Merge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu: (53 commits)
  iommu/amd: Set IOTLB invalidation timeout
  iommu/amd: Init stats for iommu=pt
  iommu/amd: Remove unnecessary cache flushes in amd_iommu_resume
  iommu/amd: Add invalidate-context call-back
  iommu/amd: Add amd_iommu_device_info() function
  iommu/amd: Adapt IOMMU driver to PCI register name changes
  iommu/amd: Add invalid_ppr callback
  iommu/amd: Implement notifiers for IOMMUv2
  iommu/amd: Implement IO page-fault handler
  iommu/amd: Add routines to bind/unbind a pasid
  iommu/amd: Implement device aquisition code for IOMMUv2
  iommu/amd: Add driver stub for AMD IOMMUv2 support
  iommu/amd: Add stat counter for IOMMUv2 events
  iommu/amd: Add device errata handling
  iommu/amd: Add function to get IOMMUv2 domain for pdev
  iommu/amd: Implement function to send PPR completions
  iommu/amd: Implement functions to manage GCR3 table
  iommu/amd: Implement IOMMUv2 TLB flushing routines
  iommu/amd: Add support for IOMMUv2 domain mode
  iommu/amd: Add amd_iommu_domain_direct_map function
  ...

Documentation/ABI/testing/sysfs-bus-pci

@@ -66,6 +66,24 @@ Description:
 		re-discover previously removed devices.
 		Depends on CONFIG_HOTPLUG.
 
+What:		/sys/bus/pci/devices/.../msi_irqs/
+Date:		September, 2011
+Contact:	Neil Horman <nhorman@tuxdriver.com>
+Description:
+		The /sys/devices/.../msi_irqs directory contains a variable set
+		of sub-directories, with each sub-directory being named after a
+		corresponding msi irq vector allocated to that device.  Each
+		numbered sub-directory N contains attributes of that irq.
+		Note that this directory is not created for device drivers which
+		do not support msi irqs
+
+What:		/sys/bus/pci/devices/.../msi_irqs/<N>/mode
+Date:		September 2011
+Contact:	Neil Horman <nhorman@tuxdriver.com>
+Description:
+		This attribute indicates the mode that the irq vector named by
+		the parent directory is in (msi vs. msix)
+
 What:		/sys/bus/pci/devices/.../remove
 Date:		January 2009
 Contact:	Linux PCI developers <linux-pci@vger.kernel.org>

Documentation/kernel-parameters.txt

@@ -329,6 +329,11 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
 				    is a lot of faster
 			off	  - do not initialize any AMD IOMMU found in
 				    the system
+			force_isolation - Force device isolation for all
+					  devices. The IOMMU driver is not
+					  allowed anymore to lift isolation
+					  requirements as needed. This option
+					  does not override iommu=pt
 
 	amijoy.map=	[HW,JOY] Amiga joystick support
 			Map of devices attached to JOY0DAT and JOY1DAT
@@ -1059,7 +1064,9 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
 		nomerge
 		forcesac
 		soft
-		pt	[x86, IA-64]
+		pt		[x86, IA-64]
+		group_mf	[x86, IA-64]
+
 
 	io7=		[HW] IO7 for Marvel based alpha systems
 			See comment before marvel_specify_io7 in

arch/arm/mach-omap2/devices.c

@@ -28,6 +28,7 @@
 #include <plat/board.h>
 #include <plat/mcbsp.h>
 #include <plat/mmc.h>
+#include <plat/iommu.h>
 #include <plat/dma.h>
 #include <plat/omap_hwmod.h>
 #include <plat/omap_device.h>
@@ -211,9 +212,15 @@ static struct platform_device omap3isp_device = {
 	.resource	= omap3isp_resources,
 };
 
+static struct omap_iommu_arch_data omap3_isp_iommu = {
+	.name = "isp",
+};
+
 int omap3_init_camera(struct isp_platform_data *pdata)
 {
 	omap3isp_device.dev.platform_data = pdata;
+	omap3isp_device.dev.archdata.iommu = &omap3_isp_iommu;
+
 	return platform_device_register(&omap3isp_device);
 }
 

arch/arm/plat-omap/include/plat/iommu.h

@@ -111,6 +111,32 @@ struct iommu_platform_data {
 	u32 da_end;
 };
 
+/**
+ * struct iommu_arch_data - omap iommu private data
+ * @name: name of the iommu device
+ * @iommu_dev: handle of the iommu device
+ *
+ * This is an omap iommu private data object, which binds an iommu user
+ * to its iommu device. This object should be placed at the iommu user's
+ * dev_archdata so generic IOMMU API can be used without having to
+ * utilize omap-specific plumbing anymore.
+ */
+struct omap_iommu_arch_data {
+	const char *name;
+	struct omap_iommu *iommu_dev;
+};
+
+/**
+ * dev_to_omap_iommu() - retrieves an omap iommu object from a user device
+ * @dev: iommu client device
+ */
+static inline struct omap_iommu *dev_to_omap_iommu(struct device *dev)
+{
+	struct omap_iommu_arch_data *arch_data = dev->archdata.iommu;
+
+	return arch_data->iommu_dev;
+}
+
 /* IOMMU errors */
 #define OMAP_IOMMU_ERR_TLB_MISS		(1 << 0)
 #define OMAP_IOMMU_ERR_TRANS_FAULT	(1 << 1)
@@ -163,8 +189,8 @@ extern int omap_iommu_set_isr(const char *name,
 				    void *priv),
 			 void *isr_priv);
 
-extern void omap_iommu_save_ctx(struct omap_iommu *obj);
-extern void omap_iommu_restore_ctx(struct omap_iommu *obj);
+extern void omap_iommu_save_ctx(struct device *dev);
+extern void omap_iommu_restore_ctx(struct device *dev);
 
 extern int omap_install_iommu_arch(const struct iommu_functions *ops);
 extern void omap_uninstall_iommu_arch(const struct iommu_functions *ops);
@@ -176,6 +202,5 @@ extern ssize_t
 omap_iommu_dump_ctx(struct omap_iommu *obj, char *buf, ssize_t len);
 extern size_t
 omap_dump_tlb_entries(struct omap_iommu *obj, char *buf, ssize_t len);
-struct device *omap_find_iommu_device(const char *name);
 
 #endif /* __MACH_IOMMU_H */

arch/arm/plat-omap/include/plat/iovmm.h

@@ -72,18 +72,18 @@ struct iovm_struct {
 #define IOVMF_DA_FIXED		(1 << (4 + IOVMF_SW_SHIFT))
 
 
-extern struct iovm_struct *omap_find_iovm_area(struct omap_iommu *obj, u32 da);
+extern struct iovm_struct *omap_find_iovm_area(struct device *dev, u32 da);
 extern u32
-omap_iommu_vmap(struct iommu_domain *domain, struct omap_iommu *obj, u32 da,
+omap_iommu_vmap(struct iommu_domain *domain, struct device *dev, u32 da,
 			const struct sg_table *sgt, u32 flags);
 extern struct sg_table *omap_iommu_vunmap(struct iommu_domain *domain,
-				struct omap_iommu *obj, u32 da);
+				struct device *dev, u32 da);
 extern u32
-omap_iommu_vmalloc(struct iommu_domain *domain, struct omap_iommu *obj,
+omap_iommu_vmalloc(struct iommu_domain *domain, struct device *dev,
 				u32 da, size_t bytes, u32 flags);
 extern void
-omap_iommu_vfree(struct iommu_domain *domain, struct omap_iommu *obj,
+omap_iommu_vfree(struct iommu_domain *domain, struct device *dev,
 				const u32 da);
-extern void *omap_da_to_va(struct omap_iommu *obj, u32 da);
+extern void *omap_da_to_va(struct device *dev, u32 da);
 
 #endif /* __IOMMU_MMAP_H */

arch/ia64/include/asm/iommu.h

@@ -11,10 +11,12 @@ extern void no_iommu_init(void);
 extern int force_iommu, no_iommu;
 extern int iommu_pass_through;
 extern int iommu_detected;
+extern int iommu_group_mf;
 #else
 #define iommu_pass_through	(0)
 #define no_iommu		(1)
 #define iommu_detected		(0)
+#define iommu_group_mf		(0)
 #endif
 extern void iommu_dma_init(void);
 extern void machvec_init(const char *name);

arch/ia64/kernel/pci-dma.c

@@ -33,6 +33,7 @@ int force_iommu __read_mostly;
 #endif
 
 int iommu_pass_through;
+int iommu_group_mf;
 
 /* Dummy device used for NULL arguments (normally ISA). Better would
    be probably a smaller DMA mask, but this is bug-to-bug compatible

arch/x86/include/asm/iommu.h

@@ -5,6 +5,7 @@ extern struct dma_map_ops nommu_dma_ops;
 extern int force_iommu, no_iommu;
 extern int iommu_detected;
 extern int iommu_pass_through;
+extern int iommu_group_mf;
 
 /* 10 seconds */
 #define DMAR_OPERATION_TIMEOUT ((cycles_t) tsc_khz*10*1000)

arch/x86/kernel/pci-dma.c

@@ -45,6 +45,15 @@ int iommu_detected __read_mostly = 0;
  */
 int iommu_pass_through __read_mostly;
 
+/*
+ * Group multi-function PCI devices into a single device-group for the
+ * iommu_device_group interface.  This tells the iommu driver to pretend
+ * it cannot distinguish between functions of a device, exposing only one
+ * group for the device.  Useful for disallowing use of individual PCI
+ * functions from userspace drivers.
+ */
+int iommu_group_mf __read_mostly;
+
 extern struct iommu_table_entry __iommu_table[], __iommu_table_end[];
 
 /* Dummy device used for NULL arguments (normally ISA). */
@@ -169,6 +178,8 @@ static __init int iommu_setup(char *p)
 #endif
 		if (!strncmp(p, "pt", 2))
 			iommu_pass_through = 1;
+		if (!strncmp(p, "group_mf", 8))
+			iommu_group_mf = 1;
 
 		gart_parse_options(p);
 

drivers/acpi/pci_root.c

@@ -596,6 +596,13 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device)
 		if (ACPI_SUCCESS(status)) {
 			dev_info(root->bus->bridge,
 				"ACPI _OSC control (0x%02x) granted\n", flags);
+			if (acpi_gbl_FADT.boot_flags & ACPI_FADT_NO_ASPM) {
+				/*
+				 * We have ASPM control, but the FADT indicates
+				 * that it's unsupported. Clear it.
+				 */
+				pcie_clear_aspm(root->bus);
+			}
 		} else {
 			dev_info(root->bus->bridge,
 				"ACPI _OSC request failed (%s), "

drivers/iommu/Kconfig

@@ -34,7 +34,9 @@ config AMD_IOMMU
 	bool "AMD IOMMU support"
 	select SWIOTLB
 	select PCI_MSI
-	select PCI_IOV
+	select PCI_ATS
+	select PCI_PRI
+	select PCI_PASID
 	select IOMMU_API
 	depends on X86_64 && PCI && ACPI
 	---help---
@@ -58,6 +60,15 @@ config AMD_IOMMU_STATS
 	  information to userspace via debugfs.
 	  If unsure, say N.
 
+config AMD_IOMMU_V2
+	tristate "AMD IOMMU Version 2 driver (EXPERIMENTAL)"
+	depends on AMD_IOMMU && PROFILING && EXPERIMENTAL
+	select MMU_NOTIFIER
+	---help---
+	  This option enables support for the AMD IOMMUv2 features of the IOMMU
+	  hardware. Select this option if you want to use devices that support
+	  the the PCI PRI and PASID interface.
+
 # Intel IOMMU support
 config DMAR_TABLE
 	bool

drivers/iommu/Makefile

@@ -1,6 +1,7 @@
 obj-$(CONFIG_IOMMU_API) += iommu.o
 obj-$(CONFIG_MSM_IOMMU) += msm_iommu.o msm_iommu_dev.o
 obj-$(CONFIG_AMD_IOMMU) += amd_iommu.o amd_iommu_init.o
+obj-$(CONFIG_AMD_IOMMU_V2) += amd_iommu_v2.o
 obj-$(CONFIG_DMAR_TABLE) += dmar.o
 obj-$(CONFIG_INTEL_IOMMU) += iova.o intel-iommu.o
 obj-$(CONFIG_IRQ_REMAP) += intr_remapping.o

drivers/iommu/amd_iommu.c

@@ -17,6 +17,7 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  */
 
+#include <linux/ratelimit.h>
 #include <linux/pci.h>
 #include <linux/pci-ats.h>
 #include <linux/bitmap.h>
@@ -28,6 +29,8 @@
 #include <linux/iommu.h>
 #include <linux/delay.h>
 #include <linux/amd-iommu.h>
+#include <linux/notifier.h>
+#include <linux/export.h>
 #include <asm/msidef.h>
 #include <asm/proto.h>
 #include <asm/iommu.h>
@@ -41,6 +44,24 @@
 
 #define LOOP_TIMEOUT	100000
 
+/*
+ * This bitmap is used to advertise the page sizes our hardware support
+ * to the IOMMU core, which will then use this information to split
+ * physically contiguous memory regions it is mapping into page sizes
+ * that we support.
+ *
+ * Traditionally the IOMMU core just handed us the mappings directly,
+ * after making sure the size is an order of a 4KiB page and that the
+ * mapping has natural alignment.
+ *
+ * To retain this behavior, we currently advertise that we support
+ * all page sizes that are an order of 4KiB.
+ *
+ * If at some point we'd like to utilize the IOMMU core's new behavior,
+ * we could change this to advertise the real page sizes we support.
+ */
+#define AMD_IOMMU_PGSIZES	(~0xFFFUL)
+
 static DEFINE_RWLOCK(amd_iommu_devtable_lock);
 
 /* A list of preallocated protection domains */
@@ -59,6 +80,9 @@ static struct protection_domain *pt_domain;
 
 static struct iommu_ops amd_iommu_ops;
 
+static ATOMIC_NOTIFIER_HEAD(ppr_notifier);
+int amd_iommu_max_glx_val = -1;
+
 /*
  * general struct to manage commands send to an IOMMU
  */
@@ -67,6 +91,7 @@ struct iommu_cmd {
 };
 
 static void update_domain(struct protection_domain *domain);
+static int __init alloc_passthrough_domain(void);
 
 /****************************************************************************
  *
@@ -147,6 +172,33 @@ static struct iommu_dev_data *get_dev_data(struct device *dev)
 	return dev->archdata.iommu;
 }
 
+static bool pci_iommuv2_capable(struct pci_dev *pdev)
+{
+	static const int caps[] = {
+		PCI_EXT_CAP_ID_ATS,
+		PCI_EXT_CAP_ID_PRI,
+		PCI_EXT_CAP_ID_PASID,
+	};
+	int i, pos;
+
+	for (i = 0; i < 3; ++i) {
+		pos = pci_find_ext_capability(pdev, caps[i]);
+		if (pos == 0)
+			return false;
+	}
+
+	return true;
+}
+
+static bool pdev_pri_erratum(struct pci_dev *pdev, u32 erratum)
+{
+	struct iommu_dev_data *dev_data;
+
+	dev_data = get_dev_data(&pdev->dev);
+
+	return dev_data->errata & (1 << erratum) ? true : false;
+}
+
 /*
  * In this function the list of preallocated protection domains is traversed to
  * find the domain for a specific device
@@ -204,6 +256,7 @@ static bool check_device(struct device *dev)
 
 static int iommu_init_device(struct device *dev)
 {
+	struct pci_dev *pdev = to_pci_dev(dev);
 	struct iommu_dev_data *dev_data;
 	u16 alias;
 
@@ -228,6 +281,13 @@ static int iommu_init_device(struct device *dev)
 		dev_data->alias_data = alias_data;
 	}
 
+	if (pci_iommuv2_capable(pdev)) {
+		struct amd_iommu *iommu;
+
+		iommu              = amd_iommu_rlookup_table[dev_data->devid];
+		dev_data->iommu_v2 = iommu->is_iommu_v2;
+	}
+
 	dev->archdata.iommu = dev_data;
 
 	return 0;
@@ -317,6 +377,11 @@ DECLARE_STATS_COUNTER(domain_flush_single);
 DECLARE_STATS_COUNTER(domain_flush_all);
 DECLARE_STATS_COUNTER(alloced_io_mem);
 DECLARE_STATS_COUNTER(total_map_requests);
+DECLARE_STATS_COUNTER(complete_ppr);
+DECLARE_STATS_COUNTER(invalidate_iotlb);
+DECLARE_STATS_COUNTER(invalidate_iotlb_all);
+DECLARE_STATS_COUNTER(pri_requests);
+
 
 static struct dentry *stats_dir;
 static struct dentry *de_fflush;
@@ -351,6 +416,10 @@ static void amd_iommu_stats_init(void)
 	amd_iommu_stats_add(&domain_flush_all);
 	amd_iommu_stats_add(&alloced_io_mem);
 	amd_iommu_stats_add(&total_map_requests);
+	amd_iommu_stats_add(&complete_ppr);
+	amd_iommu_stats_add(&invalidate_iotlb);
+	amd_iommu_stats_add(&invalidate_iotlb_all);
+	amd_iommu_stats_add(&pri_requests);
 }
 
 #endif
@@ -365,8 +434,8 @@ static void dump_dte_entry(u16 devid)
 {
 	int i;
 
-	for (i = 0; i < 8; ++i)
-		pr_err("AMD-Vi: DTE[%d]: %08x\n", i,
+	for (i = 0; i < 4; ++i)
+		pr_err("AMD-Vi: DTE[%d]: %016llx\n", i,
 			amd_iommu_dev_table[devid].data[i]);
 }
 
@@ -461,12 +530,84 @@ static void iommu_poll_events(struct amd_iommu *iommu)
 	spin_unlock_irqrestore(&iommu->lock, flags);
 }
 
+static void iommu_handle_ppr_entry(struct amd_iommu *iommu, u32 head)
+{
+	struct amd_iommu_fault fault;
+	volatile u64 *raw;
+	int i;
+
+	INC_STATS_COUNTER(pri_requests);
+
+	raw = (u64 *)(iommu->ppr_log + head);
+
+	/*
+	 * Hardware bug: Interrupt may arrive before the entry is written to
+	 * memory. If this happens we need to wait for the entry to arrive.
+	 */
+	for (i = 0; i < LOOP_TIMEOUT; ++i) {
+		if (PPR_REQ_TYPE(raw[0]) != 0)
+			break;
+		udelay(1);
+	}
+
+	if (PPR_REQ_TYPE(raw[0]) != PPR_REQ_FAULT) {
+		pr_err_ratelimited("AMD-Vi: Unknown PPR request received\n");
+		return;
+	}
+
+	fault.address   = raw[1];
+	fault.pasid     = PPR_PASID(raw[0]);
+	fault.device_id = PPR_DEVID(raw[0]);
+	fault.tag       = PPR_TAG(raw[0]);
+	fault.flags     = PPR_FLAGS(raw[0]);
+
+	/*
+	 * To detect the hardware bug we need to clear the entry
+	 * to back to zero.
+	 */
+	raw[0] = raw[1] = 0;
+
+	atomic_notifier_call_chain(&ppr_notifier, 0, &fault);
+}
+
+static void iommu_poll_ppr_log(struct amd_iommu *iommu)
+{
+	unsigned long flags;
+	u32 head, tail;
+
+	if (iommu->ppr_log == NULL)
+		return;
+
+	spin_lock_irqsave(&iommu->lock, flags);
+
+	head = readl(iommu->mmio_base + MMIO_PPR_HEAD_OFFSET);
+	tail = readl(iommu->mmio_base + MMIO_PPR_TAIL_OFFSET);
+
+	while (head != tail) {
+
+		/* Handle PPR entry */
+		iommu_handle_ppr_entry(iommu, head);
+
+		/* Update and refresh ring-buffer state*/
+		head = (head + PPR_ENTRY_SIZE) % PPR_LOG_SIZE;
+		writel(head, iommu->mmio_base + MMIO_PPR_HEAD_OFFSET);
+		tail = readl(iommu->mmio_base + MMIO_PPR_TAIL_OFFSET);
+	}
+
+	/* enable ppr interrupts again */
+	writel(MMIO_STATUS_PPR_INT_MASK, iommu->mmio_base + MMIO_STATUS_OFFSET);
+
+	spin_unlock_irqrestore(&iommu->lock, flags);
+}
+
 irqreturn_t amd_iommu_int_thread(int irq, void *data)
 {
 	struct amd_iommu *iommu;
 
-	for_each_iommu(iommu)
+	for_each_iommu(iommu) {
 		iommu_poll_events(iommu);
+		iommu_poll_ppr_log(iommu);
+	}
 
 	return IRQ_HANDLED;
 }
@@ -595,6 +736,60 @@ static void build_inv_iotlb_pages(struct iommu_cmd *cmd, u16 devid, int qdep,
 		cmd->data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK;
 }
 
+static void build_inv_iommu_pasid(struct iommu_cmd *cmd, u16 domid, int pasid,
+				  u64 address, bool size)
+{
+	memset(cmd, 0, sizeof(*cmd));
+
+	address &= ~(0xfffULL);
+
+	cmd->data[0]  = pasid & PASID_MASK;
+	cmd->data[1]  = domid;
+	cmd->data[2]  = lower_32_bits(address);
+	cmd->data[3]  = upper_32_bits(address);
+	cmd->data[2] |= CMD_INV_IOMMU_PAGES_PDE_MASK;
+	cmd->data[2] |= CMD_INV_IOMMU_PAGES_GN_MASK;
+	if (size)
+		cmd->data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK;
+	CMD_SET_TYPE(cmd, CMD_INV_IOMMU_PAGES);
+}
+
+static void build_inv_iotlb_pasid(struct iommu_cmd *cmd, u16 devid, int pasid,
+				  int qdep, u64 address, bool size)
+{
+	memset(cmd, 0, sizeof(*cmd));
+
+	address &= ~(0xfffULL);
+
+	cmd->data[0]  = devid;
+	cmd->data[0] |= (pasid & 0xff) << 16;
+	cmd->data[0] |= (qdep  & 0xff) << 24;
+	cmd->data[1]  = devid;
+	cmd->data[1] |= ((pasid >> 8) & 0xfff) << 16;
+	cmd->data[2]  = lower_32_bits(address);
+	cmd->data[2] |= CMD_INV_IOMMU_PAGES_GN_MASK;
+	cmd->data[3]  = upper_32_bits(address);
+	if (size)
+		cmd->data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK;
+	CMD_SET_TYPE(cmd, CMD_INV_IOTLB_PAGES);
+}
+
+static void build_complete_ppr(struct iommu_cmd *cmd, u16 devid, int pasid,
+			       int status, int tag, bool gn)
+{
+	memset(cmd, 0, sizeof(*cmd));
+
+	cmd->data[0]  = devid;
+	if (gn) {
+		cmd->data[1]  = pasid & PASID_MASK;
+		cmd->data[2]  = CMD_INV_IOMMU_PAGES_GN_MASK;
+	}
+	cmd->data[3]  = tag & 0x1ff;
+	cmd->data[3] |= (status & PPR_STATUS_MASK) << PPR_STATUS_SHIFT;
+
+	CMD_SET_TYPE(cmd, CMD_COMPLETE_PPR);
+}
+
 static void build_inv_all(struct iommu_cmd *cmd)
 {
 	memset(cmd, 0, sizeof(*cmd));
@@ -1496,6 +1691,48 @@ static void free_pagetable(struct protection_domain *domain)
 	domain->pt_root = NULL;
 }
 
+static void free_gcr3_tbl_level1(u64 *tbl)
+{
+	u64 *ptr;
+	int i;
+
+	for (i = 0; i < 512; ++i) {
+		if (!(tbl[i] & GCR3_VALID))
+			continue;
+
+		ptr = __va(tbl[i] & PAGE_MASK);
+
+		free_page((unsigned long)ptr);
+	}
+}
+
+static void free_gcr3_tbl_level2(u64 *tbl)
+{
+	u64 *ptr;
+	int i;
+
+	for (i = 0; i < 512; ++i) {
+		if (!(tbl[i] & GCR3_VALID))
+			continue;
+
+		ptr = __va(tbl[i] & PAGE_MASK);
+
+		free_gcr3_tbl_level1(ptr);
+	}
+}
+
+static void free_gcr3_table(struct protection_domain *domain)
+{
+	if (domain->glx == 2)
+		free_gcr3_tbl_level2(domain->gcr3_tbl);
+	else if (domain->glx == 1)
+		free_gcr3_tbl_level1(domain->gcr3_tbl);
+	else if (domain->glx != 0)
+		BUG();
+
+	free_page((unsigned long)domain->gcr3_tbl);
+}
+
 /*
  * Free a domain, only used if something went wrong in the
  * allocation path and we need to free an already allocated page table
@@ -1582,20 +1819,52 @@ static bool dma_ops_domain(struct protection_domain *domain)
 
 static void set_dte_entry(u16 devid, struct protection_domain *domain, bool ats)
 {
-	u64 pte_root = virt_to_phys(domain->pt_root);
-	u32 flags = 0;
+	u64 pte_root = 0;
+	u64 flags = 0;
+
+	if (domain->mode != PAGE_MODE_NONE)
+		pte_root = virt_to_phys(domain->pt_root);
 
 	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;
 
+	flags = amd_iommu_dev_table[devid].data[1];
+
 	if (ats)
 		flags |= DTE_FLAG_IOTLB;
 
-	amd_iommu_dev_table[devid].data[3] |= flags;
-	amd_iommu_dev_table[devid].data[2]  = domain->id;
-	amd_iommu_dev_table[devid].data[1]  = upper_32_bits(pte_root);
-	amd_iommu_dev_table[devid].data[0]  = lower_32_bits(pte_root);
+	if (domain->flags & PD_IOMMUV2_MASK) {
+		u64 gcr3 = __pa(domain->gcr3_tbl);
+		u64 glx  = domain->glx;
+		u64 tmp;
+
+		pte_root |= DTE_FLAG_GV;
+		pte_root |= (glx & DTE_GLX_MASK) << DTE_GLX_SHIFT;
+
+		/* First mask out possible old values for GCR3 table */
+		tmp = DTE_GCR3_VAL_B(~0ULL) << DTE_GCR3_SHIFT_B;
+		flags    &= ~tmp;
+
+		tmp = DTE_GCR3_VAL_C(~0ULL) << DTE_GCR3_SHIFT_C;
+		flags    &= ~tmp;
+
+		/* Encode GCR3 table into DTE */
+		tmp = DTE_GCR3_VAL_A(gcr3) << DTE_GCR3_SHIFT_A;
+		pte_root |= tmp;
+
+		tmp = DTE_GCR3_VAL_B(gcr3) << DTE_GCR3_SHIFT_B;
+		flags    |= tmp;
+
+		tmp = DTE_GCR3_VAL_C(gcr3) << DTE_GCR3_SHIFT_C;
+		flags    |= tmp;
+	}
+
+	flags &= ~(0xffffUL);
+	flags |= domain->id;
+
+	amd_iommu_dev_table[devid].data[1]  = flags;
+	amd_iommu_dev_table[devid].data[0]  = pte_root;
 }
 
 static void clear_dte_entry(u16 devid)
@@ -1603,7 +1872,6 @@ static void clear_dte_entry(u16 devid)
 	/* 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;
 
 	amd_iommu_apply_erratum_63(devid);
 }
@@ -1696,6 +1964,93 @@ out_unlock:
 	return ret;
 }
 
+
+static void pdev_iommuv2_disable(struct pci_dev *pdev)
+{
+	pci_disable_ats(pdev);
+	pci_disable_pri(pdev);
+	pci_disable_pasid(pdev);
+}
+
+/* FIXME: Change generic reset-function to do the same */
+static int pri_reset_while_enabled(struct pci_dev *pdev)
+{
+	u16 control;
+	int pos;
+
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
+	if (!pos)
+		return -EINVAL;
+
+	pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+	control |= PCI_PRI_CTRL_RESET;
+	pci_write_config_word(pdev, pos + PCI_PRI_CTRL, control);
+
+	return 0;
+}
+
+static int pdev_iommuv2_enable(struct pci_dev *pdev)
+{
+	bool reset_enable;
+	int reqs, ret;
+
+	/* FIXME: Hardcode number of outstanding requests for now */
+	reqs = 32;
+	if (pdev_pri_erratum(pdev, AMD_PRI_DEV_ERRATUM_LIMIT_REQ_ONE))
+		reqs = 1;
+	reset_enable = pdev_pri_erratum(pdev, AMD_PRI_DEV_ERRATUM_ENABLE_RESET);
+
+	/* Only allow access to user-accessible pages */
+	ret = pci_enable_pasid(pdev, 0);
+	if (ret)
+		goto out_err;
+
+	/* First reset the PRI state of the device */
+	ret = pci_reset_pri(pdev);
+	if (ret)
+		goto out_err;
+
+	/* Enable PRI */
+	ret = pci_enable_pri(pdev, reqs);
+	if (ret)
+		goto out_err;
+
+	if (reset_enable) {
+		ret = pri_reset_while_enabled(pdev);
+		if (ret)
+			goto out_err;
+	}
+
+	ret = pci_enable_ats(pdev, PAGE_SHIFT);
+	if (ret)
+		goto out_err;
+
+	return 0;
+
+out_err:
+	pci_disable_pri(pdev);
+	pci_disable_pasid(pdev);
+
+	return ret;
+}
+
+/* FIXME: Move this to PCI code */
+#define PCI_PRI_TLP_OFF		(1 << 2)
+
+bool pci_pri_tlp_required(struct pci_dev *pdev)
+{
+	u16 control;
+	int pos;
+
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
+	if (!pos)
+		return false;
+
+	pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+
+	return (control & PCI_PRI_TLP_OFF) ? true : false;
+}
+
 /*
  * If a device is not yet associated with a domain, this function does
  * assigns it visible for the hardware
@@ -1710,7 +2065,18 @@ static int attach_device(struct device *dev,
 
 	dev_data = get_dev_data(dev);
 
-	if (amd_iommu_iotlb_sup && pci_enable_ats(pdev, PAGE_SHIFT) == 0) {
+	if (domain->flags & PD_IOMMUV2_MASK) {
+		if (!dev_data->iommu_v2 || !dev_data->passthrough)
+			return -EINVAL;
+
+		if (pdev_iommuv2_enable(pdev) != 0)
+			return -EINVAL;
+
+		dev_data->ats.enabled = true;
+		dev_data->ats.qdep    = pci_ats_queue_depth(pdev);
+		dev_data->pri_tlp     = pci_pri_tlp_required(pdev);
+	} else if (amd_iommu_iotlb_sup &&
+		   pci_enable_ats(pdev, PAGE_SHIFT) == 0) {
 		dev_data->ats.enabled = true;
 		dev_data->ats.qdep    = pci_ats_queue_depth(pdev);
 	}
@@ -1760,7 +2126,7 @@ static void __detach_device(struct iommu_dev_data *dev_data)
 	 * passthrough domain if it is detached from any other domain.
 	 * Make sure we can deassign from the pt_domain itself.
 	 */
-	if (iommu_pass_through &&
+	if (dev_data->passthrough &&
 	    (dev_data->domain == NULL && domain != pt_domain))
 		__attach_device(dev_data, pt_domain);
 }
@@ -1770,20 +2136,24 @@ static void __detach_device(struct iommu_dev_data *dev_data)
  */
 static void detach_device(struct device *dev)
 {
+	struct protection_domain *domain;
 	struct iommu_dev_data *dev_data;
 	unsigned long flags;
 
 	dev_data = get_dev_data(dev);
+	domain   = dev_data->domain;
 
 	/* lock device table */
 	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
 	__detach_device(dev_data);
 	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
 
-	if (dev_data->ats.enabled) {
+	if (domain->flags & PD_IOMMUV2_MASK)
+		pdev_iommuv2_disable(to_pci_dev(dev));
+	else if (dev_data->ats.enabled)
 		pci_disable_ats(to_pci_dev(dev));
-		dev_data->ats.enabled = false;
-	}
+
+	dev_data->ats.enabled = false;
 }
 
 /*
@@ -1818,18 +2188,20 @@ static struct protection_domain *domain_for_device(struct device *dev)
 static int device_change_notifier(struct notifier_block *nb,
 				  unsigned long action, void *data)
 {
-	struct device *dev = data;
-	u16 devid;
-	struct protection_domain *domain;
 	struct dma_ops_domain *dma_domain;
+	struct protection_domain *domain;
+	struct iommu_dev_data *dev_data;
+	struct device *dev = data;
 	struct amd_iommu *iommu;
 	unsigned long flags;
+	u16 devid;
 
 	if (!check_device(dev))
 		return 0;
 
-	devid  = get_device_id(dev);
-	iommu  = amd_iommu_rlookup_table[devid];
+	devid    = get_device_id(dev);
+	iommu    = amd_iommu_rlookup_table[devid];
+	dev_data = get_dev_data(dev);
 
 	switch (action) {
 	case BUS_NOTIFY_UNBOUND_DRIVER:
@@ -1838,7 +2210,7 @@ static int device_change_notifier(struct notifier_block *nb,
 
 		if (!domain)
 			goto out;
-		if (iommu_pass_through)
+		if (dev_data->passthrough)
 			break;
 		detach_device(dev);
 		break;
@@ -2434,8 +2806,9 @@ static int amd_iommu_dma_supported(struct device *dev, u64 mask)
  */
 static void prealloc_protection_domains(void)
 {
-	struct pci_dev *dev = NULL;
+	struct iommu_dev_data *dev_data;
 	struct dma_ops_domain *dma_dom;
+	struct pci_dev *dev = NULL;
 	u16 devid;
 
 	for_each_pci_dev(dev) {
@@ -2444,6 +2817,16 @@ static void prealloc_protection_domains(void)
 		if (!check_device(&dev->dev))
 			continue;
 
+		dev_data = get_dev_data(&dev->dev);
+		if (!amd_iommu_force_isolation && dev_data->iommu_v2) {
+			/* Make sure passthrough domain is allocated */
+			alloc_passthrough_domain();
+			dev_data->passthrough = true;
+			attach_device(&dev->dev, pt_domain);
+			pr_info("AMD-Vi: Using passthough domain for device %s\n",
+				dev_name(&dev->dev));
+		}
+
 		/* Is there already any domain for it? */
 		if (domain_for_device(&dev->dev))
 			continue;
@@ -2474,6 +2857,7 @@ static struct dma_map_ops amd_iommu_dma_ops = {
 
 static unsigned device_dma_ops_init(void)
 {
+	struct iommu_dev_data *dev_data;
 	struct pci_dev *pdev = NULL;
 	unsigned unhandled = 0;
 
@@ -2483,7 +2867,12 @@ static unsigned device_dma_ops_init(void)
 			continue;
 		}
 
-		pdev->dev.archdata.dma_ops = &amd_iommu_dma_ops;
+		dev_data = get_dev_data(&pdev->dev);
+
+		if (!dev_data->passthrough)
+			pdev->dev.archdata.dma_ops = &amd_iommu_dma_ops;
+		else
+			pdev->dev.archdata.dma_ops = &nommu_dma_ops;
 	}
 
 	return unhandled;
@@ -2610,6 +2999,20 @@ out_err:
 	return NULL;
 }
 
+static int __init alloc_passthrough_domain(void)
+{
+	if (pt_domain != NULL)
+		return 0;
+
+	/* allocate passthrough domain */
+	pt_domain = protection_domain_alloc();
+	if (!pt_domain)
+		return -ENOMEM;
+
+	pt_domain->mode = PAGE_MODE_NONE;
+
+	return 0;
+}
 static int amd_iommu_domain_init(struct iommu_domain *dom)
 {
 	struct protection_domain *domain;
@@ -2623,6 +3026,8 @@ static int amd_iommu_domain_init(struct iommu_domain *dom)
 	if (!domain->pt_root)
 		goto out_free;
 
+	domain->iommu_domain = dom;
+
 	dom->priv = domain;
 
 	return 0;
@@ -2645,7 +3050,11 @@ static void amd_iommu_domain_destroy(struct iommu_domain *dom)
 
 	BUG_ON(domain->dev_cnt != 0);
 
-	free_pagetable(domain);
+	if (domain->mode != PAGE_MODE_NONE)
+		free_pagetable(domain);
+
+	if (domain->flags & PD_IOMMUV2_MASK)
+		free_gcr3_table(domain);
 
 	protection_domain_free(domain);
 
@@ -2702,13 +3111,15 @@ static int amd_iommu_attach_device(struct iommu_domain *dom,
 }
 
 static int amd_iommu_map(struct iommu_domain *dom, unsigned long iova,
-			 phys_addr_t paddr, int gfp_order, int iommu_prot)
+			 phys_addr_t paddr, size_t page_size, int iommu_prot)
 {
-	unsigned long page_size = 0x1000UL << gfp_order;
 	struct protection_domain *domain = dom->priv;
 	int prot = 0;
 	int ret;
 
+	if (domain->mode == PAGE_MODE_NONE)
+		return -EINVAL;
+
 	if (iommu_prot & IOMMU_READ)
 		prot |= IOMMU_PROT_IR;
 	if (iommu_prot & IOMMU_WRITE)
@@ -2721,13 +3132,14 @@ static int amd_iommu_map(struct iommu_domain *dom, unsigned long iova,
 	return ret;
 }
 
-static int amd_iommu_unmap(struct iommu_domain *dom, unsigned long iova,
-			   int gfp_order)
+static size_t amd_iommu_unmap(struct iommu_domain *dom, unsigned long iova,
+			   size_t page_size)
 {
 	struct protection_domain *domain = dom->priv;
-	unsigned long page_size, unmap_size;
+	size_t unmap_size;
 
-	page_size  = 0x1000UL << gfp_order;
+	if (domain->mode == PAGE_MODE_NONE)
+		return -EINVAL;
 
 	mutex_lock(&domain->api_lock);
 	unmap_size = iommu_unmap_page(domain, iova, page_size);
@@ -2735,7 +3147,7 @@ static int amd_iommu_unmap(struct iommu_domain *dom, unsigned long iova,
 
 	domain_flush_tlb_pde(domain);
 
-	return get_order(unmap_size);
+	return unmap_size;
 }
 
 static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
@@ -2746,6 +3158,9 @@ static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
 	phys_addr_t paddr;
 	u64 *pte, __pte;
 
+	if (domain->mode == PAGE_MODE_NONE)
+		return iova;
+
 	pte = fetch_pte(domain, iova);
 
 	if (!pte || !IOMMU_PTE_PRESENT(*pte))
@@ -2773,6 +3188,26 @@ static int amd_iommu_domain_has_cap(struct iommu_domain *domain,
 	return 0;
 }
 
+static int amd_iommu_device_group(struct device *dev, unsigned int *groupid)
+{
+	struct iommu_dev_data *dev_data = dev->archdata.iommu;
+	struct pci_dev *pdev = to_pci_dev(dev);
+	u16 devid;
+
+	if (!dev_data)
+		return -ENODEV;
+
+	if (pdev->is_virtfn || !iommu_group_mf)
+		devid = dev_data->devid;
+	else
+		devid = calc_devid(pdev->bus->number,
+				   PCI_DEVFN(PCI_SLOT(pdev->devfn), 0));
+
+	*groupid = amd_iommu_alias_table[devid];
+
+	return 0;
+}
+
 static struct iommu_ops amd_iommu_ops = {
 	.domain_init = amd_iommu_domain_init,
 	.domain_destroy = amd_iommu_domain_destroy,
@@ -2782,6 +3217,8 @@ static struct iommu_ops amd_iommu_ops = {
 	.unmap = amd_iommu_unmap,
 	.iova_to_phys = amd_iommu_iova_to_phys,
 	.domain_has_cap = amd_iommu_domain_has_cap,
+	.device_group = amd_iommu_device_group,
+	.pgsize_bitmap	= AMD_IOMMU_PGSIZES,
 };
 
 /*****************************************************************************
@@ -2796,21 +3233,23 @@ static struct iommu_ops amd_iommu_ops = {
 
 int __init amd_iommu_init_passthrough(void)
 {
-	struct amd_iommu *iommu;
+	struct iommu_dev_data *dev_data;
 	struct pci_dev *dev = NULL;
+	struct amd_iommu *iommu;
 	u16 devid;
+	int ret;
 
-	/* allocate passthrough domain */
-	pt_domain = protection_domain_alloc();
-	if (!pt_domain)
-		return -ENOMEM;
-
-	pt_domain->mode |= PAGE_MODE_NONE;
+	ret = alloc_passthrough_domain();
+	if (ret)
+		return ret;
 
 	for_each_pci_dev(dev) {
 		if (!check_device(&dev->dev))
 			continue;
 
+		dev_data = get_dev_data(&dev->dev);
+		dev_data->passthrough = true;
+
 		devid = get_device_id(&dev->dev);
 
 		iommu = amd_iommu_rlookup_table[devid];
@@ -2820,7 +3259,375 @@ int __init amd_iommu_init_passthrough(void)
 		attach_device(&dev->dev, pt_domain);
 	}
 
+	amd_iommu_stats_init();
+
 	pr_info("AMD-Vi: Initialized for Passthrough Mode\n");
 
 	return 0;
 }
+
+/* IOMMUv2 specific functions */
+int amd_iommu_register_ppr_notifier(struct notifier_block *nb)
+{
+	return atomic_notifier_chain_register(&ppr_notifier, nb);
+}
+EXPORT_SYMBOL(amd_iommu_register_ppr_notifier);
+
+int amd_iommu_unregister_ppr_notifier(struct notifier_block *nb)
+{
+	return atomic_notifier_chain_unregister(&ppr_notifier, nb);
+}
+EXPORT_SYMBOL(amd_iommu_unregister_ppr_notifier);
+
+void amd_iommu_domain_direct_map(struct iommu_domain *dom)
+{
+	struct protection_domain *domain = dom->priv;
+	unsigned long flags;
+
+	spin_lock_irqsave(&domain->lock, flags);
+
+	/* Update data structure */
+	domain->mode    = PAGE_MODE_NONE;
+	domain->updated = true;
+
+	/* Make changes visible to IOMMUs */
+	update_domain(domain);
+
+	/* Page-table is not visible to IOMMU anymore, so free it */
+	free_pagetable(domain);
+
+	spin_unlock_irqrestore(&domain->lock, flags);
+}
+EXPORT_SYMBOL(amd_iommu_domain_direct_map);
+
+int amd_iommu_domain_enable_v2(struct iommu_domain *dom, int pasids)
+{
+	struct protection_domain *domain = dom->priv;
+	unsigned long flags;
+	int levels, ret;
+
+	if (pasids <= 0 || pasids > (PASID_MASK + 1))
+		return -EINVAL;
+
+	/* Number of GCR3 table levels required */
+	for (levels = 0; (pasids - 1) & ~0x1ff; pasids >>= 9)
+		levels += 1;
+
+	if (levels > amd_iommu_max_glx_val)
+		return -EINVAL;
+
+	spin_lock_irqsave(&domain->lock, flags);
+
+	/*
+	 * Save us all sanity checks whether devices already in the
+	 * domain support IOMMUv2. Just force that the domain has no
+	 * devices attached when it is switched into IOMMUv2 mode.
+	 */
+	ret = -EBUSY;
+	if (domain->dev_cnt > 0 || domain->flags & PD_IOMMUV2_MASK)
+		goto out;
+
+	ret = -ENOMEM;
+	domain->gcr3_tbl = (void *)get_zeroed_page(GFP_ATOMIC);
+	if (domain->gcr3_tbl == NULL)
+		goto out;
+
+	domain->glx      = levels;
+	domain->flags   |= PD_IOMMUV2_MASK;
+	domain->updated  = true;
+
+	update_domain(domain);
+
+	ret = 0;
+
+out:
+	spin_unlock_irqrestore(&domain->lock, flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_domain_enable_v2);
+
+static int __flush_pasid(struct protection_domain *domain, int pasid,
+			 u64 address, bool size)
+{
+	struct iommu_dev_data *dev_data;
+	struct iommu_cmd cmd;
+	int i, ret;
+
+	if (!(domain->flags & PD_IOMMUV2_MASK))
+		return -EINVAL;
+
+	build_inv_iommu_pasid(&cmd, domain->id, pasid, address, size);
+
+	/*
+	 * IOMMU TLB needs to be flushed before Device TLB to
+	 * prevent device TLB refill from IOMMU TLB
+	 */
+	for (i = 0; i < amd_iommus_present; ++i) {
+		if (domain->dev_iommu[i] == 0)
+			continue;
+
+		ret = iommu_queue_command(amd_iommus[i], &cmd);
+		if (ret != 0)
+			goto out;
+	}
+
+	/* Wait until IOMMU TLB flushes are complete */
+	domain_flush_complete(domain);
+
+	/* Now flush device TLBs */
+	list_for_each_entry(dev_data, &domain->dev_list, list) {
+		struct amd_iommu *iommu;
+		int qdep;
+
+		BUG_ON(!dev_data->ats.enabled);
+
+		qdep  = dev_data->ats.qdep;
+		iommu = amd_iommu_rlookup_table[dev_data->devid];
+
+		build_inv_iotlb_pasid(&cmd, dev_data->devid, pasid,
+				      qdep, address, size);
+
+		ret = iommu_queue_command(iommu, &cmd);
+		if (ret != 0)
+			goto out;
+	}
+
+	/* Wait until all device TLBs are flushed */
+	domain_flush_complete(domain);
+
+	ret = 0;
+
+out:
+
+	return ret;
+}
+
+static int __amd_iommu_flush_page(struct protection_domain *domain, int pasid,
+				  u64 address)
+{
+	INC_STATS_COUNTER(invalidate_iotlb);
+
+	return __flush_pasid(domain, pasid, address, false);
+}
+
+int amd_iommu_flush_page(struct iommu_domain *dom, int pasid,
+			 u64 address)
+{
+	struct protection_domain *domain = dom->priv;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&domain->lock, flags);
+	ret = __amd_iommu_flush_page(domain, pasid, address);
+	spin_unlock_irqrestore(&domain->lock, flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_flush_page);
+
+static int __amd_iommu_flush_tlb(struct protection_domain *domain, int pasid)
+{
+	INC_STATS_COUNTER(invalidate_iotlb_all);
+
+	return __flush_pasid(domain, pasid, CMD_INV_IOMMU_ALL_PAGES_ADDRESS,
+			     true);
+}
+
+int amd_iommu_flush_tlb(struct iommu_domain *dom, int pasid)
+{
+	struct protection_domain *domain = dom->priv;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&domain->lock, flags);
+	ret = __amd_iommu_flush_tlb(domain, pasid);
+	spin_unlock_irqrestore(&domain->lock, flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_flush_tlb);
+
+static u64 *__get_gcr3_pte(u64 *root, int level, int pasid, bool alloc)
+{
+	int index;
+	u64 *pte;
+
+	while (true) {
+
+		index = (pasid >> (9 * level)) & 0x1ff;
+		pte   = &root[index];
+
+		if (level == 0)
+			break;
+
+		if (!(*pte & GCR3_VALID)) {
+			if (!alloc)
+				return NULL;
+
+			root = (void *)get_zeroed_page(GFP_ATOMIC);
+			if (root == NULL)
+				return NULL;
+
+			*pte = __pa(root) | GCR3_VALID;
+		}
+
+		root = __va(*pte & PAGE_MASK);
+
+		level -= 1;
+	}
+
+	return pte;
+}
+
+static int __set_gcr3(struct protection_domain *domain, int pasid,
+		      unsigned long cr3)
+{
+	u64 *pte;
+
+	if (domain->mode != PAGE_MODE_NONE)
+		return -EINVAL;
+
+	pte = __get_gcr3_pte(domain->gcr3_tbl, domain->glx, pasid, true);
+	if (pte == NULL)
+		return -ENOMEM;
+
+	*pte = (cr3 & PAGE_MASK) | GCR3_VALID;
+
+	return __amd_iommu_flush_tlb(domain, pasid);
+}
+
+static int __clear_gcr3(struct protection_domain *domain, int pasid)
+{
+	u64 *pte;
+
+	if (domain->mode != PAGE_MODE_NONE)
+		return -EINVAL;
+
+	pte = __get_gcr3_pte(domain->gcr3_tbl, domain->glx, pasid, false);
+	if (pte == NULL)
+		return 0;
+
+	*pte = 0;
+
+	return __amd_iommu_flush_tlb(domain, pasid);
+}
+
+int amd_iommu_domain_set_gcr3(struct iommu_domain *dom, int pasid,
+			      unsigned long cr3)
+{
+	struct protection_domain *domain = dom->priv;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&domain->lock, flags);
+	ret = __set_gcr3(domain, pasid, cr3);
+	spin_unlock_irqrestore(&domain->lock, flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_domain_set_gcr3);
+
+int amd_iommu_domain_clear_gcr3(struct iommu_domain *dom, int pasid)
+{
+	struct protection_domain *domain = dom->priv;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&domain->lock, flags);
+	ret = __clear_gcr3(domain, pasid);
+	spin_unlock_irqrestore(&domain->lock, flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_domain_clear_gcr3);
+
+int amd_iommu_complete_ppr(struct pci_dev *pdev, int pasid,
+			   int status, int tag)
+{
+	struct iommu_dev_data *dev_data;
+	struct amd_iommu *iommu;
+	struct iommu_cmd cmd;
+
+	INC_STATS_COUNTER(complete_ppr);
+
+	dev_data = get_dev_data(&pdev->dev);
+	iommu    = amd_iommu_rlookup_table[dev_data->devid];
+
+	build_complete_ppr(&cmd, dev_data->devid, pasid, status,
+			   tag, dev_data->pri_tlp);
+
+	return iommu_queue_command(iommu, &cmd);
+}
+EXPORT_SYMBOL(amd_iommu_complete_ppr);
+
+struct iommu_domain *amd_iommu_get_v2_domain(struct pci_dev *pdev)
+{
+	struct protection_domain *domain;
+
+	domain = get_domain(&pdev->dev);
+	if (IS_ERR(domain))
+		return NULL;
+
+	/* Only return IOMMUv2 domains */
+	if (!(domain->flags & PD_IOMMUV2_MASK))
+		return NULL;
+
+	return domain->iommu_domain;
+}
+EXPORT_SYMBOL(amd_iommu_get_v2_domain);
+
+void amd_iommu_enable_device_erratum(struct pci_dev *pdev, u32 erratum)
+{
+	struct iommu_dev_data *dev_data;
+
+	if (!amd_iommu_v2_supported())
+		return;
+
+	dev_data = get_dev_data(&pdev->dev);
+	dev_data->errata |= (1 << erratum);
+}
+EXPORT_SYMBOL(amd_iommu_enable_device_erratum);
+
+int amd_iommu_device_info(struct pci_dev *pdev,
+                          struct amd_iommu_device_info *info)
+{
+	int max_pasids;
+	int pos;
+
+	if (pdev == NULL || info == NULL)
+		return -EINVAL;
+
+	if (!amd_iommu_v2_supported())
+		return -EINVAL;
+
+	memset(info, 0, sizeof(*info));
+
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ATS);
+	if (pos)
+		info->flags |= AMD_IOMMU_DEVICE_FLAG_ATS_SUP;
+
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
+	if (pos)
+		info->flags |= AMD_IOMMU_DEVICE_FLAG_PRI_SUP;
+
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PASID);
+	if (pos) {
+		int features;
+
+		max_pasids = 1 << (9 * (amd_iommu_max_glx_val + 1));
+		max_pasids = min(max_pasids, (1 << 20));
+
+		info->flags |= AMD_IOMMU_DEVICE_FLAG_PASID_SUP;
+		info->max_pasids = min(pci_max_pasids(pdev), max_pasids);
+
+		features = pci_pasid_features(pdev);
+		if (features & PCI_PASID_CAP_EXEC)
+			info->flags |= AMD_IOMMU_DEVICE_FLAG_EXEC_SUP;
+		if (features & PCI_PASID_CAP_PRIV)
+			info->flags |= AMD_IOMMU_DEVICE_FLAG_PRIV_SUP;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(amd_iommu_device_info);

drivers/iommu/amd_iommu_init.c

@@ -25,6 +25,7 @@
 #include <linux/interrupt.h>
 #include <linux/msi.h>
 #include <linux/amd-iommu.h>
+#include <linux/export.h>
 #include <asm/pci-direct.h>
 #include <asm/iommu.h>
 #include <asm/gart.h>
@@ -141,6 +142,12 @@ int amd_iommus_present;
 bool amd_iommu_np_cache __read_mostly;
 bool amd_iommu_iotlb_sup __read_mostly = true;
 
+u32 amd_iommu_max_pasids __read_mostly = ~0;
+
+bool amd_iommu_v2_present __read_mostly;
+
+bool amd_iommu_force_isolation __read_mostly;
+
 /*
  * The ACPI table parsing functions set this variable on an error
  */
@@ -299,6 +306,16 @@ static void iommu_feature_disable(struct amd_iommu *iommu, u8 bit)
 	writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
 }
 
+static void iommu_set_inv_tlb_timeout(struct amd_iommu *iommu, int timeout)
+{
+	u32 ctrl;
+
+	ctrl = readl(iommu->mmio_base + MMIO_CONTROL_OFFSET);
+	ctrl &= ~CTRL_INV_TO_MASK;
+	ctrl |= (timeout << CONTROL_INV_TIMEOUT) & CTRL_INV_TO_MASK;
+	writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
+}
+
 /* Function to enable the hardware */
 static void iommu_enable(struct amd_iommu *iommu)
 {
@@ -581,21 +598,69 @@ static void __init free_event_buffer(struct amd_iommu *iommu)
 	free_pages((unsigned long)iommu->evt_buf, get_order(EVT_BUFFER_SIZE));
 }
 
+/* allocates the memory where the IOMMU will log its events to */
+static u8 * __init alloc_ppr_log(struct amd_iommu *iommu)
+{
+	iommu->ppr_log = (u8 *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
+						get_order(PPR_LOG_SIZE));
+
+	if (iommu->ppr_log == NULL)
+		return NULL;
+
+	return iommu->ppr_log;
+}
+
+static void iommu_enable_ppr_log(struct amd_iommu *iommu)
+{
+	u64 entry;
+
+	if (iommu->ppr_log == NULL)
+		return;
+
+	entry = (u64)virt_to_phys(iommu->ppr_log) | PPR_LOG_SIZE_512;
+
+	memcpy_toio(iommu->mmio_base + MMIO_PPR_LOG_OFFSET,
+		    &entry, sizeof(entry));
+
+	/* set head and tail to zero manually */
+	writel(0x00, iommu->mmio_base + MMIO_PPR_HEAD_OFFSET);
+	writel(0x00, iommu->mmio_base + MMIO_PPR_TAIL_OFFSET);
+
+	iommu_feature_enable(iommu, CONTROL_PPFLOG_EN);
+	iommu_feature_enable(iommu, CONTROL_PPR_EN);
+}
+
+static void __init free_ppr_log(struct amd_iommu *iommu)
+{
+	if (iommu->ppr_log == NULL)
+		return;
+
+	free_pages((unsigned long)iommu->ppr_log, get_order(PPR_LOG_SIZE));
+}
+
+static void iommu_enable_gt(struct amd_iommu *iommu)
+{
+	if (!iommu_feature(iommu, FEATURE_GT))
+		return;
+
+	iommu_feature_enable(iommu, CONTROL_GT_EN);
+}
+
 /* sets a specific bit in the device table entry. */
 static void set_dev_entry_bit(u16 devid, u8 bit)
 {
-	int i = (bit >> 5) & 0x07;
-	int _bit = bit & 0x1f;
+	int i = (bit >> 6) & 0x03;
+	int _bit = bit & 0x3f;
 
-	amd_iommu_dev_table[devid].data[i] |= (1 << _bit);
+	amd_iommu_dev_table[devid].data[i] |= (1UL << _bit);
 }
 
 static int get_dev_entry_bit(u16 devid, u8 bit)
 {
-	int i = (bit >> 5) & 0x07;
-	int _bit = bit & 0x1f;
+	int i = (bit >> 6) & 0x03;
+	int _bit = bit & 0x3f;
 
-	return (amd_iommu_dev_table[devid].data[i] & (1 << _bit)) >> _bit;
+	return (amd_iommu_dev_table[devid].data[i] & (1UL << _bit)) >> _bit;
 }
 
 
@@ -699,6 +764,32 @@ static void __init init_iommu_from_pci(struct amd_iommu *iommu)
 
 	iommu->features = ((u64)high << 32) | low;
 
+	if (iommu_feature(iommu, FEATURE_GT)) {
+		int glxval;
+		u32 pasids;
+		u64 shift;
+
+		shift   = iommu->features & FEATURE_PASID_MASK;
+		shift >>= FEATURE_PASID_SHIFT;
+		pasids  = (1 << shift);
+
+		amd_iommu_max_pasids = min(amd_iommu_max_pasids, pasids);
+
+		glxval   = iommu->features & FEATURE_GLXVAL_MASK;
+		glxval >>= FEATURE_GLXVAL_SHIFT;
+
+		if (amd_iommu_max_glx_val == -1)
+			amd_iommu_max_glx_val = glxval;
+		else
+			amd_iommu_max_glx_val = min(amd_iommu_max_glx_val, glxval);
+	}
+
+	if (iommu_feature(iommu, FEATURE_GT) &&
+	    iommu_feature(iommu, FEATURE_PPR)) {
+		iommu->is_iommu_v2   = true;
+		amd_iommu_v2_present = true;
+	}
+
 	if (!is_rd890_iommu(iommu->dev))
 		return;
 
@@ -901,6 +992,7 @@ static void __init free_iommu_one(struct amd_iommu *iommu)
 {
 	free_command_buffer(iommu);
 	free_event_buffer(iommu);
+	free_ppr_log(iommu);
 	iommu_unmap_mmio_space(iommu);
 }
 
@@ -964,6 +1056,12 @@ static int __init init_iommu_one(struct amd_iommu *iommu, struct ivhd_header *h)
 	init_iommu_from_acpi(iommu, h);
 	init_iommu_devices(iommu);
 
+	if (iommu_feature(iommu, FEATURE_PPR)) {
+		iommu->ppr_log = alloc_ppr_log(iommu);
+		if (!iommu->ppr_log)
+			return -ENOMEM;
+	}
+
 	if (iommu->cap & (1UL << IOMMU_CAP_NPCACHE))
 		amd_iommu_np_cache = true;
 
@@ -1050,6 +1148,9 @@ static int iommu_setup_msi(struct amd_iommu *iommu)
 	iommu->int_enabled = true;
 	iommu_feature_enable(iommu, CONTROL_EVT_INT_EN);
 
+	if (iommu->ppr_log != NULL)
+		iommu_feature_enable(iommu, CONTROL_PPFINT_EN);
+
 	return 0;
 }
 
@@ -1209,6 +1310,9 @@ static void iommu_init_flags(struct amd_iommu *iommu)
 	 * make IOMMU memory accesses cache coherent
 	 */
 	iommu_feature_enable(iommu, CONTROL_COHERENT_EN);
+
+	/* Set IOTLB invalidation timeout to 1s */
+	iommu_set_inv_tlb_timeout(iommu, CTRL_INV_TO_1S);
 }
 
 static void iommu_apply_resume_quirks(struct amd_iommu *iommu)
@@ -1274,6 +1378,8 @@ static void enable_iommus(void)
 		iommu_set_device_table(iommu);
 		iommu_enable_command_buffer(iommu);
 		iommu_enable_event_buffer(iommu);
+		iommu_enable_ppr_log(iommu);
+		iommu_enable_gt(iommu);
 		iommu_set_exclusion_range(iommu);
 		iommu_init_msi(iommu);
 		iommu_enable(iommu);
@@ -1303,13 +1409,6 @@ static void amd_iommu_resume(void)
 
 	/* re-load the hardware */
 	enable_iommus();
-
-	/*
-	 * we have to flush after the IOMMUs are enabled because a
-	 * disabled IOMMU will never execute the commands we send
-	 */
-	for_each_iommu(iommu)
-		iommu_flush_all_caches(iommu);
 }
 
 static int amd_iommu_suspend(void)
@@ -1560,6 +1659,8 @@ static int __init parse_amd_iommu_options(char *str)
 			amd_iommu_unmap_flush = true;
 		if (strncmp(str, "off", 3) == 0)
 			amd_iommu_disabled = true;
+		if (strncmp(str, "force_isolation", 15) == 0)
+			amd_iommu_force_isolation = true;
 	}
 
 	return 1;
@@ -1572,3 +1673,9 @@ IOMMU_INIT_FINISH(amd_iommu_detect,
 		  gart_iommu_hole_init,
 		  0,
 		  0);
+
+bool amd_iommu_v2_supported(void)
+{
+	return amd_iommu_v2_present;
+}
+EXPORT_SYMBOL(amd_iommu_v2_supported);

drivers/iommu/amd_iommu_proto.h

@@ -31,6 +31,30 @@ extern int amd_iommu_init_devices(void);
 extern void amd_iommu_uninit_devices(void);
 extern void amd_iommu_init_notifier(void);
 extern void amd_iommu_init_api(void);
+
+/* IOMMUv2 specific functions */
+struct iommu_domain;
+
+extern bool amd_iommu_v2_supported(void);
+extern int amd_iommu_register_ppr_notifier(struct notifier_block *nb);
+extern int amd_iommu_unregister_ppr_notifier(struct notifier_block *nb);
+extern void amd_iommu_domain_direct_map(struct iommu_domain *dom);
+extern int amd_iommu_domain_enable_v2(struct iommu_domain *dom, int pasids);
+extern int amd_iommu_flush_page(struct iommu_domain *dom, int pasid,
+				u64 address);
+extern int amd_iommu_flush_tlb(struct iommu_domain *dom, int pasid);
+extern int amd_iommu_domain_set_gcr3(struct iommu_domain *dom, int pasid,
+				     unsigned long cr3);
+extern int amd_iommu_domain_clear_gcr3(struct iommu_domain *dom, int pasid);
+extern struct iommu_domain *amd_iommu_get_v2_domain(struct pci_dev *pdev);
+
+#define PPR_SUCCESS			0x0
+#define PPR_INVALID			0x1
+#define PPR_FAILURE			0xf
+
+extern int amd_iommu_complete_ppr(struct pci_dev *pdev, int pasid,
+				  int status, int tag);
+
 #ifndef CONFIG_AMD_IOMMU_STATS
 
 static inline void amd_iommu_stats_init(void) { }

drivers/iommu/amd_iommu_types.h

@@ -69,11 +69,14 @@
 #define MMIO_EXCL_BASE_OFFSET   0x0020
 #define MMIO_EXCL_LIMIT_OFFSET  0x0028
 #define MMIO_EXT_FEATURES	0x0030
+#define MMIO_PPR_LOG_OFFSET	0x0038
 #define MMIO_CMD_HEAD_OFFSET	0x2000
 #define MMIO_CMD_TAIL_OFFSET	0x2008
 #define MMIO_EVT_HEAD_OFFSET	0x2010
 #define MMIO_EVT_TAIL_OFFSET	0x2018
 #define MMIO_STATUS_OFFSET	0x2020
+#define MMIO_PPR_HEAD_OFFSET	0x2030
+#define MMIO_PPR_TAIL_OFFSET	0x2038
 
 
 /* Extended Feature Bits */
@@ -87,8 +90,17 @@
 #define FEATURE_HE		(1ULL<<8)
 #define FEATURE_PC		(1ULL<<9)
 
+#define FEATURE_PASID_SHIFT	32
+#define FEATURE_PASID_MASK	(0x1fULL << FEATURE_PASID_SHIFT)
+
+#define FEATURE_GLXVAL_SHIFT	14
+#define FEATURE_GLXVAL_MASK	(0x03ULL << FEATURE_GLXVAL_SHIFT)
+
+#define PASID_MASK		0x000fffff
+
 /* MMIO status bits */
-#define MMIO_STATUS_COM_WAIT_INT_MASK	0x04
+#define MMIO_STATUS_COM_WAIT_INT_MASK	(1 << 2)
+#define MMIO_STATUS_PPR_INT_MASK	(1 << 6)
 
 /* event logging constants */
 #define EVENT_ENTRY_SIZE	0x10
@@ -115,6 +127,7 @@
 #define CONTROL_EVT_LOG_EN      0x02ULL
 #define CONTROL_EVT_INT_EN      0x03ULL
 #define CONTROL_COMWAIT_EN      0x04ULL
+#define CONTROL_INV_TIMEOUT	0x05ULL
 #define CONTROL_PASSPW_EN       0x08ULL
 #define CONTROL_RESPASSPW_EN    0x09ULL
 #define CONTROL_COHERENT_EN     0x0aULL
@@ -122,18 +135,34 @@
 #define CONTROL_CMDBUF_EN       0x0cULL
 #define CONTROL_PPFLOG_EN       0x0dULL
 #define CONTROL_PPFINT_EN       0x0eULL
+#define CONTROL_PPR_EN          0x0fULL
+#define CONTROL_GT_EN           0x10ULL
+
+#define CTRL_INV_TO_MASK	(7 << CONTROL_INV_TIMEOUT)
+#define CTRL_INV_TO_NONE	0
+#define CTRL_INV_TO_1MS		1
+#define CTRL_INV_TO_10MS	2
+#define CTRL_INV_TO_100MS	3
+#define CTRL_INV_TO_1S		4
+#define CTRL_INV_TO_10S		5
+#define CTRL_INV_TO_100S	6
 
 /* command specific defines */
 #define CMD_COMPL_WAIT          0x01
 #define CMD_INV_DEV_ENTRY       0x02
 #define CMD_INV_IOMMU_PAGES	0x03
 #define CMD_INV_IOTLB_PAGES	0x04
+#define CMD_COMPLETE_PPR	0x07
 #define CMD_INV_ALL		0x08
 
 #define CMD_COMPL_WAIT_STORE_MASK	0x01
 #define CMD_COMPL_WAIT_INT_MASK		0x02
 #define CMD_INV_IOMMU_PAGES_SIZE_MASK	0x01
 #define CMD_INV_IOMMU_PAGES_PDE_MASK	0x02
+#define CMD_INV_IOMMU_PAGES_GN_MASK	0x04
+
+#define PPR_STATUS_MASK			0xf
+#define PPR_STATUS_SHIFT		12
 
 #define CMD_INV_IOMMU_ALL_PAGES_ADDRESS	0x7fffffffffffffffULL
 
@@ -165,6 +194,23 @@
 #define EVT_BUFFER_SIZE		8192 /* 512 entries */
 #define EVT_LEN_MASK		(0x9ULL << 56)
 
+/* Constants for PPR Log handling */
+#define PPR_LOG_ENTRIES		512
+#define PPR_LOG_SIZE_SHIFT	56
+#define PPR_LOG_SIZE_512	(0x9ULL << PPR_LOG_SIZE_SHIFT)
+#define PPR_ENTRY_SIZE		16
+#define PPR_LOG_SIZE		(PPR_ENTRY_SIZE * PPR_LOG_ENTRIES)
+
+#define PPR_REQ_TYPE(x)		(((x) >> 60) & 0xfULL)
+#define PPR_FLAGS(x)		(((x) >> 48) & 0xfffULL)
+#define PPR_DEVID(x)		((x) & 0xffffULL)
+#define PPR_TAG(x)		(((x) >> 32) & 0x3ffULL)
+#define PPR_PASID1(x)		(((x) >> 16) & 0xffffULL)
+#define PPR_PASID2(x)		(((x) >> 42) & 0xfULL)
+#define PPR_PASID(x)		((PPR_PASID2(x) << 16) | PPR_PASID1(x))
+
+#define PPR_REQ_FAULT		0x01
+
 #define PAGE_MODE_NONE    0x00
 #define PAGE_MODE_1_LEVEL 0x01
 #define PAGE_MODE_2_LEVEL 0x02
@@ -230,7 +276,24 @@
 #define IOMMU_PTE_IR (1ULL << 61)
 #define IOMMU_PTE_IW (1ULL << 62)
 
-#define DTE_FLAG_IOTLB	0x01
+#define DTE_FLAG_IOTLB	(0x01UL << 32)
+#define DTE_FLAG_GV	(0x01ULL << 55)
+#define DTE_GLX_SHIFT	(56)
+#define DTE_GLX_MASK	(3)
+
+#define DTE_GCR3_VAL_A(x)	(((x) >> 12) & 0x00007ULL)
+#define DTE_GCR3_VAL_B(x)	(((x) >> 15) & 0x0ffffULL)
+#define DTE_GCR3_VAL_C(x)	(((x) >> 31) & 0xfffffULL)
+
+#define DTE_GCR3_INDEX_A	0
+#define DTE_GCR3_INDEX_B	1
+#define DTE_GCR3_INDEX_C	1
+
+#define DTE_GCR3_SHIFT_A	58
+#define DTE_GCR3_SHIFT_B	16
+#define DTE_GCR3_SHIFT_C	43
+
+#define GCR3_VALID		0x01ULL
 
 #define IOMMU_PAGE_MASK (((1ULL << 52) - 1) & ~0xfffULL)
 #define IOMMU_PTE_PRESENT(pte) ((pte) & IOMMU_PTE_P)
@@ -257,6 +320,7 @@
 					      domain for an IOMMU */
 #define PD_PASSTHROUGH_MASK	(1UL << 2) /* domain has no page
 					      translation */
+#define PD_IOMMUV2_MASK		(1UL << 3) /* domain has gcr3 table */
 
 extern bool amd_iommu_dump;
 #define DUMP_printk(format, arg...)					\
@@ -285,6 +349,29 @@ extern bool amd_iommu_iotlb_sup;
 #define APERTURE_RANGE_INDEX(a)	((a) >> APERTURE_RANGE_SHIFT)
 #define APERTURE_PAGE_INDEX(a)	(((a) >> 21) & 0x3fULL)
 
+
+/*
+ * This struct is used to pass information about
+ * incoming PPR faults around.
+ */
+struct amd_iommu_fault {
+	u64 address;    /* IO virtual address of the fault*/
+	u32 pasid;      /* Address space identifier */
+	u16 device_id;  /* Originating PCI device id */
+	u16 tag;        /* PPR tag */
+	u16 flags;      /* Fault flags */
+
+};
+
+#define PPR_FAULT_EXEC	(1 << 1)
+#define PPR_FAULT_READ  (1 << 2)
+#define PPR_FAULT_WRITE (1 << 5)
+#define PPR_FAULT_USER  (1 << 6)
+#define PPR_FAULT_RSVD  (1 << 7)
+#define PPR_FAULT_GN    (1 << 8)
+
+struct iommu_domain;
+
 /*
  * This structure contains generic data for  IOMMU protection domains
  * independent of their use.
@@ -297,11 +384,15 @@ struct protection_domain {
 	u16 id;			/* the domain id written to the device table */
 	int mode;		/* paging mode (0-6 levels) */
 	u64 *pt_root;		/* page table root pointer */
+	int glx;		/* Number of levels for GCR3 table */
+	u64 *gcr3_tbl;		/* Guest CR3 table */
 	unsigned long flags;	/* flags to find out type of domain */
 	bool updated;		/* complete domain flush required */
 	unsigned dev_cnt;	/* devices assigned to this domain */
 	unsigned dev_iommu[MAX_IOMMUS]; /* per-IOMMU reference count */
 	void *priv;		/* private data */
+	struct iommu_domain *iommu_domain; /* Pointer to generic
+					      domain structure */
 
 };
 
@@ -315,10 +406,15 @@ struct iommu_dev_data {
 	struct protection_domain *domain; /* Domain the device is bound to */
 	atomic_t bind;			  /* Domain attach reverent count */
 	u16 devid;			  /* PCI Device ID */
+	bool iommu_v2;			  /* Device can make use of IOMMUv2 */
+	bool passthrough;		  /* Default for device is pt_domain */
 	struct {
 		bool enabled;
 		int qdep;
 	} ats;				  /* ATS state */
+	bool pri_tlp;			  /* PASID TLB required for
+					     PPR completions */
+	u32 errata;			  /* Bitmap for errata to apply */
 };
 
 /*
@@ -399,6 +495,9 @@ struct amd_iommu {
 	/* Extended features */
 	u64 features;
 
+	/* IOMMUv2 */
+	bool is_iommu_v2;
+
 	/*
 	 * Capability pointer. There could be more than one IOMMU per PCI
 	 * device function if there are more than one AMD IOMMU capability
@@ -431,6 +530,9 @@ struct amd_iommu {
 	/* MSI number for event interrupt */
 	u16 evt_msi_num;
 
+	/* Base of the PPR log, if present */
+	u8 *ppr_log;
+
 	/* true if interrupts for this IOMMU are already enabled */
 	bool int_enabled;
 
@@ -484,7 +586,7 @@ extern struct list_head amd_iommu_pd_list;
  * Structure defining one entry in the device table
  */
 struct dev_table_entry {
-	u32 data[8];
+	u64 data[4];
 };
 
 /*
@@ -549,6 +651,16 @@ extern unsigned long *amd_iommu_pd_alloc_bitmap;
  */
 extern bool amd_iommu_unmap_flush;
 
+/* Smallest number of PASIDs supported by any IOMMU in the system */
+extern u32 amd_iommu_max_pasids;
+
+extern bool amd_iommu_v2_present;
+
+extern bool amd_iommu_force_isolation;
+
+/* Max levels of glxval supported */
+extern int amd_iommu_max_glx_val;
+
 /* 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)

drivers/iommu/amd_iommu_v2.c

@@ -0,0 +1,994 @@
+/*
+ * Copyright (C) 2010-2012 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/mmu_notifier.h>
+#include <linux/amd-iommu.h>
+#include <linux/mm_types.h>
+#include <linux/profile.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/iommu.h>
+#include <linux/wait.h>
+#include <linux/pci.h>
+#include <linux/gfp.h>
+
+#include "amd_iommu_types.h"
+#include "amd_iommu_proto.h"
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Joerg Roedel <joerg.roedel@amd.com>");
+
+#define MAX_DEVICES		0x10000
+#define PRI_QUEUE_SIZE		512
+
+struct pri_queue {
+	atomic_t inflight;
+	bool finish;
+	int status;
+};
+
+struct pasid_state {
+	struct list_head list;			/* For global state-list */
+	atomic_t count;				/* Reference count */
+	struct task_struct *task;		/* Task bound to this PASID */
+	struct mm_struct *mm;			/* mm_struct for the faults */
+	struct mmu_notifier mn;                 /* mmu_otifier handle */
+	struct pri_queue pri[PRI_QUEUE_SIZE];	/* PRI tag states */
+	struct device_state *device_state;	/* Link to our device_state */
+	int pasid;				/* PASID index */
+	spinlock_t lock;			/* Protect pri_queues */
+	wait_queue_head_t wq;			/* To wait for count == 0 */
+};
+
+struct device_state {
+	atomic_t count;
+	struct pci_dev *pdev;
+	struct pasid_state **states;
+	struct iommu_domain *domain;
+	int pasid_levels;
+	int max_pasids;
+	amd_iommu_invalid_ppr_cb inv_ppr_cb;
+	amd_iommu_invalidate_ctx inv_ctx_cb;
+	spinlock_t lock;
+	wait_queue_head_t wq;
+};
+
+struct fault {
+	struct work_struct work;
+	struct device_state *dev_state;
+	struct pasid_state *state;
+	struct mm_struct *mm;
+	u64 address;
+	u16 devid;
+	u16 pasid;
+	u16 tag;
+	u16 finish;
+	u16 flags;
+};
+
+struct device_state **state_table;
+static spinlock_t state_lock;
+
+/* List and lock for all pasid_states */
+static LIST_HEAD(pasid_state_list);
+static DEFINE_SPINLOCK(ps_lock);
+
+static struct workqueue_struct *iommu_wq;
+
+/*
+ * Empty page table - Used between
+ * mmu_notifier_invalidate_range_start and
+ * mmu_notifier_invalidate_range_end
+ */
+static u64 *empty_page_table;
+
+static void free_pasid_states(struct device_state *dev_state);
+static void unbind_pasid(struct device_state *dev_state, int pasid);
+static int task_exit(struct notifier_block *nb, unsigned long e, void *data);
+
+static u16 device_id(struct pci_dev *pdev)
+{
+	u16 devid;
+
+	devid = pdev->bus->number;
+	devid = (devid << 8) | pdev->devfn;
+
+	return devid;
+}
+
+static struct device_state *get_device_state(u16 devid)
+{
+	struct device_state *dev_state;
+	unsigned long flags;
+
+	spin_lock_irqsave(&state_lock, flags);
+	dev_state = state_table[devid];
+	if (dev_state != NULL)
+		atomic_inc(&dev_state->count);
+	spin_unlock_irqrestore(&state_lock, flags);
+
+	return dev_state;
+}
+
+static void free_device_state(struct device_state *dev_state)
+{
+	/*
+	 * First detach device from domain - No more PRI requests will arrive
+	 * from that device after it is unbound from the IOMMUv2 domain.
+	 */
+	iommu_detach_device(dev_state->domain, &dev_state->pdev->dev);
+
+	/* Everything is down now, free the IOMMUv2 domain */
+	iommu_domain_free(dev_state->domain);
+
+	/* Finally get rid of the device-state */
+	kfree(dev_state);
+}
+
+static void put_device_state(struct device_state *dev_state)
+{
+	if (atomic_dec_and_test(&dev_state->count))
+		wake_up(&dev_state->wq);
+}
+
+static void put_device_state_wait(struct device_state *dev_state)
+{
+	DEFINE_WAIT(wait);
+
+	prepare_to_wait(&dev_state->wq, &wait, TASK_UNINTERRUPTIBLE);
+	if (!atomic_dec_and_test(&dev_state->count))
+		schedule();
+	finish_wait(&dev_state->wq, &wait);
+
+	free_device_state(dev_state);
+}
+
+static struct notifier_block profile_nb = {
+	.notifier_call = task_exit,
+};
+
+static void link_pasid_state(struct pasid_state *pasid_state)
+{
+	spin_lock(&ps_lock);
+	list_add_tail(&pasid_state->list, &pasid_state_list);
+	spin_unlock(&ps_lock);
+}
+
+static void __unlink_pasid_state(struct pasid_state *pasid_state)
+{
+	list_del(&pasid_state->list);
+}
+
+static void unlink_pasid_state(struct pasid_state *pasid_state)
+{
+	spin_lock(&ps_lock);
+	__unlink_pasid_state(pasid_state);
+	spin_unlock(&ps_lock);
+}
+
+/* Must be called under dev_state->lock */
+static struct pasid_state **__get_pasid_state_ptr(struct device_state *dev_state,
+						  int pasid, bool alloc)
+{
+	struct pasid_state **root, **ptr;
+	int level, index;
+
+	level = dev_state->pasid_levels;
+	root  = dev_state->states;
+
+	while (true) {
+
+		index = (pasid >> (9 * level)) & 0x1ff;
+		ptr   = &root[index];
+
+		if (level == 0)
+			break;
+
+		if (*ptr == NULL) {
+			if (!alloc)
+				return NULL;
+
+			*ptr = (void *)get_zeroed_page(GFP_ATOMIC);
+			if (*ptr == NULL)
+				return NULL;
+		}
+
+		root   = (struct pasid_state **)*ptr;
+		level -= 1;
+	}
+
+	return ptr;
+}
+
+static int set_pasid_state(struct device_state *dev_state,
+			   struct pasid_state *pasid_state,
+			   int pasid)
+{
+	struct pasid_state **ptr;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&dev_state->lock, flags);
+	ptr = __get_pasid_state_ptr(dev_state, pasid, true);
+
+	ret = -ENOMEM;
+	if (ptr == NULL)
+		goto out_unlock;
+
+	ret = -ENOMEM;
+	if (*ptr != NULL)
+		goto out_unlock;
+
+	*ptr = pasid_state;
+
+	ret = 0;
+
+out_unlock:
+	spin_unlock_irqrestore(&dev_state->lock, flags);
+
+	return ret;
+}
+
+static void clear_pasid_state(struct device_state *dev_state, int pasid)
+{
+	struct pasid_state **ptr;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev_state->lock, flags);
+	ptr = __get_pasid_state_ptr(dev_state, pasid, true);
+
+	if (ptr == NULL)
+		goto out_unlock;
+
+	*ptr = NULL;
+
+out_unlock:
+	spin_unlock_irqrestore(&dev_state->lock, flags);
+}
+
+static struct pasid_state *get_pasid_state(struct device_state *dev_state,
+					   int pasid)
+{
+	struct pasid_state **ptr, *ret = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev_state->lock, flags);
+	ptr = __get_pasid_state_ptr(dev_state, pasid, false);
+
+	if (ptr == NULL)
+		goto out_unlock;
+
+	ret = *ptr;
+	if (ret)
+		atomic_inc(&ret->count);
+
+out_unlock:
+	spin_unlock_irqrestore(&dev_state->lock, flags);
+
+	return ret;
+}
+
+static void free_pasid_state(struct pasid_state *pasid_state)
+{
+	kfree(pasid_state);
+}
+
+static void put_pasid_state(struct pasid_state *pasid_state)
+{
+	if (atomic_dec_and_test(&pasid_state->count)) {
+		put_device_state(pasid_state->device_state);
+		wake_up(&pasid_state->wq);
+	}
+}
+
+static void put_pasid_state_wait(struct pasid_state *pasid_state)
+{
+	DEFINE_WAIT(wait);
+
+	prepare_to_wait(&pasid_state->wq, &wait, TASK_UNINTERRUPTIBLE);
+
+	if (atomic_dec_and_test(&pasid_state->count))
+		put_device_state(pasid_state->device_state);
+	else
+		schedule();
+
+	finish_wait(&pasid_state->wq, &wait);
+	mmput(pasid_state->mm);
+	free_pasid_state(pasid_state);
+}
+
+static void __unbind_pasid(struct pasid_state *pasid_state)
+{
+	struct iommu_domain *domain;
+
+	domain = pasid_state->device_state->domain;
+
+	amd_iommu_domain_clear_gcr3(domain, pasid_state->pasid);
+	clear_pasid_state(pasid_state->device_state, pasid_state->pasid);
+
+	/* Make sure no more pending faults are in the queue */
+	flush_workqueue(iommu_wq);
+
+	mmu_notifier_unregister(&pasid_state->mn, pasid_state->mm);
+
+	put_pasid_state(pasid_state); /* Reference taken in bind() function */
+}
+
+static void unbind_pasid(struct device_state *dev_state, int pasid)
+{
+	struct pasid_state *pasid_state;
+
+	pasid_state = get_pasid_state(dev_state, pasid);
+	if (pasid_state == NULL)
+		return;
+
+	unlink_pasid_state(pasid_state);
+	__unbind_pasid(pasid_state);
+	put_pasid_state_wait(pasid_state); /* Reference taken in this function */
+}
+
+static void free_pasid_states_level1(struct pasid_state **tbl)
+{
+	int i;
+
+	for (i = 0; i < 512; ++i) {
+		if (tbl[i] == NULL)
+			continue;
+
+		free_page((unsigned long)tbl[i]);
+	}
+}
+
+static void free_pasid_states_level2(struct pasid_state **tbl)
+{
+	struct pasid_state **ptr;
+	int i;
+
+	for (i = 0; i < 512; ++i) {
+		if (tbl[i] == NULL)
+			continue;
+
+		ptr = (struct pasid_state **)tbl[i];
+		free_pasid_states_level1(ptr);
+	}
+}
+
+static void free_pasid_states(struct device_state *dev_state)
+{
+	struct pasid_state *pasid_state;
+	int i;
+
+	for (i = 0; i < dev_state->max_pasids; ++i) {
+		pasid_state = get_pasid_state(dev_state, i);
+		if (pasid_state == NULL)
+			continue;
+
+		put_pasid_state(pasid_state);
+		unbind_pasid(dev_state, i);
+	}
+
+	if (dev_state->pasid_levels == 2)
+		free_pasid_states_level2(dev_state->states);
+	else if (dev_state->pasid_levels == 1)
+		free_pasid_states_level1(dev_state->states);
+	else if (dev_state->pasid_levels != 0)
+		BUG();
+
+	free_page((unsigned long)dev_state->states);
+}
+
+static struct pasid_state *mn_to_state(struct mmu_notifier *mn)
+{
+	return container_of(mn, struct pasid_state, mn);
+}
+
+static void __mn_flush_page(struct mmu_notifier *mn,
+			    unsigned long address)
+{
+	struct pasid_state *pasid_state;
+	struct device_state *dev_state;
+
+	pasid_state = mn_to_state(mn);
+	dev_state   = pasid_state->device_state;
+
+	amd_iommu_flush_page(dev_state->domain, pasid_state->pasid, address);
+}
+
+static int mn_clear_flush_young(struct mmu_notifier *mn,
+				struct mm_struct *mm,
+				unsigned long address)
+{
+	__mn_flush_page(mn, address);
+
+	return 0;
+}
+
+static void mn_change_pte(struct mmu_notifier *mn,
+			  struct mm_struct *mm,
+			  unsigned long address,
+			  pte_t pte)
+{
+	__mn_flush_page(mn, address);
+}
+
+static void mn_invalidate_page(struct mmu_notifier *mn,
+			       struct mm_struct *mm,
+			       unsigned long address)
+{
+	__mn_flush_page(mn, address);
+}
+
+static void mn_invalidate_range_start(struct mmu_notifier *mn,
+				      struct mm_struct *mm,
+				      unsigned long start, unsigned long end)
+{
+	struct pasid_state *pasid_state;
+	struct device_state *dev_state;
+
+	pasid_state = mn_to_state(mn);
+	dev_state   = pasid_state->device_state;
+
+	amd_iommu_domain_set_gcr3(dev_state->domain, pasid_state->pasid,
+				  __pa(empty_page_table));
+}
+
+static void mn_invalidate_range_end(struct mmu_notifier *mn,
+				    struct mm_struct *mm,
+				    unsigned long start, unsigned long end)
+{
+	struct pasid_state *pasid_state;
+	struct device_state *dev_state;
+
+	pasid_state = mn_to_state(mn);
+	dev_state   = pasid_state->device_state;
+
+	amd_iommu_domain_set_gcr3(dev_state->domain, pasid_state->pasid,
+				  __pa(pasid_state->mm->pgd));
+}
+
+static struct mmu_notifier_ops iommu_mn = {
+	.clear_flush_young      = mn_clear_flush_young,
+	.change_pte             = mn_change_pte,
+	.invalidate_page        = mn_invalidate_page,
+	.invalidate_range_start = mn_invalidate_range_start,
+	.invalidate_range_end   = mn_invalidate_range_end,
+};
+
+static void set_pri_tag_status(struct pasid_state *pasid_state,
+			       u16 tag, int status)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&pasid_state->lock, flags);
+	pasid_state->pri[tag].status = status;
+	spin_unlock_irqrestore(&pasid_state->lock, flags);
+}
+
+static void finish_pri_tag(struct device_state *dev_state,
+			   struct pasid_state *pasid_state,
+			   u16 tag)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&pasid_state->lock, flags);
+	if (atomic_dec_and_test(&pasid_state->pri[tag].inflight) &&
+	    pasid_state->pri[tag].finish) {
+		amd_iommu_complete_ppr(dev_state->pdev, pasid_state->pasid,
+				       pasid_state->pri[tag].status, tag);
+		pasid_state->pri[tag].finish = false;
+		pasid_state->pri[tag].status = PPR_SUCCESS;
+	}
+	spin_unlock_irqrestore(&pasid_state->lock, flags);
+}
+
+static void do_fault(struct work_struct *work)
+{
+	struct fault *fault = container_of(work, struct fault, work);
+	int npages, write;
+	struct page *page;
+
+	write = !!(fault->flags & PPR_FAULT_WRITE);
+
+	npages = get_user_pages(fault->state->task, fault->state->mm,
+				fault->address, 1, write, 0, &page, NULL);
+
+	if (npages == 1) {
+		put_page(page);
+	} else if (fault->dev_state->inv_ppr_cb) {
+		int status;
+
+		status = fault->dev_state->inv_ppr_cb(fault->dev_state->pdev,
+						      fault->pasid,
+						      fault->address,
+						      fault->flags);
+		switch (status) {
+		case AMD_IOMMU_INV_PRI_RSP_SUCCESS:
+			set_pri_tag_status(fault->state, fault->tag, PPR_SUCCESS);
+			break;
+		case AMD_IOMMU_INV_PRI_RSP_INVALID:
+			set_pri_tag_status(fault->state, fault->tag, PPR_INVALID);
+			break;
+		case AMD_IOMMU_INV_PRI_RSP_FAIL:
+			set_pri_tag_status(fault->state, fault->tag, PPR_FAILURE);
+			break;
+		default:
+			BUG();
+		}
+	} else {
+		set_pri_tag_status(fault->state, fault->tag, PPR_INVALID);
+	}
+
+	finish_pri_tag(fault->dev_state, fault->state, fault->tag);
+
+	put_pasid_state(fault->state);
+
+	kfree(fault);
+}
+
+static int ppr_notifier(struct notifier_block *nb, unsigned long e, void *data)
+{
+	struct amd_iommu_fault *iommu_fault;
+	struct pasid_state *pasid_state;
+	struct device_state *dev_state;
+	unsigned long flags;
+	struct fault *fault;
+	bool finish;
+	u16 tag;
+	int ret;
+
+	iommu_fault = data;
+	tag         = iommu_fault->tag & 0x1ff;
+	finish      = (iommu_fault->tag >> 9) & 1;
+
+	ret = NOTIFY_DONE;
+	dev_state = get_device_state(iommu_fault->device_id);
+	if (dev_state == NULL)
+		goto out;
+
+	pasid_state = get_pasid_state(dev_state, iommu_fault->pasid);
+	if (pasid_state == NULL) {
+		/* We know the device but not the PASID -> send INVALID */
+		amd_iommu_complete_ppr(dev_state->pdev, iommu_fault->pasid,
+				       PPR_INVALID, tag);
+		goto out_drop_state;
+	}
+
+	spin_lock_irqsave(&pasid_state->lock, flags);
+	atomic_inc(&pasid_state->pri[tag].inflight);
+	if (finish)
+		pasid_state->pri[tag].finish = true;
+	spin_unlock_irqrestore(&pasid_state->lock, flags);
+
+	fault = kzalloc(sizeof(*fault), GFP_ATOMIC);
+	if (fault == NULL) {
+		/* We are OOM - send success and let the device re-fault */
+		finish_pri_tag(dev_state, pasid_state, tag);
+		goto out_drop_state;
+	}
+
+	fault->dev_state = dev_state;
+	fault->address   = iommu_fault->address;
+	fault->state     = pasid_state;
+	fault->tag       = tag;
+	fault->finish    = finish;
+	fault->flags     = iommu_fault->flags;
+	INIT_WORK(&fault->work, do_fault);
+
+	queue_work(iommu_wq, &fault->work);
+
+	ret = NOTIFY_OK;
+
+out_drop_state:
+	put_device_state(dev_state);
+
+out:
+	return ret;
+}
+
+static struct notifier_block ppr_nb = {
+	.notifier_call = ppr_notifier,
+};
+
+static int task_exit(struct notifier_block *nb, unsigned long e, void *data)
+{
+	struct pasid_state *pasid_state;
+	struct task_struct *task;
+
+	task = data;
+
+	/*
+	 * Using this notifier is a hack - but there is no other choice
+	 * at the moment. What I really want is a sleeping notifier that
+	 * is called when an MM goes down. But such a notifier doesn't
+	 * exist yet. The notifier needs to sleep because it has to make
+	 * sure that the device does not use the PASID and the address
+	 * space anymore before it is destroyed. This includes waiting
+	 * for pending PRI requests to pass the workqueue. The
+	 * MMU-Notifiers would be a good fit, but they use RCU and so
+	 * they are not allowed to sleep. Lets see how we can solve this
+	 * in a more intelligent way in the future.
+	 */
+again:
+	spin_lock(&ps_lock);
+	list_for_each_entry(pasid_state, &pasid_state_list, list) {
+		struct device_state *dev_state;
+		int pasid;
+
+		if (pasid_state->task != task)
+			continue;
+
+		/* Drop Lock and unbind */
+		spin_unlock(&ps_lock);
+
+		dev_state = pasid_state->device_state;
+		pasid     = pasid_state->pasid;
+
+		if (pasid_state->device_state->inv_ctx_cb)
+			dev_state->inv_ctx_cb(dev_state->pdev, pasid);
+
+		unbind_pasid(dev_state, pasid);
+
+		/* Task may be in the list multiple times */
+		goto again;
+	}
+	spin_unlock(&ps_lock);
+
+	return NOTIFY_OK;
+}
+
+int amd_iommu_bind_pasid(struct pci_dev *pdev, int pasid,
+			 struct task_struct *task)
+{
+	struct pasid_state *pasid_state;
+	struct device_state *dev_state;
+	u16 devid;
+	int ret;
+
+	might_sleep();
+
+	if (!amd_iommu_v2_supported())
+		return -ENODEV;
+
+	devid     = device_id(pdev);
+	dev_state = get_device_state(devid);
+
+	if (dev_state == NULL)
+		return -EINVAL;
+
+	ret = -EINVAL;
+	if (pasid < 0 || pasid >= dev_state->max_pasids)
+		goto out;
+
+	ret = -ENOMEM;
+	pasid_state = kzalloc(sizeof(*pasid_state), GFP_KERNEL);
+	if (pasid_state == NULL)
+		goto out;
+
+	atomic_set(&pasid_state->count, 1);
+	init_waitqueue_head(&pasid_state->wq);
+	pasid_state->task         = task;
+	pasid_state->mm           = get_task_mm(task);
+	pasid_state->device_state = dev_state;
+	pasid_state->pasid        = pasid;
+	pasid_state->mn.ops       = &iommu_mn;
+
+	if (pasid_state->mm == NULL)
+		goto out_free;
+
+	mmu_notifier_register(&pasid_state->mn, pasid_state->mm);
+
+	ret = set_pasid_state(dev_state, pasid_state, pasid);
+	if (ret)
+		goto out_unregister;
+
+	ret = amd_iommu_domain_set_gcr3(dev_state->domain, pasid,
+					__pa(pasid_state->mm->pgd));
+	if (ret)
+		goto out_clear_state;
+
+	link_pasid_state(pasid_state);
+
+	return 0;
+
+out_clear_state:
+	clear_pasid_state(dev_state, pasid);
+
+out_unregister:
+	mmu_notifier_unregister(&pasid_state->mn, pasid_state->mm);
+
+out_free:
+	free_pasid_state(pasid_state);
+
+out:
+	put_device_state(dev_state);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_bind_pasid);
+
+void amd_iommu_unbind_pasid(struct pci_dev *pdev, int pasid)
+{
+	struct device_state *dev_state;
+	u16 devid;
+
+	might_sleep();
+
+	if (!amd_iommu_v2_supported())
+		return;
+
+	devid = device_id(pdev);
+	dev_state = get_device_state(devid);
+	if (dev_state == NULL)
+		return;
+
+	if (pasid < 0 || pasid >= dev_state->max_pasids)
+		goto out;
+
+	unbind_pasid(dev_state, pasid);
+
+out:
+	put_device_state(dev_state);
+}
+EXPORT_SYMBOL(amd_iommu_unbind_pasid);
+
+int amd_iommu_init_device(struct pci_dev *pdev, int pasids)
+{
+	struct device_state *dev_state;
+	unsigned long flags;
+	int ret, tmp;
+	u16 devid;
+
+	might_sleep();
+
+	if (!amd_iommu_v2_supported())
+		return -ENODEV;
+
+	if (pasids <= 0 || pasids > (PASID_MASK + 1))
+		return -EINVAL;
+
+	devid = device_id(pdev);
+
+	dev_state = kzalloc(sizeof(*dev_state), GFP_KERNEL);
+	if (dev_state == NULL)
+		return -ENOMEM;
+
+	spin_lock_init(&dev_state->lock);
+	init_waitqueue_head(&dev_state->wq);
+	dev_state->pdev = pdev;
+
+	tmp = pasids;
+	for (dev_state->pasid_levels = 0; (tmp - 1) & ~0x1ff; tmp >>= 9)
+		dev_state->pasid_levels += 1;
+
+	atomic_set(&dev_state->count, 1);
+	dev_state->max_pasids = pasids;
+
+	ret = -ENOMEM;
+	dev_state->states = (void *)get_zeroed_page(GFP_KERNEL);
+	if (dev_state->states == NULL)
+		goto out_free_dev_state;
+
+	dev_state->domain = iommu_domain_alloc(&pci_bus_type);
+	if (dev_state->domain == NULL)
+		goto out_free_states;
+
+	amd_iommu_domain_direct_map(dev_state->domain);
+
+	ret = amd_iommu_domain_enable_v2(dev_state->domain, pasids);
+	if (ret)
+		goto out_free_domain;
+
+	ret = iommu_attach_device(dev_state->domain, &pdev->dev);
+	if (ret != 0)
+		goto out_free_domain;
+
+	spin_lock_irqsave(&state_lock, flags);
+
+	if (state_table[devid] != NULL) {
+		spin_unlock_irqrestore(&state_lock, flags);
+		ret = -EBUSY;
+		goto out_free_domain;
+	}
+
+	state_table[devid] = dev_state;
+
+	spin_unlock_irqrestore(&state_lock, flags);
+
+	return 0;
+
+out_free_domain:
+	iommu_domain_free(dev_state->domain);
+
+out_free_states:
+	free_page((unsigned long)dev_state->states);
+
+out_free_dev_state:
+	kfree(dev_state);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_init_device);
+
+void amd_iommu_free_device(struct pci_dev *pdev)
+{
+	struct device_state *dev_state;
+	unsigned long flags;
+	u16 devid;
+
+	if (!amd_iommu_v2_supported())
+		return;
+
+	devid = device_id(pdev);
+
+	spin_lock_irqsave(&state_lock, flags);
+
+	dev_state = state_table[devid];
+	if (dev_state == NULL) {
+		spin_unlock_irqrestore(&state_lock, flags);
+		return;
+	}
+
+	state_table[devid] = NULL;
+
+	spin_unlock_irqrestore(&state_lock, flags);
+
+	/* Get rid of any remaining pasid states */
+	free_pasid_states(dev_state);
+
+	put_device_state_wait(dev_state);
+}
+EXPORT_SYMBOL(amd_iommu_free_device);
+
+int amd_iommu_set_invalid_ppr_cb(struct pci_dev *pdev,
+				 amd_iommu_invalid_ppr_cb cb)
+{
+	struct device_state *dev_state;
+	unsigned long flags;
+	u16 devid;
+	int ret;
+
+	if (!amd_iommu_v2_supported())
+		return -ENODEV;
+
+	devid = device_id(pdev);
+
+	spin_lock_irqsave(&state_lock, flags);
+
+	ret = -EINVAL;
+	dev_state = state_table[devid];
+	if (dev_state == NULL)
+		goto out_unlock;
+
+	dev_state->inv_ppr_cb = cb;
+
+	ret = 0;
+
+out_unlock:
+	spin_unlock_irqrestore(&state_lock, flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_set_invalid_ppr_cb);
+
+int amd_iommu_set_invalidate_ctx_cb(struct pci_dev *pdev,
+				    amd_iommu_invalidate_ctx cb)
+{
+	struct device_state *dev_state;
+	unsigned long flags;
+	u16 devid;
+	int ret;
+
+	if (!amd_iommu_v2_supported())
+		return -ENODEV;
+
+	devid = device_id(pdev);
+
+	spin_lock_irqsave(&state_lock, flags);
+
+	ret = -EINVAL;
+	dev_state = state_table[devid];
+	if (dev_state == NULL)
+		goto out_unlock;
+
+	dev_state->inv_ctx_cb = cb;
+
+	ret = 0;
+
+out_unlock:
+	spin_unlock_irqrestore(&state_lock, flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(amd_iommu_set_invalidate_ctx_cb);
+
+static int __init amd_iommu_v2_init(void)
+{
+	size_t state_table_size;
+	int ret;
+
+	pr_info("AMD IOMMUv2 driver by Joerg Roedel <joerg.roedel@amd.com>");
+
+	spin_lock_init(&state_lock);
+
+	state_table_size = MAX_DEVICES * sizeof(struct device_state *);
+	state_table = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
+					       get_order(state_table_size));
+	if (state_table == NULL)
+		return -ENOMEM;
+
+	ret = -ENOMEM;
+	iommu_wq = create_workqueue("amd_iommu_v2");
+	if (iommu_wq == NULL)
+		goto out_free;
+
+	ret = -ENOMEM;
+	empty_page_table = (u64 *)get_zeroed_page(GFP_KERNEL);
+	if (empty_page_table == NULL)
+		goto out_destroy_wq;
+
+	amd_iommu_register_ppr_notifier(&ppr_nb);
+	profile_event_register(PROFILE_TASK_EXIT, &profile_nb);
+
+	return 0;
+
+out_destroy_wq:
+	destroy_workqueue(iommu_wq);
+
+out_free:
+	free_pages((unsigned long)state_table, get_order(state_table_size));
+
+	return ret;
+}
+
+static void __exit amd_iommu_v2_exit(void)
+{
+	struct device_state *dev_state;
+	size_t state_table_size;
+	int i;
+
+	profile_event_unregister(PROFILE_TASK_EXIT, &profile_nb);
+	amd_iommu_unregister_ppr_notifier(&ppr_nb);
+
+	flush_workqueue(iommu_wq);
+
+	/*
+	 * The loop below might call flush_workqueue(), so call
+	 * destroy_workqueue() after it
+	 */
+	for (i = 0; i < MAX_DEVICES; ++i) {
+		dev_state = get_device_state(i);
+
+		if (dev_state == NULL)
+			continue;
+
+		WARN_ON_ONCE(1);
+
+		put_device_state(dev_state);
+		amd_iommu_free_device(dev_state->pdev);
+	}
+
+	destroy_workqueue(iommu_wq);
+
+	state_table_size = MAX_DEVICES * sizeof(struct device_state *);
+	free_pages((unsigned long)state_table, get_order(state_table_size));
+
+	free_page((unsigned long)empty_page_table);
+}
+
+module_init(amd_iommu_v2_init);
+module_exit(amd_iommu_v2_exit);

drivers/iommu/intel-iommu.c

@@ -79,6 +79,24 @@
 #define LEVEL_STRIDE		(9)
 #define LEVEL_MASK		(((u64)1 << LEVEL_STRIDE) - 1)
 
+/*
+ * This bitmap is used to advertise the page sizes our hardware support
+ * to the IOMMU core, which will then use this information to split
+ * physically contiguous memory regions it is mapping into page sizes
+ * that we support.
+ *
+ * Traditionally the IOMMU core just handed us the mappings directly,
+ * after making sure the size is an order of a 4KiB page and that the
+ * mapping has natural alignment.
+ *
+ * To retain this behavior, we currently advertise that we support
+ * all page sizes that are an order of 4KiB.
+ *
+ * If at some point we'd like to utilize the IOMMU core's new behavior,
+ * we could change this to advertise the real page sizes we support.
+ */
+#define INTEL_IOMMU_PGSIZES	(~0xFFFUL)
+
 static inline int agaw_to_level(int agaw)
 {
 	return agaw + 2;
@@ -3979,12 +3997,11 @@ static void intel_iommu_detach_device(struct iommu_domain *domain,
 
 static int intel_iommu_map(struct iommu_domain *domain,
 			   unsigned long iova, phys_addr_t hpa,
-			   int gfp_order, int iommu_prot)
+			   size_t size, int iommu_prot)
 {
 	struct dmar_domain *dmar_domain = domain->priv;
 	u64 max_addr;
 	int prot = 0;
-	size_t size;
 	int ret;
 
 	if (iommu_prot & IOMMU_READ)
@@ -3994,7 +4011,6 @@ static int intel_iommu_map(struct iommu_domain *domain,
 	if ((iommu_prot & IOMMU_CACHE) && dmar_domain->iommu_snooping)
 		prot |= DMA_PTE_SNP;
 
-	size     = PAGE_SIZE << gfp_order;
 	max_addr = iova + size;
 	if (dmar_domain->max_addr < max_addr) {
 		u64 end;
@@ -4017,11 +4033,10 @@ static int intel_iommu_map(struct iommu_domain *domain,
 	return ret;
 }
 
-static int intel_iommu_unmap(struct iommu_domain *domain,
-			     unsigned long iova, int gfp_order)
+static size_t intel_iommu_unmap(struct iommu_domain *domain,
+			     unsigned long iova, size_t size)
 {
 	struct dmar_domain *dmar_domain = domain->priv;
-	size_t size = PAGE_SIZE << gfp_order;
 	int order;
 
 	order = dma_pte_clear_range(dmar_domain, iova >> VTD_PAGE_SHIFT,
@@ -4030,7 +4045,7 @@ static int intel_iommu_unmap(struct iommu_domain *domain,
 	if (dmar_domain->max_addr == iova + size)
 		dmar_domain->max_addr = iova;
 
-	return order;
+	return PAGE_SIZE << order;
 }
 
 static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
@@ -4060,6 +4075,54 @@ static int intel_iommu_domain_has_cap(struct iommu_domain *domain,
 	return 0;
 }
 
+/*
+ * Group numbers are arbitrary.  Device with the same group number
+ * indicate the iommu cannot differentiate between them.  To avoid
+ * tracking used groups we just use the seg|bus|devfn of the lowest
+ * level we're able to differentiate devices
+ */
+static int intel_iommu_device_group(struct device *dev, unsigned int *groupid)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct pci_dev *bridge;
+	union {
+		struct {
+			u8 devfn;
+			u8 bus;
+			u16 segment;
+		} pci;
+		u32 group;
+	} id;
+
+	if (iommu_no_mapping(dev))
+		return -ENODEV;
+
+	id.pci.segment = pci_domain_nr(pdev->bus);
+	id.pci.bus = pdev->bus->number;
+	id.pci.devfn = pdev->devfn;
+
+	if (!device_to_iommu(id.pci.segment, id.pci.bus, id.pci.devfn))
+		return -ENODEV;
+
+	bridge = pci_find_upstream_pcie_bridge(pdev);
+	if (bridge) {
+		if (pci_is_pcie(bridge)) {
+			id.pci.bus = bridge->subordinate->number;
+			id.pci.devfn = 0;
+		} else {
+			id.pci.bus = bridge->bus->number;
+			id.pci.devfn = bridge->devfn;
+		}
+	}
+
+	if (!pdev->is_virtfn && iommu_group_mf)
+		id.pci.devfn = PCI_DEVFN(PCI_SLOT(id.pci.devfn), 0);
+
+	*groupid = id.group;
+
+	return 0;
+}
+
 static struct iommu_ops intel_iommu_ops = {
 	.domain_init	= intel_iommu_domain_init,
 	.domain_destroy = intel_iommu_domain_destroy,
@@ -4069,6 +4132,8 @@ static struct iommu_ops intel_iommu_ops = {
 	.unmap		= intel_iommu_unmap,
 	.iova_to_phys	= intel_iommu_iova_to_phys,
 	.domain_has_cap = intel_iommu_domain_has_cap,
+	.device_group	= intel_iommu_device_group,
+	.pgsize_bitmap	= INTEL_IOMMU_PGSIZES,
 };
 
 static void __devinit quirk_iommu_rwbf(struct pci_dev *dev)

drivers/iommu/iommu.c

@@ -16,6 +16,8 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  */
 
+#define pr_fmt(fmt)    "%s: " fmt, __func__
+
 #include <linux/device.h>
 #include <linux/kernel.h>
 #include <linux/bug.h>
@@ -25,8 +27,59 @@
 #include <linux/errno.h>
 #include <linux/iommu.h>
 
+static ssize_t show_iommu_group(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	unsigned int groupid;
+
+	if (iommu_device_group(dev, &groupid))
+		return 0;
+
+	return sprintf(buf, "%u", groupid);
+}
+static DEVICE_ATTR(iommu_group, S_IRUGO, show_iommu_group, NULL);
+
+static int add_iommu_group(struct device *dev, void *data)
+{
+	unsigned int groupid;
+
+	if (iommu_device_group(dev, &groupid) == 0)
+		return device_create_file(dev, &dev_attr_iommu_group);
+
+	return 0;
+}
+
+static int remove_iommu_group(struct device *dev)
+{
+	unsigned int groupid;
+
+	if (iommu_device_group(dev, &groupid) == 0)
+		device_remove_file(dev, &dev_attr_iommu_group);
+
+	return 0;
+}
+
+static int iommu_device_notifier(struct notifier_block *nb,
+				 unsigned long action, void *data)
+{
+	struct device *dev = data;
+
+	if (action == BUS_NOTIFY_ADD_DEVICE)
+		return add_iommu_group(dev, NULL);
+	else if (action == BUS_NOTIFY_DEL_DEVICE)
+		return remove_iommu_group(dev);
+
+	return 0;
+}
+
+static struct notifier_block iommu_device_nb = {
+	.notifier_call = iommu_device_notifier,
+};
+
 static void iommu_bus_init(struct bus_type *bus, struct iommu_ops *ops)
 {
+	bus_register_notifier(bus, &iommu_device_nb);
+	bus_for_each_dev(bus, NULL, NULL, add_iommu_group);
 }
 
 /**
@@ -157,32 +210,134 @@ int iommu_domain_has_cap(struct iommu_domain *domain,
 EXPORT_SYMBOL_GPL(iommu_domain_has_cap);
 
 int iommu_map(struct iommu_domain *domain, unsigned long iova,
-	      phys_addr_t paddr, int gfp_order, int prot)
+	      phys_addr_t paddr, size_t size, int prot)
 {
-	size_t size;
+	unsigned long orig_iova = iova;
+	unsigned int min_pagesz;
+	size_t orig_size = size;
+	int ret = 0;
 
 	if (unlikely(domain->ops->map == NULL))
 		return -ENODEV;
 
-	size         = PAGE_SIZE << gfp_order;
+	/* find out the minimum page size supported */
+	min_pagesz = 1 << __ffs(domain->ops->pgsize_bitmap);
+
+	/*
+	 * both the virtual address and the physical one, as well as
+	 * the size of the mapping, must be aligned (at least) to the
+	 * size of the smallest page supported by the hardware
+	 */
+	if (!IS_ALIGNED(iova | paddr | size, min_pagesz)) {
+		pr_err("unaligned: iova 0x%lx pa 0x%lx size 0x%lx min_pagesz "
+			"0x%x\n", iova, (unsigned long)paddr,
+			(unsigned long)size, min_pagesz);
+		return -EINVAL;
+	}
+
+	pr_debug("map: iova 0x%lx pa 0x%lx size 0x%lx\n", iova,
+				(unsigned long)paddr, (unsigned long)size);
+
+	while (size) {
+		unsigned long pgsize, addr_merge = iova | paddr;
+		unsigned int pgsize_idx;
+
+		/* Max page size that still fits into 'size' */
+		pgsize_idx = __fls(size);
+
+		/* need to consider alignment requirements ? */
+		if (likely(addr_merge)) {
+			/* Max page size allowed by both iova and paddr */
+			unsigned int align_pgsize_idx = __ffs(addr_merge);
+
+			pgsize_idx = min(pgsize_idx, align_pgsize_idx);
+		}
+
+		/* build a mask of acceptable page sizes */
+		pgsize = (1UL << (pgsize_idx + 1)) - 1;
 
-	BUG_ON(!IS_ALIGNED(iova | paddr, size));
+		/* throw away page sizes not supported by the hardware */
+		pgsize &= domain->ops->pgsize_bitmap;
 
-	return domain->ops->map(domain, iova, paddr, gfp_order, prot);
+		/* make sure we're still sane */
+		BUG_ON(!pgsize);
+
+		/* pick the biggest page */
+		pgsize_idx = __fls(pgsize);
+		pgsize = 1UL << pgsize_idx;
+
+		pr_debug("mapping: iova 0x%lx pa 0x%lx pgsize %lu\n", iova,
+					(unsigned long)paddr, pgsize);
+
+		ret = domain->ops->map(domain, iova, paddr, pgsize, prot);
+		if (ret)
+			break;
+
+		iova += pgsize;
+		paddr += pgsize;
+		size -= pgsize;
+	}
+
+	/* unroll mapping in case something went wrong */
+	if (ret)
+		iommu_unmap(domain, orig_iova, orig_size - size);
+
+	return ret;
 }
 EXPORT_SYMBOL_GPL(iommu_map);
 
-int iommu_unmap(struct iommu_domain *domain, unsigned long iova, int gfp_order)
+size_t iommu_unmap(struct iommu_domain *domain, unsigned long iova, size_t size)
 {
-	size_t size;
+	size_t unmapped_page, unmapped = 0;
+	unsigned int min_pagesz;
 
 	if (unlikely(domain->ops->unmap == NULL))
 		return -ENODEV;
 
-	size         = PAGE_SIZE << gfp_order;
+	/* find out the minimum page size supported */
+	min_pagesz = 1 << __ffs(domain->ops->pgsize_bitmap);
+
+	/*
+	 * The virtual address, as well as the size of the mapping, must be
+	 * aligned (at least) to the size of the smallest page supported
+	 * by the hardware
+	 */
+	if (!IS_ALIGNED(iova | size, min_pagesz)) {
+		pr_err("unaligned: iova 0x%lx size 0x%lx min_pagesz 0x%x\n",
+					iova, (unsigned long)size, min_pagesz);
+		return -EINVAL;
+	}
+
+	pr_debug("unmap this: iova 0x%lx size 0x%lx\n", iova,
+							(unsigned long)size);
+
+	/*
+	 * Keep iterating until we either unmap 'size' bytes (or more)
+	 * or we hit an area that isn't mapped.
+	 */
+	while (unmapped < size) {
+		size_t left = size - unmapped;
+
+		unmapped_page = domain->ops->unmap(domain, iova, left);
+		if (!unmapped_page)
+			break;
+
+		pr_debug("unmapped: iova 0x%lx size %lx\n", iova,
+					(unsigned long)unmapped_page);
+
+		iova += unmapped_page;
+		unmapped += unmapped_page;
+	}
+
+	return unmapped;
+}
+EXPORT_SYMBOL_GPL(iommu_unmap);
 
-	BUG_ON(!IS_ALIGNED(iova, size));
+int iommu_device_group(struct device *dev, unsigned int *groupid)
+{
+	if (iommu_present(dev->bus) && dev->bus->iommu_ops->device_group)
+		return dev->bus->iommu_ops->device_group(dev, groupid);
 
-	return domain->ops->unmap(domain, iova, gfp_order);
+	return -ENODEV;
 }
-EXPORT_SYMBOL_GPL(iommu_unmap);
+EXPORT_SYMBOL_GPL(iommu_device_group);

drivers/iommu/msm_iommu.c

@@ -42,6 +42,9 @@ __asm__ __volatile__ (							\
 #define RCP15_PRRR(reg)		MRC(reg, p15, 0, c10, c2, 0)
 #define RCP15_NMRR(reg)		MRC(reg, p15, 0, c10, c2, 1)
 
+/* bitmap of the page sizes currently supported */
+#define MSM_IOMMU_PGSIZES	(SZ_4K | SZ_64K | SZ_1M | SZ_16M)
+
 static int msm_iommu_tex_class[4];
 
 DEFINE_SPINLOCK(msm_iommu_lock);
@@ -352,7 +355,7 @@ fail:
 }
 
 static int msm_iommu_map(struct iommu_domain *domain, unsigned long va,
-			 phys_addr_t pa, int order, int prot)
+			 phys_addr_t pa, size_t len, int prot)
 {
 	struct msm_priv *priv;
 	unsigned long flags;
@@ -363,7 +366,6 @@ static int msm_iommu_map(struct iommu_domain *domain, unsigned long va,
 	unsigned long *sl_pte;
 	unsigned long sl_offset;
 	unsigned int pgprot;
-	size_t len = 0x1000UL << order;
 	int ret = 0, tex, sh;
 
 	spin_lock_irqsave(&msm_iommu_lock, flags);
@@ -463,8 +465,8 @@ fail:
 	return ret;
 }
 
-static int msm_iommu_unmap(struct iommu_domain *domain, unsigned long va,
-			    int order)
+static size_t msm_iommu_unmap(struct iommu_domain *domain, unsigned long va,
+			    size_t len)
 {
 	struct msm_priv *priv;
 	unsigned long flags;
@@ -474,7 +476,6 @@ static int msm_iommu_unmap(struct iommu_domain *domain, unsigned long va,
 	unsigned long *sl_table;
 	unsigned long *sl_pte;
 	unsigned long sl_offset;
-	size_t len = 0x1000UL << order;
 	int i, ret = 0;
 
 	spin_lock_irqsave(&msm_iommu_lock, flags);
@@ -544,15 +545,12 @@ static int msm_iommu_unmap(struct iommu_domain *domain, unsigned long va,
 
 	ret = __flush_iotlb(domain);
 
-	/*
-	 * the IOMMU API requires us to return the order of the unmapped
-	 * page (on success).
-	 */
-	if (!ret)
-		ret = order;
 fail:
 	spin_unlock_irqrestore(&msm_iommu_lock, flags);
-	return ret;
+
+	/* the IOMMU API requires us to return how many bytes were unmapped */
+	len = ret ? 0 : len;
+	return len;
 }
 
 static phys_addr_t msm_iommu_iova_to_phys(struct iommu_domain *domain,
@@ -684,7 +682,8 @@ static struct iommu_ops msm_iommu_ops = {
 	.map = msm_iommu_map,
 	.unmap = msm_iommu_unmap,
 	.iova_to_phys = msm_iommu_iova_to_phys,
-	.domain_has_cap = msm_iommu_domain_has_cap
+	.domain_has_cap = msm_iommu_domain_has_cap,
+	.pgsize_bitmap = MSM_IOMMU_PGSIZES,
 };
 
 static int __init get_tex_class(int icp, int ocp, int mt, int nos)

drivers/iommu/omap-iommu.c

@@ -33,6 +33,9 @@
 	     (__i < (n)) && (cr = __iotlb_read_cr((obj), __i), true);	\
 	     __i++)
 
+/* bitmap of the page sizes currently supported */
+#define OMAP_IOMMU_PGSIZES	(SZ_4K | SZ_64K | SZ_1M | SZ_16M)
+
 /**
  * struct omap_iommu_domain - omap iommu domain
  * @pgtable:	the page table
@@ -86,20 +89,24 @@ EXPORT_SYMBOL_GPL(omap_uninstall_iommu_arch);
 
 /**
  * omap_iommu_save_ctx - Save registers for pm off-mode support
- * @obj:	target iommu
+ * @dev:	client device
  **/
-void omap_iommu_save_ctx(struct omap_iommu *obj)
+void omap_iommu_save_ctx(struct device *dev)
 {
+	struct omap_iommu *obj = dev_to_omap_iommu(dev);
+
 	arch_iommu->save_ctx(obj);
 }
 EXPORT_SYMBOL_GPL(omap_iommu_save_ctx);
 
 /**
  * omap_iommu_restore_ctx - Restore registers for pm off-mode support
- * @obj:	target iommu
+ * @dev:	client device
  **/
-void omap_iommu_restore_ctx(struct omap_iommu *obj)
+void omap_iommu_restore_ctx(struct device *dev)
 {
+	struct omap_iommu *obj = dev_to_omap_iommu(dev);
+
 	arch_iommu->restore_ctx(obj);
 }
 EXPORT_SYMBOL_GPL(omap_iommu_restore_ctx);
@@ -819,36 +826,24 @@ static int device_match_by_alias(struct device *dev, void *data)
 	return strcmp(obj->name, name) == 0;
 }
 
-/**
- * omap_find_iommu_device() - find an omap iommu device by name
- * @name:	name of the iommu device
- *
- * The generic iommu API requires the caller to provide the device
- * he wishes to attach to a certain iommu domain.
- *
- * Drivers generally should not bother with this as it should just
- * be taken care of by the DMA-API using dev_archdata.
- *
- * This function is provided as an interim solution until the latter
- * materializes, and omap3isp is fully migrated to the DMA-API.
- */
-struct device *omap_find_iommu_device(const char *name)
-{
-	return driver_find_device(&omap_iommu_driver.driver, NULL,
-				(void *)name,
-				device_match_by_alias);
-}
-EXPORT_SYMBOL_GPL(omap_find_iommu_device);
-
 /**
  * omap_iommu_attach() - attach iommu device to an iommu domain
- * @dev:	target omap iommu device
+ * @name:	name of target omap iommu device
  * @iopgd:	page table
  **/
-static struct omap_iommu *omap_iommu_attach(struct device *dev, u32 *iopgd)
+static struct omap_iommu *omap_iommu_attach(const char *name, u32 *iopgd)
 {
 	int err = -ENOMEM;
-	struct omap_iommu *obj = to_iommu(dev);
+	struct device *dev;
+	struct omap_iommu *obj;
+
+	dev = driver_find_device(&omap_iommu_driver.driver, NULL,
+				(void *)name,
+				device_match_by_alias);
+	if (!dev)
+		return NULL;
+
+	obj = to_iommu(dev);
 
 	spin_lock(&obj->iommu_lock);
 
@@ -1019,12 +1014,11 @@ static void iopte_cachep_ctor(void *iopte)
 }
 
 static int omap_iommu_map(struct iommu_domain *domain, unsigned long da,
-			 phys_addr_t pa, int order, int prot)
+			 phys_addr_t pa, size_t bytes, int prot)
 {
 	struct omap_iommu_domain *omap_domain = domain->priv;
 	struct omap_iommu *oiommu = omap_domain->iommu_dev;
 	struct device *dev = oiommu->dev;
-	size_t bytes = PAGE_SIZE << order;
 	struct iotlb_entry e;
 	int omap_pgsz;
 	u32 ret, flags;
@@ -1049,19 +1043,16 @@ static int omap_iommu_map(struct iommu_domain *domain, unsigned long da,
 	return ret;
 }
 
-static int omap_iommu_unmap(struct iommu_domain *domain, unsigned long da,
-			    int order)
+static size_t omap_iommu_unmap(struct iommu_domain *domain, unsigned long da,
+			    size_t size)
 {
 	struct omap_iommu_domain *omap_domain = domain->priv;
 	struct omap_iommu *oiommu = omap_domain->iommu_dev;
 	struct device *dev = oiommu->dev;
-	size_t unmap_size;
-
-	dev_dbg(dev, "unmapping da 0x%lx order %d\n", da, order);
 
-	unmap_size = iopgtable_clear_entry(oiommu, da);
+	dev_dbg(dev, "unmapping da 0x%lx size %u\n", da, size);
 
-	return unmap_size ? get_order(unmap_size) : -EINVAL;
+	return iopgtable_clear_entry(oiommu, da);
 }
 
 static int
@@ -1069,6 +1060,7 @@ omap_iommu_attach_dev(struct iommu_domain *domain, struct device *dev)
 {
 	struct omap_iommu_domain *omap_domain = domain->priv;
 	struct omap_iommu *oiommu;
+	struct omap_iommu_arch_data *arch_data = dev->archdata.iommu;
 	int ret = 0;
 
 	spin_lock(&omap_domain->lock);
@@ -1081,14 +1073,14 @@ omap_iommu_attach_dev(struct iommu_domain *domain, struct device *dev)
 	}
 
 	/* get a handle to and enable the omap iommu */
-	oiommu = omap_iommu_attach(dev, omap_domain->pgtable);
+	oiommu = omap_iommu_attach(arch_data->name, omap_domain->pgtable);
 	if (IS_ERR(oiommu)) {
 		ret = PTR_ERR(oiommu);
 		dev_err(dev, "can't get omap iommu: %d\n", ret);
 		goto out;
 	}
 
-	omap_domain->iommu_dev = oiommu;
+	omap_domain->iommu_dev = arch_data->iommu_dev = oiommu;
 	oiommu->domain = domain;
 
 out:
@@ -1100,7 +1092,8 @@ static void omap_iommu_detach_dev(struct iommu_domain *domain,
 				 struct device *dev)
 {
 	struct omap_iommu_domain *omap_domain = domain->priv;
-	struct omap_iommu *oiommu = to_iommu(dev);
+	struct omap_iommu_arch_data *arch_data = dev->archdata.iommu;
+	struct omap_iommu *oiommu = dev_to_omap_iommu(dev);
 
 	spin_lock(&omap_domain->lock);
 
@@ -1114,7 +1107,7 @@ static void omap_iommu_detach_dev(struct iommu_domain *domain,
 
 	omap_iommu_detach(oiommu);
 
-	omap_domain->iommu_dev = NULL;
+	omap_domain->iommu_dev = arch_data->iommu_dev = NULL;
 
 out:
 	spin_unlock(&omap_domain->lock);
@@ -1183,14 +1176,14 @@ static phys_addr_t omap_iommu_iova_to_phys(struct iommu_domain *domain,
 		else if (iopte_is_large(*pte))
 			ret = omap_iommu_translate(*pte, da, IOLARGE_MASK);
 		else
-			dev_err(dev, "bogus pte 0x%x", *pte);
+			dev_err(dev, "bogus pte 0x%x, da 0x%lx", *pte, da);
 	} else {
 		if (iopgd_is_section(*pgd))
 			ret = omap_iommu_translate(*pgd, da, IOSECTION_MASK);
 		else if (iopgd_is_super(*pgd))
 			ret = omap_iommu_translate(*pgd, da, IOSUPER_MASK);
 		else
-			dev_err(dev, "bogus pgd 0x%x", *pgd);
+			dev_err(dev, "bogus pgd 0x%x, da 0x%lx", *pgd, da);
 	}
 
 	return ret;
@@ -1211,6 +1204,7 @@ static struct iommu_ops omap_iommu_ops = {
 	.unmap		= omap_iommu_unmap,
 	.iova_to_phys	= omap_iommu_iova_to_phys,
 	.domain_has_cap	= omap_iommu_domain_has_cap,
+	.pgsize_bitmap	= OMAP_IOMMU_PGSIZES,
 };
 
 static int __init omap_iommu_init(void)

drivers/iommu/omap-iovmm.c

@@ -231,12 +231,14 @@ static struct iovm_struct *__find_iovm_area(struct omap_iommu *obj,
 
 /**
  * omap_find_iovm_area  -  find iovma which includes @da
+ * @dev:	client device
  * @da:		iommu device virtual address
  *
  * Find the existing iovma starting at @da
  */
-struct iovm_struct *omap_find_iovm_area(struct omap_iommu *obj, u32 da)
+struct iovm_struct *omap_find_iovm_area(struct device *dev, u32 da)
 {
+	struct omap_iommu *obj = dev_to_omap_iommu(dev);
 	struct iovm_struct *area;
 
 	mutex_lock(&obj->mmap_lock);
@@ -343,14 +345,15 @@ static void free_iovm_area(struct omap_iommu *obj, struct iovm_struct *area)
 
 /**
  * omap_da_to_va - convert (d) to (v)
- * @obj:	objective iommu
+ * @dev:	client device
  * @da:		iommu device virtual address
  * @va:		mpu virtual address
  *
  * Returns mpu virtual addr which corresponds to a given device virtual addr
  */
-void *omap_da_to_va(struct omap_iommu *obj, u32 da)
+void *omap_da_to_va(struct device *dev, u32 da)
 {
+	struct omap_iommu *obj = dev_to_omap_iommu(dev);
 	void *va = NULL;
 	struct iovm_struct *area;
 
@@ -410,7 +413,6 @@ static int map_iovm_area(struct iommu_domain *domain, struct iovm_struct *new,
 	unsigned int i, j;
 	struct scatterlist *sg;
 	u32 da = new->da_start;
-	int order;
 
 	if (!domain || !sgt)
 		return -EINVAL;
@@ -429,12 +431,10 @@ static int map_iovm_area(struct iommu_domain *domain, struct iovm_struct *new,
 		if (bytes_to_iopgsz(bytes) < 0)
 			goto err_out;
 
-		order = get_order(bytes);
-
 		pr_debug("%s: [%d] %08x %08x(%x)\n", __func__,
 			 i, da, pa, bytes);
 
-		err = iommu_map(domain, da, pa, order, flags);
+		err = iommu_map(domain, da, pa, bytes, flags);
 		if (err)
 			goto err_out;
 
@@ -449,10 +449,9 @@ err_out:
 		size_t bytes;
 
 		bytes = sg->length + sg->offset;
-		order = get_order(bytes);
 
 		/* ignore failures.. we're already handling one */
-		iommu_unmap(domain, da, order);
+		iommu_unmap(domain, da, bytes);
 
 		da += bytes;
 	}
@@ -467,7 +466,8 @@ static void unmap_iovm_area(struct iommu_domain *domain, struct omap_iommu *obj,
 	size_t total = area->da_end - area->da_start;
 	const struct sg_table *sgt = area->sgt;
 	struct scatterlist *sg;
-	int i, err;
+	int i;
+	size_t unmapped;
 
 	BUG_ON(!sgtable_ok(sgt));
 	BUG_ON((!total) || !IS_ALIGNED(total, PAGE_SIZE));
@@ -475,13 +475,11 @@ static void unmap_iovm_area(struct iommu_domain *domain, struct omap_iommu *obj,
 	start = area->da_start;
 	for_each_sg(sgt->sgl, sg, sgt->nents, i) {
 		size_t bytes;
-		int order;
 
 		bytes = sg->length + sg->offset;
-		order = get_order(bytes);
 
-		err = iommu_unmap(domain, start, order);
-		if (err < 0)
+		unmapped = iommu_unmap(domain, start, bytes);
+		if (unmapped < bytes)
 			break;
 
 		dev_dbg(obj->dev, "%s: unmap %08x(%x) %08x\n",
@@ -582,16 +580,18 @@ __iommu_vmap(struct iommu_domain *domain, struct omap_iommu *obj,
 
 /**
  * omap_iommu_vmap  -  (d)-(p)-(v) address mapper
- * @obj:	objective iommu
+ * @domain:	iommu domain
+ * @dev:	client device
  * @sgt:	address of scatter gather table
  * @flags:	iovma and page property
  *
  * Creates 1-n-1 mapping with given @sgt and returns @da.
  * All @sgt element must be io page size aligned.
  */
-u32 omap_iommu_vmap(struct iommu_domain *domain, struct omap_iommu *obj, u32 da,
+u32 omap_iommu_vmap(struct iommu_domain *domain, struct device *dev, u32 da,
 		const struct sg_table *sgt, u32 flags)
 {
+	struct omap_iommu *obj = dev_to_omap_iommu(dev);
 	size_t bytes;
 	void *va = NULL;
 
@@ -622,15 +622,17 @@ EXPORT_SYMBOL_GPL(omap_iommu_vmap);
 
 /**
  * omap_iommu_vunmap  -  release virtual mapping obtained by 'omap_iommu_vmap()'
- * @obj:	objective iommu
+ * @domain:	iommu domain
+ * @dev:	client device
  * @da:		iommu device virtual address
  *
  * Free the iommu virtually contiguous memory area starting at
  * @da, which was returned by 'omap_iommu_vmap()'.
  */
 struct sg_table *
-omap_iommu_vunmap(struct iommu_domain *domain, struct omap_iommu *obj, u32 da)
+omap_iommu_vunmap(struct iommu_domain *domain, struct device *dev, u32 da)
 {
+	struct omap_iommu *obj = dev_to_omap_iommu(dev);
 	struct sg_table *sgt;
 	/*
 	 * 'sgt' is allocated before 'omap_iommu_vmalloc()' is called.
@@ -647,7 +649,7 @@ EXPORT_SYMBOL_GPL(omap_iommu_vunmap);
 
 /**
  * omap_iommu_vmalloc  -  (d)-(p)-(v) address allocator and mapper
- * @obj:	objective iommu
+ * @dev:	client device
  * @da:		contiguous iommu virtual memory
  * @bytes:	allocation size
  * @flags:	iovma and page property
@@ -656,9 +658,10 @@ EXPORT_SYMBOL_GPL(omap_iommu_vunmap);
  * @da again, which might be adjusted if 'IOVMF_DA_FIXED' is not set.
  */
 u32
-omap_iommu_vmalloc(struct iommu_domain *domain, struct omap_iommu *obj, u32 da,
+omap_iommu_vmalloc(struct iommu_domain *domain, struct device *dev, u32 da,
 						size_t bytes, u32 flags)
 {
+	struct omap_iommu *obj = dev_to_omap_iommu(dev);
 	void *va;
 	struct sg_table *sgt;
 
@@ -698,15 +701,16 @@ EXPORT_SYMBOL_GPL(omap_iommu_vmalloc);
 
 /**
  * omap_iommu_vfree  -  release memory allocated by 'omap_iommu_vmalloc()'
- * @obj:	objective iommu
+ * @dev:	client device
  * @da:		iommu device virtual address
  *
  * Frees the iommu virtually continuous memory area starting at
  * @da, as obtained from 'omap_iommu_vmalloc()'.
  */
-void omap_iommu_vfree(struct iommu_domain *domain, struct omap_iommu *obj,
+void omap_iommu_vfree(struct iommu_domain *domain, struct device *dev,
 								const u32 da)
 {
+	struct omap_iommu *obj = dev_to_omap_iommu(dev);
 	struct sg_table *sgt;
 
 	sgt = unmap_vm_area(domain, obj, da, vfree,

drivers/media/video/omap3isp/isp.c

@@ -80,13 +80,6 @@
 #include "isph3a.h"
 #include "isphist.h"
 
-/*
- * this is provided as an interim solution until omap3isp doesn't need
- * any omap-specific iommu API
- */
-#define to_iommu(dev)							\
-	(struct omap_iommu *)platform_get_drvdata(to_platform_device(dev))
-
 static unsigned int autoidle;
 module_param(autoidle, int, 0444);
 MODULE_PARM_DESC(autoidle, "Enable OMAP3ISP AUTOIDLE support");
@@ -1114,8 +1107,7 @@ isp_restore_context(struct isp_device *isp, struct isp_reg *reg_list)
 static void isp_save_ctx(struct isp_device *isp)
 {
 	isp_save_context(isp, isp_reg_list);
-	if (isp->iommu)
-		omap_iommu_save_ctx(isp->iommu);
+	omap_iommu_save_ctx(isp->dev);
 }
 
 /*
@@ -1128,8 +1120,7 @@ static void isp_save_ctx(struct isp_device *isp)
 static void isp_restore_ctx(struct isp_device *isp)
 {
 	isp_restore_context(isp, isp_reg_list);
-	if (isp->iommu)
-		omap_iommu_restore_ctx(isp->iommu);
+	omap_iommu_restore_ctx(isp->dev);
 	omap3isp_ccdc_restore_context(isp);
 	omap3isp_preview_restore_context(isp);
 }
@@ -1983,7 +1974,7 @@ static int isp_remove(struct platform_device *pdev)
 	isp_cleanup_modules(isp);
 
 	omap3isp_get(isp);
-	iommu_detach_device(isp->domain, isp->iommu_dev);
+	iommu_detach_device(isp->domain, &pdev->dev);
 	iommu_domain_free(isp->domain);
 	omap3isp_put(isp);
 
@@ -2131,17 +2122,6 @@ static int isp_probe(struct platform_device *pdev)
 		}
 	}
 
-	/* IOMMU */
-	isp->iommu_dev = omap_find_iommu_device("isp");
-	if (!isp->iommu_dev) {
-		dev_err(isp->dev, "omap_find_iommu_device failed\n");
-		ret = -ENODEV;
-		goto error_isp;
-	}
-
-	/* to be removed once iommu migration is complete */
-	isp->iommu = to_iommu(isp->iommu_dev);
-
 	isp->domain = iommu_domain_alloc(pdev->dev.bus);
 	if (!isp->domain) {
 		dev_err(isp->dev, "can't alloc iommu domain\n");
@@ -2149,7 +2129,7 @@ static int isp_probe(struct platform_device *pdev)
 		goto error_isp;
 	}
 
-	ret = iommu_attach_device(isp->domain, isp->iommu_dev);
+	ret = iommu_attach_device(isp->domain, &pdev->dev);
 	if (ret) {
 		dev_err(&pdev->dev, "can't attach iommu device: %d\n", ret);
 		goto free_domain;
@@ -2188,7 +2168,7 @@ error_modules:
 error_irq:
 	free_irq(isp->irq_num, isp);
 detach_dev:
-	iommu_detach_device(isp->domain, isp->iommu_dev);
+	iommu_detach_device(isp->domain, &pdev->dev);
 free_domain:
 	iommu_domain_free(isp->domain);
 error_isp:

drivers/media/video/omap3isp/isp.h

@@ -212,9 +212,7 @@ struct isp_device {
 	unsigned int sbl_resources;
 	unsigned int subclk_resources;
 
-	struct omap_iommu *iommu;
 	struct iommu_domain *domain;
-	struct device *iommu_dev;
 
 	struct isp_platform_callback platform_cb;
 };

drivers/media/video/omap3isp/ispccdc.c

@@ -366,7 +366,7 @@ static void ccdc_lsc_free_request(struct isp_ccdc_device *ccdc,
 		dma_unmap_sg(isp->dev, req->iovm->sgt->sgl,
 			     req->iovm->sgt->nents, DMA_TO_DEVICE);
 	if (req->table)
-		omap_iommu_vfree(isp->domain, isp->iommu, req->table);
+		omap_iommu_vfree(isp->domain, isp->dev, req->table);
 	kfree(req);
 }
 
@@ -438,7 +438,7 @@ static int ccdc_lsc_config(struct isp_ccdc_device *ccdc,
 
 		req->enable = 1;
 
-		req->table = omap_iommu_vmalloc(isp->domain, isp->iommu, 0,
+		req->table = omap_iommu_vmalloc(isp->domain, isp->dev, 0,
 					req->config.size, IOMMU_FLAG);
 		if (IS_ERR_VALUE(req->table)) {
 			req->table = 0;
@@ -446,7 +446,7 @@ static int ccdc_lsc_config(struct isp_ccdc_device *ccdc,
 			goto done;
 		}
 
-		req->iovm = omap_find_iovm_area(isp->iommu, req->table);
+		req->iovm = omap_find_iovm_area(isp->dev, req->table);
 		if (req->iovm == NULL) {
 			ret = -ENOMEM;
 			goto done;
@@ -462,7 +462,7 @@ static int ccdc_lsc_config(struct isp_ccdc_device *ccdc,
 		dma_sync_sg_for_cpu(isp->dev, req->iovm->sgt->sgl,
 				    req->iovm->sgt->nents, DMA_TO_DEVICE);
 
-		table = omap_da_to_va(isp->iommu, req->table);
+		table = omap_da_to_va(isp->dev, req->table);
 		if (copy_from_user(table, config->lsc, req->config.size)) {
 			ret = -EFAULT;
 			goto done;
@@ -734,15 +734,15 @@ static int ccdc_config(struct isp_ccdc_device *ccdc,
 			 * already done by omap_iommu_vmalloc().
 			 */
 			size = ccdc->fpc.fpnum * 4;
-			table_new = omap_iommu_vmalloc(isp->domain, isp->iommu,
+			table_new = omap_iommu_vmalloc(isp->domain, isp->dev,
 							0, size, IOMMU_FLAG);
 			if (IS_ERR_VALUE(table_new))
 				return -ENOMEM;
 
-			if (copy_from_user(omap_da_to_va(isp->iommu, table_new),
+			if (copy_from_user(omap_da_to_va(isp->dev, table_new),
 					   (__force void __user *)
 					   ccdc->fpc.fpcaddr, size)) {
-				omap_iommu_vfree(isp->domain, isp->iommu,
+				omap_iommu_vfree(isp->domain, isp->dev,
 								table_new);
 				return -EFAULT;
 			}
@@ -753,7 +753,7 @@ static int ccdc_config(struct isp_ccdc_device *ccdc,
 
 		ccdc_configure_fpc(ccdc);
 		if (table_old != 0)
-			omap_iommu_vfree(isp->domain, isp->iommu, table_old);
+			omap_iommu_vfree(isp->domain, isp->dev, table_old);
 	}
 
 	return ccdc_lsc_config(ccdc, ccdc_struct);
@@ -2309,7 +2309,7 @@ void omap3isp_ccdc_cleanup(struct isp_device *isp)
 	ccdc_lsc_free_queue(ccdc, &ccdc->lsc.free_queue);
 
 	if (ccdc->fpc.fpcaddr != 0)
-		omap_iommu_vfree(isp->domain, isp->iommu, ccdc->fpc.fpcaddr);
+		omap_iommu_vfree(isp->domain, isp->dev, ccdc->fpc.fpcaddr);
 
 	mutex_destroy(&ccdc->ioctl_lock);
 }

drivers/media/video/omap3isp/ispstat.c

@@ -366,7 +366,7 @@ static void isp_stat_bufs_free(struct ispstat *stat)
 				dma_unmap_sg(isp->dev, buf->iovm->sgt->sgl,
 					     buf->iovm->sgt->nents,
 					     DMA_FROM_DEVICE);
-			omap_iommu_vfree(isp->domain, isp->iommu,
+			omap_iommu_vfree(isp->domain, isp->dev,
 							buf->iommu_addr);
 		} else {
 			if (!buf->virt_addr)
@@ -400,7 +400,7 @@ static int isp_stat_bufs_alloc_iommu(struct ispstat *stat, unsigned int size)
 		struct iovm_struct *iovm;
 
 		WARN_ON(buf->dma_addr);
-		buf->iommu_addr = omap_iommu_vmalloc(isp->domain, isp->iommu, 0,
+		buf->iommu_addr = omap_iommu_vmalloc(isp->domain, isp->dev, 0,
 							size, IOMMU_FLAG);
 		if (IS_ERR((void *)buf->iommu_addr)) {
 			dev_err(stat->isp->dev,
@@ -410,7 +410,7 @@ static int isp_stat_bufs_alloc_iommu(struct ispstat *stat, unsigned int size)
 			return -ENOMEM;
 		}
 
-		iovm = omap_find_iovm_area(isp->iommu, buf->iommu_addr);
+		iovm = omap_find_iovm_area(isp->dev, buf->iommu_addr);
 		if (!iovm ||
 		    !dma_map_sg(isp->dev, iovm->sgt->sgl, iovm->sgt->nents,
 				DMA_FROM_DEVICE)) {
@@ -419,7 +419,7 @@ static int isp_stat_bufs_alloc_iommu(struct ispstat *stat, unsigned int size)
 		}
 		buf->iovm = iovm;
 
-		buf->virt_addr = omap_da_to_va(stat->isp->iommu,
+		buf->virt_addr = omap_da_to_va(stat->isp->dev,
 					  (u32)buf->iommu_addr);
 		buf->empty = 1;
 		dev_dbg(stat->isp->dev, "%s: buffer[%d] allocated."

drivers/media/video/omap3isp/ispvideo.c

@@ -453,7 +453,7 @@ ispmmu_vmap(struct isp_device *isp, const struct scatterlist *sglist, int sglen)
 	sgt->nents = sglen;
 	sgt->orig_nents = sglen;
 
-	da = omap_iommu_vmap(isp->domain, isp->iommu, 0, sgt, IOMMU_FLAG);
+	da = omap_iommu_vmap(isp->domain, isp->dev, 0, sgt, IOMMU_FLAG);
 	if (IS_ERR_VALUE(da))
 		kfree(sgt);
 
@@ -469,7 +469,7 @@ static void ispmmu_vunmap(struct isp_device *isp, dma_addr_t da)
 {
 	struct sg_table *sgt;
 
-	sgt = omap_iommu_vunmap(isp->domain, isp->iommu, (u32)da);
+	sgt = omap_iommu_vunmap(isp->domain, isp->dev, (u32)da);
 	kfree(sgt);
 }
 

drivers/pci/ats.c

@@ -175,21 +175,22 @@ int pci_enable_pri(struct pci_dev *pdev, u32 reqs)
 	u32 max_requests;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PRI_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
 	if (!pos)
 		return -EINVAL;
 
-	pci_read_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, &control);
-	pci_read_config_word(pdev, pos + PCI_PRI_STATUS_OFF,  &status);
-	if ((control & PCI_PRI_ENABLE) || !(status & PCI_PRI_STATUS_STOPPED))
+	pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+	pci_read_config_word(pdev, pos + PCI_PRI_STATUS, &status);
+	if ((control & PCI_PRI_CTRL_ENABLE) ||
+	    !(status & PCI_PRI_STATUS_STOPPED))
 		return -EBUSY;
 
-	pci_read_config_dword(pdev, pos + PCI_PRI_MAX_REQ_OFF, &max_requests);
+	pci_read_config_dword(pdev, pos + PCI_PRI_MAX_REQ, &max_requests);
 	reqs = min(max_requests, reqs);
-	pci_write_config_dword(pdev, pos + PCI_PRI_ALLOC_REQ_OFF, reqs);
+	pci_write_config_dword(pdev, pos + PCI_PRI_ALLOC_REQ, reqs);
 
-	control |= PCI_PRI_ENABLE;
-	pci_write_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, control);
+	control |= PCI_PRI_CTRL_ENABLE;
+	pci_write_config_word(pdev, pos + PCI_PRI_CTRL, control);
 
 	return 0;
 }
@@ -206,13 +207,13 @@ void pci_disable_pri(struct pci_dev *pdev)
 	u16 control;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PRI_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
 	if (!pos)
 		return;
 
-	pci_read_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, &control);
-	control &= ~PCI_PRI_ENABLE;
-	pci_write_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, control);
+	pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+	control &= ~PCI_PRI_CTRL_ENABLE;
+	pci_write_config_word(pdev, pos + PCI_PRI_CTRL, control);
 }
 EXPORT_SYMBOL_GPL(pci_disable_pri);
 
@@ -227,13 +228,13 @@ bool pci_pri_enabled(struct pci_dev *pdev)
 	u16 control;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PRI_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
 	if (!pos)
 		return false;
 
-	pci_read_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, &control);
+	pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
 
-	return (control & PCI_PRI_ENABLE) ? true : false;
+	return (control & PCI_PRI_CTRL_ENABLE) ? true : false;
 }
 EXPORT_SYMBOL_GPL(pci_pri_enabled);
 
@@ -249,17 +250,17 @@ int pci_reset_pri(struct pci_dev *pdev)
 	u16 control;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PRI_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
 	if (!pos)
 		return -EINVAL;
 
-	pci_read_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, &control);
-	if (control & PCI_PRI_ENABLE)
+	pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+	if (control & PCI_PRI_CTRL_ENABLE)
 		return -EBUSY;
 
-	control |= PCI_PRI_RESET;
+	control |= PCI_PRI_CTRL_RESET;
 
-	pci_write_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, control);
+	pci_write_config_word(pdev, pos + PCI_PRI_CTRL, control);
 
 	return 0;
 }
@@ -282,14 +283,14 @@ bool pci_pri_stopped(struct pci_dev *pdev)
 	u16 control, status;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PRI_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
 	if (!pos)
 		return true;
 
-	pci_read_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, &control);
-	pci_read_config_word(pdev, pos + PCI_PRI_STATUS_OFF,  &status);
+	pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+	pci_read_config_word(pdev, pos + PCI_PRI_STATUS, &status);
 
-	if (control & PCI_PRI_ENABLE)
+	if (control & PCI_PRI_CTRL_ENABLE)
 		return false;
 
 	return (status & PCI_PRI_STATUS_STOPPED) ? true : false;
@@ -311,15 +312,15 @@ int pci_pri_status(struct pci_dev *pdev)
 	u16 status, control;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PRI_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
 	if (!pos)
 		return -EINVAL;
 
-	pci_read_config_word(pdev, pos + PCI_PRI_CONTROL_OFF, &control);
-	pci_read_config_word(pdev, pos + PCI_PRI_STATUS_OFF,  &status);
+	pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+	pci_read_config_word(pdev, pos + PCI_PRI_STATUS, &status);
 
 	/* Stopped bit is undefined when enable == 1, so clear it */
-	if (control & PCI_PRI_ENABLE)
+	if (control & PCI_PRI_CTRL_ENABLE)
 		status &= ~PCI_PRI_STATUS_STOPPED;
 
 	return status;
@@ -342,25 +343,25 @@ int pci_enable_pasid(struct pci_dev *pdev, int features)
 	u16 control, supported;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PASID_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PASID);
 	if (!pos)
 		return -EINVAL;
 
-	pci_read_config_word(pdev, pos + PCI_PASID_CONTROL_OFF, &control);
-	pci_read_config_word(pdev, pos + PCI_PASID_CAP_OFF,     &supported);
+	pci_read_config_word(pdev, pos + PCI_PASID_CTRL, &control);
+	pci_read_config_word(pdev, pos + PCI_PASID_CAP, &supported);
 
-	if (!(supported & PCI_PASID_ENABLE))
+	if (control & PCI_PASID_CTRL_ENABLE)
 		return -EINVAL;
 
-	supported &= PCI_PASID_EXEC | PCI_PASID_PRIV;
+	supported &= PCI_PASID_CAP_EXEC | PCI_PASID_CAP_PRIV;
 
 	/* User wants to enable anything unsupported? */
 	if ((supported & features) != features)
 		return -EINVAL;
 
-	control = PCI_PASID_ENABLE | features;
+	control = PCI_PASID_CTRL_ENABLE | features;
 
-	pci_write_config_word(pdev, pos + PCI_PASID_CONTROL_OFF, control);
+	pci_write_config_word(pdev, pos + PCI_PASID_CTRL, control);
 
 	return 0;
 }
@@ -376,11 +377,11 @@ void pci_disable_pasid(struct pci_dev *pdev)
 	u16 control = 0;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PASID_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PASID);
 	if (!pos)
 		return;
 
-	pci_write_config_word(pdev, pos + PCI_PASID_CONTROL_OFF, control);
+	pci_write_config_word(pdev, pos + PCI_PASID_CTRL, control);
 }
 EXPORT_SYMBOL_GPL(pci_disable_pasid);
 
@@ -391,22 +392,21 @@ EXPORT_SYMBOL_GPL(pci_disable_pasid);
  * Returns a negative value when no PASI capability is present.
  * Otherwise is returns a bitmask with supported features. Current
  * features reported are:
- * PCI_PASID_ENABLE - PASID capability can be enabled
- * PCI_PASID_EXEC - Execute permission supported
- * PCI_PASID_PRIV - Priviledged mode supported
+ * PCI_PASID_CAP_EXEC - Execute permission supported
+ * PCI_PASID_CAP_PRIV - Priviledged mode supported
  */
 int pci_pasid_features(struct pci_dev *pdev)
 {
 	u16 supported;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PASID_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PASID);
 	if (!pos)
 		return -EINVAL;
 
-	pci_read_config_word(pdev, pos + PCI_PASID_CAP_OFF, &supported);
+	pci_read_config_word(pdev, pos + PCI_PASID_CAP, &supported);
 
-	supported &= PCI_PASID_ENABLE | PCI_PASID_EXEC | PCI_PASID_PRIV;
+	supported &= PCI_PASID_CAP_EXEC | PCI_PASID_CAP_PRIV;
 
 	return supported;
 }
@@ -426,11 +426,11 @@ int pci_max_pasids(struct pci_dev *pdev)
 	u16 supported;
 	int pos;
 
-	pos = pci_find_ext_capability(pdev, PCI_PASID_CAP);
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PASID);
 	if (!pos)
 		return -EINVAL;
 
-	pci_read_config_word(pdev, pos + PCI_PASID_CAP_OFF, &supported);
+	pci_read_config_word(pdev, pos + PCI_PASID_CAP, &supported);
 
 	supported = (supported & PASID_NUMBER_MASK) >> PASID_NUMBER_SHIFT;
 

drivers/pci/hotplug/pciehp.h

@@ -45,7 +45,6 @@ extern int pciehp_poll_time;
 extern int pciehp_debug;
 extern int pciehp_force;
 extern struct workqueue_struct *pciehp_wq;
-extern struct workqueue_struct *pciehp_ordered_wq;
 
 #define dbg(format, arg...)						\
 do {									\

drivers/pci/hotplug/pciehp_core.c

@@ -43,7 +43,6 @@ int pciehp_poll_mode;
 int pciehp_poll_time;
 int pciehp_force;
 struct workqueue_struct *pciehp_wq;
-struct workqueue_struct *pciehp_ordered_wq;
 
 #define DRIVER_VERSION	"0.4"
 #define DRIVER_AUTHOR	"Dan Zink <dan.zink@compaq.com>, Greg Kroah-Hartman <greg@kroah.com>, Dely Sy <dely.l.sy@intel.com>"
@@ -345,18 +344,11 @@ static int __init pcied_init(void)
 	if (!pciehp_wq)
 		return -ENOMEM;
 
-	pciehp_ordered_wq = alloc_ordered_workqueue("pciehp_ordered", 0);
-	if (!pciehp_ordered_wq) {
-		destroy_workqueue(pciehp_wq);
-		return -ENOMEM;
-	}
-
 	pciehp_firmware_init();
 	retval = pcie_port_service_register(&hpdriver_portdrv);
  	dbg("pcie_port_service_register = %d\n", retval);
   	info(DRIVER_DESC " version: " DRIVER_VERSION "\n");
  	if (retval) {
-		destroy_workqueue(pciehp_ordered_wq);
 		destroy_workqueue(pciehp_wq);
 		dbg("Failure to register service\n");
 	}
@@ -366,9 +358,8 @@ static int __init pcied_init(void)
 static void __exit pcied_cleanup(void)
 {
 	dbg("unload_pciehpd()\n");
-	destroy_workqueue(pciehp_ordered_wq);
-	destroy_workqueue(pciehp_wq);
 	pcie_port_service_unregister(&hpdriver_portdrv);
+	destroy_workqueue(pciehp_wq);
 	info(DRIVER_DESC " version: " DRIVER_VERSION " unloaded\n");
 }
 

drivers/pci/hotplug/pciehp_ctrl.c

@@ -344,7 +344,7 @@ void pciehp_queue_pushbutton_work(struct work_struct *work)
 		kfree(info);
 		goto out;
 	}
-	queue_work(pciehp_ordered_wq, &info->work);
+	queue_work(pciehp_wq, &info->work);
  out:
 	mutex_unlock(&p_slot->lock);
 }
@@ -439,7 +439,7 @@ static void handle_surprise_event(struct slot *p_slot)
 	else
 		p_slot->state = POWERON_STATE;
 
-	queue_work(pciehp_ordered_wq, &info->work);
+	queue_work(pciehp_wq, &info->work);
 }
 
 static void interrupt_event_handler(struct work_struct *work)

drivers/pci/hotplug/pciehp_hpc.c

@@ -806,7 +806,6 @@ static void pcie_cleanup_slot(struct controller *ctrl)
 	struct slot *slot = ctrl->slot;
 	cancel_delayed_work(&slot->work);
 	flush_workqueue(pciehp_wq);
-	flush_workqueue(pciehp_ordered_wq);
 	kfree(slot);
 }
 

drivers/pci/msi.c

@@ -323,6 +323,8 @@ static void free_msi_irqs(struct pci_dev *dev)
 			if (list_is_last(&entry->list, &dev->msi_list))
 				iounmap(entry->mask_base);
 		}
+		kobject_del(&entry->kobj);
+		kobject_put(&entry->kobj);
 		list_del(&entry->list);
 		kfree(entry);
 	}
@@ -403,6 +405,98 @@ void pci_restore_msi_state(struct pci_dev *dev)
 }
 EXPORT_SYMBOL_GPL(pci_restore_msi_state);
 
+
+#define to_msi_attr(obj) container_of(obj, struct msi_attribute, attr)
+#define to_msi_desc(obj) container_of(obj, struct msi_desc, kobj)
+
+struct msi_attribute {
+	struct attribute        attr;
+	ssize_t (*show)(struct msi_desc *entry, struct msi_attribute *attr,
+			char *buf);
+	ssize_t (*store)(struct msi_desc *entry, struct msi_attribute *attr,
+			 const char *buf, size_t count);
+};
+
+static ssize_t show_msi_mode(struct msi_desc *entry, struct msi_attribute *atr,
+			     char *buf)
+{
+	return sprintf(buf, "%s\n", entry->msi_attrib.is_msix ? "msix" : "msi");
+}
+
+static ssize_t msi_irq_attr_show(struct kobject *kobj,
+				 struct attribute *attr, char *buf)
+{
+	struct msi_attribute *attribute = to_msi_attr(attr);
+	struct msi_desc *entry = to_msi_desc(kobj);
+
+	if (!attribute->show)
+		return -EIO;
+
+	return attribute->show(entry, attribute, buf);
+}
+
+static const struct sysfs_ops msi_irq_sysfs_ops = {
+	.show = msi_irq_attr_show,
+};
+
+static struct msi_attribute mode_attribute =
+	__ATTR(mode, S_IRUGO, show_msi_mode, NULL);
+
+
+struct attribute *msi_irq_default_attrs[] = {
+	&mode_attribute.attr,
+	NULL
+};
+
+void msi_kobj_release(struct kobject *kobj)
+{
+	struct msi_desc *entry = to_msi_desc(kobj);
+
+	pci_dev_put(entry->dev);
+}
+
+static struct kobj_type msi_irq_ktype = {
+	.release = msi_kobj_release,
+	.sysfs_ops = &msi_irq_sysfs_ops,
+	.default_attrs = msi_irq_default_attrs,
+};
+
+static int populate_msi_sysfs(struct pci_dev *pdev)
+{
+	struct msi_desc *entry;
+	struct kobject *kobj;
+	int ret;
+	int count = 0;
+
+	pdev->msi_kset = kset_create_and_add("msi_irqs", NULL, &pdev->dev.kobj);
+	if (!pdev->msi_kset)
+		return -ENOMEM;
+
+	list_for_each_entry(entry, &pdev->msi_list, list) {
+		kobj = &entry->kobj;
+		kobj->kset = pdev->msi_kset;
+		pci_dev_get(pdev);
+		ret = kobject_init_and_add(kobj, &msi_irq_ktype, NULL,
+				     "%u", entry->irq);
+		if (ret)
+			goto out_unroll;
+
+		count++;
+	}
+
+	return 0;
+
+out_unroll:
+	list_for_each_entry(entry, &pdev->msi_list, list) {
+		if (!count)
+			break;
+		kobject_del(&entry->kobj);
+		kobject_put(&entry->kobj);
+		count--;
+	}
+	return ret;
+}
+
 /**
  * msi_capability_init - configure device's MSI capability structure
  * @dev: pointer to the pci_dev data structure of MSI device function
@@ -454,6 +548,13 @@ static int msi_capability_init(struct pci_dev *dev, int nvec)
 		return ret;
 	}
 
+	ret = populate_msi_sysfs(dev);
+	if (ret) {
+		msi_mask_irq(entry, mask, ~mask);
+		free_msi_irqs(dev);
+		return ret;
+	}
+
 	/* Set MSI enabled bits	 */
 	pci_intx_for_msi(dev, 0);
 	msi_set_enable(dev, pos, 1);
@@ -574,6 +675,12 @@ static int msix_capability_init(struct pci_dev *dev,
 
 	msix_program_entries(dev, entries);
 
+	ret = populate_msi_sysfs(dev);
+	if (ret) {
+		ret = 0;
+		goto error;
+	}
+
 	/* Set MSI-X enabled bits and unmask the function */
 	pci_intx_for_msi(dev, 0);
 	dev->msix_enabled = 1;
@@ -732,6 +839,8 @@ void pci_disable_msi(struct pci_dev *dev)
 
 	pci_msi_shutdown(dev);
 	free_msi_irqs(dev);
+	kset_unregister(dev->msi_kset);
+	dev->msi_kset = NULL;
 }
 EXPORT_SYMBOL(pci_disable_msi);
 
@@ -830,6 +939,8 @@ void pci_disable_msix(struct pci_dev *dev)
 
 	pci_msix_shutdown(dev);
 	free_msi_irqs(dev);
+	kset_unregister(dev->msi_kset);
+	dev->msi_kset = NULL;
 }
 EXPORT_SYMBOL(pci_disable_msix);
 
@@ -870,5 +981,15 @@ EXPORT_SYMBOL(pci_msi_enabled);
 
 void pci_msi_init_pci_dev(struct pci_dev *dev)
 {
+	int pos;
 	INIT_LIST_HEAD(&dev->msi_list);
+
+	/* Disable the msi hardware to avoid screaming interrupts
+	 * during boot.  This is the power on reset default so
+	 * usually this should be a noop.
+	 */
+	pos = pci_find_capability(dev, PCI_CAP_ID_MSI);
+	if (pos)
+		msi_set_enable(dev, pos, 0);
+	msix_set_enable(dev, 0);
 }

drivers/pci/pci-acpi.c

@@ -45,16 +45,20 @@ static void pci_acpi_wake_dev(acpi_handle handle, u32 event, void *context)
 {
 	struct pci_dev *pci_dev = context;
 
-	if (event == ACPI_NOTIFY_DEVICE_WAKE && pci_dev) {
+	if (event != ACPI_NOTIFY_DEVICE_WAKE || !pci_dev)
+		return;
+
+	if (!pci_dev->pm_cap || !pci_dev->pme_support
+	     || pci_check_pme_status(pci_dev)) {
 		if (pci_dev->pme_poll)
 			pci_dev->pme_poll = false;
 
 		pci_wakeup_event(pci_dev);
-		pci_check_pme_status(pci_dev);
 		pm_runtime_resume(&pci_dev->dev);
-		if (pci_dev->subordinate)
-			pci_pme_wakeup_bus(pci_dev->subordinate);
 	}
+
+	if (pci_dev->subordinate)
+		pci_pme_wakeup_bus(pci_dev->subordinate);
 }
 
 /**
@@ -395,7 +399,6 @@ static int __init acpi_pci_init(void)
 
 	if (acpi_gbl_FADT.boot_flags & ACPI_FADT_NO_ASPM) {
 		printk(KERN_INFO"ACPI FADT declares the system doesn't support PCIe ASPM, so disable it\n");
-		pcie_clear_aspm();
 		pcie_no_aspm();
 	}
 

drivers/pci/pcie/aspm.c

@@ -68,7 +68,7 @@ struct pcie_link_state {
 	struct aspm_latency acceptable[8];
 };
 
-static int aspm_disabled, aspm_force, aspm_clear_state;
+static int aspm_disabled, aspm_force;
 static bool aspm_support_enabled = true;
 static DEFINE_MUTEX(aspm_lock);
 static LIST_HEAD(link_list);
@@ -500,9 +500,6 @@ static int pcie_aspm_sanity_check(struct pci_dev *pdev)
 	int pos;
 	u32 reg32;
 
-	if (aspm_clear_state)
-		return -EINVAL;
-
 	/*
 	 * Some functions in a slot might not all be PCIe functions,
 	 * very strange. Disable ASPM for the whole slot
@@ -574,9 +571,6 @@ void pcie_aspm_init_link_state(struct pci_dev *pdev)
 	    pdev->pcie_type != PCI_EXP_TYPE_DOWNSTREAM)
 		return;
 
-	if (aspm_disabled && !aspm_clear_state)
-		return;
-
 	/* VIA has a strange chipset, root port is under a bridge */
 	if (pdev->pcie_type == PCI_EXP_TYPE_ROOT_PORT &&
 	    pdev->bus->self)
@@ -608,7 +602,7 @@ void pcie_aspm_init_link_state(struct pci_dev *pdev)
 	 * the BIOS's expectation, we'll do so once pci_enable_device() is
 	 * called.
 	 */
-	if (aspm_policy != POLICY_POWERSAVE || aspm_clear_state) {
+	if (aspm_policy != POLICY_POWERSAVE) {
 		pcie_config_aspm_path(link);
 		pcie_set_clkpm(link, policy_to_clkpm_state(link));
 	}
@@ -649,8 +643,7 @@ void pcie_aspm_exit_link_state(struct pci_dev *pdev)
 	struct pci_dev *parent = pdev->bus->self;
 	struct pcie_link_state *link, *root, *parent_link;
 
-	if ((aspm_disabled && !aspm_clear_state) || !pci_is_pcie(pdev) ||
-	    !parent || !parent->link_state)
+	if (!pci_is_pcie(pdev) || !parent || !parent->link_state)
 		return;
 	if ((parent->pcie_type != PCI_EXP_TYPE_ROOT_PORT) &&
 	    (parent->pcie_type != PCI_EXP_TYPE_DOWNSTREAM))
@@ -734,13 +727,18 @@ void pcie_aspm_powersave_config_link(struct pci_dev *pdev)
  * pci_disable_link_state - disable pci device's link state, so the link will
  * never enter specific states
  */
-static void __pci_disable_link_state(struct pci_dev *pdev, int state, bool sem)
+static void __pci_disable_link_state(struct pci_dev *pdev, int state, bool sem,
+				     bool force)
 {
 	struct pci_dev *parent = pdev->bus->self;
 	struct pcie_link_state *link;
 
-	if (aspm_disabled || !pci_is_pcie(pdev))
+	if (aspm_disabled && !force)
+		return;
+
+	if (!pci_is_pcie(pdev))
 		return;
+
 	if (pdev->pcie_type == PCI_EXP_TYPE_ROOT_PORT ||
 	    pdev->pcie_type == PCI_EXP_TYPE_DOWNSTREAM)
 		parent = pdev;
@@ -768,16 +766,31 @@ static void __pci_disable_link_state(struct pci_dev *pdev, int state, bool sem)
 
 void pci_disable_link_state_locked(struct pci_dev *pdev, int state)
 {
-	__pci_disable_link_state(pdev, state, false);
+	__pci_disable_link_state(pdev, state, false, false);
 }
 EXPORT_SYMBOL(pci_disable_link_state_locked);
 
 void pci_disable_link_state(struct pci_dev *pdev, int state)
 {
-	__pci_disable_link_state(pdev, state, true);
+	__pci_disable_link_state(pdev, state, true, false);
 }
 EXPORT_SYMBOL(pci_disable_link_state);
 
+void pcie_clear_aspm(struct pci_bus *bus)
+{
+	struct pci_dev *child;
+
+	/*
+	 * Clear any ASPM setup that the firmware has carried out on this bus
+	 */
+	list_for_each_entry(child, &bus->devices, bus_list) {
+		__pci_disable_link_state(child, PCIE_LINK_STATE_L0S |
+					 PCIE_LINK_STATE_L1 |
+					 PCIE_LINK_STATE_CLKPM,
+					 false, true);
+	}
+}
+
 static int pcie_aspm_set_policy(const char *val, struct kernel_param *kp)
 {
 	int i;
@@ -935,6 +948,7 @@ void pcie_aspm_remove_sysfs_dev_files(struct pci_dev *pdev)
 static int __init pcie_aspm_disable(char *str)
 {
 	if (!strcmp(str, "off")) {
+		aspm_policy = POLICY_DEFAULT;
 		aspm_disabled = 1;
 		aspm_support_enabled = false;
 		printk(KERN_INFO "PCIe ASPM is disabled\n");
@@ -947,16 +961,18 @@ static int __init pcie_aspm_disable(char *str)
 
 __setup("pcie_aspm=", pcie_aspm_disable);
 
-void pcie_clear_aspm(void)
-{
-	if (!aspm_force)
-		aspm_clear_state = 1;
-}
-
 void pcie_no_aspm(void)
 {
-	if (!aspm_force)
+	/*
+	 * Disabling ASPM is intended to prevent the kernel from modifying
+	 * existing hardware state, not to clear existing state. To that end:
+	 * (a) set policy to POLICY_DEFAULT in order to avoid changing state
+	 * (b) prevent userspace from changing policy
+	 */
+	if (!aspm_force) {
+		aspm_policy = POLICY_DEFAULT;
 		aspm_disabled = 1;
+	}
 }
 
 /**

include/linux/acpi.h

@@ -302,6 +302,10 @@ extern bool osc_sb_apei_support_acked;
 				OSC_PCI_EXPRESS_PME_CONTROL |		\
 				OSC_PCI_EXPRESS_AER_CONTROL |		\
 				OSC_PCI_EXPRESS_CAP_STRUCTURE_CONTROL)
+
+#define OSC_PCI_NATIVE_HOTPLUG	(OSC_PCI_EXPRESS_NATIVE_HP_CONTROL |	\
+				OSC_SHPC_NATIVE_HP_CONTROL)
+
 extern acpi_status acpi_pci_osc_control_set(acpi_handle handle,
 					     u32 *mask, u32 req);
 extern void acpi_early_init(void);

include/linux/amd-iommu.h

@@ -20,12 +20,148 @@
 #ifndef _ASM_X86_AMD_IOMMU_H
 #define _ASM_X86_AMD_IOMMU_H
 
-#include <linux/irqreturn.h>
+#include <linux/types.h>
 
 #ifdef CONFIG_AMD_IOMMU
 
+struct task_struct;
+struct pci_dev;
+
 extern int amd_iommu_detect(void);
 
+
+/**
+ * amd_iommu_enable_device_erratum() - Enable erratum workaround for device
+ *				       in the IOMMUv2 driver
+ * @pdev: The PCI device the workaround is necessary for
+ * @erratum: The erratum workaround to enable
+ *
+ * The function needs to be called before amd_iommu_init_device().
+ * Possible values for the erratum number are for now:
+ * - AMD_PRI_DEV_ERRATUM_ENABLE_RESET - Reset PRI capability when PRI
+ *					is enabled
+ * - AMD_PRI_DEV_ERRATUM_LIMIT_REQ_ONE - Limit number of outstanding PRI
+ *					 requests to one
+ */
+#define AMD_PRI_DEV_ERRATUM_ENABLE_RESET		0
+#define AMD_PRI_DEV_ERRATUM_LIMIT_REQ_ONE		1
+
+extern void amd_iommu_enable_device_erratum(struct pci_dev *pdev, u32 erratum);
+
+/**
+ * amd_iommu_init_device() - Init device for use with IOMMUv2 driver
+ * @pdev: The PCI device to initialize
+ * @pasids: Number of PASIDs to support for this device
+ *
+ * This function does all setup for the device pdev so that it can be
+ * used with IOMMUv2.
+ * Returns 0 on success or negative value on error.
+ */
+extern int amd_iommu_init_device(struct pci_dev *pdev, int pasids);
+
+/**
+ * amd_iommu_free_device() - Free all IOMMUv2 related device resources
+ *			     and disable IOMMUv2 usage for this device
+ * @pdev: The PCI device to disable IOMMUv2 usage for'
+ */
+extern void amd_iommu_free_device(struct pci_dev *pdev);
+
+/**
+ * amd_iommu_bind_pasid() - Bind a given task to a PASID on a device
+ * @pdev: The PCI device to bind the task to
+ * @pasid: The PASID on the device the task should be bound to
+ * @task: the task to bind
+ *
+ * The function returns 0 on success or a negative value on error.
+ */
+extern int amd_iommu_bind_pasid(struct pci_dev *pdev, int pasid,
+				struct task_struct *task);
+
+/**
+ * amd_iommu_unbind_pasid() - Unbind a PASID from its task on
+ *			      a device
+ * @pdev: The device of the PASID
+ * @pasid: The PASID to unbind
+ *
+ * When this function returns the device is no longer using the PASID
+ * and the PASID is no longer bound to its task.
+ */
+extern void amd_iommu_unbind_pasid(struct pci_dev *pdev, int pasid);
+
+/**
+ * amd_iommu_set_invalid_ppr_cb() - Register a call-back for failed
+ *				    PRI requests
+ * @pdev: The PCI device the call-back should be registered for
+ * @cb: The call-back function
+ *
+ * The IOMMUv2 driver invokes this call-back when it is unable to
+ * successfully handle a PRI request. The device driver can then decide
+ * which PRI response the device should see. Possible return values for
+ * the call-back are:
+ *
+ * - AMD_IOMMU_INV_PRI_RSP_SUCCESS - Send SUCCESS back to the device
+ * - AMD_IOMMU_INV_PRI_RSP_INVALID - Send INVALID back to the device
+ * - AMD_IOMMU_INV_PRI_RSP_FAIL    - Send Failure back to the device,
+ *				     the device is required to disable
+ *				     PRI when it receives this response
+ *
+ * The function returns 0 on success or negative value on error.
+ */
+#define AMD_IOMMU_INV_PRI_RSP_SUCCESS	0
+#define AMD_IOMMU_INV_PRI_RSP_INVALID	1
+#define AMD_IOMMU_INV_PRI_RSP_FAIL	2
+
+typedef int (*amd_iommu_invalid_ppr_cb)(struct pci_dev *pdev,
+					int pasid,
+					unsigned long address,
+					u16);
+
+extern int amd_iommu_set_invalid_ppr_cb(struct pci_dev *pdev,
+					amd_iommu_invalid_ppr_cb cb);
+
+/**
+ * amd_iommu_device_info() - Get information about IOMMUv2 support of a
+ *			     PCI device
+ * @pdev: PCI device to query information from
+ * @info: A pointer to an amd_iommu_device_info structure which will contain
+ *	  the information about the PCI device
+ *
+ * Returns 0 on success, negative value on error
+ */
+
+#define AMD_IOMMU_DEVICE_FLAG_ATS_SUP     0x1    /* ATS feature supported */
+#define AMD_IOMMU_DEVICE_FLAG_PRI_SUP     0x2    /* PRI feature supported */
+#define AMD_IOMMU_DEVICE_FLAG_PASID_SUP   0x4    /* PASID context supported */
+#define AMD_IOMMU_DEVICE_FLAG_EXEC_SUP    0x8    /* Device may request execution
+						    on memory pages */
+#define AMD_IOMMU_DEVICE_FLAG_PRIV_SUP   0x10    /* Device may request
+						    super-user privileges */
+
+struct amd_iommu_device_info {
+	int max_pasids;
+	u32 flags;
+};
+
+extern int amd_iommu_device_info(struct pci_dev *pdev,
+				 struct amd_iommu_device_info *info);
+
+/**
+ * amd_iommu_set_invalidate_ctx_cb() - Register a call-back for invalidating
+ *				       a pasid context. This call-back is
+ *				       invoked when the IOMMUv2 driver needs to
+ *				       invalidate a PASID context, for example
+ *				       because the task that is bound to that
+ *				       context is about to exit.
+ *
+ * @pdev: The PCI device the call-back should be registered for
+ * @cb: The call-back function
+ */
+
+typedef void (*amd_iommu_invalidate_ctx)(struct pci_dev *pdev, int pasid);
+
+extern int amd_iommu_set_invalidate_ctx_cb(struct pci_dev *pdev,
+					   amd_iommu_invalidate_ctx cb);
+
 #else
 
 static inline int amd_iommu_detect(void) { return -ENODEV; }

include/linux/iommu.h

@@ -48,19 +48,34 @@ struct iommu_domain {
 
 #ifdef CONFIG_IOMMU_API
 
+/**
+ * struct iommu_ops - iommu ops and capabilities
+ * @domain_init: init iommu domain
+ * @domain_destroy: destroy iommu domain
+ * @attach_dev: attach device to an iommu domain
+ * @detach_dev: detach device from an iommu domain
+ * @map: map a physically contiguous memory region to an iommu domain
+ * @unmap: unmap a physically contiguous memory region from an iommu domain
+ * @iova_to_phys: translate iova to physical address
+ * @domain_has_cap: domain capabilities query
+ * @commit: commit iommu domain
+ * @pgsize_bitmap: bitmap of supported page sizes
+ */
 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, int gfp_order, int prot);
-	int (*unmap)(struct iommu_domain *domain, unsigned long iova,
-		     int gfp_order);
+		   phys_addr_t paddr, size_t size, int prot);
+	size_t (*unmap)(struct iommu_domain *domain, unsigned long iova,
+		     size_t size);
 	phys_addr_t (*iova_to_phys)(struct iommu_domain *domain,
 				    unsigned long iova);
 	int (*domain_has_cap)(struct iommu_domain *domain,
 			      unsigned long cap);
+	int (*device_group)(struct device *dev, unsigned int *groupid);
+	unsigned long pgsize_bitmap;
 };
 
 extern int bus_set_iommu(struct bus_type *bus, struct iommu_ops *ops);
@@ -72,15 +87,16 @@ extern int iommu_attach_device(struct iommu_domain *domain,
 extern void iommu_detach_device(struct iommu_domain *domain,
 				struct device *dev);
 extern int iommu_map(struct iommu_domain *domain, unsigned long iova,
-		     phys_addr_t paddr, int gfp_order, int prot);
-extern int iommu_unmap(struct iommu_domain *domain, unsigned long iova,
-		       int gfp_order);
+		     phys_addr_t paddr, size_t size, int prot);
+extern size_t iommu_unmap(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);
 extern int iommu_domain_has_cap(struct iommu_domain *domain,
 				unsigned long cap);
 extern void iommu_set_fault_handler(struct iommu_domain *domain,
 					iommu_fault_handler_t handler);
+extern int iommu_device_group(struct device *dev, unsigned int *groupid);
 
 /**
  * report_iommu_fault() - report about an IOMMU fault to the IOMMU framework
@@ -179,6 +195,11 @@ static inline void iommu_set_fault_handler(struct iommu_domain *domain,
 {
 }
 
+static inline int iommu_device_group(struct device *dev, unsigned int *groupid)
+{
+	return -ENODEV;
+}
+
 #endif /* CONFIG_IOMMU_API */
 
 #endif /* __LINUX_IOMMU_H */

include/linux/msi.h

@@ -1,6 +1,7 @@
 #ifndef LINUX_MSI_H
 #define LINUX_MSI_H
 
+#include <linux/kobject.h>
 #include <linux/list.h>
 
 struct msi_msg {
@@ -44,6 +45,8 @@ struct msi_desc {
 
 	/* Last set MSI message */
 	struct msi_msg msg;
+
+	struct kobject kobj;
 };
 
 /*

include/linux/pci-aspm.h

@@ -29,7 +29,7 @@ extern void pcie_aspm_pm_state_change(struct pci_dev *pdev);
 extern void pcie_aspm_powersave_config_link(struct pci_dev *pdev);
 extern void pci_disable_link_state(struct pci_dev *pdev, int state);
 extern void pci_disable_link_state_locked(struct pci_dev *pdev, int state);
-extern void pcie_clear_aspm(void);
+extern void pcie_clear_aspm(struct pci_bus *bus);
 extern void pcie_no_aspm(void);
 #else
 static inline void pcie_aspm_init_link_state(struct pci_dev *pdev)
@@ -47,7 +47,7 @@ static inline void pcie_aspm_powersave_config_link(struct pci_dev *pdev)
 static inline void pci_disable_link_state(struct pci_dev *pdev, int state)
 {
 }
-static inline void pcie_clear_aspm(void)
+static inline void pcie_clear_aspm(struct pci_bus *bus)
 {
 }
 static inline void pcie_no_aspm(void)

include/linux/pci.h

@@ -336,6 +336,7 @@ struct pci_dev {
 	struct bin_attribute *res_attr_wc[DEVICE_COUNT_RESOURCE]; /* sysfs file for WC mapping of resources */
 #ifdef CONFIG_PCI_MSI
 	struct list_head msi_list;
+	struct kset *msi_kset;
 #endif
 	struct pci_vpd *vpd;
 #ifdef CONFIG_PCI_ATS

include/linux/pci_regs.h

@@ -537,7 +537,9 @@
 #define PCI_EXT_CAP_ID_ARI	14
 #define PCI_EXT_CAP_ID_ATS	15
 #define PCI_EXT_CAP_ID_SRIOV	16
+#define PCI_EXT_CAP_ID_PRI	19
 #define PCI_EXT_CAP_ID_LTR	24
+#define PCI_EXT_CAP_ID_PASID	27
 
 /* Advanced Error Reporting */
 #define PCI_ERR_UNCOR_STATUS	4	/* Uncorrectable Error Status */
@@ -664,24 +666,24 @@
 #define  PCI_ATS_MIN_STU	12	/* shift of minimum STU block */
 
 /* Page Request Interface */
-#define PCI_PRI_CAP		0x13    /* PRI capability ID */
-#define PCI_PRI_CONTROL_OFF	0x04	/* Offset of control register */
-#define PCI_PRI_STATUS_OFF	0x06	/* Offset of status register */
-#define PCI_PRI_ENABLE		0x0001	/* Enable mask */
-#define PCI_PRI_RESET		0x0002	/* Reset bit mask */
-#define PCI_PRI_STATUS_RF	0x0001  /* Request Failure */
-#define PCI_PRI_STATUS_UPRGI	0x0002  /* Unexpected PRG index */
-#define PCI_PRI_STATUS_STOPPED	0x0100  /* PRI Stopped */
-#define PCI_PRI_MAX_REQ_OFF	0x08	/* Cap offset for max reqs supported */
-#define PCI_PRI_ALLOC_REQ_OFF	0x0c	/* Cap offset for max reqs allowed */
+#define PCI_PRI_CTRL		0x04	/* PRI control register */
+#define  PCI_PRI_CTRL_ENABLE	0x01	/* Enable */
+#define  PCI_PRI_CTRL_RESET	0x02	/* Reset */
+#define PCI_PRI_STATUS		0x06	/* PRI status register */
+#define  PCI_PRI_STATUS_RF	0x001	/* Response Failure */
+#define  PCI_PRI_STATUS_UPRGI	0x002	/* Unexpected PRG index */
+#define  PCI_PRI_STATUS_STOPPED	0x100	/* PRI Stopped */
+#define PCI_PRI_MAX_REQ		0x08	/* PRI max reqs supported */
+#define PCI_PRI_ALLOC_REQ	0x0c	/* PRI max reqs allowed */
 
 /* PASID capability */
-#define PCI_PASID_CAP		0x1b    /* PASID capability ID */
-#define PCI_PASID_CAP_OFF	0x04    /* PASID feature register */
-#define PCI_PASID_CONTROL_OFF   0x06    /* PASID control register */
-#define PCI_PASID_ENABLE	0x01	/* Enable/Supported bit */
-#define PCI_PASID_EXEC		0x02	/* Exec permissions Enable/Supported */
-#define PCI_PASID_PRIV		0x04	/* Priviledge Mode Enable/Support */
+#define PCI_PASID_CAP		0x04    /* PASID feature register */
+#define  PCI_PASID_CAP_EXEC	0x02	/* Exec permissions Supported */
+#define  PCI_PASID_CAP_PRIV	0x04	/* Priviledge Mode Supported */
+#define PCI_PASID_CTRL		0x06    /* PASID control register */
+#define  PCI_PASID_CTRL_ENABLE	0x01	/* Enable bit */
+#define  PCI_PASID_CTRL_EXEC	0x02	/* Exec permissions Enable */
+#define  PCI_PASID_CTRL_PRIV	0x04	/* Priviledge Mode Enable */
 
 /* Single Root I/O Virtualization */
 #define PCI_SRIOV_CAP		0x04	/* SR-IOV Capabilities */

virt/kvm/iommu.c

@@ -113,7 +113,7 @@ int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot)
 
 		/* Map into IO address space */
 		r = iommu_map(domain, gfn_to_gpa(gfn), pfn_to_hpa(pfn),
-			      get_order(page_size), flags);
+			      page_size, flags);
 		if (r) {
 			printk(KERN_ERR "kvm_iommu_map_address:"
 			       "iommu failed to map pfn=%llx\n", pfn);
@@ -293,15 +293,15 @@ static void kvm_iommu_put_pages(struct kvm *kvm,
 
 	while (gfn < end_gfn) {
 		unsigned long unmap_pages;
-		int order;
+		size_t size;
 
 		/* Get physical address */
 		phys = iommu_iova_to_phys(domain, gfn_to_gpa(gfn));
 		pfn  = phys >> PAGE_SHIFT;
 
 		/* Unmap address from IO address space */
-		order       = iommu_unmap(domain, gfn_to_gpa(gfn), 0);
-		unmap_pages = 1ULL << order;
+		size       = iommu_unmap(domain, gfn_to_gpa(gfn), PAGE_SIZE);
+		unmap_pages = 1ULL << get_order(size);
 
 		/* Unpin all pages we just unmapped to not leak any memory */
 		kvm_unpin_pages(kvm, pfn, unmap_pages);