Merge branch 'drm-next-4.6' of git://people.freedesktop.org/~agd5f/linux into drm-next

A few more fixes and cleanups for 4.6:
- DCE code cleanups
- HDP flush/invalidation fixes
- GPUVM fixes
- switch to drm_vblank_[on|off]
- PX fixes
- misc bug fixes

* 'drm-next-4.6' of git://people.freedesktop.org/~agd5f/linux: (50 commits)
  drm/amdgpu: split pipeline sync out of SDMA vm_flush() as well
  drm/amdgpu: Revert "add mutex for ba_va->valids/invalids"
  drm/amdgpu: Revert "add lock for interval tree in vm"
  drm/amdgpu: Revert "add spin lock to protect freed list in vm (v3)"
  drm/amdgpu: reserve the PD during unmap and remove
  drm/amdgpu: Fix two bugs in amdgpu_vm_bo_split_mapping
  drm/radeon: Don't drop DP 2.7 Ghz link setup on some cards.
  MAINTAINERS: update radeon entry to include amdgpu as well
  drm/amdgpu: disable runtime pm on PX laptops without dGPU power control
  drm/radeon: disable runtime pm on PX laptops without dGPU power control
  drm/amd/amdgpu: Fix indentation in do_set_base() (DCEv8)
  drm/amd/amdgpu: make afmt_init cleanup if alloc fails (DCEv8)
  drm/amd/amdgpu: Move config init flag to bottom of sw_init (DCEv8)
  drm/amd/amdgpu: Don't proceed into audio_fini if audio is disabled (DCEv8)
  drm/amd/amdgpu: Fix identation in do_set_base() (DCEv10)
  drm/amd/amdgpu: Make afmt_init cleanup if alloc fails (DCEv10)
  drm/amd/amdgpu: Move initialized flag to bottom of sw_init (DCEv10)
  drm/amd/amdgpu: Don't proceed in audio_fini if disabled (DCEv10)
  drm/amd/amdgpu: Fix indentation in dce_v11_0_crtc_do_set_base()
  drm/amd/amdgpu: Make afmt_init() cleanup if alloc fails (DCEv11)
  ...

35 files changed, 738 insertions, 311 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index d048972..a80c8ce 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -287,9 +287,11 @@ struct amdgpu_ring_funcs {
 			struct amdgpu_ib *ib);
 	void (*emit_fence)(struct amdgpu_ring *ring, uint64_t addr,
 			   uint64_t seq, unsigned flags);
+	void (*emit_pipeline_sync)(struct amdgpu_ring *ring);
 	void (*emit_vm_flush)(struct amdgpu_ring *ring, unsigned vm_id,
 			      uint64_t pd_addr);
 	void (*emit_hdp_flush)(struct amdgpu_ring *ring);
+	void (*emit_hdp_invalidate)(struct amdgpu_ring *ring);
 	void (*emit_gds_switch)(struct amdgpu_ring *ring, uint32_t vmid,
 				uint32_t gds_base, uint32_t gds_size,
 				uint32_t gws_base, uint32_t gws_size,
@@ -369,9 +371,6 @@ struct amdgpu_fence {
 	struct amdgpu_ring		*ring;
 	uint64_t			seq;
 
-	/* filp or special value for fence creator */
-	void				*owner;
-
 	wait_queue_t			fence_wake;
 };
 
@@ -392,8 +391,7 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
 				   unsigned irq_type);
 void amdgpu_fence_driver_suspend(struct amdgpu_device *adev);
 void amdgpu_fence_driver_resume(struct amdgpu_device *adev);
-int amdgpu_fence_emit(struct amdgpu_ring *ring, void *owner,
-		      struct amdgpu_fence **fence);
+int amdgpu_fence_emit(struct amdgpu_ring *ring, struct fence **fence);
 void amdgpu_fence_process(struct amdgpu_ring *ring);
 int amdgpu_fence_wait_next(struct amdgpu_ring *ring);
 int amdgpu_fence_wait_empty(struct amdgpu_ring *ring);
@@ -434,6 +432,8 @@ struct amdgpu_bo_list_entry {
 	struct ttm_validate_buffer	tv;
 	struct amdgpu_bo_va		*bo_va;
 	uint32_t			priority;
+	struct page			**user_pages;
+	int				user_invalidated;
 };
 
 struct amdgpu_bo_va_mapping {
@@ -445,7 +445,6 @@ struct amdgpu_bo_va_mapping {
 
 /* bo virtual addresses in a specific vm */
 struct amdgpu_bo_va {
-	struct mutex		        mutex;
 	/* protected by bo being reserved */
 	struct list_head		bo_list;
 	struct fence		        *last_pt_update;
@@ -596,6 +595,8 @@ int amdgpu_sync_resv(struct amdgpu_device *adev,
 struct fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync);
 int amdgpu_sync_wait(struct amdgpu_sync *sync);
 void amdgpu_sync_free(struct amdgpu_sync *sync);
+int amdgpu_sync_init(void);
+void amdgpu_sync_fini(void);
 
 /*
  * GART structures, functions & helpers
@@ -726,7 +727,7 @@ struct amdgpu_ib {
 	uint32_t			length_dw;
 	uint64_t			gpu_addr;
 	uint32_t			*ptr;
-	struct amdgpu_fence		*fence;
+	struct fence			*fence;
 	struct amdgpu_user_fence        *user;
 	struct amdgpu_vm		*vm;
 	unsigned			vm_id;
@@ -845,7 +846,6 @@ struct amdgpu_vm_id {
 
 struct amdgpu_vm {
 	/* tree of virtual addresses mapped */
-	spinlock_t		it_lock;
 	struct rb_root		va;
 
 	/* protecting invalidated */
@@ -882,6 +882,13 @@ struct amdgpu_vm_manager_id {
 	struct list_head	list;
 	struct fence		*active;
 	atomic_long_t		owner;
+
+	uint32_t		gds_base;
+	uint32_t		gds_size;
+	uint32_t		gws_base;
+	uint32_t		gws_size;
+	uint32_t		oa_base;
+	uint32_t		oa_size;
 };
 
 struct amdgpu_vm_manager {
@@ -917,8 +924,11 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 		      struct amdgpu_sync *sync, struct fence *fence,
 		      unsigned *vm_id, uint64_t *vm_pd_addr);
 void amdgpu_vm_flush(struct amdgpu_ring *ring,
-		     unsigned vmid,
-		     uint64_t pd_addr);
+		     unsigned vm_id, uint64_t pd_addr,
+		     uint32_t gds_base, uint32_t gds_size,
+		     uint32_t gws_base, uint32_t gws_size,
+		     uint32_t oa_base, uint32_t oa_size);
+void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vm_id);
 uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr);
 int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
 				    struct amdgpu_vm *vm);
@@ -1006,7 +1016,7 @@ struct amdgpu_bo_list {
 	struct amdgpu_bo *gds_obj;
 	struct amdgpu_bo *gws_obj;
 	struct amdgpu_bo *oa_obj;
-	bool has_userptr;
+	unsigned first_userptr;
 	unsigned num_entries;
 	struct amdgpu_bo_list_entry *array;
 };
@@ -1135,8 +1145,7 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		  unsigned size, struct amdgpu_ib *ib);
 void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib);
 int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
-		       struct amdgpu_ib *ib, void *owner,
-		       struct fence *last_vm_update,
+		       struct amdgpu_ib *ib, struct fence *last_vm_update,
 		       struct fence **f);
 int amdgpu_ib_pool_init(struct amdgpu_device *adev);
 void amdgpu_ib_pool_fini(struct amdgpu_device *adev);
@@ -2012,7 +2021,6 @@ struct amdgpu_device {
 	struct amdgpu_sdma		sdma;
 
 	/* uvd */
-	bool				has_uvd;
 	struct amdgpu_uvd		uvd;
 
 	/* vce */
@@ -2186,10 +2194,12 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r))
 #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r))
 #define amdgpu_ring_emit_ib(r, ib) (r)->funcs->emit_ib((r), (ib))
+#define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r))
 #define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr))
 #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags))
 #define amdgpu_ring_emit_gds_switch(r, v, db, ds, wb, ws, ab, as) (r)->funcs->emit_gds_switch((r), (v), (db), (ds), (wb), (ws), (ab), (as))
 #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r))
+#define amdgpu_ring_emit_hdp_invalidate(r) (r)->funcs->emit_hdp_invalidate((r))
 #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib)))
 #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev))
 #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv))
@@ -2314,12 +2324,15 @@ int amdgpu_cs_get_ring(struct amdgpu_device *adev, u32 ip_type,
 		       struct amdgpu_ring **out_ring);
 void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *rbo, u32 domain);
 bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo);
+int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages);
 int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
 				     uint32_t flags);
 bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm);
 struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm);
 bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
 				  unsigned long end);
+bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
+				       int *last_invalidated);
 bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
 uint32_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
 				 struct ttm_mem_reg *mem);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
index fa948dc..0020a0e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
@@ -63,6 +63,10 @@ bool amdgpu_has_atpx(void) {
 	return amdgpu_atpx_priv.atpx_detected;
 }
 
+bool amdgpu_has_atpx_dgpu_power_cntl(void) {
+	return amdgpu_atpx_priv.atpx.functions.power_cntl;
+}
+
 /**
  * amdgpu_atpx_call - call an ATPX method
  *
@@ -142,10 +146,6 @@ static void amdgpu_atpx_parse_functions(struct amdgpu_atpx_functions *f, u32 mas
  */
 static int amdgpu_atpx_validate(struct amdgpu_atpx *atpx)
 {
-	/* make sure required functions are enabled */
-	/* dGPU power control is required */
-	atpx->functions.power_cntl = true;
-
 	if (atpx->functions.px_params) {
 		union acpi_object *info;
 		struct atpx_px_params output;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
index 4792f9d..eacd810 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
@@ -91,7 +91,7 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev,
 	struct amdgpu_bo *gws_obj = adev->gds.gws_gfx_bo;
 	struct amdgpu_bo *oa_obj = adev->gds.oa_gfx_bo;
 
-	bool has_userptr = false;
+	unsigned last_entry = 0, first_userptr = num_entries;
 	unsigned i;
 	int r;
 
@@ -101,8 +101,9 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev,
 	memset(array, 0, num_entries * sizeof(struct amdgpu_bo_list_entry));
 
 	for (i = 0; i < num_entries; ++i) {
-		struct amdgpu_bo_list_entry *entry = &array[i];
+		struct amdgpu_bo_list_entry *entry;
 		struct drm_gem_object *gobj;
+		struct amdgpu_bo *bo;
 		struct mm_struct *usermm;
 
 		gobj = drm_gem_object_lookup(adev->ddev, filp, info[i].bo_handle);
@@ -111,19 +112,24 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev,
 			goto error_free;
 		}
 
-		entry->robj = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
+		bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
 		drm_gem_object_unreference_unlocked(gobj);
-		entry->priority = min(info[i].bo_priority,
-				      AMDGPU_BO_LIST_MAX_PRIORITY);
-		usermm = amdgpu_ttm_tt_get_usermm(entry->robj->tbo.ttm);
+
+		usermm = amdgpu_ttm_tt_get_usermm(bo->tbo.ttm);
 		if (usermm) {
 			if (usermm != current->mm) {
-				amdgpu_bo_unref(&entry->robj);
+				amdgpu_bo_unref(&bo);
 				r = -EPERM;
 				goto error_free;
 			}
-			has_userptr = true;
+			entry = &array[--first_userptr];
+		} else {
+			entry = &array[last_entry++];
 		}
+
+		entry->robj = bo;
+		entry->priority = min(info[i].bo_priority,
+				      AMDGPU_BO_LIST_MAX_PRIORITY);
 		entry->tv.bo = &entry->robj->tbo;
 		entry->tv.shared = true;
 
@@ -145,7 +151,7 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev,
 	list->gds_obj = gds_obj;
 	list->gws_obj = gws_obj;
 	list->oa_obj = oa_obj;
-	list->has_userptr = has_userptr;
+	list->first_userptr = first_userptr;
 	list->array = array;
 	list->num_entries = num_entries;
 
@@ -194,6 +200,7 @@ void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list,
 
 		list_add_tail(&list->array[i].tv.head,
 			      &bucket[priority]);
+		list->array[i].user_pages = NULL;
 	}
 
 	/* Connect the sorted buckets in the output list. */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 52c3eb9..4f5ef41 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -111,6 +111,7 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p,
 	p->uf_entry.priority = 0;
 	p->uf_entry.tv.bo = &p->uf_entry.robj->tbo;
 	p->uf_entry.tv.shared = true;
+	p->uf_entry.user_pages = NULL;
 
 	drm_gem_object_unreference_unlocked(gobj);
 	return 0;
@@ -297,6 +298,7 @@ int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p,
 
 	list_for_each_entry(lobj, validated, tv.head) {
 		struct amdgpu_bo *bo = lobj->robj;
+		bool binding_userptr = false;
 		struct mm_struct *usermm;
 		uint32_t domain;
 
@@ -304,6 +306,15 @@ int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p,
 		if (usermm && usermm != current->mm)
 			return -EPERM;
 
+		/* Check if we have user pages and nobody bound the BO already */
+		if (lobj->user_pages && bo->tbo.ttm->state != tt_bound) {
+			size_t size = sizeof(struct page *);
+
+			size *= bo->tbo.ttm->num_pages;
+			memcpy(bo->tbo.ttm->pages, lobj->user_pages, size);
+			binding_userptr = true;
+		}
+
 		if (bo->pin_count)
 			continue;
 
@@ -334,6 +345,11 @@ int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p,
 			}
 			return r;
 		}
+
+		if (binding_userptr) {
+			drm_free_large(lobj->user_pages);
+			lobj->user_pages = NULL;
+		}
 	}
 	return 0;
 }
@@ -342,15 +358,18 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 				union drm_amdgpu_cs *cs)
 {
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
+	struct amdgpu_bo_list_entry *e;
 	struct list_head duplicates;
 	bool need_mmap_lock = false;
+	unsigned i, tries = 10;
 	int r;
 
 	INIT_LIST_HEAD(&p->validated);
 
 	p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle);
 	if (p->bo_list) {
-		need_mmap_lock = p->bo_list->has_userptr;
+		need_mmap_lock = p->bo_list->first_userptr !=
+			p->bo_list->num_entries;
 		amdgpu_bo_list_get_list(p->bo_list, &p->validated);
 	}
 
@@ -363,9 +382,81 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	if (need_mmap_lock)
 		down_read(&current->mm->mmap_sem);
 
-	r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true, &duplicates);
-	if (unlikely(r != 0))
-		goto error_reserve;
+	while (1) {
+		struct list_head need_pages;
+		unsigned i;
+
+		r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true,
+					   &duplicates);
+		if (unlikely(r != 0))
+			goto error_free_pages;
+
+		/* Without a BO list we don't have userptr BOs */
+		if (!p->bo_list)
+			break;
+
+		INIT_LIST_HEAD(&need_pages);
+		for (i = p->bo_list->first_userptr;
+		     i < p->bo_list->num_entries; ++i) {
+
+			e = &p->bo_list->array[i];
+
+			if (amdgpu_ttm_tt_userptr_invalidated(e->robj->tbo.ttm,
+				 &e->user_invalidated) && e->user_pages) {
+
+				/* We acquired a page array, but somebody
+				 * invalidated it. Free it an try again
+				 */
+				release_pages(e->user_pages,
+					      e->robj->tbo.ttm->num_pages,
+					      false);
+				drm_free_large(e->user_pages);
+				e->user_pages = NULL;
+			}
+
+			if (e->robj->tbo.ttm->state != tt_bound &&
+			    !e->user_pages) {
+				list_del(&e->tv.head);
+				list_add(&e->tv.head, &need_pages);
+
+				amdgpu_bo_unreserve(e->robj);
+			}
+		}
+
+		if (list_empty(&need_pages))
+			break;
+
+		/* Unreserve everything again. */
+		ttm_eu_backoff_reservation(&p->ticket, &p->validated);
+
+		/* We tried to often, just abort */
+		if (!--tries) {
+			r = -EDEADLK;
+			goto error_free_pages;
+		}
+
+		/* Fill the page arrays for all useptrs. */
+		list_for_each_entry(e, &need_pages, tv.head) {
+			struct ttm_tt *ttm = e->robj->tbo.ttm;
+
+			e->user_pages = drm_calloc_large(ttm->num_pages,
+							 sizeof(struct page*));
+			if (!e->user_pages) {
+				r = -ENOMEM;
+				goto error_free_pages;
+			}
+
+			r = amdgpu_ttm_tt_get_user_pages(ttm, e->user_pages);
+			if (r) {
+				drm_free_large(e->user_pages);
+				e->user_pages = NULL;
+				goto error_free_pages;
+			}
+		}
+
+		/* And try again. */
+		list_splice(&need_pages, &p->validated);
+	}
 
 	amdgpu_vm_get_pt_bos(&fpriv->vm, &duplicates);
 
@@ -397,10 +488,26 @@ error_validate:
 		ttm_eu_backoff_reservation(&p->ticket, &p->validated);
 	}
 
-error_reserve:
+error_free_pages:
+
 	if (need_mmap_lock)
 		up_read(&current->mm->mmap_sem);
 
+	if (p->bo_list) {
+		for (i = p->bo_list->first_userptr;
+		     i < p->bo_list->num_entries; ++i) {
+			e = &p->bo_list->array[i];
+
+			if (!e->user_pages)
+				continue;
+
+			release_pages(e->user_pages,
+				      e->robj->tbo.ttm->num_pages,
+				      false);
+			drm_free_large(e->user_pages);
+		}
+	}
+
 	return r;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 2139da7..61211747 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -62,6 +62,12 @@ static const char *amdgpu_asic_name[] = {
 	"LAST",
 };
 
+#if defined(CONFIG_VGA_SWITCHEROO)
+bool amdgpu_has_atpx_dgpu_power_cntl(void);
+#else
+static inline bool amdgpu_has_atpx_dgpu_power_cntl(void) { return false; }
+#endif
+
 bool amdgpu_device_is_px(struct drm_device *dev)
 {
 	struct amdgpu_device *adev = dev->dev_private;
@@ -1479,7 +1485,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 
 	if (amdgpu_runtime_pm == 1)
 		runtime = true;
-	if (amdgpu_device_is_px(ddev))
+	if (amdgpu_device_is_px(ddev) && amdgpu_has_atpx_dgpu_power_cntl())
 		runtime = true;
 	vga_switcheroo_register_client(adev->pdev, &amdgpu_switcheroo_ops, runtime);
 	if (runtime)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
index 74a2f8a..f1e17d6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
@@ -555,6 +555,7 @@ static struct pci_driver amdgpu_kms_pci_driver = {
 
 static int __init amdgpu_init(void)
 {
+	amdgpu_sync_init();
 #ifdef CONFIG_VGA_CONSOLE
 	if (vgacon_text_force()) {
 		DRM_ERROR("VGACON disables amdgpu kernel modesetting.\n");
@@ -577,6 +578,7 @@ static void __exit amdgpu_exit(void)
 	amdgpu_amdkfd_fini();
 	drm_pci_exit(driver, pdriver);
 	amdgpu_unregister_atpx_handler();
+	amdgpu_sync_fini();
 }
 
 module_init(amdgpu_init);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
index 97db196..83599f2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
@@ -91,32 +91,29 @@ static u32 amdgpu_fence_read(struct amdgpu_ring *ring)
  * amdgpu_fence_emit - emit a fence on the requested ring
  *
  * @ring: ring the fence is associated with
- * @owner: creator of the fence
- * @fence: amdgpu fence object
+ * @f: resulting fence object
  *
  * Emits a fence command on the requested ring (all asics).
  * Returns 0 on success, -ENOMEM on failure.
  */
-int amdgpu_fence_emit(struct amdgpu_ring *ring, void *owner,
-		      struct amdgpu_fence **fence)
+int amdgpu_fence_emit(struct amdgpu_ring *ring, struct fence **f)
 {
 	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_fence *fence;
 
-	/* we are protected by the ring emission mutex */
-	*fence = kmem_cache_alloc(amdgpu_fence_slab, GFP_KERNEL);
-	if ((*fence) == NULL) {
+	fence = kmem_cache_alloc(amdgpu_fence_slab, GFP_KERNEL);
+	if (fence == NULL)
 		return -ENOMEM;
-	}
-	(*fence)->seq = ++ring->fence_drv.sync_seq;
-	(*fence)->ring = ring;
-	(*fence)->owner = owner;
-	fence_init(&(*fence)->base, &amdgpu_fence_ops,
-		&ring->fence_drv.fence_queue.lock,
-		adev->fence_context + ring->idx,
-		(*fence)->seq);
+
+	fence->seq = ++ring->fence_drv.sync_seq;
+	fence->ring = ring;
+	fence_init(&fence->base, &amdgpu_fence_ops,
+		   &ring->fence_drv.fence_queue.lock,
+		   adev->fence_context + ring->idx,
+		   fence->seq);
 	amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr,
-			       (*fence)->seq,
-			       AMDGPU_FENCE_FLAG_INT);
+			       fence->seq, AMDGPU_FENCE_FLAG_INT);
+	*f = &fence->base;
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index 7a47c45..1ecdf6c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -140,25 +140,40 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj, struct drm_file *file_pri
 void amdgpu_gem_object_close(struct drm_gem_object *obj,
 			     struct drm_file *file_priv)
 {
-	struct amdgpu_bo *rbo = gem_to_amdgpu_bo(obj);
-	struct amdgpu_device *adev = rbo->adev;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+	struct amdgpu_device *adev = bo->adev;
 	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
+
+	struct amdgpu_bo_list_entry vm_pd;
+	struct list_head list, duplicates;
+	struct ttm_validate_buffer tv;
+	struct ww_acquire_ctx ticket;
 	struct amdgpu_bo_va *bo_va;
 	int r;
-	r = amdgpu_bo_reserve(rbo, true);
+
+	INIT_LIST_HEAD(&list);
+	INIT_LIST_HEAD(&duplicates);
+
+	tv.bo = &bo->tbo;
+	tv.shared = true;
+	list_add(&tv.head, &list);
+
+	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
+
+	r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates);
 	if (r) {
 		dev_err(adev->dev, "leaking bo va because "
 			"we fail to reserve bo (%d)\n", r);
 		return;
 	}
-	bo_va = amdgpu_vm_bo_find(vm, rbo);
+	bo_va = amdgpu_vm_bo_find(vm, bo);
 	if (bo_va) {
 		if (--bo_va->ref_count == 0) {
 			amdgpu_vm_bo_rmv(adev, bo_va);
 		}
 	}
-	amdgpu_bo_unreserve(rbo);
+	ttm_eu_backoff_reservation(&ticket, &list);
 }
 
 static int amdgpu_gem_handle_lockup(struct amdgpu_device *adev, int r)
@@ -274,18 +289,23 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 
 	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE) {
 		down_read(&current->mm->mmap_sem);
+
+		r = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm,
+						 bo->tbo.ttm->pages);
+		if (r)
+			goto unlock_mmap_sem;
+
 		r = amdgpu_bo_reserve(bo, true);
-		if (r) {
-			up_read(&current->mm->mmap_sem);
-			goto release_object;
-		}
+		if (r)
+			goto free_pages;
 
 		amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
 		r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false);
 		amdgpu_bo_unreserve(bo);
-		up_read(&current->mm->mmap_sem);
 		if (r)
-			goto release_object;
+			goto free_pages;
+
+		up_read(&current->mm->mmap_sem);
 	}
 
 	r = drm_gem_handle_create(filp, gobj, &handle);
@@ -297,6 +317,12 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 	args->handle = handle;
 	return 0;
 
+free_pages:
+	release_pages(bo->tbo.ttm->pages, bo->tbo.ttm->num_pages, false);
+
+unlock_mmap_sem:
+	up_read(&current->mm->mmap_sem);
+
 release_object:
 	drm_gem_object_unreference_unlocked(gobj);
 
@@ -569,11 +595,10 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	tv.shared = true;
 	list_add(&tv.head, &list);
 
-	if (args->operation == AMDGPU_VA_OP_MAP) {
-		tv_pd.bo = &fpriv->vm.page_directory->tbo;
-		tv_pd.shared = true;
-		list_add(&tv_pd.head, &list);
-	}
+	tv_pd.bo = &fpriv->vm.page_directory->tbo;
+	tv_pd.shared = true;
+	list_add(&tv_pd.head, &list);
+
 	r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates);
 	if (r) {
 		drm_gem_object_unreference_unlocked(gobj);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index db14a7b..798d466 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -90,9 +90,8 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  */
 void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib)
 {
-	amdgpu_sa_bo_free(adev, &ib->sa_bo, &ib->fence->base);
-	if (ib->fence)
-		fence_put(&ib->fence->base);
+	amdgpu_sa_bo_free(adev, &ib->sa_bo, ib->fence);
+	fence_put(ib->fence);
 }
 
 /**
@@ -101,7 +100,6 @@ void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib)
  * @adev: amdgpu_device pointer
  * @num_ibs: number of IBs to schedule
  * @ibs: IB objects to schedule
- * @owner: owner for creating the fences
  * @f: fence created during this submission
  *
  * Schedule an IB on the associated ring (all asics).
@@ -118,8 +116,7 @@ void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib)
  * to SI there was just a DE IB.
  */
 int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
-		       struct amdgpu_ib *ibs, void *owner,
-		       struct fence *last_vm_update,
+		       struct amdgpu_ib *ibs, struct fence *last_vm_update,
 		       struct fence **f)
 {
 	struct amdgpu_device *adev = ring->adev;
@@ -153,13 +150,10 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 
 	if (vm) {
 		/* do context switch */
-		amdgpu_vm_flush(ring, ib->vm_id, ib->vm_pd_addr);
-
-		if (ring->funcs->emit_gds_switch)
-			amdgpu_ring_emit_gds_switch(ring, ib->vm_id,
-						    ib->gds_base, ib->gds_size,
-						    ib->gws_base, ib->gws_size,
-						    ib->oa_base, ib->oa_size);
+		amdgpu_vm_flush(ring, ib->vm_id, ib->vm_pd_addr,
+				ib->gds_base, ib->gds_size,
+				ib->gws_base, ib->gws_size,
+				ib->oa_base, ib->oa_size);
 
 		if (ring->funcs->emit_hdp_flush)
 			amdgpu_ring_emit_hdp_flush(ring);
@@ -171,6 +165,8 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 
 		if (ib->ctx != ctx || ib->vm != vm) {
 			ring->current_ctx = old_ctx;
+			if (ib->vm_id)
+				amdgpu_vm_reset_id(adev, ib->vm_id);
 			amdgpu_ring_undo(ring);
 			return -EINVAL;
 		}
@@ -178,10 +174,17 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		ring->current_ctx = ctx;
 	}
 
-	r = amdgpu_fence_emit(ring, owner, &ib->fence);
+	if (vm) {
+		if (ring->funcs->emit_hdp_invalidate)
+			amdgpu_ring_emit_hdp_invalidate(ring);
+	}
+
+	r = amdgpu_fence_emit(ring, &ib->fence);
 	if (r) {
 		dev_err(adev->dev, "failed to emit fence (%d)\n", r);
 		ring->current_ctx = old_ctx;
+		if (ib->vm_id)
+			amdgpu_vm_reset_id(adev, ib->vm_id);
 		amdgpu_ring_undo(ring);
 		return r;
 	}
@@ -195,7 +198,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	}
 
 	if (f)
-		*f = fence_get(&ib->fence->base);
+		*f = fence_get(ib->fence);
 
 	amdgpu_ring_commit(ring);
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index 90e52f7..692b455 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -148,7 +148,7 @@ static struct fence *amdgpu_job_run(struct amd_sched_job *sched_job)
 	}
 
 	trace_amdgpu_sched_run_job(job);
-	r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job->owner,
+	r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs,
 			       job->sync.last_vm_update, &fence);
 	if (r) {
 		DRM_ERROR("Error scheduling IBs (%d)\n", r);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
index c15be00..c48b4fc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
@@ -37,6 +37,8 @@ struct amdgpu_sync_entry {
 	struct fence		*fence;
 };
 
+static struct kmem_cache *amdgpu_sync_slab;
+
 /**
  * amdgpu_sync_create - zero init sync object
  *
@@ -50,14 +52,18 @@ void amdgpu_sync_create(struct amdgpu_sync *sync)
 	sync->last_vm_update = NULL;
 }
 
+/**
+ * amdgpu_sync_same_dev - test if fence belong to us
+ *
+ * @adev: amdgpu device to use for the test
+ * @f: fence to test
+ *
+ * Test if the fence was issued by us.
+ */
 static bool amdgpu_sync_same_dev(struct amdgpu_device *adev, struct fence *f)
 {
-	struct amdgpu_fence *a_fence = to_amdgpu_fence(f);
 	struct amd_sched_fence *s_fence = to_amd_sched_fence(f);
 
-	if (a_fence)
-		return a_fence->ring->adev == adev;
-
 	if (s_fence) {
 		struct amdgpu_ring *ring;
 
@@ -68,17 +74,31 @@ static bool amdgpu_sync_same_dev(struct amdgpu_device *adev, struct fence *f)
 	return false;
 }
 
-static bool amdgpu_sync_test_owner(struct fence *f, void *owner)
+/**
+ * amdgpu_sync_get_owner - extract the owner of a fence
+ *
+ * @fence: fence get the owner from
+ *
+ * Extract who originally created the fence.
+ */
+static void *amdgpu_sync_get_owner(struct fence *f)
 {
-	struct amdgpu_fence *a_fence = to_amdgpu_fence(f);
 	struct amd_sched_fence *s_fence = to_amd_sched_fence(f);
+
 	if (s_fence)
-		return s_fence->owner == owner;
-	if (a_fence)
-		return a_fence->owner == owner;
-	return false;
+		return s_fence->owner;
+
+	return AMDGPU_FENCE_OWNER_UNDEFINED;
 }
 
+/**
+ * amdgpu_sync_keep_later - Keep the later fence
+ *
+ * @keep: existing fence to test
+ * @fence: new fence
+ *
+ * Either keep the existing fence or the new one, depending which one is later.
+ */
 static void amdgpu_sync_keep_later(struct fence **keep, struct fence *fence)
 {
 	if (*keep && fence_is_later(*keep, fence))
@@ -104,7 +124,7 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 		return 0;
 
 	if (amdgpu_sync_same_dev(adev, f) &&
-	    amdgpu_sync_test_owner(f, AMDGPU_FENCE_OWNER_VM))
+	    amdgpu_sync_get_owner(f) == AMDGPU_FENCE_OWNER_VM)
 		amdgpu_sync_keep_later(&sync->last_vm_update, f);
 
 	hash_for_each_possible(sync->fences, e, node, f->context) {
@@ -115,7 +135,7 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 		return 0;
 	}
 
-	e = kmalloc(sizeof(struct amdgpu_sync_entry), GFP_KERNEL);
+	e = kmem_cache_alloc(amdgpu_sync_slab, GFP_KERNEL);
 	if (!e)
 		return -ENOMEM;
 
@@ -124,18 +144,6 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 	return 0;
 }
 
-static void *amdgpu_sync_get_owner(struct fence *f)
-{
-	struct amdgpu_fence *a_fence = to_amdgpu_fence(f);
-	struct amd_sched_fence *s_fence = to_amd_sched_fence(f);
-
-	if (s_fence)
-		return s_fence->owner;
-	else if (a_fence)
-		return a_fence->owner;
-	return AMDGPU_FENCE_OWNER_UNDEFINED;
-}
-
 /**
  * amdgpu_sync_resv - sync to a reservation object
  *
@@ -208,7 +216,7 @@ struct fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync)
 		f = e->fence;
 
 		hash_del(&e->node);
-		kfree(e);
+		kmem_cache_free(amdgpu_sync_slab, e);
 
 		if (!fence_is_signaled(f))
 			return f;
@@ -231,7 +239,7 @@ int amdgpu_sync_wait(struct amdgpu_sync *sync)
 
 		hash_del(&e->node);
 		fence_put(e->fence);
-		kfree(e);
+		kmem_cache_free(amdgpu_sync_slab, e);
 	}
 
 	return 0;
@@ -253,8 +261,34 @@ void amdgpu_sync_free(struct amdgpu_sync *sync)
 	hash_for_each_safe(sync->fences, i, tmp, e, node) {
 		hash_del(&e->node);
 		fence_put(e->fence);
-		kfree(e);
+		kmem_cache_free(amdgpu_sync_slab, e);
 	}
 
 	fence_put(sync->last_vm_update);
 }
+
+/**
+ * amdgpu_sync_init - init sync object subsystem
+ *
+ * Allocate the slab allocator.
+ */
+int amdgpu_sync_init(void)
+{
+	amdgpu_sync_slab = kmem_cache_create(
+		"amdgpu_sync", sizeof(struct amdgpu_sync_entry), 0,
+		SLAB_HWCACHE_ALIGN, NULL);
+	if (!amdgpu_sync_slab)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/**
+ * amdgpu_sync_fini - fini sync object subsystem
+ *
+ * Free the slab allocator.
+ */
+void amdgpu_sync_fini(void)
+{
+	kmem_cache_destroy(amdgpu_sync_slab);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 9ccdd18..0f42b1a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -494,29 +494,32 @@ static void amdgpu_ttm_io_mem_free(struct ttm_bo_device *bdev, struct ttm_mem_re
 /*
  * TTM backend functions.
  */
+struct amdgpu_ttm_gup_task_list {
+	struct list_head	list;
+	struct task_struct	*task;
+};
+
 struct amdgpu_ttm_tt {
-	struct ttm_dma_tt		ttm;
-	struct amdgpu_device		*adev;
-	u64				offset;
-	uint64_t			userptr;
-	struct mm_struct		*usermm;
-	uint32_t			userflags;
+	struct ttm_dma_tt	ttm;
+	struct amdgpu_device	*adev;
+	u64			offset;
+	uint64_t		userptr;
+	struct mm_struct	*usermm;
+	uint32_t		userflags;
+	spinlock_t              guptasklock;
+	struct list_head        guptasks;
+	atomic_t		mmu_invalidations;
 };
 
-/* prepare the sg table with the user pages */
-static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm)
+int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 {
-	struct amdgpu_device *adev = amdgpu_get_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	unsigned pinned = 0, nents;
-	int r;
-
 	int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY);
-	enum dma_data_direction direction = write ?
-		DMA_BIDIRECTIONAL : DMA_TO_DEVICE;
+	unsigned pinned = 0;
+	int r;
 
 	if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) {
-		/* check that we only pin down anonymous memory
+		/* check that we only use anonymous memory
 		   to prevent problems with writeback */
 		unsigned long end = gtt->userptr + ttm->num_pages * PAGE_SIZE;
 		struct vm_area_struct *vma;
@@ -529,10 +532,21 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm)
 	do {
 		unsigned num_pages = ttm->num_pages - pinned;
 		uint64_t userptr = gtt->userptr + pinned * PAGE_SIZE;
-		struct page **pages = ttm->pages + pinned;
+		struct page **p = pages + pinned;
+		struct amdgpu_ttm_gup_task_list guptask;
+
+		guptask.task = current;
+		spin_lock(&gtt->guptasklock);
+		list_add(&guptask.list, &gtt->guptasks);
+		spin_unlock(&gtt->guptasklock);
 
 		r = get_user_pages(current, current->mm, userptr, num_pages,
-				   write, 0, pages, NULL);
+				   write, 0, p, NULL);
+
+		spin_lock(&gtt->guptasklock);
+		list_del(&guptask.list);
+		spin_unlock(&gtt->guptasklock);
+
 		if (r < 0)
 			goto release_pages;
 
@@ -540,6 +554,25 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm)
 
 	} while (pinned < ttm->num_pages);
 
+	return 0;
+
+release_pages:
+	release_pages(pages, pinned, 0);
+	return r;
+}
+
+/* prepare the sg table with the user pages */
+static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm)
+{
+	struct amdgpu_device *adev = amdgpu_get_adev(ttm->bdev);
+	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+	unsigned nents;
+	int r;
+
+	int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY);
+	enum dma_data_direction direction = write ?
+		DMA_BIDIRECTIONAL : DMA_TO_DEVICE;
+
 	r = sg_alloc_table_from_pages(ttm->sg, ttm->pages, ttm->num_pages, 0,
 				      ttm->num_pages << PAGE_SHIFT,
 				      GFP_KERNEL);
@@ -558,9 +591,6 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm)
 
 release_sg:
 	kfree(ttm->sg);
-
-release_pages:
-	release_pages(ttm->pages, pinned, 0);
 	return r;
 }
 
@@ -783,6 +813,10 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
 	gtt->userptr = addr;
 	gtt->usermm = current->mm;
 	gtt->userflags = flags;
+	spin_lock_init(&gtt->guptasklock);
+	INIT_LIST_HEAD(&gtt->guptasks);
+	atomic_set(&gtt->mmu_invalidations, 0);
+
 	return 0;
 }
 
@@ -800,21 +834,40 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
 				  unsigned long end)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+	struct amdgpu_ttm_gup_task_list *entry;
 	unsigned long size;
 
-	if (gtt == NULL)
-		return false;
-
-	if (gtt->ttm.ttm.state != tt_bound || !gtt->userptr)
+	if (gtt == NULL || !gtt->userptr)
 		return false;
 
 	size = (unsigned long)gtt->ttm.ttm.num_pages * PAGE_SIZE;
 	if (gtt->userptr > end || gtt->userptr + size <= start)
 		return false;
 
+	spin_lock(&gtt->guptasklock);
+	list_for_each_entry(entry, &gtt->guptasks, list) {
+		if (entry->task == current) {
+			spin_unlock(&gtt->guptasklock);
+			return false;
+		}
+	}
+	spin_unlock(&gtt->guptasklock);
+
+	atomic_inc(&gtt->mmu_invalidations);
+
 	return true;
 }
 
+bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
+				       int *last_invalidated)
+{
+	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+	int prev_invalidated = *last_invalidated;
+
+	*last_invalidated = atomic_read(&gtt->mmu_invalidations);
+	return prev_invalidated != *last_invalidated;
+}
+
 bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
index 1de82bf..e5f0a5e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
@@ -886,8 +886,7 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
 	ib->length_dw = 16;
 
 	if (direct) {
-		r = amdgpu_ib_schedule(ring, 1, ib,
-				       AMDGPU_FENCE_OWNER_UNDEFINED, NULL, &f);
+		r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f);
 		if (r)
 			goto err_free;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
index 39c3aa6..6d191fb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
@@ -425,8 +425,7 @@ int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 	for (i = ib->length_dw; i < ib_size_dw; ++i)
 		ib->ptr[i] = 0x0;
 
-	r = amdgpu_ib_schedule(ring, 1, ib, AMDGPU_FENCE_OWNER_UNDEFINED,
-			       NULL, &f);
+	r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f);
 	if (r)
 		goto err;
 
@@ -487,9 +486,7 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 		ib->ptr[i] = 0x0;
 
 	if (direct) {
-		r = amdgpu_ib_schedule(ring, 1, ib,
-				       AMDGPU_FENCE_OWNER_UNDEFINED,
-				       NULL, &f);
+		r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f);
 		if (r)
 			goto err;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index d9dc8be..b6c011b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -95,6 +95,7 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 	entry->priority = 0;
 	entry->tv.bo = &vm->page_directory->tbo;
 	entry->tv.shared = true;
+	entry->user_pages = NULL;
 	list_add(&entry->tv.head, validated);
 }
 
@@ -188,6 +189,13 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 		if (!is_later && owner == (long)id &&
 		    pd_addr == id->pd_gpu_addr) {
 
+			r = amdgpu_sync_fence(ring->adev, sync,
+					      id->mgr_id->active);
+			if (r) {
+				mutex_unlock(&adev->vm_manager.lock);
+				return r;
+			}
+
 			fence_put(id->mgr_id->active);
 			id->mgr_id->active = fence_get(fence);
 
@@ -234,19 +242,68 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
  * amdgpu_vm_flush - hardware flush the vm
  *
  * @ring: ring to use for flush
- * @vmid: vmid number to use
+ * @vm_id: vmid number to use
  * @pd_addr: address of the page directory
  *
  * Emit a VM flush when it is necessary.
  */
 void amdgpu_vm_flush(struct amdgpu_ring *ring,
-		     unsigned vmid,
-		     uint64_t pd_addr)
+		     unsigned vm_id, uint64_t pd_addr,
+		     uint32_t gds_base, uint32_t gds_size,
+		     uint32_t gws_base, uint32_t gws_size,
+		     uint32_t oa_base, uint32_t oa_size)
 {
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vm_manager_id *mgr_id = &adev->vm_manager.ids[vm_id];
+	bool gds_switch_needed = ring->funcs->emit_gds_switch && (
+		mgr_id->gds_base != gds_base ||
+		mgr_id->gds_size != gds_size ||
+		mgr_id->gws_base != gws_base ||
+		mgr_id->gws_size != gws_size ||
+		mgr_id->oa_base != oa_base ||
+		mgr_id->oa_size != oa_size);
+
+	if (ring->funcs->emit_pipeline_sync && (
+	    pd_addr != AMDGPU_VM_NO_FLUSH || gds_switch_needed))
+		amdgpu_ring_emit_pipeline_sync(ring);
+
 	if (pd_addr != AMDGPU_VM_NO_FLUSH) {
-		trace_amdgpu_vm_flush(pd_addr, ring->idx, vmid);
-		amdgpu_ring_emit_vm_flush(ring, vmid, pd_addr);
+		trace_amdgpu_vm_flush(pd_addr, ring->idx, vm_id);
+		amdgpu_ring_emit_vm_flush(ring, vm_id, pd_addr);
 	}
+
+	if (gds_switch_needed) {
+		mgr_id->gds_base = gds_base;
+		mgr_id->gds_size = gds_size;
+		mgr_id->gws_base = gws_base;
+		mgr_id->gws_size = gws_size;
+		mgr_id->oa_base = oa_base;
+		mgr_id->oa_size = oa_size;
+		amdgpu_ring_emit_gds_switch(ring, vm_id,
+					    gds_base, gds_size,
+					    gws_base, gws_size,
+					    oa_base, oa_size);
+	}
+}
+
+/**
+ * amdgpu_vm_reset_id - reset VMID to zero
+ *
+ * @adev: amdgpu device structure
+ * @vm_id: vmid number to use
+ *
+ * Reset saved GDW, GWS and OA to force switch on next flush.
+ */
+void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vm_id)
+{
+	struct amdgpu_vm_manager_id *mgr_id = &adev->vm_manager.ids[vm_id];
+
+	mgr_id->gds_base = 0;
+	mgr_id->gds_size = 0;
+	mgr_id->gws_base = 0;
+	mgr_id->gws_size = 0;
+	mgr_id->oa_base = 0;
+	mgr_id->oa_size = 0;
 }
 
 /**
@@ -810,7 +867,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
 	while (start != mapping->it.last + 1) {
 		uint64_t last;
 
-		last = min((uint64_t)mapping->it.last, start + max_size);
+		last = min((uint64_t)mapping->it.last, start + max_size - 1);
 		r = amdgpu_vm_bo_update_mapping(adev, gtt, gtt_flags, vm,
 						start, last, flags, addr,
 						fence);
@@ -818,7 +875,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
 			return r;
 
 		start = last + 1;
-		addr += max_size;
+		addr += max_size * AMDGPU_GPU_PAGE_SIZE;
 	}
 
 	return 0;
@@ -914,22 +971,18 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 	struct amdgpu_bo_va_mapping *mapping;
 	int r;
 
-	spin_lock(&vm->freed_lock);
 	while (!list_empty(&vm->freed)) {
 		mapping = list_first_entry(&vm->freed,
 			struct amdgpu_bo_va_mapping, list);
 		list_del(&mapping->list);
-		spin_unlock(&vm->freed_lock);
+
 		r = amdgpu_vm_bo_split_mapping(adev, NULL, 0, vm, mapping,
 					       0, NULL);
 		kfree(mapping);
 		if (r)
 			return r;
 
-		spin_lock(&vm->freed_lock);
 	}
-	spin_unlock(&vm->freed_lock);
-
 	return 0;
 
 }
@@ -956,9 +1009,8 @@ int amdgpu_vm_clear_invalids(struct amdgpu_device *adev,
 		bo_va = list_first_entry(&vm->invalidated,
 			struct amdgpu_bo_va, vm_status);
 		spin_unlock(&vm->status_lock);
-		mutex_lock(&bo_va->mutex);
+
 		r = amdgpu_vm_bo_update(adev, bo_va, NULL);
-		mutex_unlock(&bo_va->mutex);
 		if (r)
 			return r;
 
@@ -1002,7 +1054,7 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
 	INIT_LIST_HEAD(&bo_va->valids);
 	INIT_LIST_HEAD(&bo_va->invalids);
 	INIT_LIST_HEAD(&bo_va->vm_status);
-	mutex_init(&bo_va->mutex);
+
 	list_add_tail(&bo_va->bo_list, &bo->va);
 
 	return bo_va;
@@ -1054,9 +1106,7 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
 	eaddr /= AMDGPU_GPU_PAGE_SIZE;
 
-	spin_lock(&vm->it_lock);
 	it = interval_tree_iter_first(&vm->va, saddr, eaddr);
-	spin_unlock(&vm->it_lock);
 	if (it) {
 		struct amdgpu_bo_va_mapping *tmp;
 		tmp = container_of(it, struct amdgpu_bo_va_mapping, it);
@@ -1080,13 +1130,8 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	mapping->offset = offset;
 	mapping->flags = flags;
 
-	mutex_lock(&bo_va->mutex);
 	list_add(&mapping->list, &bo_va->invalids);
-	mutex_unlock(&bo_va->mutex);
-	spin_lock(&vm->it_lock);
 	interval_tree_insert(&mapping->it, &vm->va);
-	spin_unlock(&vm->it_lock);
-	trace_amdgpu_vm_bo_map(bo_va, mapping);
 
 	/* Make sure the page tables are allocated */
 	saddr >>= amdgpu_vm_block_size;
@@ -1130,6 +1175,7 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		entry->priority = 0;
 		entry->tv.bo = &entry->robj->tbo;
 		entry->tv.shared = true;
+		entry->user_pages = NULL;
 		vm->page_tables[pt_idx].addr = 0;
 	}
 
@@ -1137,9 +1183,7 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 
 error_free:
 	list_del(&mapping->list);
-	spin_lock(&vm->it_lock);
 	interval_tree_remove(&mapping->it, &vm->va);
-	spin_unlock(&vm->it_lock);
 	trace_amdgpu_vm_bo_unmap(bo_va, mapping);
 	kfree(mapping);
 
@@ -1168,7 +1212,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 	bool valid = true;
 
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
-	mutex_lock(&bo_va->mutex);
+
 	list_for_each_entry(mapping, &bo_va->valids, list) {
 		if (mapping->it.start == saddr)
 			break;
@@ -1182,25 +1226,18 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 				break;
 		}
 
-		if (&mapping->list == &bo_va->invalids) {
-			mutex_unlock(&bo_va->mutex);
+		if (&mapping->list == &bo_va->invalids)
 			return -ENOENT;
-		}
 	}
-	mutex_unlock(&bo_va->mutex);
+
 	list_del(&mapping->list);
-	spin_lock(&vm->it_lock);
 	interval_tree_remove(&mapping->it, &vm->va);
-	spin_unlock(&vm->it_lock);
 	trace_amdgpu_vm_bo_unmap(bo_va, mapping);
 
-	if (valid) {
-		spin_lock(&vm->freed_lock);
+	if (valid)
 		list_add(&mapping->list, &vm->freed);
-		spin_unlock(&vm->freed_lock);
-	} else {
+	else
 		kfree(mapping);
-	}
 
 	return 0;
 }
@@ -1229,23 +1266,17 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
 
 	list_for_each_entry_safe(mapping, next, &bo_va->valids, list) {
 		list_del(&mapping->list);
-		spin_lock(&vm->it_lock);
 		interval_tree_remove(&mapping->it, &vm->va);
-		spin_unlock(&vm->it_lock);
 		trace_amdgpu_vm_bo_unmap(bo_va, mapping);
-		spin_lock(&vm->freed_lock);
 		list_add(&mapping->list, &vm->freed);
-		spin_unlock(&vm->freed_lock);
 	}
 	list_for_each_entry_safe(mapping, next, &bo_va->invalids, list) {
 		list_del(&mapping->list);
-		spin_lock(&vm->it_lock);
 		interval_tree_remove(&mapping->it, &vm->va);
-		spin_unlock(&vm->it_lock);
 		kfree(mapping);
 	}
+
 	fence_put(bo_va->last_pt_update);
-	mutex_destroy(&bo_va->mutex);
 	kfree(bo_va);
 }
 
@@ -1298,8 +1329,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	INIT_LIST_HEAD(&vm->invalidated);
 	INIT_LIST_HEAD(&vm->cleared);
 	INIT_LIST_HEAD(&vm->freed);
-	spin_lock_init(&vm->it_lock);
-	spin_lock_init(&vm->freed_lock);
+
 	pd_size = amdgpu_vm_directory_size(adev);
 	pd_entries = amdgpu_vm_num_pdes(adev);
 
@@ -1386,6 +1416,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 
 	amdgpu_bo_unref(&vm->page_directory);
 	fence_put(vm->page_directory_fence);
+
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_vm_id *id = &vm->ids[i];
 
@@ -1410,9 +1441,11 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 	INIT_LIST_HEAD(&adev->vm_manager.ids_lru);
 
 	/* skip over VMID 0, since it is the system VM */
-	for (i = 1; i < adev->vm_manager.num_ids; ++i)
+	for (i = 1; i < adev->vm_manager.num_ids; ++i) {
+		amdgpu_vm_reset_id(adev, i);
 		list_add_tail(&adev->vm_manager.ids[i].list,
 			      &adev->vm_manager.ids_lru);
+	}
 
 	atomic_set(&adev->vm_manager.vm_pte_next_ring, 0);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
index 474ca02..1f9109d 100644
--- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
+++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
@@ -3017,7 +3017,6 @@ static int ci_populate_single_memory_level(struct amdgpu_device *adev,
 						      &memory_level->MinVddcPhases);
 
 	memory_level->EnabledForThrottle = 1;
-	memory_level->EnabledForActivity = 1;
 	memory_level->UpH = 0;
 	memory_level->DownH = 100;
 	memory_level->VoltageDownH = 0;
@@ -3376,7 +3375,6 @@ static int ci_populate_single_graphic_level(struct amdgpu_device *adev,
 	graphic_level->SpllSpreadSpectrum2 = cpu_to_be32(graphic_level->SpllSpreadSpectrum2);
 	graphic_level->CcPwrDynRm = cpu_to_be32(graphic_level->CcPwrDynRm);
 	graphic_level->CcPwrDynRm1 = cpu_to_be32(graphic_level->CcPwrDynRm1);
-	graphic_level->EnabledForActivity = 1;
 
 	return 0;
 }
@@ -3407,6 +3405,7 @@ static int ci_populate_all_graphic_levels(struct amdgpu_device *adev)
 			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 =
@@ -3450,6 +3449,8 @@ static int ci_populate_all_memory_levels(struct amdgpu_device *adev)
 			return ret;
 	}
 
+	pi->smc_state_table.MemoryLevel[0].EnabledForActivity = 1;
+
 	if ((dpm_table->mclk_table.count >= 2) &&
 	    ((adev->pdev->device == 0x67B0) || (adev->pdev->device == 0x67B1))) {
 		pi->smc_state_table.MemoryLevel[1].MinVddc =
@@ -4381,26 +4382,6 @@ static int ci_dpm_force_performance_level(struct amdgpu_device *adev,
 				}
 			}
 		}
-		if ((!pi->pcie_dpm_key_disabled) &&
-		    pi->dpm_level_enable_mask.pcie_dpm_enable_mask) {
-			levels = 0;
-			tmp = pi->dpm_level_enable_mask.pcie_dpm_enable_mask;
-			while (tmp >>= 1)
-				levels++;
-			if (levels) {
-				ret = ci_dpm_force_state_pcie(adev, level);
-				if (ret)
-					return ret;
-				for (i = 0; i < adev->usec_timeout; i++) {
-					tmp = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX_1) &
-					TARGET_AND_CURRENT_PROFILE_INDEX_1__CURR_PCIE_INDEX_MASK) >>
-					TARGET_AND_CURRENT_PROFILE_INDEX_1__CURR_PCIE_INDEX__SHIFT;
-					if (tmp == levels)
-						break;
-					udelay(1);
-				}
-			}
-		}
 	} else if (level == AMDGPU_DPM_FORCED_LEVEL_LOW) {
 		if ((!pi->sclk_dpm_key_disabled) &&
 		    pi->dpm_level_enable_mask.sclk_dpm_enable_mask) {
@@ -5395,30 +5376,6 @@ static int ci_dpm_enable(struct amdgpu_device *adev)
 
 	ci_update_current_ps(adev, boot_ps);
 
-	if (adev->irq.installed &&
-	    amdgpu_is_internal_thermal_sensor(adev->pm.int_thermal_type)) {
-#if 0
-		PPSMC_Result result;
-#endif
-		ret = ci_thermal_set_temperature_range(adev, CISLANDS_TEMP_RANGE_MIN,
-						       CISLANDS_TEMP_RANGE_MAX);
-		if (ret) {
-			DRM_ERROR("ci_thermal_set_temperature_range failed\n");
-			return ret;
-		}
-		amdgpu_irq_get(adev, &adev->pm.dpm.thermal.irq,
-			       AMDGPU_THERMAL_IRQ_LOW_TO_HIGH);
-		amdgpu_irq_get(adev, &adev->pm.dpm.thermal.irq,
-			       AMDGPU_THERMAL_IRQ_HIGH_TO_LOW);
-
-#if 0
-		result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_EnableThermalInterrupt);
-
-		if (result != PPSMC_Result_OK)
-			DRM_DEBUG_KMS("Could not enable thermal interrupts.\n");
-#endif
-	}
-
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c
index 192ab13..bddc9ba 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik.c
@@ -2028,8 +2028,6 @@ static int cik_common_early_init(void *handle)
 
 	adev->asic_funcs = &cik_asic_funcs;
 
-	adev->has_uvd = true;
-
 	adev->rev_id = cik_get_rev_id(adev);
 	adev->external_rev_id = 0xFF;
 	switch (adev->asic_type) {
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index 266db15..0212219 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -261,6 +261,13 @@ static void cik_sdma_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, (0xfff << 16) | 10); /* retry count, poll interval */
 }
 
+static void cik_sdma_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
+{
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, mmHDP_DEBUG0);
+	amdgpu_ring_write(ring, 1);
+}
+
 /**
  * cik_sdma_ring_emit_fence - emit a fence on the DMA ring
  *
@@ -636,8 +643,7 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring)
 	ib.ptr[3] = 1;
 	ib.ptr[4] = 0xDEADBEEF;
 	ib.length_dw = 5;
-	r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED,
-			       NULL, &f);
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
 	if (r)
 		goto err1;
 
@@ -816,6 +822,30 @@ static void cik_sdma_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
 }
 
 /**
+ * cik_sdma_ring_emit_pipeline_sync - sync the pipeline
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Make sure all previous operations are completed (CIK).
+ */
+static void cik_sdma_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	/* wait for idle */
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0,
+					    SDMA_POLL_REG_MEM_EXTRA_OP(0) |
+					    SDMA_POLL_REG_MEM_EXTRA_FUNC(3) | /* equal */
+					    SDMA_POLL_REG_MEM_EXTRA_M));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff);
+	amdgpu_ring_write(ring, seq); /* reference */
+	amdgpu_ring_write(ring, 0xfffffff); /* mask */
+	amdgpu_ring_write(ring, (0xfff << 16) | 4); /* retry count, poll interval */
+}
+
+/**
  * cik_sdma_ring_emit_vm_flush - cik vm flush using sDMA
  *
  * @ring: amdgpu_ring pointer
@@ -1270,8 +1300,10 @@ static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = {
 	.parse_cs = NULL,
 	.emit_ib = cik_sdma_ring_emit_ib,
 	.emit_fence = cik_sdma_ring_emit_fence,
+	.emit_pipeline_sync = cik_sdma_ring_emit_pipeline_sync,
 	.emit_vm_flush = cik_sdma_ring_emit_vm_flush,
 	.emit_hdp_flush = cik_sdma_ring_emit_hdp_flush,
+	.emit_hdp_invalidate = cik_sdma_ring_emit_hdp_invalidate,
 	.test_ring = cik_sdma_ring_test_ring,
 	.test_ib = cik_sdma_ring_test_ib,
 	.insert_nop = cik_sdma_ring_insert_nop,
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
index e3ff809..6de2ce53 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
@@ -1668,6 +1668,9 @@ static void dce_v10_0_audio_fini(struct amdgpu_device *adev)
 {
 	int i;
 
+	if (!amdgpu_audio)
+		return;
+
 	if (!adev->mode_info.audio.enabled)
 		return;
 
@@ -1973,7 +1976,7 @@ static void dce_v10_0_afmt_enable(struct drm_encoder *encoder, bool enable)
 		  enable ? "En" : "Dis", dig->afmt->offset, amdgpu_encoder->encoder_id);
 }
 
-static void dce_v10_0_afmt_init(struct amdgpu_device *adev)
+static int dce_v10_0_afmt_init(struct amdgpu_device *adev)
 {
 	int i;
 
@@ -1986,8 +1989,16 @@ static void dce_v10_0_afmt_init(struct amdgpu_device *adev)
 		if (adev->mode_info.afmt[i]) {
 			adev->mode_info.afmt[i]->offset = dig_offsets[i];
 			adev->mode_info.afmt[i]->id = i;
+		} else {
+			int j;
+			for (j = 0; j < i; j++) {
+				kfree(adev->mode_info.afmt[j]);
+				adev->mode_info.afmt[j] = NULL;
+			}
+			return -ENOMEM;
 		}
 	}
+	return 0;
 }
 
 static void dce_v10_0_afmt_fini(struct amdgpu_device *adev)
@@ -2064,8 +2075,7 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc,
 	if (atomic) {
 		amdgpu_fb = to_amdgpu_framebuffer(fb);
 		target_fb = fb;
-	}
-	else {
+	} else {
 		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
 		target_fb = crtc->primary->fb;
 	}
@@ -2079,9 +2089,9 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc,
 	if (unlikely(r != 0))
 		return r;
 
-	if (atomic)
+	if (atomic) {
 		fb_location = amdgpu_bo_gpu_offset(rbo);
-	else {
+	} else {
 		r = amdgpu_bo_pin(rbo, AMDGPU_GEM_DOMAIN_VRAM, &fb_location);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(rbo);
@@ -2700,13 +2710,13 @@ static void dce_v10_0_crtc_dpms(struct drm_crtc *crtc, int mode)
 		type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id);
 		amdgpu_irq_update(adev, &adev->crtc_irq, type);
 		amdgpu_irq_update(adev, &adev->pageflip_irq, type);
-		drm_vblank_post_modeset(dev, amdgpu_crtc->crtc_id);
+		drm_vblank_on(dev, amdgpu_crtc->crtc_id);
 		dce_v10_0_crtc_load_lut(crtc);
 		break;
 	case DRM_MODE_DPMS_STANDBY:
 	case DRM_MODE_DPMS_SUSPEND:
 	case DRM_MODE_DPMS_OFF:
-		drm_vblank_pre_modeset(dev, amdgpu_crtc->crtc_id);
+		drm_vblank_off(dev, amdgpu_crtc->crtc_id);
 		if (amdgpu_crtc->enabled) {
 			dce_v10_0_vga_enable(crtc, true);
 			amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE);
@@ -2980,8 +2990,6 @@ static int dce_v10_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	adev->mode_info.mode_config_initialized = true;
-
 	adev->ddev->mode_config.funcs = &amdgpu_mode_funcs;
 
 	adev->ddev->mode_config.max_width = 16384;
@@ -3012,7 +3020,9 @@ static int dce_v10_0_sw_init(void *handle)
 		return -EINVAL;
 
 	/* setup afmt */
-	dce_v10_0_afmt_init(adev);
+	r = dce_v10_0_afmt_init(adev);
+	if (r)
+		return r;
 
 	r = dce_v10_0_audio_init(adev);
 	if (r)
@@ -3020,7 +3030,8 @@ static int dce_v10_0_sw_init(void *handle)
 
 	drm_kms_helper_poll_init(adev->ddev);
 
-	return r;
+	adev->mode_info.mode_config_initialized = true;
+	return 0;
 }
 
 static int dce_v10_0_sw_fini(void *handle)
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
index 6b6c9b6..e9ccc6b 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
@@ -1658,6 +1658,9 @@ static void dce_v11_0_audio_fini(struct amdgpu_device *adev)
 {
 	int i;
 
+	if (!amdgpu_audio)
+		return;
+
 	if (!adev->mode_info.audio.enabled)
 		return;
 
@@ -1963,7 +1966,7 @@ static void dce_v11_0_afmt_enable(struct drm_encoder *encoder, bool enable)
 		  enable ? "En" : "Dis", dig->afmt->offset, amdgpu_encoder->encoder_id);
 }
 
-static void dce_v11_0_afmt_init(struct amdgpu_device *adev)
+static int dce_v11_0_afmt_init(struct amdgpu_device *adev)
 {
 	int i;
 
@@ -1976,8 +1979,16 @@ static void dce_v11_0_afmt_init(struct amdgpu_device *adev)
 		if (adev->mode_info.afmt[i]) {
 			adev->mode_info.afmt[i]->offset = dig_offsets[i];
 			adev->mode_info.afmt[i]->id = i;
+		} else {
+			int j;
+			for (j = 0; j < i; j++) {
+				kfree(adev->mode_info.afmt[j]);
+				adev->mode_info.afmt[j] = NULL;
+			}
+			return -ENOMEM;
 		}
 	}
+	return 0;
 }
 
 static void dce_v11_0_afmt_fini(struct amdgpu_device *adev)
@@ -2054,8 +2065,7 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc,
 	if (atomic) {
 		amdgpu_fb = to_amdgpu_framebuffer(fb);
 		target_fb = fb;
-	}
-	else {
+	} else {
 		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
 		target_fb = crtc->primary->fb;
 	}
@@ -2069,9 +2079,9 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc,
 	if (unlikely(r != 0))
 		return r;
 
-	if (atomic)
+	if (atomic) {
 		fb_location = amdgpu_bo_gpu_offset(rbo);
-	else {
+	} else {
 		r = amdgpu_bo_pin(rbo, AMDGPU_GEM_DOMAIN_VRAM, &fb_location);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(rbo);
@@ -2691,13 +2701,13 @@ static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode)
 		type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id);
 		amdgpu_irq_update(adev, &adev->crtc_irq, type);
 		amdgpu_irq_update(adev, &adev->pageflip_irq, type);
-		drm_vblank_post_modeset(dev, amdgpu_crtc->crtc_id);
+		drm_vblank_on(dev, amdgpu_crtc->crtc_id);
 		dce_v11_0_crtc_load_lut(crtc);
 		break;
 	case DRM_MODE_DPMS_STANDBY:
 	case DRM_MODE_DPMS_SUSPEND:
 	case DRM_MODE_DPMS_OFF:
-		drm_vblank_pre_modeset(dev, amdgpu_crtc->crtc_id);
+		drm_vblank_off(dev, amdgpu_crtc->crtc_id);
 		if (amdgpu_crtc->enabled) {
 			dce_v11_0_vga_enable(crtc, true);
 			amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE);
@@ -2961,7 +2971,7 @@ static int dce_v11_0_sw_init(void *handle)
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
 		r = amdgpu_irq_add_id(adev, i + 1, &adev->crtc_irq);
 		if (r)
-		return r;
+			return r;
 	}
 
 	for (i = 8; i < 20; i += 2) {
@@ -2973,9 +2983,7 @@ static int dce_v11_0_sw_init(void *handle)
 	/* HPD hotplug */
 	r = amdgpu_irq_add_id(adev, 42, &adev->hpd_irq);
 	if (r)
-	return r;
-
-	adev->mode_info.mode_config_initialized = true;
+		return r;
 
 	adev->ddev->mode_config.funcs = &amdgpu_mode_funcs;
 
@@ -2994,6 +3002,7 @@ static int dce_v11_0_sw_init(void *handle)
 	adev->ddev->mode_config.max_width = 16384;
 	adev->ddev->mode_config.max_height = 16384;
 
+
 	/* allocate crtcs */
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
 		r = dce_v11_0_crtc_init(adev, i);
@@ -3007,7 +3016,9 @@ static int dce_v11_0_sw_init(void *handle)
 		return -EINVAL;
 
 	/* setup afmt */
-	dce_v11_0_afmt_init(adev);
+	r = dce_v11_0_afmt_init(adev);
+	if (r)
+		return r;
 
 	r = dce_v11_0_audio_init(adev);
 	if (r)
@@ -3015,7 +3026,8 @@ static int dce_v11_0_sw_init(void *handle)
 
 	drm_kms_helper_poll_init(adev->ddev);
 
-	return r;
+	adev->mode_info.mode_config_initialized = true;
+	return 0;
 }
 
 static int dce_v11_0_sw_fini(void *handle)
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
index 56bea36..e56b55d 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
@@ -1639,6 +1639,9 @@ static void dce_v8_0_audio_fini(struct amdgpu_device *adev)
 {
 	int i;
 
+	if (!amdgpu_audio)
+		return;
+
 	if (!adev->mode_info.audio.enabled)
 		return;
 
@@ -1910,7 +1913,7 @@ static void dce_v8_0_afmt_enable(struct drm_encoder *encoder, bool enable)
 		  enable ? "En" : "Dis", dig->afmt->offset, amdgpu_encoder->encoder_id);
 }
 
-static void dce_v8_0_afmt_init(struct amdgpu_device *adev)
+static int dce_v8_0_afmt_init(struct amdgpu_device *adev)
 {
 	int i;
 
@@ -1923,8 +1926,16 @@ static void dce_v8_0_afmt_init(struct amdgpu_device *adev)
 		if (adev->mode_info.afmt[i]) {
 			adev->mode_info.afmt[i]->offset = dig_offsets[i];
 			adev->mode_info.afmt[i]->id = i;
+		} else {
+			int j;
+			for (j = 0; j < i; j++) {
+				kfree(adev->mode_info.afmt[j]);
+				adev->mode_info.afmt[j] = NULL;
+			}
+			return -ENOMEM;
 		}
 	}
+	return 0;
 }
 
 static void dce_v8_0_afmt_fini(struct amdgpu_device *adev)
@@ -2001,8 +2012,7 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc,
 	if (atomic) {
 		amdgpu_fb = to_amdgpu_framebuffer(fb);
 		target_fb = fb;
-	}
-	else {
+	} else {
 		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
 		target_fb = crtc->primary->fb;
 	}
@@ -2016,9 +2026,9 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc,
 	if (unlikely(r != 0))
 		return r;
 
-	if (atomic)
+	if (atomic) {
 		fb_location = amdgpu_bo_gpu_offset(rbo);
-	else {
+	} else {
 		r = amdgpu_bo_pin(rbo, AMDGPU_GEM_DOMAIN_VRAM, &fb_location);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(rbo);
@@ -2612,13 +2622,13 @@ static void dce_v8_0_crtc_dpms(struct drm_crtc *crtc, int mode)
 		type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id);
 		amdgpu_irq_update(adev, &adev->crtc_irq, type);
 		amdgpu_irq_update(adev, &adev->pageflip_irq, type);
-		drm_vblank_post_modeset(dev, amdgpu_crtc->crtc_id);
+		drm_vblank_on(dev, amdgpu_crtc->crtc_id);
 		dce_v8_0_crtc_load_lut(crtc);
 		break;
 	case DRM_MODE_DPMS_STANDBY:
 	case DRM_MODE_DPMS_SUSPEND:
 	case DRM_MODE_DPMS_OFF:
-		drm_vblank_pre_modeset(dev, amdgpu_crtc->crtc_id);
+		drm_vblank_off(dev, amdgpu_crtc->crtc_id);
 		if (amdgpu_crtc->enabled) {
 			dce_v8_0_vga_enable(crtc, true);
 			amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE);
@@ -2890,8 +2900,6 @@ static int dce_v8_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	adev->mode_info.mode_config_initialized = true;
-
 	adev->ddev->mode_config.funcs = &amdgpu_mode_funcs;
 
 	adev->ddev->mode_config.max_width = 16384;
@@ -2922,7 +2930,9 @@ static int dce_v8_0_sw_init(void *handle)
 		return -EINVAL;
 
 	/* setup afmt */
-	dce_v8_0_afmt_init(adev);
+	r = dce_v8_0_afmt_init(adev);
+	if (r)
+		return r;
 
 	r = dce_v8_0_audio_init(adev);
 	if (r)
@@ -2930,7 +2940,8 @@ static int dce_v8_0_sw_init(void *handle)
 
 	drm_kms_helper_poll_init(adev->ddev);
 
-	return r;
+	adev->mode_info.mode_config_initialized = true;
+	return 0;
 }
 
 static int dce_v8_0_sw_fini(void *handle)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 4411b947..a06045f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -1925,6 +1925,25 @@ static void gfx_v7_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 }
 
 /**
+ * gfx_v7_0_ring_emit_hdp_invalidate - emit an hdp invalidate on the cp
+ *
+ * @adev: amdgpu_device pointer
+ * @ridx: amdgpu ring index
+ *
+ * Emits an hdp invalidate on the cp.
+ */
+static void gfx_v7_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
+{
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(0) |
+				 WR_CONFIRM));
+	amdgpu_ring_write(ring, mmHDP_DEBUG0);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, 1);
+}
+
+/**
  * gfx_v7_0_ring_emit_fence_gfx - emit a fence on the gfx ring
  *
  * @adev: amdgpu_device pointer
@@ -2117,8 +2136,7 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring)
 	ib.ptr[2] = 0xDEADBEEF;
 	ib.length_dw = 3;
 
-	r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED,
-			       NULL, &f);
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
 	if (r)
 		goto err2;
 
@@ -3023,6 +3041,26 @@ static int gfx_v7_0_cp_resume(struct amdgpu_device *adev)
 	return 0;
 }
 
+/**
+ * gfx_v7_0_ring_emit_vm_flush - cik vm flush using the CP
+ *
+ * @ring: the ring to emmit the commands to
+ *
+ * Sync the command pipeline with the PFP. E.g. wait for everything
+ * to be completed.
+ */
+static void gfx_v7_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
+	if (usepfp) {
+		/* synce CE with ME to prevent CE fetch CEIB before context switch done */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+	}
+}
+
 /*
  * vm
  * VMID 0 is the physical GPU addresses as used by the kernel.
@@ -3054,14 +3092,6 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, 0xffffffff);
 	amdgpu_ring_write(ring, 4); /* poll interval */
 
-	if (usepfp) {
-		/* synce CE with ME to prevent CE fetch CEIB before context switch done */
-		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
-		amdgpu_ring_write(ring, 0);
-		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
-		amdgpu_ring_write(ring, 0);
-	}
-
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
 	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) |
 				 WRITE_DATA_DST_SEL(0)));
@@ -5142,9 +5172,11 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
 	.parse_cs = NULL,
 	.emit_ib = gfx_v7_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v7_0_ring_emit_fence_gfx,
+	.emit_pipeline_sync = gfx_v7_0_ring_emit_pipeline_sync,
 	.emit_vm_flush = gfx_v7_0_ring_emit_vm_flush,
 	.emit_gds_switch = gfx_v7_0_ring_emit_gds_switch,
 	.emit_hdp_flush = gfx_v7_0_ring_emit_hdp_flush,
+	.emit_hdp_invalidate = gfx_v7_0_ring_emit_hdp_invalidate,
 	.test_ring = gfx_v7_0_ring_test_ring,
 	.test_ib = gfx_v7_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
@@ -5158,9 +5190,11 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
 	.parse_cs = NULL,
 	.emit_ib = gfx_v7_0_ring_emit_ib_compute,
 	.emit_fence = gfx_v7_0_ring_emit_fence_compute,
+	.emit_pipeline_sync = gfx_v7_0_ring_emit_pipeline_sync,
 	.emit_vm_flush = gfx_v7_0_ring_emit_vm_flush,
 	.emit_gds_switch = gfx_v7_0_ring_emit_gds_switch,
 	.emit_hdp_flush = gfx_v7_0_ring_emit_hdp_flush,
+	.emit_hdp_invalidate = gfx_v7_0_ring_emit_hdp_invalidate,
 	.test_ring = gfx_v7_0_ring_test_ring,
 	.test_ib = gfx_v7_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index 1b85c00..509d0ba 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -706,8 +706,7 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring)
 	ib.ptr[2] = 0xDEADBEEF;
 	ib.length_dw = 3;
 
-	r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED,
-			       NULL, &f);
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
 	if (r)
 		goto err2;
 
@@ -1262,8 +1261,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
 	ib.ptr[ib.length_dw++] = EVENT_TYPE(7) | EVENT_INDEX(4);
 
 	/* shedule the ib on the ring */
-	r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED,
-			       NULL, &f);
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
 	if (r) {
 		DRM_ERROR("amdgpu: ib submit failed (%d).\n", r);
 		goto fail;
@@ -4589,6 +4587,18 @@ static void gfx_v8_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, 0x20); /* poll interval */
 }
 
+static void gfx_v8_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
+{
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(0) |
+				 WR_CONFIRM));
+	amdgpu_ring_write(ring, mmHDP_DEBUG0);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, 1);
+
+}
+
 static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 				  struct amdgpu_ib *ib)
 {
@@ -4682,8 +4692,7 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 
 }
 
-static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					unsigned vm_id, uint64_t pd_addr)
+static void gfx_v8_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
 {
 	int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
 	uint32_t seq = ring->fence_drv.sync_seq;
@@ -4706,6 +4715,12 @@ static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
 		amdgpu_ring_write(ring, 0);
 	}
+}
+
+static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					unsigned vm_id, uint64_t pd_addr)
+{
+	int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
 	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) |
@@ -5028,9 +5043,11 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
 	.parse_cs = NULL,
 	.emit_ib = gfx_v8_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v8_0_ring_emit_fence_gfx,
+	.emit_pipeline_sync = gfx_v8_0_ring_emit_pipeline_sync,
 	.emit_vm_flush = gfx_v8_0_ring_emit_vm_flush,
 	.emit_gds_switch = gfx_v8_0_ring_emit_gds_switch,
 	.emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush,
+	.emit_hdp_invalidate = gfx_v8_0_ring_emit_hdp_invalidate,
 	.test_ring = gfx_v8_0_ring_test_ring,
 	.test_ib = gfx_v8_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
@@ -5044,9 +5061,11 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
 	.parse_cs = NULL,
 	.emit_ib = gfx_v8_0_ring_emit_ib_compute,
 	.emit_fence = gfx_v8_0_ring_emit_fence_compute,
+	.emit_pipeline_sync = gfx_v8_0_ring_emit_pipeline_sync,
 	.emit_vm_flush = gfx_v8_0_ring_emit_vm_flush,
 	.emit_gds_switch = gfx_v8_0_ring_emit_gds_switch,
 	.emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush,
+	.emit_hdp_invalidate = gfx_v8_0_ring_emit_hdp_invalidate,
 	.test_ring = gfx_v8_0_ring_test_ring,
 	.test_ib = gfx_v8_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
index 711840a..82ce7d9 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
@@ -339,7 +339,7 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
 	WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
 
 	tmp = RREG32(mmHDP_MISC_CNTL);
-	tmp = REG_SET_FIELD(tmp, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 0);
 	WREG32(mmHDP_MISC_CNTL, tmp);
 
 	tmp = RREG32(mmHDP_HOST_PATH_CNTL);
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
index 757803a..29bd7b5 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
@@ -386,7 +386,7 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev)
 	WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
 
 	tmp = RREG32(mmHDP_MISC_CNTL);
-	tmp = REG_SET_FIELD(tmp, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 0);
 	WREG32(mmHDP_MISC_CNTL, tmp);
 
 	tmp = RREG32(mmHDP_HOST_PATH_CNTL);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
index dddb8d6..ab9ff89 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
@@ -300,6 +300,13 @@ static void sdma_v2_4_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */
 }
 
+static void sdma_v2_4_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
+{
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
+			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
+	amdgpu_ring_write(ring, mmHDP_DEBUG0);
+	amdgpu_ring_write(ring, 1);
+}
 /**
  * sdma_v2_4_ring_emit_fence - emit a fence on the DMA ring
  *
@@ -694,8 +701,7 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring)
 	ib.ptr[7] = SDMA_PKT_HEADER_OP(SDMA_OP_NOP);
 	ib.length_dw = 8;
 
-	r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED,
-			       NULL, &f);
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
 	if (r)
 		goto err1;
 
@@ -874,6 +880,31 @@ static void sdma_v2_4_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib
 }
 
 /**
+ * sdma_v2_4_ring_emit_pipeline_sync - sync the pipeline
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Make sure all previous operations are completed (CIK).
+ */
+static void sdma_v2_4_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	/* wait for idle */
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3) | /* equal */
+			  SDMA_PKT_POLL_REGMEM_HEADER_MEM_POLL(1));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff);
+	amdgpu_ring_write(ring, seq); /* reference */
+	amdgpu_ring_write(ring, 0xfffffff); /* mask */
+	amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */
+}
+
+/**
  * sdma_v2_4_ring_emit_vm_flush - cik vm flush using sDMA
  *
  * @ring: amdgpu_ring pointer
@@ -1274,8 +1305,10 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = {
 	.parse_cs = NULL,
 	.emit_ib = sdma_v2_4_ring_emit_ib,
 	.emit_fence = sdma_v2_4_ring_emit_fence,
+	.emit_pipeline_sync = sdma_v2_4_ring_emit_pipeline_sync,
 	.emit_vm_flush = sdma_v2_4_ring_emit_vm_flush,
 	.emit_hdp_flush = sdma_v2_4_ring_emit_hdp_flush,
+	.emit_hdp_invalidate = sdma_v2_4_ring_emit_hdp_invalidate,
 	.test_ring = sdma_v2_4_ring_test_ring,
 	.test_ib = sdma_v2_4_ring_test_ib,
 	.insert_nop = sdma_v2_4_ring_insert_nop,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
index 19e02f7..4c24c37 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
@@ -410,6 +410,14 @@ static void sdma_v3_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */
 }
 
+static void sdma_v3_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
+{
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
+			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
+	amdgpu_ring_write(ring, mmHDP_DEBUG0);
+	amdgpu_ring_write(ring, 1);
+}
+
 /**
  * sdma_v3_0_ring_emit_fence - emit a fence on the DMA ring
  *
@@ -845,8 +853,7 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring)
 	ib.ptr[7] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP);
 	ib.length_dw = 8;
 
-	r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED,
-			       NULL, &f);
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
 	if (r)
 		goto err1;
 
@@ -1024,6 +1031,31 @@ static void sdma_v3_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib
 }
 
 /**
+ * sdma_v3_0_ring_emit_pipeline_sync - sync the pipeline
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Make sure all previous operations are completed (CIK).
+ */
+static void sdma_v3_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	/* wait for idle */
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3) | /* equal */
+			  SDMA_PKT_POLL_REGMEM_HEADER_MEM_POLL(1));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff);
+	amdgpu_ring_write(ring, seq); /* reference */
+	amdgpu_ring_write(ring, 0xfffffff); /* mask */
+	amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */
+}
+
+/**
  * sdma_v3_0_ring_emit_vm_flush - cik vm flush using sDMA
  *
  * @ring: amdgpu_ring pointer
@@ -1541,8 +1573,10 @@ static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = {
 	.parse_cs = NULL,
 	.emit_ib = sdma_v3_0_ring_emit_ib,
 	.emit_fence = sdma_v3_0_ring_emit_fence,
+	.emit_pipeline_sync = sdma_v3_0_ring_emit_pipeline_sync,
 	.emit_vm_flush = sdma_v3_0_ring_emit_vm_flush,
 	.emit_hdp_flush = sdma_v3_0_ring_emit_hdp_flush,
+	.emit_hdp_invalidate = sdma_v3_0_ring_emit_hdp_invalidate,
 	.test_ring = sdma_v3_0_ring_test_ring,
 	.test_ib = sdma_v3_0_ring_test_ib,
 	.insert_nop = sdma_v3_0_ring_insert_nop,
diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c
index b72cf06..1c120ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/vi.c
+++ b/drivers/gpu/drm/amd/amdgpu/vi.c
@@ -1071,26 +1071,22 @@ static int vi_common_early_init(void *handle)
 	adev->external_rev_id = 0xFF;
 	switch (adev->asic_type) {
 	case CHIP_TOPAZ:
-		adev->has_uvd = false;
 		adev->cg_flags = 0;
 		adev->pg_flags = 0;
 		adev->external_rev_id = 0x1;
 		break;
 	case CHIP_FIJI:
-		adev->has_uvd = true;
 		adev->cg_flags = 0;
 		adev->pg_flags = 0;
 		adev->external_rev_id = adev->rev_id + 0x3c;
 		break;
 	case CHIP_TONGA:
-		adev->has_uvd = true;
 		adev->cg_flags = 0;
 		adev->pg_flags = 0;
 		adev->external_rev_id = adev->rev_id + 0x14;
 		break;
 	case CHIP_CARRIZO:
 	case CHIP_STONEY:
-		adev->has_uvd = true;
 		adev->cg_flags = 0;
 		adev->pg_flags = 0;
 		adev->external_rev_id = adev->rev_id + 0x1;
diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/cz_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/cz_hwmgr.c
index b8d6a82..727d5c9 100644
--- a/drivers/gpu/drm/amd/powerplay/hwmgr/cz_hwmgr.c
+++ b/drivers/gpu/drm/amd/powerplay/hwmgr/cz_hwmgr.c
@@ -744,8 +744,9 @@ static int cz_tf_update_sclk_limit(struct pp_hwmgr *hwmgr,
 		cz_hwmgr->sclk_dpm.soft_max_clk  = table->entries[table->count - 1].clk;
 
 	clock = hwmgr->display_config.min_core_set_clock;
+;
 	if (clock == 0)
-		printk(KERN_ERR "[ powerplay ] min_core_set_clock not set\n");
+		printk(KERN_INFO "[ powerplay ] min_core_set_clock not set\n");
 
 	if (cz_hwmgr->sclk_dpm.hard_min_clk != clock) {
 		cz_hwmgr->sclk_dpm.hard_min_clk = clock;
diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/functiontables.c b/drivers/gpu/drm/amd/powerplay/hwmgr/functiontables.c
index 9deadab..72cfecc 100644
--- a/drivers/gpu/drm/amd/powerplay/hwmgr/functiontables.c
+++ b/drivers/gpu/drm/amd/powerplay/hwmgr/functiontables.c
@@ -34,6 +34,11 @@ static int phm_run_table(struct pp_hwmgr *hwmgr,
 	int result = 0;
 	phm_table_function *function;
 
+	if (rt_table->function_list == NULL) {
+		printk(KERN_INFO "[ powerplay ] this function not implement!\n");
+		return 0;
+	}
+
 	for (function = rt_table->function_list; NULL != *function; function++) {
 		int tmp = (*function)(hwmgr, input, output, temp_storage, result);
 
@@ -57,9 +62,9 @@ int phm_dispatch_table(struct pp_hwmgr *hwmgr,
 	int result = 0;
 	void *temp_storage = NULL;
 
-	if (hwmgr == NULL || rt_table == NULL || rt_table->function_list == NULL) {
+	if (hwmgr == NULL || rt_table == NULL) {
 		printk(KERN_ERR "[ powerplay ] Invalid Parameter!\n");
-		return 0; /*temp return ture because some function not implement on some asic */
+		return -EINVAL;
 	}
 
 	if (0 != rt_table->storage_size) {
diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c
index 801dd60..e187bec 100644
--- a/drivers/gpu/drm/radeon/atombios_crtc.c
+++ b/drivers/gpu/drm/radeon/atombios_crtc.c
@@ -275,13 +275,13 @@ void atombios_crtc_dpms(struct drm_crtc *crtc, int mode)
 		if (ASIC_IS_DCE3(rdev) && !ASIC_IS_DCE6(rdev))
 			atombios_enable_crtc_memreq(crtc, ATOM_ENABLE);
 		atombios_blank_crtc(crtc, ATOM_DISABLE);
-		drm_vblank_post_modeset(dev, radeon_crtc->crtc_id);
+		drm_vblank_on(dev, radeon_crtc->crtc_id);
 		radeon_crtc_load_lut(crtc);
 		break;
 	case DRM_MODE_DPMS_STANDBY:
 	case DRM_MODE_DPMS_SUSPEND:
 	case DRM_MODE_DPMS_OFF:
-		drm_vblank_pre_modeset(dev, radeon_crtc->crtc_id);
+		drm_vblank_off(dev, radeon_crtc->crtc_id);
 		if (radeon_crtc->enabled)
 			atombios_blank_crtc(crtc, ATOM_ENABLE);
 		if (ASIC_IS_DCE3(rdev) && !ASIC_IS_DCE6(rdev))
diff --git a/drivers/gpu/drm/radeon/atombios_encoders.c b/drivers/gpu/drm/radeon/atombios_encoders.c
index 1603751..edd05cd 100644
--- a/drivers/gpu/drm/radeon/atombios_encoders.c
+++ b/drivers/gpu/drm/radeon/atombios_encoders.c
@@ -892,8 +892,6 @@ atombios_dig_encoder_setup2(struct drm_encoder *encoder, int action, int panel_m
 			else
 				args.v1.ucLaneNum = 4;
 
-			if (ENCODER_MODE_IS_DP(args.v1.ucEncoderMode) && (dp_clock == 270000))
-				args.v1.ucConfig |= ATOM_ENCODER_CONFIG_DPLINKRATE_2_70GHZ;
 			switch (radeon_encoder->encoder_id) {
 			case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
 				args.v1.ucConfig = ATOM_ENCODER_CONFIG_V2_TRANSMITTER1;
@@ -910,6 +908,10 @@ atombios_dig_encoder_setup2(struct drm_encoder *encoder, int action, int panel_m
 				args.v1.ucConfig |= ATOM_ENCODER_CONFIG_LINKB;
 			else
 				args.v1.ucConfig |= ATOM_ENCODER_CONFIG_LINKA;
+
+			if (ENCODER_MODE_IS_DP(args.v1.ucEncoderMode) && (dp_clock == 270000))
+				args.v1.ucConfig |= ATOM_ENCODER_CONFIG_DPLINKRATE_2_70GHZ;
+
 			break;
 		case 2:
 		case 3:
diff --git a/drivers/gpu/drm/radeon/radeon_atpx_handler.c b/drivers/gpu/drm/radeon/radeon_atpx_handler.c
index 56482e3..fd8c4d3 100644
--- a/drivers/gpu/drm/radeon/radeon_atpx_handler.c
+++ b/drivers/gpu/drm/radeon/radeon_atpx_handler.c
@@ -62,6 +62,10 @@ bool radeon_has_atpx(void) {
 	return radeon_atpx_priv.atpx_detected;
 }
 
+bool radeon_has_atpx_dgpu_power_cntl(void) {
+	return radeon_atpx_priv.atpx.functions.power_cntl;
+}
+
 /**
  * radeon_atpx_call - call an ATPX method
  *
@@ -141,10 +145,6 @@ static void radeon_atpx_parse_functions(struct radeon_atpx_functions *f, u32 mas
  */
 static int radeon_atpx_validate(struct radeon_atpx *atpx)
 {
-	/* make sure required functions are enabled */
-	/* dGPU power control is required */
-	atpx->functions.power_cntl = true;
-
 	if (atpx->functions.px_params) {
 		union acpi_object *info;
 		struct atpx_px_params output;
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index 4197ca1..e239633 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -103,6 +103,12 @@ static const char radeon_family_name[][16] = {
 	"LAST",
 };
 
+#if defined(CONFIG_VGA_SWITCHEROO)
+bool radeon_has_atpx_dgpu_power_cntl(void);
+#else
+static inline bool radeon_has_atpx_dgpu_power_cntl(void) { return false; }
+#endif
+
 #define RADEON_PX_QUIRK_DISABLE_PX  (1 << 0)
 #define RADEON_PX_QUIRK_LONG_WAKEUP (1 << 1)
 
@@ -1433,7 +1439,7 @@ int radeon_device_init(struct radeon_device *rdev,
 	 * ignore it */
 	vga_client_register(rdev->pdev, rdev, NULL, radeon_vga_set_decode);
 
-	if (rdev->flags & RADEON_IS_PX)
+	if ((rdev->flags & RADEON_IS_PX) && radeon_has_atpx_dgpu_power_cntl())
 		runtime = true;
 	vga_switcheroo_register_client(rdev->pdev, &radeon_switcheroo_ops, runtime);
 	if (runtime)
diff --git a/drivers/gpu/drm/radeon/radeon_legacy_crtc.c b/drivers/gpu/drm/radeon/radeon_legacy_crtc.c
index 32b338f..24152df 100644
--- a/drivers/gpu/drm/radeon/radeon_legacy_crtc.c
+++ b/drivers/gpu/drm/radeon/radeon_legacy_crtc.c
@@ -331,13 +331,13 @@ static void radeon_crtc_dpms(struct drm_crtc *crtc, int mode)
 									 RADEON_CRTC_DISP_REQ_EN_B));
 			WREG32_P(RADEON_CRTC_EXT_CNTL, crtc_ext_cntl, ~(mask | crtc_ext_cntl));
 		}
-		drm_vblank_post_modeset(dev, radeon_crtc->crtc_id);
+		drm_vblank_on(dev, radeon_crtc->crtc_id);
 		radeon_crtc_load_lut(crtc);
 		break;
 	case DRM_MODE_DPMS_STANDBY:
 	case DRM_MODE_DPMS_SUSPEND:
 	case DRM_MODE_DPMS_OFF:
-		drm_vblank_pre_modeset(dev, radeon_crtc->crtc_id);
+		drm_vblank_off(dev, radeon_crtc->crtc_id);
 		if (radeon_crtc->crtc_id)
 			WREG32_P(RADEON_CRTC2_GEN_CNTL, mask, ~(RADEON_CRTC2_EN | mask));
 		else {