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>

Author: jgunthorpe

arch/arm/mm/dma-mapping.c

@@ -984,7 +984,8 @@ __iommu_create_mapping(struct device *dev, struct page **pages, size_t size,
 
 		len = (j - i) << PAGE_SHIFT;
 		ret = iommu_map(mapping->domain, iova, phys, len,
-				__dma_info_to_prot(DMA_BIDIRECTIONAL, attrs));
+				__dma_info_to_prot(DMA_BIDIRECTIONAL, attrs),
+				GFP_KERNEL);
 		if (ret < 0)
 			goto fail;
 		iova += len;
@@ -1207,7 +1208,8 @@ static int __map_sg_chunk(struct device *dev, struct scatterlist *sg,
 
 		prot = __dma_info_to_prot(dir, attrs);
 
-		ret = iommu_map(mapping->domain, iova, phys, len, prot);
+		ret = iommu_map(mapping->domain, iova, phys, len, prot,
+				GFP_KERNEL);
 		if (ret < 0)
 			goto fail;
 		count += len >> PAGE_SHIFT;
@@ -1379,7 +1381,8 @@ static dma_addr_t arm_iommu_map_page(struct device *dev, struct page *page,
 
 	prot = __dma_info_to_prot(dir, attrs);
 
-	ret = iommu_map(mapping->domain, dma_addr, page_to_phys(page), len, prot);
+	ret = iommu_map(mapping->domain, dma_addr, page_to_phys(page), len,
+			prot, GFP_KERNEL);
 	if (ret < 0)
 		goto fail;
 
@@ -1443,7 +1446,7 @@ static dma_addr_t arm_iommu_map_resource(struct device *dev,
 
 	prot = __dma_info_to_prot(dir, attrs) | IOMMU_MMIO;
 
-	ret = iommu_map(mapping->domain, dma_addr, addr, len, prot);
+	ret = iommu_map(mapping->domain, dma_addr, addr, len, prot, GFP_KERNEL);
 	if (ret < 0)
 		goto fail;
 

arch/s390/include/asm/pci_dma.h

@@ -186,9 +186,10 @@ static inline unsigned long *get_st_pto(unsigned long entry)
 
 /* Prototypes */
 void dma_free_seg_table(unsigned long);
-unsigned long *dma_alloc_cpu_table(void);
+unsigned long *dma_alloc_cpu_table(gfp_t gfp);
 void dma_cleanup_tables(unsigned long *);
-unsigned long *dma_walk_cpu_trans(unsigned long *rto, dma_addr_t dma_addr);
+unsigned long *dma_walk_cpu_trans(unsigned long *rto, dma_addr_t dma_addr,
+				  gfp_t gfp);
 void dma_update_cpu_trans(unsigned long *entry, phys_addr_t page_addr, int flags);
 
 extern const struct dma_map_ops s390_pci_dma_ops;

arch/s390/pci/pci_dma.c

@@ -27,11 +27,11 @@ static int zpci_refresh_global(struct zpci_dev *zdev)
 				  zdev->iommu_pages * PAGE_SIZE);
 }
 
-unsigned long *dma_alloc_cpu_table(void)
+unsigned long *dma_alloc_cpu_table(gfp_t gfp)
 {
 	unsigned long *table, *entry;
 
-	table = kmem_cache_alloc(dma_region_table_cache, GFP_ATOMIC);
+	table = kmem_cache_alloc(dma_region_table_cache, gfp);
 	if (!table)
 		return NULL;
 
@@ -45,11 +45,11 @@ static void dma_free_cpu_table(void *table)
 	kmem_cache_free(dma_region_table_cache, table);
 }
 
-static unsigned long *dma_alloc_page_table(void)
+static unsigned long *dma_alloc_page_table(gfp_t gfp)
 {
 	unsigned long *table, *entry;
 
-	table = kmem_cache_alloc(dma_page_table_cache, GFP_ATOMIC);
+	table = kmem_cache_alloc(dma_page_table_cache, gfp);
 	if (!table)
 		return NULL;
 
@@ -63,7 +63,7 @@ static void dma_free_page_table(void *table)
 	kmem_cache_free(dma_page_table_cache, table);
 }
 
-static unsigned long *dma_get_seg_table_origin(unsigned long *rtep)
+static unsigned long *dma_get_seg_table_origin(unsigned long *rtep, gfp_t gfp)
 {
 	unsigned long old_rte, rte;
 	unsigned long *sto;
@@ -72,7 +72,7 @@ static unsigned long *dma_get_seg_table_origin(unsigned long *rtep)
 	if (reg_entry_isvalid(rte)) {
 		sto = get_rt_sto(rte);
 	} else {
-		sto = dma_alloc_cpu_table();
+		sto = dma_alloc_cpu_table(gfp);
 		if (!sto)
 			return NULL;
 
@@ -90,7 +90,7 @@ static unsigned long *dma_get_seg_table_origin(unsigned long *rtep)
 	return sto;
 }
 
-static unsigned long *dma_get_page_table_origin(unsigned long *step)
+static unsigned long *dma_get_page_table_origin(unsigned long *step, gfp_t gfp)
 {
 	unsigned long old_ste, ste;
 	unsigned long *pto;
@@ -99,7 +99,7 @@ static unsigned long *dma_get_page_table_origin(unsigned long *step)
 	if (reg_entry_isvalid(ste)) {
 		pto = get_st_pto(ste);
 	} else {
-		pto = dma_alloc_page_table();
+		pto = dma_alloc_page_table(gfp);
 		if (!pto)
 			return NULL;
 		set_st_pto(&ste, virt_to_phys(pto));
@@ -116,18 +116,19 @@ static unsigned long *dma_get_page_table_origin(unsigned long *step)
 	return pto;
 }
 
-unsigned long *dma_walk_cpu_trans(unsigned long *rto, dma_addr_t dma_addr)
+unsigned long *dma_walk_cpu_trans(unsigned long *rto, dma_addr_t dma_addr,
+				  gfp_t gfp)
 {
 	unsigned long *sto, *pto;
 	unsigned int rtx, sx, px;
 
 	rtx = calc_rtx(dma_addr);
-	sto = dma_get_seg_table_origin(&rto[rtx]);
+	sto = dma_get_seg_table_origin(&rto[rtx], gfp);
 	if (!sto)
 		return NULL;
 
 	sx = calc_sx(dma_addr);
-	pto = dma_get_page_table_origin(&sto[sx]);
+	pto = dma_get_page_table_origin(&sto[sx], gfp);
 	if (!pto)
 		return NULL;
 
@@ -170,7 +171,8 @@ static int __dma_update_trans(struct zpci_dev *zdev, phys_addr_t pa,
 		return -EINVAL;
 
 	for (i = 0; i < nr_pages; i++) {
-		entry = dma_walk_cpu_trans(zdev->dma_table, dma_addr);
+		entry = dma_walk_cpu_trans(zdev->dma_table, dma_addr,
+					   GFP_ATOMIC);
 		if (!entry) {
 			rc = -ENOMEM;
 			goto undo_cpu_trans;
@@ -186,7 +188,8 @@ static int __dma_update_trans(struct zpci_dev *zdev, phys_addr_t pa,
 		while (i-- > 0) {
 			page_addr -= PAGE_SIZE;
 			dma_addr -= PAGE_SIZE;
-			entry = dma_walk_cpu_trans(zdev->dma_table, dma_addr);
+			entry = dma_walk_cpu_trans(zdev->dma_table, dma_addr,
+						   GFP_ATOMIC);
 			if (!entry)
 				break;
 			dma_update_cpu_trans(entry, page_addr, flags);
@@ -576,7 +579,7 @@ int zpci_dma_init_device(struct zpci_dev *zdev)
 
 	spin_lock_init(&zdev->iommu_bitmap_lock);
 
-	zdev->dma_table = dma_alloc_cpu_table();
+	zdev->dma_table = dma_alloc_cpu_table(GFP_KERNEL);
 	if (!zdev->dma_table) {
 		rc = -ENOMEM;
 		goto out;

drivers/gpu/drm/nouveau/nvkm/subdev/instmem/gk20a.c

@@ -475,7 +475,8 @@ gk20a_instobj_ctor_iommu(struct gk20a_instmem *imem, u32 npages, u32 align,
 		u32 offset = (r->offset + i) << imem->iommu_pgshift;
 
 		ret = iommu_map(imem->domain, offset, node->dma_addrs[i],
-				PAGE_SIZE, IOMMU_READ | IOMMU_WRITE);
+				PAGE_SIZE, IOMMU_READ | IOMMU_WRITE,
+				GFP_KERNEL);
 		if (ret < 0) {
 			nvkm_error(subdev, "IOMMU mapping failure: %d\n", ret);
 

drivers/gpu/drm/tegra/drm.c

@@ -1057,7 +1057,7 @@ void *tegra_drm_alloc(struct tegra_drm *tegra, size_t size, dma_addr_t *dma)
 
 	*dma = iova_dma_addr(&tegra->carveout.domain, alloc);
 	err = iommu_map(tegra->domain, *dma, virt_to_phys(virt),
-			size, IOMMU_READ | IOMMU_WRITE);
+			size, IOMMU_READ | IOMMU_WRITE, GFP_KERNEL);
 	if (err < 0)
 		goto free_iova;
 

drivers/gpu/host1x/cdma.c

@@ -105,7 +105,7 @@ static int host1x_pushbuffer_init(struct push_buffer *pb)
 
 		pb->dma = iova_dma_addr(&host1x->iova, alloc);
 		err = iommu_map(host1x->domain, pb->dma, pb->phys, size,
-				IOMMU_READ);
+				IOMMU_READ, GFP_KERNEL);
 		if (err)
 			goto iommu_free_iova;
 	} else {

drivers/infiniband/hw/usnic/usnic_uiom.c

@@ -277,7 +277,7 @@ static int usnic_uiom_map_sorted_intervals(struct list_head *intervals,
 				usnic_dbg("va 0x%lx pa %pa size 0x%zx flags 0x%x",
 					va_start, &pa_start, size, flags);
 				err = iommu_map(pd->domain, va_start, pa_start,
-							size, flags);
+						size, flags, GFP_KERNEL);
 				if (err) {
 					usnic_err("Failed to map va 0x%lx pa %pa size 0x%zx with err %d\n",
 						va_start, &pa_start, size, err);
@@ -294,7 +294,7 @@ static int usnic_uiom_map_sorted_intervals(struct list_head *intervals,
 				usnic_dbg("va 0x%lx pa %pa size 0x%zx flags 0x%x\n",
 					va_start, &pa_start, size, flags);
 				err = iommu_map(pd->domain, va_start, pa_start,
-						size, flags);
+						size, flags, GFP_KERNEL);
 				if (err) {
 					usnic_err("Failed to map va 0x%lx pa %pa size 0x%zx with err %d\n",
 						va_start, &pa_start, size, err);

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);