1 files changed, 395 insertions, 193 deletions

diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
index 1e31cdcdae1..073b4a19a8b 100644
--- a/drivers/xen/gntdev.c
+++ b/drivers/xen/gntdev.c
@@ -19,6 +19,8 @@
 
 #undef DEBUG
 
+#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
+
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/init.h>
@@ -32,10 +34,13 @@
 #include <linux/sched.h>
 #include <linux/spinlock.h>
 #include <linux/slab.h>
+#include <linux/highmem.h>
 
 #include <xen/xen.h>
 #include <xen/grant_table.h>
+#include <xen/balloon.h>
 #include <xen/gntdev.h>
+#include <xen/events.h>
 #include <asm/xen/hypervisor.h>
 #include <asm/xen/hypercall.h>
 #include <asm/xen/page.h>
@@ -45,35 +50,52 @@ MODULE_AUTHOR("Derek G. Murray <Derek.Murray@cl.cam.ac.uk>, "
 	      "Gerd Hoffmann <kraxel@redhat.com>");
 MODULE_DESCRIPTION("User-space granted page access driver");
 
-static int limit = 1024;
+static int limit = 1024*1024;
 module_param(limit, int, 0644);
-MODULE_PARM_DESC(limit, "Maximum number of grants that may be mapped at "
-		"once by a gntdev instance");
+MODULE_PARM_DESC(limit, "Maximum number of grants that may be mapped by "
+		"the gntdev device");
+
+static atomic_t pages_mapped = ATOMIC_INIT(0);
+
+static int use_ptemod;
+#define populate_freeable_maps use_ptemod
 
 struct gntdev_priv {
+	/* maps with visible offsets in the file descriptor */
 	struct list_head maps;
-	uint32_t used;
-	uint32_t limit;
-	/* lock protects maps from concurrent changes */
+	/* maps that are not visible; will be freed on munmap.
+	 * Only populated if populate_freeable_maps == 1 */
+	struct list_head freeable_maps;
+	/* lock protects maps and freeable_maps */
 	spinlock_t lock;
 	struct mm_struct *mm;
 	struct mmu_notifier mn;
 };
 
+struct unmap_notify {
+	int flags;
+	/* Address relative to the start of the grant_map */
+	int addr;
+	int event;
+};
+
 struct grant_map {
 	struct list_head next;
-	struct gntdev_priv *priv;
 	struct vm_area_struct *vma;
 	int index;
 	int count;
 	int flags;
-	int is_mapped;
+	atomic_t users;
+	struct unmap_notify notify;
 	struct ioctl_gntdev_grant_ref *grants;
 	struct gnttab_map_grant_ref   *map_ops;
 	struct gnttab_unmap_grant_ref *unmap_ops;
+	struct gnttab_map_grant_ref   *kmap_ops;
 	struct page **pages;
 };
 
+static int unmap_grant_pages(struct grant_map *map, int offset, int pages);
+
 /* ------------------------------------------------------------------ */
 
 static void gntdev_print_maps(struct gntdev_priv *priv,
@@ -82,9 +104,7 @@ static void gntdev_print_maps(struct gntdev_priv *priv,
 #ifdef DEBUG
 	struct grant_map *map;
 
-	pr_debug("maps list (priv %p, usage %d/%d)\n",
-	       priv, priv->used, priv->limit);
-
+	pr_debug("%s: maps list (priv %p)\n", __func__, priv);
 	list_for_each_entry(map, &priv->maps, next)
 		pr_debug("  index %2d, count %2d %s\n",
 		       map->index, map->count,
@@ -92,6 +112,21 @@ static void gntdev_print_maps(struct gntdev_priv *priv,
 #endif
 }
 
+static void gntdev_free_map(struct grant_map *map)
+{
+	if (map == NULL)
+		return;
+
+	if (map->pages)
+		free_xenballooned_pages(map->count, map->pages);
+	kfree(map->pages);
+	kfree(map->grants);
+	kfree(map->map_ops);
+	kfree(map->unmap_ops);
+	kfree(map->kmap_ops);
+	kfree(map);
+}
+
 static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count)
 {
 	struct grant_map *add;
@@ -101,42 +136,35 @@ static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count)
 	if (NULL == add)
 		return NULL;
 
-	add->grants    = kzalloc(sizeof(add->grants[0])    * count, GFP_KERNEL);
-	add->map_ops   = kzalloc(sizeof(add->map_ops[0])   * count, GFP_KERNEL);
-	add->unmap_ops = kzalloc(sizeof(add->unmap_ops[0]) * count, GFP_KERNEL);
-	add->pages     = kzalloc(sizeof(add->pages[0])     * count, GFP_KERNEL);
+	add->grants    = kcalloc(count, sizeof(add->grants[0]), GFP_KERNEL);
+	add->map_ops   = kcalloc(count, sizeof(add->map_ops[0]), GFP_KERNEL);
+	add->unmap_ops = kcalloc(count, sizeof(add->unmap_ops[0]), GFP_KERNEL);
+	add->kmap_ops  = kcalloc(count, sizeof(add->kmap_ops[0]), GFP_KERNEL);
+	add->pages     = kcalloc(count, sizeof(add->pages[0]), GFP_KERNEL);
 	if (NULL == add->grants    ||
 	    NULL == add->map_ops   ||
 	    NULL == add->unmap_ops ||
+	    NULL == add->kmap_ops  ||
 	    NULL == add->pages)
 		goto err;
 
+	if (alloc_xenballooned_pages(count, add->pages, false /* lowmem */))
+		goto err;
+
 	for (i = 0; i < count; i++) {
-		add->pages[i] = alloc_page(GFP_KERNEL | __GFP_HIGHMEM);
-		if (add->pages[i] == NULL)
-			goto err;
+		add->map_ops[i].handle = -1;
+		add->unmap_ops[i].handle = -1;
+		add->kmap_ops[i].handle = -1;
 	}
 
 	add->index = 0;
 	add->count = count;
-	add->priv  = priv;
-
-	if (add->count + priv->used > priv->limit)
-		goto err;
+	atomic_set(&add->users, 1);
 
 	return add;
 
 err:
-	if (add->pages)
-		for (i = 0; i < count; i++) {
-			if (add->pages[i])
-				__free_page(add->pages[i]);
-		}
-	kfree(add->pages);
-	kfree(add->grants);
-	kfree(add->map_ops);
-	kfree(add->unmap_ops);
-	kfree(add);
+	gntdev_free_map(add);
 	return NULL;
 }
 
@@ -154,7 +182,6 @@ static void gntdev_add_map(struct gntdev_priv *priv, struct grant_map *add)
 	list_add_tail(&add->next, &priv->maps);
 
 done:
-	priv->used += add->count;
 	gntdev_print_maps(priv, "[new]", add->index);
 }
 
@@ -166,62 +193,37 @@ static struct grant_map *gntdev_find_map_index(struct gntdev_priv *priv,
 	list_for_each_entry(map, &priv->maps, next) {
 		if (map->index != index)
 			continue;
-		if (map->count != count)
-			continue;
-		return map;
-	}
-	return NULL;
-}
-
-static struct grant_map *gntdev_find_map_vaddr(struct gntdev_priv *priv,
-					       unsigned long vaddr)
-{
-	struct grant_map *map;
-
-	list_for_each_entry(map, &priv->maps, next) {
-		if (!map->vma)
-			continue;
-		if (vaddr < map->vma->vm_start)
-			continue;
-		if (vaddr >= map->vma->vm_end)
+		if (count && map->count != count)
 			continue;
 		return map;
 	}
 	return NULL;
 }
 
-static int gntdev_del_map(struct grant_map *map)
+static void gntdev_put_map(struct gntdev_priv *priv, struct grant_map *map)
 {
-	int i;
+	if (!map)
+		return;
 
-	if (map->vma)
-		return -EBUSY;
-	for (i = 0; i < map->count; i++)
-		if (map->unmap_ops[i].handle)
-			return -EBUSY;
+	if (!atomic_dec_and_test(&map->users))
+		return;
 
-	map->priv->used -= map->count;
-	list_del(&map->next);
-	return 0;
-}
+	atomic_sub(map->count, &pages_mapped);
 
-static void gntdev_free_map(struct grant_map *map)
-{
-	int i;
+	if (map->notify.flags & UNMAP_NOTIFY_SEND_EVENT) {
+		notify_remote_via_evtchn(map->notify.event);
+		evtchn_put(map->notify.event);
+	}
 
-	if (!map)
-		return;
+	if (populate_freeable_maps && priv) {
+		spin_lock(&priv->lock);
+		list_del(&map->next);
+		spin_unlock(&priv->lock);
+	}
 
-	if (map->pages)
-		for (i = 0; i < map->count; i++) {
-			if (map->pages[i])
-				__free_page(map->pages[i]);
-		}
-	kfree(map->pages);
-	kfree(map->grants);
-	kfree(map->map_ops);
-	kfree(map->unmap_ops);
-	kfree(map);
+	if (map->pages && !use_ptemod)
+		unmap_grant_pages(map, 0, map->count);
+	gntdev_free_map(map);
 }
 
 /* ------------------------------------------------------------------ */
@@ -231,18 +233,17 @@ static int find_grant_ptes(pte_t *pte, pgtable_t token,
 {
 	struct grant_map *map = data;
 	unsigned int pgnr = (addr - map->vma->vm_start) >> PAGE_SHIFT;
+	int flags = map->flags | GNTMAP_application_map | GNTMAP_contains_pte;
 	u64 pte_maddr;
 
 	BUG_ON(pgnr >= map->count);
 	pte_maddr = arbitrary_virt_to_machine(pte).maddr;
 
-	gnttab_set_map_op(&map->map_ops[pgnr], pte_maddr,
-			  GNTMAP_contains_pte | map->flags,
+	gnttab_set_map_op(&map->map_ops[pgnr], pte_maddr, flags,
 			  map->grants[pgnr].ref,
 			  map->grants[pgnr].domid);
-	gnttab_set_unmap_op(&map->unmap_ops[pgnr], pte_maddr,
-			    GNTMAP_contains_pte | map->flags,
-			    0 /* handle */);
+	gnttab_set_unmap_op(&map->unmap_ops[pgnr], pte_maddr, flags,
+			    -1 /* handle */);
 	return 0;
 }
 
@@ -250,92 +251,194 @@ static int map_grant_pages(struct grant_map *map)
 {
 	int i, err = 0;
 
+	if (!use_ptemod) {
+		/* Note: it could already be mapped */
+		if (map->map_ops[0].handle != -1)
+			return 0;
+		for (i = 0; i < map->count; i++) {
+			unsigned long addr = (unsigned long)
+				pfn_to_kaddr(page_to_pfn(map->pages[i]));
+			gnttab_set_map_op(&map->map_ops[i], addr, map->flags,
+				map->grants[i].ref,
+				map->grants[i].domid);
+			gnttab_set_unmap_op(&map->unmap_ops[i], addr,
+				map->flags, -1 /* handle */);
+		}
+	} else {
+		/*
+		 * Setup the map_ops corresponding to the pte entries pointing
+		 * to the kernel linear addresses of the struct pages.
+		 * These ptes are completely different from the user ptes dealt
+		 * with find_grant_ptes.
+		 */
+		for (i = 0; i < map->count; i++) {
+			unsigned long address = (unsigned long)
+				pfn_to_kaddr(page_to_pfn(map->pages[i]));
+			BUG_ON(PageHighMem(map->pages[i]));
+
+			gnttab_set_map_op(&map->kmap_ops[i], address,
+				map->flags | GNTMAP_host_map,
+				map->grants[i].ref,
+				map->grants[i].domid);
+		}
+	}
+
 	pr_debug("map %d+%d\n", map->index, map->count);
-	err = gnttab_map_refs(map->map_ops, map->pages, map->count);
+	err = gnttab_map_refs(map->map_ops, use_ptemod ? map->kmap_ops : NULL,
+			map->pages, map->count);
 	if (err)
 		return err;
 
 	for (i = 0; i < map->count; i++) {
 		if (map->map_ops[i].status)
 			err = -EINVAL;
-		map->unmap_ops[i].handle = map->map_ops[i].handle;
+		else {
+			BUG_ON(map->map_ops[i].handle == -1);
+			map->unmap_ops[i].handle = map->map_ops[i].handle;
+			pr_debug("map handle=%d\n", map->map_ops[i].handle);
+		}
 	}
 	return err;
 }
 
-static int unmap_grant_pages(struct grant_map *map, int offset, int pages)
+static int __unmap_grant_pages(struct grant_map *map, int offset, int pages)
 {
 	int i, err = 0;
 
-	pr_debug("map %d+%d [%d+%d]\n", map->index, map->count, offset, pages);
-	err = gnttab_unmap_refs(map->unmap_ops + offset, map->pages, pages);
+	if (map->notify.flags & UNMAP_NOTIFY_CLEAR_BYTE) {
+		int pgno = (map->notify.addr >> PAGE_SHIFT);
+		if (pgno >= offset && pgno < offset + pages) {
+			/* No need for kmap, pages are in lowmem */
+			uint8_t *tmp = pfn_to_kaddr(page_to_pfn(map->pages[pgno]));
+			tmp[map->notify.addr & (PAGE_SIZE-1)] = 0;
+			map->notify.flags &= ~UNMAP_NOTIFY_CLEAR_BYTE;
+		}
+	}
+
+	err = gnttab_unmap_refs(map->unmap_ops + offset,
+			use_ptemod ? map->kmap_ops + offset : NULL, map->pages + offset,
+			pages);
 	if (err)
 		return err;
 
 	for (i = 0; i < pages; i++) {
 		if (map->unmap_ops[offset+i].status)
 			err = -EINVAL;
-		map->unmap_ops[offset+i].handle = 0;
+		pr_debug("unmap handle=%d st=%d\n",
+			map->unmap_ops[offset+i].handle,
+			map->unmap_ops[offset+i].status);
+		map->unmap_ops[offset+i].handle = -1;
+	}
+	return err;
+}
+
+static int unmap_grant_pages(struct grant_map *map, int offset, int pages)
+{
+	int range, err = 0;
+
+	pr_debug("unmap %d+%d [%d+%d]\n", map->index, map->count, offset, pages);
+
+	/* It is possible the requested range will have a "hole" where we
+	 * already unmapped some of the grants. Only unmap valid ranges.
+	 */
+	while (pages && !err) {
+		while (pages && map->unmap_ops[offset].handle == -1) {
+			offset++;
+			pages--;
+		}
+		range = 0;
+		while (range < pages) {
+			if (map->unmap_ops[offset+range].handle == -1) {
+				range--;
+				break;
+			}
+			range++;
+		}
+		err = __unmap_grant_pages(map, offset, range);
+		offset += range;
+		pages -= range;
 	}
+
 	return err;
 }
 
 /* ------------------------------------------------------------------ */
 
-static void gntdev_vma_close(struct vm_area_struct *vma)
+static void gntdev_vma_open(struct vm_area_struct *vma)
 {
 	struct grant_map *map = vma->vm_private_data;
 
-	pr_debug("close %p\n", vma);
-	map->is_mapped = 0;
-	map->vma = NULL;
-	vma->vm_private_data = NULL;
+	pr_debug("gntdev_vma_open %p\n", vma);
+	atomic_inc(&map->users);
 }
 
-static int gntdev_vma_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+static void gntdev_vma_close(struct vm_area_struct *vma)
 {
-	pr_debug("vaddr %p, pgoff %ld (shouldn't happen)\n",
-			vmf->virtual_address, vmf->pgoff);
-	vmf->flags = VM_FAULT_ERROR;
-	return 0;
+	struct grant_map *map = vma->vm_private_data;
+	struct file *file = vma->vm_file;
+	struct gntdev_priv *priv = file->private_data;
+
+	pr_debug("gntdev_vma_close %p\n", vma);
+	if (use_ptemod) {
+		/* It is possible that an mmu notifier could be running
+		 * concurrently, so take priv->lock to ensure that the vma won't
+		 * vanishing during the unmap_grant_pages call, since we will
+		 * spin here until that completes. Such a concurrent call will
+		 * not do any unmapping, since that has been done prior to
+		 * closing the vma, but it may still iterate the unmap_ops list.
+		 */
+		spin_lock(&priv->lock);
+		map->vma = NULL;
+		spin_unlock(&priv->lock);
+	}
+	vma->vm_private_data = NULL;
+	gntdev_put_map(priv, map);
 }
 
 static struct vm_operations_struct gntdev_vmops = {
+	.open = gntdev_vma_open,
 	.close = gntdev_vma_close,
-	.fault = gntdev_vma_fault,
 };
 
 /* ------------------------------------------------------------------ */
 
+static void unmap_if_in_range(struct grant_map *map,
+			      unsigned long start, unsigned long end)
+{
+	unsigned long mstart, mend;
+	int err;
+
+	if (!map->vma)
+		return;
+	if (map->vma->vm_start >= end)
+		return;
+	if (map->vma->vm_end <= start)
+		return;
+	mstart = max(start, map->vma->vm_start);
+	mend   = min(end,   map->vma->vm_end);
+	pr_debug("map %d+%d (%lx %lx), range %lx %lx, mrange %lx %lx\n",
+			map->index, map->count,
+			map->vma->vm_start, map->vma->vm_end,
+			start, end, mstart, mend);
+	err = unmap_grant_pages(map,
+				(mstart - map->vma->vm_start) >> PAGE_SHIFT,
+				(mend - mstart) >> PAGE_SHIFT);
+	WARN_ON(err);
+}
+
 static void mn_invl_range_start(struct mmu_notifier *mn,
 				struct mm_struct *mm,
 				unsigned long start, unsigned long end)
 {
 	struct gntdev_priv *priv = container_of(mn, struct gntdev_priv, mn);
 	struct grant_map *map;
-	unsigned long mstart, mend;
-	int err;
 
 	spin_lock(&priv->lock);
 	list_for_each_entry(map, &priv->maps, next) {
-		if (!map->vma)
-			continue;
-		if (!map->is_mapped)
-			continue;
-		if (map->vma->vm_start >= end)
-			continue;
-		if (map->vma->vm_end <= start)
-			continue;
-		mstart = max(start, map->vma->vm_start);
-		mend   = min(end,   map->vma->vm_end);
-		pr_debug("map %d+%d (%lx %lx), range %lx %lx, mrange %lx %lx\n",
-				map->index, map->count,
-				map->vma->vm_start, map->vma->vm_end,
-				start, end, mstart, mend);
-		err = unmap_grant_pages(map,
-					(mstart - map->vma->vm_start) >> PAGE_SHIFT,
-					(mend - mstart) >> PAGE_SHIFT);
-		WARN_ON(err);
+		unmap_if_in_range(map, start, end);
+	}
+	list_for_each_entry(map, &priv->freeable_maps, next) {
+		unmap_if_in_range(map, start, end);
 	}
 	spin_unlock(&priv->lock);
 }
@@ -364,10 +467,19 @@ static void mn_release(struct mmu_notifier *mn,
 		err = unmap_grant_pages(map, /* offset */ 0, map->count);
 		WARN_ON(err);
 	}
+	list_for_each_entry(map, &priv->freeable_maps, next) {
+		if (!map->vma)
+			continue;
+		pr_debug("map %d+%d (%lx %lx)\n",
+				map->index, map->count,
+				map->vma->vm_start, map->vma->vm_end);
+		err = unmap_grant_pages(map, /* offset */ 0, map->count);
+		WARN_ON(err);
+	}
 	spin_unlock(&priv->lock);
 }
 
-struct mmu_notifier_ops gntdev_mmu_ops = {
+static struct mmu_notifier_ops gntdev_mmu_ops = {
 	.release                = mn_release,
 	.invalidate_page        = mn_invl_page,
 	.invalidate_range_start = mn_invl_range_start,
@@ -385,17 +497,19 @@ static int gntdev_open(struct inode *inode, struct file *flip)
 		return -ENOMEM;
 
 	INIT_LIST_HEAD(&priv->maps);
+	INIT_LIST_HEAD(&priv->freeable_maps);
 	spin_lock_init(&priv->lock);
-	priv->limit = limit;
 
-	priv->mm = get_task_mm(current);
-	if (!priv->mm) {
-		kfree(priv);
-		return -ENOMEM;
+	if (use_ptemod) {
+		priv->mm = get_task_mm(current);
+		if (!priv->mm) {
+			kfree(priv);
+			return -ENOMEM;
+		}
+		priv->mn.ops = &gntdev_mmu_ops;
+		ret = mmu_notifier_register(&priv->mn, priv->mm);
+		mmput(priv->mm);
 	}
-	priv->mn.ops = &gntdev_mmu_ops;
-	ret = mmu_notifier_register(&priv->mn, priv->mm);
-	mmput(priv->mm);
 
 	if (ret) {
 		kfree(priv);
@@ -412,21 +526,18 @@ static int gntdev_release(struct inode *inode, struct file *flip)
 {
 	struct gntdev_priv *priv = flip->private_data;
 	struct grant_map *map;
-	int err;
 
 	pr_debug("priv %p\n", priv);
 
-	spin_lock(&priv->lock);
 	while (!list_empty(&priv->maps)) {
 		map = list_entry(priv->maps.next, struct grant_map, next);
-		err = gntdev_del_map(map);
-		if (WARN_ON(err))
-			gntdev_free_map(map);
-
+		list_del(&map->next);
+		gntdev_put_map(NULL /* already removed */, map);
 	}
-	spin_unlock(&priv->lock);
+	WARN_ON(!list_empty(&priv->freeable_maps));
 
-	mmu_notifier_unregister(&priv->mn, priv->mm);
+	if (use_ptemod)
+		mmu_notifier_unregister(&priv->mn, priv->mm);
 	kfree(priv);
 	return 0;
 }
@@ -443,17 +554,22 @@ static long gntdev_ioctl_map_grant_ref(struct gntdev_priv *priv,
 	pr_debug("priv %p, add %d\n", priv, op.count);
 	if (unlikely(op.count <= 0))
 		return -EINVAL;
-	if (unlikely(op.count > priv->limit))
-		return -EINVAL;
 
 	err = -ENOMEM;
 	map = gntdev_alloc_map(priv, op.count);
 	if (!map)
 		return err;
+
+	if (unlikely(atomic_add_return(op.count, &pages_mapped) > limit)) {
+		pr_debug("can't map: over limit\n");
+		gntdev_put_map(NULL, map);
+		return err;
+	}
+
 	if (copy_from_user(map->grants, &u->refs,
 			   sizeof(map->grants[0]) * op.count) != 0) {
-		gntdev_free_map(map);
-		return err;
+		gntdev_put_map(NULL, map);
+		return -EFAULT;
 	}
 
 	spin_lock(&priv->lock);
@@ -461,13 +577,9 @@ static long gntdev_ioctl_map_grant_ref(struct gntdev_priv *priv,
 	op.index = map->index << PAGE_SHIFT;
 	spin_unlock(&priv->lock);
 
-	if (copy_to_user(u, &op, sizeof(op)) != 0) {
-		spin_lock(&priv->lock);
-		gntdev_del_map(map);
-		spin_unlock(&priv->lock);
-		gntdev_free_map(map);
-		return err;
-	}
+	if (copy_to_user(u, &op, sizeof(op)) != 0)
+		return -EFAULT;
+
 	return 0;
 }
 
@@ -484,11 +596,15 @@ static long gntdev_ioctl_unmap_grant_ref(struct gntdev_priv *priv,
 
 	spin_lock(&priv->lock);
 	map = gntdev_find_map_index(priv, op.index >> PAGE_SHIFT, op.count);
-	if (map)
-		err = gntdev_del_map(map);
+	if (map) {
+		list_del(&map->next);
+		if (populate_freeable_maps)
+			list_add_tail(&map->next, &priv->freeable_maps);
+		err = 0;
+	}
 	spin_unlock(&priv->lock);
-	if (!err)
-		gntdev_free_map(map);
+	if (map)
+		gntdev_put_map(priv, map);
 	return err;
 }
 
@@ -496,43 +612,99 @@ static long gntdev_ioctl_get_offset_for_vaddr(struct gntdev_priv *priv,
 					      struct ioctl_gntdev_get_offset_for_vaddr __user *u)
 {
 	struct ioctl_gntdev_get_offset_for_vaddr op;
+	struct vm_area_struct *vma;
 	struct grant_map *map;
+	int rv = -EINVAL;
 
 	if (copy_from_user(&op, u, sizeof(op)) != 0)
 		return -EFAULT;
 	pr_debug("priv %p, offset for vaddr %lx\n", priv, (unsigned long)op.vaddr);
 
-	spin_lock(&priv->lock);
-	map = gntdev_find_map_vaddr(priv, op.vaddr);
-	if (map == NULL ||
-	    map->vma->vm_start != op.vaddr) {
-		spin_unlock(&priv->lock);
-		return -EINVAL;
-	}
+	down_read(&current->mm->mmap_sem);
+	vma = find_vma(current->mm, op.vaddr);
+	if (!vma || vma->vm_ops != &gntdev_vmops)
+		goto out_unlock;
+
+	map = vma->vm_private_data;
+	if (!map)
+		goto out_unlock;
+
 	op.offset = map->index << PAGE_SHIFT;
 	op.count = map->count;
-	spin_unlock(&priv->lock);
+	rv = 0;
 
-	if (copy_to_user(u, &op, sizeof(op)) != 0)
+ out_unlock:
+	up_read(&current->mm->mmap_sem);
+
+	if (rv == 0 && copy_to_user(u, &op, sizeof(op)) != 0)
 		return -EFAULT;
-	return 0;
+	return rv;
 }
 
-static long gntdev_ioctl_set_max_grants(struct gntdev_priv *priv,
-					struct ioctl_gntdev_set_max_grants __user *u)
+static long gntdev_ioctl_notify(struct gntdev_priv *priv, void __user *u)
 {
-	struct ioctl_gntdev_set_max_grants op;
+	struct ioctl_gntdev_unmap_notify op;
+	struct grant_map *map;
+	int rc;
+	int out_flags;
+	unsigned int out_event;
 
-	if (copy_from_user(&op, u, sizeof(op)) != 0)
+	if (copy_from_user(&op, u, sizeof(op)))
 		return -EFAULT;
-	pr_debug("priv %p, limit %d\n", priv, op.count);
-	if (op.count > limit)
-		return -E2BIG;
+
+	if (op.action & ~(UNMAP_NOTIFY_CLEAR_BYTE|UNMAP_NOTIFY_SEND_EVENT))
+		return -EINVAL;
+
+	/* We need to grab a reference to the event channel we are going to use
+	 * to send the notify before releasing the reference we may already have
+	 * (if someone has called this ioctl twice). This is required so that
+	 * it is possible to change the clear_byte part of the notification
+	 * without disturbing the event channel part, which may now be the last
+	 * reference to that event channel.
+	 */
+	if (op.action & UNMAP_NOTIFY_SEND_EVENT) {
+		if (evtchn_get(op.event_channel_port))
+			return -EINVAL;
+	}
+
+	out_flags = op.action;
+	out_event = op.event_channel_port;
 
 	spin_lock(&priv->lock);
-	priv->limit = op.count;
+
+	list_for_each_entry(map, &priv->maps, next) {
+		uint64_t begin = map->index << PAGE_SHIFT;
+		uint64_t end = (map->index + map->count) << PAGE_SHIFT;
+		if (op.index >= begin && op.index < end)
+			goto found;
+	}
+	rc = -ENOENT;
+	goto unlock_out;
+
+ found:
+	if ((op.action & UNMAP_NOTIFY_CLEAR_BYTE) &&
+			(map->flags & GNTMAP_readonly)) {
+		rc = -EINVAL;
+		goto unlock_out;
+	}
+
+	out_flags = map->notify.flags;
+	out_event = map->notify.event;
+
+	map->notify.flags = op.action;
+	map->notify.addr = op.index - (map->index << PAGE_SHIFT);
+	map->notify.event = op.event_channel_port;
+
+	rc = 0;
+
+ unlock_out:
 	spin_unlock(&priv->lock);
-	return 0;
+
+	/* Drop the reference to the event channel we did not save in the map */
+	if (out_flags & UNMAP_NOTIFY_SEND_EVENT)
+		evtchn_put(out_event);
+
+	return rc;
 }
 
 static long gntdev_ioctl(struct file *flip,
@@ -551,8 +723,8 @@ static long gntdev_ioctl(struct file *flip,
 	case IOCTL_GNTDEV_GET_OFFSET_FOR_VADDR:
 		return gntdev_ioctl_get_offset_for_vaddr(priv, ptr);
 
-	case IOCTL_GNTDEV_SET_MAX_GRANTS:
-		return gntdev_ioctl_set_max_grants(priv, ptr);
+	case IOCTL_GNTDEV_SET_UNMAP_NOTIFY:
+		return gntdev_ioctl_notify(priv, ptr);
 
 	default:
 		pr_debug("priv %p, unknown cmd %x\n", priv, cmd);
@@ -568,7 +740,7 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
 	int index = vma->vm_pgoff;
 	int count = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
 	struct grant_map *map;
-	int err = -EINVAL;
+	int i, err = -EINVAL;
 
 	if ((vma->vm_flags & VM_WRITE) && !(vma->vm_flags & VM_SHARED))
 		return -EINVAL;
@@ -580,47 +752,75 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
 	map = gntdev_find_map_index(priv, index, count);
 	if (!map)
 		goto unlock_out;
-	if (map->vma)
+	if (use_ptemod && map->vma)
 		goto unlock_out;
-	if (priv->mm != vma->vm_mm) {
-		printk(KERN_WARNING "Huh? Other mm?\n");
+	if (use_ptemod && priv->mm != vma->vm_mm) {
+		pr_warn("Huh? Other mm?\n");
 		goto unlock_out;
 	}
 
+	atomic_inc(&map->users);
+
 	vma->vm_ops = &gntdev_vmops;
 
-	vma->vm_flags |= VM_RESERVED|VM_DONTCOPY|VM_DONTEXPAND|VM_PFNMAP;
+	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
+
+	if (use_ptemod)
+		vma->vm_flags |= VM_DONTCOPY;
 
 	vma->vm_private_data = map;
-	map->vma = vma;
 
-	map->flags = GNTMAP_host_map | GNTMAP_application_map;
-	if (!(vma->vm_flags & VM_WRITE))
-		map->flags |= GNTMAP_readonly;
+	if (use_ptemod)
+		map->vma = vma;
+
+	if (map->flags) {
+		if ((vma->vm_flags & VM_WRITE) &&
+				(map->flags & GNTMAP_readonly))
+			goto out_unlock_put;
+	} else {
+		map->flags = GNTMAP_host_map;
+		if (!(vma->vm_flags & VM_WRITE))
+			map->flags |= GNTMAP_readonly;
+	}
 
 	spin_unlock(&priv->lock);
 
-	err = apply_to_page_range(vma->vm_mm, vma->vm_start,
-				  vma->vm_end - vma->vm_start,
-				  find_grant_ptes, map);
-	if (err) {
-		printk(KERN_WARNING "find_grant_ptes() failure.\n");
-		return err;
+	if (use_ptemod) {
+		err = apply_to_page_range(vma->vm_mm, vma->vm_start,
+					  vma->vm_end - vma->vm_start,
+					  find_grant_ptes, map);
+		if (err) {
+			pr_warn("find_grant_ptes() failure.\n");
+			goto out_put_map;
+		}
 	}
 
 	err = map_grant_pages(map);
-	if (err) {
-		printk(KERN_WARNING "map_grant_pages() failure.\n");
-		return err;
-	}
+	if (err)
+		goto out_put_map;
 
-	map->is_mapped = 1;
+	if (!use_ptemod) {
+		for (i = 0; i < count; i++) {
+			err = vm_insert_page(vma, vma->vm_start + i*PAGE_SIZE,
+				map->pages[i]);
+			if (err)
+				goto out_put_map;
+		}
+	}
 
 	return 0;
 
 unlock_out:
 	spin_unlock(&priv->lock);
 	return err;
+
+out_unlock_put:
+	spin_unlock(&priv->lock);
+out_put_map:
+	if (use_ptemod)
+		map->vma = NULL;
+	gntdev_put_map(priv, map);
+	return err;
 }
 
 static const struct file_operations gntdev_fops = {
@@ -646,9 +846,11 @@ static int __init gntdev_init(void)
 	if (!xen_domain())
 		return -ENODEV;
 
+	use_ptemod = !xen_feature(XENFEAT_auto_translated_physmap);
+
 	err = misc_register(&gntdev_miscdev);
 	if (err != 0) {
-		printk(KERN_ERR "Could not register gntdev device\n");
+		pr_err("Could not register gntdev device\n");
 		return err;
 	}
 	return 0;