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

* git://git.infradead.org/iommu-2.6: (38 commits)
  intel-iommu: Don't keep freeing page zero in dma_pte_free_pagetable()
  intel-iommu: Introduce first_pte_in_page() to simplify PTE-setting loops
  intel-iommu: Use cmpxchg64_local() for setting PTEs
  intel-iommu: Warn about unmatched unmap requests
  intel-iommu: Kill superfluous mapping_lock
  intel-iommu: Ensure that PTE writes are 64-bit atomic, even on i386
  intel-iommu: Make iommu=pt work on i386 too
  intel-iommu: Performance improvement for dma_pte_free_pagetable()
  intel-iommu: Don't free too much in dma_pte_free_pagetable()
  intel-iommu: dump mappings but don't die on pte already set
  intel-iommu: Combine domain_pfn_mapping() and domain_sg_mapping()
  intel-iommu: Introduce domain_sg_mapping() to speed up intel_map_sg()
  intel-iommu: Simplify __intel_alloc_iova()
  intel-iommu: Performance improvement for domain_pfn_mapping()
  intel-iommu: Performance improvement for dma_pte_clear_range()
  intel-iommu: Clean up iommu_domain_identity_map()
  intel-iommu: Remove last use of PHYSICAL_PAGE_MASK, for reserving PCI BARs
  intel-iommu: Make iommu_flush_iotlb_psi() take pfn as argument
  intel-iommu: Change aligned_size() to aligned_nrpages()
  intel-iommu: Clean up intel_map_sg(), remove domain_page_mapping()
  ...

1 files changed, 350 insertions, 345 deletions

diff --git a/drivers/pci/intel-iommu.c b/drivers/pci/intel-iommu.c
index 420afa88728..53075424a43 100644
--- a/drivers/pci/intel-iommu.c
+++ b/drivers/pci/intel-iommu.c
@@ -56,14 +56,32 @@
 #define MAX_AGAW_WIDTH 64
 
 #define DOMAIN_MAX_ADDR(gaw) ((((u64)1) << gaw) - 1)
+#define DOMAIN_MAX_PFN(gaw)  ((((u64)1) << (gaw-VTD_PAGE_SHIFT)) - 1)
 
 #define IOVA_PFN(addr)		((addr) >> PAGE_SHIFT)
 #define DMA_32BIT_PFN		IOVA_PFN(DMA_BIT_MASK(32))
 #define DMA_64BIT_PFN		IOVA_PFN(DMA_BIT_MASK(64))
 
-#ifndef PHYSICAL_PAGE_MASK
-#define PHYSICAL_PAGE_MASK PAGE_MASK
-#endif
+
+/* VT-d pages must always be _smaller_ than MM pages. Otherwise things
+   are never going to work. */
+static inline unsigned long dma_to_mm_pfn(unsigned long dma_pfn)
+{
+	return dma_pfn >> (PAGE_SHIFT - VTD_PAGE_SHIFT);
+}
+
+static inline unsigned long mm_to_dma_pfn(unsigned long mm_pfn)
+{
+	return mm_pfn << (PAGE_SHIFT - VTD_PAGE_SHIFT);
+}
+static inline unsigned long page_to_dma_pfn(struct page *pg)
+{
+	return mm_to_dma_pfn(page_to_pfn(pg));
+}
+static inline unsigned long virt_to_dma_pfn(void *p)
+{
+	return page_to_dma_pfn(virt_to_page(p));
+}
 
 /* global iommu list, set NULL for ignored DMAR units */
 static struct intel_iommu **g_iommus;
@@ -204,12 +222,17 @@ static inline void dma_set_pte_prot(struct dma_pte *pte, unsigned long prot)
 
 static inline u64 dma_pte_addr(struct dma_pte *pte)
 {
-	return (pte->val & VTD_PAGE_MASK);
+#ifdef CONFIG_64BIT
+	return pte->val & VTD_PAGE_MASK;
+#else
+	/* Must have a full atomic 64-bit read */
+	return  __cmpxchg64(pte, 0ULL, 0ULL) & VTD_PAGE_MASK;
+#endif
 }
 
-static inline void dma_set_pte_addr(struct dma_pte *pte, u64 addr)
+static inline void dma_set_pte_pfn(struct dma_pte *pte, unsigned long pfn)
 {
-	pte->val |= (addr & VTD_PAGE_MASK);
+	pte->val |= (uint64_t)pfn << VTD_PAGE_SHIFT;
 }
 
 static inline bool dma_pte_present(struct dma_pte *pte)
@@ -217,6 +240,11 @@ static inline bool dma_pte_present(struct dma_pte *pte)
 	return (pte->val & 3) != 0;
 }
 
+static inline int first_pte_in_page(struct dma_pte *pte)
+{
+	return !((unsigned long)pte & ~VTD_PAGE_MASK);
+}
+
 /*
  * This domain is a statically identity mapping domain.
  *	1. This domain creats a static 1:1 mapping to all usable memory.
@@ -244,7 +272,6 @@ struct dmar_domain {
 	struct iova_domain iovad;	/* iova's that belong to this domain */
 
 	struct dma_pte	*pgd;		/* virtual address */
-	spinlock_t	mapping_lock;	/* page table lock */
 	int		gaw;		/* max guest address width */
 
 	/* adjusted guest address width, 0 is level 2 30-bit */
@@ -648,80 +675,78 @@ static inline int width_to_agaw(int width)
 
 static inline unsigned int level_to_offset_bits(int level)
 {
-	return (12 + (level - 1) * LEVEL_STRIDE);
+	return (level - 1) * LEVEL_STRIDE;
 }
 
-static inline int address_level_offset(u64 addr, int level)
+static inline int pfn_level_offset(unsigned long pfn, int level)
 {
-	return ((addr >> level_to_offset_bits(level)) & LEVEL_MASK);
+	return (pfn >> level_to_offset_bits(level)) & LEVEL_MASK;
 }
 
-static inline u64 level_mask(int level)
+static inline unsigned long level_mask(int level)
 {
-	return ((u64)-1 << level_to_offset_bits(level));
+	return -1UL << level_to_offset_bits(level);
 }
 
-static inline u64 level_size(int level)
+static inline unsigned long level_size(int level)
 {
-	return ((u64)1 << level_to_offset_bits(level));
+	return 1UL << level_to_offset_bits(level);
 }
 
-static inline u64 align_to_level(u64 addr, int level)
+static inline unsigned long align_to_level(unsigned long pfn, int level)
 {
-	return ((addr + level_size(level) - 1) & level_mask(level));
+	return (pfn + level_size(level) - 1) & level_mask(level);
 }
 
-static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr)
+static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain,
+				      unsigned long pfn)
 {
-	int addr_width = agaw_to_width(domain->agaw);
+	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
 	struct dma_pte *parent, *pte = NULL;
 	int level = agaw_to_level(domain->agaw);
 	int offset;
-	unsigned long flags;
 
 	BUG_ON(!domain->pgd);
-
-	addr &= (((u64)1) << addr_width) - 1;
+	BUG_ON(addr_width < BITS_PER_LONG && pfn >> addr_width);
 	parent = domain->pgd;
 
-	spin_lock_irqsave(&domain->mapping_lock, flags);
 	while (level > 0) {
 		void *tmp_page;
 
-		offset = address_level_offset(addr, level);
+		offset = pfn_level_offset(pfn, level);
 		pte = &parent[offset];
 		if (level == 1)
 			break;
 
 		if (!dma_pte_present(pte)) {
+			uint64_t pteval;
+
 			tmp_page = alloc_pgtable_page();
 
-			if (!tmp_page) {
-				spin_unlock_irqrestore(&domain->mapping_lock,
-					flags);
+			if (!tmp_page)
 				return NULL;
+
+			domain_flush_cache(domain, tmp_page, VTD_PAGE_SIZE);
+			pteval = (virt_to_dma_pfn(tmp_page) << VTD_PAGE_SHIFT) | DMA_PTE_READ | DMA_PTE_WRITE;
+			if (cmpxchg64(&pte->val, 0ULL, pteval)) {
+				/* Someone else set it while we were thinking; use theirs. */
+				free_pgtable_page(tmp_page);
+			} else {
+				dma_pte_addr(pte);
+				domain_flush_cache(domain, pte, sizeof(*pte));
 			}
-			domain_flush_cache(domain, tmp_page, PAGE_SIZE);
-			dma_set_pte_addr(pte, virt_to_phys(tmp_page));
-			/*
-			 * high level table always sets r/w, last level page
-			 * table control read/write
-			 */
-			dma_set_pte_readable(pte);
-			dma_set_pte_writable(pte);
-			domain_flush_cache(domain, pte, sizeof(*pte));
 		}
 		parent = phys_to_virt(dma_pte_addr(pte));
 		level--;
 	}
 
-	spin_unlock_irqrestore(&domain->mapping_lock, flags);
 	return pte;
 }
 
 /* return address's pte at specific level */
-static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
-		int level)
+static struct dma_pte *dma_pfn_level_pte(struct dmar_domain *domain,
+					 unsigned long pfn,
+					 int level)
 {
 	struct dma_pte *parent, *pte = NULL;
 	int total = agaw_to_level(domain->agaw);
@@ -729,7 +754,7 @@ static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
 
 	parent = domain->pgd;
 	while (level <= total) {
-		offset = address_level_offset(addr, total);
+		offset = pfn_level_offset(pfn, total);
 		pte = &parent[offset];
 		if (level == total)
 			return pte;
@@ -742,74 +767,82 @@ static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
 	return NULL;
 }
 
-/* clear one page's page table */
-static void dma_pte_clear_one(struct dmar_domain *domain, u64 addr)
-{
-	struct dma_pte *pte = NULL;
-
-	/* get last level pte */
-	pte = dma_addr_level_pte(domain, addr, 1);
-
-	if (pte) {
-		dma_clear_pte(pte);
-		domain_flush_cache(domain, pte, sizeof(*pte));
-	}
-}
-
 /* clear last level pte, a tlb flush should be followed */
-static void dma_pte_clear_range(struct dmar_domain *domain, u64 start, u64 end)
+static void dma_pte_clear_range(struct dmar_domain *domain,
+				unsigned long start_pfn,
+				unsigned long last_pfn)
 {
-	int addr_width = agaw_to_width(domain->agaw);
-	int npages;
+	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+	struct dma_pte *first_pte, *pte;
+
+	BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
+	BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
 
-	start &= (((u64)1) << addr_width) - 1;
-	end &= (((u64)1) << addr_width) - 1;
-	/* in case it's partial page */
-	start &= PAGE_MASK;
-	end = PAGE_ALIGN(end);
-	npages = (end - start) / VTD_PAGE_SIZE;
+	/* we don't need lock here; nobody else touches the iova range */
+	while (start_pfn <= last_pfn) {
+		first_pte = pte = dma_pfn_level_pte(domain, start_pfn, 1);
+		if (!pte) {
+			start_pfn = align_to_level(start_pfn + 1, 2);
+			continue;
+		}
+		do { 
+			dma_clear_pte(pte);
+			start_pfn++;
+			pte++;
+		} while (start_pfn <= last_pfn && !first_pte_in_page(pte));
 
-	/* we don't need lock here, nobody else touches the iova range */
-	while (npages--) {
-		dma_pte_clear_one(domain, start);
-		start += VTD_PAGE_SIZE;
+		domain_flush_cache(domain, first_pte,
+				   (void *)pte - (void *)first_pte);
 	}
 }
 
 /* free page table pages. last level pte should already be cleared */
 static void dma_pte_free_pagetable(struct dmar_domain *domain,
-	u64 start, u64 end)
+				   unsigned long start_pfn,
+				   unsigned long last_pfn)
 {
-	int addr_width = agaw_to_width(domain->agaw);
-	struct dma_pte *pte;
+	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+	struct dma_pte *first_pte, *pte;
 	int total = agaw_to_level(domain->agaw);
 	int level;
-	u64 tmp;
+	unsigned long tmp;
 
-	start &= (((u64)1) << addr_width) - 1;
-	end &= (((u64)1) << addr_width) - 1;
+	BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
+	BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
 
-	/* we don't need lock here, nobody else touches the iova range */
+	/* We don't need lock here; nobody else touches the iova range */
 	level = 2;
 	while (level <= total) {
-		tmp = align_to_level(start, level);
-		if (tmp >= end || (tmp + level_size(level) > end))
+		tmp = align_to_level(start_pfn, level);
+
+		/* If we can't even clear one PTE at this level, we're done */
+		if (tmp + level_size(level) - 1 > last_pfn)
 			return;
 
-		while (tmp < end) {
-			pte = dma_addr_level_pte(domain, tmp, level);
-			if (pte) {
-				free_pgtable_page(
-					phys_to_virt(dma_pte_addr(pte)));
-				dma_clear_pte(pte);
-				domain_flush_cache(domain, pte, sizeof(*pte));
+		while (tmp + level_size(level) - 1 <= last_pfn) {
+			first_pte = pte = dma_pfn_level_pte(domain, tmp, level);
+			if (!pte) {
+				tmp = align_to_level(tmp + 1, level + 1);
+				continue;
 			}
-			tmp += level_size(level);
+			do {
+				if (dma_pte_present(pte)) {
+					free_pgtable_page(phys_to_virt(dma_pte_addr(pte)));
+					dma_clear_pte(pte);
+				}
+				pte++;
+				tmp += level_size(level);
+			} while (!first_pte_in_page(pte) &&
+				 tmp + level_size(level) - 1 <= last_pfn);
+
+			domain_flush_cache(domain, first_pte,
+					   (void *)pte - (void *)first_pte);
+			
 		}
 		level++;
 	}
 	/* free pgd */
-	if (start == 0 && end >= ((((u64)1) << addr_width) - 1)) {
+	if (start_pfn == 0 && last_pfn == DOMAIN_MAX_PFN(domain->gaw)) {
 		free_pgtable_page(domain->pgd);
 		domain->pgd = NULL;
 	}
@@ -1035,11 +1068,11 @@ static void iommu_flush_dev_iotlb(struct dmar_domain *domain,
 }
 
 static void iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
-				  u64 addr, unsigned int pages)
+				  unsigned long pfn, unsigned int pages)
 {
 	unsigned int mask = ilog2(__roundup_pow_of_two(pages));
+	uint64_t addr = (uint64_t)pfn << VTD_PAGE_SHIFT;
 
-	BUG_ON(addr & (~VTD_PAGE_MASK));
 	BUG_ON(pages == 0);
 
 	/*
@@ -1054,7 +1087,12 @@ static void iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
 	else
 		iommu->flush.flush_iotlb(iommu, did, addr, mask,
 						DMA_TLB_PSI_FLUSH);
-	if (did)
+
+	/*
+	 * In caching mode, domain ID 0 is reserved for non-present to present
+	 * mapping flush. Device IOTLB doesn't need to be flushed in this case.
+	 */
+	if (!cap_caching_mode(iommu->cap) || did)
 		iommu_flush_dev_iotlb(iommu->domains[did], addr, mask);
 }
 
@@ -1279,7 +1317,6 @@ static void dmar_init_reserved_ranges(void)
 	struct pci_dev *pdev = NULL;
 	struct iova *iova;
 	int i;
-	u64 addr, size;
 
 	init_iova_domain(&reserved_iova_list, DMA_32BIT_PFN);
 
@@ -1302,12 +1339,9 @@ static void dmar_init_reserved_ranges(void)
 			r = &pdev->resource[i];
 			if (!r->flags || !(r->flags & IORESOURCE_MEM))
 				continue;
-			addr = r->start;
-			addr &= PHYSICAL_PAGE_MASK;
-			size = r->end - addr;
-			size = PAGE_ALIGN(size);
-			iova = reserve_iova(&reserved_iova_list, IOVA_PFN(addr),
-				IOVA_PFN(size + addr) - 1);
+			iova = reserve_iova(&reserved_iova_list,
+					    IOVA_PFN(r->start),
+					    IOVA_PFN(r->end));
 			if (!iova)
 				printk(KERN_ERR "Reserve iova failed\n");
 		}
@@ -1341,7 +1375,6 @@ static int domain_init(struct dmar_domain *domain, int guest_width)
 	unsigned long sagaw;
 
 	init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
-	spin_lock_init(&domain->mapping_lock);
 	spin_lock_init(&domain->iommu_lock);
 
 	domain_reserve_special_ranges(domain);
@@ -1388,7 +1421,6 @@ static void domain_exit(struct dmar_domain *domain)
 {
 	struct dmar_drhd_unit *drhd;
 	struct intel_iommu *iommu;
-	u64 end;
 
 	/* Domain 0 is reserved, so dont process it */
 	if (!domain)
@@ -1397,14 +1429,12 @@ static void domain_exit(struct dmar_domain *domain)
 	domain_remove_dev_info(domain);
 	/* destroy iovas */
 	put_iova_domain(&domain->iovad);
-	end = DOMAIN_MAX_ADDR(domain->gaw);
-	end = end & (~PAGE_MASK);
 
 	/* clear ptes */
-	dma_pte_clear_range(domain, 0, end);
+	dma_pte_clear_range(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
 
 	/* free page tables */
-	dma_pte_free_pagetable(domain, 0, end);
+	dma_pte_free_pagetable(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
 
 	for_each_active_iommu(iommu, drhd)
 		if (test_bit(iommu->seq_id, &domain->iommu_bmp))
@@ -1618,42 +1648,86 @@ static int domain_context_mapped(struct pci_dev *pdev)
 					     tmp->devfn);
 }
 
-static int
-domain_page_mapping(struct dmar_domain *domain, dma_addr_t iova,
-			u64 hpa, size_t size, int prot)
+static int __domain_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+			    struct scatterlist *sg, unsigned long phys_pfn,
+			    unsigned long nr_pages, int prot)
 {
-	u64 start_pfn, end_pfn;
-	struct dma_pte *pte;
-	int index;
-	int addr_width = agaw_to_width(domain->agaw);
+	struct dma_pte *first_pte = NULL, *pte = NULL;
+	phys_addr_t uninitialized_var(pteval);
+	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+	unsigned long sg_res;
 
-	hpa &= (((u64)1) << addr_width) - 1;
+	BUG_ON(addr_width < BITS_PER_LONG && (iov_pfn + nr_pages - 1) >> addr_width);
 
 	if ((prot & (DMA_PTE_READ|DMA_PTE_WRITE)) == 0)
 		return -EINVAL;
-	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 + VTD_PAGE_SIZE * index);
-		if (!pte)
-			return -ENOMEM;
+
+	prot &= DMA_PTE_READ | DMA_PTE_WRITE | DMA_PTE_SNP;
+
+	if (sg)
+		sg_res = 0;
+	else {
+		sg_res = nr_pages + 1;
+		pteval = ((phys_addr_t)phys_pfn << VTD_PAGE_SHIFT) | prot;
+	}
+
+	while (nr_pages--) {
+		uint64_t tmp;
+
+		if (!sg_res) {
+			sg_res = (sg->offset + sg->length + VTD_PAGE_SIZE - 1) >> VTD_PAGE_SHIFT;
+			sg->dma_address = ((dma_addr_t)iov_pfn << VTD_PAGE_SHIFT) + sg->offset;
+			sg->dma_length = sg->length;
+			pteval = page_to_phys(sg_page(sg)) | prot;
+		}
+		if (!pte) {
+			first_pte = pte = pfn_to_dma_pte(domain, iov_pfn);
+			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 << VTD_PAGE_SHIFT);
-		dma_set_pte_prot(pte, prot);
-		if (prot & DMA_PTE_SNP)
-			dma_set_pte_snp(pte);
-		domain_flush_cache(domain, pte, sizeof(*pte));
-		start_pfn++;
-		index++;
+		tmp = cmpxchg64_local(&pte->val, 0ULL, pteval);
+		if (tmp) {
+			static int dumps = 5;
+			printk(KERN_CRIT "ERROR: DMA PTE for vPFN 0x%lx already set (to %llx not %llx)\n",
+			       iov_pfn, tmp, (unsigned long long)pteval);
+			if (dumps) {
+				dumps--;
+				debug_dma_dump_mappings(NULL);
+			}
+			WARN_ON(1);
+		}
+		pte++;
+		if (!nr_pages || first_pte_in_page(pte)) {
+			domain_flush_cache(domain, first_pte,
+					   (void *)pte - (void *)first_pte);
+			pte = NULL;
+		}
+		iov_pfn++;
+		pteval += VTD_PAGE_SIZE;
+		sg_res--;
+		if (!sg_res)
+			sg = sg_next(sg);
 	}
 	return 0;
 }
 
+static inline int domain_sg_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+				    struct scatterlist *sg, unsigned long nr_pages,
+				    int prot)
+{
+	return __domain_mapping(domain, iov_pfn, sg, 0, nr_pages, prot);
+}
+
+static inline int domain_pfn_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+				     unsigned long phys_pfn, unsigned long nr_pages,
+				     int prot)
+{
+	return __domain_mapping(domain, iov_pfn, NULL, phys_pfn, nr_pages, prot);
+}
+
 static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn)
 {
 	if (!iommu)
@@ -1844,58 +1918,61 @@ error:
 
 static int iommu_identity_mapping;
 
+static int iommu_domain_identity_map(struct dmar_domain *domain,
+				     unsigned long long start,
+				     unsigned long long end)
+{
+	unsigned long first_vpfn = start >> VTD_PAGE_SHIFT;
+	unsigned long last_vpfn = end >> VTD_PAGE_SHIFT;
+
+	if (!reserve_iova(&domain->iovad, dma_to_mm_pfn(first_vpfn),
+			  dma_to_mm_pfn(last_vpfn))) {
+		printk(KERN_ERR "IOMMU: reserve iova failed\n");
+		return -ENOMEM;
+	}
+
+	pr_debug("Mapping reserved region %llx-%llx for domain %d\n",
+		 start, end, domain->id);
+	/*
+	 * RMRR range might have overlap with physical memory range,
+	 * clear it first
+	 */
+	dma_pte_clear_range(domain, first_vpfn, last_vpfn);
+
+	return domain_pfn_mapping(domain, first_vpfn, first_vpfn,
+				  last_vpfn - first_vpfn + 1,
+				  DMA_PTE_READ|DMA_PTE_WRITE);
+}
+
 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;
-	unsigned long long base;
 	int ret;
 
 	printk(KERN_INFO
-		"IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
-		pci_name(pdev), start, end);
-	if (iommu_identity_mapping)
-		domain = si_domain;
-	else
-		/* page table init */
-		domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
+	       "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
+	       pci_name(pdev), start, end);
+
+	domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
 	if (!domain)
 		return -ENOMEM;
 
-	/* The address might not be aligned */
-	base = start & PAGE_MASK;
-	size = end - base;
-	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");
-		ret = -ENOMEM;
-		goto error;
-	}
-
-	pr_debug("Mapping reserved region %lx@%llx for %s\n",
-		size, base, pci_name(pdev));
-	/*
-	 * RMRR range might have overlap with physical memory range,
-	 * clear it first
-	 */
-	dma_pte_clear_range(domain, base, base + size);
-
-	ret = domain_page_mapping(domain, base, base, size,
-		DMA_PTE_READ|DMA_PTE_WRITE);
+	ret = iommu_domain_identity_map(domain, start, end);
 	if (ret)
 		goto error;
 
 	/* context entry init */
 	ret = domain_context_mapping(domain, pdev, CONTEXT_TT_MULTI_LEVEL);
-	if (!ret)
-		return 0;
-error:
+	if (ret)
+		goto error;
+
+	return 0;
+
+ error:
 	domain_exit(domain);
 	return ret;
-
 }
 
 static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr,
@@ -1907,64 +1984,6 @@ static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr,
 		rmrr->end_address + 1);
 }
 
-struct iommu_prepare_data {
-	struct pci_dev *pdev;
-	int ret;
-};
-
-static int __init iommu_prepare_work_fn(unsigned long start_pfn,
-					 unsigned long end_pfn, void *datax)
-{
-	struct iommu_prepare_data *data;
-
-	data = (struct iommu_prepare_data *)datax;
-
-	data->ret = iommu_prepare_identity_map(data->pdev,
-				start_pfn<<PAGE_SHIFT, end_pfn<<PAGE_SHIFT);
-	return data->ret;
-
-}
-
-static int __init iommu_prepare_with_active_regions(struct pci_dev *pdev)
-{
-	int nid;
-	struct iommu_prepare_data data;
-
-	data.pdev = pdev;
-	data.ret = 0;
-
-	for_each_online_node(nid) {
-		work_with_active_regions(nid, iommu_prepare_work_fn, &data);
-		if (data.ret)
-			return data.ret;
-	}
-	return data.ret;
-}
-
-#ifdef CONFIG_DMAR_GFX_WA
-static void __init iommu_prepare_gfx_mapping(void)
-{
-	struct pci_dev *pdev = NULL;
-	int ret;
-
-	for_each_pci_dev(pdev) {
-		if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO ||
-				!IS_GFX_DEVICE(pdev))
-			continue;
-		printk(KERN_INFO "IOMMU: gfx device %s 1-1 mapping\n",
-			pci_name(pdev));
-		ret = iommu_prepare_with_active_regions(pdev);
-		if (ret)
-			printk(KERN_ERR "IOMMU: mapping reserved region failed\n");
-	}
-}
-#else /* !CONFIG_DMAR_GFX_WA */
-static inline void iommu_prepare_gfx_mapping(void)
-{
-	return;
-}
-#endif
-
 #ifdef CONFIG_DMAR_FLOPPY_WA
 static inline void iommu_prepare_isa(void)
 {
@@ -1975,12 +1994,12 @@ static inline void iommu_prepare_isa(void)
 	if (!pdev)
 		return;
 
-	printk(KERN_INFO "IOMMU: Prepare 0-16M unity mapping for LPC\n");
+	printk(KERN_INFO "IOMMU: Prepare 0-16MiB unity mapping for LPC\n");
 	ret = iommu_prepare_identity_map(pdev, 0, 16*1024*1024);
 
 	if (ret)
-		printk(KERN_ERR "IOMMU: Failed to create 0-64M identity map, "
-			"floppy might not work\n");
+		printk(KERN_ERR "IOMMU: Failed to create 0-16MiB identity map; "
+		       "floppy might not work\n");
 
 }
 #else
@@ -2008,16 +2027,30 @@ static int __init init_context_pass_through(void)
 }
 
 static int md_domain_init(struct dmar_domain *domain, int guest_width);
+
+static int __init si_domain_work_fn(unsigned long start_pfn,
+				    unsigned long end_pfn, void *datax)
+{
+	int *ret = datax;
+
+	*ret = iommu_domain_identity_map(si_domain,
+					 (uint64_t)start_pfn << PAGE_SHIFT,
+					 (uint64_t)end_pfn << PAGE_SHIFT);
+	return *ret;
+
+}
+
 static int si_domain_init(void)
 {
 	struct dmar_drhd_unit *drhd;
 	struct intel_iommu *iommu;
-	int ret = 0;
+	int nid, ret = 0;
 
 	si_domain = alloc_domain();
 	if (!si_domain)
 		return -EFAULT;
 
+	pr_debug("Identity mapping domain is domain %d\n", si_domain->id);
 
 	for_each_active_iommu(iommu, drhd) {
 		ret = iommu_attach_domain(si_domain, iommu);
@@ -2034,6 +2067,12 @@ static int si_domain_init(void)
 
 	si_domain->flags = DOMAIN_FLAG_STATIC_IDENTITY;
 
+	for_each_online_node(nid) {
+		work_with_active_regions(nid, si_domain_work_fn, &ret);
+		if (ret)
+			return ret;
+	}
+
 	return 0;
 }
 
@@ -2087,13 +2126,14 @@ static int iommu_prepare_static_identity_mapping(void)
 	if (ret)
 		return -EFAULT;
 
-	printk(KERN_INFO "IOMMU: Setting identity map:\n");
 	for_each_pci_dev(pdev) {
-		ret = iommu_prepare_with_active_regions(pdev);
-		if (ret) {
-			printk(KERN_INFO "1:1 mapping to one domain failed.\n");
-			return -EFAULT;
-		}
+		printk(KERN_INFO "IOMMU: identity mapping for device %s\n",
+		       pci_name(pdev));
+
+		ret = domain_context_mapping(si_domain, pdev,
+					     CONTEXT_TT_MULTI_LEVEL);
+		if (ret)
+			return ret;
 		ret = domain_add_dev_info(si_domain, pdev);
 		if (ret)
 			return ret;
@@ -2284,8 +2324,6 @@ int __init init_dmars(void)
 			}
 		}
 
-		iommu_prepare_gfx_mapping();
-
 		iommu_prepare_isa();
 	}
 
@@ -2330,50 +2368,40 @@ error:
 	return ret;
 }
 
-static inline u64 aligned_size(u64 host_addr, size_t size)
-{
-	u64 addr;
-	addr = (host_addr & (~PAGE_MASK)) + size;
-	return PAGE_ALIGN(addr);
-}
-
-struct iova *
-iommu_alloc_iova(struct dmar_domain *domain, size_t size, u64 end)
+static inline unsigned long aligned_nrpages(unsigned long host_addr,
+					    size_t size)
 {
-	struct iova *piova;
+	host_addr &= ~PAGE_MASK;
+	host_addr += size + PAGE_SIZE - 1;
 
-	/* Make sure it's in range */
-	end = min_t(u64, DOMAIN_MAX_ADDR(domain->gaw), end);
-	if (!size || (IOVA_START_ADDR + size > end))
-		return NULL;
-
-	piova = alloc_iova(&domain->iovad,
-			size >> PAGE_SHIFT, IOVA_PFN(end), 1);
-	return piova;
+	return host_addr >> VTD_PAGE_SHIFT;
 }
 
-static struct iova *
-__intel_alloc_iova(struct device *dev, struct dmar_domain *domain,
-		   size_t size, u64 dma_mask)
+static struct iova *intel_alloc_iova(struct device *dev,
+				     struct dmar_domain *domain,
+				     unsigned long nrpages, uint64_t dma_mask)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct iova *iova = NULL;
 
-	if (dma_mask <= DMA_BIT_MASK(32) || dmar_forcedac)
-		iova = iommu_alloc_iova(domain, size, dma_mask);
-	else {
+	/* Restrict dma_mask to the width that the iommu can handle */
+	dma_mask = min_t(uint64_t, DOMAIN_MAX_ADDR(domain->gaw), dma_mask);
+
+	if (!dmar_forcedac && dma_mask > DMA_BIT_MASK(32)) {
 		/*
 		 * First try to allocate an io virtual address in
 		 * DMA_BIT_MASK(32) and if that fails then try allocating
 		 * from higher range
 		 */
-		iova = iommu_alloc_iova(domain, size, DMA_BIT_MASK(32));
-		if (!iova)
-			iova = iommu_alloc_iova(domain, size, dma_mask);
-	}
-
-	if (!iova) {
-		printk(KERN_ERR"Allocating iova for %s failed", pci_name(pdev));
+		iova = alloc_iova(&domain->iovad, nrpages,
+				  IOVA_PFN(DMA_BIT_MASK(32)), 1);
+		if (iova)
+			return iova;
+	}
+	iova = alloc_iova(&domain->iovad, nrpages, IOVA_PFN(dma_mask), 1);
+	if (unlikely(!iova)) {
+		printk(KERN_ERR "Allocating %ld-page iova for %s failed",
+		       nrpages, pci_name(pdev));
 		return NULL;
 	}
 
@@ -2476,14 +2504,12 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
 		return 0;
 
 	iommu = domain_get_iommu(domain);
-	size = aligned_size((u64)paddr, size);
+	size = aligned_nrpages(paddr, size);
 
-	iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
+	iova = intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
 	if (!iova)
 		goto error;
 
-	start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
-
 	/*
 	 * Check if DMAR supports zero-length reads on write only
 	 * mappings..
@@ -2499,20 +2525,20 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
 	 * might have two guest_addr mapping to the same host paddr, but this
 	 * is not a big problem
 	 */
-	ret = domain_page_mapping(domain, start_paddr,
-				  ((u64)paddr) & PHYSICAL_PAGE_MASK,
-				  size, prot);
+	ret = domain_pfn_mapping(domain, mm_to_dma_pfn(iova->pfn_lo),
+				 paddr >> VTD_PAGE_SHIFT, size, prot);
 	if (ret)
 		goto error;
 
 	/* it's a non-present to present mapping. Only flush if caching mode */
 	if (cap_caching_mode(iommu->cap))
-		iommu_flush_iotlb_psi(iommu, 0, start_paddr,
-				      size >> VTD_PAGE_SHIFT);
+		iommu_flush_iotlb_psi(iommu, 0, mm_to_dma_pfn(iova->pfn_lo), size);
 	else
 		iommu_flush_write_buffer(iommu);
 
-	return start_paddr + ((u64)paddr & (~PAGE_MASK));
+	start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
+	start_paddr += paddr & ~PAGE_MASK;
+	return start_paddr;
 
 error:
 	if (iova)
@@ -2605,7 +2631,7 @@ static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr,
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct dmar_domain *domain;
-	unsigned long start_addr;
+	unsigned long start_pfn, last_pfn;
 	struct iova *iova;
 	struct intel_iommu *iommu;
 
@@ -2618,22 +2644,25 @@ static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr,
 	iommu = domain_get_iommu(domain);
 
 	iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr));
-	if (!iova)
+	if (WARN_ONCE(!iova, "Driver unmaps unmatched page at PFN %llx\n",
+		      (unsigned long long)dev_addr))
 		return;
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT;
-	size = aligned_size((u64)dev_addr, size);
+	start_pfn = mm_to_dma_pfn(iova->pfn_lo);
+	last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
 
-	pr_debug("Device %s unmapping: %zx@%llx\n",
-		pci_name(pdev), size, (unsigned long long)start_addr);
+	pr_debug("Device %s unmapping: pfn %lx-%lx\n",
+		 pci_name(pdev), start_pfn, last_pfn);
 
 	/*  clear the whole page */
-	dma_pte_clear_range(domain, start_addr, start_addr + size);
+	dma_pte_clear_range(domain, start_pfn, last_pfn);
+
 	/* free page tables */
-	dma_pte_free_pagetable(domain, start_addr, start_addr + size);
+	dma_pte_free_pagetable(domain, start_pfn, last_pfn);
+
 	if (intel_iommu_strict) {
-		iommu_flush_iotlb_psi(iommu, domain->id, start_addr,
-				      size >> VTD_PAGE_SHIFT);
+		iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
+				      last_pfn - start_pfn + 1);
 		/* free iova */
 		__free_iova(&domain->iovad, iova);
 	} else {
@@ -2691,14 +2720,10 @@ static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
 			   int nelems, enum dma_data_direction dir,
 			   struct dma_attrs *attrs)
 {
-	int i;
 	struct pci_dev *pdev = to_pci_dev(hwdev);
 	struct dmar_domain *domain;
-	unsigned long start_addr;
+	unsigned long start_pfn, last_pfn;
 	struct iova *iova;
-	size_t size = 0;
-	phys_addr_t addr;
-	struct scatterlist *sg;
 	struct intel_iommu *iommu;
 
 	if (iommu_no_mapping(pdev))
@@ -2710,22 +2735,21 @@ static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
 	iommu = domain_get_iommu(domain);
 
 	iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address));
-	if (!iova)
+	if (WARN_ONCE(!iova, "Driver unmaps unmatched sglist at PFN %llx\n",
+		      (unsigned long long)sglist[0].dma_address))
 		return;
-	for_each_sg(sglist, sg, nelems, i) {
-		addr = page_to_phys(sg_page(sg)) + sg->offset;
-		size += aligned_size((u64)addr, sg->length);
-	}
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT;
+	start_pfn = mm_to_dma_pfn(iova->pfn_lo);
+	last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
 
 	/*  clear the whole page */
-	dma_pte_clear_range(domain, start_addr, start_addr + size);
+	dma_pte_clear_range(domain, start_pfn, last_pfn);
+
 	/* free page tables */
-	dma_pte_free_pagetable(domain, start_addr, start_addr + size);
+	dma_pte_free_pagetable(domain, start_pfn, last_pfn);
 
-	iommu_flush_iotlb_psi(iommu, domain->id, start_addr,
-			      size >> VTD_PAGE_SHIFT);
+	iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
+			      (last_pfn - start_pfn + 1));
 
 	/* free iova */
 	__free_iova(&domain->iovad, iova);
@@ -2748,17 +2772,16 @@ static int intel_nontranslate_map_sg(struct device *hddev,
 static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
 			enum dma_data_direction dir, struct dma_attrs *attrs)
 {
-	phys_addr_t addr;
 	int i;
 	struct pci_dev *pdev = to_pci_dev(hwdev);
 	struct dmar_domain *domain;
 	size_t size = 0;
 	int prot = 0;
-	size_t offset = 0;
+	size_t offset_pfn = 0;
 	struct iova *iova = NULL;
 	int ret;
 	struct scatterlist *sg;
-	unsigned long start_addr;
+	unsigned long start_vpfn;
 	struct intel_iommu *iommu;
 
 	BUG_ON(dir == DMA_NONE);
@@ -2771,12 +2794,10 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int ne
 
 	iommu = domain_get_iommu(domain);
 
-	for_each_sg(sglist, sg, nelems, i) {
-		addr = page_to_phys(sg_page(sg)) + sg->offset;
-		size += aligned_size((u64)addr, sg->length);
-	}
+	for_each_sg(sglist, sg, nelems, i)
+		size += aligned_nrpages(sg->offset, sg->length);
 
-	iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
+	iova = intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
 	if (!iova) {
 		sglist->dma_length = 0;
 		return 0;
@@ -2792,35 +2813,24 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int ne
 	if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
 		prot |= DMA_PTE_WRITE;
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT;
-	offset = 0;
-	for_each_sg(sglist, sg, nelems, i) {
-		addr = page_to_phys(sg_page(sg)) + sg->offset;
-		size = aligned_size((u64)addr, sg->length);
-		ret = domain_page_mapping(domain, start_addr + offset,
-					  ((u64)addr) & PHYSICAL_PAGE_MASK,
-					  size, prot);
-		if (ret) {
-			/*  clear the page */
-			dma_pte_clear_range(domain, start_addr,
-				  start_addr + offset);
-			/* free page tables */
-			dma_pte_free_pagetable(domain, start_addr,
-				  start_addr + offset);
-			/* free iova */
-			__free_iova(&domain->iovad, iova);
-			return 0;
-		}
-		sg->dma_address = start_addr + offset +
-				((u64)addr & (~PAGE_MASK));
-		sg->dma_length = sg->length;
-		offset += size;
+	start_vpfn = mm_to_dma_pfn(iova->pfn_lo);
+
+	ret = domain_sg_mapping(domain, start_vpfn, sglist, mm_to_dma_pfn(size), prot);
+	if (unlikely(ret)) {
+		/*  clear the page */
+		dma_pte_clear_range(domain, start_vpfn,
+				    start_vpfn + size - 1);
+		/* free page tables */
+		dma_pte_free_pagetable(domain, start_vpfn,
+				       start_vpfn + size - 1);
+		/* free iova */
+		__free_iova(&domain->iovad, iova);
+		return 0;
 	}
 
 	/* it's a non-present to present mapping. Only flush if caching mode */
 	if (cap_caching_mode(iommu->cap))
-		iommu_flush_iotlb_psi(iommu, 0, start_addr,
-				      offset >> VTD_PAGE_SHIFT);
+		iommu_flush_iotlb_psi(iommu, 0, start_vpfn, offset_pfn);
 	else
 		iommu_flush_write_buffer(iommu);
 
@@ -3325,7 +3335,6 @@ static int md_domain_init(struct dmar_domain *domain, int guest_width)
 	int adjust_width;
 
 	init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
-	spin_lock_init(&domain->mapping_lock);
 	spin_lock_init(&domain->iommu_lock);
 
 	domain_reserve_special_ranges(domain);
@@ -3379,8 +3388,6 @@ static void iommu_free_vm_domain(struct dmar_domain *domain)
 
 static void vm_domain_exit(struct dmar_domain *domain)
 {
-	u64 end;
-
 	/* Domain 0 is reserved, so dont process it */
 	if (!domain)
 		return;
@@ -3388,14 +3395,12 @@ static void vm_domain_exit(struct dmar_domain *domain)
 	vm_domain_remove_all_dev_info(domain);
 	/* destroy iovas */
 	put_iova_domain(&domain->iovad);
-	end = DOMAIN_MAX_ADDR(domain->gaw);
-	end = end & (~VTD_PAGE_MASK);
 
 	/* clear ptes */
-	dma_pte_clear_range(domain, 0, end);
+	dma_pte_clear_range(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
 
 	/* free page tables */
-	dma_pte_free_pagetable(domain, 0, end);
+	dma_pte_free_pagetable(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
 
 	iommu_free_vm_domain(domain);
 	free_domain_mem(domain);
@@ -3504,7 +3509,7 @@ static int intel_iommu_map_range(struct iommu_domain *domain,
 	if ((iommu_prot & IOMMU_CACHE) && dmar_domain->iommu_snooping)
 		prot |= DMA_PTE_SNP;
 
-	max_addr = (iova & VTD_PAGE_MASK) + VTD_PAGE_ALIGN(size);
+	max_addr = iova + size;
 	if (dmar_domain->max_addr < max_addr) {
 		int min_agaw;
 		u64 end;
@@ -3522,8 +3527,11 @@ static int intel_iommu_map_range(struct iommu_domain *domain,
 		}
 		dmar_domain->max_addr = max_addr;
 	}
-
-	ret = domain_page_mapping(dmar_domain, iova, hpa, size, prot);
+	/* Round up size to next multiple of PAGE_SIZE, if it and
+	   the low bits of hpa would take us onto the next page */
+	size = aligned_nrpages(hpa, size);
+	ret = domain_pfn_mapping(dmar_domain, iova >> VTD_PAGE_SHIFT,
+				 hpa >> VTD_PAGE_SHIFT, size, prot);
 	return ret;
 }
 
@@ -3531,15 +3539,12 @@ static void intel_iommu_unmap_range(struct iommu_domain *domain,
 				    unsigned long iova, size_t size)
 {
 	struct dmar_domain *dmar_domain = domain->priv;
-	dma_addr_t base;
 
-	/* The address might not be aligned */
-	base = iova & VTD_PAGE_MASK;
-	size = VTD_PAGE_ALIGN(size);
-	dma_pte_clear_range(dmar_domain, base, base + size);
+	dma_pte_clear_range(dmar_domain, iova >> VTD_PAGE_SHIFT,
+			    (iova + size - 1) >> VTD_PAGE_SHIFT);
 
-	if (dmar_domain->max_addr == base + size)
-		dmar_domain->max_addr = base;
+	if (dmar_domain->max_addr == iova + size)
+		dmar_domain->max_addr = iova;
 }
 
 static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
@@ -3549,7 +3554,7 @@ static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
 	struct dma_pte *pte;
 	u64 phys = 0;
 
-	pte = addr_to_dma_pte(dmar_domain, iova);
+	pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT);
 	if (pte)
 		phys = dma_pte_addr(pte);