Merge branch 'iommu-memory-accounting' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/joro/iommu intoiommufd/for-next

Jason Gunthorpe says:

====================
iommufd follows the same design as KVM and uses memory cgroups to limit
the amount of kernel memory a iommufd file descriptor can pin down. The
various internal data structures already use GFP_KERNEL_ACCOUNT to charge
its own memory.

However, one of the biggest consumers of kernel memory is the IOPTEs
stored under the iommu_domain and these allocations are not tracked.

This series is the first step in fixing it.

The iommu driver contract already includes a 'gfp' argument to the
map_pages op, allowing iommufd to specify GFP_KERNEL_ACCOUNT and then
having the driver allocate the IOPTE tables with that flag will capture a
significant amount of the allocations.

Update the iommu_map() API to pass in the GFP argument, and fix all call
sites. Replace iommu_map_atomic().

Audit the "enterprise" iommu drivers to make sure they do the right thing.
Intel and S390 ignore the GFP argument and always use GFP_ATOMIC. This is
problematic for iommufd anyhow, so fix it. AMD and ARM SMMUv2/3 are
already correct.

A follow up series will be needed to capture the allocations made when the
iommu_domain itself is allocated, which will complete the job.
====================

* 'iommu-memory-accounting' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/s390: Use GFP_KERNEL in sleepable contexts
  iommu/s390: Push the gfp parameter to the kmem_cache_alloc()'s
  iommu/intel: Use GFP_KERNEL in sleepable contexts
  iommu/intel: Support the gfp argument to the map_pages op
  iommu/intel: Add a gfp parameter to alloc_pgtable_page()
  iommufd: Use GFP_KERNEL_ACCOUNT for iommu_map()
  iommu/dma: Use the gfp parameter in __iommu_dma_alloc_noncontiguous()
  iommu: Add a gfp parameter to iommu_map_sg()
  iommu: Remove iommu_map_atomic()
  iommu: Add a gfp parameter to iommu_map()

Link: https://lore.kernel.org/linux-iommu/0-v3-76b587fe28df+6e3-iommu_map_gfp_jgg@nvidia.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

7 files changed, 69 insertions, 63 deletions

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index f798c44e0903..c99e4bc55d8c 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -713,7 +713,7 @@ static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
 	if (!iova)
 		return DMA_MAPPING_ERROR;
 
-	if (iommu_map_atomic(domain, iova, phys - iova_off, size, prot)) {
+	if (iommu_map(domain, iova, phys - iova_off, size, prot, GFP_ATOMIC)) {
 		iommu_dma_free_iova(cookie, iova, size, NULL);
 		return DMA_MAPPING_ERROR;
 	}
@@ -822,7 +822,14 @@ static struct page **__iommu_dma_alloc_noncontiguous(struct device *dev,
 	if (!iova)
 		goto out_free_pages;
 
-	if (sg_alloc_table_from_pages(sgt, pages, count, 0, size, GFP_KERNEL))
+	/*
+	 * Remove the zone/policy flags from the GFP - these are applied to the
+	 * __iommu_dma_alloc_pages() but are not used for the supporting
+	 * internal allocations that follow.
+	 */
+	gfp &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM | __GFP_COMP);
+
+	if (sg_alloc_table_from_pages(sgt, pages, count, 0, size, gfp))
 		goto out_free_iova;
 
 	if (!(ioprot & IOMMU_CACHE)) {
@@ -833,7 +840,8 @@ static struct page **__iommu_dma_alloc_noncontiguous(struct device *dev,
 			arch_dma_prep_coherent(sg_page(sg), sg->length);
 	}
 
-	ret = iommu_map_sg_atomic(domain, iova, sgt->sgl, sgt->orig_nents, ioprot);
+	ret = iommu_map_sg(domain, iova, sgt->sgl, sgt->orig_nents, ioprot,
+			   gfp);
 	if (ret < 0 || ret < size)
 		goto out_free_sg;
 
@@ -1281,7 +1289,7 @@ static int iommu_dma_map_sg(struct device *dev, struct scatterlist *sg,
 	 * We'll leave any physical concatenation to the IOMMU driver's
 	 * implementation - it knows better than we do.
 	 */
-	ret = iommu_map_sg_atomic(domain, iova, sg, nents, prot);
+	ret = iommu_map_sg(domain, iova, sg, nents, prot, GFP_ATOMIC);
 	if (ret < 0 || ret < iova_len)
 		goto out_free_iova;
 
@@ -1615,7 +1623,7 @@ static struct iommu_dma_msi_page *iommu_dma_get_msi_page(struct device *dev,
 	if (!iova)
 		goto out_free_page;
 
-	if (iommu_map(domain, iova, msi_addr, size, prot))
+	if (iommu_map(domain, iova, msi_addr, size, prot, GFP_KERNEL))
 		goto out_free_iova;
 
 	INIT_LIST_HEAD(&msi_page->list);
diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c
index 7cfab5fd5e59..6fd223cf639f 100644
--- a/drivers/iommu/intel/iommu.c
+++ b/drivers/iommu/intel/iommu.c
@@ -362,12 +362,12 @@ static int __init intel_iommu_setup(char *str)
 }
 __setup("intel_iommu=", intel_iommu_setup);
 
-void *alloc_pgtable_page(int node)
+void *alloc_pgtable_page(int node, gfp_t gfp)
 {
 	struct page *page;
 	void *vaddr = NULL;
 
-	page = alloc_pages_node(node, GFP_ATOMIC | __GFP_ZERO, 0);
+	page = alloc_pages_node(node, gfp | __GFP_ZERO, 0);
 	if (page)
 		vaddr = page_address(page);
 	return vaddr;
@@ -612,7 +612,7 @@ struct context_entry *iommu_context_addr(struct intel_iommu *iommu, u8 bus,
 		if (!alloc)
 			return NULL;
 
-		context = alloc_pgtable_page(iommu->node);
+		context = alloc_pgtable_page(iommu->node, GFP_ATOMIC);
 		if (!context)
 			return NULL;
 
@@ -908,7 +908,8 @@ pgtable_walk:
 #endif
 
 static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain,
-				      unsigned long pfn, int *target_level)
+				      unsigned long pfn, int *target_level,
+				      gfp_t gfp)
 {
 	struct dma_pte *parent, *pte;
 	int level = agaw_to_level(domain->agaw);
@@ -935,7 +936,7 @@ static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain,
 		if (!dma_pte_present(pte)) {
 			uint64_t pteval;
 
-			tmp_page = alloc_pgtable_page(domain->nid);
+			tmp_page = alloc_pgtable_page(domain->nid, gfp);
 
 			if (!tmp_page)
 				return NULL;
@@ -1186,7 +1187,7 @@ static int iommu_alloc_root_entry(struct intel_iommu *iommu)
 {
 	struct root_entry *root;
 
-	root = (struct root_entry *)alloc_pgtable_page(iommu->node);
+	root = (struct root_entry *)alloc_pgtable_page(iommu->node, GFP_ATOMIC);
 	if (!root) {
 		pr_err("Allocating root entry for %s failed\n",
 			iommu->name);
@@ -2150,7 +2151,8 @@ static void switch_to_super_page(struct dmar_domain *domain,
 
 	while (start_pfn <= end_pfn) {
 		if (!pte)
-			pte = pfn_to_dma_pte(domain, start_pfn, &level);
+			pte = pfn_to_dma_pte(domain, start_pfn, &level,
+					     GFP_ATOMIC);
 
 		if (dma_pte_present(pte)) {
 			dma_pte_free_pagetable(domain, start_pfn,
@@ -2172,7 +2174,8 @@ static void switch_to_super_page(struct dmar_domain *domain,
 
 static int
 __domain_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
-		 unsigned long phys_pfn, unsigned long nr_pages, int prot)
+		 unsigned long phys_pfn, unsigned long nr_pages, int prot,
+		 gfp_t gfp)
 {
 	struct dma_pte *first_pte = NULL, *pte = NULL;
 	unsigned int largepage_lvl = 0;
@@ -2202,7 +2205,8 @@ __domain_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
 			largepage_lvl = hardware_largepage_caps(domain, iov_pfn,
 					phys_pfn, nr_pages);
 
-			pte = pfn_to_dma_pte(domain, iov_pfn, &largepage_lvl);
+			pte = pfn_to_dma_pte(domain, iov_pfn, &largepage_lvl,
+					     gfp);
 			if (!pte)
 				return -ENOMEM;
 			first_pte = pte;
@@ -2368,7 +2372,7 @@ static int iommu_domain_identity_map(struct dmar_domain *domain,
 
 	return __domain_mapping(domain, first_vpfn,
 				first_vpfn, last_vpfn - first_vpfn + 1,
-				DMA_PTE_READ|DMA_PTE_WRITE);
+				DMA_PTE_READ|DMA_PTE_WRITE, GFP_KERNEL);
 }
 
 static int md_domain_init(struct dmar_domain *domain, int guest_width);
@@ -2676,7 +2680,7 @@ static int copy_context_table(struct intel_iommu *iommu,
 			if (!old_ce)
 				goto out;
 
-			new_ce = alloc_pgtable_page(iommu->node);
+			new_ce = alloc_pgtable_page(iommu->node, GFP_KERNEL);
 			if (!new_ce)
 				goto out_unmap;
 
@@ -4136,7 +4140,7 @@ static int md_domain_init(struct dmar_domain *domain, int guest_width)
 	domain->max_addr = 0;
 
 	/* always allocate the top pgd */
-	domain->pgd = alloc_pgtable_page(domain->nid);
+	domain->pgd = alloc_pgtable_page(domain->nid, GFP_ATOMIC);
 	if (!domain->pgd)
 		return -ENOMEM;
 	domain_flush_cache(domain, domain->pgd, PAGE_SIZE);
@@ -4298,7 +4302,7 @@ static int intel_iommu_map(struct iommu_domain *domain,
 	   the low bits of hpa would take us onto the next page */
 	size = aligned_nrpages(hpa, size);
 	return __domain_mapping(dmar_domain, iova >> VTD_PAGE_SHIFT,
-				hpa >> VTD_PAGE_SHIFT, size, prot);
+				hpa >> VTD_PAGE_SHIFT, size, prot, gfp);
 }
 
 static int intel_iommu_map_pages(struct iommu_domain *domain,
@@ -4333,7 +4337,8 @@ static size_t intel_iommu_unmap(struct iommu_domain *domain,
 
 	/* Cope with horrid API which requires us to unmap more than the
 	   size argument if it happens to be a large-page mapping. */
-	BUG_ON(!pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, &level));
+	BUG_ON(!pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, &level,
+			       GFP_ATOMIC));
 
 	if (size < VTD_PAGE_SIZE << level_to_offset_bits(level))
 		size = VTD_PAGE_SIZE << level_to_offset_bits(level);
@@ -4392,7 +4397,8 @@ static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
 	int level = 0;
 	u64 phys = 0;
 
-	pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, &level);
+	pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, &level,
+			     GFP_ATOMIC);
 	if (pte && dma_pte_present(pte))
 		phys = dma_pte_addr(pte) +
 			(iova & (BIT_MASK(level_to_offset_bits(level) +
diff --git a/drivers/iommu/intel/iommu.h b/drivers/iommu/intel/iommu.h
index 06e61e474856..ca9a035e0110 100644
--- a/drivers/iommu/intel/iommu.h
+++ b/drivers/iommu/intel/iommu.h
@@ -737,7 +737,7 @@ int qi_submit_sync(struct intel_iommu *iommu, struct qi_desc *desc,
 
 extern int dmar_ir_support(void);
 
-void *alloc_pgtable_page(int node);
+void *alloc_pgtable_page(int node, gfp_t gfp);
 void free_pgtable_page(void *vaddr);
 void iommu_flush_write_buffer(struct intel_iommu *iommu);
 struct intel_iommu *device_to_iommu(struct device *dev, u8 *bus, u8 *devfn);
diff --git a/drivers/iommu/intel/pasid.c b/drivers/iommu/intel/pasid.c
index fb3c7020028d..c5bf74e9372d 100644
--- a/drivers/iommu/intel/pasid.c
+++ b/drivers/iommu/intel/pasid.c
@@ -200,7 +200,7 @@ static struct pasid_entry *intel_pasid_get_entry(struct device *dev, u32 pasid)
 retry:
 	entries = get_pasid_table_from_pde(&dir[dir_index]);
 	if (!entries) {
-		entries = alloc_pgtable_page(info->iommu->node);
+		entries = alloc_pgtable_page(info->iommu->node, GFP_ATOMIC);
 		if (!entries)
 			return NULL;
 
diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
index 834e6ecf3e51..dd9b60897c86 100644
--- a/drivers/iommu/iommu.c
+++ b/drivers/iommu/iommu.c
@@ -931,7 +931,7 @@ map_end:
 			if (map_size) {
 				ret = iommu_map(domain, addr - map_size,
 						addr - map_size, map_size,
-						entry->prot);
+						entry->prot, GFP_KERNEL);
 				if (ret)
 					goto out;
 				map_size = 0;
@@ -2384,34 +2384,27 @@ static int __iommu_map(struct iommu_domain *domain, unsigned long iova,
 	return ret;
 }
 
-static int _iommu_map(struct iommu_domain *domain, unsigned long iova,
-		      phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
+int iommu_map(struct iommu_domain *domain, unsigned long iova,
+	      phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	const struct iommu_domain_ops *ops = domain->ops;
 	int ret;
 
+	might_sleep_if(gfpflags_allow_blocking(gfp));
+
+	/* Discourage passing strange GFP flags */
+	if (WARN_ON_ONCE(gfp & (__GFP_COMP | __GFP_DMA | __GFP_DMA32 |
+				__GFP_HIGHMEM)))
+		return -EINVAL;
+
 	ret = __iommu_map(domain, iova, paddr, size, prot, gfp);
 	if (ret == 0 && ops->iotlb_sync_map)
 		ops->iotlb_sync_map(domain, iova, size);
 
 	return ret;
 }
-
-int iommu_map(struct iommu_domain *domain, unsigned long iova,
-	      phys_addr_t paddr, size_t size, int prot)
-{
-	might_sleep();
-	return _iommu_map(domain, iova, paddr, size, prot, GFP_KERNEL);
-}
 EXPORT_SYMBOL_GPL(iommu_map);
 
-int iommu_map_atomic(struct iommu_domain *domain, unsigned long iova,
-	      phys_addr_t paddr, size_t size, int prot)
-{
-	return _iommu_map(domain, iova, paddr, size, prot, GFP_ATOMIC);
-}
-EXPORT_SYMBOL_GPL(iommu_map_atomic);
-
 static size_t __iommu_unmap_pages(struct iommu_domain *domain,
 				  unsigned long iova, size_t size,
 				  struct iommu_iotlb_gather *iotlb_gather)
@@ -2501,9 +2494,9 @@ size_t iommu_unmap_fast(struct iommu_domain *domain,
 }
 EXPORT_SYMBOL_GPL(iommu_unmap_fast);
 
-static ssize_t __iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
-		struct scatterlist *sg, unsigned int nents, int prot,
-		gfp_t gfp)
+ssize_t iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
+		     struct scatterlist *sg, unsigned int nents, int prot,
+		     gfp_t gfp)
 {
 	const struct iommu_domain_ops *ops = domain->ops;
 	size_t len = 0, mapped = 0;
@@ -2511,6 +2504,13 @@ static ssize_t __iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
 	unsigned int i = 0;
 	int ret;
 
+	might_sleep_if(gfpflags_allow_blocking(gfp));
+
+	/* Discourage passing strange GFP flags */
+	if (WARN_ON_ONCE(gfp & (__GFP_COMP | __GFP_DMA | __GFP_DMA32 |
+				__GFP_HIGHMEM)))
+		return -EINVAL;
+
 	while (i <= nents) {
 		phys_addr_t s_phys = sg_phys(sg);
 
@@ -2550,21 +2550,8 @@ out_err:
 
 	return ret;
 }
-
-ssize_t iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
-		     struct scatterlist *sg, unsigned int nents, int prot)
-{
-	might_sleep();
-	return __iommu_map_sg(domain, iova, sg, nents, prot, GFP_KERNEL);
-}
 EXPORT_SYMBOL_GPL(iommu_map_sg);
 
-ssize_t iommu_map_sg_atomic(struct iommu_domain *domain, unsigned long iova,
-		    struct scatterlist *sg, unsigned int nents, int prot)
-{
-	return __iommu_map_sg(domain, iova, sg, nents, prot, GFP_ATOMIC);
-}
-
 /**
  * report_iommu_fault() - report about an IOMMU fault to the IOMMU framework
  * @domain: the iommu domain where the fault has happened
diff --git a/drivers/iommu/iommufd/pages.c b/drivers/iommu/iommufd/pages.c
index 1e1d3509efae..f8d92c9bb65b 100644
--- a/drivers/iommu/iommufd/pages.c
+++ b/drivers/iommu/iommufd/pages.c
@@ -456,7 +456,8 @@ static int batch_iommu_map_small(struct iommu_domain *domain,
 			size % PAGE_SIZE);
 
 	while (size) {
-		rc = iommu_map(domain, iova, paddr, PAGE_SIZE, prot);
+		rc = iommu_map(domain, iova, paddr, PAGE_SIZE, prot,
+			       GFP_KERNEL_ACCOUNT);
 		if (rc)
 			goto err_unmap;
 		iova += PAGE_SIZE;
@@ -500,7 +501,8 @@ static int batch_to_domain(struct pfn_batch *batch, struct iommu_domain *domain,
 		else
 			rc = iommu_map(domain, iova,
 				       PFN_PHYS(batch->pfns[cur]) + page_offset,
-				       next_iova - iova, area->iommu_prot);
+				       next_iova - iova, area->iommu_prot,
+				       GFP_KERNEL_ACCOUNT);
 		if (rc)
 			goto err_unmap;
 		iova = next_iova;
diff --git a/drivers/iommu/s390-iommu.c b/drivers/iommu/s390-iommu.c
index bb00580a30d8..c489714ddc4d 100644
--- a/drivers/iommu/s390-iommu.c
+++ b/drivers/iommu/s390-iommu.c
@@ -50,7 +50,7 @@ static struct iommu_domain *s390_domain_alloc(unsigned domain_type)
 	if (!s390_domain)
 		return NULL;
 
-	s390_domain->dma_table = dma_alloc_cpu_table();
+	s390_domain->dma_table = dma_alloc_cpu_table(GFP_KERNEL);
 	if (!s390_domain->dma_table) {
 		kfree(s390_domain);
 		return NULL;
@@ -258,7 +258,8 @@ static void s390_iommu_iotlb_sync_map(struct iommu_domain *domain,
 
 static int s390_iommu_validate_trans(struct s390_domain *s390_domain,
 				     phys_addr_t pa, dma_addr_t dma_addr,
-				     unsigned long nr_pages, int flags)
+				     unsigned long nr_pages, int flags,
+				     gfp_t gfp)
 {
 	phys_addr_t page_addr = pa & PAGE_MASK;
 	unsigned long *entry;
@@ -266,7 +267,8 @@ static int s390_iommu_validate_trans(struct s390_domain *s390_domain,
 	int rc;
 
 	for (i = 0; i < nr_pages; i++) {
-		entry = dma_walk_cpu_trans(s390_domain->dma_table, dma_addr);
+		entry = dma_walk_cpu_trans(s390_domain->dma_table, dma_addr,
+					   gfp);
 		if (unlikely(!entry)) {
 			rc = -ENOMEM;
 			goto undo_cpu_trans;
@@ -282,7 +284,7 @@ undo_cpu_trans:
 	while (i-- > 0) {
 		dma_addr -= PAGE_SIZE;
 		entry = dma_walk_cpu_trans(s390_domain->dma_table,
-					   dma_addr);
+					   dma_addr, gfp);
 		if (!entry)
 			break;
 		dma_update_cpu_trans(entry, 0, ZPCI_PTE_INVALID);
@@ -299,7 +301,8 @@ static int s390_iommu_invalidate_trans(struct s390_domain *s390_domain,
 	int rc = 0;
 
 	for (i = 0; i < nr_pages; i++) {
-		entry = dma_walk_cpu_trans(s390_domain->dma_table, dma_addr);
+		entry = dma_walk_cpu_trans(s390_domain->dma_table, dma_addr,
+					   GFP_ATOMIC);
 		if (unlikely(!entry)) {
 			rc = -EINVAL;
 			break;
@@ -337,7 +340,7 @@ static int s390_iommu_map_pages(struct iommu_domain *domain,
 		flags |= ZPCI_TABLE_PROTECTED;
 
 	rc = s390_iommu_validate_trans(s390_domain, paddr, iova,
-				       pgcount, flags);
+				       pgcount, flags, gfp);
 	if (!rc)
 		*mapped = size;