Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-iommu

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-iommu: (89 commits)
  AMD IOMMU: remove now unnecessary #ifdefs
  AMD IOMMU: prealloc_protection_domains should be static
  kvm/iommu: fix compile warning
  AMD IOMMU: add statistics about total number of map requests
  AMD IOMMU: add statistics about allocated io memory
  AMD IOMMU: add stats counter for domain tlb flushes
  AMD IOMMU: add stats counter for single iommu domain tlb flushes
  AMD IOMMU: add stats counter for cross-page request
  AMD IOMMU: add stats counter for free_coherent requests
  AMD IOMMU: add stats counter for alloc_coherent requests
  AMD IOMMU: add stats counter for unmap_sg requests
  AMD IOMMU: add stats counter for map_sg requests
  AMD IOMMU: add stats counter for unmap_single requests
  AMD IOMMU: add stats counter for map_single requests
  AMD IOMMU: add stats counter for completion wait events
  AMD IOMMU: add init code for statistic collection
  AMD IOMMU: add necessary header defines for stats counting
  AMD IOMMU: add Kconfig entry for statistic collection code
  AMD IOMMU: use dev_name in iommu_enable function
  AMD IOMMU: use calc_devid in prealloc_protection_domains
  ...

arch/ia64/Kconfig

@@ -687,3 +687,6 @@ config IRQ_PER_CPU
 
 config IOMMU_HELPER
 	def_bool (IA64_HP_ZX1 || IA64_HP_ZX1_SWIOTLB || IA64_GENERIC || SWIOTLB)
+
+config IOMMU_API
+	def_bool (DMAR)

arch/ia64/include/asm/kvm_host.h

@@ -467,7 +467,7 @@ struct kvm_arch {
 	struct kvm_sal_data rdv_sal_data;
 
 	struct list_head assigned_dev_head;
-	struct dmar_domain *intel_iommu_domain;
+	struct iommu_domain *iommu_domain;
 	struct hlist_head irq_ack_notifier_list;
 
 	unsigned long irq_sources_bitmap;

arch/ia64/kvm/Makefile

@@ -51,8 +51,8 @@ EXTRA_AFLAGS += -Ivirt/kvm -Iarch/ia64/kvm/
 common-objs = $(addprefix ../../../virt/kvm/, kvm_main.o ioapic.o \
 		coalesced_mmio.o irq_comm.o)
 
-ifeq ($(CONFIG_DMAR),y)
-common-objs += $(addprefix ../../../virt/kvm/, vtd.o)
+ifeq ($(CONFIG_IOMMU_API),y)
+common-objs += $(addprefix ../../../virt/kvm/, iommu.o)
 endif
 
 kvm-objs := $(common-objs) kvm-ia64.o kvm_fw.o

arch/ia64/kvm/kvm-ia64.c

@@ -31,6 +31,7 @@
 #include <linux/bitops.h>
 #include <linux/hrtimer.h>
 #include <linux/uaccess.h>
+#include <linux/iommu.h>
 #include <linux/intel-iommu.h>
 
 #include <asm/pgtable.h>
@@ -188,7 +189,7 @@ int kvm_dev_ioctl_check_extension(long ext)
 		r = KVM_COALESCED_MMIO_PAGE_OFFSET;
 		break;
 	case KVM_CAP_IOMMU:
-		r = intel_iommu_found();
+		r = iommu_found();
 		break;
 	default:
 		r = 0;

arch/x86/Kconfig

@@ -586,6 +586,16 @@ config AMD_IOMMU
 	  your BIOS for an option to enable it or if you have an IVRS ACPI
 	  table.
 
+config AMD_IOMMU_STATS
+	bool "Export AMD IOMMU statistics to debugfs"
+	depends on AMD_IOMMU
+	select DEBUG_FS
+	help
+	  This option enables code in the AMD IOMMU driver to collect various
+	  statistics about whats happening in the driver and exports that
+	  information to userspace via debugfs.
+	  If unsure, say N.
+
 # need this always selected by IOMMU for the VIA workaround
 config SWIOTLB
 	def_bool y if X86_64
@@ -599,6 +609,9 @@ config SWIOTLB
 config IOMMU_HELPER
 	def_bool (CALGARY_IOMMU || GART_IOMMU || SWIOTLB || AMD_IOMMU)
 
+config IOMMU_API
+	def_bool (AMD_IOMMU || DMAR)
+
 config MAXSMP
 	bool "Configure Maximum number of SMP Processors and NUMA Nodes"
 	depends on X86_64 && SMP && DEBUG_KERNEL && EXPERIMENTAL

arch/x86/include/asm/amd_iommu_types.h

@@ -190,16 +190,23 @@
 /* FIXME: move this macro to <linux/pci.h> */
 #define PCI_BUS(x) (((x) >> 8) & 0xff)
 
+/* Protection domain flags */
+#define PD_DMA_OPS_MASK		(1UL << 0) /* domain used for dma_ops */
+#define PD_DEFAULT_MASK		(1UL << 1) /* domain is a default dma_ops
+					      domain for an IOMMU */
+
 /*
  * This structure contains generic data for  IOMMU protection domains
  * independent of their use.
  */
 struct protection_domain {
-	spinlock_t lock; /* mostly used to lock the page table*/
-	u16 id;		 /* the domain id written to the device table */
-	int mode;	 /* paging mode (0-6 levels) */
-	u64 *pt_root;	 /* page table root pointer */
-	void *priv;	 /* private data */
+	spinlock_t lock;	/* mostly used to lock the page table*/
+	u16 id;			/* the domain id written to the device table */
+	int mode;		/* paging mode (0-6 levels) */
+	u64 *pt_root;		/* page table root pointer */
+	unsigned long flags;	/* flags to find out type of domain */
+	unsigned dev_cnt;	/* devices assigned to this domain */
+	void *priv;		/* private data */
 };
 
 /*
@@ -295,7 +302,7 @@ struct amd_iommu {
 	bool int_enabled;
 
 	/* if one, we need to send a completion wait command */
-	int need_sync;
+	bool need_sync;
 
 	/* default dma_ops domain for that IOMMU */
 	struct dma_ops_domain *default_dom;
@@ -374,7 +381,7 @@ extern struct protection_domain **amd_iommu_pd_table;
 extern unsigned long *amd_iommu_pd_alloc_bitmap;
 
 /* will be 1 if device isolation is enabled */
-extern int amd_iommu_isolate;
+extern bool amd_iommu_isolate;
 
 /*
  * If true, the addresses will be flushed on unmap time, not when
@@ -382,18 +389,6 @@ extern int amd_iommu_isolate;
  */
 extern bool amd_iommu_unmap_flush;
 
-/* takes a PCI device id and prints it out in a readable form */
-static inline void print_devid(u16 devid, int nl)
-{
-	int bus = devid >> 8;
-	int dev = devid >> 3 & 0x1f;
-	int fn  = devid & 0x07;
-
-	printk("%02x:%02x.%x", bus, dev, fn);
-	if (nl)
-		printk("\n");
-}
-
 /* takes bus and device/function and returns the device id
  * FIXME: should that be in generic PCI code? */
 static inline u16 calc_devid(u8 bus, u8 devfn)
@@ -401,4 +396,32 @@ static inline u16 calc_devid(u8 bus, u8 devfn)
 	return (((u16)bus) << 8) | devfn;
 }
 
+#ifdef CONFIG_AMD_IOMMU_STATS
+
+struct __iommu_counter {
+	char *name;
+	struct dentry *dent;
+	u64 value;
+};
+
+#define DECLARE_STATS_COUNTER(nm) \
+	static struct __iommu_counter nm = {	\
+		.name = #nm,			\
+	}
+
+#define INC_STATS_COUNTER(name)		name.value += 1
+#define ADD_STATS_COUNTER(name, x)	name.value += (x)
+#define SUB_STATS_COUNTER(name, x)	name.value -= (x)
+
+#else /* CONFIG_AMD_IOMMU_STATS */
+
+#define DECLARE_STATS_COUNTER(name)
+#define INC_STATS_COUNTER(name)
+#define ADD_STATS_COUNTER(name, x)
+#define SUB_STATS_COUNTER(name, x)
+
+static inline void amd_iommu_stats_init(void) { }
+
+#endif /* CONFIG_AMD_IOMMU_STATS */
+
 #endif /* _ASM_X86_AMD_IOMMU_TYPES_H */

arch/x86/include/asm/kvm_host.h

@@ -360,7 +360,7 @@ struct kvm_arch{
 	struct list_head active_mmu_pages;
 	struct list_head assigned_dev_head;
 	struct list_head oos_global_pages;
-	struct dmar_domain *intel_iommu_domain;
+	struct iommu_domain *iommu_domain;
 	struct kvm_pic *vpic;
 	struct kvm_ioapic *vioapic;
 	struct kvm_pit *vpit;

arch/x86/kernel/amd_iommu.c

@@ -20,8 +20,12 @@
 #include <linux/pci.h>
 #include <linux/gfp.h>
 #include <linux/bitops.h>
+#include <linux/debugfs.h>
 #include <linux/scatterlist.h>
 #include <linux/iommu-helper.h>
+#ifdef CONFIG_IOMMU_API
+#include <linux/iommu.h>
+#endif
 #include <asm/proto.h>
 #include <asm/iommu.h>
 #include <asm/gart.h>
@@ -38,6 +42,10 @@ static DEFINE_RWLOCK(amd_iommu_devtable_lock);
 static LIST_HEAD(iommu_pd_list);
 static DEFINE_SPINLOCK(iommu_pd_list_lock);
 
+#ifdef CONFIG_IOMMU_API
+static struct iommu_ops amd_iommu_ops;
+#endif
+
 /*
  * general struct to manage commands send to an IOMMU
  */
@@ -47,6 +55,68 @@ struct iommu_cmd {
 
 static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
 			     struct unity_map_entry *e);
+static struct dma_ops_domain *find_protection_domain(u16 devid);
+
+
+#ifdef CONFIG_AMD_IOMMU_STATS
+
+/*
+ * Initialization code for statistics collection
+ */
+
+DECLARE_STATS_COUNTER(compl_wait);
+DECLARE_STATS_COUNTER(cnt_map_single);
+DECLARE_STATS_COUNTER(cnt_unmap_single);
+DECLARE_STATS_COUNTER(cnt_map_sg);
+DECLARE_STATS_COUNTER(cnt_unmap_sg);
+DECLARE_STATS_COUNTER(cnt_alloc_coherent);
+DECLARE_STATS_COUNTER(cnt_free_coherent);
+DECLARE_STATS_COUNTER(cross_page);
+DECLARE_STATS_COUNTER(domain_flush_single);
+DECLARE_STATS_COUNTER(domain_flush_all);
+DECLARE_STATS_COUNTER(alloced_io_mem);
+DECLARE_STATS_COUNTER(total_map_requests);
+
+static struct dentry *stats_dir;
+static struct dentry *de_isolate;
+static struct dentry *de_fflush;
+
+static void amd_iommu_stats_add(struct __iommu_counter *cnt)
+{
+	if (stats_dir == NULL)
+		return;
+
+	cnt->dent = debugfs_create_u64(cnt->name, 0444, stats_dir,
+				       &cnt->value);
+}
+
+static void amd_iommu_stats_init(void)
+{
+	stats_dir = debugfs_create_dir("amd-iommu", NULL);
+	if (stats_dir == NULL)
+		return;
+
+	de_isolate = debugfs_create_bool("isolation", 0444, stats_dir,
+					 (u32 *)&amd_iommu_isolate);
+
+	de_fflush  = debugfs_create_bool("fullflush", 0444, stats_dir,
+					 (u32 *)&amd_iommu_unmap_flush);
+
+	amd_iommu_stats_add(&compl_wait);
+	amd_iommu_stats_add(&cnt_map_single);
+	amd_iommu_stats_add(&cnt_unmap_single);
+	amd_iommu_stats_add(&cnt_map_sg);
+	amd_iommu_stats_add(&cnt_unmap_sg);
+	amd_iommu_stats_add(&cnt_alloc_coherent);
+	amd_iommu_stats_add(&cnt_free_coherent);
+	amd_iommu_stats_add(&cross_page);
+	amd_iommu_stats_add(&domain_flush_single);
+	amd_iommu_stats_add(&domain_flush_all);
+	amd_iommu_stats_add(&alloced_io_mem);
+	amd_iommu_stats_add(&total_map_requests);
+}
+
+#endif
 
 /* returns !0 if the IOMMU is caching non-present entries in its TLB */
 static int iommu_has_npcache(struct amd_iommu *iommu)
@@ -189,12 +259,54 @@ static int iommu_queue_command(struct amd_iommu *iommu, struct iommu_cmd *cmd)
 	spin_lock_irqsave(&iommu->lock, flags);
 	ret = __iommu_queue_command(iommu, cmd);
 	if (!ret)
-		iommu->need_sync = 1;
+		iommu->need_sync = true;
 	spin_unlock_irqrestore(&iommu->lock, flags);
 
 	return ret;
 }
 
+/*
+ * This function waits until an IOMMU has completed a completion
+ * wait command
+ */
+static void __iommu_wait_for_completion(struct amd_iommu *iommu)
+{
+	int ready = 0;
+	unsigned status = 0;
+	unsigned long i = 0;
+
+	INC_STATS_COUNTER(compl_wait);
+
+	while (!ready && (i < EXIT_LOOP_COUNT)) {
+		++i;
+		/* wait for the bit to become one */
+		status = readl(iommu->mmio_base + MMIO_STATUS_OFFSET);
+		ready = status & MMIO_STATUS_COM_WAIT_INT_MASK;
+	}
+
+	/* set bit back to zero */
+	status &= ~MMIO_STATUS_COM_WAIT_INT_MASK;
+	writel(status, iommu->mmio_base + MMIO_STATUS_OFFSET);
+
+	if (unlikely(i == EXIT_LOOP_COUNT))
+		panic("AMD IOMMU: Completion wait loop failed\n");
+}
+
+/*
+ * This function queues a completion wait command into the command
+ * buffer of an IOMMU
+ */
+static int __iommu_completion_wait(struct amd_iommu *iommu)
+{
+	struct iommu_cmd cmd;
+
+	 memset(&cmd, 0, sizeof(cmd));
+	 cmd.data[0] = CMD_COMPL_WAIT_INT_MASK;
+	 CMD_SET_TYPE(&cmd, CMD_COMPL_WAIT);
+
+	 return __iommu_queue_command(iommu, &cmd);
+}
+
 /*
  * This function is called whenever we need to ensure that the IOMMU has
  * completed execution of all commands we sent. It sends a
@@ -204,40 +316,22 @@ static int iommu_queue_command(struct amd_iommu *iommu, struct iommu_cmd *cmd)
  */
 static int iommu_completion_wait(struct amd_iommu *iommu)
 {
-	int ret = 0, ready = 0;
-	unsigned status = 0;
-	struct iommu_cmd cmd;
-	unsigned long flags, i = 0;
-
-	memset(&cmd, 0, sizeof(cmd));
-	cmd.data[0] = CMD_COMPL_WAIT_INT_MASK;
-	CMD_SET_TYPE(&cmd, CMD_COMPL_WAIT);
+	int ret = 0;
+	unsigned long flags;
 
 	spin_lock_irqsave(&iommu->lock, flags);
 
 	if (!iommu->need_sync)
 		goto out;
 
-	iommu->need_sync = 0;
+	ret = __iommu_completion_wait(iommu);
 
-	ret = __iommu_queue_command(iommu, &cmd);
+	iommu->need_sync = false;
 
 	if (ret)
 		goto out;
 
-	while (!ready && (i < EXIT_LOOP_COUNT)) {
-		++i;
-		/* wait for the bit to become one */
-		status = readl(iommu->mmio_base + MMIO_STATUS_OFFSET);
-		ready = status & MMIO_STATUS_COM_WAIT_INT_MASK;
-	}
-
-	/* set bit back to zero */
-	status &= ~MMIO_STATUS_COM_WAIT_INT_MASK;
-	writel(status, iommu->mmio_base + MMIO_STATUS_OFFSET);
-
-	if (unlikely(i == EXIT_LOOP_COUNT))
-		panic("AMD IOMMU: Completion wait loop failed\n");
+	__iommu_wait_for_completion(iommu);
 
 out:
 	spin_unlock_irqrestore(&iommu->lock, flags);
@@ -264,6 +358,21 @@ static int iommu_queue_inv_dev_entry(struct amd_iommu *iommu, u16 devid)
 	return ret;
 }
 
+static void __iommu_build_inv_iommu_pages(struct iommu_cmd *cmd, u64 address,
+					  u16 domid, int pde, int s)
+{
+	memset(cmd, 0, sizeof(*cmd));
+	address &= PAGE_MASK;
+	CMD_SET_TYPE(cmd, CMD_INV_IOMMU_PAGES);
+	cmd->data[1] |= domid;
+	cmd->data[2] = lower_32_bits(address);
+	cmd->data[3] = upper_32_bits(address);
+	if (s) /* size bit - we flush more than one 4kb page */
+		cmd->data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK;
+	if (pde) /* PDE bit - we wan't flush everything not only the PTEs */
+		cmd->data[2] |= CMD_INV_IOMMU_PAGES_PDE_MASK;
+}
+
 /*
  * Generic command send function for invalidaing TLB entries
  */
@@ -273,16 +382,7 @@ static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu,
 	struct iommu_cmd cmd;
 	int ret;
 
-	memset(&cmd, 0, sizeof(cmd));
-	address &= PAGE_MASK;
-	CMD_SET_TYPE(&cmd, CMD_INV_IOMMU_PAGES);
-	cmd.data[1] |= domid;
-	cmd.data[2] = lower_32_bits(address);
-	cmd.data[3] = upper_32_bits(address);
-	if (s) /* size bit - we flush more than one 4kb page */
-		cmd.data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK;
-	if (pde) /* PDE bit - we wan't flush everything not only the PTEs */
-		cmd.data[2] |= CMD_INV_IOMMU_PAGES_PDE_MASK;
+	__iommu_build_inv_iommu_pages(&cmd, address, domid, pde, s);
 
 	ret = iommu_queue_command(iommu, &cmd);
 
@@ -321,9 +421,35 @@ static void iommu_flush_tlb(struct amd_iommu *iommu, u16 domid)
 {
 	u64 address = CMD_INV_IOMMU_ALL_PAGES_ADDRESS;
 
+	INC_STATS_COUNTER(domain_flush_single);
+
 	iommu_queue_inv_iommu_pages(iommu, address, domid, 0, 1);
 }
 
+/*
+ * This function is used to flush the IO/TLB for a given protection domain
+ * on every IOMMU in the system
+ */
+static void iommu_flush_domain(u16 domid)
+{
+	unsigned long flags;
+	struct amd_iommu *iommu;
+	struct iommu_cmd cmd;
+
+	INC_STATS_COUNTER(domain_flush_all);
+
+	__iommu_build_inv_iommu_pages(&cmd, CMD_INV_IOMMU_ALL_PAGES_ADDRESS,
+				      domid, 1, 1);
+
+	list_for_each_entry(iommu, &amd_iommu_list, list) {
+		spin_lock_irqsave(&iommu->lock, flags);
+		__iommu_queue_command(iommu, &cmd);
+		__iommu_completion_wait(iommu);
+		__iommu_wait_for_completion(iommu);
+		spin_unlock_irqrestore(&iommu->lock, flags);
+	}
+}
+
 /****************************************************************************
  *
  * The functions below are used the create the page table mappings for
@@ -338,10 +464,10 @@ static void iommu_flush_tlb(struct amd_iommu *iommu, u16 domid)
  * supporting all features of AMD IOMMU page tables like level skipping
  * and full 64 bit address spaces.
  */
-static int iommu_map(struct protection_domain *dom,
-		     unsigned long bus_addr,
-		     unsigned long phys_addr,
-		     int prot)
+static int iommu_map_page(struct protection_domain *dom,
+			  unsigned long bus_addr,
+			  unsigned long phys_addr,
+			  int prot)
 {
 	u64 __pte, *pte, *page;
 
@@ -388,6 +514,28 @@ static int iommu_map(struct protection_domain *dom,
 	return 0;
 }
 
+static void iommu_unmap_page(struct protection_domain *dom,
+			     unsigned long bus_addr)
+{
+	u64 *pte;
+
+	pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(bus_addr)];
+
+	if (!IOMMU_PTE_PRESENT(*pte))
+		return;
+
+	pte = IOMMU_PTE_PAGE(*pte);
+	pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)];
+
+	if (!IOMMU_PTE_PRESENT(*pte))
+		return;
+
+	pte = IOMMU_PTE_PAGE(*pte);
+	pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)];
+
+	*pte = 0;
+}
+
 /*
  * This function checks if a specific unity mapping entry is needed for
  * this specific IOMMU.
@@ -440,7 +588,7 @@ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
 
 	for (addr = e->address_start; addr < e->address_end;
 	     addr += PAGE_SIZE) {
-		ret = iommu_map(&dma_dom->domain, addr, addr, e->prot);
+		ret = iommu_map_page(&dma_dom->domain, addr, addr, e->prot);
 		if (ret)
 			return ret;
 		/*
@@ -571,6 +719,16 @@ static u16 domain_id_alloc(void)
 	return id;
 }
 
+static void domain_id_free(int id)
+{
+	unsigned long flags;
+
+	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
+	if (id > 0 && id < MAX_DOMAIN_ID)
+		__clear_bit(id, amd_iommu_pd_alloc_bitmap);
+	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+}
+
 /*
  * Used to reserve address ranges in the aperture (e.g. for exclusion
  * ranges.
@@ -587,12 +745,12 @@ static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
 	iommu_area_reserve(dom->bitmap, start_page, pages);
 }
 
-static void dma_ops_free_pagetable(struct dma_ops_domain *dma_dom)
+static void free_pagetable(struct protection_domain *domain)
 {
 	int i, j;
 	u64 *p1, *p2, *p3;
 
-	p1 = dma_dom->domain.pt_root;
+	p1 = domain->pt_root;
 
 	if (!p1)
 		return;
@@ -613,6 +771,8 @@ static void dma_ops_free_pagetable(struct dma_ops_domain *dma_dom)
 	}
 
 	free_page((unsigned long)p1);
+
+	domain->pt_root = NULL;
 }
 
 /*
@@ -624,7 +784,7 @@ static void dma_ops_domain_free(struct dma_ops_domain *dom)
 	if (!dom)
 		return;
 
-	dma_ops_free_pagetable(dom);
+	free_pagetable(&dom->domain);
 
 	kfree(dom->pte_pages);
 
@@ -663,6 +823,7 @@ static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu,
 		goto free_dma_dom;
 	dma_dom->domain.mode = PAGE_MODE_3_LEVEL;
 	dma_dom->domain.pt_root = (void *)get_zeroed_page(GFP_KERNEL);
+	dma_dom->domain.flags = PD_DMA_OPS_MASK;
 	dma_dom->domain.priv = dma_dom;
 	if (!dma_dom->domain.pt_root)
 		goto free_dma_dom;
@@ -724,6 +885,15 @@ free_dma_dom:
 	return NULL;
 }
 
+/*
+ * little helper function to check whether a given protection domain is a
+ * dma_ops domain
+ */
+static bool dma_ops_domain(struct protection_domain *domain)
+{
+	return domain->flags & PD_DMA_OPS_MASK;
+}
+
 /*
  * Find out the protection domain structure for a given PCI device. This
  * will give us the pointer to the page table root for example.
@@ -744,14 +914,15 @@ static struct protection_domain *domain_for_device(u16 devid)
  * If a device is not yet associated with a domain, this function does
  * assigns it visible for the hardware
  */
-static void set_device_domain(struct amd_iommu *iommu,
-			      struct protection_domain *domain,
-			      u16 devid)
+static void attach_device(struct amd_iommu *iommu,
+			  struct protection_domain *domain,
+			  u16 devid)
 {
 	unsigned long flags;
-
 	u64 pte_root = virt_to_phys(domain->pt_root);
 
+	domain->dev_cnt += 1;
+
 	pte_root |= (domain->mode & DEV_ENTRY_MODE_MASK)
 		    << DEV_ENTRY_MODE_SHIFT;
 	pte_root |= IOMMU_PTE_IR | IOMMU_PTE_IW | IOMMU_PTE_P | IOMMU_PTE_TV;
@@ -767,6 +938,116 @@ static void set_device_domain(struct amd_iommu *iommu,
 	iommu_queue_inv_dev_entry(iommu, devid);
 }
 
+/*
+ * Removes a device from a protection domain (unlocked)
+ */
+static void __detach_device(struct protection_domain *domain, u16 devid)
+{
+
+	/* lock domain */
+	spin_lock(&domain->lock);
+
+	/* remove domain from the lookup table */
+	amd_iommu_pd_table[devid] = NULL;
+
+	/* remove entry from the device table seen by the hardware */
+	amd_iommu_dev_table[devid].data[0] = IOMMU_PTE_P | IOMMU_PTE_TV;
+	amd_iommu_dev_table[devid].data[1] = 0;
+	amd_iommu_dev_table[devid].data[2] = 0;
+
+	/* decrease reference counter */
+	domain->dev_cnt -= 1;
+
+	/* ready */
+	spin_unlock(&domain->lock);
+}
+
+/*
+ * Removes a device from a protection domain (with devtable_lock held)
+ */
+static void detach_device(struct protection_domain *domain, u16 devid)
+{
+	unsigned long flags;
+
+	/* lock device table */
+	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
+	__detach_device(domain, devid);
+	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+}
+
+static int device_change_notifier(struct notifier_block *nb,
+				  unsigned long action, void *data)
+{
+	struct device *dev = data;
+	struct pci_dev *pdev = to_pci_dev(dev);
+	u16 devid = calc_devid(pdev->bus->number, pdev->devfn);
+	struct protection_domain *domain;
+	struct dma_ops_domain *dma_domain;
+	struct amd_iommu *iommu;
+	int order = amd_iommu_aperture_order;
+	unsigned long flags;
+
+	if (devid > amd_iommu_last_bdf)
+		goto out;
+
+	devid = amd_iommu_alias_table[devid];
+
+	iommu = amd_iommu_rlookup_table[devid];
+	if (iommu == NULL)
+		goto out;
+
+	domain = domain_for_device(devid);
+
+	if (domain && !dma_ops_domain(domain))
+		WARN_ONCE(1, "AMD IOMMU WARNING: device %s already bound "
+			  "to a non-dma-ops domain\n", dev_name(dev));
+
+	switch (action) {
+	case BUS_NOTIFY_BOUND_DRIVER:
+		if (domain)
+			goto out;
+		dma_domain = find_protection_domain(devid);
+		if (!dma_domain)
+			dma_domain = iommu->default_dom;
+		attach_device(iommu, &dma_domain->domain, devid);
+		printk(KERN_INFO "AMD IOMMU: Using protection domain %d for "
+		       "device %s\n", dma_domain->domain.id, dev_name(dev));
+		break;
+	case BUS_NOTIFY_UNBIND_DRIVER:
+		if (!domain)
+			goto out;
+		detach_device(domain, devid);
+		break;
+	case BUS_NOTIFY_ADD_DEVICE:
+		/* allocate a protection domain if a device is added */
+		dma_domain = find_protection_domain(devid);
+		if (dma_domain)
+			goto out;
+		dma_domain = dma_ops_domain_alloc(iommu, order);
+		if (!dma_domain)
+			goto out;
+		dma_domain->target_dev = devid;
+
+		spin_lock_irqsave(&iommu_pd_list_lock, flags);
+		list_add_tail(&dma_domain->list, &iommu_pd_list);
+		spin_unlock_irqrestore(&iommu_pd_list_lock, flags);
+
+		break;
+	default:
+		goto out;
+	}
+
+	iommu_queue_inv_dev_entry(iommu, devid);
+	iommu_completion_wait(iommu);
+
+out:
+	return 0;
+}
+
+struct notifier_block device_nb = {
+	.notifier_call = device_change_notifier,
+};
+
 /*****************************************************************************
  *
  * The next functions belong to the dma_ops mapping/unmapping code.
@@ -802,7 +1083,6 @@ static struct dma_ops_domain *find_protection_domain(u16 devid)
 	list_for_each_entry(entry, &iommu_pd_list, list) {
 		if (entry->target_dev == devid) {
 			ret = entry;
-			list_del(&ret->list);
 			break;
 		}
 	}
@@ -853,14 +1133,13 @@ static int get_device_resources(struct device *dev,
 		if (!dma_dom)
 			dma_dom = (*iommu)->default_dom;
 		*domain = &dma_dom->domain;
-		set_device_domain(*iommu, *domain, *bdf);
+		attach_device(*iommu, *domain, *bdf);
 		printk(KERN_INFO "AMD IOMMU: Using protection domain %d for "
-				"device ", (*domain)->id);
-		print_devid(_bdf, 1);
+				"device %s\n", (*domain)->id, dev_name(dev));
 	}
 
 	if (domain_for_device(_bdf) == NULL)
-		set_device_domain(*iommu, *domain, _bdf);
+		attach_device(*iommu, *domain, _bdf);
 
 	return 1;
 }
@@ -946,6 +1225,11 @@ static dma_addr_t __map_single(struct device *dev,
 	pages = iommu_num_pages(paddr, size, PAGE_SIZE);
 	paddr &= PAGE_MASK;
 
+	INC_STATS_COUNTER(total_map_requests);
+
+	if (pages > 1)
+		INC_STATS_COUNTER(cross_page);
+
 	if (align)
 		align_mask = (1UL << get_order(size)) - 1;
 
@@ -962,6 +1246,8 @@ static dma_addr_t __map_single(struct device *dev,
 	}
 	address += offset;
 
+	ADD_STATS_COUNTER(alloced_io_mem, size);
+
 	if (unlikely(dma_dom->need_flush && !amd_iommu_unmap_flush)) {
 		iommu_flush_tlb(iommu, dma_dom->domain.id);
 		dma_dom->need_flush = false;
@@ -998,6 +1284,8 @@ static void __unmap_single(struct amd_iommu *iommu,
 		start += PAGE_SIZE;
 	}
 
+	SUB_STATS_COUNTER(alloced_io_mem, size);
+
 	dma_ops_free_addresses(dma_dom, dma_addr, pages);
 
 	if (amd_iommu_unmap_flush || dma_dom->need_flush) {
@@ -1019,6 +1307,8 @@ static dma_addr_t map_single(struct device *dev, phys_addr_t paddr,
 	dma_addr_t addr;
 	u64 dma_mask;
 
+	INC_STATS_COUNTER(cnt_map_single);
+
 	if (!check_device(dev))
 		return bad_dma_address;
 
@@ -1030,6 +1320,9 @@ static dma_addr_t map_single(struct device *dev, phys_addr_t paddr,
 		/* device not handled by any AMD IOMMU */
 		return (dma_addr_t)paddr;
 
+	if (!dma_ops_domain(domain))
+		return bad_dma_address;
+
 	spin_lock_irqsave(&domain->lock, flags);
 	addr = __map_single(dev, iommu, domain->priv, paddr, size, dir, false,
 			    dma_mask);
@@ -1055,11 +1348,16 @@ static void unmap_single(struct device *dev, dma_addr_t dma_addr,
 	struct protection_domain *domain;
 	u16 devid;
 
+	INC_STATS_COUNTER(cnt_unmap_single);
+
 	if (!check_device(dev) ||
 	    !get_device_resources(dev, &iommu, &domain, &devid))
 		/* device not handled by any AMD IOMMU */
 		return;
 
+	if (!dma_ops_domain(domain))
+		return;
+
 	spin_lock_irqsave(&domain->lock, flags);
 
 	__unmap_single(iommu, domain->priv, dma_addr, size, dir);
@@ -1104,6 +1402,8 @@ static int map_sg(struct device *dev, struct scatterlist *sglist,
 	int mapped_elems = 0;
 	u64 dma_mask;
 
+	INC_STATS_COUNTER(cnt_map_sg);
+
 	if (!check_device(dev))
 		return 0;
 
@@ -1114,6 +1414,9 @@ static int map_sg(struct device *dev, struct scatterlist *sglist,
 	if (!iommu || !domain)
 		return map_sg_no_iommu(dev, sglist, nelems, dir);
 
+	if (!dma_ops_domain(domain))
+		return 0;
+
 	spin_lock_irqsave(&domain->lock, flags);
 
 	for_each_sg(sglist, s, nelems, i) {
@@ -1163,10 +1466,15 @@ static void unmap_sg(struct device *dev, struct scatterlist *sglist,
 	u16 devid;
 	int i;
 
+	INC_STATS_COUNTER(cnt_unmap_sg);
+
 	if (!check_device(dev) ||
 	    !get_device_resources(dev, &iommu, &domain, &devid))
 		return;
 
+	if (!dma_ops_domain(domain))
+		return;
+
 	spin_lock_irqsave(&domain->lock, flags);
 
 	for_each_sg(sglist, s, nelems, i) {
@@ -1194,6 +1502,8 @@ static void *alloc_coherent(struct device *dev, size_t size,
 	phys_addr_t paddr;
 	u64 dma_mask = dev->coherent_dma_mask;
 
+	INC_STATS_COUNTER(cnt_alloc_coherent);
+
 	if (!check_device(dev))
 		return NULL;
 
@@ -1212,6 +1522,9 @@ static void *alloc_coherent(struct device *dev, size_t size,
 		return virt_addr;
 	}
 
+	if (!dma_ops_domain(domain))
+		goto out_free;
+
 	if (!dma_mask)
 		dma_mask = *dev->dma_mask;
 
@@ -1220,18 +1533,20 @@ static void *alloc_coherent(struct device *dev, size_t size,
 	*dma_addr = __map_single(dev, iommu, domain->priv, paddr,
 				 size, DMA_BIDIRECTIONAL, true, dma_mask);
 
-	if (*dma_addr == bad_dma_address) {
-		free_pages((unsigned long)virt_addr, get_order(size));
-		virt_addr = NULL;
-		goto out;
-	}
+	if (*dma_addr == bad_dma_address)
+		goto out_free;
 
 	iommu_completion_wait(iommu);
 
-out:
 	spin_unlock_irqrestore(&domain->lock, flags);
 
 	return virt_addr;
+
+out_free:
+
+	free_pages((unsigned long)virt_addr, get_order(size));
+
+	return NULL;
 }
 
 /*
@@ -1245,6 +1560,8 @@ static void free_coherent(struct device *dev, size_t size,
 	struct protection_domain *domain;
 	u16 devid;
 
+	INC_STATS_COUNTER(cnt_free_coherent);
+
 	if (!check_device(dev))
 		return;
 
@@ -1253,6 +1570,9 @@ static void free_coherent(struct device *dev, size_t size,
 	if (!iommu || !domain)
 		goto free_mem;
 
+	if (!dma_ops_domain(domain))
+		goto free_mem;
+
 	spin_lock_irqsave(&domain->lock, flags);
 
 	__unmap_single(iommu, domain->priv, dma_addr, size, DMA_BIDIRECTIONAL);
@@ -1305,7 +1625,7 @@ static void prealloc_protection_domains(void)
 	u16 devid;
 
 	while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
-		devid = (dev->bus->number << 8) | dev->devfn;
+		devid = calc_devid(dev->bus->number, dev->devfn);
 		if (devid > amd_iommu_last_bdf)
 			continue;
 		devid = amd_iommu_alias_table[devid];
@@ -1352,6 +1672,7 @@ int __init amd_iommu_init_dma_ops(void)
 		iommu->default_dom = dma_ops_domain_alloc(iommu, order);
 		if (iommu->default_dom == NULL)
 			return -ENOMEM;
+		iommu->default_dom->domain.flags |= PD_DEFAULT_MASK;
 		ret = iommu_init_unity_mappings(iommu);
 		if (ret)
 			goto free_domains;
@@ -1375,6 +1696,12 @@ int __init amd_iommu_init_dma_ops(void)
 	/* Make the driver finally visible to the drivers */
 	dma_ops = &amd_iommu_dma_ops;
 
+	register_iommu(&amd_iommu_ops);
+
+	bus_register_notifier(&pci_bus_type, &device_nb);
+
+	amd_iommu_stats_init();
+
 	return 0;
 
 free_domains:
@@ -1386,3 +1713,224 @@ free_domains:
 
 	return ret;
 }
+
+/*****************************************************************************
+ *
+ * The following functions belong to the exported interface of AMD IOMMU
+ *
+ * This interface allows access to lower level functions of the IOMMU
+ * like protection domain handling and assignement of devices to domains
+ * which is not possible with the dma_ops interface.
+ *
+ *****************************************************************************/
+
+static void cleanup_domain(struct protection_domain *domain)
+{
+	unsigned long flags;
+	u16 devid;
+
+	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
+
+	for (devid = 0; devid <= amd_iommu_last_bdf; ++devid)
+		if (amd_iommu_pd_table[devid] == domain)
+			__detach_device(domain, devid);
+
+	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+}
+
+static int amd_iommu_domain_init(struct iommu_domain *dom)
+{
+	struct protection_domain *domain;
+
+	domain = kzalloc(sizeof(*domain), GFP_KERNEL);
+	if (!domain)
+		return -ENOMEM;
+
+	spin_lock_init(&domain->lock);
+	domain->mode = PAGE_MODE_3_LEVEL;
+	domain->id = domain_id_alloc();
+	if (!domain->id)
+		goto out_free;
+	domain->pt_root = (void *)get_zeroed_page(GFP_KERNEL);
+	if (!domain->pt_root)
+		goto out_free;
+
+	dom->priv = domain;
+
+	return 0;
+
+out_free:
+	kfree(domain);
+
+	return -ENOMEM;
+}
+
+static void amd_iommu_domain_destroy(struct iommu_domain *dom)
+{
+	struct protection_domain *domain = dom->priv;
+
+	if (!domain)
+		return;
+
+	if (domain->dev_cnt > 0)
+		cleanup_domain(domain);
+
+	BUG_ON(domain->dev_cnt != 0);
+
+	free_pagetable(domain);
+
+	domain_id_free(domain->id);
+
+	kfree(domain);
+
+	dom->priv = NULL;
+}
+
+static void amd_iommu_detach_device(struct iommu_domain *dom,
+				    struct device *dev)
+{
+	struct protection_domain *domain = dom->priv;
+	struct amd_iommu *iommu;
+	struct pci_dev *pdev;
+	u16 devid;
+
+	if (dev->bus != &pci_bus_type)
+		return;
+
+	pdev = to_pci_dev(dev);
+
+	devid = calc_devid(pdev->bus->number, pdev->devfn);
+
+	if (devid > 0)
+		detach_device(domain, devid);
+
+	iommu = amd_iommu_rlookup_table[devid];
+	if (!iommu)
+		return;
+
+	iommu_queue_inv_dev_entry(iommu, devid);
+	iommu_completion_wait(iommu);
+}
+
+static int amd_iommu_attach_device(struct iommu_domain *dom,
+				   struct device *dev)
+{
+	struct protection_domain *domain = dom->priv;
+	struct protection_domain *old_domain;
+	struct amd_iommu *iommu;
+	struct pci_dev *pdev;
+	u16 devid;
+
+	if (dev->bus != &pci_bus_type)
+		return -EINVAL;
+
+	pdev = to_pci_dev(dev);
+
+	devid = calc_devid(pdev->bus->number, pdev->devfn);
+
+	if (devid >= amd_iommu_last_bdf ||
+			devid != amd_iommu_alias_table[devid])
+		return -EINVAL;
+
+	iommu = amd_iommu_rlookup_table[devid];
+	if (!iommu)
+		return -EINVAL;
+
+	old_domain = domain_for_device(devid);
+	if (old_domain)
+		return -EBUSY;
+
+	attach_device(iommu, domain, devid);
+
+	iommu_completion_wait(iommu);
+
+	return 0;
+}
+
+static int amd_iommu_map_range(struct iommu_domain *dom,
+			       unsigned long iova, phys_addr_t paddr,
+			       size_t size, int iommu_prot)
+{
+	struct protection_domain *domain = dom->priv;
+	unsigned long i,  npages = iommu_num_pages(paddr, size, PAGE_SIZE);
+	int prot = 0;
+	int ret;
+
+	if (iommu_prot & IOMMU_READ)
+		prot |= IOMMU_PROT_IR;
+	if (iommu_prot & IOMMU_WRITE)
+		prot |= IOMMU_PROT_IW;
+
+	iova  &= PAGE_MASK;
+	paddr &= PAGE_MASK;
+
+	for (i = 0; i < npages; ++i) {
+		ret = iommu_map_page(domain, iova, paddr, prot);
+		if (ret)
+			return ret;
+
+		iova  += PAGE_SIZE;
+		paddr += PAGE_SIZE;
+	}
+
+	return 0;
+}
+
+static void amd_iommu_unmap_range(struct iommu_domain *dom,
+				  unsigned long iova, size_t size)
+{
+
+	struct protection_domain *domain = dom->priv;
+	unsigned long i,  npages = iommu_num_pages(iova, size, PAGE_SIZE);
+
+	iova  &= PAGE_MASK;
+
+	for (i = 0; i < npages; ++i) {
+		iommu_unmap_page(domain, iova);
+		iova  += PAGE_SIZE;
+	}
+
+	iommu_flush_domain(domain->id);
+}
+
+static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
+					  unsigned long iova)
+{
+	struct protection_domain *domain = dom->priv;
+	unsigned long offset = iova & ~PAGE_MASK;
+	phys_addr_t paddr;
+	u64 *pte;
+
+	pte = &domain->pt_root[IOMMU_PTE_L2_INDEX(iova)];
+
+	if (!IOMMU_PTE_PRESENT(*pte))
+		return 0;
+
+	pte = IOMMU_PTE_PAGE(*pte);
+	pte = &pte[IOMMU_PTE_L1_INDEX(iova)];
+
+	if (!IOMMU_PTE_PRESENT(*pte))
+		return 0;
+
+	pte = IOMMU_PTE_PAGE(*pte);
+	pte = &pte[IOMMU_PTE_L0_INDEX(iova)];
+
+	if (!IOMMU_PTE_PRESENT(*pte))
+		return 0;
+
+	paddr  = *pte & IOMMU_PAGE_MASK;
+	paddr |= offset;
+
+	return paddr;
+}
+
+static struct iommu_ops amd_iommu_ops = {
+	.domain_init = amd_iommu_domain_init,
+	.domain_destroy = amd_iommu_domain_destroy,
+	.attach_dev = amd_iommu_attach_device,
+	.detach_dev = amd_iommu_detach_device,
+	.map = amd_iommu_map_range,
+	.unmap = amd_iommu_unmap_range,
+	.iova_to_phys = amd_iommu_iova_to_phys,
+};
+

arch/x86/kernel/amd_iommu_init.c

@@ -122,7 +122,8 @@ u16 amd_iommu_last_bdf;			/* largest PCI device id we have
 LIST_HEAD(amd_iommu_unity_map);		/* a list of required unity mappings
 					   we find in ACPI */
 unsigned amd_iommu_aperture_order = 26; /* size of aperture in power of 2 */
-int amd_iommu_isolate = 1;		/* if 1, device isolation is enabled */
+bool amd_iommu_isolate = true;		/* if true, device isolation is
+					   enabled */
 bool amd_iommu_unmap_flush;		/* if true, flush on every unmap */
 
 LIST_HEAD(amd_iommu_list);		/* list of all AMD IOMMUs in the
@@ -245,12 +246,8 @@ static void __init iommu_feature_disable(struct amd_iommu *iommu, u8 bit)
 /* Function to enable the hardware */
 static void __init iommu_enable(struct amd_iommu *iommu)
 {
-	printk(KERN_INFO "AMD IOMMU: Enabling IOMMU "
-	       "at %02x:%02x.%x cap 0x%hx\n",
-	       iommu->dev->bus->number,
-	       PCI_SLOT(iommu->dev->devfn),
-	       PCI_FUNC(iommu->dev->devfn),
-	       iommu->cap_ptr);
+	printk(KERN_INFO "AMD IOMMU: Enabling IOMMU at %s cap 0x%hx\n",
+	       dev_name(&iommu->dev->dev), iommu->cap_ptr);
 
 	iommu_feature_enable(iommu, CONTROL_IOMMU_EN);
 }
@@ -1218,9 +1215,9 @@ static int __init parse_amd_iommu_options(char *str)
 {
 	for (; *str; ++str) {
 		if (strncmp(str, "isolate", 7) == 0)
-			amd_iommu_isolate = 1;
+			amd_iommu_isolate = true;
 		if (strncmp(str, "share", 5) == 0)
-			amd_iommu_isolate = 0;
+			amd_iommu_isolate = false;
 		if (strncmp(str, "fullflush", 9) == 0)
 			amd_iommu_unmap_flush = true;
 	}

arch/x86/kvm/Makefile

@@ -7,8 +7,8 @@ common-objs = $(addprefix ../../../virt/kvm/, kvm_main.o ioapic.o \
 ifeq ($(CONFIG_KVM_TRACE),y)
 common-objs += $(addprefix ../../../virt/kvm/, kvm_trace.o)
 endif
-ifeq ($(CONFIG_DMAR),y)
-common-objs += $(addprefix ../../../virt/kvm/, vtd.o)
+ifeq ($(CONFIG_IOMMU_API),y)
+common-objs += $(addprefix ../../../virt/kvm/, iommu.o)
 endif
 
 EXTRA_CFLAGS += -Ivirt/kvm -Iarch/x86/kvm

arch/x86/kvm/x86.c

@@ -34,6 +34,7 @@
 #include <linux/module.h>
 #include <linux/mman.h>
 #include <linux/highmem.h>
+#include <linux/iommu.h>
 #include <linux/intel-iommu.h>
 
 #include <asm/uaccess.h>
@@ -989,7 +990,7 @@ int kvm_dev_ioctl_check_extension(long ext)
 		r = !tdp_enabled;
 		break;
 	case KVM_CAP_IOMMU:
-		r = intel_iommu_found();
+		r = iommu_found();
 		break;
 	default:
 		r = 0;

drivers/base/Makefile

@@ -11,6 +11,7 @@ obj-$(CONFIG_FW_LOADER)	+= firmware_class.o
 obj-$(CONFIG_NUMA)	+= node.o
 obj-$(CONFIG_MEMORY_HOTPLUG_SPARSE) += memory.o
 obj-$(CONFIG_SMP)	+= topology.o
+obj-$(CONFIG_IOMMU_API) += iommu.o
 ifeq ($(CONFIG_SYSFS),y)
 obj-$(CONFIG_MODULES)	+= module.o
 endif

drivers/base/iommu.c

@@ -0,0 +1,100 @@
+/*
+ * Copyright (C) 2007-2008 Advanced Micro Devices, Inc.
+ * Author: Joerg Roedel <joerg.roedel@amd.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ */
+
+#include <linux/bug.h>
+#include <linux/types.h>
+#include <linux/errno.h>
+#include <linux/iommu.h>
+
+static struct iommu_ops *iommu_ops;
+
+void register_iommu(struct iommu_ops *ops)
+{
+	if (iommu_ops)
+		BUG();
+
+	iommu_ops = ops;
+}
+
+bool iommu_found()
+{
+	return iommu_ops != NULL;
+}
+EXPORT_SYMBOL_GPL(iommu_found);
+
+struct iommu_domain *iommu_domain_alloc(void)
+{
+	struct iommu_domain *domain;
+	int ret;
+
+	domain = kmalloc(sizeof(*domain), GFP_KERNEL);
+	if (!domain)
+		return NULL;
+
+	ret = iommu_ops->domain_init(domain);
+	if (ret)
+		goto out_free;
+
+	return domain;
+
+out_free:
+	kfree(domain);
+
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(iommu_domain_alloc);
+
+void iommu_domain_free(struct iommu_domain *domain)
+{
+	iommu_ops->domain_destroy(domain);
+	kfree(domain);
+}
+EXPORT_SYMBOL_GPL(iommu_domain_free);
+
+int iommu_attach_device(struct iommu_domain *domain, struct device *dev)
+{
+	return iommu_ops->attach_dev(domain, dev);
+}
+EXPORT_SYMBOL_GPL(iommu_attach_device);
+
+void iommu_detach_device(struct iommu_domain *domain, struct device *dev)
+{
+	iommu_ops->detach_dev(domain, dev);
+}
+EXPORT_SYMBOL_GPL(iommu_detach_device);
+
+int iommu_map_range(struct iommu_domain *domain, unsigned long iova,
+		    phys_addr_t paddr, size_t size, int prot)
+{
+	return iommu_ops->map(domain, iova, paddr, size, prot);
+}
+EXPORT_SYMBOL_GPL(iommu_map_range);
+
+void iommu_unmap_range(struct iommu_domain *domain, unsigned long iova,
+		      size_t size)
+{
+	iommu_ops->unmap(domain, iova, size);
+}
+EXPORT_SYMBOL_GPL(iommu_unmap_range);
+
+phys_addr_t iommu_iova_to_phys(struct iommu_domain *domain,
+			       unsigned long iova)
+{
+	return iommu_ops->iova_to_phys(domain, iova);
+}
+EXPORT_SYMBOL_GPL(iommu_iova_to_phys);

drivers/pci/dmar.c

@@ -191,26 +191,17 @@ dmar_parse_one_drhd(struct acpi_dmar_header *header)
 static int __init dmar_parse_dev(struct dmar_drhd_unit *dmaru)
 {
 	struct acpi_dmar_hardware_unit *drhd;
-	static int include_all;
 	int ret = 0;
 
 	drhd = (struct acpi_dmar_hardware_unit *) dmaru->hdr;
 
-	if (!dmaru->include_all)
-		ret = dmar_parse_dev_scope((void *)(drhd + 1),
+	if (dmaru->include_all)
+		return 0;
+
+	ret = dmar_parse_dev_scope((void *)(drhd + 1),
 				((void *)drhd) + drhd->header.length,
 				&dmaru->devices_cnt, &dmaru->devices,
 				drhd->segment);
-	else {
-		/* Only allow one INCLUDE_ALL */
-		if (include_all) {
-			printk(KERN_WARNING PREFIX "Only one INCLUDE_ALL "
-				"device scope is allowed\n");
-			ret = -EINVAL;
-		}
-		include_all = 1;
-	}
-
 	if (ret) {
 		list_del(&dmaru->list);
 		kfree(dmaru);
@@ -384,12 +375,21 @@ int dmar_pci_device_match(struct pci_dev *devices[], int cnt,
 struct dmar_drhd_unit *
 dmar_find_matched_drhd_unit(struct pci_dev *dev)
 {
-	struct dmar_drhd_unit *drhd = NULL;
+	struct dmar_drhd_unit *dmaru = NULL;
+	struct acpi_dmar_hardware_unit *drhd;
 
-	list_for_each_entry(drhd, &dmar_drhd_units, list) {
-		if (drhd->include_all || dmar_pci_device_match(drhd->devices,
-						drhd->devices_cnt, dev))
-			return drhd;
+	list_for_each_entry(dmaru, &dmar_drhd_units, list) {
+		drhd = container_of(dmaru->hdr,
+				    struct acpi_dmar_hardware_unit,
+				    header);
+
+		if (dmaru->include_all &&
+		    drhd->segment == pci_domain_nr(dev->bus))
+			return dmaru;
+
+		if (dmar_pci_device_match(dmaru->devices,
+					  dmaru->devices_cnt, dev))
+			return dmaru;
 	}
 
 	return NULL;
@@ -491,6 +491,7 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
 	int map_size;
 	u32 ver;
 	static int iommu_allocated = 0;
+	int agaw;
 
 	iommu = kzalloc(sizeof(*iommu), GFP_KERNEL);
 	if (!iommu)
@@ -506,6 +507,15 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
 	iommu->cap = dmar_readq(iommu->reg + DMAR_CAP_REG);
 	iommu->ecap = dmar_readq(iommu->reg + DMAR_ECAP_REG);
 
+	agaw = iommu_calculate_agaw(iommu);
+	if (agaw < 0) {
+		printk(KERN_ERR
+			"Cannot get a valid agaw for iommu (seq_id = %d)\n",
+			iommu->seq_id);
+		goto error;
+	}
+	iommu->agaw = agaw;
+
 	/* the registers might be more than one page */
 	map_size = max_t(int, ecap_max_iotlb_offset(iommu->ecap),
 		cap_max_fault_reg_offset(iommu->cap));

drivers/pci/intel-iommu.c

@@ -27,7 +27,6 @@
 #include <linux/slab.h>
 #include <linux/irq.h>
 #include <linux/interrupt.h>
-#include <linux/sysdev.h>
 #include <linux/spinlock.h>
 #include <linux/pci.h>
 #include <linux/dmar.h>
@@ -35,6 +34,7 @@
 #include <linux/mempool.h>
 #include <linux/timer.h>
 #include <linux/iova.h>
+#include <linux/iommu.h>
 #include <linux/intel-iommu.h>
 #include <asm/cacheflush.h>
 #include <asm/iommu.h>
@@ -54,6 +54,195 @@
 
 #define DOMAIN_MAX_ADDR(gaw) ((((u64)1) << gaw) - 1)
 
+#define IOVA_PFN(addr)		((addr) >> PAGE_SHIFT)
+#define DMA_32BIT_PFN		IOVA_PFN(DMA_32BIT_MASK)
+#define DMA_64BIT_PFN		IOVA_PFN(DMA_64BIT_MASK)
+
+/* global iommu list, set NULL for ignored DMAR units */
+static struct intel_iommu **g_iommus;
+
+/*
+ * 0: Present
+ * 1-11: Reserved
+ * 12-63: Context Ptr (12 - (haw-1))
+ * 64-127: Reserved
+ */
+struct root_entry {
+	u64	val;
+	u64	rsvd1;
+};
+#define ROOT_ENTRY_NR (VTD_PAGE_SIZE/sizeof(struct root_entry))
+static inline bool root_present(struct root_entry *root)
+{
+	return (root->val & 1);
+}
+static inline void set_root_present(struct root_entry *root)
+{
+	root->val |= 1;
+}
+static inline void set_root_value(struct root_entry *root, unsigned long value)
+{
+	root->val |= value & VTD_PAGE_MASK;
+}
+
+static inline struct context_entry *
+get_context_addr_from_root(struct root_entry *root)
+{
+	return (struct context_entry *)
+		(root_present(root)?phys_to_virt(
+		root->val & VTD_PAGE_MASK) :
+		NULL);
+}
+
+/*
+ * low 64 bits:
+ * 0: present
+ * 1: fault processing disable
+ * 2-3: translation type
+ * 12-63: address space root
+ * high 64 bits:
+ * 0-2: address width
+ * 3-6: aval
+ * 8-23: domain id
+ */
+struct context_entry {
+	u64 lo;
+	u64 hi;
+};
+
+static inline bool context_present(struct context_entry *context)
+{
+	return (context->lo & 1);
+}
+static inline void context_set_present(struct context_entry *context)
+{
+	context->lo |= 1;
+}
+
+static inline void context_set_fault_enable(struct context_entry *context)
+{
+	context->lo &= (((u64)-1) << 2) | 1;
+}
+
+#define CONTEXT_TT_MULTI_LEVEL 0
+
+static inline void context_set_translation_type(struct context_entry *context,
+						unsigned long value)
+{
+	context->lo &= (((u64)-1) << 4) | 3;
+	context->lo |= (value & 3) << 2;
+}
+
+static inline void context_set_address_root(struct context_entry *context,
+					    unsigned long value)
+{
+	context->lo |= value & VTD_PAGE_MASK;
+}
+
+static inline void context_set_address_width(struct context_entry *context,
+					     unsigned long value)
+{
+	context->hi |= value & 7;
+}
+
+static inline void context_set_domain_id(struct context_entry *context,
+					 unsigned long value)
+{
+	context->hi |= (value & ((1 << 16) - 1)) << 8;
+}
+
+static inline void context_clear_entry(struct context_entry *context)
+{
+	context->lo = 0;
+	context->hi = 0;
+}
+
+/*
+ * 0: readable
+ * 1: writable
+ * 2-6: reserved
+ * 7: super page
+ * 8-11: available
+ * 12-63: Host physcial address
+ */
+struct dma_pte {
+	u64 val;
+};
+
+static inline void dma_clear_pte(struct dma_pte *pte)
+{
+	pte->val = 0;
+}
+
+static inline void dma_set_pte_readable(struct dma_pte *pte)
+{
+	pte->val |= DMA_PTE_READ;
+}
+
+static inline void dma_set_pte_writable(struct dma_pte *pte)
+{
+	pte->val |= DMA_PTE_WRITE;
+}
+
+static inline void dma_set_pte_prot(struct dma_pte *pte, unsigned long prot)
+{
+	pte->val = (pte->val & ~3) | (prot & 3);
+}
+
+static inline u64 dma_pte_addr(struct dma_pte *pte)
+{
+	return (pte->val & VTD_PAGE_MASK);
+}
+
+static inline void dma_set_pte_addr(struct dma_pte *pte, u64 addr)
+{
+	pte->val |= (addr & VTD_PAGE_MASK);
+}
+
+static inline bool dma_pte_present(struct dma_pte *pte)
+{
+	return (pte->val & 3) != 0;
+}
+
+/* devices under the same p2p bridge are owned in one domain */
+#define DOMAIN_FLAG_P2P_MULTIPLE_DEVICES (1 << 0)
+
+/* domain represents a virtual machine, more than one devices
+ * across iommus may be owned in one domain, e.g. kvm guest.
+ */
+#define DOMAIN_FLAG_VIRTUAL_MACHINE	(1 << 1)
+
+struct dmar_domain {
+	int	id;			/* domain id */
+	unsigned long iommu_bmp;	/* bitmap of iommus this domain uses*/
+
+	struct list_head devices; 	/* all devices' list */
+	struct iova_domain iovad;	/* iova's that belong to this domain */
+
+	struct dma_pte	*pgd;		/* virtual address */
+	spinlock_t	mapping_lock;	/* page table lock */
+	int		gaw;		/* max guest address width */
+
+	/* adjusted guest address width, 0 is level 2 30-bit */
+	int		agaw;
+
+	int		flags;		/* flags to find out type of domain */
+
+	int		iommu_coherency;/* indicate coherency of iommu access */
+	int		iommu_count;	/* reference count of iommu */
+	spinlock_t	iommu_lock;	/* protect iommu set in domain */
+	u64		max_addr;	/* maximum mapped address */
+};
+
+/* PCI domain-device relationship */
+struct device_domain_info {
+	struct list_head link;	/* link to domain siblings */
+	struct list_head global; /* link to global list */
+	u8 bus;			/* PCI bus numer */
+	u8 devfn;		/* PCI devfn number */
+	struct pci_dev *dev; /* it's NULL for PCIE-to-PCI bridge */
+	struct dmar_domain *domain; /* pointer to domain */
+};
 
 static void flush_unmaps_timeout(unsigned long data);
 
@@ -88,6 +277,8 @@ static int intel_iommu_strict;
 static DEFINE_SPINLOCK(device_domain_lock);
 static LIST_HEAD(device_domain_list);
 
+static struct iommu_ops intel_iommu_ops;
+
 static int __init intel_iommu_setup(char *str)
 {
 	if (!str)
@@ -184,6 +375,87 @@ void free_iova_mem(struct iova *iova)
 	kmem_cache_free(iommu_iova_cache, iova);
 }
 
+
+static inline int width_to_agaw(int width);
+
+/* calculate agaw for each iommu.
+ * "SAGAW" may be different across iommus, use a default agaw, and
+ * get a supported less agaw for iommus that don't support the default agaw.
+ */
+int iommu_calculate_agaw(struct intel_iommu *iommu)
+{
+	unsigned long sagaw;
+	int agaw = -1;
+
+	sagaw = cap_sagaw(iommu->cap);
+	for (agaw = width_to_agaw(DEFAULT_DOMAIN_ADDRESS_WIDTH);
+	     agaw >= 0; agaw--) {
+		if (test_bit(agaw, &sagaw))
+			break;
+	}
+
+	return agaw;
+}
+
+/* in native case, each domain is related to only one iommu */
+static struct intel_iommu *domain_get_iommu(struct dmar_domain *domain)
+{
+	int iommu_id;
+
+	BUG_ON(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE);
+
+	iommu_id = find_first_bit(&domain->iommu_bmp, g_num_of_iommus);
+	if (iommu_id < 0 || iommu_id >= g_num_of_iommus)
+		return NULL;
+
+	return g_iommus[iommu_id];
+}
+
+/* "Coherency" capability may be different across iommus */
+static void domain_update_iommu_coherency(struct dmar_domain *domain)
+{
+	int i;
+
+	domain->iommu_coherency = 1;
+
+	i = find_first_bit(&domain->iommu_bmp, g_num_of_iommus);
+	for (; i < g_num_of_iommus; ) {
+		if (!ecap_coherent(g_iommus[i]->ecap)) {
+			domain->iommu_coherency = 0;
+			break;
+		}
+		i = find_next_bit(&domain->iommu_bmp, g_num_of_iommus, i+1);
+	}
+}
+
+static struct intel_iommu *device_to_iommu(u8 bus, u8 devfn)
+{
+	struct dmar_drhd_unit *drhd = NULL;
+	int i;
+
+	for_each_drhd_unit(drhd) {
+		if (drhd->ignored)
+			continue;
+
+		for (i = 0; i < drhd->devices_cnt; i++)
+			if (drhd->devices[i]->bus->number == bus &&
+			    drhd->devices[i]->devfn == devfn)
+				return drhd->iommu;
+
+		if (drhd->include_all)
+			return drhd->iommu;
+	}
+
+	return NULL;
+}
+
+static void domain_flush_cache(struct dmar_domain *domain,
+			       void *addr, int size)
+{
+	if (!domain->iommu_coherency)
+		clflush_cache_range(addr, size);
+}
+
 /* Gets context entry for a given bus and devfn */
 static struct context_entry * device_to_context_entry(struct intel_iommu *iommu,
 		u8 bus, u8 devfn)
@@ -226,7 +498,7 @@ static int device_context_mapped(struct intel_iommu *iommu, u8 bus, u8 devfn)
 		ret = 0;
 		goto out;
 	}
-	ret = context_present(context[devfn]);
+	ret = context_present(&context[devfn]);
 out:
 	spin_unlock_irqrestore(&iommu->lock, flags);
 	return ret;
@@ -242,7 +514,7 @@ static void clear_context_table(struct intel_iommu *iommu, u8 bus, u8 devfn)
 	root = &iommu->root_entry[bus];
 	context = get_context_addr_from_root(root);
 	if (context) {
-		context_clear_entry(context[devfn]);
+		context_clear_entry(&context[devfn]);
 		__iommu_flush_cache(iommu, &context[devfn], \
 			sizeof(*context));
 	}
@@ -339,7 +611,7 @@ static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr)
 		if (level == 1)
 			break;
 
-		if (!dma_pte_present(*pte)) {
+		if (!dma_pte_present(pte)) {
 			tmp_page = alloc_pgtable_page();
 
 			if (!tmp_page) {
@@ -347,18 +619,17 @@ static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr)
 					flags);
 				return NULL;
 			}
-			__iommu_flush_cache(domain->iommu, tmp_page,
-					PAGE_SIZE);
-			dma_set_pte_addr(*pte, virt_to_phys(tmp_page));
+			domain_flush_cache(domain, tmp_page, PAGE_SIZE);
+			dma_set_pte_addr(pte, virt_to_phys(tmp_page));
 			/*
 			 * high level table always sets r/w, last level page
 			 * table control read/write
 			 */
-			dma_set_pte_readable(*pte);
-			dma_set_pte_writable(*pte);
-			__iommu_flush_cache(domain->iommu, pte, sizeof(*pte));
+			dma_set_pte_readable(pte);
+			dma_set_pte_writable(pte);
+			domain_flush_cache(domain, pte, sizeof(*pte));
 		}
-		parent = phys_to_virt(dma_pte_addr(*pte));
+		parent = phys_to_virt(dma_pte_addr(pte));
 		level--;
 	}
 
@@ -381,9 +652,9 @@ static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
 		if (level == total)
 			return pte;
 
-		if (!dma_pte_present(*pte))
+		if (!dma_pte_present(pte))
 			break;
-		parent = phys_to_virt(dma_pte_addr(*pte));
+		parent = phys_to_virt(dma_pte_addr(pte));
 		total--;
 	}
 	return NULL;
@@ -398,8 +669,8 @@ static void dma_pte_clear_one(struct dmar_domain *domain, u64 addr)
 	pte = dma_addr_level_pte(domain, addr, 1);
 
 	if (pte) {
-		dma_clear_pte(*pte);
-		__iommu_flush_cache(domain->iommu, pte, sizeof(*pte));
+		dma_clear_pte(pte);
+		domain_flush_cache(domain, pte, sizeof(*pte));
 	}
 }
 
@@ -445,10 +716,9 @@ static void dma_pte_free_pagetable(struct dmar_domain *domain,
 			pte = dma_addr_level_pte(domain, tmp, level);
 			if (pte) {
 				free_pgtable_page(
-					phys_to_virt(dma_pte_addr(*pte)));
-				dma_clear_pte(*pte);
-				__iommu_flush_cache(domain->iommu,
-						pte, sizeof(*pte));
+					phys_to_virt(dma_pte_addr(pte)));
+				dma_clear_pte(pte);
+				domain_flush_cache(domain, pte, sizeof(*pte));
 			}
 			tmp += level_size(level);
 		}
@@ -950,17 +1220,28 @@ static int iommu_init_domains(struct intel_iommu *iommu)
 
 
 static void domain_exit(struct dmar_domain *domain);
+static void vm_domain_exit(struct dmar_domain *domain);
 
 void free_dmar_iommu(struct intel_iommu *iommu)
 {
 	struct dmar_domain *domain;
 	int i;
+	unsigned long flags;
 
 	i = find_first_bit(iommu->domain_ids, cap_ndoms(iommu->cap));
 	for (; i < cap_ndoms(iommu->cap); ) {
 		domain = iommu->domains[i];
 		clear_bit(i, iommu->domain_ids);
-		domain_exit(domain);
+
+		spin_lock_irqsave(&domain->iommu_lock, flags);
+		if (--domain->iommu_count == 0) {
+			if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE)
+				vm_domain_exit(domain);
+			else
+				domain_exit(domain);
+		}
+		spin_unlock_irqrestore(&domain->iommu_lock, flags);
+
 		i = find_next_bit(iommu->domain_ids,
 			cap_ndoms(iommu->cap), i+1);
 	}
@@ -978,6 +1259,17 @@ void free_dmar_iommu(struct intel_iommu *iommu)
 	kfree(iommu->domains);
 	kfree(iommu->domain_ids);
 
+	g_iommus[iommu->seq_id] = NULL;
+
+	/* if all iommus are freed, free g_iommus */
+	for (i = 0; i < g_num_of_iommus; i++) {
+		if (g_iommus[i])
+			break;
+	}
+
+	if (i == g_num_of_iommus)
+		kfree(g_iommus);
+
 	/* free context mapping */
 	free_context_table(iommu);
 }
@@ -1006,7 +1298,9 @@ static struct dmar_domain * iommu_alloc_domain(struct intel_iommu *iommu)
 
 	set_bit(num, iommu->domain_ids);
 	domain->id = num;
-	domain->iommu = iommu;
+	memset(&domain->iommu_bmp, 0, sizeof(unsigned long));
+	set_bit(iommu->seq_id, &domain->iommu_bmp);
+	domain->flags = 0;
 	iommu->domains[num] = domain;
 	spin_unlock_irqrestore(&iommu->lock, flags);
 
@@ -1016,10 +1310,13 @@ static struct dmar_domain * iommu_alloc_domain(struct intel_iommu *iommu)
 static void iommu_free_domain(struct dmar_domain *domain)
 {
 	unsigned long flags;
+	struct intel_iommu *iommu;
+
+	iommu = domain_get_iommu(domain);
 
-	spin_lock_irqsave(&domain->iommu->lock, flags);
-	clear_bit(domain->id, domain->iommu->domain_ids);
-	spin_unlock_irqrestore(&domain->iommu->lock, flags);
+	spin_lock_irqsave(&iommu->lock, flags);
+	clear_bit(domain->id, iommu->domain_ids);
+	spin_unlock_irqrestore(&iommu->lock, flags);
 }
 
 static struct iova_domain reserved_iova_list;
@@ -1094,11 +1391,12 @@ static int domain_init(struct dmar_domain *domain, int guest_width)
 
 	init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
 	spin_lock_init(&domain->mapping_lock);
+	spin_lock_init(&domain->iommu_lock);
 
 	domain_reserve_special_ranges(domain);
 
 	/* calculate AGAW */
-	iommu = domain->iommu;
+	iommu = domain_get_iommu(domain);
 	if (guest_width > cap_mgaw(iommu->cap))
 		guest_width = cap_mgaw(iommu->cap);
 	domain->gaw = guest_width;
@@ -1115,6 +1413,13 @@ static int domain_init(struct dmar_domain *domain, int guest_width)
 	domain->agaw = agaw;
 	INIT_LIST_HEAD(&domain->devices);
 
+	if (ecap_coherent(iommu->ecap))
+		domain->iommu_coherency = 1;
+	else
+		domain->iommu_coherency = 0;
+
+	domain->iommu_count = 1;
+
 	/* always allocate the top pgd */
 	domain->pgd = (struct dma_pte *)alloc_pgtable_page();
 	if (!domain->pgd)
@@ -1151,28 +1456,82 @@ static int domain_context_mapping_one(struct dmar_domain *domain,
 		u8 bus, u8 devfn)
 {
 	struct context_entry *context;
-	struct intel_iommu *iommu = domain->iommu;
 	unsigned long flags;
+	struct intel_iommu *iommu;
+	struct dma_pte *pgd;
+	unsigned long num;
+	unsigned long ndomains;
+	int id;
+	int agaw;
 
 	pr_debug("Set context mapping for %02x:%02x.%d\n",
 		bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
 	BUG_ON(!domain->pgd);
+
+	iommu = device_to_iommu(bus, devfn);
+	if (!iommu)
+		return -ENODEV;
+
 	context = device_to_context_entry(iommu, bus, devfn);
 	if (!context)
 		return -ENOMEM;
 	spin_lock_irqsave(&iommu->lock, flags);
-	if (context_present(*context)) {
+	if (context_present(context)) {
 		spin_unlock_irqrestore(&iommu->lock, flags);
 		return 0;
 	}
 
-	context_set_domain_id(*context, domain->id);
-	context_set_address_width(*context, domain->agaw);
-	context_set_address_root(*context, virt_to_phys(domain->pgd));
-	context_set_translation_type(*context, CONTEXT_TT_MULTI_LEVEL);
-	context_set_fault_enable(*context);
-	context_set_present(*context);
-	__iommu_flush_cache(iommu, context, sizeof(*context));
+	id = domain->id;
+	pgd = domain->pgd;
+
+	if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) {
+		int found = 0;
+
+		/* find an available domain id for this device in iommu */
+		ndomains = cap_ndoms(iommu->cap);
+		num = find_first_bit(iommu->domain_ids, ndomains);
+		for (; num < ndomains; ) {
+			if (iommu->domains[num] == domain) {
+				id = num;
+				found = 1;
+				break;
+			}
+			num = find_next_bit(iommu->domain_ids,
+					    cap_ndoms(iommu->cap), num+1);
+		}
+
+		if (found == 0) {
+			num = find_first_zero_bit(iommu->domain_ids, ndomains);
+			if (num >= ndomains) {
+				spin_unlock_irqrestore(&iommu->lock, flags);
+				printk(KERN_ERR "IOMMU: no free domain ids\n");
+				return -EFAULT;
+			}
+
+			set_bit(num, iommu->domain_ids);
+			iommu->domains[num] = domain;
+			id = num;
+		}
+
+		/* Skip top levels of page tables for
+		 * iommu which has less agaw than default.
+		 */
+		for (agaw = domain->agaw; agaw != iommu->agaw; agaw--) {
+			pgd = phys_to_virt(dma_pte_addr(pgd));
+			if (!dma_pte_present(pgd)) {
+				spin_unlock_irqrestore(&iommu->lock, flags);
+				return -ENOMEM;
+			}
+		}
+	}
+
+	context_set_domain_id(context, id);
+	context_set_address_width(context, iommu->agaw);
+	context_set_address_root(context, virt_to_phys(pgd));
+	context_set_translation_type(context, CONTEXT_TT_MULTI_LEVEL);
+	context_set_fault_enable(context);
+	context_set_present(context);
+	domain_flush_cache(domain, context, sizeof(*context));
 
 	/* it's a non-present to present mapping */
 	if (iommu->flush.flush_context(iommu, domain->id,
@@ -1183,6 +1542,13 @@ static int domain_context_mapping_one(struct dmar_domain *domain,
 		iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_DSI_FLUSH, 0);
 
 	spin_unlock_irqrestore(&iommu->lock, flags);
+
+	spin_lock_irqsave(&domain->iommu_lock, flags);
+	if (!test_and_set_bit(iommu->seq_id, &domain->iommu_bmp)) {
+		domain->iommu_count++;
+		domain_update_iommu_coherency(domain);
+	}
+	spin_unlock_irqrestore(&domain->iommu_lock, flags);
 	return 0;
 }
 
@@ -1218,13 +1584,17 @@ domain_context_mapping(struct dmar_domain *domain, struct pci_dev *pdev)
 			tmp->bus->number, tmp->devfn);
 }
 
-static int domain_context_mapped(struct dmar_domain *domain,
-	struct pci_dev *pdev)
+static int domain_context_mapped(struct pci_dev *pdev)
 {
 	int ret;
 	struct pci_dev *tmp, *parent;
+	struct intel_iommu *iommu;
+
+	iommu = device_to_iommu(pdev->bus->number, pdev->devfn);
+	if (!iommu)
+		return -ENODEV;
 
-	ret = device_context_mapped(domain->iommu,
+	ret = device_context_mapped(iommu,
 		pdev->bus->number, pdev->devfn);
 	if (!ret)
 		return ret;
@@ -1235,17 +1605,17 @@ static int domain_context_mapped(struct dmar_domain *domain,
 	/* Secondary interface's bus number and devfn 0 */
 	parent = pdev->bus->self;
 	while (parent != tmp) {
-		ret = device_context_mapped(domain->iommu, parent->bus->number,
+		ret = device_context_mapped(iommu, parent->bus->number,
 			parent->devfn);
 		if (!ret)
 			return ret;
 		parent = parent->bus->self;
 	}
 	if (tmp->is_pcie)
-		return device_context_mapped(domain->iommu,
+		return device_context_mapped(iommu,
 			tmp->subordinate->number, 0);
 	else
-		return device_context_mapped(domain->iommu,
+		return device_context_mapped(iommu,
 			tmp->bus->number, tmp->devfn);
 }
 
@@ -1273,22 +1643,25 @@ domain_page_mapping(struct dmar_domain *domain, dma_addr_t iova,
 		/* We don't need lock here, nobody else
 		 * touches the iova range
 		 */
-		BUG_ON(dma_pte_addr(*pte));
-		dma_set_pte_addr(*pte, start_pfn << VTD_PAGE_SHIFT);
-		dma_set_pte_prot(*pte, prot);
-		__iommu_flush_cache(domain->iommu, pte, sizeof(*pte));
+		BUG_ON(dma_pte_addr(pte));
+		dma_set_pte_addr(pte, start_pfn << VTD_PAGE_SHIFT);
+		dma_set_pte_prot(pte, prot);
+		domain_flush_cache(domain, pte, sizeof(*pte));
 		start_pfn++;
 		index++;
 	}
 	return 0;
 }
 
-static void detach_domain_for_dev(struct dmar_domain *domain, u8 bus, u8 devfn)
+static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn)
 {
-	clear_context_table(domain->iommu, bus, devfn);
-	domain->iommu->flush.flush_context(domain->iommu, 0, 0, 0,
+	if (!iommu)
+		return;
+
+	clear_context_table(iommu, bus, devfn);
+	iommu->flush.flush_context(iommu, 0, 0, 0,
 					   DMA_CCMD_GLOBAL_INVL, 0);
-	domain->iommu->flush.flush_iotlb(domain->iommu, 0, 0, 0,
+	iommu->flush.flush_iotlb(iommu, 0, 0, 0,
 					 DMA_TLB_GLOBAL_FLUSH, 0);
 }
 
@@ -1296,6 +1669,7 @@ static void domain_remove_dev_info(struct dmar_domain *domain)
 {
 	struct device_domain_info *info;
 	unsigned long flags;
+	struct intel_iommu *iommu;
 
 	spin_lock_irqsave(&device_domain_lock, flags);
 	while (!list_empty(&domain->devices)) {
@@ -1307,7 +1681,8 @@ static void domain_remove_dev_info(struct dmar_domain *domain)
 			info->dev->dev.archdata.iommu = NULL;
 		spin_unlock_irqrestore(&device_domain_lock, flags);
 
-		detach_domain_for_dev(info->domain, info->bus, info->devfn);
+		iommu = device_to_iommu(info->bus, info->devfn);
+		iommu_detach_dev(iommu, info->bus, info->devfn);
 		free_devinfo_mem(info);
 
 		spin_lock_irqsave(&device_domain_lock, flags);
@@ -1400,7 +1775,7 @@ static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw)
 		info->dev = NULL;
 		info->domain = domain;
 		/* This domain is shared by devices under p2p bridge */
-		domain->flags |= DOMAIN_FLAG_MULTIPLE_DEVICES;
+		domain->flags |= DOMAIN_FLAG_P2P_MULTIPLE_DEVICES;
 
 		/* pcie-to-pci bridge already has a domain, uses it */
 		found = NULL;
@@ -1563,6 +1938,11 @@ static void __init iommu_prepare_gfx_mapping(void)
 			printk(KERN_ERR "IOMMU: mapping reserved region failed\n");
 	}
 }
+#else /* !CONFIG_DMAR_GFX_WA */
+static inline void iommu_prepare_gfx_mapping(void)
+{
+	return;
+}
 #endif
 
 #ifdef CONFIG_DMAR_FLOPPY_WA
@@ -1590,7 +1970,7 @@ static inline void iommu_prepare_isa(void)
 }
 #endif /* !CONFIG_DMAR_FLPY_WA */
 
-int __init init_dmars(void)
+static int __init init_dmars(void)
 {
 	struct dmar_drhd_unit *drhd;
 	struct dmar_rmrr_unit *rmrr;
@@ -1613,9 +1993,18 @@ int __init init_dmars(void)
 		 */
 	}
 
+	g_iommus = kcalloc(g_num_of_iommus, sizeof(struct intel_iommu *),
+			GFP_KERNEL);
+	if (!g_iommus) {
+		printk(KERN_ERR "Allocating global iommu array failed\n");
+		ret = -ENOMEM;
+		goto error;
+	}
+
 	deferred_flush = kzalloc(g_num_of_iommus *
 		sizeof(struct deferred_flush_tables), GFP_KERNEL);
 	if (!deferred_flush) {
+		kfree(g_iommus);
 		ret = -ENOMEM;
 		goto error;
 	}
@@ -1625,6 +2014,7 @@ int __init init_dmars(void)
 			continue;
 
 		iommu = drhd->iommu;
+		g_iommus[iommu->seq_id] = iommu;
 
 		ret = iommu_init_domains(iommu);
 		if (ret)
@@ -1737,6 +2127,7 @@ error:
 		iommu = drhd->iommu;
 		free_iommu(iommu);
 	}
+	kfree(g_iommus);
 	return ret;
 }
 
@@ -1805,7 +2196,7 @@ get_valid_domain_for_dev(struct pci_dev *pdev)
 	}
 
 	/* make sure context mapping is ok */
-	if (unlikely(!domain_context_mapped(domain, pdev))) {
+	if (unlikely(!domain_context_mapped(pdev))) {
 		ret = domain_context_mapping(domain, pdev);
 		if (ret) {
 			printk(KERN_ERR
@@ -1827,6 +2218,7 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
 	struct iova *iova;
 	int prot = 0;
 	int ret;
+	struct intel_iommu *iommu;
 
 	BUG_ON(dir == DMA_NONE);
 	if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO)
@@ -1836,6 +2228,7 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
 	if (!domain)
 		return 0;
 
+	iommu = domain_get_iommu(domain);
 	size = aligned_size((u64)paddr, size);
 
 	iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
@@ -1849,7 +2242,7 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
 	 * mappings..
 	 */
 	if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \
-			!cap_zlr(domain->iommu->cap))
+			!cap_zlr(iommu->cap))
 		prot |= DMA_PTE_READ;
 	if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
 		prot |= DMA_PTE_WRITE;
@@ -1865,10 +2258,10 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
 		goto error;
 
 	/* it's a non-present to present mapping */
-	ret = iommu_flush_iotlb_psi(domain->iommu, domain->id,
+	ret = iommu_flush_iotlb_psi(iommu, domain->id,
 			start_paddr, size >> VTD_PAGE_SHIFT, 1);
 	if (ret)
-		iommu_flush_write_buffer(domain->iommu);
+		iommu_flush_write_buffer(iommu);
 
 	return start_paddr + ((u64)paddr & (~PAGE_MASK));
 
@@ -1895,10 +2288,11 @@ static void flush_unmaps(void)
 
 	/* just flush them all */
 	for (i = 0; i < g_num_of_iommus; i++) {
-		if (deferred_flush[i].next) {
-			struct intel_iommu *iommu =
-				deferred_flush[i].domain[0]->iommu;
+		struct intel_iommu *iommu = g_iommus[i];
+		if (!iommu)
+			continue;
 
+		if (deferred_flush[i].next) {
 			iommu->flush.flush_iotlb(iommu, 0, 0, 0,
 						 DMA_TLB_GLOBAL_FLUSH, 0);
 			for (j = 0; j < deferred_flush[i].next; j++) {
@@ -1925,12 +2319,14 @@ static void add_unmap(struct dmar_domain *dom, struct iova *iova)
 {
 	unsigned long flags;
 	int next, iommu_id;
+	struct intel_iommu *iommu;
 
 	spin_lock_irqsave(&async_umap_flush_lock, flags);
 	if (list_size == HIGH_WATER_MARK)
 		flush_unmaps();
 
-	iommu_id = dom->iommu->seq_id;
+	iommu = domain_get_iommu(dom);
+	iommu_id = iommu->seq_id;
 
 	next = deferred_flush[iommu_id].next;
 	deferred_flush[iommu_id].domain[next] = dom;
@@ -1952,12 +2348,15 @@ void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t size,
 	struct dmar_domain *domain;
 	unsigned long start_addr;
 	struct iova *iova;
+	struct intel_iommu *iommu;
 
 	if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO)
 		return;
 	domain = find_domain(pdev);
 	BUG_ON(!domain);
 
+	iommu = domain_get_iommu(domain);
+
 	iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr));
 	if (!iova)
 		return;
@@ -1973,9 +2372,9 @@ void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t size,
 	/* free page tables */
 	dma_pte_free_pagetable(domain, start_addr, start_addr + size);
 	if (intel_iommu_strict) {
-		if (iommu_flush_iotlb_psi(domain->iommu,
+		if (iommu_flush_iotlb_psi(iommu,
 			domain->id, start_addr, size >> VTD_PAGE_SHIFT, 0))
-			iommu_flush_write_buffer(domain->iommu);
+			iommu_flush_write_buffer(iommu);
 		/* free iova */
 		__free_iova(&domain->iovad, iova);
 	} else {
@@ -2036,11 +2435,15 @@ void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
 	size_t size = 0;
 	void *addr;
 	struct scatterlist *sg;
+	struct intel_iommu *iommu;
 
 	if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO)
 		return;
 
 	domain = find_domain(pdev);
+	BUG_ON(!domain);
+
+	iommu = domain_get_iommu(domain);
 
 	iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address));
 	if (!iova)
@@ -2057,9 +2460,9 @@ void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
 	/* free page tables */
 	dma_pte_free_pagetable(domain, start_addr, start_addr + size);
 
-	if (iommu_flush_iotlb_psi(domain->iommu, domain->id, start_addr,
+	if (iommu_flush_iotlb_psi(iommu, domain->id, start_addr,
 			size >> VTD_PAGE_SHIFT, 0))
-		iommu_flush_write_buffer(domain->iommu);
+		iommu_flush_write_buffer(iommu);
 
 	/* free iova */
 	__free_iova(&domain->iovad, iova);
@@ -2093,6 +2496,7 @@ int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
 	int ret;
 	struct scatterlist *sg;
 	unsigned long start_addr;
+	struct intel_iommu *iommu;
 
 	BUG_ON(dir == DMA_NONE);
 	if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO)
@@ -2102,6 +2506,8 @@ int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
 	if (!domain)
 		return 0;
 
+	iommu = domain_get_iommu(domain);
+
 	for_each_sg(sglist, sg, nelems, i) {
 		addr = SG_ENT_VIRT_ADDRESS(sg);
 		addr = (void *)virt_to_phys(addr);
@@ -2119,7 +2525,7 @@ int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
 	 * mappings..
 	 */
 	if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \
-			!cap_zlr(domain->iommu->cap))
+			!cap_zlr(iommu->cap))
 		prot |= DMA_PTE_READ;
 	if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
 		prot |= DMA_PTE_WRITE;
@@ -2151,9 +2557,9 @@ int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
 	}
 
 	/* it's a non-present to present mapping */
-	if (iommu_flush_iotlb_psi(domain->iommu, domain->id,
+	if (iommu_flush_iotlb_psi(iommu, domain->id,
 			start_addr, offset >> VTD_PAGE_SHIFT, 1))
-		iommu_flush_write_buffer(domain->iommu);
+		iommu_flush_write_buffer(iommu);
 	return nelems;
 }
 
@@ -2325,10 +2731,220 @@ int __init intel_iommu_init(void)
 	init_timer(&unmap_timer);
 	force_iommu = 1;
 	dma_ops = &intel_dma_ops;
+
+	register_iommu(&intel_iommu_ops);
+
+	return 0;
+}
+
+static int vm_domain_add_dev_info(struct dmar_domain *domain,
+				  struct pci_dev *pdev)
+{
+	struct device_domain_info *info;
+	unsigned long flags;
+
+	info = alloc_devinfo_mem();
+	if (!info)
+		return -ENOMEM;
+
+	info->bus = pdev->bus->number;
+	info->devfn = pdev->devfn;
+	info->dev = pdev;
+	info->domain = domain;
+
+	spin_lock_irqsave(&device_domain_lock, flags);
+	list_add(&info->link, &domain->devices);
+	list_add(&info->global, &device_domain_list);
+	pdev->dev.archdata.iommu = info;
+	spin_unlock_irqrestore(&device_domain_lock, flags);
+
+	return 0;
+}
+
+static void vm_domain_remove_one_dev_info(struct dmar_domain *domain,
+					  struct pci_dev *pdev)
+{
+	struct device_domain_info *info;
+	struct intel_iommu *iommu;
+	unsigned long flags;
+	int found = 0;
+	struct list_head *entry, *tmp;
+
+	iommu = device_to_iommu(pdev->bus->number, pdev->devfn);
+	if (!iommu)
+		return;
+
+	spin_lock_irqsave(&device_domain_lock, flags);
+	list_for_each_safe(entry, tmp, &domain->devices) {
+		info = list_entry(entry, struct device_domain_info, link);
+		if (info->bus == pdev->bus->number &&
+		    info->devfn == pdev->devfn) {
+			list_del(&info->link);
+			list_del(&info->global);
+			if (info->dev)
+				info->dev->dev.archdata.iommu = NULL;
+			spin_unlock_irqrestore(&device_domain_lock, flags);
+
+			iommu_detach_dev(iommu, info->bus, info->devfn);
+			free_devinfo_mem(info);
+
+			spin_lock_irqsave(&device_domain_lock, flags);
+
+			if (found)
+				break;
+			else
+				continue;
+		}
+
+		/* if there is no other devices under the same iommu
+		 * owned by this domain, clear this iommu in iommu_bmp
+		 * update iommu count and coherency
+		 */
+		if (device_to_iommu(info->bus, info->devfn) == iommu)
+			found = 1;
+	}
+
+	if (found == 0) {
+		unsigned long tmp_flags;
+		spin_lock_irqsave(&domain->iommu_lock, tmp_flags);
+		clear_bit(iommu->seq_id, &domain->iommu_bmp);
+		domain->iommu_count--;
+		domain_update_iommu_coherency(domain);
+		spin_unlock_irqrestore(&domain->iommu_lock, tmp_flags);
+	}
+
+	spin_unlock_irqrestore(&device_domain_lock, flags);
+}
+
+static void vm_domain_remove_all_dev_info(struct dmar_domain *domain)
+{
+	struct device_domain_info *info;
+	struct intel_iommu *iommu;
+	unsigned long flags1, flags2;
+
+	spin_lock_irqsave(&device_domain_lock, flags1);
+	while (!list_empty(&domain->devices)) {
+		info = list_entry(domain->devices.next,
+			struct device_domain_info, link);
+		list_del(&info->link);
+		list_del(&info->global);
+		if (info->dev)
+			info->dev->dev.archdata.iommu = NULL;
+
+		spin_unlock_irqrestore(&device_domain_lock, flags1);
+
+		iommu = device_to_iommu(info->bus, info->devfn);
+		iommu_detach_dev(iommu, info->bus, info->devfn);
+
+		/* clear this iommu in iommu_bmp, update iommu count
+		 * and coherency
+		 */
+		spin_lock_irqsave(&domain->iommu_lock, flags2);
+		if (test_and_clear_bit(iommu->seq_id,
+				       &domain->iommu_bmp)) {
+			domain->iommu_count--;
+			domain_update_iommu_coherency(domain);
+		}
+		spin_unlock_irqrestore(&domain->iommu_lock, flags2);
+
+		free_devinfo_mem(info);
+		spin_lock_irqsave(&device_domain_lock, flags1);
+	}
+	spin_unlock_irqrestore(&device_domain_lock, flags1);
+}
+
+/* domain id for virtual machine, it won't be set in context */
+static unsigned long vm_domid;
+
+static int vm_domain_min_agaw(struct dmar_domain *domain)
+{
+	int i;
+	int min_agaw = domain->agaw;
+
+	i = find_first_bit(&domain->iommu_bmp, g_num_of_iommus);
+	for (; i < g_num_of_iommus; ) {
+		if (min_agaw > g_iommus[i]->agaw)
+			min_agaw = g_iommus[i]->agaw;
+
+		i = find_next_bit(&domain->iommu_bmp, g_num_of_iommus, i+1);
+	}
+
+	return min_agaw;
+}
+
+static struct dmar_domain *iommu_alloc_vm_domain(void)
+{
+	struct dmar_domain *domain;
+
+	domain = alloc_domain_mem();
+	if (!domain)
+		return NULL;
+
+	domain->id = vm_domid++;
+	memset(&domain->iommu_bmp, 0, sizeof(unsigned long));
+	domain->flags = DOMAIN_FLAG_VIRTUAL_MACHINE;
+
+	return domain;
+}
+
+static int vm_domain_init(struct dmar_domain *domain, int guest_width)
+{
+	int adjust_width;
+
+	init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
+	spin_lock_init(&domain->mapping_lock);
+	spin_lock_init(&domain->iommu_lock);
+
+	domain_reserve_special_ranges(domain);
+
+	/* calculate AGAW */
+	domain->gaw = guest_width;
+	adjust_width = guestwidth_to_adjustwidth(guest_width);
+	domain->agaw = width_to_agaw(adjust_width);
+
+	INIT_LIST_HEAD(&domain->devices);
+
+	domain->iommu_count = 0;
+	domain->iommu_coherency = 0;
+	domain->max_addr = 0;
+
+	/* always allocate the top pgd */
+	domain->pgd = (struct dma_pte *)alloc_pgtable_page();
+	if (!domain->pgd)
+		return -ENOMEM;
+	domain_flush_cache(domain, domain->pgd, PAGE_SIZE);
 	return 0;
 }
 
-void intel_iommu_domain_exit(struct dmar_domain *domain)
+static void iommu_free_vm_domain(struct dmar_domain *domain)
+{
+	unsigned long flags;
+	struct dmar_drhd_unit *drhd;
+	struct intel_iommu *iommu;
+	unsigned long i;
+	unsigned long ndomains;
+
+	for_each_drhd_unit(drhd) {
+		if (drhd->ignored)
+			continue;
+		iommu = drhd->iommu;
+
+		ndomains = cap_ndoms(iommu->cap);
+		i = find_first_bit(iommu->domain_ids, ndomains);
+		for (; i < ndomains; ) {
+			if (iommu->domains[i] == domain) {
+				spin_lock_irqsave(&iommu->lock, flags);
+				clear_bit(i, iommu->domain_ids);
+				iommu->domains[i] = NULL;
+				spin_unlock_irqrestore(&iommu->lock, flags);
+				break;
+			}
+			i = find_next_bit(iommu->domain_ids, ndomains, i+1);
+		}
+	}
+}
+
+static void vm_domain_exit(struct dmar_domain *domain)
 {
 	u64 end;
 
@@ -2336,6 +2952,9 @@ void intel_iommu_domain_exit(struct dmar_domain *domain)
 	if (!domain)
 		return;
 
+	vm_domain_remove_all_dev_info(domain);
+	/* destroy iovas */
+	put_iova_domain(&domain->iovad);
 	end = DOMAIN_MAX_ADDR(domain->gaw);
 	end = end & (~VTD_PAGE_MASK);
 
@@ -2345,94 +2964,167 @@ void intel_iommu_domain_exit(struct dmar_domain *domain)
 	/* free page tables */
 	dma_pte_free_pagetable(domain, 0, end);
 
-	iommu_free_domain(domain);
+	iommu_free_vm_domain(domain);
 	free_domain_mem(domain);
 }
-EXPORT_SYMBOL_GPL(intel_iommu_domain_exit);
 
-struct dmar_domain *intel_iommu_domain_alloc(struct pci_dev *pdev)
+static int intel_iommu_domain_init(struct iommu_domain *domain)
 {
-	struct dmar_drhd_unit *drhd;
-	struct dmar_domain *domain;
-	struct intel_iommu *iommu;
-
-	drhd = dmar_find_matched_drhd_unit(pdev);
-	if (!drhd) {
-		printk(KERN_ERR "intel_iommu_domain_alloc: drhd == NULL\n");
-		return NULL;
-	}
+	struct dmar_domain *dmar_domain;
 
-	iommu = drhd->iommu;
-	if (!iommu) {
-		printk(KERN_ERR
-			"intel_iommu_domain_alloc: iommu == NULL\n");
-		return NULL;
-	}
-	domain = iommu_alloc_domain(iommu);
-	if (!domain) {
+	dmar_domain = iommu_alloc_vm_domain();
+	if (!dmar_domain) {
 		printk(KERN_ERR
-			"intel_iommu_domain_alloc: domain == NULL\n");
-		return NULL;
+			"intel_iommu_domain_init: dmar_domain == NULL\n");
+		return -ENOMEM;
 	}
-	if (domain_init(domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
+	if (vm_domain_init(dmar_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
 		printk(KERN_ERR
-			"intel_iommu_domain_alloc: domain_init() failed\n");
-		intel_iommu_domain_exit(domain);
-		return NULL;
+			"intel_iommu_domain_init() failed\n");
+		vm_domain_exit(dmar_domain);
+		return -ENOMEM;
 	}
-	return domain;
+	domain->priv = dmar_domain;
+
+	return 0;
 }
-EXPORT_SYMBOL_GPL(intel_iommu_domain_alloc);
 
-int intel_iommu_context_mapping(
-	struct dmar_domain *domain, struct pci_dev *pdev)
+static void intel_iommu_domain_destroy(struct iommu_domain *domain)
 {
-	int rc;
-	rc = domain_context_mapping(domain, pdev);
-	return rc;
+	struct dmar_domain *dmar_domain = domain->priv;
+
+	domain->priv = NULL;
+	vm_domain_exit(dmar_domain);
 }
-EXPORT_SYMBOL_GPL(intel_iommu_context_mapping);
 
-int intel_iommu_page_mapping(
-	struct dmar_domain *domain, dma_addr_t iova,
-	u64 hpa, size_t size, int prot)
+static int intel_iommu_attach_device(struct iommu_domain *domain,
+				     struct device *dev)
 {
-	int rc;
-	rc = domain_page_mapping(domain, iova, hpa, size, prot);
-	return rc;
+	struct dmar_domain *dmar_domain = domain->priv;
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct intel_iommu *iommu;
+	int addr_width;
+	u64 end;
+	int ret;
+
+	/* normally pdev is not mapped */
+	if (unlikely(domain_context_mapped(pdev))) {
+		struct dmar_domain *old_domain;
+
+		old_domain = find_domain(pdev);
+		if (old_domain) {
+			if (dmar_domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE)
+				vm_domain_remove_one_dev_info(old_domain, pdev);
+			else
+				domain_remove_dev_info(old_domain);
+		}
+	}
+
+	iommu = device_to_iommu(pdev->bus->number, pdev->devfn);
+	if (!iommu)
+		return -ENODEV;
+
+	/* check if this iommu agaw is sufficient for max mapped address */
+	addr_width = agaw_to_width(iommu->agaw);
+	end = DOMAIN_MAX_ADDR(addr_width);
+	end = end & VTD_PAGE_MASK;
+	if (end < dmar_domain->max_addr) {
+		printk(KERN_ERR "%s: iommu agaw (%d) is not "
+		       "sufficient for the mapped address (%llx)\n",
+		       __func__, iommu->agaw, dmar_domain->max_addr);
+		return -EFAULT;
+	}
+
+	ret = domain_context_mapping(dmar_domain, pdev);
+	if (ret)
+		return ret;
+
+	ret = vm_domain_add_dev_info(dmar_domain, pdev);
+	return ret;
 }
-EXPORT_SYMBOL_GPL(intel_iommu_page_mapping);
 
-void intel_iommu_detach_dev(struct dmar_domain *domain, u8 bus, u8 devfn)
+static void intel_iommu_detach_device(struct iommu_domain *domain,
+				      struct device *dev)
 {
-	detach_domain_for_dev(domain, bus, devfn);
+	struct dmar_domain *dmar_domain = domain->priv;
+	struct pci_dev *pdev = to_pci_dev(dev);
+
+	vm_domain_remove_one_dev_info(dmar_domain, pdev);
 }
-EXPORT_SYMBOL_GPL(intel_iommu_detach_dev);
 
-struct dmar_domain *
-intel_iommu_find_domain(struct pci_dev *pdev)
+static int intel_iommu_map_range(struct iommu_domain *domain,
+				 unsigned long iova, phys_addr_t hpa,
+				 size_t size, int iommu_prot)
 {
-	return find_domain(pdev);
+	struct dmar_domain *dmar_domain = domain->priv;
+	u64 max_addr;
+	int addr_width;
+	int prot = 0;
+	int ret;
+
+	if (iommu_prot & IOMMU_READ)
+		prot |= DMA_PTE_READ;
+	if (iommu_prot & IOMMU_WRITE)
+		prot |= DMA_PTE_WRITE;
+
+	max_addr = (iova & VTD_PAGE_MASK) + VTD_PAGE_ALIGN(size);
+	if (dmar_domain->max_addr < max_addr) {
+		int min_agaw;
+		u64 end;
+
+		/* check if minimum agaw is sufficient for mapped address */
+		min_agaw = vm_domain_min_agaw(dmar_domain);
+		addr_width = agaw_to_width(min_agaw);
+		end = DOMAIN_MAX_ADDR(addr_width);
+		end = end & VTD_PAGE_MASK;
+		if (end < max_addr) {
+			printk(KERN_ERR "%s: iommu agaw (%d) is not "
+			       "sufficient for the mapped address (%llx)\n",
+			       __func__, min_agaw, max_addr);
+			return -EFAULT;
+		}
+		dmar_domain->max_addr = max_addr;
+	}
+
+	ret = domain_page_mapping(dmar_domain, iova, hpa, size, prot);
+	return ret;
 }
-EXPORT_SYMBOL_GPL(intel_iommu_find_domain);
 
-int intel_iommu_found(void)
+static void intel_iommu_unmap_range(struct iommu_domain *domain,
+				    unsigned long iova, size_t size)
 {
-	return g_num_of_iommus;
+	struct dmar_domain *dmar_domain = domain->priv;
+	dma_addr_t base;
+
+	/* The address might not be aligned */
+	base = iova & VTD_PAGE_MASK;
+	size = VTD_PAGE_ALIGN(size);
+	dma_pte_clear_range(dmar_domain, base, base + size);
+
+	if (dmar_domain->max_addr == base + size)
+		dmar_domain->max_addr = base;
 }
-EXPORT_SYMBOL_GPL(intel_iommu_found);
 
-u64 intel_iommu_iova_to_pfn(struct dmar_domain *domain, u64 iova)
+static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
+					    unsigned long iova)
 {
+	struct dmar_domain *dmar_domain = domain->priv;
 	struct dma_pte *pte;
-	u64 pfn;
-
-	pfn = 0;
-	pte = addr_to_dma_pte(domain, iova);
+	u64 phys = 0;
 
+	pte = addr_to_dma_pte(dmar_domain, iova);
 	if (pte)
-		pfn = dma_pte_addr(*pte);
+		phys = dma_pte_addr(pte);
 
-	return pfn >> VTD_PAGE_SHIFT;
+	return phys;
 }
-EXPORT_SYMBOL_GPL(intel_iommu_iova_to_pfn);
+
+static struct iommu_ops intel_iommu_ops = {
+	.domain_init	= intel_iommu_domain_init,
+	.domain_destroy = intel_iommu_domain_destroy,
+	.attach_dev	= intel_iommu_attach_device,
+	.detach_dev	= intel_iommu_detach_device,
+	.map		= intel_iommu_map_range,
+	.unmap		= intel_iommu_unmap_range,
+	.iova_to_phys	= intel_iommu_iova_to_phys,
+};

include/linux/dma_remapping.h

@@ -9,148 +9,16 @@
 #define VTD_PAGE_MASK		(((u64)-1) << VTD_PAGE_SHIFT)
 #define VTD_PAGE_ALIGN(addr)	(((addr) + VTD_PAGE_SIZE - 1) & VTD_PAGE_MASK)
 
-#define IOVA_PFN(addr)		((addr) >> PAGE_SHIFT)
-#define DMA_32BIT_PFN		IOVA_PFN(DMA_32BIT_MASK)
-#define DMA_64BIT_PFN		IOVA_PFN(DMA_64BIT_MASK)
-
-
-/*
- * 0: Present
- * 1-11: Reserved
- * 12-63: Context Ptr (12 - (haw-1))
- * 64-127: Reserved
- */
-struct root_entry {
-	u64	val;
-	u64	rsvd1;
-};
-#define ROOT_ENTRY_NR (VTD_PAGE_SIZE/sizeof(struct root_entry))
-static inline bool root_present(struct root_entry *root)
-{
-	return (root->val & 1);
-}
-static inline void set_root_present(struct root_entry *root)
-{
-	root->val |= 1;
-}
-static inline void set_root_value(struct root_entry *root, unsigned long value)
-{
-	root->val |= value & VTD_PAGE_MASK;
-}
-
-struct context_entry;
-static inline struct context_entry *
-get_context_addr_from_root(struct root_entry *root)
-{
-	return (struct context_entry *)
-		(root_present(root)?phys_to_virt(
-		root->val & VTD_PAGE_MASK) :
-		NULL);
-}
-
-/*
- * low 64 bits:
- * 0: present
- * 1: fault processing disable
- * 2-3: translation type
- * 12-63: address space root
- * high 64 bits:
- * 0-2: address width
- * 3-6: aval
- * 8-23: domain id
- */
-struct context_entry {
-	u64 lo;
-	u64 hi;
-};
-#define context_present(c) ((c).lo & 1)
-#define context_fault_disable(c) (((c).lo >> 1) & 1)
-#define context_translation_type(c) (((c).lo >> 2) & 3)
-#define context_address_root(c) ((c).lo & VTD_PAGE_MASK)
-#define context_address_width(c) ((c).hi &  7)
-#define context_domain_id(c) (((c).hi >> 8) & ((1 << 16) - 1))
-
-#define context_set_present(c) do {(c).lo |= 1;} while (0)
-#define context_set_fault_enable(c) \
-	do {(c).lo &= (((u64)-1) << 2) | 1;} while (0)
-#define context_set_translation_type(c, val) \
-	do { \
-		(c).lo &= (((u64)-1) << 4) | 3; \
-		(c).lo |= ((val) & 3) << 2; \
-	} while (0)
-#define CONTEXT_TT_MULTI_LEVEL 0
-#define context_set_address_root(c, val) \
-	do {(c).lo |= (val) & VTD_PAGE_MASK; } while (0)
-#define context_set_address_width(c, val) do {(c).hi |= (val) & 7;} while (0)
-#define context_set_domain_id(c, val) \
-	do {(c).hi |= ((val) & ((1 << 16) - 1)) << 8;} while (0)
-#define context_clear_entry(c) do {(c).lo = 0; (c).hi = 0;} while (0)
-
-/*
- * 0: readable
- * 1: writable
- * 2-6: reserved
- * 7: super page
- * 8-11: available
- * 12-63: Host physcial address
- */
-struct dma_pte {
-	u64 val;
-};
-#define dma_clear_pte(p)	do {(p).val = 0;} while (0)
-
 #define DMA_PTE_READ (1)
 #define DMA_PTE_WRITE (2)
 
-#define dma_set_pte_readable(p) do {(p).val |= DMA_PTE_READ;} while (0)
-#define dma_set_pte_writable(p) do {(p).val |= DMA_PTE_WRITE;} while (0)
-#define dma_set_pte_prot(p, prot) \
-		do {(p).val = ((p).val & ~3) | ((prot) & 3); } while (0)
-#define dma_pte_addr(p) ((p).val & VTD_PAGE_MASK)
-#define dma_set_pte_addr(p, addr) do {\
-		(p).val |= ((addr) & VTD_PAGE_MASK); } while (0)
-#define dma_pte_present(p) (((p).val & 3) != 0)
-
 struct intel_iommu;
+struct dmar_domain;
+struct root_entry;
 
-struct dmar_domain {
-	int	id;			/* domain id */
-	struct intel_iommu *iommu;	/* back pointer to owning iommu */
-
-	struct list_head devices; 	/* all devices' list */
-	struct iova_domain iovad;	/* iova's that belong to this domain */
-
-	struct dma_pte	*pgd;		/* virtual address */
-	spinlock_t	mapping_lock;	/* page table lock */
-	int		gaw;		/* max guest address width */
-
-	/* adjusted guest address width, 0 is level 2 30-bit */
-	int		agaw;
-
-#define DOMAIN_FLAG_MULTIPLE_DEVICES 1
-	int		flags;
-};
-
-/* PCI domain-device relationship */
-struct device_domain_info {
-	struct list_head link;	/* link to domain siblings */
-	struct list_head global; /* link to global list */
-	u8 bus;			/* PCI bus numer */
-	u8 devfn;		/* PCI devfn number */
-	struct pci_dev *dev; /* it's NULL for PCIE-to-PCI bridge */
-	struct dmar_domain *domain; /* pointer to domain */
-};
-
-extern int init_dmars(void);
 extern void free_dmar_iommu(struct intel_iommu *iommu);
+extern int iommu_calculate_agaw(struct intel_iommu *iommu);
 
 extern int dmar_disabled;
 
-#ifndef CONFIG_DMAR_GFX_WA
-static inline void iommu_prepare_gfx_mapping(void)
-{
-	return;
-}
-#endif /* !CONFIG_DMAR_GFX_WA */
-
 #endif

include/linux/dmar.h

@@ -144,7 +144,6 @@ struct dmar_rmrr_unit {
 	list_for_each_entry(rmrr, &dmar_rmrr_units, list)
 /* Intel DMAR  initialization functions */
 extern int intel_iommu_init(void);
-extern int dmar_disabled;
 #else
 static inline int intel_iommu_init(void)
 {

include/linux/intel-iommu.h

@@ -23,8 +23,6 @@
 #define _INTEL_IOMMU_H_
 
 #include <linux/types.h>
-#include <linux/msi.h>
-#include <linux/sysdev.h>
 #include <linux/iova.h>
 #include <linux/io.h>
 #include <linux/dma_remapping.h>
@@ -289,10 +287,10 @@ struct intel_iommu {
 	void __iomem	*reg; /* Pointer to hardware regs, virtual addr */
 	u64		cap;
 	u64		ecap;
-	int		seg;
 	u32		gcmd; /* Holds TE, EAFL. Don't need SRTP, SFL, WBF */
 	spinlock_t	register_lock; /* protect register handling */
 	int		seq_id;	/* sequence id of the iommu */
+	int		agaw; /* agaw of this iommu */
 
 #ifdef CONFIG_DMAR
 	unsigned long 	*domain_ids; /* bitmap of domains */
@@ -302,8 +300,6 @@ struct intel_iommu {
 
 	unsigned int irq;
 	unsigned char name[7];    /* Device Name */
-	struct msi_msg saved_msg;
-	struct sys_device sysdev;
 	struct iommu_flush flush;
 #endif
 	struct q_inval  *qi;            /* Queued invalidation info */
@@ -334,25 +330,6 @@ extern int qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr,
 
 extern void qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu);
 
-void intel_iommu_domain_exit(struct dmar_domain *domain);
-struct dmar_domain *intel_iommu_domain_alloc(struct pci_dev *pdev);
-int intel_iommu_context_mapping(struct dmar_domain *domain,
-				struct pci_dev *pdev);
-int intel_iommu_page_mapping(struct dmar_domain *domain, dma_addr_t iova,
-			     u64 hpa, size_t size, int prot);
-void intel_iommu_detach_dev(struct dmar_domain *domain, u8 bus, u8 devfn);
-struct dmar_domain *intel_iommu_find_domain(struct pci_dev *pdev);
-u64 intel_iommu_iova_to_pfn(struct dmar_domain *domain, u64 iova);
-
-#ifdef CONFIG_DMAR
-int intel_iommu_found(void);
-#else /* CONFIG_DMAR */
-static inline int intel_iommu_found(void)
-{
-	return 0;
-}
-#endif /* CONFIG_DMAR */
-
 extern void *intel_alloc_coherent(struct device *, size_t, dma_addr_t *, gfp_t);
 extern void intel_free_coherent(struct device *, size_t, void *, dma_addr_t);
 extern dma_addr_t intel_map_single(struct device *, phys_addr_t, size_t, int);

include/linux/iommu.h

@@ -0,0 +1,112 @@
+/*
+ * Copyright (C) 2007-2008 Advanced Micro Devices, Inc.
+ * Author: Joerg Roedel <joerg.roedel@amd.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ */
+
+#ifndef __LINUX_IOMMU_H
+#define __LINUX_IOMMU_H
+
+#define IOMMU_READ	(1)
+#define IOMMU_WRITE	(2)
+
+struct device;
+
+struct iommu_domain {
+	void *priv;
+};
+
+struct iommu_ops {
+	int (*domain_init)(struct iommu_domain *domain);
+	void (*domain_destroy)(struct iommu_domain *domain);
+	int (*attach_dev)(struct iommu_domain *domain, struct device *dev);
+	void (*detach_dev)(struct iommu_domain *domain, struct device *dev);
+	int (*map)(struct iommu_domain *domain, unsigned long iova,
+		   phys_addr_t paddr, size_t size, int prot);
+	void (*unmap)(struct iommu_domain *domain, unsigned long iova,
+		      size_t size);
+	phys_addr_t (*iova_to_phys)(struct iommu_domain *domain,
+				    unsigned long iova);
+};
+
+#ifdef CONFIG_IOMMU_API
+
+extern void register_iommu(struct iommu_ops *ops);
+extern bool iommu_found(void);
+extern struct iommu_domain *iommu_domain_alloc(void);
+extern void iommu_domain_free(struct iommu_domain *domain);
+extern int iommu_attach_device(struct iommu_domain *domain,
+			       struct device *dev);
+extern void iommu_detach_device(struct iommu_domain *domain,
+				struct device *dev);
+extern int iommu_map_range(struct iommu_domain *domain, unsigned long iova,
+			   phys_addr_t paddr, size_t size, int prot);
+extern void iommu_unmap_range(struct iommu_domain *domain, unsigned long iova,
+			      size_t size);
+extern phys_addr_t iommu_iova_to_phys(struct iommu_domain *domain,
+				      unsigned long iova);
+
+#else /* CONFIG_IOMMU_API */
+
+static inline void register_iommu(struct iommu_ops *ops)
+{
+}
+
+static inline bool iommu_found(void)
+{
+	return false;
+}
+
+static inline struct iommu_domain *iommu_domain_alloc(void)
+{
+	return NULL;
+}
+
+static inline void iommu_domain_free(struct iommu_domain *domain)
+{
+}
+
+static inline int iommu_attach_device(struct iommu_domain *domain,
+				      struct device *dev)
+{
+	return -ENODEV;
+}
+
+static inline void iommu_detach_device(struct iommu_domain *domain,
+				       struct device *dev)
+{
+}
+
+static inline int iommu_map_range(struct iommu_domain *domain,
+				  unsigned long iova, phys_addr_t paddr,
+				  size_t size, int prot)
+{
+	return -ENODEV;
+}
+
+static inline void iommu_unmap_range(struct iommu_domain *domain,
+				     unsigned long iova, size_t size)
+{
+}
+
+static inline phys_addr_t iommu_iova_to_phys(struct iommu_domain *domain,
+					     unsigned long iova)
+{
+	return 0;
+}
+
+#endif /* CONFIG_IOMMU_API */
+
+#endif /* __LINUX_IOMMU_H */

include/linux/kvm_host.h

@@ -316,6 +316,7 @@ struct kvm_assigned_dev_kernel {
 #define KVM_ASSIGNED_DEV_HOST_MSI	(1 << 9)
 	unsigned long irq_requested_type;
 	int irq_source_id;
+	int flags;
 	struct pci_dev *dev;
 	struct kvm *kvm;
 };
@@ -327,13 +328,16 @@ void kvm_unregister_irq_ack_notifier(struct kvm_irq_ack_notifier *kian);
 int kvm_request_irq_source_id(struct kvm *kvm);
 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 
-#ifdef CONFIG_DMAR
+#ifdef CONFIG_IOMMU_API
 int kvm_iommu_map_pages(struct kvm *kvm, gfn_t base_gfn,
 			unsigned long npages);
-int kvm_iommu_map_guest(struct kvm *kvm,
-			struct kvm_assigned_dev_kernel *assigned_dev);
+int kvm_iommu_map_guest(struct kvm *kvm);
 int kvm_iommu_unmap_guest(struct kvm *kvm);
-#else /* CONFIG_DMAR */
+int kvm_assign_device(struct kvm *kvm,
+		      struct kvm_assigned_dev_kernel *assigned_dev);
+int kvm_deassign_device(struct kvm *kvm,
+			struct kvm_assigned_dev_kernel *assigned_dev);
+#else /* CONFIG_IOMMU_API */
 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 				      gfn_t base_gfn,
 				      unsigned long npages)
@@ -341,9 +345,7 @@ static inline int kvm_iommu_map_pages(struct kvm *kvm,
 	return 0;
 }
 
-static inline int kvm_iommu_map_guest(struct kvm *kvm,
-				      struct kvm_assigned_dev_kernel
-				      *assigned_dev)
+static inline int kvm_iommu_map_guest(struct kvm *kvm)
 {
 	return -ENODEV;
 }
@@ -352,7 +354,19 @@ static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 {
 	return 0;
 }
-#endif /* CONFIG_DMAR */
+
+static inline int kvm_assign_device(struct kvm *kvm,
+		struct kvm_assigned_dev_kernel *assigned_dev)
+{
+	return 0;
+}
+
+static inline int kvm_deassign_device(struct kvm *kvm,
+		struct kvm_assigned_dev_kernel *assigned_dev)
+{
+	return 0;
+}
+#endif /* CONFIG_IOMMU_API */
 
 static inline void kvm_guest_enter(void)
 {

virt/kvm/vtd.c → virt/kvm/iommu.c

@@ -25,6 +25,7 @@
 #include <linux/kvm_host.h>
 #include <linux/pci.h>
 #include <linux/dmar.h>
+#include <linux/iommu.h>
 #include <linux/intel-iommu.h>
 
 static int kvm_iommu_unmap_memslots(struct kvm *kvm);
@@ -37,7 +38,7 @@ int kvm_iommu_map_pages(struct kvm *kvm,
 	gfn_t gfn = base_gfn;
 	pfn_t pfn;
 	int i, r = 0;
-	struct dmar_domain *domain = kvm->arch.intel_iommu_domain;
+	struct iommu_domain *domain = kvm->arch.iommu_domain;
 
 	/* check if iommu exists and in use */
 	if (!domain)
@@ -45,20 +46,17 @@ int kvm_iommu_map_pages(struct kvm *kvm,
 
 	for (i = 0; i < npages; i++) {
 		/* check if already mapped */
-		pfn = (pfn_t)intel_iommu_iova_to_pfn(domain,
-						     gfn_to_gpa(gfn));
-		if (pfn)
+		if (iommu_iova_to_phys(domain, gfn_to_gpa(gfn)))
 			continue;
 
 		pfn = gfn_to_pfn(kvm, gfn);
-		r = intel_iommu_page_mapping(domain,
-					     gfn_to_gpa(gfn),
-					     pfn_to_hpa(pfn),
-					     PAGE_SIZE,
-					     DMA_PTE_READ |
-					     DMA_PTE_WRITE);
+		r = iommu_map_range(domain,
+				    gfn_to_gpa(gfn),
+				    pfn_to_hpa(pfn),
+				    PAGE_SIZE,
+				    IOMMU_READ | IOMMU_WRITE);
 		if (r) {
-			printk(KERN_ERR "kvm_iommu_map_pages:"
+			printk(KERN_ERR "kvm_iommu_map_address:"
 			       "iommu failed to map pfn=%lx\n", pfn);
 			goto unmap_pages;
 		}
@@ -73,7 +71,7 @@ unmap_pages:
 
 static int kvm_iommu_map_memslots(struct kvm *kvm)
 {
-	int i, r;
+	int i, r = 0;
 
 	down_read(&kvm->slots_lock);
 	for (i = 0; i < kvm->nmemslots; i++) {
@@ -86,50 +84,79 @@ static int kvm_iommu_map_memslots(struct kvm *kvm)
 	return r;
 }
 
-int kvm_iommu_map_guest(struct kvm *kvm,
-			struct kvm_assigned_dev_kernel *assigned_dev)
+int kvm_assign_device(struct kvm *kvm,
+		      struct kvm_assigned_dev_kernel *assigned_dev)
 {
 	struct pci_dev *pdev = NULL;
+	struct iommu_domain *domain = kvm->arch.iommu_domain;
 	int r;
 
-	if (!intel_iommu_found()) {
-		printk(KERN_ERR "%s: intel iommu not found\n", __func__);
+	/* check if iommu exists and in use */
+	if (!domain)
+		return 0;
+
+	pdev = assigned_dev->dev;
+	if (pdev == NULL)
 		return -ENODEV;
+
+	r = iommu_attach_device(domain, &pdev->dev);
+	if (r) {
+		printk(KERN_ERR "assign device %x:%x.%x failed",
+			pdev->bus->number,
+			PCI_SLOT(pdev->devfn),
+			PCI_FUNC(pdev->devfn));
+		return r;
 	}
 
-	printk(KERN_DEBUG "VT-d direct map: host bdf = %x:%x:%x\n",
-	       assigned_dev->host_busnr,
-	       PCI_SLOT(assigned_dev->host_devfn),
-	       PCI_FUNC(assigned_dev->host_devfn));
+	printk(KERN_DEBUG "assign device: host bdf = %x:%x:%x\n",
+		assigned_dev->host_busnr,
+		PCI_SLOT(assigned_dev->host_devfn),
+		PCI_FUNC(assigned_dev->host_devfn));
+
+	return 0;
+}
+
+int kvm_deassign_device(struct kvm *kvm,
+			struct kvm_assigned_dev_kernel *assigned_dev)
+{
+	struct iommu_domain *domain = kvm->arch.iommu_domain;
+	struct pci_dev *pdev = NULL;
+
+	/* check if iommu exists and in use */
+	if (!domain)
+		return 0;
 
 	pdev = assigned_dev->dev;
+	if (pdev == NULL)
+		return -ENODEV;
 
-	if (pdev == NULL) {
-		if (kvm->arch.intel_iommu_domain) {
-			intel_iommu_domain_exit(kvm->arch.intel_iommu_domain);
-			kvm->arch.intel_iommu_domain = NULL;
-		}
+	iommu_detach_device(domain, &pdev->dev);
+
+	printk(KERN_DEBUG "deassign device: host bdf = %x:%x:%x\n",
+		assigned_dev->host_busnr,
+		PCI_SLOT(assigned_dev->host_devfn),
+		PCI_FUNC(assigned_dev->host_devfn));
+
+	return 0;
+}
+
+int kvm_iommu_map_guest(struct kvm *kvm)
+{
+	int r;
+
+	if (!iommu_found()) {
+		printk(KERN_ERR "%s: iommu not found\n", __func__);
 		return -ENODEV;
 	}
 
-	kvm->arch.intel_iommu_domain = intel_iommu_domain_alloc(pdev);
-	if (!kvm->arch.intel_iommu_domain)
-		return -ENODEV;
+	kvm->arch.iommu_domain = iommu_domain_alloc();
+	if (!kvm->arch.iommu_domain)
+		return -ENOMEM;
 
 	r = kvm_iommu_map_memslots(kvm);
 	if (r)
 		goto out_unmap;
 
-	intel_iommu_detach_dev(kvm->arch.intel_iommu_domain,
-			       pdev->bus->number, pdev->devfn);
-
-	r = intel_iommu_context_mapping(kvm->arch.intel_iommu_domain,
-					pdev);
-	if (r) {
-		printk(KERN_ERR "Domain context map for %s failed",
-		       pci_name(pdev));
-		goto out_unmap;
-	}
 	return 0;
 
 out_unmap:
@@ -138,19 +165,26 @@ out_unmap:
 }
 
 static void kvm_iommu_put_pages(struct kvm *kvm,
-			       gfn_t base_gfn, unsigned long npages)
+				gfn_t base_gfn, unsigned long npages)
 {
 	gfn_t gfn = base_gfn;
 	pfn_t pfn;
-	struct dmar_domain *domain = kvm->arch.intel_iommu_domain;
-	int i;
+	struct iommu_domain *domain = kvm->arch.iommu_domain;
+	unsigned long i;
+	u64 phys;
+
+	/* check if iommu exists and in use */
+	if (!domain)
+		return;
 
 	for (i = 0; i < npages; i++) {
-		pfn = (pfn_t)intel_iommu_iova_to_pfn(domain,
-						     gfn_to_gpa(gfn));
+		phys = iommu_iova_to_phys(domain, gfn_to_gpa(gfn));
+		pfn = phys >> PAGE_SHIFT;
 		kvm_release_pfn_clean(pfn);
 		gfn++;
 	}
+
+	iommu_unmap_range(domain, gfn_to_gpa(base_gfn), PAGE_SIZE * npages);
 }
 
 static int kvm_iommu_unmap_memslots(struct kvm *kvm)
@@ -168,24 +202,13 @@ static int kvm_iommu_unmap_memslots(struct kvm *kvm)
 
 int kvm_iommu_unmap_guest(struct kvm *kvm)
 {
-	struct kvm_assigned_dev_kernel *entry;
-	struct dmar_domain *domain = kvm->arch.intel_iommu_domain;
+	struct iommu_domain *domain = kvm->arch.iommu_domain;
 
 	/* check if iommu exists and in use */
 	if (!domain)
 		return 0;
 
-	list_for_each_entry(entry, &kvm->arch.assigned_dev_head, list) {
-		printk(KERN_DEBUG "VT-d unmap: host bdf = %x:%x:%x\n",
-		       entry->host_busnr,
-		       PCI_SLOT(entry->host_devfn),
-		       PCI_FUNC(entry->host_devfn));
-
-		/* detach kvm dmar domain */
-		intel_iommu_detach_dev(domain, entry->host_busnr,
-				       entry->host_devfn);
-	}
 	kvm_iommu_unmap_memslots(kvm);
-	intel_iommu_domain_exit(domain);
+	iommu_domain_free(domain);
 	return 0;
 }

virt/kvm/kvm_main.c

@@ -496,6 +496,7 @@ static int kvm_vm_ioctl_assign_device(struct kvm *kvm,
 	match->assigned_dev_id = assigned_dev->assigned_dev_id;
 	match->host_busnr = assigned_dev->busnr;
 	match->host_devfn = assigned_dev->devfn;
+	match->flags = assigned_dev->flags;
 	match->dev = dev;
 	match->irq_source_id = -1;
 	match->kvm = kvm;
@@ -503,7 +504,12 @@ static int kvm_vm_ioctl_assign_device(struct kvm *kvm,
 	list_add(&match->list, &kvm->arch.assigned_dev_head);
 
 	if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) {
-		r = kvm_iommu_map_guest(kvm, match);
+		if (!kvm->arch.iommu_domain) {
+			r = kvm_iommu_map_guest(kvm);
+			if (r)
+				goto out_list_del;
+		}
+		r = kvm_assign_device(kvm, match);
 		if (r)
 			goto out_list_del;
 	}
@@ -525,6 +531,35 @@ out_free:
 }
 #endif
 
+#ifdef KVM_CAP_DEVICE_DEASSIGNMENT
+static int kvm_vm_ioctl_deassign_device(struct kvm *kvm,
+		struct kvm_assigned_pci_dev *assigned_dev)
+{
+	int r = 0;
+	struct kvm_assigned_dev_kernel *match;
+
+	mutex_lock(&kvm->lock);
+
+	match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
+				      assigned_dev->assigned_dev_id);
+	if (!match) {
+		printk(KERN_INFO "%s: device hasn't been assigned before, "
+		  "so cannot be deassigned\n", __func__);
+		r = -EINVAL;
+		goto out;
+	}
+
+	if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU)
+		kvm_deassign_device(kvm, match);
+
+	kvm_free_assigned_device(kvm, match);
+
+out:
+	mutex_unlock(&kvm->lock);
+	return r;
+}
+#endif
+
 static inline int valid_vcpu(int n)
 {
 	return likely(n >= 0 && n < KVM_MAX_VCPUS);
@@ -1857,6 +1892,19 @@ static long kvm_vm_ioctl(struct file *filp,
 			goto out;
 		break;
 	}
+#endif
+#ifdef KVM_CAP_DEVICE_DEASSIGNMENT
+	case KVM_DEASSIGN_PCI_DEVICE: {
+		struct kvm_assigned_pci_dev assigned_dev;
+
+		r = -EFAULT;
+		if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev))
+			goto out;
+		r = kvm_vm_ioctl_deassign_device(kvm, &assigned_dev);
+		if (r)
+			goto out;
+		break;
+	}
 #endif
 	default:
 		r = kvm_arch_vm_ioctl(filp, ioctl, arg);