Merge git://git.infradead.org/iommu-2.6

* git://git.infradead.org/iommu-2.6:
  Admit to maintaining VT-d, for my sins.
  dmar: fix uninitialised 'ret' variable in dmar_parse_dev()
  intel-iommu: use coherent_dma_mask in alloc_coherent
  amd_iommu: fix nasty bug that caused ILLEGAL_DEVICE_TABLE_ENTRY errors
  intel-iommu: IA64 support
  dmar: remove the quirk which disables dma-remapping when intr-remapping enabled
  dmar: Use queued invalidation interface for IOTLB and context invalidation
  dmar: context cache and IOTLB invalidation using queued invalidation
  dmar: use spin_lock_irqsave() in qi_submit_sync()

MAINTAINERS

@@ -2182,6 +2182,13 @@ M:	maciej.sosnowski@intel.com
 L:	linux-kernel@vger.kernel.org
 S:	Supported
 
+INTEL IOMMU (VT-d)
+P:	David Woodhouse
+M:	dwmw2@infradead.org
+L:	iommu@lists.linux-foundation.org
+T:	git://git.infradead.org/iommu-2.6.git
+S:	Supported
+
 INTEL IOP-ADMA DMA DRIVER
 P:	Dan Williams
 M:	dan.j.williams@intel.com

arch/x86/kernel/amd_iommu_init.c

@@ -212,7 +212,7 @@ static void __init iommu_set_exclusion_range(struct amd_iommu *iommu)
 /* Programs the physical address of the device table into the IOMMU hardware */
 static void __init iommu_set_device_table(struct amd_iommu *iommu)
 {
-	u32 entry;
+	u64 entry;
 
 	BUG_ON(iommu->mmio_base == NULL);
 

arch/x86/kernel/pci-dma.c

@@ -9,8 +9,6 @@
 #include <asm/calgary.h>
 #include <asm/amd_iommu.h>
 
-static int forbid_dac __read_mostly;
-
 struct dma_mapping_ops *dma_ops;
 EXPORT_SYMBOL(dma_ops);
 
@@ -293,17 +291,3 @@ void pci_iommu_shutdown(void)
 }
 /* Must execute after PCI subsystem */
 fs_initcall(pci_iommu_init);
-
-#ifdef CONFIG_PCI
-/* Many VIA bridges seem to corrupt data for DAC. Disable it here */
-
-static __devinit void via_no_dac(struct pci_dev *dev)
-{
-	if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI && forbid_dac == 0) {
-		printk(KERN_INFO "PCI: VIA PCI bridge detected."
-				 "Disabling DAC.\n");
-		forbid_dac = 1;
-	}
-}
-DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_VIA, PCI_ANY_ID, via_no_dac);
-#endif

drivers/pci/dmar.c

@@ -188,8 +188,7 @@ dmar_parse_one_drhd(struct acpi_dmar_header *header)
 	return 0;
 }
 
-static int __init
-dmar_parse_dev(struct dmar_drhd_unit *dmaru)
+static int __init dmar_parse_dev(struct dmar_drhd_unit *dmaru)
 {
 	struct acpi_dmar_hardware_unit *drhd;
 	static int include_all;
@@ -277,14 +276,15 @@ dmar_table_print_dmar_entry(struct acpi_dmar_header *header)
 		drhd = (struct acpi_dmar_hardware_unit *)header;
 		printk (KERN_INFO PREFIX
 			"DRHD (flags: 0x%08x)base: 0x%016Lx\n",
-			drhd->flags, drhd->address);
+			drhd->flags, (unsigned long long)drhd->address);
 		break;
 	case ACPI_DMAR_TYPE_RESERVED_MEMORY:
 		rmrr = (struct acpi_dmar_reserved_memory *)header;
 
 		printk (KERN_INFO PREFIX
 			"RMRR base: 0x%016Lx end: 0x%016Lx\n",
-			rmrr->base_address, rmrr->end_address);
+			(unsigned long long)rmrr->base_address,
+			(unsigned long long)rmrr->end_address);
 		break;
 	}
 }
@@ -328,7 +328,7 @@ parse_dmar_table(void)
 	if (!dmar)
 		return -ENODEV;
 
-	if (dmar->width < PAGE_SHIFT_4K - 1) {
+	if (dmar->width < PAGE_SHIFT - 1) {
 		printk(KERN_WARNING PREFIX "Invalid DMAR haw\n");
 		return -EINVAL;
 	}
@@ -460,8 +460,8 @@ void __init detect_intel_iommu(void)
 
 	ret = dmar_table_detect();
 
-#ifdef CONFIG_DMAR
 	{
+#ifdef CONFIG_INTR_REMAP
 		struct acpi_table_dmar *dmar;
 		/*
 		 * for now we will disable dma-remapping when interrupt
@@ -470,29 +470,17 @@ void __init detect_intel_iommu(void)
 		 * is added, we will not need this any more.
 		 */
 		dmar = (struct acpi_table_dmar *) dmar_tbl;
-		if (ret && cpu_has_x2apic && dmar->flags & 0x1) {
+		if (ret && cpu_has_x2apic && dmar->flags & 0x1)
 			printk(KERN_INFO
 			       "Queued invalidation will be enabled to support "
 			       "x2apic and Intr-remapping.\n");
-			printk(KERN_INFO
-			       "Disabling IOMMU detection, because of missing "
-			       "queued invalidation support for IOTLB "
-			       "invalidation\n");
-			printk(KERN_INFO
-			       "Use \"nox2apic\", if you want to use Intel "
-			       " IOMMU for DMA-remapping and don't care about "
-			       " x2apic support\n");
-
-			dmar_disabled = 1;
-			goto end;
-		}
-
+#endif
+#ifdef CONFIG_DMAR
 		if (ret && !no_iommu && !iommu_detected && !swiotlb &&
 		    !dmar_disabled)
 			iommu_detected = 1;
-	}
-end:
 #endif
+	}
 	dmar_tbl = NULL;
 }
 
@@ -510,7 +498,7 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
 
 	iommu->seq_id = iommu_allocated++;
 
-	iommu->reg = ioremap(drhd->reg_base_addr, PAGE_SIZE_4K);
+	iommu->reg = ioremap(drhd->reg_base_addr, VTD_PAGE_SIZE);
 	if (!iommu->reg) {
 		printk(KERN_ERR "IOMMU: can't map the region\n");
 		goto error;
@@ -521,8 +509,8 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
 	/* the registers might be more than one page */
 	map_size = max_t(int, ecap_max_iotlb_offset(iommu->ecap),
 		cap_max_fault_reg_offset(iommu->cap));
-	map_size = PAGE_ALIGN_4K(map_size);
-	if (map_size > PAGE_SIZE_4K) {
+	map_size = VTD_PAGE_ALIGN(map_size);
+	if (map_size > VTD_PAGE_SIZE) {
 		iounmap(iommu->reg);
 		iommu->reg = ioremap(drhd->reg_base_addr, map_size);
 		if (!iommu->reg) {
@@ -533,8 +521,10 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
 
 	ver = readl(iommu->reg + DMAR_VER_REG);
 	pr_debug("IOMMU %llx: ver %d:%d cap %llx ecap %llx\n",
-		drhd->reg_base_addr, DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver),
-		iommu->cap, iommu->ecap);
+		(unsigned long long)drhd->reg_base_addr,
+		DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver),
+		(unsigned long long)iommu->cap,
+		(unsigned long long)iommu->ecap);
 
 	spin_lock_init(&iommu->register_lock);
 
@@ -587,11 +577,11 @@ void qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu)
 
 	hw = qi->desc;
 
-	spin_lock(&qi->q_lock);
+	spin_lock_irqsave(&qi->q_lock, flags);
 	while (qi->free_cnt < 3) {
-		spin_unlock(&qi->q_lock);
+		spin_unlock_irqrestore(&qi->q_lock, flags);
 		cpu_relax();
-		spin_lock(&qi->q_lock);
+		spin_lock_irqsave(&qi->q_lock, flags);
 	}
 
 	index = qi->free_head;
@@ -612,15 +602,22 @@ void qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu)
 	qi->free_head = (qi->free_head + 2) % QI_LENGTH;
 	qi->free_cnt -= 2;
 
-	spin_lock_irqsave(&iommu->register_lock, flags);
+	spin_lock(&iommu->register_lock);
 	/*
 	 * update the HW tail register indicating the presence of
 	 * new descriptors.
 	 */
 	writel(qi->free_head << 4, iommu->reg + DMAR_IQT_REG);
-	spin_unlock_irqrestore(&iommu->register_lock, flags);
+	spin_unlock(&iommu->register_lock);
 
 	while (qi->desc_status[wait_index] != QI_DONE) {
+		/*
+		 * We will leave the interrupts disabled, to prevent interrupt
+		 * context to queue another cmd while a cmd is already submitted
+		 * and waiting for completion on this cpu. This is to avoid
+		 * a deadlock where the interrupt context can wait indefinitely
+		 * for free slots in the queue.
+		 */
 		spin_unlock(&qi->q_lock);
 		cpu_relax();
 		spin_lock(&qi->q_lock);
@@ -629,7 +626,7 @@ void qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu)
 	qi->desc_status[index] = QI_DONE;
 
 	reclaim_free_desc(qi);
-	spin_unlock(&qi->q_lock);
+	spin_unlock_irqrestore(&qi->q_lock, flags);
 }
 
 /*
@@ -645,6 +642,62 @@ void qi_global_iec(struct intel_iommu *iommu)
 	qi_submit_sync(&desc, iommu);
 }
 
+int qi_flush_context(struct intel_iommu *iommu, u16 did, u16 sid, u8 fm,
+		     u64 type, int non_present_entry_flush)
+{
+
+	struct qi_desc desc;
+
+	if (non_present_entry_flush) {
+		if (!cap_caching_mode(iommu->cap))
+			return 1;
+		else
+			did = 0;
+	}
+
+	desc.low = QI_CC_FM(fm) | QI_CC_SID(sid) | QI_CC_DID(did)
+			| QI_CC_GRAN(type) | QI_CC_TYPE;
+	desc.high = 0;
+
+	qi_submit_sync(&desc, iommu);
+
+	return 0;
+
+}
+
+int qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr,
+		   unsigned int size_order, u64 type,
+		   int non_present_entry_flush)
+{
+	u8 dw = 0, dr = 0;
+
+	struct qi_desc desc;
+	int ih = 0;
+
+	if (non_present_entry_flush) {
+		if (!cap_caching_mode(iommu->cap))
+			return 1;
+		else
+			did = 0;
+	}
+
+	if (cap_write_drain(iommu->cap))
+		dw = 1;
+
+	if (cap_read_drain(iommu->cap))
+		dr = 1;
+
+	desc.low = QI_IOTLB_DID(did) | QI_IOTLB_DR(dr) | QI_IOTLB_DW(dw)
+		| QI_IOTLB_GRAN(type) | QI_IOTLB_TYPE;
+	desc.high = QI_IOTLB_ADDR(addr) | QI_IOTLB_IH(ih)
+		| QI_IOTLB_AM(size_order);
+
+	qi_submit_sync(&desc, iommu);
+
+	return 0;
+
+}
+
 /*
  * Enable Queued Invalidation interface. This is a must to support
  * interrupt-remapping. Also used by DMA-remapping, which replaces

drivers/pci/intel-iommu.c

@@ -18,6 +18,7 @@
  * Author: Ashok Raj <ashok.raj@intel.com>
  * Author: Shaohua Li <shaohua.li@intel.com>
  * Author: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
+ * Author: Fenghua Yu <fenghua.yu@intel.com>
  */
 
 #include <linux/init.h>
@@ -35,11 +36,13 @@
 #include <linux/timer.h>
 #include <linux/iova.h>
 #include <linux/intel-iommu.h>
-#include <asm/proto.h> /* force_iommu in this header in x86-64*/
 #include <asm/cacheflush.h>
 #include <asm/iommu.h>
 #include "pci.h"
 
+#define ROOT_SIZE		VTD_PAGE_SIZE
+#define CONTEXT_SIZE		VTD_PAGE_SIZE
+
 #define IS_GFX_DEVICE(pdev) ((pdev->class >> 16) == PCI_BASE_CLASS_DISPLAY)
 #define IS_ISA_DEVICE(pdev) ((pdev->class >> 8) == PCI_CLASS_BRIDGE_ISA)
 
@@ -199,7 +202,7 @@ static struct context_entry * device_to_context_entry(struct intel_iommu *iommu,
 			spin_unlock_irqrestore(&iommu->lock, flags);
 			return NULL;
 		}
-		__iommu_flush_cache(iommu, (void *)context, PAGE_SIZE_4K);
+		__iommu_flush_cache(iommu, (void *)context, CONTEXT_SIZE);
 		phy_addr = virt_to_phys((void *)context);
 		set_root_value(root, phy_addr);
 		set_root_present(root);
@@ -345,7 +348,7 @@ static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr)
 				return NULL;
 			}
 			__iommu_flush_cache(domain->iommu, tmp_page,
-					PAGE_SIZE_4K);
+					PAGE_SIZE);
 			dma_set_pte_addr(*pte, virt_to_phys(tmp_page));
 			/*
 			 * high level table always sets r/w, last level page
@@ -408,13 +411,13 @@ static void dma_pte_clear_range(struct dmar_domain *domain, u64 start, u64 end)
 	start &= (((u64)1) << addr_width) - 1;
 	end &= (((u64)1) << addr_width) - 1;
 	/* in case it's partial page */
-	start = PAGE_ALIGN_4K(start);
-	end &= PAGE_MASK_4K;
+	start = PAGE_ALIGN(start);
+	end &= PAGE_MASK;
 
 	/* we don't need lock here, nobody else touches the iova range */
 	while (start < end) {
 		dma_pte_clear_one(domain, start);
-		start += PAGE_SIZE_4K;
+		start += VTD_PAGE_SIZE;
 	}
 }
 
@@ -468,7 +471,7 @@ static int iommu_alloc_root_entry(struct intel_iommu *iommu)
 	if (!root)
 		return -ENOMEM;
 
-	__iommu_flush_cache(iommu, root, PAGE_SIZE_4K);
+	__iommu_flush_cache(iommu, root, ROOT_SIZE);
 
 	spin_lock_irqsave(&iommu->lock, flags);
 	iommu->root_entry = root;
@@ -567,27 +570,6 @@ static int __iommu_flush_context(struct intel_iommu *iommu,
 	return 0;
 }
 
-static int inline iommu_flush_context_global(struct intel_iommu *iommu,
-	int non_present_entry_flush)
-{
-	return __iommu_flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL,
-		non_present_entry_flush);
-}
-
-static int inline iommu_flush_context_domain(struct intel_iommu *iommu, u16 did,
-	int non_present_entry_flush)
-{
-	return __iommu_flush_context(iommu, did, 0, 0, DMA_CCMD_DOMAIN_INVL,
-		non_present_entry_flush);
-}
-
-static int inline iommu_flush_context_device(struct intel_iommu *iommu,
-	u16 did, u16 source_id, u8 function_mask, int non_present_entry_flush)
-{
-	return __iommu_flush_context(iommu, did, source_id, function_mask,
-		DMA_CCMD_DEVICE_INVL, non_present_entry_flush);
-}
-
 /* return value determine if we need a write buffer flush */
 static int __iommu_flush_iotlb(struct intel_iommu *iommu, u16 did,
 	u64 addr, unsigned int size_order, u64 type,
@@ -655,37 +637,25 @@ static int __iommu_flush_iotlb(struct intel_iommu *iommu, u16 did,
 		printk(KERN_ERR"IOMMU: flush IOTLB failed\n");
 	if (DMA_TLB_IAIG(val) != DMA_TLB_IIRG(type))
 		pr_debug("IOMMU: tlb flush request %Lx, actual %Lx\n",
-			DMA_TLB_IIRG(type), DMA_TLB_IAIG(val));
+			(unsigned long long)DMA_TLB_IIRG(type),
+			(unsigned long long)DMA_TLB_IAIG(val));
 	/* flush iotlb entry will implicitly flush write buffer */
 	return 0;
 }
 
-static int inline iommu_flush_iotlb_global(struct intel_iommu *iommu,
-	int non_present_entry_flush)
-{
-	return __iommu_flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH,
-		non_present_entry_flush);
-}
-
-static int inline iommu_flush_iotlb_dsi(struct intel_iommu *iommu, u16 did,
-	int non_present_entry_flush)
-{
-	return __iommu_flush_iotlb(iommu, did, 0, 0, DMA_TLB_DSI_FLUSH,
-		non_present_entry_flush);
-}
-
 static int iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
 	u64 addr, unsigned int pages, int non_present_entry_flush)
 {
 	unsigned int mask;
 
-	BUG_ON(addr & (~PAGE_MASK_4K));
+	BUG_ON(addr & (~VTD_PAGE_MASK));
 	BUG_ON(pages == 0);
 
 	/* Fallback to domain selective flush if no PSI support */
 	if (!cap_pgsel_inv(iommu->cap))
-		return iommu_flush_iotlb_dsi(iommu, did,
-			non_present_entry_flush);
+		return iommu->flush.flush_iotlb(iommu, did, 0, 0,
+						DMA_TLB_DSI_FLUSH,
+						non_present_entry_flush);
 
 	/*
 	 * PSI requires page size to be 2 ^ x, and the base address is naturally
@@ -694,11 +664,12 @@ static int iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
 	mask = ilog2(__roundup_pow_of_two(pages));
 	/* Fallback to domain selective flush if size is too big */
 	if (mask > cap_max_amask_val(iommu->cap))
-		return iommu_flush_iotlb_dsi(iommu, did,
-			non_present_entry_flush);
+		return iommu->flush.flush_iotlb(iommu, did, 0, 0,
+			DMA_TLB_DSI_FLUSH, non_present_entry_flush);
 
-	return __iommu_flush_iotlb(iommu, did, addr, mask,
-		DMA_TLB_PSI_FLUSH, non_present_entry_flush);
+	return iommu->flush.flush_iotlb(iommu, did, addr, mask,
+					DMA_TLB_PSI_FLUSH,
+					non_present_entry_flush);
 }
 
 static void iommu_disable_protect_mem_regions(struct intel_iommu *iommu)
@@ -831,7 +802,7 @@ void dmar_msi_read(int irq, struct msi_msg *msg)
 }
 
 static int iommu_page_fault_do_one(struct intel_iommu *iommu, int type,
-		u8 fault_reason, u16 source_id, u64 addr)
+		u8 fault_reason, u16 source_id, unsigned long long addr)
 {
 	const char *reason;
 
@@ -1084,9 +1055,9 @@ static void dmar_init_reserved_ranges(void)
 			if (!r->flags || !(r->flags & IORESOURCE_MEM))
 				continue;
 			addr = r->start;
-			addr &= PAGE_MASK_4K;
+			addr &= PAGE_MASK;
 			size = r->end - addr;
-			size = PAGE_ALIGN_4K(size);
+			size = PAGE_ALIGN(size);
 			iova = reserve_iova(&reserved_iova_list, IOVA_PFN(addr),
 				IOVA_PFN(size + addr) - 1);
 			if (!iova)
@@ -1148,7 +1119,7 @@ static int domain_init(struct dmar_domain *domain, int guest_width)
 	domain->pgd = (struct dma_pte *)alloc_pgtable_page();
 	if (!domain->pgd)
 		return -ENOMEM;
-	__iommu_flush_cache(iommu, domain->pgd, PAGE_SIZE_4K);
+	__iommu_flush_cache(iommu, domain->pgd, PAGE_SIZE);
 	return 0;
 }
 
@@ -1164,7 +1135,7 @@ static void domain_exit(struct dmar_domain *domain)
 	/* destroy iovas */
 	put_iova_domain(&domain->iovad);
 	end = DOMAIN_MAX_ADDR(domain->gaw);
-	end = end & (~PAGE_MASK_4K);
+	end = end & (~PAGE_MASK);
 
 	/* clear ptes */
 	dma_pte_clear_range(domain, 0, end);
@@ -1204,11 +1175,13 @@ static int domain_context_mapping_one(struct dmar_domain *domain,
 	__iommu_flush_cache(iommu, context, sizeof(*context));
 
 	/* it's a non-present to present mapping */
-	if (iommu_flush_context_device(iommu, domain->id,
-			(((u16)bus) << 8) | devfn, DMA_CCMD_MASK_NOBIT, 1))
+	if (iommu->flush.flush_context(iommu, domain->id,
+		(((u16)bus) << 8) | devfn, DMA_CCMD_MASK_NOBIT,
+		DMA_CCMD_DEVICE_INVL, 1))
 		iommu_flush_write_buffer(iommu);
 	else
-		iommu_flush_iotlb_dsi(iommu, 0, 0);
+		iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_DSI_FLUSH, 0);
+
 	spin_unlock_irqrestore(&iommu->lock, flags);
 	return 0;
 }
@@ -1283,22 +1256,25 @@ domain_page_mapping(struct dmar_domain *domain, dma_addr_t iova,
 	u64 start_pfn, end_pfn;
 	struct dma_pte *pte;
 	int index;
+	int addr_width = agaw_to_width(domain->agaw);
+
+	hpa &= (((u64)1) << addr_width) - 1;
 
 	if ((prot & (DMA_PTE_READ|DMA_PTE_WRITE)) == 0)
 		return -EINVAL;
-	iova &= PAGE_MASK_4K;
-	start_pfn = ((u64)hpa) >> PAGE_SHIFT_4K;
-	end_pfn = (PAGE_ALIGN_4K(((u64)hpa) + size)) >> PAGE_SHIFT_4K;
+	iova &= PAGE_MASK;
+	start_pfn = ((u64)hpa) >> VTD_PAGE_SHIFT;
+	end_pfn = (VTD_PAGE_ALIGN(((u64)hpa) + size)) >> VTD_PAGE_SHIFT;
 	index = 0;
 	while (start_pfn < end_pfn) {
-		pte = addr_to_dma_pte(domain, iova + PAGE_SIZE_4K * index);
+		pte = addr_to_dma_pte(domain, iova + VTD_PAGE_SIZE * index);
 		if (!pte)
 			return -ENOMEM;
 		/* We don't need lock here, nobody else
 		 * touches the iova range
 		 */
 		BUG_ON(dma_pte_addr(*pte));
-		dma_set_pte_addr(*pte, start_pfn << PAGE_SHIFT_4K);
+		dma_set_pte_addr(*pte, start_pfn << VTD_PAGE_SHIFT);
 		dma_set_pte_prot(*pte, prot);
 		__iommu_flush_cache(domain->iommu, pte, sizeof(*pte));
 		start_pfn++;
@@ -1310,8 +1286,10 @@ domain_page_mapping(struct dmar_domain *domain, dma_addr_t iova,
 static void detach_domain_for_dev(struct dmar_domain *domain, u8 bus, u8 devfn)
 {
 	clear_context_table(domain->iommu, bus, devfn);
-	iommu_flush_context_global(domain->iommu, 0);
-	iommu_flush_iotlb_global(domain->iommu, 0);
+	domain->iommu->flush.flush_context(domain->iommu, 0, 0, 0,
+					   DMA_CCMD_GLOBAL_INVL, 0);
+	domain->iommu->flush.flush_iotlb(domain->iommu, 0, 0, 0,
+					 DMA_TLB_GLOBAL_FLUSH, 0);
 }
 
 static void domain_remove_dev_info(struct dmar_domain *domain)
@@ -1474,11 +1452,13 @@ static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw)
 	return find_domain(pdev);
 }
 
-static int iommu_prepare_identity_map(struct pci_dev *pdev, u64 start, u64 end)
+static int iommu_prepare_identity_map(struct pci_dev *pdev,
+				      unsigned long long start,
+				      unsigned long long end)
 {
 	struct dmar_domain *domain;
 	unsigned long size;
-	u64 base;
+	unsigned long long base;
 	int ret;
 
 	printk(KERN_INFO
@@ -1490,9 +1470,9 @@ static int iommu_prepare_identity_map(struct pci_dev *pdev, u64 start, u64 end)
 		return -ENOMEM;
 
 	/* The address might not be aligned */
-	base = start & PAGE_MASK_4K;
+	base = start & PAGE_MASK;
 	size = end - base;
-	size = PAGE_ALIGN_4K(size);
+	size = PAGE_ALIGN(size);
 	if (!reserve_iova(&domain->iovad, IOVA_PFN(base),
 			IOVA_PFN(base + size) - 1)) {
 		printk(KERN_ERR "IOMMU: reserve iova failed\n");
@@ -1662,6 +1642,28 @@ int __init init_dmars(void)
 		}
 	}
 
+	for_each_drhd_unit(drhd) {
+		if (drhd->ignored)
+			continue;
+
+		iommu = drhd->iommu;
+		if (dmar_enable_qi(iommu)) {
+			/*
+			 * Queued Invalidate not enabled, use Register Based
+			 * Invalidate
+			 */
+			iommu->flush.flush_context = __iommu_flush_context;
+			iommu->flush.flush_iotlb = __iommu_flush_iotlb;
+			printk(KERN_INFO "IOMMU 0x%Lx: using Register based "
+			       "invalidation\n", drhd->reg_base_addr);
+		} else {
+			iommu->flush.flush_context = qi_flush_context;
+			iommu->flush.flush_iotlb = qi_flush_iotlb;
+			printk(KERN_INFO "IOMMU 0x%Lx: using Queued "
+			       "invalidation\n", drhd->reg_base_addr);
+		}
+	}
+
 	/*
 	 * For each rmrr
 	 *   for each dev attached to rmrr
@@ -1714,9 +1716,10 @@ int __init init_dmars(void)
 
 		iommu_set_root_entry(iommu);
 
-		iommu_flush_context_global(iommu, 0);
-		iommu_flush_iotlb_global(iommu, 0);
-
+		iommu->flush.flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL,
+					   0);
+		iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH,
+					 0);
 		iommu_disable_protect_mem_regions(iommu);
 
 		ret = iommu_enable_translation(iommu);
@@ -1738,8 +1741,8 @@ int __init init_dmars(void)
 static inline u64 aligned_size(u64 host_addr, size_t size)
 {
 	u64 addr;
-	addr = (host_addr & (~PAGE_MASK_4K)) + size;
-	return PAGE_ALIGN_4K(addr);
+	addr = (host_addr & (~PAGE_MASK)) + size;
+	return PAGE_ALIGN(addr);
 }
 
 struct iova *
@@ -1753,20 +1756,20 @@ iommu_alloc_iova(struct dmar_domain *domain, size_t size, u64 end)
 		return NULL;
 
 	piova = alloc_iova(&domain->iovad,
-			size >> PAGE_SHIFT_4K, IOVA_PFN(end), 1);
+			size >> PAGE_SHIFT, IOVA_PFN(end), 1);
 	return piova;
 }
 
 static struct iova *
 __intel_alloc_iova(struct device *dev, struct dmar_domain *domain,
-		size_t size)
+		   size_t size, u64 dma_mask)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct iova *iova = NULL;
 
-	if ((pdev->dma_mask <= DMA_32BIT_MASK) || (dmar_forcedac)) {
-		iova = iommu_alloc_iova(domain, size, pdev->dma_mask);
-	} else  {
+	if (dma_mask <= DMA_32BIT_MASK || dmar_forcedac)
+		iova = iommu_alloc_iova(domain, size, dma_mask);
+	else {
 		/*
 		 * First try to allocate an io virtual address in
 		 * DMA_32BIT_MASK and if that fails then try allocating
@@ -1774,7 +1777,7 @@ __intel_alloc_iova(struct device *dev, struct dmar_domain *domain,
 		 */
 		iova = iommu_alloc_iova(domain, size, DMA_32BIT_MASK);
 		if (!iova)
-			iova = iommu_alloc_iova(domain, size, pdev->dma_mask);
+			iova = iommu_alloc_iova(domain, size, dma_mask);
 	}
 
 	if (!iova) {
@@ -1813,12 +1816,12 @@ get_valid_domain_for_dev(struct pci_dev *pdev)
 	return domain;
 }
 
-static dma_addr_t
-intel_map_single(struct device *hwdev, phys_addr_t paddr, size_t size, int dir)
+static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
+				     size_t size, int dir, u64 dma_mask)
 {
 	struct pci_dev *pdev = to_pci_dev(hwdev);
 	struct dmar_domain *domain;
-	unsigned long start_paddr;
+	phys_addr_t start_paddr;
 	struct iova *iova;
 	int prot = 0;
 	int ret;
@@ -1833,11 +1836,11 @@ intel_map_single(struct device *hwdev, phys_addr_t paddr, size_t size, int dir)
 
 	size = aligned_size((u64)paddr, size);
 
-	iova = __intel_alloc_iova(hwdev, domain, size);
+	iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
 	if (!iova)
 		goto error;
 
-	start_paddr = iova->pfn_lo << PAGE_SHIFT_4K;
+	start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
 
 	/*
 	 * Check if DMAR supports zero-length reads on write only
@@ -1855,30 +1858,33 @@ intel_map_single(struct device *hwdev, phys_addr_t paddr, size_t size, int dir)
 	 * is not a big problem
 	 */
 	ret = domain_page_mapping(domain, start_paddr,
-		((u64)paddr) & PAGE_MASK_4K, size, prot);
+		((u64)paddr) & PAGE_MASK, size, prot);
 	if (ret)
 		goto error;
 
-	pr_debug("Device %s request: %lx@%llx mapping: %lx@%llx, dir %d\n",
-		pci_name(pdev), size, (u64)paddr,
-		size, (u64)start_paddr, dir);
-
 	/* it's a non-present to present mapping */
 	ret = iommu_flush_iotlb_psi(domain->iommu, domain->id,
-			start_paddr, size >> PAGE_SHIFT_4K, 1);
+			start_paddr, size >> VTD_PAGE_SHIFT, 1);
 	if (ret)
 		iommu_flush_write_buffer(domain->iommu);
 
-	return (start_paddr + ((u64)paddr & (~PAGE_MASK_4K)));
+	return start_paddr + ((u64)paddr & (~PAGE_MASK));
 
 error:
 	if (iova)
 		__free_iova(&domain->iovad, iova);
 	printk(KERN_ERR"Device %s request: %lx@%llx dir %d --- failed\n",
-		pci_name(pdev), size, (u64)paddr, dir);
+		pci_name(pdev), size, (unsigned long long)paddr, dir);
 	return 0;
 }
 
+dma_addr_t intel_map_single(struct device *hwdev, phys_addr_t paddr,
+			    size_t size, int dir)
+{
+	return __intel_map_single(hwdev, paddr, size, dir,
+				  to_pci_dev(hwdev)->dma_mask);
+}
+
 static void flush_unmaps(void)
 {
 	int i, j;
@@ -1891,7 +1897,8 @@ static void flush_unmaps(void)
 			struct intel_iommu *iommu =
 				deferred_flush[i].domain[0]->iommu;
 
-			iommu_flush_iotlb_global(iommu, 0);
+			iommu->flush.flush_iotlb(iommu, 0, 0, 0,
+						 DMA_TLB_GLOBAL_FLUSH, 0);
 			for (j = 0; j < deferred_flush[i].next; j++) {
 				__free_iova(&deferred_flush[i].domain[j]->iovad,
 						deferred_flush[i].iova[j]);
@@ -1936,8 +1943,8 @@ static void add_unmap(struct dmar_domain *dom, struct iova *iova)
 	spin_unlock_irqrestore(&async_umap_flush_lock, flags);
 }
 
-static void intel_unmap_single(struct device *dev, dma_addr_t dev_addr,
-	size_t size, int dir)
+void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t size,
+			int dir)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct dmar_domain *domain;
@@ -1953,11 +1960,11 @@ static void intel_unmap_single(struct device *dev, dma_addr_t dev_addr,
 	if (!iova)
 		return;
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT_4K;
+	start_addr = iova->pfn_lo << PAGE_SHIFT;
 	size = aligned_size((u64)dev_addr, size);
 
 	pr_debug("Device %s unmapping: %lx@%llx\n",
-		pci_name(pdev), size, (u64)start_addr);
+		pci_name(pdev), size, (unsigned long long)start_addr);
 
 	/*  clear the whole page */
 	dma_pte_clear_range(domain, start_addr, start_addr + size);
@@ -1965,7 +1972,7 @@ static void intel_unmap_single(struct device *dev, dma_addr_t dev_addr,
 	dma_pte_free_pagetable(domain, start_addr, start_addr + size);
 	if (intel_iommu_strict) {
 		if (iommu_flush_iotlb_psi(domain->iommu,
-			domain->id, start_addr, size >> PAGE_SHIFT_4K, 0))
+			domain->id, start_addr, size >> VTD_PAGE_SHIFT, 0))
 			iommu_flush_write_buffer(domain->iommu);
 		/* free iova */
 		__free_iova(&domain->iovad, iova);
@@ -1978,13 +1985,13 @@ static void intel_unmap_single(struct device *dev, dma_addr_t dev_addr,
 	}
 }
 
-static void * intel_alloc_coherent(struct device *hwdev, size_t size,
-		       dma_addr_t *dma_handle, gfp_t flags)
+void *intel_alloc_coherent(struct device *hwdev, size_t size,
+			   dma_addr_t *dma_handle, gfp_t flags)
 {
 	void *vaddr;
 	int order;
 
-	size = PAGE_ALIGN_4K(size);
+	size = PAGE_ALIGN(size);
 	order = get_order(size);
 	flags &= ~(GFP_DMA | GFP_DMA32);
 
@@ -1993,19 +2000,21 @@ static void * intel_alloc_coherent(struct device *hwdev, size_t size,
 		return NULL;
 	memset(vaddr, 0, size);
 
-	*dma_handle = intel_map_single(hwdev, virt_to_bus(vaddr), size, DMA_BIDIRECTIONAL);
+	*dma_handle = __intel_map_single(hwdev, virt_to_bus(vaddr), size,
+					 DMA_BIDIRECTIONAL,
+					 hwdev->coherent_dma_mask);
 	if (*dma_handle)
 		return vaddr;
 	free_pages((unsigned long)vaddr, order);
 	return NULL;
 }
 
-static void intel_free_coherent(struct device *hwdev, size_t size,
-	void *vaddr, dma_addr_t dma_handle)
+void intel_free_coherent(struct device *hwdev, size_t size, void *vaddr,
+			 dma_addr_t dma_handle)
 {
 	int order;
 
-	size = PAGE_ALIGN_4K(size);
+	size = PAGE_ALIGN(size);
 	order = get_order(size);
 
 	intel_unmap_single(hwdev, dma_handle, size, DMA_BIDIRECTIONAL);
@@ -2013,8 +2022,9 @@ static void intel_free_coherent(struct device *hwdev, size_t size,
 }
 
 #define SG_ENT_VIRT_ADDRESS(sg)	(sg_virt((sg)))
-static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
-	int nelems, int dir)
+
+void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
+		    int nelems, int dir)
 {
 	int i;
 	struct pci_dev *pdev = to_pci_dev(hwdev);
@@ -2038,7 +2048,7 @@ static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
 		size += aligned_size((u64)addr, sg->length);
 	}
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT_4K;
+	start_addr = iova->pfn_lo << PAGE_SHIFT;
 
 	/*  clear the whole page */
 	dma_pte_clear_range(domain, start_addr, start_addr + size);
@@ -2046,7 +2056,7 @@ static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
 	dma_pte_free_pagetable(domain, start_addr, start_addr + size);
 
 	if (iommu_flush_iotlb_psi(domain->iommu, domain->id, start_addr,
-			size >> PAGE_SHIFT_4K, 0))
+			size >> VTD_PAGE_SHIFT, 0))
 		iommu_flush_write_buffer(domain->iommu);
 
 	/* free iova */
@@ -2067,8 +2077,8 @@ static int intel_nontranslate_map_sg(struct device *hddev,
 	return nelems;
 }
 
-static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist,
-				int nelems, int dir)
+int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
+		 int dir)
 {
 	void *addr;
 	int i;
@@ -2096,7 +2106,7 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist,
 		size += aligned_size((u64)addr, sg->length);
 	}
 
-	iova = __intel_alloc_iova(hwdev, domain, size);
+	iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
 	if (!iova) {
 		sglist->dma_length = 0;
 		return 0;
@@ -2112,14 +2122,14 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist,
 	if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
 		prot |= DMA_PTE_WRITE;
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT_4K;
+	start_addr = iova->pfn_lo << PAGE_SHIFT;
 	offset = 0;
 	for_each_sg(sglist, sg, nelems, i) {
 		addr = SG_ENT_VIRT_ADDRESS(sg);
 		addr = (void *)virt_to_phys(addr);
 		size = aligned_size((u64)addr, sg->length);
 		ret = domain_page_mapping(domain, start_addr + offset,
-			((u64)addr) & PAGE_MASK_4K,
+			((u64)addr) & PAGE_MASK,
 			size, prot);
 		if (ret) {
 			/*  clear the page */
@@ -2133,14 +2143,14 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist,
 			return 0;
 		}
 		sg->dma_address = start_addr + offset +
-				((u64)addr & (~PAGE_MASK_4K));
+				((u64)addr & (~PAGE_MASK));
 		sg->dma_length = sg->length;
 		offset += size;
 	}
 
 	/* it's a non-present to present mapping */
 	if (iommu_flush_iotlb_psi(domain->iommu, domain->id,
-			start_addr, offset >> PAGE_SHIFT_4K, 1))
+			start_addr, offset >> VTD_PAGE_SHIFT, 1))
 		iommu_flush_write_buffer(domain->iommu);
 	return nelems;
 }
@@ -2180,7 +2190,6 @@ static inline int iommu_devinfo_cache_init(void)
 					 sizeof(struct device_domain_info),
 					 0,
 					 SLAB_HWCACHE_ALIGN,
-
 					 NULL);
 	if (!iommu_devinfo_cache) {
 		printk(KERN_ERR "Couldn't create devinfo cache\n");
@@ -2198,7 +2207,6 @@ static inline int iommu_iova_cache_init(void)
 					 sizeof(struct iova),
 					 0,
 					 SLAB_HWCACHE_ALIGN,
-
 					 NULL);
 	if (!iommu_iova_cache) {
 		printk(KERN_ERR "Couldn't create iova cache\n");
@@ -2327,7 +2335,7 @@ void intel_iommu_domain_exit(struct dmar_domain *domain)
 		return;
 
 	end = DOMAIN_MAX_ADDR(domain->gaw);
-	end = end & (~PAGE_MASK_4K);
+	end = end & (~VTD_PAGE_MASK);
 
 	/* clear ptes */
 	dma_pte_clear_range(domain, 0, end);
@@ -2423,6 +2431,6 @@ u64 intel_iommu_iova_to_pfn(struct dmar_domain *domain, u64 iova)
 	if (pte)
 		pfn = dma_pte_addr(*pte);
 
-	return pfn >> PAGE_SHIFT_4K;
+	return pfn >> VTD_PAGE_SHIFT;
 }
 EXPORT_SYMBOL_GPL(intel_iommu_iova_to_pfn);

drivers/pci/quirks.c

@@ -43,6 +43,20 @@ static void __devinit quirk_mellanox_tavor(struct pci_dev *dev)
 DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_MELLANOX,PCI_DEVICE_ID_MELLANOX_TAVOR,quirk_mellanox_tavor);
 DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_MELLANOX,PCI_DEVICE_ID_MELLANOX_TAVOR_BRIDGE,quirk_mellanox_tavor);
 
+/* Many VIA bridges seem to corrupt data for DAC. Disable it here */
+int forbid_dac __read_mostly;
+EXPORT_SYMBOL(forbid_dac);
+
+static __devinit void via_no_dac(struct pci_dev *dev)
+{
+	if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI && forbid_dac == 0) {
+		dev_info(&dev->dev,
+			"VIA PCI bridge detected. Disabling DAC.\n");
+		forbid_dac = 1;
+	}
+}
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_VIA, PCI_ANY_ID, via_no_dac);
+
 /* Deal with broken BIOS'es that neglect to enable passive release,
    which can cause problems in combination with the 82441FX/PPro MTRRs */
 static void quirk_passive_release(struct pci_dev *dev)

include/asm-x86/iommu.h

@@ -7,9 +7,13 @@ extern struct dma_mapping_ops nommu_dma_ops;
 extern int force_iommu, no_iommu;
 extern int iommu_detected;
 extern int dmar_disabled;
+extern int forbid_dac;
 
 extern unsigned long iommu_nr_pages(unsigned long addr, unsigned long len);
 
+/* 10 seconds */
+#define DMAR_OPERATION_TIMEOUT ((cycles_t) tsc_khz*10*1000)
+
 #ifdef CONFIG_GART_IOMMU
 extern int gart_iommu_aperture;
 extern int gart_iommu_aperture_allowed;

include/linux/dma_remapping.h

@@ -2,15 +2,14 @@
 #define _DMA_REMAPPING_H
 
 /*
- * We need a fixed PAGE_SIZE of 4K irrespective of
- * arch PAGE_SIZE for IOMMU page tables.
+ * VT-d hardware uses 4KiB page size regardless of host page size.
  */
-#define PAGE_SHIFT_4K		(12)
-#define PAGE_SIZE_4K		(1UL << PAGE_SHIFT_4K)
-#define PAGE_MASK_4K		(((u64)-1) << PAGE_SHIFT_4K)
-#define PAGE_ALIGN_4K(addr)	(((addr) + PAGE_SIZE_4K - 1) & PAGE_MASK_4K)
+#define VTD_PAGE_SHIFT		(12)
+#define VTD_PAGE_SIZE		(1UL << VTD_PAGE_SHIFT)
+#define VTD_PAGE_MASK		(((u64)-1) << VTD_PAGE_SHIFT)
+#define VTD_PAGE_ALIGN(addr)	(((addr) + VTD_PAGE_SIZE - 1) & VTD_PAGE_MASK)
 
-#define IOVA_PFN(addr)		((addr) >> PAGE_SHIFT_4K)
+#define IOVA_PFN(addr)		((addr) >> PAGE_SHIFT)
 #define DMA_32BIT_PFN		IOVA_PFN(DMA_32BIT_MASK)
 #define DMA_64BIT_PFN		IOVA_PFN(DMA_64BIT_MASK)
 
@@ -25,7 +24,7 @@ struct root_entry {
 	u64	val;
 	u64	rsvd1;
 };
-#define ROOT_ENTRY_NR (PAGE_SIZE_4K/sizeof(struct root_entry))
+#define ROOT_ENTRY_NR (VTD_PAGE_SIZE/sizeof(struct root_entry))
 static inline bool root_present(struct root_entry *root)
 {
 	return (root->val & 1);
@@ -36,7 +35,7 @@ static inline void set_root_present(struct root_entry *root)
 }
 static inline void set_root_value(struct root_entry *root, unsigned long value)
 {
-	root->val |= value & PAGE_MASK_4K;
+	root->val |= value & VTD_PAGE_MASK;
 }
 
 struct context_entry;
@@ -45,7 +44,7 @@ get_context_addr_from_root(struct root_entry *root)
 {
 	return (struct context_entry *)
 		(root_present(root)?phys_to_virt(
-		root->val & PAGE_MASK_4K):
+		root->val & VTD_PAGE_MASK) :
 		NULL);
 }
 
@@ -67,7 +66,7 @@ struct context_entry {
 #define context_present(c) ((c).lo & 1)
 #define context_fault_disable(c) (((c).lo >> 1) & 1)
 #define context_translation_type(c) (((c).lo >> 2) & 3)
-#define context_address_root(c) ((c).lo & PAGE_MASK_4K)
+#define context_address_root(c) ((c).lo & VTD_PAGE_MASK)
 #define context_address_width(c) ((c).hi &  7)
 #define context_domain_id(c) (((c).hi >> 8) & ((1 << 16) - 1))
 
@@ -81,7 +80,7 @@ struct context_entry {
 	} while (0)
 #define CONTEXT_TT_MULTI_LEVEL 0
 #define context_set_address_root(c, val) \
-	do {(c).lo |= (val) & PAGE_MASK_4K;} while (0)
+	do {(c).lo |= (val) & VTD_PAGE_MASK; } while (0)
 #define context_set_address_width(c, val) do {(c).hi |= (val) & 7;} while (0)
 #define context_set_domain_id(c, val) \
 	do {(c).hi |= ((val) & ((1 << 16) - 1)) << 8;} while (0)
@@ -107,9 +106,9 @@ struct dma_pte {
 #define dma_set_pte_writable(p) do {(p).val |= DMA_PTE_WRITE;} while (0)
 #define dma_set_pte_prot(p, prot) \
 		do {(p).val = ((p).val & ~3) | ((prot) & 3); } while (0)
-#define dma_pte_addr(p) ((p).val & PAGE_MASK_4K)
+#define dma_pte_addr(p) ((p).val & VTD_PAGE_MASK)
 #define dma_set_pte_addr(p, addr) do {\
-		(p).val |= ((addr) & PAGE_MASK_4K); } while (0)
+		(p).val |= ((addr) & VTD_PAGE_MASK); } while (0)
 #define dma_pte_present(p) (((p).val & 3) != 0)
 
 struct intel_iommu;

include/linux/intel-iommu.h

@@ -29,6 +29,7 @@
 #include <linux/io.h>
 #include <linux/dma_remapping.h>
 #include <asm/cacheflush.h>
+#include <asm/iommu.h>
 
 /*
  * Intel IOMMU register specification per version 1.0 public spec.
@@ -127,6 +128,7 @@ static inline void dmar_writeq(void __iomem *addr, u64 val)
 
 
 /* IOTLB_REG */
+#define DMA_TLB_FLUSH_GRANU_OFFSET  60
 #define DMA_TLB_GLOBAL_FLUSH (((u64)1) << 60)
 #define DMA_TLB_DSI_FLUSH (((u64)2) << 60)
 #define DMA_TLB_PSI_FLUSH (((u64)3) << 60)
@@ -140,6 +142,7 @@ static inline void dmar_writeq(void __iomem *addr, u64 val)
 #define DMA_TLB_MAX_SIZE (0x3f)
 
 /* INVALID_DESC */
+#define DMA_CCMD_INVL_GRANU_OFFSET  61
 #define DMA_ID_TLB_GLOBAL_FLUSH	(((u64)1) << 3)
 #define DMA_ID_TLB_DSI_FLUSH	(((u64)2) << 3)
 #define DMA_ID_TLB_PSI_FLUSH	(((u64)3) << 3)
@@ -200,22 +203,21 @@ static inline void dmar_writeq(void __iomem *addr, u64 val)
 #define dma_frcd_type(d) ((d >> 30) & 1)
 #define dma_frcd_fault_reason(c) (c & 0xff)
 #define dma_frcd_source_id(c) (c & 0xffff)
-#define dma_frcd_page_addr(d) (d & (((u64)-1) << 12)) /* low 64 bit */
-
-#define DMAR_OPERATION_TIMEOUT ((cycles_t) tsc_khz*10*1000) /* 10sec */
-
-#define IOMMU_WAIT_OP(iommu, offset, op, cond, sts) \
-{\
-	cycles_t start_time = get_cycles();\
-	while (1) {\
-		sts = op (iommu->reg + offset);\
-		if (cond)\
-			break;\
+/* low 64 bit */
+#define dma_frcd_page_addr(d) (d & (((u64)-1) << PAGE_SHIFT))
+
+#define IOMMU_WAIT_OP(iommu, offset, op, cond, sts)			\
+do {									\
+	cycles_t start_time = get_cycles();				\
+	while (1) {							\
+		sts = op(iommu->reg + offset);				\
+		if (cond)						\
+			break;						\
 		if (DMAR_OPERATION_TIMEOUT < (get_cycles() - start_time))\
-			panic("DMAR hardware is malfunctioning\n");\
-		cpu_relax();\
-	}\
-}
+			panic("DMAR hardware is malfunctioning\n");	\
+		cpu_relax();						\
+	}								\
+} while (0)
 
 #define QI_LENGTH	256	/* queue length */
 
@@ -238,6 +240,19 @@ enum {
 #define QI_IWD_STATUS_DATA(d)	(((u64)d) << 32)
 #define QI_IWD_STATUS_WRITE	(((u64)1) << 5)
 
+#define QI_IOTLB_DID(did) 	(((u64)did) << 16)
+#define QI_IOTLB_DR(dr) 	(((u64)dr) << 7)
+#define QI_IOTLB_DW(dw) 	(((u64)dw) << 6)
+#define QI_IOTLB_GRAN(gran) 	(((u64)gran) >> (DMA_TLB_FLUSH_GRANU_OFFSET-4))
+#define QI_IOTLB_ADDR(addr)	(((u64)addr) & VTD_PAGE_MASK)
+#define QI_IOTLB_IH(ih)		(((u64)ih) << 6)
+#define QI_IOTLB_AM(am)		(((u8)am))
+
+#define QI_CC_FM(fm)		(((u64)fm) << 48)
+#define QI_CC_SID(sid)		(((u64)sid) << 32)
+#define QI_CC_DID(did)		(((u64)did) << 16)
+#define QI_CC_GRAN(gran)	(((u64)gran) >> (DMA_CCMD_INVL_GRANU_OFFSET-4))
+
 struct qi_desc {
 	u64 low, high;
 };
@@ -263,6 +278,13 @@ struct ir_table {
 };
 #endif
 
+struct iommu_flush {
+	int (*flush_context)(struct intel_iommu *iommu, u16 did, u16 sid, u8 fm,
+		u64 type, int non_present_entry_flush);
+	int (*flush_iotlb)(struct intel_iommu *iommu, u16 did, u64 addr,
+		unsigned int size_order, u64 type, int non_present_entry_flush);
+};
+
 struct intel_iommu {
 	void __iomem	*reg; /* Pointer to hardware regs, virtual addr */
 	u64		cap;
@@ -282,6 +304,7 @@ struct intel_iommu {
 	unsigned char name[7];    /* Device Name */
 	struct msi_msg saved_msg;
 	struct sys_device sysdev;
+	struct iommu_flush flush;
 #endif
 	struct q_inval  *qi;            /* Queued invalidation info */
 #ifdef CONFIG_INTR_REMAP
@@ -303,6 +326,12 @@ extern void free_iommu(struct intel_iommu *iommu);
 extern int dmar_enable_qi(struct intel_iommu *iommu);
 extern void qi_global_iec(struct intel_iommu *iommu);
 
+extern int qi_flush_context(struct intel_iommu *iommu, u16 did, u16 sid,
+			        u8 fm, u64 type, int non_present_entry_flush);
+extern int qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr,
+			  unsigned int size_order, u64 type,
+			  int non_present_entry_flush);
+
 extern void qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu);
 
 void intel_iommu_domain_exit(struct dmar_domain *domain);
@@ -324,4 +353,11 @@ static inline int intel_iommu_found(void)
 }
 #endif /* CONFIG_DMAR */
 
+extern void *intel_alloc_coherent(struct device *, size_t, dma_addr_t *, gfp_t);
+extern void intel_free_coherent(struct device *, size_t, void *, dma_addr_t);
+extern dma_addr_t intel_map_single(struct device *, phys_addr_t, size_t, int);
+extern void intel_unmap_single(struct device *, dma_addr_t, size_t, int);
+extern int intel_map_sg(struct device *, struct scatterlist *, int, int);
+extern void intel_unmap_sg(struct device *, struct scatterlist *, int, int);
+
 #endif