAT_XBGR8888: + plane_ctl |= PLANE_CTL_FORMAT_XRGB_8888 | PLANE_CTL_ORDER_RGBX; + break; + case DRM_FORMAT_XRGB8888: + plane_ctl |= PLANE_CTL_FORMAT_XRGB_8888; + break; + /* + * XXX: For ARBG/ABGR formats we default to expecting scanout buffers + * to be already pre-multiplied. We need to add a knob (or a different + * DRM_FORMAT) for user-space to configure that. + */ + case DRM_FORMAT_ABGR8888: + plane_ctl |= PLANE_CTL_FORMAT_XRGB_8888 | + PLANE_CTL_ORDER_RGBX | + PLANE_CTL_ALPHA_SW_PREMULTIPLY; + break; + case DRM_FORMAT_ARGB8888: + plane_ctl |= PLANE_CTL_FORMAT_XRGB_8888 | + PLANE_CTL_ALPHA_SW_PREMULTIPLY; + break; + case DRM_FORMAT_YUYV: + plane_ctl |= PLANE_CTL_FORMAT_YUV422 | PLANE_CTL_YUV422_YUYV; + break; + case DRM_FORMAT_YVYU: + plane_ctl |= PLANE_CTL_FORMAT_YUV422 | PLANE_CTL_YUV422_YVYU; + break; + case DRM_FORMAT_UYVY: + plane_ctl |= PLANE_CTL_FORMAT_YUV422 | PLANE_CTL_YUV422_UYVY; + break; + case DRM_FORMAT_VYUY: + plane_ctl |= PLANE_CTL_FORMAT_YUV422 | PLANE_CTL_YUV422_VYUY; + break; + default: + BUG(); + } + + switch (obj->tiling_mode) { + case I915_TILING_NONE: + stride = fb->pitches[0] >> 6; + break; + case I915_TILING_X: + plane_ctl |= PLANE_CTL_TILED_X; + stride = fb->pitches[0] >> 9; + break; + default: + BUG(); + } + + plane_ctl |= PLANE_CTL_ENABLE; + plane_ctl |= PLANE_CTL_PIPE_CSC_ENABLE; + + intel_update_sprite_watermarks(drm_plane, crtc, src_w, src_h, + pixel_size, true, + src_w != crtc_w || src_h != crtc_h); + + /* Sizes are 0 based */ + src_w--; + src_h--; + crtc_w--; + crtc_h--; + + I915_WRITE(PLANE_OFFSET(pipe, plane), (y << 16) | x); + I915_WRITE(PLANE_STRIDE(pipe, plane), stride); + I915_WRITE(PLANE_POS(pipe, plane), (crtc_y << 16) | crtc_x); + I915_WRITE(PLANE_SIZE(pipe, plane), (crtc_h << 16) | crtc_w); + I915_WRITE(PLANE_CTL(pipe, plane), plane_ctl); + I915_WRITE(PLANE_SURF(pipe, plane), i915_gem_obj_ggtt_offset(obj)); + POSTING_READ(PLANE_SURF(pipe, plane)); +} + +static void +skl_disable_plane(struct drm_plane *drm_plane, struct drm_crtc *crtc) +{ + struct drm_device *dev = drm_plane->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_plane *intel_plane = to_intel_plane(drm_plane); + const int pipe = intel_plane->pipe; + const int plane = intel_plane->plane + 1; + + I915_WRITE(PLANE_CTL(pipe, plane), + I915_READ(PLANE_CTL(pipe, plane)) & ~PLANE_CTL_ENABLE); + + /* Activate double buffered register update */ + I915_WRITE(PLANE_CTL(pipe, plane), 0); + POSTING_READ(PLANE_CTL(pipe, plane)); + + intel_update_sprite_watermarks(drm_plane, crtc, 0, 0, 0, false, false); +} + +static int +skl_update_colorkey(struct drm_plane *drm_plane, + struct drm_intel_sprite_colorkey *key) +{ + struct drm_device *dev = drm_plane->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_plane *intel_plane = to_intel_plane(drm_plane); + const int pipe = intel_plane->pipe; + const int plane = intel_plane->plane; + u32 plane_ctl; + + I915_WRITE(PLANE_KEYVAL(pipe, plane), key->min_value); + I915_WRITE(PLANE_KEYMAX(pipe, plane), key->max_value); + I915_WRITE(PLANE_KEYMSK(pipe, plane), key->channel_mask); + + plane_ctl = I915_READ(PLANE_CTL(pipe, plane)); + plane_ctl &= ~PLANE_CTL_KEY_ENABLE_MASK; + if (key->flags & I915_SET_COLORKEY_DESTINATION) + plane_ctl |= PLANE_CTL_KEY_ENABLE_DESTINATION; + else if (key->flags & I915_SET_COLORKEY_SOURCE) + plane_ctl |= PLANE_CTL_KEY_ENABLE_SOURCE; + I915_WRITE(PLANE_CTL(pipe, plane), plane_ctl); + + POSTING_READ(PLANE_CTL(pipe, plane)); + + return 0; +} + +static void +skl_get_colorkey(struct drm_plane *drm_plane, + struct drm_intel_sprite_colorkey *key) +{ + struct drm_device *dev = drm_plane->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_plane *intel_plane = to_intel_plane(drm_plane); + const int pipe = intel_plane->pipe; + const int plane = intel_plane->plane; + u32 plane_ctl; + + key->min_value = I915_READ(PLANE_KEYVAL(pipe, plane)); + key->max_value = I915_READ(PLANE_KEYMAX(pipe, plane)); + key->channel_mask = I915_READ(PLANE_KEYMSK(pipe, plane)); + + plane_ctl = I915_READ(PLANE_CTL(pipe, plane)); + + switch (plane_ctl & PLANE_CTL_KEY_ENABLE_MASK) { + case PLANE_CTL_KEY_ENABLE_DESTINATION: + key->flags = I915_SET_COLORKEY_DESTINATION; + break; + case PLANE_CTL_KEY_ENABLE_SOURCE: + key->flags = I915_SET_COLORKEY_SOURCE; + break; + default: + key->flags = I915_SET_COLORKEY_NONE; + } +} + +static void vlv_update_plane(struct drm_plane *dplane, struct drm_crtc *crtc, struct drm_framebuffer *fb, struct drm_i915_gem_object *obj, int crtc_x, int crtc_y, @@ -1305,6 +1483,18 @@ static uint32_t vlv_plane_formats[] = { DRM_FORMAT_VYUY, }; +static uint32_t skl_plane_formats[] = { + DRM_FORMAT_RGB565, + DRM_FORMAT_ABGR8888, + DRM_FORMAT_ARGB8888, + DRM_FORMAT_XBGR8888, + DRM_FORMAT_XRGB8888, + DRM_FORMAT_YUYV, + DRM_FORMAT_YVYU, + DRM_FORMAT_UYVY, + DRM_FORMAT_VYUY, +}; + int intel_plane_init(struct drm_device *dev, enum pipe pipe, int plane) { @@ -1368,7 +1558,21 @@ intel_plane_init(struct drm_device *dev, enum pipe pipe, int plane) num_plane_formats = ARRAY_SIZE(snb_plane_formats); } break; - + case 9: + /* + * FIXME: Skylake planes can be scaled (with some restrictions), + * but this is for another time. + */ + intel_plane->can_scale = false; + intel_plane->max_downscale = 1; + intel_plane->update_plane = skl_update_plane; + intel_plane->disable_plane = skl_disable_plane; + intel_plane->update_colorkey = skl_update_colorkey; + intel_plane->get_colorkey = skl_get_colorkey; + + plane_formats = skl_plane_formats; + num_plane_formats = ARRAY_SIZE(skl_plane_formats); + break; default: kfree(intel_plane); return -ENODEV; -- cgit v0.10.2 From 7feb8b88e7513d6b093539c2b7291cfb87cb917a Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 12 Mar 2014 21:05:38 +0000 Subject: drm/i915/skl: Adjust assert_sprites_disabled() Let's put to good use the new PLANE_CTL macros. Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 2087e8b..1bda97c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -1335,7 +1335,14 @@ static void assert_sprites_disabled(struct drm_i915_private *dev_priv, int reg, sprite; u32 val; - if (IS_VALLEYVIEW(dev)) { + if (INTEL_INFO(dev)->gen >= 9) { + for_each_sprite(pipe, sprite) { + val = I915_READ(PLANE_CTL(pipe, sprite)); + WARN(val & PLANE_CTL_ENABLE, + "plane %d assertion failure, should be off on pipe %c but is still active\n", + sprite, pipe_name(pipe)); + } + } else if (IS_VALLEYVIEW(dev)) { for_each_sprite(pipe, sprite) { reg = SPCNTR(pipe, sprite); val = I915_READ(reg); -- cgit v0.10.2 From 84139d1e39265c20a79daace16ef4b6a9a644d9a Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 28 Mar 2014 00:18:32 +0530 Subject: drm/i915/skl: Introduce a I915_MAX_PLANES macro This can be useful to declare structures around pipes and planes and don't have to go back auditing the code if the next platorm bump that number. Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index a8d4e23..c3dbaae 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -76,6 +76,14 @@ enum transcoder { }; #define transcoder_name(t) ((t) + 'A') +/* + * This is the maximum (across all platforms) number of planes (primary + + * sprites) that can be active at the same time on one pipe. + * + * This value doesn't count the cursor plane. + */ +#define I915_MAX_PLANES 3 + enum plane { PLANE_A = 0, PLANE_B, -- cgit v0.10.2 From 6af31a653d6c8e1708c27b7f32cafd5435b3bb90 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 28 Mar 2014 00:18:33 +0530 Subject: drm/i915/skl: Introduce intel_num_planes() It can be handy to get the number of planes for this pipe, ie including the primary plane to loop over them. Introduce a little function to do so. Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 07ce046..a09e4ad 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -731,6 +731,14 @@ hdmi_to_dig_port(struct intel_hdmi *intel_hdmi) return container_of(intel_hdmi, struct intel_digital_port, hdmi); } +/* + * Returns the number of planes for this pipe, ie the number of sprites + 1 + * (primary plane). This doesn't count the cursor plane then. + */ +static inline unsigned int intel_num_planes(struct intel_crtc *crtc) +{ + return INTEL_INFO(crtc->base.dev)->num_sprites[crtc->pipe] + 1; +} /* i915_irq.c */ bool intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, -- cgit v0.10.2 From c83155a6044341d67b85b441ba719f86058f6e2b Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 28 Mar 2014 00:18:35 +0530 Subject: drm/i915/skl: Move gen9 pm initialization into its own branch Gen9 is different enough (for instance, fetching the memory latency values is different from ILK+) to not take the HAS_PCH_SPLIT() branch, so let's prefer a clean separation. v2: Rebase on top of the broadwell_init_clock_gating() name change Reviewed-by: Rodrigo Vivi Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 4f5dcf5..81eb8bc 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7397,7 +7397,9 @@ void intel_init_pm(struct drm_device *dev) i915_ironlake_get_mem_freq(dev); /* For FIFO watermark updates */ - if (HAS_PCH_SPLIT(dev)) { + if (IS_GEN9(dev)) { + dev_priv->display.init_clock_gating = gen9_init_clock_gating; + } else if (HAS_PCH_SPLIT(dev)) { ilk_setup_wm_latency(dev); if ((IS_GEN5(dev) && dev_priv->wm.pri_latency[1] && @@ -7421,8 +7423,6 @@ void intel_init_pm(struct drm_device *dev) dev_priv->display.init_clock_gating = haswell_init_clock_gating; else if (INTEL_INFO(dev)->gen == 8) dev_priv->display.init_clock_gating = broadwell_init_clock_gating; - else if (INTEL_INFO(dev)->gen == 9) - dev_priv->display.init_clock_gating = gen9_init_clock_gating; } else if (IS_CHERRYVIEW(dev)) { dev_priv->display.update_wm = cherryview_update_wm; dev_priv->display.update_sprite_wm = valleyview_update_sprite_wm; -- cgit v0.10.2 From 7ca5a41f4da201371e131fc0641033652f76bf30 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 16 Sep 2014 19:19:07 -0400 Subject: drm/i915: Avoid re-configure panel on every PSR re-enable. The panel has to be reconfigured only when it really loose the power. The traditional enable/disable sequence already take care of this so we can minimize the time spend on every re-enable. Signed-off-by: Rodrigo Vivi Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 43b381a..c5ec821 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2121,10 +2121,7 @@ static void intel_edp_psr_do_enable(struct intel_dp *intel_dp) WARN_ON(dev_priv->psr.active); lockdep_assert_held(&dev_priv->psr.lock); - /* Enable PSR on the panel */ - intel_edp_psr_enable_sink(intel_dp); - - /* Enable PSR on the host */ + /* Enable/Re-enable PSR on the host */ intel_edp_psr_enable_source(intel_dp); dev_priv->psr.active = true; @@ -2162,6 +2159,9 @@ void intel_edp_psr_enable(struct intel_dp *intel_dp) I915_WRITE(EDP_PSR_DEBUG_CTL(dev), EDP_PSR_DEBUG_MASK_MEMUP | EDP_PSR_DEBUG_MASK_HPD | EDP_PSR_DEBUG_MASK_LPSP); + /* Enable PSR on the panel */ + intel_edp_psr_enable_sink(intel_dp); + dev_priv->psr.enabled = intel_dp; unlock: mutex_unlock(&dev_priv->psr.lock); -- cgit v0.10.2 From 1d73c2a8f218be3e8b6aa884740fc67110660b54 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 24 Sep 2014 19:50:59 -0400 Subject: drm/i915: Minimize the huge amount of unecessary fbc sw cache clean. The sw cache clean on BDW is a tempoorary workaround because we cannot set cache clean on blt ring with risk of hungs. So we are doing the cache clean on sw. However we are doing much more than needed. Not only when using blt ring. So, with this extra w/a we minimize the ammount of cache cleans and call it only on same cases that it was being called on gen7. The traditional FBC Cache clean happens over LRI on BLT ring when there is a frontbuffer touch happening. frontbuffer tracking set fbc_dirty variable to let BLT flush that it must clean FBC cache. fbc.need_sw_cache_clean works in the opposite information direction of ring->fbc_dirty telling software on frontbuffer tracking to perform the cache clean on sw side. v2: Clean it a little bit and fully check for Broadwell instead of gen8. v3: Rebase after frontbuffer organization. v4: Wiggle confused me. So fixing v3! Cc: Daniel Vetter Cc: Paulo Zanoni Reviewed-by: Paulo Zanoni Signed-off-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 999bd57..cccb776 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -666,6 +666,14 @@ struct i915_fbc { * possible. */ bool enabled; + /* On gen8 some rings cannont perform fbc clean operation so for now + * we are doing this on SW with mmio. + * This variable works in the opposite information direction + * of ring->fbc_dirty telling software on frontbuffer tracking + * to perform the cache clean on sw side. + */ + bool need_sw_cache_clean; + struct intel_fbc_work { struct delayed_work work; struct drm_crtc *crtc; @@ -2825,7 +2833,7 @@ extern void intel_modeset_setup_hw_state(struct drm_device *dev, extern void i915_redisable_vga(struct drm_device *dev); extern void i915_redisable_vga_power_on(struct drm_device *dev); extern bool intel_fbc_enabled(struct drm_device *dev); -extern void gen8_fbc_sw_flush(struct drm_device *dev, u32 value); +extern void bdw_fbc_sw_flush(struct drm_device *dev, u32 value); extern void intel_disable_fbc(struct drm_device *dev); extern bool ironlake_set_drps(struct drm_device *dev, u8 val); extern void intel_init_pch_refclk(struct drm_device *dev); diff --git a/drivers/gpu/drm/i915/intel_frontbuffer.c b/drivers/gpu/drm/i915/intel_frontbuffer.c index f74744c..7eb74a6 100644 --- a/drivers/gpu/drm/i915/intel_frontbuffer.c +++ b/drivers/gpu/drm/i915/intel_frontbuffer.c @@ -189,8 +189,10 @@ void intel_frontbuffer_flush(struct drm_device *dev, * needs to be reworked into a proper frontbuffer tracking scheme like * psr employs. */ - if (IS_BROADWELL(dev)) - gen8_fbc_sw_flush(dev, FBC_REND_CACHE_CLEAN); + if (dev_priv->fbc.need_sw_cache_clean) { + dev_priv->fbc.need_sw_cache_clean = false; + bdw_fbc_sw_flush(dev, FBC_REND_CACHE_CLEAN); + } } /** diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 6b41620..011892d 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -356,7 +356,7 @@ bool intel_fbc_enabled(struct drm_device *dev) return dev_priv->fbc.enabled; } -void gen8_fbc_sw_flush(struct drm_device *dev, u32 value) +void bdw_fbc_sw_flush(struct drm_device *dev, u32 value) { struct drm_i915_private *dev_priv = dev->dev_private; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 922d6bc..620a89d 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -2237,6 +2237,7 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, u32 invalidate, u32 flush) { struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; uint32_t cmd; int ret; @@ -2267,8 +2268,12 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, } intel_ring_advance(ring); - if (IS_GEN7(dev) && !invalidate && flush) - return gen7_ring_fbc_flush(ring, FBC_REND_CACHE_CLEAN); + if (!invalidate && flush) { + if (IS_GEN7(dev)) + return gen7_ring_fbc_flush(ring, FBC_REND_CACHE_CLEAN); + else if (IS_BROADWELL(dev)) + dev_priv->fbc.need_sw_cache_clean = true; + } return 0; } -- cgit v0.10.2 From 8d7f4fe9f561f4e3253478332192d88f6f135c41 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 24 Sep 2014 18:16:58 -0400 Subject: drm/i915: Make sure PSR is ready for been re-enabled. Let's make sure PSR is propperly disabled before to re-enabled it. According to Spec, after disabled PSR CTL, the Idle state might occur up to 24ms, that is one full frame time (1/refresh rate), plus SRD exit training time (max of 6ms), plus SRD aux channel handshake (max of 1.5ms). So if something went wrong PSR will be disabled until next full enable/disable setup. v2: The 24ms above takes in account 16ms for refresh rate on 60Hz mode. However on low frequency modes this can take longer. So let's use 50ms for safeness. v3: Move wait out of psr.lock critical area. Cc: Daniel Vetter Cc: Paulo Zanoni Signed-off-by: Rodrigo Vivi Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index c5ec821..6f42d95 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2204,6 +2204,17 @@ static void intel_edp_psr_work(struct work_struct *work) container_of(work, typeof(*dev_priv), psr.work.work); struct intel_dp *intel_dp = dev_priv->psr.enabled; + /* We have to make sure PSR is ready for re-enable + * otherwise it keeps disabled until next full enable/disable cycle. + * PSR might take some time to get fully disabled + * and be ready for re-enable. + */ + if (wait_for((I915_READ(EDP_PSR_STATUS_CTL(dev_priv->dev)) & + EDP_PSR_STATUS_STATE_MASK) == 0, 50)) { + DRM_ERROR("Timed out waiting for PSR Idle for re-enable\n"); + return; + } + mutex_lock(&dev_priv->psr.lock); intel_dp = dev_priv->psr.enabled; -- cgit v0.10.2 From 17b523ba5c3835e6d595c9d7d6a45177b379efcd Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 24 Sep 2014 20:32:43 -0400 Subject: drm/i915: Broadwell DDI Buffer translation changed to give better margin. Reference: Predator r73977 / HSD 4394389 Cc: Arthur Runyan Cc: Paulo Zanoni Signed-off-by: Rodrigo Vivi Reviewed-by: Arthur Runyan Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index b63d4fa..e3cd159 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -95,8 +95,8 @@ static const struct ddi_buf_trans bdw_ddi_translations_dp[] = { { 0x00BEFFFF, 0x00140006 }, { 0x80B2CFFF, 0x001B0002 }, { 0x00FFFFFF, 0x000E000A }, - { 0x00D75FFF, 0x00180004 }, - { 0x80CB2FFF, 0x001B0002 }, + { 0x00DB6FFF, 0x00160005 }, + { 0x00C71FFF, 0x001A0002 }, { 0x00F7DFFF, 0x00180004 }, { 0x80D75FFF, 0x001B0002 }, }; -- cgit v0.10.2 From 6805b2a7434daaede07969ed5877300b2a0783e7 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Thu, 25 Sep 2014 12:28:32 -0400 Subject: drm/i915: Broadwell DDI Buffer translation - more tuning BDW display - DP buffer translation values changed to give better margin. Further change to entry 6; set dword 0 bit 31=1. Both changes were approved already but this one didn't landed BSpec yet this is why it is in a separated patch. Making reviewer's life easier. Also alowing separated tests and any future bisect that might be needed. Reference: Predator r74080 / HSD 4394389 v2: Arthur noticed I was changing the wrong bit. Cc: Arthur Runyan Cc: Paulo Zanoni Signed-off-by: Rodrigo Vivi Reviewed-by: Arthur Runyan Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index e3cd159..2958278 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -96,7 +96,7 @@ static const struct ddi_buf_trans bdw_ddi_translations_dp[] = { { 0x80B2CFFF, 0x001B0002 }, { 0x00FFFFFF, 0x000E000A }, { 0x00DB6FFF, 0x00160005 }, - { 0x00C71FFF, 0x001A0002 }, + { 0x80C71FFF, 0x001A0002 }, { 0x00F7DFFF, 0x00180004 }, { 0x80D75FFF, 0x001B0002 }, }; -- cgit v0.10.2 From ad9dc91b6e21266bfc6f466db4b95e10211f31ee Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 16 Sep 2014 19:18:12 -0400 Subject: drm/i915: Fix Sink CRC In some cases like when PSR just got enabled the panel need more vblank times to calculate CRC. I figured that out with the new PSR test cases facing some cases that I had a green screen but a blank CRC. Even with 2 vblank waits on kernel + 2 vblank waits on test case. So let's give up to 6 vblank wait time. However we now check for TEST_CRC_COUNT that shows when panel finished to calculate CRC and has it ready. v2: Jani pointed out attempts decrements was wrong and should never reach the error condition. And Daniel pointed out that EIO is more appropriated than EGAIN. Also I realized that I have to read test_crc_count after setting test_sink v3: Rebase and adding error message Cc: Todd Previte Cc: Daniel Vetter Cc: Jani Nikula Signed-off-by: Rodrigo Vivi Reviewed-by: Todd Previte Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 6f42d95..3caac0f 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3807,21 +3807,32 @@ int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc) struct drm_device *dev = intel_dig_port->base.base.dev; struct intel_crtc *intel_crtc = to_intel_crtc(intel_dig_port->base.base.crtc); - u8 buf[1]; + u8 buf; + int test_crc_count; + int attempts = 6; - if (drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK_MISC, buf) < 0) + if (drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK_MISC, &buf) < 0) return -EIO; - if (!(buf[0] & DP_TEST_CRC_SUPPORTED)) + if (!(buf & DP_TEST_CRC_SUPPORTED)) return -ENOTTY; if (drm_dp_dpcd_writeb(&intel_dp->aux, DP_TEST_SINK, DP_TEST_SINK_START) < 0) return -EIO; - /* Wait 2 vblanks to be sure we will have the correct CRC value */ - intel_wait_for_vblank(dev, intel_crtc->pipe); - intel_wait_for_vblank(dev, intel_crtc->pipe); + drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK_MISC, &buf); + test_crc_count = buf & DP_TEST_COUNT_MASK; + + do { + drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK_MISC, &buf); + intel_wait_for_vblank(dev, intel_crtc->pipe); + } while (--attempts && (buf & DP_TEST_COUNT_MASK) == test_crc_count); + + if (attempts == 0) { + DRM_ERROR("Panel is unable to calculate CRC after 6 vblanks\n"); + return -EIO; + } if (drm_dp_dpcd_read(&intel_dp->aux, DP_TEST_CRC_R_CR, crc, 6) < 0) return -EIO; diff --git a/include/drm/drm_dp_helper.h b/include/drm/drm_dp_helper.h index 9305c71..8edeed0 100644 --- a/include/drm/drm_dp_helper.h +++ b/include/drm/drm_dp_helper.h @@ -303,7 +303,8 @@ #define DP_TEST_CRC_B_CB 0x244 #define DP_TEST_SINK_MISC 0x246 -#define DP_TEST_CRC_SUPPORTED (1 << 5) +# define DP_TEST_CRC_SUPPORTED (1 << 5) +# define DP_TEST_COUNT_MASK 0x7 #define DP_TEST_RESPONSE 0x260 # define DP_TEST_ACK (1 << 0) @@ -313,7 +314,7 @@ #define DP_TEST_EDID_CHECKSUM 0x261 #define DP_TEST_SINK 0x270 -#define DP_TEST_SINK_START (1 << 0) +# define DP_TEST_SINK_START (1 << 0) #define DP_PAYLOAD_TABLE_UPDATE_STATUS 0x2c0 /* 1.2 MST */ # define DP_PAYLOAD_TABLE_UPDATED (1 << 0) -- cgit v0.10.2 From a0fcbd95354e5ac2400be74077cc668c91429ea7 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 19 Sep 2014 20:16:26 -0400 Subject: drm/i915: Add IS_BDW_GT3 macro. It will be usefull to specify w/a that affects only BDW GT3. Signed-off-by: Rodrigo Vivi Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index cccb776..8f05258 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2111,6 +2111,8 @@ struct drm_i915_cmd_table { ((INTEL_DEVID(dev) & 0xf) == 0x2 || \ (INTEL_DEVID(dev) & 0xf) == 0x6 || \ (INTEL_DEVID(dev) & 0xf) == 0xe)) +#define IS_BDW_GT3(dev) (IS_BROADWELL(dev) && \ + (INTEL_DEVID(dev) & 0x00F0) == 0x0020) #define IS_HSW_ULT(dev) (IS_HASWELL(dev) && \ (INTEL_DEVID(dev) & 0xFF00) == 0x0A00) #define IS_ULT(dev) (IS_HSW_ULT(dev) || IS_BDW_ULT(dev)) -- cgit v0.10.2 From da09654d777c361006f6ea3452f8de4a374d5783 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 19 Sep 2014 20:16:27 -0400 Subject: drm/i915/bdw: WaDisableFenceDestinationToSLM This WA affect BDW GT3 pre-production steppings. Signed-off-by: Rodrigo Vivi Reviewed-by: Mika Kuoppala [danvet: Don't mention steppings ...] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index ad8179b..124ea60 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4836,6 +4836,7 @@ enum punit_power_well { /* GEN8 chicken */ #define HDC_CHICKEN0 0x7300 #define HDC_FORCE_NON_COHERENT (1<<4) +#define HDC_FENCE_DEST_SLM_DISABLE (1<<14) /* WaCatErrorRejectionIssue */ #define GEN7_SQ_CHICKEN_MBCUNIT_CONFIG 0x9030 diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 620a89d..c21aaad 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -740,8 +740,12 @@ static int bdw_init_workarounds(struct intel_engine_cs *ring) * workaround for for a possible hang in the unlikely event a TLB * invalidation occurs during a PSD flush. */ + /* WaDisableFenceDestinationToSLM:bdw (GT3 pre-production) */ intel_ring_emit_wa(ring, HDC_CHICKEN0, - _MASKED_BIT_ENABLE(HDC_FORCE_NON_COHERENT)); + _MASKED_BIT_ENABLE(HDC_FORCE_NON_COHERENT | + (IS_BDW_GT3(dev) ? + HDC_FENCE_DEST_SLM_DISABLE : 0) + )); /* Wa4x4STCOptimizationDisable:bdw */ intel_ring_emit_wa(ring, CACHE_MODE_1, -- cgit v0.10.2 From ce31d9f4fc05964f6c0dd3a8661dc1a1d843a1e2 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Mon, 29 Sep 2014 18:29:52 -0400 Subject: drm/i915: preserve other DP_TEST_SINK bits. Sink crc was implemented based on dp 1.1 spec that had all TEST_SINK bits reserved reading all 0s. But when reviewing my latest changes on sink crc Todd warned me that on new specs we have other valid bits on this reg that we might want to preserve. Cc: Todd Previte Signed-off-by: Rodrigo Vivi Reviewed-by: Todd Previte Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 3caac0f..342d624 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3817,8 +3817,9 @@ int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc) if (!(buf & DP_TEST_CRC_SUPPORTED)) return -ENOTTY; + drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK, &buf); if (drm_dp_dpcd_writeb(&intel_dp->aux, DP_TEST_SINK, - DP_TEST_SINK_START) < 0) + buf | DP_TEST_SINK_START) < 0) return -EIO; drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK_MISC, &buf); @@ -3837,7 +3838,10 @@ int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc) if (drm_dp_dpcd_read(&intel_dp->aux, DP_TEST_CRC_R_CR, crc, 6) < 0) return -EIO; - drm_dp_dpcd_writeb(&intel_dp->aux, DP_TEST_SINK, 0); + drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK, &buf); + drm_dp_dpcd_writeb(&intel_dp->aux, DP_TEST_SINK, + buf & ~DP_TEST_SINK_START); + return 0; } -- cgit v0.10.2 From 11c9b6c628c646894e6ef53f92cfd33a814ee553 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 22:10:52 +0200 Subject: drm/i915: Tighting frontbuffer tracking around flips So I think I've spotted a small gap in the frontbuffer tracking while discussing the logic with Paulo on irc: 1. Userspace schedules gpu rendering to the current frontbuffer. This gets tracked in dev_priv->fb_tracking.busy_bits. 2. We pageflip a fully rendered buffer before the frontbuffer rendering completes. 3. The request retiring will never clear busy_bits (since at retire time the old frontbuffer won't have obj->frontbuffer_bits set), so these bits now are stuck until someone again does a bit of frontbuffer tracking. If we clear stale busy_bits in flip_prepare this gap is closed. Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_frontbuffer.c b/drivers/gpu/drm/i915/intel_frontbuffer.c index 7eb74a6..c5a312d 100644 --- a/drivers/gpu/drm/i915/intel_frontbuffer.c +++ b/drivers/gpu/drm/i915/intel_frontbuffer.c @@ -248,8 +248,9 @@ void intel_frontbuffer_flip_prepare(struct drm_device *dev, struct drm_i915_private *dev_priv = dev->dev_private; mutex_lock(&dev_priv->fb_tracking.lock); - dev_priv->fb_tracking.flip_bits - |= frontbuffer_bits; + dev_priv->fb_tracking.flip_bits |= frontbuffer_bits; + /* Remove stale busy bits due to the old buffer. */ + dev_priv->fb_tracking.busy_bits &= ~frontbuffer_bits; mutex_unlock(&dev_priv->fb_tracking.lock); } -- cgit v0.10.2 From 5c323b2a78fade2df08a51497fc943a6fd1b868b Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 22:10:53 +0200 Subject: drm/i915: spelling fixes for frontbuffer tracking kerneldoc Oh well. v2: Fix one more spelling fail Paulo spotted. Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 072e69f..f14f456 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -819,7 +819,7 @@ void intel_frontbuffer_flip_complete(struct drm_device *dev, void intel_frontbuffer_flush(struct drm_device *dev, unsigned frontbuffer_bits); /** - * intel_frontbuffer_flip - prepare frontbuffer flip + * intel_frontbuffer_flip - synchronous frontbuffer flip * @dev: DRM device * @frontbuffer_bits: frontbuffer plane tracking bits * diff --git a/drivers/gpu/drm/i915/intel_frontbuffer.c b/drivers/gpu/drm/i915/intel_frontbuffer.c index c5a312d..58cf2e6 100644 --- a/drivers/gpu/drm/i915/intel_frontbuffer.c +++ b/drivers/gpu/drm/i915/intel_frontbuffer.c @@ -28,7 +28,7 @@ * DOC: frontbuffer tracking * * Many features require us to track changes to the currently active - * frontbuffer, especially rendering targetted at the frontbuffer. + * frontbuffer, especially rendering targeted at the frontbuffer. * * To be able to do so GEM tracks frontbuffers using a bitmask for all possible * frontbuffer slots through i915_gem_track_fb(). The function in this file are @@ -55,7 +55,7 @@ * cancelled as soon as busyness is detected. * * Note that there's also an older frontbuffer activity tracking scheme which - * just trackings general activity. This is done by the various mark_busy and + * just tracks general activity. This is done by the various mark_busy and * mark_idle functions. For display power management features using these * functions is deprecated and should be avoided. */ @@ -166,7 +166,7 @@ void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj, * * This function gets called every time rendering on the given planes has * completed and frontbuffer caching can be started again. Flushes will get - * delayed if they're blocked by some oustanding asynchronous rendering. + * delayed if they're blocked by some outstanding asynchronous rendering. * * Can be called without any locks held. */ @@ -231,7 +231,7 @@ void intel_fb_obj_flush(struct drm_i915_gem_object *obj, } /** - * intel_frontbuffer_flip_prepare - prepare asnychronous frontbuffer flip + * intel_frontbuffer_flip_prepare - prepare asynchronous frontbuffer flip * @dev: DRM device * @frontbuffer_bits: frontbuffer plane tracking bits * @@ -255,12 +255,12 @@ void intel_frontbuffer_flip_prepare(struct drm_device *dev, } /** - * intel_frontbuffer_flip_complete - complete asynchronous frontbuffer flush + * intel_frontbuffer_flip_complete - complete asynchronous frontbuffer flip * @dev: DRM device * @frontbuffer_bits: frontbuffer plane tracking bits * * This function gets called after the flip has been latched and will complete - * on the next vblank. It will execute the fush if it hasn't been cancalled yet. + * on the next vblank. It will execute the flush if it hasn't been cancelled yet. * * Can be called without any locks held. */ @@ -277,4 +277,3 @@ void intel_frontbuffer_flip_complete(struct drm_device *dev, intel_frontbuffer_flush(dev, frontbuffer_bits); } - -- cgit v0.10.2 From 970104fac6ca0cfdfbaa1a23c70c06a71208e2ac Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:37 +0200 Subject: drm/i915: Remove intel_modeset_suspend_hw Another layer of indirection for just an lpt-only w/a is a bit excessive. Reduce it. This was added in commit 7d708ee40a6b9ca1112a322e554c887df105b025 Author: Imre Deak Date: Wed Apr 17 14:04:50 2013 +0300 drm/i915: HSW: allow PCH clock gating for suspend Cc: Imre Deak Cc: Paulo Zanoni Signed-off-by: Daniel Vetter Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 6948877..ea93ff1 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -606,7 +606,7 @@ static int i915_drm_freeze(struct drm_device *dev) intel_suspend_gt_powersave(dev); - intel_modeset_suspend_hw(dev); + intel_suspend_hw(dev); } i915_gem_suspend_gtt_mappings(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 4cd2aa3..622c038 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2840,7 +2840,6 @@ static inline void intel_unregister_dsm_handler(void) { return; } /* modesetting */ extern void intel_modeset_init_hw(struct drm_device *dev); -extern void intel_modeset_suspend_hw(struct drm_device *dev); extern void intel_modeset_init(struct drm_device *dev); extern void intel_modeset_gem_init(struct drm_device *dev); extern void intel_modeset_cleanup(struct drm_device *dev); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index b8488a8..a960578 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -12830,11 +12830,6 @@ void intel_modeset_init_hw(struct drm_device *dev) intel_enable_gt_powersave(dev); } -void intel_modeset_suspend_hw(struct drm_device *dev) -{ - intel_suspend_hw(dev); -} - void intel_modeset_init(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; -- cgit v0.10.2 From 9c065a7d5b679e7fabe3cace4faadb283f2b0c1f Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:38 +0200 Subject: drm/i915: Extract intel_runtime_pm.c Geez is the audio hack ugly. Signed-off-by: Daniel Vetter Reviewed-by: Imre Deak [danvet: Rebased on top of the skl patches.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 2d8317d..3a6bce0 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -11,7 +11,9 @@ i915-y := i915_drv.o \ i915_params.o \ i915_suspend.o \ i915_sysfs.o \ - intel_pm.o + intel_pm.o \ + intel_runtime_pm.o + i915-$(CONFIG_COMPAT) += i915_ioc32.o i915-$(CONFIG_DEBUG_FS) += i915_debugfs.o diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index f14f456..9d2ee70 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1081,6 +1081,27 @@ extern struct drm_display_mode *intel_find_panel_downclock( struct drm_display_mode *fixed_mode, struct drm_connector *connector); +/* intel_runtime_pm.c */ +int intel_power_domains_init(struct drm_i915_private *); +void intel_power_domains_remove(struct drm_i915_private *); +void intel_power_domains_init_hw(struct drm_i915_private *dev_priv); +void intel_init_runtime_pm(struct drm_i915_private *dev_priv); +void intel_fini_runtime_pm(struct drm_i915_private *dev_priv); + +bool intel_display_power_enabled(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain); +bool intel_display_power_enabled_unlocked(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain); +void intel_display_power_get(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain); +void intel_display_power_put(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain); +void intel_aux_display_runtime_get(struct drm_i915_private *dev_priv); +void intel_aux_display_runtime_put(struct drm_i915_private *dev_priv); +void intel_runtime_pm_get(struct drm_i915_private *dev_priv); +void intel_runtime_pm_get_noresume(struct drm_i915_private *dev_priv); +void intel_runtime_pm_put(struct drm_i915_private *dev_priv); + /* intel_pm.c */ void intel_init_clock_gating(struct drm_device *dev); void intel_suspend_hw(struct drm_device *dev); @@ -1098,17 +1119,6 @@ bool intel_fbc_enabled(struct drm_device *dev); void intel_update_fbc(struct drm_device *dev); void intel_gpu_ips_init(struct drm_i915_private *dev_priv); void intel_gpu_ips_teardown(void); -int intel_power_domains_init(struct drm_i915_private *); -void intel_power_domains_remove(struct drm_i915_private *); -bool intel_display_power_enabled(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain); -bool intel_display_power_enabled_unlocked(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain); -void intel_display_power_get(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain); -void intel_display_power_put(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain); -void intel_power_domains_init_hw(struct drm_i915_private *dev_priv); void intel_init_gt_powersave(struct drm_device *dev); void intel_cleanup_gt_powersave(struct drm_device *dev); void intel_enable_gt_powersave(struct drm_device *dev); @@ -1119,13 +1129,6 @@ void ironlake_teardown_rc6(struct drm_device *dev); void gen6_update_ring_freq(struct drm_device *dev); void gen6_rps_idle(struct drm_i915_private *dev_priv); void gen6_rps_boost(struct drm_i915_private *dev_priv); -void intel_aux_display_runtime_get(struct drm_i915_private *dev_priv); -void intel_aux_display_runtime_put(struct drm_i915_private *dev_priv); -void intel_runtime_pm_get(struct drm_i915_private *dev_priv); -void intel_runtime_pm_get_noresume(struct drm_i915_private *dev_priv); -void intel_runtime_pm_put(struct drm_i915_private *dev_priv); -void intel_init_runtime_pm(struct drm_i915_private *dev_priv); -void intel_fini_runtime_pm(struct drm_i915_private *dev_priv); void ilk_wm_get_hw_state(struct drm_device *dev); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 043c5a8..95006e0 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -30,9 +30,6 @@ #include "intel_drv.h" #include "../../../platform/x86/intel_ips.h" #include -#include -#include -#include /** * RC6 is a special power stage which allows the GPU to enter an very @@ -6230,1163 +6227,6 @@ void intel_suspend_hw(struct drm_device *dev) lpt_suspend_hw(dev); } -#define for_each_power_well(i, power_well, domain_mask, power_domains) \ - for (i = 0; \ - i < (power_domains)->power_well_count && \ - ((power_well) = &(power_domains)->power_wells[i]); \ - i++) \ - if ((power_well)->domains & (domain_mask)) - -#define for_each_power_well_rev(i, power_well, domain_mask, power_domains) \ - for (i = (power_domains)->power_well_count - 1; \ - i >= 0 && ((power_well) = &(power_domains)->power_wells[i]);\ - i--) \ - if ((power_well)->domains & (domain_mask)) - -/** - * We should only use the power well if we explicitly asked the hardware to - * enable it, so check if it's enabled and also check if we've requested it to - * be enabled. - */ -static bool hsw_power_well_enabled(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - return I915_READ(HSW_PWR_WELL_DRIVER) == - (HSW_PWR_WELL_ENABLE_REQUEST | HSW_PWR_WELL_STATE_ENABLED); -} - -bool intel_display_power_enabled_unlocked(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain) -{ - struct i915_power_domains *power_domains; - struct i915_power_well *power_well; - bool is_enabled; - int i; - - if (dev_priv->pm.suspended) - return false; - - power_domains = &dev_priv->power_domains; - - is_enabled = true; - - for_each_power_well_rev(i, power_well, BIT(domain), power_domains) { - if (power_well->always_on) - continue; - - if (!power_well->hw_enabled) { - is_enabled = false; - break; - } - } - - return is_enabled; -} - -bool intel_display_power_enabled(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain) -{ - struct i915_power_domains *power_domains; - bool ret; - - power_domains = &dev_priv->power_domains; - - mutex_lock(&power_domains->lock); - ret = intel_display_power_enabled_unlocked(dev_priv, domain); - mutex_unlock(&power_domains->lock); - - return ret; -} - -/* - * Starting with Haswell, we have a "Power Down Well" that can be turned off - * when not needed anymore. We have 4 registers that can request the power well - * to be enabled, and it will only be disabled if none of the registers is - * requesting it to be enabled. - */ -static void hsw_power_well_post_enable(struct drm_i915_private *dev_priv) -{ - struct drm_device *dev = dev_priv->dev; - - /* - * After we re-enable the power well, if we touch VGA register 0x3d5 - * we'll get unclaimed register interrupts. This stops after we write - * anything to the VGA MSR register. The vgacon module uses this - * register all the time, so if we unbind our driver and, as a - * consequence, bind vgacon, we'll get stuck in an infinite loop at - * console_unlock(). So make here we touch the VGA MSR register, making - * sure vgacon can keep working normally without triggering interrupts - * and error messages. - */ - vga_get_uninterruptible(dev->pdev, VGA_RSRC_LEGACY_IO); - outb(inb(VGA_MSR_READ), VGA_MSR_WRITE); - vga_put(dev->pdev, VGA_RSRC_LEGACY_IO); - - if (IS_BROADWELL(dev) || (INTEL_INFO(dev)->gen >= 9)) - gen8_irq_power_well_post_enable(dev_priv); -} - -static void hsw_set_power_well(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well, bool enable) -{ - bool is_enabled, enable_requested; - uint32_t tmp; - - tmp = I915_READ(HSW_PWR_WELL_DRIVER); - is_enabled = tmp & HSW_PWR_WELL_STATE_ENABLED; - enable_requested = tmp & HSW_PWR_WELL_ENABLE_REQUEST; - - if (enable) { - if (!enable_requested) - I915_WRITE(HSW_PWR_WELL_DRIVER, - HSW_PWR_WELL_ENABLE_REQUEST); - - if (!is_enabled) { - DRM_DEBUG_KMS("Enabling power well\n"); - if (wait_for((I915_READ(HSW_PWR_WELL_DRIVER) & - HSW_PWR_WELL_STATE_ENABLED), 20)) - DRM_ERROR("Timeout enabling power well\n"); - } - - hsw_power_well_post_enable(dev_priv); - } else { - if (enable_requested) { - I915_WRITE(HSW_PWR_WELL_DRIVER, 0); - POSTING_READ(HSW_PWR_WELL_DRIVER); - DRM_DEBUG_KMS("Requesting to disable the power well\n"); - } - } -} - -static void hsw_power_well_sync_hw(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - hsw_set_power_well(dev_priv, power_well, power_well->count > 0); - - /* - * We're taking over the BIOS, so clear any requests made by it since - * the driver is in charge now. - */ - if (I915_READ(HSW_PWR_WELL_BIOS) & HSW_PWR_WELL_ENABLE_REQUEST) - I915_WRITE(HSW_PWR_WELL_BIOS, 0); -} - -static void hsw_power_well_enable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - hsw_set_power_well(dev_priv, power_well, true); -} - -static void hsw_power_well_disable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - hsw_set_power_well(dev_priv, power_well, false); -} - -static void i9xx_always_on_power_well_noop(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ -} - -static bool i9xx_always_on_power_well_enabled(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - return true; -} - -static void vlv_set_power_well(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well, bool enable) -{ - enum punit_power_well power_well_id = power_well->data; - u32 mask; - u32 state; - u32 ctrl; - - mask = PUNIT_PWRGT_MASK(power_well_id); - state = enable ? PUNIT_PWRGT_PWR_ON(power_well_id) : - PUNIT_PWRGT_PWR_GATE(power_well_id); - - mutex_lock(&dev_priv->rps.hw_lock); - -#define COND \ - ((vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_STATUS) & mask) == state) - - if (COND) - goto out; - - ctrl = vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_CTRL); - ctrl &= ~mask; - ctrl |= state; - vlv_punit_write(dev_priv, PUNIT_REG_PWRGT_CTRL, ctrl); - - if (wait_for(COND, 100)) - DRM_ERROR("timout setting power well state %08x (%08x)\n", - state, - vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_CTRL)); - -#undef COND - -out: - mutex_unlock(&dev_priv->rps.hw_lock); -} - -static void vlv_power_well_sync_hw(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - vlv_set_power_well(dev_priv, power_well, power_well->count > 0); -} - -static void vlv_power_well_enable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - vlv_set_power_well(dev_priv, power_well, true); -} - -static void vlv_power_well_disable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - vlv_set_power_well(dev_priv, power_well, false); -} - -static bool vlv_power_well_enabled(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - int power_well_id = power_well->data; - bool enabled = false; - u32 mask; - u32 state; - u32 ctrl; - - mask = PUNIT_PWRGT_MASK(power_well_id); - ctrl = PUNIT_PWRGT_PWR_ON(power_well_id); - - mutex_lock(&dev_priv->rps.hw_lock); - - state = vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_STATUS) & mask; - /* - * We only ever set the power-on and power-gate states, anything - * else is unexpected. - */ - WARN_ON(state != PUNIT_PWRGT_PWR_ON(power_well_id) && - state != PUNIT_PWRGT_PWR_GATE(power_well_id)); - if (state == ctrl) - enabled = true; - - /* - * A transient state at this point would mean some unexpected party - * is poking at the power controls too. - */ - ctrl = vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_CTRL) & mask; - WARN_ON(ctrl != state); - - mutex_unlock(&dev_priv->rps.hw_lock); - - return enabled; -} - -static void vlv_display_power_well_enable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DISP2D); - - vlv_set_power_well(dev_priv, power_well, true); - - spin_lock_irq(&dev_priv->irq_lock); - valleyview_enable_display_irqs(dev_priv); - spin_unlock_irq(&dev_priv->irq_lock); - - /* - * During driver initialization/resume we can avoid restoring the - * part of the HW/SW state that will be inited anyway explicitly. - */ - if (dev_priv->power_domains.initializing) - return; - - intel_hpd_init(dev_priv->dev); - - i915_redisable_vga_power_on(dev_priv->dev); -} - -static void vlv_display_power_well_disable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DISP2D); - - spin_lock_irq(&dev_priv->irq_lock); - valleyview_disable_display_irqs(dev_priv); - spin_unlock_irq(&dev_priv->irq_lock); - - vlv_set_power_well(dev_priv, power_well, false); - - vlv_power_sequencer_reset(dev_priv); -} - -static void vlv_dpio_cmn_power_well_enable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DPIO_CMN_BC); - - /* - * Enable the CRI clock source so we can get at the - * display and the reference clock for VGA - * hotplug / manual detection. - */ - I915_WRITE(DPLL(PIPE_B), I915_READ(DPLL(PIPE_B)) | - DPLL_REFA_CLK_ENABLE_VLV | DPLL_INTEGRATED_CRI_CLK_VLV); - udelay(1); /* >10ns for cmnreset, >0ns for sidereset */ - - vlv_set_power_well(dev_priv, power_well, true); - - /* - * From VLV2A0_DP_eDP_DPIO_driver_vbios_notes_10.docx - - * 6. De-assert cmn_reset/side_reset. Same as VLV X0. - * a. GUnit 0x2110 bit[0] set to 1 (def 0) - * b. The other bits such as sfr settings / modesel may all - * be set to 0. - * - * This should only be done on init and resume from S3 with - * both PLLs disabled, or we risk losing DPIO and PLL - * synchronization. - */ - I915_WRITE(DPIO_CTL, I915_READ(DPIO_CTL) | DPIO_CMNRST); -} - -static void vlv_dpio_cmn_power_well_disable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - enum pipe pipe; - - WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DPIO_CMN_BC); - - for_each_pipe(dev_priv, pipe) - assert_pll_disabled(dev_priv, pipe); - - /* Assert common reset */ - I915_WRITE(DPIO_CTL, I915_READ(DPIO_CTL) & ~DPIO_CMNRST); - - vlv_set_power_well(dev_priv, power_well, false); -} - -static void chv_dpio_cmn_power_well_enable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - enum dpio_phy phy; - - WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DPIO_CMN_BC && - power_well->data != PUNIT_POWER_WELL_DPIO_CMN_D); - - /* - * Enable the CRI clock source so we can get at the - * display and the reference clock for VGA - * hotplug / manual detection. - */ - if (power_well->data == PUNIT_POWER_WELL_DPIO_CMN_BC) { - phy = DPIO_PHY0; - I915_WRITE(DPLL(PIPE_B), I915_READ(DPLL(PIPE_B)) | - DPLL_REFA_CLK_ENABLE_VLV); - I915_WRITE(DPLL(PIPE_B), I915_READ(DPLL(PIPE_B)) | - DPLL_REFA_CLK_ENABLE_VLV | DPLL_INTEGRATED_CRI_CLK_VLV); - } else { - phy = DPIO_PHY1; - I915_WRITE(DPLL(PIPE_C), I915_READ(DPLL(PIPE_C)) | - DPLL_REFA_CLK_ENABLE_VLV | DPLL_INTEGRATED_CRI_CLK_VLV); - } - udelay(1); /* >10ns for cmnreset, >0ns for sidereset */ - vlv_set_power_well(dev_priv, power_well, true); - - /* Poll for phypwrgood signal */ - if (wait_for(I915_READ(DISPLAY_PHY_STATUS) & PHY_POWERGOOD(phy), 1)) - DRM_ERROR("Display PHY %d is not power up\n", phy); - - I915_WRITE(DISPLAY_PHY_CONTROL, I915_READ(DISPLAY_PHY_CONTROL) | - PHY_COM_LANE_RESET_DEASSERT(phy)); -} - -static void chv_dpio_cmn_power_well_disable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - enum dpio_phy phy; - - WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DPIO_CMN_BC && - power_well->data != PUNIT_POWER_WELL_DPIO_CMN_D); - - if (power_well->data == PUNIT_POWER_WELL_DPIO_CMN_BC) { - phy = DPIO_PHY0; - assert_pll_disabled(dev_priv, PIPE_A); - assert_pll_disabled(dev_priv, PIPE_B); - } else { - phy = DPIO_PHY1; - assert_pll_disabled(dev_priv, PIPE_C); - } - - I915_WRITE(DISPLAY_PHY_CONTROL, I915_READ(DISPLAY_PHY_CONTROL) & - ~PHY_COM_LANE_RESET_DEASSERT(phy)); - - vlv_set_power_well(dev_priv, power_well, false); -} - -static bool chv_pipe_power_well_enabled(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - enum pipe pipe = power_well->data; - bool enabled; - u32 state, ctrl; - - mutex_lock(&dev_priv->rps.hw_lock); - - state = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ) & DP_SSS_MASK(pipe); - /* - * We only ever set the power-on and power-gate states, anything - * else is unexpected. - */ - WARN_ON(state != DP_SSS_PWR_ON(pipe) && state != DP_SSS_PWR_GATE(pipe)); - enabled = state == DP_SSS_PWR_ON(pipe); - - /* - * A transient state at this point would mean some unexpected party - * is poking at the power controls too. - */ - ctrl = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ) & DP_SSC_MASK(pipe); - WARN_ON(ctrl << 16 != state); - - mutex_unlock(&dev_priv->rps.hw_lock); - - return enabled; -} - -static void chv_set_pipe_power_well(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well, - bool enable) -{ - enum pipe pipe = power_well->data; - u32 state; - u32 ctrl; - - state = enable ? DP_SSS_PWR_ON(pipe) : DP_SSS_PWR_GATE(pipe); - - mutex_lock(&dev_priv->rps.hw_lock); - -#define COND \ - ((vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ) & DP_SSS_MASK(pipe)) == state) - - if (COND) - goto out; - - ctrl = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ); - ctrl &= ~DP_SSC_MASK(pipe); - ctrl |= enable ? DP_SSC_PWR_ON(pipe) : DP_SSC_PWR_GATE(pipe); - vlv_punit_write(dev_priv, PUNIT_REG_DSPFREQ, ctrl); - - if (wait_for(COND, 100)) - DRM_ERROR("timout setting power well state %08x (%08x)\n", - state, - vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ)); - -#undef COND - -out: - mutex_unlock(&dev_priv->rps.hw_lock); -} - -static void chv_pipe_power_well_sync_hw(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - chv_set_pipe_power_well(dev_priv, power_well, power_well->count > 0); -} - -static void chv_pipe_power_well_enable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - WARN_ON_ONCE(power_well->data != PIPE_A && - power_well->data != PIPE_B && - power_well->data != PIPE_C); - - chv_set_pipe_power_well(dev_priv, power_well, true); -} - -static void chv_pipe_power_well_disable(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - WARN_ON_ONCE(power_well->data != PIPE_A && - power_well->data != PIPE_B && - power_well->data != PIPE_C); - - chv_set_pipe_power_well(dev_priv, power_well, false); -} - -static void check_power_well_state(struct drm_i915_private *dev_priv, - struct i915_power_well *power_well) -{ - bool enabled = power_well->ops->is_enabled(dev_priv, power_well); - - if (power_well->always_on || !i915.disable_power_well) { - if (!enabled) - goto mismatch; - - return; - } - - if (enabled != (power_well->count > 0)) - goto mismatch; - - return; - -mismatch: - WARN(1, "state mismatch for '%s' (always_on %d hw state %d use-count %d disable_power_well %d\n", - power_well->name, power_well->always_on, enabled, - power_well->count, i915.disable_power_well); -} - -void intel_display_power_get(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain) -{ - struct i915_power_domains *power_domains; - struct i915_power_well *power_well; - int i; - - intel_runtime_pm_get(dev_priv); - - power_domains = &dev_priv->power_domains; - - mutex_lock(&power_domains->lock); - - for_each_power_well(i, power_well, BIT(domain), power_domains) { - if (!power_well->count++) { - DRM_DEBUG_KMS("enabling %s\n", power_well->name); - power_well->ops->enable(dev_priv, power_well); - power_well->hw_enabled = true; - } - - check_power_well_state(dev_priv, power_well); - } - - power_domains->domain_use_count[domain]++; - - mutex_unlock(&power_domains->lock); -} - -void intel_display_power_put(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain) -{ - struct i915_power_domains *power_domains; - struct i915_power_well *power_well; - int i; - - power_domains = &dev_priv->power_domains; - - mutex_lock(&power_domains->lock); - - WARN_ON(!power_domains->domain_use_count[domain]); - power_domains->domain_use_count[domain]--; - - for_each_power_well_rev(i, power_well, BIT(domain), power_domains) { - WARN_ON(!power_well->count); - - if (!--power_well->count && i915.disable_power_well) { - DRM_DEBUG_KMS("disabling %s\n", power_well->name); - power_well->hw_enabled = false; - power_well->ops->disable(dev_priv, power_well); - } - - check_power_well_state(dev_priv, power_well); - } - - mutex_unlock(&power_domains->lock); - - intel_runtime_pm_put(dev_priv); -} - -static struct i915_power_domains *hsw_pwr; - -/* Display audio driver power well request */ -int i915_request_power_well(void) -{ - struct drm_i915_private *dev_priv; - - if (!hsw_pwr) - return -ENODEV; - - dev_priv = container_of(hsw_pwr, struct drm_i915_private, - power_domains); - intel_display_power_get(dev_priv, POWER_DOMAIN_AUDIO); - return 0; -} -EXPORT_SYMBOL_GPL(i915_request_power_well); - -/* Display audio driver power well release */ -int i915_release_power_well(void) -{ - struct drm_i915_private *dev_priv; - - if (!hsw_pwr) - return -ENODEV; - - dev_priv = container_of(hsw_pwr, struct drm_i915_private, - power_domains); - intel_display_power_put(dev_priv, POWER_DOMAIN_AUDIO); - return 0; -} -EXPORT_SYMBOL_GPL(i915_release_power_well); - -/* - * Private interface for the audio driver to get CDCLK in kHz. - * - * Caller must request power well using i915_request_power_well() prior to - * making the call. - */ -int i915_get_cdclk_freq(void) -{ - struct drm_i915_private *dev_priv; - - if (!hsw_pwr) - return -ENODEV; - - dev_priv = container_of(hsw_pwr, struct drm_i915_private, - power_domains); - - return intel_ddi_get_cdclk_freq(dev_priv); -} -EXPORT_SYMBOL_GPL(i915_get_cdclk_freq); - - -#define POWER_DOMAIN_MASK (BIT(POWER_DOMAIN_NUM) - 1) - -#define HSW_ALWAYS_ON_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PIPE_A) | \ - BIT(POWER_DOMAIN_TRANSCODER_EDP) | \ - BIT(POWER_DOMAIN_PORT_DDI_A_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_A_4_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_B_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_C_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_D_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_D_4_LANES) | \ - BIT(POWER_DOMAIN_PORT_CRT) | \ - BIT(POWER_DOMAIN_PLLS) | \ - BIT(POWER_DOMAIN_INIT)) -#define HSW_DISPLAY_POWER_DOMAINS ( \ - (POWER_DOMAIN_MASK & ~HSW_ALWAYS_ON_POWER_DOMAINS) | \ - BIT(POWER_DOMAIN_INIT)) - -#define BDW_ALWAYS_ON_POWER_DOMAINS ( \ - HSW_ALWAYS_ON_POWER_DOMAINS | \ - BIT(POWER_DOMAIN_PIPE_A_PANEL_FITTER)) -#define BDW_DISPLAY_POWER_DOMAINS ( \ - (POWER_DOMAIN_MASK & ~BDW_ALWAYS_ON_POWER_DOMAINS) | \ - BIT(POWER_DOMAIN_INIT)) - -#define VLV_ALWAYS_ON_POWER_DOMAINS BIT(POWER_DOMAIN_INIT) -#define VLV_DISPLAY_POWER_DOMAINS POWER_DOMAIN_MASK - -#define VLV_DPIO_CMN_BC_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_B_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_C_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ - BIT(POWER_DOMAIN_PORT_CRT) | \ - BIT(POWER_DOMAIN_INIT)) - -#define VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_B_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ - BIT(POWER_DOMAIN_INIT)) - -#define VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ - BIT(POWER_DOMAIN_INIT)) - -#define VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_C_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ - BIT(POWER_DOMAIN_INIT)) - -#define VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ - BIT(POWER_DOMAIN_INIT)) - -#define CHV_PIPE_A_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PIPE_A) | \ - BIT(POWER_DOMAIN_INIT)) - -#define CHV_PIPE_B_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PIPE_B) | \ - BIT(POWER_DOMAIN_INIT)) - -#define CHV_PIPE_C_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PIPE_C) | \ - BIT(POWER_DOMAIN_INIT)) - -#define CHV_DPIO_CMN_BC_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_B_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_C_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ - BIT(POWER_DOMAIN_INIT)) - -#define CHV_DPIO_CMN_D_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_D_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_D_4_LANES) | \ - BIT(POWER_DOMAIN_INIT)) - -#define CHV_DPIO_TX_D_LANES_01_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_D_2_LANES) | \ - BIT(POWER_DOMAIN_PORT_DDI_D_4_LANES) | \ - BIT(POWER_DOMAIN_INIT)) - -#define CHV_DPIO_TX_D_LANES_23_POWER_DOMAINS ( \ - BIT(POWER_DOMAIN_PORT_DDI_D_4_LANES) | \ - BIT(POWER_DOMAIN_INIT)) - -static const struct i915_power_well_ops i9xx_always_on_power_well_ops = { - .sync_hw = i9xx_always_on_power_well_noop, - .enable = i9xx_always_on_power_well_noop, - .disable = i9xx_always_on_power_well_noop, - .is_enabled = i9xx_always_on_power_well_enabled, -}; - -static const struct i915_power_well_ops chv_pipe_power_well_ops = { - .sync_hw = chv_pipe_power_well_sync_hw, - .enable = chv_pipe_power_well_enable, - .disable = chv_pipe_power_well_disable, - .is_enabled = chv_pipe_power_well_enabled, -}; - -static const struct i915_power_well_ops chv_dpio_cmn_power_well_ops = { - .sync_hw = vlv_power_well_sync_hw, - .enable = chv_dpio_cmn_power_well_enable, - .disable = chv_dpio_cmn_power_well_disable, - .is_enabled = vlv_power_well_enabled, -}; - -static struct i915_power_well i9xx_always_on_power_well[] = { - { - .name = "always-on", - .always_on = 1, - .domains = POWER_DOMAIN_MASK, - .ops = &i9xx_always_on_power_well_ops, - }, -}; - -static const struct i915_power_well_ops hsw_power_well_ops = { - .sync_hw = hsw_power_well_sync_hw, - .enable = hsw_power_well_enable, - .disable = hsw_power_well_disable, - .is_enabled = hsw_power_well_enabled, -}; - -static struct i915_power_well hsw_power_wells[] = { - { - .name = "always-on", - .always_on = 1, - .domains = HSW_ALWAYS_ON_POWER_DOMAINS, - .ops = &i9xx_always_on_power_well_ops, - }, - { - .name = "display", - .domains = HSW_DISPLAY_POWER_DOMAINS, - .ops = &hsw_power_well_ops, - }, -}; - -static struct i915_power_well bdw_power_wells[] = { - { - .name = "always-on", - .always_on = 1, - .domains = BDW_ALWAYS_ON_POWER_DOMAINS, - .ops = &i9xx_always_on_power_well_ops, - }, - { - .name = "display", - .domains = BDW_DISPLAY_POWER_DOMAINS, - .ops = &hsw_power_well_ops, - }, -}; - -static const struct i915_power_well_ops vlv_display_power_well_ops = { - .sync_hw = vlv_power_well_sync_hw, - .enable = vlv_display_power_well_enable, - .disable = vlv_display_power_well_disable, - .is_enabled = vlv_power_well_enabled, -}; - -static const struct i915_power_well_ops vlv_dpio_cmn_power_well_ops = { - .sync_hw = vlv_power_well_sync_hw, - .enable = vlv_dpio_cmn_power_well_enable, - .disable = vlv_dpio_cmn_power_well_disable, - .is_enabled = vlv_power_well_enabled, -}; - -static const struct i915_power_well_ops vlv_dpio_power_well_ops = { - .sync_hw = vlv_power_well_sync_hw, - .enable = vlv_power_well_enable, - .disable = vlv_power_well_disable, - .is_enabled = vlv_power_well_enabled, -}; - -static struct i915_power_well vlv_power_wells[] = { - { - .name = "always-on", - .always_on = 1, - .domains = VLV_ALWAYS_ON_POWER_DOMAINS, - .ops = &i9xx_always_on_power_well_ops, - }, - { - .name = "display", - .domains = VLV_DISPLAY_POWER_DOMAINS, - .data = PUNIT_POWER_WELL_DISP2D, - .ops = &vlv_display_power_well_ops, - }, - { - .name = "dpio-tx-b-01", - .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_B_LANES_01, - }, - { - .name = "dpio-tx-b-23", - .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_B_LANES_23, - }, - { - .name = "dpio-tx-c-01", - .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_C_LANES_01, - }, - { - .name = "dpio-tx-c-23", - .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_C_LANES_23, - }, - { - .name = "dpio-common", - .domains = VLV_DPIO_CMN_BC_POWER_DOMAINS, - .data = PUNIT_POWER_WELL_DPIO_CMN_BC, - .ops = &vlv_dpio_cmn_power_well_ops, - }, -}; - -static struct i915_power_well chv_power_wells[] = { - { - .name = "always-on", - .always_on = 1, - .domains = VLV_ALWAYS_ON_POWER_DOMAINS, - .ops = &i9xx_always_on_power_well_ops, - }, -#if 0 - { - .name = "display", - .domains = VLV_DISPLAY_POWER_DOMAINS, - .data = PUNIT_POWER_WELL_DISP2D, - .ops = &vlv_display_power_well_ops, - }, - { - .name = "pipe-a", - .domains = CHV_PIPE_A_POWER_DOMAINS, - .data = PIPE_A, - .ops = &chv_pipe_power_well_ops, - }, - { - .name = "pipe-b", - .domains = CHV_PIPE_B_POWER_DOMAINS, - .data = PIPE_B, - .ops = &chv_pipe_power_well_ops, - }, - { - .name = "pipe-c", - .domains = CHV_PIPE_C_POWER_DOMAINS, - .data = PIPE_C, - .ops = &chv_pipe_power_well_ops, - }, -#endif - { - .name = "dpio-common-bc", - /* - * XXX: cmnreset for one PHY seems to disturb the other. - * As a workaround keep both powered on at the same - * time for now. - */ - .domains = CHV_DPIO_CMN_BC_POWER_DOMAINS | CHV_DPIO_CMN_D_POWER_DOMAINS, - .data = PUNIT_POWER_WELL_DPIO_CMN_BC, - .ops = &chv_dpio_cmn_power_well_ops, - }, - { - .name = "dpio-common-d", - /* - * XXX: cmnreset for one PHY seems to disturb the other. - * As a workaround keep both powered on at the same - * time for now. - */ - .domains = CHV_DPIO_CMN_BC_POWER_DOMAINS | CHV_DPIO_CMN_D_POWER_DOMAINS, - .data = PUNIT_POWER_WELL_DPIO_CMN_D, - .ops = &chv_dpio_cmn_power_well_ops, - }, -#if 0 - { - .name = "dpio-tx-b-01", - .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_B_LANES_01, - }, - { - .name = "dpio-tx-b-23", - .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_B_LANES_23, - }, - { - .name = "dpio-tx-c-01", - .domains = VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_C_LANES_01, - }, - { - .name = "dpio-tx-c-23", - .domains = VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | - VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_C_LANES_23, - }, - { - .name = "dpio-tx-d-01", - .domains = CHV_DPIO_TX_D_LANES_01_POWER_DOMAINS | - CHV_DPIO_TX_D_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_D_LANES_01, - }, - { - .name = "dpio-tx-d-23", - .domains = CHV_DPIO_TX_D_LANES_01_POWER_DOMAINS | - CHV_DPIO_TX_D_LANES_23_POWER_DOMAINS, - .ops = &vlv_dpio_power_well_ops, - .data = PUNIT_POWER_WELL_DPIO_TX_D_LANES_23, - }, -#endif -}; - -static struct i915_power_well *lookup_power_well(struct drm_i915_private *dev_priv, - enum punit_power_well power_well_id) -{ - struct i915_power_domains *power_domains = &dev_priv->power_domains; - struct i915_power_well *power_well; - int i; - - for_each_power_well(i, power_well, POWER_DOMAIN_MASK, power_domains) { - if (power_well->data == power_well_id) - return power_well; - } - - return NULL; -} - -#define set_power_wells(power_domains, __power_wells) ({ \ - (power_domains)->power_wells = (__power_wells); \ - (power_domains)->power_well_count = ARRAY_SIZE(__power_wells); \ -}) - -int intel_power_domains_init(struct drm_i915_private *dev_priv) -{ - struct i915_power_domains *power_domains = &dev_priv->power_domains; - - mutex_init(&power_domains->lock); - - /* - * The enabling order will be from lower to higher indexed wells, - * the disabling order is reversed. - */ - if (IS_HASWELL(dev_priv->dev)) { - set_power_wells(power_domains, hsw_power_wells); - hsw_pwr = power_domains; - } else if (IS_BROADWELL(dev_priv->dev)) { - set_power_wells(power_domains, bdw_power_wells); - hsw_pwr = power_domains; - } else if (IS_CHERRYVIEW(dev_priv->dev)) { - set_power_wells(power_domains, chv_power_wells); - } else if (IS_VALLEYVIEW(dev_priv->dev)) { - set_power_wells(power_domains, vlv_power_wells); - } else { - set_power_wells(power_domains, i9xx_always_on_power_well); - } - - return 0; -} - -void intel_power_domains_remove(struct drm_i915_private *dev_priv) -{ - hsw_pwr = NULL; -} - -static void intel_power_domains_resume(struct drm_i915_private *dev_priv) -{ - struct i915_power_domains *power_domains = &dev_priv->power_domains; - struct i915_power_well *power_well; - int i; - - mutex_lock(&power_domains->lock); - for_each_power_well(i, power_well, POWER_DOMAIN_MASK, power_domains) { - power_well->ops->sync_hw(dev_priv, power_well); - power_well->hw_enabled = power_well->ops->is_enabled(dev_priv, - power_well); - } - mutex_unlock(&power_domains->lock); -} - -static void vlv_cmnlane_wa(struct drm_i915_private *dev_priv) -{ - struct i915_power_well *cmn = - lookup_power_well(dev_priv, PUNIT_POWER_WELL_DPIO_CMN_BC); - struct i915_power_well *disp2d = - lookup_power_well(dev_priv, PUNIT_POWER_WELL_DISP2D); - - /* nothing to do if common lane is already off */ - if (!cmn->ops->is_enabled(dev_priv, cmn)) - return; - - /* If the display might be already active skip this */ - if (disp2d->ops->is_enabled(dev_priv, disp2d) && - I915_READ(DPIO_CTL) & DPIO_CMNRST) - return; - - DRM_DEBUG_KMS("toggling display PHY side reset\n"); - - /* cmnlane needs DPLL registers */ - disp2d->ops->enable(dev_priv, disp2d); - - /* - * From VLV2A0_DP_eDP_HDMI_DPIO_driver_vbios_notes_11.docx: - * Need to assert and de-assert PHY SB reset by gating the - * common lane power, then un-gating it. - * Simply ungating isn't enough to reset the PHY enough to get - * ports and lanes running. - */ - cmn->ops->disable(dev_priv, cmn); -} - -void intel_power_domains_init_hw(struct drm_i915_private *dev_priv) -{ - struct drm_device *dev = dev_priv->dev; - struct i915_power_domains *power_domains = &dev_priv->power_domains; - - power_domains->initializing = true; - - if (IS_VALLEYVIEW(dev) && !IS_CHERRYVIEW(dev)) { - mutex_lock(&power_domains->lock); - vlv_cmnlane_wa(dev_priv); - mutex_unlock(&power_domains->lock); - } - - /* For now, we need the power well to be always enabled. */ - intel_display_set_init_power(dev_priv, true); - intel_power_domains_resume(dev_priv); - power_domains->initializing = false; -} - -void intel_aux_display_runtime_get(struct drm_i915_private *dev_priv) -{ - intel_runtime_pm_get(dev_priv); -} - -void intel_aux_display_runtime_put(struct drm_i915_private *dev_priv) -{ - intel_runtime_pm_put(dev_priv); -} - -void intel_runtime_pm_get(struct drm_i915_private *dev_priv) -{ - struct drm_device *dev = dev_priv->dev; - struct device *device = &dev->pdev->dev; - - if (!HAS_RUNTIME_PM(dev)) - return; - - pm_runtime_get_sync(device); - WARN(dev_priv->pm.suspended, "Device still suspended.\n"); -} - -void intel_runtime_pm_get_noresume(struct drm_i915_private *dev_priv) -{ - struct drm_device *dev = dev_priv->dev; - struct device *device = &dev->pdev->dev; - - if (!HAS_RUNTIME_PM(dev)) - return; - - WARN(dev_priv->pm.suspended, "Getting nosync-ref while suspended.\n"); - pm_runtime_get_noresume(device); -} - -void intel_runtime_pm_put(struct drm_i915_private *dev_priv) -{ - struct drm_device *dev = dev_priv->dev; - struct device *device = &dev->pdev->dev; - - if (!HAS_RUNTIME_PM(dev)) - return; - - pm_runtime_mark_last_busy(device); - pm_runtime_put_autosuspend(device); -} - -void intel_init_runtime_pm(struct drm_i915_private *dev_priv) -{ - struct drm_device *dev = dev_priv->dev; - struct device *device = &dev->pdev->dev; - - if (!HAS_RUNTIME_PM(dev)) - return; - - pm_runtime_set_active(device); - - /* - * RPM depends on RC6 to save restore the GT HW context, so make RC6 a - * requirement. - */ - if (!intel_enable_rc6(dev)) { - DRM_INFO("RC6 disabled, disabling runtime PM support\n"); - return; - } - - pm_runtime_set_autosuspend_delay(device, 10000); /* 10s */ - pm_runtime_mark_last_busy(device); - pm_runtime_use_autosuspend(device); - - pm_runtime_put_autosuspend(device); -} - -void intel_fini_runtime_pm(struct drm_i915_private *dev_priv) -{ - struct drm_device *dev = dev_priv->dev; - struct device *device = &dev->pdev->dev; - - if (!HAS_RUNTIME_PM(dev)) - return; - - if (!intel_enable_rc6(dev)) - return; - - /* Make sure we're not suspended first. */ - pm_runtime_get_sync(device); - pm_runtime_disable(device); -} - static void intel_init_fbc(struct drm_i915_private *dev_priv) { if (!HAS_FBC(dev_priv)) { diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c new file mode 100644 index 0000000..2344ecf --- /dev/null +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -0,0 +1,1190 @@ +/* + * Copyright © 2012-2014 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Eugeni Dodonov + * Daniel Vetter + * + */ + +#include +#include + +#include "i915_drv.h" +#include "intel_drv.h" +#include + +static struct i915_power_domains *hsw_pwr; + +#define for_each_power_well(i, power_well, domain_mask, power_domains) \ + for (i = 0; \ + i < (power_domains)->power_well_count && \ + ((power_well) = &(power_domains)->power_wells[i]); \ + i++) \ + if ((power_well)->domains & (domain_mask)) + +#define for_each_power_well_rev(i, power_well, domain_mask, power_domains) \ + for (i = (power_domains)->power_well_count - 1; \ + i >= 0 && ((power_well) = &(power_domains)->power_wells[i]);\ + i--) \ + if ((power_well)->domains & (domain_mask)) + +/** + * We should only use the power well if we explicitly asked the hardware to + * enable it, so check if it's enabled and also check if we've requested it to + * be enabled. + */ +static bool hsw_power_well_enabled(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + return I915_READ(HSW_PWR_WELL_DRIVER) == + (HSW_PWR_WELL_ENABLE_REQUEST | HSW_PWR_WELL_STATE_ENABLED); +} + +bool intel_display_power_enabled_unlocked(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain) +{ + struct i915_power_domains *power_domains; + struct i915_power_well *power_well; + bool is_enabled; + int i; + + if (dev_priv->pm.suspended) + return false; + + power_domains = &dev_priv->power_domains; + + is_enabled = true; + + for_each_power_well_rev(i, power_well, BIT(domain), power_domains) { + if (power_well->always_on) + continue; + + if (!power_well->hw_enabled) { + is_enabled = false; + break; + } + } + + return is_enabled; +} + +bool intel_display_power_enabled(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain) +{ + struct i915_power_domains *power_domains; + bool ret; + + power_domains = &dev_priv->power_domains; + + mutex_lock(&power_domains->lock); + ret = intel_display_power_enabled_unlocked(dev_priv, domain); + mutex_unlock(&power_domains->lock); + + return ret; +} + +/* + * Starting with Haswell, we have a "Power Down Well" that can be turned off + * when not needed anymore. We have 4 registers that can request the power well + * to be enabled, and it will only be disabled if none of the registers is + * requesting it to be enabled. + */ +static void hsw_power_well_post_enable(struct drm_i915_private *dev_priv) +{ + struct drm_device *dev = dev_priv->dev; + + /* + * After we re-enable the power well, if we touch VGA register 0x3d5 + * we'll get unclaimed register interrupts. This stops after we write + * anything to the VGA MSR register. The vgacon module uses this + * register all the time, so if we unbind our driver and, as a + * consequence, bind vgacon, we'll get stuck in an infinite loop at + * console_unlock(). So make here we touch the VGA MSR register, making + * sure vgacon can keep working normally without triggering interrupts + * and error messages. + */ + vga_get_uninterruptible(dev->pdev, VGA_RSRC_LEGACY_IO); + outb(inb(VGA_MSR_READ), VGA_MSR_WRITE); + vga_put(dev->pdev, VGA_RSRC_LEGACY_IO); + + if (IS_BROADWELL(dev) || (INTEL_INFO(dev)->gen >= 9)) + gen8_irq_power_well_post_enable(dev_priv); +} + +static void hsw_set_power_well(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well, bool enable) +{ + bool is_enabled, enable_requested; + uint32_t tmp; + + tmp = I915_READ(HSW_PWR_WELL_DRIVER); + is_enabled = tmp & HSW_PWR_WELL_STATE_ENABLED; + enable_requested = tmp & HSW_PWR_WELL_ENABLE_REQUEST; + + if (enable) { + if (!enable_requested) + I915_WRITE(HSW_PWR_WELL_DRIVER, + HSW_PWR_WELL_ENABLE_REQUEST); + + if (!is_enabled) { + DRM_DEBUG_KMS("Enabling power well\n"); + if (wait_for((I915_READ(HSW_PWR_WELL_DRIVER) & + HSW_PWR_WELL_STATE_ENABLED), 20)) + DRM_ERROR("Timeout enabling power well\n"); + } + + hsw_power_well_post_enable(dev_priv); + } else { + if (enable_requested) { + I915_WRITE(HSW_PWR_WELL_DRIVER, 0); + POSTING_READ(HSW_PWR_WELL_DRIVER); + DRM_DEBUG_KMS("Requesting to disable the power well\n"); + } + } +} + +static void hsw_power_well_sync_hw(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + hsw_set_power_well(dev_priv, power_well, power_well->count > 0); + + /* + * We're taking over the BIOS, so clear any requests made by it since + * the driver is in charge now. + */ + if (I915_READ(HSW_PWR_WELL_BIOS) & HSW_PWR_WELL_ENABLE_REQUEST) + I915_WRITE(HSW_PWR_WELL_BIOS, 0); +} + +static void hsw_power_well_enable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + hsw_set_power_well(dev_priv, power_well, true); +} + +static void hsw_power_well_disable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + hsw_set_power_well(dev_priv, power_well, false); +} + +static void i9xx_always_on_power_well_noop(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ +} + +static bool i9xx_always_on_power_well_enabled(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + return true; +} + +static void vlv_set_power_well(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well, bool enable) +{ + enum punit_power_well power_well_id = power_well->data; + u32 mask; + u32 state; + u32 ctrl; + + mask = PUNIT_PWRGT_MASK(power_well_id); + state = enable ? PUNIT_PWRGT_PWR_ON(power_well_id) : + PUNIT_PWRGT_PWR_GATE(power_well_id); + + mutex_lock(&dev_priv->rps.hw_lock); + +#define COND \ + ((vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_STATUS) & mask) == state) + + if (COND) + goto out; + + ctrl = vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_CTRL); + ctrl &= ~mask; + ctrl |= state; + vlv_punit_write(dev_priv, PUNIT_REG_PWRGT_CTRL, ctrl); + + if (wait_for(COND, 100)) + DRM_ERROR("timout setting power well state %08x (%08x)\n", + state, + vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_CTRL)); + +#undef COND + +out: + mutex_unlock(&dev_priv->rps.hw_lock); +} + +static void vlv_power_well_sync_hw(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + vlv_set_power_well(dev_priv, power_well, power_well->count > 0); +} + +static void vlv_power_well_enable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + vlv_set_power_well(dev_priv, power_well, true); +} + +static void vlv_power_well_disable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + vlv_set_power_well(dev_priv, power_well, false); +} + +static bool vlv_power_well_enabled(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + int power_well_id = power_well->data; + bool enabled = false; + u32 mask; + u32 state; + u32 ctrl; + + mask = PUNIT_PWRGT_MASK(power_well_id); + ctrl = PUNIT_PWRGT_PWR_ON(power_well_id); + + mutex_lock(&dev_priv->rps.hw_lock); + + state = vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_STATUS) & mask; + /* + * We only ever set the power-on and power-gate states, anything + * else is unexpected. + */ + WARN_ON(state != PUNIT_PWRGT_PWR_ON(power_well_id) && + state != PUNIT_PWRGT_PWR_GATE(power_well_id)); + if (state == ctrl) + enabled = true; + + /* + * A transient state at this point would mean some unexpected party + * is poking at the power controls too. + */ + ctrl = vlv_punit_read(dev_priv, PUNIT_REG_PWRGT_CTRL) & mask; + WARN_ON(ctrl != state); + + mutex_unlock(&dev_priv->rps.hw_lock); + + return enabled; +} + +static void vlv_display_power_well_enable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DISP2D); + + vlv_set_power_well(dev_priv, power_well, true); + + spin_lock_irq(&dev_priv->irq_lock); + valleyview_enable_display_irqs(dev_priv); + spin_unlock_irq(&dev_priv->irq_lock); + + /* + * During driver initialization/resume we can avoid restoring the + * part of the HW/SW state that will be inited anyway explicitly. + */ + if (dev_priv->power_domains.initializing) + return; + + intel_hpd_init(dev_priv->dev); + + i915_redisable_vga_power_on(dev_priv->dev); +} + +static void vlv_display_power_well_disable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DISP2D); + + spin_lock_irq(&dev_priv->irq_lock); + valleyview_disable_display_irqs(dev_priv); + spin_unlock_irq(&dev_priv->irq_lock); + + vlv_set_power_well(dev_priv, power_well, false); + + vlv_power_sequencer_reset(dev_priv); +} + +static void vlv_dpio_cmn_power_well_enable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DPIO_CMN_BC); + + /* + * Enable the CRI clock source so we can get at the + * display and the reference clock for VGA + * hotplug / manual detection. + */ + I915_WRITE(DPLL(PIPE_B), I915_READ(DPLL(PIPE_B)) | + DPLL_REFA_CLK_ENABLE_VLV | DPLL_INTEGRATED_CRI_CLK_VLV); + udelay(1); /* >10ns for cmnreset, >0ns for sidereset */ + + vlv_set_power_well(dev_priv, power_well, true); + + /* + * From VLV2A0_DP_eDP_DPIO_driver_vbios_notes_10.docx - + * 6. De-assert cmn_reset/side_reset. Same as VLV X0. + * a. GUnit 0x2110 bit[0] set to 1 (def 0) + * b. The other bits such as sfr settings / modesel may all + * be set to 0. + * + * This should only be done on init and resume from S3 with + * both PLLs disabled, or we risk losing DPIO and PLL + * synchronization. + */ + I915_WRITE(DPIO_CTL, I915_READ(DPIO_CTL) | DPIO_CMNRST); +} + +static void vlv_dpio_cmn_power_well_disable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + enum pipe pipe; + + WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DPIO_CMN_BC); + + for_each_pipe(dev_priv, pipe) + assert_pll_disabled(dev_priv, pipe); + + /* Assert common reset */ + I915_WRITE(DPIO_CTL, I915_READ(DPIO_CTL) & ~DPIO_CMNRST); + + vlv_set_power_well(dev_priv, power_well, false); +} + +static void chv_dpio_cmn_power_well_enable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + enum dpio_phy phy; + + WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DPIO_CMN_BC && + power_well->data != PUNIT_POWER_WELL_DPIO_CMN_D); + + /* + * Enable the CRI clock source so we can get at the + * display and the reference clock for VGA + * hotplug / manual detection. + */ + if (power_well->data == PUNIT_POWER_WELL_DPIO_CMN_BC) { + phy = DPIO_PHY0; + I915_WRITE(DPLL(PIPE_B), I915_READ(DPLL(PIPE_B)) | + DPLL_REFA_CLK_ENABLE_VLV); + I915_WRITE(DPLL(PIPE_B), I915_READ(DPLL(PIPE_B)) | + DPLL_REFA_CLK_ENABLE_VLV | DPLL_INTEGRATED_CRI_CLK_VLV); + } else { + phy = DPIO_PHY1; + I915_WRITE(DPLL(PIPE_C), I915_READ(DPLL(PIPE_C)) | + DPLL_REFA_CLK_ENABLE_VLV | DPLL_INTEGRATED_CRI_CLK_VLV); + } + udelay(1); /* >10ns for cmnreset, >0ns for sidereset */ + vlv_set_power_well(dev_priv, power_well, true); + + /* Poll for phypwrgood signal */ + if (wait_for(I915_READ(DISPLAY_PHY_STATUS) & PHY_POWERGOOD(phy), 1)) + DRM_ERROR("Display PHY %d is not power up\n", phy); + + I915_WRITE(DISPLAY_PHY_CONTROL, I915_READ(DISPLAY_PHY_CONTROL) | + PHY_COM_LANE_RESET_DEASSERT(phy)); +} + +static void chv_dpio_cmn_power_well_disable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + enum dpio_phy phy; + + WARN_ON_ONCE(power_well->data != PUNIT_POWER_WELL_DPIO_CMN_BC && + power_well->data != PUNIT_POWER_WELL_DPIO_CMN_D); + + if (power_well->data == PUNIT_POWER_WELL_DPIO_CMN_BC) { + phy = DPIO_PHY0; + assert_pll_disabled(dev_priv, PIPE_A); + assert_pll_disabled(dev_priv, PIPE_B); + } else { + phy = DPIO_PHY1; + assert_pll_disabled(dev_priv, PIPE_C); + } + + I915_WRITE(DISPLAY_PHY_CONTROL, I915_READ(DISPLAY_PHY_CONTROL) & + ~PHY_COM_LANE_RESET_DEASSERT(phy)); + + vlv_set_power_well(dev_priv, power_well, false); +} + +static bool chv_pipe_power_well_enabled(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + enum pipe pipe = power_well->data; + bool enabled; + u32 state, ctrl; + + mutex_lock(&dev_priv->rps.hw_lock); + + state = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ) & DP_SSS_MASK(pipe); + /* + * We only ever set the power-on and power-gate states, anything + * else is unexpected. + */ + WARN_ON(state != DP_SSS_PWR_ON(pipe) && state != DP_SSS_PWR_GATE(pipe)); + enabled = state == DP_SSS_PWR_ON(pipe); + + /* + * A transient state at this point would mean some unexpected party + * is poking at the power controls too. + */ + ctrl = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ) & DP_SSC_MASK(pipe); + WARN_ON(ctrl << 16 != state); + + mutex_unlock(&dev_priv->rps.hw_lock); + + return enabled; +} + +static void chv_set_pipe_power_well(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well, + bool enable) +{ + enum pipe pipe = power_well->data; + u32 state; + u32 ctrl; + + state = enable ? DP_SSS_PWR_ON(pipe) : DP_SSS_PWR_GATE(pipe); + + mutex_lock(&dev_priv->rps.hw_lock); + +#define COND \ + ((vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ) & DP_SSS_MASK(pipe)) == state) + + if (COND) + goto out; + + ctrl = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ); + ctrl &= ~DP_SSC_MASK(pipe); + ctrl |= enable ? DP_SSC_PWR_ON(pipe) : DP_SSC_PWR_GATE(pipe); + vlv_punit_write(dev_priv, PUNIT_REG_DSPFREQ, ctrl); + + if (wait_for(COND, 100)) + DRM_ERROR("timout setting power well state %08x (%08x)\n", + state, + vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ)); + +#undef COND + +out: + mutex_unlock(&dev_priv->rps.hw_lock); +} + +static void chv_pipe_power_well_sync_hw(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + chv_set_pipe_power_well(dev_priv, power_well, power_well->count > 0); +} + +static void chv_pipe_power_well_enable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + WARN_ON_ONCE(power_well->data != PIPE_A && + power_well->data != PIPE_B && + power_well->data != PIPE_C); + + chv_set_pipe_power_well(dev_priv, power_well, true); +} + +static void chv_pipe_power_well_disable(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + WARN_ON_ONCE(power_well->data != PIPE_A && + power_well->data != PIPE_B && + power_well->data != PIPE_C); + + chv_set_pipe_power_well(dev_priv, power_well, false); +} + +static void check_power_well_state(struct drm_i915_private *dev_priv, + struct i915_power_well *power_well) +{ + bool enabled = power_well->ops->is_enabled(dev_priv, power_well); + + if (power_well->always_on || !i915.disable_power_well) { + if (!enabled) + goto mismatch; + + return; + } + + if (enabled != (power_well->count > 0)) + goto mismatch; + + return; + +mismatch: + WARN(1, "state mismatch for '%s' (always_on %d hw state %d use-count %d disable_power_well %d\n", + power_well->name, power_well->always_on, enabled, + power_well->count, i915.disable_power_well); +} + +void intel_display_power_get(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain) +{ + struct i915_power_domains *power_domains; + struct i915_power_well *power_well; + int i; + + intel_runtime_pm_get(dev_priv); + + power_domains = &dev_priv->power_domains; + + mutex_lock(&power_domains->lock); + + for_each_power_well(i, power_well, BIT(domain), power_domains) { + if (!power_well->count++) { + DRM_DEBUG_KMS("enabling %s\n", power_well->name); + power_well->ops->enable(dev_priv, power_well); + power_well->hw_enabled = true; + } + + check_power_well_state(dev_priv, power_well); + } + + power_domains->domain_use_count[domain]++; + + mutex_unlock(&power_domains->lock); +} + +void intel_display_power_put(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain) +{ + struct i915_power_domains *power_domains; + struct i915_power_well *power_well; + int i; + + power_domains = &dev_priv->power_domains; + + mutex_lock(&power_domains->lock); + + WARN_ON(!power_domains->domain_use_count[domain]); + power_domains->domain_use_count[domain]--; + + for_each_power_well_rev(i, power_well, BIT(domain), power_domains) { + WARN_ON(!power_well->count); + + if (!--power_well->count && i915.disable_power_well) { + DRM_DEBUG_KMS("disabling %s\n", power_well->name); + power_well->hw_enabled = false; + power_well->ops->disable(dev_priv, power_well); + } + + check_power_well_state(dev_priv, power_well); + } + + mutex_unlock(&power_domains->lock); + + intel_runtime_pm_put(dev_priv); +} + +#define POWER_DOMAIN_MASK (BIT(POWER_DOMAIN_NUM) - 1) + +#define HSW_ALWAYS_ON_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PIPE_A) | \ + BIT(POWER_DOMAIN_TRANSCODER_EDP) | \ + BIT(POWER_DOMAIN_PORT_DDI_A_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_A_4_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_B_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_C_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_D_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_D_4_LANES) | \ + BIT(POWER_DOMAIN_PORT_CRT) | \ + BIT(POWER_DOMAIN_PLLS) | \ + BIT(POWER_DOMAIN_INIT)) +#define HSW_DISPLAY_POWER_DOMAINS ( \ + (POWER_DOMAIN_MASK & ~HSW_ALWAYS_ON_POWER_DOMAINS) | \ + BIT(POWER_DOMAIN_INIT)) + +#define BDW_ALWAYS_ON_POWER_DOMAINS ( \ + HSW_ALWAYS_ON_POWER_DOMAINS | \ + BIT(POWER_DOMAIN_PIPE_A_PANEL_FITTER)) +#define BDW_DISPLAY_POWER_DOMAINS ( \ + (POWER_DOMAIN_MASK & ~BDW_ALWAYS_ON_POWER_DOMAINS) | \ + BIT(POWER_DOMAIN_INIT)) + +#define VLV_ALWAYS_ON_POWER_DOMAINS BIT(POWER_DOMAIN_INIT) +#define VLV_DISPLAY_POWER_DOMAINS POWER_DOMAIN_MASK + +#define VLV_DPIO_CMN_BC_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_B_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_C_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ + BIT(POWER_DOMAIN_PORT_CRT) | \ + BIT(POWER_DOMAIN_INIT)) + +#define VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_B_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ + BIT(POWER_DOMAIN_INIT)) + +#define VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ + BIT(POWER_DOMAIN_INIT)) + +#define VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_C_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ + BIT(POWER_DOMAIN_INIT)) + +#define VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ + BIT(POWER_DOMAIN_INIT)) + +#define CHV_PIPE_A_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PIPE_A) | \ + BIT(POWER_DOMAIN_INIT)) + +#define CHV_PIPE_B_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PIPE_B) | \ + BIT(POWER_DOMAIN_INIT)) + +#define CHV_PIPE_C_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PIPE_C) | \ + BIT(POWER_DOMAIN_INIT)) + +#define CHV_DPIO_CMN_BC_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_B_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_B_4_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_C_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_C_4_LANES) | \ + BIT(POWER_DOMAIN_INIT)) + +#define CHV_DPIO_CMN_D_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_D_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_D_4_LANES) | \ + BIT(POWER_DOMAIN_INIT)) + +#define CHV_DPIO_TX_D_LANES_01_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_D_2_LANES) | \ + BIT(POWER_DOMAIN_PORT_DDI_D_4_LANES) | \ + BIT(POWER_DOMAIN_INIT)) + +#define CHV_DPIO_TX_D_LANES_23_POWER_DOMAINS ( \ + BIT(POWER_DOMAIN_PORT_DDI_D_4_LANES) | \ + BIT(POWER_DOMAIN_INIT)) + +static const struct i915_power_well_ops i9xx_always_on_power_well_ops = { + .sync_hw = i9xx_always_on_power_well_noop, + .enable = i9xx_always_on_power_well_noop, + .disable = i9xx_always_on_power_well_noop, + .is_enabled = i9xx_always_on_power_well_enabled, +}; + +static const struct i915_power_well_ops chv_pipe_power_well_ops = { + .sync_hw = chv_pipe_power_well_sync_hw, + .enable = chv_pipe_power_well_enable, + .disable = chv_pipe_power_well_disable, + .is_enabled = chv_pipe_power_well_enabled, +}; + +static const struct i915_power_well_ops chv_dpio_cmn_power_well_ops = { + .sync_hw = vlv_power_well_sync_hw, + .enable = chv_dpio_cmn_power_well_enable, + .disable = chv_dpio_cmn_power_well_disable, + .is_enabled = vlv_power_well_enabled, +}; + +static struct i915_power_well i9xx_always_on_power_well[] = { + { + .name = "always-on", + .always_on = 1, + .domains = POWER_DOMAIN_MASK, + .ops = &i9xx_always_on_power_well_ops, + }, +}; + +static const struct i915_power_well_ops hsw_power_well_ops = { + .sync_hw = hsw_power_well_sync_hw, + .enable = hsw_power_well_enable, + .disable = hsw_power_well_disable, + .is_enabled = hsw_power_well_enabled, +}; + +static struct i915_power_well hsw_power_wells[] = { + { + .name = "always-on", + .always_on = 1, + .domains = HSW_ALWAYS_ON_POWER_DOMAINS, + .ops = &i9xx_always_on_power_well_ops, + }, + { + .name = "display", + .domains = HSW_DISPLAY_POWER_DOMAINS, + .ops = &hsw_power_well_ops, + }, +}; + +static struct i915_power_well bdw_power_wells[] = { + { + .name = "always-on", + .always_on = 1, + .domains = BDW_ALWAYS_ON_POWER_DOMAINS, + .ops = &i9xx_always_on_power_well_ops, + }, + { + .name = "display", + .domains = BDW_DISPLAY_POWER_DOMAINS, + .ops = &hsw_power_well_ops, + }, +}; + +static const struct i915_power_well_ops vlv_display_power_well_ops = { + .sync_hw = vlv_power_well_sync_hw, + .enable = vlv_display_power_well_enable, + .disable = vlv_display_power_well_disable, + .is_enabled = vlv_power_well_enabled, +}; + +static const struct i915_power_well_ops vlv_dpio_cmn_power_well_ops = { + .sync_hw = vlv_power_well_sync_hw, + .enable = vlv_dpio_cmn_power_well_enable, + .disable = vlv_dpio_cmn_power_well_disable, + .is_enabled = vlv_power_well_enabled, +}; + +static const struct i915_power_well_ops vlv_dpio_power_well_ops = { + .sync_hw = vlv_power_well_sync_hw, + .enable = vlv_power_well_enable, + .disable = vlv_power_well_disable, + .is_enabled = vlv_power_well_enabled, +}; + +static struct i915_power_well vlv_power_wells[] = { + { + .name = "always-on", + .always_on = 1, + .domains = VLV_ALWAYS_ON_POWER_DOMAINS, + .ops = &i9xx_always_on_power_well_ops, + }, + { + .name = "display", + .domains = VLV_DISPLAY_POWER_DOMAINS, + .data = PUNIT_POWER_WELL_DISP2D, + .ops = &vlv_display_power_well_ops, + }, + { + .name = "dpio-tx-b-01", + .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_B_LANES_01, + }, + { + .name = "dpio-tx-b-23", + .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_B_LANES_23, + }, + { + .name = "dpio-tx-c-01", + .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_C_LANES_01, + }, + { + .name = "dpio-tx-c-23", + .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_C_LANES_23, + }, + { + .name = "dpio-common", + .domains = VLV_DPIO_CMN_BC_POWER_DOMAINS, + .data = PUNIT_POWER_WELL_DPIO_CMN_BC, + .ops = &vlv_dpio_cmn_power_well_ops, + }, +}; + +static struct i915_power_well chv_power_wells[] = { + { + .name = "always-on", + .always_on = 1, + .domains = VLV_ALWAYS_ON_POWER_DOMAINS, + .ops = &i9xx_always_on_power_well_ops, + }, +#if 0 + { + .name = "display", + .domains = VLV_DISPLAY_POWER_DOMAINS, + .data = PUNIT_POWER_WELL_DISP2D, + .ops = &vlv_display_power_well_ops, + }, + { + .name = "pipe-a", + .domains = CHV_PIPE_A_POWER_DOMAINS, + .data = PIPE_A, + .ops = &chv_pipe_power_well_ops, + }, + { + .name = "pipe-b", + .domains = CHV_PIPE_B_POWER_DOMAINS, + .data = PIPE_B, + .ops = &chv_pipe_power_well_ops, + }, + { + .name = "pipe-c", + .domains = CHV_PIPE_C_POWER_DOMAINS, + .data = PIPE_C, + .ops = &chv_pipe_power_well_ops, + }, +#endif + { + .name = "dpio-common-bc", + /* + * XXX: cmnreset for one PHY seems to disturb the other. + * As a workaround keep both powered on at the same + * time for now. + */ + .domains = CHV_DPIO_CMN_BC_POWER_DOMAINS | CHV_DPIO_CMN_D_POWER_DOMAINS, + .data = PUNIT_POWER_WELL_DPIO_CMN_BC, + .ops = &chv_dpio_cmn_power_well_ops, + }, + { + .name = "dpio-common-d", + /* + * XXX: cmnreset for one PHY seems to disturb the other. + * As a workaround keep both powered on at the same + * time for now. + */ + .domains = CHV_DPIO_CMN_BC_POWER_DOMAINS | CHV_DPIO_CMN_D_POWER_DOMAINS, + .data = PUNIT_POWER_WELL_DPIO_CMN_D, + .ops = &chv_dpio_cmn_power_well_ops, + }, +#if 0 + { + .name = "dpio-tx-b-01", + .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_B_LANES_01, + }, + { + .name = "dpio-tx-b-23", + .domains = VLV_DPIO_TX_B_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_B_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_B_LANES_23, + }, + { + .name = "dpio-tx-c-01", + .domains = VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_C_LANES_01, + }, + { + .name = "dpio-tx-c-23", + .domains = VLV_DPIO_TX_C_LANES_01_POWER_DOMAINS | + VLV_DPIO_TX_C_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_C_LANES_23, + }, + { + .name = "dpio-tx-d-01", + .domains = CHV_DPIO_TX_D_LANES_01_POWER_DOMAINS | + CHV_DPIO_TX_D_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_D_LANES_01, + }, + { + .name = "dpio-tx-d-23", + .domains = CHV_DPIO_TX_D_LANES_01_POWER_DOMAINS | + CHV_DPIO_TX_D_LANES_23_POWER_DOMAINS, + .ops = &vlv_dpio_power_well_ops, + .data = PUNIT_POWER_WELL_DPIO_TX_D_LANES_23, + }, +#endif +}; + +static struct i915_power_well *lookup_power_well(struct drm_i915_private *dev_priv, + enum punit_power_well power_well_id) +{ + struct i915_power_domains *power_domains = &dev_priv->power_domains; + struct i915_power_well *power_well; + int i; + + for_each_power_well(i, power_well, POWER_DOMAIN_MASK, power_domains) { + if (power_well->data == power_well_id) + return power_well; + } + + return NULL; +} + +#define set_power_wells(power_domains, __power_wells) ({ \ + (power_domains)->power_wells = (__power_wells); \ + (power_domains)->power_well_count = ARRAY_SIZE(__power_wells); \ +}) + +int intel_power_domains_init(struct drm_i915_private *dev_priv) +{ + struct i915_power_domains *power_domains = &dev_priv->power_domains; + + mutex_init(&power_domains->lock); + + /* + * The enabling order will be from lower to higher indexed wells, + * the disabling order is reversed. + */ + if (IS_HASWELL(dev_priv->dev)) { + set_power_wells(power_domains, hsw_power_wells); + hsw_pwr = power_domains; + } else if (IS_BROADWELL(dev_priv->dev)) { + set_power_wells(power_domains, bdw_power_wells); + hsw_pwr = power_domains; + } else if (IS_CHERRYVIEW(dev_priv->dev)) { + set_power_wells(power_domains, chv_power_wells); + } else if (IS_VALLEYVIEW(dev_priv->dev)) { + set_power_wells(power_domains, vlv_power_wells); + } else { + set_power_wells(power_domains, i9xx_always_on_power_well); + } + + return 0; +} + +void intel_power_domains_remove(struct drm_i915_private *dev_priv) +{ + hsw_pwr = NULL; +} + +static void intel_power_domains_resume(struct drm_i915_private *dev_priv) +{ + struct i915_power_domains *power_domains = &dev_priv->power_domains; + struct i915_power_well *power_well; + int i; + + mutex_lock(&power_domains->lock); + for_each_power_well(i, power_well, POWER_DOMAIN_MASK, power_domains) { + power_well->ops->sync_hw(dev_priv, power_well); + power_well->hw_enabled = power_well->ops->is_enabled(dev_priv, + power_well); + } + mutex_unlock(&power_domains->lock); +} + +static void vlv_cmnlane_wa(struct drm_i915_private *dev_priv) +{ + struct i915_power_well *cmn = + lookup_power_well(dev_priv, PUNIT_POWER_WELL_DPIO_CMN_BC); + struct i915_power_well *disp2d = + lookup_power_well(dev_priv, PUNIT_POWER_WELL_DISP2D); + + /* nothing to do if common lane is already off */ + if (!cmn->ops->is_enabled(dev_priv, cmn)) + return; + + /* If the display might be already active skip this */ + if (disp2d->ops->is_enabled(dev_priv, disp2d) && + I915_READ(DPIO_CTL) & DPIO_CMNRST) + return; + + DRM_DEBUG_KMS("toggling display PHY side reset\n"); + + /* cmnlane needs DPLL registers */ + disp2d->ops->enable(dev_priv, disp2d); + + /* + * From VLV2A0_DP_eDP_HDMI_DPIO_driver_vbios_notes_11.docx: + * Need to assert and de-assert PHY SB reset by gating the + * common lane power, then un-gating it. + * Simply ungating isn't enough to reset the PHY enough to get + * ports and lanes running. + */ + cmn->ops->disable(dev_priv, cmn); +} + +void intel_power_domains_init_hw(struct drm_i915_private *dev_priv) +{ + struct drm_device *dev = dev_priv->dev; + struct i915_power_domains *power_domains = &dev_priv->power_domains; + + power_domains->initializing = true; + + if (IS_VALLEYVIEW(dev) && !IS_CHERRYVIEW(dev)) { + mutex_lock(&power_domains->lock); + vlv_cmnlane_wa(dev_priv); + mutex_unlock(&power_domains->lock); + } + + /* For now, we need the power well to be always enabled. */ + intel_display_set_init_power(dev_priv, true); + intel_power_domains_resume(dev_priv); + power_domains->initializing = false; +} + +void intel_aux_display_runtime_get(struct drm_i915_private *dev_priv) +{ + intel_runtime_pm_get(dev_priv); +} + +void intel_aux_display_runtime_put(struct drm_i915_private *dev_priv) +{ + intel_runtime_pm_put(dev_priv); +} + +void intel_runtime_pm_get(struct drm_i915_private *dev_priv) +{ + struct drm_device *dev = dev_priv->dev; + struct device *device = &dev->pdev->dev; + + if (!HAS_RUNTIME_PM(dev)) + return; + + pm_runtime_get_sync(device); + WARN(dev_priv->pm.suspended, "Device still suspended.\n"); +} + +void intel_runtime_pm_get_noresume(struct drm_i915_private *dev_priv) +{ + struct drm_device *dev = dev_priv->dev; + struct device *device = &dev->pdev->dev; + + if (!HAS_RUNTIME_PM(dev)) + return; + + WARN(dev_priv->pm.suspended, "Getting nosync-ref while suspended.\n"); + pm_runtime_get_noresume(device); +} + +void intel_runtime_pm_put(struct drm_i915_private *dev_priv) +{ + struct drm_device *dev = dev_priv->dev; + struct device *device = &dev->pdev->dev; + + if (!HAS_RUNTIME_PM(dev)) + return; + + pm_runtime_mark_last_busy(device); + pm_runtime_put_autosuspend(device); +} + +void intel_init_runtime_pm(struct drm_i915_private *dev_priv) +{ + struct drm_device *dev = dev_priv->dev; + struct device *device = &dev->pdev->dev; + + if (!HAS_RUNTIME_PM(dev)) + return; + + pm_runtime_set_active(device); + + /* + * RPM depends on RC6 to save restore the GT HW context, so make RC6 a + * requirement. + */ + if (!intel_enable_rc6(dev)) { + DRM_INFO("RC6 disabled, disabling runtime PM support\n"); + return; + } + + pm_runtime_set_autosuspend_delay(device, 10000); /* 10s */ + pm_runtime_mark_last_busy(device); + pm_runtime_use_autosuspend(device); + + pm_runtime_put_autosuspend(device); +} + +void intel_fini_runtime_pm(struct drm_i915_private *dev_priv) +{ + struct drm_device *dev = dev_priv->dev; + struct device *device = &dev->pdev->dev; + + if (!HAS_RUNTIME_PM(dev)) + return; + + if (!intel_enable_rc6(dev)) + return; + + /* Make sure we're not suspended first. */ + pm_runtime_get_sync(device); + pm_runtime_disable(device); +} + +/* Display audio driver power well request */ +int i915_request_power_well(void) +{ + struct drm_i915_private *dev_priv; + + if (!hsw_pwr) + return -ENODEV; + + dev_priv = container_of(hsw_pwr, struct drm_i915_private, + power_domains); + intel_display_power_get(dev_priv, POWER_DOMAIN_AUDIO); + return 0; +} +EXPORT_SYMBOL_GPL(i915_request_power_well); + +/* Display audio driver power well release */ +int i915_release_power_well(void) +{ + struct drm_i915_private *dev_priv; + + if (!hsw_pwr) + return -ENODEV; + + dev_priv = container_of(hsw_pwr, struct drm_i915_private, + power_domains); + intel_display_power_put(dev_priv, POWER_DOMAIN_AUDIO); + return 0; +} +EXPORT_SYMBOL_GPL(i915_release_power_well); + +/* + * Private interface for the audio driver to get CDCLK in kHz. + * + * Caller must request power well using i915_request_power_well() prior to + * making the call. + */ +int i915_get_cdclk_freq(void) +{ + struct drm_i915_private *dev_priv; + + if (!hsw_pwr) + return -ENODEV; + + dev_priv = container_of(hsw_pwr, struct drm_i915_private, + power_domains); + + return intel_ddi_get_cdclk_freq(dev_priv); +} +EXPORT_SYMBOL_GPL(i915_get_cdclk_freq); -- cgit v0.10.2 From f458ebbc3329a6f064e0f40ed5909fe3a7133238 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:39 +0200 Subject: drm/i915: Bikeshed rpm functions name a bit. - fini goes with init, so call it intel_power_domains_fini. While at it shovel some of the fini code that leaked out of it back in. - give power_enabled functions the verb _is_ to make the meaning clearer. Also use a __ prefix instead of _unlocked to really discourage users. - rename runtime_pm_init/fini to enable/disable since that's what they do. Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 14c88c2..da4036d 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -716,7 +716,7 @@ static int i915_interrupt_info(struct seq_file *m, void *data) } for_each_pipe(dev_priv, pipe) { - if (!intel_display_power_enabled(dev_priv, + if (!intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_PIPE(pipe))) { seq_printf(m, "Pipe %c power disabled\n", pipe_name(pipe)); diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 1c035c4..261fc0b 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1798,12 +1798,12 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) if (IS_GEN5(dev)) intel_gpu_ips_init(dev_priv); - intel_init_runtime_pm(dev_priv); + intel_runtime_pm_enable(dev_priv); return 0; out_power_well: - intel_power_domains_remove(dev_priv); + intel_power_domains_fini(dev_priv); drm_vblank_cleanup(dev); out_gem_unload: WARN_ON(unregister_oom_notifier(&dev_priv->mm.oom_notifier)); @@ -1846,15 +1846,11 @@ int i915_driver_unload(struct drm_device *dev) return ret; } - intel_fini_runtime_pm(dev_priv); + intel_runtime_pm_disable(dev_priv); intel_gpu_ips_teardown(); - /* The i915.ko module is still not prepared to be loaded when - * the power well is not enabled, so just enable it in case - * we're going to unload/reload. */ - intel_display_set_init_power(dev_priv, true); - intel_power_domains_remove(dev_priv); + intel_power_domains_fini(dev_priv); i915_teardown_sysfs(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 622c038..5dbd5c9 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1411,7 +1411,7 @@ struct ilk_wm_values { * * Our driver uses the autosuspend delay feature, which means we'll only really * suspend if we stay with zero refcount for a certain amount of time. The - * default value is currently very conservative (see intel_init_runtime_pm), but + * default value is currently very conservative (see intel_runtime_pm_enable), but * it can be changed with the standard runtime PM files from sysfs. * * The irqs_disabled variable becomes true exactly after we disable the IRQs and diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 080981b..eef74ef 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3473,8 +3473,8 @@ static void gen8_irq_reset(struct drm_device *dev) gen8_gt_irq_reset(dev_priv); for_each_pipe(dev_priv, pipe) - if (intel_display_power_enabled(dev_priv, - POWER_DOMAIN_PIPE(pipe))) + if (intel_display_power_is_enabled(dev_priv, + POWER_DOMAIN_PIPE(pipe))) GEN8_IRQ_RESET_NDX(DE_PIPE, pipe); GEN5_IRQ_RESET(GEN8_DE_PORT_); @@ -3826,7 +3826,7 @@ static void gen8_de_irq_postinstall(struct drm_i915_private *dev_priv) dev_priv->de_irq_mask[PIPE_C] = ~de_pipe_masked; for_each_pipe(dev_priv, pipe) - if (intel_display_power_enabled(dev_priv, + if (intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_PIPE(pipe))) GEN8_IRQ_INIT_NDX(DE_PIPE, pipe, dev_priv->de_irq_mask[pipe], diff --git a/drivers/gpu/drm/i915/intel_crt.c b/drivers/gpu/drm/i915/intel_crt.c index 9212e65..dacaad5 100644 --- a/drivers/gpu/drm/i915/intel_crt.c +++ b/drivers/gpu/drm/i915/intel_crt.c @@ -72,7 +72,7 @@ static bool intel_crt_get_hw_state(struct intel_encoder *encoder, u32 tmp; power_domain = intel_display_port_power_domain(encoder); - if (!intel_display_power_enabled(dev_priv, power_domain)) + if (!intel_display_power_is_enabled(dev_priv, power_domain)) return false; tmp = I915_READ(crt->adpa_reg); diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index c9f4b3c..a151de7d 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -998,7 +998,7 @@ bool intel_ddi_connector_get_hw_state(struct intel_connector *intel_connector) uint32_t tmp; power_domain = intel_display_port_power_domain(intel_encoder); - if (!intel_display_power_enabled(dev_priv, power_domain)) + if (!intel_display_power_is_enabled(dev_priv, power_domain)) return false; if (!intel_encoder->get_hw_state(intel_encoder, &pipe)) @@ -1044,7 +1044,7 @@ bool intel_ddi_get_hw_state(struct intel_encoder *encoder, int i; power_domain = intel_display_port_power_domain(encoder); - if (!intel_display_power_enabled(dev_priv, power_domain)) + if (!intel_display_power_is_enabled(dev_priv, power_domain)) return false; tmp = I915_READ(DDI_BUF_CTL(port)); @@ -1332,7 +1332,7 @@ static bool hsw_ddi_pll_get_hw_state(struct drm_i915_private *dev_priv, { uint32_t val; - if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_PLLS)) + if (!intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_PLLS)) return false; val = I915_READ(WRPLL_CTL(pll->id)); @@ -1522,7 +1522,7 @@ void intel_ddi_get_config(struct intel_encoder *encoder, break; } - if (intel_display_power_enabled(dev_priv, POWER_DOMAIN_AUDIO)) { + if (intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_AUDIO)) { temp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); if (temp & (AUDIO_OUTPUT_ENABLE_A << (intel_crtc->pipe * 4))) pipe_config->has_audio = true; diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index a960578..ae2d67d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -1210,7 +1210,7 @@ void assert_pipe(struct drm_i915_private *dev_priv, (pipe == PIPE_B && dev_priv->quirks & QUIRK_PIPEB_FORCE)) state = true; - if (!intel_display_power_enabled(dev_priv, + if (!intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_TRANSCODER(cpu_transcoder))) { cur_state = false; } else { @@ -6493,8 +6493,8 @@ static bool i9xx_get_pipe_config(struct intel_crtc *crtc, struct drm_i915_private *dev_priv = dev->dev_private; uint32_t tmp; - if (!intel_display_power_enabled(dev_priv, - POWER_DOMAIN_PIPE(crtc->pipe))) + if (!intel_display_power_is_enabled(dev_priv, + POWER_DOMAIN_PIPE(crtc->pipe))) return false; pipe_config->cpu_transcoder = (enum transcoder) crtc->pipe; @@ -7503,8 +7503,8 @@ static bool ironlake_get_pipe_config(struct intel_crtc *crtc, struct drm_i915_private *dev_priv = dev->dev_private; uint32_t tmp; - if (!intel_display_power_enabled(dev_priv, - POWER_DOMAIN_PIPE(crtc->pipe))) + if (!intel_display_power_is_enabled(dev_priv, + POWER_DOMAIN_PIPE(crtc->pipe))) return false; pipe_config->cpu_transcoder = (enum transcoder) crtc->pipe; @@ -7902,7 +7902,7 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, enum intel_display_power_domain pfit_domain; uint32_t tmp; - if (!intel_display_power_enabled(dev_priv, + if (!intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_PIPE(crtc->pipe))) return false; @@ -7931,7 +7931,7 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, pipe_config->cpu_transcoder = TRANSCODER_EDP; } - if (!intel_display_power_enabled(dev_priv, + if (!intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_TRANSCODER(pipe_config->cpu_transcoder))) return false; @@ -7944,7 +7944,7 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, intel_get_pipe_timings(crtc, pipe_config); pfit_domain = POWER_DOMAIN_PIPE_PANEL_FITTER(crtc->pipe); - if (intel_display_power_enabled(dev_priv, pfit_domain)) + if (intel_display_power_is_enabled(dev_priv, pfit_domain)) ironlake_get_pfit_config(crtc, pipe_config); if (IS_HASWELL(dev)) @@ -11534,7 +11534,7 @@ static bool ibx_pch_dpll_get_hw_state(struct drm_i915_private *dev_priv, { uint32_t val; - if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_PLLS)) + if (!intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_PLLS)) return false; val = I915_READ(PCH_DPLL(pll->id)); @@ -13165,7 +13165,7 @@ void i915_redisable_vga(struct drm_device *dev) * level, just check if the power well is enabled instead of trying to * follow the "don't touch the power well if we don't need it" policy * the rest of the driver uses. */ - if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_VGA)) + if (!intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_VGA)) return; i915_redisable_vga_power_on(dev); @@ -13543,8 +13543,8 @@ intel_display_capture_error_state(struct drm_device *dev) for_each_pipe(dev_priv, i) { error->pipe[i].power_domain_on = - intel_display_power_enabled_unlocked(dev_priv, - POWER_DOMAIN_PIPE(i)); + __intel_display_power_is_enabled(dev_priv, + POWER_DOMAIN_PIPE(i)); if (!error->pipe[i].power_domain_on) continue; @@ -13579,7 +13579,7 @@ intel_display_capture_error_state(struct drm_device *dev) enum transcoder cpu_transcoder = transcoders[i]; error->transcoder[i].power_domain_on = - intel_display_power_enabled_unlocked(dev_priv, + __intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_TRANSCODER(cpu_transcoder)); if (!error->transcoder[i].power_domain_on) continue; diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 799918f..205363a 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1853,7 +1853,7 @@ static bool intel_dp_get_hw_state(struct intel_encoder *encoder, u32 tmp; power_domain = intel_display_port_power_domain(encoder); - if (!intel_display_power_enabled(dev_priv, power_domain)) + if (!intel_display_power_is_enabled(dev_priv, power_domain)) return false; tmp = I915_READ(intel_dp->output_reg); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 9d2ee70..dc0c6d5 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1083,15 +1083,15 @@ extern struct drm_display_mode *intel_find_panel_downclock( /* intel_runtime_pm.c */ int intel_power_domains_init(struct drm_i915_private *); -void intel_power_domains_remove(struct drm_i915_private *); +void intel_power_domains_fini(struct drm_i915_private *); void intel_power_domains_init_hw(struct drm_i915_private *dev_priv); -void intel_init_runtime_pm(struct drm_i915_private *dev_priv); -void intel_fini_runtime_pm(struct drm_i915_private *dev_priv); +void intel_runtime_pm_enable(struct drm_i915_private *dev_priv); +void intel_runtime_pm_disable(struct drm_i915_private *dev_priv); -bool intel_display_power_enabled(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain); -bool intel_display_power_enabled_unlocked(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain); +bool intel_display_power_is_enabled(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain); +bool __intel_display_power_is_enabled(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain); void intel_display_power_get(struct drm_i915_private *dev_priv, enum intel_display_power_domain domain); void intel_display_power_put(struct drm_i915_private *dev_priv, diff --git a/drivers/gpu/drm/i915/intel_dsi.c b/drivers/gpu/drm/i915/intel_dsi.c index 5bd9e09..0b18407 100644 --- a/drivers/gpu/drm/i915/intel_dsi.c +++ b/drivers/gpu/drm/i915/intel_dsi.c @@ -344,7 +344,7 @@ static bool intel_dsi_get_hw_state(struct intel_encoder *encoder, DRM_DEBUG_KMS("\n"); power_domain = intel_display_port_power_domain(encoder); - if (!intel_display_power_enabled(dev_priv, power_domain)) + if (!intel_display_power_is_enabled(dev_priv, power_domain)) return false; /* XXX: this only works for one DSI output */ diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 7fed5be..cf87def 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -690,7 +690,7 @@ static bool intel_hdmi_get_hw_state(struct intel_encoder *encoder, u32 tmp; power_domain = intel_display_port_power_domain(encoder); - if (!intel_display_power_enabled(dev_priv, power_domain)) + if (!intel_display_power_is_enabled(dev_priv, power_domain)) return false; tmp = I915_READ(intel_hdmi->hdmi_reg); diff --git a/drivers/gpu/drm/i915/intel_lvds.c b/drivers/gpu/drm/i915/intel_lvds.c index a6bd142..2b50c98 100644 --- a/drivers/gpu/drm/i915/intel_lvds.c +++ b/drivers/gpu/drm/i915/intel_lvds.c @@ -76,7 +76,7 @@ static bool intel_lvds_get_hw_state(struct intel_encoder *encoder, u32 tmp; power_domain = intel_display_port_power_domain(encoder); - if (!intel_display_power_enabled(dev_priv, power_domain)) + if (!intel_display_power_is_enabled(dev_priv, power_domain)) return false; tmp = I915_READ(lvds_encoder->reg); diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index 2344ecf..dc298c6 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -60,8 +60,8 @@ static bool hsw_power_well_enabled(struct drm_i915_private *dev_priv, (HSW_PWR_WELL_ENABLE_REQUEST | HSW_PWR_WELL_STATE_ENABLED); } -bool intel_display_power_enabled_unlocked(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain) +bool __intel_display_power_is_enabled(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain) { struct i915_power_domains *power_domains; struct i915_power_well *power_well; @@ -88,8 +88,8 @@ bool intel_display_power_enabled_unlocked(struct drm_i915_private *dev_priv, return is_enabled; } -bool intel_display_power_enabled(struct drm_i915_private *dev_priv, - enum intel_display_power_domain domain) +bool intel_display_power_is_enabled(struct drm_i915_private *dev_priv, + enum intel_display_power_domain domain) { struct i915_power_domains *power_domains; bool ret; @@ -97,7 +97,7 @@ bool intel_display_power_enabled(struct drm_i915_private *dev_priv, power_domains = &dev_priv->power_domains; mutex_lock(&power_domains->lock); - ret = intel_display_power_enabled_unlocked(dev_priv, domain); + ret = __intel_display_power_is_enabled(dev_priv, domain); mutex_unlock(&power_domains->lock); return ret; @@ -981,8 +981,13 @@ int intel_power_domains_init(struct drm_i915_private *dev_priv) return 0; } -void intel_power_domains_remove(struct drm_i915_private *dev_priv) +void intel_power_domains_fini(struct drm_i915_private *dev_priv) { + /* The i915.ko module is still not prepared to be loaded when + * the power well is not enabled, so just enable it in case + * we're going to unload/reload. */ + intel_display_set_init_power(dev_priv, true); + hsw_pwr = NULL; } @@ -1097,7 +1102,7 @@ void intel_runtime_pm_put(struct drm_i915_private *dev_priv) pm_runtime_put_autosuspend(device); } -void intel_init_runtime_pm(struct drm_i915_private *dev_priv) +void intel_runtime_pm_enable(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; struct device *device = &dev->pdev->dev; @@ -1123,7 +1128,7 @@ void intel_init_runtime_pm(struct drm_i915_private *dev_priv) pm_runtime_put_autosuspend(device); } -void intel_fini_runtime_pm(struct drm_i915_private *dev_priv) +void intel_runtime_pm_disable(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; struct device *device = &dev->pdev->dev; -- cgit v0.10.2 From d9bc89d9e39c2c1a76cd37b5e8cc1592a7898848 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:40 +0200 Subject: drm/i915: Move intel_display_set_init_power to intel_runtime_pm.c I've decided to not move intel_display_port_power_domain because that's just a hack in our design ... Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ae2d67d..7877b16 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -4560,20 +4560,6 @@ static unsigned long get_crtc_power_domains(struct drm_crtc *crtc) return mask; } -void intel_display_set_init_power(struct drm_i915_private *dev_priv, - bool enable) -{ - if (dev_priv->power_domains.init_power_on == enable) - return; - - if (enable) - intel_display_power_get(dev_priv, POWER_DOMAIN_INIT); - else - intel_display_power_put(dev_priv, POWER_DOMAIN_INIT); - - dev_priv->power_domains.init_power_on = enable; -} - static void modeset_update_crtc_power_domains(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index dc0c6d5..64fe05e 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -934,7 +934,6 @@ ironlake_check_encoder_dotclock(const struct intel_crtc_config *pipe_config, bool intel_crtc_active(struct drm_crtc *crtc); void hsw_enable_ips(struct intel_crtc *crtc); void hsw_disable_ips(struct intel_crtc *crtc); -void intel_display_set_init_power(struct drm_i915_private *dev, bool enable); enum intel_display_power_domain intel_display_port_power_domain(struct intel_encoder *intel_encoder); void intel_mode_from_pipe_config(struct drm_display_mode *mode, @@ -1102,6 +1101,8 @@ void intel_runtime_pm_get(struct drm_i915_private *dev_priv); void intel_runtime_pm_get_noresume(struct drm_i915_private *dev_priv); void intel_runtime_pm_put(struct drm_i915_private *dev_priv); +void intel_display_set_init_power(struct drm_i915_private *dev, bool enable); + /* intel_pm.c */ void intel_init_clock_gating(struct drm_device *dev); void intel_suspend_hw(struct drm_device *dev); diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index dc298c6..34f91c1 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -103,6 +103,20 @@ bool intel_display_power_is_enabled(struct drm_i915_private *dev_priv, return ret; } +void intel_display_set_init_power(struct drm_i915_private *dev_priv, + bool enable) +{ + if (dev_priv->power_domains.init_power_on == enable) + return; + + if (enable) + intel_display_power_get(dev_priv, POWER_DOMAIN_INIT); + else + intel_display_power_put(dev_priv, POWER_DOMAIN_INIT); + + dev_priv->power_domains.init_power_on = enable; +} + /* * Starting with Haswell, we have a "Power Down Well" that can be turned off * when not needed anymore. We have 4 registers that can request the power well -- cgit v0.10.2 From 41373cd53fdeb658fadd8cd0313591ce49e4342f Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:41 +0200 Subject: drm/i915: Call runtime_pm_disable directly Allows us to mark it static and so forgoe the kerneldoc for it. Note that intel_power_domains_fini is also called from failure paths in the driver load sequence. But the call to runtime_pm_disable for that is harmless since by default runtime pm is already disabled. v2: Augment the commit message as discussed with Imre on irc. Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 261fc0b..d58c711 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1846,12 +1846,10 @@ int i915_driver_unload(struct drm_device *dev) return ret; } - intel_runtime_pm_disable(dev_priv); + intel_power_domains_fini(dev_priv); intel_gpu_ips_teardown(); - intel_power_domains_fini(dev_priv); - i915_teardown_sysfs(dev); WARN_ON(unregister_oom_notifier(&dev_priv->mm.oom_notifier)); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 64fe05e..1deb5d8 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1085,7 +1085,6 @@ int intel_power_domains_init(struct drm_i915_private *); void intel_power_domains_fini(struct drm_i915_private *); void intel_power_domains_init_hw(struct drm_i915_private *dev_priv); void intel_runtime_pm_enable(struct drm_i915_private *dev_priv); -void intel_runtime_pm_disable(struct drm_i915_private *dev_priv); bool intel_display_power_is_enabled(struct drm_i915_private *dev_priv, enum intel_display_power_domain domain); diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index 34f91c1..dc63b78 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -995,8 +995,26 @@ int intel_power_domains_init(struct drm_i915_private *dev_priv) return 0; } +static void intel_runtime_pm_disable(struct drm_i915_private *dev_priv) +{ + struct drm_device *dev = dev_priv->dev; + struct device *device = &dev->pdev->dev; + + if (!HAS_RUNTIME_PM(dev)) + return; + + if (!intel_enable_rc6(dev)) + return; + + /* Make sure we're not suspended first. */ + pm_runtime_get_sync(device); + pm_runtime_disable(device); +} + void intel_power_domains_fini(struct drm_i915_private *dev_priv) { + intel_runtime_pm_disable(dev_priv); + /* The i915.ko module is still not prepared to be loaded when * the power well is not enabled, so just enable it in case * we're going to unload/reload. */ @@ -1142,22 +1160,6 @@ void intel_runtime_pm_enable(struct drm_i915_private *dev_priv) pm_runtime_put_autosuspend(device); } -void intel_runtime_pm_disable(struct drm_i915_private *dev_priv) -{ - struct drm_device *dev = dev_priv->dev; - struct device *device = &dev->pdev->dev; - - if (!HAS_RUNTIME_PM(dev)) - return; - - if (!intel_enable_rc6(dev)) - return; - - /* Make sure we're not suspended first. */ - pm_runtime_get_sync(device); - pm_runtime_disable(device); -} - /* Display audio driver power well request */ int i915_request_power_well(void) { -- cgit v0.10.2 From e4e7684fc5c5e6ef9d4fdbbc2f08917a61cbd708 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:42 +0200 Subject: drm/i915: Kerneldoc for intel_runtime_pm.c I've decided not to document the functions exported to the audio driver since really, they shouldn't exist ... v2: Improvements from Imre's review plus a few more spelling fixes I've spotted. Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index 98528b4..8627816 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -3788,6 +3788,18 @@ int num_ioctls; those have basic support through the gma500 drm driver. + Core Driver Infrastructure + + This section covers core driver infrastructure used by both the display + and the GEM parts of the driver. + + + Runtime Power Management +!Pdrivers/gpu/drm/i915/intel_runtime_pm.c runtime pm +!Idrivers/gpu/drm/i915/intel_runtime_pm.c + + + Display Hardware Handling This section covers everything related to the display hardware including diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index dc63b78..abefc98 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -33,6 +33,23 @@ #include "intel_drv.h" #include +/** + * DOC: runtime pm + * + * The i915 driver supports dynamic enabling and disabling of entire hardware + * blocks at runtime. This is especially important on the display side where + * software is supposed to control many power gates manually on recent hardware, + * since on the GT side a lot of the power management is done by the hardware. + * But even there some manual control at the device level is required. + * + * Since i915 supports a diverse set of platforms with a unified codebase and + * hardware engineers just love to shuffle functionality around between power + * domains there's a sizeable amount of indirection required. This file provides + * generic functions to the driver for grabbing and releasing references for + * abstract power domains. It then maps those to the actual power wells + * present for a given platform. + */ + static struct i915_power_domains *hsw_pwr; #define for_each_power_well(i, power_well, domain_mask, power_domains) \ @@ -48,7 +65,7 @@ static struct i915_power_domains *hsw_pwr; i--) \ if ((power_well)->domains & (domain_mask)) -/** +/* * We should only use the power well if we explicitly asked the hardware to * enable it, so check if it's enabled and also check if we've requested it to * be enabled. @@ -60,6 +77,18 @@ static bool hsw_power_well_enabled(struct drm_i915_private *dev_priv, (HSW_PWR_WELL_ENABLE_REQUEST | HSW_PWR_WELL_STATE_ENABLED); } +/** + * __intel_display_power_is_enabled - unlocked check for a power domain + * @dev_priv: i915 device instance + * @domain: power domain to check + * + * This is the unlocked version of intel_display_power_is_enabled() and should + * only be used from error capture and recovery code where deadlocks are + * possible. + * + * Returns: + * True when the power domain is enabled, false otherwise. + */ bool __intel_display_power_is_enabled(struct drm_i915_private *dev_priv, enum intel_display_power_domain domain) { @@ -88,6 +117,23 @@ bool __intel_display_power_is_enabled(struct drm_i915_private *dev_priv, return is_enabled; } +/** + * intel_display_power_is_enabled - unlocked check for a power domain + * @dev_priv: i915 device instance + * @domain: power domain to check + * + * This function can be used to check the hw power domain state. It is mostly + * used in hardware state readout functions. Everywhere else code should rely + * upon explicit power domain reference counting to ensure that the hardware + * block is powered up before accessing it. + * + * Callers must hold the relevant modesetting locks to ensure that concurrent + * threads can't disable the power well while the caller tries to read a few + * registers. + * + * Returns: + * True when the power domain is enabled, false otherwise. + */ bool intel_display_power_is_enabled(struct drm_i915_private *dev_priv, enum intel_display_power_domain domain) { @@ -103,6 +149,16 @@ bool intel_display_power_is_enabled(struct drm_i915_private *dev_priv, return ret; } +/** + * intel_display_set_init_power - set the initial power domain state + * @dev_priv: i915 device instance + * @enable: whether to enable or disable the initial power domain state + * + * For simplicity our driver load/unload and system suspend/resume code assumes + * that all power domains are always enabled. This functions controls the state + * of this little hack. While the initial power domain state is enabled runtime + * pm is effectively disabled. + */ void intel_display_set_init_power(struct drm_i915_private *dev_priv, bool enable) { @@ -556,6 +612,18 @@ mismatch: power_well->count, i915.disable_power_well); } +/** + * intel_display_power_get - grab a power domain reference + * @dev_priv: i915 device instance + * @domain: power domain to reference + * + * This function grabs a power domain reference for @domain and ensures that the + * power domain and all its parents are powered up. Therefore users should only + * grab a reference to the innermost power domain they need. + * + * Any power domain reference obtained by this function must have a symmetric + * call to intel_display_power_put() to release the reference again. + */ void intel_display_power_get(struct drm_i915_private *dev_priv, enum intel_display_power_domain domain) { @@ -584,6 +652,15 @@ void intel_display_power_get(struct drm_i915_private *dev_priv, mutex_unlock(&power_domains->lock); } +/** + * intel_display_power_put - release a power domain reference + * @dev_priv: i915 device instance + * @domain: power domain to reference + * + * This function drops the power domain reference obtained by + * intel_display_power_get() and might power down the corresponding hardware + * block right away if this is the last reference. + */ void intel_display_power_put(struct drm_i915_private *dev_priv, enum intel_display_power_domain domain) { @@ -968,6 +1045,13 @@ static struct i915_power_well *lookup_power_well(struct drm_i915_private *dev_pr (power_domains)->power_well_count = ARRAY_SIZE(__power_wells); \ }) +/** + * intel_power_domains_init - initializes the power domain structures + * @dev_priv: i915 device instance + * + * Initializes the power domain structures for @dev_priv depending upon the + * supported platform. + */ int intel_power_domains_init(struct drm_i915_private *dev_priv) { struct i915_power_domains *power_domains = &dev_priv->power_domains; @@ -1011,6 +1095,14 @@ static void intel_runtime_pm_disable(struct drm_i915_private *dev_priv) pm_runtime_disable(device); } +/** + * intel_power_domains_fini - finalizes the power domain structures + * @dev_priv: i915 device instance + * + * Finalizes the power domain structures for @dev_priv depending upon the + * supported platform. This function also disables runtime pm and ensures that + * the device stays powered up so that the driver can be reloaded. + */ void intel_power_domains_fini(struct drm_i915_private *dev_priv) { intel_runtime_pm_disable(dev_priv); @@ -1069,6 +1161,13 @@ static void vlv_cmnlane_wa(struct drm_i915_private *dev_priv) cmn->ops->disable(dev_priv, cmn); } +/** + * intel_power_domains_init_hw - initialize hardware power domain state + * @dev_priv: i915 device instance + * + * This function initializes the hardware power domain state and enables all + * power domains using intel_display_set_init_power(). + */ void intel_power_domains_init_hw(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; @@ -1088,16 +1187,46 @@ void intel_power_domains_init_hw(struct drm_i915_private *dev_priv) power_domains->initializing = false; } +/** + * intel_aux_display_runtime_get - grab an auxilliary power domain reference + * @dev_priv: i915 device instance + * + * This function grabs a power domain reference for the auxiliary power domain + * (for access to the GMBUS and DP AUX blocks) and ensures that it and all its + * parents are powered up. Therefore users should only grab a reference to the + * innermost power domain they need. + * + * Any power domain reference obtained by this function must have a symmetric + * call to intel_aux_display_runtime_put() to release the reference again. + */ void intel_aux_display_runtime_get(struct drm_i915_private *dev_priv) { intel_runtime_pm_get(dev_priv); } +/** + * intel_aux_display_runtime_put - release an auxilliary power domain reference + * @dev_priv: i915 device instance + * + * This function drops the auxilliary power domain reference obtained by + * intel_aux_display_runtime_get() and might power down the corresponding + * hardware block right away if this is the last reference. + */ void intel_aux_display_runtime_put(struct drm_i915_private *dev_priv) { intel_runtime_pm_put(dev_priv); } +/** + * intel_runtime_pm_get - grab a runtime pm reference + * @dev_priv: i915 device instance + * + * This function grabs a device-level runtime pm reference (mostly used for GEM + * code to ensure the GTT or GT is on) and ensures that it is powered up. + * + * Any runtime pm reference obtained by this function must have a symmetric + * call to intel_runtime_pm_put() to release the reference again. + */ void intel_runtime_pm_get(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; @@ -1110,6 +1239,23 @@ void intel_runtime_pm_get(struct drm_i915_private *dev_priv) WARN(dev_priv->pm.suspended, "Device still suspended.\n"); } +/** + * intel_runtime_pm_get_noresume - grab a runtime pm reference + * @dev_priv: i915 device instance + * + * This function grabs a device-level runtime pm reference (mostly used for GEM + * code to ensure the GTT or GT is on). + * + * It will _not_ power up the device but instead only check that it's powered + * on. Therefore it is only valid to call this functions from contexts where + * the device is known to be powered up and where trying to power it up would + * result in hilarity and deadlocks. That pretty much means only the system + * suspend/resume code where this is used to grab runtime pm references for + * delayed setup down in work items. + * + * Any runtime pm reference obtained by this function must have a symmetric + * call to intel_runtime_pm_put() to release the reference again. + */ void intel_runtime_pm_get_noresume(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; @@ -1122,6 +1268,14 @@ void intel_runtime_pm_get_noresume(struct drm_i915_private *dev_priv) pm_runtime_get_noresume(device); } +/** + * intel_runtime_pm_put - release a runtime pm reference + * @dev_priv: i915 device instance + * + * This function drops the device-level runtime pm reference obtained by + * intel_runtime_pm_get() and might power down the corresponding + * hardware block right away if this is the last reference. + */ void intel_runtime_pm_put(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; @@ -1134,6 +1288,16 @@ void intel_runtime_pm_put(struct drm_i915_private *dev_priv) pm_runtime_put_autosuspend(device); } +/** + * intel_runtime_pm_enable - enable runtime pm + * @dev_priv: i915 device instance + * + * This function enables runtime pm at the end of the driver load sequence. + * + * Note that this function does currently not enable runtime pm for the + * subordinate display power domains. That is only done on the first modeset + * using intel_display_set_init_power(). + */ void intel_runtime_pm_enable(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; -- cgit v0.10.2 From d8f2716a1a46bed27ce210e74c7fec3855595200 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 1 Oct 2014 01:02:04 +0200 Subject: drm/i915: Reinstate error level message for non-simulated gpu hangs This seems to have been accidentally lost in commit be62acb4cce1389a28296852737e3917d9cc5b25 Author: Mika Kuoppala Date: Fri Aug 30 16:19:28 2013 +0300 drm/i915: ban badly behaving contexts Without this real gpu hangs only log output at info level, which gets filtered away by piglit's testrunner. v2: Tune down to notice level. Note that we need to add drm/i915 so that at least the automatic igt dmesg filtering still picks it up. v3: git add and lack of coffee don't mix well. v4: Message is in between hw and sw reset, so switch verb to continuous form. v5: Use i915_stop_rings_allow_warn for consistency. For Chris' case of injecting lots of hangs I guess we need to revamp this all anyway when merging. For now this should plug the regression for piglit testing mesa. v6: Make it compile (Mika). Cc: Mika Kuoppala Cc: Chris Wilson Cc: Kenneth Graunke Reported-by: Kenneth Graunke Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index ea93ff1..66489fa 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -844,6 +844,9 @@ int i915_reset(struct drm_device *dev) } } + if (i915_stop_ring_allow_warn(dev_priv)) + pr_notice("drm/i915: Resetting chip after gpu hang\n"); + if (ret) { DRM_ERROR("Failed to reset chip: %i\n", ret); mutex_unlock(&dev->struct_mutex); -- cgit v0.10.2 From 5ca476f8874eec7eaec06d402dff2e6b47886d28 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 1 Oct 2014 16:56:56 +0300 Subject: drm/i915: De-magic the PSR AUX message MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use pack_aux() to construct the PSR exit DPMS D0 AUX message, and use the defines from dp_dp_helper.h to populate the message contents. v2: Use sizeof() for message size (Jani) Use a generic loop to write EDP_PSR_AUX_DATA registers Cc: Jani Nikula Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index c62f3eb..607d7a5 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -2503,9 +2503,7 @@ enum punit_power_well { #define EDP_PSR_AUX_CTL(dev) (EDP_PSR_BASE(dev) + 0x10) #define EDP_PSR_AUX_DATA1(dev) (EDP_PSR_BASE(dev) + 0x14) -#define EDP_PSR_DPCD_COMMAND 0x80060000 #define EDP_PSR_AUX_DATA2(dev) (EDP_PSR_BASE(dev) + 0x18) -#define EDP_PSR_DPCD_NORMAL_OPERATION (1<<24) #define EDP_PSR_AUX_DATA3(dev) (EDP_PSR_BASE(dev) + 0x1c) #define EDP_PSR_AUX_DATA4(dev) (EDP_PSR_BASE(dev) + 0x20) #define EDP_PSR_AUX_DATA5(dev) (EDP_PSR_BASE(dev) + 0x24) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 205363a..7ba3e1b 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -225,7 +225,7 @@ intel_dp_mode_valid(struct drm_connector *connector, } static uint32_t -pack_aux(uint8_t *src, int src_bytes) +pack_aux(const uint8_t *src, int src_bytes) { int i; uint32_t v = 0; @@ -2045,8 +2045,17 @@ static void intel_edp_psr_enable_sink(struct intel_dp *intel_dp) struct drm_i915_private *dev_priv = dev->dev_private; uint32_t aux_clock_divider; int precharge = 0x3; - int msg_size = 5; /* Header(4) + Message(1) */ bool only_standby = false; + static const uint8_t aux_msg[] = { + [0] = DP_AUX_NATIVE_WRITE << 4, + [1] = DP_SET_POWER >> 8, + [2] = DP_SET_POWER & 0xff, + [3] = 1 - 1, + [4] = DP_SET_POWER_D0, + }; + int i; + + BUILD_BUG_ON(sizeof(aux_msg) > 20); aux_clock_divider = intel_dp->get_aux_clock_divider(intel_dp, 0); @@ -2062,11 +2071,13 @@ static void intel_edp_psr_enable_sink(struct intel_dp *intel_dp) DP_PSR_ENABLE | DP_PSR_MAIN_LINK_ACTIVE); /* Setup AUX registers */ - I915_WRITE(EDP_PSR_AUX_DATA1(dev), EDP_PSR_DPCD_COMMAND); - I915_WRITE(EDP_PSR_AUX_DATA2(dev), EDP_PSR_DPCD_NORMAL_OPERATION); + for (i = 0; i < sizeof(aux_msg); i += 4) + I915_WRITE(EDP_PSR_AUX_DATA1(dev) + i, + pack_aux(&aux_msg[i], sizeof(aux_msg) - i)); + I915_WRITE(EDP_PSR_AUX_CTL(dev), DP_AUX_CH_CTL_TIME_OUT_400us | - (msg_size << DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT) | + (sizeof(aux_msg) << DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT) | (precharge << DP_AUX_CH_CTL_PRECHARGE_2US_SHIFT) | (aux_clock_divider << DP_AUX_CH_CTL_BIT_CLOCK_2X_SHIFT)); } -- cgit v0.10.2 From bd9f74a5e7365546c9c858ca8fa585906cba3979 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Thu, 2 Oct 2014 09:45:35 +0200 Subject: drm/i915: Constify send buffer for intel_dp_aux_ch MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Inspired by Ville constifying the send buffer for pach_aux. Cc: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 7ba3e1b..b2aedcf 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -718,7 +718,7 @@ static uint32_t skl_get_aux_send_ctl(struct intel_dp *intel_dp, static int intel_dp_aux_ch(struct intel_dp *intel_dp, - uint8_t *send, int send_bytes, + const uint8_t *send, int send_bytes, uint8_t *recv, int recv_size) { struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); -- cgit v0.10.2 From 1dda5f930451a8b9cbe915182cc64811fd67e186 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 1 Oct 2014 07:32:37 -0700 Subject: drm/i915: make sink_crc return -EIO on aux read/write failure Even though it's unliky, we should check each aux transaction not just the first one. Also commit ce31d9f4fc05964f6c0dd3a8661dc1a1d843a1e2 Author: Rodrigo Vivi Date: Mon Sep 29 18:29:52 2014 -0400 drm/i915: preserve other DP_TEST_SINK bits. added a new aux transaction before the one which was checked. Fix this. Cc: Daniel Vetter Cc: Todd Previte Signed-off-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index b2aedcf..50bcd09 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3875,16 +3875,21 @@ int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc) if (!(buf & DP_TEST_CRC_SUPPORTED)) return -ENOTTY; - drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK, &buf); + if (drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK, &buf) < 0) + return -EIO; + if (drm_dp_dpcd_writeb(&intel_dp->aux, DP_TEST_SINK, buf | DP_TEST_SINK_START) < 0) return -EIO; - drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK_MISC, &buf); + if (drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK_MISC, &buf) < 0) + return -EIO; test_crc_count = buf & DP_TEST_COUNT_MASK; do { - drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK_MISC, &buf); + if (drm_dp_dpcd_readb(&intel_dp->aux, + DP_TEST_SINK_MISC, &buf) < 0) + return -EIO; intel_wait_for_vblank(dev, intel_crtc->pipe); } while (--attempts && (buf & DP_TEST_COUNT_MASK) == test_crc_count); @@ -3896,9 +3901,11 @@ int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc) if (drm_dp_dpcd_read(&intel_dp->aux, DP_TEST_CRC_R_CR, crc, 6) < 0) return -EIO; - drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK, &buf); - drm_dp_dpcd_writeb(&intel_dp->aux, DP_TEST_SINK, - buf & ~DP_TEST_SINK_START); + if (drm_dp_dpcd_readb(&intel_dp->aux, DP_TEST_SINK, &buf) < 0) + return -EIO; + if (drm_dp_dpcd_writeb(&intel_dp->aux, DP_TEST_SINK, + buf & ~DP_TEST_SINK_START) < 0) + return -EIO; return 0; } -- cgit v0.10.2 From a02ef3c7193c942c286cef815c3dc2b2f03660b4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 18 Aug 2014 14:42:45 +0300 Subject: drm/i915: Make sure hardware uses the correct swing margin/deemph bits on chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The register can house two different swing marging/deemph settings at once. However only one gets used based on some other bits. Make sure we set those bits correctly to make the hardware use the settings we provided. Signed-off-by: Ville Syrjälä Reviewed-by: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 607d7a5..d0df76d 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -836,12 +836,31 @@ enum punit_power_well { #define _VLV_PCS_DW9_CH0 0x8224 #define _VLV_PCS_DW9_CH1 0x8424 +#define DPIO_PCS_TX2MARGIN_MASK (0x7<<13) +#define DPIO_PCS_TX2MARGIN_000 (0<<13) +#define DPIO_PCS_TX2MARGIN_101 (1<<13) +#define DPIO_PCS_TX1MARGIN_MASK (0x7<<10) +#define DPIO_PCS_TX1MARGIN_000 (0<<10) +#define DPIO_PCS_TX1MARGIN_101 (1<<10) #define VLV_PCS_DW9(ch) _PORT(ch, _VLV_PCS_DW9_CH0, _VLV_PCS_DW9_CH1) +#define _VLV_PCS01_DW9_CH0 0x224 +#define _VLV_PCS23_DW9_CH0 0x424 +#define _VLV_PCS01_DW9_CH1 0x2624 +#define _VLV_PCS23_DW9_CH1 0x2824 +#define VLV_PCS01_DW9(ch) _PORT(ch, _VLV_PCS01_DW9_CH0, _VLV_PCS01_DW9_CH1) +#define VLV_PCS23_DW9(ch) _PORT(ch, _VLV_PCS23_DW9_CH0, _VLV_PCS23_DW9_CH1) + #define _CHV_PCS_DW10_CH0 0x8228 #define _CHV_PCS_DW10_CH1 0x8428 #define DPIO_PCS_SWING_CALC_TX0_TX2 (1<<30) #define DPIO_PCS_SWING_CALC_TX1_TX3 (1<<31) +#define DPIO_PCS_TX2DEEMP_MASK (0xf<<24) +#define DPIO_PCS_TX2DEEMP_9P5 (0<<24) +#define DPIO_PCS_TX2DEEMP_6P0 (2<<24) +#define DPIO_PCS_TX1DEEMP_MASK (0xf<<16) +#define DPIO_PCS_TX1DEEMP_9P5 (0<<16) +#define DPIO_PCS_TX1DEEMP_6P0 (2<<16) #define CHV_PCS_DW10(ch) _PORT(ch, _CHV_PCS_DW10_CH0, _CHV_PCS_DW10_CH1) #define _VLV_PCS01_DW10_CH0 0x0228 diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 50bcd09..22d6c3e 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3152,12 +3152,26 @@ static uint32_t intel_chv_signal_levels(struct intel_dp *intel_dp) /* Clear calc init */ val = vlv_dpio_read(dev_priv, pipe, VLV_PCS01_DW10(ch)); val &= ~(DPIO_PCS_SWING_CALC_TX0_TX2 | DPIO_PCS_SWING_CALC_TX1_TX3); + val &= ~(DPIO_PCS_TX1DEEMP_MASK | DPIO_PCS_TX2DEEMP_MASK); + val |= DPIO_PCS_TX1DEEMP_9P5 | DPIO_PCS_TX2DEEMP_9P5; vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW10(ch), val); val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW10(ch)); val &= ~(DPIO_PCS_SWING_CALC_TX0_TX2 | DPIO_PCS_SWING_CALC_TX1_TX3); + val &= ~(DPIO_PCS_TX1DEEMP_MASK | DPIO_PCS_TX2DEEMP_MASK); + val |= DPIO_PCS_TX1DEEMP_9P5 | DPIO_PCS_TX2DEEMP_9P5; vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW10(ch), val); + val = vlv_dpio_read(dev_priv, pipe, VLV_PCS01_DW9(ch)); + val &= ~(DPIO_PCS_TX1MARGIN_MASK | DPIO_PCS_TX2MARGIN_MASK); + val |= DPIO_PCS_TX1MARGIN_000 | DPIO_PCS_TX2MARGIN_000; + vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW9(ch), val); + + val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW9(ch)); + val &= ~(DPIO_PCS_TX1MARGIN_MASK | DPIO_PCS_TX2MARGIN_MASK); + val |= DPIO_PCS_TX1MARGIN_000 | DPIO_PCS_TX2MARGIN_000; + vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW9(ch), val); + /* Program swing deemph */ for (i = 0; i < 4; i++) { val = vlv_dpio_read(dev_priv, pipe, CHV_TX_DW4(ch, i)); diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index cf87def..386932a 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -1436,12 +1436,26 @@ static void chv_hdmi_pre_enable(struct intel_encoder *encoder) /* Clear calc init */ val = vlv_dpio_read(dev_priv, pipe, VLV_PCS01_DW10(ch)); val &= ~(DPIO_PCS_SWING_CALC_TX0_TX2 | DPIO_PCS_SWING_CALC_TX1_TX3); + val &= ~(DPIO_PCS_TX1DEEMP_MASK | DPIO_PCS_TX2DEEMP_MASK); + val |= DPIO_PCS_TX1DEEMP_9P5 | DPIO_PCS_TX2DEEMP_9P5; vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW10(ch), val); val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW10(ch)); val &= ~(DPIO_PCS_SWING_CALC_TX0_TX2 | DPIO_PCS_SWING_CALC_TX1_TX3); + val &= ~(DPIO_PCS_TX1DEEMP_MASK | DPIO_PCS_TX2DEEMP_MASK); + val |= DPIO_PCS_TX1DEEMP_9P5 | DPIO_PCS_TX2DEEMP_9P5; vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW10(ch), val); + val = vlv_dpio_read(dev_priv, pipe, VLV_PCS01_DW9(ch)); + val &= ~(DPIO_PCS_TX1MARGIN_MASK | DPIO_PCS_TX2MARGIN_MASK); + val |= DPIO_PCS_TX1MARGIN_000 | DPIO_PCS_TX2MARGIN_000; + vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW9(ch), val); + + val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW9(ch)); + val &= ~(DPIO_PCS_TX1MARGIN_MASK | DPIO_PCS_TX2MARGIN_MASK); + val |= DPIO_PCS_TX1MARGIN_000 | DPIO_PCS_TX2MARGIN_000; + vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW9(ch), val); + /* FIXME: Program the support xxx V-dB */ /* Use 800mV-0dB */ for (i = 0; i < 4; i++) { -- cgit v0.10.2 From 570e2a747bc06cd8620662c5125ec2dc964c511b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 18 Aug 2014 14:42:46 +0300 Subject: drm/i915: Clear TX FIFO reset master override bits on chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Clear the override bits to make sure the hardware manages the TX FIFO reset master on its own. v2: Squash with the earlier attempt at forcing the override bits Signed-off-by: Ville Syrjälä Reviewed-by: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index d0df76d..2ed02c3 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -796,6 +796,8 @@ enum punit_power_well { #define _VLV_PCS_DW0_CH1 0x8400 #define DPIO_PCS_TX_LANE2_RESET (1<<16) #define DPIO_PCS_TX_LANE1_RESET (1<<7) +#define DPIO_LEFT_TXFIFO_RST_MASTER2 (1<<4) +#define DPIO_RIGHT_TXFIFO_RST_MASTER2 (1<<3) #define VLV_PCS_DW0(ch) _PORT(ch, _VLV_PCS_DW0_CH0, _VLV_PCS_DW0_CH1) #define _VLV_PCS01_DW0_CH0 0x200 @@ -872,8 +874,18 @@ enum punit_power_well { #define _VLV_PCS_DW11_CH0 0x822c #define _VLV_PCS_DW11_CH1 0x842c +#define DPIO_LANEDESKEW_STRAP_OVRD (1<<3) +#define DPIO_LEFT_TXFIFO_RST_MASTER (1<<1) +#define DPIO_RIGHT_TXFIFO_RST_MASTER (1<<0) #define VLV_PCS_DW11(ch) _PORT(ch, _VLV_PCS_DW11_CH0, _VLV_PCS_DW11_CH1) +#define _VLV_PCS01_DW11_CH0 0x022c +#define _VLV_PCS23_DW11_CH0 0x042c +#define _VLV_PCS01_DW11_CH1 0x262c +#define _VLV_PCS23_DW11_CH1 0x282c +#define VLV_PCS01_DW11(ch) _PORT(ch, _VLV_PCS01_DW0_CH0, _VLV_PCS01_DW0_CH1) +#define VLV_PCS23_DW11(ch) _PORT(ch, _VLV_PCS23_DW0_CH0, _VLV_PCS23_DW0_CH1) + #define _VLV_PCS_DW12_CH0 0x8230 #define _VLV_PCS_DW12_CH1 0x8430 #define VLV_PCS_DW12(ch) _PORT(ch, _VLV_PCS_DW12_CH0, _VLV_PCS_DW12_CH1) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 22d6c3e..3fed8cb 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2731,6 +2731,15 @@ static void chv_pre_enable_dp(struct intel_encoder *encoder) mutex_lock(&dev_priv->dpio_lock); + /* allow hardware to manage TX FIFO reset source */ + val = vlv_dpio_read(dev_priv, pipe, VLV_PCS01_DW11(ch)); + val &= ~DPIO_LANEDESKEW_STRAP_OVRD; + vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW11(ch), val); + + val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW11(ch)); + val &= ~DPIO_LANEDESKEW_STRAP_OVRD; + vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW11(ch), val); + /* Deassert soft data lane reset*/ val = vlv_dpio_read(dev_priv, pipe, VLV_PCS01_DW1(ch)); val |= CHV_PCS_REQ_SOFTRESET_EN; diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 386932a..9cbb7e4 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -1400,6 +1400,15 @@ static void chv_hdmi_pre_enable(struct intel_encoder *encoder) mutex_lock(&dev_priv->dpio_lock); + /* allow hardware to manage TX FIFO reset source */ + val = vlv_dpio_read(dev_priv, pipe, VLV_PCS01_DW11(ch)); + val &= ~DPIO_LANEDESKEW_STRAP_OVRD; + vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW11(ch), val); + + val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW11(ch)); + val &= ~DPIO_LANEDESKEW_STRAP_OVRD; + vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW11(ch), val); + /* Deassert soft data lane reset*/ val = vlv_dpio_read(dev_priv, pipe, VLV_PCS01_DW1(ch)); val |= CHV_PCS_REQ_SOFTRESET_EN; -- cgit v0.10.2 From 2aeb7d3a4d425be7e9185e79dd745918f7f72552 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:43 +0200 Subject: drm/i915: s/pm._irqs_disabled/pm.irqs_enabled/ Double negations just parse harder. Also this allows us to ditch some init code since clearing to 0 dtrt. Also ditch the assignment in intel_pm_setup, that's not redundant since we do the assignement now while setting up interrupts. While at it do engage in a bit of OCD and wrap up the few lines of setup/teardown code into little helper functions: intel_irq_fini for cleanup and intel_irq_init_hw for hw setup. v2: Use _install/_uninstall for the new wrapper function names as Paulo suggested. Signed-off-by: Daniel Vetter Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index d58c711..515fa6d 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1338,14 +1338,7 @@ static int i915_load_modeset_init(struct drm_device *dev) intel_power_domains_init_hw(dev_priv); - /* - * We enable some interrupt sources in our postinstall hooks, so mark - * interrupts as enabled _before_ actually enabling them to avoid - * special cases in our ordering checks. - */ - dev_priv->pm._irqs_disabled = false; - - ret = drm_irq_install(dev, dev->pdev->irq); + ret = intel_irq_install(dev_priv); if (ret) goto cleanup_gem_stolen; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 5dbd5c9..b4c4f6f 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1424,7 +1424,7 @@ struct ilk_wm_values { */ struct i915_runtime_pm { bool suspended; - bool _irqs_disabled; + bool irqs_enabled; }; enum intel_pipe_crc_source { @@ -2311,6 +2311,8 @@ void gen6_set_pm_mask(struct drm_i915_private *dev_priv, u32 pm_iir, int new_delay); extern void intel_irq_init(struct drm_device *dev); extern void intel_hpd_init(struct drm_device *dev); +int intel_irq_install(struct drm_i915_private *dev_priv); +void intel_irq_uninstall(struct drm_i915_private *dev_priv); extern void intel_uncore_sanitize(struct drm_device *dev); extern void intel_uncore_early_sanitize(struct drm_device *dev, diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index eef74ef..c07f576 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -4675,9 +4675,6 @@ void intel_irq_init(struct drm_device *dev) pm_qos_add_request(&dev_priv->pm_qos, PM_QOS_CPU_DMA_LATENCY, PM_QOS_DEFAULT_VALUE); - /* Haven't installed the IRQ handler yet */ - dev_priv->pm._irqs_disabled = true; - if (IS_GEN2(dev)) { dev->max_vblank_count = 0; dev->driver->get_vblank_counter = i8xx_get_vblank_counter; @@ -4786,13 +4783,32 @@ void intel_hpd_init(struct drm_device *dev) spin_unlock_irq(&dev_priv->irq_lock); } +int intel_irq_install(struct drm_i915_private *dev_priv) +{ + /* + * We enable some interrupt sources in our postinstall hooks, so mark + * interrupts as enabled _before_ actually enabling them to avoid + * special cases in our ordering checks. + */ + dev_priv->pm.irqs_enabled = true; + + return drm_irq_install(dev_priv->dev, dev_priv->dev->pdev->irq); +} + +void intel_irq_uninstall(struct drm_i915_private *dev_priv) +{ + drm_irq_uninstall(dev_priv->dev); + intel_hpd_cancel_work(dev_priv); + dev_priv->pm.irqs_enabled = false; +} + /* Disable interrupts so we can allow runtime PM. */ void intel_runtime_pm_disable_interrupts(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; dev->driver->irq_uninstall(dev); - dev_priv->pm._irqs_disabled = true; + dev_priv->pm.irqs_enabled = false; } /* Restore interrupts so we can recover from runtime PM. */ @@ -4800,7 +4816,7 @@ void intel_runtime_pm_restore_interrupts(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - dev_priv->pm._irqs_disabled = false; + dev_priv->pm.irqs_enabled = true; dev->driver->irq_preinstall(dev); dev->driver->irq_postinstall(dev); } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 7877b16..6fc77a1 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -13372,9 +13372,7 @@ void intel_modeset_cleanup(struct drm_device *dev) * Too much stuff here (turning of rps, connectors, ...) would * experience fancy races otherwise. */ - drm_irq_uninstall(dev); - intel_hpd_cancel_work(dev_priv); - dev_priv->pm._irqs_disabled = true; + intel_irq_uninstall(dev_priv); /* * Due to the hpd irq storm handling the hotplug work can re-arm the diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 1deb5d8..2d029d8 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -772,7 +772,7 @@ static inline bool intel_irqs_enabled(struct drm_i915_private *dev_priv) * We only use drm_irq_uninstall() at unload and VT switch, so * this is the only thing we need to check. */ - return !dev_priv->pm._irqs_disabled; + return dev_priv->pm.irqs_enabled; } int intel_get_crtc_scanline(struct intel_crtc *crtc); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 95006e0..7910805 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6531,5 +6531,4 @@ void intel_pm_setup(struct drm_device *dev) intel_gen6_powersave_work); dev_priv->pm.suspended = false; - dev_priv->pm._irqs_disabled = false; } -- cgit v0.10.2 From b963291cf9af882b2f30693fa8459afbdb0ed288 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:44 +0200 Subject: drm/i915: Use dev_priv instead of dev in irq setup functions It's the new world order! Not going full monty on these here and rolling this out throughout the subsequent call chains since this is just for the kerneldoc. Later on we can go more crazy, especially once we've embedded drm_device correctly. v2: Also frob the runtime_pm functions ... Signed-off-by: Daniel Vetter Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 515fa6d..85d14e1 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1363,7 +1363,7 @@ static int i915_load_modeset_init(struct drm_device *dev) goto cleanup_gem; /* Only enable hotplug handling once the fbdev is fully set up. */ - intel_hpd_init(dev); + intel_hpd_init(dev_priv); /* * Some ports require correctly set-up hpd registers for detection to @@ -1733,7 +1733,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) goto out_freewq; } - intel_irq_init(dev); + intel_irq_init(dev_priv); intel_uncore_sanitize(dev); /* Try to make sure MCHBAR is enabled before poking at it */ diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 66489fa..bd7978c 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -599,7 +599,7 @@ static int i915_drm_freeze(struct drm_device *dev) flush_delayed_work(&dev_priv->rps.delayed_resume_work); - intel_runtime_pm_disable_interrupts(dev); + intel_runtime_pm_disable_interrupts(dev_priv); intel_hpd_cancel_work(dev_priv); intel_suspend_encoders(dev_priv); @@ -705,7 +705,7 @@ static int __i915_drm_thaw(struct drm_device *dev, bool restore_gtt_mappings) mutex_unlock(&dev->struct_mutex); /* We need working interrupts for modeset enabling ... */ - intel_runtime_pm_restore_interrupts(dev); + intel_runtime_pm_enable_interrupts(dev_priv); intel_modeset_init_hw(dev); @@ -727,7 +727,7 @@ static int __i915_drm_thaw(struct drm_device *dev, bool restore_gtt_mappings) * bother with the tiny race here where we might loose hotplug * notifications. * */ - intel_hpd_init(dev); + intel_hpd_init(dev_priv); /* Config may have changed between suspend and resume */ drm_helper_hpd_irq_event(dev); } @@ -1473,12 +1473,12 @@ static int intel_runtime_suspend(struct device *device) * intel_mark_idle(). */ cancel_work_sync(&dev_priv->rps.work); - intel_runtime_pm_disable_interrupts(dev); + intel_runtime_pm_disable_interrupts(dev_priv); ret = intel_suspend_complete(dev_priv); if (ret) { DRM_ERROR("Runtime suspend failed, disabling it (%d)\n", ret); - intel_runtime_pm_restore_interrupts(dev); + intel_runtime_pm_enable_interrupts(dev_priv); return ret; } @@ -1538,7 +1538,7 @@ static int intel_runtime_resume(struct device *device) i915_gem_init_swizzling(dev); gen6_update_ring_freq(dev); - intel_runtime_pm_restore_interrupts(dev); + intel_runtime_pm_enable_interrupts(dev_priv); intel_reset_gt_powersave(dev); if (ret) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index b4c4f6f..9cc1329 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2309,8 +2309,8 @@ void i915_handle_error(struct drm_device *dev, bool wedged, void gen6_set_pm_mask(struct drm_i915_private *dev_priv, u32 pm_iir, int new_delay); -extern void intel_irq_init(struct drm_device *dev); -extern void intel_hpd_init(struct drm_device *dev); +extern void intel_irq_init(struct drm_i915_private *dev_priv); +extern void intel_hpd_init(struct drm_i915_private *dev_priv); int intel_irq_install(struct drm_i915_private *dev_priv); void intel_irq_uninstall(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index c07f576..97340a9 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -4650,9 +4650,9 @@ static void intel_hpd_irq_reenable_work(struct work_struct *work) intel_runtime_pm_put(dev_priv); } -void intel_irq_init(struct drm_device *dev) +void intel_irq_init(struct drm_i915_private *dev_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_device *dev = dev_priv->dev; INIT_WORK(&dev_priv->hotplug_work, i915_hotplug_work_func); INIT_WORK(&dev_priv->dig_port_work, i915_digport_work_func); @@ -4661,7 +4661,7 @@ void intel_irq_init(struct drm_device *dev) INIT_WORK(&dev_priv->l3_parity.error_work, ivybridge_parity_work); /* Let's track the enabled rps events */ - if (IS_VALLEYVIEW(dev) && !IS_CHERRYVIEW(dev)) + if (IS_VALLEYVIEW(dev_priv) && !IS_CHERRYVIEW(dev_priv)) /* WaGsvRC0ResidencyMethod:vlv */ dev_priv->pm_rps_events = GEN6_PM_RP_UP_EI_EXPIRED; else @@ -4675,10 +4675,10 @@ void intel_irq_init(struct drm_device *dev) pm_qos_add_request(&dev_priv->pm_qos, PM_QOS_CPU_DMA_LATENCY, PM_QOS_DEFAULT_VALUE); - if (IS_GEN2(dev)) { + if (IS_GEN2(dev_priv)) { dev->max_vblank_count = 0; dev->driver->get_vblank_counter = i8xx_get_vblank_counter; - } else if (IS_G4X(dev) || INTEL_INFO(dev)->gen >= 5) { + } else if (IS_G4X(dev_priv) || INTEL_INFO(dev_priv)->gen >= 5) { dev->max_vblank_count = 0xffffffff; /* full 32 bit counter */ dev->driver->get_vblank_counter = gm45_get_vblank_counter; } else { @@ -4691,7 +4691,7 @@ void intel_irq_init(struct drm_device *dev) * Gen2 doesn't have a hardware frame counter and so depends on * vblank interrupts to produce sane vblank seuquence numbers. */ - if (!IS_GEN2(dev)) + if (!IS_GEN2(dev_priv)) dev->vblank_disable_immediate = true; if (drm_core_check_feature(dev, DRIVER_MODESET)) { @@ -4699,7 +4699,7 @@ void intel_irq_init(struct drm_device *dev) dev->driver->get_scanout_position = i915_get_crtc_scanoutpos; } - if (IS_CHERRYVIEW(dev)) { + if (IS_CHERRYVIEW(dev_priv)) { dev->driver->irq_handler = cherryview_irq_handler; dev->driver->irq_preinstall = cherryview_irq_preinstall; dev->driver->irq_postinstall = cherryview_irq_postinstall; @@ -4707,7 +4707,7 @@ void intel_irq_init(struct drm_device *dev) dev->driver->enable_vblank = valleyview_enable_vblank; dev->driver->disable_vblank = valleyview_disable_vblank; dev_priv->display.hpd_irq_setup = i915_hpd_irq_setup; - } else if (IS_VALLEYVIEW(dev)) { + } else if (IS_VALLEYVIEW(dev_priv)) { dev->driver->irq_handler = valleyview_irq_handler; dev->driver->irq_preinstall = valleyview_irq_preinstall; dev->driver->irq_postinstall = valleyview_irq_postinstall; @@ -4715,7 +4715,7 @@ void intel_irq_init(struct drm_device *dev) dev->driver->enable_vblank = valleyview_enable_vblank; dev->driver->disable_vblank = valleyview_disable_vblank; dev_priv->display.hpd_irq_setup = i915_hpd_irq_setup; - } else if (INTEL_INFO(dev)->gen >= 8) { + } else if (INTEL_INFO(dev_priv)->gen >= 8) { dev->driver->irq_handler = gen8_irq_handler; dev->driver->irq_preinstall = gen8_irq_reset; dev->driver->irq_postinstall = gen8_irq_postinstall; @@ -4732,12 +4732,12 @@ void intel_irq_init(struct drm_device *dev) dev->driver->disable_vblank = ironlake_disable_vblank; dev_priv->display.hpd_irq_setup = ibx_hpd_irq_setup; } else { - if (INTEL_INFO(dev)->gen == 2) { + if (INTEL_INFO(dev_priv)->gen == 2) { dev->driver->irq_preinstall = i8xx_irq_preinstall; dev->driver->irq_postinstall = i8xx_irq_postinstall; dev->driver->irq_handler = i8xx_irq_handler; dev->driver->irq_uninstall = i8xx_irq_uninstall; - } else if (INTEL_INFO(dev)->gen == 3) { + } else if (INTEL_INFO(dev_priv)->gen == 3) { dev->driver->irq_preinstall = i915_irq_preinstall; dev->driver->irq_postinstall = i915_irq_postinstall; dev->driver->irq_uninstall = i915_irq_uninstall; @@ -4755,9 +4755,9 @@ void intel_irq_init(struct drm_device *dev) } } -void intel_hpd_init(struct drm_device *dev) +void intel_hpd_init(struct drm_i915_private *dev_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_device *dev = dev_priv->dev; struct drm_mode_config *mode_config = &dev->mode_config; struct drm_connector *connector; int i; @@ -4803,20 +4803,16 @@ void intel_irq_uninstall(struct drm_i915_private *dev_priv) } /* Disable interrupts so we can allow runtime PM. */ -void intel_runtime_pm_disable_interrupts(struct drm_device *dev) +void intel_runtime_pm_disable_interrupts(struct drm_i915_private *dev_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - - dev->driver->irq_uninstall(dev); + dev_priv->dev->driver->irq_uninstall(dev_priv->dev); dev_priv->pm.irqs_enabled = false; } /* Restore interrupts so we can recover from runtime PM. */ -void intel_runtime_pm_restore_interrupts(struct drm_device *dev) +void intel_runtime_pm_enable_interrupts(struct drm_i915_private *dev_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - dev_priv->pm.irqs_enabled = true; - dev->driver->irq_preinstall(dev); - dev->driver->irq_postinstall(dev); + dev_priv->dev->driver->irq_preinstall(dev_priv->dev); + dev_priv->dev->driver->irq_postinstall(dev_priv->dev); } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 2d029d8..ffbf38d 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -764,8 +764,8 @@ void gen6_enable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask); void gen6_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask); void gen8_enable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask); void gen8_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask); -void intel_runtime_pm_disable_interrupts(struct drm_device *dev); -void intel_runtime_pm_restore_interrupts(struct drm_device *dev); +void intel_runtime_pm_disable_interrupts(struct drm_i915_private *dev_priv); +void intel_runtime_pm_enable_interrupts(struct drm_i915_private *dev_priv); static inline bool intel_irqs_enabled(struct drm_i915_private *dev_priv) { /* diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index abefc98..36749b9 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -377,7 +377,7 @@ static void vlv_display_power_well_enable(struct drm_i915_private *dev_priv, if (dev_priv->power_domains.initializing) return; - intel_hpd_init(dev_priv->dev); + intel_hpd_init(dev_priv); i915_redisable_vga_power_on(dev_priv->dev); } -- cgit v0.10.2 From fca52a5565fbf4abc4ee4fca81842dc1f6ecdce8 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:45 +0200 Subject: drm/i915: kerneldoc for interrupt enable/disable functions Just start with the basics for now. Since there's a lot of different functionality in i915_irq.c I've decided to split it into different sections and pull in just the relevant functions. Splitting into different files looks like a lot more work since the interrupt handlers do an awful lot of reuse all over. v2: Rebase onto changed function names. Signed-off-by: Daniel Vetter Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index 8627816..d7cfc98 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -3798,6 +3798,14 @@ int num_ioctls; !Pdrivers/gpu/drm/i915/intel_runtime_pm.c runtime pm !Idrivers/gpu/drm/i915/intel_runtime_pm.c + + Interrupt Handling +!Pdrivers/gpu/drm/i915/i915_irq.c interrupt handling +!Fdrivers/gpu/drm/i915/i915_irq.c intel_irq_init intel_irq_init_hw intel_hpd_init +!Fdrivers/gpu/drm/i915/i915_irq.c intel_irq_fini +!Fdrivers/gpu/drm/i915/i915_irq.c intel_runtime_pm_disable_interrupts +!Fdrivers/gpu/drm/i915/i915_irq.c intel_runtime_pm_enable_interrupts + Display Hardware Handling @@ -3951,5 +3959,6 @@ int num_ioctls; +!Cdrivers/gpu/drm/i915/i915_irq.c diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 97340a9..737b239 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -37,6 +37,14 @@ #include "i915_trace.h" #include "intel_drv.h" +/** + * DOC: interrupt handling + * + * These functions provide the basic support for enabling and disabling the + * interrupt handling support. There's a lot more functionality in i915_irq.c + * and related files, but that will be described in separate chapters. + */ + static const u32 hpd_ibx[] = { [HPD_CRT] = SDE_CRT_HOTPLUG, [HPD_SDVO_B] = SDE_SDVOB_HOTPLUG, @@ -4650,6 +4658,13 @@ static void intel_hpd_irq_reenable_work(struct work_struct *work) intel_runtime_pm_put(dev_priv); } +/** + * intel_irq_init - initializes irq support + * @dev_priv: i915 device instance + * + * This function initializes all the irq support including work items, timers + * and all the vtables. It does not setup the interrupt itself though. + */ void intel_irq_init(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; @@ -4755,6 +4770,18 @@ void intel_irq_init(struct drm_i915_private *dev_priv) } } +/** + * intel_hpd_init - initializes and enables hpd support + * @dev_priv: i915 device instance + * + * This function enables the hotplug support. It requires that interrupts have + * already been enabled with intel_irq_init_hw(). From this point on hotplug and + * poll request can run concurrently to other code, so locking rules must be + * obeyed. + * + * This is a separate step from interrupt enabling to simplify the locking rules + * in the driver load and resume code. + */ void intel_hpd_init(struct drm_i915_private *dev_priv) { struct drm_device *dev = dev_priv->dev; @@ -4783,6 +4810,17 @@ void intel_hpd_init(struct drm_i915_private *dev_priv) spin_unlock_irq(&dev_priv->irq_lock); } +/** + * intel_irq_install - enables the hardware interrupt + * @dev_priv: i915 device instance + * + * This function enables the hardware interrupt handling, but leaves the hotplug + * handling still disabled. It is called after intel_irq_init(). + * + * In the driver load and resume code we need working interrupts in a few places + * but don't want to deal with the hassle of concurrent probe and hotplug + * workers. Hence the split into this two-stage approach. + */ int intel_irq_install(struct drm_i915_private *dev_priv) { /* @@ -4795,6 +4833,13 @@ int intel_irq_install(struct drm_i915_private *dev_priv) return drm_irq_install(dev_priv->dev, dev_priv->dev->pdev->irq); } +/** + * intel_irq_uninstall - finilizes all irq handling + * @dev_priv: i915 device instance + * + * This stops interrupt and hotplug handling and unregisters and frees all + * resources acquired in the init functions. + */ void intel_irq_uninstall(struct drm_i915_private *dev_priv) { drm_irq_uninstall(dev_priv->dev); @@ -4802,14 +4847,26 @@ void intel_irq_uninstall(struct drm_i915_private *dev_priv) dev_priv->pm.irqs_enabled = false; } -/* Disable interrupts so we can allow runtime PM. */ +/** + * intel_runtime_pm_disable_interrupts - runtime interrupt disabling + * @dev_priv: i915 device instance + * + * This function is used to disable interrupts at runtime, both in the runtime + * pm and the system suspend/resume code. + */ void intel_runtime_pm_disable_interrupts(struct drm_i915_private *dev_priv) { dev_priv->dev->driver->irq_uninstall(dev_priv->dev); dev_priv->pm.irqs_enabled = false; } -/* Restore interrupts so we can recover from runtime PM. */ +/** + * intel_runtime_pm_enable_interrupts - runtime interrupt enabling + * @dev_priv: i915 device instance + * + * This function is used to enable interrupts at runtime, both in the runtime + * pm and the system suspend/resume code. + */ void intel_runtime_pm_enable_interrupts(struct drm_i915_private *dev_priv) { dev_priv->pm.irqs_enabled = true; -- cgit v0.10.2 From 60a5372777b05c78ec25f1bc1883cb59b2418255 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 3 Oct 2014 10:29:51 +0100 Subject: drm/i915: Remove the duplicated logic between the two shrink phases We can use the same logic to walk the different bound/unbound lists during shrinker (as the unbound list is a degenerate case of the bound list), slightly compacting the code. Signed-off-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index e05e006..2719c25 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1945,7 +1945,14 @@ unsigned long i915_gem_shrink(struct drm_i915_private *dev_priv, long target, unsigned flags) { - const bool purgeable_only = flags & I915_SHRINK_PURGEABLE; + const struct { + struct list_head *list; + unsigned int bit; + } phases[] = { + { &dev_priv->mm.unbound_list, I915_SHRINK_UNBOUND }, + { &dev_priv->mm.bound_list, I915_SHRINK_BOUND }, + { NULL, 0 }, + }, *phase; unsigned long count = 0; /* @@ -1967,48 +1974,30 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, * dev->struct_mutex and so we won't ever be able to observe an * object on the bound_list with a reference count equals 0. */ - if (flags & I915_SHRINK_UNBOUND) { + for (phase = phases; phase->list; phase++) { struct list_head still_in_list; - INIT_LIST_HEAD(&still_in_list); - while (count < target && !list_empty(&dev_priv->mm.unbound_list)) { - struct drm_i915_gem_object *obj; - - obj = list_first_entry(&dev_priv->mm.unbound_list, - typeof(*obj), global_list); - list_move_tail(&obj->global_list, &still_in_list); - - if (!i915_gem_object_is_purgeable(obj) && purgeable_only) - continue; - - drm_gem_object_reference(&obj->base); - - if (i915_gem_object_put_pages(obj) == 0) - count += obj->base.size >> PAGE_SHIFT; - - drm_gem_object_unreference(&obj->base); - } - list_splice(&still_in_list, &dev_priv->mm.unbound_list); - } - - if (flags & I915_SHRINK_BOUND) { - struct list_head still_in_list; + if ((flags & phase->bit) == 0) + continue; INIT_LIST_HEAD(&still_in_list); - while (count < target && !list_empty(&dev_priv->mm.bound_list)) { + while (count < target && !list_empty(phase->list)) { struct drm_i915_gem_object *obj; struct i915_vma *vma, *v; - obj = list_first_entry(&dev_priv->mm.bound_list, + obj = list_first_entry(phase->list, typeof(*obj), global_list); list_move_tail(&obj->global_list, &still_in_list); - if (!i915_gem_object_is_purgeable(obj) && purgeable_only) + if (flags & I915_SHRINK_PURGEABLE && + !i915_gem_object_is_purgeable(obj)) continue; drm_gem_object_reference(&obj->base); - list_for_each_entry_safe(vma, v, &obj->vma_list, vma_link) + /* For the unbound phase, this should be a no-op! */ + list_for_each_entry_safe(vma, v, + &obj->vma_list, vma_link) if (i915_vma_unbind(vma)) break; @@ -2017,7 +2006,7 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, drm_gem_object_unreference(&obj->base); } - list_splice(&still_in_list, &dev_priv->mm.bound_list); + list_splice(&still_in_list, phase->list); } return count; -- cgit v0.10.2 From b727735732d5b2fde0a88911210215e03e190d11 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 3 Oct 2014 17:38:34 +0200 Subject: drm/i915: Update DRIVER_DATE to 20141003 Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 9cc1329..821ba26 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -55,7 +55,7 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20140919" +#define DRIVER_DATE "20141003" enum pipe { INVALID_PIPE = -1, -- cgit v0.10.2 From bd008e5b2953186fc0c6633a885ade95e7043800 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Oct 2014 14:13:51 +0100 Subject: drm: Implement O_NONBLOCK support on /dev/dri/cardN The implmentation is simple in the extreme: we only want to wait for events if the device was opened in blocking mode, otherwise we grab what is available and report an error if there was none. Signed-off-by: Chris Wilson Cc: dri-devel@lists.freedesktop.org Reviewed-by: Jesse Barnes Testcase: igt/kms_flip/nonblocing_read Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c index ed7bc68..91e1105 100644 --- a/drivers/gpu/drm/drm_fops.c +++ b/drivers/gpu/drm/drm_fops.c @@ -515,10 +515,12 @@ ssize_t drm_read(struct file *filp, char __user *buffer, size_t total; ssize_t ret; - ret = wait_event_interruptible(file_priv->event_wait, - !list_empty(&file_priv->event_list)); - if (ret < 0) - return ret; + if ((filp->f_flags & O_NONBLOCK) == 0) { + ret = wait_event_interruptible(file_priv->event_wait, + !list_empty(&file_priv->event_list)); + if (ret < 0) + return ret; + } total = 0; while (drm_dequeue_event(file_priv, total, count, &e)) { @@ -532,7 +534,7 @@ ssize_t drm_read(struct file *filp, char __user *buffer, e->destroy(e); } - return total; + return total ?: -EAGAIN; } EXPORT_SYMBOL(drm_read); -- cgit v0.10.2 From a1f1a79c51fd493887bb66d932ee66a23f8b1527 Mon Sep 17 00:00:00 2001 From: Joe Perches Date: Sat, 11 Oct 2014 22:08:57 -0700 Subject: drm: drm_err: Remove unnecessary __func__ argument Removing the unnecessary drm_err __func__ argument by using the equivalent %pf and __builtin_return_address(0) makes the code smaller for every use of the DRM_ERROR macro. For instance: (allmodconfig) $ size drivers/gpu/drm/i915/i915.o* text data bss dec hex filename 922447 193257 296736 1412440 158d58 drivers/gpu/drm/i915/i915.o.new 928111 193257 296736 1418104 15a378 drivers/gpu/drm/i915/i915.o.old Signed-off-by: Joe Perches Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c index bc3da32..2e5c7d9 100644 --- a/drivers/gpu/drm/drm_drv.c +++ b/drivers/gpu/drm/drm_drv.c @@ -56,7 +56,7 @@ static struct idr drm_minors_idr; struct class *drm_class; static struct dentry *drm_debugfs_root; -void drm_err(const char *func, const char *format, ...) +void drm_err(const char *format, ...) { struct va_format vaf; va_list args; @@ -66,7 +66,8 @@ void drm_err(const char *func, const char *format, ...) vaf.fmt = format; vaf.va = &args; - printk(KERN_ERR "[" DRM_NAME ":%s] *ERROR* %pV", func, &vaf); + printk(KERN_ERR "[" DRM_NAME ":%pf] *ERROR* %pV", + __builtin_return_address(0), &vaf); va_end(args); } diff --git a/include/drm/drmP.h b/include/drm/drmP.h index 53ed876..d7ebd7b 100644 --- a/include/drm/drmP.h +++ b/include/drm/drmP.h @@ -125,8 +125,8 @@ struct dma_buf_attachment; extern __printf(2, 3) void drm_ut_debug_printk(const char *function_name, const char *format, ...); -extern __printf(2, 3) -void drm_err(const char *func, const char *format, ...); +extern __printf(1, 2) +void drm_err(const char *format, ...); /***********************************************************************/ /** \name DRM template customization defaults */ @@ -155,7 +155,7 @@ void drm_err(const char *func, const char *format, ...); * \param arg arguments */ #define DRM_ERROR(fmt, ...) \ - drm_err(__func__, fmt, ##__VA_ARGS__) + drm_err(fmt, ##__VA_ARGS__) /** * Rate limited error output. Like DRM_ERROR() but won't flood the log. @@ -170,7 +170,7 @@ void drm_err(const char *func, const char *format, ...); DEFAULT_RATELIMIT_BURST); \ \ if (__ratelimit(&_rs)) \ - drm_err(__func__, fmt, ##__VA_ARGS__); \ + drm_err(fmt, ##__VA_ARGS__); \ }) #define DRM_INFO(fmt, ...) \ -- cgit v0.10.2 From d87af4d10558fedf636e3df1b10dd2954f9e2a78 Mon Sep 17 00:00:00 2001 From: Rickard Strandqvist Date: Sun, 12 Oct 2014 00:02:32 +0200 Subject: gpu: drm: drm_dp_mst_topology.c: Fix improper use of strncat Fixed wrong usage of strncat, switched to strlcpy. While sending the string size to function to reduce the potential for misuse in future. Signed-off-by: Rickard Strandqvist Reviewed-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c index b3adf14..094af1d 100644 --- a/drivers/gpu/drm/drm_dp_mst_topology.c +++ b/drivers/gpu/drm/drm_dp_mst_topology.c @@ -995,19 +995,20 @@ static void drm_dp_check_port_guid(struct drm_dp_mst_branch *mstb, static void build_mst_prop_path(struct drm_dp_mst_port *port, struct drm_dp_mst_branch *mstb, - char *proppath) + char *proppath, + size_t proppath_size) { int i; char temp[8]; - snprintf(proppath, 255, "mst:%d", mstb->mgr->conn_base_id); + snprintf(proppath, proppath_size, "mst:%d", mstb->mgr->conn_base_id); for (i = 0; i < (mstb->lct - 1); i++) { int shift = (i % 2) ? 0 : 4; int port_num = mstb->rad[i / 2] >> shift; - snprintf(temp, 8, "-%d", port_num); - strncat(proppath, temp, 255); + snprintf(temp, sizeof(temp), "-%d", port_num); + strlcat(proppath, temp, proppath_size); } - snprintf(temp, 8, "-%d", port->port_num); - strncat(proppath, temp, 255); + snprintf(temp, sizeof(temp), "-%d", port->port_num); + strlcat(proppath, temp, proppath_size); } static void drm_dp_add_port(struct drm_dp_mst_branch *mstb, @@ -1078,7 +1079,7 @@ static void drm_dp_add_port(struct drm_dp_mst_branch *mstb, if (created && !port->input) { char proppath[255]; - build_mst_prop_path(port, mstb, proppath); + build_mst_prop_path(port, mstb, proppath, sizeof(proppath)); port->connector = (*mstb->mgr->cbs->add_connector)(mstb->mgr, port, proppath); } -- cgit v0.10.2 From 32197aab0425dbecc38462a91bc5c8acf70b2036 Mon Sep 17 00:00:00 2001 From: Masanari Iida Date: Mon, 20 Oct 2014 23:53:13 +0900 Subject: gpu:drm: Fix typo in Documentation/DocBook/drm.xml This patch fix spelling typos found in drm.xml. It is because the file is generated from comments in source codes, I have to fix the typos within source files. Signed-off-by: Masanari Iida Acked-by: Randy Dunlap Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index e79c8d3..7f94989 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -3611,7 +3611,7 @@ static struct drm_property *property_create_range(struct drm_device *dev, * object with drm_object_attach_property. The returned property object must be * freed with drm_property_destroy. * - * Userspace is allowed to set any interger value in the (min, max) range + * Userspace is allowed to set any integer value in the (min, max) range * inclusive. * * Returns: diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c index 5ef03c2..3e6b582 100644 --- a/drivers/gpu/drm/drm_irq.c +++ b/drivers/gpu/drm/drm_irq.c @@ -1190,7 +1190,7 @@ EXPORT_SYMBOL(drm_crtc_vblank_off); * * This functions restores the vblank interrupt state captured with * drm_vblank_off() again. Note that calls to drm_vblank_on() and - * drm_vblank_off() can be unbalanced and so can also be unconditionaly called + * drm_vblank_off() can be unbalanced and so can also be unconditionally called * in driver load code to reflect the current hardware state of the crtc. * * This is the legacy version of drm_crtc_vblank_on(). @@ -1237,7 +1237,7 @@ EXPORT_SYMBOL(drm_vblank_on); * * This functions restores the vblank interrupt state captured with * drm_vblank_off() again. Note that calls to drm_vblank_on() and - * drm_vblank_off() can be unbalanced and so can also be unconditionaly called + * drm_vblank_off() can be unbalanced and so can also be unconditionally called * in driver load code to reflect the current hardware state of the crtc. * * This is the native kms version of drm_vblank_on(). diff --git a/drivers/gpu/drm/drm_modes.c b/drivers/gpu/drm/drm_modes.c index d1b7d20..6d8b941 100644 --- a/drivers/gpu/drm/drm_modes.c +++ b/drivers/gpu/drm/drm_modes.c @@ -914,7 +914,7 @@ EXPORT_SYMBOL(drm_mode_equal_no_clocks_no_stereo); * * This function is a helper which can be used to validate modes against size * limitations of the DRM device/connector. If a mode is too big its status - * memeber is updated with the appropriate validation failure code. The list + * member is updated with the appropriate validation failure code. The list * itself is not changed. */ void drm_mode_validate_size(struct drm_device *dev, diff --git a/drivers/gpu/drm/drm_prime.c b/drivers/gpu/drm/drm_prime.c index 78ca308..6872eca 100644 --- a/drivers/gpu/drm/drm_prime.c +++ b/drivers/gpu/drm/drm_prime.c @@ -328,7 +328,7 @@ static const struct dma_buf_ops drm_gem_prime_dmabuf_ops = { */ /** - * drm_gem_prime_export - helper library implemention of the export callback + * drm_gem_prime_export - helper library implementation of the export callback * @dev: drm_device to export from * @obj: GEM object to export * @flags: flags like DRM_CLOEXEC @@ -483,7 +483,7 @@ out_unlock: EXPORT_SYMBOL(drm_gem_prime_handle_to_fd); /** - * drm_gem_prime_import - helper library implemention of the import callback + * drm_gem_prime_import - helper library implementation of the import callback * @dev: drm_device to import into * @dma_buf: dma-buf object to import * diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c index c45856b..5506242 100644 --- a/drivers/gpu/drm/i915/i915_cmd_parser.c +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c @@ -73,7 +73,7 @@ * those commands required by the parser. This generally works because command * opcode ranges have standard command length encodings. So for commands that * the parser does not need to check, it can easily skip them. This is - * implementated via a per-ring length decoding vfunc. + * implemented via a per-ring length decoding vfunc. * * Unfortunately, there are a number of commands that do not follow the standard * length encoding for their opcode range, primarily amongst the MI_* commands. @@ -836,7 +836,7 @@ finish: * @ring: the ring in question * * Only certain platforms require software batch buffer command parsing, and - * only when enabled via module paramter. + * only when enabled via module parameter. * * Return: true if the ring requires software command parsing */ diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index b65bdfc..65821ba 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -664,7 +664,7 @@ enum punit_power_well { * need to be accessed during AUX communication, * * Generally the common lane corresponds to the pipe and - * the spline (PCS/TX) correponds to the port. + * the spline (PCS/TX) corresponds to the port. * * For dual channel PHY (VLV/CHV): * diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index bd1b28d..b7b9a09 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1639,7 +1639,7 @@ static uint32_t get_lr_context_size(struct intel_engine_cs *ring) * the creation is a deferred call: it's better to make sure first that we need to use * a given ring with the context. * - * Return: non-zero on eror. + * Return: non-zero on error. */ int intel_lr_context_deferred_create(struct intel_context *ctx, struct intel_engine_cs *ring) diff --git a/include/drm/drm_dp_mst_helper.h b/include/drm/drm_dp_mst_helper.h index 9b446ad..3ecfcc1 100644 --- a/include/drm/drm_dp_mst_helper.h +++ b/include/drm/drm_dp_mst_helper.h @@ -28,7 +28,7 @@ struct drm_dp_mst_branch; /** - * struct drm_dp_vcpi - Virtual Channel Payload Identifer + * struct drm_dp_vcpi - Virtual Channel Payload Identifier * @vcpi: Virtual channel ID. * @pbn: Payload Bandwidth Number for this channel * @aligned_pbn: PBN aligned with slot size -- cgit v0.10.2 From 209f5527b7ef06948d79f9dea636764db296db08 Mon Sep 17 00:00:00 2001 From: Chuck Ebbert Date: Wed, 8 Oct 2014 11:37:20 -0500 Subject: drm/crtc: Fix two typos MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fix: ioclt -> ioctl in comment wrong variable name in debug message Signed-off-by: Chuck Ebbert Reviewed-by: Ville Syrjälä [danvet: Frob manually generated patch to make it apply.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 7f94989..4be0189 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -2943,7 +2943,7 @@ EXPORT_SYMBOL(drm_mode_legacy_fb_format); * @file_priv: drm file for the ioctl call * * Add a new FB to the specified CRTC, given a user request. This is the - * original addfb ioclt which only supported RGB formats. + * original addfb ioctl which only supported RGB formats. * * Called by the user via ioctl. * @@ -3080,7 +3080,7 @@ static int framebuffer_check(const struct drm_mode_fb_cmd2 *r) num_planes = drm_format_num_planes(r->pixel_format); if (r->width == 0 || r->width % hsub) { - DRM_DEBUG_KMS("bad framebuffer width %u\n", r->height); + DRM_DEBUG_KMS("bad framebuffer width %u\n", r->width); return -EINVAL; } -- cgit v0.10.2 From 228f2cb32f0dbeef0b88dc97ea66a3c31b03de99 Mon Sep 17 00:00:00 2001 From: Chuck Ebbert Date: Wed, 8 Oct 2014 11:40:34 -0500 Subject: drm/crtc: Remove duplicated ioctl code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Make drm_mode_add_fb() call drm_mode_add_fb2() after converting its args to the new internal format, instead of duplicating code. Also picks up a lot more error checking, which the legacy modes should pass after being converted to the new format. Signed-off-by: Chuck Ebbert Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 4be0189..b4a6f08 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -2955,11 +2955,9 @@ int drm_mode_addfb(struct drm_device *dev, { struct drm_mode_fb_cmd *or = data; struct drm_mode_fb_cmd2 r = {}; - struct drm_mode_config *config = &dev->mode_config; - struct drm_framebuffer *fb; - int ret = 0; + int ret; - /* Use new struct with format internally */ + /* convert to new format and call new ioctl */ r.fb_id = or->fb_id; r.width = or->width; r.height = or->height; @@ -2967,26 +2965,11 @@ int drm_mode_addfb(struct drm_device *dev, r.pixel_format = drm_mode_legacy_fb_format(or->bpp, or->depth); r.handles[0] = or->handle; - if (!drm_core_check_feature(dev, DRIVER_MODESET)) - return -EINVAL; - - if ((config->min_width > r.width) || (r.width > config->max_width)) - return -EINVAL; - - if ((config->min_height > r.height) || (r.height > config->max_height)) - return -EINVAL; - - fb = dev->mode_config.funcs->fb_create(dev, file_priv, &r); - if (IS_ERR(fb)) { - DRM_DEBUG_KMS("could not create framebuffer\n"); - return PTR_ERR(fb); - } + ret = drm_mode_addfb2(dev, &r, file_priv); + if (ret) + return ret; - mutex_lock(&file_priv->fbs_lock); - or->fb_id = fb->base.id; - list_add(&fb->filp_head, &file_priv->fbs); - DRM_DEBUG_KMS("[FB:%d]\n", fb->base.id); - mutex_unlock(&file_priv->fbs_lock); + or->fb_id = r.fb_id; return ret; } -- cgit v0.10.2 From e45708976aea3a3c3aa959e3cc2b703fefbd758a Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 22 Oct 2014 11:16:36 +0200 Subject: drm/dp-helper: Move the legacy helpers to gma500 Except for gma500 all drivers are converted to the new style helpers, which have much better abstraction of the underlying hw protocols and already much more helper functions (including the entire mst library) on top of them. Since no one seems to work on converting gma500 let's just move the code away so that new drivers don't end up accidentally using this. Cc: Patrik Jakobsson Reviewed-by: Patrik Jakobsson Reviewed-by: Alan Cox [danvet: Add __deprecated as requested by Alan. Also add a short FIXME comment and drop the EXPORT_SYMBOL which is no longer needed.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_dp_helper.c b/drivers/gpu/drm/drm_dp_helper.c index 08e33b8..c088bad 100644 --- a/drivers/gpu/drm/drm_dp_helper.c +++ b/drivers/gpu/drm/drm_dp_helper.c @@ -39,198 +39,6 @@ * blocks, ... */ -/* Run a single AUX_CH I2C transaction, writing/reading data as necessary */ -static int -i2c_algo_dp_aux_transaction(struct i2c_adapter *adapter, int mode, - uint8_t write_byte, uint8_t *read_byte) -{ - struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; - int ret; - - ret = (*algo_data->aux_ch)(adapter, mode, - write_byte, read_byte); - return ret; -} - -/* - * I2C over AUX CH - */ - -/* - * Send the address. If the I2C link is running, this 'restarts' - * the connection with the new address, this is used for doing - * a write followed by a read (as needed for DDC) - */ -static int -i2c_algo_dp_aux_address(struct i2c_adapter *adapter, u16 address, bool reading) -{ - struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; - int mode = MODE_I2C_START; - int ret; - - if (reading) - mode |= MODE_I2C_READ; - else - mode |= MODE_I2C_WRITE; - algo_data->address = address; - algo_data->running = true; - ret = i2c_algo_dp_aux_transaction(adapter, mode, 0, NULL); - return ret; -} - -/* - * Stop the I2C transaction. This closes out the link, sending - * a bare address packet with the MOT bit turned off - */ -static void -i2c_algo_dp_aux_stop(struct i2c_adapter *adapter, bool reading) -{ - struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; - int mode = MODE_I2C_STOP; - - if (reading) - mode |= MODE_I2C_READ; - else - mode |= MODE_I2C_WRITE; - if (algo_data->running) { - (void) i2c_algo_dp_aux_transaction(adapter, mode, 0, NULL); - algo_data->running = false; - } -} - -/* - * Write a single byte to the current I2C address, the - * the I2C link must be running or this returns -EIO - */ -static int -i2c_algo_dp_aux_put_byte(struct i2c_adapter *adapter, u8 byte) -{ - struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; - int ret; - - if (!algo_data->running) - return -EIO; - - ret = i2c_algo_dp_aux_transaction(adapter, MODE_I2C_WRITE, byte, NULL); - return ret; -} - -/* - * Read a single byte from the current I2C address, the - * I2C link must be running or this returns -EIO - */ -static int -i2c_algo_dp_aux_get_byte(struct i2c_adapter *adapter, u8 *byte_ret) -{ - struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; - int ret; - - if (!algo_data->running) - return -EIO; - - ret = i2c_algo_dp_aux_transaction(adapter, MODE_I2C_READ, 0, byte_ret); - return ret; -} - -static int -i2c_algo_dp_aux_xfer(struct i2c_adapter *adapter, - struct i2c_msg *msgs, - int num) -{ - int ret = 0; - bool reading = false; - int m; - int b; - - for (m = 0; m < num; m++) { - u16 len = msgs[m].len; - u8 *buf = msgs[m].buf; - reading = (msgs[m].flags & I2C_M_RD) != 0; - ret = i2c_algo_dp_aux_address(adapter, msgs[m].addr, reading); - if (ret < 0) - break; - if (reading) { - for (b = 0; b < len; b++) { - ret = i2c_algo_dp_aux_get_byte(adapter, &buf[b]); - if (ret < 0) - break; - } - } else { - for (b = 0; b < len; b++) { - ret = i2c_algo_dp_aux_put_byte(adapter, buf[b]); - if (ret < 0) - break; - } - } - if (ret < 0) - break; - } - if (ret >= 0) - ret = num; - i2c_algo_dp_aux_stop(adapter, reading); - DRM_DEBUG_KMS("dp_aux_xfer return %d\n", ret); - return ret; -} - -static u32 -i2c_algo_dp_aux_functionality(struct i2c_adapter *adapter) -{ - return I2C_FUNC_I2C | I2C_FUNC_SMBUS_EMUL | - I2C_FUNC_SMBUS_READ_BLOCK_DATA | - I2C_FUNC_SMBUS_BLOCK_PROC_CALL | - I2C_FUNC_10BIT_ADDR; -} - -static const struct i2c_algorithm i2c_dp_aux_algo = { - .master_xfer = i2c_algo_dp_aux_xfer, - .functionality = i2c_algo_dp_aux_functionality, -}; - -static void -i2c_dp_aux_reset_bus(struct i2c_adapter *adapter) -{ - (void) i2c_algo_dp_aux_address(adapter, 0, false); - (void) i2c_algo_dp_aux_stop(adapter, false); -} - -static int -i2c_dp_aux_prepare_bus(struct i2c_adapter *adapter) -{ - adapter->algo = &i2c_dp_aux_algo; - adapter->retries = 3; - i2c_dp_aux_reset_bus(adapter); - return 0; -} - -/** - * i2c_dp_aux_add_bus() - register an i2c adapter using the aux ch helper - * @adapter: i2c adapter to register - * - * This registers an i2c adapter that uses dp aux channel as it's underlaying - * transport. The driver needs to fill out the &i2c_algo_dp_aux_data structure - * and store it in the algo_data member of the @adapter argument. This will be - * used by the i2c over dp aux algorithm to drive the hardware. - * - * RETURNS: - * 0 on success, -ERRNO on failure. - * - * IMPORTANT: - * This interface is deprecated, please switch to the new dp aux helpers and - * drm_dp_aux_register(). - */ -int -i2c_dp_aux_add_bus(struct i2c_adapter *adapter) -{ - int error; - - error = i2c_dp_aux_prepare_bus(adapter); - if (error) - return error; - error = i2c_add_adapter(adapter); - return error; -} -EXPORT_SYMBOL(i2c_dp_aux_add_bus); - /* Helpers for DP link training */ static u8 dp_link_status(const u8 link_status[DP_LINK_STATUS_SIZE], int r) { diff --git a/drivers/gpu/drm/gma500/cdv_intel_dp.c b/drivers/gpu/drm/gma500/cdv_intel_dp.c index 9f158ea..0fafb8e 100644 --- a/drivers/gpu/drm/gma500/cdv_intel_dp.c +++ b/drivers/gpu/drm/gma500/cdv_intel_dp.c @@ -37,6 +37,201 @@ #include "gma_display.h" #include +/** + * struct i2c_algo_dp_aux_data - driver interface structure for i2c over dp + * aux algorithm + * @running: set by the algo indicating whether an i2c is ongoing or whether + * the i2c bus is quiescent + * @address: i2c target address for the currently ongoing transfer + * @aux_ch: driver callback to transfer a single byte of the i2c payload + */ +struct i2c_algo_dp_aux_data { + bool running; + u16 address; + int (*aux_ch) (struct i2c_adapter *adapter, + int mode, uint8_t write_byte, + uint8_t *read_byte); +}; + +/* Run a single AUX_CH I2C transaction, writing/reading data as necessary */ +static int +i2c_algo_dp_aux_transaction(struct i2c_adapter *adapter, int mode, + uint8_t write_byte, uint8_t *read_byte) +{ + struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; + int ret; + + ret = (*algo_data->aux_ch)(adapter, mode, + write_byte, read_byte); + return ret; +} + +/* + * I2C over AUX CH + */ + +/* + * Send the address. If the I2C link is running, this 'restarts' + * the connection with the new address, this is used for doing + * a write followed by a read (as needed for DDC) + */ +static int +i2c_algo_dp_aux_address(struct i2c_adapter *adapter, u16 address, bool reading) +{ + struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; + int mode = MODE_I2C_START; + int ret; + + if (reading) + mode |= MODE_I2C_READ; + else + mode |= MODE_I2C_WRITE; + algo_data->address = address; + algo_data->running = true; + ret = i2c_algo_dp_aux_transaction(adapter, mode, 0, NULL); + return ret; +} + +/* + * Stop the I2C transaction. This closes out the link, sending + * a bare address packet with the MOT bit turned off + */ +static void +i2c_algo_dp_aux_stop(struct i2c_adapter *adapter, bool reading) +{ + struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; + int mode = MODE_I2C_STOP; + + if (reading) + mode |= MODE_I2C_READ; + else + mode |= MODE_I2C_WRITE; + if (algo_data->running) { + (void) i2c_algo_dp_aux_transaction(adapter, mode, 0, NULL); + algo_data->running = false; + } +} + +/* + * Write a single byte to the current I2C address, the + * the I2C link must be running or this returns -EIO + */ +static int +i2c_algo_dp_aux_put_byte(struct i2c_adapter *adapter, u8 byte) +{ + struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; + int ret; + + if (!algo_data->running) + return -EIO; + + ret = i2c_algo_dp_aux_transaction(adapter, MODE_I2C_WRITE, byte, NULL); + return ret; +} + +/* + * Read a single byte from the current I2C address, the + * I2C link must be running or this returns -EIO + */ +static int +i2c_algo_dp_aux_get_byte(struct i2c_adapter *adapter, u8 *byte_ret) +{ + struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data; + int ret; + + if (!algo_data->running) + return -EIO; + + ret = i2c_algo_dp_aux_transaction(adapter, MODE_I2C_READ, 0, byte_ret); + return ret; +} + +static int +i2c_algo_dp_aux_xfer(struct i2c_adapter *adapter, + struct i2c_msg *msgs, + int num) +{ + int ret = 0; + bool reading = false; + int m; + int b; + + for (m = 0; m < num; m++) { + u16 len = msgs[m].len; + u8 *buf = msgs[m].buf; + reading = (msgs[m].flags & I2C_M_RD) != 0; + ret = i2c_algo_dp_aux_address(adapter, msgs[m].addr, reading); + if (ret < 0) + break; + if (reading) { + for (b = 0; b < len; b++) { + ret = i2c_algo_dp_aux_get_byte(adapter, &buf[b]); + if (ret < 0) + break; + } + } else { + for (b = 0; b < len; b++) { + ret = i2c_algo_dp_aux_put_byte(adapter, buf[b]); + if (ret < 0) + break; + } + } + if (ret < 0) + break; + } + if (ret >= 0) + ret = num; + i2c_algo_dp_aux_stop(adapter, reading); + DRM_DEBUG_KMS("dp_aux_xfer return %d\n", ret); + return ret; +} + +static u32 +i2c_algo_dp_aux_functionality(struct i2c_adapter *adapter) +{ + return I2C_FUNC_I2C | I2C_FUNC_SMBUS_EMUL | + I2C_FUNC_SMBUS_READ_BLOCK_DATA | + I2C_FUNC_SMBUS_BLOCK_PROC_CALL | + I2C_FUNC_10BIT_ADDR; +} + +static const struct i2c_algorithm i2c_dp_aux_algo = { + .master_xfer = i2c_algo_dp_aux_xfer, + .functionality = i2c_algo_dp_aux_functionality, +}; + +static void +i2c_dp_aux_reset_bus(struct i2c_adapter *adapter) +{ + (void) i2c_algo_dp_aux_address(adapter, 0, false); + (void) i2c_algo_dp_aux_stop(adapter, false); +} + +static int +i2c_dp_aux_prepare_bus(struct i2c_adapter *adapter) +{ + adapter->algo = &i2c_dp_aux_algo; + adapter->retries = 3; + i2c_dp_aux_reset_bus(adapter); + return 0; +} + +/* + * FIXME: This is the old dp aux helper, gma500 is the last driver that needs to + * be ported over to the new helper code in drm_dp_helper.c like i915 or radeon. + */ +static int __deprecated +i2c_dp_aux_add_bus(struct i2c_adapter *adapter) +{ + int error; + + error = i2c_dp_aux_prepare_bus(adapter); + if (error) + return error; + error = i2c_add_adapter(adapter); + return error; +} + #define _wait_for(COND, MS, W) ({ \ unsigned long timeout__ = jiffies + msecs_to_jiffies(MS); \ int ret__ = 0; \ diff --git a/include/drm/drm_dp_helper.h b/include/drm/drm_dp_helper.h index 9305c71..afebb08 100644 --- a/include/drm/drm_dp_helper.h +++ b/include/drm/drm_dp_helper.h @@ -404,26 +404,6 @@ #define MODE_I2C_READ 4 #define MODE_I2C_STOP 8 -/** - * struct i2c_algo_dp_aux_data - driver interface structure for i2c over dp - * aux algorithm - * @running: set by the algo indicating whether an i2c is ongoing or whether - * the i2c bus is quiescent - * @address: i2c target address for the currently ongoing transfer - * @aux_ch: driver callback to transfer a single byte of the i2c payload - */ -struct i2c_algo_dp_aux_data { - bool running; - u16 address; - int (*aux_ch) (struct i2c_adapter *adapter, - int mode, uint8_t write_byte, - uint8_t *read_byte); -}; - -int -i2c_dp_aux_add_bus(struct i2c_adapter *adapter); - - #define DP_LINK_STATUS_SIZE 6 bool drm_dp_channel_eq_ok(const u8 link_status[DP_LINK_STATUS_SIZE], int lane_count); -- cgit v0.10.2 From 5a1cbdb0fb6748a52a33f4ccd5d49486d7479fbb Mon Sep 17 00:00:00 2001 From: Masanari Iida Date: Thu, 23 Oct 2014 09:06:28 +0900 Subject: gpu: drm: Fix warning caused by a parameter description in drm_crtc.c This patch fix following error while "make xmldocs" Warning(.//drivers/gpu/drm/drm_crtc.c:778): Excess function parameter 'mode' description in 'drm_connector_get_cmdline_mode' Signed-off-by: Masanari Iida Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index b4a6f08..4081d7a 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -766,7 +766,6 @@ static void drm_mode_remove(struct drm_connector *connector, /** * drm_connector_get_cmdline_mode - reads the user's cmdline mode * @connector: connector to quwery - * @mode: returned mode * * The kernel supports per-connector configration of its consoles through * use of the video= parameter. This function parses that option and -- cgit v0.10.2 From cacc6c837b799b058d59d2af02c11140640cc1d2 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 22 Oct 2014 11:18:51 +0200 Subject: Revert "drm/i915: Enable full PPGTT on gen7" This reverts commit 8c50f10d73b50139dcfe48bc22f2c8c7822c1983. It's not yet solid and Dave objected to pulling the tree in its current state. Cc: Michel Thierry Cc: Dave Airlie Cc: Chris Wilson References: http://mid.mail-archive.com/CAPM=9ty2r1MLE=wzC-_vNSUzXVqAyXiGgocpSV9qOp0gzpK3xA@mail.gmail.com References: http://lists.freedesktop.org/archives/intel-gfx/2014-October/053926.html Acked-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 273dad9..8ddc834 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -67,7 +67,7 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) return 0; } - return has_full_ppgtt ? 2 : has_aliasing_ppgtt ? 1 : 0; + return has_aliasing_ppgtt ? 1 : 0; } -- cgit v0.10.2 From 47339cd9ff07376df1639260ecc088adf1856bfe Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:46 +0200 Subject: drm/i915: Extract intel_fifo_underrun.c Prep work for some nice documentation. Requires that we export the display irq enable/disable functions on ilk/ibx. But we already export them for vlv/i915. So not more inconsistency. v2: Rebase on top of skl stage 1. Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 3a6bce0..75fd7de 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -45,6 +45,7 @@ i915-y += intel_renderstate_gen6.o \ # modesetting core code i915-y += intel_bios.o \ intel_display.o \ + intel_fifo_underrun.o \ intel_frontbuffer.o \ intel_modes.o \ intel_overlay.o \ diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 9962da2..26724c5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2312,6 +2312,17 @@ i915_disable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe, void valleyview_enable_display_irqs(struct drm_i915_private *dev_priv); void valleyview_disable_display_irqs(struct drm_i915_private *dev_priv); +void +ironlake_enable_display_irq(struct drm_i915_private *dev_priv, u32 mask); +void +ironlake_disable_display_irq(struct drm_i915_private *dev_priv, u32 mask); +void ibx_display_interrupt_update(struct drm_i915_private *dev_priv, + uint32_t interrupt_mask, + uint32_t enabled_irq_mask); +#define ibx_enable_display_interrupt(dev_priv, bits) \ + ibx_display_interrupt_update((dev_priv), (bits), (bits)) +#define ibx_disable_display_interrupt(dev_priv, bits) \ + ibx_display_interrupt_update((dev_priv), (bits), 0) /* i915_gem.c */ int i915_gem_init_ioctl(struct drm_device *dev, void *data, diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index f17bbf3..536efa2 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -139,7 +139,7 @@ static const u32 hpd_status_i915[] = { /* i915 and valleyview are the same */ } while (0) /* For display hotplug interrupt */ -static void +void ironlake_enable_display_irq(struct drm_i915_private *dev_priv, u32 mask) { assert_spin_locked(&dev_priv->irq_lock); @@ -154,7 +154,7 @@ ironlake_enable_display_irq(struct drm_i915_private *dev_priv, u32 mask) } } -static void +void ironlake_disable_display_irq(struct drm_i915_private *dev_priv, u32 mask) { assert_spin_locked(&dev_priv->irq_lock); @@ -238,24 +238,6 @@ void gen6_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask) snb_update_pm_irq(dev_priv, mask, 0); } -static bool ivb_can_enable_err_int(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_crtc *crtc; - enum pipe pipe; - - assert_spin_locked(&dev_priv->irq_lock); - - for_each_pipe(dev_priv, pipe) { - crtc = to_intel_crtc(dev_priv->pipe_to_crtc_mapping[pipe]); - - if (crtc->cpu_fifo_underrun_disabled) - return false; - } - - return true; -} - /** * bdw_update_pm_irq - update GT interrupt 2 * @dev_priv: driver private @@ -296,130 +278,15 @@ void gen8_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask) bdw_update_pm_irq(dev_priv, mask, 0); } -static bool cpt_can_enable_serr_int(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - enum pipe pipe; - struct intel_crtc *crtc; - - assert_spin_locked(&dev_priv->irq_lock); - - for_each_pipe(dev_priv, pipe) { - crtc = to_intel_crtc(dev_priv->pipe_to_crtc_mapping[pipe]); - - if (crtc->pch_fifo_underrun_disabled) - return false; - } - - return true; -} - -void i9xx_check_fifo_underruns(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_crtc *crtc; - - spin_lock_irq(&dev_priv->irq_lock); - - for_each_intel_crtc(dev, crtc) { - u32 reg = PIPESTAT(crtc->pipe); - u32 pipestat; - - if (crtc->cpu_fifo_underrun_disabled) - continue; - - pipestat = I915_READ(reg) & 0xffff0000; - if ((pipestat & PIPE_FIFO_UNDERRUN_STATUS) == 0) - continue; - - I915_WRITE(reg, pipestat | PIPE_FIFO_UNDERRUN_STATUS); - POSTING_READ(reg); - - DRM_ERROR("pipe %c underrun\n", pipe_name(crtc->pipe)); - } - - spin_unlock_irq(&dev_priv->irq_lock); -} - -static void i9xx_set_fifo_underrun_reporting(struct drm_device *dev, - enum pipe pipe, - bool enable, bool old) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - u32 reg = PIPESTAT(pipe); - u32 pipestat = I915_READ(reg) & 0xffff0000; - - assert_spin_locked(&dev_priv->irq_lock); - - if (enable) { - I915_WRITE(reg, pipestat | PIPE_FIFO_UNDERRUN_STATUS); - POSTING_READ(reg); - } else { - if (old && pipestat & PIPE_FIFO_UNDERRUN_STATUS) - DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); - } -} - -static void ironlake_set_fifo_underrun_reporting(struct drm_device *dev, - enum pipe pipe, bool enable) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - uint32_t bit = (pipe == PIPE_A) ? DE_PIPEA_FIFO_UNDERRUN : - DE_PIPEB_FIFO_UNDERRUN; - - if (enable) - ironlake_enable_display_irq(dev_priv, bit); - else - ironlake_disable_display_irq(dev_priv, bit); -} - -static void ivybridge_set_fifo_underrun_reporting(struct drm_device *dev, - enum pipe pipe, - bool enable, bool old) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - if (enable) { - I915_WRITE(GEN7_ERR_INT, ERR_INT_FIFO_UNDERRUN(pipe)); - - if (!ivb_can_enable_err_int(dev)) - return; - - ironlake_enable_display_irq(dev_priv, DE_ERR_INT_IVB); - } else { - ironlake_disable_display_irq(dev_priv, DE_ERR_INT_IVB); - - if (old && - I915_READ(GEN7_ERR_INT) & ERR_INT_FIFO_UNDERRUN(pipe)) { - DRM_ERROR("uncleared fifo underrun on pipe %c\n", - pipe_name(pipe)); - } - } -} - -static void broadwell_set_fifo_underrun_reporting(struct drm_device *dev, - enum pipe pipe, bool enable) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - assert_spin_locked(&dev_priv->irq_lock); - - if (enable) - dev_priv->de_irq_mask[pipe] &= ~GEN8_PIPE_FIFO_UNDERRUN; - else - dev_priv->de_irq_mask[pipe] |= GEN8_PIPE_FIFO_UNDERRUN; - I915_WRITE(GEN8_DE_PIPE_IMR(pipe), dev_priv->de_irq_mask[pipe]); - POSTING_READ(GEN8_DE_PIPE_IMR(pipe)); -} - /** * ibx_display_interrupt_update - update SDEIMR * @dev_priv: driver private * @interrupt_mask: mask of interrupt bits to update * @enabled_irq_mask: mask of interrupt bits to enable */ -static void ibx_display_interrupt_update(struct drm_i915_private *dev_priv, - uint32_t interrupt_mask, - uint32_t enabled_irq_mask) +void ibx_display_interrupt_update(struct drm_i915_private *dev_priv, + uint32_t interrupt_mask, + uint32_t enabled_irq_mask) { uint32_t sdeimr = I915_READ(SDEIMR); sdeimr &= ~interrupt_mask; @@ -433,160 +300,6 @@ static void ibx_display_interrupt_update(struct drm_i915_private *dev_priv, I915_WRITE(SDEIMR, sdeimr); POSTING_READ(SDEIMR); } -#define ibx_enable_display_interrupt(dev_priv, bits) \ - ibx_display_interrupt_update((dev_priv), (bits), (bits)) -#define ibx_disable_display_interrupt(dev_priv, bits) \ - ibx_display_interrupt_update((dev_priv), (bits), 0) - -static void ibx_set_fifo_underrun_reporting(struct drm_device *dev, - enum transcoder pch_transcoder, - bool enable) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - uint32_t bit = (pch_transcoder == TRANSCODER_A) ? - SDE_TRANSA_FIFO_UNDER : SDE_TRANSB_FIFO_UNDER; - - if (enable) - ibx_enable_display_interrupt(dev_priv, bit); - else - ibx_disable_display_interrupt(dev_priv, bit); -} - -static void cpt_set_fifo_underrun_reporting(struct drm_device *dev, - enum transcoder pch_transcoder, - bool enable, bool old) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - if (enable) { - I915_WRITE(SERR_INT, - SERR_INT_TRANS_FIFO_UNDERRUN(pch_transcoder)); - - if (!cpt_can_enable_serr_int(dev)) - return; - - ibx_enable_display_interrupt(dev_priv, SDE_ERROR_CPT); - } else { - ibx_disable_display_interrupt(dev_priv, SDE_ERROR_CPT); - - if (old && I915_READ(SERR_INT) & - SERR_INT_TRANS_FIFO_UNDERRUN(pch_transcoder)) { - DRM_ERROR("uncleared pch fifo underrun on pch transcoder %c\n", - transcoder_name(pch_transcoder)); - } - } -} - -/** - * intel_set_cpu_fifo_underrun_reporting - enable/disable FIFO underrun messages - * @dev: drm device - * @pipe: pipe - * @enable: true if we want to report FIFO underrun errors, false otherwise - * - * This function makes us disable or enable CPU fifo underruns for a specific - * pipe. Notice that on some Gens (e.g. IVB, HSW), disabling FIFO underrun - * reporting for one pipe may also disable all the other CPU error interruts for - * the other pipes, due to the fact that there's just one interrupt mask/enable - * bit for all the pipes. - * - * Returns the previous state of underrun reporting. - */ -static bool __intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, - enum pipe pipe, bool enable) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pipe]; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - bool old; - - assert_spin_locked(&dev_priv->irq_lock); - - old = !intel_crtc->cpu_fifo_underrun_disabled; - intel_crtc->cpu_fifo_underrun_disabled = !enable; - - if (HAS_GMCH_DISPLAY(dev)) - i9xx_set_fifo_underrun_reporting(dev, pipe, enable, old); - else if (IS_GEN5(dev) || IS_GEN6(dev)) - ironlake_set_fifo_underrun_reporting(dev, pipe, enable); - else if (IS_GEN7(dev)) - ivybridge_set_fifo_underrun_reporting(dev, pipe, enable, old); - else if (IS_GEN8(dev) || IS_GEN9(dev)) - broadwell_set_fifo_underrun_reporting(dev, pipe, enable); - - return old; -} - -bool intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, - enum pipe pipe, bool enable) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long flags; - bool ret; - - spin_lock_irqsave(&dev_priv->irq_lock, flags); - ret = __intel_set_cpu_fifo_underrun_reporting(dev, pipe, enable); - spin_unlock_irqrestore(&dev_priv->irq_lock, flags); - - return ret; -} - -static bool __cpu_fifo_underrun_reporting_enabled(struct drm_device *dev, - enum pipe pipe) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pipe]; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - - return !intel_crtc->cpu_fifo_underrun_disabled; -} - -/** - * intel_set_pch_fifo_underrun_reporting - enable/disable FIFO underrun messages - * @dev: drm device - * @pch_transcoder: the PCH transcoder (same as pipe on IVB and older) - * @enable: true if we want to report FIFO underrun errors, false otherwise - * - * This function makes us disable or enable PCH fifo underruns for a specific - * PCH transcoder. Notice that on some PCHs (e.g. CPT/PPT), disabling FIFO - * underrun reporting for one transcoder may also disable all the other PCH - * error interruts for the other transcoders, due to the fact that there's just - * one interrupt mask/enable bit for all the transcoders. - * - * Returns the previous state of underrun reporting. - */ -bool intel_set_pch_fifo_underrun_reporting(struct drm_device *dev, - enum transcoder pch_transcoder, - bool enable) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pch_transcoder]; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - unsigned long flags; - bool old; - - /* - * NOTE: Pre-LPT has a fixed cpu pipe -> pch transcoder mapping, but LPT - * has only one pch transcoder A that all pipes can use. To avoid racy - * pch transcoder -> pipe lookups from interrupt code simply store the - * underrun statistics in crtc A. Since we never expose this anywhere - * nor use it outside of the fifo underrun code here using the "wrong" - * crtc on LPT won't cause issues. - */ - - spin_lock_irqsave(&dev_priv->irq_lock, flags); - - old = !intel_crtc->pch_fifo_underrun_disabled; - intel_crtc->pch_fifo_underrun_disabled = !enable; - - if (HAS_PCH_IBX(dev)) - ibx_set_fifo_underrun_reporting(dev, pch_transcoder, enable); - else - cpt_set_fifo_underrun_reporting(dev, pch_transcoder, enable, old); - - spin_unlock_irqrestore(&dev_priv->irq_lock, flags); - return old; -} - static void __i915_enable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe, diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 94993d2..321701d 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -755,12 +755,17 @@ static inline unsigned int intel_num_planes(struct intel_crtc *crtc) return INTEL_INFO(crtc->base.dev)->num_sprites[crtc->pipe] + 1; } -/* i915_irq.c */ +/* intel_fifo_underrun.c */ bool intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, enum pipe pipe, bool enable); bool intel_set_pch_fifo_underrun_reporting(struct drm_device *dev, enum transcoder pch_transcoder, bool enable); +void i9xx_check_fifo_underruns(struct drm_device *dev); +bool __cpu_fifo_underrun_reporting_enabled(struct drm_device *dev, + enum pipe pipe); + +/* i915_irq.c */ void gen5_enable_gt_irq(struct drm_i915_private *dev_priv, uint32_t mask); void gen5_disable_gt_irq(struct drm_i915_private *dev_priv, uint32_t mask); void gen6_enable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask); @@ -779,7 +784,6 @@ static inline bool intel_irqs_enabled(struct drm_i915_private *dev_priv) } int intel_get_crtc_scanline(struct intel_crtc *crtc); -void i9xx_check_fifo_underruns(struct drm_device *dev); void gen8_irq_power_well_post_enable(struct drm_i915_private *dev_priv); /* intel_crt.c */ diff --git a/drivers/gpu/drm/i915/intel_fifo_underrun.c b/drivers/gpu/drm/i915/intel_fifo_underrun.c new file mode 100644 index 0000000..8e79d2b --- /dev/null +++ b/drivers/gpu/drm/i915/intel_fifo_underrun.c @@ -0,0 +1,311 @@ +/* + * Copyright © 2014 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Daniel Vetter + * + */ + +#include "i915_drv.h" +#include "intel_drv.h" + +static bool ivb_can_enable_err_int(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_crtc *crtc; + enum pipe pipe; + + assert_spin_locked(&dev_priv->irq_lock); + + for_each_pipe(dev_priv, pipe) { + crtc = to_intel_crtc(dev_priv->pipe_to_crtc_mapping[pipe]); + + if (crtc->cpu_fifo_underrun_disabled) + return false; + } + + return true; +} + +static bool cpt_can_enable_serr_int(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + enum pipe pipe; + struct intel_crtc *crtc; + + assert_spin_locked(&dev_priv->irq_lock); + + for_each_pipe(dev_priv, pipe) { + crtc = to_intel_crtc(dev_priv->pipe_to_crtc_mapping[pipe]); + + if (crtc->pch_fifo_underrun_disabled) + return false; + } + + return true; +} + +void i9xx_check_fifo_underruns(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_crtc *crtc; + + spin_lock_irq(&dev_priv->irq_lock); + + for_each_intel_crtc(dev, crtc) { + u32 reg = PIPESTAT(crtc->pipe); + u32 pipestat; + + if (crtc->cpu_fifo_underrun_disabled) + continue; + + pipestat = I915_READ(reg) & 0xffff0000; + if ((pipestat & PIPE_FIFO_UNDERRUN_STATUS) == 0) + continue; + + I915_WRITE(reg, pipestat | PIPE_FIFO_UNDERRUN_STATUS); + POSTING_READ(reg); + + DRM_ERROR("pipe %c underrun\n", pipe_name(crtc->pipe)); + } + + spin_unlock_irq(&dev_priv->irq_lock); +} + +static void i9xx_set_fifo_underrun_reporting(struct drm_device *dev, + enum pipe pipe, + bool enable, bool old) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + u32 reg = PIPESTAT(pipe); + u32 pipestat = I915_READ(reg) & 0xffff0000; + + assert_spin_locked(&dev_priv->irq_lock); + + if (enable) { + I915_WRITE(reg, pipestat | PIPE_FIFO_UNDERRUN_STATUS); + POSTING_READ(reg); + } else { + if (old && pipestat & PIPE_FIFO_UNDERRUN_STATUS) + DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); + } +} + +static void ironlake_set_fifo_underrun_reporting(struct drm_device *dev, + enum pipe pipe, bool enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + uint32_t bit = (pipe == PIPE_A) ? DE_PIPEA_FIFO_UNDERRUN : + DE_PIPEB_FIFO_UNDERRUN; + + if (enable) + ironlake_enable_display_irq(dev_priv, bit); + else + ironlake_disable_display_irq(dev_priv, bit); +} + +static void ivybridge_set_fifo_underrun_reporting(struct drm_device *dev, + enum pipe pipe, + bool enable, bool old) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + if (enable) { + I915_WRITE(GEN7_ERR_INT, ERR_INT_FIFO_UNDERRUN(pipe)); + + if (!ivb_can_enable_err_int(dev)) + return; + + ironlake_enable_display_irq(dev_priv, DE_ERR_INT_IVB); + } else { + ironlake_disable_display_irq(dev_priv, DE_ERR_INT_IVB); + + if (old && + I915_READ(GEN7_ERR_INT) & ERR_INT_FIFO_UNDERRUN(pipe)) { + DRM_ERROR("uncleared fifo underrun on pipe %c\n", + pipe_name(pipe)); + } + } +} + +static void broadwell_set_fifo_underrun_reporting(struct drm_device *dev, + enum pipe pipe, bool enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + assert_spin_locked(&dev_priv->irq_lock); + + if (enable) + dev_priv->de_irq_mask[pipe] &= ~GEN8_PIPE_FIFO_UNDERRUN; + else + dev_priv->de_irq_mask[pipe] |= GEN8_PIPE_FIFO_UNDERRUN; + I915_WRITE(GEN8_DE_PIPE_IMR(pipe), dev_priv->de_irq_mask[pipe]); + POSTING_READ(GEN8_DE_PIPE_IMR(pipe)); +} + +static void ibx_set_fifo_underrun_reporting(struct drm_device *dev, + enum transcoder pch_transcoder, + bool enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + uint32_t bit = (pch_transcoder == TRANSCODER_A) ? + SDE_TRANSA_FIFO_UNDER : SDE_TRANSB_FIFO_UNDER; + + if (enable) + ibx_enable_display_interrupt(dev_priv, bit); + else + ibx_disable_display_interrupt(dev_priv, bit); +} + +static void cpt_set_fifo_underrun_reporting(struct drm_device *dev, + enum transcoder pch_transcoder, + bool enable, bool old) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + if (enable) { + I915_WRITE(SERR_INT, + SERR_INT_TRANS_FIFO_UNDERRUN(pch_transcoder)); + + if (!cpt_can_enable_serr_int(dev)) + return; + + ibx_enable_display_interrupt(dev_priv, SDE_ERROR_CPT); + } else { + ibx_disable_display_interrupt(dev_priv, SDE_ERROR_CPT); + + if (old && I915_READ(SERR_INT) & + SERR_INT_TRANS_FIFO_UNDERRUN(pch_transcoder)) { + DRM_ERROR("uncleared pch fifo underrun on pch transcoder %c\n", + transcoder_name(pch_transcoder)); + } + } +} + +/** + * intel_set_cpu_fifo_underrun_reporting - enable/disable FIFO underrun messages + * @dev: drm device + * @pipe: pipe + * @enable: true if we want to report FIFO underrun errors, false otherwise + * + * This function makes us disable or enable CPU fifo underruns for a specific + * pipe. Notice that on some Gens (e.g. IVB, HSW), disabling FIFO underrun + * reporting for one pipe may also disable all the other CPU error interruts for + * the other pipes, due to the fact that there's just one interrupt mask/enable + * bit for all the pipes. + * + * Returns the previous state of underrun reporting. + */ +static bool __intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, + enum pipe pipe, bool enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pipe]; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + bool old; + + assert_spin_locked(&dev_priv->irq_lock); + + old = !intel_crtc->cpu_fifo_underrun_disabled; + intel_crtc->cpu_fifo_underrun_disabled = !enable; + + if (HAS_GMCH_DISPLAY(dev)) + i9xx_set_fifo_underrun_reporting(dev, pipe, enable, old); + else if (IS_GEN5(dev) || IS_GEN6(dev)) + ironlake_set_fifo_underrun_reporting(dev, pipe, enable); + else if (IS_GEN7(dev)) + ivybridge_set_fifo_underrun_reporting(dev, pipe, enable, old); + else if (IS_GEN8(dev) || IS_GEN9(dev)) + broadwell_set_fifo_underrun_reporting(dev, pipe, enable); + + return old; +} + +bool intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, + enum pipe pipe, bool enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + bool ret; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + ret = __intel_set_cpu_fifo_underrun_reporting(dev, pipe, enable); + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + + return ret; +} + +bool __cpu_fifo_underrun_reporting_enabled(struct drm_device *dev, + enum pipe pipe) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pipe]; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + + return !intel_crtc->cpu_fifo_underrun_disabled; +} + +/** + * intel_set_pch_fifo_underrun_reporting - enable/disable FIFO underrun messages + * @dev: drm device + * @pch_transcoder: the PCH transcoder (same as pipe on IVB and older) + * @enable: true if we want to report FIFO underrun errors, false otherwise + * + * This function makes us disable or enable PCH fifo underruns for a specific + * PCH transcoder. Notice that on some PCHs (e.g. CPT/PPT), disabling FIFO + * underrun reporting for one transcoder may also disable all the other PCH + * error interruts for the other transcoders, due to the fact that there's just + * one interrupt mask/enable bit for all the transcoders. + * + * Returns the previous state of underrun reporting. + */ +bool intel_set_pch_fifo_underrun_reporting(struct drm_device *dev, + enum transcoder pch_transcoder, + bool enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pch_transcoder]; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + unsigned long flags; + bool old; + + /* + * NOTE: Pre-LPT has a fixed cpu pipe -> pch transcoder mapping, but LPT + * has only one pch transcoder A that all pipes can use. To avoid racy + * pch transcoder -> pipe lookups from interrupt code simply store the + * underrun statistics in crtc A. Since we never expose this anywhere + * nor use it outside of the fifo underrun code here using the "wrong" + * crtc on LPT won't cause issues. + */ + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + + old = !intel_crtc->pch_fifo_underrun_disabled; + intel_crtc->pch_fifo_underrun_disabled = !enable; + + if (HAS_PCH_IBX(dev)) + ibx_set_fifo_underrun_reporting(dev, pch_transcoder, enable); + else + cpt_set_fifo_underrun_reporting(dev, pch_transcoder, enable, old); + + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + return old; +} -- cgit v0.10.2 From a72e4c9f9a11d44458a12928cc776921ae43a616 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:47 +0200 Subject: drm/i915: Use dev_priv in public intel_fifo_underrun.c functions It's the new rule! Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 536efa2..3ee8c47 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1750,7 +1750,7 @@ static void valleyview_pipestat_irq_handler(struct drm_device *dev, u32 iir) * handle. */ mask = 0; - if (__cpu_fifo_underrun_reporting_enabled(dev, pipe)) + if (__cpu_fifo_underrun_reporting_enabled(dev_priv, pipe)) mask |= PIPE_FIFO_UNDERRUN_STATUS; switch (pipe) { @@ -1797,7 +1797,8 @@ static void valleyview_pipestat_irq_handler(struct drm_device *dev, u32 iir) i9xx_pipe_crc_irq_handler(dev, pipe); if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS && - intel_set_cpu_fifo_underrun_reporting(dev, pipe, false)) + intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, + false)) DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); } @@ -1965,12 +1966,14 @@ static void ibx_irq_handler(struct drm_device *dev, u32 pch_iir) DRM_DEBUG_DRIVER("PCH transcoder CRC error interrupt\n"); if (pch_iir & SDE_TRANSA_FIFO_UNDER) - if (intel_set_pch_fifo_underrun_reporting(dev, TRANSCODER_A, + if (intel_set_pch_fifo_underrun_reporting(dev_priv, + TRANSCODER_A, false)) DRM_ERROR("PCH transcoder A FIFO underrun\n"); if (pch_iir & SDE_TRANSB_FIFO_UNDER) - if (intel_set_pch_fifo_underrun_reporting(dev, TRANSCODER_B, + if (intel_set_pch_fifo_underrun_reporting(dev_priv, + TRANSCODER_B, false)) DRM_ERROR("PCH transcoder B FIFO underrun\n"); } @@ -1986,7 +1989,7 @@ static void ivb_err_int_handler(struct drm_device *dev) for_each_pipe(dev_priv, pipe) { if (err_int & ERR_INT_FIFO_UNDERRUN(pipe)) { - if (intel_set_cpu_fifo_underrun_reporting(dev, pipe, + if (intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, false)) DRM_ERROR("Pipe %c FIFO underrun\n", pipe_name(pipe)); @@ -2012,17 +2015,17 @@ static void cpt_serr_int_handler(struct drm_device *dev) DRM_ERROR("PCH poison interrupt\n"); if (serr_int & SERR_INT_TRANS_A_FIFO_UNDERRUN) - if (intel_set_pch_fifo_underrun_reporting(dev, TRANSCODER_A, + if (intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_A, false)) DRM_ERROR("PCH transcoder A FIFO underrun\n"); if (serr_int & SERR_INT_TRANS_B_FIFO_UNDERRUN) - if (intel_set_pch_fifo_underrun_reporting(dev, TRANSCODER_B, + if (intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_B, false)) DRM_ERROR("PCH transcoder B FIFO underrun\n"); if (serr_int & SERR_INT_TRANS_C_FIFO_UNDERRUN) - if (intel_set_pch_fifo_underrun_reporting(dev, TRANSCODER_C, + if (intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_C, false)) DRM_ERROR("PCH transcoder C FIFO underrun\n"); @@ -2090,7 +2093,9 @@ static void ilk_display_irq_handler(struct drm_device *dev, u32 de_iir) intel_check_page_flip(dev, pipe); if (de_iir & DE_PIPE_FIFO_UNDERRUN(pipe)) - if (intel_set_cpu_fifo_underrun_reporting(dev, pipe, false)) + if (intel_set_cpu_fifo_underrun_reporting(dev_priv, + pipe, + false)) DRM_ERROR("Pipe %c FIFO underrun\n", pipe_name(pipe)); @@ -2312,7 +2317,8 @@ static irqreturn_t gen8_irq_handler(int irq, void *arg) hsw_pipe_crc_irq_handler(dev, pipe); if (pipe_iir & GEN8_PIPE_FIFO_UNDERRUN) { - if (intel_set_cpu_fifo_underrun_reporting(dev, pipe, + if (intel_set_cpu_fifo_underrun_reporting(dev_priv, + pipe, false)) DRM_ERROR("Pipe %c FIFO underrun\n", pipe_name(pipe)); @@ -3834,7 +3840,8 @@ static irqreturn_t i8xx_irq_handler(int irq, void *arg) i9xx_pipe_crc_irq_handler(dev, pipe); if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS && - intel_set_cpu_fifo_underrun_reporting(dev, pipe, false)) + intel_set_cpu_fifo_underrun_reporting(dev_priv, + pipe, false)) DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); } @@ -4028,7 +4035,8 @@ static irqreturn_t i915_irq_handler(int irq, void *arg) i9xx_pipe_crc_irq_handler(dev, pipe); if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS && - intel_set_cpu_fifo_underrun_reporting(dev, pipe, false)) + intel_set_cpu_fifo_underrun_reporting(dev_priv, + pipe, false)) DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); } @@ -4256,7 +4264,8 @@ static irqreturn_t i965_irq_handler(int irq, void *arg) i9xx_pipe_crc_irq_handler(dev, pipe); if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS && - intel_set_cpu_fifo_underrun_reporting(dev, pipe, false)) + intel_set_cpu_fifo_underrun_reporting(dev_priv, + pipe, false)) DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 1fc05ff..ee8156c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -4163,8 +4163,8 @@ static void ironlake_crtc_enable(struct drm_crtc *crtc) intel_crtc->active = true; - intel_set_cpu_fifo_underrun_reporting(dev, pipe, true); - intel_set_pch_fifo_underrun_reporting(dev, pipe, true); + intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, true); + intel_set_pch_fifo_underrun_reporting(dev_priv, pipe, true); for_each_encoder_on_crtc(dev, crtc, encoder) if (encoder->pre_enable) @@ -4278,13 +4278,14 @@ static void haswell_crtc_enable(struct drm_crtc *crtc) intel_crtc->active = true; - intel_set_cpu_fifo_underrun_reporting(dev, pipe, true); + intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, true); for_each_encoder_on_crtc(dev, crtc, encoder) if (encoder->pre_enable) encoder->pre_enable(encoder); if (intel_crtc->config.has_pch_encoder) { - intel_set_pch_fifo_underrun_reporting(dev, TRANSCODER_A, true); + intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_A, + true); dev_priv->display.fdi_link_train(crtc); } @@ -4360,7 +4361,7 @@ static void ironlake_crtc_disable(struct drm_crtc *crtc) encoder->disable(encoder); if (intel_crtc->config.has_pch_encoder) - intel_set_pch_fifo_underrun_reporting(dev, pipe, false); + intel_set_pch_fifo_underrun_reporting(dev_priv, pipe, false); intel_disable_pipe(intel_crtc); @@ -4374,7 +4375,7 @@ static void ironlake_crtc_disable(struct drm_crtc *crtc) ironlake_fdi_disable(crtc); ironlake_disable_pch_transcoder(dev_priv, pipe); - intel_set_pch_fifo_underrun_reporting(dev, pipe, true); + intel_set_pch_fifo_underrun_reporting(dev_priv, pipe, true); if (HAS_PCH_CPT(dev)) { /* disable TRANS_DP_CTL */ @@ -4427,7 +4428,8 @@ static void haswell_crtc_disable(struct drm_crtc *crtc) } if (intel_crtc->config.has_pch_encoder) - intel_set_pch_fifo_underrun_reporting(dev, TRANSCODER_A, false); + intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_A, + false); intel_disable_pipe(intel_crtc); if (intel_crtc->config.dp_encoder_is_mst) @@ -4441,7 +4443,8 @@ static void haswell_crtc_disable(struct drm_crtc *crtc) if (intel_crtc->config.has_pch_encoder) { lpt_disable_pch_transcoder(dev_priv); - intel_set_pch_fifo_underrun_reporting(dev, TRANSCODER_A, true); + intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_A, + true); intel_ddi_fdi_disable(crtc); } @@ -4818,6 +4821,7 @@ static void valleyview_modeset_global_resources(struct drm_device *dev) static void valleyview_crtc_enable(struct drm_crtc *crtc) { struct drm_device *dev = crtc->dev; + struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); struct intel_encoder *encoder; int pipe = intel_crtc->pipe; @@ -4846,7 +4850,7 @@ static void valleyview_crtc_enable(struct drm_crtc *crtc) intel_crtc->active = true; - intel_set_cpu_fifo_underrun_reporting(dev, pipe, true); + intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, true); for_each_encoder_on_crtc(dev, crtc, encoder) if (encoder->pre_pll_enable) @@ -4879,7 +4883,7 @@ static void valleyview_crtc_enable(struct drm_crtc *crtc) intel_crtc_enable_planes(crtc); /* Underruns don't raise interrupts, so check manually. */ - i9xx_check_fifo_underruns(dev); + i9xx_check_fifo_underruns(dev_priv); } static void i9xx_set_pll_dividers(struct intel_crtc *crtc) @@ -4894,6 +4898,7 @@ static void i9xx_set_pll_dividers(struct intel_crtc *crtc) static void i9xx_crtc_enable(struct drm_crtc *crtc) { struct drm_device *dev = crtc->dev; + struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); struct intel_encoder *encoder; int pipe = intel_crtc->pipe; @@ -4915,7 +4920,7 @@ static void i9xx_crtc_enable(struct drm_crtc *crtc) intel_crtc->active = true; if (!IS_GEN2(dev)) - intel_set_cpu_fifo_underrun_reporting(dev, pipe, true); + intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, true); for_each_encoder_on_crtc(dev, crtc, encoder) if (encoder->pre_enable) @@ -4946,10 +4951,10 @@ static void i9xx_crtc_enable(struct drm_crtc *crtc) * but leave the pipe running. */ if (IS_GEN2(dev)) - intel_set_cpu_fifo_underrun_reporting(dev, pipe, true); + intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, true); /* Underruns don't raise interrupts, so check manually. */ - i9xx_check_fifo_underruns(dev); + i9xx_check_fifo_underruns(dev_priv); } static void i9xx_pfit_disable(struct intel_crtc *crtc) @@ -4985,7 +4990,7 @@ static void i9xx_crtc_disable(struct drm_crtc *crtc) * but leave the pipe running. */ if (IS_GEN2(dev)) - intel_set_cpu_fifo_underrun_reporting(dev, pipe, false); + intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, false); /* * Vblank time updates from the shadow to live plane control register @@ -5031,7 +5036,7 @@ static void i9xx_crtc_disable(struct drm_crtc *crtc) } if (!IS_GEN2(dev)) - intel_set_cpu_fifo_underrun_reporting(dev, pipe, false); + intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, false); intel_crtc->active = false; intel_update_watermarks(crtc); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 321701d..ede9606 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -756,13 +756,13 @@ static inline unsigned int intel_num_planes(struct intel_crtc *crtc) } /* intel_fifo_underrun.c */ -bool intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, +bool intel_set_cpu_fifo_underrun_reporting(struct drm_i915_private *dev_priv, enum pipe pipe, bool enable); -bool intel_set_pch_fifo_underrun_reporting(struct drm_device *dev, +bool intel_set_pch_fifo_underrun_reporting(struct drm_i915_private *dev_priv, enum transcoder pch_transcoder, bool enable); -void i9xx_check_fifo_underruns(struct drm_device *dev); -bool __cpu_fifo_underrun_reporting_enabled(struct drm_device *dev, +void i9xx_check_fifo_underruns(struct drm_i915_private *dev_priv); +bool __cpu_fifo_underrun_reporting_enabled(struct drm_i915_private *dev_priv, enum pipe pipe); /* i915_irq.c */ diff --git a/drivers/gpu/drm/i915/intel_fifo_underrun.c b/drivers/gpu/drm/i915/intel_fifo_underrun.c index 8e79d2b..0604a11 100644 --- a/drivers/gpu/drm/i915/intel_fifo_underrun.c +++ b/drivers/gpu/drm/i915/intel_fifo_underrun.c @@ -64,14 +64,13 @@ static bool cpt_can_enable_serr_int(struct drm_device *dev) return true; } -void i9xx_check_fifo_underruns(struct drm_device *dev) +void i9xx_check_fifo_underruns(struct drm_i915_private *dev_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *crtc; spin_lock_irq(&dev_priv->irq_lock); - for_each_intel_crtc(dev, crtc) { + for_each_intel_crtc(dev_priv->dev, crtc) { u32 reg = PIPESTAT(crtc->pipe); u32 pipestat; @@ -239,24 +238,23 @@ static bool __intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, return old; } -bool intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, +bool intel_set_cpu_fifo_underrun_reporting(struct drm_i915_private *dev_priv, enum pipe pipe, bool enable) { - struct drm_i915_private *dev_priv = dev->dev_private; unsigned long flags; bool ret; spin_lock_irqsave(&dev_priv->irq_lock, flags); - ret = __intel_set_cpu_fifo_underrun_reporting(dev, pipe, enable); + ret = __intel_set_cpu_fifo_underrun_reporting(dev_priv->dev, pipe, + enable); spin_unlock_irqrestore(&dev_priv->irq_lock, flags); return ret; } -bool __cpu_fifo_underrun_reporting_enabled(struct drm_device *dev, +bool __cpu_fifo_underrun_reporting_enabled(struct drm_i915_private *dev_priv, enum pipe pipe) { - struct drm_i915_private *dev_priv = dev->dev_private; struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pipe]; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); @@ -277,11 +275,10 @@ bool __cpu_fifo_underrun_reporting_enabled(struct drm_device *dev, * * Returns the previous state of underrun reporting. */ -bool intel_set_pch_fifo_underrun_reporting(struct drm_device *dev, +bool intel_set_pch_fifo_underrun_reporting(struct drm_i915_private *dev_priv, enum transcoder pch_transcoder, bool enable) { - struct drm_i915_private *dev_priv = dev->dev_private; struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pch_transcoder]; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); unsigned long flags; @@ -301,10 +298,12 @@ bool intel_set_pch_fifo_underrun_reporting(struct drm_device *dev, old = !intel_crtc->pch_fifo_underrun_disabled; intel_crtc->pch_fifo_underrun_disabled = !enable; - if (HAS_PCH_IBX(dev)) - ibx_set_fifo_underrun_reporting(dev, pch_transcoder, enable); + if (HAS_PCH_IBX(dev_priv->dev)) + ibx_set_fifo_underrun_reporting(dev_priv->dev, pch_transcoder, + enable); else - cpt_set_fifo_underrun_reporting(dev, pch_transcoder, enable, old); + cpt_set_fifo_underrun_reporting(dev_priv->dev, pch_transcoder, + enable, old); spin_unlock_irqrestore(&dev_priv->irq_lock, flags); return old; -- cgit v0.10.2 From 1f7247c01c9d1bdc7b0e239a8f2e256972b9189b Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:48 +0200 Subject: drm/i915: Add wrappers to handle fifo underrun interrupts Way too much copypasta all over. And this also clarifies a bit what's going on since it separates the "do we have an underrun irq" from the "should we report the underrun" check. v2: Fix excessively long lines. Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 3ee8c47..1db9fb5 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1796,10 +1796,8 @@ static void valleyview_pipestat_irq_handler(struct drm_device *dev, u32 iir) if (pipe_stats[pipe] & PIPE_CRC_DONE_INTERRUPT_STATUS) i9xx_pipe_crc_irq_handler(dev, pipe); - if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS && - intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, - false)) - DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); + if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS) + intel_cpu_fifo_underrun_irq_handler(dev_priv, pipe); } if (pipe_stats[0] & PIPE_GMBUS_INTERRUPT_STATUS) @@ -1966,16 +1964,10 @@ static void ibx_irq_handler(struct drm_device *dev, u32 pch_iir) DRM_DEBUG_DRIVER("PCH transcoder CRC error interrupt\n"); if (pch_iir & SDE_TRANSA_FIFO_UNDER) - if (intel_set_pch_fifo_underrun_reporting(dev_priv, - TRANSCODER_A, - false)) - DRM_ERROR("PCH transcoder A FIFO underrun\n"); + intel_pch_fifo_underrun_irq_handler(dev_priv, TRANSCODER_A); if (pch_iir & SDE_TRANSB_FIFO_UNDER) - if (intel_set_pch_fifo_underrun_reporting(dev_priv, - TRANSCODER_B, - false)) - DRM_ERROR("PCH transcoder B FIFO underrun\n"); + intel_pch_fifo_underrun_irq_handler(dev_priv, TRANSCODER_B); } static void ivb_err_int_handler(struct drm_device *dev) @@ -1988,12 +1980,8 @@ static void ivb_err_int_handler(struct drm_device *dev) DRM_ERROR("Poison interrupt\n"); for_each_pipe(dev_priv, pipe) { - if (err_int & ERR_INT_FIFO_UNDERRUN(pipe)) { - if (intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, - false)) - DRM_ERROR("Pipe %c FIFO underrun\n", - pipe_name(pipe)); - } + if (err_int & ERR_INT_FIFO_UNDERRUN(pipe)) + intel_cpu_fifo_underrun_irq_handler(dev_priv, pipe); if (err_int & ERR_INT_PIPE_CRC_DONE(pipe)) { if (IS_IVYBRIDGE(dev)) @@ -2015,19 +2003,13 @@ static void cpt_serr_int_handler(struct drm_device *dev) DRM_ERROR("PCH poison interrupt\n"); if (serr_int & SERR_INT_TRANS_A_FIFO_UNDERRUN) - if (intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_A, - false)) - DRM_ERROR("PCH transcoder A FIFO underrun\n"); + intel_pch_fifo_underrun_irq_handler(dev_priv, TRANSCODER_A); if (serr_int & SERR_INT_TRANS_B_FIFO_UNDERRUN) - if (intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_B, - false)) - DRM_ERROR("PCH transcoder B FIFO underrun\n"); + intel_pch_fifo_underrun_irq_handler(dev_priv, TRANSCODER_B); if (serr_int & SERR_INT_TRANS_C_FIFO_UNDERRUN) - if (intel_set_pch_fifo_underrun_reporting(dev_priv, TRANSCODER_C, - false)) - DRM_ERROR("PCH transcoder C FIFO underrun\n"); + intel_pch_fifo_underrun_irq_handler(dev_priv, TRANSCODER_C); I915_WRITE(SERR_INT, serr_int); } @@ -2093,11 +2075,7 @@ static void ilk_display_irq_handler(struct drm_device *dev, u32 de_iir) intel_check_page_flip(dev, pipe); if (de_iir & DE_PIPE_FIFO_UNDERRUN(pipe)) - if (intel_set_cpu_fifo_underrun_reporting(dev_priv, - pipe, - false)) - DRM_ERROR("Pipe %c FIFO underrun\n", - pipe_name(pipe)); + intel_cpu_fifo_underrun_irq_handler(dev_priv, pipe); if (de_iir & DE_PIPE_CRC_DONE(pipe)) i9xx_pipe_crc_irq_handler(dev, pipe); @@ -2316,13 +2294,9 @@ static irqreturn_t gen8_irq_handler(int irq, void *arg) if (pipe_iir & GEN8_PIPE_CDCLK_CRC_DONE) hsw_pipe_crc_irq_handler(dev, pipe); - if (pipe_iir & GEN8_PIPE_FIFO_UNDERRUN) { - if (intel_set_cpu_fifo_underrun_reporting(dev_priv, - pipe, - false)) - DRM_ERROR("Pipe %c FIFO underrun\n", - pipe_name(pipe)); - } + if (pipe_iir & GEN8_PIPE_FIFO_UNDERRUN) + intel_cpu_fifo_underrun_irq_handler(dev_priv, + pipe); if (IS_GEN9(dev)) @@ -3839,10 +3813,9 @@ static irqreturn_t i8xx_irq_handler(int irq, void *arg) if (pipe_stats[pipe] & PIPE_CRC_DONE_INTERRUPT_STATUS) i9xx_pipe_crc_irq_handler(dev, pipe); - if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS && - intel_set_cpu_fifo_underrun_reporting(dev_priv, - pipe, false)) - DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); + if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS) + intel_cpu_fifo_underrun_irq_handler(dev_priv, + pipe); } iir = new_iir; @@ -4034,10 +4007,9 @@ static irqreturn_t i915_irq_handler(int irq, void *arg) if (pipe_stats[pipe] & PIPE_CRC_DONE_INTERRUPT_STATUS) i9xx_pipe_crc_irq_handler(dev, pipe); - if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS && - intel_set_cpu_fifo_underrun_reporting(dev_priv, - pipe, false)) - DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); + if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS) + intel_cpu_fifo_underrun_irq_handler(dev_priv, + pipe); } if (blc_event || (iir & I915_ASLE_INTERRUPT)) @@ -4263,10 +4235,8 @@ static irqreturn_t i965_irq_handler(int irq, void *arg) if (pipe_stats[pipe] & PIPE_CRC_DONE_INTERRUPT_STATUS) i9xx_pipe_crc_irq_handler(dev, pipe); - if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS && - intel_set_cpu_fifo_underrun_reporting(dev_priv, - pipe, false)) - DRM_ERROR("pipe %c underrun\n", pipe_name(pipe)); + if (pipe_stats[pipe] & PIPE_FIFO_UNDERRUN_STATUS) + intel_cpu_fifo_underrun_irq_handler(dev_priv, pipe); } if (blc_event || (iir & I915_ASLE_INTERRUPT)) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index ede9606..4ffbe4e 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -761,6 +761,10 @@ bool intel_set_cpu_fifo_underrun_reporting(struct drm_i915_private *dev_priv, bool intel_set_pch_fifo_underrun_reporting(struct drm_i915_private *dev_priv, enum transcoder pch_transcoder, bool enable); +void intel_cpu_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, + enum pipe pipe); +void intel_pch_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, + enum transcoder pch_transcoder); void i9xx_check_fifo_underruns(struct drm_i915_private *dev_priv); bool __cpu_fifo_underrun_reporting_enabled(struct drm_i915_private *dev_priv, enum pipe pipe); diff --git a/drivers/gpu/drm/i915/intel_fifo_underrun.c b/drivers/gpu/drm/i915/intel_fifo_underrun.c index 0604a11..5d66b93 100644 --- a/drivers/gpu/drm/i915/intel_fifo_underrun.c +++ b/drivers/gpu/drm/i915/intel_fifo_underrun.c @@ -308,3 +308,20 @@ bool intel_set_pch_fifo_underrun_reporting(struct drm_i915_private *dev_priv, spin_unlock_irqrestore(&dev_priv->irq_lock, flags); return old; } + +void intel_cpu_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, + enum pipe pipe) +{ + if (intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, false)) + DRM_ERROR("CPU pipe %c FIFO underrun\n", + pipe_name(pipe)); +} + +void intel_pch_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, + enum transcoder pch_transcoder) +{ + if (intel_set_pch_fifo_underrun_reporting(dev_priv, pch_transcoder, + false)) + DRM_ERROR("PCH transcoder %c FIFO underrun\n", + transcoder_name(pch_transcoder)); +} -- cgit v0.10.2 From 0f239f4cb1dbbe259b7a29fed1e2ff8a50591319 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:49 +0200 Subject: drm/i915: Filter gmch fifo underruns in the shared handler This simplifies the code in the vlv irq handler. Also this now means that we correctly filter underruns on gen2-4. And as the real upshot I need to document one less function for the fifo underrun code. v2: Shorten one long line. Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 1db9fb5..a2b013d 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1749,9 +1749,9 @@ static void valleyview_pipestat_irq_handler(struct drm_device *dev, u32 iir) * we need to be careful that we only handle what we want to * handle. */ - mask = 0; - if (__cpu_fifo_underrun_reporting_enabled(dev_priv, pipe)) - mask |= PIPE_FIFO_UNDERRUN_STATUS; + + /* fifo underruns are filterered in the underrun handler. */ + mask = PIPE_FIFO_UNDERRUN_STATUS; switch (pipe) { case PIPE_A: diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 4ffbe4e..5ab813c 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -766,8 +766,6 @@ void intel_cpu_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, void intel_pch_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, enum transcoder pch_transcoder); void i9xx_check_fifo_underruns(struct drm_i915_private *dev_priv); -bool __cpu_fifo_underrun_reporting_enabled(struct drm_i915_private *dev_priv, - enum pipe pipe); /* i915_irq.c */ void gen5_enable_gt_irq(struct drm_i915_private *dev_priv, uint32_t mask); diff --git a/drivers/gpu/drm/i915/intel_fifo_underrun.c b/drivers/gpu/drm/i915/intel_fifo_underrun.c index 5d66b93..8622ab10 100644 --- a/drivers/gpu/drm/i915/intel_fifo_underrun.c +++ b/drivers/gpu/drm/i915/intel_fifo_underrun.c @@ -252,8 +252,9 @@ bool intel_set_cpu_fifo_underrun_reporting(struct drm_i915_private *dev_priv, return ret; } -bool __cpu_fifo_underrun_reporting_enabled(struct drm_i915_private *dev_priv, - enum pipe pipe) +static bool +__cpu_fifo_underrun_reporting_enabled(struct drm_i915_private *dev_priv, + enum pipe pipe) { struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pipe]; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); @@ -312,6 +313,11 @@ bool intel_set_pch_fifo_underrun_reporting(struct drm_i915_private *dev_priv, void intel_cpu_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, enum pipe pipe) { + /* GMCH can't disable fifo underruns, filter them. */ + if (HAS_GMCH_DISPLAY(dev_priv->dev) && + !__cpu_fifo_underrun_reporting_enabled(dev_priv, pipe)) + return; + if (intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, false)) DRM_ERROR("CPU pipe %c FIFO underrun\n", pipe_name(pipe)); -- cgit v0.10.2 From ef07388e8832394f92f124e4069014d5b33cb39e Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 30 Sep 2014 10:56:50 +0200 Subject: drm/i915: kerneldoc for intel_fifo_underrun.c v2: Fix spelling fail. Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index d7cfc98..f6a9d7b 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -3831,6 +3831,11 @@ int num_ioctls; !Fdrivers/gpu/drm/i915/i915_gem.c i915_gem_track_fb + Display FIFO Underrun Reporting +!Pdrivers/gpu/drm/i915/intel_fifo_underrun.c fifo underrun handling +!Idrivers/gpu/drm/i915/intel_fifo_underrun.c + + Plane Configuration This section covers plane configuration and composition with the diff --git a/drivers/gpu/drm/i915/intel_fifo_underrun.c b/drivers/gpu/drm/i915/intel_fifo_underrun.c index 8622ab10..77af512 100644 --- a/drivers/gpu/drm/i915/intel_fifo_underrun.c +++ b/drivers/gpu/drm/i915/intel_fifo_underrun.c @@ -28,6 +28,26 @@ #include "i915_drv.h" #include "intel_drv.h" +/** + * DOC: fifo underrun handling + * + * The i915 driver checks for display fifo underruns using the interrupt signals + * provided by the hardware. This is enabled by default and fairly useful to + * debug display issues, especially watermark settings. + * + * If an underrun is detected this is logged into dmesg. To avoid flooding logs + * and occupying the cpu underrun interrupts are disabled after the first + * occurrence until the next modeset on a given pipe. + * + * Note that underrun detection on gmch platforms is a bit more ugly since there + * is no interrupt (despite that the signalling bit is in the PIPESTAT pipe + * interrupt register). Also on some other platforms underrun interrupts are + * shared, which means that if we detect an underrun we need to disable underrun + * reporting on all pipes. + * + * The code also supports underrun detection on the PCH transcoder. + */ + static bool ivb_can_enable_err_int(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -64,6 +84,14 @@ static bool cpt_can_enable_serr_int(struct drm_device *dev) return true; } +/** + * i9xx_check_fifo_underruns - check for fifo underruns + * @dev_priv: i915 device instance + * + * This function checks for fifo underruns on GMCH platforms. This needs to be + * done manually on modeset to make sure that we catch all underruns since they + * do not generate an interrupt by themselves on these platforms. + */ void i9xx_check_fifo_underruns(struct drm_i915_private *dev_priv) { struct intel_crtc *crtc; @@ -199,20 +227,6 @@ static void cpt_set_fifo_underrun_reporting(struct drm_device *dev, } } -/** - * intel_set_cpu_fifo_underrun_reporting - enable/disable FIFO underrun messages - * @dev: drm device - * @pipe: pipe - * @enable: true if we want to report FIFO underrun errors, false otherwise - * - * This function makes us disable or enable CPU fifo underruns for a specific - * pipe. Notice that on some Gens (e.g. IVB, HSW), disabling FIFO underrun - * reporting for one pipe may also disable all the other CPU error interruts for - * the other pipes, due to the fact that there's just one interrupt mask/enable - * bit for all the pipes. - * - * Returns the previous state of underrun reporting. - */ static bool __intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, enum pipe pipe, bool enable) { @@ -238,6 +252,22 @@ static bool __intel_set_cpu_fifo_underrun_reporting(struct drm_device *dev, return old; } +/** + * intel_set_cpu_fifo_underrun_reporting - set cpu fifo underrrun reporting state + * @dev_priv: i915 device instance + * @pipe: (CPU) pipe to set state for + * @enable: whether underruns should be reported or not + * + * This function sets the fifo underrun state for @pipe. It is used in the + * modeset code to avoid false positives since on many platforms underruns are + * expected when disabling or enabling the pipe. + * + * Notice that on some platforms disabling underrun reports for one pipe + * disables for all due to shared interrupts. Actual reporting is still per-pipe + * though. + * + * Returns the previous state of underrun reporting. + */ bool intel_set_cpu_fifo_underrun_reporting(struct drm_i915_private *dev_priv, enum pipe pipe, bool enable) { @@ -263,10 +293,10 @@ __cpu_fifo_underrun_reporting_enabled(struct drm_i915_private *dev_priv, } /** - * intel_set_pch_fifo_underrun_reporting - enable/disable FIFO underrun messages - * @dev: drm device + * intel_set_pch_fifo_underrun_reporting - set PCH fifo underrun reporting state + * @dev_priv: i915 device instance * @pch_transcoder: the PCH transcoder (same as pipe on IVB and older) - * @enable: true if we want to report FIFO underrun errors, false otherwise + * @enable: whether underruns should be reported or not * * This function makes us disable or enable PCH fifo underruns for a specific * PCH transcoder. Notice that on some PCHs (e.g. CPT/PPT), disabling FIFO @@ -310,6 +340,15 @@ bool intel_set_pch_fifo_underrun_reporting(struct drm_i915_private *dev_priv, return old; } +/** + * intel_pch_fifo_underrun_irq_handler - handle PCH fifo underrun interrupt + * @dev_priv: i915 device instance + * @pipe: (CPU) pipe to set state for + * + * This handles a CPU fifo underrun interrupt, generating an underrun warning + * into dmesg if underrun reporting is enabled and then disables the underrun + * interrupt to avoid an irq storm. + */ void intel_cpu_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, enum pipe pipe) { @@ -323,6 +362,15 @@ void intel_cpu_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, pipe_name(pipe)); } +/** + * intel_pch_fifo_underrun_irq_handler - handle PCH fifo underrun interrupt + * @dev_priv: i915 device instance + * @pch_transcoder: the PCH transcoder (same as pipe on IVB and older) + * + * This handles a PCH fifo underrun interrupt, generating an underrun warning + * into dmesg if underrun reporting is enabled and then disables the underrun + * interrupt to avoid an irq storm. + */ void intel_pch_fifo_underrun_irq_handler(struct drm_i915_private *dev_priv, enum transcoder pch_transcoder) { -- cgit v0.10.2 From bcef6d5aa5746e6d3955f2dbc8e00d58d2c8adbd Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 1 Oct 2014 20:04:13 +0100 Subject: drm/i915: Use IS_HSW_ULT() in a HSW specific code path No need to add the BDW pci ULT/ULX checks inside a if (IS_HASWELL(dev)) code path. Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index bd7978c..02cee51 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -463,7 +463,7 @@ void intel_detect_pch(struct drm_device *dev) dev_priv->pch_type = PCH_LPT; DRM_DEBUG_KMS("Found LynxPoint PCH\n"); WARN_ON(!IS_HASWELL(dev)); - WARN_ON(IS_ULT(dev)); + WARN_ON(IS_HSW_ULT(dev)); } else if (IS_BROADWELL(dev)) { dev_priv->pch_type = PCH_LPT; dev_priv->pch_id = @@ -474,7 +474,7 @@ void intel_detect_pch(struct drm_device *dev) dev_priv->pch_type = PCH_LPT; DRM_DEBUG_KMS("Found LynxPoint LP PCH\n"); WARN_ON(!IS_HASWELL(dev)); - WARN_ON(!IS_ULT(dev)); + WARN_ON(!IS_HSW_ULT(dev)); } else if (id == INTEL_PCH_SPT_DEVICE_ID_TYPE) { dev_priv->pch_type = PCH_SPT; DRM_DEBUG_KMS("Found SunrisePoint PCH\n"); -- cgit v0.10.2 From dbf7786efabec4d1256cc0c049d1f7d7cccfd8bf Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 1 Oct 2014 20:04:14 +0100 Subject: drm/i915: Use IS_HSW_ULT() in HAS_IPS() HAS_IPS() has a '|| IS_BROADWELL()', no need to check for IS_BDW_ULT(). Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 26724c5..b3e4c2c 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2178,7 +2178,7 @@ struct drm_i915_cmd_table { #define HAS_PIPE_CXSR(dev) (INTEL_INFO(dev)->has_pipe_cxsr) #define HAS_FBC(dev) (INTEL_INFO(dev)->has_fbc) -#define HAS_IPS(dev) (IS_ULT(dev) || IS_BROADWELL(dev)) +#define HAS_IPS(dev) (IS_HSW_ULT(dev) || IS_BROADWELL(dev)) #define HAS_DDI(dev) (INTEL_INFO(dev)->has_ddi) #define HAS_FPGA_DBG_UNCLAIMED(dev) (INTEL_INFO(dev)->has_fpga_dbg) -- cgit v0.10.2 From cf404ce4b1de77017208e73c91cd6f09524678b7 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 1 Oct 2014 20:04:15 +0100 Subject: drm/i915: Spell out IS_HSW/BDW_ULT() in intel_crt_present() The quality of being a ULT or ULX package doesn't tell anything across generations and so a global IS_ULT() macro doesn't make much sense, esp. as we're adding new products. So, spell out which ULT/ULX SKUs we are talking about here, namely HSW and BDW. Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ee8156c..1fff5b2 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -12162,7 +12162,7 @@ static bool intel_crt_present(struct drm_device *dev) if (INTEL_INFO(dev)->gen >= 9) return false; - if (IS_ULT(dev)) + if (IS_HSW_ULT(dev) || IS_BDW_ULT(dev)) return false; if (IS_CHERRYVIEW(dev)) -- cgit v0.10.2 From 95626e7caac4f0dbe285c59d9702b9f55af497eb Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 1 Oct 2014 20:04:16 +0100 Subject: drm/i915: Use IS_HSW_ULT() in HSW CDCLK clock read-out hsw_get_cdclk_freq() is really just HSW, so we can use IS_HSW_ULT() instead of IS_ULT() there. Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index a151de7d..cb5367c 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1291,7 +1291,7 @@ static int hsw_get_cdclk_freq(struct drm_i915_private *dev_priv) return 450000; else if (freq == LCPLL_CLK_FREQ_450) return 450000; - else if (IS_ULT(dev)) + else if (IS_HSW_ULT(dev)) return 337500; else return 540000; -- cgit v0.10.2 From fa4dca2cce89ce8dc95f6dde58b3791feadc4283 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 1 Oct 2014 20:04:17 +0100 Subject: drm/i915/skl: Don't check for ULT/ULX when detecting the PCH IS_ULT() wasn't taking into account SKL so we had a warn with SPT-LP. We don't realy need those checks here, and as we don't need to introduce IS_SKL_ULT/ULX() at the moment, let's just drop them. Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 02cee51..a05a1d0 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -479,12 +479,10 @@ void intel_detect_pch(struct drm_device *dev) dev_priv->pch_type = PCH_SPT; DRM_DEBUG_KMS("Found SunrisePoint PCH\n"); WARN_ON(!IS_SKYLAKE(dev)); - WARN_ON(IS_ULT(dev)); } else if (id == INTEL_PCH_SPT_LP_DEVICE_ID_TYPE) { dev_priv->pch_type = PCH_SPT; DRM_DEBUG_KMS("Found SunrisePoint LP PCH\n"); WARN_ON(!IS_SKYLAKE(dev)); - WARN_ON(!IS_ULT(dev)); } else continue; -- cgit v0.10.2 From d68a08af3d112064b35c5d6eaa582379c5057fa8 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 1 Oct 2014 20:04:18 +0100 Subject: drm/i915: Remove IS_ULT() As stated in the few previous commits, IS_ULT/ULX() is better per-platform as it has different consequences depending on the platform. We now can get rid of it. Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index b3e4c2c..1e476b5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2107,7 +2107,6 @@ struct drm_i915_cmd_table { (INTEL_DEVID(dev) & 0x00F0) == 0x0020) #define IS_HSW_ULT(dev) (IS_HASWELL(dev) && \ (INTEL_DEVID(dev) & 0xFF00) == 0x0A00) -#define IS_ULT(dev) (IS_HSW_ULT(dev) || IS_BDW_ULT(dev)) #define IS_HSW_GT3(dev) (IS_HASWELL(dev) && \ (INTEL_DEVID(dev) & 0x00F0) == 0x0020) /* ULX machines are also considered ULT. */ -- cgit v0.10.2 From ccc759dc2a0214fd8b65ed4ebe78050874a67f94 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Wed, 24 Sep 2014 14:20:22 -0300 Subject: drm/i915: Merge of visible and !visible paths for primary planes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fold intel_pipe_set_base() in the update primary plane path merging pieces of code that are common to both paths. Basically the the pin/unpin procedures are the same for both paths and some checks can also be shared (some of the were moved to the check() stage) v2: take Ville's comments: - remove unnecessary plane check - move mutex lock to inside the conditional - make the pin fail message a debug one - add a fixme for the fastboot hack - call intel_frontbuffer_flip() after FBC update v3: take more Ville's comments: - fold update code under if (intel_crtc->active), and do the visible/!visible split inside. - check ret inside the same conditional we assign it v4: don't use intel_enable_primary_hw_plane(), the primary_enabled check inside will break page flips v5: take more Ville's comments: - set primary_enabled to true and add BDW hack - unify if (old_fb) and if (old_fb != fb) v6: take more Ville's comments: - make was_primary bool and fix its check - add the BDW vblank wait comment Suggested-by: Ville Syrjälä Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 1fff5b2..0760b60 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11671,12 +11671,23 @@ intel_check_primary_plane(struct drm_plane *plane, struct drm_rect *dest = &state->dst; struct drm_rect *src = &state->src; const struct drm_rect *clip = &state->clip; + int ret; - return drm_plane_helper_check_update(plane, crtc, fb, + ret = drm_plane_helper_check_update(plane, crtc, fb, src, dest, clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, false, true, &state->visible); + if (ret) + return ret; + + /* no fb bound */ + if (state->visible && !fb) { + DRM_ERROR("No FB bound\n"); + return -EINVAL; + } + + return 0; } static int @@ -11688,6 +11699,8 @@ intel_commit_primary_plane(struct drm_plane *plane, struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + enum pipe pipe = intel_crtc->pipe; + struct drm_framebuffer *old_fb = plane->fb; struct drm_i915_gem_object *obj = intel_fb_obj(fb); struct drm_i915_gem_object *old_obj = intel_fb_obj(plane->fb); struct intel_plane *intel_plane = to_intel_plane(plane); @@ -11696,76 +11709,100 @@ intel_commit_primary_plane(struct drm_plane *plane, intel_crtc_wait_for_pending_flips(crtc); - /* - * If clipping results in a non-visible primary plane, we'll disable - * the primary plane. Note that this is a bit different than what - * happens if userspace explicitly disables the plane by passing fb=0 - * because plane->fb still gets set and pinned. - */ - if (!state->visible) { + if (intel_crtc_has_pending_flip(crtc)) { + DRM_ERROR("pipe is still busy with an old pageflip\n"); + return -EBUSY; + } + + if (plane->fb != fb) { mutex_lock(&dev->struct_mutex); + ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); + if (ret == 0) + i915_gem_track_fb(old_obj, obj, + INTEL_FRONTBUFFER_PRIMARY(pipe)); + mutex_unlock(&dev->struct_mutex); + if (ret != 0) { + DRM_DEBUG_KMS("pin & fence failed\n"); + return ret; + } + } + + crtc->primary->fb = fb; + crtc->x = src->x1; + crtc->y = src->y1; + + intel_plane->crtc_x = state->orig_dst.x1; + intel_plane->crtc_y = state->orig_dst.y1; + intel_plane->crtc_w = drm_rect_width(&state->orig_dst); + intel_plane->crtc_h = drm_rect_height(&state->orig_dst); + intel_plane->src_x = state->orig_src.x1; + intel_plane->src_y = state->orig_src.y1; + intel_plane->src_w = drm_rect_width(&state->orig_src); + intel_plane->src_h = drm_rect_height(&state->orig_src); + intel_plane->obj = obj; + if (intel_crtc->active) { /* - * Try to pin the new fb first so that we can bail out if we - * fail. + * FBC does not work on some platforms for rotated + * planes, so disable it when rotation is not 0 and + * update it when rotation is set back to 0. + * + * FIXME: This is redundant with the fbc update done in + * the primary plane enable function except that that + * one is done too late. We eventually need to unify + * this. */ - if (plane->fb != fb) { - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); - if (ret) { - mutex_unlock(&dev->struct_mutex); - return ret; - } + if (intel_crtc->primary_enabled && + INTEL_INFO(dev)->gen <= 4 && !IS_G4X(dev) && + dev_priv->fbc.plane == intel_crtc->plane && + intel_plane->rotation != BIT(DRM_ROTATE_0)) { + intel_disable_fbc(dev); } - i915_gem_track_fb(old_obj, obj, - INTEL_FRONTBUFFER_PRIMARY(intel_crtc->pipe)); - - if (intel_crtc->primary_enabled) - intel_disable_primary_hw_plane(plane, crtc); + if (state->visible) { + bool was_enabled = intel_crtc->primary_enabled; + /* FIXME: kill this fastboot hack */ + intel_update_pipe_size(intel_crtc); - if (plane->fb != fb) - if (plane->fb) - intel_unpin_fb_obj(old_obj); + intel_crtc->primary_enabled = true; - mutex_unlock(&dev->struct_mutex); + dev_priv->display.update_primary_plane(crtc, plane->fb, + crtc->x, crtc->y); - } else { - if (intel_crtc && intel_crtc->active && - intel_crtc->primary_enabled) { /* - * FBC does not work on some platforms for rotated - * planes, so disable it when rotation is not 0 and - * update it when rotation is set back to 0. - * - * FIXME: This is redundant with the fbc update done in - * the primary plane enable function except that that - * one is done too late. We eventually need to unify - * this. + * BDW signals flip done immediately if the plane + * is disabled, even if the plane enable is already + * armed to occur at the next vblank :( */ - if (INTEL_INFO(dev)->gen <= 4 && !IS_G4X(dev) && - dev_priv->fbc.plane == intel_crtc->plane && - intel_plane->rotation != BIT(DRM_ROTATE_0)) { - intel_disable_fbc(dev); - } + if (IS_BROADWELL(dev) && !was_enabled) + intel_wait_for_vblank(dev, intel_crtc->pipe); + } else { + /* + * If clipping results in a non-visible primary plane, + * we'll disable the primary plane. Note that this is + * a bit different than what happens if userspace + * explicitly disables the plane by passing fb=0 + * because plane->fb still gets set and pinned. + */ + intel_disable_primary_hw_plane(plane, crtc); } - ret = intel_pipe_set_base(crtc, src->x1, src->y1, fb); - if (ret) - return ret; - if (!intel_crtc->primary_enabled) - intel_enable_primary_hw_plane(plane, crtc); + intel_frontbuffer_flip(dev, INTEL_FRONTBUFFER_PRIMARY(pipe)); + + mutex_lock(&dev->struct_mutex); + intel_update_fbc(dev); + mutex_unlock(&dev->struct_mutex); } - intel_plane->crtc_x = state->orig_dst.x1; - intel_plane->crtc_y = state->orig_dst.y1; - intel_plane->crtc_w = drm_rect_width(&state->orig_dst); - intel_plane->crtc_h = drm_rect_height(&state->orig_dst); - intel_plane->src_x = state->orig_src.x1; - intel_plane->src_y = state->orig_src.y1; - intel_plane->src_w = drm_rect_width(&state->orig_src); - intel_plane->src_h = drm_rect_height(&state->orig_src); - intel_plane->obj = obj; + if (old_fb && old_fb != fb) { + if (intel_crtc->active) + intel_wait_for_vblank(dev, intel_crtc->pipe); + + mutex_lock(&dev->struct_mutex); + intel_unpin_fb_obj(old_obj); + mutex_unlock(&dev->struct_mutex); + } return 0; } -- cgit v0.10.2 From 5ee446383901791dc2917227de5a94ddcfadd761 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Wed, 24 Sep 2014 14:20:23 -0300 Subject: drm/i915: remove leftover from pre-universal planes days MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Now that universal planes are in place we don't need this plane unref on failures. Suggested-by: Ville Syrjälä Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 0760b60..8b85aa6 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8455,13 +8455,6 @@ static bool cursor_size_ok(struct drm_device *dev, return true; } -/* - * intel_crtc_cursor_set_obj - Set cursor to specified GEM object - * - * Note that the object's reference will be consumed if the update fails. If - * the update succeeds, the reference of the old object (if any) will be - * consumed. - */ static int intel_crtc_cursor_set_obj(struct drm_crtc *crtc, struct drm_i915_gem_object *obj, uint32_t width, uint32_t height) @@ -8491,8 +8484,7 @@ static int intel_crtc_cursor_set_obj(struct drm_crtc *crtc, stride = roundup_pow_of_two(width) * 4; if (obj->base.size < stride * height) { DRM_DEBUG_KMS("buffer is too small\n"); - ret = -ENOMEM; - goto fail; + return -ENOMEM; } /* we only need to pin inside GTT if cursor is non-phy */ @@ -8581,8 +8573,6 @@ fail_unpin: i915_gem_object_unpin_from_display_plane(obj); fail_locked: mutex_unlock(&dev->struct_mutex); -fail: - drm_gem_object_unreference_unlocked(&obj->base); return ret; } -- cgit v0.10.2 From 757f9a3e5b8a812af0c213099a5b31cb423f4d3c Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Wed, 24 Sep 2014 14:20:24 -0300 Subject: drm/i915: move check of intel_crtc_cursor_set_obj() out MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Move check inside intel_crtc_cursor_set_obj() to intel_check_cursor_plane(), we only use it there so move them out to make the merge of intel_crtc_cursor_set_obj() into intel_check_cursor_plane() easier. This is another step toward the atomic modesetting support and unification of plane operations such pin/unpin of fb objects on i915. v2: take Ville's comment: move crtc_{w,h} assignment a bit down in the code v3: take Ville's comment: kept only the restructuring changes, the rest of the code was moved to a separated patch since it is a bug fix (we weren't checking sizes when the fb was the same) Signed-off-by: Gustavo Padovan [danvet: Fixup commit message mixup.] Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 8b85aa6..81f5656 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8463,7 +8463,7 @@ static int intel_crtc_cursor_set_obj(struct drm_crtc *crtc, struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); enum pipe pipe = intel_crtc->pipe; - unsigned old_width, stride; + unsigned old_width; uint32_t addr; int ret; @@ -8475,29 +8475,11 @@ static int intel_crtc_cursor_set_obj(struct drm_crtc *crtc, goto finish; } - /* Check for which cursor types we support */ - if (!cursor_size_ok(dev, width, height)) { - DRM_DEBUG("Cursor dimension not supported\n"); - return -EINVAL; - } - - stride = roundup_pow_of_two(width) * 4; - if (obj->base.size < stride * height) { - DRM_DEBUG_KMS("buffer is too small\n"); - return -ENOMEM; - } - /* we only need to pin inside GTT if cursor is non-phy */ mutex_lock(&dev->struct_mutex); if (!INTEL_INFO(dev)->cursor_needs_physical) { unsigned alignment; - if (obj->tiling_mode) { - DRM_DEBUG_KMS("cursor cannot be tiled\n"); - ret = -EINVAL; - goto fail_locked; - } - /* * Global gtt pte registers are special registers which actually * forward writes to a chunk of system memory. Which means that @@ -11918,16 +11900,55 @@ intel_check_cursor_plane(struct drm_plane *plane, struct intel_plane_state *state) { struct drm_crtc *crtc = state->crtc; + struct drm_device *dev = crtc->dev; struct drm_framebuffer *fb = state->fb; struct drm_rect *dest = &state->dst; struct drm_rect *src = &state->src; const struct drm_rect *clip = &state->clip; + struct drm_i915_gem_object *obj = intel_fb_obj(fb); + int crtc_w, crtc_h; + unsigned stride; + int ret; - return drm_plane_helper_check_update(plane, crtc, fb, + ret = drm_plane_helper_check_update(plane, crtc, fb, src, dest, clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, true, true, &state->visible); + if (ret) + return ret; + + + /* if we want to turn off the cursor ignore width and height */ + if (!obj) + return 0; + + if (fb == crtc->cursor->fb) + return 0; + + /* Check for which cursor types we support */ + crtc_w = drm_rect_width(&state->orig_dst); + crtc_h = drm_rect_height(&state->orig_dst); + if (!cursor_size_ok(dev, crtc_w, crtc_h)) { + DRM_DEBUG("Cursor dimension not supported\n"); + return -EINVAL; + } + + stride = roundup_pow_of_two(crtc_w) * 4; + if (obj->base.size < stride * crtc_h) { + DRM_DEBUG_KMS("buffer is too small\n"); + return -ENOMEM; + } + + /* we only need to pin inside GTT if cursor is non-phy */ + mutex_lock(&dev->struct_mutex); + if (!INTEL_INFO(dev)->cursor_needs_physical && obj->tiling_mode) { + DRM_DEBUG_KMS("cursor cannot be tiled\n"); + ret = -EINVAL; + } + mutex_unlock(&dev->struct_mutex); + + return ret; } static int -- cgit v0.10.2 From e391ea882b1a04fb3f559287ac694652a3cd9da9 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Wed, 24 Sep 2014 14:20:25 -0300 Subject: drm/i915: Fix not checking cursor and object sizes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Even if the fb is the same we should still check if the sizes are valid to be set. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 81f5656..f150765 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11923,9 +11923,6 @@ intel_check_cursor_plane(struct drm_plane *plane, if (!obj) return 0; - if (fb == crtc->cursor->fb) - return 0; - /* Check for which cursor types we support */ crtc_w = drm_rect_width(&state->orig_dst); crtc_h = drm_rect_height(&state->orig_dst); @@ -11940,6 +11937,9 @@ intel_check_cursor_plane(struct drm_plane *plane, return -ENOMEM; } + if (fb == crtc->cursor->fb) + return 0; + /* we only need to pin inside GTT if cursor is non-phy */ mutex_lock(&dev->struct_mutex); if (!INTEL_INFO(dev)->cursor_needs_physical && obj->tiling_mode) { -- cgit v0.10.2 From 58abf1daae8264e5adc90733dcabeb48a8245439 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 7 Oct 2014 07:06:50 -0700 Subject: drm/i915: Do not export RC6p and RC6pp if they don't exist Avoid to expose RC6 and RC6pp to the platforms that doesn't support it. So powertop can be changed to show RC6p and RC6pp only on the platforms they are available. v2: Simplify by merging RC6p and RC6pp groups and respect the spec that mentions deep and deepest RC6 on SNB and IVB although they keep disabled by default. v3: Remove unecessary space. v4: RC6p and RC6pp is only for SNB and IVB; unify debug msg and use has_rc6p() on sanitize options instead of is gen 6 and ivb. v5: yet another fix on has_rc6p macro. final is_gen6 or is_ivb! To make sure we are excluding hsw and baytrail. References: https://bugs.freedesktop.org/show_bug.cgi?id=84524 Cc: Josh Triplett Cc: Paulo Zanoni Cc: Daniel Vetter Signed-off-by: Rodrigo Vivi Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 1e476b5..ac6232b 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2184,6 +2184,8 @@ struct drm_i915_cmd_table { #define HAS_PSR(dev) (IS_HASWELL(dev) || IS_BROADWELL(dev)) #define HAS_RUNTIME_PM(dev) (IS_GEN6(dev) || IS_HASWELL(dev) || \ IS_BROADWELL(dev) || IS_VALLEYVIEW(dev)) +#define HAS_RC6(dev) (INTEL_INFO(dev)->gen >= 6) +#define HAS_RC6p(dev) (INTEL_INFO(dev)->gen == 6 || IS_IVYBRIDGE(dev)) #define INTEL_PCH_DEVICE_ID_MASK 0xff00 #define INTEL_PCH_IBX_DEVICE_ID_TYPE 0x3b00 diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c index 503847f..4a5af69 100644 --- a/drivers/gpu/drm/i915/i915_sysfs.c +++ b/drivers/gpu/drm/i915/i915_sysfs.c @@ -139,8 +139,6 @@ static DEVICE_ATTR(rc6pp_residency_ms, S_IRUGO, show_rc6pp_ms, NULL); static struct attribute *rc6_attrs[] = { &dev_attr_rc6_enable.attr, &dev_attr_rc6_residency_ms.attr, - &dev_attr_rc6p_residency_ms.attr, - &dev_attr_rc6pp_residency_ms.attr, NULL }; @@ -148,6 +146,17 @@ static struct attribute_group rc6_attr_group = { .name = power_group_name, .attrs = rc6_attrs }; + +static struct attribute *rc6p_attrs[] = { + &dev_attr_rc6p_residency_ms.attr, + &dev_attr_rc6pp_residency_ms.attr, + NULL +}; + +static struct attribute_group rc6p_attr_group = { + .name = power_group_name, + .attrs = rc6p_attrs +}; #endif static int l3_access_valid(struct drm_device *dev, loff_t offset) @@ -595,12 +604,18 @@ void i915_setup_sysfs(struct drm_device *dev) int ret; #ifdef CONFIG_PM - if (INTEL_INFO(dev)->gen >= 6) { + if (HAS_RC6(dev)) { ret = sysfs_merge_group(&dev->primary->kdev->kobj, &rc6_attr_group); if (ret) DRM_ERROR("RC6 residency sysfs setup failed\n"); } + if (HAS_RC6p(dev)) { + ret = sysfs_merge_group(&dev->primary->kdev->kobj, + &rc6p_attr_group); + if (ret) + DRM_ERROR("RC6p residency sysfs setup failed\n"); + } #endif if (HAS_L3_DPF(dev)) { ret = device_create_bin_file(dev->primary->kdev, &dpf_attrs); @@ -640,5 +655,6 @@ void i915_teardown_sysfs(struct drm_device *dev) device_remove_bin_file(dev->primary->kdev, &dpf_attrs); #ifdef CONFIG_PM sysfs_unmerge_group(&dev->primary->kdev->kobj, &rc6_attr_group); + sysfs_unmerge_group(&dev->primary->kdev->kobj, &rc6p_attr_group); #endif } diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index a14be5d..daa99e7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3629,10 +3629,15 @@ static void intel_print_rc6_info(struct drm_device *dev, u32 mode) else mode = 0; } - DRM_DEBUG_KMS("Enabling RC6 states: RC6 %s, RC6p %s, RC6pp %s\n", - (mode & GEN6_RC_CTL_RC6_ENABLE) ? "on" : "off", - (mode & GEN6_RC_CTL_RC6p_ENABLE) ? "on" : "off", - (mode & GEN6_RC_CTL_RC6pp_ENABLE) ? "on" : "off"); + if (HAS_RC6p(dev)) + DRM_DEBUG_KMS("Enabling RC6 states: RC6 %s RC6p %s RC6pp %s\n", + (mode & GEN6_RC_CTL_RC6_ENABLE) ? "on" : "off", + (mode & GEN6_RC_CTL_RC6p_ENABLE) ? "on" : "off", + (mode & GEN6_RC_CTL_RC6pp_ENABLE) ? "on" : "off"); + + else + DRM_DEBUG_KMS("Enabling RC6 states: RC6 %s\n", + (mode & GEN6_RC_CTL_RC6_ENABLE) ? "on" : "off"); } static int sanitize_rc6_option(const struct drm_device *dev, int enable_rc6) @@ -3649,7 +3654,7 @@ static int sanitize_rc6_option(const struct drm_device *dev, int enable_rc6) if (enable_rc6 >= 0) { int mask; - if (INTEL_INFO(dev)->gen == 6 || IS_IVYBRIDGE(dev)) + if (HAS_RC6p(dev)) mask = INTEL_RC6_ENABLE | INTEL_RC6p_ENABLE | INTEL_RC6pp_ENABLE; else -- cgit v0.10.2 From 43dc52c3b43af65084d6399c4f4b723fa2374971 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 7 Oct 2014 17:41:20 +0300 Subject: drm/i915: Add missing '\n' to cdclk debug message MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index f150765..d6fc469 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -4618,7 +4618,7 @@ static void vlv_update_cdclk(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; dev_priv->vlv_cdclk_freq = dev_priv->display.get_display_clock_speed(dev); - DRM_DEBUG_DRIVER("Current CD clock rate: %d kHz", + DRM_DEBUG_DRIVER("Current CD clock rate: %d kHz\n", dev_priv->vlv_cdclk_freq); /* -- cgit v0.10.2 From 005445c5fb3a96026d6bd9580339ea8580eb6c6d Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 8 Oct 2014 11:25:16 +0100 Subject: drm/i915: Report the current number of bytes freed during oom The shrinker reports the number of pages freed, but we try to log the number of bytes - which leads to some nonsense values being reportedly freed during oom. Reported-by: Maarten Lankhorst Signed-off-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 2719c25..dd25a54 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5259,7 +5259,7 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr) struct drm_device *dev = dev_priv->dev; struct drm_i915_gem_object *obj; unsigned long timeout = msecs_to_jiffies(5000) + 1; - unsigned long pinned, bound, unbound, freed; + unsigned long pinned, bound, unbound, freed_pages; bool was_interruptible; bool unlock; @@ -5276,7 +5276,7 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr) was_interruptible = dev_priv->mm.interruptible; dev_priv->mm.interruptible = false; - freed = i915_gem_shrink_all(dev_priv); + freed_pages = i915_gem_shrink_all(dev_priv); dev_priv->mm.interruptible = was_interruptible; @@ -5308,13 +5308,13 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr) mutex_unlock(&dev->struct_mutex); pr_info("Purging GPU memory, %lu bytes freed, %lu bytes still pinned.\n", - freed, pinned); + freed_pages << PAGE_SHIFT, pinned); if (unbound || bound) pr_err("%lu and %lu bytes still available in the " "bound and unbound GPU page lists.\n", bound, unbound); - *(unsigned long *)ptr += freed; + *(unsigned long *)ptr += freed_pages; return NOTIFY_DONE; } -- cgit v0.10.2 From bb9059d3a034d56f57e037d7866afeef18876283 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 8 Oct 2014 11:25:17 +0100 Subject: drm/i915: Suppress no action noise from oom shrinker If we are not able to free anything (the shrinker leaves nothing on the global object lists), do not log anything. This is useful when other subsystems are being stress-tested for their oom behaviour and i915.ko is shouting into the logs about doing nothing. Reported-by: Dave Jones Signed-off-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index dd25a54..e9c783d 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5307,8 +5307,9 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr) if (unlock) mutex_unlock(&dev->struct_mutex); - pr_info("Purging GPU memory, %lu bytes freed, %lu bytes still pinned.\n", - freed_pages << PAGE_SHIFT, pinned); + if (freed_pages || unbound || bound) + pr_info("Purging GPU memory, %lu bytes freed, %lu bytes still pinned.\n", + freed_pages << PAGE_SHIFT, pinned); if (unbound || bound) pr_err("%lu and %lu bytes still available in the " "bound and unbound GPU page lists.\n", -- cgit v0.10.2 From 1447dde094c1ebe90289949b4735f8d6ed2f2c08 Mon Sep 17 00:00:00 2001 From: Sonika Jindal Date: Sat, 4 Oct 2014 10:53:31 +0100 Subject: drm/i915/skl: Add 180 degree HW rotation support Add support for 180 degree rotation for primary and sprite planes Signed-off-by: Sonika Jindal Reviewed-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index a56d9a7..6db369a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4579,6 +4579,9 @@ enum punit_power_well { #define PLANE_CTL_ALPHA_DISABLE ( 0 << 4) #define PLANE_CTL_ALPHA_SW_PREMULTIPLY ( 2 << 4) #define PLANE_CTL_ALPHA_HW_PREMULTIPLY ( 3 << 4) +#define PLANE_CTL_ROTATE_MASK 0x3 +#define PLANE_CTL_ROTATE_0 0x0 +#define PLANE_CTL_ROTATE_180 0x2 #define _PLANE_STRIDE_1_A 0x70188 #define _PLANE_STRIDE_2_A 0x70288 #define _PLANE_STRIDE_3_A 0x70388 diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index d6fc469..6e6f150 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2694,6 +2694,8 @@ static void skylake_update_primary_plane(struct drm_crtc *crtc, } plane_ctl |= PLANE_CTL_PLANE_GAMMA_DISABLE; + if (to_intel_plane(crtc->primary)->rotation == BIT(DRM_ROTATE_180)) + plane_ctl |= PLANE_CTL_ROTATE_180; I915_WRITE(PLANE_CTL(pipe, 0), plane_ctl); diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 750b634..2c060ad 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -162,6 +162,7 @@ skl_update_plane(struct drm_plane *drm_plane, struct drm_crtc *crtc, plane_ctl &= ~PLANE_CTL_YUV422_ORDER_MASK; plane_ctl &= ~PLANE_CTL_TILED_MASK; plane_ctl &= ~PLANE_CTL_ALPHA_MASK; + plane_ctl &= ~PLANE_CTL_ROTATE_MASK; /* Trickle feed has to be enabled */ plane_ctl &= ~PLANE_CTL_TRICKLE_FEED_DISABLE; @@ -217,6 +218,8 @@ skl_update_plane(struct drm_plane *drm_plane, struct drm_crtc *crtc, default: BUG(); } + if (intel_plane->rotation == BIT(DRM_ROTATE_180)) + plane_ctl |= PLANE_CTL_ROTATE_180; plane_ctl |= PLANE_CTL_ENABLE; plane_ctl |= PLANE_CTL_PIPE_CSC_ENABLE; -- cgit v0.10.2 From f2fbc690aee054d3e29044d66518335fe03069ca Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 24 Aug 2014 19:35:31 +0100 Subject: drm/i915: Convert a couple more INTEL_INFO-esque macros to be pointer agnostic Just a couple more macros that assume that they were being passed a struct drm_device when they want a struct drm_i915_private. Use our magic macro to ease transitioning over to using drm_i915_privates Signed-off-by: Chris Wilson Reviewed-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index ac6232b..1d6f298 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2140,7 +2140,7 @@ struct drm_i915_cmd_table { #define HAS_VEBOX(dev) (INTEL_INFO(dev)->ring_mask & VEBOX_RING) #define HAS_LLC(dev) (INTEL_INFO(dev)->has_llc) #define HAS_WT(dev) ((IS_HASWELL(dev) || IS_BROADWELL(dev)) && \ - to_i915(dev)->ellc_size) + __I915__(dev)->ellc_size) #define I915_NEED_GFX_HWS(dev) (INTEL_INFO(dev)->need_gfx_hws) #define HAS_HW_CONTEXTS(dev) (INTEL_INFO(dev)->gen >= 6) @@ -2196,7 +2196,7 @@ struct drm_i915_cmd_table { #define INTEL_PCH_SPT_DEVICE_ID_TYPE 0xA100 #define INTEL_PCH_SPT_LP_DEVICE_ID_TYPE 0x9D00 -#define INTEL_PCH_TYPE(dev) (to_i915(dev)->pch_type) +#define INTEL_PCH_TYPE(dev) (__I915__(dev)->pch_type) #define HAS_PCH_SPT(dev) (INTEL_PCH_TYPE(dev) == PCH_SPT) #define HAS_PCH_LPT(dev) (INTEL_PCH_TYPE(dev) == PCH_LPT) #define HAS_PCH_CPT(dev) (INTEL_PCH_TYPE(dev) == PCH_CPT) -- cgit v0.10.2 From 101b376d358e2f724db5e0ac4d207079b16c4754 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Thu, 9 Oct 2014 07:11:47 -0700 Subject: drm/i915/bdw: Remove BDW preproduction W/As until C stepping. Let's clean this a bit v2: Rebase after other Mika's patch that removed some BDW production workarounds. v3: Removed stepping info. Reviewed-by: Mika Kuoppala Signed-off-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index daa99e7..23d3318 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5654,16 +5654,6 @@ static void broadwell_init_clock_gating(struct drm_device *dev) I915_WRITE(WM2_LP_ILK, 0); I915_WRITE(WM1_LP_ILK, 0); - /* FIXME(BDW): Check all the w/a, some might only apply to - * pre-production hw. */ - - - I915_WRITE(GAMTARBMODE, _MASKED_BIT_ENABLE(ARB_MODE_BWGTLB_DISABLE)); - - I915_WRITE(_3D_CHICKEN3, - _MASKED_BIT_ENABLE(_3D_CHICKEN_SDE_LIMIT_FIFO_POLY_DEPTH(2))); - - /* WaSwitchSolVfFArbitrationPriority:bdw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 7c0c28c..5ebe46a 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -712,13 +712,12 @@ static int bdw_init_workarounds(struct intel_engine_cs *ring) return ret; /* WaDisablePartialInstShootdown:bdw */ - /* WaDisableThreadStallDopClockGating:bdw */ - /* FIXME: Unclear whether we really need this on production bdw. */ + /* WaDisableThreadStallDopClockGating:bdw (pre-production) */ intel_ring_emit_wa(ring, GEN8_ROW_CHICKEN, _MASKED_BIT_ENABLE(PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE | STALL_DOP_GATING_DISABLE)); - /* WaDisableDopClockGating:bdw May not be needed for production */ + /* WaDisableDopClockGating:bdw */ intel_ring_emit_wa(ring, GEN7_ROW_CHICKEN2, _MASKED_BIT_ENABLE(DOP_CLOCK_GATING_DISABLE)); -- cgit v0.10.2 From 343673818098b80e5adca6a93d4e112d907cd293 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Thu, 16 Oct 2014 12:28:18 +0200 Subject: drm/i915: Document that mmap forwarding is discouraged Too many new drm driver writers seem to look at i915 for inspiration. But we have two ways to do mmap, so discourage readers from the old, ugly version. In a new driver we'd just expose two mmap offsets per object, one for the gtt map and the other for the cpu map. v2: Make it clear that i915 does cpu mmaps this way for past cluelessness^W^W historical reasons. Asked for by Jani. Cc: "Cheng, Yao" Cc: David Herrmann Reviewed-by: David Herrmann Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index e9c783d..895f988 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1466,6 +1466,16 @@ unlock: * * While the mapping holds a reference on the contents of the object, it doesn't * imply a ref on the object itself. + * + * IMPORTANT: + * + * DRM driver writers who look a this function as an example for how to do GEM + * mmap support, please don't implement mmap support like here. The modern way + * to implement DRM mmap support is with an mmap offset ioctl (like + * i915_gem_mmap_gtt) and then using the mmap syscall on the DRM fd directly. + * That way debug tooling like valgrind will understand what's going on, hiding + * the mmap call in a driver private ioctl will break that. The i915 driver only + * does cpu mmaps this way because we didn't know better. */ int i915_gem_mmap_ioctl(struct drm_device *dev, void *data, -- cgit v0.10.2 From 6e2cc0963a962aad91184eaabcf67a106e80e815 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Mon, 20 Oct 2014 13:46:42 +0300 Subject: drm/i915: Replace some loop through encoders with intel_pipe_has_type() In the ironlake mode set code, there was two instances of a loop through encoders to find out if one of them has INTEL_OUTPUT_LVDS type. Simplify the code by deleting some lines and use intel_pipe_has_type() instead. Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 6e6f150..79d9944 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7110,18 +7110,11 @@ static bool ironlake_compute_clocks(struct drm_crtc *crtc, { struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_encoder *intel_encoder; int refclk; const intel_limit_t *limit; bool ret, is_lvds = false; - for_each_encoder_on_crtc(dev, crtc, intel_encoder) { - switch (intel_encoder->type) { - case INTEL_OUTPUT_LVDS: - is_lvds = true; - break; - } - } + is_lvds = intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS); refclk = ironlake_get_refclk(crtc); @@ -7261,23 +7254,13 @@ static int ironlake_crtc_mode_set(struct drm_crtc *crtc, { struct drm_device *dev = crtc->dev; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int num_connectors = 0; intel_clock_t clock, reduced_clock; u32 dpll = 0, fp = 0, fp2 = 0; bool ok, has_reduced_clock = false; bool is_lvds = false; - struct intel_encoder *encoder; struct intel_shared_dpll *pll; - for_each_encoder_on_crtc(dev, crtc, encoder) { - switch (encoder->type) { - case INTEL_OUTPUT_LVDS: - is_lvds = true; - break; - } - - num_connectors++; - } + is_lvds = intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS); WARN(!(HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev)), "Unexpected PCH type %d\n", INTEL_PCH_TYPE(dev)); -- cgit v0.10.2 From a919ff14e606a7a2aca62259500158413c733fe8 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Mon, 20 Oct 2014 13:46:43 +0300 Subject: drm/i915: Make *_find_best_dpll() take an intel_crtc insted of drm_crtc For consistency, since that's the rule followed for internal functions. Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 1d6f298..3a8d0cb 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -460,7 +460,7 @@ struct drm_i915_display_funcs { * Returns true on success, false on failure. */ bool (*find_dpll)(const struct intel_limit *limit, - struct drm_crtc *crtc, + struct intel_crtc *crtc, int target, int refclk, struct dpll *match_clock, struct dpll *best_clock); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 79d9944..eed3894 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -576,15 +576,15 @@ static bool intel_PLL_is_valid(struct drm_device *dev, } static bool -i9xx_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, +i9xx_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, int target, int refclk, intel_clock_t *match_clock, intel_clock_t *best_clock) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; intel_clock_t clock; int err = target; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS)) { /* * For LVDS just rely on its current settings for dual-channel. * We haven't figured out how to reliably set up different @@ -637,15 +637,15 @@ i9xx_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, } static bool -pnv_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, +pnv_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, int target, int refclk, intel_clock_t *match_clock, intel_clock_t *best_clock) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; intel_clock_t clock; int err = target; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS)) { /* * For LVDS just rely on its current settings for dual-channel. * We haven't figured out how to reliably set up different @@ -696,11 +696,11 @@ pnv_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, } static bool -g4x_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, +g4x_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, int target, int refclk, intel_clock_t *match_clock, intel_clock_t *best_clock) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; intel_clock_t clock; int max_n; bool found; @@ -708,7 +708,7 @@ g4x_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, int err_most = (target >> 8) + (target >> 9); found = false; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS)) { if (intel_is_dual_link_lvds(dev)) clock.p2 = limit->p2.p2_fast; else @@ -753,11 +753,11 @@ g4x_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, } static bool -vlv_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, +vlv_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, int target, int refclk, intel_clock_t *match_clock, intel_clock_t *best_clock) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; intel_clock_t clock; unsigned int bestppm = 1000000; /* min update 19.2 MHz */ @@ -810,11 +810,11 @@ vlv_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, } static bool -chv_find_best_dpll(const intel_limit_t *limit, struct drm_crtc *crtc, +chv_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, int target, int refclk, intel_clock_t *match_clock, intel_clock_t *best_clock) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; intel_clock_t clock; uint64_t m2; int found = false; @@ -6284,7 +6284,7 @@ static int i9xx_crtc_mode_set(struct drm_crtc *crtc, * 2) / p1 / p2. */ limit = intel_limit(crtc, refclk); - ok = dev_priv->display.find_dpll(limit, crtc, + ok = dev_priv->display.find_dpll(limit, intel_crtc, intel_crtc->config.port_clock, refclk, NULL, &clock); if (!ok) { @@ -6300,7 +6300,7 @@ static int i9xx_crtc_mode_set(struct drm_crtc *crtc, * we will disable the LVDS downclock feature. */ has_reduced_clock = - dev_priv->display.find_dpll(limit, crtc, + dev_priv->display.find_dpll(limit, intel_crtc, dev_priv->lvds_downclock, refclk, &clock, &reduced_clock); @@ -7110,6 +7110,7 @@ static bool ironlake_compute_clocks(struct drm_crtc *crtc, { struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); int refclk; const intel_limit_t *limit; bool ret, is_lvds = false; @@ -7124,8 +7125,8 @@ static bool ironlake_compute_clocks(struct drm_crtc *crtc, * reflck * (5 * (m1 + 2) + (m2 + 2)) / (n + 2) / p1 / p2. */ limit = intel_limit(crtc, refclk); - ret = dev_priv->display.find_dpll(limit, crtc, - to_intel_crtc(crtc)->config.port_clock, + ret = dev_priv->display.find_dpll(limit, intel_crtc, + intel_crtc->config.port_clock, refclk, NULL, clock); if (!ret) return false; @@ -7138,7 +7139,7 @@ static bool ironlake_compute_clocks(struct drm_crtc *crtc, * downclock feature. */ *has_reduced_clock = - dev_priv->display.find_dpll(limit, crtc, + dev_priv->display.find_dpll(limit, intel_crtc, dev_priv->lvds_downclock, refclk, clock, reduced_clock); -- cgit v0.10.2 From c7653199c01a2499ea134ead0fe03f2451c28518 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Mon, 20 Oct 2014 13:46:44 +0300 Subject: drm/i915: Make *_crtc_mode_set() take an intel_crtc insted of drm_crtc For consistency, since that's the rule followed for internal functions. Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 3a8d0cb..9f3d689 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -476,7 +476,7 @@ struct drm_i915_display_funcs { struct intel_crtc_config *); void (*get_plane_config)(struct intel_crtc *, struct intel_plane_config *); - int (*crtc_mode_set)(struct drm_crtc *crtc, + int (*crtc_mode_set)(struct intel_crtc *crtc, int x, int y, struct drm_framebuffer *old_fb); void (*crtc_enable)(struct drm_crtc *crtc); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index eed3894..b9f0f38 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -6244,13 +6244,12 @@ static void i9xx_set_pipeconf(struct intel_crtc *intel_crtc) POSTING_READ(PIPECONF(intel_crtc->pipe)); } -static int i9xx_crtc_mode_set(struct drm_crtc *crtc, +static int i9xx_crtc_mode_set(struct intel_crtc *crtc, int x, int y, struct drm_framebuffer *fb) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); int refclk, num_connectors = 0; intel_clock_t clock, reduced_clock; bool ok, has_reduced_clock = false; @@ -6258,7 +6257,7 @@ static int i9xx_crtc_mode_set(struct drm_crtc *crtc, struct intel_encoder *encoder; const intel_limit_t *limit; - for_each_encoder_on_crtc(dev, crtc, encoder) { + for_each_encoder_on_crtc(dev, &crtc->base, encoder) { switch (encoder->type) { case INTEL_OUTPUT_LVDS: is_lvds = true; @@ -6274,8 +6273,8 @@ static int i9xx_crtc_mode_set(struct drm_crtc *crtc, if (is_dsi) return 0; - if (!intel_crtc->config.clock_set) { - refclk = i9xx_get_refclk(crtc, num_connectors); + if (!crtc->config.clock_set) { + refclk = i9xx_get_refclk(&crtc->base, num_connectors); /* * Returns a set of divisors for the desired target clock with @@ -6283,9 +6282,9 @@ static int i9xx_crtc_mode_set(struct drm_crtc *crtc, * the clock equation: reflck * (5 * (m1 + 2) + (m2 + 2)) / (n + * 2) / p1 / p2. */ - limit = intel_limit(crtc, refclk); - ok = dev_priv->display.find_dpll(limit, intel_crtc, - intel_crtc->config.port_clock, + limit = intel_limit(&crtc->base, refclk); + ok = dev_priv->display.find_dpll(limit, crtc, + crtc->config.port_clock, refclk, NULL, &clock); if (!ok) { DRM_ERROR("Couldn't find PLL settings for mode!\n"); @@ -6300,29 +6299,29 @@ static int i9xx_crtc_mode_set(struct drm_crtc *crtc, * we will disable the LVDS downclock feature. */ has_reduced_clock = - dev_priv->display.find_dpll(limit, intel_crtc, + dev_priv->display.find_dpll(limit, crtc, dev_priv->lvds_downclock, refclk, &clock, &reduced_clock); } /* Compat-code for transition, will disappear. */ - intel_crtc->config.dpll.n = clock.n; - intel_crtc->config.dpll.m1 = clock.m1; - intel_crtc->config.dpll.m2 = clock.m2; - intel_crtc->config.dpll.p1 = clock.p1; - intel_crtc->config.dpll.p2 = clock.p2; + crtc->config.dpll.n = clock.n; + crtc->config.dpll.m1 = clock.m1; + crtc->config.dpll.m2 = clock.m2; + crtc->config.dpll.p1 = clock.p1; + crtc->config.dpll.p2 = clock.p2; } if (IS_GEN2(dev)) { - i8xx_update_pll(intel_crtc, + i8xx_update_pll(crtc, has_reduced_clock ? &reduced_clock : NULL, num_connectors); } else if (IS_CHERRYVIEW(dev)) { - chv_update_pll(intel_crtc); + chv_update_pll(crtc); } else if (IS_VALLEYVIEW(dev)) { - vlv_update_pll(intel_crtc); + vlv_update_pll(crtc); } else { - i9xx_update_pll(intel_crtc, + i9xx_update_pll(crtc, has_reduced_clock ? &reduced_clock : NULL, num_connectors); } @@ -7249,68 +7248,67 @@ static uint32_t ironlake_compute_dpll(struct intel_crtc *intel_crtc, return dpll | DPLL_VCO_ENABLE; } -static int ironlake_crtc_mode_set(struct drm_crtc *crtc, +static int ironlake_crtc_mode_set(struct intel_crtc *crtc, int x, int y, struct drm_framebuffer *fb) { - struct drm_device *dev = crtc->dev; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct drm_device *dev = crtc->base.dev; intel_clock_t clock, reduced_clock; u32 dpll = 0, fp = 0, fp2 = 0; bool ok, has_reduced_clock = false; bool is_lvds = false; struct intel_shared_dpll *pll; - is_lvds = intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS); + is_lvds = intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS); WARN(!(HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev)), "Unexpected PCH type %d\n", INTEL_PCH_TYPE(dev)); - ok = ironlake_compute_clocks(crtc, &clock, + ok = ironlake_compute_clocks(&crtc->base, &clock, &has_reduced_clock, &reduced_clock); - if (!ok && !intel_crtc->config.clock_set) { + if (!ok && !crtc->config.clock_set) { DRM_ERROR("Couldn't find PLL settings for mode!\n"); return -EINVAL; } /* Compat-code for transition, will disappear. */ - if (!intel_crtc->config.clock_set) { - intel_crtc->config.dpll.n = clock.n; - intel_crtc->config.dpll.m1 = clock.m1; - intel_crtc->config.dpll.m2 = clock.m2; - intel_crtc->config.dpll.p1 = clock.p1; - intel_crtc->config.dpll.p2 = clock.p2; + if (!crtc->config.clock_set) { + crtc->config.dpll.n = clock.n; + crtc->config.dpll.m1 = clock.m1; + crtc->config.dpll.m2 = clock.m2; + crtc->config.dpll.p1 = clock.p1; + crtc->config.dpll.p2 = clock.p2; } /* CPU eDP is the only output that doesn't need a PCH PLL of its own. */ - if (intel_crtc->config.has_pch_encoder) { - fp = i9xx_dpll_compute_fp(&intel_crtc->config.dpll); + if (crtc->config.has_pch_encoder) { + fp = i9xx_dpll_compute_fp(&crtc->config.dpll); if (has_reduced_clock) fp2 = i9xx_dpll_compute_fp(&reduced_clock); - dpll = ironlake_compute_dpll(intel_crtc, + dpll = ironlake_compute_dpll(crtc, &fp, &reduced_clock, has_reduced_clock ? &fp2 : NULL); - intel_crtc->config.dpll_hw_state.dpll = dpll; - intel_crtc->config.dpll_hw_state.fp0 = fp; + crtc->config.dpll_hw_state.dpll = dpll; + crtc->config.dpll_hw_state.fp0 = fp; if (has_reduced_clock) - intel_crtc->config.dpll_hw_state.fp1 = fp2; + crtc->config.dpll_hw_state.fp1 = fp2; else - intel_crtc->config.dpll_hw_state.fp1 = fp; + crtc->config.dpll_hw_state.fp1 = fp; - pll = intel_get_shared_dpll(intel_crtc); + pll = intel_get_shared_dpll(crtc); if (pll == NULL) { DRM_DEBUG_DRIVER("failed to find PLL for pipe %c\n", - pipe_name(intel_crtc->pipe)); + pipe_name(crtc->pipe)); return -EINVAL; } } else - intel_put_shared_dpll(intel_crtc); + intel_put_shared_dpll(crtc); if (is_lvds && has_reduced_clock && i915.powersave) - intel_crtc->lowfreq_avail = true; + crtc->lowfreq_avail = true; else - intel_crtc->lowfreq_avail = false; + crtc->lowfreq_avail = false; return 0; } @@ -7804,16 +7802,14 @@ static void haswell_modeset_global_resources(struct drm_device *dev) modeset_update_crtc_power_domains(dev); } -static int haswell_crtc_mode_set(struct drm_crtc *crtc, +static int haswell_crtc_mode_set(struct intel_crtc *crtc, int x, int y, struct drm_framebuffer *fb) { - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - - if (!intel_ddi_pll_select(intel_crtc)) + if (!intel_ddi_pll_select(crtc)) return -EINVAL; - intel_crtc->lowfreq_avail = false; + crtc->lowfreq_avail = false; return 0; } @@ -11004,8 +11000,7 @@ static int __intel_set_mode(struct drm_crtc *crtc, crtc->x = x; crtc->y = y; - ret = dev_priv->display.crtc_mode_set(&intel_crtc->base, - x, y, fb); + ret = dev_priv->display.crtc_mode_set(intel_crtc, x, y, fb); if (ret) goto done; } -- cgit v0.10.2 From 409ee761fdc3b39e34ffccea375e7f0f13fce9cc Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Mon, 20 Oct 2014 13:46:45 +0300 Subject: drm/i915: Make intel_pipe_has_type() and some callers take intel_crtc For consistency, since that's the rule followed for internal functions. Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index b9f0f38..103972c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -406,22 +406,22 @@ static void vlv_clock(int refclk, intel_clock_t *clock) /** * Returns whether any output on the specified pipe is of the specified type */ -static bool intel_pipe_has_type(struct drm_crtc *crtc, int type) +static bool intel_pipe_has_type(struct intel_crtc *crtc, int type) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; struct intel_encoder *encoder; - for_each_encoder_on_crtc(dev, crtc, encoder) + for_each_encoder_on_crtc(dev, &crtc->base, encoder) if (encoder->type == type) return true; return false; } -static const intel_limit_t *intel_ironlake_limit(struct drm_crtc *crtc, +static const intel_limit_t *intel_ironlake_limit(struct intel_crtc *crtc, int refclk) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; const intel_limit_t *limit; if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { @@ -442,9 +442,9 @@ static const intel_limit_t *intel_ironlake_limit(struct drm_crtc *crtc, return limit; } -static const intel_limit_t *intel_g4x_limit(struct drm_crtc *crtc) +static const intel_limit_t *intel_g4x_limit(struct intel_crtc *crtc) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; const intel_limit_t *limit; if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { @@ -463,9 +463,9 @@ static const intel_limit_t *intel_g4x_limit(struct drm_crtc *crtc) return limit; } -static const intel_limit_t *intel_limit(struct drm_crtc *crtc, int refclk) +static const intel_limit_t *intel_limit(struct intel_crtc *crtc, int refclk) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; const intel_limit_t *limit; if (HAS_PCH_SPLIT(dev)) @@ -584,7 +584,7 @@ i9xx_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, intel_clock_t clock; int err = target; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { /* * For LVDS just rely on its current settings for dual-channel. * We haven't figured out how to reliably set up different @@ -645,7 +645,7 @@ pnv_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, intel_clock_t clock; int err = target; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { /* * For LVDS just rely on its current settings for dual-channel. * We haven't figured out how to reliably set up different @@ -708,7 +708,7 @@ g4x_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, int err_most = (target >> 8) + (target >> 9); found = false; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { if (intel_is_dual_link_lvds(dev)) clock.p2 = limit->p2.p2_fast; else @@ -1567,7 +1567,7 @@ static int intel_num_dvo_pipes(struct drm_device *dev) for_each_intel_crtc(dev, crtc) count += crtc->active && - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_DVO); + intel_pipe_has_type(crtc, INTEL_OUTPUT_DVO); return count; } @@ -1646,7 +1646,7 @@ static void i9xx_disable_pll(struct intel_crtc *crtc) /* Disable DVO 2x clock on both PLLs if necessary */ if (IS_I830(dev) && - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_DVO) && + intel_pipe_has_type(crtc, INTEL_OUTPUT_DVO) && intel_num_dvo_pipes(dev) == 1) { I915_WRITE(DPLL(PIPE_B), I915_READ(DPLL(PIPE_B)) & ~DPLL_DVO_2X_MODE); @@ -1884,7 +1884,7 @@ static void ironlake_enable_pch_transcoder(struct drm_i915_private *dev_priv, val &= ~TRANS_INTERLACE_MASK; if ((pipeconf_val & PIPECONF_INTERLACE_MASK) == PIPECONF_INTERLACED_ILK) if (HAS_PCH_IBX(dev_priv->dev) && - intel_pipe_has_type(crtc, INTEL_OUTPUT_SDVO)) + intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_SDVO)) val |= TRANS_LEGACY_INTERLACED_ILK; else val |= TRANS_INTERLACED; @@ -2007,7 +2007,7 @@ static void intel_enable_pipe(struct intel_crtc *crtc) * need the check. */ if (!HAS_PCH_SPLIT(dev_priv->dev)) - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_DSI)) + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DSI)) assert_dsi_pll_enabled(dev_priv); else assert_pll_enabled(dev_priv, pipe); @@ -2846,8 +2846,8 @@ static void intel_update_pipe_size(struct intel_crtc *crtc) ((adjusted_mode->crtc_hdisplay - 1) << 16) | (adjusted_mode->crtc_vdisplay - 1)); if (!crtc->config.pch_pfit.enabled && - (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS) || - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_EDP))) { + (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) || + intel_pipe_has_type(crtc, INTEL_OUTPUT_EDP))) { I915_WRITE(PF_CTL(crtc->pipe), 0); I915_WRITE(PF_WIN_POS(crtc->pipe), 0); I915_WRITE(PF_WIN_SZ(crtc->pipe), 0); @@ -3755,8 +3755,8 @@ static void ironlake_pch_enable(struct drm_crtc *crtc) /* For PCH DP, enable TRANS_DP_CTL */ if (HAS_PCH_CPT(dev) && - (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT) || - intel_pipe_has_type(crtc, INTEL_OUTPUT_EDP))) { + (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT) || + intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_EDP))) { u32 bpc = (I915_READ(PIPECONF(pipe)) & PIPECONF_BPC_MASK) >> 5; reg = TRANS_DP_CTL(pipe); temp = I915_READ(reg); @@ -4033,7 +4033,7 @@ static void intel_crtc_load_lut(struct drm_crtc *crtc) return; if (!HAS_PCH_SPLIT(dev_priv->dev)) { - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DSI)) + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DSI)) assert_dsi_pll_enabled(dev_priv); else assert_pll_enabled(dev_priv, pipe); @@ -4834,7 +4834,7 @@ static void valleyview_crtc_enable(struct drm_crtc *crtc) if (intel_crtc->active) return; - is_dsi = intel_pipe_has_type(crtc, INTEL_OUTPUT_DSI); + is_dsi = intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DSI); if (!is_dsi) { if (IS_CHERRYVIEW(dev)) @@ -5028,7 +5028,7 @@ static void i9xx_crtc_disable(struct drm_crtc *crtc) if (encoder->post_disable) encoder->post_disable(encoder); - if (!intel_pipe_has_type(crtc, INTEL_OUTPUT_DSI)) { + if (!intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DSI)) { if (IS_CHERRYVIEW(dev)) chv_disable_pll(dev_priv, pipe); else if (IS_VALLEYVIEW(dev)) @@ -5411,7 +5411,7 @@ static int intel_crtc_compute_config(struct intel_crtc *crtc, * - LVDS dual channel mode * - Double wide pipe */ - if ((intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS) && + if ((intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) && intel_is_dual_link_lvds(dev)) || pipe_config->double_wide) pipe_config->pipe_src_w &= ~1; @@ -5599,9 +5599,9 @@ static inline bool intel_panel_use_ssc(struct drm_i915_private *dev_priv) && !(dev_priv->quirks & QUIRK_LVDS_SSC_DISABLE); } -static int i9xx_get_refclk(struct drm_crtc *crtc, int num_connectors) +static int i9xx_get_refclk(struct intel_crtc *crtc, int num_connectors) { - struct drm_device *dev = crtc->dev; + struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; int refclk; @@ -5649,7 +5649,7 @@ static void i9xx_update_pll_dividers(struct intel_crtc *crtc, crtc->config.dpll_hw_state.fp0 = fp; crtc->lowfreq_avail = false; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS) && + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) && reduced_clock && i915.powersave) { crtc->config.dpll_hw_state.fp1 = fp2; crtc->lowfreq_avail = true; @@ -5818,16 +5818,16 @@ static void vlv_prepare_pll(struct intel_crtc *crtc) /* Set HBR and RBR LPF coefficients */ if (crtc->config.port_clock == 162000 || - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_ANALOG) || - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_HDMI)) + intel_pipe_has_type(crtc, INTEL_OUTPUT_ANALOG) || + intel_pipe_has_type(crtc, INTEL_OUTPUT_HDMI)) vlv_dpio_write(dev_priv, pipe, VLV_PLL_DW10(pipe), 0x009f0003); else vlv_dpio_write(dev_priv, pipe, VLV_PLL_DW10(pipe), 0x00d0000f); - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_EDP) || - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_DISPLAYPORT)) { + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_EDP) || + intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) { /* Use SSC source */ if (pipe == PIPE_A) vlv_dpio_write(dev_priv, pipe, VLV_PLL_DW5(pipe), @@ -5847,8 +5847,8 @@ static void vlv_prepare_pll(struct intel_crtc *crtc) coreclk = vlv_dpio_read(dev_priv, pipe, VLV_PLL_DW7(pipe)); coreclk = (coreclk & 0x0000ff00) | 0x01c00000; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_DISPLAYPORT) || - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_EDP)) + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT) || + intel_pipe_has_type(crtc, INTEL_OUTPUT_EDP)) coreclk |= 0x01000000; vlv_dpio_write(dev_priv, pipe, VLV_PLL_DW7(pipe), coreclk); @@ -5918,7 +5918,7 @@ static void chv_prepare_pll(struct intel_crtc *crtc) (2 << DPIO_CHV_FEEDFWD_GAIN_SHIFT)); /* Loop filter */ - refclk = i9xx_get_refclk(&crtc->base, 0); + refclk = i9xx_get_refclk(crtc, 0); loopfilter = 5 << DPIO_CHV_PROP_COEFF_SHIFT | 2 << DPIO_CHV_GAIN_CTRL_SHIFT; if (refclk == 100000) @@ -5950,12 +5950,12 @@ static void i9xx_update_pll(struct intel_crtc *crtc, i9xx_update_pll_dividers(crtc, reduced_clock); - is_sdvo = intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_SDVO) || - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_HDMI); + is_sdvo = intel_pipe_has_type(crtc, INTEL_OUTPUT_SDVO) || + intel_pipe_has_type(crtc, INTEL_OUTPUT_HDMI); dpll = DPLL_VGA_MODE_DIS; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS)) + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) dpll |= DPLLB_MODE_LVDS; else dpll |= DPLLB_MODE_DAC_SERIAL; @@ -5968,7 +5968,7 @@ static void i9xx_update_pll(struct intel_crtc *crtc, if (is_sdvo) dpll |= DPLL_SDVO_HIGH_SPEED; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_DISPLAYPORT)) + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) dpll |= DPLL_SDVO_HIGH_SPEED; /* compute bitmask from p1 value */ @@ -5998,7 +5998,7 @@ static void i9xx_update_pll(struct intel_crtc *crtc, if (crtc->config.sdvo_tv_clock) dpll |= PLL_REF_INPUT_TVCLKINBC; - else if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS) && + else if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) && intel_panel_use_ssc(dev_priv) && num_connectors < 2) dpll |= PLLB_REF_INPUT_SPREADSPECTRUMIN; else @@ -6027,7 +6027,7 @@ static void i8xx_update_pll(struct intel_crtc *crtc, dpll = DPLL_VGA_MODE_DIS; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { dpll |= (1 << (clock->p1 - 1)) << DPLL_FPA01_P1_POST_DIV_SHIFT; } else { if (clock->p1 == 2) @@ -6038,10 +6038,10 @@ static void i8xx_update_pll(struct intel_crtc *crtc, dpll |= PLL_P2_DIVIDE_BY_4; } - if (!IS_I830(dev) && intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_DVO)) + if (!IS_I830(dev) && intel_pipe_has_type(crtc, INTEL_OUTPUT_DVO)) dpll |= DPLL_DVO_2X_MODE; - if (intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS) && + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) && intel_panel_use_ssc(dev_priv) && num_connectors < 2) dpll |= PLLB_REF_INPUT_SPREADSPECTRUMIN; else @@ -6072,7 +6072,7 @@ static void intel_set_pipe_timings(struct intel_crtc *intel_crtc) crtc_vtotal -= 1; crtc_vblank_end -= 1; - if (intel_pipe_has_type(&intel_crtc->base, INTEL_OUTPUT_SDVO)) + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_SDVO)) vsyncshift = (adjusted_mode->crtc_htotal - 1) / 2; else vsyncshift = adjusted_mode->crtc_hsync_start - @@ -6230,7 +6230,7 @@ static void i9xx_set_pipeconf(struct intel_crtc *intel_crtc) if (intel_crtc->config.adjusted_mode.flags & DRM_MODE_FLAG_INTERLACE) { if (INTEL_INFO(dev)->gen < 4 || - intel_pipe_has_type(&intel_crtc->base, INTEL_OUTPUT_SDVO)) + intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_SDVO)) pipeconf |= PIPECONF_INTERLACE_W_FIELD_INDICATION; else pipeconf |= PIPECONF_INTERLACE_W_SYNC_SHIFT; @@ -6274,7 +6274,7 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, return 0; if (!crtc->config.clock_set) { - refclk = i9xx_get_refclk(&crtc->base, num_connectors); + refclk = i9xx_get_refclk(crtc, num_connectors); /* * Returns a set of divisors for the desired target clock with @@ -6282,7 +6282,7 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, * the clock equation: reflck * (5 * (m1 + 2) + (m2 + 2)) / (n + * 2) / p1 / p2. */ - limit = intel_limit(&crtc->base, refclk); + limit = intel_limit(crtc, refclk); ok = dev_priv->display.find_dpll(limit, crtc, crtc->config.port_clock, refclk, NULL, &clock); @@ -7114,7 +7114,7 @@ static bool ironlake_compute_clocks(struct drm_crtc *crtc, const intel_limit_t *limit; bool ret, is_lvds = false; - is_lvds = intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS); + is_lvds = intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_LVDS); refclk = ironlake_get_refclk(crtc); @@ -7123,7 +7123,7 @@ static bool ironlake_compute_clocks(struct drm_crtc *crtc, * refclk, or FALSE. The returned values represent the clock equation: * reflck * (5 * (m1 + 2) + (m2 + 2)) / (n + 2) / p1 / p2. */ - limit = intel_limit(crtc, refclk); + limit = intel_limit(intel_crtc, refclk); ret = dev_priv->display.find_dpll(limit, intel_crtc, intel_crtc->config.port_clock, refclk, NULL, clock); @@ -7259,7 +7259,7 @@ static int ironlake_crtc_mode_set(struct intel_crtc *crtc, bool is_lvds = false; struct intel_shared_dpll *pll; - is_lvds = intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_LVDS); + is_lvds = intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS); WARN(!(HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev)), "Unexpected PCH type %d\n", INTEL_PCH_TYPE(dev)); @@ -8049,6 +8049,7 @@ static void haswell_write_eld(struct drm_connector *connector, struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = connector->dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); uint8_t *eld = connector->eld; uint32_t eldv; uint32_t i; @@ -8089,7 +8090,7 @@ static void haswell_write_eld(struct drm_connector *connector, eldv = AUDIO_ELD_VALID_A << (pipe * 4); - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) { + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { DRM_DEBUG_DRIVER("ELD: DisplayPort detected\n"); eld[5] |= (1 << 2); /* Conn_Type, 0x1 = DisplayPort */ I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ @@ -8132,6 +8133,7 @@ static void ironlake_write_eld(struct drm_connector *connector, struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = connector->dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); uint8_t *eld = connector->eld; uint32_t eldv; uint32_t i; @@ -8185,7 +8187,7 @@ static void ironlake_write_eld(struct drm_connector *connector, eldv = IBX_ELD_VALIDB << ((i - 1) * 4); } - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) { + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { DRM_DEBUG_DRIVER("ELD: DisplayPort detected\n"); eld[5] |= (1 << 2); /* Conn_Type, 0x1 = DisplayPort */ I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ @@ -10880,7 +10882,7 @@ static void update_scanline_offset(struct intel_crtc *crtc) crtc->scanline_offset = vtotal - 1; } else if (HAS_DDI(dev) && - intel_pipe_has_type(&crtc->base, INTEL_OUTPUT_HDMI)) { + intel_pipe_has_type(crtc, INTEL_OUTPUT_HDMI)) { crtc->scanline_offset = 2; } else crtc->scanline_offset = 1; -- cgit v0.10.2 From 0b14cbd2f58199a024acbe2994bb27533c97d756 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 10 Sep 2014 18:16:55 +0300 Subject: drm/i915: remove dead code from legacy suspend handler MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The legacy DRM suspend logic (effective in UMS) doesn't handle any S4 thaw events so we don't need to care about it either. Only S3 suspend and S4 freeze events are handled. Leave an assert behind to be sure. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index a05a1d0..f7bc3e980 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -640,9 +640,9 @@ int i915_suspend(struct drm_device *dev, pm_message_t state) return -ENODEV; } - if (state.event == PM_EVENT_PRETHAW) - return 0; - + if (WARN_ON_ONCE(state.event != PM_EVENT_SUSPEND && + state.event != PM_EVENT_FREEZE)) + return -EINVAL; if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; -- cgit v0.10.2 From 8eff426233e61332686acea22a4199288047ec64 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Tue, 21 Oct 2014 14:40:37 +0300 Subject: drm/i915: remove redundant #ifdef CONFIG_COMPAT The whole file is only built with CONFIG_COMPAT=y. Signed-off-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_ioc32.c b/drivers/gpu/drm/i915/i915_ioc32.c index 2e0613e..176de63 100644 --- a/drivers/gpu/drm/i915/i915_ioc32.c +++ b/drivers/gpu/drm/i915/i915_ioc32.c @@ -189,7 +189,6 @@ static drm_ioctl_compat_t *i915_compat_ioctls[] = { [DRM_I915_ALLOC] = compat_i915_alloc }; -#ifdef CONFIG_COMPAT /** * Called whenever a 32-bit process running under a 64-bit kernel * performs an ioctl on /dev/dri/card. @@ -218,4 +217,3 @@ long i915_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) return ret; } -#endif -- cgit v0.10.2 From 7225342ab501befdb64bcec76ded41f5897c0855 Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Tue, 7 Oct 2014 17:21:26 +0300 Subject: drm/i915: Build workaround list in ring initialization If we build the workaround list in ring initialization and decouple it from the actual writing of values, we gain the ability to decide where and how we want to apply the values. The advantage of this will become more clear when we need to initialize workarounds on older gens where it is not possible to write all the registers through ring LRIs. v2: rebase on newest bdw workarounds Cc: Arun Siluvery Cc: Damien Lespiau Signed-off-by: Mika Kuoppala Reviewed-by: Arun Siluvery [danvet: Resolve tiny conflict in comments and ocd alignments a bit.] [danvet2: Remove bogus force_wake_get call spotted by Paulo and QA.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index da4036d..62c111b 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2655,18 +2655,18 @@ static int i915_wa_registers(struct seq_file *m, void *unused) intel_runtime_pm_get(dev_priv); - seq_printf(m, "Workarounds applied: %d\n", dev_priv->num_wa_regs); - for (i = 0; i < dev_priv->num_wa_regs; ++i) { + seq_printf(m, "Workarounds applied: %d\n", dev_priv->workarounds.count); + for (i = 0; i < dev_priv->workarounds.count; ++i) { u32 addr, mask; - addr = dev_priv->intel_wa_regs[i].addr; - mask = dev_priv->intel_wa_regs[i].mask; - dev_priv->intel_wa_regs[i].value = I915_READ(addr) | mask; - if (dev_priv->intel_wa_regs[i].addr) + addr = dev_priv->workarounds.reg[i].addr; + mask = dev_priv->workarounds.reg[i].mask; + dev_priv->workarounds.reg[i].value = I915_READ(addr) | mask; + if (dev_priv->workarounds.reg[i].addr) seq_printf(m, "0x%X: 0x%08X, mask: 0x%08X\n", - dev_priv->intel_wa_regs[i].addr, - dev_priv->intel_wa_regs[i].value, - dev_priv->intel_wa_regs[i].mask); + dev_priv->workarounds.reg[i].addr, + dev_priv->workarounds.reg[i].value, + dev_priv->workarounds.reg[i].mask); } intel_runtime_pm_put(dev_priv); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 9f3d689..d1905b3 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1448,6 +1448,20 @@ struct i915_frontbuffer_tracking { unsigned flip_bits; }; +struct i915_wa_reg { + u32 addr; + u32 value; + /* bitmask representing WA bits */ + u32 mask; +}; + +#define I915_MAX_WA_REGS 16 + +struct i915_workarounds { + struct i915_wa_reg reg[I915_MAX_WA_REGS]; + u32 count; +}; + struct drm_i915_private { struct drm_device *dev; struct kmem_cache *slab; @@ -1590,19 +1604,7 @@ struct drm_i915_private { struct intel_shared_dpll shared_dplls[I915_NUM_PLLS]; int dpio_phy_iosf_port[I915_NUM_PHYS_VLV]; - /* - * workarounds are currently applied at different places and - * changes are being done to consolidate them so exact count is - * not clear at this point, use a max value for now. - */ -#define I915_MAX_WA_REGS 16 - struct { - u32 addr; - u32 value; - /* bitmask representing WA bits */ - u32 mask; - } intel_wa_regs[I915_MAX_WA_REGS]; - u32 num_wa_regs; + struct i915_workarounds workarounds; /* Reclocking support */ bool render_reclock_avail; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 5ebe46a..5f935d4 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -665,79 +665,107 @@ err: return ret; } -static inline void intel_ring_emit_wa(struct intel_engine_cs *ring, - u32 addr, u32 value) +static int intel_ring_workarounds_emit(struct intel_engine_cs *ring) { + int ret, i; struct drm_device *dev = ring->dev; struct drm_i915_private *dev_priv = dev->dev_private; + struct i915_workarounds *w = &dev_priv->workarounds; - if (WARN_ON(dev_priv->num_wa_regs >= I915_MAX_WA_REGS)) - return; + if (WARN_ON(w->count == 0)) + return 0; - intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1)); - intel_ring_emit(ring, addr); - intel_ring_emit(ring, value); + ring->gpu_caches_dirty = true; + ret = intel_ring_flush_all_caches(ring); + if (ret) + return ret; - dev_priv->intel_wa_regs[dev_priv->num_wa_regs].addr = addr; - dev_priv->intel_wa_regs[dev_priv->num_wa_regs].mask = value & 0xFFFF; - /* value is updated with the status of remaining bits of this - * register when it is read from debugfs file - */ - dev_priv->intel_wa_regs[dev_priv->num_wa_regs].value = value; - dev_priv->num_wa_regs++; + ret = intel_ring_begin(ring, w->count * 3); + if (ret) + return ret; + + for (i = 0; i < w->count; i++) { + intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1)); + intel_ring_emit(ring, w->reg[i].addr); + intel_ring_emit(ring, w->reg[i].value); + } + + intel_ring_advance(ring); + + ring->gpu_caches_dirty = true; + ret = intel_ring_flush_all_caches(ring); + if (ret) + return ret; + + DRM_DEBUG_DRIVER("Number of Workarounds emitted: %d\n", w->count); - return; + return 0; +} + +static int wa_add(struct drm_i915_private *dev_priv, + const u32 addr, const u32 val, const u32 mask) +{ + const u32 idx = dev_priv->workarounds.count; + + if (WARN_ON(idx >= I915_MAX_WA_REGS)) + return -ENOSPC; + + dev_priv->workarounds.reg[idx].addr = addr; + dev_priv->workarounds.reg[idx].value = val; + dev_priv->workarounds.reg[idx].mask = mask; + + dev_priv->workarounds.count++; + + return 0; } +#define WA_REG(addr, val, mask) { \ + const int r = wa_add(dev_priv, (addr), (val), (mask)); \ + if (r) \ + return r; \ + } + +#define WA_SET_BIT_MASKED(addr, mask) \ + WA_REG(addr, _MASKED_BIT_ENABLE(mask), (mask) & 0xffff) + +#define WA_CLR_BIT_MASKED(addr, mask) \ + WA_REG(addr, _MASKED_BIT_DISABLE(mask), (mask) & 0xffff) + +#define WA_SET_BIT(addr, mask) WA_REG(addr, I915_READ(addr) | (mask), mask) +#define WA_CLR_BIT(addr, mask) WA_REG(addr, I915_READ(addr) & ~(mask), mask) + +#define WA_WRITE(addr, val) WA_REG(addr, val, 0xffffffff) + static int bdw_init_workarounds(struct intel_engine_cs *ring) { - int ret; struct drm_device *dev = ring->dev; struct drm_i915_private *dev_priv = dev->dev_private; - /* - * workarounds applied in this fn are part of register state context, - * they need to be re-initialized followed by gpu reset, suspend/resume, - * module reload. - */ - dev_priv->num_wa_regs = 0; - memset(dev_priv->intel_wa_regs, 0, sizeof(dev_priv->intel_wa_regs)); - - /* - * update the number of dwords required based on the - * actual number of workarounds applied - */ - ret = intel_ring_begin(ring, 18); - if (ret) - return ret; - /* WaDisablePartialInstShootdown:bdw */ /* WaDisableThreadStallDopClockGating:bdw (pre-production) */ - intel_ring_emit_wa(ring, GEN8_ROW_CHICKEN, - _MASKED_BIT_ENABLE(PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE - | STALL_DOP_GATING_DISABLE)); + WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN, + PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE | + STALL_DOP_GATING_DISABLE); /* WaDisableDopClockGating:bdw */ - intel_ring_emit_wa(ring, GEN7_ROW_CHICKEN2, - _MASKED_BIT_ENABLE(DOP_CLOCK_GATING_DISABLE)); + WA_SET_BIT_MASKED(GEN7_ROW_CHICKEN2, + DOP_CLOCK_GATING_DISABLE); - intel_ring_emit_wa(ring, HALF_SLICE_CHICKEN3, - _MASKED_BIT_ENABLE(GEN8_SAMPLER_POWER_BYPASS_DIS)); + WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3, + GEN8_SAMPLER_POWER_BYPASS_DIS); /* Use Force Non-Coherent whenever executing a 3D context. This is a * workaround for for a possible hang in the unlikely event a TLB * invalidation occurs during a PSD flush. */ /* WaDisableFenceDestinationToSLM:bdw (GT3 pre-production) */ - intel_ring_emit_wa(ring, HDC_CHICKEN0, - _MASKED_BIT_ENABLE(HDC_FORCE_NON_COHERENT | - (IS_BDW_GT3(dev) ? - HDC_FENCE_DEST_SLM_DISABLE : 0) - )); + WA_SET_BIT_MASKED(HDC_CHICKEN0, + HDC_FORCE_NON_COHERENT | + (IS_BDW_GT3(dev) ? HDC_FENCE_DEST_SLM_DISABLE : 0)); /* Wa4x4STCOptimizationDisable:bdw */ - intel_ring_emit_wa(ring, CACHE_MODE_1, - _MASKED_BIT_ENABLE(GEN8_4x4_STC_OPTIMIZATION_DISABLE)); + WA_SET_BIT_MASKED(CACHE_MODE_1, + GEN8_4x4_STC_OPTIMIZATION_DISABLE); /* * BSpec recommends 8x4 when MSAA is used, @@ -747,52 +775,50 @@ static int bdw_init_workarounds(struct intel_engine_cs *ring) * disable bit, which we don't touch here, but it's good * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM). */ - intel_ring_emit_wa(ring, GEN7_GT_MODE, - GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); - - intel_ring_advance(ring); - - DRM_DEBUG_DRIVER("Number of Workarounds applied: %d\n", - dev_priv->num_wa_regs); + WA_SET_BIT_MASKED(GEN7_GT_MODE, + GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); return 0; } static int chv_init_workarounds(struct intel_engine_cs *ring) { - int ret; struct drm_device *dev = ring->dev; struct drm_i915_private *dev_priv = dev->dev_private; - /* - * workarounds applied in this fn are part of register state context, - * they need to be re-initialized followed by gpu reset, suspend/resume, - * module reload. - */ - dev_priv->num_wa_regs = 0; - memset(dev_priv->intel_wa_regs, 0, sizeof(dev_priv->intel_wa_regs)); - - ret = intel_ring_begin(ring, 12); - if (ret) - return ret; - /* WaDisablePartialInstShootdown:chv */ - intel_ring_emit_wa(ring, GEN8_ROW_CHICKEN, - _MASKED_BIT_ENABLE(PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE)); + WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN, + PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE); /* WaDisableThreadStallDopClockGating:chv */ - intel_ring_emit_wa(ring, GEN8_ROW_CHICKEN, - _MASKED_BIT_ENABLE(STALL_DOP_GATING_DISABLE)); + WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN, + STALL_DOP_GATING_DISABLE); /* WaDisableDopClockGating:chv (pre-production hw) */ - intel_ring_emit_wa(ring, GEN7_ROW_CHICKEN2, - _MASKED_BIT_ENABLE(DOP_CLOCK_GATING_DISABLE)); + WA_SET_BIT_MASKED(GEN7_ROW_CHICKEN2, + DOP_CLOCK_GATING_DISABLE); /* WaDisableSamplerPowerBypass:chv (pre-production hw) */ - intel_ring_emit_wa(ring, HALF_SLICE_CHICKEN3, - _MASKED_BIT_ENABLE(GEN8_SAMPLER_POWER_BYPASS_DIS)); + WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3, + GEN8_SAMPLER_POWER_BYPASS_DIS); - intel_ring_advance(ring); + return 0; +} + +static int init_workarounds_ring(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + + WARN_ON(ring->id != RCS); + + dev_priv->workarounds.count = 0; + + if (IS_BROADWELL(dev)) + return bdw_init_workarounds(ring); + + if (IS_CHERRYVIEW(dev)) + return chv_init_workarounds(ring); return 0; } @@ -852,7 +878,7 @@ static int init_render_ring(struct intel_engine_cs *ring) if (HAS_L3_DPF(dev)) I915_WRITE_IMR(ring, ~GT_PARITY_ERROR(dev)); - return ret; + return init_workarounds_ring(ring); } static void render_ring_cleanup(struct intel_engine_cs *ring) @@ -2298,10 +2324,8 @@ int intel_init_render_ring_buffer(struct drm_device *dev) dev_priv->semaphore_obj = obj; } } - if (IS_CHERRYVIEW(dev)) - ring->init_context = chv_init_workarounds; - else - ring->init_context = bdw_init_workarounds; + + ring->init_context = intel_ring_workarounds_emit; ring->add_request = gen6_add_request; ring->flush = gen8_render_ring_flush; ring->irq_get = gen8_ring_get_irq; -- cgit v0.10.2 From 2fa60f6d515b509b05f6a96afc83ac9629d6401b Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Tue, 7 Oct 2014 17:21:27 +0300 Subject: drm/i915: Check workaround status on dfs read time As the workaround list has the value as initialization time constant, we can do the simple checking on the go without negleting igt. Signed-off-by: Mika Kuoppala Reviewed-by: Arun Siluvery Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 62c111b..6f8c9d7 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2657,16 +2657,16 @@ static int i915_wa_registers(struct seq_file *m, void *unused) seq_printf(m, "Workarounds applied: %d\n", dev_priv->workarounds.count); for (i = 0; i < dev_priv->workarounds.count; ++i) { - u32 addr, mask; + u32 addr, mask, value, read; + bool ok; addr = dev_priv->workarounds.reg[i].addr; mask = dev_priv->workarounds.reg[i].mask; - dev_priv->workarounds.reg[i].value = I915_READ(addr) | mask; - if (dev_priv->workarounds.reg[i].addr) - seq_printf(m, "0x%X: 0x%08X, mask: 0x%08X\n", - dev_priv->workarounds.reg[i].addr, - dev_priv->workarounds.reg[i].value, - dev_priv->workarounds.reg[i].mask); + value = dev_priv->workarounds.reg[i].value; + read = I915_READ(addr); + ok = (value & mask) == (read & mask); + seq_printf(m, "0x%X: 0x%08X, mask: 0x%08X, read: 0x%08x, status: %s\n", + addr, value, mask, read, ok ? "OK" : "FAIL"); } intel_runtime_pm_put(dev_priv); -- cgit v0.10.2 From 163f53a2611c98a9f8e25cbda30ce8ce268b92de Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 10 Sep 2014 18:16:54 +0300 Subject: drm/i915: vlv: fix gunit HW state corruption during S4 suspend MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit During S4 freeze we don't call intel_suspend_complete(), which would save the gunit HW state, but during S4 thaw/restore events we call intel_resume_prepare() which restores it, thus ending up in a corrupted HW state. Fix this by calling intel_suspend_complete() from the corresponding freeze_late event handler. The issue was introduced in commit 016970beb05da6285c2f3ed2bee1c676cb75972e Author: Sagar Kamble Date: Wed Aug 13 23:07:06 2014 +0530 CC: Sagar Kamble Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index f7bc3e980..ffd672e 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1011,6 +1011,15 @@ static int i915_pm_freeze(struct device *dev) return i915_drm_freeze(drm_dev); } +static int i915_pm_freeze_late(struct device *dev) +{ + struct pci_dev *pdev = to_pci_dev(dev); + struct drm_device *drm_dev = pci_get_drvdata(pdev); + struct drm_i915_private *dev_priv = drm_dev->dev_private; + + return intel_suspend_complete(dev_priv); +} + static int i915_pm_thaw_early(struct device *dev) { struct pci_dev *pdev = to_pci_dev(dev); @@ -1595,6 +1604,7 @@ static const struct dev_pm_ops i915_pm_ops = { .resume_early = i915_pm_resume_early, .resume = i915_pm_resume, .freeze = i915_pm_freeze, + .freeze_late = i915_pm_freeze_late, .thaw_early = i915_pm_thaw_early, .thaw = i915_pm_thaw, .poweroff = i915_pm_poweroff, -- cgit v0.10.2 From d9ceb8163339134bd3ffb9fb87a0db4698283e32 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Thu, 9 Oct 2014 12:57:43 -0700 Subject: drm/i915: preserve swizzle settings if necessary v4 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Some machines (like MBAs) might use a tiled framebuffer but not enable display swizzling at boot time. We want to preserve that configuration if possible to prevent a boot time mode set. On IVB+ it shouldn't affect performance anyway since the memory controller does internal swizzling anyway. For most other configs we'll be able to enable swizzling at boot time, since the initial framebuffer won't be tiled, thus we won't see any corruption when we enable it. v2: preserve swizzling if BIOS had it set (Daniel) v3: preserve swizzling only if we inherited a tiled framebuffer (Daniel) check display swizzle setting in detect_bit_6_swizzle (Daniel) use gen6 as cutoff point (Daniel) v4: fixup swizzle preserve again, had wrong init order (Daniel) Reported-by: Kristian Høgsberg Signed-off-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index d1905b3..3212d62 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1541,6 +1541,8 @@ struct drm_i915_private { struct intel_opregion opregion; struct intel_vbt_data vbt; + bool preserve_bios_swizzle; + /* overlay */ struct intel_overlay *overlay; diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index 2cefb59..d1e7a3e 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -102,22 +102,33 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev) swizzle_x = I915_BIT_6_SWIZZLE_NONE; swizzle_y = I915_BIT_6_SWIZZLE_NONE; } else if (INTEL_INFO(dev)->gen >= 6) { - uint32_t dimm_c0, dimm_c1; - dimm_c0 = I915_READ(MAD_DIMM_C0); - dimm_c1 = I915_READ(MAD_DIMM_C1); - dimm_c0 &= MAD_DIMM_A_SIZE_MASK | MAD_DIMM_B_SIZE_MASK; - dimm_c1 &= MAD_DIMM_A_SIZE_MASK | MAD_DIMM_B_SIZE_MASK; - /* Enable swizzling when the channels are populated with - * identically sized dimms. We don't need to check the 3rd - * channel because no cpu with gpu attached ships in that - * configuration. Also, swizzling only makes sense for 2 - * channels anyway. */ - if (dimm_c0 == dimm_c1) { - swizzle_x = I915_BIT_6_SWIZZLE_9_10; - swizzle_y = I915_BIT_6_SWIZZLE_9; + if (dev_priv->preserve_bios_swizzle) { + if (I915_READ(DISP_ARB_CTL) & + DISP_TILE_SURFACE_SWIZZLING) { + swizzle_x = I915_BIT_6_SWIZZLE_9_10; + swizzle_y = I915_BIT_6_SWIZZLE_9; + } else { + swizzle_x = I915_BIT_6_SWIZZLE_NONE; + swizzle_y = I915_BIT_6_SWIZZLE_NONE; + } } else { - swizzle_x = I915_BIT_6_SWIZZLE_NONE; - swizzle_y = I915_BIT_6_SWIZZLE_NONE; + uint32_t dimm_c0, dimm_c1; + dimm_c0 = I915_READ(MAD_DIMM_C0); + dimm_c1 = I915_READ(MAD_DIMM_C1); + dimm_c0 &= MAD_DIMM_A_SIZE_MASK | MAD_DIMM_B_SIZE_MASK; + dimm_c1 &= MAD_DIMM_A_SIZE_MASK | MAD_DIMM_B_SIZE_MASK; + /* Enable swizzling when the channels are populated + * with identically sized dimms. We don't need to check + * the 3rd channel because no cpu with gpu attached + * ships in that configuration. Also, swizzling only + * makes sense for 2 channels anyway. */ + if (dimm_c0 == dimm_c1) { + swizzle_x = I915_BIT_6_SWIZZLE_9_10; + swizzle_y = I915_BIT_6_SWIZZLE_9; + } else { + swizzle_x = I915_BIT_6_SWIZZLE_NONE; + swizzle_y = I915_BIT_6_SWIZZLE_NONE; + } } } else if (IS_GEN5(dev)) { /* On Ironlake whatever DRAM config, GPU always do diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 103972c..fb9a434 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2359,6 +2359,7 @@ static void intel_find_plane_obj(struct intel_crtc *intel_crtc, struct intel_plane_config *plane_config) { struct drm_device *dev = intel_crtc->base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; struct drm_crtc *c; struct intel_crtc *i; struct drm_i915_gem_object *obj; @@ -2390,6 +2391,9 @@ static void intel_find_plane_obj(struct intel_crtc *intel_crtc, continue; if (i915_gem_obj_ggtt_offset(obj) == plane_config->base) { + if (obj->tiling_mode != I915_TILING_NONE) + dev_priv->preserve_bios_swizzle = true; + drm_framebuffer_reference(c->primary->fb); intel_crtc->base.primary->fb = c->primary->fb; obj->frontbuffer_bits |= INTEL_FRONTBUFFER_PRIMARY(intel_crtc->pipe); -- cgit v0.10.2 From e6b2627c6a8b25afc85649b54987b4bea27875b1 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Fri, 10 Oct 2014 17:53:33 +0300 Subject: drm/i915: spt does not have pch backlight override bit MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit SPT is always in the PCH override mode, and the bit MBZ. Only set override on LPT. v2: check for PCH version (Ville) Signed-off-by: Jani Nikula Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c index e3def5a..e18b3f4 100644 --- a/drivers/gpu/drm/i915/intel_panel.c +++ b/drivers/gpu/drm/i915/intel_panel.c @@ -775,8 +775,9 @@ static void bdw_enable_backlight(struct intel_connector *connector) if (panel->backlight.active_low_pwm) pch_ctl1 |= BLM_PCH_POLARITY; - /* BDW always uses the pch pwm controls. */ - pch_ctl1 |= BLM_PCH_OVERRIDE_ENABLE; + /* After LPT, override is the default. */ + if (HAS_PCH_LPT(dev_priv)) + pch_ctl1 |= BLM_PCH_OVERRIDE_ENABLE; I915_WRITE(BLC_PWM_PCH_CTL1, pch_ctl1); POSTING_READ(BLC_PWM_PCH_CTL1); -- cgit v0.10.2 From 2ebfaf5f10c6f00d4232f741c82c421eca8ae784 Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Wed, 15 Oct 2014 14:15:04 -0300 Subject: drm/i915: call drm_vblank_cleanup() earlier at unload MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In its current place, it just segfaults while trying to access the CRTC structures: [ 9132.421681] Call Trace: [ 9132.421707] [] i915_get_crtc_scanoutpos+0x1e8/0x220 [i915] [ 9132.421727] [] drm_calc_vbltimestamp_from_scanoutpos+0x94/0x330 [drm] [ 9132.421744] [] ?vblank_disable_and_save+0x40/0x1e0 [drm] [ 9132.421769] [] i915_get_vblank_timestamp+0x68/0xb0 [i915] [ 9132.421786] [] drm_get_last_vbltimestamp+0x44/0x80 [drm] [ 9132.421801] [] vblank_disable_and_save+0x1a6/0x1e0 [drm] [ 9132.421817] [] drm_vblank_cleanup+0x61/0xa0 [drm] [ 9132.421849] [] i915_driver_unload+0xde/0x290 [i915] [ 9132.421867] [] drm_dev_unregister+0x24/0xb0 [drm] [ 9132.421884] [] drm_put_dev+0x1e/0x70 [drm] [ 9132.421901] [] i915_pci_remove+0x10/0x20 [i915] [ 9132.421910] [] pci_device_remove+0x36/0xb0 [ 9132.421920] [] __device_release_driver+0x7a/0xf0 [ 9132.421928] [] driver_detach+0xb8/0xc0 [ 9132.421936] [] bus_remove_driver+0x4a/0xb0 [ 9132.421944] [] driver_unregister+0x27/0x50 [ 9132.421953] [] pci_unregister_driver+0x25/0x70 [ 9132.421971] [] drm_pci_exit+0x78/0xa0 [drm] [ 9132.422000] [] i915_exit+0x20/0x94e [i915] [ 9132.422009] [] SyS_delete_module+0x13c/0x1f0 [ 9132.422019] [] ? trace_hardirqs_on_thunk+0x3a/0x3f [ 9132.422028] [] system_call_fastpath+0x16/0x1b This means it has to be before intel_modeset_cleanup, which cleans the CRTC structures. But if we move it to before intel_fbdev_fini(), we get WARNs because intel_fbdev_fini() still tries to use the vblanks, so the only acceptable point for drm_vblank_cleanup() seems to be this place. Related commit: commit cbb47d179fb345c579cd8cd884693903fceed26a Author: Chris Wilson Date: Mon Sep 23 17:33:20 2013 -0300 drm/i915: Add some missing steps to i915_driver_load error path Testsuite: igt/drv_module_reload Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=77511 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83484 Signed-off-by: Paulo Zanoni Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 85d14e1..1b39807 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1853,8 +1853,12 @@ int i915_driver_unload(struct drm_device *dev) acpi_video_unregister(); - if (drm_core_check_feature(dev, DRIVER_MODESET)) { + if (drm_core_check_feature(dev, DRIVER_MODESET)) intel_fbdev_fini(dev); + + drm_vblank_cleanup(dev); + + if (drm_core_check_feature(dev, DRIVER_MODESET)) { intel_modeset_cleanup(dev); /* @@ -1895,8 +1899,6 @@ int i915_driver_unload(struct drm_device *dev) i915_free_hws(dev); } - drm_vblank_cleanup(dev); - intel_teardown_gmbus(dev); intel_teardown_mchbar(dev); -- cgit v0.10.2 From 8c740dcea254a1472df2c0ac5ac585412a2507ec Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Fri, 17 Oct 2014 18:42:03 -0300 Subject: drm/i915: disable IPS while getting the pipe CRCs. For some yet-undiscovered reason, when IPS gets enabled, the pipe CRC changes. Since hsw_enable_ips() doesn't really guarantees to enable IPS (it depends on package C-states), we can't really predict if IPS is enabled or disabled while running our CRC tests, so let's just completely disable IPS while pipe CRCs are being used. If we find a way to make IPS not change the pipe CRC result, we may want to fix IPS and then revert this patch. While this doesn't happen, let's merge this patch, so every IGT test relying on the CRCs can work on pipe A. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72864 Testcase: igt/kms_cursor_crc (and others) Signed-off-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 6f8c9d7..1e7a7c6 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -3255,6 +3255,8 @@ static int pipe_crc_set_source(struct drm_device *dev, enum pipe pipe, { struct drm_i915_private *dev_priv = dev->dev_private; struct intel_pipe_crc *pipe_crc = &dev_priv->pipe_crc[pipe]; + struct intel_crtc *crtc = to_intel_crtc(intel_get_crtc_for_pipe(dev, + pipe)); u32 val = 0; /* shut up gcc */ int ret; @@ -3290,6 +3292,14 @@ static int pipe_crc_set_source(struct drm_device *dev, enum pipe pipe, if (!pipe_crc->entries) return -ENOMEM; + /* + * When IPS gets enabled, the pipe CRC changes. Since IPS gets + * enabled and disabled dynamically based on package C states, + * user space can't make reliable use of the CRCs, so let's just + * completely disable it. + */ + hsw_disable_ips(crtc); + spin_lock_irq(&pipe_crc->lock); pipe_crc->head = 0; pipe_crc->tail = 0; @@ -3328,6 +3338,8 @@ static int pipe_crc_set_source(struct drm_device *dev, enum pipe pipe, vlv_undo_pipe_scramble_reset(dev, pipe); else if (IS_HASWELL(dev) && pipe == PIPE_A) hsw_undo_trans_edp_pipe_A_crc_wa(dev); + + hsw_enable_ips(crtc); } return 0; -- cgit v0.10.2 From 0039a4b357477bd8bcd495e4160974c71657b3e6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 20:52:30 +0300 Subject: drm/i915: Don't claim that we're resetting PCH ADPA register MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit intel_crt_reset() resets the ADPA register on all gen5+ platforms. However the debug message claims it's touching the PCH ADPA register which is clearly not what it does on VLV. Drop the PCH part from the debug message. Signed-off-by: Ville Syrjälä Reviewed-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_crt.c b/drivers/gpu/drm/i915/intel_crt.c index dacaad5..a9af9a4 100644 --- a/drivers/gpu/drm/i915/intel_crt.c +++ b/drivers/gpu/drm/i915/intel_crt.c @@ -775,7 +775,7 @@ static void intel_crt_reset(struct drm_connector *connector) I915_WRITE(crt->adpa_reg, adpa); POSTING_READ(crt->adpa_reg); - DRM_DEBUG_KMS("pch crt adpa set to 0x%x\n", adpa); + DRM_DEBUG_KMS("crt adpa set to 0x%x\n", adpa); crt->force_hotplug_required = 1; } -- cgit v0.10.2 From 142d2eca356af6744c7e4bb577c3dfaadee486fc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 20:52:32 +0300 Subject: drm/i915: Fix chv PCS DW11 register defines MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit I managed to fumble the per spline PCS DW11 register defines in: commit 570e2a747bc06cd8620662c5125ec2dc964c511b Author: Ville Syrjälä Date: Mon Aug 18 14:42:46 2014 +0300 drm/i915: Clear TX FIFO reset master override bits on chv Fortunately the bit in DW0 that was cleared due to this didn't have any effect as long as the bit we meant to clear was already zero. Signed-off-by: Ville Syrjälä Reviewed-by: Jani Nikula [danvet: Fix commit ref as pointed out by Jani.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 6db369a..46cfbc7 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -883,8 +883,8 @@ enum punit_power_well { #define _VLV_PCS23_DW11_CH0 0x042c #define _VLV_PCS01_DW11_CH1 0x262c #define _VLV_PCS23_DW11_CH1 0x282c -#define VLV_PCS01_DW11(ch) _PORT(ch, _VLV_PCS01_DW0_CH0, _VLV_PCS01_DW0_CH1) -#define VLV_PCS23_DW11(ch) _PORT(ch, _VLV_PCS23_DW0_CH0, _VLV_PCS23_DW0_CH1) +#define VLV_PCS01_DW11(ch) _PORT(ch, _VLV_PCS01_DW11_CH0, _VLV_PCS01_DW11_CH1) +#define VLV_PCS23_DW11(ch) _PORT(ch, _VLV_PCS23_DW11_CH0, _VLV_PCS23_DW11_CH1) #define _VLV_PCS_DW12_CH0 0x8230 #define _VLV_PCS_DW12_CH1 0x8430 -- cgit v0.10.2 From fc0412ec0ff5aa7eef694ccde47606caa7c089e0 Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Thu, 16 Oct 2014 16:13:38 +0100 Subject: drm/i915: add runtime PM get/put call in i915_execlists Otherwise we will get WARNs when we read context status registers and the machine is suspended. Testcase: igt/pm_rpm/debugfs-read Signed-off-by: Michel Thierry Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 1e7a7c6..e60d5c2 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1848,6 +1848,8 @@ static int i915_execlists(struct seq_file *m, void *data) if (ret) return ret; + intel_runtime_pm_get(dev_priv); + for_each_ring(ring, dev_priv, ring_id) { struct intel_ctx_submit_request *head_req = NULL; int count = 0; @@ -1899,6 +1901,7 @@ static int i915_execlists(struct seq_file *m, void *data) seq_putc(m, '\n'); } + intel_runtime_pm_put(dev_priv); mutex_unlock(&dev->struct_mutex); return 0; -- cgit v0.10.2 From f7f3d48ac858e4d2f0c71f8479d4ea022a35dd9c Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 22 Oct 2014 11:06:44 +0200 Subject: drm/i915: Fold in intel_mst_port_dp_detect The indirection here seems to serve no purpose. Probably leftovers from earlier revisions. Spotted while trying to review some mst patches. Cc: Dave Airlie Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index d9a7a78..b03fa90 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -278,7 +278,7 @@ static int intel_dp_mst_get_ddc_modes(struct drm_connector *connector) } static enum drm_connector_status -intel_mst_port_dp_detect(struct drm_connector *connector) +intel_dp_mst_detect(struct drm_connector *connector, bool force) { struct intel_connector *intel_connector = to_intel_connector(connector); struct intel_dp *intel_dp = intel_connector->mst_port; @@ -286,14 +286,6 @@ intel_mst_port_dp_detect(struct drm_connector *connector) return drm_dp_mst_detect_port(&intel_dp->mst_mgr, intel_connector->port); } -static enum drm_connector_status -intel_dp_mst_detect(struct drm_connector *connector, bool force) -{ - enum drm_connector_status status; - status = intel_mst_port_dp_detect(connector); - return status; -} - static int intel_dp_mst_set_property(struct drm_connector *connector, struct drm_property *property, -- cgit v0.10.2 From 5e56ba45053ab5d7a77b60924e7a15cbbb11dcad Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 17 Oct 2014 08:05:08 -0700 Subject: drm/i915/chv: Use 16 and 32 for low and high drain latency precision. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Current chv spec teels we can only use either 16 or 32 bits as precision. Although in the past VLV went from 16/32 to 32/64 and spec might not be updated, these precision values brings stability and fixes some issues Wayne was facing. Cc: Wayne Boyer Cc: Ville Syrjälä Signed-off-by: Rodrigo Vivi Tested-by: Wayne Boyer Reviewed-by: Ville Syrjälä [danvet: Sprinkle const as requested by Ville.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 46cfbc7..1e6f7fb 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4054,17 +4054,18 @@ enum punit_power_well { #define DSPFW_PLANEA_WM1_HI_MASK (1<<0) /* drain latency register values*/ +#define DRAIN_LATENCY_PRECISION_16 16 #define DRAIN_LATENCY_PRECISION_32 32 #define DRAIN_LATENCY_PRECISION_64 64 #define VLV_DDL(pipe) (VLV_DISPLAY_BASE + 0x70050 + 4 * (pipe)) -#define DDL_CURSOR_PRECISION_64 (1<<31) -#define DDL_CURSOR_PRECISION_32 (0<<31) +#define DDL_CURSOR_PRECISION_HIGH (1<<31) +#define DDL_CURSOR_PRECISION_LOW (0<<31) #define DDL_CURSOR_SHIFT 24 -#define DDL_SPRITE_PRECISION_64(sprite) (1<<(15+8*(sprite))) -#define DDL_SPRITE_PRECISION_32(sprite) (0<<(15+8*(sprite))) +#define DDL_SPRITE_PRECISION_HIGH(sprite) (1<<(15+8*(sprite))) +#define DDL_SPRITE_PRECISION_LOW(sprite) (0<<(15+8*(sprite))) #define DDL_SPRITE_SHIFT(sprite) (8+8*(sprite)) -#define DDL_PLANE_PRECISION_64 (1<<7) -#define DDL_PLANE_PRECISION_32 (0<<7) +#define DDL_PLANE_PRECISION_HIGH (1<<7) +#define DDL_PLANE_PRECISION_LOW (0<<7) #define DDL_PLANE_SHIFT 0 #define DRAIN_LATENCY_MASK 0x7f diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 23d3318..7a69eba 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -1345,6 +1345,7 @@ static bool vlv_compute_drain_latency(struct drm_crtc *crtc, int *prec_mult, int *drain_latency) { + struct drm_device *dev = crtc->dev; int entries; int clock = to_intel_crtc(crtc)->config.adjusted_mode.crtc_clock; @@ -1355,8 +1356,12 @@ static bool vlv_compute_drain_latency(struct drm_crtc *crtc, return false; entries = DIV_ROUND_UP(clock, 1000) * pixel_size; - *prec_mult = (entries > 128) ? DRAIN_LATENCY_PRECISION_64 : - DRAIN_LATENCY_PRECISION_32; + if (IS_CHERRYVIEW(dev)) + *prec_mult = (entries > 128) ? DRAIN_LATENCY_PRECISION_32 : + DRAIN_LATENCY_PRECISION_16; + else + *prec_mult = (entries > 128) ? DRAIN_LATENCY_PRECISION_64 : + DRAIN_LATENCY_PRECISION_32; *drain_latency = (64 * (*prec_mult) * 4) / entries; if (*drain_latency > DRAIN_LATENCY_MASK) @@ -1375,15 +1380,18 @@ static bool vlv_compute_drain_latency(struct drm_crtc *crtc, static void vlv_update_drain_latency(struct drm_crtc *crtc) { - struct drm_i915_private *dev_priv = crtc->dev->dev_private; + struct drm_device *dev = crtc->dev; + struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); int pixel_size; int drain_latency; enum pipe pipe = intel_crtc->pipe; int plane_prec, prec_mult, plane_dl; + const int high_precision = IS_CHERRYVIEW(dev) ? + DRAIN_LATENCY_PRECISION_32 : DRAIN_LATENCY_PRECISION_64; - plane_dl = I915_READ(VLV_DDL(pipe)) & ~(DDL_PLANE_PRECISION_64 | - DRAIN_LATENCY_MASK | DDL_CURSOR_PRECISION_64 | + plane_dl = I915_READ(VLV_DDL(pipe)) & ~(DDL_PLANE_PRECISION_HIGH | + DRAIN_LATENCY_MASK | DDL_CURSOR_PRECISION_HIGH | (DRAIN_LATENCY_MASK << DDL_CURSOR_SHIFT)); if (!intel_crtc_active(crtc)) { @@ -1394,9 +1402,9 @@ static void vlv_update_drain_latency(struct drm_crtc *crtc) /* Primary plane Drain Latency */ pixel_size = crtc->primary->fb->bits_per_pixel / 8; /* BPP */ if (vlv_compute_drain_latency(crtc, pixel_size, &prec_mult, &drain_latency)) { - plane_prec = (prec_mult == DRAIN_LATENCY_PRECISION_64) ? - DDL_PLANE_PRECISION_64 : - DDL_PLANE_PRECISION_32; + plane_prec = (prec_mult == high_precision) ? + DDL_PLANE_PRECISION_HIGH : + DDL_PLANE_PRECISION_LOW; plane_dl |= plane_prec | drain_latency; } @@ -1408,9 +1416,9 @@ static void vlv_update_drain_latency(struct drm_crtc *crtc) /* Program cursor DL only if it is enabled */ if (intel_crtc->cursor_base && vlv_compute_drain_latency(crtc, pixel_size, &prec_mult, &drain_latency)) { - plane_prec = (prec_mult == DRAIN_LATENCY_PRECISION_64) ? - DDL_CURSOR_PRECISION_64 : - DDL_CURSOR_PRECISION_32; + plane_prec = (prec_mult == high_precision) ? + DDL_CURSOR_PRECISION_HIGH : + DDL_CURSOR_PRECISION_LOW; plane_dl |= plane_prec | (drain_latency << DDL_CURSOR_SHIFT); } @@ -1578,15 +1586,17 @@ static void valleyview_update_sprite_wm(struct drm_plane *plane, int plane_prec; int sprite_dl; int prec_mult; + const int high_precision = IS_CHERRYVIEW(dev) ? + DRAIN_LATENCY_PRECISION_32 : DRAIN_LATENCY_PRECISION_64; - sprite_dl = I915_READ(VLV_DDL(pipe)) & ~(DDL_SPRITE_PRECISION_64(sprite) | + sprite_dl = I915_READ(VLV_DDL(pipe)) & ~(DDL_SPRITE_PRECISION_HIGH(sprite) | (DRAIN_LATENCY_MASK << DDL_SPRITE_SHIFT(sprite))); if (enabled && vlv_compute_drain_latency(crtc, pixel_size, &prec_mult, &drain_latency)) { - plane_prec = (prec_mult == DRAIN_LATENCY_PRECISION_64) ? - DDL_SPRITE_PRECISION_64(sprite) : - DDL_SPRITE_PRECISION_32(sprite); + plane_prec = (prec_mult == high_precision) ? + DDL_SPRITE_PRECISION_HIGH(sprite) : + DDL_SPRITE_PRECISION_LOW(sprite); sprite_dl |= plane_prec | (drain_latency << DDL_SPRITE_SHIFT(sprite)); } -- cgit v0.10.2 From 6d729bff304f3b81062e21cd333a639fda4244b3 Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Tue, 7 Oct 2014 16:11:11 -0300 Subject: drm/i915: only run hsw_power_well_post_enable when really needed MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Only run it after we actually enable the power well. When we're booting the machine there are cases where we run hsw_power_well_post_enable without really needing, and even though this is not causing any real bugs, it is unneeded and causes confusion to people debugging interrupts. Signed-off-by: Paulo Zanoni Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index 36749b9..39c33e0 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -221,9 +221,9 @@ static void hsw_set_power_well(struct drm_i915_private *dev_priv, if (wait_for((I915_READ(HSW_PWR_WELL_DRIVER) & HSW_PWR_WELL_STATE_ENABLED), 20)) DRM_ERROR("Timeout enabling power well\n"); + hsw_power_well_post_enable(dev_priv); } - hsw_power_well_post_enable(dev_priv); } else { if (enable_requested) { I915_WRITE(HSW_PWR_WELL_DRIVER, 0); -- cgit v0.10.2 From 43d942a780efe72b426d30dc112f5eb2fc7eec3a Mon Sep 17 00:00:00 2001 From: Yu Zhang Date: Thu, 23 Oct 2014 15:28:24 +0800 Subject: drm/i915: use macros to assign mmio access functions This is beautification prep work since vgt will add even more special cases. With these macros it's much easier to see what's going on really. Signed-off-by: Yu Zhang [danvet: #undef the temporary macros after the function again. And write a commit message.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 0b0f4f8..964805c 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -823,6 +823,22 @@ __gen4_write(64) #undef REG_WRITE_FOOTER #undef REG_WRITE_HEADER +#define ASSIGN_WRITE_MMIO_VFUNCS(x) \ +do { \ + dev_priv->uncore.funcs.mmio_writeb = x##_write8; \ + dev_priv->uncore.funcs.mmio_writew = x##_write16; \ + dev_priv->uncore.funcs.mmio_writel = x##_write32; \ + dev_priv->uncore.funcs.mmio_writeq = x##_write64; \ +} while (0) + +#define ASSIGN_READ_MMIO_VFUNCS(x) \ +do { \ + dev_priv->uncore.funcs.mmio_readb = x##_read8; \ + dev_priv->uncore.funcs.mmio_readw = x##_read16; \ + dev_priv->uncore.funcs.mmio_readl = x##_read32; \ + dev_priv->uncore.funcs.mmio_readq = x##_read64; \ +} while (0) + void intel_uncore_init(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -879,76 +895,42 @@ void intel_uncore_init(struct drm_device *dev) switch (INTEL_INFO(dev)->gen) { default: if (IS_CHERRYVIEW(dev)) { - dev_priv->uncore.funcs.mmio_writeb = chv_write8; - dev_priv->uncore.funcs.mmio_writew = chv_write16; - dev_priv->uncore.funcs.mmio_writel = chv_write32; - dev_priv->uncore.funcs.mmio_writeq = chv_write64; - dev_priv->uncore.funcs.mmio_readb = chv_read8; - dev_priv->uncore.funcs.mmio_readw = chv_read16; - dev_priv->uncore.funcs.mmio_readl = chv_read32; - dev_priv->uncore.funcs.mmio_readq = chv_read64; + ASSIGN_WRITE_MMIO_VFUNCS(chv); + ASSIGN_READ_MMIO_VFUNCS(chv); } else { - dev_priv->uncore.funcs.mmio_writeb = gen8_write8; - dev_priv->uncore.funcs.mmio_writew = gen8_write16; - dev_priv->uncore.funcs.mmio_writel = gen8_write32; - dev_priv->uncore.funcs.mmio_writeq = gen8_write64; - dev_priv->uncore.funcs.mmio_readb = gen6_read8; - dev_priv->uncore.funcs.mmio_readw = gen6_read16; - dev_priv->uncore.funcs.mmio_readl = gen6_read32; - dev_priv->uncore.funcs.mmio_readq = gen6_read64; + ASSIGN_WRITE_MMIO_VFUNCS(gen8); + ASSIGN_READ_MMIO_VFUNCS(gen6); } break; case 7: case 6: if (IS_HASWELL(dev)) { - dev_priv->uncore.funcs.mmio_writeb = hsw_write8; - dev_priv->uncore.funcs.mmio_writew = hsw_write16; - dev_priv->uncore.funcs.mmio_writel = hsw_write32; - dev_priv->uncore.funcs.mmio_writeq = hsw_write64; + ASSIGN_WRITE_MMIO_VFUNCS(hsw); } else { - dev_priv->uncore.funcs.mmio_writeb = gen6_write8; - dev_priv->uncore.funcs.mmio_writew = gen6_write16; - dev_priv->uncore.funcs.mmio_writel = gen6_write32; - dev_priv->uncore.funcs.mmio_writeq = gen6_write64; + ASSIGN_WRITE_MMIO_VFUNCS(gen6); } if (IS_VALLEYVIEW(dev)) { - dev_priv->uncore.funcs.mmio_readb = vlv_read8; - dev_priv->uncore.funcs.mmio_readw = vlv_read16; - dev_priv->uncore.funcs.mmio_readl = vlv_read32; - dev_priv->uncore.funcs.mmio_readq = vlv_read64; + ASSIGN_READ_MMIO_VFUNCS(vlv); } else { - dev_priv->uncore.funcs.mmio_readb = gen6_read8; - dev_priv->uncore.funcs.mmio_readw = gen6_read16; - dev_priv->uncore.funcs.mmio_readl = gen6_read32; - dev_priv->uncore.funcs.mmio_readq = gen6_read64; + ASSIGN_READ_MMIO_VFUNCS(gen6); } break; case 5: - dev_priv->uncore.funcs.mmio_writeb = gen5_write8; - dev_priv->uncore.funcs.mmio_writew = gen5_write16; - dev_priv->uncore.funcs.mmio_writel = gen5_write32; - dev_priv->uncore.funcs.mmio_writeq = gen5_write64; - dev_priv->uncore.funcs.mmio_readb = gen5_read8; - dev_priv->uncore.funcs.mmio_readw = gen5_read16; - dev_priv->uncore.funcs.mmio_readl = gen5_read32; - dev_priv->uncore.funcs.mmio_readq = gen5_read64; + ASSIGN_WRITE_MMIO_VFUNCS(gen5); + ASSIGN_READ_MMIO_VFUNCS(gen5); break; case 4: case 3: case 2: - dev_priv->uncore.funcs.mmio_writeb = gen4_write8; - dev_priv->uncore.funcs.mmio_writew = gen4_write16; - dev_priv->uncore.funcs.mmio_writel = gen4_write32; - dev_priv->uncore.funcs.mmio_writeq = gen4_write64; - dev_priv->uncore.funcs.mmio_readb = gen4_read8; - dev_priv->uncore.funcs.mmio_readw = gen4_read16; - dev_priv->uncore.funcs.mmio_readl = gen4_read32; - dev_priv->uncore.funcs.mmio_readq = gen4_read64; + ASSIGN_WRITE_MMIO_VFUNCS(gen4); + ASSIGN_READ_MMIO_VFUNCS(gen4); break; } } +#undef ASSIGN_WRITE_MMIO_VFUNCS +#undef ASSIGN_READ_MMIO_VFUNCS void intel_uncore_fini(struct drm_device *dev) { -- cgit v0.10.2 From 11b5d5112c51f8cf4ac51f9f063bb76992a5ebe7 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 29 Sep 2014 15:31:26 +0200 Subject: drm/i915: Correctly reject invalid flags for wait_ioctl Not having checks for this isn't good. I've checked igt and libdrm and they all already clear flags properly. So we're lucky and should be able to sneak this ABI clarification in. Testcase: igt/gem_wait/invalid-flags Signed-off-by: Daniel Vetter Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85280 Reviewed-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 895f988..827edb5 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2810,6 +2810,9 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file) u32 seqno = 0; int ret = 0; + if (args->flags != 0) + return -EINVAL; + ret = i915_mutex_lock_interruptible(dev); if (ret) return ret; -- cgit v0.10.2 From 4398ad454a976d69a30fd76aee21fc82b9489054 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 23 Oct 2014 07:41:34 -0700 Subject: drm/i915: Add rotation support for cursor plane (v5) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The cursor plane also supports 180 degree rotation. Add a new "cursor-rotation" property on the crtc which controls this. Unlike sprites, the cursor has a fixed size, so if you have a small cursor image with the rest of the bo filled by transparent pixels, simply flipping the rotation property will cause the visible part of the cursor to shift. This is something to keep in mind when using cursor rotation. v2: Fix gen4/vlv by offsetting the base address appropriately v3: Removing cursor-rotation property and using rotation property on cursor plane. v4: Changing the author name back to Ville. v5 (by Matt Roper): Slight tweaking to apply against latest di-nightly codebase. Cc: Sagar Kamble Signed-off-by: Ville Syrjälä Signed-off-by: Sonika Jindal Reviewed-by: Matt Roper Tested-by (IVB): Matt Roper Signed-off-by: Matt Roper Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 1e6f7fb..ea84e1e 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4208,6 +4208,7 @@ enum punit_power_well { #define MCURSOR_PIPE_A 0x00 #define MCURSOR_PIPE_B (1 << 28) #define MCURSOR_GAMMA_ENABLE (1 << 26) +#define CURSOR_ROTATE_180 (1<<15) #define CURSOR_TRICKLE_FEED_DISABLE (1 << 14) #define _CURABASE 0x70084 #define _CURAPOS 0x70088 diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index fb9a434..1d2fa7f 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8343,6 +8343,9 @@ static void i9xx_update_cursor(struct drm_crtc *crtc, u32 base) cntl |= CURSOR_PIPE_CSC_ENABLE; } + if (to_intel_plane(crtc->cursor)->rotation == BIT(DRM_ROTATE_180)) + cntl |= CURSOR_ROTATE_180; + if (intel_crtc->cursor_cntl != cntl) { I915_WRITE(CURCNTR(pipe), cntl); POSTING_READ(CURCNTR(pipe)); @@ -8400,6 +8403,13 @@ static void intel_crtc_update_cursor(struct drm_crtc *crtc, I915_WRITE(CURPOS(pipe), pos); + /* ILK+ do this automagically */ + if (HAS_GMCH_DISPLAY(dev) && + to_intel_plane(crtc->cursor)->rotation == BIT(DRM_ROTATE_180)) { + base += (intel_crtc->cursor_height * + intel_crtc->cursor_width - 1) * 4; + } + if (IS_845G(dev) || IS_I865G(dev)) i845_update_cursor(crtc, base); else @@ -12010,6 +12020,7 @@ static const struct drm_plane_funcs intel_cursor_plane_funcs = { .update_plane = intel_cursor_plane_update, .disable_plane = intel_cursor_plane_disable, .destroy = intel_plane_destroy, + .set_property = intel_plane_set_property, }; static struct drm_plane *intel_cursor_plane_create(struct drm_device *dev, @@ -12025,12 +12036,26 @@ static struct drm_plane *intel_cursor_plane_create(struct drm_device *dev, cursor->max_downscale = 1; cursor->pipe = pipe; cursor->plane = pipe; + cursor->rotation = BIT(DRM_ROTATE_0); drm_universal_plane_init(dev, &cursor->base, 0, &intel_cursor_plane_funcs, intel_cursor_formats, ARRAY_SIZE(intel_cursor_formats), DRM_PLANE_TYPE_CURSOR); + + if (INTEL_INFO(dev)->gen >= 4) { + if (!dev->mode_config.rotation_property) + dev->mode_config.rotation_property = + drm_mode_create_rotation_property(dev, + BIT(DRM_ROTATE_0) | + BIT(DRM_ROTATE_180)); + if (dev->mode_config.rotation_property) + drm_object_attach_property(&cursor->base.base, + dev->mode_config.rotation_property, + cursor->rotation); + } + return &cursor->base; } -- cgit v0.10.2 From 22a916aaa187946e8df724ab7838a0c13b45a9f4 Mon Sep 17 00:00:00 2001 From: Arun Siluvery Date: Wed, 22 Oct 2014 18:59:52 +0100 Subject: drm/i915: Emit even number of dwords when emitting LRIs The number of DWords should be even when doing ring emits as command sequences require QWord alignment. There was some discussion about the maximum length of the MI_LRI command. Quoting Mika "I did some test with bdw: "The maximum is 128 writes, resulting the 8 bit length field of the command being 0xff, thus following the spec. The 128'th write went through. "Perhaps the max command length is then less in older gens? "Perhaps WARN_ON(x > 128) in MI_LOAD_REGISTER_IMM would be in place but one needs minor tweak to command parser a bit also then. #define I915_MAX_WA_REGS 16 keeps us safe for now atleast." Ville commented that on pre-gen6 the length field seems to be restricted to 0x3f though. So for all cases we should be ok. v2: user LRI variant that can write multiple regs in one go (Damien). We can simply insert one NOP at the end instead of one per register write. Cc: Mika Kuoppala Signed-off-by: Arun Siluvery Reviewed-by: Damien Lespiau [danvet: Add a summary of the MI_LRI length discussion.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 5f935d4..603148e 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -680,15 +680,16 @@ static int intel_ring_workarounds_emit(struct intel_engine_cs *ring) if (ret) return ret; - ret = intel_ring_begin(ring, w->count * 3); + ret = intel_ring_begin(ring, (w->count * 2 + 2)); if (ret) return ret; + intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(w->count)); for (i = 0; i < w->count; i++) { - intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1)); intel_ring_emit(ring, w->reg[i].addr); intel_ring_emit(ring, w->reg[i].value); } + intel_ring_emit(ring, MI_NOOP); intel_ring_advance(ring); -- cgit v0.10.2 From c3c09c9522f806bd12d0c85dfdb8d2b8578d0da1 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:15 +0300 Subject: drm/i915: factor out i915_drm_suspend_late MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This is needed by an upcoming patch fixing the switcheroo/legacy suspend paths. No functional change. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index ffd672e..753d44a 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -630,6 +630,25 @@ static int i915_drm_freeze(struct drm_device *dev) return 0; } +static int i915_drm_suspend_late(struct drm_device *drm_dev) +{ + struct drm_i915_private *dev_priv = drm_dev->dev_private; + int ret; + + ret = intel_suspend_complete(dev_priv); + + if (ret) { + DRM_ERROR("Suspend complete failed: %d\n", ret); + + return ret; + } + + pci_disable_device(drm_dev->pdev); + pci_set_power_state(drm_dev->pdev, PCI_D3hot); + + return 0; +} + int i915_suspend(struct drm_device *dev, pm_message_t state) { int error; @@ -955,8 +974,6 @@ static int i915_pm_suspend_late(struct device *dev) { struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); - struct drm_i915_private *dev_priv = drm_dev->dev_private; - int ret; /* * We have a suspedn ordering issue with the snd-hda driver also @@ -970,16 +987,7 @@ static int i915_pm_suspend_late(struct device *dev) if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; - ret = intel_suspend_complete(dev_priv); - - if (ret) - DRM_ERROR("Suspend complete failed: %d\n", ret); - else { - pci_disable_device(pdev); - pci_set_power_state(pdev, PCI_D3hot); - } - - return ret; + return i915_drm_suspend_late(drm_dev); } static int i915_pm_resume_early(struct device *dev) -- cgit v0.10.2 From f2888fabfbbccc0acb4d21713b91d557edc1e64f Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:16 +0300 Subject: drm/i915: unify legacy S3 suspend and S4 freeze handlers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit i915_suspend() is called from the DRM legacy S3 suspend/S4 freeze paths and the switcheroo suspend path. For switcheroo we only ever need to perform a full suspend (PM_EVENT_SUSPEND) and for the DRM legacy path we can handle the S4 freeze (PM_EVENT_FREEZE) the same way as S3 suspend. The only difference atm between suspend and freeze is that during freeze we don't disable the PCI device, but there is no reason why we can't do so. So unify the two cases to reduce complexity. Note that for the DRM legacy case the thaw event is not handled, so we disable the display before creating the hibernation image and it won't get re-enabled until reboot. We could fix this leaving the display enabled for the image creation/writing (if we care enough about UMS), but this can be done as a follow-up. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 753d44a..56c7570 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -670,11 +670,9 @@ int i915_suspend(struct drm_device *dev, pm_message_t state) if (error) return error; - if (state.event == PM_EVENT_SUSPEND) { - /* Shut down the device */ - pci_disable_device(dev->pdev); - pci_set_power_state(dev->pdev, PCI_D3hot); - } + /* Shut down the device */ + pci_disable_device(dev->pdev); + pci_set_power_state(dev->pdev, PCI_D3hot); return 0; } -- cgit v0.10.2 From 50a0072f3829d36767f3359649d224f18f7987f7 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:17 +0300 Subject: drm/i915: propagate error from legacy resume handler MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 56c7570..524c3df 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -810,10 +810,13 @@ int i915_resume(struct drm_device *dev) static int i915_resume_legacy(struct drm_device *dev) { - i915_resume_early(dev); - i915_resume(dev); + int ret; - return 0; + ret = i915_resume_early(dev); + if (ret) + return ret; + + return i915_resume(dev); } /** -- cgit v0.10.2 From 5a17514ed924201f1465e4a2f5209bdb192d8dff Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:18 +0300 Subject: drm/i915: vlv: fix switcheroo/legacy suspend/resume MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit During switcheroo/legacy suspend we don't call the suspend_late handler but when resuming afterwards we call resume_early. This happened to work so far, since suspend_late only disabled the PCI device. This changed in commit 016970beb05da6285c2f3ed2bee1c676cb75972e Author: Sagar Kamble Date: Wed Aug 13 23:07:06 2014 +0530 drm/i915: Sharing platform specific sequence between runtime and system susp after which we also saved/restored the VLV Gunit HW state in suspend_late/resume_early. So now since we don't save the state during suspend a following resume will restore a corrupted state. Fix this by calling the suspend_late handler during both switcheroo and legacy suspend. CC: Sagar Kamble Signed-off-by: Imre Deak Reviewed-by: Sagar Kamble Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 524c3df..203bd82 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -670,11 +670,7 @@ int i915_suspend(struct drm_device *dev, pm_message_t state) if (error) return error; - /* Shut down the device */ - pci_disable_device(dev->pdev); - pci_set_power_state(dev->pdev, PCI_D3hot); - - return 0; + return i915_drm_suspend_late(dev); } static int i915_drm_thaw_early(struct drm_device *dev) @@ -790,7 +786,7 @@ static int i915_resume_early(struct drm_device *dev) return i915_drm_thaw_early(dev); } -int i915_resume(struct drm_device *dev) +static int i915_drm_resume(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; int ret; @@ -816,7 +812,12 @@ static int i915_resume_legacy(struct drm_device *dev) if (ret) return ret; - return i915_resume(dev); + return i915_drm_resume(dev); +} + +int i915_resume(struct drm_device *dev) +{ + return i915_resume_legacy(dev); } /** @@ -1004,7 +1005,7 @@ static int i915_pm_resume(struct device *dev) struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); - return i915_resume(drm_dev); + return i915_drm_resume(drm_dev); } static int i915_pm_freeze(struct device *dev) -- cgit v0.10.2 From 097dd83763bfe8e8643957c0b90819742fb01cd9 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:19 +0300 Subject: drm/i915: fix S4 suspend while switcheroo state is off MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If the device is suspended already through the switcheroo interface we shouldn't suspend it again or resume it after suspend. We have the corresponding check for S3 suspend already, add it for all the other S3 and S4 handlers. Also move the check from i915_resume_early() to i915_resume_legacy(), so that it's done in the high level handler for all PM events. v2: - fix the resume path too, we don't need to special case there DRM_SWITCH_POWER_OFF with the device being enabled (in which case we'd have to disable the device), since that never happens (Ville) Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 203bd82..6993632 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -766,9 +766,6 @@ static int i915_drm_thaw(struct drm_device *dev) static int i915_resume_early(struct drm_device *dev) { - if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) - return 0; - /* * We have a resume ordering issue with the snd-hda driver also * requiring our device to be power up. Due to the lack of a @@ -808,6 +805,9 @@ static int i915_resume_legacy(struct drm_device *dev) { int ret; + if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) + return 0; + ret = i915_resume_early(dev); if (ret) return ret; @@ -997,6 +997,9 @@ static int i915_pm_resume_early(struct device *dev) struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); + if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) + return 0; + return i915_resume_early(drm_dev); } @@ -1005,6 +1008,9 @@ static int i915_pm_resume(struct device *dev) struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); + if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) + return 0; + return i915_drm_resume(drm_dev); } @@ -1018,6 +1024,9 @@ static int i915_pm_freeze(struct device *dev) return -ENODEV; } + if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) + return 0; + return i915_drm_freeze(drm_dev); } @@ -1027,6 +1036,9 @@ static int i915_pm_freeze_late(struct device *dev) struct drm_device *drm_dev = pci_get_drvdata(pdev); struct drm_i915_private *dev_priv = drm_dev->dev_private; + if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) + return 0; + return intel_suspend_complete(dev_priv); } @@ -1035,6 +1047,9 @@ static int i915_pm_thaw_early(struct device *dev) struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); + if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) + return 0; + return i915_drm_thaw_early(drm_dev); } @@ -1043,6 +1058,9 @@ static int i915_pm_thaw(struct device *dev) struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); + if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) + return 0; + return i915_drm_thaw(drm_dev); } @@ -1051,6 +1069,9 @@ static int i915_pm_poweroff(struct device *dev) struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); + if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) + return 0; + return i915_drm_freeze(drm_dev); } -- cgit v0.10.2 From f4a12ead50580c17c3641ac1a453e68b5a5195dd Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:20 +0300 Subject: drm/i915: remove unused restore_gtt_mappings optimization during suspend MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The logic to skip restoring GTT mappings was added to speed up suspend/resume, but not on old GENs where not restoring them caused problems. The check for old GENs is based on the existence of OpRegion, but this doesn't work since opregion is initialized only after the check. So we end up always restoring the mappings. On my BYT - which has OpRegion - skipping restoring the mappings during suspend doesn't work, I get a GPU hang after resume. Also the logic of when to allow the optimization during S4 is reversed: we should allow it during S4 thaw but not during S4 restore, but atm we have it the other way around in the code. Since correctness wins over optimal code and since the optimization wasn't used anyway I decided not to try to fix it at this point, but just remove it. This allows us to unify the S3 and S4 handlers in the following patches. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 6993632..9c934be 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -689,12 +689,11 @@ static int i915_drm_thaw_early(struct drm_device *dev) return ret; } -static int __i915_drm_thaw(struct drm_device *dev, bool restore_gtt_mappings) +static int __i915_drm_thaw(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - if (drm_core_check_feature(dev, DRIVER_MODESET) && - restore_gtt_mappings) { + if (drm_core_check_feature(dev, DRIVER_MODESET)) { mutex_lock(&dev->struct_mutex); i915_gem_restore_gtt_mappings(dev); mutex_unlock(&dev->struct_mutex); @@ -761,7 +760,7 @@ static int i915_drm_thaw(struct drm_device *dev) if (drm_core_check_feature(dev, DRIVER_MODESET)) i915_check_and_clear_faults(dev); - return __i915_drm_thaw(dev, true); + return __i915_drm_thaw(dev); } static int i915_resume_early(struct drm_device *dev) @@ -785,15 +784,9 @@ static int i915_resume_early(struct drm_device *dev) static int i915_drm_resume(struct drm_device *dev) { - struct drm_i915_private *dev_priv = dev->dev_private; int ret; - /* - * Platforms with opregion should have sane BIOS, older ones (gen3 and - * earlier) need to restore the GTT mappings since the BIOS might clear - * all our scratch PTEs. - */ - ret = __i915_drm_thaw(dev, !dev_priv->opregion.header); + ret = __i915_drm_thaw(dev); if (ret) return ret; -- cgit v0.10.2 From ed49388384b19ed37567053de7b4087ba990813e Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:21 +0300 Subject: drm/i915: check for GT faults in all resume handlers and driver load time Checking for GT faults is not specific in any way to S4 thaw, so do it also during S3 resume, S4 restore and driver load time. This allows us to unify the Sx handlers in an upcoming patch. v2: - move the check to intel_uncore_early_sanitize(), so we check at driver load time too (Chris) Signed-off-by: Imre Deak Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 9c934be..d0f721f 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -757,9 +757,6 @@ static int __i915_drm_thaw(struct drm_device *dev) static int i915_drm_thaw(struct drm_device *dev) { - if (drm_core_check_feature(dev, DRIVER_MODESET)) - i915_check_and_clear_faults(dev); - return __i915_drm_thaw(dev); } diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 964805c..9427641 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -360,7 +360,8 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore) spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } -void intel_uncore_early_sanitize(struct drm_device *dev, bool restore_forcewake) +static void __intel_uncore_early_sanitize(struct drm_device *dev, + bool restore_forcewake) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -386,6 +387,12 @@ void intel_uncore_early_sanitize(struct drm_device *dev, bool restore_forcewake) intel_uncore_forcewake_reset(dev, restore_forcewake); } +void intel_uncore_early_sanitize(struct drm_device *dev, bool restore_forcewake) +{ + __intel_uncore_early_sanitize(dev, restore_forcewake); + i915_check_and_clear_faults(dev); +} + void intel_uncore_sanitize(struct drm_device *dev) { /* BIOS often leaves RC6 enabled, but disable it for hw init */ @@ -846,7 +853,7 @@ void intel_uncore_init(struct drm_device *dev) setup_timer(&dev_priv->uncore.force_wake_timer, gen6_force_wake_timer, (unsigned long)dev_priv); - intel_uncore_early_sanitize(dev, false); + __intel_uncore_early_sanitize(dev, false); if (IS_VALLEYVIEW(dev)) { dev_priv->uncore.funcs.force_wake_get = __vlv_force_wake_get; @@ -928,6 +935,8 @@ void intel_uncore_init(struct drm_device *dev) ASSIGN_READ_MMIO_VFUNCS(gen4); break; } + + i915_check_and_clear_faults(dev); } #undef ASSIGN_WRITE_MMIO_VFUNCS #undef ASSIGN_READ_MMIO_VFUNCS -- cgit v0.10.2 From ee6f280e7895c838d90e8ef19ba737c8003ae85d Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:22 +0300 Subject: drm/i915: enable output polling during S4 thaw MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit To avoid processing hotplug events we disable connector polling for the duration of S3 suspend. We also disable it for S4 freeze, and keep it disabled after S4 thaw. This won't prevent though hotplug processing, since we re-enable interrupts anyway. There is also no need to prevent it at that time, since we reinitialize everything during thaw, so the device is in a consistent state. So to simplify things enable polling during thaw, which will allow us to handle S4 thaw the same way as S3 resume in an upcoming patch. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index d0f721f..4524c61 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -752,6 +752,8 @@ static int __i915_drm_thaw(struct drm_device *dev) intel_opregion_notify_adapter(dev, PCI_D0); + drm_kms_helper_poll_enable(dev); + return 0; } @@ -781,14 +783,7 @@ static int i915_resume_early(struct drm_device *dev) static int i915_drm_resume(struct drm_device *dev) { - int ret; - - ret = __i915_drm_thaw(dev); - if (ret) - return ret; - - drm_kms_helper_poll_enable(dev); - return 0; + return __i915_drm_thaw(dev); } static int i915_resume_legacy(struct drm_device *dev) -- cgit v0.10.2 From f2476ae65e6159b41168bc41c630e9fbb1d72dde Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:23 +0300 Subject: drm/i915: disable/re-enable PCI device around S4 freeze/thaw MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We already disable everything during S4 freeze, except the PCI device itself. There is no reason why we couldn't disable that too and doing so allows us to unify these handlers in the next patch with the corresponding S3 suspend/resume handlers. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 4524c61..be4cb1d 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1019,12 +1019,11 @@ static int i915_pm_freeze_late(struct device *dev) { struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); - struct drm_i915_private *dev_priv = drm_dev->dev_private; if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; - return intel_suspend_complete(dev_priv); + return i915_drm_suspend_late(drm_dev); } static int i915_pm_thaw_early(struct device *dev) @@ -1035,7 +1034,7 @@ static int i915_pm_thaw_early(struct device *dev) if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; - return i915_drm_thaw_early(drm_dev); + return i915_resume_early(drm_dev); } static int i915_pm_thaw(struct device *dev) -- cgit v0.10.2 From 36d61e6751d16fa80a4355b9708ff24ab7536f89 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:24 +0300 Subject: drm/i915: unify S3 and S4 suspend/resume handlers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The S3 and S4 events are now handled the same way internally, there is no need to keep separate wrapper functions around them. Simply reuse the suspend/resume versions everywhere. No functional change. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index be4cb1d..2190d0f 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -673,22 +673,6 @@ int i915_suspend(struct drm_device *dev, pm_message_t state) return i915_drm_suspend_late(dev); } -static int i915_drm_thaw_early(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int ret; - - ret = intel_resume_prepare(dev_priv, false); - if (ret) - DRM_ERROR("Resume prepare failed: %d,Continuing resume\n", ret); - - intel_uncore_early_sanitize(dev, true); - intel_uncore_sanitize(dev); - intel_power_domains_init_hw(dev_priv); - - return ret; -} - static int __i915_drm_thaw(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -757,13 +741,11 @@ static int __i915_drm_thaw(struct drm_device *dev) return 0; } -static int i915_drm_thaw(struct drm_device *dev) -{ - return __i915_drm_thaw(dev); -} - static int i915_resume_early(struct drm_device *dev) { + struct drm_i915_private *dev_priv = dev->dev_private; + int ret; + /* * We have a resume ordering issue with the snd-hda driver also * requiring our device to be power up. Due to the lack of a @@ -778,7 +760,15 @@ static int i915_resume_early(struct drm_device *dev) pci_set_master(dev->pdev); - return i915_drm_thaw_early(dev); + ret = intel_resume_prepare(dev_priv, false); + if (ret) + DRM_ERROR("Resume prepare failed: %d,Continuing resume\n", ret); + + intel_uncore_early_sanitize(dev, true); + intel_uncore_sanitize(dev); + intel_power_domains_init_hw(dev_priv); + + return ret; } static int i915_drm_resume(struct drm_device *dev) @@ -999,66 +989,6 @@ static int i915_pm_resume(struct device *dev) return i915_drm_resume(drm_dev); } -static int i915_pm_freeze(struct device *dev) -{ - struct pci_dev *pdev = to_pci_dev(dev); - struct drm_device *drm_dev = pci_get_drvdata(pdev); - - if (!drm_dev || !drm_dev->dev_private) { - dev_err(dev, "DRM not initialized, aborting suspend.\n"); - return -ENODEV; - } - - if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) - return 0; - - return i915_drm_freeze(drm_dev); -} - -static int i915_pm_freeze_late(struct device *dev) -{ - struct pci_dev *pdev = to_pci_dev(dev); - struct drm_device *drm_dev = pci_get_drvdata(pdev); - - if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) - return 0; - - return i915_drm_suspend_late(drm_dev); -} - -static int i915_pm_thaw_early(struct device *dev) -{ - struct pci_dev *pdev = to_pci_dev(dev); - struct drm_device *drm_dev = pci_get_drvdata(pdev); - - if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) - return 0; - - return i915_resume_early(drm_dev); -} - -static int i915_pm_thaw(struct device *dev) -{ - struct pci_dev *pdev = to_pci_dev(dev); - struct drm_device *drm_dev = pci_get_drvdata(pdev); - - if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) - return 0; - - return i915_drm_thaw(drm_dev); -} - -static int i915_pm_poweroff(struct device *dev) -{ - struct pci_dev *pdev = to_pci_dev(dev); - struct drm_device *drm_dev = pci_get_drvdata(pdev); - - if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) - return 0; - - return i915_drm_freeze(drm_dev); -} - static int hsw_suspend_complete(struct drm_i915_private *dev_priv) { hsw_enable_pc8(dev_priv); @@ -1618,11 +1548,11 @@ static const struct dev_pm_ops i915_pm_ops = { .suspend_late = i915_pm_suspend_late, .resume_early = i915_pm_resume_early, .resume = i915_pm_resume, - .freeze = i915_pm_freeze, - .freeze_late = i915_pm_freeze_late, - .thaw_early = i915_pm_thaw_early, - .thaw = i915_pm_thaw, - .poweroff = i915_pm_poweroff, + .freeze = i915_pm_suspend, + .freeze_late = i915_pm_suspend_late, + .thaw_early = i915_pm_resume_early, + .thaw = i915_pm_resume, + .poweroff = i915_pm_suspend, .restore_early = i915_pm_resume_early, .restore = i915_pm_resume, .runtime_suspend = intel_runtime_suspend, -- cgit v0.10.2 From 5e365c391aeffe8b53d6952c28a68bd5fc856390 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:25 +0300 Subject: drm/i915: sanitize suspend/resume helper function names MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit By now the S4 freeze/thaw and S3 suspend/resume events are handled the same way, so we can rename the freeze/thaw internal helpers to suspend/resume accordingly to make clearer what the helpers do. Also rename i915_resume_early to i915_drm_resume_early aligning it with the rest of the helper names. No functional change. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 2190d0f..8b2b34c 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -554,7 +554,7 @@ static int intel_suspend_complete(struct drm_i915_private *dev_priv); static int intel_resume_prepare(struct drm_i915_private *dev_priv, bool rpm_resume); -static int i915_drm_freeze(struct drm_device *dev) +static int i915_drm_suspend(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; struct drm_crtc *crtc; @@ -666,14 +666,14 @@ int i915_suspend(struct drm_device *dev, pm_message_t state) if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; - error = i915_drm_freeze(dev); + error = i915_drm_suspend(dev); if (error) return error; return i915_drm_suspend_late(dev); } -static int __i915_drm_thaw(struct drm_device *dev) +static int i915_drm_resume(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -741,7 +741,7 @@ static int __i915_drm_thaw(struct drm_device *dev) return 0; } -static int i915_resume_early(struct drm_device *dev) +static int i915_drm_resume_early(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; int ret; @@ -771,11 +771,6 @@ static int i915_resume_early(struct drm_device *dev) return ret; } -static int i915_drm_resume(struct drm_device *dev) -{ - return __i915_drm_thaw(dev); -} - static int i915_resume_legacy(struct drm_device *dev) { int ret; @@ -783,7 +778,7 @@ static int i915_resume_legacy(struct drm_device *dev) if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; - ret = i915_resume_early(dev); + ret = i915_drm_resume_early(dev); if (ret) return ret; @@ -944,7 +939,7 @@ static int i915_pm_suspend(struct device *dev) if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; - return i915_drm_freeze(drm_dev); + return i915_drm_suspend(drm_dev); } static int i915_pm_suspend_late(struct device *dev) @@ -975,7 +970,7 @@ static int i915_pm_resume_early(struct device *dev) if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; - return i915_resume_early(drm_dev); + return i915_drm_resume_early(drm_dev); } static int i915_pm_resume(struct device *dev) -- cgit v0.10.2 From da2bc1b9db3351addd293e5b82757efe1f77ed1d Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:26 +0300 Subject: drm/i915: add poweroff_late handler MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The suspend_late handler saves some registers and powers off the device, so it doesn't have a big overhead. Calling it at S4 poweroff_late time makes the power off handling identical to the S3 suspend and S4 freeze handling, so do this for consistency. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 8b2b34c..808f659 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1548,6 +1548,7 @@ static const struct dev_pm_ops i915_pm_ops = { .thaw_early = i915_pm_resume_early, .thaw = i915_pm_resume, .poweroff = i915_pm_suspend, + .poweroff_late = i915_pm_suspend_late, .restore_early = i915_pm_resume_early, .restore = i915_pm_resume, .runtime_suspend = intel_runtime_suspend, -- cgit v0.10.2 From fc49b3da0e1f20484ac069982e2dcbf1a53f60af Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:27 +0300 Subject: drm/i915: unify switcheroo and legacy suspend/resume handlers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit By now we handle switcheroo and legacy suspend/resume the same way, so no need to keep separate functions for them. No functional change. Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 1b39807..9a73533 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1275,12 +1275,12 @@ static void i915_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_ dev->switch_power_state = DRM_SWITCH_POWER_CHANGING; /* i915 resume handler doesn't set to D0 */ pci_set_power_state(dev->pdev, PCI_D0); - i915_resume(dev); + i915_resume_legacy(dev); dev->switch_power_state = DRM_SWITCH_POWER_ON; } else { pr_err("switched off\n"); dev->switch_power_state = DRM_SWITCH_POWER_CHANGING; - i915_suspend(dev, pmm); + i915_suspend_legacy(dev, pmm); dev->switch_power_state = DRM_SWITCH_POWER_OFF; } } diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 808f659..bce58d3 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -649,7 +649,7 @@ static int i915_drm_suspend_late(struct drm_device *drm_dev) return 0; } -int i915_suspend(struct drm_device *dev, pm_message_t state) +int i915_suspend_legacy(struct drm_device *dev, pm_message_t state) { int error; @@ -771,7 +771,7 @@ static int i915_drm_resume_early(struct drm_device *dev) return ret; } -static int i915_resume_legacy(struct drm_device *dev) +int i915_resume_legacy(struct drm_device *dev) { int ret; @@ -785,11 +785,6 @@ static int i915_resume_legacy(struct drm_device *dev) return i915_drm_resume(dev); } -int i915_resume(struct drm_device *dev) -{ - return i915_resume_legacy(dev); -} - /** * i915_reset - reset chip after a hang * @dev: drm device to reset @@ -1592,7 +1587,7 @@ static struct drm_driver driver = { .set_busid = drm_pci_set_busid, /* Used in place of i915_pm_ops for non-DRIVER_MODESET */ - .suspend = i915_suspend, + .suspend = i915_suspend_legacy, .resume = i915_resume_legacy, .device_is_agp = i915_driver_device_is_agp, diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 3212d62..059330c 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2221,8 +2221,8 @@ struct drm_i915_cmd_table { extern const struct drm_ioctl_desc i915_ioctls[]; extern int i915_max_ioctl; -extern int i915_suspend(struct drm_device *dev, pm_message_t state); -extern int i915_resume(struct drm_device *dev); +extern int i915_suspend_legacy(struct drm_device *dev, pm_message_t state); +extern int i915_resume_legacy(struct drm_device *dev); extern int i915_master_create(struct drm_device *dev, struct drm_master *master); extern void i915_master_destroy(struct drm_device *dev, struct drm_master *master); -- cgit v0.10.2 From 5545dbbfb3bb7a5ab5ced991e08d1c312326e843 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 23 Oct 2014 19:23:28 +0300 Subject: drm/i915: add comments on what stage a given PM handler is called MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This will hopefully make it easier to navigate the code without the need to consult the full PM documentation. v2: - add a comment that the freeze handler is also called after rebooting - add a comment that the thaw handler is also called to recover from errors (Ville) - add the PM event names (PMSG_THAW etc.) for reference (Ville) - add comments that s0ix can be handled both via system and runtime suspend (Ville) Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index bce58d3..035ec94 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1534,10 +1534,30 @@ static int intel_resume_prepare(struct drm_i915_private *dev_priv, } static const struct dev_pm_ops i915_pm_ops = { + /* + * S0ix (via system suspend) and S3 event handlers [PMSG_SUSPEND, + * PMSG_RESUME] + */ .suspend = i915_pm_suspend, .suspend_late = i915_pm_suspend_late, .resume_early = i915_pm_resume_early, .resume = i915_pm_resume, + + /* + * S4 event handlers + * @freeze, @freeze_late : called (1) before creating the + * hibernation image [PMSG_FREEZE] and + * (2) after rebooting, before restoring + * the image [PMSG_QUIESCE] + * @thaw, @thaw_early : called (1) after creating the hibernation + * image, before writing it [PMSG_THAW] + * and (2) after failing to create or + * restore the image [PMSG_RECOVER] + * @poweroff, @poweroff_late: called after writing the hibernation + * image, before rebooting [PMSG_HIBERNATE] + * @restore, @restore_early : called after rebooting and restoring the + * hibernation image [PMSG_RESTORE] + */ .freeze = i915_pm_suspend, .freeze_late = i915_pm_suspend_late, .thaw_early = i915_pm_resume_early, @@ -1546,6 +1566,8 @@ static const struct dev_pm_ops i915_pm_ops = { .poweroff_late = i915_pm_suspend_late, .restore_early = i915_pm_resume_early, .restore = i915_pm_resume, + + /* S0ix (via runtime suspend) event handlers */ .runtime_suspend = intel_runtime_suspend, .runtime_resume = intel_runtime_resume, }; -- cgit v0.10.2 From 3eebaec630c2413a5e67bb7f49f0c6a53069a399 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 24 Oct 2014 16:45:21 +0200 Subject: drm/i915: Update DRIVER_DATE to 20141024 Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 059330c..583c97d 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -55,7 +55,7 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20141003" +#define DRIVER_DATE "20141024" enum pipe { INVALID_PIPE = -1, -- cgit v0.10.2 From 3b5b9932fb547a7c89169a1c43a73beee7183ed0 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 31 Oct 2014 14:39:11 +0000 Subject: drm: Add a note to drm_property_create() about property lifetime Cc: Chandra Konduru Cc: Daniel Vetter Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 4081d7a..0f3c24c 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -3417,6 +3417,10 @@ void drm_fb_release(struct drm_file *priv) * object with drm_object_attach_property. The returned property object must be * freed with drm_property_destroy. * + * Note that the DRM core keeps a per-device list of properties and that, if + * drm_mode_config_cleanup() is called, it will destroy all properties created + * by the driver. + * * Returns: * A pointer to the newly created property on success, NULL on failure. */ -- cgit v0.10.2 From 9d5fab402d3b6fc3d4b895fbb0bde4e4f2378167 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 31 Oct 2014 14:39:12 +0000 Subject: drm/i915: Don't destroy DRM properties in the driver When drm properties are created, they are added to mode_config.property_list, which is then used in drm_mode_config_cleanup() to destroy every single property created by the driver. Cc: Chandra Konduru Cc: Daniel Vetter Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_sdvo.c b/drivers/gpu/drm/i915/intel_sdvo.c index 9350edd..6d7a277 100644 --- a/drivers/gpu/drm/i915/intel_sdvo.c +++ b/drivers/gpu/drm/i915/intel_sdvo.c @@ -1991,57 +1991,10 @@ static int intel_sdvo_get_modes(struct drm_connector *connector) return !list_empty(&connector->probed_modes); } -static void -intel_sdvo_destroy_enhance_property(struct drm_connector *connector) -{ - struct intel_sdvo_connector *intel_sdvo_connector = to_intel_sdvo_connector(connector); - struct drm_device *dev = connector->dev; - - if (intel_sdvo_connector->left) - drm_property_destroy(dev, intel_sdvo_connector->left); - if (intel_sdvo_connector->right) - drm_property_destroy(dev, intel_sdvo_connector->right); - if (intel_sdvo_connector->top) - drm_property_destroy(dev, intel_sdvo_connector->top); - if (intel_sdvo_connector->bottom) - drm_property_destroy(dev, intel_sdvo_connector->bottom); - if (intel_sdvo_connector->hpos) - drm_property_destroy(dev, intel_sdvo_connector->hpos); - if (intel_sdvo_connector->vpos) - drm_property_destroy(dev, intel_sdvo_connector->vpos); - if (intel_sdvo_connector->saturation) - drm_property_destroy(dev, intel_sdvo_connector->saturation); - if (intel_sdvo_connector->contrast) - drm_property_destroy(dev, intel_sdvo_connector->contrast); - if (intel_sdvo_connector->hue) - drm_property_destroy(dev, intel_sdvo_connector->hue); - if (intel_sdvo_connector->sharpness) - drm_property_destroy(dev, intel_sdvo_connector->sharpness); - if (intel_sdvo_connector->flicker_filter) - drm_property_destroy(dev, intel_sdvo_connector->flicker_filter); - if (intel_sdvo_connector->flicker_filter_2d) - drm_property_destroy(dev, intel_sdvo_connector->flicker_filter_2d); - if (intel_sdvo_connector->flicker_filter_adaptive) - drm_property_destroy(dev, intel_sdvo_connector->flicker_filter_adaptive); - if (intel_sdvo_connector->tv_luma_filter) - drm_property_destroy(dev, intel_sdvo_connector->tv_luma_filter); - if (intel_sdvo_connector->tv_chroma_filter) - drm_property_destroy(dev, intel_sdvo_connector->tv_chroma_filter); - if (intel_sdvo_connector->dot_crawl) - drm_property_destroy(dev, intel_sdvo_connector->dot_crawl); - if (intel_sdvo_connector->brightness) - drm_property_destroy(dev, intel_sdvo_connector->brightness); -} - static void intel_sdvo_destroy(struct drm_connector *connector) { struct intel_sdvo_connector *intel_sdvo_connector = to_intel_sdvo_connector(connector); - if (intel_sdvo_connector->tv_format) - drm_property_destroy(connector->dev, - intel_sdvo_connector->tv_format); - - intel_sdvo_destroy_enhance_property(connector); drm_connector_cleanup(connector); kfree(intel_sdvo_connector); } -- cgit v0.10.2 From ba8ce31f438c6bdd1d37b07cfea36d248c7f64d0 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 31 Oct 2014 14:39:13 +0000 Subject: drm/gma500: Don't destroy DRM properties in the driver When drm properties are created, they are added to mode_config.property_list which is then used in drm_mode_config_cleanup() to destroy every single property created by the driver. Cc: Patrik Jakobsson Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/gma500/psb_intel_sdvo.c b/drivers/gpu/drm/gma500/psb_intel_sdvo.c index 0be96fd..58529ce 100644 --- a/drivers/gpu/drm/gma500/psb_intel_sdvo.c +++ b/drivers/gpu/drm/gma500/psb_intel_sdvo.c @@ -1631,57 +1631,8 @@ static int psb_intel_sdvo_get_modes(struct drm_connector *connector) return !list_empty(&connector->probed_modes); } -static void -psb_intel_sdvo_destroy_enhance_property(struct drm_connector *connector) -{ - struct psb_intel_sdvo_connector *psb_intel_sdvo_connector = to_psb_intel_sdvo_connector(connector); - struct drm_device *dev = connector->dev; - - if (psb_intel_sdvo_connector->left) - drm_property_destroy(dev, psb_intel_sdvo_connector->left); - if (psb_intel_sdvo_connector->right) - drm_property_destroy(dev, psb_intel_sdvo_connector->right); - if (psb_intel_sdvo_connector->top) - drm_property_destroy(dev, psb_intel_sdvo_connector->top); - if (psb_intel_sdvo_connector->bottom) - drm_property_destroy(dev, psb_intel_sdvo_connector->bottom); - if (psb_intel_sdvo_connector->hpos) - drm_property_destroy(dev, psb_intel_sdvo_connector->hpos); - if (psb_intel_sdvo_connector->vpos) - drm_property_destroy(dev, psb_intel_sdvo_connector->vpos); - if (psb_intel_sdvo_connector->saturation) - drm_property_destroy(dev, psb_intel_sdvo_connector->saturation); - if (psb_intel_sdvo_connector->contrast) - drm_property_destroy(dev, psb_intel_sdvo_connector->contrast); - if (psb_intel_sdvo_connector->hue) - drm_property_destroy(dev, psb_intel_sdvo_connector->hue); - if (psb_intel_sdvo_connector->sharpness) - drm_property_destroy(dev, psb_intel_sdvo_connector->sharpness); - if (psb_intel_sdvo_connector->flicker_filter) - drm_property_destroy(dev, psb_intel_sdvo_connector->flicker_filter); - if (psb_intel_sdvo_connector->flicker_filter_2d) - drm_property_destroy(dev, psb_intel_sdvo_connector->flicker_filter_2d); - if (psb_intel_sdvo_connector->flicker_filter_adaptive) - drm_property_destroy(dev, psb_intel_sdvo_connector->flicker_filter_adaptive); - if (psb_intel_sdvo_connector->tv_luma_filter) - drm_property_destroy(dev, psb_intel_sdvo_connector->tv_luma_filter); - if (psb_intel_sdvo_connector->tv_chroma_filter) - drm_property_destroy(dev, psb_intel_sdvo_connector->tv_chroma_filter); - if (psb_intel_sdvo_connector->dot_crawl) - drm_property_destroy(dev, psb_intel_sdvo_connector->dot_crawl); - if (psb_intel_sdvo_connector->brightness) - drm_property_destroy(dev, psb_intel_sdvo_connector->brightness); -} - static void psb_intel_sdvo_destroy(struct drm_connector *connector) { - struct psb_intel_sdvo_connector *psb_intel_sdvo_connector = to_psb_intel_sdvo_connector(connector); - - if (psb_intel_sdvo_connector->tv_format) - drm_property_destroy(connector->dev, - psb_intel_sdvo_connector->tv_format); - - psb_intel_sdvo_destroy_enhance_property(connector); drm_connector_unregister(connector); drm_connector_cleanup(connector); kfree(connector); -- cgit v0.10.2 From ecd5c9821c39626fa7c03e9c397586b24cb11b79 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Sun, 2 Nov 2014 12:18:29 +0200 Subject: amdkfd: Implement the Get Version IOCTL Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 3130265..64c73ba 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -123,7 +123,16 @@ static int kfd_open(struct inode *inode, struct file *filep) static long kfd_ioctl_get_version(struct file *filep, struct kfd_process *p, void __user *arg) { - return -ENODEV; + struct kfd_ioctl_get_version_args args; + int err = 0; + + args.major_version = KFD_IOCTL_MAJOR_VERSION; + args.minor_version = KFD_IOCTL_MINOR_VERSION; + + if (copy_to_user(arg, &args, sizeof(args))) + err = -EFAULT; + + return err; } static int set_queue_properties_from_user(struct queue_properties *q_properties, -- cgit v0.10.2 From 2c4cc91b0b43ae364d47b0eec560b8e5bd5af746 Mon Sep 17 00:00:00 2001 From: Peter Hurley Date: Mon, 3 Nov 2014 20:51:45 -0500 Subject: drm: Fix DRM_FORCE_ON_DIGITAL use A connector may be forced on from the command line via video= command line setting. The digital output of dual-mode connectors can also be specifically selected and forced on; eg., 'video=DVI-I-2:D'. However, in this case, the connector->status will be mistakenly set to connector_status_disconnected, and the connector will not be mode set. Fix the connector->status when connector->force is DRM_FORCE_ON_DIGITAL. Note that this seems to have been broken ever since the introduction of the connector forcing support in commit d50ba256b5f1478e15accfcfda9b72fd7a661364 Author: Dave Airlie Date: Wed Sep 23 14:44:08 2009 +1000 drm/kms: start adding command line interface using fb. Signed-off-by: Peter Hurley [danvet: Add note about that this never worked.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_probe_helper.c b/drivers/gpu/drm/drm_probe_helper.c index 6857e9a..7483a47 100644 --- a/drivers/gpu/drm/drm_probe_helper.c +++ b/drivers/gpu/drm/drm_probe_helper.c @@ -118,7 +118,8 @@ static int drm_helper_probe_single_connector_modes_merge_bits(struct drm_connect mode->status = MODE_UNVERIFIED; if (connector->force) { - if (connector->force == DRM_FORCE_ON) + if (connector->force == DRM_FORCE_ON || + connector->force == DRM_FORCE_ON_DIGITAL) connector->status = connector_status_connected; else connector->status = connector_status_disconnected; -- cgit v0.10.2 From b74c6c92f2839206a3714a00af53b4bdc24c7224 Mon Sep 17 00:00:00 2001 From: Peter Hurley Date: Mon, 3 Nov 2014 20:53:41 -0500 Subject: drm: Remove compiler BUG_ON() test modeset->num_connectors must be 0 to reach the BUG_ON() which tests for non-zero modeset->num_connectors; remove BUG_ON(). Signed-off-by: Peter Hurley Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 0c0c39b..09d47e9 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -1570,7 +1570,6 @@ static void drm_setup_crtcs(struct drm_fb_helper *fb_helper) modeset = &fb_helper->crtc_info[i].mode_set; if (modeset->num_connectors == 0) { BUG_ON(modeset->fb); - BUG_ON(modeset->num_connectors); if (modeset->mode) drm_mode_destroy(dev, modeset->mode); modeset->mode = NULL; -- cgit v0.10.2 From aff437667b93c3d65576b02628885687c72e1b3b Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Fri, 24 Oct 2014 12:42:33 +0100 Subject: drm/i915: Move flags describing VMA mappings into the VMA If these flags are on the object level it will be more difficult to allow for multiple VMAs per object. v2: Simplification and cleanup after code review comments (Chris Wilson). Signed-off-by: Tvrtko Ursulin Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index e60d5c2..61ea8da 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -116,7 +116,7 @@ static const char *get_tiling_flag(struct drm_i915_gem_object *obj) static inline const char *get_global_flag(struct drm_i915_gem_object *obj) { - return obj->has_global_gtt_mapping ? "g" : " "; + return i915_gem_obj_to_ggtt(obj) ? "g" : " "; } static void diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 583c97d..627b7e7 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1856,8 +1856,6 @@ struct drm_i915_gem_object { unsigned long gt_ro:1; unsigned int cache_level:3; - unsigned int has_aliasing_ppgtt_mapping:1; - unsigned int has_global_gtt_mapping:1; unsigned int has_dma_mapping:1; unsigned int frontbuffer_bits:INTEL_FRONTBUFFER_BITS; diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 827edb5..7e91978 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3701,7 +3701,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj, list_for_each_entry(vma, &obj->vma_list, vma_link) if (drm_mm_node_allocated(&vma->node)) vma->bind_vma(vma, cache_level, - obj->has_global_gtt_mapping ? GLOBAL_BIND : 0); + vma->bound & GLOBAL_BIND); } list_for_each_entry(vma, &obj->vma_list, vma_link) @@ -4097,7 +4097,7 @@ i915_gem_object_pin(struct drm_i915_gem_object *obj, return PTR_ERR(vma); } - if (flags & PIN_GLOBAL && !obj->has_global_gtt_mapping) + if (flags & PIN_GLOBAL && !(vma->bound & GLOBAL_BIND)) vma->bind_vma(vma, obj->cache_level, GLOBAL_BIND); vma->pin_count++; diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index a5221d8..7d32571 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -522,6 +522,7 @@ static int do_switch(struct intel_engine_cs *ring, struct intel_context *from = ring->last_context; u32 hw_flags = 0; bool uninitialized = false; + struct i915_vma *vma; int ret, i; if (from != NULL && ring == &dev_priv->ring[RCS]) { @@ -571,11 +572,10 @@ static int do_switch(struct intel_engine_cs *ring, if (ret) goto unpin_out; - if (!to->legacy_hw_ctx.rcs_state->has_global_gtt_mapping) { - struct i915_vma *vma = i915_gem_obj_to_vma(to->legacy_hw_ctx.rcs_state, - &dev_priv->gtt.base); - vma->bind_vma(vma, to->legacy_hw_ctx.rcs_state->cache_level, GLOBAL_BIND); - } + vma = i915_gem_obj_to_ggtt(to->legacy_hw_ctx.rcs_state); + if (!(vma->bound & GLOBAL_BIND)) + vma->bind_vma(vma, to->legacy_hw_ctx.rcs_state->cache_level, + GLOBAL_BIND); if (!to->legacy_hw_ctx.initialized || i915_gem_context_is_default(to)) hw_flags |= MI_RESTORE_INHIBIT; diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 1a0611b..c70b329 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -357,12 +357,9 @@ i915_gem_execbuffer_relocate_entry(struct drm_i915_gem_object *obj, * through the ppgtt for non_secure batchbuffers. */ if (unlikely(IS_GEN6(dev) && reloc->write_domain == I915_GEM_DOMAIN_INSTRUCTION && - !target_i915_obj->has_global_gtt_mapping)) { - struct i915_vma *vma = - list_first_entry(&target_i915_obj->vma_list, - typeof(*vma), vma_link); - vma->bind_vma(vma, target_i915_obj->cache_level, GLOBAL_BIND); - } + !(target_vma->bound & GLOBAL_BIND))) + target_vma->bind_vma(target_vma, target_i915_obj->cache_level, + GLOBAL_BIND); /* Validate that the target is in a valid r/w GPU domain */ if (unlikely(reloc->write_domain & (reloc->write_domain - 1))) { diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index ae82ef5..cb7adab 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -1336,7 +1336,7 @@ void i915_gem_restore_gtt_mappings(struct drm_device *dev) * Unfortunately above, we've just wiped out the mappings * without telling our object about it. So we need to fake it. */ - obj->has_global_gtt_mapping = 0; + vma->bound &= ~GLOBAL_BIND; vma->bind_vma(vma, obj->cache_level, GLOBAL_BIND); } @@ -1533,7 +1533,7 @@ static void i915_ggtt_bind_vma(struct i915_vma *vma, BUG_ON(!i915_is_ggtt(vma->vm)); intel_gtt_insert_sg_entries(vma->obj->pages, entry, flags); - vma->obj->has_global_gtt_mapping = 1; + vma->bound = GLOBAL_BIND; } static void i915_ggtt_clear_range(struct i915_address_space *vm, @@ -1552,7 +1552,7 @@ static void i915_ggtt_unbind_vma(struct i915_vma *vma) const unsigned int size = vma->obj->base.size >> PAGE_SHIFT; BUG_ON(!i915_is_ggtt(vma->vm)); - vma->obj->has_global_gtt_mapping = 0; + vma->bound = 0; intel_gtt_clear_range(first, size); } @@ -1580,24 +1580,24 @@ static void ggtt_bind_vma(struct i915_vma *vma, * flags. At all other times, the GPU will use the aliasing PPGTT. */ if (!dev_priv->mm.aliasing_ppgtt || flags & GLOBAL_BIND) { - if (!obj->has_global_gtt_mapping || + if (!(vma->bound & GLOBAL_BIND) || (cache_level != obj->cache_level)) { vma->vm->insert_entries(vma->vm, obj->pages, vma->node.start, cache_level, flags); - obj->has_global_gtt_mapping = 1; + vma->bound |= GLOBAL_BIND; } } if (dev_priv->mm.aliasing_ppgtt && - (!obj->has_aliasing_ppgtt_mapping || + (!(vma->bound & LOCAL_BIND) || (cache_level != obj->cache_level))) { struct i915_hw_ppgtt *appgtt = dev_priv->mm.aliasing_ppgtt; appgtt->base.insert_entries(&appgtt->base, vma->obj->pages, vma->node.start, cache_level, flags); - vma->obj->has_aliasing_ppgtt_mapping = 1; + vma->bound |= LOCAL_BIND; } } @@ -1607,21 +1607,21 @@ static void ggtt_unbind_vma(struct i915_vma *vma) struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_gem_object *obj = vma->obj; - if (obj->has_global_gtt_mapping) { + if (vma->bound & GLOBAL_BIND) { vma->vm->clear_range(vma->vm, vma->node.start, obj->base.size, true); - obj->has_global_gtt_mapping = 0; + vma->bound &= ~GLOBAL_BIND; } - if (obj->has_aliasing_ppgtt_mapping) { + if (vma->bound & LOCAL_BIND) { struct i915_hw_ppgtt *appgtt = dev_priv->mm.aliasing_ppgtt; appgtt->base.clear_range(&appgtt->base, vma->node.start, obj->base.size, true); - obj->has_aliasing_ppgtt_mapping = 0; + vma->bound &= ~LOCAL_BIND; } } @@ -1699,7 +1699,7 @@ int i915_gem_setup_global_gtt(struct drm_device *dev, DRM_DEBUG_KMS("Reservation failed: %i\n", ret); return ret; } - obj->has_global_gtt_mapping = 1; + vma->bound |= GLOBAL_BIND; } dev_priv->gtt.base.start = start; diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h index d5c14af..d0562d0 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.h +++ b/drivers/gpu/drm/i915/i915_gem_gtt.h @@ -123,6 +123,12 @@ struct i915_vma { struct drm_i915_gem_object *obj; struct i915_address_space *vm; + /** Flags and address space this VMA is bound to */ +#define GLOBAL_BIND (1<<0) +#define LOCAL_BIND (1<<1) +#define PTE_READ_ONLY (1<<2) + unsigned int bound : 4; + /** This object's place on the active/inactive lists */ struct list_head mm_list; @@ -155,8 +161,6 @@ struct i915_vma { * setting the valid PTE entries to a reserved scratch page. */ void (*unbind_vma)(struct i915_vma *vma); /* Map an object into an address space with the given cache flags. */ -#define GLOBAL_BIND (1<<0) -#define PTE_READ_ONLY (1<<1) void (*bind_vma)(struct i915_vma *vma, enum i915_cache_level cache_level, u32 flags); diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c index 85fda6b..c388918 100644 --- a/drivers/gpu/drm/i915/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/i915_gem_stolen.c @@ -533,7 +533,7 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_device *dev, } } - obj->has_global_gtt_mapping = 1; + vma->bound |= GLOBAL_BIND; list_add_tail(&obj->global_list, &dev_priv->mm.bound_list); list_add_tail(&vma->mm_list, &ggtt->inactive_list); diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index e664599..d17360b 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -565,6 +565,7 @@ i915_error_object_create(struct drm_i915_private *dev_priv, struct i915_address_space *vm) { struct drm_i915_error_object *dst; + struct i915_vma *vma = NULL; int num_pages; bool use_ggtt; int i = 0; @@ -585,16 +586,17 @@ i915_error_object_create(struct drm_i915_private *dev_priv, dst->gtt_offset = -1; reloc_offset = dst->gtt_offset; + if (i915_is_ggtt(vm)) + vma = i915_gem_obj_to_ggtt(src); use_ggtt = (src->cache_level == I915_CACHE_NONE && - i915_is_ggtt(vm) && - src->has_global_gtt_mapping && - reloc_offset + num_pages * PAGE_SIZE <= dev_priv->gtt.mappable_end); + vma && (vma->bound & GLOBAL_BIND) && + reloc_offset + num_pages * PAGE_SIZE <= dev_priv->gtt.mappable_end); /* Cannot access stolen address directly, try to use the aperture */ if (src->stolen) { use_ggtt = true; - if (!src->has_global_gtt_mapping) + if (!(vma && vma->bound & GLOBAL_BIND)) goto unwind; reloc_offset = i915_gem_obj_ggtt_offset(src); -- cgit v0.10.2 From a919db9015832d7e4140a38e20ddc1c305c95cb2 Mon Sep 17 00:00:00 2001 From: Sonika Jindal Date: Thu, 23 Oct 2014 07:41:33 -0700 Subject: drm/i915: Update plane parameters for cursor plane (v2) This allows the cursor plane to be updated the same way as primary and sprites, and same set_property handler is used for all of these planes. v2 (by Matt Roper): Rework to apply to latest di-nightly codebase. The switch to split check/commit plane programming changed the code flow enough that the original patch could no longer be applied. Signed-off-by: Sonika Jindal Reviewed-by: Matt Roper Tested-by (IVB): Matt Roper Signed-off-by: Matt Roper Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 1d2fa7f..910e8a4 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11955,12 +11955,24 @@ intel_commit_cursor_plane(struct drm_plane *plane, struct drm_crtc *crtc = state->crtc; struct drm_framebuffer *fb = state->fb; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_plane *intel_plane = to_intel_plane(plane); struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); struct drm_i915_gem_object *obj = intel_fb->obj; int crtc_w, crtc_h; crtc->cursor_x = state->orig_dst.x1; crtc->cursor_y = state->orig_dst.y1; + + intel_plane->crtc_x = state->orig_dst.x1; + intel_plane->crtc_y = state->orig_dst.y1; + intel_plane->crtc_w = drm_rect_width(&state->orig_dst); + intel_plane->crtc_h = drm_rect_height(&state->orig_dst); + intel_plane->src_x = state->orig_src.x1; + intel_plane->src_y = state->orig_src.y1; + intel_plane->src_w = drm_rect_width(&state->orig_src); + intel_plane->src_h = drm_rect_height(&state->orig_src); + intel_plane->obj = obj; + if (fb != crtc->cursor->fb) { crtc_w = drm_rect_width(&state->orig_dst); crtc_h = drm_rect_height(&state->orig_dst); -- cgit v0.10.2 From 42c7156af94783ff42edba6a3b07a51d35552c60 Mon Sep 17 00:00:00 2001 From: Brad Volkin Date: Thu, 16 Oct 2014 12:24:42 -0700 Subject: drm/i915: Abort command parsing for chained batches libva uses chained batch buffers in a way that the command parser can't generally handle. Fortunately, libva doesn't need to write registers from batch buffers in the way that mesa does, so this patch causes the driver to fall back to non-secure dispatch if the parser detects a chained batch buffer. Note: The 2nd hunk to munge the error code of the parser looks a bit superflous. At least until we have the batch copy code ready and can run the cmd parser in granting mode. But it isn't since we still need to let existing libva buffers pass (though not with elevated privs ofc!). Testcase: igt/gem_exec_parse/chained-batch Signed-off-by: Brad Volkin [danvet: Add note - this confused me in review and Brad clarified things (after a few mails ...).] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c index 86b3ae0..ef38915 100644 --- a/drivers/gpu/drm/i915/i915_cmd_parser.c +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c @@ -138,6 +138,11 @@ static const struct drm_i915_cmd_descriptor common_cmds[] = { .mask = MI_GLOBAL_GTT, .expected = 0, }}, ), + /* + * MI_BATCH_BUFFER_START requires some special handling. It's not + * really a 'skip' action but it doesn't seem like it's worth adding + * a new action. See i915_parse_cmds(). + */ CMD( MI_BATCH_BUFFER_START, SMI, !F, 0xFF, S ), }; @@ -955,7 +960,8 @@ static bool check_cmd(const struct intel_engine_cs *ring, * Parses the specified batch buffer looking for privilege violations as * described in the overview. * - * Return: non-zero if the parser finds violations or otherwise fails + * Return: non-zero if the parser finds violations or otherwise fails; -EACCES + * if the batch appears legal but should use hardware parsing */ int i915_parse_cmds(struct intel_engine_cs *ring, struct drm_i915_gem_object *batch_obj, @@ -1002,6 +1008,16 @@ int i915_parse_cmds(struct intel_engine_cs *ring, break; } + /* + * If the batch buffer contains a chained batch, return an + * error that tells the caller to abort and dispatch the + * workload as a non-secure batch. + */ + if (desc->cmd.value == MI_BATCH_BUFFER_START) { + ret = -EACCES; + break; + } + if (desc->flags & CMD_DESC_FIXED) length = desc->length.fixed; else diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index c70b329..4b7f5c1 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -1365,17 +1365,19 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data, batch_obj, args->batch_start_offset, file->is_master); - if (ret) - goto err; - - /* - * XXX: Actually do this when enabling batch copy... - * - * Set the DISPATCH_SECURE bit to remove the NON_SECURE bit - * from MI_BATCH_BUFFER_START commands issued in the - * dispatch_execbuffer implementations. We specifically don't - * want that set when the command parser is enabled. - */ + if (ret) { + if (ret != -EACCES) + goto err; + } else { + /* + * XXX: Actually do this when enabling batch copy... + * + * Set the DISPATCH_SECURE bit to remove the NON_SECURE bit + * from MI_BATCH_BUFFER_START commands issued in the + * dispatch_execbuffer implementations. We specifically don't + * want that set when the command parser is enabled. + */ + } } /* snb/ivb/vlv conflate the "batch in ppgtt" bit with the "non-secure -- cgit v0.10.2 From 655b5391ca5ef4095110c0b3f82f10e337f0a736 Mon Sep 17 00:00:00 2001 From: Armin Reese Date: Thu, 23 Oct 2014 08:34:27 -0700 Subject: drm/i915 Update Gen8 golden context batch buffer The file drivers/gpu/drm/i915/intel_renderstate_gen8.c is updated to the version created by IGT null_state_gen Signed-off-by: Armin Reese Reviewed-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_renderstate_gen8.c b/drivers/gpu/drm/i915/intel_renderstate_gen8.c index 75ef1b5d..78011d7 100644 --- a/drivers/gpu/drm/i915/intel_renderstate_gen8.c +++ b/drivers/gpu/drm/i915/intel_renderstate_gen8.c @@ -1,70 +1,575 @@ #include "intel_renderstate.h" static const u32 gen8_null_state_relocs[] = { - 0x00000048, - 0x00000050, - 0x00000060, - 0x000003ec, + 0x00000798, + 0x000007a4, + 0x000007ac, + 0x000007bc, -1, }; static const u32 gen8_null_state_batch[] = { + 0x7a000004, + 0x01000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, 0x69040000, - 0x61020001, + 0x78140000, + 0x04000000, + 0x7820000a, + 0x00000000, + 0x00000000, + 0x80000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78130002, + 0x00000000, + 0x00000000, + 0x02001808, + 0x781f0002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78510009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78100007, + 0x00000000, + 0x00000000, + 0x00010000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781b0007, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000800, + 0x00000000, + 0x78110008, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781e0003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781d0007, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78120002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78500003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781c0002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x780c0000, + 0x00000000, + 0x78520003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78300000, + 0x08010040, + 0x78310000, + 0x1e000000, + 0x78320000, + 0x1e000000, + 0x78330000, + 0x1e000000, + 0x79190002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x791a0002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x791b0002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79120000, + 0x00000000, + 0x79130000, + 0x00000000, + 0x79140000, + 0x00000000, + 0x79150000, + 0x00000000, + 0x79160000, + 0x00000000, + 0x78150009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78190009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781a0009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78160009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78170009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78490001, + 0x00000000, + 0x00000000, + 0x784a0000, + 0x00000000, + 0x784b0000, + 0x00000004, + 0x79170101, + 0x00000000, + 0x00000080, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, 0x00000000, 0x00000000, - 0x79120000, 0x00000000, - 0x79130000, 0x00000000, - 0x79140000, 0x00000000, - 0x79150000, 0x00000000, - 0x79160000, 0x00000000, - 0x6101000e, - 0x00000001, 0x00000000, - 0x00000001, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79180006, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79180006, + 0x20000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79180006, + 0x40000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79180006, + 0x60000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x6101000e, 0x00000001, /* reloc */ 0x00000000, + 0x00000000, 0x00000001, /* reloc */ 0x00000000, + 0x00000001, /* reloc */ 0x00000000, + 0x00000001, 0x00000000, 0x00000001, /* reloc */ 0x00000000, - 0xfffff001, 0x00001001, - 0xfffff001, 0x00001001, - 0x78230000, - 0x000006e0, - 0x78210000, - 0x00000700, - 0x78300000, - 0x08010040, - 0x78330000, - 0x08000000, - 0x78310000, - 0x08000000, - 0x78320000, - 0x08000000, - 0x78240000, - 0x00000641, - 0x780e0000, - 0x00000601, + 0x00000001, + 0x00001001, + 0x61020001, + 0x00000000, + 0x00000000, + 0x79000002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78050006, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79040002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79040002, + 0x40000000, + 0x00000000, + 0x00000000, + 0x79040002, + 0x80000000, + 0x00000000, + 0x00000000, + 0x79040002, + 0xc0000000, + 0x00000000, + 0x00000000, + 0x79080001, + 0x00000000, + 0x00000000, + 0x790a0001, + 0x00000000, + 0x00000000, + 0x78060003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78070003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78040001, + 0x00000000, + 0x00000000, + 0x79110000, + 0x00000000, 0x780d0000, 0x00000000, - 0x78180000, - 0x00000001, - 0x78520003, + 0x79060000, 0x00000000, + 0x7907001f, 0x00000000, 0x00000000, 0x00000000, - 0x78190009, 0x00000000, 0x00000000, 0x00000000, @@ -75,7 +580,6 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, - 0x781b0007, 0x00000000, 0x00000000, 0x00000000, @@ -84,26 +588,22 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, - 0x78270000, 0x00000000, - 0x782c0000, 0x00000000, - 0x781c0002, 0x00000000, 0x00000000, 0x00000000, - 0x78160009, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x7902000f, 0x00000000, 0x00000000, 0x00000000, 0x00000000, - 0x78110008, 0x00000000, 0x00000000, 0x00000000, @@ -113,12 +613,10 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, - 0x78290000, 0x00000000, - 0x782e0000, 0x00000000, - 0x781a0009, 0x00000000, + 0x790c000f, 0x00000000, 0x00000000, 0x00000000, @@ -128,7 +626,6 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, - 0x781d0007, 0x00000000, 0x00000000, 0x00000000, @@ -136,153 +633,153 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, + 0x780a0003, 0x00000000, - 0x78280000, 0x00000000, - 0x782d0000, 0x00000000, - 0x78260000, 0x00000000, - 0x782b0000, + 0x78080083, + 0x00004000, 0x00000000, - 0x78150009, 0x00000000, 0x00000000, + 0x04004000, 0x00000000, 0x00000000, 0x00000000, + 0x08004000, 0x00000000, 0x00000000, 0x00000000, + 0x0c004000, 0x00000000, 0x00000000, - 0x78100007, 0x00000000, + 0x10004000, 0x00000000, 0x00000000, 0x00000000, + 0x14004000, 0x00000000, 0x00000000, 0x00000000, + 0x18004000, 0x00000000, - 0x781e0003, 0x00000000, 0x00000000, + 0x1c004000, 0x00000000, 0x00000000, - 0x78120002, 0x00000000, + 0x20004000, 0x00000000, 0x00000000, - 0x781f0002, - 0x30400820, 0x00000000, + 0x24004000, 0x00000000, - 0x78510009, 0x00000000, 0x00000000, + 0x28004000, 0x00000000, 0x00000000, 0x00000000, + 0x2c004000, 0x00000000, 0x00000000, 0x00000000, + 0x30004000, 0x00000000, 0x00000000, - 0x78500003, - 0x00210000, 0x00000000, + 0x34004000, 0x00000000, 0x00000000, - 0x78130002, 0x00000000, + 0x38004000, 0x00000000, 0x00000000, - 0x782a0000, - 0x00000480, - 0x782f0000, - 0x00000540, - 0x78140000, - 0x00000800, - 0x78170009, 0x00000000, + 0x3c004000, 0x00000000, 0x00000000, 0x00000000, + 0x40004000, 0x00000000, 0x00000000, 0x00000000, + 0x44004000, 0x00000000, 0x00000000, 0x00000000, - 0x7820000a, - 0x00000580, + 0x48004000, 0x00000000, - 0x08080000, 0x00000000, 0x00000000, - 0x1f000002, - 0x00060000, + 0x4c004000, 0x00000000, 0x00000000, 0x00000000, + 0x50004000, 0x00000000, - 0x784d0000, - 0x40000000, - 0x784f0000, - 0x80000100, - 0x780f0000, - 0x00000740, - 0x78050006, 0x00000000, 0x00000000, + 0x54004000, 0x00000000, 0x00000000, 0x00000000, + 0x58004000, 0x00000000, 0x00000000, - 0x78070003, 0x00000000, + 0x5c004000, 0x00000000, 0x00000000, 0x00000000, - 0x78060003, + 0x60004000, 0x00000000, 0x00000000, 0x00000000, + 0x64004000, 0x00000000, - 0x78040001, 0x00000000, - 0x00000001, - 0x79000002, - 0xffffffff, + 0x00000000, + 0x68004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x6c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x70004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x74004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78004000, + 0x00000000, 0x00000000, 0x00000000, - 0x78080003, - 0x00006000, - 0x000005e0, /* reloc */ + 0x7c004000, 0x00000000, 0x00000000, - 0x78090005, + 0x00000000, + 0x80004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78090043, 0x02000000, 0x22220000, - 0x02f60000, - 0x11230000, - 0x02850004, - 0x11230000, - 0x784b0000, - 0x0000000f, - 0x78490001, 0x00000000, 0x00000000, - 0x7b000005, 0x00000000, - 0x00000003, 0x00000000, - 0x00000001, 0x00000000, 0x00000000, - 0x05000000, /* cmds end */ 0x00000000, 0x00000000, 0x00000000, @@ -297,8 +794,6 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, - 0x000004c0, /* state start */ - 0x00000500, 0x00000000, 0x00000000, 0x00000000, @@ -345,46 +840,65 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, + 0x680b0001, + 0x78260000, + 0x00000000, + 0x78270000, + 0x00000000, + 0x78280000, + 0x00000000, + 0x78290000, + 0x00000000, + 0x782a0000, + 0x00000000, + 0x780e0000, + 0x00000dc1, + 0x78240000, + 0x00000e01, + 0x784f0000, + 0x80000100, + 0x784d0000, + 0x40000000, + 0x782b0000, + 0x00000000, + 0x782c0000, + 0x00000000, + 0x782d0000, + 0x00000000, + 0x782e0000, + 0x00000000, + 0x782f0000, + 0x00000000, + 0x780f0000, 0x00000000, + 0x78230000, + 0x00000e60, + 0x78210000, + 0x00000e80, + 0x7b000005, + 0x00000004, + 0x00000001, 0x00000000, + 0x00000001, 0x00000000, - 0x00000092, 0x00000000, + 0x05000000, /* cmds end */ 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, /* state start */ 0x00000000, + 0x3f800000, + 0x3f800000, + 0x3f800000, + 0x3f800000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, - 0x0060005a, - 0x21403ae8, - 0x3a0000c0, - 0x008d0040, - 0x0060005a, - 0x21603ae8, - 0x3a0000c0, - 0x008d0080, - 0x0060005a, - 0x21803ae8, - 0x3a0000d0, - 0x008d0040, - 0x0060005a, - 0x21a03ae8, - 0x3a0000d0, - 0x008d0080, - 0x02800031, - 0x2e0022e8, - 0x0e000140, - 0x08840001, - 0x05800031, - 0x200022e0, - 0x0e000e00, - 0x90031000, 0x00000000, 0x00000000, 0x00000000, @@ -410,38 +924,6 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, - 0x06200000, - 0x00000002, 0x00000000, 0x00000000, 0x00000000, @@ -449,8 +931,6 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, - 0xf99a130c, - 0x799a130c, 0x00000000, 0x00000000, 0x00000000, @@ -466,9 +946,7 @@ static const u32 gen8_null_state_batch[] = { 0x00000000, 0x00000000, 0x00000000, - 0x3f800000, 0x00000000, - 0x3f800000, 0x00000000, 0x00000000, 0x00000000, -- cgit v0.10.2 From ff7a60f28fa2e018ecd91dc00cbb91d07ff1eb91 Mon Sep 17 00:00:00 2001 From: Armin Reese Date: Thu, 23 Oct 2014 08:34:28 -0700 Subject: drm/i915 Add golden context support for Gen9 This patch includes the Gen9 batch buffer to generate a 'golden context' for that product family. Signed-off-by: Armin Reese Reviewed-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 75fd7de..9c646c6 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -40,7 +40,8 @@ i915-y += i915_cmd_parser.o \ # autogenerated null render state i915-y += intel_renderstate_gen6.o \ intel_renderstate_gen7.o \ - intel_renderstate_gen8.o + intel_renderstate_gen8.o \ + intel_renderstate_gen9.o # modesetting core code i915-y += intel_bios.o \ diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c index a9a62d7..98dcd94 100644 --- a/drivers/gpu/drm/i915/i915_gem_render_state.c +++ b/drivers/gpu/drm/i915/i915_gem_render_state.c @@ -38,6 +38,8 @@ render_state_get_rodata(struct drm_device *dev, const int gen) return &gen7_null_state; case 8: return &gen8_null_state; + case 9: + return &gen9_null_state; } return NULL; diff --git a/drivers/gpu/drm/i915/intel_renderstate.h b/drivers/gpu/drm/i915/intel_renderstate.h index 6c792d3..5bd6985 100644 --- a/drivers/gpu/drm/i915/intel_renderstate.h +++ b/drivers/gpu/drm/i915/intel_renderstate.h @@ -29,6 +29,7 @@ extern const struct intel_renderstate_rodata gen6_null_state; extern const struct intel_renderstate_rodata gen7_null_state; extern const struct intel_renderstate_rodata gen8_null_state; +extern const struct intel_renderstate_rodata gen9_null_state; #define RO_RENDERSTATE(_g) \ const struct intel_renderstate_rodata gen ## _g ## _null_state = { \ diff --git a/drivers/gpu/drm/i915/intel_renderstate_gen9.c b/drivers/gpu/drm/i915/intel_renderstate_gen9.c new file mode 100644 index 0000000..8750753 --- /dev/null +++ b/drivers/gpu/drm/i915/intel_renderstate_gen9.c @@ -0,0 +1,974 @@ +#include "intel_renderstate.h" + +static const u32 gen9_null_state_relocs[] = { + 0x000007a8, + 0x000007b4, + 0x000007bc, + 0x000007cc, + -1, +}; + +static const u32 gen9_null_state_batch[] = { + 0x7a000004, + 0x01000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x69040300, + 0x78140000, + 0x04000000, + 0x7820000a, + 0x00000000, + 0x00000000, + 0x80000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78130002, + 0x00000000, + 0x00000000, + 0x02001808, + 0x781f0004, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78510009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78100007, + 0x00000000, + 0x00000000, + 0x00010000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781b0007, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000800, + 0x00000000, + 0x78110008, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781e0003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781d0009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78120002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78500003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781c0002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x780c0000, + 0x00000000, + 0x78520003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78300000, + 0x08010040, + 0x78310000, + 0x1e000000, + 0x78320000, + 0x1e000000, + 0x78330000, + 0x1e000000, + 0x79190002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x791a0002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x791b0002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79120000, + 0x00000000, + 0x79130000, + 0x00000000, + 0x79140000, + 0x00000000, + 0x79150000, + 0x00000000, + 0x79160000, + 0x00000000, + 0x78150009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78190009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x781a0009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78160009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78170009, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78490001, + 0x00000000, + 0x00000000, + 0x784a0000, + 0x00000000, + 0x784b0000, + 0x00000004, + 0x79170101, + 0x00000000, + 0x00000080, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79180006, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79180006, + 0x20000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79180006, + 0x40000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79180006, + 0x60000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x61010011, + 0x00000001, /* reloc */ + 0x00000000, + 0x00000000, + 0x00000001, /* reloc */ + 0x00000000, + 0x00000001, /* reloc */ + 0x00000000, + 0x00000001, + 0x00000000, + 0x00000001, /* reloc */ + 0x00000000, + 0x00001001, + 0x00001001, + 0x00000001, + 0x00001001, + 0x00000000, + 0x00000000, + 0x00000000, + 0x61020001, + 0x00000000, + 0x00000000, + 0x79000002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78050006, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79040002, + 0x00000000, + 0x00000000, + 0x00000000, + 0x79040002, + 0x40000000, + 0x00000000, + 0x00000000, + 0x79040002, + 0x80000000, + 0x00000000, + 0x00000000, + 0x79040002, + 0xc0000000, + 0x00000000, + 0x00000000, + 0x79080001, + 0x00000000, + 0x00000000, + 0x790a0001, + 0x00000000, + 0x00000000, + 0x78060003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78070003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78040001, + 0x00000000, + 0x00000000, + 0x79110000, + 0x00000000, + 0x780d0000, + 0x00000000, + 0x79060000, + 0x00000000, + 0x7907001f, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x7902000f, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x790c000f, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x780a0003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78080083, + 0x00004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x04004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x08004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x0c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x10004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x14004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x18004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x1c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x20004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x24004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x28004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x2c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x30004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x34004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x38004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x3c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x40004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x44004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x48004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x4c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x50004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x54004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x58004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x5c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x60004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x64004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x68004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x6c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x70004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x74004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x7c004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x80004000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78090043, + 0x02000000, + 0x22220000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x78550003, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x680b0001, + 0x780e0000, + 0x00000e01, + 0x78240000, + 0x00000e41, + 0x784f0000, + 0x80000100, + 0x784d0000, + 0x40000000, + 0x782b0000, + 0x00000000, + 0x782c0000, + 0x00000000, + 0x782d0000, + 0x00000000, + 0x782e0000, + 0x00000000, + 0x782f0000, + 0x00000000, + 0x780f0000, + 0x00000000, + 0x78230000, + 0x00000ea0, + 0x78210000, + 0x00000ec0, + 0x78260000, + 0x00000000, + 0x78270000, + 0x00000000, + 0x78280000, + 0x00000000, + 0x78290000, + 0x00000000, + 0x782a0000, + 0x00000000, + 0x7b000005, + 0x00000004, + 0x00000001, + 0x00000000, + 0x00000001, + 0x00000000, + 0x00000000, + 0x05000000, /* cmds end */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, /* state start */ + 0x00000000, + 0x3f800000, + 0x3f800000, + 0x3f800000, + 0x3f800000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, /* state end */ +}; + +RO_RENDERSTATE(9); -- cgit v0.10.2 From 3f20df98870236b288a90fc3b65f593d3db05b17 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 24 Oct 2014 14:51:34 +0100 Subject: drm/i915: only flip frontbuffer if crtc is active MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There is no point in flipping a buffer for a disabled crtc. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 910e8a4..140bbef 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8544,9 +8544,9 @@ static int intel_crtc_cursor_set_obj(struct drm_crtc *crtc, if (old_width != width) intel_update_watermarks(crtc); intel_crtc_update_cursor(crtc, intel_crtc->cursor_bo != NULL); - } - intel_frontbuffer_flip(dev, INTEL_FRONTBUFFER_CURSOR(pipe)); + intel_frontbuffer_flip(dev, INTEL_FRONTBUFFER_CURSOR(pipe)); + } return 0; fail_unpin: -- cgit v0.10.2 From c5d974728fafc151e740638b4cf492958be78b5b Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Sat, 25 Oct 2014 00:11:11 +0100 Subject: drm/i915: Remove unnecessary test on the gen in intel_do_mmio_flip() use_mmio_flip() makes sure we only enable MMIO flips on gen5+. So we don't need to take into account older devices. Signed-off-by: Damien Lespiau Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 140bbef..46224c6 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9567,12 +9567,11 @@ static void intel_do_mmio_flip(struct intel_crtc *intel_crtc) reg = DSPCNTR(intel_crtc->plane); dspcntr = I915_READ(reg); - if (INTEL_INFO(dev)->gen >= 4) { - if (obj->tiling_mode != I915_TILING_NONE) - dspcntr |= DISPPLANE_TILED; - else - dspcntr &= ~DISPPLANE_TILED; - } + if (obj->tiling_mode != I915_TILING_NONE) + dspcntr |= DISPPLANE_TILED; + else + dspcntr &= ~DISPPLANE_TILED; + I915_WRITE(reg, dspcntr); I915_WRITE(DSPSURF(intel_crtc->plane), -- cgit v0.10.2 From 083fe3b035d286d79071fd77d2a2c5abfdef495b Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 24 Oct 2014 19:00:17 +0100 Subject: drm: make sure visible is set to false if fb is null MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We can't let visible set true while the fb is null, some places of the code only check for visible to base its decisions. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Acked-by: Dave Airlie Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_plane_helper.c b/drivers/gpu/drm/drm_plane_helper.c index 827ec1a..fe4d1fb 100644 --- a/drivers/gpu/drm/drm_plane_helper.c +++ b/drivers/gpu/drm/drm_plane_helper.c @@ -127,6 +127,11 @@ int drm_plane_helper_check_update(struct drm_plane *plane, return -ERANGE; } + if (!fb) { + *visible = false; + return 0; + } + *visible = drm_rect_clip_scaled(src, dest, clip, hscale, vscale); if (!*visible) /* -- cgit v0.10.2 From 3ead8bb26e1e3a8b35f41a0eb65a1bd571b0204e Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 24 Oct 2014 19:00:18 +0100 Subject: drm/i915: remove unneeded visible check MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The fb check introduced to drm_plane_helper_check_update() just make this check impossible to branch in. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 46224c6..6e9d4e16 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11639,23 +11639,12 @@ intel_check_primary_plane(struct drm_plane *plane, struct drm_rect *dest = &state->dst; struct drm_rect *src = &state->src; const struct drm_rect *clip = &state->clip; - int ret; - - ret = drm_plane_helper_check_update(plane, crtc, fb, - src, dest, clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - false, true, &state->visible); - if (ret) - return ret; - /* no fb bound */ - if (state->visible && !fb) { - DRM_ERROR("No FB bound\n"); - return -EINVAL; - } - - return 0; + return drm_plane_helper_check_update(plane, crtc, fb, + src, dest, clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + false, true, &state->visible); } static int -- cgit v0.10.2 From 7c10a2b5876e014b3986d7e20b2a2894a757b138 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:43 +0200 Subject: drm/i915: add new intel audio file to group DP/HDMI audio In preparation for some additional cleanup. No functional changes. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 9c646c6..891e584 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -44,7 +44,8 @@ i915-y += intel_renderstate_gen6.o \ intel_renderstate_gen9.o # modesetting core code -i915-y += intel_bios.o \ +i915-y += intel_audio.o \ + intel_bios.o \ intel_display.o \ intel_fifo_underrun.o \ intel_frontbuffer.o \ diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c new file mode 100644 index 0000000..167903b --- /dev/null +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -0,0 +1,357 @@ +/* + * Copyright © 2014 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include + +#include +#include +#include "intel_drv.h" +#include "i915_drv.h" + +static struct { + int clock; + u32 config; +} hdmi_audio_clock[] = { + { DIV_ROUND_UP(25200 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_25175 }, + { 25200, AUD_CONFIG_PIXEL_CLOCK_HDMI_25200 }, /* default per bspec */ + { 27000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27000 }, + { 27000 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27027 }, + { 54000, AUD_CONFIG_PIXEL_CLOCK_HDMI_54000 }, + { 54000 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_54054 }, + { DIV_ROUND_UP(74250 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_74176 }, + { 74250, AUD_CONFIG_PIXEL_CLOCK_HDMI_74250 }, + { DIV_ROUND_UP(148500 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_148352 }, + { 148500, AUD_CONFIG_PIXEL_CLOCK_HDMI_148500 }, +}; + +/* get AUD_CONFIG_PIXEL_CLOCK_HDMI_* value for mode */ +static u32 audio_config_hdmi_pixel_clock(struct drm_display_mode *mode) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(hdmi_audio_clock); i++) { + if (mode->clock == hdmi_audio_clock[i].clock) + break; + } + + if (i == ARRAY_SIZE(hdmi_audio_clock)) { + DRM_DEBUG_KMS("HDMI audio pixel clock setting for %d not found, falling back to defaults\n", mode->clock); + i = 1; + } + + DRM_DEBUG_KMS("Configuring HDMI audio for pixel clock %d (0x%08x)\n", + hdmi_audio_clock[i].clock, + hdmi_audio_clock[i].config); + + return hdmi_audio_clock[i].config; +} + +static bool intel_eld_uptodate(struct drm_connector *connector, + int reg_eldv, uint32_t bits_eldv, + int reg_elda, uint32_t bits_elda, + int reg_edid) +{ + struct drm_i915_private *dev_priv = connector->dev->dev_private; + uint8_t *eld = connector->eld; + uint32_t i; + + i = I915_READ(reg_eldv); + i &= bits_eldv; + + if (!eld[0]) + return !i; + + if (!i) + return false; + + i = I915_READ(reg_elda); + i &= ~bits_elda; + I915_WRITE(reg_elda, i); + + for (i = 0; i < eld[2]; i++) + if (I915_READ(reg_edid) != *((uint32_t *)eld + i)) + return false; + + return true; +} + +static void g4x_write_eld(struct drm_connector *connector, + struct drm_crtc *crtc, + struct drm_display_mode *mode) +{ + struct drm_i915_private *dev_priv = connector->dev->dev_private; + uint8_t *eld = connector->eld; + uint32_t eldv; + uint32_t len; + uint32_t i; + + i = I915_READ(G4X_AUD_VID_DID); + + if (i == INTEL_AUDIO_DEVBLC || i == INTEL_AUDIO_DEVCL) + eldv = G4X_ELDV_DEVCL_DEVBLC; + else + eldv = G4X_ELDV_DEVCTG; + + if (intel_eld_uptodate(connector, + G4X_AUD_CNTL_ST, eldv, + G4X_AUD_CNTL_ST, G4X_ELD_ADDR, + G4X_HDMIW_HDMIEDID)) + return; + + i = I915_READ(G4X_AUD_CNTL_ST); + i &= ~(eldv | G4X_ELD_ADDR); + len = (i >> 9) & 0x1f; /* ELD buffer size */ + I915_WRITE(G4X_AUD_CNTL_ST, i); + + if (!eld[0]) + return; + + len = min_t(uint8_t, eld[2], len); + DRM_DEBUG_DRIVER("ELD size %d\n", len); + for (i = 0; i < len; i++) + I915_WRITE(G4X_HDMIW_HDMIEDID, *((uint32_t *)eld + i)); + + i = I915_READ(G4X_AUD_CNTL_ST); + i |= eldv; + I915_WRITE(G4X_AUD_CNTL_ST, i); +} + +static void haswell_write_eld(struct drm_connector *connector, + struct drm_crtc *crtc, + struct drm_display_mode *mode) +{ + struct drm_i915_private *dev_priv = connector->dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + uint8_t *eld = connector->eld; + uint32_t eldv; + uint32_t i; + int len; + int pipe = to_intel_crtc(crtc)->pipe; + int tmp; + + int hdmiw_hdmiedid = HSW_AUD_EDID_DATA(pipe); + int aud_cntl_st = HSW_AUD_DIP_ELD_CTRL(pipe); + int aud_config = HSW_AUD_CFG(pipe); + int aud_cntrl_st2 = HSW_AUD_PIN_ELD_CP_VLD; + + /* Audio output enable */ + DRM_DEBUG_DRIVER("HDMI audio: enable codec\n"); + tmp = I915_READ(aud_cntrl_st2); + tmp |= (AUDIO_OUTPUT_ENABLE_A << (pipe * 4)); + I915_WRITE(aud_cntrl_st2, tmp); + POSTING_READ(aud_cntrl_st2); + + assert_pipe_disabled(dev_priv, to_intel_crtc(crtc)->pipe); + + /* Set ELD valid state */ + tmp = I915_READ(aud_cntrl_st2); + DRM_DEBUG_DRIVER("HDMI audio: pin eld vld status=0x%08x\n", tmp); + tmp |= (AUDIO_ELD_VALID_A << (pipe * 4)); + I915_WRITE(aud_cntrl_st2, tmp); + tmp = I915_READ(aud_cntrl_st2); + DRM_DEBUG_DRIVER("HDMI audio: eld vld status=0x%08x\n", tmp); + + /* Enable HDMI mode */ + tmp = I915_READ(aud_config); + DRM_DEBUG_DRIVER("HDMI audio: audio conf: 0x%08x\n", tmp); + /* clear N_programing_enable and N_value_index */ + tmp &= ~(AUD_CONFIG_N_VALUE_INDEX | AUD_CONFIG_N_PROG_ENABLE); + I915_WRITE(aud_config, tmp); + + DRM_DEBUG_DRIVER("ELD on pipe %c\n", pipe_name(pipe)); + + eldv = AUDIO_ELD_VALID_A << (pipe * 4); + + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { + DRM_DEBUG_DRIVER("ELD: DisplayPort detected\n"); + eld[5] |= (1 << 2); /* Conn_Type, 0x1 = DisplayPort */ + I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ + } else { + I915_WRITE(aud_config, audio_config_hdmi_pixel_clock(mode)); + } + + if (intel_eld_uptodate(connector, + aud_cntrl_st2, eldv, + aud_cntl_st, IBX_ELD_ADDRESS, + hdmiw_hdmiedid)) + return; + + i = I915_READ(aud_cntrl_st2); + i &= ~eldv; + I915_WRITE(aud_cntrl_st2, i); + + if (!eld[0]) + return; + + i = I915_READ(aud_cntl_st); + i &= ~IBX_ELD_ADDRESS; + I915_WRITE(aud_cntl_st, i); + i = (i >> 29) & DIP_PORT_SEL_MASK; /* DIP_Port_Select, 0x1 = PortB */ + DRM_DEBUG_DRIVER("port num:%d\n", i); + + len = min_t(uint8_t, eld[2], 21); /* 84 bytes of hw ELD buffer */ + DRM_DEBUG_DRIVER("ELD size %d\n", len); + for (i = 0; i < len; i++) + I915_WRITE(hdmiw_hdmiedid, *((uint32_t *)eld + i)); + + i = I915_READ(aud_cntrl_st2); + i |= eldv; + I915_WRITE(aud_cntrl_st2, i); + +} + +static void ironlake_write_eld(struct drm_connector *connector, + struct drm_crtc *crtc, + struct drm_display_mode *mode) +{ + struct drm_i915_private *dev_priv = connector->dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + uint8_t *eld = connector->eld; + uint32_t eldv; + uint32_t i; + int len; + int hdmiw_hdmiedid; + int aud_config; + int aud_cntl_st; + int aud_cntrl_st2; + int pipe = to_intel_crtc(crtc)->pipe; + + if (HAS_PCH_IBX(connector->dev)) { + hdmiw_hdmiedid = IBX_HDMIW_HDMIEDID(pipe); + aud_config = IBX_AUD_CFG(pipe); + aud_cntl_st = IBX_AUD_CNTL_ST(pipe); + aud_cntrl_st2 = IBX_AUD_CNTL_ST2; + } else if (IS_VALLEYVIEW(connector->dev)) { + hdmiw_hdmiedid = VLV_HDMIW_HDMIEDID(pipe); + aud_config = VLV_AUD_CFG(pipe); + aud_cntl_st = VLV_AUD_CNTL_ST(pipe); + aud_cntrl_st2 = VLV_AUD_CNTL_ST2; + } else { + hdmiw_hdmiedid = CPT_HDMIW_HDMIEDID(pipe); + aud_config = CPT_AUD_CFG(pipe); + aud_cntl_st = CPT_AUD_CNTL_ST(pipe); + aud_cntrl_st2 = CPT_AUD_CNTRL_ST2; + } + + DRM_DEBUG_DRIVER("ELD on pipe %c\n", pipe_name(pipe)); + + if (IS_VALLEYVIEW(connector->dev)) { + struct intel_encoder *intel_encoder; + struct intel_digital_port *intel_dig_port; + + intel_encoder = intel_attached_encoder(connector); + intel_dig_port = enc_to_dig_port(&intel_encoder->base); + i = intel_dig_port->port; + } else { + i = I915_READ(aud_cntl_st); + i = (i >> 29) & DIP_PORT_SEL_MASK; + /* DIP_Port_Select, 0x1 = PortB */ + } + + if (!i) { + DRM_DEBUG_DRIVER("Audio directed to unknown port\n"); + /* operate blindly on all ports */ + eldv = IBX_ELD_VALIDB; + eldv |= IBX_ELD_VALIDB << 4; + eldv |= IBX_ELD_VALIDB << 8; + } else { + DRM_DEBUG_DRIVER("ELD on port %c\n", port_name(i)); + eldv = IBX_ELD_VALIDB << ((i - 1) * 4); + } + + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { + DRM_DEBUG_DRIVER("ELD: DisplayPort detected\n"); + eld[5] |= (1 << 2); /* Conn_Type, 0x1 = DisplayPort */ + I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ + } else { + I915_WRITE(aud_config, audio_config_hdmi_pixel_clock(mode)); + } + + if (intel_eld_uptodate(connector, + aud_cntrl_st2, eldv, + aud_cntl_st, IBX_ELD_ADDRESS, + hdmiw_hdmiedid)) + return; + + i = I915_READ(aud_cntrl_st2); + i &= ~eldv; + I915_WRITE(aud_cntrl_st2, i); + + if (!eld[0]) + return; + + i = I915_READ(aud_cntl_st); + i &= ~IBX_ELD_ADDRESS; + I915_WRITE(aud_cntl_st, i); + + len = min_t(uint8_t, eld[2], 21); /* 84 bytes of hw ELD buffer */ + DRM_DEBUG_DRIVER("ELD size %d\n", len); + for (i = 0; i < len; i++) + I915_WRITE(hdmiw_hdmiedid, *((uint32_t *)eld + i)); + + i = I915_READ(aud_cntrl_st2); + i |= eldv; + I915_WRITE(aud_cntrl_st2, i); +} + +void intel_write_eld(struct drm_encoder *encoder, + struct drm_display_mode *mode) +{ + struct drm_crtc *crtc = encoder->crtc; + struct drm_connector *connector; + struct drm_device *dev = encoder->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + + connector = drm_select_eld(encoder, mode); + if (!connector) + return; + + DRM_DEBUG_DRIVER("ELD on [CONNECTOR:%d:%s], [ENCODER:%d:%s]\n", + connector->base.id, + connector->name, + connector->encoder->base.id, + connector->encoder->name); + + connector->eld[6] = drm_av_sync_delay(connector, mode) / 2; + + if (dev_priv->display.write_eld) + dev_priv->display.write_eld(connector, crtc, mode); +} + +/** + * intel_init_audio - Set up chip specific audio functions + * @dev: drm device + */ +void intel_init_audio(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + if (IS_G4X(dev)) + dev_priv->display.write_eld = g4x_write_eld; + else if (IS_VALLEYVIEW(dev)) + dev_priv->display.write_eld = ironlake_write_eld; + else if (IS_HASWELL(dev) || INTEL_INFO(dev)->gen >= 8) + dev_priv->display.write_eld = haswell_write_eld; + else if (HAS_PCH_SPLIT(dev)) + dev_priv->display.write_eld = ironlake_write_eld; +} diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 6e9d4e16..aeda09d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -406,7 +406,7 @@ static void vlv_clock(int refclk, intel_clock_t *clock) /** * Returns whether any output on the specified pipe is of the specified type */ -static bool intel_pipe_has_type(struct intel_crtc *crtc, int type) +bool intel_pipe_has_type(struct intel_crtc *crtc, int type) { struct drm_device *dev = crtc->base.dev; struct intel_encoder *encoder; @@ -7940,316 +7940,6 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, return true; } -static struct { - int clock; - u32 config; -} hdmi_audio_clock[] = { - { DIV_ROUND_UP(25200 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_25175 }, - { 25200, AUD_CONFIG_PIXEL_CLOCK_HDMI_25200 }, /* default per bspec */ - { 27000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27000 }, - { 27000 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27027 }, - { 54000, AUD_CONFIG_PIXEL_CLOCK_HDMI_54000 }, - { 54000 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_54054 }, - { DIV_ROUND_UP(74250 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_74176 }, - { 74250, AUD_CONFIG_PIXEL_CLOCK_HDMI_74250 }, - { DIV_ROUND_UP(148500 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_148352 }, - { 148500, AUD_CONFIG_PIXEL_CLOCK_HDMI_148500 }, -}; - -/* get AUD_CONFIG_PIXEL_CLOCK_HDMI_* value for mode */ -static u32 audio_config_hdmi_pixel_clock(struct drm_display_mode *mode) -{ - int i; - - for (i = 0; i < ARRAY_SIZE(hdmi_audio_clock); i++) { - if (mode->clock == hdmi_audio_clock[i].clock) - break; - } - - if (i == ARRAY_SIZE(hdmi_audio_clock)) { - DRM_DEBUG_KMS("HDMI audio pixel clock setting for %d not found, falling back to defaults\n", mode->clock); - i = 1; - } - - DRM_DEBUG_KMS("Configuring HDMI audio for pixel clock %d (0x%08x)\n", - hdmi_audio_clock[i].clock, - hdmi_audio_clock[i].config); - - return hdmi_audio_clock[i].config; -} - -static bool intel_eld_uptodate(struct drm_connector *connector, - int reg_eldv, uint32_t bits_eldv, - int reg_elda, uint32_t bits_elda, - int reg_edid) -{ - struct drm_i915_private *dev_priv = connector->dev->dev_private; - uint8_t *eld = connector->eld; - uint32_t i; - - i = I915_READ(reg_eldv); - i &= bits_eldv; - - if (!eld[0]) - return !i; - - if (!i) - return false; - - i = I915_READ(reg_elda); - i &= ~bits_elda; - I915_WRITE(reg_elda, i); - - for (i = 0; i < eld[2]; i++) - if (I915_READ(reg_edid) != *((uint32_t *)eld + i)) - return false; - - return true; -} - -static void g4x_write_eld(struct drm_connector *connector, - struct drm_crtc *crtc, - struct drm_display_mode *mode) -{ - struct drm_i915_private *dev_priv = connector->dev->dev_private; - uint8_t *eld = connector->eld; - uint32_t eldv; - uint32_t len; - uint32_t i; - - i = I915_READ(G4X_AUD_VID_DID); - - if (i == INTEL_AUDIO_DEVBLC || i == INTEL_AUDIO_DEVCL) - eldv = G4X_ELDV_DEVCL_DEVBLC; - else - eldv = G4X_ELDV_DEVCTG; - - if (intel_eld_uptodate(connector, - G4X_AUD_CNTL_ST, eldv, - G4X_AUD_CNTL_ST, G4X_ELD_ADDR, - G4X_HDMIW_HDMIEDID)) - return; - - i = I915_READ(G4X_AUD_CNTL_ST); - i &= ~(eldv | G4X_ELD_ADDR); - len = (i >> 9) & 0x1f; /* ELD buffer size */ - I915_WRITE(G4X_AUD_CNTL_ST, i); - - if (!eld[0]) - return; - - len = min_t(uint8_t, eld[2], len); - DRM_DEBUG_DRIVER("ELD size %d\n", len); - for (i = 0; i < len; i++) - I915_WRITE(G4X_HDMIW_HDMIEDID, *((uint32_t *)eld + i)); - - i = I915_READ(G4X_AUD_CNTL_ST); - i |= eldv; - I915_WRITE(G4X_AUD_CNTL_ST, i); -} - -static void haswell_write_eld(struct drm_connector *connector, - struct drm_crtc *crtc, - struct drm_display_mode *mode) -{ - struct drm_i915_private *dev_priv = connector->dev->dev_private; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - uint8_t *eld = connector->eld; - uint32_t eldv; - uint32_t i; - int len; - int pipe = to_intel_crtc(crtc)->pipe; - int tmp; - - int hdmiw_hdmiedid = HSW_AUD_EDID_DATA(pipe); - int aud_cntl_st = HSW_AUD_DIP_ELD_CTRL(pipe); - int aud_config = HSW_AUD_CFG(pipe); - int aud_cntrl_st2 = HSW_AUD_PIN_ELD_CP_VLD; - - /* Audio output enable */ - DRM_DEBUG_DRIVER("HDMI audio: enable codec\n"); - tmp = I915_READ(aud_cntrl_st2); - tmp |= (AUDIO_OUTPUT_ENABLE_A << (pipe * 4)); - I915_WRITE(aud_cntrl_st2, tmp); - POSTING_READ(aud_cntrl_st2); - - assert_pipe_disabled(dev_priv, to_intel_crtc(crtc)->pipe); - - /* Set ELD valid state */ - tmp = I915_READ(aud_cntrl_st2); - DRM_DEBUG_DRIVER("HDMI audio: pin eld vld status=0x%08x\n", tmp); - tmp |= (AUDIO_ELD_VALID_A << (pipe * 4)); - I915_WRITE(aud_cntrl_st2, tmp); - tmp = I915_READ(aud_cntrl_st2); - DRM_DEBUG_DRIVER("HDMI audio: eld vld status=0x%08x\n", tmp); - - /* Enable HDMI mode */ - tmp = I915_READ(aud_config); - DRM_DEBUG_DRIVER("HDMI audio: audio conf: 0x%08x\n", tmp); - /* clear N_programing_enable and N_value_index */ - tmp &= ~(AUD_CONFIG_N_VALUE_INDEX | AUD_CONFIG_N_PROG_ENABLE); - I915_WRITE(aud_config, tmp); - - DRM_DEBUG_DRIVER("ELD on pipe %c\n", pipe_name(pipe)); - - eldv = AUDIO_ELD_VALID_A << (pipe * 4); - - if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { - DRM_DEBUG_DRIVER("ELD: DisplayPort detected\n"); - eld[5] |= (1 << 2); /* Conn_Type, 0x1 = DisplayPort */ - I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ - } else { - I915_WRITE(aud_config, audio_config_hdmi_pixel_clock(mode)); - } - - if (intel_eld_uptodate(connector, - aud_cntrl_st2, eldv, - aud_cntl_st, IBX_ELD_ADDRESS, - hdmiw_hdmiedid)) - return; - - i = I915_READ(aud_cntrl_st2); - i &= ~eldv; - I915_WRITE(aud_cntrl_st2, i); - - if (!eld[0]) - return; - - i = I915_READ(aud_cntl_st); - i &= ~IBX_ELD_ADDRESS; - I915_WRITE(aud_cntl_st, i); - i = (i >> 29) & DIP_PORT_SEL_MASK; /* DIP_Port_Select, 0x1 = PortB */ - DRM_DEBUG_DRIVER("port num:%d\n", i); - - len = min_t(uint8_t, eld[2], 21); /* 84 bytes of hw ELD buffer */ - DRM_DEBUG_DRIVER("ELD size %d\n", len); - for (i = 0; i < len; i++) - I915_WRITE(hdmiw_hdmiedid, *((uint32_t *)eld + i)); - - i = I915_READ(aud_cntrl_st2); - i |= eldv; - I915_WRITE(aud_cntrl_st2, i); - -} - -static void ironlake_write_eld(struct drm_connector *connector, - struct drm_crtc *crtc, - struct drm_display_mode *mode) -{ - struct drm_i915_private *dev_priv = connector->dev->dev_private; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - uint8_t *eld = connector->eld; - uint32_t eldv; - uint32_t i; - int len; - int hdmiw_hdmiedid; - int aud_config; - int aud_cntl_st; - int aud_cntrl_st2; - int pipe = to_intel_crtc(crtc)->pipe; - - if (HAS_PCH_IBX(connector->dev)) { - hdmiw_hdmiedid = IBX_HDMIW_HDMIEDID(pipe); - aud_config = IBX_AUD_CFG(pipe); - aud_cntl_st = IBX_AUD_CNTL_ST(pipe); - aud_cntrl_st2 = IBX_AUD_CNTL_ST2; - } else if (IS_VALLEYVIEW(connector->dev)) { - hdmiw_hdmiedid = VLV_HDMIW_HDMIEDID(pipe); - aud_config = VLV_AUD_CFG(pipe); - aud_cntl_st = VLV_AUD_CNTL_ST(pipe); - aud_cntrl_st2 = VLV_AUD_CNTL_ST2; - } else { - hdmiw_hdmiedid = CPT_HDMIW_HDMIEDID(pipe); - aud_config = CPT_AUD_CFG(pipe); - aud_cntl_st = CPT_AUD_CNTL_ST(pipe); - aud_cntrl_st2 = CPT_AUD_CNTRL_ST2; - } - - DRM_DEBUG_DRIVER("ELD on pipe %c\n", pipe_name(pipe)); - - if (IS_VALLEYVIEW(connector->dev)) { - struct intel_encoder *intel_encoder; - struct intel_digital_port *intel_dig_port; - - intel_encoder = intel_attached_encoder(connector); - intel_dig_port = enc_to_dig_port(&intel_encoder->base); - i = intel_dig_port->port; - } else { - i = I915_READ(aud_cntl_st); - i = (i >> 29) & DIP_PORT_SEL_MASK; - /* DIP_Port_Select, 0x1 = PortB */ - } - - if (!i) { - DRM_DEBUG_DRIVER("Audio directed to unknown port\n"); - /* operate blindly on all ports */ - eldv = IBX_ELD_VALIDB; - eldv |= IBX_ELD_VALIDB << 4; - eldv |= IBX_ELD_VALIDB << 8; - } else { - DRM_DEBUG_DRIVER("ELD on port %c\n", port_name(i)); - eldv = IBX_ELD_VALIDB << ((i - 1) * 4); - } - - if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { - DRM_DEBUG_DRIVER("ELD: DisplayPort detected\n"); - eld[5] |= (1 << 2); /* Conn_Type, 0x1 = DisplayPort */ - I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ - } else { - I915_WRITE(aud_config, audio_config_hdmi_pixel_clock(mode)); - } - - if (intel_eld_uptodate(connector, - aud_cntrl_st2, eldv, - aud_cntl_st, IBX_ELD_ADDRESS, - hdmiw_hdmiedid)) - return; - - i = I915_READ(aud_cntrl_st2); - i &= ~eldv; - I915_WRITE(aud_cntrl_st2, i); - - if (!eld[0]) - return; - - i = I915_READ(aud_cntl_st); - i &= ~IBX_ELD_ADDRESS; - I915_WRITE(aud_cntl_st, i); - - len = min_t(uint8_t, eld[2], 21); /* 84 bytes of hw ELD buffer */ - DRM_DEBUG_DRIVER("ELD size %d\n", len); - for (i = 0; i < len; i++) - I915_WRITE(hdmiw_hdmiedid, *((uint32_t *)eld + i)); - - i = I915_READ(aud_cntrl_st2); - i |= eldv; - I915_WRITE(aud_cntrl_st2, i); -} - -void intel_write_eld(struct drm_encoder *encoder, - struct drm_display_mode *mode) -{ - struct drm_crtc *crtc = encoder->crtc; - struct drm_connector *connector; - struct drm_device *dev = encoder->dev; - struct drm_i915_private *dev_priv = dev->dev_private; - - connector = drm_select_eld(encoder, mode); - if (!connector) - return; - - DRM_DEBUG_DRIVER("ELD on [CONNECTOR:%d:%s], [ENCODER:%d:%s]\n", - connector->base.id, - connector->name, - connector->encoder->base.id, - connector->encoder->name); - - connector->eld[6] = drm_av_sync_delay(connector, mode) / 2; - - if (dev_priv->display.write_eld) - dev_priv->display.write_eld(connector, crtc, mode); -} - static void i845_update_cursor(struct drm_crtc *crtc, u32 base) { struct drm_device *dev = crtc->dev; @@ -12638,33 +12328,25 @@ static void intel_init_display(struct drm_device *dev) dev_priv->display.get_display_clock_speed = i830_get_display_clock_speed; - if (IS_G4X(dev)) { - dev_priv->display.write_eld = g4x_write_eld; - } else if (IS_GEN5(dev)) { + if (IS_GEN5(dev)) { dev_priv->display.fdi_link_train = ironlake_fdi_link_train; - dev_priv->display.write_eld = ironlake_write_eld; } else if (IS_GEN6(dev)) { dev_priv->display.fdi_link_train = gen6_fdi_link_train; - dev_priv->display.write_eld = ironlake_write_eld; dev_priv->display.modeset_global_resources = snb_modeset_global_resources; } else if (IS_IVYBRIDGE(dev)) { /* FIXME: detect B0+ stepping and use auto training */ dev_priv->display.fdi_link_train = ivb_manual_fdi_link_train; - dev_priv->display.write_eld = ironlake_write_eld; dev_priv->display.modeset_global_resources = ivb_modeset_global_resources; } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { dev_priv->display.fdi_link_train = hsw_fdi_link_train; - dev_priv->display.write_eld = haswell_write_eld; dev_priv->display.modeset_global_resources = haswell_modeset_global_resources; } else if (IS_VALLEYVIEW(dev)) { dev_priv->display.modeset_global_resources = valleyview_modeset_global_resources; - dev_priv->display.write_eld = ironlake_write_eld; } else if (INTEL_INFO(dev)->gen >= 9) { - dev_priv->display.write_eld = haswell_write_eld; dev_priv->display.modeset_global_resources = haswell_modeset_global_resources; } @@ -12919,6 +12601,7 @@ void intel_modeset_init(struct drm_device *dev) return; intel_init_display(dev); + intel_init_audio(dev); if (IS_GEN2(dev)) { dev->mode_config.max_width = 2048; diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 5ab813c..3bbc4fe 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -848,6 +848,11 @@ void intel_frontbuffer_flip(struct drm_device *dev, void intel_fb_obj_flush(struct drm_i915_gem_object *obj, bool retire); +/* intel_audio.c */ +void intel_init_audio(struct drm_device *dev); +void intel_write_eld(struct drm_encoder *encoder, + struct drm_display_mode *mode); + /* intel_display.c */ const char *intel_output_name(int output); bool intel_has_pending_fb_unpin(struct drm_device *dev); @@ -873,6 +878,7 @@ int intel_get_pipe_from_crtc_id(struct drm_device *dev, void *data, struct drm_file *file_priv); enum transcoder intel_pipe_to_cpu_transcoder(struct drm_i915_private *dev_priv, enum pipe pipe); +bool intel_pipe_has_type(struct intel_crtc *crtc, int type); static inline void intel_wait_for_vblank(struct drm_device *dev, int pipe) { @@ -924,8 +930,6 @@ void assert_fdi_rx_pll(struct drm_i915_private *dev_priv, void assert_pipe(struct drm_i915_private *dev_priv, enum pipe pipe, bool state); #define assert_pipe_enabled(d, p) assert_pipe(d, p, true) #define assert_pipe_disabled(d, p) assert_pipe(d, p, false) -void intel_write_eld(struct drm_encoder *encoder, - struct drm_display_mode *mode); unsigned long intel_gen4_compute_page_offset(int *x, int *y, unsigned int tiling_mode, unsigned int bpp, -- cgit v0.10.2 From 87fcb2ad4582c6fb8fb661a943670485df0a1810 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:44 +0200 Subject: drm/i915/audio: constify hdmi audio clock struct Const is good. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 167903b..e761f2c 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -28,7 +28,7 @@ #include "intel_drv.h" #include "i915_drv.h" -static struct { +static const struct { int clock; u32 config; } hdmi_audio_clock[] = { -- cgit v0.10.2 From f9f682ae3530f9287379dd91d1b7deaef4abf391 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:45 +0200 Subject: drm/i915/audio: beat some sense into the variable types and names Most importantly, "i" need not be the universal variable used for everything. No functional changes. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index e761f2c..00e9bfc 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -73,20 +73,21 @@ static bool intel_eld_uptodate(struct drm_connector *connector, { struct drm_i915_private *dev_priv = connector->dev->dev_private; uint8_t *eld = connector->eld; - uint32_t i; + uint32_t tmp; + int i; - i = I915_READ(reg_eldv); - i &= bits_eldv; + tmp = I915_READ(reg_eldv); + tmp &= bits_eldv; if (!eld[0]) - return !i; + return !tmp; - if (!i) + if (!tmp) return false; - i = I915_READ(reg_elda); - i &= ~bits_elda; - I915_WRITE(reg_elda, i); + tmp = I915_READ(reg_elda); + tmp &= ~bits_elda; + I915_WRITE(reg_elda, tmp); for (i = 0; i < eld[2]; i++) if (I915_READ(reg_edid) != *((uint32_t *)eld + i)) @@ -102,12 +103,11 @@ static void g4x_write_eld(struct drm_connector *connector, struct drm_i915_private *dev_priv = connector->dev->dev_private; uint8_t *eld = connector->eld; uint32_t eldv; - uint32_t len; - uint32_t i; - - i = I915_READ(G4X_AUD_VID_DID); + uint32_t tmp; + int len, i; - if (i == INTEL_AUDIO_DEVBLC || i == INTEL_AUDIO_DEVCL) + tmp = I915_READ(G4X_AUD_VID_DID); + if (tmp == INTEL_AUDIO_DEVBLC || tmp == INTEL_AUDIO_DEVCL) eldv = G4X_ELDV_DEVCL_DEVBLC; else eldv = G4X_ELDV_DEVCTG; @@ -118,22 +118,22 @@ static void g4x_write_eld(struct drm_connector *connector, G4X_HDMIW_HDMIEDID)) return; - i = I915_READ(G4X_AUD_CNTL_ST); - i &= ~(eldv | G4X_ELD_ADDR); - len = (i >> 9) & 0x1f; /* ELD buffer size */ - I915_WRITE(G4X_AUD_CNTL_ST, i); + tmp = I915_READ(G4X_AUD_CNTL_ST); + tmp &= ~(eldv | G4X_ELD_ADDR); + len = (tmp >> 9) & 0x1f; /* ELD buffer size */ + I915_WRITE(G4X_AUD_CNTL_ST, tmp); if (!eld[0]) return; - len = min_t(uint8_t, eld[2], len); + len = min_t(int, eld[2], len); DRM_DEBUG_DRIVER("ELD size %d\n", len); for (i = 0; i < len; i++) I915_WRITE(G4X_HDMIW_HDMIEDID, *((uint32_t *)eld + i)); - i = I915_READ(G4X_AUD_CNTL_ST); - i |= eldv; - I915_WRITE(G4X_AUD_CNTL_ST, i); + tmp = I915_READ(G4X_AUD_CNTL_ST); + tmp |= eldv; + I915_WRITE(G4X_AUD_CNTL_ST, tmp); } static void haswell_write_eld(struct drm_connector *connector, @@ -144,11 +144,10 @@ static void haswell_write_eld(struct drm_connector *connector, struct intel_crtc *intel_crtc = to_intel_crtc(crtc); uint8_t *eld = connector->eld; uint32_t eldv; - uint32_t i; - int len; - int pipe = to_intel_crtc(crtc)->pipe; - int tmp; - + uint32_t tmp; + int len, i; + enum pipe pipe = to_intel_crtc(crtc)->pipe; + enum port port; int hdmiw_hdmiedid = HSW_AUD_EDID_DATA(pipe); int aud_cntl_st = HSW_AUD_DIP_ELD_CTRL(pipe); int aud_config = HSW_AUD_CFG(pipe); @@ -196,28 +195,27 @@ static void haswell_write_eld(struct drm_connector *connector, hdmiw_hdmiedid)) return; - i = I915_READ(aud_cntrl_st2); - i &= ~eldv; - I915_WRITE(aud_cntrl_st2, i); + tmp = I915_READ(aud_cntrl_st2); + tmp &= ~eldv; + I915_WRITE(aud_cntrl_st2, tmp); if (!eld[0]) return; - i = I915_READ(aud_cntl_st); - i &= ~IBX_ELD_ADDRESS; - I915_WRITE(aud_cntl_st, i); - i = (i >> 29) & DIP_PORT_SEL_MASK; /* DIP_Port_Select, 0x1 = PortB */ - DRM_DEBUG_DRIVER("port num:%d\n", i); + tmp = I915_READ(aud_cntl_st); + tmp &= ~IBX_ELD_ADDRESS; + I915_WRITE(aud_cntl_st, tmp); + port = (tmp >> 29) & DIP_PORT_SEL_MASK; /* DIP_Port_Select, 0x1 = PortB */ + DRM_DEBUG_DRIVER("port num:%d\n", port); - len = min_t(uint8_t, eld[2], 21); /* 84 bytes of hw ELD buffer */ + len = min_t(int, eld[2], 21); /* 84 bytes of hw ELD buffer */ DRM_DEBUG_DRIVER("ELD size %d\n", len); for (i = 0; i < len; i++) I915_WRITE(hdmiw_hdmiedid, *((uint32_t *)eld + i)); - i = I915_READ(aud_cntrl_st2); - i |= eldv; - I915_WRITE(aud_cntrl_st2, i); - + tmp = I915_READ(aud_cntrl_st2); + tmp |= eldv; + I915_WRITE(aud_cntrl_st2, tmp); } static void ironlake_write_eld(struct drm_connector *connector, @@ -228,13 +226,14 @@ static void ironlake_write_eld(struct drm_connector *connector, struct intel_crtc *intel_crtc = to_intel_crtc(crtc); uint8_t *eld = connector->eld; uint32_t eldv; - uint32_t i; - int len; + uint32_t tmp; + int len, i; int hdmiw_hdmiedid; int aud_config; int aud_cntl_st; int aud_cntrl_st2; - int pipe = to_intel_crtc(crtc)->pipe; + enum pipe pipe = to_intel_crtc(crtc)->pipe; + enum port port; if (HAS_PCH_IBX(connector->dev)) { hdmiw_hdmiedid = IBX_HDMIW_HDMIEDID(pipe); @@ -261,22 +260,22 @@ static void ironlake_write_eld(struct drm_connector *connector, intel_encoder = intel_attached_encoder(connector); intel_dig_port = enc_to_dig_port(&intel_encoder->base); - i = intel_dig_port->port; + port = intel_dig_port->port; } else { - i = I915_READ(aud_cntl_st); - i = (i >> 29) & DIP_PORT_SEL_MASK; + tmp = I915_READ(aud_cntl_st); + port = (tmp >> 29) & DIP_PORT_SEL_MASK; /* DIP_Port_Select, 0x1 = PortB */ } - if (!i) { + if (!port) { DRM_DEBUG_DRIVER("Audio directed to unknown port\n"); /* operate blindly on all ports */ eldv = IBX_ELD_VALIDB; eldv |= IBX_ELD_VALIDB << 4; eldv |= IBX_ELD_VALIDB << 8; } else { - DRM_DEBUG_DRIVER("ELD on port %c\n", port_name(i)); - eldv = IBX_ELD_VALIDB << ((i - 1) * 4); + DRM_DEBUG_DRIVER("ELD on port %c\n", port_name(port)); + eldv = IBX_ELD_VALIDB << ((port - 1) * 4); } if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { @@ -293,25 +292,25 @@ static void ironlake_write_eld(struct drm_connector *connector, hdmiw_hdmiedid)) return; - i = I915_READ(aud_cntrl_st2); - i &= ~eldv; - I915_WRITE(aud_cntrl_st2, i); + tmp = I915_READ(aud_cntrl_st2); + tmp &= ~eldv; + I915_WRITE(aud_cntrl_st2, tmp); if (!eld[0]) return; - i = I915_READ(aud_cntl_st); - i &= ~IBX_ELD_ADDRESS; - I915_WRITE(aud_cntl_st, i); + tmp = I915_READ(aud_cntl_st); + tmp &= ~IBX_ELD_ADDRESS; + I915_WRITE(aud_cntl_st, tmp); - len = min_t(uint8_t, eld[2], 21); /* 84 bytes of hw ELD buffer */ + len = min_t(int, eld[2], 21); /* 84 bytes of hw ELD buffer */ DRM_DEBUG_DRIVER("ELD size %d\n", len); for (i = 0; i < len; i++) I915_WRITE(hdmiw_hdmiedid, *((uint32_t *)eld + i)); - i = I915_READ(aud_cntrl_st2); - i |= eldv; - I915_WRITE(aud_cntrl_st2, i); + tmp = I915_READ(aud_cntrl_st2); + tmp |= eldv; + I915_WRITE(aud_cntrl_st2, tmp); } void intel_write_eld(struct drm_encoder *encoder, -- cgit v0.10.2 From 33d1e7c6f41656b62a0736a9f0eb977f5f1ff88f Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:46 +0200 Subject: drm/i915: pass intel_encoder to intel_write_eld Everything else can be derived from that. No functional changes. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 00e9bfc..829afd5 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -313,10 +313,11 @@ static void ironlake_write_eld(struct drm_connector *connector, I915_WRITE(aud_cntrl_st2, tmp); } -void intel_write_eld(struct drm_encoder *encoder, - struct drm_display_mode *mode) +void intel_write_eld(struct intel_encoder *intel_encoder) { - struct drm_crtc *crtc = encoder->crtc; + struct drm_encoder *encoder = &intel_encoder->base; + struct intel_crtc *crtc = to_intel_crtc(encoder->crtc); + struct drm_display_mode *mode = &crtc->config.adjusted_mode; struct drm_connector *connector; struct drm_device *dev = encoder->dev; struct drm_i915_private *dev_priv = dev->dev_private; @@ -334,7 +335,7 @@ void intel_write_eld(struct drm_encoder *encoder, connector->eld[6] = drm_av_sync_delay(connector, mode) / 2; if (dev_priv->display.write_eld) - dev_priv->display.write_eld(connector, crtc, mode); + dev_priv->display.write_eld(connector, encoder->crtc, mode); } /** diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index cb5367c..2688bc9 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1126,7 +1126,7 @@ static void intel_ddi_pre_enable(struct intel_encoder *intel_encoder) /* write eld */ DRM_DEBUG_DRIVER("DDI audio: write eld information\n"); - intel_write_eld(encoder, &crtc->config.adjusted_mode); + intel_write_eld(intel_encoder); } if (type == INTEL_OUTPUT_EDP) { diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 64c8e04..b474a29 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1250,7 +1250,7 @@ static void intel_dp_prepare(struct intel_encoder *encoder) DRM_DEBUG_DRIVER("Enabling DP audio on pipe %c\n", pipe_name(crtc->pipe)); intel_dp->DP |= DP_AUDIO_OUTPUT_ENABLE; - intel_write_eld(&encoder->base, adjusted_mode); + intel_write_eld(encoder); } /* Split out the IBX/CPU vs CPT settings */ diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 3bbc4fe..bf72a92 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -850,8 +850,7 @@ void intel_fb_obj_flush(struct drm_i915_gem_object *obj, bool retire); /* intel_audio.c */ void intel_init_audio(struct drm_device *dev); -void intel_write_eld(struct drm_encoder *encoder, - struct drm_display_mode *mode); +void intel_write_eld(struct intel_encoder *encoder); /* intel_display.c */ const char *intel_output_name(int output); diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 8b5f3aa..07b5ebd 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -666,7 +666,7 @@ static void intel_hdmi_prepare(struct intel_encoder *encoder) DRM_DEBUG_DRIVER("Enabling HDMI audio on pipe %c\n", pipe_name(crtc->pipe)); hdmi_val |= SDVO_AUDIO_ENABLE; - intel_write_eld(&encoder->base, adjusted_mode); + intel_write_eld(encoder); } if (HAS_PCH_CPT(dev)) -- cgit v0.10.2 From 820d2d77482810702758381808bdbb64595298e2 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:47 +0200 Subject: drm/i915/audio: pass intel_encoder on to platform specific ELD functions This will simplify things later on. No functional changes. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 627b7e7..6a73803 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -434,6 +434,7 @@ struct drm_i915_error_state { }; struct intel_connector; +struct intel_encoder; struct intel_crtc_config; struct intel_plane_config; struct intel_crtc; @@ -483,7 +484,7 @@ struct drm_i915_display_funcs { void (*crtc_disable)(struct drm_crtc *crtc); void (*off)(struct drm_crtc *crtc); void (*write_eld)(struct drm_connector *connector, - struct drm_crtc *crtc, + struct intel_encoder *encoder, struct drm_display_mode *mode); void (*fdi_link_train)(struct drm_crtc *crtc); void (*init_clock_gating)(struct drm_device *dev); @@ -2798,7 +2799,6 @@ static inline bool intel_gmbus_is_forced_bit(struct i2c_adapter *adapter) extern void intel_i2c_reset(struct drm_device *dev); /* intel_opregion.c */ -struct intel_encoder; #ifdef CONFIG_ACPI extern int intel_opregion_setup(struct drm_device *dev); extern void intel_opregion_init(struct drm_device *dev); diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 829afd5..4a384d7 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -97,7 +97,7 @@ static bool intel_eld_uptodate(struct drm_connector *connector, } static void g4x_write_eld(struct drm_connector *connector, - struct drm_crtc *crtc, + struct intel_encoder *encoder, struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = connector->dev->dev_private; @@ -137,16 +137,16 @@ static void g4x_write_eld(struct drm_connector *connector, } static void haswell_write_eld(struct drm_connector *connector, - struct drm_crtc *crtc, + struct intel_encoder *encoder, struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = connector->dev->dev_private; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); uint8_t *eld = connector->eld; uint32_t eldv; uint32_t tmp; int len, i; - enum pipe pipe = to_intel_crtc(crtc)->pipe; + enum pipe pipe = intel_crtc->pipe; enum port port; int hdmiw_hdmiedid = HSW_AUD_EDID_DATA(pipe); int aud_cntl_st = HSW_AUD_DIP_ELD_CTRL(pipe); @@ -160,7 +160,7 @@ static void haswell_write_eld(struct drm_connector *connector, I915_WRITE(aud_cntrl_st2, tmp); POSTING_READ(aud_cntrl_st2); - assert_pipe_disabled(dev_priv, to_intel_crtc(crtc)->pipe); + assert_pipe_disabled(dev_priv, pipe); /* Set ELD valid state */ tmp = I915_READ(aud_cntrl_st2); @@ -219,11 +219,11 @@ static void haswell_write_eld(struct drm_connector *connector, } static void ironlake_write_eld(struct drm_connector *connector, - struct drm_crtc *crtc, + struct intel_encoder *encoder, struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = connector->dev->dev_private; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); uint8_t *eld = connector->eld; uint32_t eldv; uint32_t tmp; @@ -232,7 +232,7 @@ static void ironlake_write_eld(struct drm_connector *connector, int aud_config; int aud_cntl_st; int aud_cntrl_st2; - enum pipe pipe = to_intel_crtc(crtc)->pipe; + enum pipe pipe = intel_crtc->pipe; enum port port; if (HAS_PCH_IBX(connector->dev)) { @@ -255,11 +255,9 @@ static void ironlake_write_eld(struct drm_connector *connector, DRM_DEBUG_DRIVER("ELD on pipe %c\n", pipe_name(pipe)); if (IS_VALLEYVIEW(connector->dev)) { - struct intel_encoder *intel_encoder; struct intel_digital_port *intel_dig_port; - intel_encoder = intel_attached_encoder(connector); - intel_dig_port = enc_to_dig_port(&intel_encoder->base); + intel_dig_port = enc_to_dig_port(&encoder->base); port = intel_dig_port->port; } else { tmp = I915_READ(aud_cntl_st); @@ -335,7 +333,7 @@ void intel_write_eld(struct intel_encoder *intel_encoder) connector->eld[6] = drm_av_sync_delay(connector, mode) / 2; if (dev_priv->display.write_eld) - dev_priv->display.write_eld(connector, encoder->crtc, mode); + dev_priv->display.write_eld(connector, intel_encoder, mode); } /** -- cgit v0.10.2 From 6847d71bfc9b2855fb5fd0517309fb1adf9203ff Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Mon, 27 Oct 2014 17:47:52 -0200 Subject: drm/i915: transform INTEL_OUTPUT_* into an enum Because I got annoyed that I had to document what values "int ddi_personality" is supposed to hold. A good side-effect of this change is that now the compilers can do some additional checks on our code, which may prevent some bugs in the future. A bad side-effect of this change is that now the compilers do some additional checks on our code and complain when a switch statement doesn't check for all possible values, so we need to add "default" cases to all those switches. Hopefully, this may help preventing confusions against DRM_MODE_CONNECTOR_* and DRM_MODE_ENCODER_*. I guess that just by looking at the patch, some people will think this change is not worth its benefits. In this case, I don't really mind dropping the patch. Also, there's probably still a few more places where we can s/int/enum intel_output_type/, but we can change that later, when we spot the places. Signed-off-by: Paulo Zanoni [danvet: Resolve conflict due to reordered patches.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 61ea8da..a79f83c 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2971,6 +2971,8 @@ static int i9xx_pipe_crc_auto_source(struct drm_device *dev, enum pipe pipe, break; } break; + default: + break; } } drm_modeset_unlock_all(dev); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index aeda09d..ee982f5 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -6269,6 +6269,8 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, case INTEL_OUTPUT_DSI: is_dsi = true; break; + default: + break; } num_connectors++; @@ -6597,6 +6599,8 @@ static void ironlake_init_pch_refclk(struct drm_device *dev) if (enc_to_dig_port(&encoder->base)->port == PORT_A) has_cpu_edp = true; break; + default: + break; } } @@ -6901,6 +6905,8 @@ static void lpt_init_pch_refclk(struct drm_device *dev) case INTEL_OUTPUT_ANALOG: has_vga = true; break; + default: + break; } } @@ -6934,6 +6940,8 @@ static int ironlake_get_refclk(struct drm_crtc *crtc) case INTEL_OUTPUT_LVDS: is_lvds = true; break; + default: + break; } num_connectors++; } @@ -7188,6 +7196,8 @@ static uint32_t ironlake_compute_dpll(struct intel_crtc *intel_crtc, case INTEL_OUTPUT_HDMI: is_sdvo = true; break; + default: + break; } num_connectors++; diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index bf72a92..1e58e64 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -94,18 +94,20 @@ /* these are outputs from the chip - integrated only external chips are via DVO or SDVO output */ -#define INTEL_OUTPUT_UNUSED 0 -#define INTEL_OUTPUT_ANALOG 1 -#define INTEL_OUTPUT_DVO 2 -#define INTEL_OUTPUT_SDVO 3 -#define INTEL_OUTPUT_LVDS 4 -#define INTEL_OUTPUT_TVOUT 5 -#define INTEL_OUTPUT_HDMI 6 -#define INTEL_OUTPUT_DISPLAYPORT 7 -#define INTEL_OUTPUT_EDP 8 -#define INTEL_OUTPUT_DSI 9 -#define INTEL_OUTPUT_UNKNOWN 10 -#define INTEL_OUTPUT_DP_MST 11 +enum intel_output_type { + INTEL_OUTPUT_UNUSED = 0, + INTEL_OUTPUT_ANALOG = 1, + INTEL_OUTPUT_DVO = 2, + INTEL_OUTPUT_SDVO = 3, + INTEL_OUTPUT_LVDS = 4, + INTEL_OUTPUT_TVOUT = 5, + INTEL_OUTPUT_HDMI = 6, + INTEL_OUTPUT_DISPLAYPORT = 7, + INTEL_OUTPUT_EDP = 8, + INTEL_OUTPUT_DSI = 9, + INTEL_OUTPUT_UNKNOWN = 10, + INTEL_OUTPUT_DP_MST = 11, +}; #define INTEL_DVO_CHIP_NONE 0 #define INTEL_DVO_CHIP_LVDS 1 @@ -136,7 +138,7 @@ struct intel_encoder { */ struct intel_crtc *new_crtc; - int type; + enum intel_output_type type; unsigned int cloneable; bool connectors_active; void (*hot_plug)(struct intel_encoder *); -- cgit v0.10.2 From c17ed5b5a46ec715620b0548a187d3ed98f4cf0b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:27 +0300 Subject: drm/i915: Warn if trying to register eDP on port != B/C on vlv/chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Only ports B and C have the power sequencer and backlight controls, so complain if we ever try to register an eDP connector on some other port. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index b474a29..3f18580 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -5203,6 +5203,11 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, if (type == DRM_MODE_CONNECTOR_eDP) intel_encoder->type = INTEL_OUTPUT_EDP; + /* eDP only on port B and/or C on vlv/chv */ + if (WARN_ON(IS_VALLEYVIEW(dev) && is_edp(intel_dp) && + port != PORT_B && port != PORT_C)) + return false; + DRM_DEBUG_KMS("Adding %s connector on port %c\n", type == DRM_MODE_CONNECTOR_eDP ? "eDP" : "DP", port_name(port)); -- cgit v0.10.2 From 7a66800e035bb6a64c2d0f8cb315edabbef819b9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:29 +0300 Subject: drm/i915: Remove high level intel_edp_vdd_{on, off}() from hpd/detect MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit want_panel_vdd is a bool so it can't cope with interleaving on/off calls from multiple threads. If we want to make that possible we'd need to convert want_panel_vdd into a proper ref count. But an easier fix is to remove the high level vdd on/off calls from detect/hpd code paths and just rely on the delayed vdd off to avoid needless vdd on<->off ping pong. After this change only the encoder enable/disable paths use the high level functions, which is fine since both the on and off low level edp vdd calls from intel_dp_aux_ch() happen without dropping pps_mutex in between and so want_panel_vdd can't change in between. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 3f18580..9076e9a 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3846,8 +3846,6 @@ intel_dp_probe_oui(struct intel_dp *intel_dp) if (!(intel_dp->dpcd[DP_DOWN_STREAM_PORT_COUNT] & DP_OUI_SUPPORT)) return; - intel_edp_panel_vdd_on(intel_dp); - if (intel_dp_dpcd_read_wake(&intel_dp->aux, DP_SINK_OUI, buf, 3) == 3) DRM_DEBUG_KMS("Sink OUI: %02hx%02hx%02hx\n", buf[0], buf[1], buf[2]); @@ -3855,8 +3853,6 @@ intel_dp_probe_oui(struct intel_dp *intel_dp) if (intel_dp_dpcd_read_wake(&intel_dp->aux, DP_BRANCH_OUI, buf, 3) == 3) DRM_DEBUG_KMS("Branch OUI: %02hx%02hx%02hx\n", buf[0], buf[1], buf[2]); - - intel_edp_panel_vdd_off(intel_dp, false); } static bool @@ -3870,7 +3866,6 @@ intel_dp_probe_mst(struct intel_dp *intel_dp) if (intel_dp->dpcd[DP_DPCD_REV] < 0x12) return false; - intel_edp_panel_vdd_on(intel_dp); if (intel_dp_dpcd_read_wake(&intel_dp->aux, DP_MSTM_CAP, buf, 1)) { if (buf[0] & DP_MST_CAP) { DRM_DEBUG_KMS("Sink is MST capable\n"); @@ -3880,7 +3875,6 @@ intel_dp_probe_mst(struct intel_dp *intel_dp) intel_dp->is_mst = false; } } - intel_edp_panel_vdd_off(intel_dp, false); drm_dp_mst_topology_mgr_set_mst(&intel_dp->mst_mgr, intel_dp->is_mst); return intel_dp->is_mst; @@ -5086,9 +5080,7 @@ static bool intel_edp_init_connector(struct intel_dp *intel_dp, intel_edp_panel_vdd_sanitize(intel_encoder); /* Cache DPCD and EDID for edp. */ - intel_edp_panel_vdd_on(intel_dp); has_dpcd = intel_dp_get_dpcd(intel_dp); - intel_edp_panel_vdd_off(intel_dp, false); if (has_dpcd) { if (intel_dp->dpcd[DP_DPCD_REV] >= 0x11) -- cgit v0.10.2 From 36b5f425dd560d37f5633cf317818377c1db70de Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:30 +0300 Subject: drm/i915: Store power sequencer delays in intel_dp MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The power seqeuncer delays are fixed for a given panel, so we can keep them around once computed. Not that on VLV/CHV we still re-compute them every time we initialize the power seqeuncer registers, but that will change soon enough. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 9076e9a..2694031 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -283,12 +283,10 @@ intel_hrawclk(struct drm_device *dev) static void intel_dp_init_panel_power_sequencer(struct drm_device *dev, - struct intel_dp *intel_dp, - struct edp_power_seq *out); + struct intel_dp *intel_dp); static void intel_dp_init_panel_power_sequencer_registers(struct drm_device *dev, - struct intel_dp *intel_dp, - struct edp_power_seq *out); + struct intel_dp *intel_dp); static void pps_lock(struct intel_dp *intel_dp) { @@ -330,7 +328,6 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) struct drm_i915_private *dev_priv = dev->dev_private; struct intel_encoder *encoder; unsigned int pipes = (1 << PIPE_A) | (1 << PIPE_B); - struct edp_power_seq power_seq; lockdep_assert_held(&dev_priv->pps_mutex); @@ -368,9 +365,8 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) port_name(intel_dig_port->port)); /* init power sequencer on this pipe and port */ - intel_dp_init_panel_power_sequencer(dev, intel_dp, &power_seq); - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, - &power_seq); + intel_dp_init_panel_power_sequencer(dev, intel_dp); + intel_dp_init_panel_power_sequencer_registers(dev, intel_dp); return intel_dp->pps_pipe; } @@ -425,7 +421,6 @@ vlv_initial_power_sequencer_setup(struct intel_dp *intel_dp) struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); struct drm_device *dev = intel_dig_port->base.base.dev; struct drm_i915_private *dev_priv = dev->dev_private; - struct edp_power_seq power_seq; enum port port = intel_dig_port->port; lockdep_assert_held(&dev_priv->pps_mutex); @@ -453,9 +448,8 @@ vlv_initial_power_sequencer_setup(struct intel_dp *intel_dp) DRM_DEBUG_KMS("initial power sequencer for port %c: pipe %c\n", port_name(port), pipe_name(intel_dp->pps_pipe)); - intel_dp_init_panel_power_sequencer(dev, intel_dp, &power_seq); - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, - &power_seq); + intel_dp_init_panel_power_sequencer(dev, intel_dp); + intel_dp_init_panel_power_sequencer_registers(dev, intel_dp); } void vlv_power_sequencer_reset(struct drm_i915_private *dev_priv) @@ -2620,7 +2614,6 @@ static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp) struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); - struct edp_power_seq power_seq; lockdep_assert_held(&dev_priv->pps_mutex); @@ -2648,9 +2641,8 @@ static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp) pipe_name(intel_dp->pps_pipe), port_name(intel_dig_port->port)); /* init power sequencer on this pipe and port */ - intel_dp_init_panel_power_sequencer(dev, intel_dp, &power_seq); - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, - &power_seq); + intel_dp_init_panel_power_sequencer(dev, intel_dp); + intel_dp_init_panel_power_sequencer_registers(dev, intel_dp); } static void vlv_pre_enable_dp(struct intel_encoder *encoder) @@ -4731,11 +4723,11 @@ static void intel_dp_init_panel_power_timestamps(struct intel_dp *intel_dp) static void intel_dp_init_panel_power_sequencer(struct drm_device *dev, - struct intel_dp *intel_dp, - struct edp_power_seq *out) + struct intel_dp *intel_dp) { struct drm_i915_private *dev_priv = dev->dev_private; - struct edp_power_seq cur, vbt, spec, final; + struct edp_power_seq cur, vbt, spec, + *final = &intel_dp->pps_delays; u32 pp_on, pp_off, pp_div, pp; int pp_ctrl_reg, pp_on_reg, pp_off_reg, pp_div_reg; @@ -4802,7 +4794,7 @@ intel_dp_init_panel_power_sequencer(struct drm_device *dev, /* Use the max of the register settings and vbt. If both are * unset, fall back to the spec limits. */ -#define assign_final(field) final.field = (max(cur.field, vbt.field) == 0 ? \ +#define assign_final(field) final->field = (max(cur.field, vbt.field) == 0 ? \ spec.field : \ max(cur.field, vbt.field)) assign_final(t1_t3); @@ -4812,7 +4804,7 @@ intel_dp_init_panel_power_sequencer(struct drm_device *dev, assign_final(t11_t12); #undef assign_final -#define get_delay(field) (DIV_ROUND_UP(final.field, 10)) +#define get_delay(field) (DIV_ROUND_UP(final->field, 10)) intel_dp->panel_power_up_delay = get_delay(t1_t3); intel_dp->backlight_on_delay = get_delay(t8); intel_dp->backlight_off_delay = get_delay(t9); @@ -4826,21 +4818,18 @@ intel_dp_init_panel_power_sequencer(struct drm_device *dev, DRM_DEBUG_KMS("backlight on delay %d, off delay %d\n", intel_dp->backlight_on_delay, intel_dp->backlight_off_delay); - - if (out) - *out = final; } static void intel_dp_init_panel_power_sequencer_registers(struct drm_device *dev, - struct intel_dp *intel_dp, - struct edp_power_seq *seq) + struct intel_dp *intel_dp) { struct drm_i915_private *dev_priv = dev->dev_private; u32 pp_on, pp_off, pp_div, port_sel = 0; int div = HAS_PCH_SPLIT(dev) ? intel_pch_rawclk(dev) : intel_hrawclk(dev); int pp_on_reg, pp_off_reg, pp_div_reg; enum port port = dp_to_dig_port(intel_dp)->port; + const struct edp_power_seq *seq = &intel_dp->pps_delays; lockdep_assert_held(&dev_priv->pps_mutex); @@ -5058,8 +5047,7 @@ void intel_edp_panel_vdd_sanitize(struct intel_encoder *intel_encoder) } static bool intel_edp_init_connector(struct intel_dp *intel_dp, - struct intel_connector *intel_connector, - struct edp_power_seq *power_seq) + struct intel_connector *intel_connector) { struct drm_connector *connector = &intel_connector->base; struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); @@ -5095,7 +5083,7 @@ static bool intel_edp_init_connector(struct intel_dp *intel_dp, /* We now know it's not a ghost, init power sequence regs. */ pps_lock(intel_dp); - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, power_seq); + intel_dp_init_panel_power_sequencer_registers(dev, intel_dp); pps_unlock(intel_dp); mutex_lock(&dev->mode_config.mutex); @@ -5156,7 +5144,6 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, struct drm_device *dev = intel_encoder->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; enum port port = intel_dig_port->port; - struct edp_power_seq power_seq = { 0 }; int type; intel_dp->pps_pipe = INVALID_PIPE; @@ -5246,8 +5233,7 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, vlv_initial_power_sequencer_setup(intel_dp); } else { intel_dp_init_panel_power_timestamps(intel_dp); - intel_dp_init_panel_power_sequencer(dev, intel_dp, - &power_seq); + intel_dp_init_panel_power_sequencer(dev, intel_dp); } pps_unlock(intel_dp); } @@ -5262,7 +5248,7 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, } } - if (!intel_edp_init_connector(intel_dp, intel_connector, &power_seq)) { + if (!intel_edp_init_connector(intel_dp, intel_connector)) { drm_dp_aux_unregister(&intel_dp->aux); if (is_edp(intel_dp)) { cancel_delayed_work_sync(&intel_dp->panel_vdd_work); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 1e58e64..b323c9b 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -592,6 +592,7 @@ struct intel_dp { * this port. Only relevant on VLV/CHV. */ enum pipe pps_pipe; + struct edp_power_seq pps_delays; bool use_tps3; bool can_mst; /* this port supports mst */ -- cgit v0.10.2 From 81ddbc69992a1f29aeb2a5b183e5228cc7c49caf Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:31 +0300 Subject: drm/i915: Don't initialize power seqeuencer delays more than once MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Since we read the current power seqeuncer delays from the registers (as well as looking at the vbt and spec values) we may end up corrupting delays we already initialized when we switch to another pipe and the power seqeuncer there has different values currently in the registers. So make sure we only initialize the delays once even if intel_dp_init_panel_power_sequencer() gets called multiple times. There was some discussion in the review about when exactly we need to unlock the pps. Quoting Bspec: "If this bit is not a zero, it activates the register write protect and writes to those registers will be ignored unless the write protect key value is set in the panel sequencing control register." Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak [danvet: Add Bspec quote per review discussion between Imre and Ville.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 2694031..e1894ce 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -4733,6 +4733,10 @@ intel_dp_init_panel_power_sequencer(struct drm_device *dev, lockdep_assert_held(&dev_priv->pps_mutex); + /* already initialized? */ + if (final->t11_t12 != 0) + return; + if (HAS_PCH_SPLIT(dev)) { pp_ctrl_reg = PCH_PP_CONTROL; pp_on_reg = PCH_PP_ON_DELAYS; -- cgit v0.10.2 From 9f0fb5bec73e4afd84d4dec040a9d54a971620ae Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:32 +0300 Subject: drm/i915: Split power sequencer panel on/off functions to locked and unlocked variants MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We'll be needing to the call the power seqeuencer functions while already holding pps_mutex, so split the locking out to small wrapper functions. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index e1894ce..c97bb5c8 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1555,23 +1555,23 @@ static void intel_edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync) pps_unlock(intel_dp); } -void intel_edp_panel_on(struct intel_dp *intel_dp) +static void edp_panel_on(struct intel_dp *intel_dp) { struct drm_device *dev = intel_dp_to_dev(intel_dp); struct drm_i915_private *dev_priv = dev->dev_private; u32 pp; u32 pp_ctrl_reg; + lockdep_assert_held(&dev_priv->pps_mutex); + if (!is_edp(intel_dp)) return; DRM_DEBUG_KMS("Turn eDP power on\n"); - pps_lock(intel_dp); - if (edp_have_panel_power(intel_dp)) { DRM_DEBUG_KMS("eDP power already on\n"); - goto out; + return; } wait_panel_power_cycle(intel_dp); @@ -1600,12 +1600,20 @@ void intel_edp_panel_on(struct intel_dp *intel_dp) I915_WRITE(pp_ctrl_reg, pp); POSTING_READ(pp_ctrl_reg); } +} - out: +void intel_edp_panel_on(struct intel_dp *intel_dp) +{ + if (!is_edp(intel_dp)) + return; + + pps_lock(intel_dp); + edp_panel_on(intel_dp); pps_unlock(intel_dp); } -void intel_edp_panel_off(struct intel_dp *intel_dp) + +static void edp_panel_off(struct intel_dp *intel_dp) { struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); struct intel_encoder *intel_encoder = &intel_dig_port->base; @@ -1615,13 +1623,13 @@ void intel_edp_panel_off(struct intel_dp *intel_dp) u32 pp; u32 pp_ctrl_reg; + lockdep_assert_held(&dev_priv->pps_mutex); + if (!is_edp(intel_dp)) return; DRM_DEBUG_KMS("Turn eDP power off\n"); - pps_lock(intel_dp); - WARN(!intel_dp->want_panel_vdd, "Need VDD to turn off panel\n"); pp = ironlake_get_pp_control(intel_dp); @@ -1643,7 +1651,15 @@ void intel_edp_panel_off(struct intel_dp *intel_dp) /* We got a reference when we enabled the VDD. */ power_domain = intel_display_port_power_domain(intel_encoder); intel_display_power_put(dev_priv, power_domain); +} +void intel_edp_panel_off(struct intel_dp *intel_dp) +{ + if (!is_edp(intel_dp)) + return; + + pps_lock(intel_dp); + edp_panel_off(intel_dp); pps_unlock(intel_dp); } -- cgit v0.10.2 From 093e3f134e2eff13503f708b81aecc2501e7aecb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:33 +0300 Subject: drm/i915: Hold the pps mutex across the whole panel power enable sequence MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Just grab the pps_mutex once and do all the pps panel startup operations while holding the mutex instead of grabbing the mutex separately for each individual step. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index c97bb5c8..cad4e11 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -113,6 +113,7 @@ static struct intel_dp *intel_attached_dp(struct drm_connector *connector) static void intel_dp_link_down(struct intel_dp *intel_dp); static bool edp_panel_vdd_on(struct intel_dp *intel_dp); static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync); +static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp); int intel_dp_max_link_bw(struct intel_dp *intel_dp) @@ -1539,22 +1540,6 @@ static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync) edp_panel_vdd_schedule_off(intel_dp); } -/* - * Must be paired with intel_edp_panel_vdd_on(). - * Nested calls to these functions are not allowed since - * we drop the lock. Caller must use some higher level - * locking to prevent nested calls from other threads. - */ -static void intel_edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync) -{ - if (!is_edp(intel_dp)) - return; - - pps_lock(intel_dp); - edp_panel_vdd_off(intel_dp, sync); - pps_unlock(intel_dp); -} - static void edp_panel_on(struct intel_dp *intel_dp) { struct drm_device *dev = intel_dp_to_dev(intel_dp); @@ -2552,10 +2537,19 @@ static void intel_enable_dp(struct intel_encoder *encoder) if (WARN_ON(dp_reg & DP_PORT_EN)) return; + pps_lock(intel_dp); + + if (IS_VALLEYVIEW(dev)) + vlv_init_panel_power_sequencer(intel_dp); + intel_dp_enable_port(intel_dp); - intel_edp_panel_vdd_on(intel_dp); - intel_edp_panel_on(intel_dp); - intel_edp_panel_vdd_off(intel_dp, true); + + edp_panel_vdd_on(intel_dp); + edp_panel_on(intel_dp); + edp_panel_vdd_off(intel_dp, true); + + pps_unlock(intel_dp); + intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_ON); intel_dp_start_link_train(intel_dp); intel_dp_complete_link_train(intel_dp); @@ -2633,6 +2627,9 @@ static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp) lockdep_assert_held(&dev_priv->pps_mutex); + if (!is_edp(intel_dp)) + return; + if (intel_dp->pps_pipe == crtc->pipe) return; @@ -2687,12 +2684,6 @@ static void vlv_pre_enable_dp(struct intel_encoder *encoder) mutex_unlock(&dev_priv->dpio_lock); - if (is_edp(intel_dp)) { - pps_lock(intel_dp); - vlv_init_panel_power_sequencer(intel_dp); - pps_unlock(intel_dp); - } - intel_enable_dp(encoder); vlv_wait_port_ready(dev_priv, dport); @@ -2787,12 +2778,6 @@ static void chv_pre_enable_dp(struct intel_encoder *encoder) mutex_unlock(&dev_priv->dpio_lock); - if (is_edp(intel_dp)) { - pps_lock(intel_dp); - vlv_init_panel_power_sequencer(intel_dp); - pps_unlock(intel_dp); - } - intel_enable_dp(encoder); vlv_wait_port_ready(dev_priv, dport); -- cgit v0.10.2 From 61234fa5e5232c35f87d44d9d596af4b10eac255 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:34 +0300 Subject: drm/i915: Wait for PHY port ready before link training on VLV/CHV MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There's no point in checking if the data lanes came out of reset after link training. If the data lanes aren't ready link training will fail anyway. Suggested-by: Todd Previte Cc: Todd Previte Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Acked-by: Todd Previte Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index cad4e11..f3b6cd5 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2550,6 +2550,9 @@ static void intel_enable_dp(struct intel_encoder *encoder) pps_unlock(intel_dp); + if (IS_VALLEYVIEW(dev)) + vlv_wait_port_ready(dev_priv, dp_to_dig_port(intel_dp)); + intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_ON); intel_dp_start_link_train(intel_dp); intel_dp_complete_link_train(intel_dp); @@ -2685,8 +2688,6 @@ static void vlv_pre_enable_dp(struct intel_encoder *encoder) mutex_unlock(&dev_priv->dpio_lock); intel_enable_dp(encoder); - - vlv_wait_port_ready(dev_priv, dport); } static void vlv_dp_pre_pll_enable(struct intel_encoder *encoder) @@ -2779,8 +2780,6 @@ static void chv_pre_enable_dp(struct intel_encoder *encoder) mutex_unlock(&dev_priv->dpio_lock); intel_enable_dp(encoder); - - vlv_wait_port_ready(dev_priv, dport); } static void chv_dp_pre_pll_enable(struct intel_encoder *encoder) -- cgit v0.10.2 From 7b713f50d78b68c04c86eb21e59faf7247aa2cd1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:35 +0300 Subject: drm/i915: Fix eDP link training when switching pipes on VLV/CHV MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When switching from one pipe to another, the power sequencer of the new pipe seems to need a bit of kicking to lock into the port. Even the vdd force bit doesn't work before the power sequencer has been sufficiently kicked, so this must be done before any AUX transactions are attempted. After extensive experimentation I've determined that it's sufficient to first write the port register with the correct values except the port must remain disabled, then we can do a second write to enable the port, after which the power sequencer is operational and allows the port to start up properly. Contrary to my earlier theories we don't need to enable the port with the idle pattern, so let's just use training pattern 1 as that's what other platforms use here. v2: Refine the kick procedure Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index f3b6cd5..8233557 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2517,14 +2517,23 @@ static void intel_dp_enable_port(struct intel_dp *intel_dp) struct drm_device *dev = intel_dp_to_dev(intel_dp); struct drm_i915_private *dev_priv = dev->dev_private; - intel_dp->DP |= DP_PORT_EN; - /* enable with pattern 1 (as per spec) */ _intel_dp_set_link_train(intel_dp, &intel_dp->DP, DP_TRAINING_PATTERN_1); I915_WRITE(intel_dp->output_reg, intel_dp->DP); POSTING_READ(intel_dp->output_reg); + + /* + * Magic for VLV/CHV. We _must_ first set up the register + * without actually enabling the port, and then do another + * write to enable the port. Otherwise link training will + * fail when the power sequencer is freshly used for this port. + */ + intel_dp->DP |= DP_PORT_EN; + + I915_WRITE(intel_dp->output_reg, intel_dp->DP); + POSTING_READ(intel_dp->output_reg); } static void intel_enable_dp(struct intel_encoder *encoder) -- cgit v0.10.2 From 961a0db00977ff98ff99d942ae4c046423e1aebc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:29:42 +0300 Subject: drm/i915: Kick the power sequencer before AUX transactions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When we pick a new power sequencer for the port but we're not doing a full modeset, the power sequencer may have locked on to another port (or no port). So kick it a bit to make sure it controls the port we want. Again just like when we attempt to actually enable the DP port, we must first write the port register with the approriate value except the enable bit, and then we must enable the port to make the power sequencer happy. In this case since we don't want the port actually enabled we just toggle it on and immediately back off. Going forward the power sequencer will keep working on that specific port until again moved to another port. v2: Refine the kick procedure Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 8233557..4f7df37 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -321,6 +321,52 @@ static void pps_unlock(struct intel_dp *intel_dp) intel_display_power_put(dev_priv, power_domain); } +static void +vlv_power_sequencer_kick(struct intel_dp *intel_dp) +{ + struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = intel_dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + enum pipe pipe = intel_dp->pps_pipe; + uint32_t DP; + + if (WARN(I915_READ(intel_dp->output_reg) & DP_PORT_EN, + "skipping pipe %c power seqeuncer kick due to port %c being active\n", + pipe_name(pipe), port_name(intel_dig_port->port))) + return; + + DRM_DEBUG_KMS("kicking pipe %c power sequencer for port %c\n", + pipe_name(pipe), port_name(intel_dig_port->port)); + + /* Preserve the BIOS-computed detected bit. This is + * supposed to be read-only. + */ + DP = I915_READ(intel_dp->output_reg) & DP_DETECTED; + DP |= DP_VOLTAGE_0_4 | DP_PRE_EMPHASIS_0; + DP |= DP_PORT_WIDTH(1); + DP |= DP_LINK_TRAIN_PAT_1; + + if (IS_CHERRYVIEW(dev)) + DP |= DP_PIPE_SELECT_CHV(pipe); + else if (pipe == PIPE_B) + DP |= DP_PIPEB_SELECT; + + /* + * Similar magic as in intel_dp_enable_port(). + * We _must_ do this port enable + disable trick + * to make this power seqeuencer lock onto the port. + * Otherwise even VDD force bit won't work. + */ + I915_WRITE(intel_dp->output_reg, DP); + POSTING_READ(intel_dp->output_reg); + + I915_WRITE(intel_dp->output_reg, DP | DP_PORT_EN); + POSTING_READ(intel_dp->output_reg); + + I915_WRITE(intel_dp->output_reg, DP & ~DP_PORT_EN); + POSTING_READ(intel_dp->output_reg); +} + static enum pipe vlv_power_sequencer_pipe(struct intel_dp *intel_dp) { @@ -369,6 +415,12 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) intel_dp_init_panel_power_sequencer(dev, intel_dp); intel_dp_init_panel_power_sequencer_registers(dev, intel_dp); + /* + * Even vdd force doesn't work until we've made + * the power sequencer lock in on the port. + */ + vlv_power_sequencer_kick(intel_dp); + return intel_dp->pps_pipe; } -- cgit v0.10.2 From 9a42356b964dee6418d8b812109298658cc8dc38 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:29:48 +0300 Subject: drm/i915: Don't kick the power seqeuncer just to check if we have vdd/panel power MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If there's no power sequencer assigned to the port currently we can't very well have vdd or panel power enabled either. If we would try to check that from the pps registers we'd need to pick a power seqeuncer and kick it. So let's skip the register read and the kick. Note that there's still a bit an issue about correctly recovering pps state from resume if the bios is nasty: With this check we'll always assume that the pps is off. But that's better done in a follow-up patch and it shouldn't be too harmful - at most we waste time enabling the pps if it's on already. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak [danvet: Add note about resume issues Imre spotted.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 4f7df37..451dd25 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -597,6 +597,10 @@ static bool edp_have_panel_power(struct intel_dp *intel_dp) lockdep_assert_held(&dev_priv->pps_mutex); + if (IS_VALLEYVIEW(dev) && + intel_dp->pps_pipe == INVALID_PIPE) + return false; + return (I915_READ(_pp_stat_reg(intel_dp)) & PP_ON) != 0; } @@ -607,6 +611,10 @@ static bool edp_have_panel_vdd(struct intel_dp *intel_dp) lockdep_assert_held(&dev_priv->pps_mutex); + if (IS_VALLEYVIEW(dev) && + intel_dp->pps_pipe == INVALID_PIPE) + return false; + return I915_READ(_pp_ctrl_reg(intel_dp)) & EDP_FORCE_VDD; } -- cgit v0.10.2 From 83b8459756659ce55446e3eb97d64b966c60bfb9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:29:51 +0300 Subject: drm/i915: Clear PPS port select when giving up the power sequencer MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit VLV gets confused if two power sequencers have the same port selected. It would seem the port doesn't start up properly in the is case and vlv_wait_port_ready() will fail as will the link training. Clearing the port select in the PP_ON_DELAYS register fixes this problem. CHV doesn't seem to need this, but it doesn't seem to hurt either so let's just do it for both to keep the code between the platforms as uniform as possible. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 451dd25..04d7bc7 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2657,6 +2657,32 @@ static void g4x_pre_enable_dp(struct intel_encoder *encoder) } } +static void vlv_detach_power_sequencer(struct intel_dp *intel_dp) +{ + struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); + struct drm_i915_private *dev_priv = intel_dig_port->base.base.dev->dev_private; + enum pipe pipe = intel_dp->pps_pipe; + int pp_on_reg = VLV_PIPE_PP_ON_DELAYS(pipe); + + edp_panel_vdd_off_sync(intel_dp); + + /* + * VLV seems to get confused when multiple power seqeuencers + * have the same port selected (even if only one has power/vdd + * enabled). The failure manifests as vlv_wait_port_ready() failing + * CHV on the other hand doesn't seem to mind having the same port + * selected in multiple power seqeuencers, but let's clear the + * port select always when logically disconnecting a power sequencer + * from a port. + */ + DRM_DEBUG_KMS("detaching pipe %c power sequencer from port %c\n", + pipe_name(pipe), port_name(intel_dig_port->port)); + I915_WRITE(pp_on_reg, 0); + POSTING_READ(pp_on_reg); + + intel_dp->pps_pipe = INVALID_PIPE; +} + static void vlv_steal_power_sequencer(struct drm_device *dev, enum pipe pipe) { @@ -2683,9 +2709,7 @@ static void vlv_steal_power_sequencer(struct drm_device *dev, pipe_name(pipe), port_name(port)); /* make sure vdd is off before we steal it */ - edp_panel_vdd_off_sync(intel_dp); - - intel_dp->pps_pipe = INVALID_PIPE; + vlv_detach_power_sequencer(intel_dp); } } @@ -2711,7 +2735,7 @@ static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp) * we still have control of it. */ if (intel_dp->pps_pipe != INVALID_PIPE) - edp_panel_vdd_off_sync(intel_dp); + vlv_detach_power_sequencer(intel_dp); /* * We may be stealing the power -- cgit v0.10.2 From ac3c12e4e5a31b75b2621f8f23336bebbe3dbff4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:29:56 +0300 Subject: drm/i915: Warn if stealing non pipe A/B power sequencer MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There's no power sequencer on pipe C on VLV/CHV so scream a bit if we try to steal one from pipes other than A and B. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 04d7bc7..7df024c 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2691,6 +2691,9 @@ static void vlv_steal_power_sequencer(struct drm_device *dev, lockdep_assert_held(&dev_priv->pps_mutex); + if (WARN_ON(pipe != PIPE_A && pipe != PIPE_B)) + return; + list_for_each_entry(encoder, &dev->mode_config.encoder_list, base.head) { struct intel_dp *intel_dp; -- cgit v0.10.2 From a8c3344e069e9c483b1fd328e14fbc6235073649 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:29:59 +0300 Subject: drm/i915: Steal power sequencer in vlv_power_sequencer_pipe() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In case we fumble something and end up picking an already used power seqeuencer in vlv_power_sequencer_pipe() at least try to steal it gracefully. In theory this should never happen though. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 7df024c..1900c4d 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -114,6 +114,8 @@ static void intel_dp_link_down(struct intel_dp *intel_dp); static bool edp_panel_vdd_on(struct intel_dp *intel_dp); static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync); static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp); +static void vlv_steal_power_sequencer(struct drm_device *dev, + enum pipe pipe); int intel_dp_max_link_bw(struct intel_dp *intel_dp) @@ -375,9 +377,13 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) struct drm_i915_private *dev_priv = dev->dev_private; struct intel_encoder *encoder; unsigned int pipes = (1 << PIPE_A) | (1 << PIPE_B); + enum pipe pipe; lockdep_assert_held(&dev_priv->pps_mutex); + /* We should never land here with regular DP ports */ + WARN_ON(!is_edp(intel_dp)); + if (intel_dp->pps_pipe != INVALID_PIPE) return intel_dp->pps_pipe; @@ -403,9 +409,12 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) * are two power sequencers and up to two eDP ports. */ if (WARN_ON(pipes == 0)) - return PIPE_A; + pipe = PIPE_A; + else + pipe = ffs(pipes) - 1; - intel_dp->pps_pipe = ffs(pipes) - 1; + vlv_steal_power_sequencer(dev, pipe); + intel_dp->pps_pipe = pipe; DRM_DEBUG_KMS("picked pipe %c power sequencer for port %c\n", pipe_name(intel_dp->pps_pipe), -- cgit v0.10.2 From 3936fcf453af4de87303f42bb34ef5e3eb13c139 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:30:02 +0300 Subject: drm/i915: Improve VDD/PPS debugs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Print the port name in the VDD/PPS debugs messages. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 1900c4d..21fb4224 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1473,7 +1473,8 @@ static bool edp_panel_vdd_on(struct intel_dp *intel_dp) power_domain = intel_display_port_power_domain(intel_encoder); intel_display_power_get(dev_priv, power_domain); - DRM_DEBUG_KMS("Turning eDP VDD on\n"); + DRM_DEBUG_KMS("Turning eDP port %c VDD on\n", + port_name(intel_dig_port->port)); if (!edp_have_panel_power(intel_dp)) wait_panel_power_cycle(intel_dp); @@ -1492,7 +1493,8 @@ static bool edp_panel_vdd_on(struct intel_dp *intel_dp) * If the panel wasn't on, delay before accessing aux channel */ if (!edp_have_panel_power(intel_dp)) { - DRM_DEBUG_KMS("eDP was not running\n"); + DRM_DEBUG_KMS("eDP port %c panel power wasn't enabled\n", + port_name(intel_dig_port->port)); msleep(intel_dp->panel_power_up_delay); } @@ -1517,7 +1519,8 @@ void intel_edp_panel_vdd_on(struct intel_dp *intel_dp) vdd = edp_panel_vdd_on(intel_dp); pps_unlock(intel_dp); - WARN(!vdd, "eDP VDD already requested on\n"); + WARN(!vdd, "eDP port %c VDD already requested on\n", + port_name(dp_to_dig_port(intel_dp)->port)); } static void edp_panel_vdd_off_sync(struct intel_dp *intel_dp) @@ -1538,7 +1541,8 @@ static void edp_panel_vdd_off_sync(struct intel_dp *intel_dp) if (!edp_have_panel_vdd(intel_dp)) return; - DRM_DEBUG_KMS("Turning eDP VDD off\n"); + DRM_DEBUG_KMS("Turning eDP port %c VDD off\n", + port_name(intel_dig_port->port)); pp = ironlake_get_pp_control(intel_dp); pp &= ~EDP_FORCE_VDD; @@ -1599,7 +1603,8 @@ static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync) if (!is_edp(intel_dp)) return; - WARN(!intel_dp->want_panel_vdd, "eDP VDD not forced on"); + WARN(!intel_dp->want_panel_vdd, "eDP port %c VDD not forced on", + port_name(dp_to_dig_port(intel_dp)->port)); intel_dp->want_panel_vdd = false; @@ -1621,7 +1626,8 @@ static void edp_panel_on(struct intel_dp *intel_dp) if (!is_edp(intel_dp)) return; - DRM_DEBUG_KMS("Turn eDP power on\n"); + DRM_DEBUG_KMS("Turn eDP port %c panel power on\n", + port_name(dp_to_dig_port(intel_dp)->port)); if (edp_have_panel_power(intel_dp)) { DRM_DEBUG_KMS("eDP power already on\n"); @@ -1682,9 +1688,11 @@ static void edp_panel_off(struct intel_dp *intel_dp) if (!is_edp(intel_dp)) return; - DRM_DEBUG_KMS("Turn eDP power off\n"); + DRM_DEBUG_KMS("Turn eDP port %c panel power off\n", + port_name(dp_to_dig_port(intel_dp)->port)); - WARN(!intel_dp->want_panel_vdd, "Need VDD to turn off panel\n"); + WARN(!intel_dp->want_panel_vdd, "Need eDP port %c VDD to turn off panel\n", + port_name(dp_to_dig_port(intel_dp)->port)); pp = ironlake_get_pp_control(intel_dp); /* We need to switch off panel power _and_ force vdd, for otherwise some -- cgit v0.10.2 From e7a89acea863ffdffa66f4c6cda272eebd893c51 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:30:07 +0300 Subject: drm/i915: Warn if panel power is already on when enabling it MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We should never enable the panel power twice. That would indicate a bug somewhere else as we would need to enable the port twice without disabling it in between. Also print the port name. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 21fb4224..9db977b 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1629,10 +1629,10 @@ static void edp_panel_on(struct intel_dp *intel_dp) DRM_DEBUG_KMS("Turn eDP port %c panel power on\n", port_name(dp_to_dig_port(intel_dp)->port)); - if (edp_have_panel_power(intel_dp)) { - DRM_DEBUG_KMS("eDP power already on\n"); + if (WARN(edp_have_panel_power(intel_dp), + "eDP port %c panel power already on\n", + port_name(dp_to_dig_port(intel_dp)->port))) return; - } wait_panel_power_cycle(intel_dp); -- cgit v0.10.2 From 034e43c6c0dc7e70a6edc96987758237cb92f0cd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 21:27:28 +0300 Subject: drm/i915: Warn if stealing power sequencer from an active eDP port MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit eDP ports need the power seqeuncer whenever the port is active. Warn if we accidentally steal the power sequener from an active eDP port. This should not happen unless there's a bug somewhere else, but it's best to scream loudly if it happens to help with debugging. Note that this only checks for active pipes and not for enabled pipes which are turned off with dpms. Which means we might run the risk that the pps might get stolen and we can't reacquire one when enabling the pipe again with dpms on. But on current platforms that's impossible since we only support two edp ports with just two panel power sequencers. So a more elaborate scheme which reserves the pps even when the pipe is inactive isn't required. Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak [danvet: Summarize my discussion with Ville about dpms on/off issues.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 9db977b..90d752d 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2728,6 +2728,10 @@ static void vlv_steal_power_sequencer(struct drm_device *dev, DRM_DEBUG_KMS("stealing pipe %c power sequencer from port %c\n", pipe_name(pipe), port_name(port)); + WARN(encoder->connectors_active, + "stealing pipe %c power sequencer from active eDP port %c\n", + pipe_name(pipe), port_name(port)); + /* make sure vdd is off before we steal it */ vlv_detach_power_sequencer(intel_dp); } -- cgit v0.10.2 From d288f65fe9472501929eb8d63c2b248813b92056 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 28 Oct 2014 13:20:22 +0200 Subject: drm/i915: Make sure DPLL is enabled when kicking the power sequencer on VLV/CHV MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The power seqeuencer kick procedure requires the DPLL to be running in order to complete successfully. In case the DPLL isn't currently running when we need to kick the power seqeuncer enable it temporarily. This can happen eg. during ->detect() when the pipe is not already active. To avoid needlessly duplicating the DPLL programming re-use the already existing functions by passing a temporary pipe config to them instead of having them consult the current pipe config at crtc->config. v2: Introduce vlv_force_pll_{on,off}() (Daniel) v3: Rebase due to drm_crtc vs. intel_crtc changes Fix a typo in commit msg (checkpatch) Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak (v1) [danvet: Appease checkpatch.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ee982f5..848f330 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -94,8 +94,10 @@ static void intel_cpu_transcoder_set_m_n(struct intel_crtc *crtc, static void ironlake_set_pipeconf(struct drm_crtc *crtc); static void haswell_set_pipeconf(struct drm_crtc *crtc); static void intel_set_pipe_csc(struct drm_crtc *crtc); -static void vlv_prepare_pll(struct intel_crtc *crtc); -static void chv_prepare_pll(struct intel_crtc *crtc); +static void vlv_prepare_pll(struct intel_crtc *crtc, + const struct intel_crtc_config *pipe_config); +static void chv_prepare_pll(struct intel_crtc *crtc, + const struct intel_crtc_config *pipe_config); static struct intel_encoder *intel_find_encoder(struct intel_connector *connector, int pipe) { @@ -1484,12 +1486,13 @@ static void intel_init_dpio(struct drm_device *dev) } } -static void vlv_enable_pll(struct intel_crtc *crtc) +static void vlv_enable_pll(struct intel_crtc *crtc, + const struct intel_crtc_config *pipe_config) { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; int reg = DPLL(crtc->pipe); - u32 dpll = crtc->config.dpll_hw_state.dpll; + u32 dpll = pipe_config->dpll_hw_state.dpll; assert_pipe_disabled(dev_priv, crtc->pipe); @@ -1507,7 +1510,7 @@ static void vlv_enable_pll(struct intel_crtc *crtc) if (wait_for(((I915_READ(reg) & DPLL_LOCK_VLV) == DPLL_LOCK_VLV), 1)) DRM_ERROR("DPLL %d failed to lock\n", crtc->pipe); - I915_WRITE(DPLL_MD(crtc->pipe), crtc->config.dpll_hw_state.dpll_md); + I915_WRITE(DPLL_MD(crtc->pipe), pipe_config->dpll_hw_state.dpll_md); POSTING_READ(DPLL_MD(crtc->pipe)); /* We do this three times for luck */ @@ -1522,7 +1525,8 @@ static void vlv_enable_pll(struct intel_crtc *crtc) udelay(150); /* wait for warmup */ } -static void chv_enable_pll(struct intel_crtc *crtc) +static void chv_enable_pll(struct intel_crtc *crtc, + const struct intel_crtc_config *pipe_config) { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; @@ -1547,14 +1551,14 @@ static void chv_enable_pll(struct intel_crtc *crtc) udelay(1); /* Enable PLL */ - I915_WRITE(DPLL(pipe), crtc->config.dpll_hw_state.dpll); + I915_WRITE(DPLL(pipe), pipe_config->dpll_hw_state.dpll); /* Check PLL is locked */ if (wait_for(((I915_READ(DPLL(pipe)) & DPLL_LOCK_VLV) == DPLL_LOCK_VLV), 1)) DRM_ERROR("PLL %d failed to lock\n", pipe); /* not sure when this should be written */ - I915_WRITE(DPLL_MD(pipe), crtc->config.dpll_hw_state.dpll_md); + I915_WRITE(DPLL_MD(pipe), pipe_config->dpll_hw_state.dpll_md); POSTING_READ(DPLL_MD(pipe)); mutex_unlock(&dev_priv->dpio_lock); @@ -4842,9 +4846,9 @@ static void valleyview_crtc_enable(struct drm_crtc *crtc) if (!is_dsi) { if (IS_CHERRYVIEW(dev)) - chv_prepare_pll(intel_crtc); + chv_prepare_pll(intel_crtc, &intel_crtc->config); else - vlv_prepare_pll(intel_crtc); + vlv_prepare_pll(intel_crtc, &intel_crtc->config); } if (intel_crtc->config.has_dp_encoder) @@ -4864,9 +4868,9 @@ static void valleyview_crtc_enable(struct drm_crtc *crtc) if (!is_dsi) { if (IS_CHERRYVIEW(dev)) - chv_enable_pll(intel_crtc); + chv_enable_pll(intel_crtc, &intel_crtc->config); else - vlv_enable_pll(intel_crtc); + vlv_enable_pll(intel_crtc, &intel_crtc->config); } for_each_encoder_on_crtc(dev, crtc, encoder) @@ -5747,7 +5751,8 @@ void intel_dp_set_m_n(struct intel_crtc *crtc) &crtc->config.dp_m2_n2); } -static void vlv_update_pll(struct intel_crtc *crtc) +static void vlv_update_pll(struct intel_crtc *crtc, + struct intel_crtc_config *pipe_config) { u32 dpll, dpll_md; @@ -5762,14 +5767,15 @@ static void vlv_update_pll(struct intel_crtc *crtc) if (crtc->pipe == PIPE_B) dpll |= DPLL_INTEGRATED_CRI_CLK_VLV; dpll |= DPLL_VCO_ENABLE; - crtc->config.dpll_hw_state.dpll = dpll; + pipe_config->dpll_hw_state.dpll = dpll; - dpll_md = (crtc->config.pixel_multiplier - 1) + dpll_md = (pipe_config->pixel_multiplier - 1) << DPLL_MD_UDI_MULTIPLIER_SHIFT; - crtc->config.dpll_hw_state.dpll_md = dpll_md; + pipe_config->dpll_hw_state.dpll_md = dpll_md; } -static void vlv_prepare_pll(struct intel_crtc *crtc) +static void vlv_prepare_pll(struct intel_crtc *crtc, + const struct intel_crtc_config *pipe_config) { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; @@ -5780,11 +5786,11 @@ static void vlv_prepare_pll(struct intel_crtc *crtc) mutex_lock(&dev_priv->dpio_lock); - bestn = crtc->config.dpll.n; - bestm1 = crtc->config.dpll.m1; - bestm2 = crtc->config.dpll.m2; - bestp1 = crtc->config.dpll.p1; - bestp2 = crtc->config.dpll.p2; + bestn = pipe_config->dpll.n; + bestm1 = pipe_config->dpll.m1; + bestm2 = pipe_config->dpll.m2; + bestp1 = pipe_config->dpll.p1; + bestp2 = pipe_config->dpll.p2; /* See eDP HDMI DPIO driver vbios notes doc */ @@ -5821,7 +5827,7 @@ static void vlv_prepare_pll(struct intel_crtc *crtc) vlv_dpio_write(dev_priv, pipe, VLV_PLL_DW3(pipe), mdiv); /* Set HBR and RBR LPF coefficients */ - if (crtc->config.port_clock == 162000 || + if (pipe_config->port_clock == 162000 || intel_pipe_has_type(crtc, INTEL_OUTPUT_ANALOG) || intel_pipe_has_type(crtc, INTEL_OUTPUT_HDMI)) vlv_dpio_write(dev_priv, pipe, VLV_PLL_DW10(pipe), @@ -5860,19 +5866,21 @@ static void vlv_prepare_pll(struct intel_crtc *crtc) mutex_unlock(&dev_priv->dpio_lock); } -static void chv_update_pll(struct intel_crtc *crtc) +static void chv_update_pll(struct intel_crtc *crtc, + struct intel_crtc_config *pipe_config) { - crtc->config.dpll_hw_state.dpll = DPLL_SSC_REF_CLOCK_CHV | + pipe_config->dpll_hw_state.dpll = DPLL_SSC_REF_CLOCK_CHV | DPLL_REFA_CLK_ENABLE_VLV | DPLL_VGA_MODE_DIS | DPLL_VCO_ENABLE; if (crtc->pipe != PIPE_A) - crtc->config.dpll_hw_state.dpll |= DPLL_INTEGRATED_CRI_CLK_VLV; + pipe_config->dpll_hw_state.dpll |= DPLL_INTEGRATED_CRI_CLK_VLV; - crtc->config.dpll_hw_state.dpll_md = - (crtc->config.pixel_multiplier - 1) << DPLL_MD_UDI_MULTIPLIER_SHIFT; + pipe_config->dpll_hw_state.dpll_md = + (pipe_config->pixel_multiplier - 1) << DPLL_MD_UDI_MULTIPLIER_SHIFT; } -static void chv_prepare_pll(struct intel_crtc *crtc) +static void chv_prepare_pll(struct intel_crtc *crtc, + const struct intel_crtc_config *pipe_config) { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; @@ -5883,18 +5891,18 @@ static void chv_prepare_pll(struct intel_crtc *crtc) u32 bestn, bestm1, bestm2, bestp1, bestp2, bestm2_frac; int refclk; - bestn = crtc->config.dpll.n; - bestm2_frac = crtc->config.dpll.m2 & 0x3fffff; - bestm1 = crtc->config.dpll.m1; - bestm2 = crtc->config.dpll.m2 >> 22; - bestp1 = crtc->config.dpll.p1; - bestp2 = crtc->config.dpll.p2; + bestn = pipe_config->dpll.n; + bestm2_frac = pipe_config->dpll.m2 & 0x3fffff; + bestm1 = pipe_config->dpll.m1; + bestm2 = pipe_config->dpll.m2 >> 22; + bestp1 = pipe_config->dpll.p1; + bestp2 = pipe_config->dpll.p2; /* * Enable Refclk and SSC */ I915_WRITE(dpll_reg, - crtc->config.dpll_hw_state.dpll & ~DPLL_VCO_ENABLE); + pipe_config->dpll_hw_state.dpll & ~DPLL_VCO_ENABLE); mutex_lock(&dev_priv->dpio_lock); @@ -5942,6 +5950,53 @@ static void chv_prepare_pll(struct intel_crtc *crtc) mutex_unlock(&dev_priv->dpio_lock); } +/** + * vlv_force_pll_on - forcibly enable just the PLL + * @dev_priv: i915 private structure + * @pipe: pipe PLL to enable + * @dpll: PLL configuration + * + * Enable the PLL for @pipe using the supplied @dpll config. To be used + * in cases where we need the PLL enabled even when @pipe is not going to + * be enabled. + */ +void vlv_force_pll_on(struct drm_device *dev, enum pipe pipe, + const struct dpll *dpll) +{ + struct intel_crtc *crtc = + to_intel_crtc(intel_get_crtc_for_pipe(dev, pipe)); + struct intel_crtc_config pipe_config = { + .pixel_multiplier = 1, + .dpll = *dpll, + }; + + if (IS_CHERRYVIEW(dev)) { + chv_update_pll(crtc, &pipe_config); + chv_prepare_pll(crtc, &pipe_config); + chv_enable_pll(crtc, &pipe_config); + } else { + vlv_update_pll(crtc, &pipe_config); + vlv_prepare_pll(crtc, &pipe_config); + vlv_enable_pll(crtc, &pipe_config); + } +} + +/** + * vlv_force_pll_off - forcibly disable just the PLL + * @dev_priv: i915 private structure + * @pipe: pipe PLL to disable + * + * Disable the PLL for @pipe. To be used in cases where we need + * the PLL enabled even when @pipe is not going to be enabled. + */ +void vlv_force_pll_off(struct drm_device *dev, enum pipe pipe) +{ + if (IS_CHERRYVIEW(dev)) + chv_disable_pll(to_i915(dev), pipe); + else + vlv_disable_pll(to_i915(dev), pipe); +} + static void i9xx_update_pll(struct intel_crtc *crtc, intel_clock_t *reduced_clock, int num_connectors) @@ -6323,9 +6378,9 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, has_reduced_clock ? &reduced_clock : NULL, num_connectors); } else if (IS_CHERRYVIEW(dev)) { - chv_update_pll(crtc); + chv_update_pll(crtc, &crtc->config); } else if (IS_VALLEYVIEW(dev)) { - vlv_update_pll(crtc); + vlv_update_pll(crtc, &crtc->config); } else { i9xx_update_pll(crtc, has_reduced_clock ? &reduced_clock : NULL, diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 90d752d..ccc690e 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -330,6 +330,7 @@ vlv_power_sequencer_kick(struct intel_dp *intel_dp) struct drm_device *dev = intel_dig_port->base.base.dev; struct drm_i915_private *dev_priv = dev->dev_private; enum pipe pipe = intel_dp->pps_pipe; + bool pll_enabled; uint32_t DP; if (WARN(I915_READ(intel_dp->output_reg) & DP_PORT_EN, @@ -353,6 +354,16 @@ vlv_power_sequencer_kick(struct intel_dp *intel_dp) else if (pipe == PIPE_B) DP |= DP_PIPEB_SELECT; + pll_enabled = I915_READ(DPLL(pipe)) & DPLL_VCO_ENABLE; + + /* + * The DPLL for the pipe must be enabled for this to work. + * So enable temporarily it if it's not already enabled. + */ + if (!pll_enabled) + vlv_force_pll_on(dev, pipe, IS_CHERRYVIEW(dev) ? + &chv_dpll[0].dpll : &vlv_dpll[0].dpll); + /* * Similar magic as in intel_dp_enable_port(). * We _must_ do this port enable + disable trick @@ -367,6 +378,9 @@ vlv_power_sequencer_kick(struct intel_dp *intel_dp) I915_WRITE(intel_dp->output_reg, DP & ~DP_PORT_EN); POSTING_READ(intel_dp->output_reg); + + if (!pll_enabled) + vlv_force_pll_off(dev, pipe); } static enum pipe diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index b323c9b..d53ac23 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -918,6 +918,10 @@ void assert_shared_dpll(struct drm_i915_private *dev_priv, struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc); void intel_put_shared_dpll(struct intel_crtc *crtc); +void vlv_force_pll_on(struct drm_device *dev, enum pipe pipe, + const struct dpll *dpll); +void vlv_force_pll_off(struct drm_device *dev, enum pipe pipe); + /* modesetting asserts */ void assert_panel_unlocked(struct drm_i915_private *dev_priv, enum pipe pipe); -- cgit v0.10.2 From 14af293f060cbf2a508463b01dc4e124bd3c6e15 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 24 Oct 2014 14:51:31 +0100 Subject: drm/i915: create a prepare step for primary planes updates MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Take out the pin_fb code so commit phase can't fail anymore. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 848f330..d678bc0 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11403,20 +11403,16 @@ intel_check_primary_plane(struct drm_plane *plane, } static int -intel_commit_primary_plane(struct drm_plane *plane, - struct intel_plane_state *state) +intel_prepare_primary_plane(struct drm_plane *plane, + struct intel_plane_state *state) { struct drm_crtc *crtc = state->crtc; struct drm_framebuffer *fb = state->fb; struct drm_device *dev = crtc->dev; - struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); enum pipe pipe = intel_crtc->pipe; - struct drm_framebuffer *old_fb = plane->fb; struct drm_i915_gem_object *obj = intel_fb_obj(fb); struct drm_i915_gem_object *old_obj = intel_fb_obj(plane->fb); - struct intel_plane *intel_plane = to_intel_plane(plane); - struct drm_rect *src = &state->src; int ret; intel_crtc_wait_for_pending_flips(crtc); @@ -11426,7 +11422,7 @@ intel_commit_primary_plane(struct drm_plane *plane, return -EBUSY; } - if (plane->fb != fb) { + if (old_obj != obj) { mutex_lock(&dev->struct_mutex); ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); if (ret == 0) @@ -11439,6 +11435,25 @@ intel_commit_primary_plane(struct drm_plane *plane, } } + return 0; +} + +static void +intel_commit_primary_plane(struct drm_plane *plane, + struct intel_plane_state *state) +{ + struct drm_crtc *crtc = state->crtc; + struct drm_framebuffer *fb = state->fb; + struct drm_device *dev = crtc->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + enum pipe pipe = intel_crtc->pipe; + struct drm_framebuffer *old_fb = plane->fb; + struct drm_i915_gem_object *obj = intel_fb_obj(fb); + struct drm_i915_gem_object *old_obj = intel_fb_obj(plane->fb); + struct intel_plane *intel_plane = to_intel_plane(plane); + struct drm_rect *src = &state->src; + crtc->primary->fb = fb; crtc->x = src->x1; crtc->y = src->y1; @@ -11515,8 +11530,6 @@ intel_commit_primary_plane(struct drm_plane *plane, intel_unpin_fb_obj(old_obj); mutex_unlock(&dev->struct_mutex); } - - return 0; } static int @@ -11557,6 +11570,10 @@ intel_primary_plane_setplane(struct drm_plane *plane, struct drm_crtc *crtc, if (ret) return ret; + ret = intel_prepare_primary_plane(plane, &state); + if (ret) + return ret; + intel_commit_primary_plane(plane, &state); return 0; -- cgit v0.10.2 From 34aa50a976c58697ad4050674b272892efea865c Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 24 Oct 2014 14:51:32 +0100 Subject: drm/i915: create a prepare phase for sprite plane updates MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit take out pin_fb code so the commit phase can't fail anymore. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 2c060ad..3631b0e 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -1192,34 +1192,18 @@ intel_check_sprite_plane(struct drm_plane *plane, } static int -intel_commit_sprite_plane(struct drm_plane *plane, - struct intel_plane_state *state) +intel_prepare_sprite_plane(struct drm_plane *plane, + struct intel_plane_state *state) { struct drm_device *dev = plane->dev; struct drm_crtc *crtc = state->crtc; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - struct intel_plane *intel_plane = to_intel_plane(plane); enum pipe pipe = intel_crtc->pipe; struct drm_framebuffer *fb = state->fb; - struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); - struct drm_i915_gem_object *obj = intel_fb->obj; - struct drm_i915_gem_object *old_obj = intel_plane->obj; - int crtc_x, crtc_y; - unsigned int crtc_w, crtc_h; - uint32_t src_x, src_y, src_w, src_h; - struct drm_rect *dst = &state->dst; - const struct drm_rect *clip = &state->clip; - bool primary_enabled; + struct drm_i915_gem_object *obj = intel_fb_obj(fb); + struct drm_i915_gem_object *old_obj = intel_fb_obj(plane->fb); int ret; - /* - * If the sprite is completely covering the primary plane, - * we can disable the primary and save power. - */ - primary_enabled = !drm_rect_equals(dst, clip) || colorkey_enabled(intel_plane); - WARN_ON(!primary_enabled && !state->visible && intel_crtc->active); - - if (old_obj != obj) { mutex_lock(&dev->struct_mutex); @@ -1238,6 +1222,36 @@ intel_commit_sprite_plane(struct drm_plane *plane, return ret; } + return 0; +} + +static void +intel_commit_sprite_plane(struct drm_plane *plane, + struct intel_plane_state *state) +{ + struct drm_device *dev = plane->dev; + struct drm_crtc *crtc = state->crtc; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_plane *intel_plane = to_intel_plane(plane); + enum pipe pipe = intel_crtc->pipe; + struct drm_framebuffer *fb = state->fb; + struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); + struct drm_i915_gem_object *obj = intel_fb->obj; + struct drm_i915_gem_object *old_obj = intel_plane->obj; + int crtc_x, crtc_y; + unsigned int crtc_w, crtc_h; + uint32_t src_x, src_y, src_w, src_h; + struct drm_rect *dst = &state->dst; + const struct drm_rect *clip = &state->clip; + bool primary_enabled; + + /* + * If the sprite is completely covering the primary plane, + * we can disable the primary and save power. + */ + primary_enabled = !drm_rect_equals(dst, clip) || colorkey_enabled(intel_plane); + WARN_ON(!primary_enabled && !state->visible && intel_crtc->active); + intel_plane->crtc_x = state->orig_dst.x1; intel_plane->crtc_y = state->orig_dst.y1; intel_plane->crtc_w = drm_rect_width(&state->orig_dst); @@ -1298,8 +1312,6 @@ intel_commit_sprite_plane(struct drm_plane *plane, intel_unpin_fb_obj(old_obj); mutex_unlock(&dev->struct_mutex); } - - return 0; } static int @@ -1339,7 +1351,12 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, if (ret) return ret; - return intel_commit_sprite_plane(plane, &state); + ret = intel_prepare_sprite_plane(plane, &state); + if (ret) + return ret; + + intel_commit_sprite_plane(plane, &state); + return 0; } static int -- cgit v0.10.2 From 77cde95217484e845743818691df026cec2534f4 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 24 Oct 2014 14:51:33 +0100 Subject: drm/i915: use intel_fb_obj() macros to assign gem objects MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use the macros makes the code cleaner and it also checks for a NULL fb. Signed-off-by: Gustavo Padovan Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 3631b0e..8b80d68 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -1032,8 +1032,7 @@ intel_check_sprite_plane(struct drm_plane *plane, struct intel_crtc *intel_crtc = to_intel_crtc(state->crtc); struct intel_plane *intel_plane = to_intel_plane(plane); struct drm_framebuffer *fb = state->fb; - struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); - struct drm_i915_gem_object *obj = intel_fb->obj; + struct drm_i915_gem_object *obj = intel_fb_obj(fb); int crtc_x, crtc_y; unsigned int crtc_w, crtc_h; uint32_t src_x, src_y, src_w, src_h; @@ -1235,9 +1234,8 @@ intel_commit_sprite_plane(struct drm_plane *plane, struct intel_plane *intel_plane = to_intel_plane(plane); enum pipe pipe = intel_crtc->pipe; struct drm_framebuffer *fb = state->fb; - struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb); - struct drm_i915_gem_object *obj = intel_fb->obj; - struct drm_i915_gem_object *old_obj = intel_plane->obj; + struct drm_i915_gem_object *obj = intel_fb_obj(fb); + struct drm_i915_gem_object *old_obj = intel_fb_obj(plane->fb); int crtc_x, crtc_y; unsigned int crtc_w, crtc_h; uint32_t src_x, src_y, src_w, src_h; -- cgit v0.10.2 From 5d93a6e5a9d225d9c38a63d2e0164ddc4228b354 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 20:52:33 +0300 Subject: drm/i915: Do vlv cmnlane toggle w/a in more cases MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In case the cmnlane power well is down but cmnreset isn't asserted we would currently skip the off+on toggle for the power well. That could leave cmnreset deasserted while cmnlane is powered down which might lead to problems with the PHY. To avoid such issues skip the cmnlane toggle only if both cmnlane and disp2d wells are up and cmnreset is already deasserted. In all other cases power down the cmnlane well which will also make sure cmnreset gets asserted correctly while cmnlane is powered down. Signed-off-by: Ville Syrjälä Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index 39c33e0..59ca54f 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -1137,12 +1137,9 @@ static void vlv_cmnlane_wa(struct drm_i915_private *dev_priv) struct i915_power_well *disp2d = lookup_power_well(dev_priv, PUNIT_POWER_WELL_DISP2D); - /* nothing to do if common lane is already off */ - if (!cmn->ops->is_enabled(dev_priv, cmn)) - return; - /* If the display might be already active skip this */ - if (disp2d->ops->is_enabled(dev_priv, disp2d) && + if (cmn->ops->is_enabled(dev_priv, cmn) && + disp2d->ops->is_enabled(dev_priv, disp2d) && I915_READ(DPIO_CTL) & DPIO_CMNRST) return; -- cgit v0.10.2 From c14b048521ed3413df939ec6cc5dd450ff618120 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Oct 2014 20:52:34 +0300 Subject: drm/i915: Initialize new chv primary plane and pipe blender registers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit CHV adds a bunch of new registers for primary plane size/position and pipe blender setup. Initialize all those registers to avoid nasty surprises. PRIMSIZE is especially important as without programming it the outout will be garbled whenever the primary plane size would not match what the BIOS set up. Also program the sprite constant alpha register to disable the constant alpha blending factor. This applies to vlv as well as chv. Signed-off-by: Ville Syrjälä Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index ea84e1e..a9a8617 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4272,9 +4272,11 @@ enum punit_power_well { #define DISPPLANE_NO_LINE_DOUBLE 0 #define DISPPLANE_STEREO_POLARITY_FIRST 0 #define DISPPLANE_STEREO_POLARITY_SECOND (1<<18) -#define DISPPLANE_ROTATE_180 (1<<15) +#define DISPPLANE_ALPHA_PREMULTIPLY (1<<16) /* CHV pipe B */ +#define DISPPLANE_ROTATE_180 (1<<15) #define DISPPLANE_TRICKLE_FEED_DISABLE (1<<14) /* Ironlake */ #define DISPPLANE_TILED (1<<10) +#define DISPPLANE_MIRROR (1<<8) /* CHV pipe B */ #define _DSPAADDR 0x70184 #define _DSPASTRIDE 0x70188 #define _DSPAPOS 0x7018C /* reserved */ @@ -4295,6 +4297,24 @@ enum punit_power_well { #define DSPOFFSET(plane) _PIPE2(plane, _DSPAOFFSET) #define DSPSURFLIVE(plane) _PIPE2(plane, _DSPASURFLIVE) +/* CHV pipe B blender and primary plane */ +#define _CHV_BLEND_A 0x60a00 +#define CHV_BLEND_LEGACY (0<<30) +#define CHV_BLEND_ANDROID (1<<30) +#define CHV_BLEND_MPO (2<<30) +#define CHV_BLEND_MASK (3<<30) +#define _CHV_CANVAS_A 0x60a04 +#define _PRIMPOS_A 0x60a08 +#define _PRIMSIZE_A 0x60a0c +#define _PRIMCNSTALPHA_A 0x60a10 +#define PRIM_CONST_ALPHA_ENABLE (1<<31) + +#define CHV_BLEND(pipe) _TRANSCODER2(pipe, _CHV_BLEND_A) +#define CHV_CANVAS(pipe) _TRANSCODER2(pipe, _CHV_CANVAS_A) +#define PRIMPOS(plane) _TRANSCODER2(plane, _PRIMPOS_A) +#define PRIMSIZE(plane) _TRANSCODER2(plane, _PRIMSIZE_A) +#define PRIMCNSTALPHA(plane) _TRANSCODER2(plane, _PRIMCNSTALPHA_A) + /* Display/Sprite base address macros */ #define DISP_BASEADDR_MASK (0xfffff000) #define I915_LO_DISPBASE(val) (val & ~DISP_BASEADDR_MASK) @@ -4496,6 +4516,7 @@ enum punit_power_well { #define SP_FORMAT_RGBA1010102 (9<<26) #define SP_FORMAT_RGBX8888 (0xe<<26) #define SP_FORMAT_RGBA8888 (0xf<<26) +#define SP_ALPHA_PREMULTIPLY (1<<23) /* CHV pipe B */ #define SP_SOURCE_KEY (1<<22) #define SP_YUV_BYTE_ORDER_MASK (3<<16) #define SP_YUV_ORDER_YUYV (0<<16) @@ -4504,6 +4525,7 @@ enum punit_power_well { #define SP_YUV_ORDER_VYUY (3<<16) #define SP_ROTATE_180 (1<<15) #define SP_TILED (1<<10) +#define SP_MIRROR (1<<8) /* CHV pipe B */ #define _SPALINOFF (VLV_DISPLAY_BASE + 0x72184) #define _SPASTRIDE (VLV_DISPLAY_BASE + 0x72188) #define _SPAPOS (VLV_DISPLAY_BASE + 0x7218c) @@ -4514,6 +4536,7 @@ enum punit_power_well { #define _SPAKEYMAXVAL (VLV_DISPLAY_BASE + 0x721a0) #define _SPATILEOFF (VLV_DISPLAY_BASE + 0x721a4) #define _SPACONSTALPHA (VLV_DISPLAY_BASE + 0x721a8) +#define SP_CONST_ALPHA_ENABLE (1<<31) #define _SPAGAMC (VLV_DISPLAY_BASE + 0x721f4) #define _SPBCNTR (VLV_DISPLAY_BASE + 0x72280) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index d678bc0..c17ec502 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2451,6 +2451,12 @@ static void i9xx_update_primary_plane(struct drm_crtc *crtc, ((intel_crtc->config.pipe_src_h - 1) << 16) | (intel_crtc->config.pipe_src_w - 1)); I915_WRITE(DSPPOS(plane), 0); + } else if (IS_CHERRYVIEW(dev) && plane == PLANE_B) { + I915_WRITE(PRIMSIZE(plane), + ((intel_crtc->config.pipe_src_h - 1) << 16) | + (intel_crtc->config.pipe_src_w - 1)); + I915_WRITE(PRIMPOS(plane), 0); + I915_WRITE(PRIMCNSTALPHA(plane), 0); } switch (fb->pixel_format) { @@ -4856,6 +4862,13 @@ static void valleyview_crtc_enable(struct drm_crtc *crtc) intel_set_pipe_timings(intel_crtc); + if (IS_CHERRYVIEW(dev) && pipe == PIPE_B) { + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_WRITE(CHV_BLEND(pipe), CHV_BLEND_LEGACY); + I915_WRITE(CHV_CANVAS(pipe), 0); + } + i9xx_set_pipeconf(intel_crtc); intel_crtc->active = true; diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 8b80d68..4912161 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -438,6 +438,8 @@ vlv_update_plane(struct drm_plane *dplane, struct drm_crtc *crtc, else I915_WRITE(SPLINOFF(pipe, plane), linear_offset); + I915_WRITE(SPCONSTALPHA(pipe, plane), 0); + I915_WRITE(SPSIZE(pipe, plane), (crtc_h << 16) | crtc_w); I915_WRITE(SPCNTR(pipe, plane), sprctl); I915_WRITE(SPSURF(pipe, plane), i915_gem_obj_ggtt_offset(obj) + -- cgit v0.10.2 From 1a5df187172a92ffdf8dc89988154d512d66a896 Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Mon, 27 Oct 2014 17:54:32 -0200 Subject: drm/i915: kill intel_resume_prepare() Because, really, the abstraction is not working for us. It is nice for VLV, but doesn't add anything useful on SNB/HSW/BDW. We want to change this code due to a recently-discovered bug, but we can't seem to find a nice solution that repects the current abstraction. So let's kill intel_resume_prepare() and its friends, and add an equivalent implementation to both its callers. Also, look at the diffstat! v2: - Rebase. Signed-off-by: Paulo Zanoni Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 035ec94..33b6fc4 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -551,8 +551,8 @@ static void intel_suspend_encoders(struct drm_i915_private *dev_priv) } static int intel_suspend_complete(struct drm_i915_private *dev_priv); -static int intel_resume_prepare(struct drm_i915_private *dev_priv, - bool rpm_resume); +static int vlv_resume_prepare(struct drm_i915_private *dev_priv, + bool rpm_resume); static int i915_drm_suspend(struct drm_device *dev) { @@ -744,7 +744,7 @@ static int i915_drm_resume(struct drm_device *dev) static int i915_drm_resume_early(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - int ret; + int ret = 0; /* * We have a resume ordering issue with the snd-hda driver also @@ -760,7 +760,10 @@ static int i915_drm_resume_early(struct drm_device *dev) pci_set_master(dev->pdev); - ret = intel_resume_prepare(dev_priv, false); + if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) + hsw_disable_pc8(dev_priv); + else if (IS_VALLEYVIEW(dev_priv)) + ret = vlv_resume_prepare(dev_priv, false); if (ret) DRM_ERROR("Resume prepare failed: %d,Continuing resume\n", ret); @@ -986,25 +989,6 @@ static int hsw_suspend_complete(struct drm_i915_private *dev_priv) return 0; } -static int snb_resume_prepare(struct drm_i915_private *dev_priv, - bool rpm_resume) -{ - struct drm_device *dev = dev_priv->dev; - - if (rpm_resume) - intel_init_pch_refclk(dev); - - return 0; -} - -static int hsw_resume_prepare(struct drm_i915_private *dev_priv, - bool rpm_resume) -{ - hsw_disable_pc8(dev_priv); - - return 0; -} - /* * Save all Gunit registers that may be lost after a D3 and a subsequent * S0i[R123] transition. The list of registers needing a save/restore is @@ -1462,7 +1446,7 @@ static int intel_runtime_resume(struct device *device) struct pci_dev *pdev = to_pci_dev(device); struct drm_device *dev = pci_get_drvdata(pdev); struct drm_i915_private *dev_priv = dev->dev_private; - int ret; + int ret = 0; if (WARN_ON_ONCE(!HAS_RUNTIME_PM(dev))) return -ENODEV; @@ -1472,7 +1456,13 @@ static int intel_runtime_resume(struct device *device) intel_opregion_notify_adapter(dev, PCI_D0); dev_priv->pm.suspended = false; - ret = intel_resume_prepare(dev_priv, true); + if (IS_GEN6(dev_priv)) + intel_init_pch_refclk(dev); + else if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) + hsw_disable_pc8(dev_priv); + else if (IS_VALLEYVIEW(dev_priv)) + ret = vlv_resume_prepare(dev_priv, true); + /* * No point of rolling back things in case of an error, as the best * we can do is to hope that things will still work (and disable RPM). @@ -1510,29 +1500,6 @@ static int intel_suspend_complete(struct drm_i915_private *dev_priv) return ret; } -/* - * This function implements common functionality of runtime and system - * resume sequence. Variable rpm_resume used for implementing different - * code paths. - */ -static int intel_resume_prepare(struct drm_i915_private *dev_priv, - bool rpm_resume) -{ - struct drm_device *dev = dev_priv->dev; - int ret; - - if (IS_GEN6(dev)) - ret = snb_resume_prepare(dev_priv, rpm_resume); - else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) - ret = hsw_resume_prepare(dev_priv, rpm_resume); - else if (IS_VALLEYVIEW(dev)) - ret = vlv_resume_prepare(dev_priv, rpm_resume); - else - ret = 0; - - return ret; -} - static const struct dev_pm_ops i915_pm_ops = { /* * S0ix (via system suspend) and S3 event handlers [PMSG_SUSPEND, -- cgit v0.10.2 From efee833a2d0a69dfa2c4ec28377b00803a0f9a60 Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Mon, 27 Oct 2014 17:54:33 -0200 Subject: drm/i915: run hsw_disable_pc8() later on resume We want to run intel_uncore_early_sanitize() before we touch any registers, because on BDW, when we resume, the FPGA_DBG_RM_NOCLAIM bit is set, so we need to clear it - through intel_uncore_early_sanitize() - before we do anything else. With the current code, we don't clear the bit before our first register access, so we print a WARN complaining about an unclaimed register error. v1: Was called "drm/i915: run intel_uncore_early_sanitize earlier on resume" v2: Was called "drm/i915: run intel_uncore_early_sanitize earlier on resume on non-VLV" v3: This one, on top of the intel_resume_prepare() rework. v4: Rebase. Signed-off-by: Paulo Zanoni Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 33b6fc4..0c7cf48 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -760,14 +760,16 @@ static int i915_drm_resume_early(struct drm_device *dev) pci_set_master(dev->pdev); - if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) - hsw_disable_pc8(dev_priv); - else if (IS_VALLEYVIEW(dev_priv)) + if (IS_VALLEYVIEW(dev_priv)) ret = vlv_resume_prepare(dev_priv, false); if (ret) DRM_ERROR("Resume prepare failed: %d,Continuing resume\n", ret); intel_uncore_early_sanitize(dev, true); + + if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) + hsw_disable_pc8(dev_priv); + intel_uncore_sanitize(dev); intel_power_domains_init_hw(dev_priv); -- cgit v0.10.2 From 6ca2aeb27b4a8451052baa61c11c33bd5de8a71c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 20 Oct 2014 19:47:53 +0300 Subject: drm/i915: Add support for CHV pipe B sprite CSC MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit CHV has a programmable CSC unit on the pipe B sprites. Program the unit appropriately for BT.601 limited range YCbCr to full range RGB color conversion. This matches the programming we currently do for sprites on the other pipes and on other platforms. It seems the CSC only works when the input data is YCbCr. For RGB pixel formats it doesn't matter what we program into the CSC registers. Doesn't make much sense to me especially since the register names give the impression that RGB input data would also work. But that's how it behaves here. In the review discussions there's been some nice math to explain the values obtained here. First about the YCbCr->RGB matrix: "I had the RGB->YCbCr matrix, inverted it and the values came out. But they should match the wikipedia article. Also keep in mind that the coefficients are in .12 in fixed point format, hence we need a 1<<12 factor. So let's try it: Kb=.114 Kr=.299 (1<<12) * 255/219 ~= 4769 -(1<<12) * 255/112*(1-Kb)*Kb/(1-Kb-Kr) ~= -1605 -(1<<12) * 255/112*(1-Kr)*Kr/(1-Kb-Kr) ~= -3330 (1<<12) * 255/112*(1-Kr) ~= 6537 (1<<12) * 255/112*(1-Kb) ~= 8263 "Looks like the same values to me." And then about the limits used for clamping: "> where did you get these min/max? "The hardware apparently deals in 10bit values, so we need to multiply everything by 4 when we start with the 8bit min/max values. Y = [16:235] * 4 = [64:940] CbCr = ([16:240] - 128) * 4 = [-112:112] * 4 = [-448:448] "The -128 being the -0.5 bias that the hardware already applied before the data entered the CSC unit." Raw data is also supplied in 10bpc in the registers. Signed-off-by: Ville Syrjälä Reviewed-by Rodrigo Vivi [danvet: Copypaste explanations&math from the review discussion.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index a9a8617..2a4bdd5 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4565,6 +4565,39 @@ enum punit_power_well { #define SPCONSTALPHA(pipe, plane) _PIPE(pipe * 2 + plane, _SPACONSTALPHA, _SPBCONSTALPHA) #define SPGAMC(pipe, plane) _PIPE(pipe * 2 + plane, _SPAGAMC, _SPBGAMC) +/* + * CHV pipe B sprite CSC + * + * |cr| |c0 c1 c2| |cr + cr_ioff| |cr_ooff| + * |yg| = |c3 c4 c5| x |yg + yg_ioff| + |yg_ooff| + * |cb| |c6 c7 c8| |cb + cr_ioff| |cb_ooff| + */ +#define SPCSCYGOFF(sprite) (VLV_DISPLAY_BASE + 0x6d900 + (sprite) * 0x1000) +#define SPCSCCBOFF(sprite) (VLV_DISPLAY_BASE + 0x6d904 + (sprite) * 0x1000) +#define SPCSCCROFF(sprite) (VLV_DISPLAY_BASE + 0x6d908 + (sprite) * 0x1000) +#define SPCSC_OOFF(x) (((x) & 0x7ff) << 16) /* s11 */ +#define SPCSC_IOFF(x) (((x) & 0x7ff) << 0) /* s11 */ + +#define SPCSCC01(sprite) (VLV_DISPLAY_BASE + 0x6d90c + (sprite) * 0x1000) +#define SPCSCC23(sprite) (VLV_DISPLAY_BASE + 0x6d910 + (sprite) * 0x1000) +#define SPCSCC45(sprite) (VLV_DISPLAY_BASE + 0x6d914 + (sprite) * 0x1000) +#define SPCSCC67(sprite) (VLV_DISPLAY_BASE + 0x6d918 + (sprite) * 0x1000) +#define SPCSCC8(sprite) (VLV_DISPLAY_BASE + 0x6d91c + (sprite) * 0x1000) +#define SPCSC_C1(x) (((x) & 0x7fff) << 16) /* s3.12 */ +#define SPCSC_C0(x) (((x) & 0x7fff) << 0) /* s3.12 */ + +#define SPCSCYGICLAMP(sprite) (VLV_DISPLAY_BASE + 0x6d920 + (sprite) * 0x1000) +#define SPCSCCBICLAMP(sprite) (VLV_DISPLAY_BASE + 0x6d924 + (sprite) * 0x1000) +#define SPCSCCRICLAMP(sprite) (VLV_DISPLAY_BASE + 0x6d928 + (sprite) * 0x1000) +#define SPCSC_IMAX(x) (((x) & 0x7ff) << 16) /* s11 */ +#define SPCSC_IMIN(x) (((x) & 0x7ff) << 0) /* s11 */ + +#define SPCSCYGOCLAMP(sprite) (VLV_DISPLAY_BASE + 0x6d92c + (sprite) * 0x1000) +#define SPCSCCBOCLAMP(sprite) (VLV_DISPLAY_BASE + 0x6d930 + (sprite) * 0x1000) +#define SPCSCCROCLAMP(sprite) (VLV_DISPLAY_BASE + 0x6d934 + (sprite) * 0x1000) +#define SPCSC_OMAX(x) ((x) << 16) /* u10 */ +#define SPCSC_OMIN(x) ((x) << 0) /* u10 */ + /* Skylake plane registers */ #define _PLANE_CTL_1_A 0x70180 diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 4912161..bf1eb15 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -37,6 +37,20 @@ #include #include "i915_drv.h" +static bool +format_is_yuv(uint32_t format) +{ + switch (format) { + case DRM_FORMAT_YUYV: + case DRM_FORMAT_UYVY: + case DRM_FORMAT_VYUY: + case DRM_FORMAT_YVYU: + return true; + default: + return false; + } +} + static int usecs_to_scanlines(const struct drm_display_mode *mode, int usecs) { /* paranoia */ @@ -320,6 +334,45 @@ skl_get_colorkey(struct drm_plane *drm_plane, } static void +chv_update_csc(struct intel_plane *intel_plane, uint32_t format) +{ + struct drm_i915_private *dev_priv = intel_plane->base.dev->dev_private; + int plane = intel_plane->plane; + + /* Seems RGB data bypasses the CSC always */ + if (!format_is_yuv(format)) + return; + + /* + * BT.601 limited range YCbCr -> full range RGB + * + * |r| | 6537 4769 0| |cr | + * |g| = |-3330 4769 -1605| x |y-64| + * |b| | 0 4769 8263| |cb | + * + * Cb and Cr apparently come in as signed already, so no + * need for any offset. For Y we need to remove the offset. + */ + I915_WRITE(SPCSCYGOFF(plane), SPCSC_OOFF(0) | SPCSC_IOFF(-64)); + I915_WRITE(SPCSCCBOFF(plane), SPCSC_OOFF(0) | SPCSC_IOFF(0)); + I915_WRITE(SPCSCCROFF(plane), SPCSC_OOFF(0) | SPCSC_IOFF(0)); + + I915_WRITE(SPCSCC01(plane), SPCSC_C1(4769) | SPCSC_C0(6537)); + I915_WRITE(SPCSCC23(plane), SPCSC_C1(-3330) | SPCSC_C0(0)); + I915_WRITE(SPCSCC45(plane), SPCSC_C1(-1605) | SPCSC_C0(4769)); + I915_WRITE(SPCSCC67(plane), SPCSC_C1(4769) | SPCSC_C0(0)); + I915_WRITE(SPCSCC8(plane), SPCSC_C0(8263)); + + I915_WRITE(SPCSCYGICLAMP(plane), SPCSC_IMAX(940) | SPCSC_IMIN(64)); + I915_WRITE(SPCSCCBICLAMP(plane), SPCSC_IMAX(448) | SPCSC_IMIN(-448)); + I915_WRITE(SPCSCCRICLAMP(plane), SPCSC_IMAX(448) | SPCSC_IMIN(-448)); + + I915_WRITE(SPCSCYGOCLAMP(plane), SPCSC_OMAX(1023) | SPCSC_OMIN(0)); + I915_WRITE(SPCSCCBOCLAMP(plane), SPCSC_OMAX(1023) | SPCSC_OMIN(0)); + I915_WRITE(SPCSCCROCLAMP(plane), SPCSC_OMAX(1023) | SPCSC_OMIN(0)); +} + +static void vlv_update_plane(struct drm_plane *dplane, struct drm_crtc *crtc, struct drm_framebuffer *fb, struct drm_i915_gem_object *obj, int crtc_x, int crtc_y, @@ -430,6 +483,9 @@ vlv_update_plane(struct drm_plane *dplane, struct drm_crtc *crtc, intel_update_primary_plane(intel_crtc); + if (IS_CHERRYVIEW(dev) && pipe == PIPE_B) + chv_update_csc(intel_plane, fb->pixel_format); + I915_WRITE(SPSTRIDE(pipe, plane), fb->pitches[0]); I915_WRITE(SPPOS(pipe, plane), (crtc_y << 16) | crtc_x); @@ -1004,20 +1060,6 @@ ilk_get_colorkey(struct drm_plane *plane, struct drm_intel_sprite_colorkey *key) key->flags = I915_SET_COLORKEY_NONE; } -static bool -format_is_yuv(uint32_t format) -{ - switch (format) { - case DRM_FORMAT_YUYV: - case DRM_FORMAT_UYVY: - case DRM_FORMAT_VYUY: - case DRM_FORMAT_YVYU: - return true; - default: - return false; - } -} - static bool colorkey_enabled(struct intel_plane *intel_plane) { struct drm_intel_sprite_colorkey key; -- cgit v0.10.2 From baa4e575d6a18dcd6f2e622784aa16ab24024f09 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 27 Oct 2014 16:07:32 +0200 Subject: drm/i915: Enable pipe-a power well on chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It seems that the pipe-a power well has replaced the disp2d power well on chv. At least that's the case with the current punit firmware. So enable the pipe-a power and expand its domains to cover everything the disp2d well ought to cover. The other power wells (apart from the cmnlane wells) still seem awol in the current punit firmware. So leave them disabled in the code. This fixes a hilarious oops during resume on bsw where intel_hdmi_get_config() would read the port register and get back 0xffffffff and thus think the port is enabled on pipe D. It would then go and index the pipe_to_crtc_mapping[] array with PIPE_D and blow up when intel_hdmi_get_config() tries to write to crtc->config. Someone really ought to replace all naked pipe_to_crtc_mapping[] uses with the appropriate function call so we could add a warning there if the pipe doesn't actually exist... We must also call the power seqeuencer state reset function from the pipe-a well disable just like we do from disp2d on vlv. Otherwise the eDP panel won't recover at resume time since the PPS has lost its hold on the port. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84903 Signed-off-by: Ville Syrjälä Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index 59ca54f..dcbecff 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -587,6 +587,9 @@ static void chv_pipe_power_well_disable(struct drm_i915_private *dev_priv, power_well->data != PIPE_C); chv_set_pipe_power_well(dev_priv, power_well, false); + + if (power_well->data == PIPE_A) + vlv_power_sequencer_reset(dev_priv); } static void check_power_well_state(struct drm_i915_private *dev_priv, @@ -938,12 +941,20 @@ static struct i915_power_well chv_power_wells[] = { .data = PUNIT_POWER_WELL_DISP2D, .ops = &vlv_display_power_well_ops, }, +#endif { .name = "pipe-a", - .domains = CHV_PIPE_A_POWER_DOMAINS, + /* + * FIXME: pipe A power well seems to be the new disp2d well. + * At least all registers seem to be housed there. Figure + * out if this a a temporary situation in pre-production + * hardware or a permanent state of affairs. + */ + .domains = CHV_PIPE_A_POWER_DOMAINS | VLV_DISPLAY_POWER_DOMAINS, .data = PIPE_A, .ops = &chv_pipe_power_well_ops, }, +#if 0 { .name = "pipe-b", .domains = CHV_PIPE_B_POWER_DOMAINS, -- cgit v0.10.2 From 6189b0369c67f5bc6687ee957b3c1317dbb78922 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Tue, 28 Oct 2014 13:53:01 +0200 Subject: drm/i915/audio: set ELD Conn_Type at one place Keep the driver modifications to ELD together. This also sets the Conn_Type for G4X DP which wasn't done before. Clean up the debugs while at it; this is all obvious from the connector name. v3: add missing ~ (Rodrigo) Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 4a384d7..537f6d8 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -181,13 +181,10 @@ static void haswell_write_eld(struct drm_connector *connector, eldv = AUDIO_ELD_VALID_A << (pipe * 4); - if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { - DRM_DEBUG_DRIVER("ELD: DisplayPort detected\n"); - eld[5] |= (1 << 2); /* Conn_Type, 0x1 = DisplayPort */ + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ - } else { + else I915_WRITE(aud_config, audio_config_hdmi_pixel_clock(mode)); - } if (intel_eld_uptodate(connector, aud_cntrl_st2, eldv, @@ -276,13 +273,10 @@ static void ironlake_write_eld(struct drm_connector *connector, eldv = IBX_ELD_VALIDB << ((port - 1) * 4); } - if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) { - DRM_DEBUG_DRIVER("ELD: DisplayPort detected\n"); - eld[5] |= (1 << 2); /* Conn_Type, 0x1 = DisplayPort */ + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ - } else { + else I915_WRITE(aud_config, audio_config_hdmi_pixel_clock(mode)); - } if (intel_eld_uptodate(connector, aud_cntrl_st2, eldv, @@ -330,6 +324,11 @@ void intel_write_eld(struct intel_encoder *intel_encoder) connector->encoder->base.id, connector->encoder->name); + /* ELD Conn_Type */ + connector->eld[5] &= ~(3 << 2); + if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) + connector->eld[5] |= (1 << 2); + connector->eld[6] = drm_av_sync_delay(connector, mode) / 2; if (dev_priv->display.write_eld) -- cgit v0.10.2 From 2aa0de39fa071129fdd952a0d99828db8434d88f Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:49 +0200 Subject: drm/i915/ddi: write ELD where it's supposed to be done The audio programming sequence states that the ELD must be written and enabled after the pipe is ready. Indeed, this should clarify the situation with commit c79057922ed6c2c6df1214e6ab4414fea1b23db2 Author: Daniel Vetter Date: Wed Apr 16 16:56:09 2014 +0200 drm/i915: Remove vblank wait from haswell_write_eld and Ville's review of it [1]. Moreover, we should not touch the relevant registers before we get the audio power domain. [1] http://mid.gmane.org/20140416155309.GK18465@intel.com Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 537f6d8..158ad60 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -160,8 +160,6 @@ static void haswell_write_eld(struct drm_connector *connector, I915_WRITE(aud_cntrl_st2, tmp); POSTING_READ(aud_cntrl_st2); - assert_pipe_disabled(dev_priv, pipe); - /* Set ELD valid state */ tmp = I915_READ(aud_cntrl_st2); DRM_DEBUG_DRIVER("HDMI audio: pin eld vld status=0x%08x\n", tmp); diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 2688bc9..56e7cb1 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1120,15 +1120,6 @@ static void intel_ddi_pre_enable(struct intel_encoder *intel_encoder) enum port port = intel_ddi_get_encoder_port(intel_encoder); int type = intel_encoder->type; - if (crtc->config.has_audio) { - DRM_DEBUG_DRIVER("Audio on pipe %c on DDI\n", - pipe_name(crtc->pipe)); - - /* write eld */ - DRM_DEBUG_DRIVER("DDI audio: write eld information\n"); - intel_write_eld(intel_encoder); - } - if (type == INTEL_OUTPUT_EDP) { struct intel_dp *intel_dp = enc_to_intel_dp(encoder); intel_edp_panel_on(intel_dp); @@ -1225,6 +1216,8 @@ static void intel_enable_ddi(struct intel_encoder *intel_encoder) if (intel_crtc->config.has_audio) { intel_display_power_get(dev_priv, POWER_DOMAIN_AUDIO); + intel_write_eld(intel_encoder); + tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); tmp |= ((AUDIO_OUTPUT_ENABLE_A | AUDIO_ELD_VALID_A) << (pipe * 4)); I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); -- cgit v0.10.2 From 69bfe1a9b4dffca482c6cd7f1e218f24a1128dcc Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:50 +0200 Subject: drm/i915: introduce intel_audio_codec_{enable, disable} Introduce functions to enable/disable the audio codec, incorporating the ELD setup within enable. The disable is initially limited to HSW, covering exactly what was done previously. The only functional difference is that ELD valid is no longer set if there is no connector with ELD, which should be the right thing to do anyway. Otherwise the sequence remains the same, with warts and all, in preparation for applying more sanity. v2: add kernel doc. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 6a73803..0344fd5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -483,9 +483,10 @@ struct drm_i915_display_funcs { void (*crtc_enable)(struct drm_crtc *crtc); void (*crtc_disable)(struct drm_crtc *crtc); void (*off)(struct drm_crtc *crtc); - void (*write_eld)(struct drm_connector *connector, - struct intel_encoder *encoder, - struct drm_display_mode *mode); + void (*audio_codec_enable)(struct drm_connector *connector, + struct intel_encoder *encoder, + struct drm_display_mode *mode); + void (*audio_codec_disable)(struct intel_encoder *encoder); void (*fdi_link_train)(struct drm_crtc *crtc); void (*init_clock_gating)(struct drm_device *dev); int (*queue_flip)(struct drm_device *dev, struct drm_crtc *crtc, diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 158ad60..f291efc 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -96,9 +96,9 @@ static bool intel_eld_uptodate(struct drm_connector *connector, return true; } -static void g4x_write_eld(struct drm_connector *connector, - struct intel_encoder *encoder, - struct drm_display_mode *mode) +static void g4x_audio_codec_enable(struct drm_connector *connector, + struct intel_encoder *encoder, + struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = connector->dev->dev_private; uint8_t *eld = connector->eld; @@ -136,9 +136,22 @@ static void g4x_write_eld(struct drm_connector *connector, I915_WRITE(G4X_AUD_CNTL_ST, tmp); } -static void haswell_write_eld(struct drm_connector *connector, - struct intel_encoder *encoder, - struct drm_display_mode *mode) +static void hsw_audio_codec_disable(struct intel_encoder *encoder) +{ + struct drm_device *dev = encoder->base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_crtc *crtc = encoder->base.crtc; + enum pipe pipe = to_intel_crtc(crtc)->pipe; + uint32_t tmp; + + tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); + tmp &= ~((AUDIO_OUTPUT_ENABLE_A | AUDIO_ELD_VALID_A) << (pipe * 4)); + I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); +} + +static void hsw_audio_codec_enable(struct drm_connector *connector, + struct intel_encoder *encoder, + struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = connector->dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); @@ -211,11 +224,16 @@ static void haswell_write_eld(struct drm_connector *connector, tmp = I915_READ(aud_cntrl_st2); tmp |= eldv; I915_WRITE(aud_cntrl_st2, tmp); + + /* XXX: Transitional */ + tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); + tmp |= ((AUDIO_OUTPUT_ENABLE_A | AUDIO_ELD_VALID_A) << (pipe * 4)); + I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); } -static void ironlake_write_eld(struct drm_connector *connector, - struct intel_encoder *encoder, - struct drm_display_mode *mode) +static void ilk_audio_codec_enable(struct drm_connector *connector, + struct intel_encoder *encoder, + struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = connector->dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); @@ -303,7 +321,14 @@ static void ironlake_write_eld(struct drm_connector *connector, I915_WRITE(aud_cntrl_st2, tmp); } -void intel_write_eld(struct intel_encoder *intel_encoder) +/** + * intel_audio_codec_enable - Enable the audio codec for HD audio + * @intel_encoder: encoder on which to enable audio + * + * The enable sequences may only be performed after enabling the transcoder and + * port, and after completed link training. + */ +void intel_audio_codec_enable(struct intel_encoder *intel_encoder) { struct drm_encoder *encoder = &intel_encoder->base; struct intel_crtc *crtc = to_intel_crtc(encoder->crtc); @@ -329,8 +354,24 @@ void intel_write_eld(struct intel_encoder *intel_encoder) connector->eld[6] = drm_av_sync_delay(connector, mode) / 2; - if (dev_priv->display.write_eld) - dev_priv->display.write_eld(connector, intel_encoder, mode); + if (dev_priv->display.audio_codec_enable) + dev_priv->display.audio_codec_enable(connector, intel_encoder, mode); +} + +/** + * intel_audio_codec_disable - Disable the audio codec for HD audio + * @encoder: encoder on which to disable audio + * + * The disable sequences must be performed before disabling the transcoder or + * port. + */ +void intel_audio_codec_disable(struct intel_encoder *encoder) +{ + struct drm_device *dev = encoder->base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + + if (dev_priv->display.audio_codec_disable) + dev_priv->display.audio_codec_disable(encoder); } /** @@ -341,12 +382,14 @@ void intel_init_audio(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - if (IS_G4X(dev)) - dev_priv->display.write_eld = g4x_write_eld; - else if (IS_VALLEYVIEW(dev)) - dev_priv->display.write_eld = ironlake_write_eld; - else if (IS_HASWELL(dev) || INTEL_INFO(dev)->gen >= 8) - dev_priv->display.write_eld = haswell_write_eld; - else if (HAS_PCH_SPLIT(dev)) - dev_priv->display.write_eld = ironlake_write_eld; + if (IS_G4X(dev)) { + dev_priv->display.audio_codec_enable = g4x_audio_codec_enable; + } else if (IS_VALLEYVIEW(dev)) { + dev_priv->display.audio_codec_enable = ilk_audio_codec_enable; + } else if (IS_HASWELL(dev) || INTEL_INFO(dev)->gen >= 8) { + dev_priv->display.audio_codec_enable = hsw_audio_codec_enable; + dev_priv->display.audio_codec_disable = hsw_audio_codec_disable; + } else if (HAS_PCH_SPLIT(dev)) { + dev_priv->display.audio_codec_enable = ilk_audio_codec_enable; + } } diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 56e7cb1..b182b9b 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1186,12 +1186,10 @@ static void intel_enable_ddi(struct intel_encoder *intel_encoder) struct drm_encoder *encoder = &intel_encoder->base; struct drm_crtc *crtc = encoder->crtc; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; struct drm_device *dev = encoder->dev; struct drm_i915_private *dev_priv = dev->dev_private; enum port port = intel_ddi_get_encoder_port(intel_encoder); int type = intel_encoder->type; - uint32_t tmp; if (type == INTEL_OUTPUT_HDMI) { struct intel_digital_port *intel_dig_port = @@ -1216,11 +1214,7 @@ static void intel_enable_ddi(struct intel_encoder *intel_encoder) if (intel_crtc->config.has_audio) { intel_display_power_get(dev_priv, POWER_DOMAIN_AUDIO); - intel_write_eld(intel_encoder); - - tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); - tmp |= ((AUDIO_OUTPUT_ENABLE_A | AUDIO_ELD_VALID_A) << (pipe * 4)); - I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); + intel_audio_codec_enable(intel_encoder); } } @@ -1229,19 +1223,12 @@ static void intel_disable_ddi(struct intel_encoder *intel_encoder) struct drm_encoder *encoder = &intel_encoder->base; struct drm_crtc *crtc = encoder->crtc; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; int type = intel_encoder->type; struct drm_device *dev = encoder->dev; struct drm_i915_private *dev_priv = dev->dev_private; - uint32_t tmp; - /* We can't touch HSW_AUD_PIN_ELD_CP_VLD uncionditionally because this - * register is part of the power well on Haswell. */ if (intel_crtc->config.has_audio) { - tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); - tmp &= ~((AUDIO_OUTPUT_ENABLE_A | AUDIO_ELD_VALID_A) << - (pipe * 4)); - I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); + intel_audio_codec_disable(intel_encoder); intel_display_power_put(dev_priv, POWER_DOMAIN_AUDIO); } diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index ccc690e..f1a5fe2 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1328,7 +1328,7 @@ static void intel_dp_prepare(struct intel_encoder *encoder) DRM_DEBUG_DRIVER("Enabling DP audio on pipe %c\n", pipe_name(crtc->pipe)); intel_dp->DP |= DP_AUDIO_OUTPUT_ENABLE; - intel_write_eld(encoder); + intel_audio_codec_enable(encoder); } /* Split out the IBX/CPU vs CPT settings */ diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index d53ac23..134ab08 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -853,7 +853,8 @@ void intel_fb_obj_flush(struct drm_i915_gem_object *obj, bool retire); /* intel_audio.c */ void intel_init_audio(struct drm_device *dev); -void intel_write_eld(struct intel_encoder *encoder); +void intel_audio_codec_enable(struct intel_encoder *encoder); +void intel_audio_codec_disable(struct intel_encoder *encoder); /* intel_display.c */ const char *intel_output_name(int output); diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 07b5ebd..f29026a 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -666,7 +666,7 @@ static void intel_hdmi_prepare(struct intel_encoder *encoder) DRM_DEBUG_DRIVER("Enabling HDMI audio on pipe %c\n", pipe_name(crtc->pipe)); hdmi_val |= SDVO_AUDIO_ENABLE; - intel_write_eld(encoder); + intel_audio_codec_enable(encoder); } if (HAS_PCH_CPT(dev)) -- cgit v0.10.2 From d806fbf5d38f23fc1b259fee181b749665d72735 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:51 +0200 Subject: drm/i915/audio: remove misleading checks for !eld[0] We'll never end up in the hooks with eld[0] unset, as that's checked by drm_select_eld(). Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index f291efc..7580b38 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -79,9 +79,6 @@ static bool intel_eld_uptodate(struct drm_connector *connector, tmp = I915_READ(reg_eldv); tmp &= bits_eldv; - if (!eld[0]) - return !tmp; - if (!tmp) return false; @@ -123,9 +120,6 @@ static void g4x_audio_codec_enable(struct drm_connector *connector, len = (tmp >> 9) & 0x1f; /* ELD buffer size */ I915_WRITE(G4X_AUD_CNTL_ST, tmp); - if (!eld[0]) - return; - len = min_t(int, eld[2], len); DRM_DEBUG_DRIVER("ELD size %d\n", len); for (i = 0; i < len; i++) @@ -207,9 +201,6 @@ static void hsw_audio_codec_enable(struct drm_connector *connector, tmp &= ~eldv; I915_WRITE(aud_cntrl_st2, tmp); - if (!eld[0]) - return; - tmp = I915_READ(aud_cntl_st); tmp &= ~IBX_ELD_ADDRESS; I915_WRITE(aud_cntl_st, tmp); @@ -304,9 +295,6 @@ static void ilk_audio_codec_enable(struct drm_connector *connector, tmp &= ~eldv; I915_WRITE(aud_cntrl_st2, tmp); - if (!eld[0]) - return; - tmp = I915_READ(aud_cntl_st); tmp &= ~IBX_ELD_ADDRESS; I915_WRITE(aud_cntl_st, tmp); -- cgit v0.10.2 From 1e74a324465e5a4f8e3ee7c423aef92c54de8a72 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 28 Oct 2014 16:15:51 +0200 Subject: drm/i915: Initialize PPS timestamps on vlv/chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The pps timestamp initialization was accidentally lost on vlv/chv in commit a4a5d2f8a96e09844a91469e889f15bd5e927399 Author: Ville Syrjälä Date: Thu Sep 4 14:54:20 2014 +0300 drm/i915: Track which port is using which pipe's power sequencer Restore it so that we avoid introducing random delays into the pps operations during/after driver init time. Cc: Imre Deak Reported-by: Imre Deak Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index f1a5fe2..45c0a66 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -5364,12 +5364,11 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, if (is_edp(intel_dp)) { pps_lock(intel_dp); - if (IS_VALLEYVIEW(dev)) { + intel_dp_init_panel_power_timestamps(intel_dp); + if (IS_VALLEYVIEW(dev)) vlv_initial_power_sequencer_setup(intel_dp); - } else { - intel_dp_init_panel_power_timestamps(intel_dp); + else intel_dp_init_panel_power_sequencer(dev, intel_dp); - } pps_unlock(intel_dp); } -- cgit v0.10.2 From 49e6bc51bc9e22c8a433ba32a4e45a5818de3850 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 28 Oct 2014 16:15:52 +0200 Subject: drm/i915: Read out the power sequencer port assignment on resume on vlv/chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When we suspend we turn everything off so the pps should be idle, and we also (or at least should) disable all power wells which will reset the power sequencer port assignment. So when we resume all power sequencers should be in their reset state. However it's at least theoretically possible that the BIOS would touch the power seuqencer(s), so to be safe we ought to read out the current port assignment like we do at driver init time. To do that we can simply call vlv_initial_power_sequencer_setup() from the encoder ->reset() hook before calling intel_edp_panel_vdd_sanitize(). There's no danger or clobbering the pps delays since we now have those stored within intel_dp and we don't change them once initialized. This will make sure that the vdd state gets correctly tracked post-resume in case the BIOS enabled it. We need to shuffle things around a bit to get the locking right, and while at it, make intel_edp_panel_vdd_sanitize() static and move it around a bit to avoid a forward declaration. Cc: Imre Deak Signed-off-by: Ville Syrjälä Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 45c0a66..ec80e82 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -4677,9 +4677,52 @@ static void intel_dp_encoder_suspend(struct intel_encoder *intel_encoder) pps_unlock(intel_dp); } +static void intel_edp_panel_vdd_sanitize(struct intel_dp *intel_dp) +{ + struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = intel_dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + enum intel_display_power_domain power_domain; + + lockdep_assert_held(&dev_priv->pps_mutex); + + if (!edp_have_panel_vdd(intel_dp)) + return; + + /* + * The VDD bit needs a power domain reference, so if the bit is + * already enabled when we boot or resume, grab this reference and + * schedule a vdd off, so we don't hold on to the reference + * indefinitely. + */ + DRM_DEBUG_KMS("VDD left on by BIOS, adjusting state tracking\n"); + power_domain = intel_display_port_power_domain(&intel_dig_port->base); + intel_display_power_get(dev_priv, power_domain); + + edp_panel_vdd_schedule_off(intel_dp); +} + static void intel_dp_encoder_reset(struct drm_encoder *encoder) { - intel_edp_panel_vdd_sanitize(to_intel_encoder(encoder)); + struct intel_dp *intel_dp; + + if (to_intel_encoder(encoder)->type != INTEL_OUTPUT_EDP) + return; + + intel_dp = enc_to_intel_dp(encoder); + + pps_lock(intel_dp); + + /* + * Read out the current power sequencer assignment, + * in case the BIOS did something with it. + */ + if (IS_VALLEYVIEW(encoder->dev)) + vlv_initial_power_sequencer_setup(intel_dp); + + intel_edp_panel_vdd_sanitize(intel_dp); + + pps_unlock(intel_dp); } static const struct drm_connector_funcs intel_dp_connector_funcs = { @@ -5150,37 +5193,6 @@ intel_dp_drrs_init(struct intel_digital_port *intel_dig_port, return downclock_mode; } -void intel_edp_panel_vdd_sanitize(struct intel_encoder *intel_encoder) -{ - struct drm_device *dev = intel_encoder->base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_dp *intel_dp; - enum intel_display_power_domain power_domain; - - if (intel_encoder->type != INTEL_OUTPUT_EDP) - return; - - intel_dp = enc_to_intel_dp(&intel_encoder->base); - - pps_lock(intel_dp); - - if (!edp_have_panel_vdd(intel_dp)) - goto out; - /* - * The VDD bit needs a power domain reference, so if the bit is - * already enabled when we boot or resume, grab this reference and - * schedule a vdd off, so we don't hold on to the reference - * indefinitely. - */ - DRM_DEBUG_KMS("VDD left on by BIOS, adjusting state tracking\n"); - power_domain = intel_display_port_power_domain(intel_encoder); - intel_display_power_get(dev_priv, power_domain); - - edp_panel_vdd_schedule_off(intel_dp); - out: - pps_unlock(intel_dp); -} - static bool intel_edp_init_connector(struct intel_dp *intel_dp, struct intel_connector *intel_connector) { @@ -5200,7 +5212,9 @@ static bool intel_edp_init_connector(struct intel_dp *intel_dp, if (!is_edp(intel_dp)) return true; - intel_edp_panel_vdd_sanitize(intel_encoder); + pps_lock(intel_dp); + intel_edp_panel_vdd_sanitize(intel_dp); + pps_unlock(intel_dp); /* Cache DPCD and EDID for edp. */ has_dpcd = intel_dp_get_dpcd(intel_dp); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 134ab08..82d8916 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -981,7 +981,6 @@ bool intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, void intel_edp_backlight_on(struct intel_dp *intel_dp); void intel_edp_backlight_off(struct intel_dp *intel_dp); void intel_edp_panel_vdd_on(struct intel_dp *intel_dp); -void intel_edp_panel_vdd_sanitize(struct intel_encoder *intel_encoder); void intel_edp_panel_on(struct intel_dp *intel_dp); void intel_edp_panel_off(struct intel_dp *intel_dp); void intel_edp_psr_enable(struct intel_dp *intel_dp); -- cgit v0.10.2 From c883ef1b1c998d2d66866772fd0fc34afa45641e Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Tue, 28 Oct 2014 17:32:30 +0200 Subject: drm/i915: Redefine WARN_ON to include the condition When looking at the bug report logs with triggered WARN_ON, the person doing bug triaging will have to find exact kernel source and match file/line. Attach the condition that triggered the WARN_ON to kernel log. In most cases the context is self evident and this way we can save developer time. The drawback is ~16kbytes bigger i915.ko Signed-off-by: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 0344fd5..0e12c6a 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -57,6 +57,9 @@ #define DRIVER_DESC "Intel Graphics" #define DRIVER_DATE "20141024" +#undef WARN_ON +#define WARN_ON(x) WARN(x, "WARN_ON(" #x ")") + enum pipe { INVALID_PIPE = -1, PIPE_A = 0, -- cgit v0.10.2 From 6402c330a62685c77c32d1bdfa882759c9f8e8a7 Mon Sep 17 00:00:00 2001 From: John Harrison Date: Fri, 31 Oct 2014 12:00:26 +0000 Subject: drm/i915: Fix null pointer dereference in ring cleanup code If a ring failed to initialise for any reason then the error path would try to clean up all rings including those that had not yet been allocated. The ring clean up code did a check that the ring was valid before starting its work. Unfortunately, that was after it had already dereferenced the ring to obtain a dev_private pointer. Signed-off-by: John Harrison Reviewed-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 803fc38..d7525bd 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1214,11 +1214,13 @@ static int gen8_emit_request(struct intel_ringbuffer *ringbuf) */ void intel_logical_ring_cleanup(struct intel_engine_cs *ring) { - struct drm_i915_private *dev_priv = ring->dev->dev_private; + struct drm_i915_private *dev_priv; if (!intel_ring_initialized(ring)) return; + dev_priv = ring->dev->dev_private; + intel_logical_ring_stop(ring); WARN_ON((I915_READ_MODE(ring) & MODE_IDLE) == 0); ring->preallocated_lazy_request = NULL; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index a8f72e8..f457146 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1845,12 +1845,15 @@ error: void intel_cleanup_ring_buffer(struct intel_engine_cs *ring) { - struct drm_i915_private *dev_priv = to_i915(ring->dev); - struct intel_ringbuffer *ringbuf = ring->buffer; + struct drm_i915_private *dev_priv; + struct intel_ringbuffer *ringbuf; if (!intel_ring_initialized(ring)) return; + dev_priv = to_i915(ring->dev); + ringbuf = ring->buffer; + intel_stop_ring_buffer(ring); WARN_ON(!IS_GEN2(ring->dev) && (I915_READ_MODE(ring) & MODE_IDLE) == 0); -- cgit v0.10.2 From cd0707cb1dbdd96b0358e5b10eba89c3bbe5a690 Mon Sep 17 00:00:00 2001 From: Dave Gordon Date: Thu, 30 Oct 2014 15:41:56 +0000 Subject: drm/i915: Remove redundant return value and WARN_ON execlists_submit_context() always returns 0, which is redundant. And its name is inaccurate, since it actually submits (up to) TWO contextS. So we rename it, change it to "void", and remove the WARN_ON() testing its return value. Change-Id: Ie225b0eca7754c6093c8b8bd15550b251b6feb82 Signed-off-by: Dave Gordon Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index d7525bd..70a6c63 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -356,9 +356,9 @@ static int execlists_ctx_write_tail(struct drm_i915_gem_object *ctx_obj, u32 tai return 0; } -static int execlists_submit_context(struct intel_engine_cs *ring, - struct intel_context *to0, u32 tail0, - struct intel_context *to1, u32 tail1) +static void execlists_submit_contexts(struct intel_engine_cs *ring, + struct intel_context *to0, u32 tail0, + struct intel_context *to1, u32 tail1) { struct drm_i915_gem_object *ctx_obj0; struct drm_i915_gem_object *ctx_obj1 = NULL; @@ -378,8 +378,6 @@ static int execlists_submit_context(struct intel_engine_cs *ring, } execlists_elsp_write(ring, ctx_obj0, ctx_obj1); - - return 0; } static void execlists_context_unqueue(struct intel_engine_cs *ring) @@ -413,9 +411,9 @@ static void execlists_context_unqueue(struct intel_engine_cs *ring) WARN_ON(req1 && req1->elsp_submitted); - WARN_ON(execlists_submit_context(ring, req0->ctx, req0->tail, - req1 ? req1->ctx : NULL, - req1 ? req1->tail : 0)); + execlists_submit_contexts(ring, req0->ctx, req0->tail, + req1 ? req1->ctx : NULL, + req1 ? req1->tail : 0); req0->elsp_submitted++; if (req1) -- cgit v0.10.2 From d0737e1d5944f570d5151ae6213952058c18a903 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:30 +0200 Subject: drm/i915: Make *_crtc_mode_set work on new_config This shouldn't change the behavior of those functions, since they are called after the new_config is made effective and that points to the current config. In a follow up patch, the mode set sequence will be changed so this is called before disabling crtcs, and in that case those functions should work on the staged config. Signed-off-by: Ander Conselvan de Oliveira [danvet: Flatten if by moving the check into the WARN.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index b182b9b..d417688 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -459,6 +459,27 @@ intel_ddi_get_crtc_encoder(struct drm_crtc *crtc) return ret; } +static struct intel_encoder * +intel_ddi_get_crtc_new_encoder(struct intel_crtc *crtc) +{ + struct drm_device *dev = crtc->base.dev; + struct intel_encoder *intel_encoder, *ret = NULL; + int num_encoders = 0; + + for_each_intel_encoder(dev, intel_encoder) { + if (intel_encoder->new_crtc == crtc) { + ret = intel_encoder; + num_encoders++; + } + } + + WARN(num_encoders != 1, "%d encoders on crtc for pipe %c\n", num_encoders, + pipe_name(crtc->pipe)); + + BUG_ON(ret == NULL); + return ret; +} + #define LC_FREQ 2700 #define LC_FREQ_2K U64_C(LC_FREQ * 2000) @@ -792,7 +813,7 @@ hsw_ddi_pll_select(struct intel_crtc *intel_crtc, WRPLL_DIVIDER_REFERENCE(r2) | WRPLL_DIVIDER_FEEDBACK(n2) | WRPLL_DIVIDER_POST(p); - intel_crtc->config.dpll_hw_state.wrpll = val; + intel_crtc->new_config->dpll_hw_state.wrpll = val; pll = intel_get_shared_dpll(intel_crtc); if (pll == NULL) { @@ -801,7 +822,7 @@ hsw_ddi_pll_select(struct intel_crtc *intel_crtc, return false; } - intel_crtc->config.ddi_pll_sel = PORT_CLK_SEL_WRPLL(pll->id); + intel_crtc->new_config->ddi_pll_sel = PORT_CLK_SEL_WRPLL(pll->id); } return true; @@ -817,9 +838,9 @@ hsw_ddi_pll_select(struct intel_crtc *intel_crtc, */ bool intel_ddi_pll_select(struct intel_crtc *intel_crtc) { - struct drm_crtc *crtc = &intel_crtc->base; - struct intel_encoder *intel_encoder = intel_ddi_get_crtc_encoder(crtc); - int clock = intel_crtc->config.port_clock; + struct intel_encoder *intel_encoder = + intel_ddi_get_crtc_new_encoder(intel_crtc); + int clock = intel_crtc->new_config->port_clock; intel_put_shared_dpll(intel_crtc); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index c17ec502..6bd290d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -420,13 +420,31 @@ bool intel_pipe_has_type(struct intel_crtc *crtc, int type) return false; } +/** + * Returns whether any output on the specified pipe will have the specified + * type after a staged modeset is complete, i.e., the same as + * intel_pipe_has_type() but looking at encoder->new_crtc instead of + * encoder->crtc. + */ +static bool intel_pipe_will_have_type(struct intel_crtc *crtc, int type) +{ + struct drm_device *dev = crtc->base.dev; + struct intel_encoder *encoder; + + for_each_intel_encoder(dev, encoder) + if (encoder->new_crtc == crtc && encoder->type == type) + return true; + + return false; +} + static const intel_limit_t *intel_ironlake_limit(struct intel_crtc *crtc, int refclk) { struct drm_device *dev = crtc->base.dev; const intel_limit_t *limit; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) { if (intel_is_dual_link_lvds(dev)) { if (refclk == 100000) limit = &intel_limits_ironlake_dual_lvds_100m; @@ -449,15 +467,15 @@ static const intel_limit_t *intel_g4x_limit(struct intel_crtc *crtc) struct drm_device *dev = crtc->base.dev; const intel_limit_t *limit; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) { if (intel_is_dual_link_lvds(dev)) limit = &intel_limits_g4x_dual_channel_lvds; else limit = &intel_limits_g4x_single_channel_lvds; - } else if (intel_pipe_has_type(crtc, INTEL_OUTPUT_HDMI) || - intel_pipe_has_type(crtc, INTEL_OUTPUT_ANALOG)) { + } else if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_HDMI) || + intel_pipe_will_have_type(crtc, INTEL_OUTPUT_ANALOG)) { limit = &intel_limits_g4x_hdmi; - } else if (intel_pipe_has_type(crtc, INTEL_OUTPUT_SDVO)) { + } else if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_SDVO)) { limit = &intel_limits_g4x_sdvo; } else /* The option is for other outputs */ limit = &intel_limits_i9xx_sdvo; @@ -475,7 +493,7 @@ static const intel_limit_t *intel_limit(struct intel_crtc *crtc, int refclk) else if (IS_G4X(dev)) { limit = intel_g4x_limit(crtc); } else if (IS_PINEVIEW(dev)) { - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) limit = &intel_limits_pineview_lvds; else limit = &intel_limits_pineview_sdvo; @@ -484,14 +502,14 @@ static const intel_limit_t *intel_limit(struct intel_crtc *crtc, int refclk) } else if (IS_VALLEYVIEW(dev)) { limit = &intel_limits_vlv; } else if (!IS_GEN2(dev)) { - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) limit = &intel_limits_i9xx_lvds; else limit = &intel_limits_i9xx_sdvo; } else { - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) limit = &intel_limits_i8xx_lvds; - else if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DVO)) + else if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_DVO)) limit = &intel_limits_i8xx_dvo; else limit = &intel_limits_i8xx_dac; @@ -586,7 +604,7 @@ i9xx_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, intel_clock_t clock; int err = target; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) { /* * For LVDS just rely on its current settings for dual-channel. * We haven't figured out how to reliably set up different @@ -647,7 +665,7 @@ pnv_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, intel_clock_t clock; int err = target; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) { /* * For LVDS just rely on its current settings for dual-channel. * We haven't figured out how to reliably set up different @@ -710,7 +728,7 @@ g4x_find_best_dpll(const intel_limit_t *limit, struct intel_crtc *crtc, int err_most = (target >> 8) + (target >> 9); found = false; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) { if (intel_is_dual_link_lvds(dev)) clock.p2 = limit->p2.p2_fast; else @@ -5628,7 +5646,7 @@ static int i9xx_get_refclk(struct intel_crtc *crtc, int num_connectors) if (IS_VALLEYVIEW(dev)) { refclk = 100000; - } else if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) && + } else if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS) && intel_panel_use_ssc(dev_priv) && num_connectors < 2) { refclk = dev_priv->vbt.lvds_ssc_freq; DRM_DEBUG_KMS("using SSC reference clock of %d kHz\n", refclk); @@ -6018,29 +6036,29 @@ static void i9xx_update_pll(struct intel_crtc *crtc, struct drm_i915_private *dev_priv = dev->dev_private; u32 dpll; bool is_sdvo; - struct dpll *clock = &crtc->config.dpll; + struct dpll *clock = &crtc->new_config->dpll; i9xx_update_pll_dividers(crtc, reduced_clock); - is_sdvo = intel_pipe_has_type(crtc, INTEL_OUTPUT_SDVO) || - intel_pipe_has_type(crtc, INTEL_OUTPUT_HDMI); + is_sdvo = intel_pipe_will_have_type(crtc, INTEL_OUTPUT_SDVO) || + intel_pipe_will_have_type(crtc, INTEL_OUTPUT_HDMI); dpll = DPLL_VGA_MODE_DIS; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) dpll |= DPLLB_MODE_LVDS; else dpll |= DPLLB_MODE_DAC_SERIAL; if (IS_I945G(dev) || IS_I945GM(dev) || IS_G33(dev)) { - dpll |= (crtc->config.pixel_multiplier - 1) + dpll |= (crtc->new_config->pixel_multiplier - 1) << SDVO_MULTIPLIER_SHIFT_HIRES; } if (is_sdvo) dpll |= DPLL_SDVO_HIGH_SPEED; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) dpll |= DPLL_SDVO_HIGH_SPEED; /* compute bitmask from p1 value */ @@ -6068,21 +6086,21 @@ static void i9xx_update_pll(struct intel_crtc *crtc, if (INTEL_INFO(dev)->gen >= 4) dpll |= (6 << PLL_LOAD_PULSE_PHASE_SHIFT); - if (crtc->config.sdvo_tv_clock) + if (crtc->new_config->sdvo_tv_clock) dpll |= PLL_REF_INPUT_TVCLKINBC; - else if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) && + else if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS) && intel_panel_use_ssc(dev_priv) && num_connectors < 2) dpll |= PLLB_REF_INPUT_SPREADSPECTRUMIN; else dpll |= PLL_REF_INPUT_DREFCLK; dpll |= DPLL_VCO_ENABLE; - crtc->config.dpll_hw_state.dpll = dpll; + crtc->new_config->dpll_hw_state.dpll = dpll; if (INTEL_INFO(dev)->gen >= 4) { - u32 dpll_md = (crtc->config.pixel_multiplier - 1) + u32 dpll_md = (crtc->new_config->pixel_multiplier - 1) << DPLL_MD_UDI_MULTIPLIER_SHIFT; - crtc->config.dpll_hw_state.dpll_md = dpll_md; + crtc->new_config->dpll_hw_state.dpll_md = dpll_md; } } @@ -6093,13 +6111,13 @@ static void i8xx_update_pll(struct intel_crtc *crtc, struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; u32 dpll; - struct dpll *clock = &crtc->config.dpll; + struct dpll *clock = &crtc->new_config->dpll; i9xx_update_pll_dividers(crtc, reduced_clock); dpll = DPLL_VGA_MODE_DIS; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) { + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS)) { dpll |= (1 << (clock->p1 - 1)) << DPLL_FPA01_P1_POST_DIV_SHIFT; } else { if (clock->p1 == 2) @@ -6110,17 +6128,17 @@ static void i8xx_update_pll(struct intel_crtc *crtc, dpll |= PLL_P2_DIVIDE_BY_4; } - if (!IS_I830(dev) && intel_pipe_has_type(crtc, INTEL_OUTPUT_DVO)) + if (!IS_I830(dev) && intel_pipe_will_have_type(crtc, INTEL_OUTPUT_DVO)) dpll |= DPLL_DVO_2X_MODE; - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) && + if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_LVDS) && intel_panel_use_ssc(dev_priv) && num_connectors < 2) dpll |= PLLB_REF_INPUT_SPREADSPECTRUMIN; else dpll |= PLL_REF_INPUT_DREFCLK; dpll |= DPLL_VCO_ENABLE; - crtc->config.dpll_hw_state.dpll = dpll; + crtc->new_config->dpll_hw_state.dpll = dpll; } static void intel_set_pipe_timings(struct intel_crtc *intel_crtc) @@ -6329,7 +6347,10 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, struct intel_encoder *encoder; const intel_limit_t *limit; - for_each_encoder_on_crtc(dev, &crtc->base, encoder) { + for_each_intel_encoder(dev, encoder) { + if (encoder->new_crtc != crtc) + continue; + switch (encoder->type) { case INTEL_OUTPUT_LVDS: is_lvds = true; @@ -6347,7 +6368,7 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, if (is_dsi) return 0; - if (!crtc->config.clock_set) { + if (!crtc->new_config->clock_set) { refclk = i9xx_get_refclk(crtc, num_connectors); /* @@ -6358,7 +6379,7 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, */ limit = intel_limit(crtc, refclk); ok = dev_priv->display.find_dpll(limit, crtc, - crtc->config.port_clock, + crtc->new_config->port_clock, refclk, NULL, &clock); if (!ok) { DRM_ERROR("Couldn't find PLL settings for mode!\n"); @@ -6379,11 +6400,11 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, &reduced_clock); } /* Compat-code for transition, will disappear. */ - crtc->config.dpll.n = clock.n; - crtc->config.dpll.m1 = clock.m1; - crtc->config.dpll.m2 = clock.m2; - crtc->config.dpll.p1 = clock.p1; - crtc->config.dpll.p2 = clock.p2; + crtc->new_config->dpll.n = clock.n; + crtc->new_config->dpll.m1 = clock.m1; + crtc->new_config->dpll.m2 = clock.m2; + crtc->new_config->dpll.p1 = clock.p1; + crtc->new_config->dpll.p2 = clock.p2; } if (IS_GEN2(dev)) { @@ -6391,9 +6412,9 @@ static int i9xx_crtc_mode_set(struct intel_crtc *crtc, has_reduced_clock ? &reduced_clock : NULL, num_connectors); } else if (IS_CHERRYVIEW(dev)) { - chv_update_pll(crtc, &crtc->config); + chv_update_pll(crtc, crtc->new_config); } else if (IS_VALLEYVIEW(dev)) { - vlv_update_pll(crtc, &crtc->config); + vlv_update_pll(crtc, crtc->new_config); } else { i9xx_update_pll(crtc, has_reduced_clock ? &reduced_clock : NULL, @@ -7003,7 +7024,10 @@ static int ironlake_get_refclk(struct drm_crtc *crtc) int num_connectors = 0; bool is_lvds = false; - for_each_encoder_on_crtc(dev, crtc, encoder) { + for_each_intel_encoder(dev, encoder) { + if (encoder->new_crtc != to_intel_crtc(crtc)) + continue; + switch (encoder->type) { case INTEL_OUTPUT_LVDS: is_lvds = true; @@ -7194,7 +7218,7 @@ static bool ironlake_compute_clocks(struct drm_crtc *crtc, const intel_limit_t *limit; bool ret, is_lvds = false; - is_lvds = intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_LVDS); + is_lvds = intel_pipe_will_have_type(intel_crtc, INTEL_OUTPUT_LVDS); refclk = ironlake_get_refclk(crtc); @@ -7205,7 +7229,7 @@ static bool ironlake_compute_clocks(struct drm_crtc *crtc, */ limit = intel_limit(intel_crtc, refclk); ret = dev_priv->display.find_dpll(limit, intel_crtc, - intel_crtc->config.port_clock, + intel_crtc->new_config->port_clock, refclk, NULL, clock); if (!ret) return false; @@ -7255,7 +7279,10 @@ static uint32_t ironlake_compute_dpll(struct intel_crtc *intel_crtc, int factor, num_connectors = 0; bool is_lvds = false, is_sdvo = false; - for_each_encoder_on_crtc(dev, crtc, intel_encoder) { + for_each_intel_encoder(dev, intel_encoder) { + if (intel_encoder->new_crtc != to_intel_crtc(crtc)) + continue; + switch (intel_encoder->type) { case INTEL_OUTPUT_LVDS: is_lvds = true; @@ -7278,10 +7305,10 @@ static uint32_t ironlake_compute_dpll(struct intel_crtc *intel_crtc, dev_priv->vbt.lvds_ssc_freq == 100000) || (HAS_PCH_IBX(dev) && intel_is_dual_link_lvds(dev))) factor = 25; - } else if (intel_crtc->config.sdvo_tv_clock) + } else if (intel_crtc->new_config->sdvo_tv_clock) factor = 20; - if (ironlake_needs_fb_cb_tune(&intel_crtc->config.dpll, factor)) + if (ironlake_needs_fb_cb_tune(&intel_crtc->new_config->dpll, factor)) *fp |= FP_CB_TUNE; if (fp2 && (reduced_clock->m < factor * reduced_clock->n)) @@ -7294,20 +7321,20 @@ static uint32_t ironlake_compute_dpll(struct intel_crtc *intel_crtc, else dpll |= DPLLB_MODE_DAC_SERIAL; - dpll |= (intel_crtc->config.pixel_multiplier - 1) + dpll |= (intel_crtc->new_config->pixel_multiplier - 1) << PLL_REF_SDVO_HDMI_MULTIPLIER_SHIFT; if (is_sdvo) dpll |= DPLL_SDVO_HIGH_SPEED; - if (intel_crtc->config.has_dp_encoder) + if (intel_crtc->new_config->has_dp_encoder) dpll |= DPLL_SDVO_HIGH_SPEED; /* compute bitmask from p1 value */ - dpll |= (1 << (intel_crtc->config.dpll.p1 - 1)) << DPLL_FPA01_P1_POST_DIV_SHIFT; + dpll |= (1 << (intel_crtc->new_config->dpll.p1 - 1)) << DPLL_FPA01_P1_POST_DIV_SHIFT; /* also FPA1 */ - dpll |= (1 << (intel_crtc->config.dpll.p1 - 1)) << DPLL_FPA1_P1_POST_DIV_SHIFT; + dpll |= (1 << (intel_crtc->new_config->dpll.p1 - 1)) << DPLL_FPA1_P1_POST_DIV_SHIFT; - switch (intel_crtc->config.dpll.p2) { + switch (intel_crtc->new_config->dpll.p2) { case 5: dpll |= DPLL_DAC_SERIAL_P2_CLOCK_DIV_5; break; @@ -7348,22 +7375,22 @@ static int ironlake_crtc_mode_set(struct intel_crtc *crtc, ok = ironlake_compute_clocks(&crtc->base, &clock, &has_reduced_clock, &reduced_clock); - if (!ok && !crtc->config.clock_set) { + if (!ok && !crtc->new_config->clock_set) { DRM_ERROR("Couldn't find PLL settings for mode!\n"); return -EINVAL; } /* Compat-code for transition, will disappear. */ - if (!crtc->config.clock_set) { - crtc->config.dpll.n = clock.n; - crtc->config.dpll.m1 = clock.m1; - crtc->config.dpll.m2 = clock.m2; - crtc->config.dpll.p1 = clock.p1; - crtc->config.dpll.p2 = clock.p2; + if (!crtc->new_config->clock_set) { + crtc->new_config->dpll.n = clock.n; + crtc->new_config->dpll.m1 = clock.m1; + crtc->new_config->dpll.m2 = clock.m2; + crtc->new_config->dpll.p1 = clock.p1; + crtc->new_config->dpll.p2 = clock.p2; } /* CPU eDP is the only output that doesn't need a PCH PLL of its own. */ - if (crtc->config.has_pch_encoder) { - fp = i9xx_dpll_compute_fp(&crtc->config.dpll); + if (crtc->new_config->has_pch_encoder) { + fp = i9xx_dpll_compute_fp(&crtc->new_config->dpll); if (has_reduced_clock) fp2 = i9xx_dpll_compute_fp(&reduced_clock); @@ -7371,12 +7398,12 @@ static int ironlake_crtc_mode_set(struct intel_crtc *crtc, &fp, &reduced_clock, has_reduced_clock ? &fp2 : NULL); - crtc->config.dpll_hw_state.dpll = dpll; - crtc->config.dpll_hw_state.fp0 = fp; + crtc->new_config->dpll_hw_state.dpll = dpll; + crtc->new_config->dpll_hw_state.fp0 = fp; if (has_reduced_clock) - crtc->config.dpll_hw_state.fp1 = fp2; + crtc->new_config->dpll_hw_state.fp1 = fp2; else - crtc->config.dpll_hw_state.fp1 = fp; + crtc->new_config->dpll_hw_state.fp1 = fp; pll = intel_get_shared_dpll(crtc); if (pll == NULL) { -- cgit v0.10.2 From 3cb9ae4fd82ebc9759bc59998c88c385d35496fe Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 29 Oct 2014 10:03:57 +0100 Subject: drm: Move drm_crtc_init from drm_crtc.h to drm_plane_helper.h Just a bit of OCD cleanup on headers - this function isn't the core interface any more but just a helper for drivers who haven't yet transitioned to universal planes. Put the declaration at the right spot and sprinkle necessary #includes over all drivers. Maybe this helps to encourage driver maintainers to do the switch. v2: Fix #include ordering for tegra, reported by 0-day builder. v3: Include required headers, reported by Thierry. Cc: Matt Roper Cc: Thierry Reding Reviewed-by: Matt Roper Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/armada/armada_crtc.c b/drivers/gpu/drm/armada/armada_crtc.c index 9a0cc09..0b164fb 100644 --- a/drivers/gpu/drm/armada/armada_crtc.c +++ b/drivers/gpu/drm/armada/armada_crtc.c @@ -12,6 +12,7 @@ #include #include #include +#include #include "armada_crtc.h" #include "armada_drm.h" #include "armada_fb.h" diff --git a/drivers/gpu/drm/ast/ast_mode.c b/drivers/gpu/drm/ast/ast_mode.c index 9dc0fd5..b7ee263 100644 --- a/drivers/gpu/drm/ast/ast_mode.c +++ b/drivers/gpu/drm/ast/ast_mode.c @@ -31,6 +31,7 @@ #include #include #include +#include #include "ast_drv.h" #include "ast_tables.h" diff --git a/drivers/gpu/drm/bochs/bochs_kms.c b/drivers/gpu/drm/bochs/bochs_kms.c index 6b7efcf3..5ffd489 100644 --- a/drivers/gpu/drm/bochs/bochs_kms.c +++ b/drivers/gpu/drm/bochs/bochs_kms.c @@ -6,6 +6,7 @@ */ #include "bochs.h" +#include static int defx = 1024; static int defy = 768; diff --git a/drivers/gpu/drm/cirrus/cirrus_mode.c b/drivers/gpu/drm/cirrus/cirrus_mode.c index c7c5a9d..99d4a74 100644 --- a/drivers/gpu/drm/cirrus/cirrus_mode.c +++ b/drivers/gpu/drm/cirrus/cirrus_mode.c @@ -16,6 +16,7 @@ */ #include #include +#include #include + KMS Data Structures +!Iinclude/drm/drm_crtc.h + + KMS Locking !Pdrivers/gpu/drm/drm_modeset_lock.c kms locking !Iinclude/drm/drm_modeset_lock.h diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index c3baaee..3a80605 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -142,8 +142,8 @@ struct drm_framebuffer_funcs { int (*create_handle)(struct drm_framebuffer *fb, struct drm_file *file_priv, unsigned int *handle); - /** - * Optinal callback for the dirty fb ioctl. + /* + * Optional callback for the dirty fb ioctl. * * Userspace can notify the driver via this callback * that a area of the framebuffer has changed and should @@ -226,7 +226,7 @@ struct drm_plane; struct drm_bridge; /** - * drm_crtc_funcs - control CRTCs for a given device + * struct drm_crtc_funcs - control CRTCs for a given device * @save: save CRTC state * @restore: restore CRTC state * @reset: reset CRTC after state has been invalidated (e.g. resume) @@ -290,7 +290,7 @@ struct drm_crtc_funcs { }; /** - * drm_crtc - central CRTC control structure + * struct drm_crtc - central CRTC control structure * @dev: parent DRM device * @head: list management * @mutex: per-CRTC locking @@ -322,7 +322,7 @@ struct drm_crtc { struct device_node *port; struct list_head head; - /** + /* * crtc mutex * * This provides a read lock for the overall crtc state (mode, dpms @@ -377,7 +377,7 @@ struct drm_crtc { /** - * drm_connector_funcs - control connectors on a given device + * struct drm_connector_funcs - control connectors on a given device * @dpms: set power state (see drm_crtc_funcs above) * @save: save connector state * @restore: restore connector state @@ -414,7 +414,7 @@ struct drm_connector_funcs { }; /** - * drm_encoder_funcs - encoder controls + * struct drm_encoder_funcs - encoder controls * @reset: reset state (e.g. at init or resume time) * @destroy: cleanup and free associated data * @@ -428,7 +428,7 @@ struct drm_encoder_funcs { #define DRM_CONNECTOR_MAX_ENCODER 3 /** - * drm_encoder - central DRM encoder structure + * struct drm_encoder - central DRM encoder structure * @dev: parent DRM device * @head: list management * @base: base KMS object @@ -472,7 +472,7 @@ struct drm_encoder { #define MAX_ELD_BYTES 128 /** - * drm_connector - central DRM connector control structure + * struct drm_connector - central DRM connector control structure * @dev: parent DRM device * @kdev: kernel device for sysfs attributes * @attr: sysfs attributes @@ -566,7 +566,7 @@ struct drm_connector { }; /** - * drm_plane_funcs - driver plane control functions + * struct drm_plane_funcs - driver plane control functions * @update_plane: update the plane configuration * @disable_plane: shut down the plane * @destroy: clean up plane resources @@ -594,7 +594,7 @@ enum drm_plane_type { }; /** - * drm_plane - central DRM plane control structure + * struct drm_plane - central DRM plane control structure * @dev: DRM device this plane belongs to * @head: for list management * @base: base mode object @@ -632,7 +632,7 @@ struct drm_plane { }; /** - * drm_bridge_funcs - drm_bridge control functions + * struct drm_bridge_funcs - drm_bridge control functions * @mode_fixup: Try to fixup (or reject entirely) proposed mode for this bridge * @disable: Called right before encoder prepare, disables the bridge * @post_disable: Called right after encoder prepare, for lockstepped disable @@ -656,7 +656,7 @@ struct drm_bridge_funcs { }; /** - * drm_bridge - central DRM bridge control structure + * struct drm_bridge - central DRM bridge control structure * @dev: DRM device this bridge belongs to * @head: list management * @base: base mode object @@ -674,7 +674,7 @@ struct drm_bridge { }; /** - * drm_mode_set - new values for a CRTC config change + * struct drm_mode_set - new values for a CRTC config change * @head: list management * @fb: framebuffer to use for new config * @crtc: CRTC whose configuration we're about to change -- cgit v0.10.2 From 2c0c33d41e0ecb46c37583a6179eeb1f16767296 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 27 Oct 2014 20:19:38 +0100 Subject: drm: fixup kerneldoc in drm_crtc.h I've tried to cc all the people who have recently added new stuff but forgotten to update documentation. I've also decided not to bother documenting the massive property list in struct drm_mode_config. If that beast keeps on growing we might want to extract it into a separate structure which we won't document. Cc: Thomas Wood Cc: Sean Paul Cc: Russell King Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 3a80605..125bb7f 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -231,6 +231,7 @@ struct drm_bridge; * @restore: restore CRTC state * @reset: reset CRTC after state has been invalidated (e.g. resume) * @cursor_set: setup the cursor + * @cursor_set2: setup the cursor with hotspot, superseeds @cursor_set if set * @cursor_move: move the cursor * @gamma_set: specify color ramp for CRTC * @destroy: deinit and free object @@ -292,11 +293,14 @@ struct drm_crtc_funcs { /** * struct drm_crtc - central CRTC control structure * @dev: parent DRM device + * @port: OF node used by drm_of_find_possible_crtcs() * @head: list management * @mutex: per-CRTC locking * @base: base KMS object for ID tracking etc. * @primary: primary plane for this CRTC * @cursor: cursor plane for this CRTC + * @cursor_x: current x position of the cursor, used for universal cursor planes + * @cursor_y: current y position of the cursor, used for universal cursor planes * @enabled: is this CRTC enabled? * @mode: current mode timings * @hwmode: mode timings as programmed to hw regs @@ -309,10 +313,12 @@ struct drm_crtc_funcs { * @gamma_size: size of gamma ramp * @gamma_store: gamma ramp values * @framedur_ns: precise frame timing - * @framedur_ns: precise line timing + * @linedur_ns: precise line timing * @pixeldur_ns: precise pixel timing * @helper_private: mid-layer private data * @properties: property tracking for this CRTC + * @acquire_ctx: per-CRTC implicit acquire context used by atomic drivers for + * legacy ioctls * * Each CRTC may have one or more connectors associated with it. This structure * allows the CRTC to be controlled. @@ -483,6 +489,7 @@ struct drm_encoder { * @connector_type_id: index into connector type enum * @interlace_allowed: can this connector handle interlaced modes? * @doublescan_allowed: can this connector handle doublescan? + * @stereo_allowed: can this connector handle stereo modes? * @modes: modes available on this connector (from fill_modes() + user) * @status: one of the drm_connector_status enums (connected, not, or unknown) * @probed_modes: list of modes derived directly from the display @@ -490,10 +497,13 @@ struct drm_encoder { * @funcs: connector control functions * @edid_blob_ptr: DRM property containing EDID if present * @properties: property tracking for this connector + * @path_blob_ptr: DRM blob property data for the DP MST path property * @polled: a %DRM_CONNECTOR_POLL_ value for core driven polling * @dpms: current dpms state * @helper_private: mid-layer private data + * @cmdline_mode: mode line parsed from the kernel cmdline for this connector * @force: a %DRM_FORCE_ state for forced mode sets + * @override_edid: has the EDID been overwritten through debugfs for testing? * @encoder_ids: valid encoders for this connector * @encoder: encoder driving this connector, if any * @eld: EDID-like data, if present @@ -503,6 +513,8 @@ struct drm_encoder { * @video_latency: video latency info from ELD, if found * @audio_latency: audio latency info from ELD, if found * @null_edid_counter: track sinks that give us all zeros for the EDID + * @bad_edid_counter: track sinks that give us an EDID with invalid checksum + * @debugfs_entry: debugfs directory for this connector * * Each connector may be connected to one or more CRTCs, or may be clonable by * another connector if they can share a CRTC. Each connector also has a specific @@ -570,6 +582,7 @@ struct drm_connector { * @update_plane: update the plane configuration * @disable_plane: shut down the plane * @destroy: clean up plane resources + * @reset: reset plane after state has been invalidated (e.g. resume) * @set_property: called when a property is changed */ struct drm_plane_funcs { @@ -603,6 +616,8 @@ enum drm_plane_type { * @format_count: number of formats supported * @crtc: currently bound CRTC * @fb: currently bound fb + * @old_fb: Temporary tracking of the old fb while a modeset is ongoing. Used by + * drm_mode_set_config_internal() to implement correct refcounting. * @funcs: helper functions * @properties: property tracking for this plane * @type: type of plane (overlay, primary, cursor) @@ -620,8 +635,6 @@ struct drm_plane { struct drm_crtc *crtc; struct drm_framebuffer *fb; - /* Temporary tracking of the old fb while a modeset is ongoing. Used - * by drm_mode_set_config_internal to implement correct refcounting. */ struct drm_framebuffer *old_fb; const struct drm_plane_funcs *funcs; @@ -675,7 +688,6 @@ struct drm_bridge { /** * struct drm_mode_set - new values for a CRTC config change - * @head: list management * @fb: framebuffer to use for new config * @crtc: CRTC whose configuration we're about to change * @mode: mode timings to use @@ -717,10 +729,11 @@ struct drm_mode_config_funcs { }; /** - * drm_mode_group - group of mode setting resources for potential sub-grouping + * struct drm_mode_group - group of mode setting resources for potential sub-grouping * @num_crtcs: CRTC count * @num_encoders: encoder count * @num_connectors: connector count + * @num_bridges: bridge count * @id_list: list of KMS object IDs in this group * * Currently this simply tracks the global mode setting state. But in the @@ -740,10 +753,14 @@ struct drm_mode_group { }; /** - * drm_mode_config - Mode configuration control structure + * struct drm_mode_config - Mode configuration control structure * @mutex: mutex protecting KMS related lists and structures + * @connection_mutex: ww mutex protecting connector state and routing + * @acquire_ctx: global implicit acquire context used by atomic drivers for + * legacy ioctls * @idr_mutex: mutex for KMS ID allocation and management * @crtc_idr: main KMS ID tracking object + * @fb_lock: mutex to protect fb state and lists * @num_fb: number of fbs available * @fb_list: list of framebuffers available * @num_connector: number of connectors on this device @@ -752,17 +769,28 @@ struct drm_mode_group { * @bridge_list: list of bridge objects * @num_encoder: number of encoders on this device * @encoder_list: list of encoder objects + * @num_overlay_plane: number of overlay planes on this device + * @num_total_plane: number of universal (i.e. with primary/curso) planes on this device + * @plane_list: list of plane objects * @num_crtc: number of CRTCs on this device * @crtc_list: list of CRTC objects + * @property_list: list of property objects * @min_width: minimum pixel width on this device * @min_height: minimum pixel height on this device * @max_width: maximum pixel width on this device * @max_height: maximum pixel height on this device * @funcs: core driver provided mode setting functions * @fb_base: base address of the framebuffer - * @poll_enabled: track polling status for this device + * @poll_enabled: track polling support for this device + * @poll_running: track polling status for this device * @output_poll_work: delayed work for polling in process context + * @property_blob_list: list of all the blob property objects * @*_property: core property tracking + * @preferred_depth: preferred RBG pixel depth, used by fb helpers + * @prefer_shadow: hint to userspace to prefer shadow-fb rendering + * @async_page_flip: does this device support async flips on the primary plane? + * @cursor_width: hint to userspace for max cursor width + * @cursor_height: hint to userspace for max cursor height * * Core mode resource tracking structure. All CRTC, encoders, and connectors * enumerated by the driver are added here, as are global properties. Some @@ -776,14 +804,7 @@ struct drm_mode_config { struct idr crtc_idr; /* use this idr for all IDs, fb, crtc, connector, modes - just makes life easier */ /* this is limited to one for now */ - - /** - * fb_lock - mutex to protect fb state - * - * Besides the global fb list his also protects the fbs list in the - * file_priv - */ - struct mutex fb_lock; + struct mutex fb_lock; /* proctects global and per-file fb lists */ int num_fb; struct list_head fb_list; -- cgit v0.10.2 From e9cf6194ab0cdb8f40f1eca7b230751c2289557e Mon Sep 17 00:00:00 2001 From: Todd Previte Date: Tue, 4 Nov 2014 15:17:35 -0700 Subject: drm/dp: Add counters in the drm_dp_aux struct for I2C NACKs and DEFERs These counters are used for Displayort compliance testing to detect error conditions when executing tests 4.2.2.4 and 4.2.2.5 in the Displayport Link CTS specificaiton. They determine whether to use the preferred/requested mode or the failsafe mode during these tests. V2: - Addressed previous review feedback - Updated commit message - Changed from uint8_t to uint32_t Cc: dri-devel@lists.freedesktop.org Signed-off-by: Todd Previte [danvet: s/uint32_t/unsigned/ for clearer intent. Also drop the i915 from the subject, it's all core stuff.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_dp_helper.c b/drivers/gpu/drm/drm_dp_helper.c index c088bad..959e207 100644 --- a/drivers/gpu/drm/drm_dp_helper.c +++ b/drivers/gpu/drm/drm_dp_helper.c @@ -462,10 +462,12 @@ static int drm_dp_i2c_do_msg(struct drm_dp_aux *aux, struct drm_dp_aux_msg *msg) case DP_AUX_I2C_REPLY_NACK: DRM_DEBUG_KMS("I2C nack\n"); + aux->i2c_nack_count++; return -EREMOTEIO; case DP_AUX_I2C_REPLY_DEFER: DRM_DEBUG_KMS("I2C defer\n"); + aux->i2c_defer_count++; usleep_range(400, 500); continue; diff --git a/include/drm/drm_dp_helper.h b/include/drm/drm_dp_helper.h index afebb08..6041ad8 100644 --- a/include/drm/drm_dp_helper.h +++ b/include/drm/drm_dp_helper.h @@ -530,6 +530,7 @@ struct drm_dp_aux { struct mutex hw_mutex; ssize_t (*transfer)(struct drm_dp_aux *aux, struct drm_dp_aux_msg *msg); + unsigned i2c_nack_count, i2c_defer_count; }; ssize_t drm_dp_dpcd_read(struct drm_dp_aux *aux, unsigned int offset, -- cgit v0.10.2 From babc94936b7a48f0411c85a63074c0e378b55761 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Tue, 28 Oct 2014 16:20:47 +0200 Subject: drm/edid: add #defines and helpers for ELD In the interest of reducing magic numbers and having to cross check with the specs all the time. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/include/drm/drm_edid.h b/include/drm/drm_edid.h index b96031d..c2f1bfa 100644 --- a/include/drm/drm_edid.h +++ b/include/drm/drm_edid.h @@ -207,6 +207,61 @@ struct detailed_timing { #define DRM_EDID_HDMI_DC_30 (1 << 4) #define DRM_EDID_HDMI_DC_Y444 (1 << 3) +/* ELD Header Block */ +#define DRM_ELD_HEADER_BLOCK_SIZE 4 + +#define DRM_ELD_VER 0 +# define DRM_ELD_VER_SHIFT 3 +# define DRM_ELD_VER_MASK (0x1f << 3) + +#define DRM_ELD_BASELINE_ELD_LEN 2 /* in dwords! */ + +/* ELD Baseline Block for ELD_Ver == 2 */ +#define DRM_ELD_CEA_EDID_VER_MNL 4 +# define DRM_ELD_CEA_EDID_VER_SHIFT 5 +# define DRM_ELD_CEA_EDID_VER_MASK (7 << 5) +# define DRM_ELD_CEA_EDID_VER_NONE (0 << 5) +# define DRM_ELD_CEA_EDID_VER_CEA861 (1 << 5) +# define DRM_ELD_CEA_EDID_VER_CEA861A (2 << 5) +# define DRM_ELD_CEA_EDID_VER_CEA861BCD (3 << 5) +# define DRM_ELD_MNL_SHIFT 0 +# define DRM_ELD_MNL_MASK (0x1f << 0) + +#define DRM_ELD_SAD_COUNT_CONN_TYPE 5 +# define DRM_ELD_SAD_COUNT_SHIFT 4 +# define DRM_ELD_SAD_COUNT_MASK (0xf << 4) +# define DRM_ELD_CONN_TYPE_SHIFT 2 +# define DRM_ELD_CONN_TYPE_MASK (3 << 2) +# define DRM_ELD_CONN_TYPE_HDMI (0 << 2) +# define DRM_ELD_CONN_TYPE_DP (1 << 2) +# define DRM_ELD_SUPPORTS_AI (1 << 1) +# define DRM_ELD_SUPPORTS_HDCP (1 << 0) + +#define DRM_ELD_AUD_SYNCH_DELAY 6 /* in units of 2 ms */ +# define DRM_ELD_AUD_SYNCH_DELAY_MAX 0xfa /* 500 ms */ + +#define DRM_ELD_SPEAKER 7 +# define DRM_ELD_SPEAKER_RLRC (1 << 6) +# define DRM_ELD_SPEAKER_FLRC (1 << 5) +# define DRM_ELD_SPEAKER_RC (1 << 4) +# define DRM_ELD_SPEAKER_RLR (1 << 3) +# define DRM_ELD_SPEAKER_FC (1 << 2) +# define DRM_ELD_SPEAKER_LFE (1 << 1) +# define DRM_ELD_SPEAKER_FLR (1 << 0) + +#define DRM_ELD_PORT_ID 8 /* offsets 8..15 inclusive */ +# define DRM_ELD_PORT_ID_LEN 8 + +#define DRM_ELD_MANUFACTURER_NAME0 16 +#define DRM_ELD_MANUFACTURER_NAME1 17 + +#define DRM_ELD_PRODUCT_CODE0 18 +#define DRM_ELD_PRODUCT_CODE1 19 + +#define DRM_ELD_MONITOR_NAME_STRING 20 /* offsets 20..(20+mnl-1) inclusive */ + +#define DRM_ELD_CEA_SAD(mnl, sad) (20 + (mnl) + 3 * (sad)) + struct edid { u8 header[8]; /* Vendor & product info */ @@ -279,4 +334,51 @@ int drm_hdmi_vendor_infoframe_from_display_mode(struct hdmi_vendor_infoframe *frame, const struct drm_display_mode *mode); +/** + * drm_eld_mnl - Get ELD monitor name length in bytes. + * @eld: pointer to an eld memory structure with mnl set + */ +static inline int drm_eld_mnl(const uint8_t *eld) +{ + return (eld[DRM_ELD_CEA_EDID_VER_MNL] & DRM_ELD_MNL_MASK) >> DRM_ELD_MNL_SHIFT; +} + +/** + * drm_eld_sad_count - Get ELD SAD count. + * @eld: pointer to an eld memory structure with sad_count set + */ +static inline int drm_eld_sad_count(const uint8_t *eld) +{ + return (eld[DRM_ELD_SAD_COUNT_CONN_TYPE] & DRM_ELD_SAD_COUNT_MASK) >> + DRM_ELD_SAD_COUNT_SHIFT; +} + +/** + * drm_eld_calc_baseline_block_size - Calculate baseline block size in bytes + * @eld: pointer to an eld memory structure with mnl and sad_count set + * + * This is a helper for determining the payload size of the baseline block, in + * bytes, for e.g. setting the Baseline_ELD_Len field in the ELD header block. + */ +static inline int drm_eld_calc_baseline_block_size(const uint8_t *eld) +{ + return DRM_ELD_MONITOR_NAME_STRING - DRM_ELD_HEADER_BLOCK_SIZE + + drm_eld_mnl(eld) + drm_eld_sad_count(eld) * 3; +} + +/** + * drm_eld_size - Get ELD size in bytes + * @eld: pointer to a complete eld memory structure + * + * The returned value does not include the vendor block. It's vendor specific, + * and comprises of the remaining bytes in the ELD memory buffer after + * drm_eld_size() bytes of header and baseline block. + * + * The returned value is guaranteed to be a multiple of 4. + */ +static inline int drm_eld_size(const uint8_t *eld) +{ + return DRM_ELD_HEADER_BLOCK_SIZE + eld[DRM_ELD_BASELINE_ELD_LEN] * 4; +} + #endif /* __DRM_EDID_H__ */ -- cgit v0.10.2 From b7a1aafda61c27d54230ce7c8a4806a749ecf2f3 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 27 Oct 2014 20:37:37 +0100 Subject: drm/modeset_lock: document trylock_only in kerneldoc I've forgotten to do this in: commit cb597bb3a2fbfc871cc1c703fb330d247bd21394 Author: Daniel Vetter Date: Sun Jul 27 19:09:33 2014 +0200 drm: trylock modest locking for fbdev panics Oops, fix this asap. In my defense kerneldoc is really awful and there's no way it can pick up structured comments per struct member. Which means we need both since people won't scroll up even a few lines. Reviewed-by: Sean Paul Reviewed-by: Thierry Reding Signed-off-by: Daniel Vetter diff --git a/include/drm/drm_modeset_lock.h b/include/drm/drm_modeset_lock.h index 75a5c45..28931a2 100644 --- a/include/drm/drm_modeset_lock.h +++ b/include/drm/drm_modeset_lock.h @@ -33,6 +33,7 @@ struct drm_modeset_lock; * @ww_ctx: base acquire ctx * @contended: used internally for -EDEADLK handling * @locked: list of held locks + * @trylock_only: trylock mode used in atomic contexts/panic notifiers * * Each thread competing for a set of locks must use one acquire * ctx. And if any lock fxn returns -EDEADLK, it must backoff and -- cgit v0.10.2 From 144ecb97cd57d2a61cc455730a3337e413499cae Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 27 Oct 2014 20:28:44 +0100 Subject: drm: Add atomic driver interface definitions for objects Heavily based upon Rob Clark's atomic series. - Dropped the connector state from the crtc state, instead opting for a full-blown connector state. The only thing it has is the desired crtc, but drivers which have connector properties have now a data-structure to subclass. - Rename create_state to duplicate_state. Especially for legacy ioctls we want updates on top of existing state, so we need a way to get at the current state. We need to be careful to clear the backpointers to the global state correctly though. - Drop property values. Drivers with properties simply need to subclass the datastructures and track the decoded values in there. I also think that common properties (like rotation) should be decoded and stored in the core structures. - Create a new set of ->atomic_set_prop functions, for smoother transitions from legacy to atomic operations. - Pass the ->atomic_set_prop ioctl the right structure to avoid chasing pointers in drivers. - Drop temporary boolean state for now until we resurrect them with the helper functions. - Drop invert_dimensions. For now we don't need any checking since that's done by the higher-level legacy ioctls. But even then we should also add rotation/flip tracking to the core drm_crtc_state, not just whether the dimensions are inverted. - Track crtc state with an enable/disable. That's equivalent to mode_valid, but a bit clearer that it means the entire crtc. The global interface will follow in subsequent patches. v2: We need to allow drivers to somehow set up the initial state and clear it on resume. So add a plane->reset callback for that. Helpers will be provided with default behaviour for all these. v3: Split out the plane->reset into a separate patch. v4: Improve kerneldoc in drm_crtc.h v5: Remove unused inline functions for handling state objects, those callbacks are now mandatory for full atomic support. v6: Fix commit message nit Sean noticed. Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 125bb7f..3554868 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -224,6 +224,25 @@ struct drm_encoder; struct drm_pending_vblank_event; struct drm_plane; struct drm_bridge; +struct drm_atomic_state; + +/** + * struct drm_crtc_state - mutable crtc state + * @enable: whether the CRTC should be enabled, gates all other state + * @mode: current mode timings + * @event: optional pointer to a DRM event to signal upon completion of the + * state update + * @state: backpointer to global drm_atomic_state + */ +struct drm_crtc_state { + bool enable : 1; + + struct drm_display_mode mode; + + struct drm_pending_vblank_event *event; + + struct drm_atomic_state *state; +}; /** * struct drm_crtc_funcs - control CRTCs for a given device @@ -238,6 +257,9 @@ struct drm_bridge; * @set_property: called when a property is changed * @set_config: apply a new CRTC configuration * @page_flip: initiate a page flip + * @atomic_duplicate_state: duplicate the atomic state for this CRTC + * @atomic_destroy_state: destroy an atomic state for this CRTC + * @atomic_set_property: set a property on an atomic state for this CRTC * * The drm_crtc_funcs structure is the central CRTC management structure * in the DRM. Each CRTC controls one or more connectors (note that the name @@ -288,6 +310,15 @@ struct drm_crtc_funcs { int (*set_property)(struct drm_crtc *crtc, struct drm_property *property, uint64_t val); + + /* atomic update handling */ + struct drm_crtc_state *(*atomic_duplicate_state)(struct drm_crtc *crtc); + void (*atomic_destroy_state)(struct drm_crtc *crtc, + struct drm_crtc_state *cstate); + int (*atomic_set_property)(struct drm_crtc *crtc, + struct drm_crtc_state *state, + struct drm_property *property, + uint64_t val); }; /** @@ -317,6 +348,7 @@ struct drm_crtc_funcs { * @pixeldur_ns: precise pixel timing * @helper_private: mid-layer private data * @properties: property tracking for this CRTC + * @state: current atomic state for this CRTC * @acquire_ctx: per-CRTC implicit acquire context used by atomic drivers for * legacy ioctls * @@ -374,6 +406,8 @@ struct drm_crtc { struct drm_object_properties properties; + struct drm_crtc_state *state; + /* * For legacy crtc ioctls so that atomic drivers can get at the locking * acquire context. @@ -381,6 +415,16 @@ struct drm_crtc { struct drm_modeset_acquire_ctx *acquire_ctx; }; +/** + * struct drm_connector_state - mutable connector state + * @crtc: crtc to connect connector to, NULL if disabled + * @state: backpointer to global drm_atomic_state + */ +struct drm_connector_state { + struct drm_crtc *crtc; + + struct drm_atomic_state *state; +}; /** * struct drm_connector_funcs - control connectors on a given device @@ -393,6 +437,10 @@ struct drm_crtc { * @set_property: property for this connector may need an update * @destroy: make object go away * @force: notify the driver that the connector is forced on + * @atomic_duplicate_state: duplicate the atomic state for this connector + * @atomic_destroy_state: destroy an atomic state for this connector + * @atomic_set_property: set a property on an atomic state for this connector + * * * Each CRTC may have one or more connectors attached to it. The functions * below allow the core DRM code to control connectors, enumerate available modes, @@ -417,6 +465,15 @@ struct drm_connector_funcs { uint64_t val); void (*destroy)(struct drm_connector *connector); void (*force)(struct drm_connector *connector); + + /* atomic update handling */ + struct drm_connector_state *(*atomic_duplicate_state)(struct drm_connector *connector); + void (*atomic_destroy_state)(struct drm_connector *connector, + struct drm_connector_state *cstate); + int (*atomic_set_property)(struct drm_connector *connector, + struct drm_connector_state *state, + struct drm_property *property, + uint64_t val); }; /** @@ -515,6 +572,7 @@ struct drm_encoder { * @null_edid_counter: track sinks that give us all zeros for the EDID * @bad_edid_counter: track sinks that give us an EDID with invalid checksum * @debugfs_entry: debugfs directory for this connector + * @state: current atomic state for this connector * * Each connector may be connected to one or more CRTCs, or may be clonable by * another connector if they can share a CRTC. Each connector also has a specific @@ -575,8 +633,42 @@ struct drm_connector { unsigned bad_edid_counter; struct dentry *debugfs_entry; + + struct drm_connector_state *state; +}; + +/** + * struct drm_plane_state - mutable plane state + * @crtc: currently bound CRTC, NULL if disabled + * @fb: currently bound fb + * @crtc_x: left position of visible portion of plane on crtc + * @crtc_y: upper position of visible portion of plane on crtc + * @crtc_w: width of visible portion of plane on crtc + * @crtc_h: height of visible portion of plane on crtc + * @src_x: left position of visible portion of plane within + * plane (in 16.16) + * @src_y: upper position of visible portion of plane within + * plane (in 16.16) + * @src_w: width of visible portion of plane (in 16.16) + * @src_h: height of visible portion of plane (in 16.16) + * @state: backpointer to global drm_atomic_state + */ +struct drm_plane_state { + struct drm_crtc *crtc; + struct drm_framebuffer *fb; + + /* Signed dest location allows it to be partially off screen */ + int32_t crtc_x, crtc_y; + uint32_t crtc_w, crtc_h; + + /* Source values are 16.16 fixed point */ + uint32_t src_x, src_y; + uint32_t src_h, src_w; + + struct drm_atomic_state *state; }; + /** * struct drm_plane_funcs - driver plane control functions * @update_plane: update the plane configuration @@ -584,6 +676,9 @@ struct drm_connector { * @destroy: clean up plane resources * @reset: reset plane after state has been invalidated (e.g. resume) * @set_property: called when a property is changed + * @atomic_duplicate_state: duplicate the atomic state for this plane + * @atomic_destroy_state: destroy an atomic state for this plane + * @atomic_set_property: set a property on an atomic state for this plane */ struct drm_plane_funcs { int (*update_plane)(struct drm_plane *plane, @@ -598,6 +693,15 @@ struct drm_plane_funcs { int (*set_property)(struct drm_plane *plane, struct drm_property *property, uint64_t val); + + /* atomic update handling */ + struct drm_plane_state *(*atomic_duplicate_state)(struct drm_plane *plane); + void (*atomic_destroy_state)(struct drm_plane *plane, + struct drm_plane_state *cstate); + int (*atomic_set_property)(struct drm_plane *plane, + struct drm_plane_state *state, + struct drm_property *property, + uint64_t val); }; enum drm_plane_type { @@ -621,6 +725,7 @@ enum drm_plane_type { * @funcs: helper functions * @properties: property tracking for this plane * @type: type of plane (overlay, primary, cursor) + * @state: current atomic state for this plane */ struct drm_plane { struct drm_device *dev; @@ -642,6 +747,8 @@ struct drm_plane { struct drm_object_properties properties; enum drm_plane_type type; + + struct drm_plane_state *state; }; /** -- cgit v0.10.2 From cc4ceb484b37b9369e0d4e8682b7ae1849ae4579 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 25 Jul 2014 21:30:38 +0200 Subject: drm: Global atomic state handling Some differences compared to Rob's patches again: - Dropped the committed and checked booleans. Checking will be internally enforced by always calling ->atomic_check before ->atomic_commit. And async handling needs to be solved differently because the current scheme completely side-steps ww mutex deadlock avoidance (and so either reinvents a new deadlock avoidance wheel or like the current code just deadlocks). - State for connectors needed to be added, since now they have a full-blown drm_connector_state (so that drivers have something to attach their own stuff to). - Refcounting is gone. I plane to solve async updates differently, since the lock-passing scheme doesn't cut it (since it abuses ww mutexes). Essentially what we need for async is a simple ownership transfer from the caller to the driver. That doesn't need full-blown refcounting. - The acquire ctx is a pointer. Real atomic callers should have that on their stack, legacy entry points need to put the right one (obtained by drm_modeset_legacy_acuire_ctx) in there. - I've dropped all hooks except check/commit. All the begin/end handling is done by core functions and is the same. - commit/check are just thin wrappers that ensure that ->check is always called. - To help out with locking in the legacy implementations I've added a helper to just grab all locks in the backoff case. v2: Add notices that check/commit can fail with EDEADLK. v3: - More consistent naming for state_alloc. - Add state_clear which is needed for backoff and retry. v4: Planes/connectors can switch between crtcs, and we need to be careful that we grab the state (and locks) for both the old and new crtc. Improve the interface functions to ensure this. v5: Add functions to grab affected connectors for a crtc and to recompute the crtc->enable state. This is useful for both helper and atomic ioctl code when e.g. removing a connector. v6: Squash in fixup from Fengguang to use ERR_CAST. v7: Add debug output. v8: Make checkpatch happy about kcalloc argument ordering. v9: Improve kerneldoc in drm_crtc.h v10: - Fix another kcalloc argument misorder I've missed. - More polish for kerneldoc. v11: Clarify the ownership rules for the state object. The new rule is that a successful drm_atomic_commit (whether synchronous or asnyc) always inherits the state and is responsible for the clean-up. That way async and sync ->commit functions are more similar. v12: A few bugfixes: - Assign state->state pointers correctly when grabbing state objects - we need to link them up with the global state. - Handle a NULL crtc in set_crtc_for_plane to simplify code flow a bit for the callers of this function. v13: Review from Sean: - kerneldoc spelling fixes - Don't overallocate states->planes. - Handle NULL crtc in set_crtc_for_connector. v14: Sprinkle __must_check over all functions which do wait/wound locking to make sure callers don't forget this. Since I have ;-) v15: Be more explicit in the kerneldoc when functions can return -EDEADLK what to do. And that every other -errno is fatal. v16: Indent with tabs instead of space, spotted by Ander. v17: Review from Thierry, small kerneldoc and other naming polish. Cc: Thierry Reding Cc: Ander Conselvan de Oliveira Cc: Daniel Thompson Cc: Fengguang Wu Cc: Sean Paul Cc: Matt Roper Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index f83d3b6..748513b 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -996,6 +996,10 @@ int max_width, max_height; !Edrivers/gpu/drm/drm_modes.c + Atomic Mode Setting Function Reference +!Edrivers/gpu/drm/drm_atomic.c + + Frame Buffer Creation struct drm_framebuffer *(*fb_create)(struct drm_device *dev, struct drm_file *file_priv, diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile index 9292a76..c5e37dc 100644 --- a/drivers/gpu/drm/Makefile +++ b/drivers/gpu/drm/Makefile @@ -14,7 +14,7 @@ drm-y := drm_auth.o drm_bufs.o drm_cache.o \ drm_info.o drm_debugfs.o drm_encoder_slave.o \ drm_trace_points.o drm_global.o drm_prime.o \ drm_rect.o drm_vma_manager.o drm_flip_work.o \ - drm_modeset_lock.o + drm_modeset_lock.o drm_atomic.o drm-$(CONFIG_COMPAT) += drm_ioc32.o drm-$(CONFIG_DRM_GEM_CMA_HELPER) += drm_gem_cma_helper.o diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c new file mode 100644 index 0000000..ad15a88 --- /dev/null +++ b/drivers/gpu/drm/drm_atomic.c @@ -0,0 +1,600 @@ +/* + * Copyright (C) 2014 Red Hat + * Copyright (C) 2014 Intel Corp. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Rob Clark + * Daniel Vetter + */ + + +#include +#include +#include + +static void kfree_state(struct drm_atomic_state *state) +{ + kfree(state->connectors); + kfree(state->connector_states); + kfree(state->crtcs); + kfree(state->crtc_states); + kfree(state->planes); + kfree(state->plane_states); + kfree(state); +} + +/** + * drm_atomic_state_alloc - allocate atomic state + * @dev: DRM device + * + * This allocates an empty atomic state to track updates. + */ +struct drm_atomic_state * +drm_atomic_state_alloc(struct drm_device *dev) +{ + struct drm_atomic_state *state; + + state = kzalloc(sizeof(*state), GFP_KERNEL); + if (!state) + return NULL; + + state->crtcs = kcalloc(dev->mode_config.num_crtc, + sizeof(*state->crtcs), GFP_KERNEL); + if (!state->crtcs) + goto fail; + state->crtc_states = kcalloc(dev->mode_config.num_crtc, + sizeof(*state->crtc_states), GFP_KERNEL); + if (!state->crtc_states) + goto fail; + state->planes = kcalloc(dev->mode_config.num_total_plane, + sizeof(*state->planes), GFP_KERNEL); + if (!state->planes) + goto fail; + state->plane_states = kcalloc(dev->mode_config.num_total_plane, + sizeof(*state->plane_states), GFP_KERNEL); + if (!state->plane_states) + goto fail; + state->connectors = kcalloc(dev->mode_config.num_connector, + sizeof(*state->connectors), + GFP_KERNEL); + if (!state->connectors) + goto fail; + state->connector_states = kcalloc(dev->mode_config.num_connector, + sizeof(*state->connector_states), + GFP_KERNEL); + if (!state->connector_states) + goto fail; + + state->dev = dev; + + DRM_DEBUG_KMS("Allocate atomic state %p\n", state); + + return state; +fail: + kfree_state(state); + + return NULL; +} +EXPORT_SYMBOL(drm_atomic_state_alloc); + +/** + * drm_atomic_state_clear - clear state object + * @state: atomic state + * + * When the w/w mutex algorithm detects a deadlock we need to back off and drop + * all locks. So someone else could sneak in and change the current modeset + * configuration. Which means that all the state assembled in @state is no + * longer an atomic update to the current state, but to some arbitrary earlier + * state. Which could break assumptions the driver's ->atomic_check likely + * relies on. + * + * Hence we must clear all cached state and completely start over, using this + * function. + */ +void drm_atomic_state_clear(struct drm_atomic_state *state) +{ + struct drm_device *dev = state->dev; + int i; + + DRM_DEBUG_KMS("Clearing atomic state %p\n", state); + + for (i = 0; i < dev->mode_config.num_connector; i++) { + struct drm_connector *connector = state->connectors[i]; + + if (!connector) + continue; + + connector->funcs->atomic_destroy_state(connector, + state->connector_states[i]); + } + + for (i = 0; i < dev->mode_config.num_crtc; i++) { + struct drm_crtc *crtc = state->crtcs[i]; + + if (!crtc) + continue; + + crtc->funcs->atomic_destroy_state(crtc, + state->crtc_states[i]); + } + + for (i = 0; i < dev->mode_config.num_total_plane; i++) { + struct drm_plane *plane = state->planes[i]; + + if (!plane) + continue; + + plane->funcs->atomic_destroy_state(plane, + state->plane_states[i]); + } +} +EXPORT_SYMBOL(drm_atomic_state_clear); + +/** + * drm_atomic_state_free - free all memory for an atomic state + * @state: atomic state to deallocate + * + * This frees all memory associated with an atomic state, including all the + * per-object state for planes, crtcs and connectors. + */ +void drm_atomic_state_free(struct drm_atomic_state *state) +{ + drm_atomic_state_clear(state); + + DRM_DEBUG_KMS("Freeing atomic state %p\n", state); + + kfree_state(state); +} +EXPORT_SYMBOL(drm_atomic_state_free); + +/** + * drm_atomic_get_crtc_state - get crtc state + * @state: global atomic state object + * @crtc: crtc to get state object for + * + * This function returns the crtc state for the given crtc, allocating it if + * needed. It will also grab the relevant crtc lock to make sure that the state + * is consistent. + * + * Returns: + * + * Either the allocated state or the error code encoded into the pointer. When + * the error is EDEADLK then the w/w mutex code has detected a deadlock and the + * entire atomic sequence must be restarted. All other errors are fatal. + */ +struct drm_crtc_state * +drm_atomic_get_crtc_state(struct drm_atomic_state *state, + struct drm_crtc *crtc) +{ + int ret, index; + struct drm_crtc_state *crtc_state; + + index = drm_crtc_index(crtc); + + if (state->crtc_states[index]) + return state->crtc_states[index]; + + ret = drm_modeset_lock(&crtc->mutex, state->acquire_ctx); + if (ret) + return ERR_PTR(ret); + + crtc_state = crtc->funcs->atomic_duplicate_state(crtc); + if (!crtc_state) + return ERR_PTR(-ENOMEM); + + state->crtc_states[index] = crtc_state; + state->crtcs[index] = crtc; + crtc_state->state = state; + + DRM_DEBUG_KMS("Added [CRTC:%d] %p state to %p\n", + crtc->base.id, crtc_state, state); + + return crtc_state; +} +EXPORT_SYMBOL(drm_atomic_get_crtc_state); + +/** + * drm_atomic_get_plane_state - get plane state + * @state: global atomic state object + * @plane: plane to get state object for + * + * This function returns the plane state for the given plane, allocating it if + * needed. It will also grab the relevant plane lock to make sure that the state + * is consistent. + * + * Returns: + * + * Either the allocated state or the error code encoded into the pointer. When + * the error is EDEADLK then the w/w mutex code has detected a deadlock and the + * entire atomic sequence must be restarted. All other errors are fatal. + */ +struct drm_plane_state * +drm_atomic_get_plane_state(struct drm_atomic_state *state, + struct drm_plane *plane) +{ + int ret, index; + struct drm_plane_state *plane_state; + + index = drm_plane_index(plane); + + if (state->plane_states[index]) + return state->plane_states[index]; + + /* + * TODO: We currently don't have per-plane mutexes. So instead of trying + * crazy tricks with deferring plane->crtc and hoping for the best just + * grab all crtc locks. Once we have per-plane locks we must update this + * to only take the plane mutex. + */ + ret = drm_modeset_lock_all_crtcs(state->dev, state->acquire_ctx); + if (ret) + return ERR_PTR(ret); + + plane_state = plane->funcs->atomic_duplicate_state(plane); + if (!plane_state) + return ERR_PTR(-ENOMEM); + + state->plane_states[index] = plane_state; + state->planes[index] = plane; + plane_state->state = state; + + DRM_DEBUG_KMS("Added [PLANE:%d] %p state to %p\n", + plane->base.id, plane_state, state); + + if (plane_state->crtc) { + struct drm_crtc_state *crtc_state; + + crtc_state = drm_atomic_get_crtc_state(state, + plane_state->crtc); + if (IS_ERR(crtc_state)) + return ERR_CAST(crtc_state); + } + + return plane_state; +} +EXPORT_SYMBOL(drm_atomic_get_plane_state); + +/** + * drm_atomic_get_connector_state - get connector state + * @state: global atomic state object + * @connector: connector to get state object for + * + * This function returns the connector state for the given connector, + * allocating it if needed. It will also grab the relevant connector lock to + * make sure that the state is consistent. + * + * Returns: + * + * Either the allocated state or the error code encoded into the pointer. When + * the error is EDEADLK then the w/w mutex code has detected a deadlock and the + * entire atomic sequence must be restarted. All other errors are fatal. + */ +struct drm_connector_state * +drm_atomic_get_connector_state(struct drm_atomic_state *state, + struct drm_connector *connector) +{ + int ret, index; + struct drm_mode_config *config = &connector->dev->mode_config; + struct drm_connector_state *connector_state; + + index = drm_connector_index(connector); + + if (state->connector_states[index]) + return state->connector_states[index]; + + ret = drm_modeset_lock(&config->connection_mutex, state->acquire_ctx); + if (ret) + return ERR_PTR(ret); + + connector_state = connector->funcs->atomic_duplicate_state(connector); + if (!connector_state) + return ERR_PTR(-ENOMEM); + + state->connector_states[index] = connector_state; + state->connectors[index] = connector; + connector_state->state = state; + + DRM_DEBUG_KMS("Added [CONNECTOR:%d] %p state to %p\n", + connector->base.id, connector_state, state); + + if (connector_state->crtc) { + struct drm_crtc_state *crtc_state; + + crtc_state = drm_atomic_get_crtc_state(state, + connector_state->crtc); + if (IS_ERR(crtc_state)) + return ERR_CAST(crtc_state); + } + + return connector_state; +} +EXPORT_SYMBOL(drm_atomic_get_connector_state); + +/** + * drm_atomic_set_crtc_for_plane - set crtc for plane + * @plane_state: atomic state object for the plane + * @crtc: crtc to use for the plane + * + * Changing the assigned crtc for a plane requires us to grab the lock and state + * for the new crtc, as needed. This function takes care of all these details + * besides updating the pointer in the state object itself. + * + * Returns: + * 0 on success or can fail with -EDEADLK or -ENOMEM. When the error is EDEADLK + * then the w/w mutex code has detected a deadlock and the entire atomic + * sequence must be restarted. All other errors are fatal. + */ +int +drm_atomic_set_crtc_for_plane(struct drm_plane_state *plane_state, + struct drm_crtc *crtc) +{ + struct drm_crtc_state *crtc_state; + + if (crtc) { + crtc_state = drm_atomic_get_crtc_state(plane_state->state, + crtc); + if (IS_ERR(crtc_state)) + return PTR_ERR(crtc_state); + } + + plane_state->crtc = crtc; + + if (crtc) + DRM_DEBUG_KMS("Link plane state %p to [CRTC:%d]\n", + plane_state, crtc->base.id); + else + DRM_DEBUG_KMS("Link plane state %p to [NOCRTC]\n", plane_state); + + return 0; +} +EXPORT_SYMBOL(drm_atomic_set_crtc_for_plane); + +/** + * drm_atomic_set_crtc_for_connector - set crtc for connector + * @conn_state: atomic state object for the connector + * @crtc: crtc to use for the connector + * + * Changing the assigned crtc for a connector requires us to grab the lock and + * state for the new crtc, as needed. This function takes care of all these + * details besides updating the pointer in the state object itself. + * + * Returns: + * 0 on success or can fail with -EDEADLK or -ENOMEM. When the error is EDEADLK + * then the w/w mutex code has detected a deadlock and the entire atomic + * sequence must be restarted. All other errors are fatal. + */ +int +drm_atomic_set_crtc_for_connector(struct drm_connector_state *conn_state, + struct drm_crtc *crtc) +{ + struct drm_crtc_state *crtc_state; + + if (crtc) { + crtc_state = drm_atomic_get_crtc_state(conn_state->state, crtc); + if (IS_ERR(crtc_state)) + return PTR_ERR(crtc_state); + } + + conn_state->crtc = crtc; + + if (crtc) + DRM_DEBUG_KMS("Link connector state %p to [CRTC:%d]\n", + conn_state, crtc->base.id); + else + DRM_DEBUG_KMS("Link connector state %p to [NOCRTC]\n", + conn_state); + + return 0; +} +EXPORT_SYMBOL(drm_atomic_set_crtc_for_connector); + +/** + * drm_atomic_add_affected_connectors - add connectors for crtc + * @state: atomic state + * @crtc: DRM crtc + * + * This function walks the current configuration and adds all connectors + * currently using @crtc to the atomic configuration @state. Note that this + * function must acquire the connection mutex. This can potentially cause + * unneeded seralization if the update is just for the planes on one crtc. Hence + * drivers and helpers should only call this when really needed (e.g. when a + * full modeset needs to happen due to some change). + * + * Returns: + * 0 on success or can fail with -EDEADLK or -ENOMEM. When the error is EDEADLK + * then the w/w mutex code has detected a deadlock and the entire atomic + * sequence must be restarted. All other errors are fatal. + */ +int +drm_atomic_add_affected_connectors(struct drm_atomic_state *state, + struct drm_crtc *crtc) +{ + struct drm_mode_config *config = &state->dev->mode_config; + struct drm_connector *connector; + struct drm_connector_state *conn_state; + int ret; + + ret = drm_modeset_lock(&config->connection_mutex, state->acquire_ctx); + if (ret) + return ret; + + DRM_DEBUG_KMS("Adding all current connectors for [CRTC:%d] to %p\n", + crtc->base.id, state); + + /* + * Changed connectors are already in @state, so only need to look at the + * current configuration. + */ + list_for_each_entry(connector, &config->connector_list, head) { + if (connector->state->crtc != crtc) + continue; + + conn_state = drm_atomic_get_connector_state(state, connector); + if (IS_ERR(conn_state)) + return PTR_ERR(conn_state); + } + + return 0; +} +EXPORT_SYMBOL(drm_atomic_add_affected_connectors); + +/** + * drm_atomic_connectors_for_crtc - count number of connected outputs + * @state: atomic state + * @crtc: DRM crtc + * + * This function counts all connectors which will be connected to @crtc + * according to @state. Useful to recompute the enable state for @crtc. + */ +int +drm_atomic_connectors_for_crtc(struct drm_atomic_state *state, + struct drm_crtc *crtc) +{ + int nconnectors = state->dev->mode_config.num_connector; + int i, num_connected_connectors = 0; + + for (i = 0; i < nconnectors; i++) { + struct drm_connector_state *conn_state; + + conn_state = state->connector_states[i]; + + if (conn_state && conn_state->crtc == crtc) + num_connected_connectors++; + } + + DRM_DEBUG_KMS("State %p has %i connectors for [CRTC:%d]\n", + state, num_connected_connectors, crtc->base.id); + + return num_connected_connectors; +} +EXPORT_SYMBOL(drm_atomic_connectors_for_crtc); + +/** + * drm_atomic_legacy_backoff - locking backoff for legacy ioctls + * @state: atomic state + * + * This function should be used by legacy entry points which don't understand + * -EDEADLK semantics. For simplicity this one will grab all modeset locks after + * the slowpath completed. + */ +void drm_atomic_legacy_backoff(struct drm_atomic_state *state) +{ + int ret; + +retry: + drm_modeset_backoff(state->acquire_ctx); + + ret = drm_modeset_lock(&state->dev->mode_config.connection_mutex, + state->acquire_ctx); + if (ret) + goto retry; + ret = drm_modeset_lock_all_crtcs(state->dev, + state->acquire_ctx); + if (ret) + goto retry; +} +EXPORT_SYMBOL(drm_atomic_legacy_backoff); + +/** + * drm_atomic_check_only - check whether a given config would work + * @state: atomic configuration to check + * + * Note that this function can return -EDEADLK if the driver needed to acquire + * more locks but encountered a deadlock. The caller must then do the usual w/w + * backoff dance and restart. All other errors are fatal. + * + * Returns: + * 0 on success, negative error code on failure. + */ +int drm_atomic_check_only(struct drm_atomic_state *state) +{ + struct drm_mode_config *config = &state->dev->mode_config; + + DRM_DEBUG_KMS("checking %p\n", state); + + if (config->funcs->atomic_check) + return config->funcs->atomic_check(state->dev, state); + else + return 0; +} +EXPORT_SYMBOL(drm_atomic_check_only); + +/** + * drm_atomic_commit - commit configuration atomically + * @state: atomic configuration to check + * + * Note that this function can return -EDEADLK if the driver needed to acquire + * more locks but encountered a deadlock. The caller must then do the usual w/w + * backoff dance and restart. All other errors are fatal. + * + * Also note that on successful execution ownership of @state is transferred + * from the caller of this function to the function itself. The caller must not + * free or in any other way access @state. If the function fails then the caller + * must clean up @state itself. + * + * Returns: + * 0 on success, negative error code on failure. + */ +int drm_atomic_commit(struct drm_atomic_state *state) +{ + struct drm_mode_config *config = &state->dev->mode_config; + int ret; + + ret = drm_atomic_check_only(state); + if (ret) + return ret; + + DRM_DEBUG_KMS("commiting %p\n", state); + + return config->funcs->atomic_commit(state->dev, state, false); +} +EXPORT_SYMBOL(drm_atomic_commit); + +/** + * drm_atomic_async_commit - atomic&async configuration commit + * @state: atomic configuration to check + * + * Note that this function can return -EDEADLK if the driver needed to acquire + * more locks but encountered a deadlock. The caller must then do the usual w/w + * backoff dance and restart. All other errors are fatal. + * + * Also note that on successful execution ownership of @state is transferred + * from the caller of this function to the function itself. The caller must not + * free or in any other way access @state. If the function fails then the caller + * must clean up @state itself. + * + * Returns: + * 0 on success, negative error code on failure. + */ +int drm_atomic_async_commit(struct drm_atomic_state *state) +{ + struct drm_mode_config *config = &state->dev->mode_config; + int ret; + + ret = drm_atomic_check_only(state); + if (ret) + return ret; + + DRM_DEBUG_KMS("commiting %p asynchronously\n", state); + + return config->funcs->atomic_commit(state->dev, state, true); +} +EXPORT_SYMBOL(drm_atomic_async_commit); diff --git a/include/drm/drm_atomic.h b/include/drm/drm_atomic.h new file mode 100644 index 0000000..5bb15f5 --- /dev/null +++ b/include/drm/drm_atomic.h @@ -0,0 +1,65 @@ +/* + * Copyright (C) 2014 Red Hat + * Copyright (C) 2014 Intel Corp. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Rob Clark + * Daniel Vetter + */ + +#ifndef DRM_ATOMIC_H_ +#define DRM_ATOMIC_H_ + +struct drm_atomic_state * __must_check +drm_atomic_state_alloc(struct drm_device *dev); +void drm_atomic_state_clear(struct drm_atomic_state *state); +void drm_atomic_state_free(struct drm_atomic_state *state); + +struct drm_crtc_state * __must_check +drm_atomic_get_crtc_state(struct drm_atomic_state *state, + struct drm_crtc *crtc); +struct drm_plane_state * __must_check +drm_atomic_get_plane_state(struct drm_atomic_state *state, + struct drm_plane *plane); +struct drm_connector_state * __must_check +drm_atomic_get_connector_state(struct drm_atomic_state *state, + struct drm_connector *connector); + +int __must_check +drm_atomic_set_crtc_for_plane(struct drm_plane_state *plane_state, + struct drm_crtc *crtc); +int __must_check +drm_atomic_set_crtc_for_connector(struct drm_connector_state *conn_state, + struct drm_crtc *crtc); +int __must_check +drm_atomic_add_affected_connectors(struct drm_atomic_state *state, + struct drm_crtc *crtc); +int +drm_atomic_connectors_for_crtc(struct drm_atomic_state *state, + struct drm_crtc *crtc); + +void drm_atomic_legacy_backoff(struct drm_atomic_state *state); + +int __must_check drm_atomic_check_only(struct drm_atomic_state *state); +int __must_check drm_atomic_commit(struct drm_atomic_state *state); +int __must_check drm_atomic_async_commit(struct drm_atomic_state *state); + +#endif /* DRM_ATOMIC_H_ */ diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 3554868..c7a3400 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -227,7 +227,7 @@ struct drm_bridge; struct drm_atomic_state; /** - * struct drm_crtc_state - mutable crtc state + * struct drm_crtc_state - mutable CRTC state * @enable: whether the CRTC should be enabled, gates all other state * @mode: current mode timings * @event: optional pointer to a DRM event to signal upon completion of the @@ -235,7 +235,7 @@ struct drm_atomic_state; * @state: backpointer to global drm_atomic_state */ struct drm_crtc_state { - bool enable : 1; + bool enable; struct drm_display_mode mode; @@ -314,7 +314,7 @@ struct drm_crtc_funcs { /* atomic update handling */ struct drm_crtc_state *(*atomic_duplicate_state)(struct drm_crtc *crtc); void (*atomic_destroy_state)(struct drm_crtc *crtc, - struct drm_crtc_state *cstate); + struct drm_crtc_state *state); int (*atomic_set_property)(struct drm_crtc *crtc, struct drm_crtc_state *state, struct drm_property *property, @@ -417,7 +417,7 @@ struct drm_crtc { /** * struct drm_connector_state - mutable connector state - * @crtc: crtc to connect connector to, NULL if disabled + * @crtc: CRTC to connect connector to, NULL if disabled * @state: backpointer to global drm_atomic_state */ struct drm_connector_state { @@ -441,7 +441,6 @@ struct drm_connector_state { * @atomic_destroy_state: destroy an atomic state for this connector * @atomic_set_property: set a property on an atomic state for this connector * - * * Each CRTC may have one or more connectors attached to it. The functions * below allow the core DRM code to control connectors, enumerate available modes, * etc. @@ -469,7 +468,7 @@ struct drm_connector_funcs { /* atomic update handling */ struct drm_connector_state *(*atomic_duplicate_state)(struct drm_connector *connector); void (*atomic_destroy_state)(struct drm_connector *connector, - struct drm_connector_state *cstate); + struct drm_connector_state *state); int (*atomic_set_property)(struct drm_connector *connector, struct drm_connector_state *state, struct drm_property *property, @@ -640,7 +639,7 @@ struct drm_connector { /** * struct drm_plane_state - mutable plane state * @crtc: currently bound CRTC, NULL if disabled - * @fb: currently bound fb + * @fb: currently bound framebuffer * @crtc_x: left position of visible portion of plane on crtc * @crtc_y: upper position of visible portion of plane on crtc * @crtc_w: width of visible portion of plane on crtc @@ -697,7 +696,7 @@ struct drm_plane_funcs { /* atomic update handling */ struct drm_plane_state *(*atomic_duplicate_state)(struct drm_plane *plane); void (*atomic_destroy_state)(struct drm_plane *plane, - struct drm_plane_state *cstate); + struct drm_plane_state *state); int (*atomic_set_property)(struct drm_plane *plane, struct drm_plane_state *state, struct drm_property *property, @@ -794,6 +793,32 @@ struct drm_bridge { }; /** + * struct struct drm_atomic_state - the global state object for atomic updates + * @dev: parent DRM device + * @flags: state flags like async update + * @planes: pointer to array of plane pointers + * @plane_states: pointer to array of plane states pointers + * @crtcs: pointer to array of CRTC pointers + * @crtc_states: pointer to array of CRTC states pointers + * @connectors: pointer to array of connector pointers + * @connector_states: pointer to array of connector states pointers + * @acquire_ctx: acquire context for this atomic modeset state update + */ +struct drm_atomic_state { + struct drm_device *dev; + uint32_t flags; + struct drm_plane **planes; + struct drm_plane_state **plane_states; + struct drm_crtc **crtcs; + struct drm_crtc_state **crtc_states; + struct drm_connector **connectors; + struct drm_connector_state **connector_states; + + struct drm_modeset_acquire_ctx *acquire_ctx; +}; + + +/** * struct drm_mode_set - new values for a CRTC config change * @fb: framebuffer to use for new config * @crtc: CRTC whose configuration we're about to change @@ -824,6 +849,9 @@ struct drm_mode_set { * struct drm_mode_config_funcs - basic driver provided mode setting functions * @fb_create: create a new framebuffer object * @output_poll_changed: function to handle output configuration changes + * @atomic_check: check whether a give atomic state update is possible + * @atomic_commit: commit an atomic state update previously verified with + * atomic_check() * * Some global (i.e. not per-CRTC, connector, etc) mode setting functions that * involve drivers. @@ -833,6 +861,12 @@ struct drm_mode_config_funcs { struct drm_file *file_priv, struct drm_mode_fb_cmd2 *mode_cmd); void (*output_poll_changed)(struct drm_device *dev); + + int (*atomic_check)(struct drm_device *dev, + struct drm_atomic_state *a); + int (*atomic_commit)(struct drm_device *dev, + struct drm_atomic_state *a, + bool async); }; /** -- cgit v0.10.2 From c2fcd274bce521cd824d96d7f26dfa57ea06478c Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 5 Nov 2014 00:14:14 +0100 Subject: drm: Add atomic/plane helpers This is the first cut of atomic helper code. As-is it's only useful to implement a pure atomic interface for plane updates. Later patches will integrate this with the crtc helpers so that full atomic updates are possible. We also need a pile of helpers to aid drivers in transitioning from the legacy world to the shiny new atomic age. Finally we need helpers to implement legacy ioctls on top of the atomic interface. The design of the overall helpers<->driver interaction is fairly simple, but has an unfortunate large interface: - We have ->atomic_check callbacks for crtcs and planes. The idea is that connectors don't need any checking, and if they do they can adjust the relevant crtc driver-private state. So no connector hooks should be needed. Also the crtc helpers integration will do the ->best_encoder checks, so no need for that. - Framebuffer pinning needs to be done before we can commit to the hw state. This is especially important for async updates where we must pin all buffers before returning to userspace, so that really only hw failures can happen in the asynchronous worker. Hence we add ->prepare_fb and ->cleanup_fb hooks for this resources management. - The actual atomic plane commit can't fail (except hw woes), so has void return type. It has three stages: 1. Prepare all affected crtcs with crtc->atomic_begin. Drivers can use this to unset the GO bit or similar latches to prevent plane updates. 2. Update plane state by looping over all changed planes and calling plane->atomic_update. Presuming the hardware is sane and has GO bits drivers can simply bash the state into the hardware in this function. Other drivers might use this to precompute hw state for the final step. 3. Finally latch the update for the next vblank with crtc->atomic_flush. Note that this function doesn't need to wait for the vblank to happen even for the synchronous case. v2: Clear drm__state->state to NULL when swapping in state. v3: Add TODO that we don't short-circuit plane updates for now. Likely no one will care. v4: Squash in a bit of polish that somehow landed in the wrong (later) patche. v5: Integrate atomic functions into the drm docbook and fixup the kerneldoc. v6: Fixup fixup patch squashing fumble. v7: Don't touch the legacy plane state plane->fb and plane->crtc. This is only used by the legacy ioctl code in the drm core, and that code already takes care of updating the pointers in all relevant cases. This is in stark contrast to connector->encoder->crtc links on the modeset side, which we still need to set since the core doesn't touch them. Also some more kerneldoc polish. v8: Drop outdated comment. v9: Handle the state->state pointer correctly: Only clearing the ->state pointer when assigning the state to the kms object isn't good enough. We also need to re-link the swapped out state into the drm_atomic_state structure. v10: Shuffle the misplaced docbook template hunk around that Sean spotted. Cc: Sean Paul Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile index c5e37dc..c3cf64c 100644 --- a/drivers/gpu/drm/Makefile +++ b/drivers/gpu/drm/Makefile @@ -23,7 +23,7 @@ drm-$(CONFIG_DRM_PANEL) += drm_panel.o drm-$(CONFIG_OF) += drm_of.o drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \ - drm_plane_helper.o drm_dp_mst_topology.o + drm_plane_helper.o drm_dp_mst_topology.o drm_atomic_helper.o drm_kms_helper-$(CONFIG_DRM_LOAD_EDID_FIRMWARE) += drm_edid_load.o drm_kms_helper-$(CONFIG_DRM_KMS_FB_HELPER) += drm_fb_helper.o drm_kms_helper-$(CONFIG_DRM_KMS_CMA_HELPER) += drm_fb_cma_helper.o diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c new file mode 100644 index 0000000..55a8eb2 --- /dev/null +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -0,0 +1,355 @@ +/* + * Copyright (C) 2014 Red Hat + * Copyright (C) 2014 Intel Corp. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Rob Clark + * Daniel Vetter + */ + +#include +#include +#include +#include + +static void +drm_atomic_helper_plane_changed(struct drm_atomic_state *state, + struct drm_plane_state *plane_state, + struct drm_plane *plane) +{ + struct drm_crtc_state *crtc_state; + + if (plane->state->crtc) { + crtc_state = state->crtc_states[drm_crtc_index(plane->crtc)]; + + if (WARN_ON(!crtc_state)) + return; + + crtc_state->planes_changed = true; + } + + if (plane_state->crtc) { + crtc_state = + state->crtc_states[drm_crtc_index(plane_state->crtc)]; + + if (WARN_ON(!crtc_state)) + return; + + crtc_state->planes_changed = true; + } +} + +/** + * drm_atomic_helper_check - validate state object + * @dev: DRM device + * @state: the driver state object + * + * Check the state object to see if the requested state is physically possible. + * Only crtcs and planes have check callbacks, so for any additional (global) + * checking that a driver needs it can simply wrap that around this function. + * Drivers without such needs can directly use this as their ->atomic_check() + * callback. + * + * RETURNS + * Zero for success or -errno + */ +int drm_atomic_helper_check(struct drm_device *dev, + struct drm_atomic_state *state) +{ + int nplanes = dev->mode_config.num_total_plane; + int ncrtcs = dev->mode_config.num_crtc; + int i, ret = 0; + + for (i = 0; i < nplanes; i++) { + struct drm_plane_helper_funcs *funcs; + struct drm_plane *plane = state->planes[i]; + struct drm_plane_state *plane_state = state->plane_states[i]; + + if (!plane) + continue; + + funcs = plane->helper_private; + + drm_atomic_helper_plane_changed(state, plane_state, plane); + + if (!funcs || !funcs->atomic_check) + continue; + + ret = funcs->atomic_check(plane, plane_state); + if (ret) { + DRM_DEBUG_KMS("[PLANE:%d] atomic check failed\n", + plane->base.id); + return ret; + } + } + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc_helper_funcs *funcs; + struct drm_crtc *crtc = state->crtcs[i]; + + if (!crtc) + continue; + + funcs = crtc->helper_private; + + if (!funcs || !funcs->atomic_check) + continue; + + ret = funcs->atomic_check(crtc, state->crtc_states[i]); + if (ret) { + DRM_DEBUG_KMS("[CRTC:%d] atomic check failed\n", + crtc->base.id); + return ret; + } + } + + return ret; +} +EXPORT_SYMBOL(drm_atomic_helper_check); + +/** + * drm_atomic_helper_prepare_planes - prepare plane resources after commit + * @dev: DRM device + * @state: atomic state object with old state structures + * + * This function prepares plane state, specifically framebuffers, for the new + * configuration. If any failure is encountered this function will call + * ->cleanup_fb on any already successfully prepared framebuffer. + * + * Returns: + * 0 on success, negative error code on failure. + */ +int drm_atomic_helper_prepare_planes(struct drm_device *dev, + struct drm_atomic_state *state) +{ + int nplanes = dev->mode_config.num_total_plane; + int ret, i; + + for (i = 0; i < nplanes; i++) { + struct drm_plane_helper_funcs *funcs; + struct drm_plane *plane = state->planes[i]; + struct drm_framebuffer *fb; + + if (!plane) + continue; + + funcs = plane->helper_private; + + fb = state->plane_states[i]->fb; + + if (fb && funcs->prepare_fb) { + ret = funcs->prepare_fb(plane, fb); + if (ret) + goto fail; + } + } + + return 0; + +fail: + for (i--; i >= 0; i--) { + struct drm_plane_helper_funcs *funcs; + struct drm_plane *plane = state->planes[i]; + struct drm_framebuffer *fb; + + if (!plane) + continue; + + funcs = plane->helper_private; + + fb = state->plane_states[i]->fb; + + if (fb && funcs->cleanup_fb) + funcs->cleanup_fb(plane, fb); + + } + + return ret; +} +EXPORT_SYMBOL(drm_atomic_helper_prepare_planes); + +/** + * drm_atomic_helper_commit_planes - commit plane state + * @dev: DRM device + * @state: atomic state + * + * This function commits the new plane state using the plane and atomic helper + * functions for planes and crtcs. It assumes that the atomic state has already + * been pushed into the relevant object state pointers, since this step can no + * longer fail. + * + * It still requires the global state object @state to know which planes and + * crtcs need to be updated though. + */ +void drm_atomic_helper_commit_planes(struct drm_device *dev, + struct drm_atomic_state *state) +{ + int nplanes = dev->mode_config.num_total_plane; + int ncrtcs = dev->mode_config.num_crtc; + int i; + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc_helper_funcs *funcs; + struct drm_crtc *crtc = state->crtcs[i]; + + if (!crtc) + continue; + + funcs = crtc->helper_private; + + if (!funcs || !funcs->atomic_begin) + continue; + + funcs->atomic_begin(crtc); + } + + for (i = 0; i < nplanes; i++) { + struct drm_plane_helper_funcs *funcs; + struct drm_plane *plane = state->planes[i]; + + if (!plane) + continue; + + funcs = plane->helper_private; + + if (!funcs || !funcs->atomic_update) + continue; + + funcs->atomic_update(plane); + } + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc_helper_funcs *funcs; + struct drm_crtc *crtc = state->crtcs[i]; + + if (!crtc) + continue; + + funcs = crtc->helper_private; + + if (!funcs || !funcs->atomic_flush) + continue; + + funcs->atomic_flush(crtc); + } +} +EXPORT_SYMBOL(drm_atomic_helper_commit_planes); + +/** + * drm_atomic_helper_cleanup_planes - cleanup plane resources after commit + * @dev: DRM device + * @old_state: atomic state object with old state structures + * + * This function cleans up plane state, specifically framebuffers, from the old + * configuration. Hence the old configuration must be perserved in @old_state to + * be able to call this function. + * + * This function must also be called on the new state when the atomic update + * fails at any point after calling drm_atomic_helper_prepare_planes(). + */ +void drm_atomic_helper_cleanup_planes(struct drm_device *dev, + struct drm_atomic_state *old_state) +{ + int nplanes = dev->mode_config.num_total_plane; + int i; + + for (i = 0; i < nplanes; i++) { + struct drm_plane_helper_funcs *funcs; + struct drm_plane *plane = old_state->planes[i]; + struct drm_framebuffer *old_fb; + + if (!plane) + continue; + + funcs = plane->helper_private; + + old_fb = old_state->plane_states[i]->fb; + + if (old_fb && funcs->cleanup_fb) + funcs->cleanup_fb(plane, old_fb); + } +} +EXPORT_SYMBOL(drm_atomic_helper_cleanup_planes); + +/** + * drm_atomic_helper_swap_state - store atomic state into current sw state + * @dev: DRM device + * @state: atomic state + * + * This function stores the atomic state into the current state pointers in all + * driver objects. It should be called after all failing steps have been done + * and succeeded, but before the actual hardware state is committed. + * + * For cleanup and error recovery the current state for all changed objects will + * be swaped into @state. + * + * With that sequence it fits perfectly into the plane prepare/cleanup sequence: + * + * 1. Call drm_atomic_helper_prepare_planes() with the staged atomic state. + * + * 2. Do any other steps that might fail. + * + * 3. Put the staged state into the current state pointers with this function. + * + * 4. Actually commit the hardware state. + * + * 5. Call drm_atomic_helper_cleanup_planes with @state, which since step 3 + * contains the old state. Also do any other cleanup required with that state. + */ +void drm_atomic_helper_swap_state(struct drm_device *dev, + struct drm_atomic_state *state) +{ + int i; + + for (i = 0; i < dev->mode_config.num_connector; i++) { + struct drm_connector *connector = state->connectors[i]; + + if (!connector) + continue; + + connector->state->state = state; + swap(state->connector_states[i], connector->state); + connector->state->state = NULL; + } + + for (i = 0; i < dev->mode_config.num_crtc; i++) { + struct drm_crtc *crtc = state->crtcs[i]; + + if (!crtc) + continue; + + crtc->state->state = state; + swap(state->crtc_states[i], crtc->state); + crtc->state->state = NULL; + } + + for (i = 0; i < dev->mode_config.num_total_plane; i++) { + struct drm_plane *plane = state->planes[i]; + + if (!plane) + continue; + + plane->state->state = state; + swap(state->plane_states[i], plane->state); + plane->state->state = NULL; + } +} +EXPORT_SYMBOL(drm_atomic_helper_swap_state); diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h new file mode 100644 index 0000000..79938c6 --- /dev/null +++ b/include/drm/drm_atomic_helper.h @@ -0,0 +1,44 @@ +/* + * Copyright (C) 2014 Red Hat + * Copyright (C) 2014 Intel Corp. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Rob Clark + * Daniel Vetter + */ + +#ifndef DRM_ATOMIC_HELPER_H_ +#define DRM_ATOMIC_HELPER_H_ + +int drm_atomic_helper_check(struct drm_device *dev, + struct drm_atomic_state *state); + +int drm_atomic_helper_prepare_planes(struct drm_device *dev, + struct drm_atomic_state *state); +void drm_atomic_helper_commit_planes(struct drm_device *dev, + struct drm_atomic_state *state); +void drm_atomic_helper_cleanup_planes(struct drm_device *dev, + struct drm_atomic_state *old_state); + +void drm_atomic_helper_swap_state(struct drm_device *dev, + struct drm_atomic_state *state); + +#endif /* DRM_ATOMIC_HELPER_H_ */ diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index c7a3400..5614740 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -229,6 +229,7 @@ struct drm_atomic_state; /** * struct drm_crtc_state - mutable CRTC state * @enable: whether the CRTC should be enabled, gates all other state + * @planes_changed: for use by helpers and drivers when computing state updates * @mode: current mode timings * @event: optional pointer to a DRM event to signal upon completion of the * state update @@ -237,6 +238,9 @@ struct drm_atomic_state; struct drm_crtc_state { bool enable; + /* computed state bits used by helpers and drivers */ + bool planes_changed : 1; + struct drm_display_mode mode; struct drm_pending_vblank_event *event; @@ -747,6 +751,8 @@ struct drm_plane { enum drm_plane_type type; + void *helper_private; + struct drm_plane_state *state; }; diff --git a/include/drm/drm_crtc_helper.h b/include/drm/drm_crtc_helper.h index a3d75fe..adec48b 100644 --- a/include/drm/drm_crtc_helper.h +++ b/include/drm/drm_crtc_helper.h @@ -81,6 +81,12 @@ struct drm_crtc_helper_funcs { /* disable crtc when not in use - more explicit than dpms off */ void (*disable)(struct drm_crtc *crtc); + + /* atomic helpers */ + int (*atomic_check)(struct drm_crtc *crtc, + struct drm_crtc_state *state); + void (*atomic_begin)(struct drm_crtc *crtc); + void (*atomic_flush)(struct drm_crtc *crtc); }; /** diff --git a/include/drm/drm_plane_helper.h b/include/drm/drm_plane_helper.h index fa60d19..8577ec8 100644 --- a/include/drm/drm_plane_helper.h +++ b/include/drm/drm_plane_helper.h @@ -47,6 +47,28 @@ extern int drm_crtc_init(struct drm_device *dev, struct drm_crtc *crtc, const struct drm_crtc_funcs *funcs); +/** + * drm_plane_helper_funcs - helper operations for CRTCs + * + * The helper operations are called by the mid-layer CRTC helper. + */ +struct drm_plane_helper_funcs { + int (*prepare_fb)(struct drm_plane *plane, + struct drm_framebuffer *fb); + void (*cleanup_fb)(struct drm_plane *plane, + struct drm_framebuffer *fb); + + int (*atomic_check)(struct drm_plane *plane, + struct drm_plane_state *state); + void (*atomic_update)(struct drm_plane *plane); +}; + +static inline void drm_plane_helper_add(struct drm_plane *plane, + const struct drm_plane_helper_funcs *funcs) +{ + plane->helper_private = (void *)funcs; +} + extern int drm_plane_helper_check_update(struct drm_plane *plane, struct drm_crtc *crtc, struct drm_framebuffer *fb, -- cgit v0.10.2 From acf24a395c5a9290189b080383564437101d411c Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 29 Jul 2014 15:33:05 +0200 Subject: drm/plane-helper: transitional atomic plane helpers Converting a driver to the atomic interface can be a daunting undertaking. One of the prerequisites is to have full universal planes support. To make that transition a bit easier this patch provides plane helpers which use the new atomic helper callbacks just only for the plane changes. This way the plane update functionality can be tested without being forced to convert everything at once. Of course a real atomic update capable driver will implement the all plane properties through the atomic interface, so these helpers are mostly transitional. But they can be used to enable proper universal plane support, especially once the crtc helpers have also been adapted. v2: Use ->atomic_duplicate_state if available. v3: Don't forget to call ->atomic_destroy_state if available. v4: Fixup kerneldoc, reported by Paulo. v5: Extract a common plane_commit helper and fix some bugs in the plane_state setup of the plane_disable implementation. v6: Fix issues with the cleanup of the old fb. Since transitional helpers can be mixed we need to assume that the old fb has been set up by a legacy path (e.g. set_config or page_flip when the primary plane is converted to use these functions already). Hence pass an additional old_fb parameter to plane_commit to do that cleanup work correctly. v7: - Fix spurious WARNING (crtc helpers really love to disable stuff harder) and fix array index bonghits. - Correctly handle the lack of plane->state object, necessary for transitional use. - Don't indicate failure if drm_vblank_get doesn't work - that's expected when the pipe is in dpms off mode. v8: Review from Sean: - s/fail/out/ to make the meaning of a label more clear. - spelling fix in the commit message. Cc: Paulo Zanoni Cc: Sean Paul Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_plane_helper.c b/drivers/gpu/drm/drm_plane_helper.c index 827ec1a..a202b89 100644 --- a/drivers/gpu/drm/drm_plane_helper.c +++ b/drivers/gpu/drm/drm_plane_helper.c @@ -27,7 +27,7 @@ #include #include #include -#include +#include #define SUBPIXEL_MASK 0xffff @@ -369,3 +369,173 @@ int drm_crtc_init(struct drm_device *dev, struct drm_crtc *crtc, return drm_crtc_init_with_planes(dev, crtc, primary, NULL, funcs); } EXPORT_SYMBOL(drm_crtc_init); + +static int +plane_commit(struct drm_plane *plane, struct drm_plane_state *plane_state, + struct drm_framebuffer *old_fb) +{ + struct drm_plane_helper_funcs *plane_funcs; + struct drm_crtc *crtc[2]; + struct drm_crtc_helper_funcs *crtc_funcs[2]; + int i, ret = 0; + + plane_funcs = plane->helper_private; + + /* Since this is a transitional helper we can't assume that plane->state + * is always valid. Hence we need to use plane->crtc instead of + * plane->state->crtc as the old crtc. */ + crtc[0] = plane->crtc; + crtc[1] = crtc[0] != plane_state->crtc ? plane_state->crtc : NULL; + + for (i = 0; i < 2; i++) + crtc_funcs[i] = crtc[i] ? crtc[i]->helper_private : NULL; + + if (plane_funcs->atomic_check) { + ret = plane_funcs->atomic_check(plane, plane_state); + if (ret) + goto out; + } + + if (plane_funcs->prepare_fb && plane_state->fb) { + ret = plane_funcs->prepare_fb(plane, plane_state->fb); + if (ret) + goto out; + } + + /* Point of no return, commit sw state. */ + swap(plane->state, plane_state); + + for (i = 0; i < 2; i++) { + if (crtc_funcs[i] && crtc_funcs[i]->atomic_begin) + crtc_funcs[i]->atomic_begin(crtc[i]); + } + + plane_funcs->atomic_update(plane); + + for (i = 0; i < 2; i++) { + if (crtc_funcs[i] && crtc_funcs[i]->atomic_flush) + crtc_funcs[i]->atomic_flush(crtc[i]); + } + + for (i = 0; i < 2; i++) { + if (!crtc[i]) + continue; + + /* There's no other way to figure out whether the crtc is running. */ + ret = drm_crtc_vblank_get(crtc[i]); + if (ret == 0) { + drm_crtc_wait_one_vblank(crtc[i]); + drm_crtc_vblank_put(crtc[i]); + } + + ret = 0; + } + + if (plane_funcs->cleanup_fb && old_fb) + plane_funcs->cleanup_fb(plane, old_fb); +out: + if (plane_state) { + if (plane->funcs->atomic_destroy_state) + plane->funcs->atomic_destroy_state(plane, plane_state); + else + kfree(plane_state); + } + + return ret; +} + +/** + * drm_plane_helper_update() - Helper for primary plane update + * @plane: plane object to update + * @crtc: owning CRTC of owning plane + * @fb: framebuffer to flip onto plane + * @crtc_x: x offset of primary plane on crtc + * @crtc_y: y offset of primary plane on crtc + * @crtc_w: width of primary plane rectangle on crtc + * @crtc_h: height of primary plane rectangle on crtc + * @src_x: x offset of @fb for panning + * @src_y: y offset of @fb for panning + * @src_w: width of source rectangle in @fb + * @src_h: height of source rectangle in @fb + * + * Provides a default plane update handler using the atomic plane update + * functions. It is fully left to the driver to check plane constraints and + * handle corner-cases like a fully occluded or otherwise invisible plane. + * + * This is useful for piecewise transitioning of a driver to the atomic helpers. + * + * RETURNS: + * Zero on success, error code on failure + */ +int drm_plane_helper_update(struct drm_plane *plane, struct drm_crtc *crtc, + struct drm_framebuffer *fb, + int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t src_x, uint32_t src_y, + uint32_t src_w, uint32_t src_h) +{ + struct drm_plane_state *plane_state; + + if (plane->funcs->atomic_duplicate_state) + plane_state = plane->funcs->atomic_duplicate_state(plane); + else if (plane->state) + plane_state = kmemdup(plane->state, sizeof(*plane_state), + GFP_KERNEL); + else + plane_state = kzalloc(sizeof(*plane_state), GFP_KERNEL); + if (!plane_state) + return -ENOMEM; + + plane_state->crtc = crtc; + plane_state->fb = fb; + plane_state->crtc_x = crtc_x; + plane_state->crtc_y = crtc_y; + plane_state->crtc_h = crtc_h; + plane_state->crtc_w = crtc_w; + plane_state->src_x = src_x; + plane_state->src_y = src_y; + plane_state->src_h = src_h; + plane_state->src_w = src_w; + + return plane_commit(plane, plane_state, plane->fb); +} +EXPORT_SYMBOL(drm_plane_helper_update); + +/** + * drm_plane_helper_disable() - Helper for primary plane disable + * @plane: plane to disable + * + * Provides a default plane disable handler using the atomic plane update + * functions. It is fully left to the driver to check plane constraints and + * handle corner-cases like a fully occluded or otherwise invisible plane. + * + * This is useful for piecewise transitioning of a driver to the atomic helpers. + * + * RETURNS: + * Zero on success, error code on failure + */ +int drm_plane_helper_disable(struct drm_plane *plane) +{ + struct drm_plane_state *plane_state; + + /* crtc helpers love to call disable functions for already disabled hw + * functions. So cope with that. */ + if (!plane->crtc) + return 0; + + if (plane->funcs->atomic_duplicate_state) + plane_state = plane->funcs->atomic_duplicate_state(plane); + else if (plane->state) + plane_state = kmemdup(plane->state, sizeof(*plane_state), + GFP_KERNEL); + else + plane_state = kzalloc(sizeof(*plane_state), GFP_KERNEL); + if (!plane_state) + return -ENOMEM; + + plane_state->crtc = NULL; + plane_state->fb = NULL; + + return plane_commit(plane, plane_state, plane->fb); +} +EXPORT_SYMBOL(drm_plane_helper_disable); diff --git a/include/drm/drm_plane_helper.h b/include/drm/drm_plane_helper.h index 8577ec8..92bcecc 100644 --- a/include/drm/drm_plane_helper.h +++ b/include/drm/drm_plane_helper.h @@ -95,4 +95,12 @@ extern struct drm_plane *drm_primary_helper_create_plane(struct drm_device *dev, int num_formats); +int drm_plane_helper_update(struct drm_plane *plane, struct drm_crtc *crtc, + struct drm_framebuffer *fb, + int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t src_x, uint32_t src_y, + uint32_t src_w, uint32_t src_h); +int drm_plane_helper_disable(struct drm_plane *plane); + #endif -- cgit v0.10.2 From 2f324b42b7b24a48fe3f8a7af60ec3c9024255fa Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 29 Oct 2014 11:13:47 +0100 Subject: drm/crtc-helper: Transitional functions using atomic plane helpers These two functions allow drivers to reuse their atomic plane helpers functions for the primary plane to implement the interfaces required by the crtc helpers for the legacy ->set_config callback. This is purely transitional and won't be used once the driver is fully converted. But it allows partial conversions to the atomic plane helpers which are functional. v2: - Use ->atomic_duplicate_state if available. - Don't forget to run crtc_funcs->atomic_check. v3: Shift source coordinates correctly for 16.16 fixed point. v4: Don't forget to call ->atomic_destroy_state if available. v5: Fixup kerneldoc. v6: Reuse the plane_commit function from the transitional plane helpers to avoid too much duplication. v7: - Remove some stale comment. - Correctly handle the lack of plane->state object, necessary for transitional use. v8: Fixup an embarrassing h/vdisplay mixup. Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc_helper.c b/drivers/gpu/drm/drm_crtc_helper.c index 6c65a0a..95ecbb1 100644 --- a/drivers/gpu/drm/drm_crtc_helper.c +++ b/drivers/gpu/drm/drm_crtc_helper.c @@ -38,6 +38,7 @@ #include #include #include +#include #include MODULE_AUTHOR("David Airlie, Jesse Barnes"); @@ -888,3 +889,112 @@ void drm_helper_resume_force_mode(struct drm_device *dev) drm_modeset_unlock_all(dev); } EXPORT_SYMBOL(drm_helper_resume_force_mode); + +/** + * drm_helper_crtc_mode_set - mode_set implementation for atomic plane helpers + * @crtc: DRM CRTC + * @mode: DRM display mode which userspace requested + * @adjusted_mode: DRM display mode adjusted by ->mode_fixup callbacks + * @x: x offset of the CRTC scanout area on the underlying framebuffer + * @y: y offset of the CRTC scanout area on the underlying framebuffer + * @old_fb: previous framebuffer + * + * This function implements a callback useable as the ->mode_set callback + * required by the crtc helpers. Besides the atomic plane helper functions for + * the primary plane the driver must also provide the ->mode_set_nofb callback + * to set up the crtc. + * + * This is a transitional helper useful for converting drivers to the atomic + * interfaces. + */ +int drm_helper_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode, int x, int y, + struct drm_framebuffer *old_fb) +{ + struct drm_crtc_state *crtc_state; + struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private; + int ret; + + if (crtc->funcs->atomic_duplicate_state) + crtc_state = crtc->funcs->atomic_duplicate_state(crtc); + else if (crtc->state) + crtc_state = kmemdup(crtc->state, sizeof(*crtc_state), + GFP_KERNEL); + else + crtc_state = kzalloc(sizeof(*crtc_state), GFP_KERNEL); + if (!crtc_state) + return -ENOMEM; + + crtc_state->enable = true; + crtc_state->planes_changed = true; + drm_mode_copy(&crtc_state->mode, mode); + drm_mode_copy(&crtc_state->adjusted_mode, adjusted_mode); + + if (crtc_funcs->atomic_check) { + ret = crtc_funcs->atomic_check(crtc, crtc_state); + if (ret) { + kfree(crtc_state); + + return ret; + } + } + + swap(crtc->state, crtc_state); + + crtc_funcs->mode_set_nofb(crtc); + + if (crtc_state) { + if (crtc->funcs->atomic_destroy_state) + crtc->funcs->atomic_destroy_state(crtc, crtc_state); + else + kfree(crtc_state); + } + + return drm_helper_crtc_mode_set_base(crtc, x, y, old_fb); +} +EXPORT_SYMBOL(drm_helper_crtc_mode_set); + +/** + * drm_helper_crtc_mode_set_base - mode_set_base implementation for atomic plane helpers + * @crtc: DRM CRTC + * @x: x offset of the CRTC scanout area on the underlying framebuffer + * @y: y offset of the CRTC scanout area on the underlying framebuffer + * @old_fb: previous framebuffer + * + * This function implements a callback useable as the ->mode_set_base used + * required by the crtc helpers. The driver must provide the atomic plane helper + * functions for the primary plane. + * + * This is a transitional helper useful for converting drivers to the atomic + * interfaces. + */ +int drm_helper_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, + struct drm_framebuffer *old_fb) +{ + struct drm_plane_state *plane_state; + struct drm_plane *plane = crtc->primary; + + if (plane->funcs->atomic_duplicate_state) + plane_state = plane->funcs->atomic_duplicate_state(plane); + else if (plane->state) + plane_state = kmemdup(plane->state, sizeof(*plane_state), + GFP_KERNEL); + else + plane_state = kzalloc(sizeof(*plane_state), GFP_KERNEL); + if (!plane_state) + return -ENOMEM; + + plane_state->crtc = crtc; + plane_state->fb = crtc->primary->fb; + plane_state->crtc_x = 0; + plane_state->crtc_y = 0; + plane_state->crtc_h = crtc->mode.vdisplay; + plane_state->crtc_w = crtc->mode.hdisplay; + plane_state->src_x = x << 16; + plane_state->src_y = y << 16; + plane_state->src_h = crtc->mode.vdisplay << 16; + plane_state->src_w = crtc->mode.hdisplay << 16; + + return drm_plane_helper_commit(plane, plane_state, old_fb); +} +EXPORT_SYMBOL(drm_helper_crtc_mode_set_base); diff --git a/drivers/gpu/drm/drm_plane_helper.c b/drivers/gpu/drm/drm_plane_helper.c index a202b89..a5a2958 100644 --- a/drivers/gpu/drm/drm_plane_helper.c +++ b/drivers/gpu/drm/drm_plane_helper.c @@ -370,9 +370,9 @@ int drm_crtc_init(struct drm_device *dev, struct drm_crtc *crtc, } EXPORT_SYMBOL(drm_crtc_init); -static int -plane_commit(struct drm_plane *plane, struct drm_plane_state *plane_state, - struct drm_framebuffer *old_fb) +int drm_plane_helper_commit(struct drm_plane *plane, + struct drm_plane_state *plane_state, + struct drm_framebuffer *old_fb) { struct drm_plane_helper_funcs *plane_funcs; struct drm_crtc *crtc[2]; @@ -497,7 +497,7 @@ int drm_plane_helper_update(struct drm_plane *plane, struct drm_crtc *crtc, plane_state->src_h = src_h; plane_state->src_w = src_w; - return plane_commit(plane, plane_state, plane->fb); + return drm_plane_helper_commit(plane, plane_state, plane->fb); } EXPORT_SYMBOL(drm_plane_helper_update); @@ -536,6 +536,6 @@ int drm_plane_helper_disable(struct drm_plane *plane) plane_state->crtc = NULL; plane_state->fb = NULL; - return plane_commit(plane, plane_state, plane->fb); + return drm_plane_helper_commit(plane, plane_state, plane->fb); } EXPORT_SYMBOL(drm_plane_helper_disable); diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 5614740..53c8638 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -230,6 +230,7 @@ struct drm_atomic_state; * struct drm_crtc_state - mutable CRTC state * @enable: whether the CRTC should be enabled, gates all other state * @planes_changed: for use by helpers and drivers when computing state updates + * @adjusted_mode: for use by helpers and drivers to compute adjusted mode timings * @mode: current mode timings * @event: optional pointer to a DRM event to signal upon completion of the * state update @@ -241,6 +242,9 @@ struct drm_crtc_state { /* computed state bits used by helpers and drivers */ bool planes_changed : 1; + /* adjusted_mode: for use by helpers and drivers */ + struct drm_display_mode adjusted_mode; + struct drm_display_mode mode; struct drm_pending_vblank_event *event; diff --git a/include/drm/drm_crtc_helper.h b/include/drm/drm_crtc_helper.h index adec48b..7adbb65 100644 --- a/include/drm/drm_crtc_helper.h +++ b/include/drm/drm_crtc_helper.h @@ -68,6 +68,7 @@ struct drm_crtc_helper_funcs { int (*mode_set)(struct drm_crtc *crtc, struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode, int x, int y, struct drm_framebuffer *old_fb); + void (*mode_set_nofb)(struct drm_crtc *crtc); /* Move the crtc on the current fb to the given position *optional* */ int (*mode_set_base)(struct drm_crtc *crtc, int x, int y, @@ -167,6 +168,12 @@ static inline void drm_connector_helper_add(struct drm_connector *connector, extern void drm_helper_resume_force_mode(struct drm_device *dev); +int drm_helper_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode, int x, int y, + struct drm_framebuffer *old_fb); +int drm_helper_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, + struct drm_framebuffer *old_fb); + /* drm_probe_helper.c */ extern int drm_helper_probe_single_connector_modes(struct drm_connector *connector, uint32_t maxX, diff --git a/include/drm/drm_plane_helper.h b/include/drm/drm_plane_helper.h index 92bcecc..c48f14d 100644 --- a/include/drm/drm_plane_helper.h +++ b/include/drm/drm_plane_helper.h @@ -103,4 +103,8 @@ int drm_plane_helper_update(struct drm_plane *plane, struct drm_crtc *crtc, uint32_t src_w, uint32_t src_h); int drm_plane_helper_disable(struct drm_plane *plane); +/* For use by drm_crtc_helper.c */ +int drm_plane_helper_commit(struct drm_plane *plane, + struct drm_plane_state *plane_state, + struct drm_framebuffer *old_fb); #endif -- cgit v0.10.2 From 8a8cc83cbd081bd38673220d4c2816baa89cc788 Mon Sep 17 00:00:00 2001 From: Alexandre Courbot Date: Thu, 23 Oct 2014 17:16:17 +0900 Subject: drm/panel: ld9040: Update calls to gpiod_get*() Add the new flags argument to calls of (devm_)gpiod_get*() and remove any direction setting code afterwards. Currently both forms (with or without the flags argument) are valid thanks to transitional macros in . These macros will be removed once all consumers are updated and the flags argument will become compulsary. Signed-off-by: Alexandre Courbot Acked-by: Andrzej Hajda Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/panel/panel-ld9040.c b/drivers/gpu/drm/panel/panel-ld9040.c index 42ac67b..dd1d3f2 100644 --- a/drivers/gpu/drm/panel/panel-ld9040.c +++ b/drivers/gpu/drm/panel/panel-ld9040.c @@ -336,17 +336,12 @@ static int ld9040_probe(struct spi_device *spi) if (ret < 0) return ret; - ctx->reset_gpio = devm_gpiod_get(dev, "reset"); + ctx->reset_gpio = devm_gpiod_get(dev, "reset", GPIOD_OUT_HIGH); if (IS_ERR(ctx->reset_gpio)) { dev_err(dev, "cannot get reset-gpios %ld\n", PTR_ERR(ctx->reset_gpio)); return PTR_ERR(ctx->reset_gpio); } - ret = gpiod_direction_output(ctx->reset_gpio, 1); - if (ret < 0) { - dev_err(dev, "cannot configure reset-gpios %d\n", ret); - return ret; - } spi->bits_per_word = 9; ret = spi_setup(spi); -- cgit v0.10.2 From 26c2ad8a49dffcb6bdae447dc7c02db81323dbf7 Mon Sep 17 00:00:00 2001 From: Alexandre Courbot Date: Thu, 23 Oct 2014 17:16:57 +0900 Subject: drm/panel: s6e8aa0: Update calls to gpiod_get*() Add the new flags argument to calls of (devm_)gpiod_get*() and remove any direction setting code afterwards. Currently both forms (with or without the flags argument) are valid thanks to transitional macros in . These macros will be removed once all consumers are updated and the flags argument will become compulsary. Signed-off-by: Alexandre Courbot Acked-by: Andrzej Hajda Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/panel/panel-s6e8aa0.c b/drivers/gpu/drm/panel/panel-s6e8aa0.c index b5217fe..6427aa1 100644 --- a/drivers/gpu/drm/panel/panel-s6e8aa0.c +++ b/drivers/gpu/drm/panel/panel-s6e8aa0.c @@ -1019,17 +1019,12 @@ static int s6e8aa0_probe(struct mipi_dsi_device *dsi) return ret; } - ctx->reset_gpio = devm_gpiod_get(dev, "reset"); + ctx->reset_gpio = devm_gpiod_get(dev, "reset", GPIOD_OUT_HIGH); if (IS_ERR(ctx->reset_gpio)) { dev_err(dev, "cannot get reset-gpios %ld\n", PTR_ERR(ctx->reset_gpio)); return PTR_ERR(ctx->reset_gpio); } - ret = gpiod_direction_output(ctx->reset_gpio, 1); - if (ret < 0) { - dev_err(dev, "cannot configure reset-gpios %d\n", ret); - return ret; - } ctx->brightness = GAMMA_LEVEL_NUM - 1; -- cgit v0.10.2 From a61400d85c3ba4c8e83fd7a6e610b2af0136a94a Mon Sep 17 00:00:00 2001 From: Alexandre Courbot Date: Thu, 23 Oct 2014 17:16:58 +0900 Subject: drm/panel: simple: Update calls to gpiod_get*() Add the new flags argument to calls of (devm_)gpiod_get*() and remove any direction setting code afterwards. Currently both forms (with or without the flags argument) are valid thanks to transitional macros in . These macros will be removed once all consumers are updated and the flags argument will become compulsary. Signed-off-by: Alexandre Courbot Acked-by: Andrzej Hajda Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 23de22f..83d9306 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -247,21 +247,14 @@ static int panel_simple_probe(struct device *dev, const struct panel_desc *desc) if (IS_ERR(panel->supply)) return PTR_ERR(panel->supply); - panel->enable_gpio = devm_gpiod_get_optional(dev, "enable"); + panel->enable_gpio = devm_gpiod_get_optional(dev, "enable", + GPIOD_OUT_LOW); if (IS_ERR(panel->enable_gpio)) { err = PTR_ERR(panel->enable_gpio); dev_err(dev, "failed to request GPIO: %d\n", err); return err; } - if (panel->enable_gpio) { - err = gpiod_direction_output(panel->enable_gpio, 0); - if (err < 0) { - dev_err(dev, "failed to setup GPIO: %d\n", err); - return err; - } - } - backlight = of_parse_phandle(dev->of_node, "backlight", 0); if (backlight) { panel->backlight = of_find_backlight_by_node(backlight); -- cgit v0.10.2 From 623369e533e8a5f85999d605723efa4523554bae Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 16 Sep 2014 17:50:47 +0200 Subject: drm: Atomic crtc/connector updates using crtc/plane helper interfaces So this is finally the integration of the crtc and plane helper interfaces into the atomic helper functions. In the check function we now have a few steps: - First we update the output routing and figure out which crtcs need a full mode set. Suitable encoders are selected using ->best_encoder, with the same semantics as the crtc helpers of implicitly disabling all connectors currently using the encoder. - Then we pull all other connectors into the state update which feed from a crtc which changes. This must be done do catch mode changes and similar updates - atomic updates are differences on top of the current state. - Then we call all the various ->mode_fixup to compute the adjusted mode. Note that here we have a slight semantic difference compared to the crtc helpers: We have not yet updated the encoder->crtc link when calling the encoder's ->mode_fixup function. But that's a requirement when converting to atomic since we want to prepare the entire state completely contained with the over drm_atomic_state structure. So this must be carefully checked when converting drivers over to atomic helpers. - Finally we do call the atomic_check functions on planes and crtcs. The commit function is also quite a beast: - The only step that can fail is done first, namely pinning the framebuffers. After that we cross the point of no return, an async commit would push all that into the worker thread. - The disabling of encoders and connectors is a bit tricky, since depending upon the final state we need to select different crtc helper functions. - Software tracking is a bit clarified compared to the crtc helpers: We commit the software state before starting to touch the hardware, like crtc helpers. But since we just swap them we still have the old state (i.e. the current hw state) around, which is really handy to write simple disable functions. So no more drm_crtc_helper_disable_all_unused_functions kind of fun because we're leaving unused crtcs/encoders behind. Everything gets shut down in-order now, which is one of the key differences of the i915 helpers compared to crtc helpers and a really nice additional guarantee. - Like with the plane helpers the atomic commit function waits for one vblank to pass before calling the framebuffer cleanup function. Compared to Rob's helper approach there's a bunch of upsides: - All the interfaces which can fail are called in the ->check hook (i.e. ->best_match and the various ->mode_fixup hooks). This means that drivers can just reuse those functions and don't need to move everything into ->atomic_check callbacks. If drivers have no need for additional constraint checking beyong their existing crtc helper callbacks they don't need to do anything. - The actual commit operation is properly stage: First we prepare framebuffers, which can potentially still fail (due to memory exhausting). This is important for the async case, where this must be done synchronously to correctly return errors. - The output configuration changes (done with crtc helper functions) and the plane update (using atomic plane helpers) are correctly interleaved: First we shut down any crtcs that need changing, then we update planes and finally we enable everything again. Hardware without GO bits must be more careful with ordering, which this sequence enables. - Also for hardware with shared output resources (like display PLLs) we first must shut down the old configuration before we can enable the new one. Otherwise we can hit an impossible intermediate state where there's not enough PLLs (which is the point behind atomic updates). v2: - Ensure that users of ->check update crtc_state->enable correctly. - Update the legacy state in crtc/plane structures. Eventually we want to remove that, but for now the drm core still expects this (especially the plane->fb pointer). v3: A few changes for better async handling: - Reorder the software side state commit so that it happens all before we touch the hardware. This way async support becomes very easy since we can punt all the actual hw touching to a worker thread. And as long as we synchronize with that thread (flushing or cancelling, depending upon what the driver can handle) before we commit the next software state there's no need for any locking in the worker thread at all. Which greatly simplifies things. And as long as we synchronize with all relevant threads we can have a lot of them (e.g. per-crtc for per-crtc updates) running in parallel. - Expose pre/post plane commit steps separately. We need to expose the actual hw commit step anyway for drivers to be able to implement asynchronous commit workers. But if we expose pre/post and plane commit steps individually we allow drivers to selectively use atomic helpers. - I've forgotten to call encoder/bridge ->mode_set functions, fix this. v4: Add debug output and fix a mixup between current and new state that resulted in crtcs not getting updated correctly. And in an Oops ... v5: - Be kind to driver writers in the vblank wait functions.. if thing aren't working yet, and vblank irq will never come, then let's not block forever.. especially under console-lock. - Correctly clear connector_state->best_encoder when disabling. Spotted while trying to understand a report from Rob Clark. - Only steal encoder if it actually changed, otherwise hilarity ensues if we steal from the current connector and so set the ->crtc pointer unexpectedly to NULL. Reported by Rob Clark. - Bail out in disable_outputs if an output currently doesn't have a best_encoder - this means it's already disabled. v6: Fixupe kerneldoc as reported by Paulo. And also fix up kerneldoc in drm_crtc.h. v7: Take ownership of the atomic state and clean it up with drm_atomic_state_free(). v8 Various improvements all over: - Polish code comments and kerneldoc. - Improve debug output to make sure all failure cases are logged. - Treat enabled crtc with no connectors as invalid input from userspace. - Don't ignore the return value from mode_fixup(). v9: - Improve debug output for crtc_state->mode_changed. v10: - Fixup the vblank waiting code to properly balance the vblank_get/put calls. - Better comments when checking/computing crtc->mode_changed v11: Fixup the encoder stealing logic: We can't look at encoder->crtc since that's not in the atomic state structures and might be updated asynchronously in and async commit. Instead we need to inspect all the connector states and check whether the encoder is currently in used and if so, on which crtc. v12: Review from Sean: - A few spelling fixes. - Flatten control flow indent by converting if blocks to early continue/return in 2 places. - Capture connectors_for_crtc return value in int num_connectors instead of bool has_connectors and do an explicit int->bool conversion with !!. I think the helper is more useful for drivers if it returns the number of connectors (e.g. to detect cloning configurations), so decided to keep that return value. Cc: Sean Paul Cc: Paulo Zanoni Cc: Rob Clark Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 55a8eb2..c4936bc 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -29,6 +29,7 @@ #include #include #include +#include static void drm_atomic_helper_plane_changed(struct drm_atomic_state *state, @@ -57,6 +58,332 @@ drm_atomic_helper_plane_changed(struct drm_atomic_state *state, } } +static struct drm_crtc * +get_current_crtc_for_encoder(struct drm_device *dev, + struct drm_encoder *encoder) +{ + struct drm_mode_config *config = &dev->mode_config; + struct drm_connector *connector; + + WARN_ON(!drm_modeset_is_locked(&config->connection_mutex)); + + list_for_each_entry(connector, &config->connector_list, head) { + if (connector->state->best_encoder != encoder) + continue; + + return connector->state->crtc; + } + + return NULL; +} + +static int +steal_encoder(struct drm_atomic_state *state, + struct drm_encoder *encoder, + struct drm_crtc *encoder_crtc) +{ + struct drm_mode_config *config = &state->dev->mode_config; + struct drm_crtc_state *crtc_state; + struct drm_connector *connector; + struct drm_connector_state *connector_state; + + /* + * We can only steal an encoder coming from a connector, which means we + * must already hold the connection_mutex. + */ + WARN_ON(!drm_modeset_is_locked(&config->connection_mutex)); + + DRM_DEBUG_KMS("[ENCODER:%d:%s] in use on [CRTC:%d], stealing it\n", + encoder->base.id, encoder->name, + encoder_crtc->base.id); + + crtc_state = drm_atomic_get_crtc_state(state, encoder_crtc); + if (IS_ERR(crtc_state)) + return PTR_ERR(crtc_state); + + crtc_state->mode_changed = true; + + list_for_each_entry(connector, &config->connector_list, head) { + if (connector->state->best_encoder != encoder) + continue; + + DRM_DEBUG_KMS("Stealing encoder from [CONNECTOR:%d:%s]\n", + connector->base.id, + connector->name); + + connector_state = drm_atomic_get_connector_state(state, + connector); + if (IS_ERR(connector_state)) + return PTR_ERR(connector_state); + + connector_state->crtc = NULL; + connector_state->best_encoder = NULL; + } + + return 0; +} + +static int +update_connector_routing(struct drm_atomic_state *state, int conn_idx) +{ + struct drm_connector_helper_funcs *funcs; + struct drm_encoder *new_encoder; + struct drm_crtc *encoder_crtc; + struct drm_connector *connector; + struct drm_connector_state *connector_state; + struct drm_crtc_state *crtc_state; + int idx, ret; + + connector = state->connectors[conn_idx]; + connector_state = state->connector_states[conn_idx]; + + if (!connector) + return 0; + + DRM_DEBUG_KMS("Updating routing for [CONNECTOR:%d:%s]\n", + connector->base.id, + connector->name); + + if (connector->state->crtc != connector_state->crtc) { + if (connector->state->crtc) { + idx = drm_crtc_index(connector->state->crtc); + + crtc_state = state->crtc_states[idx]; + crtc_state->mode_changed = true; + } + + if (connector_state->crtc) { + idx = drm_crtc_index(connector_state->crtc); + + crtc_state = state->crtc_states[idx]; + crtc_state->mode_changed = true; + } + } + + if (!connector_state->crtc) { + DRM_DEBUG_KMS("Disabling [CONNECTOR:%d:%s]\n", + connector->base.id, + connector->name); + + connector_state->best_encoder = NULL; + + return 0; + } + + funcs = connector->helper_private; + new_encoder = funcs->best_encoder(connector); + + if (!new_encoder) { + DRM_DEBUG_KMS("No suitable encoder found for [CONNECTOR:%d:%s]\n", + connector->base.id, + connector->name); + return -EINVAL; + } + + if (new_encoder == connector_state->best_encoder) { + DRM_DEBUG_KMS("[CONNECTOR:%d:%s] keeps [ENCODER:%d:%s], now on [CRTC:%d]\n", + connector->base.id, + connector->name, + new_encoder->base.id, + new_encoder->name, + connector_state->crtc->base.id); + + return 0; + } + + encoder_crtc = get_current_crtc_for_encoder(state->dev, + new_encoder); + + if (encoder_crtc) { + ret = steal_encoder(state, new_encoder, encoder_crtc); + if (ret) { + DRM_DEBUG_KMS("Encoder stealing failed for [CONNECTOR:%d:%s]\n", + connector->base.id, + connector->name); + return ret; + } + } + + connector_state->best_encoder = new_encoder; + idx = drm_crtc_index(connector_state->crtc); + + crtc_state = state->crtc_states[idx]; + crtc_state->mode_changed = true; + + DRM_DEBUG_KMS("[CONNECTOR:%d:%s] using [ENCODER:%d:%s] on [CRTC:%d]\n", + connector->base.id, + connector->name, + new_encoder->base.id, + new_encoder->name, + connector_state->crtc->base.id); + + return 0; +} + +static int +mode_fixup(struct drm_atomic_state *state) +{ + int ncrtcs = state->dev->mode_config.num_crtc; + int nconnectors = state->dev->mode_config.num_connector; + struct drm_crtc_state *crtc_state; + struct drm_connector_state *conn_state; + int i; + bool ret; + + for (i = 0; i < ncrtcs; i++) { + crtc_state = state->crtc_states[i]; + + if (!crtc_state || !crtc_state->mode_changed) + continue; + + drm_mode_copy(&crtc_state->adjusted_mode, &crtc_state->mode); + } + + for (i = 0; i < nconnectors; i++) { + struct drm_encoder_helper_funcs *funcs; + struct drm_encoder *encoder; + + conn_state = state->connector_states[i]; + + if (!conn_state) + continue; + + WARN_ON(!!conn_state->best_encoder != !!conn_state->crtc); + + if (!conn_state->crtc || !conn_state->best_encoder) + continue; + + crtc_state = + state->crtc_states[drm_crtc_index(conn_state->crtc)]; + + /* + * Each encoder has at most one connector (since we always steal + * it away), so we won't call ->mode_fixup twice. + */ + encoder = conn_state->best_encoder; + funcs = encoder->helper_private; + + if (encoder->bridge && encoder->bridge->funcs->mode_fixup) { + ret = encoder->bridge->funcs->mode_fixup( + encoder->bridge, &crtc_state->mode, + &crtc_state->adjusted_mode); + if (!ret) { + DRM_DEBUG_KMS("Bridge fixup failed\n"); + return -EINVAL; + } + } + + + ret = funcs->mode_fixup(encoder, &crtc_state->mode, + &crtc_state->adjusted_mode); + if (!ret) { + DRM_DEBUG_KMS("[ENCODER:%d:%s] fixup failed\n", + encoder->base.id, encoder->name); + return -EINVAL; + } + } + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc_helper_funcs *funcs; + struct drm_crtc *crtc; + + crtc_state = state->crtc_states[i]; + crtc = state->crtcs[i]; + + if (!crtc_state || !crtc_state->mode_changed) + continue; + + funcs = crtc->helper_private; + ret = funcs->mode_fixup(crtc, &crtc_state->mode, + &crtc_state->adjusted_mode); + if (!ret) { + DRM_DEBUG_KMS("[CRTC:%d] fixup failed\n", + crtc->base.id); + return -EINVAL; + } + } + + return 0; +} + +static int +drm_atomic_helper_check_prepare(struct drm_device *dev, + struct drm_atomic_state *state) +{ + int ncrtcs = dev->mode_config.num_crtc; + int nconnectors = dev->mode_config.num_connector; + struct drm_crtc *crtc; + struct drm_crtc_state *crtc_state; + int i, ret; + + for (i = 0; i < ncrtcs; i++) { + crtc = state->crtcs[i]; + crtc_state = state->crtc_states[i]; + + if (!crtc) + continue; + + if (!drm_mode_equal(&crtc->state->mode, &crtc_state->mode)) { + DRM_DEBUG_KMS("[CRTC:%d] mode changed\n", + crtc->base.id); + crtc_state->mode_changed = true; + } + + if (crtc->state->enable != crtc_state->enable) { + DRM_DEBUG_KMS("[CRTC:%d] enable changed\n", + crtc->base.id); + crtc_state->mode_changed = true; + } + } + + for (i = 0; i < nconnectors; i++) { + /* + * This only sets crtc->mode_changed for routing changes, + * drivers must set crtc->mode_changed themselves when connector + * properties need to be updated. + */ + ret = update_connector_routing(state, i); + if (ret) + return ret; + } + + /* + * After all the routing has been prepared we need to add in any + * connector which is itself unchanged, but who's crtc changes it's + * configuration. This must be done before calling mode_fixup in case a + * crtc only changed its mode but has the same set of connectors. + */ + for (i = 0; i < ncrtcs; i++) { + int num_connectors; + + crtc = state->crtcs[i]; + crtc_state = state->crtc_states[i]; + + if (!crtc || !crtc_state->mode_changed) + continue; + + DRM_DEBUG_KMS("[CRTC:%d] needs full modeset, enable: %c\n", + crtc->base.id, + crtc_state->enable ? 'y' : 'n'); + + ret = drm_atomic_add_affected_connectors(state, crtc); + if (ret != 0) + return ret; + + num_connectors = drm_atomic_connectors_for_crtc(state, + crtc); + + if (crtc_state->enable != !!num_connectors) { + DRM_DEBUG_KMS("[CRTC:%d] enabled/connectors mismatch\n", + crtc->base.id); + + return -EINVAL; + } + } + + return mode_fixup(state); +} + /** * drm_atomic_helper_check - validate state object * @dev: DRM device @@ -78,6 +405,10 @@ int drm_atomic_helper_check(struct drm_device *dev, int ncrtcs = dev->mode_config.num_crtc; int i, ret = 0; + ret = drm_atomic_helper_check_prepare(dev, state); + if (ret) + return ret; + for (i = 0; i < nplanes; i++) { struct drm_plane_helper_funcs *funcs; struct drm_plane *plane = state->planes[i]; @@ -125,6 +456,372 @@ int drm_atomic_helper_check(struct drm_device *dev, } EXPORT_SYMBOL(drm_atomic_helper_check); +static void +disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state) +{ + int ncrtcs = old_state->dev->mode_config.num_crtc; + int nconnectors = old_state->dev->mode_config.num_connector; + int i; + + for (i = 0; i < nconnectors; i++) { + struct drm_connector_state *old_conn_state; + struct drm_connector *connector; + struct drm_encoder_helper_funcs *funcs; + struct drm_encoder *encoder; + + old_conn_state = old_state->connector_states[i]; + connector = old_state->connectors[i]; + + /* Shut down everything that's in the changeset and currently + * still on. So need to check the old, saved state. */ + if (!old_conn_state || !old_conn_state->crtc) + continue; + + encoder = connector->state->best_encoder; + + if (!encoder) + continue; + + funcs = encoder->helper_private; + + /* + * Each encoder has at most one connector (since we always steal + * it away), so we won't call call disable hooks twice. + */ + if (encoder->bridge) + encoder->bridge->funcs->disable(encoder->bridge); + + /* Right function depends upon target state. */ + if (connector->state->crtc) + funcs->prepare(encoder); + else if (funcs->disable) + funcs->disable(encoder); + else + funcs->dpms(encoder, DRM_MODE_DPMS_OFF); + + if (encoder->bridge) + encoder->bridge->funcs->post_disable(encoder->bridge); + } + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc_helper_funcs *funcs; + struct drm_crtc *crtc; + + crtc = old_state->crtcs[i]; + + /* Shut down everything that needs a full modeset. */ + if (!crtc || !crtc->state->mode_changed) + continue; + + funcs = crtc->helper_private; + + /* Right function depends upon target state. */ + if (crtc->state->enable) + funcs->prepare(crtc); + else if (funcs->disable) + funcs->disable(crtc); + else + funcs->dpms(crtc, DRM_MODE_DPMS_OFF); + } +} + +static void +set_routing_links(struct drm_device *dev, struct drm_atomic_state *old_state) +{ + int nconnectors = dev->mode_config.num_connector; + int ncrtcs = old_state->dev->mode_config.num_crtc; + int i; + + /* clear out existing links */ + for (i = 0; i < nconnectors; i++) { + struct drm_connector *connector; + + connector = old_state->connectors[i]; + + if (!connector || !connector->encoder) + continue; + + WARN_ON(!connector->encoder->crtc); + + connector->encoder->crtc = NULL; + connector->encoder = NULL; + } + + /* set new links */ + for (i = 0; i < nconnectors; i++) { + struct drm_connector *connector; + + connector = old_state->connectors[i]; + + if (!connector || !connector->state->crtc) + continue; + + if (WARN_ON(!connector->state->best_encoder)) + continue; + + connector->encoder = connector->state->best_encoder; + connector->encoder->crtc = connector->state->crtc; + } + + /* set legacy state in the crtc structure */ + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc *crtc; + + crtc = old_state->crtcs[i]; + + if (!crtc) + continue; + + crtc->mode = crtc->state->mode; + crtc->enabled = crtc->state->enable; + crtc->x = crtc->primary->state->src_x >> 16; + crtc->y = crtc->primary->state->src_y >> 16; + } +} + +static void +crtc_set_mode(struct drm_device *dev, struct drm_atomic_state *old_state) +{ + int ncrtcs = old_state->dev->mode_config.num_crtc; + int nconnectors = old_state->dev->mode_config.num_connector; + int i; + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc_helper_funcs *funcs; + struct drm_crtc *crtc; + + crtc = old_state->crtcs[i]; + + if (!crtc || !crtc->state->mode_changed) + continue; + + funcs = crtc->helper_private; + + if (crtc->state->enable) + funcs->mode_set_nofb(crtc); + } + + for (i = 0; i < nconnectors; i++) { + struct drm_connector *connector; + struct drm_crtc_state *new_crtc_state; + struct drm_encoder_helper_funcs *funcs; + struct drm_encoder *encoder; + struct drm_display_mode *mode, *adjusted_mode; + + connector = old_state->connectors[i]; + + if (!connector || !connector->state->best_encoder) + continue; + + encoder = connector->state->best_encoder; + funcs = encoder->helper_private; + new_crtc_state = connector->state->crtc->state; + mode = &new_crtc_state->mode; + adjusted_mode = &new_crtc_state->adjusted_mode; + + /* + * Each encoder has at most one connector (since we always steal + * it away), so we won't call call mode_set hooks twice. + */ + funcs->mode_set(encoder, mode, adjusted_mode); + + if (encoder->bridge && encoder->bridge->funcs->mode_set) + encoder->bridge->funcs->mode_set(encoder->bridge, + mode, adjusted_mode); + } +} + +/** + * drm_atomic_helper_commit_pre_planes - modeset commit before plane updates + * @dev: DRM device + * @state: atomic state + * + * This function commits the modeset changes that need to be committed before + * updating planes. It shuts down all the outputs that need to be shut down and + * prepares them (if required) with the new mode. + */ +void drm_atomic_helper_commit_pre_planes(struct drm_device *dev, + struct drm_atomic_state *state) +{ + disable_outputs(dev, state); + set_routing_links(dev, state); + crtc_set_mode(dev, state); +} +EXPORT_SYMBOL(drm_atomic_helper_commit_pre_planes); + +/** + * drm_atomic_helper_commit_post_planes - modeset commit after plane updates + * @dev: DRM device + * @old_state: atomic state object with old state structures + * + * This function commits the modeset changes that need to be committed after + * updating planes: It enables all the outputs with the new configuration which + * had to be turned off for the update. + */ +void drm_atomic_helper_commit_post_planes(struct drm_device *dev, + struct drm_atomic_state *old_state) +{ + int ncrtcs = old_state->dev->mode_config.num_crtc; + int nconnectors = old_state->dev->mode_config.num_connector; + int i; + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc_helper_funcs *funcs; + struct drm_crtc *crtc; + + crtc = old_state->crtcs[i]; + + /* Need to filter out CRTCs where only planes change. */ + if (!crtc || !crtc->state->mode_changed) + continue; + + funcs = crtc->helper_private; + + if (crtc->state->enable) + funcs->commit(crtc); + } + + for (i = 0; i < nconnectors; i++) { + struct drm_connector *connector; + struct drm_encoder_helper_funcs *funcs; + struct drm_encoder *encoder; + + connector = old_state->connectors[i]; + + if (!connector || !connector->state->best_encoder) + continue; + + encoder = connector->state->best_encoder; + funcs = encoder->helper_private; + + /* + * Each encoder has at most one connector (since we always steal + * it away), so we won't call call enable hooks twice. + */ + if (encoder->bridge) + encoder->bridge->funcs->pre_enable(encoder->bridge); + + funcs->commit(encoder); + + if (encoder->bridge) + encoder->bridge->funcs->enable(encoder->bridge); + } +} +EXPORT_SYMBOL(drm_atomic_helper_commit_post_planes); + +static void +wait_for_vblanks(struct drm_device *dev, struct drm_atomic_state *old_state) +{ + struct drm_crtc *crtc; + struct drm_crtc_state *old_crtc_state; + int ncrtcs = old_state->dev->mode_config.num_crtc; + int i, ret; + + for (i = 0; i < ncrtcs; i++) { + crtc = old_state->crtcs[i]; + old_crtc_state = old_state->crtc_states[i]; + + if (!crtc) + continue; + + /* No one cares about the old state, so abuse it for tracking + * and store whether we hold a vblank reference (and should do a + * vblank wait) in the ->enable boolean. */ + old_crtc_state->enable = false; + + if (!crtc->state->enable) + continue; + + ret = drm_crtc_vblank_get(crtc); + if (ret != 0) + continue; + + old_crtc_state->enable = true; + old_crtc_state->last_vblank_count = drm_vblank_count(dev, i); + } + + for (i = 0; i < ncrtcs; i++) { + crtc = old_state->crtcs[i]; + old_crtc_state = old_state->crtc_states[i]; + + if (!crtc || !old_crtc_state->enable) + continue; + + ret = wait_event_timeout(dev->vblank[i].queue, + old_crtc_state->last_vblank_count != + drm_vblank_count(dev, i), + msecs_to_jiffies(50)); + + drm_crtc_vblank_put(crtc); + } +} + +/** + * drm_atomic_helper_commit - commit validated state object + * @dev: DRM device + * @state: the driver state object + * @async: asynchronous commit + * + * This function commits a with drm_atomic_helper_check() pre-validated state + * object. This can still fail when e.g. the framebuffer reservation fails. For + * now this doesn't implement asynchronous commits. + * + * RETURNS + * Zero for success or -errno. + */ +int drm_atomic_helper_commit(struct drm_device *dev, + struct drm_atomic_state *state, + bool async) +{ + int ret; + + if (async) + return -EBUSY; + + ret = drm_atomic_helper_prepare_planes(dev, state); + if (ret) + return ret; + + /* + * This is the point of no return - everything below never fails except + * when the hw goes bonghits. Which means we can commit the new state on + * the software side now. + */ + + drm_atomic_helper_swap_state(dev, state); + + /* + * Everything below can be run asynchronously without the need to grab + * any modeset locks at all under one conditions: It must be guaranteed + * that the asynchronous work has either been cancelled (if the driver + * supports it, which at least requires that the framebuffers get + * cleaned up with drm_atomic_helper_cleanup_planes()) or completed + * before the new state gets committed on the software side with + * drm_atomic_helper_swap_state(). + * + * This scheme allows new atomic state updates to be prepared and + * checked in parallel to the asynchronous completion of the previous + * update. Which is important since compositors need to figure out the + * composition of the next frame right after having submitted the + * current layout. + */ + + drm_atomic_helper_commit_pre_planes(dev, state); + + drm_atomic_helper_commit_planes(dev, state); + + drm_atomic_helper_commit_post_planes(dev, state); + + wait_for_vblanks(dev, state); + + drm_atomic_helper_cleanup_planes(dev, state); + + drm_atomic_state_free(state); + + return 0; +} +EXPORT_SYMBOL(drm_atomic_helper_commit); + /** * drm_atomic_helper_prepare_planes - prepare plane resources after commit * @dev: DRM device diff --git a/drivers/gpu/drm/drm_crtc_helper.c b/drivers/gpu/drm/drm_crtc_helper.c index 95ecbb1..46728a8 100644 --- a/drivers/gpu/drm/drm_crtc_helper.c +++ b/drivers/gpu/drm/drm_crtc_helper.c @@ -927,6 +927,7 @@ int drm_helper_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mod crtc_state->enable = true; crtc_state->planes_changed = true; + crtc_state->mode_changed = true; drm_mode_copy(&crtc_state->mode, mode); drm_mode_copy(&crtc_state->adjusted_mode, adjusted_mode); diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h index 79938c6..9781ce7 100644 --- a/include/drm/drm_atomic_helper.h +++ b/include/drm/drm_atomic_helper.h @@ -30,6 +30,14 @@ int drm_atomic_helper_check(struct drm_device *dev, struct drm_atomic_state *state); +int drm_atomic_helper_commit(struct drm_device *dev, + struct drm_atomic_state *state, + bool async); + +void drm_atomic_helper_commit_pre_planes(struct drm_device *dev, + struct drm_atomic_state *state); +void drm_atomic_helper_commit_post_planes(struct drm_device *dev, + struct drm_atomic_state *old_state); int drm_atomic_helper_prepare_planes(struct drm_device *dev, struct drm_atomic_state *state); diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 53c8638..bc47d11 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -229,6 +229,9 @@ struct drm_atomic_state; /** * struct drm_crtc_state - mutable CRTC state * @enable: whether the CRTC should be enabled, gates all other state + * @mode_changed: for use by helpers and drivers when computing state updates + * @last_vblank_count: for helpers and drivers to capture the vblank of the + * update to ensure framebuffer cleanup isn't done too early * @planes_changed: for use by helpers and drivers when computing state updates * @adjusted_mode: for use by helpers and drivers to compute adjusted mode timings * @mode: current mode timings @@ -241,6 +244,10 @@ struct drm_crtc_state { /* computed state bits used by helpers and drivers */ bool planes_changed : 1; + bool mode_changed : 1; + + /* last_vblank_count: for vblank waits before cleanup */ + u32 last_vblank_count; /* adjusted_mode: for use by helpers and drivers */ struct drm_display_mode adjusted_mode; @@ -426,11 +433,14 @@ struct drm_crtc { /** * struct drm_connector_state - mutable connector state * @crtc: CRTC to connect connector to, NULL if disabled + * @best_encoder: can be used by helpers and drivers to select the encoder * @state: backpointer to global drm_atomic_state */ struct drm_connector_state { struct drm_crtc *crtc; + struct drm_encoder *best_encoder; + struct drm_atomic_state *state; }; -- cgit v0.10.2 From 042652ed95996a9ef6dcddddc53b5d8bc7fa887e Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Sun, 27 Jul 2014 13:46:52 +0200 Subject: drm/atomic-helper: implementatations for legacy interfaces Well, except page_flip since that requires async commit, which isn't there yet. For the functions which changes planes there's a bit of trickery involved to keep the fb refcounting working. But otherwise fairly straight-forward atomic updates. The property setting functions are still a bit incomplete. Once we have generic properties (e.g. rotation, but also all the properties needed by the atomic ioctl) we need to filter those out and parse them in the helper. Preferrably with the same function as used by the real atomic ioctl implementation. v2: Fixup kerneldoc, reported by Paulo. v3: Add missing EXPORT_SYMBOL. v4: We need to look at the crtc of the modeset, not some random leftover one from a previous loop when udpating the connector->crtc routing. Also push some local variables into inner loops to avoid these kinds of bugs. v5: Adjust semantics - drivers now own the atomic state upon successfully synchronous commit. v6: Use the set_crtc_for_plane function to assign the crtc, since otherwise the book-keeping is off. v7: - Improve comments. - Filter out the crtc of the ->set_config call when recomputing crtc_state->enabled: We should compute the same state, but not doing so will give us a good chance to catch bugs and inconsistencies - the atomic helper's atomic_check function re-validates this again. - Fix the set_config implementation logic when disabling the crtc: We still need to update the output routing to disable all the connectors properly in the state. Caught by the atomic_check functions, so at least that part worked ;-) Also add some WARN_ONs to ensure ->set_config preconditions all apply. v8: Fixup an embarrassing h/vdisplay mixup. v9: Shuffled bad squash to the right patch, spotted by Daniel v10: Use set_crtc_for_connector as suggested by Sean. v11: Daniel Thompson noticed that my error handling is inconsistent and that in a few cases I didn't handle fatal errors (i.e. not -EDEADLK). Fix this by consolidate the ww mutex backoff handling into one check in the fail: block and flatten the error control flow everywhere else. v12: Review and discussion with Sean: - One spelling fix. - Correctly skip the crtc from the set_config set when recomputing ->enable state. That should allow us to catch any bugs in higher levels in computing that state (which is supplied to the ->set_config implementation). I've screwed this up and Sean spotted that the current code is pointless. Cc: Sean Paul Cc: Daniel Thompson Cc: Sean Paul Cc: Daniel Thompson Cc: Paulo Zanoni Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index c4936bc..077c792 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -86,6 +86,7 @@ steal_encoder(struct drm_atomic_state *state, struct drm_crtc_state *crtc_state; struct drm_connector *connector; struct drm_connector_state *connector_state; + int ret; /* * We can only steal an encoder coming from a connector, which means we @@ -116,7 +117,9 @@ steal_encoder(struct drm_atomic_state *state, if (IS_ERR(connector_state)) return PTR_ERR(connector_state); - connector_state->crtc = NULL; + ret = drm_atomic_set_crtc_for_connector(connector_state, NULL); + if (ret) + return ret; connector_state->best_encoder = NULL; } @@ -1050,3 +1053,509 @@ void drm_atomic_helper_swap_state(struct drm_device *dev, } } EXPORT_SYMBOL(drm_atomic_helper_swap_state); + +/** + * drm_atomic_helper_update_plane - Helper for primary plane update using atomic + * @plane: plane object to update + * @crtc: owning CRTC of owning plane + * @fb: framebuffer to flip onto plane + * @crtc_x: x offset of primary plane on crtc + * @crtc_y: y offset of primary plane on crtc + * @crtc_w: width of primary plane rectangle on crtc + * @crtc_h: height of primary plane rectangle on crtc + * @src_x: x offset of @fb for panning + * @src_y: y offset of @fb for panning + * @src_w: width of source rectangle in @fb + * @src_h: height of source rectangle in @fb + * + * Provides a default plane update handler using the atomic driver interface. + * + * RETURNS: + * Zero on success, error code on failure + */ +int drm_atomic_helper_update_plane(struct drm_plane *plane, + struct drm_crtc *crtc, + struct drm_framebuffer *fb, + int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t src_x, uint32_t src_y, + uint32_t src_w, uint32_t src_h) +{ + struct drm_atomic_state *state; + struct drm_plane_state *plane_state; + int ret = 0; + + state = drm_atomic_state_alloc(plane->dev); + if (!state) + return -ENOMEM; + + state->acquire_ctx = drm_modeset_legacy_acquire_ctx(crtc); +retry: + plane_state = drm_atomic_get_plane_state(state, plane); + if (IS_ERR(plane_state)) { + ret = PTR_ERR(plane_state); + goto fail; + } + + ret = drm_atomic_set_crtc_for_plane(plane_state, crtc); + if (ret != 0) + goto fail; + plane_state->fb = fb; + plane_state->crtc_x = crtc_x; + plane_state->crtc_y = crtc_y; + plane_state->crtc_h = crtc_h; + plane_state->crtc_w = crtc_w; + plane_state->src_x = src_x; + plane_state->src_y = src_y; + plane_state->src_h = src_h; + plane_state->src_w = src_w; + + ret = drm_atomic_commit(state); + if (ret != 0) + goto fail; + + /* Driver takes ownership of state on successful commit. */ + return 0; +fail: + if (ret == -EDEADLK) + goto backoff; + + drm_atomic_state_free(state); + + return ret; +backoff: + drm_atomic_legacy_backoff(state); + drm_atomic_state_clear(state); + + /* + * Someone might have exchanged the framebuffer while we dropped locks + * in the backoff code. We need to fix up the fb refcount tracking the + * core does for us. + */ + plane->old_fb = plane->fb; + + goto retry; +} +EXPORT_SYMBOL(drm_atomic_helper_update_plane); + +/** + * drm_atomic_helper_disable_plane - Helper for primary plane disable using * atomic + * @plane: plane to disable + * + * Provides a default plane disable handler using the atomic driver interface. + * + * RETURNS: + * Zero on success, error code on failure + */ +int drm_atomic_helper_disable_plane(struct drm_plane *plane) +{ + struct drm_atomic_state *state; + struct drm_plane_state *plane_state; + int ret = 0; + + state = drm_atomic_state_alloc(plane->dev); + if (!state) + return -ENOMEM; + + state->acquire_ctx = drm_modeset_legacy_acquire_ctx(plane->crtc); +retry: + plane_state = drm_atomic_get_plane_state(state, plane); + if (IS_ERR(plane_state)) { + ret = PTR_ERR(plane_state); + goto fail; + } + + ret = drm_atomic_set_crtc_for_plane(plane_state, NULL); + if (ret != 0) + goto fail; + plane_state->fb = NULL; + plane_state->crtc_x = 0; + plane_state->crtc_y = 0; + plane_state->crtc_h = 0; + plane_state->crtc_w = 0; + plane_state->src_x = 0; + plane_state->src_y = 0; + plane_state->src_h = 0; + plane_state->src_w = 0; + + ret = drm_atomic_commit(state); + if (ret != 0) + goto fail; + + /* Driver takes ownership of state on successful commit. */ + return 0; +fail: + if (ret == -EDEADLK) + goto backoff; + + drm_atomic_state_free(state); + + return ret; +backoff: + drm_atomic_legacy_backoff(state); + drm_atomic_state_clear(state); + + /* + * Someone might have exchanged the framebuffer while we dropped locks + * in the backoff code. We need to fix up the fb refcount tracking the + * core does for us. + */ + plane->old_fb = plane->fb; + + goto retry; +} +EXPORT_SYMBOL(drm_atomic_helper_disable_plane); + +static int update_output_state(struct drm_atomic_state *state, + struct drm_mode_set *set) +{ + struct drm_device *dev = set->crtc->dev; + struct drm_connector_state *conn_state; + int nconnectors = state->dev->mode_config.num_connector; + int ncrtcs = state->dev->mode_config.num_crtc; + int ret, i, j; + + ret = drm_modeset_lock(&dev->mode_config.connection_mutex, + state->acquire_ctx); + if (ret) + return ret; + + /* First grab all affected connector/crtc states. */ + for (i = 0; i < set->num_connectors; i++) { + conn_state = drm_atomic_get_connector_state(state, + set->connectors[i]); + if (IS_ERR(conn_state)) + return PTR_ERR(conn_state); + } + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc *crtc = state->crtcs[i]; + + if (!crtc) + continue; + + ret = drm_atomic_add_affected_connectors(state, crtc); + if (ret) + return ret; + } + + /* Then recompute connector->crtc links and crtc enabling state. */ + for (i = 0; i < nconnectors; i++) { + struct drm_connector *connector; + + connector = state->connectors[i]; + conn_state = state->connector_states[i]; + + if (!connector) + continue; + + if (conn_state->crtc == set->crtc) { + ret = drm_atomic_set_crtc_for_connector(conn_state, + NULL); + if (ret) + return ret; + } + + for (j = 0; j < set->num_connectors; j++) { + if (set->connectors[j] == connector) { + ret = drm_atomic_set_crtc_for_connector(conn_state, + set->crtc); + if (ret) + return ret; + break; + } + } + } + + for (i = 0; i < ncrtcs; i++) { + struct drm_crtc *crtc = state->crtcs[i]; + struct drm_crtc_state *crtc_state = state->crtc_states[i]; + + if (!crtc) + continue; + + /* Don't update ->enable for the CRTC in the set_config request, + * since a mismatch would indicate a bug in the upper layers. + * The actual modeset code later on will catch any + * inconsistencies here. */ + if (crtc == set->crtc) + continue; + + crtc_state->enable = + drm_atomic_connectors_for_crtc(state, crtc); + } + + return 0; +} + +/** + * drm_atomic_helper_set_config - set a new config from userspace + * @set: mode set configuration + * + * Provides a default crtc set_config handler using the atomic driver interface. + * + * Returns: + * Returns 0 on success, negative errno numbers on failure. + */ +int drm_atomic_helper_set_config(struct drm_mode_set *set) +{ + struct drm_atomic_state *state; + struct drm_crtc *crtc = set->crtc; + struct drm_crtc_state *crtc_state; + struct drm_plane_state *primary_state; + int ret = 0; + + state = drm_atomic_state_alloc(crtc->dev); + if (!state) + return -ENOMEM; + + state->acquire_ctx = drm_modeset_legacy_acquire_ctx(crtc); +retry: + crtc_state = drm_atomic_get_crtc_state(state, crtc); + if (IS_ERR(crtc_state)) { + ret = PTR_ERR(crtc_state); + goto fail; + } + + if (!set->mode) { + WARN_ON(set->fb); + WARN_ON(set->num_connectors); + + crtc_state->enable = false; + goto commit; + } + + WARN_ON(!set->fb); + WARN_ON(!set->num_connectors); + + crtc_state->enable = true; + drm_mode_copy(&crtc_state->mode, set->mode); + + primary_state = drm_atomic_get_plane_state(state, crtc->primary); + if (IS_ERR(primary_state)) { + ret = PTR_ERR(primary_state); + goto fail; + } + + ret = drm_atomic_set_crtc_for_plane(primary_state, crtc); + if (ret != 0) + goto fail; + primary_state->fb = set->fb; + primary_state->crtc_x = 0; + primary_state->crtc_y = 0; + primary_state->crtc_h = set->mode->vdisplay; + primary_state->crtc_w = set->mode->hdisplay; + primary_state->src_x = set->x << 16; + primary_state->src_y = set->y << 16; + primary_state->src_h = set->mode->vdisplay << 16; + primary_state->src_w = set->mode->hdisplay << 16; + +commit: + ret = update_output_state(state, set); + if (ret) + goto fail; + + ret = drm_atomic_commit(state); + if (ret != 0) + goto fail; + + /* Driver takes ownership of state on successful commit. */ + return 0; +fail: + if (ret == -EDEADLK) + goto backoff; + + drm_atomic_state_free(state); + + return ret; +backoff: + drm_atomic_legacy_backoff(state); + drm_atomic_state_clear(state); + + /* + * Someone might have exchanged the framebuffer while we dropped locks + * in the backoff code. We need to fix up the fb refcount tracking the + * core does for us. + */ + crtc->primary->old_fb = crtc->primary->fb; + + goto retry; +} +EXPORT_SYMBOL(drm_atomic_helper_set_config); + +/** + * drm_atomic_helper_crtc_set_property - helper for crtc prorties + * @crtc: DRM crtc + * @property: DRM property + * @val: value of property + * + * Provides a default plane disablle handler using the atomic driver interface. + * + * RETURNS: + * Zero on success, error code on failure + */ +int +drm_atomic_helper_crtc_set_property(struct drm_crtc *crtc, + struct drm_property *property, + uint64_t val) +{ + struct drm_atomic_state *state; + struct drm_crtc_state *crtc_state; + int ret = 0; + + state = drm_atomic_state_alloc(crtc->dev); + if (!state) + return -ENOMEM; + + /* ->set_property is always called with all locks held. */ + state->acquire_ctx = crtc->dev->mode_config.acquire_ctx; +retry: + crtc_state = drm_atomic_get_crtc_state(state, crtc); + if (IS_ERR(crtc_state)) { + ret = PTR_ERR(crtc_state); + goto fail; + } + + ret = crtc->funcs->atomic_set_property(crtc, crtc_state, + property, val); + if (ret) + goto fail; + + ret = drm_atomic_commit(state); + if (ret != 0) + goto fail; + + /* Driver takes ownership of state on successful commit. */ + return 0; +fail: + if (ret == -EDEADLK) + goto backoff; + + drm_atomic_state_free(state); + + return ret; +backoff: + drm_atomic_legacy_backoff(state); + drm_atomic_state_clear(state); + + goto retry; +} +EXPORT_SYMBOL(drm_atomic_helper_crtc_set_property); + +/** + * drm_atomic_helper_plane_set_property - helper for plane prorties + * @plane: DRM plane + * @property: DRM property + * @val: value of property + * + * Provides a default plane disable handler using the atomic driver interface. + * + * RETURNS: + * Zero on success, error code on failure + */ +int +drm_atomic_helper_plane_set_property(struct drm_plane *plane, + struct drm_property *property, + uint64_t val) +{ + struct drm_atomic_state *state; + struct drm_plane_state *plane_state; + int ret = 0; + + state = drm_atomic_state_alloc(plane->dev); + if (!state) + return -ENOMEM; + + /* ->set_property is always called with all locks held. */ + state->acquire_ctx = plane->dev->mode_config.acquire_ctx; +retry: + plane_state = drm_atomic_get_plane_state(state, plane); + if (IS_ERR(plane_state)) { + ret = PTR_ERR(plane_state); + goto fail; + } + + ret = plane->funcs->atomic_set_property(plane, plane_state, + property, val); + if (ret) + goto fail; + + ret = drm_atomic_commit(state); + if (ret != 0) + goto fail; + + /* Driver takes ownership of state on successful commit. */ + return 0; +fail: + if (ret == -EDEADLK) + goto backoff; + + drm_atomic_state_free(state); + + return ret; +backoff: + drm_atomic_legacy_backoff(state); + drm_atomic_state_clear(state); + + goto retry; +} +EXPORT_SYMBOL(drm_atomic_helper_plane_set_property); + +/** + * drm_atomic_helper_connector_set_property - helper for connector prorties + * @connector: DRM connector + * @property: DRM property + * @val: value of property + * + * Provides a default plane disablle handler using the atomic driver interface. + * + * RETURNS: + * Zero on success, error code on failure + */ +int +drm_atomic_helper_connector_set_property(struct drm_connector *connector, + struct drm_property *property, + uint64_t val) +{ + struct drm_atomic_state *state; + struct drm_connector_state *connector_state; + int ret = 0; + + state = drm_atomic_state_alloc(connector->dev); + if (!state) + return -ENOMEM; + + /* ->set_property is always called with all locks held. */ + state->acquire_ctx = connector->dev->mode_config.acquire_ctx; +retry: + connector_state = drm_atomic_get_connector_state(state, connector); + if (IS_ERR(connector_state)) { + ret = PTR_ERR(connector_state); + goto fail; + } + + ret = connector->funcs->atomic_set_property(connector, connector_state, + property, val); + if (ret) + goto fail; + + ret = drm_atomic_commit(state); + if (ret != 0) + goto fail; + + /* Driver takes ownership of state on successful commit. */ + return 0; +fail: + if (ret == -EDEADLK) + goto backoff; + + drm_atomic_state_free(state); + + return ret; +backoff: + drm_atomic_legacy_backoff(state); + drm_atomic_state_clear(state); + + goto retry; +} +EXPORT_SYMBOL(drm_atomic_helper_connector_set_property); diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h index 9781ce7..8cd6fe7 100644 --- a/include/drm/drm_atomic_helper.h +++ b/include/drm/drm_atomic_helper.h @@ -49,4 +49,25 @@ void drm_atomic_helper_cleanup_planes(struct drm_device *dev, void drm_atomic_helper_swap_state(struct drm_device *dev, struct drm_atomic_state *state); +/* implementations for legacy interfaces */ +int drm_atomic_helper_update_plane(struct drm_plane *plane, + struct drm_crtc *crtc, + struct drm_framebuffer *fb, + int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t src_x, uint32_t src_y, + uint32_t src_w, uint32_t src_h); +int drm_atomic_helper_disable_plane(struct drm_plane *plane); +int drm_atomic_helper_set_config(struct drm_mode_set *set); + +int drm_atomic_helper_crtc_set_property(struct drm_crtc *crtc, + struct drm_property *property, + uint64_t val); +int drm_atomic_helper_plane_set_property(struct drm_plane *plane, + struct drm_property *property, + uint64_t val); +int drm_atomic_helper_connector_set_property(struct drm_connector *connector, + struct drm_property *property, + uint64_t val); + #endif /* DRM_ATOMIC_HELPER_H_ */ -- cgit v0.10.2 From e2330f0719515e41090a4d9685b931888b7c01d6 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 29 Oct 2014 11:34:56 +0100 Subject: drm/atomic: Integrate fence support This patch is for enabling async commits. It replaces an earlier approach which added an async boolean paramter to the ->prepare_fb callbacks. The idea is that prepare_fb picks up the right fence to synchronize against, which is then used by the synchronous commit helper. For async commits drivers can either register a callback to the fence or simply do the synchronous wait in their async work queue. v2: Remove unused variable. v3: Only wait for fences after the point of no return in the part of the commit function which can be run asynchronously. This is after the atomic state has been swapped in, hence now check plane->state->fence. Also add a WARN_ON to make sure we don't try to wait on a fence when there's no fb, just as a sanity check. Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 077c792..4808e71 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -30,6 +30,7 @@ #include #include #include +#include static void drm_atomic_helper_plane_changed(struct drm_atomic_state *state, @@ -712,6 +713,26 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev, } EXPORT_SYMBOL(drm_atomic_helper_commit_post_planes); +static void wait_for_fences(struct drm_device *dev, + struct drm_atomic_state *state) +{ + int nplanes = dev->mode_config.num_total_plane; + int i; + + for (i = 0; i < nplanes; i++) { + struct drm_plane *plane = state->planes[i]; + + if (!plane || !plane->state->fence) + continue; + + WARN_ON(!plane->state->fb); + + fence_wait(plane->state->fence, false); + fence_put(plane->state->fence); + plane->state->fence = NULL; + } +} + static void wait_for_vblanks(struct drm_device *dev, struct drm_atomic_state *old_state) { @@ -809,6 +830,8 @@ int drm_atomic_helper_commit(struct drm_device *dev, * current layout. */ + wait_for_fences(dev, state); + drm_atomic_helper_commit_pre_planes(dev, state); drm_atomic_helper_commit_planes(dev, state); diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index bc47d11..bc1cc3c 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -42,6 +42,7 @@ struct drm_object_properties; struct drm_file; struct drm_clip_rect; struct device_node; +struct fence; #define DRM_MODE_OBJECT_CRTC 0xcccccccc #define DRM_MODE_OBJECT_CONNECTOR 0xc0c0c0c0 @@ -658,6 +659,7 @@ struct drm_connector { * struct drm_plane_state - mutable plane state * @crtc: currently bound CRTC, NULL if disabled * @fb: currently bound framebuffer + * @fence: optional fence to wait for before scanning out @fb * @crtc_x: left position of visible portion of plane on crtc * @crtc_y: upper position of visible portion of plane on crtc * @crtc_w: width of visible portion of plane on crtc @@ -673,6 +675,7 @@ struct drm_connector { struct drm_plane_state { struct drm_crtc *crtc; struct drm_framebuffer *fb; + struct fence *fence; /* Signed dest location allows it to be partially off screen */ int32_t crtc_x, crtc_y; -- cgit v0.10.2 From e8c833a7dfdbf47934537e19c3a61c2c98409afd Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Sun, 27 Jul 2014 18:30:19 +0200 Subject: drm/atomic-helpers: document how to implement async commit No helper function to do it all yet provided since no driver has support for driver core fences yet. Which we'd need to make the implementation really generic. v2: Clarify async howto a bit per the discussion With Rob Clark. Cc: Rob Clark Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 4808e71..6668d2c 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -849,6 +849,44 @@ int drm_atomic_helper_commit(struct drm_device *dev, EXPORT_SYMBOL(drm_atomic_helper_commit); /** + * DOC: implementing async commit + * + * For now the atomic helpers don't support async commit directly. If there is + * real need it could be added though, using the dma-buf fence infrastructure + * for generic synchronization with outstanding rendering. + * + * For now drivers have to implement async commit themselves, with the following + * sequence being the recommended one: + * + * 1. Run drm_atomic_helper_prepare_planes() first. This is the only function + * which commit needs to call which can fail, so we want to run it first and + * synchronously. + * + * 2. Synchronize with any outstanding asynchronous commit worker threads which + * might be affected the new state update. This can be done by either cancelling + * or flushing the work items, depending upon whether the driver can deal with + * cancelled updates. Note that it is important to ensure that the framebuffer + * cleanup is still done when cancelling. + * + * For sufficient parallelism it is recommended to have a work item per crtc + * (for updates which don't touch global state) and a global one. Then we only + * need to synchronize with the crtc work items for changed crtcs and the global + * work item, which allows nice concurrent updates on disjoint sets of crtcs. + * + * 3. The software state is updated synchronously with + * drm_atomic_helper_swap_state. Doing this under the protection of all modeset + * locks means concurrent callers never see inconsistent state. And doing this + * while it's guaranteed that no relevant async worker runs means that async + * workers do not need grab any locks. Actually they must not grab locks, for + * otherwise the work flushing will deadlock. + * + * 4. Schedule a work item to do all subsequent steps, using the split-out + * commit helpers: a) pre-plane commit b) plane commit c) post-plane commit and + * then cleaning up the framebuffers after the old framebuffer is no longer + * being displayed. + */ + +/** * drm_atomic_helper_prepare_planes - prepare plane resources after commit * @dev: DRM device * @state: atomic state object with old state structures -- cgit v0.10.2 From 8bc0f3126cc5d37c98d213171e71626536fd1e86 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Sun, 27 Jul 2014 18:42:37 +0200 Subject: drm/atomic-helper: implement ->page_flip Currently there is no way to implement async flips using atomic, that essentially requires us to be able to cancel pending requests mid-flight. To be able to do that (and I guess we want this since vblank synced updates which opportunistically cancel still pending updates seem to be wanted) we'd need to add a mandatory cancellation mode. Depending upon the exact semantics we decide upon that could mean that userspace will not get completion events, or will get them all stacked up. So reject async updates for now. Also async updates usually means not vblank synced at all, and I guess for drivers which want to support this they should simply add a special pageflip handler (since usually you need a special flip cmd to achieve this). That kind of async flip is pretty much exclusively just used for games and benchmarks where dropping just one frame means you'll get a headshot or something bad like that ... And so slight amounts of tearing is acceptable. v2: Fixup kerneldoc, reported by Paulo. v3: Use the set_crtc_for_plane function to assign the crtc, since otherwise the book-keeping is off. v4: Update crtc->primary->fb since ->page_flip is the only driver callback where the core won't do this itself. We might want to fix this inconsistency eventually. v5: Use set_crtc_for_connector as suggested by Sean. v6: Daniel Thompson noticed that my error handling is inconsistent and that in a few cases I didn't handle fatal errors (i.e. not -EDEADLK). Fix this by consolidate the ww mutex backoff handling into one check in the fail: block and flatten the error control flow everywhere else. v7: Fix spelling mistake in the commit message (Sean). Cc: Daniel Thompson Cc: Sean Paul Cc: Paulo Zanoni Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 6668d2c..a3216a7 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -1620,3 +1620,89 @@ backoff: goto retry; } EXPORT_SYMBOL(drm_atomic_helper_connector_set_property); + +/** + * drm_atomic_helper_page_flip - execute a legacy page flip + * @crtc: DRM crtc + * @fb: DRM framebuffer + * @event: optional DRM event to signal upon completion + * @flags: flip flags for non-vblank sync'ed updates + * + * Provides a default page flip implementation using the atomic driver interface. + * + * Note that for now so called async page flips (i.e. updates which are not + * synchronized to vblank) are not supported, since the atomic interfaces have + * no provisions for this yet. + * + * Returns: + * Returns 0 on success, negative errno numbers on failure. + */ +int drm_atomic_helper_page_flip(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_pending_vblank_event *event, + uint32_t flags) +{ + struct drm_plane *plane = crtc->primary; + struct drm_atomic_state *state; + struct drm_plane_state *plane_state; + struct drm_crtc_state *crtc_state; + int ret = 0; + + if (flags & DRM_MODE_PAGE_FLIP_ASYNC) + return -EINVAL; + + state = drm_atomic_state_alloc(plane->dev); + if (!state) + return -ENOMEM; + + state->acquire_ctx = drm_modeset_legacy_acquire_ctx(crtc); +retry: + crtc_state = drm_atomic_get_crtc_state(state, crtc); + if (IS_ERR(crtc_state)) { + ret = PTR_ERR(crtc_state); + goto fail; + } + crtc_state->event = event; + + plane_state = drm_atomic_get_plane_state(state, plane); + if (IS_ERR(plane_state)) { + ret = PTR_ERR(plane_state); + goto fail; + } + + ret = drm_atomic_set_crtc_for_plane(plane_state, crtc); + if (ret != 0) + goto fail; + plane_state->fb = fb; + + ret = drm_atomic_async_commit(state); + if (ret != 0) + goto fail; + + /* TODO: ->page_flip is the only driver callback where the core + * doesn't update plane->fb. For now patch it up here. */ + plane->fb = plane->state->fb; + + /* Driver takes ownership of state on successful async commit. */ + return 0; +fail: + if (ret == -EDEADLK) + goto backoff; + + drm_atomic_state_free(state); + + return ret; +backoff: + drm_atomic_legacy_backoff(state); + drm_atomic_state_clear(state); + + /* + * Someone might have exchanged the framebuffer while we dropped locks + * in the backoff code. We need to fix up the fb refcount tracking the + * core does for us. + */ + plane->old_fb = plane->fb; + + goto retry; +} +EXPORT_SYMBOL(drm_atomic_helper_page_flip); diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h index 8cd6fe7..28a2f3a 100644 --- a/include/drm/drm_atomic_helper.h +++ b/include/drm/drm_atomic_helper.h @@ -69,5 +69,10 @@ int drm_atomic_helper_plane_set_property(struct drm_plane *plane, int drm_atomic_helper_connector_set_property(struct drm_connector *connector, struct drm_property *property, uint64_t val); +int drm_atomic_helper_page_flip(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_pending_vblank_event *event, + uint32_t flags); + #endif /* DRM_ATOMIC_HELPER_H_ */ -- cgit v0.10.2 From d461701c551efe92893b31d05906d51e2f1f8244 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 3 Nov 2014 15:56:43 +0100 Subject: drm/atomic-helpers: functions for state duplicate/destroy/reset The atomic users and helpers assume that there is always a obj->state structure around. Which means drivers need to somehow create that at driver load time. Also it should obviously reset hardware state, so needs to be reset upon resume. Finally the destroy/duplicate_state functions are an awful lot of boilerplate if the driver doesn't need anything beyond the default state objects. So add helper functions for all of this. v2: Somehow the plane/connector versions got lost in the first version. v3: Add kerneldoc. v4: Make duplicate_state functions a bit more robust, which is useful for debugging state tracking issues when transitioning to atomic. v5: Clear temporary variables in the crtc state when duplicating it, like ->mode_changed or ->planes_changed. If we don't do this stale values for these might pollute the next atomic modeset. v6: Also clear crtc_state->event in case the driver didn't (yet) clear this out. v7: Split out wrong squashed commit. Also improve the kerneldoc to mention that obj->state can be NULL and when. Both suggested by Daniel Thompson. Cc: Daniel Thompson Cc: Sean Paul Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index a3216a7..2c38bda 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -1706,3 +1706,152 @@ backoff: goto retry; } EXPORT_SYMBOL(drm_atomic_helper_page_flip); + +/** + * drm_atomic_helper_crtc_reset - default ->reset hook for CRTCs + * @crtc: drm CRTC + * + * Resets the atomic state for @crtc by freeing the state pointer (which might + * be NULL, e.g. at driver load time) and allocating a new empty state object. + */ +void drm_atomic_helper_crtc_reset(struct drm_crtc *crtc) +{ + kfree(crtc->state); + crtc->state = kzalloc(sizeof(*crtc->state), GFP_KERNEL); +} +EXPORT_SYMBOL(drm_atomic_helper_crtc_reset); + +/** + * drm_atomic_helper_crtc_duplicate_state - default state duplicate hook + * @crtc: drm CRTC + * + * Default CRTC state duplicate hook for drivers which don't have their own + * subclassed CRTC state structure. + */ +struct drm_crtc_state * +drm_atomic_helper_crtc_duplicate_state(struct drm_crtc *crtc) +{ + struct drm_crtc_state *state; + + if (WARN_ON(!crtc->state)) + return NULL; + + state = kmemdup(crtc->state, sizeof(*crtc->state), GFP_KERNEL); + + if (state) { + state->mode_changed = false; + state->planes_changed = false; + state->event = NULL; + } + + return state; +} +EXPORT_SYMBOL(drm_atomic_helper_crtc_duplicate_state); + +/** + * drm_atomic_helper_crtc_destroy_state - default state destroy hook + * @crtc: drm CRTC + * @state: CRTC state object to release + * + * Default CRTC state destroy hook for drivers which don't have their own + * subclassed CRTC state structure. + */ +void drm_atomic_helper_crtc_destroy_state(struct drm_crtc *crtc, + struct drm_crtc_state *state) +{ + kfree(state); +} +EXPORT_SYMBOL(drm_atomic_helper_crtc_destroy_state); + +/** + * drm_atomic_helper_plane_reset - default ->reset hook for planes + * @plane: drm plane + * + * Resets the atomic state for @plane by freeing the state pointer (which might + * be NULL, e.g. at driver load time) and allocating a new empty state object. + */ +void drm_atomic_helper_plane_reset(struct drm_plane *plane) +{ + kfree(plane->state); + plane->state = kzalloc(sizeof(*plane->state), GFP_KERNEL); +} +EXPORT_SYMBOL(drm_atomic_helper_plane_reset); + +/** + * drm_atomic_helper_plane_duplicate_state - default state duplicate hook + * @plane: drm plane + * + * Default plane state duplicate hook for drivers which don't have their own + * subclassed plane state structure. + */ +struct drm_plane_state * +drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane) +{ + if (WARN_ON(!plane->state)) + return NULL; + + return kmemdup(plane->state, sizeof(*plane->state), GFP_KERNEL); +} +EXPORT_SYMBOL(drm_atomic_helper_plane_duplicate_state); + +/** + * drm_atomic_helper_plane_destroy_state - default state destroy hook + * @plane: drm plane + * @state: plane state object to release + * + * Default plane state destroy hook for drivers which don't have their own + * subclassed plane state structure. + */ +void drm_atomic_helper_plane_destroy_state(struct drm_plane *plane, + struct drm_plane_state *state) +{ + kfree(state); +} +EXPORT_SYMBOL(drm_atomic_helper_plane_destroy_state); + +/** + * drm_atomic_helper_connector_reset - default ->reset hook for connectors + * @connector: drm connector + * + * Resets the atomic state for @connector by freeing the state pointer (which + * might be NULL, e.g. at driver load time) and allocating a new empty state + * object. + */ +void drm_atomic_helper_connector_reset(struct drm_connector *connector) +{ + kfree(connector->state); + connector->state = kzalloc(sizeof(*connector->state), GFP_KERNEL); +} +EXPORT_SYMBOL(drm_atomic_helper_connector_reset); + +/** + * drm_atomic_helper_connector_duplicate_state - default state duplicate hook + * @connector: drm connector + * + * Default connector state duplicate hook for drivers which don't have their own + * subclassed connector state structure. + */ +struct drm_connector_state * +drm_atomic_helper_connector_duplicate_state(struct drm_connector *connector) +{ + if (WARN_ON(!connector->state)) + return NULL; + + return kmemdup(connector->state, sizeof(*connector->state), GFP_KERNEL); +} +EXPORT_SYMBOL(drm_atomic_helper_connector_duplicate_state); + +/** + * drm_atomic_helper_connector_destroy_state - default state destroy hook + * @connector: drm connector + * @state: connector state object to release + * + * Default connector state destroy hook for drivers which don't have their own + * subclassed connector state structure. + */ +void drm_atomic_helper_connector_destroy_state(struct drm_connector *connector, + struct drm_connector_state *state) +{ + kfree(state); +} +EXPORT_SYMBOL(drm_atomic_helper_connector_destroy_state); diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h index 28a2f3a..67e3c46 100644 --- a/include/drm/drm_atomic_helper.h +++ b/include/drm/drm_atomic_helper.h @@ -74,5 +74,24 @@ int drm_atomic_helper_page_flip(struct drm_crtc *crtc, struct drm_pending_vblank_event *event, uint32_t flags); +/* default implementations for state handling */ +void drm_atomic_helper_crtc_reset(struct drm_crtc *crtc); +struct drm_crtc_state * +drm_atomic_helper_crtc_duplicate_state(struct drm_crtc *crtc); +void drm_atomic_helper_crtc_destroy_state(struct drm_crtc *crtc, + struct drm_crtc_state *state); + +void drm_atomic_helper_plane_reset(struct drm_plane *plane); +struct drm_plane_state * +drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane); +void drm_atomic_helper_plane_destroy_state(struct drm_plane *plane, + struct drm_plane_state *state); + +void drm_atomic_helper_connector_reset(struct drm_connector *connector); +struct drm_connector_state * +drm_atomic_helper_connector_duplicate_state(struct drm_connector *connector); +void drm_atomic_helper_connector_destroy_state(struct drm_connector *connector, + struct drm_connector_state *state); + #endif /* DRM_ATOMIC_HELPER_H_ */ -- cgit v0.10.2 From 3150c7d0c686ffee9e414ccce705c34ebcbf4e30 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Thu, 6 Nov 2014 20:53:29 +0100 Subject: drm: Docbook integration and over sections for all the new helpers In all cases the text requires that new drivers are converted to the atomic interfaces. v2: Add overview for state handling. v3: Review from Sean: Some spelling fixes and drop the misguided hunk to remove rgba8888 from the plane helpers compat list. Cc: Sean Paul Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index 748513b..f276c2c 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -2324,8 +2324,25 @@ void intel_crt_init(struct drm_device *dev) + Atomic Modeset Helper Functions Reference + + Overview +!Pdrivers/gpu/drm/drm_atomic_helper.c overview + + + Implementing Asynchronous Atomic Commit +!Pdrivers/gpu/drm/drm_atomic_helper.c implementing async commit + + + Atomic State Reset and Initialization +!Pdrivers/gpu/drm/drm_atomic_helper.c atomic state reset and initialization + +!Edrivers/gpu/drm/drm_atomic_helper.c + + Modeset Helper Functions Reference !Edrivers/gpu/drm/drm_crtc_helper.c +!Pdrivers/gpu/drm/drm_crtc_helper.c overview Output Probing Helper Functions Reference @@ -2379,7 +2396,8 @@ void intel_crt_init(struct drm_device *dev) Plane Helper Reference -!Edrivers/gpu/drm/drm_plane_helper.c Plane Helpers +!Edrivers/gpu/drm/drm_plane_helper.c +!Pdrivers/gpu/drm/drm_plane_helper.c overview diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 2c38bda..2b1db0c 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -32,6 +32,27 @@ #include #include +/** + * DOC: overview + * + * This helper library provides implementations of check and commit functions on + * top of the CRTC modeset helper callbacks and the plane helper callbacks. It + * also provides convenience implementations for the atomic state handling + * callbacks for drivers which don't need to subclass the drm core structures to + * add their own additional internal state. + * + * This library also provides default implementations for the check callback in + * drm_atomic_helper_check and for the commit callback with + * drm_atomic_helper_commit. But the individual stages and callbacks are expose + * to allow drivers to mix and match and e.g. use the plane helpers only + * together with a driver private modeset implementation. + * + * This library also provides implementations for all the legacy driver + * interfaces on top of the atomic interface. See drm_atomic_helper_set_config, + * drm_atomic_helper_disable_plane, drm_atomic_helper_disable_plane and the + * various functions to implement set_property callbacks. New drivers must not + * implement these functions themselves but must use the provided helpers. + */ static void drm_atomic_helper_plane_changed(struct drm_atomic_state *state, struct drm_plane_state *plane_state, @@ -1708,6 +1729,21 @@ backoff: EXPORT_SYMBOL(drm_atomic_helper_page_flip); /** + * DOC: atomic state reset and initialization + * + * Both the drm core and the atomic helpers assume that there is always the full + * and correct atomic software state for all connectors, CRTCs and planes + * available. Which is a bit a problem on driver load and also after system + * suspend. One way to solve this is to have a hardware state read-out + * infrastructure which reconstructs the full software state (e.g. the i915 + * driver). + * + * The simpler solution is to just reset the software state to everything off, + * which is easiest to do by calling drm_mode_config_reset(). To facilitate this + * the atomic helpers provide default reset implementations for all hooks. + */ + +/** * drm_atomic_helper_crtc_reset - default ->reset hook for CRTCs * @crtc: drm CRTC * diff --git a/drivers/gpu/drm/drm_crtc_helper.c b/drivers/gpu/drm/drm_crtc_helper.c index 46728a8..33195e9 100644 --- a/drivers/gpu/drm/drm_crtc_helper.c +++ b/drivers/gpu/drm/drm_crtc_helper.c @@ -41,6 +41,26 @@ #include #include +/** + * DOC: overview + * + * The CRTC modeset helper library provides a default set_config implementation + * in drm_crtc_helper_set_config(). Plus a few other convenience functions using + * the same callbacks which drivers can use to e.g. restore the modeset + * configuration on resume with drm_helper_resume_force_mode(). + * + * The driver callbacks are mostly compatible with the atomic modeset helpers, + * except for the handling of the primary plane: Atomic helpers require that the + * primary plane is implemented as a real standalone plane and not directly tied + * to the CRTC state. For easier transition this library provides functions to + * implement the old semantics required by the CRTC helpers using the new plane + * and atomic helper callbacks. + * + * Drivers are strongly urged to convert to the atomic helpers (by way of first + * converting to the plane helpers). New drivers must not use these functions + * but need to implement the atomic interface instead, potentially using the + * atomic helpers for that. + */ MODULE_AUTHOR("David Airlie, Jesse Barnes"); MODULE_DESCRIPTION("DRM KMS helper"); MODULE_LICENSE("GPL and additional rights"); diff --git a/drivers/gpu/drm/drm_plane_helper.c b/drivers/gpu/drm/drm_plane_helper.c index a5a2958..fa56bb5 100644 --- a/drivers/gpu/drm/drm_plane_helper.c +++ b/drivers/gpu/drm/drm_plane_helper.c @@ -31,6 +31,32 @@ #define SUBPIXEL_MASK 0xffff +/** + * DOC: overview + * + * This helper library has two parts. The first part has support to implement + * primary plane support on top of the normal CRTC configuration interface. + * Since the legacy ->set_config interface ties the primary plane together with + * the CRTC state this does not allow userspace to disable the primary plane + * itself. To avoid too much duplicated code use + * drm_plane_helper_check_update() which can be used to enforce the same + * restrictions as primary planes had thus. The default primary plane only + * expose XRBG8888 and ARGB8888 as valid pixel formats for the attached + * framebuffer. + * + * Drivers are highly recommended to implement proper support for primary + * planes, and newly merged drivers must not rely upon these transitional + * helpers. + * + * The second part also implements transitional helpers which allow drivers to + * gradually switch to the atomic helper infrastructure for plane updates. Once + * that switch is complete drivers shouldn't use these any longer, instead using + * the proper legacy implementations for update and disable plane hooks provided + * by the atomic helpers. + * + * Again drivers are strongly urged to switch to the new interfaces. + */ + /* * This is the minimal list of formats that seem to be safe for modeset use * with all current DRM drivers. Most hardware can actually support more -- cgit v0.10.2 From 321ebf04dc7ab5c54d658f93db0ffe35277664ab Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 4 Nov 2014 22:57:27 +0100 Subject: drm/atomic: Refcounting for plane_state->fb So my original plan was that the drm core refcounts framebuffers like with the legacy ioctls. But that doesn't work for a bunch of reasons: - State objects might live longer than until the next fb change happens for a plane. For example delayed cleanup work only happens _after_ the pageflip ioctl has completed. So this definitely doesn't work without the plane state holding its own references. - The other issue is transition from legacy to atomic implementations, where the driver works under a mix of both worlds. Which means legacy paths might not properly update the ->fb pointer under plane->state->fb. Which is a bit a problem when then someone comes around and _does_ try to clean it up when it's long gone. The second issue is just a bit a transition bug, since drivers should update plane->state->fb in all the paths that aren't converted yet. But a bit more robustness for the transition can't hurt - we pull similar tricks with cleaning up the old fb in the transitional helpers already. The pattern for drivers that transition is if (plane->state) drm_atomic_set_fb_for_plane(plane->state, plane->fb); inserted after the fb update has logically completed at the end of ->set_config (or ->set_base/mode_set if using the crtc helpers), ->page_flip, ->update_plane or any other entry point which updates plane->fb. v2: Update kerneldoc - copypasta fail. v3: Fix spelling in the commit message (Sean). Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index ad15a88..ed991ba 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -368,6 +368,34 @@ drm_atomic_set_crtc_for_plane(struct drm_plane_state *plane_state, EXPORT_SYMBOL(drm_atomic_set_crtc_for_plane); /** + * drm_atomic_set_fb_for_plane - set crtc for plane + * @plane_state: atomic state object for the plane + * @fb: fb to use for the plane + * + * Changing the assigned framebuffer for a plane requires us to grab a reference + * to the new fb and drop the reference to the old fb, if there is one. This + * function takes care of all these details besides updating the pointer in the + * state object itself. + */ +void +drm_atomic_set_fb_for_plane(struct drm_plane_state *plane_state, + struct drm_framebuffer *fb) +{ + if (plane_state->fb) + drm_framebuffer_unreference(plane_state->fb); + if (fb) + drm_framebuffer_reference(fb); + plane_state->fb = fb; + + if (fb) + DRM_DEBUG_KMS("Set [FB:%d] for plane state %p\n", + fb->base.id, plane_state); + else + DRM_DEBUG_KMS("Set [NOFB] for plane state %p\n", plane_state); +} +EXPORT_SYMBOL(drm_atomic_set_fb_for_plane); + +/** * drm_atomic_set_crtc_for_connector - set crtc for connector * @conn_state: atomic state object for the connector * @crtc: crtc to use for the connector diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 2b1db0c..ca839bd 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -1182,7 +1182,7 @@ retry: ret = drm_atomic_set_crtc_for_plane(plane_state, crtc); if (ret != 0) goto fail; - plane_state->fb = fb; + drm_atomic_set_fb_for_plane(plane_state, fb); plane_state->crtc_x = crtc_x; plane_state->crtc_y = crtc_y; plane_state->crtc_h = crtc_h; @@ -1250,7 +1250,7 @@ retry: ret = drm_atomic_set_crtc_for_plane(plane_state, NULL); if (ret != 0) goto fail; - plane_state->fb = NULL; + drm_atomic_set_fb_for_plane(plane_state, NULL); plane_state->crtc_x = 0; plane_state->crtc_y = 0; plane_state->crtc_h = 0; @@ -1422,7 +1422,7 @@ retry: ret = drm_atomic_set_crtc_for_plane(primary_state, crtc); if (ret != 0) goto fail; - primary_state->fb = set->fb; + drm_atomic_set_fb_for_plane(primary_state, set->fb); primary_state->crtc_x = 0; primary_state->crtc_y = 0; primary_state->crtc_h = set->mode->vdisplay; @@ -1694,7 +1694,7 @@ retry: ret = drm_atomic_set_crtc_for_plane(plane_state, crtc); if (ret != 0) goto fail; - plane_state->fb = fb; + drm_atomic_set_fb_for_plane(plane_state, fb); ret = drm_atomic_async_commit(state); if (ret != 0) @@ -1808,6 +1808,9 @@ EXPORT_SYMBOL(drm_atomic_helper_crtc_destroy_state); */ void drm_atomic_helper_plane_reset(struct drm_plane *plane) { + if (plane->state && plane->state->fb) + drm_framebuffer_unreference(plane->state->fb); + kfree(plane->state); plane->state = kzalloc(sizeof(*plane->state), GFP_KERNEL); } @@ -1823,10 +1826,17 @@ EXPORT_SYMBOL(drm_atomic_helper_plane_reset); struct drm_plane_state * drm_atomic_helper_plane_duplicate_state(struct drm_plane *plane) { + struct drm_plane_state *state; + if (WARN_ON(!plane->state)) return NULL; - return kmemdup(plane->state, sizeof(*plane->state), GFP_KERNEL); + state = kmemdup(plane->state, sizeof(*plane->state), GFP_KERNEL); + + if (state && state->fb) + drm_framebuffer_reference(state->fb); + + return state; } EXPORT_SYMBOL(drm_atomic_helper_plane_duplicate_state); @@ -1839,8 +1849,11 @@ EXPORT_SYMBOL(drm_atomic_helper_plane_duplicate_state); * subclassed plane state structure. */ void drm_atomic_helper_plane_destroy_state(struct drm_plane *plane, - struct drm_plane_state *state) + struct drm_plane_state *state) { + if (state->fb) + drm_framebuffer_unreference(state->fb); + kfree(state); } EXPORT_SYMBOL(drm_atomic_helper_plane_destroy_state); diff --git a/drivers/gpu/drm/drm_crtc_helper.c b/drivers/gpu/drm/drm_crtc_helper.c index 33195e9..d552708 100644 --- a/drivers/gpu/drm/drm_crtc_helper.c +++ b/drivers/gpu/drm/drm_crtc_helper.c @@ -34,11 +34,13 @@ #include #include +#include #include #include #include #include #include +#include #include /** @@ -998,15 +1000,14 @@ int drm_helper_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, if (plane->funcs->atomic_duplicate_state) plane_state = plane->funcs->atomic_duplicate_state(plane); else if (plane->state) - plane_state = kmemdup(plane->state, sizeof(*plane_state), - GFP_KERNEL); + plane_state = drm_atomic_helper_plane_duplicate_state(plane); else plane_state = kzalloc(sizeof(*plane_state), GFP_KERNEL); if (!plane_state) return -ENOMEM; plane_state->crtc = crtc; - plane_state->fb = crtc->primary->fb; + drm_atomic_set_fb_for_plane(plane_state, crtc->primary->fb); plane_state->crtc_x = 0; plane_state->crtc_y = 0; plane_state->crtc_h = crtc->mode.vdisplay; diff --git a/drivers/gpu/drm/drm_plane_helper.c b/drivers/gpu/drm/drm_plane_helper.c index fa56bb5..d99c452 100644 --- a/drivers/gpu/drm/drm_plane_helper.c +++ b/drivers/gpu/drm/drm_plane_helper.c @@ -27,7 +27,9 @@ #include #include #include +#include #include +#include #define SUBPIXEL_MASK 0xffff @@ -464,7 +466,7 @@ out: if (plane->funcs->atomic_destroy_state) plane->funcs->atomic_destroy_state(plane, plane_state); else - kfree(plane_state); + drm_atomic_helper_plane_destroy_state(plane, plane_state); } return ret; @@ -505,15 +507,14 @@ int drm_plane_helper_update(struct drm_plane *plane, struct drm_crtc *crtc, if (plane->funcs->atomic_duplicate_state) plane_state = plane->funcs->atomic_duplicate_state(plane); else if (plane->state) - plane_state = kmemdup(plane->state, sizeof(*plane_state), - GFP_KERNEL); + plane_state = drm_atomic_helper_plane_duplicate_state(plane); else plane_state = kzalloc(sizeof(*plane_state), GFP_KERNEL); if (!plane_state) return -ENOMEM; plane_state->crtc = crtc; - plane_state->fb = fb; + drm_atomic_set_fb_for_plane(plane_state, fb); plane_state->crtc_x = crtc_x; plane_state->crtc_y = crtc_y; plane_state->crtc_h = crtc_h; @@ -552,15 +553,14 @@ int drm_plane_helper_disable(struct drm_plane *plane) if (plane->funcs->atomic_duplicate_state) plane_state = plane->funcs->atomic_duplicate_state(plane); else if (plane->state) - plane_state = kmemdup(plane->state, sizeof(*plane_state), - GFP_KERNEL); + plane_state = drm_atomic_helper_plane_duplicate_state(plane); else plane_state = kzalloc(sizeof(*plane_state), GFP_KERNEL); if (!plane_state) return -ENOMEM; plane_state->crtc = NULL; - plane_state->fb = NULL; + drm_atomic_set_fb_for_plane(plane_state, NULL); return drm_plane_helper_commit(plane, plane_state, plane->fb); } diff --git a/include/drm/drm_atomic.h b/include/drm/drm_atomic.h index 5bb15f5..9d91916 100644 --- a/include/drm/drm_atomic.h +++ b/include/drm/drm_atomic.h @@ -46,6 +46,8 @@ drm_atomic_get_connector_state(struct drm_atomic_state *state, int __must_check drm_atomic_set_crtc_for_plane(struct drm_plane_state *plane_state, struct drm_crtc *crtc); +void drm_atomic_set_fb_for_plane(struct drm_plane_state *plane_state, + struct drm_framebuffer *fb); int __must_check drm_atomic_set_crtc_for_connector(struct drm_connector_state *conn_state, struct drm_crtc *crtc); -- cgit v0.10.2 From 5fa2704e011e6d0ebc144188cc5fb02363e12f05 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 5 Nov 2014 11:02:26 +1000 Subject: drm: drop README.drm, ancient scrolls This stuff is ancient, we have docs now in the kernel, lets just drop it. Pointed out by Glenn Reviewed-by: Thierry Reding Reviewed-by: Alex Deucher Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/README.drm b/drivers/gpu/drm/README.drm deleted file mode 100644 index b5b3327..0000000 --- a/drivers/gpu/drm/README.drm +++ /dev/null @@ -1,43 +0,0 @@ -************************************************************ -* For the very latest on DRI development, please see: * -* http://dri.freedesktop.org/ * -************************************************************ - -The Direct Rendering Manager (drm) is a device-independent kernel-level -device driver that provides support for the XFree86 Direct Rendering -Infrastructure (DRI). - -The DRM supports the Direct Rendering Infrastructure (DRI) in four major -ways: - - 1. The DRM provides synchronized access to the graphics hardware via - the use of an optimized two-tiered lock. - - 2. The DRM enforces the DRI security policy for access to the graphics - hardware by only allowing authenticated X11 clients access to - restricted regions of memory. - - 3. The DRM provides a generic DMA engine, complete with multiple - queues and the ability to detect the need for an OpenGL context - switch. - - 4. The DRM is extensible via the use of small device-specific modules - that rely extensively on the API exported by the DRM module. - - -Documentation on the DRI is available from: - http://dri.freedesktop.org/wiki/Documentation - http://sourceforge.net/project/showfiles.php?group_id=387 - http://dri.sourceforge.net/doc/ - -For specific information about kernel-level support, see: - - The Direct Rendering Manager, Kernel Support for the Direct Rendering - Infrastructure - http://dri.sourceforge.net/doc/drm_low_level.html - - Hardware Locking for the Direct Rendering Infrastructure - http://dri.sourceforge.net/doc/hardware_locking_low_level.html - - A Security Analysis of the Direct Rendering Infrastructure - http://dri.sourceforge.net/doc/security_low_level.html -- cgit v0.10.2 From 929059250a6795da9a8bd91b8d7a4d9379030f49 Mon Sep 17 00:00:00 2001 From: Philipp Zabel Date: Thu, 23 Oct 2014 16:31:05 +0200 Subject: of: Add vendor prefix for HannStar Display Corporation Signed-off-by: Philipp Zabel Acked-by: Rob Herring Signed-off-by: Thierry Reding diff --git a/Documentation/devicetree/bindings/vendor-prefixes.txt b/Documentation/devicetree/bindings/vendor-prefixes.txt index 723999d..9863039 100644 --- a/Documentation/devicetree/bindings/vendor-prefixes.txt +++ b/Documentation/devicetree/bindings/vendor-prefixes.txt @@ -64,6 +64,7 @@ gmt Global Mixed-mode Technology, Inc. google Google, Inc. gumstix Gumstix, Inc. gw Gateworks Corporation +hannstar HannStar Display Corporation haoyu Haoyu Microelectronic Co. Ltd. hisilicon Hisilicon Limited. honeywell Honeywell -- cgit v0.10.2 From a853205efb741951382edfdb1efa36886eeb90c3 Mon Sep 17 00:00:00 2001 From: Philipp Zabel Date: Thu, 23 Oct 2014 16:31:06 +0200 Subject: drm/panel: simple: Add HannStar HSD070PWW1 7.0" WXGA TFT LCD panel This patch adds support for the HannStar Display Corp. HSD070PWW1 7.0" WXGA TFT LCD panel to the simple-panel driver. The binding documentation is included. This panel is connected via LVDS and uses the data enable signal for timing. Since HSYNC/VSYNC are ignored, the split between sync length and porches is arbitrary, as long as the complete horizontal blanking interval is 160 clocks, and the vertical blanking interval is 23 lines. Signed-off-by: Philipp Zabel Signed-off-by: Thierry Reding diff --git a/Documentation/devicetree/bindings/panel/hannstar,hsd070pww1.txt b/Documentation/devicetree/bindings/panel/hannstar,hsd070pww1.txt new file mode 100644 index 0000000..7da1d5c --- /dev/null +++ b/Documentation/devicetree/bindings/panel/hannstar,hsd070pww1.txt @@ -0,0 +1,7 @@ +HannStar Display Corp. HSD070PWW1 7.0" WXGA TFT LCD panel + +Required properties: +- compatible: should be "hannstar,hsd070pww1" + +This binding is compatible with the simple-panel binding, which is specified +in simple-panel.txt in this directory. diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 83d9306..de78d3e 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -535,6 +535,29 @@ static const struct panel_desc foxlink_fl500wvr00_a0t = { }, }; +static const struct drm_display_mode hannstar_hsd070pww1_mode = { + .clock = 71100, + .hdisplay = 1280, + .hsync_start = 1280 + 1, + .hsync_end = 1280 + 1 + 158, + .htotal = 1280 + 1 + 158 + 1, + .vdisplay = 800, + .vsync_start = 800 + 1, + .vsync_end = 800 + 1 + 21, + .vtotal = 800 + 1 + 21 + 1, + .vrefresh = 60, +}; + +static const struct panel_desc hannstar_hsd070pww1 = { + .modes = &hannstar_hsd070pww1_mode, + .num_modes = 1, + .bpc = 6, + .size = { + .width = 151, + .height = 94, + }, +}; + static const struct drm_display_mode innolux_n116bge_mode = { .clock = 71000, .hdisplay = 1366, @@ -660,6 +683,9 @@ static const struct of_device_id platform_of_match[] = { .compatible = "foxlink,fl500wvr00-a0t", .data = &foxlink_fl500wvr00_a0t, }, { + .compatible = "hannstar,hsd070pww1", + .data = &hannstar_hsd070pww1, + }, { .compatible = "innolux,n116bge", .data = &innolux_n116bge, }, { -- cgit v0.10.2 From e35e305eff0fda4ddd5cfbd8eea06948faa503af Mon Sep 17 00:00:00 2001 From: Ajay Kumar Date: Mon, 1 Sep 2014 15:40:02 +0530 Subject: drm/panel: simple: Add AUO B116XW03 panel support The AUO B116XW03 is a 11.6" HD TFT LCD panel connecting to a LVDS interface and with an integrated LED backlight unit. This panel is used on the Samsung Chromebook(XE303C12). Signed-off-by: Ajay Kumar [treding@nvidia.com: add missing .bpc field] Signed-off-by: Thierry Reding diff --git a/Documentation/devicetree/bindings/panel/auo,b116xw03.txt b/Documentation/devicetree/bindings/panel/auo,b116xw03.txt new file mode 100644 index 0000000..690d0a5 --- /dev/null +++ b/Documentation/devicetree/bindings/panel/auo,b116xw03.txt @@ -0,0 +1,7 @@ +AU Optronics Corporation 11.6" HD (1366x768) color TFT-LCD panel + +Required properties: +- compatible: should be "auo,b116xw03" + +This binding is compatible with the simple-panel binding, which is specified +in simple-panel.txt in this directory. diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index de78d3e..f6d6ebc 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -369,6 +369,29 @@ static const struct panel_desc auo_b101xtn01 = { }, }; +static const struct drm_display_mode auo_b116xw03_mode = { + .clock = 70589, + .hdisplay = 1366, + .hsync_start = 1366 + 40, + .hsync_end = 1366 + 40 + 40, + .htotal = 1366 + 40 + 40 + 32, + .vdisplay = 768, + .vsync_start = 768 + 10, + .vsync_end = 768 + 10 + 12, + .vtotal = 768 + 10 + 12 + 6, + .vrefresh = 60, +}; + +static const struct panel_desc auo_b116xw03 = { + .modes = &auo_b116xw03_mode, + .num_modes = 1, + .bpc = 6, + .size = { + .width = 256, + .height = 144, + }, +}; + static const struct drm_display_mode auo_b133xtn01_mode = { .clock = 69500, .hdisplay = 1366, @@ -659,6 +682,9 @@ static const struct of_device_id platform_of_match[] = { .compatible = "auo,b101xtn01", .data = &auo_b101xtn01, }, { + .compatible = "auo,b116xw03", + .data = &auo_b116xw03, + }, { .compatible = "auo,b133htn01", .data = &auo_b133htn01, }, { -- cgit v0.10.2 From d7a839cde9acaa7252c75178bdbacce3cf12419c Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Thu, 6 Nov 2014 10:11:01 +0100 Subject: drm/panel: simple: Add missing .bpc fields Various panels were missing the .bpc field which encodes the number of bits per color. Not every display driver relies on this value, but since the panels can be used with any display engine it must be specified so that if a driver knows how to differentiate based on this field it can do so. Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index f6d6ebc..bc6a566 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -431,6 +431,7 @@ static const struct drm_display_mode auo_b133htn01_mode = { static const struct panel_desc auo_b133htn01 = { .modes = &auo_b133htn01_mode, .num_modes = 1, + .bpc = 6, .size = { .width = 293, .height = 165, @@ -552,6 +553,7 @@ static const struct drm_display_mode foxlink_fl500wvr00_a0t_mode = { static const struct panel_desc foxlink_fl500wvr00_a0t = { .modes = &foxlink_fl500wvr00_a0t_mode, .num_modes = 1, + .bpc = 8, .size = { .width = 108, .height = 65, @@ -786,6 +788,7 @@ static const struct panel_desc_dsi lg_ld070wx3_sl01 = { .desc = { .modes = &lg_ld070wx3_sl01_mode, .num_modes = 1, + .bpc = 8, .size = { .width = 94, .height = 151, @@ -813,6 +816,7 @@ static const struct panel_desc_dsi lg_lh500wx1_sd03 = { .desc = { .modes = &lg_lh500wx1_sd03_mode, .num_modes = 1, + .bpc = 8, .size = { .width = 62, .height = 110, @@ -840,6 +844,7 @@ static const struct panel_desc_dsi panasonic_vvx10f004b00 = { .desc = { .modes = &panasonic_vvx10f004b00_mode, .num_modes = 1, + .bpc = 8, .size = { .width = 217, .height = 136, -- cgit v0.10.2 From d731f661b5d2dd5f7748d2d8a4ed939b9d88128d Mon Sep 17 00:00:00 2001 From: Lucas Stach Date: Thu, 6 Nov 2014 17:44:33 +0100 Subject: drm/panel: simple: Add support for Innolux G121I1-L01 The Innolux G121I1-L01 is a 12.1" TFT LCD panel and can be supported by the simple-panel driver. This panel is connected via LVDS and uses the data enable signal for timing. Since HSYNC/VSYNC are ignored, the split between sync length and porches is arbitrary, as long as the complete horizontal blanking interval is 160 clocks, and the vertical blanking interval is 24 lines. Signed-off-by: Lucas Stach Signed-off-by: Thierry Reding diff --git a/Documentation/devicetree/bindings/panel/innolux,g121i1-l01.txt b/Documentation/devicetree/bindings/panel/innolux,g121i1-l01.txt new file mode 100644 index 0000000..2743b07 --- /dev/null +++ b/Documentation/devicetree/bindings/panel/innolux,g121i1-l01.txt @@ -0,0 +1,7 @@ +Innolux Corporation 12.1" WXGA (1280x800) TFT LCD panel + +Required properties: +- compatible: should be "innolux,g121i1-l01" + +This binding is compatible with the simple-panel binding, which is specified +in simple-panel.txt in this directory. diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index bc6a566..be535e7 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -583,6 +583,29 @@ static const struct panel_desc hannstar_hsd070pww1 = { }, }; +static const struct drm_display_mode innolux_g121i1_l01_mode = { + .clock = 71000, + .hdisplay = 1280, + .hsync_start = 1280 + 64, + .hsync_end = 1280 + 64 + 32, + .htotal = 1280 + 64 + 32 + 64, + .vdisplay = 800, + .vsync_start = 800 + 9, + .vsync_end = 800 + 9 + 6, + .vtotal = 800 + 9 + 6 + 9, + .vrefresh = 60, +}; + +static const struct panel_desc innolux_g121i1_l01 = { + .modes = &innolux_g121i1_l01_mode, + .num_modes = 1, + .bpc = 6, + .size = { + .width = 261, + .height = 163, + }, +}; + static const struct drm_display_mode innolux_n116bge_mode = { .clock = 71000, .hdisplay = 1366, @@ -714,6 +737,9 @@ static const struct of_device_id platform_of_match[] = { .compatible = "hannstar,hsd070pww1", .data = &hannstar_hsd070pww1, }, { + .compatible ="innolux,g121i1-l01", + .data = &innolux_g121i1_l01 + }, { .compatible = "innolux,n116bge", .data = &innolux_n116bge, }, { -- cgit v0.10.2 From 071964ae2b27e645fe8324557fe2754ec77b6417 Mon Sep 17 00:00:00 2001 From: Lucas Stach Date: Thu, 6 Nov 2014 17:44:34 +0100 Subject: of: Add vendor prefix for Hitachi Ltd. Corporation Signed-off-by: Lucas Stach Acked-by: Rob Herring Signed-off-by: Thierry Reding diff --git a/Documentation/devicetree/bindings/vendor-prefixes.txt b/Documentation/devicetree/bindings/vendor-prefixes.txt index 9863039..0b2fc73 100644 --- a/Documentation/devicetree/bindings/vendor-prefixes.txt +++ b/Documentation/devicetree/bindings/vendor-prefixes.txt @@ -67,6 +67,7 @@ gw Gateworks Corporation hannstar HannStar Display Corporation haoyu Haoyu Microelectronic Co. Ltd. hisilicon Hisilicon Limited. +hit Hitachi Ltd. honeywell Honeywell hp Hewlett Packard i2se I2SE GmbH -- cgit v0.10.2 From 61ac0bf89d74e92b086eb147e5a5aed5b37ad396 Mon Sep 17 00:00:00 2001 From: Lucas Stach Date: Thu, 6 Nov 2014 17:44:35 +0100 Subject: drm/panel: simple: Add support for Hitachi TX23D38VM0CAA The Hitachi TX23D38VM0CAA is a 9" WVGA TFT LCD panel and can be supported by the simple-panel driver. This panel is connected via LVDS and uses the data enable signal for timing. Since HSYNC/VSYNC are ignored, the split between sync length and porches is arbitrary, as long as the complete horizontal blanking interval is 256 clocks, and the vertical blanking interval is 45 lines. Signed-off-by: Lucas Stach Signed-off-by: Thierry Reding diff --git a/Documentation/devicetree/bindings/panel/hit,tx23d38vm0caa.txt b/Documentation/devicetree/bindings/panel/hit,tx23d38vm0caa.txt new file mode 100644 index 0000000..04caaae --- /dev/null +++ b/Documentation/devicetree/bindings/panel/hit,tx23d38vm0caa.txt @@ -0,0 +1,7 @@ +Hitachi Ltd. Corporation 9" WVGA (800x480) TFT LCD panel + +Required properties: +- compatible: should be "hit,tx23d38vm0caa" + +This binding is compatible with the simple-panel binding, which is specified +in simple-panel.txt in this directory. diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index be535e7..965114b 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -583,6 +583,29 @@ static const struct panel_desc hannstar_hsd070pww1 = { }, }; +static const struct drm_display_mode hitachi_tx23d38vm0caa_mode = { + .clock = 33333, + .hdisplay = 800, + .hsync_start = 800 + 85, + .hsync_end = 800 + 85 + 86, + .htotal = 800 + 85 + 86 + 85, + .vdisplay = 480, + .vsync_start = 480 + 16, + .vsync_end = 480 + 16 + 13, + .vtotal = 480 + 16 + 13 + 16, + .vrefresh = 60, +}; + +static const struct panel_desc hitachi_tx23d38vm0caa = { + .modes = &hitachi_tx23d38vm0caa_mode, + .num_modes = 1, + .bpc = 6, + .size = { + .width = 195, + .height = 117, + }, +}; + static const struct drm_display_mode innolux_g121i1_l01_mode = { .clock = 71000, .hdisplay = 1280, @@ -737,6 +760,9 @@ static const struct of_device_id platform_of_match[] = { .compatible = "hannstar,hsd070pww1", .data = &hannstar_hsd070pww1, }, { + .compatible = "hit,tx23d38vm0caa", + .data = &hitachi_tx23d38vm0caa + }, { .compatible ="innolux,g121i1-l01", .data = &innolux_g121i1_l01 }, { -- cgit v0.10.2 From 7fe8c7778f7382400544a5a4252a07b5c093d187 Mon Sep 17 00:00:00 2001 From: Daniel Kurtz Date: Tue, 2 Sep 2014 10:56:46 +0800 Subject: drm/panel: simple: Update Innolux N116BGE timings There are several different models of N116BGE. According to commit 0a2288c06aab ("drm/panel: simple: Add Innolux N116BGE panel support"), the video timings are for the eDP variant. The clock and htotal values added by that patch are out of spec according to the datasheets I have seen for the eDP N116BGE (-EA2 and -EB2). This patch changes the values to the "Typ" values on the datasheet. Signed-off-by: Daniel Kurtz [tested that these timings work with the Tegra132 Norrin panel] Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 965114b..ce4eb25 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -630,15 +630,15 @@ static const struct panel_desc innolux_g121i1_l01 = { }; static const struct drm_display_mode innolux_n116bge_mode = { - .clock = 71000, + .clock = 76420, .hdisplay = 1366, - .hsync_start = 1366 + 64, - .hsync_end = 1366 + 64 + 6, - .htotal = 1366 + 64 + 6 + 64, + .hsync_start = 1366 + 136, + .hsync_end = 1366 + 136 + 30, + .htotal = 1366 + 136 + 30 + 60, .vdisplay = 768, .vsync_start = 768 + 8, - .vsync_end = 768 + 8 + 4, - .vtotal = 768 + 8 + 4 + 8, + .vsync_end = 768 + 8 + 12, + .vtotal = 768 + 8 + 12 + 12, .vrefresh = 60, .flags = DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC, }; -- cgit v0.10.2 From eccda2d1e1848f215a6c7be4833ca1139e5ffd92 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Fri, 7 Nov 2014 14:41:05 +0100 Subject: drm/panel: ld9040: Fix build warnings on 64-bit The %* format specifier expects an integer, which works fine with size_t arguments on 32-bit because the types match. However on 64-bit, size_t is typedef'd to unsigned long and will cause a build warning. Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/panel/panel-ld9040.c b/drivers/gpu/drm/panel/panel-ld9040.c index dd1d3f2..08cf2c5 100644 --- a/drivers/gpu/drm/panel/panel-ld9040.c +++ b/drivers/gpu/drm/panel/panel-ld9040.c @@ -145,7 +145,7 @@ static void ld9040_dcs_write(struct ld9040 *ctx, const u8 *data, size_t len) if (ctx->error < 0 || len == 0) return; - dev_dbg(ctx->dev, "writing dcs seq: %*ph\n", len, data); + dev_dbg(ctx->dev, "writing dcs seq: %*ph\n", (int)len, data); ret = ld9040_spi_write_word(ctx, *data); while (!ret && --len) { @@ -154,8 +154,8 @@ static void ld9040_dcs_write(struct ld9040 *ctx, const u8 *data, size_t len) } if (ret) { - dev_err(ctx->dev, "error %d writing dcs seq: %*ph\n", ret, len, - data); + dev_err(ctx->dev, "error %d writing dcs seq: %*ph\n", ret, + (int)len, data); ctx->error = ret; } -- cgit v0.10.2 From d85a1609e62c22730aa4e3ef09884f04606b086e Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Fri, 7 Nov 2014 14:42:16 +0100 Subject: drm/panel: s6e8aa0: Fix build warnings on 64-bit The %* format specifier expects an integer, which works fine with size_t arguments on 32-bit because the types match. However on 64-bit, size_t is typedef'd to unsigned long and will cause a build warning. Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/panel/panel-s6e8aa0.c b/drivers/gpu/drm/panel/panel-s6e8aa0.c index 6427aa1..96f4476 100644 --- a/drivers/gpu/drm/panel/panel-s6e8aa0.c +++ b/drivers/gpu/drm/panel/panel-s6e8aa0.c @@ -143,8 +143,8 @@ static void s6e8aa0_dcs_write(struct s6e8aa0 *ctx, const void *data, size_t len) ret = mipi_dsi_dcs_write(dsi, data, len); if (ret < 0) { - dev_err(ctx->dev, "error %zd writing dcs seq: %*ph\n", ret, len, - data); + dev_err(ctx->dev, "error %zd writing dcs seq: %*ph\n", ret, + (int)len, data); ctx->error = ret; } } -- cgit v0.10.2 From 0a88818d093b4d5746d7b3af3acc2b2bddd59808 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 3 Nov 2014 14:37:38 +0100 Subject: drm/i915: Check pipe_config.has_dp_encoder instead of encoder types More concise. Noticed while reviewing Ander's patch which touched a lot of the pipe_has_type checks. v2: Use new_config in one place Ander spotted. Cc: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 6bd290d..96663eb 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -3786,9 +3786,7 @@ static void ironlake_pch_enable(struct drm_crtc *crtc) intel_fdi_normal_train(crtc); /* For PCH DP, enable TRANS_DP_CTL */ - if (HAS_PCH_CPT(dev) && - (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT) || - intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_EDP))) { + if (HAS_PCH_CPT(dev) && intel_crtc->config.has_dp_encoder) { u32 bpc = (I915_READ(PIPECONF(pipe)) & PIPECONF_BPC_MASK) >> 5; reg = TRANS_DP_CTL(pipe); temp = I915_READ(reg); @@ -5867,8 +5865,7 @@ static void vlv_prepare_pll(struct intel_crtc *crtc, vlv_dpio_write(dev_priv, pipe, VLV_PLL_DW10(pipe), 0x00d0000f); - if (intel_pipe_has_type(crtc, INTEL_OUTPUT_EDP) || - intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) { + if (crtc->config.has_dp_encoder) { /* Use SSC source */ if (pipe == PIPE_A) vlv_dpio_write(dev_priv, pipe, VLV_PLL_DW5(pipe), @@ -6058,7 +6055,7 @@ static void i9xx_update_pll(struct intel_crtc *crtc, if (is_sdvo) dpll |= DPLL_SDVO_HIGH_SPEED; - if (intel_pipe_will_have_type(crtc, INTEL_OUTPUT_DISPLAYPORT)) + if (crtc->new_config->has_dp_encoder) dpll |= DPLL_SDVO_HIGH_SPEED; /* compute bitmask from p1 value */ -- cgit v0.10.2 From 1e6f2ddc88fec990ef4fb1aa3b7cb9941df0321b Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:31 +0200 Subject: drm/i915: Convert shared dpll reference count to a crtc mask This will be used in a follow up patch to properly release shared DPLLs without relying on the shared_dpll field in pipe_config. v2: Fix white space error (Ville) Use hweight32() (Ville) Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index a79f83c..ce32ae7 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2630,8 +2630,8 @@ static int i915_shared_dplls_info(struct seq_file *m, void *unused) struct intel_shared_dpll *pll = &dev_priv->shared_dplls[i]; seq_printf(m, "DPLL%i: %s, id: %i\n", i, pll->name, pll->id); - seq_printf(m, " refcount: %i, active: %i, on: %s\n", pll->refcount, - pll->active, yesno(pll->on)); + seq_printf(m, " crtc_mask: 0x%08x, active: %d, on: %s\n", + pll->crtc_mask, pll->active, yesno(pll->on)); seq_printf(m, " tracked hardware state:\n"); seq_printf(m, " dpll: 0x%08x\n", pll->hw_state.dpll); seq_printf(m, " dpll_md: 0x%08x\n", pll->hw_state.dpll_md); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 0e12c6a..4f4eef7 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -230,7 +230,7 @@ struct intel_dpll_hw_state { }; struct intel_shared_dpll { - int refcount; /* count of number of CRTCs sharing this PLL */ + unsigned crtc_mask; /* mask of CRTCs sharing this PLL */ int active; /* count of number of active CRTCs (i.e. DPMS on) */ bool on; /* is the PLL actually active? Disabled during modeset */ const char *name; diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 96663eb..5219b29 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -1779,7 +1779,7 @@ static void intel_prepare_shared_dpll(struct intel_crtc *crtc) if (WARN_ON(pll == NULL)) return; - WARN_ON(!pll->refcount); + WARN_ON(!pll->crtc_mask); if (pll->active == 0) { DRM_DEBUG_DRIVER("setting up %s\n", pll->name); WARN_ON(pll->on); @@ -1806,7 +1806,7 @@ static void intel_enable_shared_dpll(struct intel_crtc *crtc) if (WARN_ON(pll == NULL)) return; - if (WARN_ON(pll->refcount == 0)) + if (WARN_ON(pll->crtc_mask == 0)) return; DRM_DEBUG_KMS("enable %s (active %d, on? %d) for crtc %d\n", @@ -1838,7 +1838,7 @@ static void intel_disable_shared_dpll(struct intel_crtc *crtc) if (WARN_ON(pll == NULL)) return; - if (WARN_ON(pll->refcount == 0)) + if (WARN_ON(pll->crtc_mask == 0)) return; DRM_DEBUG_KMS("disable %s (active %d, on? %d) for crtc %d\n", @@ -3846,12 +3846,13 @@ void intel_put_shared_dpll(struct intel_crtc *crtc) if (pll == NULL) return; - if (pll->refcount == 0) { - WARN(1, "bad %s refcount\n", pll->name); + if (!(pll->crtc_mask & (1 << crtc->pipe))) { + WARN(1, "bad %s crtc mask\n", pll->name); return; } - if (--pll->refcount == 0) { + pll->crtc_mask &= ~(1 << crtc->pipe); + if (pll->crtc_mask == 0) { WARN_ON(pll->on); WARN_ON(pll->active); } @@ -3879,7 +3880,7 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) DRM_DEBUG_KMS("CRTC:%d using pre-allocated %s\n", crtc->base.base.id, pll->name); - WARN_ON(pll->refcount); + WARN_ON(pll->crtc_mask); goto found; } @@ -3888,14 +3889,15 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) pll = &dev_priv->shared_dplls[i]; /* Only want to check enabled timings first */ - if (pll->refcount == 0) + if (pll->crtc_mask == 0) continue; if (memcmp(&crtc->config.dpll_hw_state, &pll->hw_state, sizeof(pll->hw_state)) == 0) { - DRM_DEBUG_KMS("CRTC:%d sharing existing %s (refcount %d, ative %d)\n", - crtc->base.base.id, - pll->name, pll->refcount, pll->active); + DRM_DEBUG_KMS("CRTC:%d sharing existing %s " + "(crtc_mask 0x%08x, active %d)\n", + crtc->base.base.id, pll->name, + pll->crtc_mask, pll->active); goto found; } @@ -3904,7 +3906,7 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) /* Ok no matching timings, maybe there's a free one? */ for (i = 0; i < dev_priv->num_shared_dpll; i++) { pll = &dev_priv->shared_dplls[i]; - if (pll->refcount == 0) { + if (pll->crtc_mask == 0) { DRM_DEBUG_KMS("CRTC:%d allocated %s\n", crtc->base.base.id, pll->name); goto found; @@ -3914,14 +3916,14 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) return NULL; found: - if (pll->refcount == 0) + if (pll->crtc_mask == 0) pll->hw_state = crtc->config.dpll_hw_state; crtc->config.shared_dpll = i; DRM_DEBUG_DRIVER("using %s for pipe %c\n", pll->name, pipe_name(crtc->pipe)); - pll->refcount++; + pll->crtc_mask |= 1 << crtc->pipe; return pll; } @@ -10604,9 +10606,9 @@ check_shared_dpll_state(struct drm_device *dev) active = pll->get_hw_state(dev_priv, pll, &dpll_hw_state); - WARN(pll->active > pll->refcount, + WARN(pll->active > hweight32(pll->crtc_mask), "more active pll users than references: %i vs %i\n", - pll->active, pll->refcount); + pll->active, hweight32(pll->crtc_mask)); WARN(pll->active && !pll->on, "pll in active use but not on in sw tracking\n"); WARN(pll->on && !pll->active, @@ -10624,9 +10626,9 @@ check_shared_dpll_state(struct drm_device *dev) WARN(pll->active != active_crtcs, "pll active crtcs mismatch (expected %i, found %i)\n", pll->active, active_crtcs); - WARN(pll->refcount != enabled_crtcs, + WARN(hweight32(pll->crtc_mask) != enabled_crtcs, "pll enabled crtcs mismatch (expected %i, found %i)\n", - pll->refcount, enabled_crtcs); + hweight32(pll->crtc_mask), enabled_crtcs); WARN(pll->on && memcmp(&pll->hw_state, &dpll_hw_state, sizeof(dpll_hw_state)), @@ -13077,16 +13079,18 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) pll->on = pll->get_hw_state(dev_priv, pll, &pll->hw_state); pll->active = 0; + pll->crtc_mask = 0; for_each_intel_crtc(dev, crtc) { - if (crtc->active && intel_crtc_to_shared_dpll(crtc) == pll) + if (crtc->active && intel_crtc_to_shared_dpll(crtc) == pll) { pll->active++; + pll->crtc_mask |= 1 << crtc->pipe; + } } - pll->refcount = pll->active; - DRM_DEBUG_KMS("%s hw state readout: refcount %i, on %i\n", - pll->name, pll->refcount, pll->on); + DRM_DEBUG_KMS("%s hw state readout: crtc_mask 0x%08x, on %i\n", + pll->name, pll->crtc_mask, pll->on); - if (pll->refcount) + if (pll->crtc_mask) intel_display_power_get(dev_priv, POWER_DOMAIN_PLLS); } -- cgit v0.10.2 From 3e369b76ceb625dffabef6f1c52e32fc0998843e Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:32 +0200 Subject: drm/i915: Move dpll crtc_mask and hw_state fields into separate struct The new struct will be used in a follow up patch to allow a current and a staged config to exist for the same shared DPLL. v2: Rebase on by mask_to_refcount()->hweight32() change. (Damien) Signed-off-by: Ander Conselvan de Oliveira Reviewed-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index ce32ae7..0a69813 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2631,13 +2631,14 @@ static int i915_shared_dplls_info(struct seq_file *m, void *unused) seq_printf(m, "DPLL%i: %s, id: %i\n", i, pll->name, pll->id); seq_printf(m, " crtc_mask: 0x%08x, active: %d, on: %s\n", - pll->crtc_mask, pll->active, yesno(pll->on)); + pll->config.crtc_mask, pll->active, yesno(pll->on)); seq_printf(m, " tracked hardware state:\n"); - seq_printf(m, " dpll: 0x%08x\n", pll->hw_state.dpll); - seq_printf(m, " dpll_md: 0x%08x\n", pll->hw_state.dpll_md); - seq_printf(m, " fp0: 0x%08x\n", pll->hw_state.fp0); - seq_printf(m, " fp1: 0x%08x\n", pll->hw_state.fp1); - seq_printf(m, " wrpll: 0x%08x\n", pll->hw_state.wrpll); + seq_printf(m, " dpll: 0x%08x\n", pll->config.hw_state.dpll); + seq_printf(m, " dpll_md: 0x%08x\n", + pll->config.hw_state.dpll_md); + seq_printf(m, " fp0: 0x%08x\n", pll->config.hw_state.fp0); + seq_printf(m, " fp1: 0x%08x\n", pll->config.hw_state.fp1); + seq_printf(m, " wrpll: 0x%08x\n", pll->config.hw_state.wrpll); } drm_modeset_unlock_all(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 4f4eef7..62a8e05 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -229,14 +229,18 @@ struct intel_dpll_hw_state { uint32_t wrpll; }; -struct intel_shared_dpll { +struct intel_shared_dpll_config { unsigned crtc_mask; /* mask of CRTCs sharing this PLL */ + struct intel_dpll_hw_state hw_state; +}; + +struct intel_shared_dpll { + struct intel_shared_dpll_config config; int active; /* count of number of active CRTCs (i.e. DPMS on) */ bool on; /* is the PLL actually active? Disabled during modeset */ const char *name; /* should match the index in the dev_priv->shared_dplls array */ enum intel_dpll_id id; - struct intel_dpll_hw_state hw_state; /* The mode_set hook is optional and should be used together with the * intel_prepare_shared_dpll function. */ void (*mode_set)(struct drm_i915_private *dev_priv, diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index d417688..6f9712b 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1312,7 +1312,7 @@ int intel_ddi_get_cdclk_freq(struct drm_i915_private *dev_priv) static void hsw_ddi_pll_enable(struct drm_i915_private *dev_priv, struct intel_shared_dpll *pll) { - I915_WRITE(WRPLL_CTL(pll->id), pll->hw_state.wrpll); + I915_WRITE(WRPLL_CTL(pll->id), pll->config.hw_state.wrpll); POSTING_READ(WRPLL_CTL(pll->id)); udelay(20); } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5219b29..5eb4754 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -1779,7 +1779,7 @@ static void intel_prepare_shared_dpll(struct intel_crtc *crtc) if (WARN_ON(pll == NULL)) return; - WARN_ON(!pll->crtc_mask); + WARN_ON(!pll->config.crtc_mask); if (pll->active == 0) { DRM_DEBUG_DRIVER("setting up %s\n", pll->name); WARN_ON(pll->on); @@ -1806,7 +1806,7 @@ static void intel_enable_shared_dpll(struct intel_crtc *crtc) if (WARN_ON(pll == NULL)) return; - if (WARN_ON(pll->crtc_mask == 0)) + if (WARN_ON(pll->config.crtc_mask == 0)) return; DRM_DEBUG_KMS("enable %s (active %d, on? %d) for crtc %d\n", @@ -1838,7 +1838,7 @@ static void intel_disable_shared_dpll(struct intel_crtc *crtc) if (WARN_ON(pll == NULL)) return; - if (WARN_ON(pll->crtc_mask == 0)) + if (WARN_ON(pll->config.crtc_mask == 0)) return; DRM_DEBUG_KMS("disable %s (active %d, on? %d) for crtc %d\n", @@ -3846,13 +3846,13 @@ void intel_put_shared_dpll(struct intel_crtc *crtc) if (pll == NULL) return; - if (!(pll->crtc_mask & (1 << crtc->pipe))) { + if (!(pll->config.crtc_mask & (1 << crtc->pipe))) { WARN(1, "bad %s crtc mask\n", pll->name); return; } - pll->crtc_mask &= ~(1 << crtc->pipe); - if (pll->crtc_mask == 0) { + pll->config.crtc_mask &= ~(1 << crtc->pipe); + if (pll->config.crtc_mask == 0) { WARN_ON(pll->on); WARN_ON(pll->active); } @@ -3880,7 +3880,7 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) DRM_DEBUG_KMS("CRTC:%d using pre-allocated %s\n", crtc->base.base.id, pll->name); - WARN_ON(pll->crtc_mask); + WARN_ON(pll->config.crtc_mask); goto found; } @@ -3889,15 +3889,16 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) pll = &dev_priv->shared_dplls[i]; /* Only want to check enabled timings first */ - if (pll->crtc_mask == 0) + if (pll->config.crtc_mask == 0) continue; - if (memcmp(&crtc->config.dpll_hw_state, &pll->hw_state, - sizeof(pll->hw_state)) == 0) { + if (memcmp(&crtc->config.dpll_hw_state, + &pll->config.hw_state, + sizeof(pll->config.hw_state)) == 0) { DRM_DEBUG_KMS("CRTC:%d sharing existing %s " "(crtc_mask 0x%08x, active %d)\n", crtc->base.base.id, pll->name, - pll->crtc_mask, pll->active); + pll->config.crtc_mask, pll->active); goto found; } @@ -3906,7 +3907,7 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) /* Ok no matching timings, maybe there's a free one? */ for (i = 0; i < dev_priv->num_shared_dpll; i++) { pll = &dev_priv->shared_dplls[i]; - if (pll->crtc_mask == 0) { + if (pll->config.crtc_mask == 0) { DRM_DEBUG_KMS("CRTC:%d allocated %s\n", crtc->base.base.id, pll->name); goto found; @@ -3916,14 +3917,14 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) return NULL; found: - if (pll->crtc_mask == 0) - pll->hw_state = crtc->config.dpll_hw_state; + if (pll->config.crtc_mask == 0) + pll->config.hw_state = crtc->config.dpll_hw_state; crtc->config.shared_dpll = i; DRM_DEBUG_DRIVER("using %s for pipe %c\n", pll->name, pipe_name(crtc->pipe)); - pll->crtc_mask |= 1 << crtc->pipe; + pll->config.crtc_mask |= 1 << crtc->pipe; return pll; } @@ -10606,9 +10607,9 @@ check_shared_dpll_state(struct drm_device *dev) active = pll->get_hw_state(dev_priv, pll, &dpll_hw_state); - WARN(pll->active > hweight32(pll->crtc_mask), + WARN(pll->active > hweight32(pll->config.crtc_mask), "more active pll users than references: %i vs %i\n", - pll->active, hweight32(pll->crtc_mask)); + pll->active, hweight32(pll->config.crtc_mask)); WARN(pll->active && !pll->on, "pll in active use but not on in sw tracking\n"); WARN(pll->on && !pll->active, @@ -10626,11 +10627,11 @@ check_shared_dpll_state(struct drm_device *dev) WARN(pll->active != active_crtcs, "pll active crtcs mismatch (expected %i, found %i)\n", pll->active, active_crtcs); - WARN(hweight32(pll->crtc_mask) != enabled_crtcs, + WARN(hweight32(pll->config.crtc_mask) != enabled_crtcs, "pll enabled crtcs mismatch (expected %i, found %i)\n", - hweight32(pll->crtc_mask), enabled_crtcs); + hweight32(pll->config.crtc_mask), enabled_crtcs); - WARN(pll->on && memcmp(&pll->hw_state, &dpll_hw_state, + WARN(pll->on && memcmp(&pll->config.hw_state, &dpll_hw_state, sizeof(dpll_hw_state)), "pll hw state mismatch\n"); } @@ -11305,8 +11306,8 @@ static bool ibx_pch_dpll_get_hw_state(struct drm_i915_private *dev_priv, static void ibx_pch_dpll_mode_set(struct drm_i915_private *dev_priv, struct intel_shared_dpll *pll) { - I915_WRITE(PCH_FP0(pll->id), pll->hw_state.fp0); - I915_WRITE(PCH_FP1(pll->id), pll->hw_state.fp1); + I915_WRITE(PCH_FP0(pll->id), pll->config.hw_state.fp0); + I915_WRITE(PCH_FP1(pll->id), pll->config.hw_state.fp1); } static void ibx_pch_dpll_enable(struct drm_i915_private *dev_priv, @@ -11315,7 +11316,7 @@ static void ibx_pch_dpll_enable(struct drm_i915_private *dev_priv, /* PCH refclock must be enabled first */ ibx_assert_pch_refclk_enabled(dev_priv); - I915_WRITE(PCH_DPLL(pll->id), pll->hw_state.dpll); + I915_WRITE(PCH_DPLL(pll->id), pll->config.hw_state.dpll); /* Wait for the clocks to stabilize. */ POSTING_READ(PCH_DPLL(pll->id)); @@ -11326,7 +11327,7 @@ static void ibx_pch_dpll_enable(struct drm_i915_private *dev_priv, * * So write it again. */ - I915_WRITE(PCH_DPLL(pll->id), pll->hw_state.dpll); + I915_WRITE(PCH_DPLL(pll->id), pll->config.hw_state.dpll); POSTING_READ(PCH_DPLL(pll->id)); udelay(200); } @@ -13077,20 +13078,21 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) for (i = 0; i < dev_priv->num_shared_dpll; i++) { struct intel_shared_dpll *pll = &dev_priv->shared_dplls[i]; - pll->on = pll->get_hw_state(dev_priv, pll, &pll->hw_state); + pll->on = pll->get_hw_state(dev_priv, pll, + &pll->config.hw_state); pll->active = 0; - pll->crtc_mask = 0; + pll->config.crtc_mask = 0; for_each_intel_crtc(dev, crtc) { if (crtc->active && intel_crtc_to_shared_dpll(crtc) == pll) { pll->active++; - pll->crtc_mask |= 1 << crtc->pipe; + pll->config.crtc_mask |= 1 << crtc->pipe; } } DRM_DEBUG_KMS("%s hw state readout: crtc_mask 0x%08x, on %i\n", - pll->name, pll->crtc_mask, pll->on); + pll->name, pll->config.crtc_mask, pll->on); - if (pll->crtc_mask) + if (pll->config.crtc_mask) intel_display_power_get(dev_priv, POWER_DOMAIN_PLLS); } -- cgit v0.10.2 From 8bd31e67c9f737216cb25bcc008fed172b8a4375 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:33 +0200 Subject: drm/i915: Add infrastructure for choosing DPLLs before disabling crtcs It is possible for a mode set to fail if there aren't shared DPLLS that match the new configuration requirement or other errors in clock computation. If that step is executed after disabling crtcs, in the failure case the hardware configuration is changed and needs to be restored. Doing those things early will allow the mode set to fail before actually touching the hardware. Follow up patches will convert different platforms to use the new infrastructure. v2: Keep pll->new_config valid only during mode set (Ville) Use kmemdup() in i915_shared_dpll_start_config() (Ville) Restore old pll config if something fails before commit (Ville) Don't set compute_clock hooks since dev_priv is kzalloc()'d (Ville) Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 62a8e05..4fb93e2 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -236,6 +236,8 @@ struct intel_shared_dpll_config { struct intel_shared_dpll { struct intel_shared_dpll_config config; + struct intel_shared_dpll_config *new_config; + int active; /* count of number of active CRTCs (i.e. DPMS on) */ bool on; /* is the PLL actually active? Disabled during modeset */ const char *name; @@ -484,6 +486,7 @@ struct drm_i915_display_funcs { struct intel_crtc_config *); void (*get_plane_config)(struct intel_crtc *, struct intel_plane_config *); + int (*crtc_compute_clock)(struct intel_crtc *crtc); int (*crtc_mode_set)(struct intel_crtc *crtc, int x, int y, struct drm_framebuffer *old_fb); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5eb4754..4253f12 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -3863,15 +3863,9 @@ void intel_put_shared_dpll(struct intel_crtc *crtc) struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) { struct drm_i915_private *dev_priv = crtc->base.dev->dev_private; - struct intel_shared_dpll *pll = intel_crtc_to_shared_dpll(crtc); + struct intel_shared_dpll *pll; enum intel_dpll_id i; - if (pll) { - DRM_DEBUG_KMS("CRTC:%d dropping existing %s\n", - crtc->base.base.id, pll->name); - intel_put_shared_dpll(crtc); - } - if (HAS_PCH_IBX(dev_priv->dev)) { /* Ironlake PCH has a fixed PLL->PCH pipe mapping. */ i = (enum intel_dpll_id) crtc->pipe; @@ -3880,7 +3874,7 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) DRM_DEBUG_KMS("CRTC:%d using pre-allocated %s\n", crtc->base.base.id, pll->name); - WARN_ON(pll->config.crtc_mask); + WARN_ON(pll->new_config->crtc_mask); goto found; } @@ -3889,17 +3883,16 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) pll = &dev_priv->shared_dplls[i]; /* Only want to check enabled timings first */ - if (pll->config.crtc_mask == 0) + if (pll->new_config->crtc_mask == 0) continue; - if (memcmp(&crtc->config.dpll_hw_state, - &pll->config.hw_state, - sizeof(pll->config.hw_state)) == 0) { - DRM_DEBUG_KMS("CRTC:%d sharing existing %s " - "(crtc_mask 0x%08x, active %d)\n", + if (memcmp(&crtc->new_config->dpll_hw_state, + &pll->new_config->hw_state, + sizeof(pll->new_config->hw_state)) == 0) { + DRM_DEBUG_KMS("CRTC:%d sharing existing %s (crtc mask 0x%08x, ative %d)\n", crtc->base.base.id, pll->name, - pll->config.crtc_mask, pll->active); - + pll->new_config->crtc_mask, + pll->active); goto found; } } @@ -3907,7 +3900,7 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) /* Ok no matching timings, maybe there's a free one? */ for (i = 0; i < dev_priv->num_shared_dpll; i++) { pll = &dev_priv->shared_dplls[i]; - if (pll->config.crtc_mask == 0) { + if (pll->new_config->crtc_mask == 0) { DRM_DEBUG_KMS("CRTC:%d allocated %s\n", crtc->base.base.id, pll->name); goto found; @@ -3917,18 +3910,85 @@ struct intel_shared_dpll *intel_get_shared_dpll(struct intel_crtc *crtc) return NULL; found: - if (pll->config.crtc_mask == 0) - pll->config.hw_state = crtc->config.dpll_hw_state; + if (pll->new_config->crtc_mask == 0) + pll->new_config->hw_state = crtc->new_config->dpll_hw_state; - crtc->config.shared_dpll = i; + crtc->new_config->shared_dpll = i; DRM_DEBUG_DRIVER("using %s for pipe %c\n", pll->name, pipe_name(crtc->pipe)); - pll->config.crtc_mask |= 1 << crtc->pipe; + pll->new_config->crtc_mask |= 1 << crtc->pipe; return pll; } +/** + * intel_shared_dpll_start_config - start a new PLL staged config + * @dev_priv: DRM device + * @clear_pipes: mask of pipes that will have their PLLs freed + * + * Starts a new PLL staged config, copying the current config but + * releasing the references of pipes specified in clear_pipes. + */ +static int intel_shared_dpll_start_config(struct drm_i915_private *dev_priv, + unsigned clear_pipes) +{ + struct intel_shared_dpll *pll; + enum intel_dpll_id i; + + for (i = 0; i < dev_priv->num_shared_dpll; i++) { + pll = &dev_priv->shared_dplls[i]; + + pll->new_config = kmemdup(&pll->config, sizeof pll->config, + GFP_KERNEL); + if (!pll->new_config) + goto cleanup; + + pll->new_config->crtc_mask &= ~clear_pipes; + } + + return 0; + +cleanup: + while (--i >= 0) { + pll = &dev_priv->shared_dplls[i]; + pll->new_config = NULL; + } + + return -ENOMEM; +} + +static void intel_shared_dpll_commit(struct drm_i915_private *dev_priv) +{ + struct intel_shared_dpll *pll; + enum intel_dpll_id i; + + for (i = 0; i < dev_priv->num_shared_dpll; i++) { + pll = &dev_priv->shared_dplls[i]; + + WARN_ON(pll->new_config == &pll->config); + + pll->config = *pll->new_config; + kfree(pll->new_config); + pll->new_config = NULL; + } +} + +static void intel_shared_dpll_abort_config(struct drm_i915_private *dev_priv) +{ + struct intel_shared_dpll *pll; + enum intel_dpll_id i; + + for (i = 0; i < dev_priv->num_shared_dpll; i++) { + pll = &dev_priv->shared_dplls[i]; + + WARN_ON(pll->new_config == &pll->config); + + kfree(pll->new_config); + pll->new_config = NULL; + } +} + static void cpt_verify_modeset(struct drm_device *dev, int pipe) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -5420,11 +5480,11 @@ static int intel_crtc_compute_config(struct intel_crtc *crtc, struct intel_crtc_config *pipe_config) { struct drm_device *dev = crtc->base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; struct drm_display_mode *adjusted_mode = &pipe_config->adjusted_mode; /* FIXME should check pixel clock limits on all platforms */ if (INTEL_INFO(dev)->gen < 4) { - struct drm_i915_private *dev_priv = dev->dev_private; int clock_limit = dev_priv->display.get_display_clock_speed(dev); @@ -5474,10 +5534,11 @@ static int intel_crtc_compute_config(struct intel_crtc *crtc, hsw_compute_ips_config(crtc, pipe_config); /* - * XXX: PCH/WRPLL clock sharing is done in ->mode_set, so make sure the - * old clock survives for now. + * XXX: PCH/WRPLL clock sharing is done in ->mode_set if ->compute_clock is not + * set, so make sure the old clock survives for now. */ - if (HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev) || HAS_DDI(dev)) + if (dev_priv->display.crtc_compute_clock == NULL && + (HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev) || HAS_DDI(dev))) pipe_config->shared_dpll = crtc->config.shared_dpll; if (pipe_config->has_pch_encoder) @@ -7405,6 +7466,9 @@ static int ironlake_crtc_mode_set(struct intel_crtc *crtc, else crtc->new_config->dpll_hw_state.fp1 = fp; + if (intel_crtc_to_shared_dpll(crtc)) + intel_put_shared_dpll(crtc); + pll = intel_get_shared_dpll(crtc); if (pll == NULL) { DRM_DEBUG_DRIVER("failed to find PLL for pipe %c\n", @@ -10749,6 +10813,22 @@ static int __intel_set_mode(struct drm_crtc *crtc, prepare_pipes &= ~disable_pipes; } + if (dev_priv->display.crtc_compute_clock) { + unsigned clear_pipes = modeset_pipes | disable_pipes; + + ret = intel_shared_dpll_start_config(dev_priv, clear_pipes); + if (ret) + goto done; + + for_each_intel_crtc_masked(dev, modeset_pipes, intel_crtc) { + ret = dev_priv->display.crtc_compute_clock(intel_crtc); + if (ret) { + intel_shared_dpll_abort_config(dev_priv); + goto done; + } + } + } + for_each_intel_crtc_masked(dev, disable_pipes, intel_crtc) intel_crtc_disable(&intel_crtc->base); @@ -10776,6 +10856,9 @@ static int __intel_set_mode(struct drm_crtc *crtc, &pipe_config->adjusted_mode); } + if (dev_priv->display.crtc_compute_clock) + intel_shared_dpll_commit(dev_priv); + /* Only after disabling all output pipelines that will be changed can we * update the the output configuration. */ intel_modeset_update_state(dev, prepare_pipes); @@ -10810,9 +10893,12 @@ static int __intel_set_mode(struct drm_crtc *crtc, crtc->x = x; crtc->y = y; - ret = dev_priv->display.crtc_mode_set(intel_crtc, x, y, fb); - if (ret) - goto done; + if (dev_priv->display.crtc_mode_set) { + ret = dev_priv->display.crtc_mode_set(intel_crtc, + x, y, fb); + if (ret) + goto done; + } } /* Now enable the clocks, plane, pipe, and connectors that we set up. */ -- cgit v0.10.2 From 797d02592304b15359f407bd0e52fa5143cb4c9f Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:34 +0200 Subject: drm/i915: Covert HSW+ to choose DPLLS before disabling CRTCs Use the infrastructure added in a previous patch to choose shared DPLLs and calculate clocks before touching the hardware. v2: Don't set mode_set hooks since dev_priv is kzalloc()'d (Ville) Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 6f9712b..a50df35 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -842,8 +842,6 @@ bool intel_ddi_pll_select(struct intel_crtc *intel_crtc) intel_ddi_get_crtc_new_encoder(intel_crtc); int clock = intel_crtc->new_config->port_clock; - intel_put_shared_dpll(intel_crtc); - return hsw_ddi_pll_select(intel_crtc, intel_encoder, clock); } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 4253f12..e4bff0b 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7975,9 +7975,7 @@ static void haswell_modeset_global_resources(struct drm_device *dev) modeset_update_crtc_power_domains(dev); } -static int haswell_crtc_mode_set(struct intel_crtc *crtc, - int x, int y, - struct drm_framebuffer *fb) +static int haswell_crtc_compute_clock(struct intel_crtc *crtc) { if (!intel_ddi_pll_select(crtc)) return -EINVAL; @@ -12468,7 +12466,8 @@ static void intel_init_display(struct drm_device *dev) if (HAS_DDI(dev)) { dev_priv->display.get_pipe_config = haswell_get_pipe_config; dev_priv->display.get_plane_config = ironlake_get_plane_config; - dev_priv->display.crtc_mode_set = haswell_crtc_mode_set; + dev_priv->display.crtc_compute_clock = + haswell_crtc_compute_clock; dev_priv->display.crtc_enable = haswell_crtc_enable; dev_priv->display.crtc_disable = haswell_crtc_disable; dev_priv->display.off = ironlake_crtc_off; -- cgit v0.10.2 From 3fb37703683543725dec3d8bc5c5b3a930325d04 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:35 +0200 Subject: drm/i915: Covert ILK-IVB to choose DPLLS before disabling CRTCs Use the infrastructure added in a previous patch to choose shared DPLLs and calculate clocks before touching the hardware. v2: Don't set mode_set hooks since dev_priv is kzalloc()'d (Ville) Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index e4bff0b..ee5eba7 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7418,9 +7418,7 @@ static uint32_t ironlake_compute_dpll(struct intel_crtc *intel_crtc, return dpll | DPLL_VCO_ENABLE; } -static int ironlake_crtc_mode_set(struct intel_crtc *crtc, - int x, int y, - struct drm_framebuffer *fb) +static int ironlake_crtc_compute_clock(struct intel_crtc *crtc) { struct drm_device *dev = crtc->base.dev; intel_clock_t clock, reduced_clock; @@ -7466,17 +7464,13 @@ static int ironlake_crtc_mode_set(struct intel_crtc *crtc, else crtc->new_config->dpll_hw_state.fp1 = fp; - if (intel_crtc_to_shared_dpll(crtc)) - intel_put_shared_dpll(crtc); - pll = intel_get_shared_dpll(crtc); if (pll == NULL) { DRM_DEBUG_DRIVER("failed to find PLL for pipe %c\n", pipe_name(crtc->pipe)); return -EINVAL; } - } else - intel_put_shared_dpll(crtc); + } if (is_lvds && has_reduced_clock && i915.powersave) crtc->lowfreq_avail = true; @@ -12480,7 +12474,8 @@ static void intel_init_display(struct drm_device *dev) } else if (HAS_PCH_SPLIT(dev)) { dev_priv->display.get_pipe_config = ironlake_get_pipe_config; dev_priv->display.get_plane_config = ironlake_get_plane_config; - dev_priv->display.crtc_mode_set = ironlake_crtc_mode_set; + dev_priv->display.crtc_compute_clock = + ironlake_crtc_compute_clock; dev_priv->display.crtc_enable = ironlake_crtc_enable; dev_priv->display.crtc_disable = ironlake_crtc_disable; dev_priv->display.off = ironlake_crtc_off; -- cgit v0.10.2 From d6dfee7a038ed1d160a557de50926b04dbcfbb6a Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:36 +0200 Subject: drm/i915: Covert remaining platforms to choose DPLLS before disabling CRTCs Use the infrastructure added in a previous patch to choose shared DPLLs and calculate clocks before touching the hardware. v2: Don't set mode_set hooks since dev_priv is kzalloc()'d (Ville) Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ee5eba7..5af00c4 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -6395,9 +6395,7 @@ static void i9xx_set_pipeconf(struct intel_crtc *intel_crtc) POSTING_READ(PIPECONF(intel_crtc->pipe)); } -static int i9xx_crtc_mode_set(struct intel_crtc *crtc, - int x, int y, - struct drm_framebuffer *fb) +static int i9xx_crtc_compute_clock(struct intel_crtc *crtc) { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; @@ -12484,7 +12482,7 @@ static void intel_init_display(struct drm_device *dev) } else if (IS_VALLEYVIEW(dev)) { dev_priv->display.get_pipe_config = i9xx_get_pipe_config; dev_priv->display.get_plane_config = i9xx_get_plane_config; - dev_priv->display.crtc_mode_set = i9xx_crtc_mode_set; + dev_priv->display.crtc_compute_clock = i9xx_crtc_compute_clock; dev_priv->display.crtc_enable = valleyview_crtc_enable; dev_priv->display.crtc_disable = i9xx_crtc_disable; dev_priv->display.off = i9xx_crtc_off; @@ -12493,7 +12491,7 @@ static void intel_init_display(struct drm_device *dev) } else { dev_priv->display.get_pipe_config = i9xx_get_pipe_config; dev_priv->display.get_plane_config = i9xx_get_plane_config; - dev_priv->display.crtc_mode_set = i9xx_crtc_mode_set; + dev_priv->display.crtc_compute_clock = i9xx_crtc_compute_clock; dev_priv->display.crtc_enable = i9xx_crtc_enable; dev_priv->display.crtc_disable = i9xx_crtc_disable; dev_priv->display.off = i9xx_crtc_off; -- cgit v0.10.2 From f3019a4d92f08b2dd92443a4b567a066a51c6ec0 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:37 +0200 Subject: drm/i915: Remove crtc_mode_set() hook There's no users left after the conversion to calculate clocks before disabling crtcs during mode set. Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 4fb93e2..03be342 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -487,9 +487,6 @@ struct drm_i915_display_funcs { void (*get_plane_config)(struct intel_crtc *, struct intel_plane_config *); int (*crtc_compute_clock)(struct intel_crtc *crtc); - int (*crtc_mode_set)(struct intel_crtc *crtc, - int x, int y, - struct drm_framebuffer *old_fb); void (*crtc_enable)(struct drm_crtc *crtc); void (*crtc_disable)(struct drm_crtc *crtc); void (*off)(struct drm_crtc *crtc); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5af00c4..200fcb1 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -10882,13 +10882,6 @@ static int __intel_set_mode(struct drm_crtc *crtc, crtc->primary->fb = fb; crtc->x = x; crtc->y = y; - - if (dev_priv->display.crtc_mode_set) { - ret = dev_priv->display.crtc_mode_set(intel_crtc, - x, y, fb); - if (ret) - goto done; - } } /* Now enable the clocks, plane, pipe, and connectors that we set up. */ -- cgit v0.10.2 From e9f1af393477bd4a60e23531b579b325cf1918da Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Wed, 29 Oct 2014 11:32:38 +0200 Subject: drm/i915: Don't store current shared DPLL in the new pipe_config Now that shared DPLLs configuration is staged, there's no need to track the current ones in the new pipe_config since those are released before making the new pipe_config effective. Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 200fcb1..be59617 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -5533,14 +5533,6 @@ static int intel_crtc_compute_config(struct intel_crtc *crtc, if (HAS_IPS(dev)) hsw_compute_ips_config(crtc, pipe_config); - /* - * XXX: PCH/WRPLL clock sharing is done in ->mode_set if ->compute_clock is not - * set, so make sure the old clock survives for now. - */ - if (dev_priv->display.crtc_compute_clock == NULL && - (HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev) || HAS_DDI(dev))) - pipe_config->shared_dpll = crtc->config.shared_dpll; - if (pipe_config->has_pch_encoder) return ironlake_fdi_compute_config(crtc, pipe_config); -- cgit v0.10.2 From ba41c0dec7786ca2d5f630d0f364c5f5331b34c1 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 3 Nov 2014 15:04:55 +0100 Subject: drm/i915: Move pll state commit into intel_modeset_update_state It's really part of the "push all new_* state into current state pointers" done in that function. So let's move it there to make this clear. Also, with the conversion done the num_shared_dpll check the function does in it's loop is enough, so we can drop the check for the dpll compute callback, too. Cc: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index be59617..a05a857 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -10231,10 +10231,13 @@ static bool intel_crtc_in_use(struct drm_crtc *crtc) static void intel_modeset_update_state(struct drm_device *dev, unsigned prepare_pipes) { + struct drm_i915_private *dev_priv = dev->dev_private; struct intel_encoder *intel_encoder; struct intel_crtc *intel_crtc; struct drm_connector *connector; + intel_shared_dpll_commit(dev_priv); + for_each_intel_encoder(dev, intel_encoder) { if (!intel_encoder->base.crtc) continue; @@ -10838,9 +10841,6 @@ static int __intel_set_mode(struct drm_crtc *crtc, &pipe_config->adjusted_mode); } - if (dev_priv->display.crtc_compute_clock) - intel_shared_dpll_commit(dev_priv); - /* Only after disabling all output pipelines that will be changed can we * update the the output configuration. */ intel_modeset_update_state(dev, prepare_pipes); -- cgit v0.10.2 From 4093561b70adccec7d3c8fbb19a80d636fe0f562 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 29 Oct 2014 11:16:59 +0000 Subject: drm/i915: Make intel_pipe_has_type() take an output type enum As Paulo said when introducing the enum, having more types is really good to document what should go where (int foo(int, int, bool, bool). Cc: Paulo Zanoni Signed-off-by: Damien Lespiau Reviewed-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index a05a857..d795142 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -408,7 +408,7 @@ static void vlv_clock(int refclk, intel_clock_t *clock) /** * Returns whether any output on the specified pipe is of the specified type */ -bool intel_pipe_has_type(struct intel_crtc *crtc, int type) +bool intel_pipe_has_type(struct intel_crtc *crtc, enum intel_output_type type) { struct drm_device *dev = crtc->base.dev; struct intel_encoder *encoder; diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 82d8916..7fab2d8 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -881,7 +881,7 @@ int intel_get_pipe_from_crtc_id(struct drm_device *dev, void *data, struct drm_file *file_priv); enum transcoder intel_pipe_to_cpu_transcoder(struct drm_i915_private *dev_priv, enum pipe pipe); -bool intel_pipe_has_type(struct intel_crtc *crtc, int type); +bool intel_pipe_has_type(struct intel_crtc *crtc, enum intel_output_type type); static inline void intel_wait_for_vblank(struct drm_device *dev, int pipe) { -- cgit v0.10.2 From 7637b6bd4e20b877f77a4e0bd1ac48b1b1b064e1 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Wed, 29 Oct 2014 12:03:45 +0000 Subject: drm/i915: Remove unused WATCH_GTT define Chris removed the code using it in: commit be2d599b5da3936ca92e0187ff50b34b6b8ff997 Author: Chris Wilson Date: Wed Sep 10 19:52:18 2014 +0100 drm/i915: Remove dead code, i915_gem_verify_gtt Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 03be342..f4c2db4 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -284,7 +284,6 @@ void intel_link_compute_m_n(int bpp, int nlanes, #define DRIVER_PATCHLEVEL 0 #define WATCH_LISTS 0 -#define WATCH_GTT 0 struct opregion_header; struct opregion_acpi; -- cgit v0.10.2 From d6faadb7a2e9b56cf83e45205a97dde17aab71af Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 3 Nov 2014 11:39:24 +0100 Subject: drm/i915/dp: Don't stop the link when retraining On pre-ddi platforms we don't shut down the link when changing link training parameters. Except when clock recovery fails too hard and we restart with channel eq training. Which doesn't make a lot of sense really, since just stopping/restarting the DP port at this point violates the modeset sequence documented in the Bspec. So let's tempt fate and try this. This patch is motivated by a WARN_ON triggered by commit bc76e320f21f8bd790a72bd5dc06909617432352 Author: Daniel Vetter Date: Tue May 20 22:46:50 2014 +0200 drm/i915: Drop now misleading DDI comment from dp_link_down References: https://bugs.freedesktop.org/show_bug.cgi?id=85670 Signed-off-by: Daniel Vetter Acked-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index ec80e82..5f2c206 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3812,7 +3812,6 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp) /* Try 5 times, then try clock recovery if that fails */ if (tries > 5) { - intel_dp_link_down(intel_dp); intel_dp_start_link_train(intel_dp); intel_dp_set_link_train(intel_dp, &DP, training_pattern | -- cgit v0.10.2 From 816a360489b08ccba323fa0d02fc03f85bd70321 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Sat, 1 Nov 2014 05:05:56 +0000 Subject: drm/i915: Removed orphaned prototype intel_dp_handle_hpd_irq() The function was removed in: commit 0e32b39ceed665bfa4a77a4bc307b6652b991632 Author: Dave Airlie Date: Fri May 2 14:02:48 2014 +1000 drm/i915: add DP 1.2 MST support (v0.7) Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 7fab2d8..dc53cfe 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -992,7 +992,6 @@ void intel_edp_psr_flush(struct drm_device *dev, unsigned frontbuffer_bits); void intel_edp_psr_init(struct drm_device *dev); -int intel_dp_handle_hpd_irq(struct intel_digital_port *digport, bool long_hpd); void intel_dp_add_properties(struct intel_dp *intel_dp, struct drm_connector *connector); void intel_dp_mst_suspend(struct drm_device *dev); void intel_dp_mst_resume(struct drm_device *dev); -- cgit v0.10.2 From 7f518dd09e246e96603533167d13b2ede27c20f7 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Sat, 1 Nov 2014 05:05:57 +0000 Subject: drm/i915: Remove orphaned prototype gen6_set_pm_mask() The function was removed in: commit 037bde19a43e299d30f0490bba9be32ab355975c Author: Chris Wilson Date: Thu Mar 27 08:24:19 2014 +0000 Revert "drm/i915: Disable/Enable PM Intrrupts based on the current freq." Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index f4c2db4..54691bc 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2298,8 +2298,6 @@ __printf(3, 4) void i915_handle_error(struct drm_device *dev, bool wedged, const char *fmt, ...); -void gen6_set_pm_mask(struct drm_i915_private *dev_priv, u32 pm_iir, - int new_delay); extern void intel_irq_init(struct drm_i915_private *dev_priv); extern void intel_hpd_init(struct drm_i915_private *dev_priv); int intel_irq_install(struct drm_i915_private *dev_priv); -- cgit v0.10.2 From 1df06b75f0235c6c4f836e5c03fb499ba70a54bd Mon Sep 17 00:00:00 2001 From: Thomas Daniel Date: Wed, 29 Oct 2014 09:52:51 +0000 Subject: drm/i915/bdw: Setup global hardware status page in execlists mode Write HWS_PGA address even in execlists mode as the global hardware status page is still required. This address was previously uninitialized and HWSP writes would clobber whatever buffer happened to reside at GGTT address 0. v2: Break out hardware status page setup into a separate function. Issue: VIZ-2020 Signed-off-by: Thomas Daniel Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 70a6c63..292beb0 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1649,6 +1649,27 @@ static uint32_t get_lr_context_size(struct intel_engine_cs *ring) return ret; } +static int lrc_setup_hardware_status_page(struct intel_engine_cs *ring, + struct drm_i915_gem_object *default_ctx_obj) +{ + struct drm_i915_private *dev_priv = ring->dev->dev_private; + + /* The status page is offset 0 from the default context object + * in LRC mode. */ + ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(default_ctx_obj); + ring->status_page.page_addr = + kmap(sg_page(default_ctx_obj->pages->sgl)); + if (ring->status_page.page_addr == NULL) + return -ENOMEM; + ring->status_page.obj = default_ctx_obj; + + I915_WRITE(RING_HWS_PGA(ring->mmio_base), + (u32)ring->status_page.gfx_addr); + POSTING_READ(RING_HWS_PGA(ring->mmio_base)); + + return 0; +} + /** * intel_lr_context_deferred_create() - create the LRC specific bits of a context * @ctx: LR context to create. @@ -1734,14 +1755,11 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, ctx->engine[ring->id].state = ctx_obj; if (ctx == ring->default_context) { - /* The status page is offset 0 from the default context object - * in LRC mode. */ - ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(ctx_obj); - ring->status_page.page_addr = - kmap(sg_page(ctx_obj->pages->sgl)); - if (ring->status_page.page_addr == NULL) - return -ENOMEM; - ring->status_page.obj = ctx_obj; + ret = lrc_setup_hardware_status_page(ring, ctx_obj); + if (ret) { + DRM_ERROR("Failed to setup hardware status page\n"); + goto error; + } } if (ring->id == RCS && !ctx->rcs_initialized) { -- cgit v0.10.2 From 850c4cdc6c223d02748f20c8ee1b677fb42da1aa Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Thu, 30 Oct 2014 16:39:38 +0000 Subject: drm/i915: Make intel_pin_and_fence_fb_obj take plane and framebuffer It will help future code if this function knows something about of the context of the display setup object is being pinned for. Signed-off-by: Tvrtko Ursulin Cc: Daniel Vetter Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index d795142..44060fb 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2194,11 +2194,13 @@ static int intel_align_height(struct drm_device *dev, int height, bool tiled) } int -intel_pin_and_fence_fb_obj(struct drm_device *dev, - struct drm_i915_gem_object *obj, +intel_pin_and_fence_fb_obj(struct drm_plane *plane, + struct drm_framebuffer *fb, struct intel_engine_cs *pipelined) { + struct drm_device *dev = fb->dev; struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj = intel_fb_obj(fb); u32 alignment; int ret; @@ -2897,7 +2899,6 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y, struct intel_crtc *intel_crtc = to_intel_crtc(crtc); enum pipe pipe = intel_crtc->pipe; struct drm_framebuffer *old_fb = crtc->primary->fb; - struct drm_i915_gem_object *obj = intel_fb_obj(fb); struct drm_i915_gem_object *old_obj = intel_fb_obj(old_fb); int ret; @@ -2920,9 +2921,9 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y, } mutex_lock(&dev->struct_mutex); - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); + ret = intel_pin_and_fence_fb_obj(crtc->primary, fb, NULL); if (ret == 0) - i915_gem_track_fb(old_obj, obj, + i915_gem_track_fb(old_obj, intel_fb_obj(fb), INTEL_FRONTBUFFER_PRIMARY(pipe)); mutex_unlock(&dev->struct_mutex); if (ret != 0) { @@ -9689,7 +9690,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, ring = &dev_priv->ring[RCS]; } - ret = intel_pin_and_fence_fb_obj(dev, obj, ring); + ret = intel_pin_and_fence_fb_obj(crtc->primary, fb, ring); if (ret) goto cleanup_pending; @@ -10857,9 +10858,7 @@ static int __intel_set_mode(struct drm_crtc *crtc, struct drm_i915_gem_object *obj = intel_fb_obj(fb); mutex_lock(&dev->struct_mutex); - ret = intel_pin_and_fence_fb_obj(dev, - obj, - NULL); + ret = intel_pin_and_fence_fb_obj(crtc->primary, fb, NULL); if (ret != 0) { DRM_ERROR("pin & fence failed\n"); mutex_unlock(&dev->struct_mutex); @@ -11525,7 +11524,7 @@ intel_prepare_primary_plane(struct drm_plane *plane, if (old_obj != obj) { mutex_lock(&dev->struct_mutex); - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); + ret = intel_pin_and_fence_fb_obj(plane, fb, NULL); if (ret == 0) i915_gem_track_fb(old_obj, obj, INTEL_FRONTBUFFER_PRIMARY(pipe)); @@ -13293,7 +13292,9 @@ void intel_modeset_gem_init(struct drm_device *dev) if (obj == NULL) continue; - if (intel_pin_and_fence_fb_obj(dev, obj, NULL)) { + if (intel_pin_and_fence_fb_obj(c->primary, + c->primary->fb, + NULL)) { DRM_ERROR("failed to pin boot fb on pipe %d\n", to_intel_crtc(c)->pipe); drm_framebuffer_unreference(c->primary->fb); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index dc53cfe..ae75b99 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -896,8 +896,8 @@ bool intel_get_load_detect_pipe(struct drm_connector *connector, struct drm_modeset_acquire_ctx *ctx); void intel_release_load_detect_pipe(struct drm_connector *connector, struct intel_load_detect_pipe *old); -int intel_pin_and_fence_fb_obj(struct drm_device *dev, - struct drm_i915_gem_object *obj, +int intel_pin_and_fence_fb_obj(struct drm_plane *plane, + struct drm_framebuffer *fb, struct intel_engine_cs *pipelined); void intel_unpin_fb_obj(struct drm_i915_gem_object *obj); struct drm_framebuffer * diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c index 9b584f3..f2183b5 100644 --- a/drivers/gpu/drm/i915/intel_fbdev.c +++ b/drivers/gpu/drm/i915/intel_fbdev.c @@ -119,25 +119,25 @@ static int intelfb_alloc(struct drm_fb_helper *helper, goto out; } - /* Flush everything out, we'll be doing GTT only from now on */ - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); - if (ret) { - DRM_ERROR("failed to pin obj: %d\n", ret); - goto out_unref; - } - fb = __intel_framebuffer_create(dev, &mode_cmd, obj); if (IS_ERR(fb)) { ret = PTR_ERR(fb); - goto out_unpin; + goto out_unref; + } + + /* Flush everything out, we'll be doing GTT only from now on */ + ret = intel_pin_and_fence_fb_obj(NULL, fb, NULL); + if (ret) { + DRM_ERROR("failed to pin obj: %d\n", ret); + goto out_fb; } ifbdev->fb = to_intel_framebuffer(fb); return 0; -out_unpin: - i915_gem_object_ggtt_unpin(obj); +out_fb: + drm_framebuffer_remove(fb); out_unref: drm_gem_object_unreference(&obj->base); out: diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index bf1eb15..8c70c54 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -1256,7 +1256,7 @@ intel_prepare_sprite_plane(struct drm_plane *plane, * the sprite planes only require 128KiB alignment and 32 PTE * padding. */ - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); + ret = intel_pin_and_fence_fb_obj(plane, fb, NULL); if (ret == 0) i915_gem_track_fb(old_obj, obj, INTEL_FRONTBUFFER_SPRITE(pipe)); -- cgit v0.10.2 From 7d1bd53931e5afe5b2549c045ed4afad71ee6e8f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:50 +0200 Subject: drm/i915: Apply some ocd for IMR vs. IER order during irq enable MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When disabling interrupts we do the writes in this order: IMR,IER,IIR,IIR. But when enabling interrupts we don't do use the mirrored order, and instead do IIR,IIR,IMR,IER. I like consistency unless there's a good reason against it, which I can't think of here, so change the enable order to IIR,IIR,IER,IMR. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index a2b013d..98a8d65 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -126,16 +126,16 @@ static const u32 hpd_status_i915[] = { /* i915 and valleyview are the same */ #define GEN8_IRQ_INIT_NDX(type, which, imr_val, ier_val) do { \ GEN5_ASSERT_IIR_IS_ZERO(GEN8_##type##_IIR(which)); \ - I915_WRITE(GEN8_##type##_IMR(which), (imr_val)); \ I915_WRITE(GEN8_##type##_IER(which), (ier_val)); \ - POSTING_READ(GEN8_##type##_IER(which)); \ + I915_WRITE(GEN8_##type##_IMR(which), (imr_val)); \ + POSTING_READ(GEN8_##type##_IMR(which)); \ } while (0) #define GEN5_IRQ_INIT(type, imr_val, ier_val) do { \ GEN5_ASSERT_IIR_IS_ZERO(type##IIR); \ - I915_WRITE(type##IMR, (imr_val)); \ I915_WRITE(type##IER, (ier_val)); \ - POSTING_READ(type##IER); \ + I915_WRITE(type##IMR, (imr_val)); \ + POSTING_READ(type##IMR); \ } while (0) /* For display hotplug interrupt */ -- cgit v0.10.2 From 7c4cde39684cffc6aaae4f34f2bcd987a86befe6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:51 +0200 Subject: drm/i915: Use DPINVGTT_STATUS_MASK MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Some has given a name for the DPINVGTT status bitmask, so let's use it instead of the magic number. Looks more like the chv code now. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 98a8d65..e41272d 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3122,7 +3122,7 @@ static void valleyview_irq_preinstall(struct drm_device *dev) gen5_gt_irq_reset(dev); - I915_WRITE(DPINVGTT, 0xff); + I915_WRITE(DPINVGTT, DPINVGTT_STATUS_MASK); I915_WRITE(PORT_HOTPLUG_EN, 0); I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); -- cgit v0.10.2 From a2c30fbafceb937eb3c3535a21810d4458628e62 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:52 +0200 Subject: drm/i915: Use gen8_gt_irq_reset() in cherryview_irq_uninstall() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Replace the hand rolled macros with gen8_gt_irq_reset() and GEN5_IRQ_RESET() in cherryview_irq_uninstall(). Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index e41272d..1ec4ebb 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3625,33 +3625,9 @@ static void cherryview_irq_uninstall(struct drm_device *dev) I915_WRITE(GEN8_MASTER_IRQ, 0); POSTING_READ(GEN8_MASTER_IRQ); -#define GEN8_IRQ_FINI_NDX(type, which) \ -do { \ - I915_WRITE(GEN8_##type##_IMR(which), 0xffffffff); \ - I915_WRITE(GEN8_##type##_IER(which), 0); \ - I915_WRITE(GEN8_##type##_IIR(which), 0xffffffff); \ - POSTING_READ(GEN8_##type##_IIR(which)); \ - I915_WRITE(GEN8_##type##_IIR(which), 0xffffffff); \ -} while (0) - -#define GEN8_IRQ_FINI(type) \ -do { \ - I915_WRITE(GEN8_##type##_IMR, 0xffffffff); \ - I915_WRITE(GEN8_##type##_IER, 0); \ - I915_WRITE(GEN8_##type##_IIR, 0xffffffff); \ - POSTING_READ(GEN8_##type##_IIR); \ - I915_WRITE(GEN8_##type##_IIR, 0xffffffff); \ -} while (0) - - GEN8_IRQ_FINI_NDX(GT, 0); - GEN8_IRQ_FINI_NDX(GT, 1); - GEN8_IRQ_FINI_NDX(GT, 2); - GEN8_IRQ_FINI_NDX(GT, 3); - - GEN8_IRQ_FINI(PCU); + gen8_gt_irq_reset(dev_priv); -#undef GEN8_IRQ_FINI -#undef GEN8_IRQ_FINI_NDX + GEN5_IRQ_RESET(GEN8_PCU_); I915_WRITE(PORT_HOTPLUG_EN, 0); I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); -- cgit v0.10.2 From 8e20599ae1327ef3f0cf6a080fc3a39c882f4718 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:53 +0200 Subject: drm/i915: Drop the extra GEN8_PCU_IIR posting read from cherryview_irq_preinstall() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Looks like a leftover POSTING_READ(GEN8_PCU_IIR) in cherryview_irq_preinstall() from some earlier age. GEN5_IRQ_RESET() already does the posting read so this changes nothing, so kill it. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 1ec4ebb..1e4062d 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3188,8 +3188,6 @@ static void cherryview_irq_preinstall(struct drm_device *dev) GEN5_IRQ_RESET(GEN8_PCU_); - POSTING_READ(GEN8_PCU_IIR); - I915_WRITE(DPINVGTT, DPINVGTT_STATUS_MASK_CHV); I915_WRITE(PORT_HOTPLUG_EN, 0); -- cgit v0.10.2 From 76e418605448597159c5decf4e060d41d3618831 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:54 +0200 Subject: drm/i915: Use a consistent order between IIR, IER, IMR writes on vlv/chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Follow the same ordering rules for the IIR,IER,IMR writes on vlv/chv that we do on other gen5+ platforms. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 1e4062d..589ae51 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3128,10 +3128,11 @@ static void valleyview_irq_preinstall(struct drm_device *dev) I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); for_each_pipe(dev_priv, pipe) I915_WRITE(PIPESTAT(pipe), 0xffff); - I915_WRITE(VLV_IIR, 0xffffffff); I915_WRITE(VLV_IMR, 0xffffffff); I915_WRITE(VLV_IER, 0x0); - POSTING_READ(VLV_IER); + I915_WRITE(VLV_IIR, 0xffffffff); + I915_WRITE(VLV_IIR, 0xffffffff); + POSTING_READ(VLV_IIR); } static void gen8_gt_irq_reset(struct drm_i915_private *dev_priv) @@ -3199,6 +3200,7 @@ static void cherryview_irq_preinstall(struct drm_device *dev) I915_WRITE(VLV_IMR, 0xffffffff); I915_WRITE(VLV_IER, 0x0); I915_WRITE(VLV_IIR, 0xffffffff); + I915_WRITE(VLV_IIR, 0xffffffff); POSTING_READ(VLV_IIR); } @@ -3362,9 +3364,9 @@ static void valleyview_display_irqs_install(struct drm_i915_private *dev_priv) I915_WRITE(VLV_IIR, iir_mask); I915_WRITE(VLV_IIR, iir_mask); - I915_WRITE(VLV_IMR, dev_priv->irq_mask); I915_WRITE(VLV_IER, ~dev_priv->irq_mask); - POSTING_READ(VLV_IER); + I915_WRITE(VLV_IMR, dev_priv->irq_mask); + POSTING_READ(VLV_IMR); } static void valleyview_display_irqs_uninstall(struct drm_i915_private *dev_priv) @@ -3377,8 +3379,8 @@ static void valleyview_display_irqs_uninstall(struct drm_i915_private *dev_priv) I915_DISPLAY_PIPE_B_EVENT_INTERRUPT; dev_priv->irq_mask |= iir_mask; - I915_WRITE(VLV_IER, ~dev_priv->irq_mask); I915_WRITE(VLV_IMR, dev_priv->irq_mask); + I915_WRITE(VLV_IER, ~dev_priv->irq_mask); I915_WRITE(VLV_IIR, iir_mask); I915_WRITE(VLV_IIR, iir_mask); POSTING_READ(VLV_IIR); @@ -3432,10 +3434,11 @@ static int valleyview_irq_postinstall(struct drm_device *dev) I915_WRITE(PORT_HOTPLUG_EN, 0); POSTING_READ(PORT_HOTPLUG_EN); - I915_WRITE(VLV_IMR, dev_priv->irq_mask); - I915_WRITE(VLV_IER, ~dev_priv->irq_mask); I915_WRITE(VLV_IIR, 0xffffffff); - POSTING_READ(VLV_IER); + I915_WRITE(VLV_IIR, 0xffffffff); + I915_WRITE(VLV_IER, ~dev_priv->irq_mask); + I915_WRITE(VLV_IMR, dev_priv->irq_mask); + POSTING_READ(VLV_IMR); /* Interrupt setup is already guaranteed to be single-threaded, this is * just to make the assert_spin_locked check happy. */ @@ -3559,8 +3562,10 @@ static int cherryview_irq_postinstall(struct drm_device *dev) spin_unlock_irq(&dev_priv->irq_lock); I915_WRITE(VLV_IIR, 0xffffffff); - I915_WRITE(VLV_IMR, dev_priv->irq_mask); + I915_WRITE(VLV_IIR, 0xffffffff); I915_WRITE(VLV_IER, enable_mask); + I915_WRITE(VLV_IMR, dev_priv->irq_mask); + POSTING_READ(VLV_IMR); gen8_gt_irq_postinstall(dev_priv); @@ -3606,10 +3611,11 @@ static void valleyview_irq_uninstall(struct drm_device *dev) dev_priv->irq_mask = 0; - I915_WRITE(VLV_IIR, 0xffffffff); I915_WRITE(VLV_IMR, 0xffffffff); I915_WRITE(VLV_IER, 0x0); - POSTING_READ(VLV_IER); + I915_WRITE(VLV_IIR, 0xffffffff); + I915_WRITE(VLV_IIR, 0xffffffff); + POSTING_READ(VLV_IIR); } static void cherryview_irq_uninstall(struct drm_device *dev) @@ -3636,6 +3642,7 @@ static void cherryview_irq_uninstall(struct drm_device *dev) I915_WRITE(VLV_IMR, 0xffffffff); I915_WRITE(VLV_IER, 0x0); I915_WRITE(VLV_IIR, 0xffffffff); + I915_WRITE(VLV_IIR, 0xffffffff); POSTING_READ(VLV_IIR); } -- cgit v0.10.2 From 23a09c76cc37c132844345cec94abc1edf570b53 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:55 +0200 Subject: drm/i915: Use GEN5_IRQ_RESET() on vlv/chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Replace the hand rolled IIR,IER,IMR disable sequences with GEN5_IRQ_RESET(). Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 589ae51..c106bba 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3128,11 +3128,8 @@ static void valleyview_irq_preinstall(struct drm_device *dev) I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); for_each_pipe(dev_priv, pipe) I915_WRITE(PIPESTAT(pipe), 0xffff); - I915_WRITE(VLV_IMR, 0xffffffff); - I915_WRITE(VLV_IER, 0x0); - I915_WRITE(VLV_IIR, 0xffffffff); - I915_WRITE(VLV_IIR, 0xffffffff); - POSTING_READ(VLV_IIR); + + GEN5_IRQ_RESET(VLV_); } static void gen8_gt_irq_reset(struct drm_i915_private *dev_priv) @@ -3197,11 +3194,7 @@ static void cherryview_irq_preinstall(struct drm_device *dev) for_each_pipe(dev_priv, pipe) I915_WRITE(PIPESTAT(pipe), 0xffff); - I915_WRITE(VLV_IMR, 0xffffffff); - I915_WRITE(VLV_IER, 0x0); - I915_WRITE(VLV_IIR, 0xffffffff); - I915_WRITE(VLV_IIR, 0xffffffff); - POSTING_READ(VLV_IIR); + GEN5_IRQ_RESET(VLV_); } static void ibx_hpd_irq_setup(struct drm_device *dev) @@ -3611,11 +3604,7 @@ static void valleyview_irq_uninstall(struct drm_device *dev) dev_priv->irq_mask = 0; - I915_WRITE(VLV_IMR, 0xffffffff); - I915_WRITE(VLV_IER, 0x0); - I915_WRITE(VLV_IIR, 0xffffffff); - I915_WRITE(VLV_IIR, 0xffffffff); - POSTING_READ(VLV_IIR); + GEN5_IRQ_RESET(VLV_); } static void cherryview_irq_uninstall(struct drm_device *dev) @@ -3639,11 +3628,7 @@ static void cherryview_irq_uninstall(struct drm_device *dev) for_each_pipe(dev_priv, pipe) I915_WRITE(PIPESTAT(pipe), 0xffff); - I915_WRITE(VLV_IMR, 0xffffffff); - I915_WRITE(VLV_IER, 0x0); - I915_WRITE(VLV_IIR, 0xffffffff); - I915_WRITE(VLV_IIR, 0xffffffff); - POSTING_READ(VLV_IIR); + GEN5_IRQ_RESET(VLV_); } static void ironlake_irq_uninstall(struct drm_device *dev) -- cgit v0.10.2 From 893fce8edae9feabb5515b982285e5af72fcfa95 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:56 +0200 Subject: drm/i915: Call gen5_gt_irq_reset() from valleyview_irq_uninstall() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Looks like we forgot to call gen5_gt_irq_reset() for vlv in the uninstall phase. Do so. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index c106bba..67c046b 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3588,6 +3588,8 @@ static void valleyview_irq_uninstall(struct drm_device *dev) I915_WRITE(VLV_MASTER_IER, 0); + gen5_gt_irq_reset(dev); + for_each_pipe(dev_priv, pipe) I915_WRITE(PIPESTAT(pipe), 0xffff); -- cgit v0.10.2 From 120dda4f638fe65d69674a674c3c1d639b31b4f1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:57 +0200 Subject: drm/i915: Make valleyview_display_irqs_(un)install() work for chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Genralize valleyview_display_irqs_install() and valleyview_display_irqs_uninstall() enough so that they work on chv. The only difference to vlv here being the third pipe that chv brings. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 67c046b..50431f6 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3335,24 +3335,27 @@ static void valleyview_display_irqs_install(struct drm_i915_private *dev_priv) { u32 pipestat_mask; u32 iir_mask; + enum pipe pipe; pipestat_mask = PIPESTAT_INT_STATUS_MASK | PIPE_FIFO_UNDERRUN_STATUS; - I915_WRITE(PIPESTAT(PIPE_A), pipestat_mask); - I915_WRITE(PIPESTAT(PIPE_B), pipestat_mask); + for_each_pipe(dev_priv, pipe) + I915_WRITE(PIPESTAT(pipe), pipestat_mask); POSTING_READ(PIPESTAT(PIPE_A)); pipestat_mask = PLANE_FLIP_DONE_INT_STATUS_VLV | PIPE_CRC_DONE_INTERRUPT_STATUS; - i915_enable_pipestat(dev_priv, PIPE_A, pipestat_mask | - PIPE_GMBUS_INTERRUPT_STATUS); - i915_enable_pipestat(dev_priv, PIPE_B, pipestat_mask); + i915_enable_pipestat(dev_priv, PIPE_A, PIPE_GMBUS_INTERRUPT_STATUS); + for_each_pipe(dev_priv, pipe) + i915_enable_pipestat(dev_priv, pipe, pipestat_mask); iir_mask = I915_DISPLAY_PORT_INTERRUPT | I915_DISPLAY_PIPE_A_EVENT_INTERRUPT | I915_DISPLAY_PIPE_B_EVENT_INTERRUPT; + if (IS_CHERRYVIEW(dev_priv)) + iir_mask |= I915_DISPLAY_PIPE_C_EVENT_INTERRUPT; dev_priv->irq_mask &= ~iir_mask; I915_WRITE(VLV_IIR, iir_mask); @@ -3366,10 +3369,13 @@ static void valleyview_display_irqs_uninstall(struct drm_i915_private *dev_priv) { u32 pipestat_mask; u32 iir_mask; + enum pipe pipe; iir_mask = I915_DISPLAY_PORT_INTERRUPT | I915_DISPLAY_PIPE_A_EVENT_INTERRUPT | I915_DISPLAY_PIPE_B_EVENT_INTERRUPT; + if (IS_CHERRYVIEW(dev_priv)) + iir_mask |= I915_DISPLAY_PIPE_C_EVENT_INTERRUPT; dev_priv->irq_mask |= iir_mask; I915_WRITE(VLV_IMR, dev_priv->irq_mask); @@ -3381,14 +3387,15 @@ static void valleyview_display_irqs_uninstall(struct drm_i915_private *dev_priv) pipestat_mask = PLANE_FLIP_DONE_INT_STATUS_VLV | PIPE_CRC_DONE_INTERRUPT_STATUS; - i915_disable_pipestat(dev_priv, PIPE_A, pipestat_mask | - PIPE_GMBUS_INTERRUPT_STATUS); - i915_disable_pipestat(dev_priv, PIPE_B, pipestat_mask); + i915_disable_pipestat(dev_priv, PIPE_A, PIPE_GMBUS_INTERRUPT_STATUS); + for_each_pipe(dev_priv, pipe) + i915_disable_pipestat(dev_priv, pipe, pipestat_mask); pipestat_mask = PIPESTAT_INT_STATUS_MASK | PIPE_FIFO_UNDERRUN_STATUS; - I915_WRITE(PIPESTAT(PIPE_A), pipestat_mask); - I915_WRITE(PIPESTAT(PIPE_B), pipestat_mask); + + for_each_pipe(dev_priv, pipe) + I915_WRITE(PIPESTAT(pipe), pipestat_mask); POSTING_READ(PIPESTAT(PIPE_A)); } -- cgit v0.10.2 From 70591a4101a4e5c1b89dc98a706406368a23c8ec Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:42:58 +0200 Subject: drm/i915: Refactor vlv_display_irq_reset() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Pull the vlv display irq reset code to a new functions. The aim is to share the code with chv. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 50431f6..38e57dd 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3105,10 +3105,22 @@ static void ironlake_irq_reset(struct drm_device *dev) ibx_irq_reset(dev); } +static void vlv_display_irq_reset(struct drm_i915_private *dev_priv) +{ + enum pipe pipe; + + I915_WRITE(PORT_HOTPLUG_EN, 0); + I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); + + for_each_pipe(dev_priv, pipe) + I915_WRITE(PIPESTAT(pipe), 0xffff); + + GEN5_IRQ_RESET(VLV_); +} + static void valleyview_irq_preinstall(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - int pipe; /* VLV magic */ I915_WRITE(VLV_IMR, 0); @@ -3124,12 +3136,7 @@ static void valleyview_irq_preinstall(struct drm_device *dev) I915_WRITE(DPINVGTT, DPINVGTT_STATUS_MASK); - I915_WRITE(PORT_HOTPLUG_EN, 0); - I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); - for_each_pipe(dev_priv, pipe) - I915_WRITE(PIPESTAT(pipe), 0xffff); - - GEN5_IRQ_RESET(VLV_); + vlv_display_irq_reset(dev_priv); } static void gen8_gt_irq_reset(struct drm_i915_private *dev_priv) @@ -3177,7 +3184,6 @@ void gen8_irq_power_well_post_enable(struct drm_i915_private *dev_priv) static void cherryview_irq_preinstall(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - int pipe; I915_WRITE(GEN8_MASTER_IRQ, 0); POSTING_READ(GEN8_MASTER_IRQ); @@ -3188,13 +3194,7 @@ static void cherryview_irq_preinstall(struct drm_device *dev) I915_WRITE(DPINVGTT, DPINVGTT_STATUS_MASK_CHV); - I915_WRITE(PORT_HOTPLUG_EN, 0); - I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); - - for_each_pipe(dev_priv, pipe) - I915_WRITE(PIPESTAT(pipe), 0xffff); - - GEN5_IRQ_RESET(VLV_); + vlv_display_irq_reset(dev_priv); } static void ibx_hpd_irq_setup(struct drm_device *dev) @@ -3588,7 +3588,6 @@ static void gen8_irq_uninstall(struct drm_device *dev) static void valleyview_irq_uninstall(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - int pipe; if (!dev_priv) return; @@ -3597,12 +3596,7 @@ static void valleyview_irq_uninstall(struct drm_device *dev) gen5_gt_irq_reset(dev); - for_each_pipe(dev_priv, pipe) - I915_WRITE(PIPESTAT(pipe), 0xffff); - I915_WRITE(HWSTAM, 0xffffffff); - I915_WRITE(PORT_HOTPLUG_EN, 0); - I915_WRITE(PORT_HOTPLUG_STAT, I915_READ(PORT_HOTPLUG_STAT)); /* Interrupt setup is already guaranteed to be single-threaded, this is * just to make the assert_spin_locked check happy. */ @@ -3611,9 +3605,9 @@ static void valleyview_irq_uninstall(struct drm_device *dev) valleyview_display_irqs_uninstall(dev_priv); spin_unlock_irq(&dev_priv->irq_lock); - dev_priv->irq_mask = 0; + vlv_display_irq_reset(dev_priv); - GEN5_IRQ_RESET(VLV_); + dev_priv->irq_mask = 0; } static void cherryview_irq_uninstall(struct drm_device *dev) -- cgit v0.10.2 From 0e6c9a9eb85b1cb38a7194630cb14c293614da8d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:43:00 +0200 Subject: drm/i914: Refactor vlv_display_irq_postinstall() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Split the vlv display irq postinstall code to a separate function so that we can share it with chv. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 38e57dd..297c9b1 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3425,10 +3425,8 @@ void valleyview_disable_display_irqs(struct drm_i915_private *dev_priv) valleyview_display_irqs_uninstall(dev_priv); } -static int valleyview_irq_postinstall(struct drm_device *dev) +static void vlv_display_irq_postinstall(struct drm_i915_private *dev_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - dev_priv->irq_mask = ~0; I915_WRITE(PORT_HOTPLUG_EN, 0); @@ -3449,6 +3447,13 @@ static int valleyview_irq_postinstall(struct drm_device *dev) I915_WRITE(VLV_IIR, 0xffffffff); I915_WRITE(VLV_IIR, 0xffffffff); +} + +static int valleyview_irq_postinstall(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + vlv_display_irq_postinstall(dev_priv); gen5_gt_irq_postinstall(dev); -- cgit v0.10.2 From 9b2e8c9411a81daf4772fbe30e0953f781f9b440 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:43:01 +0200 Subject: drm/i915: Drop useless VLV_IIR writes from vlv_display_irq_postinstall() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The extra VLV_IIR writes at the end of vlv_display_irq_postinstall() serve no purpose. Remove them. The VLV_IMR/IER/IIR setup at the start of the function also seems a bit pointless since it doesn't unmask/enable anything. But leave it be for now. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 297c9b1..f2a45f1 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3444,9 +3444,6 @@ static void vlv_display_irq_postinstall(struct drm_i915_private *dev_priv) if (dev_priv->display_irqs_enabled) valleyview_display_irqs_install(dev_priv); spin_unlock_irq(&dev_priv->irq_lock); - - I915_WRITE(VLV_IIR, 0xffffffff); - I915_WRITE(VLV_IIR, 0xffffffff); } static int valleyview_irq_postinstall(struct drm_device *dev) -- cgit v0.10.2 From 5eba929d1bb8fb662594b44d02f54fbaa7557a3b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 31 Oct 2014 11:53:06 +0200 Subject: drm/i915: Kill leftover GTIIR writes from valleyview_irq_preinstall() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There are two leftover GTIIR writes in valleyview_irq_preinstall(). Looks like the were originally left behind by: commit d18ea1b58a5003eb6fca03aff03c4c01321e6cb1 Author: Daniel Vetter Date: Fri Jul 12 22:43:25 2013 +0200 drm/i915: unify PM interrupt preinstall sequence and then the GTIIR reset was added back here: commit f86f3fb005d0c907285fa8685badcb24ec31ee59 Author: Paulo Zanoni Date: Tue Apr 1 15:37:14 2014 -0300 drm/i915: properly clear IIR at irq_uninstall on Gen5+ so we can kill the leftovers from the vlv code. Cc: Paulo Zanoni Suggested-by: Paulo Zanoni Signed-off-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index f2a45f1..318a6a0 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3128,10 +3128,6 @@ static void valleyview_irq_preinstall(struct drm_device *dev) I915_WRITE(RING_IMR(GEN6_BSD_RING_BASE), 0); I915_WRITE(RING_IMR(BLT_RING_BASE), 0); - /* and GT */ - I915_WRITE(GTIIR, I915_READ(GTIIR)); - I915_WRITE(GTIIR, I915_READ(GTIIR)); - gen5_gt_irq_reset(dev); I915_WRITE(DPINVGTT, DPINVGTT_STATUS_MASK); -- cgit v0.10.2 From 59a5d2907f652249209453af9748a8ca8ccf8655 Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Thu, 30 Oct 2014 15:52:45 -0200 Subject: drm/i915: fix "Unexpected fault" error message line break Fix the message, not the fault :) This is what I see: [ 282.108597] [drm:i915_check_and_clear_faults] Unexpected fault [ 282.108597] Addr: 0x00000000\n Address space: PPGTT [ 282.108597] Source ID: 24 [ 282.108597] Type: 0 Signed-off-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index cb7adab..de12017 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -1266,7 +1266,7 @@ void i915_check_and_clear_faults(struct drm_device *dev) fault_reg = I915_READ(RING_FAULT_REG(ring)); if (fault_reg & RING_FAULT_VALID) { DRM_DEBUG_DRIVER("Unexpected fault\n" - "\tAddr: 0x%08lx\\n" + "\tAddr: 0x%08lx\n" "\tAddress space: %s\n" "\tSource ID: %d\n" "\tType: %d\n", -- cgit v0.10.2 From fac6adb06a530ead523a60c975bae633e44335ca Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Thu, 30 Oct 2014 15:59:31 -0200 Subject: drm/i915: fix RPS on runtime suspend With this patch, the RPS sequence for runtime suspend/resume is exactly like the sequence for S3 suspend/resume: - flush_delayed_work(&dev_priv->rps.delayed_resume_work) - intel_runtime_pm_disable_interrupts() - intel_suspend_gt_powersave() (suspended) - intel_runtime_pm_enable_interrupts() - intel_enable_gt_powersave() With this, we get rid of WARNs that are currently intermittently triggered by the system-suspend-execbuf subtest of runtime PM. Notice that these WARNs could also be triggered in other ways that involved doing lots of RPM suspend/resume cycles just after a system S3 resume. Testcase: igt/pm_rpm/system-suspend-execbuf Signed-off-by: Paulo Zanoni Reference: https://bugs.freedesktop.org/show_bug.cgi?id=82939 Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 0c7cf48..2404b2b 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1395,13 +1395,9 @@ static int intel_runtime_suspend(struct device *device) i915_gem_release_all_mmaps(dev_priv); mutex_unlock(&dev->struct_mutex); - /* - * rps.work can't be rearmed here, since we get here only after making - * sure the GPU is idle and the RPS freq is set to the minimum. See - * intel_mark_idle(). - */ - cancel_work_sync(&dev_priv->rps.work); + flush_delayed_work(&dev_priv->rps.delayed_resume_work); intel_runtime_pm_disable_interrupts(dev_priv); + intel_suspend_gt_powersave(dev); ret = intel_suspend_complete(dev_priv); if (ret) { @@ -1473,7 +1469,7 @@ static int intel_runtime_resume(struct device *device) gen6_update_ring_freq(dev); intel_runtime_pm_enable_interrupts(dev_priv); - intel_reset_gt_powersave(dev); + intel_enable_gt_powersave(dev); if (ret) DRM_ERROR("Runtime resume failed, disabling it (%d)\n", ret); -- cgit v0.10.2 From 8e63954903779f0f70b4564d5cb0accb4b065c64 Mon Sep 17 00:00:00 2001 From: John Harrison Date: Thu, 30 Oct 2014 18:40:53 +0000 Subject: drm/i915: Remove redundant parameter to i915_gem_object_wait_rendering__tail() An earlier commit (c8725f3dc0911d4354315a65150aecd8b7d0d74a: Do not call retire_requests from wait_for_rendering) removed the use of the ring parameter within wait_rendering__tail() but did not remove the parameter itself. As the plan is to remove obj->ring which is where this parameter comes from, it is simpler to just remove the parameter completely than to update it with a new source. For: VIZ-4377 Signed-off-by: John Harrison CC: Chris Wilson CC: Brad Volkin Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 7e91978..c2456c2 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1281,8 +1281,7 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno) } static int -i915_gem_object_wait_rendering__tail(struct drm_i915_gem_object *obj, - struct intel_engine_cs *ring) +i915_gem_object_wait_rendering__tail(struct drm_i915_gem_object *obj) { if (!obj->active) return 0; @@ -1319,7 +1318,7 @@ i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj, if (ret) return ret; - return i915_gem_object_wait_rendering__tail(obj, ring); + return i915_gem_object_wait_rendering__tail(obj); } /* A nonblocking variant of the above wait. This is a highly dangerous routine @@ -1359,7 +1358,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj, if (ret) return ret; - return i915_gem_object_wait_rendering__tail(obj, ring); + return i915_gem_object_wait_rendering__tail(obj); } /** -- cgit v0.10.2 From 26ff2762108010976e2785c605cdc7ac6882da11 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Tue, 28 Oct 2014 15:10:12 +0200 Subject: drm/i915: Add kerneldoc for intel_pipe_update_{start, end} Note that a later patch will use these functions in some other file and drop the static. Hence the kerneldoc looks appropriate. Signed-off-by: Ander Conselvan de Oliveira Reviewed-by: Paulo Zanoni [danvet: Add comment that the functions will become non-static shortly.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 8c70c54..c692ade 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -60,6 +60,22 @@ static int usecs_to_scanlines(const struct drm_display_mode *mode, int usecs) return DIV_ROUND_UP(usecs * mode->crtc_clock, 1000 * mode->crtc_htotal); } +/** + * intel_pipe_update_start() - start update of a set of display registers + * @crtc: the crtc of which the registers are going to be updated + * @start_vbl_count: vblank counter return pointer used for error checking + * + * Mark the start of an update to pipe registers that should be updated + * atomically regarding vblank. If the next vblank will happens within + * the next 100 us, this function waits until the vblank passes. + * + * After a successful call to this function, interrupts will be disabled + * until a subsequent call to intel_pipe_update_end(). That is done to + * avoid random delays. The value written to @start_vbl_count should be + * supplied to intel_pipe_update_end() for error checking. + * + * Return: true if the call was successful + */ static bool intel_pipe_update_start(struct intel_crtc *crtc, uint32_t *start_vbl_count) { struct drm_device *dev = crtc->base.dev; @@ -126,6 +142,15 @@ static bool intel_pipe_update_start(struct intel_crtc *crtc, uint32_t *start_vbl return true; } +/** + * intel_pipe_update_end() - end update of a set of display registers + * @crtc: the crtc of which the registers were updated + * @start_vbl_count: start vblank counter (used for error checking) + * + * Mark the end of an update started with intel_pipe_update_start(). This + * re-enables interrupts and verifies the update was actually completed + * before a vblank using the value of @start_vbl_count. + */ static void intel_pipe_update_end(struct intel_crtc *crtc, u32 start_vbl_count) { struct drm_device *dev = crtc->base.dev; -- cgit v0.10.2 From 0594a3d9c0072f198462e8dbc56cb64b7d872d90 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Tue, 28 Oct 2014 15:10:13 +0200 Subject: drm/i915: Remove modeset lock check from intel_pipe_update_start() A follow up patch will call this funcion from a work context for the mmio flip, in which case we cannot acquire the modeset locks. That's not a problem though, since the check is there to protect vblank and the mode, but the code that changes that waits for pending flips first. Signed-off-by: Ander Conselvan de Oliveira Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index c692ade..0068fef 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -86,8 +86,6 @@ static bool intel_pipe_update_start(struct intel_crtc *crtc, uint32_t *start_vbl wait_queue_head_t *wq = drm_crtc_vblank_waitqueue(&crtc->base); DEFINE_WAIT(wait); - WARN_ON(!drm_modeset_is_locked(&crtc->base.mutex)); - vblank_start = mode->crtc_vblank_start; if (mode->flags & DRM_MODE_FLAG_INTERLACE) vblank_start = DIV_ROUND_UP(vblank_start, 2); -- cgit v0.10.2 From 9362c7c576d39ebca81f82e931ab109c065d42be Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Tue, 28 Oct 2014 15:10:14 +0200 Subject: drm/i915: Use vblank evade mechanism in mmio_flip Currently we program just DPSCNTR and DSPSTRIDE directly from the ring interrupt handler, which is fine since the hardware guarantees that those are update atomically. When we have atomic page flips we'll want to be able to update also the offset registers, and then we need to use the vblank evade mechanism to guarantee atomicity. Since that mechanism introduces a wait, we need to do the actual register write from a work when it is triggered by the ring interrupt. v2: Explain the need for mmio_flip.work in the commit message (Paulo) Initialize the mmio_flip work in intel_crtc_init() (Paulo) Prevent new flips the previous flip work finishes (Paulo) Don't acquire modeset locks for mmio flip work Note: Paulo had reservations about the work item leaking over a plane disable. But insofar as we do lack these checks that issue is already present with the existing code. Signed-off-by: Ander Conselvan de Oliveira Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 44060fb..84f4f05 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9401,11 +9401,15 @@ static void intel_do_mmio_flip(struct intel_crtc *intel_crtc) struct intel_framebuffer *intel_fb = to_intel_framebuffer(intel_crtc->base.primary->fb); struct drm_i915_gem_object *obj = intel_fb->obj; + bool atomic_update; + u32 start_vbl_count; u32 dspcntr; u32 reg; intel_mark_page_flip_active(intel_crtc); + atomic_update = intel_pipe_update_start(intel_crtc, &start_vbl_count); + reg = DSPCNTR(intel_crtc->plane); dspcntr = I915_READ(reg); @@ -9419,6 +9423,21 @@ static void intel_do_mmio_flip(struct intel_crtc *intel_crtc) I915_WRITE(DSPSURF(intel_crtc->plane), intel_crtc->unpin_work->gtt_offset); POSTING_READ(DSPSURF(intel_crtc->plane)); + + if (atomic_update) + intel_pipe_update_end(intel_crtc, start_vbl_count); + + spin_lock_irq(&dev_priv->mmio_flip_lock); + intel_crtc->mmio_flip.status = INTEL_MMIO_FLIP_IDLE; + spin_unlock_irq(&dev_priv->mmio_flip_lock); +} + +static void intel_mmio_flip_work_func(struct work_struct *work) +{ + struct intel_crtc *intel_crtc = + container_of(work, struct intel_crtc, mmio_flip.work); + + intel_do_mmio_flip(intel_crtc); } static int intel_postpone_flip(struct drm_i915_gem_object *obj) @@ -9461,15 +9480,15 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring) struct intel_mmio_flip *mmio_flip; mmio_flip = &intel_crtc->mmio_flip; - if (mmio_flip->seqno == 0) + if (mmio_flip->status != INTEL_MMIO_FLIP_WAIT_RING) continue; if (ring->id != mmio_flip->ring_id) continue; if (i915_seqno_passed(seqno, mmio_flip->seqno)) { - intel_do_mmio_flip(intel_crtc); - mmio_flip->seqno = 0; + schedule_work(&intel_crtc->mmio_flip.work); + mmio_flip->status = INTEL_MMIO_FLIP_WORK_SCHEDULED; ring->irq_put(ring); } } @@ -9487,7 +9506,7 @@ static int intel_queue_mmio_flip(struct drm_device *dev, struct intel_crtc *intel_crtc = to_intel_crtc(crtc); int ret; - if (WARN_ON(intel_crtc->mmio_flip.seqno)) + if (WARN_ON(intel_crtc->mmio_flip.status != INTEL_MMIO_FLIP_IDLE)) return -EBUSY; ret = intel_postpone_flip(obj); @@ -9499,6 +9518,7 @@ static int intel_queue_mmio_flip(struct drm_device *dev, } spin_lock_irq(&dev_priv->mmio_flip_lock); + intel_crtc->mmio_flip.status = INTEL_MMIO_FLIP_WAIT_RING; intel_crtc->mmio_flip.seqno = obj->last_write_seqno; intel_crtc->mmio_flip.ring_id = obj->ring->id; spin_unlock_irq(&dev_priv->mmio_flip_lock); @@ -11983,6 +12003,8 @@ static void intel_crtc_init(struct drm_device *dev, int pipe) dev_priv->plane_to_crtc_mapping[intel_crtc->plane] = &intel_crtc->base; dev_priv->pipe_to_crtc_mapping[intel_crtc->pipe] = &intel_crtc->base; + INIT_WORK(&intel_crtc->mmio_flip.work, intel_mmio_flip_work_func); + drm_crtc_helper_add(&intel_crtc->base, &intel_helper_funcs); WARN_ON(drm_crtc_index(&intel_crtc->base) != intel_crtc->pipe); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index ae75b99..c8ddde6 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -399,9 +399,17 @@ struct intel_pipe_wm { bool sprites_scaled; }; +enum intel_mmio_flip_status { + INTEL_MMIO_FLIP_IDLE = 0, + INTEL_MMIO_FLIP_WAIT_RING, + INTEL_MMIO_FLIP_WORK_SCHEDULED, +}; + struct intel_mmio_flip { u32 seqno; u32 ring_id; + enum intel_mmio_flip_status status; + struct work_struct work; }; struct intel_crtc { @@ -1167,7 +1175,9 @@ int intel_sprite_set_colorkey(struct drm_device *dev, void *data, struct drm_file *file_priv); int intel_sprite_get_colorkey(struct drm_device *dev, void *data, struct drm_file *file_priv); - +bool intel_pipe_update_start(struct intel_crtc *crtc, + uint32_t *start_vbl_count); +void intel_pipe_update_end(struct intel_crtc *crtc, u32 start_vbl_count); /* intel_tv.c */ void intel_tv_init(struct drm_device *dev); diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 0068fef..6407655 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -76,7 +76,7 @@ static int usecs_to_scanlines(const struct drm_display_mode *mode, int usecs) * * Return: true if the call was successful */ -static bool intel_pipe_update_start(struct intel_crtc *crtc, uint32_t *start_vbl_count) +bool intel_pipe_update_start(struct intel_crtc *crtc, uint32_t *start_vbl_count) { struct drm_device *dev = crtc->base.dev; const struct drm_display_mode *mode = &crtc->config.adjusted_mode; @@ -149,7 +149,7 @@ static bool intel_pipe_update_start(struct intel_crtc *crtc, uint32_t *start_vbl * re-enables interrupts and verifies the update was actually completed * before a vblank using the value of @start_vbl_count. */ -static void intel_pipe_update_end(struct intel_crtc *crtc, u32 start_vbl_count) +void intel_pipe_update_end(struct intel_crtc *crtc, u32 start_vbl_count) { struct drm_device *dev = crtc->base.dev; enum pipe pipe = crtc->pipe; -- cgit v0.10.2 From ef79e17ccef32599f3cec3842f81d70117181291 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 31 Oct 2014 13:53:52 +0000 Subject: drm/i915: Only mark as map-and-fenceable when bound into the GGTT We use the obj->map_and_fenceable hint for when we already have a valid mapping of this object in the aperture. This hint can only apply to the GGTT and not to the aliasing-ppGTT. One user of the hint is execbuffer relocation, which began to fail when it tried to follow the hint and perform the relocate through the non-existent GGTT mapping. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85671 Signed-off-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index c2456c2..0c6ef66 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3476,20 +3476,6 @@ search_free: list_move_tail(&obj->global_list, &dev_priv->mm.bound_list); list_add_tail(&vma->mm_list, &vm->inactive_list); - if (i915_is_ggtt(vm)) { - bool mappable, fenceable; - - fenceable = (vma->node.size == fence_size && - (vma->node.start & (fence_alignment - 1)) == 0); - - mappable = (vma->node.start + obj->base.size <= - dev_priv->gtt.mappable_end); - - obj->map_and_fenceable = mappable && fenceable; - } - - WARN_ON(flags & PIN_MAPPABLE && !obj->map_and_fenceable); - trace_i915_vma_bind(vma, flags); vma->bind_vma(vma, obj->cache_level, flags & (PIN_MAPPABLE | PIN_GLOBAL) ? GLOBAL_BIND : 0); @@ -4061,6 +4047,7 @@ i915_gem_object_pin(struct drm_i915_gem_object *obj, { struct drm_i915_private *dev_priv = obj->base.dev->dev_private; struct i915_vma *vma; + unsigned bound; int ret; if (WARN_ON(vm == &dev_priv->mm.aliasing_ppgtt->base)) @@ -4090,6 +4077,7 @@ i915_gem_object_pin(struct drm_i915_gem_object *obj, } } + bound = vma ? vma->bound : 0; if (vma == NULL || !drm_mm_node_allocated(&vma->node)) { vma = i915_gem_object_bind_to_vm(obj, vm, alignment, flags); if (IS_ERR(vma)) @@ -4099,6 +4087,29 @@ i915_gem_object_pin(struct drm_i915_gem_object *obj, if (flags & PIN_GLOBAL && !(vma->bound & GLOBAL_BIND)) vma->bind_vma(vma, obj->cache_level, GLOBAL_BIND); + if ((bound ^ vma->bound) & GLOBAL_BIND) { + bool mappable, fenceable; + u32 fence_size, fence_alignment; + + fence_size = i915_gem_get_gtt_size(obj->base.dev, + obj->base.size, + obj->tiling_mode); + fence_alignment = i915_gem_get_gtt_alignment(obj->base.dev, + obj->base.size, + obj->tiling_mode, + true); + + fenceable = (vma->node.size == fence_size && + (vma->node.start & (fence_alignment - 1)) == 0); + + mappable = (vma->node.start + obj->base.size <= + dev_priv->gtt.mappable_end); + + obj->map_and_fenceable = mappable && fenceable; + } + + WARN_ON(flags & PIN_MAPPABLE && !obj->map_and_fenceable); + vma->pin_count++; if (flags & PIN_MAPPABLE) obj->pin_mappable |= true; -- cgit v0.10.2 From c826c4493851913e57dd47a7fbfe07156ac9f91d Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 31 Oct 2014 13:53:53 +0000 Subject: drm/i915: Request PIN_GLOBAL when pinning a vma for GTT relocations Always require PIN_GLOBAL when we want a mappable offset (PIN_MAPPABLE). This causes the pin to fixup the global binding in cases were the vma was already bound (and due to the proceeding bug, we considered it to be already mappable). References: https://bugs.freedesktop.org/show_bug.cgi?id=85671 Signed-off-by: Chris Wilson [danvet: Add WARN_ON to check that PIN_MAP implies PIN_GLOBAL as discussed on irc.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 0c6ef66..1de94cc 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3478,7 +3478,7 @@ search_free: trace_i915_vma_bind(vma, flags); vma->bind_vma(vma, obj->cache_level, - flags & (PIN_MAPPABLE | PIN_GLOBAL) ? GLOBAL_BIND : 0); + flags & PIN_GLOBAL ? GLOBAL_BIND : 0); return vma; @@ -4056,6 +4056,9 @@ i915_gem_object_pin(struct drm_i915_gem_object *obj, if (WARN_ON(flags & (PIN_GLOBAL | PIN_MAPPABLE) && !i915_is_ggtt(vm))) return -EINVAL; + if (WARN_ON((flags & (PIN_MAPPABLE | PIN_GLOBAL)) == PIN_MAPPABLE)) + return -EINVAL; + vma = i915_gem_obj_to_vma(obj, vm); if (vma) { if (WARN_ON(vma->pin_count == DRM_I915_GEM_OBJECT_MAX_PIN_COUNT)) diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 4b7f5c1..e1ed85a 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -528,7 +528,7 @@ i915_gem_execbuffer_reserve_vma(struct i915_vma *vma, flags = 0; if (entry->flags & __EXEC_OBJECT_NEEDS_MAP) - flags |= PIN_MAPPABLE; + flags |= PIN_GLOBAL | PIN_MAPPABLE; if (entry->flags & EXEC_OBJECT_NEEDS_GTT) flags |= PIN_GLOBAL; if (entry->flags & __EXEC_OBJECT_NEEDS_BIAS) -- cgit v0.10.2 From 70f2f5c70440feff01d9ba7c8b4432eb72bd69eb Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 24 Oct 2014 12:11:11 +0100 Subject: drm/i915: Report the actual swizzling back to userspace Userspace cares about whether or not swizzling depends on the page address for its direct access into bound objects. Extend the get_tiling ioctl to report the physical swizzling value in addition to the logical swizzling value so that userspace can accurately determine when it is possible for manual detiling. Signed-off-by: Chris Wilson Cc: Akash Goel Cc: Daniel Vetter Testcase: igt/gem_tiled_wc Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index d1e7a3e..749ab48 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -458,6 +458,7 @@ i915_gem_get_tiling(struct drm_device *dev, void *data, } /* Hide bit 17 from the user -- see comment in i915_gem_set_tiling */ + args->phys_swizzle_mode = args->swizzle_mode; if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_17) args->swizzle_mode = I915_BIT_6_SWIZZLE_9; if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_10_17) diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index ff57f07..2ec0efc 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -876,6 +876,12 @@ struct drm_i915_gem_get_tiling { * mmap mapping. */ __u32 swizzle_mode; + + /** + * Returned address bit 6 swizzling required for CPU access through + * mmap mapping whilst bound. + */ + __u32 phys_swizzle_mode; }; struct drm_i915_gem_get_aperture { -- cgit v0.10.2 From c46f111f5171b63572fae3f79bcfe578bc96c603 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:52 +0200 Subject: drm/i915: clean up and clarify audio related register defines Make audio related register defines conform to existing style: Add _MASK where relevant, indent the defines for register contents, don't indent the defines for register addresses, prefix pipe specific register address defines with underscores, drop self explanatory comments. No functional changes. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 2a4bdd5..cfc6554 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -6011,57 +6011,58 @@ enum punit_power_well { #define GEN8_CENTROID_PIXEL_OPT_DIS (1<<8) #define GEN8_SAMPLER_POWER_BYPASS_DIS (1<<1) +/* Audio */ #define G4X_AUD_VID_DID (dev_priv->info.display_mmio_offset + 0x62020) -#define INTEL_AUDIO_DEVCL 0x808629FB -#define INTEL_AUDIO_DEVBLC 0x80862801 -#define INTEL_AUDIO_DEVCTG 0x80862802 +#define INTEL_AUDIO_DEVCL 0x808629FB +#define INTEL_AUDIO_DEVBLC 0x80862801 +#define INTEL_AUDIO_DEVCTG 0x80862802 #define G4X_AUD_CNTL_ST 0x620B4 -#define G4X_ELDV_DEVCL_DEVBLC (1 << 13) -#define G4X_ELDV_DEVCTG (1 << 14) -#define G4X_ELD_ADDR (0xf << 5) -#define G4X_ELD_ACK (1 << 4) +#define G4X_ELDV_DEVCL_DEVBLC (1 << 13) +#define G4X_ELDV_DEVCTG (1 << 14) +#define G4X_ELD_ADDR_MASK (0xf << 5) +#define G4X_ELD_ACK (1 << 4) #define G4X_HDMIW_HDMIEDID 0x6210C -#define IBX_HDMIW_HDMIEDID_A 0xE2050 -#define IBX_HDMIW_HDMIEDID_B 0xE2150 +#define _IBX_HDMIW_HDMIEDID_A 0xE2050 +#define _IBX_HDMIW_HDMIEDID_B 0xE2150 #define IBX_HDMIW_HDMIEDID(pipe) _PIPE(pipe, \ - IBX_HDMIW_HDMIEDID_A, \ - IBX_HDMIW_HDMIEDID_B) -#define IBX_AUD_CNTL_ST_A 0xE20B4 -#define IBX_AUD_CNTL_ST_B 0xE21B4 + _IBX_HDMIW_HDMIEDID_A, \ + _IBX_HDMIW_HDMIEDID_B) +#define _IBX_AUD_CNTL_ST_A 0xE20B4 +#define _IBX_AUD_CNTL_ST_B 0xE21B4 #define IBX_AUD_CNTL_ST(pipe) _PIPE(pipe, \ - IBX_AUD_CNTL_ST_A, \ - IBX_AUD_CNTL_ST_B) -#define IBX_ELD_BUFFER_SIZE (0x1f << 10) -#define IBX_ELD_ADDRESS (0x1f << 5) -#define IBX_ELD_ACK (1 << 4) + _IBX_AUD_CNTL_ST_A, \ + _IBX_AUD_CNTL_ST_B) +#define IBX_ELD_BUFFER_SIZE_MASK (0x1f << 10) +#define IBX_ELD_ADDRESS_MASK (0x1f << 5) +#define IBX_ELD_ACK (1 << 4) #define IBX_AUD_CNTL_ST2 0xE20C0 -#define IBX_ELD_VALIDB (1 << 0) -#define IBX_CP_READYB (1 << 1) +#define IBX_ELD_VALIDB (1 << 0) +#define IBX_CP_READYB (1 << 1) -#define CPT_HDMIW_HDMIEDID_A 0xE5050 -#define CPT_HDMIW_HDMIEDID_B 0xE5150 +#define _CPT_HDMIW_HDMIEDID_A 0xE5050 +#define _CPT_HDMIW_HDMIEDID_B 0xE5150 #define CPT_HDMIW_HDMIEDID(pipe) _PIPE(pipe, \ - CPT_HDMIW_HDMIEDID_A, \ - CPT_HDMIW_HDMIEDID_B) -#define CPT_AUD_CNTL_ST_A 0xE50B4 -#define CPT_AUD_CNTL_ST_B 0xE51B4 + _CPT_HDMIW_HDMIEDID_A, \ + _CPT_HDMIW_HDMIEDID_B) +#define _CPT_AUD_CNTL_ST_A 0xE50B4 +#define _CPT_AUD_CNTL_ST_B 0xE51B4 #define CPT_AUD_CNTL_ST(pipe) _PIPE(pipe, \ - CPT_AUD_CNTL_ST_A, \ - CPT_AUD_CNTL_ST_B) + _CPT_AUD_CNTL_ST_A, \ + _CPT_AUD_CNTL_ST_B) #define CPT_AUD_CNTRL_ST2 0xE50C0 -#define VLV_HDMIW_HDMIEDID_A (VLV_DISPLAY_BASE + 0x62050) -#define VLV_HDMIW_HDMIEDID_B (VLV_DISPLAY_BASE + 0x62150) +#define _VLV_HDMIW_HDMIEDID_A (VLV_DISPLAY_BASE + 0x62050) +#define _VLV_HDMIW_HDMIEDID_B (VLV_DISPLAY_BASE + 0x62150) #define VLV_HDMIW_HDMIEDID(pipe) _PIPE(pipe, \ - VLV_HDMIW_HDMIEDID_A, \ - VLV_HDMIW_HDMIEDID_B) -#define VLV_AUD_CNTL_ST_A (VLV_DISPLAY_BASE + 0x620B4) -#define VLV_AUD_CNTL_ST_B (VLV_DISPLAY_BASE + 0x621B4) + _VLV_HDMIW_HDMIEDID_A, \ + _VLV_HDMIW_HDMIEDID_B) +#define _VLV_AUD_CNTL_ST_A (VLV_DISPLAY_BASE + 0x620B4) +#define _VLV_AUD_CNTL_ST_B (VLV_DISPLAY_BASE + 0x621B4) #define VLV_AUD_CNTL_ST(pipe) _PIPE(pipe, \ - VLV_AUD_CNTL_ST_A, \ - VLV_AUD_CNTL_ST_B) + _VLV_AUD_CNTL_ST_A, \ + _VLV_AUD_CNTL_ST_B) #define VLV_AUD_CNTL_ST2 (VLV_DISPLAY_BASE + 0x620C0) /* These are the 4 32-bit write offset registers for each stream @@ -6070,28 +6071,28 @@ enum punit_power_well { */ #define GEN7_SO_WRITE_OFFSET(n) (0x5280 + (n) * 4) -#define IBX_AUD_CONFIG_A 0xe2000 -#define IBX_AUD_CONFIG_B 0xe2100 +#define _IBX_AUD_CONFIG_A 0xe2000 +#define _IBX_AUD_CONFIG_B 0xe2100 #define IBX_AUD_CFG(pipe) _PIPE(pipe, \ - IBX_AUD_CONFIG_A, \ - IBX_AUD_CONFIG_B) -#define CPT_AUD_CONFIG_A 0xe5000 -#define CPT_AUD_CONFIG_B 0xe5100 + _IBX_AUD_CONFIG_A, \ + _IBX_AUD_CONFIG_B) +#define _CPT_AUD_CONFIG_A 0xe5000 +#define _CPT_AUD_CONFIG_B 0xe5100 #define CPT_AUD_CFG(pipe) _PIPE(pipe, \ - CPT_AUD_CONFIG_A, \ - CPT_AUD_CONFIG_B) -#define VLV_AUD_CONFIG_A (VLV_DISPLAY_BASE + 0x62000) -#define VLV_AUD_CONFIG_B (VLV_DISPLAY_BASE + 0x62100) + _CPT_AUD_CONFIG_A, \ + _CPT_AUD_CONFIG_B) +#define _VLV_AUD_CONFIG_A (VLV_DISPLAY_BASE + 0x62000) +#define _VLV_AUD_CONFIG_B (VLV_DISPLAY_BASE + 0x62100) #define VLV_AUD_CFG(pipe) _PIPE(pipe, \ - VLV_AUD_CONFIG_A, \ - VLV_AUD_CONFIG_B) + _VLV_AUD_CONFIG_A, \ + _VLV_AUD_CONFIG_B) #define AUD_CONFIG_N_VALUE_INDEX (1 << 29) #define AUD_CONFIG_N_PROG_ENABLE (1 << 28) #define AUD_CONFIG_UPPER_N_SHIFT 20 -#define AUD_CONFIG_UPPER_N_VALUE (0xff << 20) +#define AUD_CONFIG_UPPER_N_MASK (0xff << 20) #define AUD_CONFIG_LOWER_N_SHIFT 4 -#define AUD_CONFIG_LOWER_N_VALUE (0xfff << 4) +#define AUD_CONFIG_LOWER_N_MASK (0xfff << 4) #define AUD_CONFIG_PIXEL_CLOCK_HDMI_SHIFT 16 #define AUD_CONFIG_PIXEL_CLOCK_HDMI_MASK (0xf << 16) #define AUD_CONFIG_PIXEL_CLOCK_HDMI_25175 (0 << 16) @@ -6107,40 +6108,40 @@ enum punit_power_well { #define AUD_CONFIG_DISABLE_NCTS (1 << 3) /* HSW Audio */ -#define HSW_AUD_CONFIG_A 0x65000 /* Audio Configuration Transcoder A */ -#define HSW_AUD_CONFIG_B 0x65100 /* Audio Configuration Transcoder B */ -#define HSW_AUD_CFG(pipe) _PIPE(pipe, \ - HSW_AUD_CONFIG_A, \ - HSW_AUD_CONFIG_B) - -#define HSW_AUD_MISC_CTRL_A 0x65010 /* Audio Misc Control Convert 1 */ -#define HSW_AUD_MISC_CTRL_B 0x65110 /* Audio Misc Control Convert 2 */ -#define HSW_AUD_MISC_CTRL(pipe) _PIPE(pipe, \ - HSW_AUD_MISC_CTRL_A, \ - HSW_AUD_MISC_CTRL_B) - -#define HSW_AUD_DIP_ELD_CTRL_ST_A 0x650b4 /* Audio DIP and ELD Control State Transcoder A */ -#define HSW_AUD_DIP_ELD_CTRL_ST_B 0x651b4 /* Audio DIP and ELD Control State Transcoder B */ -#define HSW_AUD_DIP_ELD_CTRL(pipe) _PIPE(pipe, \ - HSW_AUD_DIP_ELD_CTRL_ST_A, \ - HSW_AUD_DIP_ELD_CTRL_ST_B) +#define _HSW_AUD_CONFIG_A 0x65000 +#define _HSW_AUD_CONFIG_B 0x65100 +#define HSW_AUD_CFG(pipe) _PIPE(pipe, \ + _HSW_AUD_CONFIG_A, \ + _HSW_AUD_CONFIG_B) + +#define _HSW_AUD_MISC_CTRL_A 0x65010 +#define _HSW_AUD_MISC_CTRL_B 0x65110 +#define HSW_AUD_MISC_CTRL(pipe) _PIPE(pipe, \ + _HSW_AUD_MISC_CTRL_A, \ + _HSW_AUD_MISC_CTRL_B) + +#define _HSW_AUD_DIP_ELD_CTRL_ST_A 0x650b4 +#define _HSW_AUD_DIP_ELD_CTRL_ST_B 0x651b4 +#define HSW_AUD_DIP_ELD_CTRL(pipe) _PIPE(pipe, \ + _HSW_AUD_DIP_ELD_CTRL_ST_A, \ + _HSW_AUD_DIP_ELD_CTRL_ST_B) /* Audio Digital Converter */ -#define HSW_AUD_DIG_CNVT_1 0x65080 /* Audio Converter 1 */ -#define HSW_AUD_DIG_CNVT_2 0x65180 /* Audio Converter 1 */ -#define AUD_DIG_CNVT(pipe) _PIPE(pipe, \ - HSW_AUD_DIG_CNVT_1, \ - HSW_AUD_DIG_CNVT_2) -#define DIP_PORT_SEL_MASK 0x3 - -#define HSW_AUD_EDID_DATA_A 0x65050 -#define HSW_AUD_EDID_DATA_B 0x65150 -#define HSW_AUD_EDID_DATA(pipe) _PIPE(pipe, \ - HSW_AUD_EDID_DATA_A, \ - HSW_AUD_EDID_DATA_B) - -#define HSW_AUD_PIPE_CONV_CFG 0x6507c /* Audio pipe and converter configs */ -#define HSW_AUD_PIN_ELD_CP_VLD 0x650c0 /* Audio ELD and CP Ready Status */ +#define _HSW_AUD_DIG_CNVT_1 0x65080 +#define _HSW_AUD_DIG_CNVT_2 0x65180 +#define AUD_DIG_CNVT(pipe) _PIPE(pipe, \ + _HSW_AUD_DIG_CNVT_1, \ + _HSW_AUD_DIG_CNVT_2) +#define DIP_PORT_SEL_MASK 0x3 + +#define _HSW_AUD_EDID_DATA_A 0x65050 +#define _HSW_AUD_EDID_DATA_B 0x65150 +#define HSW_AUD_EDID_DATA(pipe) _PIPE(pipe, \ + _HSW_AUD_EDID_DATA_A, \ + _HSW_AUD_EDID_DATA_B) + +#define HSW_AUD_PIPE_CONV_CFG 0x6507c +#define HSW_AUD_PIN_ELD_CP_VLD 0x650c0 #define AUDIO_INACTIVE_C (1<<11) #define AUDIO_INACTIVE_B (1<<7) #define AUDIO_INACTIVE_A (1<<3) diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 7580b38..821514c 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -111,12 +111,12 @@ static void g4x_audio_codec_enable(struct drm_connector *connector, if (intel_eld_uptodate(connector, G4X_AUD_CNTL_ST, eldv, - G4X_AUD_CNTL_ST, G4X_ELD_ADDR, + G4X_AUD_CNTL_ST, G4X_ELD_ADDR_MASK, G4X_HDMIW_HDMIEDID)) return; tmp = I915_READ(G4X_AUD_CNTL_ST); - tmp &= ~(eldv | G4X_ELD_ADDR); + tmp &= ~(eldv | G4X_ELD_ADDR_MASK); len = (tmp >> 9) & 0x1f; /* ELD buffer size */ I915_WRITE(G4X_AUD_CNTL_ST, tmp); @@ -193,7 +193,7 @@ static void hsw_audio_codec_enable(struct drm_connector *connector, if (intel_eld_uptodate(connector, aud_cntrl_st2, eldv, - aud_cntl_st, IBX_ELD_ADDRESS, + aud_cntl_st, IBX_ELD_ADDRESS_MASK, hdmiw_hdmiedid)) return; @@ -202,7 +202,7 @@ static void hsw_audio_codec_enable(struct drm_connector *connector, I915_WRITE(aud_cntrl_st2, tmp); tmp = I915_READ(aud_cntl_st); - tmp &= ~IBX_ELD_ADDRESS; + tmp &= ~IBX_ELD_ADDRESS_MASK; I915_WRITE(aud_cntl_st, tmp); port = (tmp >> 29) & DIP_PORT_SEL_MASK; /* DIP_Port_Select, 0x1 = PortB */ DRM_DEBUG_DRIVER("port num:%d\n", port); @@ -287,7 +287,7 @@ static void ilk_audio_codec_enable(struct drm_connector *connector, if (intel_eld_uptodate(connector, aud_cntrl_st2, eldv, - aud_cntl_st, IBX_ELD_ADDRESS, + aud_cntl_st, IBX_ELD_ADDRESS_MASK, hdmiw_hdmiedid)) return; @@ -296,7 +296,7 @@ static void ilk_audio_codec_enable(struct drm_connector *connector, I915_WRITE(aud_cntrl_st2, tmp); tmp = I915_READ(aud_cntl_st); - tmp &= ~IBX_ELD_ADDRESS; + tmp &= ~IBX_ELD_ADDRESS_MASK; I915_WRITE(aud_cntl_st, tmp); len = min_t(int, eld[2], 21); /* 84 bytes of hw ELD buffer */ -- cgit v0.10.2 From 5fad84a7530f8e7664cdc6f490cb90653fed1266 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Tue, 4 Nov 2014 10:30:23 +0200 Subject: drm/i915: rewrite hsw/bdw audio codec enable/disable sequences There's some serious confusion regarding ELD valid bit that gets set and cleared back and forth etc. Rewrite it all based on the documented audio codec enable/disable sequences. v3: replace vblank wait with a comment v4: expand the comment on what should be done with the vblank wait Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 821514c..4dcfc46 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -132,14 +132,26 @@ static void g4x_audio_codec_enable(struct drm_connector *connector, static void hsw_audio_codec_disable(struct intel_encoder *encoder) { - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_crtc *crtc = encoder->base.crtc; - enum pipe pipe = to_intel_crtc(crtc)->pipe; + struct drm_i915_private *dev_priv = encoder->base.dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); + enum pipe pipe = intel_crtc->pipe; uint32_t tmp; + DRM_DEBUG_KMS("Disable audio codec on pipe %c\n", pipe_name(pipe)); + + /* Disable timestamps */ + tmp = I915_READ(HSW_AUD_CFG(pipe)); + tmp &= ~AUD_CONFIG_N_VALUE_INDEX; + tmp |= AUD_CONFIG_N_PROG_ENABLE; + tmp &= ~AUD_CONFIG_UPPER_N_MASK; + tmp &= ~AUD_CONFIG_LOWER_N_MASK; + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) + tmp |= AUD_CONFIG_N_VALUE_INDEX; + I915_WRITE(HSW_AUD_CFG(pipe), tmp); + + /* Invalidate ELD */ tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); - tmp &= ~((AUDIO_OUTPUT_ENABLE_A | AUDIO_ELD_VALID_A) << (pipe * 4)); + tmp &= ~(AUDIO_ELD_VALID_A << (pipe * 4)); I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); } @@ -149,77 +161,52 @@ static void hsw_audio_codec_enable(struct drm_connector *connector, { struct drm_i915_private *dev_priv = connector->dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); - uint8_t *eld = connector->eld; - uint32_t eldv; + enum pipe pipe = intel_crtc->pipe; + const uint8_t *eld = connector->eld; uint32_t tmp; int len, i; - enum pipe pipe = intel_crtc->pipe; - enum port port; - int hdmiw_hdmiedid = HSW_AUD_EDID_DATA(pipe); - int aud_cntl_st = HSW_AUD_DIP_ELD_CTRL(pipe); - int aud_config = HSW_AUD_CFG(pipe); - int aud_cntrl_st2 = HSW_AUD_PIN_ELD_CP_VLD; - /* Audio output enable */ - DRM_DEBUG_DRIVER("HDMI audio: enable codec\n"); - tmp = I915_READ(aud_cntrl_st2); - tmp |= (AUDIO_OUTPUT_ENABLE_A << (pipe * 4)); - I915_WRITE(aud_cntrl_st2, tmp); - POSTING_READ(aud_cntrl_st2); - - /* Set ELD valid state */ - tmp = I915_READ(aud_cntrl_st2); - DRM_DEBUG_DRIVER("HDMI audio: pin eld vld status=0x%08x\n", tmp); - tmp |= (AUDIO_ELD_VALID_A << (pipe * 4)); - I915_WRITE(aud_cntrl_st2, tmp); - tmp = I915_READ(aud_cntrl_st2); - DRM_DEBUG_DRIVER("HDMI audio: eld vld status=0x%08x\n", tmp); - - /* Enable HDMI mode */ - tmp = I915_READ(aud_config); - DRM_DEBUG_DRIVER("HDMI audio: audio conf: 0x%08x\n", tmp); - /* clear N_programing_enable and N_value_index */ - tmp &= ~(AUD_CONFIG_N_VALUE_INDEX | AUD_CONFIG_N_PROG_ENABLE); - I915_WRITE(aud_config, tmp); + DRM_DEBUG_KMS("Enable audio codec on pipe %c, %u bytes ELD\n", + pipe_name(pipe), eld[2]); - DRM_DEBUG_DRIVER("ELD on pipe %c\n", pipe_name(pipe)); - - eldv = AUDIO_ELD_VALID_A << (pipe * 4); - - if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) - I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ - else - I915_WRITE(aud_config, audio_config_hdmi_pixel_clock(mode)); - - if (intel_eld_uptodate(connector, - aud_cntrl_st2, eldv, - aud_cntl_st, IBX_ELD_ADDRESS_MASK, - hdmiw_hdmiedid)) - return; + /* Enable audio presence detect, invalidate ELD */ + tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); + tmp |= AUDIO_OUTPUT_ENABLE_A << (pipe * 4); + tmp &= ~(AUDIO_ELD_VALID_A << (pipe * 4)); + I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); - tmp = I915_READ(aud_cntrl_st2); - tmp &= ~eldv; - I915_WRITE(aud_cntrl_st2, tmp); + /* + * FIXME: We're supposed to wait for vblank here, but we have vblanks + * disabled during the mode set. The proper fix would be to push the + * rest of the setup into a vblank work item, queued here, but the + * infrastructure is not there yet. + */ - tmp = I915_READ(aud_cntl_st); + /* Reset ELD write address */ + tmp = I915_READ(HSW_AUD_DIP_ELD_CTRL(pipe)); tmp &= ~IBX_ELD_ADDRESS_MASK; - I915_WRITE(aud_cntl_st, tmp); - port = (tmp >> 29) & DIP_PORT_SEL_MASK; /* DIP_Port_Select, 0x1 = PortB */ - DRM_DEBUG_DRIVER("port num:%d\n", port); + I915_WRITE(HSW_AUD_DIP_ELD_CTRL(pipe), tmp); - len = min_t(int, eld[2], 21); /* 84 bytes of hw ELD buffer */ - DRM_DEBUG_DRIVER("ELD size %d\n", len); + /* Up to 84 bytes of hw ELD buffer */ + len = min_t(int, eld[2], 21); for (i = 0; i < len; i++) - I915_WRITE(hdmiw_hdmiedid, *((uint32_t *)eld + i)); - - tmp = I915_READ(aud_cntrl_st2); - tmp |= eldv; - I915_WRITE(aud_cntrl_st2, tmp); + I915_WRITE(HSW_AUD_EDID_DATA(pipe), *((uint32_t *)eld + i)); - /* XXX: Transitional */ + /* ELD valid */ tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); - tmp |= ((AUDIO_OUTPUT_ENABLE_A | AUDIO_ELD_VALID_A) << (pipe * 4)); + tmp |= AUDIO_ELD_VALID_A << (pipe * 4); I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); + + /* Enable timestamps */ + tmp = I915_READ(HSW_AUD_CFG(pipe)); + tmp &= ~AUD_CONFIG_N_VALUE_INDEX; + tmp &= ~AUD_CONFIG_N_PROG_ENABLE; + tmp &= ~AUD_CONFIG_PIXEL_CLOCK_HDMI_MASK; + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) + tmp |= AUD_CONFIG_N_VALUE_INDEX; + else + tmp |= audio_config_hdmi_pixel_clock(mode); + I915_WRITE(HSW_AUD_CFG(pipe), tmp); } static void ilk_audio_codec_enable(struct drm_connector *connector, -- cgit v0.10.2 From 2af30a5c40bff540f13106961662d000b1bc8cd7 Mon Sep 17 00:00:00 2001 From: Pradeep Bhat Date: Tue, 4 Nov 2014 17:06:38 +0000 Subject: drm/i915/skl: Read the Memory Latency Values for WM computation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This patch reads the memory latency values for all the 8 levels for SKL. These values are needed for the Watermark computation. v2: Incorporated the review comments from Damien on register indentation. v3: Updated the code to use the sandybridge_pcode_read for reading memory latencies for GEN9. v4: Don't put gen 9 in the middle of an ordered list of ifs (Damien) v5: take the rps.hw_lock around sandybridge_pcode_read() (Damien) v6: Use gen >= 9 in the pcode_read() function for data1. Move the defines near the gen6 ones and prefix them with PCODE. Remove unused timeout define (the pcode_read() code has a larger timeout already). Reviewed-by: Ville Syrjälä Signed-off-by: Pradeep Bhat Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 54691bc..05fcbe5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1678,6 +1678,12 @@ struct drm_i915_private { uint16_t spr_latency[5]; /* cursor */ uint16_t cur_latency[5]; + /* + * Raw watermark memory latency values + * for SKL for all 8 levels + * in 1us units. + */ + uint16_t skl_latency[8]; /* current hardware state */ struct ilk_wm_values hw; diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index cfc6554..db7d8fd 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5957,6 +5957,13 @@ enum punit_power_well { #define GEN6_PCODE_FREQ_IA_RATIO_SHIFT 8 #define GEN6_PCODE_FREQ_RING_RATIO_SHIFT 16 +#define GEN9_PCODE_DATA1 0x13812C +#define GEN9_PCODE_READ_MEM_LATENCY 0x6 +#define GEN9_MEM_LATENCY_LEVEL_MASK 0xFF +#define GEN9_MEM_LATENCY_LEVEL_1_5_SHIFT 8 +#define GEN9_MEM_LATENCY_LEVEL_2_6_SHIFT 16 +#define GEN9_MEM_LATENCY_LEVEL_3_7_SHIFT 24 + #define GEN6_GT_CORE_STATUS 0x138060 #define GEN6_CORE_CPD_STATE_MASK (7<<4) #define GEN6_RCn_MASK 7 diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..761c884 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -2271,11 +2271,56 @@ hsw_compute_linetime_wm(struct drm_device *dev, struct drm_crtc *crtc) PIPE_WM_LINETIME_TIME(linetime); } -static void intel_read_wm_latency(struct drm_device *dev, uint16_t wm[5]) +static void intel_read_wm_latency(struct drm_device *dev, uint16_t wm[8]) { struct drm_i915_private *dev_priv = dev->dev_private; - if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { + if (IS_GEN9(dev)) { + uint32_t val; + int ret; + + /* read the first set of memory latencies[0:3] */ + val = 0; /* data0 to be programmed to 0 for first set */ + mutex_lock(&dev_priv->rps.hw_lock); + ret = sandybridge_pcode_read(dev_priv, + GEN9_PCODE_READ_MEM_LATENCY, + &val); + mutex_unlock(&dev_priv->rps.hw_lock); + + if (ret) { + DRM_ERROR("SKL Mailbox read error = %d\n", ret); + return; + } + + wm[0] = val & GEN9_MEM_LATENCY_LEVEL_MASK; + wm[1] = (val >> GEN9_MEM_LATENCY_LEVEL_1_5_SHIFT) & + GEN9_MEM_LATENCY_LEVEL_MASK; + wm[2] = (val >> GEN9_MEM_LATENCY_LEVEL_2_6_SHIFT) & + GEN9_MEM_LATENCY_LEVEL_MASK; + wm[3] = (val >> GEN9_MEM_LATENCY_LEVEL_3_7_SHIFT) & + GEN9_MEM_LATENCY_LEVEL_MASK; + + /* read the second set of memory latencies[4:7] */ + val = 1; /* data0 to be programmed to 1 for second set */ + mutex_lock(&dev_priv->rps.hw_lock); + ret = sandybridge_pcode_read(dev_priv, + GEN9_PCODE_READ_MEM_LATENCY, + &val); + mutex_unlock(&dev_priv->rps.hw_lock); + if (ret) { + DRM_ERROR("SKL Mailbox read error = %d\n", ret); + return; + } + + wm[4] = val & GEN9_MEM_LATENCY_LEVEL_MASK; + wm[5] = (val >> GEN9_MEM_LATENCY_LEVEL_1_5_SHIFT) & + GEN9_MEM_LATENCY_LEVEL_MASK; + wm[6] = (val >> GEN9_MEM_LATENCY_LEVEL_2_6_SHIFT) & + GEN9_MEM_LATENCY_LEVEL_MASK; + wm[7] = (val >> GEN9_MEM_LATENCY_LEVEL_3_7_SHIFT) & + GEN9_MEM_LATENCY_LEVEL_MASK; + + } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { uint64_t sskpd = I915_READ64(MCH_SSKPD); wm[0] = (sskpd >> 56) & 0xFF; @@ -2323,7 +2368,9 @@ static void intel_fixup_cur_wm_latency(struct drm_device *dev, uint16_t wm[5]) int ilk_wm_max_level(const struct drm_device *dev) { /* how many WM levels are we expecting */ - if (IS_HASWELL(dev) || IS_BROADWELL(dev)) + if (IS_GEN9(dev)) + return 7; + else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) return 4; else if (INTEL_INFO(dev)->gen >= 6) return 3; @@ -2333,7 +2380,7 @@ int ilk_wm_max_level(const struct drm_device *dev) static void intel_print_wm_latency(struct drm_device *dev, const char *name, - const uint16_t wm[5]) + const uint16_t wm[8]) { int level, max_level = ilk_wm_max_level(dev); @@ -2346,8 +2393,13 @@ static void intel_print_wm_latency(struct drm_device *dev, continue; } - /* WM1+ latency values in 0.5us units */ - if (level > 0) + /* + * - latencies are in us on gen9. + * - before then, WM1+ latency values are in 0.5us units + */ + if (IS_GEN9(dev)) + latency *= 10; + else if (level > 0) latency *= 5; DRM_DEBUG_KMS("%s WM%d latency %u (%u.%u usec)\n", @@ -2415,6 +2467,14 @@ static void ilk_setup_wm_latency(struct drm_device *dev) snb_wm_latency_quirk(dev); } +static void skl_setup_wm_latency(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + intel_read_wm_latency(dev, dev_priv->wm.skl_latency); + intel_print_wm_latency(dev, "Gen9 Plane", dev_priv->wm.skl_latency); +} + static void ilk_compute_wm_parameters(struct drm_crtc *crtc, struct ilk_pipe_wm_parameters *p) { @@ -6127,6 +6187,8 @@ void intel_init_pm(struct drm_device *dev) /* For FIFO watermark updates */ if (IS_GEN9(dev)) { + skl_setup_wm_latency(dev); + dev_priv->display.init_clock_gating = gen9_init_clock_gating; } else if (HAS_PCH_SPLIT(dev)) { ilk_setup_wm_latency(dev); @@ -6219,6 +6281,8 @@ int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u8 mbox, u32 *val) } I915_WRITE(GEN6_PCODE_DATA, *val); + if (INTEL_INFO(dev_priv)->gen >= 9) + I915_WRITE(GEN9_PCODE_DATA1, 0); I915_WRITE(GEN6_PCODE_MAILBOX, GEN6_PCODE_READY | mbox); if (wait_for((I915_READ(GEN6_PCODE_MAILBOX) & GEN6_PCODE_READY) == 0, -- cgit v0.10.2 From fae1267df8a14140544204c65258867d72f287bb Mon Sep 17 00:00:00 2001 From: Pradeep Bhat Date: Tue, 4 Nov 2014 17:06:39 +0000 Subject: drm/i915/skl: Register definitions and macros for SKL Watermark regs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This patch defines SKL specific PLANE_WM Watermark registers. It also defines macros to get the addresses of different LP levels within a pipe. v2: Reworked the register definitions and associated macros to make it more generic and be able to use for_each_pipe in values computation. Incorporated Damien's review comments and indentation. v3: Added default values for lines and blocks. Provided mask for blocks. v4: Prefix intermedidate (internal-only) macros with _ (Ville) v5: Remove the lines and block defaults value (Ville) Reviewed-by: Ville Syrjälä (v4) Signed-off-by: Pradeep Bhat Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index db7d8fd..ccbaeb3 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4102,6 +4102,41 @@ enum punit_power_well { #define I965_CURSOR_MAX_WM 32 #define I965_CURSOR_DFT_WM 8 +/* Watermark register definitions for SKL */ +#define CUR_WM_A_0 0x70140 +#define CUR_WM_B_0 0x71140 +#define PLANE_WM_1_A_0 0x70240 +#define PLANE_WM_1_B_0 0x71240 +#define PLANE_WM_2_A_0 0x70340 +#define PLANE_WM_2_B_0 0x71340 +#define PLANE_WM_TRANS_1_A_0 0x70268 +#define PLANE_WM_TRANS_1_B_0 0x71268 +#define PLANE_WM_TRANS_2_A_0 0x70368 +#define PLANE_WM_TRANS_2_B_0 0x71368 +#define CUR_WM_TRANS_A_0 0x70168 +#define CUR_WM_TRANS_B_0 0x71168 +#define PLANE_WM_EN (1 << 31) +#define PLANE_WM_LINES_SHIFT 14 +#define PLANE_WM_LINES_MASK 0x1f +#define PLANE_WM_BLOCKS_MASK 0x3ff + +#define CUR_WM_0(pipe) _PIPE(pipe, CUR_WM_A_0, CUR_WM_B_0) +#define CUR_WM(pipe, level) (CUR_WM_0(pipe) + ((4) * (level))) +#define CUR_WM_TRANS(pipe) _PIPE(pipe, CUR_WM_TRANS_A_0, CUR_WM_TRANS_B_0) + +#define _PLANE_WM_1(pipe) _PIPE(pipe, PLANE_WM_1_A_0, PLANE_WM_1_B_0) +#define _PLANE_WM_2(pipe) _PIPE(pipe, PLANE_WM_2_A_0, PLANE_WM_2_B_0) +#define _PLANE_WM_BASE(pipe, plane) \ + _PLANE(plane, _PLANE_WM_1(pipe), _PLANE_WM_2(pipe)) +#define PLANE_WM(pipe, plane, level) \ + (_PLANE_WM_BASE(pipe, plane) + ((4) * (level))) +#define _PLANE_WM_TRANS_1(pipe) \ + _PIPE(pipe, PLANE_WM_TRANS_1_A_0, PLANE_WM_TRANS_1_B_0) +#define _PLANE_WM_TRANS_2(pipe) \ + _PIPE(pipe, PLANE_WM_TRANS_2_A_0, PLANE_WM_TRANS_2_B_0) +#define PLANE_WM_TRANS(pipe, plane) \ + _PLANE(plane, _PLANE_WM_TRANS_1(pipe), _PLANE_WM_TRANS_2(pipe)) + /* define the Watermark register on Ironlake */ #define WM0_PIPEA_ILK 0x45100 #define WM0_PIPE_PLANE_MASK (0xffff<<16) -- cgit v0.10.2 From 2ac96d2a6e102ec64a568a470a6d9c50b9c02e59 Mon Sep 17 00:00:00 2001 From: Pradeep Bhat Date: Tue, 4 Nov 2014 17:06:40 +0000 Subject: drm/i915/skl: Definition of SKL WM param structs for pipe/plane MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This patch defines the structures needed for computation of watermarks of pipes and planes for SKL. v2: Incorporated Damien's review comments and removed unused fields in structs for future features like rotation, drrs and scaling. The skl_wm_values struct is now made more generic across planes and cursor planes for all pipes. v3: implemented the plane/cursor split. v4: Change the wm union back to a structure (Ville, Daniel) Reviewed-by: Ville Syrjälä Signed-off-by: Pradeep Bhat Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 05fcbe5..19bc2d0 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1387,6 +1387,24 @@ struct ilk_wm_values { enum intel_ddb_partitioning partitioning; }; +struct skl_wm_values { + bool dirty[I915_MAX_PIPES]; + uint32_t wm_linetime[I915_MAX_PIPES]; + uint32_t plane[I915_MAX_PIPES][I915_MAX_PLANES][8]; + uint32_t cursor[I915_MAX_PIPES][8]; + uint32_t plane_trans[I915_MAX_PIPES][I915_MAX_PLANES]; + uint32_t cursor_trans[I915_MAX_PIPES]; +}; + +struct skl_wm_level { + bool plane_en[I915_MAX_PLANES]; + uint16_t plane_res_b[I915_MAX_PLANES]; + uint8_t plane_res_l[I915_MAX_PLANES]; + bool cursor_en; + uint16_t cursor_res_b; + uint8_t cursor_res_l; +}; + /* * This struct helps tracking the state needed for runtime PM, which puts the * device in PCI D3 state. Notice that when this happens, nothing on the diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index c8ddde6..446be27 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -412,6 +412,12 @@ struct intel_mmio_flip { struct work_struct work; }; +struct skl_pipe_wm { + struct skl_wm_level wm[8]; + struct skl_wm_level trans_wm; + uint32_t linetime; +}; + struct intel_crtc { struct drm_crtc base; enum pipe pipe; @@ -459,6 +465,8 @@ struct intel_crtc { struct { /* watermarks currently being used */ struct intel_pipe_wm active; + /* SKL wm values currently in use */ + struct skl_pipe_wm skl_active; } wm; int scanline_offset; diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 761c884..ca2080b 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -1960,6 +1960,14 @@ static uint32_t ilk_wm_fbc(uint32_t pri_val, uint32_t horiz_pixels, return DIV_ROUND_UP(pri_val * 64, horiz_pixels * bytes_per_pixel) + 2; } +struct skl_pipe_wm_parameters { + bool active; + uint32_t pipe_htotal; + uint32_t pixel_rate; /* in KHz */ + struct intel_plane_wm_parameters plane[I915_MAX_PLANES]; + struct intel_plane_wm_parameters cursor; +}; + struct ilk_pipe_wm_parameters { bool active; uint32_t pipe_htotal; -- cgit v0.10.2 From c193924e9cef42995cbd204253520efc9d8ad1ba Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:41 +0000 Subject: drm/i915/skl: Add DDB allocation management structures MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We now need to allocate space in the DDB for planes being scanned out ourselves. The data structure to represent an allocation mirrors what we'll need to write in the registers later on: (start, end). We add that allocation datat to the skl_wm_values structure as part of the values to program the hardware with. v2: Split planes and cursor for consistency. v3: Make the skl_ddb_entry_size() parameter const Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 19bc2d0..aa8607a 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1387,8 +1387,27 @@ struct ilk_wm_values { enum intel_ddb_partitioning partitioning; }; +struct skl_ddb_entry { + uint16_t start, end; /* in number of blocks */ +}; + +static inline uint16_t skl_ddb_entry_size(const struct skl_ddb_entry *entry) +{ + /* end not set, clearly no allocation here. start can be 0 though */ + if (entry->end == 0) + return 0; + + return entry->end - entry->start + 1; +} + +struct skl_ddb_allocation { + struct skl_ddb_entry plane[I915_MAX_PIPES][I915_MAX_PLANES]; + struct skl_ddb_entry cursor[I915_MAX_PIPES]; +}; + struct skl_wm_values { bool dirty[I915_MAX_PIPES]; + struct skl_ddb_allocation ddb; uint32_t wm_linetime[I915_MAX_PIPES]; uint32_t plane[I915_MAX_PIPES][I915_MAX_PLANES][8]; uint32_t cursor[I915_MAX_PIPES][8]; -- cgit v0.10.2 From 2d41c0b59afcdd87d689212cdb929709ffc27d08 Mon Sep 17 00:00:00 2001 From: Pradeep Bhat Date: Tue, 4 Nov 2014 17:06:42 +0000 Subject: drm/i915/skl: SKL Watermark Computation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This patch implements the watermark algorithm and its necessary functions. Two function pointers skl_update_wm and skl_update_sprite_wm are provided. The skl_update_wm will update the watermarks for the crtc provided as an argument and then checks for change in DDB allocation for other active pipes and recomputes the watermarks for those Pipes and planes as well. Finally it does the register programming for all dirty pipes. The trigger of the Watermark double buffer registers will have to be once the plane configurations are done by the caller. v2: fixed the divide-by-0 error in the results computation func. Also reworked the PLANE_WM register values computation func to make it more compact. Incorporated all other review comments from Damien. v3: Changed the skl_compute_plane_wm function to now return success or failure. Also the result blocks and lines are computed here instead of in skl_compute_wm_results function. v4: Adjust skl_ddb_alloc_changed() to the new planes/cursor split (Damien) v5: Reworked the affected functions to implement new plane/cursor split. v6: Rework the logic that triggers the DDB allocation and WM computation of skl_update_other_pipe_wm() to not depend on non-computed DDB values. Always give a valid cursor_width (at boot it's 0) to keep the invariant that we consider the cursor plane always enabled. Otherwise we end up dividing by 0 in skl_compute_plane_wm() (Damien Lespiau) v7: Spell out allocation skl_ddb_ functions should have the ddb as first argument Make the skl_ddb_alloc_changed() parameters const (Damien) v8: Rebase on top of the crtc->primary changes v9: Split the staging results structure to not exceed the 1Kb stack allocation in skl_update_wm() v10: Make skl_pipe_pixel_rate() take a pointer to the pipe config Add a comment about overflow considerations for skl_wm_method1() Various additions of const Various use of sizeof(variable) instead of sizeof(type) Various move of variable definitons to a narrower scope Zero initialize some stack allocated structures to make sure we don't have garbage in case we don't write all the values (Ville) v11: Remove non-necessary default number of blocks/lines when the plane is disabled (Ville) Reviewed-by: Ville Syrjälä Signed-off-by: Pradeep Bhat Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index aa8607a..fbe18ae 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1722,8 +1722,18 @@ struct drm_i915_private { */ uint16_t skl_latency[8]; + /* + * The skl_wm_values structure is a bit too big for stack + * allocation, so we keep the staging struct where we store + * intermediate results here instead. + */ + struct skl_wm_values skl_results; + /* current hardware state */ - struct ilk_wm_values hw; + union { + struct ilk_wm_values hw; + struct skl_wm_values skl_hw; + }; } wm; struct i915_runtime_pm pm; diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index ca2080b..5500c53 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -2973,6 +2973,426 @@ static bool ilk_disable_lp_wm(struct drm_device *dev) return _ilk_disable_lp_wm(dev_priv, WM_DIRTY_LP_ALL); } +static uint32_t skl_pipe_pixel_rate(const struct intel_crtc_config *config) +{ + /* TODO: Take into account the scalers once we support them */ + return config->adjusted_mode.crtc_clock; +} + +/* + * The max latency should be 257 (max the punit can code is 255 and we add 2us + * for the read latency) and bytes_per_pixel should always be <= 8, so that + * should allow pixel_rate up to ~2 GHz which seems sufficient since max + * 2xcdclk is 1350 MHz and the pixel rate should never exceed that. +*/ +static uint32_t skl_wm_method1(uint32_t pixel_rate, uint8_t bytes_per_pixel, + uint32_t latency) +{ + uint32_t wm_intermediate_val, ret; + + if (latency == 0) + return UINT_MAX; + + wm_intermediate_val = latency * pixel_rate * bytes_per_pixel; + ret = DIV_ROUND_UP(wm_intermediate_val, 1000); + + return ret; +} + +static uint32_t skl_wm_method2(uint32_t pixel_rate, uint32_t pipe_htotal, + uint32_t horiz_pixels, uint8_t bytes_per_pixel, + uint32_t latency) +{ + uint32_t ret, plane_bytes_per_line, wm_intermediate_val; + + if (latency == 0) + return UINT_MAX; + + plane_bytes_per_line = horiz_pixels * bytes_per_pixel; + wm_intermediate_val = latency * pixel_rate; + ret = DIV_ROUND_UP(wm_intermediate_val, pipe_htotal * 1000) * + plane_bytes_per_line; + + return ret; +} + +static void skl_compute_transition_wm(struct drm_crtc *crtc, + struct skl_pipe_wm_parameters *params, + struct skl_pipe_wm *pipe_wm) +{ + /* + * For now it is suggested to use the LP0 wm val of corresponding + * plane as transition wm val. This is done while computing results. + */ + if (!params->active) + return; +} + +static uint32_t +skl_compute_linetime_wm(struct drm_crtc *crtc, struct skl_pipe_wm_parameters *p) +{ + if (!intel_crtc_active(crtc)) + return 0; + + return DIV_ROUND_UP(8 * p->pipe_htotal * 1000, p->pixel_rate); + +} + +static bool skl_ddb_allocation_changed(const struct skl_ddb_allocation *new_ddb, + const struct intel_crtc *intel_crtc) +{ + struct drm_device *dev = intel_crtc->base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + const struct skl_ddb_allocation *cur_ddb = &dev_priv->wm.skl_hw.ddb; + enum pipe pipe = intel_crtc->pipe; + + if (memcmp(new_ddb->plane[pipe], cur_ddb->plane[pipe], + sizeof(new_ddb->plane[pipe]))) + return true; + + if (memcmp(&new_ddb->cursor[pipe], &cur_ddb->cursor[pipe], + sizeof(new_ddb->cursor[pipe]))) + return true; + + return false; +} + +static void skl_compute_wm_global_parameters(struct drm_device *dev, + struct intel_wm_config *config) +{ + struct drm_crtc *crtc; + struct drm_plane *plane; + + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) + config->num_pipes_active += intel_crtc_active(crtc); + + /* FIXME: I don't think we need those two global parameters on SKL */ + list_for_each_entry(plane, &dev->mode_config.plane_list, head) { + struct intel_plane *intel_plane = to_intel_plane(plane); + + config->sprites_enabled |= intel_plane->wm.enabled; + config->sprites_scaled |= intel_plane->wm.scaled; + } +} + +static void skl_compute_wm_pipe_parameters(struct drm_crtc *crtc, + struct skl_pipe_wm_parameters *p) +{ + struct drm_device *dev = crtc->dev; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + enum pipe pipe = intel_crtc->pipe; + struct drm_plane *plane; + int i = 1; /* Index for sprite planes start */ + + p->active = intel_crtc_active(crtc); + if (p->active) { + p->pipe_htotal = intel_crtc->config.adjusted_mode.crtc_htotal; + p->pixel_rate = skl_pipe_pixel_rate(&intel_crtc->config); + + /* + * For now, assume primary and cursor planes are always enabled. + */ + p->plane[0].enabled = true; + p->plane[0].bytes_per_pixel = + crtc->primary->fb->bits_per_pixel / 8; + p->plane[0].horiz_pixels = intel_crtc->config.pipe_src_w; + p->plane[0].vert_pixels = intel_crtc->config.pipe_src_h; + + p->cursor.enabled = true; + p->cursor.bytes_per_pixel = 4; + p->cursor.horiz_pixels = intel_crtc->cursor_width ? + intel_crtc->cursor_width : 64; + } + + list_for_each_entry(plane, &dev->mode_config.plane_list, head) { + struct intel_plane *intel_plane = to_intel_plane(plane); + + if (intel_plane->pipe == pipe) + p->plane[i++] = intel_plane->wm; + } +} + +static bool skl_compute_plane_wm(struct skl_pipe_wm_parameters *p, + struct intel_plane_wm_parameters *p_params, + uint16_t max_page_buff_alloc, + uint32_t mem_value, + uint16_t *res_blocks, /* out */ + uint8_t *res_lines /* out */) +{ + uint32_t method1, method2, plane_bytes_per_line; + uint32_t result_bytes; + + if (!p->active || !p_params->enabled) + return false; + + method1 = skl_wm_method1(p->pixel_rate, + p_params->bytes_per_pixel, + mem_value); + method2 = skl_wm_method2(p->pixel_rate, + p->pipe_htotal, + p_params->horiz_pixels, + p_params->bytes_per_pixel, + mem_value); + + plane_bytes_per_line = p_params->horiz_pixels * + p_params->bytes_per_pixel; + + /* For now xtile and linear */ + if (((max_page_buff_alloc * 512) / plane_bytes_per_line) >= 1) + result_bytes = min(method1, method2); + else + result_bytes = method1; + + *res_blocks = DIV_ROUND_UP(result_bytes, 512) + 1; + *res_lines = DIV_ROUND_UP(result_bytes, plane_bytes_per_line); + + return true; +} + +static void skl_compute_wm_level(const struct drm_i915_private *dev_priv, + struct skl_ddb_allocation *ddb, + struct skl_pipe_wm_parameters *p, + enum pipe pipe, + int level, + int num_planes, + struct skl_wm_level *result) +{ + uint16_t latency = dev_priv->wm.skl_latency[level]; + uint16_t ddb_blocks; + int i; + + for (i = 0; i < num_planes; i++) { + ddb_blocks = skl_ddb_entry_size(&ddb->plane[pipe][i]); + + result->plane_en[i] = skl_compute_plane_wm(p, &p->plane[i], + ddb_blocks, + latency, + &result->plane_res_b[i], + &result->plane_res_l[i]); + } + + ddb_blocks = skl_ddb_entry_size(&ddb->cursor[pipe]); + result->cursor_en = skl_compute_plane_wm(p, &p->cursor, ddb_blocks, + latency, &result->cursor_res_b, + &result->cursor_res_l); +} + +static void skl_compute_pipe_wm(struct drm_crtc *crtc, + struct skl_ddb_allocation *ddb, + struct skl_pipe_wm_parameters *params, + struct skl_pipe_wm *pipe_wm) +{ + struct drm_device *dev = crtc->dev; + const struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + int level, max_level = ilk_wm_max_level(dev); + + for (level = 0; level <= max_level; level++) { + skl_compute_wm_level(dev_priv, ddb, params, intel_crtc->pipe, + level, intel_num_planes(intel_crtc), + &pipe_wm->wm[level]); + } + pipe_wm->linetime = skl_compute_linetime_wm(crtc, params); + + skl_compute_transition_wm(crtc, params, pipe_wm); +} + +static void skl_compute_wm_results(struct drm_device *dev, + struct skl_pipe_wm_parameters *p, + struct skl_pipe_wm *p_wm, + struct skl_wm_values *r, + struct intel_crtc *intel_crtc) +{ + int level, max_level = ilk_wm_max_level(dev); + enum pipe pipe = intel_crtc->pipe; + + for (level = 0; level <= max_level; level++) { + uint16_t ddb_blocks; + uint32_t temp; + int i; + + for (i = 0; i < intel_num_planes(intel_crtc); i++) { + temp = 0; + ddb_blocks = skl_ddb_entry_size(&r->ddb.plane[pipe][i]); + + if ((p_wm->wm[level].plane_res_b[i] > ddb_blocks) || + (p_wm->wm[level].plane_res_l[i] > 31)) + p_wm->wm[level].plane_en[i] = false; + + temp |= p_wm->wm[level].plane_res_l[i] << + PLANE_WM_LINES_SHIFT; + temp |= p_wm->wm[level].plane_res_b[i]; + if (p_wm->wm[level].plane_en[i]) + temp |= PLANE_WM_EN; + + r->plane[pipe][i][level] = temp; + /* Use the LP0 WM value for transition WM for now. */ + if (level == 0) + r->plane_trans[pipe][i] = temp; + } + + temp = 0; + ddb_blocks = skl_ddb_entry_size(&r->ddb.cursor[pipe]); + + if ((p_wm->wm[level].cursor_res_b > ddb_blocks) || + (p_wm->wm[level].cursor_res_l > 31)) + p_wm->wm[level].cursor_en = false; + + temp |= p_wm->wm[level].cursor_res_l << PLANE_WM_LINES_SHIFT; + temp |= p_wm->wm[level].cursor_res_b; + + if (p_wm->wm[level].cursor_en) + temp |= PLANE_WM_EN; + + r->cursor[pipe][level] = temp; + /* Use the LP0 WM value for transition WM for now. */ + if (level == 0) + r->cursor_trans[pipe] = temp; + + } + + r->wm_linetime[pipe] = p_wm->linetime; +} + +static void skl_write_wm_values(struct drm_i915_private *dev_priv, + const struct skl_wm_values *new) +{ + struct drm_device *dev = dev_priv->dev; + struct intel_crtc *crtc; + + list_for_each_entry(crtc, &dev->mode_config.crtc_list, base.head) { + int i, level, max_level = ilk_wm_max_level(dev); + enum pipe pipe = crtc->pipe; + + if (new->dirty[pipe]) { + I915_WRITE(PIPE_WM_LINETIME(pipe), + new->wm_linetime[pipe]); + + for (level = 0; level <= max_level; level++) { + for (i = 0; i < intel_num_planes(crtc); i++) + I915_WRITE(PLANE_WM(pipe, i, level), + new->plane[pipe][i][level]); + I915_WRITE(CUR_WM(pipe, level), + new->cursor[pipe][level]); + } + for (i = 0; i < intel_num_planes(crtc); i++) + I915_WRITE(PLANE_WM_TRANS(pipe, i), + new->plane_trans[pipe][i]); + I915_WRITE(CUR_WM_TRANS(pipe), new->cursor_trans[pipe]); + } + } + + dev_priv->wm.skl_hw = *new; +} + +static bool skl_update_pipe_wm(struct drm_crtc *crtc, + struct skl_pipe_wm_parameters *params, + struct intel_wm_config *config, + struct skl_ddb_allocation *ddb, /* out */ + struct skl_pipe_wm *pipe_wm /* out */) +{ + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + + skl_compute_wm_pipe_parameters(crtc, params); + skl_compute_pipe_wm(crtc, ddb, params, pipe_wm); + + if (!memcmp(&intel_crtc->wm.skl_active, pipe_wm, sizeof(*pipe_wm))) + return false; + + intel_crtc->wm.skl_active = *pipe_wm; + return true; +} + +static void skl_update_other_pipe_wm(struct drm_device *dev, + struct drm_crtc *crtc, + struct intel_wm_config *config, + struct skl_wm_values *r) +{ + struct intel_crtc *intel_crtc; + struct intel_crtc *this_crtc = to_intel_crtc(crtc); + + /* + * If the WM update hasn't changed the allocation for this_crtc (the + * crtc we are currently computing the new WM values for), other + * enabled crtcs will keep the same allocation and we don't need to + * recompute anything for them. + */ + if (!skl_ddb_allocation_changed(&r->ddb, this_crtc)) + return; + + /* + * Otherwise, because of this_crtc being freshly enabled/disabled, the + * other active pipes need new DDB allocation and WM values. + */ + list_for_each_entry(intel_crtc, &dev->mode_config.crtc_list, + base.head) { + struct skl_pipe_wm_parameters params = {}; + struct skl_pipe_wm pipe_wm = {}; + bool wm_changed; + + if (this_crtc->pipe == intel_crtc->pipe) + continue; + + if (!intel_crtc->active) + continue; + + wm_changed = skl_update_pipe_wm(&intel_crtc->base, + ¶ms, config, + &r->ddb, &pipe_wm); + + /* + * If we end up re-computing the other pipe WM values, it's + * because it was really needed, so we expect the WM values to + * be different. + */ + WARN_ON(!wm_changed); + + skl_compute_wm_results(dev, ¶ms, &pipe_wm, r, intel_crtc); + r->dirty[intel_crtc->pipe] = true; + } +} + +static void skl_update_wm(struct drm_crtc *crtc) +{ + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct drm_device *dev = crtc->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct skl_pipe_wm_parameters params = {}; + struct skl_wm_values *results = &dev_priv->wm.skl_results; + struct skl_pipe_wm pipe_wm = {}; + struct intel_wm_config config = {}; + + memset(results, 0, sizeof(*results)); + + skl_compute_wm_global_parameters(dev, &config); + + if (!skl_update_pipe_wm(crtc, ¶ms, &config, + &results->ddb, &pipe_wm)) + return; + + skl_compute_wm_results(dev, ¶ms, &pipe_wm, results, intel_crtc); + results->dirty[intel_crtc->pipe] = true; + + skl_update_other_pipe_wm(dev, crtc, &config, results); + skl_write_wm_values(dev_priv, results); +} + +static void +skl_update_sprite_wm(struct drm_plane *plane, struct drm_crtc *crtc, + uint32_t sprite_width, uint32_t sprite_height, + int pixel_size, bool enabled, bool scaled) +{ + struct intel_plane *intel_plane = to_intel_plane(plane); + + intel_plane->wm.enabled = enabled; + intel_plane->wm.scaled = scaled; + intel_plane->wm.horiz_pixels = sprite_width; + intel_plane->wm.vert_pixels = sprite_height; + intel_plane->wm.bytes_per_pixel = pixel_size; + + skl_update_wm(crtc); +} + static void ilk_update_wm(struct drm_crtc *crtc) { struct intel_crtc *intel_crtc = to_intel_crtc(crtc); @@ -6198,6 +6618,8 @@ void intel_init_pm(struct drm_device *dev) skl_setup_wm_latency(dev); dev_priv->display.init_clock_gating = gen9_init_clock_gating; + dev_priv->display.update_wm = skl_update_wm; + dev_priv->display.update_sprite_wm = skl_update_sprite_wm; } else if (HAS_PCH_SPLIT(dev)) { ilk_setup_wm_latency(dev); -- cgit v0.10.2 From b9cec07585cf1f554810097b8088b8e1f3f5137b Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:43 +0000 Subject: drm/i915/skl: Allocate DDB portions for display planes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit v2: Fix the 3rd plane/cursor logic (Pradeep Bhat) v3: Fix one-by-one error in the DDB allocation code v4: Rebase on top of the skl_pipe_pixel_rate() argument change v5: Replace the available/start/end output parameters of skl_ddb_get_pipe_allocation_limits() by a single ddb entry constify a few arguments Make nth_active_pipe 0 indexed Use sizeof(variable) instead of sizeof(type) (Ville) v6: Use the for_each_crtc() macro instead of list_for_each_entry() Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 5500c53..b5c7c4b 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -2973,6 +2973,153 @@ static bool ilk_disable_lp_wm(struct drm_device *dev) return _ilk_disable_lp_wm(dev_priv, WM_DIRTY_LP_ALL); } +/* + * On gen9, we need to allocate Display Data Buffer (DDB) portions to the + * different active planes. + */ + +#define SKL_DDB_SIZE 896 /* in blocks */ + +static void +skl_ddb_get_pipe_allocation_limits(struct drm_device *dev, + struct drm_crtc *for_crtc, + const struct intel_wm_config *config, + const struct skl_pipe_wm_parameters *params, + struct skl_ddb_entry *alloc /* out */) +{ + struct drm_crtc *crtc; + unsigned int pipe_size, ddb_size; + int nth_active_pipe; + + if (!params->active) { + alloc->start = 0; + alloc->end = 0; + return; + } + + ddb_size = SKL_DDB_SIZE; + + ddb_size -= 4; /* 4 blocks for bypass path allocation */ + + nth_active_pipe = 0; + for_each_crtc(dev, crtc) { + if (!intel_crtc_active(crtc)) + continue; + + if (crtc == for_crtc) + break; + + nth_active_pipe++; + } + + pipe_size = ddb_size / config->num_pipes_active; + alloc->start = nth_active_pipe * ddb_size / config->num_pipes_active; + alloc->end = alloc->start + pipe_size - 1; +} + +static unsigned int skl_cursor_allocation(const struct intel_wm_config *config) +{ + if (config->num_pipes_active == 1) + return 32; + + return 8; +} + +static unsigned int +skl_plane_relative_data_rate(const struct intel_plane_wm_parameters *p) +{ + return p->horiz_pixels * p->vert_pixels * p->bytes_per_pixel; +} + +/* + * We don't overflow 32 bits. Worst case is 3 planes enabled, each fetching + * a 8192x4096@32bpp framebuffer: + * 3 * 4096 * 8192 * 4 < 2^32 + */ +static unsigned int +skl_get_total_relative_data_rate(struct intel_crtc *intel_crtc, + const struct skl_pipe_wm_parameters *params) +{ + unsigned int total_data_rate = 0; + int plane; + + for (plane = 0; plane < intel_num_planes(intel_crtc); plane++) { + const struct intel_plane_wm_parameters *p; + + p = ¶ms->plane[plane]; + if (!p->enabled) + continue; + + total_data_rate += skl_plane_relative_data_rate(p); + } + + return total_data_rate; +} + +static void +skl_allocate_pipe_ddb(struct drm_crtc *crtc, + const struct intel_wm_config *config, + const struct skl_pipe_wm_parameters *params, + struct skl_ddb_allocation *ddb /* out */) +{ + struct drm_device *dev = crtc->dev; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + enum pipe pipe = intel_crtc->pipe; + struct skl_ddb_entry alloc; + uint16_t alloc_size, start, cursor_blocks; + unsigned int total_data_rate; + int plane; + + skl_ddb_get_pipe_allocation_limits(dev, crtc, config, params, &alloc); + alloc_size = skl_ddb_entry_size(&alloc); + if (alloc_size == 0) { + memset(ddb->plane[pipe], 0, sizeof(ddb->plane[pipe])); + memset(&ddb->cursor[pipe], 0, sizeof(ddb->cursor[pipe])); + return; + } + + cursor_blocks = skl_cursor_allocation(config); + ddb->cursor[pipe].start = alloc.end - cursor_blocks + 1; + ddb->cursor[pipe].end = alloc.end; + + alloc_size -= cursor_blocks; + alloc.end -= cursor_blocks; + + /* + * Each active plane get a portion of the remaining space, in + * proportion to the amount of data they need to fetch from memory. + * + * FIXME: we may not allocate every single block here. + */ + total_data_rate = skl_get_total_relative_data_rate(intel_crtc, params); + + start = alloc.start; + for (plane = 0; plane < intel_num_planes(intel_crtc); plane++) { + const struct intel_plane_wm_parameters *p; + unsigned int data_rate; + uint16_t plane_blocks; + + p = ¶ms->plane[plane]; + if (!p->enabled) + continue; + + data_rate = skl_plane_relative_data_rate(p); + + /* + * promote the expression to 64 bits to avoid overflowing, the + * result is < available as data_rate / total_data_rate < 1 + */ + plane_blocks = div_u64((uint64_t)alloc_size * data_rate, + total_data_rate); + + ddb->plane[pipe][plane].start = start; + ddb->plane[pipe][plane].end = start + plane_blocks - 1; + + start += plane_blocks; + } + +} + static uint32_t skl_pipe_pixel_rate(const struct intel_crtc_config *config) { /* TODO: Take into account the scalers once we support them */ @@ -3294,6 +3441,7 @@ static bool skl_update_pipe_wm(struct drm_crtc *crtc, struct intel_crtc *intel_crtc = to_intel_crtc(crtc); skl_compute_wm_pipe_parameters(crtc, params); + skl_allocate_pipe_ddb(crtc, config, params, ddb); skl_compute_pipe_wm(crtc, ddb, params, pipe_wm); if (!memcmp(&intel_crtc->wm.skl_active, pipe_wm, sizeof(*pipe_wm))) -- cgit v0.10.2 From 8211bd5bdf5e1cf860616138ed71600a24a827bd Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:44 +0000 Subject: drm/i915/skl: Program the DDB allocation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit v2: Adapt to the planes/cursor split Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index ccbaeb3..ea22e09 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4696,6 +4696,8 @@ enum punit_power_well { #define _PLANE_KEYMSK_2_A 0x70298 #define _PLANE_KEYMAX_1_A 0x701a0 #define _PLANE_KEYMAX_2_A 0x702a0 +#define _PLANE_BUF_CFG_1_A 0x7027c +#define _PLANE_BUF_CFG_2_A 0x7037c #define _PLANE_CTL_1_B 0x71180 #define _PLANE_CTL_2_B 0x71280 @@ -4773,6 +4775,20 @@ enum punit_power_well { #define PLANE_KEYMAX(pipe, plane) \ _PLANE(plane, _PLANE_KEYMAX_1(pipe), _PLANE_KEYMAX_2(pipe)) +#define _PLANE_BUF_CFG_1_B 0x7127c +#define _PLANE_BUF_CFG_2_B 0x7137c +#define _PLANE_BUF_CFG_1(pipe) \ + _PIPE(pipe, _PLANE_BUF_CFG_1_A, _PLANE_BUF_CFG_1_B) +#define _PLANE_BUF_CFG_2(pipe) \ + _PIPE(pipe, _PLANE_BUF_CFG_2_A, _PLANE_BUF_CFG_2_B) +#define PLANE_BUF_CFG(pipe, plane) \ + _PLANE(plane, _PLANE_BUF_CFG_1(pipe), _PLANE_BUF_CFG_2(pipe)) + +/* SKL new cursor registers */ +#define _CUR_BUF_CFG_A 0x7017c +#define _CUR_BUF_CFG_B 0x7117c +#define CUR_BUF_CFG(pipe) _PIPE(pipe, _CUR_BUF_CFG_A, _CUR_BUF_CFG_B) + /* VBIOS regs */ #define VGACNTRL 0x71400 # define VGA_DISP_DISABLE (1 << 31) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index b5c7c4b..c2f1441 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3426,6 +3426,15 @@ static void skl_write_wm_values(struct drm_i915_private *dev_priv, I915_WRITE(PLANE_WM_TRANS(pipe, i), new->plane_trans[pipe][i]); I915_WRITE(CUR_WM_TRANS(pipe), new->cursor_trans[pipe]); + + for (i = 0; i < intel_num_planes(crtc); i++) + I915_WRITE(PLANE_BUF_CFG(pipe, i), + new->ddb.plane[pipe][i].end << 16 | + new->ddb.plane[pipe][i].start); + + I915_WRITE(CUR_BUF_CFG(pipe), + new->ddb.cursor[pipe].end << 16 | + new->ddb.cursor[pipe].start); } } -- cgit v0.10.2 From 3078999f2a4ee4aaabf9df2166e4e43581bb357a Mon Sep 17 00:00:00 2001 From: Pradeep Bhat Date: Tue, 4 Nov 2014 17:06:45 +0000 Subject: drm/i915/skl: Read the pipe WM HW state MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This patch provides the implementation for reading the pipe wm HW state. v2: Incorporated Damien's review comments and also made modifications to incorporate the plane/cursor split. v3: No need to ident a line that was fitting 80 chars Return early instead of indenting the remaining of a function (Damien) v4: Rebase on top of nightly (minor conflict in intel_drv.h) v5: Rebase on top of nightly (minor conflict in intel_drv.h) v6: Rebase on top of nightly (minor conflict in intel_drv.h) Reviewed-by: Ville Syrjälä Signed-off-by: Pradeep Bhat Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 84f4f05..01bc558 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -13266,7 +13266,9 @@ void intel_modeset_setup_hw_state(struct drm_device *dev, pll->on = false; } - if (HAS_PCH_SPLIT(dev)) + if (IS_GEN9(dev)) + skl_wm_get_hw_state(dev); + else if (HAS_PCH_SPLIT(dev)) ilk_wm_get_hw_state(dev); if (force_restore) { diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 446be27..1cdd6f9 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1164,6 +1164,7 @@ void gen6_update_ring_freq(struct drm_device *dev); void gen6_rps_idle(struct drm_i915_private *dev_priv); void gen6_rps_boost(struct drm_i915_private *dev_priv); void ilk_wm_get_hw_state(struct drm_device *dev); +void skl_wm_get_hw_state(struct drm_device *dev); /* intel_sdvo.c */ diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index c2f1441..09a81b3 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3624,6 +3624,110 @@ ilk_update_sprite_wm(struct drm_plane *plane, ilk_update_wm(crtc); } +static void skl_pipe_wm_active_state(uint32_t val, + struct skl_pipe_wm *active, + bool is_transwm, + bool is_cursor, + int i, + int level) +{ + bool is_enabled = (val & PLANE_WM_EN) != 0; + + if (!is_transwm) { + if (!is_cursor) { + active->wm[level].plane_en[i] = is_enabled; + active->wm[level].plane_res_b[i] = + val & PLANE_WM_BLOCKS_MASK; + active->wm[level].plane_res_l[i] = + (val >> PLANE_WM_LINES_SHIFT) & + PLANE_WM_LINES_MASK; + } else { + active->wm[level].cursor_en = is_enabled; + active->wm[level].cursor_res_b = + val & PLANE_WM_BLOCKS_MASK; + active->wm[level].cursor_res_l = + (val >> PLANE_WM_LINES_SHIFT) & + PLANE_WM_LINES_MASK; + } + } else { + if (!is_cursor) { + active->trans_wm.plane_en[i] = is_enabled; + active->trans_wm.plane_res_b[i] = + val & PLANE_WM_BLOCKS_MASK; + active->trans_wm.plane_res_l[i] = + (val >> PLANE_WM_LINES_SHIFT) & + PLANE_WM_LINES_MASK; + } else { + active->trans_wm.cursor_en = is_enabled; + active->trans_wm.cursor_res_b = + val & PLANE_WM_BLOCKS_MASK; + active->trans_wm.cursor_res_l = + (val >> PLANE_WM_LINES_SHIFT) & + PLANE_WM_LINES_MASK; + } + } +} + +static void skl_pipe_wm_get_hw_state(struct drm_crtc *crtc) +{ + struct drm_device *dev = crtc->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct skl_wm_values *hw = &dev_priv->wm.skl_hw; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct skl_pipe_wm *active = &intel_crtc->wm.skl_active; + enum pipe pipe = intel_crtc->pipe; + int level, i, max_level; + uint32_t temp; + + max_level = ilk_wm_max_level(dev); + + hw->wm_linetime[pipe] = I915_READ(PIPE_WM_LINETIME(pipe)); + + for (level = 0; level <= max_level; level++) { + for (i = 0; i < intel_num_planes(intel_crtc); i++) + hw->plane[pipe][i][level] = + I915_READ(PLANE_WM(pipe, i, level)); + hw->cursor[pipe][level] = I915_READ(CUR_WM(pipe, level)); + } + + for (i = 0; i < intel_num_planes(intel_crtc); i++) + hw->plane_trans[pipe][i] = I915_READ(PLANE_WM_TRANS(pipe, i)); + hw->cursor_trans[pipe] = I915_READ(CUR_WM_TRANS(pipe)); + + if (!intel_crtc_active(crtc)) + return; + + hw->dirty[pipe] = true; + + active->linetime = hw->wm_linetime[pipe]; + + for (level = 0; level <= max_level; level++) { + for (i = 0; i < intel_num_planes(intel_crtc); i++) { + temp = hw->plane[pipe][i][level]; + skl_pipe_wm_active_state(temp, active, false, + false, i, level); + } + temp = hw->cursor[pipe][level]; + skl_pipe_wm_active_state(temp, active, false, true, i, level); + } + + for (i = 0; i < intel_num_planes(intel_crtc); i++) { + temp = hw->plane_trans[pipe][i]; + skl_pipe_wm_active_state(temp, active, true, false, i, 0); + } + + temp = hw->cursor_trans[pipe]; + skl_pipe_wm_active_state(temp, active, true, true, i, 0); +} + +void skl_wm_get_hw_state(struct drm_device *dev) +{ + struct drm_crtc *crtc; + + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) + skl_pipe_wm_get_hw_state(crtc); +} + static void ilk_pipe_wm_get_hw_state(struct drm_crtc *crtc) { struct drm_device *dev = crtc->dev; -- cgit v0.10.2 From 367294be7c257dd4d8d8f0e296aa9824eae18568 Mon Sep 17 00:00:00 2001 From: Vandana Kannan Date: Tue, 4 Nov 2014 17:06:46 +0000 Subject: drm/i915/gen9: Add 2us read latency to WM level MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit According to the updated Bspec, The mailbox response data is not currently accounting for memory read latency. Add 2 microseconds to the result for each level. This patch adds 2us to latency of level 0 for all cases and for all other levels (1-7) only if latency[level] > 0. v2: Slightly rework the patch and add a big comment (Damien) v3: Rebase on top of the renames of the memory latency defines Signed-off-by: Vandana Kannan (v1) Signed-off-by: Damien Lespiau Reviewed-by: Ville Syrjälä (v2) Reviewed-by: M, Satheeshakrishna (v1) Cc: Lespiau, Damien Cc: M, Satheeshakrishna Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 09a81b3..44fecfc 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -2286,6 +2286,7 @@ static void intel_read_wm_latency(struct drm_device *dev, uint16_t wm[8]) if (IS_GEN9(dev)) { uint32_t val; int ret; + int level, max_level = ilk_wm_max_level(dev); /* read the first set of memory latencies[0:3] */ val = 0; /* data0 to be programmed to 0 for first set */ @@ -2328,6 +2329,21 @@ static void intel_read_wm_latency(struct drm_device *dev, uint16_t wm[8]) wm[7] = (val >> GEN9_MEM_LATENCY_LEVEL_3_7_SHIFT) & GEN9_MEM_LATENCY_LEVEL_MASK; + /* + * punit doesn't take into account the read latency so we need + * to add 2us to the various latency levels we retrieve from + * the punit. + * - W0 is a bit special in that it's the only level that + * can't be disabled if we want to have display working, so + * we always add 2us there. + * - For levels >=1, punit returns 0us latency when they are + * disabled, so we respect that and don't add 2us then + */ + wm[0] += 2; + for (level = 1; level <= max_level; level++) + if (wm[level] != 0) + wm[level] += 2; + } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { uint64_t sskpd = I915_READ64(MCH_SSKPD); -- cgit v0.10.2 From 4f94738674b813fe0ef8e7dbef8a24aeb8c2271a Mon Sep 17 00:00:00 2001 From: Vandana Kannan Date: Tue, 4 Nov 2014 17:06:47 +0000 Subject: drm/i915/gen9: Disable WM if corresponding latency is 0 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit According to updated BSpec, If level 1 or any higher level has a value of 0x00, that level and any higher levels are unused and the associated watermark registers must not be enabled. This patch checks for latency 0 for level >=1 and does not enable WM corresponding to level m | m>=n, if level n (n != 0) has a 0us latency. v2: Satheesh's review comments - zero-out latency values (for all higher levels if latency of given level is zero ) in read_wm_latency() function itself v3: removed redundant check as per Satheesh's observation. v4: rebase on top before merging (Damien) v5: Rebase on top of the default value removal (Ville) Reviewed-by: Satheeshakrishna M (v3) Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Vandana Kannan Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 44fecfc..bf2cd65 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -2285,7 +2285,7 @@ static void intel_read_wm_latency(struct drm_device *dev, uint16_t wm[8]) if (IS_GEN9(dev)) { uint32_t val; - int ret; + int ret, i; int level, max_level = ilk_wm_max_level(dev); /* read the first set of memory latencies[0:3] */ @@ -2338,12 +2338,22 @@ static void intel_read_wm_latency(struct drm_device *dev, uint16_t wm[8]) * we always add 2us there. * - For levels >=1, punit returns 0us latency when they are * disabled, so we respect that and don't add 2us then + * + * Additionally, if a level n (n > 1) has a 0us latency, all + * levels m (m >= n) need to be disabled. We make sure to + * sanitize the values out of the punit to satisfy this + * requirement. */ wm[0] += 2; for (level = 1; level <= max_level; level++) if (wm[level] != 0) wm[level] += 2; + else { + for (i = level + 1; i <= max_level; i++) + wm[i] = 0; + break; + } } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { uint64_t sskpd = I915_READ64(MCH_SSKPD); @@ -3285,7 +3295,7 @@ static bool skl_compute_plane_wm(struct skl_pipe_wm_parameters *p, uint32_t method1, method2, plane_bytes_per_line; uint32_t result_bytes; - if (!p->active || !p_params->enabled) + if (mem_value == 0 || !p->active || !p_params->enabled) return false; method1 = skl_wm_method1(p->pixel_rate, -- cgit v0.10.2 From 53b0deb4cba206c6f8cf9225080ac49e02183376 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:48 +0000 Subject: drm/i915/skl: Store the new WM state at the very end of the update MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We're going to add a new step, let's not hide the copy of the new WM state inside one inner function, but as a 1st level operation in the WM update. v2: Split the staging results structure to not exceed the 1Kb stack allocation in skl_update_wm() Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index bf2cd65..a4294b1 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3463,8 +3463,6 @@ static void skl_write_wm_values(struct drm_i915_private *dev_priv, new->ddb.cursor[pipe].start); } } - - dev_priv->wm.skl_hw = *new; } static bool skl_update_pipe_wm(struct drm_crtc *crtc, @@ -3558,6 +3556,9 @@ static void skl_update_wm(struct drm_crtc *crtc) skl_update_other_pipe_wm(dev, crtc, &config, results); skl_write_wm_values(dev_priv, results); + + /* store the new configuration */ + dev_priv->wm.skl_hw = *results; } static void -- cgit v0.10.2 From a269c5839bc4e3a997228314cb2efbddc9da604a Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:49 +0000 Subject: drm/i915/skl: Read back the DDB allocation hw state MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This logically belongs to the WM state, so do it there. Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index a4294b1..8b7abc7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3051,6 +3051,32 @@ static unsigned int skl_cursor_allocation(const struct intel_wm_config *config) return 8; } +static void skl_ddb_entry_init_from_hw(struct skl_ddb_entry *entry, u32 reg) +{ + entry->start = reg & 0x3ff; + entry->end = (reg >> 16) & 0x3ff; +} + +static void skl_ddb_get_hw_state(struct drm_i915_private *dev_priv, + struct skl_ddb_allocation *ddb /* out */) +{ + struct drm_device *dev = dev_priv->dev; + enum pipe pipe; + int plane; + u32 val; + + for_each_pipe(dev_priv, pipe) { + for_each_plane(pipe, plane) { + val = I915_READ(PLANE_BUF_CFG(pipe, plane)); + skl_ddb_entry_init_from_hw(&ddb->plane[pipe][plane], + val); + } + + val = I915_READ(CUR_BUF_CFG(pipe)); + skl_ddb_entry_init_from_hw(&ddb->cursor[pipe], val); + } +} + static unsigned int skl_plane_relative_data_rate(const struct intel_plane_wm_parameters *p) { @@ -3749,8 +3775,11 @@ static void skl_pipe_wm_get_hw_state(struct drm_crtc *crtc) void skl_wm_get_hw_state(struct drm_device *dev) { + struct drm_i915_private *dev_priv = dev->dev_private; + struct skl_ddb_allocation *ddb = &dev_priv->wm.skl_hw.ddb; struct drm_crtc *crtc; + skl_ddb_get_hw_state(dev_priv, ddb); list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) skl_pipe_wm_get_hw_state(crtc); } -- cgit v0.10.2 From 97e94b22ed4fe116beadbdbe4667f2a2f9aa9fd0 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:50 +0000 Subject: drm/i915/skl: Augment the latency debugfs files for SKL MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit v2: Use the gen >= 9 in the debugfs file condition (Ville) Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 0a69813..b1cbfbf 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -3523,7 +3523,7 @@ static const struct file_operations i915_display_crc_ctl_fops = { .write = display_crc_ctl_write }; -static void wm_latency_show(struct seq_file *m, const uint16_t wm[5]) +static void wm_latency_show(struct seq_file *m, const uint16_t wm[8]) { struct drm_device *dev = m->private; int num_levels = ilk_wm_max_level(dev) + 1; @@ -3534,13 +3534,17 @@ static void wm_latency_show(struct seq_file *m, const uint16_t wm[5]) for (level = 0; level < num_levels; level++) { unsigned int latency = wm[level]; - /* WM1+ latency values in 0.5us units */ - if (level > 0) + /* + * - WM1+ latency values in 0.5us units + * - latencies are in us on gen9 + */ + if (INTEL_INFO(dev)->gen >= 9) + latency *= 10; + else if (level > 0) latency *= 5; seq_printf(m, "WM%d %u (%u.%u usec)\n", - level, wm[level], - latency / 10, latency % 10); + level, wm[level], latency / 10, latency % 10); } drm_modeset_unlock_all(dev); @@ -3549,8 +3553,15 @@ static void wm_latency_show(struct seq_file *m, const uint16_t wm[5]) static int pri_wm_latency_show(struct seq_file *m, void *data) { struct drm_device *dev = m->private; + struct drm_i915_private *dev_priv = dev->dev_private; + const uint16_t *latencies; + + if (INTEL_INFO(dev)->gen >= 9) + latencies = dev_priv->wm.skl_latency; + else + latencies = to_i915(dev)->wm.pri_latency; - wm_latency_show(m, to_i915(dev)->wm.pri_latency); + wm_latency_show(m, latencies); return 0; } @@ -3558,8 +3569,15 @@ static int pri_wm_latency_show(struct seq_file *m, void *data) static int spr_wm_latency_show(struct seq_file *m, void *data) { struct drm_device *dev = m->private; + struct drm_i915_private *dev_priv = dev->dev_private; + const uint16_t *latencies; + + if (INTEL_INFO(dev)->gen >= 9) + latencies = dev_priv->wm.skl_latency; + else + latencies = to_i915(dev)->wm.spr_latency; - wm_latency_show(m, to_i915(dev)->wm.spr_latency); + wm_latency_show(m, latencies); return 0; } @@ -3567,8 +3585,15 @@ static int spr_wm_latency_show(struct seq_file *m, void *data) static int cur_wm_latency_show(struct seq_file *m, void *data) { struct drm_device *dev = m->private; + struct drm_i915_private *dev_priv = dev->dev_private; + const uint16_t *latencies; + + if (INTEL_INFO(dev)->gen >= 9) + latencies = dev_priv->wm.skl_latency; + else + latencies = to_i915(dev)->wm.cur_latency; - wm_latency_show(m, to_i915(dev)->wm.cur_latency); + wm_latency_show(m, latencies); return 0; } @@ -3604,11 +3629,11 @@ static int cur_wm_latency_open(struct inode *inode, struct file *file) } static ssize_t wm_latency_write(struct file *file, const char __user *ubuf, - size_t len, loff_t *offp, uint16_t wm[5]) + size_t len, loff_t *offp, uint16_t wm[8]) { struct seq_file *m = file->private_data; struct drm_device *dev = m->private; - uint16_t new[5] = { 0 }; + uint16_t new[8] = { 0 }; int num_levels = ilk_wm_max_level(dev) + 1; int level; int ret; @@ -3622,7 +3647,9 @@ static ssize_t wm_latency_write(struct file *file, const char __user *ubuf, tmp[len] = '\0'; - ret = sscanf(tmp, "%hu %hu %hu %hu %hu", &new[0], &new[1], &new[2], &new[3], &new[4]); + ret = sscanf(tmp, "%hu %hu %hu %hu %hu %hu %hu %hu", + &new[0], &new[1], &new[2], &new[3], + &new[4], &new[5], &new[6], &new[7]); if (ret != num_levels) return -EINVAL; @@ -3642,8 +3669,15 @@ static ssize_t pri_wm_latency_write(struct file *file, const char __user *ubuf, { struct seq_file *m = file->private_data; struct drm_device *dev = m->private; + struct drm_i915_private *dev_priv = dev->dev_private; + uint16_t *latencies; - return wm_latency_write(file, ubuf, len, offp, to_i915(dev)->wm.pri_latency); + if (INTEL_INFO(dev)->gen >= 9) + latencies = dev_priv->wm.skl_latency; + else + latencies = to_i915(dev)->wm.pri_latency; + + return wm_latency_write(file, ubuf, len, offp, latencies); } static ssize_t spr_wm_latency_write(struct file *file, const char __user *ubuf, @@ -3651,8 +3685,15 @@ static ssize_t spr_wm_latency_write(struct file *file, const char __user *ubuf, { struct seq_file *m = file->private_data; struct drm_device *dev = m->private; + struct drm_i915_private *dev_priv = dev->dev_private; + uint16_t *latencies; - return wm_latency_write(file, ubuf, len, offp, to_i915(dev)->wm.spr_latency); + if (INTEL_INFO(dev)->gen >= 9) + latencies = dev_priv->wm.skl_latency; + else + latencies = to_i915(dev)->wm.spr_latency; + + return wm_latency_write(file, ubuf, len, offp, latencies); } static ssize_t cur_wm_latency_write(struct file *file, const char __user *ubuf, @@ -3660,8 +3701,15 @@ static ssize_t cur_wm_latency_write(struct file *file, const char __user *ubuf, { struct seq_file *m = file->private_data; struct drm_device *dev = m->private; + struct drm_i915_private *dev_priv = dev->dev_private; + uint16_t *latencies; + + if (INTEL_INFO(dev)->gen >= 9) + latencies = dev_priv->wm.skl_latency; + else + latencies = to_i915(dev)->wm.cur_latency; - return wm_latency_write(file, ubuf, len, offp, to_i915(dev)->wm.cur_latency); + return wm_latency_write(file, ubuf, len, offp, latencies); } static const struct file_operations i915_pri_wm_latency_fops = { -- cgit v0.10.2 From c5511e44e8a643cf6400e8dbf9e4f6d348786112 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:51 +0000 Subject: drm/i915/skl: Add a debugfs file to dump the DDB allocation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit v2: minor conflict in i915_debugfs.c v3: Rebase on top of the for_each_pipe() change adding dev_priv as first argument. v4: minor conflict in the i915_debugfs_files array v5: minor conflict in the i915_debugfs_files array Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index b1cbfbf..319da61 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2679,6 +2679,42 @@ static int i915_wa_registers(struct seq_file *m, void *unused) return 0; } +static int i915_ddb_info(struct seq_file *m, void *unused) +{ + struct drm_info_node *node = m->private; + struct drm_device *dev = node->minor->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct skl_ddb_allocation *ddb; + struct skl_ddb_entry *entry; + enum pipe pipe; + int plane; + + drm_modeset_lock_all(dev); + + ddb = &dev_priv->wm.skl_hw.ddb; + + seq_printf(m, "%-15s%8s%8s%8s\n", "", "Start", "End", "Size"); + + for_each_pipe(dev_priv, pipe) { + seq_printf(m, "Pipe %c\n", pipe_name(pipe)); + + for_each_plane(pipe, plane) { + entry = &ddb->plane[pipe][plane]; + seq_printf(m, " Plane%-8d%8u%8u%8u\n", plane + 1, + entry->start, entry->end, + skl_ddb_entry_size(entry)); + } + + entry = &ddb->cursor[pipe]; + seq_printf(m, " %-13s%8u%8u%8u\n", "Cursor", entry->start, + entry->end, skl_ddb_entry_size(entry)); + } + + drm_modeset_unlock_all(dev); + + return 0; +} + struct pipe_crc_info { const char *name; struct drm_device *dev; @@ -4252,6 +4288,7 @@ static const struct drm_info_list i915_debugfs_list[] = { {"i915_shared_dplls_info", i915_shared_dplls_info, 0}, {"i915_dp_mst_info", i915_dp_mst_info, 0}, {"i915_wa_registers", i915_wa_registers, 0}, + {"i915_ddb_info", i915_ddb_info, 0}, }; #define I915_DEBUGFS_ENTRIES ARRAY_SIZE(i915_debugfs_list) -- cgit v0.10.2 From 08db665203d5e587efc391afc405a9317d29fe17 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:52 +0000 Subject: drm/i915/skl: Check the DDB state at modeset MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit v2: Don't check DDB on pre-SKL platforms Don't check DDB state on disabled pipes v3: Squash "Expose skl_ddb_get_hw_state()" Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index fbe18ae..ee744a2 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1400,6 +1400,15 @@ static inline uint16_t skl_ddb_entry_size(const struct skl_ddb_entry *entry) return entry->end - entry->start + 1; } +static inline bool skl_ddb_entry_equal(const struct skl_ddb_entry *e1, + const struct skl_ddb_entry *e2) +{ + if (e1->start == e2->start && e1->end == e2->end) + return true; + + return false; +} + struct skl_ddb_allocation { struct skl_ddb_entry plane[I915_MAX_PIPES][I915_MAX_PLANES]; struct skl_ddb_entry cursor[I915_MAX_PIPES]; diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 01bc558..07fb0ff 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -10511,6 +10511,56 @@ intel_pipe_config_compare(struct drm_device *dev, return true; } +static void check_wm_state(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct skl_ddb_allocation hw_ddb, *sw_ddb; + struct intel_crtc *intel_crtc; + int plane; + + if (INTEL_INFO(dev)->gen < 9) + return; + + skl_ddb_get_hw_state(dev_priv, &hw_ddb); + sw_ddb = &dev_priv->wm.skl_hw.ddb; + + for_each_intel_crtc(dev, intel_crtc) { + struct skl_ddb_entry *hw_entry, *sw_entry; + const enum pipe pipe = intel_crtc->pipe; + + if (!intel_crtc->active) + continue; + + /* planes */ + for_each_plane(pipe, plane) { + hw_entry = &hw_ddb.plane[pipe][plane]; + sw_entry = &sw_ddb->plane[pipe][plane]; + + if (skl_ddb_entry_equal(hw_entry, sw_entry)) + continue; + + DRM_ERROR("mismatch in DDB state pipe %c plane %d " + "(expected (%u,%u), found (%u,%u))\n", + pipe_name(pipe), plane + 1, + sw_entry->start, sw_entry->end, + hw_entry->start, hw_entry->end); + } + + /* cursor */ + hw_entry = &hw_ddb.cursor[pipe]; + sw_entry = &sw_ddb->cursor[pipe]; + + if (skl_ddb_entry_equal(hw_entry, sw_entry)) + continue; + + DRM_ERROR("mismatch in DDB state pipe %c cursor " + "(expected (%u,%u), found (%u,%u))\n", + pipe_name(pipe), + sw_entry->start, sw_entry->end, + hw_entry->start, hw_entry->end); + } +} + static void check_connector_state(struct drm_device *dev) { @@ -10710,6 +10760,7 @@ check_shared_dpll_state(struct drm_device *dev) void intel_modeset_check_state(struct drm_device *dev) { + check_wm_state(dev); check_connector_state(dev); check_encoder_state(dev); check_crtc_state(dev); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 1cdd6f9..cb0e9db 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1165,6 +1165,8 @@ void gen6_rps_idle(struct drm_i915_private *dev_priv); void gen6_rps_boost(struct drm_i915_private *dev_priv); void ilk_wm_get_hw_state(struct drm_device *dev); void skl_wm_get_hw_state(struct drm_device *dev); +void skl_ddb_get_hw_state(struct drm_i915_private *dev_priv, + struct skl_ddb_allocation *ddb /* out */); /* intel_sdvo.c */ diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 8b7abc7..159c2e0 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3057,8 +3057,8 @@ static void skl_ddb_entry_init_from_hw(struct skl_ddb_entry *entry, u32 reg) entry->end = (reg >> 16) & 0x3ff; } -static void skl_ddb_get_hw_state(struct drm_i915_private *dev_priv, - struct skl_ddb_allocation *ddb /* out */) +void skl_ddb_get_hw_state(struct drm_i915_private *dev_priv, + struct skl_ddb_allocation *ddb /* out */) { struct drm_device *dev = dev_priv->dev; enum pipe pipe; -- cgit v0.10.2 From 16160e3dd32343de9d6115353408728adecdecfb Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:53 +0000 Subject: drm/i915/skl: Make 'end' of the DDB allocation entry exclusive MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Ville suggested that we should use the same semantics as C arrays to reduce the number of those pesky +1/-1 in the allocation code. This patch leaves the debugfs file as is, showing the internal DDB allocation structure, not the values written in the registers. v2: Remove the test on ->end in skl_ddb_entry_size() (Ville) Reviewed-by: Ville Syrjälä Suggested-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index ee744a2..84e8cdb 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1388,16 +1388,12 @@ struct ilk_wm_values { }; struct skl_ddb_entry { - uint16_t start, end; /* in number of blocks */ + uint16_t start, end; /* in number of blocks, 'end' is exclusive */ }; static inline uint16_t skl_ddb_entry_size(const struct skl_ddb_entry *entry) { - /* end not set, clearly no allocation here. start can be 0 though */ - if (entry->end == 0) - return 0; - - return entry->end - entry->start + 1; + return entry->end - entry->start; } static inline bool skl_ddb_entry_equal(const struct skl_ddb_entry *e1, diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 159c2e0..e417a6a 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3040,7 +3040,7 @@ skl_ddb_get_pipe_allocation_limits(struct drm_device *dev, pipe_size = ddb_size / config->num_pipes_active; alloc->start = nth_active_pipe * ddb_size / config->num_pipes_active; - alloc->end = alloc->start + pipe_size - 1; + alloc->end = alloc->start + pipe_size; } static unsigned int skl_cursor_allocation(const struct intel_wm_config *config) @@ -3055,6 +3055,8 @@ static void skl_ddb_entry_init_from_hw(struct skl_ddb_entry *entry, u32 reg) { entry->start = reg & 0x3ff; entry->end = (reg >> 16) & 0x3ff; + if (entry->end) + entry->end += 1; } void skl_ddb_get_hw_state(struct drm_i915_private *dev_priv, @@ -3131,7 +3133,7 @@ skl_allocate_pipe_ddb(struct drm_crtc *crtc, } cursor_blocks = skl_cursor_allocation(config); - ddb->cursor[pipe].start = alloc.end - cursor_blocks + 1; + ddb->cursor[pipe].start = alloc.end - cursor_blocks; ddb->cursor[pipe].end = alloc.end; alloc_size -= cursor_blocks; @@ -3165,7 +3167,7 @@ skl_allocate_pipe_ddb(struct drm_crtc *crtc, total_data_rate); ddb->plane[pipe][plane].start = start; - ddb->plane[pipe][plane].end = start + plane_blocks - 1; + ddb->plane[pipe][plane].end = start + plane_blocks; start += plane_blocks; } @@ -3453,6 +3455,15 @@ static void skl_compute_wm_results(struct drm_device *dev, r->wm_linetime[pipe] = p_wm->linetime; } +static void skl_ddb_entry_write(struct drm_i915_private *dev_priv, uint32_t reg, + const struct skl_ddb_entry *entry) +{ + if (entry->end) + I915_WRITE(reg, (entry->end - 1) << 16 | entry->start); + else + I915_WRITE(reg, 0); +} + static void skl_write_wm_values(struct drm_i915_private *dev_priv, const struct skl_wm_values *new) { @@ -3480,13 +3491,12 @@ static void skl_write_wm_values(struct drm_i915_private *dev_priv, I915_WRITE(CUR_WM_TRANS(pipe), new->cursor_trans[pipe]); for (i = 0; i < intel_num_planes(crtc); i++) - I915_WRITE(PLANE_BUF_CFG(pipe, i), - new->ddb.plane[pipe][i].end << 16 | - new->ddb.plane[pipe][i].start); + skl_ddb_entry_write(dev_priv, + PLANE_BUF_CFG(pipe, i), + &new->ddb.plane[pipe][i]); - I915_WRITE(CUR_BUF_CFG(pipe), - new->ddb.cursor[pipe].end << 16 | - new->ddb.cursor[pipe].start); + skl_ddb_entry_write(dev_priv, CUR_BUF_CFG(pipe), + &new->ddb.cursor[pipe]); } } } -- cgit v0.10.2 From 21fca258bc09b951b66ff5c501a37b56be4d4310 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:54 +0000 Subject: drm/i915/skl: Use a more descriptive parameter name in skl_compute_plane_wm() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit What we're talking about here is the DDB allocation (in blocks). That's more descriptive than 'max_page_buff_alloc'. Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index e417a6a..0142ad1 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3315,7 +3315,7 @@ static void skl_compute_wm_pipe_parameters(struct drm_crtc *crtc, static bool skl_compute_plane_wm(struct skl_pipe_wm_parameters *p, struct intel_plane_wm_parameters *p_params, - uint16_t max_page_buff_alloc, + uint16_t ddb_allocation, uint32_t mem_value, uint16_t *res_blocks, /* out */ uint8_t *res_lines /* out */) @@ -3339,7 +3339,7 @@ static bool skl_compute_plane_wm(struct skl_pipe_wm_parameters *p, p_params->bytes_per_pixel; /* For now xtile and linear */ - if (((max_page_buff_alloc * 512) / plane_bytes_per_line) >= 1) + if (((ddb_allocation * 512) / plane_bytes_per_line) >= 1) result_bytes = min(method1, method2); else result_bytes = method1; -- cgit v0.10.2 From e6d6617152def4f9d46fb9c5cba3e322bedc27c2 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:55 +0000 Subject: drm/i915/skl: Make res_blocks/lines intermediate values 32 bits MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit To align with the ilk WM code and because it makes sense to test against the upper bounds as soon as possible on variables that are bigger than the number of bits in the register, let's move the maximum checks from skl_compute_wm_results() to skl_compute_plane_wm(). v2: Leave the result values to 0 when overflowing the limits (Ville) Use 32 bits intermediate variables (Damien) Instead of using the 16 and 8 bits space we have in the result structure, use 32 bits local variables until we're sure they fit into the constraints. Suggested-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 0142ad1..8580298 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3317,10 +3317,10 @@ static bool skl_compute_plane_wm(struct skl_pipe_wm_parameters *p, struct intel_plane_wm_parameters *p_params, uint16_t ddb_allocation, uint32_t mem_value, - uint16_t *res_blocks, /* out */ - uint8_t *res_lines /* out */) + uint16_t *out_blocks, /* out */ + uint8_t *out_lines /* out */) { - uint32_t method1, method2, plane_bytes_per_line; + uint32_t method1, method2, plane_bytes_per_line, res_blocks, res_lines; uint32_t result_bytes; if (mem_value == 0 || !p->active || !p_params->enabled) @@ -3344,8 +3344,14 @@ static bool skl_compute_plane_wm(struct skl_pipe_wm_parameters *p, else result_bytes = method1; - *res_blocks = DIV_ROUND_UP(result_bytes, 512) + 1; - *res_lines = DIV_ROUND_UP(result_bytes, plane_bytes_per_line); + res_blocks = DIV_ROUND_UP(result_bytes, 512) + 1; + res_lines = DIV_ROUND_UP(result_bytes, plane_bytes_per_line); + + if (res_blocks > ddb_allocation || res_lines > 31) + return false; + + *out_blocks = res_blocks; + *out_lines = res_lines; return true; } @@ -3408,17 +3414,11 @@ static void skl_compute_wm_results(struct drm_device *dev, enum pipe pipe = intel_crtc->pipe; for (level = 0; level <= max_level; level++) { - uint16_t ddb_blocks; uint32_t temp; int i; for (i = 0; i < intel_num_planes(intel_crtc); i++) { temp = 0; - ddb_blocks = skl_ddb_entry_size(&r->ddb.plane[pipe][i]); - - if ((p_wm->wm[level].plane_res_b[i] > ddb_blocks) || - (p_wm->wm[level].plane_res_l[i] > 31)) - p_wm->wm[level].plane_en[i] = false; temp |= p_wm->wm[level].plane_res_l[i] << PLANE_WM_LINES_SHIFT; @@ -3433,11 +3433,6 @@ static void skl_compute_wm_results(struct drm_device *dev, } temp = 0; - ddb_blocks = skl_ddb_entry_size(&r->ddb.cursor[pipe]); - - if ((p_wm->wm[level].cursor_res_b > ddb_blocks) || - (p_wm->wm[level].cursor_res_l > 31)) - p_wm->wm[level].cursor_en = false; temp |= p_wm->wm[level].cursor_res_l << PLANE_WM_LINES_SHIFT; temp |= p_wm->wm[level].cursor_res_b; -- cgit v0.10.2 From b99f58dabb9dae18c97a9e3812ba9961b207fd45 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:56 +0000 Subject: drm/i915/skl: Reduce the number of holes in struct skl_wm_level MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 84e8cdb..cf84b6d 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1422,9 +1422,9 @@ struct skl_wm_values { struct skl_wm_level { bool plane_en[I915_MAX_PLANES]; + bool cursor_en; uint16_t plane_res_b[I915_MAX_PLANES]; uint8_t plane_res_l[I915_MAX_PLANES]; - bool cursor_en; uint16_t cursor_res_b; uint8_t cursor_res_l; }; -- cgit v0.10.2 From 407b50f31ba19317cdc30530d30220182e42fa3d Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:57 +0000 Subject: drm/i915/skl: Move all the WM compute functions in one place MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The DDB allocation code managed to split in two the compute functions. Bring back skl_compute_transition_wm() and skl_compute_linetime_wm() with their little friends. Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 8580298..66f142b 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3217,28 +3217,6 @@ static uint32_t skl_wm_method2(uint32_t pixel_rate, uint32_t pipe_htotal, return ret; } -static void skl_compute_transition_wm(struct drm_crtc *crtc, - struct skl_pipe_wm_parameters *params, - struct skl_pipe_wm *pipe_wm) -{ - /* - * For now it is suggested to use the LP0 wm val of corresponding - * plane as transition wm val. This is done while computing results. - */ - if (!params->active) - return; -} - -static uint32_t -skl_compute_linetime_wm(struct drm_crtc *crtc, struct skl_pipe_wm_parameters *p) -{ - if (!intel_crtc_active(crtc)) - return 0; - - return DIV_ROUND_UP(8 * p->pipe_htotal * 1000, p->pixel_rate); - -} - static bool skl_ddb_allocation_changed(const struct skl_ddb_allocation *new_ddb, const struct intel_crtc *intel_crtc) { @@ -3384,6 +3362,28 @@ static void skl_compute_wm_level(const struct drm_i915_private *dev_priv, &result->cursor_res_l); } +static uint32_t +skl_compute_linetime_wm(struct drm_crtc *crtc, struct skl_pipe_wm_parameters *p) +{ + if (!intel_crtc_active(crtc)) + return 0; + + return DIV_ROUND_UP(8 * p->pipe_htotal * 1000, p->pixel_rate); + +} + +static void skl_compute_transition_wm(struct drm_crtc *crtc, + struct skl_pipe_wm_parameters *params, + struct skl_pipe_wm *pipe_wm) +{ + /* + * For now it is suggested to use the LP0 wm val of corresponding + * plane as transition wm val. + */ + if (!params->active) + return; +} + static void skl_compute_pipe_wm(struct drm_crtc *crtc, struct skl_ddb_allocation *ddb, struct skl_pipe_wm_parameters *params, -- cgit v0.10.2 From 9414f563f3857018d87e480b7cd13636b0c00ce6 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:58 +0000 Subject: drm/i915/skl: Rework when the transition WMs are computed MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The transition WMs code was doing a shortcut and the values were copied from the WM0 ones at compute_wm_results() time. Going forward, we want to compute them like the other WMs and resolve their final register values in the same way as well. This patch does just that and isolate the transtion WM compute code in skl_compute_transition_wm() while skl_compute_wm_results() takes care of the register values. We also take the opportunity to disable the transition WMs for now. We've noticed underruns and they seem to be the culprit. Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 66f142b..41cfb26 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3374,14 +3374,18 @@ skl_compute_linetime_wm(struct drm_crtc *crtc, struct skl_pipe_wm_parameters *p) static void skl_compute_transition_wm(struct drm_crtc *crtc, struct skl_pipe_wm_parameters *params, - struct skl_pipe_wm *pipe_wm) + struct skl_wm_level *trans_wm /* out */) { - /* - * For now it is suggested to use the LP0 wm val of corresponding - * plane as transition wm val. - */ + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + int i; + if (!params->active) return; + + /* Until we know more, just disable transition WMs */ + for (i = 0; i < intel_num_planes(intel_crtc); i++) + trans_wm->plane_en[i] = false; + trans_wm->cursor_en = false; } static void skl_compute_pipe_wm(struct drm_crtc *crtc, @@ -3401,7 +3405,7 @@ static void skl_compute_pipe_wm(struct drm_crtc *crtc, } pipe_wm->linetime = skl_compute_linetime_wm(crtc, params); - skl_compute_transition_wm(crtc, params, pipe_wm); + skl_compute_transition_wm(crtc, params, &pipe_wm->trans_wm); } static void skl_compute_wm_results(struct drm_device *dev, @@ -3412,11 +3416,10 @@ static void skl_compute_wm_results(struct drm_device *dev, { int level, max_level = ilk_wm_max_level(dev); enum pipe pipe = intel_crtc->pipe; + uint32_t temp; + int i; for (level = 0; level <= max_level; level++) { - uint32_t temp; - int i; - for (i = 0; i < intel_num_planes(intel_crtc); i++) { temp = 0; @@ -3427,9 +3430,6 @@ static void skl_compute_wm_results(struct drm_device *dev, temp |= PLANE_WM_EN; r->plane[pipe][i][level] = temp; - /* Use the LP0 WM value for transition WM for now. */ - if (level == 0) - r->plane_trans[pipe][i] = temp; } temp = 0; @@ -3441,12 +3441,28 @@ static void skl_compute_wm_results(struct drm_device *dev, temp |= PLANE_WM_EN; r->cursor[pipe][level] = temp; - /* Use the LP0 WM value for transition WM for now. */ - if (level == 0) - r->cursor_trans[pipe] = temp; } + /* transition WMs */ + for (i = 0; i < intel_num_planes(intel_crtc); i++) { + temp = 0; + temp |= p_wm->trans_wm.plane_res_l[i] << PLANE_WM_LINES_SHIFT; + temp |= p_wm->trans_wm.plane_res_b[i]; + if (p_wm->trans_wm.plane_en[i]) + temp |= PLANE_WM_EN; + + r->plane_trans[pipe][i] = temp; + } + + temp = 0; + temp |= p_wm->trans_wm.cursor_res_l << PLANE_WM_LINES_SHIFT; + temp |= p_wm->trans_wm.cursor_res_b; + if (p_wm->trans_wm.cursor_en) + temp |= PLANE_WM_EN; + + r->cursor_trans[pipe] = temp; + r->wm_linetime[pipe] = p_wm->linetime; } -- cgit v0.10.2 From afb024aa656b7d13392430a254b341728806c540 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:06:59 +0000 Subject: drm/i915/skl: Correctly align skl_compute_plane_wm() arguments MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 41cfb26..176e9ef 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3292,11 +3292,11 @@ static void skl_compute_wm_pipe_parameters(struct drm_crtc *crtc, } static bool skl_compute_plane_wm(struct skl_pipe_wm_parameters *p, - struct intel_plane_wm_parameters *p_params, - uint16_t ddb_allocation, - uint32_t mem_value, - uint16_t *out_blocks, /* out */ - uint8_t *out_lines /* out */) + struct intel_plane_wm_parameters *p_params, + uint16_t ddb_allocation, + uint32_t mem_value, + uint16_t *out_blocks, /* out */ + uint8_t *out_lines /* out */) { uint32_t method1, method2, plane_bytes_per_line, res_blocks, res_lines; uint32_t result_bytes; -- cgit v0.10.2 From 5d374d9638924ff8c91a87f55dc895be7ab163f8 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:07:00 +0000 Subject: drm/i915/skl: Reduce the indentation level in skl_write_wm_values() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We can reduce the indentation level by continuing early. Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 176e9ef..7fdccba 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3485,30 +3485,30 @@ static void skl_write_wm_values(struct drm_i915_private *dev_priv, int i, level, max_level = ilk_wm_max_level(dev); enum pipe pipe = crtc->pipe; - if (new->dirty[pipe]) { - I915_WRITE(PIPE_WM_LINETIME(pipe), - new->wm_linetime[pipe]); - - for (level = 0; level <= max_level; level++) { - for (i = 0; i < intel_num_planes(crtc); i++) - I915_WRITE(PLANE_WM(pipe, i, level), - new->plane[pipe][i][level]); - I915_WRITE(CUR_WM(pipe, level), - new->cursor[pipe][level]); - } - for (i = 0; i < intel_num_planes(crtc); i++) - I915_WRITE(PLANE_WM_TRANS(pipe, i), - new->plane_trans[pipe][i]); - I915_WRITE(CUR_WM_TRANS(pipe), new->cursor_trans[pipe]); + if (!new->dirty[pipe]) + continue; - for (i = 0; i < intel_num_planes(crtc); i++) - skl_ddb_entry_write(dev_priv, - PLANE_BUF_CFG(pipe, i), - &new->ddb.plane[pipe][i]); + I915_WRITE(PIPE_WM_LINETIME(pipe), new->wm_linetime[pipe]); - skl_ddb_entry_write(dev_priv, CUR_BUF_CFG(pipe), - &new->ddb.cursor[pipe]); + for (level = 0; level <= max_level; level++) { + for (i = 0; i < intel_num_planes(crtc); i++) + I915_WRITE(PLANE_WM(pipe, i, level), + new->plane[pipe][i][level]); + I915_WRITE(CUR_WM(pipe, level), + new->cursor[pipe][level]); } + for (i = 0; i < intel_num_planes(crtc); i++) + I915_WRITE(PLANE_WM_TRANS(pipe, i), + new->plane_trans[pipe][i]); + I915_WRITE(CUR_WM_TRANS(pipe), new->cursor_trans[pipe]); + + for (i = 0; i < intel_num_planes(crtc); i++) + skl_ddb_entry_write(dev_priv, + PLANE_BUF_CFG(pipe, i), + &new->ddb.plane[pipe][i]); + + skl_ddb_entry_write(dev_priv, CUR_BUF_CFG(pipe), + &new->ddb.cursor[pipe]); } } -- cgit v0.10.2 From 34bb56af7fae9d4d6a33bf6f5f915e75a4023e02 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:07:01 +0000 Subject: drm/i915/skl: Stage the pipe DDB allocation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit To correctly flush the new DDB allocation we need to know about the pipe allocation layout inside the DDB in order to sequence the re-allocation to not cause a newly allocated pipe to fetch from a space that was previously allocated to another pipe. This patch preserves the per-pipe (start,end) allocation to be used in the flush. Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index cf84b6d..01d741d 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1406,6 +1406,7 @@ static inline bool skl_ddb_entry_equal(const struct skl_ddb_entry *e1, } struct skl_ddb_allocation { + struct skl_ddb_entry pipe[I915_MAX_PIPES]; struct skl_ddb_entry plane[I915_MAX_PIPES][I915_MAX_PLANES]; struct skl_ddb_entry cursor[I915_MAX_PIPES]; }; diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7fdccba..cd2b335 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3119,13 +3119,13 @@ skl_allocate_pipe_ddb(struct drm_crtc *crtc, struct drm_device *dev = crtc->dev; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); enum pipe pipe = intel_crtc->pipe; - struct skl_ddb_entry alloc; + struct skl_ddb_entry *alloc = &ddb->pipe[pipe]; uint16_t alloc_size, start, cursor_blocks; unsigned int total_data_rate; int plane; - skl_ddb_get_pipe_allocation_limits(dev, crtc, config, params, &alloc); - alloc_size = skl_ddb_entry_size(&alloc); + skl_ddb_get_pipe_allocation_limits(dev, crtc, config, params, alloc); + alloc_size = skl_ddb_entry_size(alloc); if (alloc_size == 0) { memset(ddb->plane[pipe], 0, sizeof(ddb->plane[pipe])); memset(&ddb->cursor[pipe], 0, sizeof(ddb->cursor[pipe])); @@ -3133,11 +3133,11 @@ skl_allocate_pipe_ddb(struct drm_crtc *crtc, } cursor_blocks = skl_cursor_allocation(config); - ddb->cursor[pipe].start = alloc.end - cursor_blocks; - ddb->cursor[pipe].end = alloc.end; + ddb->cursor[pipe].start = alloc->end - cursor_blocks; + ddb->cursor[pipe].end = alloc->end; alloc_size -= cursor_blocks; - alloc.end -= cursor_blocks; + alloc->end -= cursor_blocks; /* * Each active plane get a portion of the remaining space, in @@ -3147,7 +3147,7 @@ skl_allocate_pipe_ddb(struct drm_crtc *crtc, */ total_data_rate = skl_get_total_relative_data_rate(intel_crtc, params); - start = alloc.start; + start = alloc->start; for (plane = 0; plane < intel_num_planes(intel_crtc); plane++) { const struct intel_plane_wm_parameters *p; unsigned int data_rate; -- cgit v0.10.2 From 0e8fb7ba7ca5410b2a8adc7f19eeb229ed01a32e Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:07:02 +0000 Subject: drm/i915/skl: Flush the WM configuration MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When we write new values for the DDB allocation and WM parameters, we now need to trigger the double buffer update for the pipe to take the new configuration into account. As the DDB is a global resource shared between planes, enabling or disabling one plane will result in changes for all planes that are currently in use, thus the need write PLANE_SURF/CUR_BASE for more than the plane we're touching. v2: Don't wait for pipes that are off v3: Split the staging results structure to not exceed the 1Kb stack allocation in skl_update_wm() v4: Rework and document the algorithm after Ville found that it was all wrong. Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index cd2b335..ec7b86b 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3512,6 +3512,140 @@ static void skl_write_wm_values(struct drm_i915_private *dev_priv, } } +/* + * When setting up a new DDB allocation arrangement, we need to correctly + * sequence the times at which the new allocations for the pipes are taken into + * account or we'll have pipes fetching from space previously allocated to + * another pipe. + * + * Roughly the sequence looks like: + * 1. re-allocate the pipe(s) with the allocation being reduced and not + * overlapping with a previous light-up pipe (another way to put it is: + * pipes with their new allocation strickly included into their old ones). + * 2. re-allocate the other pipes that get their allocation reduced + * 3. allocate the pipes having their allocation increased + * + * Steps 1. and 2. are here to take care of the following case: + * - Initially DDB looks like this: + * | B | C | + * - enable pipe A. + * - pipe B has a reduced DDB allocation that overlaps with the old pipe C + * allocation + * | A | B | C | + * + * We need to sequence the re-allocation: C, B, A (and not B, C, A). + */ + +static void skl_wm_flush_pipe(struct drm_i915_private *dev_priv, enum pipe pipe) +{ + struct drm_device *dev = dev_priv->dev; + int plane; + + for_each_plane(pipe, plane) { + I915_WRITE(PLANE_SURF(pipe, plane), + I915_READ(PLANE_SURF(pipe, plane))); + } + I915_WRITE(CURBASE(pipe), I915_READ(CURBASE(pipe))); +} + +static bool +skl_ddb_allocation_included(const struct skl_ddb_allocation *old, + const struct skl_ddb_allocation *new, + enum pipe pipe) +{ + uint16_t old_size, new_size; + + old_size = skl_ddb_entry_size(&old->pipe[pipe]); + new_size = skl_ddb_entry_size(&new->pipe[pipe]); + + return old_size != new_size && + new->pipe[pipe].start >= old->pipe[pipe].start && + new->pipe[pipe].end <= old->pipe[pipe].end; +} + +static void skl_flush_wm_values(struct drm_i915_private *dev_priv, + struct skl_wm_values *new_values) +{ + struct drm_device *dev = dev_priv->dev; + struct skl_ddb_allocation *cur_ddb, *new_ddb; + bool reallocated[I915_MAX_PIPES] = {false, false, false}; + struct intel_crtc *crtc; + enum pipe pipe; + + new_ddb = &new_values->ddb; + cur_ddb = &dev_priv->wm.skl_hw.ddb; + + /* + * First pass: flush the pipes with the new allocation contained into + * the old space. + * + * We'll wait for the vblank on those pipes to ensure we can safely + * re-allocate the freed space without this pipe fetching from it. + */ + for_each_intel_crtc(dev, crtc) { + if (!crtc->active) + continue; + + pipe = crtc->pipe; + + if (!skl_ddb_allocation_included(cur_ddb, new_ddb, pipe)) + continue; + + skl_wm_flush_pipe(dev_priv, pipe); + intel_wait_for_vblank(dev, pipe); + + reallocated[pipe] = true; + } + + + /* + * Second pass: flush the pipes that are having their allocation + * reduced, but overlapping with a previous allocation. + * + * Here as well we need to wait for the vblank to make sure the freed + * space is not used anymore. + */ + for_each_intel_crtc(dev, crtc) { + if (!crtc->active) + continue; + + pipe = crtc->pipe; + + if (reallocated[pipe]) + continue; + + if (skl_ddb_entry_size(&new_ddb->pipe[pipe]) < + skl_ddb_entry_size(&cur_ddb->pipe[pipe])) { + skl_wm_flush_pipe(dev_priv, pipe); + intel_wait_for_vblank(dev, pipe); + } + + reallocated[pipe] = true; + } + + /* + * Third pass: flush the pipes that got more space allocated. + * + * We don't need to actively wait for the update here, next vblank + * will just get more DDB space with the correct WM values. + */ + for_each_intel_crtc(dev, crtc) { + if (!crtc->active) + continue; + + pipe = crtc->pipe; + + /* + * At this point, only the pipes more space than before are + * left to re-allocate. + */ + if (reallocated[pipe]) + continue; + + skl_wm_flush_pipe(dev_priv, pipe); + } +} + static bool skl_update_pipe_wm(struct drm_crtc *crtc, struct skl_pipe_wm_parameters *params, struct intel_wm_config *config, @@ -3603,6 +3737,7 @@ static void skl_update_wm(struct drm_crtc *crtc) skl_update_other_pipe_wm(dev, crtc, &config, results); skl_write_wm_values(dev_priv, results); + skl_flush_wm_values(dev_priv, results); /* store the new configuration */ dev_priv->wm.skl_hw = *results; -- cgit v0.10.2 From d21b795c419a7f59b4921b1ab5ff40d9a5eef05a Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Tue, 4 Nov 2014 17:07:03 +0000 Subject: drm/i915/skl: Log the order in which we flush the pipes in the WM code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Reviewed-by: Ville Syrjälä Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index ec7b86b..d0dbd65 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3536,11 +3536,14 @@ static void skl_write_wm_values(struct drm_i915_private *dev_priv, * We need to sequence the re-allocation: C, B, A (and not B, C, A). */ -static void skl_wm_flush_pipe(struct drm_i915_private *dev_priv, enum pipe pipe) +static void +skl_wm_flush_pipe(struct drm_i915_private *dev_priv, enum pipe pipe, int pass) { struct drm_device *dev = dev_priv->dev; int plane; + DRM_DEBUG_KMS("flush pipe %c (pass %d)\n", pipe_name(pipe), pass); + for_each_plane(pipe, plane) { I915_WRITE(PLANE_SURF(pipe, plane), I915_READ(PLANE_SURF(pipe, plane))); @@ -3591,7 +3594,7 @@ static void skl_flush_wm_values(struct drm_i915_private *dev_priv, if (!skl_ddb_allocation_included(cur_ddb, new_ddb, pipe)) continue; - skl_wm_flush_pipe(dev_priv, pipe); + skl_wm_flush_pipe(dev_priv, pipe, 1); intel_wait_for_vblank(dev, pipe); reallocated[pipe] = true; @@ -3616,7 +3619,7 @@ static void skl_flush_wm_values(struct drm_i915_private *dev_priv, if (skl_ddb_entry_size(&new_ddb->pipe[pipe]) < skl_ddb_entry_size(&cur_ddb->pipe[pipe])) { - skl_wm_flush_pipe(dev_priv, pipe); + skl_wm_flush_pipe(dev_priv, pipe, 2); intel_wait_for_vblank(dev, pipe); } @@ -3642,7 +3645,7 @@ static void skl_flush_wm_values(struct drm_i915_private *dev_priv, if (reallocated[pipe]) continue; - skl_wm_flush_pipe(dev_priv, pipe); + skl_wm_flush_pipe(dev_priv, pipe, 3); } } -- cgit v0.10.2 From 38cff0b15756dae3c7452618e6673f77d87e7220 Mon Sep 17 00:00:00 2001 From: Zhe Wang Date: Tue, 4 Nov 2014 17:07:04 +0000 Subject: drm/i915/skl: Gen9 Forcewake Implement common forcewake functions shared by Gen9 features. v2: Make the focewake_{get,put} functions static (Mika) Small coding style fix in the function definition (Damien) Reviewed-by: Mika Kuoppala Signed-off-by: Zhe Wang (v1) Signed-off-by: Damien Lespiau (v2) Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 01d741d..0f00e58 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -549,6 +549,7 @@ struct intel_uncore { unsigned fw_rendercount; unsigned fw_mediacount; + unsigned fw_blittercount; struct timer_list force_wake_timer; }; @@ -2979,7 +2980,9 @@ int vlv_freq_opcode(struct drm_i915_private *dev_priv, int val); #define FORCEWAKE_RENDER (1 << 0) #define FORCEWAKE_MEDIA (1 << 1) -#define FORCEWAKE_ALL (FORCEWAKE_RENDER | FORCEWAKE_MEDIA) +#define FORCEWAKE_BLITTER (1 << 2) +#define FORCEWAKE_ALL (FORCEWAKE_RENDER | FORCEWAKE_MEDIA | \ + FORCEWAKE_BLITTER) #define I915_READ8(reg) dev_priv->uncore.funcs.mmio_readb(dev_priv, (reg), true) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index ea22e09..77461ce 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5833,6 +5833,12 @@ enum punit_power_well { #define VLV_GTLC_PW_MEDIA_STATUS_MASK (1 << 5) #define VLV_GTLC_PW_RENDER_STATUS_MASK (1 << 7) #define FORCEWAKE_MT 0xa188 /* multi-threaded */ +#define FORCEWAKE_MEDIA_GEN9 0xa270 +#define FORCEWAKE_RENDER_GEN9 0xa278 +#define FORCEWAKE_BLITTER_GEN9 0xa188 +#define FORCEWAKE_ACK_MEDIA_GEN9 0x0D88 +#define FORCEWAKE_ACK_RENDER_GEN9 0x0D84 +#define FORCEWAKE_ACK_BLITTER_GEN9 0x130044 #define FORCEWAKE_KERNEL 0x1 #define FORCEWAKE_USER 0x2 #define FORCEWAKE_MT_ACK 0x130040 diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 9427641..68e722b 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -296,6 +296,154 @@ static void vlv_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine) spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } +static void __gen9_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv) +{ + __raw_i915_write32(dev_priv, FORCEWAKE_RENDER_GEN9, + _MASKED_BIT_DISABLE(0xffff)); + + __raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_GEN9, + _MASKED_BIT_DISABLE(0xffff)); + + __raw_i915_write32(dev_priv, FORCEWAKE_BLITTER_GEN9, + _MASKED_BIT_DISABLE(0xffff)); +} + +static void +__gen9_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) +{ + /* Check for Render Engine */ + if (FORCEWAKE_RENDER & fw_engine) { + if (wait_for_atomic((__raw_i915_read32(dev_priv, + FORCEWAKE_ACK_RENDER_GEN9) & + FORCEWAKE_KERNEL) == 0, + FORCEWAKE_ACK_TIMEOUT_MS)) + DRM_ERROR("Timed out: Render forcewake old ack to clear.\n"); + + __raw_i915_write32(dev_priv, FORCEWAKE_RENDER_GEN9, + _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL)); + + if (wait_for_atomic((__raw_i915_read32(dev_priv, + FORCEWAKE_ACK_RENDER_GEN9) & + FORCEWAKE_KERNEL), + FORCEWAKE_ACK_TIMEOUT_MS)) + DRM_ERROR("Timed out: waiting for Render to ack.\n"); + } + + /* Check for Media Engine */ + if (FORCEWAKE_MEDIA & fw_engine) { + if (wait_for_atomic((__raw_i915_read32(dev_priv, + FORCEWAKE_ACK_MEDIA_GEN9) & + FORCEWAKE_KERNEL) == 0, + FORCEWAKE_ACK_TIMEOUT_MS)) + DRM_ERROR("Timed out: Media forcewake old ack to clear.\n"); + + __raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_GEN9, + _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL)); + + if (wait_for_atomic((__raw_i915_read32(dev_priv, + FORCEWAKE_ACK_MEDIA_GEN9) & + FORCEWAKE_KERNEL), + FORCEWAKE_ACK_TIMEOUT_MS)) + DRM_ERROR("Timed out: waiting for Media to ack.\n"); + } + + /* Check for Blitter Engine */ + if (FORCEWAKE_BLITTER & fw_engine) { + if (wait_for_atomic((__raw_i915_read32(dev_priv, + FORCEWAKE_ACK_BLITTER_GEN9) & + FORCEWAKE_KERNEL) == 0, + FORCEWAKE_ACK_TIMEOUT_MS)) + DRM_ERROR("Timed out: Blitter forcewake old ack to clear.\n"); + + __raw_i915_write32(dev_priv, FORCEWAKE_BLITTER_GEN9, + _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL)); + + if (wait_for_atomic((__raw_i915_read32(dev_priv, + FORCEWAKE_ACK_BLITTER_GEN9) & + FORCEWAKE_KERNEL), + FORCEWAKE_ACK_TIMEOUT_MS)) + DRM_ERROR("Timed out: waiting for Blitter to ack.\n"); + } +} + +static void +__gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine) +{ + /* Check for Render Engine */ + if (FORCEWAKE_RENDER & fw_engine) + __raw_i915_write32(dev_priv, FORCEWAKE_RENDER_GEN9, + _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL)); + + /* Check for Media Engine */ + if (FORCEWAKE_MEDIA & fw_engine) + __raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_GEN9, + _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL)); + + /* Check for Blitter Engine */ + if (FORCEWAKE_BLITTER & fw_engine) + __raw_i915_write32(dev_priv, FORCEWAKE_BLITTER_GEN9, + _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL)); +} + +static void +gen9_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) +{ + unsigned long irqflags; + + spin_lock_irqsave(&dev_priv->uncore.lock, irqflags); + + if (FORCEWAKE_RENDER & fw_engine) { + if (dev_priv->uncore.fw_rendercount++ == 0) + dev_priv->uncore.funcs.force_wake_get(dev_priv, + FORCEWAKE_RENDER); + } + + if (FORCEWAKE_MEDIA & fw_engine) { + if (dev_priv->uncore.fw_mediacount++ == 0) + dev_priv->uncore.funcs.force_wake_get(dev_priv, + FORCEWAKE_MEDIA); + } + + if (FORCEWAKE_BLITTER & fw_engine) { + if (dev_priv->uncore.fw_blittercount++ == 0) + dev_priv->uncore.funcs.force_wake_get(dev_priv, + FORCEWAKE_BLITTER); + } + + spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); +} + +static void +gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine) +{ + unsigned long irqflags; + + spin_lock_irqsave(&dev_priv->uncore.lock, irqflags); + + if (FORCEWAKE_RENDER & fw_engine) { + WARN_ON(dev_priv->uncore.fw_rendercount == 0); + if (--dev_priv->uncore.fw_rendercount == 0) + dev_priv->uncore.funcs.force_wake_put(dev_priv, + FORCEWAKE_RENDER); + } + + if (FORCEWAKE_MEDIA & fw_engine) { + WARN_ON(dev_priv->uncore.fw_mediacount == 0); + if (--dev_priv->uncore.fw_mediacount == 0) + dev_priv->uncore.funcs.force_wake_put(dev_priv, + FORCEWAKE_MEDIA); + } + + if (FORCEWAKE_BLITTER & fw_engine) { + WARN_ON(dev_priv->uncore.fw_blittercount == 0); + if (--dev_priv->uncore.fw_blittercount == 0) + dev_priv->uncore.funcs.force_wake_put(dev_priv, + FORCEWAKE_BLITTER); + } + + spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); +} + static void gen6_force_wake_timer(unsigned long arg) { struct drm_i915_private *dev_priv = (void *)arg; @@ -334,6 +482,9 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore) if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev) || IS_BROADWELL(dev)) __gen7_gt_force_wake_mt_reset(dev_priv); + if (IS_GEN9(dev)) + __gen9_gt_force_wake_mt_reset(dev_priv); + if (restore) { /* If reset with a user forcewake, try to restore */ unsigned fw = 0; @@ -343,6 +494,15 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore) if (dev_priv->uncore.fw_mediacount) fw |= FORCEWAKE_MEDIA; + } else if (IS_GEN9(dev)) { + if (dev_priv->uncore.fw_rendercount) + fw |= FORCEWAKE_RENDER; + + if (dev_priv->uncore.fw_mediacount) + fw |= FORCEWAKE_MEDIA; + + if (dev_priv->uncore.fw_blittercount) + fw |= FORCEWAKE_BLITTER; } else { if (dev_priv->uncore.forcewake_count) fw = FORCEWAKE_ALL; @@ -414,6 +574,10 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) intel_runtime_pm_get(dev_priv); + /* Redirect to Gen9 specific routine */ + if (IS_GEN9(dev_priv->dev)) + return gen9_force_wake_get(dev_priv, fw_engine); + /* Redirect to VLV specific routine */ if (IS_VALLEYVIEW(dev_priv->dev)) return vlv_force_wake_get(dev_priv, fw_engine); @@ -435,6 +599,12 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine) if (!dev_priv->uncore.funcs.force_wake_put) return; + /* Redirect to Gen9 specific routine */ + if (IS_GEN9(dev_priv->dev)) { + gen9_force_wake_put(dev_priv, fw_engine); + goto out; + } + /* Redirect to VLV specific routine */ if (IS_VALLEYVIEW(dev_priv->dev)) { vlv_force_wake_put(dev_priv, fw_engine); @@ -855,7 +1025,10 @@ void intel_uncore_init(struct drm_device *dev) __intel_uncore_early_sanitize(dev, false); - if (IS_VALLEYVIEW(dev)) { + if (IS_GEN9(dev)) { + dev_priv->uncore.funcs.force_wake_get = __gen9_force_wake_get; + dev_priv->uncore.funcs.force_wake_put = __gen9_force_wake_put; + } else if (IS_VALLEYVIEW(dev)) { dev_priv->uncore.funcs.force_wake_get = __vlv_force_wake_get; dev_priv->uncore.funcs.force_wake_put = __vlv_force_wake_put; } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { -- cgit v0.10.2 From 20e4936693ea8a727e35d0f2b840670a0576b706 Mon Sep 17 00:00:00 2001 From: Zhe Wang Date: Tue, 4 Nov 2014 17:07:05 +0000 Subject: drm/i915/skl: Enable Gen9 RC6 Configure and enable RC6 for Gen9. v2: Rebase on top of BDW rc6 support (Damien) Reviewed-by: Mika Kuoppala Signed-off-by: Zhe Wang Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index d0dbd65..300d7e5 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4539,6 +4539,13 @@ static void gen8_disable_rps_interrupts(struct drm_device *dev) I915_WRITE(GEN8_GT_IIR(2), dev_priv->pm_rps_events); } +static void gen9_disable_rps(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_WRITE(GEN6_RC_CONTROL, 0); +} + static void gen6_disable_rps_interrupts(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -4699,6 +4706,45 @@ static void parse_rp_state_cap(struct drm_i915_private *dev_priv, u32 rp_state_c dev_priv->rps.min_freq_softlimit = dev_priv->rps.min_freq; } +static void gen9_enable_rps(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *ring; + uint32_t rc6_mask = 0; + int unused; + + /* 1a: Software RC state - RC0 */ + I915_WRITE(GEN6_RC_STATE, 0); + + /* 1b: Get forcewake during program sequence. Although the driver + * hasn't enabled a state yet where we need forcewake, BIOS may have.*/ + gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL); + + /* 2a: Disable RC states. */ + I915_WRITE(GEN6_RC_CONTROL, 0); + + /* 2b: Program RC6 thresholds.*/ + I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16); + I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + for_each_ring(ring, dev_priv, unused) + I915_WRITE(RING_MAX_IDLE(ring->mmio_base), 10); + I915_WRITE(GEN6_RC_SLEEP, 0); + I915_WRITE(GEN6_RC6_THRESHOLD, 37500); /* 37.5/125ms per EI */ + + /* 3a: Enable RC6 */ + if (intel_enable_rc6(dev) & INTEL_RC6_ENABLE) + rc6_mask = GEN6_RC_CTL_RC6_ENABLE; + DRM_INFO("RC6 %s\n", (rc6_mask & GEN6_RC_CTL_RC6_ENABLE) ? + "on" : "off"); + I915_WRITE(GEN6_RC_CONTROL, GEN6_RC_CTL_HW_ENABLE | + GEN6_RC_CTL_EI_MODE(1) | + rc6_mask); + + gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL); + +} + static void gen8_enable_rps(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -6233,7 +6279,9 @@ void intel_disable_gt_powersave(struct drm_device *dev) intel_suspend_gt_powersave(dev); mutex_lock(&dev_priv->rps.hw_lock); - if (IS_CHERRYVIEW(dev)) + if (INTEL_INFO(dev)->gen >= 9) + gen9_disable_rps(dev); + else if (IS_CHERRYVIEW(dev)) cherryview_disable_rps(dev); else if (IS_VALLEYVIEW(dev)) valleyview_disable_rps(dev); @@ -6257,6 +6305,8 @@ static void intel_gen6_powersave_work(struct work_struct *work) cherryview_enable_rps(dev); } else if (IS_VALLEYVIEW(dev)) { valleyview_enable_rps(dev); + } else if (INTEL_INFO(dev)->gen >= 9) { + gen9_enable_rps(dev); } else if (IS_BROADWELL(dev)) { gen8_enable_rps(dev); __gen6_update_ring_freq(dev); -- cgit v0.10.2 From c6bde93b922586245ba1083c35b1139d6498eabb Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Tue, 4 Nov 2014 10:31:28 +0200 Subject: drm/i915/audio: rewrite vlv/chv and gen 5-7 audio codec enable sequence Similar to the hsw/bdw enable sequence rewrite. v3: replace vblank wait with a comment v4: expand the comment on what should be done with the vblank wait Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 4dcfc46..8759fc5 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -215,6 +215,10 @@ static void ilk_audio_codec_enable(struct drm_connector *connector, { struct drm_i915_private *dev_priv = connector->dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); + struct intel_digital_port *intel_dig_port = + enc_to_dig_port(&encoder->base); + enum port port = intel_dig_port->port; + enum pipe pipe = intel_crtc->pipe; uint8_t *eld = connector->eld; uint32_t eldv; uint32_t tmp; @@ -223,8 +227,16 @@ static void ilk_audio_codec_enable(struct drm_connector *connector, int aud_config; int aud_cntl_st; int aud_cntrl_st2; - enum pipe pipe = intel_crtc->pipe; - enum port port; + + DRM_DEBUG_KMS("Enable audio codec on port %c, pipe %c, %u bytes ELD\n", + port_name(port), pipe_name(pipe), eld[2]); + + /* + * FIXME: We're supposed to wait for vblank here, but we have vblanks + * disabled during the mode set. The proper fix would be to push the + * rest of the setup into a vblank work item, queued here, but the + * infrastructure is not there yet. + */ if (HAS_PCH_IBX(connector->dev)) { hdmiw_hdmiedid = IBX_HDMIW_HDMIEDID(pipe); @@ -243,57 +255,44 @@ static void ilk_audio_codec_enable(struct drm_connector *connector, aud_cntrl_st2 = CPT_AUD_CNTRL_ST2; } - DRM_DEBUG_DRIVER("ELD on pipe %c\n", pipe_name(pipe)); - - if (IS_VALLEYVIEW(connector->dev)) { - struct intel_digital_port *intel_dig_port; - - intel_dig_port = enc_to_dig_port(&encoder->base); - port = intel_dig_port->port; - } else { - tmp = I915_READ(aud_cntl_st); - port = (tmp >> 29) & DIP_PORT_SEL_MASK; - /* DIP_Port_Select, 0x1 = PortB */ - } - - if (!port) { - DRM_DEBUG_DRIVER("Audio directed to unknown port\n"); - /* operate blindly on all ports */ + if (WARN_ON(!port)) { eldv = IBX_ELD_VALIDB; eldv |= IBX_ELD_VALIDB << 4; eldv |= IBX_ELD_VALIDB << 8; } else { - DRM_DEBUG_DRIVER("ELD on port %c\n", port_name(port)); eldv = IBX_ELD_VALIDB << ((port - 1) * 4); } - if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) - I915_WRITE(aud_config, AUD_CONFIG_N_VALUE_INDEX); /* 0x1 = DP */ - else - I915_WRITE(aud_config, audio_config_hdmi_pixel_clock(mode)); - - if (intel_eld_uptodate(connector, - aud_cntrl_st2, eldv, - aud_cntl_st, IBX_ELD_ADDRESS_MASK, - hdmiw_hdmiedid)) - return; - + /* Invalidate ELD */ tmp = I915_READ(aud_cntrl_st2); tmp &= ~eldv; I915_WRITE(aud_cntrl_st2, tmp); + /* Reset ELD write address */ tmp = I915_READ(aud_cntl_st); tmp &= ~IBX_ELD_ADDRESS_MASK; I915_WRITE(aud_cntl_st, tmp); - len = min_t(int, eld[2], 21); /* 84 bytes of hw ELD buffer */ - DRM_DEBUG_DRIVER("ELD size %d\n", len); + /* Up to 84 bytes of hw ELD buffer */ + len = min_t(int, eld[2], 21); for (i = 0; i < len; i++) I915_WRITE(hdmiw_hdmiedid, *((uint32_t *)eld + i)); + /* ELD valid */ tmp = I915_READ(aud_cntrl_st2); tmp |= eldv; I915_WRITE(aud_cntrl_st2, tmp); + + /* Enable timestamps */ + tmp = I915_READ(aud_config); + tmp &= ~AUD_CONFIG_N_VALUE_INDEX; + tmp &= ~AUD_CONFIG_N_PROG_ENABLE; + tmp &= ~AUD_CONFIG_PIXEL_CLOCK_HDMI_MASK; + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) + tmp |= AUD_CONFIG_N_VALUE_INDEX; + else + tmp |= audio_config_hdmi_pixel_clock(mode); + I915_WRITE(aud_config, tmp); } /** -- cgit v0.10.2 From 495a5bb81d3469495d81dbec6290dbb1dfebd673 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:55 +0200 Subject: drm/i915/audio: add vlv/chv/gen5-7 audio codec disable sequence Add support for disabling the audio codec on vlv/chv/gen5-7, similar to hsw/bdw. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 8759fc5..5c2d06c 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -209,6 +209,56 @@ static void hsw_audio_codec_enable(struct drm_connector *connector, I915_WRITE(HSW_AUD_CFG(pipe), tmp); } +static void ilk_audio_codec_disable(struct intel_encoder *encoder) +{ + struct drm_i915_private *dev_priv = encoder->base.dev->dev_private; + struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); + struct intel_digital_port *intel_dig_port = + enc_to_dig_port(&encoder->base); + enum port port = intel_dig_port->port; + enum pipe pipe = intel_crtc->pipe; + uint32_t tmp, eldv; + int aud_config; + int aud_cntrl_st2; + + DRM_DEBUG_KMS("Disable audio codec on port %c, pipe %c\n", + port_name(port), pipe_name(pipe)); + + if (HAS_PCH_IBX(dev_priv->dev)) { + aud_config = IBX_AUD_CFG(pipe); + aud_cntrl_st2 = IBX_AUD_CNTL_ST2; + } else if (IS_VALLEYVIEW(dev_priv)) { + aud_config = VLV_AUD_CFG(pipe); + aud_cntrl_st2 = VLV_AUD_CNTL_ST2; + } else { + aud_config = CPT_AUD_CFG(pipe); + aud_cntrl_st2 = CPT_AUD_CNTRL_ST2; + } + + /* Disable timestamps */ + tmp = I915_READ(aud_config); + tmp &= ~AUD_CONFIG_N_VALUE_INDEX; + tmp |= AUD_CONFIG_N_PROG_ENABLE; + tmp &= ~AUD_CONFIG_UPPER_N_MASK; + tmp &= ~AUD_CONFIG_LOWER_N_MASK; + if (intel_pipe_has_type(intel_crtc, INTEL_OUTPUT_DISPLAYPORT)) + tmp |= AUD_CONFIG_N_VALUE_INDEX; + I915_WRITE(aud_config, tmp); + + if (WARN_ON(!port)) { + eldv = IBX_ELD_VALIDB; + eldv |= IBX_ELD_VALIDB << 4; + eldv |= IBX_ELD_VALIDB << 8; + } else { + eldv = IBX_ELD_VALIDB << ((port - 1) * 4); + } + + /* Invalidate ELD */ + tmp = I915_READ(aud_cntrl_st2); + tmp &= ~eldv; + I915_WRITE(aud_cntrl_st2, tmp); +} + static void ilk_audio_codec_enable(struct drm_connector *connector, struct intel_encoder *encoder, struct drm_display_mode *mode) @@ -360,10 +410,12 @@ void intel_init_audio(struct drm_device *dev) dev_priv->display.audio_codec_enable = g4x_audio_codec_enable; } else if (IS_VALLEYVIEW(dev)) { dev_priv->display.audio_codec_enable = ilk_audio_codec_enable; + dev_priv->display.audio_codec_disable = ilk_audio_codec_disable; } else if (IS_HASWELL(dev) || INTEL_INFO(dev)->gen >= 8) { dev_priv->display.audio_codec_enable = hsw_audio_codec_enable; dev_priv->display.audio_codec_disable = hsw_audio_codec_disable; } else if (HAS_PCH_SPLIT(dev)) { dev_priv->display.audio_codec_enable = ilk_audio_codec_enable; + dev_priv->display.audio_codec_disable = ilk_audio_codec_disable; } } diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 5f2c206..f48a1aae 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2454,6 +2454,10 @@ static void intel_disable_dp(struct intel_encoder *encoder) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct drm_device *dev = encoder->base.dev; + struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); + + if (crtc->config.has_audio) + intel_audio_codec_disable(encoder); /* Make sure the panel is off before trying to change the mode. But also * ensure that we have vdd while we switch off the panel. */ diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index f29026a..f89f71e 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -802,9 +802,13 @@ static void intel_disable_hdmi(struct intel_encoder *encoder) struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; struct intel_hdmi *intel_hdmi = enc_to_intel_hdmi(&encoder->base); + struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); u32 temp; u32 enable_bits = SDVO_ENABLE | SDVO_AUDIO_ENABLE; + if (crtc->config.has_audio) + intel_audio_codec_disable(encoder); + temp = I915_READ(intel_hdmi->hdmi_reg); /* HW workaround for IBX, we need to move the port to transcoder A -- cgit v0.10.2 From c1dec79aaee69042986247658ffdf51ee81017c2 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:56 +0200 Subject: drm/i915: enable audio codec after port As per spec, and similar to DDI. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index f48a1aae..464eea0 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1324,12 +1324,8 @@ static void intel_dp_prepare(struct intel_encoder *encoder) intel_dp->DP |= DP_VOLTAGE_0_4 | DP_PRE_EMPHASIS_0; intel_dp->DP |= DP_PORT_WIDTH(intel_dp->lane_count); - if (crtc->config.has_audio) { - DRM_DEBUG_DRIVER("Enabling DP audio on pipe %c\n", - pipe_name(crtc->pipe)); + if (crtc->config.has_audio) intel_dp->DP |= DP_AUDIO_OUTPUT_ENABLE; - intel_audio_codec_enable(encoder); - } /* Split out the IBX/CPU vs CPT settings */ @@ -2636,6 +2632,7 @@ static void intel_enable_dp(struct intel_encoder *encoder) struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); uint32_t dp_reg = I915_READ(intel_dp->output_reg); if (WARN_ON(dp_reg & DP_PORT_EN)) @@ -2661,6 +2658,12 @@ static void intel_enable_dp(struct intel_encoder *encoder) intel_dp_start_link_train(intel_dp); intel_dp_complete_link_train(intel_dp); intel_dp_stop_link_train(intel_dp); + + if (crtc->config.has_audio) { + DRM_DEBUG_DRIVER("Enabling DP audio on pipe %c\n", + pipe_name(crtc->pipe)); + intel_audio_codec_enable(encoder); + } } static void g4x_enable_dp(struct intel_encoder *encoder) diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index f89f71e..29baa53 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -661,14 +661,6 @@ static void intel_hdmi_prepare(struct intel_encoder *encoder) if (crtc->config.has_hdmi_sink) hdmi_val |= HDMI_MODE_SELECT_HDMI; - if (crtc->config.has_audio) { - WARN_ON(!crtc->config.has_hdmi_sink); - DRM_DEBUG_DRIVER("Enabling HDMI audio on pipe %c\n", - pipe_name(crtc->pipe)); - hdmi_val |= SDVO_AUDIO_ENABLE; - intel_audio_codec_enable(encoder); - } - if (HAS_PCH_CPT(dev)) hdmi_val |= SDVO_PIPE_SEL_CPT(crtc->pipe); else if (IS_CHERRYVIEW(dev)) @@ -791,6 +783,13 @@ static void intel_enable_hdmi(struct intel_encoder *encoder) I915_WRITE(intel_hdmi->hdmi_reg, temp); POSTING_READ(intel_hdmi->hdmi_reg); } + + if (intel_crtc->config.has_audio) { + WARN_ON(!intel_crtc->config.has_hdmi_sink); + DRM_DEBUG_DRIVER("Enabling HDMI audio on pipe %c\n", + pipe_name(intel_crtc->pipe)); + intel_audio_codec_enable(encoder); + } } static void vlv_enable_hdmi(struct intel_encoder *encoder) -- cgit v0.10.2 From 76d8d3e5b57d7938613e4018eddbf0a16913338a Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:57 +0200 Subject: drm/i915/audio: add audio codec disable on g4x This not based on any documentation... Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 5c2d06c..fa0eecf 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -93,6 +93,25 @@ static bool intel_eld_uptodate(struct drm_connector *connector, return true; } +static void g4x_audio_codec_disable(struct intel_encoder *encoder) +{ + struct drm_i915_private *dev_priv = encoder->base.dev->dev_private; + uint32_t eldv, tmp; + + DRM_DEBUG_KMS("Disable audio codec\n"); + + tmp = I915_READ(G4X_AUD_VID_DID); + if (tmp == INTEL_AUDIO_DEVBLC || tmp == INTEL_AUDIO_DEVCL) + eldv = G4X_ELDV_DEVCL_DEVBLC; + else + eldv = G4X_ELDV_DEVCTG; + + /* Invalidate ELD */ + tmp = I915_READ(G4X_AUD_CNTL_ST); + tmp &= ~eldv; + I915_WRITE(G4X_AUD_CNTL_ST, tmp); +} + static void g4x_audio_codec_enable(struct drm_connector *connector, struct intel_encoder *encoder, struct drm_display_mode *mode) @@ -408,6 +427,7 @@ void intel_init_audio(struct drm_device *dev) if (IS_G4X(dev)) { dev_priv->display.audio_codec_enable = g4x_audio_codec_enable; + dev_priv->display.audio_codec_disable = g4x_audio_codec_disable; } else if (IS_VALLEYVIEW(dev)) { dev_priv->display.audio_codec_enable = ilk_audio_codec_enable; dev_priv->display.audio_codec_disable = ilk_audio_codec_disable; -- cgit v0.10.2 From d5ee08de1bbeac867818c37b5b8413f68c742703 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:58 +0200 Subject: drm/i915/audio: add audio codec enable debug log for g4x Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index fa0eecf..c4c036a 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -122,6 +122,8 @@ static void g4x_audio_codec_enable(struct drm_connector *connector, uint32_t tmp; int len, i; + DRM_DEBUG_KMS("Enable audio codec, %u bytes ELD\n", eld[2]); + tmp = I915_READ(G4X_AUD_VID_DID); if (tmp == INTEL_AUDIO_DEVBLC || tmp == INTEL_AUDIO_DEVCL) eldv = G4X_ELDV_DEVCL_DEVBLC; -- cgit v0.10.2 From 82910ac6d575d4ce4bd4dcefb61923623b263e00 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:26:59 +0200 Subject: drm/i915: make pipe/port based audio valid accessors easier to use Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77461ce..b250d96 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -6102,8 +6102,8 @@ enum punit_power_well { #define IBX_ELD_ADDRESS_MASK (0x1f << 5) #define IBX_ELD_ACK (1 << 4) #define IBX_AUD_CNTL_ST2 0xE20C0 -#define IBX_ELD_VALIDB (1 << 0) -#define IBX_CP_READYB (1 << 1) +#define IBX_CP_READY(port) ((1 << 1) << (((port) - 1) * 4)) +#define IBX_ELD_VALID(port) ((1 << 0) << (((port) - 1) * 4)) #define _CPT_HDMIW_HDMIEDID_A 0xE5050 #define _CPT_HDMIW_HDMIEDID_B 0xE5150 @@ -6206,18 +6206,10 @@ enum punit_power_well { #define HSW_AUD_PIPE_CONV_CFG 0x6507c #define HSW_AUD_PIN_ELD_CP_VLD 0x650c0 -#define AUDIO_INACTIVE_C (1<<11) -#define AUDIO_INACTIVE_B (1<<7) -#define AUDIO_INACTIVE_A (1<<3) -#define AUDIO_OUTPUT_ENABLE_A (1<<2) -#define AUDIO_OUTPUT_ENABLE_B (1<<6) -#define AUDIO_OUTPUT_ENABLE_C (1<<10) -#define AUDIO_ELD_VALID_A (1<<0) -#define AUDIO_ELD_VALID_B (1<<4) -#define AUDIO_ELD_VALID_C (1<<8) -#define AUDIO_CP_READY_A (1<<1) -#define AUDIO_CP_READY_B (1<<5) -#define AUDIO_CP_READY_C (1<<9) +#define AUDIO_INACTIVE(trans) ((1 << 3) << ((trans) * 4)) +#define AUDIO_OUTPUT_ENABLE(trans) ((1 << 2) << ((trans) * 4)) +#define AUDIO_CP_READY(trans) ((1 << 1) << ((trans) * 4)) +#define AUDIO_ELD_VALID(trans) ((1 << 0) << ((trans) * 4)) /* HSW Power Wells */ #define HSW_PWR_WELL_BIOS 0x45400 /* CTL1 */ diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index c4c036a..9181b85 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -172,7 +172,7 @@ static void hsw_audio_codec_disable(struct intel_encoder *encoder) /* Invalidate ELD */ tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); - tmp &= ~(AUDIO_ELD_VALID_A << (pipe * 4)); + tmp &= ~AUDIO_ELD_VALID(pipe); I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); } @@ -192,8 +192,8 @@ static void hsw_audio_codec_enable(struct drm_connector *connector, /* Enable audio presence detect, invalidate ELD */ tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); - tmp |= AUDIO_OUTPUT_ENABLE_A << (pipe * 4); - tmp &= ~(AUDIO_ELD_VALID_A << (pipe * 4)); + tmp |= AUDIO_OUTPUT_ENABLE(pipe); + tmp &= ~AUDIO_ELD_VALID(pipe); I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); /* @@ -215,7 +215,7 @@ static void hsw_audio_codec_enable(struct drm_connector *connector, /* ELD valid */ tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); - tmp |= AUDIO_ELD_VALID_A << (pipe * 4); + tmp |= AUDIO_ELD_VALID(pipe); I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); /* Enable timestamps */ @@ -267,11 +267,10 @@ static void ilk_audio_codec_disable(struct intel_encoder *encoder) I915_WRITE(aud_config, tmp); if (WARN_ON(!port)) { - eldv = IBX_ELD_VALIDB; - eldv |= IBX_ELD_VALIDB << 4; - eldv |= IBX_ELD_VALIDB << 8; + eldv = IBX_ELD_VALID(PORT_B) | IBX_ELD_VALID(PORT_C) | + IBX_ELD_VALID(PORT_D); } else { - eldv = IBX_ELD_VALIDB << ((port - 1) * 4); + eldv = IBX_ELD_VALID(port); } /* Invalidate ELD */ @@ -327,11 +326,10 @@ static void ilk_audio_codec_enable(struct drm_connector *connector, } if (WARN_ON(!port)) { - eldv = IBX_ELD_VALIDB; - eldv |= IBX_ELD_VALIDB << 4; - eldv |= IBX_ELD_VALIDB << 8; + eldv = IBX_ELD_VALID(PORT_B) | IBX_ELD_VALID(PORT_C) | + IBX_ELD_VALID(PORT_D); } else { - eldv = IBX_ELD_VALIDB << ((port - 1) * 4); + eldv = IBX_ELD_VALID(port); } /* Invalidate ELD */ diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index a50df35..68703ce 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1523,7 +1523,7 @@ void intel_ddi_get_config(struct intel_encoder *encoder, if (intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_AUDIO)) { temp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); - if (temp & (AUDIO_OUTPUT_ENABLE_A << (intel_crtc->pipe * 4))) + if (temp & AUDIO_OUTPUT_ENABLE(intel_crtc->pipe)) pipe_config->has_audio = true; } -- cgit v0.10.2 From 28855d2ac349b3fbb285c38493429dd1aa503523 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Mon, 27 Oct 2014 16:27:00 +0200 Subject: drm/i915/audio: add DOC comment describing HDA over HDMI/DP v2: include the section in the drm docbook. Signed-off-by: Jani Nikula Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index f6a9d7b..01b8ca5 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -3855,6 +3855,11 @@ int num_ioctls; + High Definition Audio +!Pdrivers/gpu/drm/i915/intel_audio.c High Definition Audio over HDMI and Display Port +!Idrivers/gpu/drm/i915/intel_audio.c + + DPIO !Pdrivers/gpu/drm/i915/i915_reg.h DPIO diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 9181b85..44c49df 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -28,6 +28,27 @@ #include "intel_drv.h" #include "i915_drv.h" +/** + * DOC: High Definition Audio over HDMI and Display Port + * + * The graphics and audio drivers together support High Definition Audio over + * HDMI and Display Port. The audio programming sequences are divided into audio + * codec and controller enable and disable sequences. The graphics driver + * handles the audio codec sequences, while the audio driver handles the audio + * controller sequences. + * + * The disable sequences must be performed before disabling the transcoder or + * port. The enable sequences may only be performed after enabling the + * transcoder and port, and after completed link training. + * + * The codec and controller sequences could be done either parallel or serial, + * but generally the ELDV/PD change in the codec sequence indicates to the audio + * driver that the controller sequence should start. Indeed, most of the + * co-operation between the graphics and audio drivers is handled via audio + * related registers. (The notable exception is the power management, not + * covered here.) + */ + static const struct { int clock; u32 config; -- cgit v0.10.2 From 50f6e502718fd187f8e11a7b5048b21f85ac77e4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 6 Nov 2014 14:49:12 +0200 Subject: drm/i915: Move the .global_resources() hook call into modeset_update_crtc_power_domains() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We may need to access various hardware bits in the .global_resources() hook, so move the call to occur after enabling all the newly required power wells, but before disabling all the now unneeded wells. This should guarantee that we have all the sufficient hardware resources available during the .global_resources() call. And if not, any additional resources must be explicitly acquired by the .global_resorces() hook. For instance on VLV/CHV we need to access the gunit mailbox so that we can talk to punit/cck over sideband. In addition some PFI credit reprogramming may need to be addes as well, which may require the disp2d well. This should also make the power domain refcounts consistent on platforms which don't have a .global_resource() hook since now they too will call modeset_update_crtc_power_domains() which will drop the init power. Previously init power was just left enabled for such platforms. Cc: Imre Deak Signed-off-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 07fb0ff..18ad2bc 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -4683,6 +4683,9 @@ static void modeset_update_crtc_power_domains(struct drm_device *dev) intel_display_power_get(dev_priv, domain); } + if (dev_priv->display.modeset_global_resources) + dev_priv->display.modeset_global_resources(dev); + for_each_intel_crtc(dev, crtc) { enum intel_display_power_domain domain; @@ -4910,8 +4913,6 @@ static void valleyview_modeset_global_resources(struct drm_device *dev) else valleyview_set_cdclk(dev, req_cdclk); } - - modeset_update_crtc_power_domains(dev); } static void valleyview_crtc_enable(struct drm_crtc *crtc) @@ -7950,16 +7951,6 @@ void hsw_disable_pc8(struct drm_i915_private *dev_priv) intel_prepare_ddi(dev); } -static void snb_modeset_global_resources(struct drm_device *dev) -{ - modeset_update_crtc_power_domains(dev); -} - -static void haswell_modeset_global_resources(struct drm_device *dev) -{ - modeset_update_crtc_power_domains(dev); -} - static int haswell_crtc_compute_clock(struct intel_crtc *crtc) { if (!intel_ddi_pll_select(crtc)) @@ -10917,8 +10908,7 @@ static int __intel_set_mode(struct drm_crtc *crtc, * update the the output configuration. */ intel_modeset_update_state(dev, prepare_pipes); - if (dev_priv->display.modeset_global_resources) - dev_priv->display.modeset_global_resources(dev); + modeset_update_crtc_power_domains(dev); /* Set up the DPLL and any encoders state that needs to adjust or depend * on the DPLL. @@ -12589,8 +12579,6 @@ static void intel_init_display(struct drm_device *dev) dev_priv->display.fdi_link_train = ironlake_fdi_link_train; } else if (IS_GEN6(dev)) { dev_priv->display.fdi_link_train = gen6_fdi_link_train; - dev_priv->display.modeset_global_resources = - snb_modeset_global_resources; } else if (IS_IVYBRIDGE(dev)) { /* FIXME: detect B0+ stepping and use auto training */ dev_priv->display.fdi_link_train = ivb_manual_fdi_link_train; @@ -12598,14 +12586,9 @@ static void intel_init_display(struct drm_device *dev) ivb_modeset_global_resources; } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { dev_priv->display.fdi_link_train = hsw_fdi_link_train; - dev_priv->display.modeset_global_resources = - haswell_modeset_global_resources; } else if (IS_VALLEYVIEW(dev)) { dev_priv->display.modeset_global_resources = valleyview_modeset_global_resources; - } else if (INTEL_INFO(dev)->gen >= 9) { - dev_priv->display.modeset_global_resources = - haswell_modeset_global_resources; } /* Default just returns -ENODEV to indicate unsupported */ -- cgit v0.10.2 From 16e9a21f33cda9d48b70ff54e1ed2e4a77497bc6 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Thu, 6 Nov 2014 09:26:38 +0200 Subject: drm/i915: Make __wait_seqno non-static and rename to __i915_wait_seqno So that it can be used by the flip code to wait for rendering without holding any locks. Signed-off-by: Ander Conselvan de Oliveira Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 0f00e58..2dc888c 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2595,6 +2595,11 @@ int __i915_add_request(struct intel_engine_cs *ring, u32 *seqno); #define i915_add_request(ring, seqno) \ __i915_add_request(ring, NULL, NULL, seqno) +int __i915_wait_seqno(struct intel_engine_cs *ring, u32 seqno, + unsigned reset_counter, + bool interruptible, + s64 *timeout, + struct drm_i915_file_private *file_priv); int __must_check i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno); int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 1de94cc..3e0cabe 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1134,7 +1134,7 @@ static bool can_wait_boost(struct drm_i915_file_private *file_priv) } /** - * __wait_seqno - wait until execution of seqno has finished + * __i915_wait_seqno - wait until execution of seqno has finished * @ring: the ring expected to report seqno * @seqno: duh! * @reset_counter: reset sequence associated with the given seqno @@ -1151,7 +1151,7 @@ static bool can_wait_boost(struct drm_i915_file_private *file_priv) * Returns 0 if the seqno was found within the alloted time. Else returns the * errno with remaining time filled in timeout argument. */ -static int __wait_seqno(struct intel_engine_cs *ring, u32 seqno, +int __i915_wait_seqno(struct intel_engine_cs *ring, u32 seqno, unsigned reset_counter, bool interruptible, s64 *timeout, @@ -1262,6 +1262,7 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno) struct drm_device *dev = ring->dev; struct drm_i915_private *dev_priv = dev->dev_private; bool interruptible = dev_priv->mm.interruptible; + unsigned reset_counter; int ret; BUG_ON(!mutex_is_locked(&dev->struct_mutex)); @@ -1275,9 +1276,9 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno) if (ret) return ret; - return __wait_seqno(ring, seqno, - atomic_read(&dev_priv->gpu_error.reset_counter), - interruptible, NULL, NULL); + reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter); + return __i915_wait_seqno(ring, seqno, reset_counter, interruptible, + NULL, NULL); } static int @@ -1353,7 +1354,8 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj, reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter); mutex_unlock(&dev->struct_mutex); - ret = __wait_seqno(ring, seqno, reset_counter, true, NULL, file_priv); + ret = __i915_wait_seqno(ring, seqno, reset_counter, true, NULL, + file_priv); mutex_lock(&dev->struct_mutex); if (ret) return ret; @@ -2847,8 +2849,8 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file) reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter); mutex_unlock(&dev->struct_mutex); - return __wait_seqno(ring, seqno, reset_counter, true, &args->timeout_ns, - file->driver_priv); + return __i915_wait_seqno(ring, seqno, reset_counter, true, + &args->timeout_ns, file->driver_priv); out: drm_gem_object_unreference(&obj->base); @@ -4013,7 +4015,7 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file) if (seqno == 0) return 0; - ret = __wait_seqno(ring, seqno, reset_counter, true, NULL, NULL); + ret = __i915_wait_seqno(ring, seqno, reset_counter, true, NULL, NULL); if (ret == 0) queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work, 0); -- cgit v0.10.2 From 536f5b5e86b225dab94c7ff8061ae482b6077387 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Thu, 6 Nov 2014 11:03:40 +0200 Subject: drm/i915: Make mmio flip wait for seqno in the work function This simplifies the code quite a bit compared to iterating over all rings during the ring interrupt. Also, it allows us to drop the mmio_flip spinlock, since the mmio_flip struct is only accessed in two places. The first is when the flip is queued and the other when the mmio writes are done. Since a flip cannot be queued while there is a pending flip, the two paths shouldn't ever run in parallel. We might need to revisit that if support for replacing flips is implemented though. v2: Don't hold dev->struct_mutext while waiting (Chris) v3: Make the wait uninterruptable (Chris) Signed-off-by: Ander Conselvan de Oliveira Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 318a6a0..5fff287 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -979,9 +979,6 @@ static void notify_ring(struct drm_device *dev, trace_i915_gem_request_complete(ring); - if (drm_core_check_feature(dev, DRIVER_MODESET)) - intel_notify_mmio_flip(ring); - wake_up_all(&ring->irq_queue); i915_queue_hangcheck(dev); } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 18ad2bc..cb0f32f 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9417,73 +9417,24 @@ static void intel_do_mmio_flip(struct intel_crtc *intel_crtc) if (atomic_update) intel_pipe_update_end(intel_crtc, start_vbl_count); - - spin_lock_irq(&dev_priv->mmio_flip_lock); - intel_crtc->mmio_flip.status = INTEL_MMIO_FLIP_IDLE; - spin_unlock_irq(&dev_priv->mmio_flip_lock); } static void intel_mmio_flip_work_func(struct work_struct *work) { struct intel_crtc *intel_crtc = container_of(work, struct intel_crtc, mmio_flip.work); - - intel_do_mmio_flip(intel_crtc); -} - -static int intel_postpone_flip(struct drm_i915_gem_object *obj) -{ struct intel_engine_cs *ring; - int ret; - - lockdep_assert_held(&obj->base.dev->struct_mutex); - - if (!obj->last_write_seqno) - return 0; - - ring = obj->ring; - - if (i915_seqno_passed(ring->get_seqno(ring, true), - obj->last_write_seqno)) - return 0; - - ret = i915_gem_check_olr(ring, obj->last_write_seqno); - if (ret) - return ret; - - if (WARN_ON(!ring->irq_get(ring))) - return 0; - - return 1; -} + uint32_t seqno; -void intel_notify_mmio_flip(struct intel_engine_cs *ring) -{ - struct drm_i915_private *dev_priv = to_i915(ring->dev); - struct intel_crtc *intel_crtc; - unsigned long irq_flags; - u32 seqno; - - seqno = ring->get_seqno(ring, false); + seqno = intel_crtc->mmio_flip.seqno; + ring = intel_crtc->mmio_flip.ring; - spin_lock_irqsave(&dev_priv->mmio_flip_lock, irq_flags); - for_each_intel_crtc(ring->dev, intel_crtc) { - struct intel_mmio_flip *mmio_flip; + if (seqno) + WARN_ON(__i915_wait_seqno(ring, seqno, + intel_crtc->reset_counter, + false, NULL, NULL) != 0); - mmio_flip = &intel_crtc->mmio_flip; - if (mmio_flip->status != INTEL_MMIO_FLIP_WAIT_RING) - continue; - - if (ring->id != mmio_flip->ring_id) - continue; - - if (i915_seqno_passed(seqno, mmio_flip->seqno)) { - schedule_work(&intel_crtc->mmio_flip.work); - mmio_flip->status = INTEL_MMIO_FLIP_WORK_SCHEDULED; - ring->irq_put(ring); - } - } - spin_unlock_irqrestore(&dev_priv->mmio_flip_lock, irq_flags); + intel_do_mmio_flip(intel_crtc); } static int intel_queue_mmio_flip(struct drm_device *dev, @@ -9493,32 +9444,13 @@ static int intel_queue_mmio_flip(struct drm_device *dev, struct intel_engine_cs *ring, uint32_t flags) { - struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int ret; - - if (WARN_ON(intel_crtc->mmio_flip.status != INTEL_MMIO_FLIP_IDLE)) - return -EBUSY; - - ret = intel_postpone_flip(obj); - if (ret < 0) - return ret; - if (ret == 0) { - intel_do_mmio_flip(intel_crtc); - return 0; - } - spin_lock_irq(&dev_priv->mmio_flip_lock); - intel_crtc->mmio_flip.status = INTEL_MMIO_FLIP_WAIT_RING; intel_crtc->mmio_flip.seqno = obj->last_write_seqno; - intel_crtc->mmio_flip.ring_id = obj->ring->id; - spin_unlock_irq(&dev_priv->mmio_flip_lock); + intel_crtc->mmio_flip.ring = obj->ring; + + schedule_work(&intel_crtc->mmio_flip.work); - /* - * Double check to catch cases where irq fired before - * mmio flip data was ready - */ - intel_notify_mmio_flip(obj->ring); return 0; } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index cb0e9db..5c622ad 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -399,16 +399,9 @@ struct intel_pipe_wm { bool sprites_scaled; }; -enum intel_mmio_flip_status { - INTEL_MMIO_FLIP_IDLE = 0, - INTEL_MMIO_FLIP_WAIT_RING, - INTEL_MMIO_FLIP_WORK_SCHEDULED, -}; - struct intel_mmio_flip { u32 seqno; - u32 ring_id; - enum intel_mmio_flip_status status; + struct intel_engine_cs *ring; struct work_struct work; }; -- cgit v0.10.2 From 9500986159d9017f9277f9e42a1f9b13b5e0e666 Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Wed, 5 Nov 2014 17:30:52 +0200 Subject: Revert "drm/i915/vlv: Remove check for Old Ack during forcewake" MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This reverts commit 5cb13c07dae73380d8b3ddc792740487b8742938. While the relevance for WaRsDontPollForAckOnClearingFWBits is under investigation, revert this as regression. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85684 Tested-by: Tested-by: lu hua Cc: Chris Wilson Cc: Ville Syrjälä Cc: S, Deepak Signed-off-by: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 68e722b..6a0c3fb 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -194,15 +194,13 @@ static void vlv_force_wake_reset(struct drm_i915_private *dev_priv) static void __vlv_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) { - /* - * WaRsDontPollForAckOnClearingFWBits:vlv - * Hardware clears ack bits lazily (only when all ack - * bits become 0) so don't poll for individiual ack - * bits to be clear here like on other platforms. - */ - /* Check for Render Engine */ if (FORCEWAKE_RENDER & fw_engine) { + if (wait_for_atomic((__raw_i915_read32(dev_priv, + FORCEWAKE_ACK_VLV) & + FORCEWAKE_KERNEL) == 0, + FORCEWAKE_ACK_TIMEOUT_MS)) + DRM_ERROR("Timed out: Render forcewake old ack to clear.\n"); __raw_i915_write32(dev_priv, FORCEWAKE_VLV, _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL)); @@ -216,6 +214,11 @@ static void __vlv_force_wake_get(struct drm_i915_private *dev_priv, /* Check for Media Engine */ if (FORCEWAKE_MEDIA & fw_engine) { + if (wait_for_atomic((__raw_i915_read32(dev_priv, + FORCEWAKE_ACK_MEDIA_VLV) & + FORCEWAKE_KERNEL) == 0, + FORCEWAKE_ACK_TIMEOUT_MS)) + DRM_ERROR("Timed out: Media forcewake old ack to clear.\n"); __raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_VLV, _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL)); -- cgit v0.10.2 From 6bcda4f0dfcbe78349b5b283af892913299c8b2a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 7 Oct 2014 17:41:22 +0300 Subject: drm/i915: Cache HPLL frequency on VLV/CHV MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We need the HPLL frequency when calculating cdclk. Currently we read that out from the hardware every single time, which isn't going to fly very well if the device is runtime suspended. So cache the HPLL frequency in dev_priv and use the cached value. Signed-off-by: Ville Syrjälä Reference: https://bugs.freedesktop.org/show_bug.cgi?id=82939 Reviewed-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 2dc888c..8fb8eba 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1613,6 +1613,7 @@ struct drm_i915_private { unsigned int fsb_freq, mem_freq, is_ddr3; unsigned int vlv_cdclk_freq; + unsigned int hpll_freq; /** * wq - Driver workqueue for GEM. diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index cb0f32f..3f4a5b8 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -4756,10 +4756,9 @@ static void valleyview_set_cdclk(struct drm_device *dev, int cdclk) mutex_unlock(&dev_priv->rps.hw_lock); if (cdclk == 400000) { - u32 divider, vco; + u32 divider; - vco = valleyview_get_vco(dev_priv); - divider = DIV_ROUND_CLOSEST(vco << 1, cdclk) - 1; + divider = DIV_ROUND_CLOSEST(dev_priv->hpll_freq << 1, cdclk) - 1; mutex_lock(&dev_priv->dpio_lock); /* adjust cdclk divider */ @@ -4838,8 +4837,7 @@ static void cherryview_set_cdclk(struct drm_device *dev, int cdclk) static int valleyview_calc_cdclk(struct drm_i915_private *dev_priv, int max_pixclk) { - int vco = valleyview_get_vco(dev_priv); - int freq_320 = (vco << 1) % 320000 != 0 ? 333333 : 320000; + int freq_320 = (dev_priv->hpll_freq << 1) % 320000 != 0 ? 333333 : 320000; /* FIXME: Punit isn't quite ready yet */ if (IS_CHERRYVIEW(dev_priv->dev)) @@ -5544,7 +5542,6 @@ static int intel_crtc_compute_config(struct intel_crtc *crtc, static int valleyview_get_display_clock_speed(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - int vco = valleyview_get_vco(dev_priv); u32 val; int divider; @@ -5552,6 +5549,9 @@ static int valleyview_get_display_clock_speed(struct drm_device *dev) if (IS_CHERRYVIEW(dev)) return 400000; + if (dev_priv->hpll_freq == 0) + dev_priv->hpll_freq = valleyview_get_vco(dev_priv); + mutex_lock(&dev_priv->dpio_lock); val = vlv_cck_read(dev_priv, CCK_DISPLAY_CLOCK_CONTROL); mutex_unlock(&dev_priv->dpio_lock); @@ -5562,7 +5562,7 @@ static int valleyview_get_display_clock_speed(struct drm_device *dev) (divider << DISPLAY_FREQUENCY_STATUS_SHIFT), "cdclk change in progress\n"); - return DIV_ROUND_CLOSEST(vco << 1, divider + 1); + return DIV_ROUND_CLOSEST(dev_priv->hpll_freq << 1, divider + 1); } static int i945_get_display_clock_speed(struct drm_device *dev) -- cgit v0.10.2 From 0b5492d6b53251acab99b1906a328fac56e08be3 Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Thu, 6 Nov 2014 13:03:46 +0200 Subject: drm/i915: Add gen to the gpu hang ecode for the Brothers in Triage Signed-off-by: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index d17360b..89a2f3d 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -1242,7 +1242,8 @@ static void i915_error_capture_msg(struct drm_device *dev, ecode = i915_error_generate_code(dev_priv, error, &ring_id); len = scnprintf(error->error_msg, sizeof(error->error_msg), - "GPU HANG: ecode %d:0x%08x", ring_id, ecode); + "GPU HANG: ecode %d:%d:0x%08x", + INTEL_INFO(dev)->gen, ring_id, ecode); if (ring_id != -1 && error->ring[ring_id].pid != -1) len += scnprintf(error->error_msg + len, -- cgit v0.10.2 From 69f627f56fac212c1e49cd0d285e3f8cd264dd0c Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 7 Nov 2014 19:03:19 +0100 Subject: drm/i915: Update DRIVER_DATE to 20141107 Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 8fb8eba..f830596 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -55,7 +55,7 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20141024" +#define DRIVER_DATE "20141107" #undef WARN_ON #define WARN_ON(x) WARN(x, "WARN_ON(" #x ")") -- cgit v0.10.2 From 4307d8f6e5105d7f4baa8cc4f0b56aca048eb024 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 20 Nov 2014 15:37:13 +0200 Subject: amdkfd: Fix sparse warnings in kfd_chardev.c Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 64c73ba..3b3fce7 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -149,7 +149,9 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties, } if ((args->ring_base_address) && - (!access_ok(VERIFY_WRITE, args->ring_base_address, sizeof(uint64_t)))) { + (!access_ok(VERIFY_WRITE, + (const void __user *) args->ring_base_address, + sizeof(uint64_t)))) { pr_err("kfd: can't access ring base address\n"); return -EFAULT; } @@ -159,12 +161,16 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties, return -EINVAL; } - if (!access_ok(VERIFY_WRITE, args->read_pointer_address, sizeof(uint32_t))) { + if (!access_ok(VERIFY_WRITE, + (const void __user *) args->read_pointer_address, + sizeof(uint32_t))) { pr_err("kfd: can't access read pointer\n"); return -EFAULT; } - if (!access_ok(VERIFY_WRITE, args->write_pointer_address, sizeof(uint32_t))) { + if (!access_ok(VERIFY_WRITE, + (const void __user *) args->write_pointer_address, + sizeof(uint32_t))) { pr_err("kfd: can't access write pointer\n"); return -EFAULT; } @@ -325,7 +331,9 @@ static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p, } if ((args.ring_base_address) && - (!access_ok(VERIFY_WRITE, args.ring_base_address, sizeof(uint64_t)))) { + (!access_ok(VERIFY_WRITE, + (const void __user *) args.ring_base_address, + sizeof(uint64_t)))) { pr_err("kfd: can't access ring base address\n"); return -EFAULT; } -- cgit v0.10.2 From 16b9201c620a83ebb4613161c974fd7926f22a31 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 20 Nov 2014 15:41:33 +0200 Subject: amdkfd: Fix sparse warnings in kfd_topology.c Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index 77cd7d5..5733e28 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -96,7 +96,7 @@ static int kfd_topology_get_crat_acpi(void *crat_image, size_t *size) return -EINVAL; } - if (*size >= crat_table->length && crat_image != 0) + if (*size >= crat_table->length && crat_image != NULL) memcpy(crat_image, crat_table, crat_table->length); *size = crat_table->length; @@ -183,7 +183,7 @@ static int kfd_parse_subtype_mem(struct crat_subtype_memory *mem) list_for_each_entry(dev, &topology_device_list, list) { if (mem->promixity_domain == i) { props = kfd_alloc_struct(props); - if (props == 0) + if (props == NULL) return -ENOMEM; if (dev->node_props.cpu_cores_count == 0) @@ -231,7 +231,7 @@ static int kfd_parse_subtype_cache(struct crat_subtype_cache *cache) if (id == dev->node_props.cpu_core_id_base || id == dev->node_props.simd_id_base) { props = kfd_alloc_struct(props); - if (props == 0) + if (props == NULL) return -ENOMEM; props->processor_id_low = id; @@ -282,7 +282,7 @@ static int kfd_parse_subtype_iolink(struct crat_subtype_iolink *iolink) list_for_each_entry(dev, &topology_device_list, list) { if (id_from == i) { props = kfd_alloc_struct(props); - if (props == 0) + if (props == NULL) return -ENOMEM; props->node_from = id_from; @@ -415,9 +415,9 @@ static struct kfd_topology_device *kfd_create_topology_device(void) struct kfd_topology_device *dev; dev = kfd_alloc_struct(dev); - if (dev == 0) { + if (dev == NULL) { pr_err("No memory to allocate a topology device"); - return 0; + return NULL; } INIT_LIST_HEAD(&dev->mem_props); @@ -428,7 +428,7 @@ static struct kfd_topology_device *kfd_create_topology_device(void) sys_props.num_devices++; return dev; - } +} static int kfd_parse_crat_table(void *crat_image) { @@ -752,11 +752,11 @@ static void kfd_remove_sysfs_node_entry(struct kfd_topology_device *dev) if (iolink->kobj) { kfd_remove_sysfs_file(iolink->kobj, &iolink->attr); - iolink->kobj = 0; + iolink->kobj = NULL; } kobject_del(dev->kobj_iolink); kobject_put(dev->kobj_iolink); - dev->kobj_iolink = 0; + dev->kobj_iolink = NULL; } if (dev->kobj_cache) { @@ -764,22 +764,22 @@ static void kfd_remove_sysfs_node_entry(struct kfd_topology_device *dev) if (cache->kobj) { kfd_remove_sysfs_file(cache->kobj, &cache->attr); - cache->kobj = 0; + cache->kobj = NULL; } kobject_del(dev->kobj_cache); kobject_put(dev->kobj_cache); - dev->kobj_cache = 0; + dev->kobj_cache = NULL; } if (dev->kobj_mem) { list_for_each_entry(mem, &dev->mem_props, list) if (mem->kobj) { kfd_remove_sysfs_file(mem->kobj, &mem->attr); - mem->kobj = 0; + mem->kobj = NULL; } kobject_del(dev->kobj_mem); kobject_put(dev->kobj_mem); - dev->kobj_mem = 0; + dev->kobj_mem = NULL; } if (dev->kobj_node) { @@ -788,7 +788,7 @@ static void kfd_remove_sysfs_node_entry(struct kfd_topology_device *dev) sysfs_remove_file(dev->kobj_node, &dev->attr_props); kobject_del(dev->kobj_node); kobject_put(dev->kobj_node); - dev->kobj_node = 0; + dev->kobj_node = NULL; } } @@ -939,7 +939,7 @@ static int kfd_topology_update_sysfs(void) int ret; pr_info("Creating topology SYSFS entries\n"); - if (sys_props.kobj_topology == 0) { + if (sys_props.kobj_topology == NULL) { sys_props.kobj_topology = kfd_alloc_struct(sys_props.kobj_topology); if (!sys_props.kobj_topology) @@ -989,17 +989,17 @@ static void kfd_topology_release_sysfs(void) if (sys_props.kobj_nodes) { kobject_del(sys_props.kobj_nodes); kobject_put(sys_props.kobj_nodes); - sys_props.kobj_nodes = 0; + sys_props.kobj_nodes = NULL; } kobject_del(sys_props.kobj_topology); kobject_put(sys_props.kobj_topology); - sys_props.kobj_topology = 0; + sys_props.kobj_topology = NULL; } } int kfd_topology_init(void) { - void *crat_image = 0; + void *crat_image = NULL; size_t image_size = 0; int ret; @@ -1094,12 +1094,12 @@ static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu) static struct kfd_topology_device *kfd_assign_gpu(struct kfd_dev *gpu) { struct kfd_topology_device *dev; - struct kfd_topology_device *out_dev = 0; + struct kfd_topology_device *out_dev = NULL; BUG_ON(!gpu); list_for_each_entry(dev, &topology_device_list, list) - if (dev->gpu == 0 && dev->node_props.simd_count > 0) { + if (dev->gpu == NULL && dev->node_props.simd_count > 0) { dev->gpu = gpu; out_dev = dev; break; -- cgit v0.10.2 From 5ef360eab7e6e634c22b5821468aae3af205d2fb Mon Sep 17 00:00:00 2001 From: kbuild test robot Date: Thu, 20 Nov 2014 16:08:14 +0800 Subject: amdkfd: test_kq() can be static Reviewed-by: Alex Deucher Signed-off-by: Fengguang Wu Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c index 555af45..424ddcc 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -321,7 +321,7 @@ void kernel_queue_uninit(struct kernel_queue *kq) kfree(kq); } -void test_kq(struct kfd_dev *dev) +static __attribute__((unused)) void test_kq(struct kfd_dev *dev) { struct kernel_queue *kq; uint32_t *buffer, i; -- cgit v0.10.2 From 7347a6cbf1b5672a87975e58c8b99d882b96e9b9 Mon Sep 17 00:00:00 2001 From: kbuild test robot Date: Thu, 20 Nov 2014 17:16:23 +0800 Subject: amdkfd: pqm_get_kernel_queue() can be static Reviewed-by: Alex Deucher Signed-off-by: Fengguang Wu Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index c7859fc..de2c163 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -325,7 +325,8 @@ int pqm_update_queue(struct process_queue_manager *pqm, unsigned int qid, return 0; } -struct kernel_queue *pqm_get_kernel_queue(struct process_queue_manager *pqm, +static __attribute__((unused)) struct kernel_queue *pqm_get_kernel_queue( + struct process_queue_manager *pqm, unsigned int qid) { struct process_queue_node *pqn; -- cgit v0.10.2 From 585dbf3842a37149410932f5fac2bfe68b9d86c4 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 20 Nov 2014 15:49:49 +0200 Subject: amdkfd: Fix sparse warnings in kfd_flat_memory.c Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c index 2dfc4c0..66df4da 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c @@ -276,21 +276,22 @@ */ #define MAKE_GPUVM_APP_BASE(gpu_num) \ - (((uint64_t)(gpu_num) << 61) + 0x1000000000000) + (((uint64_t)(gpu_num) << 61) + 0x1000000000000L) #define MAKE_GPUVM_APP_LIMIT(base) \ - (((uint64_t)(base) & 0xFFFFFF0000000000) | 0xFFFFFFFFFF) + (((uint64_t)(base) & \ + 0xFFFFFF0000000000UL) | 0xFFFFFFFFFFL) #define MAKE_SCRATCH_APP_BASE(gpu_num) \ - (((uint64_t)(gpu_num) << 61) + 0x100000000) + (((uint64_t)(gpu_num) << 61) + 0x100000000L) #define MAKE_SCRATCH_APP_LIMIT(base) \ - (((uint64_t)base & 0xFFFFFFFF00000000) | 0xFFFFFFFF) + (((uint64_t)base & 0xFFFFFFFF00000000UL) | 0xFFFFFFFF) #define MAKE_LDS_APP_BASE(gpu_num) \ (((uint64_t)(gpu_num) << 61) + 0x0) #define MAKE_LDS_APP_LIMIT(base) \ - (((uint64_t)(base) & 0xFFFFFFFF00000000) | 0xFFFFFFFF) + (((uint64_t)(base) & 0xFFFFFFFF00000000UL) | 0xFFFFFFFF) int kfd_init_apertures(struct kfd_process *process) { -- cgit v0.10.2 From 20981e68010c66187450b4a5a78b56b99d1f6f59 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 20 Nov 2014 15:50:53 +0200 Subject: amdkfd: is_occupied() can be static Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c index 59d2407..adc3147 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c @@ -179,9 +179,9 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd, pipe_id, queue_id); } -bool is_occupied(struct mqd_manager *mm, void *mqd, - uint64_t queue_address, uint32_t pipe_id, - uint32_t queue_id) +static bool is_occupied(struct mqd_manager *mm, void *mqd, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) { return kfd2kgd->hqd_is_occupies(mm->dev->kgd, queue_address, -- cgit v0.10.2 From d80d19bd5064692d40e9190ebadbd381f20d2969 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 20 Nov 2014 15:54:05 +0200 Subject: amdkfd: fence_wait_timeout() can be static Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 8c40d04..718f50e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -789,8 +789,9 @@ out: return retval; } -int fence_wait_timeout(unsigned int *fence_addr, unsigned int fence_value, - unsigned long timeout) +static int fence_wait_timeout(unsigned int *fence_addr, + unsigned int fence_value, + unsigned long timeout) { BUG_ON(!fence_addr); timeout += jiffies; -- cgit v0.10.2 From 5cd78de5267f36c090043836649b31c12533bbe9 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 20 Nov 2014 16:14:56 +0200 Subject: amdkfd: add __iomem attribute to doorbell_ptr This patch was done due to sparse warning. It changes the definition of doorbell_ptr in queue_properties to be with __iomem attribute, so it would match the type which the doorbell module functions are returning. Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c index 424ddcc..5055fc9 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -66,8 +66,7 @@ static bool initialize(struct kernel_queue *kq, struct kfd_dev *dev, if (kq->mqd == NULL) return false; - prop.doorbell_ptr = - (uint32_t *)kfd_get_kernel_doorbell(dev, &prop.doorbell_off); + prop.doorbell_ptr = kfd_get_kernel_doorbell(dev, &prop.doorbell_off); if (prop.doorbell_ptr == NULL) goto err_get_kernel_doorbell; @@ -172,7 +171,7 @@ err_rptr_allocate_vidmem: kfd2kgd->free_mem(dev->kgd, (struct kgd_mem *) kq->pq); err_pq_allocate_vidmem: pr_err("kfd: error init pq\n"); - kfd_release_kernel_doorbell(dev, (u32 *)prop.doorbell_ptr); + kfd_release_kernel_doorbell(dev, prop.doorbell_ptr); err_get_kernel_doorbell: pr_err("kfd: error init doorbell"); return false; @@ -195,7 +194,7 @@ static void uninitialize(struct kernel_queue *kq) kfd2kgd->free_mem(kq->dev->kgd, (struct kgd_mem *) kq->wptr_mem); kfd2kgd->free_mem(kq->dev->kgd, (struct kgd_mem *) kq->pq); kfd_release_kernel_doorbell(kq->dev, - (u32 *)kq->queue->properties.doorbell_ptr); + kq->queue->properties.doorbell_ptr); uninit_queue(kq->queue); } @@ -255,7 +254,7 @@ static void submit_packet(struct kernel_queue *kq) #endif *kq->wptr_kernel = kq->pending_wptr; - write_kernel_doorbell((u32 *)kq->queue->properties.doorbell_ptr, + write_kernel_doorbell(kq->queue->properties.doorbell_ptr, kq->pending_wptr); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 41e608d..d0bcafc 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -279,7 +279,7 @@ struct queue_properties { uint32_t queue_percent; uint32_t *read_ptr; uint32_t *write_ptr; - uint32_t *doorbell_ptr; + uint32_t __iomem *doorbell_ptr; uint32_t doorbell_off; bool is_interop; bool is_active; -- cgit v0.10.2 From f5d896bbd03a03bef436d9358599b50962095f38 Mon Sep 17 00:00:00 2001 From: Jay Cornwall Date: Thu, 20 Nov 2014 11:52:16 -0600 Subject: amdkfd: Fix memory leak on process deregistration struct device_process_node was allocated during process registration but not released at process deregistration. Reviewed-by: Alex Deucher Signed-off-by: Jay Cornwall Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 718f50e..bc8961c3 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -409,6 +409,7 @@ static int unregister_process_nocpsch(struct device_queue_manager *dqm, list_for_each_entry_safe(cur, next, &dqm->queues, list) { if (qpd == cur->qpd) { list_del(&cur->list); + kfree(cur); dqm->processes_count--; goto out; } -- cgit v0.10.2 From 9a5634a72934a566b3e5fdb0e8f345846162f7d3 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Mon, 17 Nov 2014 13:18:32 +0200 Subject: amdkfd: use schedule() in sync_with_hw amdkfd uses cpu_relax() in its sync_with_hw() function. Because cpu_relax() is defined as 'REP; NOP' on x86_64, it will block the CPU from servicing IOMMU PPR requests. This may cause a deadlock, because sync_with_hw() won't be completed until the PPR request has been served. Therefore, we need to use schedule() instead of cpu_relax() as it is the minimum requirement to allow other threads to execute. Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c index 5055fc9..9abac48 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -25,6 +25,7 @@ #include #include #include +#include #include "kfd_kernel_queue.h" #include "kfd_priv.h" #include "kfd_device_queue_manager.h" @@ -274,7 +275,7 @@ static int sync_with_hw(struct kernel_queue *kq, unsigned long timeout_ms) *kq->wptr_kernel, *kq->rptr_kernel); return -ETIME; } - cpu_relax(); + schedule(); } return 0; -- cgit v0.10.2 From 52a5fdce136bcaabc719be6e8f141d68426b3c74 Mon Sep 17 00:00:00 2001 From: Alexey Skidanov Date: Wed, 19 Nov 2014 17:07:00 +0200 Subject: amdkfd: Instead of using get function, use container_of Reviewed-by: Alex Deucher Signed-off-by: Alexey Skidanov Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index bc8961c3..904eb38 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -67,26 +67,21 @@ static inline unsigned int get_pipes_num_cpsch(void) return PIPE_PER_ME_CP_SCHEDULING; } -static unsigned int get_sh_mem_bases_nybble_64(struct kfd_process *process, - struct kfd_dev *dev) +static inline unsigned int +get_sh_mem_bases_nybble_64(struct kfd_process_device *pdd) { - struct kfd_process_device *pdd; uint32_t nybble; - pdd = kfd_get_process_device_data(dev, process, 1); nybble = (pdd->lds_base >> 60) & 0x0E; return nybble; } -static unsigned int get_sh_mem_bases_32(struct kfd_process *process, - struct kfd_dev *dev) +static inline unsigned int get_sh_mem_bases_32(struct kfd_process_device *pdd) { - struct kfd_process_device *pdd; unsigned int shared_base; - pdd = kfd_get_process_device_data(dev, process, 1); shared_base = (pdd->lds_base >> 16) & 0xFF; return shared_base; @@ -96,10 +91,13 @@ static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble); static void init_process_memory(struct device_queue_manager *dqm, struct qcm_process_device *qpd) { + struct kfd_process_device *pdd; unsigned int temp; BUG_ON(!dqm || !qpd); + pdd = qpd_to_pdd(qpd); + /* check if sh_mem_config register already configured */ if (qpd->sh_mem_config == 0) { qpd->sh_mem_config = @@ -111,11 +109,11 @@ static void init_process_memory(struct device_queue_manager *dqm, } if (qpd->pqm->process->is_32bit_user_mode) { - temp = get_sh_mem_bases_32(qpd->pqm->process, dqm->dev); + temp = get_sh_mem_bases_32(pdd); qpd->sh_mem_bases = SHARED_BASE(temp); qpd->sh_mem_config |= PTR32; } else { - temp = get_sh_mem_bases_nybble_64(qpd->pqm->process, dqm->dev); + temp = get_sh_mem_bases_nybble_64(pdd); qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp); } @@ -707,8 +705,7 @@ static int stop_cpsch(struct device_queue_manager *dqm) destroy_queues_cpsch(dqm, true); list_for_each_entry(node, &dqm->queues, list) { - pdd = kfd_get_process_device_data(dqm->dev, - node->qpd->pqm->process, 1); + pdd = qpd_to_pdd(node->qpd); pdd->bound = false; } kfd2kgd->free_mem(dqm->dev->kgd, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index d0bcafc..f9fb81e3 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -414,6 +414,8 @@ struct kfd_process_device { bool bound; }; +#define qpd_to_pdd(x) container_of(x, struct kfd_process_device, qpd) + /* Process data */ struct kfd_process { /* -- cgit v0.10.2 From abc9d3e3b9dc77d7b887ade9f5845a93ce743fd4 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Sun, 9 Nov 2014 22:36:22 +0200 Subject: amdkfd: Clear ctx cb before suspend Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 9beb6f7..43884eb 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -267,6 +267,7 @@ void kgd2kfd_suspend(struct kfd_dev *kfd) if (kfd->init_complete) { kfd->dqm->stop(kfd->dqm); + amd_iommu_set_invalidate_ctx_cb(kfd->pdev, NULL); amd_iommu_free_device(kfd->pdev); } } -- cgit v0.10.2 From cc7096fb6d1dfbdac5e7e2675c046fd40646cc66 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 22 Oct 2014 12:03:04 +1000 Subject: drm/mode: document path property and function to set it. (v1.1) These two didn't get documented properly, do so. Pointed out by Daniel. v1.1: add missing boilerplate (Daniel) Reviewed-by: Daniel Vetter Signed-off-by: Dave Airlie diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index 64d9c1e..b4d80a2 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -2534,7 +2534,7 @@ void intel_crt_init(struct drm_device *dev) - + @@ -2549,6 +2549,13 @@ void intel_crt_init(struct drm_device *dev) + + + + + + + diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 0f3c24c..e6c1691 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -4005,6 +4005,19 @@ done: return ret; } +/** + * drm_mode_connector_set_path_property - set tile property on connector + * @connector: connector to set property on. + * @path: path to use for property. + * + * This creates a property to expose to userspace to specify a + * connector path. This is mainly used for DisplayPort MST where + * connectors have a topology and we want to allow userspace to give + * them more meaningful names. + * + * Returns: + * Zero on success, errno on failure. + */ int drm_mode_connector_set_path_property(struct drm_connector *connector, char *path) { -- cgit v0.10.2 From 5ee3229c87d396cab3c2dfc335b90320cc4a2f42 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Tue, 11 Nov 2014 19:38:59 -0500 Subject: drm: export atomic wait_for_vblanks helper (v2) v1: original v2: danvet's kerneldoc nitpicks Signed-off-by: Rob Clark Reviewed-by: Daniel Vetter Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index ca839bd..fad2b93 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -754,8 +754,18 @@ static void wait_for_fences(struct drm_device *dev, } } -static void -wait_for_vblanks(struct drm_device *dev, struct drm_atomic_state *old_state) +/** + * drm_atomic_helper_wait_for_vblanks - wait for vblank on crtcs + * @dev: DRM device + * @old_state: atomic state object with old state structures + * + * Helper to, after atomic commit, wait for vblanks on all effected + * crtcs (ie. before cleaning up old framebuffers using + * drm_atomic_helper_cleanup_planes()) + */ +void +drm_atomic_helper_wait_for_vblanks(struct drm_device *dev, + struct drm_atomic_state *old_state) { struct drm_crtc *crtc; struct drm_crtc_state *old_crtc_state; @@ -800,6 +810,7 @@ wait_for_vblanks(struct drm_device *dev, struct drm_atomic_state *old_state) drm_crtc_vblank_put(crtc); } } +EXPORT_SYMBOL(drm_atomic_helper_wait_for_vblanks); /** * drm_atomic_helper_commit - commit validated state object @@ -859,7 +870,7 @@ int drm_atomic_helper_commit(struct drm_device *dev, drm_atomic_helper_commit_post_planes(dev, state); - wait_for_vblanks(dev, state); + drm_atomic_helper_wait_for_vblanks(dev, state); drm_atomic_helper_cleanup_planes(dev, state); diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h index 67e3c46..64b4e91 100644 --- a/include/drm/drm_atomic_helper.h +++ b/include/drm/drm_atomic_helper.h @@ -34,6 +34,9 @@ int drm_atomic_helper_commit(struct drm_device *dev, struct drm_atomic_state *state, bool async); +void drm_atomic_helper_wait_for_vblanks(struct drm_device *dev, + struct drm_atomic_state *old_state); + void drm_atomic_helper_commit_pre_planes(struct drm_device *dev, struct drm_atomic_state *state); void drm_atomic_helper_commit_post_planes(struct drm_device *dev, -- cgit v0.10.2 From 4d02e2de0e80a786452e70d7f3a20a50641e6620 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 11 Nov 2014 10:12:00 +0100 Subject: drm: Per-plane locking Turned out to be much simpler on top of my latest atomic stuff than what I've feared. Some details: - Drop the modeset_lock_all snakeoil in drm_plane_init. Same justification as for the equivalent change in drm_crtc_init done in commit d0fa1af40e784aaf7ebb7ba8a17b229bb3fa4c21 Author: Daniel Vetter Date: Mon Sep 8 09:02:49 2014 +0200 drm: Drop modeset locking from crtc init function Without these the drm_modeset_lock_init would fall over the exact same way. - Since the atomic core code wraps the locking switching it to per-plane locks was a one-line change. - For the legacy ioctls add a plane argument to the locking helper so that we can grab the right plane lock (cursor or primary). Since the universal cursor plane might not be there, or someone really crazy might forgoe the primary plane even accept NULL. - Add some locking WARN_ON to the atomic helpers for good paranoid measure and to check that it all works out. Tested on my exynos atomic hackfest with full lockdep checks and ww backoff injection. v2: I've forgotten about the load-detect code in i915. v3: Thierry reported that in latest 3.18-rc vmwgfx doesn't compile any more due to commit 21e88620aa21b48d4f62d29275e3e2944a5ea2b5 Author: Rob Clark Date: Thu Oct 30 13:39:04 2014 -0400 drm/vmwgfx: fix lock breakage Rebased and fix this up. Cc: Thierry Reding Signed-off-by: Daniel Vetter Reviewed-by: Sean Paul Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index ed991ba..ed22a71 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -244,7 +244,7 @@ drm_atomic_get_plane_state(struct drm_atomic_state *state, * grab all crtc locks. Once we have per-plane locks we must update this * to only take the plane mutex. */ - ret = drm_modeset_lock_all_crtcs(state->dev, state->acquire_ctx); + ret = drm_modeset_lock(&plane->mutex, state->acquire_ctx); if (ret) return ERR_PTR(ret); diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index fad2b93..c0a07cb 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -1006,6 +1006,8 @@ void drm_atomic_helper_commit_planes(struct drm_device *dev, if (!crtc) continue; + WARN_ON(!drm_modeset_is_locked(&crtc->mutex)); + funcs = crtc->helper_private; if (!funcs || !funcs->atomic_begin) @@ -1021,6 +1023,8 @@ void drm_atomic_helper_commit_planes(struct drm_device *dev, if (!plane) continue; + WARN_ON(!drm_modeset_is_locked(&plane->mutex)); + funcs = plane->helper_private; if (!funcs || !funcs->atomic_update) diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index e6c1691..3652ed8 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -1152,12 +1152,12 @@ int drm_universal_plane_init(struct drm_device *dev, struct drm_plane *plane, { int ret; - drm_modeset_lock_all(dev); - ret = drm_mode_object_get(dev, &plane->base, DRM_MODE_OBJECT_PLANE); if (ret) goto out; + drm_modeset_lock_init(&plane->mutex); + plane->base.properties = &plane->properties; plane->dev = dev; plane->funcs = funcs; @@ -1185,7 +1185,6 @@ int drm_universal_plane_init(struct drm_device *dev, struct drm_plane *plane, plane->type); out: - drm_modeset_unlock_all(dev); return ret; } @@ -2809,7 +2808,7 @@ static int drm_mode_cursor_common(struct drm_device *dev, * If this crtc has a universal cursor plane, call that plane's update * handler rather than using legacy cursor handlers. */ - drm_modeset_lock_crtc(crtc); + drm_modeset_lock_crtc(crtc, crtc->cursor); if (crtc->cursor) { ret = drm_mode_cursor_universal(crtc, req, file_priv); goto out; @@ -4598,7 +4597,7 @@ int drm_mode_page_flip_ioctl(struct drm_device *dev, if (!crtc) return -ENOENT; - drm_modeset_lock_crtc(crtc); + drm_modeset_lock_crtc(crtc, crtc->primary); if (crtc->primary->fb == NULL) { /* The framebuffer is currently unbound, presumably * due to a hotplug event, that userspace has not diff --git a/drivers/gpu/drm/drm_modeset_lock.c b/drivers/gpu/drm/drm_modeset_lock.c index 474e4d1..51cc47d 100644 --- a/drivers/gpu/drm/drm_modeset_lock.c +++ b/drivers/gpu/drm/drm_modeset_lock.c @@ -157,14 +157,20 @@ void drm_modeset_unlock_all(struct drm_device *dev) EXPORT_SYMBOL(drm_modeset_unlock_all); /** - * drm_modeset_lock_crtc - lock crtc with hidden acquire ctx - * @crtc: drm crtc + * drm_modeset_lock_crtc - lock crtc with hidden acquire ctx for a plane update + * @crtc: DRM CRTC + * @plane: DRM plane to be updated on @crtc + * + * This function locks the given crtc and plane (which should be either the + * primary or cursor plane) using a hidden acquire context. This is necessary so + * that drivers internally using the atomic interfaces can grab further locks + * with the lock acquire context. * - * This function locks the given crtc using a hidden acquire context. This is - * necessary so that drivers internally using the atomic interfaces can grab - * further locks with the lock acquire context. + * Note that @plane can be NULL, e.g. when the cursor support hasn't yet been + * converted to universal planes yet. */ -void drm_modeset_lock_crtc(struct drm_crtc *crtc) +void drm_modeset_lock_crtc(struct drm_crtc *crtc, + struct drm_plane *plane) { struct drm_modeset_acquire_ctx *ctx; int ret; @@ -180,6 +186,18 @@ retry: if (ret) goto fail; + if (plane) { + ret = drm_modeset_lock(&plane->mutex, ctx); + if (ret) + goto fail; + + if (plane->crtc) { + ret = drm_modeset_lock(&plane->crtc->mutex, ctx); + if (ret) + goto fail; + } + } + WARN_ON(crtc->acquire_ctx); /* now we hold the locks, so now that it is safe, stash the @@ -437,15 +455,14 @@ void drm_modeset_unlock(struct drm_modeset_lock *lock) } EXPORT_SYMBOL(drm_modeset_unlock); -/* Temporary.. until we have sufficiently fine grained locking, there - * are a couple scenarios where it is convenient to grab all crtc locks. - * It is planned to remove this: - */ +/* In some legacy codepaths it's convenient to just grab all the crtc and plane + * related locks. */ int drm_modeset_lock_all_crtcs(struct drm_device *dev, struct drm_modeset_acquire_ctx *ctx) { struct drm_mode_config *config = &dev->mode_config; struct drm_crtc *crtc; + struct drm_plane *plane; int ret = 0; list_for_each_entry(crtc, &config->crtc_list, head) { @@ -454,6 +471,12 @@ int drm_modeset_lock_all_crtcs(struct drm_device *dev, return ret; } + list_for_each_entry(plane, &config->plane_list, head) { + ret = drm_modeset_lock(&plane->mutex, ctx); + if (ret) + return ret; + } + return 0; } EXPORT_SYMBOL(drm_modeset_lock_all_crtcs); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 31bb1d7..5fae12c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8730,6 +8730,9 @@ retry: ret = drm_modeset_lock(&crtc->mutex, ctx); if (ret) goto fail_unlock; + ret = drm_modeset_lock(&crtc->primary->mutex, ctx); + if (ret) + goto fail_unlock; old->dpms_mode = connector->dpms; old->load_detect_temp = false; @@ -8767,6 +8770,9 @@ retry: ret = drm_modeset_lock(&crtc->mutex, ctx); if (ret) goto fail_unlock; + ret = drm_modeset_lock(&crtc->primary->mutex, ctx); + if (ret) + goto fail_unlock; intel_encoder->new_crtc = to_intel_crtc(crtc); to_intel_connector(connector)->new_encoder = intel_encoder; diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c index 941a7bc..3725b52 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c @@ -252,7 +252,7 @@ int vmw_du_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv, ret = 0; out: drm_modeset_unlock_all(dev_priv->dev); - drm_modeset_lock_crtc(crtc); + drm_modeset_lock_crtc(crtc, crtc->cursor); return ret; } @@ -281,7 +281,7 @@ int vmw_du_crtc_cursor_move(struct drm_crtc *crtc, int x, int y) du->cursor_y + du->hotspot_y); drm_modeset_unlock_all(dev_priv->dev); - drm_modeset_lock_crtc(crtc); + drm_modeset_lock_crtc(crtc, crtc->cursor); return 0; } diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index bc1cc3c..cbb4756 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -751,6 +751,8 @@ struct drm_plane { struct drm_device *dev; struct list_head head; + struct drm_modeset_lock mutex; + struct drm_mode_object base; uint32_t possible_crtcs; diff --git a/include/drm/drm_modeset_lock.h b/include/drm/drm_modeset_lock.h index 28931a2..70595ff 100644 --- a/include/drm/drm_modeset_lock.h +++ b/include/drm/drm_modeset_lock.h @@ -127,11 +127,13 @@ void drm_modeset_unlock(struct drm_modeset_lock *lock); struct drm_device; struct drm_crtc; +struct drm_plane; void drm_modeset_lock_all(struct drm_device *dev); int __drm_modeset_lock_all(struct drm_device *dev, bool trylock); void drm_modeset_unlock_all(struct drm_device *dev); -void drm_modeset_lock_crtc(struct drm_crtc *crtc); +void drm_modeset_lock_crtc(struct drm_crtc *crtc, + struct drm_plane *plane); void drm_modeset_unlock_crtc(struct drm_crtc *crtc); void drm_warn_on_modeset_not_all_locked(struct drm_device *dev); struct drm_modeset_acquire_ctx * -- cgit v0.10.2 From fcf93f6948c3c29fd414bc00f1f713c501dfd22e Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 12 Nov 2014 08:45:01 +0100 Subject: drm: More specific locking for get* ioctls Motivated by the per-plane locking I've gone through all the get* ioctls and reduced the locking to the bare minimum required. v2: Rebase and make it compile ... v3: Review from Sean: - Simplify return handling in getplane_res. - Add a comment to getplane_res that the plane list is invariant and can be walked locklessly. v4: Actually git add. Cc: Sean Paul Reviewed-by: Sean Paul Signed-off-by: Daniel Vetter Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 3652ed8..2beb219 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -1743,7 +1743,9 @@ int drm_mode_getresources(struct drm_device *dev, void *data, card_res->count_fbs = fb_count; mutex_unlock(&file_priv->fbs_lock); - drm_modeset_lock_all(dev); + /* mode_config.mutex protects the connector list against e.g. DP MST + * connector hot-adding. CRTC/Plane lists are invariant. */ + mutex_lock(&dev->mode_config.mutex); if (!drm_is_primary_client(file_priv)) { mode_group = NULL; @@ -1863,7 +1865,7 @@ int drm_mode_getresources(struct drm_device *dev, void *data, card_res->count_connectors, card_res->count_encoders); out: - drm_modeset_unlock_all(dev); + mutex_unlock(&dev->mode_config.mutex); return ret; } @@ -1890,14 +1892,11 @@ int drm_mode_getcrtc(struct drm_device *dev, if (!drm_core_check_feature(dev, DRIVER_MODESET)) return -EINVAL; - drm_modeset_lock_all(dev); - crtc = drm_crtc_find(dev, crtc_resp->crtc_id); - if (!crtc) { - ret = -ENOENT; - goto out; - } + if (!crtc) + return -ENOENT; + drm_modeset_lock_crtc(crtc, crtc->primary); crtc_resp->x = crtc->x; crtc_resp->y = crtc->y; crtc_resp->gamma_size = crtc->gamma_size; @@ -1914,9 +1913,8 @@ int drm_mode_getcrtc(struct drm_device *dev, } else { crtc_resp->mode_valid = 0; } + drm_modeset_unlock_crtc(crtc); -out: - drm_modeset_unlock_all(dev); return ret; } @@ -2100,24 +2098,22 @@ int drm_mode_getencoder(struct drm_device *dev, void *data, if (!drm_core_check_feature(dev, DRIVER_MODESET)) return -EINVAL; - drm_modeset_lock_all(dev); encoder = drm_encoder_find(dev, enc_resp->encoder_id); - if (!encoder) { - ret = -ENOENT; - goto out; - } + if (!encoder) + return -ENOENT; + drm_modeset_lock(&dev->mode_config.connection_mutex, NULL); if (encoder->crtc) enc_resp->crtc_id = encoder->crtc->base.id; else enc_resp->crtc_id = 0; + drm_modeset_unlock(&dev->mode_config.connection_mutex); + enc_resp->encoder_type = encoder->encoder_type; enc_resp->encoder_id = encoder->base.id; enc_resp->possible_crtcs = encoder->possible_crtcs; enc_resp->possible_clones = encoder->possible_clones; -out: - drm_modeset_unlock_all(dev); return ret; } @@ -2141,13 +2137,12 @@ int drm_mode_getplane_res(struct drm_device *dev, void *data, struct drm_mode_config *config; struct drm_plane *plane; uint32_t __user *plane_ptr; - int copied = 0, ret = 0; + int copied = 0; unsigned num_planes; if (!drm_core_check_feature(dev, DRIVER_MODESET)) return -EINVAL; - drm_modeset_lock_all(dev); config = &dev->mode_config; if (file_priv->universal_planes) @@ -2163,6 +2158,7 @@ int drm_mode_getplane_res(struct drm_device *dev, void *data, (plane_resp->count_planes >= num_planes)) { plane_ptr = (uint32_t __user *)(unsigned long)plane_resp->plane_id_ptr; + /* Plane lists are invariant, no locking needed. */ list_for_each_entry(plane, &config->plane_list, head) { /* * Unless userspace set the 'universal planes' @@ -2172,18 +2168,14 @@ int drm_mode_getplane_res(struct drm_device *dev, void *data, !file_priv->universal_planes) continue; - if (put_user(plane->base.id, plane_ptr + copied)) { - ret = -EFAULT; - goto out; - } + if (put_user(plane->base.id, plane_ptr + copied)) + return -EFAULT; copied++; } } plane_resp->count_planes = num_planes; -out: - drm_modeset_unlock_all(dev); - return ret; + return 0; } /** @@ -2210,13 +2202,11 @@ int drm_mode_getplane(struct drm_device *dev, void *data, if (!drm_core_check_feature(dev, DRIVER_MODESET)) return -EINVAL; - drm_modeset_lock_all(dev); plane = drm_plane_find(dev, plane_resp->plane_id); - if (!plane) { - ret = -ENOENT; - goto out; - } + if (!plane) + return -ENOENT; + drm_modeset_lock(&plane->mutex, NULL); if (plane->crtc) plane_resp->crtc_id = plane->crtc->base.id; else @@ -2226,6 +2216,7 @@ int drm_mode_getplane(struct drm_device *dev, void *data, plane_resp->fb_id = plane->fb->base.id; else plane_resp->fb_id = 0; + drm_modeset_unlock(&plane->mutex); plane_resp->plane_id = plane->base.id; plane_resp->possible_crtcs = plane->possible_crtcs; @@ -2241,14 +2232,11 @@ int drm_mode_getplane(struct drm_device *dev, void *data, if (copy_to_user(format_ptr, plane->format_types, sizeof(uint32_t) * plane->format_count)) { - ret = -EFAULT; - goto out; + return -EFAULT; } } plane_resp->count_format_types = plane->format_count; -out: - drm_modeset_unlock_all(dev); return ret; } -- cgit v0.10.2 From f354d733f680ec9f9afed0d0fa6429ed259f0a52 Mon Sep 17 00:00:00 2001 From: Ander Conselvan de Oliveira Date: Fri, 7 Nov 2014 14:07:41 +0200 Subject: drm/i915: Plug memory leak in intel_shared_dpll_start_config() The cleanup path would reset pll->new_config to NULL but wouldn't free the allocated memory. Signed-off-by: Ander Conselvan de Oliveira Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 3f4a5b8..ea607f3 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -3953,6 +3953,7 @@ static int intel_shared_dpll_start_config(struct drm_i915_private *dev_priv, cleanup: while (--i >= 0) { pll = &dev_priv->shared_dplls[i]; + kfree(pll->new_config); pll->new_config = NULL; } -- cgit v0.10.2 From c9da4a4b3820782753c1c455242eca7e5a69eedb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 10 Oct 2014 12:28:36 +0900 Subject: drm/radeon: Try placing NO_CPU_ACCESS BOs outside of CPU accessible VRAM MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This avoids them getting in the way of BOs which might be accessed by the CPU. They can still go to the CPU accessible part of VRAM though if there's no space outside of it. Signed-off-by: Michel Dänzer Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index a9717b3..73288c6 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -474,7 +474,7 @@ struct radeon_bo { struct list_head list; /* Protected by tbo.reserved */ u32 initial_domain; - struct ttm_place placements[3]; + struct ttm_place placements[4]; struct ttm_placement placement; struct ttm_buffer_object tbo; struct ttm_bo_kmap_obj kmap; diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index 99a960a..7f3b1e1 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -99,22 +99,39 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) rbo->placement.placement = rbo->placements; rbo->placement.busy_placement = rbo->placements; - if (domain & RADEON_GEM_DOMAIN_VRAM) + if (domain & RADEON_GEM_DOMAIN_VRAM) { + /* Try placing BOs which don't need CPU access outside of the + * CPU accessible part of VRAM + */ + if ((rbo->flags & RADEON_GEM_NO_CPU_ACCESS) && + rbo->rdev->mc.visible_vram_size < rbo->rdev->mc.real_vram_size) { + rbo->placements[c].fpfn = + rbo->rdev->mc.visible_vram_size >> PAGE_SHIFT; + rbo->placements[c++].flags = TTM_PL_FLAG_WC | + TTM_PL_FLAG_UNCACHED | + TTM_PL_FLAG_VRAM; + } + + rbo->placements[c].fpfn = 0; rbo->placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_VRAM; + } if (domain & RADEON_GEM_DOMAIN_GTT) { if (rbo->flags & RADEON_GEM_GTT_UC) { + rbo->placements[c].fpfn = 0; rbo->placements[c++].flags = TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_TT; } else if ((rbo->flags & RADEON_GEM_GTT_WC) || (rbo->rdev->flags & RADEON_IS_AGP)) { + rbo->placements[c].fpfn = 0; rbo->placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_TT; } else { + rbo->placements[c].fpfn = 0; rbo->placements[c++].flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_TT; } @@ -122,30 +139,35 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) if (domain & RADEON_GEM_DOMAIN_CPU) { if (rbo->flags & RADEON_GEM_GTT_UC) { + rbo->placements[c].fpfn = 0; rbo->placements[c++].flags = TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_SYSTEM; } else if ((rbo->flags & RADEON_GEM_GTT_WC) || rbo->rdev->flags & RADEON_IS_AGP) { + rbo->placements[c].fpfn = 0; rbo->placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_SYSTEM; } else { + rbo->placements[c].fpfn = 0; rbo->placements[c++].flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_SYSTEM; } } - if (!c) + if (!c) { + rbo->placements[c].fpfn = 0; rbo->placements[c++].flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM; + } rbo->placement.num_placement = c; rbo->placement.num_busy_placement = c; for (i = 0; i < c; ++i) { - rbo->placements[i].fpfn = 0; if ((rbo->flags & RADEON_GEM_CPU_ACCESS) && - (rbo->placements[i].flags & TTM_PL_FLAG_VRAM)) + (rbo->placements[i].flags & TTM_PL_FLAG_VRAM) && + !rbo->placements[i].fpfn) rbo->placements[i].lpfn = rbo->rdev->mc.visible_vram_size >> PAGE_SHIFT; else @@ -743,8 +765,8 @@ int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo) { struct radeon_device *rdev; struct radeon_bo *rbo; - unsigned long offset, size; - int r; + unsigned long offset, size, lpfn; + int i, r; if (!radeon_ttm_bo_is_radeon_bo(bo)) return 0; @@ -761,7 +783,13 @@ int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo) /* hurrah the memory is not visible ! */ radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM); - rbo->placements[0].lpfn = rdev->mc.visible_vram_size >> PAGE_SHIFT; + lpfn = rdev->mc.visible_vram_size >> PAGE_SHIFT; + for (i = 0; i < rbo->placement.num_placement; i++) { + /* Force into visible VRAM */ + if ((rbo->placements[i].flags & TTM_PL_FLAG_VRAM) && + (!rbo->placements[i].lpfn || rbo->placements[i].lpfn > lpfn)) + rbo->placements[i].lpfn = lpfn; + } r = ttm_bo_validate(bo, &rbo->placement, false, false); if (unlikely(r == -ENOMEM)) { radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_GTT); -- cgit v0.10.2 From 2a85aedd117cab38fd29c2c1a15b4c7338850d53 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 9 Oct 2014 18:55:04 +0900 Subject: drm/radeon: Try evicting from CPU accessible to inaccessible VRAM first MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Michel Dänzer Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c index 8624979..cbe7b32 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -198,7 +198,30 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo, case TTM_PL_VRAM: if (rbo->rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready == false) radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU); - else + else if (rbo->rdev->mc.visible_vram_size < rbo->rdev->mc.real_vram_size && + bo->mem.start < (rbo->rdev->mc.visible_vram_size >> PAGE_SHIFT)) { + unsigned fpfn = rbo->rdev->mc.visible_vram_size >> PAGE_SHIFT; + int i; + + /* Try evicting to the CPU inaccessible part of VRAM + * first, but only set GTT as busy placement, so this + * BO will be evicted to GTT rather than causing other + * BOs to be evicted from VRAM + */ + radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM | + RADEON_GEM_DOMAIN_GTT); + rbo->placement.num_busy_placement = 0; + for (i = 0; i < rbo->placement.num_placement; i++) { + if (rbo->placements[i].flags & TTM_PL_FLAG_VRAM) { + if (rbo->placements[0].fpfn < fpfn) + rbo->placements[0].fpfn = fpfn; + } else { + rbo->placement.busy_placement = + &rbo->placements[i]; + rbo->placement.num_busy_placement = 1; + } + } + } else radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_GTT); break; case TTM_PL_TT: -- cgit v0.10.2 From a8b5ebe6b56cca0f12f0f53cdae2f03950f4100f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 28 Oct 2014 18:35:02 +0900 Subject: drm/radeon: Set TTM_PL_FLAG_TOPDOWN also for RADEON_GEM_CPU_ACCESS BOs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit I wasn't sure if TTM_PL_FLAG_TOPDOWN works correctly with non-0 lpfn, but AFAICT it does. Reviewed-by: Lauri Kasanen Signed-off-by: Michel Dänzer Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index 7f3b1e1..33e6c7a 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -179,9 +179,7 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) * improve fragmentation quality. * 512kb was measured as the most optimal number. */ - if (!((rbo->flags & RADEON_GEM_CPU_ACCESS) && - (rbo->placements[i].flags & TTM_PL_FLAG_VRAM)) && - rbo->tbo.mem.size > 512 * 1024) { + if (rbo->tbo.mem.size > 512 * 1024) { for (i = 0; i < c; i++) { rbo->placements[i].flags |= TTM_PL_FLAG_TOPDOWN; } -- cgit v0.10.2 From c165812cbf6cf4bdd62f174115c01017f55f0933 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 28 Oct 2014 18:35:03 +0900 Subject: drm/ttm: Add DRM_MM_SEARCH_BELOW for TTM_PL_FLAG_TOPDOWN MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If the BO should be placed at the top of the area, we should start looking for holes from the top. Reviewed-by: Lauri Kasanen Signed-off-by: Michel Dänzer Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/ttm/ttm_bo_manager.c b/drivers/gpu/drm/ttm/ttm_bo_manager.c index 964387f..1e93f6c 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_manager.c +++ b/drivers/gpu/drm/ttm/ttm_bo_manager.c @@ -55,6 +55,7 @@ static int ttm_bo_man_get_node(struct ttm_mem_type_manager *man, struct ttm_range_manager *rman = (struct ttm_range_manager *) man->priv; struct drm_mm *mm = &rman->mm; struct drm_mm_node *node = NULL; + enum drm_mm_search_flags sflags = DRM_MM_SEARCH_BEST; enum drm_mm_allocator_flags aflags = DRM_MM_CREATE_DEFAULT; unsigned long lpfn; int ret; @@ -67,15 +68,16 @@ static int ttm_bo_man_get_node(struct ttm_mem_type_manager *man, if (!node) return -ENOMEM; - if (place->flags & TTM_PL_FLAG_TOPDOWN) + if (place->flags & TTM_PL_FLAG_TOPDOWN) { + sflags |= DRM_MM_SEARCH_BELOW; aflags = DRM_MM_CREATE_TOP; + } spin_lock(&rman->lock); ret = drm_mm_insert_node_in_range_generic(mm, node, mem->num_pages, mem->page_alignment, 0, place->fpfn, lpfn, - DRM_MM_SEARCH_BEST, - aflags); + sflags, aflags); spin_unlock(&rman->lock); if (unlikely(ret)) { -- cgit v0.10.2 From 507d0ca71bcbefd8623eb20f4de1b5f4e103a48b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 28 Oct 2014 18:35:04 +0900 Subject: drm/ttm: Use only DRM_MM_SEARCH_BELOW for TTM_PL_FLAG_TOPDOWN MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit DRM_MM_SEARCH_BEST gets the smallest hole which can fit the BO. That seems against the idea of TTM_PL_FLAG_TOPDOWN: * The smallest hole may be in the overall bottom of the area * If the hole isn't much larger than the BO, it doesn't make much difference whether the BO is placed at the bottom or at the top of the hole Reviewed-by: Lauri Kasanen Signed-off-by: Michel Dänzer Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/ttm/ttm_bo_manager.c b/drivers/gpu/drm/ttm/ttm_bo_manager.c index 1e93f6c..aa0bd05 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_manager.c +++ b/drivers/gpu/drm/ttm/ttm_bo_manager.c @@ -69,7 +69,7 @@ static int ttm_bo_man_get_node(struct ttm_mem_type_manager *man, return -ENOMEM; if (place->flags & TTM_PL_FLAG_TOPDOWN) { - sflags |= DRM_MM_SEARCH_BELOW; + sflags = DRM_MM_SEARCH_BELOW; aflags = DRM_MM_CREATE_TOP; } -- cgit v0.10.2 From 9b92d1ec62d515c28bb579466e2397ab20c1735a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 8 Sep 2014 02:51:49 -0400 Subject: drm/radeon/dpm: grab fan info from vbios Required for fan control support. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 73288c6..908f349 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1623,6 +1623,11 @@ struct radeon_pm { /* internal thermal controller on rv6xx+ */ enum radeon_int_thermal_type int_thermal_type; struct device *int_hwmon_dev; + /* fan control parameters */ + bool no_fan; + u8 fan_pulses_per_revolution; + u8 fan_min_rpm; + u8 fan_max_rpm; /* dpm */ bool dpm_enabled; struct radeon_dpm dpm; diff --git a/drivers/gpu/drm/radeon/radeon_atombios.c b/drivers/gpu/drm/radeon/radeon_atombios.c index df69b92..9beadf8 100644 --- a/drivers/gpu/drm/radeon/radeon_atombios.c +++ b/drivers/gpu/drm/radeon/radeon_atombios.c @@ -2248,6 +2248,14 @@ static void radeon_atombios_add_pplib_thermal_controller(struct radeon_device *r /* add the i2c bus for thermal/fan chip */ if (controller->ucType > 0) { + if (controller->ucFanParameters & ATOM_PP_FANPARAMETERS_NOFAN) + rdev->pm.no_fan = true; + rdev->pm.fan_pulses_per_revolution = + controller->ucFanParameters & ATOM_PP_FANPARAMETERS_TACHOMETER_PULSES_PER_REVOLUTION_MASK; + if (rdev->pm.fan_pulses_per_revolution) { + rdev->pm.fan_min_rpm = controller->ucFanMinRPM; + rdev->pm.fan_max_rpm = controller->ucFanMaxRPM; + } if (controller->ucType == ATOM_PP_THERMALCONTROLLER_RV6xx) { DRM_INFO("Internal thermal controller %s fan control\n", (controller->ucFanParameters & -- cgit v0.10.2 From 2271e2e2a2525f826549f1aa19db36e0b80082fe Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 8 Sep 2014 03:35:17 -0400 Subject: drm/radeon: rework SI dpm thermal setup In preparation for fan control. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/si_dpm.c b/drivers/gpu/drm/radeon/si_dpm.c index 676e6c2..b59e1d6 100644 --- a/drivers/gpu/drm/radeon/si_dpm.c +++ b/drivers/gpu/drm/radeon/si_dpm.c @@ -5817,8 +5817,33 @@ void si_dpm_setup_asic(struct radeon_device *rdev) si_enable_acpi_power_management(rdev); } -static int si_set_thermal_temperature_range(struct radeon_device *rdev, - int min_temp, int max_temp) +static int si_thermal_enable_alert(struct radeon_device *rdev, + bool enable) +{ + u32 thermal_int = RREG32(CG_THERMAL_INT); + + if (enable) { + PPSMC_Result result; + + thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW; + rdev->irq.dpm_thermal = true; + result = si_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + if (result != PPSMC_Result_OK) { + DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); + return -EINVAL; + } + } else { + thermal_int &= ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); + rdev->irq.dpm_thermal = false; + } + + WREG32(CG_THERMAL_INT, thermal_int); + + return 0; +} + +static int si_thermal_set_temperature_range(struct radeon_device *rdev, + int min_temp, int max_temp) { int low_temp = 0 * 1000; int high_temp = 255 * 1000; @@ -5959,26 +5984,32 @@ int si_dpm_enable(struct radeon_device *rdev) return 0; } -int si_dpm_late_enable(struct radeon_device *rdev) +static int si_set_temperature_range(struct radeon_device *rdev) { int ret; - if (rdev->irq.installed && - r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { - PPSMC_Result result; + ret = si_thermal_enable_alert(rdev, false); + if (ret) + return ret; + ret = si_thermal_set_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + ret = si_thermal_enable_alert(rdev, true); + if (ret) + return ret; - ret = si_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); - if (ret) - return ret; - rdev->irq.dpm_thermal = true; - radeon_irq_set(rdev); - result = si_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + return ret; +} - if (result != PPSMC_Result_OK) - DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); - } +int si_dpm_late_enable(struct radeon_device *rdev) +{ + int ret; - return 0; + ret = si_set_temperature_range(rdev); + if (ret) + return ret; + + return ret; } void si_dpm_disable(struct radeon_device *rdev) -- cgit v0.10.2 From 1955f107a7bf5c9f718bf5025a24f5ff1607d7b2 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Sun, 14 Sep 2014 23:45:30 -0400 Subject: drm/radeon: rework CI dpm thermal setup In preparation for fan control. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 11a55e9..51bf798 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -814,7 +814,7 @@ static void ci_apply_state_adjust_rules(struct radeon_device *rdev, } } -static int ci_set_thermal_temperature_range(struct radeon_device *rdev, +static int ci_thermal_set_temperature_range(struct radeon_device *rdev, int min_temp, int max_temp) { int low_temp = 0 * 1000; @@ -850,6 +850,35 @@ static int ci_set_thermal_temperature_range(struct radeon_device *rdev, return 0; } +static int ci_thermal_enable_alert(struct radeon_device *rdev, + bool enable) +{ + u32 thermal_int = RREG32_SMC(CG_THERMAL_INT); + PPSMC_Result result; + + if (enable) { + thermal_int &= ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); + rdev->irq.dpm_thermal = false; + result = ci_send_msg_to_smc(rdev, PPSMC_MSG_Thermal_Cntl_Enable); + if (result != PPSMC_Result_OK) { + DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); + return -EINVAL; + } + } else { + thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW; + rdev->irq.dpm_thermal = true; + result = ci_send_msg_to_smc(rdev, PPSMC_MSG_Thermal_Cntl_Disable); + if (result != PPSMC_Result_OK) { + DRM_DEBUG_KMS("Could not disable thermal interrupts.\n"); + return -EINVAL; + } + } + + WREG32_SMC(CG_THERMAL_INT, thermal_int); + + return 0; +} + #if 0 static int ci_read_smc_soft_register(struct radeon_device *rdev, u16 reg_offset, u32 *value) @@ -4682,29 +4711,30 @@ int ci_dpm_enable(struct radeon_device *rdev) return 0; } -int ci_dpm_late_enable(struct radeon_device *rdev) +static int ci_set_temperature_range(struct radeon_device *rdev) { int ret; - if (rdev->irq.installed && - r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { -#if 0 - PPSMC_Result result; -#endif - ret = ci_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); - if (ret) { - DRM_ERROR("ci_set_thermal_temperature_range failed\n"); - return ret; - } - rdev->irq.dpm_thermal = true; - radeon_irq_set(rdev); -#if 0 - result = ci_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + ret = ci_thermal_enable_alert(rdev, false); + if (ret) + return ret; + ret = ci_thermal_set_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + ret = ci_thermal_enable_alert(rdev, true); + if (ret) + return ret; - if (result != PPSMC_Result_OK) - DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); -#endif - } + return ret; +} + +int ci_dpm_late_enable(struct radeon_device *rdev) +{ + int ret; + + ret = ci_set_temperature_range(rdev); + if (ret) + return ret; ci_dpm_powergate_uvd(rdev, true); diff --git a/drivers/gpu/drm/radeon/ppsmc.h b/drivers/gpu/drm/radeon/ppsmc.h index 5670b82..11c0e4d 100644 --- a/drivers/gpu/drm/radeon/ppsmc.h +++ b/drivers/gpu/drm/radeon/ppsmc.h @@ -106,6 +106,7 @@ typedef uint8_t PPSMC_Result; #define PPSMC_MSG_SAMUDPM_SetEnabledMask ((uint16_t) 0x130) #define PPSMC_MSG_MCLKDPM_ForceState ((uint16_t) 0x131) #define PPSMC_MSG_MCLKDPM_NoForcedLevel ((uint16_t) 0x132) +#define PPSMC_MSG_Thermal_Cntl_Disable ((uint16_t) 0x133) #define PPSMC_MSG_Voltage_Cntl_Disable ((uint16_t) 0x135) #define PPSMC_MSG_PCIeDPM_Enable ((uint16_t) 0x136) #define PPSMC_MSG_PCIeDPM_Disable ((uint16_t) 0x13d) @@ -157,10 +158,11 @@ typedef uint8_t PPSMC_Result; #define PPSMC_MSG_DPM_Config ((uint32_t) 0x102) #define PPSMC_MSG_DPM_ForceState ((uint32_t) 0x104) #define PPSMC_MSG_PG_SIMD_Config ((uint32_t) 0x108) -#define PPSMC_MSG_DPM_N_LevelsDisabled ((uint32_t) 0x112) +#define PPSMC_MSG_Thermal_Cntl_Enable ((uint32_t) 0x10a) #define PPSMC_MSG_Voltage_Cntl_Enable ((uint32_t) 0x109) #define PPSMC_MSG_VCEPowerOFF ((uint32_t) 0x10e) #define PPSMC_MSG_VCEPowerON ((uint32_t) 0x10f) +#define PPSMC_MSG_DPM_N_LevelsDisabled ((uint32_t) 0x112) #define PPSMC_MSG_DCE_RemoveVoltageAdjustment ((uint32_t) 0x11d) #define PPSMC_MSG_DCE_AllowVoltageAdjustment ((uint32_t) 0x11e) #define PPSMC_MSG_EnableBAPM ((uint32_t) 0x120) -- cgit v0.10.2 From 129acb7c0b682512e89c4f65c33593d50f2f49a9 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 11:05:04 -0500 Subject: drm/radeon: fix typo in CI dpm disable Need to disable DS, not enable it when disabling dpm. Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 51bf798..7092034 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -4759,7 +4759,7 @@ void ci_dpm_disable(struct radeon_device *rdev) ci_enable_spread_spectrum(rdev, false); ci_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, false); ci_stop_dpm(rdev); - ci_enable_ds_master_switch(rdev, true); + ci_enable_ds_master_switch(rdev, false); ci_enable_ulv(rdev, false); ci_clear_vc(rdev); ci_reset_to_default(rdev); -- cgit v0.10.2 From 09e619c0c60013921678994a9888c8abe2a6e10b Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 11:16:25 -0500 Subject: drm/radeon: export radeon_atombios_lookup_gpio We need it for dpm. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_atombios.c b/drivers/gpu/drm/radeon/radeon_atombios.c index 9beadf8..7ef3d9e 100644 --- a/drivers/gpu/drm/radeon/radeon_atombios.c +++ b/drivers/gpu/drm/radeon/radeon_atombios.c @@ -196,8 +196,8 @@ void radeon_atombios_i2c_init(struct radeon_device *rdev) } } -static struct radeon_gpio_rec radeon_lookup_gpio(struct radeon_device *rdev, - u8 id) +struct radeon_gpio_rec radeon_atombios_lookup_gpio(struct radeon_device *rdev, + u8 id) { struct atom_context *ctx = rdev->mode_info.atom_context; struct radeon_gpio_rec gpio; @@ -801,7 +801,7 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev) hpd_record = (ATOM_HPD_INT_RECORD *) record; - gpio = radeon_lookup_gpio(rdev, + gpio = radeon_atombios_lookup_gpio(rdev, hpd_record->ucHPDIntGPIOID); hpd = radeon_atom_get_hpd_info_from_gpio(rdev, &gpio); hpd.plugged_state = hpd_record->ucPlugged_PinState; @@ -2128,7 +2128,7 @@ static int radeon_atombios_parse_power_table_1_3(struct radeon_device *rdev) rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_GPIO; rdev->pm.power_state[state_index].clock_info[0].voltage.gpio = - radeon_lookup_gpio(rdev, + radeon_atombios_lookup_gpio(rdev, power_info->info.asPowerPlayInfo[i].ucVoltageDropIndex); if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH) rdev->pm.power_state[state_index].clock_info[0].voltage.active_high = @@ -2164,7 +2164,7 @@ static int radeon_atombios_parse_power_table_1_3(struct radeon_device *rdev) rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_GPIO; rdev->pm.power_state[state_index].clock_info[0].voltage.gpio = - radeon_lookup_gpio(rdev, + radeon_atombios_lookup_gpio(rdev, power_info->info_2.asPowerPlayInfo[i].ucVoltageDropIndex); if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH) rdev->pm.power_state[state_index].clock_info[0].voltage.active_high = @@ -2200,7 +2200,7 @@ static int radeon_atombios_parse_power_table_1_3(struct radeon_device *rdev) rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_GPIO; rdev->pm.power_state[state_index].clock_info[0].voltage.gpio = - radeon_lookup_gpio(rdev, + radeon_atombios_lookup_gpio(rdev, power_info->info_3.asPowerPlayInfo[i].ucVoltageDropIndex); if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH) rdev->pm.power_state[state_index].clock_info[0].voltage.active_high = diff --git a/drivers/gpu/drm/radeon/radeon_mode.h b/drivers/gpu/drm/radeon/radeon_mode.h index 04db2fd..0ef1413 100644 --- a/drivers/gpu/drm/radeon/radeon_mode.h +++ b/drivers/gpu/drm/radeon/radeon_mode.h @@ -748,6 +748,8 @@ extern bool radeon_atombios_get_ppll_ss_info(struct radeon_device *rdev, extern bool radeon_atombios_get_asic_ss_info(struct radeon_device *rdev, struct radeon_atom_ss *ss, int id, u32 clock); +extern struct radeon_gpio_rec radeon_atombios_lookup_gpio(struct radeon_device *rdev, + u8 id); extern void radeon_compute_pll_legacy(struct radeon_pll *pll, uint64_t freq, -- cgit v0.10.2 From 727b3d25be4116315f4488ee1918ec64927b3eb4 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 11:34:57 -0500 Subject: drm/radeon: store the gpio shift as well We need this in the dpm code. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_atombios.c b/drivers/gpu/drm/radeon/radeon_atombios.c index 7ef3d9e..dbc94f3 100644 --- a/drivers/gpu/drm/radeon/radeon_atombios.c +++ b/drivers/gpu/drm/radeon/radeon_atombios.c @@ -221,6 +221,7 @@ struct radeon_gpio_rec radeon_atombios_lookup_gpio(struct radeon_device *rdev, if (id == pin->ucGPIO_ID) { gpio.id = pin->ucGPIO_ID; gpio.reg = le16_to_cpu(pin->usGpioPin_AIndex) * 4; + gpio.shift = pin->ucGpioPinBitShift; gpio.mask = (1 << pin->ucGpioPinBitShift); gpio.valid = true; break; diff --git a/drivers/gpu/drm/radeon/radeon_mode.h b/drivers/gpu/drm/radeon/radeon_mode.h index 0ef1413..64eba7e 100644 --- a/drivers/gpu/drm/radeon/radeon_mode.h +++ b/drivers/gpu/drm/radeon/radeon_mode.h @@ -462,6 +462,7 @@ struct radeon_gpio_rec { u8 id; u32 reg; u32 mask; + u32 shift; }; struct radeon_hpd { -- cgit v0.10.2 From 34fc0b58d91f7eef2884b5ae798f760323831ae7 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 11:52:12 -0500 Subject: drm/radeon/ci: handle gpio controlled dpm features properly Certain feature enablement depends on entries in the atom gpio pin table. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 7092034..9a3c71c0 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -5098,6 +5098,8 @@ void ci_dpm_fini(struct radeon_device *rdev) int ci_dpm_init(struct radeon_device *rdev) { int index = GetIndexIntoMasterTable(DATA, ASIC_InternalSS_Info); + SMU7_Discrete_DpmTable *dpm_table; + struct radeon_gpio_rec gpio; u16 data_offset, size; u8 frev, crev; struct ci_power_info *pi; @@ -5231,6 +5233,55 @@ int ci_dpm_init(struct radeon_device *rdev) pi->uvd_enabled = false; + dpm_table = &pi->smc_state_table; + + gpio = radeon_atombios_lookup_gpio(rdev, VDDC_VRHOT_GPIO_PINID); + if (gpio.valid) { + dpm_table->VRHotGpio = gpio.shift; + rdev->pm.dpm.platform_caps |= ATOM_PP_PLATFORM_CAP_REGULATOR_HOT; + } else { + dpm_table->VRHotGpio = CISLANDS_UNUSED_GPIO_PIN; + rdev->pm.dpm.platform_caps &= ~ATOM_PP_PLATFORM_CAP_REGULATOR_HOT; + } + + gpio = radeon_atombios_lookup_gpio(rdev, PP_AC_DC_SWITCH_GPIO_PINID); + if (gpio.valid) { + dpm_table->AcDcGpio = gpio.shift; + rdev->pm.dpm.platform_caps |= ATOM_PP_PLATFORM_CAP_HARDWAREDC; + } else { + dpm_table->AcDcGpio = CISLANDS_UNUSED_GPIO_PIN; + rdev->pm.dpm.platform_caps &= ~ATOM_PP_PLATFORM_CAP_HARDWAREDC; + } + + gpio = radeon_atombios_lookup_gpio(rdev, VDDC_PCC_GPIO_PINID); + if (gpio.valid) { + u32 tmp = RREG32_SMC(CNB_PWRMGT_CNTL); + + switch (gpio.shift) { + case 0: + tmp &= ~GNB_SLOW_MODE_MASK; + tmp |= GNB_SLOW_MODE(1); + break; + case 1: + tmp &= ~GNB_SLOW_MODE_MASK; + tmp |= GNB_SLOW_MODE(2); + break; + case 2: + tmp |= GNB_SLOW; + break; + case 3: + tmp |= FORCE_NB_PS1; + break; + case 4: + tmp |= DPM_ENABLED; + break; + default: + DRM_ERROR("Invalid PCC GPIO!"); + break; + } + WREG32_SMC(CNB_PWRMGT_CNTL, tmp); + } + pi->voltage_control = CISLANDS_VOLTAGE_CONTROL_NONE; pi->vddci_control = CISLANDS_VOLTAGE_CONTROL_NONE; pi->mvdd_control = CISLANDS_VOLTAGE_CONTROL_NONE; diff --git a/drivers/gpu/drm/radeon/ci_dpm.h b/drivers/gpu/drm/radeon/ci_dpm.h index 93bbed9..615cb2c 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.h +++ b/drivers/gpu/drm/radeon/ci_dpm.h @@ -33,6 +33,8 @@ #define CISLANDS_MAX_HARDWARE_POWERLEVELS 2 +#define CISLANDS_UNUSED_GPIO_PIN 0x7F + struct ci_pl { u32 mclk; u32 sclk; -- cgit v0.10.2 From 9feb3dda5ce0a9d2155f20506384d2e359f151ce Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 12:00:25 -0500 Subject: drm/radeon: fix for memory training on bonaire 0x6649 Workaround for memory link training on certain variants of 0x6649. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 377afa50..57a359d 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -1806,7 +1806,7 @@ int ci_mc_load_microcode(struct radeon_device *rdev) { const __be32 *fw_data = NULL; const __le32 *new_fw_data = NULL; - u32 running, blackout = 0; + u32 running, blackout = 0, tmp; u32 *io_mc_regs = NULL; const __le32 *new_io_mc_regs = NULL; int i, regs_size, ucode_size; @@ -1866,6 +1866,15 @@ int ci_mc_load_microcode(struct radeon_device *rdev) WREG32(MC_SEQ_IO_DEBUG_DATA, io_mc_regs[(i << 1) + 1]); } } + + tmp = RREG32(MC_SEQ_MISC0); + if ((rdev->pdev->device == 0x6649) && ((tmp & 0xff00) == 0x5600)) { + WREG32(MC_SEQ_IO_DEBUG_INDEX, 5); + WREG32(MC_SEQ_IO_DEBUG_DATA, 0x00000023); + WREG32(MC_SEQ_IO_DEBUG_INDEX, 9); + WREG32(MC_SEQ_IO_DEBUG_DATA, 0x000001f0); + } + /* load the MC ucode */ for (i = 0; i < ucode_size; i++) { if (rdev->new_fw) -- cgit v0.10.2 From 1c52279f578f8d1dc164c0d27f8f00b5aa06f2c1 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 12:06:56 -0500 Subject: drm/radeon: switch force state commands for CI Use the preferred SMC commands for forcing state on CI. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 9a3c71c0..893ed68 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -1282,7 +1282,7 @@ static int ci_dpm_force_state_sclk(struct radeon_device *rdev, u32 n) if (!pi->sclk_dpm_key_disabled) { PPSMC_Result smc_result = - ci_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_DPM_ForceState, n); + ci_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SCLKDPM_SetEnabledMask, 1 << n); if (smc_result != PPSMC_Result_OK) return -EINVAL; } @@ -1296,7 +1296,7 @@ static int ci_dpm_force_state_mclk(struct radeon_device *rdev, u32 n) if (!pi->mclk_dpm_key_disabled) { PPSMC_Result smc_result = - ci_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_MCLKDPM_ForceState, n); + ci_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_MCLKDPM_SetEnabledMask, 1 << n); if (smc_result != PPSMC_Result_OK) return -EINVAL; } @@ -3736,7 +3736,6 @@ int ci_dpm_force_performance_level(struct radeon_device *rdev, enum radeon_dpm_forced_level level) { struct ci_power_info *pi = ci_get_pi(rdev); - PPSMC_Result smc_result; u32 tmp, levels, i; int ret; @@ -3845,21 +3844,9 @@ int ci_dpm_force_performance_level(struct radeon_device *rdev, } } } else if (level == RADEON_DPM_FORCED_LEVEL_AUTO) { - if (!pi->sclk_dpm_key_disabled) { - smc_result = ci_send_msg_to_smc(rdev, PPSMC_MSG_NoForcedLevel); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - if (!pi->mclk_dpm_key_disabled) { - smc_result = ci_send_msg_to_smc(rdev, PPSMC_MSG_MCLKDPM_NoForcedLevel); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - if (!pi->pcie_dpm_key_disabled) { - smc_result = ci_send_msg_to_smc(rdev, PPSMC_MSG_PCIeDPM_UnForceLevel); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } + ret = ci_upload_dpm_level_enable_mask(rdev); + if (ret) + return ret; } rdev->pm.dpm.forced_level = level; -- cgit v0.10.2 From 21b8a369046fbccb39f66cf065c43289376e63fa Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 12:51:04 -0500 Subject: drm/radeon: fix dram timing for certain hawaii boards Certain memory configurations need a fix. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 893ed68..4dd8f357 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -2071,6 +2071,33 @@ static int ci_force_switch_to_arb_f0(struct radeon_device *rdev) return ni_copy_and_switch_arb_sets(rdev, tmp, MC_CG_ARB_FREQ_F0); } +static void ci_register_patching_mc_arb(struct radeon_device *rdev, + const u32 engine_clock, + const u32 memory_clock, + u32 *dram_timimg2) +{ + bool patch; + u32 tmp, tmp2; + + tmp = RREG32(MC_SEQ_MISC0); + patch = ((tmp & 0x0000f00) == 0x300) ? true : false; + + if (patch && + ((rdev->pdev->device == 0x67B0) || + (rdev->pdev->device == 0x67B1))) { + if ((memory_clock > 100000) && (memory_clock <= 125000)) { + tmp2 = (((0x31 * engine_clock) / 125000) - 1) & 0xff; + *dram_timimg2 &= ~0x00ff0000; + *dram_timimg2 |= tmp2 << 16; + } else if ((memory_clock > 125000) && (memory_clock <= 137500)) { + tmp2 = (((0x36 * engine_clock) / 137500) - 1) & 0xff; + *dram_timimg2 &= ~0x00ff0000; + *dram_timimg2 |= tmp2 << 16; + } + } +} + + static int ci_populate_memory_timing_parameters(struct radeon_device *rdev, u32 sclk, u32 mclk, @@ -2086,6 +2113,8 @@ static int ci_populate_memory_timing_parameters(struct radeon_device *rdev, dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); burst_time = RREG32(MC_ARB_BURST_TIME) & STATE0_MASK; + ci_register_patching_mc_arb(rdev, sclk, mclk, &dram_timing2); + arb_regs->McArbDramTiming = cpu_to_be32(dram_timing); arb_regs->McArbDramTiming2 = cpu_to_be32(dram_timing2); arb_regs->McArbBurstTime = (u8)burst_time; -- cgit v0.10.2 From c0392f8f0971fee63fa736c800e7c32b59773ba7 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 12:56:00 -0500 Subject: drm/radeon: improve mclk param calcuations for ci dpm Properly take into account the post divider. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 4dd8f357..3705d13 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -2409,10 +2409,10 @@ static int ci_calculate_mclk_params(struct radeon_device *rdev, u32 tmp; u32 reference_clock = rdev->clock.mpll.reference_freq; - if (pi->mem_gddr5) - freq_nom = memory_clock * 4; + if (mpll_param.qdr == 1) + freq_nom = memory_clock * 4 * (1 << mpll_param.post_div); else - freq_nom = memory_clock * 2; + freq_nom = memory_clock * 2 * (1 << mpll_param.post_div); tmp = (freq_nom / reference_clock); tmp = tmp * tmp; -- cgit v0.10.2 From d3052b8ce8a308d2086519fa5f7c4966257ea184 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 13:04:47 -0500 Subject: drm/radeon: fix activity settings for sclk and mclk for CI Only need to be enabled on the first level. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 3705d13..9d04e68 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -2492,7 +2492,6 @@ static int ci_populate_single_memory_level(struct radeon_device *rdev, &memory_level->MinVddcPhases); memory_level->EnabledForThrottle = 1; - memory_level->EnabledForActivity = 1; memory_level->UpH = 0; memory_level->DownH = 100; memory_level->VoltageDownH = 0; @@ -2825,7 +2824,6 @@ static int ci_populate_single_graphic_level(struct radeon_device *rdev, graphic_level->CcPwrDynRm = 0; graphic_level->CcPwrDynRm1 = 0; - graphic_level->EnabledForActivity = 1; graphic_level->EnabledForThrottle = 1; graphic_level->UpH = 0; graphic_level->DownH = 0; @@ -2878,6 +2876,7 @@ static int ci_populate_all_graphic_levels(struct radeon_device *rdev) pi->smc_state_table.GraphicsLevel[i].DisplayWatermark = PPSMC_DISPLAY_WATERMARK_HIGH; } + pi->smc_state_table.GraphicsLevel[0].EnabledForActivity = 1; pi->smc_state_table.GraphicsDpmLevelCount = (u8)dpm_table->sclk_table.count; pi->dpm_level_enable_mask.sclk_dpm_enable_mask = @@ -2921,6 +2920,8 @@ static int ci_populate_all_memory_levels(struct radeon_device *rdev) return ret; } + pi->smc_state_table.MemoryLevel[0].EnabledForActivity = 1; + pi->smc_state_table.MemoryLevel[0].ActivityLevel = cpu_to_be16(0x1F); pi->smc_state_table.MemoryDpmLevelCount = (u8)dpm_table->mclk_table.count; -- cgit v0.10.2 From 489ba72c1ef9c6ceb1480be1550f235d9591d59a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 13:06:57 -0500 Subject: drm/radeon: fix sclk DS enablement Only enable it for levels 0 and 1. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 9d04e68..324e31d 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -2872,6 +2872,8 @@ static int ci_populate_all_graphic_levels(struct radeon_device *rdev) &pi->smc_state_table.GraphicsLevel[i]); if (ret) return ret; + if (i > 1) + pi->smc_state_table.GraphicsLevel[i].DeepSleepDivId = 0; if (i == (dpm_table->sclk_table.count - 1)) pi->smc_state_table.GraphicsLevel[i].DisplayWatermark = PPSMC_DISPLAY_WATERMARK_HIGH; -- cgit v0.10.2 From 127e056e2a82406fd3455a9da20d02f192ae8a05 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 13:11:39 -0500 Subject: drm/radeon: fix mclk vddc configuration for cards for hawaii Need to use vddc0 for vdcc1 for certain hawaii configurations. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 324e31d..a33e2388 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -2924,6 +2924,14 @@ static int ci_populate_all_memory_levels(struct radeon_device *rdev) pi->smc_state_table.MemoryLevel[0].EnabledForActivity = 1; + if ((dpm_table->mclk_table.count >= 2) && + ((rdev->pdev->device == 0x67B0) || (rdev->pdev->device == 0x67B1))) { + pi->smc_state_table.MemoryLevel[1].MinVddc = + pi->smc_state_table.MemoryLevel[0].MinVddc; + pi->smc_state_table.MemoryLevel[1].MinVddcPhases = + pi->smc_state_table.MemoryLevel[0].MinVddcPhases; + } + pi->smc_state_table.MemoryLevel[0].ActivityLevel = cpu_to_be16(0x1F); pi->smc_state_table.MemoryDpmLevelCount = (u8)dpm_table->mclk_table.count; -- cgit v0.10.2 From 36654dd4b9b9dd65143095ca48f5b9c846b13e0b Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 13:19:17 -0500 Subject: drm/radeon: workaround a hw bug in bonaire pcie dpm Some boards get stuck in pcie x1 otherwise. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index a33e2388..aad6477 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -2988,9 +2988,14 @@ static int ci_setup_default_pcie_tables(struct radeon_device *rdev) &pi->dpm_table.pcie_speed_table, SMU7_MAX_LEVELS_LINK); - ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 0, - pi->pcie_gen_powersaving.min, - pi->pcie_lane_powersaving.min); + if (rdev->family == CHIP_BONAIRE) + ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 0, + pi->pcie_gen_powersaving.min, + pi->pcie_lane_powersaving.max); + else + ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 0, + pi->pcie_gen_powersaving.min, + pi->pcie_lane_powersaving.min); ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 1, pi->pcie_gen_performance.min, pi->pcie_lane_performance.min); -- cgit v0.10.2 From b6b41cf3b64f88472f08276a90248a6fa36ee8c7 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 13:24:54 -0500 Subject: drm/radeon: fix default dpm state setup Only enable the first levels for mclk and sclk. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index aad6477..dbc6d2a 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -3062,19 +3062,21 @@ static int ci_setup_default_dpm_tables(struct radeon_device *rdev) allowed_sclk_vddc_table->entries[i].clk)) { pi->dpm_table.sclk_table.dpm_levels[pi->dpm_table.sclk_table.count].value = allowed_sclk_vddc_table->entries[i].clk; - pi->dpm_table.sclk_table.dpm_levels[pi->dpm_table.sclk_table.count].enabled = true; + pi->dpm_table.sclk_table.dpm_levels[pi->dpm_table.sclk_table.count].enabled = + (i == 0) ? true : false; pi->dpm_table.sclk_table.count++; } } pi->dpm_table.mclk_table.count = 0; for (i = 0; i < allowed_mclk_table->count; i++) { - if ((i==0) || + if ((i == 0) || (pi->dpm_table.mclk_table.dpm_levels[pi->dpm_table.mclk_table.count-1].value != allowed_mclk_table->entries[i].clk)) { pi->dpm_table.mclk_table.dpm_levels[pi->dpm_table.mclk_table.count].value = allowed_mclk_table->entries[i].clk; - pi->dpm_table.mclk_table.dpm_levels[pi->dpm_table.mclk_table.count].enabled = true; + pi->dpm_table.mclk_table.dpm_levels[pi->dpm_table.mclk_table.count].enabled = + (i == 0) ? true : false; pi->dpm_table.mclk_table.count++; } } -- cgit v0.10.2 From 4e21518c3d635ad8cad7c07198bb3f592b25395c Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 13:34:28 -0500 Subject: drm/radeon: set bootup pcie level to max for ci dpm Avoids problems when re-loading the driver. Does not affect power saving when dpm is enabled. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index dbc6d2a..befa84c 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -3242,7 +3242,7 @@ static int ci_init_smc_table(struct radeon_device *rdev) table->VddcVddciDelta = 4000; table->PhaseResponseTime = 0; table->MemoryThermThrottleEnable = 1; - table->PCIeBootLinkLevel = 0; + table->PCIeBootLinkLevel = pi->dpm_table.pcie_speed_table.count - 1; table->PCIeGenInterval = 1; if (pi->voltage_control == CISLANDS_VOLTAGE_CONTROL_BY_SVID2) table->SVI2Enable = 1; -- cgit v0.10.2 From 90b2fee35cb9c143f049091ff7ea87500c5c7c46 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 13:58:11 -0500 Subject: drm/radeon: fix dpm mc init for certain hawaii boards Needs special overrides for certain vram configurations. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index befa84c..d9071ae 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -4099,6 +4099,96 @@ static int ci_copy_vbios_mc_reg_table(const struct atom_mc_reg_table *table, return 0; } +static int ci_register_patching_mc_seq(struct radeon_device *rdev, + struct ci_mc_reg_table *table) +{ + u8 i, k; + u32 tmp; + bool patch; + + tmp = RREG32(MC_SEQ_MISC0); + patch = ((tmp & 0x0000f00) == 0x300) ? true : false; + + if (patch && + ((rdev->pdev->device == 0x67B0) || + (rdev->pdev->device == 0x67B1))) { + for (i = 0; i < table->last; i++) { + if (table->last >= SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + switch(table->mc_reg_address[i].s1 >> 2) { + case MC_SEQ_MISC1: + for (k = 0; k < table->num_entries; k++) { + if ((table->mc_reg_table_entry[k].mclk_max == 125000) || + (table->mc_reg_table_entry[k].mclk_max == 137500)) + table->mc_reg_table_entry[k].mc_data[i] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFFFFF8) | + 0x00000007; + } + break; + case MC_SEQ_WR_CTL_D0: + for (k = 0; k < table->num_entries; k++) { + if ((table->mc_reg_table_entry[k].mclk_max == 125000) || + (table->mc_reg_table_entry[k].mclk_max == 137500)) + table->mc_reg_table_entry[k].mc_data[i] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFF0F00) | + 0x0000D0DD; + } + break; + case MC_SEQ_WR_CTL_D1: + for (k = 0; k < table->num_entries; k++) { + if ((table->mc_reg_table_entry[k].mclk_max == 125000) || + (table->mc_reg_table_entry[k].mclk_max == 137500)) + table->mc_reg_table_entry[k].mc_data[i] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFF0F00) | + 0x0000D0DD; + } + break; + case MC_SEQ_WR_CTL_2: + for (k = 0; k < table->num_entries; k++) { + if ((table->mc_reg_table_entry[k].mclk_max == 125000) || + (table->mc_reg_table_entry[k].mclk_max == 137500)) + table->mc_reg_table_entry[k].mc_data[i] = 0; + } + break; + case MC_SEQ_CAS_TIMING: + for (k = 0; k < table->num_entries; k++) { + if (table->mc_reg_table_entry[k].mclk_max == 125000) + table->mc_reg_table_entry[k].mc_data[i] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xFFE0FE0F) | + 0x000C0140; + else if (table->mc_reg_table_entry[k].mclk_max == 137500) + table->mc_reg_table_entry[k].mc_data[i] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xFFE0FE0F) | + 0x000C0150; + } + break; + case MC_SEQ_MISC_TIMING: + for (k = 0; k < table->num_entries; k++) { + if (table->mc_reg_table_entry[k].mclk_max == 125000) + table->mc_reg_table_entry[k].mc_data[i] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFFFFE0) | + 0x00000030; + else if (table->mc_reg_table_entry[k].mclk_max == 137500) + table->mc_reg_table_entry[k].mc_data[i] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFFFFE0) | + 0x00000035; + } + break; + default: + break; + } + } + + WREG32(MC_SEQ_IO_DEBUG_INDEX, 3); + tmp = RREG32(MC_SEQ_IO_DEBUG_DATA); + tmp = (tmp & 0xFFF8FFFF) | (1 << 16); + WREG32(MC_SEQ_IO_DEBUG_INDEX, 3); + WREG32(MC_SEQ_IO_DEBUG_DATA, tmp); + } + + return 0; +} + static int ci_initialize_mc_reg_table(struct radeon_device *rdev) { struct ci_power_info *pi = ci_get_pi(rdev); @@ -4142,6 +4232,10 @@ static int ci_initialize_mc_reg_table(struct radeon_device *rdev) ci_set_s0_mc_reg_index(ci_table); + ret = ci_register_patching_mc_seq(rdev, ci_table); + if (ret) + goto init_mc_done; + ret = ci_set_mc_special_registers(rdev, ci_table); if (ret) goto init_mc_done; -- cgit v0.10.2 From 542b379b55513ac1c564de7d8dac1cee8b3ba7e8 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 18:31:39 -0500 Subject: drm/radeon: powertune fixes for hawaii - bapm is not available on hawaii - update pt defaults Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index d9071ae..4d88090 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -46,15 +46,15 @@ static const struct ci_pt_defaults defaults_hawaii_xt = { 1, 0xF, 0xFD, 0x19, 5, 0x14, 0, 0xB0000, - { 0x84, 0x0, 0x0, 0x7F, 0x0, 0x0, 0x5A, 0x60, 0x51, 0x8E, 0x79, 0x6B, 0x5F, 0x90, 0x79 }, - { 0x1EA, 0x1EA, 0x1EA, 0x224, 0x224, 0x224, 0x24F, 0x24F, 0x24F, 0x28E, 0x28E, 0x28E, 0x2BC, 0x2BC, 0x2BC } + { 0x2E, 0x00, 0x00, 0x88, 0x00, 0x00, 0x72, 0x60, 0x51, 0xA7, 0x79, 0x6B, 0x90, 0xBD, 0x79 }, + { 0x217, 0x217, 0x217, 0x242, 0x242, 0x242, 0x269, 0x269, 0x269, 0x2A1, 0x2A1, 0x2A1, 0x2C9, 0x2C9, 0x2C9 } }; static const struct ci_pt_defaults defaults_hawaii_pro = { 1, 0xF, 0xFD, 0x19, 5, 0x14, 0, 0x65062, - { 0x93, 0x0, 0x0, 0x97, 0x0, 0x0, 0x6B, 0x60, 0x51, 0x95, 0x79, 0x6B, 0x5F, 0x90, 0x79 }, - { 0x1EA, 0x1EA, 0x1EA, 0x224, 0x224, 0x224, 0x24F, 0x24F, 0x24F, 0x28E, 0x28E, 0x28E, 0x2BC, 0x2BC, 0x2BC } + { 0x2E, 0x00, 0x00, 0x88, 0x00, 0x00, 0x72, 0x60, 0x51, 0xA7, 0x79, 0x6B, 0x90, 0xBD, 0x79 }, + { 0x217, 0x217, 0x217, 0x242, 0x242, 0x242, 0x269, 0x269, 0x269, 0x2A1, 0x2A1, 0x2A1, 0x2C9, 0x2C9, 0x2C9 } }; static const struct ci_pt_defaults defaults_bonaire_xt = @@ -249,7 +249,10 @@ static void ci_initialize_powertune_defaults(struct radeon_device *rdev) if (pi->caps_power_containment) { pi->caps_cac = true; - pi->enable_bapm_feature = true; + if (rdev->family == CHIP_HAWAII) + pi->enable_bapm_feature = false; + else + pi->enable_bapm_feature = true; pi->enable_tdc_limit_feature = true; pi->enable_pkg_pwr_tracking_feature = true; } @@ -700,13 +703,11 @@ static int ci_power_control_set_level(struct radeon_device *rdev) int ret = 0; bool adjust_polarity = false; /* ??? */ - if (pi->caps_power_containment && - (pi->power_containment_features & POWERCONTAINMENT_FEATURE_BAPM)) { + if (pi->caps_power_containment) { adjust_percent = adjust_polarity ? rdev->pm.dpm.tdp_adjustment : (-1 * rdev->pm.dpm.tdp_adjustment); target_tdp = ((100 + adjust_percent) * (s32)cac_tdp_table->configurable_tdp) / 100; - target_tdp *= 256; ret = ci_set_overdrive_target_tdp(rdev, (u32)target_tdp); } -- cgit v0.10.2 From b94b95e7e3710a6137f8bc1311938cab3b067a0e Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 7 Nov 2014 14:12:14 -0500 Subject: drm/radeon: set power control in ci dpm enable Necessary for poper operation. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 4d88090..9dbc52f 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -4833,6 +4833,12 @@ int ci_dpm_enable(struct radeon_device *rdev) return ret; } + ret = ci_power_control_set_level(rdev); + if (ret) { + DRM_ERROR("ci_power_control_set_level failed\n"); + return ret; + } + ci_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); ci_update_current_ps(rdev, boot_ps); @@ -4963,11 +4969,6 @@ int ci_dpm_set_power_state(struct radeon_device *rdev) return 0; } -int ci_dpm_power_control_set_level(struct radeon_device *rdev) -{ - return ci_power_control_set_level(rdev); -} - void ci_dpm_reset_asic(struct radeon_device *rdev) { ci_set_boot_state(rdev); -- cgit v0.10.2 From c81b99423bd9d3fc35ac8752ca5fb4c50eab063c Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 10 Nov 2014 16:51:08 +1000 Subject: drm/radeon/si/ci: make u8 static arrays constant These two arrays don't change, just make them constant, reduces data segment by a few bytes. Signed-off-by: Dave Airlie Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_smc.c b/drivers/gpu/drm/radeon/ci_smc.c index b630edc..e78bcad 100644 --- a/drivers/gpu/drm/radeon/ci_smc.c +++ b/drivers/gpu/drm/radeon/ci_smc.c @@ -129,7 +129,7 @@ void ci_reset_smc(struct radeon_device *rdev) int ci_program_jump_on_start(struct radeon_device *rdev) { - static u8 data[] = { 0xE0, 0x00, 0x80, 0x40 }; + static const u8 data[] = { 0xE0, 0x00, 0x80, 0x40 }; return ci_copy_bytes_to_smc(rdev, 0x0, data, 4, sizeof(data)+1); } diff --git a/drivers/gpu/drm/radeon/si_smc.c b/drivers/gpu/drm/radeon/si_smc.c index 73dbc79..e5bb92f 100644 --- a/drivers/gpu/drm/radeon/si_smc.c +++ b/drivers/gpu/drm/radeon/si_smc.c @@ -135,7 +135,7 @@ void si_reset_smc(struct radeon_device *rdev) int si_program_jump_on_start(struct radeon_device *rdev) { - static u8 data[] = { 0x0E, 0x00, 0x40, 0x40 }; + static const u8 data[] = { 0x0E, 0x00, 0x40, 0x40 }; return si_copy_bytes_to_smc(rdev, 0x0, data, 4, sizeof(data)+1); } -- cgit v0.10.2 From ecbbe59bbb1cd2973e031c5b6ba28653d66a17de Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 13 May 2014 11:36:13 +0200 Subject: drm: Use size_t for blob property sizes size_t is the standard type when dealing with sizes of all kinds. Use it consistently when instantiating DRM blob properties. Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index e79c8d3..d4d7834 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -3938,8 +3938,9 @@ done: return ret; } -static struct drm_property_blob *drm_property_create_blob(struct drm_device *dev, int length, - void *data) +static struct drm_property_blob * +drm_property_create_blob(struct drm_device *dev, size_t length, + void *data) { struct drm_property_blob *blob; int ret; @@ -4023,8 +4024,8 @@ int drm_mode_connector_set_path_property(struct drm_connector *connector, char *path) { struct drm_device *dev = connector->dev; - int ret, size; - size = strlen(path) + 1; + size_t size = strlen(path) + 1; + int ret; connector->path_blob_ptr = drm_property_create_blob(connector->dev, size, path); @@ -4053,7 +4054,8 @@ int drm_mode_connector_update_edid_property(struct drm_connector *connector, struct edid *edid) { struct drm_device *dev = connector->dev; - int ret, size; + size_t size; + int ret; /* ignore requests to set edid when overridden */ if (connector->override_edid) diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index c40070a..628369c 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -196,7 +196,7 @@ struct drm_framebuffer { struct drm_property_blob { struct drm_mode_object base; struct list_head head; - unsigned int length; + size_t length; unsigned char data[]; }; -- cgit v0.10.2 From 12e6cecd55e541d3e8110f7dfbb6a601e81733ff Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 13 May 2014 11:38:36 +0200 Subject: drm: Use const data when creating blob properties Creating a blob property will always copy the input data so the data that is passed in can be const. Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index d4d7834..3ea31bc 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -3940,7 +3940,7 @@ done: static struct drm_property_blob * drm_property_create_blob(struct drm_device *dev, size_t length, - void *data) + const void *data) { struct drm_property_blob *blob; int ret; @@ -4021,7 +4021,7 @@ done: } int drm_mode_connector_set_path_property(struct drm_connector *connector, - char *path) + const char *path) { struct drm_device *dev = connector->dev; size_t size = strlen(path) + 1; @@ -4051,7 +4051,7 @@ EXPORT_SYMBOL(drm_mode_connector_set_path_property); * Zero on success, errno on failure. */ int drm_mode_connector_update_edid_property(struct drm_connector *connector, - struct edid *edid) + const struct edid *edid) { struct drm_device *dev = connector->dev; size_t size; diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index d9a7a78..eb023d6 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -393,7 +393,7 @@ static void intel_connector_remove_from_fbdev(struct intel_connector *connector) #endif } -static struct drm_connector *intel_dp_add_mst_connector(struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port, char *pathprop) +static struct drm_connector *intel_dp_add_mst_connector(struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port, const char *pathprop) { struct intel_dp *intel_dp = container_of(mgr, struct intel_dp, mst_mgr); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 628369c..ce0dd6c 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -978,9 +978,9 @@ extern void drm_mode_config_reset(struct drm_device *dev); extern void drm_mode_config_cleanup(struct drm_device *dev); extern int drm_mode_connector_set_path_property(struct drm_connector *connector, - char *path); + const char *path); extern int drm_mode_connector_update_edid_property(struct drm_connector *connector, - struct edid *edid); + const struct edid *edid); static inline bool drm_property_type_is(struct drm_property *property, uint32_t type) diff --git a/include/drm/drm_dp_mst_helper.h b/include/drm/drm_dp_mst_helper.h index 338fc10..fdcd7f2 100644 --- a/include/drm/drm_dp_mst_helper.h +++ b/include/drm/drm_dp_mst_helper.h @@ -371,7 +371,7 @@ struct drm_dp_sideband_msg_tx { struct drm_dp_mst_topology_mgr; struct drm_dp_mst_topology_cbs { /* create a connector for a port */ - struct drm_connector *(*add_connector)(struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port, char *path); + struct drm_connector *(*add_connector)(struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port, const char *path); void (*destroy_connector)(struct drm_dp_mst_topology_mgr *mgr, struct drm_connector *connector); void (*hotplug)(struct drm_dp_mst_topology_mgr *mgr); -- cgit v0.10.2 From c6a843256a2523c621eb109770c2868ebc29c508 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Thu, 2 Oct 2014 14:45:55 +0200 Subject: drm/gem: Fix typo in kerneldoc The function being documented is drm_gem_object_handle_free(), not drm_gem_object_free(). Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c index f6ca512..973a9b6 100644 --- a/drivers/gpu/drm/drm_gem.c +++ b/drivers/gpu/drm/drm_gem.c @@ -188,7 +188,7 @@ drm_gem_remove_prime_handles(struct drm_gem_object *obj, struct drm_file *filp) } /** - * drm_gem_object_free - release resources bound to userspace handles + * drm_gem_object_handle_free - release resources bound to userspace handles * @obj: GEM object to clean up. * * Called after the last handle to the object has been closed -- cgit v0.10.2 From 34eab43ed2483e69bc79fd2e3aaf5adfae771907 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Wed, 4 Jun 2014 09:18:29 +0200 Subject: drm/prime: Use unsigned type for number of pages The number of pages can never be negative, so an unsigned type is enough. This also matches the type of the n_pages argument of the sg_alloc_table_from_pages() function. Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/drm_prime.c b/drivers/gpu/drm/drm_prime.c index 78ca308..e15882f 100644 --- a/drivers/gpu/drm/drm_prime.c +++ b/drivers/gpu/drm/drm_prime.c @@ -669,7 +669,7 @@ int drm_prime_fd_to_handle_ioctl(struct drm_device *dev, void *data, * the driver is responsible for mapping the pages into the * importers address space for use with dma_buf itself. */ -struct sg_table *drm_prime_pages_to_sg(struct page **pages, int nr_pages) +struct sg_table *drm_prime_pages_to_sg(struct page **pages, unsigned int nr_pages) { struct sg_table *sg = NULL; int ret; diff --git a/include/drm/drmP.h b/include/drm/drmP.h index 53ed876..75b2594 100644 --- a/include/drm/drmP.h +++ b/include/drm/drmP.h @@ -986,7 +986,7 @@ extern void drm_gem_dmabuf_release(struct dma_buf *dma_buf); extern int drm_prime_sg_to_page_addr_arrays(struct sg_table *sgt, struct page **pages, dma_addr_t *addrs, int max_pages); -extern struct sg_table *drm_prime_pages_to_sg(struct page **pages, int nr_pages); +extern struct sg_table *drm_prime_pages_to_sg(struct page **pages, unsigned int nr_pages); extern void drm_prime_gem_destroy(struct drm_gem_object *obj, struct sg_table *sg); -- cgit v0.10.2 From a9e3c90c9fa39ed00e3223dc8f93c9cd12abb750 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 5 Aug 2014 13:37:47 +0200 Subject: drm: Implement drm_get_pci_dev() dummy for !PCI Implementing a dummy of this function allows drivers that use it to be built on platforms that don't have PCI. This can happen for example if the nouveau driver is built on Tegra without PCI enabled (or on 64-bit ARM where PCI is not yet implemented). Signed-off-by: Thierry Reding diff --git a/include/drm/drmP.h b/include/drm/drmP.h index 75b2594..87aac39 100644 --- a/include/drm/drmP.h +++ b/include/drm/drmP.h @@ -1028,10 +1028,25 @@ void drm_pci_agp_destroy(struct drm_device *dev); extern int drm_pci_init(struct drm_driver *driver, struct pci_driver *pdriver); extern void drm_pci_exit(struct drm_driver *driver, struct pci_driver *pdriver); +#ifdef CONFIG_PCI extern int drm_get_pci_dev(struct pci_dev *pdev, const struct pci_device_id *ent, struct drm_driver *driver); extern int drm_pci_set_busid(struct drm_device *dev, struct drm_master *master); +#else +static inline int drm_get_pci_dev(struct pci_dev *pdev, + const struct pci_device_id *ent, + struct drm_driver *driver) +{ + return -ENOSYS; +} + +static inline int drm_pci_set_busid(struct drm_device *dev, + struct drm_master *master) +{ + return -ENOSYS; +} +#endif #define DRM_PCIE_SPEED_25 1 #define DRM_PCIE_SPEED_50 2 -- cgit v0.10.2 From 2f7633125a1ca8a03b63bf91b5eca60551141ddb Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Mon, 13 Oct 2014 12:45:57 +0200 Subject: drm: Make drm_mode_create_tv_properties() signature consistent The prototype and the function implementation differ in their signature. Make them consistent and use an unsigned integer for the number of modes while at it. Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 3ea31bc..776ec41 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -1388,12 +1388,13 @@ EXPORT_SYMBOL(drm_mode_create_dvi_i_properties); * responsible for allocating a list of format names and passing them to * this routine. */ -int drm_mode_create_tv_properties(struct drm_device *dev, int num_modes, +int drm_mode_create_tv_properties(struct drm_device *dev, + unsigned int num_modes, char *modes[]) { struct drm_property *tv_selector; struct drm_property *tv_subconnector; - int i; + unsigned int i; if (dev->mode_config.tv_select_subconnector_property) return 0; @@ -2274,7 +2275,7 @@ static int __setplane_internal(struct drm_plane *plane, { int ret = 0; unsigned int fb_width, fb_height; - int i; + unsigned int i; /* No fb means shut it down */ if (!fb) { diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index ce0dd6c..4825507 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -1041,8 +1041,9 @@ extern void drm_property_destroy(struct drm_device *dev, struct drm_property *pr extern int drm_property_add_enum(struct drm_property *property, int index, uint64_t value, const char *name); extern int drm_mode_create_dvi_i_properties(struct drm_device *dev); -extern int drm_mode_create_tv_properties(struct drm_device *dev, int num_formats, - char *formats[]); +extern int drm_mode_create_tv_properties(struct drm_device *dev, + unsigned int num_modes, + char *modes[]); extern int drm_mode_create_scaling_mode_property(struct drm_device *dev); extern int drm_mode_create_aspect_ratio_property(struct drm_device *dev); extern int drm_mode_create_dirty_info_property(struct drm_device *dev); -- cgit v0.10.2 From 8446956ed2e877f8d725d7e3b59cbf35ee1081f0 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 5 Aug 2014 08:22:48 +0200 Subject: drm/gma500: mdfld: Reuse video/mipi_display.h The GMA500 driver redefines many constants already found in the generic header. Replace uses of the custom defines by the standard ones and get rid of the duplicate defininitions. Acked-by: Alan Cox Signed-off-by: Thierry Reding diff --git a/drivers/gpu/drm/gma500/mdfld_dsi_pkg_sender.c b/drivers/gpu/drm/gma500/mdfld_dsi_pkg_sender.c index 87885d8..6b43ae3 100644 --- a/drivers/gpu/drm/gma500/mdfld_dsi_pkg_sender.c +++ b/drivers/gpu/drm/gma500/mdfld_dsi_pkg_sender.c @@ -25,6 +25,7 @@ */ #include +#include - + @@ -2677,6 +2677,21 @@ void intel_crt_init(struct drm_device *dev) + + + + + + + + + + + + + + + diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 5a4be03..15f8b3b 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -1529,6 +1529,30 @@ int drm_mode_create_dirty_info_property(struct drm_device *dev) } EXPORT_SYMBOL(drm_mode_create_dirty_info_property); +/** + * drm_mode_create_suggested_offset_properties - create suggests offset properties + * @dev: DRM device + * + * Create the the suggested x/y offset property for connectors. + */ +int drm_mode_create_suggested_offset_properties(struct drm_device *dev) +{ + if (dev->mode_config.suggested_x_property && dev->mode_config.suggested_y_property) + return 0; + + dev->mode_config.suggested_x_property = + drm_property_create_range(dev, DRM_MODE_PROP_IMMUTABLE, "suggested X", 0, 0xffffffff); + + dev->mode_config.suggested_y_property = + drm_property_create_range(dev, DRM_MODE_PROP_IMMUTABLE, "suggested Y", 0, 0xffffffff); + + if (dev->mode_config.suggested_x_property == NULL || + dev->mode_config.suggested_y_property == NULL) + return -ENOMEM; + return 0; +} +EXPORT_SYMBOL(drm_mode_create_suggested_offset_properties); + static int drm_mode_group_init(struct drm_device *dev, struct drm_mode_group *group) { uint32_t total_objects = 0; diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index c959410..7b28ab0 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -1038,6 +1038,10 @@ struct drm_mode_config { struct drm_property *aspect_ratio_property; struct drm_property *dirty_info_property; + /* properties for virtual machine layout */ + struct drm_property *suggested_x_property; + struct drm_property *suggested_y_property; + /* dumb ioctl parameters */ uint32_t preferred_depth, prefer_shadow; @@ -1231,6 +1235,7 @@ extern int drm_mode_create_tv_properties(struct drm_device *dev, extern int drm_mode_create_scaling_mode_property(struct drm_device *dev); extern int drm_mode_create_aspect_ratio_property(struct drm_device *dev); extern int drm_mode_create_dirty_info_property(struct drm_device *dev); +extern int drm_mode_create_suggested_offset_properties(struct drm_device *dev); extern int drm_mode_connector_attach_encoder(struct drm_connector *connector, struct drm_encoder *encoder); -- cgit v0.10.2 From 7dea0941f8806e79fed562256822564d5f903edc Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Tue, 28 Oct 2014 11:28:44 +1000 Subject: drm/qxl: use suggested x/y offset properties to pass guest prefs This passes the guest preferences for a where to place the outputs through to userspace. Userspace would need to be updated to take note of this information, X server and GNOME. Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/qxl/qxl_display.c b/drivers/gpu/drm/qxl/qxl_display.c index 8b78928..b8cf556 100644 --- a/drivers/gpu/drm/qxl/qxl_display.c +++ b/drivers/gpu/drm/qxl/qxl_display.c @@ -101,14 +101,37 @@ static int qxl_display_copy_rom_client_monitors_config(struct qxl_device *qdev) return 0; } +static void qxl_update_offset_props(struct qxl_device *qdev) +{ + struct drm_device *dev = qdev->ddev; + struct drm_connector *connector; + struct qxl_output *output; + struct qxl_head *head; + + list_for_each_entry(connector, &dev->mode_config.connector_list, head) { + output = drm_connector_to_qxl_output(connector); + + head = &qdev->client_monitors_config->heads[output->index]; + + drm_object_property_set_value(&connector->base, + dev->mode_config.suggested_x_property, head->x); + drm_object_property_set_value(&connector->base, + dev->mode_config.suggested_y_property, head->y); + } +} + void qxl_display_read_client_monitors_config(struct qxl_device *qdev) { + struct drm_device *dev = qdev->ddev; while (qxl_display_copy_rom_client_monitors_config(qdev)) { qxl_io_log(qdev, "failed crc check for client_monitors_config," " retrying\n"); } + drm_modeset_lock_all(dev); + qxl_update_offset_props(qdev); + drm_modeset_unlock_all(dev); if (!drm_helper_hpd_irq_event(qdev->ddev)) { /* notify that the monitor configuration changed, to adjust at the arbitrary resolution */ @@ -952,6 +975,10 @@ static int qdev_output_init(struct drm_device *dev, int num_output) drm_object_attach_property(&connector->base, qdev->hotplug_mode_update_property, 0); + drm_object_attach_property(&connector->base, + dev->mode_config.suggested_x_property, 0); + drm_object_attach_property(&connector->base, + dev->mode_config.suggested_y_property, 0); drm_connector_register(connector); return 0; } @@ -1065,6 +1092,7 @@ int qxl_modeset_init(struct qxl_device *qdev) qdev->ddev->mode_config.fb_base = qdev->vram_base; + drm_mode_create_suggested_offset_properties(qdev->ddev); qxl_mode_create_hotplug_mode_update_property(qdev); for (i = 0 ; i < qxl_num_crtc; ++i) { -- cgit v0.10.2 From 967b8e04d7993e93a44ffb63cc6504b6eded899a Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Sat, 15 Nov 2014 10:15:30 +1000 Subject: drm/qxl: drop unused mode private pointer This was pointless, forgot to remove the code. Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/qxl/qxl_display.c b/drivers/gpu/drm/qxl/qxl_display.c index b8cf556..4a0a8b2 100644 --- a/drivers/gpu/drm/qxl/qxl_display.c +++ b/drivers/gpu/drm/qxl/qxl_display.c @@ -592,7 +592,6 @@ static int qxl_crtc_mode_set(struct drm_crtc *crtc, { struct drm_device *dev = crtc->dev; struct qxl_device *qdev = dev->dev_private; - struct qxl_mode *m = (void *)mode->private; struct qxl_framebuffer *qfb; struct qxl_bo *bo, *old_bo = NULL; struct qxl_crtc *qcrtc = to_qxl_crtc(crtc); @@ -610,12 +609,6 @@ static int qxl_crtc_mode_set(struct drm_crtc *crtc, } qfb = to_qxl_framebuffer(crtc->primary->fb); bo = gem_to_qxl_bo(qfb->obj); - if (!m) - /* and do we care? */ - DRM_DEBUG("%dx%d: not a native mode\n", x, y); - else - DRM_DEBUG("%dx%d: qxl id %d\n", - mode->hdisplay, mode->vdisplay, m->id); DRM_DEBUG("+%d+%d (%d,%d) => (%d,%d)\n", x, y, mode->hdisplay, mode->vdisplay, -- cgit v0.10.2 From b7bbd6406ae82f4faa8a5e571866e44f4c49255d Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 31 Oct 2014 12:19:40 -0400 Subject: drm/msm: select REGULATOR MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fixes a potential error, spotted by Felipe with randconfig: ----- drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c: In function ‘mdp4_kms_init’: drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c:384:2: error: implicit declaration \ of function ‘devm_regulator_get_exclusive’ [-Werror=implicit-function-declaration] mdp4_kms->vdd = devm_regulator_get_exclusive(&pdev->dev, "vdd"); ^ drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c:384:16: error: assignment makes \ pointer from integer without a cast [-Werror] mdp4_kms->vdd = devm_regulator_get_exclusive(&pdev->dev, "vdd"); ^ ----- Also add a brief comment explaining the use of _get_exclusive() Reported-by: Felipe Balbi Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/Kconfig b/drivers/gpu/drm/msm/Kconfig index 9d907c5..5b2a1ff 100644 --- a/drivers/gpu/drm/msm/Kconfig +++ b/drivers/gpu/drm/msm/Kconfig @@ -3,6 +3,7 @@ config DRM_MSM tristate "MSM DRM" depends on DRM depends on ARCH_QCOM || (ARM && COMPILE_TEST) + select REGULATOR select DRM_KMS_HELPER select DRM_PANEL select SHMEM diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c index 79d804e..ac20472 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c @@ -381,6 +381,10 @@ struct msm_kms *mdp4_kms_init(struct drm_device *dev) if (IS_ERR(mdp4_kms->dsi_pll_vddio)) mdp4_kms->dsi_pll_vddio = NULL; + /* NOTE: driver for this regulator still missing upstream.. use + * _get_exclusive() and ignore the error if it does not exist + * (and hope that the bootloader left it on for us) + */ mdp4_kms->vdd = devm_regulator_get_exclusive(&pdev->dev, "vdd"); if (IS_ERR(mdp4_kms->vdd)) mdp4_kms->vdd = NULL; -- cgit v0.10.2 From 0122f96fc2c1f911cb5b4db0a6dc98ac367e7e44 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 31 Oct 2014 11:50:55 -0400 Subject: drm/msm/adreno: slight init order cleanup Move anything that can fail after call to base class msm_gpu_init(). This way, if we fail, active_list has already been initialized so we don't trip 'WARN_ON(!list_empty(&gpu->active_list))' in msm_gpu_cleanup(). Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c index 6afa291..dc807b2 100644 --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c @@ -319,6 +319,12 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev, DBG("fast_rate=%u, slow_rate=%u, bus_freq=%u", gpu->fast_rate, gpu->slow_rate, gpu->bus_freq); + ret = msm_gpu_init(drm, pdev, &adreno_gpu->base, &funcs->base, + adreno_gpu->info->name, "kgsl_3d0_reg_memory", "kgsl_3d0_irq", + RB_SIZE); + if (ret) + return ret; + ret = request_firmware(&adreno_gpu->pm4, adreno_gpu->info->pm4fw, drm->dev); if (ret) { dev_err(drm->dev, "failed to load %s PM4 firmware: %d\n", @@ -333,12 +339,6 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev, return ret; } - ret = msm_gpu_init(drm, pdev, &adreno_gpu->base, &funcs->base, - adreno_gpu->info->name, "kgsl_3d0_reg_memory", "kgsl_3d0_irq", - RB_SIZE); - if (ret) - return ret; - mmu = gpu->mmu; if (mmu) { ret = mmu->funcs->attach(mmu, iommu_ports, -- cgit v0.10.2 From bc00ae02e4a0b0d34a03bb25a8285b80e4f628c4 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 31 Oct 2014 12:54:25 -0400 Subject: drm/msm: update generated headers Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/adreno/a2xx.xml.h b/drivers/gpu/drm/msm/adreno/a2xx.xml.h index a310459..22882cc 100644 --- a/drivers/gpu/drm/msm/adreno/a2xx.xml.h +++ b/drivers/gpu/drm/msm/adreno/a2xx.xml.h @@ -11,10 +11,10 @@ The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 364 bytes, from 2013-11-30 14:47:15) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) - /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2014-06-02 15:21:30) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 9859 bytes, from 2014-06-02 15:21:30) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14960 bytes, from 2014-07-27 17:22:13) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 58020 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 41068 bytes, from 2014-08-01 12:22:48) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10551 bytes, from 2014-11-13 22:44:30) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15053 bytes, from 2014-11-09 15:45:47) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 63169 bytes, from 2014-11-13 22:44:18) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 49097 bytes, from 2014-11-14 15:38:00) Copyright (C) 2013-2014 by the following authors: - Rob Clark (robclark) @@ -926,11 +926,11 @@ static inline uint32_t A2XX_VGT_DRAW_INITIATOR_INDEX_SIZE(enum pc_di_index_size #define A2XX_VGT_DRAW_INITIATOR_NOT_EOP 0x00001000 #define A2XX_VGT_DRAW_INITIATOR_SMALL_INDEX 0x00002000 #define A2XX_VGT_DRAW_INITIATOR_PRE_DRAW_INITIATOR_ENABLE 0x00004000 -#define A2XX_VGT_DRAW_INITIATOR_NUM_INDICES__MASK 0xffff0000 -#define A2XX_VGT_DRAW_INITIATOR_NUM_INDICES__SHIFT 16 -static inline uint32_t A2XX_VGT_DRAW_INITIATOR_NUM_INDICES(uint32_t val) +#define A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__MASK 0xff000000 +#define A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__SHIFT 24 +static inline uint32_t A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES(uint32_t val) { - return ((val) << A2XX_VGT_DRAW_INITIATOR_NUM_INDICES__SHIFT) & A2XX_VGT_DRAW_INITIATOR_NUM_INDICES__MASK; + return ((val) << A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__SHIFT) & A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__MASK; } #define REG_A2XX_VGT_IMMED_DATA 0x000021fd @@ -1243,13 +1243,13 @@ static inline uint32_t A2XX_CLEAR_COLOR_ALPHA(uint32_t val) #define A2XX_PA_SU_POINT_SIZE_HEIGHT__SHIFT 0 static inline uint32_t A2XX_PA_SU_POINT_SIZE_HEIGHT(float val) { - return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_POINT_SIZE_HEIGHT__SHIFT) & A2XX_PA_SU_POINT_SIZE_HEIGHT__MASK; + return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_POINT_SIZE_HEIGHT__SHIFT) & A2XX_PA_SU_POINT_SIZE_HEIGHT__MASK; } #define A2XX_PA_SU_POINT_SIZE_WIDTH__MASK 0xffff0000 #define A2XX_PA_SU_POINT_SIZE_WIDTH__SHIFT 16 static inline uint32_t A2XX_PA_SU_POINT_SIZE_WIDTH(float val) { - return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_POINT_SIZE_WIDTH__SHIFT) & A2XX_PA_SU_POINT_SIZE_WIDTH__MASK; + return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_POINT_SIZE_WIDTH__SHIFT) & A2XX_PA_SU_POINT_SIZE_WIDTH__MASK; } #define REG_A2XX_PA_SU_POINT_MINMAX 0x00002281 @@ -1257,13 +1257,13 @@ static inline uint32_t A2XX_PA_SU_POINT_SIZE_WIDTH(float val) #define A2XX_PA_SU_POINT_MINMAX_MIN__SHIFT 0 static inline uint32_t A2XX_PA_SU_POINT_MINMAX_MIN(float val) { - return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_POINT_MINMAX_MIN__SHIFT) & A2XX_PA_SU_POINT_MINMAX_MIN__MASK; + return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_POINT_MINMAX_MIN__SHIFT) & A2XX_PA_SU_POINT_MINMAX_MIN__MASK; } #define A2XX_PA_SU_POINT_MINMAX_MAX__MASK 0xffff0000 #define A2XX_PA_SU_POINT_MINMAX_MAX__SHIFT 16 static inline uint32_t A2XX_PA_SU_POINT_MINMAX_MAX(float val) { - return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_POINT_MINMAX_MAX__SHIFT) & A2XX_PA_SU_POINT_MINMAX_MAX__MASK; + return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_POINT_MINMAX_MAX__SHIFT) & A2XX_PA_SU_POINT_MINMAX_MAX__MASK; } #define REG_A2XX_PA_SU_LINE_CNTL 0x00002282 @@ -1271,7 +1271,7 @@ static inline uint32_t A2XX_PA_SU_POINT_MINMAX_MAX(float val) #define A2XX_PA_SU_LINE_CNTL_WIDTH__SHIFT 0 static inline uint32_t A2XX_PA_SU_LINE_CNTL_WIDTH(float val) { - return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_LINE_CNTL_WIDTH__SHIFT) & A2XX_PA_SU_LINE_CNTL_WIDTH__MASK; + return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_LINE_CNTL_WIDTH__SHIFT) & A2XX_PA_SU_LINE_CNTL_WIDTH__MASK; } #define REG_A2XX_PA_SC_LINE_STIPPLE 0x00002283 diff --git a/drivers/gpu/drm/msm/adreno/a3xx.xml.h b/drivers/gpu/drm/msm/adreno/a3xx.xml.h index 82d0152..109e9a2 100644 --- a/drivers/gpu/drm/msm/adreno/a3xx.xml.h +++ b/drivers/gpu/drm/msm/adreno/a3xx.xml.h @@ -11,10 +11,10 @@ The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 364 bytes, from 2013-11-30 14:47:15) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) - /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2014-06-02 15:21:30) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 9859 bytes, from 2014-06-02 15:21:30) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14960 bytes, from 2014-07-27 17:22:13) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 58020 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 41068 bytes, from 2014-08-01 12:22:48) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10551 bytes, from 2014-11-13 22:44:30) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15053 bytes, from 2014-11-09 15:45:47) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 63169 bytes, from 2014-11-13 22:44:18) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 49097 bytes, from 2014-11-14 15:38:00) Copyright (C) 2013-2014 by the following authors: - Rob Clark (robclark) @@ -86,6 +86,14 @@ enum a3xx_vtx_fmt { VFMT_NORM_USHORT_16_16 = 29, VFMT_NORM_USHORT_16_16_16 = 30, VFMT_NORM_USHORT_16_16_16_16 = 31, + VFMT_UINT_32 = 32, + VFMT_UINT_32_32 = 33, + VFMT_UINT_32_32_32 = 34, + VFMT_UINT_32_32_32_32 = 35, + VFMT_INT_32 = 36, + VFMT_INT_32_32 = 37, + VFMT_INT_32_32_32 = 38, + VFMT_INT_32_32_32_32 = 39, VFMT_UBYTE_8 = 40, VFMT_UBYTE_8_8 = 41, VFMT_UBYTE_8_8_8 = 42, @@ -112,7 +120,9 @@ enum a3xx_tex_fmt { TFMT_NORM_USHORT_565 = 4, TFMT_NORM_USHORT_5551 = 6, TFMT_NORM_USHORT_4444 = 7, + TFMT_NORM_USHORT_Z16 = 9, TFMT_NORM_UINT_X8Z24 = 10, + TFMT_FLOAT_Z32 = 11, TFMT_NORM_UINT_NV12_UV_TILED = 17, TFMT_NORM_UINT_NV12_Y_TILED = 19, TFMT_NORM_UINT_NV12_UV = 21, @@ -121,18 +131,38 @@ enum a3xx_tex_fmt { TFMT_NORM_UINT_I420_U = 26, TFMT_NORM_UINT_I420_V = 27, TFMT_NORM_UINT_2_10_10_10 = 41, + TFMT_FLOAT_9_9_9_E5 = 42, + TFMT_FLOAT_10_11_11 = 43, TFMT_NORM_UINT_A8 = 44, TFMT_NORM_UINT_L8_A8 = 47, TFMT_NORM_UINT_8 = 48, TFMT_NORM_UINT_8_8 = 49, TFMT_NORM_UINT_8_8_8 = 50, TFMT_NORM_UINT_8_8_8_8 = 51, + TFMT_NORM_SINT_8_8 = 53, + TFMT_NORM_SINT_8_8_8_8 = 55, + TFMT_UINT_8_8 = 57, + TFMT_UINT_8_8_8_8 = 59, + TFMT_SINT_8_8 = 61, + TFMT_SINT_8_8_8_8 = 63, TFMT_FLOAT_16 = 64, TFMT_FLOAT_16_16 = 65, TFMT_FLOAT_16_16_16_16 = 67, + TFMT_UINT_16 = 68, + TFMT_UINT_16_16 = 69, + TFMT_UINT_16_16_16_16 = 71, + TFMT_SINT_16 = 72, + TFMT_SINT_16_16 = 73, + TFMT_SINT_16_16_16_16 = 75, TFMT_FLOAT_32 = 84, TFMT_FLOAT_32_32 = 85, TFMT_FLOAT_32_32_32_32 = 87, + TFMT_UINT_32 = 88, + TFMT_UINT_32_32 = 89, + TFMT_UINT_32_32_32_32 = 91, + TFMT_SINT_32 = 92, + TFMT_SINT_32_32 = 93, + TFMT_SINT_32_32_32_32 = 95, }; enum a3xx_tex_fetchsize { @@ -145,19 +175,34 @@ enum a3xx_tex_fetchsize { }; enum a3xx_color_fmt { + RB_R5G6B5_UNORM = 0, + RB_R5G5B5A1_UNORM = 1, + RB_R4G4B4A4_UNORM = 3, RB_R8G8B8_UNORM = 4, RB_R8G8B8A8_UNORM = 8, - RB_Z16_UNORM = 12, + RB_R8G8B8A8_UINT = 10, + RB_R8G8B8A8_SINT = 11, + RB_R8G8_UNORM = 12, + RB_R8_UINT = 14, + RB_R8_SINT = 15, + RB_R10G10B10A2_UNORM = 16, RB_A8_UNORM = 20, + RB_R8_UNORM = 21, RB_R16G16B16A16_FLOAT = 27, + RB_R11G11B10_FLOAT = 28, + RB_R16_SINT = 40, + RB_R16G16_SINT = 41, + RB_R16G16B16A16_SINT = 43, + RB_R16_UINT = 44, + RB_R16G16_UINT = 45, + RB_R16G16B16A16_UINT = 47, RB_R32G32B32A32_FLOAT = 51, -}; - -enum a3xx_color_swap { - WZYX = 0, - WXYZ = 1, - ZYXW = 2, - XYZW = 3, + RB_R32_SINT = 52, + RB_R32G32_SINT = 53, + RB_R32G32B32A32_SINT = 55, + RB_R32_UINT = 56, + RB_R32G32_UINT = 57, + RB_R32G32B32A32_UINT = 59, }; enum a3xx_sp_perfcounter_select { @@ -194,6 +239,11 @@ enum a3xx_rb_blend_opcode { BLEND_MAX_DST_SRC = 4, }; +enum a3xx_intp_mode { + SMOOTH = 0, + FLAT = 1, +}; + enum a3xx_tex_filter { A3XX_TEX_NEAREST = 0, A3XX_TEX_LINEAR = 1, @@ -536,6 +586,10 @@ enum a3xx_tex_type { #define REG_A3XX_CP_MEQ_DATA 0x000001db +#define REG_A3XX_CP_WFI_PEND_CTR 0x000001f5 + +#define REG_A3XX_RBBM_PM_OVERRIDE2 0x0000039d + #define REG_A3XX_CP_PERFCOUNTER_SELECT 0x00000445 #define REG_A3XX_CP_HW_FAULT 0x0000045c @@ -550,6 +604,12 @@ static inline uint32_t REG_A3XX_CP_PROTECT_REG(uint32_t i0) { return 0x00000460 #define REG_A3XX_CP_AHB_FAULT 0x0000054d +#define REG_A3XX_SQ_GPR_MANAGEMENT 0x00000d00 + +#define REG_A3XX_SQ_INST_STORE_MANAGMENT 0x00000d02 + +#define REG_A3XX_TP0_CHICKEN 0x00000e1e + #define REG_A3XX_SP_GLOBAL_MEM_SIZE 0x00000e22 #define REG_A3XX_SP_GLOBAL_MEM_ADDR 0x00000e23 @@ -632,13 +692,13 @@ static inline uint32_t A3XX_GRAS_CL_VPORT_ZSCALE(float val) #define A3XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT 0 static inline uint32_t A3XX_GRAS_SU_POINT_MINMAX_MIN(float val) { - return ((((uint32_t)(val * 8.0))) << A3XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT) & A3XX_GRAS_SU_POINT_MINMAX_MIN__MASK; + return ((((uint32_t)(val * 16.0))) << A3XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT) & A3XX_GRAS_SU_POINT_MINMAX_MIN__MASK; } #define A3XX_GRAS_SU_POINT_MINMAX_MAX__MASK 0xffff0000 #define A3XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT 16 static inline uint32_t A3XX_GRAS_SU_POINT_MINMAX_MAX(float val) { - return ((((uint32_t)(val * 8.0))) << A3XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT) & A3XX_GRAS_SU_POINT_MINMAX_MAX__MASK; + return ((((uint32_t)(val * 16.0))) << A3XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT) & A3XX_GRAS_SU_POINT_MINMAX_MAX__MASK; } #define REG_A3XX_GRAS_SU_POINT_SIZE 0x00002069 @@ -646,7 +706,7 @@ static inline uint32_t A3XX_GRAS_SU_POINT_MINMAX_MAX(float val) #define A3XX_GRAS_SU_POINT_SIZE__SHIFT 0 static inline uint32_t A3XX_GRAS_SU_POINT_SIZE(float val) { - return ((((uint32_t)(val * 8.0))) << A3XX_GRAS_SU_POINT_SIZE__SHIFT) & A3XX_GRAS_SU_POINT_SIZE__MASK; + return ((((int32_t)(val * 16.0))) << A3XX_GRAS_SU_POINT_SIZE__SHIFT) & A3XX_GRAS_SU_POINT_SIZE__MASK; } #define REG_A3XX_GRAS_SU_POLY_OFFSET_SCALE 0x0000206c @@ -654,7 +714,7 @@ static inline uint32_t A3XX_GRAS_SU_POINT_SIZE(float val) #define A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__SHIFT 0 static inline uint32_t A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL(float val) { - return ((((uint32_t)(val * 28.0))) << A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__SHIFT) & A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__MASK; + return ((((int32_t)(val * 16384.0))) << A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__SHIFT) & A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__MASK; } #define REG_A3XX_GRAS_SU_POLY_OFFSET_OFFSET 0x0000206d @@ -662,7 +722,7 @@ static inline uint32_t A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL(float val) #define A3XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT 0 static inline uint32_t A3XX_GRAS_SU_POLY_OFFSET_OFFSET(float val) { - return ((((uint32_t)(val * 28.0))) << A3XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT) & A3XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK; + return ((((int32_t)(val * 16384.0))) << A3XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT) & A3XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK; } #define REG_A3XX_GRAS_SU_MODE_CONTROL 0x00002070 @@ -673,7 +733,7 @@ static inline uint32_t A3XX_GRAS_SU_POLY_OFFSET_OFFSET(float val) #define A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT 3 static inline uint32_t A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH(float val) { - return ((((uint32_t)(val * 4.0))) << A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT) & A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__MASK; + return ((((int32_t)(val * 4.0))) << A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT) & A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__MASK; } #define A3XX_GRAS_SU_MODE_CONTROL_POLY_OFFSET 0x00000800 @@ -863,6 +923,7 @@ static inline uint32_t A3XX_RB_MRT_BUF_INFO_COLOR_SWAP(enum a3xx_color_swap val) { return ((val) << A3XX_RB_MRT_BUF_INFO_COLOR_SWAP__SHIFT) & A3XX_RB_MRT_BUF_INFO_COLOR_SWAP__MASK; } +#define A3XX_RB_MRT_BUF_INFO_COLOR_SRGB 0x00004000 #define A3XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__MASK 0xfffe0000 #define A3XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__SHIFT 17 static inline uint32_t A3XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH(uint32_t val) @@ -1001,6 +1062,7 @@ static inline uint32_t A3XX_RB_COPY_CONTROL_FASTCLEAR(uint32_t val) { return ((val) << A3XX_RB_COPY_CONTROL_FASTCLEAR__SHIFT) & A3XX_RB_COPY_CONTROL_FASTCLEAR__MASK; } +#define A3XX_RB_COPY_CONTROL_UNK12 0x00001000 #define A3XX_RB_COPY_CONTROL_GMEM_BASE__MASK 0xffffc000 #define A3XX_RB_COPY_CONTROL_GMEM_BASE__SHIFT 14 static inline uint32_t A3XX_RB_COPY_CONTROL_GMEM_BASE(uint32_t val) @@ -1079,7 +1141,7 @@ static inline uint32_t A3XX_RB_DEPTH_CONTROL_ZFUNC(enum adreno_compare_func val) #define REG_A3XX_RB_DEPTH_CLEAR 0x00002101 #define REG_A3XX_RB_DEPTH_INFO 0x00002102 -#define A3XX_RB_DEPTH_INFO_DEPTH_FORMAT__MASK 0x00000001 +#define A3XX_RB_DEPTH_INFO_DEPTH_FORMAT__MASK 0x00000003 #define A3XX_RB_DEPTH_INFO_DEPTH_FORMAT__SHIFT 0 static inline uint32_t A3XX_RB_DEPTH_INFO_DEPTH_FORMAT(enum adreno_rb_depth_format val) { @@ -1265,6 +1327,7 @@ static inline uint32_t A3XX_PC_PRIM_VTX_CNTL_POLYMODE_BACK_PTYPE(enum adreno_pa_ { return ((val) << A3XX_PC_PRIM_VTX_CNTL_POLYMODE_BACK_PTYPE__SHIFT) & A3XX_PC_PRIM_VTX_CNTL_POLYMODE_BACK_PTYPE__MASK; } +#define A3XX_PC_PRIM_VTX_CNTL_PRIMITIVE_RESTART 0x00100000 #define A3XX_PC_PRIM_VTX_CNTL_PROVOKING_VTX_LAST 0x02000000 #define A3XX_PC_PRIM_VTX_CNTL_PSIZE 0x04000000 @@ -1281,7 +1344,12 @@ static inline uint32_t A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(enum a3xx_threadsize #define A3XX_HLSQ_CONTROL_0_REG_SPSHADERRESTART 0x00000200 #define A3XX_HLSQ_CONTROL_0_REG_RESERVED2 0x00000400 #define A3XX_HLSQ_CONTROL_0_REG_CHUNKDISABLE 0x04000000 -#define A3XX_HLSQ_CONTROL_0_REG_CONSTSWITCHMODE 0x08000000 +#define A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK 0x08000000 +#define A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT 27 +static inline uint32_t A3XX_HLSQ_CONTROL_0_REG_CONSTMODE(uint32_t val) +{ + return ((val) << A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT) & A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK; +} #define A3XX_HLSQ_CONTROL_0_REG_LAZYUPDATEDISABLE 0x10000000 #define A3XX_HLSQ_CONTROL_0_REG_SPCONSTFULLUPDATE 0x20000000 #define A3XX_HLSQ_CONTROL_0_REG_TPFULLUPDATE 0x40000000 @@ -1484,6 +1552,8 @@ static inline uint32_t A3XX_VFD_CONTROL_1_REGID4INST(uint32_t val) #define REG_A3XX_VFD_INDEX_OFFSET 0x00002245 +#define REG_A3XX_VFD_INDEX_OFFSET 0x00002245 + static inline uint32_t REG_A3XX_VFD_FETCH(uint32_t i0) { return 0x00002246 + 0x2*i0; } static inline uint32_t REG_A3XX_VFD_FETCH_INSTR_0(uint32_t i0) { return 0x00002246 + 0x2*i0; } @@ -1537,6 +1607,7 @@ static inline uint32_t A3XX_VFD_DECODE_INSTR_REGID(uint32_t val) { return ((val) << A3XX_VFD_DECODE_INSTR_REGID__SHIFT) & A3XX_VFD_DECODE_INSTR_REGID__MASK; } +#define A3XX_VFD_DECODE_INSTR_INT 0x00100000 #define A3XX_VFD_DECODE_INSTR_SWAP__MASK 0x00c00000 #define A3XX_VFD_DECODE_INSTR_SWAP__SHIFT 22 static inline uint32_t A3XX_VFD_DECODE_INSTR_SWAP(enum a3xx_color_swap val) @@ -1604,6 +1675,102 @@ static inline uint32_t A3XX_VPC_PACK_NUMNONPOSVSVAR(uint32_t val) static inline uint32_t REG_A3XX_VPC_VARYING_INTERP(uint32_t i0) { return 0x00002282 + 0x1*i0; } static inline uint32_t REG_A3XX_VPC_VARYING_INTERP_MODE(uint32_t i0) { return 0x00002282 + 0x1*i0; } +#define A3XX_VPC_VARYING_INTERP_MODE_C0__MASK 0x00000003 +#define A3XX_VPC_VARYING_INTERP_MODE_C0__SHIFT 0 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C0(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C0__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C0__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C1__MASK 0x0000000c +#define A3XX_VPC_VARYING_INTERP_MODE_C1__SHIFT 2 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C1(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C1__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C1__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C2__MASK 0x00000030 +#define A3XX_VPC_VARYING_INTERP_MODE_C2__SHIFT 4 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C2(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C2__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C2__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C3__MASK 0x000000c0 +#define A3XX_VPC_VARYING_INTERP_MODE_C3__SHIFT 6 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C3(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C3__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C3__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C4__MASK 0x00000300 +#define A3XX_VPC_VARYING_INTERP_MODE_C4__SHIFT 8 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C4(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C4__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C4__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C5__MASK 0x00000c00 +#define A3XX_VPC_VARYING_INTERP_MODE_C5__SHIFT 10 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C5(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C5__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C5__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C6__MASK 0x00003000 +#define A3XX_VPC_VARYING_INTERP_MODE_C6__SHIFT 12 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C6(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C6__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C6__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C7__MASK 0x0000c000 +#define A3XX_VPC_VARYING_INTERP_MODE_C7__SHIFT 14 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C7(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C7__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C7__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C8__MASK 0x00030000 +#define A3XX_VPC_VARYING_INTERP_MODE_C8__SHIFT 16 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C8(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C8__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C8__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_C9__MASK 0x000c0000 +#define A3XX_VPC_VARYING_INTERP_MODE_C9__SHIFT 18 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C9(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C9__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C9__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_CA__MASK 0x00300000 +#define A3XX_VPC_VARYING_INTERP_MODE_CA__SHIFT 20 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CA(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CA__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CA__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_CB__MASK 0x00c00000 +#define A3XX_VPC_VARYING_INTERP_MODE_CB__SHIFT 22 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CB(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CB__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CB__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_CC__MASK 0x03000000 +#define A3XX_VPC_VARYING_INTERP_MODE_CC__SHIFT 24 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CC(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CC__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CC__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_CD__MASK 0x0c000000 +#define A3XX_VPC_VARYING_INTERP_MODE_CD__SHIFT 26 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CD(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CD__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CD__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_CE__MASK 0x30000000 +#define A3XX_VPC_VARYING_INTERP_MODE_CE__SHIFT 28 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CE(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CE__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CE__MASK; +} +#define A3XX_VPC_VARYING_INTERP_MODE_CF__MASK 0xc0000000 +#define A3XX_VPC_VARYING_INTERP_MODE_CF__SHIFT 30 +static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CF(enum a3xx_intp_mode val) +{ + return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CF__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CF__MASK; +} static inline uint32_t REG_A3XX_VPC_VARYING_PS_REPL(uint32_t i0) { return 0x00002286 + 0x1*i0; } @@ -1928,6 +2095,8 @@ static inline uint32_t A3XX_SP_FS_MRT_REG_REGID(uint32_t val) return ((val) << A3XX_SP_FS_MRT_REG_REGID__SHIFT) & A3XX_SP_FS_MRT_REG_REGID__MASK; } #define A3XX_SP_FS_MRT_REG_HALF_PRECISION 0x00000100 +#define A3XX_SP_FS_MRT_REG_SINT 0x00000400 +#define A3XX_SP_FS_MRT_REG_UINT 0x00000800 static inline uint32_t REG_A3XX_SP_FS_IMAGE_OUTPUT(uint32_t i0) { return 0x000022f4 + 0x1*i0; } @@ -1947,6 +2116,8 @@ static inline uint32_t A3XX_SP_FS_LENGTH_REG_SHADERLENGTH(uint32_t val) return ((val) << A3XX_SP_FS_LENGTH_REG_SHADERLENGTH__SHIFT) & A3XX_SP_FS_LENGTH_REG_SHADERLENGTH__MASK; } +#define REG_A3XX_PA_SC_AA_CONFIG 0x00002301 + #define REG_A3XX_TPL1_TP_VS_TEX_OFFSET 0x00002340 #define A3XX_TPL1_TP_VS_TEX_OFFSET_SAMPLEROFFSET__MASK 0x000000ff #define A3XX_TPL1_TP_VS_TEX_OFFSET_SAMPLEROFFSET__SHIFT 0 @@ -2297,11 +2468,11 @@ static inline uint32_t A3XX_VGT_DRAW_INITIATOR_INDEX_SIZE(enum pc_di_index_size #define A3XX_VGT_DRAW_INITIATOR_NOT_EOP 0x00001000 #define A3XX_VGT_DRAW_INITIATOR_SMALL_INDEX 0x00002000 #define A3XX_VGT_DRAW_INITIATOR_PRE_DRAW_INITIATOR_ENABLE 0x00004000 -#define A3XX_VGT_DRAW_INITIATOR_NUM_INDICES__MASK 0xffff0000 -#define A3XX_VGT_DRAW_INITIATOR_NUM_INDICES__SHIFT 16 -static inline uint32_t A3XX_VGT_DRAW_INITIATOR_NUM_INDICES(uint32_t val) +#define A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__MASK 0xff000000 +#define A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__SHIFT 24 +static inline uint32_t A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES(uint32_t val) { - return ((val) << A3XX_VGT_DRAW_INITIATOR_NUM_INDICES__SHIFT) & A3XX_VGT_DRAW_INITIATOR_NUM_INDICES__MASK; + return ((val) << A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__SHIFT) & A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__MASK; } #define REG_A3XX_VGT_IMMED_DATA 0x000021fd @@ -2347,17 +2518,23 @@ static inline uint32_t A3XX_TEX_SAMP_0_COMPARE_FUNC(enum adreno_compare_func val #define A3XX_TEX_SAMP_0_UNNORM_COORDS 0x80000000 #define REG_A3XX_TEX_SAMP_1 0x00000001 +#define A3XX_TEX_SAMP_1_LOD_BIAS__MASK 0x000007ff +#define A3XX_TEX_SAMP_1_LOD_BIAS__SHIFT 0 +static inline uint32_t A3XX_TEX_SAMP_1_LOD_BIAS(float val) +{ + return ((((int32_t)(val * 64.0))) << A3XX_TEX_SAMP_1_LOD_BIAS__SHIFT) & A3XX_TEX_SAMP_1_LOD_BIAS__MASK; +} #define A3XX_TEX_SAMP_1_MAX_LOD__MASK 0x003ff000 #define A3XX_TEX_SAMP_1_MAX_LOD__SHIFT 12 static inline uint32_t A3XX_TEX_SAMP_1_MAX_LOD(float val) { - return ((((uint32_t)(val * 12.0))) << A3XX_TEX_SAMP_1_MAX_LOD__SHIFT) & A3XX_TEX_SAMP_1_MAX_LOD__MASK; + return ((((uint32_t)(val * 64.0))) << A3XX_TEX_SAMP_1_MAX_LOD__SHIFT) & A3XX_TEX_SAMP_1_MAX_LOD__MASK; } #define A3XX_TEX_SAMP_1_MIN_LOD__MASK 0xffc00000 #define A3XX_TEX_SAMP_1_MIN_LOD__SHIFT 22 static inline uint32_t A3XX_TEX_SAMP_1_MIN_LOD(float val) { - return ((((uint32_t)(val * 12.0))) << A3XX_TEX_SAMP_1_MIN_LOD__SHIFT) & A3XX_TEX_SAMP_1_MIN_LOD__MASK; + return ((((uint32_t)(val * 64.0))) << A3XX_TEX_SAMP_1_MIN_LOD__SHIFT) & A3XX_TEX_SAMP_1_MIN_LOD__MASK; } #define REG_A3XX_TEX_CONST_0 0x00000000 @@ -2448,6 +2625,24 @@ static inline uint32_t A3XX_TEX_CONST_2_SWAP(enum a3xx_color_swap val) } #define REG_A3XX_TEX_CONST_3 0x00000003 +#define A3XX_TEX_CONST_3_LAYERSZ1__MASK 0x0000000f +#define A3XX_TEX_CONST_3_LAYERSZ1__SHIFT 0 +static inline uint32_t A3XX_TEX_CONST_3_LAYERSZ1(uint32_t val) +{ + return ((val >> 12) << A3XX_TEX_CONST_3_LAYERSZ1__SHIFT) & A3XX_TEX_CONST_3_LAYERSZ1__MASK; +} +#define A3XX_TEX_CONST_3_DEPTH__MASK 0x0ffe0000 +#define A3XX_TEX_CONST_3_DEPTH__SHIFT 17 +static inline uint32_t A3XX_TEX_CONST_3_DEPTH(uint32_t val) +{ + return ((val) << A3XX_TEX_CONST_3_DEPTH__SHIFT) & A3XX_TEX_CONST_3_DEPTH__MASK; +} +#define A3XX_TEX_CONST_3_LAYERSZ2__MASK 0xf0000000 +#define A3XX_TEX_CONST_3_LAYERSZ2__SHIFT 28 +static inline uint32_t A3XX_TEX_CONST_3_LAYERSZ2(uint32_t val) +{ + return ((val >> 12) << A3XX_TEX_CONST_3_LAYERSZ2__SHIFT) & A3XX_TEX_CONST_3_LAYERSZ2__MASK; +} #endif /* A3XX_XML */ diff --git a/drivers/gpu/drm/msm/adreno/a4xx.xml.h b/drivers/gpu/drm/msm/adreno/a4xx.xml.h new file mode 100644 index 0000000..5a24c41 --- /dev/null +++ b/drivers/gpu/drm/msm/adreno/a4xx.xml.h @@ -0,0 +1,2144 @@ +#ifndef A4XX_XML +#define A4XX_XML + +/* Autogenerated file, DO NOT EDIT manually! + +This file was generated by the rules-ng-ng headergen tool in this git repository: +http://github.com/freedreno/envytools/ +git clone https://github.com/freedreno/envytools.git + +The rules-ng-ng source files this header was generated from are: +- /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 364 bytes, from 2013-11-30 14:47:15) +- /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2014-06-02 15:21:30) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10551 bytes, from 2014-11-13 22:44:30) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15053 bytes, from 2014-11-09 15:45:47) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 63169 bytes, from 2014-11-13 22:44:18) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 49097 bytes, from 2014-11-14 15:38:00) + +Copyright (C) 2013-2014 by the following authors: +- Rob Clark (robclark) + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +*/ + + +enum a4xx_color_fmt { + RB4_A8_UNORM = 1, + RB4_R5G6R5_UNORM = 14, + RB4_Z16_UNORM = 15, + RB4_R8G8B8_UNORM = 25, + RB4_R8G8B8A8_UNORM = 26, +}; + +enum a4xx_tile_mode { + TILE4_LINEAR = 0, + TILE4_3 = 3, +}; + +enum a4xx_rb_blend_opcode { + BLEND_DST_PLUS_SRC = 0, + BLEND_SRC_MINUS_DST = 1, + BLEND_DST_MINUS_SRC = 2, + BLEND_MIN_DST_SRC = 3, + BLEND_MAX_DST_SRC = 4, +}; + +enum a4xx_vtx_fmt { + VFMT4_FLOAT_32 = 1, + VFMT4_FLOAT_32_32 = 2, + VFMT4_FLOAT_32_32_32 = 3, + VFMT4_FLOAT_32_32_32_32 = 4, + VFMT4_FLOAT_16 = 5, + VFMT4_FLOAT_16_16 = 6, + VFMT4_FLOAT_16_16_16 = 7, + VFMT4_FLOAT_16_16_16_16 = 8, + VFMT4_FIXED_32 = 9, + VFMT4_FIXED_32_32 = 10, + VFMT4_FIXED_32_32_32 = 11, + VFMT4_FIXED_32_32_32_32 = 12, + VFMT4_SHORT_16 = 16, + VFMT4_SHORT_16_16 = 17, + VFMT4_SHORT_16_16_16 = 18, + VFMT4_SHORT_16_16_16_16 = 19, + VFMT4_USHORT_16 = 20, + VFMT4_USHORT_16_16 = 21, + VFMT4_USHORT_16_16_16 = 22, + VFMT4_USHORT_16_16_16_16 = 23, + VFMT4_NORM_SHORT_16 = 24, + VFMT4_NORM_SHORT_16_16 = 25, + VFMT4_NORM_SHORT_16_16_16 = 26, + VFMT4_NORM_SHORT_16_16_16_16 = 27, + VFMT4_NORM_USHORT_16 = 28, + VFMT4_NORM_USHORT_16_16 = 29, + VFMT4_NORM_USHORT_16_16_16 = 30, + VFMT4_NORM_USHORT_16_16_16_16 = 31, + VFMT4_UBYTE_8 = 40, + VFMT4_UBYTE_8_8 = 41, + VFMT4_UBYTE_8_8_8 = 42, + VFMT4_UBYTE_8_8_8_8 = 43, + VFMT4_NORM_UBYTE_8 = 44, + VFMT4_NORM_UBYTE_8_8 = 45, + VFMT4_NORM_UBYTE_8_8_8 = 46, + VFMT4_NORM_UBYTE_8_8_8_8 = 47, + VFMT4_BYTE_8 = 48, + VFMT4_BYTE_8_8 = 49, + VFMT4_BYTE_8_8_8 = 50, + VFMT4_BYTE_8_8_8_8 = 51, + VFMT4_NORM_BYTE_8 = 52, + VFMT4_NORM_BYTE_8_8 = 53, + VFMT4_NORM_BYTE_8_8_8 = 54, + VFMT4_NORM_BYTE_8_8_8_8 = 55, + VFMT4_UINT_10_10_10_2 = 60, + VFMT4_NORM_UINT_10_10_10_2 = 61, + VFMT4_INT_10_10_10_2 = 62, + VFMT4_NORM_INT_10_10_10_2 = 63, +}; + +enum a4xx_tex_fmt { + TFMT4_NORM_USHORT_565 = 11, + TFMT4_NORM_USHORT_5551 = 10, + TFMT4_NORM_USHORT_4444 = 8, + TFMT4_NORM_UINT_X8Z24 = 71, + TFMT4_NORM_UINT_2_10_10_10 = 33, + TFMT4_NORM_UINT_A8 = 3, + TFMT4_NORM_UINT_L8_A8 = 13, + TFMT4_NORM_UINT_8 = 4, + TFMT4_NORM_UINT_8_8_8_8 = 28, + TFMT4_FLOAT_16 = 20, + TFMT4_FLOAT_16_16 = 40, + TFMT4_FLOAT_16_16_16_16 = 53, + TFMT4_FLOAT_32 = 43, + TFMT4_FLOAT_32_32 = 56, + TFMT4_FLOAT_32_32_32_32 = 63, +}; + +enum a4xx_depth_format { + DEPTH4_NONE = 0, + DEPTH4_16 = 1, + DEPTH4_24_8 = 2, +}; + +enum a4xx_tex_filter { + A4XX_TEX_NEAREST = 0, + A4XX_TEX_LINEAR = 1, +}; + +enum a4xx_tex_clamp { + A4XX_TEX_REPEAT = 0, + A4XX_TEX_CLAMP_TO_EDGE = 1, + A4XX_TEX_MIRROR_REPEAT = 2, + A4XX_TEX_CLAMP_NONE = 3, +}; + +enum a4xx_tex_swiz { + A4XX_TEX_X = 0, + A4XX_TEX_Y = 1, + A4XX_TEX_Z = 2, + A4XX_TEX_W = 3, + A4XX_TEX_ZERO = 4, + A4XX_TEX_ONE = 5, +}; + +enum a4xx_tex_type { + A4XX_TEX_1D = 0, + A4XX_TEX_2D = 1, + A4XX_TEX_CUBE = 2, + A4XX_TEX_3D = 3, +}; + +#define A4XX_CGC_HLSQ_EARLY_CYC__MASK 0x00700000 +#define A4XX_CGC_HLSQ_EARLY_CYC__SHIFT 20 +static inline uint32_t A4XX_CGC_HLSQ_EARLY_CYC(uint32_t val) +{ + return ((val) << A4XX_CGC_HLSQ_EARLY_CYC__SHIFT) & A4XX_CGC_HLSQ_EARLY_CYC__MASK; +} +#define A4XX_INT0_RBBM_GPU_IDLE 0x00000001 +#define A4XX_INT0_RBBM_AHB_ERROR 0x00000002 +#define A4XX_INT0_RBBM_REG_TIMEOUT 0x00000004 +#define A4XX_INT0_RBBM_ME_MS_TIMEOUT 0x00000008 +#define A4XX_INT0_RBBM_PFP_MS_TIMEOUT 0x00000010 +#define A4XX_INT0_RBBM_ATB_BUS_OVERFLOW 0x00000020 +#define A4XX_INT0_VFD_ERROR 0x00000040 +#define A4XX_INT0_CP_SW_INT 0x00000080 +#define A4XX_INT0_CP_T0_PACKET_IN_IB 0x00000100 +#define A4XX_INT0_CP_OPCODE_ERROR 0x00000200 +#define A4XX_INT0_CP_RESERVED_BIT_ERROR 0x00000400 +#define A4XX_INT0_CP_HW_FAULT 0x00000800 +#define A4XX_INT0_CP_DMA 0x00001000 +#define A4XX_INT0_CP_IB2_INT 0x00002000 +#define A4XX_INT0_CP_IB1_INT 0x00004000 +#define A4XX_INT0_CP_RB_INT 0x00008000 +#define A4XX_INT0_CP_REG_PROTECT_FAULT 0x00010000 +#define A4XX_INT0_CP_RB_DONE_TS 0x00020000 +#define A4XX_INT0_CP_VS_DONE_TS 0x00040000 +#define A4XX_INT0_CP_PS_DONE_TS 0x00080000 +#define A4XX_INT0_CACHE_FLUSH_TS 0x00100000 +#define A4XX_INT0_CP_AHB_ERROR_HALT 0x00200000 +#define A4XX_INT0_MISC_HANG_DETECT 0x01000000 +#define A4XX_INT0_UCHE_OOB_ACCESS 0x02000000 +#define REG_A4XX_RB_GMEM_BASE_ADDR 0x00000cc0 + +#define REG_A4XX_RB_PERFCTR_RB_SEL_0 0x00000cc7 + +#define REG_A4XX_RB_PERFCTR_RB_SEL_1 0x00000cc8 + +#define REG_A4XX_RB_PERFCTR_RB_SEL_2 0x00000cc9 + +#define REG_A4XX_RB_PERFCTR_RB_SEL_3 0x00000cca + +#define REG_A4XX_RB_PERFCTR_RB_SEL_4 0x00000ccb + +#define REG_A4XX_RB_PERFCTR_RB_SEL_5 0x00000ccc + +#define REG_A4XX_RB_PERFCTR_RB_SEL_6 0x00000ccd + +#define REG_A4XX_RB_PERFCTR_RB_SEL_7 0x00000cce + +#define REG_A4XX_RB_PERFCTR_CCU_SEL_3 0x00000cd2 + +#define REG_A4XX_RB_FRAME_BUFFER_DIMENSION 0x00000ce0 +#define A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH__MASK 0x00003fff +#define A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH__SHIFT 0 +static inline uint32_t A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH(uint32_t val) +{ + return ((val) << A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH__SHIFT) & A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH__MASK; +} +#define A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT__MASK 0x3fff0000 +#define A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT__SHIFT 16 +static inline uint32_t A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT(uint32_t val) +{ + return ((val) << A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT__SHIFT) & A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT__MASK; +} + +#define REG_A4XX_RB_CLEAR_COLOR_DW0 0x000020cc + +#define REG_A4XX_RB_CLEAR_COLOR_DW1 0x000020cd + +#define REG_A4XX_RB_CLEAR_COLOR_DW2 0x000020ce + +#define REG_A4XX_RB_CLEAR_COLOR_DW3 0x000020cf + +#define REG_A4XX_RB_MODE_CONTROL 0x000020a0 +#define A4XX_RB_MODE_CONTROL_WIDTH__MASK 0x0000003f +#define A4XX_RB_MODE_CONTROL_WIDTH__SHIFT 0 +static inline uint32_t A4XX_RB_MODE_CONTROL_WIDTH(uint32_t val) +{ + return ((val >> 5) << A4XX_RB_MODE_CONTROL_WIDTH__SHIFT) & A4XX_RB_MODE_CONTROL_WIDTH__MASK; +} +#define A4XX_RB_MODE_CONTROL_HEIGHT__MASK 0x00003f00 +#define A4XX_RB_MODE_CONTROL_HEIGHT__SHIFT 8 +static inline uint32_t A4XX_RB_MODE_CONTROL_HEIGHT(uint32_t val) +{ + return ((val >> 5) << A4XX_RB_MODE_CONTROL_HEIGHT__SHIFT) & A4XX_RB_MODE_CONTROL_HEIGHT__MASK; +} + +#define REG_A4XX_RB_RENDER_CONTROL 0x000020a1 +#define A4XX_RB_RENDER_CONTROL_BINNING_PASS 0x00000001 +#define A4XX_RB_RENDER_CONTROL_DISABLE_COLOR_PIPE 0x00000020 + +#define REG_A4XX_RB_MSAA_CONTROL 0x000020a2 +#define A4XX_RB_MSAA_CONTROL_DISABLE 0x00001000 +#define A4XX_RB_MSAA_CONTROL_SAMPLES__MASK 0x0000e000 +#define A4XX_RB_MSAA_CONTROL_SAMPLES__SHIFT 13 +static inline uint32_t A4XX_RB_MSAA_CONTROL_SAMPLES(uint32_t val) +{ + return ((val) << A4XX_RB_MSAA_CONTROL_SAMPLES__SHIFT) & A4XX_RB_MSAA_CONTROL_SAMPLES__MASK; +} + +#define REG_A4XX_RB_MSAA_CONTROL2 0x000020a3 +#define A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES__MASK 0x00000380 +#define A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES__SHIFT 7 +static inline uint32_t A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES(uint32_t val) +{ + return ((val) << A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES__SHIFT) & A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES__MASK; +} +#define A4XX_RB_MSAA_CONTROL2_VARYING 0x00001000 + +static inline uint32_t REG_A4XX_RB_MRT(uint32_t i0) { return 0x000020a4 + 0x5*i0; } + +static inline uint32_t REG_A4XX_RB_MRT_CONTROL(uint32_t i0) { return 0x000020a4 + 0x5*i0; } +#define A4XX_RB_MRT_CONTROL_READ_DEST_ENABLE 0x00000008 +#define A4XX_RB_MRT_CONTROL_BLEND 0x00000010 +#define A4XX_RB_MRT_CONTROL_BLEND2 0x00000020 +#define A4XX_RB_MRT_CONTROL_FASTCLEAR 0x00000400 +#define A4XX_RB_MRT_CONTROL_B11 0x00000800 +#define A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__MASK 0x0f000000 +#define A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__SHIFT 24 +static inline uint32_t A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE(uint32_t val) +{ + return ((val) << A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__SHIFT) & A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__MASK; +} + +static inline uint32_t REG_A4XX_RB_MRT_BUF_INFO(uint32_t i0) { return 0x000020a5 + 0x5*i0; } +#define A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT__MASK 0x0000003f +#define A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT__SHIFT 0 +static inline uint32_t A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT(enum a4xx_color_fmt val) +{ + return ((val) << A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT__SHIFT) & A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT__MASK; +} +#define A4XX_RB_MRT_BUF_INFO_DITHER_MODE__MASK 0x00000600 +#define A4XX_RB_MRT_BUF_INFO_DITHER_MODE__SHIFT 9 +static inline uint32_t A4XX_RB_MRT_BUF_INFO_DITHER_MODE(enum adreno_rb_dither_mode val) +{ + return ((val) << A4XX_RB_MRT_BUF_INFO_DITHER_MODE__SHIFT) & A4XX_RB_MRT_BUF_INFO_DITHER_MODE__MASK; +} +#define A4XX_RB_MRT_BUF_INFO_COLOR_SWAP__MASK 0x00001800 +#define A4XX_RB_MRT_BUF_INFO_COLOR_SWAP__SHIFT 11 +static inline uint32_t A4XX_RB_MRT_BUF_INFO_COLOR_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A4XX_RB_MRT_BUF_INFO_COLOR_SWAP__SHIFT) & A4XX_RB_MRT_BUF_INFO_COLOR_SWAP__MASK; +} +#define A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__MASK 0x007fc000 +#define A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__SHIFT 14 +static inline uint32_t A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH(uint32_t val) +{ + return ((val >> 4) << A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__SHIFT) & A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__MASK; +} + +static inline uint32_t REG_A4XX_RB_MRT_BASE(uint32_t i0) { return 0x000020a6 + 0x5*i0; } + +static inline uint32_t REG_A4XX_RB_MRT_CONTROL3(uint32_t i0) { return 0x000020a7 + 0x5*i0; } +#define A4XX_RB_MRT_CONTROL3_STRIDE__MASK 0x0001fff8 +#define A4XX_RB_MRT_CONTROL3_STRIDE__SHIFT 3 +static inline uint32_t A4XX_RB_MRT_CONTROL3_STRIDE(uint32_t val) +{ + return ((val) << A4XX_RB_MRT_CONTROL3_STRIDE__SHIFT) & A4XX_RB_MRT_CONTROL3_STRIDE__MASK; +} + +static inline uint32_t REG_A4XX_RB_MRT_BLEND_CONTROL(uint32_t i0) { return 0x000020a8 + 0x5*i0; } +#define A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__MASK 0x0000001f +#define A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__SHIFT 0 +static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR(enum adreno_rb_blend_factor val) +{ + return ((val) << A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__MASK; +} +#define A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__MASK 0x000000e0 +#define A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__SHIFT 5 +static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE(enum a4xx_rb_blend_opcode val) +{ + return ((val) << A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__MASK; +} +#define A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__MASK 0x00001f00 +#define A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__SHIFT 8 +static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR(enum adreno_rb_blend_factor val) +{ + return ((val) << A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__MASK; +} +#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__MASK 0x001f0000 +#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__SHIFT 16 +static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR(enum adreno_rb_blend_factor val) +{ + return ((val) << A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__MASK; +} +#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__MASK 0x00e00000 +#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__SHIFT 21 +static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE(enum a4xx_rb_blend_opcode val) +{ + return ((val) << A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__MASK; +} +#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__MASK 0x1f000000 +#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__SHIFT 24 +static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR(enum adreno_rb_blend_factor val) +{ + return ((val) << A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__MASK; +} + +#define REG_A4XX_RB_ALPHA_CONTROL 0x000020f8 +#define A4XX_RB_ALPHA_CONTROL_ALPHA_TEST 0x00000100 +#define A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__MASK 0x00000e00 +#define A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__SHIFT 9 +static inline uint32_t A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC(enum adreno_compare_func val) +{ + return ((val) << A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__SHIFT) & A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__MASK; +} + +#define REG_A4XX_RB_FS_OUTPUT 0x000020f9 +#define A4XX_RB_FS_OUTPUT_ENABLE_COLOR_PIPE 0x00000001 +#define A4XX_RB_FS_OUTPUT_FAST_CLEAR 0x00000100 +#define A4XX_RB_FS_OUTPUT_SAMPLE_MASK__MASK 0xffff0000 +#define A4XX_RB_FS_OUTPUT_SAMPLE_MASK__SHIFT 16 +static inline uint32_t A4XX_RB_FS_OUTPUT_SAMPLE_MASK(uint32_t val) +{ + return ((val) << A4XX_RB_FS_OUTPUT_SAMPLE_MASK__SHIFT) & A4XX_RB_FS_OUTPUT_SAMPLE_MASK__MASK; +} + +#define REG_A4XX_RB_RENDER_CONTROL3 0x000020fb +#define A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE__MASK 0x0000001f +#define A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE__SHIFT 0 +static inline uint32_t A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE(uint32_t val) +{ + return ((val) << A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE__SHIFT) & A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE__MASK; +} + +#define REG_A4XX_RB_COPY_CONTROL 0x000020fc +#define A4XX_RB_COPY_CONTROL_MSAA_RESOLVE__MASK 0x00000003 +#define A4XX_RB_COPY_CONTROL_MSAA_RESOLVE__SHIFT 0 +static inline uint32_t A4XX_RB_COPY_CONTROL_MSAA_RESOLVE(enum a3xx_msaa_samples val) +{ + return ((val) << A4XX_RB_COPY_CONTROL_MSAA_RESOLVE__SHIFT) & A4XX_RB_COPY_CONTROL_MSAA_RESOLVE__MASK; +} +#define A4XX_RB_COPY_CONTROL_MODE__MASK 0x00000070 +#define A4XX_RB_COPY_CONTROL_MODE__SHIFT 4 +static inline uint32_t A4XX_RB_COPY_CONTROL_MODE(enum adreno_rb_copy_control_mode val) +{ + return ((val) << A4XX_RB_COPY_CONTROL_MODE__SHIFT) & A4XX_RB_COPY_CONTROL_MODE__MASK; +} +#define A4XX_RB_COPY_CONTROL_FASTCLEAR__MASK 0x00000f00 +#define A4XX_RB_COPY_CONTROL_FASTCLEAR__SHIFT 8 +static inline uint32_t A4XX_RB_COPY_CONTROL_FASTCLEAR(uint32_t val) +{ + return ((val) << A4XX_RB_COPY_CONTROL_FASTCLEAR__SHIFT) & A4XX_RB_COPY_CONTROL_FASTCLEAR__MASK; +} +#define A4XX_RB_COPY_CONTROL_GMEM_BASE__MASK 0xffffc000 +#define A4XX_RB_COPY_CONTROL_GMEM_BASE__SHIFT 14 +static inline uint32_t A4XX_RB_COPY_CONTROL_GMEM_BASE(uint32_t val) +{ + return ((val >> 14) << A4XX_RB_COPY_CONTROL_GMEM_BASE__SHIFT) & A4XX_RB_COPY_CONTROL_GMEM_BASE__MASK; +} + +#define REG_A4XX_RB_COPY_DEST_BASE 0x000020fd +#define A4XX_RB_COPY_DEST_BASE_BASE__MASK 0xfffffff0 +#define A4XX_RB_COPY_DEST_BASE_BASE__SHIFT 4 +static inline uint32_t A4XX_RB_COPY_DEST_BASE_BASE(uint32_t val) +{ + return ((val >> 4) << A4XX_RB_COPY_DEST_BASE_BASE__SHIFT) & A4XX_RB_COPY_DEST_BASE_BASE__MASK; +} + +#define REG_A4XX_RB_COPY_DEST_PITCH 0x000020fe +#define A4XX_RB_COPY_DEST_PITCH_PITCH__MASK 0xffffffff +#define A4XX_RB_COPY_DEST_PITCH_PITCH__SHIFT 0 +static inline uint32_t A4XX_RB_COPY_DEST_PITCH_PITCH(uint32_t val) +{ + return ((val >> 5) << A4XX_RB_COPY_DEST_PITCH_PITCH__SHIFT) & A4XX_RB_COPY_DEST_PITCH_PITCH__MASK; +} + +#define REG_A4XX_RB_COPY_DEST_INFO 0x000020ff +#define A4XX_RB_COPY_DEST_INFO_FORMAT__MASK 0x000000fc +#define A4XX_RB_COPY_DEST_INFO_FORMAT__SHIFT 2 +static inline uint32_t A4XX_RB_COPY_DEST_INFO_FORMAT(enum a4xx_color_fmt val) +{ + return ((val) << A4XX_RB_COPY_DEST_INFO_FORMAT__SHIFT) & A4XX_RB_COPY_DEST_INFO_FORMAT__MASK; +} +#define A4XX_RB_COPY_DEST_INFO_SWAP__MASK 0x00000300 +#define A4XX_RB_COPY_DEST_INFO_SWAP__SHIFT 8 +static inline uint32_t A4XX_RB_COPY_DEST_INFO_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A4XX_RB_COPY_DEST_INFO_SWAP__SHIFT) & A4XX_RB_COPY_DEST_INFO_SWAP__MASK; +} +#define A4XX_RB_COPY_DEST_INFO_DITHER_MODE__MASK 0x00000c00 +#define A4XX_RB_COPY_DEST_INFO_DITHER_MODE__SHIFT 10 +static inline uint32_t A4XX_RB_COPY_DEST_INFO_DITHER_MODE(enum adreno_rb_dither_mode val) +{ + return ((val) << A4XX_RB_COPY_DEST_INFO_DITHER_MODE__SHIFT) & A4XX_RB_COPY_DEST_INFO_DITHER_MODE__MASK; +} +#define A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE__MASK 0x0003c000 +#define A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE__SHIFT 14 +static inline uint32_t A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE(uint32_t val) +{ + return ((val) << A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE__SHIFT) & A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE__MASK; +} +#define A4XX_RB_COPY_DEST_INFO_ENDIAN__MASK 0x001c0000 +#define A4XX_RB_COPY_DEST_INFO_ENDIAN__SHIFT 18 +static inline uint32_t A4XX_RB_COPY_DEST_INFO_ENDIAN(enum adreno_rb_surface_endian val) +{ + return ((val) << A4XX_RB_COPY_DEST_INFO_ENDIAN__SHIFT) & A4XX_RB_COPY_DEST_INFO_ENDIAN__MASK; +} +#define A4XX_RB_COPY_DEST_INFO_TILE__MASK 0x03000000 +#define A4XX_RB_COPY_DEST_INFO_TILE__SHIFT 24 +static inline uint32_t A4XX_RB_COPY_DEST_INFO_TILE(enum a4xx_tile_mode val) +{ + return ((val) << A4XX_RB_COPY_DEST_INFO_TILE__SHIFT) & A4XX_RB_COPY_DEST_INFO_TILE__MASK; +} + +#define REG_A4XX_RB_FS_OUTPUT_REG 0x00002100 +#define A4XX_RB_FS_OUTPUT_REG_COLOR_PIPE_ENABLE 0x00000001 +#define A4XX_RB_FS_OUTPUT_REG_FRAG_WRITES_Z 0x00000020 + +#define REG_A4XX_RB_DEPTH_CONTROL 0x00002101 +#define A4XX_RB_DEPTH_CONTROL_FRAG_WRITES_Z 0x00000001 +#define A4XX_RB_DEPTH_CONTROL_Z_ENABLE 0x00000002 +#define A4XX_RB_DEPTH_CONTROL_Z_WRITE_ENABLE 0x00000004 +#define A4XX_RB_DEPTH_CONTROL_ZFUNC__MASK 0x00000070 +#define A4XX_RB_DEPTH_CONTROL_ZFUNC__SHIFT 4 +static inline uint32_t A4XX_RB_DEPTH_CONTROL_ZFUNC(enum adreno_compare_func val) +{ + return ((val) << A4XX_RB_DEPTH_CONTROL_ZFUNC__SHIFT) & A4XX_RB_DEPTH_CONTROL_ZFUNC__MASK; +} +#define A4XX_RB_DEPTH_CONTROL_BF_ENABLE 0x00000080 +#define A4XX_RB_DEPTH_CONTROL_EARLY_Z_DISABLE 0x00010000 +#define A4XX_RB_DEPTH_CONTROL_Z_TEST_ENABLE 0x80000000 + +#define REG_A4XX_RB_DEPTH_CLEAR 0x00002102 + +#define REG_A4XX_RB_DEPTH_INFO 0x00002103 +#define A4XX_RB_DEPTH_INFO_DEPTH_FORMAT__MASK 0x00000003 +#define A4XX_RB_DEPTH_INFO_DEPTH_FORMAT__SHIFT 0 +static inline uint32_t A4XX_RB_DEPTH_INFO_DEPTH_FORMAT(enum a4xx_depth_format val) +{ + return ((val) << A4XX_RB_DEPTH_INFO_DEPTH_FORMAT__SHIFT) & A4XX_RB_DEPTH_INFO_DEPTH_FORMAT__MASK; +} +#define A4XX_RB_DEPTH_INFO_DEPTH_BASE__MASK 0xfffff000 +#define A4XX_RB_DEPTH_INFO_DEPTH_BASE__SHIFT 12 +static inline uint32_t A4XX_RB_DEPTH_INFO_DEPTH_BASE(uint32_t val) +{ + return ((val >> 12) << A4XX_RB_DEPTH_INFO_DEPTH_BASE__SHIFT) & A4XX_RB_DEPTH_INFO_DEPTH_BASE__MASK; +} + +#define REG_A4XX_RB_DEPTH_PITCH 0x00002104 +#define A4XX_RB_DEPTH_PITCH__MASK 0xffffffff +#define A4XX_RB_DEPTH_PITCH__SHIFT 0 +static inline uint32_t A4XX_RB_DEPTH_PITCH(uint32_t val) +{ + return ((val >> 4) << A4XX_RB_DEPTH_PITCH__SHIFT) & A4XX_RB_DEPTH_PITCH__MASK; +} + +#define REG_A4XX_RB_DEPTH_PITCH2 0x00002105 +#define A4XX_RB_DEPTH_PITCH2__MASK 0xffffffff +#define A4XX_RB_DEPTH_PITCH2__SHIFT 0 +static inline uint32_t A4XX_RB_DEPTH_PITCH2(uint32_t val) +{ + return ((val >> 4) << A4XX_RB_DEPTH_PITCH2__SHIFT) & A4XX_RB_DEPTH_PITCH2__MASK; +} + +#define REG_A4XX_RB_STENCIL_CONTROL 0x00002106 +#define A4XX_RB_STENCIL_CONTROL_STENCIL_ENABLE 0x00000001 +#define A4XX_RB_STENCIL_CONTROL_STENCIL_ENABLE_BF 0x00000002 +#define A4XX_RB_STENCIL_CONTROL_STENCIL_READ 0x00000004 +#define A4XX_RB_STENCIL_CONTROL_FUNC__MASK 0x00000700 +#define A4XX_RB_STENCIL_CONTROL_FUNC__SHIFT 8 +static inline uint32_t A4XX_RB_STENCIL_CONTROL_FUNC(enum adreno_compare_func val) +{ + return ((val) << A4XX_RB_STENCIL_CONTROL_FUNC__SHIFT) & A4XX_RB_STENCIL_CONTROL_FUNC__MASK; +} +#define A4XX_RB_STENCIL_CONTROL_FAIL__MASK 0x00003800 +#define A4XX_RB_STENCIL_CONTROL_FAIL__SHIFT 11 +static inline uint32_t A4XX_RB_STENCIL_CONTROL_FAIL(enum adreno_stencil_op val) +{ + return ((val) << A4XX_RB_STENCIL_CONTROL_FAIL__SHIFT) & A4XX_RB_STENCIL_CONTROL_FAIL__MASK; +} +#define A4XX_RB_STENCIL_CONTROL_ZPASS__MASK 0x0001c000 +#define A4XX_RB_STENCIL_CONTROL_ZPASS__SHIFT 14 +static inline uint32_t A4XX_RB_STENCIL_CONTROL_ZPASS(enum adreno_stencil_op val) +{ + return ((val) << A4XX_RB_STENCIL_CONTROL_ZPASS__SHIFT) & A4XX_RB_STENCIL_CONTROL_ZPASS__MASK; +} +#define A4XX_RB_STENCIL_CONTROL_ZFAIL__MASK 0x000e0000 +#define A4XX_RB_STENCIL_CONTROL_ZFAIL__SHIFT 17 +static inline uint32_t A4XX_RB_STENCIL_CONTROL_ZFAIL(enum adreno_stencil_op val) +{ + return ((val) << A4XX_RB_STENCIL_CONTROL_ZFAIL__SHIFT) & A4XX_RB_STENCIL_CONTROL_ZFAIL__MASK; +} +#define A4XX_RB_STENCIL_CONTROL_FUNC_BF__MASK 0x00700000 +#define A4XX_RB_STENCIL_CONTROL_FUNC_BF__SHIFT 20 +static inline uint32_t A4XX_RB_STENCIL_CONTROL_FUNC_BF(enum adreno_compare_func val) +{ + return ((val) << A4XX_RB_STENCIL_CONTROL_FUNC_BF__SHIFT) & A4XX_RB_STENCIL_CONTROL_FUNC_BF__MASK; +} +#define A4XX_RB_STENCIL_CONTROL_FAIL_BF__MASK 0x03800000 +#define A4XX_RB_STENCIL_CONTROL_FAIL_BF__SHIFT 23 +static inline uint32_t A4XX_RB_STENCIL_CONTROL_FAIL_BF(enum adreno_stencil_op val) +{ + return ((val) << A4XX_RB_STENCIL_CONTROL_FAIL_BF__SHIFT) & A4XX_RB_STENCIL_CONTROL_FAIL_BF__MASK; +} +#define A4XX_RB_STENCIL_CONTROL_ZPASS_BF__MASK 0x1c000000 +#define A4XX_RB_STENCIL_CONTROL_ZPASS_BF__SHIFT 26 +static inline uint32_t A4XX_RB_STENCIL_CONTROL_ZPASS_BF(enum adreno_stencil_op val) +{ + return ((val) << A4XX_RB_STENCIL_CONTROL_ZPASS_BF__SHIFT) & A4XX_RB_STENCIL_CONTROL_ZPASS_BF__MASK; +} +#define A4XX_RB_STENCIL_CONTROL_ZFAIL_BF__MASK 0xe0000000 +#define A4XX_RB_STENCIL_CONTROL_ZFAIL_BF__SHIFT 29 +static inline uint32_t A4XX_RB_STENCIL_CONTROL_ZFAIL_BF(enum adreno_stencil_op val) +{ + return ((val) << A4XX_RB_STENCIL_CONTROL_ZFAIL_BF__SHIFT) & A4XX_RB_STENCIL_CONTROL_ZFAIL_BF__MASK; +} + +#define REG_A4XX_RB_STENCIL_CONTROL2 0x00002107 +#define A4XX_RB_STENCIL_CONTROL2_STENCIL_BUFFER 0x00000001 + +#define REG_A4XX_RB_STENCILREFMASK 0x0000210b +#define A4XX_RB_STENCILREFMASK_STENCILREF__MASK 0x000000ff +#define A4XX_RB_STENCILREFMASK_STENCILREF__SHIFT 0 +static inline uint32_t A4XX_RB_STENCILREFMASK_STENCILREF(uint32_t val) +{ + return ((val) << A4XX_RB_STENCILREFMASK_STENCILREF__SHIFT) & A4XX_RB_STENCILREFMASK_STENCILREF__MASK; +} +#define A4XX_RB_STENCILREFMASK_STENCILMASK__MASK 0x0000ff00 +#define A4XX_RB_STENCILREFMASK_STENCILMASK__SHIFT 8 +static inline uint32_t A4XX_RB_STENCILREFMASK_STENCILMASK(uint32_t val) +{ + return ((val) << A4XX_RB_STENCILREFMASK_STENCILMASK__SHIFT) & A4XX_RB_STENCILREFMASK_STENCILMASK__MASK; +} +#define A4XX_RB_STENCILREFMASK_STENCILWRITEMASK__MASK 0x00ff0000 +#define A4XX_RB_STENCILREFMASK_STENCILWRITEMASK__SHIFT 16 +static inline uint32_t A4XX_RB_STENCILREFMASK_STENCILWRITEMASK(uint32_t val) +{ + return ((val) << A4XX_RB_STENCILREFMASK_STENCILWRITEMASK__SHIFT) & A4XX_RB_STENCILREFMASK_STENCILWRITEMASK__MASK; +} + +#define REG_A4XX_RB_STENCILREFMASK_BF 0x0000210c +#define A4XX_RB_STENCILREFMASK_BF_STENCILREF__MASK 0x000000ff +#define A4XX_RB_STENCILREFMASK_BF_STENCILREF__SHIFT 0 +static inline uint32_t A4XX_RB_STENCILREFMASK_BF_STENCILREF(uint32_t val) +{ + return ((val) << A4XX_RB_STENCILREFMASK_BF_STENCILREF__SHIFT) & A4XX_RB_STENCILREFMASK_BF_STENCILREF__MASK; +} +#define A4XX_RB_STENCILREFMASK_BF_STENCILMASK__MASK 0x0000ff00 +#define A4XX_RB_STENCILREFMASK_BF_STENCILMASK__SHIFT 8 +static inline uint32_t A4XX_RB_STENCILREFMASK_BF_STENCILMASK(uint32_t val) +{ + return ((val) << A4XX_RB_STENCILREFMASK_BF_STENCILMASK__SHIFT) & A4XX_RB_STENCILREFMASK_BF_STENCILMASK__MASK; +} +#define A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__MASK 0x00ff0000 +#define A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__SHIFT 16 +static inline uint32_t A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK(uint32_t val) +{ + return ((val) << A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__SHIFT) & A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__MASK; +} + +#define REG_A4XX_RB_BIN_OFFSET 0x0000210d +#define A4XX_RB_BIN_OFFSET_WINDOW_OFFSET_DISABLE 0x80000000 +#define A4XX_RB_BIN_OFFSET_X__MASK 0x00007fff +#define A4XX_RB_BIN_OFFSET_X__SHIFT 0 +static inline uint32_t A4XX_RB_BIN_OFFSET_X(uint32_t val) +{ + return ((val) << A4XX_RB_BIN_OFFSET_X__SHIFT) & A4XX_RB_BIN_OFFSET_X__MASK; +} +#define A4XX_RB_BIN_OFFSET_Y__MASK 0x7fff0000 +#define A4XX_RB_BIN_OFFSET_Y__SHIFT 16 +static inline uint32_t A4XX_RB_BIN_OFFSET_Y(uint32_t val) +{ + return ((val) << A4XX_RB_BIN_OFFSET_Y__SHIFT) & A4XX_RB_BIN_OFFSET_Y__MASK; +} + +#define REG_A4XX_RB_VPORT_Z_CLAMP_MAX_15 0x0000213f + +#define REG_A4XX_RBBM_HW_VERSION 0x00000000 + +#define REG_A4XX_RBBM_HW_CONFIGURATION 0x00000002 + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_TP(uint32_t i0) { return 0x00000004 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_TP_REG(uint32_t i0) { return 0x00000004 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_TP(uint32_t i0) { return 0x00000008 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_TP_REG(uint32_t i0) { return 0x00000008 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_TP(uint32_t i0) { return 0x0000000c + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_TP_REG(uint32_t i0) { return 0x0000000c + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_TP(uint32_t i0) { return 0x00000010 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_TP_REG(uint32_t i0) { return 0x00000010 + 0x1*i0; } + +#define REG_A4XX_RBBM_CLOCK_CTL_UCHE 0x00000014 + +#define REG_A4XX_RBBM_CLOCK_CTL2_UCHE 0x00000015 + +#define REG_A4XX_RBBM_CLOCK_CTL3_UCHE 0x00000016 + +#define REG_A4XX_RBBM_CLOCK_CTL4_UCHE 0x00000017 + +#define REG_A4XX_RBBM_CLOCK_HYST_UCHE 0x00000018 + +#define REG_A4XX_RBBM_CLOCK_DELAY_UCHE 0x00000019 + +#define REG_A4XX_RBBM_CLOCK_MODE_GPC 0x0000001a + +#define REG_A4XX_RBBM_CLOCK_DELAY_GPC 0x0000001b + +#define REG_A4XX_RBBM_CLOCK_HYST_GPC 0x0000001c + +#define REG_A4XX_RBBM_CLOCK_CTL_TSE_RAS_RBBM 0x0000001d + +#define REG_A4XX_RBBM_CLOCK_HYST_TSE_RAS_RBBM 0x0000001e + +#define REG_A4XX_RBBM_CLOCK_DELAY_TSE_RAS_RBBM 0x0000001f + +#define REG_A4XX_RBBM_CLOCK_CTL 0x00000020 + +#define REG_A4XX_RBBM_SP_HYST_CNT 0x00000021 + +#define REG_A4XX_RBBM_SW_RESET_CMD 0x00000022 + +#define REG_A4XX_RBBM_AHB_CTL0 0x00000023 + +#define REG_A4XX_RBBM_AHB_CTL1 0x00000024 + +#define REG_A4XX_RBBM_AHB_CMD 0x00000025 + +#define REG_A4XX_RBBM_RB_SUB_BLOCK_SEL_CTL 0x00000026 + +#define REG_A4XX_RBBM_RAM_ACC_63_32 0x00000028 + +#define REG_A4XX_RBBM_WAIT_IDLE_CLOCKS_CTL 0x0000002b + +#define REG_A4XX_RBBM_INTERFACE_HANG_INT_CTL 0x0000002f + +#define REG_A4XX_RBBM_INTERFACE_HANG_MASK_CTL4 0x00000034 + +#define REG_A4XX_RBBM_INT_CLEAR_CMD 0x00000036 + +#define REG_A4XX_RBBM_INT_0_MASK 0x00000037 + +#define REG_A4XX_RBBM_RBBM_CTL 0x0000003e + +#define REG_A4XX_RBBM_AHB_DEBUG_CTL 0x0000003f + +#define REG_A4XX_RBBM_VBIF_DEBUG_CTL 0x00000041 + +#define REG_A4XX_RBBM_CLOCK_CTL2 0x00000042 + +#define REG_A4XX_RBBM_BLOCK_SW_RESET_CMD 0x00000045 + +#define REG_A4XX_RBBM_RESET_CYCLES 0x00000047 + +#define REG_A4XX_RBBM_EXT_TRACE_BUS_CTL 0x00000049 + +#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_A 0x0000004a + +#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_B 0x0000004b + +#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_C 0x0000004c + +#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_D 0x0000004d + +#define REG_A4XX_RBBM_PERFCTR_CP_0_LO 0x0000009c + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_SP(uint32_t i0) { return 0x00000068 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_SP_REG(uint32_t i0) { return 0x00000068 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_SP(uint32_t i0) { return 0x0000006c + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_SP_REG(uint32_t i0) { return 0x0000006c + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_SP(uint32_t i0) { return 0x00000070 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_SP_REG(uint32_t i0) { return 0x00000070 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_SP(uint32_t i0) { return 0x00000074 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_SP_REG(uint32_t i0) { return 0x00000074 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_RB(uint32_t i0) { return 0x00000078 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_RB_REG(uint32_t i0) { return 0x00000078 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_RB(uint32_t i0) { return 0x0000007c + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_RB_REG(uint32_t i0) { return 0x0000007c + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_MARB_CCU(uint32_t i0) { return 0x00000082 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_MARB_CCU_REG(uint32_t i0) { return 0x00000082 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_RB_MARB_CCU(uint32_t i0) { return 0x00000086 + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_RB_MARB_CCU_REG(uint32_t i0) { return 0x00000086 + 0x1*i0; } + +#define REG_A4XX_RBBM_CLOCK_HYST_COM_DCOM 0x00000080 + +#define REG_A4XX_RBBM_CLOCK_CTL_COM_DCOM 0x00000081 + +#define REG_A4XX_RBBM_CLOCK_CTL_HLSQ 0x0000008a + +#define REG_A4XX_RBBM_CLOCK_HYST_HLSQ 0x0000008b + +#define REG_A4XX_RBBM_CLOCK_DELAY_HLSQ 0x0000008c + +#define REG_A4XX_RBBM_CLOCK_DELAY_COM_DCOM 0x0000008d + +static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_RB_MARB_CCU_L1(uint32_t i0) { return 0x0000008e + 0x1*i0; } + +static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_RB_MARB_CCU_L1_REG(uint32_t i0) { return 0x0000008e + 0x1*i0; } + +#define REG_A4XX_RBBM_PERFCTR_PWR_1_LO 0x00000168 + +#define REG_A4XX_RBBM_PERFCTR_CTL 0x00000170 + +#define REG_A4XX_RBBM_PERFCTR_LOAD_CMD0 0x00000171 + +#define REG_A4XX_RBBM_PERFCTR_LOAD_CMD1 0x00000172 + +#define REG_A4XX_RBBM_PERFCTR_LOAD_CMD2 0x00000173 + +#define REG_A4XX_RBBM_PERFCTR_LOAD_VALUE_LO 0x00000174 + +#define REG_A4XX_RBBM_PERFCTR_LOAD_VALUE_HI 0x00000175 + +#define REG_A4XX_RBBM_GPU_BUSY_MASKED 0x0000017a + +#define REG_A4XX_RBBM_INT_0_STATUS 0x0000017d + +#define REG_A4XX_RBBM_CLOCK_STATUS 0x00000182 + +#define REG_A4XX_RBBM_AHB_STATUS 0x00000189 + +#define REG_A4XX_RBBM_AHB_ME_SPLIT_STATUS 0x0000018c + +#define REG_A4XX_RBBM_AHB_PFP_SPLIT_STATUS 0x0000018d + +#define REG_A4XX_RBBM_AHB_ERROR_STATUS 0x0000018f + +#define REG_A4XX_RBBM_STATUS 0x00000191 +#define A4XX_RBBM_STATUS_HI_BUSY 0x00000001 +#define A4XX_RBBM_STATUS_CP_ME_BUSY 0x00000002 +#define A4XX_RBBM_STATUS_CP_PFP_BUSY 0x00000004 +#define A4XX_RBBM_STATUS_CP_NRT_BUSY 0x00004000 +#define A4XX_RBBM_STATUS_VBIF_BUSY 0x00008000 +#define A4XX_RBBM_STATUS_TSE_BUSY 0x00010000 +#define A4XX_RBBM_STATUS_RAS_BUSY 0x00020000 +#define A4XX_RBBM_STATUS_RB_BUSY 0x00040000 +#define A4XX_RBBM_STATUS_PC_DCALL_BUSY 0x00080000 +#define A4XX_RBBM_STATUS_PC_VSD_BUSY 0x00100000 +#define A4XX_RBBM_STATUS_VFD_BUSY 0x00200000 +#define A4XX_RBBM_STATUS_VPC_BUSY 0x00400000 +#define A4XX_RBBM_STATUS_UCHE_BUSY 0x00800000 +#define A4XX_RBBM_STATUS_SP_BUSY 0x01000000 +#define A4XX_RBBM_STATUS_TPL1_BUSY 0x02000000 +#define A4XX_RBBM_STATUS_MARB_BUSY 0x04000000 +#define A4XX_RBBM_STATUS_VSC_BUSY 0x08000000 +#define A4XX_RBBM_STATUS_ARB_BUSY 0x10000000 +#define A4XX_RBBM_STATUS_HLSQ_BUSY 0x20000000 +#define A4XX_RBBM_STATUS_GPU_BUSY_NOHC 0x40000000 +#define A4XX_RBBM_STATUS_GPU_BUSY 0x80000000 + +#define REG_A4XX_RBBM_INTERFACE_RRDY_STATUS5 0x0000019f + +#define REG_A4XX_CP_SCRATCH_UMASK 0x00000228 + +#define REG_A4XX_CP_SCRATCH_ADDR 0x00000229 + +#define REG_A4XX_CP_RB_BASE 0x00000200 + +#define REG_A4XX_CP_RB_CNTL 0x00000201 + +#define REG_A4XX_CP_RB_WPTR 0x00000205 + +#define REG_A4XX_CP_RB_RPTR_ADDR 0x00000203 + +#define REG_A4XX_CP_RB_RPTR 0x00000204 + +#define REG_A4XX_CP_IB1_BASE 0x00000206 + +#define REG_A4XX_CP_IB1_BUFSZ 0x00000207 + +#define REG_A4XX_CP_IB2_BASE 0x00000208 + +#define REG_A4XX_CP_IB2_BUFSZ 0x00000209 + +#define REG_A4XX_CP_ME_RB_DONE_DATA 0x00000217 + +#define REG_A4XX_CP_QUEUE_THRESH2 0x00000219 + +#define REG_A4XX_CP_MERCIU_SIZE 0x0000021b + +#define REG_A4XX_CP_ROQ_ADDR 0x0000021c + +#define REG_A4XX_CP_ROQ_DATA 0x0000021d + +#define REG_A4XX_CP_MEQ_ADDR 0x0000021e + +#define REG_A4XX_CP_MEQ_DATA 0x0000021f + +#define REG_A4XX_CP_MERCIU_ADDR 0x00000220 + +#define REG_A4XX_CP_MERCIU_DATA 0x00000221 + +#define REG_A4XX_CP_MERCIU_DATA2 0x00000222 + +#define REG_A4XX_CP_PFP_UCODE_ADDR 0x00000223 + +#define REG_A4XX_CP_PFP_UCODE_DATA 0x00000224 + +#define REG_A4XX_CP_ME_RAM_WADDR 0x00000225 + +#define REG_A4XX_CP_ME_RAM_RADDR 0x00000226 + +#define REG_A4XX_CP_ME_RAM_DATA 0x00000227 + +#define REG_A4XX_CP_PREEMPT 0x0000022a + +#define REG_A4XX_CP_CNTL 0x0000022c + +#define REG_A4XX_CP_ME_CNTL 0x0000022d + +#define REG_A4XX_CP_DEBUG 0x0000022e + +#define REG_A4XX_CP_DEBUG_ECO_CONTROL 0x00000231 + +#define REG_A4XX_CP_DRAW_STATE_ADDR 0x00000232 + +#define REG_A4XX_CP_PROTECT_REG_0 0x00000240 + +static inline uint32_t REG_A4XX_CP_PROTECT(uint32_t i0) { return 0x00000240 + 0x1*i0; } + +static inline uint32_t REG_A4XX_CP_PROTECT_REG(uint32_t i0) { return 0x00000240 + 0x1*i0; } + +#define REG_A4XX_CP_PROTECT_CTRL 0x00000250 + +#define REG_A4XX_CP_ST_BASE 0x000004c0 + +#define REG_A4XX_CP_STQ_AVAIL 0x000004ce + +#define REG_A4XX_CP_MERCIU_STAT 0x000004d0 + +#define REG_A4XX_CP_WFI_PEND_CTR 0x000004d2 + +#define REG_A4XX_CP_HW_FAULT 0x000004d8 + +#define REG_A4XX_CP_PROTECT_STATUS 0x000004da + +#define REG_A4XX_CP_EVENTS_IN_FLIGHT 0x000004dd + +#define REG_A4XX_CP_PERFCTR_CP_SEL_0 0x00000500 + +#define REG_A4XX_CP_PERFCOMBINER_SELECT 0x0000050b + +static inline uint32_t REG_A4XX_CP_SCRATCH(uint32_t i0) { return 0x00000578 + 0x1*i0; } + +static inline uint32_t REG_A4XX_CP_SCRATCH_REG(uint32_t i0) { return 0x00000578 + 0x1*i0; } + +#define REG_A4XX_SP_VS_STATUS 0x00000ec0 + +#define REG_A4XX_SP_PERFCTR_SP_SEL_11 0x00000ecf + +#define REG_A4XX_SP_SP_CTRL_REG 0x000022c0 +#define A4XX_SP_SP_CTRL_REG_BINNING_PASS 0x00080000 + +#define REG_A4XX_SP_INSTR_CACHE_CTRL 0x000022c1 + +#define REG_A4XX_SP_VS_CTRL_REG0 0x000022c4 +#define A4XX_SP_VS_CTRL_REG0_THREADMODE__MASK 0x00000001 +#define A4XX_SP_VS_CTRL_REG0_THREADMODE__SHIFT 0 +static inline uint32_t A4XX_SP_VS_CTRL_REG0_THREADMODE(enum a3xx_threadmode val) +{ + return ((val) << A4XX_SP_VS_CTRL_REG0_THREADMODE__SHIFT) & A4XX_SP_VS_CTRL_REG0_THREADMODE__MASK; +} +#define A4XX_SP_VS_CTRL_REG0_VARYING 0x00000002 +#define A4XX_SP_VS_CTRL_REG0_CACHEINVALID 0x00000004 +#define A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__MASK 0x000003f0 +#define A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT 4 +static inline uint32_t A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val) +{ + return ((val) << A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__MASK; +} +#define A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0003fc00 +#define A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT 10 +static inline uint32_t A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val) +{ + return ((val) << A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK; +} +#define A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__MASK 0x000c0000 +#define A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__SHIFT 18 +static inline uint32_t A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP(uint32_t val) +{ + return ((val) << A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__SHIFT) & A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__MASK; +} +#define A4XX_SP_VS_CTRL_REG0_THREADSIZE__MASK 0x00100000 +#define A4XX_SP_VS_CTRL_REG0_THREADSIZE__SHIFT 20 +static inline uint32_t A4XX_SP_VS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A4XX_SP_VS_CTRL_REG0_THREADSIZE__SHIFT) & A4XX_SP_VS_CTRL_REG0_THREADSIZE__MASK; +} +#define A4XX_SP_VS_CTRL_REG0_SUPERTHREADMODE 0x00200000 +#define A4XX_SP_VS_CTRL_REG0_PIXLODENABLE 0x00400000 + +#define REG_A4XX_SP_VS_CTRL_REG1 0x000022c5 +#define A4XX_SP_VS_CTRL_REG1_CONSTLENGTH__MASK 0x000000ff +#define A4XX_SP_VS_CTRL_REG1_CONSTLENGTH__SHIFT 0 +static inline uint32_t A4XX_SP_VS_CTRL_REG1_CONSTLENGTH(uint32_t val) +{ + return ((val) << A4XX_SP_VS_CTRL_REG1_CONSTLENGTH__SHIFT) & A4XX_SP_VS_CTRL_REG1_CONSTLENGTH__MASK; +} +#define A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING__MASK 0x7f000000 +#define A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING__SHIFT 24 +static inline uint32_t A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING(uint32_t val) +{ + return ((val) << A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING__SHIFT) & A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING__MASK; +} + +#define REG_A4XX_SP_VS_PARAM_REG 0x000022c6 +#define A4XX_SP_VS_PARAM_REG_POSREGID__MASK 0x000000ff +#define A4XX_SP_VS_PARAM_REG_POSREGID__SHIFT 0 +static inline uint32_t A4XX_SP_VS_PARAM_REG_POSREGID(uint32_t val) +{ + return ((val) << A4XX_SP_VS_PARAM_REG_POSREGID__SHIFT) & A4XX_SP_VS_PARAM_REG_POSREGID__MASK; +} +#define A4XX_SP_VS_PARAM_REG_PSIZEREGID__MASK 0x0000ff00 +#define A4XX_SP_VS_PARAM_REG_PSIZEREGID__SHIFT 8 +static inline uint32_t A4XX_SP_VS_PARAM_REG_PSIZEREGID(uint32_t val) +{ + return ((val) << A4XX_SP_VS_PARAM_REG_PSIZEREGID__SHIFT) & A4XX_SP_VS_PARAM_REG_PSIZEREGID__MASK; +} +#define A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__MASK 0xfff00000 +#define A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__SHIFT 20 +static inline uint32_t A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR(uint32_t val) +{ + return ((val) << A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__SHIFT) & A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__MASK; +} + +static inline uint32_t REG_A4XX_SP_VS_OUT(uint32_t i0) { return 0x000022c7 + 0x1*i0; } + +static inline uint32_t REG_A4XX_SP_VS_OUT_REG(uint32_t i0) { return 0x000022c7 + 0x1*i0; } +#define A4XX_SP_VS_OUT_REG_A_REGID__MASK 0x000001ff +#define A4XX_SP_VS_OUT_REG_A_REGID__SHIFT 0 +static inline uint32_t A4XX_SP_VS_OUT_REG_A_REGID(uint32_t val) +{ + return ((val) << A4XX_SP_VS_OUT_REG_A_REGID__SHIFT) & A4XX_SP_VS_OUT_REG_A_REGID__MASK; +} +#define A4XX_SP_VS_OUT_REG_A_COMPMASK__MASK 0x00001e00 +#define A4XX_SP_VS_OUT_REG_A_COMPMASK__SHIFT 9 +static inline uint32_t A4XX_SP_VS_OUT_REG_A_COMPMASK(uint32_t val) +{ + return ((val) << A4XX_SP_VS_OUT_REG_A_COMPMASK__SHIFT) & A4XX_SP_VS_OUT_REG_A_COMPMASK__MASK; +} +#define A4XX_SP_VS_OUT_REG_B_REGID__MASK 0x01ff0000 +#define A4XX_SP_VS_OUT_REG_B_REGID__SHIFT 16 +static inline uint32_t A4XX_SP_VS_OUT_REG_B_REGID(uint32_t val) +{ + return ((val) << A4XX_SP_VS_OUT_REG_B_REGID__SHIFT) & A4XX_SP_VS_OUT_REG_B_REGID__MASK; +} +#define A4XX_SP_VS_OUT_REG_B_COMPMASK__MASK 0x1e000000 +#define A4XX_SP_VS_OUT_REG_B_COMPMASK__SHIFT 25 +static inline uint32_t A4XX_SP_VS_OUT_REG_B_COMPMASK(uint32_t val) +{ + return ((val) << A4XX_SP_VS_OUT_REG_B_COMPMASK__SHIFT) & A4XX_SP_VS_OUT_REG_B_COMPMASK__MASK; +} + +static inline uint32_t REG_A4XX_SP_VS_VPC_DST(uint32_t i0) { return 0x000022d8 + 0x1*i0; } + +static inline uint32_t REG_A4XX_SP_VS_VPC_DST_REG(uint32_t i0) { return 0x000022d8 + 0x1*i0; } +#define A4XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK 0x000000ff +#define A4XX_SP_VS_VPC_DST_REG_OUTLOC0__SHIFT 0 +static inline uint32_t A4XX_SP_VS_VPC_DST_REG_OUTLOC0(uint32_t val) +{ + return ((val) << A4XX_SP_VS_VPC_DST_REG_OUTLOC0__SHIFT) & A4XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK; +} +#define A4XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK 0x0000ff00 +#define A4XX_SP_VS_VPC_DST_REG_OUTLOC1__SHIFT 8 +static inline uint32_t A4XX_SP_VS_VPC_DST_REG_OUTLOC1(uint32_t val) +{ + return ((val) << A4XX_SP_VS_VPC_DST_REG_OUTLOC1__SHIFT) & A4XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK; +} +#define A4XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK 0x00ff0000 +#define A4XX_SP_VS_VPC_DST_REG_OUTLOC2__SHIFT 16 +static inline uint32_t A4XX_SP_VS_VPC_DST_REG_OUTLOC2(uint32_t val) +{ + return ((val) << A4XX_SP_VS_VPC_DST_REG_OUTLOC2__SHIFT) & A4XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK; +} +#define A4XX_SP_VS_VPC_DST_REG_OUTLOC3__MASK 0xff000000 +#define A4XX_SP_VS_VPC_DST_REG_OUTLOC3__SHIFT 24 +static inline uint32_t A4XX_SP_VS_VPC_DST_REG_OUTLOC3(uint32_t val) +{ + return ((val) << A4XX_SP_VS_VPC_DST_REG_OUTLOC3__SHIFT) & A4XX_SP_VS_VPC_DST_REG_OUTLOC3__MASK; +} + +#define REG_A4XX_SP_VS_OBJ_OFFSET_REG 0x000022e0 +#define A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK 0x01ff0000 +#define A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT 16 +static inline uint32_t A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK 0xfe000000 +#define A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT 25 +static inline uint32_t A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK; +} + +#define REG_A4XX_SP_VS_OBJ_START 0x000022e1 + +#define REG_A4XX_SP_VS_PVT_MEM_PARAM 0x000022e2 + +#define REG_A4XX_SP_VS_PVT_MEM_ADDR 0x000022e3 + +#define REG_A4XX_SP_VS_LENGTH_REG 0x000022e5 + +#define REG_A4XX_SP_FS_CTRL_REG0 0x000022e8 +#define A4XX_SP_FS_CTRL_REG0_THREADMODE__MASK 0x00000001 +#define A4XX_SP_FS_CTRL_REG0_THREADMODE__SHIFT 0 +static inline uint32_t A4XX_SP_FS_CTRL_REG0_THREADMODE(enum a3xx_threadmode val) +{ + return ((val) << A4XX_SP_FS_CTRL_REG0_THREADMODE__SHIFT) & A4XX_SP_FS_CTRL_REG0_THREADMODE__MASK; +} +#define A4XX_SP_FS_CTRL_REG0_VARYING 0x00000002 +#define A4XX_SP_FS_CTRL_REG0_CACHEINVALID 0x00000004 +#define A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__MASK 0x000003f0 +#define A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT 4 +static inline uint32_t A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val) +{ + return ((val) << A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__MASK; +} +#define A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0003fc00 +#define A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT 10 +static inline uint32_t A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val) +{ + return ((val) << A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK; +} +#define A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__MASK 0x000c0000 +#define A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__SHIFT 18 +static inline uint32_t A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP(uint32_t val) +{ + return ((val) << A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__SHIFT) & A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__MASK; +} +#define A4XX_SP_FS_CTRL_REG0_THREADSIZE__MASK 0x00100000 +#define A4XX_SP_FS_CTRL_REG0_THREADSIZE__SHIFT 20 +static inline uint32_t A4XX_SP_FS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A4XX_SP_FS_CTRL_REG0_THREADSIZE__SHIFT) & A4XX_SP_FS_CTRL_REG0_THREADSIZE__MASK; +} +#define A4XX_SP_FS_CTRL_REG0_SUPERTHREADMODE 0x00200000 +#define A4XX_SP_FS_CTRL_REG0_PIXLODENABLE 0x00400000 + +#define REG_A4XX_SP_FS_CTRL_REG1 0x000022e9 +#define A4XX_SP_FS_CTRL_REG1_CONSTLENGTH__MASK 0x000000ff +#define A4XX_SP_FS_CTRL_REG1_CONSTLENGTH__SHIFT 0 +static inline uint32_t A4XX_SP_FS_CTRL_REG1_CONSTLENGTH(uint32_t val) +{ + return ((val) << A4XX_SP_FS_CTRL_REG1_CONSTLENGTH__SHIFT) & A4XX_SP_FS_CTRL_REG1_CONSTLENGTH__MASK; +} +#define A4XX_SP_FS_CTRL_REG1_VARYING 0x00100000 + +#define REG_A4XX_SP_FS_OBJ_OFFSET_REG 0x000022ea +#define A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK 0x01ff0000 +#define A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT 16 +static inline uint32_t A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK 0xfe000000 +#define A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT 25 +static inline uint32_t A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK; +} + +#define REG_A4XX_SP_FS_OBJ_START 0x000022eb + +#define REG_A4XX_SP_FS_PVT_MEM_PARAM 0x000022ec + +#define REG_A4XX_SP_FS_PVT_MEM_ADDR 0x000022ed + +#define REG_A4XX_SP_FS_LENGTH_REG 0x000022ef + +#define REG_A4XX_SP_FS_OUTPUT_REG 0x000022f0 +#define A4XX_SP_FS_OUTPUT_REG_DEPTH_ENABLE 0x00000080 +#define A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID__MASK 0x0000ff00 +#define A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID__SHIFT 8 +static inline uint32_t A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID(uint32_t val) +{ + return ((val) << A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID__SHIFT) & A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID__MASK; +} + +static inline uint32_t REG_A4XX_SP_FS_MRT(uint32_t i0) { return 0x000022f1 + 0x1*i0; } + +static inline uint32_t REG_A4XX_SP_FS_MRT_REG(uint32_t i0) { return 0x000022f1 + 0x1*i0; } +#define A4XX_SP_FS_MRT_REG_REGID__MASK 0x000000ff +#define A4XX_SP_FS_MRT_REG_REGID__SHIFT 0 +static inline uint32_t A4XX_SP_FS_MRT_REG_REGID(uint32_t val) +{ + return ((val) << A4XX_SP_FS_MRT_REG_REGID__SHIFT) & A4XX_SP_FS_MRT_REG_REGID__MASK; +} +#define A4XX_SP_FS_MRT_REG_HALF_PRECISION 0x00000100 +#define A4XX_SP_FS_MRT_REG_MRTFORMAT__MASK 0x0003f000 +#define A4XX_SP_FS_MRT_REG_MRTFORMAT__SHIFT 12 +static inline uint32_t A4XX_SP_FS_MRT_REG_MRTFORMAT(enum a4xx_color_fmt val) +{ + return ((val) << A4XX_SP_FS_MRT_REG_MRTFORMAT__SHIFT) & A4XX_SP_FS_MRT_REG_MRTFORMAT__MASK; +} + +#define REG_A4XX_SP_HS_OBJ_OFFSET_REG 0x0000230d +#define A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK 0x01ff0000 +#define A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT 16 +static inline uint32_t A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK 0xfe000000 +#define A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT 25 +static inline uint32_t A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK; +} + +#define REG_A4XX_SP_DS_OBJ_OFFSET_REG 0x00002334 +#define A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK 0x01ff0000 +#define A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT 16 +static inline uint32_t A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK 0xfe000000 +#define A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT 25 +static inline uint32_t A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK; +} + +#define REG_A4XX_SP_GS_OBJ_OFFSET_REG 0x0000235b +#define A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK 0x01ff0000 +#define A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT 16 +static inline uint32_t A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK 0xfe000000 +#define A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT 25 +static inline uint32_t A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK; +} + +#define REG_A4XX_SP_GS_LENGTH_REG 0x00002360 + +#define REG_A4XX_VPC_DEBUG_RAM_SEL 0x00000e60 + +#define REG_A4XX_VPC_DEBUG_RAM_READ 0x00000e61 + +#define REG_A4XX_VPC_DEBUG_ECO_CONTROL 0x00000e64 + +#define REG_A4XX_VPC_PERFCTR_VPC_SEL_3 0x00000e68 + +#define REG_A4XX_VPC_ATTR 0x00002140 +#define A4XX_VPC_ATTR_TOTALATTR__MASK 0x000001ff +#define A4XX_VPC_ATTR_TOTALATTR__SHIFT 0 +static inline uint32_t A4XX_VPC_ATTR_TOTALATTR(uint32_t val) +{ + return ((val) << A4XX_VPC_ATTR_TOTALATTR__SHIFT) & A4XX_VPC_ATTR_TOTALATTR__MASK; +} +#define A4XX_VPC_ATTR_PSIZE 0x00000200 +#define A4XX_VPC_ATTR_THRDASSIGN__MASK 0x00003000 +#define A4XX_VPC_ATTR_THRDASSIGN__SHIFT 12 +static inline uint32_t A4XX_VPC_ATTR_THRDASSIGN(uint32_t val) +{ + return ((val) << A4XX_VPC_ATTR_THRDASSIGN__SHIFT) & A4XX_VPC_ATTR_THRDASSIGN__MASK; +} +#define A4XX_VPC_ATTR_ENABLE 0x02000000 + +#define REG_A4XX_VPC_PACK 0x00002141 +#define A4XX_VPC_PACK_NUMBYPASSVAR__MASK 0x000000ff +#define A4XX_VPC_PACK_NUMBYPASSVAR__SHIFT 0 +static inline uint32_t A4XX_VPC_PACK_NUMBYPASSVAR(uint32_t val) +{ + return ((val) << A4XX_VPC_PACK_NUMBYPASSVAR__SHIFT) & A4XX_VPC_PACK_NUMBYPASSVAR__MASK; +} +#define A4XX_VPC_PACK_NUMFPNONPOSVAR__MASK 0x0000ff00 +#define A4XX_VPC_PACK_NUMFPNONPOSVAR__SHIFT 8 +static inline uint32_t A4XX_VPC_PACK_NUMFPNONPOSVAR(uint32_t val) +{ + return ((val) << A4XX_VPC_PACK_NUMFPNONPOSVAR__SHIFT) & A4XX_VPC_PACK_NUMFPNONPOSVAR__MASK; +} +#define A4XX_VPC_PACK_NUMNONPOSVSVAR__MASK 0x00ff0000 +#define A4XX_VPC_PACK_NUMNONPOSVSVAR__SHIFT 16 +static inline uint32_t A4XX_VPC_PACK_NUMNONPOSVSVAR(uint32_t val) +{ + return ((val) << A4XX_VPC_PACK_NUMNONPOSVSVAR__SHIFT) & A4XX_VPC_PACK_NUMNONPOSVSVAR__MASK; +} + +static inline uint32_t REG_A4XX_VPC_VARYING_INTERP(uint32_t i0) { return 0x00002142 + 0x1*i0; } + +static inline uint32_t REG_A4XX_VPC_VARYING_INTERP_MODE(uint32_t i0) { return 0x00002142 + 0x1*i0; } + +static inline uint32_t REG_A4XX_VPC_VARYING_PS_REPL(uint32_t i0) { return 0x0000214a + 0x1*i0; } + +static inline uint32_t REG_A4XX_VPC_VARYING_PS_REPL_MODE(uint32_t i0) { return 0x0000214a + 0x1*i0; } + +#define REG_A4XX_VPC_SO_FLUSH_WADDR_3 0x0000216e + +#define REG_A4XX_VSC_BIN_SIZE 0x00000c00 +#define A4XX_VSC_BIN_SIZE_WIDTH__MASK 0x0000001f +#define A4XX_VSC_BIN_SIZE_WIDTH__SHIFT 0 +static inline uint32_t A4XX_VSC_BIN_SIZE_WIDTH(uint32_t val) +{ + return ((val >> 5) << A4XX_VSC_BIN_SIZE_WIDTH__SHIFT) & A4XX_VSC_BIN_SIZE_WIDTH__MASK; +} +#define A4XX_VSC_BIN_SIZE_HEIGHT__MASK 0x000003e0 +#define A4XX_VSC_BIN_SIZE_HEIGHT__SHIFT 5 +static inline uint32_t A4XX_VSC_BIN_SIZE_HEIGHT(uint32_t val) +{ + return ((val >> 5) << A4XX_VSC_BIN_SIZE_HEIGHT__SHIFT) & A4XX_VSC_BIN_SIZE_HEIGHT__MASK; +} + +#define REG_A4XX_VSC_SIZE_ADDRESS 0x00000c01 + +#define REG_A4XX_VSC_SIZE_ADDRESS2 0x00000c02 + +#define REG_A4XX_VSC_DEBUG_ECO_CONTROL 0x00000c03 + +static inline uint32_t REG_A4XX_VSC_PIPE_CONFIG(uint32_t i0) { return 0x00000c08 + 0x1*i0; } + +static inline uint32_t REG_A4XX_VSC_PIPE_CONFIG_REG(uint32_t i0) { return 0x00000c08 + 0x1*i0; } +#define A4XX_VSC_PIPE_CONFIG_REG_X__MASK 0x000003ff +#define A4XX_VSC_PIPE_CONFIG_REG_X__SHIFT 0 +static inline uint32_t A4XX_VSC_PIPE_CONFIG_REG_X(uint32_t val) +{ + return ((val) << A4XX_VSC_PIPE_CONFIG_REG_X__SHIFT) & A4XX_VSC_PIPE_CONFIG_REG_X__MASK; +} +#define A4XX_VSC_PIPE_CONFIG_REG_Y__MASK 0x000ffc00 +#define A4XX_VSC_PIPE_CONFIG_REG_Y__SHIFT 10 +static inline uint32_t A4XX_VSC_PIPE_CONFIG_REG_Y(uint32_t val) +{ + return ((val) << A4XX_VSC_PIPE_CONFIG_REG_Y__SHIFT) & A4XX_VSC_PIPE_CONFIG_REG_Y__MASK; +} +#define A4XX_VSC_PIPE_CONFIG_REG_W__MASK 0x00f00000 +#define A4XX_VSC_PIPE_CONFIG_REG_W__SHIFT 20 +static inline uint32_t A4XX_VSC_PIPE_CONFIG_REG_W(uint32_t val) +{ + return ((val) << A4XX_VSC_PIPE_CONFIG_REG_W__SHIFT) & A4XX_VSC_PIPE_CONFIG_REG_W__MASK; +} +#define A4XX_VSC_PIPE_CONFIG_REG_H__MASK 0x0f000000 +#define A4XX_VSC_PIPE_CONFIG_REG_H__SHIFT 24 +static inline uint32_t A4XX_VSC_PIPE_CONFIG_REG_H(uint32_t val) +{ + return ((val) << A4XX_VSC_PIPE_CONFIG_REG_H__SHIFT) & A4XX_VSC_PIPE_CONFIG_REG_H__MASK; +} + +static inline uint32_t REG_A4XX_VSC_PIPE_DATA_ADDRESS(uint32_t i0) { return 0x00000c10 + 0x1*i0; } + +static inline uint32_t REG_A4XX_VSC_PIPE_DATA_ADDRESS_REG(uint32_t i0) { return 0x00000c10 + 0x1*i0; } + +static inline uint32_t REG_A4XX_VSC_PIPE_DATA_LENGTH(uint32_t i0) { return 0x00000c18 + 0x1*i0; } + +static inline uint32_t REG_A4XX_VSC_PIPE_DATA_LENGTH_REG(uint32_t i0) { return 0x00000c18 + 0x1*i0; } + +#define REG_A4XX_VSC_PIPE_PARTIAL_POSN_1 0x00000c41 + +#define REG_A4XX_VSC_PERFCTR_VSC_SEL_0 0x00000c50 + +#define REG_A4XX_VSC_PERFCTR_VSC_SEL_1 0x00000c51 + +#define REG_A4XX_VFD_DEBUG_CONTROL 0x00000e40 + +#define REG_A4XX_VFD_PERFCTR_VFD_SEL_7 0x00000e4a + +#define REG_A4XX_VFD_CONTROL_0 0x00002200 +#define A4XX_VFD_CONTROL_0_TOTALATTRTOVS__MASK 0x000000ff +#define A4XX_VFD_CONTROL_0_TOTALATTRTOVS__SHIFT 0 +static inline uint32_t A4XX_VFD_CONTROL_0_TOTALATTRTOVS(uint32_t val) +{ + return ((val) << A4XX_VFD_CONTROL_0_TOTALATTRTOVS__SHIFT) & A4XX_VFD_CONTROL_0_TOTALATTRTOVS__MASK; +} +#define A4XX_VFD_CONTROL_0_BYPASSATTROVS__MASK 0x0001fe00 +#define A4XX_VFD_CONTROL_0_BYPASSATTROVS__SHIFT 9 +static inline uint32_t A4XX_VFD_CONTROL_0_BYPASSATTROVS(uint32_t val) +{ + return ((val) << A4XX_VFD_CONTROL_0_BYPASSATTROVS__SHIFT) & A4XX_VFD_CONTROL_0_BYPASSATTROVS__MASK; +} +#define A4XX_VFD_CONTROL_0_STRMDECINSTRCNT__MASK 0x03f00000 +#define A4XX_VFD_CONTROL_0_STRMDECINSTRCNT__SHIFT 20 +static inline uint32_t A4XX_VFD_CONTROL_0_STRMDECINSTRCNT(uint32_t val) +{ + return ((val) << A4XX_VFD_CONTROL_0_STRMDECINSTRCNT__SHIFT) & A4XX_VFD_CONTROL_0_STRMDECINSTRCNT__MASK; +} +#define A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT__MASK 0xfc000000 +#define A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT__SHIFT 26 +static inline uint32_t A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT(uint32_t val) +{ + return ((val) << A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT__SHIFT) & A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT__MASK; +} + +#define REG_A4XX_VFD_CONTROL_1 0x00002201 +#define A4XX_VFD_CONTROL_1_MAXSTORAGE__MASK 0x0000ffff +#define A4XX_VFD_CONTROL_1_MAXSTORAGE__SHIFT 0 +static inline uint32_t A4XX_VFD_CONTROL_1_MAXSTORAGE(uint32_t val) +{ + return ((val) << A4XX_VFD_CONTROL_1_MAXSTORAGE__SHIFT) & A4XX_VFD_CONTROL_1_MAXSTORAGE__MASK; +} +#define A4XX_VFD_CONTROL_1_REGID4VTX__MASK 0x00ff0000 +#define A4XX_VFD_CONTROL_1_REGID4VTX__SHIFT 16 +static inline uint32_t A4XX_VFD_CONTROL_1_REGID4VTX(uint32_t val) +{ + return ((val) << A4XX_VFD_CONTROL_1_REGID4VTX__SHIFT) & A4XX_VFD_CONTROL_1_REGID4VTX__MASK; +} +#define A4XX_VFD_CONTROL_1_REGID4INST__MASK 0xff000000 +#define A4XX_VFD_CONTROL_1_REGID4INST__SHIFT 24 +static inline uint32_t A4XX_VFD_CONTROL_1_REGID4INST(uint32_t val) +{ + return ((val) << A4XX_VFD_CONTROL_1_REGID4INST__SHIFT) & A4XX_VFD_CONTROL_1_REGID4INST__MASK; +} + +#define REG_A4XX_VFD_CONTROL_2 0x00002202 + +#define REG_A4XX_VFD_CONTROL_3 0x00002203 + +#define REG_A4XX_VFD_CONTROL_4 0x00002204 + +#define REG_A4XX_VFD_INDEX_OFFSET 0x00002208 + +static inline uint32_t REG_A4XX_VFD_FETCH(uint32_t i0) { return 0x0000220a + 0x4*i0; } + +static inline uint32_t REG_A4XX_VFD_FETCH_INSTR_0(uint32_t i0) { return 0x0000220a + 0x4*i0; } +#define A4XX_VFD_FETCH_INSTR_0_FETCHSIZE__MASK 0x0000007f +#define A4XX_VFD_FETCH_INSTR_0_FETCHSIZE__SHIFT 0 +static inline uint32_t A4XX_VFD_FETCH_INSTR_0_FETCHSIZE(uint32_t val) +{ + return ((val) << A4XX_VFD_FETCH_INSTR_0_FETCHSIZE__SHIFT) & A4XX_VFD_FETCH_INSTR_0_FETCHSIZE__MASK; +} +#define A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE__MASK 0x0001ff80 +#define A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE__SHIFT 7 +static inline uint32_t A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE(uint32_t val) +{ + return ((val) << A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE__SHIFT) & A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE__MASK; +} +#define A4XX_VFD_FETCH_INSTR_0_SWITCHNEXT 0x00080000 +#define A4XX_VFD_FETCH_INSTR_0_STEPRATE__MASK 0xff000000 +#define A4XX_VFD_FETCH_INSTR_0_STEPRATE__SHIFT 24 +static inline uint32_t A4XX_VFD_FETCH_INSTR_0_STEPRATE(uint32_t val) +{ + return ((val) << A4XX_VFD_FETCH_INSTR_0_STEPRATE__SHIFT) & A4XX_VFD_FETCH_INSTR_0_STEPRATE__MASK; +} + +static inline uint32_t REG_A4XX_VFD_FETCH_INSTR_1(uint32_t i0) { return 0x0000220b + 0x4*i0; } + +static inline uint32_t REG_A4XX_VFD_FETCH_INSTR_2(uint32_t i0) { return 0x0000220c + 0x4*i0; } +#define A4XX_VFD_FETCH_INSTR_2_SIZE__MASK 0xfffffff0 +#define A4XX_VFD_FETCH_INSTR_2_SIZE__SHIFT 4 +static inline uint32_t A4XX_VFD_FETCH_INSTR_2_SIZE(uint32_t val) +{ + return ((val >> 4) << A4XX_VFD_FETCH_INSTR_2_SIZE__SHIFT) & A4XX_VFD_FETCH_INSTR_2_SIZE__MASK; +} + +static inline uint32_t REG_A4XX_VFD_FETCH_INSTR_3(uint32_t i0) { return 0x0000220d + 0x4*i0; } + +static inline uint32_t REG_A4XX_VFD_DECODE(uint32_t i0) { return 0x0000228a + 0x1*i0; } + +static inline uint32_t REG_A4XX_VFD_DECODE_INSTR(uint32_t i0) { return 0x0000228a + 0x1*i0; } +#define A4XX_VFD_DECODE_INSTR_WRITEMASK__MASK 0x0000000f +#define A4XX_VFD_DECODE_INSTR_WRITEMASK__SHIFT 0 +static inline uint32_t A4XX_VFD_DECODE_INSTR_WRITEMASK(uint32_t val) +{ + return ((val) << A4XX_VFD_DECODE_INSTR_WRITEMASK__SHIFT) & A4XX_VFD_DECODE_INSTR_WRITEMASK__MASK; +} +#define A4XX_VFD_DECODE_INSTR_CONSTFILL 0x00000010 +#define A4XX_VFD_DECODE_INSTR_FORMAT__MASK 0x00000fc0 +#define A4XX_VFD_DECODE_INSTR_FORMAT__SHIFT 6 +static inline uint32_t A4XX_VFD_DECODE_INSTR_FORMAT(enum a4xx_vtx_fmt val) +{ + return ((val) << A4XX_VFD_DECODE_INSTR_FORMAT__SHIFT) & A4XX_VFD_DECODE_INSTR_FORMAT__MASK; +} +#define A4XX_VFD_DECODE_INSTR_REGID__MASK 0x000ff000 +#define A4XX_VFD_DECODE_INSTR_REGID__SHIFT 12 +static inline uint32_t A4XX_VFD_DECODE_INSTR_REGID(uint32_t val) +{ + return ((val) << A4XX_VFD_DECODE_INSTR_REGID__SHIFT) & A4XX_VFD_DECODE_INSTR_REGID__MASK; +} +#define A4XX_VFD_DECODE_INSTR_SWAP__MASK 0x00c00000 +#define A4XX_VFD_DECODE_INSTR_SWAP__SHIFT 22 +static inline uint32_t A4XX_VFD_DECODE_INSTR_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A4XX_VFD_DECODE_INSTR_SWAP__SHIFT) & A4XX_VFD_DECODE_INSTR_SWAP__MASK; +} +#define A4XX_VFD_DECODE_INSTR_SHIFTCNT__MASK 0x1f000000 +#define A4XX_VFD_DECODE_INSTR_SHIFTCNT__SHIFT 24 +static inline uint32_t A4XX_VFD_DECODE_INSTR_SHIFTCNT(uint32_t val) +{ + return ((val) << A4XX_VFD_DECODE_INSTR_SHIFTCNT__SHIFT) & A4XX_VFD_DECODE_INSTR_SHIFTCNT__MASK; +} +#define A4XX_VFD_DECODE_INSTR_LASTCOMPVALID 0x20000000 +#define A4XX_VFD_DECODE_INSTR_SWITCHNEXT 0x40000000 + +#define REG_A4XX_TPL1_DEBUG_ECO_CONTROL 0x00000f00 + +#define REG_A4XX_TPL1_PERFCTR_TP_SEL_7 0x00000f0b + +#define REG_A4XX_TPL1_TP_TEX_OFFSET 0x00002380 + +#define REG_A4XX_TPL1_TP_CS_TEXMEMOBJ_BASE_ADDR 0x000023a6 + +#define REG_A4XX_GRAS_TSE_STATUS 0x00000c80 + +#define REG_A4XX_GRAS_DEBUG_ECO_CONTROL 0x00000c81 + +#define REG_A4XX_GRAS_PERFCTR_TSE_SEL_0 0x00000c88 + +#define REG_A4XX_GRAS_PERFCTR_TSE_SEL_3 0x00000c8b + +#define REG_A4XX_GRAS_CL_CLIP_CNTL 0x00002000 + +#define REG_A4XX_GRAS_CLEAR_CNTL 0x00002003 +#define A4XX_GRAS_CLEAR_CNTL_NOT_FASTCLEAR 0x00000001 + +#define REG_A4XX_GRAS_CL_GB_CLIP_ADJ 0x00002004 +#define A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ__MASK 0x000003ff +#define A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ__SHIFT 0 +static inline uint32_t A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ(uint32_t val) +{ + return ((val) << A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ__SHIFT) & A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ__MASK; +} +#define A4XX_GRAS_CL_GB_CLIP_ADJ_VERT__MASK 0x000ffc00 +#define A4XX_GRAS_CL_GB_CLIP_ADJ_VERT__SHIFT 10 +static inline uint32_t A4XX_GRAS_CL_GB_CLIP_ADJ_VERT(uint32_t val) +{ + return ((val) << A4XX_GRAS_CL_GB_CLIP_ADJ_VERT__SHIFT) & A4XX_GRAS_CL_GB_CLIP_ADJ_VERT__MASK; +} + +#define REG_A4XX_GRAS_CL_VPORT_XOFFSET_0 0x00002008 +#define A4XX_GRAS_CL_VPORT_XOFFSET_0__MASK 0xffffffff +#define A4XX_GRAS_CL_VPORT_XOFFSET_0__SHIFT 0 +static inline uint32_t A4XX_GRAS_CL_VPORT_XOFFSET_0(float val) +{ + return ((fui(val)) << A4XX_GRAS_CL_VPORT_XOFFSET_0__SHIFT) & A4XX_GRAS_CL_VPORT_XOFFSET_0__MASK; +} + +#define REG_A4XX_GRAS_CL_VPORT_XSCALE_0 0x00002009 +#define A4XX_GRAS_CL_VPORT_XSCALE_0__MASK 0xffffffff +#define A4XX_GRAS_CL_VPORT_XSCALE_0__SHIFT 0 +static inline uint32_t A4XX_GRAS_CL_VPORT_XSCALE_0(float val) +{ + return ((fui(val)) << A4XX_GRAS_CL_VPORT_XSCALE_0__SHIFT) & A4XX_GRAS_CL_VPORT_XSCALE_0__MASK; +} + +#define REG_A4XX_GRAS_CL_VPORT_YOFFSET_0 0x0000200a +#define A4XX_GRAS_CL_VPORT_YOFFSET_0__MASK 0xffffffff +#define A4XX_GRAS_CL_VPORT_YOFFSET_0__SHIFT 0 +static inline uint32_t A4XX_GRAS_CL_VPORT_YOFFSET_0(float val) +{ + return ((fui(val)) << A4XX_GRAS_CL_VPORT_YOFFSET_0__SHIFT) & A4XX_GRAS_CL_VPORT_YOFFSET_0__MASK; +} + +#define REG_A4XX_GRAS_CL_VPORT_YSCALE_0 0x0000200b +#define A4XX_GRAS_CL_VPORT_YSCALE_0__MASK 0xffffffff +#define A4XX_GRAS_CL_VPORT_YSCALE_0__SHIFT 0 +static inline uint32_t A4XX_GRAS_CL_VPORT_YSCALE_0(float val) +{ + return ((fui(val)) << A4XX_GRAS_CL_VPORT_YSCALE_0__SHIFT) & A4XX_GRAS_CL_VPORT_YSCALE_0__MASK; +} + +#define REG_A4XX_GRAS_CL_VPORT_ZOFFSET_0 0x0000200c +#define A4XX_GRAS_CL_VPORT_ZOFFSET_0__MASK 0xffffffff +#define A4XX_GRAS_CL_VPORT_ZOFFSET_0__SHIFT 0 +static inline uint32_t A4XX_GRAS_CL_VPORT_ZOFFSET_0(float val) +{ + return ((fui(val)) << A4XX_GRAS_CL_VPORT_ZOFFSET_0__SHIFT) & A4XX_GRAS_CL_VPORT_ZOFFSET_0__MASK; +} + +#define REG_A4XX_GRAS_CL_VPORT_ZSCALE_0 0x0000200d +#define A4XX_GRAS_CL_VPORT_ZSCALE_0__MASK 0xffffffff +#define A4XX_GRAS_CL_VPORT_ZSCALE_0__SHIFT 0 +static inline uint32_t A4XX_GRAS_CL_VPORT_ZSCALE_0(float val) +{ + return ((fui(val)) << A4XX_GRAS_CL_VPORT_ZSCALE_0__SHIFT) & A4XX_GRAS_CL_VPORT_ZSCALE_0__MASK; +} + +#define REG_A4XX_GRAS_SU_POINT_MINMAX 0x00002070 +#define A4XX_GRAS_SU_POINT_MINMAX_MIN__MASK 0x0000ffff +#define A4XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT 0 +static inline uint32_t A4XX_GRAS_SU_POINT_MINMAX_MIN(float val) +{ + return ((((uint32_t)(val * 16.0))) << A4XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT) & A4XX_GRAS_SU_POINT_MINMAX_MIN__MASK; +} +#define A4XX_GRAS_SU_POINT_MINMAX_MAX__MASK 0xffff0000 +#define A4XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT 16 +static inline uint32_t A4XX_GRAS_SU_POINT_MINMAX_MAX(float val) +{ + return ((((uint32_t)(val * 16.0))) << A4XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT) & A4XX_GRAS_SU_POINT_MINMAX_MAX__MASK; +} + +#define REG_A4XX_GRAS_SU_POINT_SIZE 0x00002071 +#define A4XX_GRAS_SU_POINT_SIZE__MASK 0xffffffff +#define A4XX_GRAS_SU_POINT_SIZE__SHIFT 0 +static inline uint32_t A4XX_GRAS_SU_POINT_SIZE(float val) +{ + return ((((int32_t)(val * 16.0))) << A4XX_GRAS_SU_POINT_SIZE__SHIFT) & A4XX_GRAS_SU_POINT_SIZE__MASK; +} + +#define REG_A4XX_GRAS_ALPHA_CONTROL 0x00002073 +#define A4XX_GRAS_ALPHA_CONTROL_ALPHA_TEST_ENABLE 0x00000004 + +#define REG_A4XX_GRAS_SU_POLY_OFFSET_SCALE 0x00002074 +#define A4XX_GRAS_SU_POLY_OFFSET_SCALE__MASK 0xffffffff +#define A4XX_GRAS_SU_POLY_OFFSET_SCALE__SHIFT 0 +static inline uint32_t A4XX_GRAS_SU_POLY_OFFSET_SCALE(float val) +{ + return ((fui(val)) << A4XX_GRAS_SU_POLY_OFFSET_SCALE__SHIFT) & A4XX_GRAS_SU_POLY_OFFSET_SCALE__MASK; +} + +#define REG_A4XX_GRAS_SU_POLY_OFFSET_OFFSET 0x00002075 +#define A4XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK 0xffffffff +#define A4XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT 0 +static inline uint32_t A4XX_GRAS_SU_POLY_OFFSET_OFFSET(float val) +{ + return ((fui(val)) << A4XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT) & A4XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK; +} + +#define REG_A4XX_GRAS_SC_EXTENT_WINDOW_TL 0x0000209f + +#define REG_A4XX_GRAS_SC_SCREEN_SCISSOR_TL 0x0000207c +#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_WINDOW_OFFSET_DISABLE 0x80000000 +#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X__MASK 0x00007fff +#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X__SHIFT 0 +static inline uint32_t A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X__SHIFT) & A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X__MASK; +} +#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y__MASK 0x7fff0000 +#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y__SHIFT 16 +static inline uint32_t A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y__SHIFT) & A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y__MASK; +} + +#define REG_A4XX_GRAS_SC_SCREEN_SCISSOR_BR 0x0000207d +#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_WINDOW_OFFSET_DISABLE 0x80000000 +#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X__MASK 0x00007fff +#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X__SHIFT 0 +static inline uint32_t A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X__SHIFT) & A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X__MASK; +} +#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y__MASK 0x7fff0000 +#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y__SHIFT 16 +static inline uint32_t A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y__SHIFT) & A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y__MASK; +} + +#define REG_A4XX_GRAS_SC_WINDOW_SCISSOR_BR 0x0000209c +#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_WINDOW_OFFSET_DISABLE 0x80000000 +#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X__MASK 0x00007fff +#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X__SHIFT 0 +static inline uint32_t A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X__SHIFT) & A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X__MASK; +} +#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__MASK 0x7fff0000 +#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__SHIFT 16 +static inline uint32_t A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__SHIFT) & A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__MASK; +} + +#define REG_A4XX_GRAS_SC_WINDOW_SCISSOR_TL 0x0000209d +#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_WINDOW_OFFSET_DISABLE 0x80000000 +#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X__MASK 0x00007fff +#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X__SHIFT 0 +static inline uint32_t A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X__SHIFT) & A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X__MASK; +} +#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__MASK 0x7fff0000 +#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__SHIFT 16 +static inline uint32_t A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__SHIFT) & A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__MASK; +} + +#define REG_A4XX_GRAS_DEPTH_CONTROL 0x00002077 +#define A4XX_GRAS_DEPTH_CONTROL_FORMAT__MASK 0x00000003 +#define A4XX_GRAS_DEPTH_CONTROL_FORMAT__SHIFT 0 +static inline uint32_t A4XX_GRAS_DEPTH_CONTROL_FORMAT(enum a4xx_depth_format val) +{ + return ((val) << A4XX_GRAS_DEPTH_CONTROL_FORMAT__SHIFT) & A4XX_GRAS_DEPTH_CONTROL_FORMAT__MASK; +} + +#define REG_A4XX_GRAS_SU_MODE_CONTROL 0x00002078 +#define A4XX_GRAS_SU_MODE_CONTROL_CULL_FRONT 0x00000001 +#define A4XX_GRAS_SU_MODE_CONTROL_CULL_BACK 0x00000002 +#define A4XX_GRAS_SU_MODE_CONTROL_FRONT_CW 0x00000004 +#define A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__MASK 0x000007f8 +#define A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT 3 +static inline uint32_t A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH(float val) +{ + return ((((int32_t)(val * 4.0))) << A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT) & A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__MASK; +} +#define A4XX_GRAS_SU_MODE_CONTROL_POLY_OFFSET 0x00000800 +#define A4XX_GRAS_SU_MODE_CONTROL_RENDERING_PASS 0x00100000 + +#define REG_A4XX_GRAS_SC_CONTROL 0x0000207b +#define A4XX_GRAS_SC_CONTROL_RENDER_MODE__MASK 0x0000000c +#define A4XX_GRAS_SC_CONTROL_RENDER_MODE__SHIFT 2 +static inline uint32_t A4XX_GRAS_SC_CONTROL_RENDER_MODE(enum a3xx_render_mode val) +{ + return ((val) << A4XX_GRAS_SC_CONTROL_RENDER_MODE__SHIFT) & A4XX_GRAS_SC_CONTROL_RENDER_MODE__MASK; +} +#define A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES__MASK 0x00000380 +#define A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES__SHIFT 7 +static inline uint32_t A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES__SHIFT) & A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES__MASK; +} +#define A4XX_GRAS_SC_CONTROL_MSAA_DISABLE 0x00000800 +#define A4XX_GRAS_SC_CONTROL_RASTER_MODE__MASK 0x0000f000 +#define A4XX_GRAS_SC_CONTROL_RASTER_MODE__SHIFT 12 +static inline uint32_t A4XX_GRAS_SC_CONTROL_RASTER_MODE(uint32_t val) +{ + return ((val) << A4XX_GRAS_SC_CONTROL_RASTER_MODE__SHIFT) & A4XX_GRAS_SC_CONTROL_RASTER_MODE__MASK; +} + +#define REG_A4XX_UCHE_CACHE_MODE_CONTROL 0x00000e80 + +#define REG_A4XX_UCHE_TRAP_BASE_LO 0x00000e83 + +#define REG_A4XX_UCHE_TRAP_BASE_HI 0x00000e84 + +#define REG_A4XX_UCHE_CACHE_STATUS 0x00000e88 + +#define REG_A4XX_UCHE_INVALIDATE0 0x00000e8a + +#define REG_A4XX_UCHE_INVALIDATE1 0x00000e8b + +#define REG_A4XX_UCHE_CACHE_WAYS_VFD 0x00000e8c + +#define REG_A4XX_UCHE_PERFCTR_UCHE_SEL_7 0x00000e95 + +#define REG_A4XX_HLSQ_TIMEOUT_THRESHOLD 0x00000e00 + +#define REG_A4XX_HLSQ_DEBUG_ECO_CONTROL 0x00000e04 + +#define REG_A4XX_HLSQ_PERF_PIPE_MASK 0x00000e0e + +#define REG_A4XX_HLSQ_CONTROL_0_REG 0x000023c0 +#define A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK 0x00000010 +#define A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__SHIFT 4 +static inline uint32_t A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__SHIFT) & A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK; +} +#define A4XX_HLSQ_CONTROL_0_REG_FSSUPERTHREADENABLE 0x00000040 +#define A4XX_HLSQ_CONTROL_0_REG_SPSHADERRESTART 0x00000200 +#define A4XX_HLSQ_CONTROL_0_REG_RESERVED2 0x00000400 +#define A4XX_HLSQ_CONTROL_0_REG_CHUNKDISABLE 0x04000000 +#define A4XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK 0x08000000 +#define A4XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT 27 +static inline uint32_t A4XX_HLSQ_CONTROL_0_REG_CONSTMODE(uint32_t val) +{ + return ((val) << A4XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT) & A4XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK; +} +#define A4XX_HLSQ_CONTROL_0_REG_LAZYUPDATEDISABLE 0x10000000 +#define A4XX_HLSQ_CONTROL_0_REG_SPCONSTFULLUPDATE 0x20000000 +#define A4XX_HLSQ_CONTROL_0_REG_TPFULLUPDATE 0x40000000 +#define A4XX_HLSQ_CONTROL_0_REG_SINGLECONTEXT 0x80000000 + +#define REG_A4XX_HLSQ_CONTROL_1_REG 0x000023c1 +#define A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__MASK 0x00000040 +#define A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__SHIFT 6 +static inline uint32_t A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__SHIFT) & A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__MASK; +} +#define A4XX_HLSQ_CONTROL_1_REG_VSSUPERTHREADENABLE 0x00000100 +#define A4XX_HLSQ_CONTROL_1_REG_RESERVED1 0x00000200 +#define A4XX_HLSQ_CONTROL_1_REG_ZWCOORD 0x02000000 + +#define REG_A4XX_HLSQ_CONTROL_2_REG 0x000023c2 +#define A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__MASK 0xfc000000 +#define A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__SHIFT 26 +static inline uint32_t A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD(uint32_t val) +{ + return ((val) << A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__SHIFT) & A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__MASK; +} + +#define REG_A4XX_HLSQ_CONTROL_3_REG 0x000023c3 +#define A4XX_HLSQ_CONTROL_3_REG_REGID__MASK 0x000000ff +#define A4XX_HLSQ_CONTROL_3_REG_REGID__SHIFT 0 +static inline uint32_t A4XX_HLSQ_CONTROL_3_REG_REGID(uint32_t val) +{ + return ((val) << A4XX_HLSQ_CONTROL_3_REG_REGID__SHIFT) & A4XX_HLSQ_CONTROL_3_REG_REGID__MASK; +} + +#define REG_A4XX_HLSQ_VS_CONTROL_REG 0x000023c5 +#define A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__MASK 0x000000ff +#define A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__SHIFT 0 +static inline uint32_t A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__MASK; +} +#define A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET__MASK 0x0000ff00 +#define A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT 8 +static inline uint32_t A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET__MASK 0x00fe0000 +#define A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET__SHIFT 17 +static inline uint32_t A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET__MASK; +} +#define A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH__MASK 0xff000000 +#define A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH__SHIFT 24 +static inline uint32_t A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH__MASK; +} + +#define REG_A4XX_HLSQ_FS_CONTROL_REG 0x000023c6 +#define A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__MASK 0x000000ff +#define A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__SHIFT 0 +static inline uint32_t A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__MASK; +} +#define A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET__MASK 0x0000ff00 +#define A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT 8 +static inline uint32_t A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET__MASK 0x00fe0000 +#define A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET__SHIFT 17 +static inline uint32_t A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET__MASK; +} +#define A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH__MASK 0xff000000 +#define A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH__SHIFT 24 +static inline uint32_t A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH__MASK; +} + +#define REG_A4XX_HLSQ_HS_CONTROL_REG 0x000023c7 +#define A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH__MASK 0x000000ff +#define A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH__SHIFT 0 +static inline uint32_t A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH__MASK; +} +#define A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET__MASK 0x0000ff00 +#define A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT 8 +static inline uint32_t A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET__MASK 0x00fe0000 +#define A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET__SHIFT 17 +static inline uint32_t A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET__MASK; +} +#define A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH__MASK 0xff000000 +#define A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH__SHIFT 24 +static inline uint32_t A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH__MASK; +} + +#define REG_A4XX_HLSQ_DS_CONTROL_REG 0x000023c8 +#define A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH__MASK 0x000000ff +#define A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH__SHIFT 0 +static inline uint32_t A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH__MASK; +} +#define A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET__MASK 0x0000ff00 +#define A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT 8 +static inline uint32_t A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET__MASK 0x00fe0000 +#define A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET__SHIFT 17 +static inline uint32_t A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET__MASK; +} +#define A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH__MASK 0xff000000 +#define A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH__SHIFT 24 +static inline uint32_t A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH__MASK; +} + +#define REG_A4XX_HLSQ_GS_CONTROL_REG 0x000023c9 +#define A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH__MASK 0x000000ff +#define A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH__SHIFT 0 +static inline uint32_t A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH__MASK; +} +#define A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET__MASK 0x0000ff00 +#define A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT 8 +static inline uint32_t A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET__MASK; +} +#define A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET__MASK 0x00fe0000 +#define A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET__SHIFT 17 +static inline uint32_t A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET__MASK; +} +#define A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH__MASK 0xff000000 +#define A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH__SHIFT 24 +static inline uint32_t A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH(uint32_t val) +{ + return ((val) << A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH__MASK; +} + +#define REG_A4XX_HLSQ_UPDATE_CONTROL 0x000023db + +#define REG_A4XX_PC_BINNING_COMMAND 0x00000d00 +#define A4XX_PC_BINNING_COMMAND_BINNING_ENABLE 0x00000001 + +#define REG_A4XX_PC_DRAWCALL_SETUP_OVERRIDE 0x00000d0c + +#define REG_A4XX_PC_PERFCTR_PC_SEL_0 0x00000d10 + +#define REG_A4XX_PC_PERFCTR_PC_SEL_7 0x00000d17 + +#define REG_A4XX_PC_BIN_BASE 0x000021c0 + +#define REG_A4XX_PC_PRIM_VTX_CNTL 0x000021c4 +#define A4XX_PC_PRIM_VTX_CNTL_VAROUT 0x00000001 +#define A4XX_PC_PRIM_VTX_CNTL_PROVOKING_VTX_LAST 0x02000000 +#define A4XX_PC_PRIM_VTX_CNTL_PSIZE 0x04000000 + +#define REG_A4XX_UNKNOWN_21C5 0x000021c5 + +#define REG_A4XX_PC_RESTART_INDEX 0x000021c6 + +#define REG_A4XX_PC_GS_PARAM 0x000021e5 + +#define REG_A4XX_PC_HS_PARAM 0x000021e7 + +#define REG_A4XX_VBIF_VERSION 0x00003000 + +#define REG_A4XX_VBIF_CLKON 0x00003001 +#define A4XX_VBIF_CLKON_FORCE_ON_TESTBUS 0x00000001 + +#define REG_A4XX_VBIF_ABIT_SORT 0x0000301c + +#define REG_A4XX_VBIF_ABIT_SORT_CONF 0x0000301d + +#define REG_A4XX_VBIF_GATE_OFF_WRREQ_EN 0x0000302a + +#define REG_A4XX_VBIF_IN_RD_LIM_CONF0 0x0000302c + +#define REG_A4XX_VBIF_IN_RD_LIM_CONF1 0x0000302d + +#define REG_A4XX_VBIF_IN_WR_LIM_CONF0 0x00003030 + +#define REG_A4XX_VBIF_IN_WR_LIM_CONF1 0x00003031 + +#define REG_A4XX_VBIF_ROUND_ROBIN_QOS_ARB 0x00003049 + +#define REG_A4XX_UNKNOWN_0CC5 0x00000cc5 + +#define REG_A4XX_UNKNOWN_0CC6 0x00000cc6 + +#define REG_A4XX_UNKNOWN_0D01 0x00000d01 + +#define REG_A4XX_UNKNOWN_0E05 0x00000e05 + +#define REG_A4XX_UNKNOWN_0E42 0x00000e42 + +#define REG_A4XX_UNKNOWN_0EC2 0x00000ec2 + +#define REG_A4XX_UNKNOWN_0EC3 0x00000ec3 + +#define REG_A4XX_UNKNOWN_0F03 0x00000f03 + +#define REG_A4XX_UNKNOWN_2001 0x00002001 + +#define REG_A4XX_UNKNOWN_209B 0x0000209b + +#define REG_A4XX_UNKNOWN_20EF 0x000020ef + +#define REG_A4XX_UNKNOWN_20F0 0x000020f0 + +#define REG_A4XX_UNKNOWN_20F1 0x000020f1 + +#define REG_A4XX_UNKNOWN_20F2 0x000020f2 + +#define REG_A4XX_UNKNOWN_20F3 0x000020f3 + +#define REG_A4XX_UNKNOWN_20F4 0x000020f4 + +#define REG_A4XX_UNKNOWN_20F5 0x000020f5 + +#define REG_A4XX_UNKNOWN_20F6 0x000020f6 + +#define REG_A4XX_UNKNOWN_20F7 0x000020f7 + +#define REG_A4XX_UNKNOWN_2152 0x00002152 + +#define REG_A4XX_UNKNOWN_2153 0x00002153 + +#define REG_A4XX_UNKNOWN_2154 0x00002154 + +#define REG_A4XX_UNKNOWN_2155 0x00002155 + +#define REG_A4XX_UNKNOWN_2156 0x00002156 + +#define REG_A4XX_UNKNOWN_2157 0x00002157 + +#define REG_A4XX_UNKNOWN_21C3 0x000021c3 + +#define REG_A4XX_UNKNOWN_21E6 0x000021e6 + +#define REG_A4XX_UNKNOWN_2209 0x00002209 + +#define REG_A4XX_UNKNOWN_22D7 0x000022d7 + +#define REG_A4XX_UNKNOWN_2381 0x00002381 + +#define REG_A4XX_UNKNOWN_23A0 0x000023a0 + +#define REG_A4XX_TEX_SAMP_0 0x00000000 +#define A4XX_TEX_SAMP_0_XY_MAG__MASK 0x00000006 +#define A4XX_TEX_SAMP_0_XY_MAG__SHIFT 1 +static inline uint32_t A4XX_TEX_SAMP_0_XY_MAG(enum a4xx_tex_filter val) +{ + return ((val) << A4XX_TEX_SAMP_0_XY_MAG__SHIFT) & A4XX_TEX_SAMP_0_XY_MAG__MASK; +} +#define A4XX_TEX_SAMP_0_XY_MIN__MASK 0x00000018 +#define A4XX_TEX_SAMP_0_XY_MIN__SHIFT 3 +static inline uint32_t A4XX_TEX_SAMP_0_XY_MIN(enum a4xx_tex_filter val) +{ + return ((val) << A4XX_TEX_SAMP_0_XY_MIN__SHIFT) & A4XX_TEX_SAMP_0_XY_MIN__MASK; +} +#define A4XX_TEX_SAMP_0_WRAP_S__MASK 0x000000e0 +#define A4XX_TEX_SAMP_0_WRAP_S__SHIFT 5 +static inline uint32_t A4XX_TEX_SAMP_0_WRAP_S(enum a4xx_tex_clamp val) +{ + return ((val) << A4XX_TEX_SAMP_0_WRAP_S__SHIFT) & A4XX_TEX_SAMP_0_WRAP_S__MASK; +} +#define A4XX_TEX_SAMP_0_WRAP_T__MASK 0x00000700 +#define A4XX_TEX_SAMP_0_WRAP_T__SHIFT 8 +static inline uint32_t A4XX_TEX_SAMP_0_WRAP_T(enum a4xx_tex_clamp val) +{ + return ((val) << A4XX_TEX_SAMP_0_WRAP_T__SHIFT) & A4XX_TEX_SAMP_0_WRAP_T__MASK; +} +#define A4XX_TEX_SAMP_0_WRAP_R__MASK 0x00003800 +#define A4XX_TEX_SAMP_0_WRAP_R__SHIFT 11 +static inline uint32_t A4XX_TEX_SAMP_0_WRAP_R(enum a4xx_tex_clamp val) +{ + return ((val) << A4XX_TEX_SAMP_0_WRAP_R__SHIFT) & A4XX_TEX_SAMP_0_WRAP_R__MASK; +} + +#define REG_A4XX_TEX_SAMP_1 0x00000001 +#define A4XX_TEX_SAMP_1_COMPARE_FUNC__MASK 0x0000000e +#define A4XX_TEX_SAMP_1_COMPARE_FUNC__SHIFT 1 +static inline uint32_t A4XX_TEX_SAMP_1_COMPARE_FUNC(enum adreno_compare_func val) +{ + return ((val) << A4XX_TEX_SAMP_1_COMPARE_FUNC__SHIFT) & A4XX_TEX_SAMP_1_COMPARE_FUNC__MASK; +} +#define A4XX_TEX_SAMP_1_MAX_LOD__MASK 0x000fff00 +#define A4XX_TEX_SAMP_1_MAX_LOD__SHIFT 8 +static inline uint32_t A4XX_TEX_SAMP_1_MAX_LOD(float val) +{ + return ((((uint32_t)(val * 64.0))) << A4XX_TEX_SAMP_1_MAX_LOD__SHIFT) & A4XX_TEX_SAMP_1_MAX_LOD__MASK; +} +#define A4XX_TEX_SAMP_1_MIN_LOD__MASK 0xfff00000 +#define A4XX_TEX_SAMP_1_MIN_LOD__SHIFT 20 +static inline uint32_t A4XX_TEX_SAMP_1_MIN_LOD(float val) +{ + return ((((uint32_t)(val * 64.0))) << A4XX_TEX_SAMP_1_MIN_LOD__SHIFT) & A4XX_TEX_SAMP_1_MIN_LOD__MASK; +} + +#define REG_A4XX_TEX_CONST_0 0x00000000 +#define A4XX_TEX_CONST_0_TILED 0x00000001 +#define A4XX_TEX_CONST_0_SWIZ_X__MASK 0x00000070 +#define A4XX_TEX_CONST_0_SWIZ_X__SHIFT 4 +static inline uint32_t A4XX_TEX_CONST_0_SWIZ_X(enum a4xx_tex_swiz val) +{ + return ((val) << A4XX_TEX_CONST_0_SWIZ_X__SHIFT) & A4XX_TEX_CONST_0_SWIZ_X__MASK; +} +#define A4XX_TEX_CONST_0_SWIZ_Y__MASK 0x00000380 +#define A4XX_TEX_CONST_0_SWIZ_Y__SHIFT 7 +static inline uint32_t A4XX_TEX_CONST_0_SWIZ_Y(enum a4xx_tex_swiz val) +{ + return ((val) << A4XX_TEX_CONST_0_SWIZ_Y__SHIFT) & A4XX_TEX_CONST_0_SWIZ_Y__MASK; +} +#define A4XX_TEX_CONST_0_SWIZ_Z__MASK 0x00001c00 +#define A4XX_TEX_CONST_0_SWIZ_Z__SHIFT 10 +static inline uint32_t A4XX_TEX_CONST_0_SWIZ_Z(enum a4xx_tex_swiz val) +{ + return ((val) << A4XX_TEX_CONST_0_SWIZ_Z__SHIFT) & A4XX_TEX_CONST_0_SWIZ_Z__MASK; +} +#define A4XX_TEX_CONST_0_SWIZ_W__MASK 0x0000e000 +#define A4XX_TEX_CONST_0_SWIZ_W__SHIFT 13 +static inline uint32_t A4XX_TEX_CONST_0_SWIZ_W(enum a4xx_tex_swiz val) +{ + return ((val) << A4XX_TEX_CONST_0_SWIZ_W__SHIFT) & A4XX_TEX_CONST_0_SWIZ_W__MASK; +} +#define A4XX_TEX_CONST_0_FMT__MASK 0x1fc00000 +#define A4XX_TEX_CONST_0_FMT__SHIFT 22 +static inline uint32_t A4XX_TEX_CONST_0_FMT(enum a4xx_tex_fmt val) +{ + return ((val) << A4XX_TEX_CONST_0_FMT__SHIFT) & A4XX_TEX_CONST_0_FMT__MASK; +} +#define A4XX_TEX_CONST_0_TYPE__MASK 0x60000000 +#define A4XX_TEX_CONST_0_TYPE__SHIFT 29 +static inline uint32_t A4XX_TEX_CONST_0_TYPE(enum a4xx_tex_type val) +{ + return ((val) << A4XX_TEX_CONST_0_TYPE__SHIFT) & A4XX_TEX_CONST_0_TYPE__MASK; +} + +#define REG_A4XX_TEX_CONST_1 0x00000001 +#define A4XX_TEX_CONST_1_HEIGHT__MASK 0x00007fff +#define A4XX_TEX_CONST_1_HEIGHT__SHIFT 0 +static inline uint32_t A4XX_TEX_CONST_1_HEIGHT(uint32_t val) +{ + return ((val) << A4XX_TEX_CONST_1_HEIGHT__SHIFT) & A4XX_TEX_CONST_1_HEIGHT__MASK; +} +#define A4XX_TEX_CONST_1_WIDTH__MASK 0x1fff8000 +#define A4XX_TEX_CONST_1_WIDTH__SHIFT 15 +static inline uint32_t A4XX_TEX_CONST_1_WIDTH(uint32_t val) +{ + return ((val) << A4XX_TEX_CONST_1_WIDTH__SHIFT) & A4XX_TEX_CONST_1_WIDTH__MASK; +} + +#define REG_A4XX_TEX_CONST_2 0x00000002 +#define A4XX_TEX_CONST_2_PITCH__MASK 0x3ffffe00 +#define A4XX_TEX_CONST_2_PITCH__SHIFT 9 +static inline uint32_t A4XX_TEX_CONST_2_PITCH(uint32_t val) +{ + return ((val) << A4XX_TEX_CONST_2_PITCH__SHIFT) & A4XX_TEX_CONST_2_PITCH__MASK; +} +#define A4XX_TEX_CONST_2_SWAP__MASK 0xc0000000 +#define A4XX_TEX_CONST_2_SWAP__SHIFT 30 +static inline uint32_t A4XX_TEX_CONST_2_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A4XX_TEX_CONST_2_SWAP__SHIFT) & A4XX_TEX_CONST_2_SWAP__MASK; +} + +#define REG_A4XX_TEX_CONST_3 0x00000003 +#define A4XX_TEX_CONST_3_LAYERSZ__MASK 0x0000000f +#define A4XX_TEX_CONST_3_LAYERSZ__SHIFT 0 +static inline uint32_t A4XX_TEX_CONST_3_LAYERSZ(uint32_t val) +{ + return ((val >> 12) << A4XX_TEX_CONST_3_LAYERSZ__SHIFT) & A4XX_TEX_CONST_3_LAYERSZ__MASK; +} + +#define REG_A4XX_TEX_CONST_4 0x00000004 +#define A4XX_TEX_CONST_4_BASE__MASK 0xffffffff +#define A4XX_TEX_CONST_4_BASE__SHIFT 0 +static inline uint32_t A4XX_TEX_CONST_4_BASE(uint32_t val) +{ + return ((val) << A4XX_TEX_CONST_4_BASE__SHIFT) & A4XX_TEX_CONST_4_BASE__MASK; +} + +#define REG_A4XX_TEX_CONST_5 0x00000005 + +#define REG_A4XX_TEX_CONST_6 0x00000006 + +#define REG_A4XX_TEX_CONST_7 0x00000007 + + +#endif /* A4XX_XML */ diff --git a/drivers/gpu/drm/msm/adreno/adreno_common.xml.h b/drivers/gpu/drm/msm/adreno/adreno_common.xml.h index cc341bc..a4b33af 100644 --- a/drivers/gpu/drm/msm/adreno/adreno_common.xml.h +++ b/drivers/gpu/drm/msm/adreno/adreno_common.xml.h @@ -11,10 +11,10 @@ The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 364 bytes, from 2013-11-30 14:47:15) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) - /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2014-06-02 15:21:30) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 9859 bytes, from 2014-06-02 15:21:30) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14960 bytes, from 2014-07-27 17:22:13) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 58020 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 41068 bytes, from 2014-08-01 12:22:48) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10551 bytes, from 2014-11-13 22:44:30) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15053 bytes, from 2014-11-09 15:45:47) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 63169 bytes, from 2014-11-13 22:44:18) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 49097 bytes, from 2014-11-14 15:38:00) Copyright (C) 2013-2014 by the following authors: - Rob Clark (robclark) @@ -105,6 +105,7 @@ enum adreno_rb_dither_mode { enum adreno_rb_depth_format { DEPTHX_16 = 0, DEPTHX_24_8 = 1, + DEPTHX_32 = 2, }; enum adreno_rb_copy_control_mode { @@ -132,6 +133,7 @@ enum a3xx_threadmode { }; enum a3xx_instrbuffermode { + CACHE = 0, BUFFER = 1, }; @@ -140,6 +142,13 @@ enum a3xx_threadsize { FOUR_QUADS = 1, }; +enum a3xx_color_swap { + WZYX = 0, + WXYZ = 1, + ZYXW = 2, + XYZW = 3, +}; + #define REG_AXXX_CP_RB_BASE 0x000001c0 #define REG_AXXX_CP_RB_CNTL 0x000001c1 diff --git a/drivers/gpu/drm/msm/adreno/adreno_pm4.xml.h b/drivers/gpu/drm/msm/adreno/adreno_pm4.xml.h index 6ef43f6..6a75cee 100644 --- a/drivers/gpu/drm/msm/adreno/adreno_pm4.xml.h +++ b/drivers/gpu/drm/msm/adreno/adreno_pm4.xml.h @@ -11,10 +11,10 @@ The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 364 bytes, from 2013-11-30 14:47:15) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) - /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2014-06-02 15:21:30) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 9859 bytes, from 2014-06-02 15:21:30) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14960 bytes, from 2014-07-27 17:22:13) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 58020 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 41068 bytes, from 2014-08-01 12:22:48) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10551 bytes, from 2014-11-13 22:44:30) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15053 bytes, from 2014-11-09 15:45:47) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 63169 bytes, from 2014-11-13 22:44:18) +- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 49097 bytes, from 2014-11-14 15:38:00) Copyright (C) 2013-2014 by the following authors: - Rob Clark (robclark) @@ -157,6 +157,7 @@ enum adreno_pm4_type3_packets { CP_IM_STORE = 44, CP_SET_DRAW_INIT_FLAGS = 75, CP_SET_PROTECTED_MODE = 95, + CP_BOOTSTRAP_UCODE = 111, CP_LOAD_STATE = 48, CP_COND_INDIRECT_BUFFER_PFE = 58, CP_COND_INDIRECT_BUFFER_PFD = 50, @@ -278,11 +279,11 @@ static inline uint32_t CP_DRAW_INDX_1_INDEX_SIZE(enum pc_di_index_size val) #define CP_DRAW_INDX_1_NOT_EOP 0x00001000 #define CP_DRAW_INDX_1_SMALL_INDEX 0x00002000 #define CP_DRAW_INDX_1_PRE_DRAW_INITIATOR_ENABLE 0x00004000 -#define CP_DRAW_INDX_1_NUM_INDICES__MASK 0xffff0000 -#define CP_DRAW_INDX_1_NUM_INDICES__SHIFT 16 -static inline uint32_t CP_DRAW_INDX_1_NUM_INDICES(uint32_t val) +#define CP_DRAW_INDX_1_NUM_INSTANCES__MASK 0xff000000 +#define CP_DRAW_INDX_1_NUM_INSTANCES__SHIFT 24 +static inline uint32_t CP_DRAW_INDX_1_NUM_INSTANCES(uint32_t val) { - return ((val) << CP_DRAW_INDX_1_NUM_INDICES__SHIFT) & CP_DRAW_INDX_1_NUM_INDICES__MASK; + return ((val) << CP_DRAW_INDX_1_NUM_INSTANCES__SHIFT) & CP_DRAW_INDX_1_NUM_INSTANCES__MASK; } #define REG_CP_DRAW_INDX_2 0x00000002 @@ -293,20 +294,20 @@ static inline uint32_t CP_DRAW_INDX_2_NUM_INDICES(uint32_t val) return ((val) << CP_DRAW_INDX_2_NUM_INDICES__SHIFT) & CP_DRAW_INDX_2_NUM_INDICES__MASK; } -#define REG_CP_DRAW_INDX_2 0x00000002 -#define CP_DRAW_INDX_2_INDX_BASE__MASK 0xffffffff -#define CP_DRAW_INDX_2_INDX_BASE__SHIFT 0 -static inline uint32_t CP_DRAW_INDX_2_INDX_BASE(uint32_t val) +#define REG_CP_DRAW_INDX_3 0x00000003 +#define CP_DRAW_INDX_3_INDX_BASE__MASK 0xffffffff +#define CP_DRAW_INDX_3_INDX_BASE__SHIFT 0 +static inline uint32_t CP_DRAW_INDX_3_INDX_BASE(uint32_t val) { - return ((val) << CP_DRAW_INDX_2_INDX_BASE__SHIFT) & CP_DRAW_INDX_2_INDX_BASE__MASK; + return ((val) << CP_DRAW_INDX_3_INDX_BASE__SHIFT) & CP_DRAW_INDX_3_INDX_BASE__MASK; } -#define REG_CP_DRAW_INDX_2 0x00000002 -#define CP_DRAW_INDX_2_INDX_SIZE__MASK 0xffffffff -#define CP_DRAW_INDX_2_INDX_SIZE__SHIFT 0 -static inline uint32_t CP_DRAW_INDX_2_INDX_SIZE(uint32_t val) +#define REG_CP_DRAW_INDX_4 0x00000004 +#define CP_DRAW_INDX_4_INDX_SIZE__MASK 0xffffffff +#define CP_DRAW_INDX_4_INDX_SIZE__SHIFT 0 +static inline uint32_t CP_DRAW_INDX_4_INDX_SIZE(uint32_t val) { - return ((val) << CP_DRAW_INDX_2_INDX_SIZE__SHIFT) & CP_DRAW_INDX_2_INDX_SIZE__MASK; + return ((val) << CP_DRAW_INDX_4_INDX_SIZE__SHIFT) & CP_DRAW_INDX_4_INDX_SIZE__MASK; } #define REG_CP_DRAW_INDX_2_0 0x00000000 @@ -345,11 +346,11 @@ static inline uint32_t CP_DRAW_INDX_2_1_INDEX_SIZE(enum pc_di_index_size val) #define CP_DRAW_INDX_2_1_NOT_EOP 0x00001000 #define CP_DRAW_INDX_2_1_SMALL_INDEX 0x00002000 #define CP_DRAW_INDX_2_1_PRE_DRAW_INITIATOR_ENABLE 0x00004000 -#define CP_DRAW_INDX_2_1_NUM_INDICES__MASK 0xffff0000 -#define CP_DRAW_INDX_2_1_NUM_INDICES__SHIFT 16 -static inline uint32_t CP_DRAW_INDX_2_1_NUM_INDICES(uint32_t val) +#define CP_DRAW_INDX_2_1_NUM_INSTANCES__MASK 0xff000000 +#define CP_DRAW_INDX_2_1_NUM_INSTANCES__SHIFT 24 +static inline uint32_t CP_DRAW_INDX_2_1_NUM_INSTANCES(uint32_t val) { - return ((val) << CP_DRAW_INDX_2_1_NUM_INDICES__SHIFT) & CP_DRAW_INDX_2_1_NUM_INDICES__MASK; + return ((val) << CP_DRAW_INDX_2_1_NUM_INSTANCES__SHIFT) & CP_DRAW_INDX_2_1_NUM_INSTANCES__MASK; } #define REG_CP_DRAW_INDX_2_2 0x00000002 @@ -388,11 +389,11 @@ static inline uint32_t CP_DRAW_INDX_OFFSET_0_INDEX_SIZE(enum pc_di_index_size va #define CP_DRAW_INDX_OFFSET_0_NOT_EOP 0x00001000 #define CP_DRAW_INDX_OFFSET_0_SMALL_INDEX 0x00002000 #define CP_DRAW_INDX_OFFSET_0_PRE_DRAW_INITIATOR_ENABLE 0x00004000 -#define CP_DRAW_INDX_OFFSET_0_NUM_INDICES__MASK 0xffff0000 -#define CP_DRAW_INDX_OFFSET_0_NUM_INDICES__SHIFT 16 -static inline uint32_t CP_DRAW_INDX_OFFSET_0_NUM_INDICES(uint32_t val) +#define CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES__MASK 0xffff0000 +#define CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES__SHIFT 16 +static inline uint32_t CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES(uint32_t val) { - return ((val) << CP_DRAW_INDX_OFFSET_0_NUM_INDICES__SHIFT) & CP_DRAW_INDX_OFFSET_0_NUM_INDICES__MASK; + return ((val) << CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES__SHIFT) & CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES__MASK; } #define REG_CP_DRAW_INDX_OFFSET_1 0x00000001 @@ -405,20 +406,22 @@ static inline uint32_t CP_DRAW_INDX_OFFSET_2_NUM_INDICES(uint32_t val) return ((val) << CP_DRAW_INDX_OFFSET_2_NUM_INDICES__SHIFT) & CP_DRAW_INDX_OFFSET_2_NUM_INDICES__MASK; } -#define REG_CP_DRAW_INDX_OFFSET_2 0x00000002 -#define CP_DRAW_INDX_OFFSET_2_INDX_BASE__MASK 0xffffffff -#define CP_DRAW_INDX_OFFSET_2_INDX_BASE__SHIFT 0 -static inline uint32_t CP_DRAW_INDX_OFFSET_2_INDX_BASE(uint32_t val) +#define REG_CP_DRAW_INDX_OFFSET_3 0x00000003 + +#define REG_CP_DRAW_INDX_OFFSET_4 0x00000004 +#define CP_DRAW_INDX_OFFSET_4_INDX_BASE__MASK 0xffffffff +#define CP_DRAW_INDX_OFFSET_4_INDX_BASE__SHIFT 0 +static inline uint32_t CP_DRAW_INDX_OFFSET_4_INDX_BASE(uint32_t val) { - return ((val) << CP_DRAW_INDX_OFFSET_2_INDX_BASE__SHIFT) & CP_DRAW_INDX_OFFSET_2_INDX_BASE__MASK; + return ((val) << CP_DRAW_INDX_OFFSET_4_INDX_BASE__SHIFT) & CP_DRAW_INDX_OFFSET_4_INDX_BASE__MASK; } -#define REG_CP_DRAW_INDX_OFFSET_2 0x00000002 -#define CP_DRAW_INDX_OFFSET_2_INDX_SIZE__MASK 0xffffffff -#define CP_DRAW_INDX_OFFSET_2_INDX_SIZE__SHIFT 0 -static inline uint32_t CP_DRAW_INDX_OFFSET_2_INDX_SIZE(uint32_t val) +#define REG_CP_DRAW_INDX_OFFSET_5 0x00000005 +#define CP_DRAW_INDX_OFFSET_5_INDX_SIZE__MASK 0xffffffff +#define CP_DRAW_INDX_OFFSET_5_INDX_SIZE__SHIFT 0 +static inline uint32_t CP_DRAW_INDX_OFFSET_5_INDX_SIZE(uint32_t val) { - return ((val) << CP_DRAW_INDX_OFFSET_2_INDX_SIZE__SHIFT) & CP_DRAW_INDX_OFFSET_2_INDX_SIZE__MASK; + return ((val) << CP_DRAW_INDX_OFFSET_5_INDX_SIZE__SHIFT) & CP_DRAW_INDX_OFFSET_5_INDX_SIZE__MASK; } #define REG_CP_SET_DRAW_STATE_0 0x00000000 diff --git a/drivers/gpu/drm/msm/dsi/dsi.xml.h b/drivers/gpu/drm/msm/dsi/dsi.xml.h index e965898..448438b7 100644 --- a/drivers/gpu/drm/msm/dsi/dsi.xml.h +++ b/drivers/gpu/drm/msm/dsi/dsi.xml.h @@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/msm.xml ( 647 bytes, from 2013-11-30 14:45:35) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20457 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1615 bytes, from 2014-07-17 15:34:33) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 22517 bytes, from 2014-07-17 15:34:33) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20136 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1940 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 23963 bytes, from 2014-10-31 16:51:46) - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml ( 11712 bytes, from 2013-08-17 17:13:43) - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml ( 344 bytes, from 2013-08-11 19:26:32) -- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-08-01 12:23:53) +- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-10-31 16:48:57) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml ( 600 bytes, from 2013-07-05 19:21:12) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml ( 23613 bytes, from 2014-07-17 15:33:30) diff --git a/drivers/gpu/drm/msm/dsi/mmss_cc.xml.h b/drivers/gpu/drm/msm/dsi/mmss_cc.xml.h index f2bdda9..c102a7f 100644 --- a/drivers/gpu/drm/msm/dsi/mmss_cc.xml.h +++ b/drivers/gpu/drm/msm/dsi/mmss_cc.xml.h @@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/msm.xml ( 647 bytes, from 2013-11-30 14:45:35) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20457 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1615 bytes, from 2014-07-17 15:34:33) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 22517 bytes, from 2014-07-17 15:34:33) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20136 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1940 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 23963 bytes, from 2014-10-31 16:51:46) - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml ( 11712 bytes, from 2013-08-17 17:13:43) - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml ( 344 bytes, from 2013-08-11 19:26:32) -- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-08-01 12:23:53) +- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-10-31 16:48:57) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml ( 600 bytes, from 2013-07-05 19:21:12) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml ( 23613 bytes, from 2014-07-17 15:33:30) diff --git a/drivers/gpu/drm/msm/dsi/sfpb.xml.h b/drivers/gpu/drm/msm/dsi/sfpb.xml.h index e5b071f..a900134 100644 --- a/drivers/gpu/drm/msm/dsi/sfpb.xml.h +++ b/drivers/gpu/drm/msm/dsi/sfpb.xml.h @@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/msm.xml ( 647 bytes, from 2013-11-30 14:45:35) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20457 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1615 bytes, from 2014-07-17 15:34:33) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 22517 bytes, from 2014-07-17 15:34:33) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20136 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1940 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 23963 bytes, from 2014-10-31 16:51:46) - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml ( 11712 bytes, from 2013-08-17 17:13:43) - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml ( 344 bytes, from 2013-08-11 19:26:32) -- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-08-01 12:23:53) +- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-10-31 16:48:57) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml ( 600 bytes, from 2013-07-05 19:21:12) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml ( 23613 bytes, from 2014-07-17 15:33:30) diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.xml.h b/drivers/gpu/drm/msm/hdmi/hdmi.xml.h index 76fd0cf..5b0844b 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi.xml.h +++ b/drivers/gpu/drm/msm/hdmi/hdmi.xml.h @@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/msm.xml ( 647 bytes, from 2013-11-30 14:45:35) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20457 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1615 bytes, from 2014-07-17 15:34:33) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 22517 bytes, from 2014-07-17 15:34:33) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20136 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1940 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 23963 bytes, from 2014-10-31 16:51:46) - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml ( 11712 bytes, from 2013-08-17 17:13:43) - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml ( 344 bytes, from 2013-08-11 19:26:32) -- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-08-01 12:23:53) +- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-10-31 16:48:57) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml ( 600 bytes, from 2013-07-05 19:21:12) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml ( 23613 bytes, from 2014-07-17 15:33:30) diff --git a/drivers/gpu/drm/msm/hdmi/qfprom.xml.h b/drivers/gpu/drm/msm/hdmi/qfprom.xml.h index d53c293..29bd796 100644 --- a/drivers/gpu/drm/msm/hdmi/qfprom.xml.h +++ b/drivers/gpu/drm/msm/hdmi/qfprom.xml.h @@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/msm.xml ( 647 bytes, from 2013-11-30 14:45:35) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20457 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1615 bytes, from 2014-07-17 15:34:33) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 22517 bytes, from 2014-07-17 15:34:33) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20136 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1940 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 23963 bytes, from 2014-10-31 16:51:46) - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml ( 11712 bytes, from 2013-08-17 17:13:43) - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml ( 344 bytes, from 2013-08-11 19:26:32) -- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-08-01 12:23:53) +- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-10-31 16:48:57) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml ( 600 bytes, from 2013-07-05 19:21:12) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml ( 23613 bytes, from 2014-07-17 15:33:30) diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4.xml.h b/drivers/gpu/drm/msm/mdp/mdp4/mdp4.xml.h index 03c0bd9..a4a7f8c 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4.xml.h +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4.xml.h @@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/msm.xml ( 647 bytes, from 2013-11-30 14:45:35) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20457 bytes, from 2014-08-01 12:22:48) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1615 bytes, from 2014-07-17 15:34:33) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 22517 bytes, from 2014-07-17 15:34:33) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20136 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1940 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 23963 bytes, from 2014-10-31 16:51:46) - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml ( 11712 bytes, from 2013-08-17 17:13:43) - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml ( 344 bytes, from 2013-08-11 19:26:32) -- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-08-01 12:23:53) +- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-10-31 16:48:57) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml ( 600 bytes, from 2013-07-05 19:21:12) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml ( 23613 bytes, from 2014-07-17 15:33:30) diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5.xml.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5.xml.h index 67f4f89..e87ef55 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5.xml.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5.xml.h @@ -10,14 +10,14 @@ git clone https://github.com/freedreno/envytools.git The rules-ng-ng source files this header was generated from are: - /home/robclark/src/freedreno/envytools/rnndb/msm.xml ( 647 bytes, from 2013-11-30 14:45:35) - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2013-03-31 16:51:27) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 17996 bytes, from 2013-12-01 19:10:31) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1615 bytes, from 2013-11-30 15:00:52) -- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 22517 bytes, from 2014-06-25 12:55:02) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml ( 20136 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml ( 1940 bytes, from 2014-10-31 16:51:39) +- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml ( 23963 bytes, from 2014-10-31 16:51:46) - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml ( 11712 bytes, from 2013-08-17 17:13:43) - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml ( 344 bytes, from 2013-08-11 19:26:32) -- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1544 bytes, from 2013-08-16 19:17:05) +- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml ( 1686 bytes, from 2014-10-31 16:48:57) - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml ( 600 bytes, from 2013-07-05 19:21:12) -- /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml ( 23613 bytes, from 2014-06-25 12:53:44) +- /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml ( 23613 bytes, from 2014-07-17 15:33:30) Copyright (C) 2013-2014 by the following authors: - Rob Clark (robclark) -- cgit v0.10.2 From 067fef372c7356f64e4d307218df0fae49f9c88e Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Tue, 4 Nov 2014 13:33:14 -0500 Subject: drm/msm/hdmi: refactor bind/init Split up hdmi_init() into hdmi_init() (done at hdmi sub-device bind/probe time) and hdmi_modeset_init() done from master driver's modeset_init(). Anything that can fail due to dependencies on other drivers which may be missing or not probed yet should go in hdmi_init(), so that devm error/cleanup paths work properly. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.c b/drivers/gpu/drm/msm/hdmi/hdmi.c index 9d00dcb..9007761 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi.c +++ b/drivers/gpu/drm/msm/hdmi/hdmi.c @@ -68,24 +68,17 @@ void hdmi_destroy(struct kref *kref) platform_set_drvdata(hdmi->pdev, NULL); } -/* initialize connector */ -struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) +/* construct hdmi at bind/probe time, grab all the resources. If + * we are to EPROBE_DEFER we want to do it here, rather than later + * at modeset_init() time + */ +static struct hdmi *hdmi_init(struct platform_device *pdev) { + struct hdmi_platform_config *config = pdev->dev.platform_data; struct hdmi *hdmi = NULL; - struct msm_drm_private *priv = dev->dev_private; - struct platform_device *pdev = priv->hdmi_pdev; - struct hdmi_platform_config *config; int i, ret; - if (!pdev) { - dev_err(dev->dev, "no hdmi device\n"); - ret = -ENXIO; - goto fail; - } - - config = pdev->dev.platform_data; - - hdmi = kzalloc(sizeof(*hdmi), GFP_KERNEL); + hdmi = devm_kzalloc(&pdev->dev, sizeof(*hdmi), GFP_KERNEL); if (!hdmi) { ret = -ENOMEM; goto fail; @@ -93,12 +86,8 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) kref_init(&hdmi->refcount); - hdmi->dev = dev; hdmi->pdev = pdev; hdmi->config = config; - hdmi->encoder = encoder; - - hdmi_audio_infoframe_init(&hdmi->audio.infoframe); /* not sure about which phy maps to which msm.. probably I miss some */ if (config->phy_init) @@ -108,7 +97,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) if (IS_ERR(hdmi->phy)) { ret = PTR_ERR(hdmi->phy); - dev_err(dev->dev, "failed to load phy: %d\n", ret); + dev_err(&pdev->dev, "failed to load phy: %d\n", ret); hdmi->phy = NULL; goto fail; } @@ -127,7 +116,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) config->hpd_reg_names[i]); if (IS_ERR(reg)) { ret = PTR_ERR(reg); - dev_err(dev->dev, "failed to get hpd regulator: %s (%d)\n", + dev_err(&pdev->dev, "failed to get hpd regulator: %s (%d)\n", config->hpd_reg_names[i], ret); goto fail; } @@ -143,7 +132,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) config->pwr_reg_names[i]); if (IS_ERR(reg)) { ret = PTR_ERR(reg); - dev_err(dev->dev, "failed to get pwr regulator: %s (%d)\n", + dev_err(&pdev->dev, "failed to get pwr regulator: %s (%d)\n", config->pwr_reg_names[i], ret); goto fail; } @@ -158,7 +147,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) clk = devm_clk_get(&pdev->dev, config->hpd_clk_names[i]); if (IS_ERR(clk)) { ret = PTR_ERR(clk); - dev_err(dev->dev, "failed to get hpd clk: %s (%d)\n", + dev_err(&pdev->dev, "failed to get hpd clk: %s (%d)\n", config->hpd_clk_names[i], ret); goto fail; } @@ -173,7 +162,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) clk = devm_clk_get(&pdev->dev, config->pwr_clk_names[i]); if (IS_ERR(clk)) { ret = PTR_ERR(clk); - dev_err(dev->dev, "failed to get pwr clk: %s (%d)\n", + dev_err(&pdev->dev, "failed to get pwr clk: %s (%d)\n", config->pwr_clk_names[i], ret); goto fail; } @@ -184,11 +173,41 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) hdmi->i2c = hdmi_i2c_init(hdmi); if (IS_ERR(hdmi->i2c)) { ret = PTR_ERR(hdmi->i2c); - dev_err(dev->dev, "failed to get i2c: %d\n", ret); + dev_err(&pdev->dev, "failed to get i2c: %d\n", ret); hdmi->i2c = NULL; goto fail; } + return hdmi; + +fail: + if (hdmi) + hdmi_destroy(&hdmi->refcount); + + return ERR_PTR(ret); +} + +/* Second part of initialization, the drm/kms level modeset_init, + * constructs/initializes mode objects, etc, is called from master + * driver (not hdmi sub-device's probe/bind!) + * + * Any resource (regulator/clk/etc) which could be missing at boot + * should be handled in hdmi_init() so that failure happens from + * hdmi sub-device's probe. + */ +int hdmi_modeset_init(struct hdmi *hdmi, + struct drm_device *dev, struct drm_encoder *encoder) +{ + struct msm_drm_private *priv = dev->dev_private; + struct platform_device *pdev = hdmi->pdev; + struct hdmi_platform_config *config = pdev->dev.platform_data; + int ret; + + hdmi->dev = dev; + hdmi->encoder = encoder; + + hdmi_audio_infoframe_init(&hdmi->audio.infoframe); + hdmi->bridge = hdmi_bridge_init(hdmi); if (IS_ERR(hdmi->bridge)) { ret = PTR_ERR(hdmi->bridge); @@ -230,19 +249,20 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder) platform_set_drvdata(pdev, hdmi); - return hdmi; + return 0; fail: - if (hdmi) { - /* bridge/connector are normally destroyed by drm: */ - if (hdmi->bridge) - hdmi->bridge->funcs->destroy(hdmi->bridge); - if (hdmi->connector) - hdmi->connector->funcs->destroy(hdmi->connector); - hdmi_destroy(&hdmi->refcount); + /* bridge/connector are normally destroyed by drm: */ + if (hdmi->bridge) { + hdmi->bridge->funcs->destroy(hdmi->bridge); + hdmi->bridge = NULL; + } + if (hdmi->connector) { + hdmi->connector->funcs->destroy(hdmi->connector); + hdmi->connector = NULL; } - return ERR_PTR(ret); + return ret; } /* @@ -251,11 +271,10 @@ fail: #include -static void set_hdmi_pdev(struct drm_device *dev, - struct platform_device *pdev) +static void set_hdmi(struct drm_device *dev, struct hdmi *hdmi) { struct msm_drm_private *priv = dev->dev_private; - priv->hdmi_pdev = pdev; + priv->hdmi = hdmi; } #ifdef CONFIG_OF @@ -279,6 +298,7 @@ static int get_gpio(struct device *dev, struct device_node *of_node, const char static int hdmi_bind(struct device *dev, struct device *master, void *data) { static struct hdmi_platform_config config = {}; + struct hdmi *hdmi; #ifdef CONFIG_OF struct device_node *of_node = dev->of_node; @@ -369,14 +389,17 @@ static int hdmi_bind(struct device *dev, struct device *master, void *data) } #endif dev->platform_data = &config; - set_hdmi_pdev(dev_get_drvdata(master), to_platform_device(dev)); + hdmi = hdmi_init(to_platform_device(dev)); + if (IS_ERR(hdmi)) + return PTR_ERR(hdmi); + set_hdmi(dev_get_drvdata(master), hdmi); return 0; } static void hdmi_unbind(struct device *dev, struct device *master, void *data) { - set_hdmi_pdev(dev_get_drvdata(master), NULL); + set_hdmi(dev_get_drvdata(master), NULL); } static const struct component_ops hdmi_ops = { diff --git a/drivers/gpu/drm/msm/hdmi/hdmi_phy_8960.c b/drivers/gpu/drm/msm/hdmi/hdmi_phy_8960.c index f408b69..eeed006 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi_phy_8960.c +++ b/drivers/gpu/drm/msm/hdmi/hdmi_phy_8960.c @@ -510,7 +510,7 @@ struct hdmi_phy *hdmi_phy_8960_init(struct hdmi *hdmi) #ifdef CONFIG_COMMON_CLK phy_8960->pll_hw.init = &pll_init; - phy_8960->pll = devm_clk_register(hdmi->dev->dev, &phy_8960->pll_hw); + phy_8960->pll = devm_clk_register(&hdmi->pdev->dev, &phy_8960->pll_hw); if (IS_ERR(phy_8960->pll)) { ret = PTR_ERR(phy_8960->pll); phy_8960->pll = NULL; diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c index ac20472..a62109e4 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c @@ -228,7 +228,6 @@ static int modeset_init(struct mdp4_kms *mdp4_kms) struct drm_encoder *encoder; struct drm_connector *connector; struct drm_panel *panel; - struct hdmi *hdmi; int ret; /* construct non-private planes: */ @@ -326,11 +325,13 @@ static int modeset_init(struct mdp4_kms *mdp4_kms) priv->crtcs[priv->num_crtcs++] = crtc; priv->encoders[priv->num_encoders++] = encoder; - hdmi = hdmi_init(dev, encoder); - if (IS_ERR(hdmi)) { - ret = PTR_ERR(hdmi); - dev_err(dev->dev, "failed to initialize HDMI: %d\n", ret); - goto fail; + if (priv->hdmi) { + /* Construct bridge/connector for HDMI: */ + ret = hdmi_modeset_init(priv->hdmi, dev, encoder); + if (ret) { + dev_err(dev->dev, "failed to initialize HDMI: %d\n", ret); + goto fail; + } } return 0; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c index f2b985b..812c59b 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c @@ -82,6 +82,7 @@ irqreturn_t mdp5_irq(struct msm_kms *kms) { struct mdp_kms *mdp_kms = to_mdp_kms(kms); struct mdp5_kms *mdp5_kms = to_mdp5_kms(mdp_kms); + struct msm_drm_private *priv = mdp5_kms->dev->dev_private; uint32_t intr; intr = mdp5_read(mdp5_kms, REG_MDP5_HW_INTR_STATUS); @@ -92,7 +93,7 @@ irqreturn_t mdp5_irq(struct msm_kms *kms) mdp5_irq_mdp(mdp_kms); if (intr & MDP5_HW_INTR_STATUS_INTR_HDMI) - hdmi_irq(0, mdp5_kms->hdmi); + hdmi_irq(0, priv->hdmi); return IRQ_HANDLED; } diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c index 31a2c63..ce03081 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c @@ -324,11 +324,12 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) priv->encoders[priv->num_encoders++] = encoder; /* Construct bridge/connector for HDMI: */ - mdp5_kms->hdmi = hdmi_init(dev, encoder); - if (IS_ERR(mdp5_kms->hdmi)) { - ret = PTR_ERR(mdp5_kms->hdmi); - dev_err(dev->dev, "failed to initialize HDMI: %d\n", ret); - goto fail; + if (priv->hdmi) { + ret = hdmi_modeset_init(priv->hdmi, dev, encoder); + if (ret) { + dev_err(dev->dev, "failed to initialize HDMI: %d\n", ret); + goto fail; + } } return 0; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h index 5bf340d..c91101d 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h @@ -71,8 +71,6 @@ struct mdp5_kms { struct clk *lut_clk; struct clk *vsync_clk; - struct hdmi *hdmi; - struct mdp_irq error_handler; }; #define to_mdp5_kms(x) container_of(x, struct mdp5_kms, base) diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h index 67f9d0a..a0398b7 100644 --- a/drivers/gpu/drm/msm/msm_drv.h +++ b/drivers/gpu/drm/msm/msm_drv.h @@ -75,7 +75,12 @@ struct msm_drm_private { struct msm_kms *kms; /* subordinate devices, if present: */ - struct platform_device *hdmi_pdev, *gpu_pdev; + struct platform_device *gpu_pdev; + + /* possibly this should be in the kms component, but it is + * shared by both mdp4 and mdp5.. + */ + struct hdmi *hdmi; /* when we have more than one 'msm_gpu' these need to be an array: */ struct msm_gpu *gpu; @@ -202,7 +207,8 @@ struct drm_framebuffer *msm_framebuffer_create(struct drm_device *dev, struct drm_fb_helper *msm_fbdev_init(struct drm_device *dev); struct hdmi; -struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder); +int hdmi_modeset_init(struct hdmi *hdmi, struct drm_device *dev, + struct drm_encoder *encoder); irqreturn_t hdmi_irq(int irq, void *dev_id); void __init hdmi_register(void); void __exit hdmi_unregister(void); -- cgit v0.10.2 From 77a147e742afe33a612d93d61ba153df668de428 Mon Sep 17 00:00:00 2001 From: Daniel Thompson Date: Wed, 12 Nov 2014 11:38:14 +0000 Subject: drm/msm: Allow exported dma-bufs to be mapped Currently msm does not implement gem_prime_mmap. Without this it is not possible to draw onto a dma-buf from userspace (making its very hard to implement the Android rendering model). Fixing this is just a matter of adding a little boilerplate. Signed-off-by: Daniel Thompson Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c index b67ef59..5717d4e 100644 --- a/drivers/gpu/drm/msm/msm_drv.c +++ b/drivers/gpu/drm/msm/msm_drv.c @@ -832,6 +832,7 @@ static struct drm_driver msm_driver = { .gem_prime_import_sg_table = msm_gem_prime_import_sg_table, .gem_prime_vmap = msm_gem_prime_vmap, .gem_prime_vunmap = msm_gem_prime_vunmap, + .gem_prime_mmap = msm_gem_prime_mmap, #ifdef CONFIG_DEBUG_FS .debugfs_init = msm_debugfs_init, .debugfs_cleanup = msm_debugfs_cleanup, diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h index a0398b7..c763c194 100644 --- a/drivers/gpu/drm/msm/msm_drv.h +++ b/drivers/gpu/drm/msm/msm_drv.h @@ -159,6 +159,8 @@ void msm_update_fence(struct drm_device *dev, uint32_t fence); int msm_ioctl_gem_submit(struct drm_device *dev, void *data, struct drm_file *file); +int msm_gem_mmap_obj(struct drm_gem_object *obj, + struct vm_area_struct *vma); int msm_gem_mmap(struct file *filp, struct vm_area_struct *vma); int msm_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf); uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj); @@ -175,6 +177,7 @@ int msm_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev, struct sg_table *msm_gem_prime_get_sg_table(struct drm_gem_object *obj); void *msm_gem_prime_vmap(struct drm_gem_object *obj); void msm_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); +int msm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma); struct drm_gem_object *msm_gem_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sg); int msm_gem_prime_pin(struct drm_gem_object *obj); diff --git a/drivers/gpu/drm/msm/msm_gem_prime.c b/drivers/gpu/drm/msm/msm_gem_prime.c index ad772fe..dd7a7ab 100644 --- a/drivers/gpu/drm/msm/msm_gem_prime.c +++ b/drivers/gpu/drm/msm/msm_gem_prime.c @@ -37,6 +37,19 @@ void msm_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr) /* TODO msm_gem_vunmap() */ } +int msm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) +{ + int ret; + + mutex_lock(&obj->dev->struct_mutex); + ret = drm_gem_mmap_obj(obj, obj->size, vma); + mutex_unlock(&obj->dev->struct_mutex); + if (ret < 0) + return ret; + + return msm_gem_mmap_obj(vma->vm_private_data, vma); +} + struct drm_gem_object *msm_gem_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sg) { -- cgit v0.10.2 From 95a9b8d10cb0ee78c1eb81ea51eeaf94be479416 Mon Sep 17 00:00:00 2001 From: Daniel Thompson Date: Fri, 17 Oct 2014 16:48:54 +0100 Subject: drm/msm: Fix fbdev for 16- and 24-bit modes. Currently forcing the video mode from the kernel command line (for example video=HDMI-A-1:1280x720-16@60) does not correctly set the number of bits per pixel. This is due to a rather aggressive override in msm_fbdev_create(). This is a particular problem for Android bring up because the software EGL fallbacks don't support 32bpp. Since the overrides are actually the default values anyway then this problem can be trivially fixed by removing the overrides completely. Change was tested by dd'ing a test image to /dev/fb0 with no video= (still 32bpp), video=1920x1080-32@60, video=1920x1080-24@60 and video=1920x1080-16@60 . Signed-off-by: Daniel Thompson Cc: David Airlie Cc: Rob Clark Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/msm_fbdev.c b/drivers/gpu/drm/msm/msm_fbdev.c index ab5bfd2..94d55e5 100644 --- a/drivers/gpu/drm/msm/msm_fbdev.c +++ b/drivers/gpu/drm/msm/msm_fbdev.c @@ -93,9 +93,6 @@ static int msm_fbdev_create(struct drm_fb_helper *helper, uint32_t paddr; int ret, size; - sizes->surface_bpp = 32; - sizes->surface_depth = 24; - DBG("create fbdev: %dx%d@%d (%dx%d)", sizes->surface_width, sizes->surface_height, sizes->surface_bpp, sizes->fb_width, sizes->fb_height); -- cgit v0.10.2 From 466c2686717ad643193c7a7dffe74697c841e77d Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 7 Nov 2014 13:04:51 -0500 Subject: drm/msm/mdp4: drop private primary ptr Since primary-plane support in core, we can just use crtc->primary. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c index 1ca87ae..5d2913f 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c @@ -25,7 +25,6 @@ struct mdp4_crtc { struct drm_crtc base; char name[8]; - struct drm_plane *plane; struct drm_plane *planes[8]; int id; int ovlp; @@ -188,7 +187,7 @@ static void pageflip_cb(struct msm_fence_cb *cb) return; drm_framebuffer_reference(fb); - mdp4_plane_set_scanout(mdp4_crtc->plane, fb); + mdp4_plane_set_scanout(crtc->primary, fb); update_scanout(crtc, fb); } @@ -353,7 +352,7 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc, /* grab extra ref for update_scanout() */ drm_framebuffer_reference(crtc->primary->fb); - ret = mdp4_plane_mode_set(mdp4_crtc->plane, crtc, crtc->primary->fb, + ret = mdp4_plane_mode_set(crtc->primary, crtc, crtc->primary->fb, 0, 0, mode->hdisplay, mode->vdisplay, x << 16, y << 16, mode->hdisplay << 16, mode->vdisplay << 16); @@ -419,8 +418,7 @@ static void mdp4_crtc_commit(struct drm_crtc *crtc) static int mdp4_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, struct drm_framebuffer *old_fb) { - struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); - struct drm_plane *plane = mdp4_crtc->plane; + struct drm_plane *plane = crtc->primary; struct drm_display_mode *mode = &crtc->mode; int ret; @@ -729,7 +727,7 @@ static void set_attach(struct drm_crtc *crtc, enum mdp4_pipe pipe_id, mdp4_crtc->planes[pipe_id] = plane; blend_setup(crtc); - if (mdp4_crtc->enabled && (plane != mdp4_crtc->plane)) + if (mdp4_crtc->enabled && (plane != crtc->primary)) crtc_flush(crtc); } @@ -741,7 +739,7 @@ void mdp4_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane) void mdp4_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane) { /* don't actually detatch our primary plane: */ - if (to_mdp4_crtc(crtc)->plane == plane) + if (crtc->primary == plane) return; set_attach(crtc, mdp4_plane_pipe(plane), NULL); } @@ -764,7 +762,6 @@ struct drm_crtc *mdp4_crtc_init(struct drm_device *dev, crtc = &mdp4_crtc->base; - mdp4_crtc->plane = plane; mdp4_crtc->id = id; mdp4_crtc->ovlp = ovlp_id; @@ -791,7 +788,7 @@ struct drm_crtc *mdp4_crtc_init(struct drm_device *dev, drm_crtc_init_with_planes(dev, crtc, plane, NULL, &mdp4_crtc_funcs); drm_crtc_helper_add(crtc, &mdp4_crtc_helper_funcs); - mdp4_plane_install_properties(mdp4_crtc->plane, &crtc->base); + mdp4_plane_install_properties(plane, &crtc->base); return crtc; } -- cgit v0.10.2 From 8845ef80413a6a63a7d91bf15ccbcc9df3466b0c Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Wed, 12 Nov 2014 11:22:19 -0500 Subject: drm/msm/mdp5: drop private primary ptr Since primary-plane support in core, we can just use crtc->primary. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c index 2979a2c..1241447 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c @@ -25,7 +25,6 @@ struct mdp5_crtc { struct drm_crtc base; char name[8]; - struct drm_plane *plane; struct drm_plane *planes[8]; int id; bool enabled; @@ -176,7 +175,7 @@ static void pageflip_cb(struct msm_fence_cb *cb) return; drm_framebuffer_reference(fb); - mdp5_plane_set_scanout(mdp5_crtc->plane, fb); + mdp5_plane_set_scanout(crtc->primary, fb); update_scanout(crtc, fb); } @@ -289,7 +288,7 @@ static int mdp5_crtc_mode_set(struct drm_crtc *crtc, /* grab extra ref for update_scanout() */ drm_framebuffer_reference(crtc->primary->fb); - ret = mdp5_plane_mode_set(mdp5_crtc->plane, crtc, crtc->primary->fb, + ret = mdp5_plane_mode_set(crtc->primary, crtc, crtc->primary->fb, 0, 0, mode->hdisplay, mode->vdisplay, x << 16, y << 16, mode->hdisplay << 16, mode->vdisplay << 16); @@ -330,8 +329,7 @@ static void mdp5_crtc_commit(struct drm_crtc *crtc) static int mdp5_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, struct drm_framebuffer *old_fb) { - struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); - struct drm_plane *plane = mdp5_crtc->plane; + struct drm_plane *plane = crtc->primary; struct drm_display_mode *mode = &crtc->mode; int ret; @@ -511,7 +509,7 @@ static void set_attach(struct drm_crtc *crtc, enum mdp5_pipe pipe_id, mdp5_crtc->planes[pipe_id] = plane; blend_setup(crtc); - if (mdp5_crtc->enabled && (plane != mdp5_crtc->plane)) + if (mdp5_crtc->enabled && (plane != crtc->primary)) crtc_flush(crtc); } @@ -523,7 +521,7 @@ void mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane) void mdp5_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane) { /* don't actually detatch our primary plane: */ - if (to_mdp5_crtc(crtc)->plane == plane) + if (crtc->primary == plane) return; set_attach(crtc, mdp5_plane_pipe(plane), NULL); } @@ -541,7 +539,6 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev, crtc = &mdp5_crtc->base; - mdp5_crtc->plane = plane; mdp5_crtc->id = id; mdp5_crtc->vblank.irq = mdp5_crtc_vblank_irq; @@ -558,7 +555,7 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev, drm_crtc_init_with_planes(dev, crtc, plane, NULL, &mdp5_crtc_funcs); drm_crtc_helper_add(crtc, &mdp5_crtc_helper_funcs); - mdp5_plane_install_properties(mdp5_crtc->plane, &crtc->base); + mdp5_plane_install_properties(plane, &crtc->base); return crtc; } -- cgit v0.10.2 From 88ff1c2f3b5d4826bbe3b6ec3cf12a223dc0cb22 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 7 Nov 2014 13:06:54 -0500 Subject: drm/msm/mdp4: don't care about fb in crtc Since we are configuring things via MDP4_PIPE regs in the plane, it seems like setting the dimensions of the primary plane on the OVLP/DMA regs in crtc is unnecessary. This will make life easier when we want to do a nofb modeset. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c index 5d2913f..d0e6c27 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c @@ -369,8 +369,7 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc, /* take data from pipe: */ mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_BASE(dma), 0); - mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_STRIDE(dma), - crtc->primary->fb->pitches[0]); + mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_STRIDE(dma), 0); mdp4_write(mdp4_kms, REG_MDP4_DMA_DST_SIZE(dma), MDP4_DMA_DST_SIZE_WIDTH(0) | MDP4_DMA_DST_SIZE_HEIGHT(0)); @@ -379,8 +378,7 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc, mdp4_write(mdp4_kms, REG_MDP4_OVLP_SIZE(ovlp), MDP4_OVLP_SIZE_WIDTH(mode->hdisplay) | MDP4_OVLP_SIZE_HEIGHT(mode->vdisplay)); - mdp4_write(mdp4_kms, REG_MDP4_OVLP_STRIDE(ovlp), - crtc->primary->fb->pitches[0]); + mdp4_write(mdp4_kms, REG_MDP4_OVLP_STRIDE(ovlp), 0); mdp4_write(mdp4_kms, REG_MDP4_OVLP_CFG(ovlp), 1); -- cgit v0.10.2 From bb6c018d3556dd683e3f16d24f2e9d60bcaed1b2 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 7 Nov 2014 13:20:50 -0500 Subject: drm/msm/mdp4: drop attached planes table Simplify things a bit for atomic, gets rid of some bookkeeping, and makes the code cleaner. TODO move iterator macro somewhere common. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c index d0e6c27..a28b161 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c @@ -25,7 +25,6 @@ struct mdp4_crtc { struct drm_crtc base; char name[8]; - struct drm_plane *planes[8]; int id; int ovlp; enum mdp4_dma dma; @@ -96,15 +95,14 @@ static void crtc_flush(struct drm_crtc *crtc) { struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); struct mdp4_kms *mdp4_kms = get_kms(crtc); - uint32_t i, flush = 0; + struct drm_plane *plane; + uint32_t flush = 0; - for (i = 0; i < ARRAY_SIZE(mdp4_crtc->planes); i++) { - struct drm_plane *plane = mdp4_crtc->planes[i]; - if (plane) { - enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane); - flush |= pipe2flush(pipe_id); - } + for_each_plane_on_crtc(crtc, plane) { + enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane); + flush |= pipe2flush(pipe_id); } + flush |= ovlp2flush(mdp4_crtc->ovlp); DBG("%s: flush=%08x", mdp4_crtc->name, flush); @@ -254,6 +252,7 @@ static void blend_setup(struct drm_crtc *crtc) { struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); struct mdp4_kms *mdp4_kms = get_kms(crtc); + struct drm_plane *plane; int i, ovlp = mdp4_crtc->ovlp; uint32_t mixer_cfg = 0; static const enum mdp_mixer_stage_id stages[] = { @@ -283,19 +282,16 @@ static void blend_setup(struct drm_crtc *crtc) mdp4_write(mdp4_kms, REG_MDP4_OVLP_TRANSP_HIGH0(ovlp), 0); mdp4_write(mdp4_kms, REG_MDP4_OVLP_TRANSP_HIGH1(ovlp), 0); - for (i = 0; i < ARRAY_SIZE(mdp4_crtc->planes); i++) { - struct drm_plane *plane = mdp4_crtc->planes[i]; - if (plane) { - enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane); - int idx = idxs[pipe_id]; - if (idx > 0) { - const struct mdp_format *format = + for_each_plane_on_crtc(crtc, plane) { + enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane); + int idx = idxs[pipe_id]; + if (idx > 0) { + const struct mdp_format *format = to_mdp_format(msm_framebuffer_format(plane->fb)); - alpha[idx-1] = format->alpha_enable; - } - mixer_cfg = mixercfg(mixer_cfg, mdp4_crtc->mixer, - pipe_id, stages[idx]); + alpha[idx-1] = format->alpha_enable; } + mixer_cfg = mixercfg(mixer_cfg, mdp4_crtc->mixer, + pipe_id, stages[idx]); } /* this shouldn't happen.. and seems to cause underflow: */ @@ -718,12 +714,6 @@ static void set_attach(struct drm_crtc *crtc, enum mdp4_pipe pipe_id, { struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); - BUG_ON(pipe_id >= ARRAY_SIZE(mdp4_crtc->planes)); - - if (mdp4_crtc->planes[pipe_id] == plane) - return; - - mdp4_crtc->planes[pipe_id] = plane; blend_setup(crtc); if (mdp4_crtc->enabled && (plane != crtc->primary)) crtc_flush(crtc); @@ -785,6 +775,7 @@ struct drm_crtc *mdp4_crtc_init(struct drm_device *dev, drm_crtc_init_with_planes(dev, crtc, plane, NULL, &mdp4_crtc_funcs); drm_crtc_helper_add(crtc, &mdp4_crtc_helper_funcs); + plane->crtc = crtc; mdp4_plane_install_properties(plane, &crtc->base); diff --git a/drivers/gpu/drm/msm/msm_kms.h b/drivers/gpu/drm/msm/msm_kms.h index 0643774..15a0fec9 100644 --- a/drivers/gpu/drm/msm/msm_kms.h +++ b/drivers/gpu/drm/msm/msm_kms.h @@ -65,4 +65,9 @@ static inline void msm_kms_init(struct msm_kms *kms, struct msm_kms *mdp4_kms_init(struct drm_device *dev); struct msm_kms *mdp5_kms_init(struct drm_device *dev); +/* TODO move these helper iterator macro somewhere common: */ +#define for_each_plane_on_crtc(_crtc, _plane) \ + list_for_each_entry((_plane), &(_crtc)->dev->mode_config.plane_list, head) \ + if ((_plane)->crtc == (_crtc)) + #endif /* __MSM_KMS_H__ */ -- cgit v0.10.2 From a8cecf33249b80d8a3c2ca5df1c45eb3b5231b28 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Wed, 12 Nov 2014 11:37:12 -0500 Subject: drm/msm/mdp5: drop attached planes table Simplify things a bit for atomic, gets rid of some bookkeeping, and makes the code cleaner. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c index 1241447..b7b32c4 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c @@ -25,7 +25,6 @@ struct mdp5_crtc { struct drm_crtc base; char name[8]; - struct drm_plane *planes[8]; int id; bool enabled; @@ -77,15 +76,14 @@ static void crtc_flush(struct drm_crtc *crtc) struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); struct mdp5_kms *mdp5_kms = get_kms(crtc); int id = mdp5_crtc->id; - uint32_t i, flush = 0; + struct drm_plane *plane; + uint32_t flush = 0; - for (i = 0; i < ARRAY_SIZE(mdp5_crtc->planes); i++) { - struct drm_plane *plane = mdp5_crtc->planes[i]; - if (plane) { - enum mdp5_pipe pipe = mdp5_plane_pipe(plane); - flush |= pipe2flush(pipe); - } + for_each_plane_on_crtc(crtc, plane) { + enum mdp5_pipe pipe = mdp5_plane_pipe(plane); + flush |= pipe2flush(pipe); } + flush |= mixer2flush(mdp5_crtc->id); flush |= MDP5_CTL_FLUSH_CTL; @@ -141,7 +139,8 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file) struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); struct drm_device *dev = crtc->dev; struct drm_pending_vblank_event *event; - unsigned long flags, i; + struct drm_plane *plane; + unsigned long flags; spin_lock_irqsave(&dev->event_lock, flags); event = mdp5_crtc->event; @@ -157,11 +156,8 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file) } spin_unlock_irqrestore(&dev->event_lock, flags); - for (i = 0; i < ARRAY_SIZE(mdp5_crtc->planes); i++) { - struct drm_plane *plane = mdp5_crtc->planes[i]; - if (plane) - mdp5_plane_complete_flip(plane); - } + for_each_plane_on_crtc(crtc, plane) + mdp5_plane_complete_flip(plane); } static void pageflip_cb(struct msm_fence_cb *cb) @@ -502,12 +498,6 @@ static void set_attach(struct drm_crtc *crtc, enum mdp5_pipe pipe_id, { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); - BUG_ON(pipe_id >= ARRAY_SIZE(mdp5_crtc->planes)); - - if (mdp5_crtc->planes[pipe_id] == plane) - return; - - mdp5_crtc->planes[pipe_id] = plane; blend_setup(crtc); if (mdp5_crtc->enabled && (plane != crtc->primary)) crtc_flush(crtc); @@ -554,6 +544,7 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev, drm_crtc_init_with_planes(dev, crtc, plane, NULL, &mdp5_crtc_funcs); drm_crtc_helper_add(crtc, &mdp5_crtc_helper_funcs); + plane->crtc = crtc; mdp5_plane_install_properties(plane, &crtc->base); -- cgit v0.10.2 From 69193e5060dd1c8f7cd614bad05f33f3e042ae6d Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 7 Nov 2014 18:10:04 -0500 Subject: drm/msm: small fence cleanup Give ourselves a way to wait for certain fence #.. makes it easier to wait on a set of bo's, which we'll need for atomic. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c index 5717d4e..1456b1c 100644 --- a/drivers/gpu/drm/msm/msm_drv.c +++ b/drivers/gpu/drm/msm/msm_drv.c @@ -619,6 +619,26 @@ int msm_wait_fence_interruptable(struct drm_device *dev, uint32_t fence, return ret; } +int msm_queue_fence_cb(struct drm_device *dev, + struct msm_fence_cb *cb, uint32_t fence) +{ + struct msm_drm_private *priv = dev->dev_private; + int ret = 0; + + mutex_lock(&dev->struct_mutex); + if (!list_empty(&cb->work.entry)) { + ret = -EINVAL; + } else if (fence > priv->completed_fence) { + cb->fence = fence; + list_add_tail(&cb->work.entry, &priv->fence_cbs); + } else { + queue_work(priv->wq, &cb->work); + } + mutex_unlock(&dev->struct_mutex); + + return ret; +} + /* called from workqueue */ void msm_update_fence(struct drm_device *dev, uint32_t fence) { diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h index c763c194..5901be4 100644 --- a/drivers/gpu/drm/msm/msm_drv.h +++ b/drivers/gpu/drm/msm/msm_drv.h @@ -154,6 +154,8 @@ int msm_register_mmu(struct drm_device *dev, struct msm_mmu *mmu); int msm_wait_fence_interruptable(struct drm_device *dev, uint32_t fence, struct timespec *timeout); +int msm_queue_fence_cb(struct drm_device *dev, + struct msm_fence_cb *cb, uint32_t fence); void msm_update_fence(struct drm_device *dev, uint32_t fence); int msm_ioctl_gem_submit(struct drm_device *dev, void *data, diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c index 4b1b82a..8207862 100644 --- a/drivers/gpu/drm/msm/msm_gem.c +++ b/drivers/gpu/drm/msm/msm_gem.c @@ -397,23 +397,10 @@ void *msm_gem_vaddr(struct drm_gem_object *obj) int msm_gem_queue_inactive_cb(struct drm_gem_object *obj, struct msm_fence_cb *cb) { - struct drm_device *dev = obj->dev; - struct msm_drm_private *priv = dev->dev_private; struct msm_gem_object *msm_obj = to_msm_bo(obj); - int ret = 0; - - mutex_lock(&dev->struct_mutex); - if (!list_empty(&cb->work.entry)) { - ret = -EINVAL; - } else if (is_active(msm_obj)) { - cb->fence = max(msm_obj->read_fence, msm_obj->write_fence); - list_add_tail(&cb->work.entry, &priv->fence_cbs); - } else { - queue_work(priv->wq, &cb->work); - } - mutex_unlock(&dev->struct_mutex); - - return ret; + uint32_t fence = msm_gem_fence(msm_obj, + MSM_PREP_READ | MSM_PREP_WRITE); + return msm_queue_fence_cb(obj->dev, cb, fence); } void msm_gem_move_to_active(struct drm_gem_object *obj, @@ -452,12 +439,8 @@ int msm_gem_cpu_prep(struct drm_gem_object *obj, uint32_t op, int ret = 0; if (is_active(msm_obj)) { - uint32_t fence = 0; + uint32_t fence = msm_gem_fence(msm_obj, op); - if (op & MSM_PREP_READ) - fence = msm_obj->write_fence; - if (op & MSM_PREP_WRITE) - fence = max(fence, msm_obj->read_fence); if (op & MSM_PREP_NOSYNC) timeout = NULL; diff --git a/drivers/gpu/drm/msm/msm_gem.h b/drivers/gpu/drm/msm/msm_gem.h index bfb0526..8fbbd05 100644 --- a/drivers/gpu/drm/msm/msm_gem.h +++ b/drivers/gpu/drm/msm/msm_gem.h @@ -70,6 +70,19 @@ static inline bool is_active(struct msm_gem_object *msm_obj) return msm_obj->gpu != NULL; } +static inline uint32_t msm_gem_fence(struct msm_gem_object *msm_obj, + uint32_t op) +{ + uint32_t fence = 0; + + if (op & MSM_PREP_READ) + fence = msm_obj->write_fence; + if (op & MSM_PREP_WRITE) + fence = max(fence, msm_obj->read_fence); + + return fence; +} + #define MAX_CMDS 4 /* Created per submit-ioctl, to track bo's and cmdstream bufs, etc, -- cgit v0.10.2 From 032f8d5c41082ceb457a601f9c464233a3876291 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 7 Nov 2014 18:11:02 -0500 Subject: drm/msm: remove unused compile-test stub Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h index 5901be4..5ab6c3d 100644 --- a/drivers/gpu/drm/msm/msm_drv.h +++ b/drivers/gpu/drm/msm/msm_drv.h @@ -32,15 +32,6 @@ #include #include - -#if defined(CONFIG_COMPILE_TEST) && !defined(CONFIG_ARCH_QCOM) -/* stubs we need for compile-test: */ -static inline struct device *msm_iommu_get_ctx(const char *ctx_name) -{ - return NULL; -} -#endif - #ifndef CONFIG_OF #include #include -- cgit v0.10.2 From 2638d90abbe7dc23448f6b6b723cb91344cf93aa Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Sat, 8 Nov 2014 09:13:37 -0500 Subject: drm/msm: fb prepare/cleanup Atomic wants to split the prepare/pin from where we actually program the scanout address (so that any part that can fail is done synchronously). Add some fb/gem apis to make this easier to use from the kms parts. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h index 5ab6c3d..4a7e8b8 100644 --- a/drivers/gpu/drm/msm/msm_drv.h +++ b/drivers/gpu/drm/msm/msm_drv.h @@ -160,6 +160,7 @@ uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj); int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id, uint32_t *iova); int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint32_t *iova); +uint32_t msm_gem_iova(struct drm_gem_object *obj, int id); struct page **msm_gem_get_pages(struct drm_gem_object *obj); void msm_gem_put_pages(struct drm_gem_object *obj); void msm_gem_put_iova(struct drm_gem_object *obj, int id); @@ -193,6 +194,9 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev, struct drm_gem_object *msm_gem_import(struct drm_device *dev, uint32_t size, struct sg_table *sgt); +int msm_framebuffer_prepare(struct drm_framebuffer *fb, int id); +void msm_framebuffer_cleanup(struct drm_framebuffer *fb, int id); +uint32_t msm_framebuffer_iova(struct drm_framebuffer *fb, int id, int plane); struct drm_gem_object *msm_framebuffer_bo(struct drm_framebuffer *fb, int plane); const struct msm_format *msm_framebuffer_format(struct drm_framebuffer *fb); struct drm_framebuffer *msm_framebuffer_init(struct drm_device *dev, diff --git a/drivers/gpu/drm/msm/msm_fb.c b/drivers/gpu/drm/msm/msm_fb.c index 81bafdf..acf34ec 100644 --- a/drivers/gpu/drm/msm/msm_fb.c +++ b/drivers/gpu/drm/msm/msm_fb.c @@ -87,6 +87,42 @@ void msm_framebuffer_describe(struct drm_framebuffer *fb, struct seq_file *m) } #endif +/* prepare/pin all the fb's bo's for scanout. Note that it is not valid + * to prepare an fb more multiple different initiator 'id's. But that + * should be fine, since only the scanout (mdpN) side of things needs + * this, the gpu doesn't care about fb's. + */ +int msm_framebuffer_prepare(struct drm_framebuffer *fb, int id) +{ + struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb); + int ret, i, n = drm_format_num_planes(fb->pixel_format); + uint32_t iova; + + for (i = 0; i < n; i++) { + ret = msm_gem_get_iova(msm_fb->planes[i], id, &iova); + DBG("FB[%u]: iova[%d]: %08x (%d)", fb->base.id, i, iova, ret); + if (ret) + return ret; + } + + return 0; +} + +void msm_framebuffer_cleanup(struct drm_framebuffer *fb, int id) +{ + struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb); + int i, n = drm_format_num_planes(fb->pixel_format); + + for (i = 0; i < n; i++) + msm_gem_put_iova(msm_fb->planes[i], id); +} + +uint32_t msm_framebuffer_iova(struct drm_framebuffer *fb, int id, int plane) +{ + struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb); + return msm_gem_iova(msm_fb->planes[plane], id); +} + struct drm_gem_object *msm_framebuffer_bo(struct drm_framebuffer *fb, int plane) { struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb); diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c index 8207862..aaa7898 100644 --- a/drivers/gpu/drm/msm/msm_gem.c +++ b/drivers/gpu/drm/msm/msm_gem.c @@ -309,6 +309,7 @@ int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id, return ret; } +/* get iova, taking a reference. Should have a matching put */ int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint32_t *iova) { struct msm_gem_object *msm_obj = to_msm_bo(obj); @@ -328,6 +329,16 @@ int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint32_t *iova) return ret; } +/* get iova without taking a reference, used in places where you have + * already done a 'msm_gem_get_iova()'. + */ +uint32_t msm_gem_iova(struct drm_gem_object *obj, int id) +{ + struct msm_gem_object *msm_obj = to_msm_bo(obj); + WARN_ON(!msm_obj->domain[id].iova); + return msm_obj->domain[id].iova; +} + void msm_gem_put_iova(struct drm_gem_object *obj, int id) { // XXX TODO .. -- cgit v0.10.2 From 10291bff0a5f811310ac2f6d0e382d00907a2943 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Sat, 8 Nov 2014 09:20:28 -0500 Subject: drm/msm: bit of fb error checking It's a problem that can't happen yet, since we don't support any multi-planar formats yet. But let's avoid nasty surprises when the time comes. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/msm_fb.c b/drivers/gpu/drm/msm/msm_fb.c index acf34ec..f4e42d5 100644 --- a/drivers/gpu/drm/msm/msm_fb.c +++ b/drivers/gpu/drm/msm/msm_fb.c @@ -24,7 +24,7 @@ struct msm_framebuffer { struct drm_framebuffer base; const struct msm_format *format; - struct drm_gem_object *planes[2]; + struct drm_gem_object *planes[3]; }; #define to_msm_framebuffer(x) container_of(x, struct msm_framebuffer, base) @@ -202,6 +202,11 @@ struct drm_framebuffer *msm_framebuffer_init(struct drm_device *dev, msm_fb->format = format; + if (n > ARRAY_SIZE(msm_fb->planes)) { + ret = -EINVAL; + goto fail; + } + for (i = 0; i < n; i++) { unsigned int width = mode_cmd->width / (i ? hsub : 1); unsigned int height = mode_cmd->height / (i ? vsub : 1); -- cgit v0.10.2 From cf3a7e4ce08e6876cdcb80390876647f28a7cf8f Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Sat, 8 Nov 2014 13:21:06 -0500 Subject: drm/msm: atomic core bits The core parts for async commit. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/Makefile b/drivers/gpu/drm/msm/Makefile index 6283dcb..415a145 100644 --- a/drivers/gpu/drm/msm/Makefile +++ b/drivers/gpu/drm/msm/Makefile @@ -30,6 +30,7 @@ msm-y := \ mdp/mdp5/mdp5_kms.o \ mdp/mdp5/mdp5_plane.o \ mdp/mdp5/mdp5_smp.o \ + msm_atomic.o \ msm_drv.o \ msm_fb.o \ msm_gem.o \ diff --git a/drivers/gpu/drm/msm/msm_atomic.c b/drivers/gpu/drm/msm/msm_atomic.c new file mode 100644 index 0000000..8cf3361 --- /dev/null +++ b/drivers/gpu/drm/msm/msm_atomic.c @@ -0,0 +1,163 @@ +/* + * Copyright (C) 2014 Red Hat + * Author: Rob Clark + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program. If not, see . + */ + +#include "msm_drv.h" +#include "msm_kms.h" +#include "msm_gem.h" + +struct msm_commit { + struct drm_atomic_state *state; + uint32_t fence; + struct msm_fence_cb fence_cb; +}; + +static void fence_cb(struct msm_fence_cb *cb); + +static struct msm_commit *new_commit(struct drm_atomic_state *state) +{ + struct msm_commit *c = kzalloc(sizeof(*c), GFP_KERNEL); + + if (!c) + return NULL; + + c->state = state; + /* TODO we might need a way to indicate to run the cb on a + * different wq so wait_for_vblanks() doesn't block retiring + * bo's.. + */ + INIT_FENCE_CB(&c->fence_cb, fence_cb); + + return c; +} + +/* The (potentially) asynchronous part of the commit. At this point + * nothing can fail short of armageddon. + */ +static void complete_commit(struct msm_commit *c) +{ + struct drm_atomic_state *state = c->state; + struct drm_device *dev = state->dev; + + drm_atomic_helper_commit_pre_planes(dev, state); + + drm_atomic_helper_commit_planes(dev, state); + + drm_atomic_helper_commit_post_planes(dev, state); + + drm_atomic_helper_wait_for_vblanks(dev, state); + + drm_atomic_helper_cleanup_planes(dev, state); + + drm_atomic_state_free(state); + + kfree(c); +} + +static void fence_cb(struct msm_fence_cb *cb) +{ + struct msm_commit *c = + container_of(cb, struct msm_commit, fence_cb); + complete_commit(c); +} + +static void add_fb(struct msm_commit *c, struct drm_framebuffer *fb) +{ + struct drm_gem_object *obj = msm_framebuffer_bo(fb, 0); + c->fence = max(c->fence, msm_gem_fence(to_msm_bo(obj), MSM_PREP_READ)); +} + + +/** + * drm_atomic_helper_commit - commit validated state object + * @dev: DRM device + * @state: the driver state object + * @async: asynchronous commit + * + * This function commits a with drm_atomic_helper_check() pre-validated state + * object. This can still fail when e.g. the framebuffer reservation fails. For + * now this doesn't implement asynchronous commits. + * + * RETURNS + * Zero for success or -errno. + */ +int msm_atomic_commit(struct drm_device *dev, + struct drm_atomic_state *state, bool async) +{ + struct msm_commit *c; + int nplanes = dev->mode_config.num_total_plane; + int i, ret; + + ret = drm_atomic_helper_prepare_planes(dev, state); + if (ret) + return ret; + + c = new_commit(state); + + /* + * Figure out what fence to wait for: + */ + for (i = 0; i < nplanes; i++) { + struct drm_plane *plane = state->planes[i]; + struct drm_plane_state *new_state = state->plane_states[i]; + + if (!plane) + continue; + + if (plane->state->fb != new_state->fb) + add_fb(c, new_state->fb); + } + + /* + * This is the point of no return - everything below never fails except + * when the hw goes bonghits. Which means we can commit the new state on + * the software side now. + */ + + drm_atomic_helper_swap_state(dev, state); + + /* + * Everything below can be run asynchronously without the need to grab + * any modeset locks at all under one conditions: It must be guaranteed + * that the asynchronous work has either been cancelled (if the driver + * supports it, which at least requires that the framebuffers get + * cleaned up with drm_atomic_helper_cleanup_planes()) or completed + * before the new state gets committed on the software side with + * drm_atomic_helper_swap_state(). + * + * This scheme allows new atomic state updates to be prepared and + * checked in parallel to the asynchronous completion of the previous + * update. Which is important since compositors need to figure out the + * composition of the next frame right after having submitted the + * current layout. + */ + + if (async) { + msm_queue_fence_cb(dev, &c->fence_cb, c->fence); + return 0; + } + + ret = msm_wait_fence_interruptable(dev, c->fence, NULL); + if (ret) { + WARN_ON(ret); // TODO unswap state back? or?? + kfree(c); + return ret; + } + + complete_commit(c); + + return 0; +} diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c index 1456b1c..d3b791b 100644 --- a/drivers/gpu/drm/msm/msm_drv.c +++ b/drivers/gpu/drm/msm/msm_drv.c @@ -29,6 +29,8 @@ static void msm_fb_output_poll_changed(struct drm_device *dev) static const struct drm_mode_config_funcs mode_config_funcs = { .fb_create = msm_framebuffer_create, .output_poll_changed = msm_fb_output_poll_changed, + .atomic_check = drm_atomic_helper_check, + .atomic_commit = msm_atomic_commit, }; int msm_register_mmu(struct drm_device *dev, struct msm_mmu *mmu) @@ -294,6 +296,8 @@ static int msm_load(struct drm_device *dev, unsigned long flags) goto fail; } + drm_mode_config_reset(dev); + #ifdef CONFIG_DRM_MSM_FBDEV priv->fbdev = msm_fbdev_init(dev); #endif diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h index 4a7e8b8..4b52d75 100644 --- a/drivers/gpu/drm/msm/msm_drv.h +++ b/drivers/gpu/drm/msm/msm_drv.h @@ -39,7 +39,10 @@ #endif #include +#include +#include #include +#include #include #include #include @@ -141,6 +144,9 @@ void __msm_fence_worker(struct work_struct *work); (_cb)->func = _func; \ } while (0) +int msm_atomic_commit(struct drm_device *dev, + struct drm_atomic_state *state, bool async); + int msm_register_mmu(struct drm_device *dev, struct msm_mmu *mmu); int msm_wait_fence_interruptable(struct drm_device *dev, uint32_t fence, -- cgit v0.10.2 From 3e7849efe0371fd63226a63be0ca87cd31705a37 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Sat, 8 Nov 2014 13:23:07 -0500 Subject: drm/msm/hdmi: atomic Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/hdmi/hdmi_connector.c b/drivers/gpu/drm/msm/hdmi/hdmi_connector.c index 4aca2a3..0aecb25 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi_connector.c +++ b/drivers/gpu/drm/msm/hdmi/hdmi_connector.c @@ -401,6 +401,9 @@ static const struct drm_connector_funcs hdmi_connector_funcs = { .detect = hdmi_connector_detect, .fill_modes = drm_helper_probe_single_connector_modes, .destroy = hdmi_connector_destroy, + .reset = drm_atomic_helper_connector_reset, + .atomic_duplicate_state = drm_atomic_helper_connector_duplicate_state, + .atomic_destroy_state = drm_atomic_helper_connector_destroy_state, }; static const struct drm_connector_helper_funcs hdmi_connector_helper_funcs = { -- cgit v0.10.2 From e27c54ffd2b30f9e09f13211869449ec97c6e42f Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Sat, 8 Nov 2014 13:24:08 -0500 Subject: drm/msm/mdp4: atomic Convert mdp4 display controller backend to atomic helpers. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c index a28b161..fef22e8 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c @@ -50,25 +50,11 @@ struct mdp4_crtc { /* if there is a pending flip, these will be non-null: */ struct drm_pending_vblank_event *event; - struct msm_fence_cb pageflip_cb; #define PENDING_CURSOR 0x1 #define PENDING_FLIP 0x2 atomic_t pending; - /* the fb that we logically (from PoV of KMS API) hold a ref - * to. Which we may not yet be scanning out (we may still - * be scanning out previous in case of page_flip while waiting - * for gpu rendering to complete: - */ - struct drm_framebuffer *fb; - - /* the fb that we currently hold a scanout ref to: */ - struct drm_framebuffer *scanout_fb; - - /* for unref'ing framebuffers after scanout completes: */ - struct drm_flip_work unref_fb_work; - /* for unref'ing cursor bo's after scanout completes: */ struct drm_flip_work unref_cursor_work; @@ -110,47 +96,6 @@ static void crtc_flush(struct drm_crtc *crtc) mdp4_write(mdp4_kms, REG_MDP4_OVERLAY_FLUSH, flush); } -static void update_fb(struct drm_crtc *crtc, struct drm_framebuffer *new_fb) -{ - struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); - struct drm_framebuffer *old_fb = mdp4_crtc->fb; - - /* grab reference to incoming scanout fb: */ - drm_framebuffer_reference(new_fb); - mdp4_crtc->base.primary->fb = new_fb; - mdp4_crtc->fb = new_fb; - - if (old_fb) - drm_flip_work_queue(&mdp4_crtc->unref_fb_work, old_fb); -} - -/* unlike update_fb(), take a ref to the new scanout fb *before* updating - * plane, then call this. Needed to ensure we don't unref the buffer that - * is actually still being scanned out. - * - * Note that this whole thing goes away with atomic.. since we can defer - * calling into driver until rendering is done. - */ -static void update_scanout(struct drm_crtc *crtc, struct drm_framebuffer *fb) -{ - struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); - - /* flush updates, to make sure hw is updated to new scanout fb, - * so that we can safely queue unref to current fb (ie. next - * vblank we know hw is done w/ previous scanout_fb). - */ - crtc_flush(crtc); - - if (mdp4_crtc->scanout_fb) - drm_flip_work_queue(&mdp4_crtc->unref_fb_work, - mdp4_crtc->scanout_fb); - - mdp4_crtc->scanout_fb = fb; - - /* enable vblank to complete flip: */ - request_pending(crtc, PENDING_FLIP); -} - /* if file!=NULL, this is preclose potential cancel-flip path */ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file) { @@ -168,38 +113,13 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file) */ if (!file || (event->base.file_priv == file)) { mdp4_crtc->event = NULL; + DBG("%s: send event: %p", mdp4_crtc->name, event); drm_send_vblank_event(dev, mdp4_crtc->id, event); } } spin_unlock_irqrestore(&dev->event_lock, flags); } -static void pageflip_cb(struct msm_fence_cb *cb) -{ - struct mdp4_crtc *mdp4_crtc = - container_of(cb, struct mdp4_crtc, pageflip_cb); - struct drm_crtc *crtc = &mdp4_crtc->base; - struct drm_framebuffer *fb = crtc->primary->fb; - - if (!fb) - return; - - drm_framebuffer_reference(fb); - mdp4_plane_set_scanout(crtc->primary, fb); - update_scanout(crtc, fb); -} - -static void unref_fb_worker(struct drm_flip_work *work, void *val) -{ - struct mdp4_crtc *mdp4_crtc = - container_of(work, struct mdp4_crtc, unref_fb_work); - struct drm_device *dev = mdp4_crtc->base.dev; - - mutex_lock(&dev->mode_config.mutex); - drm_framebuffer_unreference(val); - mutex_unlock(&dev->mode_config.mutex); -} - static void unref_cursor_worker(struct drm_flip_work *work, void *val) { struct mdp4_crtc *mdp4_crtc = @@ -215,7 +135,6 @@ static void mdp4_crtc_destroy(struct drm_crtc *crtc) struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); drm_crtc_cleanup(crtc); - drm_flip_work_cleanup(&mdp4_crtc->unref_fb_work); drm_flip_work_cleanup(&mdp4_crtc->unref_cursor_work); kfree(mdp4_crtc); @@ -323,18 +242,18 @@ static void blend_setup(struct drm_crtc *crtc) mdp4_write(mdp4_kms, REG_MDP4_LAYERMIXER_IN_CFG, mixer_cfg); } -static int mdp4_crtc_mode_set(struct drm_crtc *crtc, - struct drm_display_mode *mode, - struct drm_display_mode *adjusted_mode, - int x, int y, - struct drm_framebuffer *old_fb) +static void mdp4_crtc_mode_set_nofb(struct drm_crtc *crtc) { struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); struct mdp4_kms *mdp4_kms = get_kms(crtc); enum mdp4_dma dma = mdp4_crtc->dma; - int ret, ovlp = mdp4_crtc->ovlp; + int ovlp = mdp4_crtc->ovlp; + struct drm_display_mode *mode; + + if (WARN_ON(!crtc->state)) + return; - mode = adjusted_mode; + mode = &crtc->state->adjusted_mode; DBG("%s: set mode: %d:\"%s\" %d %d %d %d %d %d %d %d %d %d 0x%x 0x%x", mdp4_crtc->name, mode->base.id, mode->name, @@ -345,20 +264,6 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc, mode->vsync_end, mode->vtotal, mode->type, mode->flags); - /* grab extra ref for update_scanout() */ - drm_framebuffer_reference(crtc->primary->fb); - - ret = mdp4_plane_mode_set(crtc->primary, crtc, crtc->primary->fb, - 0, 0, mode->hdisplay, mode->vdisplay, - x << 16, y << 16, - mode->hdisplay << 16, mode->vdisplay << 16); - if (ret) { - drm_framebuffer_unreference(crtc->primary->fb); - dev_err(crtc->dev->dev, "%s: failed to set mode on plane: %d\n", - mdp4_crtc->name, ret); - return ret; - } - mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_SIZE(dma), MDP4_DMA_SRC_SIZE_WIDTH(mode->hdisplay) | MDP4_DMA_SRC_SIZE_HEIGHT(mode->vdisplay)); @@ -383,11 +288,6 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc, mdp4_write(mdp4_kms, REG_MDP4_DMA_E_QUANT(1), 0x00ff0000); mdp4_write(mdp4_kms, REG_MDP4_DMA_E_QUANT(2), 0x00ff0000); } - - update_fb(crtc, crtc->primary->fb); - update_scanout(crtc, crtc->primary->fb); - - return 0; } static void mdp4_crtc_prepare(struct drm_crtc *crtc) @@ -409,59 +309,51 @@ static void mdp4_crtc_commit(struct drm_crtc *crtc) drm_crtc_vblank_put(crtc); } -static int mdp4_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, - struct drm_framebuffer *old_fb) +static void mdp4_crtc_load_lut(struct drm_crtc *crtc) +{ +} + +static int mdp4_crtc_atomic_check(struct drm_crtc *crtc, + struct drm_crtc_state *state) { - struct drm_plane *plane = crtc->primary; - struct drm_display_mode *mode = &crtc->mode; - int ret; + struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); + struct drm_device *dev = crtc->dev; - /* grab extra ref for update_scanout() */ - drm_framebuffer_reference(crtc->primary->fb); + DBG("%s: check", mdp4_crtc->name); - ret = mdp4_plane_mode_set(plane, crtc, crtc->primary->fb, - 0, 0, mode->hdisplay, mode->vdisplay, - x << 16, y << 16, - mode->hdisplay << 16, mode->vdisplay << 16); - if (ret) { - drm_framebuffer_unreference(crtc->primary->fb); - return ret; + if (mdp4_crtc->event) { + dev_err(dev->dev, "already pending flip!\n"); + return -EBUSY; } - update_fb(crtc, crtc->primary->fb); - update_scanout(crtc, crtc->primary->fb); + // TODO anything else to check? return 0; } -static void mdp4_crtc_load_lut(struct drm_crtc *crtc) +static void mdp4_crtc_atomic_begin(struct drm_crtc *crtc) { + struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); + DBG("%s: begin", mdp4_crtc->name); } -static int mdp4_crtc_page_flip(struct drm_crtc *crtc, - struct drm_framebuffer *new_fb, - struct drm_pending_vblank_event *event, - uint32_t page_flip_flags) +static void mdp4_crtc_atomic_flush(struct drm_crtc *crtc) { struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); struct drm_device *dev = crtc->dev; - struct drm_gem_object *obj; unsigned long flags; - if (mdp4_crtc->event) { - dev_err(dev->dev, "already pending flip!\n"); - return -EBUSY; - } + DBG("%s: flush", mdp4_crtc->name); - obj = msm_framebuffer_bo(new_fb, 0); + WARN_ON(mdp4_crtc->event); spin_lock_irqsave(&dev->event_lock, flags); - mdp4_crtc->event = event; + mdp4_crtc->event = crtc->state->event; spin_unlock_irqrestore(&dev->event_lock, flags); - update_fb(crtc, new_fb); - - return msm_gem_queue_inactive_cb(obj, &mdp4_crtc->pageflip_cb); + blend_setup(crtc); + crtc_flush(crtc); + request_pending(crtc, PENDING_FLIP); } static int mdp4_crtc_set_property(struct drm_crtc *crtc, @@ -599,22 +491,29 @@ static int mdp4_crtc_cursor_move(struct drm_crtc *crtc, int x, int y) } static const struct drm_crtc_funcs mdp4_crtc_funcs = { - .set_config = drm_crtc_helper_set_config, + .set_config = drm_atomic_helper_set_config, .destroy = mdp4_crtc_destroy, - .page_flip = mdp4_crtc_page_flip, + .page_flip = drm_atomic_helper_page_flip, .set_property = mdp4_crtc_set_property, .cursor_set = mdp4_crtc_cursor_set, .cursor_move = mdp4_crtc_cursor_move, + .reset = drm_atomic_helper_crtc_reset, + .atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state, + .atomic_destroy_state = drm_atomic_helper_crtc_destroy_state, }; static const struct drm_crtc_helper_funcs mdp4_crtc_helper_funcs = { .dpms = mdp4_crtc_dpms, .mode_fixup = mdp4_crtc_mode_fixup, - .mode_set = mdp4_crtc_mode_set, + .mode_set_nofb = mdp4_crtc_mode_set_nofb, + .mode_set = drm_helper_crtc_mode_set, + .mode_set_base = drm_helper_crtc_mode_set_base, .prepare = mdp4_crtc_prepare, .commit = mdp4_crtc_commit, - .mode_set_base = mdp4_crtc_mode_set_base, .load_lut = mdp4_crtc_load_lut, + .atomic_check = mdp4_crtc_atomic_check, + .atomic_begin = mdp4_crtc_atomic_begin, + .atomic_flush = mdp4_crtc_atomic_flush, }; static void mdp4_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus) @@ -630,7 +529,6 @@ static void mdp4_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus) if (pending & PENDING_FLIP) { complete_flip(crtc, NULL); - drm_flip_work_commit(&mdp4_crtc->unref_fb_work, priv->wq); } if (pending & PENDING_CURSOR) { @@ -655,7 +553,8 @@ uint32_t mdp4_crtc_vblank(struct drm_crtc *crtc) void mdp4_crtc_cancel_pending_flip(struct drm_crtc *crtc, struct drm_file *file) { - DBG("cancel: %p", file); + struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); + DBG("%s: cancel: %p", mdp4_crtc->name, file); complete_flip(crtc, file); } @@ -709,29 +608,6 @@ void mdp4_crtc_set_intf(struct drm_crtc *crtc, enum mdp4_intf intf, int mixer) mdp4_write(mdp4_kms, REG_MDP4_DISP_INTF_SEL, intf_sel); } -static void set_attach(struct drm_crtc *crtc, enum mdp4_pipe pipe_id, - struct drm_plane *plane) -{ - struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); - - blend_setup(crtc); - if (mdp4_crtc->enabled && (plane != crtc->primary)) - crtc_flush(crtc); -} - -void mdp4_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane) -{ - set_attach(crtc, mdp4_plane_pipe(plane), plane); -} - -void mdp4_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane) -{ - /* don't actually detatch our primary plane: */ - if (crtc->primary == plane) - return; - set_attach(crtc, mdp4_plane_pipe(plane), NULL); -} - static const char *dma_names[] = { "DMA_P", "DMA_S", "DMA_E", }; @@ -766,13 +642,9 @@ struct drm_crtc *mdp4_crtc_init(struct drm_device *dev, spin_lock_init(&mdp4_crtc->cursor.lock); - drm_flip_work_init(&mdp4_crtc->unref_fb_work, - "unref fb", unref_fb_worker); drm_flip_work_init(&mdp4_crtc->unref_cursor_work, "unref cursor", unref_cursor_worker); - INIT_FENCE_CB(&mdp4_crtc->pageflip_cb, pageflip_cb); - drm_crtc_init_with_planes(dev, crtc, plane, NULL, &mdp4_crtc_funcs); drm_crtc_helper_add(crtc, &mdp4_crtc_helper_funcs); plane->crtc = crtc; diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h index 9ff6e7c..7706452 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h @@ -194,14 +194,6 @@ uint32_t mdp4_get_formats(enum mdp4_pipe pipe_id, uint32_t *pixel_formats, void mdp4_plane_install_properties(struct drm_plane *plane, struct drm_mode_object *obj); -void mdp4_plane_set_scanout(struct drm_plane *plane, - struct drm_framebuffer *fb); -int mdp4_plane_mode_set(struct drm_plane *plane, - struct drm_crtc *crtc, struct drm_framebuffer *fb, - int crtc_x, int crtc_y, - unsigned int crtc_w, unsigned int crtc_h, - uint32_t src_x, uint32_t src_y, - uint32_t src_w, uint32_t src_h); enum mdp4_pipe mdp4_plane_pipe(struct drm_plane *plane); struct drm_plane *mdp4_plane_init(struct drm_device *dev, enum mdp4_pipe pipe_id, bool private_plane); @@ -210,8 +202,6 @@ uint32_t mdp4_crtc_vblank(struct drm_crtc *crtc); void mdp4_crtc_cancel_pending_flip(struct drm_crtc *crtc, struct drm_file *file); void mdp4_crtc_set_config(struct drm_crtc *crtc, uint32_t config); void mdp4_crtc_set_intf(struct drm_crtc *crtc, enum mdp4_intf intf, int mixer); -void mdp4_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane); -void mdp4_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane); struct drm_crtc *mdp4_crtc_init(struct drm_device *dev, struct drm_plane *plane, int id, int ovlp_id, enum mdp4_dma dma_id); diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_lvds_connector.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_lvds_connector.c index 3100346..4ddc28e1 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_lvds_connector.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_lvds_connector.c @@ -98,6 +98,9 @@ static const struct drm_connector_funcs mdp4_lvds_connector_funcs = { .detect = mdp4_lvds_connector_detect, .fill_modes = drm_helper_probe_single_connector_modes, .destroy = mdp4_lvds_connector_destroy, + .reset = drm_atomic_helper_connector_reset, + .atomic_duplicate_state = drm_atomic_helper_connector_duplicate_state, + .atomic_destroy_state = drm_atomic_helper_connector_destroy_state, }; static const struct drm_connector_helper_funcs mdp4_lvds_connector_helper_funcs = { diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c index 66f33db..76d0a40 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c @@ -31,47 +31,26 @@ struct mdp4_plane { }; #define to_mdp4_plane(x) container_of(x, struct mdp4_plane, base) -static struct mdp4_kms *get_kms(struct drm_plane *plane) -{ - struct msm_drm_private *priv = plane->dev->dev_private; - return to_mdp4_kms(to_mdp_kms(priv->kms)); -} - -static int mdp4_plane_update(struct drm_plane *plane, +static void mdp4_plane_set_scanout(struct drm_plane *plane, + struct drm_framebuffer *fb); +static int mdp4_plane_mode_set(struct drm_plane *plane, struct drm_crtc *crtc, struct drm_framebuffer *fb, int crtc_x, int crtc_y, unsigned int crtc_w, unsigned int crtc_h, uint32_t src_x, uint32_t src_y, - uint32_t src_w, uint32_t src_h) -{ - struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane); - - mdp4_plane->enabled = true; - - if (plane->fb) - drm_framebuffer_unreference(plane->fb); - - drm_framebuffer_reference(fb); - - return mdp4_plane_mode_set(plane, crtc, fb, - crtc_x, crtc_y, crtc_w, crtc_h, - src_x, src_y, src_w, src_h); -} + uint32_t src_w, uint32_t src_h); -static int mdp4_plane_disable(struct drm_plane *plane) +static struct mdp4_kms *get_kms(struct drm_plane *plane) { - struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane); - DBG("%s: disable", mdp4_plane->name); - if (plane->crtc) - mdp4_crtc_detach(plane->crtc, plane); - return 0; + struct msm_drm_private *priv = plane->dev->dev_private; + return to_mdp4_kms(to_mdp_kms(priv->kms)); } static void mdp4_plane_destroy(struct drm_plane *plane) { struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane); - mdp4_plane_disable(plane); + drm_plane_helper_disable(plane); drm_plane_cleanup(plane); kfree(mdp4_plane); @@ -92,19 +71,74 @@ int mdp4_plane_set_property(struct drm_plane *plane, } static const struct drm_plane_funcs mdp4_plane_funcs = { - .update_plane = mdp4_plane_update, - .disable_plane = mdp4_plane_disable, + .update_plane = drm_atomic_helper_update_plane, + .disable_plane = drm_atomic_helper_disable_plane, .destroy = mdp4_plane_destroy, .set_property = mdp4_plane_set_property, + .reset = drm_atomic_helper_plane_reset, + .atomic_duplicate_state = drm_atomic_helper_plane_duplicate_state, + .atomic_destroy_state = drm_atomic_helper_plane_destroy_state, }; -void mdp4_plane_set_scanout(struct drm_plane *plane, +static int mdp4_plane_prepare_fb(struct drm_plane *plane, + struct drm_framebuffer *fb) +{ + struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane); + struct mdp4_kms *mdp4_kms = get_kms(plane); + + DBG("%s: prepare: FB[%u]", mdp4_plane->name, fb->base.id); + return msm_framebuffer_prepare(fb, mdp4_kms->id); +} + +static void mdp4_plane_cleanup_fb(struct drm_plane *plane, + struct drm_framebuffer *fb) +{ + struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane); + struct mdp4_kms *mdp4_kms = get_kms(plane); + + DBG("%s: cleanup: FB[%u]", mdp4_plane->name, fb->base.id); + msm_framebuffer_cleanup(fb, mdp4_kms->id); +} + + +static int mdp4_plane_atomic_check(struct drm_plane *plane, + struct drm_plane_state *state) +{ + return 0; +} + +static void mdp4_plane_atomic_update(struct drm_plane *plane) +{ + struct drm_plane_state *state = plane->state; + int ret; + + ret = mdp4_plane_mode_set(plane, + state->crtc, state->fb, + state->crtc_x, state->crtc_y, + state->crtc_w, state->crtc_h, + state->src_x, state->src_y, + state->src_w, state->src_h); + /* atomic_check should have ensured that this doesn't fail */ + WARN_ON(ret < 0); +} + +static const struct drm_plane_helper_funcs mdp4_plane_helper_funcs = { + .prepare_fb = mdp4_plane_prepare_fb, + .cleanup_fb = mdp4_plane_cleanup_fb, + .atomic_check = mdp4_plane_atomic_check, + .atomic_update = mdp4_plane_atomic_update, +}; + +static void mdp4_plane_set_scanout(struct drm_plane *plane, struct drm_framebuffer *fb) { struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane); struct mdp4_kms *mdp4_kms = get_kms(plane); enum mdp4_pipe pipe = mdp4_plane->pipe; - uint32_t iova; + uint32_t iova = msm_framebuffer_iova(fb, mdp4_kms->id, 0); + + DBG("%s: set_scanout: %08x (%u)", mdp4_plane->name, + iova, fb->pitches[0]); mdp4_write(mdp4_kms, REG_MDP4_PIPE_SRC_STRIDE_A(pipe), MDP4_PIPE_SRC_STRIDE_A_P0(fb->pitches[0]) | @@ -114,7 +148,6 @@ void mdp4_plane_set_scanout(struct drm_plane *plane, MDP4_PIPE_SRC_STRIDE_B_P2(fb->pitches[2]) | MDP4_PIPE_SRC_STRIDE_B_P3(fb->pitches[3])); - msm_gem_get_iova(msm_framebuffer_bo(fb, 0), mdp4_kms->id, &iova); mdp4_write(mdp4_kms, REG_MDP4_PIPE_SRCP0_BASE(pipe), iova); plane->fb = fb; @@ -122,7 +155,7 @@ void mdp4_plane_set_scanout(struct drm_plane *plane, #define MDP4_VG_PHASE_STEP_DEFAULT 0x20000000 -int mdp4_plane_mode_set(struct drm_plane *plane, +static int mdp4_plane_mode_set(struct drm_plane *plane, struct drm_crtc *crtc, struct drm_framebuffer *fb, int crtc_x, int crtc_y, unsigned int crtc_w, unsigned int crtc_h, @@ -137,6 +170,11 @@ int mdp4_plane_mode_set(struct drm_plane *plane, uint32_t phasex_step = MDP4_VG_PHASE_STEP_DEFAULT; uint32_t phasey_step = MDP4_VG_PHASE_STEP_DEFAULT; + if (!(crtc && fb)) { + DBG("%s: disabled!", mdp4_plane->name); + return 0; + } + /* src values are in Q16 fixed point, convert to integer: */ src_x = src_x >> 16; src_y = src_y >> 16; @@ -197,9 +235,6 @@ int mdp4_plane_mode_set(struct drm_plane *plane, mdp4_write(mdp4_kms, REG_MDP4_PIPE_PHASEX_STEP(pipe), phasex_step); mdp4_write(mdp4_kms, REG_MDP4_PIPE_PHASEY_STEP(pipe), phasey_step); - /* TODO detach from old crtc (if we had more than one) */ - mdp4_crtc_attach(crtc, plane); - return 0; } @@ -239,9 +274,12 @@ struct drm_plane *mdp4_plane_init(struct drm_device *dev, ARRAY_SIZE(mdp4_plane->formats)); type = private_plane ? DRM_PLANE_TYPE_PRIMARY : DRM_PLANE_TYPE_OVERLAY; - drm_universal_plane_init(dev, plane, 0xff, &mdp4_plane_funcs, - mdp4_plane->formats, mdp4_plane->nformats, - type); + ret = drm_universal_plane_init(dev, plane, 0xff, &mdp4_plane_funcs, + mdp4_plane->formats, mdp4_plane->nformats, type); + if (ret) + goto fail; + + drm_plane_helper_add(plane, &mdp4_plane_helper_funcs); mdp4_plane_install_properties(plane, &plane->base); -- cgit v0.10.2 From 257d06f7043a432774a315223053730f8b170b0e Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Wed, 12 Nov 2014 15:25:50 -0500 Subject: drm/msm: small mmap offset cleanups Use pre-computed iova when unmapping, to reduce the places we assume iova and mmap offset are (at the moment) the same. And get rid of an extra drm_gem_free_mmap_offset() call (since it is already called from drm_gem_object_release()) Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c index aaa7898..4a6f0e4 100644 --- a/drivers/gpu/drm/msm/msm_gem.c +++ b/drivers/gpu/drm/msm/msm_gem.c @@ -519,13 +519,11 @@ void msm_gem_free_object(struct drm_gem_object *obj) for (id = 0; id < ARRAY_SIZE(msm_obj->domain); id++) { struct msm_mmu *mmu = priv->mmus[id]; if (mmu && msm_obj->domain[id].iova) { - uint32_t offset = (uint32_t)mmap_offset(obj); + uint32_t offset = msm_obj->domain[id].iova; mmu->funcs->unmap(mmu, offset, msm_obj->sgt, obj->size); } } - drm_gem_free_mmap_offset(obj); - if (obj->import_attach) { if (msm_obj->vaddr) dma_buf_vunmap(obj->import_attach->dmabuf, msm_obj->vaddr); -- cgit v0.10.2 From 91b74e9761d785f41156383ad5ec7c437dfb2960 Mon Sep 17 00:00:00 2001 From: Aravind Ganesan Date: Mon, 8 Sep 2014 10:57:28 -0600 Subject: drm/msm: Handle register offset differences between a3xx and a4xx Register offsets have changed between a3xx and a4xx GPUs. To be able access these registers in common code, we create a lookup table, and set of read-write APIs to access the register through the lookup table. Signed-off-by: Aravind Ganesan [robclark: remove REG_ADRENO_UNDEFINED, just use zero, and minor tweaks for latest generated headers] Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/adreno/a3xx_gpu.c b/drivers/gpu/drm/msm/adreno/a3xx_gpu.c index 218c5b0..b66c53b 100644 --- a/drivers/gpu/drm/msm/adreno/a3xx_gpu.c +++ b/drivers/gpu/drm/msm/adreno/a3xx_gpu.c @@ -2,6 +2,8 @@ * Copyright (C) 2013 Red Hat * Author: Rob Clark * + * Copyright (c) 2014 The Linux Foundation. All rights reserved. + * * This program is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 as published by * the Free Software Foundation. @@ -406,6 +408,94 @@ static void a3xx_dump(struct msm_gpu *gpu) gpu_read(gpu, REG_A3XX_RBBM_STATUS)); adreno_dump(gpu); } +/* Register offset defines for A3XX */ +static const unsigned int a3xx_register_offsets[REG_ADRENO_REGISTER_MAX] = { + REG_ADRENO_DEFINE(REG_ADRENO_CP_DEBUG, REG_AXXX_CP_DEBUG), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_WADDR, REG_AXXX_CP_ME_RAM_WADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_DATA, REG_AXXX_CP_ME_RAM_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_PFP_UCODE_DATA, + REG_A3XX_CP_PFP_UCODE_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_PFP_UCODE_ADDR, + REG_A3XX_CP_PFP_UCODE_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_WFI_PEND_CTR, REG_A3XX_CP_WFI_PEND_CTR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_BASE, REG_AXXX_CP_RB_BASE), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_RPTR_ADDR, REG_AXXX_CP_RB_RPTR_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_RPTR, REG_AXXX_CP_RB_RPTR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_WPTR, REG_AXXX_CP_RB_WPTR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_PROTECT_CTRL, REG_A3XX_CP_PROTECT_CTRL), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_CNTL, REG_AXXX_CP_ME_CNTL), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_CNTL, REG_AXXX_CP_RB_CNTL), + REG_ADRENO_DEFINE(REG_ADRENO_CP_IB1_BASE, REG_AXXX_CP_IB1_BASE), + REG_ADRENO_DEFINE(REG_ADRENO_CP_IB1_BUFSZ, REG_AXXX_CP_IB1_BUFSZ), + REG_ADRENO_DEFINE(REG_ADRENO_CP_IB2_BASE, REG_AXXX_CP_IB2_BASE), + REG_ADRENO_DEFINE(REG_ADRENO_CP_IB2_BUFSZ, REG_AXXX_CP_IB2_BUFSZ), + REG_ADRENO_DEFINE(REG_ADRENO_CP_TIMESTAMP, REG_AXXX_CP_SCRATCH_REG0), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_RADDR, REG_AXXX_CP_ME_RAM_RADDR), + REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_ADDR, REG_AXXX_SCRATCH_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_UMSK, REG_AXXX_SCRATCH_UMSK), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ROQ_ADDR, REG_A3XX_CP_ROQ_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ROQ_DATA, REG_A3XX_CP_ROQ_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_ADDR, REG_A3XX_CP_MERCIU_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_DATA, REG_A3XX_CP_MERCIU_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_DATA2, REG_A3XX_CP_MERCIU_DATA2), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MEQ_ADDR, REG_A3XX_CP_MEQ_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MEQ_DATA, REG_A3XX_CP_MEQ_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_HW_FAULT, REG_A3XX_CP_HW_FAULT), + REG_ADRENO_DEFINE(REG_ADRENO_CP_PROTECT_STATUS, + REG_A3XX_CP_PROTECT_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_STATUS, REG_A3XX_RBBM_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_CTL, + REG_A3XX_RBBM_PERFCTR_CTL), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD0, + REG_A3XX_RBBM_PERFCTR_LOAD_CMD0), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD1, + REG_A3XX_RBBM_PERFCTR_LOAD_CMD1), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_PWR_1_LO, + REG_A3XX_RBBM_PERFCTR_PWR_1_LO), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_0_MASK, REG_A3XX_RBBM_INT_0_MASK), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_0_STATUS, + REG_A3XX_RBBM_INT_0_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_ERROR_STATUS, + REG_A3XX_RBBM_AHB_ERROR_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_CMD, REG_A3XX_RBBM_AHB_CMD), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_CLEAR_CMD, + REG_A3XX_RBBM_INT_CLEAR_CMD), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_CLOCK_CTL, REG_A3XX_RBBM_CLOCK_CTL), + REG_ADRENO_DEFINE(REG_ADRENO_VPC_DEBUG_RAM_SEL, + REG_A3XX_VPC_VPC_DEBUG_RAM_SEL), + REG_ADRENO_DEFINE(REG_ADRENO_VPC_DEBUG_RAM_READ, + REG_A3XX_VPC_VPC_DEBUG_RAM_READ), + REG_ADRENO_DEFINE(REG_ADRENO_VSC_SIZE_ADDRESS, + REG_A3XX_VSC_SIZE_ADDRESS), + REG_ADRENO_DEFINE(REG_ADRENO_VFD_CONTROL_0, REG_A3XX_VFD_CONTROL_0), + REG_ADRENO_DEFINE(REG_ADRENO_VFD_INDEX_MAX, REG_A3XX_VFD_INDEX_MAX), + REG_ADRENO_DEFINE(REG_ADRENO_SP_VS_PVT_MEM_ADDR_REG, + REG_A3XX_SP_VS_PVT_MEM_ADDR_REG), + REG_ADRENO_DEFINE(REG_ADRENO_SP_FS_PVT_MEM_ADDR_REG, + REG_A3XX_SP_FS_PVT_MEM_ADDR_REG), + REG_ADRENO_DEFINE(REG_ADRENO_SP_VS_OBJ_START_REG, + REG_A3XX_SP_VS_OBJ_START_REG), + REG_ADRENO_DEFINE(REG_ADRENO_SP_FS_OBJ_START_REG, + REG_A3XX_SP_FS_OBJ_START_REG), + REG_ADRENO_DEFINE(REG_ADRENO_PA_SC_AA_CONFIG, REG_A3XX_PA_SC_AA_CONFIG), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PM_OVERRIDE2, + REG_A3XX_RBBM_PM_OVERRIDE2), + REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_REG2, REG_AXXX_CP_SCRATCH_REG2), + REG_ADRENO_DEFINE(REG_ADRENO_SQ_GPR_MANAGEMENT, + REG_A3XX_SQ_GPR_MANAGEMENT), + REG_ADRENO_DEFINE(REG_ADRENO_SQ_INST_STORE_MANAGMENT, + REG_A3XX_SQ_INST_STORE_MANAGMENT), + REG_ADRENO_DEFINE(REG_ADRENO_TP0_CHICKEN, REG_A3XX_TP0_CHICKEN), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_RBBM_CTL, REG_A3XX_RBBM_RBBM_CTL), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_SW_RESET_CMD, + REG_A3XX_RBBM_SW_RESET_CMD), + REG_ADRENO_DEFINE(REG_ADRENO_UCHE_INVALIDATE0, + REG_A3XX_UCHE_CACHE_INVALIDATE0_REG), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_LO, + REG_A3XX_RBBM_PERFCTR_LOAD_VALUE_LO), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_HI, + REG_A3XX_RBBM_PERFCTR_LOAD_VALUE_HI), +}; static const struct adreno_gpu_funcs funcs = { .base = { @@ -463,6 +553,7 @@ struct msm_gpu *a3xx_gpu_init(struct drm_device *dev) gpu->num_perfcntrs = ARRAY_SIZE(perfcntrs); adreno_gpu->registers = a3xx_registers; + adreno_gpu->reg_offsets = a3xx_register_offsets; ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs); if (ret) diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c index dc807b2..28e5847 100644 --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c @@ -2,6 +2,8 @@ * Copyright (C) 2013 Red Hat * Author: Rob Clark * + * Copyright (c) 2014 The Linux Foundation. All rights reserved. + * * This program is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 as published by * the Free Software Foundation. @@ -63,19 +65,21 @@ int adreno_hw_init(struct msm_gpu *gpu) } /* Setup REG_CP_RB_CNTL: */ - gpu_write(gpu, REG_AXXX_CP_RB_CNTL, + adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_CNTL, /* size is log2(quad-words): */ AXXX_CP_RB_CNTL_BUFSZ(ilog2(gpu->rb->size / 8)) | AXXX_CP_RB_CNTL_BLKSZ(ilog2(RB_BLKSIZE / 8))); /* Setup ringbuffer address: */ - gpu_write(gpu, REG_AXXX_CP_RB_BASE, gpu->rb_iova); - gpu_write(gpu, REG_AXXX_CP_RB_RPTR_ADDR, rbmemptr(adreno_gpu, rptr)); + adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_BASE, gpu->rb_iova); + adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_RPTR_ADDR, + rbmemptr(adreno_gpu, rptr)); /* Setup scratch/timestamp: */ - gpu_write(gpu, REG_AXXX_SCRATCH_ADDR, rbmemptr(adreno_gpu, fence)); + adreno_gpu_write(adreno_gpu, REG_ADRENO_SCRATCH_ADDR, + rbmemptr(adreno_gpu, fence)); - gpu_write(gpu, REG_AXXX_SCRATCH_UMSK, 0x1); + adreno_gpu_write(adreno_gpu, REG_ADRENO_SCRATCH_UMSK, 0x1); return 0; } @@ -188,12 +192,13 @@ int adreno_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit, void adreno_flush(struct msm_gpu *gpu) { + struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu); uint32_t wptr = get_wptr(gpu->rb); /* ensure writes to ringbuffer have hit system memory: */ mb(); - gpu_write(gpu, REG_AXXX_CP_RB_WPTR, wptr); + adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_WPTR, wptr); } void adreno_idle(struct msm_gpu *gpu) diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h index 52f0515..b9c05e3 100644 --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h @@ -2,6 +2,8 @@ * Copyright (C) 2013 Red Hat * Author: Rob Clark * + * Copyright (c) 2014 The Linux Foundation. All rights reserved. + * * This program is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 as published by * the Free Software Foundation. @@ -25,6 +27,81 @@ #include "adreno_common.xml.h" #include "adreno_pm4.xml.h" +#define REG_ADRENO_DEFINE(_offset, _reg) [_offset] = (_reg) + 1 +/** + * adreno_regs: List of registers that are used in across all + * 3D devices. Each device type has different offset value for the same + * register, so an array of register offsets are declared for every device + * and are indexed by the enumeration values defined in this enum + */ +enum adreno_regs { + REG_ADRENO_CP_DEBUG, + REG_ADRENO_CP_ME_RAM_WADDR, + REG_ADRENO_CP_ME_RAM_DATA, + REG_ADRENO_CP_PFP_UCODE_DATA, + REG_ADRENO_CP_PFP_UCODE_ADDR, + REG_ADRENO_CP_WFI_PEND_CTR, + REG_ADRENO_CP_RB_BASE, + REG_ADRENO_CP_RB_RPTR_ADDR, + REG_ADRENO_CP_RB_RPTR, + REG_ADRENO_CP_RB_WPTR, + REG_ADRENO_CP_PROTECT_CTRL, + REG_ADRENO_CP_ME_CNTL, + REG_ADRENO_CP_RB_CNTL, + REG_ADRENO_CP_IB1_BASE, + REG_ADRENO_CP_IB1_BUFSZ, + REG_ADRENO_CP_IB2_BASE, + REG_ADRENO_CP_IB2_BUFSZ, + REG_ADRENO_CP_TIMESTAMP, + REG_ADRENO_CP_ME_RAM_RADDR, + REG_ADRENO_CP_ROQ_ADDR, + REG_ADRENO_CP_ROQ_DATA, + REG_ADRENO_CP_MERCIU_ADDR, + REG_ADRENO_CP_MERCIU_DATA, + REG_ADRENO_CP_MERCIU_DATA2, + REG_ADRENO_CP_MEQ_ADDR, + REG_ADRENO_CP_MEQ_DATA, + REG_ADRENO_CP_HW_FAULT, + REG_ADRENO_CP_PROTECT_STATUS, + REG_ADRENO_SCRATCH_ADDR, + REG_ADRENO_SCRATCH_UMSK, + REG_ADRENO_SCRATCH_REG2, + REG_ADRENO_RBBM_STATUS, + REG_ADRENO_RBBM_PERFCTR_CTL, + REG_ADRENO_RBBM_PERFCTR_LOAD_CMD0, + REG_ADRENO_RBBM_PERFCTR_LOAD_CMD1, + REG_ADRENO_RBBM_PERFCTR_LOAD_CMD2, + REG_ADRENO_RBBM_PERFCTR_PWR_1_LO, + REG_ADRENO_RBBM_INT_0_MASK, + REG_ADRENO_RBBM_INT_0_STATUS, + REG_ADRENO_RBBM_AHB_ERROR_STATUS, + REG_ADRENO_RBBM_PM_OVERRIDE2, + REG_ADRENO_RBBM_AHB_CMD, + REG_ADRENO_RBBM_INT_CLEAR_CMD, + REG_ADRENO_RBBM_SW_RESET_CMD, + REG_ADRENO_RBBM_CLOCK_CTL, + REG_ADRENO_RBBM_AHB_ME_SPLIT_STATUS, + REG_ADRENO_RBBM_AHB_PFP_SPLIT_STATUS, + REG_ADRENO_VPC_DEBUG_RAM_SEL, + REG_ADRENO_VPC_DEBUG_RAM_READ, + REG_ADRENO_VSC_SIZE_ADDRESS, + REG_ADRENO_VFD_CONTROL_0, + REG_ADRENO_VFD_INDEX_MAX, + REG_ADRENO_SP_VS_PVT_MEM_ADDR_REG, + REG_ADRENO_SP_FS_PVT_MEM_ADDR_REG, + REG_ADRENO_SP_VS_OBJ_START_REG, + REG_ADRENO_SP_FS_OBJ_START_REG, + REG_ADRENO_PA_SC_AA_CONFIG, + REG_ADRENO_SQ_GPR_MANAGEMENT, + REG_ADRENO_SQ_INST_STORE_MANAGMENT, + REG_ADRENO_TP0_CHICKEN, + REG_ADRENO_RBBM_RBBM_CTL, + REG_ADRENO_UCHE_INVALIDATE0, + REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_LO, + REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_HI, + REG_ADRENO_REGISTER_MAX, +}; + struct adreno_rev { uint8_t core; uint8_t major; @@ -76,6 +153,13 @@ struct adreno_gpu { struct adreno_rbmemptrs *memptrs; struct drm_gem_object *memptrs_bo; uint32_t memptrs_iova; + + /* + * Register offsets are different between some GPUs. + * GPU specific offsets will be exported by GPU specific + * code (a3xx_gpu.c) and stored in this common location. + */ + const unsigned int *reg_offsets; }; #define to_adreno_gpu(x) container_of(x, struct adreno_gpu, base) @@ -171,5 +255,37 @@ OUT_PKT3(struct msm_ringbuffer *ring, uint8_t opcode, uint16_t cnt) OUT_RING(ring, CP_TYPE3_PKT | ((cnt-1) << 16) | ((opcode & 0xFF) << 8)); } +/* + * adreno_checkreg_off() - Checks the validity of a register enum + * @gpu: Pointer to struct adreno_gpu + * @offset_name: The register enum that is checked + */ +static inline bool adreno_reg_check(struct adreno_gpu *gpu, + enum adreno_regs offset_name) +{ + if (offset_name >= REG_ADRENO_REGISTER_MAX || + !gpu->reg_offsets[offset_name]) { + BUG(); + } + return true; +} + +static inline u32 adreno_gpu_read(struct adreno_gpu *gpu, + enum adreno_regs offset_name) +{ + u32 reg = gpu->reg_offsets[offset_name]; + u32 val = 0; + if(adreno_reg_check(gpu,offset_name)) + val = gpu_read(&gpu->base, reg - 1); + return val; +} + +static inline void adreno_gpu_write(struct adreno_gpu *gpu, + enum adreno_regs offset_name, u32 data) +{ + u32 reg = gpu->reg_offsets[offset_name]; + if(adreno_reg_check(gpu, offset_name)) + gpu_write(&gpu->base, reg - 1, data); +} #endif /* __ADRENO_GPU_H__ */ -- cgit v0.10.2 From 23bd62fd419755b439152915f4df8ff26346f2b7 Mon Sep 17 00:00:00 2001 From: Aravind Ganesan Date: Mon, 8 Sep 2014 13:40:16 -0600 Subject: drm/msm: a4xx support for msm-drm Added a4xx GPU support. Signed-off-by: Aravind Ganesan Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/Makefile b/drivers/gpu/drm/msm/Makefile index 415a145..0d96132 100644 --- a/drivers/gpu/drm/msm/Makefile +++ b/drivers/gpu/drm/msm/Makefile @@ -7,6 +7,7 @@ msm-y := \ adreno/adreno_device.o \ adreno/adreno_gpu.o \ adreno/a3xx_gpu.o \ + adreno/a4xx_gpu.o \ hdmi/hdmi.o \ hdmi/hdmi_audio.o \ hdmi/hdmi_bridge.o \ diff --git a/drivers/gpu/drm/msm/adreno/a4xx_gpu.c b/drivers/gpu/drm/msm/adreno/a4xx_gpu.c new file mode 100644 index 0000000..9122183 --- /dev/null +++ b/drivers/gpu/drm/msm/adreno/a4xx_gpu.c @@ -0,0 +1,604 @@ +/* Copyright (c) 2014 The Linux Foundation. All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 and + * only version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + */ +#include "a4xx_gpu.h" +#ifdef CONFIG_MSM_OCMEM +# include +#endif + +#define A4XX_INT0_MASK \ + (A4XX_INT0_RBBM_AHB_ERROR | \ + A4XX_INT0_RBBM_ATB_BUS_OVERFLOW | \ + A4XX_INT0_CP_T0_PACKET_IN_IB | \ + A4XX_INT0_CP_OPCODE_ERROR | \ + A4XX_INT0_CP_RESERVED_BIT_ERROR | \ + A4XX_INT0_CP_HW_FAULT | \ + A4XX_INT0_CP_IB1_INT | \ + A4XX_INT0_CP_IB2_INT | \ + A4XX_INT0_CP_RB_INT | \ + A4XX_INT0_CP_REG_PROTECT_FAULT | \ + A4XX_INT0_CP_AHB_ERROR_HALT | \ + A4XX_INT0_UCHE_OOB_ACCESS) + +extern bool hang_debug; +static void a4xx_dump(struct msm_gpu *gpu); + +/* + * a4xx_enable_hwcg() - Program the clock control registers + * @device: The adreno device pointer + */ +static void a4xx_enable_hwcg(struct msm_gpu *gpu) +{ + struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu); + unsigned int i; + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_TP(i), 0x02222202); + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_TP(i), 0x00002222); + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_TP(i), 0x0E739CE7); + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_TP(i), 0x00111111); + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_SP(i), 0x22222222); + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_SP(i), 0x00222222); + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_SP(i), 0x00000104); + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_SP(i), 0x00000081); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_UCHE, 0x22222222); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_UCHE, 0x02222222); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL3_UCHE, 0x00000000); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL4_UCHE, 0x00000000); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_UCHE, 0x00004444); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_UCHE, 0x00001112); + for (i = 0; i < 4; i++) + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_RB(i), 0x22222222); + + /* Disable L1 clocking in A420 due to CCU issues with it */ + for (i = 0; i < 4; i++) { + if (adreno_is_a420(adreno_gpu)) { + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_RB(i), + 0x00002020); + } else { + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_RB(i), + 0x00022020); + } + } + + for (i = 0; i < 4; i++) { + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_MARB_CCU(i), + 0x00000922); + } + + for (i = 0; i < 4; i++) { + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_RB_MARB_CCU(i), + 0x00000000); + } + + for (i = 0; i < 4; i++) { + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_RB_MARB_CCU_L1(i), + 0x00000001); + } + + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_MODE_GPC, 0x02222222); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_GPC, 0x04100104); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_GPC, 0x00022222); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_COM_DCOM, 0x00000022); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_COM_DCOM, 0x0000010F); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_COM_DCOM, 0x00000022); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_TSE_RAS_RBBM, 0x00222222); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_TSE_RAS_RBBM, 0x00004104); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_TSE_RAS_RBBM, 0x00000222); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_HLSQ , 0x00000000); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_HLSQ, 0x00000000); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ, 0x00020000); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL, 0xAAAAAAAA); + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2, 0); +} + +static void a4xx_me_init(struct msm_gpu *gpu) +{ + struct msm_ringbuffer *ring = gpu->rb; + + OUT_PKT3(ring, CP_ME_INIT, 17); + OUT_RING(ring, 0x000003f7); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000080); + OUT_RING(ring, 0x00000100); + OUT_RING(ring, 0x00000180); + OUT_RING(ring, 0x00006600); + OUT_RING(ring, 0x00000150); + OUT_RING(ring, 0x0000014e); + OUT_RING(ring, 0x00000154); + OUT_RING(ring, 0x00000001); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + gpu->funcs->flush(gpu); + gpu->funcs->idle(gpu); +} + +static int a4xx_hw_init(struct msm_gpu *gpu) +{ + struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu); + struct a4xx_gpu *a4xx_gpu = to_a4xx_gpu(adreno_gpu); + uint32_t *ptr, len; + int i, ret; + + if (adreno_is_a4xx(adreno_gpu)) { + gpu_write(gpu, REG_A4XX_VBIF_ABIT_SORT, 0x0001001F); + gpu_write(gpu, REG_A4XX_VBIF_ABIT_SORT_CONF, 0x000000A4); + gpu_write(gpu, REG_A4XX_VBIF_GATE_OFF_WRREQ_EN, 0x00000001); + gpu_write(gpu, REG_A4XX_VBIF_IN_RD_LIM_CONF0, 0x18181818); + gpu_write(gpu, REG_A4XX_VBIF_IN_RD_LIM_CONF1, 0x00000018); + gpu_write(gpu, REG_A4XX_VBIF_IN_WR_LIM_CONF0, 0x18181818); + gpu_write(gpu, REG_A4XX_VBIF_IN_WR_LIM_CONF1, 0x00000018); + gpu_write(gpu, REG_A4XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x00000003); + } else { + BUG(); + } + + /* Make all blocks contribute to the GPU BUSY perf counter */ + gpu_write(gpu, REG_A4XX_RBBM_GPU_BUSY_MASKED, 0xffffffff); + + /* Tune the hystersis counters for SP and CP idle detection */ + gpu_write(gpu, REG_A4XX_RBBM_SP_HYST_CNT, 0x10); + gpu_write(gpu, REG_A4XX_RBBM_WAIT_IDLE_CLOCKS_CTL, 0x10); + + /* Enable the RBBM error reporting bits */ + gpu_write(gpu, REG_A4XX_RBBM_AHB_CTL0, 0x00000001); + + /* Enable AHB error reporting*/ + gpu_write(gpu, REG_A4XX_RBBM_AHB_CTL1, 0xa6ffffff); + + /* Enable power counters*/ + gpu_write(gpu, REG_A4XX_RBBM_RBBM_CTL, 0x00000030); + + /* + * Turn on hang detection - this spews a lot of useful information + * into the RBBM registers on a hang: + */ + gpu_write(gpu, REG_A4XX_RBBM_INTERFACE_HANG_INT_CTL, + (1 << 30) | 0xFFFF); + + gpu_write(gpu, REG_A4XX_RB_GMEM_BASE_ADDR, + (unsigned int)(a4xx_gpu->ocmem_base >> 14)); + + /* Turn on performance counters: */ + gpu_write(gpu, REG_A4XX_RBBM_PERFCTR_CTL, 0x01); + + /* Disable L2 bypass to avoid UCHE out of bounds errors */ + gpu_write(gpu, REG_A4XX_UCHE_TRAP_BASE_LO, 0xffff0000); + gpu_write(gpu, REG_A4XX_UCHE_TRAP_BASE_HI, 0xffff0000); + + gpu_write(gpu, REG_A4XX_CP_DEBUG, (1 << 25) | + (adreno_is_a420(adreno_gpu) ? (1 << 29) : 0)); + + a4xx_enable_hwcg(gpu); + + /* + * For A420 set RBBM_CLOCK_DELAY_HLSQ.CGC_HLSQ_TP_EARLY_CYC >= 2 + * due to timing issue with HLSQ_TP_CLK_EN + */ + if (adreno_is_a420(adreno_gpu)) { + unsigned int val; + val = gpu_read(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ); + val &= ~A4XX_CGC_HLSQ_EARLY_CYC__MASK; + val |= 2 << A4XX_CGC_HLSQ_EARLY_CYC__SHIFT; + gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ, val); + } + + ret = adreno_hw_init(gpu); + if (ret) + return ret; + + /* setup access protection: */ + gpu_write(gpu, REG_A4XX_CP_PROTECT_CTRL, 0x00000007); + + /* RBBM registers */ + gpu_write(gpu, REG_A4XX_CP_PROTECT(0), 0x62000010); + gpu_write(gpu, REG_A4XX_CP_PROTECT(1), 0x63000020); + gpu_write(gpu, REG_A4XX_CP_PROTECT(2), 0x64000040); + gpu_write(gpu, REG_A4XX_CP_PROTECT(3), 0x65000080); + gpu_write(gpu, REG_A4XX_CP_PROTECT(4), 0x66000100); + gpu_write(gpu, REG_A4XX_CP_PROTECT(5), 0x64000200); + + /* CP registers */ + gpu_write(gpu, REG_A4XX_CP_PROTECT(6), 0x67000800); + gpu_write(gpu, REG_A4XX_CP_PROTECT(7), 0x64001600); + + + /* RB registers */ + gpu_write(gpu, REG_A4XX_CP_PROTECT(8), 0x60003300); + + /* HLSQ registers */ + gpu_write(gpu, REG_A4XX_CP_PROTECT(9), 0x60003800); + + /* VPC registers */ + gpu_write(gpu, REG_A4XX_CP_PROTECT(10), 0x61003980); + + /* SMMU registers */ + gpu_write(gpu, REG_A4XX_CP_PROTECT(11), 0x6e010000); + + gpu_write(gpu, REG_A4XX_RBBM_INT_0_MASK, A4XX_INT0_MASK); + + ret = adreno_hw_init(gpu); + if (ret) + return ret; + + /* Load PM4: */ + ptr = (uint32_t *)(adreno_gpu->pm4->data); + len = adreno_gpu->pm4->size / 4; + DBG("loading PM4 ucode version: %u", ptr[0]); + gpu_write(gpu, REG_A4XX_CP_ME_RAM_WADDR, 0); + for (i = 1; i < len; i++) + gpu_write(gpu, REG_A4XX_CP_ME_RAM_DATA, ptr[i]); + + /* Load PFP: */ + ptr = (uint32_t *)(adreno_gpu->pfp->data); + len = adreno_gpu->pfp->size / 4; + DBG("loading PFP ucode version: %u", ptr[0]); + + gpu_write(gpu, REG_A4XX_CP_PFP_UCODE_ADDR, 0); + for (i = 1; i < len; i++) + gpu_write(gpu, REG_A4XX_CP_PFP_UCODE_DATA, ptr[i]); + + /* clear ME_HALT to start micro engine */ + gpu_write(gpu, REG_A4XX_CP_ME_CNTL, 0); + + a4xx_me_init(gpu); + return 0; +} + +static void a4xx_recover(struct msm_gpu *gpu) +{ + /* dump registers before resetting gpu, if enabled: */ + if (hang_debug) + a4xx_dump(gpu); + + gpu_write(gpu, REG_A4XX_RBBM_SW_RESET_CMD, 1); + gpu_read(gpu, REG_A4XX_RBBM_SW_RESET_CMD); + gpu_write(gpu, REG_A4XX_RBBM_SW_RESET_CMD, 0); + adreno_recover(gpu); +} + +static void a4xx_destroy(struct msm_gpu *gpu) +{ + struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu); + struct a4xx_gpu *a4xx_gpu = to_a4xx_gpu(adreno_gpu); + + DBG("%s", gpu->name); + + adreno_gpu_cleanup(adreno_gpu); + +#ifdef CONFIG_MSM_OCMEM + if (a4xx_gpu->ocmem_base) + ocmem_free(OCMEM_GRAPHICS, a4xx_gpu->ocmem_hdl); +#endif + + kfree(a4xx_gpu); +} + +static void a4xx_idle(struct msm_gpu *gpu) +{ + /* wait for ringbuffer to drain: */ + adreno_idle(gpu); + + /* then wait for GPU to finish: */ + if (spin_until(!(gpu_read(gpu, REG_A4XX_RBBM_STATUS) & + A4XX_RBBM_STATUS_GPU_BUSY))) + DRM_ERROR("%s: timeout waiting for GPU to idle!\n", gpu->name); + + /* TODO maybe we need to reset GPU here to recover from hang? */ +} + +static irqreturn_t a4xx_irq(struct msm_gpu *gpu) +{ + uint32_t status; + + status = gpu_read(gpu, REG_A4XX_RBBM_INT_0_STATUS); + DBG("%s: Int status %08x", gpu->name, status); + + gpu_write(gpu, REG_A4XX_RBBM_INT_CLEAR_CMD, status); + + msm_gpu_retire(gpu); + + return IRQ_HANDLED; +} + +static const unsigned int a4xx_registers[] = { + /* RBBM */ + 0x0000, 0x0002, 0x0004, 0x0021, 0x0023, 0x0024, 0x0026, 0x0026, + 0x0028, 0x002B, 0x002E, 0x0034, 0x0037, 0x0044, 0x0047, 0x0066, + 0x0068, 0x0095, 0x009C, 0x0170, 0x0174, 0x01AF, + /* CP */ + 0x0200, 0x0233, 0x0240, 0x0250, 0x04C0, 0x04DD, 0x0500, 0x050B, + 0x0578, 0x058F, + /* VSC */ + 0x0C00, 0x0C03, 0x0C08, 0x0C41, 0x0C50, 0x0C51, + /* GRAS */ + 0x0C80, 0x0C81, 0x0C88, 0x0C8F, + /* RB */ + 0x0CC0, 0x0CC0, 0x0CC4, 0x0CD2, + /* PC */ + 0x0D00, 0x0D0C, 0x0D10, 0x0D17, 0x0D20, 0x0D23, + /* VFD */ + 0x0E40, 0x0E4A, + /* VPC */ + 0x0E60, 0x0E61, 0x0E63, 0x0E68, + /* UCHE */ + 0x0E80, 0x0E84, 0x0E88, 0x0E95, + /* VMIDMT */ + 0x1000, 0x1000, 0x1002, 0x1002, 0x1004, 0x1004, 0x1008, 0x100A, + 0x100C, 0x100D, 0x100F, 0x1010, 0x1012, 0x1016, 0x1024, 0x1024, + 0x1027, 0x1027, 0x1100, 0x1100, 0x1102, 0x1102, 0x1104, 0x1104, + 0x1110, 0x1110, 0x1112, 0x1116, 0x1124, 0x1124, 0x1300, 0x1300, + 0x1380, 0x1380, + /* GRAS CTX 0 */ + 0x2000, 0x2004, 0x2008, 0x2067, 0x2070, 0x2078, 0x207B, 0x216E, + /* PC CTX 0 */ + 0x21C0, 0x21C6, 0x21D0, 0x21D0, 0x21D9, 0x21D9, 0x21E5, 0x21E7, + /* VFD CTX 0 */ + 0x2200, 0x2204, 0x2208, 0x22A9, + /* GRAS CTX 1 */ + 0x2400, 0x2404, 0x2408, 0x2467, 0x2470, 0x2478, 0x247B, 0x256E, + /* PC CTX 1 */ + 0x25C0, 0x25C6, 0x25D0, 0x25D0, 0x25D9, 0x25D9, 0x25E5, 0x25E7, + /* VFD CTX 1 */ + 0x2600, 0x2604, 0x2608, 0x26A9, + /* XPU */ + 0x2C00, 0x2C01, 0x2C10, 0x2C10, 0x2C12, 0x2C16, 0x2C1D, 0x2C20, + 0x2C28, 0x2C28, 0x2C30, 0x2C30, 0x2C32, 0x2C36, 0x2C40, 0x2C40, + 0x2C50, 0x2C50, 0x2C52, 0x2C56, 0x2C80, 0x2C80, 0x2C94, 0x2C95, + /* VBIF */ + 0x3000, 0x3007, 0x300C, 0x3014, 0x3018, 0x301D, 0x3020, 0x3022, + 0x3024, 0x3026, 0x3028, 0x302A, 0x302C, 0x302D, 0x3030, 0x3031, + 0x3034, 0x3036, 0x3038, 0x3038, 0x303C, 0x303D, 0x3040, 0x3040, + 0x3049, 0x3049, 0x3058, 0x3058, 0x305B, 0x3061, 0x3064, 0x3068, + 0x306C, 0x306D, 0x3080, 0x3088, 0x308B, 0x308C, 0x3090, 0x3094, + 0x3098, 0x3098, 0x309C, 0x309C, 0x30C0, 0x30C0, 0x30C8, 0x30C8, + 0x30D0, 0x30D0, 0x30D8, 0x30D8, 0x30E0, 0x30E0, 0x3100, 0x3100, + 0x3108, 0x3108, 0x3110, 0x3110, 0x3118, 0x3118, 0x3120, 0x3120, + 0x3124, 0x3125, 0x3129, 0x3129, 0x3131, 0x3131, 0x330C, 0x330C, + 0x3310, 0x3310, 0x3400, 0x3401, 0x3410, 0x3410, 0x3412, 0x3416, + 0x341D, 0x3420, 0x3428, 0x3428, 0x3430, 0x3430, 0x3432, 0x3436, + 0x3440, 0x3440, 0x3450, 0x3450, 0x3452, 0x3456, 0x3480, 0x3480, + 0x3494, 0x3495, 0x4000, 0x4000, 0x4002, 0x4002, 0x4004, 0x4004, + 0x4008, 0x400A, 0x400C, 0x400D, 0x400F, 0x4012, 0x4014, 0x4016, + 0x401D, 0x401D, 0x4020, 0x4027, 0x4060, 0x4062, 0x4200, 0x4200, + 0x4300, 0x4300, 0x4400, 0x4400, 0x4500, 0x4500, 0x4800, 0x4802, + 0x480F, 0x480F, 0x4811, 0x4811, 0x4813, 0x4813, 0x4815, 0x4816, + 0x482B, 0x482B, 0x4857, 0x4857, 0x4883, 0x4883, 0x48AF, 0x48AF, + 0x48C5, 0x48C5, 0x48E5, 0x48E5, 0x4905, 0x4905, 0x4925, 0x4925, + 0x4945, 0x4945, 0x4950, 0x4950, 0x495B, 0x495B, 0x4980, 0x498E, + 0x4B00, 0x4B00, 0x4C00, 0x4C00, 0x4D00, 0x4D00, 0x4E00, 0x4E00, + 0x4E80, 0x4E80, 0x4F00, 0x4F00, 0x4F08, 0x4F08, 0x4F10, 0x4F10, + 0x4F18, 0x4F18, 0x4F20, 0x4F20, 0x4F30, 0x4F30, 0x4F60, 0x4F60, + 0x4F80, 0x4F81, 0x4F88, 0x4F89, 0x4FEE, 0x4FEE, 0x4FF3, 0x4FF3, + 0x6000, 0x6001, 0x6008, 0x600F, 0x6014, 0x6016, 0x6018, 0x601B, + 0x61FD, 0x61FD, 0x623C, 0x623C, 0x6380, 0x6380, 0x63A0, 0x63A0, + 0x63C0, 0x63C1, 0x63C8, 0x63C9, 0x63D0, 0x63D4, 0x63D6, 0x63D6, + 0x63EE, 0x63EE, 0x6400, 0x6401, 0x6408, 0x640F, 0x6414, 0x6416, + 0x6418, 0x641B, 0x65FD, 0x65FD, 0x663C, 0x663C, 0x6780, 0x6780, + 0x67A0, 0x67A0, 0x67C0, 0x67C1, 0x67C8, 0x67C9, 0x67D0, 0x67D4, + 0x67D6, 0x67D6, 0x67EE, 0x67EE, 0x6800, 0x6801, 0x6808, 0x680F, + 0x6814, 0x6816, 0x6818, 0x681B, 0x69FD, 0x69FD, 0x6A3C, 0x6A3C, + 0x6B80, 0x6B80, 0x6BA0, 0x6BA0, 0x6BC0, 0x6BC1, 0x6BC8, 0x6BC9, + 0x6BD0, 0x6BD4, 0x6BD6, 0x6BD6, 0x6BEE, 0x6BEE, + ~0 /* sentinel */ +}; + +#ifdef CONFIG_DEBUG_FS +static void a4xx_show(struct msm_gpu *gpu, struct seq_file *m) +{ + gpu->funcs->pm_resume(gpu); + + seq_printf(m, "status: %08x\n", + gpu_read(gpu, REG_A4XX_RBBM_STATUS)); + gpu->funcs->pm_suspend(gpu); + + adreno_show(gpu, m); + +} +#endif + +/* Register offset defines for A4XX, in order of enum adreno_regs */ +static const unsigned int a4xx_register_offsets[REG_ADRENO_REGISTER_MAX] = { + REG_ADRENO_DEFINE(REG_ADRENO_CP_DEBUG, REG_A4XX_CP_DEBUG), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_WADDR, REG_A4XX_CP_ME_RAM_WADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_DATA, REG_A4XX_CP_ME_RAM_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_PFP_UCODE_DATA, + REG_A4XX_CP_PFP_UCODE_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_PFP_UCODE_ADDR, + REG_A4XX_CP_PFP_UCODE_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_WFI_PEND_CTR, REG_A4XX_CP_WFI_PEND_CTR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_BASE, REG_A4XX_CP_RB_BASE), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_RPTR_ADDR, REG_A4XX_CP_RB_RPTR_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_RPTR, REG_A4XX_CP_RB_RPTR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_WPTR, REG_A4XX_CP_RB_WPTR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_PROTECT_CTRL, REG_A4XX_CP_PROTECT_CTRL), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_CNTL, REG_A4XX_CP_ME_CNTL), + REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_CNTL, REG_A4XX_CP_RB_CNTL), + REG_ADRENO_DEFINE(REG_ADRENO_CP_IB1_BASE, REG_A4XX_CP_IB1_BASE), + REG_ADRENO_DEFINE(REG_ADRENO_CP_IB1_BUFSZ, REG_A4XX_CP_IB1_BUFSZ), + REG_ADRENO_DEFINE(REG_ADRENO_CP_IB2_BASE, REG_A4XX_CP_IB2_BASE), + REG_ADRENO_DEFINE(REG_ADRENO_CP_IB2_BUFSZ, REG_A4XX_CP_IB2_BUFSZ), + REG_ADRENO_DEFINE(REG_ADRENO_CP_TIMESTAMP, REG_AXXX_CP_SCRATCH_REG0), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_RADDR, REG_A4XX_CP_ME_RAM_RADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ROQ_ADDR, REG_A4XX_CP_ROQ_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_ROQ_DATA, REG_A4XX_CP_ROQ_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_ADDR, REG_A4XX_CP_MERCIU_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_DATA, REG_A4XX_CP_MERCIU_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_DATA2, REG_A4XX_CP_MERCIU_DATA2), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MEQ_ADDR, REG_A4XX_CP_MEQ_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_CP_MEQ_DATA, REG_A4XX_CP_MEQ_DATA), + REG_ADRENO_DEFINE(REG_ADRENO_CP_HW_FAULT, REG_A4XX_CP_HW_FAULT), + REG_ADRENO_DEFINE(REG_ADRENO_CP_PROTECT_STATUS, + REG_A4XX_CP_PROTECT_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_ADDR, REG_A4XX_CP_SCRATCH_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_UMSK, REG_A4XX_CP_SCRATCH_UMASK), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_STATUS, REG_A4XX_RBBM_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_CTL, + REG_A4XX_RBBM_PERFCTR_CTL), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD0, + REG_A4XX_RBBM_PERFCTR_LOAD_CMD0), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD1, + REG_A4XX_RBBM_PERFCTR_LOAD_CMD1), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD2, + REG_A4XX_RBBM_PERFCTR_LOAD_CMD2), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_PWR_1_LO, + REG_A4XX_RBBM_PERFCTR_PWR_1_LO), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_0_MASK, REG_A4XX_RBBM_INT_0_MASK), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_0_STATUS, + REG_A4XX_RBBM_INT_0_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_ERROR_STATUS, + REG_A4XX_RBBM_AHB_ERROR_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_CMD, REG_A4XX_RBBM_AHB_CMD), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_CLOCK_CTL, REG_A4XX_RBBM_CLOCK_CTL), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_ME_SPLIT_STATUS, + REG_A4XX_RBBM_AHB_ME_SPLIT_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_PFP_SPLIT_STATUS, + REG_A4XX_RBBM_AHB_PFP_SPLIT_STATUS), + REG_ADRENO_DEFINE(REG_ADRENO_VPC_DEBUG_RAM_SEL, + REG_A4XX_VPC_DEBUG_RAM_SEL), + REG_ADRENO_DEFINE(REG_ADRENO_VPC_DEBUG_RAM_READ, + REG_A4XX_VPC_DEBUG_RAM_READ), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_CLEAR_CMD, + REG_A4XX_RBBM_INT_CLEAR_CMD), + REG_ADRENO_DEFINE(REG_ADRENO_VSC_SIZE_ADDRESS, + REG_A4XX_VSC_SIZE_ADDRESS), + REG_ADRENO_DEFINE(REG_ADRENO_VFD_CONTROL_0, REG_A4XX_VFD_CONTROL_0), + REG_ADRENO_DEFINE(REG_ADRENO_SP_VS_PVT_MEM_ADDR_REG, + REG_A4XX_SP_VS_PVT_MEM_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_SP_FS_PVT_MEM_ADDR_REG, + REG_A4XX_SP_FS_PVT_MEM_ADDR), + REG_ADRENO_DEFINE(REG_ADRENO_SP_VS_OBJ_START_REG, + REG_A4XX_SP_VS_OBJ_START), + REG_ADRENO_DEFINE(REG_ADRENO_SP_FS_OBJ_START_REG, + REG_A4XX_SP_FS_OBJ_START), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_RBBM_CTL, REG_A4XX_RBBM_RBBM_CTL), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_SW_RESET_CMD, + REG_A4XX_RBBM_SW_RESET_CMD), + REG_ADRENO_DEFINE(REG_ADRENO_UCHE_INVALIDATE0, + REG_A4XX_UCHE_INVALIDATE0), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_LO, + REG_A4XX_RBBM_PERFCTR_LOAD_VALUE_LO), + REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_HI, + REG_A4XX_RBBM_PERFCTR_LOAD_VALUE_HI), +}; + +static void a4xx_dump(struct msm_gpu *gpu) +{ + adreno_dump(gpu); + printk("status: %08x\n", + gpu_read(gpu, REG_A4XX_RBBM_STATUS)); + adreno_dump(gpu); +} + +static const struct adreno_gpu_funcs funcs = { + .base = { + .get_param = adreno_get_param, + .hw_init = a4xx_hw_init, + .pm_suspend = msm_gpu_pm_suspend, + .pm_resume = msm_gpu_pm_resume, + .recover = a4xx_recover, + .last_fence = adreno_last_fence, + .submit = adreno_submit, + .flush = adreno_flush, + .idle = a4xx_idle, + .irq = a4xx_irq, + .destroy = a4xx_destroy, +#ifdef CONFIG_DEBUG_FS + .show = a4xx_show, +#endif + }, +}; + +struct msm_gpu *a4xx_gpu_init(struct drm_device *dev) +{ + struct a4xx_gpu *a4xx_gpu = NULL; + struct adreno_gpu *adreno_gpu; + struct msm_gpu *gpu; + struct msm_drm_private *priv = dev->dev_private; + struct platform_device *pdev = priv->gpu_pdev; + int ret; + + if (!pdev) { + dev_err(dev->dev, "no a4xx device\n"); + ret = -ENXIO; + goto fail; + } + + a4xx_gpu = kzalloc(sizeof(*a4xx_gpu), GFP_KERNEL); + if (!a4xx_gpu) { + ret = -ENOMEM; + goto fail; + } + + adreno_gpu = &a4xx_gpu->base; + gpu = &adreno_gpu->base; + + a4xx_gpu->pdev = pdev; + + gpu->perfcntrs = NULL; + gpu->num_perfcntrs = 0; + + adreno_gpu->registers = a4xx_registers; + adreno_gpu->reg_offsets = a4xx_register_offsets; + + ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs); + if (ret) + goto fail; + + /* if needed, allocate gmem: */ + if (adreno_is_a4xx(adreno_gpu)) { +#ifdef CONFIG_MSM_OCMEM + /* TODO this is different/missing upstream: */ + struct ocmem_buf *ocmem_hdl = + ocmem_allocate(OCMEM_GRAPHICS, adreno_gpu->gmem); + + a4xx_gpu->ocmem_hdl = ocmem_hdl; + a4xx_gpu->ocmem_base = ocmem_hdl->addr; + adreno_gpu->gmem = ocmem_hdl->len; + DBG("using %dK of OCMEM at 0x%08x", adreno_gpu->gmem / 1024, + a4xx_gpu->ocmem_base); +#endif + } + + if (!gpu->mmu) { + /* TODO we think it is possible to configure the GPU to + * restrict access to VRAM carveout. But the required + * registers are unknown. For now just bail out and + * limp along with just modesetting. If it turns out + * to not be possible to restrict access, then we must + * implement a cmdstream validator. + */ + dev_err(dev->dev, "No memory protection without IOMMU\n"); + ret = -ENXIO; + goto fail; + } + + return gpu; + +fail: + if (a4xx_gpu) + a4xx_destroy(&a4xx_gpu->base.base); + + return ERR_PTR(ret); +} diff --git a/drivers/gpu/drm/msm/adreno/a4xx_gpu.h b/drivers/gpu/drm/msm/adreno/a4xx_gpu.h new file mode 100644 index 0000000..0124720 --- /dev/null +++ b/drivers/gpu/drm/msm/adreno/a4xx_gpu.h @@ -0,0 +1,34 @@ +/* Copyright (c) 2014 The Linux Foundation. All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 and + * only version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + */ +#ifndef __A4XX_GPU_H__ +#define __A4XX_GPU_H__ + +#include "adreno_gpu.h" + +/* arrg, somehow fb.h is getting pulled in: */ +#undef ROP_COPY +#undef ROP_XOR + +#include "a4xx.xml.h" + +struct a4xx_gpu { + struct adreno_gpu base; + struct platform_device *pdev; + + /* if OCMEM is used for GMEM: */ + uint32_t ocmem_base; + void *ocmem_hdl; +}; +#define to_a4xx_gpu(x) container_of(x, struct a4xx_gpu, base) + +#endif /* __A4XX_GPU_H__ */ diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c index 7ab85af..be83dee 100644 --- a/drivers/gpu/drm/msm/adreno/adreno_device.c +++ b/drivers/gpu/drm/msm/adreno/adreno_device.c @@ -2,6 +2,8 @@ * Copyright (C) 2013-2014 Red Hat * Author: Rob Clark * + * Copyright (c) 2014 The Linux Foundation. All rights reserved. + * * This program is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 as published by * the Free Software Foundation. @@ -28,6 +30,7 @@ MODULE_PARM_DESC(hang_debug, "Dump registers when hang is detected (can be slow! module_param_named(hang_debug, hang_debug, bool, 0600); struct msm_gpu *a3xx_gpu_init(struct drm_device *dev); +struct msm_gpu *a4xx_gpu_init(struct drm_device *dev); static const struct adreno_info gpulist[] = { { @@ -54,6 +57,14 @@ static const struct adreno_info gpulist[] = { .pfpfw = "a330_pfp.fw", .gmem = SZ_1M, .init = a3xx_gpu_init, + }, { + .rev = ADRENO_REV(4, 2, 0, ANY_ID), + .revn = 420, + .name = "A420", + .pm4fw = "a420_pm4.fw", + .pfpfw = "a420_pfp.fw", + .gmem = (SZ_1M + SZ_512K), + .init = a4xx_gpu_init, }, }; @@ -61,6 +72,8 @@ MODULE_FIRMWARE("a300_pm4.fw"); MODULE_FIRMWARE("a300_pfp.fw"); MODULE_FIRMWARE("a330_pm4.fw"); MODULE_FIRMWARE("a330_pfp.fw"); +MODULE_FIRMWARE("a420_pm4.fw"); +MODULE_FIRMWARE("a420_pfp.fw"); static inline bool _rev_match(uint8_t entry, uint8_t id) { diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c index 28e5847..aa87304 100644 --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c @@ -155,7 +155,7 @@ int adreno_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit, OUT_PKT0(ring, REG_AXXX_CP_SCRATCH_REG2, 1); OUT_RING(ring, submit->fence); - if (adreno_is_a3xx(adreno_gpu)) { + if (adreno_is_a3xx(adreno_gpu) || adreno_is_a4xx(adreno_gpu)) { /* Flush HLSQ lazy updates to make sure there is nothing * pending for indirect loads after the timestamp has * passed: diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h index b9c05e3..a0cc309 100644 --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h @@ -212,6 +212,16 @@ static inline bool adreno_is_a330v2(struct adreno_gpu *gpu) return adreno_is_a330(gpu) && (gpu->rev.patchid > 0); } +static inline bool adreno_is_a4xx(struct adreno_gpu *gpu) +{ + return (gpu->revn >= 400) && (gpu->revn < 500); +} + +static inline int adreno_is_a420(struct adreno_gpu *gpu) +{ + return gpu->revn == 420; +} + int adreno_get_param(struct msm_gpu *gpu, uint32_t param, uint64_t *value); int adreno_hw_init(struct msm_gpu *gpu); uint32_t adreno_last_fence(struct msm_gpu *gpu); -- cgit v0.10.2 From afd6275dd12381a7bd3750c125b3b68b5f8b03ca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 30 Oct 2014 19:43:03 +0200 Subject: drm/i915: Reinit display irqs and hpd from chv pipe-a power well MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit On chv the pipe-a power well is the new disp2d well, and it kills pretty much everything in the display block. So we need to do the the same dance that vlv does wrt. display irqs and hpd when the power well goes up or down. Signed-off-by: Ville Syrjälä Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index dcbecff..f5a78d5 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -577,6 +577,23 @@ static void chv_pipe_power_well_enable(struct drm_i915_private *dev_priv, power_well->data != PIPE_C); chv_set_pipe_power_well(dev_priv, power_well, true); + + if (power_well->data == PIPE_A) { + spin_lock_irq(&dev_priv->irq_lock); + valleyview_enable_display_irqs(dev_priv); + spin_unlock_irq(&dev_priv->irq_lock); + + /* + * During driver initialization/resume we can avoid restoring the + * part of the HW/SW state that will be inited anyway explicitly. + */ + if (dev_priv->power_domains.initializing) + return; + + intel_hpd_init(dev_priv); + + i915_redisable_vga_power_on(dev_priv->dev); + } } static void chv_pipe_power_well_disable(struct drm_i915_private *dev_priv, @@ -586,6 +603,12 @@ static void chv_pipe_power_well_disable(struct drm_i915_private *dev_priv, power_well->data != PIPE_B && power_well->data != PIPE_C); + if (power_well->data == PIPE_A) { + spin_lock_irq(&dev_priv->irq_lock); + valleyview_disable_display_irqs(dev_priv); + spin_unlock_irq(&dev_priv->irq_lock); + } + chv_set_pipe_power_well(dev_priv, power_well, false); if (power_well->data == PIPE_A) -- cgit v0.10.2 From 9a3b9c7a645d2781ead28360dec7ceabef0daa74 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 7 Nov 2014 21:33:42 +0200 Subject: drm/i915: Silence valleyview_set_rps() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Even with the rps debug messages signficantly recuced by commit 67956867aa07c59d6d83628bbc9ee4bd9799a1e1 Author: Ville Syrjälä Date: Tue Sep 2 15:12:17 2014 +0300 drm/i915: Don't spam dmesg with rps messages on vlv/chv we still get an inordinate amount of spam from this. Just kill the debug print. If someone wants to observe it they can just use the tracepoint. Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 87bff16..fcbe335 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4504,14 +4504,8 @@ void valleyview_set_rps(struct drm_device *dev, u8 val) "Odd GPU freq value\n")) val &= ~1; - if (val != dev_priv->rps.cur_freq) { - DRM_DEBUG_DRIVER("GPU freq request from %d MHz (%u) to %d MHz (%u)\n", - vlv_gpu_freq(dev_priv, dev_priv->rps.cur_freq), - dev_priv->rps.cur_freq, - vlv_gpu_freq(dev_priv, val), val); - + if (val != dev_priv->rps.cur_freq) vlv_punit_write(dev_priv, PUNIT_REG_GPU_FREQ_REQ, val); - } I915_WRITE(GEN6_PMINTRMSK, gen6_rps_pm_mask(dev_priv, val)); -- cgit v0.10.2 From c8e9627d2ad9efed125bcee8587ea27889fb06d2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 7 Nov 2014 21:33:44 +0200 Subject: drm/i915: Add a name for the Punit GPLLENABLE bit MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Remove the magic number for the GPLLENABLE bit by adding a name for it. Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index a143127..faf61f5 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -576,6 +576,7 @@ enum punit_power_well { #define PUNIT_REG_GPU_LFM 0xd3 #define PUNIT_REG_GPU_FREQ_REQ 0xd4 #define PUNIT_REG_GPU_FREQ_STS 0xd8 +#define GPLLENABLE (1<<4) #define GENFREQSTATUS (1<<0) #define PUNIT_REG_MEDIA_TURBO_FREQ_REQ 0xdc #define PUNIT_REG_CZ_TIMESTAMP 0xce diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index fcbe335..1f2441f 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5333,7 +5333,7 @@ static void cherryview_enable_rps(struct drm_device *dev) val = vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS); - DRM_DEBUG_DRIVER("GPLL enabled? %s\n", val & 0x10 ? "yes" : "no"); + DRM_DEBUG_DRIVER("GPLL enabled? %s\n", val & GPLLENABLE ? "yes" : "no"); DRM_DEBUG_DRIVER("GPU status: 0x%08x\n", val); dev_priv->rps.cur_freq = (val >> 8) & 0xff; @@ -5413,7 +5413,7 @@ static void valleyview_enable_rps(struct drm_device *dev) val = vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS); - DRM_DEBUG_DRIVER("GPLL enabled? %s\n", val & 0x10 ? "yes" : "no"); + DRM_DEBUG_DRIVER("GPLL enabled? %s\n", val & GPLLENABLE ? "yes" : "no"); DRM_DEBUG_DRIVER("GPU status: 0x%08x\n", val); dev_priv->rps.cur_freq = (val >> 8) & 0xff; -- cgit v0.10.2 From 8d40c3ae51483059733187ba8bc6e031764fafec Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 7 Nov 2014 21:33:45 +0200 Subject: drm/i915: Warn if GPLL isn't used on vlv/chv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Our freq<->opcode conversions assume that GPLL is always used. Apparently that should be the case always, but let's scream if we ever encounter something different. Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 1f2441f..cbf84cc 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5333,6 +5333,9 @@ static void cherryview_enable_rps(struct drm_device *dev) val = vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS); + /* RPS code assumes GPLL is used */ + WARN_ONCE((val & GPLLENABLE) == 0, "GPLL not enabled\n"); + DRM_DEBUG_DRIVER("GPLL enabled? %s\n", val & GPLLENABLE ? "yes" : "no"); DRM_DEBUG_DRIVER("GPU status: 0x%08x\n", val); @@ -5413,6 +5416,9 @@ static void valleyview_enable_rps(struct drm_device *dev) val = vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS); + /* RPS code assumes GPLL is used */ + WARN_ONCE((val & GPLLENABLE) == 0, "GPLL not enabled\n"); + DRM_DEBUG_DRIVER("GPLL enabled? %s\n", val & GPLLENABLE ? "yes" : "no"); DRM_DEBUG_DRIVER("GPU status: 0x%08x\n", val); -- cgit v0.10.2 From ce611ef81ff73fe64962fb1b22d8279aab205ad3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 7 Nov 2014 21:33:46 +0200 Subject: drm/i915: Improve PCBR debug information MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Always print the final PCBR register value on both vlv and chv, and also tell us whether the BIOS was a good citizen or not. Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index cbf84cc..eaae801 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5062,12 +5062,15 @@ static void cherryview_setup_pctx(struct drm_device *dev) pcbr = I915_READ(VLV_PCBR); if ((pcbr >> VLV_PCBR_ADDR_SHIFT) == 0) { + DRM_DEBUG_DRIVER("BIOS didn't set up PCBR, fixing up\n"); paddr = (dev_priv->mm.stolen_base + (gtt->stolen_size - pctx_size)); pctx_paddr = (paddr & (~4095)); I915_WRITE(VLV_PCBR, pctx_paddr); } + + DRM_DEBUG_DRIVER("PCBR: 0x%08x\n", I915_READ(VLV_PCBR)); } static void valleyview_setup_pctx(struct drm_device *dev) @@ -5093,6 +5096,8 @@ static void valleyview_setup_pctx(struct drm_device *dev) goto out; } + DRM_DEBUG_DRIVER("BIOS didn't set up PCBR, fixing up\n"); + /* * From the Gunit register HAS: * The Gfx driver is expected to program this register and ensure @@ -5111,6 +5116,7 @@ static void valleyview_setup_pctx(struct drm_device *dev) I915_WRITE(VLV_PCBR, pctx_paddr); out: + DRM_DEBUG_DRIVER("PCBR: 0x%08x\n", I915_READ(VLV_PCBR)); dev_priv->vlv_pctx = pctx; } @@ -5302,8 +5308,6 @@ static void cherryview_enable_rps(struct drm_device *dev) /* For now we assume BIOS is allocating and populating the PCBR */ pcbr = I915_READ(VLV_PCBR); - DRM_DEBUG_DRIVER("PCBR offset : 0x%x\n", pcbr); - /* 3: Enable RC6 */ if ((intel_enable_rc6(dev) & INTEL_RC6_ENABLE) && (pcbr >> VLV_PCBR_ADDR_SHIFT)) -- cgit v0.10.2 From dd06f88cd0c730421b575f423d8e75f6ed348804 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 10 Nov 2014 22:55:12 +0200 Subject: drm/i915: Refactor vlv/chv GPU frequency divider setup MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The divider used in the GPU frequency calculations is compatible between vlv and chv. vlv just wants doubled values compared to chv. Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index eaae801..feda281 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7206,99 +7206,65 @@ int sandybridge_pcode_write(struct drm_i915_private *dev_priv, u8 mbox, u32 val) return 0; } -static int byt_gpu_freq(struct drm_i915_private *dev_priv, int val) +static int vlv_gpu_freq_div(unsigned int czclk_freq) { - int div; - - /* 4 x czclk */ - switch (dev_priv->mem_freq) { - case 800: - div = 10; - break; - case 1066: - div = 12; - break; - case 1333: - div = 16; - break; + switch (czclk_freq) { + case 200: + return 10; + case 267: + return 12; + case 320: + case 333: + case 400: + return 16; default: return -1; } +} - return DIV_ROUND_CLOSEST(dev_priv->mem_freq * (val + 6 - 0xbd), 4 * div); +static int byt_gpu_freq(struct drm_i915_private *dev_priv, int val) +{ + int div, czclk_freq = DIV_ROUND_CLOSEST(dev_priv->mem_freq, 4); + + div = vlv_gpu_freq_div(czclk_freq); + if (div < 0) + return div; + + return DIV_ROUND_CLOSEST(czclk_freq * (val + 6 - 0xbd), div); } static int byt_freq_opcode(struct drm_i915_private *dev_priv, int val) { - int mul; + int mul, czclk_freq = DIV_ROUND_CLOSEST(dev_priv->mem_freq, 4); - /* 4 x czclk */ - switch (dev_priv->mem_freq) { - case 800: - mul = 10; - break; - case 1066: - mul = 12; - break; - case 1333: - mul = 16; - break; - default: - return -1; - } + mul = vlv_gpu_freq_div(czclk_freq); + if (mul < 0) + return mul; - return DIV_ROUND_CLOSEST(4 * mul * val, dev_priv->mem_freq) + 0xbd - 6; + return DIV_ROUND_CLOSEST(mul * val, czclk_freq) + 0xbd - 6; } static int chv_gpu_freq(struct drm_i915_private *dev_priv, int val) { - int div, freq; - - switch (dev_priv->rps.cz_freq) { - case 200: - div = 5; - break; - case 267: - div = 6; - break; - case 320: - case 333: - case 400: - div = 8; - break; - default: - return -1; - } + int div, czclk_freq = dev_priv->rps.cz_freq; - freq = (DIV_ROUND_CLOSEST((dev_priv->rps.cz_freq * val), 2 * div) / 2); + div = vlv_gpu_freq_div(czclk_freq) / 2; + if (div < 0) + return div; - return freq; + return DIV_ROUND_CLOSEST(czclk_freq * val, 2 * div) / 2; } static int chv_freq_opcode(struct drm_i915_private *dev_priv, int val) { - int mul, opcode; + int mul, czclk_freq = dev_priv->rps.cz_freq; - switch (dev_priv->rps.cz_freq) { - case 200: - mul = 5; - break; - case 267: - mul = 6; - break; - case 320: - case 333: - case 400: - mul = 8; - break; - default: - return -1; - } + mul = vlv_gpu_freq_div(czclk_freq) / 2; + if (mul < 0) + return mul; /* CHV needs even values */ - opcode = (DIV_ROUND_CLOSEST((val * 2 * mul), dev_priv->rps.cz_freq) * 2); - - return opcode; + return DIV_ROUND_CLOSEST(val * 2 * mul, czclk_freq) * 2; } int vlv_gpu_freq(struct drm_i915_private *dev_priv, int val) -- cgit v0.10.2 From 80b83b621732c3bb0ae617b4bb9cbb2f1436121d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 10 Nov 2014 22:55:14 +0200 Subject: drm/i915: Add missing newline to 'DDR speed' debug messages MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index feda281..290af15 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5153,7 +5153,7 @@ static void valleyview_init_gt_powersave(struct drm_device *dev) dev_priv->mem_freq = 1333; break; } - DRM_DEBUG_DRIVER("DDR speed: %d MHz", dev_priv->mem_freq); + DRM_DEBUG_DRIVER("DDR speed: %d MHz\n", dev_priv->mem_freq); dev_priv->rps.max_freq = valleyview_rps_max_freq(dev_priv); dev_priv->rps.rp0_freq = dev_priv->rps.max_freq; @@ -5222,7 +5222,7 @@ static void cherryview_init_gt_powersave(struct drm_device *dev) dev_priv->mem_freq = 1600; break; } - DRM_DEBUG_DRIVER("DDR speed: %d MHz", dev_priv->mem_freq); + DRM_DEBUG_DRIVER("DDR speed: %d MHz\n", dev_priv->mem_freq); dev_priv->rps.max_freq = cherryview_rps_max_freq(dev_priv); dev_priv->rps.rp0_freq = dev_priv->rps.max_freq; -- cgit v0.10.2 From ab3fb15730fd7a3d2d00c926bc95a81bb283a90f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 10 Nov 2014 22:55:15 +0200 Subject: drm/i915: Change CHV SKU400 GPU freq divider to 10 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit According to "Cherryview_GFXclocks_y14w36d1.xlsx" the GPU frequency divider should be 10 in when the CZ clock is 400 MHz. Change the code to agree so that we report the correct frequencies. Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 290af15..929648f 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7215,8 +7215,9 @@ static int vlv_gpu_freq_div(unsigned int czclk_freq) return 12; case 320: case 333: - case 400: return 16; + case 400: + return 20; default: return -1; } -- cgit v0.10.2 From f5f7d3c6e46ec8d4f8a38055267de89ec57b42b9 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 14 Nov 2014 17:24:32 +0000 Subject: drm/i915/skl: Remove spurious warn in get_ddi_pll() When reading out a DDI config that uses a PLL that is not part of the shared_dpll scheme (DPLL0), it's totally normal to end up in the default: case of that switch. Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index c42a040..7d58a64 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8029,8 +8029,6 @@ static void skylake_get_ddi_pll(struct drm_i915_private *dev_priv, case SKL_DPLL3: pipe_config->shared_dpll = DPLL_ID_SKL_DPLL3; break; - default: - WARN(1, "Unknown DPLL programmed\n"); } } -- cgit v0.10.2 From eb88bd1b47afd4d1f2754bbcc4bd1a5e57f31ebd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 13 Nov 2014 22:12:52 +0200 Subject: drm/i915: Drop the HSW special case from __gen6_gt_wait_for_thread_c0() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Bits [18:16] of GEN6_GT_THREAD_STATUS_REG have always had the same meaning since SNB. So treating them as something special for HSW doesn't make sense to me. Also the bits *seem* to work exactly the same way on IVB, HSW GT2 and HSW GT3. At least intel_reg_read gives the identical results on all platforms with and without forcewake. Also the HSW PM guide rev 0.99 (ww05 2013) doesn't say anything about those bits. It just says to poll for bits [2:0]. As does the more recent BDW PM guide. So just drop the HSW special case and treat all platforms the same way. Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index faf61f5..7a77cd5 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -2326,7 +2326,6 @@ enum punit_power_well { #define GEN6_GT_THREAD_STATUS_REG 0x13805c #define GEN6_GT_THREAD_STATUS_CORE_MASK 0x7 -#define GEN6_GT_THREAD_STATUS_CORE_MASK_HSW (0x7 | (0x07 << 16)) #define GEN6_GT_PERF_STATUS (MCHBAR_MIRROR_BASE_SNB + 0x5948) #define GEN6_RP_STATE_LIMITS (MCHBAR_MIRROR_BASE_SNB + 0x5994) diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 86a755a..5b3a49a 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -49,17 +49,11 @@ assert_device_not_suspended(struct drm_i915_private *dev_priv) static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv) { - u32 gt_thread_status_mask; - - if (IS_HASWELL(dev_priv->dev)) - gt_thread_status_mask = GEN6_GT_THREAD_STATUS_CORE_MASK_HSW; - else - gt_thread_status_mask = GEN6_GT_THREAD_STATUS_CORE_MASK; - /* w/a for a sporadic read returning 0 by waiting for the GT * thread to wake up. */ - if (wait_for_atomic_us((__raw_i915_read32(dev_priv, GEN6_GT_THREAD_STATUS_REG) & gt_thread_status_mask) == 0, 500)) + if (wait_for_atomic_us((__raw_i915_read32(dev_priv, GEN6_GT_THREAD_STATUS_REG) & + GEN6_GT_THREAD_STATUS_CORE_MASK) == 0, 500)) DRM_ERROR("GT thread status wait timed out\n"); } -- cgit v0.10.2 From 2fe486c7eae9fc5b52297aed0e4008df6b66b01c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 13 Nov 2014 22:12:53 +0200 Subject: drm/i915: Drop WaRsForcewakeWaitTC0:vlv MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit GEN6_GT_THREAD_STATUS_REG doesn't seem to exist on VLV. Reads just give 0x0 no matter what the state of the render and media wells. There was also some hint in the Gunit HAS that thread status not being needed on VLV, and hence dropped when bringing stuff over from the IVB design. Not really a definite comment about the specific register itself though. Also the w/a itself is no longer listed for VLV in the database. It was there some time ago in the past, but I guess someone figured out the mistake and dropped it. So let's just drop it from the code as well. Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 5b3a49a..c98421d 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -222,10 +222,6 @@ static void __vlv_force_wake_get(struct drm_i915_private *dev_priv, FORCEWAKE_ACK_TIMEOUT_MS)) DRM_ERROR("Timed out: waiting for media to ack.\n"); } - - /* WaRsForcewakeWaitTC0:vlv */ - if (!IS_CHERRYVIEW(dev_priv->dev)) - __gen6_gt_wait_for_thread_c0(dev_priv); } static void __vlv_force_wake_put(struct drm_i915_private *dev_priv, -- cgit v0.10.2 From a4f1289eaacdc2651355201b1ddded91710c89a2 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 14 Nov 2014 08:52:27 -0800 Subject: drm/i915: Make dp aux pack/unpack public outside intel_dp.c No functional change. Just making it public for use outside intel_dp.c Allowing split psr functions. Signed-off-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index a16193f..a89585d 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -227,8 +227,7 @@ intel_dp_mode_valid(struct drm_connector *connector, return MODE_OK; } -static uint32_t -pack_aux(const uint8_t *src, int src_bytes) +uint32_t intel_dp_pack_aux(const uint8_t *src, int src_bytes) { int i; uint32_t v = 0; @@ -240,8 +239,7 @@ pack_aux(const uint8_t *src, int src_bytes) return v; } -static void -unpack_aux(uint32_t src, uint8_t *dst, int dst_bytes) +void intel_dp_unpack_aux(uint32_t src, uint8_t *dst, int dst_bytes) { int i; if (dst_bytes > 4) @@ -863,7 +861,8 @@ intel_dp_aux_ch(struct intel_dp *intel_dp, /* Load the send data into the aux channel data registers */ for (i = 0; i < send_bytes; i += 4) I915_WRITE(ch_data + i, - pack_aux(send + i, send_bytes - i)); + intel_dp_pack_aux(send + i, + send_bytes - i)); /* Send the command and wait for it to complete */ I915_WRITE(ch_ctl, send_ctl); @@ -917,8 +916,8 @@ intel_dp_aux_ch(struct intel_dp *intel_dp, recv_bytes = recv_size; for (i = 0; i < recv_bytes; i += 4) - unpack_aux(I915_READ(ch_data + i), - recv + i, recv_bytes - i); + intel_dp_unpack_aux(I915_READ(ch_data + i), + recv + i, recv_bytes - i); ret = recv_bytes; out: @@ -2159,7 +2158,7 @@ static void intel_edp_psr_enable_sink(struct intel_dp *intel_dp) /* Setup AUX registers */ for (i = 0; i < sizeof(aux_msg); i += 4) I915_WRITE(EDP_PSR_AUX_DATA1(dev) + i, - pack_aux(&aux_msg[i], sizeof(aux_msg) - i)); + intel_dp_pack_aux(&aux_msg[i], sizeof(aux_msg) - i)); I915_WRITE(EDP_PSR_AUX_CTL(dev), DP_AUX_CH_CTL_TIME_OUT_400us | -- cgit v0.10.2 From 0bc12bcb1b9686d7011f16410ba17ed0740167c3 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 14 Nov 2014 08:52:28 -0800 Subject: drm/i915: Introduce intel_psr.c No functional changes. Just cleaning and reorganizing it. v2: Rebase it puting it to begin of psr rework. This helps to blame easily at least latest changes. Signed-off-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 891e584..e4083e4 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -51,6 +51,7 @@ i915-y += intel_audio.o \ intel_frontbuffer.o \ intel_modes.o \ intel_overlay.o \ + intel_psr.o \ intel_sideband.o \ intel_sprite.o i915-$(CONFIG_ACPI) += intel_acpi.o intel_opregion.o diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index ca33ee9..3e4c63c 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1583,7 +1583,7 @@ static void intel_enable_ddi(struct intel_encoder *intel_encoder) intel_dp_stop_link_train(intel_dp); intel_edp_backlight_on(intel_dp); - intel_edp_psr_enable(intel_dp); + intel_psr_enable(intel_dp); } if (intel_crtc->config.has_audio) { @@ -1609,7 +1609,7 @@ static void intel_disable_ddi(struct intel_encoder *intel_encoder) if (type == INTEL_OUTPUT_EDP) { struct intel_dp *intel_dp = enc_to_intel_dp(encoder); - intel_edp_psr_disable(intel_dp); + intel_psr_disable(intel_dp); intel_edp_backlight_off(intel_dp); } } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 7d58a64..2632f2a 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -12449,7 +12449,7 @@ static void intel_setup_outputs(struct drm_device *dev) if (SUPPORTS_TV(dev)) intel_tv_init(dev); - intel_edp_psr_init(dev); + intel_psr_init(dev); for_each_intel_encoder(dev, encoder) { encoder->base.possible_crtcs = encoder->crtc_mask; diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index a89585d..7369e6f 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2066,385 +2066,6 @@ static void intel_dp_get_config(struct intel_encoder *encoder, } } -static bool is_edp_psr(struct intel_dp *intel_dp) -{ - return intel_dp->psr_dpcd[0] & DP_PSR_IS_SUPPORTED; -} - -static bool intel_edp_is_psr_enabled(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - if (!HAS_PSR(dev)) - return false; - - return I915_READ(EDP_PSR_CTL(dev)) & EDP_PSR_ENABLE; -} - -static void intel_edp_psr_write_vsc(struct intel_dp *intel_dp, - struct edp_vsc_psr *vsc_psr) -{ - struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = dig_port->base.base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_crtc *crtc = to_intel_crtc(dig_port->base.base.crtc); - u32 ctl_reg = HSW_TVIDEO_DIP_CTL(crtc->config.cpu_transcoder); - u32 data_reg = HSW_TVIDEO_DIP_VSC_DATA(crtc->config.cpu_transcoder); - uint32_t *data = (uint32_t *) vsc_psr; - unsigned int i; - - /* As per BSPec (Pipe Video Data Island Packet), we need to disable - the video DIP being updated before program video DIP data buffer - registers for DIP being updated. */ - I915_WRITE(ctl_reg, 0); - POSTING_READ(ctl_reg); - - for (i = 0; i < VIDEO_DIP_VSC_DATA_SIZE; i += 4) { - if (i < sizeof(struct edp_vsc_psr)) - I915_WRITE(data_reg + i, *data++); - else - I915_WRITE(data_reg + i, 0); - } - - I915_WRITE(ctl_reg, VIDEO_DIP_ENABLE_VSC_HSW); - POSTING_READ(ctl_reg); -} - -static void intel_edp_psr_setup_vsc(struct intel_dp *intel_dp) -{ - struct edp_vsc_psr psr_vsc; - - /* Prepare VSC packet as per EDP 1.3 spec, Table 3.10 */ - memset(&psr_vsc, 0, sizeof(psr_vsc)); - psr_vsc.sdp_header.HB0 = 0; - psr_vsc.sdp_header.HB1 = 0x7; - psr_vsc.sdp_header.HB2 = 0x2; - psr_vsc.sdp_header.HB3 = 0x8; - intel_edp_psr_write_vsc(intel_dp, &psr_vsc); -} - -static void intel_edp_psr_enable_sink(struct intel_dp *intel_dp) -{ - struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = dig_port->base.base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - uint32_t aux_clock_divider; - int precharge = 0x3; - bool only_standby = false; - static const uint8_t aux_msg[] = { - [0] = DP_AUX_NATIVE_WRITE << 4, - [1] = DP_SET_POWER >> 8, - [2] = DP_SET_POWER & 0xff, - [3] = 1 - 1, - [4] = DP_SET_POWER_D0, - }; - int i; - - BUILD_BUG_ON(sizeof(aux_msg) > 20); - - aux_clock_divider = intel_dp->get_aux_clock_divider(intel_dp, 0); - - if (IS_BROADWELL(dev) && dig_port->port != PORT_A) - only_standby = true; - - /* Enable PSR in sink */ - if (intel_dp->psr_dpcd[1] & DP_PSR_NO_TRAIN_ON_EXIT || only_standby) - drm_dp_dpcd_writeb(&intel_dp->aux, DP_PSR_EN_CFG, - DP_PSR_ENABLE & ~DP_PSR_MAIN_LINK_ACTIVE); - else - drm_dp_dpcd_writeb(&intel_dp->aux, DP_PSR_EN_CFG, - DP_PSR_ENABLE | DP_PSR_MAIN_LINK_ACTIVE); - - /* Setup AUX registers */ - for (i = 0; i < sizeof(aux_msg); i += 4) - I915_WRITE(EDP_PSR_AUX_DATA1(dev) + i, - intel_dp_pack_aux(&aux_msg[i], sizeof(aux_msg) - i)); - - I915_WRITE(EDP_PSR_AUX_CTL(dev), - DP_AUX_CH_CTL_TIME_OUT_400us | - (sizeof(aux_msg) << DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT) | - (precharge << DP_AUX_CH_CTL_PRECHARGE_2US_SHIFT) | - (aux_clock_divider << DP_AUX_CH_CTL_BIT_CLOCK_2X_SHIFT)); -} - -static void intel_edp_psr_enable_source(struct intel_dp *intel_dp) -{ - struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = dig_port->base.base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - uint32_t max_sleep_time = 0x1f; - uint32_t idle_frames = 1; - uint32_t val = 0x0; - const uint32_t link_entry_time = EDP_PSR_MIN_LINK_ENTRY_TIME_8_LINES; - bool only_standby = false; - - if (IS_BROADWELL(dev) && dig_port->port != PORT_A) - only_standby = true; - - if (intel_dp->psr_dpcd[1] & DP_PSR_NO_TRAIN_ON_EXIT || only_standby) { - val |= EDP_PSR_LINK_STANDBY; - val |= EDP_PSR_TP2_TP3_TIME_0us; - val |= EDP_PSR_TP1_TIME_0us; - val |= EDP_PSR_SKIP_AUX_EXIT; - val |= IS_BROADWELL(dev) ? BDW_PSR_SINGLE_FRAME : 0; - } else - val |= EDP_PSR_LINK_DISABLE; - - I915_WRITE(EDP_PSR_CTL(dev), val | - (IS_BROADWELL(dev) ? 0 : link_entry_time) | - max_sleep_time << EDP_PSR_MAX_SLEEP_TIME_SHIFT | - idle_frames << EDP_PSR_IDLE_FRAME_SHIFT | - EDP_PSR_ENABLE); -} - -static bool intel_edp_psr_match_conditions(struct intel_dp *intel_dp) -{ - struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = dig_port->base.base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_crtc *crtc = dig_port->base.base.crtc; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - - lockdep_assert_held(&dev_priv->psr.lock); - WARN_ON(!drm_modeset_is_locked(&dev->mode_config.connection_mutex)); - WARN_ON(!drm_modeset_is_locked(&crtc->mutex)); - - dev_priv->psr.source_ok = false; - - if (IS_HASWELL(dev) && dig_port->port != PORT_A) { - DRM_DEBUG_KMS("HSW ties PSR to DDI A (eDP)\n"); - return false; - } - - if (!i915.enable_psr) { - DRM_DEBUG_KMS("PSR disable by flag\n"); - return false; - } - - /* Below limitations aren't valid for Broadwell */ - if (IS_BROADWELL(dev)) - goto out; - - if (I915_READ(HSW_STEREO_3D_CTL(intel_crtc->config.cpu_transcoder)) & - S3D_ENABLE) { - DRM_DEBUG_KMS("PSR condition failed: Stereo 3D is Enabled\n"); - return false; - } - - if (intel_crtc->config.adjusted_mode.flags & DRM_MODE_FLAG_INTERLACE) { - DRM_DEBUG_KMS("PSR condition failed: Interlaced is Enabled\n"); - return false; - } - - out: - dev_priv->psr.source_ok = true; - return true; -} - -static void intel_edp_psr_do_enable(struct intel_dp *intel_dp) -{ - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = dev->dev_private; - - WARN_ON(I915_READ(EDP_PSR_CTL(dev)) & EDP_PSR_ENABLE); - WARN_ON(dev_priv->psr.active); - lockdep_assert_held(&dev_priv->psr.lock); - - /* Enable/Re-enable PSR on the host */ - intel_edp_psr_enable_source(intel_dp); - - dev_priv->psr.active = true; -} - -void intel_edp_psr_enable(struct intel_dp *intel_dp) -{ - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = dev->dev_private; - - if (!HAS_PSR(dev)) { - DRM_DEBUG_KMS("PSR not supported on this platform\n"); - return; - } - - if (!is_edp_psr(intel_dp)) { - DRM_DEBUG_KMS("PSR not supported by this panel\n"); - return; - } - - mutex_lock(&dev_priv->psr.lock); - if (dev_priv->psr.enabled) { - DRM_DEBUG_KMS("PSR already in use\n"); - goto unlock; - } - - if (!intel_edp_psr_match_conditions(intel_dp)) - goto unlock; - - dev_priv->psr.busy_frontbuffer_bits = 0; - - intel_edp_psr_setup_vsc(intel_dp); - - /* Avoid continuous PSR exit by masking memup and hpd */ - I915_WRITE(EDP_PSR_DEBUG_CTL(dev), EDP_PSR_DEBUG_MASK_MEMUP | - EDP_PSR_DEBUG_MASK_HPD | EDP_PSR_DEBUG_MASK_LPSP); - - /* Enable PSR on the panel */ - intel_edp_psr_enable_sink(intel_dp); - - dev_priv->psr.enabled = intel_dp; -unlock: - mutex_unlock(&dev_priv->psr.lock); -} - -void intel_edp_psr_disable(struct intel_dp *intel_dp) -{ - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = dev->dev_private; - - mutex_lock(&dev_priv->psr.lock); - if (!dev_priv->psr.enabled) { - mutex_unlock(&dev_priv->psr.lock); - return; - } - - if (dev_priv->psr.active) { - I915_WRITE(EDP_PSR_CTL(dev), - I915_READ(EDP_PSR_CTL(dev)) & ~EDP_PSR_ENABLE); - - /* Wait till PSR is idle */ - if (_wait_for((I915_READ(EDP_PSR_STATUS_CTL(dev)) & - EDP_PSR_STATUS_STATE_MASK) == 0, 2000, 10)) - DRM_ERROR("Timed out waiting for PSR Idle State\n"); - - dev_priv->psr.active = false; - } else { - WARN_ON(I915_READ(EDP_PSR_CTL(dev)) & EDP_PSR_ENABLE); - } - - dev_priv->psr.enabled = NULL; - mutex_unlock(&dev_priv->psr.lock); - - cancel_delayed_work_sync(&dev_priv->psr.work); -} - -static void intel_edp_psr_work(struct work_struct *work) -{ - struct drm_i915_private *dev_priv = - container_of(work, typeof(*dev_priv), psr.work.work); - struct intel_dp *intel_dp = dev_priv->psr.enabled; - - /* We have to make sure PSR is ready for re-enable - * otherwise it keeps disabled until next full enable/disable cycle. - * PSR might take some time to get fully disabled - * and be ready for re-enable. - */ - if (wait_for((I915_READ(EDP_PSR_STATUS_CTL(dev_priv->dev)) & - EDP_PSR_STATUS_STATE_MASK) == 0, 50)) { - DRM_ERROR("Timed out waiting for PSR Idle for re-enable\n"); - return; - } - - mutex_lock(&dev_priv->psr.lock); - intel_dp = dev_priv->psr.enabled; - - if (!intel_dp) - goto unlock; - - /* - * The delayed work can race with an invalidate hence we need to - * recheck. Since psr_flush first clears this and then reschedules we - * won't ever miss a flush when bailing out here. - */ - if (dev_priv->psr.busy_frontbuffer_bits) - goto unlock; - - intel_edp_psr_do_enable(intel_dp); -unlock: - mutex_unlock(&dev_priv->psr.lock); -} - -static void intel_edp_psr_do_exit(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - if (dev_priv->psr.active) { - u32 val = I915_READ(EDP_PSR_CTL(dev)); - - WARN_ON(!(val & EDP_PSR_ENABLE)); - - I915_WRITE(EDP_PSR_CTL(dev), val & ~EDP_PSR_ENABLE); - - dev_priv->psr.active = false; - } - -} - -void intel_edp_psr_invalidate(struct drm_device *dev, - unsigned frontbuffer_bits) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_crtc *crtc; - enum pipe pipe; - - mutex_lock(&dev_priv->psr.lock); - if (!dev_priv->psr.enabled) { - mutex_unlock(&dev_priv->psr.lock); - return; - } - - crtc = dp_to_dig_port(dev_priv->psr.enabled)->base.base.crtc; - pipe = to_intel_crtc(crtc)->pipe; - - intel_edp_psr_do_exit(dev); - - frontbuffer_bits &= INTEL_FRONTBUFFER_ALL_MASK(pipe); - - dev_priv->psr.busy_frontbuffer_bits |= frontbuffer_bits; - mutex_unlock(&dev_priv->psr.lock); -} - -void intel_edp_psr_flush(struct drm_device *dev, - unsigned frontbuffer_bits) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_crtc *crtc; - enum pipe pipe; - - mutex_lock(&dev_priv->psr.lock); - if (!dev_priv->psr.enabled) { - mutex_unlock(&dev_priv->psr.lock); - return; - } - - crtc = dp_to_dig_port(dev_priv->psr.enabled)->base.base.crtc; - pipe = to_intel_crtc(crtc)->pipe; - dev_priv->psr.busy_frontbuffer_bits &= ~frontbuffer_bits; - - /* - * On Haswell sprite plane updates don't result in a psr invalidating - * signal in the hardware. Which means we need to manually fake this in - * software for all flushes, not just when we've seen a preceding - * invalidation through frontbuffer rendering. - */ - if (IS_HASWELL(dev) && - (frontbuffer_bits & INTEL_FRONTBUFFER_SPRITE(pipe))) - intel_edp_psr_do_exit(dev); - - if (!dev_priv->psr.active && !dev_priv->psr.busy_frontbuffer_bits) - schedule_delayed_work(&dev_priv->psr.work, - msecs_to_jiffies(100)); - mutex_unlock(&dev_priv->psr.lock); -} - -void intel_edp_psr_init(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - INIT_DELAYED_WORK(&dev_priv->psr.work, intel_edp_psr_work); - mutex_init(&dev_priv->psr.lock); -} - static void intel_disable_dp(struct intel_encoder *encoder) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); @@ -5095,7 +4716,7 @@ void intel_dp_set_drrs_state(struct drm_device *dev, int refresh_rate) * hard to tell without seeing the user of this function of this code. * Check locking and ordering once that lands. */ - if (INTEL_INFO(dev)->gen < 8 && intel_edp_is_psr_enabled(dev)) { + if (INTEL_INFO(dev)->gen < 8 && intel_psr_is_enabled(dev)) { DRM_DEBUG_KMS("DRRS is disabled as PSR is enabled\n"); return; } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index d67c59b..d1f9b63 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -999,21 +999,16 @@ void intel_edp_backlight_off(struct intel_dp *intel_dp); void intel_edp_panel_vdd_on(struct intel_dp *intel_dp); void intel_edp_panel_on(struct intel_dp *intel_dp); void intel_edp_panel_off(struct intel_dp *intel_dp); -void intel_edp_psr_enable(struct intel_dp *intel_dp); -void intel_edp_psr_disable(struct intel_dp *intel_dp); void intel_dp_set_drrs_state(struct drm_device *dev, int refresh_rate); -void intel_edp_psr_invalidate(struct drm_device *dev, - unsigned frontbuffer_bits); -void intel_edp_psr_flush(struct drm_device *dev, - unsigned frontbuffer_bits); -void intel_edp_psr_init(struct drm_device *dev); - void intel_dp_add_properties(struct intel_dp *intel_dp, struct drm_connector *connector); void intel_dp_mst_suspend(struct drm_device *dev); void intel_dp_mst_resume(struct drm_device *dev); int intel_dp_max_link_bw(struct intel_dp *intel_dp); void intel_dp_hot_plug(struct intel_encoder *intel_encoder); void vlv_power_sequencer_reset(struct drm_i915_private *dev_priv); +uint32_t intel_dp_pack_aux(const uint8_t *src, int src_bytes); +void intel_dp_unpack_aux(uint32_t src, uint8_t *dst, int dst_bytes); + /* intel_dp_mst.c */ int intel_dp_mst_encoder_init(struct intel_digital_port *intel_dig_port, int conn_id); void intel_dp_mst_encoder_cleanup(struct intel_digital_port *intel_dig_port); @@ -1117,6 +1112,16 @@ void intel_backlight_register(struct drm_device *dev); void intel_backlight_unregister(struct drm_device *dev); +/* intel_psr.c */ +bool intel_psr_is_enabled(struct drm_device *dev); +void intel_psr_enable(struct intel_dp *intel_dp); +void intel_psr_disable(struct intel_dp *intel_dp); +void intel_psr_invalidate(struct drm_device *dev, + unsigned frontbuffer_bits); +void intel_psr_flush(struct drm_device *dev, + unsigned frontbuffer_bits); +void intel_psr_init(struct drm_device *dev); + /* intel_runtime_pm.c */ int intel_power_domains_init(struct drm_i915_private *); void intel_power_domains_fini(struct drm_i915_private *); diff --git a/drivers/gpu/drm/i915/intel_frontbuffer.c b/drivers/gpu/drm/i915/intel_frontbuffer.c index 58cf2e6..79f6d72 100644 --- a/drivers/gpu/drm/i915/intel_frontbuffer.c +++ b/drivers/gpu/drm/i915/intel_frontbuffer.c @@ -156,7 +156,7 @@ void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj, intel_mark_fb_busy(dev, obj->frontbuffer_bits, ring); - intel_edp_psr_invalidate(dev, obj->frontbuffer_bits); + intel_psr_invalidate(dev, obj->frontbuffer_bits); } /** @@ -182,7 +182,7 @@ void intel_frontbuffer_flush(struct drm_device *dev, intel_mark_fb_busy(dev, frontbuffer_bits, NULL); - intel_edp_psr_flush(dev, frontbuffer_bits); + intel_psr_flush(dev, frontbuffer_bits); /* * FIXME: Unconditional fbc flushing here is a rather gross hack and diff --git a/drivers/gpu/drm/i915/intel_psr.c b/drivers/gpu/drm/i915/intel_psr.c new file mode 100644 index 0000000..7b3ed91 --- /dev/null +++ b/drivers/gpu/drm/i915/intel_psr.c @@ -0,0 +1,408 @@ +/* + * Copyright © 2014 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include + +#include "intel_drv.h" +#include "i915_drv.h" + +static bool is_edp_psr(struct intel_dp *intel_dp) +{ + return intel_dp->psr_dpcd[0] & DP_PSR_IS_SUPPORTED; +} + +bool intel_psr_is_enabled(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + if (!HAS_PSR(dev)) + return false; + + return I915_READ(EDP_PSR_CTL(dev)) & EDP_PSR_ENABLE; +} + +static void intel_psr_write_vsc(struct intel_dp *intel_dp, + struct edp_vsc_psr *vsc_psr) +{ + struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_crtc *crtc = to_intel_crtc(dig_port->base.base.crtc); + u32 ctl_reg = HSW_TVIDEO_DIP_CTL(crtc->config.cpu_transcoder); + u32 data_reg = HSW_TVIDEO_DIP_VSC_DATA(crtc->config.cpu_transcoder); + uint32_t *data = (uint32_t *) vsc_psr; + unsigned int i; + + /* As per BSPec (Pipe Video Data Island Packet), we need to disable + the video DIP being updated before program video DIP data buffer + registers for DIP being updated. */ + I915_WRITE(ctl_reg, 0); + POSTING_READ(ctl_reg); + + for (i = 0; i < VIDEO_DIP_VSC_DATA_SIZE; i += 4) { + if (i < sizeof(struct edp_vsc_psr)) + I915_WRITE(data_reg + i, *data++); + else + I915_WRITE(data_reg + i, 0); + } + + I915_WRITE(ctl_reg, VIDEO_DIP_ENABLE_VSC_HSW); + POSTING_READ(ctl_reg); +} + +static void intel_psr_setup_vsc(struct intel_dp *intel_dp) +{ + struct edp_vsc_psr psr_vsc; + + /* Prepare VSC packet as per EDP 1.3 spec, Table 3.10 */ + memset(&psr_vsc, 0, sizeof(psr_vsc)); + psr_vsc.sdp_header.HB0 = 0; + psr_vsc.sdp_header.HB1 = 0x7; + psr_vsc.sdp_header.HB2 = 0x2; + psr_vsc.sdp_header.HB3 = 0x8; + intel_psr_write_vsc(intel_dp, &psr_vsc); +} + +static void intel_psr_enable_sink(struct intel_dp *intel_dp) +{ + struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + uint32_t aux_clock_divider; + int precharge = 0x3; + bool only_standby = false; + static const uint8_t aux_msg[] = { + [0] = DP_AUX_NATIVE_WRITE << 4, + [1] = DP_SET_POWER >> 8, + [2] = DP_SET_POWER & 0xff, + [3] = 1 - 1, + [4] = DP_SET_POWER_D0, + }; + int i; + + BUILD_BUG_ON(sizeof(aux_msg) > 20); + + aux_clock_divider = intel_dp->get_aux_clock_divider(intel_dp, 0); + + if (IS_BROADWELL(dev) && dig_port->port != PORT_A) + only_standby = true; + + /* Enable PSR in sink */ + if (intel_dp->psr_dpcd[1] & DP_PSR_NO_TRAIN_ON_EXIT || only_standby) + drm_dp_dpcd_writeb(&intel_dp->aux, DP_PSR_EN_CFG, + DP_PSR_ENABLE & ~DP_PSR_MAIN_LINK_ACTIVE); + else + drm_dp_dpcd_writeb(&intel_dp->aux, DP_PSR_EN_CFG, + DP_PSR_ENABLE | DP_PSR_MAIN_LINK_ACTIVE); + + /* Setup AUX registers */ + for (i = 0; i < sizeof(aux_msg); i += 4) + I915_WRITE(EDP_PSR_AUX_DATA1(dev) + i, + intel_dp_pack_aux(&aux_msg[i], sizeof(aux_msg) - i)); + + I915_WRITE(EDP_PSR_AUX_CTL(dev), + DP_AUX_CH_CTL_TIME_OUT_400us | + (sizeof(aux_msg) << DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT) | + (precharge << DP_AUX_CH_CTL_PRECHARGE_2US_SHIFT) | + (aux_clock_divider << DP_AUX_CH_CTL_BIT_CLOCK_2X_SHIFT)); +} + +static void intel_psr_enable_source(struct intel_dp *intel_dp) +{ + struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + uint32_t max_sleep_time = 0x1f; + uint32_t idle_frames = 1; + uint32_t val = 0x0; + const uint32_t link_entry_time = EDP_PSR_MIN_LINK_ENTRY_TIME_8_LINES; + bool only_standby = false; + + if (IS_BROADWELL(dev) && dig_port->port != PORT_A) + only_standby = true; + + if (intel_dp->psr_dpcd[1] & DP_PSR_NO_TRAIN_ON_EXIT || only_standby) { + val |= EDP_PSR_LINK_STANDBY; + val |= EDP_PSR_TP2_TP3_TIME_0us; + val |= EDP_PSR_TP1_TIME_0us; + val |= EDP_PSR_SKIP_AUX_EXIT; + val |= IS_BROADWELL(dev) ? BDW_PSR_SINGLE_FRAME : 0; + } else + val |= EDP_PSR_LINK_DISABLE; + + I915_WRITE(EDP_PSR_CTL(dev), val | + (IS_BROADWELL(dev) ? 0 : link_entry_time) | + max_sleep_time << EDP_PSR_MAX_SLEEP_TIME_SHIFT | + idle_frames << EDP_PSR_IDLE_FRAME_SHIFT | + EDP_PSR_ENABLE); +} + +static bool intel_psr_match_conditions(struct intel_dp *intel_dp) +{ + struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_crtc *crtc = dig_port->base.base.crtc; + struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + + lockdep_assert_held(&dev_priv->psr.lock); + WARN_ON(!drm_modeset_is_locked(&dev->mode_config.connection_mutex)); + WARN_ON(!drm_modeset_is_locked(&crtc->mutex)); + + dev_priv->psr.source_ok = false; + + if (IS_HASWELL(dev) && dig_port->port != PORT_A) { + DRM_DEBUG_KMS("HSW ties PSR to DDI A (eDP)\n"); + return false; + } + + if (!i915.enable_psr) { + DRM_DEBUG_KMS("PSR disable by flag\n"); + return false; + } + + /* Below limitations aren't valid for Broadwell */ + if (IS_BROADWELL(dev)) + goto out; + + if (I915_READ(HSW_STEREO_3D_CTL(intel_crtc->config.cpu_transcoder)) & + S3D_ENABLE) { + DRM_DEBUG_KMS("PSR condition failed: Stereo 3D is Enabled\n"); + return false; + } + + if (intel_crtc->config.adjusted_mode.flags & DRM_MODE_FLAG_INTERLACE) { + DRM_DEBUG_KMS("PSR condition failed: Interlaced is Enabled\n"); + return false; + } + + out: + dev_priv->psr.source_ok = true; + return true; +} + +static void intel_psr_do_enable(struct intel_dp *intel_dp) +{ + struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = intel_dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + + WARN_ON(I915_READ(EDP_PSR_CTL(dev)) & EDP_PSR_ENABLE); + WARN_ON(dev_priv->psr.active); + lockdep_assert_held(&dev_priv->psr.lock); + + /* Enable/Re-enable PSR on the host */ + intel_psr_enable_source(intel_dp); + + dev_priv->psr.active = true; +} + +void intel_psr_enable(struct intel_dp *intel_dp) +{ + struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = intel_dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + + if (!HAS_PSR(dev)) { + DRM_DEBUG_KMS("PSR not supported on this platform\n"); + return; + } + + if (!is_edp_psr(intel_dp)) { + DRM_DEBUG_KMS("PSR not supported by this panel\n"); + return; + } + + mutex_lock(&dev_priv->psr.lock); + if (dev_priv->psr.enabled) { + DRM_DEBUG_KMS("PSR already in use\n"); + goto unlock; + } + + if (!intel_psr_match_conditions(intel_dp)) + goto unlock; + + dev_priv->psr.busy_frontbuffer_bits = 0; + + intel_psr_setup_vsc(intel_dp); + + /* Avoid continuous PSR exit by masking memup and hpd */ + I915_WRITE(EDP_PSR_DEBUG_CTL(dev), EDP_PSR_DEBUG_MASK_MEMUP | + EDP_PSR_DEBUG_MASK_HPD | EDP_PSR_DEBUG_MASK_LPSP); + + /* Enable PSR on the panel */ + intel_psr_enable_sink(intel_dp); + + dev_priv->psr.enabled = intel_dp; +unlock: + mutex_unlock(&dev_priv->psr.lock); +} + +void intel_psr_disable(struct intel_dp *intel_dp) +{ + struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); + struct drm_device *dev = intel_dig_port->base.base.dev; + struct drm_i915_private *dev_priv = dev->dev_private; + + mutex_lock(&dev_priv->psr.lock); + if (!dev_priv->psr.enabled) { + mutex_unlock(&dev_priv->psr.lock); + return; + } + + if (dev_priv->psr.active) { + I915_WRITE(EDP_PSR_CTL(dev), + I915_READ(EDP_PSR_CTL(dev)) & ~EDP_PSR_ENABLE); + + /* Wait till PSR is idle */ + if (_wait_for((I915_READ(EDP_PSR_STATUS_CTL(dev)) & + EDP_PSR_STATUS_STATE_MASK) == 0, 2000, 10)) + DRM_ERROR("Timed out waiting for PSR Idle State\n"); + + dev_priv->psr.active = false; + } else { + WARN_ON(I915_READ(EDP_PSR_CTL(dev)) & EDP_PSR_ENABLE); + } + + dev_priv->psr.enabled = NULL; + mutex_unlock(&dev_priv->psr.lock); + + cancel_delayed_work_sync(&dev_priv->psr.work); +} + +static void intel_psr_work(struct work_struct *work) +{ + struct drm_i915_private *dev_priv = + container_of(work, typeof(*dev_priv), psr.work.work); + struct intel_dp *intel_dp = dev_priv->psr.enabled; + + /* We have to make sure PSR is ready for re-enable + * otherwise it keeps disabled until next full enable/disable cycle. + * PSR might take some time to get fully disabled + * and be ready for re-enable. + */ + if (wait_for((I915_READ(EDP_PSR_STATUS_CTL(dev_priv->dev)) & + EDP_PSR_STATUS_STATE_MASK) == 0, 50)) { + DRM_ERROR("Timed out waiting for PSR Idle for re-enable\n"); + return; + } + + mutex_lock(&dev_priv->psr.lock); + intel_dp = dev_priv->psr.enabled; + + if (!intel_dp) + goto unlock; + + /* + * The delayed work can race with an invalidate hence we need to + * recheck. Since psr_flush first clears this and then reschedules we + * won't ever miss a flush when bailing out here. + */ + if (dev_priv->psr.busy_frontbuffer_bits) + goto unlock; + + intel_psr_do_enable(intel_dp); +unlock: + mutex_unlock(&dev_priv->psr.lock); +} + +static void intel_psr_exit(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + if (dev_priv->psr.active) { + u32 val = I915_READ(EDP_PSR_CTL(dev)); + + WARN_ON(!(val & EDP_PSR_ENABLE)); + + I915_WRITE(EDP_PSR_CTL(dev), val & ~EDP_PSR_ENABLE); + + dev_priv->psr.active = false; + } + +} + +void intel_psr_invalidate(struct drm_device *dev, + unsigned frontbuffer_bits) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_crtc *crtc; + enum pipe pipe; + + mutex_lock(&dev_priv->psr.lock); + if (!dev_priv->psr.enabled) { + mutex_unlock(&dev_priv->psr.lock); + return; + } + + crtc = dp_to_dig_port(dev_priv->psr.enabled)->base.base.crtc; + pipe = to_intel_crtc(crtc)->pipe; + + intel_psr_exit(dev); + + frontbuffer_bits &= INTEL_FRONTBUFFER_ALL_MASK(pipe); + + dev_priv->psr.busy_frontbuffer_bits |= frontbuffer_bits; + mutex_unlock(&dev_priv->psr.lock); +} + +void intel_psr_flush(struct drm_device *dev, + unsigned frontbuffer_bits) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_crtc *crtc; + enum pipe pipe; + + mutex_lock(&dev_priv->psr.lock); + if (!dev_priv->psr.enabled) { + mutex_unlock(&dev_priv->psr.lock); + return; + } + + crtc = dp_to_dig_port(dev_priv->psr.enabled)->base.base.crtc; + pipe = to_intel_crtc(crtc)->pipe; + dev_priv->psr.busy_frontbuffer_bits &= ~frontbuffer_bits; + + /* + * On Haswell sprite plane updates don't result in a psr invalidating + * signal in the hardware. Which means we need to manually fake this in + * software for all flushes, not just when we've seen a preceding + * invalidation through frontbuffer rendering. + */ + if (IS_HASWELL(dev) && + (frontbuffer_bits & INTEL_FRONTBUFFER_SPRITE(pipe))) + intel_psr_exit(dev); + + if (!dev_priv->psr.active && !dev_priv->psr.busy_frontbuffer_bits) + schedule_delayed_work(&dev_priv->psr.work, + msecs_to_jiffies(100)); + mutex_unlock(&dev_priv->psr.lock); +} + +void intel_psr_init(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + INIT_DELAYED_WORK(&dev_priv->psr.work, intel_psr_work); + mutex_init(&dev_priv->psr.lock); +} -- cgit v0.10.2 From b2b89f55dfe3f19c8028d6c907d1ab53766b534b Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 14 Nov 2014 08:52:29 -0800 Subject: drm/i915: Add PSR docbook Let's document PSR a bit. No functional changes. v2: Add actual DocBook entry and accept Daniel's improvements. Cc: Daniel Vetter Signed-off-by: Rodrigo Vivi Signed-off-by: Daniel Vetter diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index 9449cd6..a1168a8 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -3893,6 +3893,11 @@ int num_ioctls; !Idrivers/gpu/drm/i915/intel_audio.c + Panel Self Refresh PSR (PSR/SRD) +!Pdrivers/gpu/drm/i915/intel_psr.c Panel Self Refresh (PSR/SRD) +!Idrivers/gpu/drm/i915/intel_psr.c + + DPIO !Pdrivers/gpu/drm/i915/i915_reg.h DPIO
DRMGenericGeneric “EDID” BLOB | IMMUTABLE 0Contains DPMS operation mode value.
“PATH”BLOB | IMMUTABLE0ConnectorContains topology path to a connector.
Plane “type” ENUM | IMMUTABLEDescription/Restrictions
DRMDRM Generic “EDID” BLOB | IMMUTABLETBD
Virtual GPU“suggested X”RANGEMin=0, Max=0xffffffffConnectorproperty to suggest an X offset for a connector
“suggested Y”RANGEMin=0, Max=0xffffffffConnectorproperty to suggest an Y offset for a connector
Optional “scaling mode” ENUM
diff --git a/drivers/gpu/drm/i915/intel_psr.c b/drivers/gpu/drm/i915/intel_psr.c index 7b3ed91..716b8a9 100644 --- a/drivers/gpu/drm/i915/intel_psr.c +++ b/drivers/gpu/drm/i915/intel_psr.c @@ -21,6 +21,36 @@ * DEALINGS IN THE SOFTWARE. */ +/** + * DOC: Panel Self Refresh (PSR/SRD) + * + * Since Haswell Display controller supports Panel Self-Refresh on display + * panels witch have a remote frame buffer (RFB) implemented according to PSR + * spec in eDP1.3. PSR feature allows the display to go to lower standby states + * when system is idle but display is on as it eliminates display refresh + * request to DDR memory completely as long as the frame buffer for that + * display is unchanged. + * + * Panel Self Refresh must be supported by both Hardware (source) and + * Panel (sink). + * + * PSR saves power by caching the framebuffer in the panel RFB, which allows us + * to power down the link and memory controller. For DSI panels the same idea + * is called "manual mode". + * + * The implementation uses the hardware-based PSR support which automatically + * enters/exits self-refresh mode. The hardware takes care of sending the + * required DP aux message and could even retrain the link (that part isn't + * enabled yet though). The hardware also keeps track of any frontbuffer + * changes to know when to exit self-refresh mode again. Unfortunately that + * part doesn't work too well, hence why the i915 PSR support uses the + * software frontbuffer tracking to make sure it doesn't miss a screen + * update. For this integration intel_psr_invalidate() and intel_psr_flush() + * get called by the frontbuffer tracking code. Note that because of locking + * issues the self-refresh re-enable code is done from a work queue, which + * must be correctly synchronized/cancelled when shutting down the pipe." + */ + #include #include "intel_drv.h" @@ -217,6 +247,12 @@ static void intel_psr_do_enable(struct intel_dp *intel_dp) dev_priv->psr.active = true; } +/** + * intel_psr_enable - Enable PSR + * @intel_dp: Intel DP + * + * This function can only be called after the pipe is fully trained and enabled. + */ void intel_psr_enable(struct intel_dp *intel_dp) { struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); @@ -258,6 +294,12 @@ unlock: mutex_unlock(&dev_priv->psr.lock); } +/** + * intel_psr_disable - Disable PSR + * @intel_dp: Intel DP + * + * This function needs to be called before disabling pipe. + */ void intel_psr_disable(struct intel_dp *intel_dp) { struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); @@ -342,6 +384,18 @@ static void intel_psr_exit(struct drm_device *dev) } +/** + * intel_psr_invalidate - Invalidade PSR + * @dev: DRM device + * @frontbuffer_bits: frontbuffer plane tracking bits + * + * Since the hardware frontbuffer tracking has gaps we need to integrate + * with the software frontbuffer tracking. This function gets called every + * time frontbuffer rendering starts and a buffer gets dirtied. PSR must be + * disabled if the frontbuffer mask contains a buffer relevant to PSR. + * + * Dirty frontbuffers relevant to PSR are tracked in busy_frontbuffer_bits." + */ void intel_psr_invalidate(struct drm_device *dev, unsigned frontbuffer_bits) { @@ -366,6 +420,18 @@ void intel_psr_invalidate(struct drm_device *dev, mutex_unlock(&dev_priv->psr.lock); } +/** + * intel_psr_flush - Flush PSR + * @dev: DRM device + * @frontbuffer_bits: frontbuffer plane tracking bits + * + * Since the hardware frontbuffer tracking has gaps we need to integrate + * with the software frontbuffer tracking. This function gets called every + * time frontbuffer rendering has completed and flushed out to memory. PSR + * can be enabled again if no other frontbuffer relevant to PSR is dirty. + * + * Dirty frontbuffers relevant to PSR are tracked in busy_frontbuffer_bits. + */ void intel_psr_flush(struct drm_device *dev, unsigned frontbuffer_bits) { @@ -399,6 +465,13 @@ void intel_psr_flush(struct drm_device *dev, mutex_unlock(&dev_priv->psr.lock); } +/** + * intel_psr_init - Init basic PSR work and mutex. + * @dev: DRM device + * + * This function is called only once at driver load to initialize basic + * PSR stuff. + */ void intel_psr_init(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; -- cgit v0.10.2 From 5416d871136d7cc674c508889b8abbf2469dd115 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 14 Nov 2014 17:24:33 +0000 Subject: drm/i915/skl: Set the eDP link rate on DPLL0 On SKL DPLL0 is used to derive CDCLK but can also be used to drive an eDP port (as long as we don't want SSC). DPLL0 is special enough to not be handled by the shared DPLL framework (drives CDCLK, not supposed to enable the HDMI mode), So we need to compute the configuration separately from the other DPLLs. Note that we don't need to reprogram DPLL0 (which would mean bringing down CDCLK) to support the various eDP 1.3 link rates as they all share the same VCO (8100). Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 3e4c63c..83eeecb 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1484,6 +1484,25 @@ static void intel_ddi_pre_enable(struct intel_encoder *intel_encoder) uint32_t dpll = crtc->config.ddi_pll_sel; uint32_t val; + /* + * DPLL0 is used for eDP and is the only "private" DPLL (as + * opposed to shared) on SKL + */ + if (type == INTEL_OUTPUT_EDP) { + WARN_ON(dpll != SKL_DPLL0); + + val = I915_READ(DPLL_CTRL1); + + val &= ~(DPLL_CTRL1_HDMI_MODE(dpll) | + DPLL_CTRL1_SSC(dpll) | + DPLL_CRTL1_LINK_RATE_MASK(dpll)); + val |= crtc->config.dpll_hw_state.ctrl1 << (dpll * 6); + + I915_WRITE(DPLL_CTRL1, val); + POSTING_READ(DPLL_CTRL1); + } + + /* DDI -> PLL mapping */ val = I915_READ(DPLL_CTRL2); val &= ~(DPLL_CTRL2_DDI_CLK_OFF(port) | @@ -1492,6 +1511,7 @@ static void intel_ddi_pre_enable(struct intel_encoder *intel_encoder) DPLL_CTRL2_DDI_SEL_OVERRIDE(port)); I915_WRITE(DPLL_CTRL2, val); + } else { WARN_ON(crtc->config.ddi_pll_sel == PORT_CLK_SEL_NONE); I915_WRITE(PORT_CLK_SEL(port), crtc->config.ddi_pll_sel); diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 7369e6f..4e49caa 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -1074,6 +1074,33 @@ intel_dp_connector_unregister(struct intel_connector *intel_connector) } static void +skl_edp_set_pll_config(struct intel_crtc_config *pipe_config, int link_bw) +{ + u32 ctrl1; + + pipe_config->ddi_pll_sel = SKL_DPLL0; + pipe_config->dpll_hw_state.cfgcr1 = 0; + pipe_config->dpll_hw_state.cfgcr2 = 0; + + ctrl1 = DPLL_CTRL1_OVERRIDE(SKL_DPLL0); + switch (link_bw) { + case DP_LINK_BW_1_62: + ctrl1 |= DPLL_CRTL1_LINK_RATE(DPLL_CRTL1_LINK_RATE_810, + SKL_DPLL0); + break; + case DP_LINK_BW_2_7: + ctrl1 |= DPLL_CRTL1_LINK_RATE(DPLL_CRTL1_LINK_RATE_1350, + SKL_DPLL0); + break; + case DP_LINK_BW_5_4: + ctrl1 |= DPLL_CRTL1_LINK_RATE(DPLL_CRTL1_LINK_RATE_2700, + SKL_DPLL0); + break; + } + pipe_config->dpll_hw_state.ctrl1 = ctrl1; +} + +static void hsw_dp_set_ddi_pll_sel(struct intel_crtc_config *pipe_config, int link_bw) { switch (link_bw) { @@ -1250,7 +1277,9 @@ found: &pipe_config->dp_m2_n2); } - if (IS_HASWELL(dev) || IS_BROADWELL(dev)) + if (IS_SKYLAKE(dev) && is_edp(intel_dp)) + skl_edp_set_pll_config(pipe_config, intel_dp->link_bw); + else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) hsw_dp_set_ddi_pll_sel(pipe_config, intel_dp->link_bw); else intel_dp_set_clock(encoder, pipe_config, intel_dp->link_bw); -- cgit v0.10.2 From 134ffa44d1ad244216a114506a67dcfc7c80a90f Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 14 Nov 2014 17:24:34 +0000 Subject: drm/i915/skl: Use the pipe config DPLL tracking to query the link clock Signed-off-by: Damien Lespiau Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 83eeecb..86745da 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -735,14 +735,10 @@ static void skl_ddi_clock_get(struct intel_encoder *encoder, struct intel_crtc_config *pipe_config) { struct drm_i915_private *dev_priv = encoder->base.dev->dev_private; - enum port port = intel_ddi_get_encoder_port(encoder); int link_clock = 0; uint32_t dpll_ctl1, dpll; - /* FIXME: This should be tracked in the pipe config. */ - dpll = I915_READ(DPLL_CTRL2); - dpll &= DPLL_CTRL2_DDI_CLK_SEL_MASK(port); - dpll >>= DPLL_CTRL2_DDI_CLK_SEL_SHIFT(port); + dpll = pipe_config->ddi_pll_sel; dpll_ctl1 = I915_READ(DPLL_CTRL1); -- cgit v0.10.2 From 6ac0483b35bfc7546398d1198b65ed7be17f3cc5 Mon Sep 17 00:00:00 2001 From: Matt Roper Date: Mon, 17 Nov 2014 09:59:28 -0800 Subject: drm/i915: Propagate invalid setcrtc cloning errors back to userspace MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When invalid cloning configurations were detected during modeset, we never copied the error code into the return value variable, leading us to return 0 (success) to userspace. This regression has been introduced in commit 50f5275698df4490046cc5b4ed2018abb642a803 Author: Jesse Barnes Date: Fri Nov 7 13:11:00 2014 -0800 drm/i915: use compute_config in set_config v4 Testcase: igt/kms_setmode Signed-off-by: Matt Roper Reviewed-by: Ville Syrjälä Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86226 Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 2632f2a..f84738d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11469,6 +11469,7 @@ static int intel_crtc_set_config(struct drm_mode_set *set) &prepare_pipes, &disable_pipes); if (IS_ERR(pipe_config)) { + ret = PTR_ERR(pipe_config); goto fail; } else if (pipe_config) { if (to_intel_crtc(set->crtc)->new_config->has_audio != -- cgit v0.10.2 From 70b0ea86568fc91e29742010c5dcf4b5b5661939 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 18 Nov 2014 09:09:32 +0100 Subject: drm/i915: Drop return value from lrc_setup_hardware_status_page kmap never fails. Spotted-by: Chris Wilson Cc: Chris Wilson Cc: Arun Siluvery Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index b7c4c9a..3cf15c4 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1706,7 +1706,7 @@ static uint32_t get_lr_context_size(struct intel_engine_cs *ring) return ret; } -static int lrc_setup_hardware_status_page(struct intel_engine_cs *ring, +static void lrc_setup_hardware_status_page(struct intel_engine_cs *ring, struct drm_i915_gem_object *default_ctx_obj) { struct drm_i915_private *dev_priv = ring->dev->dev_private; @@ -1716,15 +1716,11 @@ static int lrc_setup_hardware_status_page(struct intel_engine_cs *ring, ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(default_ctx_obj); ring->status_page.page_addr = kmap(sg_page(default_ctx_obj->pages->sgl)); - if (ring->status_page.page_addr == NULL) - return -ENOMEM; ring->status_page.obj = default_ctx_obj; I915_WRITE(RING_HWS_PGA(ring->mmio_base), (u32)ring->status_page.gfx_addr); POSTING_READ(RING_HWS_PGA(ring->mmio_base)); - - return 0; } /** @@ -1811,13 +1807,8 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, ctx->engine[ring->id].ringbuf = ringbuf; ctx->engine[ring->id].state = ctx_obj; - if (ctx == ring->default_context) { - ret = lrc_setup_hardware_status_page(ring, ctx_obj); - if (ret) { - DRM_ERROR("Failed to setup hardware status page\n"); - goto error; - } - } + if (ctx == ring->default_context) + lrc_setup_hardware_status_page(ring, ctx_obj); if (ring->id == RCS && !ctx->rcs_initialized) { if (ring->init_context) { -- cgit v0.10.2 From cbc572a9a5a90aac54ff1a5c7356b6f6bda78264 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Mon, 17 Nov 2014 13:08:47 -0800 Subject: drm/i915/ddi: add break in DDI mode select switch The lack of a break here wasn't for falling through to some other important code, so made me do a double take. Add a break just to make things a little less confusing. Signed-off-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 86745da..07c5625 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2062,6 +2062,7 @@ void intel_ddi_get_config(struct intel_encoder *encoder, switch (temp & TRANS_DDI_MODE_SELECT_MASK) { case TRANS_DDI_MODE_SELECT_HDMI: pipe_config->has_hdmi_sink = true; + break; case TRANS_DDI_MODE_SELECT_DVI: case TRANS_DDI_MODE_SELECT_FDI: break; -- cgit v0.10.2 From eb45fa0b93e03b03848cd048dcc57648409c8125 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Tue, 18 Nov 2014 12:11:29 +0200 Subject: drm/i915/audio: fix monitor presence indication after disable Indicate the monitor has been disconnected on disable. The regression has been introduced in commit 5fad84a7530f8e7664cdc6f490cb90653fed1266 Author: Jani Nikula Date: Tue Nov 4 10:30:23 2014 +0200 drm/i915: rewrite hsw/bdw audio codec enable/disable sequences Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86424 Cc: Rodrigo Vivi Signed-off-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 87750ef..2c7ed5c 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -194,6 +194,7 @@ static void hsw_audio_codec_disable(struct intel_encoder *encoder) /* Invalidate ELD */ tmp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); tmp &= ~AUDIO_ELD_VALID(pipe); + tmp &= ~AUDIO_OUTPUT_ENABLE(pipe); I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); } -- cgit v0.10.2 From 77c1aa84de0096792de673aa1c64c36b38553cf5 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 18 Nov 2014 13:27:07 +0100 Subject: drm/i915: Don't print header in error state for non-existing CS This goes back to commit 362b8af7ad1d91266aa4931e62be45c1e5cf753b Author: Ben Widawsky Date: Thu Jan 30 00:19:38 2014 -0800 drm/i915: Move per ring error state to ring_error Spotted while reading error states. Cc: Ben Widawsky Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 89a2f3d..82111b8 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -242,11 +242,15 @@ static const char *hangcheck_action_to_str(enum intel_ring_hangcheck_action a) static void i915_ring_error_state(struct drm_i915_error_state_buf *m, struct drm_device *dev, - struct drm_i915_error_ring *ring) + struct drm_i915_error_state *error, + int ring_idx) { + struct drm_i915_error_ring *ring = &error->ring[ring_idx]; + if (!ring->valid) return; + err_printf(m, "%s command stream:\n", ring_str(ring_idx)); err_printf(m, " HEAD: 0x%08x\n", ring->head); err_printf(m, " TAIL: 0x%08x\n", ring->tail); err_printf(m, " CTL: 0x%08x\n", ring->ctl); @@ -388,10 +392,8 @@ int i915_error_state_to_str(struct drm_i915_error_state_buf *m, if (INTEL_INFO(dev)->gen == 7) err_printf(m, "ERR_INT: 0x%08x\n", error->err_int); - for (i = 0; i < ARRAY_SIZE(error->ring); i++) { - err_printf(m, "%s command stream:\n", ring_str(i)); - i915_ring_error_state(m, dev, &error->ring[i]); - } + for (i = 0; i < ARRAY_SIZE(error->ring); i++) + i915_ring_error_state(m, dev, error, i); for (i = 0; i < error->vm_count; i++) { err_printf(m, "vm[%d]\n", i); -- cgit v0.10.2 From 672e7b7c1849c904b2c55185906b3940843c55c6 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 19 Nov 2014 09:47:19 +0000 Subject: drm/i915: Don't continually defer the hangcheck With multiple rings, we may continue to render on the blitter whilst executing an infinite shader on the render ring. As we currently, rearm the timer with each execbuf, in this scenario the hangcheck will never fire and we will never detect the lockup on the render ring. Instead, only arm the timer once per hangcheck, so that hangcheck runs more frequently. v2: Rearrange code to avoid triggering a BUG_ON in add_timer from softirq context. Testcase: igt/gem_reset_stats/defer-hangcheck* Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86225 Signed-off-by: Chris Wilson Cc: Mika Kuoppala Reviewed-by: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 5e5d3f7..e298518 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -988,7 +988,6 @@ static void notify_ring(struct drm_device *dev, trace_i915_gem_request_complete(ring); wake_up_all(&ring->irq_queue); - i915_queue_hangcheck(dev); } static u32 vlv_c0_residency(struct drm_i915_private *dev_priv, @@ -3041,11 +3040,15 @@ static void i915_hangcheck_elapsed(unsigned long data) void i915_queue_hangcheck(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; + struct timer_list *timer = &dev_priv->gpu_error.hangcheck_timer; + if (!i915.enable_hangcheck) return; - mod_timer(&dev_priv->gpu_error.hangcheck_timer, - round_jiffies_up(jiffies + DRM_I915_HANGCHECK_JIFFIES)); + /* Don't continually defer the hangcheck, but make sure it is active */ + if (!timer_pending(timer)) + timer->expires = round_jiffies_up(jiffies + DRM_I915_HANGCHECK_JIFFIES); + mod_timer(timer, timer->expires); } static void ibx_irq_reset(struct drm_device *dev) -- cgit v0.10.2 From 90bd1f46caa9ad0d51ec48c91f3c3a616d6d6b8c Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 11:18:47 +0100 Subject: drm/i915: Tune down sink crc timeout dmesg output For whatever reasons this can happen. For real testcases the test will notice the -EIO and fall over, but we also have some testcases that just read all debugfs files. And that shouldn't cause dmesg spam. So tune it down a bit so that we still have the information for debugging. And change the errno so that real testcases can easily differentiate. Cc: Rodrigo Vivi Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84890 Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 4e49caa..5654790 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3693,8 +3693,8 @@ int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc) } while (--attempts && (buf & DP_TEST_COUNT_MASK) == test_crc_count); if (attempts == 0) { - DRM_ERROR("Panel is unable to calculate CRC after 6 vblanks\n"); - return -EIO; + DRM_DEBUG_KMS("Panel is unable to calculate CRC after 6 vblanks\n"); + return -ETIMEDOUT; } if (drm_dp_dpcd_read(&intel_dp->aux, DP_TEST_CRC_R_CR, crc, 6) < 0) -- cgit v0.10.2 From 151a49d0792203ab735020b5777bb636420a2a11 Mon Sep 17 00:00:00 2001 From: Tom O'Rourke Date: Thu, 13 Nov 2014 18:50:10 -0800 Subject: drm/i915: Extend pcode mailbox interface MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In sandybridge_pcode_read and sandybridge_pcode_write, extend the mbox parameter from u8 to u32. On Haswell and Sandybridge, bits 7:0 encode the mailbox command and bits 28:8 are used for address control for specific commands. Based on suggestion from Ville Syrjälä. Signed-off-by: Tom O'Rourke Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index c4f2cb6..a8cfb14 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2975,8 +2975,8 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine); void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine); void assert_force_wake_inactive(struct drm_i915_private *dev_priv); -int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u8 mbox, u32 *val); -int sandybridge_pcode_write(struct drm_i915_private *dev_priv, u8 mbox, u32 val); +int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u32 mbox, u32 *val); +int sandybridge_pcode_write(struct drm_i915_private *dev_priv, u32 mbox, u32 val); /* intel_sideband.c */ u32 vlv_punit_read(struct drm_i915_private *dev_priv, u8 addr); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 929648f..4e39d74 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7158,7 +7158,7 @@ void intel_init_pm(struct drm_device *dev) } } -int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u8 mbox, u32 *val) +int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u32 mbox, u32 *val) { WARN_ON(!mutex_is_locked(&dev_priv->rps.hw_lock)); @@ -7183,7 +7183,7 @@ int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u8 mbox, u32 *val) return 0; } -int sandybridge_pcode_write(struct drm_i915_private *dev_priv, u8 mbox, u32 val) +int sandybridge_pcode_write(struct drm_i915_private *dev_priv, u32 mbox, u32 val) { WARN_ON(!mutex_is_locked(&dev_priv->rps.hw_lock)); -- cgit v0.10.2 From f061b9be74dbb83f383daa7c95bfa22ae011a81e Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Tue, 18 Nov 2014 09:45:52 -0800 Subject: drm/i915/ddi: set has_infoframe flag on DDI too v2 Just like we do in the HDMI code, set the infoframe flag if we detect that infoframes are enabled. v2: check for actual infoframe status as in hdmi code (Daniel) Signed-off-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 07c5625..24110c9 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2075,6 +2075,14 @@ void intel_ddi_get_config(struct intel_encoder *encoder, break; } + if (encoder->type == INTEL_OUTPUT_HDMI) { + struct intel_hdmi *intel_hdmi = + enc_to_intel_hdmi(&encoder->base); + + if (intel_hdmi->infoframe_enabled(&encoder->base)) + pipe_config->has_infoframe = true; + } + if (intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_AUDIO)) { temp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); if (temp & AUDIO_OUTPUT_ENABLE(intel_crtc->pipe)) -- cgit v0.10.2 From 9dc806fc1a9445c721aacf15a141c2758e518e90 Mon Sep 17 00:00:00 2001 From: Matt Roper Date: Mon, 17 Nov 2014 18:10:38 -0800 Subject: drm/i915: Don't store panning coordinates as 16.16 fixed point MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When using the universal plane interface, the source rectangle coordinates define the panning offset for the primary plane, which needs to be stored in crtc->{x,y}. The original universal plane code negelected to set these panning offset fields, which was partially remedied in: commit ccc759dc2a0214fd8b65ed4ebe78050874a67f94 Author: Gustavo Padovan Date: Wed Sep 24 14:20:22 2014 -0300 drm/i915: Merge of visible and !visible paths for primary planes However the plane source coordinates are provided in 16.16 fixed point format and the above commit forgot to convert back to integer coordinates before saving the values. When we replace intel_pipe_set_base() with plane->funcs->update_plane() in a future patch, this bug becomes visible via the set_config entrypoint as well as update_plane. Cc: Gustavo Padovan Testcase: igt/kms_plane Signed-off-by: Matt Roper Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index f84738d..32cec9d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11765,8 +11765,8 @@ intel_commit_primary_plane(struct drm_plane *plane, struct drm_rect *src = &state->src; crtc->primary->fb = fb; - crtc->x = src->x1; - crtc->y = src->y1; + crtc->x = src->x1 >> 16; + crtc->y = src->y1 >> 16; intel_plane->crtc_x = state->orig_dst.x1; intel_plane->crtc_y = state->orig_dst.y1; -- cgit v0.10.2 From 4a74de821af510fcb3224d16ceed6a6609730d9c Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 19 Nov 2014 15:30:01 +0200 Subject: drm/i915: WARN if we receive any rps interrupts on gen>9 This extends commit 132f3f1767dbabfb01f3c9bd63098c65d91eeac9 Author: Imre Deak Date: Mon Nov 10 15:34:33 2014 +0200 drm/i915: WARN if we receive any gen9 rps interrupts to GEN>9 platforms as suggested by Paulo. Signed-off-by: Imre Deak Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index e298518..21c579e 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1689,9 +1689,9 @@ static void i9xx_pipe_crc_irq_handler(struct drm_device *dev, enum pipe pipe) * the work queue. */ static void gen6_rps_irq_handler(struct drm_i915_private *dev_priv, u32 pm_iir) { - /* TODO: RPS on GEN9 is not supported yet. */ - if (WARN_ONCE(INTEL_INFO(dev_priv)->gen == 9, - "GEN9: unexpected RPS IRQ\n")) + /* TODO: RPS on GEN9+ is not supported yet. */ + if (WARN_ONCE(INTEL_INFO(dev_priv)->gen >= 9, + "GEN9+: unexpected RPS IRQ\n")) return; if (pm_iir & dev_priv->pm_rps_events) { -- cgit v0.10.2 From e534770addd798e17b95372b4cd61682d1bf9296 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 19 Nov 2014 15:30:02 +0200 Subject: drm/i915: move rps irq disable one level up We disable the RPS interrupts for all platforms at the same spot, so move it one level up in the callstack to simplify things. No functional change. v2: - rebase on the GEN9 patches where RPS isn't supported yet, so we don't need to disable RPS interrupts on it (Paulo) v3: - avoid disabling the interrupts on GEN>9 too (Paulo) Signed-off-by: Imre Deak Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 4e39d74..cdd00f1 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4526,8 +4526,6 @@ static void gen6_disable_rps(struct drm_device *dev) I915_WRITE(GEN6_RC_CONTROL, 0); I915_WRITE(GEN6_RPNSWREQ, 1 << 31); - - gen6_disable_rps_interrupts(dev); } static void cherryview_disable_rps(struct drm_device *dev) @@ -4535,8 +4533,6 @@ static void cherryview_disable_rps(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; I915_WRITE(GEN6_RC_CONTROL, 0); - - gen6_disable_rps_interrupts(dev); } static void valleyview_disable_rps(struct drm_device *dev) @@ -4550,8 +4546,6 @@ static void valleyview_disable_rps(struct drm_device *dev) I915_WRITE(GEN6_RC_CONTROL, 0); gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL); - - gen6_disable_rps_interrupts(dev); } static void intel_print_rc6_info(struct drm_device *dev, u32 mode) @@ -6230,6 +6224,14 @@ void intel_disable_gt_powersave(struct drm_device *dev) valleyview_disable_rps(dev); else gen6_disable_rps(dev); + + /* + * TODO: disable RPS interrupts on GEN9+ too once RPS support + * is added for it. + */ + if (INTEL_INFO(dev)->gen < 9) + gen6_disable_rps_interrupts(dev); + dev_priv->rps.enabled = false; mutex_unlock(&dev_priv->rps.hw_lock); } -- cgit v0.10.2 From 3cc134e3ee09055d5a87193fc7eb0ecf4a59eaa1 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 19 Nov 2014 15:30:03 +0200 Subject: drm/i915: sanitize rps irq enabling Atm we first enable the RPS interrupts then we clear any pending ones. By this we could lose an interrupt arriving after we unmasked it. This may not be a problem as the caller should handle such a race, but logic still calls for the opposite order. Also we can delay enabling the interrupts until after all the RPS initialization is ready with the following order: 1. disable left-over RPS (earlier via intel_uncore_sanitize) 2. clear any pending RPS interrupts 3. initialize RPS 4. enable RPS interrupts This also allows us to do the 2. and 4. step the same way for all platforms, so let's follow this order to simplifying things. Also make sure any queued interrupts are also cleared. v2: - rebase on the GEN9 patches where we don't support RPS yet, so we musn't enable RPS interrupts on it (Paulo) v3: - avoid enabling RPS interrupts on GEN>9 too (Paulo) - clarify the RPS init sequence in the log message (Chris) - add POSTING_READ to gen6_reset_rps_interrupts() (Paulo) - WARN if any PM_IIR bits are set in gen6_enable_rps_interrupts() (Paulo) Signed-off-by: Imre Deak Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 21c579e..56b3053 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -255,14 +255,26 @@ void gen6_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask) snb_update_pm_irq(dev_priv, mask, 0); } +void gen6_reset_rps_interrupts(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + uint32_t reg = gen6_pm_iir(dev_priv); + + spin_lock_irq(&dev_priv->irq_lock); + I915_WRITE(reg, dev_priv->pm_rps_events); + I915_WRITE(reg, dev_priv->pm_rps_events); + POSTING_READ(reg); + spin_unlock_irq(&dev_priv->irq_lock); +} + void gen6_enable_rps_interrupts(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; spin_lock_irq(&dev_priv->irq_lock); WARN_ON(dev_priv->rps.pm_iir); + WARN_ON(I915_READ(gen6_pm_iir(dev_priv)) & dev_priv->pm_rps_events); gen6_enable_pm_irq(dev_priv, dev_priv->pm_rps_events); - I915_WRITE(gen6_pm_iir(dev_priv), dev_priv->pm_rps_events); spin_unlock_irq(&dev_priv->irq_lock); } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index d1f9b63..f0a46ec 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -791,6 +791,7 @@ void gen5_enable_gt_irq(struct drm_i915_private *dev_priv, uint32_t mask); void gen5_disable_gt_irq(struct drm_i915_private *dev_priv, uint32_t mask); void gen6_enable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask); void gen6_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask); +void gen6_reset_rps_interrupts(struct drm_device *dev); void gen6_enable_rps_interrupts(struct drm_device *dev); void gen6_disable_rps_interrupts(struct drm_device *dev); void intel_runtime_pm_disable_interrupts(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index cdd00f1..c5b2636 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4747,8 +4747,6 @@ static void gen8_enable_rps(struct drm_device *dev) gen6_set_rps(dev, (I915_READ(GEN6_GT_PERF_STATUS) & 0xff00) >> 8); - gen6_enable_rps_interrupts(dev); - gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL); } @@ -4845,8 +4843,6 @@ static void gen6_enable_rps(struct drm_device *dev) dev_priv->rps.power = HIGH_POWER; /* force a reset */ gen6_set_rps(dev_priv->dev, dev_priv->rps.min_freq_softlimit); - gen6_enable_rps_interrupts(dev); - rc6vids = 0; ret = sandybridge_pcode_read(dev_priv, GEN6_PCODE_READ_RC6VIDS, &rc6vids); if (IS_GEN6(dev) && ret) { @@ -5348,8 +5344,6 @@ static void cherryview_enable_rps(struct drm_device *dev) valleyview_set_rps(dev_priv->dev, dev_priv->rps.efficient_freq); - gen6_enable_rps_interrupts(dev); - gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL); } @@ -5431,8 +5425,6 @@ static void valleyview_enable_rps(struct drm_device *dev) valleyview_set_rps(dev_priv->dev, dev_priv->rps.efficient_freq); - gen6_enable_rps_interrupts(dev); - gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL); } @@ -6246,6 +6238,13 @@ static void intel_gen6_powersave_work(struct work_struct *work) mutex_lock(&dev_priv->rps.hw_lock); + /* + * TODO: reset/enable RPS interrupts on GEN9+ too, once RPS support is + * added for it. + */ + if (INTEL_INFO(dev)->gen < 9) + gen6_reset_rps_interrupts(dev); + if (IS_CHERRYVIEW(dev)) { cherryview_enable_rps(dev); } else if (IS_VALLEYVIEW(dev)) { @@ -6260,6 +6259,10 @@ static void intel_gen6_powersave_work(struct work_struct *work) __gen6_update_ring_freq(dev); } dev_priv->rps.enabled = true; + + if (INTEL_INFO(dev)->gen < 9) + gen6_enable_rps_interrupts(dev); + mutex_unlock(&dev_priv->rps.hw_lock); intel_runtime_pm_put(dev_priv); -- cgit v0.10.2 From d4d70aa5960a7fbf3d887663f144c324a10619ba Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 19 Nov 2014 15:30:04 +0200 Subject: drm/i915: sanitize rps irq disabling When disabling the RPS interrupts there is a tricky dependency between the thread disabling the interrupts, the RPS interrupt handler and the corresponding RPS work. The RPS work can reenable the interrupts, so there is no straightforward order in the disabling thread to (1) make sure that any RPS work is flushed and to (2) disable all RPS interrupts. Currently this is solved by masking the interrupts using two separate mask registers (first level display IMR and PM IMR) and doing the disabling when all first level interrupts are disabled. This works, but the requirement to run with all first level interrupts disabled is unnecessary making the suspend / unload time ordering of RPS disabling wrt. other unitialization steps difficult and error prone. Removing this restriction allows us to disable RPS early during suspend / unload and forget about it for the rest of the sequence. By adding a more explicit method for avoiding the above race, it also becomes easier to prove its correctness. Finally currently we can hit the WARN in snb_update_pm_irq(), when a final RPS work runs with the first level interrupts already disabled. This won't lead to any problem (due to the separate interrupt masks), but with the change in this and the next patch we can get rid of the WARN, while leaving it in place for other scenarios. To address the above points, add a new RPS interrupts_enabled flag and use this during RPS disabling to avoid requeuing the RPS work and reenabling of the RPS interrupts. Since the interrupt disabling happens now in intel_suspend_gt_powersave(), we will disable RPS interrupts explicitly during suspend (and not just through the first level mask), but there is no problem doing so, it's also more consistent and allows us to unify more of the RPS disabling during suspend and unload time in the next patch. v2/v3: - rebase on patch "drm/i915: move rps irq disable one level up" in the patchset Signed-off-by: Imre Deak Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index a8cfb14..4f08160 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -998,8 +998,12 @@ struct intel_rps_ei { }; struct intel_gen6_power_mgmt { - /* work and pm_iir are protected by dev_priv->irq_lock */ + /* + * work, interrupts_enabled and pm_iir are protected by + * dev_priv->irq_lock + */ struct work_struct work; + bool interrupts_enabled; u32 pm_iir; /* Frequencies are stored in potentially platform dependent multiples. diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 56b3053..283756f 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -274,6 +274,7 @@ void gen6_enable_rps_interrupts(struct drm_device *dev) spin_lock_irq(&dev_priv->irq_lock); WARN_ON(dev_priv->rps.pm_iir); WARN_ON(I915_READ(gen6_pm_iir(dev_priv)) & dev_priv->pm_rps_events); + dev_priv->rps.interrupts_enabled = true; gen6_enable_pm_irq(dev_priv, dev_priv->pm_rps_events); spin_unlock_irq(&dev_priv->irq_lock); } @@ -282,14 +283,16 @@ void gen6_disable_rps_interrupts(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; + spin_lock_irq(&dev_priv->irq_lock); + dev_priv->rps.interrupts_enabled = false; + spin_unlock_irq(&dev_priv->irq_lock); + + cancel_work_sync(&dev_priv->rps.work); + I915_WRITE(GEN6_PMINTRMSK, INTEL_INFO(dev_priv)->gen >= 8 ? ~GEN8_PMINTR_REDIRECT_TO_NON_DISP : ~0); I915_WRITE(gen6_pm_ier(dev_priv), I915_READ(gen6_pm_ier(dev_priv)) & ~dev_priv->pm_rps_events); - /* Complete PM interrupt masking here doesn't race with the rps work - * item again unmasking PM interrupts because that is using a different - * register (PMIMR) to mask PM interrupts. The only risk is in leaving - * stale bits in PMIIR and PMIMR which gen6_enable_rps will clean up. */ spin_lock_irq(&dev_priv->irq_lock); dev_priv->rps.pm_iir = 0; @@ -1135,6 +1138,11 @@ static void gen6_pm_rps_work(struct work_struct *work) int new_delay, adj; spin_lock_irq(&dev_priv->irq_lock); + /* Speed up work cancelation during disabling rps interrupts. */ + if (!dev_priv->rps.interrupts_enabled) { + spin_unlock_irq(&dev_priv->irq_lock); + return; + } pm_iir = dev_priv->rps.pm_iir; dev_priv->rps.pm_iir = 0; /* Make sure not to corrupt PMIMR state used by ringbuffer on GEN6 */ @@ -1708,11 +1716,12 @@ static void gen6_rps_irq_handler(struct drm_i915_private *dev_priv, u32 pm_iir) if (pm_iir & dev_priv->pm_rps_events) { spin_lock(&dev_priv->irq_lock); - dev_priv->rps.pm_iir |= pm_iir & dev_priv->pm_rps_events; gen6_disable_pm_irq(dev_priv, pm_iir & dev_priv->pm_rps_events); + if (dev_priv->rps.interrupts_enabled) { + dev_priv->rps.pm_iir |= pm_iir & dev_priv->pm_rps_events; + queue_work(dev_priv->wq, &dev_priv->rps.work); + } spin_unlock(&dev_priv->irq_lock); - - queue_work(dev_priv->wq, &dev_priv->rps.work); } if (INTEL_INFO(dev_priv)->gen >= 8) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index c5b2636..96a948a 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6186,9 +6186,17 @@ void intel_suspend_gt_powersave(struct drm_device *dev) /* Interrupts should be disabled already to avoid re-arming. */ WARN_ON(intel_irqs_enabled(dev_priv)); + if (INTEL_INFO(dev)->gen < 6) + return; + flush_delayed_work(&dev_priv->rps.delayed_resume_work); - cancel_work_sync(&dev_priv->rps.work); + /* + * TODO: disable RPS interrupts on GEN9+ too once RPS support + * is added for it. + */ + if (INTEL_INFO(dev)->gen < 9) + gen6_disable_rps_interrupts(dev); /* Force GPU to min freq during suspend */ gen6_rps_idle(dev_priv); @@ -6217,13 +6225,6 @@ void intel_disable_gt_powersave(struct drm_device *dev) else gen6_disable_rps(dev); - /* - * TODO: disable RPS interrupts on GEN9+ too once RPS support - * is added for it. - */ - if (INTEL_INFO(dev)->gen < 9) - gen6_disable_rps_interrupts(dev); - dev_priv->rps.enabled = false; mutex_unlock(&dev_priv->rps.hw_lock); } -- cgit v0.10.2 From 2eb5252e2fffc52745a672152c7df597f4041045 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 19 Nov 2014 15:30:05 +0200 Subject: drm/i915: disable rps irqs earlier during suspend/unload After the previous patch RPS disabling doesn't depend any more on the first level interrupts being disabled, so we can move it everywhere earlier. Doing so let's us think about the uninitialization steps afterwards independently of any asynchronous RPS events that can happen atm. It also makes the system/runtime suspend time RPS disabling more uniform. Finally this gets rid of the WARN in intel_suspend_gt_powersave(), which we can hit if a final RPS work runs after we disabled the first level interrupts. Testcase: igt/pm_rpm Reference: https://bugs.freedesktop.org/show_bug.cgi?id=82939 Signed-off-by: Imre Deak Reviewed-by: Paulo Zanoni Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 122ea95..19558d6 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -584,6 +584,8 @@ static int i915_drm_suspend(struct drm_device *dev) return error; } + intel_suspend_gt_powersave(dev); + /* * Disable CRTCs directly since we want to preserve sw state * for _thaw. Also, power gate the CRTC power wells. @@ -595,15 +597,11 @@ static int i915_drm_suspend(struct drm_device *dev) intel_dp_mst_suspend(dev); - flush_delayed_work(&dev_priv->rps.delayed_resume_work); - intel_runtime_pm_disable_interrupts(dev_priv); intel_hpd_cancel_work(dev_priv); intel_suspend_encoders(dev_priv); - intel_suspend_gt_powersave(dev); - intel_suspend_hw(dev); } @@ -1393,9 +1391,8 @@ static int intel_runtime_suspend(struct device *device) i915_gem_release_all_mmaps(dev_priv); mutex_unlock(&dev->struct_mutex); - flush_delayed_work(&dev_priv->rps.delayed_resume_work); - intel_runtime_pm_disable_interrupts(dev_priv); intel_suspend_gt_powersave(dev); + intel_runtime_pm_disable_interrupts(dev_priv); ret = intel_suspend_complete(dev_priv); if (ret) { diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 32cec9d..ca4b787 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -13540,11 +13540,13 @@ void intel_modeset_cleanup(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; struct drm_connector *connector; + intel_disable_gt_powersave(dev); + intel_backlight_unregister(dev); /* * Interrupts and polling as the first thing to avoid creating havoc. - * Too much stuff here (turning of rps, connectors, ...) would + * Too much stuff here (turning of connectors, ...) would * experience fancy races otherwise. */ intel_irq_uninstall(dev_priv); @@ -13561,8 +13563,6 @@ void intel_modeset_cleanup(struct drm_device *dev) intel_disable_fbc(dev); - intel_disable_gt_powersave(dev); - ironlake_teardown_rc6(dev); mutex_unlock(&dev->struct_mutex); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 96a948a..41c51fe 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6183,9 +6183,6 @@ void intel_suspend_gt_powersave(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - /* Interrupts should be disabled already to avoid re-arming. */ - WARN_ON(intel_irqs_enabled(dev_priv)); - if (INTEL_INFO(dev)->gen < 6) return; -- cgit v0.10.2 From c86ee3a9f8cddcf2e637da19d6e7c05bdea11a96 Mon Sep 17 00:00:00 2001 From: Thomas Daniel Date: Thu, 13 Nov 2014 10:27:05 +0000 Subject: drm/i915/bdw: Clean up execlist queue items in retire_work No longer create a work item to clean each execlist queue item. Instead, move retired execlist requests to a queue and clean up the items during retire_requests. v2: Fix legacy ring path broken during overzealous cleanup v3: Update idle detection to take execlists queue into account v4: Grab execlist lock when checking queue state v5: Fix leaking requests by freeing in execlists_retire_requests. Issue: VIZ-4274 Signed-off-by: Thomas Daniel Reviewed-by: Deepak S Reviewed-by: Akash Goel Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 86cf428..2e85ef1 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2800,6 +2800,15 @@ i915_gem_retire_requests(struct drm_device *dev) for_each_ring(ring, dev_priv, i) { i915_gem_retire_requests_ring(ring); idle &= list_empty(&ring->request_list); + if (i915.enable_execlists) { + unsigned long flags; + + spin_lock_irqsave(&ring->execlist_lock, flags); + idle &= list_empty(&ring->execlist_queue); + spin_unlock_irqrestore(&ring->execlist_lock, flags); + + intel_execlists_retire_requests(ring); + } } if (idle) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 3cf15c4..c855051 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -399,7 +399,6 @@ static void execlists_context_unqueue(struct intel_engine_cs *ring) { struct intel_ctx_submit_request *req0 = NULL, *req1 = NULL; struct intel_ctx_submit_request *cursor = NULL, *tmp = NULL; - struct drm_i915_private *dev_priv = ring->dev->dev_private; assert_spin_locked(&ring->execlist_lock); @@ -416,7 +415,8 @@ static void execlists_context_unqueue(struct intel_engine_cs *ring) * will update tail past first request's workload */ cursor->elsp_submitted = req0->elsp_submitted; list_del(&req0->execlist_link); - queue_work(dev_priv->wq, &req0->work); + list_add_tail(&req0->execlist_link, + &ring->execlist_retired_req_list); req0 = cursor; } else { req1 = cursor; @@ -438,7 +438,6 @@ static void execlists_context_unqueue(struct intel_engine_cs *ring) static bool execlists_check_remove_request(struct intel_engine_cs *ring, u32 request_id) { - struct drm_i915_private *dev_priv = ring->dev->dev_private; struct intel_ctx_submit_request *head_req; assert_spin_locked(&ring->execlist_lock); @@ -456,7 +455,8 @@ static bool execlists_check_remove_request(struct intel_engine_cs *ring, if (--head_req->elsp_submitted <= 0) { list_del(&head_req->execlist_link); - queue_work(dev_priv->wq, &head_req->work); + list_add_tail(&head_req->execlist_link, + &ring->execlist_retired_req_list); return true; } } @@ -525,22 +525,6 @@ void intel_execlists_handle_ctx_events(struct intel_engine_cs *ring) ((u32)ring->next_context_status_buffer & 0x07) << 8); } -static void execlists_free_request_task(struct work_struct *work) -{ - struct intel_ctx_submit_request *req = - container_of(work, struct intel_ctx_submit_request, work); - struct drm_device *dev = req->ring->dev; - struct drm_i915_private *dev_priv = dev->dev_private; - - intel_runtime_pm_put(dev_priv); - - mutex_lock(&dev->struct_mutex); - i915_gem_context_unreference(req->ctx); - mutex_unlock(&dev->struct_mutex); - - kfree(req); -} - static int execlists_context_queue(struct intel_engine_cs *ring, struct intel_context *to, u32 tail) @@ -557,7 +541,6 @@ static int execlists_context_queue(struct intel_engine_cs *ring, i915_gem_context_reference(req->ctx); req->ring = ring; req->tail = tail; - INIT_WORK(&req->work, execlists_free_request_task); intel_runtime_pm_get(dev_priv); @@ -578,7 +561,8 @@ static int execlists_context_queue(struct intel_engine_cs *ring, WARN(tail_req->elsp_submitted != 0, "More than 2 already-submitted reqs queued\n"); list_del(&tail_req->execlist_link); - queue_work(dev_priv->wq, &tail_req->work); + list_add_tail(&tail_req->execlist_link, + &ring->execlist_retired_req_list); } } @@ -746,6 +730,30 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file, return 0; } +void intel_execlists_retire_requests(struct intel_engine_cs *ring) +{ + struct intel_ctx_submit_request *req, *tmp; + struct drm_i915_private *dev_priv = ring->dev->dev_private; + unsigned long flags; + struct list_head retired_list; + + WARN_ON(!mutex_is_locked(&ring->dev->struct_mutex)); + if (list_empty(&ring->execlist_retired_req_list)) + return; + + INIT_LIST_HEAD(&retired_list); + spin_lock_irqsave(&ring->execlist_lock, flags); + list_replace_init(&ring->execlist_retired_req_list, &retired_list); + spin_unlock_irqrestore(&ring->execlist_lock, flags); + + list_for_each_entry_safe(req, tmp, &retired_list, execlist_link) { + intel_runtime_pm_put(dev_priv); + i915_gem_context_unreference(req->ctx); + list_del(&req->execlist_link); + kfree(req); + } +} + void intel_logical_ring_stop(struct intel_engine_cs *ring) { struct drm_i915_private *dev_priv = ring->dev->dev_private; @@ -1301,6 +1309,7 @@ static int logical_ring_init(struct drm_device *dev, struct intel_engine_cs *rin init_waitqueue_head(&ring->irq_queue); INIT_LIST_HEAD(&ring->execlist_queue); + INIT_LIST_HEAD(&ring->execlist_retired_req_list); spin_lock_init(&ring->execlist_lock); ring->next_context_status_buffer = 0; diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h index 33c3b4b..84bbf19 100644 --- a/drivers/gpu/drm/i915/intel_lrc.h +++ b/drivers/gpu/drm/i915/intel_lrc.h @@ -104,11 +104,11 @@ struct intel_ctx_submit_request { u32 tail; struct list_head execlist_link; - struct work_struct work; int elsp_submitted; }; void intel_execlists_handle_ctx_events(struct intel_engine_cs *ring); +void intel_execlists_retire_requests(struct intel_engine_cs *ring); #endif /* _INTEL_LRC_H_ */ diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index aab2e2f..8515656 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -236,6 +236,7 @@ struct intel_engine_cs { /* Execlists */ spinlock_t execlist_lock; struct list_head execlist_queue; + struct list_head execlist_retired_req_list; u8 next_context_status_buffer; u32 irq_keep_mask; /* bitmask for interrupts that should not be masked */ int (*emit_request)(struct intel_ringbuffer *ringbuf); -- cgit v0.10.2 From dcb4c12a687710ab745c2cdee8298c3e97f6f707 Mon Sep 17 00:00:00 2001 From: Oscar Mateo Date: Thu, 13 Nov 2014 10:28:10 +0000 Subject: drm/i915/bdw: Pin the context backing objects to GGTT on-demand Up until now, we have pinned every logical ring context backing object during creation, and left it pinned until destruction. This made my life easier, but it's a harmful thing to do, because we cause fragmentation of the GGTT (and, eventually, we would run out of space). This patch makes the pinning on-demand: the backing objects of the two contexts that are written to the ELSP are pinned right before submission and unpinned once the hardware is done with them. The only context that is still pinned regardless is the global default one, so that the HWS can still be accessed in the same way (ring->status_page). v2: In the early version of this patch, we were pinning the context as we put it into the ELSP: on the one hand, this is very efficient because only a maximum two contexts are pinned at any given time, but on the other hand, we cannot really pin in interrupt time :( v3: Use a mutex rather than atomic_t to protect pin count to avoid races. Do not unpin default context in free_request. v4: Break out pin and unpin into functions. Fix style problems reported by checkpatch v5: Remove unpin_lock as all pinning and unpinning is done with the struct mutex already locked. Add WARN_ONs to make sure this is the case in future. Issue: VIZ-4277 Signed-off-by: Oscar Mateo Signed-off-by: Thomas Daniel Reviewed-by: Akash Goel Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 319da61..bb14583 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1799,10 +1799,16 @@ static int i915_dump_lrc(struct seq_file *m, void *unused) continue; if (ctx_obj) { - struct page *page = i915_gem_object_get_page(ctx_obj, 1); - uint32_t *reg_state = kmap_atomic(page); + struct page *page; + uint32_t *reg_state; int j; + i915_gem_obj_ggtt_pin(ctx_obj, + GEN8_LR_CONTEXT_ALIGN, 0); + + page = i915_gem_object_get_page(ctx_obj, 1); + reg_state = kmap_atomic(page); + seq_printf(m, "CONTEXT: %s %u\n", ring->name, intel_execlists_ctx_id(ctx_obj)); @@ -1814,6 +1820,8 @@ static int i915_dump_lrc(struct seq_file *m, void *unused) } kunmap_atomic(reg_state); + i915_gem_object_ggtt_unpin(ctx_obj); + seq_putc(m, '\n'); } } diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 4f08160..1097d20 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -685,6 +685,7 @@ struct intel_context { struct { struct drm_i915_gem_object *state; struct intel_ringbuffer *ringbuf; + int unpin_count; } engine[I915_NUM_RINGS]; struct list_head link; diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 2e85ef1..7f95d8f 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2576,12 +2576,18 @@ static void i915_set_reset_status(struct drm_i915_private *dev_priv, static void i915_gem_free_request(struct drm_i915_gem_request *request) { + struct intel_context *ctx = request->ctx; + list_del(&request->list); i915_gem_request_remove_from_client(request); - if (request->ctx) - i915_gem_context_unreference(request->ctx); + if (i915.enable_execlists && ctx) { + struct intel_engine_cs *ring = request->ring; + if (ctx != ring->default_context) + intel_lr_context_unpin(ring, ctx); + i915_gem_context_unreference(ctx); + } kfree(request); } @@ -2636,6 +2642,23 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, } /* + * Clear the execlists queue up before freeing the requests, as those + * are the ones that keep the context and ringbuffer backing objects + * pinned in place. + */ + while (!list_empty(&ring->execlist_queue)) { + struct intel_ctx_submit_request *submit_req; + + submit_req = list_first_entry(&ring->execlist_queue, + struct intel_ctx_submit_request, + execlist_link); + list_del(&submit_req->execlist_link); + intel_runtime_pm_put(dev_priv); + i915_gem_context_unreference(submit_req->ctx); + kfree(submit_req); + } + + /* * We must free the requests after all the corresponding objects have * been moved off active lists. Which is the same order as the normal * retire_requests function does. This is important if object hold @@ -2652,18 +2675,6 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, i915_gem_free_request(request); } - while (!list_empty(&ring->execlist_queue)) { - struct intel_ctx_submit_request *submit_req; - - submit_req = list_first_entry(&ring->execlist_queue, - struct intel_ctx_submit_request, - execlist_link); - list_del(&submit_req->execlist_link); - intel_runtime_pm_put(dev_priv); - i915_gem_context_unreference(submit_req->ctx); - kfree(submit_req); - } - /* These may not have been flush before the reset, do so now */ kfree(ring->preallocated_lazy_request); ring->preallocated_lazy_request = NULL; diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index c855051..9560e63 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -140,8 +140,6 @@ #define GEN8_LR_CONTEXT_RENDER_SIZE (20 * PAGE_SIZE) #define GEN8_LR_CONTEXT_OTHER_SIZE (2 * PAGE_SIZE) -#define GEN8_LR_CONTEXT_ALIGN 4096 - #define RING_EXECLIST_QFULL (1 << 0x2) #define RING_EXECLIST1_VALID (1 << 0x3) #define RING_EXECLIST0_VALID (1 << 0x4) @@ -814,9 +812,40 @@ void intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf) execlists_context_queue(ring, ctx, ringbuf->tail); } +static int intel_lr_context_pin(struct intel_engine_cs *ring, + struct intel_context *ctx) +{ + struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state; + int ret = 0; + + WARN_ON(!mutex_is_locked(&ring->dev->struct_mutex)); + if (ctx->engine[ring->id].unpin_count++ == 0) { + ret = i915_gem_obj_ggtt_pin(ctx_obj, + GEN8_LR_CONTEXT_ALIGN, 0); + if (ret) + ctx->engine[ring->id].unpin_count = 0; + } + + return ret; +} + +void intel_lr_context_unpin(struct intel_engine_cs *ring, + struct intel_context *ctx) +{ + struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state; + + if (ctx_obj) { + WARN_ON(!mutex_is_locked(&ring->dev->struct_mutex)); + if (--ctx->engine[ring->id].unpin_count == 0) + i915_gem_object_ggtt_unpin(ctx_obj); + } +} + static int logical_ring_alloc_seqno(struct intel_engine_cs *ring, struct intel_context *ctx) { + int ret; + if (ring->outstanding_lazy_seqno) return 0; @@ -827,6 +856,14 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring, if (request == NULL) return -ENOMEM; + if (ctx != ring->default_context) { + ret = intel_lr_context_pin(ring, ctx); + if (ret) { + kfree(request); + return ret; + } + } + /* Hold a reference to the context this request belongs to * (we will need it when the time comes to emit/retire the * request). @@ -1680,12 +1717,16 @@ void intel_lr_context_free(struct intel_context *ctx) for (i = 0; i < I915_NUM_RINGS; i++) { struct drm_i915_gem_object *ctx_obj = ctx->engine[i].state; - struct intel_ringbuffer *ringbuf = ctx->engine[i].ringbuf; if (ctx_obj) { + struct intel_ringbuffer *ringbuf = + ctx->engine[i].ringbuf; + struct intel_engine_cs *ring = ringbuf->ring; + intel_destroy_ringbuffer_obj(ringbuf); kfree(ringbuf); - i915_gem_object_ggtt_unpin(ctx_obj); + if (ctx == ring->default_context) + i915_gem_object_ggtt_unpin(ctx_obj); drm_gem_object_unreference(&ctx_obj->base); } } @@ -1748,6 +1789,7 @@ static void lrc_setup_hardware_status_page(struct intel_engine_cs *ring, int intel_lr_context_deferred_create(struct intel_context *ctx, struct intel_engine_cs *ring) { + const bool is_global_default_ctx = (ctx == ring->default_context); struct drm_device *dev = ring->dev; struct drm_i915_gem_object *ctx_obj; uint32_t context_size; @@ -1767,18 +1809,22 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, return ret; } - ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN, 0); - if (ret) { - DRM_DEBUG_DRIVER("Pin LRC backing obj failed: %d\n", ret); - drm_gem_object_unreference(&ctx_obj->base); - return ret; + if (is_global_default_ctx) { + ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN, 0); + if (ret) { + DRM_DEBUG_DRIVER("Pin LRC backing obj failed: %d\n", + ret); + drm_gem_object_unreference(&ctx_obj->base); + return ret; + } } ringbuf = kzalloc(sizeof(*ringbuf), GFP_KERNEL); if (!ringbuf) { DRM_DEBUG_DRIVER("Failed to allocate ringbuffer %s\n", ring->name); - i915_gem_object_ggtt_unpin(ctx_obj); + if (is_global_default_ctx) + i915_gem_object_ggtt_unpin(ctx_obj); drm_gem_object_unreference(&ctx_obj->base); ret = -ENOMEM; return ret; @@ -1841,7 +1887,8 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, error: kfree(ringbuf); - i915_gem_object_ggtt_unpin(ctx_obj); + if (is_global_default_ctx) + i915_gem_object_ggtt_unpin(ctx_obj); drm_gem_object_unreference(&ctx_obj->base); return ret; } diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h index 84bbf19..14b216b 100644 --- a/drivers/gpu/drm/i915/intel_lrc.h +++ b/drivers/gpu/drm/i915/intel_lrc.h @@ -24,6 +24,8 @@ #ifndef _INTEL_LRC_H_ #define _INTEL_LRC_H_ +#define GEN8_LR_CONTEXT_ALIGN 4096 + /* Execlists regs */ #define RING_ELSP(ring) ((ring)->mmio_base+0x230) #define RING_EXECLIST_STATUS(ring) ((ring)->mmio_base+0x234) @@ -67,6 +69,8 @@ int intel_lr_context_render_state_init(struct intel_engine_cs *ring, void intel_lr_context_free(struct intel_context *ctx); int intel_lr_context_deferred_create(struct intel_context *ctx, struct intel_engine_cs *ring); +void intel_lr_context_unpin(struct intel_engine_cs *ring, + struct intel_context *ctx); /* Execlists */ int intel_sanitize_enable_execlists(struct drm_device *dev, int enable_execlists); -- cgit v0.10.2 From 7ba717cf365d79f2b284e508205ec3d4a05fc41b Mon Sep 17 00:00:00 2001 From: Thomas Daniel Date: Thu, 13 Nov 2014 10:28:56 +0000 Subject: drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo Signed-off-by: Thomas Daniel Reviewed-by: Akash Goel Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 9560e63..e588376 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -203,6 +203,9 @@ enum { }; #define GEN8_CTX_ID_SHIFT 32 +static int intel_lr_context_pin(struct intel_engine_cs *ring, + struct intel_context *ctx); + /** * intel_sanitize_enable_execlists() - sanitize i915.enable_execlists * @dev: DRM device. @@ -354,7 +357,9 @@ static void execlists_elsp_write(struct intel_engine_cs *ring, spin_unlock_irqrestore(&dev_priv->uncore.lock, flags); } -static int execlists_ctx_write_tail(struct drm_i915_gem_object *ctx_obj, u32 tail) +static int execlists_update_context(struct drm_i915_gem_object *ctx_obj, + struct drm_i915_gem_object *ring_obj, + u32 tail) { struct page *page; uint32_t *reg_state; @@ -363,6 +368,7 @@ static int execlists_ctx_write_tail(struct drm_i915_gem_object *ctx_obj, u32 tai reg_state = kmap_atomic(page); reg_state[CTX_RING_TAIL+1] = tail; + reg_state[CTX_RING_BUFFER_START+1] = i915_gem_obj_ggtt_offset(ring_obj); kunmap_atomic(reg_state); @@ -373,21 +379,25 @@ static void execlists_submit_contexts(struct intel_engine_cs *ring, struct intel_context *to0, u32 tail0, struct intel_context *to1, u32 tail1) { - struct drm_i915_gem_object *ctx_obj0; + struct drm_i915_gem_object *ctx_obj0 = to0->engine[ring->id].state; + struct intel_ringbuffer *ringbuf0 = to0->engine[ring->id].ringbuf; struct drm_i915_gem_object *ctx_obj1 = NULL; + struct intel_ringbuffer *ringbuf1 = NULL; - ctx_obj0 = to0->engine[ring->id].state; BUG_ON(!ctx_obj0); WARN_ON(!i915_gem_obj_is_pinned(ctx_obj0)); + WARN_ON(!i915_gem_obj_is_pinned(ringbuf0->obj)); - execlists_ctx_write_tail(ctx_obj0, tail0); + execlists_update_context(ctx_obj0, ringbuf0->obj, tail0); if (to1) { + ringbuf1 = to1->engine[ring->id].ringbuf; ctx_obj1 = to1->engine[ring->id].state; BUG_ON(!ctx_obj1); WARN_ON(!i915_gem_obj_is_pinned(ctx_obj1)); + WARN_ON(!i915_gem_obj_is_pinned(ringbuf1->obj)); - execlists_ctx_write_tail(ctx_obj1, tail1); + execlists_update_context(ctx_obj1, ringbuf1->obj, tail1); } execlists_elsp_write(ring, ctx_obj0, ctx_obj1); @@ -537,6 +547,10 @@ static int execlists_context_queue(struct intel_engine_cs *ring, return -ENOMEM; req->ctx = to; i915_gem_context_reference(req->ctx); + + if (to != ring->default_context) + intel_lr_context_pin(ring, to); + req->ring = ring; req->tail = tail; @@ -557,7 +571,7 @@ static int execlists_context_queue(struct intel_engine_cs *ring, if (to == tail_req->ctx) { WARN(tail_req->elsp_submitted != 0, - "More than 2 already-submitted reqs queued\n"); + "More than 2 already-submitted reqs queued\n"); list_del(&tail_req->execlist_link); list_add_tail(&tail_req->execlist_link, &ring->execlist_retired_req_list); @@ -745,6 +759,12 @@ void intel_execlists_retire_requests(struct intel_engine_cs *ring) spin_unlock_irqrestore(&ring->execlist_lock, flags); list_for_each_entry_safe(req, tmp, &retired_list, execlist_link) { + struct intel_context *ctx = req->ctx; + struct drm_i915_gem_object *ctx_obj = + ctx->engine[ring->id].state; + + if (ctx_obj && (ctx != ring->default_context)) + intel_lr_context_unpin(ring, ctx); intel_runtime_pm_put(dev_priv); i915_gem_context_unreference(req->ctx); list_del(&req->execlist_link); @@ -816,6 +836,7 @@ static int intel_lr_context_pin(struct intel_engine_cs *ring, struct intel_context *ctx) { struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state; + struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf; int ret = 0; WARN_ON(!mutex_is_locked(&ring->dev->struct_mutex)); @@ -823,21 +844,35 @@ static int intel_lr_context_pin(struct intel_engine_cs *ring, ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN, 0); if (ret) - ctx->engine[ring->id].unpin_count = 0; + goto reset_unpin_count; + + ret = intel_pin_and_map_ringbuffer_obj(ring->dev, ringbuf); + if (ret) + goto unpin_ctx_obj; } return ret; + +unpin_ctx_obj: + i915_gem_object_ggtt_unpin(ctx_obj); +reset_unpin_count: + ctx->engine[ring->id].unpin_count = 0; + + return ret; } void intel_lr_context_unpin(struct intel_engine_cs *ring, struct intel_context *ctx) { struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state; + struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf; if (ctx_obj) { WARN_ON(!mutex_is_locked(&ring->dev->struct_mutex)); - if (--ctx->engine[ring->id].unpin_count == 0) + if (--ctx->engine[ring->id].unpin_count == 0) { + intel_unpin_ringbuffer_obj(ringbuf); i915_gem_object_ggtt_unpin(ctx_obj); + } } } @@ -1595,7 +1630,6 @@ populate_lr_context(struct intel_context *ctx, struct drm_i915_gem_object *ctx_o { struct drm_device *dev = ring->dev; struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_gem_object *ring_obj = ringbuf->obj; struct i915_hw_ppgtt *ppgtt = ctx->ppgtt; struct page *page; uint32_t *reg_state; @@ -1641,7 +1675,9 @@ populate_lr_context(struct intel_context *ctx, struct drm_i915_gem_object *ctx_o reg_state[CTX_RING_TAIL] = RING_TAIL(ring->mmio_base); reg_state[CTX_RING_TAIL+1] = 0; reg_state[CTX_RING_BUFFER_START] = RING_START(ring->mmio_base); - reg_state[CTX_RING_BUFFER_START+1] = i915_gem_obj_ggtt_offset(ring_obj); + /* Ring buffer start address is not known until the buffer is pinned. + * It is written to the context image in execlists_update_context() + */ reg_state[CTX_RING_BUFFER_CONTROL] = RING_CTL(ring->mmio_base); reg_state[CTX_RING_BUFFER_CONTROL+1] = ((ringbuf->size - PAGE_SIZE) & RING_NR_PAGES) | RING_VALID; @@ -1723,10 +1759,12 @@ void intel_lr_context_free(struct intel_context *ctx) ctx->engine[i].ringbuf; struct intel_engine_cs *ring = ringbuf->ring; + if (ctx == ring->default_context) { + intel_unpin_ringbuffer_obj(ringbuf); + i915_gem_object_ggtt_unpin(ctx_obj); + } intel_destroy_ringbuffer_obj(ringbuf); kfree(ringbuf); - if (ctx == ring->default_context) - i915_gem_object_ggtt_unpin(ctx_obj); drm_gem_object_unreference(&ctx_obj->base); } } @@ -1823,11 +1861,8 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, if (!ringbuf) { DRM_DEBUG_DRIVER("Failed to allocate ringbuffer %s\n", ring->name); - if (is_global_default_ctx) - i915_gem_object_ggtt_unpin(ctx_obj); - drm_gem_object_unreference(&ctx_obj->base); ret = -ENOMEM; - return ret; + goto error_unpin_ctx; } ringbuf->ring = ring; @@ -1840,22 +1875,30 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, ringbuf->space = ringbuf->size; ringbuf->last_retired_head = -1; - /* TODO: For now we put this in the mappable region so that we can reuse - * the existing ringbuffer code which ioremaps it. When we start - * creating many contexts, this will no longer work and we must switch - * to a kmapish interface. - */ - ret = intel_alloc_ringbuffer_obj(dev, ringbuf); - if (ret) { - DRM_DEBUG_DRIVER("Failed to allocate ringbuffer obj %s: %d\n", + if (ringbuf->obj == NULL) { + ret = intel_alloc_ringbuffer_obj(dev, ringbuf); + if (ret) { + DRM_DEBUG_DRIVER( + "Failed to allocate ringbuffer obj %s: %d\n", ring->name, ret); - goto error; + goto error_free_rbuf; + } + + if (is_global_default_ctx) { + ret = intel_pin_and_map_ringbuffer_obj(dev, ringbuf); + if (ret) { + DRM_ERROR( + "Failed to pin and map ringbuffer %s: %d\n", + ring->name, ret); + goto error_destroy_rbuf; + } + } + } ret = populate_lr_context(ctx, ctx_obj, ring, ringbuf); if (ret) { DRM_DEBUG_DRIVER("Failed to populate LRC: %d\n", ret); - intel_destroy_ringbuffer_obj(ringbuf); goto error; } @@ -1877,7 +1920,6 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, DRM_ERROR("Init render state failed: %d\n", ret); ctx->engine[ring->id].ringbuf = NULL; ctx->engine[ring->id].state = NULL; - intel_destroy_ringbuffer_obj(ringbuf); goto error; } ctx->rcs_initialized = true; @@ -1886,7 +1928,13 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, return 0; error: + if (is_global_default_ctx) + intel_unpin_ringbuffer_obj(ringbuf); +error_destroy_rbuf: + intel_destroy_ringbuffer_obj(ringbuf); +error_free_rbuf: kfree(ringbuf); +error_unpin_ctx: if (is_global_default_ctx) i915_gem_object_ggtt_unpin(ctx_obj); drm_gem_object_unreference(&ctx_obj->base); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index ae09258..0a4f35e 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1722,13 +1722,42 @@ static int init_phys_status_page(struct intel_engine_cs *ring) return 0; } -void intel_destroy_ringbuffer_obj(struct intel_ringbuffer *ringbuf) +void intel_unpin_ringbuffer_obj(struct intel_ringbuffer *ringbuf) { - if (!ringbuf->obj) - return; - iounmap(ringbuf->virtual_start); + ringbuf->virtual_start = NULL; i915_gem_object_ggtt_unpin(ringbuf->obj); +} + +int intel_pin_and_map_ringbuffer_obj(struct drm_device *dev, + struct intel_ringbuffer *ringbuf) +{ + struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_gem_object *obj = ringbuf->obj; + int ret; + + ret = i915_gem_obj_ggtt_pin(obj, PAGE_SIZE, PIN_MAPPABLE); + if (ret) + return ret; + + ret = i915_gem_object_set_to_gtt_domain(obj, true); + if (ret) { + i915_gem_object_ggtt_unpin(obj); + return ret; + } + + ringbuf->virtual_start = ioremap_wc(dev_priv->gtt.mappable_base + + i915_gem_obj_ggtt_offset(obj), ringbuf->size); + if (ringbuf->virtual_start == NULL) { + i915_gem_object_ggtt_unpin(obj); + return -EINVAL; + } + + return 0; +} + +void intel_destroy_ringbuffer_obj(struct intel_ringbuffer *ringbuf) +{ drm_gem_object_unreference(&ringbuf->obj->base); ringbuf->obj = NULL; } @@ -1736,12 +1765,7 @@ void intel_destroy_ringbuffer_obj(struct intel_ringbuffer *ringbuf) int intel_alloc_ringbuffer_obj(struct drm_device *dev, struct intel_ringbuffer *ringbuf) { - struct drm_i915_private *dev_priv = to_i915(dev); struct drm_i915_gem_object *obj; - int ret; - - if (ringbuf->obj) - return 0; obj = NULL; if (!HAS_LLC(dev)) @@ -1754,30 +1778,9 @@ int intel_alloc_ringbuffer_obj(struct drm_device *dev, /* mark ring buffers as read-only from GPU side by default */ obj->gt_ro = 1; - ret = i915_gem_obj_ggtt_pin(obj, PAGE_SIZE, PIN_MAPPABLE); - if (ret) - goto err_unref; - - ret = i915_gem_object_set_to_gtt_domain(obj, true); - if (ret) - goto err_unpin; - - ringbuf->virtual_start = - ioremap_wc(dev_priv->gtt.mappable_base + i915_gem_obj_ggtt_offset(obj), - ringbuf->size); - if (ringbuf->virtual_start == NULL) { - ret = -EINVAL; - goto err_unpin; - } - ringbuf->obj = obj; - return 0; -err_unpin: - i915_gem_object_ggtt_unpin(obj); -err_unref: - drm_gem_object_unreference(&obj->base); - return ret; + return 0; } static int intel_init_ring_buffer(struct drm_device *dev, @@ -1814,10 +1817,21 @@ static int intel_init_ring_buffer(struct drm_device *dev, goto error; } - ret = intel_alloc_ringbuffer_obj(dev, ringbuf); - if (ret) { - DRM_ERROR("Failed to allocate ringbuffer %s: %d\n", ring->name, ret); - goto error; + if (ringbuf->obj == NULL) { + ret = intel_alloc_ringbuffer_obj(dev, ringbuf); + if (ret) { + DRM_ERROR("Failed to allocate ringbuffer %s: %d\n", + ring->name, ret); + goto error; + } + + ret = intel_pin_and_map_ringbuffer_obj(dev, ringbuf); + if (ret) { + DRM_ERROR("Failed to pin and map ringbuffer %s: %d\n", + ring->name, ret); + intel_destroy_ringbuffer_obj(ringbuf); + goto error; + } } /* Workaround an erratum on the i830 which causes a hang if @@ -1858,6 +1872,7 @@ void intel_cleanup_ring_buffer(struct intel_engine_cs *ring) intel_stop_ring_buffer(ring); WARN_ON(!IS_GEN2(ring->dev) && (I915_READ_MODE(ring) & MODE_IDLE) == 0); + intel_unpin_ringbuffer_obj(ringbuf); intel_destroy_ringbuffer_obj(ringbuf); ring->preallocated_lazy_request = NULL; ring->outstanding_lazy_seqno = 0; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 8515656..9eb3188 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -383,6 +383,9 @@ intel_write_status_page(struct intel_engine_cs *ring, #define I915_GEM_HWS_SCRATCH_INDEX 0x30 #define I915_GEM_HWS_SCRATCH_ADDR (I915_GEM_HWS_SCRATCH_INDEX << MI_STORE_DWORD_INDEX_SHIFT) +void intel_unpin_ringbuffer_obj(struct intel_ringbuffer *ringbuf); +int intel_pin_and_map_ringbuffer_obj(struct drm_device *dev, + struct intel_ringbuffer *ringbuf); void intel_destroy_ringbuffer_obj(struct intel_ringbuffer *ringbuf); int intel_alloc_ringbuffer_obj(struct drm_device *dev, struct intel_ringbuffer *ringbuf); -- cgit v0.10.2 From 5c6c600354adac5f95fd41b178b084ac0182e14c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sat, 6 Sep 2014 10:28:27 +0100 Subject: drm/i915: Remove DRI1 ring accessors and API With the deprecation of UMS, and by association DRI1, we have a tough choice when updating the ring access routines. We either rewrite the DRI1 routines blindly without testing (so likely to be broken) or take the liberty of declaring them no longer supported and remove them entirely. This takes the latter approach. v2: Also remove the DRI1 sarea updates Signed-off-by: Chris Wilson [danvet: Fix rebase conflicts.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 56f6211..d156228 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -50,883 +50,58 @@ #include #include -#define LP_RING(d) (&((struct drm_i915_private *)(d))->ring[RCS]) - -#define BEGIN_LP_RING(n) \ - intel_ring_begin(LP_RING(dev_priv), (n)) - -#define OUT_RING(x) \ - intel_ring_emit(LP_RING(dev_priv), x) - -#define ADVANCE_LP_RING() \ - __intel_ring_advance(LP_RING(dev_priv)) - -/** - * Lock test for when it's just for synchronization of ring access. - * - * In that case, we don't need to do it when GEM is initialized as nobody else - * has access to the ring. - */ -#define RING_LOCK_TEST_WITH_RETURN(dev, file) do { \ - if (LP_RING(dev->dev_private)->buffer->obj == NULL) \ - LOCK_TEST_WITH_RETURN(dev, file); \ -} while (0) - -static inline u32 -intel_read_legacy_status_page(struct drm_i915_private *dev_priv, int reg) -{ - if (I915_NEED_GFX_HWS(dev_priv->dev)) - return ioread32(dev_priv->dri1.gfx_hws_cpu_addr + reg); - else - return intel_read_status_page(LP_RING(dev_priv), reg); -} - -#define READ_HWSP(dev_priv, reg) intel_read_legacy_status_page(dev_priv, reg) -#define READ_BREADCRUMB(dev_priv) READ_HWSP(dev_priv, I915_BREADCRUMB_INDEX) -#define I915_BREADCRUMB_INDEX 0x21 - -void i915_update_dri1_breadcrumb(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv; - - /* - * The dri breadcrumb update races against the drm master disappearing. - * Instead of trying to fix this (this is by far not the only ums issue) - * just don't do the update in kms mode. - */ - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return; - - if (dev->primary->master) { - master_priv = dev->primary->master->driver_priv; - if (master_priv->sarea_priv) - master_priv->sarea_priv->last_dispatch = - READ_BREADCRUMB(dev_priv); - } -} - -static void i915_write_hws_pga(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - u32 addr; - - addr = dev_priv->status_page_dmah->busaddr; - if (INTEL_INFO(dev)->gen >= 4) - addr |= (dev_priv->status_page_dmah->busaddr >> 28) & 0xf0; - I915_WRITE(HWS_PGA, addr); -} - -/** - * Frees the hardware status page, whether it's a physical address or a virtual - * address set up by the X Server. - */ -static void i915_free_hws(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_engine_cs *ring = LP_RING(dev_priv); - - if (dev_priv->status_page_dmah) { - drm_pci_free(dev, dev_priv->status_page_dmah); - dev_priv->status_page_dmah = NULL; - } - - if (ring->status_page.gfx_addr) { - ring->status_page.gfx_addr = 0; - iounmap(dev_priv->dri1.gfx_hws_cpu_addr); - } - - /* Need to rewrite hardware status page */ - I915_WRITE(HWS_PGA, 0x1ffff000); -} - -void i915_kernel_lost_context(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv; - struct intel_engine_cs *ring = LP_RING(dev_priv); - struct intel_ringbuffer *ringbuf = ring->buffer; - - /* - * We should never lose context on the ring with modesetting - * as we don't expose it to userspace - */ - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return; - - ringbuf->head = I915_READ_HEAD(ring) & HEAD_ADDR; - ringbuf->tail = I915_READ_TAIL(ring) & TAIL_ADDR; - ringbuf->space = ringbuf->head - (ringbuf->tail + I915_RING_FREE_SPACE); - if (ringbuf->space < 0) - ringbuf->space += ringbuf->size; - - if (!dev->primary->master) - return; - - master_priv = dev->primary->master->driver_priv; - if (ringbuf->head == ringbuf->tail && master_priv->sarea_priv) - master_priv->sarea_priv->perf_boxes |= I915_BOX_RING_EMPTY; -} - -static int i915_dma_cleanup(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int i; - - /* Make sure interrupts are disabled here because the uninstall ioctl - * may not have been called from userspace and after dev_private - * is freed, it's too late. - */ - if (dev->irq_enabled) - drm_irq_uninstall(dev); - - mutex_lock(&dev->struct_mutex); - for (i = 0; i < I915_NUM_RINGS; i++) - intel_cleanup_ring_buffer(&dev_priv->ring[i]); - mutex_unlock(&dev->struct_mutex); - - /* Clear the HWS virtual address at teardown */ - if (I915_NEED_GFX_HWS(dev)) - i915_free_hws(dev); - - return 0; -} - -static int i915_initialize(struct drm_device *dev, drm_i915_init_t *init) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; - int ret; - - master_priv->sarea = drm_legacy_getsarea(dev); - if (master_priv->sarea) { - master_priv->sarea_priv = (drm_i915_sarea_t *) - ((u8 *)master_priv->sarea->handle + init->sarea_priv_offset); - } else { - DRM_DEBUG_DRIVER("sarea not found assuming DRI2 userspace\n"); - } - - if (init->ring_size != 0) { - if (LP_RING(dev_priv)->buffer->obj != NULL) { - i915_dma_cleanup(dev); - DRM_ERROR("Client tried to initialize ringbuffer in " - "GEM mode\n"); - return -EINVAL; - } - - ret = intel_render_ring_init_dri(dev, - init->ring_start, - init->ring_size); - if (ret) { - i915_dma_cleanup(dev); - return ret; - } - } - - dev_priv->dri1.cpp = init->cpp; - dev_priv->dri1.back_offset = init->back_offset; - dev_priv->dri1.front_offset = init->front_offset; - dev_priv->dri1.current_page = 0; - if (master_priv->sarea_priv) - master_priv->sarea_priv->pf_current_page = 0; - - /* Allow hardware batchbuffers unless told otherwise. - */ - dev_priv->dri1.allow_batchbuffer = 1; - - return 0; -} - -static int i915_dma_resume(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_engine_cs *ring = LP_RING(dev_priv); - - DRM_DEBUG_DRIVER("%s\n", __func__); - - if (ring->buffer->virtual_start == NULL) { - DRM_ERROR("can not ioremap virtual address for" - " ring buffer\n"); - return -ENOMEM; - } - - /* Program Hardware Status Page */ - if (!ring->status_page.page_addr) { - DRM_ERROR("Can not find hardware status page\n"); - return -EINVAL; - } - DRM_DEBUG_DRIVER("hw status page @ %p\n", - ring->status_page.page_addr); - if (ring->status_page.gfx_addr != 0) - intel_ring_setup_status_page(ring); - else - i915_write_hws_pga(dev); - - DRM_DEBUG_DRIVER("Enabled hardware status page\n"); - - return 0; -} - static int i915_dma_init(struct drm_device *dev, void *data, struct drm_file *file_priv) { - drm_i915_init_t *init = data; - int retcode = 0; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - switch (init->func) { - case I915_INIT_DMA: - retcode = i915_initialize(dev, init); - break; - case I915_CLEANUP_DMA: - retcode = i915_dma_cleanup(dev); - break; - case I915_RESUME_DMA: - retcode = i915_dma_resume(dev); - break; - default: - retcode = -EINVAL; - break; - } - - return retcode; -} - -/* Implement basically the same security restrictions as hardware does - * for MI_BATCH_NON_SECURE. These can be made stricter at any time. - * - * Most of the calculations below involve calculating the size of a - * particular instruction. It's important to get the size right as - * that tells us where the next instruction to check is. Any illegal - * instruction detected will be given a size of zero, which is a - * signal to abort the rest of the buffer. - */ -static int validate_cmd(int cmd) -{ - switch (((cmd >> 29) & 0x7)) { - case 0x0: - switch ((cmd >> 23) & 0x3f) { - case 0x0: - return 1; /* MI_NOOP */ - case 0x4: - return 1; /* MI_FLUSH */ - default: - return 0; /* disallow everything else */ - } - break; - case 0x1: - return 0; /* reserved */ - case 0x2: - return (cmd & 0xff) + 2; /* 2d commands */ - case 0x3: - if (((cmd >> 24) & 0x1f) <= 0x18) - return 1; - - switch ((cmd >> 24) & 0x1f) { - case 0x1c: - return 1; - case 0x1d: - switch ((cmd >> 16) & 0xff) { - case 0x3: - return (cmd & 0x1f) + 2; - case 0x4: - return (cmd & 0xf) + 2; - default: - return (cmd & 0xffff) + 2; - } - case 0x1e: - if (cmd & (1 << 23)) - return (cmd & 0xffff) + 1; - else - return 1; - case 0x1f: - if ((cmd & (1 << 23)) == 0) /* inline vertices */ - return (cmd & 0x1ffff) + 2; - else if (cmd & (1 << 17)) /* indirect random */ - if ((cmd & 0xffff) == 0) - return 0; /* unknown length, too hard */ - else - return (((cmd & 0xffff) + 1) / 2) + 1; - else - return 2; /* indirect sequential */ - default: - return 0; - } - default: - return 0; - } - - return 0; -} - -static int i915_emit_cmds(struct drm_device *dev, int *buffer, int dwords) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int i, ret; - - if ((dwords+1) * sizeof(int) >= LP_RING(dev_priv)->buffer->size - 8) - return -EINVAL; - - for (i = 0; i < dwords;) { - int sz = validate_cmd(buffer[i]); - - if (sz == 0 || i + sz > dwords) - return -EINVAL; - i += sz; - } - - ret = BEGIN_LP_RING((dwords+1)&~1); - if (ret) - return ret; - - for (i = 0; i < dwords; i++) - OUT_RING(buffer[i]); - if (dwords & 1) - OUT_RING(0); - - ADVANCE_LP_RING(); - - return 0; -} - -int -i915_emit_box(struct drm_device *dev, - struct drm_clip_rect *box, - int DR1, int DR4) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int ret; - - if (box->y2 <= box->y1 || box->x2 <= box->x1 || - box->y2 <= 0 || box->x2 <= 0) { - DRM_ERROR("Bad box %d,%d..%d,%d\n", - box->x1, box->y1, box->x2, box->y2); - return -EINVAL; - } - - if (INTEL_INFO(dev)->gen >= 4) { - ret = BEGIN_LP_RING(4); - if (ret) - return ret; - - OUT_RING(GFX_OP_DRAWRECT_INFO_I965); - OUT_RING((box->x1 & 0xffff) | (box->y1 << 16)); - OUT_RING(((box->x2 - 1) & 0xffff) | ((box->y2 - 1) << 16)); - OUT_RING(DR4); - } else { - ret = BEGIN_LP_RING(6); - if (ret) - return ret; - - OUT_RING(GFX_OP_DRAWRECT_INFO); - OUT_RING(DR1); - OUT_RING((box->x1 & 0xffff) | (box->y1 << 16)); - OUT_RING(((box->x2 - 1) & 0xffff) | ((box->y2 - 1) << 16)); - OUT_RING(DR4); - OUT_RING(0); - } - ADVANCE_LP_RING(); - - return 0; -} - -/* XXX: Emitting the counter should really be moved to part of the IRQ - * emit. For now, do it in both places: - */ - -static void i915_emit_breadcrumb(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; - - dev_priv->dri1.counter++; - if (dev_priv->dri1.counter > 0x7FFFFFFFUL) - dev_priv->dri1.counter = 0; - if (master_priv->sarea_priv) - master_priv->sarea_priv->last_enqueue = dev_priv->dri1.counter; - - if (BEGIN_LP_RING(4) == 0) { - OUT_RING(MI_STORE_DWORD_INDEX); - OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - OUT_RING(dev_priv->dri1.counter); - OUT_RING(0); - ADVANCE_LP_RING(); - } -} - -static int i915_dispatch_cmdbuffer(struct drm_device *dev, - drm_i915_cmdbuffer_t *cmd, - struct drm_clip_rect *cliprects, - void *cmdbuf) -{ - int nbox = cmd->num_cliprects; - int i = 0, count, ret; - - if (cmd->sz & 0x3) { - DRM_ERROR("alignment"); - return -EINVAL; - } - - i915_kernel_lost_context(dev); - - count = nbox ? nbox : 1; - - for (i = 0; i < count; i++) { - if (i < nbox) { - ret = i915_emit_box(dev, &cliprects[i], - cmd->DR1, cmd->DR4); - if (ret) - return ret; - } - - ret = i915_emit_cmds(dev, cmdbuf, cmd->sz / 4); - if (ret) - return ret; - } - - i915_emit_breadcrumb(dev); - return 0; -} - -static int i915_dispatch_batchbuffer(struct drm_device *dev, - drm_i915_batchbuffer_t *batch, - struct drm_clip_rect *cliprects) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int nbox = batch->num_cliprects; - int i, count, ret; - - if ((batch->start | batch->used) & 0x7) { - DRM_ERROR("alignment"); - return -EINVAL; - } - - i915_kernel_lost_context(dev); - - count = nbox ? nbox : 1; - for (i = 0; i < count; i++) { - if (i < nbox) { - ret = i915_emit_box(dev, &cliprects[i], - batch->DR1, batch->DR4); - if (ret) - return ret; - } - - if (!IS_I830(dev) && !IS_845G(dev)) { - ret = BEGIN_LP_RING(2); - if (ret) - return ret; - - if (INTEL_INFO(dev)->gen >= 4) { - OUT_RING(MI_BATCH_BUFFER_START | (2 << 6) | MI_BATCH_NON_SECURE_I965); - OUT_RING(batch->start); - } else { - OUT_RING(MI_BATCH_BUFFER_START | (2 << 6)); - OUT_RING(batch->start | MI_BATCH_NON_SECURE); - } - } else { - ret = BEGIN_LP_RING(4); - if (ret) - return ret; - - OUT_RING(MI_BATCH_BUFFER); - OUT_RING(batch->start | MI_BATCH_NON_SECURE); - OUT_RING(batch->start + batch->used - 4); - OUT_RING(0); - } - ADVANCE_LP_RING(); - } - - - if (IS_G4X(dev) || IS_GEN5(dev)) { - if (BEGIN_LP_RING(2) == 0) { - OUT_RING(MI_FLUSH | MI_NO_WRITE_FLUSH | MI_INVALIDATE_ISP); - OUT_RING(MI_NOOP); - ADVANCE_LP_RING(); - } - } - - i915_emit_breadcrumb(dev); - return 0; -} - -static int i915_dispatch_flip(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv = - dev->primary->master->driver_priv; - int ret; - - if (!master_priv->sarea_priv) - return -EINVAL; - - DRM_DEBUG_DRIVER("%s: page=%d pfCurrentPage=%d\n", - __func__, - dev_priv->dri1.current_page, - master_priv->sarea_priv->pf_current_page); - - i915_kernel_lost_context(dev); - - ret = BEGIN_LP_RING(10); - if (ret) - return ret; - - OUT_RING(MI_FLUSH | MI_READ_FLUSH); - OUT_RING(0); - - OUT_RING(CMD_OP_DISPLAYBUFFER_INFO | ASYNC_FLIP); - OUT_RING(0); - if (dev_priv->dri1.current_page == 0) { - OUT_RING(dev_priv->dri1.back_offset); - dev_priv->dri1.current_page = 1; - } else { - OUT_RING(dev_priv->dri1.front_offset); - dev_priv->dri1.current_page = 0; - } - OUT_RING(0); - - OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP); - OUT_RING(0); - - ADVANCE_LP_RING(); - - master_priv->sarea_priv->last_enqueue = dev_priv->dri1.counter++; - - if (BEGIN_LP_RING(4) == 0) { - OUT_RING(MI_STORE_DWORD_INDEX); - OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - OUT_RING(dev_priv->dri1.counter); - OUT_RING(0); - ADVANCE_LP_RING(); - } - - master_priv->sarea_priv->pf_current_page = dev_priv->dri1.current_page; - return 0; -} - -static int i915_quiescent(struct drm_device *dev) -{ - i915_kernel_lost_context(dev); - return intel_ring_idle(LP_RING(dev->dev_private)); + return -ENODEV; } static int i915_flush_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) { - int ret; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - RING_LOCK_TEST_WITH_RETURN(dev, file_priv); - - mutex_lock(&dev->struct_mutex); - ret = i915_quiescent(dev); - mutex_unlock(&dev->struct_mutex); - - return ret; + return -ENODEV; } static int i915_batchbuffer(struct drm_device *dev, void *data, struct drm_file *file_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv; - drm_i915_sarea_t *sarea_priv; - drm_i915_batchbuffer_t *batch = data; - int ret; - struct drm_clip_rect *cliprects = NULL; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - master_priv = dev->primary->master->driver_priv; - sarea_priv = (drm_i915_sarea_t *) master_priv->sarea_priv; - - if (!dev_priv->dri1.allow_batchbuffer) { - DRM_ERROR("Batchbuffer ioctl disabled\n"); - return -EINVAL; - } - - DRM_DEBUG_DRIVER("i915 batchbuffer, start %x used %d cliprects %d\n", - batch->start, batch->used, batch->num_cliprects); - - RING_LOCK_TEST_WITH_RETURN(dev, file_priv); - - if (batch->num_cliprects < 0) - return -EINVAL; - - if (batch->num_cliprects) { - cliprects = kcalloc(batch->num_cliprects, - sizeof(*cliprects), - GFP_KERNEL); - if (cliprects == NULL) - return -ENOMEM; - - ret = copy_from_user(cliprects, batch->cliprects, - batch->num_cliprects * - sizeof(struct drm_clip_rect)); - if (ret != 0) { - ret = -EFAULT; - goto fail_free; - } - } - - mutex_lock(&dev->struct_mutex); - ret = i915_dispatch_batchbuffer(dev, batch, cliprects); - mutex_unlock(&dev->struct_mutex); - - if (sarea_priv) - sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - -fail_free: - kfree(cliprects); - - return ret; + return -ENODEV; } static int i915_cmdbuffer(struct drm_device *dev, void *data, struct drm_file *file_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv; - drm_i915_sarea_t *sarea_priv; - drm_i915_cmdbuffer_t *cmdbuf = data; - struct drm_clip_rect *cliprects = NULL; - void *batch_data; - int ret; - - DRM_DEBUG_DRIVER("i915 cmdbuffer, buf %p sz %d cliprects %d\n", - cmdbuf->buf, cmdbuf->sz, cmdbuf->num_cliprects); - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - master_priv = dev->primary->master->driver_priv; - sarea_priv = (drm_i915_sarea_t *) master_priv->sarea_priv; - - RING_LOCK_TEST_WITH_RETURN(dev, file_priv); - - if (cmdbuf->num_cliprects < 0) - return -EINVAL; - - batch_data = kmalloc(cmdbuf->sz, GFP_KERNEL); - if (batch_data == NULL) - return -ENOMEM; - - ret = copy_from_user(batch_data, cmdbuf->buf, cmdbuf->sz); - if (ret != 0) { - ret = -EFAULT; - goto fail_batch_free; - } - - if (cmdbuf->num_cliprects) { - cliprects = kcalloc(cmdbuf->num_cliprects, - sizeof(*cliprects), GFP_KERNEL); - if (cliprects == NULL) { - ret = -ENOMEM; - goto fail_batch_free; - } - - ret = copy_from_user(cliprects, cmdbuf->cliprects, - cmdbuf->num_cliprects * - sizeof(struct drm_clip_rect)); - if (ret != 0) { - ret = -EFAULT; - goto fail_clip_free; - } - } - - mutex_lock(&dev->struct_mutex); - ret = i915_dispatch_cmdbuffer(dev, cmdbuf, cliprects, batch_data); - mutex_unlock(&dev->struct_mutex); - if (ret) { - DRM_ERROR("i915_dispatch_cmdbuffer failed\n"); - goto fail_clip_free; - } - - if (sarea_priv) - sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - -fail_clip_free: - kfree(cliprects); -fail_batch_free: - kfree(batch_data); - - return ret; -} - -static int i915_emit_irq(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; - - i915_kernel_lost_context(dev); - - DRM_DEBUG_DRIVER("\n"); - - dev_priv->dri1.counter++; - if (dev_priv->dri1.counter > 0x7FFFFFFFUL) - dev_priv->dri1.counter = 1; - if (master_priv->sarea_priv) - master_priv->sarea_priv->last_enqueue = dev_priv->dri1.counter; - - if (BEGIN_LP_RING(4) == 0) { - OUT_RING(MI_STORE_DWORD_INDEX); - OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - OUT_RING(dev_priv->dri1.counter); - OUT_RING(MI_USER_INTERRUPT); - ADVANCE_LP_RING(); - } - - return dev_priv->dri1.counter; -} - -static int i915_wait_irq(struct drm_device *dev, int irq_nr) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; - int ret = 0; - struct intel_engine_cs *ring = LP_RING(dev_priv); - - DRM_DEBUG_DRIVER("irq_nr=%d breadcrumb=%d\n", irq_nr, - READ_BREADCRUMB(dev_priv)); - - if (READ_BREADCRUMB(dev_priv) >= irq_nr) { - if (master_priv->sarea_priv) - master_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - return 0; - } - - if (master_priv->sarea_priv) - master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; - - if (ring->irq_get(ring)) { - DRM_WAIT_ON(ret, ring->irq_queue, 3 * HZ, - READ_BREADCRUMB(dev_priv) >= irq_nr); - ring->irq_put(ring); - } else if (wait_for(READ_BREADCRUMB(dev_priv) >= irq_nr, 3000)) - ret = -EBUSY; - - if (ret == -EBUSY) { - DRM_ERROR("EBUSY -- rec: %d emitted: %d\n", - READ_BREADCRUMB(dev_priv), (int)dev_priv->dri1.counter); - } - - return ret; + return -ENODEV; } -/* Needs the lock as it touches the ring. - */ static int i915_irq_emit(struct drm_device *dev, void *data, struct drm_file *file_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - drm_i915_irq_emit_t *emit = data; - int result; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - if (!dev_priv || !LP_RING(dev_priv)->buffer->virtual_start) { - DRM_ERROR("called with no initialization\n"); - return -EINVAL; - } - - RING_LOCK_TEST_WITH_RETURN(dev, file_priv); - - mutex_lock(&dev->struct_mutex); - result = i915_emit_irq(dev); - mutex_unlock(&dev->struct_mutex); - - if (copy_to_user(emit->irq_seq, &result, sizeof(int))) { - DRM_ERROR("copy_to_user\n"); - return -EFAULT; - } - - return 0; + return -ENODEV; } -/* Doesn't need the hardware lock. - */ static int i915_irq_wait(struct drm_device *dev, void *data, struct drm_file *file_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - drm_i915_irq_wait_t *irqwait = data; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - if (!dev_priv) { - DRM_ERROR("called with no initialization\n"); - return -EINVAL; - } - - return i915_wait_irq(dev, irqwait->irq_seq); + return -ENODEV; } static int i915_vblank_pipe_get(struct drm_device *dev, void *data, struct drm_file *file_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - drm_i915_vblank_pipe_t *pipe = data; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - if (!dev_priv) { - DRM_ERROR("called with no initialization\n"); - return -EINVAL; - } - - pipe->pipe = DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B; - - return 0; + return -ENODEV; } -/** - * Schedule buffer swap at given vertical blank. - */ static int i915_vblank_swap(struct drm_device *dev, void *data, struct drm_file *file_priv) { - /* The delayed swap mechanism was fundamentally racy, and has been - * removed. The model was that the client requested a delayed flip/swap - * from the kernel, then waited for vblank before continuing to perform - * rendering. The problem was that the kernel might wake the client - * up before it dispatched the vblank swap (since the lock has to be - * held while touching the ringbuffer), in which case the client would - * clear and start the next frame before the swap occurred, and - * flicker would occur in addition to likely missing the vblank. - * - * In the absence of this ioctl, userland falls back to a correct path - * of waiting for a vblank, then dispatching the swap on its own. - * Context switching to userland and back is plenty fast enough for - * meeting the requirements of vblank swapping. - */ - return -EINVAL; + return -ENODEV; } static int i915_flip_bufs(struct drm_device *dev, void *data, struct drm_file *file_priv) { - int ret; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - DRM_DEBUG_DRIVER("%s\n", __func__); - - RING_LOCK_TEST_WITH_RETURN(dev, file_priv); - - mutex_lock(&dev->struct_mutex); - ret = i915_dispatch_flip(dev); - mutex_unlock(&dev->struct_mutex); - - return ret; + return -ENODEV; } static int i915_getparam(struct drm_device *dev, void *data, @@ -943,14 +118,11 @@ static int i915_getparam(struct drm_device *dev, void *data, switch (param->param) { case I915_PARAM_IRQ_ACTIVE: - value = dev->pdev->irq ? 1 : 0; - break; + return -ENODEV; case I915_PARAM_ALLOW_BATCHBUFFER: - value = dev_priv->dri1.allow_batchbuffer ? 1 : 0; - break; + return -ENODEV; case I915_PARAM_LAST_DISPATCH: - value = READ_BREADCRUMB(dev_priv); - break; + return -ENODEV; case I915_PARAM_CHIPSET_ID: value = dev->pdev->device; break; @@ -1056,12 +228,10 @@ static int i915_setparam(struct drm_device *dev, void *data, switch (param->param) { case I915_SETPARAM_USE_MI_BATCHBUFFER_START: - break; case I915_SETPARAM_TEX_LRU_LOG_GRANULARITY: - break; case I915_SETPARAM_ALLOW_BATCHBUFFER: - dev_priv->dri1.allow_batchbuffer = param->value ? 1 : 0; - break; + return -ENODEV; + case I915_SETPARAM_NUM_USED_FENCES: if (param->value > dev_priv->num_fence_regs || param->value < 0) @@ -1081,49 +251,7 @@ static int i915_setparam(struct drm_device *dev, void *data, static int i915_set_status_page(struct drm_device *dev, void *data, struct drm_file *file_priv) { - struct drm_i915_private *dev_priv = dev->dev_private; - drm_i915_hws_addr_t *hws = data; - struct intel_engine_cs *ring; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - if (!I915_NEED_GFX_HWS(dev)) - return -EINVAL; - - if (!dev_priv) { - DRM_ERROR("called with no initialization\n"); - return -EINVAL; - } - - if (drm_core_check_feature(dev, DRIVER_MODESET)) { - WARN(1, "tried to set status page when mode setting active\n"); - return 0; - } - - DRM_DEBUG_DRIVER("set status page addr 0x%08x\n", (u32)hws->addr); - - ring = LP_RING(dev_priv); - ring->status_page.gfx_addr = hws->addr & (0x1ffff<<12); - - dev_priv->dri1.gfx_hws_cpu_addr = - ioremap_wc(dev_priv->gtt.mappable_base + hws->addr, 4096); - if (dev_priv->dri1.gfx_hws_cpu_addr == NULL) { - i915_dma_cleanup(dev); - ring->status_page.gfx_addr = 0; - DRM_ERROR("can not ioremap virtual address for" - " G33 hw status page\n"); - return -ENOMEM; - } - - memset_io(dev_priv->dri1.gfx_hws_cpu_addr, 0, PAGE_SIZE); - I915_WRITE(HWS_PGA, ring->status_page.gfx_addr); - - DRM_DEBUG_DRIVER("load hws HWS_PGA with gfx mem 0x%x\n", - ring->status_page.gfx_addr); - DRM_DEBUG_DRIVER("load hws at %p\n", - ring->status_page.page_addr); - return 0; + return -ENODEV; } static int i915_get_bridge_dev(struct drm_device *dev) @@ -1401,30 +529,6 @@ out: return ret; } -int i915_master_create(struct drm_device *dev, struct drm_master *master) -{ - struct drm_i915_master_private *master_priv; - - master_priv = kzalloc(sizeof(*master_priv), GFP_KERNEL); - if (!master_priv) - return -ENOMEM; - - master->driver_priv = master_priv; - return 0; -} - -void i915_master_destroy(struct drm_device *dev, struct drm_master *master) -{ - struct drm_i915_master_private *master_priv = master->driver_priv; - - if (!master_priv) - return; - - kfree(master_priv); - - master->driver_priv = NULL; -} - #if IS_ENABLED(CONFIG_FB) static int i915_kick_out_firmware_fb(struct drm_i915_private *dev_priv) { @@ -1899,9 +1003,6 @@ int i915_driver_unload(struct drm_device *dev) i915_gem_context_fini(dev); mutex_unlock(&dev->struct_mutex); i915_gem_cleanup_stolen(dev); - - if (!I915_NEED_GFX_HWS(dev)) - i915_free_hws(dev); } intel_teardown_gmbus(dev); @@ -1966,8 +1067,6 @@ void i915_driver_lastclose(struct drm_device *dev) } i915_gem_lastclose(dev); - - i915_dma_cleanup(dev); } void i915_driver_preclose(struct drm_device *dev, struct drm_file *file) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 19558d6..c9f248a 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1573,8 +1573,6 @@ static struct drm_driver driver = { .resume = i915_resume_legacy, .device_is_agp = i915_driver_device_is_agp, - .master_create = i915_master_create, - .master_destroy = i915_master_destroy, #if defined(CONFIG_DEBUG_FS) .debugfs_init = i915_debugfs_init, .debugfs_cleanup = i915_debugfs_cleanup, diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 1097d20..1002964 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -327,12 +327,6 @@ struct intel_opregion { struct intel_overlay; struct intel_overlay_error_state; -struct drm_local_map; - -struct drm_i915_master_private { - struct drm_local_map *sarea; - struct _drm_i915_sarea *sarea_priv; -}; #define I915_FENCE_REG_NONE -1 #define I915_MAX_NUM_FENCES 32 /* 32 fences + sign bit for FENCE_REG_NONE */ @@ -1127,19 +1121,6 @@ struct i915_power_domains { struct i915_power_well *power_wells; }; -struct i915_dri1_state { - unsigned allow_batchbuffer : 1; - u32 __iomem *gfx_hws_cpu_addr; - - unsigned int cpp; - int back_offset; - int front_offset; - int current_page; - int page_flipping; - - uint32_t counter; -}; - struct i915_ums_state { /** * Flag if the X Server, and thus DRM, is not currently in @@ -1787,9 +1768,6 @@ struct drm_i915_private { uint32_t bios_vgacntr; - /* Old dri1 support infrastructure, beware the dragons ya fools entering - * here! */ - struct i915_dri1_state dri1; /* Old ums support infrastructure, same warning applies. */ struct i915_ums_state ums; @@ -2351,8 +2329,6 @@ struct i915_params { extern struct i915_params i915 __read_mostly; /* i915_dma.c */ -void i915_update_dri1_breadcrumb(struct drm_device *dev); -extern void i915_kernel_lost_context(struct drm_device * dev); extern int i915_driver_load(struct drm_device *, unsigned long flags); extern int i915_driver_unload(struct drm_device *); extern int i915_driver_open(struct drm_device *dev, struct drm_file *file); @@ -2366,9 +2342,6 @@ extern int i915_driver_device_is_agp(struct drm_device * dev); extern long i915_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); #endif -extern int i915_emit_box(struct drm_device *dev, - struct drm_clip_rect *box, - int DR1, int DR4); extern int intel_gpu_reset(struct drm_device *dev); extern int i915_reset(struct drm_device *dev); extern unsigned long i915_chipset_val(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 7f95d8f..7985f7b 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4664,7 +4664,6 @@ i915_gem_suspend(struct drm_device *dev) if (!drm_core_check_feature(dev, DRIVER_MODESET)) i915_gem_evict_everything(dev); - i915_kernel_lost_context(dev); i915_gem_stop_ringbuffers(dev); /* Hack! Don't let anybody do execbuf while we don't control the chip. @@ -4963,9 +4962,6 @@ int i915_gem_init(struct drm_device *dev) } mutex_unlock(&dev->struct_mutex); - /* Allow hardware batchbuffers unless told otherwise, but not for KMS. */ - if (!drm_core_check_feature(dev, DRIVER_MODESET)) - dev_priv->dri1.allow_batchbuffer = 1; return ret; } diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index e1ed85a..b16eee0 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -1020,6 +1020,47 @@ i915_reset_gen7_sol_offsets(struct drm_device *dev, return 0; } +static int +i915_emit_box(struct intel_engine_cs *ring, + struct drm_clip_rect *box, + int DR1, int DR4) +{ + int ret; + + if (box->y2 <= box->y1 || box->x2 <= box->x1 || + box->y2 <= 0 || box->x2 <= 0) { + DRM_ERROR("Bad box %d,%d..%d,%d\n", + box->x1, box->y1, box->x2, box->y2); + return -EINVAL; + } + + if (INTEL_INFO(ring->dev)->gen >= 4) { + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_DRAWRECT_INFO_I965); + intel_ring_emit(ring, (box->x1 & 0xffff) | box->y1 << 16); + intel_ring_emit(ring, ((box->x2 - 1) & 0xffff) | (box->y2 - 1) << 16); + intel_ring_emit(ring, DR4); + } else { + ret = intel_ring_begin(ring, 6); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_DRAWRECT_INFO); + intel_ring_emit(ring, DR1); + intel_ring_emit(ring, (box->x1 & 0xffff) | box->y1 << 16); + intel_ring_emit(ring, ((box->x2 - 1) & 0xffff) | (box->y2 - 1) << 16); + intel_ring_emit(ring, DR4); + intel_ring_emit(ring, 0); + } + intel_ring_advance(ring); + + return 0; +} + + int i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file, struct intel_engine_cs *ring, @@ -1148,7 +1189,7 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file, exec_len = args->batch_len; if (cliprects) { for (i = 0; i < args->num_cliprects; i++) { - ret = i915_emit_box(dev, &cliprects[i], + ret = i915_emit_box(ring, &cliprects[i], args->DR1, args->DR4); if (ret) goto error; diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 283756f..8d169e1 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3764,8 +3764,6 @@ static irqreturn_t i8xx_irq_handler(int irq, void *arg) I915_WRITE16(IIR, iir & ~flip_mask); new_iir = I915_READ16(IIR); /* Flush posted writes */ - i915_update_dri1_breadcrumb(dev); - if (iir & I915_USER_INTERRUPT) notify_ring(dev, &dev_priv->ring[RCS]); @@ -4002,8 +4000,6 @@ static irqreturn_t i915_irq_handler(int irq, void *arg) iir = new_iir; } while (iir & ~flip_mask); - i915_update_dri1_breadcrumb(dev); - return ret; } @@ -4231,8 +4227,6 @@ static irqreturn_t i965_irq_handler(int irq, void *arg) iir = new_iir; } - i915_update_dri1_breadcrumb(dev); - return ret; } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 6060b71..9c983e4 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -5185,36 +5185,6 @@ static void i9xx_crtc_off(struct drm_crtc *crtc) { } -static void intel_crtc_update_sarea(struct drm_crtc *crtc, - bool enabled) -{ - struct drm_device *dev = crtc->dev; - struct drm_i915_master_private *master_priv; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; - - if (!dev->primary->master) - return; - - master_priv = dev->primary->master->driver_priv; - if (!master_priv->sarea_priv) - return; - - switch (pipe) { - case 0: - master_priv->sarea_priv->pipeA_w = enabled ? crtc->mode.hdisplay : 0; - master_priv->sarea_priv->pipeA_h = enabled ? crtc->mode.vdisplay : 0; - break; - case 1: - master_priv->sarea_priv->pipeB_w = enabled ? crtc->mode.hdisplay : 0; - master_priv->sarea_priv->pipeB_h = enabled ? crtc->mode.vdisplay : 0; - break; - default: - DRM_ERROR("Can't update pipe %c in SAREA\n", pipe_name(pipe)); - break; - } -} - /* Master function to enable/disable CRTC and corresponding power wells */ void intel_crtc_control(struct drm_crtc *crtc, bool enable) { @@ -5258,8 +5228,6 @@ void intel_crtc_update_dpms(struct drm_crtc *crtc) enable |= intel_encoder->connectors_active; intel_crtc_control(crtc, enable); - - intel_crtc_update_sarea(crtc, enable); } static void intel_crtc_disable(struct drm_crtc *crtc) @@ -5274,7 +5242,6 @@ static void intel_crtc_disable(struct drm_crtc *crtc) WARN_ON(!crtc->enabled); dev_priv->display.crtc_disable(crtc); - intel_crtc_update_sarea(crtc, false); dev_priv->display.off(crtc); if (crtc->primary->fb) { @@ -8369,7 +8336,7 @@ static int intel_crtc_cursor_set_obj(struct drm_crtc *crtc, uint32_t width, uint32_t height) { struct drm_device *dev = crtc->dev; - struct drm_i915_private *dev_priv = dev->dev_private; + struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); enum pipe pipe = intel_crtc->pipe; unsigned old_width; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 0a4f35e..1d01b51 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -589,14 +589,10 @@ static int init_ring_common(struct intel_engine_cs *ring) goto out; } - if (!drm_core_check_feature(ring->dev, DRIVER_MODESET)) - i915_kernel_lost_context(ring->dev); - else { - ringbuf->head = I915_READ_HEAD(ring); - ringbuf->tail = I915_READ_TAIL(ring) & TAIL_ADDR; - ringbuf->space = intel_ring_space(ringbuf); - ringbuf->last_retired_head = -1; - } + ringbuf->head = I915_READ_HEAD(ring); + ringbuf->tail = I915_READ_TAIL(ring) & TAIL_ADDR; + ringbuf->space = intel_ring_space(ringbuf); + ringbuf->last_retired_head = -1; memset(&ring->hangcheck, 0, sizeof(ring->hangcheck)); @@ -1958,13 +1954,6 @@ static int ring_wait_for_space(struct intel_engine_cs *ring, int n) break; } - if (!drm_core_check_feature(dev, DRIVER_MODESET) && - dev->primary->master) { - struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; - if (master_priv->sarea_priv) - master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; - } - msleep(1); if (dev_priv->mm.interruptible && signal_pending(current)) { @@ -2455,91 +2444,6 @@ int intel_init_render_ring_buffer(struct drm_device *dev) return intel_init_ring_buffer(dev, ring); } -int intel_render_ring_init_dri(struct drm_device *dev, u64 start, u32 size) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_engine_cs *ring = &dev_priv->ring[RCS]; - struct intel_ringbuffer *ringbuf = ring->buffer; - int ret; - - if (ringbuf == NULL) { - ringbuf = kzalloc(sizeof(*ringbuf), GFP_KERNEL); - if (!ringbuf) - return -ENOMEM; - ring->buffer = ringbuf; - } - - ring->name = "render ring"; - ring->id = RCS; - ring->mmio_base = RENDER_RING_BASE; - - if (INTEL_INFO(dev)->gen >= 6) { - /* non-kms not supported on gen6+ */ - ret = -ENODEV; - goto err_ringbuf; - } - - /* Note: gem is not supported on gen5/ilk without kms (the corresponding - * gem_init ioctl returns with -ENODEV). Hence we do not need to set up - * the special gen5 functions. */ - ring->add_request = i9xx_add_request; - if (INTEL_INFO(dev)->gen < 4) - ring->flush = gen2_render_ring_flush; - else - ring->flush = gen4_render_ring_flush; - ring->get_seqno = ring_get_seqno; - ring->set_seqno = ring_set_seqno; - if (IS_GEN2(dev)) { - ring->irq_get = i8xx_ring_get_irq; - ring->irq_put = i8xx_ring_put_irq; - } else { - ring->irq_get = i9xx_ring_get_irq; - ring->irq_put = i9xx_ring_put_irq; - } - ring->irq_enable_mask = I915_USER_INTERRUPT; - ring->write_tail = ring_write_tail; - if (INTEL_INFO(dev)->gen >= 4) - ring->dispatch_execbuffer = i965_dispatch_execbuffer; - else if (IS_I830(dev) || IS_845G(dev)) - ring->dispatch_execbuffer = i830_dispatch_execbuffer; - else - ring->dispatch_execbuffer = i915_dispatch_execbuffer; - ring->init = init_render_ring; - ring->cleanup = render_ring_cleanup; - - ring->dev = dev; - INIT_LIST_HEAD(&ring->active_list); - INIT_LIST_HEAD(&ring->request_list); - - ringbuf->size = size; - ringbuf->effective_size = ringbuf->size; - if (IS_I830(ring->dev) || IS_845G(ring->dev)) - ringbuf->effective_size -= 2 * CACHELINE_BYTES; - - ringbuf->virtual_start = ioremap_wc(start, size); - if (ringbuf->virtual_start == NULL) { - DRM_ERROR("can not ioremap virtual address for" - " ring buffer\n"); - ret = -ENOMEM; - goto err_ringbuf; - } - - if (!I915_NEED_GFX_HWS(dev)) { - ret = init_phys_status_page(ring); - if (ret) - goto err_vstart; - } - - return 0; - -err_vstart: - iounmap(ringbuf->virtual_start); -err_ringbuf: - kfree(ringbuf); - ring->buffer = NULL; - return ret; -} - int intel_init_bsd_ring_buffer(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 9eb3188..fe426cf 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -448,7 +448,4 @@ static inline void i915_trace_irq_get(struct intel_engine_cs *ring, u32 seqno) ring->trace_irq_seqno = seqno; } -/* DRI warts */ -int intel_render_ring_init_dri(struct drm_device *dev, u64 start, u32 size); - #endif /* _INTEL_RINGBUFFER_H_ */ -- cgit v0.10.2 From 77f31815e1d0a9ae2cb073fc5f9ef33e5f3df059 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 21:23:55 +0100 Subject: drm/i915: Replace dri1 functions with drm_noop Whether we'll reject them or no-op doesn't really matter ... Acked-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index d156228..887df56 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -50,59 +50,6 @@ #include #include -static int i915_dma_init(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - -static int i915_flush_ioctl(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - -static int i915_batchbuffer(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - -static int i915_cmdbuffer(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - -static int i915_irq_emit(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - -static int i915_irq_wait(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - -static int i915_vblank_pipe_get(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - -static int i915_vblank_swap(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - -static int i915_flip_bufs(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} static int i915_getparam(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -248,12 +195,6 @@ static int i915_setparam(struct drm_device *dev, void *data, return 0; } -static int i915_set_status_page(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - return -ENODEV; -} - static int i915_get_bridge_dev(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -1090,23 +1031,23 @@ void i915_driver_postclose(struct drm_device *dev, struct drm_file *file) } const struct drm_ioctl_desc i915_ioctls[] = { - DRM_IOCTL_DEF_DRV(I915_INIT, i915_dma_init, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), - DRM_IOCTL_DEF_DRV(I915_FLUSH, i915_flush_ioctl, DRM_AUTH), - DRM_IOCTL_DEF_DRV(I915_FLIP, i915_flip_bufs, DRM_AUTH), - DRM_IOCTL_DEF_DRV(I915_BATCHBUFFER, i915_batchbuffer, DRM_AUTH), - DRM_IOCTL_DEF_DRV(I915_IRQ_EMIT, i915_irq_emit, DRM_AUTH), - DRM_IOCTL_DEF_DRV(I915_IRQ_WAIT, i915_irq_wait, DRM_AUTH), + DRM_IOCTL_DEF_DRV(I915_INIT, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), + DRM_IOCTL_DEF_DRV(I915_FLUSH, drm_noop, DRM_AUTH), + DRM_IOCTL_DEF_DRV(I915_FLIP, drm_noop, DRM_AUTH), + DRM_IOCTL_DEF_DRV(I915_BATCHBUFFER, drm_noop, DRM_AUTH), + DRM_IOCTL_DEF_DRV(I915_IRQ_EMIT, drm_noop, DRM_AUTH), + DRM_IOCTL_DEF_DRV(I915_IRQ_WAIT, drm_noop, DRM_AUTH), DRM_IOCTL_DEF_DRV(I915_GETPARAM, i915_getparam, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(I915_SETPARAM, i915_setparam, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF_DRV(I915_ALLOC, drm_noop, DRM_AUTH), DRM_IOCTL_DEF_DRV(I915_FREE, drm_noop, DRM_AUTH), DRM_IOCTL_DEF_DRV(I915_INIT_HEAP, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), - DRM_IOCTL_DEF_DRV(I915_CMDBUFFER, i915_cmdbuffer, DRM_AUTH), + DRM_IOCTL_DEF_DRV(I915_CMDBUFFER, drm_noop, DRM_AUTH), DRM_IOCTL_DEF_DRV(I915_DESTROY_HEAP, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF_DRV(I915_SET_VBLANK_PIPE, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), - DRM_IOCTL_DEF_DRV(I915_GET_VBLANK_PIPE, i915_vblank_pipe_get, DRM_AUTH), - DRM_IOCTL_DEF_DRV(I915_VBLANK_SWAP, i915_vblank_swap, DRM_AUTH), - DRM_IOCTL_DEF_DRV(I915_HWS_ADDR, i915_set_status_page, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), + DRM_IOCTL_DEF_DRV(I915_GET_VBLANK_PIPE, drm_noop, DRM_AUTH), + DRM_IOCTL_DEF_DRV(I915_VBLANK_SWAP, drm_noop, DRM_AUTH), + DRM_IOCTL_DEF_DRV(I915_HWS_ADDR, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF_DRV(I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY|DRM_UNLOCKED), DRM_IOCTL_DEF_DRV(I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH|DRM_UNLOCKED), DRM_IOCTL_DEF_DRV(I915_GEM_EXECBUFFER2, i915_gem_execbuffer2, DRM_AUTH|DRM_UNLOCKED|DRM_RENDER_ALLOW), -- cgit v0.10.2 From ac883c84e45f44a671e6ebe2a4be41b3167c0261 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 21:24:54 +0100 Subject: drm/i915: Drop checks for initialization KMS always intializes, this was only a valid check when userspace was still in control of the kernel driver. v2: Comment that we outright reject all dri1/ums params. Acked-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 887df56..69d6c76 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -58,17 +58,11 @@ static int i915_getparam(struct drm_device *dev, void *data, drm_i915_getparam_t *param = data; int value; - if (!dev_priv) { - DRM_ERROR("called with no initialization\n"); - return -EINVAL; - } - switch (param->param) { case I915_PARAM_IRQ_ACTIVE: - return -ENODEV; case I915_PARAM_ALLOW_BATCHBUFFER: - return -ENODEV; case I915_PARAM_LAST_DISPATCH: + /* Reject all old ums/dri params. */ return -ENODEV; case I915_PARAM_CHIPSET_ID: value = dev->pdev->device; @@ -168,15 +162,11 @@ static int i915_setparam(struct drm_device *dev, void *data, struct drm_i915_private *dev_priv = dev->dev_private; drm_i915_setparam_t *param = data; - if (!dev_priv) { - DRM_ERROR("called with no initialization\n"); - return -EINVAL; - } - switch (param->param) { case I915_SETPARAM_USE_MI_BATCHBUFFER_START: case I915_SETPARAM_TEX_LRU_LOG_GRANULARITY: case I915_SETPARAM_ALLOW_BATCHBUFFER: + /* Reject all old ums/dri params. */ return -ENODEV; case I915_SETPARAM_NUM_USED_FENCES: -- cgit v0.10.2 From 71b14ab6184cc618d4c3557205113d3df51e2197 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 20:36:47 +0100 Subject: drm/i915: No-Op enter/leave vt gem ioctl We've killed ums support by now, it's time to reap the benefits. This one here is getting in the way of doing some ring init cleanup. Acked-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 69d6c76..6c96253 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1047,8 +1047,8 @@ const struct drm_ioctl_desc i915_ioctls[] = { DRM_IOCTL_DEF_DRV(I915_GEM_SET_CACHING, i915_gem_set_caching_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(I915_GEM_GET_CACHING, i915_gem_get_caching_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(I915_GEM_THROTTLE, i915_gem_throttle_ioctl, DRM_AUTH|DRM_UNLOCKED|DRM_RENDER_ALLOW), - DRM_IOCTL_DEF_DRV(I915_GEM_ENTERVT, i915_gem_entervt_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY|DRM_UNLOCKED), - DRM_IOCTL_DEF_DRV(I915_GEM_LEAVEVT, i915_gem_leavevt_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY|DRM_UNLOCKED), + DRM_IOCTL_DEF_DRV(I915_GEM_ENTERVT, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY|DRM_UNLOCKED), + DRM_IOCTL_DEF_DRV(I915_GEM_LEAVEVT, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY|DRM_UNLOCKED), DRM_IOCTL_DEF_DRV(I915_GEM_CREATE, i915_gem_create_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(I915_GEM_PREAD, i915_gem_pread_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(I915_GEM_PWRITE, i915_gem_pwrite_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW), diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 1002964..a52fba1 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2441,10 +2441,6 @@ int i915_gem_throttle_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv); int i915_gem_madvise_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv); -int i915_gem_entervt_ioctl(struct drm_device *dev, void *data, - struct drm_file *file_priv); -int i915_gem_leavevt_ioctl(struct drm_device *dev, void *data, - struct drm_file *file_priv); int i915_gem_set_tiling(struct drm_device *dev, void *data, struct drm_file *file_priv); int i915_gem_get_tiling(struct drm_device *dev, void *data, diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 7985f7b..bd17109 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4976,61 +4976,6 @@ i915_gem_cleanup_ringbuffer(struct drm_device *dev) dev_priv->gt.cleanup_ring(ring); } -int -i915_gem_entervt_ioctl(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int ret; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return 0; - - if (i915_reset_in_progress(&dev_priv->gpu_error)) { - DRM_ERROR("Reenabling wedged hardware, good luck\n"); - atomic_set(&dev_priv->gpu_error.reset_counter, 0); - } - - mutex_lock(&dev->struct_mutex); - dev_priv->ums.mm_suspended = 0; - - ret = i915_gem_init_hw(dev); - if (ret != 0) { - mutex_unlock(&dev->struct_mutex); - return ret; - } - - BUG_ON(!list_empty(&dev_priv->gtt.base.active_list)); - - ret = drm_irq_install(dev, dev->pdev->irq); - if (ret) - goto cleanup_ringbuffer; - mutex_unlock(&dev->struct_mutex); - - return 0; - -cleanup_ringbuffer: - i915_gem_cleanup_ringbuffer(dev); - dev_priv->ums.mm_suspended = 1; - mutex_unlock(&dev->struct_mutex); - - return ret; -} - -int -i915_gem_leavevt_ioctl(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return 0; - - mutex_lock(&dev->struct_mutex); - drm_irq_uninstall(dev); - mutex_unlock(&dev->struct_mutex); - - return i915_gem_suspend(dev); -} - void i915_gem_lastclose(struct drm_device *dev) { -- cgit v0.10.2 From fbd2f9fe97726944f98d9a5813d08372446b6a37 Mon Sep 17 00:00:00 2001 From: Gerd Hoffmann Date: Wed, 19 Nov 2014 12:28:09 +0100 Subject: bochs: little cleanup Drop some leftover, commented code. Signed-off-by: Gerd Hoffmann Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/bochs/bochs_hw.c b/drivers/gpu/drm/bochs/bochs_hw.c index dbe619e..e8b1dacc 100644 --- a/drivers/gpu/drm/bochs/bochs_hw.c +++ b/drivers/gpu/drm/bochs/bochs_hw.c @@ -54,8 +54,7 @@ int bochs_hw_init(struct drm_device *dev, uint32_t flags) unsigned long addr, size, mem, ioaddr, iosize; u16 id; - if (/* (ent->driver_data == BOCHS_QEMU_STDVGA) && */ - (pdev->resource[2].flags & IORESOURCE_MEM)) { + if (pdev->resource[2].flags & IORESOURCE_MEM) { /* mmio bar with vga and bochs registers present */ if (pci_request_region(pdev, 2, "bochs-drm") != 0) { DRM_ERROR("Cannot request mmio region\n"); -- cgit v0.10.2 From 9ecdb039b7517dc10b8c3e6dbeb40859178ac28e Mon Sep 17 00:00:00 2001 From: Gerd Hoffmann Date: Wed, 19 Nov 2014 12:28:10 +0100 Subject: bochs: add endian switching support Recently (qemu 2.2+) the qemu stdvga got a register to switch the vga framebuffer endianness. This patch adds code to explicitly set the endianness of the framebuffer. In most cases this has no effect as the default is guest architecture endianness. It is needed though in case a architecture supports both big and little endian, i.e. for ppc64le. Signed-off-by: Gerd Hoffmann Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/bochs/bochs_hw.c b/drivers/gpu/drm/bochs/bochs_hw.c index e8b1dacc..4603897 100644 --- a/drivers/gpu/drm/bochs/bochs_hw.c +++ b/drivers/gpu/drm/bochs/bochs_hw.c @@ -51,7 +51,7 @@ int bochs_hw_init(struct drm_device *dev, uint32_t flags) { struct bochs_device *bochs = dev->dev_private; struct pci_dev *pdev = dev->pdev; - unsigned long addr, size, mem, ioaddr, iosize; + unsigned long addr, size, mem, ioaddr, iosize, qext_size; u16 id; if (pdev->resource[2].flags & IORESOURCE_MEM) { @@ -115,6 +115,24 @@ int bochs_hw_init(struct drm_device *dev, uint32_t flags) size / 1024, addr, bochs->ioports ? "ioports" : "mmio", ioaddr); + + if (bochs->mmio && pdev->revision >= 2) { + qext_size = readl(bochs->mmio + 0x600); + if (qext_size < 4 || qext_size > iosize) + goto noext; + DRM_DEBUG("Found qemu ext regs, size %ld\n", qext_size); + if (qext_size >= 8) { +#ifdef __BIG_ENDIAN + writel(0xbebebebe, bochs->mmio + 0x604); +#else + writel(0x1e1e1e1e, bochs->mmio + 0x604); +#endif + DRM_DEBUG(" qext endian: 0x%x\n", + readl(bochs->mmio + 0x604)); + } + } + +noext: return 0; } -- cgit v0.10.2 From 49b70a31c166a6da12f60ec85da608c064555406 Mon Sep 17 00:00:00 2001 From: Gerd Hoffmann Date: Wed, 19 Nov 2014 12:28:11 +0100 Subject: bochs: fix bochsdrmfb mmap Remove the mapping offset from the bo backing the fbdev framebuffer. Wire up fbdev mmap function to map the backing bo using ttm_fbdev_mmap. With that patch in place mmap(/dev/fb0) works as expected. Signed-off-by: Gerd Hoffmann Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/bochs/bochs_fbdev.c b/drivers/gpu/drm/bochs/bochs_fbdev.c index fe95d31..61dbf09 100644 --- a/drivers/gpu/drm/bochs/bochs_fbdev.c +++ b/drivers/gpu/drm/bochs/bochs_fbdev.c @@ -9,6 +9,17 @@ /* ---------------------------------------------------------------------- */ +static int bochsfb_mmap(struct fb_info *info, + struct vm_area_struct *vma) +{ + struct drm_fb_helper *fb_helper = info->par; + struct bochs_device *bochs = + container_of(fb_helper, struct bochs_device, fb.helper); + struct bochs_bo *bo = gem_to_bochs_bo(bochs->fb.gfb.obj); + + return ttm_fbdev_mmap(vma, &bo->bo); +} + static struct fb_ops bochsfb_ops = { .owner = THIS_MODULE, .fb_check_var = drm_fb_helper_check_var, @@ -19,6 +30,7 @@ static struct fb_ops bochsfb_ops = { .fb_pan_display = drm_fb_helper_pan_display, .fb_blank = drm_fb_helper_blank, .fb_setcmap = drm_fb_helper_setcmap, + .fb_mmap = bochsfb_mmap, }; static int bochsfb_create_object(struct bochs_device *bochs, @@ -123,11 +135,9 @@ static int bochsfb_create(struct drm_fb_helper *helper, info->screen_base = bo->kmap.virtual; info->screen_size = size; -#if 0 - /* FIXME: get this right for mmap(/dev/fb0) */ - info->fix.smem_start = bochs_bo_mmap_offset(bo); + drm_vma_offset_remove(&bo->bo.bdev->vma_manager, &bo->bo.vma_node); + info->fix.smem_start = 0; info->fix.smem_len = size; -#endif ret = fb_alloc_cmap(&info->cmap, 256, 0); if (ret) { -- cgit v0.10.2 From 3ce733b909764a3ee17b35d7a2c940fd0bc45d5b Mon Sep 17 00:00:00 2001 From: Gerd Hoffmann Date: Wed, 19 Nov 2014 12:28:12 +0100 Subject: bochs: add page_flip Implement crtc page_flip callback for bochsdrm. The qemu stdvga has no vblank signaling, so we have to fake it. We do so by instantly calling drm_send_vblank_event. Tested with kmscon. Signed-off-by: Gerd Hoffmann Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/bochs/bochs_kms.c b/drivers/gpu/drm/bochs/bochs_kms.c index 5ffd489..85f0f8cf 100644 --- a/drivers/gpu/drm/bochs/bochs_kms.c +++ b/drivers/gpu/drm/bochs/bochs_kms.c @@ -109,11 +109,32 @@ static void bochs_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, { } +static int bochs_crtc_page_flip(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_pending_vblank_event *event, + uint32_t page_flip_flags) +{ + struct bochs_device *bochs = + container_of(crtc, struct bochs_device, crtc); + struct drm_framebuffer *old_fb = crtc->primary->fb; + unsigned long irqflags; + + crtc->primary->fb = fb; + bochs_crtc_mode_set_base(crtc, 0, 0, old_fb); + if (event) { + spin_lock_irqsave(&bochs->dev->event_lock, irqflags); + drm_send_vblank_event(bochs->dev, -1, event); + spin_unlock_irqrestore(&bochs->dev->event_lock, irqflags); + } + return 0; +} + /* These provide the minimum set of functions required to handle a CRTC */ static const struct drm_crtc_funcs bochs_crtc_funcs = { .gamma_set = bochs_crtc_gamma_set, .set_config = drm_crtc_helper_set_config, .destroy = drm_crtc_cleanup, + .page_flip = bochs_crtc_page_flip, }; static const struct drm_crtc_helper_funcs bochs_helper_funcs = { -- cgit v0.10.2 From aeb24cc0a31646eb9233a05834440fdf16a713bd Mon Sep 17 00:00:00 2001 From: Boris BREZILLON Date: Mon, 17 Nov 2014 11:21:08 +0100 Subject: drm: omapdrm: remove unused variable Commit f9b9faf6d94dd29eab8c128905c7d091f955481d "drm: flip-work: change drm_flip_work_init prototype" changed the drm_flip_work_init prototype to a void function, which makes 'ret' an unused variable. Signed-off-by: Boris Brezillon Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/omapdrm/omap_plane.c b/drivers/gpu/drm/omapdrm/omap_plane.c index 0ad7401..ee8e2b3 100644 --- a/drivers/gpu/drm/omapdrm/omap_plane.c +++ b/drivers/gpu/drm/omapdrm/omap_plane.c @@ -388,7 +388,6 @@ struct drm_plane *omap_plane_init(struct drm_device *dev, struct drm_plane *plane = NULL; struct omap_plane *omap_plane; struct omap_overlay_info *info; - int ret; DBG("%s: priv=%d", plane_names[id], private_plane); -- cgit v0.10.2 From 2b0a3c400033c23ef83d2e9191d36c250289a79e Mon Sep 17 00:00:00 2001 From: Fabian Frederick Date: Sun, 16 Nov 2014 11:21:25 +0100 Subject: drm: remove unnecessary sizeof(u8) sizeof(u8) is always 1. Signed-off-by: Fabian Frederick Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/r128/r128_state.c b/drivers/gpu/drm/r128/r128_state.c index 575e986..8fd2d9f 100644 --- a/drivers/gpu/drm/r128/r128_state.c +++ b/drivers/gpu/drm/r128/r128_state.c @@ -905,7 +905,7 @@ static int r128_cce_dispatch_write_span(struct drm_device *dev, if (IS_ERR(buffer)) return PTR_ERR(buffer); - mask_size = depth->n * sizeof(u8); + mask_size = depth->n; if (depth->mask) { mask = memdup_user(depth->mask, mask_size); if (IS_ERR(mask)) { @@ -1010,7 +1010,7 @@ static int r128_cce_dispatch_write_pixels(struct drm_device *dev, } if (depth->mask) { - mask_size = depth->n * sizeof(u8); + mask_size = depth->n; mask = memdup_user(depth->mask, mask_size); if (IS_ERR(mask)) { kfree(x); -- cgit v0.10.2 From 881fdaa5e4cb0d68e52acab0ad4e1820e2bfffa4 Mon Sep 17 00:00:00 2001 From: Tetsuo Handa Date: Thu, 13 Nov 2014 22:43:23 +0900 Subject: drm/ttm: Avoid memory allocation from shrinker functions. Andrew Morton wrote: > On Wed, 12 Nov 2014 13:08:55 +0900 Tetsuo Handa wrote: > > > Andrew Morton wrote: > > > Poor ttm guys - this is a bit of a trap we set for them. > > > > Commit a91576d7916f6cce ("drm/ttm: Pass GFP flags in order to avoid deadlock.") > > changed to use sc->gfp_mask rather than GFP_KERNEL. > > > > - pages_to_free = kmalloc(npages_to_free * sizeof(struct page *), > > - GFP_KERNEL); > > + pages_to_free = kmalloc(npages_to_free * sizeof(struct page *), gfp); > > > > But this bug is caused by sc->gfp_mask containing some flags which are not > > in GFP_KERNEL, right? Then, I think > > > > - pages_to_free = kmalloc(npages_to_free * sizeof(struct page *), gfp); > > + pages_to_free = kmalloc(npages_to_free * sizeof(struct page *), gfp & GFP_KERNEL); > > > > would hide this bug. > > > > But I think we should use GFP_ATOMIC (or drop __GFP_WAIT flag) > > Well no - ttm_page_pool_free() should stop calling kmalloc altogether. > Just do > > struct page *pages_to_free[16]; > > and rework the code to free 16 pages at a time. Easy. Well, ttm code wants to process 512 pages at a time for performance. Memory footprint increased by 512 * sizeof(struct page *) buffer is only 4096 bytes. What about using static buffer like below? ---------- >From d3cb5393c9c8099d6b37e769f78c31af1541fe8c Mon Sep 17 00:00:00 2001 From: Tetsuo Handa Date: Thu, 13 Nov 2014 22:21:54 +0900 Subject: [PATCH] drm/ttm: Avoid memory allocation from shrinker functions. Commit a91576d7916f6cce ("drm/ttm: Pass GFP flags in order to avoid deadlock.") caused BUG_ON() due to sc->gfp_mask containing flags which are not in GFP_KERNEL. https://bugzilla.kernel.org/show_bug.cgi?id=87891 Changing from sc->gfp_mask to (sc->gfp_mask & GFP_KERNEL) would avoid the BUG_ON(), but avoiding memory allocation from shrinker function is better and reliable fix. Shrinker function is already serialized by global lock, and clean up function is called after shrinker function is unregistered. Thus, we can use static buffer when called from shrinker function and clean up function. Signed-off-by: Tetsuo Handa Cc: stable [2.6.35+] Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index 09874d6..025c429 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -297,11 +297,12 @@ static void ttm_pool_update_free_locked(struct ttm_page_pool *pool, * * @pool: to free the pages from * @free_all: If set to true will free all pages in pool - * @gfp: GFP flags. + * @use_static: Safe to use static buffer **/ static int ttm_page_pool_free(struct ttm_page_pool *pool, unsigned nr_free, - gfp_t gfp) + bool use_static) { + static struct page *static_buf[NUM_PAGES_TO_ALLOC]; unsigned long irq_flags; struct page *p; struct page **pages_to_free; @@ -311,7 +312,11 @@ static int ttm_page_pool_free(struct ttm_page_pool *pool, unsigned nr_free, if (NUM_PAGES_TO_ALLOC < nr_free) npages_to_free = NUM_PAGES_TO_ALLOC; - pages_to_free = kmalloc(npages_to_free * sizeof(struct page *), gfp); + if (use_static) + pages_to_free = static_buf; + else + pages_to_free = kmalloc(npages_to_free * sizeof(struct page *), + GFP_KERNEL); if (!pages_to_free) { pr_err("Failed to allocate memory for pool free operation\n"); return 0; @@ -374,7 +379,8 @@ restart: if (freed_pages) ttm_pages_put(pages_to_free, freed_pages); out: - kfree(pages_to_free); + if (pages_to_free != static_buf) + kfree(pages_to_free); return nr_free; } @@ -383,8 +389,6 @@ out: * * XXX: (dchinner) Deadlock warning! * - * We need to pass sc->gfp_mask to ttm_page_pool_free(). - * * This code is crying out for a shrinker per pool.... */ static unsigned long @@ -407,8 +411,8 @@ ttm_pool_shrink_scan(struct shrinker *shrink, struct shrink_control *sc) if (shrink_pages == 0) break; pool = &_manager->pools[(i + pool_offset)%NUM_POOLS]; - shrink_pages = ttm_page_pool_free(pool, nr_free, - sc->gfp_mask); + /* OK to use static buffer since global mutex is held. */ + shrink_pages = ttm_page_pool_free(pool, nr_free, true); freed += nr_free - shrink_pages; } mutex_unlock(&lock); @@ -710,7 +714,7 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags, } spin_unlock_irqrestore(&pool->lock, irq_flags); if (npages) - ttm_page_pool_free(pool, npages, GFP_KERNEL); + ttm_page_pool_free(pool, npages, false); } /* @@ -849,9 +853,9 @@ void ttm_page_alloc_fini(void) pr_info("Finalizing pool allocator\n"); ttm_pool_mm_shrink_fini(_manager); + /* OK to use static buffer since global mutex is no longer used. */ for (i = 0; i < NUM_POOLS; ++i) - ttm_page_pool_free(&_manager->pools[i], FREE_ALL_PAGES, - GFP_KERNEL); + ttm_page_pool_free(&_manager->pools[i], FREE_ALL_PAGES, true); kobject_put(&_manager->kobj); _manager = NULL; diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c index c96db43..01e1d27 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c @@ -411,11 +411,12 @@ static void ttm_dma_page_put(struct dma_pool *pool, struct dma_page *d_page) * * @pool: to free the pages from * @nr_free: If set to true will free all pages in pool - * @gfp: GFP flags. + * @use_static: Safe to use static buffer **/ static unsigned ttm_dma_page_pool_free(struct dma_pool *pool, unsigned nr_free, - gfp_t gfp) + bool use_static) { + static struct page *static_buf[NUM_PAGES_TO_ALLOC]; unsigned long irq_flags; struct dma_page *dma_p, *tmp; struct page **pages_to_free; @@ -432,7 +433,11 @@ static unsigned ttm_dma_page_pool_free(struct dma_pool *pool, unsigned nr_free, npages_to_free, nr_free); } #endif - pages_to_free = kmalloc(npages_to_free * sizeof(struct page *), gfp); + if (use_static) + pages_to_free = static_buf; + else + pages_to_free = kmalloc(npages_to_free * sizeof(struct page *), + GFP_KERNEL); if (!pages_to_free) { pr_err("%s: Failed to allocate memory for pool free operation\n", @@ -502,7 +507,8 @@ restart: if (freed_pages) ttm_dma_pages_put(pool, &d_pages, pages_to_free, freed_pages); out: - kfree(pages_to_free); + if (pages_to_free != static_buf) + kfree(pages_to_free); return nr_free; } @@ -531,7 +537,8 @@ static void ttm_dma_free_pool(struct device *dev, enum pool_type type) if (pool->type != type) continue; /* Takes a spinlock.. */ - ttm_dma_page_pool_free(pool, FREE_ALL_PAGES, GFP_KERNEL); + /* OK to use static buffer since global mutex is held. */ + ttm_dma_page_pool_free(pool, FREE_ALL_PAGES, true); WARN_ON(((pool->npages_in_use + pool->npages_free) != 0)); /* This code path is called after _all_ references to the * struct device has been dropped - so nobody should be @@ -986,7 +993,7 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev) /* shrink pool if necessary (only on !is_cached pools)*/ if (npages) - ttm_dma_page_pool_free(pool, npages, GFP_KERNEL); + ttm_dma_page_pool_free(pool, npages, false); ttm->state = tt_unpopulated; } EXPORT_SYMBOL_GPL(ttm_dma_unpopulate); @@ -996,8 +1003,6 @@ EXPORT_SYMBOL_GPL(ttm_dma_unpopulate); * * XXX: (dchinner) Deadlock warning! * - * We need to pass sc->gfp_mask to ttm_dma_page_pool_free(). - * * I'm getting sadder as I hear more pathetical whimpers about needing per-pool * shrinkers */ @@ -1030,8 +1035,8 @@ ttm_dma_pool_shrink_scan(struct shrinker *shrink, struct shrink_control *sc) if (++idx < pool_offset) continue; nr_free = shrink_pages; - shrink_pages = ttm_dma_page_pool_free(p->pool, nr_free, - sc->gfp_mask); + /* OK to use static buffer since global mutex is held. */ + shrink_pages = ttm_dma_page_pool_free(p->pool, nr_free, true); freed += nr_free - shrink_pages; pr_debug("%s: (%s:%d) Asked to shrink %d, have %d more to go\n", -- cgit v0.10.2 From 7f907bf284ba7bb8d271f094b226699d3fef2142 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Sat, 8 Nov 2014 10:16:19 -0500 Subject: drm/irq: BUG_ON() -> WARN_ON() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Let's make things a bit easier to debug when things go bad (potentially under console_lock). Signed-off-by: Rob Clark Reviewed-by: Michel Dänzer Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c index 3e6b582..0e47df4 100644 --- a/drivers/gpu/drm/drm_irq.c +++ b/drivers/gpu/drm/drm_irq.c @@ -1029,7 +1029,8 @@ void drm_vblank_put(struct drm_device *dev, int crtc) { struct drm_vblank_crtc *vblank = &dev->vblank[crtc]; - BUG_ON(atomic_read(&vblank->refcount) == 0); + if (WARN_ON(atomic_read(&vblank->refcount) == 0)) + return; if (WARN_ON(crtc >= dev->num_crtcs)) return; -- cgit v0.10.2 From c7eb76f4931b505e57c6510133df19f4ec1ce3e5 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 18:38:06 +0100 Subject: drm/atomic: Ensure that drm_connector_index is stable I've totally forgotten that with DP MST connectors can now be hotplugged. And failed to adapt Rob's drm_atomic_state code (which predates connector hotplugging) to the new realities. The first step is to make sure that the connector indices used to access the arrays of pointers are stable. The connection mutex gives us enough guarantees for that, which means we won't unecessarily block on concurrent modesets or background probing. So add a locking WARN_ON and shuffle the code slightly to make sure we always hold the right lock. Signed-off-by: Daniel Vetter Reviewed-by: Rob Clark Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index ed22a71..90b2d16 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -295,15 +295,15 @@ drm_atomic_get_connector_state(struct drm_atomic_state *state, struct drm_mode_config *config = &connector->dev->mode_config; struct drm_connector_state *connector_state; + ret = drm_modeset_lock(&config->connection_mutex, state->acquire_ctx); + if (ret) + return ERR_PTR(ret); + index = drm_connector_index(connector); if (state->connector_states[index]) return state->connector_states[index]; - ret = drm_modeset_lock(&config->connection_mutex, state->acquire_ctx); - if (ret) - return ERR_PTR(ret); - connector_state = connector->funcs->atomic_duplicate_state(connector); if (!connector_state) return ERR_PTR(-ENOMEM); diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 56737e7..5c878f1 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -867,6 +867,8 @@ int drm_connector_init(struct drm_device *dev, drm_connector_get_cmdline_mode(connector); + /* We should add connectors at the end to avoid upsetting the connector + * index too much. */ list_add_tail(&connector->head, &dev->mode_config.connector_list); dev->mode_config.num_connector++; @@ -930,6 +932,9 @@ unsigned int drm_connector_index(struct drm_connector *connector) { unsigned int index = 0; struct drm_connector *tmp; + struct drm_mode_config *config = &connector->dev->mode_config; + + WARN_ON(!drm_modeset_is_locked(&config->connection_mutex)); list_for_each_entry(tmp, &connector->dev->mode_config.connector_list, head) { if (tmp == connector) -- cgit v0.10.2 From 6f75cea66c8dd043ced282016b21a639af176642 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 18:38:07 +0100 Subject: drm/atomic: Only destroy connector states with connection mutex held Otherwise the connector might have been unplugged and destroyed while we didn't look. Yet another fallout from DP MST hotplugging that I didn't consider. To make sure we get this right add an appropriate WARN_ON to drm_atomic_state_clear (obviously only when we actually have a state to clear up). And reorder all the state_clear and backoff calls to make it work out properly. Signed-off-by: Daniel Vetter Reviewed-by: Rob Clark Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index 90b2d16..67c1dc8 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -112,21 +112,24 @@ EXPORT_SYMBOL(drm_atomic_state_alloc); void drm_atomic_state_clear(struct drm_atomic_state *state) { struct drm_device *dev = state->dev; + struct drm_mode_config *config = &dev->mode_config; int i; DRM_DEBUG_KMS("Clearing atomic state %p\n", state); - for (i = 0; i < dev->mode_config.num_connector; i++) { + for (i = 0; i < config->num_connector; i++) { struct drm_connector *connector = state->connectors[i]; if (!connector) continue; + WARN_ON(!drm_modeset_is_locked(&config->connection_mutex)); + connector->funcs->atomic_destroy_state(connector, state->connector_states[i]); } - for (i = 0; i < dev->mode_config.num_crtc; i++) { + for (i = 0; i < config->num_crtc; i++) { struct drm_crtc *crtc = state->crtcs[i]; if (!crtc) @@ -136,7 +139,7 @@ void drm_atomic_state_clear(struct drm_atomic_state *state) state->crtc_states[i]); } - for (i = 0; i < dev->mode_config.num_total_plane; i++) { + for (i = 0; i < config->num_total_plane; i++) { struct drm_plane *plane = state->planes[i]; if (!plane) diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index fad2b93..0cd0546 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -1217,8 +1217,8 @@ fail: return ret; backoff: - drm_atomic_legacy_backoff(state); drm_atomic_state_clear(state); + drm_atomic_legacy_backoff(state); /* * Someone might have exchanged the framebuffer while we dropped locks @@ -1285,8 +1285,8 @@ fail: return ret; backoff: - drm_atomic_legacy_backoff(state); drm_atomic_state_clear(state); + drm_atomic_legacy_backoff(state); /* * Someone might have exchanged the framebuffer while we dropped locks @@ -1462,8 +1462,8 @@ fail: return ret; backoff: - drm_atomic_legacy_backoff(state); drm_atomic_state_clear(state); + drm_atomic_legacy_backoff(state); /* * Someone might have exchanged the framebuffer while we dropped locks @@ -1528,8 +1528,8 @@ fail: return ret; backoff: - drm_atomic_legacy_backoff(state); drm_atomic_state_clear(state); + drm_atomic_legacy_backoff(state); goto retry; } @@ -1587,8 +1587,8 @@ fail: return ret; backoff: - drm_atomic_legacy_backoff(state); drm_atomic_state_clear(state); + drm_atomic_legacy_backoff(state); goto retry; } @@ -1646,8 +1646,8 @@ fail: return ret; backoff: - drm_atomic_legacy_backoff(state); drm_atomic_state_clear(state); + drm_atomic_legacy_backoff(state); goto retry; } @@ -1725,8 +1725,8 @@ fail: return ret; backoff: - drm_atomic_legacy_backoff(state); drm_atomic_state_clear(state); + drm_atomic_legacy_backoff(state); /* * Someone might have exchanged the framebuffer while we dropped locks -- cgit v0.10.2 From f52b69f1ecfdd7ef6867a257620258c09e569552 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 18:38:08 +0100 Subject: drm/atomic: Don't overrun the connector array when hotplugging Yet another fallout from not considering DP MST hotplug. With the previous patches we have stable indices, but it might still happen that a connector gets added between when we allocate the array and when we actually add a connector. Especially when we back off due to ww mutex contention or similar issues. So store the sizes of the arrays in struct drm_atomic_state and double check them. We don't really care about races except that we want to use a consistent value, so ACCESS_ONCE is all we need. And if we indeed notice that we'd overrun the array then just give up and restart the entire ioctl. Signed-off-by: Daniel Vetter Reviewed-by: Rob Clark Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index 67c1dc8..3624632 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -56,6 +56,8 @@ drm_atomic_state_alloc(struct drm_device *dev) if (!state) return NULL; + state->num_connector = ACCESS_ONCE(dev->mode_config.num_connector); + state->crtcs = kcalloc(dev->mode_config.num_crtc, sizeof(*state->crtcs), GFP_KERNEL); if (!state->crtcs) @@ -72,12 +74,12 @@ drm_atomic_state_alloc(struct drm_device *dev) sizeof(*state->plane_states), GFP_KERNEL); if (!state->plane_states) goto fail; - state->connectors = kcalloc(dev->mode_config.num_connector, + state->connectors = kcalloc(state->num_connector, sizeof(*state->connectors), GFP_KERNEL); if (!state->connectors) goto fail; - state->connector_states = kcalloc(dev->mode_config.num_connector, + state->connector_states = kcalloc(state->num_connector, sizeof(*state->connector_states), GFP_KERNEL); if (!state->connector_states) @@ -117,7 +119,7 @@ void drm_atomic_state_clear(struct drm_atomic_state *state) DRM_DEBUG_KMS("Clearing atomic state %p\n", state); - for (i = 0; i < config->num_connector; i++) { + for (i = 0; i < state->num_connector; i++) { struct drm_connector *connector = state->connectors[i]; if (!connector) @@ -304,6 +306,21 @@ drm_atomic_get_connector_state(struct drm_atomic_state *state, index = drm_connector_index(connector); + /* + * Construction of atomic state updates can race with a connector + * hot-add which might overflow. In this case flip the table and just + * restart the entire ioctl - no one is fast enough to livelock a cpu + * with physical hotplug events anyway. + * + * Note that we only grab the indexes once we have the right lock to + * prevent hotplug/unplugging of connectors. So removal is no problem, + * at most the array is a bit too large. + */ + if (index >= state->num_connector) { + DRM_DEBUG_KMS("Hot-added connector would overflow state array, restarting\n"); + return -EAGAIN; + } + if (state->connector_states[index]) return state->connector_states[index]; @@ -499,10 +516,9 @@ int drm_atomic_connectors_for_crtc(struct drm_atomic_state *state, struct drm_crtc *crtc) { - int nconnectors = state->dev->mode_config.num_connector; int i, num_connected_connectors = 0; - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < state->num_connector; i++) { struct drm_connector_state *conn_state; conn_state = state->connector_states[i]; diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 0cd0546..99095ef 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -249,7 +249,6 @@ static int mode_fixup(struct drm_atomic_state *state) { int ncrtcs = state->dev->mode_config.num_crtc; - int nconnectors = state->dev->mode_config.num_connector; struct drm_crtc_state *crtc_state; struct drm_connector_state *conn_state; int i; @@ -264,7 +263,7 @@ mode_fixup(struct drm_atomic_state *state) drm_mode_copy(&crtc_state->adjusted_mode, &crtc_state->mode); } - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < state->num_connector; i++) { struct drm_encoder_helper_funcs *funcs; struct drm_encoder *encoder; @@ -336,7 +335,6 @@ drm_atomic_helper_check_prepare(struct drm_device *dev, struct drm_atomic_state *state) { int ncrtcs = dev->mode_config.num_crtc; - int nconnectors = dev->mode_config.num_connector; struct drm_crtc *crtc; struct drm_crtc_state *crtc_state; int i, ret; @@ -361,7 +359,7 @@ drm_atomic_helper_check_prepare(struct drm_device *dev, } } - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < state->num_connector; i++) { /* * This only sets crtc->mode_changed for routing changes, * drivers must set crtc->mode_changed themselves when connector @@ -485,10 +483,9 @@ static void disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state) { int ncrtcs = old_state->dev->mode_config.num_crtc; - int nconnectors = old_state->dev->mode_config.num_connector; int i; - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < old_state->num_connector; i++) { struct drm_connector_state *old_conn_state; struct drm_connector *connector; struct drm_encoder_helper_funcs *funcs; @@ -553,12 +550,11 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state) static void set_routing_links(struct drm_device *dev, struct drm_atomic_state *old_state) { - int nconnectors = dev->mode_config.num_connector; int ncrtcs = old_state->dev->mode_config.num_crtc; int i; /* clear out existing links */ - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < old_state->num_connector; i++) { struct drm_connector *connector; connector = old_state->connectors[i]; @@ -573,7 +569,7 @@ set_routing_links(struct drm_device *dev, struct drm_atomic_state *old_state) } /* set new links */ - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < old_state->num_connector; i++) { struct drm_connector *connector; connector = old_state->connectors[i]; @@ -608,7 +604,6 @@ static void crtc_set_mode(struct drm_device *dev, struct drm_atomic_state *old_state) { int ncrtcs = old_state->dev->mode_config.num_crtc; - int nconnectors = old_state->dev->mode_config.num_connector; int i; for (i = 0; i < ncrtcs; i++) { @@ -626,7 +621,7 @@ crtc_set_mode(struct drm_device *dev, struct drm_atomic_state *old_state) funcs->mode_set_nofb(crtc); } - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < old_state->num_connector; i++) { struct drm_connector *connector; struct drm_crtc_state *new_crtc_state; struct drm_encoder_helper_funcs *funcs; @@ -687,7 +682,6 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev, struct drm_atomic_state *old_state) { int ncrtcs = old_state->dev->mode_config.num_crtc; - int nconnectors = old_state->dev->mode_config.num_connector; int i; for (i = 0; i < ncrtcs; i++) { @@ -706,7 +700,7 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev, funcs->commit(crtc); } - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < old_state->num_connector; i++) { struct drm_connector *connector; struct drm_encoder_helper_funcs *funcs; struct drm_encoder *encoder; @@ -1304,7 +1298,6 @@ static int update_output_state(struct drm_atomic_state *state, { struct drm_device *dev = set->crtc->dev; struct drm_connector_state *conn_state; - int nconnectors = state->dev->mode_config.num_connector; int ncrtcs = state->dev->mode_config.num_crtc; int ret, i, j; @@ -1333,7 +1326,7 @@ static int update_output_state(struct drm_atomic_state *state, } /* Then recompute connector->crtc links and crtc enabling state. */ - for (i = 0; i < nconnectors; i++) { + for (i = 0; i < state->num_connector; i++) { struct drm_connector *connector; connector = state->connectors[i]; diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 7b28ab0..f3142c6 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -825,6 +825,7 @@ struct drm_bridge { * @plane_states: pointer to array of plane states pointers * @crtcs: pointer to array of CRTC pointers * @crtc_states: pointer to array of CRTC states pointers + * @num_connector: size of the @connectors and @connector_states arrays * @connectors: pointer to array of connector pointers * @connector_states: pointer to array of connector states pointers * @acquire_ctx: acquire context for this atomic modeset state update @@ -836,6 +837,7 @@ struct drm_atomic_state { struct drm_plane_state **plane_states; struct drm_crtc **crtcs; struct drm_crtc_state **crtc_states; + int num_connector; struct drm_connector **connectors; struct drm_connector_state **connector_states; -- cgit v0.10.2 From 1a49863349929f5a85fd7c7ffb06213d4fdc8c15 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 18:38:09 +0100 Subject: drm/crtc: Polish kerneldoc - Make it clear that it's a negative errno (more in line with everything else). - Clean up the confusion around get_properties vs. getproperty ioctls: One reads per-obj property values, the other reads property metadata. Signed-off-by: Daniel Vetter Reviewed-by: Rob Clark Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 5c878f1..8c55030 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -1490,7 +1490,7 @@ EXPORT_SYMBOL(drm_mode_create_scaling_mode_property); * connectors. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_create_aspect_ratio_property(struct drm_device *dev) { @@ -1674,7 +1674,7 @@ static void drm_crtc_convert_to_umode(struct drm_mode_modeinfo *out, * the caller. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ static int drm_crtc_convert_umode(struct drm_display_mode *out, const struct drm_mode_modeinfo *in) @@ -1717,7 +1717,7 @@ static int drm_crtc_convert_umode(struct drm_display_mode *out, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getresources(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -1905,7 +1905,7 @@ out: * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getcrtc(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -1966,7 +1966,7 @@ static bool drm_mode_expose_to_userspace(const struct drm_display_mode *mode, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getconnector(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -2110,7 +2110,7 @@ out: * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getencoder(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -2151,7 +2151,7 @@ int drm_mode_getencoder(struct drm_device *dev, void *data, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getplane_res(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -2212,7 +2212,7 @@ int drm_mode_getplane_res(struct drm_device *dev, void *data, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getplane(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -2386,7 +2386,7 @@ static int setplane_internal(struct drm_plane *plane, * valid crtc). * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_setplane(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -2461,7 +2461,7 @@ int drm_mode_setplane(struct drm_device *dev, void *data, * interface. The only thing it adds is correct refcounting dance. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_set_config_internal(struct drm_mode_set *set) { @@ -2554,7 +2554,7 @@ EXPORT_SYMBOL(drm_crtc_check_viewport); * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_setcrtc(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -2717,7 +2717,7 @@ out: * userspace wants to make use of these capabilities. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ static int drm_mode_cursor_universal(struct drm_crtc *crtc, struct drm_mode_cursor2 *req, @@ -2865,7 +2865,7 @@ out: * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_cursor_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -2892,7 +2892,7 @@ int drm_mode_cursor_ioctl(struct drm_device *dev, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_cursor2_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -2956,7 +2956,7 @@ EXPORT_SYMBOL(drm_mode_legacy_fb_format); * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_addfb(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -3161,7 +3161,7 @@ static struct drm_framebuffer *add_framebuffer_internal(struct drm_device *dev, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_addfb2(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -3189,7 +3189,7 @@ int drm_mode_addfb2(struct drm_device *dev, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_rmfb(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -3243,7 +3243,7 @@ fail_lookup: * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getfb(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -3304,7 +3304,7 @@ int drm_mode_getfb(struct drm_device *dev, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_dirtyfb_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -3384,7 +3384,7 @@ out_err1: * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ void drm_fb_release(struct drm_file *priv) { @@ -3818,17 +3818,20 @@ int drm_object_property_get_value(struct drm_mode_object *obj, EXPORT_SYMBOL(drm_object_property_get_value); /** - * drm_mode_getproperty_ioctl - get the current value of a connector's property + * drm_mode_getproperty_ioctl - get the property metadata * @dev: DRM device * @data: ioctl data * @file_priv: DRM file info * - * This function retrieves the current value for an connectors's property. + * This function retrieves the metadata for a given property, like the different + * possible values for an enum property or the limits for a range property. + * + * Blob properties are special * * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getproperty_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -3981,7 +3984,7 @@ static void drm_property_destroy_blob(struct drm_device *dev, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_getblob_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4026,7 +4029,7 @@ done: * them more meaningful names. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_connector_set_path_property(struct drm_connector *connector, const char *path) @@ -4056,7 +4059,7 @@ EXPORT_SYMBOL(drm_mode_connector_set_path_property); * connector's edid property. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_connector_update_edid_property(struct drm_connector *connector, const struct edid *edid) @@ -4153,7 +4156,7 @@ static bool drm_property_change_is_valid(struct drm_property *property, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_connector_property_set_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4236,7 +4239,7 @@ int drm_mode_plane_set_obj_prop(struct drm_plane *plane, EXPORT_SYMBOL(drm_mode_plane_set_obj_prop); /** - * drm_mode_getproperty_ioctl - get the current value of a object's property + * drm_mode_obj_get_properties_ioctl - get the current value of a object's property * @dev: DRM device * @data: ioctl data * @file_priv: DRM file info @@ -4248,7 +4251,7 @@ EXPORT_SYMBOL(drm_mode_plane_set_obj_prop); * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_obj_get_properties_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4320,7 +4323,7 @@ out: * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_obj_set_property_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4392,7 +4395,7 @@ out: * possible_clones and possible_crtcs bitmasks. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_connector_attach_encoder(struct drm_connector *connector, struct drm_encoder *encoder) @@ -4419,7 +4422,7 @@ EXPORT_SYMBOL(drm_mode_connector_attach_encoder); * fixed gamma table size. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_crtc_set_gamma_size(struct drm_crtc *crtc, int gamma_size) @@ -4448,7 +4451,7 @@ EXPORT_SYMBOL(drm_mode_crtc_set_gamma_size); * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_gamma_set_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4520,7 +4523,7 @@ out: * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_gamma_get_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4586,7 +4589,7 @@ out: * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_page_flip_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4752,7 +4755,7 @@ EXPORT_SYMBOL(drm_mode_config_reset); * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_create_dumb_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4804,7 +4807,7 @@ int drm_mode_create_dumb_ioctl(struct drm_device *dev, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_mmap_dumb_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -4831,7 +4834,7 @@ int drm_mode_mmap_dumb_ioctl(struct drm_device *dev, * Called by the user via ioctl. * * Returns: - * Zero on success, errno on failure. + * Zero on success, negative errno on failure. */ int drm_mode_destroy_dumb_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) -- cgit v0.10.2 From 3758b34193638f664177565f1692faa1bec7d9ed Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 18:38:10 +0100 Subject: drm: s/enum_blob_list/enum_list/ in drm_property I guess for hysterical raisins this was meant to be the way to read blob properties. But that's done with the two-stage approach which uses separate blob kms object and the special-purpose get_blob ioctl. Shipping userspace seems to have never relied on this, and the kernel also never put any blob thing onto that property. And nowadays it would blow up, e.g. in drm_property_destroy. Also it makes no sense to return values in an ioctl that only returns metadata about everything. So let's ditch all the internal code for the blob list, rename the list to be unambiguous and sprinkle comments all over the place to explain this peculiar piece of api. v2: Squash in fixup from Rob to remove now unused variables. Cc: Rob Clark Signed-off-by: Daniel Vetter Reviewed-by: Rob Clark Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 8c55030..589a921 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -3457,7 +3457,7 @@ struct drm_property *drm_property_create(struct drm_device *dev, int flags, property->flags = flags; property->num_values = num_values; - INIT_LIST_HEAD(&property->enum_blob_list); + INIT_LIST_HEAD(&property->enum_list); if (name) { strncpy(property->name, name, DRM_PROP_NAME_LEN); @@ -3679,8 +3679,8 @@ int drm_property_add_enum(struct drm_property *property, int index, (value > 63)) return -EINVAL; - if (!list_empty(&property->enum_blob_list)) { - list_for_each_entry(prop_enum, &property->enum_blob_list, head) { + if (!list_empty(&property->enum_list)) { + list_for_each_entry(prop_enum, &property->enum_list, head) { if (prop_enum->value == value) { strncpy(prop_enum->name, name, DRM_PROP_NAME_LEN); prop_enum->name[DRM_PROP_NAME_LEN-1] = '\0'; @@ -3698,7 +3698,7 @@ int drm_property_add_enum(struct drm_property *property, int index, prop_enum->value = value; property->values[index] = value; - list_add_tail(&prop_enum->head, &property->enum_blob_list); + list_add_tail(&prop_enum->head, &property->enum_list); return 0; } EXPORT_SYMBOL(drm_property_add_enum); @@ -3715,7 +3715,7 @@ void drm_property_destroy(struct drm_device *dev, struct drm_property *property) { struct drm_property_enum *prop_enum, *pt; - list_for_each_entry_safe(prop_enum, pt, &property->enum_blob_list, head) { + list_for_each_entry_safe(prop_enum, pt, &property->enum_list, head) { list_del(&prop_enum->head); kfree(prop_enum); } @@ -3839,16 +3839,12 @@ int drm_mode_getproperty_ioctl(struct drm_device *dev, struct drm_mode_get_property *out_resp = data; struct drm_property *property; int enum_count = 0; - int blob_count = 0; int value_count = 0; int ret = 0, i; int copied; struct drm_property_enum *prop_enum; struct drm_mode_property_enum __user *enum_ptr; - struct drm_property_blob *prop_blob; - uint32_t __user *blob_id_ptr; uint64_t __user *values_ptr; - uint32_t __user *blob_length_ptr; if (!drm_core_check_feature(dev, DRIVER_MODESET)) return -EINVAL; @@ -3862,11 +3858,8 @@ int drm_mode_getproperty_ioctl(struct drm_device *dev, if (drm_property_type_is(property, DRM_MODE_PROP_ENUM) || drm_property_type_is(property, DRM_MODE_PROP_BITMASK)) { - list_for_each_entry(prop_enum, &property->enum_blob_list, head) + list_for_each_entry(prop_enum, &property->enum_list, head) enum_count++; - } else if (drm_property_type_is(property, DRM_MODE_PROP_BLOB)) { - list_for_each_entry(prop_blob, &property->enum_blob_list, head) - blob_count++; } value_count = property->num_values; @@ -3891,7 +3884,7 @@ int drm_mode_getproperty_ioctl(struct drm_device *dev, if ((out_resp->count_enum_blobs >= enum_count) && enum_count) { copied = 0; enum_ptr = (struct drm_mode_property_enum __user *)(unsigned long)out_resp->enum_blob_ptr; - list_for_each_entry(prop_enum, &property->enum_blob_list, head) { + list_for_each_entry(prop_enum, &property->enum_list, head) { if (copy_to_user(&enum_ptr[copied].value, &prop_enum->value, sizeof(uint64_t))) { ret = -EFAULT; @@ -3909,28 +3902,16 @@ int drm_mode_getproperty_ioctl(struct drm_device *dev, out_resp->count_enum_blobs = enum_count; } - if (drm_property_type_is(property, DRM_MODE_PROP_BLOB)) { - if ((out_resp->count_enum_blobs >= blob_count) && blob_count) { - copied = 0; - blob_id_ptr = (uint32_t __user *)(unsigned long)out_resp->enum_blob_ptr; - blob_length_ptr = (uint32_t __user *)(unsigned long)out_resp->values_ptr; - - list_for_each_entry(prop_blob, &property->enum_blob_list, head) { - if (put_user(prop_blob->base.id, blob_id_ptr + copied)) { - ret = -EFAULT; - goto done; - } - - if (put_user(prop_blob->length, blob_length_ptr + copied)) { - ret = -EFAULT; - goto done; - } - - copied++; - } - } - out_resp->count_enum_blobs = blob_count; - } + /* + * NOTE: The idea seems to have been to use this to read all the blob + * property values. But nothing ever added them to the corresponding + * list, userspace always used the special-purpose get_blob ioctl to + * read the value for a blob property. It also doesn't make a lot of + * sense to return values here when everything else is just metadata for + * the property itself. + */ + if (drm_property_type_is(property, DRM_MODE_PROP_BLOB)) + out_resp->count_enum_blobs = 0; done: drm_modeset_unlock_all(dev); return ret; diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index f3142c6..b459e8f 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -216,7 +216,7 @@ struct drm_property { uint64_t *values; struct drm_device *dev; - struct list_head enum_blob_list; + struct list_head enum_list; }; struct drm_crtc; diff --git a/include/uapi/drm/drm_mode.h b/include/uapi/drm/drm_mode.h index a0db2d4a..86574b0 100644 --- a/include/uapi/drm/drm_mode.h +++ b/include/uapi/drm/drm_mode.h @@ -286,6 +286,8 @@ struct drm_mode_get_property { char name[DRM_PROP_NAME_LEN]; __u32 count_values; + /* This is only used to count enum values, not blobs. The _blobs is + * simply because of a historical reason, i.e. backwards compat. */ __u32 count_enum_blobs; }; -- cgit v0.10.2 From b0fcfc899513a4d4729914b6872ab0564c0f523a Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 18:38:11 +0100 Subject: drm/atomic_helper: Make it clear that commit_planes gets the old state Oversight from my kerneldoc cleanup when doing the original atomic helper series - I've only applied this clarification to the modeset related helpers, and not the plane update code. Remedy this asap. Signed-off-by: Daniel Vetter Reviewed-by: Rob Clark Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 99095ef..6903600 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -976,18 +976,18 @@ EXPORT_SYMBOL(drm_atomic_helper_prepare_planes); /** * drm_atomic_helper_commit_planes - commit plane state * @dev: DRM device - * @state: atomic state + * @old_state: atomic state object with old state structures * * This function commits the new plane state using the plane and atomic helper * functions for planes and crtcs. It assumes that the atomic state has already * been pushed into the relevant object state pointers, since this step can no * longer fail. * - * It still requires the global state object @state to know which planes and + * It still requires the global state object @old_state to know which planes and * crtcs need to be updated though. */ void drm_atomic_helper_commit_planes(struct drm_device *dev, - struct drm_atomic_state *state) + struct drm_atomic_state *old_state) { int nplanes = dev->mode_config.num_total_plane; int ncrtcs = dev->mode_config.num_crtc; @@ -995,7 +995,7 @@ void drm_atomic_helper_commit_planes(struct drm_device *dev, for (i = 0; i < ncrtcs; i++) { struct drm_crtc_helper_funcs *funcs; - struct drm_crtc *crtc = state->crtcs[i]; + struct drm_crtc *crtc = old_state->crtcs[i]; if (!crtc) continue; @@ -1010,7 +1010,7 @@ void drm_atomic_helper_commit_planes(struct drm_device *dev, for (i = 0; i < nplanes; i++) { struct drm_plane_helper_funcs *funcs; - struct drm_plane *plane = state->planes[i]; + struct drm_plane *plane = old_state->planes[i]; if (!plane) continue; @@ -1025,7 +1025,7 @@ void drm_atomic_helper_commit_planes(struct drm_device *dev, for (i = 0; i < ncrtcs; i++) { struct drm_crtc_helper_funcs *funcs; - struct drm_crtc *crtc = state->crtcs[i]; + struct drm_crtc *crtc = old_state->crtcs[i]; if (!crtc) continue; -- cgit v0.10.2 From a7ca52e108e549669940b11779bb491a931f8c65 Mon Sep 17 00:00:00 2001 From: Zach Reizner Date: Mon, 17 Nov 2014 17:19:41 -0800 Subject: drm/cirrus: fix leaky driver load error handling MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Before this patch, cirrus_device_init could have failed while cirrus_mm_init succeeded and the driver would have reported overall success on load. This patch causes cirrus_device_init to return on the first error encountered. Reviewed-by: Stéphane Marchesin Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/cirrus/cirrus_main.c b/drivers/gpu/drm/cirrus/cirrus_main.c index 99c1983..ab7cb54 100644 --- a/drivers/gpu/drm/cirrus/cirrus_main.c +++ b/drivers/gpu/drm/cirrus/cirrus_main.c @@ -179,17 +179,22 @@ int cirrus_driver_load(struct drm_device *dev, unsigned long flags) } r = cirrus_mm_init(cdev); - if (r) + if (r) { dev_err(&dev->pdev->dev, "fatal err on mm init\n"); + goto out; + } r = cirrus_modeset_init(cdev); - if (r) + if (r) { dev_err(&dev->pdev->dev, "Fatal error during modeset init: %d\n", r); + goto out; + } dev->mode_config.funcs = (void *)&cirrus_mode_funcs; + + return 0; out: - if (r) - cirrus_driver_unload(dev); + cirrus_driver_unload(dev); return r; } -- cgit v0.10.2 From 09a58da052111263e31f0050ad775b62166fafba Mon Sep 17 00:00:00 2001 From: Haixia Shi Date: Wed, 12 Nov 2014 18:33:52 -0800 Subject: drm/udl: add cache flags definitions for udl_gem_object MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit By default set udl_gem_object as cacheable, but set WC flag when attaching dmabuf. In udl_gem_mmap() update cache attributes based on the flags, similar to exynos_drm_gem_mmap(). Signed-off-by: Haixia Shi Reviewed-by: Sonny Rao Reviewed-by: Olof Johansson Reviewed-by: Stéphane Marchesin Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/udl/udl_drv.h b/drivers/gpu/drm/udl/udl_drv.h index c7490a2..3082780 100644 --- a/drivers/gpu/drm/udl/udl_drv.h +++ b/drivers/gpu/drm/udl/udl_drv.h @@ -25,6 +25,9 @@ #define DRIVER_MINOR 0 #define DRIVER_PATCHLEVEL 1 +#define UDL_BO_CACHEABLE (1 << 0) +#define UDL_BO_WC (1 << 1) + struct udl_device; struct urb_node { @@ -69,6 +72,7 @@ struct udl_gem_object { struct page **pages; void *vmapping; struct sg_table *sg; + unsigned int flags; }; #define to_udl_bo(x) container_of(x, struct udl_gem_object, base) diff --git a/drivers/gpu/drm/udl/udl_gem.c b/drivers/gpu/drm/udl/udl_gem.c index 8044f5f..e00459d 100644 --- a/drivers/gpu/drm/udl/udl_gem.c +++ b/drivers/gpu/drm/udl/udl_gem.c @@ -25,6 +25,7 @@ struct udl_gem_object *udl_gem_alloc_object(struct drm_device *dev, return NULL; } + obj->flags = UDL_BO_CACHEABLE; return obj; } @@ -56,6 +57,23 @@ udl_gem_create(struct drm_file *file, return 0; } +static void update_vm_cache_attr(struct udl_gem_object *obj, + struct vm_area_struct *vma) +{ + DRM_DEBUG_KMS("flags = 0x%x\n", obj->flags); + + /* non-cacheable as default. */ + if (obj->flags & UDL_BO_CACHEABLE) { + vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); + } else if (obj->flags & UDL_BO_WC) { + vma->vm_page_prot = + pgprot_writecombine(vm_get_page_prot(vma->vm_flags)); + } else { + vma->vm_page_prot = + pgprot_noncached(vm_get_page_prot(vma->vm_flags)); + } +} + int udl_dumb_create(struct drm_file *file, struct drm_device *dev, struct drm_mode_create_dumb *args) @@ -77,6 +95,8 @@ int udl_drm_gem_mmap(struct file *filp, struct vm_area_struct *vma) vma->vm_flags &= ~VM_PFNMAP; vma->vm_flags |= VM_MIXEDMAP; + update_vm_cache_attr(to_udl_bo(vma->vm_private_data), vma); + return ret; } @@ -279,6 +299,7 @@ struct drm_gem_object *udl_gem_prime_import(struct drm_device *dev, } uobj->base.import_attach = attach; + uobj->flags = UDL_BO_WC; return &uobj->base; -- cgit v0.10.2 From ebfdd6d5e837bd1c170b9bdd749fee2a4183a7f5 Mon Sep 17 00:00:00 2001 From: Haixia Shi Date: Wed, 12 Nov 2014 18:33:53 -0800 Subject: drm/udl: add support to export a handle to a FD on UDL. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Only importing an FD to a handle is currently supported on UDL, but the exporting functionality is equally useful. Signed-off-by: Haixia Shi Reviewed-by: Stéphane Marchesin Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/udl/Makefile b/drivers/gpu/drm/udl/Makefile index 05c7481..195bcac 100644 --- a/drivers/gpu/drm/udl/Makefile +++ b/drivers/gpu/drm/udl/Makefile @@ -1,6 +1,6 @@ ccflags-y := -Iinclude/drm -udl-y := udl_drv.o udl_modeset.o udl_connector.o udl_encoder.o udl_main.o udl_fb.o udl_transfer.o udl_gem.o +udl-y := udl_drv.o udl_modeset.o udl_connector.o udl_encoder.o udl_main.o udl_fb.o udl_transfer.o udl_gem.o udl_dmabuf.o obj-$(CONFIG_DRM_UDL) := udl.o diff --git a/drivers/gpu/drm/udl/udl_dmabuf.c b/drivers/gpu/drm/udl/udl_dmabuf.c new file mode 100644 index 0000000..1d85c3a --- /dev/null +++ b/drivers/gpu/drm/udl/udl_dmabuf.c @@ -0,0 +1,273 @@ +/* + * udl_dmabuf.c + * + * Copyright (c) 2014 The Chromium OS Authors + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License as published by the + * Free Software Foundation; either version 2 of the License, or (at your + * option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program. If not, see . + */ + +#include +#include "udl_drv.h" +#include +#include + +struct udl_drm_dmabuf_attachment { + struct sg_table sgt; + enum dma_data_direction dir; + bool is_mapped; +}; + +static int udl_attach_dma_buf(struct dma_buf *dmabuf, + struct device *dev, + struct dma_buf_attachment *attach) +{ + struct udl_drm_dmabuf_attachment *udl_attach; + + DRM_DEBUG_PRIME("[DEV:%s] size:%zd\n", dev_name(attach->dev), + attach->dmabuf->size); + + udl_attach = kzalloc(sizeof(*udl_attach), GFP_KERNEL); + if (!udl_attach) + return -ENOMEM; + + udl_attach->dir = DMA_NONE; + attach->priv = udl_attach; + + return 0; +} + +static void udl_detach_dma_buf(struct dma_buf *dmabuf, + struct dma_buf_attachment *attach) +{ + struct udl_drm_dmabuf_attachment *udl_attach = attach->priv; + struct sg_table *sgt; + + if (!udl_attach) + return; + + DRM_DEBUG_PRIME("[DEV:%s] size:%zd\n", dev_name(attach->dev), + attach->dmabuf->size); + + sgt = &udl_attach->sgt; + + if (udl_attach->dir != DMA_NONE) + dma_unmap_sg(attach->dev, sgt->sgl, sgt->nents, + udl_attach->dir); + + sg_free_table(sgt); + kfree(udl_attach); + attach->priv = NULL; +} + +static struct sg_table *udl_map_dma_buf(struct dma_buf_attachment *attach, + enum dma_data_direction dir) +{ + struct udl_drm_dmabuf_attachment *udl_attach = attach->priv; + struct udl_gem_object *obj = to_udl_bo(attach->dmabuf->priv); + struct drm_device *dev = obj->base.dev; + struct scatterlist *rd, *wr; + struct sg_table *sgt = NULL; + unsigned int i; + int page_count; + int nents, ret; + + DRM_DEBUG_PRIME("[DEV:%s] size:%zd dir=%d\n", dev_name(attach->dev), + attach->dmabuf->size, dir); + + /* just return current sgt if already requested. */ + if (udl_attach->dir == dir && udl_attach->is_mapped) + return &udl_attach->sgt; + + if (!obj->pages) { + DRM_ERROR("pages is null.\n"); + return ERR_PTR(-ENOMEM); + } + + page_count = obj->base.size / PAGE_SIZE; + obj->sg = drm_prime_pages_to_sg(obj->pages, page_count); + if (!obj->sg) { + DRM_ERROR("sg is null.\n"); + return ERR_PTR(-ENOMEM); + } + + sgt = &udl_attach->sgt; + + ret = sg_alloc_table(sgt, obj->sg->orig_nents, GFP_KERNEL); + if (ret) { + DRM_ERROR("failed to alloc sgt.\n"); + return ERR_PTR(-ENOMEM); + } + + mutex_lock(&dev->struct_mutex); + + rd = obj->sg->sgl; + wr = sgt->sgl; + for (i = 0; i < sgt->orig_nents; ++i) { + sg_set_page(wr, sg_page(rd), rd->length, rd->offset); + rd = sg_next(rd); + wr = sg_next(wr); + } + + if (dir != DMA_NONE) { + nents = dma_map_sg(attach->dev, sgt->sgl, sgt->orig_nents, dir); + if (!nents) { + DRM_ERROR("failed to map sgl with iommu.\n"); + sg_free_table(sgt); + sgt = ERR_PTR(-EIO); + goto err_unlock; + } + } + + udl_attach->is_mapped = true; + udl_attach->dir = dir; + attach->priv = udl_attach; + +err_unlock: + mutex_unlock(&dev->struct_mutex); + return sgt; +} + +static void udl_unmap_dma_buf(struct dma_buf_attachment *attach, + struct sg_table *sgt, + enum dma_data_direction dir) +{ + /* Nothing to do. */ + DRM_DEBUG_PRIME("[DEV:%s] size:%zd dir:%d\n", dev_name(attach->dev), + attach->dmabuf->size, dir); +} + +static void *udl_dmabuf_kmap(struct dma_buf *dma_buf, unsigned long page_num) +{ + /* TODO */ + + return NULL; +} + +static void *udl_dmabuf_kmap_atomic(struct dma_buf *dma_buf, + unsigned long page_num) +{ + /* TODO */ + + return NULL; +} + +static void udl_dmabuf_kunmap(struct dma_buf *dma_buf, + unsigned long page_num, void *addr) +{ + /* TODO */ +} + +static void udl_dmabuf_kunmap_atomic(struct dma_buf *dma_buf, + unsigned long page_num, + void *addr) +{ + /* TODO */ +} + +static int udl_dmabuf_mmap(struct dma_buf *dma_buf, + struct vm_area_struct *vma) +{ + /* TODO */ + + return -EINVAL; +} + +static struct dma_buf_ops udl_dmabuf_ops = { + .attach = udl_attach_dma_buf, + .detach = udl_detach_dma_buf, + .map_dma_buf = udl_map_dma_buf, + .unmap_dma_buf = udl_unmap_dma_buf, + .kmap = udl_dmabuf_kmap, + .kmap_atomic = udl_dmabuf_kmap_atomic, + .kunmap = udl_dmabuf_kunmap, + .kunmap_atomic = udl_dmabuf_kunmap_atomic, + .mmap = udl_dmabuf_mmap, + .release = drm_gem_dmabuf_release, +}; + +struct dma_buf *udl_gem_prime_export(struct drm_device *dev, + struct drm_gem_object *obj, int flags) +{ + return dma_buf_export(obj, &udl_dmabuf_ops, obj->size, flags, NULL); +} + +static int udl_prime_create(struct drm_device *dev, + size_t size, + struct sg_table *sg, + struct udl_gem_object **obj_p) +{ + struct udl_gem_object *obj; + int npages; + + npages = size / PAGE_SIZE; + + *obj_p = NULL; + obj = udl_gem_alloc_object(dev, npages * PAGE_SIZE); + if (!obj) + return -ENOMEM; + + obj->sg = sg; + obj->pages = drm_malloc_ab(npages, sizeof(struct page *)); + if (obj->pages == NULL) { + DRM_ERROR("obj pages is NULL %d\n", npages); + return -ENOMEM; + } + + drm_prime_sg_to_page_addr_arrays(sg, obj->pages, NULL, npages); + + *obj_p = obj; + return 0; +} + +struct drm_gem_object *udl_gem_prime_import(struct drm_device *dev, + struct dma_buf *dma_buf) +{ + struct dma_buf_attachment *attach; + struct sg_table *sg; + struct udl_gem_object *uobj; + int ret; + + /* need to attach */ + get_device(dev->dev); + attach = dma_buf_attach(dma_buf, dev->dev); + if (IS_ERR(attach)) { + put_device(dev->dev); + return ERR_CAST(attach); + } + + get_dma_buf(dma_buf); + + sg = dma_buf_map_attachment(attach, DMA_BIDIRECTIONAL); + if (IS_ERR(sg)) { + ret = PTR_ERR(sg); + goto fail_detach; + } + + ret = udl_prime_create(dev, dma_buf->size, sg, &uobj); + if (ret) + goto fail_unmap; + + uobj->base.import_attach = attach; + uobj->flags = UDL_BO_WC; + + return &uobj->base; + +fail_unmap: + dma_buf_unmap_attachment(attach, sg, DMA_BIDIRECTIONAL); +fail_detach: + dma_buf_detach(dma_buf, attach); + dma_buf_put(dma_buf); + put_device(dev->dev); + return ERR_PTR(ret); +} diff --git a/drivers/gpu/drm/udl/udl_drv.c b/drivers/gpu/drm/udl/udl_drv.c index 8607e9e..d5728ec 100644 --- a/drivers/gpu/drm/udl/udl_drv.c +++ b/drivers/gpu/drm/udl/udl_drv.c @@ -51,7 +51,9 @@ static struct drm_driver driver = { .dumb_destroy = drm_gem_dumb_destroy, .fops = &udl_driver_fops, + .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, + .gem_prime_export = udl_gem_prime_export, .gem_prime_import = udl_gem_prime_import, .name = DRIVER_NAME, diff --git a/drivers/gpu/drm/udl/udl_drv.h b/drivers/gpu/drm/udl/udl_drv.h index 3082780..1b132d7 100644 --- a/drivers/gpu/drm/udl/udl_drv.h +++ b/drivers/gpu/drm/udl/udl_drv.h @@ -124,6 +124,8 @@ int udl_gem_mmap(struct drm_file *file_priv, struct drm_device *dev, void udl_gem_free_object(struct drm_gem_object *gem_obj); struct udl_gem_object *udl_gem_alloc_object(struct drm_device *dev, size_t size); +struct dma_buf *udl_gem_prime_export(struct drm_device *dev, + struct drm_gem_object *obj, int flags); struct drm_gem_object *udl_gem_prime_import(struct drm_device *dev, struct dma_buf *dma_buf); diff --git a/drivers/gpu/drm/udl/udl_gem.c b/drivers/gpu/drm/udl/udl_gem.c index e00459d..692d6f2 100644 --- a/drivers/gpu/drm/udl/udl_gem.c +++ b/drivers/gpu/drm/udl/udl_gem.c @@ -240,74 +240,3 @@ unlock: mutex_unlock(&dev->struct_mutex); return ret; } - -static int udl_prime_create(struct drm_device *dev, - size_t size, - struct sg_table *sg, - struct udl_gem_object **obj_p) -{ - struct udl_gem_object *obj; - int npages; - - npages = size / PAGE_SIZE; - - *obj_p = NULL; - obj = udl_gem_alloc_object(dev, npages * PAGE_SIZE); - if (!obj) - return -ENOMEM; - - obj->sg = sg; - obj->pages = drm_malloc_ab(npages, sizeof(struct page *)); - if (obj->pages == NULL) { - DRM_ERROR("obj pages is NULL %d\n", npages); - return -ENOMEM; - } - - drm_prime_sg_to_page_addr_arrays(sg, obj->pages, NULL, npages); - - *obj_p = obj; - return 0; -} - -struct drm_gem_object *udl_gem_prime_import(struct drm_device *dev, - struct dma_buf *dma_buf) -{ - struct dma_buf_attachment *attach; - struct sg_table *sg; - struct udl_gem_object *uobj; - int ret; - - /* need to attach */ - get_device(dev->dev); - attach = dma_buf_attach(dma_buf, dev->dev); - if (IS_ERR(attach)) { - put_device(dev->dev); - return ERR_CAST(attach); - } - - get_dma_buf(dma_buf); - - sg = dma_buf_map_attachment(attach, DMA_BIDIRECTIONAL); - if (IS_ERR(sg)) { - ret = PTR_ERR(sg); - goto fail_detach; - } - - ret = udl_prime_create(dev, dma_buf->size, sg, &uobj); - if (ret) { - goto fail_unmap; - } - - uobj->base.import_attach = attach; - uobj->flags = UDL_BO_WC; - - return &uobj->base; - -fail_unmap: - dma_buf_unmap_attachment(attach, sg, DMA_BIDIRECTIONAL); -fail_detach: - dma_buf_detach(dma_buf, attach); - dma_buf_put(dma_buf); - put_device(dev->dev); - return ERR_PTR(ret); -} -- cgit v0.10.2 From 8975626ea35adcca561f8a81dedccfbc5dd8ec72 Mon Sep 17 00:00:00 2001 From: Zach Reizner Date: Wed, 29 Oct 2014 11:04:24 -0700 Subject: drm/cirrus: allow 32bpp framebuffers for cirrus drm This patch allows framebuffers for cirrus to be created with 32bpp pixel formats provided that they do not violate certain restrictions of the cirrus hardware. v2: Use pci resource length for vram size. Signed-off-by: Zach Reizner Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/cirrus/cirrus_drv.h b/drivers/gpu/drm/cirrus/cirrus_drv.h index d44e69d..693a456 100644 --- a/drivers/gpu/drm/cirrus/cirrus_drv.h +++ b/drivers/gpu/drm/cirrus/cirrus_drv.h @@ -210,6 +210,9 @@ int cirrus_framebuffer_init(struct drm_device *dev, struct drm_mode_fb_cmd2 *mode_cmd, struct drm_gem_object *obj); +bool cirrus_check_framebuffer(struct cirrus_device *cdev, int width, int height, + int bpp, int pitch); + /* cirrus_display.c */ int cirrus_modeset_init(struct cirrus_device *cdev); void cirrus_modeset_fini(struct cirrus_device *cdev); diff --git a/drivers/gpu/drm/cirrus/cirrus_fbdev.c b/drivers/gpu/drm/cirrus/cirrus_fbdev.c index d231b1c..502a89e 100644 --- a/drivers/gpu/drm/cirrus/cirrus_fbdev.c +++ b/drivers/gpu/drm/cirrus/cirrus_fbdev.c @@ -139,6 +139,7 @@ static int cirrusfb_create_object(struct cirrus_fbdev *afbdev, struct drm_gem_object **gobj_p) { struct drm_device *dev = afbdev->helper.dev; + struct cirrus_device *cdev = dev->dev_private; u32 bpp, depth; u32 size; struct drm_gem_object *gobj; @@ -146,8 +147,10 @@ static int cirrusfb_create_object(struct cirrus_fbdev *afbdev, int ret = 0; drm_fb_get_bpp_depth(mode_cmd->pixel_format, &depth, &bpp); - if (bpp > 24) + if (!cirrus_check_framebuffer(cdev, mode_cmd->width, mode_cmd->height, + bpp, mode_cmd->pitches[0])) return -EINVAL; + size = mode_cmd->pitches[0] * mode_cmd->height; ret = cirrus_gem_create(dev, size, true, &gobj); if (ret) diff --git a/drivers/gpu/drm/cirrus/cirrus_main.c b/drivers/gpu/drm/cirrus/cirrus_main.c index ab7cb54..4c2d68e 100644 --- a/drivers/gpu/drm/cirrus/cirrus_main.c +++ b/drivers/gpu/drm/cirrus/cirrus_main.c @@ -49,14 +49,16 @@ cirrus_user_framebuffer_create(struct drm_device *dev, struct drm_file *filp, struct drm_mode_fb_cmd2 *mode_cmd) { + struct cirrus_device *cdev = dev->dev_private; struct drm_gem_object *obj; struct cirrus_framebuffer *cirrus_fb; int ret; u32 bpp, depth; drm_fb_get_bpp_depth(mode_cmd->pixel_format, &depth, &bpp); - /* cirrus can't handle > 24bpp framebuffers at all */ - if (bpp > 24) + + if (!cirrus_check_framebuffer(cdev, mode_cmd->width, mode_cmd->height, + bpp, mode_cmd->pitches[0])) return ERR_PTR(-EINVAL); obj = drm_gem_object_lookup(dev, filp, mode_cmd->handles[0]); @@ -96,8 +98,7 @@ static int cirrus_vram_init(struct cirrus_device *cdev) { /* BAR 0 is VRAM */ cdev->mc.vram_base = pci_resource_start(cdev->dev->pdev, 0); - /* We have 4MB of VRAM */ - cdev->mc.vram_size = 4 * 1024 * 1024; + cdev->mc.vram_size = pci_resource_len(cdev->dev->pdev, 0); if (!request_mem_region(cdev->mc.vram_base, cdev->mc.vram_size, "cirrusdrmfb_vram")) { @@ -312,3 +313,21 @@ out_unlock: return ret; } + +bool cirrus_check_framebuffer(struct cirrus_device *cdev, int width, int height, + int bpp, int pitch) +{ + const int max_pitch = 0x1FF << 3; /* (4096 - 1) & ~111b bytes */ + const int max_size = cdev->mc.vram_size; + + if (bpp > 32) + return false; + + if (pitch > max_pitch) + return false; + + if (pitch * height > max_size) + return false; + + return true; +} -- cgit v0.10.2 From 87255483070c1e5a37b39cdda4d1e98fa3a13516 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 20:36:48 +0100 Subject: drm/i915: Ditch dev_priv->ums.mm_suspend Again just complicates gem init functions and makes a general mess out of everything. Good riddance! v2: In my enthusiasm to start removing dri1/ums crud I went overboard a bit and killed parts of hangcheck. Resurrect it. Acked-by: Chris Wilson Cc: Chris Wilson Cc: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 6c96253..c20fa30 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -815,9 +815,6 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) DRM_ERROR("failed to init modeset\n"); goto out_power_well; } - } else { - /* Start out suspended in ums mode. */ - dev_priv->ums.mm_suspended = 1; } i915_setup_sysfs(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index c9f248a..1df4079 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -852,10 +852,7 @@ int i915_reset(struct drm_device *dev) * was running at the time of the reset (i.e. we weren't VT * switched away). */ - if (drm_core_check_feature(dev, DRIVER_MODESET) || - !dev_priv->ums.mm_suspended) { - dev_priv->ums.mm_suspended = 0; - + if (drm_core_check_feature(dev, DRIVER_MODESET)) { /* Used to prevent gem_check_wedged returning -EAGAIN during gpu reset */ dev_priv->gpu_error.reload_in_reset = true; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index a52fba1..3d572a6 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1121,18 +1121,6 @@ struct i915_power_domains { struct i915_power_well *power_wells; }; -struct i915_ums_state { - /** - * Flag if the X Server, and thus DRM, is not currently in - * control of the device. - * - * This is set between LeaveVT and EnterVT. It needs to be - * replaced with a semaphore. It also needs to be - * transitioned away from for kernel modesetting. - */ - int mm_suspended; -}; - #define MAX_L3_SLICES 2 struct intel_l3_parity { u32 *remap_info[MAX_L3_SLICES]; @@ -1768,9 +1756,6 @@ struct drm_i915_private { uint32_t bios_vgacntr; - /* Old ums support infrastructure, same warning applies. */ - struct i915_ums_state ums; - /* Abstract the submission mechanism (legacy ringbuffer or execlists) away */ struct { int (*do_execbuf)(struct drm_device *dev, struct drm_file *file, diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index bd17109..af30283 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2501,15 +2501,13 @@ int __i915_add_request(struct intel_engine_cs *ring, ring->outstanding_lazy_seqno = 0; ring->preallocated_lazy_request = NULL; - if (!dev_priv->ums.mm_suspended) { - i915_queue_hangcheck(ring->dev); + i915_queue_hangcheck(ring->dev); - cancel_delayed_work_sync(&dev_priv->mm.idle_work); - queue_delayed_work(dev_priv->wq, - &dev_priv->mm.retire_work, - round_jiffies_up_relative(HZ)); - intel_mark_busy(dev_priv->dev); - } + cancel_delayed_work_sync(&dev_priv->mm.idle_work); + queue_delayed_work(dev_priv->wq, + &dev_priv->mm.retire_work, + round_jiffies_up_relative(HZ)); + intel_mark_busy(dev_priv->dev); if (out_seqno) *out_seqno = request->seqno; @@ -4651,9 +4649,6 @@ i915_gem_suspend(struct drm_device *dev) int ret = 0; mutex_lock(&dev->struct_mutex); - if (dev_priv->ums.mm_suspended) - goto err; - ret = i915_gpu_idle(dev); if (ret) goto err; @@ -4665,13 +4660,6 @@ i915_gem_suspend(struct drm_device *dev) i915_gem_evict_everything(dev); i915_gem_stop_ringbuffers(dev); - - /* Hack! Don't let anybody do execbuf while we don't control the chip. - * We need to replace this with a semaphore, or something. - * And not confound ums.mm_suspended! - */ - dev_priv->ums.mm_suspended = !drm_core_check_feature(dev, - DRIVER_MODESET); mutex_unlock(&dev->struct_mutex); del_timer_sync(&dev_priv->gpu_error.hangcheck_timer); diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index b16eee0..1173831 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -1338,12 +1338,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data, if (ret) goto pre_mutex_err; - if (dev_priv->ums.mm_suspended) { - mutex_unlock(&dev->struct_mutex); - ret = -EBUSY; - goto pre_mutex_err; - } - ctx = i915_gem_validate_context(dev, file, ring, ctx_id); if (IS_ERR(ctx)) { mutex_unlock(&dev->struct_mutex); -- cgit v0.10.2 From 377e91b2049352aed54e6dc3edc186f26da372ed Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 20:36:49 +0100 Subject: drm/i915: Sanitize ->lastclose With this all the ums nonsense around gem setup/teardown has disappeared, yay! Acked-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index c20fa30..f5ded5c 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -980,21 +980,8 @@ int i915_driver_open(struct drm_device *dev, struct drm_file *file) */ void i915_driver_lastclose(struct drm_device *dev) { - struct drm_i915_private *dev_priv = dev->dev_private; - - /* On gen6+ we refuse to init without kms enabled, but then the drm core - * goes right around and calls lastclose. Check for this and don't clean - * up anything. */ - if (!dev_priv) - return; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) { - intel_fbdev_restore_mode(dev); - vga_switcheroo_process_delayed_switch(); - return; - } - - i915_gem_lastclose(dev); + intel_fbdev_restore_mode(dev); + vga_switcheroo_process_delayed_switch(); } void i915_driver_preclose(struct drm_device *dev, struct drm_file *file) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 3d572a6..230a172 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2468,7 +2468,6 @@ int __must_check i915_vma_unbind(struct i915_vma *vma); int i915_gem_object_put_pages(struct drm_i915_gem_object *obj); void i915_gem_release_all_mmaps(struct drm_i915_private *dev_priv); void i915_gem_release_mmap(struct drm_i915_gem_object *obj); -void i915_gem_lastclose(struct drm_device *dev); int i915_gem_obj_prepare_shmem_read(struct drm_i915_gem_object *obj, int *needs_clflush); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index af30283..f2d7ce5 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4964,19 +4964,6 @@ i915_gem_cleanup_ringbuffer(struct drm_device *dev) dev_priv->gt.cleanup_ring(ring); } -void -i915_gem_lastclose(struct drm_device *dev) -{ - int ret; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return; - - ret = i915_gem_suspend(dev); - if (ret) - DRM_ERROR("failed to idle hardware: %d\n", ret); -} - static void init_ring_lists(struct intel_engine_cs *ring) { -- cgit v0.10.2 From f548c0e9d4200a660e607bb4d8824c21c6d0bd74 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 21:40:13 +0100 Subject: drm/i915: Can i915_gem_init_ioctl Found one more! With this we can clear up the ggtt init code a bit, yay! Acked-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index f5ded5c..ecee3bc 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1022,7 +1022,7 @@ const struct drm_ioctl_desc i915_ioctls[] = { DRM_IOCTL_DEF_DRV(I915_GET_VBLANK_PIPE, drm_noop, DRM_AUTH), DRM_IOCTL_DEF_DRV(I915_VBLANK_SWAP, drm_noop, DRM_AUTH), DRM_IOCTL_DEF_DRV(I915_HWS_ADDR, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), - DRM_IOCTL_DEF_DRV(I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY|DRM_UNLOCKED), + DRM_IOCTL_DEF_DRV(I915_GEM_INIT, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY|DRM_UNLOCKED), DRM_IOCTL_DEF_DRV(I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH|DRM_UNLOCKED), DRM_IOCTL_DEF_DRV(I915_GEM_EXECBUFFER2, i915_gem_execbuffer2, DRM_AUTH|DRM_UNLOCKED|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY|DRM_UNLOCKED), diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 230a172..02b3cb3 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2378,8 +2378,6 @@ void ibx_display_interrupt_update(struct drm_i915_private *dev_priv, ibx_display_interrupt_update((dev_priv), (bits), 0) /* i915_gem.c */ -int i915_gem_init_ioctl(struct drm_device *dev, void *data, - struct drm_file *file_priv); int i915_gem_create_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv); int i915_gem_pread_ioctl(struct drm_device *dev, void *data, diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index f2d7ce5..480d00e 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -160,33 +160,6 @@ i915_gem_object_is_inactive(struct drm_i915_gem_object *obj) } int -i915_gem_init_ioctl(struct drm_device *dev, void *data, - struct drm_file *file) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_gem_init *args = data; - - if (drm_core_check_feature(dev, DRIVER_MODESET)) - return -ENODEV; - - if (args->gtt_start >= args->gtt_end || - (args->gtt_end | args->gtt_start) & (PAGE_SIZE - 1)) - return -EINVAL; - - /* GEM with user mode setting was never supported on ilk and later. */ - if (INTEL_INFO(dev)->gen >= 5) - return -ENODEV; - - mutex_lock(&dev->struct_mutex); - i915_gem_setup_global_gtt(dev, args->gtt_start, args->gtt_end, - args->gtt_end); - dev_priv->gtt.mappable_end = args->gtt_end; - mutex_unlock(&dev->struct_mutex); - - return 0; -} - -int i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data, struct drm_file *file) { diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index bee5b0a..171f6ea 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -1664,10 +1664,10 @@ static void i915_gtt_color_adjust(struct drm_mm_node *node, } } -int i915_gem_setup_global_gtt(struct drm_device *dev, - unsigned long start, - unsigned long mappable_end, - unsigned long end) +static int i915_gem_setup_global_gtt(struct drm_device *dev, + unsigned long start, + unsigned long mappable_end, + unsigned long end) { /* Let GEM Manage all of the aperture. * diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h index d0562d0..beaf4bc 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.h +++ b/drivers/gpu/drm/i915/i915_gem_gtt.h @@ -274,8 +274,6 @@ struct i915_hw_ppgtt { int i915_gem_gtt_init(struct drm_device *dev); void i915_gem_init_global_gtt(struct drm_device *dev); -int i915_gem_setup_global_gtt(struct drm_device *dev, unsigned long start, - unsigned long mappable_end, unsigned long end); void i915_global_gtt_cleanup(struct drm_device *dev); -- cgit v0.10.2 From 93ee29203f506582cca2bcec5f05041526d9ab0a Mon Sep 17 00:00:00 2001 From: Tom O'Rourke Date: Wed, 19 Nov 2014 14:21:52 -0800 Subject: drm/i915: Use efficient frequency for HSW/BDW Added gen6_init_rps_frequencies() to initialize the rps frequency values. This function replaces parse_rp_state_cap(). In addition to reading RPn, RP0, and RP1 from RP_STATE_CAP register, the new function reads efficient frequency (aka RPe) from pcode for Haswell and Broadwell and sets the turbo softlimits. The turbo minimum frequency softlimit is set to RPe for Haswell and Broadwell and to RPn otherwise. For RPe, the efficiency is based on the frequency/power ratio (MHz/W); this is considering GT power and not package power. The efficent frequency is the highest frequency for which the frequency/power ratio is within some threshold of the highest frequency/power ratio. A fixed decrease in frequency results in smaller decrease in power at frequencies less than RPe than at frequencies above RPe. v2: Following suggestions from Chris Wilson and Daniel Vetter to extend and rename parse_rp_state_cap and to open-code a poorly named function. Signed-off-by: Tom O'Rourke Reviewed-by: Chris Wilson [danvet: Remove unused variables.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 7a77cd5..3a51c05 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -6038,6 +6038,7 @@ enum punit_power_well { #define GEN6_ENCODE_RC6_VID(mv) (((mv) - 245) / 5) #define GEN6_DECODE_RC6_VID(vids) (((vids) * 5) + 245) #define DISPLAY_IPS_CONTROL 0x19 +#define HSW_PCODE_DYNAMIC_DUTY_CYCLE_CONTROL 0x1A #define GEN6_PCODE_DATA 0x138128 #define GEN6_PCODE_FREQ_IA_RATIO_SHIFT 8 #define GEN6_PCODE_FREQ_RING_RATIO_SHIFT 16 diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 417ba88..eb2a0a8 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4609,25 +4609,45 @@ int intel_enable_rc6(const struct drm_device *dev) return i915.enable_rc6; } -static void parse_rp_state_cap(struct drm_i915_private *dev_priv, u32 rp_state_cap) +static void gen6_init_rps_frequencies(struct drm_device *dev) { + struct drm_i915_private *dev_priv = dev->dev_private; + uint32_t rp_state_cap; + u32 ddcc_status = 0; + int ret; + + rp_state_cap = I915_READ(GEN6_RP_STATE_CAP); /* All of these values are in units of 50MHz */ dev_priv->rps.cur_freq = 0; - /* static values from HW: RP0 < RPe < RP1 < RPn (min_freq) */ - dev_priv->rps.rp1_freq = (rp_state_cap >> 8) & 0xff; + /* static values from HW: RP0 > RP1 > RPn (min_freq) */ dev_priv->rps.rp0_freq = (rp_state_cap >> 0) & 0xff; + dev_priv->rps.rp1_freq = (rp_state_cap >> 8) & 0xff; dev_priv->rps.min_freq = (rp_state_cap >> 16) & 0xff; - /* XXX: only BYT has a special efficient freq */ - dev_priv->rps.efficient_freq = dev_priv->rps.rp1_freq; /* hw_max = RP0 until we check for overclocking */ dev_priv->rps.max_freq = dev_priv->rps.rp0_freq; + dev_priv->rps.efficient_freq = dev_priv->rps.rp1_freq; + if (IS_HASWELL(dev) || IS_BROADWELL(dev)) { + ret = sandybridge_pcode_read(dev_priv, + HSW_PCODE_DYNAMIC_DUTY_CYCLE_CONTROL, + &ddcc_status); + if (0 == ret) + dev_priv->rps.efficient_freq = + (ddcc_status >> 8) & 0xff; + } + /* Preserve min/max settings in case of re-init */ if (dev_priv->rps.max_freq_softlimit == 0) dev_priv->rps.max_freq_softlimit = dev_priv->rps.max_freq; - if (dev_priv->rps.min_freq_softlimit == 0) - dev_priv->rps.min_freq_softlimit = dev_priv->rps.min_freq; + if (dev_priv->rps.min_freq_softlimit == 0) { + if (IS_HASWELL(dev) || IS_BROADWELL(dev)) + dev_priv->rps.min_freq_softlimit = + dev_priv->rps.efficient_freq; + else + dev_priv->rps.min_freq_softlimit = + dev_priv->rps.min_freq; + } } static void gen9_enable_rps(struct drm_device *dev) @@ -4673,7 +4693,7 @@ static void gen8_enable_rps(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; struct intel_engine_cs *ring; - uint32_t rc6_mask = 0, rp_state_cap; + uint32_t rc6_mask = 0; int unused; /* 1a: Software RC state - RC0 */ @@ -4686,8 +4706,8 @@ static void gen8_enable_rps(struct drm_device *dev) /* 2a: Disable RC states. */ I915_WRITE(GEN6_RC_CONTROL, 0); - rp_state_cap = I915_READ(GEN6_RP_STATE_CAP); - parse_rp_state_cap(dev_priv, rp_state_cap); + /* Initialize rps frequencies */ + gen6_init_rps_frequencies(dev); /* 2b: Program RC6 thresholds.*/ I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16); @@ -4754,7 +4774,6 @@ static void gen6_enable_rps(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; struct intel_engine_cs *ring; - u32 rp_state_cap; u32 rc6vids, pcu_mbox = 0, rc6_mask = 0; u32 gtfifodbg; int rc6_mode; @@ -4778,9 +4797,8 @@ static void gen6_enable_rps(struct drm_device *dev) gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL); - rp_state_cap = I915_READ(GEN6_RP_STATE_CAP); - - parse_rp_state_cap(dev_priv, rp_state_cap); + /* Initialize rps frequencies */ + gen6_init_rps_frequencies(dev); /* disable the counters and set deterministic thresholds */ I915_WRITE(GEN6_RC_CONTROL, 0); -- cgit v0.10.2 From f4ab408c4b55fcbc447cf0288456613bad824626 Mon Sep 17 00:00:00 2001 From: Tom O'Rourke Date: Wed, 19 Nov 2014 14:21:53 -0800 Subject: drm/i915: Keep min freq above floor on HSW/BDW Set the min_freq_softlimit to max(RPe, 450MHz). Setting a floor can ensure a minimum experience level. The 450MHz value came from a power and performance study of various types of workloads (3D, Media, GPGPU, idle, etc). v2: rebased Signed-off-by: Tom O'Rourke Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index eb2a0a8..6d0795d 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4643,7 +4643,8 @@ static void gen6_init_rps_frequencies(struct drm_device *dev) if (dev_priv->rps.min_freq_softlimit == 0) { if (IS_HASWELL(dev) || IS_BROADWELL(dev)) dev_priv->rps.min_freq_softlimit = - dev_priv->rps.efficient_freq; + /* max(RPe, 450 MHz) */ + max(dev_priv->rps.efficient_freq, (u8) 9); else dev_priv->rps.min_freq_softlimit = dev_priv->rps.min_freq; -- cgit v0.10.2 From c7f3153a61b543830d397cda8918ef2d6330c6c7 Mon Sep 17 00:00:00 2001 From: Tom O'Rourke Date: Wed, 19 Nov 2014 14:21:54 -0800 Subject: drm/i915: change initial rps frequency for gen8 In gen8_enable_rps, change the initial rps setting to the min_freq_softlimit (same as gen6_enable_rps). Signed-off-by: Tom O'Rourke Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 6d0795d..70e7547 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4766,7 +4766,8 @@ static void gen8_enable_rps(struct drm_device *dev) /* 6: Ring frequency + overclocking (our driver does this later */ - gen6_set_rps(dev, (I915_READ(GEN6_GT_PERF_STATUS) & 0xff00) >> 8); + dev_priv->rps.power = HIGH_POWER; /* force a reset */ + gen6_set_rps(dev_priv->dev, dev_priv->rps.min_freq_softlimit); gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL); } -- cgit v0.10.2 From 6985b352183fc00df22a769892a24bc438db5357 Mon Sep 17 00:00:00 2001 From: Tom O'Rourke Date: Wed, 19 Nov 2014 14:21:55 -0800 Subject: drm/i915: Update ring freq for full gpu freq range In __gen6_update_ring_freq, use the full range of possible gpu frequencies from max_freq to min_freq. The actual gpu frequency could be outside the range from max_freq_softlimit to min_freq_softlimit due to power/thermal constraints. Signed-off-by: Tom O'Rourke Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 70e7547..f103fff 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4915,9 +4915,9 @@ static void __gen6_update_ring_freq(struct drm_device *dev) * to use for memory access. We do this by specifying the IA frequency * the PCU should use as a reference to determine the ring frequency. */ - for (gpu_freq = dev_priv->rps.max_freq_softlimit; gpu_freq >= dev_priv->rps.min_freq_softlimit; + for (gpu_freq = dev_priv->rps.max_freq; gpu_freq >= dev_priv->rps.min_freq; gpu_freq--) { - int diff = dev_priv->rps.max_freq_softlimit - gpu_freq; + int diff = dev_priv->rps.max_freq - gpu_freq; unsigned int ia_freq = 0, ring_freq = 0; if (INTEL_INFO(dev)->gen >= 8) { -- cgit v0.10.2 From 656bfa3afc14e45e2d9e1624bf60d79b3beb12f2 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Thu, 20 Nov 2014 09:26:30 +0100 Subject: drm/i915: Pin tiled objects for L-shaped configs Let's just throw in the towel on this one and take the cheap way out. Based on a patch from Chris Wilson, but checking for a different bit. Chris' patch checked for even bank layout, this one here for a magic bit. Given the evidence we've gathered (not much) both work I think, but checking for the magic bit might be more accurate. Anyway, works on my gm45 here. For paranoi restrict to gen4 (and mobile), since we've only ever seen this on gm45 and i965gm. Also add some debugfs output so that we can skip the tiled swapping tests properly in these cases. v2: Clean up the quirk'ed pin count in free_object to avoid upsetting the WARN_ON. Spotted by Chris. Cc: Chris Wilson Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=28813 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=45092 Signed-off-by: Daniel Vetter Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index bb14583..a27b1e4 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1983,6 +1983,8 @@ static int i915_swizzle_info(struct seq_file *m, void *data) if (IS_GEN3(dev) || IS_GEN4(dev)) { seq_printf(m, "DDC = 0x%08x\n", I915_READ(DCC)); + seq_printf(m, "DDC2 = 0x%08x\n", + I915_READ(DCC2)); seq_printf(m, "C0DRB3 = 0x%04x\n", I915_READ16(C0DRB3)); seq_printf(m, "C1DRB3 = 0x%04x\n", @@ -2005,6 +2007,10 @@ static int i915_swizzle_info(struct seq_file *m, void *data) seq_printf(m, "DISP_ARB_CTL = 0x%08x\n", I915_READ(DISP_ARB_CTL)); } + + if (dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES) + seq_puts(m, "L-shaped memory detected\n"); + intel_runtime_pm_put(dev_priv); mutex_unlock(&dev->struct_mutex); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 02b3cb3..5448ce9 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -764,6 +764,7 @@ enum intel_sbi_destination { #define QUIRK_INVERT_BRIGHTNESS (1<<2) #define QUIRK_BACKLIGHT_PRESENT (1<<3) #define QUIRK_PIPEB_FORCE (1<<4) +#define QUIRK_PIN_SWIZZLED_PAGES (1<<5) struct intel_fbdev; struct intel_fbc_work; diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 480d00e..3c64eb6 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2176,6 +2176,10 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj) if (i915_gem_object_needs_bit17_swizzle(obj)) i915_gem_object_do_bit_17_swizzle(obj); + if (obj->tiling_mode != I915_TILING_NONE && + dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES) + i915_gem_object_pin_pages(obj); + return 0; err_pages: @@ -4374,6 +4378,7 @@ int i915_gem_madvise_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) { + struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_gem_madvise *args = data; struct drm_i915_gem_object *obj; int ret; @@ -4401,6 +4406,15 @@ i915_gem_madvise_ioctl(struct drm_device *dev, void *data, goto out; } + if (obj->pages && + obj->tiling_mode != I915_TILING_NONE && + dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES) { + if (obj->madv == I915_MADV_WILLNEED) + i915_gem_object_unpin_pages(obj); + if (args->madv == I915_MADV_WILLNEED) + i915_gem_object_pin_pages(obj); + } + if (obj->madv != __I915_MADV_PURGED) obj->madv = args->madv; @@ -4550,6 +4564,11 @@ void i915_gem_free_object(struct drm_gem_object *gem_obj) WARN_ON(obj->frontbuffer_bits); + if (obj->pages && obj->madv == I915_MADV_WILLNEED && + dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES && + obj->tiling_mode != I915_TILING_NONE) + i915_gem_object_unpin_pages(obj); + if (WARN_ON(obj->pages_pin_count)) obj->pages_pin_count = 0; if (discard_backing_storage(obj)) diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index cd7f473..4727a4e 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -178,6 +178,15 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev) } break; } + + /* check for L-shaped memory aka modified enhanced addressing */ + if (IS_GEN4(dev)) { + uint32_t ddc2 = I915_READ(DCC2); + + if (!(ddc2 & DCC2_MODIFIED_ENHANCED_DISABLE)) + dev_priv->quirks |= QUIRK_PIN_SWIZZLED_PAGES; + } + if (dcc == 0xffffffff) { DRM_ERROR("Couldn't read from MCHBAR. " "Disabling tiling.\n"); @@ -380,6 +389,15 @@ i915_gem_set_tiling(struct drm_device *dev, void *data, ret = i915_gem_object_ggtt_unbind(obj); if (ret == 0) { + if (obj->pages && + obj->madv == I915_MADV_WILLNEED && + dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES) { + if (args->tiling_mode == I915_TILING_NONE) + i915_gem_object_unpin_pages(obj); + if (obj->tiling_mode == I915_TILING_NONE) + i915_gem_object_pin_pages(obj); + } + obj->fence_dirty = obj->last_fenced_seqno || obj->fence_reg != I915_FENCE_REG_NONE; diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 3a51c05..3102907 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -2043,6 +2043,8 @@ enum punit_power_well { #define DCC_ADDRESSING_MODE_MASK (3 << 0) #define DCC_CHANNEL_XOR_DISABLE (1 << 10) #define DCC_CHANNEL_XOR_BIT_17 (1 << 9) +#define DCC2 0x10204 +#define DCC2_MODIFIED_ENHANCED_DISABLE (1 << 20) /* Pineview MCH register contains DDR3 setting */ #define CSHRDDR3CTL 0x101a8 -- cgit v0.10.2 From 6b312cd3668b066467228b51142d05b94c2ad7d8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 19 Nov 2014 20:07:42 +0200 Subject: drm/i915: Read power well status before other registers for drpc info MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Trying to read the status of the power wells right after taking forcewake for the other register reads makes little sense. Most of the time the power wells will still be up due to the recent forcewake. Instead do the power well status read first, and only then read the register needing forcewake. This way the reported power well status can actually reflect what's going on in the system. Cc: Deepak S Signed-off-by: Ville Syrjälä Reviewed-by: Deepak S Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index a27b1e4..f91e7f7 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1240,11 +1240,12 @@ static int vlv_drpc_info(struct seq_file *m) struct drm_info_node *node = m->private; struct drm_device *dev = node->minor->dev; struct drm_i915_private *dev_priv = dev->dev_private; - u32 rpmodectl1, rcctl1; + u32 rpmodectl1, rcctl1, pw_status; unsigned fw_rendercount = 0, fw_mediacount = 0; intel_runtime_pm_get(dev_priv); + pw_status = I915_READ(VLV_GTLC_PW_STATUS); rpmodectl1 = I915_READ(GEN6_RP_CONTROL); rcctl1 = I915_READ(GEN6_RC_CONTROL); @@ -1263,11 +1264,9 @@ static int vlv_drpc_info(struct seq_file *m) yesno(rcctl1 & (GEN7_RC_CTL_TO_MODE | GEN6_RC_CTL_EI_MODE(1)))); seq_printf(m, "Render Power Well: %s\n", - (I915_READ(VLV_GTLC_PW_STATUS) & - VLV_GTLC_PW_RENDER_STATUS_MASK) ? "Up" : "Down"); + (pw_status & VLV_GTLC_PW_RENDER_STATUS_MASK) ? "Up" : "Down"); seq_printf(m, "Media Power Well: %s\n", - (I915_READ(VLV_GTLC_PW_STATUS) & - VLV_GTLC_PW_MEDIA_STATUS_MASK) ? "Up" : "Down"); + (pw_status & VLV_GTLC_PW_MEDIA_STATUS_MASK) ? "Up" : "Down"); seq_printf(m, "Render RC6 residency since boot: %u\n", I915_READ(VLV_GT_RENDER_RC6)); -- cgit v0.10.2 From 4597a88a1eba0f513583517cbdb5d90bada68f0f Mon Sep 17 00:00:00 2001 From: Zhe Wang Date: Thu, 20 Nov 2014 13:42:55 +0000 Subject: drm/i915/skl: Gen9 multi-engine forcewake Enable multi-engine forcewake for Gen9. v2: (Damien) - Rebase on top of nightly - Move the register range definitions to intel_uncore.c - Whitespace fixes v3: (Addressing Mika's comment, done by Damien) - Use REG_RANGE() (introduced after the patch was written) - Add a SKL_NEEDS_FORCE_WAKE() macro that gets rid of a useless comparison to FORCEWAKE (reg 0xa18c is not used on SKL) v4: (Damien) - Use newly introduced ASSIGN_READ/WRITE_MMIO_VFUNCS() macros Reviewed-by: Mika Kuoppala Signed-off-by: Zhe Wang Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index c98421d..4c10faa 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -670,6 +670,34 @@ void assert_force_wake_inactive(struct drm_i915_private *dev_priv) REG_RANGE((reg), 0x14000, 0x14400) || \ REG_RANGE((reg), 0x22000, 0x24000)) +#define FORCEWAKE_GEN9_UNCORE_RANGE_OFFSET(reg) \ + REG_RANGE((reg), 0xC00, 0x2000) + +#define FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg) \ + (REG_RANGE((reg), 0x2000, 0x4000) || \ + REG_RANGE((reg), 0x5200, 0x8000) || \ + REG_RANGE((reg), 0x8300, 0x8500) || \ + REG_RANGE((reg), 0x8C00, 0x8D00) || \ + REG_RANGE((reg), 0xB000, 0xB480) || \ + REG_RANGE((reg), 0xE000, 0xE800)) + +#define FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg) \ + (REG_RANGE((reg), 0x8800, 0x8A00) || \ + REG_RANGE((reg), 0xD000, 0xD800) || \ + REG_RANGE((reg), 0x12000, 0x14000) || \ + REG_RANGE((reg), 0x1A000, 0x1EA00) || \ + REG_RANGE((reg), 0x30000, 0x40000)) + +#define FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg) \ + REG_RANGE((reg), 0x9400, 0x9800) + +#define FORCEWAKE_GEN9_BLITTER_RANGE_OFFSET(reg) \ + ((reg) < 0x40000 &&\ + !FORCEWAKE_GEN9_UNCORE_RANGE_OFFSET(reg) && \ + !FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg) && \ + !FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg) && \ + !FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) + static void ilk_dummy_write(struct drm_i915_private *dev_priv) { @@ -800,6 +828,45 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \ REG_READ_FOOTER; \ } +#define SKL_NEEDS_FORCE_WAKE(dev_priv, reg) \ + ((reg) < 0x40000 && !FORCEWAKE_GEN9_UNCORE_RANGE_OFFSET(reg)) + +#define __gen9_read(x) \ +static u##x \ +gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \ + REG_READ_HEADER(x); \ + if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg))) { \ + val = __raw_i915_read##x(dev_priv, reg); \ + } else { \ + unsigned fwengine = 0; \ + if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg)) { \ + if (dev_priv->uncore.fw_rendercount == 0) \ + fwengine = FORCEWAKE_RENDER; \ + } else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) { \ + if (dev_priv->uncore.fw_mediacount == 0) \ + fwengine = FORCEWAKE_MEDIA; \ + } else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) { \ + if (dev_priv->uncore.fw_rendercount == 0) \ + fwengine |= FORCEWAKE_RENDER; \ + if (dev_priv->uncore.fw_mediacount == 0) \ + fwengine |= FORCEWAKE_MEDIA; \ + } else { \ + if (dev_priv->uncore.fw_blittercount == 0) \ + fwengine = FORCEWAKE_BLITTER; \ + } \ + if (fwengine) \ + dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \ + val = __raw_i915_read##x(dev_priv, reg); \ + if (fwengine) \ + dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \ + } \ + REG_READ_FOOTER; \ +} + +__gen9_read(8) +__gen9_read(16) +__gen9_read(32) +__gen9_read(64) __chv_read(8) __chv_read(16) __chv_read(32) @@ -821,6 +888,7 @@ __gen4_read(16) __gen4_read(32) __gen4_read(64) +#undef __gen9_read #undef __chv_read #undef __vlv_read #undef __gen6_read @@ -958,6 +1026,45 @@ chv_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) REG_WRITE_FOOTER; \ } +#define __gen9_write(x) \ +static void \ +gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \ + bool trace) { \ + REG_WRITE_HEADER; \ + if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg))) { \ + __raw_i915_write##x(dev_priv, reg, val); \ + } else { \ + unsigned fwengine = 0; \ + if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg)) { \ + if (dev_priv->uncore.fw_rendercount == 0) \ + fwengine = FORCEWAKE_RENDER; \ + } else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) { \ + if (dev_priv->uncore.fw_mediacount == 0) \ + fwengine = FORCEWAKE_MEDIA; \ + } else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) { \ + if (dev_priv->uncore.fw_rendercount == 0) \ + fwengine |= FORCEWAKE_RENDER; \ + if (dev_priv->uncore.fw_mediacount == 0) \ + fwengine |= FORCEWAKE_MEDIA; \ + } else { \ + if (dev_priv->uncore.fw_blittercount == 0) \ + fwengine = FORCEWAKE_BLITTER; \ + } \ + if (fwengine) \ + dev_priv->uncore.funcs.force_wake_get(dev_priv, \ + fwengine); \ + __raw_i915_write##x(dev_priv, reg, val); \ + if (fwengine) \ + dev_priv->uncore.funcs.force_wake_put(dev_priv, \ + fwengine); \ + } \ + REG_WRITE_FOOTER; \ +} + +__gen9_write(8) +__gen9_write(16) +__gen9_write(32) +__gen9_write(64) __chv_write(8) __chv_write(16) __chv_write(32) @@ -983,6 +1090,7 @@ __gen4_write(16) __gen4_write(32) __gen4_write(64) +#undef __gen9_write #undef __chv_write #undef __gen8_write #undef __hsw_write @@ -1066,6 +1174,13 @@ void intel_uncore_init(struct drm_device *dev) switch (INTEL_INFO(dev)->gen) { default: + WARN_ON(1); + return; + case 9: + ASSIGN_WRITE_MMIO_VFUNCS(gen9); + ASSIGN_READ_MMIO_VFUNCS(gen9); + break; + case 8: if (IS_CHERRYVIEW(dev)) { ASSIGN_WRITE_MMIO_VFUNCS(chv); ASSIGN_READ_MMIO_VFUNCS(chv); -- cgit v0.10.2 From 7c859007fdae545b71db7a2d58a9214da5806800 Mon Sep 17 00:00:00 2001 From: Zhe Wang Date: Thu, 20 Nov 2014 13:42:56 +0000 Subject: drm/i915: Gen9 shadowed registers For MMIO registers which are shadowed, force wake is not needed to write to these registers. v2: Rebase on top of nightly (Damien) v3: Rebase on top of "Gen9 multiple-engine forcewake" changes v4: (Mika, Bob, done by Damien) - Reorder the shadowed registers by popularity Reviewed-by: Mika Kuoppala Signed-off-by: Zhe Wang Signed-off-by: Damien Lespiau Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 4c10faa..1a3e485 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -1026,12 +1026,36 @@ chv_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) REG_WRITE_FOOTER; \ } +static const u32 gen9_shadowed_regs[] = { + RING_TAIL(RENDER_RING_BASE), + RING_TAIL(GEN6_BSD_RING_BASE), + RING_TAIL(VEBOX_RING_BASE), + RING_TAIL(BLT_RING_BASE), + FORCEWAKE_BLITTER_GEN9, + FORCEWAKE_RENDER_GEN9, + FORCEWAKE_MEDIA_GEN9, + GEN6_RPNSWREQ, + GEN6_RC_VIDEO_FREQ, + /* TODO: Other registers are not yet used */ +}; + +static bool is_gen9_shadowed(struct drm_i915_private *dev_priv, u32 reg) +{ + int i; + for (i = 0; i < ARRAY_SIZE(gen9_shadowed_regs); i++) + if (reg == gen9_shadowed_regs[i]) + return true; + + return false; +} + #define __gen9_write(x) \ static void \ gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \ bool trace) { \ REG_WRITE_HEADER; \ - if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg))) { \ + if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) || \ + is_gen9_shadowed(dev_priv, reg)) { \ __raw_i915_write##x(dev_priv, reg, val); \ } else { \ unsigned fwengine = 0; \ -- cgit v0.10.2 From 55072d194ca1bee7dcbf2e24d30728c0141085f6 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Thu, 20 Nov 2014 16:10:28 +0100 Subject: drm/i915: Dump hdmi pipe_config state Makes it easier to debug infoframe mismatches. Cc: Paulo Zanoni Cc: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 9c983e4..83af88c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -10001,6 +10001,10 @@ static void intel_dump_pipe_config(struct intel_crtc *crtc, pipe_config->dp_m2_n2.link_n, pipe_config->dp_m2_n2.tu); + DRM_DEBUG_KMS("audio: %i, infoframes: %i\n", + pipe_config->has_audio, + pipe_config->has_infoframe); + DRM_DEBUG_KMS("requested mode:\n"); drm_mode_debug_printmodeline(&pipe_config->requested_mode); DRM_DEBUG_KMS("adjusted mode:\n"); -- cgit v0.10.2 From 738c05c07d74769a1c5704636969b9b21cc96835 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 19 Nov 2014 16:25:37 +0200 Subject: drm/i915: vlv: fix cdclk setting during modeset while suspended MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Currently after doing DPMS-OFF on all outputs CDCLK won't be set to its minimum value as it should. A subsequent modeset to turn off all outputs will thus run with all power domains disabled, and notice that it needs to change CDCLK to its minimum value. Since the power domains are disabled this will emit a register-access-while-suspended WARN and fail to set the minimum freq. The proper solution for this is to set the minimum frequency during DPMS-OFF. That needs a bigger rework that would take into account the user DPMS setting too during the calculation of the new modesetting configuration. Until that's done this stop-gap solution gets the PIPE-A power domain during setting the CDCLK; this domain covers the HW blocks needed for this. Idea to use PIPE-A domain from Ville. Testcase: igt/pm_rpm Reference: https://bugs.freedesktop.org/show_bug.cgi?id=82939 Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 83af88c..d4aa2dd 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -4939,10 +4939,23 @@ static void valleyview_modeset_global_resources(struct drm_device *dev) int req_cdclk = valleyview_calc_cdclk(dev_priv, max_pixclk); if (req_cdclk != dev_priv->vlv_cdclk_freq) { + /* + * FIXME: We can end up here with all power domains off, yet + * with a CDCLK frequency other than the minimum. To account + * for this take the PIPE-A power domain, which covers the HW + * blocks needed for the following programming. This can be + * removed once it's guaranteed that we get here either with + * the minimum CDCLK set, or the required power domains + * enabled. + */ + intel_display_power_get(dev_priv, POWER_DOMAIN_PIPE_A); + if (IS_CHERRYVIEW(dev)) cherryview_set_cdclk(dev, req_cdclk); else valleyview_set_cdclk(dev, req_cdclk); + + intel_display_power_put(dev_priv, POWER_DOMAIN_PIPE_A); } } -- cgit v0.10.2 From 2837ac40698d0931727b957a40c8c8ea27c3bcb2 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 19 Nov 2014 16:25:38 +0200 Subject: drm/i915: vlv: increase timeout when setting idle GPU freq MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit I saw punit timeouts in vlv_set_rps_idle() while running various subtests of pm_rpm. Increasing the timeout to 100ms got rid of the issue. Testcase: igt/pm_rpm Reference: https://bugs.freedesktop.org/show_bug.cgi?id=82939 Signed-off-by: Imre Deak Tested-by: Guo Jinxian Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index f103fff..14f34f0 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4451,7 +4451,7 @@ static void vlv_set_rps_idle(struct drm_i915_private *dev_priv) dev_priv->rps.min_freq_softlimit); if (wait_for(((vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS)) - & GENFREQSTATUS) == 0, 5)) + & GENFREQSTATUS) == 0, 100)) DRM_ERROR("timed out waiting for Punit\n"); vlv_force_gfx_clock(dev_priv, false); -- cgit v0.10.2 From cc1df8a3fe39357fa5b727780129f50a68bff8f9 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 19 Nov 2014 18:38:39 +0100 Subject: drm/i915: Use ggtt error obj capture helper for gen8 semaphores Spotted while reading and trying to understand how our error capture code deals with full ppgtt. Cc: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 82111b8..cdaee6c 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -809,9 +809,8 @@ static void gen8_record_semaphore_state(struct drm_i915_private *dev_priv, if (!error->semaphore_obj) error->semaphore_obj = - i915_error_object_create(dev_priv, - dev_priv->semaphore_obj, - &dev_priv->gtt.base); + i915_error_ggtt_object_create(dev_priv, + dev_priv->semaphore_obj); for_each_ring(to, dev_priv, i) { int idx; -- cgit v0.10.2 From 99990f1b0b95b3c6056d7d80669ab9a8644fed34 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Thu, 20 Nov 2014 17:59:22 +0200 Subject: drm/i915: remove the IRQs enabled WARN from intel_disable_gt_powersave This function can be called now with i915 interrupts enabled, so the corresponding WARN is incorrect, remove it. I think this was spotted by Paulo during his review, but since I already removed the same WARN from intel_suspend_gt_powersave() I missed then his point. Spotted-by: Paulo Zanoni Signed-off-by: Imre Deak Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 14f34f0..9af0af4 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6223,9 +6223,6 @@ void intel_disable_gt_powersave(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - /* Interrupts should be disabled already to avoid re-arming. */ - WARN_ON(intel_irqs_enabled(dev_priv)); - if (IS_IRONLAKE_M(dev)) { ironlake_disable_drps(dev); ironlake_disable_rc6(dev); -- cgit v0.10.2 From 78b1a6010b46a69bcd47b723a80f92693f26d17b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 18 Nov 2014 18:00:08 +0900 Subject: drm/radeon: Use cursor_set2 hook for enabling / disabling the HW cursor MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The cursor_set2 hook provides the cursor hotspot position within the cursor image. When the hotspot position changes, we can adjust the cursor position such that the hotspot doesn't move on the screen. This prevents the cursor from appearing to intermittently jump around on the screen when the position of the hotspot within the cursor image changes. Reviewed-by: Alex Deucher Signed-off-by: Michel Dänzer Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_cursor.c b/drivers/gpu/drm/radeon/radeon_cursor.c index 9630e8d..fd4bddf 100644 --- a/drivers/gpu/drm/radeon/radeon_cursor.c +++ b/drivers/gpu/drm/radeon/radeon_cursor.c @@ -117,8 +117,10 @@ static void radeon_show_cursor(struct drm_crtc *crtc) } } +static int radeon_cursor_move_locked(struct drm_crtc *crtc, int x, int y); + static void radeon_set_cursor(struct drm_crtc *crtc, struct drm_gem_object *obj, - uint64_t gpu_addr) + uint64_t gpu_addr, int hot_x, int hot_y) { struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); struct radeon_device *rdev = crtc->dev->dev_private; @@ -142,13 +144,28 @@ static void radeon_set_cursor(struct drm_crtc *crtc, struct drm_gem_object *obj, /* offset is from DISP(2)_BASE_ADDRESS */ WREG32(RADEON_CUR_OFFSET + radeon_crtc->crtc_offset, radeon_crtc->legacy_cursor_offset); } + + if (hot_x != radeon_crtc->cursor_hot_x || + hot_y != radeon_crtc->cursor_hot_y) { + int x, y; + + x = radeon_crtc->cursor_x + radeon_crtc->cursor_hot_x - hot_x; + y = radeon_crtc->cursor_y + radeon_crtc->cursor_hot_y - hot_y; + + radeon_cursor_move_locked(crtc, x, y); + + radeon_crtc->cursor_hot_x = hot_x; + radeon_crtc->cursor_hot_y = hot_y; + } } -int radeon_crtc_cursor_set(struct drm_crtc *crtc, - struct drm_file *file_priv, - uint32_t handle, - uint32_t width, - uint32_t height) +int radeon_crtc_cursor_set2(struct drm_crtc *crtc, + struct drm_file *file_priv, + uint32_t handle, + uint32_t width, + uint32_t height, + int32_t hot_x, + int32_t hot_y) { struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); struct radeon_device *rdev = crtc->dev->dev_private; @@ -192,7 +209,7 @@ int radeon_crtc_cursor_set(struct drm_crtc *crtc, radeon_crtc->cursor_height = height; radeon_lock_cursor(crtc, true); - radeon_set_cursor(crtc, obj, gpu_addr); + radeon_set_cursor(crtc, obj, gpu_addr, hot_x, hot_y); radeon_show_cursor(crtc); radeon_lock_cursor(crtc, false); @@ -215,8 +232,7 @@ fail: return ret; } -int radeon_crtc_cursor_move(struct drm_crtc *crtc, - int x, int y) +static int radeon_cursor_move_locked(struct drm_crtc *crtc, int x, int y) { struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); struct radeon_device *rdev = crtc->dev->dev_private; @@ -281,7 +297,6 @@ int radeon_crtc_cursor_move(struct drm_crtc *crtc, } } - radeon_lock_cursor(crtc, true); if (ASIC_IS_DCE4(rdev)) { WREG32(EVERGREEN_CUR_POSITION + radeon_crtc->crtc_offset, (x << 16) | y); WREG32(EVERGREEN_CUR_HOT_SPOT + radeon_crtc->crtc_offset, (xorigin << 16) | yorigin); @@ -308,7 +323,21 @@ int radeon_crtc_cursor_move(struct drm_crtc *crtc, WREG32(RADEON_CUR_OFFSET + radeon_crtc->crtc_offset, (radeon_crtc->legacy_cursor_offset + (yorigin * 256))); } - radeon_lock_cursor(crtc, false); + + radeon_crtc->cursor_x = x; + radeon_crtc->cursor_y = y; return 0; } + +int radeon_crtc_cursor_move(struct drm_crtc *crtc, + int x, int y) +{ + int ret; + + radeon_lock_cursor(crtc, true); + ret = radeon_cursor_move_locked(crtc, x, y); + radeon_lock_cursor(crtc, false); + + return ret; +} diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c index f1b0fa1..1021169 100644 --- a/drivers/gpu/drm/radeon/radeon_display.c +++ b/drivers/gpu/drm/radeon/radeon_display.c @@ -635,7 +635,7 @@ radeon_crtc_set_config(struct drm_mode_set *set) return ret; } static const struct drm_crtc_funcs radeon_crtc_funcs = { - .cursor_set = radeon_crtc_cursor_set, + .cursor_set2 = radeon_crtc_cursor_set2, .cursor_move = radeon_crtc_cursor_move, .gamma_set = radeon_crtc_gamma_set, .set_config = radeon_crtc_set_config, diff --git a/drivers/gpu/drm/radeon/radeon_mode.h b/drivers/gpu/drm/radeon/radeon_mode.h index 64eba7e..f3d87cd 100644 --- a/drivers/gpu/drm/radeon/radeon_mode.h +++ b/drivers/gpu/drm/radeon/radeon_mode.h @@ -321,6 +321,10 @@ struct radeon_crtc { uint32_t crtc_offset; struct drm_gem_object *cursor_bo; uint64_t cursor_addr; + int cursor_x; + int cursor_y; + int cursor_hot_x; + int cursor_hot_y; int cursor_width; int cursor_height; int max_cursor_width; @@ -805,11 +809,13 @@ extern int radeon_crtc_set_base_atomic(struct drm_crtc *crtc, extern int radeon_crtc_do_set_base(struct drm_crtc *crtc, struct drm_framebuffer *fb, int x, int y, int atomic); -extern int radeon_crtc_cursor_set(struct drm_crtc *crtc, - struct drm_file *file_priv, - uint32_t handle, - uint32_t width, - uint32_t height); +extern int radeon_crtc_cursor_set2(struct drm_crtc *crtc, + struct drm_file *file_priv, + uint32_t handle, + uint32_t width, + uint32_t height, + int32_t hot_x, + int32_t hot_y); extern int radeon_crtc_cursor_move(struct drm_crtc *crtc, int x, int y); -- cgit v0.10.2 From 3feba08d79c32777a845c3c8a4ab93092bdf6f19 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 18 Nov 2014 18:00:09 +0900 Subject: drm/radeon: Move radeon_cursor_move(_locked) to replace forward declaration MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit No functional change. Signed-off-by: Michel Dänzer Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_cursor.c b/drivers/gpu/drm/radeon/radeon_cursor.c index fd4bddf..85f38ee 100644 --- a/drivers/gpu/drm/radeon/radeon_cursor.c +++ b/drivers/gpu/drm/radeon/radeon_cursor.c @@ -117,7 +117,115 @@ static void radeon_show_cursor(struct drm_crtc *crtc) } } -static int radeon_cursor_move_locked(struct drm_crtc *crtc, int x, int y); +static int radeon_cursor_move_locked(struct drm_crtc *crtc, int x, int y) +{ + struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); + struct radeon_device *rdev = crtc->dev->dev_private; + int xorigin = 0, yorigin = 0; + int w = radeon_crtc->cursor_width; + + if (ASIC_IS_AVIVO(rdev)) { + /* avivo cursor are offset into the total surface */ + x += crtc->x; + y += crtc->y; + } + DRM_DEBUG("x %d y %d c->x %d c->y %d\n", x, y, crtc->x, crtc->y); + + if (x < 0) { + xorigin = min(-x, radeon_crtc->max_cursor_width - 1); + x = 0; + } + if (y < 0) { + yorigin = min(-y, radeon_crtc->max_cursor_height - 1); + y = 0; + } + + /* fixed on DCE6 and newer */ + if (ASIC_IS_AVIVO(rdev) && !ASIC_IS_DCE6(rdev)) { + int i = 0; + struct drm_crtc *crtc_p; + + /* + * avivo cursor image can't end on 128 pixel boundary or + * go past the end of the frame if both crtcs are enabled + * + * NOTE: It is safe to access crtc->enabled of other crtcs + * without holding either the mode_config lock or the other + * crtc's lock as long as write access to this flag _always_ + * grabs all locks. + */ + list_for_each_entry(crtc_p, &crtc->dev->mode_config.crtc_list, head) { + if (crtc_p->enabled) + i++; + } + if (i > 1) { + int cursor_end, frame_end; + + cursor_end = x - xorigin + w; + frame_end = crtc->x + crtc->mode.crtc_hdisplay; + if (cursor_end >= frame_end) { + w = w - (cursor_end - frame_end); + if (!(frame_end & 0x7f)) + w--; + } else { + if (!(cursor_end & 0x7f)) + w--; + } + if (w <= 0) { + w = 1; + cursor_end = x - xorigin + w; + if (!(cursor_end & 0x7f)) { + x--; + WARN_ON_ONCE(x < 0); + } + } + } + } + + if (ASIC_IS_DCE4(rdev)) { + WREG32(EVERGREEN_CUR_POSITION + radeon_crtc->crtc_offset, (x << 16) | y); + WREG32(EVERGREEN_CUR_HOT_SPOT + radeon_crtc->crtc_offset, (xorigin << 16) | yorigin); + WREG32(EVERGREEN_CUR_SIZE + radeon_crtc->crtc_offset, + ((w - 1) << 16) | (radeon_crtc->cursor_height - 1)); + } else if (ASIC_IS_AVIVO(rdev)) { + WREG32(AVIVO_D1CUR_POSITION + radeon_crtc->crtc_offset, (x << 16) | y); + WREG32(AVIVO_D1CUR_HOT_SPOT + radeon_crtc->crtc_offset, (xorigin << 16) | yorigin); + WREG32(AVIVO_D1CUR_SIZE + radeon_crtc->crtc_offset, + ((w - 1) << 16) | (radeon_crtc->cursor_height - 1)); + } else { + if (crtc->mode.flags & DRM_MODE_FLAG_DBLSCAN) + y *= 2; + + WREG32(RADEON_CUR_HORZ_VERT_OFF + radeon_crtc->crtc_offset, + (RADEON_CUR_LOCK + | (xorigin << 16) + | yorigin)); + WREG32(RADEON_CUR_HORZ_VERT_POSN + radeon_crtc->crtc_offset, + (RADEON_CUR_LOCK + | (x << 16) + | y)); + /* offset is from DISP(2)_BASE_ADDRESS */ + WREG32(RADEON_CUR_OFFSET + radeon_crtc->crtc_offset, (radeon_crtc->legacy_cursor_offset + + (yorigin * 256))); + } + + radeon_crtc->cursor_x = x; + radeon_crtc->cursor_y = y; + + return 0; +} + +int radeon_crtc_cursor_move(struct drm_crtc *crtc, + int x, int y) +{ + int ret; + + radeon_lock_cursor(crtc, true); + ret = radeon_cursor_move_locked(crtc, x, y); + radeon_lock_cursor(crtc, false); + + return ret; +} static void radeon_set_cursor(struct drm_crtc *crtc, struct drm_gem_object *obj, uint64_t gpu_addr, int hot_x, int hot_y) @@ -231,113 +339,3 @@ fail: return ret; } - -static int radeon_cursor_move_locked(struct drm_crtc *crtc, int x, int y) -{ - struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); - struct radeon_device *rdev = crtc->dev->dev_private; - int xorigin = 0, yorigin = 0; - int w = radeon_crtc->cursor_width; - - if (ASIC_IS_AVIVO(rdev)) { - /* avivo cursor are offset into the total surface */ - x += crtc->x; - y += crtc->y; - } - DRM_DEBUG("x %d y %d c->x %d c->y %d\n", x, y, crtc->x, crtc->y); - - if (x < 0) { - xorigin = min(-x, radeon_crtc->max_cursor_width - 1); - x = 0; - } - if (y < 0) { - yorigin = min(-y, radeon_crtc->max_cursor_height - 1); - y = 0; - } - - /* fixed on DCE6 and newer */ - if (ASIC_IS_AVIVO(rdev) && !ASIC_IS_DCE6(rdev)) { - int i = 0; - struct drm_crtc *crtc_p; - - /* - * avivo cursor image can't end on 128 pixel boundary or - * go past the end of the frame if both crtcs are enabled - * - * NOTE: It is safe to access crtc->enabled of other crtcs - * without holding either the mode_config lock or the other - * crtc's lock as long as write access to this flag _always_ - * grabs all locks. - */ - list_for_each_entry(crtc_p, &crtc->dev->mode_config.crtc_list, head) { - if (crtc_p->enabled) - i++; - } - if (i > 1) { - int cursor_end, frame_end; - - cursor_end = x - xorigin + w; - frame_end = crtc->x + crtc->mode.crtc_hdisplay; - if (cursor_end >= frame_end) { - w = w - (cursor_end - frame_end); - if (!(frame_end & 0x7f)) - w--; - } else { - if (!(cursor_end & 0x7f)) - w--; - } - if (w <= 0) { - w = 1; - cursor_end = x - xorigin + w; - if (!(cursor_end & 0x7f)) { - x--; - WARN_ON_ONCE(x < 0); - } - } - } - } - - if (ASIC_IS_DCE4(rdev)) { - WREG32(EVERGREEN_CUR_POSITION + radeon_crtc->crtc_offset, (x << 16) | y); - WREG32(EVERGREEN_CUR_HOT_SPOT + radeon_crtc->crtc_offset, (xorigin << 16) | yorigin); - WREG32(EVERGREEN_CUR_SIZE + radeon_crtc->crtc_offset, - ((w - 1) << 16) | (radeon_crtc->cursor_height - 1)); - } else if (ASIC_IS_AVIVO(rdev)) { - WREG32(AVIVO_D1CUR_POSITION + radeon_crtc->crtc_offset, (x << 16) | y); - WREG32(AVIVO_D1CUR_HOT_SPOT + radeon_crtc->crtc_offset, (xorigin << 16) | yorigin); - WREG32(AVIVO_D1CUR_SIZE + radeon_crtc->crtc_offset, - ((w - 1) << 16) | (radeon_crtc->cursor_height - 1)); - } else { - if (crtc->mode.flags & DRM_MODE_FLAG_DBLSCAN) - y *= 2; - - WREG32(RADEON_CUR_HORZ_VERT_OFF + radeon_crtc->crtc_offset, - (RADEON_CUR_LOCK - | (xorigin << 16) - | yorigin)); - WREG32(RADEON_CUR_HORZ_VERT_POSN + radeon_crtc->crtc_offset, - (RADEON_CUR_LOCK - | (x << 16) - | y)); - /* offset is from DISP(2)_BASE_ADDRESS */ - WREG32(RADEON_CUR_OFFSET + radeon_crtc->crtc_offset, (radeon_crtc->legacy_cursor_offset + - (yorigin * 256))); - } - - radeon_crtc->cursor_x = x; - radeon_crtc->cursor_y = y; - - return 0; -} - -int radeon_crtc_cursor_move(struct drm_crtc *crtc, - int x, int y) -{ - int ret; - - radeon_lock_cursor(crtc, true); - ret = radeon_cursor_move_locked(crtc, x, y); - radeon_lock_cursor(crtc, false); - - return ret; -} -- cgit v0.10.2 From 4bb62c95a7e781a238b2ab374f34b1bf91e01ddc Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 17 Nov 2014 15:08:17 -0500 Subject: drm/radeon: work around a hw bug in MGCG on CIK Always need to set bit 0 of RLC_CGTT_MGCG_OVERRIDE to avoid unreliable doorbell updates in some cases. Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index bce73b6..cef4cb7 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -6344,6 +6344,7 @@ static void cik_enable_mgcg(struct radeon_device *rdev, bool enable) } orig = data = RREG32(RLC_CGTT_MGCG_OVERRIDE); + data |= 0x00000001; data &= 0xfffffffd; if (orig != data) WREG32(RLC_CGTT_MGCG_OVERRIDE, data); @@ -6377,7 +6378,7 @@ static void cik_enable_mgcg(struct radeon_device *rdev, bool enable) } } else { orig = data = RREG32(RLC_CGTT_MGCG_OVERRIDE); - data |= 0x00000002; + data |= 0x00000003; if (orig != data) WREG32(RLC_CGTT_MGCG_OVERRIDE, data); -- cgit v0.10.2 From 39471ad39de827657e6ab69da96496eb0943295e Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Sun, 14 Sep 2014 21:14:14 -0400 Subject: drm/radeon/dpm: add smc fan control for SI (v2) Enable smc fan control for SI boards. Should reduce the fan noise on systems with a higher default fan profile. v2: disable by default, add rpm controls bug: https://bugs.freedesktop.org/show_bug.cgi?id=73338 Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ppsmc.h b/drivers/gpu/drm/radeon/ppsmc.h index 11c0e4d..0c4eaa6 100644 --- a/drivers/gpu/drm/radeon/ppsmc.h +++ b/drivers/gpu/drm/radeon/ppsmc.h @@ -56,6 +56,9 @@ #define PPSMC_STATEFLAG_DEEPSLEEP_THROTTLE 0x20 #define PPSMC_STATEFLAG_DEEPSLEEP_BYPASS 0x40 +#define FDO_MODE_HARDWARE 0 +#define FDO_MODE_PIECE_WISE_LINEAR 1 + #define PPSMC_Result_OK ((uint8_t)0x01) #define PPSMC_Result_Failed ((uint8_t)0xFF) @@ -79,6 +82,8 @@ typedef uint8_t PPSMC_Result; #define PPSMC_MSG_DisableCac ((uint8_t)0x54) #define PPSMC_TDPClampingActive ((uint8_t)0x59) #define PPSMC_TDPClampingInactive ((uint8_t)0x5A) +#define PPSMC_StartFanControl ((uint8_t)0x5B) +#define PPSMC_StopFanControl ((uint8_t)0x5C) #define PPSMC_MSG_NoDisplay ((uint8_t)0x5D) #define PPSMC_MSG_HasDisplay ((uint8_t)0x5E) #define PPSMC_MSG_UVDPowerOFF ((uint8_t)0x60) diff --git a/drivers/gpu/drm/radeon/r600_dpm.h b/drivers/gpu/drm/radeon/r600_dpm.h index 46b9d2a..bd499d7 100644 --- a/drivers/gpu/drm/radeon/r600_dpm.h +++ b/drivers/gpu/drm/radeon/r600_dpm.h @@ -96,6 +96,9 @@ #define R600_TEMP_RANGE_MIN (90 * 1000) #define R600_TEMP_RANGE_MAX (120 * 1000) +#define FDO_PWM_MODE_STATIC 1 +#define FDO_PWM_MODE_STATIC_RPM 5 + enum r600_power_level { R600_POWER_LEVEL_LOW = 0, R600_POWER_LEVEL_MEDIUM = 1, diff --git a/drivers/gpu/drm/radeon/si_dpm.c b/drivers/gpu/drm/radeon/si_dpm.c index b59e1d6..cf4c420 100644 --- a/drivers/gpu/drm/radeon/si_dpm.c +++ b/drivers/gpu/drm/radeon/si_dpm.c @@ -3398,6 +3398,15 @@ static int si_process_firmware_header(struct radeon_device *rdev) ret = si_read_smc_sram_dword(rdev, SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_fanTable, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->fan_table_start = tmp; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + SISLANDS_SMC_FIRMWARE_HEADER_mcArbDramAutoRefreshTable, &tmp, si_pi->sram_end); if (ret) @@ -5825,20 +5834,20 @@ static int si_thermal_enable_alert(struct radeon_device *rdev, if (enable) { PPSMC_Result result; - thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW; - rdev->irq.dpm_thermal = true; + thermal_int &= ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); + WREG32(CG_THERMAL_INT, thermal_int); + rdev->irq.dpm_thermal = false; result = si_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); if (result != PPSMC_Result_OK) { DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); return -EINVAL; } } else { - thermal_int &= ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); - rdev->irq.dpm_thermal = false; + thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW; + WREG32(CG_THERMAL_INT, thermal_int); + rdev->irq.dpm_thermal = true; } - WREG32(CG_THERMAL_INT, thermal_int); - return 0; } @@ -5867,6 +5876,309 @@ static int si_thermal_set_temperature_range(struct radeon_device *rdev, return 0; } +static void si_fan_ctrl_set_static_mode(struct radeon_device *rdev, u32 mode) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 tmp; + + if (si_pi->fan_ctrl_is_in_default_mode) { + tmp = (RREG32(CG_FDO_CTRL2) & FDO_PWM_MODE_MASK) >> FDO_PWM_MODE_SHIFT; + si_pi->fan_ctrl_default_mode = tmp; + tmp = (RREG32(CG_FDO_CTRL2) & TMIN_MASK) >> TMIN_SHIFT; + si_pi->t_min = tmp; + si_pi->fan_ctrl_is_in_default_mode = false; + } + + tmp = RREG32(CG_FDO_CTRL2) & ~TMIN_MASK; + tmp |= TMIN(0); + WREG32(CG_FDO_CTRL2, tmp); + + tmp = RREG32(CG_FDO_CTRL2) & FDO_PWM_MODE_MASK; + tmp |= FDO_PWM_MODE(mode); + WREG32(CG_FDO_CTRL2, tmp); +} + +static int si_thermal_setup_fan_table(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + PP_SIslands_FanTable fan_table = { FDO_MODE_HARDWARE }; + u32 duty100; + u32 t_diff1, t_diff2, pwm_diff1, pwm_diff2; + u16 fdo_min, slope1, slope2; + u32 reference_clock, tmp; + int ret; + u64 tmp64; + + if (!si_pi->fan_table_start) { + rdev->pm.dpm.fan.ucode_fan_control = false; + return 0; + } + + duty100 = (RREG32(CG_FDO_CTRL1) & FMAX_DUTY100_MASK) >> FMAX_DUTY100_SHIFT; + + if (duty100 == 0) { + rdev->pm.dpm.fan.ucode_fan_control = false; + return 0; + } + + tmp64 = (u64)rdev->pm.dpm.fan.pwm_min * duty100; + do_div(tmp64, 10000); + fdo_min = (u16)tmp64; + + t_diff1 = rdev->pm.dpm.fan.t_med - rdev->pm.dpm.fan.t_min; + t_diff2 = rdev->pm.dpm.fan.t_high - rdev->pm.dpm.fan.t_med; + + pwm_diff1 = rdev->pm.dpm.fan.pwm_med - rdev->pm.dpm.fan.pwm_min; + pwm_diff2 = rdev->pm.dpm.fan.pwm_high - rdev->pm.dpm.fan.pwm_med; + + slope1 = (u16)((50 + ((16 * duty100 * pwm_diff1) / t_diff1)) / 100); + slope2 = (u16)((50 + ((16 * duty100 * pwm_diff2) / t_diff2)) / 100); + + fan_table.slope1 = cpu_to_be16(slope1); + fan_table.slope2 = cpu_to_be16(slope2); + + fan_table.fdo_min = cpu_to_be16(fdo_min); + + fan_table.hys_down = cpu_to_be16(rdev->pm.dpm.fan.t_hyst); + + fan_table.hys_up = cpu_to_be16(1); + + fan_table.hys_slope = cpu_to_be16(1); + + fan_table.temp_resp_lim = cpu_to_be16(5); + + reference_clock = radeon_get_xclk(rdev); + + fan_table.refresh_period = cpu_to_be32((rdev->pm.dpm.fan.cycle_delay * + reference_clock) / 1600); + + fan_table.fdo_max = cpu_to_be16((u16)duty100); + + tmp = (RREG32(CG_MULT_THERMAL_CTRL) & TEMP_SEL_MASK) >> TEMP_SEL_SHIFT; + fan_table.temp_src = (uint8_t)tmp; + + ret = si_copy_bytes_to_smc(rdev, + si_pi->fan_table_start, + (u8 *)(&fan_table), + sizeof(fan_table), + si_pi->sram_end); + + if (ret) { + DRM_ERROR("Failed to load fan table to the SMC."); + rdev->pm.dpm.fan.ucode_fan_control = false; + } + + return 0; +} + +static int si_fan_ctrl_start_smc_fan_control(struct radeon_device *rdev) +{ + PPSMC_Result ret; + + ret = si_send_msg_to_smc(rdev, PPSMC_StartFanControl); + if (ret == PPSMC_Result_OK) + return 0; + else + return -EINVAL; +} + +static int si_fan_ctrl_stop_smc_fan_control(struct radeon_device *rdev) +{ + PPSMC_Result ret; + + ret = si_send_msg_to_smc(rdev, PPSMC_StopFanControl); + if (ret == PPSMC_Result_OK) + return 0; + else + return -EINVAL; +} + +#if 0 +static int si_fan_ctrl_get_fan_speed_percent(struct radeon_device *rdev, + u32 *speed) +{ + u32 duty, duty100; + u64 tmp64; + + if (rdev->pm.no_fan) + return -ENOENT; + + duty100 = (RREG32(CG_FDO_CTRL1) & FMAX_DUTY100_MASK) >> FMAX_DUTY100_SHIFT; + duty = (RREG32(CG_THERMAL_STATUS) & FDO_PWM_DUTY_MASK) >> FDO_PWM_DUTY_SHIFT; + + if (duty100 == 0) + return -EINVAL; + + tmp64 = (u64)duty * 100; + do_div(tmp64, duty100); + *speed = (u32)tmp64; + + if (*speed > 100) + *speed = 100; + + return 0; +} + +static int si_fan_ctrl_set_fan_speed_percent(struct radeon_device *rdev, + u32 speed) +{ + u32 tmp; + u32 duty, duty100; + u64 tmp64; + + if (rdev->pm.no_fan) + return -ENOENT; + + if (speed > 100) + return -EINVAL; + + if (rdev->pm.dpm.fan.ucode_fan_control) + si_fan_ctrl_stop_smc_fan_control(rdev); + + duty100 = (RREG32(CG_FDO_CTRL1) & FMAX_DUTY100_MASK) >> FMAX_DUTY100_SHIFT; + + if (duty100 == 0) + return -EINVAL; + + tmp64 = (u64)speed * duty100; + do_div(tmp64, 100); + duty = (u32)tmp64; + + tmp = RREG32(CG_FDO_CTRL0) & ~FDO_STATIC_DUTY_MASK; + tmp |= FDO_STATIC_DUTY(duty); + WREG32(CG_FDO_CTRL0, tmp); + + si_fan_ctrl_set_static_mode(rdev, FDO_PWM_MODE_STATIC); + + return 0; +} + +static int si_fan_ctrl_get_fan_speed_rpm(struct radeon_device *rdev, + u32 *speed) +{ + u32 tach_period; + u32 xclk = radeon_get_xclk(rdev); + + if (rdev->pm.no_fan) + return -ENOENT; + + if (rdev->pm.fan_pulses_per_revolution == 0) + return -ENOENT; + + tach_period = (RREG32(CG_TACH_STATUS) & TACH_PERIOD_MASK) >> TACH_PERIOD_SHIFT; + if (tach_period == 0) + return -ENOENT; + + *speed = 60 * xclk * 10000 / tach_period; + + return 0; +} + +static int si_fan_ctrl_set_fan_speed_rpm(struct radeon_device *rdev, + u32 speed) +{ + u32 tach_period, tmp; + u32 xclk = radeon_get_xclk(rdev); + + if (rdev->pm.no_fan) + return -ENOENT; + + if (rdev->pm.fan_pulses_per_revolution == 0) + return -ENOENT; + + if ((speed < rdev->pm.fan_min_rpm) || + (speed > rdev->pm.fan_max_rpm)) + return -EINVAL; + + if (rdev->pm.dpm.fan.ucode_fan_control) + si_fan_ctrl_stop_smc_fan_control(rdev); + + tach_period = 60 * xclk * 10000 / (8 * speed); + tmp = RREG32(CG_TACH_CTRL) & ~TARGET_PERIOD_MASK; + tmp |= TARGET_PERIOD(tach_period); + WREG32(CG_TACH_CTRL, tmp); + + si_fan_ctrl_set_static_mode(rdev, FDO_PWM_MODE_STATIC); + + return 0; +} +#endif + +static void si_fan_ctrl_set_default_mode(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 tmp; + + if (!si_pi->fan_ctrl_is_in_default_mode) { + tmp = RREG32(CG_FDO_CTRL2) & ~FDO_PWM_MODE_MASK; + tmp |= FDO_PWM_MODE(si_pi->fan_ctrl_default_mode); + WREG32(CG_FDO_CTRL2, tmp); + + tmp = RREG32(CG_FDO_CTRL2) & TMIN_MASK; + tmp |= TMIN(si_pi->t_min); + WREG32(CG_FDO_CTRL2, tmp); + si_pi->fan_ctrl_is_in_default_mode = true; + } +} + +static void si_thermal_start_smc_fan_control(struct radeon_device *rdev) +{ + if (rdev->pm.dpm.fan.ucode_fan_control) { + si_fan_ctrl_start_smc_fan_control(rdev); + si_fan_ctrl_set_static_mode(rdev, FDO_PWM_MODE_STATIC); + } +} + +static void si_thermal_initialize(struct radeon_device *rdev) +{ + u32 tmp; + + if (rdev->pm.fan_pulses_per_revolution) { + tmp = RREG32(CG_TACH_CTRL) & ~EDGE_PER_REV_MASK; + tmp |= EDGE_PER_REV(rdev->pm.fan_pulses_per_revolution -1); + WREG32(CG_TACH_CTRL, tmp); + } + + tmp = RREG32(CG_FDO_CTRL2) & ~TACH_PWM_RESP_RATE_MASK; + tmp |= TACH_PWM_RESP_RATE(0x28); + WREG32(CG_FDO_CTRL2, tmp); +} + +static int si_thermal_start_thermal_controller(struct radeon_device *rdev) +{ + int ret; + + si_thermal_initialize(rdev); + ret = si_thermal_set_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + ret = si_thermal_enable_alert(rdev, true); + if (ret) + return ret; + if (rdev->pm.dpm.fan.ucode_fan_control) { + ret = si_halt_smc(rdev); + if (ret) + return ret; + ret = si_thermal_setup_fan_table(rdev); + if (ret) + return ret; + ret = si_resume_smc(rdev); + if (ret) + return ret; + si_thermal_start_smc_fan_control(rdev); + } + + return 0; +} + +static void si_thermal_stop_thermal_controller(struct radeon_device *rdev) +{ + if (!rdev->pm.no_fan) { + si_fan_ctrl_set_default_mode(rdev); + si_fan_ctrl_stop_smc_fan_control(rdev); + } +} + int si_dpm_enable(struct radeon_device *rdev) { struct rv7xx_power_info *pi = rv770_get_pi(rdev); @@ -5979,6 +6291,8 @@ int si_dpm_enable(struct radeon_device *rdev) si_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + si_thermal_start_thermal_controller(rdev); + ni_update_current_ps(rdev, boot_ps); return 0; @@ -6019,6 +6333,7 @@ void si_dpm_disable(struct radeon_device *rdev) if (!si_is_smc_running(rdev)) return; + si_thermal_stop_thermal_controller(rdev); si_disable_ulv(rdev); si_clear_vc(rdev); if (pi->thermal_protection) @@ -6557,6 +6872,9 @@ int si_dpm_init(struct radeon_device *rdev) rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc = rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac; + si_pi->fan_ctrl_is_in_default_mode = true; + rdev->pm.dpm.fan.ucode_fan_control = false; + return 0; } diff --git a/drivers/gpu/drm/radeon/si_dpm.h b/drivers/gpu/drm/radeon/si_dpm.h index 8b5c06a..d16bb1b 100644 --- a/drivers/gpu/drm/radeon/si_dpm.h +++ b/drivers/gpu/drm/radeon/si_dpm.h @@ -182,6 +182,7 @@ struct si_power_info { u32 dte_table_start; u32 spll_table_start; u32 papm_cfg_table_start; + u32 fan_table_start; /* CAC stuff */ const struct si_cac_config_reg *cac_weights; const struct si_cac_config_reg *lcac_config; @@ -197,6 +198,10 @@ struct si_power_info { /* SVI2 */ u8 svd_gpio_id; u8 svc_gpio_id; + /* fan control */ + bool fan_ctrl_is_in_default_mode; + u32 t_min; + u32 fan_ctrl_default_mode; }; #define SISLANDS_INITIAL_STATE_ARB_INDEX 0 diff --git a/drivers/gpu/drm/radeon/sid.h b/drivers/gpu/drm/radeon/sid.h index 6635da9..c549c16a 100644 --- a/drivers/gpu/drm/radeon/sid.h +++ b/drivers/gpu/drm/radeon/sid.h @@ -180,7 +180,10 @@ #define DIG_THERM_DPM(x) ((x) << 14) #define DIG_THERM_DPM_MASK 0x003FC000 #define DIG_THERM_DPM_SHIFT 14 - +#define CG_THERMAL_STATUS 0x704 +#define FDO_PWM_DUTY(x) ((x) << 9) +#define FDO_PWM_DUTY_MASK (0xff << 9) +#define FDO_PWM_DUTY_SHIFT 9 #define CG_THERMAL_INT 0x708 #define DIG_THERM_INTH(x) ((x) << 8) #define DIG_THERM_INTH_MASK 0x0000FF00 @@ -191,6 +194,10 @@ #define THERM_INT_MASK_HIGH (1 << 24) #define THERM_INT_MASK_LOW (1 << 25) +#define CG_MULT_THERMAL_CTRL 0x710 +#define TEMP_SEL(x) ((x) << 20) +#define TEMP_SEL_MASK (0xff << 20) +#define TEMP_SEL_SHIFT 20 #define CG_MULT_THERMAL_STATUS 0x714 #define ASIC_MAX_TEMP(x) ((x) << 0) #define ASIC_MAX_TEMP_MASK 0x000001ff @@ -199,6 +206,37 @@ #define CTF_TEMP_MASK 0x0003fe00 #define CTF_TEMP_SHIFT 9 +#define CG_FDO_CTRL0 0x754 +#define FDO_STATIC_DUTY(x) ((x) << 0) +#define FDO_STATIC_DUTY_MASK 0x0000000F +#define FDO_STATIC_DUTY_SHIFT 0 +#define CG_FDO_CTRL1 0x758 +#define FMAX_DUTY100(x) ((x) << 0) +#define FMAX_DUTY100_MASK 0x0000000F +#define FMAX_DUTY100_SHIFT 0 +#define CG_FDO_CTRL2 0x75C +#define TMIN(x) ((x) << 0) +#define TMIN_MASK 0x0000000F +#define TMIN_SHIFT 0 +#define FDO_PWM_MODE(x) ((x) << 11) +#define FDO_PWM_MODE_MASK (3 << 11) +#define FDO_PWM_MODE_SHIFT 11 +#define TACH_PWM_RESP_RATE(x) ((x) << 25) +#define TACH_PWM_RESP_RATE_MASK (0x7f << 25) +#define TACH_PWM_RESP_RATE_SHIFT 25 + +#define CG_TACH_CTRL 0x770 +# define EDGE_PER_REV(x) ((x) << 0) +# define EDGE_PER_REV_MASK (0x7 << 0) +# define EDGE_PER_REV_SHIFT 0 +# define TARGET_PERIOD(x) ((x) << 3) +# define TARGET_PERIOD_MASK 0xfffffff8 +# define TARGET_PERIOD_SHIFT 3 +#define CG_TACH_STATUS 0x774 +# define TACH_PERIOD(x) ((x) << 0) +# define TACH_PERIOD_MASK 0xffffffff +# define TACH_PERIOD_SHIFT 0 + #define GENERAL_PWRMGT 0x780 # define GLOBAL_PWRMGT_EN (1 << 0) # define STATIC_PM_EN (1 << 1) diff --git a/drivers/gpu/drm/radeon/sislands_smc.h b/drivers/gpu/drm/radeon/sislands_smc.h index 623a0b1..3c77983 100644 --- a/drivers/gpu/drm/radeon/sislands_smc.h +++ b/drivers/gpu/drm/radeon/sislands_smc.h @@ -245,6 +245,31 @@ typedef struct SISLANDS_SMC_STATETABLE SISLANDS_SMC_STATETABLE; #define SI_SMC_SOFT_REGISTER_svi_rework_gpio_id_svd 0x11c #define SI_SMC_SOFT_REGISTER_svi_rework_gpio_id_svc 0x120 +struct PP_SIslands_FanTable +{ + uint8_t fdo_mode; + uint8_t padding; + int16_t temp_min; + int16_t temp_med; + int16_t temp_max; + int16_t slope1; + int16_t slope2; + int16_t fdo_min; + int16_t hys_up; + int16_t hys_down; + int16_t hys_slope; + int16_t temp_resp_lim; + int16_t temp_curr; + int16_t slope_curr; + int16_t pwm_curr; + uint32_t refresh_period; + int16_t fdo_max; + uint8_t temp_src; + int8_t padding2; +}; + +typedef struct PP_SIslands_FanTable PP_SIslands_FanTable; + #define SMC_SISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES 16 #define SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES 32 -- cgit v0.10.2 From e03cea367f992e683435c41344197cd7b223b62c Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 15 Sep 2014 00:15:22 -0400 Subject: drm/radeon/dpm: add smc fan control for CI (v2) Enable smc fan control for CI boards. Should reduce the fan noise on systems with a higher default fan profile. v2: disable by default, add additional fan setup, rpm control bug: https://bugs.freedesktop.org/show_bug.cgi?id=73338 Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 9dbc52f..4581d6c 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -184,6 +184,9 @@ static int ci_set_overdrive_target_tdp(struct radeon_device *rdev, u32 target_tdp); static int ci_update_uvd_dpm(struct radeon_device *rdev, bool gate); +static PPSMC_Result ci_send_msg_to_smc_with_parameter(struct radeon_device *rdev, + PPSMC_Msg msg, u32 parameter); + static struct ci_power_info *ci_get_pi(struct radeon_device *rdev) { struct ci_power_info *pi = rdev->pm.dpm.priv; @@ -355,6 +358,21 @@ static int ci_populate_dw8(struct radeon_device *rdev) return 0; } +static int ci_populate_fuzzy_fan(struct radeon_device *rdev) +{ + struct ci_power_info *pi = ci_get_pi(rdev); + + if ((rdev->pm.dpm.fan.fan_output_sensitivity & (1 << 15)) || + (rdev->pm.dpm.fan.fan_output_sensitivity == 0)) + rdev->pm.dpm.fan.fan_output_sensitivity = + rdev->pm.dpm.fan.default_fan_output_sensitivity; + + pi->smc_powertune_table.FuzzyFan_PwmSetDelta = + cpu_to_be16(rdev->pm.dpm.fan.fan_output_sensitivity); + + return 0; +} + static int ci_min_max_v_gnbl_pm_lid_from_bapm_vddc(struct radeon_device *rdev) { struct ci_power_info *pi = ci_get_pi(rdev); @@ -480,6 +498,9 @@ static int ci_populate_pm_base(struct radeon_device *rdev) ret = ci_populate_dw8(rdev); if (ret) return ret; + ret = ci_populate_fuzzy_fan(rdev); + if (ret) + return ret; ret = ci_min_max_v_gnbl_pm_lid_from_bapm_vddc(rdev); if (ret) return ret; @@ -859,6 +880,7 @@ static int ci_thermal_enable_alert(struct radeon_device *rdev, if (enable) { thermal_int &= ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); + WREG32_SMC(CG_THERMAL_INT, thermal_int); rdev->irq.dpm_thermal = false; result = ci_send_msg_to_smc(rdev, PPSMC_MSG_Thermal_Cntl_Enable); if (result != PPSMC_Result_OK) { @@ -867,6 +889,7 @@ static int ci_thermal_enable_alert(struct radeon_device *rdev, } } else { thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW; + WREG32_SMC(CG_THERMAL_INT, thermal_int); rdev->irq.dpm_thermal = true; result = ci_send_msg_to_smc(rdev, PPSMC_MSG_Thermal_Cntl_Disable); if (result != PPSMC_Result_OK) { @@ -875,12 +898,325 @@ static int ci_thermal_enable_alert(struct radeon_device *rdev, } } - WREG32_SMC(CG_THERMAL_INT, thermal_int); + return 0; +} + +static void ci_fan_ctrl_set_static_mode(struct radeon_device *rdev, u32 mode) +{ + struct ci_power_info *pi = ci_get_pi(rdev); + u32 tmp; + + if (pi->fan_ctrl_is_in_default_mode) { + tmp = (RREG32_SMC(CG_FDO_CTRL2) & FDO_PWM_MODE_MASK) >> FDO_PWM_MODE_SHIFT; + pi->fan_ctrl_default_mode = tmp; + tmp = (RREG32_SMC(CG_FDO_CTRL2) & TMIN_MASK) >> TMIN_SHIFT; + pi->t_min = tmp; + pi->fan_ctrl_is_in_default_mode = false; + } + + tmp = RREG32_SMC(CG_FDO_CTRL2) & ~TMIN_MASK; + tmp |= TMIN(0); + WREG32_SMC(CG_FDO_CTRL2, tmp); + + tmp = RREG32_SMC(CG_FDO_CTRL2) & FDO_PWM_MODE_MASK; + tmp |= FDO_PWM_MODE(mode); + WREG32_SMC(CG_FDO_CTRL2, tmp); +} + +static int ci_thermal_setup_fan_table(struct radeon_device *rdev) +{ + struct ci_power_info *pi = ci_get_pi(rdev); + SMU7_Discrete_FanTable fan_table = { FDO_MODE_HARDWARE }; + u32 duty100; + u32 t_diff1, t_diff2, pwm_diff1, pwm_diff2; + u16 fdo_min, slope1, slope2; + u32 reference_clock, tmp; + int ret; + u64 tmp64; + + if (!pi->fan_table_start) { + rdev->pm.dpm.fan.ucode_fan_control = false; + return 0; + } + + duty100 = (RREG32_SMC(CG_FDO_CTRL1) & FMAX_DUTY100_MASK) >> FMAX_DUTY100_SHIFT; + + if (duty100 == 0) { + rdev->pm.dpm.fan.ucode_fan_control = false; + return 0; + } + + tmp64 = (u64)rdev->pm.dpm.fan.pwm_min * duty100; + do_div(tmp64, 10000); + fdo_min = (u16)tmp64; + + t_diff1 = rdev->pm.dpm.fan.t_med - rdev->pm.dpm.fan.t_min; + t_diff2 = rdev->pm.dpm.fan.t_high - rdev->pm.dpm.fan.t_med; + + pwm_diff1 = rdev->pm.dpm.fan.pwm_med - rdev->pm.dpm.fan.pwm_min; + pwm_diff2 = rdev->pm.dpm.fan.pwm_high - rdev->pm.dpm.fan.pwm_med; + + slope1 = (u16)((50 + ((16 * duty100 * pwm_diff1) / t_diff1)) / 100); + slope2 = (u16)((50 + ((16 * duty100 * pwm_diff2) / t_diff2)) / 100); + + fan_table.TempMin = cpu_to_be16((50 + rdev->pm.dpm.fan.t_min) / 100); + fan_table.TempMed = cpu_to_be16((50 + rdev->pm.dpm.fan.t_med) / 100); + fan_table.TempMax = cpu_to_be16((50 + rdev->pm.dpm.fan.t_max) / 100); + + fan_table.Slope1 = cpu_to_be16(slope1); + fan_table.Slope2 = cpu_to_be16(slope2); + + fan_table.FdoMin = cpu_to_be16(fdo_min); + + fan_table.HystDown = cpu_to_be16(rdev->pm.dpm.fan.t_hyst); + + fan_table.HystUp = cpu_to_be16(1); + + fan_table.HystSlope = cpu_to_be16(1); + + fan_table.TempRespLim = cpu_to_be16(5); + + reference_clock = radeon_get_xclk(rdev); + + fan_table.RefreshPeriod = cpu_to_be32((rdev->pm.dpm.fan.cycle_delay * + reference_clock) / 1600); + + fan_table.FdoMax = cpu_to_be16((u16)duty100); + + tmp = (RREG32_SMC(CG_MULT_THERMAL_CTRL) & TEMP_SEL_MASK) >> TEMP_SEL_SHIFT; + fan_table.TempSrc = (uint8_t)tmp; + + ret = ci_copy_bytes_to_smc(rdev, + pi->fan_table_start, + (u8 *)(&fan_table), + sizeof(fan_table), + pi->sram_end); + + if (ret) { + DRM_ERROR("Failed to load fan table to the SMC."); + rdev->pm.dpm.fan.ucode_fan_control = false; + } + + return 0; +} + +static int ci_fan_ctrl_start_smc_fan_control(struct radeon_device *rdev) +{ + struct ci_power_info *pi = ci_get_pi(rdev); + PPSMC_Result ret; + + if (pi->caps_od_fuzzy_fan_control_support) { + ret = ci_send_msg_to_smc_with_parameter(rdev, + PPSMC_StartFanControl, + FAN_CONTROL_FUZZY); + if (ret != PPSMC_Result_OK) + return -EINVAL; + ret = ci_send_msg_to_smc_with_parameter(rdev, + PPSMC_MSG_SetFanPwmMax, + rdev->pm.dpm.fan.default_max_fan_pwm); + if (ret != PPSMC_Result_OK) + return -EINVAL; + } else { + ret = ci_send_msg_to_smc_with_parameter(rdev, + PPSMC_StartFanControl, + FAN_CONTROL_TABLE); + if (ret != PPSMC_Result_OK) + return -EINVAL; + } return 0; } #if 0 +static int ci_fan_ctrl_stop_smc_fan_control(struct radeon_device *rdev) +{ + PPSMC_Result ret; + + ret = ci_send_msg_to_smc(rdev, PPSMC_StopFanControl); + if (ret == PPSMC_Result_OK) + return 0; + else + return -EINVAL; +} + +static int ci_fan_ctrl_get_fan_speed_percent(struct radeon_device *rdev, + u32 *speed) +{ + u32 duty, duty100; + u64 tmp64; + + if (rdev->pm.no_fan) + return -ENOENT; + + duty100 = (RREG32_SMC(CG_FDO_CTRL1) & FMAX_DUTY100_MASK) >> FMAX_DUTY100_SHIFT; + duty = (RREG32_SMC(CG_THERMAL_STATUS) & FDO_PWM_DUTY_MASK) >> FDO_PWM_DUTY_SHIFT; + + if (duty100 == 0) + return -EINVAL; + + tmp64 = (u64)duty * 100; + do_div(tmp64, duty100); + *speed = (u32)tmp64; + + if (*speed > 100) + *speed = 100; + + return 0; +} + +static int ci_fan_ctrl_set_fan_speed_percent(struct radeon_device *rdev, + u32 speed) +{ + u32 tmp; + u32 duty, duty100; + u64 tmp64; + + if (rdev->pm.no_fan) + return -ENOENT; + + if (speed > 100) + return -EINVAL; + + if (rdev->pm.dpm.fan.ucode_fan_control) + ci_fan_ctrl_stop_smc_fan_control(rdev); + + duty100 = (RREG32_SMC(CG_FDO_CTRL1) & FMAX_DUTY100_MASK) >> FMAX_DUTY100_SHIFT; + + if (duty100 == 0) + return -EINVAL; + + tmp64 = (u64)speed * duty100; + do_div(tmp64, 100); + duty = (u32)tmp64; + + tmp = RREG32_SMC(CG_FDO_CTRL0) & ~FDO_STATIC_DUTY_MASK; + tmp |= FDO_STATIC_DUTY(duty); + WREG32_SMC(CG_FDO_CTRL0, tmp); + + ci_fan_ctrl_set_static_mode(rdev, FDO_PWM_MODE_STATIC); + + return 0; +} + +static int ci_fan_ctrl_get_fan_speed_rpm(struct radeon_device *rdev, + u32 *speed) +{ + u32 tach_period; + u32 xclk = radeon_get_xclk(rdev); + + if (rdev->pm.no_fan) + return -ENOENT; + + if (rdev->pm.fan_pulses_per_revolution == 0) + return -ENOENT; + + tach_period = (RREG32_SMC(CG_TACH_STATUS) & TACH_PERIOD_MASK) >> TACH_PERIOD_SHIFT; + if (tach_period == 0) + return -ENOENT; + + *speed = 60 * xclk * 10000 / tach_period; + + return 0; +} + +static int ci_fan_ctrl_set_fan_speed_rpm(struct radeon_device *rdev, + u32 speed) +{ + u32 tach_period, tmp; + u32 xclk = radeon_get_xclk(rdev); + + if (rdev->pm.no_fan) + return -ENOENT; + + if (rdev->pm.fan_pulses_per_revolution == 0) + return -ENOENT; + + if ((speed < rdev->pm.fan_min_rpm) || + (speed > rdev->pm.fan_max_rpm)) + return -EINVAL; + + if (rdev->pm.dpm.fan.ucode_fan_control) + ci_fan_ctrl_stop_smc_fan_control(rdev); + + tach_period = 60 * xclk * 10000 / (8 * speed); + tmp = RREG32_SMC(CG_TACH_CTRL) & ~TARGET_PERIOD_MASK; + tmp |= TARGET_PERIOD(tach_period); + WREG32_SMC(CG_TACH_CTRL, tmp); + + ci_fan_ctrl_set_static_mode(rdev, FDO_PWM_MODE_STATIC); + + return 0; +} +#endif + +static void ci_fan_ctrl_set_default_mode(struct radeon_device *rdev) +{ + struct ci_power_info *pi = ci_get_pi(rdev); + u32 tmp; + + if (!pi->fan_ctrl_is_in_default_mode) { + tmp = RREG32_SMC(CG_FDO_CTRL2) & ~FDO_PWM_MODE_MASK; + tmp |= FDO_PWM_MODE(pi->fan_ctrl_default_mode); + WREG32_SMC(CG_FDO_CTRL2, tmp); + + tmp = RREG32_SMC(CG_FDO_CTRL2) & TMIN_MASK; + tmp |= TMIN(pi->t_min); + WREG32_SMC(CG_FDO_CTRL2, tmp); + pi->fan_ctrl_is_in_default_mode = true; + } +} + +static void ci_thermal_start_smc_fan_control(struct radeon_device *rdev) +{ + if (rdev->pm.dpm.fan.ucode_fan_control) { + ci_fan_ctrl_start_smc_fan_control(rdev); + ci_fan_ctrl_set_static_mode(rdev, FDO_PWM_MODE_STATIC); + } +} + +static void ci_thermal_initialize(struct radeon_device *rdev) +{ + u32 tmp; + + if (rdev->pm.fan_pulses_per_revolution) { + tmp = RREG32_SMC(CG_TACH_CTRL) & ~EDGE_PER_REV_MASK; + tmp |= EDGE_PER_REV(rdev->pm.fan_pulses_per_revolution -1); + WREG32_SMC(CG_TACH_CTRL, tmp); + } + + tmp = RREG32_SMC(CG_FDO_CTRL2) & ~TACH_PWM_RESP_RATE_MASK; + tmp |= TACH_PWM_RESP_RATE(0x28); + WREG32_SMC(CG_FDO_CTRL2, tmp); +} + +static int ci_thermal_start_thermal_controller(struct radeon_device *rdev) +{ + int ret; + + ci_thermal_initialize(rdev); + ret = ci_thermal_set_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + ret = ci_thermal_enable_alert(rdev, true); + if (ret) + return ret; + if (rdev->pm.dpm.fan.ucode_fan_control) { + ret = ci_thermal_setup_fan_table(rdev); + if (ret) + return ret; + ci_thermal_start_smc_fan_control(rdev); + } + + return 0; +} + +static void ci_thermal_stop_thermal_controller(struct radeon_device *rdev) +{ + if (!rdev->pm.no_fan) + ci_fan_ctrl_set_default_mode(rdev); +} + +#if 0 static int ci_read_smc_soft_register(struct radeon_device *rdev, u16 reg_offset, u32 *value) { @@ -4841,6 +5177,8 @@ int ci_dpm_enable(struct radeon_device *rdev) ci_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + ci_thermal_start_thermal_controller(rdev); + ci_update_current_ps(rdev, boot_ps); return 0; @@ -4886,6 +5224,8 @@ void ci_dpm_disable(struct radeon_device *rdev) if (!ci_is_smc_running(rdev)) return; + ci_thermal_stop_thermal_controller(rdev); + if (pi->thermal_protection) ci_enable_thermal_protection(rdev, false); ci_enable_power_containment(rdev, false); @@ -5473,6 +5813,9 @@ int ci_dpm_init(struct radeon_device *rdev) rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc = rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac; + pi->fan_ctrl_is_in_default_mode = true; + rdev->pm.dpm.fan.ucode_fan_control = false; + return 0; } diff --git a/drivers/gpu/drm/radeon/ci_dpm.h b/drivers/gpu/drm/radeon/ci_dpm.h index 615cb2c..bb19fbf 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.h +++ b/drivers/gpu/drm/radeon/ci_dpm.h @@ -266,6 +266,7 @@ struct ci_power_info { bool caps_automatic_dc_transition; bool caps_sclk_throttle_low_notification; bool caps_dynamic_ac_timing; + bool caps_od_fuzzy_fan_control_support; /* flags */ bool thermal_protection; bool pcie_performance_request; @@ -287,6 +288,10 @@ struct ci_power_info { struct ci_ps current_ps; struct radeon_ps requested_rps; struct ci_ps requested_ps; + /* fan control */ + bool fan_ctrl_is_in_default_mode; + u32 t_min; + u32 fan_ctrl_default_mode; }; #define CISLANDS_VOLTAGE_CONTROL_NONE 0x0 diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h index 068cbb0..e4e88ca 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -186,7 +186,10 @@ #define DIG_THERM_DPM(x) ((x) << 14) #define DIG_THERM_DPM_MASK 0x003FC000 #define DIG_THERM_DPM_SHIFT 14 - +#define CG_THERMAL_STATUS 0xC0300008 +#define FDO_PWM_DUTY(x) ((x) << 9) +#define FDO_PWM_DUTY_MASK (0xff << 9) +#define FDO_PWM_DUTY_SHIFT 9 #define CG_THERMAL_INT 0xC030000C #define CI_DIG_THERM_INTH(x) ((x) << 8) #define CI_DIG_THERM_INTH_MASK 0x0000FF00 @@ -196,7 +199,10 @@ #define CI_DIG_THERM_INTL_SHIFT 16 #define THERM_INT_MASK_HIGH (1 << 24) #define THERM_INT_MASK_LOW (1 << 25) - +#define CG_MULT_THERMAL_CTRL 0xC0300010 +#define TEMP_SEL(x) ((x) << 20) +#define TEMP_SEL_MASK (0xff << 20) +#define TEMP_SEL_SHIFT 20 #define CG_MULT_THERMAL_STATUS 0xC0300014 #define ASIC_MAX_TEMP(x) ((x) << 0) #define ASIC_MAX_TEMP_MASK 0x000001ff @@ -205,6 +211,36 @@ #define CTF_TEMP_MASK 0x0003fe00 #define CTF_TEMP_SHIFT 9 +#define CG_FDO_CTRL0 0xC0300064 +#define FDO_STATIC_DUTY(x) ((x) << 0) +#define FDO_STATIC_DUTY_MASK 0x0000000F +#define FDO_STATIC_DUTY_SHIFT 0 +#define CG_FDO_CTRL1 0xC0300068 +#define FMAX_DUTY100(x) ((x) << 0) +#define FMAX_DUTY100_MASK 0x0000000F +#define FMAX_DUTY100_SHIFT 0 +#define CG_FDO_CTRL2 0xC030006C +#define TMIN(x) ((x) << 0) +#define TMIN_MASK 0x0000000F +#define TMIN_SHIFT 0 +#define FDO_PWM_MODE(x) ((x) << 11) +#define FDO_PWM_MODE_MASK (3 << 11) +#define FDO_PWM_MODE_SHIFT 11 +#define TACH_PWM_RESP_RATE(x) ((x) << 25) +#define TACH_PWM_RESP_RATE_MASK (0x7f << 25) +#define TACH_PWM_RESP_RATE_SHIFT 25 +#define CG_TACH_CTRL 0xC0300070 +# define EDGE_PER_REV(x) ((x) << 0) +# define EDGE_PER_REV_MASK (0x7 << 0) +# define EDGE_PER_REV_SHIFT 0 +# define TARGET_PERIOD(x) ((x) << 3) +# define TARGET_PERIOD_MASK 0xfffffff8 +# define TARGET_PERIOD_SHIFT 3 +#define CG_TACH_STATUS 0xC0300074 +# define TACH_PERIOD(x) ((x) << 0) +# define TACH_PERIOD_MASK 0xffffffff +# define TACH_PERIOD_SHIFT 0 + #define CG_ECLK_CNTL 0xC05000AC # define ECLK_DIVIDER_MASK 0x7f # define ECLK_DIR_CNTL_EN (1 << 8) diff --git a/drivers/gpu/drm/radeon/ppsmc.h b/drivers/gpu/drm/radeon/ppsmc.h index 0c4eaa6..ff698b0 100644 --- a/drivers/gpu/drm/radeon/ppsmc.h +++ b/drivers/gpu/drm/radeon/ppsmc.h @@ -59,6 +59,11 @@ #define FDO_MODE_HARDWARE 0 #define FDO_MODE_PIECE_WISE_LINEAR 1 +enum FAN_CONTROL { + FAN_CONTROL_FUZZY, + FAN_CONTROL_TABLE +}; + #define PPSMC_Result_OK ((uint8_t)0x01) #define PPSMC_Result_Failed ((uint8_t)0xFF) @@ -155,6 +160,7 @@ typedef uint8_t PPSMC_Result; #define PPSMC_MSG_MASTER_DeepSleep_ON ((uint16_t) 0x18F) #define PPSMC_MSG_MASTER_DeepSleep_OFF ((uint16_t) 0x190) #define PPSMC_MSG_Remove_DC_Clamp ((uint16_t) 0x191) +#define PPSMC_MSG_SetFanPwmMax ((uint16_t) 0x19A) #define PPSMC_MSG_API_GetSclkFrequency ((uint16_t) 0x200) #define PPSMC_MSG_API_GetMclkFrequency ((uint16_t) 0x201) diff --git a/drivers/gpu/drm/radeon/pptable.h b/drivers/gpu/drm/radeon/pptable.h index 2d53299..4c2eec4 100644 --- a/drivers/gpu/drm/radeon/pptable.h +++ b/drivers/gpu/drm/radeon/pptable.h @@ -96,6 +96,14 @@ typedef struct _ATOM_PPLIB_FANTABLE2 USHORT usTMax; // The max temperature } ATOM_PPLIB_FANTABLE2; +typedef struct _ATOM_PPLIB_FANTABLE3 +{ + ATOM_PPLIB_FANTABLE2 basicTable2; + UCHAR ucFanControlMode; + USHORT usFanPWMMax; + USHORT usFanOutputSensitivity; +} ATOM_PPLIB_FANTABLE3; + typedef struct _ATOM_PPLIB_EXTENDEDHEADER { USHORT usSize; diff --git a/drivers/gpu/drm/radeon/r600_dpm.c b/drivers/gpu/drm/radeon/r600_dpm.c index f6309bd..76c6a17 100644 --- a/drivers/gpu/drm/radeon/r600_dpm.c +++ b/drivers/gpu/drm/radeon/r600_dpm.c @@ -811,6 +811,7 @@ union power_info { union fan_info { struct _ATOM_PPLIB_FANTABLE fan; struct _ATOM_PPLIB_FANTABLE2 fan2; + struct _ATOM_PPLIB_FANTABLE3 fan3; }; static int r600_parse_clk_voltage_dep_table(struct radeon_clock_voltage_dependency_table *radeon_table, @@ -900,6 +901,14 @@ int r600_parse_extended_power_table(struct radeon_device *rdev) else rdev->pm.dpm.fan.t_max = 10900; rdev->pm.dpm.fan.cycle_delay = 100000; + if (fan_info->fan.ucFanTableFormat >= 3) { + rdev->pm.dpm.fan.control_mode = fan_info->fan3.ucFanControlMode; + rdev->pm.dpm.fan.default_max_fan_pwm = + le16_to_cpu(fan_info->fan3.usFanPWMMax); + rdev->pm.dpm.fan.default_fan_output_sensitivity = 4836; + rdev->pm.dpm.fan.fan_output_sensitivity = + le16_to_cpu(fan_info->fan3.usFanOutputSensitivity); + } rdev->pm.dpm.fan.ucode_fan_control = true; } } diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 1f61ff0..5aabbe0 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1494,6 +1494,10 @@ struct radeon_dpm_fan { u8 t_hyst; u32 cycle_delay; u16 t_max; + u8 control_mode; + u16 default_max_fan_pwm; + u16 default_fan_output_sensitivity; + u16 fan_output_sensitivity; bool ucode_fan_control; }; diff --git a/drivers/gpu/drm/radeon/smu7_discrete.h b/drivers/gpu/drm/radeon/smu7_discrete.h index 82f70c9..0b0b404 100644 --- a/drivers/gpu/drm/radeon/smu7_discrete.h +++ b/drivers/gpu/drm/radeon/smu7_discrete.h @@ -431,6 +431,31 @@ struct SMU7_Discrete_MCRegisters typedef struct SMU7_Discrete_MCRegisters SMU7_Discrete_MCRegisters; +struct SMU7_Discrete_FanTable +{ + uint16_t FdoMode; + int16_t TempMin; + int16_t TempMed; + int16_t TempMax; + int16_t Slope1; + int16_t Slope2; + int16_t FdoMin; + int16_t HystUp; + int16_t HystDown; + int16_t HystSlope; + int16_t TempRespLim; + int16_t TempCurr; + int16_t SlopeCurr; + int16_t PwmCurr; + uint32_t RefreshPeriod; + int16_t FdoMax; + uint8_t TempSrc; + int8_t Padding; +}; + +typedef struct SMU7_Discrete_FanTable SMU7_Discrete_FanTable; + + struct SMU7_Discrete_PmFuses { // dw0-dw1 uint8_t BapmVddCVidHiSidd[8]; @@ -462,7 +487,10 @@ struct SMU7_Discrete_PmFuses { uint8_t BapmVddCVidHiSidd2[8]; // dw11-dw12 - uint32_t Reserved6[2]; + int16_t FuzzyFan_ErrorSetDelta; + int16_t FuzzyFan_ErrorRateSetDelta; + int16_t FuzzyFan_PwmSetDelta; + uint16_t CalcMeasPowerBlend; // dw13-dw16 uint8_t GnbLPML[16]; -- cgit v0.10.2 From 0e4ed1c1908e48ee66d9e955ce824bc1f34d103e Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 15 Sep 2014 02:15:04 -0400 Subject: drm/radeon/dpm: add thermal dpm support for CI Not currently used. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 4581d6c..8312140 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -714,6 +714,25 @@ static int ci_enable_smc_cac(struct radeon_device *rdev, bool enable) return ret; } +static int ci_enable_thermal_based_sclk_dpm(struct radeon_device *rdev, + bool enable) +{ + struct ci_power_info *pi = ci_get_pi(rdev); + PPSMC_Result smc_result = PPSMC_Result_OK; + + if (pi->thermal_sclk_dpm_enabled) { + if (enable) + smc_result = ci_send_msg_to_smc(rdev, PPSMC_MSG_ENABLE_THERMAL_DPM); + else + smc_result = ci_send_msg_to_smc(rdev, PPSMC_MSG_DISABLE_THERMAL_DPM); + } + + if (smc_result == PPSMC_Result_OK) + return 0; + else + return -EINVAL; +} + static int ci_power_control_set_level(struct radeon_device *rdev) { struct ci_power_info *pi = ci_get_pi(rdev); @@ -5177,6 +5196,12 @@ int ci_dpm_enable(struct radeon_device *rdev) ci_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + ret = ci_enable_thermal_based_sclk_dpm(rdev, true); + if (ret) { + DRM_ERROR("ci_enable_thermal_based_sclk_dpm failed\n"); + return ret; + } + ci_thermal_start_thermal_controller(rdev); ci_update_current_ps(rdev, boot_ps); @@ -5240,6 +5265,7 @@ void ci_dpm_disable(struct radeon_device *rdev) ci_reset_to_default(rdev); ci_dpm_stop_smc(rdev); ci_force_switch_to_arb_f0(rdev); + ci_enable_thermal_based_sclk_dpm(rdev, false); ci_update_current_ps(rdev, boot_ps); } @@ -5639,6 +5665,7 @@ int ci_dpm_init(struct radeon_device *rdev) pi->sclk_dpm_key_disabled = 0; pi->mclk_dpm_key_disabled = 0; pi->pcie_dpm_key_disabled = 0; + pi->thermal_sclk_dpm_enabled = 0; /* mclk dpm is unstable on some R7 260X cards with the old mc ucode */ if ((rdev->pdev->device == 0x6658) && diff --git a/drivers/gpu/drm/radeon/ci_dpm.h b/drivers/gpu/drm/radeon/ci_dpm.h index bb19fbf..84e3d3b 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.h +++ b/drivers/gpu/drm/radeon/ci_dpm.h @@ -239,6 +239,7 @@ struct ci_power_info { u32 sclk_dpm_key_disabled; u32 mclk_dpm_key_disabled; u32 pcie_dpm_key_disabled; + u32 thermal_sclk_dpm_enabled; struct ci_pcie_perf_range pcie_gen_performance; struct ci_pcie_perf_range pcie_lane_performance; struct ci_pcie_perf_range pcie_gen_powersaving; diff --git a/drivers/gpu/drm/radeon/ppsmc.h b/drivers/gpu/drm/radeon/ppsmc.h index ff698b0..7e5724a 100644 --- a/drivers/gpu/drm/radeon/ppsmc.h +++ b/drivers/gpu/drm/radeon/ppsmc.h @@ -162,6 +162,9 @@ typedef uint8_t PPSMC_Result; #define PPSMC_MSG_Remove_DC_Clamp ((uint16_t) 0x191) #define PPSMC_MSG_SetFanPwmMax ((uint16_t) 0x19A) +#define PPSMC_MSG_ENABLE_THERMAL_DPM ((uint16_t) 0x19C) +#define PPSMC_MSG_DISABLE_THERMAL_DPM ((uint16_t) 0x19D) + #define PPSMC_MSG_API_GetSclkFrequency ((uint16_t) 0x200) #define PPSMC_MSG_API_GetMclkFrequency ((uint16_t) 0x201) -- cgit v0.10.2 From 66648b81ce6a1ea713e299d2074d81d36cff4c94 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 17 Nov 2014 11:52:48 -0500 Subject: drm/radeon: fix PCC debugging message for CI DPM Add missing newline and print the bad gpio shift. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 8312140..c9963b5 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -5773,7 +5773,7 @@ int ci_dpm_init(struct radeon_device *rdev) tmp |= DPM_ENABLED; break; default: - DRM_ERROR("Invalid PCC GPIO!"); + DRM_ERROR("Invalid PCC GPIO: %u!\n", gpio.shift); break; } WREG32_SMC(CNB_PWRMGT_CNTL, tmp); -- cgit v0.10.2 From f281d0a37c423cb4cc8ef91ba7136848744c955f Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 18 Nov 2014 14:32:14 -0500 Subject: drm/radeon/ci: apply disp voltage changes before clk changes Apply voltage changes for displays before changing clocks. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index c9963b5..e781628 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -3752,6 +3752,8 @@ static int ci_upload_dpm_level_enable_mask(struct radeon_device *rdev) struct ci_power_info *pi = ci_get_pi(rdev); PPSMC_Result result; + ci_apply_disp_minimum_voltage_request(rdev); + if (!pi->sclk_dpm_key_disabled) { if (pi->dpm_level_enable_mask.sclk_dpm_enable_mask) { result = ci_send_msg_to_smc_with_parameter(rdev, @@ -3782,8 +3784,6 @@ static int ci_upload_dpm_level_enable_mask(struct radeon_device *rdev) } } - ci_apply_disp_minimum_voltage_request(rdev); - return 0; } -- cgit v0.10.2 From e745c3c9aa5e30ae94a1be04a1d2f326fb213a56 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 18 Nov 2014 14:36:59 -0500 Subject: drm/radeon/ci: use different smc command for pcie dpm Use unforce levels rather than enable mask instruction. This is the preferred method. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index e781628..73f8c4b 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -3773,7 +3773,7 @@ static int ci_upload_dpm_level_enable_mask(struct radeon_device *rdev) return -EINVAL; } } - +#if 0 if (!pi->pcie_dpm_key_disabled) { if (pi->dpm_level_enable_mask.pcie_dpm_enable_mask) { result = ci_send_msg_to_smc_with_parameter(rdev, @@ -3783,7 +3783,7 @@ static int ci_upload_dpm_level_enable_mask(struct radeon_device *rdev) return -EINVAL; } } - +#endif return 0; } @@ -4247,6 +4247,14 @@ int ci_dpm_force_performance_level(struct radeon_device *rdev, } } } else if (level == RADEON_DPM_FORCED_LEVEL_AUTO) { + if (!pi->pcie_dpm_key_disabled) { + PPSMC_Result smc_result; + + smc_result = ci_send_msg_to_smc(rdev, + PPSMC_MSG_PCIeDPM_UnForceLevel); + if (smc_result != PPSMC_Result_OK) + return -EINVAL; + } ret = ci_upload_dpm_level_enable_mask(rdev); if (ret) return ret; -- cgit v0.10.2 From 7f6233ca8769a92cf4f23a0bc18bf241e9c50606 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 18 Nov 2014 14:40:26 -0500 Subject: drm/radeon/ci: force pcie level before sclk and mclk Preferred ordering. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 73f8c4b..630434c 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -4143,57 +4143,57 @@ int ci_dpm_force_performance_level(struct radeon_device *rdev, int ret; if (level == RADEON_DPM_FORCED_LEVEL_HIGH) { - if ((!pi->sclk_dpm_key_disabled) && - pi->dpm_level_enable_mask.sclk_dpm_enable_mask) { + if ((!pi->pcie_dpm_key_disabled) && + pi->dpm_level_enable_mask.pcie_dpm_enable_mask) { levels = 0; - tmp = pi->dpm_level_enable_mask.sclk_dpm_enable_mask; + tmp = pi->dpm_level_enable_mask.pcie_dpm_enable_mask; while (tmp >>= 1) levels++; if (levels) { - ret = ci_dpm_force_state_sclk(rdev, levels); + ret = ci_dpm_force_state_pcie(rdev, level); if (ret) return ret; for (i = 0; i < rdev->usec_timeout; i++) { - tmp = (RREG32_SMC(TARGET_AND_CURRENT_PROFILE_INDEX) & - CURR_SCLK_INDEX_MASK) >> CURR_SCLK_INDEX_SHIFT; + tmp = (RREG32_SMC(TARGET_AND_CURRENT_PROFILE_INDEX_1) & + CURR_PCIE_INDEX_MASK) >> CURR_PCIE_INDEX_SHIFT; if (tmp == levels) break; udelay(1); } } } - if ((!pi->mclk_dpm_key_disabled) && - pi->dpm_level_enable_mask.mclk_dpm_enable_mask) { + if ((!pi->sclk_dpm_key_disabled) && + pi->dpm_level_enable_mask.sclk_dpm_enable_mask) { levels = 0; - tmp = pi->dpm_level_enable_mask.mclk_dpm_enable_mask; + tmp = pi->dpm_level_enable_mask.sclk_dpm_enable_mask; while (tmp >>= 1) levels++; if (levels) { - ret = ci_dpm_force_state_mclk(rdev, levels); + ret = ci_dpm_force_state_sclk(rdev, levels); if (ret) return ret; for (i = 0; i < rdev->usec_timeout; i++) { tmp = (RREG32_SMC(TARGET_AND_CURRENT_PROFILE_INDEX) & - CURR_MCLK_INDEX_MASK) >> CURR_MCLK_INDEX_SHIFT; + CURR_SCLK_INDEX_MASK) >> CURR_SCLK_INDEX_SHIFT; if (tmp == levels) break; udelay(1); } } } - if ((!pi->pcie_dpm_key_disabled) && - pi->dpm_level_enable_mask.pcie_dpm_enable_mask) { + if ((!pi->mclk_dpm_key_disabled) && + pi->dpm_level_enable_mask.mclk_dpm_enable_mask) { levels = 0; - tmp = pi->dpm_level_enable_mask.pcie_dpm_enable_mask; + tmp = pi->dpm_level_enable_mask.mclk_dpm_enable_mask; while (tmp >>= 1) levels++; if (levels) { - ret = ci_dpm_force_state_pcie(rdev, level); + ret = ci_dpm_force_state_mclk(rdev, levels); if (ret) return ret; for (i = 0; i < rdev->usec_timeout; i++) { - tmp = (RREG32_SMC(TARGET_AND_CURRENT_PROFILE_INDEX_1) & - CURR_PCIE_INDEX_MASK) >> CURR_PCIE_INDEX_SHIFT; + tmp = (RREG32_SMC(TARGET_AND_CURRENT_PROFILE_INDEX) & + CURR_MCLK_INDEX_MASK) >> CURR_MCLK_INDEX_SHIFT; if (tmp == levels) break; udelay(1); -- cgit v0.10.2 From d967be9b80a5aa3ba228a9c2d3fea91ae99e4a07 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 18 Nov 2014 15:07:33 -0500 Subject: drm/radeon/ci: disable needless sclk changes The current code always reprogrammed the sclk levels, but we don't currently handle disp sclk requirements so just skip it. Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 630434c..3f898d0 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -3809,7 +3809,7 @@ static void ci_find_dpm_states_clocks_in_dpm_table(struct radeon_device *rdev, pi->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_SCLK; } else { /* XXX check display min clock requirements */ - if (0 != CISLAND_MINIMUM_ENGINE_CLOCK) + if (CISLAND_MINIMUM_ENGINE_CLOCK != CISLAND_MINIMUM_ENGINE_CLOCK) pi->need_update_smu7_dpm_table |= DPMTABLE_UPDATE_SCLK; } -- cgit v0.10.2 From faffaf620f32c6e907d06570f8f75845ecb1349f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:19 +0100 Subject: drm/radeon: rework vm_flush parameters MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use ring structure instead of index and provide vm_id and pd_addr separately. Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index cef4cb7..3dc2be0 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -5982,26 +5982,23 @@ static void cik_vm_decode_fault(struct radeon_device *rdev, * Update the page table base and flush the VM TLB * using the CP (CIK). */ -void cik_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +void cik_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr) { - struct radeon_ring *ring = &rdev->ring[ridx]; - int usepfp = (ridx == RADEON_RING_TYPE_GFX_INDEX); - - if (vm == NULL) - return; + int usepfp = (ring->idx == RADEON_RING_TYPE_GFX_INDEX); radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) | WRITE_DATA_DST_SEL(0))); - if (vm->id < 8) { + if (vm_id < 8) { radeon_ring_write(ring, - (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2); + (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm_id << 2)) >> 2); } else { radeon_ring_write(ring, - (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2); + (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm_id - 8) << 2)) >> 2); } radeon_ring_write(ring, 0); - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + radeon_ring_write(ring, pd_addr >> 12); /* update SH_MEM_* regs */ radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); @@ -6009,7 +6006,7 @@ void cik_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) WRITE_DATA_DST_SEL(0))); radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); radeon_ring_write(ring, 0); - radeon_ring_write(ring, VMID(vm->id)); + radeon_ring_write(ring, VMID(vm_id)); radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 6)); radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) | @@ -6030,7 +6027,7 @@ void cik_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) radeon_ring_write(ring, VMID(0)); /* HDP flush */ - cik_hdp_flush_cp_ring_emit(rdev, ridx); + cik_hdp_flush_cp_ring_emit(rdev, ring->idx); /* bits 0-15 are the VM contexts0-15 */ radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); @@ -6038,7 +6035,7 @@ void cik_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) WRITE_DATA_DST_SEL(0))); radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2); radeon_ring_write(ring, 0); - radeon_ring_write(ring, 1 << vm->id); + radeon_ring_write(ring, 1 << vm_id); /* compute doesn't have PFP */ if (usepfp) { diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c index 4e8432d..7470a2e 100644 --- a/drivers/gpu/drm/radeon/cik_sdma.c +++ b/drivers/gpu/drm/radeon/cik_sdma.c @@ -901,25 +901,21 @@ void cik_sdma_vm_pad_ib(struct radeon_ib *ib) * Update the page table base and flush the VM TLB * using sDMA (CIK). */ -void cik_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +void cik_dma_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr) { - struct radeon_ring *ring = &rdev->ring[ridx]; - - if (vm == NULL) - return; - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - if (vm->id < 8) { - radeon_ring_write(ring, (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2); + if (vm_id < 8) { + radeon_ring_write(ring, (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm_id << 2)) >> 2); } else { - radeon_ring_write(ring, (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2); + radeon_ring_write(ring, (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm_id - 8) << 2)) >> 2); } - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + radeon_ring_write(ring, pd_addr >> 12); /* update SH_MEM_* regs */ radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); - radeon_ring_write(ring, VMID(vm->id)); + radeon_ring_write(ring, VMID(vm_id)); radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); radeon_ring_write(ring, SH_MEM_BASES >> 2); @@ -942,11 +938,11 @@ void cik_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm radeon_ring_write(ring, VMID(0)); /* flush HDP */ - cik_sdma_hdp_flush_ring_emit(rdev, ridx); + cik_sdma_hdp_flush_ring_emit(rdev, ring->idx); /* flush TLB */ radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2); - radeon_ring_write(ring, 1 << vm->id); + radeon_ring_write(ring, 1 << vm_id); } diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 3faee58..bee432d 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -2502,15 +2502,11 @@ void cayman_vm_decode_fault(struct radeon_device *rdev, * Update the page table base and flush the VM TLB * using the CP (cayman-si). */ -void cayman_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +void cayman_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr) { - struct radeon_ring *ring = &rdev->ring[ridx]; - - if (vm == NULL) - return; - - radeon_ring_write(ring, PACKET0(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2), 0)); - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + radeon_ring_write(ring, PACKET0(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm_id << 2), 0)); + radeon_ring_write(ring, pd_addr >> 12); /* flush hdp cache */ radeon_ring_write(ring, PACKET0(HDP_MEM_COHERENCY_FLUSH_CNTL, 0)); @@ -2518,7 +2514,7 @@ void cayman_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) /* bits 0-7 are the VM contexts0-7 */ radeon_ring_write(ring, PACKET0(VM_INVALIDATE_REQUEST, 0)); - radeon_ring_write(ring, 1 << vm->id); + radeon_ring_write(ring, 1 << vm_id); /* sync PFP to ME, otherwise we might get invalid PFP reads */ radeon_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0)); diff --git a/drivers/gpu/drm/radeon/ni_dma.c b/drivers/gpu/drm/radeon/ni_dma.c index f26f0a9..5a72404 100644 --- a/drivers/gpu/drm/radeon/ni_dma.c +++ b/drivers/gpu/drm/radeon/ni_dma.c @@ -446,16 +446,12 @@ void cayman_dma_vm_pad_ib(struct radeon_ib *ib) ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0); } -void cayman_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +void cayman_dma_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr) { - struct radeon_ring *ring = &rdev->ring[ridx]; - - if (vm == NULL) - return; - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); - radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2)); - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm_id << 2)) >> 2)); + radeon_ring_write(ring, pd_addr >> 12); /* flush hdp cache */ radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); @@ -465,6 +461,6 @@ void cayman_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm /* bits 0-7 are the VM contexts0-7 */ radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); radeon_ring_write(ring, (0xf << 16) | (VM_INVALIDATE_REQUEST >> 2)); - radeon_ring_write(ring, 1 << vm->id); + radeon_ring_write(ring, 1 << vm_id); } diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 5aabbe0..39f7716 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1798,7 +1798,8 @@ struct radeon_asic_ring { void (*hdp_flush)(struct radeon_device *rdev, struct radeon_ring *ring); bool (*emit_semaphore)(struct radeon_device *rdev, struct radeon_ring *cp, struct radeon_semaphore *semaphore, bool emit_wait); - void (*vm_flush)(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); + void (*vm_flush)(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr); /* testing functions */ int (*ring_test)(struct radeon_device *rdev, struct radeon_ring *cp); @@ -2850,7 +2851,7 @@ static inline void radeon_ring_write(struct radeon_ring *ring, uint32_t v) #define radeon_ring_ib_execute(rdev, r, ib) (rdev)->asic->ring[(r)]->ib_execute((rdev), (ib)) #define radeon_ring_ib_parse(rdev, r, ib) (rdev)->asic->ring[(r)]->ib_parse((rdev), (ib)) #define radeon_ring_is_lockup(rdev, r, cp) (rdev)->asic->ring[(r)]->is_lockup((rdev), (cp)) -#define radeon_ring_vm_flush(rdev, r, vm) (rdev)->asic->ring[(r)]->vm_flush((rdev), (r), (vm)) +#define radeon_ring_vm_flush(rdev, r, vm_id, pd_addr) (rdev)->asic->ring[(r)->idx]->vm_flush((rdev), (r), (vm_id), (pd_addr)) #define radeon_ring_get_rptr(rdev, r) (rdev)->asic->ring[(r)->idx]->get_rptr((rdev), (r)) #define radeon_ring_get_wptr(rdev, r) (rdev)->asic->ring[(r)->idx]->get_wptr((rdev), (r)) #define radeon_ring_set_wptr(rdev, r) (rdev)->asic->ring[(r)->idx]->set_wptr((rdev), (r)) diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index d8ace5b..2a45d54 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -599,7 +599,8 @@ int cayman_asic_reset(struct radeon_device *rdev); void cayman_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); int cayman_vm_init(struct radeon_device *rdev); void cayman_vm_fini(struct radeon_device *rdev); -void cayman_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); +void cayman_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr); uint32_t cayman_vm_page_flags(struct radeon_device *rdev, uint32_t flags); int evergreen_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib); int evergreen_dma_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib); @@ -624,7 +625,8 @@ void cayman_dma_vm_set_pages(struct radeon_device *rdev, uint32_t incr, uint32_t flags); void cayman_dma_vm_pad_ib(struct radeon_ib *ib); -void cayman_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); +void cayman_dma_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr); u32 cayman_gfx_get_rptr(struct radeon_device *rdev, struct radeon_ring *ring); @@ -699,7 +701,8 @@ int si_irq_set(struct radeon_device *rdev); int si_irq_process(struct radeon_device *rdev); int si_vm_init(struct radeon_device *rdev); void si_vm_fini(struct radeon_device *rdev); -void si_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); +void si_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr); int si_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib); struct radeon_fence *si_copy_dma(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, @@ -721,7 +724,8 @@ void si_dma_vm_set_pages(struct radeon_device *rdev, uint64_t addr, unsigned count, uint32_t incr, uint32_t flags); -void si_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); +void si_dma_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr); u32 si_get_xclk(struct radeon_device *rdev); uint64_t si_get_gpu_clock_counter(struct radeon_device *rdev); int si_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk); @@ -793,7 +797,8 @@ int cik_irq_set(struct radeon_device *rdev); int cik_irq_process(struct radeon_device *rdev); int cik_vm_init(struct radeon_device *rdev); void cik_vm_fini(struct radeon_device *rdev); -void cik_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); +void cik_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr); void cik_sdma_vm_copy_pages(struct radeon_device *rdev, struct radeon_ib *ib, @@ -811,7 +816,8 @@ void cik_sdma_vm_set_pages(struct radeon_device *rdev, uint32_t incr, uint32_t flags); void cik_sdma_vm_pad_ib(struct radeon_ib *ib); -void cik_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); +void cik_dma_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr); int cik_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib); u32 cik_gfx_get_rptr(struct radeon_device *rdev, struct radeon_ring *ring); diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index dfde266..9d0f87b 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -243,7 +243,8 @@ void radeon_vm_flush(struct radeon_device *rdev, if (!vm->last_flush || pd_addr != vm->pd_gpu_addr) { trace_radeon_vm_flush(pd_addr, ring, vm->id); vm->pd_gpu_addr = pd_addr; - radeon_ring_vm_flush(rdev, ring, vm); + radeon_ring_vm_flush(rdev, &rdev->ring[ring], + vm->id, vm->pd_gpu_addr); } } diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index eeea5b6..e91968b 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -5020,27 +5020,23 @@ static void si_vm_decode_fault(struct radeon_device *rdev, block, mc_id); } -void si_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +void si_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr) { - struct radeon_ring *ring = &rdev->ring[ridx]; - - if (vm == NULL) - return; - /* write new base address */ radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(1) | WRITE_DATA_DST_SEL(0))); - if (vm->id < 8) { + if (vm_id < 8) { radeon_ring_write(ring, - (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2); + (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm_id << 2)) >> 2); } else { radeon_ring_write(ring, - (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2); + (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm_id - 8) << 2)) >> 2); } radeon_ring_write(ring, 0); - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + radeon_ring_write(ring, pd_addr >> 12); /* flush hdp cache */ radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); @@ -5056,7 +5052,7 @@ void si_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) WRITE_DATA_DST_SEL(0))); radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2); radeon_ring_write(ring, 0); - radeon_ring_write(ring, 1 << vm->id); + radeon_ring_write(ring, 1 << vm_id); /* sync PFP to ME, otherwise we might get invalid PFP reads */ radeon_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0)); diff --git a/drivers/gpu/drm/radeon/si_dma.c b/drivers/gpu/drm/radeon/si_dma.c index b58f12b..e8bc0a5 100644 --- a/drivers/gpu/drm/radeon/si_dma.c +++ b/drivers/gpu/drm/radeon/si_dma.c @@ -185,20 +185,17 @@ void si_dma_vm_set_pages(struct radeon_device *rdev, } } -void si_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) -{ - struct radeon_ring *ring = &rdev->ring[ridx]; - - if (vm == NULL) - return; +void si_dma_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring, + unsigned vm_id, uint64_t pd_addr) +{ radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); - if (vm->id < 8) { - radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2)); + if (vm_id < 8) { + radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm_id << 2)) >> 2)); } else { - radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2)); + radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm_id - 8) << 2)) >> 2)); } - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + radeon_ring_write(ring, pd_addr >> 12); /* flush hdp cache */ radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); @@ -208,7 +205,7 @@ void si_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) /* bits 0-7 are the VM contexts0-7 */ radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); radeon_ring_write(ring, (0xf << 16) | (VM_INVALIDATE_REQUEST >> 2)); - radeon_ring_write(ring, 1 << vm->id); + radeon_ring_write(ring, 1 << vm_id); } /** -- cgit v0.10.2 From 85761f605a21ef7f3feda9d272565f97ecf1aa56 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:20 +0100 Subject: drm/radeon: stop re-reserving the BO in radeon_vm_bo_set_addr MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit That's useless when all callers drop the reservation immediately after calling the function. Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index c194497..f752c7f 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -601,6 +601,7 @@ int radeon_gem_va_ioctl(struct drm_device *dev, void *data, if (bo_va->it.start) { args->operation = RADEON_VA_RESULT_VA_EXIST; args->offset = bo_va->it.start * RADEON_GPU_PAGE_SIZE; + radeon_bo_unreserve(rbo); goto out; } r = radeon_vm_bo_set_addr(rdev, bo_va, args->offset, args->flags); @@ -616,7 +617,6 @@ int radeon_gem_va_ioctl(struct drm_device *dev, void *data, args->operation = RADEON_VA_RESULT_ERROR; } out: - radeon_bo_unreserve(rbo); drm_gem_object_unreference_unlocked(gobj); return r; } diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 6eb561d..f4dd26a 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -628,8 +628,6 @@ int radeon_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) RADEON_VA_IB_OFFSET, RADEON_VM_PAGE_READABLE | RADEON_VM_PAGE_SNOOPED); - - radeon_bo_unreserve(rdev->ring_tmp_bo.bo); if (r) { radeon_vm_fini(rdev, vm); kfree(fpriv); diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index 9d0f87b..db0ed3a 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -450,7 +450,7 @@ error: * Validate and set the offset requested within the vm address space. * Returns 0 for success, error for failure. * - * Object has to be reserved! + * Object has to be reserved and gets unreserved by this function! */ int radeon_vm_bo_set_addr(struct radeon_device *rdev, struct radeon_bo_va *bo_va, @@ -576,7 +576,7 @@ int radeon_vm_bo_set_addr(struct radeon_device *rdev, } mutex_unlock(&vm->mutex); - return radeon_bo_reserve(bo_va->bo, false); + return 0; } /** -- cgit v0.10.2 From e0602c35fe3bcbe5fe9179be78882b3101780781 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:21 +0100 Subject: drm/radeon: remove unnecessary VM syncs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The PD/PTs reservation object now contains everything needed. Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index db0ed3a..d8c49ad 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -701,7 +701,6 @@ int radeon_vm_update_page_directory(struct radeon_device *rdev, radeon_asic_vm_pad_ib(rdev, &ib); radeon_semaphore_sync_resv(rdev, ib.semaphore, pd->tbo.resv, false); - radeon_semaphore_sync_fence(ib.semaphore, vm->last_id_use); WARN_ON(ib.length_dw > ndw); r = radeon_ib_schedule(rdev, &ib, NULL, false); if (r) { @@ -969,7 +968,6 @@ int radeon_vm_bo_update(struct radeon_device *rdev, radeon_asic_vm_pad_ib(rdev, &ib); WARN_ON(ib.length_dw > ndw); - radeon_semaphore_sync_fence(ib.semaphore, vm->fence); r = radeon_ib_schedule(rdev, &ib, NULL, false); if (r) { radeon_ib_free(rdev, &ib); -- cgit v0.10.2 From 975700d2cc84408efa9b2360e38b1ab95368556f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:22 +0100 Subject: drm/radeon: split semaphore and sync object handling v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Previously we just allocated space for four hardware semaphores in each software semaphore object. Make software semaphore objects represent only one hardware semaphore address again by splitting the sync code into it's own object. v2: fix typo in comment Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index bad6caa..12bc212 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -80,7 +80,8 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \ r600_dpm.o rs780_dpm.o rv6xx_dpm.o rv770_dpm.o rv730_dpm.o rv740_dpm.o \ rv770_smc.o cypress_dpm.o btc_dpm.o sumo_dpm.o sumo_smc.o trinity_dpm.o \ trinity_smc.o ni_dpm.o si_smc.o si_dpm.o kv_smc.o kv_dpm.o ci_smc.o \ - ci_dpm.o dce6_afmt.o radeon_vm.o radeon_ucode.o radeon_ib.o radeon_mn.o + ci_dpm.o dce6_afmt.o radeon_vm.o radeon_ucode.o radeon_ib.o radeon_mn.o \ + radeon_sync.o # add async DMA block radeon-y += \ diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 3dc2be0..755923b 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -3994,31 +3994,27 @@ struct radeon_fence *cik_copy_cpdma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.blit_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes, control; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); r = radeon_ring_lock(rdev, ring, num_loops * 7 + 18); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(rdev, sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(rdev, &sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_bytes = size_in_bytes; @@ -4042,12 +4038,12 @@ struct radeon_fence *cik_copy_cpdma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c index 7470a2e..604e2e7 100644 --- a/drivers/gpu/drm/radeon/cik_sdma.c +++ b/drivers/gpu/drm/radeon/cik_sdma.c @@ -541,31 +541,27 @@ struct radeon_fence *cik_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); r = radeon_ring_lock(rdev, ring, num_loops * 7 + 14); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(rdev, sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(rdev, &sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_bytes = size_in_bytes; @@ -586,12 +582,12 @@ struct radeon_fence *cik_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/evergreen_dma.c b/drivers/gpu/drm/radeon/evergreen_dma.c index 66bcfad..96535aa 100644 --- a/drivers/gpu/drm/radeon/evergreen_dma.c +++ b/drivers/gpu/drm/radeon/evergreen_dma.c @@ -110,31 +110,27 @@ struct radeon_fence *evergreen_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_dw, cur_size_in_dw; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; num_loops = DIV_ROUND_UP(size_in_dw, 0xfffff); r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(rdev, sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(rdev, &sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_dw = size_in_dw; @@ -153,12 +149,12 @@ struct radeon_fence *evergreen_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 56b0292..ef5d606 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2889,31 +2889,27 @@ struct radeon_fence *r600_copy_cpdma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.blit_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes, tmp; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); r = radeon_ring_lock(rdev, ring, num_loops * 6 + 24); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(rdev, sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(rdev, &sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); radeon_ring_write(ring, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); @@ -2942,12 +2938,12 @@ struct radeon_fence *r600_copy_cpdma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/r600_dma.c b/drivers/gpu/drm/radeon/r600_dma.c index aabc343..3a58b80 100644 --- a/drivers/gpu/drm/radeon/r600_dma.c +++ b/drivers/gpu/drm/radeon/r600_dma.c @@ -441,31 +441,27 @@ struct radeon_fence *r600_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_dw, cur_size_in_dw; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFE); r = radeon_ring_lock(rdev, ring, num_loops * 4 + 8); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(rdev, sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(rdev, &sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_dw = size_in_dw; @@ -484,12 +480,12 @@ struct radeon_fence *r600_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 39f7716..7cda6d7 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -150,9 +150,6 @@ extern int radeon_backlight; /* number of hw syncs before falling back on blocking */ #define RADEON_NUM_SYNCS 4 -/* number of hw syncs before falling back on blocking */ -#define RADEON_NUM_SYNCS 4 - /* hardcode those limit for now */ #define RADEON_VA_IB_OFFSET (1 << 20) #define RADEON_VA_RESERVED_SIZE (8 << 20) @@ -576,10 +573,9 @@ int radeon_mode_dumb_mmap(struct drm_file *filp, * Semaphores. */ struct radeon_semaphore { - struct radeon_sa_bo *sa_bo; - signed waiters; - uint64_t gpu_addr; - struct radeon_fence *sync_to[RADEON_NUM_RINGS]; + struct radeon_sa_bo *sa_bo; + signed waiters; + uint64_t gpu_addr; }; int radeon_semaphore_create(struct radeon_device *rdev, @@ -588,20 +584,32 @@ bool radeon_semaphore_emit_signal(struct radeon_device *rdev, int ring, struct radeon_semaphore *semaphore); bool radeon_semaphore_emit_wait(struct radeon_device *rdev, int ring, struct radeon_semaphore *semaphore); -void radeon_semaphore_sync_fence(struct radeon_semaphore *semaphore, - struct radeon_fence *fence); -int radeon_semaphore_sync_resv(struct radeon_device *rdev, - struct radeon_semaphore *semaphore, - struct reservation_object *resv, - bool shared); -int radeon_semaphore_sync_rings(struct radeon_device *rdev, - struct radeon_semaphore *semaphore, - int waiting_ring); void radeon_semaphore_free(struct radeon_device *rdev, struct radeon_semaphore **semaphore, struct radeon_fence *fence); /* + * Synchronization + */ +struct radeon_sync { + struct radeon_semaphore *semaphores[RADEON_NUM_SYNCS]; + struct radeon_fence *sync_to[RADEON_NUM_RINGS]; +}; + +void radeon_sync_create(struct radeon_sync *sync); +void radeon_sync_fence(struct radeon_sync *sync, + struct radeon_fence *fence); +int radeon_sync_resv(struct radeon_device *rdev, + struct radeon_sync *sync, + struct reservation_object *resv, + bool shared); +int radeon_sync_rings(struct radeon_device *rdev, + struct radeon_sync *sync, + int waiting_ring); +void radeon_sync_free(struct radeon_device *rdev, struct radeon_sync *sync, + struct radeon_fence *fence); + +/* * GART structures, functions & helpers */ struct radeon_mc; @@ -818,7 +826,7 @@ struct radeon_ib { struct radeon_fence *fence; struct radeon_vm *vm; bool is_const_ib; - struct radeon_semaphore *semaphore; + struct radeon_sync sync; }; struct radeon_ring { diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index a3e7aed..30437aa 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -260,8 +260,8 @@ static int radeon_cs_sync_rings(struct radeon_cs_parser *p) continue; resv = p->relocs[i].robj->tbo.resv; - r = radeon_semaphore_sync_resv(p->rdev, p->ib.semaphore, resv, - p->relocs[i].tv.shared); + r = radeon_sync_resv(p->rdev, &p->ib.sync, resv, + p->relocs[i].tv.shared); if (r) break; @@ -285,9 +285,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data) INIT_LIST_HEAD(&p->validated); p->idx = 0; p->ib.sa_bo = NULL; - p->ib.semaphore = NULL; p->const_ib.sa_bo = NULL; - p->const_ib.semaphore = NULL; p->chunk_ib_idx = -1; p->chunk_relocs_idx = -1; p->chunk_flags_idx = -1; @@ -582,7 +580,7 @@ static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev, DRM_ERROR("Failed to sync rings: %i\n", r); goto out; } - radeon_semaphore_sync_fence(parser->ib.semaphore, vm->fence); + radeon_sync_fence(&parser->ib.sync, vm->fence); if ((rdev->family >= CHIP_TAHITI) && (parser->chunk_const_ib_idx != -1)) { diff --git a/drivers/gpu/drm/radeon/radeon_ib.c b/drivers/gpu/drm/radeon/radeon_ib.c index 3f39fcc..56a1704 100644 --- a/drivers/gpu/drm/radeon/radeon_ib.c +++ b/drivers/gpu/drm/radeon/radeon_ib.c @@ -64,10 +64,7 @@ int radeon_ib_get(struct radeon_device *rdev, int ring, return r; } - r = radeon_semaphore_create(rdev, &ib->semaphore); - if (r) { - return r; - } + radeon_sync_create(&ib->sync); ib->ring = ring; ib->fence = NULL; @@ -96,7 +93,7 @@ int radeon_ib_get(struct radeon_device *rdev, int ring, */ void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib *ib) { - radeon_semaphore_free(rdev, &ib->semaphore, ib->fence); + radeon_sync_free(rdev, &ib->sync, ib->fence); radeon_sa_bo_free(rdev, &ib->sa_bo, ib->fence); radeon_fence_unref(&ib->fence); } @@ -145,11 +142,11 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib, if (ib->vm) { struct radeon_fence *vm_id_fence; vm_id_fence = radeon_vm_grab_id(rdev, ib->vm, ib->ring); - radeon_semaphore_sync_fence(ib->semaphore, vm_id_fence); + radeon_sync_fence(&ib->sync, vm_id_fence); } /* sync with other rings */ - r = radeon_semaphore_sync_rings(rdev, ib->semaphore, ib->ring); + r = radeon_sync_rings(rdev, &ib->sync, ib->ring); if (r) { dev_err(rdev->dev, "failed to sync rings (%d)\n", r); radeon_ring_unlock_undo(rdev, ring); @@ -161,7 +158,7 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib, if (const_ib) { radeon_ring_ib_execute(rdev, const_ib->ring, const_ib); - radeon_semaphore_free(rdev, &const_ib->semaphore, NULL); + radeon_sync_free(rdev, &const_ib->sync, NULL); } radeon_ring_ib_execute(rdev, ib->ring, ib); r = radeon_fence_emit(rdev, &ib->fence, ib->ring); diff --git a/drivers/gpu/drm/radeon/radeon_semaphore.c b/drivers/gpu/drm/radeon/radeon_semaphore.c index 6deb08f..e6ad54c 100644 --- a/drivers/gpu/drm/radeon/radeon_semaphore.c +++ b/drivers/gpu/drm/radeon/radeon_semaphore.c @@ -34,15 +34,14 @@ int radeon_semaphore_create(struct radeon_device *rdev, struct radeon_semaphore **semaphore) { - uint64_t *cpu_addr; - int i, r; + int r; *semaphore = kmalloc(sizeof(struct radeon_semaphore), GFP_KERNEL); if (*semaphore == NULL) { return -ENOMEM; } - r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, &(*semaphore)->sa_bo, - 8 * RADEON_NUM_SYNCS, 8); + r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, + &(*semaphore)->sa_bo, 8, 8); if (r) { kfree(*semaphore); *semaphore = NULL; @@ -51,12 +50,7 @@ int radeon_semaphore_create(struct radeon_device *rdev, (*semaphore)->waiters = 0; (*semaphore)->gpu_addr = radeon_sa_bo_gpu_addr((*semaphore)->sa_bo); - cpu_addr = radeon_sa_bo_cpu_addr((*semaphore)->sa_bo); - for (i = 0; i < RADEON_NUM_SYNCS; ++i) - cpu_addr[i] = 0; - - for (i = 0; i < RADEON_NUM_RINGS; ++i) - (*semaphore)->sync_to[i] = NULL; + *((uint64_t *)radeon_sa_bo_cpu_addr((*semaphore)->sa_bo)) = 0; return 0; } @@ -95,146 +89,6 @@ bool radeon_semaphore_emit_wait(struct radeon_device *rdev, int ridx, return false; } -/** - * radeon_semaphore_sync_fence - use the semaphore to sync to a fence - * - * @semaphore: semaphore object to add fence to - * @fence: fence to sync to - * - * Sync to the fence using this semaphore object - */ -void radeon_semaphore_sync_fence(struct radeon_semaphore *semaphore, - struct radeon_fence *fence) -{ - struct radeon_fence *other; - - if (!fence) - return; - - other = semaphore->sync_to[fence->ring]; - semaphore->sync_to[fence->ring] = radeon_fence_later(fence, other); -} - -/** - * radeon_semaphore_sync_to - use the semaphore to sync to a reservation object - * - * @sema: semaphore object to add fence from reservation object to - * @resv: reservation object with embedded fence - * @shared: true if we should onyl sync to the exclusive fence - * - * Sync to the fence using this semaphore object - */ -int radeon_semaphore_sync_resv(struct radeon_device *rdev, - struct radeon_semaphore *sema, - struct reservation_object *resv, - bool shared) -{ - struct reservation_object_list *flist; - struct fence *f; - struct radeon_fence *fence; - unsigned i; - int r = 0; - - /* always sync to the exclusive fence */ - f = reservation_object_get_excl(resv); - fence = f ? to_radeon_fence(f) : NULL; - if (fence && fence->rdev == rdev) - radeon_semaphore_sync_fence(sema, fence); - else if (f) - r = fence_wait(f, true); - - flist = reservation_object_get_list(resv); - if (shared || !flist || r) - return r; - - for (i = 0; i < flist->shared_count; ++i) { - f = rcu_dereference_protected(flist->shared[i], - reservation_object_held(resv)); - fence = to_radeon_fence(f); - if (fence && fence->rdev == rdev) - radeon_semaphore_sync_fence(sema, fence); - else - r = fence_wait(f, true); - - if (r) - break; - } - return r; -} - -/** - * radeon_semaphore_sync_rings - sync ring to all registered fences - * - * @rdev: radeon_device pointer - * @semaphore: semaphore object to use for sync - * @ring: ring that needs sync - * - * Ensure that all registered fences are signaled before letting - * the ring continue. The caller must hold the ring lock. - */ -int radeon_semaphore_sync_rings(struct radeon_device *rdev, - struct radeon_semaphore *semaphore, - int ring) -{ - unsigned count = 0; - int i, r; - - for (i = 0; i < RADEON_NUM_RINGS; ++i) { - struct radeon_fence *fence = semaphore->sync_to[i]; - - /* check if we really need to sync */ - if (!radeon_fence_need_sync(fence, ring)) - continue; - - /* prevent GPU deadlocks */ - if (!rdev->ring[i].ready) { - dev_err(rdev->dev, "Syncing to a disabled ring!"); - return -EINVAL; - } - - if (++count > RADEON_NUM_SYNCS) { - /* not enough room, wait manually */ - r = radeon_fence_wait(fence, false); - if (r) - return r; - continue; - } - - /* allocate enough space for sync command */ - r = radeon_ring_alloc(rdev, &rdev->ring[i], 16); - if (r) { - return r; - } - - /* emit the signal semaphore */ - if (!radeon_semaphore_emit_signal(rdev, i, semaphore)) { - /* signaling wasn't successful wait manually */ - radeon_ring_undo(&rdev->ring[i]); - r = radeon_fence_wait(fence, false); - if (r) - return r; - continue; - } - - /* we assume caller has already allocated space on waiters ring */ - if (!radeon_semaphore_emit_wait(rdev, ring, semaphore)) { - /* waiting wasn't successful wait manually */ - radeon_ring_undo(&rdev->ring[i]); - r = radeon_fence_wait(fence, false); - if (r) - return r; - continue; - } - - radeon_ring_commit(rdev, &rdev->ring[i], false); - radeon_fence_note_sync(fence, ring); - - semaphore->gpu_addr += 8; - } - - return 0; -} - void radeon_semaphore_free(struct radeon_device *rdev, struct radeon_semaphore **semaphore, struct radeon_fence *fence) diff --git a/drivers/gpu/drm/radeon/radeon_sync.c b/drivers/gpu/drm/radeon/radeon_sync.c new file mode 100644 index 0000000..ddd88fb --- /dev/null +++ b/drivers/gpu/drm/radeon/radeon_sync.c @@ -0,0 +1,213 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + */ +/* + * Authors: + * Christian König + */ + +#include +#include "radeon.h" +#include "radeon_trace.h" + +/** + * radeon_sync_create - zero init sync object + * + * @sync: sync object to initialize + * + * Just clear the sync object for now. + */ +void radeon_sync_create(struct radeon_sync *sync) +{ + unsigned i; + + for (i = 0; i < RADEON_NUM_SYNCS; ++i) + sync->semaphores[i] = NULL; + + for (i = 0; i < RADEON_NUM_RINGS; ++i) + sync->sync_to[i] = NULL; +} + +/** + * radeon_sync_fence - use the semaphore to sync to a fence + * + * @sync: sync object to add fence to + * @fence: fence to sync to + * + * Sync to the fence using the semaphore objects + */ +void radeon_sync_fence(struct radeon_sync *sync, + struct radeon_fence *fence) +{ + struct radeon_fence *other; + + if (!fence) + return; + + other = sync->sync_to[fence->ring]; + sync->sync_to[fence->ring] = radeon_fence_later(fence, other); +} + +/** + * radeon_sync_resv - use the semaphores to sync to a reservation object + * + * @sync: sync object to add fences from reservation object to + * @resv: reservation object with embedded fence + * @shared: true if we should only sync to the exclusive fence + * + * Sync to the fence using the semaphore objects + */ +int radeon_sync_resv(struct radeon_device *rdev, + struct radeon_sync *sync, + struct reservation_object *resv, + bool shared) +{ + struct reservation_object_list *flist; + struct fence *f; + struct radeon_fence *fence; + unsigned i; + int r = 0; + + /* always sync to the exclusive fence */ + f = reservation_object_get_excl(resv); + fence = f ? to_radeon_fence(f) : NULL; + if (fence && fence->rdev == rdev) + radeon_sync_fence(sync, fence); + else if (f) + r = fence_wait(f, true); + + flist = reservation_object_get_list(resv); + if (shared || !flist || r) + return r; + + for (i = 0; i < flist->shared_count; ++i) { + f = rcu_dereference_protected(flist->shared[i], + reservation_object_held(resv)); + fence = to_radeon_fence(f); + if (fence && fence->rdev == rdev) + radeon_sync_fence(sync, fence); + else + r = fence_wait(f, true); + + if (r) + break; + } + return r; +} + +/** + * radeon_sync_rings - sync ring to all registered fences + * + * @rdev: radeon_device pointer + * @sync: sync object to use + * @ring: ring that needs sync + * + * Ensure that all registered fences are signaled before letting + * the ring continue. The caller must hold the ring lock. + */ +int radeon_sync_rings(struct radeon_device *rdev, + struct radeon_sync *sync, + int ring) +{ + unsigned count = 0; + int i, r; + + for (i = 0; i < RADEON_NUM_RINGS; ++i) { + struct radeon_fence *fence = sync->sync_to[i]; + struct radeon_semaphore *semaphore; + + /* check if we really need to sync */ + if (!radeon_fence_need_sync(fence, ring)) + continue; + + /* prevent GPU deadlocks */ + if (!rdev->ring[i].ready) { + dev_err(rdev->dev, "Syncing to a disabled ring!"); + return -EINVAL; + } + + if (count >= RADEON_NUM_SYNCS) { + /* not enough room, wait manually */ + r = radeon_fence_wait(fence, false); + if (r) + return r; + continue; + } + r = radeon_semaphore_create(rdev, &semaphore); + if (r) + return r; + + sync->semaphores[count++] = semaphore; + + /* allocate enough space for sync command */ + r = radeon_ring_alloc(rdev, &rdev->ring[i], 16); + if (r) + return r; + + /* emit the signal semaphore */ + if (!radeon_semaphore_emit_signal(rdev, i, semaphore)) { + /* signaling wasn't successful wait manually */ + radeon_ring_undo(&rdev->ring[i]); + r = radeon_fence_wait(fence, false); + if (r) + return r; + continue; + } + + /* we assume caller has already allocated space on waiters ring */ + if (!radeon_semaphore_emit_wait(rdev, ring, semaphore)) { + /* waiting wasn't successful wait manually */ + radeon_ring_undo(&rdev->ring[i]); + r = radeon_fence_wait(fence, false); + if (r) + return r; + continue; + } + + radeon_ring_commit(rdev, &rdev->ring[i], false); + radeon_fence_note_sync(fence, ring); + } + + return 0; +} + +/** + * radeon_sync_free - free the sync object + * + * @rdev: radeon_device pointer + * @sync: sync object to use + * @fence: fence to use for the free + * + * Free the sync object by freeing all semaphores in it. + */ +void radeon_sync_free(struct radeon_device *rdev, + struct radeon_sync *sync, + struct radeon_fence *fence) +{ + unsigned i; + + for (i = 0; i < RADEON_NUM_SYNCS; ++i) + radeon_semaphore_free(rdev, &sync->semaphores[i], fence); +} diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index d8c49ad..20ef826 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -700,7 +700,7 @@ int radeon_vm_update_page_directory(struct radeon_device *rdev, if (ib.length_dw != 0) { radeon_asic_vm_pad_ib(rdev, &ib); - radeon_semaphore_sync_resv(rdev, ib.semaphore, pd->tbo.resv, false); + radeon_sync_resv(rdev, &ib.sync, pd->tbo.resv, false); WARN_ON(ib.length_dw > ndw); r = radeon_ib_schedule(rdev, &ib, NULL, false); if (r) { @@ -826,7 +826,7 @@ static void radeon_vm_update_ptes(struct radeon_device *rdev, unsigned nptes; uint64_t pte; - radeon_semaphore_sync_resv(rdev, ib->semaphore, pt->tbo.resv, false); + radeon_sync_resv(rdev, &ib->sync, pt->tbo.resv, false); if ((addr & ~mask) == (end & ~mask)) nptes = end - addr; diff --git a/drivers/gpu/drm/radeon/rv770_dma.c b/drivers/gpu/drm/radeon/rv770_dma.c index 7f34bad..acff6e0 100644 --- a/drivers/gpu/drm/radeon/rv770_dma.c +++ b/drivers/gpu/drm/radeon/rv770_dma.c @@ -44,31 +44,27 @@ struct radeon_fence *rv770_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_dw, cur_size_in_dw; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFF); r = radeon_ring_lock(rdev, ring, num_loops * 5 + 8); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(rdev, sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(rdev, &sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_dw = size_in_dw; @@ -87,12 +83,12 @@ struct radeon_fence *rv770_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/si_dma.c b/drivers/gpu/drm/radeon/si_dma.c index e8bc0a5..f5cc777 100644 --- a/drivers/gpu/drm/radeon/si_dma.c +++ b/drivers/gpu/drm/radeon/si_dma.c @@ -226,31 +226,27 @@ struct radeon_fence *si_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); num_loops = DIV_ROUND_UP(size_in_bytes, 0xfffff); r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(rdev, sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(rdev, &sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_bytes = size_in_bytes; @@ -269,12 +265,12 @@ struct radeon_fence *si_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } -- cgit v0.10.2 From 587cdda8f739f4c57c91d3f73a1d5b2851a86cb8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:23 +0100 Subject: drm/radeon: fence PT updates manually v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This allows us to add the real execution fence as shared. v2: fix typo Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index 33e6c7a..686e450 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -818,3 +818,22 @@ int radeon_bo_wait(struct radeon_bo *bo, u32 *mem_type, bool no_wait) ttm_bo_unreserve(&bo->tbo); return r; } + +/** + * radeon_bo_fence - add fence to buffer object + * + * @bo: buffer object in question + * @fence: fence to add + * @shared: true if fence should be added shared + * + */ +void radeon_bo_fence(struct radeon_bo *bo, struct radeon_fence *fence, + bool shared) +{ + struct reservation_object *resv = bo->tbo.resv; + + if (shared) + reservation_object_add_shared_fence(resv, &fence->base); + else + reservation_object_add_excl_fence(resv, &fence->base); +} diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h index 1b8ec79..3b0b377 100644 --- a/drivers/gpu/drm/radeon/radeon_object.h +++ b/drivers/gpu/drm/radeon/radeon_object.h @@ -155,6 +155,8 @@ extern void radeon_bo_move_notify(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem); extern int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo); extern int radeon_bo_get_surface_reg(struct radeon_bo *bo); +extern void radeon_bo_fence(struct radeon_bo *bo, struct radeon_fence *fence, + bool shared); /* * sub allocation diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index 20ef826..4ca2779 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -143,7 +143,7 @@ struct radeon_cs_reloc *radeon_vm_get_bos(struct radeon_device *rdev, list[0].prefered_domains = RADEON_GEM_DOMAIN_VRAM; list[0].allowed_domains = RADEON_GEM_DOMAIN_VRAM; list[0].tv.bo = &vm->page_directory->tbo; - list[0].tv.shared = false; + list[0].tv.shared = true; list[0].tiling_flags = 0; list[0].handle = 0; list_add(&list[0].tv.head, head); @@ -157,7 +157,7 @@ struct radeon_cs_reloc *radeon_vm_get_bos(struct radeon_device *rdev, list[idx].prefered_domains = RADEON_GEM_DOMAIN_VRAM; list[idx].allowed_domains = RADEON_GEM_DOMAIN_VRAM; list[idx].tv.bo = &list[idx].robj->tbo; - list[idx].tv.shared = false; + list[idx].tv.shared = true; list[idx].tiling_flags = 0; list[idx].handle = 0; list_add(&list[idx++].tv.head, head); @@ -388,35 +388,25 @@ static void radeon_vm_set_pages(struct radeon_device *rdev, static int radeon_vm_clear_bo(struct radeon_device *rdev, struct radeon_bo *bo) { - struct ttm_validate_buffer tv; - struct ww_acquire_ctx ticket; - struct list_head head; struct radeon_ib ib; unsigned entries; uint64_t addr; int r; - memset(&tv, 0, sizeof(tv)); - tv.bo = &bo->tbo; - tv.shared = false; - - INIT_LIST_HEAD(&head); - list_add(&tv.head, &head); - - r = ttm_eu_reserve_buffers(&ticket, &head, true); - if (r) + r = radeon_bo_reserve(bo, false); + if (r) return r; - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); - if (r) - goto error; + r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + if (r) + goto error_unreserve; addr = radeon_bo_gpu_offset(bo); entries = radeon_bo_size(bo) / 8; r = radeon_ib_get(rdev, R600_RING_TYPE_DMA_INDEX, &ib, NULL, 256); if (r) - goto error; + goto error_unreserve; ib.length_dw = 0; @@ -426,15 +416,15 @@ static int radeon_vm_clear_bo(struct radeon_device *rdev, r = radeon_ib_schedule(rdev, &ib, NULL, false); if (r) - goto error; + goto error_free; - ttm_eu_fence_buffer_objects(&ticket, &head, &ib.fence->base); - radeon_ib_free(rdev, &ib); + radeon_bo_fence(bo, ib.fence, false); - return 0; +error_free: + radeon_ib_free(rdev, &ib); -error: - ttm_eu_backoff_reservation(&ticket, &head); +error_unreserve: + radeon_bo_unreserve(bo); return r; } @@ -707,6 +697,7 @@ int radeon_vm_update_page_directory(struct radeon_device *rdev, radeon_ib_free(rdev, &ib); return r; } + radeon_bo_fence(pd, ib.fence, false); radeon_fence_unref(&vm->fence); vm->fence = radeon_fence_ref(ib.fence); radeon_fence_unref(&vm->last_flush); @@ -863,6 +854,31 @@ static void radeon_vm_update_ptes(struct radeon_device *rdev, } /** + * radeon_vm_fence_pts - fence page tables after an update + * + * @vm: requested vm + * @start: start of GPU address range + * @end: end of GPU address range + * @fence: fence to use + * + * Fence the page tables in the range @start - @end (cayman+). + * + * Global and local mutex must be locked! + */ +static void radeon_vm_fence_pts(struct radeon_vm *vm, + uint64_t start, uint64_t end, + struct radeon_fence *fence) +{ + unsigned i; + + start >>= radeon_vm_block_size; + end >>= radeon_vm_block_size; + + for (i = start; i <= end; ++i) + radeon_bo_fence(vm->page_tables[i].bo, fence, false); +} + +/** * radeon_vm_bo_update - map a bo into the vm page table * * @rdev: radeon_device pointer @@ -973,6 +989,7 @@ int radeon_vm_bo_update(struct radeon_device *rdev, radeon_ib_free(rdev, &ib); return r; } + radeon_vm_fence_pts(vm, bo_va->it.start, bo_va->it.last + 1, ib.fence); radeon_fence_unref(&vm->fence); vm->fence = radeon_fence_ref(ib.fence); radeon_ib_free(rdev, &ib); -- cgit v0.10.2 From ad1a58a45ae9a35aaf68b27d21600889360680f3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:24 +0100 Subject: drm/radeon: track VM update fences separately MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Note for each fence if it's a VM page table update or not. This allows us to determine the last VM update in a sync object and so to figure out if we need to flush the TLB or not. Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 7cda6d7..61b2eea 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -360,14 +360,15 @@ struct radeon_fence_driver { }; struct radeon_fence { - struct fence base; + struct fence base; - struct radeon_device *rdev; - uint64_t seq; + struct radeon_device *rdev; + uint64_t seq; /* RB, DMA, etc. */ - unsigned ring; + unsigned ring; + bool is_vm_update; - wait_queue_t fence_wake; + wait_queue_t fence_wake; }; int radeon_fence_driver_start_ring(struct radeon_device *rdev, int ring); @@ -594,6 +595,7 @@ void radeon_semaphore_free(struct radeon_device *rdev, struct radeon_sync { struct radeon_semaphore *semaphores[RADEON_NUM_SYNCS]; struct radeon_fence *sync_to[RADEON_NUM_RINGS]; + struct radeon_fence *last_vm_update; }; void radeon_sync_create(struct radeon_sync *sync); @@ -926,8 +928,8 @@ struct radeon_vm { struct mutex mutex; /* last fence for cs using this vm */ struct radeon_fence *fence; - /* last flush or NULL if we still need to flush */ - struct radeon_fence *last_flush; + /* last flushed PD/PT update */ + struct radeon_fence *flushed_updates; /* last use of vmid */ struct radeon_fence *last_id_use; }; @@ -2975,7 +2977,7 @@ struct radeon_fence *radeon_vm_grab_id(struct radeon_device *rdev, struct radeon_vm *vm, int ring); void radeon_vm_flush(struct radeon_device *rdev, struct radeon_vm *vm, - int ring); + int ring, struct radeon_fence *fence); void radeon_vm_fence(struct radeon_device *rdev, struct radeon_vm *vm, struct radeon_fence *fence); diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index 9951670..d13d1b5 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -140,6 +140,7 @@ int radeon_fence_emit(struct radeon_device *rdev, (*fence)->rdev = rdev; (*fence)->seq = seq; (*fence)->ring = ring; + (*fence)->is_vm_update = false; fence_init(&(*fence)->base, &radeon_fence_ops, &rdev->fence_queue.lock, rdev->fence_context + ring, seq); radeon_fence_ring_emit(rdev, ring, *fence); diff --git a/drivers/gpu/drm/radeon/radeon_ib.c b/drivers/gpu/drm/radeon/radeon_ib.c index 56a1704..c39ce1f 100644 --- a/drivers/gpu/drm/radeon/radeon_ib.c +++ b/drivers/gpu/drm/radeon/radeon_ib.c @@ -154,7 +154,8 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib, } if (ib->vm) - radeon_vm_flush(rdev, ib->vm, ib->ring); + radeon_vm_flush(rdev, ib->vm, ib->ring, + ib->sync.last_vm_update); if (const_ib) { radeon_ring_ib_execute(rdev, const_ib->ring, const_ib); diff --git a/drivers/gpu/drm/radeon/radeon_sync.c b/drivers/gpu/drm/radeon/radeon_sync.c index ddd88fb..02ac8a1 100644 --- a/drivers/gpu/drm/radeon/radeon_sync.c +++ b/drivers/gpu/drm/radeon/radeon_sync.c @@ -48,6 +48,8 @@ void radeon_sync_create(struct radeon_sync *sync) for (i = 0; i < RADEON_NUM_RINGS; ++i) sync->sync_to[i] = NULL; + + sync->last_vm_update = NULL; } /** @@ -68,6 +70,11 @@ void radeon_sync_fence(struct radeon_sync *sync, other = sync->sync_to[fence->ring]; sync->sync_to[fence->ring] = radeon_fence_later(fence, other); + + if (fence->is_vm_update) { + other = sync->last_vm_update; + sync->last_vm_update = radeon_fence_later(fence, other); + } } /** diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index 4ca2779..6ff5741 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -190,7 +190,7 @@ struct radeon_fence *radeon_vm_grab_id(struct radeon_device *rdev, return NULL; /* we definately need to flush */ - radeon_fence_unref(&vm->last_flush); + vm->pd_gpu_addr = ~0ll; /* skip over VMID 0, since it is the system VM */ for (i = 1; i < rdev->vm_manager.nvm; ++i) { @@ -228,6 +228,7 @@ struct radeon_fence *radeon_vm_grab_id(struct radeon_device *rdev, * @rdev: radeon_device pointer * @vm: vm we want to flush * @ring: ring to use for flush + * @updates: last vm update that is waited for * * Flush the vm (cayman+). * @@ -235,13 +236,16 @@ struct radeon_fence *radeon_vm_grab_id(struct radeon_device *rdev, */ void radeon_vm_flush(struct radeon_device *rdev, struct radeon_vm *vm, - int ring) + int ring, struct radeon_fence *updates) { uint64_t pd_addr = radeon_bo_gpu_offset(vm->page_directory); - /* if we can't remember our last VM flush then flush now! */ - if (!vm->last_flush || pd_addr != vm->pd_gpu_addr) { + if (pd_addr != vm->pd_gpu_addr || !vm->flushed_updates || + radeon_fence_is_earlier(vm->flushed_updates, updates)) { + trace_radeon_vm_flush(pd_addr, ring, vm->id); + radeon_fence_unref(&vm->flushed_updates); + vm->flushed_updates = radeon_fence_ref(updates); vm->pd_gpu_addr = pd_addr; radeon_ring_vm_flush(rdev, &rdev->ring[ring], vm->id, vm->pd_gpu_addr); @@ -272,10 +276,6 @@ void radeon_vm_fence(struct radeon_device *rdev, radeon_fence_unref(&vm->last_id_use); vm->last_id_use = radeon_fence_ref(fence); - - /* we just flushed the VM, remember that */ - if (!vm->last_flush) - vm->last_flush = radeon_fence_ref(fence); } /** @@ -418,6 +418,7 @@ static int radeon_vm_clear_bo(struct radeon_device *rdev, if (r) goto error_free; + ib.fence->is_vm_update = true; radeon_bo_fence(bo, ib.fence, false); error_free: @@ -697,10 +698,10 @@ int radeon_vm_update_page_directory(struct radeon_device *rdev, radeon_ib_free(rdev, &ib); return r; } + ib.fence->is_vm_update = true; radeon_bo_fence(pd, ib.fence, false); radeon_fence_unref(&vm->fence); vm->fence = radeon_fence_ref(ib.fence); - radeon_fence_unref(&vm->last_flush); } radeon_ib_free(rdev, &ib); @@ -989,11 +990,11 @@ int radeon_vm_bo_update(struct radeon_device *rdev, radeon_ib_free(rdev, &ib); return r; } + ib.fence->is_vm_update = true; radeon_vm_fence_pts(vm, bo_va->it.start, bo_va->it.last + 1, ib.fence); radeon_fence_unref(&vm->fence); vm->fence = radeon_fence_ref(ib.fence); radeon_ib_free(rdev, &ib); - radeon_fence_unref(&vm->last_flush); return 0; } @@ -1124,7 +1125,7 @@ int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm) vm->id = 0; vm->ib_bo_va = NULL; vm->fence = NULL; - vm->last_flush = NULL; + vm->flushed_updates = NULL; vm->last_id_use = NULL; mutex_init(&vm->mutex); vm->va = RB_ROOT; @@ -1196,7 +1197,7 @@ void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm) radeon_bo_unref(&vm->page_directory); radeon_fence_unref(&vm->fence); - radeon_fence_unref(&vm->last_flush); + radeon_fence_unref(&vm->flushed_updates); radeon_fence_unref(&vm->last_id_use); mutex_destroy(&vm->mutex); -- cgit v0.10.2 From 7c42bc1aa23fc061a6ff9c2bd9208817bd54ea04 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:25 +0100 Subject: drm/radeon: use one VMID for each ring MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use multiple VMIDs for each VM, one for each ring. That allows us to execute flushes separately on each ring, still not ideal cause in a lot of cases rings can share IDs. Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 755923b..3deeed3 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -4066,6 +4066,7 @@ struct radeon_fence *cik_copy_cpdma(struct radeon_device *rdev, void cik_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) { struct radeon_ring *ring = &rdev->ring[ib->ring]; + unsigned vm_id = ib->vm ? ib->vm->ids[ib->ring].id : 0; u32 header, control = INDIRECT_BUFFER_VALID; if (ib->is_const_ib) { @@ -4094,8 +4095,7 @@ void cik_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); } - control |= ib->length_dw | - (ib->vm ? (ib->vm->id << 24) : 0); + control |= ib->length_dw | (vm_id << 24); radeon_ring_write(ring, header); radeon_ring_write(ring, diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c index 604e2e7..54b9837 100644 --- a/drivers/gpu/drm/radeon/cik_sdma.c +++ b/drivers/gpu/drm/radeon/cik_sdma.c @@ -134,7 +134,7 @@ void cik_sdma_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) { struct radeon_ring *ring = &rdev->ring[ib->ring]; - u32 extra_bits = (ib->vm ? ib->vm->id : 0) & 0xf; + u32 extra_bits = (ib->vm ? ib->vm->ids[ib->ring].id : 0) & 0xf; if (rdev->wb.enabled) { u32 next_rptr = ring->wptr + 5; diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index bee432d..360de9f 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -1373,6 +1373,7 @@ void cayman_fence_ring_emit(struct radeon_device *rdev, void cayman_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) { struct radeon_ring *ring = &rdev->ring[ib->ring]; + unsigned vm_id = ib->vm ? ib->vm->ids[ib->ring].id : 0; u32 cp_coher_cntl = PACKET3_FULL_CACHE_ENA | PACKET3_TC_ACTION_ENA | PACKET3_SH_ACTION_ENA; @@ -1395,15 +1396,14 @@ void cayman_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) #endif (ib->gpu_addr & 0xFFFFFFFC)); radeon_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFF); - radeon_ring_write(ring, ib->length_dw | - (ib->vm ? (ib->vm->id << 24) : 0)); + radeon_ring_write(ring, ib->length_dw | (vm_id << 24)); /* flush read cache over gart for this vmid */ radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3)); radeon_ring_write(ring, PACKET3_ENGINE_ME | cp_coher_cntl); radeon_ring_write(ring, 0xFFFFFFFF); radeon_ring_write(ring, 0); - radeon_ring_write(ring, ((ib->vm ? ib->vm->id : 0) << 24) | 10); /* poll interval */ + radeon_ring_write(ring, (vm_id << 24) | 10); /* poll interval */ } static void cayman_cp_enable(struct radeon_device *rdev, bool enable) diff --git a/drivers/gpu/drm/radeon/ni_dma.c b/drivers/gpu/drm/radeon/ni_dma.c index 5a72404..50f8861 100644 --- a/drivers/gpu/drm/radeon/ni_dma.c +++ b/drivers/gpu/drm/radeon/ni_dma.c @@ -123,6 +123,7 @@ void cayman_dma_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) { struct radeon_ring *ring = &rdev->ring[ib->ring]; + unsigned vm_id = ib->vm ? ib->vm->ids[ib->ring].id : 0; if (rdev->wb.enabled) { u32 next_rptr = ring->wptr + 4; @@ -140,7 +141,7 @@ void cayman_dma_ring_ib_execute(struct radeon_device *rdev, */ while ((ring->wptr & 7) != 5) radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); - radeon_ring_write(ring, DMA_IB_PACKET(DMA_PACKET_INDIRECT_BUFFER, ib->vm ? ib->vm->id : 0, 0)); + radeon_ring_write(ring, DMA_IB_PACKET(DMA_PACKET_INDIRECT_BUFFER, vm_id, 0)); radeon_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0)); radeon_ring_write(ring, (ib->length_dw << 12) | (upper_32_bits(ib->gpu_addr) & 0xFF)); diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 61b2eea..79f5f5b 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -905,33 +905,39 @@ struct radeon_vm_pt { uint64_t addr; }; +struct radeon_vm_id { + unsigned id; + uint64_t pd_gpu_addr; + /* last flushed PD/PT update */ + struct radeon_fence *flushed_updates; + /* last use of vmid */ + struct radeon_fence *last_id_use; +}; + struct radeon_vm { - struct rb_root va; - unsigned id; + struct rb_root va; /* BOs moved, but not yet updated in the PT */ - struct list_head invalidated; + struct list_head invalidated; /* BOs freed, but not yet updated in the PT */ - struct list_head freed; + struct list_head freed; /* contains the page directory */ - struct radeon_bo *page_directory; - uint64_t pd_gpu_addr; - unsigned max_pde_used; + struct radeon_bo *page_directory; + unsigned max_pde_used; /* array of page tables, one for each page directory entry */ - struct radeon_vm_pt *page_tables; + struct radeon_vm_pt *page_tables; - struct radeon_bo_va *ib_bo_va; + struct radeon_bo_va *ib_bo_va; - struct mutex mutex; + struct mutex mutex; /* last fence for cs using this vm */ - struct radeon_fence *fence; - /* last flushed PD/PT update */ - struct radeon_fence *flushed_updates; - /* last use of vmid */ - struct radeon_fence *last_id_use; + struct radeon_fence *fence; + + /* for id and flush management per ring */ + struct radeon_vm_id ids[RADEON_NUM_RINGS]; }; struct radeon_vm_manager { diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index 6ff5741..e38efe4 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -182,15 +182,18 @@ struct radeon_fence *radeon_vm_grab_id(struct radeon_device *rdev, struct radeon_vm *vm, int ring) { struct radeon_fence *best[RADEON_NUM_RINGS] = {}; + struct radeon_vm_id *vm_id = &vm->ids[ring]; + unsigned choices[2] = {}; unsigned i; /* check if the id is still valid */ - if (vm->last_id_use && vm->last_id_use == rdev->vm_manager.active[vm->id]) + if (vm_id->id && vm_id->last_id_use && + vm_id->last_id_use == rdev->vm_manager.active[vm_id->id]) return NULL; /* we definately need to flush */ - vm->pd_gpu_addr = ~0ll; + vm_id->pd_gpu_addr = ~0ll; /* skip over VMID 0, since it is the system VM */ for (i = 1; i < rdev->vm_manager.nvm; ++i) { @@ -198,8 +201,8 @@ struct radeon_fence *radeon_vm_grab_id(struct radeon_device *rdev, if (fence == NULL) { /* found a free one */ - vm->id = i; - trace_radeon_vm_grab_id(vm->id, ring); + vm_id->id = i; + trace_radeon_vm_grab_id(i, ring); return NULL; } @@ -211,8 +214,8 @@ struct radeon_fence *radeon_vm_grab_id(struct radeon_device *rdev, for (i = 0; i < 2; ++i) { if (choices[i]) { - vm->id = choices[i]; - trace_radeon_vm_grab_id(vm->id, ring); + vm_id->id = choices[i]; + trace_radeon_vm_grab_id(choices[i], ring); return rdev->vm_manager.active[choices[i]]; } } @@ -239,16 +242,18 @@ void radeon_vm_flush(struct radeon_device *rdev, int ring, struct radeon_fence *updates) { uint64_t pd_addr = radeon_bo_gpu_offset(vm->page_directory); + struct radeon_vm_id *vm_id = &vm->ids[ring]; - if (pd_addr != vm->pd_gpu_addr || !vm->flushed_updates || - radeon_fence_is_earlier(vm->flushed_updates, updates)) { + if (pd_addr != vm_id->pd_gpu_addr || !vm_id->flushed_updates || + radeon_fence_is_earlier(vm_id->flushed_updates, updates)) { - trace_radeon_vm_flush(pd_addr, ring, vm->id); - radeon_fence_unref(&vm->flushed_updates); - vm->flushed_updates = radeon_fence_ref(updates); - vm->pd_gpu_addr = pd_addr; + trace_radeon_vm_flush(pd_addr, ring, vm->ids[ring].id); + radeon_fence_unref(&vm_id->flushed_updates); + vm_id->flushed_updates = radeon_fence_ref(updates); + vm_id->pd_gpu_addr = pd_addr; radeon_ring_vm_flush(rdev, &rdev->ring[ring], - vm->id, vm->pd_gpu_addr); + vm_id->id, vm_id->pd_gpu_addr); + } } @@ -268,14 +273,16 @@ void radeon_vm_fence(struct radeon_device *rdev, struct radeon_vm *vm, struct radeon_fence *fence) { + unsigned vm_id = vm->ids[fence->ring].id; + radeon_fence_unref(&vm->fence); vm->fence = radeon_fence_ref(fence); - radeon_fence_unref(&rdev->vm_manager.active[vm->id]); - rdev->vm_manager.active[vm->id] = radeon_fence_ref(fence); + radeon_fence_unref(&rdev->vm_manager.active[vm_id]); + rdev->vm_manager.active[vm_id] = radeon_fence_ref(fence); - radeon_fence_unref(&vm->last_id_use); - vm->last_id_use = radeon_fence_ref(fence); + radeon_fence_unref(&vm->ids[fence->ring].last_id_use); + vm->ids[fence->ring].last_id_use = radeon_fence_ref(fence); } /** @@ -1120,13 +1127,16 @@ int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm) const unsigned align = min(RADEON_VM_PTB_ALIGN_SIZE, RADEON_VM_PTE_COUNT * 8); unsigned pd_size, pd_entries, pts_size; - int r; + int i, r; - vm->id = 0; vm->ib_bo_va = NULL; vm->fence = NULL; - vm->flushed_updates = NULL; - vm->last_id_use = NULL; + + for (i = 0; i < RADEON_NUM_RINGS; ++i) { + vm->ids[i].id = 0; + vm->ids[i].flushed_updates = NULL; + vm->ids[i].last_id_use = NULL; + } mutex_init(&vm->mutex); vm->va = RB_ROOT; INIT_LIST_HEAD(&vm->invalidated); @@ -1197,8 +1207,11 @@ void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm) radeon_bo_unref(&vm->page_directory); radeon_fence_unref(&vm->fence); - radeon_fence_unref(&vm->flushed_updates); - radeon_fence_unref(&vm->last_id_use); + + for (i = 0; i < RADEON_NUM_RINGS; ++i) { + radeon_fence_unref(&vm->ids[i].flushed_updates); + radeon_fence_unref(&vm->ids[i].last_id_use); + } mutex_destroy(&vm->mutex); } diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index e91968b..14896ce 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -3362,6 +3362,7 @@ void si_fence_ring_emit(struct radeon_device *rdev, void si_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) { struct radeon_ring *ring = &rdev->ring[ib->ring]; + unsigned vm_id = ib->vm ? ib->vm->ids[ib->ring].id : 0; u32 header; if (ib->is_const_ib) { @@ -3397,14 +3398,13 @@ void si_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) #endif (ib->gpu_addr & 0xFFFFFFFC)); radeon_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF); - radeon_ring_write(ring, ib->length_dw | - (ib->vm ? (ib->vm->id << 24) : 0)); + radeon_ring_write(ring, ib->length_dw | (vm_id << 24)); if (!ib->is_const_ib) { /* flush read cache over gart for this vmid */ radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); radeon_ring_write(ring, (CP_COHER_CNTL2 - PACKET3_SET_CONFIG_REG_START) >> 2); - radeon_ring_write(ring, ib->vm ? ib->vm->id : 0); + radeon_ring_write(ring, vm_id); radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3)); radeon_ring_write(ring, PACKET3_TCL1_ACTION_ENA | PACKET3_TC_ACTION_ENA | -- cgit v0.10.2 From 94214635f09c9211023730acdff9342fdf100aee Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:26 +0100 Subject: drm/radeon: fence BO_VAs manually MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This allows us to finally remove the VM fence and so allow concurrent use of it from different engines. Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 79f5f5b..3207bb6 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -456,6 +456,7 @@ struct radeon_bo_va { struct list_head bo_list; uint32_t flags; uint64_t addr; + struct radeon_fence *last_pt_update; unsigned ref_count; /* protected by vm mutex */ @@ -915,6 +916,8 @@ struct radeon_vm_id { }; struct radeon_vm { + struct mutex mutex; + struct rb_root va; /* BOs moved, but not yet updated in the PT */ @@ -932,10 +935,6 @@ struct radeon_vm { struct radeon_bo_va *ib_bo_va; - struct mutex mutex; - /* last fence for cs using this vm */ - struct radeon_fence *fence; - /* for id and flush management per ring */ struct radeon_vm_id ids[RADEON_NUM_RINGS]; }; diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index 30437aa..75f22e5 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -505,6 +505,9 @@ static int radeon_bo_vm_update_pte(struct radeon_cs_parser *p, if (r) return r; + radeon_sync_resv(p->rdev, &p->ib.sync, vm->page_directory->tbo.resv, + true); + r = radeon_vm_clear_freed(rdev, vm); if (r) return r; @@ -536,6 +539,8 @@ static int radeon_bo_vm_update_pte(struct radeon_cs_parser *p, r = radeon_vm_bo_update(rdev, bo_va, &bo->tbo.mem); if (r) return r; + + radeon_sync_fence(&p->ib.sync, bo_va->last_pt_update); } return radeon_vm_clear_invalids(rdev, vm); @@ -580,7 +585,6 @@ static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev, DRM_ERROR("Failed to sync rings: %i\n", r); goto out; } - radeon_sync_fence(&parser->ib.sync, vm->fence); if ((rdev->family >= CHIP_TAHITI) && (parser->chunk_const_ib_idx != -1)) { diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index e38efe4..f457614 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -275,9 +275,6 @@ void radeon_vm_fence(struct radeon_device *rdev, { unsigned vm_id = vm->ids[fence->ring].id; - radeon_fence_unref(&vm->fence); - vm->fence = radeon_fence_ref(fence); - radeon_fence_unref(&rdev->vm_manager.active[vm_id]); rdev->vm_manager.active[vm_id] = radeon_fence_ref(fence); @@ -707,8 +704,6 @@ int radeon_vm_update_page_directory(struct radeon_device *rdev, } ib.fence->is_vm_update = true; radeon_bo_fence(pd, ib.fence, false); - radeon_fence_unref(&vm->fence); - vm->fence = radeon_fence_ref(ib.fence); } radeon_ib_free(rdev, &ib); @@ -999,8 +994,8 @@ int radeon_vm_bo_update(struct radeon_device *rdev, } ib.fence->is_vm_update = true; radeon_vm_fence_pts(vm, bo_va->it.start, bo_va->it.last + 1, ib.fence); - radeon_fence_unref(&vm->fence); - vm->fence = radeon_fence_ref(ib.fence); + radeon_fence_unref(&bo_va->last_pt_update); + bo_va->last_pt_update = radeon_fence_ref(ib.fence); radeon_ib_free(rdev, &ib); return 0; @@ -1026,6 +1021,7 @@ int radeon_vm_clear_freed(struct radeon_device *rdev, list_for_each_entry_safe(bo_va, tmp, &vm->freed, vm_status) { r = radeon_vm_bo_update(rdev, bo_va, NULL); radeon_bo_unref(&bo_va->bo); + radeon_fence_unref(&bo_va->last_pt_update); kfree(bo_va); if (r) return r; @@ -1084,6 +1080,7 @@ void radeon_vm_bo_rmv(struct radeon_device *rdev, bo_va->bo = radeon_bo_ref(bo_va->bo); list_add(&bo_va->vm_status, &vm->freed); } else { + radeon_fence_unref(&bo_va->last_pt_update); kfree(bo_va); } @@ -1130,8 +1127,6 @@ int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm) int i, r; vm->ib_bo_va = NULL; - vm->fence = NULL; - for (i = 0; i < RADEON_NUM_RINGS; ++i) { vm->ids[i].id = 0; vm->ids[i].flushed_updates = NULL; @@ -1192,11 +1187,13 @@ void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm) if (!r) { list_del_init(&bo_va->bo_list); radeon_bo_unreserve(bo_va->bo); + radeon_fence_unref(&bo_va->last_pt_update); kfree(bo_va); } } list_for_each_entry_safe(bo_va, tmp, &vm->freed, vm_status) { radeon_bo_unref(&bo_va->bo); + radeon_fence_unref(&bo_va->last_pt_update); kfree(bo_va); } @@ -1206,8 +1203,6 @@ void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm) radeon_bo_unref(&vm->page_directory); - radeon_fence_unref(&vm->fence); - for (i = 0; i < RADEON_NUM_RINGS; ++i) { radeon_fence_unref(&vm->ids[i].flushed_updates); radeon_fence_unref(&vm->ids[i].last_id_use); -- cgit v0.10.2 From 43ac885764a589626018cc66973507213cef96ca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:27 +0100 Subject: drm/radeon: sync PD updates as shared MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We never invalidate PD entries and making them valid can run with other users in parallel. Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index f457614..470451c 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -695,7 +695,7 @@ int radeon_vm_update_page_directory(struct radeon_device *rdev, if (ib.length_dw != 0) { radeon_asic_vm_pad_ib(rdev, &ib); - radeon_sync_resv(rdev, &ib.sync, pd->tbo.resv, false); + radeon_sync_resv(rdev, &ib.sync, pd->tbo.resv, true); WARN_ON(ib.length_dw > ndw); r = radeon_ib_schedule(rdev, &ib, NULL, false); if (r) { -- cgit v0.10.2 From d1968e1db892e53b5783570a5738477b2a51080b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 19 Nov 2014 14:01:28 +0100 Subject: drm/radeon: sync PT updates as shared v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Only invalidating PTEs needs to be executed synchronized to using the PT. v2: fix sync to uses Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index 470451c..0b10f3a 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -820,7 +820,7 @@ static void radeon_vm_update_ptes(struct radeon_device *rdev, unsigned nptes; uint64_t pte; - radeon_sync_resv(rdev, &ib->sync, pt->tbo.resv, false); + radeon_sync_resv(rdev, &ib->sync, pt->tbo.resv, true); if ((addr & ~mask) == (end & ~mask)) nptes = end - addr; @@ -980,6 +980,13 @@ int radeon_vm_bo_update(struct radeon_device *rdev, return r; ib.length_dw = 0; + if (!(bo_va->flags & RADEON_VM_PAGE_VALID)) { + unsigned i; + + for (i = 0; i < RADEON_NUM_RINGS; ++i) + radeon_sync_fence(&ib.sync, vm->ids[i].last_id_use); + } + radeon_vm_update_ptes(rdev, vm, &ib, bo_va->it.start, bo_va->it.last + 1, addr, radeon_vm_page_flags(bo_va->flags)); -- cgit v0.10.2 From 2f2624c23511b4bf0dd3d4c5ae167715513f351d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Fri, 12 Sep 2014 12:25:45 +0200 Subject: drm/radeon: update the VM after setting BO address v4 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This way the necessary VM update is kicked off immediately if all BOs involved are in GPU accessible memory. v2: fix vm lock v3: immediately update unmaps as well v4: use drm_free_large instead of kfree Tested-by: Kai Wasserbäch Signed-off-by: Christian König Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index f752c7f..a748a64 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -518,6 +518,68 @@ out: return r; } +/** + * radeon_gem_va_update_vm -update the bo_va in its VM + * + * @rdev: radeon_device pointer + * @bo_va: bo_va to update + * + * Update the bo_va directly after setting it's address. Errors are not + * vital here, so they are not reported back to userspace. + */ +static void radeon_gem_va_update_vm(struct radeon_device *rdev, + struct radeon_bo_va *bo_va) +{ + struct ttm_validate_buffer tv, *entry; + struct radeon_cs_reloc *vm_bos; + struct ww_acquire_ctx ticket; + struct list_head list; + unsigned domain; + int r; + + INIT_LIST_HEAD(&list); + + tv.bo = &bo_va->bo->tbo; + tv.shared = true; + list_add(&tv.head, &list); + + vm_bos = radeon_vm_get_bos(rdev, bo_va->vm, &list); + if (!vm_bos) + return; + + r = ttm_eu_reserve_buffers(&ticket, &list, true); + if (r) + goto error_free; + + list_for_each_entry(entry, &list, head) { + domain = radeon_mem_type_to_domain(entry->bo->mem.mem_type); + /* if anything is swapped out don't swap it in here, + just abort and wait for the next CS */ + if (domain == RADEON_GEM_DOMAIN_CPU) + goto error_unreserve; + } + + mutex_lock(&bo_va->vm->mutex); + r = radeon_vm_clear_freed(rdev, bo_va->vm); + if (r) + goto error_unlock; + + if (bo_va->it.start) + r = radeon_vm_bo_update(rdev, bo_va, &bo_va->bo->tbo.mem); + +error_unlock: + mutex_unlock(&bo_va->vm->mutex); + +error_unreserve: + ttm_eu_backoff_reservation(&ticket, &list); + +error_free: + drm_free_large(vm_bos); + + if (r) + DRM_ERROR("Couldn't update BO_VA (%d)\n", r); +} + int radeon_gem_va_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { @@ -612,6 +674,8 @@ int radeon_gem_va_ioctl(struct drm_device *dev, void *data, default: break; } + if (!r) + radeon_gem_va_update_vm(rdev, bo_va); args->operation = RADEON_VA_RESULT_OK; if (r) { args->operation = RADEON_VA_RESULT_ERROR; -- cgit v0.10.2 From bbd440fb81338d8e8d58193867f1404c4e6cef7a Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Thu, 20 Nov 2014 22:33:59 +0100 Subject: drm/i915: Don't rely upon encoder->type for infoframe hw state readout encoder->type can change underneath us and doesn't need to reflect actual hw state (since we don't construct it from hw state like e.g. encoder->crtc crtc->config). And this can indeed happen: 1) Boot with plugged-in hdmi screen. Since we only set ->type in the probe functions this means we won't detect any infoframes since type is still unkown. 2) First probe sets type to HDMI. 3) If the first modeset now does _not_ happen on the HDMI pipe with infoframes encoder->get_config suddenly sees infoframes and the state checker gets angry. Fix this by only relying on actual hw state when figuring out whether the ddi port is in hdmi mode and sends infoframes. Cc: Jesse Barnes Cc: Paulo Zanoni Reported-by: Paulo Zanoni Reviewed-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 24110c9..e6b45cd 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2027,6 +2027,7 @@ void intel_ddi_get_config(struct intel_encoder *encoder, struct drm_i915_private *dev_priv = encoder->base.dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); enum transcoder cpu_transcoder = intel_crtc->config.cpu_transcoder; + struct intel_hdmi *intel_hdmi; u32 temp, flags = 0; struct drm_device *dev = dev_priv->dev; @@ -2062,6 +2063,10 @@ void intel_ddi_get_config(struct intel_encoder *encoder, switch (temp & TRANS_DDI_MODE_SELECT_MASK) { case TRANS_DDI_MODE_SELECT_HDMI: pipe_config->has_hdmi_sink = true; + intel_hdmi = enc_to_intel_hdmi(&encoder->base); + + if (intel_hdmi->infoframe_enabled(&encoder->base)) + pipe_config->has_infoframe = true; break; case TRANS_DDI_MODE_SELECT_DVI: case TRANS_DDI_MODE_SELECT_FDI: @@ -2075,14 +2080,6 @@ void intel_ddi_get_config(struct intel_encoder *encoder, break; } - if (encoder->type == INTEL_OUTPUT_HDMI) { - struct intel_hdmi *intel_hdmi = - enc_to_intel_hdmi(&encoder->base); - - if (intel_hdmi->infoframe_enabled(&encoder->base)) - pipe_config->has_infoframe = true; - } - if (intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_AUDIO)) { temp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); if (temp & AUDIO_OUTPUT_ENABLE(intel_crtc->pipe)) -- cgit v0.10.2 From d9e600b2e4a5e9f1dfe80cfcb453c8f5067a2a8a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 20 Nov 2014 20:10:33 +0000 Subject: drm/i915: Only call mod_timer() if not already pending The final arrangement of updating timer->expires and calling mod_timer() used in commit 672e7b7c1849c904b2c55185906b3940843c55c6 Author: Chris Wilson Date: Wed Nov 19 09:47:19 2014 +0000 drm/i915: Don't continually defer the hangcheck turns out to be very unsafe. Try again. Signed-off-by: Chris Wilson Cc: Mika Kuoppala Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 8d169e1..5908580d 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3067,9 +3067,10 @@ void i915_queue_hangcheck(struct drm_device *dev) return; /* Don't continually defer the hangcheck, but make sure it is active */ - if (!timer_pending(timer)) - timer->expires = round_jiffies_up(jiffies + DRM_I915_HANGCHECK_JIFFIES); - mod_timer(timer, timer->expires); + if (timer_pending(timer)) + return; + mod_timer(timer, + round_jiffies_up(jiffies + DRM_I915_HANGCHECK_JIFFIES)); } static void ibx_irq_reset(struct drm_device *dev) -- cgit v0.10.2 From 89a35ecdc6aa5a88165313ca5cfd52b8e8e7fbbd Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Thu, 20 Nov 2014 13:24:13 -0800 Subject: drm/i915/g4x: fix g4x infoframe readout Need to check the port too. Reported-by: Daniel Vetter Signed-off-by: Jesse Barnes Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index f58e883..ec87333 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -170,9 +170,13 @@ static bool g4x_infoframe_enabled(struct drm_encoder *encoder) { struct drm_device *dev = encoder->dev; struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_digital_port *intel_dig_port = enc_to_dig_port(encoder); u32 val = I915_READ(VIDEO_DIP_CTL); - return val & VIDEO_DIP_ENABLE; + if (VIDEO_DIP_PORT(intel_dig_port->port) == (val & VIDEO_DIP_PORT_MASK)) + return val & VIDEO_DIP_ENABLE; + + return false; } static void ibx_write_infoframe(struct drm_encoder *encoder, -- cgit v0.10.2 From 355a70183848f21198e9f6296bd646df3478a26d Mon Sep 17 00:00:00 2001 From: Thomas Hellstrom Date: Thu, 20 Nov 2014 09:56:25 +0100 Subject: drm/gem: Warn on illegal use of the dumb buffer interface v2 It happens on occasion that developers of generic user-space applications abuse the dumb buffer API to get hold of drm buffers that they can both mmap() and use for GPU acceleration, using the assumptions that dumb buffers and buffers available for GPU are a) The same type and can be aribtrarily type-casted. b) fully coherent. This patch makes the most widely used drivers warn nicely when that happens, the next step will be to fail. v2: Move drmP.h changes to drm_gem.h. Fix Radeon dumb mmap breakage. Signed-off-by: Thomas Hellstrom Acked-by: Daniel Vetter Acked-by: Alex Deucher Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 2404b2b..c743908 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1593,7 +1593,7 @@ static struct drm_driver driver = { .gem_prime_import = i915_gem_prime_import, .dumb_create = i915_gem_dumb_create, - .dumb_map_offset = i915_gem_mmap_gtt, + .dumb_map_offset = i915_gem_dumb_map_offset, .dumb_destroy = drm_gem_dumb_destroy, .ioctls = i915_ioctls, .fops = &i915_driver_fops, diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index f830596..4ba1aca 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2523,8 +2523,9 @@ void i915_vma_move_to_active(struct i915_vma *vma, int i915_gem_dumb_create(struct drm_file *file_priv, struct drm_device *dev, struct drm_mode_create_dumb *args); -int i915_gem_mmap_gtt(struct drm_file *file_priv, struct drm_device *dev, - uint32_t handle, uint64_t *offset); +int i915_gem_dumb_map_offset(struct drm_file *file_priv, + struct drm_device *dev, uint32_t handle, + uint64_t *offset); /** * Returns true if seq1 is later than seq2. */ diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 3e0cabe..50b8422 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -346,6 +346,7 @@ static int i915_gem_create(struct drm_file *file, struct drm_device *dev, uint64_t size, + bool dumb, uint32_t *handle_p) { struct drm_i915_gem_object *obj; @@ -361,6 +362,7 @@ i915_gem_create(struct drm_file *file, if (obj == NULL) return -ENOMEM; + obj->base.dumb = dumb; ret = drm_gem_handle_create(file, &obj->base, &handle); /* drop reference from allocate - handle holds it now */ drm_gem_object_unreference_unlocked(&obj->base); @@ -380,7 +382,7 @@ i915_gem_dumb_create(struct drm_file *file, args->pitch = ALIGN(args->width * DIV_ROUND_UP(args->bpp, 8), 64); args->size = args->pitch * args->height; return i915_gem_create(file, dev, - args->size, &args->handle); + args->size, true, &args->handle); } /** @@ -393,7 +395,7 @@ i915_gem_create_ioctl(struct drm_device *dev, void *data, struct drm_i915_gem_create *args = data; return i915_gem_create(file, dev, - args->size, &args->handle); + args->size, false, &args->handle); } static inline int @@ -1773,10 +1775,10 @@ static void i915_gem_object_free_mmap_offset(struct drm_i915_gem_object *obj) drm_gem_free_mmap_offset(&obj->base); } -int +static int i915_gem_mmap_gtt(struct drm_file *file, struct drm_device *dev, - uint32_t handle, + uint32_t handle, bool dumb, uint64_t *offset) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -1793,6 +1795,13 @@ i915_gem_mmap_gtt(struct drm_file *file, goto unlock; } + /* + * We don't allow dumb mmaps on objects created using another + * interface. + */ + WARN_ONCE(dumb && !(obj->base.dumb || obj->base.import_attach), + "Illegal dumb map of accelerated buffer.\n"); + if (obj->base.size > dev_priv->gtt.mappable_end) { ret = -E2BIG; goto out; @@ -1817,6 +1826,15 @@ unlock: return ret; } +int +i915_gem_dumb_map_offset(struct drm_file *file, + struct drm_device *dev, + uint32_t handle, + uint64_t *offset) +{ + return i915_gem_mmap_gtt(file, dev, handle, true, offset); +} + /** * i915_gem_mmap_gtt_ioctl - prepare an object for GTT mmap'ing * @dev: DRM device @@ -1838,7 +1856,7 @@ i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data, { struct drm_i915_gem_mmap_gtt *args = data; - return i915_gem_mmap_gtt(file, dev, args->handle, &args->offset); + return i915_gem_mmap_gtt(file, dev, args->handle, false, &args->offset); } static inline int diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index e1ed85a..2b02fcf 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -121,6 +121,9 @@ eb_lookup_vmas(struct eb_vmas *eb, goto err; } + WARN_ONCE(obj->base.dumb, + "GPU use of dumb buffer is illegal.\n"); + drm_gem_object_reference(&obj->base); list_add_tail(&obj->obj_exec_link, &objects); } diff --git a/drivers/gpu/drm/nouveau/nouveau_display.c b/drivers/gpu/drm/nouveau/nouveau_display.c index a88e692..2640fcf 100644 --- a/drivers/gpu/drm/nouveau/nouveau_display.c +++ b/drivers/gpu/drm/nouveau/nouveau_display.c @@ -871,6 +871,7 @@ nouveau_display_dumb_create(struct drm_file *file_priv, struct drm_device *dev, if (ret) return ret; + bo->gem.dumb = true; ret = drm_gem_handle_create(file_priv, &bo->gem, &args->handle); drm_gem_object_unreference_unlocked(&bo->gem); return ret; @@ -886,6 +887,14 @@ nouveau_display_dumb_map_offset(struct drm_file *file_priv, gem = drm_gem_object_lookup(dev, file_priv, handle); if (gem) { struct nouveau_bo *bo = nouveau_gem_object(gem); + + /* + * We don't allow dumb mmaps on objects created using another + * interface. + */ + WARN_ONCE(!(gem->dumb || gem->import_attach), + "Illegal dumb map of accelerated buffer.\n"); + *poffset = drm_vma_node_offset_addr(&bo->bo.vma_node); drm_gem_object_unreference_unlocked(gem); return 0; diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c index 36951ee..ebba9de 100644 --- a/drivers/gpu/drm/nouveau/nouveau_gem.c +++ b/drivers/gpu/drm/nouveau/nouveau_gem.c @@ -444,6 +444,9 @@ validate_list(struct nouveau_channel *chan, struct nouveau_cli *cli, list_for_each_entry(nvbo, list, entry) { struct drm_nouveau_gem_pushbuf_bo *b = &pbbo[nvbo->pbbo_index]; + WARN_ONCE(nvbo->gem.dumb, + "GPU use of dumb buffer is illegal.\n"); + ret = nouveau_gem_set_domain(&nvbo->gem, b->read_domains, b->write_domains, b->valid_domains); diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index c194497..429213b 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -394,9 +394,10 @@ int radeon_gem_set_domain_ioctl(struct drm_device *dev, void *data, return r; } -int radeon_mode_dumb_mmap(struct drm_file *filp, - struct drm_device *dev, - uint32_t handle, uint64_t *offset_p) +static int radeon_mode_mmap(struct drm_file *filp, + struct drm_device *dev, + uint32_t handle, bool dumb, + uint64_t *offset_p) { struct drm_gem_object *gobj; struct radeon_bo *robj; @@ -405,6 +406,14 @@ int radeon_mode_dumb_mmap(struct drm_file *filp, if (gobj == NULL) { return -ENOENT; } + + /* + * We don't allow dumb mmaps on objects created using another + * interface. + */ + WARN_ONCE(dumb && !(gobj->dumb || gobj->import_attach), + "Illegal dumb map of GPU buffer.\n"); + robj = gem_to_radeon_bo(gobj); if (radeon_ttm_tt_has_userptr(robj->tbo.ttm)) { drm_gem_object_unreference_unlocked(gobj); @@ -415,12 +424,20 @@ int radeon_mode_dumb_mmap(struct drm_file *filp, return 0; } +int radeon_mode_dumb_mmap(struct drm_file *filp, + struct drm_device *dev, + uint32_t handle, uint64_t *offset_p) +{ + return radeon_mode_mmap(filp, dev, handle, true, offset_p); +} + int radeon_gem_mmap_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { struct drm_radeon_gem_mmap *args = data; - return radeon_mode_dumb_mmap(filp, dev, args->handle, &args->addr_ptr); + return radeon_mode_mmap(filp, dev, args->handle, false, + &args->addr_ptr); } int radeon_gem_busy_ioctl(struct drm_device *dev, void *data, @@ -682,6 +699,7 @@ int radeon_mode_dumb_create(struct drm_file *file_priv, return -ENOMEM; r = drm_gem_handle_create(file_priv, gobj, &handle); + gobj->dumb = true; /* drop reference from allocate - handle holds it now */ drm_gem_object_unreference_unlocked(gobj); if (r) { diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index 33e6c7a..76eedd6 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -521,6 +521,9 @@ int radeon_bo_list_validate(struct radeon_device *rdev, u32 current_domain = radeon_mem_type_to_domain(bo->tbo.mem.mem_type); + WARN_ONCE(bo->gem_base.dumb, + "GPU use of dumb buffer is illegal.\n"); + /* Check if this buffer will be moved and don't move it * if we have moved too many buffers for this IB already. * diff --git a/include/drm/drm_gem.h b/include/drm/drm_gem.h index 1e6ae14..780511a 100644 --- a/include/drm/drm_gem.h +++ b/include/drm/drm_gem.h @@ -119,6 +119,13 @@ struct drm_gem_object { * simply leave it as NULL. */ struct dma_buf_attachment *import_attach; + + /** + * dumb - created as dumb buffer + * Whether the gem object was created using the dumb buffer interface + * as such it may not be used for GPU rendering. + */ + bool dumb; }; void drm_gem_object_release(struct drm_gem_object *obj); -- cgit v0.10.2 From fc2d2bc1f1f8d8351f884bb65b5cc4ad2411b2a7 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Thu, 20 Nov 2014 09:53:35 +0100 Subject: drm/atomic: Add missing ERR_PTR casting This is an oversight from commit f52b69f1ecfdd7ef6867a257620258c09e569552 Author: Daniel Vetter Date: Wed Nov 19 18:38:08 2014 +0100 drm/atomic: Don't overrun the connector array when hotplugging Cc: Dave Airlie Cc: Rob Clark Signed-off-by: Daniel Vetter Reviewed-by: Thierry Reding Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index 3624632..d3b4674 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -318,7 +318,7 @@ drm_atomic_get_connector_state(struct drm_atomic_state *state, */ if (index >= state->num_connector) { DRM_DEBUG_KMS("Hot-added connector would overflow state array, restarting\n"); - return -EAGAIN; + return ERR_PTR(-EAGAIN); } if (state->connector_states[index]) -- cgit v0.10.2 From 9084acf19bb98afb0ed90337fd39541103fe3494 Mon Sep 17 00:00:00 2001 From: Markus Elfring Date: Wed, 19 Nov 2014 16:33:17 +0100 Subject: drm/edid: Deletion of an unnecessary check before the function call "release_firmware" The release_firmware() function tests whether its argument is NULL and then returns immediately. Thus the test around the call is not needed. This issue was detected by using the Coccinelle software. Signed-off-by: Markus Elfring Reviewed-by: Thierry Reding Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_edid_load.c b/drivers/gpu/drm/drm_edid_load.c index 0a235fe..732cb6f 100644 --- a/drivers/gpu/drm/drm_edid_load.c +++ b/drivers/gpu/drm/drm_edid_load.c @@ -254,8 +254,7 @@ static void *edid_load(struct drm_connector *connector, const char *name, name, connector_name); out: - if (fw) - release_firmware(fw); + release_firmware(fw); return edid; } -- cgit v0.10.2 From c08448173545428ed9761308fbaf90f8be7498c5 Mon Sep 17 00:00:00 2001 From: Markus Elfring Date: Wed, 19 Nov 2014 17:05:20 +0100 Subject: drm/tilcdc: Deletion of an unnecessary check before the function call "drm_fbdev_cma_hotplug_event" The drm_fbdev_cma_hotplug_event() function tests whether its argument is NULL and then returns immediately. Thus the test around the call is not needed. This issue was detected by using the Coccinelle software. Signed-off-by: Markus Elfring Reviewed-by: Thierry Reding Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/tilcdc/tilcdc_drv.c b/drivers/gpu/drm/tilcdc/tilcdc_drv.c index 79a34cb..d56d3f8 100644 --- a/drivers/gpu/drm/tilcdc/tilcdc_drv.c +++ b/drivers/gpu/drm/tilcdc/tilcdc_drv.c @@ -58,8 +58,7 @@ static struct drm_framebuffer *tilcdc_fb_create(struct drm_device *dev, static void tilcdc_fb_output_poll_changed(struct drm_device *dev) { struct tilcdc_drm_private *priv = dev->dev_private; - if (priv->fbdev) - drm_fbdev_cma_hotplug_event(priv->fbdev); + drm_fbdev_cma_hotplug_event(priv->fbdev); } static const struct drm_mode_config_funcs mode_config_funcs = { -- cgit v0.10.2 From e9196d238597da0189b035f37c473c411275b0dd Mon Sep 17 00:00:00 2001 From: Markus Elfring Date: Wed, 19 Nov 2014 17:33:32 +0100 Subject: drm/udl: Deletion of an unnecessary check before the function call "vunmap" The vunmap() function performes also input parameter validation. Thus the test around the call is not needed. This issue was detected by using the Coccinelle software. Signed-off-by: Markus Elfring Reviewed-by: Thierry Reding Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/udl/udl_gem.c b/drivers/gpu/drm/udl/udl_gem.c index 692d6f2..cd3482d 100644 --- a/drivers/gpu/drm/udl/udl_gem.c +++ b/drivers/gpu/drm/udl/udl_gem.c @@ -184,8 +184,7 @@ void udl_gem_vunmap(struct udl_gem_object *obj) return; } - if (obj->vmapping) - vunmap(obj->vmapping); + vunmap(obj->vmapping); udl_gem_put_pages(obj); } -- cgit v0.10.2 From a3a1a6674f7fd6c0f7e673c4fc1b1aea3d8fce53 Mon Sep 17 00:00:00 2001 From: Markus Elfring Date: Wed, 19 Nov 2014 17:50:19 +0100 Subject: drm/vmwgfx: Deletion of an unnecessary check before the function call "vfree" The vfree() function performes also input parameter validation. Thus the test around the call is not needed. This issue was detected by using the Coccinelle software. Signed-off-by: Markus Elfring Reviewed-by: Thierry Reding Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c index 25f3c25..db76218 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c @@ -889,8 +889,7 @@ static int vmw_driver_unload(struct drm_device *dev) if (dev_priv->ctx.res_ht_initialized) drm_ht_remove(&dev_priv->ctx.res_ht); - if (dev_priv->ctx.cmd_bounce) - vfree(dev_priv->ctx.cmd_bounce); + vfree(dev_priv->ctx.cmd_bounce); if (dev_priv->enable_fb) { vmw_fb_close(dev_priv); vmw_kms_restore_vga(dev_priv); -- cgit v0.10.2 From e7f1d0b735fd4bbec225a83de2aa8f52a0a2e95c Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 21 Nov 2014 10:37:14 +0100 Subject: drm/i915: Update DRIVER_DATE to 20141121 Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 5448ce9..b6d8d2d 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -55,7 +55,7 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20141107" +#define DRIVER_DATE "20141121" #undef WARN_ON #define WARN_ON(x) WARN(x, "WARN_ON(" #x ")") -- cgit v0.10.2 From f6a8eaca0ea10fc5c5ae0d6b0067759164e633a0 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Mon, 17 Nov 2014 15:28:07 -0500 Subject: drm/msm/mdp5: use irqdomains For mdp5, the irqs of hdmi/eDP/dsi0/dsi1 blocks get routed through the mdp block. In order to decouple hdmi/eDP/etc, register an irq domain in mdp5. When hdmi/dsi/etc are used with mdp4, they can directly setup their irqs in their DT nodes as normal. When used with mdp5, instead set the mdp device as the interrupt-parent, as in: mdp: qcom,mdss_mdp@fd900000 { compatible = "qcom,mdss_mdp"; interrupt-controller; #interrupt-cells = <1>; ... }; hdmi: qcom,hdmi_tx@fd922100 { compatible = "qcom,hdmi-tx-8074"; interrupt-parent = <&mdp>; interrupts = <8 0>; /* MDP5_HW_INTR_STATUS.INTR_HDMI */ ... }; There is a slight awkwardness, in that we cannot disable child irqs at the mdp level, they can only be cleared in the child block. So you must not use threaded irq handlers in the child. I'm not sure if there is a better way to deal with that. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.c b/drivers/gpu/drm/msm/hdmi/hdmi.c index 9007761..db8c3b4 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi.c +++ b/drivers/gpu/drm/msm/hdmi/hdmi.c @@ -15,6 +15,7 @@ * this program. If not, see . */ +#include #include "hdmi.h" void hdmi_set_mode(struct hdmi *hdmi, bool power_on) @@ -39,7 +40,7 @@ void hdmi_set_mode(struct hdmi *hdmi, bool power_on) power_on ? "Enable" : "Disable", ctrl); } -irqreturn_t hdmi_irq(int irq, void *dev_id) +static irqreturn_t hdmi_irq(int irq, void *dev_id) { struct hdmi *hdmi = dev_id; @@ -200,7 +201,6 @@ int hdmi_modeset_init(struct hdmi *hdmi, { struct msm_drm_private *priv = dev->dev_private; struct platform_device *pdev = hdmi->pdev; - struct hdmi_platform_config *config = pdev->dev.platform_data; int ret; hdmi->dev = dev; @@ -224,22 +224,20 @@ int hdmi_modeset_init(struct hdmi *hdmi, goto fail; } - if (!config->shared_irq) { - hdmi->irq = platform_get_irq(pdev, 0); - if (hdmi->irq < 0) { - ret = hdmi->irq; - dev_err(dev->dev, "failed to get irq: %d\n", ret); - goto fail; - } + hdmi->irq = irq_of_parse_and_map(pdev->dev.of_node, 0); + if (hdmi->irq < 0) { + ret = hdmi->irq; + dev_err(dev->dev, "failed to get irq: %d\n", ret); + goto fail; + } - ret = devm_request_threaded_irq(&pdev->dev, hdmi->irq, - NULL, hdmi_irq, IRQF_TRIGGER_HIGH | IRQF_ONESHOT, - "hdmi_isr", hdmi); - if (ret < 0) { - dev_err(dev->dev, "failed to request IRQ%u: %d\n", - hdmi->irq, ret); - goto fail; - } + ret = devm_request_irq(&pdev->dev, hdmi->irq, + hdmi_irq, IRQF_TRIGGER_HIGH | IRQF_ONESHOT, + "hdmi_isr", hdmi); + if (ret < 0) { + dev_err(dev->dev, "failed to request IRQ%u: %d\n", + hdmi->irq, ret); + goto fail; } encoder->bridge = hdmi->bridge; @@ -318,7 +316,6 @@ static int hdmi_bind(struct device *dev, struct device *master, void *data) config.hpd_clk_cnt = ARRAY_SIZE(hpd_clk_names); config.pwr_clk_names = pwr_clk_names; config.pwr_clk_cnt = ARRAY_SIZE(pwr_clk_names); - config.shared_irq = true; } else if (of_device_is_compatible(of_node, "qcom,hdmi-tx-8960")) { static const char *hpd_clk_names[] = {"core_clk", "master_iface_clk", "slave_iface_clk"}; static const char *hpd_reg_names[] = {"core-vdda", "hdmi-mux"}; diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.h b/drivers/gpu/drm/msm/hdmi/hdmi.h index b981995..0a6f538 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi.h +++ b/drivers/gpu/drm/msm/hdmi/hdmi.h @@ -97,9 +97,6 @@ struct hdmi_platform_config { /* gpio's: */ int ddc_clk_gpio, ddc_data_gpio, hpd_gpio, mux_en_gpio, mux_sel_gpio; int mux_lpm_gpio; - - /* older devices had their own irq, mdp5+ it is shared w/ mdp: */ - bool shared_irq; }; void hdmi_set_mode(struct hdmi *hdmi, bool power_on); diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c index 812c59b..70ac81e 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c @@ -15,6 +15,8 @@ * this program. If not, see . */ +#include +#include #include "msm_drv.h" #include "mdp5_kms.h" @@ -82,18 +84,23 @@ irqreturn_t mdp5_irq(struct msm_kms *kms) { struct mdp_kms *mdp_kms = to_mdp_kms(kms); struct mdp5_kms *mdp5_kms = to_mdp5_kms(mdp_kms); - struct msm_drm_private *priv = mdp5_kms->dev->dev_private; uint32_t intr; intr = mdp5_read(mdp5_kms, REG_MDP5_HW_INTR_STATUS); VERB("intr=%08x", intr); - if (intr & MDP5_HW_INTR_STATUS_INTR_MDP) + if (intr & MDP5_HW_INTR_STATUS_INTR_MDP) { mdp5_irq_mdp(mdp_kms); + intr &= ~MDP5_HW_INTR_STATUS_INTR_MDP; + } - if (intr & MDP5_HW_INTR_STATUS_INTR_HDMI) - hdmi_irq(0, priv->hdmi); + while (intr) { + irq_hw_number_t hwirq = fls(intr) - 1; + generic_handle_irq(irq_find_mapping( + mdp5_kms->irqcontroller.domain, hwirq)); + intr &= ~(1 << hwirq); + } return IRQ_HANDLED; } @@ -110,3 +117,82 @@ void mdp5_disable_vblank(struct msm_kms *kms, struct drm_crtc *crtc) mdp_update_vblank_mask(to_mdp_kms(kms), mdp5_crtc_vblank(crtc), false); } + +/* + * interrupt-controller implementation, so sub-blocks (hdmi/eDP/dsi/etc) + * can register to get their irq's delivered + */ + +#define VALID_IRQS (MDP5_HW_INTR_STATUS_INTR_DSI0 | \ + MDP5_HW_INTR_STATUS_INTR_DSI1 | \ + MDP5_HW_INTR_STATUS_INTR_HDMI | \ + MDP5_HW_INTR_STATUS_INTR_EDP) + +static void mdp5_hw_mask_irq(struct irq_data *irqd) +{ + struct mdp5_kms *mdp5_kms = irq_data_get_irq_chip_data(irqd); + smp_mb__before_atomic(); + clear_bit(irqd->hwirq, &mdp5_kms->irqcontroller.enabled_mask); + smp_mb__after_atomic(); +} + +static void mdp5_hw_unmask_irq(struct irq_data *irqd) +{ + struct mdp5_kms *mdp5_kms = irq_data_get_irq_chip_data(irqd); + smp_mb__before_atomic(); + set_bit(irqd->hwirq, &mdp5_kms->irqcontroller.enabled_mask); + smp_mb__after_atomic(); +} + +static struct irq_chip mdp5_hw_irq_chip = { + .name = "mdp5", + .irq_mask = mdp5_hw_mask_irq, + .irq_unmask = mdp5_hw_unmask_irq, +}; + +static int mdp5_hw_irqdomain_map(struct irq_domain *d, + unsigned int irq, irq_hw_number_t hwirq) +{ + struct mdp5_kms *mdp5_kms = d->host_data; + + if (!(VALID_IRQS & (1 << hwirq))) + return -EPERM; + + irq_set_chip_and_handler(irq, &mdp5_hw_irq_chip, handle_level_irq); + irq_set_chip_data(irq, mdp5_kms); + set_irq_flags(irq, IRQF_VALID); + + return 0; +} + +static struct irq_domain_ops mdp5_hw_irqdomain_ops = { + .map = mdp5_hw_irqdomain_map, + .xlate = irq_domain_xlate_onecell, +}; + + +int mdp5_irq_domain_init(struct mdp5_kms *mdp5_kms) +{ + struct device *dev = mdp5_kms->dev->dev; + struct irq_domain *d; + + d = irq_domain_add_linear(dev->of_node, 32, + &mdp5_hw_irqdomain_ops, mdp5_kms); + if (!d) { + dev_err(dev, "mdp5 irq domain add failed\n"); + return -ENXIO; + } + + mdp5_kms->irqcontroller.enabled_mask = 0; + mdp5_kms->irqcontroller.domain = d; + + return 0; +} + +void mdp5_irq_domain_fini(struct mdp5_kms *mdp5_kms) +{ + if (mdp5_kms->irqcontroller.domain) { + irq_domain_remove(mdp5_kms->irqcontroller.domain); + mdp5_kms->irqcontroller.domain = NULL; + } +} diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c index ce03081..6c414db 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c @@ -221,10 +221,13 @@ static void mdp5_destroy(struct msm_kms *kms) struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms)); struct msm_mmu *mmu = mdp5_kms->mmu; + mdp5_irq_domain_fini(mdp5_kms); + if (mmu) { mmu->funcs->detach(mmu, iommu_ports, ARRAY_SIZE(iommu_ports)); mmu->funcs->destroy(mmu); } + kfree(mdp5_kms); } @@ -279,6 +282,13 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) struct drm_encoder *encoder; int i, ret; + /* register our interrupt-controller for hdmi/eDP/dsi/etc + * to use for irqs routed through mdp: + */ + ret = mdp5_irq_domain_init(mdp5_kms); + if (ret) + goto fail; + /* construct CRTCs: */ for (i = 0; i < mdp5_kms->hw_cfg->pipe_rgb.count; i++) { struct drm_plane *plane; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h index c91101d..0e9e3f7 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h @@ -72,6 +72,11 @@ struct mdp5_kms { struct clk *vsync_clk; struct mdp_irq error_handler; + + struct { + volatile unsigned long enabled_mask; + struct irq_domain *domain; + } irqcontroller; }; #define to_mdp5_kms(x) container_of(x, struct mdp5_kms, base) @@ -195,6 +200,8 @@ void mdp5_irq_uninstall(struct msm_kms *kms); irqreturn_t mdp5_irq(struct msm_kms *kms); int mdp5_enable_vblank(struct msm_kms *kms, struct drm_crtc *crtc); void mdp5_disable_vblank(struct msm_kms *kms, struct drm_crtc *crtc); +int mdp5_irq_domain_init(struct mdp5_kms *mdp5_kms); +void mdp5_irq_domain_fini(struct mdp5_kms *mdp5_kms); static inline uint32_t mdp5_get_formats(enum mdp5_pipe pipe, uint32_t *pixel_formats, diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h index 4b52d75..1363038 100644 --- a/drivers/gpu/drm/msm/msm_drv.h +++ b/drivers/gpu/drm/msm/msm_drv.h @@ -215,7 +215,6 @@ struct drm_fb_helper *msm_fbdev_init(struct drm_device *dev); struct hdmi; int hdmi_modeset_init(struct hdmi *hdmi, struct drm_device *dev, struct drm_encoder *encoder); -irqreturn_t hdmi_irq(int irq, void *dev_id); void __init hdmi_register(void); void __exit hdmi_unregister(void); -- cgit v0.10.2 From 3f307963fc07953b56b51b18ed47416bf340320f Mon Sep 17 00:00:00 2001 From: Stephane Viau Date: Mon, 17 Nov 2014 13:39:34 -0500 Subject: drm/msm/mdp5: get the core clock rate from MDP5 config The core clock rate depends on the hw configuration. Once we have read the hardware revision, we can set the core clock to its maximum value. Before then, the clock is set at a rate supported by all MDP5 revisions. Signed-off-by: Stephane Viau Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c index 6c414db..1cb91be 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c @@ -62,6 +62,7 @@ static const struct mdp5_config msm8x74_config = { .count = 4, .base = { 0x12500, 0x12700, 0x12900, 0x12b00 }, }, + .max_clk = 200000000, }; static const struct mdp5_config apq8084_config = { @@ -99,6 +100,7 @@ static const struct mdp5_config apq8084_config = { .count = 5, .base = { 0x12500, 0x12700, 0x12900, 0x12b00, 0x12d00 }, }, + .max_clk = 320000000, }; struct mdp5_config_entry { @@ -427,12 +429,13 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) if (ret) goto fail; - ret = clk_set_rate(mdp5_kms->src_clk, config->max_clk); - ret = mdp5_select_hw_cfg(kms); if (ret) goto fail; + /* TODO: compute core clock rate at runtime */ + clk_set_rate(mdp5_kms->src_clk, mdp5_kms->hw_cfg->max_clk); + /* make sure things are off before attaching iommu (bootloader could * have left things on, in which case we'll start getting faults if * we don't disable): @@ -493,8 +496,6 @@ static struct mdp5_platform_config *mdp5_get_config(struct platform_device *dev) /* TODO */ #endif config.iommu = iommu_domain_alloc(&platform_bus_type); - /* TODO hard-coded in downstream mdss, but should it be? */ - config.max_clk = 200000000; /* TODO get from DT: */ config.smp_blk_cnt = 22; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h index 0e9e3f7..70aa8d4 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h @@ -37,6 +37,8 @@ struct mdp5_config { struct mdp5_sub_block dspp; struct mdp5_sub_block ad; struct mdp5_sub_block intf; + + uint32_t max_clk; }; extern const struct mdp5_config *mdp5_cfg; #include "mdp5.xml.h" @@ -83,7 +85,6 @@ struct mdp5_kms { /* platform config data (ie. from DT, or pdata) */ struct mdp5_platform_config { struct iommu_domain *iommu; - uint32_t max_clk; int smp_blk_cnt; }; -- cgit v0.10.2 From d1a717bd2719d6416fdc5e9b65786f17c6c3facb Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Tue, 18 Nov 2014 08:40:44 -0500 Subject: drm/msm/hdmi: remove useless kref A left-over from prior to component framework. The original intent was to deal with hdmi getting unloaded before the master component, but that isn't really going to work anyways. These days with the component framework taking care to unload the master component first, we don't have to worry about this. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.c b/drivers/gpu/drm/msm/hdmi/hdmi.c index db8c3b4..062c687 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi.c +++ b/drivers/gpu/drm/msm/hdmi/hdmi.c @@ -55,9 +55,8 @@ static irqreturn_t hdmi_irq(int irq, void *dev_id) return IRQ_HANDLED; } -void hdmi_destroy(struct kref *kref) +static void hdmi_destroy(struct hdmi *hdmi) { - struct hdmi *hdmi = container_of(kref, struct hdmi, refcount); struct hdmi_phy *phy = hdmi->phy; if (phy) @@ -85,8 +84,6 @@ static struct hdmi *hdmi_init(struct platform_device *pdev) goto fail; } - kref_init(&hdmi->refcount); - hdmi->pdev = pdev; hdmi->config = config; @@ -183,7 +180,7 @@ static struct hdmi *hdmi_init(struct platform_device *pdev) fail: if (hdmi) - hdmi_destroy(&hdmi->refcount); + hdmi_destroy(hdmi); return ERR_PTR(ret); } @@ -269,12 +266,6 @@ fail: #include -static void set_hdmi(struct drm_device *dev, struct hdmi *hdmi) -{ - struct msm_drm_private *priv = dev->dev_private; - priv->hdmi = hdmi; -} - #ifdef CONFIG_OF static int get_gpio(struct device *dev, struct device_node *of_node, const char *name) { @@ -295,6 +286,8 @@ static int get_gpio(struct device *dev, struct device_node *of_node, const char static int hdmi_bind(struct device *dev, struct device *master, void *data) { + struct drm_device *drm = dev_get_drvdata(master); + struct msm_drm_private *priv = drm->dev_private; static struct hdmi_platform_config config = {}; struct hdmi *hdmi; #ifdef CONFIG_OF @@ -389,14 +382,19 @@ static int hdmi_bind(struct device *dev, struct device *master, void *data) hdmi = hdmi_init(to_platform_device(dev)); if (IS_ERR(hdmi)) return PTR_ERR(hdmi); - set_hdmi(dev_get_drvdata(master), hdmi); + priv->hdmi = hdmi; return 0; } static void hdmi_unbind(struct device *dev, struct device *master, void *data) { - set_hdmi(dev_get_drvdata(master), NULL); + struct drm_device *drm = dev_get_drvdata(master); + struct msm_drm_private *priv = drm->dev_private; + if (priv->hdmi) { + hdmi_destroy(priv->hdmi); + priv->hdmi = NULL; + } } static const struct component_ops hdmi_ops = { diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.h b/drivers/gpu/drm/msm/hdmi/hdmi.h index 0a6f538..43e654f 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi.h +++ b/drivers/gpu/drm/msm/hdmi/hdmi.h @@ -38,8 +38,6 @@ struct hdmi_audio { }; struct hdmi { - struct kref refcount; - struct drm_device *dev; struct platform_device *pdev; @@ -100,7 +98,6 @@ struct hdmi_platform_config { }; void hdmi_set_mode(struct hdmi *hdmi, bool power_on); -void hdmi_destroy(struct kref *kref); static inline void hdmi_write(struct hdmi *hdmi, u32 reg, u32 data) { @@ -112,17 +109,6 @@ static inline u32 hdmi_read(struct hdmi *hdmi, u32 reg) return msm_readl(hdmi->mmio + reg); } -static inline struct hdmi * hdmi_reference(struct hdmi *hdmi) -{ - kref_get(&hdmi->refcount); - return hdmi; -} - -static inline void hdmi_unreference(struct hdmi *hdmi) -{ - kref_put(&hdmi->refcount, hdmi_destroy); -} - /* * The phy appears to be different, for example between 8960 and 8x60, * so split the phy related functions out and load the correct one at diff --git a/drivers/gpu/drm/msm/hdmi/hdmi_bridge.c b/drivers/gpu/drm/msm/hdmi/hdmi_bridge.c index f6cf745..6902ad6 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi_bridge.c +++ b/drivers/gpu/drm/msm/hdmi/hdmi_bridge.c @@ -26,7 +26,6 @@ struct hdmi_bridge { static void hdmi_bridge_destroy(struct drm_bridge *bridge) { struct hdmi_bridge *hdmi_bridge = to_hdmi_bridge(bridge); - hdmi_unreference(hdmi_bridge->hdmi); drm_bridge_cleanup(bridge); kfree(hdmi_bridge); } @@ -218,7 +217,7 @@ struct drm_bridge *hdmi_bridge_init(struct hdmi *hdmi) goto fail; } - hdmi_bridge->hdmi = hdmi_reference(hdmi); + hdmi_bridge->hdmi = hdmi; bridge = &hdmi_bridge->base; diff --git a/drivers/gpu/drm/msm/hdmi/hdmi_connector.c b/drivers/gpu/drm/msm/hdmi/hdmi_connector.c index 0aecb25..fbebb04 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi_connector.c +++ b/drivers/gpu/drm/msm/hdmi/hdmi_connector.c @@ -330,8 +330,6 @@ static void hdmi_connector_destroy(struct drm_connector *connector) drm_connector_unregister(connector); drm_connector_cleanup(connector); - hdmi_unreference(hdmi_connector->hdmi); - kfree(hdmi_connector); } @@ -425,7 +423,7 @@ struct drm_connector *hdmi_connector_init(struct hdmi *hdmi) goto fail; } - hdmi_connector->hdmi = hdmi_reference(hdmi); + hdmi_connector->hdmi = hdmi; INIT_WORK(&hdmi_connector->hpd_work, hotplug_work); connector = &hdmi_connector->base; -- cgit v0.10.2 From bfcdfb0e62639732339c85371a8c07e915bf0941 Mon Sep 17 00:00:00 2001 From: Stephane Viau Date: Tue, 18 Nov 2014 12:49:47 -0500 Subject: drm/msm/mdp5: make SMP module dynamically configurable The Shared Memory Pool (SMP) has its own limitation, features and state. Some examples are: - the number of Memory Macro Block (MMB) and their size - the number of lines that can be fetched - the state of MMB currently allocated - the computation of number of blocks required per plane - client IDs ... In order to avoid private data to be overwritten by other modules, let's make these private to the SMP module. Some of these depend on the hardware configuration, let's add them to the mdp5_config struct. In some hw configurations, some MMBs are statically tied to RGB pipes and cannot be re-allocated dynamically. This change introduces the concept of MMB static usage and makes sure that dynamic MMB requests are dimensioned accordingly. A note on passing a pipe pointer, instead of client IDs: Client IDs are SMP-related information. Passing PIPE information to SMP lets SMP module to find out which SMP client(s) are used. This allows the SMP module to access the PIPE pointer, which can be used for FIFO watermark configuration. By the way, even though REG_MDP5_PIPE_REQPRIO_FIFO_WM_* registers are part of the PIPE registers, their functionality is to reflect the behavior of the SMP block. These registers access is now restricted to the SMP module. Signed-off-by: Stephane Viau Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c index 1cb91be..f852fa4 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c @@ -30,6 +30,10 @@ const struct mdp5_config *mdp5_cfg; static const struct mdp5_config msm8x74_config = { .name = "msm8x74", + .smp = { + .mmb_count = 22, + .mmb_size = 4096, + }, .ctl = { .count = 5, .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 }, @@ -67,6 +71,15 @@ static const struct mdp5_config msm8x74_config = { static const struct mdp5_config apq8084_config = { .name = "apq8084", + .smp = { + .mmb_count = 44, + .mmb_size = 8192, + .reserved_state[0] = GENMASK(7, 0), /* first 8 MMBs */ + .reserved[CID_RGB0] = 2, + .reserved[CID_RGB1] = 2, + .reserved[CID_RGB2] = 2, + .reserved[CID_RGB3] = 2, + }, .ctl = { .count = 5, .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 }, @@ -222,6 +235,7 @@ static void mdp5_destroy(struct msm_kms *kms) { struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms)); struct msm_mmu *mmu = mdp5_kms->mmu; + void *smp = mdp5_kms->smp_priv; mdp5_irq_domain_fini(mdp5_kms); @@ -230,6 +244,9 @@ static void mdp5_destroy(struct msm_kms *kms) mmu->funcs->destroy(mmu); } + if (smp) + mdp5_smp_destroy(smp); + kfree(mdp5_kms); } @@ -370,6 +387,7 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) struct mdp5_kms *mdp5_kms; struct msm_kms *kms = NULL; struct msm_mmu *mmu; + void *priv; int i, ret; mdp5_kms = kzalloc(sizeof(*mdp5_kms), GFP_KERNEL); @@ -384,7 +402,6 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) kms = &mdp5_kms->base.base; mdp5_kms->dev = dev; - mdp5_kms->smp_blk_cnt = config->smp_blk_cnt; mdp5_kms->mmio = msm_ioremap(pdev, "mdp_phys", "MDP5"); if (IS_ERR(mdp5_kms->mmio)) { @@ -436,6 +453,13 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) /* TODO: compute core clock rate at runtime */ clk_set_rate(mdp5_kms->src_clk, mdp5_kms->hw_cfg->max_clk); + priv = mdp5_smp_init(mdp5_kms->dev, &mdp5_kms->hw_cfg->smp); + if (IS_ERR(priv)) { + ret = PTR_ERR(priv); + goto fail; + } + mdp5_kms->smp_priv = priv; + /* make sure things are off before attaching iommu (bootloader could * have left things on, in which case we'll start getting faults if * we don't disable): @@ -496,8 +520,6 @@ static struct mdp5_platform_config *mdp5_get_config(struct platform_device *dev) /* TODO */ #endif config.iommu = iommu_domain_alloc(&platform_bus_type); - /* TODO get from DT: */ - config.smp_blk_cnt = 22; return &config; } diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h index 70aa8d4..3f7aa49 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h @@ -23,12 +23,22 @@ #include "mdp/mdp_kms.h" /* dynamic offsets used by mdp5.xml.h (initialized in mdp5_kms.c) */ #define MDP5_MAX_BASES 8 +#define MAX_SMP_BLOCKS 44 +#define MAX_CLIENTS 32 +typedef DECLARE_BITMAP(mdp5_smp_state_t, MAX_SMP_BLOCKS); struct mdp5_sub_block { int count; uint32_t base[MDP5_MAX_BASES]; }; +struct mdp5_smp_block { + int mmb_count; /* number of SMP MMBs */ + int mmb_size; /* MMB: size in bytes */ + mdp5_smp_state_t reserved_state;/* SMP MMBs statically allocated */ + int reserved[MAX_CLIENTS]; /* # of MMBs reserved per client */ +}; struct mdp5_config { char *name; + struct mdp5_smp_block smp; struct mdp5_sub_block ctl; struct mdp5_sub_block pipe_vig; struct mdp5_sub_block pipe_rgb; @@ -56,10 +66,7 @@ struct mdp5_kms { int id; struct msm_mmu *mmu; - /* for tracking smp allocation amongst pipes: */ - mdp5_smp_state_t smp_state; - struct mdp5_client_smp_state smp_client_state[CID_MAX]; - int smp_blk_cnt; + void *smp_priv; /* io/register spaces: */ void __iomem *mmio, *vbif; @@ -85,7 +92,6 @@ struct mdp5_kms { /* platform config data (ie. from DT, or pdata) */ struct mdp5_platform_config { struct iommu_domain *iommu; - int smp_blk_cnt; }; static inline void mdp5_write(struct mdp5_kms *mdp5_kms, u32 reg, u32 data) @@ -141,24 +147,6 @@ static inline int pipe2nclients(enum mdp5_pipe pipe) } } -static inline enum mdp5_client_id pipe2client(enum mdp5_pipe pipe, int plane) -{ - WARN_ON(plane >= pipe2nclients(pipe)); - switch (pipe) { - case SSPP_VIG0: return CID_VIG0_Y + plane; - case SSPP_VIG1: return CID_VIG1_Y + plane; - case SSPP_VIG2: return CID_VIG2_Y + plane; - case SSPP_RGB0: return CID_RGB0; - case SSPP_RGB1: return CID_RGB1; - case SSPP_RGB2: return CID_RGB2; - case SSPP_DMA0: return CID_DMA0_Y + plane; - case SSPP_DMA1: return CID_DMA1_Y + plane; - case SSPP_VIG3: return CID_VIG3_Y + plane; - case SSPP_RGB3: return CID_RGB3; - default: return CID_UNUSED; - } -} - static inline uint32_t mixer2flush(int lm) { switch (lm) { diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c index f3daec4..633ca08 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c @@ -63,13 +63,13 @@ static int mdp5_plane_disable(struct drm_plane *plane) struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); struct mdp5_kms *mdp5_kms = get_kms(plane); enum mdp5_pipe pipe = mdp5_plane->pipe; - int i; DBG("%s: disable", mdp5_plane->name); - /* update our SMP request to zero (release all our blks): */ - for (i = 0; i < pipe2nclients(pipe); i++) - mdp5_smp_request(mdp5_kms, pipe2client(pipe, i), 0); + if (mdp5_kms) { + /* Release the memory we requested earlier from the SMP: */ + mdp5_smp_release(mdp5_kms->smp_priv, pipe); + } /* TODO detaching now will cause us not to get the last * vblank and mdp5_smp_commit().. so other planes will @@ -149,68 +149,6 @@ void mdp5_plane_set_scanout(struct drm_plane *plane, plane->fb = fb; } -/* NOTE: looks like if horizontal decimation is used (if we supported that) - * then the width used to calculate SMP block requirements is the post- - * decimated width. Ie. SMP buffering sits downstream of decimation (which - * presumably happens during the dma from scanout buffer). - */ -static int request_smp_blocks(struct drm_plane *plane, uint32_t format, - uint32_t nplanes, uint32_t width) -{ - struct drm_device *dev = plane->dev; - struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); - struct mdp5_kms *mdp5_kms = get_kms(plane); - enum mdp5_pipe pipe = mdp5_plane->pipe; - int i, hsub, nlines, nblks, ret; - - hsub = drm_format_horz_chroma_subsampling(format); - - /* different if BWC (compressed framebuffer?) enabled: */ - nlines = 2; - - for (i = 0, nblks = 0; i < nplanes; i++) { - int n, fetch_stride, cpp; - - cpp = drm_format_plane_cpp(format, i); - fetch_stride = width * cpp / (i ? hsub : 1); - - n = DIV_ROUND_UP(fetch_stride * nlines, SMP_BLK_SIZE); - - /* for hw rev v1.00 */ - if (mdp5_kms->rev == 0) - n = roundup_pow_of_two(n); - - DBG("%s[%d]: request %d SMP blocks", mdp5_plane->name, i, n); - ret = mdp5_smp_request(mdp5_kms, pipe2client(pipe, i), n); - if (ret) { - dev_err(dev->dev, "Could not allocate %d SMP blocks: %d\n", - n, ret); - return ret; - } - - nblks += n; - } - - /* in success case, return total # of blocks allocated: */ - return nblks; -} - -static void set_fifo_thresholds(struct drm_plane *plane, int nblks) -{ - struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); - struct mdp5_kms *mdp5_kms = get_kms(plane); - enum mdp5_pipe pipe = mdp5_plane->pipe; - uint32_t val; - - /* 1/4 of SMP pool that is being fetched */ - val = (nblks * SMP_ENTRIES_PER_BLK) / 4; - - mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_0(pipe), val * 1); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_1(pipe), val * 2); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_2(pipe), val * 3); - -} - int mdp5_plane_mode_set(struct drm_plane *plane, struct drm_crtc *crtc, struct drm_framebuffer *fb, int crtc_x, int crtc_y, @@ -225,7 +163,7 @@ int mdp5_plane_mode_set(struct drm_plane *plane, uint32_t nplanes, config = 0; uint32_t phasex_step = 0, phasey_step = 0; uint32_t hdecm = 0, vdecm = 0; - int i, nblks; + int ret; nplanes = drm_format_num_planes(fb->pixel_format); @@ -243,12 +181,11 @@ int mdp5_plane_mode_set(struct drm_plane *plane, fb->base.id, src_x, src_y, src_w, src_h, crtc->base.id, crtc_x, crtc_y, crtc_w, crtc_h); - /* - * Calculate and request required # of smp blocks: - */ - nblks = request_smp_blocks(plane, fb->pixel_format, nplanes, src_w); - if (nblks < 0) - return nblks; + /* Request some memory from the SMP: */ + ret = mdp5_smp_request(mdp5_kms->smp_priv, + mdp5_plane->pipe, fb->pixel_format, src_w); + if (ret) + return ret; /* * Currently we update the hw for allocations/requests immediately, @@ -256,8 +193,7 @@ int mdp5_plane_mode_set(struct drm_plane *plane, * would move into atomic->check_plane_state(), while updating the * hw would remain here: */ - for (i = 0; i < pipe2nclients(pipe); i++) - mdp5_smp_configure(mdp5_kms, pipe2client(pipe, i)); + mdp5_smp_configure(mdp5_kms->smp_priv, pipe); if (src_w != crtc_w) { config |= MDP5_PIPE_SCALE_CONFIG_SCALEX_EN; @@ -330,8 +266,6 @@ int mdp5_plane_mode_set(struct drm_plane *plane, MDP5_PIPE_SCALE_CONFIG_SCALEX_MAX_FILTER(SCALE_FILTER_NEAREST) | MDP5_PIPE_SCALE_CONFIG_SCALEY_MAX_FILTER(SCALE_FILTER_NEAREST)); - set_fifo_thresholds(plane, nblks); - /* TODO detach from old crtc (if we had more than one) */ mdp5_crtc_attach(crtc, plane); @@ -342,10 +276,8 @@ void mdp5_plane_complete_flip(struct drm_plane *plane) { struct mdp5_kms *mdp5_kms = get_kms(plane); enum mdp5_pipe pipe = to_mdp5_plane(plane)->pipe; - int i; - for (i = 0; i < pipe2nclients(pipe); i++) - mdp5_smp_commit(mdp5_kms, pipe2client(pipe, i)); + mdp5_smp_commit(mdp5_kms->smp_priv, pipe); } enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane) diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c index 2d0236b..e61e1cf 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c @@ -1,4 +1,5 @@ /* + * Copyright (c) 2014, The Linux Foundation. All rights reserved. * Copyright (C) 2013 Red Hat * Author: Rob Clark * @@ -29,8 +30,11 @@ * Based on the size of the attached scanout buffer, a certain # of * blocks must be allocated to that client out of the shared pool. * - * For each block, it can be either free, or pending/in-use by a - * client. The updates happen in three steps: + * In some hw, some blocks are statically allocated for certain pipes + * and CANNOT be re-allocated (eg: MMB0 and MMB1 both tied to RGB0). + * + * For each block that can be dynamically allocated, it can be either + * free, or pending/in-use by a client. The updates happen in three steps: * * 1) mdp5_smp_request(): * When plane scanout is setup, calculate required number of @@ -61,21 +65,64 @@ * inuse and pending state of all clients.. */ -static DEFINE_SPINLOCK(smp_lock); +struct mdp5_smp { + struct drm_device *dev; + + int blk_cnt; + int blk_size; + + spinlock_t state_lock; + mdp5_smp_state_t state; /* to track smp allocation amongst pipes: */ + + struct mdp5_client_smp_state client_state[CID_MAX]; +}; +static inline +struct mdp5_kms *get_kms(struct mdp5_smp *smp) +{ + struct msm_drm_private *priv = smp->dev->dev_private; + + return to_mdp5_kms(to_mdp_kms(priv->kms)); +} + +static inline enum mdp5_client_id pipe2client(enum mdp5_pipe pipe, int plane) +{ + WARN_ON(plane >= pipe2nclients(pipe)); + switch (pipe) { + case SSPP_VIG0: return CID_VIG0_Y + plane; + case SSPP_VIG1: return CID_VIG1_Y + plane; + case SSPP_VIG2: return CID_VIG2_Y + plane; + case SSPP_RGB0: return CID_RGB0; + case SSPP_RGB1: return CID_RGB1; + case SSPP_RGB2: return CID_RGB2; + case SSPP_DMA0: return CID_DMA0_Y + plane; + case SSPP_DMA1: return CID_DMA1_Y + plane; + case SSPP_VIG3: return CID_VIG3_Y + plane; + case SSPP_RGB3: return CID_RGB3; + default: return CID_UNUSED; + } +} /* step #1: update # of blocks pending for the client: */ -int mdp5_smp_request(struct mdp5_kms *mdp5_kms, +static int smp_request_block(struct mdp5_smp *smp, enum mdp5_client_id cid, int nblks) { - struct mdp5_client_smp_state *ps = &mdp5_kms->smp_client_state[cid]; - int i, ret, avail, cur_nblks, cnt = mdp5_kms->smp_blk_cnt; + struct mdp5_kms *mdp5_kms = get_kms(smp); + struct mdp5_client_smp_state *ps = &smp->client_state[cid]; + int i, ret, avail, cur_nblks, cnt = smp->blk_cnt; + int reserved = mdp5_kms->hw_cfg->smp.reserved[cid]; unsigned long flags; - spin_lock_irqsave(&smp_lock, flags); + spin_lock_irqsave(&smp->state_lock, flags); + + nblks -= reserved; + if (reserved) + DBG("%d MMBs allocated (%d reserved)", nblks, reserved); - avail = cnt - bitmap_weight(mdp5_kms->smp_state, cnt); + avail = cnt - bitmap_weight(smp->state, cnt); if (nblks > avail) { + dev_err(mdp5_kms->dev->dev, "out of blks (req=%d > avail=%d)\n", + nblks, avail); ret = -ENOSPC; goto fail; } @@ -84,9 +131,9 @@ int mdp5_smp_request(struct mdp5_kms *mdp5_kms, if (nblks > cur_nblks) { /* grow the existing pending reservation: */ for (i = cur_nblks; i < nblks; i++) { - int blk = find_first_zero_bit(mdp5_kms->smp_state, cnt); + int blk = find_first_zero_bit(smp->state, cnt); set_bit(blk, ps->pending); - set_bit(blk, mdp5_kms->smp_state); + set_bit(blk, smp->state); } } else { /* shrink the existing pending reservation: */ @@ -98,15 +145,89 @@ int mdp5_smp_request(struct mdp5_kms *mdp5_kms, } fail: - spin_unlock_irqrestore(&smp_lock, flags); + spin_unlock_irqrestore(&smp->state_lock, flags); + return 0; +} + +static void set_fifo_thresholds(struct mdp5_smp *smp, + enum mdp5_pipe pipe, int nblks) +{ + struct mdp5_kms *mdp5_kms = get_kms(smp); + u32 smp_entries_per_blk = smp->blk_size / (128 / BITS_PER_BYTE); + u32 val; + + /* 1/4 of SMP pool that is being fetched */ + val = (nblks * smp_entries_per_blk) / 4; + + mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_0(pipe), val * 1); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_1(pipe), val * 2); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_2(pipe), val * 3); +} + +/* + * NOTE: looks like if horizontal decimation is used (if we supported that) + * then the width used to calculate SMP block requirements is the post- + * decimated width. Ie. SMP buffering sits downstream of decimation (which + * presumably happens during the dma from scanout buffer). + */ +int mdp5_smp_request(void *handler, enum mdp5_pipe pipe, u32 fmt, u32 width) +{ + struct mdp5_smp *smp = handler; + struct mdp5_kms *mdp5_kms = get_kms(smp); + struct drm_device *dev = mdp5_kms->dev; + int i, hsub, nplanes, nlines, nblks, ret; + + nplanes = drm_format_num_planes(fmt); + hsub = drm_format_horz_chroma_subsampling(fmt); + + /* different if BWC (compressed framebuffer?) enabled: */ + nlines = 2; + + for (i = 0, nblks = 0; i < nplanes; i++) { + int n, fetch_stride, cpp; + + cpp = drm_format_plane_cpp(fmt, i); + fetch_stride = width * cpp / (i ? hsub : 1); + + n = DIV_ROUND_UP(fetch_stride * nlines, smp->blk_size); + + /* for hw rev v1.00 */ + if (mdp5_kms->rev == 0) + n = roundup_pow_of_two(n); + + DBG("%s[%d]: request %d SMP blocks", pipe2name(pipe), i, n); + ret = smp_request_block(smp, pipe2client(pipe, i), n); + if (ret) { + dev_err(dev->dev, "Cannot allocate %d SMP blocks: %d\n", + n, ret); + return ret; + } + + nblks += n; + } + + set_fifo_thresholds(smp, pipe, nblks); + return 0; } -static void update_smp_state(struct mdp5_kms *mdp5_kms, +/* Release SMP blocks for all clients of the pipe */ +void mdp5_smp_release(void *handler, enum mdp5_pipe pipe) +{ + struct mdp5_smp *smp = handler; + int i, nblks; + + for (i = 0, nblks = 0; i < pipe2nclients(pipe); i++) + smp_request_block(smp, pipe2client(pipe, i), 0); + set_fifo_thresholds(smp, pipe, 0); +} + +static void update_smp_state(struct mdp5_smp *smp, enum mdp5_client_id cid, mdp5_smp_state_t *assigned) { - int cnt = mdp5_kms->smp_blk_cnt; - uint32_t blk, val; + struct mdp5_kms *mdp5_kms = get_kms(smp); + int cnt = smp->blk_cnt; + u32 blk, val; for_each_set_bit(blk, *assigned, cnt) { int idx = blk / 3; @@ -135,39 +256,84 @@ static void update_smp_state(struct mdp5_kms *mdp5_kms, } /* step #2: configure hw for union(pending, inuse): */ -void mdp5_smp_configure(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid) +void mdp5_smp_configure(void *handler, enum mdp5_pipe pipe) { - struct mdp5_client_smp_state *ps = &mdp5_kms->smp_client_state[cid]; - int cnt = mdp5_kms->smp_blk_cnt; + struct mdp5_smp *smp = handler; + int cnt = smp->blk_cnt; mdp5_smp_state_t assigned; + int i; + + for (i = 0; i < pipe2nclients(pipe); i++) { + enum mdp5_client_id cid = pipe2client(pipe, i); + struct mdp5_client_smp_state *ps = &smp->client_state[cid]; - bitmap_or(assigned, ps->inuse, ps->pending, cnt); - update_smp_state(mdp5_kms, cid, &assigned); + bitmap_or(assigned, ps->inuse, ps->pending, cnt); + update_smp_state(smp, cid, &assigned); + } } /* step #3: after vblank, copy pending -> inuse: */ -void mdp5_smp_commit(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid) +void mdp5_smp_commit(void *handler, enum mdp5_pipe pipe) { - struct mdp5_client_smp_state *ps = &mdp5_kms->smp_client_state[cid]; - int cnt = mdp5_kms->smp_blk_cnt; + struct mdp5_smp *smp = handler; + int cnt = smp->blk_cnt; mdp5_smp_state_t released; + int i; + + for (i = 0; i < pipe2nclients(pipe); i++) { + enum mdp5_client_id cid = pipe2client(pipe, i); + struct mdp5_client_smp_state *ps = &smp->client_state[cid]; + + /* + * Figure out if there are any blocks we where previously + * using, which can be released and made available to other + * clients: + */ + if (bitmap_andnot(released, ps->inuse, ps->pending, cnt)) { + unsigned long flags; + + spin_lock_irqsave(&smp->state_lock, flags); + /* clear released blocks: */ + bitmap_andnot(smp->state, smp->state, released, cnt); + spin_unlock_irqrestore(&smp->state_lock, flags); + + update_smp_state(smp, CID_UNUSED, &released); + } - /* - * Figure out if there are any blocks we where previously - * using, which can be released and made available to other - * clients: - */ - if (bitmap_andnot(released, ps->inuse, ps->pending, cnt)) { - unsigned long flags; - - spin_lock_irqsave(&smp_lock, flags); - /* clear released blocks: */ - bitmap_andnot(mdp5_kms->smp_state, mdp5_kms->smp_state, - released, cnt); - spin_unlock_irqrestore(&smp_lock, flags); - - update_smp_state(mdp5_kms, CID_UNUSED, &released); + bitmap_copy(ps->inuse, ps->pending, cnt); } +} + +void mdp5_smp_destroy(void *handler) +{ + struct mdp5_smp *smp = handler; + + kfree(smp); +} + +void *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg) +{ + struct mdp5_smp *smp = NULL; + int ret; + + smp = kzalloc(sizeof(*smp), GFP_KERNEL); + if (unlikely(!smp)) { + ret = -ENOMEM; + goto fail; + } + + smp->dev = dev; + smp->blk_cnt = cfg->mmb_count; + smp->blk_size = cfg->mmb_size; + + /* statically tied MMBs cannot be re-allocated: */ + bitmap_copy(smp->state, cfg->reserved_state, smp->blk_cnt); + spin_lock_init(&smp->state_lock); + + return smp; +fail: + if (smp) + mdp5_smp_destroy(smp); - bitmap_copy(ps->inuse, ps->pending, cnt); + return ERR_PTR(ret); } diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h index 0ab739e..3e3c753 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h @@ -1,4 +1,5 @@ /* + * Copyright (c) 2014, The Linux Foundation. All rights reserved. * Copyright (C) 2013 Red Hat * Author: Rob Clark * @@ -20,12 +21,6 @@ #include "msm_drv.h" -#define MAX_SMP_BLOCKS 22 -#define SMP_BLK_SIZE 4096 -#define SMP_ENTRIES_PER_BLK (SMP_BLK_SIZE / 16) - -typedef DECLARE_BITMAP(mdp5_smp_state_t, MAX_SMP_BLOCKS); - struct mdp5_client_smp_state { mdp5_smp_state_t inuse; mdp5_smp_state_t pending; @@ -33,9 +28,18 @@ struct mdp5_client_smp_state { struct mdp5_kms; -int mdp5_smp_request(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid, int nblks); -void mdp5_smp_configure(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid); -void mdp5_smp_commit(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid); +/* + * SMP module prototypes: + * mdp5_smp_init() returns a SMP @handler, + * which is then used to call the other mdp5_smp_*(handler, ...) functions. + */ + +void *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg); +void mdp5_smp_destroy(void *handler); +int mdp5_smp_request(void *handler, enum mdp5_pipe pipe, u32 fmt, u32 width); +void mdp5_smp_configure(void *handler, enum mdp5_pipe pipe); +void mdp5_smp_commit(void *handler, enum mdp5_pipe pipe); +void mdp5_smp_release(void *handler, enum mdp5_pipe pipe); #endif /* __MDP5_SMP_H__ */ -- cgit v0.10.2 From 2e362e1772b8978428f087007fc4d6c4990efd41 Mon Sep 17 00:00:00 2001 From: Stephane Viau Date: Tue, 18 Nov 2014 12:49:48 -0500 Subject: drm/msm/mdp5: introduce mdp5_cfg module The hardware configuration modification from a version to another is quite consequent. Introducing a configuration module (mdp5_cfg) may make things more clear and easier to access when a new hardware version comes up. Signed-off-by: Stephane Viau Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/Makefile b/drivers/gpu/drm/msm/Makefile index 0d96132..dda3852 100644 --- a/drivers/gpu/drm/msm/Makefile +++ b/drivers/gpu/drm/msm/Makefile @@ -25,6 +25,7 @@ msm-y := \ mdp/mdp4/mdp4_irq.o \ mdp/mdp4/mdp4_kms.o \ mdp/mdp4/mdp4_plane.o \ + mdp/mdp5/mdp5_cfg.o \ mdp/mdp5/mdp5_crtc.o \ mdp/mdp5/mdp5_encoder.o \ mdp/mdp5/mdp5_irq.o \ diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c new file mode 100644 index 0000000..62e77d1 --- /dev/null +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c @@ -0,0 +1,215 @@ +/* + * Copyright (c) 2014 The Linux Foundation. All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 and + * only version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + */ + +#include "mdp5_kms.h" +#include "mdp5_cfg.h" + +struct mdp5_cfg_handler { + int revision; + struct mdp5_cfg config; +}; + +/* mdp5_cfg must be exposed (used in mdp5.xml.h) */ +const struct mdp5_cfg_hw *mdp5_cfg = NULL; + +const struct mdp5_cfg_hw msm8x74_config = { + .name = "msm8x74", + .smp = { + .mmb_count = 22, + .mmb_size = 4096, + }, + .ctl = { + .count = 5, + .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 }, + }, + .pipe_vig = { + .count = 3, + .base = { 0x01200, 0x01600, 0x01a00 }, + }, + .pipe_rgb = { + .count = 3, + .base = { 0x01e00, 0x02200, 0x02600 }, + }, + .pipe_dma = { + .count = 2, + .base = { 0x02a00, 0x02e00 }, + }, + .lm = { + .count = 5, + .base = { 0x03200, 0x03600, 0x03a00, 0x03e00, 0x04200 }, + .nb_stages = 5, + }, + .dspp = { + .count = 3, + .base = { 0x04600, 0x04a00, 0x04e00 }, + }, + .ad = { + .count = 2, + .base = { 0x13100, 0x13300 }, /* NOTE: no ad in v1.0 */ + }, + .intf = { + .count = 4, + .base = { 0x12500, 0x12700, 0x12900, 0x12b00 }, + }, + .max_clk = 200000000, +}; + +const struct mdp5_cfg_hw apq8084_config = { + .name = "apq8084", + .smp = { + .mmb_count = 44, + .mmb_size = 8192, + .reserved_state[0] = GENMASK(7, 0), /* first 8 MMBs */ + .reserved[CID_RGB0] = 2, + .reserved[CID_RGB1] = 2, + .reserved[CID_RGB2] = 2, + .reserved[CID_RGB3] = 2, + }, + .ctl = { + .count = 5, + .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 }, + }, + .pipe_vig = { + .count = 4, + .base = { 0x01200, 0x01600, 0x01a00, 0x01e00 }, + }, + .pipe_rgb = { + .count = 4, + .base = { 0x02200, 0x02600, 0x02a00, 0x02e00 }, + }, + .pipe_dma = { + .count = 2, + .base = { 0x03200, 0x03600 }, + }, + .lm = { + .count = 6, + .base = { 0x03a00, 0x03e00, 0x04200, 0x04600, 0x04a00, 0x04e00 }, + .nb_stages = 5, + }, + .dspp = { + .count = 4, + .base = { 0x05200, 0x05600, 0x05a00, 0x05e00 }, + + }, + .ad = { + .count = 3, + .base = { 0x13500, 0x13700, 0x13900 }, + }, + .intf = { + .count = 5, + .base = { 0x12500, 0x12700, 0x12900, 0x12b00, 0x12d00 }, + }, + .max_clk = 320000000, +}; + +static const struct mdp5_cfg_handler cfg_handlers[] = { + { .revision = 0, .config = { .hw = &msm8x74_config } }, + { .revision = 2, .config = { .hw = &msm8x74_config } }, + { .revision = 3, .config = { .hw = &apq8084_config } }, +}; + + +static struct mdp5_cfg_platform *mdp5_get_config(struct platform_device *dev); + +const struct mdp5_cfg_hw *mdp5_cfg_get_hw_config(void *cfg_hnd) +{ + struct mdp5_cfg_handler *cfg_handler = cfg_hnd; + + return cfg_handler->config.hw; +} + +struct mdp5_cfg *mdp5_cfg_get_config(void *cfg_hnd) +{ + struct mdp5_cfg_handler *cfg_handler = cfg_hnd; + + return &cfg_handler->config; +} + +int mdp5_cfg_get_hw_rev(void *cfg_hnd) +{ + struct mdp5_cfg_handler *cfg_handler = cfg_hnd; + + return cfg_handler->revision; +} + +void mdp5_cfg_destroy(void *cfg_hnd) +{ + struct mdp5_cfg_handler *cfg_handler = cfg_hnd; + + kfree(cfg_handler); +} + +void *mdp5_cfg_init(struct mdp5_kms *mdp5_kms, + uint32_t major, uint32_t minor) +{ + struct drm_device *dev = mdp5_kms->dev; + struct platform_device *pdev = dev->platformdev; + struct mdp5_cfg_handler *cfg_handler; + struct mdp5_cfg_platform *pconfig; + int i, ret = 0; + + cfg_handler = kzalloc(sizeof(*cfg_handler), GFP_KERNEL); + if (unlikely(!cfg_handler)) { + ret = -ENOMEM; + goto fail; + } + + if (major != 1) { + dev_err(dev->dev, "unexpected MDP major version: v%d.%d\n", + major, minor); + ret = -ENXIO; + goto fail; + } + + /* only after mdp5_cfg global pointer's init can we access the hw */ + for (i = 0; i < ARRAY_SIZE(cfg_handlers); i++) { + if (cfg_handlers[i].revision != minor) + continue; + mdp5_cfg = cfg_handlers[i].config.hw; + + break; + } + if (unlikely(!mdp5_cfg)) { + dev_err(dev->dev, "unexpected MDP minor revision: v%d.%d\n", + major, minor); + ret = -ENXIO; + goto fail; + } + + cfg_handler->revision = minor; + cfg_handler->config.hw = mdp5_cfg; + + pconfig = mdp5_get_config(pdev); + memcpy(&cfg_handler->config.platform, pconfig, sizeof(*pconfig)); + + DBG("MDP5: %s hw config selected", mdp5_cfg->name); + + return cfg_handler; + +fail: + if (cfg_handler) + mdp5_cfg_destroy(cfg_handler); + + return NULL; +} + +static struct mdp5_cfg_platform *mdp5_get_config(struct platform_device *dev) +{ + static struct mdp5_cfg_platform config = {}; +#ifdef CONFIG_OF + /* TODO */ +#endif + config.iommu = iommu_domain_alloc(&platform_bus_type); + + return &config; +} diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h new file mode 100644 index 0000000..00c8271 --- /dev/null +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h @@ -0,0 +1,88 @@ +/* + * Copyright (c) 2014 The Linux Foundation. All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 and + * only version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + */ + +#ifndef __MDP5_CFG_H__ +#define __MDP5_CFG_H__ + +#include "msm_drv.h" + +/* + * mdp5_cfg + * + * This module configures the dynamic offsets used by mdp5.xml.h + * (initialized in mdp5_cfg.c) + */ +extern const struct mdp5_cfg_hw *mdp5_cfg; + +#define MAX_BASES 8 +#define MAX_SMP_BLOCKS 44 +#define MAX_CLIENTS 32 + +typedef DECLARE_BITMAP(mdp5_smp_state_t, MAX_SMP_BLOCKS); + +#define MDP5_SUB_BLOCK_DEFINITION \ + int count; \ + uint32_t base[MAX_BASES] + +struct mdp5_sub_block { + MDP5_SUB_BLOCK_DEFINITION; +}; + +struct mdp5_lm_block { + MDP5_SUB_BLOCK_DEFINITION; + uint32_t nb_stages; /* number of stages per blender */ +}; + +struct mdp5_smp_block { + int mmb_count; /* number of SMP MMBs */ + int mmb_size; /* MMB: size in bytes */ + mdp5_smp_state_t reserved_state;/* SMP MMBs statically allocated */ + int reserved[MAX_CLIENTS]; /* # of MMBs allocated per client */ +}; + +struct mdp5_cfg_hw { + char *name; + + struct mdp5_smp_block smp; + struct mdp5_sub_block ctl; + struct mdp5_sub_block pipe_vig; + struct mdp5_sub_block pipe_rgb; + struct mdp5_sub_block pipe_dma; + struct mdp5_lm_block lm; + struct mdp5_sub_block dspp; + struct mdp5_sub_block ad; + struct mdp5_sub_block intf; + + uint32_t max_clk; +}; + +/* platform config data (ie. from DT, or pdata) */ +struct mdp5_cfg_platform { + struct iommu_domain *iommu; +}; + +struct mdp5_cfg { + const struct mdp5_cfg_hw *hw; + struct mdp5_cfg_platform platform; +}; + +struct mdp5_kms; + +const struct mdp5_cfg_hw *mdp5_cfg_get_hw_config(void *cfg_hnd); +struct mdp5_cfg *mdp5_cfg_get_config(void *cfg_hnd); +int mdp5_cfg_get_hw_rev(void *cfg_hnd); + +void *mdp5_cfg_init(struct mdp5_kms *mdp5_kms, uint32_t major, uint32_t minor); +void mdp5_cfg_destroy(void *cfg_hnd); + +#endif /* __MDP5_CFG_H__ */ diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c index f852fa4..64de0f9 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c @@ -1,4 +1,5 @@ /* + * Copyright (c) 2014, The Linux Foundation. All rights reserved. * Copyright (C) 2013 Red Hat * Author: Rob Clark * @@ -24,158 +25,10 @@ static const char *iommu_ports[] = { "mdp_0", }; -static struct mdp5_platform_config *mdp5_get_config(struct platform_device *dev); - -const struct mdp5_config *mdp5_cfg; - -static const struct mdp5_config msm8x74_config = { - .name = "msm8x74", - .smp = { - .mmb_count = 22, - .mmb_size = 4096, - }, - .ctl = { - .count = 5, - .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 }, - }, - .pipe_vig = { - .count = 3, - .base = { 0x01200, 0x01600, 0x01a00 }, - }, - .pipe_rgb = { - .count = 3, - .base = { 0x01e00, 0x02200, 0x02600 }, - }, - .pipe_dma = { - .count = 2, - .base = { 0x02a00, 0x02e00 }, - }, - .lm = { - .count = 5, - .base = { 0x03200, 0x03600, 0x03a00, 0x03e00, 0x04200 }, - }, - .dspp = { - .count = 3, - .base = { 0x04600, 0x04a00, 0x04e00 }, - }, - .ad = { - .count = 2, - .base = { 0x13100, 0x13300 }, /* NOTE: no ad in v1.0 */ - }, - .intf = { - .count = 4, - .base = { 0x12500, 0x12700, 0x12900, 0x12b00 }, - }, - .max_clk = 200000000, -}; - -static const struct mdp5_config apq8084_config = { - .name = "apq8084", - .smp = { - .mmb_count = 44, - .mmb_size = 8192, - .reserved_state[0] = GENMASK(7, 0), /* first 8 MMBs */ - .reserved[CID_RGB0] = 2, - .reserved[CID_RGB1] = 2, - .reserved[CID_RGB2] = 2, - .reserved[CID_RGB3] = 2, - }, - .ctl = { - .count = 5, - .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 }, - }, - .pipe_vig = { - .count = 4, - .base = { 0x01200, 0x01600, 0x01a00, 0x01e00 }, - }, - .pipe_rgb = { - .count = 4, - .base = { 0x02200, 0x02600, 0x02a00, 0x02e00 }, - }, - .pipe_dma = { - .count = 2, - .base = { 0x03200, 0x03600 }, - }, - .lm = { - .count = 6, - .base = { 0x03a00, 0x03e00, 0x04200, 0x04600, 0x04a00, 0x04e00 }, - }, - .dspp = { - .count = 4, - .base = { 0x05200, 0x05600, 0x05a00, 0x05e00 }, - - }, - .ad = { - .count = 3, - .base = { 0x13500, 0x13700, 0x13900 }, - }, - .intf = { - .count = 5, - .base = { 0x12500, 0x12700, 0x12900, 0x12b00, 0x12d00 }, - }, - .max_clk = 320000000, -}; - -struct mdp5_config_entry { - int revision; - const struct mdp5_config *config; -}; - -static const struct mdp5_config_entry mdp5_configs[] = { - { .revision = 0, .config = &msm8x74_config }, - { .revision = 2, .config = &msm8x74_config }, - { .revision = 3, .config = &apq8084_config }, -}; - -static int mdp5_select_hw_cfg(struct msm_kms *kms) -{ - struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms)); - struct drm_device *dev = mdp5_kms->dev; - uint32_t version, major, minor; - int i, ret = 0; - - mdp5_enable(mdp5_kms); - version = mdp5_read(mdp5_kms, REG_MDP5_MDP_VERSION); - mdp5_disable(mdp5_kms); - - major = FIELD(version, MDP5_MDP_VERSION_MAJOR); - minor = FIELD(version, MDP5_MDP_VERSION_MINOR); - - DBG("found MDP5 version v%d.%d", major, minor); - - if (major != 1) { - dev_err(dev->dev, "unexpected MDP major version: v%d.%d\n", - major, minor); - ret = -ENXIO; - goto out; - } - - mdp5_kms->rev = minor; - - /* only after mdp5_cfg global pointer's init can we access the hw */ - for (i = 0; i < ARRAY_SIZE(mdp5_configs); i++) { - if (mdp5_configs[i].revision != minor) - continue; - mdp5_kms->hw_cfg = mdp5_cfg = mdp5_configs[i].config; - break; - } - if (unlikely(!mdp5_kms->hw_cfg)) { - dev_err(dev->dev, "unexpected MDP minor revision: v%d.%d\n", - major, minor); - ret = -ENXIO; - goto out; - } - - DBG("MDP5: %s config selected", mdp5_kms->hw_cfg->name); - - return 0; -out: - return ret; -} - static int mdp5_hw_init(struct msm_kms *kms) { struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms)); + const struct mdp5_cfg_hw *hw_cfg; struct drm_device *dev = mdp5_kms->dev; int i; @@ -207,7 +60,9 @@ static int mdp5_hw_init(struct msm_kms *kms) mdp5_write(mdp5_kms, REG_MDP5_DISP_INTF_SEL, 0); - for (i = 0; i < mdp5_kms->hw_cfg->ctl.count; i++) + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); + + for (i = 0; i < hw_cfg->ctl.count; i++) mdp5_write(mdp5_kms, REG_MDP5_CTL_OP(i), 0); pm_runtime_put_sync(dev->dev); @@ -236,6 +91,7 @@ static void mdp5_destroy(struct msm_kms *kms) struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms)); struct msm_mmu *mmu = mdp5_kms->mmu; void *smp = mdp5_kms->smp_priv; + void *cfg = mdp5_kms->cfg_priv; mdp5_irq_domain_fini(mdp5_kms); @@ -246,6 +102,8 @@ static void mdp5_destroy(struct msm_kms *kms) if (smp) mdp5_smp_destroy(smp); + if (cfg) + mdp5_cfg_destroy(cfg); kfree(mdp5_kms); } @@ -299,8 +157,11 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) struct drm_device *dev = mdp5_kms->dev; struct msm_drm_private *priv = dev->dev_private; struct drm_encoder *encoder; + const struct mdp5_cfg_hw *hw_cfg; int i, ret; + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); + /* register our interrupt-controller for hdmi/eDP/dsi/etc * to use for irqs routed through mdp: */ @@ -309,7 +170,7 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) goto fail; /* construct CRTCs: */ - for (i = 0; i < mdp5_kms->hw_cfg->pipe_rgb.count; i++) { + for (i = 0; i < hw_cfg->pipe_rgb.count; i++) { struct drm_plane *plane; struct drm_crtc *crtc; @@ -367,6 +228,21 @@ fail: return ret; } +static void read_hw_revision(struct mdp5_kms *mdp5_kms, + uint32_t *major, uint32_t *minor) +{ + uint32_t version; + + mdp5_enable(mdp5_kms); + version = mdp5_read(mdp5_kms, REG_MDP5_MDP_VERSION); + mdp5_disable(mdp5_kms); + + *major = FIELD(version, MDP5_MDP_VERSION_MAJOR); + *minor = FIELD(version, MDP5_MDP_VERSION_MINOR); + + DBG("MDP5 version v%d.%d", *major, *minor); +} + static int get_clk(struct platform_device *pdev, struct clk **clkp, const char *name) { @@ -383,10 +259,11 @@ static int get_clk(struct platform_device *pdev, struct clk **clkp, struct msm_kms *mdp5_kms_init(struct drm_device *dev) { struct platform_device *pdev = dev->platformdev; - struct mdp5_platform_config *config = mdp5_get_config(pdev); + struct mdp5_cfg *config; struct mdp5_kms *mdp5_kms; struct msm_kms *kms = NULL; struct msm_mmu *mmu; + uint32_t major, minor; void *priv; int i, ret; @@ -446,14 +323,19 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) if (ret) goto fail; - ret = mdp5_select_hw_cfg(kms); - if (ret) + read_hw_revision(mdp5_kms, &major, &minor); + priv = mdp5_cfg_init(mdp5_kms, major, minor); + if (IS_ERR(priv)) { + ret = PTR_ERR(priv); goto fail; + } + mdp5_kms->cfg_priv = priv; + config = mdp5_cfg_get_config(mdp5_kms->cfg_priv); /* TODO: compute core clock rate at runtime */ - clk_set_rate(mdp5_kms->src_clk, mdp5_kms->hw_cfg->max_clk); + clk_set_rate(mdp5_kms->src_clk, config->hw->max_clk); - priv = mdp5_smp_init(mdp5_kms->dev, &mdp5_kms->hw_cfg->smp); + priv = mdp5_smp_init(mdp5_kms->dev, &config->hw->smp); if (IS_ERR(priv)) { ret = PTR_ERR(priv); goto fail; @@ -465,13 +347,13 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) * we don't disable): */ mdp5_enable(mdp5_kms); - for (i = 0; i < mdp5_kms->hw_cfg->intf.count; i++) + for (i = 0; i < config->hw->intf.count; i++) mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(i), 0); mdp5_disable(mdp5_kms); mdelay(16); - if (config->iommu) { - mmu = msm_iommu_new(&pdev->dev, config->iommu); + if (config->platform.iommu) { + mmu = msm_iommu_new(&pdev->dev, config->platform.iommu); if (IS_ERR(mmu)) { ret = PTR_ERR(mmu); dev_err(dev->dev, "failed to init iommu: %d\n", ret); @@ -512,14 +394,3 @@ fail: mdp5_destroy(kms); return ERR_PTR(ret); } - -static struct mdp5_platform_config *mdp5_get_config(struct platform_device *dev) -{ - static struct mdp5_platform_config config = {}; -#ifdef CONFIG_OF - /* TODO */ -#endif - config.iommu = iommu_domain_alloc(&platform_bus_type); - - return &config; -} diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h index 3f7aa49..daca8da 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h @@ -21,36 +21,7 @@ #include "msm_drv.h" #include "msm_kms.h" #include "mdp/mdp_kms.h" -/* dynamic offsets used by mdp5.xml.h (initialized in mdp5_kms.c) */ -#define MDP5_MAX_BASES 8 -#define MAX_SMP_BLOCKS 44 -#define MAX_CLIENTS 32 -typedef DECLARE_BITMAP(mdp5_smp_state_t, MAX_SMP_BLOCKS); -struct mdp5_sub_block { - int count; - uint32_t base[MDP5_MAX_BASES]; -}; -struct mdp5_smp_block { - int mmb_count; /* number of SMP MMBs */ - int mmb_size; /* MMB: size in bytes */ - mdp5_smp_state_t reserved_state;/* SMP MMBs statically allocated */ - int reserved[MAX_CLIENTS]; /* # of MMBs reserved per client */ -}; -struct mdp5_config { - char *name; - struct mdp5_smp_block smp; - struct mdp5_sub_block ctl; - struct mdp5_sub_block pipe_vig; - struct mdp5_sub_block pipe_rgb; - struct mdp5_sub_block pipe_dma; - struct mdp5_sub_block lm; - struct mdp5_sub_block dspp; - struct mdp5_sub_block ad; - struct mdp5_sub_block intf; - - uint32_t max_clk; -}; -extern const struct mdp5_config *mdp5_cfg; +#include "mdp5_cfg.h" /* must be included before mdp5.xml.h */ #include "mdp5.xml.h" #include "mdp5_smp.h" @@ -59,8 +30,7 @@ struct mdp5_kms { struct drm_device *dev; - int rev; - const struct mdp5_config *hw_cfg; + void *cfg_priv; /* mapper-id used to request GEM buffer mapped for scanout: */ int id; @@ -89,11 +59,6 @@ struct mdp5_kms { }; #define to_mdp5_kms(x) container_of(x, struct mdp5_kms, base) -/* platform config data (ie. from DT, or pdata) */ -struct mdp5_platform_config { - struct iommu_domain *iommu; -}; - static inline void mdp5_write(struct mdp5_kms *mdp5_kms, u32 reg, u32 data) { msm_writel(data, mdp5_kms->mmio + reg); diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c index e61e1cf..04996ca 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c @@ -108,11 +108,15 @@ static int smp_request_block(struct mdp5_smp *smp, enum mdp5_client_id cid, int nblks) { struct mdp5_kms *mdp5_kms = get_kms(smp); + const struct mdp5_cfg_hw *hw_cfg; struct mdp5_client_smp_state *ps = &smp->client_state[cid]; int i, ret, avail, cur_nblks, cnt = smp->blk_cnt; - int reserved = mdp5_kms->hw_cfg->smp.reserved[cid]; + int reserved; unsigned long flags; + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); + reserved = hw_cfg->smp.reserved[cid]; + spin_lock_irqsave(&smp->state_lock, flags); nblks -= reserved; @@ -175,6 +179,7 @@ int mdp5_smp_request(void *handler, enum mdp5_pipe pipe, u32 fmt, u32 width) struct mdp5_smp *smp = handler; struct mdp5_kms *mdp5_kms = get_kms(smp); struct drm_device *dev = mdp5_kms->dev; + int rev = mdp5_cfg_get_hw_rev(mdp5_kms->cfg_priv); int i, hsub, nplanes, nlines, nblks, ret; nplanes = drm_format_num_planes(fmt); @@ -192,7 +197,7 @@ int mdp5_smp_request(void *handler, enum mdp5_pipe pipe, u32 fmt, u32 width) n = DIV_ROUND_UP(fetch_stride * nlines, smp->blk_size); /* for hw rev v1.00 */ - if (mdp5_kms->rev == 0) + if (rev == 0) n = roundup_pow_of_two(n); DBG("%s[%d]: request %d SMP blocks", pipe2name(pipe), i, n); -- cgit v0.10.2 From ac7a570406417e9d837f81c3a6b83fc8d629e583 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Thu, 20 Nov 2014 17:08:06 -0500 Subject: drm/msm/mdp5: set rate before enabling clk Set a "safe" rate at first, in order to read out the hw revision. And then after set the optimal value. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c index 64de0f9..adb4541 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c @@ -323,6 +323,12 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) if (ret) goto fail; + /* we need to set a default rate before enabling. Set a safe + * rate first, then figure out hw revision, and then set a + * more optimal rate: + */ + clk_set_rate(mdp5_kms->src_clk, 200000000); + read_hw_revision(mdp5_kms, &major, &minor); priv = mdp5_cfg_init(mdp5_kms, major, minor); if (IS_ERR(priv)) { -- cgit v0.10.2 From 0deed25b65aaf495e36818481cfc9f58dfa5cd3f Mon Sep 17 00:00:00 2001 From: Stephane Viau Date: Tue, 18 Nov 2014 12:49:49 -0500 Subject: drm/msm: add multiple CRTC and overlay support MDP5 currently support one single CRTC with its private pipe. This change allows the configuration of multiple CRTCs with the possibility to attach several public planes to these CRTCs. Signed-off-by: Stephane Viau Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/Makefile b/drivers/gpu/drm/msm/Makefile index dda3852..143d988 100644 --- a/drivers/gpu/drm/msm/Makefile +++ b/drivers/gpu/drm/msm/Makefile @@ -26,6 +26,7 @@ msm-y := \ mdp/mdp4/mdp4_kms.o \ mdp/mdp4/mdp4_plane.o \ mdp/mdp5/mdp5_cfg.o \ + mdp/mdp5/mdp5_ctl.o \ mdp/mdp5/mdp5_crtc.o \ mdp/mdp5/mdp5_encoder.o \ mdp/mdp5/mdp5_irq.o \ diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h index 00c8271..d0c98f9 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h @@ -24,6 +24,7 @@ */ extern const struct mdp5_cfg_hw *mdp5_cfg; +#define MAX_CTL 8 #define MAX_BASES 8 #define MAX_SMP_BLOCKS 44 #define MAX_CLIENTS 32 diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c index b7b32c4..85f2fb4 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c @@ -1,4 +1,5 @@ /* + * Copyright (c) 2014 The Linux Foundation. All rights reserved. * Copyright (C) 2013 Red Hat * Author: Rob Clark * @@ -22,14 +23,21 @@ #include "drm_crtc_helper.h" #include "drm_flip_work.h" +#define SSPP_MAX (SSPP_RGB3 + 1) /* TODO: Add SSPP_MAX in mdp5.xml.h */ + struct mdp5_crtc { struct drm_crtc base; char name[8]; int id; bool enabled; - /* which mixer/encoder we route output to: */ - int mixer; + /* layer mixer used for this CRTC (+ its lock): */ +#define GET_LM_ID(crtc_id) ((crtc_id == 3) ? 5 : crtc_id) + int lm; + spinlock_t lm_lock; /* protect REG_MDP5_LM_* registers */ + + /* CTL used for this CRTC: */ + void *ctl; /* if there is a pending flip, these will be non-null: */ struct drm_pending_vblank_event *event; @@ -71,25 +79,38 @@ static void request_pending(struct drm_crtc *crtc, uint32_t pending) mdp_irq_register(&get_kms(crtc)->base, &mdp5_crtc->vblank); } -static void crtc_flush(struct drm_crtc *crtc) +#define mdp5_lm_get_flush(lm) mdp_ctl_flush_mask_lm(lm) + +static void crtc_flush(struct drm_crtc *crtc, u32 flush_mask) +{ + struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); + + DBG("%s: flush=%08x", mdp5_crtc->name, flush_mask); + mdp5_ctl_commit(mdp5_crtc->ctl, flush_mask); +} + +/* + * flush updates, to make sure hw is updated to new scanout fb, + * so that we can safely queue unref to current fb (ie. next + * vblank we know hw is done w/ previous scanout_fb). + */ +static void crtc_flush_all(struct drm_crtc *crtc) { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); - struct mdp5_kms *mdp5_kms = get_kms(crtc); - int id = mdp5_crtc->id; struct drm_plane *plane; - uint32_t flush = 0; + uint32_t flush_mask = 0; + + /* we could have already released CTL in the disable path: */ + if (!mdp5_crtc->ctl) + return; for_each_plane_on_crtc(crtc, plane) { - enum mdp5_pipe pipe = mdp5_plane_pipe(plane); - flush |= pipe2flush(pipe); + flush_mask |= mdp5_plane_get_flush(plane); } + flush_mask |= mdp5_ctl_get_flush(mdp5_crtc->ctl); + flush_mask |= mdp5_lm_get_flush(mdp5_crtc->lm); - flush |= mixer2flush(mdp5_crtc->id); - flush |= MDP5_CTL_FLUSH_CTL; - - DBG("%s: flush=%08x", mdp5_crtc->name, flush); - - mdp5_write(mdp5_kms, REG_MDP5_CTL_FLUSH(id), flush); + crtc_flush(crtc, flush_mask); } static void update_fb(struct drm_crtc *crtc, struct drm_framebuffer *new_fb) @@ -117,12 +138,6 @@ static void update_scanout(struct drm_crtc *crtc, struct drm_framebuffer *fb) { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); - /* flush updates, to make sure hw is updated to new scanout fb, - * so that we can safely queue unref to current fb (ie. next - * vblank we know hw is done w/ previous scanout_fb). - */ - crtc_flush(crtc); - if (mdp5_crtc->scanout_fb) drm_flip_work_queue(&mdp5_crtc->unref_fb_work, mdp5_crtc->scanout_fb); @@ -173,6 +188,7 @@ static void pageflip_cb(struct msm_fence_cb *cb) drm_framebuffer_reference(fb); mdp5_plane_set_scanout(crtc->primary, fb); update_scanout(crtc, fb); + crtc_flush_all(crtc); } static void unref_fb_worker(struct drm_flip_work *work, void *val) @@ -223,41 +239,68 @@ static bool mdp5_crtc_mode_fixup(struct drm_crtc *crtc, return true; } +/* + * blend_setup() - blend all the planes of a CRTC + * + * When border is enabled, the border color will ALWAYS be the base layer. + * Therefore, the first plane (private RGB pipe) will start at STAGE0. + * If disabled, the first plane starts at STAGE_BASE. + * + * Note: + * Border is not enabled here because the private plane is exactly + * the CRTC resolution. + */ static void blend_setup(struct drm_crtc *crtc) { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); struct mdp5_kms *mdp5_kms = get_kms(crtc); - int id = mdp5_crtc->id; + struct drm_plane *plane; + const struct mdp5_cfg_hw *hw_cfg; + uint32_t lm = mdp5_crtc->lm, blend_cfg = 0; + enum mdp_mixer_stage_id stage; + unsigned long flags; +#define blender(stage) ((stage) - STAGE_BASE) - /* - * Hard-coded setup for now until I figure out how the - * layer-mixer works - */ + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); - /* LM[id]: */ - mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(id), - MDP5_LM_BLEND_COLOR_OUT_STAGE0_FG_ALPHA); - mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_OP_MODE(id, 0), - MDP5_LM_BLEND_OP_MODE_FG_ALPHA(FG_CONST) | - MDP5_LM_BLEND_OP_MODE_BG_ALPHA(FG_PIXEL) | - MDP5_LM_BLEND_OP_MODE_BG_INV_ALPHA); - mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_FG_ALPHA(id, 0), 0xff); - mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_BG_ALPHA(id, 0), 0x00); - - /* NOTE: seems that LM[n] and CTL[m], we do not need n==m.. but - * we want to be setting CTL[m].LAYER[n]. Not sure what the - * point of having CTL[m].LAYER[o] (for o!=n).. maybe that is - * used when chaining up mixers for high resolution displays? - */ + spin_lock_irqsave(&mdp5_crtc->lm_lock, flags); + + /* ctl could be released already when we are shutting down: */ + if (!mdp5_crtc->ctl) + goto out; + + for_each_plane_on_crtc(crtc, plane) { + struct mdp5_overlay_info *overlay; + + overlay = mdp5_plane_get_overlay_info(plane); + stage = overlay->zorder; - /* CTL[id]: */ - mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 0), - MDP5_CTL_LAYER_REG_RGB0(STAGE0) | - MDP5_CTL_LAYER_REG_BORDER_COLOR); - mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 1), 0); - mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 2), 0); - mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 3), 0); - mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 4), 0); + /* + * Note: This cannot happen with current implementation but + * we need to check this condition once z property is added + */ + BUG_ON(stage > hw_cfg->lm.nb_stages); + + /* LM */ + mdp5_write(mdp5_kms, + REG_MDP5_LM_BLEND_OP_MODE(lm, blender(stage)), + MDP5_LM_BLEND_OP_MODE_FG_ALPHA(FG_CONST) | + MDP5_LM_BLEND_OP_MODE_BG_ALPHA(BG_CONST)); + mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_FG_ALPHA(lm, + blender(stage)), 0xff); + mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_BG_ALPHA(lm, + blender(stage)), 0x00); + /* CTL */ + blend_cfg |= mdp_ctl_blend_mask(mdp5_plane_pipe(plane), stage); + DBG("%s: blending pipe %s on stage=%d", mdp5_crtc->name, + pipe2name(mdp5_plane_pipe(plane)), stage); + } + + DBG("%s: lm%d: blend config = 0x%08x", mdp5_crtc->name, lm, blend_cfg); + mdp5_ctl_blend(mdp5_crtc->ctl, lm, blend_cfg); + +out: + spin_unlock_irqrestore(&mdp5_crtc->lm_lock, flags); } static int mdp5_crtc_mode_set(struct drm_crtc *crtc, @@ -268,6 +311,7 @@ static int mdp5_crtc_mode_set(struct drm_crtc *crtc, { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); struct mdp5_kms *mdp5_kms = get_kms(crtc); + unsigned long flags; int ret; mode = adjusted_mode; @@ -281,6 +325,13 @@ static int mdp5_crtc_mode_set(struct drm_crtc *crtc, mode->vsync_end, mode->vtotal, mode->type, mode->flags); + /* request a free CTL, if none is already allocated for this CRTC */ + if (!mdp5_crtc->ctl) { + mdp5_crtc->ctl = mdp5_ctl_request(mdp5_kms->ctl_priv, crtc); + if (!mdp5_crtc->ctl) + return -EBUSY; + } + /* grab extra ref for update_scanout() */ drm_framebuffer_reference(crtc->primary->fb); @@ -295,12 +346,15 @@ static int mdp5_crtc_mode_set(struct drm_crtc *crtc, return ret; } - mdp5_write(mdp5_kms, REG_MDP5_LM_OUT_SIZE(mdp5_crtc->id), + spin_lock_irqsave(&mdp5_crtc->lm_lock, flags); + mdp5_write(mdp5_kms, REG_MDP5_LM_OUT_SIZE(mdp5_crtc->lm), MDP5_LM_OUT_SIZE_WIDTH(mode->hdisplay) | MDP5_LM_OUT_SIZE_HEIGHT(mode->vdisplay)); + spin_unlock_irqrestore(&mdp5_crtc->lm_lock, flags); update_fb(crtc, crtc->primary->fb); update_scanout(crtc, crtc->primary->fb); + /* crtc_flush_all(crtc) will be called in _commit callback */ return 0; } @@ -317,7 +371,7 @@ static void mdp5_crtc_prepare(struct drm_crtc *crtc) static void mdp5_crtc_commit(struct drm_crtc *crtc) { mdp5_crtc_dpms(crtc, DRM_MODE_DPMS_ON); - crtc_flush(crtc); + crtc_flush_all(crtc); /* drop the ref to mdp clk's that we got in prepare: */ mdp5_disable(get_kms(crtc)); } @@ -343,6 +397,7 @@ static int mdp5_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, update_fb(crtc, crtc->primary->fb); update_scanout(crtc, crtc->primary->fb); + crtc_flush_all(crtc); return 0; } @@ -351,6 +406,19 @@ static void mdp5_crtc_load_lut(struct drm_crtc *crtc) { } +static void mdp5_crtc_disable(struct drm_crtc *crtc) +{ + struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); + + DBG("%s", mdp5_crtc->name); + + if (mdp5_crtc->ctl) { + mdp5_ctl_release(mdp5_crtc->ctl); + mdp5_crtc->ctl = NULL; + } +} + + static int mdp5_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *new_fb, struct drm_pending_vblank_event *event, @@ -399,6 +467,7 @@ static const struct drm_crtc_helper_funcs mdp5_crtc_helper_funcs = { .commit = mdp5_crtc_commit, .mode_set_base = mdp5_crtc_mode_set_base, .load_lut = mdp5_crtc_load_lut, + .disable = mdp5_crtc_disable, }; static void mdp5_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus) @@ -421,9 +490,8 @@ static void mdp5_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus) static void mdp5_crtc_err_irq(struct mdp_irq *irq, uint32_t irqstatus) { struct mdp5_crtc *mdp5_crtc = container_of(irq, struct mdp5_crtc, err); - struct drm_crtc *crtc = &mdp5_crtc->base; + DBG("%s: error: %08x", mdp5_crtc->name, irqstatus); - crtc_flush(crtc); } uint32_t mdp5_crtc_vblank(struct drm_crtc *crtc) @@ -444,10 +512,9 @@ void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf, { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); struct mdp5_kms *mdp5_kms = get_kms(crtc); - static const enum mdp5_intfnum intfnum[] = { - INTF0, INTF1, INTF2, INTF3, - }; + uint32_t flush_mask = 0; uint32_t intf_sel; + unsigned long flags; /* now that we know what irq's we want: */ mdp5_crtc->err.irqmask = intf2err(intf); @@ -457,6 +524,7 @@ void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf, if (!mdp5_kms) return; + spin_lock_irqsave(&mdp5_kms->resource_lock, flags); intf_sel = mdp5_read(mdp5_kms, REG_MDP5_DISP_INTF_SEL); switch (intf) { @@ -481,16 +549,24 @@ void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf, break; } - blend_setup(crtc); + mdp5_write(mdp5_kms, REG_MDP5_DISP_INTF_SEL, intf_sel); + spin_unlock_irqrestore(&mdp5_kms->resource_lock, flags); DBG("%s: intf_sel=%08x", mdp5_crtc->name, intf_sel); + mdp5_ctl_set_intf(mdp5_crtc->ctl, intf); + flush_mask |= mdp5_ctl_get_flush(mdp5_crtc->ctl); + flush_mask |= mdp5_lm_get_flush(mdp5_crtc->lm); - mdp5_write(mdp5_kms, REG_MDP5_DISP_INTF_SEL, intf_sel); - mdp5_write(mdp5_kms, REG_MDP5_CTL_OP(mdp5_crtc->id), - MDP5_CTL_OP_MODE(MODE_NONE) | - MDP5_CTL_OP_INTF_NUM(intfnum[intf])); + crtc_flush(crtc, flush_mask); +} - crtc_flush(crtc); +static int count_planes(struct drm_crtc *crtc) +{ + struct drm_plane *plane; + int cnt = 0; + for_each_plane_on_crtc(crtc, plane) + cnt++; + return cnt; } static void set_attach(struct drm_crtc *crtc, enum mdp5_pipe pipe_id, @@ -498,14 +574,68 @@ static void set_attach(struct drm_crtc *crtc, enum mdp5_pipe pipe_id, { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); + if (plane) + plane->crtc = crtc; + + DBG("%s: %d planes attached", mdp5_crtc->name, count_planes(crtc)); + blend_setup(crtc); - if (mdp5_crtc->enabled && (plane != crtc->primary)) - crtc_flush(crtc); + if (mdp5_crtc->enabled) + crtc_flush_all(crtc); } -void mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane) +int mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane) { + struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); + struct mdp5_kms *mdp5_kms = get_kms(crtc); + struct device *dev = crtc->dev->dev; + const struct mdp5_cfg_hw *hw_cfg; + bool private_plane = (plane == crtc->primary); + struct mdp5_overlay_info overlay_info; + enum mdp_mixer_stage_id stage = STAGE_BASE; + int max_nb_planes; + + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); + max_nb_planes = hw_cfg->lm.nb_stages; + + if (count_planes(crtc) >= max_nb_planes) { + dev_err(dev, "%s: max # of planes (%d) reached\n", + mdp5_crtc->name, max_nb_planes); + return -EBUSY; + } + + /* + * Set default z-ordering depending on the type of plane + * private -> lower stage + * public -> topmost stage + * + * TODO: add a property to give userspace an API to change this... + * (will come in a subsequent patch) + */ + if (private_plane) { + stage = STAGE_BASE; + } else { + struct drm_plane *attached_plane; + for_each_plane_on_crtc(crtc, attached_plane) { + struct mdp5_overlay_info *overlay; + + if (!attached_plane) + continue; + overlay = mdp5_plane_get_overlay_info(attached_plane); + stage = max(stage, overlay->zorder); + } + stage++; + } + overlay_info.zorder = stage; + mdp5_plane_set_overlay_info(plane, &overlay_info); + + DBG("%s: %s plane %s set to stage %d by default", mdp5_crtc->name, + private_plane ? "private" : "public", + pipe2name(mdp5_plane_pipe(plane)), overlay_info.zorder); + set_attach(crtc, mdp5_plane_pipe(plane), plane); + + return 0; } void mdp5_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane) @@ -516,6 +646,16 @@ void mdp5_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane) set_attach(crtc, mdp5_plane_pipe(plane), NULL); } +int mdp5_crtc_get_lm(struct drm_crtc *crtc) +{ + struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); + + if (WARN_ON(!crtc)) + return -EINVAL; + + return mdp5_crtc->lm; +} + /* initialize crtc */ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev, struct drm_plane *plane, int id) @@ -530,6 +670,9 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev, crtc = &mdp5_crtc->base; mdp5_crtc->id = id; + mdp5_crtc->lm = GET_LM_ID(id); + + spin_lock_init(&mdp5_crtc->lm_lock); mdp5_crtc->vblank.irq = mdp5_crtc_vblank_irq; mdp5_crtc->err.irq = mdp5_crtc_err_irq; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c new file mode 100644 index 0000000..a6155b7 --- /dev/null +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c @@ -0,0 +1,325 @@ +/* + * Copyright (c) 2014 The Linux Foundation. All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 and + * only version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + */ + +#include "mdp5_kms.h" +#include "mdp5_ctl.h" + +/* + * CTL - MDP Control Pool Manager + * + * Controls are shared between all CRTCs. + * + * They are intended to be used for data path configuration. + * The top level register programming describes the complete data path for + * a specific data path ID - REG_MDP5_CTL_*(, ...) + * + * Hardware capabilities determine the number of concurrent data paths + * + * In certain use cases (high-resolution dual pipe), one single CTL can be + * shared across multiple CRTCs. + * + * Because the number of CTLs can be less than the number of CRTCs, + * CTLs are dynamically allocated from a pool of CTLs, only once a CRTC is + * requested by the client (in mdp5_crtc_mode_set()). + */ + +struct mdp5_ctl { + u32 id; + + /* whether this CTL has been allocated or not: */ + bool busy; + + /* memory output connection (@see mdp5_ctl_mode): */ + u32 mode; + + /* REG_MDP5_CTL_*() registers access info + lock: */ + spinlock_t hw_lock; + u32 reg_offset; + + /* flush mask used to commit CTL registers */ + u32 flush_mask; + + bool cursor_on; + void *crtc; +}; + +struct mdp5_ctl_manager { + struct drm_device *dev; + + /* number of CTL / Layer Mixers in this hw config: */ + u32 nlm; + u32 nctl; + + /* pool of CTLs + lock to protect resource allocation (ctls[i].busy) */ + spinlock_t pool_lock; + struct mdp5_ctl ctls[MAX_CTL]; +}; + +static struct mdp5_ctl_manager mdp5_ctl_mgr; + +static inline +struct mdp5_kms *get_kms(struct mdp5_ctl_manager *ctl_mgr) +{ + struct msm_drm_private *priv = ctl_mgr->dev->dev_private; + + return to_mdp5_kms(to_mdp_kms(priv->kms)); +} + +static inline +void ctl_write(struct mdp5_ctl *ctl, u32 reg, u32 data) +{ + struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_kms *mdp5_kms = get_kms(ctl_mgr); + + (void)ctl->reg_offset; /* TODO use this instead of mdp5_write */ + mdp5_write(mdp5_kms, reg, data); +} + +static inline +u32 ctl_read(struct mdp5_ctl *ctl, u32 reg) +{ + struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_kms *mdp5_kms = get_kms(ctl_mgr); + + (void)ctl->reg_offset; /* TODO use this instead of mdp5_write */ + return mdp5_read(mdp5_kms, reg); +} + + +int mdp5_ctl_set_intf(void *c, enum mdp5_intf intf) +{ + struct mdp5_ctl *ctl = c; + unsigned long flags; + static const enum mdp5_intfnum intfnum[] = { + INTF0, INTF1, INTF2, INTF3, + }; + + spin_lock_irqsave(&ctl->hw_lock, flags); + ctl_write(ctl, REG_MDP5_CTL_OP(ctl->id), + MDP5_CTL_OP_MODE(ctl->mode) | + MDP5_CTL_OP_INTF_NUM(intfnum[intf])); + spin_unlock_irqrestore(&ctl->hw_lock, flags); + + return 0; +} + +int mdp5_ctl_set_cursor(void *c, bool enable) +{ + struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_ctl *ctl = c; + unsigned long flags; + u32 blend_cfg; + int lm; + + lm = mdp5_crtc_get_lm(ctl->crtc); + if (unlikely(WARN_ON(lm < 0))) { + dev_err(ctl_mgr->dev->dev, "CTL %d cannot find LM: %d", + ctl->id, lm); + return -EINVAL; + } + + spin_lock_irqsave(&ctl->hw_lock, flags); + + blend_cfg = ctl_read(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, lm)); + + if (enable) + blend_cfg |= MDP5_CTL_LAYER_REG_CURSOR_OUT; + else + blend_cfg &= ~MDP5_CTL_LAYER_REG_CURSOR_OUT; + + ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, lm), blend_cfg); + + spin_unlock_irqrestore(&ctl->hw_lock, flags); + + ctl->cursor_on = enable; + + return 0; +} + + +int mdp5_ctl_blend(void *c, u32 lm, u32 blend_cfg) +{ + struct mdp5_ctl *ctl = c; + unsigned long flags; + + if (ctl->cursor_on) + blend_cfg |= MDP5_CTL_LAYER_REG_CURSOR_OUT; + else + blend_cfg &= ~MDP5_CTL_LAYER_REG_CURSOR_OUT; + + spin_lock_irqsave(&ctl->hw_lock, flags); + ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, lm), blend_cfg); + spin_unlock_irqrestore(&ctl->hw_lock, flags); + + return 0; +} + +int mdp5_ctl_commit(void *c, u32 flush_mask) +{ + struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_ctl *ctl = c; + unsigned long flags; + + if (flush_mask & MDP5_CTL_FLUSH_CURSOR_DUMMY) { + int lm = mdp5_crtc_get_lm(ctl->crtc); + + if (unlikely(WARN_ON(lm < 0))) { + dev_err(ctl_mgr->dev->dev, "CTL %d cannot find LM: %d", + ctl->id, lm); + return -EINVAL; + } + + /* for current targets, cursor bit is the same as LM bit */ + flush_mask |= mdp_ctl_flush_mask_lm(lm); + } + + spin_lock_irqsave(&ctl->hw_lock, flags); + ctl_write(ctl, REG_MDP5_CTL_FLUSH(ctl->id), flush_mask); + spin_unlock_irqrestore(&ctl->hw_lock, flags); + + return 0; +} + +u32 mdp5_ctl_get_flush(void *c) +{ + struct mdp5_ctl *ctl = c; + + return ctl->flush_mask; +} + +void mdp5_ctl_release(void *c) +{ + struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_ctl *ctl = c; + unsigned long flags; + + if (unlikely(WARN_ON(ctl->id >= MAX_CTL) || !ctl->busy)) { + dev_err(ctl_mgr->dev->dev, "CTL %d in bad state (%d)", + ctl->id, ctl->busy); + return; + } + + spin_lock_irqsave(&ctl_mgr->pool_lock, flags); + ctl->busy = false; + spin_unlock_irqrestore(&ctl_mgr->pool_lock, flags); + + DBG("CTL %d released", ctl->id); +} + +/* + * mdp5_ctl_request() - CTL dynamic allocation + * + * Note: Current implementation considers that we can only have one CRTC per CTL + * + * @return first free CTL + */ +void *mdp5_ctl_request(void *ctlm, void *crtc) +{ + struct mdp5_ctl_manager *ctl_mgr = ctlm; + struct mdp5_ctl *ctl = NULL; + unsigned long flags; + int c; + + spin_lock_irqsave(&ctl_mgr->pool_lock, flags); + + for (c = 0; c < ctl_mgr->nctl; c++) + if (!ctl_mgr->ctls[c].busy) + break; + + if (unlikely(c >= ctl_mgr->nctl)) { + dev_err(ctl_mgr->dev->dev, "No more CTL available!"); + goto unlock; + } + + ctl = &ctl_mgr->ctls[c]; + + ctl->crtc = crtc; + ctl->busy = true; + DBG("CTL %d allocated", ctl->id); + +unlock: + spin_unlock_irqrestore(&ctl_mgr->pool_lock, flags); + return ctl; +} + +void mdp5_ctlm_hw_reset(void *ctlm) +{ + struct mdp5_ctl_manager *ctl_mgr = ctlm; + unsigned long flags; + int c; + + for (c = 0; c < ctl_mgr->nctl; c++) { + struct mdp5_ctl *ctl = &ctl_mgr->ctls[c]; + + spin_lock_irqsave(&ctl->hw_lock, flags); + ctl_write(ctl, REG_MDP5_CTL_OP(ctl->id), 0); + spin_unlock_irqrestore(&ctl->hw_lock, flags); + } +} + +void mdp5_ctlm_destroy(void *ctlm) +{ + struct mdp5_ctl_manager *ctl_mgr = ctlm; + + kfree(ctl_mgr); +} + +void *mdp5_ctlm_init(struct drm_device *dev, void __iomem *mmio_base, + const struct mdp5_cfg_hw *hw_cfg) +{ + struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + const struct mdp5_sub_block *ctl_cfg = &hw_cfg->ctl; + unsigned long flags; + int c, ret; + + if (unlikely(WARN_ON(ctl_cfg->count > MAX_CTL))) { + dev_err(dev->dev, "Increase static pool size to at least %d\n", + ctl_cfg->count); + ret = -ENOSPC; + goto fail; + } + + /* initialize the CTL manager: */ + ctl_mgr->dev = dev; + ctl_mgr->nlm = hw_cfg->lm.count; + ctl_mgr->nctl = ctl_cfg->count; + spin_lock_init(&ctl_mgr->pool_lock); + + /* initialize each CTL of the pool: */ + spin_lock_irqsave(&ctl_mgr->pool_lock, flags); + for (c = 0; c < ctl_mgr->nctl; c++) { + struct mdp5_ctl *ctl = &ctl_mgr->ctls[c]; + + if (WARN_ON(!ctl_cfg->base[c])) { + dev_err(dev->dev, "CTL_%d: base is null!\n", c); + ret = -EINVAL; + goto fail; + } + ctl->id = c; + ctl->mode = MODE_NONE; + ctl->reg_offset = ctl_cfg->base[c]; + ctl->flush_mask = MDP5_CTL_FLUSH_CTL; + ctl->busy = false; + spin_lock_init(&ctl->hw_lock); + } + spin_unlock_irqrestore(&ctl_mgr->pool_lock, flags); + DBG("Pool of %d CTLs created.", ctl_mgr->nctl); + + return ctl_mgr; + +fail: + if (ctl_mgr) + mdp5_ctlm_destroy(ctl_mgr); + + return ERR_PTR(ret); +} diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h new file mode 100644 index 0000000..dbe1cae --- /dev/null +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h @@ -0,0 +1,121 @@ +/* + * Copyright (c) 2014 The Linux Foundation. All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 and + * only version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + */ + +#ifndef __MDP5_CTL_H__ +#define __MDP5_CTL_H__ + +#include "msm_drv.h" + +/* + * CTL Manager prototypes: + * mdp5_ctlm_init() returns a ctlm (CTL Manager) handler, + * which is then used to call the other mdp5_ctlm_*(ctlm, ...) functions. + */ +void *mdp5_ctlm_init(struct drm_device *dev, void __iomem *mmio_base, + const struct mdp5_cfg_hw *hw_cfg); +void mdp5_ctlm_hw_reset(void *ctlm); +void mdp5_ctlm_destroy(void *ctlm); + +/* + * CTL prototypes: + * mdp5_ctl_request(ctlm, ...) returns a ctl (CTL resource) handler, + * which is then used to call the other mdp5_ctl_*(ctl, ...) functions. + */ +void *mdp5_ctl_request(void *ctlm, void *crtc); + +int mdp5_ctl_set_intf(void *ctl, enum mdp5_intf intf); + +int mdp5_ctl_set_cursor(void *ctl, bool enable); + +/* @blend_cfg: see LM blender config definition below */ +int mdp5_ctl_blend(void *ctl, u32 lm, u32 blend_cfg); + +/* @flush_mask: see CTL flush masks definitions below */ +int mdp5_ctl_commit(void *ctl, u32 flush_mask); +u32 mdp5_ctl_get_flush(void *ctl); + +void mdp5_ctl_release(void *ctl); + +/* + * blend_cfg (LM blender config): + * + * The function below allows the caller of mdp5_ctl_blend() to specify how pipes + * are being blended according to their stage (z-order), through @blend_cfg arg. + */ +static inline u32 mdp_ctl_blend_mask(enum mdp5_pipe pipe, + enum mdp_mixer_stage_id stage) +{ + switch (pipe) { + case SSPP_VIG0: return MDP5_CTL_LAYER_REG_VIG0(stage); + case SSPP_VIG1: return MDP5_CTL_LAYER_REG_VIG1(stage); + case SSPP_VIG2: return MDP5_CTL_LAYER_REG_VIG2(stage); + case SSPP_RGB0: return MDP5_CTL_LAYER_REG_RGB0(stage); + case SSPP_RGB1: return MDP5_CTL_LAYER_REG_RGB1(stage); + case SSPP_RGB2: return MDP5_CTL_LAYER_REG_RGB2(stage); + case SSPP_DMA0: return MDP5_CTL_LAYER_REG_DMA0(stage); + case SSPP_DMA1: return MDP5_CTL_LAYER_REG_DMA1(stage); + case SSPP_VIG3: return MDP5_CTL_LAYER_REG_VIG3(stage); + case SSPP_RGB3: return MDP5_CTL_LAYER_REG_RGB3(stage); + default: return 0; + } +} + +/* + * flush_mask (CTL flush masks): + * + * The following functions allow each DRM entity to get and store + * their own flush mask. + * Once stored, these masks will then be accessed through each DRM's + * interface and used by the caller of mdp5_ctl_commit() to specify + * which block(s) need to be flushed through @flush_mask parameter. + */ + +#define MDP5_CTL_FLUSH_CURSOR_DUMMY 0x80000000 + +static inline u32 mdp_ctl_flush_mask_cursor(int cursor_id) +{ + /* TODO: use id once multiple cursor support is present */ + (void)cursor_id; + + return MDP5_CTL_FLUSH_CURSOR_DUMMY; +} + +static inline u32 mdp_ctl_flush_mask_lm(int lm) +{ + switch (lm) { + case 0: return MDP5_CTL_FLUSH_LM0; + case 1: return MDP5_CTL_FLUSH_LM1; + case 2: return MDP5_CTL_FLUSH_LM2; + case 5: return MDP5_CTL_FLUSH_LM5; + default: return 0; + } +} + +static inline u32 mdp_ctl_flush_mask_pipe(enum mdp5_pipe pipe) +{ + switch (pipe) { + case SSPP_VIG0: return MDP5_CTL_FLUSH_VIG0; + case SSPP_VIG1: return MDP5_CTL_FLUSH_VIG1; + case SSPP_VIG2: return MDP5_CTL_FLUSH_VIG2; + case SSPP_RGB0: return MDP5_CTL_FLUSH_RGB0; + case SSPP_RGB1: return MDP5_CTL_FLUSH_RGB1; + case SSPP_RGB2: return MDP5_CTL_FLUSH_RGB2; + case SSPP_DMA0: return MDP5_CTL_FLUSH_DMA0; + case SSPP_DMA1: return MDP5_CTL_FLUSH_DMA1; + case SSPP_VIG3: return MDP5_CTL_FLUSH_VIG3; + case SSPP_RGB3: return MDP5_CTL_FLUSH_RGB3; + default: return 0; + } +} + +#endif /* __MDP5_CTL_H__ */ diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c index edec7bf..25c2fcb 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c @@ -24,6 +24,7 @@ struct mdp5_encoder { struct drm_encoder base; int intf; enum mdp5_intf intf_id; + spinlock_t intf_lock; /* protect REG_MDP5_INTF_* registers */ bool enabled; uint32_t bsc; }; @@ -115,6 +116,7 @@ static void mdp5_encoder_dpms(struct drm_encoder *encoder, int mode) struct mdp5_kms *mdp5_kms = get_kms(encoder); int intf = mdp5_encoder->intf; bool enabled = (mode == DRM_MODE_DPMS_ON); + unsigned long flags; DBG("mode=%d", mode); @@ -123,9 +125,13 @@ static void mdp5_encoder_dpms(struct drm_encoder *encoder, int mode) if (enabled) { bs_set(mdp5_encoder, 1); + spin_lock_irqsave(&mdp5_encoder->intf_lock, flags); mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(intf), 1); + spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags); } else { + spin_lock_irqsave(&mdp5_encoder->intf_lock, flags); mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(intf), 0); + spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags); bs_set(mdp5_encoder, 0); } @@ -150,6 +156,7 @@ static void mdp5_encoder_mode_set(struct drm_encoder *encoder, uint32_t display_v_start, display_v_end; uint32_t hsync_start_x, hsync_end_x; uint32_t format; + unsigned long flags; mode = adjusted_mode; @@ -180,6 +187,8 @@ static void mdp5_encoder_mode_set(struct drm_encoder *encoder, display_v_start = (mode->vtotal - mode->vsync_start) * mode->htotal + dtv_hsync_skew; display_v_end = vsync_period - ((mode->vsync_start - mode->vdisplay) * mode->htotal) + dtv_hsync_skew - 1; + spin_lock_irqsave(&mdp5_encoder->intf_lock, flags); + mdp5_write(mdp5_kms, REG_MDP5_INTF_HSYNC_CTL(intf), MDP5_INTF_HSYNC_CTL_PULSEW(mode->hsync_end - mode->hsync_start) | MDP5_INTF_HSYNC_CTL_PERIOD(mode->htotal)); @@ -201,6 +210,8 @@ static void mdp5_encoder_mode_set(struct drm_encoder *encoder, mdp5_write(mdp5_kms, REG_MDP5_INTF_ACTIVE_VEND_F0(intf), 0); mdp5_write(mdp5_kms, REG_MDP5_INTF_PANEL_FORMAT(intf), format); mdp5_write(mdp5_kms, REG_MDP5_INTF_FRAME_LINE_COUNT_EN(intf), 0x3); /* frame+line? */ + + spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags); } static void mdp5_encoder_prepare(struct drm_encoder *encoder) @@ -242,6 +253,8 @@ struct drm_encoder *mdp5_encoder_init(struct drm_device *dev, int intf, mdp5_encoder->intf_id = intf_id; encoder = &mdp5_encoder->base; + spin_lock_init(&mdp5_encoder->intf_lock); + drm_encoder_init(dev, encoder, &mdp5_encoder_funcs, DRM_MODE_ENCODER_TMDS); drm_encoder_helper_add(encoder, &mdp5_encoder_helper_funcs); diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c index adb4541..da248c2 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c @@ -28,9 +28,8 @@ static const char *iommu_ports[] = { static int mdp5_hw_init(struct msm_kms *kms) { struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms)); - const struct mdp5_cfg_hw *hw_cfg; struct drm_device *dev = mdp5_kms->dev; - int i; + unsigned long flags; pm_runtime_get_sync(dev->dev); @@ -58,12 +57,11 @@ static int mdp5_hw_init(struct msm_kms *kms) * care. */ + spin_lock_irqsave(&mdp5_kms->resource_lock, flags); mdp5_write(mdp5_kms, REG_MDP5_DISP_INTF_SEL, 0); + spin_unlock_irqrestore(&mdp5_kms->resource_lock, flags); - hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); - - for (i = 0; i < hw_cfg->ctl.count; i++) - mdp5_write(mdp5_kms, REG_MDP5_CTL_OP(i), 0); + mdp5_ctlm_hw_reset(mdp5_kms->ctl_priv); pm_runtime_put_sync(dev->dev); @@ -92,6 +90,7 @@ static void mdp5_destroy(struct msm_kms *kms) struct msm_mmu *mmu = mdp5_kms->mmu; void *smp = mdp5_kms->smp_priv; void *cfg = mdp5_kms->cfg_priv; + void *ctl = mdp5_kms->ctl_priv; mdp5_irq_domain_fini(mdp5_kms); @@ -99,7 +98,8 @@ static void mdp5_destroy(struct msm_kms *kms) mmu->funcs->detach(mmu, iommu_ports, ARRAY_SIZE(iommu_ports)); mmu->funcs->destroy(mmu); } - + if (ctl) + mdp5_ctlm_destroy(ctl); if (smp) mdp5_smp_destroy(smp); if (cfg) @@ -154,6 +154,9 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) static const enum mdp5_pipe crtcs[] = { SSPP_RGB0, SSPP_RGB1, SSPP_RGB2, SSPP_RGB3, }; + static const enum mdp5_pipe pub_planes[] = { + SSPP_VIG0, SSPP_VIG1, SSPP_VIG2, SSPP_VIG3, + }; struct drm_device *dev = mdp5_kms->dev; struct msm_drm_private *priv = dev->dev_private; struct drm_encoder *encoder; @@ -169,12 +172,13 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) if (ret) goto fail; - /* construct CRTCs: */ + /* construct CRTCs and their private planes: */ for (i = 0; i < hw_cfg->pipe_rgb.count; i++) { struct drm_plane *plane; struct drm_crtc *crtc; - plane = mdp5_plane_init(dev, crtcs[i], true); + plane = mdp5_plane_init(dev, crtcs[i], true, + hw_cfg->pipe_rgb.base[i]); if (IS_ERR(plane)) { ret = PTR_ERR(plane); dev_err(dev->dev, "failed to construct plane for %s (%d)\n", @@ -192,6 +196,20 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) priv->crtcs[priv->num_crtcs++] = crtc; } + /* Construct public planes: */ + for (i = 0; i < hw_cfg->pipe_vig.count; i++) { + struct drm_plane *plane; + + plane = mdp5_plane_init(dev, pub_planes[i], false, + hw_cfg->pipe_vig.base[i]); + if (IS_ERR(plane)) { + ret = PTR_ERR(plane); + dev_err(dev->dev, "failed to construct %s plane: %d\n", + pipe2name(pub_planes[i]), ret); + goto fail; + } + } + /* Construct encoder for HDMI: */ encoder = mdp5_encoder_init(dev, 3, INTF_HDMI); if (IS_ERR(encoder)) { @@ -274,6 +292,8 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) goto fail; } + spin_lock_init(&mdp5_kms->resource_lock); + mdp_kms_init(&mdp5_kms->base, &kms_funcs); kms = &mdp5_kms->base.base; @@ -348,6 +368,13 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) } mdp5_kms->smp_priv = priv; + priv = mdp5_ctlm_init(dev, mdp5_kms->mmio, config->hw); + if (IS_ERR(priv)) { + ret = PTR_ERR(priv); + goto fail; + } + mdp5_kms->ctl_priv = priv; + /* make sure things are off before attaching iommu (bootloader could * have left things on, in which case we'll start getting faults if * we don't disable): diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h index daca8da..77fd43e 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h @@ -23,6 +23,7 @@ #include "mdp/mdp_kms.h" #include "mdp5_cfg.h" /* must be included before mdp5.xml.h */ #include "mdp5.xml.h" +#include "mdp5_ctl.h" #include "mdp5_smp.h" struct mdp5_kms { @@ -37,6 +38,7 @@ struct mdp5_kms { struct msm_mmu *mmu; void *smp_priv; + void *ctl_priv; /* io/register spaces: */ void __iomem *mmio, *vbif; @@ -50,6 +52,12 @@ struct mdp5_kms { struct clk *lut_clk; struct clk *vsync_clk; + /* + * lock to protect access to global resources: ie., following register: + * - REG_MDP5_DISP_INTF_SEL + */ + spinlock_t resource_lock; + struct mdp_irq error_handler; struct { @@ -59,6 +67,10 @@ struct mdp5_kms { }; #define to_mdp5_kms(x) container_of(x, struct mdp5_kms, base) +struct mdp5_overlay_info { + enum mdp_mixer_stage_id zorder; +}; + static inline void mdp5_write(struct mdp5_kms *mdp5_kms, u32 reg, u32 data) { msm_writel(data, mdp5_kms->mmio + reg); @@ -82,23 +94,6 @@ static inline const char *pipe2name(enum mdp5_pipe pipe) return names[pipe]; } -static inline uint32_t pipe2flush(enum mdp5_pipe pipe) -{ - switch (pipe) { - case SSPP_VIG0: return MDP5_CTL_FLUSH_VIG0; - case SSPP_VIG1: return MDP5_CTL_FLUSH_VIG1; - case SSPP_VIG2: return MDP5_CTL_FLUSH_VIG2; - case SSPP_RGB0: return MDP5_CTL_FLUSH_RGB0; - case SSPP_RGB1: return MDP5_CTL_FLUSH_RGB1; - case SSPP_RGB2: return MDP5_CTL_FLUSH_RGB2; - case SSPP_DMA0: return MDP5_CTL_FLUSH_DMA0; - case SSPP_DMA1: return MDP5_CTL_FLUSH_DMA1; - case SSPP_VIG3: return MDP5_CTL_FLUSH_VIG3; - case SSPP_RGB3: return MDP5_CTL_FLUSH_RGB3; - default: return 0; - } -} - static inline int pipe2nclients(enum mdp5_pipe pipe) { switch (pipe) { @@ -112,16 +107,6 @@ static inline int pipe2nclients(enum mdp5_pipe pipe) } } -static inline uint32_t mixer2flush(int lm) -{ - switch (lm) { - case 0: return MDP5_CTL_FLUSH_LM0; - case 1: return MDP5_CTL_FLUSH_LM1; - case 2: return MDP5_CTL_FLUSH_LM2; - default: return 0; - } -} - static inline uint32_t intf2err(int intf) { switch (intf) { @@ -169,6 +154,10 @@ uint32_t mdp5_get_formats(enum mdp5_pipe pipe, uint32_t *pixel_formats, void mdp5_plane_install_properties(struct drm_plane *plane, struct drm_mode_object *obj); +void mdp5_plane_set_overlay_info(struct drm_plane *plane, + const struct mdp5_overlay_info *overlay_info); +struct mdp5_overlay_info *mdp5_plane_get_overlay_info(struct drm_plane *plane); +uint32_t mdp5_plane_get_flush(struct drm_plane *plane); void mdp5_plane_set_scanout(struct drm_plane *plane, struct drm_framebuffer *fb); int mdp5_plane_mode_set(struct drm_plane *plane, @@ -180,14 +169,15 @@ int mdp5_plane_mode_set(struct drm_plane *plane, void mdp5_plane_complete_flip(struct drm_plane *plane); enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane); struct drm_plane *mdp5_plane_init(struct drm_device *dev, - enum mdp5_pipe pipe, bool private_plane); + enum mdp5_pipe pipe, bool private_plane, uint32_t reg_offset); uint32_t mdp5_crtc_vblank(struct drm_crtc *crtc); +int mdp5_crtc_get_lm(struct drm_crtc *crtc); void mdp5_crtc_cancel_pending_flip(struct drm_crtc *crtc, struct drm_file *file); void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf, enum mdp5_intf intf_id); -void mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane); +int mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane); void mdp5_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane); struct drm_crtc *mdp5_crtc_init(struct drm_device *dev, struct drm_plane *plane, int id); diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c index 633ca08..59703fa 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c @@ -1,4 +1,5 @@ /* + * Copyright (c) 2014 The Linux Foundation. All rights reserved. * Copyright (C) 2013 Red Hat * Author: Rob Clark * @@ -17,6 +18,7 @@ #include "mdp5_kms.h" +#define MAX_PLANE 4 struct mdp5_plane { struct drm_plane base; @@ -24,6 +26,13 @@ struct mdp5_plane { enum mdp5_pipe pipe; + spinlock_t pipe_lock; /* protect REG_MDP5_PIPE_* registers */ + uint32_t reg_offset; + + uint32_t flush_mask; /* used to commit pipe registers */ + + struct mdp5_overlay_info overlay_info; + uint32_t nformats; uint32_t formats[32]; @@ -95,6 +104,22 @@ static void mdp5_plane_destroy(struct drm_plane *plane) kfree(mdp5_plane); } +void mdp5_plane_set_overlay_info(struct drm_plane *plane, + const struct mdp5_overlay_info *overlay_info) +{ + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); + + memcpy(&mdp5_plane->overlay_info, overlay_info, sizeof(*overlay_info)); +} + +struct mdp5_overlay_info *mdp5_plane_get_overlay_info( + struct drm_plane *plane) +{ + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); + + return &mdp5_plane->overlay_info; +} + /* helper to install properties which are common to planes and crtcs */ void mdp5_plane_install_properties(struct drm_plane *plane, struct drm_mode_object *obj) @@ -116,35 +141,58 @@ static const struct drm_plane_funcs mdp5_plane_funcs = { .set_property = mdp5_plane_set_property, }; -void mdp5_plane_set_scanout(struct drm_plane *plane, - struct drm_framebuffer *fb) +static int get_fb_addr(struct drm_plane *plane, struct drm_framebuffer *fb, + uint32_t iova[MAX_PLANE]) { - struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); struct mdp5_kms *mdp5_kms = get_kms(plane); - enum mdp5_pipe pipe = mdp5_plane->pipe; uint32_t nplanes = drm_format_num_planes(fb->pixel_format); - uint32_t iova[4]; int i; for (i = 0; i < nplanes; i++) { struct drm_gem_object *bo = msm_framebuffer_bo(fb, i); msm_gem_get_iova(bo, mdp5_kms->id, &iova[i]); } - for (; i < 4; i++) + for (; i < MAX_PLANE; i++) iova[i] = 0; + return 0; +} + +static void set_scanout_locked(struct drm_plane *plane, + uint32_t pitches[MAX_PLANE], uint32_t src_addr[MAX_PLANE]) +{ + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); + struct mdp5_kms *mdp5_kms = get_kms(plane); + enum mdp5_pipe pipe = mdp5_plane->pipe; + + WARN_ON(!spin_is_locked(&mdp5_plane->pipe_lock)); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_A(pipe), - MDP5_PIPE_SRC_STRIDE_A_P0(fb->pitches[0]) | - MDP5_PIPE_SRC_STRIDE_A_P1(fb->pitches[1])); + MDP5_PIPE_SRC_STRIDE_A_P0(pitches[0]) | + MDP5_PIPE_SRC_STRIDE_A_P1(pitches[1])); mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_B(pipe), - MDP5_PIPE_SRC_STRIDE_B_P2(fb->pitches[2]) | - MDP5_PIPE_SRC_STRIDE_B_P3(fb->pitches[3])); + MDP5_PIPE_SRC_STRIDE_B_P2(pitches[2]) | + MDP5_PIPE_SRC_STRIDE_B_P3(pitches[3])); + + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC0_ADDR(pipe), src_addr[0]); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC1_ADDR(pipe), src_addr[1]); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC2_ADDR(pipe), src_addr[2]); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC3_ADDR(pipe), src_addr[3]); +} + +void mdp5_plane_set_scanout(struct drm_plane *plane, + struct drm_framebuffer *fb) +{ + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); + uint32_t src_addr[MAX_PLANE]; + unsigned long flags; - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC0_ADDR(pipe), iova[0]); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC1_ADDR(pipe), iova[1]); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC2_ADDR(pipe), iova[2]); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC3_ADDR(pipe), iova[3]); + get_fb_addr(plane, fb, src_addr); + + spin_lock_irqsave(&mdp5_plane->pipe_lock, flags); + set_scanout_locked(plane, fb->pitches, src_addr); + spin_unlock_irqrestore(&mdp5_plane->pipe_lock, flags); plane->fb = fb; } @@ -163,6 +211,8 @@ int mdp5_plane_mode_set(struct drm_plane *plane, uint32_t nplanes, config = 0; uint32_t phasex_step = 0, phasey_step = 0; uint32_t hdecm = 0, vdecm = 0; + uint32_t src_addr[MAX_PLANE]; + unsigned long flags; int ret; nplanes = drm_format_num_planes(fb->pixel_format); @@ -205,6 +255,12 @@ int mdp5_plane_mode_set(struct drm_plane *plane, /* TODO calc phasey_step, vdecm */ } + ret = get_fb_addr(plane, fb, src_addr); + if (ret) + return ret; + + spin_lock_irqsave(&mdp5_plane->pipe_lock, flags); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_IMG_SIZE(pipe), MDP5_PIPE_SRC_IMG_SIZE_WIDTH(src_w) | MDP5_PIPE_SRC_IMG_SIZE_HEIGHT(src_h)); @@ -225,8 +281,6 @@ int mdp5_plane_mode_set(struct drm_plane *plane, MDP5_PIPE_OUT_XY_X(crtc_x) | MDP5_PIPE_OUT_XY_Y(crtc_y)); - mdp5_plane_set_scanout(plane, fb); - format = to_mdp_format(msm_framebuffer_format(fb)); mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_FORMAT(pipe), @@ -266,10 +320,14 @@ int mdp5_plane_mode_set(struct drm_plane *plane, MDP5_PIPE_SCALE_CONFIG_SCALEX_MAX_FILTER(SCALE_FILTER_NEAREST) | MDP5_PIPE_SCALE_CONFIG_SCALEY_MAX_FILTER(SCALE_FILTER_NEAREST)); + set_scanout_locked(plane, fb->pitches, src_addr); + + spin_unlock_irqrestore(&mdp5_plane->pipe_lock, flags); + /* TODO detach from old crtc (if we had more than one) */ - mdp5_crtc_attach(crtc, plane); + ret = mdp5_crtc_attach(crtc, plane); - return 0; + return ret; } void mdp5_plane_complete_flip(struct drm_plane *plane) @@ -286,9 +344,16 @@ enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane) return mdp5_plane->pipe; } +uint32_t mdp5_plane_get_flush(struct drm_plane *plane) +{ + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); + + return mdp5_plane->flush_mask; +} + /* initialize plane */ struct drm_plane *mdp5_plane_init(struct drm_device *dev, - enum mdp5_pipe pipe, bool private_plane) + enum mdp5_pipe pipe, bool private_plane, uint32_t reg_offset) { struct drm_plane *plane = NULL; struct mdp5_plane *mdp5_plane; @@ -309,6 +374,10 @@ struct drm_plane *mdp5_plane_init(struct drm_device *dev, mdp5_plane->nformats = mdp5_get_formats(pipe, mdp5_plane->formats, ARRAY_SIZE(mdp5_plane->formats)); + mdp5_plane->flush_mask = mdp_ctl_flush_mask_pipe(pipe); + mdp5_plane->reg_offset = reg_offset; + spin_lock_init(&mdp5_plane->pipe_lock); + type = private_plane ? DRM_PLANE_TYPE_PRIMARY : DRM_PLANE_TYPE_OVERLAY; drm_universal_plane_init(dev, plane, 0xff, &mdp5_plane_funcs, mdp5_plane->formats, mdp5_plane->nformats, -- cgit v0.10.2 From 42238da8b10320420be19f056cc8e0d95cc606a3 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Tue, 18 Nov 2014 14:28:43 -0500 Subject: drm/msm/mdp5: don't use void * for opaque types For example, use 'struct mdp5_smp *' everywhere instead of 'void *', but only declare it as 'struct mdp5_smp;' in common headers, so the struct body is still private. The accomplishes the desired modularity while still letting the compiler provide some type checking for us. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c index 62e77d1..b0a4431 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c @@ -121,35 +121,27 @@ static const struct mdp5_cfg_handler cfg_handlers[] = { static struct mdp5_cfg_platform *mdp5_get_config(struct platform_device *dev); -const struct mdp5_cfg_hw *mdp5_cfg_get_hw_config(void *cfg_hnd) +const struct mdp5_cfg_hw *mdp5_cfg_get_hw_config(struct mdp5_cfg_handler *cfg_handler) { - struct mdp5_cfg_handler *cfg_handler = cfg_hnd; - return cfg_handler->config.hw; } -struct mdp5_cfg *mdp5_cfg_get_config(void *cfg_hnd) +struct mdp5_cfg *mdp5_cfg_get_config(struct mdp5_cfg_handler *cfg_handler) { - struct mdp5_cfg_handler *cfg_handler = cfg_hnd; - return &cfg_handler->config; } -int mdp5_cfg_get_hw_rev(void *cfg_hnd) +int mdp5_cfg_get_hw_rev(struct mdp5_cfg_handler *cfg_handler) { - struct mdp5_cfg_handler *cfg_handler = cfg_hnd; - return cfg_handler->revision; } -void mdp5_cfg_destroy(void *cfg_hnd) +void mdp5_cfg_destroy(struct mdp5_cfg_handler *cfg_handler) { - struct mdp5_cfg_handler *cfg_handler = cfg_hnd; - kfree(cfg_handler); } -void *mdp5_cfg_init(struct mdp5_kms *mdp5_kms, +struct mdp5_cfg_handler *mdp5_cfg_init(struct mdp5_kms *mdp5_kms, uint32_t major, uint32_t minor) { struct drm_device *dev = mdp5_kms->dev; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h index d0c98f9..dba4d52 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h @@ -78,12 +78,14 @@ struct mdp5_cfg { }; struct mdp5_kms; +struct mdp5_cfg_handler; -const struct mdp5_cfg_hw *mdp5_cfg_get_hw_config(void *cfg_hnd); -struct mdp5_cfg *mdp5_cfg_get_config(void *cfg_hnd); -int mdp5_cfg_get_hw_rev(void *cfg_hnd); +const struct mdp5_cfg_hw *mdp5_cfg_get_hw_config(struct mdp5_cfg_handler *cfg_hnd); +struct mdp5_cfg *mdp5_cfg_get_config(struct mdp5_cfg_handler *cfg_hnd); +int mdp5_cfg_get_hw_rev(struct mdp5_cfg_handler *cfg_hnd); -void *mdp5_cfg_init(struct mdp5_kms *mdp5_kms, uint32_t major, uint32_t minor); -void mdp5_cfg_destroy(void *cfg_hnd); +struct mdp5_cfg_handler *mdp5_cfg_init(struct mdp5_kms *mdp5_kms, + uint32_t major, uint32_t minor); +void mdp5_cfg_destroy(struct mdp5_cfg_handler *cfg_hnd); #endif /* __MDP5_CFG_H__ */ diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c index 85f2fb4..11c38fa 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c @@ -37,7 +37,7 @@ struct mdp5_crtc { spinlock_t lm_lock; /* protect REG_MDP5_LM_* registers */ /* CTL used for this CRTC: */ - void *ctl; + struct mdp5_ctl *ctl; /* if there is a pending flip, these will be non-null: */ struct drm_pending_vblank_event *event; @@ -261,7 +261,7 @@ static void blend_setup(struct drm_crtc *crtc) unsigned long flags; #define blender(stage) ((stage) - STAGE_BASE) - hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg); spin_lock_irqsave(&mdp5_crtc->lm_lock, flags); @@ -327,7 +327,7 @@ static int mdp5_crtc_mode_set(struct drm_crtc *crtc, /* request a free CTL, if none is already allocated for this CRTC */ if (!mdp5_crtc->ctl) { - mdp5_crtc->ctl = mdp5_ctl_request(mdp5_kms->ctl_priv, crtc); + mdp5_crtc->ctl = mdp5_ctlm_request(mdp5_kms->ctlm, crtc); if (!mdp5_crtc->ctl) return -EBUSY; } @@ -595,7 +595,7 @@ int mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane) enum mdp_mixer_stage_id stage = STAGE_BASE; int max_nb_planes; - hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg); max_nb_planes = hw_cfg->lm.nb_stages; if (count_planes(crtc) >= max_nb_planes) { diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c index a6155b7..c5347a7 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c @@ -50,7 +50,8 @@ struct mdp5_ctl { u32 flush_mask; bool cursor_on; - void *crtc; + + struct drm_crtc *crtc; }; struct mdp5_ctl_manager { @@ -96,9 +97,8 @@ u32 ctl_read(struct mdp5_ctl *ctl, u32 reg) } -int mdp5_ctl_set_intf(void *c, enum mdp5_intf intf) +int mdp5_ctl_set_intf(struct mdp5_ctl *ctl, enum mdp5_intf intf) { - struct mdp5_ctl *ctl = c; unsigned long flags; static const enum mdp5_intfnum intfnum[] = { INTF0, INTF1, INTF2, INTF3, @@ -113,10 +113,9 @@ int mdp5_ctl_set_intf(void *c, enum mdp5_intf intf) return 0; } -int mdp5_ctl_set_cursor(void *c, bool enable) +int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, bool enable) { struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; - struct mdp5_ctl *ctl = c; unsigned long flags; u32 blend_cfg; int lm; @@ -147,9 +146,8 @@ int mdp5_ctl_set_cursor(void *c, bool enable) } -int mdp5_ctl_blend(void *c, u32 lm, u32 blend_cfg) +int mdp5_ctl_blend(struct mdp5_ctl *ctl, u32 lm, u32 blend_cfg) { - struct mdp5_ctl *ctl = c; unsigned long flags; if (ctl->cursor_on) @@ -164,10 +162,9 @@ int mdp5_ctl_blend(void *c, u32 lm, u32 blend_cfg) return 0; } -int mdp5_ctl_commit(void *c, u32 flush_mask) +int mdp5_ctl_commit(struct mdp5_ctl *ctl, u32 flush_mask) { struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; - struct mdp5_ctl *ctl = c; unsigned long flags; if (flush_mask & MDP5_CTL_FLUSH_CURSOR_DUMMY) { @@ -190,17 +187,14 @@ int mdp5_ctl_commit(void *c, u32 flush_mask) return 0; } -u32 mdp5_ctl_get_flush(void *c) +u32 mdp5_ctl_get_flush(struct mdp5_ctl *ctl) { - struct mdp5_ctl *ctl = c; - return ctl->flush_mask; } -void mdp5_ctl_release(void *c) +void mdp5_ctl_release(struct mdp5_ctl *ctl) { struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; - struct mdp5_ctl *ctl = c; unsigned long flags; if (unlikely(WARN_ON(ctl->id >= MAX_CTL) || !ctl->busy)) { @@ -223,9 +217,9 @@ void mdp5_ctl_release(void *c) * * @return first free CTL */ -void *mdp5_ctl_request(void *ctlm, void *crtc) +struct mdp5_ctl *mdp5_ctlm_request(struct mdp5_ctl_manager *ctl_mgr, + struct drm_crtc *crtc) { - struct mdp5_ctl_manager *ctl_mgr = ctlm; struct mdp5_ctl *ctl = NULL; unsigned long flags; int c; @@ -252,9 +246,8 @@ unlock: return ctl; } -void mdp5_ctlm_hw_reset(void *ctlm) +void mdp5_ctlm_hw_reset(struct mdp5_ctl_manager *ctl_mgr) { - struct mdp5_ctl_manager *ctl_mgr = ctlm; unsigned long flags; int c; @@ -267,15 +260,13 @@ void mdp5_ctlm_hw_reset(void *ctlm) } } -void mdp5_ctlm_destroy(void *ctlm) +void mdp5_ctlm_destroy(struct mdp5_ctl_manager *ctl_mgr) { - struct mdp5_ctl_manager *ctl_mgr = ctlm; - kfree(ctl_mgr); } -void *mdp5_ctlm_init(struct drm_device *dev, void __iomem *mmio_base, - const struct mdp5_cfg_hw *hw_cfg) +struct mdp5_ctl_manager *mdp5_ctlm_init(struct drm_device *dev, + void __iomem *mmio_base, const struct mdp5_cfg_hw *hw_cfg) { struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; const struct mdp5_sub_block *ctl_cfg = &hw_cfg->ctl; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h index dbe1cae..1018519 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h @@ -21,30 +21,31 @@ * mdp5_ctlm_init() returns a ctlm (CTL Manager) handler, * which is then used to call the other mdp5_ctlm_*(ctlm, ...) functions. */ -void *mdp5_ctlm_init(struct drm_device *dev, void __iomem *mmio_base, - const struct mdp5_cfg_hw *hw_cfg); -void mdp5_ctlm_hw_reset(void *ctlm); -void mdp5_ctlm_destroy(void *ctlm); +struct mdp5_ctl_manager; +struct mdp5_ctl_manager *mdp5_ctlm_init(struct drm_device *dev, + void __iomem *mmio_base, const struct mdp5_cfg_hw *hw_cfg); +void mdp5_ctlm_hw_reset(struct mdp5_ctl_manager *ctlm); +void mdp5_ctlm_destroy(struct mdp5_ctl_manager *ctlm); /* * CTL prototypes: * mdp5_ctl_request(ctlm, ...) returns a ctl (CTL resource) handler, * which is then used to call the other mdp5_ctl_*(ctl, ...) functions. */ -void *mdp5_ctl_request(void *ctlm, void *crtc); +struct mdp5_ctl *mdp5_ctlm_request(struct mdp5_ctl_manager *ctlm, struct drm_crtc *crtc); -int mdp5_ctl_set_intf(void *ctl, enum mdp5_intf intf); +int mdp5_ctl_set_intf(struct mdp5_ctl *ctl, enum mdp5_intf intf); -int mdp5_ctl_set_cursor(void *ctl, bool enable); +int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, bool enable); /* @blend_cfg: see LM blender config definition below */ -int mdp5_ctl_blend(void *ctl, u32 lm, u32 blend_cfg); +int mdp5_ctl_blend(struct mdp5_ctl *ctl, u32 lm, u32 blend_cfg); /* @flush_mask: see CTL flush masks definitions below */ -int mdp5_ctl_commit(void *ctl, u32 flush_mask); -u32 mdp5_ctl_get_flush(void *ctl); +int mdp5_ctl_commit(struct mdp5_ctl *ctl, u32 flush_mask); +u32 mdp5_ctl_get_flush(struct mdp5_ctl *ctl); -void mdp5_ctl_release(void *ctl); +void mdp5_ctl_release(struct mdp5_ctl *ctl); /* * blend_cfg (LM blender config): diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c index da248c2..a11f1b8 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c @@ -61,7 +61,7 @@ static int mdp5_hw_init(struct msm_kms *kms) mdp5_write(mdp5_kms, REG_MDP5_DISP_INTF_SEL, 0); spin_unlock_irqrestore(&mdp5_kms->resource_lock, flags); - mdp5_ctlm_hw_reset(mdp5_kms->ctl_priv); + mdp5_ctlm_hw_reset(mdp5_kms->ctlm); pm_runtime_put_sync(dev->dev); @@ -88,9 +88,6 @@ static void mdp5_destroy(struct msm_kms *kms) { struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms)); struct msm_mmu *mmu = mdp5_kms->mmu; - void *smp = mdp5_kms->smp_priv; - void *cfg = mdp5_kms->cfg_priv; - void *ctl = mdp5_kms->ctl_priv; mdp5_irq_domain_fini(mdp5_kms); @@ -98,12 +95,13 @@ static void mdp5_destroy(struct msm_kms *kms) mmu->funcs->detach(mmu, iommu_ports, ARRAY_SIZE(iommu_ports)); mmu->funcs->destroy(mmu); } - if (ctl) - mdp5_ctlm_destroy(ctl); - if (smp) - mdp5_smp_destroy(smp); - if (cfg) - mdp5_cfg_destroy(cfg); + + if (mdp5_kms->ctlm) + mdp5_ctlm_destroy(mdp5_kms->ctlm); + if (mdp5_kms->smp) + mdp5_smp_destroy(mdp5_kms->smp); + if (mdp5_kms->cfg) + mdp5_cfg_destroy(mdp5_kms->cfg); kfree(mdp5_kms); } @@ -163,7 +161,7 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) const struct mdp5_cfg_hw *hw_cfg; int i, ret; - hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg); /* register our interrupt-controller for hdmi/eDP/dsi/etc * to use for irqs routed through mdp: @@ -282,7 +280,6 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) struct msm_kms *kms = NULL; struct msm_mmu *mmu; uint32_t major, minor; - void *priv; int i, ret; mdp5_kms = kzalloc(sizeof(*mdp5_kms), GFP_KERNEL); @@ -350,30 +347,32 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev) clk_set_rate(mdp5_kms->src_clk, 200000000); read_hw_revision(mdp5_kms, &major, &minor); - priv = mdp5_cfg_init(mdp5_kms, major, minor); - if (IS_ERR(priv)) { - ret = PTR_ERR(priv); + + mdp5_kms->cfg = mdp5_cfg_init(mdp5_kms, major, minor); + if (IS_ERR(mdp5_kms->cfg)) { + ret = PTR_ERR(mdp5_kms->cfg); + mdp5_kms->cfg = NULL; goto fail; } - mdp5_kms->cfg_priv = priv; - config = mdp5_cfg_get_config(mdp5_kms->cfg_priv); + + config = mdp5_cfg_get_config(mdp5_kms->cfg); /* TODO: compute core clock rate at runtime */ clk_set_rate(mdp5_kms->src_clk, config->hw->max_clk); - priv = mdp5_smp_init(mdp5_kms->dev, &config->hw->smp); - if (IS_ERR(priv)) { - ret = PTR_ERR(priv); + mdp5_kms->smp = mdp5_smp_init(mdp5_kms->dev, &config->hw->smp); + if (IS_ERR(mdp5_kms->smp)) { + ret = PTR_ERR(mdp5_kms->smp); + mdp5_kms->smp = NULL; goto fail; } - mdp5_kms->smp_priv = priv; - priv = mdp5_ctlm_init(dev, mdp5_kms->mmio, config->hw); - if (IS_ERR(priv)) { - ret = PTR_ERR(priv); + mdp5_kms->ctlm = mdp5_ctlm_init(dev, mdp5_kms->mmio, config->hw); + if (IS_ERR(mdp5_kms->ctlm)) { + ret = PTR_ERR(mdp5_kms->ctlm); + mdp5_kms->ctlm = NULL; goto fail; } - mdp5_kms->ctl_priv = priv; /* make sure things are off before attaching iommu (bootloader could * have left things on, in which case we'll start getting faults if diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h index 77fd43e..414447c 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h @@ -31,14 +31,14 @@ struct mdp5_kms { struct drm_device *dev; - void *cfg_priv; + struct mdp5_cfg_handler *cfg; /* mapper-id used to request GEM buffer mapped for scanout: */ int id; struct msm_mmu *mmu; - void *smp_priv; - void *ctl_priv; + struct mdp5_smp *smp; + struct mdp5_ctl_manager *ctlm; /* io/register spaces: */ void __iomem *mmio, *vbif; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c index 59703fa..49c10a9 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c @@ -77,7 +77,7 @@ static int mdp5_plane_disable(struct drm_plane *plane) if (mdp5_kms) { /* Release the memory we requested earlier from the SMP: */ - mdp5_smp_release(mdp5_kms->smp_priv, pipe); + mdp5_smp_release(mdp5_kms->smp, pipe); } /* TODO detaching now will cause us not to get the last @@ -232,7 +232,7 @@ int mdp5_plane_mode_set(struct drm_plane *plane, crtc->base.id, crtc_x, crtc_y, crtc_w, crtc_h); /* Request some memory from the SMP: */ - ret = mdp5_smp_request(mdp5_kms->smp_priv, + ret = mdp5_smp_request(mdp5_kms->smp, mdp5_plane->pipe, fb->pixel_format, src_w); if (ret) return ret; @@ -243,7 +243,7 @@ int mdp5_plane_mode_set(struct drm_plane *plane, * would move into atomic->check_plane_state(), while updating the * hw would remain here: */ - mdp5_smp_configure(mdp5_kms->smp_priv, pipe); + mdp5_smp_configure(mdp5_kms->smp, pipe); if (src_w != crtc_w) { config |= MDP5_PIPE_SCALE_CONFIG_SCALEX_EN; @@ -335,7 +335,7 @@ void mdp5_plane_complete_flip(struct drm_plane *plane) struct mdp5_kms *mdp5_kms = get_kms(plane); enum mdp5_pipe pipe = to_mdp5_plane(plane)->pipe; - mdp5_smp_commit(mdp5_kms->smp_priv, pipe); + mdp5_smp_commit(mdp5_kms->smp, pipe); } enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane) diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c index 04996ca..bf55188 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c @@ -114,7 +114,7 @@ static int smp_request_block(struct mdp5_smp *smp, int reserved; unsigned long flags; - hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg_priv); + hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg); reserved = hw_cfg->smp.reserved[cid]; spin_lock_irqsave(&smp->state_lock, flags); @@ -174,12 +174,11 @@ static void set_fifo_thresholds(struct mdp5_smp *smp, * decimated width. Ie. SMP buffering sits downstream of decimation (which * presumably happens during the dma from scanout buffer). */ -int mdp5_smp_request(void *handler, enum mdp5_pipe pipe, u32 fmt, u32 width) +int mdp5_smp_request(struct mdp5_smp *smp, enum mdp5_pipe pipe, u32 fmt, u32 width) { - struct mdp5_smp *smp = handler; struct mdp5_kms *mdp5_kms = get_kms(smp); struct drm_device *dev = mdp5_kms->dev; - int rev = mdp5_cfg_get_hw_rev(mdp5_kms->cfg_priv); + int rev = mdp5_cfg_get_hw_rev(mdp5_kms->cfg); int i, hsub, nplanes, nlines, nblks, ret; nplanes = drm_format_num_planes(fmt); @@ -217,9 +216,8 @@ int mdp5_smp_request(void *handler, enum mdp5_pipe pipe, u32 fmt, u32 width) } /* Release SMP blocks for all clients of the pipe */ -void mdp5_smp_release(void *handler, enum mdp5_pipe pipe) +void mdp5_smp_release(struct mdp5_smp *smp, enum mdp5_pipe pipe) { - struct mdp5_smp *smp = handler; int i, nblks; for (i = 0, nblks = 0; i < pipe2nclients(pipe); i++) @@ -261,9 +259,8 @@ static void update_smp_state(struct mdp5_smp *smp, } /* step #2: configure hw for union(pending, inuse): */ -void mdp5_smp_configure(void *handler, enum mdp5_pipe pipe) +void mdp5_smp_configure(struct mdp5_smp *smp, enum mdp5_pipe pipe) { - struct mdp5_smp *smp = handler; int cnt = smp->blk_cnt; mdp5_smp_state_t assigned; int i; @@ -278,9 +275,8 @@ void mdp5_smp_configure(void *handler, enum mdp5_pipe pipe) } /* step #3: after vblank, copy pending -> inuse: */ -void mdp5_smp_commit(void *handler, enum mdp5_pipe pipe) +void mdp5_smp_commit(struct mdp5_smp *smp, enum mdp5_pipe pipe) { - struct mdp5_smp *smp = handler; int cnt = smp->blk_cnt; mdp5_smp_state_t released; int i; @@ -309,14 +305,12 @@ void mdp5_smp_commit(void *handler, enum mdp5_pipe pipe) } } -void mdp5_smp_destroy(void *handler) +void mdp5_smp_destroy(struct mdp5_smp *smp) { - struct mdp5_smp *smp = handler; - kfree(smp); } -void *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg) +struct mdp5_smp *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg) { struct mdp5_smp *smp = NULL; int ret; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h index 3e3c753..e47179f 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h @@ -27,6 +27,7 @@ struct mdp5_client_smp_state { }; struct mdp5_kms; +struct mdp5_smp; /* * SMP module prototypes: @@ -34,12 +35,12 @@ struct mdp5_kms; * which is then used to call the other mdp5_smp_*(handler, ...) functions. */ -void *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg); -void mdp5_smp_destroy(void *handler); +struct mdp5_smp *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg); +void mdp5_smp_destroy(struct mdp5_smp *smp); -int mdp5_smp_request(void *handler, enum mdp5_pipe pipe, u32 fmt, u32 width); -void mdp5_smp_configure(void *handler, enum mdp5_pipe pipe); -void mdp5_smp_commit(void *handler, enum mdp5_pipe pipe); -void mdp5_smp_release(void *handler, enum mdp5_pipe pipe); +int mdp5_smp_request(struct mdp5_smp *smp, enum mdp5_pipe pipe, u32 fmt, u32 width); +void mdp5_smp_configure(struct mdp5_smp *smp, enum mdp5_pipe pipe); +void mdp5_smp_commit(struct mdp5_smp *smp, enum mdp5_pipe pipe); +void mdp5_smp_release(struct mdp5_smp *smp, enum mdp5_pipe pipe); #endif /* __MDP5_SMP_H__ */ -- cgit v0.10.2 From 32c0e3e24c885ffeecb6d3a67ecb607b8fe63129 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Tue, 18 Nov 2014 17:22:51 -0500 Subject: drm/msm/mdp5: remove global mdp5_ctl_mgr A bit cleaner.. and won't resulting in an attempt to kfree() a static global in unload path. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c index c5347a7..dea4505 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c @@ -34,6 +34,8 @@ */ struct mdp5_ctl { + struct mdp5_ctl_manager *ctlm; + u32 id; /* whether this CTL has been allocated or not: */ @@ -66,8 +68,6 @@ struct mdp5_ctl_manager { struct mdp5_ctl ctls[MAX_CTL]; }; -static struct mdp5_ctl_manager mdp5_ctl_mgr; - static inline struct mdp5_kms *get_kms(struct mdp5_ctl_manager *ctl_mgr) { @@ -79,8 +79,7 @@ struct mdp5_kms *get_kms(struct mdp5_ctl_manager *ctl_mgr) static inline void ctl_write(struct mdp5_ctl *ctl, u32 reg, u32 data) { - struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; - struct mdp5_kms *mdp5_kms = get_kms(ctl_mgr); + struct mdp5_kms *mdp5_kms = get_kms(ctl->ctlm); (void)ctl->reg_offset; /* TODO use this instead of mdp5_write */ mdp5_write(mdp5_kms, reg, data); @@ -89,8 +88,7 @@ void ctl_write(struct mdp5_ctl *ctl, u32 reg, u32 data) static inline u32 ctl_read(struct mdp5_ctl *ctl, u32 reg) { - struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; - struct mdp5_kms *mdp5_kms = get_kms(ctl_mgr); + struct mdp5_kms *mdp5_kms = get_kms(ctl->ctlm); (void)ctl->reg_offset; /* TODO use this instead of mdp5_write */ return mdp5_read(mdp5_kms, reg); @@ -115,7 +113,7 @@ int mdp5_ctl_set_intf(struct mdp5_ctl *ctl, enum mdp5_intf intf) int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, bool enable) { - struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm; unsigned long flags; u32 blend_cfg; int lm; @@ -164,7 +162,7 @@ int mdp5_ctl_blend(struct mdp5_ctl *ctl, u32 lm, u32 blend_cfg) int mdp5_ctl_commit(struct mdp5_ctl *ctl, u32 flush_mask) { - struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm; unsigned long flags; if (flush_mask & MDP5_CTL_FLUSH_CURSOR_DUMMY) { @@ -194,7 +192,7 @@ u32 mdp5_ctl_get_flush(struct mdp5_ctl *ctl) void mdp5_ctl_release(struct mdp5_ctl *ctl) { - struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm; unsigned long flags; if (unlikely(WARN_ON(ctl->id >= MAX_CTL) || !ctl->busy)) { @@ -268,11 +266,18 @@ void mdp5_ctlm_destroy(struct mdp5_ctl_manager *ctl_mgr) struct mdp5_ctl_manager *mdp5_ctlm_init(struct drm_device *dev, void __iomem *mmio_base, const struct mdp5_cfg_hw *hw_cfg) { - struct mdp5_ctl_manager *ctl_mgr = &mdp5_ctl_mgr; + struct mdp5_ctl_manager *ctl_mgr; const struct mdp5_sub_block *ctl_cfg = &hw_cfg->ctl; unsigned long flags; int c, ret; + ctl_mgr = kzalloc(sizeof(*ctl_mgr), GFP_KERNEL); + if (!ctl_mgr) { + dev_err(dev->dev, "failed to allocate CTL manager\n"); + ret = -ENOMEM; + goto fail; + } + if (unlikely(WARN_ON(ctl_cfg->count > MAX_CTL))) { dev_err(dev->dev, "Increase static pool size to at least %d\n", ctl_cfg->count); @@ -296,6 +301,7 @@ struct mdp5_ctl_manager *mdp5_ctlm_init(struct drm_device *dev, ret = -EINVAL; goto fail; } + ctl->ctlm = ctl_mgr; ctl->id = c; ctl->mode = MODE_NONE; ctl->reg_offset = ctl_cfg->base[c]; -- cgit v0.10.2 From 3e2f29e4137c78a3bd7e1b85cf287ff375f6367e Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Wed, 19 Nov 2014 12:29:33 -0500 Subject: drm/msm: atomic fixes Fixes for a couple little issues found in testing. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/msm_atomic.c b/drivers/gpu/drm/msm/msm_atomic.c index 8cf3361..f0de412 100644 --- a/drivers/gpu/drm/msm/msm_atomic.c +++ b/drivers/gpu/drm/msm/msm_atomic.c @@ -117,7 +117,7 @@ int msm_atomic_commit(struct drm_device *dev, if (!plane) continue; - if (plane->state->fb != new_state->fb) + if ((plane->state->fb != new_state->fb) && new_state->fb) add_fb(c, new_state->fb); } diff --git a/drivers/gpu/drm/msm/msm_fb.c b/drivers/gpu/drm/msm/msm_fb.c index f4e42d5..84dec16 100644 --- a/drivers/gpu/drm/msm/msm_fb.c +++ b/drivers/gpu/drm/msm/msm_fb.c @@ -120,6 +120,8 @@ void msm_framebuffer_cleanup(struct drm_framebuffer *fb, int id) uint32_t msm_framebuffer_iova(struct drm_framebuffer *fb, int id, int plane) { struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb); + if (!msm_fb->planes[plane]) + return 0; return msm_gem_iova(msm_fb->planes[plane], id); } diff --git a/drivers/gpu/drm/msm/msm_kms.h b/drivers/gpu/drm/msm/msm_kms.h index 15a0fec9..1e9d8c2 100644 --- a/drivers/gpu/drm/msm/msm_kms.h +++ b/drivers/gpu/drm/msm/msm_kms.h @@ -68,6 +68,11 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev); /* TODO move these helper iterator macro somewhere common: */ #define for_each_plane_on_crtc(_crtc, _plane) \ list_for_each_entry((_plane), &(_crtc)->dev->mode_config.plane_list, head) \ - if ((_plane)->crtc == (_crtc)) + if ((_plane)->state->crtc == (_crtc)) + +#define for_each_pending_plane_on_crtc(_state, _crtc, _plane) \ + list_for_each_entry((_plane), &(_crtc)->dev->mode_config.plane_list, head) \ + if (({struct drm_plane_state *_ps = (_state)->plane_states[drm_plane_index(_plane)]; \ + _ps && _ps->crtc == (_crtc);})) #endif /* __MSM_KMS_H__ */ -- cgit v0.10.2 From ed8519636ef6c4b6ffbc4a449ebf39502e80ee74 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Wed, 19 Nov 2014 12:31:03 -0500 Subject: drm/msm/mdp5: atomic Convert mdp5 over to atomic helpers. Extend/wrap drm_plane_state to track plane zpos and to keep track of the needed when applying the atomic update. In mdp5's plane->atomic_check() we also need to check for updates which require SMP reallocation, in order to trigger full modeset. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c index 11c38fa..a6728a8 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c @@ -18,6 +18,7 @@ #include "mdp5_kms.h" +#include #include #include "drm_crtc.h" #include "drm_crtc_helper.h" @@ -41,25 +42,11 @@ struct mdp5_crtc { /* if there is a pending flip, these will be non-null: */ struct drm_pending_vblank_event *event; - struct msm_fence_cb pageflip_cb; #define PENDING_CURSOR 0x1 #define PENDING_FLIP 0x2 atomic_t pending; - /* the fb that we logically (from PoV of KMS API) hold a ref - * to. Which we may not yet be scanning out (we may still - * be scanning out previous in case of page_flip while waiting - * for gpu rendering to complete: - */ - struct drm_framebuffer *fb; - - /* the fb that we currently hold a scanout ref to: */ - struct drm_framebuffer *scanout_fb; - - /* for unref'ing framebuffers after scanout completes: */ - struct drm_flip_work unref_fb_work; - struct mdp_irq vblank; struct mdp_irq err; }; @@ -113,41 +100,6 @@ static void crtc_flush_all(struct drm_crtc *crtc) crtc_flush(crtc, flush_mask); } -static void update_fb(struct drm_crtc *crtc, struct drm_framebuffer *new_fb) -{ - struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); - struct drm_framebuffer *old_fb = mdp5_crtc->fb; - - /* grab reference to incoming scanout fb: */ - drm_framebuffer_reference(new_fb); - mdp5_crtc->base.primary->fb = new_fb; - mdp5_crtc->fb = new_fb; - - if (old_fb) - drm_flip_work_queue(&mdp5_crtc->unref_fb_work, old_fb); -} - -/* unlike update_fb(), take a ref to the new scanout fb *before* updating - * plane, then call this. Needed to ensure we don't unref the buffer that - * is actually still being scanned out. - * - * Note that this whole thing goes away with atomic.. since we can defer - * calling into driver until rendering is done. - */ -static void update_scanout(struct drm_crtc *crtc, struct drm_framebuffer *fb) -{ - struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); - - if (mdp5_crtc->scanout_fb) - drm_flip_work_queue(&mdp5_crtc->unref_fb_work, - mdp5_crtc->scanout_fb); - - mdp5_crtc->scanout_fb = fb; - - /* enable vblank to complete flip: */ - request_pending(crtc, PENDING_FLIP); -} - /* if file!=NULL, this is preclose potential cancel-flip path */ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file) { @@ -166,6 +118,7 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file) */ if (!file || (event->base.file_priv == file)) { mdp5_crtc->event = NULL; + DBG("%s: send event: %p", mdp5_crtc->name, event); drm_send_vblank_event(dev, mdp5_crtc->id, event); } } @@ -175,39 +128,11 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file) mdp5_plane_complete_flip(plane); } -static void pageflip_cb(struct msm_fence_cb *cb) -{ - struct mdp5_crtc *mdp5_crtc = - container_of(cb, struct mdp5_crtc, pageflip_cb); - struct drm_crtc *crtc = &mdp5_crtc->base; - struct drm_framebuffer *fb = mdp5_crtc->fb; - - if (!fb) - return; - - drm_framebuffer_reference(fb); - mdp5_plane_set_scanout(crtc->primary, fb); - update_scanout(crtc, fb); - crtc_flush_all(crtc); -} - -static void unref_fb_worker(struct drm_flip_work *work, void *val) -{ - struct mdp5_crtc *mdp5_crtc = - container_of(work, struct mdp5_crtc, unref_fb_work); - struct drm_device *dev = mdp5_crtc->base.dev; - - mutex_lock(&dev->mode_config.mutex); - drm_framebuffer_unreference(val); - mutex_unlock(&dev->mode_config.mutex); -} - static void mdp5_crtc_destroy(struct drm_crtc *crtc) { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); drm_crtc_cleanup(crtc); - drm_flip_work_cleanup(&mdp5_crtc->unref_fb_work); kfree(mdp5_crtc); } @@ -257,7 +182,6 @@ static void blend_setup(struct drm_crtc *crtc) struct drm_plane *plane; const struct mdp5_cfg_hw *hw_cfg; uint32_t lm = mdp5_crtc->lm, blend_cfg = 0; - enum mdp_mixer_stage_id stage; unsigned long flags; #define blender(stage) ((stage) - STAGE_BASE) @@ -270,10 +194,8 @@ static void blend_setup(struct drm_crtc *crtc) goto out; for_each_plane_on_crtc(crtc, plane) { - struct mdp5_overlay_info *overlay; - - overlay = mdp5_plane_get_overlay_info(plane); - stage = overlay->zorder; + enum mdp_mixer_stage_id stage = + to_mdp5_plane_state(plane->state)->stage; /* * Note: This cannot happen with current implementation but @@ -303,18 +225,17 @@ out: spin_unlock_irqrestore(&mdp5_crtc->lm_lock, flags); } -static int mdp5_crtc_mode_set(struct drm_crtc *crtc, - struct drm_display_mode *mode, - struct drm_display_mode *adjusted_mode, - int x, int y, - struct drm_framebuffer *old_fb) +static void mdp5_crtc_mode_set_nofb(struct drm_crtc *crtc) { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); struct mdp5_kms *mdp5_kms = get_kms(crtc); unsigned long flags; - int ret; + struct drm_display_mode *mode; + + if (WARN_ON(!crtc->state)) + return; - mode = adjusted_mode; + mode = &crtc->state->adjusted_mode; DBG("%s: set mode: %d:\"%s\" %d %d %d %d %d %d %d %d %d %d 0x%x 0x%x", mdp5_crtc->name, mode->base.id, mode->name, @@ -325,38 +246,11 @@ static int mdp5_crtc_mode_set(struct drm_crtc *crtc, mode->vsync_end, mode->vtotal, mode->type, mode->flags); - /* request a free CTL, if none is already allocated for this CRTC */ - if (!mdp5_crtc->ctl) { - mdp5_crtc->ctl = mdp5_ctlm_request(mdp5_kms->ctlm, crtc); - if (!mdp5_crtc->ctl) - return -EBUSY; - } - - /* grab extra ref for update_scanout() */ - drm_framebuffer_reference(crtc->primary->fb); - - ret = mdp5_plane_mode_set(crtc->primary, crtc, crtc->primary->fb, - 0, 0, mode->hdisplay, mode->vdisplay, - x << 16, y << 16, - mode->hdisplay << 16, mode->vdisplay << 16); - if (ret) { - drm_framebuffer_unreference(crtc->primary->fb); - dev_err(crtc->dev->dev, "%s: failed to set mode on plane: %d\n", - mdp5_crtc->name, ret); - return ret; - } - spin_lock_irqsave(&mdp5_crtc->lm_lock, flags); mdp5_write(mdp5_kms, REG_MDP5_LM_OUT_SIZE(mdp5_crtc->lm), MDP5_LM_OUT_SIZE_WIDTH(mode->hdisplay) | MDP5_LM_OUT_SIZE_HEIGHT(mode->vdisplay)); spin_unlock_irqrestore(&mdp5_crtc->lm_lock, flags); - - update_fb(crtc, crtc->primary->fb); - update_scanout(crtc, crtc->primary->fb); - /* crtc_flush_all(crtc) will be called in _commit callback */ - - return 0; } static void mdp5_crtc_prepare(struct drm_crtc *crtc) @@ -370,79 +264,119 @@ static void mdp5_crtc_prepare(struct drm_crtc *crtc) static void mdp5_crtc_commit(struct drm_crtc *crtc) { + struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); + DBG("%s", mdp5_crtc->name); mdp5_crtc_dpms(crtc, DRM_MODE_DPMS_ON); crtc_flush_all(crtc); /* drop the ref to mdp clk's that we got in prepare: */ mdp5_disable(get_kms(crtc)); } -static int mdp5_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, - struct drm_framebuffer *old_fb) +static void mdp5_crtc_load_lut(struct drm_crtc *crtc) { - struct drm_plane *plane = crtc->primary; - struct drm_display_mode *mode = &crtc->mode; - int ret; - - /* grab extra ref for update_scanout() */ - drm_framebuffer_reference(crtc->primary->fb); - - ret = mdp5_plane_mode_set(plane, crtc, crtc->primary->fb, - 0, 0, mode->hdisplay, mode->vdisplay, - x << 16, y << 16, - mode->hdisplay << 16, mode->vdisplay << 16); - if (ret) { - drm_framebuffer_unreference(crtc->primary->fb); - return ret; - } - - update_fb(crtc, crtc->primary->fb); - update_scanout(crtc, crtc->primary->fb); - crtc_flush_all(crtc); - - return 0; } -static void mdp5_crtc_load_lut(struct drm_crtc *crtc) +struct plane_state { + struct drm_plane *plane; + struct mdp5_plane_state *state; +}; + +static int pstate_cmp(const void *a, const void *b) { + struct plane_state *pa = (struct plane_state *)a; + struct plane_state *pb = (struct plane_state *)b; + return pa->state->zpos - pb->state->zpos; } -static void mdp5_crtc_disable(struct drm_crtc *crtc) +static int mdp5_crtc_atomic_check(struct drm_crtc *crtc, + struct drm_crtc_state *state) { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); + struct mdp5_kms *mdp5_kms = get_kms(crtc); + struct drm_plane *plane; + struct drm_device *dev = crtc->dev; + struct plane_state pstates[STAGE3 + 1]; + int cnt = 0, i; - DBG("%s", mdp5_crtc->name); + DBG("%s: check", mdp5_crtc->name); - if (mdp5_crtc->ctl) { - mdp5_ctl_release(mdp5_crtc->ctl); - mdp5_crtc->ctl = NULL; + if (mdp5_crtc->event) { + dev_err(dev->dev, "already pending flip!\n"); + return -EBUSY; + } + + /* request a free CTL, if none is already allocated for this CRTC */ + if (state->enable && !mdp5_crtc->ctl) { + mdp5_crtc->ctl = mdp5_ctlm_request(mdp5_kms->ctlm, crtc); + if (WARN_ON(!mdp5_crtc->ctl)) + return -EINVAL; } + + /* verify that there are not too many planes attached to crtc + * and that we don't have conflicting mixer stages: + */ + for_each_pending_plane_on_crtc(state->state, crtc, plane) { + struct drm_plane_state *pstate; + + if (cnt >= ARRAY_SIZE(pstates)) { + dev_err(dev->dev, "too many planes!\n"); + return -EINVAL; + } + + pstate = state->state->plane_states[drm_plane_index(plane)]; + + /* plane might not have changed, in which case take + * current state: + */ + if (!pstate) + pstate = plane->state; + + pstates[cnt].plane = plane; + pstates[cnt].state = to_mdp5_plane_state(pstate); + + cnt++; + } + + sort(pstates, cnt, sizeof(pstates[0]), pstate_cmp, NULL); + + for (i = 0; i < cnt; i++) { + pstates[i].state->stage = STAGE_BASE + i; + DBG("%s: assign pipe %s on stage=%d", mdp5_crtc->name, + pipe2name(mdp5_plane_pipe(pstates[i].plane)), + pstates[i].state->stage); + } + + return 0; } +static void mdp5_crtc_atomic_begin(struct drm_crtc *crtc) +{ + struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); + DBG("%s: begin", mdp5_crtc->name); +} -static int mdp5_crtc_page_flip(struct drm_crtc *crtc, - struct drm_framebuffer *new_fb, - struct drm_pending_vblank_event *event, - uint32_t page_flip_flags) +static void mdp5_crtc_atomic_flush(struct drm_crtc *crtc) { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); struct drm_device *dev = crtc->dev; - struct drm_gem_object *obj; unsigned long flags; - if (mdp5_crtc->event) { - dev_err(dev->dev, "already pending flip!\n"); - return -EBUSY; - } + DBG("%s: flush", mdp5_crtc->name); - obj = msm_framebuffer_bo(new_fb, 0); + WARN_ON(mdp5_crtc->event); spin_lock_irqsave(&dev->event_lock, flags); - mdp5_crtc->event = event; + mdp5_crtc->event = crtc->state->event; spin_unlock_irqrestore(&dev->event_lock, flags); - update_fb(crtc, new_fb); + blend_setup(crtc); + crtc_flush_all(crtc); + request_pending(crtc, PENDING_FLIP); - return msm_gem_queue_inactive_cb(obj, &mdp5_crtc->pageflip_cb); + if (mdp5_crtc->ctl && !crtc->state->enable) { + mdp5_ctl_release(mdp5_crtc->ctl); + mdp5_crtc->ctl = NULL; + } } static int mdp5_crtc_set_property(struct drm_crtc *crtc, @@ -453,28 +387,33 @@ static int mdp5_crtc_set_property(struct drm_crtc *crtc, } static const struct drm_crtc_funcs mdp5_crtc_funcs = { - .set_config = drm_crtc_helper_set_config, + .set_config = drm_atomic_helper_set_config, .destroy = mdp5_crtc_destroy, - .page_flip = mdp5_crtc_page_flip, + .page_flip = drm_atomic_helper_page_flip, .set_property = mdp5_crtc_set_property, + .reset = drm_atomic_helper_crtc_reset, + .atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state, + .atomic_destroy_state = drm_atomic_helper_crtc_destroy_state, }; static const struct drm_crtc_helper_funcs mdp5_crtc_helper_funcs = { .dpms = mdp5_crtc_dpms, .mode_fixup = mdp5_crtc_mode_fixup, - .mode_set = mdp5_crtc_mode_set, + .mode_set_nofb = mdp5_crtc_mode_set_nofb, + .mode_set = drm_helper_crtc_mode_set, + .mode_set_base = drm_helper_crtc_mode_set_base, .prepare = mdp5_crtc_prepare, .commit = mdp5_crtc_commit, - .mode_set_base = mdp5_crtc_mode_set_base, .load_lut = mdp5_crtc_load_lut, - .disable = mdp5_crtc_disable, + .atomic_check = mdp5_crtc_atomic_check, + .atomic_begin = mdp5_crtc_atomic_begin, + .atomic_flush = mdp5_crtc_atomic_flush, }; static void mdp5_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus) { struct mdp5_crtc *mdp5_crtc = container_of(irq, struct mdp5_crtc, vblank); struct drm_crtc *crtc = &mdp5_crtc->base; - struct msm_drm_private *priv = crtc->dev->dev_private; unsigned pending; mdp_irq_unregister(&get_kms(crtc)->base, &mdp5_crtc->vblank); @@ -483,7 +422,6 @@ static void mdp5_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus) if (pending & PENDING_FLIP) { complete_flip(crtc, NULL); - drm_flip_work_commit(&mdp5_crtc->unref_fb_work, priv->wq); } } @@ -560,92 +498,6 @@ void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf, crtc_flush(crtc, flush_mask); } -static int count_planes(struct drm_crtc *crtc) -{ - struct drm_plane *plane; - int cnt = 0; - for_each_plane_on_crtc(crtc, plane) - cnt++; - return cnt; -} - -static void set_attach(struct drm_crtc *crtc, enum mdp5_pipe pipe_id, - struct drm_plane *plane) -{ - struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); - - if (plane) - plane->crtc = crtc; - - DBG("%s: %d planes attached", mdp5_crtc->name, count_planes(crtc)); - - blend_setup(crtc); - if (mdp5_crtc->enabled) - crtc_flush_all(crtc); -} - -int mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane) -{ - struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); - struct mdp5_kms *mdp5_kms = get_kms(crtc); - struct device *dev = crtc->dev->dev; - const struct mdp5_cfg_hw *hw_cfg; - bool private_plane = (plane == crtc->primary); - struct mdp5_overlay_info overlay_info; - enum mdp_mixer_stage_id stage = STAGE_BASE; - int max_nb_planes; - - hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg); - max_nb_planes = hw_cfg->lm.nb_stages; - - if (count_planes(crtc) >= max_nb_planes) { - dev_err(dev, "%s: max # of planes (%d) reached\n", - mdp5_crtc->name, max_nb_planes); - return -EBUSY; - } - - /* - * Set default z-ordering depending on the type of plane - * private -> lower stage - * public -> topmost stage - * - * TODO: add a property to give userspace an API to change this... - * (will come in a subsequent patch) - */ - if (private_plane) { - stage = STAGE_BASE; - } else { - struct drm_plane *attached_plane; - for_each_plane_on_crtc(crtc, attached_plane) { - struct mdp5_overlay_info *overlay; - - if (!attached_plane) - continue; - overlay = mdp5_plane_get_overlay_info(attached_plane); - stage = max(stage, overlay->zorder); - } - stage++; - } - overlay_info.zorder = stage; - mdp5_plane_set_overlay_info(plane, &overlay_info); - - DBG("%s: %s plane %s set to stage %d by default", mdp5_crtc->name, - private_plane ? "private" : "public", - pipe2name(mdp5_plane_pipe(plane)), overlay_info.zorder); - - set_attach(crtc, mdp5_plane_pipe(plane), plane); - - return 0; -} - -void mdp5_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane) -{ - /* don't actually detatch our primary plane: */ - if (crtc->primary == plane) - return; - set_attach(crtc, mdp5_plane_pipe(plane), NULL); -} - int mdp5_crtc_get_lm(struct drm_crtc *crtc) { struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc); @@ -680,11 +532,6 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev, snprintf(mdp5_crtc->name, sizeof(mdp5_crtc->name), "%s:%d", pipe2name(mdp5_plane_pipe(plane)), id); - drm_flip_work_init(&mdp5_crtc->unref_fb_work, - "unref fb", unref_fb_worker); - - INIT_FENCE_CB(&mdp5_crtc->pageflip_cb, pageflip_cb); - drm_crtc_init_with_planes(dev, crtc, plane, NULL, &mdp5_crtc_funcs); drm_crtc_helper_add(crtc, &mdp5_crtc_helper_funcs); plane->crtc = crtc; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h index 414447c..dd69c77 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h @@ -67,9 +67,32 @@ struct mdp5_kms { }; #define to_mdp5_kms(x) container_of(x, struct mdp5_kms, base) -struct mdp5_overlay_info { - enum mdp_mixer_stage_id zorder; +struct mdp5_plane_state { + struct drm_plane_state base; + + /* "virtual" zpos.. we calculate actual mixer-stage at runtime + * by sorting the attached planes by zpos and then assigning + * mixer stage lowest to highest. Private planes get default + * zpos of zero, and public planes a unique value that is + * greater than zero. This way, things work out if a naive + * userspace assigns planes to a crtc without setting zpos. + */ + int zpos; + + /* the actual mixer stage, calculated in crtc->atomic_check() + * NOTE: this should move to mdp5_crtc_state, when that exists + */ + enum mdp_mixer_stage_id stage; + + /* some additional transactional status to help us know in the + * apply path whether we need to update SMP allocation, and + * whether current update is still pending: + */ + bool mode_changed : 1; + bool pending : 1; }; +#define to_mdp5_plane_state(x) \ + container_of(x, struct mdp5_plane_state, base) static inline void mdp5_write(struct mdp5_kms *mdp5_kms, u32 reg, u32 data) { @@ -154,18 +177,7 @@ uint32_t mdp5_get_formats(enum mdp5_pipe pipe, uint32_t *pixel_formats, void mdp5_plane_install_properties(struct drm_plane *plane, struct drm_mode_object *obj); -void mdp5_plane_set_overlay_info(struct drm_plane *plane, - const struct mdp5_overlay_info *overlay_info); -struct mdp5_overlay_info *mdp5_plane_get_overlay_info(struct drm_plane *plane); uint32_t mdp5_plane_get_flush(struct drm_plane *plane); -void mdp5_plane_set_scanout(struct drm_plane *plane, - struct drm_framebuffer *fb); -int mdp5_plane_mode_set(struct drm_plane *plane, - struct drm_crtc *crtc, struct drm_framebuffer *fb, - int crtc_x, int crtc_y, - unsigned int crtc_w, unsigned int crtc_h, - uint32_t src_x, uint32_t src_y, - uint32_t src_w, uint32_t src_h); void mdp5_plane_complete_flip(struct drm_plane *plane); enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane); struct drm_plane *mdp5_plane_init(struct drm_device *dev, @@ -177,8 +189,6 @@ int mdp5_crtc_get_lm(struct drm_crtc *crtc); void mdp5_crtc_cancel_pending_flip(struct drm_crtc *crtc, struct drm_file *file); void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf, enum mdp5_intf intf_id); -int mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane); -void mdp5_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane); struct drm_crtc *mdp5_crtc_init(struct drm_device *dev, struct drm_plane *plane, int id); diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c index 49c10a9..533df7c 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c @@ -31,8 +31,6 @@ struct mdp5_plane { uint32_t flush_mask; /* used to commit pipe registers */ - struct mdp5_overlay_info overlay_info; - uint32_t nformats; uint32_t formats[32]; @@ -40,31 +38,24 @@ struct mdp5_plane { }; #define to_mdp5_plane(x) container_of(x, struct mdp5_plane, base) +static int mdp5_plane_mode_set(struct drm_plane *plane, + struct drm_crtc *crtc, struct drm_framebuffer *fb, + int crtc_x, int crtc_y, + unsigned int crtc_w, unsigned int crtc_h, + uint32_t src_x, uint32_t src_y, + uint32_t src_w, uint32_t src_h); +static void set_scanout_locked(struct drm_plane *plane, + struct drm_framebuffer *fb); + static struct mdp5_kms *get_kms(struct drm_plane *plane) { struct msm_drm_private *priv = plane->dev->dev_private; return to_mdp5_kms(to_mdp_kms(priv->kms)); } -static int mdp5_plane_update(struct drm_plane *plane, - struct drm_crtc *crtc, struct drm_framebuffer *fb, - int crtc_x, int crtc_y, - unsigned int crtc_w, unsigned int crtc_h, - uint32_t src_x, uint32_t src_y, - uint32_t src_w, uint32_t src_h) +static bool plane_enabled(struct drm_plane_state *state) { - struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); - - mdp5_plane->enabled = true; - - if (plane->fb) - drm_framebuffer_unreference(plane->fb); - - drm_framebuffer_reference(fb); - - return mdp5_plane_mode_set(plane, crtc, fb, - crtc_x, crtc_y, crtc_w, crtc_h, - src_x, src_y, src_w, src_h); + return state->fb && state->crtc; } static int mdp5_plane_disable(struct drm_plane *plane) @@ -80,46 +71,19 @@ static int mdp5_plane_disable(struct drm_plane *plane) mdp5_smp_release(mdp5_kms->smp, pipe); } - /* TODO detaching now will cause us not to get the last - * vblank and mdp5_smp_commit().. so other planes will - * still see smp blocks previously allocated to us as - * in-use.. - */ - if (plane->crtc) - mdp5_crtc_detach(plane->crtc, plane); - return 0; } static void mdp5_plane_destroy(struct drm_plane *plane) { struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); - struct msm_drm_private *priv = plane->dev->dev_private; - - if (priv->kms) - mdp5_plane_disable(plane); + drm_plane_helper_disable(plane); drm_plane_cleanup(plane); kfree(mdp5_plane); } -void mdp5_plane_set_overlay_info(struct drm_plane *plane, - const struct mdp5_overlay_info *overlay_info) -{ - struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); - - memcpy(&mdp5_plane->overlay_info, overlay_info, sizeof(*overlay_info)); -} - -struct mdp5_overlay_info *mdp5_plane_get_overlay_info( - struct drm_plane *plane) -{ - struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); - - return &mdp5_plane->overlay_info; -} - /* helper to install properties which are common to planes and crtcs */ void mdp5_plane_install_properties(struct drm_plane *plane, struct drm_mode_object *obj) @@ -134,70 +98,185 @@ int mdp5_plane_set_property(struct drm_plane *plane, return -EINVAL; } +static void mdp5_plane_reset(struct drm_plane *plane) +{ + struct mdp5_plane_state *mdp5_state; + + if (plane->state && plane->state->fb) + drm_framebuffer_unreference(plane->state->fb); + + kfree(to_mdp5_plane_state(plane->state)); + mdp5_state = kzalloc(sizeof(*mdp5_state), GFP_KERNEL); + + if (plane->type == DRM_PLANE_TYPE_PRIMARY) { + mdp5_state->zpos = 0; + } else { + mdp5_state->zpos = 1 + drm_plane_index(plane); + } + + plane->state = &mdp5_state->base; +} + +static struct drm_plane_state * +mdp5_plane_duplicate_state(struct drm_plane *plane) +{ + struct mdp5_plane_state *mdp5_state; + + if (WARN_ON(!plane->state)) + return NULL; + + mdp5_state = kmemdup(to_mdp5_plane_state(plane->state), + sizeof(*mdp5_state), GFP_KERNEL); + + if (mdp5_state && mdp5_state->base.fb) + drm_framebuffer_reference(mdp5_state->base.fb); + + mdp5_state->mode_changed = false; + mdp5_state->pending = false; + + return &mdp5_state->base; +} + +static void mdp5_plane_destroy_state(struct drm_plane *plane, + struct drm_plane_state *state) +{ + if (state->fb) + drm_framebuffer_unreference(state->fb); + + kfree(to_mdp5_plane_state(state)); +} + static const struct drm_plane_funcs mdp5_plane_funcs = { - .update_plane = mdp5_plane_update, - .disable_plane = mdp5_plane_disable, + .update_plane = drm_atomic_helper_update_plane, + .disable_plane = drm_atomic_helper_disable_plane, .destroy = mdp5_plane_destroy, .set_property = mdp5_plane_set_property, + .reset = mdp5_plane_reset, + .atomic_duplicate_state = mdp5_plane_duplicate_state, + .atomic_destroy_state = mdp5_plane_destroy_state, }; -static int get_fb_addr(struct drm_plane *plane, struct drm_framebuffer *fb, - uint32_t iova[MAX_PLANE]) +static int mdp5_plane_prepare_fb(struct drm_plane *plane, + struct drm_framebuffer *fb) { + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); struct mdp5_kms *mdp5_kms = get_kms(plane); - uint32_t nplanes = drm_format_num_planes(fb->pixel_format); - int i; - for (i = 0; i < nplanes; i++) { - struct drm_gem_object *bo = msm_framebuffer_bo(fb, i); - msm_gem_get_iova(bo, mdp5_kms->id, &iova[i]); - } - for (; i < MAX_PLANE; i++) - iova[i] = 0; - - return 0; + DBG("%s: prepare: FB[%u]", mdp5_plane->name, fb->base.id); + return msm_framebuffer_prepare(fb, mdp5_kms->id); } -static void set_scanout_locked(struct drm_plane *plane, - uint32_t pitches[MAX_PLANE], uint32_t src_addr[MAX_PLANE]) +static void mdp5_plane_cleanup_fb(struct drm_plane *plane, + struct drm_framebuffer *fb) { struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); struct mdp5_kms *mdp5_kms = get_kms(plane); - enum mdp5_pipe pipe = mdp5_plane->pipe; - WARN_ON(!spin_is_locked(&mdp5_plane->pipe_lock)); + DBG("%s: cleanup: FB[%u]", mdp5_plane->name, fb->base.id); + msm_framebuffer_cleanup(fb, mdp5_kms->id); +} - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_A(pipe), - MDP5_PIPE_SRC_STRIDE_A_P0(pitches[0]) | - MDP5_PIPE_SRC_STRIDE_A_P1(pitches[1])); +static int mdp5_plane_atomic_check(struct drm_plane *plane, + struct drm_plane_state *state) +{ + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); + struct drm_plane_state *old_state = plane->state; + + DBG("%s: check (%d -> %d)", mdp5_plane->name, + plane_enabled(old_state), plane_enabled(state)); + + if (plane_enabled(state) && plane_enabled(old_state)) { + /* we cannot change SMP block configuration during scanout: */ + bool full_modeset = false; + if (state->fb->pixel_format != old_state->fb->pixel_format) { + DBG("%s: pixel_format change!", mdp5_plane->name); + full_modeset = true; + } + if (state->src_w != old_state->src_w) { + DBG("%s: src_w change!", mdp5_plane->name); + full_modeset = true; + } + if (to_mdp5_plane_state(old_state)->pending) { + DBG("%s: still pending!", mdp5_plane->name); + full_modeset = true; + } + if (full_modeset) { + struct drm_crtc_state *crtc_state = + drm_atomic_get_crtc_state(state->state, state->crtc); + crtc_state->mode_changed = true; + to_mdp5_plane_state(state)->mode_changed = true; + } + } else { + to_mdp5_plane_state(state)->mode_changed = true; + } - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_B(pipe), - MDP5_PIPE_SRC_STRIDE_B_P2(pitches[2]) | - MDP5_PIPE_SRC_STRIDE_B_P3(pitches[3])); + return 0; +} + +static void mdp5_plane_atomic_update(struct drm_plane *plane) +{ + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); + struct drm_plane_state *state = plane->state; + + DBG("%s: update", mdp5_plane->name); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC0_ADDR(pipe), src_addr[0]); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC1_ADDR(pipe), src_addr[1]); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC2_ADDR(pipe), src_addr[2]); - mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC3_ADDR(pipe), src_addr[3]); + if (!plane_enabled(state)) { + to_mdp5_plane_state(state)->pending = true; + mdp5_plane_disable(plane); + } else if (to_mdp5_plane_state(state)->mode_changed) { + int ret; + to_mdp5_plane_state(state)->pending = true; + ret = mdp5_plane_mode_set(plane, + state->crtc, state->fb, + state->crtc_x, state->crtc_y, + state->crtc_w, state->crtc_h, + state->src_x, state->src_y, + state->src_w, state->src_h); + /* atomic_check should have ensured that this doesn't fail */ + WARN_ON(ret < 0); + } else { + unsigned long flags; + spin_lock_irqsave(&mdp5_plane->pipe_lock, flags); + set_scanout_locked(plane, state->fb); + spin_unlock_irqrestore(&mdp5_plane->pipe_lock, flags); + } } -void mdp5_plane_set_scanout(struct drm_plane *plane, +static const struct drm_plane_helper_funcs mdp5_plane_helper_funcs = { + .prepare_fb = mdp5_plane_prepare_fb, + .cleanup_fb = mdp5_plane_cleanup_fb, + .atomic_check = mdp5_plane_atomic_check, + .atomic_update = mdp5_plane_atomic_update, +}; + +static void set_scanout_locked(struct drm_plane *plane, struct drm_framebuffer *fb) { struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); - uint32_t src_addr[MAX_PLANE]; - unsigned long flags; + struct mdp5_kms *mdp5_kms = get_kms(plane); + enum mdp5_pipe pipe = mdp5_plane->pipe; - get_fb_addr(plane, fb, src_addr); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_A(pipe), + MDP5_PIPE_SRC_STRIDE_A_P0(fb->pitches[0]) | + MDP5_PIPE_SRC_STRIDE_A_P1(fb->pitches[1])); - spin_lock_irqsave(&mdp5_plane->pipe_lock, flags); - set_scanout_locked(plane, fb->pitches, src_addr); - spin_unlock_irqrestore(&mdp5_plane->pipe_lock, flags); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_B(pipe), + MDP5_PIPE_SRC_STRIDE_B_P2(fb->pitches[2]) | + MDP5_PIPE_SRC_STRIDE_B_P3(fb->pitches[3])); + + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC0_ADDR(pipe), + msm_framebuffer_iova(fb, mdp5_kms->id, 0)); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC1_ADDR(pipe), + msm_framebuffer_iova(fb, mdp5_kms->id, 1)); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC2_ADDR(pipe), + msm_framebuffer_iova(fb, mdp5_kms->id, 2)); + mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC3_ADDR(pipe), + msm_framebuffer_iova(fb, mdp5_kms->id, 4)); plane->fb = fb; } -int mdp5_plane_mode_set(struct drm_plane *plane, +static int mdp5_plane_mode_set(struct drm_plane *plane, struct drm_crtc *crtc, struct drm_framebuffer *fb, int crtc_x, int crtc_y, unsigned int crtc_w, unsigned int crtc_h, @@ -211,7 +290,6 @@ int mdp5_plane_mode_set(struct drm_plane *plane, uint32_t nplanes, config = 0; uint32_t phasex_step = 0, phasey_step = 0; uint32_t hdecm = 0, vdecm = 0; - uint32_t src_addr[MAX_PLANE]; unsigned long flags; int ret; @@ -255,10 +333,6 @@ int mdp5_plane_mode_set(struct drm_plane *plane, /* TODO calc phasey_step, vdecm */ } - ret = get_fb_addr(plane, fb, src_addr); - if (ret) - return ret; - spin_lock_irqsave(&mdp5_plane->pipe_lock, flags); mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_IMG_SIZE(pipe), @@ -320,22 +394,24 @@ int mdp5_plane_mode_set(struct drm_plane *plane, MDP5_PIPE_SCALE_CONFIG_SCALEX_MAX_FILTER(SCALE_FILTER_NEAREST) | MDP5_PIPE_SCALE_CONFIG_SCALEY_MAX_FILTER(SCALE_FILTER_NEAREST)); - set_scanout_locked(plane, fb->pitches, src_addr); + set_scanout_locked(plane, fb); spin_unlock_irqrestore(&mdp5_plane->pipe_lock, flags); - /* TODO detach from old crtc (if we had more than one) */ - ret = mdp5_crtc_attach(crtc, plane); - return ret; } void mdp5_plane_complete_flip(struct drm_plane *plane) { struct mdp5_kms *mdp5_kms = get_kms(plane); - enum mdp5_pipe pipe = to_mdp5_plane(plane)->pipe; + struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); + enum mdp5_pipe pipe = mdp5_plane->pipe; + + DBG("%s: complete flip", mdp5_plane->name); mdp5_smp_commit(mdp5_kms->smp, pipe); + + to_mdp5_plane_state(plane->state)->pending = false; } enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane) @@ -379,9 +455,13 @@ struct drm_plane *mdp5_plane_init(struct drm_device *dev, spin_lock_init(&mdp5_plane->pipe_lock); type = private_plane ? DRM_PLANE_TYPE_PRIMARY : DRM_PLANE_TYPE_OVERLAY; - drm_universal_plane_init(dev, plane, 0xff, &mdp5_plane_funcs, + ret = drm_universal_plane_init(dev, plane, 0xff, &mdp5_plane_funcs, mdp5_plane->formats, mdp5_plane->nformats, type); + if (ret) + goto fail; + + drm_plane_helper_add(plane, &mdp5_plane_helper_funcs); mdp5_plane_install_properties(plane, &plane->base); diff --git a/drivers/gpu/drm/msm/msm_kms.h b/drivers/gpu/drm/msm/msm_kms.h index 1e9d8c2..7fb4876 100644 --- a/drivers/gpu/drm/msm/msm_kms.h +++ b/drivers/gpu/drm/msm/msm_kms.h @@ -70,9 +70,22 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev); list_for_each_entry((_plane), &(_crtc)->dev->mode_config.plane_list, head) \ if ((_plane)->state->crtc == (_crtc)) +static inline bool +__plane_will_be_attached_to_crtc(struct drm_atomic_state *state, + struct drm_plane *plane, struct drm_crtc *crtc) +{ + int idx = drm_plane_index(plane); + + /* if plane is modified in incoming state, use the new state: */ + if (state->plane_states[idx]) + return state->plane_states[idx]->crtc == crtc; + + /* otherwise, current state: */ + return plane->state->crtc == crtc; +} + #define for_each_pending_plane_on_crtc(_state, _crtc, _plane) \ list_for_each_entry((_plane), &(_crtc)->dev->mode_config.plane_list, head) \ - if (({struct drm_plane_state *_ps = (_state)->plane_states[drm_plane_index(_plane)]; \ - _ps && _ps->crtc == (_crtc);})) + if (__plane_will_be_attached_to_crtc((_state), (_plane), (_crtc))) #endif /* __MSM_KMS_H__ */ -- cgit v0.10.2 From d6ac4a84e7a3713929ec473cb8c10acc1dec41a2 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Thu, 20 Nov 2014 17:05:04 -0500 Subject: drm/msm/mdp5: dpms(OFF) cleanups When disabling the interface (INTF), the change doesn't latch until next vblank, so we need to wait for vblank. Also, to be pedantic, in the crtc, set all the mixer stages to unused. It shouldn't really matter, since at this point we have already disabled the INTF and waited for necessary vblank. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c index a6728a8..0598bde 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c @@ -150,6 +150,8 @@ static void mdp5_crtc_dpms(struct drm_crtc *crtc, int mode) mdp5_enable(mdp5_kms); mdp_irq_register(&mdp5_kms->base, &mdp5_crtc->err); } else { + /* set STAGE_UNUSED for all layers */ + mdp5_ctl_blend(mdp5_crtc->ctl, mdp5_crtc->lm, 0x00000000); mdp_irq_unregister(&mdp5_kms->base, &mdp5_crtc->err); mdp5_disable(mdp5_kms); } diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c index 25c2fcb..0254bfd 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c @@ -132,6 +132,17 @@ static void mdp5_encoder_dpms(struct drm_encoder *encoder, int mode) spin_lock_irqsave(&mdp5_encoder->intf_lock, flags); mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(intf), 0); spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags); + + /* + * Wait for a vsync so we know the ENABLE=0 latched before + * the (connector) source of the vsync's gets disabled, + * otherwise we end up in a funny state if we re-enable + * before the disable latches, which results that some of + * the settings changes for the new modeset (like new + * scanout buffer) don't latch properly.. + */ + mdp_irq_wait(&mdp5_kms->base, intf2vblank(intf)); + bs_set(mdp5_encoder, 0); } -- cgit v0.10.2 From 824cb7d13638ebacb3f5c85b01a2aeec6fe14081 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Fri, 21 Nov 2014 22:04:44 +0200 Subject: amdkfd: explicitely include io.h in kfd_doorbell.c This patch fixes a compilation error when using certain configuration by including the file io.h in kfd_doorbell.c Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c index 0dcb787..b5791a5 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c @@ -23,6 +23,7 @@ #include #include #include +#include /* * This extension supports a kernel level doorbells management for -- cgit v0.10.2 From 4dd14fe6fb8119d1d8c87127ebc74c63deb2ffc8 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 21 Nov 2014 11:18:01 -0500 Subject: drm/msm/mdp4: fix mixer setup for multi-crtc + planes On mdp4 there is a single global LAYERMIXER_IN_CFG register. The previous logic to share that between multiple crtcs didn't actually handle plane-disable very well. Easier just to look at all of the crtcs each time. Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c index fef22e8..6781aa9 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c @@ -167,34 +167,54 @@ static bool mdp4_crtc_mode_fixup(struct drm_crtc *crtc, return true; } -static void blend_setup(struct drm_crtc *crtc) +/* statically (for now) map planes to mixer stage (z-order): */ +static const int idxs[] = { + [VG1] = 1, + [VG2] = 2, + [RGB1] = 0, + [RGB2] = 0, + [RGB3] = 0, + [VG3] = 3, + [VG4] = 4, + +}; + +/* setup mixer config, for which we need to consider all crtc's and + * the planes attached to them + * + * TODO may possibly need some extra locking here + */ +static void setup_mixer(struct mdp4_kms *mdp4_kms) { - struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); - struct mdp4_kms *mdp4_kms = get_kms(crtc); - struct drm_plane *plane; - int i, ovlp = mdp4_crtc->ovlp; + struct drm_mode_config *config = &mdp4_kms->dev->mode_config; + struct drm_crtc *crtc; uint32_t mixer_cfg = 0; static const enum mdp_mixer_stage_id stages[] = { STAGE_BASE, STAGE0, STAGE1, STAGE2, STAGE3, }; - /* statically (for now) map planes to mixer stage (z-order): */ - static const int idxs[] = { - [VG1] = 1, - [VG2] = 2, - [RGB1] = 0, - [RGB2] = 0, - [RGB3] = 0, - [VG3] = 3, - [VG4] = 4, - }; - bool alpha[4]= { false, false, false, false }; + list_for_each_entry(crtc, &config->crtc_list, head) { + struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); + struct drm_plane *plane; - /* Don't rely on value read back from hw, but instead use our - * own shadowed value. Possibly disable/reenable looses the - * previous value and goes back to power-on default? - */ - mixer_cfg = mdp4_kms->mixer_cfg; + for_each_plane_on_crtc(crtc, plane) { + enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane); + int idx = idxs[pipe_id]; + mixer_cfg = mixercfg(mixer_cfg, mdp4_crtc->mixer, + pipe_id, stages[idx]); + } + } + + mdp4_write(mdp4_kms, REG_MDP4_LAYERMIXER_IN_CFG, mixer_cfg); +} + +static void blend_setup(struct drm_crtc *crtc) +{ + struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc); + struct mdp4_kms *mdp4_kms = get_kms(crtc); + struct drm_plane *plane; + int i, ovlp = mdp4_crtc->ovlp; + bool alpha[4]= { false, false, false, false }; mdp4_write(mdp4_kms, REG_MDP4_OVLP_TRANSP_LOW0(ovlp), 0); mdp4_write(mdp4_kms, REG_MDP4_OVLP_TRANSP_LOW1(ovlp), 0); @@ -209,13 +229,8 @@ static void blend_setup(struct drm_crtc *crtc) to_mdp_format(msm_framebuffer_format(plane->fb)); alpha[idx-1] = format->alpha_enable; } - mixer_cfg = mixercfg(mixer_cfg, mdp4_crtc->mixer, - pipe_id, stages[idx]); } - /* this shouldn't happen.. and seems to cause underflow: */ - WARN_ON(!mixer_cfg); - for (i = 0; i < 4; i++) { uint32_t op; @@ -238,8 +253,7 @@ static void blend_setup(struct drm_crtc *crtc) mdp4_write(mdp4_kms, REG_MDP4_OVLP_STAGE_TRANSP_HIGH1(ovlp, i), 0); } - mdp4_kms->mixer_cfg = mixer_cfg; - mdp4_write(mdp4_kms, REG_MDP4_LAYERMIXER_IN_CFG, mixer_cfg); + setup_mixer(mdp4_kms); } static void mdp4_crtc_mode_set_nofb(struct drm_crtc *crtc) diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h index 7706452..cbd77bc 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h @@ -32,13 +32,6 @@ struct mdp4_kms { int rev; - /* Shadow value for MDP4_LAYERMIXER_IN_CFG.. since setup for all - * crtcs/encoders is in one shared register, we need to update it - * via read/modify/write. But to avoid getting confused by power- - * on-default values after resume, use this shadow value instead: - */ - uint32_t mixer_cfg; - /* mapper-id used to request GEM buffer mapped for scanout: */ int id; -- cgit v0.10.2 From a49493b548db635084abbdbdda2c95da9c25356c Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Fri, 21 Nov 2014 22:36:09 +0200 Subject: amdkfd: Remove DRM_AMDGPU dependency from Kconfig This patch removes the dependency of amdkfd upon DRM_AMDGPU symbol in amdkfd's Kconfig file. This is done because amdgpu driver is not yet upstreamed and therefore, DRM_AMDGPU symbol is not present in any Kconfig file. Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig index e13c67c..8dfac37 100644 --- a/drivers/gpu/drm/amd/amdkfd/Kconfig +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig @@ -4,6 +4,6 @@ config HSA_AMD tristate "HSA kernel driver for AMD GPU devices" - depends on (DRM_RADEON || DRM_AMDGPU) && AMD_IOMMU_V2 && X86_64 + depends on DRM_RADEON && AMD_IOMMU_V2 && X86_64 help Enable this if you want to use HSA features on AMD GPU devices. -- cgit v0.10.2 From 934ce1c23624526d9d784e0499190bb48113e6f4 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Wed, 19 Nov 2014 16:41:33 -0500 Subject: drm/atomic: check mode_changed *after* atomic_check The intention is that drivers can set crtc_state->mode_changed in their atomic_check() fxns if they encounter a scenario that requires full modeset. Signed-off-by: Rob Clark Reviewed-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 6903600..3a74e81 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -331,7 +331,7 @@ mode_fixup(struct drm_atomic_state *state) } static int -drm_atomic_helper_check_prepare(struct drm_device *dev, +drm_atomic_helper_check_modeset(struct drm_device *dev, struct drm_atomic_state *state) { int ncrtcs = dev->mode_config.num_crtc; @@ -428,10 +428,6 @@ int drm_atomic_helper_check(struct drm_device *dev, int ncrtcs = dev->mode_config.num_crtc; int i, ret = 0; - ret = drm_atomic_helper_check_prepare(dev, state); - if (ret) - return ret; - for (i = 0; i < nplanes; i++) { struct drm_plane_helper_funcs *funcs; struct drm_plane *plane = state->planes[i]; @@ -475,6 +471,10 @@ int drm_atomic_helper_check(struct drm_device *dev, } } + ret = drm_atomic_helper_check_modeset(dev, state); + if (ret) + return ret; + return ret; } EXPORT_SYMBOL(drm_atomic_helper_check); -- cgit v0.10.2 From 46df9adb2e7709e56ab8aacaff2fc997a6d17239 Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Thu, 20 Nov 2014 15:40:36 -0500 Subject: drm/atomic: shutdown *current* encoder In disable_outputs() we need to shut down the outgoing encoder, not the incoming one (we have already swapped-state at this point). Without this, we end up telling the driver to crtc->dpms(OFF) without first encoder->dpms(OFF), and that makes some hw quite unhappy. v2: missing WARN_ON() hunk and comment Reviewed-by: Daniel Vetter Signed-off-by: Rob Clark diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 3a74e81..a17b8e9 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -499,9 +499,12 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state) if (!old_conn_state || !old_conn_state->crtc) continue; - encoder = connector->state->best_encoder; + encoder = old_conn_state->best_encoder; - if (!encoder) + /* We shouldn't get this far if we didn't previously have + * an encoder.. but WARN_ON() rather than explode. + */ + if (WARN_ON(!encoder)) continue; funcs = encoder->helper_private; -- cgit v0.10.2 From 5cbb37df378dbedfa2b5f542b6d50937736f2427 Mon Sep 17 00:00:00 2001 From: Inki Dae Date: Thu, 6 Nov 2014 19:23:35 +0900 Subject: drm/exynos: resolve infinite loop issue on multi-platform This patch resolves temporarily infinite loop issue incurred when Exynos drm driver is enabled and multi-platform kernel is used by registering Exynos drm device object only in case of Exynos SoC. So this patch will be replaced with more generic way later. Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index c57466e..d41aae0 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -741,6 +741,18 @@ static int exynos_drm_init(void) { int ret; + /* + * Register device object only in case of Exynos SoC. + * + * Below codes resolves temporarily infinite loop issue incurred + * by Exynos drm driver when using multi-platform kernel. + * So these codes will be replaced with more generic way later. + */ + if (!of_machine_is_compatible("samsung,exynos3") && + !of_machine_is_compatible("samsung,exynos4") && + !of_machine_is_compatible("samsung,exynos5")) + return -ENODEV; + exynos_drm_pdev = platform_device_register_simple("exynos-drm", -1, NULL, 0); if (IS_ERR(exynos_drm_pdev)) -- cgit v0.10.2 From fbdf093d62de4974cd30f170cf76aa14e26b13e4 Mon Sep 17 00:00:00 2001 From: Inki Dae Date: Thu, 6 Nov 2014 23:00:37 +0900 Subject: drm/exynos: resolve infinite loop issue on non multi-platform This patch resovles the infinite loop issue incurred when Exyno drm driver is enabled but all kms drivers are disabled on Exynos board by returning -EPROBE_DEFER only in case that there is kms device registered. Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index d41aae0..b8abbc4 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -495,6 +495,12 @@ static struct component_match *exynos_drm_match_add(struct device *dev) mutex_lock(&drm_component_lock); + /* Do not retry to probe if there is no any kms driver regitered. */ + if (list_empty(&drm_component_list)) { + mutex_unlock(&drm_component_lock); + return ERR_PTR(-ENODEV); + } + list_for_each_entry(cdev, &drm_component_list, list) { /* * Add components to master only in case that crtc and -- cgit v0.10.2 From 2d15118750fd69a50cf286d3fc275546fa0b5857 Mon Sep 17 00:00:00 2001 From: Inki Dae Date: Fri, 7 Nov 2014 20:31:08 +0900 Subject: drm/exynos: g2d: fix null pointer dereference This patch fixes a null pointer dereference issue incurred by calling g2d_remove when exynos_drm_platform_probe is failed. cmdlist_pool of g2d is allocated when g2d sub driver is probed. So if exynos_drm_platform_probe is failed, the g2d sub driver is not probed and the cmdlist_pool is still NULL. Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_g2d.c b/drivers/gpu/drm/exynos/exynos_drm_g2d.c index df7a77d..6ff8599 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_g2d.c +++ b/drivers/gpu/drm/exynos/exynos_drm_g2d.c @@ -302,9 +302,12 @@ static void g2d_fini_cmdlist(struct g2d_data *g2d) struct exynos_drm_subdrv *subdrv = &g2d->subdrv; kfree(g2d->cmdlist_node); - dma_free_attrs(subdrv->drm_dev->dev, G2D_CMDLIST_POOL_SIZE, - g2d->cmdlist_pool_virt, - g2d->cmdlist_pool, &g2d->cmdlist_dma_attrs); + + if (g2d->cmdlist_pool_virt && g2d->cmdlist_pool) { + dma_free_attrs(subdrv->drm_dev->dev, G2D_CMDLIST_POOL_SIZE, + g2d->cmdlist_pool_virt, + g2d->cmdlist_pool, &g2d->cmdlist_dma_attrs); + } } static struct g2d_cmdlist_node *g2d_get_cmdlist(struct g2d_data *g2d) -- cgit v0.10.2 From e9fbdcb45a36c775383dd98d53b4452cdb5e53c0 Mon Sep 17 00:00:00 2001 From: Inki Dae Date: Fri, 7 Nov 2014 21:32:34 +0900 Subject: drm/exynos: fix possible infinite loop issue This patch fixes possible infinite loop issue by postponing registration to non kms drivers after component_master_add_with_match call, which can be incurred in all cases that non kms driver is probed and then component bind is failed This patch should be applied on top of below patches, http://comments.gmane.org/gmane.comp.video.dri.devel/117740 http://www.spinics.net/lists/linux-samsung-soc/msg38624.html Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index b8abbc4..e5c4c6c 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -591,10 +591,21 @@ static int exynos_drm_platform_probe(struct platform_device *pdev) goto err_unregister_mixer_drv; #endif + match = exynos_drm_match_add(&pdev->dev); + if (IS_ERR(match)) { + ret = PTR_ERR(match); + goto err_unregister_hdmi_drv; + } + + ret = component_master_add_with_match(&pdev->dev, &exynos_drm_ops, + match); + if (ret < 0) + goto err_unregister_hdmi_drv; + #ifdef CONFIG_DRM_EXYNOS_G2D ret = platform_driver_register(&g2d_driver); if (ret < 0) - goto err_unregister_hdmi_drv; + goto err_del_component_master; #endif #ifdef CONFIG_DRM_EXYNOS_FIMC @@ -625,23 +636,9 @@ static int exynos_drm_platform_probe(struct platform_device *pdev) goto err_unregister_ipp_drv; #endif - match = exynos_drm_match_add(&pdev->dev); - if (IS_ERR(match)) { - ret = PTR_ERR(match); - goto err_unregister_resources; - } - - ret = component_master_add_with_match(&pdev->dev, &exynos_drm_ops, - match); - if (ret < 0) - goto err_unregister_resources; - return ret; -err_unregister_resources: - #ifdef CONFIG_DRM_EXYNOS_IPP - exynos_platform_device_ipp_unregister(); err_unregister_ipp_drv: platform_driver_unregister(&ipp_driver); err_unregister_gsc_drv: @@ -664,9 +661,11 @@ err_unregister_g2d_drv: #ifdef CONFIG_DRM_EXYNOS_G2D platform_driver_unregister(&g2d_driver); -err_unregister_hdmi_drv: +err_del_component_master: #endif + component_master_del(&pdev->dev, &exynos_drm_ops); +err_unregister_hdmi_drv: #ifdef CONFIG_DRM_EXYNOS_HDMI platform_driver_unregister(&hdmi_driver); err_unregister_mixer_drv: -- cgit v0.10.2 From 421ee18d4e040d673f5ce2972efa75c2ce0826d9 Mon Sep 17 00:00:00 2001 From: Inki Dae Date: Thu, 13 Nov 2014 16:05:12 +0900 Subject: drm/exynos: fix null pointer dereference issue This patch fixes null pointer dereference issue incurred when ipp driver is enabled and Exynos drm driver is closed. Non kms driver should register its own sub driver to setup necessary resources, which is done by load(). So null pointer dereference occurs when ipp driver is enabled and Exynos drm driver is closed because ipp core device is registered after component_master_add_with_match call. This patch makes exynos_drm_device_subdrv_probe() to be called after all non kms drivers are registered. Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index e5c4c6c..f0bcdc6 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -108,11 +108,6 @@ static int exynos_drm_load(struct drm_device *dev, unsigned long flags) if (ret) goto err_unbind_all; - /* Probe non kms sub drivers and virtual display driver. */ - ret = exynos_drm_device_subdrv_probe(dev); - if (ret) - goto err_cleanup_vblank; - /* * enable drm irq mode. * - with irq_enabled = true, we can use the vblank feature. @@ -138,8 +133,6 @@ static int exynos_drm_load(struct drm_device *dev, unsigned long flags) return 0; -err_cleanup_vblank: - drm_vblank_cleanup(dev); err_unbind_all: component_unbind_all(dev->dev, dev); err_mode_config_cleanup: @@ -153,8 +146,6 @@ err_free_private: static int exynos_drm_unload(struct drm_device *dev) { - exynos_drm_device_subdrv_remove(dev); - exynos_drm_fbdev_fini(dev); drm_kms_helper_poll_fini(dev); @@ -636,9 +627,16 @@ static int exynos_drm_platform_probe(struct platform_device *pdev) goto err_unregister_ipp_drv; #endif + /* Probe non kms sub drivers and virtual display driver. */ + ret = exynos_drm_device_subdrv_probe(platform_get_drvdata(pdev)); + if (ret) + goto err_unregister_resources; + return ret; +err_unregister_resources: #ifdef CONFIG_DRM_EXYNOS_IPP + exynos_platform_device_ipp_unregister(); err_unregister_ipp_drv: platform_driver_unregister(&ipp_driver); err_unregister_gsc_drv: @@ -691,6 +689,8 @@ err_unregister_fimd_drv: static int exynos_drm_platform_remove(struct platform_device *pdev) { + exynos_drm_device_subdrv_remove(platform_get_drvdata(pdev)); + #ifdef CONFIG_DRM_EXYNOS_IPP exynos_platform_device_ipp_unregister(); platform_driver_unregister(&ipp_driver); -- cgit v0.10.2 From 7239067795dc03be5b152034558bed0bb9c73ab3 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Thu, 13 Nov 2014 16:37:57 +0900 Subject: drm/exynos: remove ifdeferry from initialization code The patch replaces separate calls to driver (de)registration by loops over the array of drivers. As a result it significantly decreases number of ifdefs. Additionally it moves device registration related ifdefs to header file. Changelog v2: - Rebased. - Consider non kms driver in respect to infinite loop issue. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index f0bcdc6..8aee629 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -547,85 +547,74 @@ static const struct component_master_ops exynos_drm_ops = { .unbind = exynos_drm_unbind, }; -static int exynos_drm_platform_probe(struct platform_device *pdev) -{ - struct component_match *match; - int ret; - - pdev->dev.coherent_dma_mask = DMA_BIT_MASK(32); - exynos_drm_driver.num_ioctls = ARRAY_SIZE(exynos_ioctls); - +static struct platform_driver *const exynos_drm_kms_drivers[] = { #ifdef CONFIG_DRM_EXYNOS_FIMD - ret = platform_driver_register(&fimd_driver); - if (ret < 0) - return ret; + &fimd_driver, #endif - #ifdef CONFIG_DRM_EXYNOS_DP - ret = platform_driver_register(&dp_driver); - if (ret < 0) - goto err_unregister_fimd_drv; + &dp_driver, #endif - #ifdef CONFIG_DRM_EXYNOS_DSI - ret = platform_driver_register(&dsi_driver); - if (ret < 0) - goto err_unregister_dp_drv; + &dsi_driver, #endif - #ifdef CONFIG_DRM_EXYNOS_HDMI - ret = platform_driver_register(&mixer_driver); - if (ret < 0) - goto err_unregister_dsi_drv; - ret = platform_driver_register(&hdmi_driver); - if (ret < 0) - goto err_unregister_mixer_drv; + &mixer_driver, + &hdmi_driver, #endif +}; - match = exynos_drm_match_add(&pdev->dev); - if (IS_ERR(match)) { - ret = PTR_ERR(match); - goto err_unregister_hdmi_drv; - } - - ret = component_master_add_with_match(&pdev->dev, &exynos_drm_ops, - match); - if (ret < 0) - goto err_unregister_hdmi_drv; - +static struct platform_driver *const exynos_drm_non_kms_drivers[] = { #ifdef CONFIG_DRM_EXYNOS_G2D - ret = platform_driver_register(&g2d_driver); - if (ret < 0) - goto err_del_component_master; + &g2d_driver, #endif - #ifdef CONFIG_DRM_EXYNOS_FIMC - ret = platform_driver_register(&fimc_driver); - if (ret < 0) - goto err_unregister_g2d_drv; + &fimc_driver, #endif - #ifdef CONFIG_DRM_EXYNOS_ROTATOR - ret = platform_driver_register(&rotator_driver); - if (ret < 0) - goto err_unregister_fimc_drv; + &rotator_driver, #endif - #ifdef CONFIG_DRM_EXYNOS_GSC - ret = platform_driver_register(&gsc_driver); - if (ret < 0) - goto err_unregister_rotator_drv; + &gsc_driver, #endif - #ifdef CONFIG_DRM_EXYNOS_IPP - ret = platform_driver_register(&ipp_driver); + &ipp_driver, +#endif +}; + +static int exynos_drm_platform_probe(struct platform_device *pdev) +{ + struct component_match *match; + int ret, i, j; + + pdev->dev.coherent_dma_mask = DMA_BIT_MASK(32); + exynos_drm_driver.num_ioctls = ARRAY_SIZE(exynos_ioctls); + + for (i = 0; i < ARRAY_SIZE(exynos_drm_kms_drivers); ++i) { + ret = platform_driver_register(exynos_drm_kms_drivers[i]); + if (ret < 0) + goto err_unregister_kms_drivers; + } + + match = exynos_drm_match_add(&pdev->dev); + if (IS_ERR(match)) { + ret = PTR_ERR(match); + goto err_unregister_kms_drivers; + } + + ret = component_master_add_with_match(&pdev->dev, &exynos_drm_ops, + match); if (ret < 0) - goto err_unregister_gsc_drv; + goto err_unregister_kms_drivers; + + for (j = 0; j < ARRAY_SIZE(exynos_drm_non_kms_drivers); ++j) { + ret = platform_driver_register(exynos_drm_non_kms_drivers[j]); + if (ret < 0) + goto err_del_component_master; + } ret = exynos_platform_device_ipp_register(); if (ret < 0) - goto err_unregister_ipp_drv; -#endif + goto err_unregister_non_kms_drivers; /* Probe non kms sub drivers and virtual display driver. */ ret = exynos_drm_device_subdrv_probe(platform_get_drvdata(pdev)); @@ -637,98 +626,39 @@ static int exynos_drm_platform_probe(struct platform_device *pdev) err_unregister_resources: #ifdef CONFIG_DRM_EXYNOS_IPP exynos_platform_device_ipp_unregister(); -err_unregister_ipp_drv: - platform_driver_unregister(&ipp_driver); -err_unregister_gsc_drv: #endif +err_unregister_non_kms_drivers: + while (--j >= 0) + platform_driver_unregister(exynos_drm_non_kms_drivers[j]); -#ifdef CONFIG_DRM_EXYNOS_GSC - platform_driver_unregister(&gsc_driver); -err_unregister_rotator_drv: -#endif - -#ifdef CONFIG_DRM_EXYNOS_ROTATOR - platform_driver_unregister(&rotator_driver); -err_unregister_fimc_drv: -#endif - -#ifdef CONFIG_DRM_EXYNOS_FIMC - platform_driver_unregister(&fimc_driver); -err_unregister_g2d_drv: -#endif - -#ifdef CONFIG_DRM_EXYNOS_G2D - platform_driver_unregister(&g2d_driver); err_del_component_master: -#endif component_master_del(&pdev->dev, &exynos_drm_ops); -err_unregister_hdmi_drv: -#ifdef CONFIG_DRM_EXYNOS_HDMI - platform_driver_unregister(&hdmi_driver); -err_unregister_mixer_drv: - platform_driver_unregister(&mixer_driver); -err_unregister_dsi_drv: -#endif +err_unregister_kms_drivers: + while (--i >= 0) + platform_driver_unregister(exynos_drm_kms_drivers[i]); -#ifdef CONFIG_DRM_EXYNOS_DSI - platform_driver_unregister(&dsi_driver); -err_unregister_dp_drv: -#endif - -#ifdef CONFIG_DRM_EXYNOS_DP - platform_driver_unregister(&dp_driver); -err_unregister_fimd_drv: -#endif - -#ifdef CONFIG_DRM_EXYNOS_FIMD - platform_driver_unregister(&fimd_driver); -#endif return ret; } static int exynos_drm_platform_remove(struct platform_device *pdev) { + int i; + exynos_drm_device_subdrv_remove(platform_get_drvdata(pdev)); #ifdef CONFIG_DRM_EXYNOS_IPP exynos_platform_device_ipp_unregister(); - platform_driver_unregister(&ipp_driver); -#endif - -#ifdef CONFIG_DRM_EXYNOS_GSC - platform_driver_unregister(&gsc_driver); #endif -#ifdef CONFIG_DRM_EXYNOS_ROTATOR - platform_driver_unregister(&rotator_driver); -#endif - -#ifdef CONFIG_DRM_EXYNOS_FIMC - platform_driver_unregister(&fimc_driver); -#endif - -#ifdef CONFIG_DRM_EXYNOS_G2D - platform_driver_unregister(&g2d_driver); -#endif - -#ifdef CONFIG_DRM_EXYNOS_HDMI - platform_driver_unregister(&mixer_driver); - platform_driver_unregister(&hdmi_driver); -#endif + for (i = ARRAY_SIZE(exynos_drm_non_kms_drivers) - 1; i >= 0; --i) + platform_driver_unregister(exynos_drm_non_kms_drivers[i]); -#ifdef CONFIG_DRM_EXYNOS_FIMD - platform_driver_unregister(&fimd_driver); -#endif + component_master_del(&pdev->dev, &exynos_drm_ops); -#ifdef CONFIG_DRM_EXYNOS_DSI - platform_driver_unregister(&dsi_driver); -#endif + for (i = ARRAY_SIZE(exynos_drm_kms_drivers) - 1; i >= 0; --i) + platform_driver_unregister(exynos_drm_kms_drivers[i]); -#ifdef CONFIG_DRM_EXYNOS_DP - platform_driver_unregister(&dp_driver); -#endif - component_master_del(&pdev->dev, &exynos_drm_ops); return 0; } @@ -763,11 +693,9 @@ static int exynos_drm_init(void) if (IS_ERR(exynos_drm_pdev)) return PTR_ERR(exynos_drm_pdev); -#ifdef CONFIG_DRM_EXYNOS_VIDI ret = exynos_drm_probe_vidi(); if (ret < 0) goto err_unregister_pd; -#endif ret = platform_driver_register(&exynos_drm_platform_driver); if (ret) @@ -776,11 +704,9 @@ static int exynos_drm_init(void) return 0; err_remove_vidi: -#ifdef CONFIG_DRM_EXYNOS_VIDI exynos_drm_remove_vidi(); err_unregister_pd: -#endif platform_device_unregister(exynos_drm_pdev); return ret; @@ -789,9 +715,9 @@ err_unregister_pd: static void exynos_drm_exit(void) { platform_driver_unregister(&exynos_drm_platform_driver); -#ifdef CONFIG_DRM_EXYNOS_VIDI + exynos_drm_remove_vidi(); -#endif + platform_device_unregister(exynos_drm_pdev); } diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index d22e640..3c81c4b 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -323,15 +323,14 @@ int exynos_platform_device_hdmi_register(void); */ void exynos_platform_device_hdmi_unregister(void); -/* - * this function registers exynos drm ipp platform device. - */ +#ifdef CONFIG_DRM_EXYNOS_IPP int exynos_platform_device_ipp_register(void); - -/* - * this function unregisters exynos drm ipp platform device if it exists. - */ void exynos_platform_device_ipp_unregister(void); +#else +static inline int exynos_platform_device_ipp_register(void) { return 0; } +static inline void exynos_platform_device_ipp_unregister(void) {} +#endif + #ifdef CONFIG_DRM_EXYNOS_DPI struct exynos_drm_display * exynos_dpi_probe(struct device *dev); @@ -342,15 +341,13 @@ exynos_dpi_probe(struct device *dev) { return NULL; } static inline int exynos_dpi_remove(struct device *dev) { return 0; } #endif -/* - * this function registers exynos drm vidi platform device/driver. - */ +#ifdef CONFIG_DRM_EXYNOS_VIDI int exynos_drm_probe_vidi(void); - -/* - * this function unregister exynos drm vidi platform device/driver. - */ void exynos_drm_remove_vidi(void); +#else +static inline int exynos_drm_probe_vidi(void) { return 0; } +static inline void exynos_drm_remove_vidi(void) {} +#endif /* This function creates a encoder and a connector, and initializes them. */ int exynos_drm_create_enc_conn(struct drm_device *dev, -- cgit v0.10.2 From 2900c69c52079a0340db0f012d2b2e92db7544c4 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Tue, 7 Oct 2014 14:01:08 +0200 Subject: drm/exynos: dsi: remove global variable exynos_dsi_display exynos_dsi_display is used by internal Exynos DRM framework for representing pair encoder->connecter. As it should be mapped 1:1 to dsi private context it seems more reasonable to embed it directly in that context. As a result further code simplification will be possible. Moreover it will be possible to handle multiple DSI devices in the system. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c index f43d258..c0141b60 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c @@ -268,6 +268,7 @@ struct exynos_dsi_driver_data { }; struct exynos_dsi { + struct exynos_drm_display display; struct mipi_dsi_host dsi_host; struct drm_connector connector; struct drm_encoder *encoder; @@ -1531,10 +1532,6 @@ static struct exynos_drm_display_ops exynos_dsi_display_ops = { .dpms = exynos_dsi_dpms }; -static struct exynos_drm_display exynos_dsi_display = { - .type = EXYNOS_DISPLAY_TYPE_LCD, - .ops = &exynos_dsi_display_ops, -}; MODULE_DEVICE_TABLE(of, exynos_dsi_of_match); /* of_* functions will be removed after merge of of_graph patches */ @@ -1640,28 +1637,28 @@ end: static int exynos_dsi_bind(struct device *dev, struct device *master, void *data) { + struct exynos_drm_display *display = dev_get_drvdata(dev); + struct exynos_dsi *dsi = display->ctx; struct drm_device *drm_dev = data; - struct exynos_dsi *dsi; int ret; - ret = exynos_drm_create_enc_conn(drm_dev, &exynos_dsi_display); + ret = exynos_drm_create_enc_conn(drm_dev, display); if (ret) { DRM_ERROR("Encoder create [%d] failed with %d\n", - exynos_dsi_display.type, ret); + display->type, ret); return ret; } - dsi = exynos_dsi_display.ctx; - return mipi_dsi_host_register(&dsi->dsi_host); } static void exynos_dsi_unbind(struct device *dev, struct device *master, void *data) { - struct exynos_dsi *dsi = exynos_dsi_display.ctx; + struct exynos_drm_display *display = dev_get_drvdata(dev); + struct exynos_dsi *dsi = display->ctx; - exynos_dsi_dpms(&exynos_dsi_display, DRM_MODE_DPMS_OFF); + exynos_dsi_dpms(display, DRM_MODE_DPMS_OFF); mipi_dsi_host_unregister(&dsi->dsi_host); } @@ -1673,22 +1670,23 @@ static const struct component_ops exynos_dsi_component_ops = { static int exynos_dsi_probe(struct platform_device *pdev) { + struct device *dev = &pdev->dev; struct resource *res; struct exynos_dsi *dsi; int ret; - ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR, - exynos_dsi_display.type); + dsi = devm_kzalloc(dev, sizeof(*dsi), GFP_KERNEL); + if (!dsi) + return -ENOMEM; + + dsi->display.type = EXYNOS_DISPLAY_TYPE_LCD; + dsi->display.ops = &exynos_dsi_display_ops; + + ret = exynos_drm_component_add(dev, EXYNOS_DEVICE_TYPE_CONNECTOR, + dsi->display.type); if (ret) return ret; - dsi = devm_kzalloc(&pdev->dev, sizeof(*dsi), GFP_KERNEL); - if (!dsi) { - dev_err(&pdev->dev, "failed to allocate dsi object.\n"); - ret = -ENOMEM; - goto err_del_component; - } - /* To be checked as invalid one */ dsi->te_gpio = -ENOENT; -- cgit v0.10.2 From e2d2a1e0a264725fd0a62b91422d33ba2263a341 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Tue, 7 Oct 2014 14:01:09 +0200 Subject: drm/exynos: dsi: simplify device pointer evaluation The patch replaces multiple evaluation of device address with local variable. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c index c0141b60..926682c 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c @@ -1695,9 +1695,9 @@ static int exynos_dsi_probe(struct platform_device *pdev) INIT_LIST_HEAD(&dsi->transfer_list); dsi->dsi_host.ops = &exynos_dsi_ops; - dsi->dsi_host.dev = &pdev->dev; + dsi->dsi_host.dev = dev; - dsi->dev = &pdev->dev; + dsi->dev = dev; dsi->driver_data = exynos_dsi_get_driver_data(pdev); ret = exynos_dsi_parse_dt(dsi); @@ -1706,70 +1706,70 @@ static int exynos_dsi_probe(struct platform_device *pdev) dsi->supplies[0].supply = "vddcore"; dsi->supplies[1].supply = "vddio"; - ret = devm_regulator_bulk_get(&pdev->dev, ARRAY_SIZE(dsi->supplies), + ret = devm_regulator_bulk_get(dev, ARRAY_SIZE(dsi->supplies), dsi->supplies); if (ret) { - dev_info(&pdev->dev, "failed to get regulators: %d\n", ret); + dev_info(dev, "failed to get regulators: %d\n", ret); return -EPROBE_DEFER; } - dsi->pll_clk = devm_clk_get(&pdev->dev, "pll_clk"); + dsi->pll_clk = devm_clk_get(dev, "pll_clk"); if (IS_ERR(dsi->pll_clk)) { - dev_info(&pdev->dev, "failed to get dsi pll input clock\n"); + dev_info(dev, "failed to get dsi pll input clock\n"); ret = PTR_ERR(dsi->pll_clk); goto err_del_component; } - dsi->bus_clk = devm_clk_get(&pdev->dev, "bus_clk"); + dsi->bus_clk = devm_clk_get(dev, "bus_clk"); if (IS_ERR(dsi->bus_clk)) { - dev_info(&pdev->dev, "failed to get dsi bus clock\n"); + dev_info(dev, "failed to get dsi bus clock\n"); ret = PTR_ERR(dsi->bus_clk); goto err_del_component; } res = platform_get_resource(pdev, IORESOURCE_MEM, 0); - dsi->reg_base = devm_ioremap_resource(&pdev->dev, res); + dsi->reg_base = devm_ioremap_resource(dev, res); if (IS_ERR(dsi->reg_base)) { - dev_err(&pdev->dev, "failed to remap io region\n"); + dev_err(dev, "failed to remap io region\n"); ret = PTR_ERR(dsi->reg_base); goto err_del_component; } - dsi->phy = devm_phy_get(&pdev->dev, "dsim"); + dsi->phy = devm_phy_get(dev, "dsim"); if (IS_ERR(dsi->phy)) { - dev_info(&pdev->dev, "failed to get dsim phy\n"); + dev_info(dev, "failed to get dsim phy\n"); ret = PTR_ERR(dsi->phy); goto err_del_component; } dsi->irq = platform_get_irq(pdev, 0); if (dsi->irq < 0) { - dev_err(&pdev->dev, "failed to request dsi irq resource\n"); + dev_err(dev, "failed to request dsi irq resource\n"); ret = dsi->irq; goto err_del_component; } irq_set_status_flags(dsi->irq, IRQ_NOAUTOEN); - ret = devm_request_threaded_irq(&pdev->dev, dsi->irq, NULL, + ret = devm_request_threaded_irq(dev, dsi->irq, NULL, exynos_dsi_irq, IRQF_ONESHOT, - dev_name(&pdev->dev), dsi); + dev_name(dev), dsi); if (ret) { - dev_err(&pdev->dev, "failed to request dsi irq\n"); + dev_err(dev, "failed to request dsi irq\n"); goto err_del_component; } - exynos_dsi_display.ctx = dsi; + dsi->display.ctx = dsi; - platform_set_drvdata(pdev, &exynos_dsi_display); + platform_set_drvdata(pdev, &dsi->display); - ret = component_add(&pdev->dev, &exynos_dsi_component_ops); + ret = component_add(dev, &exynos_dsi_component_ops); if (ret) goto err_del_component; return ret; err_del_component: - exynos_drm_component_del(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR); + exynos_drm_component_del(dev, EXYNOS_DEVICE_TYPE_CONNECTOR); return ret; } -- cgit v0.10.2 From e5169723da013eec25e1d52faa8c5058abe1b23e Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Tue, 7 Oct 2014 14:01:10 +0200 Subject: drm/exynos: dsi: remove redundant encoder field The patch removes redundant encoder field from private DSI context. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c index 926682c..c9d1a00 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c @@ -271,7 +271,6 @@ struct exynos_dsi { struct exynos_drm_display display; struct mipi_dsi_host dsi_host; struct drm_connector connector; - struct drm_encoder *encoder; struct device_node *panel_node; struct drm_panel *panel; struct device *dev; @@ -1105,7 +1104,7 @@ static irqreturn_t exynos_dsi_irq(int irq, void *dev_id) static irqreturn_t exynos_dsi_te_irq_handler(int irq, void *dev_id) { struct exynos_dsi *dsi = (struct exynos_dsi *)dev_id; - struct drm_encoder *encoder = dsi->encoder; + struct drm_encoder *encoder = dsi->display.encoder; if (dsi->state & DSIM_STATE_ENABLED) exynos_drm_crtc_te_handler(encoder->crtc); @@ -1475,7 +1474,7 @@ exynos_dsi_best_encoder(struct drm_connector *connector) { struct exynos_dsi *dsi = connector_to_dsi(connector); - return dsi->encoder; + return dsi->display.encoder; } static struct drm_connector_helper_funcs exynos_dsi_connector_helper_funcs = { @@ -1491,8 +1490,6 @@ static int exynos_dsi_create_connector(struct exynos_drm_display *display, struct drm_connector *connector = &dsi->connector; int ret; - dsi->encoder = encoder; - connector->polled = DRM_CONNECTOR_POLL_HPD; ret = drm_connector_init(encoder->dev, connector, -- cgit v0.10.2 From 5cd5db80402c91d4b5f4ea5f5cf4be5b21864935 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Tue, 7 Oct 2014 14:01:11 +0200 Subject: drm/exynos: dsi: stop using display->ctx pointer The patch replaces accesses to display->ctx pointer by container_of construct. It will allow to remove ctx field in the future. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c index c9d1a00..5e38d15 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c @@ -304,6 +304,11 @@ struct exynos_dsi { #define host_to_dsi(host) container_of(host, struct exynos_dsi, dsi_host) #define connector_to_dsi(c) container_of(c, struct exynos_dsi, connector) +static inline struct exynos_dsi *display_to_dsi(struct exynos_drm_display *d) +{ + return container_of(d, struct exynos_dsi, display); +} + static struct exynos_dsi_driver_data exynos3_dsi_driver_data = { .plltmr_reg = 0x50, .has_freqband = 1, @@ -1397,7 +1402,7 @@ static void exynos_dsi_disable(struct exynos_dsi *dsi) static void exynos_dsi_dpms(struct exynos_drm_display *display, int mode) { - struct exynos_dsi *dsi = display->ctx; + struct exynos_dsi *dsi = display_to_dsi(display); if (dsi->panel) { switch (mode) { @@ -1486,7 +1491,7 @@ static struct drm_connector_helper_funcs exynos_dsi_connector_helper_funcs = { static int exynos_dsi_create_connector(struct exynos_drm_display *display, struct drm_encoder *encoder) { - struct exynos_dsi *dsi = display->ctx; + struct exynos_dsi *dsi = display_to_dsi(display); struct drm_connector *connector = &dsi->connector; int ret; @@ -1510,7 +1515,7 @@ static int exynos_dsi_create_connector(struct exynos_drm_display *display, static void exynos_dsi_mode_set(struct exynos_drm_display *display, struct drm_display_mode *mode) { - struct exynos_dsi *dsi = display->ctx; + struct exynos_dsi *dsi = display_to_dsi(display); struct videomode *vm = &dsi->vm; vm->hactive = mode->hdisplay; @@ -1635,7 +1640,7 @@ static int exynos_dsi_bind(struct device *dev, struct device *master, void *data) { struct exynos_drm_display *display = dev_get_drvdata(dev); - struct exynos_dsi *dsi = display->ctx; + struct exynos_dsi *dsi = display_to_dsi(display); struct drm_device *drm_dev = data; int ret; @@ -1653,7 +1658,7 @@ static void exynos_dsi_unbind(struct device *dev, struct device *master, void *data) { struct exynos_drm_display *display = dev_get_drvdata(dev); - struct exynos_dsi *dsi = display->ctx; + struct exynos_dsi *dsi = display_to_dsi(display); exynos_dsi_dpms(display, DRM_MODE_DPMS_OFF); @@ -1755,8 +1760,6 @@ static int exynos_dsi_probe(struct platform_device *pdev) goto err_del_component; } - dsi->display.ctx = dsi; - platform_set_drvdata(pdev, &dsi->display); ret = component_add(dev, &exynos_dsi_component_ops); -- cgit v0.10.2 From ff9c87ce9b9d10d50f42127c0753654af6332157 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 31 Oct 2014 14:17:35 +0000 Subject: drm/exynos: remove uneeded declaration of struct dma_iommu_mapping It is not even used in this header anymore. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_iommu.h b/drivers/gpu/drm/exynos/exynos_drm_iommu.h index 72376d4..35d2588 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_iommu.h +++ b/drivers/gpu/drm/exynos/exynos_drm_iommu.h @@ -40,7 +40,6 @@ static inline bool is_drm_iommu_supported(struct drm_device *drm_dev) #else -struct dma_iommu_mapping; static inline int drm_create_iommu_mapping(struct drm_device *drm_dev) { return 0; -- cgit v0.10.2 From b5b1a4bb1598e3c11d5c74adfcab2055954ebee7 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 31 Oct 2014 14:17:36 +0000 Subject: drm/exynos: remove extra declaration of struct exynos_drm_manager The struct is defined in the same file, declare it here is just unnecessary. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_encoder.h b/drivers/gpu/drm/exynos/exynos_drm_encoder.h index b7a1620..26305d8 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_encoder.h +++ b/drivers/gpu/drm/exynos/exynos_drm_encoder.h @@ -14,8 +14,6 @@ #ifndef _EXYNOS_DRM_ENCODER_H_ #define _EXYNOS_DRM_ENCODER_H_ -struct exynos_drm_manager; - void exynos_drm_encoder_setup(struct drm_device *dev); struct drm_encoder *exynos_drm_encoder_create(struct drm_device *dev, struct exynos_drm_display *mgr, -- cgit v0.10.2 From f859d6e8594e0a9e11b76407b7f4834f612f0c6d Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 31 Oct 2014 14:17:37 +0000 Subject: drm/exynos: remove extra declaration of struct exynos_overlay The struct is defined in the same file, declare it here is just unnecessary Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index 3c81c4b..3905e30 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -37,7 +37,6 @@ #define wait_for(COND, MS) _wait_for(COND, MS) struct drm_device; -struct exynos_drm_overlay; struct drm_connector; /* This enumerates device type. */ -- cgit v0.10.2 From d2c1bba3f6fca60c924996c51d7453c5ddb5cedd Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 31 Oct 2014 14:17:38 +0000 Subject: drm/exynos: Replace repeated declaration by include Re-declare struct is not a good practice, let's use the original drm declarations. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index 3905e30..7806981 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -15,6 +15,7 @@ #ifndef _EXYNOS_DRM_DRV_H_ #define _EXYNOS_DRM_DRV_H_ +#include #include #define MAX_CRTC 3 @@ -36,9 +37,6 @@ #define wait_for(COND, MS) _wait_for(COND, MS) -struct drm_device; -struct drm_connector; - /* This enumerates device type. */ enum exynos_drm_device_type { EXYNOS_DEVICE_TYPE_NONE, -- cgit v0.10.2 From cc2d861c5f3421a1b0ba498cb9f0ecb5b3a3eb1c Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Thu, 13 Nov 2014 17:24:06 +0900 Subject: drm/exynos: Replace repeated declarations by #include "exynos_drm_drv.h" Re-declare struct is not a good practice, let's use the original drm and exynos declarations. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_crtc.h b/drivers/gpu/drm/exynos/exynos_drm_crtc.h index 690dcdd..e353d35 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_crtc.h +++ b/drivers/gpu/drm/exynos/exynos_drm_crtc.h @@ -15,10 +15,7 @@ #ifndef _EXYNOS_DRM_CRTC_H_ #define _EXYNOS_DRM_CRTC_H_ -struct drm_device; -struct drm_crtc; -struct exynos_drm_manager; -struct exynos_drm_overlay; +#include "exynos_drm_drv.h" int exynos_drm_crtc_create(struct exynos_drm_manager *manager); int exynos_drm_crtc_enable_vblank(struct drm_device *dev, int pipe); -- cgit v0.10.2 From 5d0e6fec7f012a9518ff042fd6102dda6c79089e Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 31 Oct 2014 14:17:40 +0000 Subject: drm/exynos: remove unused wait_for macro This is a leftover, all code using this macro have been removed/ changed already. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index 7806981..9e4a7e1 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -23,20 +23,6 @@ #define MAX_FB_BUFFER 4 #define DEFAULT_ZPOS -1 -#define _wait_for(COND, MS) ({ \ - unsigned long timeout__ = jiffies + msecs_to_jiffies(MS); \ - int ret__ = 0; \ - while (!(COND)) { \ - if (time_after(jiffies, timeout__)) { \ - ret__ = -ETIMEDOUT; \ - break; \ - } \ - } \ - ret__; \ -}) - -#define wait_for(COND, MS) _wait_for(COND, MS) - /* This enumerates device type. */ enum exynos_drm_device_type { EXYNOS_DEVICE_TYPE_NONE, -- cgit v0.10.2 From 4a5827fbc26edfce0d8d0839fdf22cd436f0ae5b Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Thu, 13 Nov 2014 17:25:53 +0900 Subject: drm/exynos: Save up space using bool var as bitfields Save a few bytes by compiling them all in the same byte. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index 9e4a7e1..f77e6aa 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -66,10 +66,10 @@ enum exynos_drm_output_type { * @dma_addr: array of bus(accessed by dma) address to the memory region * allocated for a overlay. * @zpos: order of overlay layer(z position). - * @default_win: a window to be enabled. - * @color_key: color key on or off. * @index_color: if using color key feature then this value would be used * as index color. + * @default_win: a window to be enabled. + * @color_key: color key on or off. * @local_path: in case of lcd type, local path mode on or off. * @transparency: transparency on or off. * @activated: activated or not. @@ -97,13 +97,13 @@ struct exynos_drm_overlay { uint32_t pixel_format; dma_addr_t dma_addr[MAX_FB_BUFFER]; int zpos; - - bool default_win; - bool color_key; unsigned int index_color; - bool local_path; - bool transparency; - bool activated; + + bool default_win:1; + bool color_key:1; + bool local_path:1; + bool transparency:1; + bool activated:1; }; /* -- cgit v0.10.2 From 60cb85862e7c12eff855fece28034f7cd3fa6edd Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 31 Oct 2014 14:17:42 +0000 Subject: drm/exynos: update documentation to reflect code changes Description of the @create_connector callback was missing, and the @manager was no longer needed. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index f77e6aa..e762cbb 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -110,6 +110,7 @@ struct exynos_drm_overlay { * Exynos DRM Display Structure. * - this structure is common to analog tv, digital tv and lcd panel. * + * @create_connector: initialize and register a new connector * @remove: cleans up the display for removal * @mode_fixup: fix mode data comparing to hw specific display mode. * @mode_set: convert drm_display_mode to hw specific display mode and @@ -262,8 +263,6 @@ struct exynos_drm_private { * @dev: pointer to device object for subdrv device driver. * @drm_dev: pointer to drm_device and this pointer would be set * when sub driver calls exynos_drm_subdrv_register(). - * @manager: subdrv has its own manager to control a hardware appropriately - * and we can access a hardware drawing on this manager. * @probe: this callback would be called by exynos drm driver after * subdrv is registered to it. * @remove: this callback is used to release resources created -- cgit v0.10.2 From 975588477220cfb32375ace1caafbdbbb9d75e70 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Fri, 31 Oct 2014 14:17:43 +0000 Subject: drm/exynos: remove leftover hdmi function declarations They are not implemented anywhere, so wipe them out. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index e762cbb..262a459 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -294,17 +294,6 @@ int exynos_drm_device_subdrv_remove(struct drm_device *dev); int exynos_drm_subdrv_open(struct drm_device *dev, struct drm_file *file); void exynos_drm_subdrv_close(struct drm_device *dev, struct drm_file *file); -/* - * this function registers exynos drm hdmi platform device. It ensures only one - * instance of the device is created. - */ -int exynos_platform_device_hdmi_register(void); - -/* - * this function unregisters exynos drm hdmi platform device if it exists. - */ -void exynos_platform_device_hdmi_unregister(void); - #ifdef CONFIG_DRM_EXYNOS_IPP int exynos_platform_device_ipp_register(void); void exynos_platform_device_ipp_unregister(void); -- cgit v0.10.2 From b128aefe0b0409c77123586e5243bee415f84f5b Mon Sep 17 00:00:00 2001 From: Vivek Gautam Date: Wed, 12 Nov 2014 15:12:10 +0530 Subject: drm/exynos: dp: Remove support for unused dptx-phy Now that we have moved to generic phy based bindings, we don't need to have any code related to older dptx-phy. Nobody is using this dptx-phy anymore, so removing the same. Signed-off-by: Vivek Gautam Acked-by: Jingoo Han Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_dp_core.c b/drivers/gpu/drm/exynos/exynos_dp_core.c index 6adb1e5..ed818b9 100644 --- a/drivers/gpu/drm/exynos/exynos_dp_core.c +++ b/drivers/gpu/drm/exynos/exynos_dp_core.c @@ -1052,28 +1052,14 @@ static int exynos_dp_create_connector(struct exynos_drm_display *display, static void exynos_dp_phy_init(struct exynos_dp_device *dp) { - if (dp->phy) { + if (dp->phy) phy_power_on(dp->phy); - } else if (dp->phy_addr) { - u32 reg; - - reg = __raw_readl(dp->phy_addr); - reg |= dp->enable_mask; - __raw_writel(reg, dp->phy_addr); - } } static void exynos_dp_phy_exit(struct exynos_dp_device *dp) { - if (dp->phy) { + if (dp->phy) phy_power_off(dp->phy); - } else if (dp->phy_addr) { - u32 reg; - - reg = __raw_readl(dp->phy_addr); - reg &= ~(dp->enable_mask); - __raw_writel(reg, dp->phy_addr); - } } static void exynos_dp_poweron(struct exynos_drm_display *display) @@ -1210,44 +1196,6 @@ static struct video_info *exynos_dp_dt_parse_pdata(struct device *dev) return dp_video_config; } -static int exynos_dp_dt_parse_phydata(struct exynos_dp_device *dp) -{ - struct device_node *dp_phy_node = of_node_get(dp->dev->of_node); - u32 phy_base; - int ret = 0; - - dp_phy_node = of_find_node_by_name(dp_phy_node, "dptx-phy"); - if (!dp_phy_node) { - dp->phy = devm_phy_get(dp->dev, "dp"); - return PTR_ERR_OR_ZERO(dp->phy); - } - - if (of_property_read_u32(dp_phy_node, "reg", &phy_base)) { - dev_err(dp->dev, "failed to get reg for dptx-phy\n"); - ret = -EINVAL; - goto err; - } - - if (of_property_read_u32(dp_phy_node, "samsung,enable-mask", - &dp->enable_mask)) { - dev_err(dp->dev, "failed to get enable-mask for dptx-phy\n"); - ret = -EINVAL; - goto err; - } - - dp->phy_addr = ioremap(phy_base, SZ_4); - if (!dp->phy_addr) { - dev_err(dp->dev, "failed to ioremap dp-phy\n"); - ret = -ENOMEM; - goto err; - } - -err: - of_node_put(dp_phy_node); - - return ret; -} - static int exynos_dp_dt_parse_panel(struct exynos_dp_device *dp) { int ret; @@ -1277,9 +1225,21 @@ static int exynos_dp_bind(struct device *dev, struct device *master, void *data) if (IS_ERR(dp->video_info)) return PTR_ERR(dp->video_info); - ret = exynos_dp_dt_parse_phydata(dp); - if (ret) - return ret; + dp->phy = devm_phy_get(dp->dev, "dp"); + if (IS_ERR(dp->phy)) { + dev_err(dp->dev, "no DP phy configured\n"); + ret = PTR_ERR(dp->phy); + if (ret) { + /* + * phy itself is not enabled, so we can move forward + * assigning NULL to phy pointer. + */ + if (ret == -ENOSYS || ret == -ENODEV) + dp->phy = NULL; + else + return ret; + } + } if (!dp->panel) { ret = exynos_dp_dt_parse_panel(dp); diff --git a/drivers/gpu/drm/exynos/exynos_dp_core.h b/drivers/gpu/drm/exynos/exynos_dp_core.h index a1aee69..6426201 100644 --- a/drivers/gpu/drm/exynos/exynos_dp_core.h +++ b/drivers/gpu/drm/exynos/exynos_dp_core.h @@ -153,8 +153,6 @@ struct exynos_dp_device { struct clk *clock; unsigned int irq; void __iomem *reg_base; - void __iomem *phy_addr; - unsigned int enable_mask; struct video_info *video_info; struct link_train link_train; -- cgit v0.10.2 From 4bc6d6445e0c0b724d4232fcc1f127bde3a4ddbd Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Fri, 7 Nov 2014 15:12:24 +0900 Subject: drm/exynos: dsi: support Exynos4415 SoC This patch supports Exynos4415 SoC. Signed-off-by: YoungJun Cho Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/Documentation/devicetree/bindings/video/exynos_dsim.txt b/Documentation/devicetree/bindings/video/exynos_dsim.txt index e74243b..ca2b4aa 100644 --- a/Documentation/devicetree/bindings/video/exynos_dsim.txt +++ b/Documentation/devicetree/bindings/video/exynos_dsim.txt @@ -4,6 +4,7 @@ Required properties: - compatible: value should be one of the following "samsung,exynos3250-mipi-dsi" /* for Exynos3250/3472 SoCs */ "samsung,exynos4210-mipi-dsi" /* for Exynos4 SoCs */ + "samsung,exynos4415-mipi-dsi" /* for Exynos4415 SoC */ "samsung,exynos5410-mipi-dsi" /* for Exynos5410/5420/5440 SoCs */ - reg: physical base address and length of the registers set for the device - interrupts: should contain DSI interrupt diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c index 5e38d15..e0e8c38 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c @@ -321,6 +321,11 @@ static struct exynos_dsi_driver_data exynos4_dsi_driver_data = { .has_clklane_stop = 1, }; +static struct exynos_dsi_driver_data exynos4415_dsi_driver_data = { + .plltmr_reg = 0x58, + .has_clklane_stop = 1, +}; + static struct exynos_dsi_driver_data exynos5_dsi_driver_data = { .plltmr_reg = 0x58, }; @@ -330,6 +335,8 @@ static struct of_device_id exynos_dsi_of_match[] = { .data = &exynos3_dsi_driver_data }, { .compatible = "samsung,exynos4210-mipi-dsi", .data = &exynos4_dsi_driver_data }, + { .compatible = "samsung,exynos4415-mipi-dsi", + .data = &exynos4415_dsi_driver_data }, { .compatible = "samsung,exynos5410-mipi-dsi", .data = &exynos5_dsi_driver_data }, { } -- cgit v0.10.2 From dcb622aa882b1108c005ebf629014acbf22690e3 Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Fri, 7 Nov 2014 15:12:25 +0900 Subject: drm/exynos: fimd: support Exynos4415 SoC This patch supports Exynos4415 SoC. Signed-off-by: YoungJun Cho Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/Documentation/devicetree/bindings/video/samsung-fimd.txt b/Documentation/devicetree/bindings/video/samsung-fimd.txt index 4e6c77c..cf1af63 100644 --- a/Documentation/devicetree/bindings/video/samsung-fimd.txt +++ b/Documentation/devicetree/bindings/video/samsung-fimd.txt @@ -11,6 +11,7 @@ Required properties: "samsung,s5pv210-fimd"; /* for S5PV210 SoC */ "samsung,exynos3250-fimd"; /* for Exynos3250/3472 SoCs */ "samsung,exynos4210-fimd"; /* for Exynos4 SoCs */ + "samsung,exynos4415-fimd"; /* for Exynos4415 SoC */ "samsung,exynos5250-fimd"; /* for Exynos5 SoCs */ - reg: physical base address and length of the FIMD registers set. diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 085b066..5dfbbdb 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -120,6 +120,15 @@ static struct fimd_driver_data exynos4_fimd_driver_data = { .has_shadowcon = 1, }; +static struct fimd_driver_data exynos4415_fimd_driver_data = { + .timing_base = 0x20000, + .lcdblk_offset = 0x210, + .lcdblk_vt_shift = 10, + .lcdblk_bypass_shift = 1, + .has_shadowcon = 1, + .has_vidoutcon = 1, +}; + static struct fimd_driver_data exynos5_fimd_driver_data = { .timing_base = 0x20000, .lcdblk_offset = 0x214, @@ -180,6 +189,8 @@ static const struct of_device_id fimd_driver_dt_match[] = { .data = &exynos3_fimd_driver_data }, { .compatible = "samsung,exynos4210-fimd", .data = &exynos4_fimd_driver_data }, + { .compatible = "samsung,exynos4415-fimd", + .data = &exynos4415_fimd_driver_data }, { .compatible = "samsung,exynos5250-fimd", .data = &exynos5_fimd_driver_data }, {}, -- cgit v0.10.2 From bd953de94393ef2a00e893aa948cc7c4cb190d00 Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Wed, 1 Oct 2014 15:19:07 +0900 Subject: drm/exynos: fimd: remove unnecessary waiting vblank routine The exynos_drm_crtc_dpms() waits until pended page flip queue is empty, calls the drm_vblank_off() then calls manager->ops->dpms() when mode is DRM_MODE_DPMS_OFF. The fimd_dpms() is one of manager->ops->dpms()s and finally calls fimd_window_suspend(). But there is no active window and vblank is already off when it is called. So addtional waiting vblank is not necessary any more. Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 5dfbbdb..1f46c89 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -814,7 +814,6 @@ static void fimd_window_suspend(struct exynos_drm_manager *mgr) if (win_data->enabled) fimd_win_disable(mgr, i); } - fimd_wait_for_vblank(mgr); } static void fimd_window_resume(struct exynos_drm_manager *mgr) -- cgit v0.10.2 From b301ae24bb22cb6cf762b174426105c85621b1b7 Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Wed, 1 Oct 2014 15:19:10 +0900 Subject: drm/exynos: fimd: move handle vblank position in TE handler For providing VBLANK information, drm_handle_vblank() should be called properly, but it is blocked by wait_vsync_event condition which is set by manager_ops->wait_for_vblank(). So moves it out from wait_vsync_event routine. Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Acked-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 1f46c89..7593f62 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -982,10 +982,10 @@ static void fimd_te_handler(struct exynos_drm_manager *mgr) if (atomic_read(&ctx->wait_vsync_event)) { atomic_set(&ctx->wait_vsync_event, 0); wake_up(&ctx->wait_vsync_queue); - - if (!atomic_read(&ctx->triggering)) - drm_handle_vblank(ctx->drm_dev, ctx->pipe); } + + if (!atomic_read(&ctx->triggering)) + drm_handle_vblank(ctx->drm_dev, ctx->pipe); } static struct exynos_drm_manager_ops fimd_manager_ops = { -- cgit v0.10.2 From d41bb38f228d98a2e411d59be14a2aee876d8db1 Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Wed, 1 Oct 2014 15:19:13 +0900 Subject: drm/exynos: dsi: move DSIM_STATE_ENABLED set position The command mode panel should draw image earlier than the display on command execution to prevent showing garbage GRAM screen data. So should set dsi->state as DSIM_STATE_ENABLED between calling exynos_dsi_set_display_enable() and drm_panel_enable() to transmit image data before executing display on command. And moves the display on command execution routine from prepare() to enable() in drm_panel_funcs also. Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c index e0e8c38..44461aa 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c @@ -1381,16 +1381,17 @@ static int exynos_dsi_enable(struct exynos_dsi *dsi) exynos_dsi_set_display_mode(dsi); exynos_dsi_set_display_enable(dsi, true); + dsi->state |= DSIM_STATE_ENABLED; + ret = drm_panel_enable(dsi->panel); if (ret < 0) { + dsi->state &= ~DSIM_STATE_ENABLED; exynos_dsi_set_display_enable(dsi, false); drm_panel_unprepare(dsi->panel); exynos_dsi_poweroff(dsi); return ret; } - dsi->state |= DSIM_STATE_ENABLED; - return 0; } -- cgit v0.10.2 From 030794a368946b98a8252d3172f5f2a1b0e4fb0e Mon Sep 17 00:00:00 2001 From: Krzysztof Kozlowski Date: Fri, 7 Nov 2014 14:53:58 +0100 Subject: drm/exynos: Fix DSI resuming fail because power domain being off During system resume from suspend to RAM the Exynos DRM driver forced CRTC mode thus turning display on (DPMS_ON). This lead to runtime resuming of DSI which failed because whole LCD power domain was off and it was not allowed to turn on because of system resume in progress. Forcing mode should not be needed and removing it solves this particular problem. This necessary fix for following scenario reproduced on Exynos DRM: 1. Power domain is off before suspending the system. 2. System is suspended to RAM. 3. Resuming starts. The Exynos DRM driver resume callback is called. 4. The Exynos DRM driver calls drm_helper_resume_force_mode() which turns on the screen by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. 5. The Exynos DSI driver calls pm_runtime_get. The driver runtime resumes and this should turn LCD power domain on. 6. Unfortunately the domain cannot be turned on because system resume is in progress and genpd->prepared_count is positive. Steps to reproduce: 1. Add runtime PM to Exynos DSI driver. 2. Build Exynos DRM/FB without FRAMEBUFFER_CONSOLE. 3. Enable the connector and screen (e.g. with modeset-vsync). 4. echo 3 > /sys/devices/platform/exynos-drm/graphics/fb0/blank 5. echo mem > /sys/power/state 6. Resume. [ 77.712469] PM: early resume of devices complete after 3.854 msecs [ 77.712739] exynos-dsi 11c80000.dsi: pm_genpd_resume() [ 77.712758] exynos4-fimc 11800000.fimc: pm_genpd_resume() [ 77.712774] exynos4-fimc 11810000.fimc: pm_genpd_resume() [ 77.712787] exynos-drm-fimc 11820000.fimc: pm_genpd_resume() [ 77.712802] exynos-drm-fimc 11830000.fimc: pm_genpd_resume() [ 77.712815] s5p-mipi-csis 11880000.csis: pm_genpd_resume() [ 77.712829] s5p-mipi-csis 11890000.csis: pm_genpd_resume() [ 77.712843] exynos-fimc-lite 12390000.fimc-lite: pm_genpd_resume() [ 77.712856] exynos-fimc-lite 123a0000.fimc-lite: pm_genpd_resume() [ 77.713788] exynos4-fb 11c00000.fimd: pm_genpd_resume() [ 77.713912] wake disabled for irq 184 [ 77.713923] wake disabled for irq 185 [ 77.714082] wake disabled for irq 173 [ 77.715676] wake disabled for irq 176 [ 77.718540] exynos4-fb 11c00000.fimd: pm_genpd_runtime_resume() [ 77.718567] exynos4-fb 11c00000.fimd: state restore latency exceeded, new value 1708 ns [ 77.718636] exynos-dsi 11c80000.dsi: pm_genpd_runtime_resume() [ 77.892366] exynos-dsi 11c80000.dsi: PLL failed to stabilize [ 77.892377] exynos-dsi 11c80000.dsi: failed to configure DSI PLL [ 78.192168] exynos-dsi 11c80000.dsi: timeout waiting for reset [ 78.211578] exynos-dsi 11c80000.dsi: waiting for bus lanes timed out [ 78.307173] exynos-dsi 11c80000.dsi: xfer timed out: d1 00 (null) [ 78.307190] panel_s6e8aa0 11c80000.dsi.0: error -110 reading dcs seq(0xd1) [ 78.307199] panel_s6e8aa0 11c80000.dsi.0: read id failed Signed-off-by: Krzysztof Kozlowski Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index 8aee629..eab12f0 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -194,8 +194,6 @@ static int exynos_drm_resume(struct drm_device *dev) } drm_modeset_unlock_all(dev); - drm_helper_resume_force_mode(dev); - return 0; } -- cgit v0.10.2 From 3c3c9c1d904dca30f89882b97822fdf025b115c9 Mon Sep 17 00:00:00 2001 From: Joonyoung Shim Date: Fri, 14 Nov 2014 11:36:02 +0900 Subject: drm/exynos: add has_vtsel flag The exynos fimd provides video type selection bits from system register but exynos3 series don't has it, so needs has_vtsel flag and we can distinguish whether set video type selection bits. Signed-off-by: Joonyoung Shim Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 7593f62..0673a39 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -96,6 +96,7 @@ struct fimd_driver_data { unsigned int has_clksel:1; unsigned int has_limited_fmt:1; unsigned int has_vidoutcon:1; + unsigned int has_vtsel:1; }; static struct fimd_driver_data s3c64xx_fimd_driver_data = { @@ -118,6 +119,7 @@ static struct fimd_driver_data exynos4_fimd_driver_data = { .lcdblk_vt_shift = 10, .lcdblk_bypass_shift = 1, .has_shadowcon = 1, + .has_vtsel = 1, }; static struct fimd_driver_data exynos4415_fimd_driver_data = { @@ -127,6 +129,7 @@ static struct fimd_driver_data exynos4415_fimd_driver_data = { .lcdblk_bypass_shift = 1, .has_shadowcon = 1, .has_vidoutcon = 1, + .has_vtsel = 1, }; static struct fimd_driver_data exynos5_fimd_driver_data = { @@ -136,6 +139,7 @@ static struct fimd_driver_data exynos5_fimd_driver_data = { .lcdblk_bypass_shift = 15, .has_shadowcon = 1, .has_vidoutcon = 1, + .has_vtsel = 1, }; struct fimd_win_data { @@ -354,7 +358,8 @@ static void fimd_commit(struct exynos_drm_manager *mgr) writel(0, timing_base + I80IFCONFBx(0)); /* set video type selection to I80 interface */ - if (ctx->sysreg && regmap_update_bits(ctx->sysreg, + if (driver_data->has_vtsel && ctx->sysreg && + regmap_update_bits(ctx->sysreg, driver_data->lcdblk_offset, 0x3 << driver_data->lcdblk_vt_shift, 0x1 << driver_data->lcdblk_vt_shift)) { -- cgit v0.10.2 From 9b67eb7365725ed00dde4714d71d4afe205329b2 Mon Sep 17 00:00:00 2001 From: Joonyoung Shim Date: Mon, 17 Nov 2014 22:00:08 +0900 Subject: drm/exynos: move triggering checking It's better to be checking whether triggerring in fimd_trigger function. Also it will return if in triggerring on fimd_te_handler, then it can't execute remain codes. Signed-off-by: Joonyoung Shim Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 0673a39..ec2d170 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -949,6 +949,13 @@ static void fimd_trigger(struct device *dev) void *timing_base = ctx->regs + driver_data->timing_base; u32 reg; + /* + * Skips to trigger if in triggering state, because multiple triggering + * requests can cause panel reset. + */ + if (atomic_read(&ctx->triggering)) + return; + atomic_set(&ctx->triggering, 1); reg = readl(ctx->regs + VIDINTCON0); @@ -969,13 +976,6 @@ static void fimd_te_handler(struct exynos_drm_manager *mgr) if (ctx->pipe < 0 || !ctx->drm_dev) return; - /* - * Skips to trigger if in triggering state, because multiple triggering - * requests can cause panel reset. - */ - if (atomic_read(&ctx->triggering)) - return; - /* * If there is a page flip request, triggers and handles the page flip * event so that current fb can be updated into panel GRAM. -- cgit v0.10.2 From 74944a58f9def188d506d3eef19fd34f293219f7 Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Mon, 17 Nov 2014 22:00:09 +0900 Subject: drm/exynos: fimd: move shadow unprotection position The C#_EN_F in SHADOWCON register is updated per frame. So it should be protected by fimd_shadow_protect_win(). Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index ec2d170..77ba961 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -751,15 +751,15 @@ static void fimd_win_commit(struct exynos_drm_manager *mgr, int zpos) val |= WINCONx_ENWIN; writel(val, ctx->regs + WINCON(win)); - /* Enable DMA channel and unprotect windows */ - fimd_shadow_protect_win(ctx, win, false); - if (ctx->driver_data->has_shadowcon) { val = readl(ctx->regs + SHADOWCON); val |= SHADOWCON_CHx_ENABLE(win); writel(val, ctx->regs + SHADOWCON); } + /* Enable DMA channel and unprotect windows */ + fimd_shadow_protect_win(ctx, win, false); + win_data->enabled = true; if (ctx->i80_if) -- cgit v0.10.2 From f181a543c76fdcf240ef39d24520141070701bdc Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Mon, 17 Nov 2014 22:00:10 +0900 Subject: drm/exynos: fimd: add fimd_enable_video_output() to cleanup This bit is used for video output and logic signal control. So it is better for readability. Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 77ba961..5cfd251 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -229,6 +229,19 @@ static void fimd_wait_for_vblank(struct exynos_drm_manager *mgr) DRM_DEBUG_KMS("vblank wait timed out.\n"); } +static void fimd_enable_video_output(struct fimd_context *ctx, int win, + bool enable) +{ + u32 val = readl(ctx->regs + WINCON(win)); + + if (enable) + val |= WINCONx_ENWIN; + else + val &= ~WINCONx_ENWIN; + + writel(val, ctx->regs + WINCON(win)); +} + static void fimd_clear_channel(struct exynos_drm_manager *mgr) { struct fimd_context *ctx = mgr->ctx; @@ -241,9 +254,7 @@ static void fimd_clear_channel(struct exynos_drm_manager *mgr) u32 val = readl(ctx->regs + WINCON(win)); if (val & WINCONx_ENWIN) { - /* wincon */ - val &= ~WINCONx_ENWIN; - writel(val, ctx->regs + WINCON(win)); + fimd_enable_video_output(ctx, win, false); /* unprotect windows */ if (ctx->driver_data->has_shadowcon) { @@ -746,10 +757,7 @@ static void fimd_win_commit(struct exynos_drm_manager *mgr, int zpos) if (win != 0) fimd_win_set_colkey(ctx, win); - /* wincon */ - val = readl(ctx->regs + WINCON(win)); - val |= WINCONx_ENWIN; - writel(val, ctx->regs + WINCON(win)); + fimd_enable_video_output(ctx, win, true); if (ctx->driver_data->has_shadowcon) { val = readl(ctx->regs + SHADOWCON); @@ -790,10 +798,7 @@ static void fimd_win_disable(struct exynos_drm_manager *mgr, int zpos) /* protect windows */ fimd_shadow_protect_win(ctx, win, true); - /* wincon */ - val = readl(ctx->regs + WINCON(win)); - val &= ~WINCONx_ENWIN; - writel(val, ctx->regs + WINCON(win)); + fimd_enable_video_output(ctx, win, false); /* unprotect windows */ if (ctx->driver_data->has_shadowcon) { -- cgit v0.10.2 From 999d8b31fac3629b8179d7a40b7930deedadeb99 Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Mon, 17 Nov 2014 22:00:11 +0900 Subject: drm/exynos: fimd: add fimd_enable_shadow_channel_path() to cleanup This function is valid only the SoC has SHADOWCON register and it should be used together with fimd_enable_video_output() to match the ENWIN_F bit in WINCON# and C#_EN_F bit in SHADOWCON. Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 5cfd251..fd7b469 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -242,6 +242,19 @@ static void fimd_enable_video_output(struct fimd_context *ctx, int win, writel(val, ctx->regs + WINCON(win)); } +static void fimd_enable_shadow_channel_path(struct fimd_context *ctx, int win, + bool enable) +{ + u32 val = readl(ctx->regs + SHADOWCON); + + if (enable) + val |= SHADOWCON_CHx_ENABLE(win); + else + val &= ~SHADOWCON_CHx_ENABLE(win); + + writel(val, ctx->regs + SHADOWCON); +} + static void fimd_clear_channel(struct exynos_drm_manager *mgr) { struct fimd_context *ctx = mgr->ctx; @@ -256,12 +269,10 @@ static void fimd_clear_channel(struct exynos_drm_manager *mgr) if (val & WINCONx_ENWIN) { fimd_enable_video_output(ctx, win, false); - /* unprotect windows */ - if (ctx->driver_data->has_shadowcon) { - val = readl(ctx->regs + SHADOWCON); - val &= ~SHADOWCON_CHx_ENABLE(win); - writel(val, ctx->regs + SHADOWCON); - } + if (ctx->driver_data->has_shadowcon) + fimd_enable_shadow_channel_path(ctx, win, + false); + ch_enabled = 1; } } @@ -759,11 +770,8 @@ static void fimd_win_commit(struct exynos_drm_manager *mgr, int zpos) fimd_enable_video_output(ctx, win, true); - if (ctx->driver_data->has_shadowcon) { - val = readl(ctx->regs + SHADOWCON); - val |= SHADOWCON_CHx_ENABLE(win); - writel(val, ctx->regs + SHADOWCON); - } + if (ctx->driver_data->has_shadowcon) + fimd_enable_shadow_channel_path(ctx, win, true); /* Enable DMA channel and unprotect windows */ fimd_shadow_protect_win(ctx, win, false); @@ -779,7 +787,6 @@ static void fimd_win_disable(struct exynos_drm_manager *mgr, int zpos) struct fimd_context *ctx = mgr->ctx; struct fimd_win_data *win_data; int win = zpos; - u32 val; if (win == DEFAULT_ZPOS) win = ctx->default_win; @@ -800,13 +807,10 @@ static void fimd_win_disable(struct exynos_drm_manager *mgr, int zpos) fimd_enable_video_output(ctx, win, false); - /* unprotect windows */ - if (ctx->driver_data->has_shadowcon) { - val = readl(ctx->regs + SHADOWCON); - val &= ~SHADOWCON_CHx_ENABLE(win); - writel(val, ctx->regs + SHADOWCON); - } + if (ctx->driver_data->has_shadowcon) + fimd_enable_shadow_channel_path(ctx, win, false); + /* unprotect windows */ fimd_shadow_protect_win(ctx, win, false); win_data->enabled = false; -- cgit v0.10.2 From 1c905d9508932ed85eee5469519aedc2d820458e Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Mon, 17 Nov 2014 22:00:12 +0900 Subject: drm/exynos: fimd: modify I80 i/f irq relevant routine For the I80 interface, the video interrupt pending register(VIDINTCON1) should be handled in fimd_irq_handler() and the video interrupt control register(VIDINTCON0) should be handled in fimd_enable_vblank() and fimd_disable_vblank() like RGB interface. So this patch moves each set / unset routines into proper positions. Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index fd7b469..3c63237 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -469,12 +469,19 @@ static int fimd_enable_vblank(struct exynos_drm_manager *mgr) val = readl(ctx->regs + VIDINTCON0); val |= VIDINTCON0_INT_ENABLE; - val |= VIDINTCON0_INT_FRAME; - val &= ~VIDINTCON0_FRAMESEL0_MASK; - val |= VIDINTCON0_FRAMESEL0_VSYNC; - val &= ~VIDINTCON0_FRAMESEL1_MASK; - val |= VIDINTCON0_FRAMESEL1_NONE; + if (ctx->i80_if) { + val |= VIDINTCON0_INT_I80IFDONE; + val |= VIDINTCON0_INT_SYSMAINCON; + val &= ~VIDINTCON0_INT_SYSSUBCON; + } else { + val |= VIDINTCON0_INT_FRAME; + + val &= ~VIDINTCON0_FRAMESEL0_MASK; + val |= VIDINTCON0_FRAMESEL0_VSYNC; + val &= ~VIDINTCON0_FRAMESEL1_MASK; + val |= VIDINTCON0_FRAMESEL1_NONE; + } writel(val, ctx->regs + VIDINTCON0); } @@ -493,9 +500,15 @@ static void fimd_disable_vblank(struct exynos_drm_manager *mgr) if (test_and_clear_bit(0, &ctx->irq_flags)) { val = readl(ctx->regs + VIDINTCON0); - val &= ~VIDINTCON0_INT_FRAME; val &= ~VIDINTCON0_INT_ENABLE; + if (ctx->i80_if) { + val &= ~VIDINTCON0_INT_I80IFDONE; + val &= ~VIDINTCON0_INT_SYSMAINCON; + val &= ~VIDINTCON0_INT_SYSSUBCON; + } else + val &= ~VIDINTCON0_INT_FRAME; + writel(val, ctx->regs + VIDINTCON0); } } @@ -959,19 +972,15 @@ static void fimd_trigger(struct device *dev) u32 reg; /* - * Skips to trigger if in triggering state, because multiple triggering - * requests can cause panel reset. - */ + * Skips triggering if in triggering state, because multiple triggering + * requests can cause panel reset. + */ if (atomic_read(&ctx->triggering)) return; + /* Enters triggering mode */ atomic_set(&ctx->triggering, 1); - reg = readl(ctx->regs + VIDINTCON0); - reg |= (VIDINTCON0_INT_ENABLE | VIDINTCON0_INT_I80IFDONE | - VIDINTCON0_INT_SYSMAINCON); - writel(reg, ctx->regs + VIDINTCON0); - reg = readl(timing_base + TRIGCON); reg |= (TRGMODE_I80_RGB_ENABLE_I80 | SWTRGCMD_I80_RGB_ENABLE); writel(reg, timing_base + TRIGCON); @@ -1036,21 +1045,13 @@ static irqreturn_t fimd_irq_handler(int irq, void *dev_id) if (ctx->pipe < 0 || !ctx->drm_dev) goto out; - if (ctx->i80_if) { - /* unset I80 frame done interrupt */ - val = readl(ctx->regs + VIDINTCON0); - val &= ~(VIDINTCON0_INT_I80IFDONE | VIDINTCON0_INT_SYSMAINCON); - writel(val, ctx->regs + VIDINTCON0); + drm_handle_vblank(ctx->drm_dev, ctx->pipe); + exynos_drm_crtc_finish_pageflip(ctx->drm_dev, ctx->pipe); - /* exit triggering mode */ + if (ctx->i80_if) { + /* Exits triggering mode */ atomic_set(&ctx->triggering, 0); - - drm_handle_vblank(ctx->drm_dev, ctx->pipe); - exynos_drm_crtc_finish_pageflip(ctx->drm_dev, ctx->pipe); } else { - drm_handle_vblank(ctx->drm_dev, ctx->pipe); - exynos_drm_crtc_finish_pageflip(ctx->drm_dev, ctx->pipe); - /* set wait vsync event to zero and wake up queue. */ if (atomic_read(&ctx->wait_vsync_event)) { atomic_set(&ctx->wait_vsync_event, 0); -- cgit v0.10.2 From 87ab85b3ccbc266154268407f85777a74f8bf820 Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Mon, 17 Nov 2014 22:00:13 +0900 Subject: drm/exynos: fimd: add triggering unset routine in fimd_trigger() There is a case like set config which requires triggering but vblank is not enabled yet. So triggering unset routine is required to exit from triggering mode. Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 3c63237..e488b80 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -984,6 +984,13 @@ static void fimd_trigger(struct device *dev) reg = readl(timing_base + TRIGCON); reg |= (TRGMODE_I80_RGB_ENABLE_I80 | SWTRGCMD_I80_RGB_ENABLE); writel(reg, timing_base + TRIGCON); + + /* + * Exits triggering mode if vblank is not enabled yet, because when the + * VIDINTCON0 register is not set, it can not exit from triggering mode. + */ + if (!test_bit(0, &ctx->irq_flags)) + atomic_set(&ctx->triggering, 0); } static void fimd_te_handler(struct exynos_drm_manager *mgr) -- cgit v0.10.2 From adf67abff09110b527431512bf05461a60ef2a72 Mon Sep 17 00:00:00 2001 From: Joonyoung Shim Date: Mon, 17 Nov 2014 22:00:14 +0900 Subject: drm/exynos: use irq_flags instead of triggering The drm_handle_vblank should be called whenever be vsync, te interrupt means vsync on i80 interface. Signed-off-by: Joonyoung Shim Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index e488b80..a8ab3ec 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -1014,7 +1014,7 @@ static void fimd_te_handler(struct exynos_drm_manager *mgr) wake_up(&ctx->wait_vsync_queue); } - if (!atomic_read(&ctx->triggering)) + if (test_bit(0, &ctx->irq_flags)) drm_handle_vblank(ctx->drm_dev, ctx->pipe); } @@ -1052,13 +1052,15 @@ static irqreturn_t fimd_irq_handler(int irq, void *dev_id) if (ctx->pipe < 0 || !ctx->drm_dev) goto out; - drm_handle_vblank(ctx->drm_dev, ctx->pipe); - exynos_drm_crtc_finish_pageflip(ctx->drm_dev, ctx->pipe); - if (ctx->i80_if) { + exynos_drm_crtc_finish_pageflip(ctx->drm_dev, ctx->pipe); + /* Exits triggering mode */ atomic_set(&ctx->triggering, 0); } else { + drm_handle_vblank(ctx->drm_dev, ctx->pipe); + exynos_drm_crtc_finish_pageflip(ctx->drm_dev, ctx->pipe); + /* set wait vsync event to zero and wake up queue. */ if (atomic_read(&ctx->wait_vsync_event)) { atomic_set(&ctx->wait_vsync_event, 0); -- cgit v0.10.2 From ecb84157b5716f80e6569cd379f3bf560116a803 Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Mon, 17 Nov 2014 22:00:15 +0900 Subject: drm/exynos: dsi: move TE irq handler registration position The drm_helper_hpd_irq_event() does dpms control and the panel is initialized and displayed on by it. So the exynos_dsi_te_irq_handler() should be registered beforehand. Signed-off-by: YoungJun Cho Acked-by: Inki Dae Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c index 44461aa..156a638 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c @@ -1207,9 +1207,6 @@ static int exynos_dsi_host_attach(struct mipi_dsi_host *host, dsi->mode_flags = device->mode_flags; dsi->panel_node = device->dev.of_node; - if (dsi->connector.dev) - drm_helper_hpd_irq_event(dsi->connector.dev); - /* * This is a temporary solution and should be made by more generic way. * @@ -1223,6 +1220,9 @@ static int exynos_dsi_host_attach(struct mipi_dsi_host *host, return ret; } + if (dsi->connector.dev) + drm_helper_hpd_irq_event(dsi->connector.dev); + return 0; } -- cgit v0.10.2 From 0cef83a5fd3aa72ed7bd8661042cd8405062fcef Mon Sep 17 00:00:00 2001 From: YoungJun Cho Date: Mon, 17 Nov 2014 22:00:16 +0900 Subject: drm/exynos: dsi: set TE GPIO IRQ status as IRQ_NOAUTOEN The exynos_dsi_te_irq_handler() works only dsi(DPMS) is on. So it is enough to enable and disable TE GPIO IRQ in exynos_dsi_enable(disable)_irq() like DSI IRQ. Signed-off-by: YoungJun Cho Acked-by: Inki Dae a Acked-by: Kyungmin Park Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c index 156a638..05fe93d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c @@ -1155,6 +1155,7 @@ static int exynos_dsi_init(struct exynos_dsi *dsi) static int exynos_dsi_register_te_irq(struct exynos_dsi *dsi) { int ret; + int te_gpio_irq; dsi->te_gpio = of_get_named_gpio(dsi->panel_node, "te-gpios", 0); if (!gpio_is_valid(dsi->te_gpio)) { @@ -1169,14 +1170,10 @@ static int exynos_dsi_register_te_irq(struct exynos_dsi *dsi) goto out; } - /* - * This TE GPIO IRQ should not be set to IRQ_NOAUTOEN, because panel - * calls drm_panel_init() first then calls mipi_dsi_attach() in probe(). - * It means that te_gpio is invalid when exynos_dsi_enable_irq() is - * called by drm_panel_init() before panel is attached. - */ - ret = request_threaded_irq(gpio_to_irq(dsi->te_gpio), - exynos_dsi_te_irq_handler, NULL, + te_gpio_irq = gpio_to_irq(dsi->te_gpio); + + irq_set_status_flags(te_gpio_irq, IRQ_NOAUTOEN); + ret = request_threaded_irq(te_gpio_irq, exynos_dsi_te_irq_handler, NULL, IRQF_TRIGGER_RISING, "TE", dsi); if (ret) { dev_err(dsi->dev, "request interrupt failed with %d\n", ret); -- cgit v0.10.2 From 8103ef1b5f471fe5e50cd4e1fc17cf3be4682d79 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 24 Nov 2014 14:12:46 +0900 Subject: drm/exynos/mixer: embed manager into private context exynos_drm_manager is used by internal Exynos DRM framework for representing crtc. As it should be mapped 1:1 to fimd private context it seems more reasonable to embed it directly in that context. As a result further code simplification will be possible. Moreover it will be possible to handle multiple mixer devices in the system. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_mixer.c b/drivers/gpu/drm/exynos/exynos_mixer.c index a41c84e..1b1fd82 100644 --- a/drivers/gpu/drm/exynos/exynos_mixer.c +++ b/drivers/gpu/drm/exynos/exynos_mixer.c @@ -40,8 +40,6 @@ #include "exynos_drm_iommu.h" #include "exynos_mixer.h" -#define get_mixer_manager(dev) platform_get_drvdata(to_platform_device(dev)) - #define MIXER_WIN_NR 3 #define MIXER_DEFAULT_WIN 0 @@ -86,6 +84,7 @@ enum mixer_version_id { }; struct mixer_context { + struct exynos_drm_manager manager; struct platform_device *pdev; struct device *dev; struct drm_device *drm_dev; @@ -1187,11 +1186,6 @@ static struct exynos_drm_manager_ops mixer_manager_ops = { .win_disable = mixer_win_disable, }; -static struct exynos_drm_manager mixer_manager = { - .type = EXYNOS_DISPLAY_TYPE_HDMI, - .ops = &mixer_manager_ops, -}; - static struct mixer_drv_data exynos5420_mxr_drv_data = { .version = MXR_VER_128_0_0_184, .is_vp_enabled = 0, @@ -1249,48 +1243,17 @@ MODULE_DEVICE_TABLE(of, mixer_match_types); static int mixer_bind(struct device *dev, struct device *manager, void *data) { - struct platform_device *pdev = to_platform_device(dev); + struct mixer_context *ctx = dev_get_drvdata(dev); struct drm_device *drm_dev = data; - struct mixer_context *ctx; - struct mixer_drv_data *drv; int ret; - dev_info(dev, "probe start\n"); - - ctx = devm_kzalloc(&pdev->dev, sizeof(*ctx), GFP_KERNEL); - if (!ctx) { - DRM_ERROR("failed to alloc mixer context.\n"); - return -ENOMEM; - } - - mutex_init(&ctx->mixer_mutex); - - if (dev->of_node) { - const struct of_device_id *match; - match = of_match_node(mixer_match_types, dev->of_node); - drv = (struct mixer_drv_data *)match->data; - } else { - drv = (struct mixer_drv_data *) - platform_get_device_id(pdev)->driver_data; - } - - ctx->pdev = pdev; - ctx->dev = dev; - ctx->vp_enabled = drv->is_vp_enabled; - ctx->has_sclk = drv->has_sclk; - ctx->mxr_ver = drv->version; - init_waitqueue_head(&ctx->wait_vsync_queue); - atomic_set(&ctx->wait_vsync_event, 0); - - mixer_manager.ctx = ctx; - ret = mixer_initialize(&mixer_manager, drm_dev); + ret = mixer_initialize(&ctx->manager, drm_dev); if (ret) return ret; - platform_set_drvdata(pdev, &mixer_manager); - ret = exynos_drm_crtc_create(&mixer_manager); + ret = exynos_drm_crtc_create(&ctx->manager); if (ret) { - mixer_mgr_remove(&mixer_manager); + mixer_mgr_remove(&ctx->manager); return ret; } @@ -1301,11 +1264,9 @@ static int mixer_bind(struct device *dev, struct device *manager, void *data) static void mixer_unbind(struct device *dev, struct device *master, void *data) { - struct exynos_drm_manager *mgr = dev_get_drvdata(dev); + struct mixer_context *ctx = dev_get_drvdata(dev); - dev_info(dev, "remove successful\n"); - - mixer_mgr_remove(mgr); + mixer_mgr_remove(&ctx->manager); pm_runtime_disable(dev); } @@ -1317,22 +1278,63 @@ static const struct component_ops mixer_component_ops = { static int mixer_probe(struct platform_device *pdev) { + struct device *dev = &pdev->dev; + struct mixer_drv_data *drv; + struct mixer_context *ctx; int ret; + ctx = devm_kzalloc(&pdev->dev, sizeof(*ctx), GFP_KERNEL); + if (!ctx) { + DRM_ERROR("failed to alloc mixer context.\n"); + return -ENOMEM; + } + + mutex_init(&ctx->mixer_mutex); + + ctx->manager.type = EXYNOS_DISPLAY_TYPE_HDMI; + ctx->manager.ops = &mixer_manager_ops; + + if (dev->of_node) { + const struct of_device_id *match; + + match = of_match_node(mixer_match_types, dev->of_node); + drv = (struct mixer_drv_data *)match->data; + } else { + drv = (struct mixer_drv_data *) + platform_get_device_id(pdev)->driver_data; + } + + ctx->pdev = pdev; + ctx->dev = dev; + ctx->vp_enabled = drv->is_vp_enabled; + ctx->has_sclk = drv->has_sclk; + ctx->mxr_ver = drv->version; + init_waitqueue_head(&ctx->wait_vsync_queue); + atomic_set(&ctx->wait_vsync_event, 0); + ctx->manager.ctx = ctx; + + platform_set_drvdata(pdev, ctx); + ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CRTC, - mixer_manager.type); + ctx->manager.type); if (ret) return ret; ret = component_add(&pdev->dev, &mixer_component_ops); - if (ret) + if (ret) { exynos_drm_component_del(&pdev->dev, EXYNOS_DEVICE_TYPE_CRTC); + return ret; + } + + pm_runtime_enable(dev); return ret; } static int mixer_remove(struct platform_device *pdev) { + pm_runtime_disable(&pdev->dev); + component_del(&pdev->dev, &mixer_component_ops); exynos_drm_component_del(&pdev->dev, EXYNOS_DEVICE_TYPE_CRTC); -- cgit v0.10.2 From 8f0be8304defc33d32ed26a7b1eec54270b995a7 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 24 Nov 2014 14:14:49 +0900 Subject: drm/exynos/mixer: stop using manager->ctx pointer The patch replaces accesses to manager->ctx pointer by container_of construct. It will allow to remove ctx field in the future. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_mixer.c b/drivers/gpu/drm/exynos/exynos_mixer.c index 1b1fd82..820b762 100644 --- a/drivers/gpu/drm/exynos/exynos_mixer.c +++ b/drivers/gpu/drm/exynos/exynos_mixer.c @@ -103,6 +103,11 @@ struct mixer_context { atomic_t wait_vsync_event; }; +static inline struct mixer_context *mgr_to_mixer(struct exynos_drm_manager *mgr) +{ + return container_of(mgr, struct mixer_context, manager); +} + struct mixer_drv_data { enum mixer_version_id version; bool is_vp_enabled; @@ -853,7 +858,7 @@ static int mixer_initialize(struct exynos_drm_manager *mgr, struct drm_device *drm_dev) { int ret; - struct mixer_context *mixer_ctx = mgr->ctx; + struct mixer_context *mixer_ctx = mgr_to_mixer(mgr); struct exynos_drm_private *priv; priv = drm_dev->dev_private; @@ -884,7 +889,7 @@ static int mixer_initialize(struct exynos_drm_manager *mgr, static void mixer_mgr_remove(struct exynos_drm_manager *mgr) { - struct mixer_context *mixer_ctx = mgr->ctx; + struct mixer_context *mixer_ctx = mgr_to_mixer(mgr); if (is_drm_iommu_supported(mixer_ctx->drm_dev)) drm_iommu_detach_device(mixer_ctx->drm_dev, mixer_ctx->dev); @@ -892,7 +897,7 @@ static void mixer_mgr_remove(struct exynos_drm_manager *mgr) static int mixer_enable_vblank(struct exynos_drm_manager *mgr) { - struct mixer_context *mixer_ctx = mgr->ctx; + struct mixer_context *mixer_ctx = mgr_to_mixer(mgr); struct mixer_resources *res = &mixer_ctx->mixer_res; if (!mixer_ctx->powered) { @@ -909,7 +914,7 @@ static int mixer_enable_vblank(struct exynos_drm_manager *mgr) static void mixer_disable_vblank(struct exynos_drm_manager *mgr) { - struct mixer_context *mixer_ctx = mgr->ctx; + struct mixer_context *mixer_ctx = mgr_to_mixer(mgr); struct mixer_resources *res = &mixer_ctx->mixer_res; /* disable vsync interrupt */ @@ -919,7 +924,7 @@ static void mixer_disable_vblank(struct exynos_drm_manager *mgr) static void mixer_win_mode_set(struct exynos_drm_manager *mgr, struct exynos_drm_overlay *overlay) { - struct mixer_context *mixer_ctx = mgr->ctx; + struct mixer_context *mixer_ctx = mgr_to_mixer(mgr); struct hdmi_win_data *win_data; int win; @@ -970,7 +975,7 @@ static void mixer_win_mode_set(struct exynos_drm_manager *mgr, static void mixer_win_commit(struct exynos_drm_manager *mgr, int zpos) { - struct mixer_context *mixer_ctx = mgr->ctx; + struct mixer_context *mixer_ctx = mgr_to_mixer(mgr); int win = zpos == DEFAULT_ZPOS ? MIXER_DEFAULT_WIN : zpos; DRM_DEBUG_KMS("win: %d\n", win); @@ -992,7 +997,7 @@ static void mixer_win_commit(struct exynos_drm_manager *mgr, int zpos) static void mixer_win_disable(struct exynos_drm_manager *mgr, int zpos) { - struct mixer_context *mixer_ctx = mgr->ctx; + struct mixer_context *mixer_ctx = mgr_to_mixer(mgr); struct mixer_resources *res = &mixer_ctx->mixer_res; int win = zpos == DEFAULT_ZPOS ? MIXER_DEFAULT_WIN : zpos; unsigned long flags; @@ -1020,7 +1025,7 @@ static void mixer_win_disable(struct exynos_drm_manager *mgr, int zpos) static void mixer_wait_for_vblank(struct exynos_drm_manager *mgr) { - struct mixer_context *mixer_ctx = mgr->ctx; + struct mixer_context *mixer_ctx = mgr_to_mixer(mgr); mutex_lock(&mixer_ctx->mixer_mutex); if (!mixer_ctx->powered) { @@ -1047,7 +1052,7 @@ static void mixer_wait_for_vblank(struct exynos_drm_manager *mgr) static void mixer_window_suspend(struct exynos_drm_manager *mgr) { - struct mixer_context *ctx = mgr->ctx; + struct mixer_context *ctx = mgr_to_mixer(mgr); struct hdmi_win_data *win_data; int i; @@ -1061,7 +1066,7 @@ static void mixer_window_suspend(struct exynos_drm_manager *mgr) static void mixer_window_resume(struct exynos_drm_manager *mgr) { - struct mixer_context *ctx = mgr->ctx; + struct mixer_context *ctx = mgr_to_mixer(mgr); struct hdmi_win_data *win_data; int i; @@ -1076,7 +1081,7 @@ static void mixer_window_resume(struct exynos_drm_manager *mgr) static void mixer_poweron(struct exynos_drm_manager *mgr) { - struct mixer_context *ctx = mgr->ctx; + struct mixer_context *ctx = mgr_to_mixer(mgr); struct mixer_resources *res = &ctx->mixer_res; mutex_lock(&ctx->mixer_mutex); @@ -1110,7 +1115,7 @@ static void mixer_poweron(struct exynos_drm_manager *mgr) static void mixer_poweroff(struct exynos_drm_manager *mgr) { - struct mixer_context *ctx = mgr->ctx; + struct mixer_context *ctx = mgr_to_mixer(mgr); struct mixer_resources *res = &ctx->mixer_res; mutex_lock(&ctx->mixer_mutex); @@ -1311,7 +1316,6 @@ static int mixer_probe(struct platform_device *pdev) ctx->mxr_ver = drv->version; init_waitqueue_head(&ctx->wait_vsync_queue); atomic_set(&ctx->wait_vsync_event, 0); - ctx->manager.ctx = ctx; platform_set_drvdata(pdev, ctx); -- cgit v0.10.2 From f01833cd2b36767d350c4f8b3eaaf48c4d2f7da3 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:16 +0100 Subject: drm/exynos/vidi: embed manager into private context exynos_drm_manager is used by internal Exynos DRM framework for representing crtc. As it should be mapped 1:1 to vidi private context it seems more reasonable to embed it directly in that context. As a result further code simplification will be possible. Moreover it will be possible to handle multiple mixer devices in the system. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_vidi.c b/drivers/gpu/drm/exynos/exynos_drm_vidi.c index 50faf91..f47939c 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_vidi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_vidi.c @@ -28,7 +28,6 @@ /* vidi has totally three virtual windows. */ #define WINDOWS_NR 3 -#define get_vidi_mgr(dev) platform_get_drvdata(to_platform_device(dev)) #define ctx_from_connector(c) container_of(c, struct vidi_context, \ connector) @@ -47,6 +46,7 @@ struct vidi_win_data { }; struct vidi_context { + struct exynos_drm_manager manager; struct drm_device *drm_dev; struct drm_crtc *crtc; struct drm_encoder *encoder; @@ -316,11 +316,6 @@ static struct exynos_drm_manager_ops vidi_manager_ops = { .win_disable = vidi_win_disable, }; -static struct exynos_drm_manager vidi_manager = { - .type = EXYNOS_DISPLAY_TYPE_VIDI, - .ops = &vidi_manager_ops, -}; - static void vidi_fake_vblank_handler(struct work_struct *work) { struct vidi_context *ctx = container_of(work, struct vidi_context, @@ -349,9 +344,8 @@ static void vidi_fake_vblank_handler(struct work_struct *work) static int vidi_show_connection(struct device *dev, struct device_attribute *attr, char *buf) { + struct vidi_context *ctx = dev_get_drvdata(dev); int rc; - struct exynos_drm_manager *mgr = get_vidi_mgr(dev); - struct vidi_context *ctx = mgr->ctx; mutex_lock(&ctx->lock); @@ -366,8 +360,7 @@ static int vidi_store_connection(struct device *dev, struct device_attribute *attr, const char *buf, size_t len) { - struct exynos_drm_manager *mgr = get_vidi_mgr(dev); - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = dev_get_drvdata(dev); int ret; ret = kstrtoint(buf, 0, &ctx->connected); @@ -563,14 +556,13 @@ static struct exynos_drm_display vidi_display = { static int vidi_subdrv_probe(struct drm_device *drm_dev, struct device *dev) { - struct exynos_drm_manager *mgr = get_vidi_mgr(dev); - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = dev_get_drvdata(dev); struct drm_crtc *crtc = ctx->crtc; int ret; - vidi_mgr_initialize(mgr, drm_dev); + vidi_mgr_initialize(&ctx->manager, drm_dev); - ret = exynos_drm_crtc_create(&vidi_manager); + ret = exynos_drm_crtc_create(&ctx->manager); if (ret) { DRM_ERROR("failed to create crtc.\n"); return ret; @@ -596,16 +588,18 @@ static int vidi_probe(struct platform_device *pdev) if (!ctx) return -ENOMEM; + ctx->manager.type = EXYNOS_DISPLAY_TYPE_VIDI; + ctx->manager.ops = &vidi_manager_ops; ctx->default_win = 0; INIT_WORK(&ctx->work, vidi_fake_vblank_handler); - vidi_manager.ctx = ctx; + ctx->manager.ctx = ctx; vidi_display.ctx = ctx; mutex_init(&ctx->lock); - platform_set_drvdata(pdev, &vidi_manager); + platform_set_drvdata(pdev, ctx); subdrv = &ctx->subdrv; subdrv->dev = &pdev->dev; @@ -628,8 +622,7 @@ static int vidi_probe(struct platform_device *pdev) static int vidi_remove(struct platform_device *pdev) { - struct exynos_drm_manager *mgr = platform_get_drvdata(pdev); - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = platform_get_drvdata(pdev); if (ctx->raw_edid != (struct edid *)fake_edid_info) { kfree(ctx->raw_edid); @@ -668,12 +661,19 @@ int exynos_drm_probe_vidi(void) return ret; } +static int exynos_drm_remove_vidi_device(struct device *dev, void *data) +{ + platform_device_unregister(to_platform_device(dev)); + + return 0; +} + void exynos_drm_remove_vidi(void) { - struct vidi_context *ctx = vidi_manager.ctx; - struct exynos_drm_subdrv *subdrv = &ctx->subdrv; - struct platform_device *pdev = to_platform_device(subdrv->dev); + int ret = driver_for_each_device(&vidi_driver.driver, NULL, NULL, + exynos_drm_remove_vidi_device); + /* silence compiler warning */ + (void)ret; platform_driver_unregister(&vidi_driver); - platform_device_unregister(pdev); } -- cgit v0.10.2 From e1819aad14ad6bd799ae5c1f97412b1dd55d2065 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:17 +0100 Subject: drm/exynos/vidi: stop using manager->ctx pointer The patch replaces accesses to manager->ctx pointer by container_of construct. It will allow to remove ctx field in the future. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_vidi.c b/drivers/gpu/drm/exynos/exynos_drm_vidi.c index f47939c..f048a90 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_vidi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_vidi.c @@ -66,6 +66,11 @@ struct vidi_context { int pipe; }; +static inline struct vidi_context *manager_to_vidi(struct exynos_drm_manager *m) +{ + return container_of(m, struct vidi_context, manager); +} + static const char fake_edid_info[] = { 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00, 0x4c, 0x2d, 0x05, 0x05, 0x00, 0x00, 0x00, 0x00, 0x30, 0x12, 0x01, 0x03, 0x80, 0x10, 0x09, 0x78, @@ -93,7 +98,7 @@ static const char fake_edid_info[] = { static void vidi_apply(struct exynos_drm_manager *mgr) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); struct exynos_drm_manager_ops *mgr_ops = mgr->ops; struct vidi_win_data *win_data; int i; @@ -110,7 +115,7 @@ static void vidi_apply(struct exynos_drm_manager *mgr) static void vidi_commit(struct exynos_drm_manager *mgr) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); if (ctx->suspended) return; @@ -118,7 +123,7 @@ static void vidi_commit(struct exynos_drm_manager *mgr) static int vidi_enable_vblank(struct exynos_drm_manager *mgr) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); if (ctx->suspended) return -EPERM; @@ -140,7 +145,7 @@ static int vidi_enable_vblank(struct exynos_drm_manager *mgr) static void vidi_disable_vblank(struct exynos_drm_manager *mgr) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); if (ctx->suspended) return; @@ -152,7 +157,7 @@ static void vidi_disable_vblank(struct exynos_drm_manager *mgr) static void vidi_win_mode_set(struct exynos_drm_manager *mgr, struct exynos_drm_overlay *overlay) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); struct vidi_win_data *win_data; int win; unsigned long offset; @@ -204,7 +209,7 @@ static void vidi_win_mode_set(struct exynos_drm_manager *mgr, static void vidi_win_commit(struct exynos_drm_manager *mgr, int zpos) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); struct vidi_win_data *win_data; int win = zpos; @@ -229,7 +234,7 @@ static void vidi_win_commit(struct exynos_drm_manager *mgr, int zpos) static void vidi_win_disable(struct exynos_drm_manager *mgr, int zpos) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); struct vidi_win_data *win_data; int win = zpos; @@ -247,7 +252,7 @@ static void vidi_win_disable(struct exynos_drm_manager *mgr, int zpos) static int vidi_power_on(struct exynos_drm_manager *mgr, bool enable) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); DRM_DEBUG_KMS("%s\n", __FILE__); @@ -271,7 +276,7 @@ static int vidi_power_on(struct exynos_drm_manager *mgr, bool enable) static void vidi_dpms(struct exynos_drm_manager *mgr, int mode) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); DRM_DEBUG_KMS("%d\n", mode); @@ -297,7 +302,7 @@ static void vidi_dpms(struct exynos_drm_manager *mgr, int mode) static int vidi_mgr_initialize(struct exynos_drm_manager *mgr, struct drm_device *drm_dev) { - struct vidi_context *ctx = mgr->ctx; + struct vidi_context *ctx = manager_to_vidi(mgr); struct exynos_drm_private *priv = drm_dev->dev_private; mgr->drm_dev = ctx->drm_dev = drm_dev; @@ -594,7 +599,6 @@ static int vidi_probe(struct platform_device *pdev) INIT_WORK(&ctx->work, vidi_fake_vblank_handler); - ctx->manager.ctx = ctx; vidi_display.ctx = ctx; mutex_init(&ctx->lock); -- cgit v0.10.2 From e152dbd77bc6388c910a3fd2d2acd2439d24631a Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:18 +0100 Subject: drm/exynos/fimd: embed manager into private context exynos_drm_manager is used by internal Exynos DRM framework for representing crtc. As it should be mapped 1:1 to fimd private context it seems more reasonable to embed it directly in that context. As a result further code simplification will be possible. Moreover it will be possible to handle multiple FIMD devices in the system. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index a8ab3ec..9a5e999 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -84,8 +84,6 @@ /* FIMD has totally five hardware windows. */ #define WINDOWS_NR 5 -#define get_fimd_manager(mgr) platform_get_drvdata(to_platform_device(dev)) - struct fimd_driver_data { unsigned int timing_base; unsigned int lcdblk_offset; @@ -159,6 +157,7 @@ struct fimd_win_data { }; struct fimd_context { + struct exynos_drm_manager manager; struct device *dev; struct drm_device *drm_dev; struct clk *bus_clk; @@ -965,8 +964,7 @@ static void fimd_dpms(struct exynos_drm_manager *mgr, int mode) static void fimd_trigger(struct device *dev) { - struct exynos_drm_manager *mgr = get_fimd_manager(dev); - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = dev_get_drvdata(dev); struct fimd_driver_data *driver_data = ctx->driver_data; void *timing_base = ctx->regs + driver_data->timing_base; u32 reg; @@ -1032,11 +1030,6 @@ static struct exynos_drm_manager_ops fimd_manager_ops = { .te_handler = fimd_te_handler, }; -static struct exynos_drm_manager fimd_manager = { - .type = EXYNOS_DISPLAY_TYPE_LCD, - .ops = &fimd_manager_ops, -}; - static irqreturn_t fimd_irq_handler(int irq, void *dev_id) { struct fimd_context *ctx = (struct fimd_context *)dev_id; @@ -1074,11 +1067,11 @@ out: static int fimd_bind(struct device *dev, struct device *master, void *data) { - struct fimd_context *ctx = fimd_manager.ctx; + struct fimd_context *ctx = dev_get_drvdata(dev); struct drm_device *drm_dev = data; - fimd_mgr_initialize(&fimd_manager, drm_dev); - exynos_drm_crtc_create(&fimd_manager); + fimd_mgr_initialize(&ctx->manager, drm_dev); + exynos_drm_crtc_create(&ctx->manager); if (ctx->display) exynos_drm_create_enc_conn(drm_dev, ctx->display); @@ -1089,15 +1082,14 @@ static int fimd_bind(struct device *dev, struct device *master, void *data) static void fimd_unbind(struct device *dev, struct device *master, void *data) { - struct exynos_drm_manager *mgr = dev_get_drvdata(dev); - struct fimd_context *ctx = fimd_manager.ctx; + struct fimd_context *ctx = dev_get_drvdata(dev); - fimd_dpms(mgr, DRM_MODE_DPMS_OFF); + fimd_dpms(&ctx->manager, DRM_MODE_DPMS_OFF); if (ctx->display) exynos_dpi_remove(dev); - fimd_mgr_remove(mgr); + fimd_mgr_remove(&ctx->manager); } static const struct component_ops fimd_component_ops = { @@ -1113,21 +1105,20 @@ static int fimd_probe(struct platform_device *pdev) struct resource *res; int ret = -EINVAL; - ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CRTC, - fimd_manager.type); - if (ret) - return ret; - - if (!dev->of_node) { - ret = -ENODEV; - goto err_del_component; - } + if (!dev->of_node) + return -ENODEV; ctx = devm_kzalloc(dev, sizeof(*ctx), GFP_KERNEL); - if (!ctx) { - ret = -ENOMEM; - goto err_del_component; - } + if (!ctx) + return -ENOMEM; + + ctx->manager.type = EXYNOS_DISPLAY_TYPE_LCD; + ctx->manager.ops = &fimd_manager_ops; + + ret = exynos_drm_component_add(dev, EXYNOS_DEVICE_TYPE_CRTC, + ctx->manager.type); + if (ret) + return ret; ctx->dev = dev; ctx->suspended = true; @@ -1215,28 +1206,27 @@ static int fimd_probe(struct platform_device *pdev) init_waitqueue_head(&ctx->wait_vsync_queue); atomic_set(&ctx->wait_vsync_event, 0); + ctx->manager.ctx = ctx; - platform_set_drvdata(pdev, &fimd_manager); - - fimd_manager.ctx = ctx; + platform_set_drvdata(pdev, ctx); ctx->display = exynos_dpi_probe(dev); if (IS_ERR(ctx->display)) return PTR_ERR(ctx->display); - pm_runtime_enable(&pdev->dev); + pm_runtime_enable(dev); - ret = component_add(&pdev->dev, &fimd_component_ops); + ret = component_add(dev, &fimd_component_ops); if (ret) goto err_disable_pm_runtime; return ret; err_disable_pm_runtime: - pm_runtime_disable(&pdev->dev); + pm_runtime_disable(dev); err_del_component: - exynos_drm_component_del(&pdev->dev, EXYNOS_DEVICE_TYPE_CRTC); + exynos_drm_component_del(dev, EXYNOS_DEVICE_TYPE_CRTC); return ret; } -- cgit v0.10.2 From 400c8ac8fc78f51dc7eb187dcf2320fb616c8215 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:19 +0100 Subject: drm/exynos/fimd: stop using manager->ctx pointer The patch replaces accesses to manager->ctx pointer by container_of construct. As fimd was the last user of ctx the patch removes this field as well. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index 262a459..f408e49 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -211,7 +211,6 @@ struct exynos_drm_manager { struct drm_crtc *crtc; int pipe; struct exynos_drm_manager_ops *ops; - void *ctx; }; struct exynos_drm_g2d_private { diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index 9a5e999..b944661 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -185,6 +185,11 @@ struct fimd_context { struct exynos_drm_display *display; }; +static inline struct fimd_context *mgr_to_fimd(struct exynos_drm_manager *mgr) +{ + return container_of(mgr, struct fimd_context, manager); +} + static const struct of_device_id fimd_driver_dt_match[] = { { .compatible = "samsung,s3c6400-fimd", .data = &s3c64xx_fimd_driver_data }, @@ -211,7 +216,7 @@ static inline struct fimd_driver_data *drm_fimd_get_driver_data( static void fimd_wait_for_vblank(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); if (ctx->suspended) return; @@ -256,7 +261,7 @@ static void fimd_enable_shadow_channel_path(struct fimd_context *ctx, int win, static void fimd_clear_channel(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); int win, ch_enabled = 0; DRM_DEBUG_KMS("%s\n", __FILE__); @@ -289,7 +294,7 @@ static void fimd_clear_channel(struct exynos_drm_manager *mgr) static int fimd_mgr_initialize(struct exynos_drm_manager *mgr, struct drm_device *drm_dev) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); struct exynos_drm_private *priv; priv = drm_dev->dev_private; @@ -311,7 +316,7 @@ static int fimd_mgr_initialize(struct exynos_drm_manager *mgr, static void fimd_mgr_remove(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); /* detach this sub driver from iommu mapping if supported. */ if (is_drm_iommu_supported(ctx->drm_dev)) @@ -351,14 +356,14 @@ static bool fimd_mode_fixup(struct exynos_drm_manager *mgr, static void fimd_mode_set(struct exynos_drm_manager *mgr, const struct drm_display_mode *in_mode) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); drm_mode_copy(&ctx->mode, in_mode); } static void fimd_commit(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); struct drm_display_mode *mode = &ctx->mode; struct fimd_driver_data *driver_data = ctx->driver_data; void *timing_base = ctx->regs + driver_data->timing_base; @@ -458,7 +463,7 @@ static void fimd_commit(struct exynos_drm_manager *mgr) static int fimd_enable_vblank(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); u32 val; if (ctx->suspended) @@ -490,7 +495,7 @@ static int fimd_enable_vblank(struct exynos_drm_manager *mgr) static void fimd_disable_vblank(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); u32 val; if (ctx->suspended) @@ -515,7 +520,7 @@ static void fimd_disable_vblank(struct exynos_drm_manager *mgr) static void fimd_win_mode_set(struct exynos_drm_manager *mgr, struct exynos_drm_overlay *overlay) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); struct fimd_win_data *win_data; int win; unsigned long offset; @@ -673,7 +678,7 @@ static void fimd_shadow_protect_win(struct fimd_context *ctx, static void fimd_win_commit(struct exynos_drm_manager *mgr, int zpos) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); struct fimd_win_data *win_data; int win = zpos; unsigned long val, alpha, size; @@ -796,7 +801,7 @@ static void fimd_win_commit(struct exynos_drm_manager *mgr, int zpos) static void fimd_win_disable(struct exynos_drm_manager *mgr, int zpos) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); struct fimd_win_data *win_data; int win = zpos; @@ -830,7 +835,7 @@ static void fimd_win_disable(struct exynos_drm_manager *mgr, int zpos) static void fimd_window_suspend(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); struct fimd_win_data *win_data; int i; @@ -844,7 +849,7 @@ static void fimd_window_suspend(struct exynos_drm_manager *mgr) static void fimd_window_resume(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); struct fimd_win_data *win_data; int i; @@ -857,7 +862,7 @@ static void fimd_window_resume(struct exynos_drm_manager *mgr) static void fimd_apply(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); struct fimd_win_data *win_data; int i; @@ -874,7 +879,7 @@ static void fimd_apply(struct exynos_drm_manager *mgr) static int fimd_poweron(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); int ret; if (!ctx->suspended) @@ -922,7 +927,7 @@ bus_clk_err: static int fimd_poweroff(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); if (ctx->suspended) return 0; @@ -993,7 +998,7 @@ static void fimd_trigger(struct device *dev) static void fimd_te_handler(struct exynos_drm_manager *mgr) { - struct fimd_context *ctx = mgr->ctx; + struct fimd_context *ctx = mgr_to_fimd(mgr); /* Checks the crtc is detached already from encoder */ if (ctx->pipe < 0 || !ctx->drm_dev) @@ -1206,7 +1211,6 @@ static int fimd_probe(struct platform_device *pdev) init_waitqueue_head(&ctx->wait_vsync_queue); atomic_set(&ctx->wait_vsync_event, 0); - ctx->manager.ctx = ctx; platform_set_drvdata(pdev, ctx); -- cgit v0.10.2 From 930865fbe290411a84da0617a3927e78ba939311 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:20 +0100 Subject: drm/exynos/hdmi: embed display into private context exynos_drm_display is used by internal Exynos DRM framework for representing encoder:connector pair. As it should be mapped 1:1 to hdmi private context it seems more reasonable to embed it directly in that context. As a result further code simplification will be possible. Moreover it will be possible to handle multiple hdmi devices in the system. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_hdmi.c b/drivers/gpu/drm/exynos/exynos_hdmi.c index 563a19e..2851f89 100644 --- a/drivers/gpu/drm/exynos/exynos_hdmi.c +++ b/drivers/gpu/drm/exynos/exynos_hdmi.c @@ -49,7 +49,6 @@ #include #include -#define get_hdmi_display(dev) platform_get_drvdata(to_platform_device(dev)) #define ctx_from_connector(c) container_of(c, struct hdmi_context, connector) #define HOTPLUG_DEBOUNCE_MS 1100 @@ -182,6 +181,7 @@ struct hdmi_conf_regs { }; struct hdmi_context { + struct exynos_drm_display display; struct device *dev; struct drm_device *drm_dev; struct drm_connector connector; @@ -2143,11 +2143,6 @@ static struct exynos_drm_display_ops hdmi_display_ops = { .commit = hdmi_commit, }; -static struct exynos_drm_display hdmi_display = { - .type = EXYNOS_DISPLAY_TYPE_HDMI, - .ops = &hdmi_display_ops, -}; - static void hdmi_hotplug_work_func(struct work_struct *work) { struct hdmi_context *hdata; @@ -2302,12 +2297,11 @@ MODULE_DEVICE_TABLE (of, hdmi_match_types); static int hdmi_bind(struct device *dev, struct device *master, void *data) { struct drm_device *drm_dev = data; - struct hdmi_context *hdata; + struct hdmi_context *hdata = dev_get_drvdata(dev); - hdata = hdmi_display.ctx; hdata->drm_dev = drm_dev; - return exynos_drm_create_enc_conn(drm_dev, &hdmi_display); + return exynos_drm_create_enc_conn(drm_dev, &hdata->display); } static void hdmi_unbind(struct device *dev, struct device *master, void *data) @@ -2349,31 +2343,28 @@ static int hdmi_probe(struct platform_device *pdev) struct resource *res; int ret; - ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR, - hdmi_display.type); - if (ret) - return ret; - - if (!dev->of_node) { - ret = -ENODEV; - goto err_del_component; - } + if (!dev->of_node) + return -ENODEV; pdata = drm_hdmi_dt_parse_pdata(dev); - if (!pdata) { - ret = -EINVAL; - goto err_del_component; - } + if (!pdata) + return -EINVAL; hdata = devm_kzalloc(dev, sizeof(struct hdmi_context), GFP_KERNEL); - if (!hdata) { - ret = -ENOMEM; - goto err_del_component; - } + if (!hdata) + return -ENOMEM; + + hdata->display.type = EXYNOS_DISPLAY_TYPE_HDMI; + hdata->display.ops = &hdmi_display_ops; + + ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR, + hdata->display.type); + if (ret) + return ret; mutex_init(&hdata->hdmi_mutex); - platform_set_drvdata(pdev, &hdmi_display); + platform_set_drvdata(pdev, hdata); match = of_match_node(hdmi_match_types, dev->of_node); if (!match) { @@ -2485,7 +2476,7 @@ out_get_phy_port: } pm_runtime_enable(dev); - hdmi_display.ctx = hdata; + hdata->display.ctx = hdata; ret = component_add(&pdev->dev, &hdmi_component_ops); if (ret) @@ -2510,7 +2501,7 @@ err_del_component: static int hdmi_remove(struct platform_device *pdev) { - struct hdmi_context *hdata = hdmi_display.ctx; + struct hdmi_context *hdata = platform_get_drvdata(pdev); cancel_delayed_work_sync(&hdata->hotplug_work); -- cgit v0.10.2 From 0d8424f83c31f1839826ea17ffc3ddf3f1b332a4 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:21 +0100 Subject: drm/exynos/hdmi: stop using display->ctx pointer The patch replaces accesses to display->ctx pointer by container_of construct. It will allow to remove ctx field in the future. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_hdmi.c b/drivers/gpu/drm/exynos/exynos_hdmi.c index 2851f89..5765a16 100644 --- a/drivers/gpu/drm/exynos/exynos_hdmi.c +++ b/drivers/gpu/drm/exynos/exynos_hdmi.c @@ -213,6 +213,11 @@ struct hdmi_context { enum hdmi_type type; }; +static inline struct hdmi_context *display_to_hdmi(struct exynos_drm_display *d) +{ + return container_of(d, struct hdmi_context, display); +} + struct hdmiphy_config { int pixel_clock; u8 conf[32]; @@ -1123,7 +1128,7 @@ static struct drm_connector_helper_funcs hdmi_connector_helper_funcs = { static int hdmi_create_connector(struct exynos_drm_display *display, struct drm_encoder *encoder) { - struct hdmi_context *hdata = display->ctx; + struct hdmi_context *hdata = display_to_hdmi(display); struct drm_connector *connector = &hdata->connector; int ret; @@ -2000,7 +2005,7 @@ static void hdmi_v14_mode_set(struct hdmi_context *hdata, static void hdmi_mode_set(struct exynos_drm_display *display, struct drm_display_mode *mode) { - struct hdmi_context *hdata = display->ctx; + struct hdmi_context *hdata = display_to_hdmi(display); struct drm_display_mode *m = mode; DRM_DEBUG_KMS("xres=%d, yres=%d, refresh=%d, intl=%s\n", @@ -2019,7 +2024,7 @@ static void hdmi_mode_set(struct exynos_drm_display *display, static void hdmi_commit(struct exynos_drm_display *display) { - struct hdmi_context *hdata = display->ctx; + struct hdmi_context *hdata = display_to_hdmi(display); mutex_lock(&hdata->hdmi_mutex); if (!hdata->powered) { @@ -2033,7 +2038,7 @@ static void hdmi_commit(struct exynos_drm_display *display) static void hdmi_poweron(struct exynos_drm_display *display) { - struct hdmi_context *hdata = display->ctx; + struct hdmi_context *hdata = display_to_hdmi(display); struct hdmi_resources *res = &hdata->res; mutex_lock(&hdata->hdmi_mutex); @@ -2064,7 +2069,7 @@ static void hdmi_poweron(struct exynos_drm_display *display) static void hdmi_poweroff(struct exynos_drm_display *display) { - struct hdmi_context *hdata = display->ctx; + struct hdmi_context *hdata = display_to_hdmi(display); struct hdmi_resources *res = &hdata->res; mutex_lock(&hdata->hdmi_mutex); @@ -2099,7 +2104,7 @@ out: static void hdmi_dpms(struct exynos_drm_display *display, int mode) { - struct hdmi_context *hdata = display->ctx; + struct hdmi_context *hdata = display_to_hdmi(display); struct drm_encoder *encoder = hdata->encoder; struct drm_crtc *crtc = encoder->crtc; struct drm_crtc_helper_funcs *funcs = NULL; @@ -2476,7 +2481,6 @@ out_get_phy_port: } pm_runtime_enable(dev); - hdata->display.ctx = hdata; ret = component_add(&pdev->dev, &hdmi_component_ops); if (ret) -- cgit v0.10.2 From 7340426affacb4b5988f9cf1c3dbfc28e9679360 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:22 +0100 Subject: drm/exynos/vidi: embed display into private context exynos_drm_display is used by internal Exynos DRM framework for representing encoder:connector pair. As it should be mapped 1:1 to vidi private context it seems more reasonable to embed it directly in that context. As a result further code simplification will be possible. Moreover it will be possible to handle multiple vidi devices in the system. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_vidi.c b/drivers/gpu/drm/exynos/exynos_drm_vidi.c index f048a90..f58dd52 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_vidi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_vidi.c @@ -47,6 +47,7 @@ struct vidi_win_data { struct vidi_context { struct exynos_drm_manager manager; + struct exynos_drm_display display; struct drm_device *drm_dev; struct drm_crtc *crtc; struct drm_encoder *encoder; @@ -554,11 +555,6 @@ static struct exynos_drm_display_ops vidi_display_ops = { .create_connector = vidi_create_connector, }; -static struct exynos_drm_display vidi_display = { - .type = EXYNOS_DISPLAY_TYPE_VIDI, - .ops = &vidi_display_ops, -}; - static int vidi_subdrv_probe(struct drm_device *drm_dev, struct device *dev) { struct vidi_context *ctx = dev_get_drvdata(dev); @@ -573,7 +569,7 @@ static int vidi_subdrv_probe(struct drm_device *drm_dev, struct device *dev) return ret; } - ret = exynos_drm_create_enc_conn(drm_dev, &vidi_display); + ret = exynos_drm_create_enc_conn(drm_dev, &ctx->display); if (ret) { crtc->funcs->destroy(crtc); DRM_ERROR("failed to create encoder and connector.\n"); @@ -595,11 +591,13 @@ static int vidi_probe(struct platform_device *pdev) ctx->manager.type = EXYNOS_DISPLAY_TYPE_VIDI; ctx->manager.ops = &vidi_manager_ops; + ctx->display.type = EXYNOS_DISPLAY_TYPE_VIDI; + ctx->display.ops = &vidi_display_ops; ctx->default_win = 0; INIT_WORK(&ctx->work, vidi_fake_vblank_handler); - vidi_display.ctx = ctx; + ctx->display.ctx = ctx; mutex_init(&ctx->lock); -- cgit v0.10.2 From 2f26bd7227b80d002dfcd3f60e71bfad168e6517 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:23 +0100 Subject: drm/exynos/vidi: stop using display->ctx pointer The patch replaces accesses to display->ctx pointer by container_of construct. It will allow to remove ctx field in the future. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_vidi.c b/drivers/gpu/drm/exynos/exynos_drm_vidi.c index f58dd52..3b6fdd6 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_vidi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_vidi.c @@ -72,6 +72,11 @@ static inline struct vidi_context *manager_to_vidi(struct exynos_drm_manager *m) return container_of(m, struct vidi_context, manager); } +static inline struct vidi_context *display_to_vidi(struct exynos_drm_display *d) +{ + return container_of(d, struct vidi_context, display); +} + static const char fake_edid_info[] = { 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00, 0x4c, 0x2d, 0x05, 0x05, 0x00, 0x00, 0x00, 0x00, 0x30, 0x12, 0x01, 0x03, 0x80, 0x10, 0x09, 0x78, @@ -419,7 +424,7 @@ int vidi_connection_ioctl(struct drm_device *drm_dev, void *data, display = exynos_drm_get_display(encoder); if (display->type == EXYNOS_DISPLAY_TYPE_VIDI) { - ctx = display->ctx; + ctx = display_to_vidi(display); break; } } @@ -529,7 +534,7 @@ static struct drm_connector_helper_funcs vidi_connector_helper_funcs = { static int vidi_create_connector(struct exynos_drm_display *display, struct drm_encoder *encoder) { - struct vidi_context *ctx = display->ctx; + struct vidi_context *ctx = display_to_vidi(display); struct drm_connector *connector = &ctx->connector; int ret; @@ -597,8 +602,6 @@ static int vidi_probe(struct platform_device *pdev) INIT_WORK(&ctx->work, vidi_fake_vblank_handler); - ctx->display.ctx = ctx; - mutex_init(&ctx->lock); platform_set_drvdata(pdev, ctx); -- cgit v0.10.2 From 1df6e5fb79f614141f4358587b18695d7acda024 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:24 +0100 Subject: drm/exynos/dp: embed display into private context exynos_drm_display is used by internal Exynos DRM framework for representing encoder:connector pair. As it should be mapped 1:1 to dp private context it seems more reasonable to embed it directly in that context. As a result further code simplification will be possible. Moreover it will be possible to handle multiple dp devices in the system. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_dp_core.c b/drivers/gpu/drm/exynos/exynos_dp_core.c index ed818b9..47c6e47 100644 --- a/drivers/gpu/drm/exynos/exynos_dp_core.c +++ b/drivers/gpu/drm/exynos/exynos_dp_core.c @@ -30,7 +30,6 @@ #include #include -#include "exynos_drm_drv.h" #include "exynos_dp_core.h" #define ctx_from_connector(c) container_of(c, struct exynos_dp_device, \ @@ -1133,11 +1132,6 @@ static struct exynos_drm_display_ops exynos_dp_display_ops = { .commit = exynos_dp_commit, }; -static struct exynos_drm_display exynos_dp_display = { - .type = EXYNOS_DISPLAY_TYPE_LCD, - .ops = &exynos_dp_display_ops, -}; - static struct video_info *exynos_dp_dt_parse_pdata(struct device *dev) { struct device_node *dp_node = dev->of_node; @@ -1211,10 +1205,10 @@ static int exynos_dp_dt_parse_panel(struct exynos_dp_device *dp) static int exynos_dp_bind(struct device *dev, struct device *master, void *data) { + struct exynos_dp_device *dp = dev_get_drvdata(dev); struct platform_device *pdev = to_platform_device(dev); struct drm_device *drm_dev = data; struct resource *res; - struct exynos_dp_device *dp = exynos_dp_display.ctx; unsigned int irq_flags; int ret = 0; @@ -1306,17 +1300,15 @@ static int exynos_dp_bind(struct device *dev, struct device *master, void *data) dp->drm_dev = drm_dev; - platform_set_drvdata(pdev, &exynos_dp_display); - - return exynos_drm_create_enc_conn(drm_dev, &exynos_dp_display); + return exynos_drm_create_enc_conn(drm_dev, &dp->display); } static void exynos_dp_unbind(struct device *dev, struct device *master, void *data) { - struct exynos_drm_display *display = dev_get_drvdata(dev); + struct exynos_dp_device *dp = dev_get_drvdata(dev); - exynos_dp_dpms(display, DRM_MODE_DPMS_OFF); + exynos_dp_dpms(&dp->display, DRM_MODE_DPMS_OFF); } static const struct component_ops exynos_dp_ops = { @@ -1331,16 +1323,20 @@ static int exynos_dp_probe(struct platform_device *pdev) struct exynos_dp_device *dp; int ret; - ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR, - exynos_dp_display.type); - if (ret) - return ret; - dp = devm_kzalloc(&pdev->dev, sizeof(struct exynos_dp_device), GFP_KERNEL); if (!dp) return -ENOMEM; + dp->display.type = EXYNOS_DISPLAY_TYPE_LCD; + dp->display.ops = &exynos_dp_display_ops; + platform_set_drvdata(pdev, dp); + + ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR, + dp->display.type); + if (ret) + return ret; + panel_node = of_parse_phandle(dev->of_node, "panel", 0); if (panel_node) { dp->panel = of_drm_find_panel(panel_node); @@ -1349,7 +1345,7 @@ static int exynos_dp_probe(struct platform_device *pdev) return -EPROBE_DEFER; } - exynos_dp_display.ctx = dp; + dp->display.ctx = dp; ret = component_add(&pdev->dev, &exynos_dp_ops); if (ret) @@ -1370,19 +1366,17 @@ static int exynos_dp_remove(struct platform_device *pdev) #ifdef CONFIG_PM_SLEEP static int exynos_dp_suspend(struct device *dev) { - struct platform_device *pdev = to_platform_device(dev); - struct exynos_drm_display *display = platform_get_drvdata(pdev); + struct exynos_dp_device *dp = dev_get_drvdata(dev); - exynos_dp_dpms(display, DRM_MODE_DPMS_OFF); + exynos_dp_dpms(&dp->display, DRM_MODE_DPMS_OFF); return 0; } static int exynos_dp_resume(struct device *dev) { - struct platform_device *pdev = to_platform_device(dev); - struct exynos_drm_display *display = platform_get_drvdata(pdev); + struct exynos_dp_device *dp = dev_get_drvdata(dev); - exynos_dp_dpms(display, DRM_MODE_DPMS_ON); + exynos_dp_dpms(&dp->display, DRM_MODE_DPMS_ON); return 0; } #endif diff --git a/drivers/gpu/drm/exynos/exynos_dp_core.h b/drivers/gpu/drm/exynos/exynos_dp_core.h index 6426201..164f171 100644 --- a/drivers/gpu/drm/exynos/exynos_dp_core.h +++ b/drivers/gpu/drm/exynos/exynos_dp_core.h @@ -17,6 +17,8 @@ #include #include +#include "exynos_drm_drv.h" + #define DP_TIMEOUT_LOOP_COUNT 100 #define MAX_CR_LOOP 5 #define MAX_EQ_LOOP 5 @@ -145,6 +147,7 @@ struct link_train { }; struct exynos_dp_device { + struct exynos_drm_display display; struct device *dev; struct drm_device *drm_dev; struct drm_connector connector; -- cgit v0.10.2 From 63b3be327048402a39068f188726e3729e061fda Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:25 +0100 Subject: drm/exynos/dp: stop using display->ctx pointer The patch replaces accesses to display->ctx pointer by container_of construct. It will allow to remove ctx field in the future. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_dp_core.c b/drivers/gpu/drm/exynos/exynos_dp_core.c index 47c6e47..34d46aa 100644 --- a/drivers/gpu/drm/exynos/exynos_dp_core.c +++ b/drivers/gpu/drm/exynos/exynos_dp_core.c @@ -35,6 +35,12 @@ #define ctx_from_connector(c) container_of(c, struct exynos_dp_device, \ connector) +static inline struct exynos_dp_device * +display_to_dp(struct exynos_drm_display *d) +{ + return container_of(d, struct exynos_dp_device, display); +} + struct bridge_init { struct i2c_client *client; struct device_node *node; @@ -881,7 +887,7 @@ static void exynos_dp_hotplug(struct work_struct *work) static void exynos_dp_commit(struct exynos_drm_display *display) { - struct exynos_dp_device *dp = display->ctx; + struct exynos_dp_device *dp = display_to_dp(display); int ret; /* Keep the panel disabled while we configure video */ @@ -1019,7 +1025,7 @@ static int exynos_drm_attach_lcd_bridge(struct drm_device *dev, static int exynos_dp_create_connector(struct exynos_drm_display *display, struct drm_encoder *encoder) { - struct exynos_dp_device *dp = display->ctx; + struct exynos_dp_device *dp = display_to_dp(display); struct drm_connector *connector = &dp->connector; int ret; @@ -1063,7 +1069,7 @@ static void exynos_dp_phy_exit(struct exynos_dp_device *dp) static void exynos_dp_poweron(struct exynos_drm_display *display) { - struct exynos_dp_device *dp = display->ctx; + struct exynos_dp_device *dp = display_to_dp(display); if (dp->dpms_mode == DRM_MODE_DPMS_ON) return; @@ -1084,7 +1090,7 @@ static void exynos_dp_poweron(struct exynos_drm_display *display) static void exynos_dp_poweroff(struct exynos_drm_display *display) { - struct exynos_dp_device *dp = display->ctx; + struct exynos_dp_device *dp = display_to_dp(display); if (dp->dpms_mode != DRM_MODE_DPMS_ON) return; @@ -1109,7 +1115,7 @@ static void exynos_dp_poweroff(struct exynos_drm_display *display) static void exynos_dp_dpms(struct exynos_drm_display *display, int mode) { - struct exynos_dp_device *dp = display->ctx; + struct exynos_dp_device *dp = display_to_dp(display); switch (mode) { case DRM_MODE_DPMS_ON: @@ -1345,8 +1351,6 @@ static int exynos_dp_probe(struct platform_device *pdev) return -EPROBE_DEFER; } - dp->display.ctx = dp; - ret = component_add(&pdev->dev, &exynos_dp_ops); if (ret) exynos_drm_component_del(&pdev->dev, -- cgit v0.10.2 From 4cfde1f2af076547ebe86f8632e65119f28438b8 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:26 +0100 Subject: drm/exynos/dpi: embed display into private context exynos_drm_display is used by internal Exynos DRM framework for representing encoder:connector pair. As it should be mapped 1:1 to dpi private context it seems more reasonable to embed it directly in that context. As a result further code simplification will be possible. Moreover it will be possible to handle multiple dpi devices in the system. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dpi.c b/drivers/gpu/drm/exynos/exynos_drm_dpi.c index 3dc678e..3acfc28 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dpi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dpi.c @@ -22,6 +22,7 @@ #include "exynos_drm_drv.h" struct exynos_dpi { + struct exynos_drm_display display; struct device *dev; struct device_node *panel_node; @@ -35,6 +36,11 @@ struct exynos_dpi { #define connector_to_dpi(c) container_of(c, struct exynos_dpi, connector) +static inline struct exynos_dpi *display_to_dpi(struct exynos_drm_display *d) +{ + return container_of(d, struct exynos_dpi, display); +} + static enum drm_connector_status exynos_dpi_detect(struct drm_connector *connector, bool force) { @@ -165,11 +171,6 @@ static struct exynos_drm_display_ops exynos_dpi_display_ops = { .dpms = exynos_dpi_dpms }; -static struct exynos_drm_display exynos_dpi_display = { - .type = EXYNOS_DISPLAY_TYPE_LCD, - .ops = &exynos_dpi_display_ops, -}; - /* of_* functions will be removed after merge of of_graph patches */ static struct device_node * of_get_child_by_name_reg(struct device_node *parent, const char *name, u32 reg) @@ -299,20 +300,22 @@ struct exynos_drm_display *exynos_dpi_probe(struct device *dev) struct exynos_dpi *ctx; int ret; - ret = exynos_drm_component_add(dev, - EXYNOS_DEVICE_TYPE_CONNECTOR, - exynos_dpi_display.type); - if (ret) - return ERR_PTR(ret); - ctx = devm_kzalloc(dev, sizeof(*ctx), GFP_KERNEL); if (!ctx) - goto err_del_component; + return ERR_PTR(-ENOMEM); + ctx->display.type = EXYNOS_DISPLAY_TYPE_LCD; + ctx->display.ops = &exynos_dpi_display_ops; ctx->dev = dev; - exynos_dpi_display.ctx = ctx; + ctx->display.ctx = ctx; ctx->dpms_mode = DRM_MODE_DPMS_OFF; + ret = exynos_drm_component_add(dev, + EXYNOS_DEVICE_TYPE_CONNECTOR, + ctx->display.type); + if (ret) + return ERR_PTR(ret); + ret = exynos_dpi_parse_dt(ctx); if (ret < 0) { devm_kfree(dev, ctx); @@ -328,7 +331,7 @@ struct exynos_drm_display *exynos_dpi_probe(struct device *dev) } } - return &exynos_dpi_display; + return &ctx->display; err_del_component: exynos_drm_component_del(dev, EXYNOS_DEVICE_TYPE_CONNECTOR); @@ -336,16 +339,16 @@ err_del_component: return NULL; } -int exynos_dpi_remove(struct device *dev) +int exynos_dpi_remove(struct exynos_drm_display *display) { - struct exynos_dpi *ctx = exynos_dpi_display.ctx; + struct exynos_dpi *ctx = display_to_dpi(display); - exynos_dpi_dpms(&exynos_dpi_display, DRM_MODE_DPMS_OFF); + exynos_dpi_dpms(&ctx->display, DRM_MODE_DPMS_OFF); if (ctx->panel) drm_panel_detach(ctx->panel); - exynos_drm_component_del(dev, EXYNOS_DEVICE_TYPE_CONNECTOR); + exynos_drm_component_del(ctx->dev, EXYNOS_DEVICE_TYPE_CONNECTOR); return 0; } diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index f408e49..b023f5f 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -304,7 +304,7 @@ static inline void exynos_platform_device_ipp_unregister(void) {} #ifdef CONFIG_DRM_EXYNOS_DPI struct exynos_drm_display * exynos_dpi_probe(struct device *dev); -int exynos_dpi_remove(struct device *dev); +int exynos_dpi_remove(struct exynos_drm_display *display); #else static inline struct exynos_drm_display * exynos_dpi_probe(struct device *dev) { return NULL; } diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index b944661..ef80a35 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -1092,7 +1092,7 @@ static void fimd_unbind(struct device *dev, struct device *master, fimd_dpms(&ctx->manager, DRM_MODE_DPMS_OFF); if (ctx->display) - exynos_dpi_remove(dev); + exynos_dpi_remove(ctx->display); fimd_mgr_remove(&ctx->manager); } -- cgit v0.10.2 From 5af3d9bb78f130819204ef44ac9d2b635e49fc35 Mon Sep 17 00:00:00 2001 From: Andrzej Hajda Date: Mon, 17 Nov 2014 09:54:27 +0100 Subject: drm/exynos/dpi: stop using display->ctx pointer The patch replaces accesses to display->ctx pointer by container_of construct. The field is removed as well as dpi was the last user of it. Signed-off-by: Andrzej Hajda Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_dpi.c b/drivers/gpu/drm/exynos/exynos_drm_dpi.c index 3acfc28..37678cf 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dpi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dpi.c @@ -106,7 +106,7 @@ static struct drm_connector_helper_funcs exynos_dpi_connector_helper_funcs = { static int exynos_dpi_create_connector(struct exynos_drm_display *display, struct drm_encoder *encoder) { - struct exynos_dpi *ctx = display->ctx; + struct exynos_dpi *ctx = display_to_dpi(display); struct drm_connector *connector = &ctx->connector; int ret; @@ -147,7 +147,7 @@ static void exynos_dpi_poweroff(struct exynos_dpi *ctx) static void exynos_dpi_dpms(struct exynos_drm_display *display, int mode) { - struct exynos_dpi *ctx = display->ctx; + struct exynos_dpi *ctx = display_to_dpi(display); switch (mode) { case DRM_MODE_DPMS_ON: @@ -307,7 +307,6 @@ struct exynos_drm_display *exynos_dpi_probe(struct device *dev) ctx->display.type = EXYNOS_DISPLAY_TYPE_LCD; ctx->display.ops = &exynos_dpi_display_ops; ctx->dev = dev; - ctx->display.ctx = ctx; ctx->dpms_mode = DRM_MODE_DPMS_OFF; ret = exynos_drm_component_add(dev, diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index b023f5f..2e4e91b 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -152,7 +152,6 @@ struct exynos_drm_display { struct drm_encoder *encoder; struct drm_connector *connector; struct exynos_drm_display_ops *ops; - void *ctx; }; /* -- cgit v0.10.2 From b67139571ec0956eb94a354572df73cf24d4521c Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Thu, 20 Nov 2014 21:42:55 -0200 Subject: Revert "drm/exynos: fix null pointer dereference issue" This reverts commit cea24824ab432f8acabb254d6805e9aa756de6af. Moving subdriver probe to exynos_drm_platform_probe() was making exynos_drm_device_subdrv_probe() fail because the platform data wasn't set yet. It only gets set in exynos_drm_load. We need to find a smarter way to fix this issue. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index eab12f0..2ca0c5d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -108,6 +108,11 @@ static int exynos_drm_load(struct drm_device *dev, unsigned long flags) if (ret) goto err_unbind_all; + /* Probe non kms sub drivers and virtual display driver. */ + ret = exynos_drm_device_subdrv_probe(dev); + if (ret) + goto err_cleanup_vblank; + /* * enable drm irq mode. * - with irq_enabled = true, we can use the vblank feature. @@ -133,6 +138,8 @@ static int exynos_drm_load(struct drm_device *dev, unsigned long flags) return 0; +err_cleanup_vblank: + drm_vblank_cleanup(dev); err_unbind_all: component_unbind_all(dev->dev, dev); err_mode_config_cleanup: @@ -146,6 +153,8 @@ err_free_private: static int exynos_drm_unload(struct drm_device *dev) { + exynos_drm_device_subdrv_remove(dev); + exynos_drm_fbdev_fini(dev); drm_kms_helper_poll_fini(dev); @@ -614,14 +623,8 @@ static int exynos_drm_platform_probe(struct platform_device *pdev) if (ret < 0) goto err_unregister_non_kms_drivers; - /* Probe non kms sub drivers and virtual display driver. */ - ret = exynos_drm_device_subdrv_probe(platform_get_drvdata(pdev)); - if (ret) - goto err_unregister_resources; - return ret; -err_unregister_resources: #ifdef CONFIG_DRM_EXYNOS_IPP exynos_platform_device_ipp_unregister(); #endif @@ -643,8 +646,6 @@ static int exynos_drm_platform_remove(struct platform_device *pdev) { int i; - exynos_drm_device_subdrv_remove(platform_get_drvdata(pdev)); - #ifdef CONFIG_DRM_EXYNOS_IPP exynos_platform_device_ipp_unregister(); #endif -- cgit v0.10.2 From 820687befec471aff3bb59bd69302d34a776e807 Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Mon, 24 Nov 2014 16:37:26 +0900 Subject: drm/exynos: move Exynos platform drivers registration to init Registering the Exynos DRM subdevices platform drivers in the probe function is causing an infinite loop. Fix this by moving it to the exynos_drm_init() function to register the drivers on module init. Registering drivers in the probe functions causes a deadlock in the parent device lock. See Grant Likely explanation on the topic: "I think the problem is that exynos_drm_init() is registering a normal (non-OF) platform device, so the parent will be /sys/devices/platform. It immediately gets bound against exynos_drm_platform_driver which calls the exynos drm_platform_probe() hook. The driver core obtains device_lock() on the device *and on the device parent*. Inside the probe hook, additional platform_drivers get registered. Each time one does, it tries to bind against every platform device in the system, which includes the ones created by OF. When it attempts to bind, it obtains device_lock() on the device *and on the device parent*. Before the change to move of-generated platform devices into /sys/devices/platform, the devices had different parents. Now both devices have /sys/devices/platform as the parent, so yes they are going to deadlock. The real problem is registering drivers from within a probe hook. That is completely wrong for the above deadlock reason. __driver_attach() will deadlock. Those registrations must be pulled out of .probe(). Registering devices in .probe() is okay because __device_attach() doesn't try to obtain device_lock() on the parent." INFO: task swapper/0:1 blocked for more than 120 seconds. Not tainted 3.18.0-rc3-next-20141105 #794 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. swapper/0 D c052534c 0 1 0 0x00000000 [] (__schedule) from [] (schedule_preempt_disabled+0x14/0x20) [] (schedule_preempt_disabled) from [] (mutex_lock_nested+0x1c4/0x464 [] (mutex_lock_nested) from [] (__driver_attach+0x48/0x98) [] (__driver_attach) from [] (bus_for_each_dev+0x54/0x88) [] (bus_for_each_dev) from [] (bus_add_driver+0xe4/0x200) [] (bus_add_driver) from [] (driver_register+0x78/0xf4) [] (driver_register) from [] (exynos_drm_platform_probe+0x34/0x234) [] (exynos_drm_platform_probe) from [] (platform_drv_probe+0x48/0xa4) [] (platform_drv_probe) from [] (driver_probe_device+0x13c/0x37c) [] (driver_probe_device) from [] (__driver_attach+0x94/0x98) [] (__driver_attach) from [] (bus_for_each_dev+0x54/0x88) [] (bus_for_each_dev) from [] (bus_add_driver+0xe4/0x200) [] (bus_add_driver) from [] (driver_register+0x78/0xf4) [] (driver_register) from [] (exynos_drm_init+0x70/0xa0) [] (exynos_drm_init) from [] (do_one_initcall+0xac/0x1f0) [] (do_one_initcall) from [] (kernel_init_freeable+0x10c/0x1d8) [] (kernel_init_freeable) from [] (kernel_init+0x8/0xec) [] (kernel_init) from [] (ret_from_fork+0x14/0x2c) 3 locks held by swapper/0/1: #0: (&dev->mutex){......}, at: [] __driver_attach+0x48/0x98 #1: (&dev->mutex){......}, at: [] __driver_attach+0x58/0x98 #2: (&dev->mutex){......}, at: [] __driver_attach+0x48/0x98 Changelog v2: - call platform_driver_register after all kms and non kms drivers are registered - rebased it to exynos-drm-next Signed-off-by: Javier Martinez Canillas Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index 2ca0c5d..c5cb8b6 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -591,73 +591,22 @@ static struct platform_driver *const exynos_drm_non_kms_drivers[] = { static int exynos_drm_platform_probe(struct platform_device *pdev) { struct component_match *match; - int ret, i, j; pdev->dev.coherent_dma_mask = DMA_BIT_MASK(32); exynos_drm_driver.num_ioctls = ARRAY_SIZE(exynos_ioctls); - for (i = 0; i < ARRAY_SIZE(exynos_drm_kms_drivers); ++i) { - ret = platform_driver_register(exynos_drm_kms_drivers[i]); - if (ret < 0) - goto err_unregister_kms_drivers; - } - match = exynos_drm_match_add(&pdev->dev); if (IS_ERR(match)) { - ret = PTR_ERR(match); - goto err_unregister_kms_drivers; + return PTR_ERR(match); } - ret = component_master_add_with_match(&pdev->dev, &exynos_drm_ops, - match); - if (ret < 0) - goto err_unregister_kms_drivers; - - for (j = 0; j < ARRAY_SIZE(exynos_drm_non_kms_drivers); ++j) { - ret = platform_driver_register(exynos_drm_non_kms_drivers[j]); - if (ret < 0) - goto err_del_component_master; - } - - ret = exynos_platform_device_ipp_register(); - if (ret < 0) - goto err_unregister_non_kms_drivers; - - return ret; - -#ifdef CONFIG_DRM_EXYNOS_IPP - exynos_platform_device_ipp_unregister(); -#endif -err_unregister_non_kms_drivers: - while (--j >= 0) - platform_driver_unregister(exynos_drm_non_kms_drivers[j]); - -err_del_component_master: - component_master_del(&pdev->dev, &exynos_drm_ops); - -err_unregister_kms_drivers: - while (--i >= 0) - platform_driver_unregister(exynos_drm_kms_drivers[i]); - - return ret; + return component_master_add_with_match(&pdev->dev, &exynos_drm_ops, + match); } static int exynos_drm_platform_remove(struct platform_device *pdev) { - int i; - -#ifdef CONFIG_DRM_EXYNOS_IPP - exynos_platform_device_ipp_unregister(); -#endif - - for (i = ARRAY_SIZE(exynos_drm_non_kms_drivers) - 1; i >= 0; --i) - platform_driver_unregister(exynos_drm_non_kms_drivers[i]); - component_master_del(&pdev->dev, &exynos_drm_ops); - - for (i = ARRAY_SIZE(exynos_drm_kms_drivers) - 1; i >= 0; --i) - platform_driver_unregister(exynos_drm_kms_drivers[i]); - return 0; } @@ -673,7 +622,7 @@ static struct platform_driver exynos_drm_platform_driver = { static int exynos_drm_init(void) { - int ret; + int ret, i, j; /* * Register device object only in case of Exynos SoC. @@ -696,13 +645,43 @@ static int exynos_drm_init(void) if (ret < 0) goto err_unregister_pd; + for (i = 0; i < ARRAY_SIZE(exynos_drm_kms_drivers); ++i) { + ret = platform_driver_register(exynos_drm_kms_drivers[i]); + if (ret < 0) + goto err_unregister_kms_drivers; + } + + for (j = 0; j < ARRAY_SIZE(exynos_drm_non_kms_drivers); ++j) { + ret = platform_driver_register(exynos_drm_non_kms_drivers[j]); + if (ret < 0) + goto err_unregister_non_kms_drivers; + } + +#ifdef CONFIG_DRM_EXYNOS_IPP + ret = exynos_platform_device_ipp_register(); + if (ret < 0) + goto err_unregister_non_kms_drivers; +#endif + ret = platform_driver_register(&exynos_drm_platform_driver); if (ret) - goto err_remove_vidi; + goto err_unregister_resources; return 0; -err_remove_vidi: +err_unregister_resources: +#ifdef CONFIG_DRM_EXYNOS_IPP + exynos_platform_device_ipp_unregister(); +#endif + +err_unregister_non_kms_drivers: + while (--j >= 0) + platform_driver_unregister(exynos_drm_non_kms_drivers[j]); + +err_unregister_kms_drivers: + while (--i >= 0) + platform_driver_unregister(exynos_drm_kms_drivers[i]); + exynos_drm_remove_vidi(); err_unregister_pd: @@ -713,6 +692,18 @@ err_unregister_pd: static void exynos_drm_exit(void) { + int i; + +#ifdef CONFIG_DRM_EXYNOS_IPP + exynos_platform_device_ipp_unregister(); +#endif + + for (i = ARRAY_SIZE(exynos_drm_non_kms_drivers) - 1; i >= 0; --i) + platform_driver_unregister(exynos_drm_non_kms_drivers[i]); + + for (i = ARRAY_SIZE(exynos_drm_kms_drivers) - 1; i >= 0; --i) + platform_driver_unregister(exynos_drm_kms_drivers[i]); + platform_driver_unregister(&exynos_drm_platform_driver); exynos_drm_remove_vidi(); -- cgit v0.10.2 From 4846e452084945891a770809f94b23f33eebcd8c Mon Sep 17 00:00:00 2001 From: Inki Dae Date: Mon, 24 Nov 2014 17:08:00 +0900 Subject: drm/exynos: clean up machine compatible string check Use 'for' statemant instead of hard-coded 'if' statement. Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index c5cb8b6..495826f 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -610,6 +610,12 @@ static int exynos_drm_platform_remove(struct platform_device *pdev) return 0; } +static const char * const strings[] = { + "samsung,exynos3", + "samsung,exynos4", + "samsung,exynos5", +}; + static struct platform_driver exynos_drm_platform_driver = { .probe = exynos_drm_platform_probe, .remove = exynos_drm_platform_remove, @@ -622,6 +628,7 @@ static struct platform_driver exynos_drm_platform_driver = { static int exynos_drm_init(void) { + bool is_exynos = false; int ret, i, j; /* @@ -631,9 +638,14 @@ static int exynos_drm_init(void) * by Exynos drm driver when using multi-platform kernel. * So these codes will be replaced with more generic way later. */ - if (!of_machine_is_compatible("samsung,exynos3") && - !of_machine_is_compatible("samsung,exynos4") && - !of_machine_is_compatible("samsung,exynos5")) + for (i = 0; i < ARRAY_SIZE(strings); i++) { + if (of_machine_is_compatible(strings[i])) { + is_exynos = true; + break; + } + } + + if (!is_exynos) return -ENODEV; exynos_drm_pdev = platform_device_register_simple("exynos-drm", -1, -- cgit v0.10.2 From be19d9336995241f5c98d0abebff440fef03455e Mon Sep 17 00:00:00 2001 From: Julia Lawall Date: Sun, 23 Nov 2014 14:11:15 +0100 Subject: drm/exynos/ipp: fix error return code Propagate the returned error code on failure. A simplified version of the semantic match that finds this problem is as follows: (http://coccinelle.lip6.fr/) // @@ identifier ret; expression e1,e2; @@ ( if (\(ret < 0\|ret != 0\)) { ... return ret; } | ret = 0 ) ... when != ret = e1 when != &ret *if(...) { ... when != ret = e2 when forall return ret; } // Signed-off-by: Julia Lawall Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_ipp.c b/drivers/gpu/drm/exynos/exynos_drm_ipp.c index 00d74b1..d5ad17d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_ipp.c +++ b/drivers/gpu/drm/exynos/exynos_drm_ipp.c @@ -426,18 +426,21 @@ int exynos_drm_ipp_set_property(struct drm_device *drm_dev, void *data, c_node->start_work = ipp_create_cmd_work(); if (IS_ERR(c_node->start_work)) { DRM_ERROR("failed to create start work.\n"); + ret = PTR_ERR(c_node->start_work); goto err_remove_id; } c_node->stop_work = ipp_create_cmd_work(); if (IS_ERR(c_node->stop_work)) { DRM_ERROR("failed to create stop work.\n"); + ret = PTR_ERR(c_node->stop_work); goto err_free_start; } c_node->event_work = ipp_create_event_work(); if (IS_ERR(c_node->event_work)) { DRM_ERROR("failed to create event work.\n"); + ret = PTR_ERR(c_node->event_work); goto err_free_stop; } -- cgit v0.10.2 From 33e2192fb1e329c26847c92edbd90f7e93eecacf Mon Sep 17 00:00:00 2001 From: Inki Dae Date: Mon, 24 Nov 2014 16:58:48 +0900 Subject: drm/exynos: fix exynos_drm_component_del This patch resolves the issue that component object isn't removed correctly. A given component object couldn't be placed to head of drm_component_list so all component objects added to the drm_component_list should be checked to remove the given component object. Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c index 495826f..d71fb54 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c @@ -473,8 +473,6 @@ void exynos_drm_component_del(struct device *dev, list_del(&cdev->list); kfree(cdev); } - - break; } mutex_unlock(&drm_component_lock); -- cgit v0.10.2 From 1d50aa9c6fb01356238ec21039f145fc2d81f7e7 Mon Sep 17 00:00:00 2001 From: Inki Dae Date: Mon, 24 Nov 2014 14:55:41 +0900 Subject: drm/exynos: vidi: add component support This patch adds component support for vidi driver. vidi driver is a kms driver so it doesn't need to be registered to exynos_drm_subdrv_list. For this, it changes for the component framework to be used for vidi driver. This patch fixes below error also, # echo 1 > /sys/devices/platform/exynos-drm-vidi/connection [ 55.618529] ------------[ cut here ]------------ [ 55.621960] WARNING: CPU: 0 PID: 1397 at drivers/gpu/drm/drm_irq.c:1203 exynos_drm_crtc_dpms+0x88/0x17c() [ 55.631268] Modules linked in: [ 55.634278] CPU: 0 PID: 1397 Comm: sh Not tainted 3.18.0-rc2-146253-g31449d7 #1154 [ 55.641885] [] (unwind_backtrace) from [] (show_stack+0x10/0x14) [ 55.649597] [] (show_stack) from [] (dump_stack+0x84/0xc4) [ 55.656802] [] (dump_stack) from [] (warn_slowpath_common+0x6c/0x88) [ 55.664866] [] (warn_slowpath_common) from [] (warn_slowpath_null+0x1c/0x24) [ 55.673632] [] (warn_slowpath_null) from [] (exynos_drm_crtc_dpms+0x88/0x17c) [ 55.682482] [] (exynos_drm_crtc_dpms) from [] (exynos_drm_crtc_commit+0x14/0x44) [ 55.691622] [] (exynos_drm_crtc_commit) from [] (drm_crtc_helper_set_mode+0x3d0/0x51c) [ 55.701233] [] (drm_crtc_helper_set_mode) from [] (drm_crtc_helper_set_config+0x87c/0x9dc) [ 55.711230] [] (drm_crtc_helper_set_config) from [] (drm_mode_set_config_internal+0x58/0xd4) [ 55.721380] [] (drm_mode_set_config_internal) from [] (restore_fbdev_mode+0xcc/0xec) [ 55.730834] [] (restore_fbdev_mode) from [] (drm_fb_helper_restore_fbdev_mode_unlocked+0x1c/0x30) [ 55.741424] [] (drm_fb_helper_restore_fbdev_mode_unlocked) from [] (drm_fb_helper_set_par+0x1c/0x60) [ 55.752271] [] (drm_fb_helper_set_par) from [] (drm_fb_helper_hotplug_event+0x88/0xc4) [ 55.761906] [] (drm_fb_helper_hotplug_event) from [] (drm_helper_hpd_irq_event+0xc8/0x134) [ 55.771898] [] (drm_helper_hpd_irq_event) from [] (vidi_store_connection+0x90/0xc8) [ 55.781268] [] (vidi_store_connection) from [] (kernfs_fop_write+0xc0/0x180) [ 55.790045] [] (kernfs_fop_write) from [] (vfs_write+0xa0/0x1ac) [ 55.797757] [] (vfs_write) from [] (SyS_write+0x44/0x9c) [ 55.804790] [] (SyS_write) from [] (ret_fast_syscall+0x0/0x30) [ 55.812328] ---[ end trace 3c0fe4386702d4dd ]--- This issue occurs when modeset to vidi is tried in case that drm_vblank_init is called prior to crtc creation of vidi driver. In this case, crtc number of vidi is invalid so any requests with the crtc number will fail. This patch guarantees drm_vblank_init to be called after all kms drivers are ready by using component framework. Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_vidi.c b/drivers/gpu/drm/exynos/exynos_drm_vidi.c index 3b6fdd6..45899fb 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_vidi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_vidi.c @@ -14,6 +14,7 @@ #include #include +#include #include @@ -48,11 +49,11 @@ struct vidi_win_data { struct vidi_context { struct exynos_drm_manager manager; struct exynos_drm_display display; + struct platform_device *pdev; struct drm_device *drm_dev; struct drm_crtc *crtc; struct drm_encoder *encoder; struct drm_connector connector; - struct exynos_drm_subdrv subdrv; struct vidi_win_data win_data[WINDOWS_NR]; struct edid *raw_edid; unsigned int clkdiv; @@ -560,9 +561,10 @@ static struct exynos_drm_display_ops vidi_display_ops = { .create_connector = vidi_create_connector, }; -static int vidi_subdrv_probe(struct drm_device *drm_dev, struct device *dev) +static int vidi_bind(struct device *dev, struct device *master, void *data) { struct vidi_context *ctx = dev_get_drvdata(dev); + struct drm_device *drm_dev = data; struct drm_crtc *crtc = ctx->crtc; int ret; @@ -584,9 +586,18 @@ static int vidi_subdrv_probe(struct drm_device *drm_dev, struct device *dev) return 0; } + +static void vidi_unbind(struct device *dev, struct device *master, void *data) +{ +} + +static const struct component_ops vidi_component_ops = { + .bind = vidi_bind, + .unbind = vidi_unbind, +}; + static int vidi_probe(struct platform_device *pdev) { - struct exynos_drm_subdrv *subdrv; struct vidi_context *ctx; int ret; @@ -599,6 +610,17 @@ static int vidi_probe(struct platform_device *pdev) ctx->display.type = EXYNOS_DISPLAY_TYPE_VIDI; ctx->display.ops = &vidi_display_ops; ctx->default_win = 0; + ctx->pdev = pdev; + + ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CRTC, + ctx->manager.type); + if (ret) + return ret; + + ret = exynos_drm_component_add(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR, + ctx->display.type); + if (ret) + goto err_del_crtc_component; INIT_WORK(&ctx->work, vidi_fake_vblank_handler); @@ -606,23 +628,26 @@ static int vidi_probe(struct platform_device *pdev) platform_set_drvdata(pdev, ctx); - subdrv = &ctx->subdrv; - subdrv->dev = &pdev->dev; - subdrv->probe = vidi_subdrv_probe; - - ret = exynos_drm_subdrv_register(subdrv); - if (ret < 0) { - dev_err(&pdev->dev, "failed to register drm vidi device\n"); - return ret; - } - ret = device_create_file(&pdev->dev, &dev_attr_connection); if (ret < 0) { - exynos_drm_subdrv_unregister(subdrv); - DRM_INFO("failed to create connection sysfs.\n"); + DRM_ERROR("failed to create connection sysfs.\n"); + goto err_del_conn_component; } - return 0; + ret = component_add(&pdev->dev, &vidi_component_ops); + if (ret) + goto err_remove_file; + + return ret; + +err_remove_file: + device_remove_file(&pdev->dev, &dev_attr_connection); +err_del_conn_component: + exynos_drm_component_del(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR); +err_del_crtc_component: + exynos_drm_component_del(&pdev->dev, EXYNOS_DEVICE_TYPE_CRTC); + + return ret; } static int vidi_remove(struct platform_device *pdev) @@ -636,6 +661,10 @@ static int vidi_remove(struct platform_device *pdev) return -EINVAL; } + component_del(&pdev->dev, &vidi_component_ops); + exynos_drm_component_del(&pdev->dev, EXYNOS_DEVICE_TYPE_CONNECTOR); + exynos_drm_component_del(&pdev->dev, EXYNOS_DEVICE_TYPE_CRTC); + return 0; } -- cgit v0.10.2 From 1c9ff4ab43a83f2b412f81ad4db862e5533c745d Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Mon, 24 Nov 2014 15:19:49 -0200 Subject: drm/exynos: Fix exynos_dpi_remove() parameter exynos_dpi_remove() should receive a exynos_drm_display but when DRM_EXYNOS_DPI was disabled it was receiving a struct device resulting in ia compiler warning. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index 2e4e91b..2e50634 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -307,7 +307,10 @@ int exynos_dpi_remove(struct exynos_drm_display *display); #else static inline struct exynos_drm_display * exynos_dpi_probe(struct device *dev) { return NULL; } -static inline int exynos_dpi_remove(struct device *dev) { return 0; } +static inline int exynos_dpi_remove(struct exynos_drm_display *display) +{ + return 0; +} #endif #ifdef CONFIG_DRM_EXYNOS_VIDI -- cgit v0.10.2 From 5baf5d44fbcde002d7f3f8148e69305f520770dd Mon Sep 17 00:00:00 2001 From: Gustavo Padovan Date: Mon, 24 Nov 2014 16:23:30 -0200 Subject: drm/exynos: avoid leak if exynos_dpi_probe() fails The component must be deleted if the probe fails. Signed-off-by: Gustavo Padovan Signed-off-by: Inki Dae diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index ef80a35..e5810d1 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -1215,8 +1215,10 @@ static int fimd_probe(struct platform_device *pdev) platform_set_drvdata(pdev, ctx); ctx->display = exynos_dpi_probe(dev); - if (IS_ERR(ctx->display)) - return PTR_ERR(ctx->display); + if (IS_ERR(ctx->display)) { + ret = PTR_ERR(ctx->display); + goto err_del_component; + } pm_runtime_enable(dev); -- cgit v0.10.2 From 66333cb3d7d4ff853b6945f01e68d8986d821235 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Tue, 25 Nov 2014 13:21:30 +0300 Subject: amdkfd: fix some error handling in ioctl There is a typo here so the errors from kfd_bind_process_to_device() are not detected. Reviewed-by: Oded Gabbay Signed-off-by: Dan Carpenter Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 3b3fce7..102cd36 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -242,7 +242,7 @@ static long kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p, mutex_lock(&p->mutex); pdd = kfd_bind_process_to_device(dev, p); - if (IS_ERR(pdd) < 0) { + if (IS_ERR(pdd)) { err = PTR_ERR(pdd); goto err_bind_process; } @@ -389,7 +389,7 @@ static long kfd_ioctl_set_memory_policy(struct file *filep, mutex_lock(&p->mutex); pdd = kfd_bind_process_to_device(dev, p); - if (IS_ERR(pdd) < 0) { + if (IS_ERR(pdd)) { err = PTR_ERR(pdd); goto out; } -- cgit v0.10.2 From e048a0b26033b35eb36c20e4d36fb466a2570c27 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Tue, 25 Nov 2014 13:24:51 +0300 Subject: amdkfd: fix an error handling bug in pqm_create_queue() The call to kernel_queue_uninit(NULL) will trigger a BUG(), and also the error code is incorrect. Fixes: 45102048f77e ('amdkfd: Add process queue manager module') Reviewed-by: Oded Gabbay Signed-off-by: Dan Carpenter Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index de2c163..4752678 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -208,7 +208,7 @@ int pqm_create_queue(struct process_queue_manager *pqm, case KFD_QUEUE_TYPE_DIQ: kq = kernel_queue_init(dev, KFD_QUEUE_TYPE_DIQ); if (kq == NULL) { - kernel_queue_uninit(kq); + retval = -ENOMEM; goto err_create_queue; } kq->queue->properties.queue_id = *qid; -- cgit v0.10.2 From b0ff4b93f646c39900ead9f6f7b6da78b5978273 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 24 Nov 2014 20:01:58 +0100 Subject: drm: Document that drm_dev_alloc doesn't need a parent Possible for purely virtual debug devices. Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c index 2e5c7d9..4f41377 100644 --- a/drivers/gpu/drm/drm_drv.c +++ b/drivers/gpu/drm/drm_drv.c @@ -535,6 +535,8 @@ static void drm_fs_inode_free(struct inode *inode) * The initial ref-count of the object is 1. Use drm_dev_ref() and * drm_dev_unref() to take and drop further ref-counts. * + * Note that for purely virtual devices @parent can be NULL. + * * RETURNS: * Pointer to new DRM device, or NULL if out of memory. */ -- cgit v0.10.2 From ab58e3384b9f9863bfd029b458ff337d381bf6d2 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 24 Nov 2014 20:42:42 +0100 Subject: drm/atomic-helper: Skip vblank waits for unchanged fbs Especially with legacy cursor ioctls existing userspace assumes that you can pile up lots of updates in one go. The super-proper way to support this would be a special commit mode which overwrites the last update. But getting there will be quite a bit of work. Meanwhile do what pretty much all the drivers have done for the plane update functions: Simply skip the vblank wait for the buffer cleanup if the buffer is the same. Since the universal cursor plane code will not recreate framebuffers needlessly this allows us to not slow down legacy pageflip events while someone moves the cursor around. v2: Drop the async plane update hunk from a previous attempt at this issue. v3: Fix up kerneldoc. v4: Don't oops so badly. Reported by Jasper. Cc: Rob Clark Cc: "Jasper St. Pierre" Reviewed-by: Rob Clark Reviewed-by: Jasper St. Pierre Tested-by: Jasper St. Pierre Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index a17b8e9..4368581 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -751,6 +751,33 @@ static void wait_for_fences(struct drm_device *dev, } } +static bool framebuffer_changed(struct drm_device *dev, + struct drm_atomic_state *old_state, + struct drm_crtc *crtc) +{ + struct drm_plane *plane; + struct drm_plane_state *old_plane_state; + int nplanes = old_state->dev->mode_config.num_total_plane; + int i; + + for (i = 0; i < nplanes; i++) { + plane = old_state->planes[i]; + old_plane_state = old_state->plane_states[i]; + + if (!plane) + continue; + + if (plane->state->crtc != crtc && + old_plane_state->crtc != crtc) + continue; + + if (plane->state->fb != old_plane_state->fb) + return true; + } + + return false; +} + /** * drm_atomic_helper_wait_for_vblanks - wait for vblank on crtcs * @dev: DRM device @@ -758,7 +785,9 @@ static void wait_for_fences(struct drm_device *dev, * * Helper to, after atomic commit, wait for vblanks on all effected * crtcs (ie. before cleaning up old framebuffers using - * drm_atomic_helper_cleanup_planes()) + * drm_atomic_helper_cleanup_planes()). It will only wait on crtcs where the + * framebuffers have actually changed to optimize for the legacy cursor and + * plane update use-case. */ void drm_atomic_helper_wait_for_vblanks(struct drm_device *dev, @@ -784,6 +813,9 @@ drm_atomic_helper_wait_for_vblanks(struct drm_device *dev, if (!crtc->state->enable) continue; + if (!framebuffer_changed(dev, old_state, crtc)) + continue; + ret = drm_crtc_vblank_get(crtc); if (ret != 0) continue; -- cgit v0.10.2 From 9c04b7e3698a2e77b3473ef91a26ecb384459a04 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 24 Nov 2014 20:51:21 +0100 Subject: drm/atomic: Drop per-plane locking TODO I've forgotten to remove that in my per-plane locking patch. Reported-by: Rob Clark Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index d3b4674..ba49b5c 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -243,12 +243,6 @@ drm_atomic_get_plane_state(struct drm_atomic_state *state, if (state->plane_states[index]) return state->plane_states[index]; - /* - * TODO: We currently don't have per-plane mutexes. So instead of trying - * crazy tricks with deferring plane->crtc and hoping for the best just - * grab all crtc locks. Once we have per-plane locks we must update this - * to only take the plane mutex. - */ ret = drm_modeset_lock(&plane->mutex, state->acquire_ctx); if (ret) return ERR_PTR(ret); -- cgit v0.10.2 From aa54e2ee80b4f653f75b9139ae7500ee8cd5ad5f Mon Sep 17 00:00:00 2001 From: "Jasper St. Pierre" Date: Thu, 20 Nov 2014 19:59:15 -0800 Subject: drm/atomic_helper: Cope with plane->crtc == NULL in disable helper The drm core can call the plane disable hook multiple times, which means it can get called when plane->crtc is already NULL. That in turn means we can't get at the implicit acquire ctx we use in the atomic helpers for legacy entries points. We could try to pass drm_modeset_legacy_acquire_ctx a drm_device pointer so that it can cope with a NULL crtc. But that still doesn't work since the cursor ioctls (remapped with the universal cursor plane support code) only grabs the crtc locks. So the global acquire context isn't set eitehr. The real solution here would be to bite the bullet and wire up explicit acquire context parameters to all relevant functions. We need to do that anyway (to be able to get rid of some small allocations which we can't cope with failing). But that's a lot of work and better done once atomic has settled a bit. So meanwhile just catch this case in the helper and bail out. Signed-off-by: Jasper St. Pierre Reviewed-by: Rob Clark Cc: Daniel Vetter [danvet: Completely rewrite commit message and comment but keep Jasper's logic and author credits since his patch is the only short-term solution that works.] Tested-by: Thierry Reding Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 4368581..d981d07 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -1275,6 +1275,17 @@ int drm_atomic_helper_disable_plane(struct drm_plane *plane) struct drm_plane_state *plane_state; int ret = 0; + /* + * FIXME: Without plane->crtc set we can't get at the implicit legacy + * acquire context. The real fix will be to wire the acquire ctx through + * everywhere we need it, but meanwhile prevent chaos by just skipping + * this noop. The critical case is the cursor ioctls which a) only grab + * crtc/cursor-plane locks (so we need the crtc to get at the right + * acquire context) and b) can try to disable the plane multiple times. + */ + if (!plane->crtc) + return 0; + state = drm_atomic_state_alloc(plane->dev); if (!state) return -ENOMEM; -- cgit v0.10.2 From f1c37e1adc6eca1fb492c74d466141d9b01e0428 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 25 Nov 2014 12:09:44 +0100 Subject: drm/plane: Pass old state to ->atomic_update() In most situations it will be useful to have the old state passed to the ->atomic_update() callback. For example if a plane is being disabled the new state's .crtc field will be NULL, but some drivers may rely on this field to program the CRTCs registers. v2: rename variable to old_plane_state and remove redundant comment as suggested by Daniel Vetter, remove an Exynos hunk that doesn't apply to drm-next and add a hunk for pending MSM mdp5 changes Reviewed-by: Daniel Vetter Signed-off-by: Thierry Reding Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index d981d07..2fa0840 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -1046,6 +1046,7 @@ void drm_atomic_helper_commit_planes(struct drm_device *dev, for (i = 0; i < nplanes; i++) { struct drm_plane_helper_funcs *funcs; struct drm_plane *plane = old_state->planes[i]; + struct drm_plane_state *old_plane_state; if (!plane) continue; @@ -1055,7 +1056,9 @@ void drm_atomic_helper_commit_planes(struct drm_device *dev, if (!funcs || !funcs->atomic_update) continue; - funcs->atomic_update(plane); + old_plane_state = old_state->plane_states[i]; + + funcs->atomic_update(plane, old_plane_state); } for (i = 0; i < ncrtcs; i++) { diff --git a/drivers/gpu/drm/drm_plane_helper.c b/drivers/gpu/drm/drm_plane_helper.c index 93c6533..18a1ac6 100644 --- a/drivers/gpu/drm/drm_plane_helper.c +++ b/drivers/gpu/drm/drm_plane_helper.c @@ -443,7 +443,7 @@ int drm_plane_helper_commit(struct drm_plane *plane, crtc_funcs[i]->atomic_begin(crtc[i]); } - plane_funcs->atomic_update(plane); + plane_funcs->atomic_update(plane, plane_state); for (i = 0; i < 2; i++) { if (crtc_funcs[i] && crtc_funcs[i]->atomic_flush) diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c index 76d0a40..1e5ebe8 100644 --- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c @@ -107,7 +107,8 @@ static int mdp4_plane_atomic_check(struct drm_plane *plane, return 0; } -static void mdp4_plane_atomic_update(struct drm_plane *plane) +static void mdp4_plane_atomic_update(struct drm_plane *plane, + struct drm_plane_state *old_state) { struct drm_plane_state *state = plane->state; int ret; diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c index 533df7c..26e5fde 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c @@ -213,7 +213,8 @@ static int mdp5_plane_atomic_check(struct drm_plane *plane, return 0; } -static void mdp5_plane_atomic_update(struct drm_plane *plane) +static void mdp5_plane_atomic_update(struct drm_plane *plane, + struct drm_plane_state *old_state) { struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane); struct drm_plane_state *state = plane->state; diff --git a/include/drm/drm_plane_helper.h b/include/drm/drm_plane_helper.h index c48f14d..fcfa969 100644 --- a/include/drm/drm_plane_helper.h +++ b/include/drm/drm_plane_helper.h @@ -60,7 +60,8 @@ struct drm_plane_helper_funcs { int (*atomic_check)(struct drm_plane *plane, struct drm_plane_state *state); - void (*atomic_update)(struct drm_plane *plane); + void (*atomic_update)(struct drm_plane *plane, + struct drm_plane_state *old_state); }; static inline void drm_plane_helper_add(struct drm_plane *plane, -- cgit v0.10.2 From 5d571690154cbe7e7b4d24ea8af2393146609b0e Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 25 Nov 2014 12:09:45 +0100 Subject: drm/plane: Add missing kerneldoc The plane helpers aren't pulled into the DocBook yet, so these weren't noticed. Reviewed-by: Daniel Vetter Signed-off-by: Thierry Reding Signed-off-by: Daniel Vetter diff --git a/include/drm/drm_plane_helper.h b/include/drm/drm_plane_helper.h index fcfa969..a185392 100644 --- a/include/drm/drm_plane_helper.h +++ b/include/drm/drm_plane_helper.h @@ -49,6 +49,10 @@ extern int drm_crtc_init(struct drm_device *dev, /** * drm_plane_helper_funcs - helper operations for CRTCs + * @prepare_fb: prepare a framebuffer for use by the plane + * @cleanup_fb: cleanup a framebuffer when it's no longer used by the plane + * @atomic_check: check that a given atomic state is valid and can be applied + * @atomic_update: apply an atomic state to the plane * * The helper operations are called by the mid-layer CRTC helper. */ -- cgit v0.10.2 From 0254951d9c581ea21f989f7c8248079fb414a152 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 25 Nov 2014 12:09:47 +0100 Subject: drm: Make drm_atomic_helper.h standalone includible This header uses a bunch of declarations from the drm/drm_crtc.h header, so make sure to include that as well so that drm_atomic_helper.h can be included standalone. Reviewed-by: Daniel Vetter Signed-off-by: Thierry Reding Signed-off-by: Daniel Vetter diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h index 64b4e91..70a8319 100644 --- a/include/drm/drm_atomic_helper.h +++ b/include/drm/drm_atomic_helper.h @@ -28,6 +28,8 @@ #ifndef DRM_ATOMIC_HELPER_H_ #define DRM_ATOMIC_HELPER_H_ +#include + int drm_atomic_helper_check(struct drm_device *dev, struct drm_atomic_state *state); int drm_atomic_helper_commit(struct drm_device *dev, -- cgit v0.10.2 From 37cc0148773abbd8e7dbf8b2cb3399e887233fc2 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 25 Nov 2014 12:09:48 +0100 Subject: drm: Make drm_atomic.h standalone includible This header file makes use of a bunch of structures declared in the drm_crtc.h header file. Include that to make sure the drm_atomic.h header can be included standalone. Signed-off-by: Thierry Reding Signed-off-by: Daniel Vetter diff --git a/include/drm/drm_atomic.h b/include/drm/drm_atomic.h index 9d91916..e224ccf 100644 --- a/include/drm/drm_atomic.h +++ b/include/drm/drm_atomic.h @@ -28,6 +28,8 @@ #ifndef DRM_ATOMIC_H_ #define DRM_ATOMIC_H_ +#include + struct drm_atomic_state * __must_check drm_atomic_state_alloc(struct drm_device *dev); void drm_atomic_state_clear(struct drm_atomic_state *state); -- cgit v0.10.2 From 3009c0377f25c29852b218a6933a969d02cbdc5d Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Tue, 25 Nov 2014 12:09:49 +0100 Subject: drm: Free atomic state during cleanup The current state of CRTCs, planes and connectors currently leaks during DRM driver ->unload() unless drivers explicitly clean it up. Since there is nothing driver-specific about it, that cleanup can be done within the DRM core. Reviewed-by: Daniel Vetter Signed-off-by: Thierry Reding Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 589a921..e4e7b92 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -721,6 +721,10 @@ void drm_crtc_cleanup(struct drm_crtc *crtc) drm_mode_object_put(dev, &crtc->base); list_del(&crtc->head); dev->mode_config.num_crtc--; + + WARN_ON(crtc->state && !crtc->funcs->atomic_destroy_state); + if (crtc->state && crtc->funcs->atomic_destroy_state) + crtc->funcs->atomic_destroy_state(crtc, crtc->state); } EXPORT_SYMBOL(drm_crtc_cleanup); @@ -918,6 +922,11 @@ void drm_connector_cleanup(struct drm_connector *connector) connector->name = NULL; list_del(&connector->head); dev->mode_config.num_connector--; + + WARN_ON(connector->state && !connector->funcs->atomic_destroy_state); + if (connector->state && connector->funcs->atomic_destroy_state) + connector->funcs->atomic_destroy_state(connector, + connector->state); } EXPORT_SYMBOL(drm_connector_cleanup); @@ -1244,6 +1253,10 @@ void drm_plane_cleanup(struct drm_plane *plane) if (plane->type == DRM_PLANE_TYPE_OVERLAY) dev->mode_config.num_overlay_plane--; drm_modeset_unlock_all(dev); + + WARN_ON(plane->state && !plane->funcs->atomic_destroy_state); + if (plane->state && plane->funcs->atomic_destroy_state) + plane->funcs->atomic_destroy_state(plane, plane->state); } EXPORT_SYMBOL(drm_plane_cleanup); -- cgit v0.10.2 From 6f9d54fd6e0125b17fb286e1e54e934dd11f60d3 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Tue, 25 Nov 2014 15:16:38 +0200 Subject: amdkfd: Fix memory leak of mqds on dqm fini The mqds array members are not freed when dqm is uninitialized. Reviewed-by: Ben Goz Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 904eb38..924e90c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -575,11 +575,15 @@ static int initialize_nocpsch(struct device_queue_manager *dqm) static void uninitialize_nocpsch(struct device_queue_manager *dqm) { + int i; + BUG_ON(!dqm); BUG_ON(dqm->queue_count > 0 || dqm->processes_count > 0); kfree(dqm->allocated_queues); + for (i = 0 ; i < KFD_MQD_TYPE_MAX ; i++) + kfree(dqm->mqds[i]); mutex_destroy(&dqm->lock); kfd2kgd->free_mem(dqm->dev->kgd, (struct kgd_mem *) dqm->pipeline_mem); -- cgit v0.10.2 From 9cf4a281314c9345c13c7bbb5786d7b8a72553f0 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Tue, 25 Nov 2014 19:43:29 +0300 Subject: amdkfd: delete some dead code This is dead code. We don't need to unbind here, we can just return directly. Reviewed-by: Oded Gabbay Signed-off-by: Dan Carpenter Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index a17e2dd..b4f49ac 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -348,11 +348,6 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, if (err < 0) return ERR_PTR(err); - if (err < 0) { - amd_iommu_unbind_pasid(dev->pdev, p->pasid); - return ERR_PTR(err); - } - pdd->bound = true; return pdd; -- cgit v0.10.2 From 6556f7f82b9c401950d703072c0d8137b6f9f516 Mon Sep 17 00:00:00 2001 From: Philipp Zabel Date: Mon, 24 Nov 2014 16:33:34 +0100 Subject: drm: imx: Move imx-drm driver out of staging The imx-drm driver was put into staging mostly for the following reasons, all of which have been addressed or superseded: - convert the irq driver to use linear irq domains - work out the device tree bindings, this lead to the common of_graph bindings being used - factor out common helper functions, this mostly resulted in the component framework and drm of_graph helpers. Before adding new fixes, and certainly before adding new features, move it into its proper place below drivers/gpu/drm. Signed-off-by: Philipp Zabel Signed-off-by: Dave Airlie diff --git a/Documentation/devicetree/bindings/drm/imx/fsl-imx-drm.txt b/Documentation/devicetree/bindings/drm/imx/fsl-imx-drm.txt new file mode 100644 index 0000000..e75f0e5 --- /dev/null +++ b/Documentation/devicetree/bindings/drm/imx/fsl-imx-drm.txt @@ -0,0 +1,83 @@ +Freescale i.MX DRM master device +================================ + +The freescale i.MX DRM master device is a virtual device needed to list all +IPU or other display interface nodes that comprise the graphics subsystem. + +Required properties: +- compatible: Should be "fsl,imx-display-subsystem" +- ports: Should contain a list of phandles pointing to display interface ports + of IPU devices + +example: + +display-subsystem { + compatible = "fsl,display-subsystem"; + ports = <&ipu_di0>; +}; + + +Freescale i.MX IPUv3 +==================== + +Required properties: +- compatible: Should be "fsl,-ipu" +- reg: should be register base and length as documented in the + datasheet +- interrupts: Should contain sync interrupt and error interrupt, + in this order. +- resets: phandle pointing to the system reset controller and + reset line index, see reset/fsl,imx-src.txt for details +Optional properties: +- port@[0-3]: Port nodes with endpoint definitions as defined in + Documentation/devicetree/bindings/media/video-interfaces.txt. + Ports 0 and 1 should correspond to CSI0 and CSI1, + ports 2 and 3 should correspond to DI0 and DI1, respectively. + +example: + +ipu: ipu@18000000 { + #address-cells = <1>; + #size-cells = <0>; + compatible = "fsl,imx53-ipu"; + reg = <0x18000000 0x080000000>; + interrupts = <11 10>; + resets = <&src 2>; + + ipu_di0: port@2 { + reg = <2>; + + ipu_di0_disp0: endpoint { + remote-endpoint = <&display_in>; + }; + }; +}; + +Parallel display support +======================== + +Required properties: +- compatible: Should be "fsl,imx-parallel-display" +Optional properties: +- interface_pix_fmt: How this display is connected to the + display interface. Currently supported types: "rgb24", "rgb565", "bgr666" + and "lvds666". +- edid: verbatim EDID data block describing attached display. +- ddc: phandle describing the i2c bus handling the display data + channel +- port: A port node with endpoint definitions as defined in + Documentation/devicetree/bindings/media/video-interfaces.txt. + +example: + +display@di0 { + compatible = "fsl,imx-parallel-display"; + edid = [edid-data]; + interface-pix-fmt = "rgb24"; + + port { + display_in: endpoint { + remote-endpoint = <&ipu_di0_disp0>; + }; + }; +}; diff --git a/Documentation/devicetree/bindings/drm/imx/hdmi.txt b/Documentation/devicetree/bindings/drm/imx/hdmi.txt new file mode 100644 index 0000000..1b756cf --- /dev/null +++ b/Documentation/devicetree/bindings/drm/imx/hdmi.txt @@ -0,0 +1,58 @@ +Device-Tree bindings for HDMI Transmitter + +HDMI Transmitter +================ + +The HDMI Transmitter is a Synopsys DesignWare HDMI 1.4 TX controller IP +with accompanying PHY IP. + +Required properties: + - #address-cells : should be <1> + - #size-cells : should be <0> + - compatible : should be "fsl,imx6q-hdmi" or "fsl,imx6dl-hdmi". + - gpr : should be <&gpr>. + The phandle points to the iomuxc-gpr region containing the HDMI + multiplexer control register. + - clocks, clock-names : phandles to the HDMI iahb and isrf clocks, as described + in Documentation/devicetree/bindings/clock/clock-bindings.txt and + Documentation/devicetree/bindings/clock/imx6q-clock.txt. + - port@[0-4]: Up to four port nodes with endpoint definitions as defined in + Documentation/devicetree/bindings/media/video-interfaces.txt, + corresponding to the four inputs to the HDMI multiplexer. + +Optional properties: + - ddc-i2c-bus: phandle of an I2C controller used for DDC EDID probing + +example: + + gpr: iomuxc-gpr@020e0000 { + /* ... */ + }; + + hdmi: hdmi@0120000 { + #address-cells = <1>; + #size-cells = <0>; + compatible = "fsl,imx6q-hdmi"; + reg = <0x00120000 0x9000>; + interrupts = <0 115 0x04>; + gpr = <&gpr>; + clocks = <&clks 123>, <&clks 124>; + clock-names = "iahb", "isfr"; + ddc-i2c-bus = <&i2c2>; + + port@0 { + reg = <0>; + + hdmi_mux_0: endpoint { + remote-endpoint = <&ipu1_di0_hdmi>; + }; + }; + + port@1 { + reg = <1>; + + hdmi_mux_1: endpoint { + remote-endpoint = <&ipu1_di1_hdmi>; + }; + }; + }; diff --git a/Documentation/devicetree/bindings/drm/imx/ldb.txt b/Documentation/devicetree/bindings/drm/imx/ldb.txt new file mode 100644 index 0000000..443bcb6 --- /dev/null +++ b/Documentation/devicetree/bindings/drm/imx/ldb.txt @@ -0,0 +1,122 @@ +Device-Tree bindings for LVDS Display Bridge (ldb) + +LVDS Display Bridge +=================== + +The LVDS Display Bridge device tree node contains up to two lvds-channel +nodes describing each of the two LVDS encoder channels of the bridge. + +Required properties: + - #address-cells : should be <1> + - #size-cells : should be <0> + - compatible : should be "fsl,imx53-ldb" or "fsl,imx6q-ldb". + Both LDB versions are similar, but i.MX6 has an additional + multiplexer in the front to select any of the four IPU display + interfaces as input for each LVDS channel. + - gpr : should be <&gpr> on i.MX53 and i.MX6q. + The phandle points to the iomuxc-gpr region containing the LVDS + control register. +- clocks, clock-names : phandles to the LDB divider and selector clocks and to + the display interface selector clocks, as described in + Documentation/devicetree/bindings/clock/clock-bindings.txt + The following clocks are expected on i.MX53: + "di0_pll" - LDB LVDS channel 0 mux + "di1_pll" - LDB LVDS channel 1 mux + "di0" - LDB LVDS channel 0 gate + "di1" - LDB LVDS channel 1 gate + "di0_sel" - IPU1 DI0 mux + "di1_sel" - IPU1 DI1 mux + On i.MX6q the following additional clocks are needed: + "di2_sel" - IPU2 DI0 mux + "di3_sel" - IPU2 DI1 mux + The needed clock numbers for each are documented in + Documentation/devicetree/bindings/clock/imx5-clock.txt, and in + Documentation/devicetree/bindings/clock/imx6q-clock.txt. + +Optional properties: + - pinctrl-names : should be "default" on i.MX53, not used on i.MX6q + - pinctrl-0 : a phandle pointing to LVDS pin settings on i.MX53, + not used on i.MX6q + - fsl,dual-channel : boolean. if it exists, only LVDS channel 0 should + be configured - one input will be distributed on both outputs in dual + channel mode + +LVDS Channel +============ + +Each LVDS Channel has to contain a display-timings node that describes the +video timings for the connected LVDS display. For detailed information, also +have a look at Documentation/devicetree/bindings/video/display-timing.txt. + +Required properties: + - reg : should be <0> or <1> + - fsl,data-mapping : should be "spwg" or "jeida" + This describes how the color bits are laid out in the + serialized LVDS signal. + - fsl,data-width : should be <18> or <24> + - port: A port node with endpoint definitions as defined in + Documentation/devicetree/bindings/media/video-interfaces.txt. + On i.MX5, the internal two-input-multiplexer is used. + Due to hardware limitations, only one port (port@[0,1]) + can be used for each channel (lvds-channel@[0,1], respectively) + On i.MX6, there should be four ports (port@[0-3]) that correspond + to the four LVDS multiplexer inputs. + +example: + +gpr: iomuxc-gpr@53fa8000 { + /* ... */ +}; + +ldb: ldb@53fa8008 { + #address-cells = <1>; + #size-cells = <0>; + compatible = "fsl,imx53-ldb"; + gpr = <&gpr>; + clocks = <&clks 122>, <&clks 120>, + <&clks 115>, <&clks 116>, + <&clks 123>, <&clks 85>; + clock-names = "di0_pll", "di1_pll", + "di0_sel", "di1_sel", + "di0", "di1"; + + lvds-channel@0 { + #address-cells = <1>; + #size-cells = <0>; + reg = <0>; + fsl,data-mapping = "spwg"; + fsl,data-width = <24>; + + display-timings { + /* ... */ + }; + + port@0 { + reg = <0>; + + lvds0_in: endpoint { + remote-endpoint = <&ipu_di0_lvds0>; + }; + }; + }; + + lvds-channel@1 { + #address-cells = <1>; + #size-cells = <0>; + reg = <1>; + fsl,data-mapping = "spwg"; + fsl,data-width = <24>; + + display-timings { + /* ... */ + }; + + port@1 { + reg = <1>; + + lvds1_in: endpoint { + remote-endpoint = <&ipu_di1_lvds1>; + }; + }; + }; +}; diff --git a/Documentation/devicetree/bindings/staging/imx-drm/fsl-imx-drm.txt b/Documentation/devicetree/bindings/staging/imx-drm/fsl-imx-drm.txt deleted file mode 100644 index e75f0e5..0000000 --- a/Documentation/devicetree/bindings/staging/imx-drm/fsl-imx-drm.txt +++ /dev/null @@ -1,83 +0,0 @@ -Freescale i.MX DRM master device -================================ - -The freescale i.MX DRM master device is a virtual device needed to list all -IPU or other display interface nodes that comprise the graphics subsystem. - -Required properties: -- compatible: Should be "fsl,imx-display-subsystem" -- ports: Should contain a list of phandles pointing to display interface ports - of IPU devices - -example: - -display-subsystem { - compatible = "fsl,display-subsystem"; - ports = <&ipu_di0>; -}; - - -Freescale i.MX IPUv3 -==================== - -Required properties: -- compatible: Should be "fsl,-ipu" -- reg: should be register base and length as documented in the - datasheet -- interrupts: Should contain sync interrupt and error interrupt, - in this order. -- resets: phandle pointing to the system reset controller and - reset line index, see reset/fsl,imx-src.txt for details -Optional properties: -- port@[0-3]: Port nodes with endpoint definitions as defined in - Documentation/devicetree/bindings/media/video-interfaces.txt. - Ports 0 and 1 should correspond to CSI0 and CSI1, - ports 2 and 3 should correspond to DI0 and DI1, respectively. - -example: - -ipu: ipu@18000000 { - #address-cells = <1>; - #size-cells = <0>; - compatible = "fsl,imx53-ipu"; - reg = <0x18000000 0x080000000>; - interrupts = <11 10>; - resets = <&src 2>; - - ipu_di0: port@2 { - reg = <2>; - - ipu_di0_disp0: endpoint { - remote-endpoint = <&display_in>; - }; - }; -}; - -Parallel display support -======================== - -Required properties: -- compatible: Should be "fsl,imx-parallel-display" -Optional properties: -- interface_pix_fmt: How this display is connected to the - display interface. Currently supported types: "rgb24", "rgb565", "bgr666" - and "lvds666". -- edid: verbatim EDID data block describing attached display. -- ddc: phandle describing the i2c bus handling the display data - channel -- port: A port node with endpoint definitions as defined in - Documentation/devicetree/bindings/media/video-interfaces.txt. - -example: - -display@di0 { - compatible = "fsl,imx-parallel-display"; - edid = [edid-data]; - interface-pix-fmt = "rgb24"; - - port { - display_in: endpoint { - remote-endpoint = <&ipu_di0_disp0>; - }; - }; -}; diff --git a/Documentation/devicetree/bindings/staging/imx-drm/hdmi.txt b/Documentation/devicetree/bindings/staging/imx-drm/hdmi.txt deleted file mode 100644 index 1b756cf..0000000 --- a/Documentation/devicetree/bindings/staging/imx-drm/hdmi.txt +++ /dev/null @@ -1,58 +0,0 @@ -Device-Tree bindings for HDMI Transmitter - -HDMI Transmitter -================ - -The HDMI Transmitter is a Synopsys DesignWare HDMI 1.4 TX controller IP -with accompanying PHY IP. - -Required properties: - - #address-cells : should be <1> - - #size-cells : should be <0> - - compatible : should be "fsl,imx6q-hdmi" or "fsl,imx6dl-hdmi". - - gpr : should be <&gpr>. - The phandle points to the iomuxc-gpr region containing the HDMI - multiplexer control register. - - clocks, clock-names : phandles to the HDMI iahb and isrf clocks, as described - in Documentation/devicetree/bindings/clock/clock-bindings.txt and - Documentation/devicetree/bindings/clock/imx6q-clock.txt. - - port@[0-4]: Up to four port nodes with endpoint definitions as defined in - Documentation/devicetree/bindings/media/video-interfaces.txt, - corresponding to the four inputs to the HDMI multiplexer. - -Optional properties: - - ddc-i2c-bus: phandle of an I2C controller used for DDC EDID probing - -example: - - gpr: iomuxc-gpr@020e0000 { - /* ... */ - }; - - hdmi: hdmi@0120000 { - #address-cells = <1>; - #size-cells = <0>; - compatible = "fsl,imx6q-hdmi"; - reg = <0x00120000 0x9000>; - interrupts = <0 115 0x04>; - gpr = <&gpr>; - clocks = <&clks 123>, <&clks 124>; - clock-names = "iahb", "isfr"; - ddc-i2c-bus = <&i2c2>; - - port@0 { - reg = <0>; - - hdmi_mux_0: endpoint { - remote-endpoint = <&ipu1_di0_hdmi>; - }; - }; - - port@1 { - reg = <1>; - - hdmi_mux_1: endpoint { - remote-endpoint = <&ipu1_di1_hdmi>; - }; - }; - }; diff --git a/Documentation/devicetree/bindings/staging/imx-drm/ldb.txt b/Documentation/devicetree/bindings/staging/imx-drm/ldb.txt deleted file mode 100644 index 443bcb6..0000000 --- a/Documentation/devicetree/bindings/staging/imx-drm/ldb.txt +++ /dev/null @@ -1,122 +0,0 @@ -Device-Tree bindings for LVDS Display Bridge (ldb) - -LVDS Display Bridge -=================== - -The LVDS Display Bridge device tree node contains up to two lvds-channel -nodes describing each of the two LVDS encoder channels of the bridge. - -Required properties: - - #address-cells : should be <1> - - #size-cells : should be <0> - - compatible : should be "fsl,imx53-ldb" or "fsl,imx6q-ldb". - Both LDB versions are similar, but i.MX6 has an additional - multiplexer in the front to select any of the four IPU display - interfaces as input for each LVDS channel. - - gpr : should be <&gpr> on i.MX53 and i.MX6q. - The phandle points to the iomuxc-gpr region containing the LVDS - control register. -- clocks, clock-names : phandles to the LDB divider and selector clocks and to - the display interface selector clocks, as described in - Documentation/devicetree/bindings/clock/clock-bindings.txt - The following clocks are expected on i.MX53: - "di0_pll" - LDB LVDS channel 0 mux - "di1_pll" - LDB LVDS channel 1 mux - "di0" - LDB LVDS channel 0 gate - "di1" - LDB LVDS channel 1 gate - "di0_sel" - IPU1 DI0 mux - "di1_sel" - IPU1 DI1 mux - On i.MX6q the following additional clocks are needed: - "di2_sel" - IPU2 DI0 mux - "di3_sel" - IPU2 DI1 mux - The needed clock numbers for each are documented in - Documentation/devicetree/bindings/clock/imx5-clock.txt, and in - Documentation/devicetree/bindings/clock/imx6q-clock.txt. - -Optional properties: - - pinctrl-names : should be "default" on i.MX53, not used on i.MX6q - - pinctrl-0 : a phandle pointing to LVDS pin settings on i.MX53, - not used on i.MX6q - - fsl,dual-channel : boolean. if it exists, only LVDS channel 0 should - be configured - one input will be distributed on both outputs in dual - channel mode - -LVDS Channel -============ - -Each LVDS Channel has to contain a display-timings node that describes the -video timings for the connected LVDS display. For detailed information, also -have a look at Documentation/devicetree/bindings/video/display-timing.txt. - -Required properties: - - reg : should be <0> or <1> - - fsl,data-mapping : should be "spwg" or "jeida" - This describes how the color bits are laid out in the - serialized LVDS signal. - - fsl,data-width : should be <18> or <24> - - port: A port node with endpoint definitions as defined in - Documentation/devicetree/bindings/media/video-interfaces.txt. - On i.MX5, the internal two-input-multiplexer is used. - Due to hardware limitations, only one port (port@[0,1]) - can be used for each channel (lvds-channel@[0,1], respectively) - On i.MX6, there should be four ports (port@[0-3]) that correspond - to the four LVDS multiplexer inputs. - -example: - -gpr: iomuxc-gpr@53fa8000 { - /* ... */ -}; - -ldb: ldb@53fa8008 { - #address-cells = <1>; - #size-cells = <0>; - compatible = "fsl,imx53-ldb"; - gpr = <&gpr>; - clocks = <&clks 122>, <&clks 120>, - <&clks 115>, <&clks 116>, - <&clks 123>, <&clks 85>; - clock-names = "di0_pll", "di1_pll", - "di0_sel", "di1_sel", - "di0", "di1"; - - lvds-channel@0 { - #address-cells = <1>; - #size-cells = <0>; - reg = <0>; - fsl,data-mapping = "spwg"; - fsl,data-width = <24>; - - display-timings { - /* ... */ - }; - - port@0 { - reg = <0>; - - lvds0_in: endpoint { - remote-endpoint = <&ipu_di0_lvds0>; - }; - }; - }; - - lvds-channel@1 { - #address-cells = <1>; - #size-cells = <0>; - reg = <1>; - fsl,data-mapping = "spwg"; - fsl,data-width = <24>; - - display-timings { - /* ... */ - }; - - port@1 { - reg = <1>; - - lvds1_in: endpoint { - remote-endpoint = <&ipu_di1_lvds1>; - }; - }; - }; -}; diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig index 37c5a6e..24c2d7c 100644 --- a/drivers/gpu/drm/Kconfig +++ b/drivers/gpu/drm/Kconfig @@ -202,3 +202,5 @@ source "drivers/gpu/drm/panel/Kconfig" source "drivers/gpu/drm/sti/Kconfig" source "drivers/gpu/drm/amd/amdkfd/Kconfig" + +source "drivers/gpu/drm/imx/Kconfig" diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile index dd9d35b..47d8986 100644 --- a/drivers/gpu/drm/Makefile +++ b/drivers/gpu/drm/Makefile @@ -62,6 +62,7 @@ obj-$(CONFIG_DRM_BOCHS) += bochs/ obj-$(CONFIG_DRM_MSM) += msm/ obj-$(CONFIG_DRM_TEGRA) += tegra/ obj-$(CONFIG_DRM_STI) += sti/ +obj-$(CONFIG_DRM_IMX) += imx/ obj-y += i2c/ obj-y += panel/ obj-y += bridge/ diff --git a/drivers/gpu/drm/imx/Kconfig b/drivers/gpu/drm/imx/Kconfig new file mode 100644 index 0000000..82fb758 --- /dev/null +++ b/drivers/gpu/drm/imx/Kconfig @@ -0,0 +1,53 @@ +config DRM_IMX + tristate "DRM Support for Freescale i.MX" + select DRM_KMS_HELPER + select DRM_KMS_FB_HELPER + select VIDEOMODE_HELPERS + select DRM_GEM_CMA_HELPER + select DRM_KMS_CMA_HELPER + depends on DRM && (ARCH_MXC || ARCH_MULTIPLATFORM) + help + enable i.MX graphics support + +config DRM_IMX_FB_HELPER + tristate "provide legacy framebuffer /dev/fb0" + select DRM_KMS_CMA_HELPER + depends on DRM_IMX + help + The DRM framework can provide a legacy /dev/fb0 framebuffer + for your device. This is necessary to get a framebuffer console + and also for applications using the legacy framebuffer API + +config DRM_IMX_PARALLEL_DISPLAY + tristate "Support for parallel displays" + select DRM_PANEL + depends on DRM_IMX + select VIDEOMODE_HELPERS + +config DRM_IMX_TVE + tristate "Support for TV and VGA displays" + depends on DRM_IMX + select REGMAP_MMIO + help + Choose this to enable the internal Television Encoder (TVe) + found on i.MX53 processors. + +config DRM_IMX_LDB + tristate "Support for LVDS displays" + depends on DRM_IMX && MFD_SYSCON + help + Choose this to enable the internal LVDS Display Bridge (LDB) + found on i.MX53 and i.MX6 processors. + +config DRM_IMX_IPUV3 + tristate "DRM Support for i.MX IPUv3" + depends on DRM_IMX + depends on IMX_IPUV3_CORE + help + Choose this if you have a i.MX5 or i.MX6 processor. + +config DRM_IMX_HDMI + tristate "Freescale i.MX DRM HDMI" + depends on DRM_IMX + help + Choose this if you want to use HDMI on i.MX6. diff --git a/drivers/gpu/drm/imx/Makefile b/drivers/gpu/drm/imx/Makefile new file mode 100644 index 0000000..582c438 --- /dev/null +++ b/drivers/gpu/drm/imx/Makefile @@ -0,0 +1,12 @@ + +imxdrm-objs := imx-drm-core.o + +obj-$(CONFIG_DRM_IMX) += imxdrm.o + +obj-$(CONFIG_DRM_IMX_PARALLEL_DISPLAY) += parallel-display.o +obj-$(CONFIG_DRM_IMX_TVE) += imx-tve.o +obj-$(CONFIG_DRM_IMX_LDB) += imx-ldb.o + +imx-ipuv3-crtc-objs := ipuv3-crtc.o ipuv3-plane.o +obj-$(CONFIG_DRM_IMX_IPUV3) += imx-ipuv3-crtc.o +obj-$(CONFIG_DRM_IMX_HDMI) += imx-hdmi.o diff --git a/drivers/gpu/drm/imx/imx-drm-core.c b/drivers/gpu/drm/imx/imx-drm-core.c new file mode 100644 index 0000000..2f80072 --- /dev/null +++ b/drivers/gpu/drm/imx/imx-drm-core.c @@ -0,0 +1,705 @@ +/* + * Freescale i.MX drm driver + * + * Copyright (C) 2011 Sascha Hauer, Pengutronix + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "imx-drm.h" + +#define MAX_CRTC 4 + +struct imx_drm_crtc; + +struct imx_drm_component { + struct device_node *of_node; + struct list_head list; +}; + +struct imx_drm_device { + struct drm_device *drm; + struct imx_drm_crtc *crtc[MAX_CRTC]; + int pipes; + struct drm_fbdev_cma *fbhelper; +}; + +struct imx_drm_crtc { + struct drm_crtc *crtc; + int pipe; + struct imx_drm_crtc_helper_funcs imx_drm_helper_funcs; + struct device_node *port; +}; + +static int legacyfb_depth = 16; +module_param(legacyfb_depth, int, 0444); + +int imx_drm_crtc_id(struct imx_drm_crtc *crtc) +{ + return crtc->pipe; +} +EXPORT_SYMBOL_GPL(imx_drm_crtc_id); + +static void imx_drm_driver_lastclose(struct drm_device *drm) +{ +#if IS_ENABLED(CONFIG_DRM_IMX_FB_HELPER) + struct imx_drm_device *imxdrm = drm->dev_private; + + if (imxdrm->fbhelper) + drm_fbdev_cma_restore_mode(imxdrm->fbhelper); +#endif +} + +static int imx_drm_driver_unload(struct drm_device *drm) +{ +#if IS_ENABLED(CONFIG_DRM_IMX_FB_HELPER) + struct imx_drm_device *imxdrm = drm->dev_private; +#endif + + drm_kms_helper_poll_fini(drm); + +#if IS_ENABLED(CONFIG_DRM_IMX_FB_HELPER) + if (imxdrm->fbhelper) + drm_fbdev_cma_fini(imxdrm->fbhelper); +#endif + + component_unbind_all(drm->dev, drm); + + drm_vblank_cleanup(drm); + drm_mode_config_cleanup(drm); + + platform_set_drvdata(drm->platformdev, NULL); + + return 0; +} + +static struct imx_drm_crtc *imx_drm_find_crtc(struct drm_crtc *crtc) +{ + struct imx_drm_device *imxdrm = crtc->dev->dev_private; + unsigned i; + + for (i = 0; i < MAX_CRTC; i++) + if (imxdrm->crtc[i] && imxdrm->crtc[i]->crtc == crtc) + return imxdrm->crtc[i]; + + return NULL; +} + +int imx_drm_panel_format_pins(struct drm_encoder *encoder, + u32 interface_pix_fmt, int hsync_pin, int vsync_pin) +{ + struct imx_drm_crtc_helper_funcs *helper; + struct imx_drm_crtc *imx_crtc; + + imx_crtc = imx_drm_find_crtc(encoder->crtc); + if (!imx_crtc) + return -EINVAL; + + helper = &imx_crtc->imx_drm_helper_funcs; + if (helper->set_interface_pix_fmt) + return helper->set_interface_pix_fmt(encoder->crtc, + encoder->encoder_type, interface_pix_fmt, + hsync_pin, vsync_pin); + return 0; +} +EXPORT_SYMBOL_GPL(imx_drm_panel_format_pins); + +int imx_drm_panel_format(struct drm_encoder *encoder, u32 interface_pix_fmt) +{ + return imx_drm_panel_format_pins(encoder, interface_pix_fmt, 2, 3); +} +EXPORT_SYMBOL_GPL(imx_drm_panel_format); + +int imx_drm_crtc_vblank_get(struct imx_drm_crtc *imx_drm_crtc) +{ + return drm_vblank_get(imx_drm_crtc->crtc->dev, imx_drm_crtc->pipe); +} +EXPORT_SYMBOL_GPL(imx_drm_crtc_vblank_get); + +void imx_drm_crtc_vblank_put(struct imx_drm_crtc *imx_drm_crtc) +{ + drm_vblank_put(imx_drm_crtc->crtc->dev, imx_drm_crtc->pipe); +} +EXPORT_SYMBOL_GPL(imx_drm_crtc_vblank_put); + +void imx_drm_handle_vblank(struct imx_drm_crtc *imx_drm_crtc) +{ + drm_handle_vblank(imx_drm_crtc->crtc->dev, imx_drm_crtc->pipe); +} +EXPORT_SYMBOL_GPL(imx_drm_handle_vblank); + +static int imx_drm_enable_vblank(struct drm_device *drm, int crtc) +{ + struct imx_drm_device *imxdrm = drm->dev_private; + struct imx_drm_crtc *imx_drm_crtc = imxdrm->crtc[crtc]; + int ret; + + if (!imx_drm_crtc) + return -EINVAL; + + if (!imx_drm_crtc->imx_drm_helper_funcs.enable_vblank) + return -ENOSYS; + + ret = imx_drm_crtc->imx_drm_helper_funcs.enable_vblank( + imx_drm_crtc->crtc); + + return ret; +} + +static void imx_drm_disable_vblank(struct drm_device *drm, int crtc) +{ + struct imx_drm_device *imxdrm = drm->dev_private; + struct imx_drm_crtc *imx_drm_crtc = imxdrm->crtc[crtc]; + + if (!imx_drm_crtc) + return; + + if (!imx_drm_crtc->imx_drm_helper_funcs.disable_vblank) + return; + + imx_drm_crtc->imx_drm_helper_funcs.disable_vblank(imx_drm_crtc->crtc); +} + +static void imx_drm_driver_preclose(struct drm_device *drm, + struct drm_file *file) +{ + int i; + + if (!file->is_master) + return; + + for (i = 0; i < MAX_CRTC; i++) + imx_drm_disable_vblank(drm, i); +} + +static const struct file_operations imx_drm_driver_fops = { + .owner = THIS_MODULE, + .open = drm_open, + .release = drm_release, + .unlocked_ioctl = drm_ioctl, + .mmap = drm_gem_cma_mmap, + .poll = drm_poll, + .read = drm_read, + .llseek = noop_llseek, +}; + +void imx_drm_connector_destroy(struct drm_connector *connector) +{ + drm_connector_unregister(connector); + drm_connector_cleanup(connector); +} +EXPORT_SYMBOL_GPL(imx_drm_connector_destroy); + +void imx_drm_encoder_destroy(struct drm_encoder *encoder) +{ + drm_encoder_cleanup(encoder); +} +EXPORT_SYMBOL_GPL(imx_drm_encoder_destroy); + +static void imx_drm_output_poll_changed(struct drm_device *drm) +{ +#if IS_ENABLED(CONFIG_DRM_IMX_FB_HELPER) + struct imx_drm_device *imxdrm = drm->dev_private; + + drm_fbdev_cma_hotplug_event(imxdrm->fbhelper); +#endif +} + +static struct drm_mode_config_funcs imx_drm_mode_config_funcs = { + .fb_create = drm_fb_cma_create, + .output_poll_changed = imx_drm_output_poll_changed, +}; + +/* + * Main DRM initialisation. This binds, initialises and registers + * with DRM the subcomponents of the driver. + */ +static int imx_drm_driver_load(struct drm_device *drm, unsigned long flags) +{ + struct imx_drm_device *imxdrm; + struct drm_connector *connector; + int ret; + + imxdrm = devm_kzalloc(drm->dev, sizeof(*imxdrm), GFP_KERNEL); + if (!imxdrm) + return -ENOMEM; + + imxdrm->drm = drm; + + drm->dev_private = imxdrm; + + /* + * enable drm irq mode. + * - with irq_enabled = true, we can use the vblank feature. + * + * P.S. note that we wouldn't use drm irq handler but + * just specific driver own one instead because + * drm framework supports only one irq handler and + * drivers can well take care of their interrupts + */ + drm->irq_enabled = true; + + /* + * set max width and height as default value(4096x4096). + * this value would be used to check framebuffer size limitation + * at drm_mode_addfb(). + */ + drm->mode_config.min_width = 64; + drm->mode_config.min_height = 64; + drm->mode_config.max_width = 4096; + drm->mode_config.max_height = 4096; + drm->mode_config.funcs = &imx_drm_mode_config_funcs; + + drm_mode_config_init(drm); + + ret = drm_vblank_init(drm, MAX_CRTC); + if (ret) + goto err_kms; + + /* + * with vblank_disable_allowed = true, vblank interrupt will be + * disabled by drm timer once a current process gives up ownership + * of vblank event. (after drm_vblank_put function is called) + */ + drm->vblank_disable_allowed = true; + + platform_set_drvdata(drm->platformdev, drm); + + /* Now try and bind all our sub-components */ + ret = component_bind_all(drm->dev, drm); + if (ret) + goto err_vblank; + + /* + * All components are now added, we can publish the connector sysfs + * entries to userspace. This will generate hotplug events and so + * userspace will expect to be able to access DRM at this point. + */ + list_for_each_entry(connector, &drm->mode_config.connector_list, head) { + ret = drm_connector_register(connector); + if (ret) { + dev_err(drm->dev, + "[CONNECTOR:%d:%s] drm_connector_register failed: %d\n", + connector->base.id, + connector->name, ret); + goto err_unbind; + } + } + + /* + * All components are now initialised, so setup the fb helper. + * The fb helper takes copies of key hardware information, so the + * crtcs/connectors/encoders must not change after this point. + */ +#if IS_ENABLED(CONFIG_DRM_IMX_FB_HELPER) + if (legacyfb_depth != 16 && legacyfb_depth != 32) { + dev_warn(drm->dev, "Invalid legacyfb_depth. Defaulting to 16bpp\n"); + legacyfb_depth = 16; + } + imxdrm->fbhelper = drm_fbdev_cma_init(drm, legacyfb_depth, + drm->mode_config.num_crtc, MAX_CRTC); + if (IS_ERR(imxdrm->fbhelper)) { + ret = PTR_ERR(imxdrm->fbhelper); + imxdrm->fbhelper = NULL; + goto err_unbind; + } +#endif + + drm_kms_helper_poll_init(drm); + + return 0; + +err_unbind: + component_unbind_all(drm->dev, drm); +err_vblank: + drm_vblank_cleanup(drm); +err_kms: + drm_mode_config_cleanup(drm); + + return ret; +} + +/* + * imx_drm_add_crtc - add a new crtc + */ +int imx_drm_add_crtc(struct drm_device *drm, struct drm_crtc *crtc, + struct imx_drm_crtc **new_crtc, + const struct imx_drm_crtc_helper_funcs *imx_drm_helper_funcs, + struct device_node *port) +{ + struct imx_drm_device *imxdrm = drm->dev_private; + struct imx_drm_crtc *imx_drm_crtc; + int ret; + + /* + * The vblank arrays are dimensioned by MAX_CRTC - we can't + * pass IDs greater than this to those functions. + */ + if (imxdrm->pipes >= MAX_CRTC) + return -EINVAL; + + if (imxdrm->drm->open_count) + return -EBUSY; + + imx_drm_crtc = kzalloc(sizeof(*imx_drm_crtc), GFP_KERNEL); + if (!imx_drm_crtc) + return -ENOMEM; + + imx_drm_crtc->imx_drm_helper_funcs = *imx_drm_helper_funcs; + imx_drm_crtc->pipe = imxdrm->pipes++; + imx_drm_crtc->port = port; + imx_drm_crtc->crtc = crtc; + + imxdrm->crtc[imx_drm_crtc->pipe] = imx_drm_crtc; + + *new_crtc = imx_drm_crtc; + + ret = drm_mode_crtc_set_gamma_size(imx_drm_crtc->crtc, 256); + if (ret) + goto err_register; + + drm_crtc_helper_add(crtc, + imx_drm_crtc->imx_drm_helper_funcs.crtc_helper_funcs); + + drm_crtc_init(drm, crtc, + imx_drm_crtc->imx_drm_helper_funcs.crtc_funcs); + + return 0; + +err_register: + imxdrm->crtc[imx_drm_crtc->pipe] = NULL; + kfree(imx_drm_crtc); + return ret; +} +EXPORT_SYMBOL_GPL(imx_drm_add_crtc); + +/* + * imx_drm_remove_crtc - remove a crtc + */ +int imx_drm_remove_crtc(struct imx_drm_crtc *imx_drm_crtc) +{ + struct imx_drm_device *imxdrm = imx_drm_crtc->crtc->dev->dev_private; + + drm_crtc_cleanup(imx_drm_crtc->crtc); + + imxdrm->crtc[imx_drm_crtc->pipe] = NULL; + + kfree(imx_drm_crtc); + + return 0; +} +EXPORT_SYMBOL_GPL(imx_drm_remove_crtc); + +/* + * Find the DRM CRTC possible mask for the connected endpoint. + * + * The encoder possible masks are defined by their position in the + * mode_config crtc_list. This means that CRTCs must not be added + * or removed once the DRM device has been fully initialised. + */ +static uint32_t imx_drm_find_crtc_mask(struct imx_drm_device *imxdrm, + struct device_node *endpoint) +{ + struct device_node *port; + unsigned i; + + port = of_graph_get_remote_port(endpoint); + if (!port) + return 0; + of_node_put(port); + + for (i = 0; i < MAX_CRTC; i++) { + struct imx_drm_crtc *imx_drm_crtc = imxdrm->crtc[i]; + + if (imx_drm_crtc && imx_drm_crtc->port == port) + return drm_crtc_mask(imx_drm_crtc->crtc); + } + + return 0; +} + +static struct device_node *imx_drm_of_get_next_endpoint( + const struct device_node *parent, struct device_node *prev) +{ + struct device_node *node = of_graph_get_next_endpoint(parent, prev); + + of_node_put(prev); + return node; +} + +int imx_drm_encoder_parse_of(struct drm_device *drm, + struct drm_encoder *encoder, struct device_node *np) +{ + struct imx_drm_device *imxdrm = drm->dev_private; + struct device_node *ep = NULL; + uint32_t crtc_mask = 0; + int i; + + for (i = 0; ; i++) { + u32 mask; + + ep = imx_drm_of_get_next_endpoint(np, ep); + if (!ep) + break; + + mask = imx_drm_find_crtc_mask(imxdrm, ep); + + /* + * If we failed to find the CRTC(s) which this encoder is + * supposed to be connected to, it's because the CRTC has + * not been registered yet. Defer probing, and hope that + * the required CRTC is added later. + */ + if (mask == 0) + return -EPROBE_DEFER; + + crtc_mask |= mask; + } + + of_node_put(ep); + if (i == 0) + return -ENOENT; + + encoder->possible_crtcs = crtc_mask; + + /* FIXME: this is the mask of outputs which can clone this output. */ + encoder->possible_clones = ~0; + + return 0; +} +EXPORT_SYMBOL_GPL(imx_drm_encoder_parse_of); + +/* + * @node: device tree node containing encoder input ports + * @encoder: drm_encoder + */ +int imx_drm_encoder_get_mux_id(struct device_node *node, + struct drm_encoder *encoder) +{ + struct imx_drm_crtc *imx_crtc = imx_drm_find_crtc(encoder->crtc); + struct device_node *ep = NULL; + struct of_endpoint endpoint; + struct device_node *port; + int ret; + + if (!node || !imx_crtc) + return -EINVAL; + + do { + ep = imx_drm_of_get_next_endpoint(node, ep); + if (!ep) + break; + + port = of_graph_get_remote_port(ep); + of_node_put(port); + if (port == imx_crtc->port) { + ret = of_graph_parse_endpoint(ep, &endpoint); + return ret ? ret : endpoint.port; + } + } while (ep); + + return -EINVAL; +} +EXPORT_SYMBOL_GPL(imx_drm_encoder_get_mux_id); + +static const struct drm_ioctl_desc imx_drm_ioctls[] = { + /* none so far */ +}; + +static struct drm_driver imx_drm_driver = { + .driver_features = DRIVER_MODESET | DRIVER_GEM | DRIVER_PRIME, + .load = imx_drm_driver_load, + .unload = imx_drm_driver_unload, + .lastclose = imx_drm_driver_lastclose, + .preclose = imx_drm_driver_preclose, + .set_busid = drm_platform_set_busid, + .gem_free_object = drm_gem_cma_free_object, + .gem_vm_ops = &drm_gem_cma_vm_ops, + .dumb_create = drm_gem_cma_dumb_create, + .dumb_map_offset = drm_gem_cma_dumb_map_offset, + .dumb_destroy = drm_gem_dumb_destroy, + + .prime_handle_to_fd = drm_gem_prime_handle_to_fd, + .prime_fd_to_handle = drm_gem_prime_fd_to_handle, + .gem_prime_import = drm_gem_prime_import, + .gem_prime_export = drm_gem_prime_export, + .gem_prime_get_sg_table = drm_gem_cma_prime_get_sg_table, + .gem_prime_import_sg_table = drm_gem_cma_prime_import_sg_table, + .gem_prime_vmap = drm_gem_cma_prime_vmap, + .gem_prime_vunmap = drm_gem_cma_prime_vunmap, + .gem_prime_mmap = drm_gem_cma_prime_mmap, + .get_vblank_counter = drm_vblank_count, + .enable_vblank = imx_drm_enable_vblank, + .disable_vblank = imx_drm_disable_vblank, + .ioctls = imx_drm_ioctls, + .num_ioctls = ARRAY_SIZE(imx_drm_ioctls), + .fops = &imx_drm_driver_fops, + .name = "imx-drm", + .desc = "i.MX DRM graphics", + .date = "20120507", + .major = 1, + .minor = 0, + .patchlevel = 0, +}; + +static int compare_of(struct device *dev, void *data) +{ + struct device_node *np = data; + + /* Special case for LDB, one device for two channels */ + if (of_node_cmp(np->name, "lvds-channel") == 0) { + np = of_get_parent(np); + of_node_put(np); + } + + return dev->of_node == np; +} + +static int imx_drm_bind(struct device *dev) +{ + return drm_platform_init(&imx_drm_driver, to_platform_device(dev)); +} + +static void imx_drm_unbind(struct device *dev) +{ + drm_put_dev(dev_get_drvdata(dev)); +} + +static const struct component_master_ops imx_drm_ops = { + .bind = imx_drm_bind, + .unbind = imx_drm_unbind, +}; + +static int imx_drm_platform_probe(struct platform_device *pdev) +{ + struct device_node *ep, *port, *remote; + struct component_match *match = NULL; + int ret; + int i; + + /* + * Bind the IPU display interface ports first, so that + * imx_drm_encoder_parse_of called from encoder .bind callbacks + * works as expected. + */ + for (i = 0; ; i++) { + port = of_parse_phandle(pdev->dev.of_node, "ports", i); + if (!port) + break; + + component_match_add(&pdev->dev, &match, compare_of, port); + } + + if (i == 0) { + dev_err(&pdev->dev, "missing 'ports' property\n"); + return -ENODEV; + } + + /* Then bind all encoders */ + for (i = 0; ; i++) { + port = of_parse_phandle(pdev->dev.of_node, "ports", i); + if (!port) + break; + + for_each_child_of_node(port, ep) { + remote = of_graph_get_remote_port_parent(ep); + if (!remote || !of_device_is_available(remote)) { + of_node_put(remote); + continue; + } else if (!of_device_is_available(remote->parent)) { + dev_warn(&pdev->dev, "parent device of %s is not available\n", + remote->full_name); + of_node_put(remote); + continue; + } + + component_match_add(&pdev->dev, &match, compare_of, remote); + of_node_put(remote); + } + of_node_put(port); + } + + ret = dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(32)); + if (ret) + return ret; + + return component_master_add_with_match(&pdev->dev, &imx_drm_ops, match); +} + +static int imx_drm_platform_remove(struct platform_device *pdev) +{ + component_master_del(&pdev->dev, &imx_drm_ops); + return 0; +} + +#ifdef CONFIG_PM_SLEEP +static int imx_drm_suspend(struct device *dev) +{ + struct drm_device *drm_dev = dev_get_drvdata(dev); + + /* The drm_dev is NULL before .load hook is called */ + if (drm_dev == NULL) + return 0; + + drm_kms_helper_poll_disable(drm_dev); + + return 0; +} + +static int imx_drm_resume(struct device *dev) +{ + struct drm_device *drm_dev = dev_get_drvdata(dev); + + if (drm_dev == NULL) + return 0; + + drm_helper_resume_force_mode(drm_dev); + drm_kms_helper_poll_enable(drm_dev); + + return 0; +} +#endif + +static SIMPLE_DEV_PM_OPS(imx_drm_pm_ops, imx_drm_suspend, imx_drm_resume); + +static const struct of_device_id imx_drm_dt_ids[] = { + { .compatible = "fsl,imx-display-subsystem", }, + { /* sentinel */ }, +}; +MODULE_DEVICE_TABLE(of, imx_drm_dt_ids); + +static struct platform_driver imx_drm_pdrv = { + .probe = imx_drm_platform_probe, + .remove = imx_drm_platform_remove, + .driver = { + .owner = THIS_MODULE, + .name = "imx-drm", + .pm = &imx_drm_pm_ops, + .of_match_table = imx_drm_dt_ids, + }, +}; +module_platform_driver(imx_drm_pdrv); + +MODULE_AUTHOR("Sascha Hauer "); +MODULE_DESCRIPTION("i.MX drm driver core"); +MODULE_LICENSE("GPL"); diff --git a/drivers/gpu/drm/imx/imx-drm.h b/drivers/gpu/drm/imx/imx-drm.h new file mode 100644 index 0000000..7453ae0 --- /dev/null +++ b/drivers/gpu/drm/imx/imx-drm.h @@ -0,0 +1,56 @@ +#ifndef _IMX_DRM_H_ +#define _IMX_DRM_H_ + +struct device_node; +struct drm_crtc; +struct drm_connector; +struct drm_device; +struct drm_display_mode; +struct drm_encoder; +struct drm_fbdev_cma; +struct drm_framebuffer; +struct imx_drm_crtc; +struct platform_device; + +int imx_drm_crtc_id(struct imx_drm_crtc *crtc); + +struct imx_drm_crtc_helper_funcs { + int (*enable_vblank)(struct drm_crtc *crtc); + void (*disable_vblank)(struct drm_crtc *crtc); + int (*set_interface_pix_fmt)(struct drm_crtc *crtc, u32 encoder_type, + u32 pix_fmt, int hsync_pin, int vsync_pin); + const struct drm_crtc_helper_funcs *crtc_helper_funcs; + const struct drm_crtc_funcs *crtc_funcs; +}; + +int imx_drm_add_crtc(struct drm_device *drm, struct drm_crtc *crtc, + struct imx_drm_crtc **new_crtc, + const struct imx_drm_crtc_helper_funcs *imx_helper_funcs, + struct device_node *port); +int imx_drm_remove_crtc(struct imx_drm_crtc *); +int imx_drm_init_drm(struct platform_device *pdev, + int preferred_bpp); +int imx_drm_exit_drm(void); + +int imx_drm_crtc_vblank_get(struct imx_drm_crtc *imx_drm_crtc); +void imx_drm_crtc_vblank_put(struct imx_drm_crtc *imx_drm_crtc); +void imx_drm_handle_vblank(struct imx_drm_crtc *imx_drm_crtc); + +void imx_drm_mode_config_init(struct drm_device *drm); + +struct drm_gem_cma_object *imx_drm_fb_get_obj(struct drm_framebuffer *fb); + +int imx_drm_panel_format_pins(struct drm_encoder *encoder, + u32 interface_pix_fmt, int hsync_pin, int vsync_pin); +int imx_drm_panel_format(struct drm_encoder *encoder, + u32 interface_pix_fmt); + +int imx_drm_encoder_get_mux_id(struct device_node *node, + struct drm_encoder *encoder); +int imx_drm_encoder_parse_of(struct drm_device *drm, + struct drm_encoder *encoder, struct device_node *np); + +void imx_drm_connector_destroy(struct drm_connector *connector); +void imx_drm_encoder_destroy(struct drm_encoder *encoder); + +#endif /* _IMX_DRM_H_ */ diff --git a/drivers/gpu/drm/imx/imx-hdmi.c b/drivers/gpu/drm/imx/imx-hdmi.c new file mode 100644 index 0000000..aaec6b2 --- /dev/null +++ b/drivers/gpu/drm/imx/imx-hdmi.c @@ -0,0 +1,1767 @@ +/* + * Copyright (C) 2011-2013 Freescale Semiconductor, Inc. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * SH-Mobile High-Definition Multimedia Interface (HDMI) driver + * for SLISHDMI13T and SLIPHDMIT IP cores + * + * Copyright (C) 2010, Guennadi Liakhovetski + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include - + -- cgit v0.10.2 From a550bb3d5313c3d894cac552e13037394ac78fd4 Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Thu, 4 Dec 2014 14:09:02 +0200 Subject: amdkfd: Set *buffer_ptr to NULL in case of error In function acquire_packet_buffer() we may return -ENOMEM. In that case, we should set the *buffer_ptr to NULL, so that calling functions which check the *buffer_ptr value as a criteria for success, will know that acquire_packet_buffer() failed. Reviewed-by: Alexey Skidanov Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c index 9abac48..9350714 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -221,8 +221,14 @@ static int acquire_packet_buffer(struct kernel_queue *kq, queue_size_dwords; if (packet_size_in_dwords >= queue_size_dwords || - packet_size_in_dwords >= available_size) + packet_size_in_dwords >= available_size) { + /* + * make sure calling functions know + * acquire_packet_buffer() failed + */ + *buffer_ptr = NULL; return -ENOMEM; + } if (wptr + packet_size_in_dwords >= queue_size_dwords) { while (wptr > 0) { -- cgit v0.10.2 From a0a0bde32a4668656cdffac677032a101c170c6b Mon Sep 17 00:00:00 2001 From: Takashi Iwai Date: Thu, 4 Dec 2014 11:56:42 +0100 Subject: drm: Fix memory leak at error path of drm_read() Note that the read manpages explicitly states that the read position is undefined on error. Since EFAULT is just a userspace bug we are therefore fine with just dropping the event on the floor. Signed-off-by: Takashi Iwai [danvet: Add note that just dropping the event is ok.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c index 91e1105..0b9514b 100644 --- a/drivers/gpu/drm/drm_fops.c +++ b/drivers/gpu/drm/drm_fops.c @@ -527,6 +527,7 @@ ssize_t drm_read(struct file *filp, char __user *buffer, if (copy_to_user(buffer + total, e->event, e->event->length)) { total = -EFAULT; + e->destroy(e); break; } -- cgit v0.10.2 From af15d2ce5d497ba6535dead782cfadbcc83d90fe Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Mon, 1 Dec 2014 09:54:28 -0800 Subject: drm/i915: don't always do full mode sets when infoframes are enabled Partial revert of commit 206645910b9796bff13fcdb67bdca166b724ba62 Author: Jesse Barnes Date: Wed Nov 5 14:26:09 2014 -0800 drm/i915: check for audio and infoframe changes across mode sets v2 References: https://bugs.freedesktop.org/show_bug.cgi?id=86683 Signed-off-by: Jesse Barnes Tested-by: Li Xu Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 6289bab..03d0b0c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11583,10 +11583,12 @@ static int intel_crtc_set_config(struct drm_mode_set *set) to_intel_crtc(set->crtc)->config.has_audio) config->mode_changed = true; - /* Force mode sets for any infoframe stuff */ - if (pipe_config->has_infoframe || - to_intel_crtc(set->crtc)->config.has_infoframe) - config->mode_changed = true; + /* + * Note we have an issue here with infoframes: current code + * only updates them on the full mode set path per hw + * requirements. So here we should be checking for any + * required changes and forcing a mode set. + */ } /* set_mode will free it in the mode_changed case */ -- cgit v0.10.2 From 7bd0e226e313547add65c7e4b0ecf65cb981a3c6 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Thu, 4 Dec 2014 11:12:54 +0100 Subject: drm/i915: compute wait_ioctl timeout correctly We've lost the +1 required for correct timeouts in commit 5ed0bdf21a85d78e04f89f15ccf227562177cbd9 Author: Thomas Gleixner Date: Wed Jul 16 21:05:06 2014 +0000 drm: i915: Use nsec based interfaces Use ktime_get_raw_ns() and get rid of the back and forth timespec conversions. Signed-off-by: Thomas Gleixner Acked-by: Daniel Vetter Signed-off-by: John Stultz So fix this up by reinstating our handrolled _timeout function. While at it bother with handling MAX_JIFFIES. v2: Convert to usecs (we don't care about the accuracy anyway) first to avoid overflow issues Dave Gordon spotted. v3: Drop the explicit MAX_JIFFY_OFFSET check, usecs_to_jiffies should take care of that already. It might be a bit too enthusiastic about it though. v4: Chris has a much nicer color, so use his implementation. This requires to export nsec_to_jiffies from time.c. Cc: Chris Wilson Cc: Dave Gordon Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82749 Cc: Thomas Gleixner Cc: John Stultz Signed-off-by: Daniel Vetter Acked-by: John Stultz Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index bb1892d..be910d2 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3032,6 +3032,11 @@ static inline unsigned long msecs_to_jiffies_timeout(const unsigned int m) return min_t(unsigned long, MAX_JIFFY_OFFSET, j + 1); } +static inline unsigned long nsecs_to_jiffies_timeout(const u64 n) +{ + return min_t(u64, MAX_JIFFY_OFFSET, nsecs_to_jiffies64(n) + 1); +} + static inline unsigned long timespec_to_jiffies_timeout(const struct timespec *value) { diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index d2ba315..820bc98 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1227,7 +1227,8 @@ int __i915_wait_seqno(struct intel_engine_cs *ring, u32 seqno, if (i915_seqno_passed(ring->get_seqno(ring, true), seqno)) return 0; - timeout_expire = timeout ? jiffies + nsecs_to_jiffies((u64)*timeout) : 0; + timeout_expire = timeout ? + jiffies + nsecs_to_jiffies_timeout((u64)*timeout) : 0; if (INTEL_INFO(dev)->gen >= 6 && ring->id == RCS && can_wait_boost(file_priv)) { gen6_rps_boost(dev_priv); diff --git a/kernel/time/time.c b/kernel/time/time.c index a9ae20f..8fae82c 100644 --- a/kernel/time/time.c +++ b/kernel/time/time.c @@ -745,6 +745,7 @@ u64 nsecs_to_jiffies64(u64 n) return div_u64(n * 9, (9ull * NSEC_PER_SEC + HZ / 2) / HZ); #endif } +EXPORT_SYMBOL(nsecs_to_jiffies64); /** * nsecs_to_jiffies - Convert nsecs in u64 to jiffies -- cgit v0.10.2 From 9cca30688043eb1b6569f794041dfeec218dffa5 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 28 Nov 2014 10:29:55 +0100 Subject: drm/i915: Handle inaccurate time conversion issues So apparently jiffies<->nsec<->ktime isn't accurate or something. At elast if we timeout there's occasionally still a few hundred us left (in a 2 second timeout). Stuff I've tried and thrown out again: - Sampling the before timestamp before jiffies. Doesn't improve test path rate at all. - Using jiffies. Way to inaccurate, which means way too much drift with signals plus automatic ioctl restarting in userspace. In hindsight we should have used an absolute timeout, but hey we need something for v3 of the i915 gem wait interfaces ;-) - Trying to figure out where accuracy gets lost. gl testcase really don't care all that much about this (as long as isn't not massively off), it's just that the testcase gets a bit upset if it receives an EITME with timeout > 0. So as long as we're in the ballbark it's good enough. So patch everything up if we're at most one jiffies off. I get's me a solid test again. This regression is probably introduced in commit 5ed0bdf21a85d78e04f89f15ccf227562177cbd9 Author: Thomas Gleixner Date: Wed Jul 16 21:05:06 2014 +0000 drm: i915: Use nsec based interfaces Use ktime_get_raw_ns() and get rid of the back and forth timespec conversions. Signed-off-by: Thomas Gleixner Acked-by: Daniel Vetter Signed-off-by: John Stultz Probably because I'm too lazy to confirm myself and still waiting for QA ;-) Cc: Thomas Gleixner Cc: John Stultz Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82749 Signed-off-by: Daniel Vetter Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 820bc98..4a9faea6 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1304,6 +1304,16 @@ int __i915_wait_seqno(struct intel_engine_cs *ring, u32 seqno, s64 tres = *timeout - (now - before); *timeout = tres < 0 ? 0 : tres; + + /* + * Apparently ktime isn't accurate enough and occasionally has a + * bit of mismatch in the jiffies<->nsecs<->ktime loop. So patch + * things up to make the test happy. We allow up to 1 jiffy. + * + * This is a regrssion from the timespec->ktime conversion. + */ + if (ret == -ETIME && *timeout < jiffies_to_usecs(1)*1000) + *timeout = 0; } return ret; -- cgit v0.10.2 From a18069c132cb0d065ff23c05977d06ea0c78404f Mon Sep 17 00:00:00 2001 From: Oded Gabbay Date: Fri, 5 Dec 2014 10:40:34 +0200 Subject: amdkfd: Disable support for 32-bit user processes This patch checks if the process that opens the /dev/kfd device is 32-bit process. If so, it returns -EPERM and prints a warning message in dmesg. This is done to prevent 32-bit user processes from using amdkfd, and hence, HSA features. AMD's HSA userspace stack will also support only 64-bit processes on Linux. Reviewed-by: Alexey Skidanov Signed-off-by: Oded Gabbay diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 102cd36..4f7b275 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -102,15 +102,26 @@ struct device *kfd_chardev(void) static int kfd_open(struct inode *inode, struct file *filep) { struct kfd_process *process; + bool is_32bit_user_mode; if (iminor(inode) != 0) return -ENODEV; + is_32bit_user_mode = is_compat_task(); + + if (is_32bit_user_mode == true) { + dev_warn(kfd_device, + "Process %d (32-bit) failed to open /dev/kfd\n" + "32-bit processes are not supported by amdkfd\n", + current->pid); + return -EPERM; + } + process = kfd_create_process(current); if (IS_ERR(process)) return PTR_ERR(process); - process->is_32bit_user_mode = is_compat_task(); + process->is_32bit_user_mode = is_32bit_user_mode; dev_dbg(kfd_device, "process %d opened, compat mode (32 bit) - %d\n", process->pasid, process->is_32bit_user_mode); -- cgit v0.10.2 From e7d6f7d708290da1b7c92f533444b042c79412e0 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 8 Dec 2014 13:23:37 +1000 Subject: drm/i915: resume MST after reading back hw state Otherwise the MST resume paths can hit DPMS paths which hit state checker paths, which hit WARN_ON, because the state checker is inconsistent with the hw. This fixes a bunch of WARN_ON's on resume after undocking. Signed-off-by: Dave Airlie Reviewed-by: Daniel Vetter Cc: stable@vger.kernel.org Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 1e9c136..f990ab4 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -706,11 +706,12 @@ static int i915_drm_resume(struct drm_device *dev) dev_priv->display.hpd_irq_setup(dev); spin_unlock_irq(&dev_priv->irq_lock); - intel_dp_mst_resume(dev); drm_modeset_lock_all(dev); intel_modeset_setup_hw_state(dev, true); drm_modeset_unlock_all(dev); + intel_dp_mst_resume(dev); + /* * ... but also need to make sure that hotplug processing * doesn't cause havoc. Like in the driver load code we don't -- cgit v0.10.2 From 19a93f042fc241ecdf98543cedfe7c171f8cdf53 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 26 Nov 2014 13:13:09 +1000 Subject: drm/dp: retry AUX transactions 32 times (v1.1) At least on two MST devices I've tested with, when they are link training downstream, they are totally unable to handle aux ch msgs, so they defer like nuts. I tried 16, it wasn't enough, 32 seems better. This fixes one Dell 4k monitor and one of the MST hubs. v1.1: fixup comment (Tom). Acked-by: Alex Deucher Cc: stable@vger.kernel.org Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_dp_helper.c b/drivers/gpu/drm/drm_dp_helper.c index 959e207..79968e3 100644 --- a/drivers/gpu/drm/drm_dp_helper.c +++ b/drivers/gpu/drm/drm_dp_helper.c @@ -186,10 +186,11 @@ static int drm_dp_dpcd_access(struct drm_dp_aux *aux, u8 request, /* * The specification doesn't give any recommendation on how often to - * retry native transactions, so retry 7 times like for I2C-over-AUX - * transactions. + * retry native transactions. We used to retry 7 times like for + * aux i2c transactions but real world devices this wasn't + * sufficient, bump to 32 which makes Dell 4k monitors happier. */ - for (retry = 0; retry < 7; retry++) { + for (retry = 0; retry < 32; retry++) { mutex_lock(&aux->hw_mutex); err = aux->transfer(aux, &msg); -- cgit v0.10.2 From e2809c7db818df6bbd0edf843e1beb2fbc9d8541 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 26 Nov 2014 13:15:24 +1000 Subject: drm/fb_helper: move deferred fb checking into restore mode (v2) On MST systems the monitors don't appear when we set the fb up, but plymouth opens the drm device and holds it open while they come up, when plymouth finishes and lastclose gets called we don't do the delayed fb probe, so the monitor never appears on the console. Fix this by moving the delayed checking into the mode restore. v2: Daniel suggested that ->delayed_hotplug is set under the mode_config mutex, so we should check it under that as well, while we are in the area. Reviewed-by: Daniel Vetter Cc: stable@vger.kernel.org Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 09d47e9..a467460 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -347,9 +347,18 @@ bool drm_fb_helper_restore_fbdev_mode_unlocked(struct drm_fb_helper *fb_helper) { struct drm_device *dev = fb_helper->dev; bool ret; + bool do_delayed = false; + drm_modeset_lock_all(dev); ret = restore_fbdev_mode(fb_helper); + + do_delayed = fb_helper->delayed_hotplug; + if (do_delayed) + fb_helper->delayed_hotplug = false; drm_modeset_unlock_all(dev); + + if (do_delayed) + drm_fb_helper_hotplug_event(fb_helper); return ret; } EXPORT_SYMBOL(drm_fb_helper_restore_fbdev_mode_unlocked); @@ -888,10 +897,6 @@ int drm_fb_helper_set_par(struct fb_info *info) drm_fb_helper_restore_fbdev_mode_unlocked(fb_helper); - if (fb_helper->delayed_hotplug) { - fb_helper->delayed_hotplug = false; - drm_fb_helper_hotplug_event(fb_helper); - } return 0; } EXPORT_SYMBOL(drm_fb_helper_set_par); -- cgit v0.10.2 From 0391359ddf79b52bb7e7bb9ace08e34fb08b0e76 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 8 Dec 2014 22:55:22 +0100 Subject: drm/dp-mst: Remove branches before dropping the reference When we unplug a dp mst branch we unreference the entire tree from the root towards the leaves. Which is ok, since that's the way the pointers and so also the refcounts go. But when we drop the reference we must make sure that we remove the branches/ports from the lists/pointers before dropping the reference. Otherwise the get_validated functions will still return it instead of returning NULL (which indicates a potentially on-going unplug). The mst branch destroy gets this right for ports: First it deletes the port from the ports list, then it unrefs. But the ports destroy function gets it wrong: First it unrefs, then it drops the ref. Which means a zombie mst branch can still be validate with get_validated_mstb_ref when it shouldn't. Fix this. This should address a backtrace Dave dug out somewhere on unplug: [] drm_dp_mst_get_validated_mstb_ref_locked+0x92/0xa0 [drm_kms_helper] [] drm_dp_mst_get_validated_mstb_ref_locked+0x41/0xa0 [drm_kms_helper] [] drm_dp_get_validated_mstb_ref+0x3a/0x60 [drm_kms_helper] [] drm_dp_payload_send_msg.isra.14+0x2b/0x100 [drm_kms_helper] [] drm_dp_update_payload_part1+0x177/0x360 [drm_kms_helper] [] intel_mst_disable_dp+0x3e/0x80 [i915] [] haswell_crtc_disable+0x1cb/0x340 [i915] [] intel_crtc_control+0x49/0x100 [i915] [] intel_crtc_update_dpms+0x67/0x80 [i915] [] intel_connector_dpms+0x59/0x70 [i915] [] intel_dp_destroy_mst_connector+0x32/0xc0 [i915] [] drm_dp_destroy_port+0x6b/0xa0 [drm_kms_helper] [] drm_dp_destroy_mst_branch_device+0x108/0x130 [drm_kms_helper] [] drm_dp_port_teardown_pdt+0x3d/0x50 [drm_kms_helper] [] drm_dp_mst_handle_up_req+0x499/0x540 [drm_kms_helper] [] ? trace_hardirqs_on_caller+0x15d/0x200 [] drm_dp_mst_hpd_irq+0x53/0xa00 [drm_kms_helper] [] ? drm_dp_dpcd_read+0x1b/0x20 [drm_kms_helper] [] ? intel_dp_dpcd_read_wake+0x38/0x70 [i915] [] intel_dp_check_mst_status+0xb5/0x250 [i915] [] intel_dp_hpd_pulse+0x181/0x210 [i915] [] i915_digport_work_func+0x96/0x120 [i915] Signed-off-by: Daniel Vetter Cc: stable@vger.kernel.org Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c index 5682d7e..9798bb7 100644 --- a/drivers/gpu/drm/drm_dp_mst_topology.c +++ b/drivers/gpu/drm/drm_dp_mst_topology.c @@ -839,6 +839,8 @@ static void drm_dp_put_mst_branch_device(struct drm_dp_mst_branch *mstb) static void drm_dp_port_teardown_pdt(struct drm_dp_mst_port *port, int old_pdt) { + struct drm_dp_mst_branch *mstb; + switch (old_pdt) { case DP_PEER_DEVICE_DP_LEGACY_CONV: case DP_PEER_DEVICE_SST_SINK: @@ -846,8 +848,9 @@ static void drm_dp_port_teardown_pdt(struct drm_dp_mst_port *port, int old_pdt) drm_dp_mst_unregister_i2c_bus(&port->aux); break; case DP_PEER_DEVICE_MST_BRANCHING: - drm_dp_put_mst_branch_device(port->mstb); + mstb = port->mstb; port->mstb = NULL; + drm_dp_put_mst_branch_device(mstb); break; } } -- cgit v0.10.2 From b49b55bd4fba6aa9bb25459e0ac3991c60c4aba4 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 20 Oct 2014 16:13:19 +1000 Subject: drm/displayid: add displayid defines and edid extension (v2) These are just taken from the DisplayID v1.3 spec, and the DDC spec. v2: use __packed (Jani) Signed-off-by: Dave Airlie diff --git a/include/drm/drm_displayid.h b/include/drm/drm_displayid.h new file mode 100644 index 0000000..623b4e9 --- /dev/null +++ b/include/drm/drm_displayid.h @@ -0,0 +1,76 @@ +/* + * Copyright © 2014 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef DRM_DISPLAYID_H +#define DRM_DISPLAYID_H + +#define DATA_BLOCK_PRODUCT_ID 0x00 +#define DATA_BLOCK_DISPLAY_PARAMETERS 0x01 +#define DATA_BLOCK_COLOR_CHARACTERISTICS 0x02 +#define DATA_BLOCK_TYPE_1_DETAILED_TIMING 0x03 +#define DATA_BLOCK_TYPE_2_DETAILED_TIMING 0x04 +#define DATA_BLOCK_TYPE_3_SHORT_TIMING 0x05 +#define DATA_BLOCK_TYPE_4_DMT_TIMING 0x06 +#define DATA_BLOCK_VESA_TIMING 0x07 +#define DATA_BLOCK_CEA_TIMING 0x08 +#define DATA_BLOCK_VIDEO_TIMING_RANGE 0x09 +#define DATA_BLOCK_PRODUCT_SERIAL_NUMBER 0x0a +#define DATA_BLOCK_GP_ASCII_STRING 0x0b +#define DATA_BLOCK_DISPLAY_DEVICE_DATA 0x0c +#define DATA_BLOCK_INTERFACE_POWER_SEQUENCING 0x0d +#define DATA_BLOCK_TRANSFER_CHARACTERISTICS 0x0e +#define DATA_BLOCK_DISPLAY_INTERFACE 0x0f +#define DATA_BLOCK_STEREO_DISPLAY_INTERFACE 0x10 +#define DATA_BLOCK_TILED_DISPLAY 0x12 + +#define DATA_BLOCK_VENDOR_SPECIFIC 0x7f + +#define PRODUCT_TYPE_EXTENSION 0 +#define PRODUCT_TYPE_TEST 1 +#define PRODUCT_TYPE_PANEL 2 +#define PRODUCT_TYPE_MONITOR 3 +#define PRODUCT_TYPE_TV 4 +#define PRODUCT_TYPE_REPEATER 5 +#define PRODUCT_TYPE_DIRECT_DRIVE 6 + +struct displayid_hdr { + u8 rev; + u8 bytes; + u8 prod_id; + u8 ext_count; +} __packed; + +struct displayid_block { + u8 tag; + u8 rev; + u8 num_bytes; +} __packed; + +struct displayid_tiled_block { + struct displayid_block base; + u8 tile_cap; + u8 topo[3]; + u8 tile_size[4]; + u8 tile_pixel_bezel[5]; + u8 topology_id[8]; +} __packed; + +#endif diff --git a/include/drm/drm_edid.h b/include/drm/drm_edid.h index d59240f..87d85e8 100644 --- a/include/drm/drm_edid.h +++ b/include/drm/drm_edid.h @@ -27,12 +27,14 @@ #define EDID_LENGTH 128 #define DDC_ADDR 0x50 +#define DDC_ADDR2 0x52 /* E-DDC 1.2 - where DisplayID can hide */ #define CEA_EXT 0x02 #define VTB_EXT 0x10 #define DI_EXT 0x40 #define LS_EXT 0x50 #define MI_EXT 0x60 +#define DISPLAYID_EXT 0x70 struct est_timings { u8 t1; -- cgit v0.10.2 From 138f9ebb9755a8cf09fd6a9ff8d011aaf5fb478f Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 20 Oct 2014 16:17:17 +1000 Subject: drm: add tile_group support. (v3) A tile group is an identifier shared by a single monitor, DisplayID topology has 8 bytes we can use for this, just use those for now until something else comes up in the future. We assign these to an idr and use the idr to tell userspace what connectors are in the same tile group. DisplayID v1.3 says the serial number must be unique for displays from the same manufacturer. v2: destroy idr (dvdhrm) add docbook (danvet) airlied:- not sure how to make docbook add fns to tile group section. v3: fix missing unlock. Reviewed-by: Daniel Vetter Signed-off-by: Dave Airlie diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index 56e2a9b..bc4b5ab 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -2412,6 +2412,10 @@ void intel_crt_init(struct drm_device *dev) !Edrivers/gpu/drm/drm_plane_helper.c !Pdrivers/gpu/drm/drm_plane_helper.c overview + + Tile group +!Pdrivers/gpu/drm/drm_crtc.c Tile group + diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index de79283..0d1eaa9 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -5152,6 +5152,7 @@ void drm_mode_config_init(struct drm_device *dev) INIT_LIST_HEAD(&dev->mode_config.property_blob_list); INIT_LIST_HEAD(&dev->mode_config.plane_list); idr_init(&dev->mode_config.crtc_idr); + idr_init(&dev->mode_config.tile_idr); drm_modeset_lock_all(dev); drm_mode_create_standard_connector_properties(dev); @@ -5239,6 +5240,7 @@ void drm_mode_config_cleanup(struct drm_device *dev) crtc->funcs->destroy(crtc); } + idr_destroy(&dev->mode_config.tile_idr); idr_destroy(&dev->mode_config.crtc_idr); drm_modeset_lock_fini(&dev->mode_config.connection_mutex); } @@ -5261,3 +5263,100 @@ struct drm_property *drm_mode_create_rotation_property(struct drm_device *dev, supported_rotations); } EXPORT_SYMBOL(drm_mode_create_rotation_property); + +/** + * DOC: Tile group + * + * Tile groups are used to represent tiled monitors with a unique + * integer identifier. Tiled monitors using DisplayID v1.3 have + * a unique 8-byte handle, we store this in a tile group, so we + * have a common identifier for all tiles in a monitor group. + */ +static void drm_tile_group_free(struct kref *kref) +{ + struct drm_tile_group *tg = container_of(kref, struct drm_tile_group, refcount); + struct drm_device *dev = tg->dev; + mutex_lock(&dev->mode_config.idr_mutex); + idr_remove(&dev->mode_config.tile_idr, tg->id); + mutex_unlock(&dev->mode_config.idr_mutex); + kfree(tg); +} + +/** + * drm_mode_put_tile_group - drop a reference to a tile group. + * @dev: DRM device + * @tg: tile group to drop reference to. + * + * drop reference to tile group and free if 0. + */ +void drm_mode_put_tile_group(struct drm_device *dev, + struct drm_tile_group *tg) +{ + kref_put(&tg->refcount, drm_tile_group_free); +} + +/** + * drm_mode_get_tile_group - get a reference to an existing tile group + * @dev: DRM device + * @topology: 8-bytes unique per monitor. + * + * Use the unique bytes to get a reference to an existing tile group. + * + * RETURNS: + * tile group or NULL if not found. + */ +struct drm_tile_group *drm_mode_get_tile_group(struct drm_device *dev, + char topology[8]) +{ + struct drm_tile_group *tg; + int id; + mutex_lock(&dev->mode_config.idr_mutex); + idr_for_each_entry(&dev->mode_config.tile_idr, tg, id) { + if (!memcmp(tg->group_data, topology, 8)) { + if (!kref_get_unless_zero(&tg->refcount)) + tg = NULL; + mutex_unlock(&dev->mode_config.idr_mutex); + return tg; + } + } + mutex_unlock(&dev->mode_config.idr_mutex); + return NULL; +} + +/** + * drm_mode_create_tile_group - create a tile group from a displayid description + * @dev: DRM device + * @topology: 8-bytes unique per monitor. + * + * Create a tile group for the unique monitor, and get a unique + * identifier for the tile group. + * + * RETURNS: + * new tile group or error. + */ +struct drm_tile_group *drm_mode_create_tile_group(struct drm_device *dev, + char topology[8]) +{ + struct drm_tile_group *tg; + int ret; + + tg = kzalloc(sizeof(*tg), GFP_KERNEL); + if (!tg) + return ERR_PTR(-ENOMEM); + + kref_init(&tg->refcount); + memcpy(tg->group_data, topology, 8); + tg->dev = dev; + + mutex_lock(&dev->mode_config.idr_mutex); + ret = idr_alloc(&dev->mode_config.tile_idr, tg, 1, 0, GFP_KERNEL); + if (ret >= 0) { + tg->id = ret; + } else { + kfree(tg); + tg = ERR_PTR(ret); + } + + mutex_unlock(&dev->mode_config.idr_mutex); + return tg; +} diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index dd2c16e..8f760a2 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -137,6 +137,14 @@ struct drm_display_info { u8 cea_rev; }; +/* data corresponds to displayid vend/prod/serial */ +struct drm_tile_group { + struct kref refcount; + struct drm_device *dev; + int id; + u8 group_data[8]; +}; + struct drm_framebuffer_funcs { /* note: use drm_framebuffer_remove() */ void (*destroy)(struct drm_framebuffer *framebuffer); @@ -978,6 +986,7 @@ struct drm_mode_config { struct drm_modeset_acquire_ctx *acquire_ctx; /* for legacy _lock_all() / _unlock_all() */ struct mutex idr_mutex; /* for IDR management */ struct idr crtc_idr; /* use this idr for all IDs, fb, crtc, connector, modes - just makes life easier */ + struct idr tile_idr; /* use this idr for all IDs, fb, crtc, connector, modes - just makes life easier */ /* this is limited to one for now */ struct mutex fb_lock; /* proctects global and per-file fb lists */ @@ -1326,6 +1335,13 @@ extern void drm_set_preferred_mode(struct drm_connector *connector, extern int drm_edid_header_is_valid(const u8 *raw_edid); extern bool drm_edid_block_valid(u8 *raw_edid, int block, bool print_bad_edid); extern bool drm_edid_is_valid(struct edid *edid); + +extern struct drm_tile_group *drm_mode_create_tile_group(struct drm_device *dev, + char topology[8]); +extern struct drm_tile_group *drm_mode_get_tile_group(struct drm_device *dev, + char topology[8]); +extern void drm_mode_put_tile_group(struct drm_device *dev, + struct drm_tile_group *tg); struct drm_display_mode *drm_mode_find_dmt(struct drm_device *dev, int hsize, int vsize, int fresh, bool rb); -- cgit v0.10.2 From c6a0aed4d493936f61cd153db84531026705c94d Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 20 Oct 2014 16:28:02 +1000 Subject: drm/mst: cached EDID for logical ports (v2) Logical ports are never going to have EDID changes, they are used for the internal ports on MST monitors. We cache the EDIDs from these to save time at MST probe. v2: drop misplace tile property line, meant for other patch. Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c index 9798bb7..0a9d3aa 100644 --- a/drivers/gpu/drm/drm_dp_mst_topology.c +++ b/drivers/gpu/drm/drm_dp_mst_topology.c @@ -861,6 +861,8 @@ static void drm_dp_destroy_port(struct kref *kref) struct drm_dp_mst_topology_mgr *mgr = port->mgr; if (!port->input) { port->vcpi.num_slots = 0; + + kfree(port->cached_edid); if (port->connector) (*port->mgr->cbs->destroy_connector)(mgr, port->connector); drm_dp_port_teardown_pdt(port, port->pdt); @@ -1100,6 +1102,10 @@ static void drm_dp_add_port(struct drm_dp_mst_branch *mstb, char proppath[255]; build_mst_prop_path(port, mstb, proppath, sizeof(proppath)); port->connector = (*mstb->mgr->cbs->add_connector)(mstb->mgr, port, proppath); + + if (port->port_num >= 8) { + port->cached_edid = drm_get_edid(port->connector, &port->aux.ddc); + } } /* put reference to this port */ @@ -2170,7 +2176,8 @@ EXPORT_SYMBOL(drm_dp_mst_hpd_irq); * This returns the current connection state for a port. It validates the * port pointer still exists so the caller doesn't require a reference */ -enum drm_connector_status drm_dp_mst_detect_port(struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port) +enum drm_connector_status drm_dp_mst_detect_port(struct drm_connector *connector, + struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port) { enum drm_connector_status status = connector_status_disconnected; @@ -2189,6 +2196,10 @@ enum drm_connector_status drm_dp_mst_detect_port(struct drm_dp_mst_topology_mgr case DP_PEER_DEVICE_SST_SINK: status = connector_status_connected; + /* for logical ports - cache the EDID */ + if (port->port_num >= 8 && !port->cached_edid) { + port->cached_edid = drm_get_edid(connector, &port->aux.ddc); + } break; case DP_PEER_DEVICE_DP_LEGACY_CONV: if (port->ldps) @@ -2220,7 +2231,11 @@ struct edid *drm_dp_mst_get_edid(struct drm_connector *connector, struct drm_dp_ if (!port) return NULL; - edid = drm_get_edid(connector, &port->aux.ddc); + if (port->cached_edid) + edid = drm_edid_duplicate(port->cached_edid); + else + edid = drm_get_edid(connector, &port->aux.ddc); + drm_dp_put_port(port); return edid; } diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index bfe3595..428bb30 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -283,7 +283,7 @@ intel_dp_mst_detect(struct drm_connector *connector, bool force) struct intel_connector *intel_connector = to_intel_connector(connector); struct intel_dp *intel_dp = intel_connector->mst_port; - return drm_dp_mst_detect_port(&intel_dp->mst_mgr, intel_connector->port); + return drm_dp_mst_detect_port(connector, &intel_dp->mst_mgr, intel_connector->port); } static int diff --git a/include/drm/drm_dp_mst_helper.h b/include/drm/drm_dp_mst_helper.h index cec6383..00c1da9 100644 --- a/include/drm/drm_dp_mst_helper.h +++ b/include/drm/drm_dp_mst_helper.h @@ -92,6 +92,8 @@ struct drm_dp_mst_port { struct drm_dp_vcpi vcpi; struct drm_connector *connector; struct drm_dp_mst_topology_mgr *mgr; + + struct edid *cached_edid; /* for DP logical ports - make tiling work */ }; /** @@ -474,7 +476,7 @@ int drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr *mgr, bool ms int drm_dp_mst_hpd_irq(struct drm_dp_mst_topology_mgr *mgr, u8 *esi, bool *handled); -enum drm_connector_status drm_dp_mst_detect_port(struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port); +enum drm_connector_status drm_dp_mst_detect_port(struct drm_connector *connector, struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port); struct edid *drm_dp_mst_get_edid(struct drm_connector *connector, struct drm_dp_mst_topology_mgr *mgr, struct drm_dp_mst_port *port); -- cgit v0.10.2 From 40d9b043a89e2301e1f97ade055a73ecc28e9afe Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 20 Oct 2014 16:29:33 +1000 Subject: drm/connector: store tile information from displayid (v3) This creates a tile group from DisplayID block, and stores the pieces of parsed info from the DisplayID block into the connector. v2: add missing signoff, add new connector bits to docs. v3: remove some debugging. Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 0d1eaa9..eb89327 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -908,6 +908,11 @@ void drm_connector_cleanup(struct drm_connector *connector) struct drm_device *dev = connector->dev; struct drm_display_mode *mode, *t; + if (connector->tile_group) { + drm_mode_put_tile_group(dev, connector->tile_group); + connector->tile_group = NULL; + } + list_for_each_entry_safe(mode, t, &connector->probed_modes, head) drm_mode_remove(connector, mode); diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index a7b5a71..72fd882 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -34,6 +34,7 @@ #include #include #include +#include #define version_greater(edid, maj, min) \ (((edid)->version > (maj)) || \ @@ -1014,6 +1015,8 @@ module_param_named(edid_fixup, edid_fixup, int, 0400); MODULE_PARM_DESC(edid_fixup, "Minimum number of valid EDID header bytes (0-8, default 6)"); +static void drm_get_displayid(struct drm_connector *connector, + struct edid *edid); /** * drm_edid_block_valid - Sanity check the EDID block (base or extension) * @raw_edid: pointer to raw EDID block @@ -1308,10 +1311,15 @@ EXPORT_SYMBOL(drm_probe_ddc); struct edid *drm_get_edid(struct drm_connector *connector, struct i2c_adapter *adapter) { + struct edid *edid; + if (!drm_probe_ddc(adapter)) return NULL; - return drm_do_get_edid(connector, drm_do_probe_ddc_edid, adapter); + edid = drm_do_get_edid(connector, drm_do_probe_ddc_edid, adapter); + if (edid) + drm_get_displayid(connector, edid); + return edid; } EXPORT_SYMBOL(drm_get_edid); @@ -2406,7 +2414,7 @@ add_detailed_modes(struct drm_connector *connector, struct edid *edid, /* * Search EDID for CEA extension block. */ -static u8 *drm_find_cea_extension(struct edid *edid) +static u8 *drm_find_edid_extension(struct edid *edid, int ext_id) { u8 *edid_ext = NULL; int i; @@ -2418,7 +2426,7 @@ static u8 *drm_find_cea_extension(struct edid *edid) /* Find CEA extension */ for (i = 0; i < edid->extensions; i++) { edid_ext = (u8 *)edid + EDID_LENGTH * (i + 1); - if (edid_ext[0] == CEA_EXT) + if (edid_ext[0] == ext_id) break; } @@ -2428,6 +2436,16 @@ static u8 *drm_find_cea_extension(struct edid *edid) return edid_ext; } +static u8 *drm_find_cea_extension(struct edid *edid) +{ + return drm_find_edid_extension(edid, CEA_EXT); +} + +static u8 *drm_find_displayid_extension(struct edid *edid) +{ + return drm_find_edid_extension(edid, DISPLAYID_EXT); +} + /* * Calculate the alternate clock for the CEA mode * (60Hz vs. 59.94Hz etc.) @@ -3888,3 +3906,123 @@ drm_hdmi_vendor_infoframe_from_display_mode(struct hdmi_vendor_infoframe *frame, return 0; } EXPORT_SYMBOL(drm_hdmi_vendor_infoframe_from_display_mode); + +static int drm_parse_display_id(struct drm_connector *connector, + u8 *displayid, int length, + bool is_edid_extension) +{ + /* if this is an EDID extension the first byte will be 0x70 */ + int idx = 0; + struct displayid_hdr *base; + struct displayid_block *block; + u8 csum = 0; + int i; + + if (is_edid_extension) + idx = 1; + + base = (struct displayid_hdr *)&displayid[idx]; + + DRM_DEBUG_KMS("base revision 0x%x, length %d, %d %d\n", + base->rev, base->bytes, base->prod_id, base->ext_count); + + if (base->bytes + 5 > length - idx) + return -EINVAL; + + for (i = idx; i <= base->bytes + 5; i++) { + csum += displayid[i]; + } + if (csum) { + DRM_ERROR("DisplayID checksum invalid, remainder is %d\n", csum); + return -EINVAL; + } + + block = (struct displayid_block *)&displayid[idx + 4]; + DRM_DEBUG_KMS("block id %d, rev %d, len %d\n", + block->tag, block->rev, block->num_bytes); + + switch (block->tag) { + case DATA_BLOCK_TILED_DISPLAY: { + struct displayid_tiled_block *tile = (struct displayid_tiled_block *)block; + + u16 w, h; + u8 tile_v_loc, tile_h_loc; + u8 num_v_tile, num_h_tile; + struct drm_tile_group *tg; + + w = tile->tile_size[0] | tile->tile_size[1] << 8; + h = tile->tile_size[2] | tile->tile_size[3] << 8; + + num_v_tile = (tile->topo[0] & 0xf) | (tile->topo[2] & 0x30); + num_h_tile = (tile->topo[0] >> 4) | ((tile->topo[2] >> 2) & 0x30); + tile_v_loc = (tile->topo[1] & 0xf) | ((tile->topo[2] & 0x3) << 4); + tile_h_loc = (tile->topo[1] >> 4) | (((tile->topo[2] >> 2) & 0x3) << 4); + + connector->has_tile = true; + if (tile->tile_cap & 0x80) + connector->tile_is_single_monitor = true; + + connector->num_h_tile = num_h_tile + 1; + connector->num_v_tile = num_v_tile + 1; + connector->tile_h_loc = tile_h_loc; + connector->tile_v_loc = tile_v_loc; + connector->tile_h_size = w + 1; + connector->tile_v_size = h + 1; + + DRM_DEBUG_KMS("tile cap 0x%x\n", tile->tile_cap); + DRM_DEBUG_KMS("tile_size %d x %d\n", w + 1, h + 1); + DRM_DEBUG_KMS("topo num tiles %dx%d, location %dx%d\n", + num_h_tile + 1, num_v_tile + 1, tile_h_loc, tile_v_loc); + DRM_DEBUG_KMS("vend %c%c%c\n", tile->topology_id[0], tile->topology_id[1], tile->topology_id[2]); + + tg = drm_mode_get_tile_group(connector->dev, tile->topology_id); + if (!tg) { + tg = drm_mode_create_tile_group(connector->dev, tile->topology_id); + } + if (!tg) + return -ENOMEM; + + if (connector->tile_group != tg) { + /* if we haven't got a pointer, + take the reference, drop ref to old tile group */ + if (connector->tile_group) { + drm_mode_put_tile_group(connector->dev, connector->tile_group); + } + connector->tile_group = tg; + } else + /* if same tile group, then release the ref we just took. */ + drm_mode_put_tile_group(connector->dev, tg); + } + break; + default: + printk("unknown displayid tag %d\n", block->tag); + break; + } + return 0; +} + +static void drm_get_displayid(struct drm_connector *connector, + struct edid *edid) +{ + void *displayid = NULL; + int ret; + connector->has_tile = false; + displayid = drm_find_displayid_extension(edid); + if (!displayid) { + /* drop reference to any tile group we had */ + goto out_drop_ref; + } + + ret = drm_parse_display_id(connector, displayid, EDID_LENGTH, true); + if (ret < 0) + goto out_drop_ref; + if (!connector->has_tile) + goto out_drop_ref; + return; +out_drop_ref: + if (connector->tile_group) { + drm_mode_put_tile_group(connector->dev, connector->tile_group); + connector->tile_group = NULL; + } + return; +} diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 8f760a2..01744ed 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -607,6 +607,15 @@ struct drm_encoder { * @bad_edid_counter: track sinks that give us an EDID with invalid checksum * @debugfs_entry: debugfs directory for this connector * @state: current atomic state for this connector + * @has_tile: is this connector connected to a tiled monitor + * @tile_group: tile group for the connected monitor + * @tile_is_single_monitor: whether the tile is one monitor housing + * @num_h_tile: number of horizontal tiles in the tile group + * @num_v_tile: number of vertical tiles in the tile group + * @tile_h_loc: horizontal location of this tile + * @tile_v_loc: vertical location of this tile + * @tile_h_size: horizontal size of this tile. + * @tile_v_size: vertical size of this tile. * * Each connector may be connected to one or more CRTCs, or may be clonable by * another connector if they can share a CRTC. Each connector also has a specific @@ -669,6 +678,15 @@ struct drm_connector { struct dentry *debugfs_entry; struct drm_connector_state *state; + + /* DisplayID bits */ + bool has_tile; + struct drm_tile_group *tile_group; + bool tile_is_single_monitor; + + uint8_t num_h_tile, num_v_tile; + uint8_t tile_h_loc, tile_v_loc; + uint16_t tile_h_size, tile_v_size; }; /** -- cgit v0.10.2 From 6f134d7bb4347ab4c66ef123efb838fedb54186f Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 20 Oct 2014 16:30:50 +1000 Subject: drm/tile: expose the tile property to userspace (v3) This takes the tiling info from the connector and exposes it to userspace, as a blob object in a connector property. The contents of the blob is ABI. v2: add property + function documentation. v3: move property setup from previous patch. add boilerplate + fix long line (Daniel) Reviewed-by: Daniel Vetter Signed-off-by: Dave Airlie diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index bc4b5ab..60c1063d 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -2551,7 +2551,7 @@ void intel_crt_init(struct drm_device *dev) - + @@ -2573,6 +2573,13 @@ void intel_crt_init(struct drm_device *dev) + + + + + + + diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index eb89327..4a44f89 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -1344,6 +1344,11 @@ static int drm_mode_create_standard_connector_properties(struct drm_device *dev) "PATH", 0); dev->mode_config.path_property = dev_path; + dev->mode_config.tile_property = drm_property_create(dev, + DRM_MODE_PROP_BLOB | + DRM_MODE_PROP_IMMUTABLE, + "TILE", 0); + return 0; } @@ -4088,6 +4093,52 @@ int drm_mode_connector_set_path_property(struct drm_connector *connector, EXPORT_SYMBOL(drm_mode_connector_set_path_property); /** + * drm_mode_connector_set_tile_property - set tile property on connector + * @connector: connector to set property on. + * + * This looks up the tile information for a connector, and creates a + * property for userspace to parse if it exists. The property is of + * the form of 8 integers using ':' as a separator. + * + * Returns: + * Zero on success, errno on failure. + */ +int drm_mode_connector_set_tile_property(struct drm_connector *connector) +{ + struct drm_device *dev = connector->dev; + int ret, size; + char tile[256]; + + if (connector->tile_blob_ptr) + drm_property_destroy_blob(dev, connector->tile_blob_ptr); + + if (!connector->has_tile) { + connector->tile_blob_ptr = NULL; + ret = drm_object_property_set_value(&connector->base, + dev->mode_config.tile_property, 0); + return ret; + } + + snprintf(tile, 256, "%d:%d:%d:%d:%d:%d:%d:%d", + connector->tile_group->id, connector->tile_is_single_monitor, + connector->num_h_tile, connector->num_v_tile, + connector->tile_h_loc, connector->tile_v_loc, + connector->tile_h_size, connector->tile_v_size); + size = strlen(tile) + 1; + + connector->tile_blob_ptr = drm_property_create_blob(connector->dev, + size, tile); + if (!connector->tile_blob_ptr) + return -EINVAL; + + ret = drm_object_property_set_value(&connector->base, + dev->mode_config.tile_property, + connector->tile_blob_ptr->base.id); + return ret; +} +EXPORT_SYMBOL(drm_mode_connector_set_tile_property); + +/** * drm_mode_connector_update_edid_property - update the edid property of a connector * @connector: drm connector * @edid: new value of the edid property diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c index 0a9d3aa..9a5b687 100644 --- a/drivers/gpu/drm/drm_dp_mst_topology.c +++ b/drivers/gpu/drm/drm_dp_mst_topology.c @@ -2236,6 +2236,7 @@ struct edid *drm_dp_mst_get_edid(struct drm_connector *connector, struct drm_dp_ else edid = drm_get_edid(connector, &port->aux.ddc); + drm_mode_connector_set_tile_property(connector); drm_dp_put_port(port); return edid; } diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index 428bb30..7f8c6a6 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -414,6 +414,8 @@ static struct drm_connector *intel_dp_add_mst_connector(struct drm_dp_mst_topolo intel_dp_add_properties(intel_dp, connector); drm_object_attach_property(&connector->base, dev->mode_config.path_property, 0); + drm_object_attach_property(&connector->base, dev->mode_config.tile_property, 0); + drm_mode_connector_set_path_property(connector, pathprop); drm_reinit_primary_mode_group(dev); mutex_lock(&dev->mode_config.mutex); diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index 01744ed..b863298 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -651,6 +651,8 @@ struct drm_connector { struct drm_property_blob *path_blob_ptr; + struct drm_property_blob *tile_blob_ptr; + uint8_t polled; /* DRM_CONNECTOR_POLL_* */ /* requested DPMS state */ @@ -1048,6 +1050,7 @@ struct drm_mode_config { struct drm_property *edid_property; struct drm_property *dpms_property; struct drm_property *path_property; + struct drm_property *tile_property; struct drm_property *plane_type_property; struct drm_property *rotation_property; @@ -1217,6 +1220,7 @@ extern void drm_mode_config_cleanup(struct drm_device *dev); extern int drm_mode_connector_set_path_property(struct drm_connector *connector, const char *path); +int drm_mode_connector_set_tile_property(struct drm_connector *connector); extern int drm_mode_connector_update_edid_property(struct drm_connector *connector, const struct edid *edid); -- cgit v0.10.2 From b0ee9e7fa5b461a91f24d1d03b10c6bf162c86f9 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 20 Oct 2014 16:31:53 +1000 Subject: drm/fb: add support for tiled monitor configurations. (v2) This adds fbdev/con support for tiled monitors, so that we only set a mode on the correct half of the monitor, or span the two halves if needed. v2: remove unneeded ERROR, fix | vs || Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index a467460..52ce26d 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -1000,19 +1000,21 @@ static int drm_fb_helper_single_fb_probe(struct drm_fb_helper *fb_helper, crtc_count = 0; for (i = 0; i < fb_helper->crtc_count; i++) { struct drm_display_mode *desired_mode; + int x, y; desired_mode = fb_helper->crtc_info[i].desired_mode; - + x = fb_helper->crtc_info[i].x; + y = fb_helper->crtc_info[i].y; if (desired_mode) { if (gamma_size == 0) gamma_size = fb_helper->crtc_info[i].mode_set.crtc->gamma_size; - if (desired_mode->hdisplay < sizes.fb_width) - sizes.fb_width = desired_mode->hdisplay; - if (desired_mode->vdisplay < sizes.fb_height) - sizes.fb_height = desired_mode->vdisplay; - if (desired_mode->hdisplay > sizes.surface_width) - sizes.surface_width = desired_mode->hdisplay; - if (desired_mode->vdisplay > sizes.surface_height) - sizes.surface_height = desired_mode->vdisplay; + if (desired_mode->hdisplay + x < sizes.fb_width) + sizes.fb_width = desired_mode->hdisplay + x; + if (desired_mode->vdisplay + y < sizes.fb_height) + sizes.fb_height = desired_mode->vdisplay + y; + if (desired_mode->hdisplay + x > sizes.surface_width) + sizes.surface_width = desired_mode->hdisplay + x; + if (desired_mode->vdisplay + y > sizes.surface_height) + sizes.surface_height = desired_mode->vdisplay + y; crtc_count++; } } @@ -1312,6 +1314,7 @@ static void drm_enable_connectors(struct drm_fb_helper *fb_helper, static bool drm_target_cloned(struct drm_fb_helper *fb_helper, struct drm_display_mode **modes, + struct drm_fb_offset *offsets, bool *enabled, int width, int height) { int count, i, j; @@ -1383,27 +1386,88 @@ static bool drm_target_cloned(struct drm_fb_helper *fb_helper, return false; } +static int drm_get_tile_offsets(struct drm_fb_helper *fb_helper, + struct drm_display_mode **modes, + struct drm_fb_offset *offsets, + int idx, + int h_idx, int v_idx) +{ + struct drm_fb_helper_connector *fb_helper_conn; + int i; + int hoffset = 0, voffset = 0; + + for (i = 0; i < fb_helper->connector_count; i++) { + fb_helper_conn = fb_helper->connector_info[i]; + if (!fb_helper_conn->connector->has_tile) + continue; + + if (!modes[i] && (h_idx || v_idx)) { + DRM_DEBUG_KMS("no modes for connector tiled %d %d\n", i, + fb_helper_conn->connector->base.id); + continue; + } + if (fb_helper_conn->connector->tile_h_loc < h_idx) + hoffset += modes[i]->hdisplay; + + if (fb_helper_conn->connector->tile_v_loc < v_idx) + voffset += modes[i]->vdisplay; + } + offsets[idx].x = hoffset; + offsets[idx].y = voffset; + DRM_DEBUG_KMS("returned %d %d for %d %d\n", hoffset, voffset, h_idx, v_idx); + return 0; +} + static bool drm_target_preferred(struct drm_fb_helper *fb_helper, struct drm_display_mode **modes, + struct drm_fb_offset *offsets, bool *enabled, int width, int height) { struct drm_fb_helper_connector *fb_helper_conn; int i; - + uint64_t conn_configured = 0, mask; + int tile_pass = 0; + mask = (1 << fb_helper->connector_count) - 1; +retry: for (i = 0; i < fb_helper->connector_count; i++) { fb_helper_conn = fb_helper->connector_info[i]; - if (enabled[i] == false) + if (conn_configured & (1 << i)) + continue; + + if (enabled[i] == false) { + conn_configured |= (1 << i); + continue; + } + + /* first pass over all the untiled connectors */ + if (tile_pass == 0 && fb_helper_conn->connector->has_tile) continue; + if (tile_pass == 1) { + if (fb_helper_conn->connector->tile_h_loc != 0 || + fb_helper_conn->connector->tile_v_loc != 0) + continue; + + } else { + if (fb_helper_conn->connector->tile_h_loc != tile_pass -1 && + fb_helper_conn->connector->tile_v_loc != tile_pass - 1) + /* if this tile_pass doesn't cover any of the tiles - keep going */ + continue; + + /* find the tile offsets for this pass - need + to find all tiles left and above */ + drm_get_tile_offsets(fb_helper, modes, offsets, + i, fb_helper_conn->connector->tile_h_loc, fb_helper_conn->connector->tile_v_loc); + } DRM_DEBUG_KMS("looking for cmdline mode on connector %d\n", fb_helper_conn->connector->base.id); /* got for command line mode first */ modes[i] = drm_pick_cmdline_mode(fb_helper_conn, width, height); if (!modes[i]) { - DRM_DEBUG_KMS("looking for preferred mode on connector %d\n", - fb_helper_conn->connector->base.id); + DRM_DEBUG_KMS("looking for preferred mode on connector %d %d\n", + fb_helper_conn->connector->base.id, fb_helper_conn->connector->tile_group ? fb_helper_conn->connector->tile_group->id : 0); modes[i] = drm_has_preferred_mode(fb_helper_conn, width, height); } /* No preferred modes, pick one off the list */ @@ -1413,6 +1477,12 @@ static bool drm_target_preferred(struct drm_fb_helper *fb_helper, } DRM_DEBUG_KMS("found mode %s\n", modes[i] ? modes[i]->name : "none"); + conn_configured |= (1 << i); + } + + if ((conn_configured & mask) != mask) { + tile_pass++; + goto retry; } return true; } @@ -1502,6 +1572,7 @@ static void drm_setup_crtcs(struct drm_fb_helper *fb_helper) struct drm_device *dev = fb_helper->dev; struct drm_fb_helper_crtc **crtcs; struct drm_display_mode **modes; + struct drm_fb_offset *offsets; struct drm_mode_set *modeset; bool *enabled; int width, height; @@ -1516,9 +1587,11 @@ static void drm_setup_crtcs(struct drm_fb_helper *fb_helper) sizeof(struct drm_fb_helper_crtc *), GFP_KERNEL); modes = kcalloc(dev->mode_config.num_connector, sizeof(struct drm_display_mode *), GFP_KERNEL); + offsets = kcalloc(dev->mode_config.num_connector, + sizeof(struct drm_fb_offset), GFP_KERNEL); enabled = kcalloc(dev->mode_config.num_connector, sizeof(bool), GFP_KERNEL); - if (!crtcs || !modes || !enabled) { + if (!crtcs || !modes || !enabled || !offsets) { DRM_ERROR("Memory allocation failed\n"); goto out; } @@ -1528,14 +1601,16 @@ static void drm_setup_crtcs(struct drm_fb_helper *fb_helper) if (!(fb_helper->funcs->initial_config && fb_helper->funcs->initial_config(fb_helper, crtcs, modes, + offsets, enabled, width, height))) { memset(modes, 0, dev->mode_config.num_connector*sizeof(modes[0])); memset(crtcs, 0, dev->mode_config.num_connector*sizeof(crtcs[0])); + memset(offsets, 0, dev->mode_config.num_connector*sizeof(offsets[0])); - if (!drm_target_cloned(fb_helper, - modes, enabled, width, height) && - !drm_target_preferred(fb_helper, - modes, enabled, width, height)) + if (!drm_target_cloned(fb_helper, modes, offsets, + enabled, width, height) && + !drm_target_preferred(fb_helper, modes, offsets, + enabled, width, height)) DRM_ERROR("Unable to find initial modes\n"); DRM_DEBUG_KMS("picking CRTCs for %dx%d config\n", @@ -1555,18 +1630,23 @@ static void drm_setup_crtcs(struct drm_fb_helper *fb_helper) for (i = 0; i < fb_helper->connector_count; i++) { struct drm_display_mode *mode = modes[i]; struct drm_fb_helper_crtc *fb_crtc = crtcs[i]; + struct drm_fb_offset *offset = &offsets[i]; modeset = &fb_crtc->mode_set; if (mode && fb_crtc) { - DRM_DEBUG_KMS("desired mode %s set on crtc %d\n", - mode->name, fb_crtc->mode_set.crtc->base.id); + DRM_DEBUG_KMS("desired mode %s set on crtc %d (%d,%d)\n", + mode->name, fb_crtc->mode_set.crtc->base.id, offset->x, offset->y); fb_crtc->desired_mode = mode; + fb_crtc->x = offset->x; + fb_crtc->y = offset->y; if (modeset->mode) drm_mode_destroy(dev, modeset->mode); modeset->mode = drm_mode_duplicate(dev, fb_crtc->desired_mode); modeset->connectors[modeset->num_connectors++] = fb_helper->connector_info[i]->connector; modeset->fb = fb_helper->fb; + modeset->x = offset->x; + modeset->y = offset->y; } } @@ -1583,6 +1663,7 @@ static void drm_setup_crtcs(struct drm_fb_helper *fb_helper) out: kfree(crtcs); kfree(modes); + kfree(offsets); kfree(enabled); } diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c index f2183b5..850cf7d 100644 --- a/drivers/gpu/drm/i915/intel_fbdev.c +++ b/drivers/gpu/drm/i915/intel_fbdev.c @@ -324,6 +324,7 @@ intel_fb_helper_crtc(struct drm_fb_helper *fb_helper, struct drm_crtc *crtc) static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, struct drm_fb_helper_crtc **crtcs, struct drm_display_mode **modes, + struct drm_fb_offset *offsets, bool *enabled, int width, int height) { struct drm_device *dev = fb_helper->dev; @@ -332,6 +333,8 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, bool fallback = true; int num_connectors_enabled = 0; int num_connectors_detected = 0; + uint64_t conn_configured = 0, mask; + int pass = 0; save_enabled = kcalloc(dev->mode_config.num_connector, sizeof(bool), GFP_KERNEL); @@ -339,7 +342,8 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, return false; memcpy(save_enabled, enabled, dev->mode_config.num_connector); - + mask = (1 << fb_helper->connector_count) - 1; +retry: for (i = 0; i < fb_helper->connector_count; i++) { struct drm_fb_helper_connector *fb_conn; struct drm_connector *connector; @@ -349,12 +353,19 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, fb_conn = fb_helper->connector_info[i]; connector = fb_conn->connector; + if (conn_configured & (1 << i)) + continue; + + if (pass == 0 && !connector->has_tile) + continue; + if (connector->status == connector_status_connected) num_connectors_detected++; if (!enabled[i]) { DRM_DEBUG_KMS("connector %s not enabled, skipping\n", connector->name); + conn_configured |= (1 << i); continue; } @@ -373,6 +384,7 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, DRM_DEBUG_KMS("connector %s has no encoder or crtc, skipping\n", connector->name); enabled[i] = false; + conn_configured |= (1 << i); continue; } @@ -400,8 +412,8 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, /* try for preferred next */ if (!modes[i]) { - DRM_DEBUG_KMS("looking for preferred mode on connector %s\n", - connector->name); + DRM_DEBUG_KMS("looking for preferred mode on connector %s %d\n", + connector->name, connector->has_tile); modes[i] = drm_has_preferred_mode(fb_conn, width, height); } @@ -444,6 +456,12 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, modes[i]->flags & DRM_MODE_FLAG_INTERLACE ? "i" :""); fallback = false; + conn_configured |= (1 << i); + } + + if ((conn_configured & mask) != mask) { + pass++; + goto retry; } /* diff --git a/include/drm/drm_fb_helper.h b/include/drm/drm_fb_helper.h index f4ad254..b597068 100644 --- a/include/drm/drm_fb_helper.h +++ b/include/drm/drm_fb_helper.h @@ -34,9 +34,14 @@ struct drm_fb_helper; #include +struct drm_fb_offset { + int x, y; +}; + struct drm_fb_helper_crtc { struct drm_mode_set mode_set; struct drm_display_mode *desired_mode; + int x, y; }; struct drm_fb_helper_surface_size { @@ -72,6 +77,7 @@ struct drm_fb_helper_funcs { bool (*initial_config)(struct drm_fb_helper *fb_helper, struct drm_fb_helper_crtc **crtcs, struct drm_display_mode **modes, + struct drm_fb_offset *offsets, bool *enabled, int width, int height); }; -- cgit v0.10.2 From e2db726b560c963c6da49005d24d1310eeed22c5 Mon Sep 17 00:00:00 2001 From: Martin Peres Date: Tue, 9 Dec 2014 07:24:04 +0100 Subject: drm: fix a word repetition in a comment Spotted while reviewing the DRM changes in Linux 3.18 for LinuxFR. CC: Daniel Vetter Signed-off-by: Martin Peres Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index de79283..3ce5882 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -3444,7 +3444,7 @@ void drm_fb_release(struct drm_file *priv) /* * When the file gets released that means no one else can access the fb - * list any more, so no need to grab fpriv->fbs_lock. And we need to to + * list any more, so no need to grab fpriv->fbs_lock. And we need to * avoid upsetting lockdep since the universal cursor code adds a * framebuffer while holding mutex locks. * -- cgit v0.10.2 From 393e872e1daa4958f862fdbd321bb9ba7f7acb0d Mon Sep 17 00:00:00 2001 From: Martin Peres Date: Tue, 9 Dec 2014 08:13:09 +0100 Subject: drm: fix a typo in a comment MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Spotted while reviewing the DRM changes in Linux 3.18 for LinuxFR. Cc: Ville Syrjälä Signed-off-by: Martin Peres Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c index 0e47df4..f5a5f18 100644 --- a/drivers/gpu/drm/drm_irq.c +++ b/drivers/gpu/drm/drm_irq.c @@ -166,7 +166,7 @@ static void vblank_disable_and_save(struct drm_device *dev, int crtc) spin_lock_irqsave(&dev->vblank_time_lock, irqflags); /* - * If the vblank interrupt was already disbled update the count + * If the vblank interrupt was already disabled update the count * and timestamp to maintain the appearance that the counter * has been ticking all along until this time. This makes the * count account for the entire time between drm_vblank_on() and -- cgit v0.10.2 From 0b6d24c01932db99fc95304235e751e7f7625c41 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 11 Apr 2014 15:55:17 +0200 Subject: drm/i915: Don't complain about stolen conflicts on gen3 Apparently stuff works that way on those machines. I agree with Chris' concern that this is a bit risky but imo worth a shot in -next just for fun. Afaics all these machines have the pci resources allocated like that by the BIOS, so I suspect that it's all ok. This regression goes back to commit eaba1b8f3379b5d100bd146b9a41d28348bdfd09 Author: Chris Wilson Date: Thu Jul 4 12:28:35 2013 +0100 drm/i915: Verify that our stolen memory doesn't conflict Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76983 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=71031 Tested-by: lu hua Signed-off-by: Daniel Vetter Reviewed-by: Jesse Barnes Tested-by: Paul Menzel Cc: stable@vger.kernel.org Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c index c388918..a204584 100644 --- a/drivers/gpu/drm/i915/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/i915_gem_stolen.c @@ -137,7 +137,11 @@ static unsigned long i915_stolen_to_physical(struct drm_device *dev) r = devm_request_mem_region(dev->dev, base + 1, dev_priv->gtt.stolen_size - 1, "Graphics Stolen Memory"); - if (r == NULL) { + /* + * GEN3 firmware likes to smash pci bridges into the stolen + * range. Apparently this works. + */ + if (r == NULL && !IS_GEN3(dev)) { DRM_ERROR("conflict detected with stolen region: [0x%08x - 0x%08x]\n", base, base + (uint32_t)dev_priv->gtt.stolen_size); base = 0; -- cgit v0.10.2 From 98533251b0bbfa5f24c502b9ab2f01ccb25c26b8 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Mon, 8 Dec 2014 17:33:51 +0000 Subject: drm/i915/bdw: Fix the write setting up the WIZ hashing mode I was playing with clang and oh surprise! a warning trigerred by -Wshift-overflow (gcc doesn't have this one): WA_SET_BIT_MASKED(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); drivers/gpu/drm/i915/intel_ringbuffer.c:786:2: warning: signed shift result (0x28002000000) requires 43 bits to represent, but 'int' only has 32 bits [-Wshift-overflow] WA_SET_BIT_MASKED(GEN7_GT_MODE, ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ drivers/gpu/drm/i915/intel_ringbuffer.c:737:15: note: expanded from macro 'WA_SET_BIT_MASKED' WA_REG(addr, _MASKED_BIT_ENABLE(mask), (mask) & 0xffff) Turned out GEN6_WIZ_HASHING_MASK was already shifted by 16, and we were trying to shift it a bit more. The other thing is that it's not the usual case of setting WA bits here, we need to have separate mask and value. To fix this, I've introduced a new _MASKED_FIELD() macro that takes both the (unshifted) mask and the desired value and the rest of the patch ripples through from it. This bug was introduced when reworking the WA emission in: Commit 7225342ab501befdb64bcec76ded41f5897c0855 Author: Mika Kuoppala Date: Tue Oct 7 17:21:26 2014 +0300 drm/i915: Build workaround list in ring initialization v2: Invert the order of the mask and value arguments (Daniel Vetter) Rewrite _MASKED_BIT_ENABLE() and _MASKED_BIT_DISABLE() with _MASKED_FIELD() (Jani Nikula) Make sure we only evaluate 'a' once in _MASKED_BIT_ENABLE() (Dave Gordon) Add check to ensure the value is within the mask boundaries (Chris Wilson) v3: Ensure the the value and mask are 16 bits (Dave Gordon) Cc: Mika Kuoppala Cc: Arun Siluvery Signed-off-by: Damien Lespiau Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 5446758..b607bbe 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -32,8 +32,19 @@ #define _PIPE3(pipe, a, b, c) ((pipe) == PIPE_A ? (a) : \ (pipe) == PIPE_B ? (b) : (c)) -#define _MASKED_BIT_ENABLE(a) (((a) << 16) | (a)) -#define _MASKED_BIT_DISABLE(a) ((a) << 16) +#define _MASKED_FIELD(mask, value) ({ \ + if (__builtin_constant_p(mask)) \ + BUILD_BUG_ON_MSG(((mask) & 0xffff0000), "Incorrect mask"); \ + if (__builtin_constant_p(value)) \ + BUILD_BUG_ON_MSG((value) & 0xffff0000, "Incorrect value"); \ + if (__builtin_constant_p(mask) && __builtin_constant_p(value)) \ + BUILD_BUG_ON_MSG((value) & ~(mask), \ + "Incorrect value for mask"); \ + (mask) << 16 | (value); }) +#define _MASKED_BIT_ENABLE(a) ({ typeof(a) _a = (a); _MASKED_FIELD(_a, _a); }) +#define _MASKED_BIT_DISABLE(a) (_MASKED_FIELD((a), 0)) + + /* PCI config space */ @@ -1282,7 +1293,7 @@ enum punit_power_well { #define GEN6_WIZ_HASHING_8x8 GEN6_WIZ_HASHING(0, 0) #define GEN6_WIZ_HASHING_8x4 GEN6_WIZ_HASHING(0, 1) #define GEN6_WIZ_HASHING_16x4 GEN6_WIZ_HASHING(1, 0) -#define GEN6_WIZ_HASHING_MASK (GEN6_WIZ_HASHING(1, 1) << 16) +#define GEN6_WIZ_HASHING_MASK GEN6_WIZ_HASHING(1, 1) #define GEN6_TD_FOUR_ROW_DISPATCH_DISABLE (1 << 5) #define GFX_MODE 0x02520 diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 9af0af4..1f4b56e 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6508,7 +6508,7 @@ static void gen6_init_clock_gating(struct drm_device *dev) * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM). */ I915_WRITE(GEN6_GT_MODE, - GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + _MASKED_FIELD(GEN6_WIZ_HASHING_MASK, GEN6_WIZ_HASHING_16x4)); ilk_init_lp_watermarks(dev); @@ -6706,7 +6706,7 @@ static void haswell_init_clock_gating(struct drm_device *dev) * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM). */ I915_WRITE(GEN7_GT_MODE, - GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + _MASKED_FIELD(GEN6_WIZ_HASHING_MASK, GEN6_WIZ_HASHING_16x4)); /* WaSwitchSolVfFArbitrationPriority:hsw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); @@ -6803,7 +6803,7 @@ static void ivybridge_init_clock_gating(struct drm_device *dev) * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM). */ I915_WRITE(GEN7_GT_MODE, - GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + _MASKED_FIELD(GEN6_WIZ_HASHING_MASK, GEN6_WIZ_HASHING_16x4)); snpcr = I915_READ(GEN6_MBCUNIT_SNPCR); snpcr &= ~GEN6_MBC_SNPCR_MASK; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 1d01b51..28db934 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -729,6 +729,9 @@ static int wa_add(struct drm_i915_private *dev_priv, #define WA_CLR_BIT_MASKED(addr, mask) \ WA_REG(addr, _MASKED_BIT_DISABLE(mask), (mask) & 0xffff) +#define WA_SET_FIELD_MASKED(addr, mask, value) \ + WA_REG(addr, _MASKED_FIELD(mask, value), mask) + #define WA_SET_BIT(addr, mask) WA_REG(addr, I915_READ(addr) | (mask), mask) #define WA_CLR_BIT(addr, mask) WA_REG(addr, I915_READ(addr) & ~(mask), mask) @@ -773,8 +776,9 @@ static int bdw_init_workarounds(struct intel_engine_cs *ring) * disable bit, which we don't touch here, but it's good * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM). */ - WA_SET_BIT_MASKED(GEN7_GT_MODE, - GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + WA_SET_FIELD_MASKED(GEN7_GT_MODE, + GEN6_WIZ_HASHING_MASK, + GEN6_WIZ_HASHING_16x4); return 0; } -- cgit v0.10.2 From a18c0af171bfb875012da26f23df051004726973 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Wed, 10 Dec 2014 11:38:49 +0100 Subject: drm: Zero out DRM object memory upon cleanup Drivers where the DRM objects have a lifetime that extends beyond that of the DRM device need to zero out the DRM object memory to void stale data such as properties. The DRM core code expects to operate on newly allocated and zeroed out objects and will behave unexpectedly, such as add duplicate properties, otherwise. Signed-off-by: Thierry Reding Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 3ce5882..632e6ea 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -725,6 +725,8 @@ void drm_crtc_cleanup(struct drm_crtc *crtc) WARN_ON(crtc->state && !crtc->funcs->atomic_destroy_state); if (crtc->state && crtc->funcs->atomic_destroy_state) crtc->funcs->atomic_destroy_state(crtc, crtc->state); + + memset(crtc, 0, sizeof(*crtc)); } EXPORT_SYMBOL(drm_crtc_cleanup); @@ -927,6 +929,8 @@ void drm_connector_cleanup(struct drm_connector *connector) if (connector->state && connector->funcs->atomic_destroy_state) connector->funcs->atomic_destroy_state(connector, connector->state); + + memset(connector, 0, sizeof(*connector)); } EXPORT_SYMBOL(drm_connector_cleanup); @@ -1068,6 +1072,8 @@ void drm_bridge_cleanup(struct drm_bridge *bridge) list_del(&bridge->head); dev->mode_config.num_bridge--; drm_modeset_unlock_all(dev); + + memset(bridge, 0, sizeof(*bridge)); } EXPORT_SYMBOL(drm_bridge_cleanup); @@ -1134,10 +1140,11 @@ void drm_encoder_cleanup(struct drm_encoder *encoder) drm_modeset_lock_all(dev); drm_mode_object_put(dev, &encoder->base); kfree(encoder->name); - encoder->name = NULL; list_del(&encoder->head); dev->mode_config.num_encoder--; drm_modeset_unlock_all(dev); + + memset(encoder, 0, sizeof(*encoder)); } EXPORT_SYMBOL(drm_encoder_cleanup); @@ -1257,6 +1264,8 @@ void drm_plane_cleanup(struct drm_plane *plane) WARN_ON(plane->state && !plane->funcs->atomic_destroy_state); if (plane->state && plane->funcs->atomic_destroy_state) plane->funcs->atomic_destroy_state(plane, plane->state); + + memset(plane, 0, sizeof(*plane)); } EXPORT_SYMBOL(drm_plane_cleanup); -- cgit v0.10.2 From cf4b0de6a3f6e1814c45206a8b175d09b265bb16 Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Mon, 8 Dec 2014 17:35:37 +0000 Subject: drm/i915: Invert the mask and val arguments in wa_add() and WA_REG() While trying to unify the order of those arguments throughout the driver, Daniel noticed what we were inverting them in this part of the code. Suggested-by: Daniel Vetter Cc: Daniel Vetter Signed-off-by: Damien Lespiau Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 28db934..ef05af0 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -701,7 +701,7 @@ static int intel_ring_workarounds_emit(struct intel_engine_cs *ring, } static int wa_add(struct drm_i915_private *dev_priv, - const u32 addr, const u32 val, const u32 mask) + const u32 addr, const u32 mask, const u32 val) { const u32 idx = dev_priv->workarounds.count; @@ -717,25 +717,25 @@ static int wa_add(struct drm_i915_private *dev_priv, return 0; } -#define WA_REG(addr, val, mask) { \ - const int r = wa_add(dev_priv, (addr), (val), (mask)); \ +#define WA_REG(addr, mask, val) { \ + const int r = wa_add(dev_priv, (addr), (mask), (val)); \ if (r) \ return r; \ } #define WA_SET_BIT_MASKED(addr, mask) \ - WA_REG(addr, _MASKED_BIT_ENABLE(mask), (mask) & 0xffff) + WA_REG(addr, (mask) & 0xffff, _MASKED_BIT_ENABLE(mask)) #define WA_CLR_BIT_MASKED(addr, mask) \ - WA_REG(addr, _MASKED_BIT_DISABLE(mask), (mask) & 0xffff) + WA_REG(addr, (mask) & 0xffff, _MASKED_BIT_DISABLE(mask)) #define WA_SET_FIELD_MASKED(addr, mask, value) \ - WA_REG(addr, _MASKED_FIELD(mask, value), mask) + WA_REG(addr, mask, _MASKED_FIELD(mask, value)) -#define WA_SET_BIT(addr, mask) WA_REG(addr, I915_READ(addr) | (mask), mask) -#define WA_CLR_BIT(addr, mask) WA_REG(addr, I915_READ(addr) & ~(mask), mask) +#define WA_SET_BIT(addr, mask) WA_REG(addr, mask, I915_READ(addr) | (mask)) +#define WA_CLR_BIT(addr, mask) WA_REG(addr, mask, I915_READ(addr) & ~(mask)) -#define WA_WRITE(addr, val) WA_REG(addr, val, 0xffffffff) +#define WA_WRITE(addr, val) WA_REG(addr, 0xffffffff, val) static int bdw_init_workarounds(struct intel_engine_cs *ring) { -- cgit v0.10.2 From 26459343e0fac1ea4cb5192c490e3e519eed74dd Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Mon, 8 Dec 2014 17:35:38 +0000 Subject: drm/i915: Remove '& 0xffff' from the mask given to WA_REG() We may be hidding bugs by doing that, so let remove it and have the actual mask value shine through, for better or worse. Signed-off-by: Damien Lespiau Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index ef05af0..9f445e9 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -724,10 +724,10 @@ static int wa_add(struct drm_i915_private *dev_priv, } #define WA_SET_BIT_MASKED(addr, mask) \ - WA_REG(addr, (mask) & 0xffff, _MASKED_BIT_ENABLE(mask)) + WA_REG(addr, (mask), _MASKED_BIT_ENABLE(mask)) #define WA_CLR_BIT_MASKED(addr, mask) \ - WA_REG(addr, (mask) & 0xffff, _MASKED_BIT_DISABLE(mask)) + WA_REG(addr, (mask), _MASKED_BIT_DISABLE(mask)) #define WA_SET_FIELD_MASKED(addr, mask, value) \ WA_REG(addr, mask, _MASKED_FIELD(mask, value)) -- cgit v0.10.2 From f41c2581bc2b6b21f774596845952a7cb4c15c74 Mon Sep 17 00:00:00 2001 From: Laurent Pinchart Date: Wed, 10 Dec 2014 21:11:31 +0200 Subject: drm/doc: Document drm_add_modes_noedid() usage And fix a spelling mistake. Signed-off-by: Laurent Pinchart Reviewed-by: Alex Deucher Signed-off-by: Dave Airlie diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl index 7a44d9d..4b592ff 100644 --- a/Documentation/DocBook/drm.tmpl +++ b/Documentation/DocBook/drm.tmpl @@ -1947,10 +1947,16 @@ void intel_crt_init(struct drm_device *dev) and then retrieves a list of modes by calling the connector get_modes helper operation. + + If the helper operation returns no mode, and if the connector status + is connector_status_connected, standard VESA DMT modes up to + 1024x768 are automatically added to the modes list by a call to + drm_add_modes_noedid. + - The function filters out modes larger than + The function then filters out modes larger than max_width and max_height - if specified. It then calls the optional connector + if specified. It finally calls the optional connector mode_valid helper operation for each mode in the probed list to check whether the mode is valid for the connector. @@ -2090,12 +2096,20 @@ void intel_crt_init(struct drm_device *dev) int (*get_modes)(struct drm_connector *connector); Fill the connector's probed_modes list - by parsing EDID data with drm_add_edid_modes or - calling drm_mode_probed_add directly for every + by parsing EDID data with drm_add_edid_modes, + adding standard VESA DMT modes with drm_add_modes_noedid, + or calling drm_mode_probed_add directly for every supported mode and return the number of modes it has detected. This operation is mandatory. + Note that the caller function will automatically add standard VESA + DMT modes up to 1024x768 if the get_modes + helper operation returns no mode and if the connector status is + connector_status_connected. There is no need to call + drm_add_edid_modes manually in that case. + + When adding modes manually the driver creates each mode with a call to drm_mode_create and must fill the following fields. @@ -2292,7 +2306,7 @@ void intel_crt_init(struct drm_device *dev) drm_helper_probe_single_connector_modes. - When parsing EDID data, drm_add_edid_modes fill the + When parsing EDID data, drm_add_edid_modes fills the connector display_info width_mm and height_mm fields. When creating modes -- cgit v0.10.2 From 41a14623bd7345017b62f167110cf95808a4891a Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Mon, 8 Sep 2014 15:52:08 +0200 Subject: drm: sti: allow to change hdmi ddc i2c adapter Depending of the board configuration i2c for ddc could change, this patch allow to use a phandle to specify which i2c controller to use. Signed-off-by: Benjamin Gaignard diff --git a/Documentation/devicetree/bindings/gpu/st,stih4xx.txt b/Documentation/devicetree/bindings/gpu/st,stih4xx.txt index 2d150c3..8885d9e 100644 --- a/Documentation/devicetree/bindings/gpu/st,stih4xx.txt +++ b/Documentation/devicetree/bindings/gpu/st,stih4xx.txt @@ -69,6 +69,7 @@ STMicroelectronics stih4xx platforms - clock-names: names of the clocks listed in clocks property in the same order. - hdmi,hpd-gpio: gpio id to detect if an hdmi cable is plugged or not. + - ddc: phandle of an I2C controller used for DDC EDID probing sti-hda: Required properties: diff --git a/drivers/gpu/drm/sti/sti_hdmi.c b/drivers/gpu/drm/sti/sti_hdmi.c index b22968c..fed1b5f 100644 --- a/drivers/gpu/drm/sti/sti_hdmi.c +++ b/drivers/gpu/drm/sti/sti_hdmi.c @@ -480,17 +480,15 @@ static const struct drm_bridge_funcs sti_hdmi_bridge_funcs = { static int sti_hdmi_connector_get_modes(struct drm_connector *connector) { - struct i2c_adapter *i2c_adap; + struct sti_hdmi_connector *hdmi_connector + = to_sti_hdmi_connector(connector); + struct sti_hdmi *hdmi = hdmi_connector->hdmi; struct edid *edid; int count; DRM_DEBUG_DRIVER("\n"); - i2c_adap = i2c_get_adapter(1); - if (!i2c_adap) - goto fail; - - edid = drm_get_edid(connector, i2c_adap); + edid = drm_get_edid(connector, hdmi->ddc_adapt); if (!edid) goto fail; @@ -603,29 +601,38 @@ static int sti_hdmi_bind(struct device *dev, struct device *master, void *data) struct sti_hdmi_connector *connector; struct drm_connector *drm_connector; struct drm_bridge *bridge; - struct i2c_adapter *i2c_adap; + struct device_node *ddc; int err; - i2c_adap = i2c_get_adapter(1); - if (!i2c_adap) - return -EPROBE_DEFER; + ddc = of_parse_phandle(dev->of_node, "ddc", 0); + if (ddc) { + hdmi->ddc_adapt = of_find_i2c_adapter_by_node(ddc); + if (!hdmi->ddc_adapt) { + err = -EPROBE_DEFER; + of_node_put(ddc); + return err; + } + + of_node_put(ddc); + } /* Set the drm device handle */ hdmi->drm_dev = drm_dev; encoder = sti_hdmi_find_encoder(drm_dev); if (!encoder) - return -ENOMEM; + goto err_adapt; connector = devm_kzalloc(dev, sizeof(*connector), GFP_KERNEL); if (!connector) - return -ENOMEM; + goto err_adapt; + connector->hdmi = hdmi; bridge = devm_kzalloc(dev, sizeof(*bridge), GFP_KERNEL); if (!bridge) - return -ENOMEM; + goto err_adapt; bridge->driver_private = hdmi; drm_bridge_init(drm_dev, bridge, &sti_hdmi_bridge_funcs); @@ -662,6 +669,8 @@ err_sysfs: err_connector: drm_bridge_cleanup(bridge); drm_connector_cleanup(drm_connector); +err_adapt: + put_device(&hdmi->ddc_adapt->dev); return -EINVAL; } @@ -788,6 +797,11 @@ static int sti_hdmi_probe(struct platform_device *pdev) static int sti_hdmi_remove(struct platform_device *pdev) { + struct sti_hdmi *hdmi = dev_get_drvdata(&pdev->dev); + + if (hdmi->ddc_adapt) + put_device(&hdmi->ddc_adapt->dev); + component_del(&pdev->dev, &sti_hdmi_ops); return 0; } diff --git a/drivers/gpu/drm/sti/sti_hdmi.h b/drivers/gpu/drm/sti/sti_hdmi.h index 61bec65..d00a3e0 100644 --- a/drivers/gpu/drm/sti/sti_hdmi.h +++ b/drivers/gpu/drm/sti/sti_hdmi.h @@ -62,6 +62,7 @@ struct sti_hdmi { wait_queue_head_t wait_event; bool event_received; struct reset_control *reset; + struct i2c_adapter *ddc_adapt; }; u32 hdmi_read(struct sti_hdmi *hdmi, int offset); -- cgit v0.10.2 From 765692078f08d0229e545d3c1a50bddbc16c800c Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 9 Oct 2014 08:53:35 +0200 Subject: drm: sti: remove gpio for HDMI hot plug detection gpio used for HDMI hot plug detection is useless, HDMI_STI register contains an hot plug detection status bit. Fix binding documentation. Signed-off-by: Benjamin Gaignard diff --git a/Documentation/devicetree/bindings/gpu/st,stih4xx.txt b/Documentation/devicetree/bindings/gpu/st,stih4xx.txt index 8885d9e..32cfc7b 100644 --- a/Documentation/devicetree/bindings/gpu/st,stih4xx.txt +++ b/Documentation/devicetree/bindings/gpu/st,stih4xx.txt @@ -68,7 +68,6 @@ STMicroelectronics stih4xx platforms number of clocks may depend of the SoC type. - clock-names: names of the clocks listed in clocks property in the same order. - - hdmi,hpd-gpio: gpio id to detect if an hdmi cable is plugged or not. - ddc: phandle of an I2C controller used for DDC EDID probing sti-hda: @@ -174,7 +173,6 @@ Example: interrupt-names = "irq"; clock-names = "pix", "tmds", "phy", "audio"; clocks = <&clockgen_c_vcc CLK_S_PIX_HDMI>, <&clockgen_c_vcc CLK_S_TMDS_HDMI>, <&clockgen_c_vcc CLK_S_HDMI_REJECT_PLL>, <&clockgen_b1 CLK_S_PCM_0>; - hdmi,hpd-gpio = <&PIO2 5>; }; sti-hda@fe85a000 { diff --git a/drivers/gpu/drm/sti/sti_hdmi.c b/drivers/gpu/drm/sti/sti_hdmi.c index fed1b5f..1921197 100644 --- a/drivers/gpu/drm/sti/sti_hdmi.c +++ b/drivers/gpu/drm/sti/sti_hdmi.c @@ -130,8 +130,7 @@ static irqreturn_t hdmi_irq_thread(int irq, void *arg) /* Hot plug/unplug IRQ */ if (hdmi->irq_status & HDMI_INT_HOT_PLUG) { - /* read gpio to get the status */ - hdmi->hpd = gpio_get_value(hdmi->hpd_gpio); + hdmi->hpd = readl(hdmi->regs + HDMI_STA) & HDMI_STA_HOT_PLUG; if (hdmi->drm_dev) drm_helper_hpd_irq_event(hdmi->drm_dev); } @@ -766,13 +765,7 @@ static int sti_hdmi_probe(struct platform_device *pdev) return PTR_ERR(hdmi->clk_audio); } - hdmi->hpd_gpio = of_get_named_gpio(np, "hdmi,hpd-gpio", 0); - if (hdmi->hpd_gpio < 0) { - DRM_ERROR("Failed to get hdmi hpd-gpio\n"); - return -EIO; - } - - hdmi->hpd = gpio_get_value(hdmi->hpd_gpio); + hdmi->hpd = readl(hdmi->regs + HDMI_STA) & HDMI_STA_HOT_PLUG; init_waitqueue_head(&hdmi->wait_event); diff --git a/drivers/gpu/drm/sti/sti_hdmi.h b/drivers/gpu/drm/sti/sti_hdmi.h index d00a3e0..3d22390 100644 --- a/drivers/gpu/drm/sti/sti_hdmi.h +++ b/drivers/gpu/drm/sti/sti_hdmi.h @@ -14,6 +14,9 @@ #define HDMI_STA 0x0010 #define HDMI_STA_DLL_LCK BIT(5) +#define HDMI_STA_HOT_PLUG_SHIFT 4 +#define HDMI_STA_HOT_PLUG (1 << HDMI_STA_HOT_PLUG_SHIFT) + struct sti_hdmi; struct hdmi_phy_ops { @@ -37,7 +40,6 @@ struct hdmi_phy_ops { * @irq_status: interrupt status register * @phy_ops: phy start/stop operations * @enabled: true if hdmi is enabled else false - * @hpd_gpio: hdmi hot plug detect gpio number * @hpd: hot plug detect status * @wait_event: wait event * @event_received: wait event status @@ -57,7 +59,6 @@ struct sti_hdmi { u32 irq_status; struct hdmi_phy_ops *phy_ops; bool enabled; - int hpd_gpio; bool hpd; wait_queue_head_t wait_event; bool event_received; -- cgit v0.10.2 From 2f7d0e82ce9fdbb450613abe28c3bbe1bda069a6 Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 4 Dec 2014 11:17:32 +0100 Subject: drm: sti: clear all mixer control Make sure that mixer control register is correctly reset before use it. Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_drm_crtc.c b/drivers/gpu/drm/sti/sti_drm_crtc.c index 36a1ad3..d42c63f 100644 --- a/drivers/gpu/drm/sti/sti_drm_crtc.c +++ b/drivers/gpu/drm/sti/sti_drm_crtc.c @@ -38,6 +38,8 @@ static void sti_drm_crtc_prepare(struct drm_crtc *crtc) if (clk_prepare_enable(compo->clk_compo_aux)) DRM_INFO("Failed to prepare/enable compo_aux clk\n"); } + + sti_mixer_clear_all_layers(mixer); } static void sti_drm_crtc_commit(struct drm_crtc *crtc) diff --git a/drivers/gpu/drm/sti/sti_mixer.c b/drivers/gpu/drm/sti/sti_mixer.c index 79f369d..6bcdf3f 100644 --- a/drivers/gpu/drm/sti/sti_mixer.c +++ b/drivers/gpu/drm/sti/sti_mixer.c @@ -215,6 +215,15 @@ int sti_mixer_set_layer_status(struct sti_mixer *mixer, return 0; } +void sti_mixer_clear_all_layers(struct sti_mixer *mixer) +{ + u32 val; + + DRM_DEBUG_DRIVER("%s clear all layer\n", sti_mixer_to_str(mixer)); + val = sti_mixer_reg_read(mixer, GAM_MIXER_CTL) & 0xFFFF0000; + sti_mixer_reg_write(mixer, GAM_MIXER_CTL, val); +} + void sti_mixer_set_matrix(struct sti_mixer *mixer) { unsigned int i; diff --git a/drivers/gpu/drm/sti/sti_mixer.h b/drivers/gpu/drm/sti/sti_mixer.h index 8743721..750e1fd 100644 --- a/drivers/gpu/drm/sti/sti_mixer.h +++ b/drivers/gpu/drm/sti/sti_mixer.h @@ -39,6 +39,7 @@ struct sti_mixer *sti_mixer_create(struct device *dev, int id, int sti_mixer_set_layer_status(struct sti_mixer *mixer, struct sti_layer *layer, bool status); +void sti_mixer_clear_all_layers(struct sti_mixer *mixer); int sti_mixer_set_layer_depth(struct sti_mixer *mixer, struct sti_layer *layer); int sti_mixer_active_video_area(struct sti_mixer *mixer, struct drm_display_mode *mode); -- cgit v0.10.2 From a51fe84d1d36122bdd7feeebd1d9d85e80ea16e7 Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 4 Dec 2014 11:21:48 +0100 Subject: drm: sti: simplify gdp code Store the physical address at node creation time to avoid use of virt_to_dma and dma_to_virt everywhere Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_gdp.c b/drivers/gpu/drm/sti/sti_gdp.c index 4e30b74..1b903ff 100644 --- a/drivers/gpu/drm/sti/sti_gdp.c +++ b/drivers/gpu/drm/sti/sti_gdp.c @@ -73,7 +73,9 @@ struct sti_gdp_node { struct sti_gdp_node_list { struct sti_gdp_node *top_field; + dma_addr_t top_field_paddr; struct sti_gdp_node *btm_field; + dma_addr_t btm_field_paddr; }; /** @@ -168,7 +170,6 @@ static int sti_gdp_get_alpharange(int format) static struct sti_gdp_node_list *sti_gdp_get_free_nodes(struct sti_layer *layer) { int hw_nvn; - void *virt_nvn; struct sti_gdp *gdp = to_sti_gdp(layer); unsigned int i; @@ -176,11 +177,9 @@ static struct sti_gdp_node_list *sti_gdp_get_free_nodes(struct sti_layer *layer) if (!hw_nvn) goto end; - virt_nvn = dma_to_virt(layer->dev, (dma_addr_t) hw_nvn); - for (i = 0; i < GDP_NODE_NB_BANK; i++) - if ((virt_nvn != gdp->node_list[i].btm_field) && - (virt_nvn != gdp->node_list[i].top_field)) + if ((hw_nvn != gdp->node_list[i].btm_field_paddr) && + (hw_nvn != gdp->node_list[i].top_field_paddr)) return &gdp->node_list[i]; /* in hazardious cases restart with the first node */ @@ -204,7 +203,6 @@ static struct sti_gdp_node_list *sti_gdp_get_current_nodes(struct sti_layer *layer) { int hw_nvn; - void *virt_nvn; struct sti_gdp *gdp = to_sti_gdp(layer); unsigned int i; @@ -212,11 +210,9 @@ struct sti_gdp_node_list *sti_gdp_get_current_nodes(struct sti_layer *layer) if (!hw_nvn) goto end; - virt_nvn = dma_to_virt(layer->dev, (dma_addr_t) hw_nvn); - for (i = 0; i < GDP_NODE_NB_BANK; i++) - if ((virt_nvn == gdp->node_list[i].btm_field) || - (virt_nvn == gdp->node_list[i].top_field)) + if ((hw_nvn == gdp->node_list[i].btm_field_paddr) || + (hw_nvn == gdp->node_list[i].top_field_paddr)) return &gdp->node_list[i]; end: @@ -292,8 +288,8 @@ static int sti_gdp_prepare_layer(struct sti_layer *layer, bool first_prepare) /* Same content and chained together */ memcpy(btm_field, top_field, sizeof(*btm_field)); - top_field->gam_gdp_nvn = virt_to_dma(dev, btm_field); - btm_field->gam_gdp_nvn = virt_to_dma(dev, top_field); + top_field->gam_gdp_nvn = list->btm_field_paddr; + btm_field->gam_gdp_nvn = list->top_field_paddr; /* Interlaced mode */ if (layer->mode->flags & DRM_MODE_FLAG_INTERLACE) @@ -349,8 +345,8 @@ static int sti_gdp_commit_layer(struct sti_layer *layer) struct sti_gdp_node *updated_top_node = updated_list->top_field; struct sti_gdp_node *updated_btm_node = updated_list->btm_field; struct sti_gdp *gdp = to_sti_gdp(layer); - u32 dma_updated_top = virt_to_dma(layer->dev, updated_top_node); - u32 dma_updated_btm = virt_to_dma(layer->dev, updated_btm_node); + u32 dma_updated_top = updated_list->top_field_paddr; + u32 dma_updated_btm = updated_list->btm_field_paddr; struct sti_gdp_node_list *curr_list = sti_gdp_get_current_nodes(layer); dev_dbg(layer->dev, "%s %s top/btm_node:0x%p/0x%p\n", __func__, @@ -461,16 +457,16 @@ static void sti_gdp_init(struct sti_layer *layer) { struct sti_gdp *gdp = to_sti_gdp(layer); struct device_node *np = layer->dev->of_node; - dma_addr_t dma; + dma_addr_t dma_addr; void *base; unsigned int i, size; /* Allocate all the nodes within a single memory page */ size = sizeof(struct sti_gdp_node) * GDP_NODE_PER_FIELD * GDP_NODE_NB_BANK; - base = dma_alloc_writecombine(layer->dev, - size, &dma, GFP_KERNEL | GFP_DMA); + size, &dma_addr, GFP_KERNEL | GFP_DMA); + if (!base) { DRM_ERROR("Failed to allocate memory for GDP node\n"); return; @@ -478,21 +474,26 @@ static void sti_gdp_init(struct sti_layer *layer) memset(base, 0, size); for (i = 0; i < GDP_NODE_NB_BANK; i++) { - if (virt_to_dma(layer->dev, base) & 0xF) { + if (dma_addr & 0xF) { DRM_ERROR("Mem alignment failed\n"); return; } gdp->node_list[i].top_field = base; + gdp->node_list[i].top_field_paddr = dma_addr; + DRM_DEBUG_DRIVER("node[%d].top_field=%p\n", i, base); base += sizeof(struct sti_gdp_node); + dma_addr += sizeof(struct sti_gdp_node); - if (virt_to_dma(layer->dev, base) & 0xF) { + if (dma_addr & 0xF) { DRM_ERROR("Mem alignment failed\n"); return; } gdp->node_list[i].btm_field = base; + gdp->node_list[i].btm_field_paddr = dma_addr; DRM_DEBUG_DRIVER("node[%d].btm_field=%p\n", i, base); base += sizeof(struct sti_gdp_node); + dma_addr += sizeof(struct sti_gdp_node); } if (of_device_is_compatible(np, "st,stih407-compositor")) { -- cgit v0.10.2 From eb929dc4d36db7881bbf90d5532b024615f64c0f Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 4 Dec 2014 12:42:37 +0100 Subject: drm: sti: remove event lock while disabling vblank Stop use event_lock in vblank disable function. This was creating a dead lock. Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_drm_crtc.c b/drivers/gpu/drm/sti/sti_drm_crtc.c index d42c63f..19e20fc 100644 --- a/drivers/gpu/drm/sti/sti_drm_crtc.c +++ b/drivers/gpu/drm/sti/sti_drm_crtc.c @@ -366,7 +366,6 @@ void sti_drm_crtc_disable_vblank(struct drm_device *dev, int crtc) struct sti_drm_private *priv = dev->dev_private; struct sti_compositor *compo = priv->compo; struct notifier_block *vtg_vblank_nb = &compo->vtg_vblank_nb; - unsigned long flags; DRM_DEBUG_DRIVER("\n"); @@ -375,13 +374,10 @@ void sti_drm_crtc_disable_vblank(struct drm_device *dev, int crtc) DRM_DEBUG_DRIVER("Warning: cannot unregister VTG notifier\n"); /* free the resources of the pending requests */ - spin_lock_irqsave(&dev->event_lock, flags); if (compo->mixer[crtc]->pending_event) { drm_vblank_put(dev, crtc); compo->mixer[crtc]->pending_event = NULL; } - spin_unlock_irqrestore(&dev->event_lock, flags); - } EXPORT_SYMBOL(sti_drm_crtc_disable_vblank); -- cgit v0.10.2 From 589b347b54932b2cd2f50845166c95f3981aa7db Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 4 Dec 2014 11:26:59 +0100 Subject: drm: sti: fix hdmi avi infoframe The hardware expect to have the infoframe checksum in the first byte. In consequence shift all infoframe on one byte. Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_hdmi.c b/drivers/gpu/drm/sti/sti_hdmi.c index 1921197..d032e02 100644 --- a/drivers/gpu/drm/sti/sti_hdmi.c +++ b/drivers/gpu/drm/sti/sti_hdmi.c @@ -272,31 +272,32 @@ static int hdmi_avi_infoframe_config(struct sti_hdmi *hdmi) hdmi_write(hdmi, val, HDMI_SW_DI_CFG); /* Infoframe header */ - val = buffer[0x0]; - val |= buffer[0x1] << 8; - val |= buffer[0x2] << 16; + val = buffer[0]; + val |= buffer[1] << 8; + val |= buffer[2] << 16; hdmi_write(hdmi, val, HDMI_SW_DI_N_HEAD_WORD(HDMI_IFRAME_SLOT_AVI)); /* Infoframe packet bytes */ - val = frame[0x0]; - val |= frame[0x1] << 8; - val |= frame[0x2] << 16; - val |= frame[0x3] << 24; + val = buffer[3]; + val |= *(frame++) << 8; + val |= *(frame++) << 16; + val |= *(frame++) << 24; hdmi_write(hdmi, val, HDMI_SW_DI_N_PKT_WORD0(HDMI_IFRAME_SLOT_AVI)); - val = frame[0x4]; - val |= frame[0x5] << 8; - val |= frame[0x6] << 16; - val |= frame[0x7] << 24; + val = *(frame++); + val |= *(frame++) << 8; + val |= *(frame++) << 16; + val |= *(frame++) << 24; hdmi_write(hdmi, val, HDMI_SW_DI_N_PKT_WORD1(HDMI_IFRAME_SLOT_AVI)); - val = frame[0x8]; - val |= frame[0x9] << 8; - val |= frame[0xA] << 16; - val |= frame[0xB] << 24; + val = *(frame++); + val |= *(frame++) << 8; + val |= *(frame++) << 16; + val |= *(frame++) << 24; hdmi_write(hdmi, val, HDMI_SW_DI_N_PKT_WORD2(HDMI_IFRAME_SLOT_AVI)); - val = frame[0xC]; + val = *(frame++); + val |= *(frame) << 8; hdmi_write(hdmi, val, HDMI_SW_DI_N_PKT_WORD3(HDMI_IFRAME_SLOT_AVI)); /* Enable transmission slot for AVI infoframe -- cgit v0.10.2 From ca614aadd7567ff074f185f3ac5f6a67466010b7 Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 4 Dec 2014 11:27:45 +0100 Subject: drm: sti: use drm_crtc_vblank_{on/off} instead of drm_vblank_{on/off} Make sure that vblank is enabled when crtc commit is call. Replace drm_vblank_off() by drm_crtc_vblank_off() Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_drm_crtc.c b/drivers/gpu/drm/sti/sti_drm_crtc.c index 19e20fc..534cd2c 100644 --- a/drivers/gpu/drm/sti/sti_drm_crtc.c +++ b/drivers/gpu/drm/sti/sti_drm_crtc.c @@ -64,6 +64,8 @@ static void sti_drm_crtc_commit(struct drm_crtc *crtc) /* Enable layer on mixer */ if (sti_mixer_set_layer_status(mixer, layer, true)) DRM_ERROR("Can not enable layer at mixer\n"); + + drm_crtc_vblank_on(crtc); } static bool sti_drm_crtc_mode_fixup(struct drm_crtc *crtc, @@ -224,7 +226,7 @@ static void sti_drm_crtc_disable(struct drm_crtc *crtc) /* Then disable layer itself */ sti_layer_disable(layer); - drm_vblank_off(crtc->dev, mixer->id); + drm_crtc_vblank_off(crtc); /* Disable pixel clock and compo IP clocks */ if (mixer->id == STI_MIXER_MAIN) { -- cgit v0.10.2 From ca279601fb988bae37339f5af064e53d8878f46b Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 4 Dec 2014 11:31:37 +0100 Subject: drm: sti: prepare sti_tvout to support auxiliary crtc Change some functions prototype to prepare the introduction of auxiliary crtc. It will also help to have a DVO encoder. Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_tvout.c b/drivers/gpu/drm/sti/sti_tvout.c index b8afe49..604e574 100644 --- a/drivers/gpu/drm/sti/sti_tvout.c +++ b/drivers/gpu/drm/sti/sti_tvout.c @@ -149,14 +149,15 @@ static void tvout_write(struct sti_tvout *tvout, u32 val, int offset) * Set the clipping mode of a VIP * * @tvout: tvout structure + * @reg: register to set * @cr_r: * @y_g: * @cb_b: */ -static void tvout_vip_set_color_order(struct sti_tvout *tvout, +static void tvout_vip_set_color_order(struct sti_tvout *tvout, int reg, u32 cr_r, u32 y_g, u32 cb_b) { - u32 val = tvout_read(tvout, TVO_VIP_HDMI); + u32 val = tvout_read(tvout, reg); val &= ~(TVO_VIP_REORDER_MASK << TVO_VIP_REORDER_R_SHIFT); val &= ~(TVO_VIP_REORDER_MASK << TVO_VIP_REORDER_G_SHIFT); @@ -165,52 +166,58 @@ static void tvout_vip_set_color_order(struct sti_tvout *tvout, val |= y_g << TVO_VIP_REORDER_G_SHIFT; val |= cb_b << TVO_VIP_REORDER_B_SHIFT; - tvout_write(tvout, val, TVO_VIP_HDMI); + tvout_write(tvout, val, reg); } /** * Set the clipping mode of a VIP * * @tvout: tvout structure + * @reg: register to set * @range: clipping range */ -static void tvout_vip_set_clip_mode(struct sti_tvout *tvout, u32 range) +static void tvout_vip_set_clip_mode(struct sti_tvout *tvout, int reg, u32 range) { - u32 val = tvout_read(tvout, TVO_VIP_HDMI); + u32 val = tvout_read(tvout, reg); val &= ~(TVO_VIP_CLIP_MASK << TVO_VIP_CLIP_SHIFT); val |= range << TVO_VIP_CLIP_SHIFT; - tvout_write(tvout, val, TVO_VIP_HDMI); + tvout_write(tvout, val, reg); } /** * Set the rounded value of a VIP * * @tvout: tvout structure + * @reg: register to set * @rnd: rounded val per component */ -static void tvout_vip_set_rnd(struct sti_tvout *tvout, u32 rnd) +static void tvout_vip_set_rnd(struct sti_tvout *tvout, int reg, u32 rnd) { - u32 val = tvout_read(tvout, TVO_VIP_HDMI); + u32 val = tvout_read(tvout, reg); val &= ~(TVO_VIP_RND_MASK << TVO_VIP_RND_SHIFT); val |= rnd << TVO_VIP_RND_SHIFT; - tvout_write(tvout, val, TVO_VIP_HDMI); + tvout_write(tvout, val, reg); } /** * Select the VIP input * * @tvout: tvout structure + * @reg: register to set + * @main_path: main or auxiliary path + * @sel_input_logic_inverted: need to invert the logic * @sel_input: selected_input (main/aux + conv) */ static void tvout_vip_set_sel_input(struct sti_tvout *tvout, + int reg, bool main_path, bool sel_input_logic_inverted, enum sti_tvout_video_out_type video_out) { u32 sel_input; - u32 val = tvout_read(tvout, TVO_VIP_HDMI); + u32 val = tvout_read(tvout, reg); if (main_path) sel_input = TVO_VIP_SEL_INPUT_MAIN; @@ -232,22 +239,24 @@ static void tvout_vip_set_sel_input(struct sti_tvout *tvout, val &= ~TVO_VIP_SEL_INPUT_MASK; val |= sel_input; - tvout_write(tvout, val, TVO_VIP_HDMI); + tvout_write(tvout, val, reg); } /** * Select the input video signed or unsigned * * @tvout: tvout structure + * @reg: register to set * @in_vid_signed: used video input format */ -static void tvout_vip_set_in_vid_fmt(struct sti_tvout *tvout, u32 in_vid_fmt) +static void tvout_vip_set_in_vid_fmt(struct sti_tvout *tvout, + int reg, u32 in_vid_fmt) { - u32 val = tvout_read(tvout, TVO_VIP_HDMI); + u32 val = tvout_read(tvout, reg); val &= ~TVO_IN_FMT_SIGNED; val |= in_vid_fmt; - tvout_write(tvout, val, TVO_MAIN_IN_VID_FORMAT); + tvout_write(tvout, val, reg); } /** @@ -261,6 +270,7 @@ static void tvout_hdmi_start(struct sti_tvout *tvout, bool main_path) { struct device_node *node = tvout->dev->of_node; bool sel_input_logic_inverted = false; + u32 tvo_in_vid_format; dev_dbg(tvout->dev, "%s\n", __func__); @@ -268,33 +278,36 @@ static void tvout_hdmi_start(struct sti_tvout *tvout, bool main_path) DRM_DEBUG_DRIVER("main vip for hdmi\n"); /* select the input sync for hdmi = VTG set 1 */ tvout_write(tvout, TVO_SYNC_MAIN_VTG_SET_1, TVO_HDMI_SYNC_SEL); + tvo_in_vid_format = TVO_MAIN_IN_VID_FORMAT; } else { DRM_DEBUG_DRIVER("aux vip for hdmi\n"); /* select the input sync for hdmi = VTG set 1 */ tvout_write(tvout, TVO_SYNC_AUX_VTG_SET_1, TVO_HDMI_SYNC_SEL); + tvo_in_vid_format = TVO_AUX_IN_VID_FORMAT; } /* set color channel order */ - tvout_vip_set_color_order(tvout, + tvout_vip_set_color_order(tvout, TVO_VIP_HDMI, TVO_VIP_REORDER_CR_R_SEL, TVO_VIP_REORDER_Y_G_SEL, TVO_VIP_REORDER_CB_B_SEL); /* set clipping mode (Limited range RGB/Y) */ - tvout_vip_set_clip_mode(tvout, TVO_VIP_CLIP_LIMITED_RANGE_RGB_Y); + tvout_vip_set_clip_mode(tvout, TVO_VIP_HDMI, + TVO_VIP_CLIP_LIMITED_RANGE_RGB_Y); /* set round mode (rounded to 8-bit per component) */ - tvout_vip_set_rnd(tvout, TVO_VIP_RND_8BIT_ROUNDED); + tvout_vip_set_rnd(tvout, TVO_VIP_HDMI, TVO_VIP_RND_8BIT_ROUNDED); if (of_device_is_compatible(node, "st,stih407-tvout")) { /* set input video format */ - tvout_vip_set_in_vid_fmt(tvout->regs + TVO_MAIN_IN_VID_FORMAT, - TVO_IN_FMT_SIGNED); + tvout_vip_set_in_vid_fmt(tvout, tvo_in_vid_format, + TVO_IN_FMT_SIGNED); sel_input_logic_inverted = true; } /* input selection */ - tvout_vip_set_sel_input(tvout, main_path, + tvout_vip_set_sel_input(tvout, TVO_VIP_HDMI, main_path, sel_input_logic_inverted, STI_TVOUT_VIDEO_OUT_RGB); } @@ -309,48 +322,47 @@ static void tvout_hda_start(struct sti_tvout *tvout, bool main_path) { struct device_node *node = tvout->dev->of_node; bool sel_input_logic_inverted = false; + u32 tvo_in_vid_format; + int val; dev_dbg(tvout->dev, "%s\n", __func__); - if (!main_path) { - DRM_ERROR("HD Analog on aux not implemented\n"); - return; + if (main_path) { + val = TVO_SYNC_MAIN_VTG_SET_2 << TVO_SYNC_HD_DCS_SHIFT; + val |= TVO_SYNC_MAIN_VTG_SET_3; + tvout_write(tvout, val, TVO_HD_SYNC_SEL); + tvo_in_vid_format = TVO_MAIN_IN_VID_FORMAT; + } else { + val = TVO_SYNC_AUX_VTG_SET_2 << TVO_SYNC_HD_DCS_SHIFT; + val |= TVO_SYNC_AUX_VTG_SET_3; + tvout_write(tvout, val, TVO_HD_SYNC_SEL); + tvo_in_vid_format = TVO_AUX_IN_VID_FORMAT; } - DRM_DEBUG_DRIVER("main vip for HDF\n"); - /* set color channel order */ - tvout_vip_set_color_order(tvout->regs + TVO_VIP_HDF, + tvout_vip_set_color_order(tvout, TVO_VIP_HDF, TVO_VIP_REORDER_CR_R_SEL, TVO_VIP_REORDER_Y_G_SEL, TVO_VIP_REORDER_CB_B_SEL); - /* set clipping mode (Limited range RGB/Y) */ - tvout_vip_set_clip_mode(tvout->regs + TVO_VIP_HDF, - TVO_VIP_CLIP_LIMITED_RANGE_CB_CR); + /* set clipping mode (EAV/SAV clipping) */ + tvout_vip_set_clip_mode(tvout, TVO_VIP_HDF, TVO_VIP_CLIP_EAV_SAV); /* set round mode (rounded to 10-bit per component) */ - tvout_vip_set_rnd(tvout->regs + TVO_VIP_HDF, TVO_VIP_RND_10BIT_ROUNDED); + tvout_vip_set_rnd(tvout, TVO_VIP_HDF, TVO_VIP_RND_10BIT_ROUNDED); if (of_device_is_compatible(node, "st,stih407-tvout")) { /* set input video format */ - tvout_vip_set_in_vid_fmt(tvout, TVO_IN_FMT_SIGNED); + tvout_vip_set_in_vid_fmt(tvout, + tvo_in_vid_format, TVO_IN_FMT_SIGNED); sel_input_logic_inverted = true; } /* Input selection */ - tvout_vip_set_sel_input(tvout->regs + TVO_VIP_HDF, - main_path, + tvout_vip_set_sel_input(tvout, TVO_VIP_HDF, main_path, sel_input_logic_inverted, STI_TVOUT_VIDEO_OUT_YUV); - /* select the input sync for HD analog = VTG set 3 - * and HD DCS = VTG set 2 */ - tvout_write(tvout, - (TVO_SYNC_MAIN_VTG_SET_2 << TVO_SYNC_HD_DCS_SHIFT) - | TVO_SYNC_MAIN_VTG_SET_3, - TVO_HD_SYNC_SEL); - /* power up HD DAC */ tvout_write(tvout, 0, TVO_HD_DAC_CFG_OFF); } -- cgit v0.10.2 From 7f2d479c017473b655b68acdf1f7335e1fac7bc5 Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Fri, 5 Dec 2014 10:14:11 +0100 Subject: drm: sti: fix delay in VTG programming The HDMI path introduce a delay of 6 pixels. This delay should be take into account while programming VTG for the HDMI. Without this delay, the HDMI active window area is shift of 6 pixel on the right. Set also timing for DVO output. Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_vtg.c b/drivers/gpu/drm/sti/sti_vtg.c index 740d6e3..9564f25 100644 --- a/drivers/gpu/drm/sti/sti_vtg.c +++ b/drivers/gpu/drm/sti/sti_vtg.c @@ -51,10 +51,19 @@ #define VTG_TOP_V_HD_3 0x010C #define VTG_BOT_V_HD_3 0x0110 +#define VTG_H_HD_4 0x0120 +#define VTG_TOP_V_VD_4 0x0124 +#define VTG_BOT_V_VD_4 0x0128 +#define VTG_TOP_V_HD_4 0x012c +#define VTG_BOT_V_HD_4 0x0130 + #define VTG_IRQ_BOTTOM BIT(0) #define VTG_IRQ_TOP BIT(1) #define VTG_IRQ_MASK (VTG_IRQ_TOP | VTG_IRQ_BOTTOM) +/* Delay introduced by the HDMI in nb of pixel */ +#define HDMI_DELAY (6) + /* delay introduced by the Arbitrary Waveform Generator in nb of pixels */ #define AWG_DELAY_HD (-9) #define AWG_DELAY_ED (-8) @@ -133,10 +142,10 @@ static void vtg_set_mode(struct sti_vtg *vtg, writel(tmp, vtg->regs + VTG_VID_TFS); writel(tmp, vtg->regs + VTG_VID_BFS); - /* prepare VTG set 1 and 2 for HDMI and VTG set 3 for HD DAC */ - tmp = (mode->hsync_end - mode->hsync_start) << 16; + /* prepare VTG set 1 for HDMI */ + tmp = (mode->hsync_end - mode->hsync_start + HDMI_DELAY) << 16; + tmp |= HDMI_DELAY; writel(tmp, vtg->regs + VTG_H_HD_1); - writel(tmp, vtg->regs + VTG_H_HD_2); tmp = (mode->vsync_end - mode->vsync_start + 1) << 16; tmp |= 1; @@ -146,6 +155,11 @@ static void vtg_set_mode(struct sti_vtg *vtg, writel(0, vtg->regs + VTG_BOT_V_HD_1); /* prepare VTG set 2 for for HD DCS */ + tmp = (mode->hsync_end - mode->hsync_start) << 16; + writel(tmp, vtg->regs + VTG_H_HD_2); + + tmp = (mode->vsync_end - mode->vsync_start + 1) << 16; + tmp |= 1; writel(tmp, vtg->regs + VTG_TOP_V_VD_2); writel(tmp, vtg->regs + VTG_BOT_V_VD_2); writel(0, vtg->regs + VTG_TOP_V_HD_2); @@ -166,6 +180,17 @@ static void vtg_set_mode(struct sti_vtg *vtg, writel(tmp, vtg->regs + VTG_TOP_V_HD_3); writel(tmp, vtg->regs + VTG_BOT_V_HD_3); + /* Prepare VTG set 4 for DVO */ + tmp = (mode->hsync_end - mode->hsync_start) << 16; + writel(tmp, vtg->regs + VTG_H_HD_4); + + tmp = (mode->vsync_end - mode->vsync_start + 1) << 16; + tmp |= 1; + writel(tmp, vtg->regs + VTG_TOP_V_VD_4); + writel(tmp, vtg->regs + VTG_BOT_V_VD_4); + writel(0, vtg->regs + VTG_TOP_V_HD_4); + writel(0, vtg->regs + VTG_BOT_V_HD_4); + /* mode */ writel(type, vtg->regs + VTG_MODE); } -- cgit v0.10.2 From 5e03abc52cd16c852552f9eaae497c9d6e55e5d0 Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Mon, 8 Dec 2014 17:32:36 +0100 Subject: drm: sti: enable auxiliary CRTC For stih407 SoC enable the second mixer to get two CRTC. Allow GPD planes and encoders to be connected to this new CRTC. Cursor plane can only be set on first CRTC. GPD clocks needed change the parent clock depending on which CRTC GPD are used. Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_compositor.c b/drivers/gpu/drm/sti/sti_compositor.c index 390d93e..bbf8462 100644 --- a/drivers/gpu/drm/sti/sti_compositor.c +++ b/drivers/gpu/drm/sti/sti_compositor.c @@ -24,14 +24,15 @@ * stiH407 compositor properties */ struct sti_compositor_data stih407_compositor_data = { - .nb_subdev = 6, + .nb_subdev = 7, .subdev_desc = { {STI_GPD_SUBDEV, (int)STI_GDP_0, 0x100}, {STI_GPD_SUBDEV, (int)STI_GDP_1, 0x200}, {STI_GPD_SUBDEV, (int)STI_GDP_2, 0x300}, {STI_GPD_SUBDEV, (int)STI_GDP_3, 0x400}, {STI_VID_SUBDEV, (int)STI_VID_0, 0x700}, - {STI_MIXER_MAIN_SUBDEV, STI_MIXER_MAIN, 0xC00} + {STI_MIXER_MAIN_SUBDEV, STI_MIXER_MAIN, 0xC00}, + {STI_MIXER_AUX_SUBDEV, STI_MIXER_AUX, 0xD00}, }, }; @@ -102,21 +103,21 @@ static int sti_compositor_bind(struct device *dev, struct device *master, enum sti_layer_type type = desc & STI_LAYER_TYPE_MASK; enum drm_plane_type plane_type = DRM_PLANE_TYPE_OVERLAY; - if (compo->mixer[crtc]) + if (crtc < compo->nb_mixers) plane_type = DRM_PLANE_TYPE_PRIMARY; switch (type) { case STI_CUR: cursor = sti_drm_plane_init(drm_dev, compo->layer[i], - (1 << crtc) - 1, - DRM_PLANE_TYPE_CURSOR); + 1, DRM_PLANE_TYPE_CURSOR); break; case STI_GDP: case STI_VID: primary = sti_drm_plane_init(drm_dev, compo->layer[i], - (1 << crtc) - 1, plane_type); + (1 << compo->nb_mixers) - 1, + plane_type); plane++; break; case STI_BCK: @@ -124,7 +125,7 @@ static int sti_compositor_bind(struct device *dev, struct device *master, } /* The first planes are reserved for primary planes*/ - if (compo->mixer[crtc]) { + if (crtc < compo->nb_mixers) { sti_drm_crtc_init(drm_dev, compo->mixer[crtc], primary, cursor); crtc++; diff --git a/drivers/gpu/drm/sti/sti_compositor.h b/drivers/gpu/drm/sti/sti_compositor.h index 3ea19db..019eb44 100644 --- a/drivers/gpu/drm/sti/sti_compositor.h +++ b/drivers/gpu/drm/sti/sti_compositor.h @@ -64,7 +64,6 @@ struct sti_compositor_data { * @layer: array of layers * @nb_mixers: number of mixers for this compositor * @nb_layers: number of layers (GDP,VID,...) for this compositor - * @enable: true if compositor is enable else false * @vtg_vblank_nb: callback for VTG VSYNC notification */ struct sti_compositor { @@ -83,7 +82,6 @@ struct sti_compositor { struct sti_layer *layer[STI_MAX_LAYER]; int nb_mixers; int nb_layers; - bool enable; struct notifier_block vtg_vblank_nb; }; diff --git a/drivers/gpu/drm/sti/sti_drm_crtc.c b/drivers/gpu/drm/sti/sti_drm_crtc.c index 534cd2c..928b44f 100644 --- a/drivers/gpu/drm/sti/sti_drm_crtc.c +++ b/drivers/gpu/drm/sti/sti_drm_crtc.c @@ -28,7 +28,7 @@ static void sti_drm_crtc_prepare(struct drm_crtc *crtc) struct device *dev = mixer->dev; struct sti_compositor *compo = dev_get_drvdata(dev); - compo->enable = true; + mixer->enabled = true; /* Prepare and enable the compo IP clock */ if (mixer->id == STI_MIXER_MAIN) { @@ -200,7 +200,7 @@ static void sti_drm_crtc_disable(struct drm_crtc *crtc) struct sti_compositor *compo = dev_get_drvdata(dev); struct sti_layer *layer; - if (!compo->enable) + if (!mixer->enabled) return; DRM_DEBUG_KMS("CRTC:%d (%s)\n", crtc->base.id, sti_mixer_to_str(mixer)); @@ -237,7 +237,7 @@ static void sti_drm_crtc_disable(struct drm_crtc *crtc) clk_disable_unprepare(compo->clk_compo_aux); } - compo->enable = false; + mixer->enabled = false; } static struct drm_crtc_helper_funcs sti_crtc_helper_funcs = { @@ -399,6 +399,7 @@ bool sti_drm_crtc_is_main(struct drm_crtc *crtc) return false; } +EXPORT_SYMBOL(sti_drm_crtc_is_main); int sti_drm_crtc_init(struct drm_device *drm_dev, struct sti_mixer *mixer, struct drm_plane *primary, struct drm_plane *cursor) diff --git a/drivers/gpu/drm/sti/sti_gdp.c b/drivers/gpu/drm/sti/sti_gdp.c index 1b903ff..32448d1 100644 --- a/drivers/gpu/drm/sti/sti_gdp.c +++ b/drivers/gpu/drm/sti/sti_gdp.c @@ -83,6 +83,8 @@ struct sti_gdp_node_list { * * @layer: layer structure * @clk_pix: pixel clock for the current gdp + * @clk_main_parent: gdp parent clock if main path used + * @clk_aux_parent: gdp parent clock if aux path used * @vtg_field_nb: callback for VTG FIELD (top or bottom) notification * @is_curr_top: true if the current node processed is the top field * @node_list: array of node list @@ -90,6 +92,8 @@ struct sti_gdp_node_list { struct sti_gdp { struct sti_layer layer; struct clk *clk_pix; + struct clk *clk_main_parent; + struct clk *clk_aux_parent; struct notifier_block vtg_field_nb; bool is_curr_top; struct sti_gdp_node_list node_list[GDP_NODE_NB_BANK]; @@ -307,6 +311,17 @@ static int sti_gdp_prepare_layer(struct sti_layer *layer, bool first_prepare) /* Set and enable gdp clock */ if (gdp->clk_pix) { + struct clk *clkp; + /* According to the mixer used, the gdp pixel clock + * should have a different parent clock. */ + if (layer->mixer_id == STI_MIXER_MAIN) + clkp = gdp->clk_main_parent; + else + clkp = gdp->clk_aux_parent; + + if (clkp) + clk_set_parent(gdp->clk_pix, clkp); + res = clk_set_rate(gdp->clk_pix, rate); if (res < 0) { DRM_ERROR("Cannot set rate (%dHz) for gdp\n", @@ -521,6 +536,14 @@ static void sti_gdp_init(struct sti_layer *layer) gdp->clk_pix = devm_clk_get(layer->dev, clk_name); if (IS_ERR(gdp->clk_pix)) DRM_ERROR("Cannot get %s clock\n", clk_name); + + gdp->clk_main_parent = devm_clk_get(layer->dev, "main_parent"); + if (IS_ERR(gdp->clk_main_parent)) + DRM_ERROR("Cannot get main_parent clock\n"); + + gdp->clk_aux_parent = devm_clk_get(layer->dev, "aux_parent"); + if (IS_ERR(gdp->clk_aux_parent)) + DRM_ERROR("Cannot get aux_parent clock\n"); } } diff --git a/drivers/gpu/drm/sti/sti_mixer.h b/drivers/gpu/drm/sti/sti_mixer.h index 750e1fd..b972821 100644 --- a/drivers/gpu/drm/sti/sti_mixer.h +++ b/drivers/gpu/drm/sti/sti_mixer.h @@ -23,6 +23,7 @@ * @id: id of the mixer * @drm_crtc: crtc object link to the mixer * @pending_event: set if a flip event is pending on crtc + * @enabled: to know if the mixer is active or not */ struct sti_mixer { struct device *dev; @@ -30,6 +31,7 @@ struct sti_mixer { int id; struct drm_crtc drm_crtc; struct drm_pending_vblank_event *pending_event; + bool enabled; }; const char *sti_mixer_to_str(struct sti_mixer *mixer); diff --git a/drivers/gpu/drm/sti/sti_tvout.c b/drivers/gpu/drm/sti/sti_tvout.c index 604e574..cb924aa 100644 --- a/drivers/gpu/drm/sti/sti_tvout.c +++ b/drivers/gpu/drm/sti/sti_tvout.c @@ -16,6 +16,8 @@ #include #include +#include "sti_drm_crtc.h" + /* glue registers */ #define TVO_CSC_MAIN_M0 0x000 #define TVO_CSC_MAIN_M1 0x004 @@ -96,7 +98,7 @@ #define TVO_SYNC_HD_DCS_SHIFT 8 -#define ENCODER_MAIN_CRTC_MASK BIT(0) +#define ENCODER_CRTC_MASK (BIT(0) | BIT(1)) /* enum listing the supported output data format */ enum sti_tvout_video_out_type { @@ -404,7 +406,7 @@ static void sti_hda_encoder_commit(struct drm_encoder *encoder) { struct sti_tvout *tvout = to_sti_tvout(encoder); - tvout_hda_start(tvout, true); + tvout_hda_start(tvout, sti_drm_crtc_is_main(encoder->crtc)); } static void sti_hda_encoder_disable(struct drm_encoder *encoder) @@ -441,7 +443,7 @@ static struct drm_encoder *sti_tvout_create_hda_encoder(struct drm_device *dev, drm_encoder = (struct drm_encoder *) encoder; - drm_encoder->possible_crtcs = ENCODER_MAIN_CRTC_MASK; + drm_encoder->possible_crtcs = ENCODER_CRTC_MASK; drm_encoder->possible_clones = 1 << 0; drm_encoder_init(dev, drm_encoder, @@ -456,7 +458,7 @@ static void sti_hdmi_encoder_commit(struct drm_encoder *encoder) { struct sti_tvout *tvout = to_sti_tvout(encoder); - tvout_hdmi_start(tvout, true); + tvout_hdmi_start(tvout, sti_drm_crtc_is_main(encoder->crtc)); } static void sti_hdmi_encoder_disable(struct drm_encoder *encoder) @@ -490,7 +492,7 @@ static struct drm_encoder *sti_tvout_create_hdmi_encoder(struct drm_device *dev, drm_encoder = (struct drm_encoder *) encoder; - drm_encoder->possible_crtcs = ENCODER_MAIN_CRTC_MASK; + drm_encoder->possible_crtcs = ENCODER_CRTC_MASK; drm_encoder->possible_clones = 1 << 1; drm_encoder_init(dev, drm_encoder, -- cgit v0.10.2 From 96006a770df80d69cbde4802e1e7cdfd70c510b3 Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 11 Dec 2014 13:34:42 +0100 Subject: drm: sti: add cursor plane stih407 SoC have a dedicated hardware cursor plane, this patch enable it. The hardware have a color look up table, fix it to be able to use ARGB8888. Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/Makefile b/drivers/gpu/drm/sti/Makefile index 04ac2ce..d6128f7f 100644 --- a/drivers/gpu/drm/sti/Makefile +++ b/drivers/gpu/drm/sti/Makefile @@ -3,6 +3,7 @@ sticompositor-y := \ sti_mixer.o \ sti_gdp.o \ sti_vid.o \ + sti_cursor.o \ sti_compositor.o \ sti_drm_crtc.o \ sti_drm_plane.o diff --git a/drivers/gpu/drm/sti/sti_compositor.c b/drivers/gpu/drm/sti/sti_compositor.c index bbf8462..b9415b3 100644 --- a/drivers/gpu/drm/sti/sti_compositor.c +++ b/drivers/gpu/drm/sti/sti_compositor.c @@ -24,8 +24,9 @@ * stiH407 compositor properties */ struct sti_compositor_data stih407_compositor_data = { - .nb_subdev = 7, + .nb_subdev = 8, .subdev_desc = { + {STI_CURSOR_SUBDEV, (int)STI_CURSOR, 0x000}, {STI_GPD_SUBDEV, (int)STI_GDP_0, 0x100}, {STI_GPD_SUBDEV, (int)STI_GDP_1, 0x200}, {STI_GPD_SUBDEV, (int)STI_GDP_2, 0x300}, @@ -68,11 +69,11 @@ static int sti_compositor_init_subdev(struct sti_compositor *compo, break; case STI_GPD_SUBDEV: case STI_VID_SUBDEV: + case STI_CURSOR_SUBDEV: compo->layer[layer_id++] = sti_layer_create(compo->dev, desc[i].id, compo->regs + desc[i].offset); break; - /* case STI_CURSOR_SUBDEV : TODO */ default: DRM_ERROR("Unknow subdev compoment type\n"); return 1; @@ -125,11 +126,12 @@ static int sti_compositor_bind(struct device *dev, struct device *master, } /* The first planes are reserved for primary planes*/ - if (crtc < compo->nb_mixers) { + if (crtc < compo->nb_mixers && primary) { sti_drm_crtc_init(drm_dev, compo->mixer[crtc], primary, cursor); crtc++; cursor = NULL; + primary = NULL; } } } diff --git a/drivers/gpu/drm/sti/sti_cursor.c b/drivers/gpu/drm/sti/sti_cursor.c new file mode 100644 index 0000000..010eaee --- /dev/null +++ b/drivers/gpu/drm/sti/sti_cursor.c @@ -0,0 +1,242 @@ +/* + * Copyright (C) STMicroelectronics SA 2014 + * Authors: Vincent Abriou + * Fabien Dessenne + * for STMicroelectronics. + * License terms: GNU General Public License (GPL), version 2 + */ +#include + +#include "sti_cursor.h" +#include "sti_layer.h" +#include "sti_vtg.h" + +/* Registers */ +#define CUR_CTL 0x00 +#define CUR_VPO 0x0C +#define CUR_PML 0x14 +#define CUR_PMP 0x18 +#define CUR_SIZE 0x1C +#define CUR_CML 0x20 +#define CUR_AWS 0x28 +#define CUR_AWE 0x2C + +#define CUR_CTL_CLUT_UPDATE BIT(1) + +#define STI_CURS_MIN_SIZE 1 +#define STI_CURS_MAX_SIZE 128 + +/* + * pixmap dma buffer stucture + * + * @paddr: physical address + * @size: buffer size + * @base: virtual address + */ +struct dma_pixmap { + dma_addr_t paddr; + size_t size; + void *base; +}; + +/** + * STI Cursor structure + * + * @layer: layer structure + * @width: cursor width + * @height: cursor height + * @clut: color look up table + * @clut_paddr: color look up table physical address + * @pixmap: pixmap dma buffer (clut8-format cursor) + */ +struct sti_cursor { + struct sti_layer layer; + unsigned int width; + unsigned int height; + unsigned short *clut; + dma_addr_t clut_paddr; + struct dma_pixmap pixmap; +}; + +static const uint32_t cursor_supported_formats[] = { + DRM_FORMAT_ARGB8888, +}; + +#define to_sti_cursor(x) container_of(x, struct sti_cursor, layer) + +static const uint32_t *sti_cursor_get_formats(struct sti_layer *layer) +{ + return cursor_supported_formats; +} + +static unsigned int sti_cursor_get_nb_formats(struct sti_layer *layer) +{ + return ARRAY_SIZE(cursor_supported_formats); +} + +static void sti_cursor_argb8888_to_clut8(struct sti_layer *layer) +{ + struct sti_cursor *cursor = to_sti_cursor(layer); + u32 *src = layer->vaddr; + u8 *dst = cursor->pixmap.base; + unsigned int i, j; + u32 a, r, g, b; + + for (i = 0; i < cursor->height; i++) { + for (j = 0; j < cursor->width; j++) { + /* Pick the 2 higher bits of each component */ + a = (*src >> 30) & 3; + r = (*src >> 22) & 3; + g = (*src >> 14) & 3; + b = (*src >> 6) & 3; + *dst = a << 6 | r << 4 | g << 2 | b; + src++; + dst++; + } + } +} + +static int sti_cursor_prepare_layer(struct sti_layer *layer, bool first_prepare) +{ + struct sti_cursor *cursor = to_sti_cursor(layer); + struct drm_display_mode *mode = layer->mode; + u32 y, x; + u32 val; + + DRM_DEBUG_DRIVER("\n"); + + dev_dbg(layer->dev, "%s %s\n", __func__, sti_layer_to_str(layer)); + + if (layer->src_w < STI_CURS_MIN_SIZE || + layer->src_h < STI_CURS_MIN_SIZE || + layer->src_w > STI_CURS_MAX_SIZE || + layer->src_h > STI_CURS_MAX_SIZE) { + DRM_ERROR("Invalid cursor size (%dx%d)\n", + layer->src_w, layer->src_h); + return -EINVAL; + } + + /* If the cursor size has changed, re-allocated the pixmap */ + if (!cursor->pixmap.base || + (cursor->width != layer->src_w) || + (cursor->height != layer->src_h)) { + cursor->width = layer->src_w; + cursor->height = layer->src_h; + + if (cursor->pixmap.base) + dma_free_writecombine(layer->dev, + cursor->pixmap.size, + cursor->pixmap.base, + cursor->pixmap.paddr); + + cursor->pixmap.size = cursor->width * cursor->height; + + cursor->pixmap.base = dma_alloc_writecombine(layer->dev, + cursor->pixmap.size, + &cursor->pixmap.paddr, + GFP_KERNEL | GFP_DMA); + if (!cursor->pixmap.base) { + DRM_ERROR("Failed to allocate memory for pixmap\n"); + return -ENOMEM; + } + } + + /* Convert ARGB8888 to CLUT8 */ + sti_cursor_argb8888_to_clut8(layer); + + /* AWS and AWE depend on the mode */ + y = sti_vtg_get_line_number(*mode, 0); + x = sti_vtg_get_pixel_number(*mode, 0); + val = y << 16 | x; + writel(val, layer->regs + CUR_AWS); + y = sti_vtg_get_line_number(*mode, mode->vdisplay - 1); + x = sti_vtg_get_pixel_number(*mode, mode->hdisplay - 1); + val = y << 16 | x; + writel(val, layer->regs + CUR_AWE); + + if (first_prepare) { + /* Set and fetch CLUT */ + writel(cursor->clut_paddr, layer->regs + CUR_CML); + writel(CUR_CTL_CLUT_UPDATE, layer->regs + CUR_CTL); + } + + return 0; +} + +static int sti_cursor_commit_layer(struct sti_layer *layer) +{ + struct sti_cursor *cursor = to_sti_cursor(layer); + struct drm_display_mode *mode = layer->mode; + u32 ydo, xdo; + + dev_dbg(layer->dev, "%s %s\n", __func__, sti_layer_to_str(layer)); + + /* Set memory location, size, and position */ + writel(cursor->pixmap.paddr, layer->regs + CUR_PML); + writel(cursor->width, layer->regs + CUR_PMP); + writel(cursor->height << 16 | cursor->width, layer->regs + CUR_SIZE); + + ydo = sti_vtg_get_line_number(*mode, layer->dst_y); + xdo = sti_vtg_get_pixel_number(*mode, layer->dst_y); + writel((ydo << 16) | xdo, layer->regs + CUR_VPO); + + return 0; +} + +static int sti_cursor_disable_layer(struct sti_layer *layer) +{ + return 0; +} + +static void sti_cursor_init(struct sti_layer *layer) +{ + struct sti_cursor *cursor = to_sti_cursor(layer); + unsigned short *base = cursor->clut; + unsigned int a, r, g, b; + + /* Assign CLUT values, ARGB444 format */ + for (a = 0; a < 4; a++) + for (r = 0; r < 4; r++) + for (g = 0; g < 4; g++) + for (b = 0; b < 4; b++) + *base++ = (a * 5) << 12 | + (r * 5) << 8 | + (g * 5) << 4 | + (b * 5); +} + +static const struct sti_layer_funcs cursor_ops = { + .get_formats = sti_cursor_get_formats, + .get_nb_formats = sti_cursor_get_nb_formats, + .init = sti_cursor_init, + .prepare = sti_cursor_prepare_layer, + .commit = sti_cursor_commit_layer, + .disable = sti_cursor_disable_layer, +}; + +struct sti_layer *sti_cursor_create(struct device *dev) +{ + struct sti_cursor *cursor; + + cursor = devm_kzalloc(dev, sizeof(*cursor), GFP_KERNEL); + if (!cursor) { + DRM_ERROR("Failed to allocate memory for cursor\n"); + return NULL; + } + + /* Allocate clut buffer */ + cursor->clut = dma_alloc_writecombine(dev, + 0x100 * sizeof(unsigned short), + &cursor->clut_paddr, + GFP_KERNEL | GFP_DMA); + + if (!cursor->clut) { + DRM_ERROR("Failed to allocate memory for cursor clut\n"); + devm_kfree(dev, cursor); + return NULL; + } + + cursor->layer.ops = &cursor_ops; + + return (struct sti_layer *)cursor; +} diff --git a/drivers/gpu/drm/sti/sti_cursor.h b/drivers/gpu/drm/sti/sti_cursor.h new file mode 100644 index 0000000..3c98274 --- /dev/null +++ b/drivers/gpu/drm/sti/sti_cursor.h @@ -0,0 +1,12 @@ +/* + * Copyright (C) STMicroelectronics SA 2013 + * Authors: Vincent Abriou for STMicroelectronics. + * License terms: GNU General Public License (GPL), version 2 + */ + +#ifndef _STI_CURSOR_H_ +#define _STI_CURSOR_H_ + +struct sti_layer *sti_cursor_create(struct device *dev); + +#endif diff --git a/drivers/gpu/drm/sti/sti_layer.c b/drivers/gpu/drm/sti/sti_layer.c index 06a587c..5051b4c 100644 --- a/drivers/gpu/drm/sti/sti_layer.c +++ b/drivers/gpu/drm/sti/sti_layer.c @@ -11,6 +11,7 @@ #include #include "sti_compositor.h" +#include "sti_cursor.h" #include "sti_gdp.h" #include "sti_layer.h" #include "sti_vid.h" @@ -50,6 +51,9 @@ struct sti_layer *sti_layer_create(struct device *dev, int desc, case STI_VID: layer = sti_vid_create(dev); break; + case STI_CUR: + layer = sti_cursor_create(dev); + break; } if (!layer) { @@ -100,6 +104,7 @@ int sti_layer_prepare(struct sti_layer *layer, struct drm_framebuffer *fb, layer->src_w = src_w; layer->src_h = src_h; layer->format = fb->pixel_format; + layer->vaddr = cma_obj->vaddr; layer->paddr = cma_obj->paddr; for (i = 0; i < 4; i++) { layer->pitches[i] = fb->pitches[i]; diff --git a/drivers/gpu/drm/sti/sti_layer.h b/drivers/gpu/drm/sti/sti_layer.h index 198c377..68bfdca 100644 --- a/drivers/gpu/drm/sti/sti_layer.h +++ b/drivers/gpu/drm/sti/sti_layer.h @@ -82,6 +82,7 @@ struct sti_layer_funcs { * @format: format * @pitches: pitch of 'planes' (eg: Y, U, V) * @offsets: offset of 'planes' + * @vaddr: virtual address of the input buffer * @paddr: physical address of the input buffer */ struct sti_layer { @@ -102,6 +103,7 @@ struct sti_layer { uint32_t format; unsigned int pitches[4]; unsigned int offsets[4]; + void *vaddr; dma_addr_t paddr; }; diff --git a/drivers/gpu/drm/sti/sti_mixer.c b/drivers/gpu/drm/sti/sti_mixer.c index 6bcdf3f..9a4ce74 100644 --- a/drivers/gpu/drm/sti/sti_mixer.c +++ b/drivers/gpu/drm/sti/sti_mixer.c @@ -45,6 +45,7 @@ static const u32 mixerColorSpaceMatIdentity[] = { #define GAM_CTL_GDP1_MASK BIT(4) #define GAM_CTL_GDP2_MASK BIT(5) #define GAM_CTL_GDP3_MASK BIT(6) +#define GAM_CTL_CURSOR_MASK BIT(9) const char *sti_mixer_to_str(struct sti_mixer *mixer) { @@ -127,6 +128,9 @@ int sti_mixer_set_layer_depth(struct sti_mixer *mixer, struct sti_layer *layer) case STI_VID_1: layer_id = GAM_DEPTH_VID1_ID; break; + case STI_CURSOR: + /* no need to set depth for cursor */ + return 0; default: DRM_ERROR("Unknown layer %d\n", layer->desc); return 1; @@ -188,6 +192,8 @@ static u32 sti_mixer_get_layer_mask(struct sti_layer *layer) return GAM_CTL_VID0_MASK; case STI_VID_1: return GAM_CTL_VID1_MASK; + case STI_CURSOR: + return GAM_CTL_CURSOR_MASK; default: return 0; } -- cgit v0.10.2 From 4fdbc678fe4dc18cbf1d97e1b45660f91c3089b4 Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 11 Dec 2014 11:38:59 +0100 Subject: drm: sti: add HQVDP plane High Quality Video Data Plane is hardware IP dedicated to video rendering. Compare to GPD (graphic planes) it have better scaler capabilities. HQVDP use VID layer to push data into hardware compositor without going into DDR. From data flow point of view HQVDP and VID are nested so HQVPD update/disable VID. Signed-off-by: Benjamin Gaignard diff --git a/Documentation/devicetree/bindings/gpu/st,stih4xx.txt b/Documentation/devicetree/bindings/gpu/st,stih4xx.txt index 32cfc7b..c99eb34 100644 --- a/Documentation/devicetree/bindings/gpu/st,stih4xx.txt +++ b/Documentation/devicetree/bindings/gpu/st,stih4xx.txt @@ -83,6 +83,22 @@ sti-hda: - clock-names: names of the clocks listed in clocks property in the same order. +sti-hqvdp: + must be a child of sti-display-subsystem + Required properties: + - compatible: "st,stih-hqvdp" + - reg: Physical base address of the IP registers and length of memory mapped region. + - clocks: from common clock binding: handle hardware IP needed clocks, the + number of clocks may depend of the SoC type. + See ../clocks/clock-bindings.txt for details. + - clock-names: names of the clocks listed in clocks property in the same + order. + - resets: resets to be used by the device + See ../reset/reset.txt for details. + - reset-names: names of the resets listed in resets property in the same + order. + - st,vtg: phandle on vtg main device node. + Example: / { @@ -183,6 +199,16 @@ Example: clocks = <&clockgen_c_vcc CLK_S_PIX_HD>, <&clockgen_c_vcc CLK_S_HDDAC>; }; }; + + sti-hqvdp@9c000000 { + compatible = "st,stih407-hqvdp"; + reg = <0x9C00000 0x100000>; + clock-names = "hqvdp", "pix_main"; + clocks = <&clk_s_c0_flexgen CLK_MAIN_DISP>, <&clk_s_d2_flexgen CLK_PIX_MAIN_DISP>; + reset-names = "hqvdp"; + resets = <&softreset STIH407_HDQVDP_SOFTRESET>; + st,vtg = <&vtg_main>; + }; }; ... }; diff --git a/drivers/gpu/drm/sti/Kconfig b/drivers/gpu/drm/sti/Kconfig index ae8850f..d6d6b70 100644 --- a/drivers/gpu/drm/sti/Kconfig +++ b/drivers/gpu/drm/sti/Kconfig @@ -5,6 +5,7 @@ config DRM_STI select DRM_KMS_HELPER select DRM_GEM_CMA_HELPER select DRM_KMS_CMA_HELPER + select FW_LOADER_USER_HELPER_FALLBACK help Choose this option to enable DRM on STM stiH41x chipset diff --git a/drivers/gpu/drm/sti/Makefile b/drivers/gpu/drm/sti/Makefile index d6128f7f..6ba9d27 100644 --- a/drivers/gpu/drm/sti/Makefile +++ b/drivers/gpu/drm/sti/Makefile @@ -19,4 +19,5 @@ obj-$(CONFIG_DRM_STI) = \ sti_hda.o \ sti_tvout.o \ sticompositor.o \ - sti_drm_drv.o \ No newline at end of file + sti_hqvdp.o \ + sti_drm_drv.o diff --git a/drivers/gpu/drm/sti/sti_compositor.c b/drivers/gpu/drm/sti/sti_compositor.c index b9415b3..c5cf4ae 100644 --- a/drivers/gpu/drm/sti/sti_compositor.c +++ b/drivers/gpu/drm/sti/sti_compositor.c @@ -122,6 +122,7 @@ static int sti_compositor_bind(struct device *dev, struct device *master, plane++; break; case STI_BCK: + case STI_VDP: break; } diff --git a/drivers/gpu/drm/sti/sti_drm_crtc.c b/drivers/gpu/drm/sti/sti_drm_crtc.c index 928b44f..4c651c2 100644 --- a/drivers/gpu/drm/sti/sti_drm_crtc.c +++ b/drivers/gpu/drm/sti/sti_drm_crtc.c @@ -148,7 +148,8 @@ sti_drm_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode, w = crtc->primary->fb->width - x; h = crtc->primary->fb->height - y; - return sti_layer_prepare(layer, crtc->primary->fb, &crtc->mode, + return sti_layer_prepare(layer, crtc, + crtc->primary->fb, &crtc->mode, mixer->id, 0, 0, w, h, x, y, w, h); } @@ -175,7 +176,8 @@ static int sti_drm_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, w = crtc->primary->fb->width - crtc->x; h = crtc->primary->fb->height - crtc->y; - ret = sti_layer_prepare(layer, crtc->primary->fb, &crtc->mode, + ret = sti_layer_prepare(layer, crtc, + crtc->primary->fb, &crtc->mode, mixer->id, 0, 0, w, h, crtc->x, crtc->y, w, h); if (ret) { diff --git a/drivers/gpu/drm/sti/sti_drm_plane.c b/drivers/gpu/drm/sti/sti_drm_plane.c index f4118d4..c9dd0e5 100644 --- a/drivers/gpu/drm/sti/sti_drm_plane.c +++ b/drivers/gpu/drm/sti/sti_drm_plane.c @@ -45,7 +45,8 @@ sti_drm_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, } /* src_x are in 16.16 format. */ - res = sti_layer_prepare(layer, fb, &crtc->mode, mixer->id, + res = sti_layer_prepare(layer, crtc, fb, + &crtc->mode, mixer->id, crtc_x, crtc_y, crtc_w, crtc_h, src_x >> 16, src_y >> 16, src_w >> 16, src_h >> 16); diff --git a/drivers/gpu/drm/sti/sti_hqvdp.c b/drivers/gpu/drm/sti/sti_hqvdp.c new file mode 100644 index 0000000..200d020 --- /dev/null +++ b/drivers/gpu/drm/sti/sti_hqvdp.c @@ -0,0 +1,1072 @@ +/* + * Copyright (C) STMicroelectronics SA 2014 + * Authors: Fabien Dessenne for STMicroelectronics. + * License terms: GNU General Public License (GPL), version 2 + */ + +#include +#include +#include +#include +#include +#include + +#include + +#include "sti_drm_plane.h" +#include "sti_hqvdp.h" +#include "sti_hqvdp_lut.h" +#include "sti_layer.h" +#include "sti_vtg.h" + +/* Firmware name */ +#define HQVDP_FMW_NAME "hqvdp-stih407.bin" + +/* Regs address */ +#define HQVDP_DMEM 0x00000000 /* 0x00000000 */ +#define HQVDP_PMEM 0x00040000 /* 0x00040000 */ +#define HQVDP_RD_PLUG 0x000E0000 /* 0x000E0000 */ +#define HQVDP_RD_PLUG_CONTROL (HQVDP_RD_PLUG + 0x1000) /* 0x000E1000 */ +#define HQVDP_RD_PLUG_PAGE_SIZE (HQVDP_RD_PLUG + 0x1004) /* 0x000E1004 */ +#define HQVDP_RD_PLUG_MIN_OPC (HQVDP_RD_PLUG + 0x1008) /* 0x000E1008 */ +#define HQVDP_RD_PLUG_MAX_OPC (HQVDP_RD_PLUG + 0x100C) /* 0x000E100C */ +#define HQVDP_RD_PLUG_MAX_CHK (HQVDP_RD_PLUG + 0x1010) /* 0x000E1010 */ +#define HQVDP_RD_PLUG_MAX_MSG (HQVDP_RD_PLUG + 0x1014) /* 0x000E1014 */ +#define HQVDP_RD_PLUG_MIN_SPACE (HQVDP_RD_PLUG + 0x1018) /* 0x000E1018 */ +#define HQVDP_WR_PLUG 0x000E2000 /* 0x000E2000 */ +#define HQVDP_WR_PLUG_CONTROL (HQVDP_WR_PLUG + 0x1000) /* 0x000E3000 */ +#define HQVDP_WR_PLUG_PAGE_SIZE (HQVDP_WR_PLUG + 0x1004) /* 0x000E3004 */ +#define HQVDP_WR_PLUG_MIN_OPC (HQVDP_WR_PLUG + 0x1008) /* 0x000E3008 */ +#define HQVDP_WR_PLUG_MAX_OPC (HQVDP_WR_PLUG + 0x100C) /* 0x000E300C */ +#define HQVDP_WR_PLUG_MAX_CHK (HQVDP_WR_PLUG + 0x1010) /* 0x000E3010 */ +#define HQVDP_WR_PLUG_MAX_MSG (HQVDP_WR_PLUG + 0x1014) /* 0x000E3014 */ +#define HQVDP_WR_PLUG_MIN_SPACE (HQVDP_WR_PLUG + 0x1018) /* 0x000E3018 */ +#define HQVDP_MBX 0x000E4000 /* 0x000E4000 */ +#define HQVDP_MBX_IRQ_TO_XP70 (HQVDP_MBX + 0x0000) /* 0x000E4000 */ +#define HQVDP_MBX_INFO_HOST (HQVDP_MBX + 0x0004) /* 0x000E4004 */ +#define HQVDP_MBX_IRQ_TO_HOST (HQVDP_MBX + 0x0008) /* 0x000E4008 */ +#define HQVDP_MBX_INFO_XP70 (HQVDP_MBX + 0x000C) /* 0x000E400C */ +#define HQVDP_MBX_SW_RESET_CTRL (HQVDP_MBX + 0x0010) /* 0x000E4010 */ +#define HQVDP_MBX_STARTUP_CTRL1 (HQVDP_MBX + 0x0014) /* 0x000E4014 */ +#define HQVDP_MBX_STARTUP_CTRL2 (HQVDP_MBX + 0x0018) /* 0x000E4018 */ +#define HQVDP_MBX_GP_STATUS (HQVDP_MBX + 0x001C) /* 0x000E401C */ +#define HQVDP_MBX_NEXT_CMD (HQVDP_MBX + 0x0020) /* 0x000E4020 */ +#define HQVDP_MBX_CURRENT_CMD (HQVDP_MBX + 0x0024) /* 0x000E4024 */ +#define HQVDP_MBX_SOFT_VSYNC (HQVDP_MBX + 0x0028) /* 0x000E4028 */ + +/* Plugs config */ +#define PLUG_CONTROL_ENABLE 0x00000001 +#define PLUG_PAGE_SIZE_256 0x00000002 +#define PLUG_MIN_OPC_8 0x00000003 +#define PLUG_MAX_OPC_64 0x00000006 +#define PLUG_MAX_CHK_2X 0x00000001 +#define PLUG_MAX_MSG_1X 0x00000000 +#define PLUG_MIN_SPACE_1 0x00000000 + +/* SW reset CTRL */ +#define SW_RESET_CTRL_FULL BIT(0) +#define SW_RESET_CTRL_CORE BIT(1) + +/* Startup ctrl 1 */ +#define STARTUP_CTRL1_RST_DONE BIT(0) +#define STARTUP_CTRL1_AUTH_IDLE BIT(2) + +/* Startup ctrl 2 */ +#define STARTUP_CTRL2_FETCH_EN BIT(1) + +/* Info xP70 */ +#define INFO_XP70_FW_READY BIT(15) +#define INFO_XP70_FW_PROCESSING BIT(14) +#define INFO_XP70_FW_INITQUEUES BIT(13) + +/* SOFT_VSYNC */ +#define SOFT_VSYNC_HW 0x00000000 +#define SOFT_VSYNC_SW_CMD 0x00000001 +#define SOFT_VSYNC_SW_CTRL_IRQ 0x00000003 + +/* Reset & boot poll config */ +#define POLL_MAX_ATTEMPT 50 +#define POLL_DELAY_MS 20 + +#define SCALE_FACTOR 8192 +#define SCALE_MAX_FOR_LEG_LUT_F 4096 +#define SCALE_MAX_FOR_LEG_LUT_E 4915 +#define SCALE_MAX_FOR_LEG_LUT_D 6654 +#define SCALE_MAX_FOR_LEG_LUT_C 8192 + +enum sti_hvsrc_orient { + HVSRC_HORI, + HVSRC_VERT +}; + +/* Command structures */ +struct sti_hqvdp_top { + u32 config; + u32 mem_format; + u32 current_luma; + u32 current_enh_luma; + u32 current_right_luma; + u32 current_enh_right_luma; + u32 current_chroma; + u32 current_enh_chroma; + u32 current_right_chroma; + u32 current_enh_right_chroma; + u32 output_luma; + u32 output_chroma; + u32 luma_src_pitch; + u32 luma_enh_src_pitch; + u32 luma_right_src_pitch; + u32 luma_enh_right_src_pitch; + u32 chroma_src_pitch; + u32 chroma_enh_src_pitch; + u32 chroma_right_src_pitch; + u32 chroma_enh_right_src_pitch; + u32 luma_processed_pitch; + u32 chroma_processed_pitch; + u32 input_frame_size; + u32 input_viewport_ori; + u32 input_viewport_ori_right; + u32 input_viewport_size; + u32 left_view_border_width; + u32 right_view_border_width; + u32 left_view_3d_offset_width; + u32 right_view_3d_offset_width; + u32 side_stripe_color; + u32 crc_reset_ctrl; +}; + +/* Configs for interlaced : no IT, no pass thru, 3 fields */ +#define TOP_CONFIG_INTER_BTM 0x00000000 +#define TOP_CONFIG_INTER_TOP 0x00000002 + +/* Config for progressive : no IT, no pass thru, 3 fields */ +#define TOP_CONFIG_PROGRESSIVE 0x00000001 + +/* Default MemFormat: in=420_raster_dual out=444_raster;opaque Mem2Tv mode */ +#define TOP_MEM_FORMAT_DFLT 0x00018060 + +/* Min/Max size */ +#define MAX_WIDTH 0x1FFF +#define MAX_HEIGHT 0x0FFF +#define MIN_WIDTH 0x0030 +#define MIN_HEIGHT 0x0010 + +struct sti_hqvdp_vc1re { + u32 ctrl_prv_csdi; + u32 ctrl_cur_csdi; + u32 ctrl_nxt_csdi; + u32 ctrl_cur_fmd; + u32 ctrl_nxt_fmd; +}; + +struct sti_hqvdp_fmd { + u32 config; + u32 viewport_ori; + u32 viewport_size; + u32 next_next_luma; + u32 next_next_right_luma; + u32 next_next_next_luma; + u32 next_next_next_right_luma; + u32 threshold_scd; + u32 threshold_rfd; + u32 threshold_move; + u32 threshold_cfd; +}; + +struct sti_hqvdp_csdi { + u32 config; + u32 config2; + u32 dcdi_config; + u32 prev_luma; + u32 prev_enh_luma; + u32 prev_right_luma; + u32 prev_enh_right_luma; + u32 next_luma; + u32 next_enh_luma; + u32 next_right_luma; + u32 next_enh_right_luma; + u32 prev_chroma; + u32 prev_enh_chroma; + u32 prev_right_chroma; + u32 prev_enh_right_chroma; + u32 next_chroma; + u32 next_enh_chroma; + u32 next_right_chroma; + u32 next_enh_right_chroma; + u32 prev_motion; + u32 prev_right_motion; + u32 cur_motion; + u32 cur_right_motion; + u32 next_motion; + u32 next_right_motion; +}; + +/* Config for progressive: by pass */ +#define CSDI_CONFIG_PROG 0x00000000 +/* Config for directional deinterlacing without motion */ +#define CSDI_CONFIG_INTER_DIR 0x00000016 +/* Additional configs for fader, blender, motion,... deinterlace algorithms */ +#define CSDI_CONFIG2_DFLT 0x000001B3 +#define CSDI_DCDI_CONFIG_DFLT 0x00203803 + +struct sti_hqvdp_hvsrc { + u32 hor_panoramic_ctrl; + u32 output_picture_size; + u32 init_horizontal; + u32 init_vertical; + u32 param_ctrl; + u32 yh_coef[NB_COEF]; + u32 ch_coef[NB_COEF]; + u32 yv_coef[NB_COEF]; + u32 cv_coef[NB_COEF]; + u32 hori_shift; + u32 vert_shift; +}; + +/* Default ParamCtrl: all controls enabled */ +#define HVSRC_PARAM_CTRL_DFLT 0xFFFFFFFF + +struct sti_hqvdp_iqi { + u32 config; + u32 demo_wind_size; + u32 pk_config; + u32 coeff0_coeff1; + u32 coeff2_coeff3; + u32 coeff4; + u32 pk_lut; + u32 pk_gain; + u32 pk_coring_level; + u32 cti_config; + u32 le_config; + u32 le_lut[64]; + u32 con_bri; + u32 sat_gain; + u32 pxf_conf; + u32 default_color; +}; + +/* Default Config : IQI bypassed */ +#define IQI_CONFIG_DFLT 0x00000001 +/* Default Contrast & Brightness gain = 256 */ +#define IQI_CON_BRI_DFLT 0x00000100 +/* Default Saturation gain = 256 */ +#define IQI_SAT_GAIN_DFLT 0x00000100 +/* Default PxfConf : P2I bypassed */ +#define IQI_PXF_CONF_DFLT 0x00000001 + +struct sti_hqvdp_top_status { + u32 processing_time; + u32 input_y_crc; + u32 input_uv_crc; +}; + +struct sti_hqvdp_fmd_status { + u32 fmd_repeat_move_status; + u32 fmd_scene_count_status; + u32 cfd_sum; + u32 field_sum; + u32 next_y_fmd_crc; + u32 next_next_y_fmd_crc; + u32 next_next_next_y_fmd_crc; +}; + +struct sti_hqvdp_csdi_status { + u32 prev_y_csdi_crc; + u32 cur_y_csdi_crc; + u32 next_y_csdi_crc; + u32 prev_uv_csdi_crc; + u32 cur_uv_csdi_crc; + u32 next_uv_csdi_crc; + u32 y_csdi_crc; + u32 uv_csdi_crc; + u32 uv_cup_crc; + u32 mot_csdi_crc; + u32 mot_cur_csdi_crc; + u32 mot_prev_csdi_crc; +}; + +struct sti_hqvdp_hvsrc_status { + u32 y_hvsrc_crc; + u32 u_hvsrc_crc; + u32 v_hvsrc_crc; +}; + +struct sti_hqvdp_iqi_status { + u32 pxf_it_status; + u32 y_iqi_crc; + u32 u_iqi_crc; + u32 v_iqi_crc; +}; + +/* Main commands. We use 2 commands one being processed by the firmware, one + * ready to be fetched upon next Vsync*/ +#define NB_VDP_CMD 2 + +struct sti_hqvdp_cmd { + struct sti_hqvdp_top top; + struct sti_hqvdp_vc1re vc1re; + struct sti_hqvdp_fmd fmd; + struct sti_hqvdp_csdi csdi; + struct sti_hqvdp_hvsrc hvsrc; + struct sti_hqvdp_iqi iqi; + struct sti_hqvdp_top_status top_status; + struct sti_hqvdp_fmd_status fmd_status; + struct sti_hqvdp_csdi_status csdi_status; + struct sti_hqvdp_hvsrc_status hvsrc_status; + struct sti_hqvdp_iqi_status iqi_status; +}; + +/* + * STI HQVDP structure + * + * @dev: driver device + * @drm_dev: the drm device + * @regs: registers + * @layer: layer structure for hqvdp it self + * @vid_plane: VID plug used as link with compositor IP + * @clk: IP clock + * @clk_pix_main: pix main clock + * @reset: reset control + * @vtg_nb: notifier to handle VTG Vsync + * @btm_field_pending: is there any bottom field (interlaced frame) to display + * @curr_field_count: number of field updates + * @last_field_count: number of field updates since last fps measure + * @hqvdp_cmd: buffer of commands + * @hqvdp_cmd_paddr: physical address of hqvdp_cmd + * @vtg: vtg for main data path + */ +struct sti_hqvdp { + struct device *dev; + struct drm_device *drm_dev; + void __iomem *regs; + struct sti_layer layer; + struct drm_plane *vid_plane; + struct clk *clk; + struct clk *clk_pix_main; + struct reset_control *reset; + struct notifier_block vtg_nb; + bool btm_field_pending; + unsigned int curr_field_count; + unsigned int last_field_count; + void *hqvdp_cmd; + dma_addr_t hqvdp_cmd_paddr; + struct sti_vtg *vtg; +}; + +#define to_sti_hqvdp(x) container_of(x, struct sti_hqvdp, layer) + +static const uint32_t hqvdp_supported_formats[] = { + DRM_FORMAT_NV12, +}; + +static const uint32_t *sti_hqvdp_get_formats(struct sti_layer *layer) +{ + return hqvdp_supported_formats; +} + +static unsigned int sti_hqvdp_get_nb_formats(struct sti_layer *layer) +{ + return ARRAY_SIZE(hqvdp_supported_formats); +} + +/** + * sti_hqvdp_get_free_cmd + * @hqvdp: hqvdp structure + * + * Look for a hqvdp_cmd that is not being used (or about to be used) by the FW. + * + * RETURNS: + * the offset of the command to be used. + * -1 in error cases + */ +static int sti_hqvdp_get_free_cmd(struct sti_hqvdp *hqvdp) +{ + int curr_cmd, next_cmd; + dma_addr_t cmd = hqvdp->hqvdp_cmd_paddr; + int i; + + curr_cmd = readl(hqvdp->regs + HQVDP_MBX_CURRENT_CMD); + next_cmd = readl(hqvdp->regs + HQVDP_MBX_NEXT_CMD); + + for (i = 0; i < NB_VDP_CMD; i++) { + if ((cmd != curr_cmd) && (cmd != next_cmd)) + return i * sizeof(struct sti_hqvdp_cmd); + cmd += sizeof(struct sti_hqvdp_cmd); + } + + return -1; +} + +/** + * sti_hqvdp_get_curr_cmd + * @hqvdp: hqvdp structure + * + * Look for the hqvdp_cmd that is being used by the FW. + * + * RETURNS: + * the offset of the command to be used. + * -1 in error cases + */ +static int sti_hqvdp_get_curr_cmd(struct sti_hqvdp *hqvdp) +{ + int curr_cmd; + dma_addr_t cmd = hqvdp->hqvdp_cmd_paddr; + unsigned int i; + + curr_cmd = readl(hqvdp->regs + HQVDP_MBX_CURRENT_CMD); + + for (i = 0; i < NB_VDP_CMD; i++) { + if (cmd == curr_cmd) + return i * sizeof(struct sti_hqvdp_cmd); + + cmd += sizeof(struct sti_hqvdp_cmd); + } + + return -1; +} + +/** + * sti_hqvdp_update_hvsrc + * @orient: horizontal or vertical + * @scale: scaling/zoom factor + * @hvsrc: the structure containing the LUT coef + * + * Update the Y and C Lut coef, as well as the shift param + * + * RETURNS: + * None. + */ +static void sti_hqvdp_update_hvsrc(enum sti_hvsrc_orient orient, int scale, + struct sti_hqvdp_hvsrc *hvsrc) +{ + const int *coef_c, *coef_y; + int shift_c, shift_y; + + /* Get the appropriate coef tables */ + if (scale < SCALE_MAX_FOR_LEG_LUT_F) { + coef_y = coef_lut_f_y_legacy; + coef_c = coef_lut_f_c_legacy; + shift_y = SHIFT_LUT_F_Y_LEGACY; + shift_c = SHIFT_LUT_F_C_LEGACY; + } else if (scale < SCALE_MAX_FOR_LEG_LUT_E) { + coef_y = coef_lut_e_y_legacy; + coef_c = coef_lut_e_c_legacy; + shift_y = SHIFT_LUT_E_Y_LEGACY; + shift_c = SHIFT_LUT_E_C_LEGACY; + } else if (scale < SCALE_MAX_FOR_LEG_LUT_D) { + coef_y = coef_lut_d_y_legacy; + coef_c = coef_lut_d_c_legacy; + shift_y = SHIFT_LUT_D_Y_LEGACY; + shift_c = SHIFT_LUT_D_C_LEGACY; + } else if (scale < SCALE_MAX_FOR_LEG_LUT_C) { + coef_y = coef_lut_c_y_legacy; + coef_c = coef_lut_c_c_legacy; + shift_y = SHIFT_LUT_C_Y_LEGACY; + shift_c = SHIFT_LUT_C_C_LEGACY; + } else if (scale == SCALE_MAX_FOR_LEG_LUT_C) { + coef_y = coef_c = coef_lut_b; + shift_y = shift_c = SHIFT_LUT_B; + } else { + coef_y = coef_c = coef_lut_a_legacy; + shift_y = shift_c = SHIFT_LUT_A_LEGACY; + } + + if (orient == HVSRC_HORI) { + hvsrc->hori_shift = (shift_c << 16) | shift_y; + memcpy(hvsrc->yh_coef, coef_y, sizeof(hvsrc->yh_coef)); + memcpy(hvsrc->ch_coef, coef_c, sizeof(hvsrc->ch_coef)); + } else { + hvsrc->vert_shift = (shift_c << 16) | shift_y; + memcpy(hvsrc->yv_coef, coef_y, sizeof(hvsrc->yv_coef)); + memcpy(hvsrc->cv_coef, coef_c, sizeof(hvsrc->cv_coef)); + } +} + +/** + * sti_hqvdp_check_hw_scaling + * @layer: hqvdp layer + * + * Check if the HW is able to perform the scaling request + * The firmware scaling limitation is "CEIL(1/Zy) <= FLOOR(LFW)" where: + * Zy = OutputHeight / InputHeight + * LFW = (Tx * IPClock) / (MaxNbCycles * Cp) + * Tx : Total video mode horizontal resolution + * IPClock : HQVDP IP clock (Mhz) + * MaxNbCycles: max(InputWidth, OutputWidth) + * Cp: Video mode pixel clock (Mhz) + * + * RETURNS: + * True if the HW can scale. + */ +static bool sti_hqvdp_check_hw_scaling(struct sti_layer *layer) +{ + struct sti_hqvdp *hqvdp = to_sti_hqvdp(layer); + unsigned long lfw; + unsigned int inv_zy; + + lfw = layer->mode->htotal * (clk_get_rate(hqvdp->clk) / 1000000); + lfw /= max(layer->src_w, layer->dst_w) * layer->mode->clock / 1000; + + inv_zy = DIV_ROUND_UP(layer->src_h, layer->dst_h); + + return (inv_zy <= lfw) ? true : false; +} + +/** + * sti_hqvdp_prepare_layer + * @layer: hqvdp layer + * @first_prepare: true if it is the first time this function is called + * + * Prepares a command for the firmware + * + * RETURNS: + * 0 on success. + */ +static int sti_hqvdp_prepare_layer(struct sti_layer *layer, bool first_prepare) +{ + struct sti_hqvdp *hqvdp = to_sti_hqvdp(layer); + struct sti_hqvdp_cmd *cmd; + int scale_h, scale_v; + int cmd_offset; + + dev_dbg(hqvdp->dev, "%s %s\n", __func__, sti_layer_to_str(layer)); + + /* prepare and commit VID plane */ + hqvdp->vid_plane->funcs->update_plane(hqvdp->vid_plane, + layer->crtc, layer->fb, + layer->dst_x, layer->dst_y, + layer->dst_w, layer->dst_h, + layer->src_x, layer->src_y, + layer->src_w, layer->src_h); + + cmd_offset = sti_hqvdp_get_free_cmd(hqvdp); + if (cmd_offset == -1) { + DRM_ERROR("No available hqvdp_cmd now\n"); + return -EBUSY; + } + cmd = hqvdp->hqvdp_cmd + cmd_offset; + + if (!sti_hqvdp_check_hw_scaling(layer)) { + DRM_ERROR("Scaling beyond HW capabilities\n"); + return -EINVAL; + } + + /* Static parameters, defaulting to progressive mode */ + cmd->top.config = TOP_CONFIG_PROGRESSIVE; + cmd->top.mem_format = TOP_MEM_FORMAT_DFLT; + cmd->hvsrc.param_ctrl = HVSRC_PARAM_CTRL_DFLT; + cmd->csdi.config = CSDI_CONFIG_PROG; + + /* VC1RE, FMD bypassed : keep everything set to 0 + * IQI/P2I bypassed */ + cmd->iqi.config = IQI_CONFIG_DFLT; + cmd->iqi.con_bri = IQI_CON_BRI_DFLT; + cmd->iqi.sat_gain = IQI_SAT_GAIN_DFLT; + cmd->iqi.pxf_conf = IQI_PXF_CONF_DFLT; + + /* Buffer planes address */ + cmd->top.current_luma = (u32) layer->paddr + layer->offsets[0]; + cmd->top.current_chroma = (u32) layer->paddr + layer->offsets[1]; + + /* Pitches */ + cmd->top.luma_processed_pitch = cmd->top.luma_src_pitch = + layer->pitches[0]; + cmd->top.chroma_processed_pitch = cmd->top.chroma_src_pitch = + layer->pitches[1]; + + /* Input / output size + * Align to upper even value */ + layer->dst_w = ALIGN(layer->dst_w, 2); + layer->dst_h = ALIGN(layer->dst_h, 2); + + if ((layer->src_w > MAX_WIDTH) || (layer->src_w < MIN_WIDTH) || + (layer->src_h > MAX_HEIGHT) || (layer->src_h < MIN_HEIGHT) || + (layer->dst_w > MAX_WIDTH) || (layer->dst_w < MIN_WIDTH) || + (layer->dst_h > MAX_HEIGHT) || (layer->dst_h < MIN_HEIGHT)) { + DRM_ERROR("Invalid in/out size %dx%d -> %dx%d\n", + layer->src_w, layer->src_h, + layer->dst_w, layer->dst_h); + return -EINVAL; + } + cmd->top.input_viewport_size = cmd->top.input_frame_size = + layer->src_h << 16 | layer->src_w; + cmd->hvsrc.output_picture_size = layer->dst_h << 16 | layer->dst_w; + cmd->top.input_viewport_ori = layer->src_y << 16 | layer->src_x; + + /* Handle interlaced */ + if (layer->fb->flags & DRM_MODE_FB_INTERLACED) { + /* Top field to display */ + cmd->top.config = TOP_CONFIG_INTER_TOP; + + /* Update pitches and vert size */ + cmd->top.input_frame_size = (layer->src_h / 2) << 16 | + layer->src_w; + cmd->top.luma_processed_pitch *= 2; + cmd->top.luma_src_pitch *= 2; + cmd->top.chroma_processed_pitch *= 2; + cmd->top.chroma_src_pitch *= 2; + + /* Enable directional deinterlacing processing */ + cmd->csdi.config = CSDI_CONFIG_INTER_DIR; + cmd->csdi.config2 = CSDI_CONFIG2_DFLT; + cmd->csdi.dcdi_config = CSDI_DCDI_CONFIG_DFLT; + } + + /* Update hvsrc lut coef */ + scale_h = SCALE_FACTOR * layer->dst_w / layer->src_w; + sti_hqvdp_update_hvsrc(HVSRC_HORI, scale_h, &cmd->hvsrc); + + scale_v = SCALE_FACTOR * layer->dst_h / layer->src_h; + sti_hqvdp_update_hvsrc(HVSRC_VERT, scale_v, &cmd->hvsrc); + + if (first_prepare) { + /* Prevent VTG shutdown */ + if (clk_prepare_enable(hqvdp->clk_pix_main)) { + DRM_ERROR("Failed to prepare/enable pix main clk\n"); + return -ENXIO; + } + + /* Register VTG Vsync callback to handle bottom fields */ + if ((layer->fb->flags & DRM_MODE_FB_INTERLACED) && + sti_vtg_register_client(hqvdp->vtg, + &hqvdp->vtg_nb, layer->mixer_id)) { + DRM_ERROR("Cannot register VTG notifier\n"); + return -ENXIO; + } + } + + return 0; +} + +static int sti_hqvdp_commit_layer(struct sti_layer *layer) +{ + struct sti_hqvdp *hqvdp = to_sti_hqvdp(layer); + int cmd_offset; + + dev_dbg(hqvdp->dev, "%s %s\n", __func__, sti_layer_to_str(layer)); + + cmd_offset = sti_hqvdp_get_free_cmd(hqvdp); + if (cmd_offset == -1) { + DRM_ERROR("No available hqvdp_cmd now\n"); + return -EBUSY; + } + + writel(hqvdp->hqvdp_cmd_paddr + cmd_offset, + hqvdp->regs + HQVDP_MBX_NEXT_CMD); + + hqvdp->curr_field_count++; + + /* Interlaced : get ready to display the bottom field at next Vsync */ + if (layer->fb->flags & DRM_MODE_FB_INTERLACED) + hqvdp->btm_field_pending = true; + + dev_dbg(hqvdp->dev, "%s Posted command:0x%x\n", + __func__, hqvdp->hqvdp_cmd_paddr + cmd_offset); + + return 0; +} + +static int sti_hqvdp_disable_layer(struct sti_layer *layer) +{ + struct sti_hqvdp *hqvdp = to_sti_hqvdp(layer); + int i; + + DRM_DEBUG_DRIVER("%s\n", sti_layer_to_str(layer)); + + /* Unregister VTG Vsync callback */ + if ((layer->fb->flags & DRM_MODE_FB_INTERLACED) && + sti_vtg_unregister_client(hqvdp->vtg, &hqvdp->vtg_nb)) + DRM_DEBUG_DRIVER("Warning: cannot unregister VTG notifier\n"); + + /* Set next cmd to NULL */ + writel(0, hqvdp->regs + HQVDP_MBX_NEXT_CMD); + + for (i = 0; i < POLL_MAX_ATTEMPT; i++) { + if (readl(hqvdp->regs + HQVDP_MBX_INFO_XP70) + & INFO_XP70_FW_READY) + break; + msleep(POLL_DELAY_MS); + } + + /* VTG can stop now */ + clk_disable_unprepare(hqvdp->clk_pix_main); + + if (i == POLL_MAX_ATTEMPT) { + DRM_ERROR("XP70 could not revert to idle\n"); + return -ENXIO; + } + + /* disable VID plane */ + hqvdp->vid_plane->funcs->disable_plane(hqvdp->vid_plane); + + return 0; +} + +/** + * sti_vdp_vtg_cb + * @nb: notifier block + * @evt: event message + * @data: private data + * + * Handle VTG Vsync event, display pending bottom field + * + * RETURNS: + * 0 on success. + */ +int sti_hqvdp_vtg_cb(struct notifier_block *nb, unsigned long evt, void *data) +{ + struct sti_hqvdp *hqvdp = container_of(nb, struct sti_hqvdp, vtg_nb); + int btm_cmd_offset, top_cmd_offest; + struct sti_hqvdp_cmd *btm_cmd, *top_cmd; + + if ((evt != VTG_TOP_FIELD_EVENT) && (evt != VTG_BOTTOM_FIELD_EVENT)) { + DRM_DEBUG_DRIVER("Unknown event\n"); + return 0; + } + + if (hqvdp->btm_field_pending) { + /* Create the btm field command from the current one */ + btm_cmd_offset = sti_hqvdp_get_free_cmd(hqvdp); + top_cmd_offest = sti_hqvdp_get_curr_cmd(hqvdp); + if ((btm_cmd_offset == -1) || (top_cmd_offest == -1)) { + DRM_ERROR("Cannot get cmds, skip btm field\n"); + return -EBUSY; + } + + btm_cmd = hqvdp->hqvdp_cmd + btm_cmd_offset; + top_cmd = hqvdp->hqvdp_cmd + top_cmd_offest; + + memcpy(btm_cmd, top_cmd, sizeof(*btm_cmd)); + + btm_cmd->top.config = TOP_CONFIG_INTER_BTM; + btm_cmd->top.current_luma += + btm_cmd->top.luma_src_pitch / 2; + btm_cmd->top.current_chroma += + btm_cmd->top.chroma_src_pitch / 2; + + /* Post the command to mailbox */ + writel(hqvdp->hqvdp_cmd_paddr + btm_cmd_offset, + hqvdp->regs + HQVDP_MBX_NEXT_CMD); + + hqvdp->curr_field_count++; + hqvdp->btm_field_pending = false; + + dev_dbg(hqvdp->dev, "%s Posted command:0x%x\n", + __func__, hqvdp->hqvdp_cmd_paddr); + } + + return 0; +} + +static struct drm_plane *sti_hqvdp_find_vid(struct drm_device *dev, int id) +{ + struct drm_plane *plane; + + list_for_each_entry(plane, &dev->mode_config.plane_list, head) { + struct sti_layer *layer = to_sti_layer(plane); + + if (layer->desc == id) + return plane; + } + + return NULL; +} + +static void sti_hqvd_init(struct sti_layer *layer) +{ + struct sti_hqvdp *hqvdp = to_sti_hqvdp(layer); + int size; + + /* find the plane macthing with vid 0 */ + hqvdp->vid_plane = sti_hqvdp_find_vid(hqvdp->drm_dev, STI_VID_0); + if (!hqvdp->vid_plane) { + DRM_ERROR("Cannot find Main video layer\n"); + return; + } + + hqvdp->vtg_nb.notifier_call = sti_hqvdp_vtg_cb; + + /* Allocate memory for the VDP commands */ + size = NB_VDP_CMD * sizeof(struct sti_hqvdp_cmd); + hqvdp->hqvdp_cmd = dma_alloc_writecombine(hqvdp->dev, size, + &hqvdp->hqvdp_cmd_paddr, + GFP_KERNEL | GFP_DMA); + if (!hqvdp->hqvdp_cmd) { + DRM_ERROR("Failed to allocate memory for VDP cmd\n"); + return; + } + + memset(hqvdp->hqvdp_cmd, 0, size); +} + +static const struct sti_layer_funcs hqvdp_ops = { + .get_formats = sti_hqvdp_get_formats, + .get_nb_formats = sti_hqvdp_get_nb_formats, + .init = sti_hqvd_init, + .prepare = sti_hqvdp_prepare_layer, + .commit = sti_hqvdp_commit_layer, + .disable = sti_hqvdp_disable_layer, +}; + +struct sti_layer *sti_hqvdp_create(struct device *dev) +{ + struct sti_hqvdp *hqvdp = dev_get_drvdata(dev); + + hqvdp->layer.ops = &hqvdp_ops; + + return &hqvdp->layer; +} + +static void sti_hqvdp_init_plugs(struct sti_hqvdp *hqvdp) +{ + /* Configure Plugs (same for RD & WR) */ + writel(PLUG_PAGE_SIZE_256, hqvdp->regs + HQVDP_RD_PLUG_PAGE_SIZE); + writel(PLUG_MIN_OPC_8, hqvdp->regs + HQVDP_RD_PLUG_MIN_OPC); + writel(PLUG_MAX_OPC_64, hqvdp->regs + HQVDP_RD_PLUG_MAX_OPC); + writel(PLUG_MAX_CHK_2X, hqvdp->regs + HQVDP_RD_PLUG_MAX_CHK); + writel(PLUG_MAX_MSG_1X, hqvdp->regs + HQVDP_RD_PLUG_MAX_MSG); + writel(PLUG_MIN_SPACE_1, hqvdp->regs + HQVDP_RD_PLUG_MIN_SPACE); + writel(PLUG_CONTROL_ENABLE, hqvdp->regs + HQVDP_RD_PLUG_CONTROL); + + writel(PLUG_PAGE_SIZE_256, hqvdp->regs + HQVDP_WR_PLUG_PAGE_SIZE); + writel(PLUG_MIN_OPC_8, hqvdp->regs + HQVDP_WR_PLUG_MIN_OPC); + writel(PLUG_MAX_OPC_64, hqvdp->regs + HQVDP_WR_PLUG_MAX_OPC); + writel(PLUG_MAX_CHK_2X, hqvdp->regs + HQVDP_WR_PLUG_MAX_CHK); + writel(PLUG_MAX_MSG_1X, hqvdp->regs + HQVDP_WR_PLUG_MAX_MSG); + writel(PLUG_MIN_SPACE_1, hqvdp->regs + HQVDP_WR_PLUG_MIN_SPACE); + writel(PLUG_CONTROL_ENABLE, hqvdp->regs + HQVDP_WR_PLUG_CONTROL); +} + +/** + * sti_hqvdp_start_xp70 + * @firmware: firmware found + * @ctxt: hqvdp structure + * + * Run the xP70 initialization sequence + */ +static void sti_hqvdp_start_xp70(const struct firmware *firmware, void *ctxt) +{ + struct sti_hqvdp *hqvdp = ctxt; + u32 *fw_rd_plug, *fw_wr_plug, *fw_pmem, *fw_dmem; + u8 *data; + int i; + struct fw_header { + int rd_size; + int wr_size; + int pmem_size; + int dmem_size; + } *header; + + DRM_DEBUG_DRIVER("\n"); + /* Check firmware parts */ + if (!firmware) { + DRM_ERROR("Firmware not available\n"); + return; + } + + header = (struct fw_header *) firmware->data; + if (firmware->size < sizeof(*header)) { + DRM_ERROR("Invalid firmware size (%d)\n", firmware->size); + goto out; + } + if ((sizeof(*header) + header->rd_size + header->wr_size + + header->pmem_size + header->dmem_size) != firmware->size) { + DRM_ERROR("Invalid fmw structure (%d+%d+%d+%d+%d != %d)\n", + sizeof(*header), header->rd_size, header->wr_size, + header->pmem_size, header->dmem_size, + firmware->size); + goto out; + } + + data = (u8 *) firmware->data; + data += sizeof(*header); + fw_rd_plug = (void *) data; + data += header->rd_size; + fw_wr_plug = (void *) data; + data += header->wr_size; + fw_pmem = (void *) data; + data += header->pmem_size; + fw_dmem = (void *) data; + + /* Enable clock */ + if (clk_prepare_enable(hqvdp->clk)) + DRM_ERROR("Failed to prepare/enable HQVDP clk\n"); + + /* Reset */ + writel(SW_RESET_CTRL_FULL, hqvdp->regs + HQVDP_MBX_SW_RESET_CTRL); + + for (i = 0; i < POLL_MAX_ATTEMPT; i++) { + if (readl(hqvdp->regs + HQVDP_MBX_STARTUP_CTRL1) + & STARTUP_CTRL1_RST_DONE) + break; + msleep(POLL_DELAY_MS); + } + if (i == POLL_MAX_ATTEMPT) { + DRM_ERROR("Could not reset\n"); + goto out; + } + + /* Init Read & Write plugs */ + for (i = 0; i < header->rd_size / 4; i++) + writel(fw_rd_plug[i], hqvdp->regs + HQVDP_RD_PLUG + i * 4); + for (i = 0; i < header->wr_size / 4; i++) + writel(fw_wr_plug[i], hqvdp->regs + HQVDP_WR_PLUG + i * 4); + + sti_hqvdp_init_plugs(hqvdp); + + /* Authorize Idle Mode */ + writel(STARTUP_CTRL1_AUTH_IDLE, hqvdp->regs + HQVDP_MBX_STARTUP_CTRL1); + + /* Prevent VTG interruption during the boot */ + writel(SOFT_VSYNC_SW_CTRL_IRQ, hqvdp->regs + HQVDP_MBX_SOFT_VSYNC); + writel(0, hqvdp->regs + HQVDP_MBX_NEXT_CMD); + + /* Download PMEM & DMEM */ + for (i = 0; i < header->pmem_size / 4; i++) + writel(fw_pmem[i], hqvdp->regs + HQVDP_PMEM + i * 4); + for (i = 0; i < header->dmem_size / 4; i++) + writel(fw_dmem[i], hqvdp->regs + HQVDP_DMEM + i * 4); + + /* Enable fetch */ + writel(STARTUP_CTRL2_FETCH_EN, hqvdp->regs + HQVDP_MBX_STARTUP_CTRL2); + + /* Wait end of boot */ + for (i = 0; i < POLL_MAX_ATTEMPT; i++) { + if (readl(hqvdp->regs + HQVDP_MBX_INFO_XP70) + & INFO_XP70_FW_READY) + break; + msleep(POLL_DELAY_MS); + } + if (i == POLL_MAX_ATTEMPT) { + DRM_ERROR("Could not boot\n"); + goto out; + } + + /* Launch Vsync */ + writel(SOFT_VSYNC_HW, hqvdp->regs + HQVDP_MBX_SOFT_VSYNC); + + DRM_INFO("HQVDP XP70 started\n"); +out: + release_firmware(firmware); +} + +int sti_hqvdp_bind(struct device *dev, struct device *master, void *data) +{ + struct sti_hqvdp *hqvdp = dev_get_drvdata(dev); + struct drm_device *drm_dev = data; + struct sti_layer *layer; + int err; + + DRM_DEBUG_DRIVER("\n"); + + hqvdp->drm_dev = drm_dev; + + /* Request for firmware */ + err = request_firmware_nowait(THIS_MODULE, FW_ACTION_HOTPLUG, + HQVDP_FMW_NAME, hqvdp->dev, + GFP_KERNEL, hqvdp, sti_hqvdp_start_xp70); + if (err) { + DRM_ERROR("Can't get HQVDP firmware\n"); + return err; + } + + layer = sti_layer_create(hqvdp->dev, STI_HQVDP_0, hqvdp->regs); + if (!layer) { + DRM_ERROR("Can't create HQVDP plane\n"); + return -ENOMEM; + } + + sti_drm_plane_init(drm_dev, layer, 1, DRM_PLANE_TYPE_OVERLAY); + + return 0; +} + +static void sti_hqvdp_unbind(struct device *dev, + struct device *master, void *data) +{ + /* do nothing */ +} + +static const struct component_ops sti_hqvdp_ops = { + .bind = sti_hqvdp_bind, + .unbind = sti_hqvdp_unbind, +}; + +static int sti_hqvdp_probe(struct platform_device *pdev) +{ + struct device *dev = &pdev->dev; + struct device_node *vtg_np; + struct sti_hqvdp *hqvdp; + struct resource *res; + + DRM_DEBUG_DRIVER("\n"); + + hqvdp = devm_kzalloc(dev, sizeof(*hqvdp), GFP_KERNEL); + if (!hqvdp) { + DRM_ERROR("Failed to allocate HQVDP context\n"); + return -ENOMEM; + } + + hqvdp->dev = dev; + + /* Get Memory resources */ + res = platform_get_resource(pdev, IORESOURCE_MEM, 0); + if (res == NULL) { + DRM_ERROR("Get memory resource failed\n"); + return -ENXIO; + } + hqvdp->regs = devm_ioremap(dev, res->start, resource_size(res)); + if (hqvdp->regs == NULL) { + DRM_ERROR("Register mapping failed\n"); + return -ENXIO; + } + + /* Get clock resources */ + hqvdp->clk = devm_clk_get(dev, "hqvdp"); + hqvdp->clk_pix_main = devm_clk_get(dev, "pix_main"); + if (IS_ERR(hqvdp->clk) || IS_ERR(hqvdp->clk)) { + DRM_ERROR("Cannot get clocks\n"); + return -ENXIO; + } + + /* Get reset resources */ + hqvdp->reset = devm_reset_control_get(dev, "hqvdp"); + if (!IS_ERR(hqvdp->reset)) + reset_control_deassert(hqvdp->reset); + + vtg_np = of_parse_phandle(pdev->dev.of_node, "st,vtg", 0); + if (vtg_np) + hqvdp->vtg = of_vtg_find(vtg_np); + + platform_set_drvdata(pdev, hqvdp); + + return component_add(&pdev->dev, &sti_hqvdp_ops); +} + +static int sti_hqvdp_remove(struct platform_device *pdev) +{ + component_del(&pdev->dev, &sti_hqvdp_ops); + return 0; +} + +static struct of_device_id hqvdp_of_match[] = { + { .compatible = "st,stih407-hqvdp", }, + { /* end node */ } +}; +MODULE_DEVICE_TABLE(of, hqvdp_of_match); + +struct platform_driver sti_hqvdp_driver = { + .driver = { + .name = "sti-hqvdp", + .owner = THIS_MODULE, + .of_match_table = hqvdp_of_match, + }, + .probe = sti_hqvdp_probe, + .remove = sti_hqvdp_remove, +}; + +module_platform_driver(sti_hqvdp_driver); + +MODULE_AUTHOR("Benjamin Gaignard "); +MODULE_DESCRIPTION("STMicroelectronics SoC DRM driver"); +MODULE_LICENSE("GPL"); diff --git a/drivers/gpu/drm/sti/sti_hqvdp.h b/drivers/gpu/drm/sti/sti_hqvdp.h new file mode 100644 index 0000000..cd5ecd0 --- /dev/null +++ b/drivers/gpu/drm/sti/sti_hqvdp.h @@ -0,0 +1,12 @@ +/* + * Copyright (C) STMicroelectronics SA 2014 + * Authors: Fabien Dessenne for STMicroelectronics. + * License terms: GNU General Public License (GPL), version 2 + */ + +#ifndef _STI_HQVDP_H_ +#define _STI_HQVDP_H_ + +struct sti_layer *sti_hqvdp_create(struct device *dev); + +#endif diff --git a/drivers/gpu/drm/sti/sti_hqvdp_lut.h b/drivers/gpu/drm/sti/sti_hqvdp_lut.h new file mode 100644 index 0000000..619af7f --- /dev/null +++ b/drivers/gpu/drm/sti/sti_hqvdp_lut.h @@ -0,0 +1,373 @@ +/* + * Copyright (C) STMicroelectronics SA 2014 + * Authors: Fabien Dessenne for STMicroelectronics. + * License terms: GNU General Public License (GPL), version 2 + */ + +#ifndef _STI_HQVDP_LUT_H_ +#define _STI_HQVDP_LUT_H_ + +#define NB_COEF 128 + +#define SHIFT_LUT_A_LEGACY 8 +#define SHIFT_LUT_B 8 +#define SHIFT_LUT_C_Y_LEGACY 8 +#define SHIFT_LUT_C_C_LEGACY 8 +#define SHIFT_LUT_D_Y_LEGACY 8 +#define SHIFT_LUT_D_C_LEGACY 8 +#define SHIFT_LUT_E_Y_LEGACY 8 +#define SHIFT_LUT_E_C_LEGACY 8 +#define SHIFT_LUT_F_Y_LEGACY 8 +#define SHIFT_LUT_F_C_LEGACY 8 + +static const u32 coef_lut_a_legacy[NB_COEF] = { + 0x0000ffff, 0x00010000, 0x000100ff, 0x00000000, + 0x00000000, 0x00050000, 0xfffc00ff, 0x00000000, + 0x00000000, 0x00090000, 0xfff900fe, 0x00000000, + 0x00000000, 0x0010ffff, 0xfff600fb, 0x00000000, + 0x00000000, 0x0017fffe, 0xfff400f7, 0x00000000, + 0x00000000, 0x001ffffd, 0xfff200f2, 0x00000000, + 0x00000000, 0x0027fffc, 0xfff100ec, 0x00000000, + 0x00000000, 0x0030fffb, 0xfff000e5, 0x00000000, + 0x00000000, 0x003afffa, 0xffee00de, 0x00000000, + 0x00000000, 0x0044fff9, 0xffed00d6, 0x00000000, + 0x00000000, 0x004efff8, 0xffed00cd, 0x00000000, + 0x00000000, 0x0059fff6, 0xffed00c4, 0x00000000, + 0x00000000, 0x0064fff5, 0xffed00ba, 0x00000000, + 0x00000000, 0x006ffff3, 0xffee00b0, 0x00000000, + 0x00000000, 0x007afff2, 0xffee00a6, 0x00000000, + 0x00000000, 0x0085fff1, 0xffef009b, 0x00000000, + 0x00000000, 0x0090fff0, 0xfff00090, 0x00000000, + 0x00000000, 0x009bffef, 0xfff10085, 0x00000000, + 0x00000000, 0x00a6ffee, 0xfff2007a, 0x00000000, + 0x00000000, 0x00b0ffee, 0xfff3006f, 0x00000000, + 0x00000000, 0x00baffed, 0xfff50064, 0x00000000, + 0x00000000, 0x00c4ffed, 0xfff60059, 0x00000000, + 0x00000000, 0x00cdffed, 0xfff8004e, 0x00000000, + 0x00000000, 0x00d6ffed, 0xfff90044, 0x00000000, + 0x00000000, 0x00deffee, 0xfffa003a, 0x00000000, + 0x00000000, 0x00e5fff0, 0xfffb0030, 0x00000000, + 0x00000000, 0x00ecfff1, 0xfffc0027, 0x00000000, + 0x00000000, 0x00f2fff2, 0xfffd001f, 0x00000000, + 0x00000000, 0x00f7fff4, 0xfffe0017, 0x00000000, + 0x00000000, 0x00fbfff6, 0xffff0010, 0x00000000, + 0x00000000, 0x00fefff9, 0x00000009, 0x00000000, + 0x00000000, 0x00fffffc, 0x00000005, 0x00000000 +}; + +static const u32 coef_lut_b[NB_COEF] = { + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000, + 0x00000000, 0x00000000, 0x00000100, 0x00000000 +}; + +static const u32 coef_lut_c_y_legacy[NB_COEF] = { + 0x00060004, 0x0038ffe1, 0x003800be, 0x0006ffe1, + 0x00050005, 0x0042ffe1, 0x003800b3, 0x0007ffe1, + 0x00040006, 0x0046ffe1, 0x003300b2, 0x0008ffe2, + 0x00030007, 0x004cffe1, 0x002e00b1, 0x0008ffe2, + 0x00020006, 0x0051ffe2, 0x002900b0, 0x0009ffe3, + 0x00010008, 0x0056ffe2, 0x002400ae, 0x0009ffe4, + 0xffff0008, 0x005cffe3, 0x001f00ad, 0x000affe4, + 0xfffe0008, 0x0062ffe4, 0x001a00ab, 0x000affe5, + 0xfffd000a, 0x0066ffe5, 0x001500a8, 0x000bffe6, + 0xfffc0009, 0x006bffe7, 0x001100a5, 0x000bffe8, + 0xfffa000a, 0x0070ffe8, 0x000d00a3, 0x000bffe9, + 0xfff9000b, 0x0076ffea, 0x0008009f, 0x000bffea, + 0xfff7000b, 0x007affec, 0x0005009b, 0x000cffec, + 0xfff6000b, 0x007effef, 0x00010098, 0x000cffed, + 0xfff4000b, 0x0084fff1, 0xfffd0095, 0x000cffee, + 0xfff3000b, 0x0088fff4, 0xfffa0090, 0x000cfff0, + 0xfff1000b, 0x008dfff7, 0xfff7008d, 0x000bfff1, + 0xfff0000c, 0x0090fffa, 0xfff40088, 0x000bfff3, + 0xffee000c, 0x0095fffd, 0xfff10084, 0x000bfff4, + 0xffed000c, 0x00980001, 0xffef007e, 0x000bfff6, + 0xffec000c, 0x009b0005, 0xffec007a, 0x000bfff7, + 0xffea000b, 0x009f0008, 0xffea0076, 0x000bfff9, + 0xffe9000b, 0x00a3000d, 0xffe80070, 0x000afffa, + 0xffe8000b, 0x00a50011, 0xffe7006b, 0x0009fffc, + 0xffe6000b, 0x00a80015, 0xffe50066, 0x000afffd, + 0xffe5000a, 0x00ab001a, 0xffe40062, 0x0008fffe, + 0xffe4000a, 0x00ad001f, 0xffe3005c, 0x0008ffff, + 0xffe40009, 0x00ae0024, 0xffe20056, 0x00080001, + 0xffe30009, 0x00b00029, 0xffe20051, 0x00060002, + 0xffe20008, 0x00b1002e, 0xffe1004c, 0x00070003, + 0xffe20008, 0x00b20033, 0xffe10046, 0x00060004, + 0xffe10007, 0x00b30038, 0xffe10042, 0x00050005 +}; + +static const u32 coef_lut_c_c_legacy[NB_COEF] = { + 0x0001fff3, 0x003afffb, 0x003a00a1, 0x0001fffb, + 0x0001fff5, 0x0041fffb, 0x0038009a, 0x0001fffb, + 0x0001fff5, 0x0046fffb, 0x00340099, 0x0001fffb, + 0x0001fff7, 0x0049fffb, 0x00300098, 0x0001fffb, + 0x0001fff9, 0x004cfffb, 0x002d0096, 0x0001fffb, + 0x0001fffa, 0x004ffffc, 0x00290095, 0x0001fffb, + 0x0001fff9, 0x0054fffd, 0x00250093, 0x0001fffc, + 0x0001fffa, 0x0058fffd, 0x00220092, 0x0000fffc, + 0x0001fffb, 0x005bfffe, 0x001f0090, 0x0000fffc, + 0x0001fffd, 0x005effff, 0x001c008c, 0x0000fffd, + 0x0001fffd, 0x00620000, 0x0019008a, 0x0000fffd, + 0x0001fffe, 0x00660001, 0x00160088, 0xfffffffd, + 0x0000fffe, 0x006a0003, 0x00130085, 0xfffffffe, + 0x0000fffe, 0x006e0004, 0x00100083, 0xfffffffe, + 0x0000fffe, 0x00710006, 0x000e007f, 0xffffffff, + 0x0000fffe, 0x00750008, 0x000c007c, 0xfffeffff, + 0xfffffffe, 0x0079000a, 0x000a0079, 0xfffeffff, + 0xfffffffe, 0x007c000c, 0x00080075, 0xfffe0000, + 0xffffffff, 0x007f000e, 0x00060071, 0xfffe0000, + 0xfffeffff, 0x00830010, 0x0004006e, 0xfffe0000, + 0xfffeffff, 0x00850013, 0x0003006a, 0xfffe0000, + 0xfffdffff, 0x00880016, 0x00010066, 0xfffe0001, + 0xfffd0000, 0x008a0019, 0x00000062, 0xfffd0001, + 0xfffd0000, 0x008c001c, 0xffff005e, 0xfffd0001, + 0xfffc0000, 0x0090001f, 0xfffe005b, 0xfffb0001, + 0xfffc0000, 0x00920022, 0xfffd0058, 0xfffa0001, + 0xfffc0001, 0x00930025, 0xfffd0054, 0xfff90001, + 0xfffb0001, 0x00950029, 0xfffc004f, 0xfffa0001, + 0xfffb0001, 0x0096002d, 0xfffb004c, 0xfff90001, + 0xfffb0001, 0x00980030, 0xfffb0049, 0xfff70001, + 0xfffb0001, 0x00990034, 0xfffb0046, 0xfff50001, + 0xfffb0001, 0x009a0038, 0xfffb0041, 0xfff50001 +}; + +static const u32 coef_lut_d_y_legacy[NB_COEF] = { + 0xfff80009, 0x0046ffec, 0x004600a3, 0xfff8ffec, + 0xfff70009, 0x004effed, 0x0044009d, 0xfff9ffeb, + 0xfff6000a, 0x0052ffee, 0x003f009d, 0xfffaffea, + 0xfff50009, 0x0057ffef, 0x003b009d, 0xfffbffe9, + 0xfff50008, 0x005bfff0, 0x0037009c, 0xfffcffe9, + 0xfff40008, 0x005ffff2, 0x0033009b, 0xfffcffe9, + 0xfff30007, 0x0064fff3, 0x002f009b, 0xfffdffe8, + 0xfff20007, 0x0068fff5, 0x002b0099, 0xfffeffe8, + 0xfff10008, 0x006bfff7, 0x00270097, 0xffffffe8, + 0xfff00007, 0x006ffff9, 0x00230097, 0xffffffe8, + 0xffef0006, 0x0073fffb, 0x00200095, 0x0000ffe8, + 0xffee0005, 0x0077fffe, 0x001c0093, 0x0000ffe9, + 0xffee0005, 0x007a0000, 0x00180091, 0x0001ffe9, + 0xffed0005, 0x007d0003, 0x0015008e, 0x0002ffe9, + 0xffec0005, 0x00800006, 0x0012008b, 0x0002ffea, + 0xffeb0004, 0x00840008, 0x000e008a, 0x0003ffea, + 0xffeb0003, 0x0087000b, 0x000b0087, 0x0003ffeb, + 0xffea0003, 0x008a000e, 0x00080084, 0x0004ffeb, + 0xffea0002, 0x008b0012, 0x00060080, 0x0005ffec, + 0xffe90002, 0x008e0015, 0x0003007d, 0x0005ffed, + 0xffe90001, 0x00910018, 0x0000007a, 0x0005ffee, + 0xffe90000, 0x0093001c, 0xfffe0077, 0x0005ffee, + 0xffe80000, 0x00950020, 0xfffb0073, 0x0006ffef, + 0xffe8ffff, 0x00970023, 0xfff9006f, 0x0007fff0, + 0xffe8ffff, 0x00970027, 0xfff7006b, 0x0008fff1, + 0xffe8fffe, 0x0099002b, 0xfff50068, 0x0007fff2, + 0xffe8fffd, 0x009b002f, 0xfff30064, 0x0007fff3, + 0xffe9fffc, 0x009b0033, 0xfff2005f, 0x0008fff4, + 0xffe9fffc, 0x009c0037, 0xfff0005b, 0x0008fff5, + 0xffe9fffb, 0x009d003b, 0xffef0057, 0x0009fff5, + 0xffeafffa, 0x009d003f, 0xffee0052, 0x000afff6, + 0xffebfff9, 0x009d0044, 0xffed004e, 0x0009fff7 +}; + +static const u32 coef_lut_d_c_legacy[NB_COEF] = { + 0xfffeffff, 0x003fffff, 0x003f0089, 0xfffeffff, + 0xfffe0000, 0x00460000, 0x0042007d, 0xfffffffe, + 0xfffe0000, 0x00490001, 0x003f007d, 0xfffffffd, + 0xfffd0001, 0x004b0002, 0x003c007d, 0x0000fffc, + 0xfffd0001, 0x004e0003, 0x0039007c, 0x0000fffc, + 0xfffc0001, 0x00510005, 0x0036007c, 0x0000fffb, + 0xfffc0001, 0x00540006, 0x0033007b, 0x0001fffa, + 0xfffc0003, 0x00550008, 0x00310078, 0x0001fffa, + 0xfffb0003, 0x00580009, 0x002e0078, 0x0001fffa, + 0xfffb0002, 0x005b000b, 0x002b0077, 0x0002fff9, + 0xfffa0003, 0x005e000d, 0x00280075, 0x0002fff9, + 0xfffa0002, 0x0060000f, 0x00260074, 0x0002fff9, + 0xfffa0004, 0x00610011, 0x00230072, 0x0002fff9, + 0xfffa0004, 0x00640013, 0x00200070, 0x0002fff9, + 0xfff90004, 0x00660015, 0x001e006e, 0x0003fff9, + 0xfff90004, 0x00680017, 0x001c006c, 0x0003fff9, + 0xfff90003, 0x006b0019, 0x0019006b, 0x0003fff9, + 0xfff90003, 0x006c001c, 0x00170068, 0x0004fff9, + 0xfff90003, 0x006e001e, 0x00150066, 0x0004fff9, + 0xfff90002, 0x00700020, 0x00130064, 0x0004fffa, + 0xfff90002, 0x00720023, 0x00110061, 0x0004fffa, + 0xfff90002, 0x00740026, 0x000f0060, 0x0002fffa, + 0xfff90002, 0x00750028, 0x000d005e, 0x0003fffa, + 0xfff90002, 0x0077002b, 0x000b005b, 0x0002fffb, + 0xfffa0001, 0x0078002e, 0x00090058, 0x0003fffb, + 0xfffa0001, 0x00780031, 0x00080055, 0x0003fffc, + 0xfffa0001, 0x007b0033, 0x00060054, 0x0001fffc, + 0xfffb0000, 0x007c0036, 0x00050051, 0x0001fffc, + 0xfffc0000, 0x007c0039, 0x0003004e, 0x0001fffd, + 0xfffc0000, 0x007d003c, 0x0002004b, 0x0001fffd, + 0xfffdffff, 0x007d003f, 0x00010049, 0x0000fffe, + 0xfffeffff, 0x007d0042, 0x00000046, 0x0000fffe +}; + +static const u32 coef_lut_e_y_legacy[NB_COEF] = { + 0xfff10001, 0x00490004, 0x00490083, 0xfff10004, + 0xfff10000, 0x00500006, 0x004b007b, 0xfff10002, + 0xfff10000, 0x00530007, 0x0048007b, 0xfff10001, + 0xfff10000, 0x00550009, 0x0046007a, 0xfff10000, + 0xfff1fffe, 0x0058000b, 0x0043007b, 0xfff2fffe, + 0xfff1ffff, 0x005a000d, 0x0040007a, 0xfff2fffd, + 0xfff1fffd, 0x005d000f, 0x003e007a, 0xfff2fffc, + 0xfff1fffd, 0x005f0011, 0x003b0079, 0xfff3fffb, + 0xfff1fffc, 0x00610013, 0x00390079, 0xfff3fffa, + 0xfff1fffb, 0x00640015, 0x00360079, 0xfff3fff9, + 0xfff1fffa, 0x00660017, 0x00340078, 0xfff4fff8, + 0xfff1fffb, 0x00680019, 0x00310077, 0xfff4fff7, + 0xfff2fff9, 0x006a001b, 0x002f0076, 0xfff5fff6, + 0xfff2fff9, 0x006c001e, 0x002c0075, 0xfff5fff5, + 0xfff2fff9, 0x006d0020, 0x002a0073, 0xfff6fff5, + 0xfff3fff7, 0x00700022, 0x00270073, 0xfff6fff4, + 0xfff3fff7, 0x00710025, 0x00250071, 0xfff7fff3, + 0xfff4fff6, 0x00730027, 0x00220070, 0xfff7fff3, + 0xfff5fff6, 0x0073002a, 0x0020006d, 0xfff9fff2, + 0xfff5fff5, 0x0075002c, 0x001e006c, 0xfff9fff2, + 0xfff6fff5, 0x0076002f, 0x001b006a, 0xfff9fff2, + 0xfff7fff4, 0x00770031, 0x00190068, 0xfffbfff1, + 0xfff8fff4, 0x00780034, 0x00170066, 0xfffafff1, + 0xfff9fff3, 0x00790036, 0x00150064, 0xfffbfff1, + 0xfffafff3, 0x00790039, 0x00130061, 0xfffcfff1, + 0xfffbfff3, 0x0079003b, 0x0011005f, 0xfffdfff1, + 0xfffcfff2, 0x007a003e, 0x000f005d, 0xfffdfff1, + 0xfffdfff2, 0x007a0040, 0x000d005a, 0xfffffff1, + 0xfffefff2, 0x007b0043, 0x000b0058, 0xfffefff1, + 0x0000fff1, 0x007a0046, 0x00090055, 0x0000fff1, + 0x0001fff1, 0x007b0048, 0x00070053, 0x0000fff1, + 0x0002fff1, 0x007b004b, 0x00060050, 0x0000fff1 +}; + +static const u32 coef_lut_e_c_legacy[NB_COEF] = { + 0xfffa0001, 0x003f0010, 0x003f006d, 0xfffa0010, + 0xfffb0002, 0x00440011, 0x00440062, 0xfffa000e, + 0xfffb0001, 0x00460013, 0x00420062, 0xfffa000d, + 0xfffb0000, 0x00480014, 0x00410062, 0xfffa000c, + 0xfffb0001, 0x00490015, 0x003f0061, 0xfffb000b, + 0xfffb0000, 0x004b0017, 0x003d0061, 0xfffb000a, + 0xfffb0000, 0x004d0018, 0x003b0062, 0xfffb0008, + 0xfffcffff, 0x004f001a, 0x00390061, 0xfffb0007, + 0xfffc0000, 0x004f001c, 0x00380060, 0xfffb0006, + 0xfffcffff, 0x0052001d, 0x00360060, 0xfffb0005, + 0xfffdfffe, 0x0053001f, 0x00340060, 0xfffb0004, + 0xfffdfffe, 0x00540021, 0x0032005e, 0xfffc0004, + 0xfffeffff, 0x00550022, 0x0030005d, 0xfffc0003, + 0xfffeffff, 0x00560024, 0x002f005c, 0xfffc0002, + 0xfffffffd, 0x00580026, 0x002d005c, 0xfffc0001, + 0xfffffffd, 0x005a0027, 0x002b005c, 0xfffc0000, + 0x0000fffd, 0x005a0029, 0x0029005a, 0xfffd0000, + 0x0000fffc, 0x005c002b, 0x0027005a, 0xfffdffff, + 0x0001fffc, 0x005c002d, 0x00260058, 0xfffdffff, + 0x0002fffc, 0x005c002f, 0x00240056, 0xfffffffe, + 0x0003fffc, 0x005d0030, 0x00220055, 0xfffffffe, + 0x0004fffc, 0x005e0032, 0x00210054, 0xfffefffd, + 0x0004fffb, 0x00600034, 0x001f0053, 0xfffefffd, + 0x0005fffb, 0x00600036, 0x001d0052, 0xfffffffc, + 0x0006fffb, 0x00600038, 0x001c004f, 0x0000fffc, + 0x0007fffb, 0x00610039, 0x001a004f, 0xfffffffc, + 0x0008fffb, 0x0062003b, 0x0018004d, 0x0000fffb, + 0x000afffb, 0x0061003d, 0x0017004b, 0x0000fffb, + 0x000bfffb, 0x0061003f, 0x00150049, 0x0001fffb, + 0x000cfffa, 0x00620041, 0x00140048, 0x0000fffb, + 0x000dfffa, 0x00620042, 0x00130046, 0x0001fffb, + 0x000efffa, 0x00620044, 0x00110044, 0x0002fffb +}; + +static const u32 coef_lut_f_y_legacy[NB_COEF] = { + 0xfff6fff0, 0x00490012, 0x0049006e, 0xfff60012, + 0xfff7fff1, 0x004e0013, 0x00490068, 0xfff60010, + 0xfff7fff2, 0x004f0015, 0x00470067, 0xfff6000f, + 0xfff7fff5, 0x004f0017, 0x00450065, 0xfff6000e, + 0xfff8fff5, 0x00500018, 0x00440065, 0xfff6000c, + 0xfff8fff6, 0x0051001a, 0x00420064, 0xfff6000b, + 0xfff8fff6, 0x0052001c, 0x00400064, 0xfff6000a, + 0xfff9fff6, 0x0054001d, 0x003e0064, 0xfff60008, + 0xfff9fff8, 0x0054001f, 0x003c0063, 0xfff60007, + 0xfffafff8, 0x00550021, 0x003a0062, 0xfff60006, + 0xfffbfff7, 0x00560022, 0x00390062, 0xfff60005, + 0xfffbfff8, 0x00570024, 0x00370061, 0xfff60004, + 0xfffcfff8, 0x00580026, 0x00350060, 0xfff60003, + 0xfffdfff8, 0x00590028, 0x0033005f, 0xfff60002, + 0xfffdfff7, 0x005b002a, 0x0031005f, 0xfff60001, + 0xfffefff7, 0x005c002c, 0x002f005e, 0xfff60000, + 0xfffffff6, 0x005e002d, 0x002d005e, 0xfff6ffff, + 0x0000fff6, 0x005e002f, 0x002c005c, 0xfff7fffe, + 0x0001fff6, 0x005f0031, 0x002a005b, 0xfff7fffd, + 0x0002fff6, 0x005f0033, 0x00280059, 0xfff8fffd, + 0x0003fff6, 0x00600035, 0x00260058, 0xfff8fffc, + 0x0004fff6, 0x00610037, 0x00240057, 0xfff8fffb, + 0x0005fff6, 0x00620039, 0x00220056, 0xfff7fffb, + 0x0006fff6, 0x0062003a, 0x00210055, 0xfff8fffa, + 0x0007fff6, 0x0063003c, 0x001f0054, 0xfff8fff9, + 0x0008fff6, 0x0064003e, 0x001d0054, 0xfff6fff9, + 0x000afff6, 0x00640040, 0x001c0052, 0xfff6fff8, + 0x000bfff6, 0x00640042, 0x001a0051, 0xfff6fff8, + 0x000cfff6, 0x00650044, 0x00180050, 0xfff5fff8, + 0x000efff6, 0x00650045, 0x0017004f, 0xfff5fff7, + 0x000ffff6, 0x00670047, 0x0015004f, 0xfff2fff7, + 0x0010fff6, 0x00680049, 0x0013004e, 0xfff1fff7 +}; + +static const u32 coef_lut_f_c_legacy[NB_COEF] = { + 0x0000fffb, 0x003a001a, 0x003a005d, 0x0000001a, + 0x0001fffb, 0x003f001b, 0x00400051, 0x00000019, + 0x0001fffc, 0x0040001c, 0x003f0051, 0x00000017, + 0x0002fffb, 0x0042001d, 0x003e0051, 0xffff0016, + 0x0002fffb, 0x0043001e, 0x003d0051, 0xffff0015, + 0x0003fffc, 0x00430020, 0x003b0050, 0xffff0014, + 0x0003fffb, 0x00450021, 0x003a0051, 0xfffe0013, + 0x0004fffc, 0x00450022, 0x00390050, 0xfffe0012, + 0x0005fffc, 0x00460023, 0x0038004f, 0xfffe0011, + 0x0005fffb, 0x00480025, 0x00360050, 0xfffd0010, + 0x0006fffc, 0x00480026, 0x0035004f, 0xfffd000f, + 0x0006fffc, 0x00490027, 0x0034004f, 0xfffd000e, + 0x0007fffd, 0x00490028, 0x0033004e, 0xfffd000d, + 0x0008fffc, 0x004a002a, 0x0031004d, 0xfffd000d, + 0x0009fffd, 0x004a002b, 0x0030004d, 0xfffc000c, + 0x0009fffc, 0x004c002c, 0x002f004d, 0xfffc000b, + 0x000afffc, 0x004c002e, 0x002e004c, 0xfffc000a, + 0x000bfffc, 0x004d002f, 0x002c004c, 0xfffc0009, + 0x000cfffc, 0x004d0030, 0x002b004a, 0xfffd0009, + 0x000dfffd, 0x004d0031, 0x002a004a, 0xfffc0008, + 0x000dfffd, 0x004e0033, 0x00280049, 0xfffd0007, + 0x000efffd, 0x004f0034, 0x00270049, 0xfffc0006, + 0x000ffffd, 0x004f0035, 0x00260048, 0xfffc0006, + 0x0010fffd, 0x00500036, 0x00250048, 0xfffb0005, + 0x0011fffe, 0x004f0038, 0x00230046, 0xfffc0005, + 0x0012fffe, 0x00500039, 0x00220045, 0xfffc0004, + 0x0013fffe, 0x0051003a, 0x00210045, 0xfffb0003, + 0x0014ffff, 0x0050003b, 0x00200043, 0xfffc0003, + 0x0015ffff, 0x0051003d, 0x001e0043, 0xfffb0002, + 0x0016ffff, 0x0051003e, 0x001d0042, 0xfffb0002, + 0x00170000, 0x0051003f, 0x001c0040, 0xfffc0001, + 0x00190000, 0x00510040, 0x001b003f, 0xfffb0001 +}; + +#endif diff --git a/drivers/gpu/drm/sti/sti_layer.c b/drivers/gpu/drm/sti/sti_layer.c index 5051b4c..480ec1c 100644 --- a/drivers/gpu/drm/sti/sti_layer.c +++ b/drivers/gpu/drm/sti/sti_layer.c @@ -13,6 +13,7 @@ #include "sti_compositor.h" #include "sti_cursor.h" #include "sti_gdp.h" +#include "sti_hqvdp.h" #include "sti_layer.h" #include "sti_vid.h" @@ -33,6 +34,8 @@ const char *sti_layer_to_str(struct sti_layer *layer) return "VID1"; case STI_CURSOR: return "CURSOR"; + case STI_HQVDP_0: + return "HQVDP0"; default: return ""; } @@ -54,6 +57,9 @@ struct sti_layer *sti_layer_create(struct device *dev, int desc, case STI_CUR: layer = sti_cursor_create(dev); break; + case STI_VDP: + layer = sti_hqvdp_create(dev); + break; } if (!layer) { @@ -72,7 +78,9 @@ struct sti_layer *sti_layer_create(struct device *dev, int desc, return layer; } -int sti_layer_prepare(struct sti_layer *layer, struct drm_framebuffer *fb, +int sti_layer_prepare(struct sti_layer *layer, + struct drm_crtc *crtc, + struct drm_framebuffer *fb, struct drm_display_mode *mode, int mixer_id, int dest_x, int dest_y, int dest_w, int dest_h, int src_x, int src_y, int src_w, int src_h) @@ -92,6 +100,7 @@ int sti_layer_prepare(struct sti_layer *layer, struct drm_framebuffer *fb, return 1; } + layer->crtc = crtc; layer->fb = fb; layer->mode = mode; layer->mixer_id = mixer_id; diff --git a/drivers/gpu/drm/sti/sti_layer.h b/drivers/gpu/drm/sti/sti_layer.h index 68bfdca..ceff497 100644 --- a/drivers/gpu/drm/sti/sti_layer.h +++ b/drivers/gpu/drm/sti/sti_layer.h @@ -22,7 +22,8 @@ enum sti_layer_type { STI_GDP = 1 << STI_LAYER_TYPE_SHIFT, STI_VID = 2 << STI_LAYER_TYPE_SHIFT, STI_CUR = 3 << STI_LAYER_TYPE_SHIFT, - STI_BCK = 4 << STI_LAYER_TYPE_SHIFT + STI_BCK = 4 << STI_LAYER_TYPE_SHIFT, + STI_VDP = 5 << STI_LAYER_TYPE_SHIFT }; enum sti_layer_id_of_type { @@ -39,6 +40,7 @@ enum sti_layer_desc { STI_GDP_3 = STI_GDP | STI_ID_3, STI_VID_0 = STI_VID | STI_ID_0, STI_VID_1 = STI_VID | STI_ID_1, + STI_HQVDP_0 = STI_VDP | STI_ID_0, STI_CURSOR = STI_CUR, STI_BACK = STI_BCK }; @@ -67,6 +69,7 @@ struct sti_layer_funcs { * * @plane: drm plane it is bound to (if any) * @fb: drm fb it is bound to + * @crtc: crtc it is bound to * @mode: display mode * @desc: layer type & id * @device: driver device @@ -88,6 +91,7 @@ struct sti_layer_funcs { struct sti_layer { struct drm_plane plane; struct drm_framebuffer *fb; + struct drm_crtc *crtc; struct drm_display_mode *mode; enum sti_layer_desc desc; struct device *dev; @@ -109,7 +113,9 @@ struct sti_layer { struct sti_layer *sti_layer_create(struct device *dev, int desc, void __iomem *baseaddr); -int sti_layer_prepare(struct sti_layer *layer, struct drm_framebuffer *fb, +int sti_layer_prepare(struct sti_layer *layer, + struct drm_crtc *crtc, + struct drm_framebuffer *fb, struct drm_display_mode *mode, int mixer_id, int dest_x, int dest_y, diff --git a/drivers/gpu/drm/sti/sti_mixer.c b/drivers/gpu/drm/sti/sti_mixer.c index 9a4ce74..13a4b84 100644 --- a/drivers/gpu/drm/sti/sti_mixer.c +++ b/drivers/gpu/drm/sti/sti_mixer.c @@ -123,6 +123,7 @@ int sti_mixer_set_layer_depth(struct sti_mixer *mixer, struct sti_layer *layer) layer_id = GAM_DEPTH_GDP3_ID; break; case STI_VID_0: + case STI_HQVDP_0: layer_id = GAM_DEPTH_VID0_ID; break; case STI_VID_1: @@ -189,6 +190,7 @@ static u32 sti_mixer_get_layer_mask(struct sti_layer *layer) case STI_GDP_3: return GAM_CTL_GDP3_MASK; case STI_VID_0: + case STI_HQVDP_0: return GAM_CTL_VID0_MASK; case STI_VID_1: return GAM_CTL_VID1_MASK; -- cgit v0.10.2 From f78e772a2c7aa4e96398de9e3fa0be750c705d3a Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Thu, 11 Dec 2014 13:35:29 +0100 Subject: drm: sti: correctly cleanup CRTC and planes When bind failed make sure that CRTC and planes are completely clean up to avoid properties duplication. Signed-off-by: Benjamin Gaignard diff --git a/drivers/gpu/drm/sti/sti_drm_drv.c b/drivers/gpu/drm/sti/sti_drm_drv.c index 223d93c..9ce7574 100644 --- a/drivers/gpu/drm/sti/sti_drm_drv.c +++ b/drivers/gpu/drm/sti/sti_drm_drv.c @@ -67,8 +67,12 @@ static int sti_drm_load(struct drm_device *dev, unsigned long flags) sti_drm_mode_config_init(dev); ret = component_bind_all(dev->dev, dev); - if (ret) + if (ret) { + drm_kms_helper_poll_fini(dev); + drm_mode_config_cleanup(dev); + kfree(private); return ret; + } drm_helper_disable_unused_functions(dev); -- cgit v0.10.2 From 9f49c37635d5c2a801f7670d5fbf0b25ec461f2c Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Wed, 10 Dec 2014 12:16:05 -0800 Subject: drm/i915: save/restore GMBUS freq across suspend/resume on gen4 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Should probably just init this in the GMbus code all the time, based on the cdclk and HPLL like we do on newer platforms. Ville has code for that in a rework branch, but until then we can fix this bug fairly easily. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76301 Signed-off-by: Jesse Barnes Reviewed-by: Ville Syrjälä Tested-by: Nikolay Cc: stable@vger.kernel.org Signed-off-by: Jani Nikula diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index be910d2..63bcda5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -924,6 +924,7 @@ struct i915_suspend_saved_registers { u32 savePIPEB_LINK_N1; u32 saveMCHBAR_RENDER_STANDBY; u32 savePCH_PORT_HOTPLUG; + u16 saveGCDGMBUS; }; struct vlv_s0ix_state { diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index b607bbe..eefdc23 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -85,6 +85,7 @@ #define I915_GC_RENDER_CLOCK_166_MHZ (0 << 0) #define I915_GC_RENDER_CLOCK_200_MHZ (1 << 0) #define I915_GC_RENDER_CLOCK_333_MHZ (4 << 0) +#define GCDGMBUS 0xcc #define PCI_LBPC 0xf4 /* legacy/combination backlight modes, also called LBB */ diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c index dfe6617..2636882 100644 --- a/drivers/gpu/drm/i915/i915_suspend.c +++ b/drivers/gpu/drm/i915/i915_suspend.c @@ -303,6 +303,10 @@ int i915_save_state(struct drm_device *dev) } } + if (IS_GEN4(dev)) + pci_read_config_word(dev->pdev, GCDGMBUS, + &dev_priv->regfile.saveGCDGMBUS); + /* Cache mode state */ if (INTEL_INFO(dev)->gen < 7) dev_priv->regfile.saveCACHE_MODE_0 = I915_READ(CACHE_MODE_0); @@ -331,6 +335,10 @@ int i915_restore_state(struct drm_device *dev) mutex_lock(&dev->struct_mutex); i915_gem_restore_fences(dev); + + if (IS_GEN4(dev)) + pci_write_config_word(dev->pdev, GCDGMBUS, + dev_priv->regfile.saveGCDGMBUS); i915_restore_display(dev); if (!drm_core_check_feature(dev, DRIVER_MODESET)) { -- cgit v0.10.2 From 4e0cd68115620bc3236ff4e58e4c073948629b41 Mon Sep 17 00:00:00 2001 From: "benjamin.gaignard@linaro.org" Date: Sat, 13 Dec 2014 07:59:31 +0100 Subject: drm: sti: fix module compilation issue When compiling in module some symbol aren't missing, export them correctly. Signed-off-by: Benjamin Gaignard Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/sti/sti_drm_plane.c b/drivers/gpu/drm/sti/sti_drm_plane.c index c9dd0e5..bb6a293 100644 --- a/drivers/gpu/drm/sti/sti_drm_plane.c +++ b/drivers/gpu/drm/sti/sti_drm_plane.c @@ -194,3 +194,4 @@ struct drm_plane *sti_drm_plane_init(struct drm_device *dev, return &layer->plane; } +EXPORT_SYMBOL(sti_drm_plane_init); diff --git a/drivers/gpu/drm/sti/sti_hqvdp.c b/drivers/gpu/drm/sti/sti_hqvdp.c index 200d020..f3db05d 100644 --- a/drivers/gpu/drm/sti/sti_hqvdp.c +++ b/drivers/gpu/drm/sti/sti_hqvdp.c @@ -816,6 +816,7 @@ struct sti_layer *sti_hqvdp_create(struct device *dev) return &hqvdp->layer; } +EXPORT_SYMBOL(sti_hqvdp_create); static void sti_hqvdp_init_plugs(struct sti_hqvdp *hqvdp) { diff --git a/drivers/gpu/drm/sti/sti_layer.c b/drivers/gpu/drm/sti/sti_layer.c index 480ec1c..899104f 100644 --- a/drivers/gpu/drm/sti/sti_layer.c +++ b/drivers/gpu/drm/sti/sti_layer.c @@ -40,6 +40,7 @@ const char *sti_layer_to_str(struct sti_layer *layer) return ""; } } +EXPORT_SYMBOL(sti_layer_to_str); struct sti_layer *sti_layer_create(struct device *dev, int desc, void __iomem *baseaddr) @@ -77,6 +78,7 @@ struct sti_layer *sti_layer_create(struct device *dev, int desc, return layer; } +EXPORT_SYMBOL(sti_layer_create); int sti_layer_prepare(struct sti_layer *layer, struct drm_crtc *crtc, -- cgit v0.10.2
Description/Restrictions
DRMDRM Generic “EDID” BLOB | IMMUTABLE
DRMGenericGeneric “EDID” BLOB | IMMUTABLE 0Contains topology path to a connector.
“TILE”BLOB | IMMUTABLE0ConnectorContains tiling information for a connector.
Plane “type” ENUM | IMMUTABLE