Merge tag 'vfio-v3.11' of git://github.com/awilliam/linux-vfio

Pull vfio updates from Alex Williamson:
 "Largely hugepage support for vfio/type1 iommu and surrounding cleanups
  and fixes"

* tag 'vfio-v3.11' of git://github.com/awilliam/linux-vfio:
  vfio/type1: Fix leak on error path
  vfio: Limit group opens
  vfio/type1: Fix missed frees and zero sized removes
  vfio: fix documentation
  vfio: Provide module option to disable vfio_iommu_type1 hugepage support
  vfio: hugepage support for vfio_iommu_type1
  vfio: Convert type1 iommu to use rbtree

4 files changed, 424 insertions, 230 deletions

diff --git a/Documentation/vfio.txt b/Documentation/vfio.txt
index c55533c0adb3..d7993dcf8537 100644
--- a/Documentation/vfio.txt
+++ b/Documentation/vfio.txt
@@ -172,12 +172,12 @@ group and can access them as follows:
 	struct vfio_device_info device_info = { .argsz = sizeof(device_info) };
 
 	/* Create a new container */
-	container = open("/dev/vfio/vfio, O_RDWR);
+	container = open("/dev/vfio/vfio", O_RDWR);
 
 	if (ioctl(container, VFIO_GET_API_VERSION) != VFIO_API_VERSION)
 		/* Unknown API version */
 
-	if (!ioctl(container, VFIO_CHECK_EXTENSION, VFIO_X86_IOMMU))
+	if (!ioctl(container, VFIO_CHECK_EXTENSION, VFIO_TYPE1_IOMMU))
 		/* Doesn't support the IOMMU driver we want. */
 
 	/* Open the group */
@@ -193,7 +193,7 @@ group and can access them as follows:
 	ioctl(group, VFIO_GROUP_SET_CONTAINER, &container);
 
 	/* Enable the IOMMU model we want */
-	ioctl(container, VFIO_SET_IOMMU, VFIO_X86_IOMMU)
+	ioctl(container, VFIO_SET_IOMMU, VFIO_TYPE1_IOMMU)
 
 	/* Get addition IOMMU info */
 	ioctl(container, VFIO_IOMMU_GET_INFO, &iommu_info);
diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
index 259ad282ae5d..c488da5db7c7 100644
--- a/drivers/vfio/vfio.c
+++ b/drivers/vfio/vfio.c
@@ -76,6 +76,7 @@ struct vfio_group {
 	struct notifier_block		nb;
 	struct list_head		vfio_next;
 	struct list_head		container_next;
+	atomic_t			opened;
 };
 
 struct vfio_device {
@@ -206,6 +207,7 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group)
 	INIT_LIST_HEAD(&group->device_list);
 	mutex_init(&group->device_lock);
 	atomic_set(&group->container_users, 0);
+	atomic_set(&group->opened, 0);
 	group->iommu_group = iommu_group;
 
 	group->nb.notifier_call = vfio_iommu_group_notifier;
@@ -1236,12 +1238,22 @@ static long vfio_group_fops_compat_ioctl(struct file *filep,
 static int vfio_group_fops_open(struct inode *inode, struct file *filep)
 {
 	struct vfio_group *group;
+	int opened;
 
 	group = vfio_group_get_from_minor(iminor(inode));
 	if (!group)
 		return -ENODEV;
 
+	/* Do we need multiple instances of the group open?  Seems not. */
+	opened = atomic_cmpxchg(&group->opened, 0, 1);
+	if (opened) {
+		vfio_group_put(group);
+		return -EBUSY;
+	}
+
+	/* Is something still in use from a previous open? */
 	if (group->container) {
+		atomic_dec(&group->opened);
 		vfio_group_put(group);
 		return -EBUSY;
 	}
@@ -1259,6 +1271,8 @@ static int vfio_group_fops_release(struct inode *inode, struct file *filep)
 
 	vfio_group_try_dissolve_container(group);
 
+	atomic_dec(&group->opened);
+
 	vfio_group_put(group);
 
 	return 0;
diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
index 6f3fbc48a6c7..a9807dea3887 100644
--- a/drivers/vfio/vfio_iommu_type1.c
+++ b/drivers/vfio/vfio_iommu_type1.c
@@ -31,6 +31,7 @@
 #include <linux/module.h>
 #include <linux/mm.h>
 #include <linux/pci.h>		/* pci_bus_type */
+#include <linux/rbtree.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
@@ -47,19 +48,25 @@ module_param_named(allow_unsafe_interrupts,
 MODULE_PARM_DESC(allow_unsafe_interrupts,
 		 "Enable VFIO IOMMU support for on platforms without interrupt remapping support.");
 
+static bool disable_hugepages;
+module_param_named(disable_hugepages,
+		   disable_hugepages, bool, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(disable_hugepages,
+		 "Disable VFIO IOMMU support for IOMMU hugepages.");
+
 struct vfio_iommu {
 	struct iommu_domain	*domain;
 	struct mutex		lock;
-	struct list_head	dma_list;
+	struct rb_root		dma_list;
 	struct list_head	group_list;
 	bool			cache;
 };
 
 struct vfio_dma {
-	struct list_head	next;
+	struct rb_node		node;
 	dma_addr_t		iova;		/* Device address */
 	unsigned long		vaddr;		/* Process virtual addr */
-	long			npage;		/* Number of pages */
+	size_t			size;		/* Map size (bytes) */
 	int			prot;		/* IOMMU_READ/WRITE */
 };
 
@@ -73,7 +80,48 @@ struct vfio_group {
  * into DMA'ble space using the IOMMU
  */
 
-#define NPAGE_TO_SIZE(npage)	((size_t)(npage) << PAGE_SHIFT)
+static struct vfio_dma *vfio_find_dma(struct vfio_iommu *iommu,
+				      dma_addr_t start, size_t size)
+{
+	struct rb_node *node = iommu->dma_list.rb_node;
+
+	while (node) {
+		struct vfio_dma *dma = rb_entry(node, struct vfio_dma, node);
+
+		if (start + size <= dma->iova)
+			node = node->rb_left;
+		else if (start >= dma->iova + dma->size)
+			node = node->rb_right;
+		else
+			return dma;
+	}
+
+	return NULL;
+}
+
+static void vfio_insert_dma(struct vfio_iommu *iommu, struct vfio_dma *new)
+{
+	struct rb_node **link = &iommu->dma_list.rb_node, *parent = NULL;
+	struct vfio_dma *dma;
+
+	while (*link) {
+		parent = *link;
+		dma = rb_entry(parent, struct vfio_dma, node);
+
+		if (new->iova + new->size <= dma->iova)
+			link = &(*link)->rb_left;
+		else
+			link = &(*link)->rb_right;
+	}
+
+	rb_link_node(&new->node, parent, link);
+	rb_insert_color(&new->node, &iommu->dma_list);
+}
+
+static void vfio_remove_dma(struct vfio_iommu *iommu, struct vfio_dma *old)
+{
+	rb_erase(&old->node, &iommu->dma_list);
+}
 
 struct vwork {
 	struct mm_struct	*mm;
@@ -100,8 +148,8 @@ static void vfio_lock_acct(long npage)
 	struct vwork *vwork;
 	struct mm_struct *mm;
 
-	if (!current->mm)
-		return; /* process exited */
+	if (!current->mm || !npage)
+		return; /* process exited or nothing to do */
 
 	if (down_write_trylock(&current->mm->mmap_sem)) {
 		current->mm->locked_vm += npage;
@@ -173,33 +221,6 @@ static int put_pfn(unsigned long pfn, int prot)
 	return 0;
 }
 
-/* Unmap DMA region */
-static long __vfio_dma_do_unmap(struct vfio_iommu *iommu, dma_addr_t iova,
-			     long npage, int prot)
-{
-	long i, unlocked = 0;
-
-	for (i = 0; i < npage; i++, iova += PAGE_SIZE) {
-		unsigned long pfn;
-
-		pfn = iommu_iova_to_phys(iommu->domain, iova) >> PAGE_SHIFT;
-		if (pfn) {
-			iommu_unmap(iommu->domain, iova, PAGE_SIZE);
-			unlocked += put_pfn(pfn, prot);
-		}
-	}
-	return unlocked;
-}
-
-static void vfio_dma_unmap(struct vfio_iommu *iommu, dma_addr_t iova,
-			   long npage, int prot)
-{
-	long unlocked;
-
-	unlocked = __vfio_dma_do_unmap(iommu, iova, npage, prot);
-	vfio_lock_acct(-unlocked);
-}
-
 static int vaddr_get_pfn(unsigned long vaddr, int prot, unsigned long *pfn)
 {
 	struct page *page[1];
@@ -226,198 +247,306 @@ static int vaddr_get_pfn(unsigned long vaddr, int prot, unsigned long *pfn)
 	return ret;
 }
 
-/* Map DMA region */
-static int __vfio_dma_map(struct vfio_iommu *iommu, dma_addr_t iova,
-			  unsigned long vaddr, long npage, int prot)
+/*
+ * Attempt to pin pages.  We really don't want to track all the pfns and
+ * the iommu can only map chunks of consecutive pfns anyway, so get the
+ * first page and all consecutive pages with the same locking.
+ */
+static long vfio_pin_pages(unsigned long vaddr, long npage,
+			   int prot, unsigned long *pfn_base)
 {
-	dma_addr_t start = iova;
-	long i, locked = 0;
-	int ret;
+	unsigned long limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
+	bool lock_cap = capable(CAP_IPC_LOCK);
+	long ret, i;
 
-	/* Verify that pages are not already mapped */
-	for (i = 0; i < npage; i++, iova += PAGE_SIZE)
-		if (iommu_iova_to_phys(iommu->domain, iova))
-			return -EBUSY;
+	if (!current->mm)
+		return -ENODEV;
 
-	iova = start;
+	ret = vaddr_get_pfn(vaddr, prot, pfn_base);
+	if (ret)
+		return ret;
 
-	if (iommu->cache)
-		prot |= IOMMU_CACHE;
+	if (is_invalid_reserved_pfn(*pfn_base))
+		return 1;
 
-	/*
-	 * XXX We break mappings into pages and use get_user_pages_fast to
-	 * pin the pages in memory.  It's been suggested that mlock might
-	 * provide a more efficient mechanism, but nothing prevents the
-	 * user from munlocking the pages, which could then allow the user
-	 * access to random host memory.  We also have no guarantee from the
-	 * IOMMU API that the iommu driver can unmap sub-pages of previous
-	 * mappings.  This means we might lose an entire range if a single
-	 * page within it is unmapped.  Single page mappings are inefficient,
-	 * but provide the most flexibility for now.
-	 */
-	for (i = 0; i < npage; i++, iova += PAGE_SIZE, vaddr += PAGE_SIZE) {
+	if (!lock_cap && current->mm->locked_vm + 1 > limit) {
+		put_pfn(*pfn_base, prot);
+		pr_warn("%s: RLIMIT_MEMLOCK (%ld) exceeded\n", __func__,
+			limit << PAGE_SHIFT);
+		return -ENOMEM;
+	}
+
+	if (unlikely(disable_hugepages)) {
+		vfio_lock_acct(1);
+		return 1;
+	}
+
+	/* Lock all the consecutive pages from pfn_base */
+	for (i = 1, vaddr += PAGE_SIZE; i < npage; i++, vaddr += PAGE_SIZE) {
 		unsigned long pfn = 0;
 
 		ret = vaddr_get_pfn(vaddr, prot, &pfn);
-		if (ret) {
-			__vfio_dma_do_unmap(iommu, start, i, prot);
-			return ret;
-		}
+		if (ret)
+			break;
 
-		/*
-		 * Only add actual locked pages to accounting
-		 * XXX We're effectively marking a page locked for every
-		 * IOVA page even though it's possible the user could be
-		 * backing multiple IOVAs with the same vaddr.  This over-
-		 * penalizes the user process, but we currently have no
-		 * easy way to do this properly.
-		 */
-		if (!is_invalid_reserved_pfn(pfn))
-			locked++;
+		if (pfn != *pfn_base + i || is_invalid_reserved_pfn(pfn)) {
+			put_pfn(pfn, prot);
+			break;
+		}
 
-		ret = iommu_map(iommu->domain, iova,
-				(phys_addr_t)pfn << PAGE_SHIFT,
-				PAGE_SIZE, prot);
-		if (ret) {
-			/* Back out mappings on error */
+		if (!lock_cap && current->mm->locked_vm + i + 1 > limit) {
 			put_pfn(pfn, prot);
-			__vfio_dma_do_unmap(iommu, start, i, prot);
-			return ret;
+			pr_warn("%s: RLIMIT_MEMLOCK (%ld) exceeded\n",
+				__func__, limit << PAGE_SHIFT);
+			break;
 		}
 	}
-	vfio_lock_acct(locked);
-	return 0;
+
+	vfio_lock_acct(i);
+
+	return i;
 }
 
-static inline bool ranges_overlap(dma_addr_t start1, size_t size1,
-				  dma_addr_t start2, size_t size2)
+static long vfio_unpin_pages(unsigned long pfn, long npage,
+			     int prot, bool do_accounting)
 {
-	if (start1 < start2)
-		return (start2 - start1 < size1);
-	else if (start2 < start1)
-		return (start1 - start2 < size2);
-	return (size1 > 0 && size2 > 0);
+	unsigned long unlocked = 0;
+	long i;
+
+	for (i = 0; i < npage; i++)
+		unlocked += put_pfn(pfn++, prot);
+
+	if (do_accounting)
+		vfio_lock_acct(-unlocked);
+
+	return unlocked;
 }
 
-static struct vfio_dma *vfio_find_dma(struct vfio_iommu *iommu,
-						dma_addr_t start, size_t size)
+static int vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma,
+			    dma_addr_t iova, size_t *size)
 {
-	struct vfio_dma *dma;
+	dma_addr_t start = iova, end = iova + *size;
+	long unlocked = 0;
 
-	list_for_each_entry(dma, &iommu->dma_list, next) {
-		if (ranges_overlap(dma->iova, NPAGE_TO_SIZE(dma->npage),
-				   start, size))
-			return dma;
+	while (iova < end) {
+		size_t unmapped;
+		phys_addr_t phys;
+
+		/*
+		 * We use the IOMMU to track the physical address.  This
+		 * saves us from having a lot more entries in our mapping
+		 * tree.  The downside is that we don't track the size
+		 * used to do the mapping.  We request unmap of a single
+		 * page, but expect IOMMUs that support large pages to
+		 * unmap a larger chunk.
+		 */
+		phys = iommu_iova_to_phys(iommu->domain, iova);
+		if (WARN_ON(!phys)) {
+			iova += PAGE_SIZE;
+			continue;
+		}
+
+		unmapped = iommu_unmap(iommu->domain, iova, PAGE_SIZE);
+		if (!unmapped)
+			break;
+
+		unlocked += vfio_unpin_pages(phys >> PAGE_SHIFT,
+					     unmapped >> PAGE_SHIFT,
+					     dma->prot, false);
+		iova += unmapped;
 	}
-	return NULL;
+
+	vfio_lock_acct(-unlocked);
+
+	*size = iova - start;
+
+	return 0;
 }
 
-static long vfio_remove_dma_overlap(struct vfio_iommu *iommu, dma_addr_t start,
-				    size_t size, struct vfio_dma *dma)
+static int vfio_remove_dma_overlap(struct vfio_iommu *iommu, dma_addr_t start,
+				   size_t *size, struct vfio_dma *dma)
 {
+	size_t offset, overlap, tmp;
 	struct vfio_dma *split;
-	long npage_lo, npage_hi;
-
-	/* Existing dma region is completely covered, unmap all */
-	if (start <= dma->iova &&
-	    start + size >= dma->iova + NPAGE_TO_SIZE(dma->npage)) {
-		vfio_dma_unmap(iommu, dma->iova, dma->npage, dma->prot);
-		list_del(&dma->next);
-		npage_lo = dma->npage;
+	int ret;
+
+	if (!*size)
+		return 0;
+
+	/*
+	 * Existing dma region is completely covered, unmap all.  This is
+	 * the likely case since userspace tends to map and unmap buffers
+	 * in one shot rather than multiple mappings within a buffer.
+	 */
+	if (likely(start <= dma->iova &&
+		   start + *size >= dma->iova + dma->size)) {
+		*size = dma->size;
+		ret = vfio_unmap_unpin(iommu, dma, dma->iova, size);
+		if (ret)
+			return ret;
+
+		/*
+		 * Did we remove more than we have?  Should never happen
+		 * since a vfio_dma is contiguous in iova and vaddr.
+		 */
+		WARN_ON(*size != dma->size);
+
+		vfio_remove_dma(iommu, dma);
 		kfree(dma);
-		return npage_lo;
+		return 0;
 	}
 
 	/* Overlap low address of existing range */
 	if (start <= dma->iova) {
-		size_t overlap;
+		overlap = start + *size - dma->iova;
+		ret = vfio_unmap_unpin(iommu, dma, dma->iova, &overlap);
+		if (ret)
+			return ret;
 
-		overlap = start + size - dma->iova;
-		npage_lo = overlap >> PAGE_SHIFT;
+		vfio_remove_dma(iommu, dma);
 
-		vfio_dma_unmap(iommu, dma->iova, npage_lo, dma->prot);
-		dma->iova += overlap;
-		dma->vaddr += overlap;
-		dma->npage -= npage_lo;
-		return npage_lo;
+		/*
+		 * Check, we may have removed to whole vfio_dma.  If not
+		 * fixup and re-insert.
+		 */
+		if (overlap < dma->size) {
+			dma->iova += overlap;
+			dma->vaddr += overlap;
+			dma->size -= overlap;
+			vfio_insert_dma(iommu, dma);
+		} else
+			kfree(dma);
+
+		*size = overlap;
+		return 0;
 	}
 
 	/* Overlap high address of existing range */
-	if (start + size >= dma->iova + NPAGE_TO_SIZE(dma->npage)) {
-		size_t overlap;
+	if (start + *size >= dma->iova + dma->size) {
+		offset = start - dma->iova;
+		overlap = dma->size - offset;
 
-		overlap = dma->iova + NPAGE_TO_SIZE(dma->npage) - start;
-		npage_hi = overlap >> PAGE_SHIFT;
+		ret = vfio_unmap_unpin(iommu, dma, start, &overlap);
+		if (ret)
+			return ret;
 
-		vfio_dma_unmap(iommu, start, npage_hi, dma->prot);
-		dma->npage -= npage_hi;
-		return npage_hi;
+		dma->size -= overlap;
+		*size = overlap;
+		return 0;
 	}
 
 	/* Split existing */
-	npage_lo = (start - dma->iova) >> PAGE_SHIFT;
-	npage_hi = dma->npage - (size >> PAGE_SHIFT) - npage_lo;
 
-	split = kzalloc(sizeof *split, GFP_KERNEL);
+	/*
+	 * Allocate our tracking structure early even though it may not
+	 * be used.  An Allocation failure later loses track of pages and
+	 * is more difficult to unwind.
+	 */
+	split = kzalloc(sizeof(*split), GFP_KERNEL);
 	if (!split)
 		return -ENOMEM;
 
-	vfio_dma_unmap(iommu, start, size >> PAGE_SHIFT, dma->prot);
+	offset = start - dma->iova;
+
+	ret = vfio_unmap_unpin(iommu, dma, start, size);
+	if (ret || !*size) {
+		kfree(split);
+		return ret;
+	}
+
+	tmp = dma->size;
 
-	dma->npage = npage_lo;
+	/* Resize the lower vfio_dma in place, before the below insert */
+	dma->size = offset;
 
-	split->npage = npage_hi;
-	split->iova = start + size;
-	split->vaddr = dma->vaddr + NPAGE_TO_SIZE(npage_lo) + size;
-	split->prot = dma->prot;
-	list_add(&split->next, &iommu->dma_list);
-	return size >> PAGE_SHIFT;
+	/* Insert new for remainder, assuming it didn't all get unmapped */
+	if (likely(offset + *size < tmp)) {
+		split->size = tmp - offset - *size;
+		split->iova = dma->iova + offset + *size;
+		split->vaddr = dma->vaddr + offset + *size;
+		split->prot = dma->prot;
+		vfio_insert_dma(iommu, split);
+	} else
+		kfree(split);
+
+	return 0;
 }
 
 static int vfio_dma_do_unmap(struct vfio_iommu *iommu,
 			     struct vfio_iommu_type1_dma_unmap *unmap)
 {
-	long ret = 0, npage = unmap->size >> PAGE_SHIFT;
-	struct vfio_dma *dma, *tmp;
 	uint64_t mask;
+	struct vfio_dma *dma;
+	size_t unmapped = 0, size;
+	int ret = 0;
 
 	mask = ((uint64_t)1 << __ffs(iommu->domain->ops->pgsize_bitmap)) - 1;
 
 	if (unmap->iova & mask)
 		return -EINVAL;
-	if (unmap->size & mask)
+	if (!unmap->size || unmap->size & mask)
 		return -EINVAL;
 
-	/* XXX We still break these down into PAGE_SIZE */
 	WARN_ON(mask & PAGE_MASK);
 
 	mutex_lock(&iommu->lock);
 
-	list_for_each_entry_safe(dma, tmp, &iommu->dma_list, next) {
-		if (ranges_overlap(dma->iova, NPAGE_TO_SIZE(dma->npage),
-				   unmap->iova, unmap->size)) {
-			ret = vfio_remove_dma_overlap(iommu, unmap->iova,
-						      unmap->size, dma);
-			if (ret > 0)
-				npage -= ret;
-			if (ret < 0 || npage == 0)
-				break;
-		}
+	while ((dma = vfio_find_dma(iommu, unmap->iova, unmap->size))) {
+		size = unmap->size;
+		ret = vfio_remove_dma_overlap(iommu, unmap->iova, &size, dma);
+		if (ret || !size)
+			break;
+		unmapped += size;
 	}
+
 	mutex_unlock(&iommu->lock);
-	return ret > 0 ? 0 : (int)ret;
+
+	/*
+	 * We may unmap more than requested, update the unmap struct so
+	 * userspace can know.
+	 */
+	unmap->size = unmapped;
+
+	return ret;
+}
+
+/*
+ * Turns out AMD IOMMU has a page table bug where it won't map large pages
+ * to a region that previously mapped smaller pages.  This should be fixed
+ * soon, so this is just a temporary workaround to break mappings down into
+ * PAGE_SIZE.  Better to map smaller pages than nothing.
+ */
+static int map_try_harder(struct vfio_iommu *iommu, dma_addr_t iova,
+			  unsigned long pfn, long npage, int prot)
+{
+	long i;
+	int ret;
+
+	for (i = 0; i < npage; i++, pfn++, iova += PAGE_SIZE) {
+		ret = iommu_map(iommu->domain, iova,
+				(phys_addr_t)pfn << PAGE_SHIFT,
+				PAGE_SIZE, prot);
+		if (ret)
+			break;
+	}
+
+	for (; i < npage && i > 0; i--, iova -= PAGE_SIZE)
+		iommu_unmap(iommu->domain, iova, PAGE_SIZE);
+
+	return ret;
 }
 
 static int vfio_dma_do_map(struct vfio_iommu *iommu,
 			   struct vfio_iommu_type1_dma_map *map)
 {
-	struct vfio_dma *dma, *pdma = NULL;
-	dma_addr_t iova = map->iova;
-	unsigned long locked, lock_limit, vaddr = map->vaddr;
+	dma_addr_t end, iova;
+	unsigned long vaddr = map->vaddr;
 	size_t size = map->size;
+	long npage;
 	int ret = 0, prot = 0;
 	uint64_t mask;
-	long npage;
+
+	end = map->iova + map->size;
 
 	mask = ((uint64_t)1 << __ffs(iommu->domain->ops->pgsize_bitmap)) - 1;
 
@@ -430,104 +559,144 @@ static int vfio_dma_do_map(struct vfio_iommu *iommu,
 	if (!prot)
 		return -EINVAL; /* No READ/WRITE? */
 
+	if (iommu->cache)
+		prot |= IOMMU_CACHE;
+
 	if (vaddr & mask)
 		return -EINVAL;
-	if (iova & mask)
+	if (map->iova & mask)
 		return -EINVAL;
-	if (size & mask)
+	if (!map->size || map->size & mask)
 		return -EINVAL;
 
-	/* XXX We still break these down into PAGE_SIZE */
 	WARN_ON(mask & PAGE_MASK);
 
 	/* Don't allow IOVA wrap */
-	if (iova + size && iova + size < iova)
+	if (end && end < map->iova)
 		return -EINVAL;
 
 	/* Don't allow virtual address wrap */
-	if (vaddr + size && vaddr + size < vaddr)
-		return -EINVAL;
-
-	npage = size >> PAGE_SHIFT;
-	if (!npage)
+	if (vaddr + map->size && vaddr + map->size < vaddr)
 		return -EINVAL;
 
 	mutex_lock(&iommu->lock);
 
-	if (vfio_find_dma(iommu, iova, size)) {
-		ret = -EBUSY;
-		goto out_lock;
+	if (vfio_find_dma(iommu, map->iova, map->size)) {
+		mutex_unlock(&iommu->lock);
+		return -EEXIST;
 	}
 
-	/* account for locked pages */
-	locked = current->mm->locked_vm + npage;
-	lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
-	if (locked > lock_limit && !capable(CAP_IPC_LOCK)) {
-		pr_warn("%s: RLIMIT_MEMLOCK (%ld) exceeded\n",
-			__func__, rlimit(RLIMIT_MEMLOCK));
-		ret = -ENOMEM;
-		goto out_lock;
-	}
+	for (iova = map->iova; iova < end; iova += size, vaddr += size) {
+		struct vfio_dma *dma = NULL;
+		unsigned long pfn;
+		long i;
+
+		/* Pin a contiguous chunk of memory */
+		npage = vfio_pin_pages(vaddr, (end - iova) >> PAGE_SHIFT,
+				       prot, &pfn);
+		if (npage <= 0) {
+			WARN_ON(!npage);
+			ret = (int)npage;
+			break;
+		}
 
-	ret = __vfio_dma_map(iommu, iova, vaddr, npage, prot);
-	if (ret)
-		goto out_lock;
+		/* Verify pages are not already mapped */
+		for (i = 0; i < npage; i++) {
+			if (iommu_iova_to_phys(iommu->domain,
+					       iova + (i << PAGE_SHIFT))) {
+				vfio_unpin_pages(pfn, npage, prot, true);
+				ret = -EBUSY;
+				break;
+			}
+		}
 
-	/* Check if we abut a region below - nothing below 0 */
-	if (iova) {
-		dma = vfio_find_dma(iommu, iova - 1, 1);
-		if (dma && dma->prot == prot &&
-		    dma->vaddr + NPAGE_TO_SIZE(dma->npage) == vaddr) {
+		ret = iommu_map(iommu->domain, iova,
+				(phys_addr_t)pfn << PAGE_SHIFT,
+				npage << PAGE_SHIFT, prot);
+		if (ret) {
+			if (ret != -EBUSY ||
+			    map_try_harder(iommu, iova, pfn, npage, prot)) {
+				vfio_unpin_pages(pfn, npage, prot, true);
+				break;
+			}
+		}
 
-			dma->npage += npage;
-			iova = dma->iova;
-			vaddr = dma->vaddr;
-			npage = dma->npage;
-			size = NPAGE_TO_SIZE(npage);
+		size = npage << PAGE_SHIFT;
 
-			pdma = dma;
+		/*
+		 * Check if we abut a region below - nothing below 0.
+		 * This is the most likely case when mapping chunks of
+		 * physically contiguous regions within a virtual address
+		 * range.  Update the abutting entry in place since iova
+		 * doesn't change.
+		 */
+		if (likely(iova)) {
+			struct vfio_dma *tmp;
+			tmp = vfio_find_dma(iommu, iova - 1, 1);
+			if (tmp && tmp->prot == prot &&
+			    tmp->vaddr + tmp->size == vaddr) {
+				tmp->size += size;
+				iova = tmp->iova;
+				size = tmp->size;
+				vaddr = tmp->vaddr;
+				dma = tmp;
+			}
+		}
+
+		/*
+		 * Check if we abut a region above - nothing above ~0 + 1.
+		 * If we abut above and below, remove and free.  If only
+		 * abut above, remove, modify, reinsert.
+		 */
+		if (likely(iova + size)) {
+			struct vfio_dma *tmp;
+			tmp = vfio_find_dma(iommu, iova + size, 1);
+			if (tmp && tmp->prot == prot &&
+			    tmp->vaddr == vaddr + size) {
+				vfio_remove_dma(iommu, tmp);
+				if (dma) {
+					dma->size += tmp->size;
+					kfree(tmp);
+				} else {
+					size += tmp->size;
+					tmp->size = size;
+					tmp->iova = iova;
+					tmp->vaddr = vaddr;
+					vfio_insert_dma(iommu, tmp);
+					dma = tmp;
+				}
+			}
 		}
-	}
 
-	/* Check if we abut a region above - nothing above ~0 + 1 */
-	if (iova + size) {
-		dma = vfio_find_dma(iommu, iova + size, 1);
-		if (dma && dma->prot == prot &&
-		    dma->vaddr == vaddr + size) {
+		if (!dma) {
+			dma = kzalloc(sizeof(*dma), GFP_KERNEL);
+			if (!dma) {
+				iommu_unmap(iommu->domain, iova, size);
+				vfio_unpin_pages(pfn, npage, prot, true);
+				ret = -ENOMEM;
+				break;
+			}
 
-			dma->npage += npage;
+			dma->size = size;
 			dma->iova = iova;
 			dma->vaddr = vaddr;
-
-			/*
-			 * If merged above and below, remove previously
-			 * merged entry.  New entry covers it.
-			 */
-			if (pdma) {
-				list_del(&pdma->next);
-				kfree(pdma);
-			}
-			pdma = dma;
+			dma->prot = prot;
+			vfio_insert_dma(iommu, dma);
 		}
 	}
 
-	/* Isolated, new region */
-	if (!pdma) {
-		dma = kzalloc(sizeof *dma, GFP_KERNEL);
-		if (!dma) {
-			ret = -ENOMEM;
-			vfio_dma_unmap(iommu, iova, npage, prot);
-			goto out_lock;
+	if (ret) {
+		struct vfio_dma *tmp;
+		iova = map->iova;
+		size = map->size;
+		while ((tmp = vfio_find_dma(iommu, iova, size))) {
+			int r = vfio_remove_dma_overlap(iommu, iova,
+							&size, tmp);
+			if (WARN_ON(r || !size))
+				break;
 		}
-
-		dma->npage = npage;
-		dma->iova = iova;
-		dma->vaddr = vaddr;
-		dma->prot = prot;
-		list_add(&dma->next, &iommu->dma_list);
 	}
 
-out_lock:
 	mutex_unlock(&iommu->lock);
 	return ret;
 }
@@ -606,7 +775,7 @@ static void *vfio_iommu_type1_open(unsigned long arg)
 		return ERR_PTR(-ENOMEM);
 
 	INIT_LIST_HEAD(&iommu->group_list);
-	INIT_LIST_HEAD(&iommu->dma_list);
+	iommu->dma_list = RB_ROOT;
 	mutex_init(&iommu->lock);
 
 	/*
@@ -640,7 +809,7 @@ static void vfio_iommu_type1_release(void *iommu_data)
 {
 	struct vfio_iommu *iommu = iommu_data;
 	struct vfio_group *group, *group_tmp;
-	struct vfio_dma *dma, *dma_tmp;
+	struct rb_node *node;
 
 	list_for_each_entry_safe(group, group_tmp, &iommu->group_list, next) {
 		iommu_detach_group(iommu->domain, group->iommu_group);
@@ -648,10 +817,12 @@ static void vfio_iommu_type1_release(void *iommu_data)
 		kfree(group);
 	}
 
-	list_for_each_entry_safe(dma, dma_tmp, &iommu->dma_list, next) {
-		vfio_dma_unmap(iommu, dma->iova, dma->npage, dma->prot);
-		list_del(&dma->next);
-		kfree(dma);
+	while ((node = rb_first(&iommu->dma_list))) {
+		struct vfio_dma *dma = rb_entry(node, struct vfio_dma, node);
+		size_t size = dma->size;
+		vfio_remove_dma_overlap(iommu, dma->iova, &size, dma);
+		if (WARN_ON(!size))
+			break;
 	}
 
 	iommu_domain_free(iommu->domain);
@@ -706,6 +877,7 @@ static long vfio_iommu_type1_ioctl(void *iommu_data,
 
 	} else if (cmd == VFIO_IOMMU_UNMAP_DMA) {
 		struct vfio_iommu_type1_dma_unmap unmap;
+		long ret;
 
 		minsz = offsetofend(struct vfio_iommu_type1_dma_unmap, size);
 
@@ -715,7 +887,11 @@ static long vfio_iommu_type1_ioctl(void *iommu_data,
 		if (unmap.argsz < minsz || unmap.flags)
 			return -EINVAL;
 
-		return vfio_dma_do_unmap(iommu, &unmap);
+		ret = vfio_dma_do_unmap(iommu, &unmap);
+		if (ret)
+			return ret;
+
+		return copy_to_user((void __user *)arg, &unmap, minsz);
 	}
 
 	return -ENOTTY;
diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h
index 87ee4f4cff25..916e444e6f74 100644
--- a/include/uapi/linux/vfio.h
+++ b/include/uapi/linux/vfio.h
@@ -362,10 +362,14 @@ struct vfio_iommu_type1_dma_map {
 #define VFIO_IOMMU_MAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 13)
 
 /**
- * VFIO_IOMMU_UNMAP_DMA - _IOW(VFIO_TYPE, VFIO_BASE + 14, struct vfio_dma_unmap)
+ * VFIO_IOMMU_UNMAP_DMA - _IOWR(VFIO_TYPE, VFIO_BASE + 14,
+ *							struct vfio_dma_unmap)
  *
  * Unmap IO virtual addresses using the provided struct vfio_dma_unmap.
- * Caller sets argsz.
+ * Caller sets argsz.  The actual unmapped size is returned in the size
+ * field.  No guarantee is made to the user that arbitrary unmaps of iova
+ * or size different from those used in the original mapping call will
+ * succeed.
  */
 struct vfio_iommu_type1_dma_unmap {
 	__u32	argsz;