diff options
Diffstat (limited to 'drivers/xen')
28 files changed, 5737 insertions, 186 deletions
diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig index a59638b..8795480 100644 --- a/drivers/xen/Kconfig +++ b/drivers/xen/Kconfig @@ -9,6 +9,53 @@ config XEN_BALLOON the system to expand the domain's memory allocation, or alternatively return unneeded memory to the system. +config XEN_SELFBALLOONING + bool "Dynamically self-balloon kernel memory to target" + depends on XEN && XEN_BALLOON && CLEANCACHE && SWAP && XEN_TMEM + default n + help + Self-ballooning dynamically balloons available kernel memory driven + by the current usage of anonymous memory ("committed AS") and + controlled by various sysfs-settable parameters. Configuring + FRONTSWAP is highly recommended; if it is not configured, self- + ballooning is disabled by default but can be enabled with the + 'selfballooning' kernel boot parameter. If FRONTSWAP is configured, + frontswap-selfshrinking is enabled by default but can be disabled + with the 'noselfshrink' kernel boot parameter; and self-ballooning + is enabled by default but can be disabled with the 'noselfballooning' + kernel boot parameter. Note that systems without a sufficiently + large swap device should not enable self-ballooning. + +config XEN_BALLOON_MEMORY_HOTPLUG + bool "Memory hotplug support for Xen balloon driver" + default n + depends on XEN_BALLOON && MEMORY_HOTPLUG + help + Memory hotplug support for Xen balloon driver allows expanding memory + available for the system above limit declared at system startup. + It is very useful on critical systems which require long + run without rebooting. + + Memory could be hotplugged in following steps: + + 1) dom0: xl mem-max <domU> <maxmem> + where <maxmem> is >= requested memory size, + + 2) dom0: xl mem-set <domU> <memory> + where <memory> is requested memory size; alternatively memory + could be added by writing proper value to + /sys/devices/system/xen_memory/xen_memory0/target or + /sys/devices/system/xen_memory/xen_memory0/target_kb on dumU, + + 3) domU: for i in /sys/devices/system/memory/memory*/state; do \ + [ "`cat "$i"`" = offline ] && echo online > "$i"; done + + Memory could be onlined automatically on domU by adding following line to udev rules: + + SUBSYSTEM=="memory", ACTION=="add", RUN+="/bin/sh -c '[ -f /sys$devpath/state ] && echo online > /sys$devpath/state'" + + In that case step 3 should be omitted. + config XEN_SCRUB_PAGES bool "Scrub pages before returning them to system" depends on XEN_BALLOON @@ -90,19 +137,38 @@ config XEN_GRANT_DEV_ALLOC to other domains. This can be used to implement frontend drivers or as part of an inter-domain shared memory channel. -config XEN_PLATFORM_PCI - tristate "xen platform pci device driver" - depends on XEN_PVHVM && PCI - default m - help - Driver for the Xen PCI Platform device: it is responsible for - initializing xenbus and grant_table when running in a Xen HVM - domain. As a consequence this driver is required to run any Xen PV - frontend on Xen HVM. - config SWIOTLB_XEN def_bool y depends on PCI select SWIOTLB +config XEN_TMEM + bool + default y if (CLEANCACHE || FRONTSWAP) + help + Shim to interface in-kernel Transcendent Memory hooks + (e.g. cleancache and frontswap) to Xen tmem hypercalls. + +config XEN_PCIDEV_BACKEND + tristate "Xen PCI-device backend driver" + depends on PCI && X86 && XEN + depends on XEN_BACKEND + default m + help + The PCI device backend driver allows the kernel to export arbitrary + PCI devices to other guests. If you select this to be a module, you + will need to make sure no other driver has bound to the device(s) + you want to make visible to other guests. + + The parameter "passthrough" allows you specify how you want the PCI + devices to appear in the guest. You can choose the default (0) where + PCI topology starts at 00.00.0, or (1) for passthrough if you want + the PCI devices topology appear the same as in the host. + + The "hide" parameter (only applicable if backend driver is compiled + into the kernel) allows you to bind the PCI devices to this module + from the default device drivers. The argument is the list of PCI BDFs: + xen-pciback.hide=(03:00.0)(04:00.0) + + If in doubt, say m. endmenu diff --git a/drivers/xen/Makefile b/drivers/xen/Makefile index bbc1825..974fffd 100644 --- a/drivers/xen/Makefile +++ b/drivers/xen/Makefile @@ -1,6 +1,5 @@ obj-y += grant-table.o features.o events.o manage.o balloon.o obj-y += xenbus/ -obj-y += tmem.o nostackp := $(call cc-option, -fno-stack-protector) CFLAGS_features.o := $(nostackp) @@ -9,17 +8,18 @@ obj-$(CONFIG_BLOCK) += biomerge.o obj-$(CONFIG_HOTPLUG_CPU) += cpu_hotplug.o obj-$(CONFIG_XEN_XENCOMM) += xencomm.o obj-$(CONFIG_XEN_BALLOON) += xen-balloon.o +obj-$(CONFIG_XEN_SELFBALLOONING) += xen-selfballoon.o obj-$(CONFIG_XEN_DEV_EVTCHN) += xen-evtchn.o obj-$(CONFIG_XEN_GNTDEV) += xen-gntdev.o obj-$(CONFIG_XEN_GRANT_DEV_ALLOC) += xen-gntalloc.o obj-$(CONFIG_XENFS) += xenfs/ obj-$(CONFIG_XEN_SYS_HYPERVISOR) += sys-hypervisor.o -obj-$(CONFIG_XEN_PLATFORM_PCI) += xen-platform-pci.o +obj-$(CONFIG_XEN_PVHVM) += platform-pci.o +obj-$(CONFIG_XEN_TMEM) += tmem.o obj-$(CONFIG_SWIOTLB_XEN) += swiotlb-xen.o obj-$(CONFIG_XEN_DOM0) += pci.o +obj-$(CONFIG_XEN_PCIDEV_BACKEND) += xen-pciback/ xen-evtchn-y := evtchn.o xen-gntdev-y := gntdev.o xen-gntalloc-y := gntalloc.o - -xen-platform-pci-y := platform-pci.o diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c index f54290b..5876e1a 100644 --- a/drivers/xen/balloon.c +++ b/drivers/xen/balloon.c @@ -4,6 +4,12 @@ * Copyright (c) 2003, B Dragovic * Copyright (c) 2003-2004, M Williamson, K Fraser * Copyright (c) 2005 Dan M. Smith, IBM Corporation + * Copyright (c) 2010 Daniel Kiper + * + * Memory hotplug support was written by Daniel Kiper. Work on + * it was sponsored by Google under Google Summer of Code 2010 + * program. Jeremy Fitzhardinge from Citrix was the mentor for + * this project. * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public License version 2 @@ -40,6 +46,9 @@ #include <linux/mutex.h> #include <linux/list.h> #include <linux/gfp.h> +#include <linux/notifier.h> +#include <linux/memory.h> +#include <linux/memory_hotplug.h> #include <asm/page.h> #include <asm/pgalloc.h> @@ -194,6 +203,87 @@ static enum bp_state update_schedule(enum bp_state state) return BP_EAGAIN; } +#ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG +static long current_credit(void) +{ + return balloon_stats.target_pages - balloon_stats.current_pages - + balloon_stats.hotplug_pages; +} + +static bool balloon_is_inflated(void) +{ + if (balloon_stats.balloon_low || balloon_stats.balloon_high || + balloon_stats.balloon_hotplug) + return true; + else + return false; +} + +/* + * reserve_additional_memory() adds memory region of size >= credit above + * max_pfn. New region is section aligned and size is modified to be multiple + * of section size. Those features allow optimal use of address space and + * establish proper alignment when this function is called first time after + * boot (last section not fully populated at boot time contains unused memory + * pages with PG_reserved bit not set; online_pages_range() does not allow page + * onlining in whole range if first onlined page does not have PG_reserved + * bit set). Real size of added memory is established at page onlining stage. + */ + +static enum bp_state reserve_additional_memory(long credit) +{ + int nid, rc; + u64 hotplug_start_paddr; + unsigned long balloon_hotplug = credit; + + hotplug_start_paddr = PFN_PHYS(SECTION_ALIGN_UP(max_pfn)); + balloon_hotplug = round_up(balloon_hotplug, PAGES_PER_SECTION); + nid = memory_add_physaddr_to_nid(hotplug_start_paddr); + + rc = add_memory(nid, hotplug_start_paddr, balloon_hotplug << PAGE_SHIFT); + + if (rc) { + pr_info("xen_balloon: %s: add_memory() failed: %i\n", __func__, rc); + return BP_EAGAIN; + } + + balloon_hotplug -= credit; + + balloon_stats.hotplug_pages += credit; + balloon_stats.balloon_hotplug = balloon_hotplug; + + return BP_DONE; +} + +static void xen_online_page(struct page *page) +{ + __online_page_set_limits(page); + + mutex_lock(&balloon_mutex); + + __balloon_append(page); + + if (balloon_stats.hotplug_pages) + --balloon_stats.hotplug_pages; + else + --balloon_stats.balloon_hotplug; + + mutex_unlock(&balloon_mutex); +} + +static int xen_memory_notifier(struct notifier_block *nb, unsigned long val, void *v) +{ + if (val == MEM_ONLINE) + schedule_delayed_work(&balloon_worker, 0); + + return NOTIFY_OK; +} + +static struct notifier_block xen_memory_nb = { + .notifier_call = xen_memory_notifier, + .priority = 0 +}; +#else static long current_credit(void) { unsigned long target = balloon_stats.target_pages; @@ -206,6 +296,21 @@ static long current_credit(void) return target - balloon_stats.current_pages; } +static bool balloon_is_inflated(void) +{ + if (balloon_stats.balloon_low || balloon_stats.balloon_high) + return true; + else + return false; +} + +static enum bp_state reserve_additional_memory(long credit) +{ + balloon_stats.target_pages = balloon_stats.current_pages; + return BP_DONE; +} +#endif /* CONFIG_XEN_BALLOON_MEMORY_HOTPLUG */ + static enum bp_state increase_reservation(unsigned long nr_pages) { int rc; @@ -217,6 +322,15 @@ static enum bp_state increase_reservation(unsigned long nr_pages) .domid = DOMID_SELF }; +#ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG + if (!balloon_stats.balloon_low && !balloon_stats.balloon_high) { + nr_pages = min(nr_pages, balloon_stats.balloon_hotplug); + balloon_stats.hotplug_pages += nr_pages; + balloon_stats.balloon_hotplug -= nr_pages; + return BP_DONE; + } +#endif + if (nr_pages > ARRAY_SIZE(frame_list)) nr_pages = ARRAY_SIZE(frame_list); @@ -279,6 +393,15 @@ static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp) .domid = DOMID_SELF }; +#ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG + if (balloon_stats.hotplug_pages) { + nr_pages = min(nr_pages, balloon_stats.hotplug_pages); + balloon_stats.hotplug_pages -= nr_pages; + balloon_stats.balloon_hotplug += nr_pages; + return BP_DONE; + } +#endif + if (nr_pages > ARRAY_SIZE(frame_list)) nr_pages = ARRAY_SIZE(frame_list); @@ -340,8 +463,12 @@ static void balloon_process(struct work_struct *work) do { credit = current_credit(); - if (credit > 0) - state = increase_reservation(credit); + if (credit > 0) { + if (balloon_is_inflated()) + state = increase_reservation(credit); + else + state = reserve_additional_memory(credit); + } if (credit < 0) state = decrease_reservation(-credit, GFP_BALLOON); @@ -374,20 +501,24 @@ EXPORT_SYMBOL_GPL(balloon_set_new_target); * alloc_xenballooned_pages - get pages that have been ballooned out * @nr_pages: Number of pages to get * @pages: pages returned + * @highmem: highmem or lowmem pages * @return 0 on success, error otherwise */ -int alloc_xenballooned_pages(int nr_pages, struct page** pages) +int alloc_xenballooned_pages(int nr_pages, struct page **pages, bool highmem) { int pgno = 0; struct page* page; mutex_lock(&balloon_mutex); while (pgno < nr_pages) { - page = balloon_retrieve(true); - if (page) { + page = balloon_retrieve(highmem); + if (page && PageHighMem(page) == highmem) { pages[pgno++] = page; } else { enum bp_state st; - st = decrease_reservation(nr_pages - pgno, GFP_HIGHUSER); + if (page) + balloon_append(page); + st = decrease_reservation(nr_pages - pgno, + highmem ? GFP_HIGHUSER : GFP_USER); if (st != BP_DONE) goto out_undo; } @@ -428,17 +559,40 @@ void free_xenballooned_pages(int nr_pages, struct page** pages) } EXPORT_SYMBOL(free_xenballooned_pages); -static int __init balloon_init(void) +static void __init balloon_add_region(unsigned long start_pfn, + unsigned long pages) { unsigned long pfn, extra_pfn_end; struct page *page; + /* + * If the amount of usable memory has been limited (e.g., with + * the 'mem' command line parameter), don't add pages beyond + * this limit. + */ + extra_pfn_end = min(max_pfn, start_pfn + pages); + + for (pfn = start_pfn; pfn < extra_pfn_end; pfn++) { + page = pfn_to_page(pfn); + /* totalram_pages and totalhigh_pages do not + include the boot-time balloon extension, so + don't subtract from it. */ + __balloon_append(page); + } +} + +static int __init balloon_init(void) +{ + int i; + if (!xen_domain()) return -ENODEV; pr_info("xen/balloon: Initialising balloon driver.\n"); - balloon_stats.current_pages = xen_pv_domain() ? min(xen_start_info->nr_pages, max_pfn) : max_pfn; + balloon_stats.current_pages = xen_pv_domain() + ? min(xen_start_info->nr_pages - xen_released_pages, max_pfn) + : max_pfn; balloon_stats.target_pages = balloon_stats.current_pages; balloon_stats.balloon_low = 0; balloon_stats.balloon_high = 0; @@ -448,25 +602,22 @@ static int __init balloon_init(void) balloon_stats.retry_count = 1; balloon_stats.max_retry_count = RETRY_UNLIMITED; +#ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG + balloon_stats.hotplug_pages = 0; + balloon_stats.balloon_hotplug = 0; + + set_online_page_callback(&xen_online_page); + register_memory_notifier(&xen_memory_nb); +#endif + /* - * Initialise the balloon with excess memory space. We need - * to make sure we don't add memory which doesn't exist or - * logically exist. The E820 map can be trimmed to be smaller - * than the amount of physical memory due to the mem= command - * line parameter. And if this is a 32-bit non-HIGHMEM kernel - * on a system with memory which requires highmem to access, - * don't try to use it. + * Initialize the balloon with pages from the extra memory + * regions (see arch/x86/xen/setup.c). */ - extra_pfn_end = min(min(max_pfn, e820_end_of_ram_pfn()), - (unsigned long)PFN_DOWN(xen_extra_mem_start + xen_extra_mem_size)); - for (pfn = PFN_UP(xen_extra_mem_start); - pfn < extra_pfn_end; - pfn++) { - page = pfn_to_page(pfn); - /* totalram_pages and totalhigh_pages do not include the boot-time - balloon extension, so don't subtract from it. */ - __balloon_append(page); - } + for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) + if (xen_extra_mem[i].size) + balloon_add_region(PFN_UP(xen_extra_mem[i].start), + PFN_DOWN(xen_extra_mem[i].size)); return 0; } diff --git a/drivers/xen/events.c b/drivers/xen/events.c index 30df85d..7a55b29 100644 --- a/drivers/xen/events.c +++ b/drivers/xen/events.c @@ -54,7 +54,7 @@ * This lock protects updates to the following mapping and reference-count * arrays. The lock does not need to be acquired to read the mapping tables. */ -static DEFINE_SPINLOCK(irq_mapping_update_lock); +static DEFINE_MUTEX(irq_mapping_update_lock); static LIST_HEAD(xen_irq_list_head); @@ -432,7 +432,8 @@ static int __must_check xen_allocate_irq_dynamic(void) irq = irq_alloc_desc_from(first, -1); - xen_irq_init(irq); + if (irq >= 0) + xen_irq_init(irq); return irq; } @@ -615,11 +616,6 @@ static int find_irq_by_gsi(unsigned gsi) return -1; } -int xen_allocate_pirq_gsi(unsigned gsi) -{ - return gsi; -} - /* * Do not make any assumptions regarding the relationship between the * IRQ number returned here and the Xen pirq argument. @@ -636,7 +632,7 @@ int xen_bind_pirq_gsi_to_irq(unsigned gsi, int irq = -1; struct physdev_irq irq_op; - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); irq = find_irq_by_gsi(gsi); if (irq != -1) { @@ -689,7 +685,7 @@ int xen_bind_pirq_gsi_to_irq(unsigned gsi, handle_edge_irq, name); out: - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); return irq; } @@ -715,10 +711,10 @@ int xen_bind_pirq_msi_to_irq(struct pci_dev *dev, struct msi_desc *msidesc, { int irq, ret; - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); irq = xen_allocate_irq_dynamic(); - if (irq == -1) + if (irq < 0) goto out; irq_set_chip_and_handler_name(irq, &xen_pirq_chip, handle_edge_irq, @@ -729,12 +725,12 @@ int xen_bind_pirq_msi_to_irq(struct pci_dev *dev, struct msi_desc *msidesc, if (ret < 0) goto error_irq; out: - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); return irq; error_irq: - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); xen_free_irq(irq); - return -1; + return ret; } #endif @@ -745,7 +741,7 @@ int xen_destroy_irq(int irq) struct irq_info *info = info_for_irq(irq); int rc = -ENOENT; - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); desc = irq_to_desc(irq); if (!desc) @@ -771,7 +767,7 @@ int xen_destroy_irq(int irq) xen_free_irq(irq); out: - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); return rc; } @@ -781,10 +777,10 @@ int xen_irq_from_pirq(unsigned pirq) struct irq_info *info; - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); list_for_each_entry(info, &xen_irq_list_head, list) { - if (info == NULL || info->type != IRQT_PIRQ) + if (info->type != IRQT_PIRQ) continue; irq = info->irq; if (info->u.pirq.pirq == pirq) @@ -792,7 +788,7 @@ int xen_irq_from_pirq(unsigned pirq) } irq = -1; out: - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); return irq; } @@ -807,7 +803,7 @@ int bind_evtchn_to_irq(unsigned int evtchn) { int irq; - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); irq = evtchn_to_irq[evtchn]; @@ -823,7 +819,7 @@ int bind_evtchn_to_irq(unsigned int evtchn) } out: - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); return irq; } @@ -834,7 +830,7 @@ static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu) struct evtchn_bind_ipi bind_ipi; int evtchn, irq; - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); irq = per_cpu(ipi_to_irq, cpu)[ipi]; @@ -858,7 +854,7 @@ static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu) } out: - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); return irq; } @@ -877,13 +873,34 @@ static int bind_interdomain_evtchn_to_irq(unsigned int remote_domain, return err ? : bind_evtchn_to_irq(bind_interdomain.local_port); } +static int find_virq(unsigned int virq, unsigned int cpu) +{ + struct evtchn_status status; + int port, rc = -ENOENT; + + memset(&status, 0, sizeof(status)); + for (port = 0; port <= NR_EVENT_CHANNELS; port++) { + status.dom = DOMID_SELF; + status.port = port; + rc = HYPERVISOR_event_channel_op(EVTCHNOP_status, &status); + if (rc < 0) + continue; + if (status.status != EVTCHNSTAT_virq) + continue; + if (status.u.virq == virq && status.vcpu == cpu) { + rc = port; + break; + } + } + return rc; +} int bind_virq_to_irq(unsigned int virq, unsigned int cpu) { struct evtchn_bind_virq bind_virq; - int evtchn, irq; + int evtchn, irq, ret; - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); irq = per_cpu(virq_to_irq, cpu)[virq]; @@ -897,10 +914,16 @@ int bind_virq_to_irq(unsigned int virq, unsigned int cpu) bind_virq.virq = virq; bind_virq.vcpu = cpu; - if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq, - &bind_virq) != 0) - BUG(); - evtchn = bind_virq.port; + ret = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq, + &bind_virq); + if (ret == 0) + evtchn = bind_virq.port; + else { + if (ret == -EEXIST) + ret = find_virq(virq, cpu); + BUG_ON(ret < 0); + evtchn = ret; + } xen_irq_info_virq_init(cpu, irq, evtchn, virq); @@ -908,7 +931,7 @@ int bind_virq_to_irq(unsigned int virq, unsigned int cpu) } out: - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); return irq; } @@ -918,7 +941,7 @@ static void unbind_from_irq(unsigned int irq) struct evtchn_close close; int evtchn = evtchn_from_irq(irq); - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); if (VALID_EVTCHN(evtchn)) { close.port = evtchn; @@ -948,7 +971,7 @@ static void unbind_from_irq(unsigned int irq) xen_free_irq(irq); - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); } int bind_evtchn_to_irqhandler(unsigned int evtchn, @@ -1284,7 +1307,7 @@ void rebind_evtchn_irq(int evtchn, int irq) will also be masked. */ disable_irq(irq); - spin_lock(&irq_mapping_update_lock); + mutex_lock(&irq_mapping_update_lock); /* After resume the irq<->evtchn mappings are all cleared out */ BUG_ON(evtchn_to_irq[evtchn] != -1); @@ -1294,7 +1317,7 @@ void rebind_evtchn_irq(int evtchn, int irq) xen_irq_info_evtchn_init(irq, evtchn); - spin_unlock(&irq_mapping_update_lock); + mutex_unlock(&irq_mapping_update_lock); /* new event channels are always bound to cpu 0 */ irq_set_affinity(irq, cpumask_of(0)); @@ -1675,6 +1698,7 @@ void __init xen_init_IRQ(void) evtchn_to_irq = kcalloc(NR_EVENT_CHANNELS, sizeof(*evtchn_to_irq), GFP_KERNEL); + BUG_ON(!evtchn_to_irq); for (i = 0; i < NR_EVENT_CHANNELS; i++) evtchn_to_irq[i] = -1; @@ -1693,6 +1717,6 @@ void __init xen_init_IRQ(void) } else { irq_ctx_init(smp_processor_id()); if (xen_initial_domain()) - xen_setup_pirqs(); + pci_xen_initial_domain(); } } diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c index f914b26..880798a 100644 --- a/drivers/xen/gntdev.c +++ b/drivers/xen/gntdev.c @@ -83,6 +83,7 @@ struct grant_map { struct ioctl_gntdev_grant_ref *grants; struct gnttab_map_grant_ref *map_ops; struct gnttab_unmap_grant_ref *unmap_ops; + struct gnttab_map_grant_ref *kmap_ops; struct page **pages; }; @@ -116,19 +117,22 @@ static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count) add->grants = kzalloc(sizeof(add->grants[0]) * count, GFP_KERNEL); add->map_ops = kzalloc(sizeof(add->map_ops[0]) * count, GFP_KERNEL); add->unmap_ops = kzalloc(sizeof(add->unmap_ops[0]) * count, GFP_KERNEL); + add->kmap_ops = kzalloc(sizeof(add->kmap_ops[0]) * count, GFP_KERNEL); add->pages = kzalloc(sizeof(add->pages[0]) * count, GFP_KERNEL); if (NULL == add->grants || NULL == add->map_ops || NULL == add->unmap_ops || + NULL == add->kmap_ops || NULL == add->pages) goto err; - if (alloc_xenballooned_pages(count, add->pages)) + if (alloc_xenballooned_pages(count, add->pages, false /* lowmem */)) goto err; for (i = 0; i < count; i++) { add->map_ops[i].handle = -1; add->unmap_ops[i].handle = -1; + add->kmap_ops[i].handle = -1; } add->index = 0; @@ -142,6 +146,7 @@ err: kfree(add->grants); kfree(add->map_ops); kfree(add->unmap_ops); + kfree(add->kmap_ops); kfree(add); return NULL; } @@ -243,10 +248,35 @@ static int map_grant_pages(struct grant_map *map) gnttab_set_unmap_op(&map->unmap_ops[i], addr, map->flags, -1 /* handle */); } + } else { + /* + * Setup the map_ops corresponding to the pte entries pointing + * to the kernel linear addresses of the struct pages. + * These ptes are completely different from the user ptes dealt + * with find_grant_ptes. + */ + for (i = 0; i < map->count; i++) { + unsigned level; + unsigned long address = (unsigned long) + pfn_to_kaddr(page_to_pfn(map->pages[i])); + pte_t *ptep; + u64 pte_maddr = 0; + BUG_ON(PageHighMem(map->pages[i])); + + ptep = lookup_address(address, &level); + pte_maddr = arbitrary_virt_to_machine(ptep).maddr; + gnttab_set_map_op(&map->kmap_ops[i], pte_maddr, + map->flags | + GNTMAP_host_map | + GNTMAP_contains_pte, + map->grants[i].ref, + map->grants[i].domid); + } } pr_debug("map %d+%d\n", map->index, map->count); - err = gnttab_map_refs(map->map_ops, map->pages, map->count); + err = gnttab_map_refs(map->map_ops, use_ptemod ? map->kmap_ops : NULL, + map->pages, map->count); if (err) return err; @@ -462,13 +492,11 @@ static int gntdev_release(struct inode *inode, struct file *flip) pr_debug("priv %p\n", priv); - spin_lock(&priv->lock); while (!list_empty(&priv->maps)) { map = list_entry(priv->maps.next, struct grant_map, next); list_del(&map->next); gntdev_put_map(map); } - spin_unlock(&priv->lock); if (use_ptemod) mmu_notifier_unregister(&priv->mn, priv->mm); @@ -532,10 +560,11 @@ static long gntdev_ioctl_unmap_grant_ref(struct gntdev_priv *priv, map = gntdev_find_map_index(priv, op.index >> PAGE_SHIFT, op.count); if (map) { list_del(&map->next); - gntdev_put_map(map); err = 0; } spin_unlock(&priv->lock); + if (map) + gntdev_put_map(map); return err; } diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c index fd725cd..8c71ab80 100644 --- a/drivers/xen/grant-table.c +++ b/drivers/xen/grant-table.c @@ -82,7 +82,7 @@ static inline grant_ref_t *__gnttab_entry(grant_ref_t entry) static int get_free_entries(unsigned count) { unsigned long flags; - int ref, rc; + int ref, rc = 0; grant_ref_t head; spin_lock_irqsave(&gnttab_list_lock, flags); @@ -448,7 +448,8 @@ unsigned int gnttab_max_grant_frames(void) EXPORT_SYMBOL_GPL(gnttab_max_grant_frames); int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops, - struct page **pages, unsigned int count) + struct gnttab_map_grant_ref *kmap_ops, + struct page **pages, unsigned int count) { int i, ret; pte_t *pte; @@ -488,8 +489,7 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops, */ return -EOPNOTSUPP; } - ret = m2p_add_override(mfn, pages[i], - map_ops[i].flags & GNTMAP_contains_pte); + ret = m2p_add_override(mfn, pages[i], &kmap_ops[i]); if (ret) return ret; } diff --git a/drivers/xen/tmem.c b/drivers/xen/tmem.c index 816a449..d369965 100644 --- a/drivers/xen/tmem.c +++ b/drivers/xen/tmem.c @@ -1,7 +1,7 @@ /* * Xen implementation for transcendent memory (tmem) * - * Copyright (C) 2009-2010 Oracle Corp. All rights reserved. + * Copyright (C) 2009-2011 Oracle Corp. All rights reserved. * Author: Dan Magenheimer */ @@ -9,8 +9,14 @@ #include <linux/types.h> #include <linux/init.h> #include <linux/pagemap.h> +#include <linux/module.h> #include <linux/cleancache.h> +/* temporary ifdef until include/linux/frontswap.h is upstream */ +#ifdef CONFIG_FRONTSWAP +#include <linux/frontswap.h> +#endif + #include <xen/xen.h> #include <xen/interface/xen.h> #include <asm/xen/hypercall.h> @@ -122,14 +128,8 @@ static int xen_tmem_flush_object(u32 pool_id, struct tmem_oid oid) return xen_tmem_op(TMEM_FLUSH_OBJECT, pool_id, oid, 0, 0, 0, 0, 0); } -static int xen_tmem_destroy_pool(u32 pool_id) -{ - struct tmem_oid oid = { { 0 } }; - - return xen_tmem_op(TMEM_DESTROY_POOL, pool_id, oid, 0, 0, 0, 0, 0); -} - -int tmem_enabled; +int tmem_enabled __read_mostly; +EXPORT_SYMBOL(tmem_enabled); static int __init enable_tmem(char *s) { @@ -139,6 +139,14 @@ static int __init enable_tmem(char *s) __setup("tmem", enable_tmem); +#ifdef CONFIG_CLEANCACHE +static int xen_tmem_destroy_pool(u32 pool_id) +{ + struct tmem_oid oid = { { 0 } }; + + return xen_tmem_op(TMEM_DESTROY_POOL, pool_id, oid, 0, 0, 0, 0, 0); +} + /* cleancache ops */ static void tmem_cleancache_put_page(int pool, struct cleancache_filekey key, @@ -240,18 +248,156 @@ static struct cleancache_ops tmem_cleancache_ops = { .init_shared_fs = tmem_cleancache_init_shared_fs, .init_fs = tmem_cleancache_init_fs }; +#endif -static int __init xen_tmem_init(void) +#ifdef CONFIG_FRONTSWAP +/* frontswap tmem operations */ + +/* a single tmem poolid is used for all frontswap "types" (swapfiles) */ +static int tmem_frontswap_poolid; + +/* + * Swizzling increases objects per swaptype, increasing tmem concurrency + * for heavy swaploads. Later, larger nr_cpus -> larger SWIZ_BITS + */ +#define SWIZ_BITS 4 +#define SWIZ_MASK ((1 << SWIZ_BITS) - 1) +#define _oswiz(_type, _ind) ((_type << SWIZ_BITS) | (_ind & SWIZ_MASK)) +#define iswiz(_ind) (_ind >> SWIZ_BITS) + +static inline struct tmem_oid oswiz(unsigned type, u32 ind) { - struct cleancache_ops old_ops; + struct tmem_oid oid = { .oid = { 0 } }; + oid.oid[0] = _oswiz(type, ind); + return oid; +} +/* returns 0 if the page was successfully put into frontswap, -1 if not */ +static int tmem_frontswap_put_page(unsigned type, pgoff_t offset, + struct page *page) +{ + u64 ind64 = (u64)offset; + u32 ind = (u32)offset; + unsigned long pfn = page_to_pfn(page); + int pool = tmem_frontswap_poolid; + int ret; + + if (pool < 0) + return -1; + if (ind64 != ind) + return -1; + mb(); /* ensure page is quiescent; tmem may address it with an alias */ + ret = xen_tmem_put_page(pool, oswiz(type, ind), iswiz(ind), pfn); + /* translate Xen tmem return values to linux semantics */ + if (ret == 1) + return 0; + else + return -1; +} + +/* + * returns 0 if the page was successfully gotten from frontswap, -1 if + * was not present (should never happen!) + */ +static int tmem_frontswap_get_page(unsigned type, pgoff_t offset, + struct page *page) +{ + u64 ind64 = (u64)offset; + u32 ind = (u32)offset; + unsigned long pfn = page_to_pfn(page); + int pool = tmem_frontswap_poolid; + int ret; + + if (pool < 0) + return -1; + if (ind64 != ind) + return -1; + ret = xen_tmem_get_page(pool, oswiz(type, ind), iswiz(ind), pfn); + /* translate Xen tmem return values to linux semantics */ + if (ret == 1) + return 0; + else + return -1; +} + +/* flush a single page from frontswap */ +static void tmem_frontswap_flush_page(unsigned type, pgoff_t offset) +{ + u64 ind64 = (u64)offset; + u32 ind = (u32)offset; + int pool = tmem_frontswap_poolid; + + if (pool < 0) + return; + if (ind64 != ind) + return; + (void) xen_tmem_flush_page(pool, oswiz(type, ind), iswiz(ind)); +} + +/* flush all pages from the passed swaptype */ +static void tmem_frontswap_flush_area(unsigned type) +{ + int pool = tmem_frontswap_poolid; + int ind; + + if (pool < 0) + return; + for (ind = SWIZ_MASK; ind >= 0; ind--) + (void)xen_tmem_flush_object(pool, oswiz(type, ind)); +} + +static void tmem_frontswap_init(unsigned ignored) +{ + struct tmem_pool_uuid private = TMEM_POOL_PRIVATE_UUID; + + /* a single tmem poolid is used for all frontswap "types" (swapfiles) */ + if (tmem_frontswap_poolid < 0) + tmem_frontswap_poolid = + xen_tmem_new_pool(private, TMEM_POOL_PERSIST, PAGE_SIZE); +} + +static int __initdata use_frontswap = 1; + +static int __init no_frontswap(char *s) +{ + use_frontswap = 0; + return 1; +} + +__setup("nofrontswap", no_frontswap); + +static struct frontswap_ops tmem_frontswap_ops = { + .put_page = tmem_frontswap_put_page, + .get_page = tmem_frontswap_get_page, + .flush_page = tmem_frontswap_flush_page, + .flush_area = tmem_frontswap_flush_area, + .init = tmem_frontswap_init +}; +#endif + +static int __init xen_tmem_init(void) +{ if (!xen_domain()) return 0; +#ifdef CONFIG_FRONTSWAP + if (tmem_enabled && use_frontswap) { + char *s = ""; + struct frontswap_ops old_ops = + frontswap_register_ops(&tmem_frontswap_ops); + + tmem_frontswap_poolid = -1; + if (old_ops.init != NULL) + s = " (WARNING: frontswap_ops overridden)"; + printk(KERN_INFO "frontswap enabled, RAM provided by " + "Xen Transcendent Memory\n"); + } +#endif #ifdef CONFIG_CLEANCACHE BUG_ON(sizeof(struct cleancache_filekey) != sizeof(struct tmem_oid)); if (tmem_enabled && use_cleancache) { char *s = ""; - old_ops = cleancache_register_ops(&tmem_cleancache_ops); + struct cleancache_ops old_ops = + cleancache_register_ops(&tmem_cleancache_ops); if (old_ops.init_fs != NULL) s = " (WARNING: cleancache_ops overridden)"; printk(KERN_INFO "cleancache enabled, RAM provided by " diff --git a/drivers/xen/xen-balloon.c b/drivers/xen/xen-balloon.c index a4ff225..5c9dc43 100644 --- a/drivers/xen/xen-balloon.c +++ b/drivers/xen/xen-balloon.c @@ -98,6 +98,8 @@ static int __init balloon_init(void) register_balloon(&balloon_sysdev); + register_xen_selfballooning(&balloon_sysdev); + target_watch.callback = watch_target; xenstore_notifier.notifier_call = balloon_init_watcher; diff --git a/drivers/xen/xen-pciback/Makefile b/drivers/xen/xen-pciback/Makefile new file mode 100644 index 0000000..ffe0ad3 --- /dev/null +++ b/drivers/xen/xen-pciback/Makefile @@ -0,0 +1,7 @@ +obj-$(CONFIG_XEN_PCIDEV_BACKEND) += xen-pciback.o + +xen-pciback-y := pci_stub.o pciback_ops.o xenbus.o +xen-pciback-y += conf_space.o conf_space_header.o \ + conf_space_capability.o \ + conf_space_quirks.o vpci.o \ + passthrough.o diff --git a/drivers/xen/xen-pciback/conf_space.c b/drivers/xen/xen-pciback/conf_space.c new file mode 100644 index 0000000..444345a --- /dev/null +++ b/drivers/xen/xen-pciback/conf_space.c @@ -0,0 +1,437 @@ +/* + * PCI Backend - Functions for creating a virtual configuration space for + * exported PCI Devices. + * It's dangerous to allow PCI Driver Domains to change their + * device's resources (memory, i/o ports, interrupts). We need to + * restrict changes to certain PCI Configuration registers: + * BARs, INTERRUPT_PIN, most registers in the header... + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ + +#include <linux/kernel.h> +#include <linux/pci.h> +#include "pciback.h" +#include "conf_space.h" +#include "conf_space_quirks.h" + +static int permissive; +module_param(permissive, bool, 0644); + +/* This is where xen_pcibk_read_config_byte, xen_pcibk_read_config_word, + * xen_pcibk_write_config_word, and xen_pcibk_write_config_byte are created. */ +#define DEFINE_PCI_CONFIG(op, size, type) \ +int xen_pcibk_##op##_config_##size \ +(struct pci_dev *dev, int offset, type value, void *data) \ +{ \ + return pci_##op##_config_##size(dev, offset, value); \ +} + +DEFINE_PCI_CONFIG(read, byte, u8 *) +DEFINE_PCI_CONFIG(read, word, u16 *) +DEFINE_PCI_CONFIG(read, dword, u32 *) + +DEFINE_PCI_CONFIG(write, byte, u8) +DEFINE_PCI_CONFIG(write, word, u16) +DEFINE_PCI_CONFIG(write, dword, u32) + +static int conf_space_read(struct pci_dev *dev, + const struct config_field_entry *entry, + int offset, u32 *value) +{ + int ret = 0; + const struct config_field *field = entry->field; + + *value = 0; + + switch (field->size) { + case 1: + if (field->u.b.read) + ret = field->u.b.read(dev, offset, (u8 *) value, + entry->data); + break; + case 2: + if (field->u.w.read) + ret = field->u.w.read(dev, offset, (u16 *) value, + entry->data); + break; + case 4: + if (field->u.dw.read) + ret = field->u.dw.read(dev, offset, value, entry->data); + break; + } + return ret; +} + +static int conf_space_write(struct pci_dev *dev, + const struct config_field_entry *entry, + int offset, u32 value) +{ + int ret = 0; + const struct config_field *field = entry->field; + + switch (field->size) { + case 1: + if (field->u.b.write) + ret = field->u.b.write(dev, offset, (u8) value, + entry->data); + break; + case 2: + if (field->u.w.write) + ret = field->u.w.write(dev, offset, (u16) value, + entry->data); + break; + case 4: + if (field->u.dw.write) + ret = field->u.dw.write(dev, offset, value, + entry->data); + break; + } + return ret; +} + +static inline u32 get_mask(int size) +{ + if (size == 1) + return 0xff; + else if (size == 2) + return 0xffff; + else + return 0xffffffff; +} + +static inline int valid_request(int offset, int size) +{ + /* Validate request (no un-aligned requests) */ + if ((size == 1 || size == 2 || size == 4) && (offset % size) == 0) + return 1; + return 0; +} + +static inline u32 merge_value(u32 val, u32 new_val, u32 new_val_mask, + int offset) +{ + if (offset >= 0) { + new_val_mask <<= (offset * 8); + new_val <<= (offset * 8); + } else { + new_val_mask >>= (offset * -8); + new_val >>= (offset * -8); + } + val = (val & ~new_val_mask) | (new_val & new_val_mask); + + return val; +} + +static int pcibios_err_to_errno(int err) +{ + switch (err) { + case PCIBIOS_SUCCESSFUL: + return XEN_PCI_ERR_success; + case PCIBIOS_DEVICE_NOT_FOUND: + return XEN_PCI_ERR_dev_not_found; + case PCIBIOS_BAD_REGISTER_NUMBER: + return XEN_PCI_ERR_invalid_offset; + case PCIBIOS_FUNC_NOT_SUPPORTED: + return XEN_PCI_ERR_not_implemented; + case PCIBIOS_SET_FAILED: + return XEN_PCI_ERR_access_denied; + } + return err; +} + +int xen_pcibk_config_read(struct pci_dev *dev, int offset, int size, + u32 *ret_val) +{ + int err = 0; + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + const struct config_field_entry *cfg_entry; + const struct config_field *field; + int req_start, req_end, field_start, field_end; + /* if read fails for any reason, return 0 + * (as if device didn't respond) */ + u32 value = 0, tmp_val; + + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: read %d bytes at 0x%x\n", + pci_name(dev), size, offset); + + if (!valid_request(offset, size)) { + err = XEN_PCI_ERR_invalid_offset; + goto out; + } + + /* Get the real value first, then modify as appropriate */ + switch (size) { + case 1: + err = pci_read_config_byte(dev, offset, (u8 *) &value); + break; + case 2: + err = pci_read_config_word(dev, offset, (u16 *) &value); + break; + case 4: + err = pci_read_config_dword(dev, offset, &value); + break; + } + + list_for_each_entry(cfg_entry, &dev_data->config_fields, list) { + field = cfg_entry->field; + + req_start = offset; + req_end = offset + size; + field_start = OFFSET(cfg_entry); + field_end = OFFSET(cfg_entry) + field->size; + + if ((req_start >= field_start && req_start < field_end) + || (req_end > field_start && req_end <= field_end)) { + err = conf_space_read(dev, cfg_entry, field_start, + &tmp_val); + if (err) + goto out; + + value = merge_value(value, tmp_val, + get_mask(field->size), + field_start - req_start); + } + } + +out: + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: read %d bytes at 0x%x = %x\n", + pci_name(dev), size, offset, value); + + *ret_val = value; + return pcibios_err_to_errno(err); +} + +int xen_pcibk_config_write(struct pci_dev *dev, int offset, int size, u32 value) +{ + int err = 0, handled = 0; + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + const struct config_field_entry *cfg_entry; + const struct config_field *field; + u32 tmp_val; + int req_start, req_end, field_start, field_end; + + if (unlikely(verbose_request)) + printk(KERN_DEBUG + DRV_NAME ": %s: write request %d bytes at 0x%x = %x\n", + pci_name(dev), size, offset, value); + + if (!valid_request(offset, size)) + return XEN_PCI_ERR_invalid_offset; + + list_for_each_entry(cfg_entry, &dev_data->config_fields, list) { + field = cfg_entry->field; + + req_start = offset; + req_end = offset + size; + field_start = OFFSET(cfg_entry); + field_end = OFFSET(cfg_entry) + field->size; + + if ((req_start >= field_start && req_start < field_end) + || (req_end > field_start && req_end <= field_end)) { + tmp_val = 0; + + err = xen_pcibk_config_read(dev, field_start, + field->size, &tmp_val); + if (err) + break; + + tmp_val = merge_value(tmp_val, value, get_mask(size), + req_start - field_start); + + err = conf_space_write(dev, cfg_entry, field_start, + tmp_val); + + /* handled is set true here, but not every byte + * may have been written! Properly detecting if + * every byte is handled is unnecessary as the + * flag is used to detect devices that need + * special helpers to work correctly. + */ + handled = 1; + } + } + + if (!handled && !err) { + /* By default, anything not specificially handled above is + * read-only. The permissive flag changes this behavior so + * that anything not specifically handled above is writable. + * This means that some fields may still be read-only because + * they have entries in the config_field list that intercept + * the write and do nothing. */ + if (dev_data->permissive || permissive) { + switch (size) { + case 1: + err = pci_write_config_byte(dev, offset, + (u8) value); + break; + case 2: + err = pci_write_config_word(dev, offset, + (u16) value); + break; + case 4: + err = pci_write_config_dword(dev, offset, + (u32) value); + break; + } + } else if (!dev_data->warned_on_write) { + dev_data->warned_on_write = 1; + dev_warn(&dev->dev, "Driver tried to write to a " + "read-only configuration space field at offset" + " 0x%x, size %d. This may be harmless, but if " + "you have problems with your device:\n" + "1) see permissive attribute in sysfs\n" + "2) report problems to the xen-devel " + "mailing list along with details of your " + "device obtained from lspci.\n", offset, size); + } + } + + return pcibios_err_to_errno(err); +} + +void xen_pcibk_config_free_dyn_fields(struct pci_dev *dev) +{ + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + struct config_field_entry *cfg_entry, *t; + const struct config_field *field; + + dev_dbg(&dev->dev, "free-ing dynamically allocated virtual " + "configuration space fields\n"); + if (!dev_data) + return; + + list_for_each_entry_safe(cfg_entry, t, &dev_data->config_fields, list) { + field = cfg_entry->field; + + if (field->clean) { + field->clean((struct config_field *)field); + + kfree(cfg_entry->data); + + list_del(&cfg_entry->list); + kfree(cfg_entry); + } + + } +} + +void xen_pcibk_config_reset_dev(struct pci_dev *dev) +{ + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + const struct config_field_entry *cfg_entry; + const struct config_field *field; + + dev_dbg(&dev->dev, "resetting virtual configuration space\n"); + if (!dev_data) + return; + + list_for_each_entry(cfg_entry, &dev_data->config_fields, list) { + field = cfg_entry->field; + + if (field->reset) + field->reset(dev, OFFSET(cfg_entry), cfg_entry->data); + } +} + +void xen_pcibk_config_free_dev(struct pci_dev *dev) +{ + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + struct config_field_entry *cfg_entry, *t; + const struct config_field *field; + + dev_dbg(&dev->dev, "free-ing virtual configuration space fields\n"); + if (!dev_data) + return; + + list_for_each_entry_safe(cfg_entry, t, &dev_data->config_fields, list) { + list_del(&cfg_entry->list); + + field = cfg_entry->field; + + if (field->release) + field->release(dev, OFFSET(cfg_entry), cfg_entry->data); + + kfree(cfg_entry); + } +} + +int xen_pcibk_config_add_field_offset(struct pci_dev *dev, + const struct config_field *field, + unsigned int base_offset) +{ + int err = 0; + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + struct config_field_entry *cfg_entry; + void *tmp; + + cfg_entry = kmalloc(sizeof(*cfg_entry), GFP_KERNEL); + if (!cfg_entry) { + err = -ENOMEM; + goto out; + } + + cfg_entry->data = NULL; + cfg_entry->field = field; + cfg_entry->base_offset = base_offset; + + /* silently ignore duplicate fields */ + err = xen_pcibk_field_is_dup(dev, OFFSET(cfg_entry)); + if (err) + goto out; + + if (field->init) { + tmp = field->init(dev, OFFSET(cfg_entry)); + + if (IS_ERR(tmp)) { + err = PTR_ERR(tmp); + goto out; + } + + cfg_entry->data = tmp; + } + + dev_dbg(&dev->dev, "added config field at offset 0x%02x\n", + OFFSET(cfg_entry)); + list_add_tail(&cfg_entry->list, &dev_data->config_fields); + +out: + if (err) + kfree(cfg_entry); + + return err; +} + +/* This sets up the device's virtual configuration space to keep track of + * certain registers (like the base address registers (BARs) so that we can + * keep the client from manipulating them directly. + */ +int xen_pcibk_config_init_dev(struct pci_dev *dev) +{ + int err = 0; + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + + dev_dbg(&dev->dev, "initializing virtual configuration space\n"); + + INIT_LIST_HEAD(&dev_data->config_fields); + + err = xen_pcibk_config_header_add_fields(dev); + if (err) + goto out; + + err = xen_pcibk_config_capability_add_fields(dev); + if (err) + goto out; + + err = xen_pcibk_config_quirks_init(dev); + +out: + return err; +} + +int xen_pcibk_config_init(void) +{ + return xen_pcibk_config_capability_init(); +} diff --git a/drivers/xen/xen-pciback/conf_space.h b/drivers/xen/xen-pciback/conf_space.h new file mode 100644 index 0000000..e56c934 --- /dev/null +++ b/drivers/xen/xen-pciback/conf_space.h @@ -0,0 +1,126 @@ +/* + * PCI Backend - Common data structures for overriding the configuration space + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ + +#ifndef __XEN_PCIBACK_CONF_SPACE_H__ +#define __XEN_PCIBACK_CONF_SPACE_H__ + +#include <linux/list.h> +#include <linux/err.h> + +/* conf_field_init can return an errno in a ptr with ERR_PTR() */ +typedef void *(*conf_field_init) (struct pci_dev *dev, int offset); +typedef void (*conf_field_reset) (struct pci_dev *dev, int offset, void *data); +typedef void (*conf_field_free) (struct pci_dev *dev, int offset, void *data); + +typedef int (*conf_dword_write) (struct pci_dev *dev, int offset, u32 value, + void *data); +typedef int (*conf_word_write) (struct pci_dev *dev, int offset, u16 value, + void *data); +typedef int (*conf_byte_write) (struct pci_dev *dev, int offset, u8 value, + void *data); +typedef int (*conf_dword_read) (struct pci_dev *dev, int offset, u32 *value, + void *data); +typedef int (*conf_word_read) (struct pci_dev *dev, int offset, u16 *value, + void *data); +typedef int (*conf_byte_read) (struct pci_dev *dev, int offset, u8 *value, + void *data); + +/* These are the fields within the configuration space which we + * are interested in intercepting reads/writes to and changing their + * values. + */ +struct config_field { + unsigned int offset; + unsigned int size; + unsigned int mask; + conf_field_init init; + conf_field_reset reset; + conf_field_free release; + void (*clean) (struct config_field *field); + union { + struct { + conf_dword_write write; + conf_dword_read read; + } dw; + struct { + conf_word_write write; + conf_word_read read; + } w; + struct { + conf_byte_write write; + conf_byte_read read; + } b; + } u; + struct list_head list; +}; + +struct config_field_entry { + struct list_head list; + const struct config_field *field; + unsigned int base_offset; + void *data; +}; + +#define OFFSET(cfg_entry) ((cfg_entry)->base_offset+(cfg_entry)->field->offset) + +/* Add fields to a device - the add_fields macro expects to get a pointer to + * the first entry in an array (of which the ending is marked by size==0) + */ +int xen_pcibk_config_add_field_offset(struct pci_dev *dev, + const struct config_field *field, + unsigned int offset); + +static inline int xen_pcibk_config_add_field(struct pci_dev *dev, + const struct config_field *field) +{ + return xen_pcibk_config_add_field_offset(dev, field, 0); +} + +static inline int xen_pcibk_config_add_fields(struct pci_dev *dev, + const struct config_field *field) +{ + int i, err = 0; + for (i = 0; field[i].size != 0; i++) { + err = xen_pcibk_config_add_field(dev, &field[i]); + if (err) + break; + } + return err; +} + +static inline int xen_pcibk_config_add_fields_offset(struct pci_dev *dev, + const struct config_field *field, + unsigned int offset) +{ + int i, err = 0; + for (i = 0; field[i].size != 0; i++) { + err = xen_pcibk_config_add_field_offset(dev, &field[i], offset); + if (err) + break; + } + return err; +} + +/* Read/Write the real configuration space */ +int xen_pcibk_read_config_byte(struct pci_dev *dev, int offset, u8 *value, + void *data); +int xen_pcibk_read_config_word(struct pci_dev *dev, int offset, u16 *value, + void *data); +int xen_pcibk_read_config_dword(struct pci_dev *dev, int offset, u32 *value, + void *data); +int xen_pcibk_write_config_byte(struct pci_dev *dev, int offset, u8 value, + void *data); +int xen_pcibk_write_config_word(struct pci_dev *dev, int offset, u16 value, + void *data); +int xen_pcibk_write_config_dword(struct pci_dev *dev, int offset, u32 value, + void *data); + +int xen_pcibk_config_capability_init(void); + +int xen_pcibk_config_header_add_fields(struct pci_dev *dev); +int xen_pcibk_config_capability_add_fields(struct pci_dev *dev); + +#endif /* __XEN_PCIBACK_CONF_SPACE_H__ */ diff --git a/drivers/xen/xen-pciback/conf_space_capability.c b/drivers/xen/xen-pciback/conf_space_capability.c new file mode 100644 index 0000000..7f83e90 --- /dev/null +++ b/drivers/xen/xen-pciback/conf_space_capability.c @@ -0,0 +1,207 @@ +/* + * PCI Backend - Handles the virtual fields found on the capability lists + * in the configuration space. + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ + +#include <linux/kernel.h> +#include <linux/pci.h> +#include "pciback.h" +#include "conf_space.h" + +static LIST_HEAD(capabilities); +struct xen_pcibk_config_capability { + struct list_head cap_list; + + int capability; + + /* If the device has the capability found above, add these fields */ + const struct config_field *fields; +}; + +static const struct config_field caplist_header[] = { + { + .offset = PCI_CAP_LIST_ID, + .size = 2, /* encompass PCI_CAP_LIST_ID & PCI_CAP_LIST_NEXT */ + .u.w.read = xen_pcibk_read_config_word, + .u.w.write = NULL, + }, + {} +}; + +static inline void register_capability(struct xen_pcibk_config_capability *cap) +{ + list_add_tail(&cap->cap_list, &capabilities); +} + +int xen_pcibk_config_capability_add_fields(struct pci_dev *dev) +{ + int err = 0; + struct xen_pcibk_config_capability *cap; + int cap_offset; + + list_for_each_entry(cap, &capabilities, cap_list) { + cap_offset = pci_find_capability(dev, cap->capability); + if (cap_offset) { + dev_dbg(&dev->dev, "Found capability 0x%x at 0x%x\n", + cap->capability, cap_offset); + + err = xen_pcibk_config_add_fields_offset(dev, + caplist_header, + cap_offset); + if (err) + goto out; + err = xen_pcibk_config_add_fields_offset(dev, + cap->fields, + cap_offset); + if (err) + goto out; + } + } + +out: + return err; +} + +static int vpd_address_write(struct pci_dev *dev, int offset, u16 value, + void *data) +{ + /* Disallow writes to the vital product data */ + if (value & PCI_VPD_ADDR_F) + return PCIBIOS_SET_FAILED; + else + return pci_write_config_word(dev, offset, value); +} + +static const struct config_field caplist_vpd[] = { + { + .offset = PCI_VPD_ADDR, + .size = 2, + .u.w.read = xen_pcibk_read_config_word, + .u.w.write = vpd_address_write, + }, + { + .offset = PCI_VPD_DATA, + .size = 4, + .u.dw.read = xen_pcibk_read_config_dword, + .u.dw.write = NULL, + }, + {} +}; + +static int pm_caps_read(struct pci_dev *dev, int offset, u16 *value, + void *data) +{ + int err; + u16 real_value; + + err = pci_read_config_word(dev, offset, &real_value); + if (err) + goto out; + + *value = real_value & ~PCI_PM_CAP_PME_MASK; + +out: + return err; +} + +/* PM_OK_BITS specifies the bits that the driver domain is allowed to change. + * Can't allow driver domain to enable PMEs - they're shared */ +#define PM_OK_BITS (PCI_PM_CTRL_PME_STATUS|PCI_PM_CTRL_DATA_SEL_MASK) + +static int pm_ctrl_write(struct pci_dev *dev, int offset, u16 new_value, + void *data) +{ + int err; + u16 old_value; + pci_power_t new_state, old_state; + + err = pci_read_config_word(dev, offset, &old_value); + if (err) + goto out; + + old_state = (pci_power_t)(old_value & PCI_PM_CTRL_STATE_MASK); + new_state = (pci_power_t)(new_value & PCI_PM_CTRL_STATE_MASK); + + new_value &= PM_OK_BITS; + if ((old_value & PM_OK_BITS) != new_value) { + new_value = (old_value & ~PM_OK_BITS) | new_value; + err = pci_write_config_word(dev, offset, new_value); + if (err) + goto out; + } + + /* Let pci core handle the power management change */ + dev_dbg(&dev->dev, "set power state to %x\n", new_state); + err = pci_set_power_state(dev, new_state); + if (err) { + err = PCIBIOS_SET_FAILED; + goto out; + } + + out: + return err; +} + +/* Ensure PMEs are disabled */ +static void *pm_ctrl_init(struct pci_dev *dev, int offset) +{ + int err; + u16 value; + + err = pci_read_config_word(dev, offset, &value); + if (err) + goto out; + + if (value & PCI_PM_CTRL_PME_ENABLE) { + value &= ~PCI_PM_CTRL_PME_ENABLE; + err = pci_write_config_word(dev, offset, value); + } + +out: + return ERR_PTR(err); +} + +static const struct config_field caplist_pm[] = { + { + .offset = PCI_PM_PMC, + .size = 2, + .u.w.read = pm_caps_read, + }, + { + .offset = PCI_PM_CTRL, + .size = 2, + .init = pm_ctrl_init, + .u.w.read = xen_pcibk_read_config_word, + .u.w.write = pm_ctrl_write, + }, + { + .offset = PCI_PM_PPB_EXTENSIONS, + .size = 1, + .u.b.read = xen_pcibk_read_config_byte, + }, + { + .offset = PCI_PM_DATA_REGISTER, + .size = 1, + .u.b.read = xen_pcibk_read_config_byte, + }, + {} +}; + +static struct xen_pcibk_config_capability xen_pcibk_config_capability_pm = { + .capability = PCI_CAP_ID_PM, + .fields = caplist_pm, +}; +static struct xen_pcibk_config_capability xen_pcibk_config_capability_vpd = { + .capability = PCI_CAP_ID_VPD, + .fields = caplist_vpd, +}; + +int xen_pcibk_config_capability_init(void) +{ + register_capability(&xen_pcibk_config_capability_vpd); + register_capability(&xen_pcibk_config_capability_pm); + + return 0; +} diff --git a/drivers/xen/xen-pciback/conf_space_header.c b/drivers/xen/xen-pciback/conf_space_header.c new file mode 100644 index 0000000..3daf862 --- /dev/null +++ b/drivers/xen/xen-pciback/conf_space_header.c @@ -0,0 +1,385 @@ +/* + * PCI Backend - Handles the virtual fields in the configuration space headers. + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ + +#include <linux/kernel.h> +#include <linux/pci.h> +#include "pciback.h" +#include "conf_space.h" + +struct pci_bar_info { + u32 val; + u32 len_val; + int which; +}; + +#define is_enable_cmd(value) ((value)&(PCI_COMMAND_MEMORY|PCI_COMMAND_IO)) +#define is_master_cmd(value) ((value)&PCI_COMMAND_MASTER) + +static int command_read(struct pci_dev *dev, int offset, u16 *value, void *data) +{ + int i; + int ret; + + ret = xen_pcibk_read_config_word(dev, offset, value, data); + if (!pci_is_enabled(dev)) + return ret; + + for (i = 0; i < PCI_ROM_RESOURCE; i++) { + if (dev->resource[i].flags & IORESOURCE_IO) + *value |= PCI_COMMAND_IO; + if (dev->resource[i].flags & IORESOURCE_MEM) + *value |= PCI_COMMAND_MEMORY; + } + + return ret; +} + +static int command_write(struct pci_dev *dev, int offset, u16 value, void *data) +{ + struct xen_pcibk_dev_data *dev_data; + int err; + + dev_data = pci_get_drvdata(dev); + if (!pci_is_enabled(dev) && is_enable_cmd(value)) { + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: enable\n", + pci_name(dev)); + err = pci_enable_device(dev); + if (err) + return err; + if (dev_data) + dev_data->enable_intx = 1; + } else if (pci_is_enabled(dev) && !is_enable_cmd(value)) { + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: disable\n", + pci_name(dev)); + pci_disable_device(dev); + if (dev_data) + dev_data->enable_intx = 0; + } + + if (!dev->is_busmaster && is_master_cmd(value)) { + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: set bus master\n", + pci_name(dev)); + pci_set_master(dev); + } + + if (value & PCI_COMMAND_INVALIDATE) { + if (unlikely(verbose_request)) + printk(KERN_DEBUG + DRV_NAME ": %s: enable memory-write-invalidate\n", + pci_name(dev)); + err = pci_set_mwi(dev); + if (err) { + printk(KERN_WARNING + DRV_NAME ": %s: cannot enable " + "memory-write-invalidate (%d)\n", + pci_name(dev), err); + value &= ~PCI_COMMAND_INVALIDATE; + } + } + + return pci_write_config_word(dev, offset, value); +} + +static int rom_write(struct pci_dev *dev, int offset, u32 value, void *data) +{ + struct pci_bar_info *bar = data; + + if (unlikely(!bar)) { + printk(KERN_WARNING DRV_NAME ": driver data not found for %s\n", + pci_name(dev)); + return XEN_PCI_ERR_op_failed; + } + + /* A write to obtain the length must happen as a 32-bit write. + * This does not (yet) support writing individual bytes + */ + if (value == ~PCI_ROM_ADDRESS_ENABLE) + bar->which = 1; + else { + u32 tmpval; + pci_read_config_dword(dev, offset, &tmpval); + if (tmpval != bar->val && value == bar->val) { + /* Allow restoration of bar value. */ + pci_write_config_dword(dev, offset, bar->val); + } + bar->which = 0; + } + + /* Do we need to support enabling/disabling the rom address here? */ + + return 0; +} + +/* For the BARs, only allow writes which write ~0 or + * the correct resource information + * (Needed for when the driver probes the resource usage) + */ +static int bar_write(struct pci_dev *dev, int offset, u32 value, void *data) +{ + struct pci_bar_info *bar = data; + + if (unlikely(!bar)) { + printk(KERN_WARNING DRV_NAME ": driver data not found for %s\n", + pci_name(dev)); + return XEN_PCI_ERR_op_failed; + } + + /* A write to obtain the length must happen as a 32-bit write. + * This does not (yet) support writing individual bytes + */ + if (value == ~0) + bar->which = 1; + else { + u32 tmpval; + pci_read_config_dword(dev, offset, &tmpval); + if (tmpval != bar->val && value == bar->val) { + /* Allow restoration of bar value. */ + pci_write_config_dword(dev, offset, bar->val); + } + bar->which = 0; + } + + return 0; +} + +static int bar_read(struct pci_dev *dev, int offset, u32 * value, void *data) +{ + struct pci_bar_info *bar = data; + + if (unlikely(!bar)) { + printk(KERN_WARNING DRV_NAME ": driver data not found for %s\n", + pci_name(dev)); + return XEN_PCI_ERR_op_failed; + } + + *value = bar->which ? bar->len_val : bar->val; + + return 0; +} + +static inline void read_dev_bar(struct pci_dev *dev, + struct pci_bar_info *bar_info, int offset, + u32 len_mask) +{ + int pos; + struct resource *res = dev->resource; + + if (offset == PCI_ROM_ADDRESS || offset == PCI_ROM_ADDRESS1) + pos = PCI_ROM_RESOURCE; + else { + pos = (offset - PCI_BASE_ADDRESS_0) / 4; + if (pos && ((res[pos - 1].flags & (PCI_BASE_ADDRESS_SPACE | + PCI_BASE_ADDRESS_MEM_TYPE_MASK)) == + (PCI_BASE_ADDRESS_SPACE_MEMORY | + PCI_BASE_ADDRESS_MEM_TYPE_64))) { + bar_info->val = res[pos - 1].start >> 32; + bar_info->len_val = res[pos - 1].end >> 32; + return; + } + } + + bar_info->val = res[pos].start | + (res[pos].flags & PCI_REGION_FLAG_MASK); + bar_info->len_val = resource_size(&res[pos]); +} + +static void *bar_init(struct pci_dev *dev, int offset) +{ + struct pci_bar_info *bar = kmalloc(sizeof(*bar), GFP_KERNEL); + + if (!bar) + return ERR_PTR(-ENOMEM); + + read_dev_bar(dev, bar, offset, ~0); + bar->which = 0; + + return bar; +} + +static void *rom_init(struct pci_dev *dev, int offset) +{ + struct pci_bar_info *bar = kmalloc(sizeof(*bar), GFP_KERNEL); + + if (!bar) + return ERR_PTR(-ENOMEM); + + read_dev_bar(dev, bar, offset, ~PCI_ROM_ADDRESS_ENABLE); + bar->which = 0; + + return bar; +} + +static void bar_reset(struct pci_dev *dev, int offset, void *data) +{ + struct pci_bar_info *bar = data; + + bar->which = 0; +} + +static void bar_release(struct pci_dev *dev, int offset, void *data) +{ + kfree(data); +} + +static int xen_pcibk_read_vendor(struct pci_dev *dev, int offset, + u16 *value, void *data) +{ + *value = dev->vendor; + + return 0; +} + +static int xen_pcibk_read_device(struct pci_dev *dev, int offset, + u16 *value, void *data) +{ + *value = dev->device; + + return 0; +} + +static int interrupt_read(struct pci_dev *dev, int offset, u8 * value, + void *data) +{ + *value = (u8) dev->irq; + + return 0; +} + +static int bist_write(struct pci_dev *dev, int offset, u8 value, void *data) +{ + u8 cur_value; + int err; + + err = pci_read_config_byte(dev, offset, &cur_value); + if (err) + goto out; + + if ((cur_value & ~PCI_BIST_START) == (value & ~PCI_BIST_START) + || value == PCI_BIST_START) + err = pci_write_config_byte(dev, offset, value); + +out: + return err; +} + +static const struct config_field header_common[] = { + { + .offset = PCI_VENDOR_ID, + .size = 2, + .u.w.read = xen_pcibk_read_vendor, + }, + { + .offset = PCI_DEVICE_ID, + .size = 2, + .u.w.read = xen_pcibk_read_device, + }, + { + .offset = PCI_COMMAND, + .size = 2, + .u.w.read = command_read, + .u.w.write = command_write, + }, + { + .offset = PCI_INTERRUPT_LINE, + .size = 1, + .u.b.read = interrupt_read, + }, + { + .offset = PCI_INTERRUPT_PIN, + .size = 1, + .u.b.read = xen_pcibk_read_config_byte, + }, + { + /* Any side effects of letting driver domain control cache line? */ + .offset = PCI_CACHE_LINE_SIZE, + .size = 1, + .u.b.read = xen_pcibk_read_config_byte, + .u.b.write = xen_pcibk_write_config_byte, + }, + { + .offset = PCI_LATENCY_TIMER, + .size = 1, + .u.b.read = xen_pcibk_read_config_byte, + }, + { + .offset = PCI_BIST, + .size = 1, + .u.b.read = xen_pcibk_read_config_byte, + .u.b.write = bist_write, + }, + {} +}; + +#define CFG_FIELD_BAR(reg_offset) \ + { \ + .offset = reg_offset, \ + .size = 4, \ + .init = bar_init, \ + .reset = bar_reset, \ + .release = bar_release, \ + .u.dw.read = bar_read, \ + .u.dw.write = bar_write, \ + } + +#define CFG_FIELD_ROM(reg_offset) \ + { \ + .offset = reg_offset, \ + .size = 4, \ + .init = rom_init, \ + .reset = bar_reset, \ + .release = bar_release, \ + .u.dw.read = bar_read, \ + .u.dw.write = rom_write, \ + } + +static const struct config_field header_0[] = { + CFG_FIELD_BAR(PCI_BASE_ADDRESS_0), + CFG_FIELD_BAR(PCI_BASE_ADDRESS_1), + CFG_FIELD_BAR(PCI_BASE_ADDRESS_2), + CFG_FIELD_BAR(PCI_BASE_ADDRESS_3), + CFG_FIELD_BAR(PCI_BASE_ADDRESS_4), + CFG_FIELD_BAR(PCI_BASE_ADDRESS_5), + CFG_FIELD_ROM(PCI_ROM_ADDRESS), + {} +}; + +static const struct config_field header_1[] = { + CFG_FIELD_BAR(PCI_BASE_ADDRESS_0), + CFG_FIELD_BAR(PCI_BASE_ADDRESS_1), + CFG_FIELD_ROM(PCI_ROM_ADDRESS1), + {} +}; + +int xen_pcibk_config_header_add_fields(struct pci_dev *dev) +{ + int err; + + err = xen_pcibk_config_add_fields(dev, header_common); + if (err) + goto out; + + switch (dev->hdr_type) { + case PCI_HEADER_TYPE_NORMAL: + err = xen_pcibk_config_add_fields(dev, header_0); + break; + + case PCI_HEADER_TYPE_BRIDGE: + err = xen_pcibk_config_add_fields(dev, header_1); + break; + + default: + err = -EINVAL; + printk(KERN_ERR DRV_NAME ": %s: Unsupported header type %d!\n", + pci_name(dev), dev->hdr_type); + break; + } + +out: + return err; +} diff --git a/drivers/xen/xen-pciback/conf_space_quirks.c b/drivers/xen/xen-pciback/conf_space_quirks.c new file mode 100644 index 0000000..7476791 --- /dev/null +++ b/drivers/xen/xen-pciback/conf_space_quirks.c @@ -0,0 +1,139 @@ +/* + * PCI Backend - Handle special overlays for broken devices. + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + * Author: Chris Bookholt <hap10@epoch.ncsc.mil> + */ + +#include <linux/kernel.h> +#include <linux/pci.h> +#include "pciback.h" +#include "conf_space.h" +#include "conf_space_quirks.h" + +LIST_HEAD(xen_pcibk_quirks); +static inline const struct pci_device_id * +match_one_device(const struct pci_device_id *id, const struct pci_dev *dev) +{ + if ((id->vendor == PCI_ANY_ID || id->vendor == dev->vendor) && + (id->device == PCI_ANY_ID || id->device == dev->device) && + (id->subvendor == PCI_ANY_ID || + id->subvendor == dev->subsystem_vendor) && + (id->subdevice == PCI_ANY_ID || + id->subdevice == dev->subsystem_device) && + !((id->class ^ dev->class) & id->class_mask)) + return id; + return NULL; +} + +static struct xen_pcibk_config_quirk *xen_pcibk_find_quirk(struct pci_dev *dev) +{ + struct xen_pcibk_config_quirk *tmp_quirk; + + list_for_each_entry(tmp_quirk, &xen_pcibk_quirks, quirks_list) + if (match_one_device(&tmp_quirk->devid, dev) != NULL) + goto out; + tmp_quirk = NULL; + printk(KERN_DEBUG DRV_NAME + ": quirk didn't match any device known\n"); +out: + return tmp_quirk; +} + +static inline void register_quirk(struct xen_pcibk_config_quirk *quirk) +{ + list_add_tail(&quirk->quirks_list, &xen_pcibk_quirks); +} + +int xen_pcibk_field_is_dup(struct pci_dev *dev, unsigned int reg) +{ + int ret = 0; + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + struct config_field_entry *cfg_entry; + + list_for_each_entry(cfg_entry, &dev_data->config_fields, list) { + if (OFFSET(cfg_entry) == reg) { + ret = 1; + break; + } + } + return ret; +} + +int xen_pcibk_config_quirks_add_field(struct pci_dev *dev, struct config_field + *field) +{ + int err = 0; + + switch (field->size) { + case 1: + field->u.b.read = xen_pcibk_read_config_byte; + field->u.b.write = xen_pcibk_write_config_byte; + break; + case 2: + field->u.w.read = xen_pcibk_read_config_word; + field->u.w.write = xen_pcibk_write_config_word; + break; + case 4: + field->u.dw.read = xen_pcibk_read_config_dword; + field->u.dw.write = xen_pcibk_write_config_dword; + break; + default: + err = -EINVAL; + goto out; + } + + xen_pcibk_config_add_field(dev, field); + +out: + return err; +} + +int xen_pcibk_config_quirks_init(struct pci_dev *dev) +{ + struct xen_pcibk_config_quirk *quirk; + int ret = 0; + + quirk = kzalloc(sizeof(*quirk), GFP_ATOMIC); + if (!quirk) { + ret = -ENOMEM; + goto out; + } + + quirk->devid.vendor = dev->vendor; + quirk->devid.device = dev->device; + quirk->devid.subvendor = dev->subsystem_vendor; + quirk->devid.subdevice = dev->subsystem_device; + quirk->devid.class = 0; + quirk->devid.class_mask = 0; + quirk->devid.driver_data = 0UL; + + quirk->pdev = dev; + + register_quirk(quirk); +out: + return ret; +} + +void xen_pcibk_config_field_free(struct config_field *field) +{ + kfree(field); +} + +int xen_pcibk_config_quirk_release(struct pci_dev *dev) +{ + struct xen_pcibk_config_quirk *quirk; + int ret = 0; + + quirk = xen_pcibk_find_quirk(dev); + if (!quirk) { + ret = -ENXIO; + goto out; + } + + list_del(&quirk->quirks_list); + kfree(quirk); + +out: + return ret; +} diff --git a/drivers/xen/xen-pciback/conf_space_quirks.h b/drivers/xen/xen-pciback/conf_space_quirks.h new file mode 100644 index 0000000..cfcc517 --- /dev/null +++ b/drivers/xen/xen-pciback/conf_space_quirks.h @@ -0,0 +1,33 @@ +/* + * PCI Backend - Data structures for special overlays for broken devices. + * + * Ryan Wilson <hap9@epoch.ncsc.mil> + * Chris Bookholt <hap10@epoch.ncsc.mil> + */ + +#ifndef __XEN_PCIBACK_CONF_SPACE_QUIRKS_H__ +#define __XEN_PCIBACK_CONF_SPACE_QUIRKS_H__ + +#include <linux/pci.h> +#include <linux/list.h> + +struct xen_pcibk_config_quirk { + struct list_head quirks_list; + struct pci_device_id devid; + struct pci_dev *pdev; +}; + +int xen_pcibk_config_quirks_add_field(struct pci_dev *dev, struct config_field + *field); + +int xen_pcibk_config_quirks_remove_field(struct pci_dev *dev, int reg); + +int xen_pcibk_config_quirks_init(struct pci_dev *dev); + +void xen_pcibk_config_field_free(struct config_field *field); + +int xen_pcibk_config_quirk_release(struct pci_dev *dev); + +int xen_pcibk_field_is_dup(struct pci_dev *dev, unsigned int reg); + +#endif diff --git a/drivers/xen/xen-pciback/passthrough.c b/drivers/xen/xen-pciback/passthrough.c new file mode 100644 index 0000000..828dddc --- /dev/null +++ b/drivers/xen/xen-pciback/passthrough.c @@ -0,0 +1,188 @@ +/* + * PCI Backend - Provides restricted access to the real PCI bus topology + * to the frontend + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ + +#include <linux/list.h> +#include <linux/pci.h> +#include <linux/mutex.h> +#include "pciback.h" + +struct passthrough_dev_data { + /* Access to dev_list must be protected by lock */ + struct list_head dev_list; + struct mutex lock; +}; + +static struct pci_dev *__xen_pcibk_get_pci_dev(struct xen_pcibk_device *pdev, + unsigned int domain, + unsigned int bus, + unsigned int devfn) +{ + struct passthrough_dev_data *dev_data = pdev->pci_dev_data; + struct pci_dev_entry *dev_entry; + struct pci_dev *dev = NULL; + + mutex_lock(&dev_data->lock); + + list_for_each_entry(dev_entry, &dev_data->dev_list, list) { + if (domain == (unsigned int)pci_domain_nr(dev_entry->dev->bus) + && bus == (unsigned int)dev_entry->dev->bus->number + && devfn == dev_entry->dev->devfn) { + dev = dev_entry->dev; + break; + } + } + + mutex_unlock(&dev_data->lock); + + return dev; +} + +static int __xen_pcibk_add_pci_dev(struct xen_pcibk_device *pdev, + struct pci_dev *dev, + int devid, publish_pci_dev_cb publish_cb) +{ + struct passthrough_dev_data *dev_data = pdev->pci_dev_data; + struct pci_dev_entry *dev_entry; + unsigned int domain, bus, devfn; + int err; + + dev_entry = kmalloc(sizeof(*dev_entry), GFP_KERNEL); + if (!dev_entry) + return -ENOMEM; + dev_entry->dev = dev; + + mutex_lock(&dev_data->lock); + list_add_tail(&dev_entry->list, &dev_data->dev_list); + mutex_unlock(&dev_data->lock); + + /* Publish this device. */ + domain = (unsigned int)pci_domain_nr(dev->bus); + bus = (unsigned int)dev->bus->number; + devfn = dev->devfn; + err = publish_cb(pdev, domain, bus, devfn, devid); + + return err; +} + +static void __xen_pcibk_release_pci_dev(struct xen_pcibk_device *pdev, + struct pci_dev *dev) +{ + struct passthrough_dev_data *dev_data = pdev->pci_dev_data; + struct pci_dev_entry *dev_entry, *t; + struct pci_dev *found_dev = NULL; + + mutex_lock(&dev_data->lock); + + list_for_each_entry_safe(dev_entry, t, &dev_data->dev_list, list) { + if (dev_entry->dev == dev) { + list_del(&dev_entry->list); + found_dev = dev_entry->dev; + kfree(dev_entry); + } + } + + mutex_unlock(&dev_data->lock); + + if (found_dev) + pcistub_put_pci_dev(found_dev); +} + +static int __xen_pcibk_init_devices(struct xen_pcibk_device *pdev) +{ + struct passthrough_dev_data *dev_data; + + dev_data = kmalloc(sizeof(*dev_data), GFP_KERNEL); + if (!dev_data) + return -ENOMEM; + + mutex_init(&dev_data->lock); + + INIT_LIST_HEAD(&dev_data->dev_list); + + pdev->pci_dev_data = dev_data; + + return 0; +} + +static int __xen_pcibk_publish_pci_roots(struct xen_pcibk_device *pdev, + publish_pci_root_cb publish_root_cb) +{ + int err = 0; + struct passthrough_dev_data *dev_data = pdev->pci_dev_data; + struct pci_dev_entry *dev_entry, *e; + struct pci_dev *dev; + int found; + unsigned int domain, bus; + + mutex_lock(&dev_data->lock); + + list_for_each_entry(dev_entry, &dev_data->dev_list, list) { + /* Only publish this device as a root if none of its + * parent bridges are exported + */ + found = 0; + dev = dev_entry->dev->bus->self; + for (; !found && dev != NULL; dev = dev->bus->self) { + list_for_each_entry(e, &dev_data->dev_list, list) { + if (dev == e->dev) { + found = 1; + break; + } + } + } + + domain = (unsigned int)pci_domain_nr(dev_entry->dev->bus); + bus = (unsigned int)dev_entry->dev->bus->number; + + if (!found) { + err = publish_root_cb(pdev, domain, bus); + if (err) + break; + } + } + + mutex_unlock(&dev_data->lock); + + return err; +} + +static void __xen_pcibk_release_devices(struct xen_pcibk_device *pdev) +{ + struct passthrough_dev_data *dev_data = pdev->pci_dev_data; + struct pci_dev_entry *dev_entry, *t; + + list_for_each_entry_safe(dev_entry, t, &dev_data->dev_list, list) { + list_del(&dev_entry->list); + pcistub_put_pci_dev(dev_entry->dev); + kfree(dev_entry); + } + + kfree(dev_data); + pdev->pci_dev_data = NULL; +} + +static int __xen_pcibk_get_pcifront_dev(struct pci_dev *pcidev, + struct xen_pcibk_device *pdev, + unsigned int *domain, unsigned int *bus, + unsigned int *devfn) +{ + *domain = pci_domain_nr(pcidev->bus); + *bus = pcidev->bus->number; + *devfn = pcidev->devfn; + return 1; +} + +const struct xen_pcibk_backend xen_pcibk_passthrough_backend = { + .name = "passthrough", + .init = __xen_pcibk_init_devices, + .free = __xen_pcibk_release_devices, + .find = __xen_pcibk_get_pcifront_dev, + .publish = __xen_pcibk_publish_pci_roots, + .release = __xen_pcibk_release_pci_dev, + .add = __xen_pcibk_add_pci_dev, + .get = __xen_pcibk_get_pci_dev, +}; diff --git a/drivers/xen/xen-pciback/pci_stub.c b/drivers/xen/xen-pciback/pci_stub.c new file mode 100644 index 0000000..8f06e1e --- /dev/null +++ b/drivers/xen/xen-pciback/pci_stub.c @@ -0,0 +1,1371 @@ +/* + * PCI Stub Driver - Grabs devices in backend to be exported later + * + * Ryan Wilson <hap9@epoch.ncsc.mil> + * Chris Bookholt <hap10@epoch.ncsc.mil> + */ +#include <linux/module.h> +#include <linux/init.h> +#include <linux/rwsem.h> +#include <linux/list.h> +#include <linux/spinlock.h> +#include <linux/kref.h> +#include <linux/pci.h> +#include <linux/wait.h> +#include <linux/sched.h> +#include <linux/atomic.h> +#include <xen/events.h> +#include <asm/xen/pci.h> +#include <asm/xen/hypervisor.h> +#include "pciback.h" +#include "conf_space.h" +#include "conf_space_quirks.h" + +static char *pci_devs_to_hide; +wait_queue_head_t xen_pcibk_aer_wait_queue; +/*Add sem for sync AER handling and xen_pcibk remove/reconfigue ops, +* We want to avoid in middle of AER ops, xen_pcibk devices is being removed +*/ +static DECLARE_RWSEM(pcistub_sem); +module_param_named(hide, pci_devs_to_hide, charp, 0444); + +struct pcistub_device_id { + struct list_head slot_list; + int domain; + unsigned char bus; + unsigned int devfn; +}; +static LIST_HEAD(pcistub_device_ids); +static DEFINE_SPINLOCK(device_ids_lock); + +struct pcistub_device { + struct kref kref; + struct list_head dev_list; + spinlock_t lock; + + struct pci_dev *dev; + struct xen_pcibk_device *pdev;/* non-NULL if struct pci_dev is in use */ +}; + +/* Access to pcistub_devices & seized_devices lists and the initialize_devices + * flag must be locked with pcistub_devices_lock + */ +static DEFINE_SPINLOCK(pcistub_devices_lock); +static LIST_HEAD(pcistub_devices); + +/* wait for device_initcall before initializing our devices + * (see pcistub_init_devices_late) + */ +static int initialize_devices; +static LIST_HEAD(seized_devices); + +static struct pcistub_device *pcistub_device_alloc(struct pci_dev *dev) +{ + struct pcistub_device *psdev; + + dev_dbg(&dev->dev, "pcistub_device_alloc\n"); + + psdev = kzalloc(sizeof(*psdev), GFP_ATOMIC); + if (!psdev) + return NULL; + + psdev->dev = pci_dev_get(dev); + if (!psdev->dev) { + kfree(psdev); + return NULL; + } + + kref_init(&psdev->kref); + spin_lock_init(&psdev->lock); + + return psdev; +} + +/* Don't call this directly as it's called by pcistub_device_put */ +static void pcistub_device_release(struct kref *kref) +{ + struct pcistub_device *psdev; + + psdev = container_of(kref, struct pcistub_device, kref); + + dev_dbg(&psdev->dev->dev, "pcistub_device_release\n"); + + xen_unregister_device_domain_owner(psdev->dev); + + /* Clean-up the device */ + xen_pcibk_reset_device(psdev->dev); + xen_pcibk_config_free_dyn_fields(psdev->dev); + xen_pcibk_config_free_dev(psdev->dev); + kfree(pci_get_drvdata(psdev->dev)); + pci_set_drvdata(psdev->dev, NULL); + + pci_dev_put(psdev->dev); + + kfree(psdev); +} + +static inline void pcistub_device_get(struct pcistub_device *psdev) +{ + kref_get(&psdev->kref); +} + +static inline void pcistub_device_put(struct pcistub_device *psdev) +{ + kref_put(&psdev->kref, pcistub_device_release); +} + +static struct pcistub_device *pcistub_device_find(int domain, int bus, + int slot, int func) +{ + struct pcistub_device *psdev = NULL; + unsigned long flags; + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + list_for_each_entry(psdev, &pcistub_devices, dev_list) { + if (psdev->dev != NULL + && domain == pci_domain_nr(psdev->dev->bus) + && bus == psdev->dev->bus->number + && PCI_DEVFN(slot, func) == psdev->dev->devfn) { + pcistub_device_get(psdev); + goto out; + } + } + + /* didn't find it */ + psdev = NULL; + +out: + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + return psdev; +} + +static struct pci_dev *pcistub_device_get_pci_dev(struct xen_pcibk_device *pdev, + struct pcistub_device *psdev) +{ + struct pci_dev *pci_dev = NULL; + unsigned long flags; + + pcistub_device_get(psdev); + + spin_lock_irqsave(&psdev->lock, flags); + if (!psdev->pdev) { + psdev->pdev = pdev; + pci_dev = psdev->dev; + } + spin_unlock_irqrestore(&psdev->lock, flags); + + if (!pci_dev) + pcistub_device_put(psdev); + + return pci_dev; +} + +struct pci_dev *pcistub_get_pci_dev_by_slot(struct xen_pcibk_device *pdev, + int domain, int bus, + int slot, int func) +{ + struct pcistub_device *psdev; + struct pci_dev *found_dev = NULL; + unsigned long flags; + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + list_for_each_entry(psdev, &pcistub_devices, dev_list) { + if (psdev->dev != NULL + && domain == pci_domain_nr(psdev->dev->bus) + && bus == psdev->dev->bus->number + && PCI_DEVFN(slot, func) == psdev->dev->devfn) { + found_dev = pcistub_device_get_pci_dev(pdev, psdev); + break; + } + } + + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + return found_dev; +} + +struct pci_dev *pcistub_get_pci_dev(struct xen_pcibk_device *pdev, + struct pci_dev *dev) +{ + struct pcistub_device *psdev; + struct pci_dev *found_dev = NULL; + unsigned long flags; + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + list_for_each_entry(psdev, &pcistub_devices, dev_list) { + if (psdev->dev == dev) { + found_dev = pcistub_device_get_pci_dev(pdev, psdev); + break; + } + } + + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + return found_dev; +} + +void pcistub_put_pci_dev(struct pci_dev *dev) +{ + struct pcistub_device *psdev, *found_psdev = NULL; + unsigned long flags; + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + list_for_each_entry(psdev, &pcistub_devices, dev_list) { + if (psdev->dev == dev) { + found_psdev = psdev; + break; + } + } + + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + if (WARN_ON(!found_psdev)) + return; + + /*hold this lock for avoiding breaking link between + * pcistub and xen_pcibk when AER is in processing + */ + down_write(&pcistub_sem); + /* Cleanup our device + * (so it's ready for the next domain) + */ + xen_pcibk_reset_device(found_psdev->dev); + xen_pcibk_config_free_dyn_fields(found_psdev->dev); + xen_pcibk_config_reset_dev(found_psdev->dev); + + spin_lock_irqsave(&found_psdev->lock, flags); + found_psdev->pdev = NULL; + spin_unlock_irqrestore(&found_psdev->lock, flags); + + pcistub_device_put(found_psdev); + up_write(&pcistub_sem); +} + +static int __devinit pcistub_match_one(struct pci_dev *dev, + struct pcistub_device_id *pdev_id) +{ + /* Match the specified device by domain, bus, slot, func and also if + * any of the device's parent bridges match. + */ + for (; dev != NULL; dev = dev->bus->self) { + if (pci_domain_nr(dev->bus) == pdev_id->domain + && dev->bus->number == pdev_id->bus + && dev->devfn == pdev_id->devfn) + return 1; + + /* Sometimes topmost bridge links to itself. */ + if (dev == dev->bus->self) + break; + } + + return 0; +} + +static int __devinit pcistub_match(struct pci_dev *dev) +{ + struct pcistub_device_id *pdev_id; + unsigned long flags; + int found = 0; + + spin_lock_irqsave(&device_ids_lock, flags); + list_for_each_entry(pdev_id, &pcistub_device_ids, slot_list) { + if (pcistub_match_one(dev, pdev_id)) { + found = 1; + break; + } + } + spin_unlock_irqrestore(&device_ids_lock, flags); + + return found; +} + +static int __devinit pcistub_init_device(struct pci_dev *dev) +{ + struct xen_pcibk_dev_data *dev_data; + int err = 0; + + dev_dbg(&dev->dev, "initializing...\n"); + + /* The PCI backend is not intended to be a module (or to work with + * removable PCI devices (yet). If it were, xen_pcibk_config_free() + * would need to be called somewhere to free the memory allocated + * here and then to call kfree(pci_get_drvdata(psdev->dev)). + */ + dev_data = kzalloc(sizeof(*dev_data) + strlen(DRV_NAME "[]") + + strlen(pci_name(dev)) + 1, GFP_ATOMIC); + if (!dev_data) { + err = -ENOMEM; + goto out; + } + pci_set_drvdata(dev, dev_data); + + /* + * Setup name for fake IRQ handler. It will only be enabled + * once the device is turned on by the guest. + */ + sprintf(dev_data->irq_name, DRV_NAME "[%s]", pci_name(dev)); + + dev_dbg(&dev->dev, "initializing config\n"); + + init_waitqueue_head(&xen_pcibk_aer_wait_queue); + err = xen_pcibk_config_init_dev(dev); + if (err) + goto out; + + /* HACK: Force device (& ACPI) to determine what IRQ it's on - we + * must do this here because pcibios_enable_device may specify + * the pci device's true irq (and possibly its other resources) + * if they differ from what's in the configuration space. + * This makes the assumption that the device's resources won't + * change after this point (otherwise this code may break!) + */ + dev_dbg(&dev->dev, "enabling device\n"); + err = pci_enable_device(dev); + if (err) + goto config_release; + + /* Now disable the device (this also ensures some private device + * data is setup before we export) + */ + dev_dbg(&dev->dev, "reset device\n"); + xen_pcibk_reset_device(dev); + + return 0; + +config_release: + xen_pcibk_config_free_dev(dev); + +out: + pci_set_drvdata(dev, NULL); + kfree(dev_data); + return err; +} + +/* + * Because some initialization still happens on + * devices during fs_initcall, we need to defer + * full initialization of our devices until + * device_initcall. + */ +static int __init pcistub_init_devices_late(void) +{ + struct pcistub_device *psdev; + unsigned long flags; + int err = 0; + + pr_debug(DRV_NAME ": pcistub_init_devices_late\n"); + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + while (!list_empty(&seized_devices)) { + psdev = container_of(seized_devices.next, + struct pcistub_device, dev_list); + list_del(&psdev->dev_list); + + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + + err = pcistub_init_device(psdev->dev); + if (err) { + dev_err(&psdev->dev->dev, + "error %d initializing device\n", err); + kfree(psdev); + psdev = NULL; + } + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + if (psdev) + list_add_tail(&psdev->dev_list, &pcistub_devices); + } + + initialize_devices = 1; + + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + + return 0; +} + +static int __devinit pcistub_seize(struct pci_dev *dev) +{ + struct pcistub_device *psdev; + unsigned long flags; + int err = 0; + + psdev = pcistub_device_alloc(dev); + if (!psdev) + return -ENOMEM; + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + if (initialize_devices) { + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + + /* don't want irqs disabled when calling pcistub_init_device */ + err = pcistub_init_device(psdev->dev); + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + if (!err) + list_add(&psdev->dev_list, &pcistub_devices); + } else { + dev_dbg(&dev->dev, "deferring initialization\n"); + list_add(&psdev->dev_list, &seized_devices); + } + + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + + if (err) + pcistub_device_put(psdev); + + return err; +} + +static int __devinit pcistub_probe(struct pci_dev *dev, + const struct pci_device_id *id) +{ + int err = 0; + + dev_dbg(&dev->dev, "probing...\n"); + + if (pcistub_match(dev)) { + + if (dev->hdr_type != PCI_HEADER_TYPE_NORMAL + && dev->hdr_type != PCI_HEADER_TYPE_BRIDGE) { + dev_err(&dev->dev, "can't export pci devices that " + "don't have a normal (0) or bridge (1) " + "header type!\n"); + err = -ENODEV; + goto out; + } + + dev_info(&dev->dev, "seizing device\n"); + err = pcistub_seize(dev); + } else + /* Didn't find the device */ + err = -ENODEV; + +out: + return err; +} + +static void pcistub_remove(struct pci_dev *dev) +{ + struct pcistub_device *psdev, *found_psdev = NULL; + unsigned long flags; + + dev_dbg(&dev->dev, "removing\n"); + + spin_lock_irqsave(&pcistub_devices_lock, flags); + + xen_pcibk_config_quirk_release(dev); + + list_for_each_entry(psdev, &pcistub_devices, dev_list) { + if (psdev->dev == dev) { + found_psdev = psdev; + break; + } + } + + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + + if (found_psdev) { + dev_dbg(&dev->dev, "found device to remove - in use? %p\n", + found_psdev->pdev); + + if (found_psdev->pdev) { + printk(KERN_WARNING DRV_NAME ": ****** removing device " + "%s while still in-use! ******\n", + pci_name(found_psdev->dev)); + printk(KERN_WARNING DRV_NAME ": ****** driver domain may" + " still access this device's i/o resources!\n"); + printk(KERN_WARNING DRV_NAME ": ****** shutdown driver " + "domain before binding device\n"); + printk(KERN_WARNING DRV_NAME ": ****** to other drivers " + "or domains\n"); + + xen_pcibk_release_pci_dev(found_psdev->pdev, + found_psdev->dev); + } + + spin_lock_irqsave(&pcistub_devices_lock, flags); + list_del(&found_psdev->dev_list); + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + + /* the final put for releasing from the list */ + pcistub_device_put(found_psdev); + } +} + +static DEFINE_PCI_DEVICE_TABLE(pcistub_ids) = { + { + .vendor = PCI_ANY_ID, + .device = PCI_ANY_ID, + .subvendor = PCI_ANY_ID, + .subdevice = PCI_ANY_ID, + }, + {0,}, +}; + +#define PCI_NODENAME_MAX 40 +static void kill_domain_by_device(struct pcistub_device *psdev) +{ + struct xenbus_transaction xbt; + int err; + char nodename[PCI_NODENAME_MAX]; + + BUG_ON(!psdev); + snprintf(nodename, PCI_NODENAME_MAX, "/local/domain/0/backend/pci/%d/0", + psdev->pdev->xdev->otherend_id); + +again: + err = xenbus_transaction_start(&xbt); + if (err) { + dev_err(&psdev->dev->dev, + "error %d when start xenbus transaction\n", err); + return; + } + /*PV AER handlers will set this flag*/ + xenbus_printf(xbt, nodename, "aerState" , "aerfail"); + err = xenbus_transaction_end(xbt, 0); + if (err) { + if (err == -EAGAIN) + goto again; + dev_err(&psdev->dev->dev, + "error %d when end xenbus transaction\n", err); + return; + } +} + +/* For each aer recovery step error_detected, mmio_enabled, etc, front_end and + * backend need to have cooperation. In xen_pcibk, those steps will do similar + * jobs: send service request and waiting for front_end response. +*/ +static pci_ers_result_t common_process(struct pcistub_device *psdev, + pci_channel_state_t state, int aer_cmd, + pci_ers_result_t result) +{ + pci_ers_result_t res = result; + struct xen_pcie_aer_op *aer_op; + int ret; + + /*with PV AER drivers*/ + aer_op = &(psdev->pdev->sh_info->aer_op); + aer_op->cmd = aer_cmd ; + /*useful for error_detected callback*/ + aer_op->err = state; + /*pcifront_end BDF*/ + ret = xen_pcibk_get_pcifront_dev(psdev->dev, psdev->pdev, + &aer_op->domain, &aer_op->bus, &aer_op->devfn); + if (!ret) { + dev_err(&psdev->dev->dev, + DRV_NAME ": failed to get pcifront device\n"); + return PCI_ERS_RESULT_NONE; + } + wmb(); + + dev_dbg(&psdev->dev->dev, + DRV_NAME ": aer_op %x dom %x bus %x devfn %x\n", + aer_cmd, aer_op->domain, aer_op->bus, aer_op->devfn); + /*local flag to mark there's aer request, xen_pcibk callback will use + * this flag to judge whether we need to check pci-front give aer + * service ack signal + */ + set_bit(_PCIB_op_pending, (unsigned long *)&psdev->pdev->flags); + + /*It is possible that a pcifront conf_read_write ops request invokes + * the callback which cause the spurious execution of wake_up. + * Yet it is harmless and better than a spinlock here + */ + set_bit(_XEN_PCIB_active, + (unsigned long *)&psdev->pdev->sh_info->flags); + wmb(); + notify_remote_via_irq(psdev->pdev->evtchn_irq); + + ret = wait_event_timeout(xen_pcibk_aer_wait_queue, + !(test_bit(_XEN_PCIB_active, (unsigned long *) + &psdev->pdev->sh_info->flags)), 300*HZ); + + if (!ret) { + if (test_bit(_XEN_PCIB_active, + (unsigned long *)&psdev->pdev->sh_info->flags)) { + dev_err(&psdev->dev->dev, + "pcifront aer process not responding!\n"); + clear_bit(_XEN_PCIB_active, + (unsigned long *)&psdev->pdev->sh_info->flags); + aer_op->err = PCI_ERS_RESULT_NONE; + return res; + } + } + clear_bit(_PCIB_op_pending, (unsigned long *)&psdev->pdev->flags); + + if (test_bit(_XEN_PCIF_active, + (unsigned long *)&psdev->pdev->sh_info->flags)) { + dev_dbg(&psdev->dev->dev, + "schedule pci_conf service in " DRV_NAME "\n"); + xen_pcibk_test_and_schedule_op(psdev->pdev); + } + + res = (pci_ers_result_t)aer_op->err; + return res; +} + +/* +* xen_pcibk_slot_reset: it will send the slot_reset request to pcifront in case +* of the device driver could provide this service, and then wait for pcifront +* ack. +* @dev: pointer to PCI devices +* return value is used by aer_core do_recovery policy +*/ +static pci_ers_result_t xen_pcibk_slot_reset(struct pci_dev *dev) +{ + struct pcistub_device *psdev; + pci_ers_result_t result; + + result = PCI_ERS_RESULT_RECOVERED; + dev_dbg(&dev->dev, "xen_pcibk_slot_reset(bus:%x,devfn:%x)\n", + dev->bus->number, dev->devfn); + + down_write(&pcistub_sem); + psdev = pcistub_device_find(pci_domain_nr(dev->bus), + dev->bus->number, + PCI_SLOT(dev->devfn), + PCI_FUNC(dev->devfn)); + + if (!psdev || !psdev->pdev) { + dev_err(&dev->dev, + DRV_NAME " device is not found/assigned\n"); + goto end; + } + + if (!psdev->pdev->sh_info) { + dev_err(&dev->dev, DRV_NAME " device is not connected or owned" + " by HVM, kill it\n"); + kill_domain_by_device(psdev); + goto release; + } + + if (!test_bit(_XEN_PCIB_AERHANDLER, + (unsigned long *)&psdev->pdev->sh_info->flags)) { + dev_err(&dev->dev, + "guest with no AER driver should have been killed\n"); + goto release; + } + result = common_process(psdev, 1, XEN_PCI_OP_aer_slotreset, result); + + if (result == PCI_ERS_RESULT_NONE || + result == PCI_ERS_RESULT_DISCONNECT) { + dev_dbg(&dev->dev, + "No AER slot_reset service or disconnected!\n"); + kill_domain_by_device(psdev); + } +release: + pcistub_device_put(psdev); +end: + up_write(&pcistub_sem); + return result; + +} + + +/*xen_pcibk_mmio_enabled: it will send the mmio_enabled request to pcifront +* in case of the device driver could provide this service, and then wait +* for pcifront ack +* @dev: pointer to PCI devices +* return value is used by aer_core do_recovery policy +*/ + +static pci_ers_result_t xen_pcibk_mmio_enabled(struct pci_dev *dev) +{ + struct pcistub_device *psdev; + pci_ers_result_t result; + + result = PCI_ERS_RESULT_RECOVERED; + dev_dbg(&dev->dev, "xen_pcibk_mmio_enabled(bus:%x,devfn:%x)\n", + dev->bus->number, dev->devfn); + + down_write(&pcistub_sem); + psdev = pcistub_device_find(pci_domain_nr(dev->bus), + dev->bus->number, + PCI_SLOT(dev->devfn), + PCI_FUNC(dev->devfn)); + + if (!psdev || !psdev->pdev) { + dev_err(&dev->dev, + DRV_NAME " device is not found/assigned\n"); + goto end; + } + + if (!psdev->pdev->sh_info) { + dev_err(&dev->dev, DRV_NAME " device is not connected or owned" + " by HVM, kill it\n"); + kill_domain_by_device(psdev); + goto release; + } + + if (!test_bit(_XEN_PCIB_AERHANDLER, + (unsigned long *)&psdev->pdev->sh_info->flags)) { + dev_err(&dev->dev, + "guest with no AER driver should have been killed\n"); + goto release; + } + result = common_process(psdev, 1, XEN_PCI_OP_aer_mmio, result); + + if (result == PCI_ERS_RESULT_NONE || + result == PCI_ERS_RESULT_DISCONNECT) { + dev_dbg(&dev->dev, + "No AER mmio_enabled service or disconnected!\n"); + kill_domain_by_device(psdev); + } +release: + pcistub_device_put(psdev); +end: + up_write(&pcistub_sem); + return result; +} + +/*xen_pcibk_error_detected: it will send the error_detected request to pcifront +* in case of the device driver could provide this service, and then wait +* for pcifront ack. +* @dev: pointer to PCI devices +* @error: the current PCI connection state +* return value is used by aer_core do_recovery policy +*/ + +static pci_ers_result_t xen_pcibk_error_detected(struct pci_dev *dev, + pci_channel_state_t error) +{ + struct pcistub_device *psdev; + pci_ers_result_t result; + + result = PCI_ERS_RESULT_CAN_RECOVER; + dev_dbg(&dev->dev, "xen_pcibk_error_detected(bus:%x,devfn:%x)\n", + dev->bus->number, dev->devfn); + + down_write(&pcistub_sem); + psdev = pcistub_device_find(pci_domain_nr(dev->bus), + dev->bus->number, + PCI_SLOT(dev->devfn), + PCI_FUNC(dev->devfn)); + + if (!psdev || !psdev->pdev) { + dev_err(&dev->dev, + DRV_NAME " device is not found/assigned\n"); + goto end; + } + + if (!psdev->pdev->sh_info) { + dev_err(&dev->dev, DRV_NAME " device is not connected or owned" + " by HVM, kill it\n"); + kill_domain_by_device(psdev); + goto release; + } + + /*Guest owns the device yet no aer handler regiested, kill guest*/ + if (!test_bit(_XEN_PCIB_AERHANDLER, + (unsigned long *)&psdev->pdev->sh_info->flags)) { + dev_dbg(&dev->dev, "guest may have no aer driver, kill it\n"); + kill_domain_by_device(psdev); + goto release; + } + result = common_process(psdev, error, XEN_PCI_OP_aer_detected, result); + + if (result == PCI_ERS_RESULT_NONE || + result == PCI_ERS_RESULT_DISCONNECT) { + dev_dbg(&dev->dev, + "No AER error_detected service or disconnected!\n"); + kill_domain_by_device(psdev); + } +release: + pcistub_device_put(psdev); +end: + up_write(&pcistub_sem); + return result; +} + +/*xen_pcibk_error_resume: it will send the error_resume request to pcifront +* in case of the device driver could provide this service, and then wait +* for pcifront ack. +* @dev: pointer to PCI devices +*/ + +static void xen_pcibk_error_resume(struct pci_dev *dev) +{ + struct pcistub_device *psdev; + + dev_dbg(&dev->dev, "xen_pcibk_error_resume(bus:%x,devfn:%x)\n", + dev->bus->number, dev->devfn); + + down_write(&pcistub_sem); + psdev = pcistub_device_find(pci_domain_nr(dev->bus), + dev->bus->number, + PCI_SLOT(dev->devfn), + PCI_FUNC(dev->devfn)); + + if (!psdev || !psdev->pdev) { + dev_err(&dev->dev, + DRV_NAME " device is not found/assigned\n"); + goto end; + } + + if (!psdev->pdev->sh_info) { + dev_err(&dev->dev, DRV_NAME " device is not connected or owned" + " by HVM, kill it\n"); + kill_domain_by_device(psdev); + goto release; + } + + if (!test_bit(_XEN_PCIB_AERHANDLER, + (unsigned long *)&psdev->pdev->sh_info->flags)) { + dev_err(&dev->dev, + "guest with no AER driver should have been killed\n"); + kill_domain_by_device(psdev); + goto release; + } + common_process(psdev, 1, XEN_PCI_OP_aer_resume, + PCI_ERS_RESULT_RECOVERED); +release: + pcistub_device_put(psdev); +end: + up_write(&pcistub_sem); + return; +} + +/*add xen_pcibk AER handling*/ +static struct pci_error_handlers xen_pcibk_error_handler = { + .error_detected = xen_pcibk_error_detected, + .mmio_enabled = xen_pcibk_mmio_enabled, + .slot_reset = xen_pcibk_slot_reset, + .resume = xen_pcibk_error_resume, +}; + +/* + * Note: There is no MODULE_DEVICE_TABLE entry here because this isn't + * for a normal device. I don't want it to be loaded automatically. + */ + +static struct pci_driver xen_pcibk_pci_driver = { + /* The name should be xen_pciback, but until the tools are updated + * we will keep it as pciback. */ + .name = "pciback", + .id_table = pcistub_ids, + .probe = pcistub_probe, + .remove = pcistub_remove, + .err_handler = &xen_pcibk_error_handler, +}; + +static inline int str_to_slot(const char *buf, int *domain, int *bus, + int *slot, int *func) +{ + int err; + + err = sscanf(buf, " %x:%x:%x.%x", domain, bus, slot, func); + if (err == 4) + return 0; + else if (err < 0) + return -EINVAL; + + /* try again without domain */ + *domain = 0; + err = sscanf(buf, " %x:%x.%x", bus, slot, func); + if (err == 3) + return 0; + + return -EINVAL; +} + +static inline int str_to_quirk(const char *buf, int *domain, int *bus, int + *slot, int *func, int *reg, int *size, int *mask) +{ + int err; + + err = + sscanf(buf, " %04x:%02x:%02x.%1x-%08x:%1x:%08x", domain, bus, slot, + func, reg, size, mask); + if (err == 7) + return 0; + return -EINVAL; +} + +static int pcistub_device_id_add(int domain, int bus, int slot, int func) +{ + struct pcistub_device_id *pci_dev_id; + unsigned long flags; + + pci_dev_id = kmalloc(sizeof(*pci_dev_id), GFP_KERNEL); + if (!pci_dev_id) + return -ENOMEM; + + pci_dev_id->domain = domain; + pci_dev_id->bus = bus; + pci_dev_id->devfn = PCI_DEVFN(slot, func); + + pr_debug(DRV_NAME ": wants to seize %04x:%02x:%02x.%01x\n", + domain, bus, slot, func); + + spin_lock_irqsave(&device_ids_lock, flags); + list_add_tail(&pci_dev_id->slot_list, &pcistub_device_ids); + spin_unlock_irqrestore(&device_ids_lock, flags); + + return 0; +} + +static int pcistub_device_id_remove(int domain, int bus, int slot, int func) +{ + struct pcistub_device_id *pci_dev_id, *t; + int devfn = PCI_DEVFN(slot, func); + int err = -ENOENT; + unsigned long flags; + + spin_lock_irqsave(&device_ids_lock, flags); + list_for_each_entry_safe(pci_dev_id, t, &pcistub_device_ids, + slot_list) { + if (pci_dev_id->domain == domain + && pci_dev_id->bus == bus && pci_dev_id->devfn == devfn) { + /* Don't break; here because it's possible the same + * slot could be in the list more than once + */ + list_del(&pci_dev_id->slot_list); + kfree(pci_dev_id); + + err = 0; + + pr_debug(DRV_NAME ": removed %04x:%02x:%02x.%01x from " + "seize list\n", domain, bus, slot, func); + } + } + spin_unlock_irqrestore(&device_ids_lock, flags); + + return err; +} + +static int pcistub_reg_add(int domain, int bus, int slot, int func, int reg, + int size, int mask) +{ + int err = 0; + struct pcistub_device *psdev; + struct pci_dev *dev; + struct config_field *field; + + psdev = pcistub_device_find(domain, bus, slot, func); + if (!psdev || !psdev->dev) { + err = -ENODEV; + goto out; + } + dev = psdev->dev; + + field = kzalloc(sizeof(*field), GFP_ATOMIC); + if (!field) { + err = -ENOMEM; + goto out; + } + + field->offset = reg; + field->size = size; + field->mask = mask; + field->init = NULL; + field->reset = NULL; + field->release = NULL; + field->clean = xen_pcibk_config_field_free; + + err = xen_pcibk_config_quirks_add_field(dev, field); + if (err) + kfree(field); +out: + return err; +} + +static ssize_t pcistub_slot_add(struct device_driver *drv, const char *buf, + size_t count) +{ + int domain, bus, slot, func; + int err; + + err = str_to_slot(buf, &domain, &bus, &slot, &func); + if (err) + goto out; + + err = pcistub_device_id_add(domain, bus, slot, func); + +out: + if (!err) + err = count; + return err; +} +static DRIVER_ATTR(new_slot, S_IWUSR, NULL, pcistub_slot_add); + +static ssize_t pcistub_slot_remove(struct device_driver *drv, const char *buf, + size_t count) +{ + int domain, bus, slot, func; + int err; + + err = str_to_slot(buf, &domain, &bus, &slot, &func); + if (err) + goto out; + + err = pcistub_device_id_remove(domain, bus, slot, func); + +out: + if (!err) + err = count; + return err; +} +static DRIVER_ATTR(remove_slot, S_IWUSR, NULL, pcistub_slot_remove); + +static ssize_t pcistub_slot_show(struct device_driver *drv, char *buf) +{ + struct pcistub_device_id *pci_dev_id; + size_t count = 0; + unsigned long flags; + + spin_lock_irqsave(&device_ids_lock, flags); + list_for_each_entry(pci_dev_id, &pcistub_device_ids, slot_list) { + if (count >= PAGE_SIZE) + break; + + count += scnprintf(buf + count, PAGE_SIZE - count, + "%04x:%02x:%02x.%01x\n", + pci_dev_id->domain, pci_dev_id->bus, + PCI_SLOT(pci_dev_id->devfn), + PCI_FUNC(pci_dev_id->devfn)); + } + spin_unlock_irqrestore(&device_ids_lock, flags); + + return count; +} +static DRIVER_ATTR(slots, S_IRUSR, pcistub_slot_show, NULL); + +static ssize_t pcistub_irq_handler_show(struct device_driver *drv, char *buf) +{ + struct pcistub_device *psdev; + struct xen_pcibk_dev_data *dev_data; + size_t count = 0; + unsigned long flags; + + spin_lock_irqsave(&pcistub_devices_lock, flags); + list_for_each_entry(psdev, &pcistub_devices, dev_list) { + if (count >= PAGE_SIZE) + break; + if (!psdev->dev) + continue; + dev_data = pci_get_drvdata(psdev->dev); + if (!dev_data) + continue; + count += + scnprintf(buf + count, PAGE_SIZE - count, + "%s:%s:%sing:%ld\n", + pci_name(psdev->dev), + dev_data->isr_on ? "on" : "off", + dev_data->ack_intr ? "ack" : "not ack", + dev_data->handled); + } + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + return count; +} +static DRIVER_ATTR(irq_handlers, S_IRUSR, pcistub_irq_handler_show, NULL); + +static ssize_t pcistub_irq_handler_switch(struct device_driver *drv, + const char *buf, + size_t count) +{ + struct pcistub_device *psdev; + struct xen_pcibk_dev_data *dev_data; + int domain, bus, slot, func; + int err = -ENOENT; + + err = str_to_slot(buf, &domain, &bus, &slot, &func); + if (err) + goto out; + + psdev = pcistub_device_find(domain, bus, slot, func); + + if (!psdev) + goto out; + + dev_data = pci_get_drvdata(psdev->dev); + if (!dev_data) + goto out; + + dev_dbg(&psdev->dev->dev, "%s fake irq handler: %d->%d\n", + dev_data->irq_name, dev_data->isr_on, + !dev_data->isr_on); + + dev_data->isr_on = !(dev_data->isr_on); + if (dev_data->isr_on) + dev_data->ack_intr = 1; +out: + if (!err) + err = count; + return err; +} +static DRIVER_ATTR(irq_handler_state, S_IWUSR, NULL, + pcistub_irq_handler_switch); + +static ssize_t pcistub_quirk_add(struct device_driver *drv, const char *buf, + size_t count) +{ + int domain, bus, slot, func, reg, size, mask; + int err; + + err = str_to_quirk(buf, &domain, &bus, &slot, &func, ®, &size, + &mask); + if (err) + goto out; + + err = pcistub_reg_add(domain, bus, slot, func, reg, size, mask); + +out: + if (!err) + err = count; + return err; +} + +static ssize_t pcistub_quirk_show(struct device_driver *drv, char *buf) +{ + int count = 0; + unsigned long flags; + struct xen_pcibk_config_quirk *quirk; + struct xen_pcibk_dev_data *dev_data; + const struct config_field *field; + const struct config_field_entry *cfg_entry; + + spin_lock_irqsave(&device_ids_lock, flags); + list_for_each_entry(quirk, &xen_pcibk_quirks, quirks_list) { + if (count >= PAGE_SIZE) + goto out; + + count += scnprintf(buf + count, PAGE_SIZE - count, + "%02x:%02x.%01x\n\t%04x:%04x:%04x:%04x\n", + quirk->pdev->bus->number, + PCI_SLOT(quirk->pdev->devfn), + PCI_FUNC(quirk->pdev->devfn), + quirk->devid.vendor, quirk->devid.device, + quirk->devid.subvendor, + quirk->devid.subdevice); + + dev_data = pci_get_drvdata(quirk->pdev); + + list_for_each_entry(cfg_entry, &dev_data->config_fields, list) { + field = cfg_entry->field; + if (count >= PAGE_SIZE) + goto out; + + count += scnprintf(buf + count, PAGE_SIZE - count, + "\t\t%08x:%01x:%08x\n", + cfg_entry->base_offset + + field->offset, field->size, + field->mask); + } + } + +out: + spin_unlock_irqrestore(&device_ids_lock, flags); + + return count; +} +static DRIVER_ATTR(quirks, S_IRUSR | S_IWUSR, pcistub_quirk_show, + pcistub_quirk_add); + +static ssize_t permissive_add(struct device_driver *drv, const char *buf, + size_t count) +{ + int domain, bus, slot, func; + int err; + struct pcistub_device *psdev; + struct xen_pcibk_dev_data *dev_data; + err = str_to_slot(buf, &domain, &bus, &slot, &func); + if (err) + goto out; + psdev = pcistub_device_find(domain, bus, slot, func); + if (!psdev) { + err = -ENODEV; + goto out; + } + if (!psdev->dev) { + err = -ENODEV; + goto release; + } + dev_data = pci_get_drvdata(psdev->dev); + /* the driver data for a device should never be null at this point */ + if (!dev_data) { + err = -ENXIO; + goto release; + } + if (!dev_data->permissive) { + dev_data->permissive = 1; + /* Let user know that what they're doing could be unsafe */ + dev_warn(&psdev->dev->dev, "enabling permissive mode " + "configuration space accesses!\n"); + dev_warn(&psdev->dev->dev, + "permissive mode is potentially unsafe!\n"); + } +release: + pcistub_device_put(psdev); +out: + if (!err) + err = count; + return err; +} + +static ssize_t permissive_show(struct device_driver *drv, char *buf) +{ + struct pcistub_device *psdev; + struct xen_pcibk_dev_data *dev_data; + size_t count = 0; + unsigned long flags; + spin_lock_irqsave(&pcistub_devices_lock, flags); + list_for_each_entry(psdev, &pcistub_devices, dev_list) { + if (count >= PAGE_SIZE) + break; + if (!psdev->dev) + continue; + dev_data = pci_get_drvdata(psdev->dev); + if (!dev_data || !dev_data->permissive) + continue; + count += + scnprintf(buf + count, PAGE_SIZE - count, "%s\n", + pci_name(psdev->dev)); + } + spin_unlock_irqrestore(&pcistub_devices_lock, flags); + return count; +} +static DRIVER_ATTR(permissive, S_IRUSR | S_IWUSR, permissive_show, + permissive_add); + +static void pcistub_exit(void) +{ + driver_remove_file(&xen_pcibk_pci_driver.driver, &driver_attr_new_slot); + driver_remove_file(&xen_pcibk_pci_driver.driver, + &driver_attr_remove_slot); + driver_remove_file(&xen_pcibk_pci_driver.driver, &driver_attr_slots); + driver_remove_file(&xen_pcibk_pci_driver.driver, &driver_attr_quirks); + driver_remove_file(&xen_pcibk_pci_driver.driver, + &driver_attr_permissive); + driver_remove_file(&xen_pcibk_pci_driver.driver, + &driver_attr_irq_handlers); + driver_remove_file(&xen_pcibk_pci_driver.driver, + &driver_attr_irq_handler_state); + pci_unregister_driver(&xen_pcibk_pci_driver); +} + +static int __init pcistub_init(void) +{ + int pos = 0; + int err = 0; + int domain, bus, slot, func; + int parsed; + + if (pci_devs_to_hide && *pci_devs_to_hide) { + do { + parsed = 0; + + err = sscanf(pci_devs_to_hide + pos, + " (%x:%x:%x.%x) %n", + &domain, &bus, &slot, &func, &parsed); + if (err != 4) { + domain = 0; + err = sscanf(pci_devs_to_hide + pos, + " (%x:%x.%x) %n", + &bus, &slot, &func, &parsed); + if (err != 3) + goto parse_error; + } + + err = pcistub_device_id_add(domain, bus, slot, func); + if (err) + goto out; + + /* if parsed<=0, we've reached the end of the string */ + pos += parsed; + } while (parsed > 0 && pci_devs_to_hide[pos]); + } + + /* If we're the first PCI Device Driver to register, we're the + * first one to get offered PCI devices as they become + * available (and thus we can be the first to grab them) + */ + err = pci_register_driver(&xen_pcibk_pci_driver); + if (err < 0) + goto out; + + err = driver_create_file(&xen_pcibk_pci_driver.driver, + &driver_attr_new_slot); + if (!err) + err = driver_create_file(&xen_pcibk_pci_driver.driver, + &driver_attr_remove_slot); + if (!err) + err = driver_create_file(&xen_pcibk_pci_driver.driver, + &driver_attr_slots); + if (!err) + err = driver_create_file(&xen_pcibk_pci_driver.driver, + &driver_attr_quirks); + if (!err) + err = driver_create_file(&xen_pcibk_pci_driver.driver, + &driver_attr_permissive); + + if (!err) + err = driver_create_file(&xen_pcibk_pci_driver.driver, + &driver_attr_irq_handlers); + if (!err) + err = driver_create_file(&xen_pcibk_pci_driver.driver, + &driver_attr_irq_handler_state); + if (err) + pcistub_exit(); + +out: + return err; + +parse_error: + printk(KERN_ERR DRV_NAME ": Error parsing pci_devs_to_hide at \"%s\"\n", + pci_devs_to_hide + pos); + return -EINVAL; +} + +#ifndef MODULE +/* + * fs_initcall happens before device_initcall + * so xen_pcibk *should* get called first (b/c we + * want to suck up any device before other drivers + * get a chance by being the first pci device + * driver to register) + */ +fs_initcall(pcistub_init); +#endif + +static int __init xen_pcibk_init(void) +{ + int err; + + if (!xen_initial_domain()) + return -ENODEV; + + err = xen_pcibk_config_init(); + if (err) + return err; + +#ifdef MODULE + err = pcistub_init(); + if (err < 0) + return err; +#endif + + pcistub_init_devices_late(); + err = xen_pcibk_xenbus_register(); + if (err) + pcistub_exit(); + + return err; +} + +static void __exit xen_pcibk_cleanup(void) +{ + xen_pcibk_xenbus_unregister(); + pcistub_exit(); +} + +module_init(xen_pcibk_init); +module_exit(xen_pcibk_cleanup); + +MODULE_LICENSE("Dual BSD/GPL"); +MODULE_ALIAS("xen-backend:pci"); diff --git a/drivers/xen/xen-pciback/pciback.h b/drivers/xen/xen-pciback/pciback.h new file mode 100644 index 0000000..e9b4011 --- /dev/null +++ b/drivers/xen/xen-pciback/pciback.h @@ -0,0 +1,191 @@ +/* + * PCI Backend Common Data Structures & Function Declarations + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ +#ifndef __XEN_PCIBACK_H__ +#define __XEN_PCIBACK_H__ + +#include <linux/pci.h> +#include <linux/interrupt.h> +#include <xen/xenbus.h> +#include <linux/list.h> +#include <linux/spinlock.h> +#include <linux/workqueue.h> +#include <linux/atomic.h> +#include <xen/interface/io/pciif.h> + +#define DRV_NAME "xen-pciback" + +struct pci_dev_entry { + struct list_head list; + struct pci_dev *dev; +}; + +#define _PDEVF_op_active (0) +#define PDEVF_op_active (1<<(_PDEVF_op_active)) +#define _PCIB_op_pending (1) +#define PCIB_op_pending (1<<(_PCIB_op_pending)) + +struct xen_pcibk_device { + void *pci_dev_data; + struct mutex dev_lock; + struct xenbus_device *xdev; + struct xenbus_watch be_watch; + u8 be_watching; + int evtchn_irq; + struct xen_pci_sharedinfo *sh_info; + unsigned long flags; + struct work_struct op_work; +}; + +struct xen_pcibk_dev_data { + struct list_head config_fields; + unsigned int permissive:1; + unsigned int warned_on_write:1; + unsigned int enable_intx:1; + unsigned int isr_on:1; /* Whether the IRQ handler is installed. */ + unsigned int ack_intr:1; /* .. and ACK-ing */ + unsigned long handled; + unsigned int irq; /* Saved in case device transitions to MSI/MSI-X */ + char irq_name[0]; /* xen-pcibk[000:04:00.0] */ +}; + +/* Used by XenBus and xen_pcibk_ops.c */ +extern wait_queue_head_t xen_pcibk_aer_wait_queue; +extern struct workqueue_struct *xen_pcibk_wq; +/* Used by pcistub.c and conf_space_quirks.c */ +extern struct list_head xen_pcibk_quirks; + +/* Get/Put PCI Devices that are hidden from the PCI Backend Domain */ +struct pci_dev *pcistub_get_pci_dev_by_slot(struct xen_pcibk_device *pdev, + int domain, int bus, + int slot, int func); +struct pci_dev *pcistub_get_pci_dev(struct xen_pcibk_device *pdev, + struct pci_dev *dev); +void pcistub_put_pci_dev(struct pci_dev *dev); + +/* Ensure a device is turned off or reset */ +void xen_pcibk_reset_device(struct pci_dev *pdev); + +/* Access a virtual configuration space for a PCI device */ +int xen_pcibk_config_init(void); +int xen_pcibk_config_init_dev(struct pci_dev *dev); +void xen_pcibk_config_free_dyn_fields(struct pci_dev *dev); +void xen_pcibk_config_reset_dev(struct pci_dev *dev); +void xen_pcibk_config_free_dev(struct pci_dev *dev); +int xen_pcibk_config_read(struct pci_dev *dev, int offset, int size, + u32 *ret_val); +int xen_pcibk_config_write(struct pci_dev *dev, int offset, int size, + u32 value); + +/* Handle requests for specific devices from the frontend */ +typedef int (*publish_pci_dev_cb) (struct xen_pcibk_device *pdev, + unsigned int domain, unsigned int bus, + unsigned int devfn, unsigned int devid); +typedef int (*publish_pci_root_cb) (struct xen_pcibk_device *pdev, + unsigned int domain, unsigned int bus); + +/* Backend registration for the two types of BDF representation: + * vpci - BDFs start at 00 + * passthrough - BDFs are exactly like in the host. + */ +struct xen_pcibk_backend { + const char *name; + int (*init)(struct xen_pcibk_device *pdev); + void (*free)(struct xen_pcibk_device *pdev); + int (*find)(struct pci_dev *pcidev, struct xen_pcibk_device *pdev, + unsigned int *domain, unsigned int *bus, + unsigned int *devfn); + int (*publish)(struct xen_pcibk_device *pdev, publish_pci_root_cb cb); + void (*release)(struct xen_pcibk_device *pdev, struct pci_dev *dev); + int (*add)(struct xen_pcibk_device *pdev, struct pci_dev *dev, + int devid, publish_pci_dev_cb publish_cb); + struct pci_dev *(*get)(struct xen_pcibk_device *pdev, + unsigned int domain, unsigned int bus, + unsigned int devfn); +}; + +extern const struct xen_pcibk_backend xen_pcibk_vpci_backend; +extern const struct xen_pcibk_backend xen_pcibk_passthrough_backend; +extern const struct xen_pcibk_backend *xen_pcibk_backend; + +static inline int xen_pcibk_add_pci_dev(struct xen_pcibk_device *pdev, + struct pci_dev *dev, + int devid, + publish_pci_dev_cb publish_cb) +{ + if (xen_pcibk_backend && xen_pcibk_backend->add) + return xen_pcibk_backend->add(pdev, dev, devid, publish_cb); + return -1; +} + +static inline void xen_pcibk_release_pci_dev(struct xen_pcibk_device *pdev, + struct pci_dev *dev) +{ + if (xen_pcibk_backend && xen_pcibk_backend->free) + return xen_pcibk_backend->release(pdev, dev); +} + +static inline struct pci_dev * +xen_pcibk_get_pci_dev(struct xen_pcibk_device *pdev, unsigned int domain, + unsigned int bus, unsigned int devfn) +{ + if (xen_pcibk_backend && xen_pcibk_backend->get) + return xen_pcibk_backend->get(pdev, domain, bus, devfn); + return NULL; +} + +/** +* Add for domain0 PCIE-AER handling. Get guest domain/bus/devfn in xen_pcibk +* before sending aer request to pcifront, so that guest could identify +* device, coopearte with xen_pcibk to finish aer recovery job if device driver +* has the capability +*/ +static inline int xen_pcibk_get_pcifront_dev(struct pci_dev *pcidev, + struct xen_pcibk_device *pdev, + unsigned int *domain, + unsigned int *bus, + unsigned int *devfn) +{ + if (xen_pcibk_backend && xen_pcibk_backend->find) + return xen_pcibk_backend->find(pcidev, pdev, domain, bus, + devfn); + return -1; +} + +static inline int xen_pcibk_init_devices(struct xen_pcibk_device *pdev) +{ + if (xen_pcibk_backend && xen_pcibk_backend->init) + return xen_pcibk_backend->init(pdev); + return -1; +} + +static inline int xen_pcibk_publish_pci_roots(struct xen_pcibk_device *pdev, + publish_pci_root_cb cb) +{ + if (xen_pcibk_backend && xen_pcibk_backend->publish) + return xen_pcibk_backend->publish(pdev, cb); + return -1; +} + +static inline void xen_pcibk_release_devices(struct xen_pcibk_device *pdev) +{ + if (xen_pcibk_backend && xen_pcibk_backend->free) + return xen_pcibk_backend->free(pdev); +} + +/* Handles events from front-end */ +irqreturn_t xen_pcibk_handle_event(int irq, void *dev_id); +void xen_pcibk_do_op(struct work_struct *data); + +int xen_pcibk_xenbus_register(void); +void xen_pcibk_xenbus_unregister(void); + +extern int verbose_request; + +void xen_pcibk_test_and_schedule_op(struct xen_pcibk_device *pdev); +#endif + +/* Handles shared IRQs that can to device domain and control domain. */ +void xen_pcibk_irq_handler(struct pci_dev *dev, int reset); diff --git a/drivers/xen/xen-pciback/pciback_ops.c b/drivers/xen/xen-pciback/pciback_ops.c new file mode 100644 index 0000000..63616d7 --- /dev/null +++ b/drivers/xen/xen-pciback/pciback_ops.c @@ -0,0 +1,383 @@ +/* + * PCI Backend Operations - respond to PCI requests from Frontend + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ +#include <linux/module.h> +#include <linux/wait.h> +#include <linux/bitops.h> +#include <xen/events.h> +#include <linux/sched.h> +#include "pciback.h" + +int verbose_request; +module_param(verbose_request, int, 0644); + +static irqreturn_t xen_pcibk_guest_interrupt(int irq, void *dev_id); + +/* Ensure a device is has the fake IRQ handler "turned on/off" and is + * ready to be exported. This MUST be run after xen_pcibk_reset_device + * which does the actual PCI device enable/disable. + */ +static void xen_pcibk_control_isr(struct pci_dev *dev, int reset) +{ + struct xen_pcibk_dev_data *dev_data; + int rc; + int enable = 0; + + dev_data = pci_get_drvdata(dev); + if (!dev_data) + return; + + /* We don't deal with bridges */ + if (dev->hdr_type != PCI_HEADER_TYPE_NORMAL) + return; + + if (reset) { + dev_data->enable_intx = 0; + dev_data->ack_intr = 0; + } + enable = dev_data->enable_intx; + + /* Asked to disable, but ISR isn't runnig */ + if (!enable && !dev_data->isr_on) + return; + + /* Squirrel away the IRQs in the dev_data. We need this + * b/c when device transitions to MSI, the dev->irq is + * overwritten with the MSI vector. + */ + if (enable) + dev_data->irq = dev->irq; + + /* + * SR-IOV devices in all use MSI-X and have no legacy + * interrupts, so inhibit creating a fake IRQ handler for them. + */ + if (dev_data->irq == 0) + goto out; + + dev_dbg(&dev->dev, "%s: #%d %s %s%s %s-> %s\n", + dev_data->irq_name, + dev_data->irq, + pci_is_enabled(dev) ? "on" : "off", + dev->msi_enabled ? "MSI" : "", + dev->msix_enabled ? "MSI/X" : "", + dev_data->isr_on ? "enable" : "disable", + enable ? "enable" : "disable"); + + if (enable) { + rc = request_irq(dev_data->irq, + xen_pcibk_guest_interrupt, IRQF_SHARED, + dev_data->irq_name, dev); + if (rc) { + dev_err(&dev->dev, "%s: failed to install fake IRQ " \ + "handler for IRQ %d! (rc:%d)\n", + dev_data->irq_name, dev_data->irq, rc); + goto out; + } + } else { + free_irq(dev_data->irq, dev); + dev_data->irq = 0; + } + dev_data->isr_on = enable; + dev_data->ack_intr = enable; +out: + dev_dbg(&dev->dev, "%s: #%d %s %s%s %s\n", + dev_data->irq_name, + dev_data->irq, + pci_is_enabled(dev) ? "on" : "off", + dev->msi_enabled ? "MSI" : "", + dev->msix_enabled ? "MSI/X" : "", + enable ? (dev_data->isr_on ? "enabled" : "failed to enable") : + (dev_data->isr_on ? "failed to disable" : "disabled")); +} + +/* Ensure a device is "turned off" and ready to be exported. + * (Also see xen_pcibk_config_reset to ensure virtual configuration space is + * ready to be re-exported) + */ +void xen_pcibk_reset_device(struct pci_dev *dev) +{ + u16 cmd; + + xen_pcibk_control_isr(dev, 1 /* reset device */); + + /* Disable devices (but not bridges) */ + if (dev->hdr_type == PCI_HEADER_TYPE_NORMAL) { +#ifdef CONFIG_PCI_MSI + /* The guest could have been abruptly killed without + * disabling MSI/MSI-X interrupts.*/ + if (dev->msix_enabled) + pci_disable_msix(dev); + if (dev->msi_enabled) + pci_disable_msi(dev); +#endif + pci_disable_device(dev); + + pci_write_config_word(dev, PCI_COMMAND, 0); + + dev->is_busmaster = 0; + } else { + pci_read_config_word(dev, PCI_COMMAND, &cmd); + if (cmd & (PCI_COMMAND_INVALIDATE)) { + cmd &= ~(PCI_COMMAND_INVALIDATE); + pci_write_config_word(dev, PCI_COMMAND, cmd); + + dev->is_busmaster = 0; + } + } +} + +#ifdef CONFIG_PCI_MSI +static +int xen_pcibk_enable_msi(struct xen_pcibk_device *pdev, + struct pci_dev *dev, struct xen_pci_op *op) +{ + struct xen_pcibk_dev_data *dev_data; + int otherend = pdev->xdev->otherend_id; + int status; + + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: enable MSI\n", pci_name(dev)); + + status = pci_enable_msi(dev); + + if (status) { + printk(KERN_ERR "error enable msi for guest %x status %x\n", + otherend, status); + op->value = 0; + return XEN_PCI_ERR_op_failed; + } + + /* The value the guest needs is actually the IDT vector, not the + * the local domain's IRQ number. */ + + op->value = dev->irq ? xen_pirq_from_irq(dev->irq) : 0; + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: MSI: %d\n", pci_name(dev), + op->value); + + dev_data = pci_get_drvdata(dev); + if (dev_data) + dev_data->ack_intr = 0; + + return 0; +} + +static +int xen_pcibk_disable_msi(struct xen_pcibk_device *pdev, + struct pci_dev *dev, struct xen_pci_op *op) +{ + struct xen_pcibk_dev_data *dev_data; + + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: disable MSI\n", + pci_name(dev)); + pci_disable_msi(dev); + + op->value = dev->irq ? xen_pirq_from_irq(dev->irq) : 0; + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: MSI: %d\n", pci_name(dev), + op->value); + dev_data = pci_get_drvdata(dev); + if (dev_data) + dev_data->ack_intr = 1; + return 0; +} + +static +int xen_pcibk_enable_msix(struct xen_pcibk_device *pdev, + struct pci_dev *dev, struct xen_pci_op *op) +{ + struct xen_pcibk_dev_data *dev_data; + int i, result; + struct msix_entry *entries; + + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: enable MSI-X\n", + pci_name(dev)); + if (op->value > SH_INFO_MAX_VEC) + return -EINVAL; + + entries = kmalloc(op->value * sizeof(*entries), GFP_KERNEL); + if (entries == NULL) + return -ENOMEM; + + for (i = 0; i < op->value; i++) { + entries[i].entry = op->msix_entries[i].entry; + entries[i].vector = op->msix_entries[i].vector; + } + + result = pci_enable_msix(dev, entries, op->value); + + if (result == 0) { + for (i = 0; i < op->value; i++) { + op->msix_entries[i].entry = entries[i].entry; + if (entries[i].vector) + op->msix_entries[i].vector = + xen_pirq_from_irq(entries[i].vector); + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: " \ + "MSI-X[%d]: %d\n", + pci_name(dev), i, + op->msix_entries[i].vector); + } + } else { + printk(KERN_WARNING DRV_NAME ": %s: failed to enable MSI-X: err %d!\n", + pci_name(dev), result); + } + kfree(entries); + + op->value = result; + dev_data = pci_get_drvdata(dev); + if (dev_data) + dev_data->ack_intr = 0; + + return result; +} + +static +int xen_pcibk_disable_msix(struct xen_pcibk_device *pdev, + struct pci_dev *dev, struct xen_pci_op *op) +{ + struct xen_pcibk_dev_data *dev_data; + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: disable MSI-X\n", + pci_name(dev)); + pci_disable_msix(dev); + + /* + * SR-IOV devices (which don't have any legacy IRQ) have + * an undefined IRQ value of zero. + */ + op->value = dev->irq ? xen_pirq_from_irq(dev->irq) : 0; + if (unlikely(verbose_request)) + printk(KERN_DEBUG DRV_NAME ": %s: MSI-X: %d\n", pci_name(dev), + op->value); + dev_data = pci_get_drvdata(dev); + if (dev_data) + dev_data->ack_intr = 1; + return 0; +} +#endif +/* +* Now the same evtchn is used for both pcifront conf_read_write request +* as well as pcie aer front end ack. We use a new work_queue to schedule +* xen_pcibk conf_read_write service for avoiding confict with aer_core +* do_recovery job which also use the system default work_queue +*/ +void xen_pcibk_test_and_schedule_op(struct xen_pcibk_device *pdev) +{ + /* Check that frontend is requesting an operation and that we are not + * already processing a request */ + if (test_bit(_XEN_PCIF_active, (unsigned long *)&pdev->sh_info->flags) + && !test_and_set_bit(_PDEVF_op_active, &pdev->flags)) { + queue_work(xen_pcibk_wq, &pdev->op_work); + } + /*_XEN_PCIB_active should have been cleared by pcifront. And also make + sure xen_pcibk is waiting for ack by checking _PCIB_op_pending*/ + if (!test_bit(_XEN_PCIB_active, (unsigned long *)&pdev->sh_info->flags) + && test_bit(_PCIB_op_pending, &pdev->flags)) { + wake_up(&xen_pcibk_aer_wait_queue); + } +} + +/* Performing the configuration space reads/writes must not be done in atomic + * context because some of the pci_* functions can sleep (mostly due to ACPI + * use of semaphores). This function is intended to be called from a work + * queue in process context taking a struct xen_pcibk_device as a parameter */ + +void xen_pcibk_do_op(struct work_struct *data) +{ + struct xen_pcibk_device *pdev = + container_of(data, struct xen_pcibk_device, op_work); + struct pci_dev *dev; + struct xen_pcibk_dev_data *dev_data = NULL; + struct xen_pci_op *op = &pdev->sh_info->op; + int test_intx = 0; + + dev = xen_pcibk_get_pci_dev(pdev, op->domain, op->bus, op->devfn); + + if (dev == NULL) + op->err = XEN_PCI_ERR_dev_not_found; + else { + dev_data = pci_get_drvdata(dev); + if (dev_data) + test_intx = dev_data->enable_intx; + switch (op->cmd) { + case XEN_PCI_OP_conf_read: + op->err = xen_pcibk_config_read(dev, + op->offset, op->size, &op->value); + break; + case XEN_PCI_OP_conf_write: + op->err = xen_pcibk_config_write(dev, + op->offset, op->size, op->value); + break; +#ifdef CONFIG_PCI_MSI + case XEN_PCI_OP_enable_msi: + op->err = xen_pcibk_enable_msi(pdev, dev, op); + break; + case XEN_PCI_OP_disable_msi: + op->err = xen_pcibk_disable_msi(pdev, dev, op); + break; + case XEN_PCI_OP_enable_msix: + op->err = xen_pcibk_enable_msix(pdev, dev, op); + break; + case XEN_PCI_OP_disable_msix: + op->err = xen_pcibk_disable_msix(pdev, dev, op); + break; +#endif + default: + op->err = XEN_PCI_ERR_not_implemented; + break; + } + } + if (!op->err && dev && dev_data) { + /* Transition detected */ + if ((dev_data->enable_intx != test_intx)) + xen_pcibk_control_isr(dev, 0 /* no reset */); + } + /* Tell the driver domain that we're done. */ + wmb(); + clear_bit(_XEN_PCIF_active, (unsigned long *)&pdev->sh_info->flags); + notify_remote_via_irq(pdev->evtchn_irq); + + /* Mark that we're done. */ + smp_mb__before_clear_bit(); /* /after/ clearing PCIF_active */ + clear_bit(_PDEVF_op_active, &pdev->flags); + smp_mb__after_clear_bit(); /* /before/ final check for work */ + + /* Check to see if the driver domain tried to start another request in + * between clearing _XEN_PCIF_active and clearing _PDEVF_op_active. + */ + xen_pcibk_test_and_schedule_op(pdev); +} + +irqreturn_t xen_pcibk_handle_event(int irq, void *dev_id) +{ + struct xen_pcibk_device *pdev = dev_id; + + xen_pcibk_test_and_schedule_op(pdev); + + return IRQ_HANDLED; +} +static irqreturn_t xen_pcibk_guest_interrupt(int irq, void *dev_id) +{ + struct pci_dev *dev = (struct pci_dev *)dev_id; + struct xen_pcibk_dev_data *dev_data = pci_get_drvdata(dev); + + if (dev_data->isr_on && dev_data->ack_intr) { + dev_data->handled++; + if ((dev_data->handled % 1000) == 0) { + if (xen_test_irq_shared(irq)) { + printk(KERN_INFO "%s IRQ line is not shared " + "with other domains. Turning ISR off\n", + dev_data->irq_name); + dev_data->ack_intr = 0; + } + } + return IRQ_HANDLED; + } + return IRQ_NONE; +} diff --git a/drivers/xen/xen-pciback/vpci.c b/drivers/xen/xen-pciback/vpci.c new file mode 100644 index 0000000..46d140b --- /dev/null +++ b/drivers/xen/xen-pciback/vpci.c @@ -0,0 +1,254 @@ +/* + * PCI Backend - Provides a Virtual PCI bus (with real devices) + * to the frontend + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ + +#include <linux/list.h> +#include <linux/slab.h> +#include <linux/pci.h> +#include <linux/mutex.h> +#include "pciback.h" + +#define PCI_SLOT_MAX 32 + +struct vpci_dev_data { + /* Access to dev_list must be protected by lock */ + struct list_head dev_list[PCI_SLOT_MAX]; + struct mutex lock; +}; + +static inline struct list_head *list_first(struct list_head *head) +{ + return head->next; +} + +static struct pci_dev *__xen_pcibk_get_pci_dev(struct xen_pcibk_device *pdev, + unsigned int domain, + unsigned int bus, + unsigned int devfn) +{ + struct pci_dev_entry *entry; + struct pci_dev *dev = NULL; + struct vpci_dev_data *vpci_dev = pdev->pci_dev_data; + + if (domain != 0 || bus != 0) + return NULL; + + if (PCI_SLOT(devfn) < PCI_SLOT_MAX) { + mutex_lock(&vpci_dev->lock); + + list_for_each_entry(entry, + &vpci_dev->dev_list[PCI_SLOT(devfn)], + list) { + if (PCI_FUNC(entry->dev->devfn) == PCI_FUNC(devfn)) { + dev = entry->dev; + break; + } + } + + mutex_unlock(&vpci_dev->lock); + } + return dev; +} + +static inline int match_slot(struct pci_dev *l, struct pci_dev *r) +{ + if (pci_domain_nr(l->bus) == pci_domain_nr(r->bus) + && l->bus == r->bus && PCI_SLOT(l->devfn) == PCI_SLOT(r->devfn)) + return 1; + + return 0; +} + +static int __xen_pcibk_add_pci_dev(struct xen_pcibk_device *pdev, + struct pci_dev *dev, int devid, + publish_pci_dev_cb publish_cb) +{ + int err = 0, slot, func = -1; + struct pci_dev_entry *t, *dev_entry; + struct vpci_dev_data *vpci_dev = pdev->pci_dev_data; + + if ((dev->class >> 24) == PCI_BASE_CLASS_BRIDGE) { + err = -EFAULT; + xenbus_dev_fatal(pdev->xdev, err, + "Can't export bridges on the virtual PCI bus"); + goto out; + } + + dev_entry = kmalloc(sizeof(*dev_entry), GFP_KERNEL); + if (!dev_entry) { + err = -ENOMEM; + xenbus_dev_fatal(pdev->xdev, err, + "Error adding entry to virtual PCI bus"); + goto out; + } + + dev_entry->dev = dev; + + mutex_lock(&vpci_dev->lock); + + /* Keep multi-function devices together on the virtual PCI bus */ + for (slot = 0; slot < PCI_SLOT_MAX; slot++) { + if (!list_empty(&vpci_dev->dev_list[slot])) { + t = list_entry(list_first(&vpci_dev->dev_list[slot]), + struct pci_dev_entry, list); + + if (match_slot(dev, t->dev)) { + pr_info(DRV_NAME ": vpci: %s: " + "assign to virtual slot %d func %d\n", + pci_name(dev), slot, + PCI_FUNC(dev->devfn)); + list_add_tail(&dev_entry->list, + &vpci_dev->dev_list[slot]); + func = PCI_FUNC(dev->devfn); + goto unlock; + } + } + } + + /* Assign to a new slot on the virtual PCI bus */ + for (slot = 0; slot < PCI_SLOT_MAX; slot++) { + if (list_empty(&vpci_dev->dev_list[slot])) { + printk(KERN_INFO DRV_NAME + ": vpci: %s: assign to virtual slot %d\n", + pci_name(dev), slot); + list_add_tail(&dev_entry->list, + &vpci_dev->dev_list[slot]); + func = PCI_FUNC(dev->devfn); + goto unlock; + } + } + + err = -ENOMEM; + xenbus_dev_fatal(pdev->xdev, err, + "No more space on root virtual PCI bus"); + +unlock: + mutex_unlock(&vpci_dev->lock); + + /* Publish this device. */ + if (!err) + err = publish_cb(pdev, 0, 0, PCI_DEVFN(slot, func), devid); + +out: + return err; +} + +static void __xen_pcibk_release_pci_dev(struct xen_pcibk_device *pdev, + struct pci_dev *dev) +{ + int slot; + struct vpci_dev_data *vpci_dev = pdev->pci_dev_data; + struct pci_dev *found_dev = NULL; + + mutex_lock(&vpci_dev->lock); + + for (slot = 0; slot < PCI_SLOT_MAX; slot++) { + struct pci_dev_entry *e; + + list_for_each_entry(e, &vpci_dev->dev_list[slot], list) { + if (e->dev == dev) { + list_del(&e->list); + found_dev = e->dev; + kfree(e); + goto out; + } + } + } + +out: + mutex_unlock(&vpci_dev->lock); + + if (found_dev) + pcistub_put_pci_dev(found_dev); +} + +static int __xen_pcibk_init_devices(struct xen_pcibk_device *pdev) +{ + int slot; + struct vpci_dev_data *vpci_dev; + + vpci_dev = kmalloc(sizeof(*vpci_dev), GFP_KERNEL); + if (!vpci_dev) + return -ENOMEM; + + mutex_init(&vpci_dev->lock); + + for (slot = 0; slot < PCI_SLOT_MAX; slot++) + INIT_LIST_HEAD(&vpci_dev->dev_list[slot]); + + pdev->pci_dev_data = vpci_dev; + + return 0; +} + +static int __xen_pcibk_publish_pci_roots(struct xen_pcibk_device *pdev, + publish_pci_root_cb publish_cb) +{ + /* The Virtual PCI bus has only one root */ + return publish_cb(pdev, 0, 0); +} + +static void __xen_pcibk_release_devices(struct xen_pcibk_device *pdev) +{ + int slot; + struct vpci_dev_data *vpci_dev = pdev->pci_dev_data; + + for (slot = 0; slot < PCI_SLOT_MAX; slot++) { + struct pci_dev_entry *e, *tmp; + list_for_each_entry_safe(e, tmp, &vpci_dev->dev_list[slot], + list) { + list_del(&e->list); + pcistub_put_pci_dev(e->dev); + kfree(e); + } + } + + kfree(vpci_dev); + pdev->pci_dev_data = NULL; +} + +static int __xen_pcibk_get_pcifront_dev(struct pci_dev *pcidev, + struct xen_pcibk_device *pdev, + unsigned int *domain, unsigned int *bus, + unsigned int *devfn) +{ + struct pci_dev_entry *entry; + struct pci_dev *dev = NULL; + struct vpci_dev_data *vpci_dev = pdev->pci_dev_data; + int found = 0, slot; + + mutex_lock(&vpci_dev->lock); + for (slot = 0; slot < PCI_SLOT_MAX; slot++) { + list_for_each_entry(entry, + &vpci_dev->dev_list[slot], + list) { + dev = entry->dev; + if (dev && dev->bus->number == pcidev->bus->number + && pci_domain_nr(dev->bus) == + pci_domain_nr(pcidev->bus) + && dev->devfn == pcidev->devfn) { + found = 1; + *domain = 0; + *bus = 0; + *devfn = PCI_DEVFN(slot, + PCI_FUNC(pcidev->devfn)); + } + } + } + mutex_unlock(&vpci_dev->lock); + return found; +} + +const struct xen_pcibk_backend xen_pcibk_vpci_backend = { + .name = "vpci", + .init = __xen_pcibk_init_devices, + .free = __xen_pcibk_release_devices, + .find = __xen_pcibk_get_pcifront_dev, + .publish = __xen_pcibk_publish_pci_roots, + .release = __xen_pcibk_release_pci_dev, + .add = __xen_pcibk_add_pci_dev, + .get = __xen_pcibk_get_pci_dev, +}; diff --git a/drivers/xen/xen-pciback/xenbus.c b/drivers/xen/xen-pciback/xenbus.c new file mode 100644 index 0000000..474d52e --- /dev/null +++ b/drivers/xen/xen-pciback/xenbus.c @@ -0,0 +1,743 @@ +/* + * PCI Backend Xenbus Setup - handles setup with frontend and xend + * + * Author: Ryan Wilson <hap9@epoch.ncsc.mil> + */ +#include <linux/module.h> +#include <linux/init.h> +#include <linux/list.h> +#include <linux/vmalloc.h> +#include <linux/workqueue.h> +#include <xen/xenbus.h> +#include <xen/events.h> +#include <asm/xen/pci.h> +#include "pciback.h" + +#define INVALID_EVTCHN_IRQ (-1) +struct workqueue_struct *xen_pcibk_wq; + +static int __read_mostly passthrough; +module_param(passthrough, bool, S_IRUGO); +MODULE_PARM_DESC(passthrough, + "Option to specify how to export PCI topology to guest:\n"\ + " 0 - (default) Hide the true PCI topology and makes the frontend\n"\ + " there is a single PCI bus with only the exported devices on it.\n"\ + " For example, a device at 03:05.0 will be re-assigned to 00:00.0\n"\ + " while second device at 02:1a.1 will be re-assigned to 00:01.1.\n"\ + " 1 - Passthrough provides a real view of the PCI topology to the\n"\ + " frontend (for example, a device at 06:01.b will still appear at\n"\ + " 06:01.b to the frontend). This is similar to how Xen 2.0.x\n"\ + " exposed PCI devices to its driver domains. This may be required\n"\ + " for drivers which depend on finding their hardward in certain\n"\ + " bus/slot locations."); + +static struct xen_pcibk_device *alloc_pdev(struct xenbus_device *xdev) +{ + struct xen_pcibk_device *pdev; + + pdev = kzalloc(sizeof(struct xen_pcibk_device), GFP_KERNEL); + if (pdev == NULL) + goto out; + dev_dbg(&xdev->dev, "allocated pdev @ 0x%p\n", pdev); + + pdev->xdev = xdev; + dev_set_drvdata(&xdev->dev, pdev); + + mutex_init(&pdev->dev_lock); + + pdev->sh_info = NULL; + pdev->evtchn_irq = INVALID_EVTCHN_IRQ; + pdev->be_watching = 0; + + INIT_WORK(&pdev->op_work, xen_pcibk_do_op); + + if (xen_pcibk_init_devices(pdev)) { + kfree(pdev); + pdev = NULL; + } +out: + return pdev; +} + +static void xen_pcibk_disconnect(struct xen_pcibk_device *pdev) +{ + mutex_lock(&pdev->dev_lock); + /* Ensure the guest can't trigger our handler before removing devices */ + if (pdev->evtchn_irq != INVALID_EVTCHN_IRQ) { + unbind_from_irqhandler(pdev->evtchn_irq, pdev); + pdev->evtchn_irq = INVALID_EVTCHN_IRQ; + } + + /* If the driver domain started an op, make sure we complete it + * before releasing the shared memory */ + + /* Note, the workqueue does not use spinlocks at all.*/ + flush_workqueue(xen_pcibk_wq); + + if (pdev->sh_info != NULL) { + xenbus_unmap_ring_vfree(pdev->xdev, pdev->sh_info); + pdev->sh_info = NULL; + } + mutex_unlock(&pdev->dev_lock); +} + +static void free_pdev(struct xen_pcibk_device *pdev) +{ + if (pdev->be_watching) { + unregister_xenbus_watch(&pdev->be_watch); + pdev->be_watching = 0; + } + + xen_pcibk_disconnect(pdev); + + xen_pcibk_release_devices(pdev); + + dev_set_drvdata(&pdev->xdev->dev, NULL); + pdev->xdev = NULL; + + kfree(pdev); +} + +static int xen_pcibk_do_attach(struct xen_pcibk_device *pdev, int gnt_ref, + int remote_evtchn) +{ + int err = 0; + void *vaddr; + + dev_dbg(&pdev->xdev->dev, + "Attaching to frontend resources - gnt_ref=%d evtchn=%d\n", + gnt_ref, remote_evtchn); + + err = xenbus_map_ring_valloc(pdev->xdev, gnt_ref, &vaddr); + if (err < 0) { + xenbus_dev_fatal(pdev->xdev, err, + "Error mapping other domain page in ours."); + goto out; + } + + pdev->sh_info = vaddr; + + err = bind_interdomain_evtchn_to_irqhandler( + pdev->xdev->otherend_id, remote_evtchn, xen_pcibk_handle_event, + 0, DRV_NAME, pdev); + if (err < 0) { + xenbus_dev_fatal(pdev->xdev, err, + "Error binding event channel to IRQ"); + goto out; + } + pdev->evtchn_irq = err; + err = 0; + + dev_dbg(&pdev->xdev->dev, "Attached!\n"); +out: + return err; +} + +static int xen_pcibk_attach(struct xen_pcibk_device *pdev) +{ + int err = 0; + int gnt_ref, remote_evtchn; + char *magic = NULL; + + + mutex_lock(&pdev->dev_lock); + /* Make sure we only do this setup once */ + if (xenbus_read_driver_state(pdev->xdev->nodename) != + XenbusStateInitialised) + goto out; + + /* Wait for frontend to state that it has published the configuration */ + if (xenbus_read_driver_state(pdev->xdev->otherend) != + XenbusStateInitialised) + goto out; + + dev_dbg(&pdev->xdev->dev, "Reading frontend config\n"); + + err = xenbus_gather(XBT_NIL, pdev->xdev->otherend, + "pci-op-ref", "%u", &gnt_ref, + "event-channel", "%u", &remote_evtchn, + "magic", NULL, &magic, NULL); + if (err) { + /* If configuration didn't get read correctly, wait longer */ + xenbus_dev_fatal(pdev->xdev, err, + "Error reading configuration from frontend"); + goto out; + } + + if (magic == NULL || strcmp(magic, XEN_PCI_MAGIC) != 0) { + xenbus_dev_fatal(pdev->xdev, -EFAULT, + "version mismatch (%s/%s) with pcifront - " + "halting " DRV_NAME, + magic, XEN_PCI_MAGIC); + goto out; + } + + err = xen_pcibk_do_attach(pdev, gnt_ref, remote_evtchn); + if (err) + goto out; + + dev_dbg(&pdev->xdev->dev, "Connecting...\n"); + + err = xenbus_switch_state(pdev->xdev, XenbusStateConnected); + if (err) + xenbus_dev_fatal(pdev->xdev, err, + "Error switching to connected state!"); + + dev_dbg(&pdev->xdev->dev, "Connected? %d\n", err); +out: + mutex_unlock(&pdev->dev_lock); + + kfree(magic); + + return err; +} + +static int xen_pcibk_publish_pci_dev(struct xen_pcibk_device *pdev, + unsigned int domain, unsigned int bus, + unsigned int devfn, unsigned int devid) +{ + int err; + int len; + char str[64]; + + len = snprintf(str, sizeof(str), "vdev-%d", devid); + if (unlikely(len >= (sizeof(str) - 1))) { + err = -ENOMEM; + goto out; + } + + err = xenbus_printf(XBT_NIL, pdev->xdev->nodename, str, + "%04x:%02x:%02x.%02x", domain, bus, + PCI_SLOT(devfn), PCI_FUNC(devfn)); + +out: + return err; +} + +static int xen_pcibk_export_device(struct xen_pcibk_device *pdev, + int domain, int bus, int slot, int func, + int devid) +{ + struct pci_dev *dev; + int err = 0; + + dev_dbg(&pdev->xdev->dev, "exporting dom %x bus %x slot %x func %x\n", + domain, bus, slot, func); + + dev = pcistub_get_pci_dev_by_slot(pdev, domain, bus, slot, func); + if (!dev) { + err = -EINVAL; + xenbus_dev_fatal(pdev->xdev, err, + "Couldn't locate PCI device " + "(%04x:%02x:%02x.%01x)! " + "perhaps already in-use?", + domain, bus, slot, func); + goto out; + } + + err = xen_pcibk_add_pci_dev(pdev, dev, devid, + xen_pcibk_publish_pci_dev); + if (err) + goto out; + + dev_dbg(&dev->dev, "registering for %d\n", pdev->xdev->otherend_id); + if (xen_register_device_domain_owner(dev, + pdev->xdev->otherend_id) != 0) { + dev_err(&dev->dev, "device has been assigned to another " \ + "domain! Over-writting the ownership, but beware.\n"); + xen_unregister_device_domain_owner(dev); + xen_register_device_domain_owner(dev, pdev->xdev->otherend_id); + } + + /* TODO: It'd be nice to export a bridge and have all of its children + * get exported with it. This may be best done in xend (which will + * have to calculate resource usage anyway) but we probably want to + * put something in here to ensure that if a bridge gets given to a + * driver domain, that all devices under that bridge are not given + * to other driver domains (as he who controls the bridge can disable + * it and stop the other devices from working). + */ +out: + return err; +} + +static int xen_pcibk_remove_device(struct xen_pcibk_device *pdev, + int domain, int bus, int slot, int func) +{ + int err = 0; + struct pci_dev *dev; + + dev_dbg(&pdev->xdev->dev, "removing dom %x bus %x slot %x func %x\n", + domain, bus, slot, func); + + dev = xen_pcibk_get_pci_dev(pdev, domain, bus, PCI_DEVFN(slot, func)); + if (!dev) { + err = -EINVAL; + dev_dbg(&pdev->xdev->dev, "Couldn't locate PCI device " + "(%04x:%02x:%02x.%01x)! not owned by this domain\n", + domain, bus, slot, func); + goto out; + } + + dev_dbg(&dev->dev, "unregistering for %d\n", pdev->xdev->otherend_id); + xen_unregister_device_domain_owner(dev); + + xen_pcibk_release_pci_dev(pdev, dev); + +out: + return err; +} + +static int xen_pcibk_publish_pci_root(struct xen_pcibk_device *pdev, + unsigned int domain, unsigned int bus) +{ + unsigned int d, b; + int i, root_num, len, err; + char str[64]; + + dev_dbg(&pdev->xdev->dev, "Publishing pci roots\n"); + + err = xenbus_scanf(XBT_NIL, pdev->xdev->nodename, + "root_num", "%d", &root_num); + if (err == 0 || err == -ENOENT) + root_num = 0; + else if (err < 0) + goto out; + + /* Verify that we haven't already published this pci root */ + for (i = 0; i < root_num; i++) { + len = snprintf(str, sizeof(str), "root-%d", i); + if (unlikely(len >= (sizeof(str) - 1))) { + err = -ENOMEM; + goto out; + } + + err = xenbus_scanf(XBT_NIL, pdev->xdev->nodename, + str, "%x:%x", &d, &b); + if (err < 0) + goto out; + if (err != 2) { + err = -EINVAL; + goto out; + } + + if (d == domain && b == bus) { + err = 0; + goto out; + } + } + + len = snprintf(str, sizeof(str), "root-%d", root_num); + if (unlikely(len >= (sizeof(str) - 1))) { + err = -ENOMEM; + goto out; + } + + dev_dbg(&pdev->xdev->dev, "writing root %d at %04x:%02x\n", + root_num, domain, bus); + + err = xenbus_printf(XBT_NIL, pdev->xdev->nodename, str, + "%04x:%02x", domain, bus); + if (err) + goto out; + + err = xenbus_printf(XBT_NIL, pdev->xdev->nodename, + "root_num", "%d", (root_num + 1)); + +out: + return err; +} + +static int xen_pcibk_reconfigure(struct xen_pcibk_device *pdev) +{ + int err = 0; + int num_devs; + int domain, bus, slot, func; + int substate; + int i, len; + char state_str[64]; + char dev_str[64]; + + + dev_dbg(&pdev->xdev->dev, "Reconfiguring device ...\n"); + + mutex_lock(&pdev->dev_lock); + /* Make sure we only reconfigure once */ + if (xenbus_read_driver_state(pdev->xdev->nodename) != + XenbusStateReconfiguring) + goto out; + + err = xenbus_scanf(XBT_NIL, pdev->xdev->nodename, "num_devs", "%d", + &num_devs); + if (err != 1) { + if (err >= 0) + err = -EINVAL; + xenbus_dev_fatal(pdev->xdev, err, + "Error reading number of devices"); + goto out; + } + + for (i = 0; i < num_devs; i++) { + len = snprintf(state_str, sizeof(state_str), "state-%d", i); + if (unlikely(len >= (sizeof(state_str) - 1))) { + err = -ENOMEM; + xenbus_dev_fatal(pdev->xdev, err, + "String overflow while reading " + "configuration"); + goto out; + } + err = xenbus_scanf(XBT_NIL, pdev->xdev->nodename, state_str, + "%d", &substate); + if (err != 1) + substate = XenbusStateUnknown; + + switch (substate) { + case XenbusStateInitialising: + dev_dbg(&pdev->xdev->dev, "Attaching dev-%d ...\n", i); + + len = snprintf(dev_str, sizeof(dev_str), "dev-%d", i); + if (unlikely(len >= (sizeof(dev_str) - 1))) { + err = -ENOMEM; + xenbus_dev_fatal(pdev->xdev, err, + "String overflow while " + "reading configuration"); + goto out; + } + err = xenbus_scanf(XBT_NIL, pdev->xdev->nodename, + dev_str, "%x:%x:%x.%x", + &domain, &bus, &slot, &func); + if (err < 0) { + xenbus_dev_fatal(pdev->xdev, err, + "Error reading device " + "configuration"); + goto out; + } + if (err != 4) { + err = -EINVAL; + xenbus_dev_fatal(pdev->xdev, err, + "Error parsing pci device " + "configuration"); + goto out; + } + + err = xen_pcibk_export_device(pdev, domain, bus, slot, + func, i); + if (err) + goto out; + + /* Publish pci roots. */ + err = xen_pcibk_publish_pci_roots(pdev, + xen_pcibk_publish_pci_root); + if (err) { + xenbus_dev_fatal(pdev->xdev, err, + "Error while publish PCI root" + "buses for frontend"); + goto out; + } + + err = xenbus_printf(XBT_NIL, pdev->xdev->nodename, + state_str, "%d", + XenbusStateInitialised); + if (err) { + xenbus_dev_fatal(pdev->xdev, err, + "Error switching substate of " + "dev-%d\n", i); + goto out; + } + break; + + case XenbusStateClosing: + dev_dbg(&pdev->xdev->dev, "Detaching dev-%d ...\n", i); + + len = snprintf(dev_str, sizeof(dev_str), "vdev-%d", i); + if (unlikely(len >= (sizeof(dev_str) - 1))) { + err = -ENOMEM; + xenbus_dev_fatal(pdev->xdev, err, + "String overflow while " + "reading configuration"); + goto out; + } + err = xenbus_scanf(XBT_NIL, pdev->xdev->nodename, + dev_str, "%x:%x:%x.%x", + &domain, &bus, &slot, &func); + if (err < 0) { + xenbus_dev_fatal(pdev->xdev, err, + "Error reading device " + "configuration"); + goto out; + } + if (err != 4) { + err = -EINVAL; + xenbus_dev_fatal(pdev->xdev, err, + "Error parsing pci device " + "configuration"); + goto out; + } + + err = xen_pcibk_remove_device(pdev, domain, bus, slot, + func); + if (err) + goto out; + + /* TODO: If at some point we implement support for pci + * root hot-remove on pcifront side, we'll need to + * remove unnecessary xenstore nodes of pci roots here. + */ + + break; + + default: + break; + } + } + + err = xenbus_switch_state(pdev->xdev, XenbusStateReconfigured); + if (err) { + xenbus_dev_fatal(pdev->xdev, err, + "Error switching to reconfigured state!"); + goto out; + } + +out: + mutex_unlock(&pdev->dev_lock); + return 0; +} + +static void xen_pcibk_frontend_changed(struct xenbus_device *xdev, + enum xenbus_state fe_state) +{ + struct xen_pcibk_device *pdev = dev_get_drvdata(&xdev->dev); + + dev_dbg(&xdev->dev, "fe state changed %d\n", fe_state); + + switch (fe_state) { + case XenbusStateInitialised: + xen_pcibk_attach(pdev); + break; + + case XenbusStateReconfiguring: + xen_pcibk_reconfigure(pdev); + break; + + case XenbusStateConnected: + /* pcifront switched its state from reconfiguring to connected. + * Then switch to connected state. + */ + xenbus_switch_state(xdev, XenbusStateConnected); + break; + + case XenbusStateClosing: + xen_pcibk_disconnect(pdev); + xenbus_switch_state(xdev, XenbusStateClosing); + break; + + case XenbusStateClosed: + xen_pcibk_disconnect(pdev); + xenbus_switch_state(xdev, XenbusStateClosed); + if (xenbus_dev_is_online(xdev)) + break; + /* fall through if not online */ + case XenbusStateUnknown: + dev_dbg(&xdev->dev, "frontend is gone! unregister device\n"); + device_unregister(&xdev->dev); + break; + + default: + break; + } +} + +static int xen_pcibk_setup_backend(struct xen_pcibk_device *pdev) +{ + /* Get configuration from xend (if available now) */ + int domain, bus, slot, func; + int err = 0; + int i, num_devs; + char dev_str[64]; + char state_str[64]; + + mutex_lock(&pdev->dev_lock); + /* It's possible we could get the call to setup twice, so make sure + * we're not already connected. + */ + if (xenbus_read_driver_state(pdev->xdev->nodename) != + XenbusStateInitWait) + goto out; + + dev_dbg(&pdev->xdev->dev, "getting be setup\n"); + + err = xenbus_scanf(XBT_NIL, pdev->xdev->nodename, "num_devs", "%d", + &num_devs); + if (err != 1) { + if (err >= 0) + err = -EINVAL; + xenbus_dev_fatal(pdev->xdev, err, + "Error reading number of devices"); + goto out; + } + + for (i = 0; i < num_devs; i++) { + int l = snprintf(dev_str, sizeof(dev_str), "dev-%d", i); + if (unlikely(l >= (sizeof(dev_str) - 1))) { + err = -ENOMEM; + xenbus_dev_fatal(pdev->xdev, err, + "String overflow while reading " + "configuration"); + goto out; + } + + err = xenbus_scanf(XBT_NIL, pdev->xdev->nodename, dev_str, + "%x:%x:%x.%x", &domain, &bus, &slot, &func); + if (err < 0) { + xenbus_dev_fatal(pdev->xdev, err, + "Error reading device configuration"); + goto out; + } + if (err != 4) { + err = -EINVAL; + xenbus_dev_fatal(pdev->xdev, err, + "Error parsing pci device " + "configuration"); + goto out; + } + + err = xen_pcibk_export_device(pdev, domain, bus, slot, func, i); + if (err) + goto out; + + /* Switch substate of this device. */ + l = snprintf(state_str, sizeof(state_str), "state-%d", i); + if (unlikely(l >= (sizeof(state_str) - 1))) { + err = -ENOMEM; + xenbus_dev_fatal(pdev->xdev, err, + "String overflow while reading " + "configuration"); + goto out; + } + err = xenbus_printf(XBT_NIL, pdev->xdev->nodename, state_str, + "%d", XenbusStateInitialised); + if (err) { + xenbus_dev_fatal(pdev->xdev, err, "Error switching " + "substate of dev-%d\n", i); + goto out; + } + } + + err = xen_pcibk_publish_pci_roots(pdev, xen_pcibk_publish_pci_root); + if (err) { + xenbus_dev_fatal(pdev->xdev, err, + "Error while publish PCI root buses " + "for frontend"); + goto out; + } + + err = xenbus_switch_state(pdev->xdev, XenbusStateInitialised); + if (err) + xenbus_dev_fatal(pdev->xdev, err, + "Error switching to initialised state!"); + +out: + mutex_unlock(&pdev->dev_lock); + if (!err) + /* see if pcifront is already configured (if not, we'll wait) */ + xen_pcibk_attach(pdev); + return err; +} + +static void xen_pcibk_be_watch(struct xenbus_watch *watch, + const char **vec, unsigned int len) +{ + struct xen_pcibk_device *pdev = + container_of(watch, struct xen_pcibk_device, be_watch); + + switch (xenbus_read_driver_state(pdev->xdev->nodename)) { + case XenbusStateInitWait: + xen_pcibk_setup_backend(pdev); + break; + + default: + break; + } +} + +static int xen_pcibk_xenbus_probe(struct xenbus_device *dev, + const struct xenbus_device_id *id) +{ + int err = 0; + struct xen_pcibk_device *pdev = alloc_pdev(dev); + + if (pdev == NULL) { + err = -ENOMEM; + xenbus_dev_fatal(dev, err, + "Error allocating xen_pcibk_device struct"); + goto out; + } + + /* wait for xend to configure us */ + err = xenbus_switch_state(dev, XenbusStateInitWait); + if (err) + goto out; + + /* watch the backend node for backend configuration information */ + err = xenbus_watch_path(dev, dev->nodename, &pdev->be_watch, + xen_pcibk_be_watch); + if (err) + goto out; + + pdev->be_watching = 1; + + /* We need to force a call to our callback here in case + * xend already configured us! + */ + xen_pcibk_be_watch(&pdev->be_watch, NULL, 0); + +out: + return err; +} + +static int xen_pcibk_xenbus_remove(struct xenbus_device *dev) +{ + struct xen_pcibk_device *pdev = dev_get_drvdata(&dev->dev); + + if (pdev != NULL) + free_pdev(pdev); + + return 0; +} + +static const struct xenbus_device_id xenpci_ids[] = { + {"pci"}, + {""}, +}; + +static struct xenbus_driver xenbus_xen_pcibk_driver = { + .name = DRV_NAME, + .owner = THIS_MODULE, + .ids = xenpci_ids, + .probe = xen_pcibk_xenbus_probe, + .remove = xen_pcibk_xenbus_remove, + .otherend_changed = xen_pcibk_frontend_changed, +}; + +const struct xen_pcibk_backend *__read_mostly xen_pcibk_backend; + +int __init xen_pcibk_xenbus_register(void) +{ + xen_pcibk_wq = create_workqueue("xen_pciback_workqueue"); + if (!xen_pcibk_wq) { + printk(KERN_ERR "%s: create" + "xen_pciback_workqueue failed\n", __func__); + return -EFAULT; + } + xen_pcibk_backend = &xen_pcibk_vpci_backend; + if (passthrough) + xen_pcibk_backend = &xen_pcibk_passthrough_backend; + pr_info(DRV_NAME ": backend is %s\n", xen_pcibk_backend->name); + return xenbus_register_backend(&xenbus_xen_pcibk_driver); +} + +void __exit xen_pcibk_xenbus_unregister(void) +{ + destroy_workqueue(xen_pcibk_wq); + xenbus_unregister_driver(&xenbus_xen_pcibk_driver); +} diff --git a/drivers/xen/xen-selfballoon.c b/drivers/xen/xen-selfballoon.c new file mode 100644 index 0000000..d93c708 --- /dev/null +++ b/drivers/xen/xen-selfballoon.c @@ -0,0 +1,545 @@ +/****************************************************************************** + * Xen selfballoon driver (and optional frontswap self-shrinking driver) + * + * Copyright (c) 2009-2011, Dan Magenheimer, Oracle Corp. + * + * This code complements the cleancache and frontswap patchsets to optimize + * support for Xen Transcendent Memory ("tmem"). The policy it implements + * is rudimentary and will likely improve over time, but it does work well + * enough today. + * + * Two functionalities are implemented here which both use "control theory" + * (feedback) to optimize memory utilization. In a virtualized environment + * such as Xen, RAM is often a scarce resource and we would like to ensure + * that each of a possibly large number of virtual machines is using RAM + * efficiently, i.e. using as little as possible when under light load + * and obtaining as much as possible when memory demands are high. + * Since RAM needs vary highly dynamically and sometimes dramatically, + * "hysteresis" is used, that is, memory target is determined not just + * on current data but also on past data stored in the system. + * + * "Selfballooning" creates memory pressure by managing the Xen balloon + * driver to decrease and increase available kernel memory, driven + * largely by the target value of "Committed_AS" (see /proc/meminfo). + * Since Committed_AS does not account for clean mapped pages (i.e. pages + * in RAM that are identical to pages on disk), selfballooning has the + * affect of pushing less frequently used clean pagecache pages out of + * kernel RAM and, presumably using cleancache, into Xen tmem where + * Xen can more efficiently optimize RAM utilization for such pages. + * + * When kernel memory demand unexpectedly increases faster than Xen, via + * the selfballoon driver, is able to (or chooses to) provide usable RAM, + * the kernel may invoke swapping. In most cases, frontswap is able + * to absorb this swapping into Xen tmem. However, due to the fact + * that the kernel swap subsystem assumes swapping occurs to a disk, + * swapped pages may sit on the disk for a very long time; even if + * the kernel knows the page will never be used again. This is because + * the disk space costs very little and can be overwritten when + * necessary. When such stale pages are in frontswap, however, they + * are taking up valuable real estate. "Frontswap selfshrinking" works + * to resolve this: When frontswap activity is otherwise stable + * and the guest kernel is not under memory pressure, the "frontswap + * selfshrinking" accounts for this by providing pressure to remove some + * pages from frontswap and return them to kernel memory. + * + * For both "selfballooning" and "frontswap-selfshrinking", a worker + * thread is used and sysfs tunables are provided to adjust the frequency + * and rate of adjustments to achieve the goal, as well as to disable one + * or both functions independently. + * + * While some argue that this functionality can and should be implemented + * in userspace, it has been observed that bad things happen (e.g. OOMs). + * + * System configuration note: Selfballooning should not be enabled on + * systems without a sufficiently large swap device configured; for best + * results, it is recommended that total swap be increased by the size + * of the guest memory. Also, while technically not required to be + * configured, it is highly recommended that frontswap also be configured + * and enabled when selfballooning is running. So, selfballooning + * is disabled by default if frontswap is not configured and can only + * be enabled with the "selfballooning" kernel boot option; similarly + * selfballooning is enabled by default if frontswap is configured and + * can be disabled with the "noselfballooning" kernel boot option. Finally, + * when frontswap is configured, frontswap-selfshrinking can be disabled + * with the "noselfshrink" kernel boot option. + * + * Selfballooning is disallowed in domain0 and force-disabled. + * + */ + +#include <linux/kernel.h> +#include <linux/bootmem.h> +#include <linux/swap.h> +#include <linux/mm.h> +#include <linux/mman.h> +#include <linux/module.h> +#include <linux/workqueue.h> +#include <xen/balloon.h> +#include <xen/tmem.h> +#include <xen/xen.h> + +/* Enable/disable with sysfs. */ +static int xen_selfballooning_enabled __read_mostly; + +/* + * Controls rate at which memory target (this iteration) approaches + * ultimate goal when memory need is increasing (up-hysteresis) or + * decreasing (down-hysteresis). Higher values of hysteresis cause + * slower increases/decreases. The default values for the various + * parameters were deemed reasonable by experimentation, may be + * workload-dependent, and can all be adjusted via sysfs. + */ +static unsigned int selfballoon_downhysteresis __read_mostly = 8; +static unsigned int selfballoon_uphysteresis __read_mostly = 1; + +/* In HZ, controls frequency of worker invocation. */ +static unsigned int selfballoon_interval __read_mostly = 5; + +/* + * Minimum usable RAM in MB for selfballooning target for balloon. + * If non-zero, it is added to totalreserve_pages and self-ballooning + * will not balloon below the sum. If zero, a piecewise linear function + * is calculated as a minimum and added to totalreserve_pages. Note that + * setting this value indiscriminately may cause OOMs and crashes. + */ +static unsigned int selfballoon_min_usable_mb; + +static void selfballoon_process(struct work_struct *work); +static DECLARE_DELAYED_WORK(selfballoon_worker, selfballoon_process); + +#ifdef CONFIG_FRONTSWAP +#include <linux/frontswap.h> + +/* Enable/disable with sysfs. */ +static bool frontswap_selfshrinking __read_mostly; + +/* Enable/disable with kernel boot option. */ +static bool use_frontswap_selfshrink __initdata = true; + +/* + * The default values for the following parameters were deemed reasonable + * by experimentation, may be workload-dependent, and can all be + * adjusted via sysfs. + */ + +/* Control rate for frontswap shrinking. Higher hysteresis is slower. */ +static unsigned int frontswap_hysteresis __read_mostly = 20; + +/* + * Number of selfballoon worker invocations to wait before observing that + * frontswap selfshrinking should commence. Note that selfshrinking does + * not use a separate worker thread. + */ +static unsigned int frontswap_inertia __read_mostly = 3; + +/* Countdown to next invocation of frontswap_shrink() */ +static unsigned long frontswap_inertia_counter; + +/* + * Invoked by the selfballoon worker thread, uses current number of pages + * in frontswap (frontswap_curr_pages()), previous status, and control + * values (hysteresis and inertia) to determine if frontswap should be + * shrunk and what the new frontswap size should be. Note that + * frontswap_shrink is essentially a partial swapoff that immediately + * transfers pages from the "swap device" (frontswap) back into kernel + * RAM; despite the name, frontswap "shrinking" is very different from + * the "shrinker" interface used by the kernel MM subsystem to reclaim + * memory. + */ +static void frontswap_selfshrink(void) +{ + static unsigned long cur_frontswap_pages; + static unsigned long last_frontswap_pages; + static unsigned long tgt_frontswap_pages; + + last_frontswap_pages = cur_frontswap_pages; + cur_frontswap_pages = frontswap_curr_pages(); + if (!cur_frontswap_pages || + (cur_frontswap_pages > last_frontswap_pages)) { + frontswap_inertia_counter = frontswap_inertia; + return; + } + if (frontswap_inertia_counter && --frontswap_inertia_counter) + return; + if (cur_frontswap_pages <= frontswap_hysteresis) + tgt_frontswap_pages = 0; + else + tgt_frontswap_pages = cur_frontswap_pages - + (cur_frontswap_pages / frontswap_hysteresis); + frontswap_shrink(tgt_frontswap_pages); +} + +static int __init xen_nofrontswap_selfshrink_setup(char *s) +{ + use_frontswap_selfshrink = false; + return 1; +} + +__setup("noselfshrink", xen_nofrontswap_selfshrink_setup); + +/* Disable with kernel boot option. */ +static bool use_selfballooning __initdata = true; + +static int __init xen_noselfballooning_setup(char *s) +{ + use_selfballooning = false; + return 1; +} + +__setup("noselfballooning", xen_noselfballooning_setup); +#else /* !CONFIG_FRONTSWAP */ +/* Enable with kernel boot option. */ +static bool use_selfballooning __initdata = false; + +static int __init xen_selfballooning_setup(char *s) +{ + use_selfballooning = true; + return 1; +} + +__setup("selfballooning", xen_selfballooning_setup); +#endif /* CONFIG_FRONTSWAP */ + +#define MB2PAGES(mb) ((mb) << (20 - PAGE_SHIFT)) + +/* + * Use current balloon size, the goal (vm_committed_as), and hysteresis + * parameters to set a new target balloon size + */ +static void selfballoon_process(struct work_struct *work) +{ + unsigned long cur_pages, goal_pages, tgt_pages, floor_pages; + unsigned long useful_pages; + bool reset_timer = false; + + if (xen_selfballooning_enabled) { + cur_pages = totalram_pages; + tgt_pages = cur_pages; /* default is no change */ + goal_pages = percpu_counter_read_positive(&vm_committed_as) + + totalreserve_pages; +#ifdef CONFIG_FRONTSWAP + /* allow space for frontswap pages to be repatriated */ + if (frontswap_selfshrinking && frontswap_enabled) + goal_pages += frontswap_curr_pages(); +#endif + if (cur_pages > goal_pages) + tgt_pages = cur_pages - + ((cur_pages - goal_pages) / + selfballoon_downhysteresis); + else if (cur_pages < goal_pages) + tgt_pages = cur_pages + + ((goal_pages - cur_pages) / + selfballoon_uphysteresis); + /* else if cur_pages == goal_pages, no change */ + useful_pages = max_pfn - totalreserve_pages; + if (selfballoon_min_usable_mb != 0) + floor_pages = totalreserve_pages + + MB2PAGES(selfballoon_min_usable_mb); + /* piecewise linear function ending in ~3% slope */ + else if (useful_pages < MB2PAGES(16)) + floor_pages = max_pfn; /* not worth ballooning */ + else if (useful_pages < MB2PAGES(64)) + floor_pages = totalreserve_pages + MB2PAGES(16) + + ((useful_pages - MB2PAGES(16)) >> 1); + else if (useful_pages < MB2PAGES(512)) + floor_pages = totalreserve_pages + MB2PAGES(40) + + ((useful_pages - MB2PAGES(40)) >> 3); + else /* useful_pages >= MB2PAGES(512) */ + floor_pages = totalreserve_pages + MB2PAGES(99) + + ((useful_pages - MB2PAGES(99)) >> 5); + if (tgt_pages < floor_pages) + tgt_pages = floor_pages; + balloon_set_new_target(tgt_pages + + balloon_stats.current_pages - totalram_pages); + reset_timer = true; + } +#ifdef CONFIG_FRONTSWAP + if (frontswap_selfshrinking && frontswap_enabled) { + frontswap_selfshrink(); + reset_timer = true; + } +#endif + if (reset_timer) + schedule_delayed_work(&selfballoon_worker, + selfballoon_interval * HZ); +} + +#ifdef CONFIG_SYSFS + +#include <linux/sysdev.h> +#include <linux/capability.h> + +#define SELFBALLOON_SHOW(name, format, args...) \ + static ssize_t show_##name(struct sys_device *dev, \ + struct sysdev_attribute *attr, \ + char *buf) \ + { \ + return sprintf(buf, format, ##args); \ + } + +SELFBALLOON_SHOW(selfballooning, "%d\n", xen_selfballooning_enabled); + +static ssize_t store_selfballooning(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + bool was_enabled = xen_selfballooning_enabled; + unsigned long tmp; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + + err = strict_strtoul(buf, 10, &tmp); + if (err || ((tmp != 0) && (tmp != 1))) + return -EINVAL; + + xen_selfballooning_enabled = !!tmp; + if (!was_enabled && xen_selfballooning_enabled) + schedule_delayed_work(&selfballoon_worker, + selfballoon_interval * HZ); + + return count; +} + +static SYSDEV_ATTR(selfballooning, S_IRUGO | S_IWUSR, + show_selfballooning, store_selfballooning); + +SELFBALLOON_SHOW(selfballoon_interval, "%d\n", selfballoon_interval); + +static ssize_t store_selfballoon_interval(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + unsigned long val; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + err = strict_strtoul(buf, 10, &val); + if (err || val == 0) + return -EINVAL; + selfballoon_interval = val; + return count; +} + +static SYSDEV_ATTR(selfballoon_interval, S_IRUGO | S_IWUSR, + show_selfballoon_interval, store_selfballoon_interval); + +SELFBALLOON_SHOW(selfballoon_downhys, "%d\n", selfballoon_downhysteresis); + +static ssize_t store_selfballoon_downhys(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + unsigned long val; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + err = strict_strtoul(buf, 10, &val); + if (err || val == 0) + return -EINVAL; + selfballoon_downhysteresis = val; + return count; +} + +static SYSDEV_ATTR(selfballoon_downhysteresis, S_IRUGO | S_IWUSR, + show_selfballoon_downhys, store_selfballoon_downhys); + + +SELFBALLOON_SHOW(selfballoon_uphys, "%d\n", selfballoon_uphysteresis); + +static ssize_t store_selfballoon_uphys(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + unsigned long val; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + err = strict_strtoul(buf, 10, &val); + if (err || val == 0) + return -EINVAL; + selfballoon_uphysteresis = val; + return count; +} + +static SYSDEV_ATTR(selfballoon_uphysteresis, S_IRUGO | S_IWUSR, + show_selfballoon_uphys, store_selfballoon_uphys); + +SELFBALLOON_SHOW(selfballoon_min_usable_mb, "%d\n", + selfballoon_min_usable_mb); + +static ssize_t store_selfballoon_min_usable_mb(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + unsigned long val; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + err = strict_strtoul(buf, 10, &val); + if (err || val == 0) + return -EINVAL; + selfballoon_min_usable_mb = val; + return count; +} + +static SYSDEV_ATTR(selfballoon_min_usable_mb, S_IRUGO | S_IWUSR, + show_selfballoon_min_usable_mb, + store_selfballoon_min_usable_mb); + + +#ifdef CONFIG_FRONTSWAP +SELFBALLOON_SHOW(frontswap_selfshrinking, "%d\n", frontswap_selfshrinking); + +static ssize_t store_frontswap_selfshrinking(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + bool was_enabled = frontswap_selfshrinking; + unsigned long tmp; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + err = strict_strtoul(buf, 10, &tmp); + if (err || ((tmp != 0) && (tmp != 1))) + return -EINVAL; + frontswap_selfshrinking = !!tmp; + if (!was_enabled && !xen_selfballooning_enabled && + frontswap_selfshrinking) + schedule_delayed_work(&selfballoon_worker, + selfballoon_interval * HZ); + + return count; +} + +static SYSDEV_ATTR(frontswap_selfshrinking, S_IRUGO | S_IWUSR, + show_frontswap_selfshrinking, store_frontswap_selfshrinking); + +SELFBALLOON_SHOW(frontswap_inertia, "%d\n", frontswap_inertia); + +static ssize_t store_frontswap_inertia(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + unsigned long val; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + err = strict_strtoul(buf, 10, &val); + if (err || val == 0) + return -EINVAL; + frontswap_inertia = val; + frontswap_inertia_counter = val; + return count; +} + +static SYSDEV_ATTR(frontswap_inertia, S_IRUGO | S_IWUSR, + show_frontswap_inertia, store_frontswap_inertia); + +SELFBALLOON_SHOW(frontswap_hysteresis, "%d\n", frontswap_hysteresis); + +static ssize_t store_frontswap_hysteresis(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + unsigned long val; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + err = strict_strtoul(buf, 10, &val); + if (err || val == 0) + return -EINVAL; + frontswap_hysteresis = val; + return count; +} + +static SYSDEV_ATTR(frontswap_hysteresis, S_IRUGO | S_IWUSR, + show_frontswap_hysteresis, store_frontswap_hysteresis); + +#endif /* CONFIG_FRONTSWAP */ + +static struct attribute *selfballoon_attrs[] = { + &attr_selfballooning.attr, + &attr_selfballoon_interval.attr, + &attr_selfballoon_downhysteresis.attr, + &attr_selfballoon_uphysteresis.attr, + &attr_selfballoon_min_usable_mb.attr, +#ifdef CONFIG_FRONTSWAP + &attr_frontswap_selfshrinking.attr, + &attr_frontswap_hysteresis.attr, + &attr_frontswap_inertia.attr, +#endif + NULL +}; + +static struct attribute_group selfballoon_group = { + .name = "selfballoon", + .attrs = selfballoon_attrs +}; +#endif + +int register_xen_selfballooning(struct sys_device *sysdev) +{ + int error = -1; + +#ifdef CONFIG_SYSFS + error = sysfs_create_group(&sysdev->kobj, &selfballoon_group); +#endif + return error; +} +EXPORT_SYMBOL(register_xen_selfballooning); + +static int __init xen_selfballoon_init(void) +{ + bool enable = false; + + if (!xen_domain()) + return -ENODEV; + + if (xen_initial_domain()) { + pr_info("xen/balloon: Xen selfballooning driver " + "disabled for domain0.\n"); + return -ENODEV; + } + + xen_selfballooning_enabled = tmem_enabled && use_selfballooning; + if (xen_selfballooning_enabled) { + pr_info("xen/balloon: Initializing Xen " + "selfballooning driver.\n"); + enable = true; + } +#ifdef CONFIG_FRONTSWAP + frontswap_selfshrinking = tmem_enabled && use_frontswap_selfshrink; + if (frontswap_selfshrinking) { + pr_info("xen/balloon: Initializing frontswap " + "selfshrinking driver.\n"); + enable = true; + } +#endif + if (!enable) + return -ENODEV; + + schedule_delayed_work(&selfballoon_worker, selfballoon_interval * HZ); + + return 0; +} + +subsys_initcall(xen_selfballoon_init); + +MODULE_LICENSE("GPL"); diff --git a/drivers/xen/xenbus/xenbus_comms.c b/drivers/xen/xenbus/xenbus_comms.c index 090c61e..2eff7a6 100644 --- a/drivers/xen/xenbus/xenbus_comms.c +++ b/drivers/xen/xenbus/xenbus_comms.c @@ -212,7 +212,9 @@ int xb_init_comms(void) printk(KERN_WARNING "XENBUS response ring is not quiescent " "(%08x:%08x): fixing up\n", intf->rsp_cons, intf->rsp_prod); - intf->rsp_cons = intf->rsp_prod; + /* breaks kdump */ + if (!reset_devices) + intf->rsp_cons = intf->rsp_prod; } if (xenbus_irq) { diff --git a/drivers/xen/xenbus/xenbus_probe.c b/drivers/xen/xenbus/xenbus_probe.c index 7397695..cef9b0b 100644 --- a/drivers/xen/xenbus/xenbus_probe.c +++ b/drivers/xen/xenbus/xenbus_probe.c @@ -378,26 +378,32 @@ static void xenbus_dev_release(struct device *dev) kfree(to_xenbus_device(dev)); } -static ssize_t xendev_show_nodename(struct device *dev, - struct device_attribute *attr, char *buf) +static ssize_t nodename_show(struct device *dev, + struct device_attribute *attr, char *buf) { return sprintf(buf, "%s\n", to_xenbus_device(dev)->nodename); } -static DEVICE_ATTR(nodename, S_IRUSR | S_IRGRP | S_IROTH, xendev_show_nodename, NULL); -static ssize_t xendev_show_devtype(struct device *dev, - struct device_attribute *attr, char *buf) +static ssize_t devtype_show(struct device *dev, + struct device_attribute *attr, char *buf) { return sprintf(buf, "%s\n", to_xenbus_device(dev)->devicetype); } -static DEVICE_ATTR(devtype, S_IRUSR | S_IRGRP | S_IROTH, xendev_show_devtype, NULL); -static ssize_t xendev_show_modalias(struct device *dev, - struct device_attribute *attr, char *buf) +static ssize_t modalias_show(struct device *dev, + struct device_attribute *attr, char *buf) { - return sprintf(buf, "xen:%s\n", to_xenbus_device(dev)->devicetype); + return sprintf(buf, "%s:%s\n", dev->bus->name, + to_xenbus_device(dev)->devicetype); } -static DEVICE_ATTR(modalias, S_IRUSR | S_IRGRP | S_IROTH, xendev_show_modalias, NULL); + +struct device_attribute xenbus_dev_attrs[] = { + __ATTR_RO(nodename), + __ATTR_RO(devtype), + __ATTR_RO(modalias), + __ATTR_NULL +}; +EXPORT_SYMBOL_GPL(xenbus_dev_attrs); int xenbus_probe_node(struct xen_bus_type *bus, const char *type, @@ -449,25 +455,7 @@ int xenbus_probe_node(struct xen_bus_type *bus, if (err) goto fail; - err = device_create_file(&xendev->dev, &dev_attr_nodename); - if (err) - goto fail_unregister; - - err = device_create_file(&xendev->dev, &dev_attr_devtype); - if (err) - goto fail_remove_nodename; - - err = device_create_file(&xendev->dev, &dev_attr_modalias); - if (err) - goto fail_remove_devtype; - return 0; -fail_remove_devtype: - device_remove_file(&xendev->dev, &dev_attr_devtype); -fail_remove_nodename: - device_remove_file(&xendev->dev, &dev_attr_nodename); -fail_unregister: - device_unregister(&xendev->dev); fail: kfree(xendev); return err; @@ -696,64 +684,74 @@ static int __init xenbus_probe_initcall(void) device_initcall(xenbus_probe_initcall); -static int __init xenbus_init(void) +/* Set up event channel for xenstored which is run as a local process + * (this is normally used only in dom0) + */ +static int __init xenstored_local_init(void) { int err = 0; unsigned long page = 0; + struct evtchn_alloc_unbound alloc_unbound; - DPRINTK(""); + /* Allocate Xenstore page */ + page = get_zeroed_page(GFP_KERNEL); + if (!page) + goto out_err; - err = -ENODEV; - if (!xen_domain()) - return err; + xen_store_mfn = xen_start_info->store_mfn = + pfn_to_mfn(virt_to_phys((void *)page) >> + PAGE_SHIFT); - /* - * Domain0 doesn't have a store_evtchn or store_mfn yet. - */ - if (xen_initial_domain()) { - struct evtchn_alloc_unbound alloc_unbound; + /* Next allocate a local port which xenstored can bind to */ + alloc_unbound.dom = DOMID_SELF; + alloc_unbound.remote_dom = DOMID_SELF; - /* Allocate Xenstore page */ - page = get_zeroed_page(GFP_KERNEL); - if (!page) - goto out_error; + err = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound, + &alloc_unbound); + if (err == -ENOSYS) + goto out_err; - xen_store_mfn = xen_start_info->store_mfn = - pfn_to_mfn(virt_to_phys((void *)page) >> - PAGE_SHIFT); + BUG_ON(err); + xen_store_evtchn = xen_start_info->store_evtchn = + alloc_unbound.port; - /* Next allocate a local port which xenstored can bind to */ - alloc_unbound.dom = DOMID_SELF; - alloc_unbound.remote_dom = 0; + return 0; - err = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound, - &alloc_unbound); - if (err == -ENOSYS) - goto out_error; + out_err: + if (page != 0) + free_page(page); + return err; +} - BUG_ON(err); - xen_store_evtchn = xen_start_info->store_evtchn = - alloc_unbound.port; +static int __init xenbus_init(void) +{ + int err = 0; - xen_store_interface = mfn_to_virt(xen_store_mfn); + if (!xen_domain()) + return -ENODEV; + + if (xen_hvm_domain()) { + uint64_t v = 0; + err = hvm_get_parameter(HVM_PARAM_STORE_EVTCHN, &v); + if (err) + goto out_error; + xen_store_evtchn = (int)v; + err = hvm_get_parameter(HVM_PARAM_STORE_PFN, &v); + if (err) + goto out_error; + xen_store_mfn = (unsigned long)v; + xen_store_interface = ioremap(xen_store_mfn << PAGE_SHIFT, PAGE_SIZE); } else { - if (xen_hvm_domain()) { - uint64_t v = 0; - err = hvm_get_parameter(HVM_PARAM_STORE_EVTCHN, &v); - if (err) - goto out_error; - xen_store_evtchn = (int)v; - err = hvm_get_parameter(HVM_PARAM_STORE_PFN, &v); + xen_store_evtchn = xen_start_info->store_evtchn; + xen_store_mfn = xen_start_info->store_mfn; + if (xen_store_evtchn) + xenstored_ready = 1; + else { + err = xenstored_local_init(); if (err) goto out_error; - xen_store_mfn = (unsigned long)v; - xen_store_interface = ioremap(xen_store_mfn << PAGE_SHIFT, PAGE_SIZE); - } else { - xen_store_evtchn = xen_start_info->store_evtchn; - xen_store_mfn = xen_start_info->store_mfn; - xen_store_interface = mfn_to_virt(xen_store_mfn); - xenstored_ready = 1; } + xen_store_interface = mfn_to_virt(xen_store_mfn); } /* Initialize the interface to xenstore. */ @@ -772,12 +770,7 @@ static int __init xenbus_init(void) proc_mkdir("xen", NULL); #endif - return 0; - - out_error: - if (page != 0) - free_page(page); - + out_error: return err; } diff --git a/drivers/xen/xenbus/xenbus_probe.h b/drivers/xen/xenbus/xenbus_probe.h index 888b990..b814935 100644 --- a/drivers/xen/xenbus/xenbus_probe.h +++ b/drivers/xen/xenbus/xenbus_probe.h @@ -48,6 +48,8 @@ struct xen_bus_type struct bus_type bus; }; +extern struct device_attribute xenbus_dev_attrs[]; + extern int xenbus_match(struct device *_dev, struct device_driver *_drv); extern int xenbus_dev_probe(struct device *_dev); extern int xenbus_dev_remove(struct device *_dev); diff --git a/drivers/xen/xenbus/xenbus_probe_backend.c b/drivers/xen/xenbus/xenbus_probe_backend.c index 6cf467b..32417b5 100644 --- a/drivers/xen/xenbus/xenbus_probe_backend.c +++ b/drivers/xen/xenbus/xenbus_probe_backend.c @@ -104,8 +104,9 @@ static int xenbus_uevent_backend(struct device *dev, xdev = to_xenbus_device(dev); bus = container_of(xdev->dev.bus, struct xen_bus_type, bus); - if (xdev == NULL) - return -ENODEV; + + if (add_uevent_var(env, "MODALIAS=xen-backend:%s", xdev->devicetype)) + return -ENOMEM; /* stuff we want to pass to /sbin/hotplug */ if (add_uevent_var(env, "XENBUS_TYPE=%s", xdev->devicetype)) @@ -183,10 +184,6 @@ static void frontend_changed(struct xenbus_watch *watch, xenbus_otherend_changed(watch, vec, len, 0); } -static struct device_attribute xenbus_backend_dev_attrs[] = { - __ATTR_NULL -}; - static struct xen_bus_type xenbus_backend = { .root = "backend", .levels = 3, /* backend/type/<frontend>/<id> */ @@ -200,7 +197,7 @@ static struct xen_bus_type xenbus_backend = { .probe = xenbus_dev_probe, .remove = xenbus_dev_remove, .shutdown = xenbus_dev_shutdown, - .dev_attrs = xenbus_backend_dev_attrs, + .dev_attrs = xenbus_dev_attrs, }, }; diff --git a/drivers/xen/xenbus/xenbus_probe_frontend.c b/drivers/xen/xenbus/xenbus_probe_frontend.c index b6a2690..540587e 100644 --- a/drivers/xen/xenbus/xenbus_probe_frontend.c +++ b/drivers/xen/xenbus/xenbus_probe_frontend.c @@ -81,10 +81,6 @@ static void backend_changed(struct xenbus_watch *watch, xenbus_otherend_changed(watch, vec, len, 1); } -static struct device_attribute xenbus_frontend_dev_attrs[] = { - __ATTR_NULL -}; - static const struct dev_pm_ops xenbus_pm_ops = { .suspend = xenbus_dev_suspend, .resume = xenbus_dev_resume, @@ -106,7 +102,7 @@ static struct xen_bus_type xenbus_frontend = { .probe = xenbus_dev_probe, .remove = xenbus_dev_remove, .shutdown = xenbus_dev_shutdown, - .dev_attrs = xenbus_frontend_dev_attrs, + .dev_attrs = xenbus_dev_attrs, .pm = &xenbus_pm_ops, }, @@ -252,10 +248,131 @@ int __xenbus_register_frontend(struct xenbus_driver *drv, } EXPORT_SYMBOL_GPL(__xenbus_register_frontend); +static DECLARE_WAIT_QUEUE_HEAD(backend_state_wq); +static int backend_state; + +static void xenbus_reset_backend_state_changed(struct xenbus_watch *w, + const char **v, unsigned int l) +{ + xenbus_scanf(XBT_NIL, v[XS_WATCH_PATH], "", "%i", &backend_state); + printk(KERN_DEBUG "XENBUS: backend %s %s\n", + v[XS_WATCH_PATH], xenbus_strstate(backend_state)); + wake_up(&backend_state_wq); +} + +static void xenbus_reset_wait_for_backend(char *be, int expected) +{ + long timeout; + timeout = wait_event_interruptible_timeout(backend_state_wq, + backend_state == expected, 5 * HZ); + if (timeout <= 0) + printk(KERN_INFO "XENBUS: backend %s timed out.\n", be); +} + +/* + * Reset frontend if it is in Connected or Closed state. + * Wait for backend to catch up. + * State Connected happens during kdump, Closed after kexec. + */ +static void xenbus_reset_frontend(char *fe, char *be, int be_state) +{ + struct xenbus_watch be_watch; + + printk(KERN_DEBUG "XENBUS: backend %s %s\n", + be, xenbus_strstate(be_state)); + + memset(&be_watch, 0, sizeof(be_watch)); + be_watch.node = kasprintf(GFP_NOIO | __GFP_HIGH, "%s/state", be); + if (!be_watch.node) + return; + + be_watch.callback = xenbus_reset_backend_state_changed; + backend_state = XenbusStateUnknown; + + printk(KERN_INFO "XENBUS: triggering reconnect on %s\n", be); + register_xenbus_watch(&be_watch); + + /* fall through to forward backend to state XenbusStateInitialising */ + switch (be_state) { + case XenbusStateConnected: + xenbus_printf(XBT_NIL, fe, "state", "%d", XenbusStateClosing); + xenbus_reset_wait_for_backend(be, XenbusStateClosing); + + case XenbusStateClosing: + xenbus_printf(XBT_NIL, fe, "state", "%d", XenbusStateClosed); + xenbus_reset_wait_for_backend(be, XenbusStateClosed); + + case XenbusStateClosed: + xenbus_printf(XBT_NIL, fe, "state", "%d", XenbusStateInitialising); + xenbus_reset_wait_for_backend(be, XenbusStateInitWait); + } + + unregister_xenbus_watch(&be_watch); + printk(KERN_INFO "XENBUS: reconnect done on %s\n", be); + kfree(be_watch.node); +} + +static void xenbus_check_frontend(char *class, char *dev) +{ + int be_state, fe_state, err; + char *backend, *frontend; + + frontend = kasprintf(GFP_NOIO | __GFP_HIGH, "device/%s/%s", class, dev); + if (!frontend) + return; + + err = xenbus_scanf(XBT_NIL, frontend, "state", "%i", &fe_state); + if (err != 1) + goto out; + + switch (fe_state) { + case XenbusStateConnected: + case XenbusStateClosed: + printk(KERN_DEBUG "XENBUS: frontend %s %s\n", + frontend, xenbus_strstate(fe_state)); + backend = xenbus_read(XBT_NIL, frontend, "backend", NULL); + if (!backend || IS_ERR(backend)) + goto out; + err = xenbus_scanf(XBT_NIL, backend, "state", "%i", &be_state); + if (err == 1) + xenbus_reset_frontend(frontend, backend, be_state); + kfree(backend); + break; + default: + break; + } +out: + kfree(frontend); +} + +static void xenbus_reset_state(void) +{ + char **devclass, **dev; + int devclass_n, dev_n; + int i, j; + + devclass = xenbus_directory(XBT_NIL, "device", "", &devclass_n); + if (IS_ERR(devclass)) + return; + + for (i = 0; i < devclass_n; i++) { + dev = xenbus_directory(XBT_NIL, "device", devclass[i], &dev_n); + if (IS_ERR(dev)) + continue; + for (j = 0; j < dev_n; j++) + xenbus_check_frontend(devclass[i], dev[j]); + kfree(dev); + } + kfree(devclass); +} + static int frontend_probe_and_watch(struct notifier_block *notifier, unsigned long event, void *data) { + /* reset devices in Connected or Closed state */ + if (xen_hvm_domain()) + xenbus_reset_state(); /* Enumerate devices in xenstore and watch for changes. */ xenbus_probe_devices(&xenbus_frontend); register_xenbus_watch(&fe_watch); diff --git a/drivers/xen/xenbus/xenbus_xs.c b/drivers/xen/xenbus/xenbus_xs.c index 5534690..b3b8f2f 100644 --- a/drivers/xen/xenbus/xenbus_xs.c +++ b/drivers/xen/xenbus/xenbus_xs.c @@ -45,6 +45,7 @@ #include <linux/module.h> #include <linux/mutex.h> #include <xen/xenbus.h> +#include <xen/xen.h> #include "xenbus_comms.h" struct xs_stored_msg { @@ -620,6 +621,15 @@ static struct xenbus_watch *find_watch(const char *token) return NULL; } +static void xs_reset_watches(void) +{ + int err; + + err = xs_error(xs_single(XBT_NIL, XS_RESET_WATCHES, "", NULL)); + if (err && err != -EEXIST) + printk(KERN_WARNING "xs_reset_watches failed: %d\n", err); +} + /* Register callback to watch this node. */ int register_xenbus_watch(struct xenbus_watch *watch) { @@ -638,8 +648,7 @@ int register_xenbus_watch(struct xenbus_watch *watch) err = xs_watch(watch->node, token); - /* Ignore errors due to multiple registration. */ - if ((err != 0) && (err != -EEXIST)) { + if (err) { spin_lock(&watches_lock); list_del(&watch->list); spin_unlock(&watches_lock); @@ -897,5 +906,9 @@ int xs_init(void) if (IS_ERR(task)) return PTR_ERR(task); + /* shutdown watches for kexec boot */ + if (xen_hvm_domain()) + xs_reset_watches(); + return 0; } |