[mirror_ubuntu-bionic-kernel.git] / drivers / xen / gntdev.c

/******************************************************************************
 * gntdev.c
 *
 * Device for accessing (in user-space) pages that have been granted by other
 * domains.
 *
 * Copyright (c) 2006-2007, D G Murray.
 *           (c) 2009 Gerd Hoffmann <kraxel@redhat.com>
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */

#undef DEBUG

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/miscdevice.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/mmu_notifier.h>
#include <linux/types.h>
#include <linux/uaccess.h>
#include <linux/sched.h>
#include <linux/spinlock.h>
#include <linux/slab.h>

#include <xen/xen.h>
#include <xen/grant_table.h>
#include <xen/gntdev.h>
#include <asm/xen/hypervisor.h>
#include <asm/xen/hypercall.h>
#include <asm/xen/page.h>

MODULE_LICENSE("GPL");
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*1024;
module_param(limit, int, 0644);
MODULE_PARM_DESC(limit, "Maximum number of grants that may be mapped by "
		"the gntdev device");

static atomic_t pages_mapped = ATOMIC_INIT(0);

struct gntdev_priv {
	struct list_head maps;
	/* lock protects maps from concurrent changes */
	spinlock_t lock;
	struct mm_struct *mm;
	struct mmu_notifier mn;
};

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;
	struct ioctl_gntdev_grant_ref *grants;
	struct gnttab_map_grant_ref   *map_ops;
	struct gnttab_unmap_grant_ref *unmap_ops;
	struct page **pages;
};

/* ------------------------------------------------------------------ */

static void gntdev_print_maps(struct gntdev_priv *priv,
			      char *text, int text_index)
{
#ifdef DEBUG
	struct grant_map *map;

	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,
		       map->index == text_index && text ? text : "");
#endif
}

static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count)
{
	struct grant_map *add;
	int i;

	add = kzalloc(sizeof(struct grant_map), GFP_KERNEL);
	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);
	if (NULL == add->grants    ||
	    NULL == add->map_ops   ||
	    NULL == add->unmap_ops ||
	    NULL == add->pages)
		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->index = 0;
	add->count = count;
	add->priv  = priv;

	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);
	return NULL;
}

static void gntdev_add_map(struct gntdev_priv *priv, struct grant_map *add)
{
	struct grant_map *map;

	list_for_each_entry(map, &priv->maps, next) {
		if (add->index + add->count < map->index) {
			list_add_tail(&add->next, &map->next);
			goto done;
		}
		add->index = map->index + map->count;
	}
	list_add_tail(&add->next, &priv->maps);

done:
	gntdev_print_maps(priv, "[new]", add->index);
}

static struct grant_map *gntdev_find_map_index(struct gntdev_priv *priv,
		int index, int count)
{
	struct grant_map *map;

	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)
			continue;
		return map;
	}
	return NULL;
}

static int gntdev_del_map(struct grant_map *map)
{
	int i;

	if (map->vma)
		return -EBUSY;
	for (i = 0; i < map->count; i++)
		if (map->unmap_ops[i].handle)
			return -EBUSY;

	atomic_sub(map->count, &pages_mapped);
	list_del(&map->next);
	return 0;
}

static void gntdev_free_map(struct grant_map *map)
{
	int i;

	if (!map)
		return;

	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);
}

/* ------------------------------------------------------------------ */

static int find_grant_ptes(pte_t *pte, pgtable_t token,
		unsigned long addr, void *data)
{
	struct grant_map *map = data;
	unsigned int pgnr = (addr - map->vma->vm_start) >> PAGE_SHIFT;
	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,
			  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 */);
	return 0;
}

static int map_grant_pages(struct grant_map *map)
{
	int i, err = 0;

	pr_debug("map %d+%d\n", map->index, map->count);
	err = gnttab_map_refs(map->map_ops, 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;
	}
	return err;
}

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 (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;
	}
	return err;
}

/* ------------------------------------------------------------------ */

static void gntdev_vma_close(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;
}

static int gntdev_vma_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
	pr_debug("vaddr %p, pgoff %ld (shouldn't happen)\n",
			vmf->virtual_address, vmf->pgoff);
	vmf->flags = VM_FAULT_ERROR;
	return 0;
}

static struct vm_operations_struct gntdev_vmops = {
	.close = gntdev_vma_close,
	.fault = gntdev_vma_fault,
};

/* ------------------------------------------------------------------ */

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);
	}
	spin_unlock(&priv->lock);
}

static void mn_invl_page(struct mmu_notifier *mn,
			 struct mm_struct *mm,
			 unsigned long address)
{
	mn_invl_range_start(mn, mm, address, address + PAGE_SIZE);
}

static void mn_release(struct mmu_notifier *mn,
		       struct mm_struct *mm)
{
	struct gntdev_priv *priv = container_of(mn, struct gntdev_priv, mn);
	struct grant_map *map;
	int err;

	spin_lock(&priv->lock);
	list_for_each_entry(map, &priv->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 = {
	.release                = mn_release,
	.invalidate_page        = mn_invl_page,
	.invalidate_range_start = mn_invl_range_start,
};

/* ------------------------------------------------------------------ */

static int gntdev_open(struct inode *inode, struct file *flip)
{
	struct gntdev_priv *priv;
	int ret = 0;

	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
	if (!priv)
		return -ENOMEM;

	INIT_LIST_HEAD(&priv->maps);
	spin_lock_init(&priv->lock);

	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);

	if (ret) {
		kfree(priv);
		return ret;
	}

	flip->private_data = priv;
	pr_debug("priv %p\n", priv);

	return 0;
}

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);

	}
	spin_unlock(&priv->lock);

	mmu_notifier_unregister(&priv->mn, priv->mm);
	kfree(priv);
	return 0;
}

static long gntdev_ioctl_map_grant_ref(struct gntdev_priv *priv,
				       struct ioctl_gntdev_map_grant_ref __user *u)
{
	struct ioctl_gntdev_map_grant_ref op;
	struct grant_map *map;
	int err;

	if (copy_from_user(&op, u, sizeof(op)) != 0)
		return -EFAULT;
	pr_debug("priv %p, add %d\n", priv, op.count);
	if (unlikely(op.count <= 0))
		return -EINVAL;

	err = -ENOMEM;
	map = gntdev_alloc_map(priv, op.count);
	if (!map)
		return err;

	if (copy_from_user(map->grants, &u->refs,
			   sizeof(map->grants[0]) * op.count) != 0) {
		gntdev_free_map(map);
		return err;
	}

	if (unlikely(atomic_add_return(op.count, &pages_mapped) > limit)) {
		pr_debug("can't map: over limit\n");
		gntdev_free_map(map);
		return err;
	}

	spin_lock(&priv->lock);
	gntdev_add_map(priv, map);
	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;
	}
	return 0;
}

static long gntdev_ioctl_unmap_grant_ref(struct gntdev_priv *priv,
					 struct ioctl_gntdev_unmap_grant_ref __user *u)
{
	struct ioctl_gntdev_unmap_grant_ref op;
	struct grant_map *map;
	int err = -ENOENT;

	if (copy_from_user(&op, u, sizeof(op)) != 0)
		return -EFAULT;
	pr_debug("priv %p, del %d+%d\n", priv, (int)op.index, (int)op.count);

	spin_lock(&priv->lock);
	map = gntdev_find_map_index(priv, op.index >> PAGE_SHIFT, op.count);
	if (map)
		err = gntdev_del_map(map);
	spin_unlock(&priv->lock);
	if (!err)
		gntdev_free_map(map);
	return err;
}

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 grant_map *map;

	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;
	}
	op.offset = map->index << PAGE_SHIFT;
	op.count = map->count;
	spin_unlock(&priv->lock);

	if (copy_to_user(u, &op, sizeof(op)) != 0)
		return -EFAULT;
	return 0;
}

static long gntdev_ioctl(struct file *flip,
			 unsigned int cmd, unsigned long arg)
{
	struct gntdev_priv *priv = flip->private_data;
	void __user *ptr = (void __user *)arg;

	switch (cmd) {
	case IOCTL_GNTDEV_MAP_GRANT_REF:
		return gntdev_ioctl_map_grant_ref(priv, ptr);

	case IOCTL_GNTDEV_UNMAP_GRANT_REF:
		return gntdev_ioctl_unmap_grant_ref(priv, ptr);

	case IOCTL_GNTDEV_GET_OFFSET_FOR_VADDR:
		return gntdev_ioctl_get_offset_for_vaddr(priv, ptr);

	default:
		pr_debug("priv %p, unknown cmd %x\n", priv, cmd);
		return -ENOIOCTLCMD;
	}

	return 0;
}

static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
{
	struct gntdev_priv *priv = flip->private_data;
	int index = vma->vm_pgoff;
	int count = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
	struct grant_map *map;
	int err = -EINVAL;

	if ((vma->vm_flags & VM_WRITE) && !(vma->vm_flags & VM_SHARED))
		return -EINVAL;

	pr_debug("map %d+%d at %lx (pgoff %lx)\n",
			index, count, vma->vm_start, vma->vm_pgoff);

	spin_lock(&priv->lock);
	map = gntdev_find_map_index(priv, index, count);
	if (!map)
		goto unlock_out;
	if (map->vma)
		goto unlock_out;
	if (priv->mm != vma->vm_mm) {
		printk(KERN_WARNING "Huh? Other mm?\n");
		goto unlock_out;
	}

	vma->vm_ops = &gntdev_vmops;

	vma->vm_flags |= VM_RESERVED|VM_DONTCOPY|VM_DONTEXPAND|VM_PFNMAP;

	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;

	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;
	}

	err = map_grant_pages(map);
	if (err) {
		printk(KERN_WARNING "map_grant_pages() failure.\n");
		return err;
	}

	map->is_mapped = 1;

	return 0;

unlock_out:
	spin_unlock(&priv->lock);
	return err;
}

static const struct file_operations gntdev_fops = {
	.owner = THIS_MODULE,
	.open = gntdev_open,
	.release = gntdev_release,
	.mmap = gntdev_mmap,
	.unlocked_ioctl = gntdev_ioctl
};

static struct miscdevice gntdev_miscdev = {
	.minor        = MISC_DYNAMIC_MINOR,
	.name         = "xen/gntdev",
	.fops         = &gntdev_fops,
};

/* ------------------------------------------------------------------ */

static int __init gntdev_init(void)
{
	int err;

	if (!xen_domain())
		return -ENODEV;

	err = misc_register(&gntdev_miscdev);
	if (err != 0) {
		printk(KERN_ERR "Could not register gntdev device\n");
		return err;
	}
	return 0;
}

static void __exit gntdev_exit(void)
{
	misc_deregister(&gntdev_miscdev);
}

module_init(gntdev_init);
module_exit(gntdev_exit);

/* ------------------------------------------------------------------ */
Commit	Line	Data
ab31523c GH	1	/******************************************************************************
	2	* gntdev.c
	3	*
	4	* Device for accessing (in user-space) pages that have been granted by other
	5	* domains.
	6	*
	7	* Copyright (c) 2006-2007, D G Murray.
	8	* (c) 2009 Gerd Hoffmann <kraxel@redhat.com>
	9	*
	10	* This program is distributed in the hope that it will be useful,
	11	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	12	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	13	* GNU General Public License for more details.
	14	*
	15	* You should have received a copy of the GNU General Public License
	16	* along with this program; if not, write to the Free Software
	17	* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
	18	*/
	19
	20	#undef DEBUG
	21
	22	#include <linux/module.h>
	23	#include <linux/kernel.h>
	24	#include <linux/init.h>
	25	#include <linux/miscdevice.h>
	26	#include <linux/fs.h>
	27	#include <linux/mm.h>
	28	#include <linux/mman.h>
	29	#include <linux/mmu_notifier.h>
	30	#include <linux/types.h>
	31	#include <linux/uaccess.h>
	32	#include <linux/sched.h>
	33	#include <linux/spinlock.h>
	34	#include <linux/slab.h>
	35
	36	#include <xen/xen.h>
	37	#include <xen/grant_table.h>
	38	#include <xen/gntdev.h>
	39	#include <asm/xen/hypervisor.h>
	40	#include <asm/xen/hypercall.h>
	41	#include <asm/xen/page.h>
	42
	43	MODULE_LICENSE("GPL");
	44	MODULE_AUTHOR("Derek G. Murray <Derek.Murray@cl.cam.ac.uk>, "
	45	"Gerd Hoffmann <kraxel@redhat.com>");
	46	MODULE_DESCRIPTION("User-space granted page access driver");
	47
ef91082e	48	static int limit = 1024*1024;
ab31523c	49	module_param(limit, int, 0644);
ef91082e DDG	50	MODULE_PARM_DESC(limit, "Maximum number of grants that may be mapped by "
	51	"the gntdev device");
	52
	53	static atomic_t pages_mapped = ATOMIC_INIT(0);
ab31523c GH	54
	55	struct gntdev_priv {
	56	struct list_head maps;
ab31523c GH	57	/* lock protects maps from concurrent changes */
	58	spinlock_t lock;
	59	struct mm_struct *mm;
	60	struct mmu_notifier mn;
	61	};
	62
	63	struct grant_map {
	64	struct list_head next;
	65	struct gntdev_priv *priv;
	66	struct vm_area_struct *vma;
	67	int index;
	68	int count;
	69	int flags;
	70	int is_mapped;
	71	struct ioctl_gntdev_grant_ref *grants;
	72	struct gnttab_map_grant_ref *map_ops;
	73	struct gnttab_unmap_grant_ref *unmap_ops;
a12b4eb3	74	struct page **pages;
ab31523c GH	75	};
	76
	77	/* ------------------------------------------------------------------ */
	78
	79	static void gntdev_print_maps(struct gntdev_priv *priv,
	80	char *text, int text_index)
	81	{
	82	#ifdef DEBUG
	83	struct grant_map *map;
	84
ef91082e	85	pr_debug("%s: maps list (priv %p)\n", __func__, priv);
ab31523c GH	86	list_for_each_entry(map, &priv->maps, next)
	87	pr_debug(" index %2d, count %2d %s\n",
	88	map->index, map->count,
	89	map->index == text_index && text ? text : "");
	90	#endif
	91	}
	92
	93	static struct grant_map gntdev_alloc_map(struct gntdev_priv priv, int count)
	94	{
	95	struct grant_map *add;
a12b4eb3	96	int i;
ab31523c GH	97
	98	add = kzalloc(sizeof(struct grant_map), GFP_KERNEL);
	99	if (NULL == add)
	100	return NULL;
	101
	102	add->grants = kzalloc(sizeof(add->grants[0]) * count, GFP_KERNEL);
	103	add->map_ops = kzalloc(sizeof(add->map_ops[0]) * count, GFP_KERNEL);
	104	add->unmap_ops = kzalloc(sizeof(add->unmap_ops[0]) * count, GFP_KERNEL);
a12b4eb3 SS	105	add->pages = kzalloc(sizeof(add->pages[0]) * count, GFP_KERNEL);
	106	if (NULL == add->grants \|\|
	107	NULL == add->map_ops \|\|
	108	NULL == add->unmap_ops \|\|
	109	NULL == add->pages)
ab31523c GH	110	goto err;
ab31523c GH	111
a12b4eb3 SS	112	for (i = 0; i < count; i++) {
	113	add->pages[i] = alloc_page(GFP_KERNEL \| __GFP_HIGHMEM);
	114	if (add->pages[i] == NULL)
	115	goto err;
	116	}
	117
ab31523c GH	118	add->index = 0;
	119	add->count = count;
	120	add->priv = priv;
	121
ab31523c GH	122	return add;
	123
	124	err:
a12b4eb3 SS	125	if (add->pages)
	126	for (i = 0; i < count; i++) {
	127	if (add->pages[i])
	128	__free_page(add->pages[i]);
	129	}
	130	kfree(add->pages);
ab31523c GH	131	kfree(add->grants);
	132	kfree(add->map_ops);
	133	kfree(add->unmap_ops);
	134	kfree(add);
	135	return NULL;
	136	}
	137
	138	static void gntdev_add_map(struct gntdev_priv priv, struct grant_map add)
	139	{
	140	struct grant_map *map;
	141
	142	list_for_each_entry(map, &priv->maps, next) {
	143	if (add->index + add->count < map->index) {
	144	list_add_tail(&add->next, &map->next);
	145	goto done;
	146	}
	147	add->index = map->index + map->count;
	148	}
	149	list_add_tail(&add->next, &priv->maps);
	150
	151	done:
ab31523c GH	152	gntdev_print_maps(priv, "[new]", add->index);
	153	}
	154
	155	static struct grant_map gntdev_find_map_index(struct gntdev_priv priv,
	156	int index, int count)
	157	{
	158	struct grant_map *map;
	159
	160	list_for_each_entry(map, &priv->maps, next) {
	161	if (map->index != index)
	162	continue;
	163	if (map->count != count)
	164	continue;
	165	return map;
	166	}
	167	return NULL;
	168	}
	169
	170	static struct grant_map gntdev_find_map_vaddr(struct gntdev_priv priv,
	171	unsigned long vaddr)
	172	{
	173	struct grant_map *map;
	174
	175	list_for_each_entry(map, &priv->maps, next) {
	176	if (!map->vma)
	177	continue;
	178	if (vaddr < map->vma->vm_start)
	179	continue;
	180	if (vaddr >= map->vma->vm_end)
	181	continue;
	182	return map;
	183	}
	184	return NULL;
	185	}
	186
	187	static int gntdev_del_map(struct grant_map *map)
	188	{
	189	int i;
	190
	191	if (map->vma)
	192	return -EBUSY;
	193	for (i = 0; i < map->count; i++)
	194	if (map->unmap_ops[i].handle)
	195	return -EBUSY;
	196
ef91082e	197	atomic_sub(map->count, &pages_mapped);
ab31523c GH	198	list_del(&map->next);
	199	return 0;
	200	}
	201
	202	static void gntdev_free_map(struct grant_map *map)
	203	{
a12b4eb3 SS	204	int i;
a12b4eb3 SS	205
ab31523c GH	206	if (!map)
ab31523c GH	207	return;
a12b4eb3 SS	208
	209	if (map->pages)
	210	for (i = 0; i < map->count; i++) {
	211	if (map->pages[i])
	212	__free_page(map->pages[i]);
	213	}
	214	kfree(map->pages);
ab31523c GH	215	kfree(map->grants);
	216	kfree(map->map_ops);
	217	kfree(map->unmap_ops);
	218	kfree(map);
	219	}
	220
	221	/* ------------------------------------------------------------------ */
	222
	223	static int find_grant_ptes(pte_t *pte, pgtable_t token,
	224	unsigned long addr, void *data)
	225	{
	226	struct grant_map *map = data;
	227	unsigned int pgnr = (addr - map->vma->vm_start) >> PAGE_SHIFT;
	228	u64 pte_maddr;
	229
	230	BUG_ON(pgnr >= map->count);
ba5d1012 JF	231	pte_maddr = arbitrary_virt_to_machine(pte).maddr;
ba5d1012 JF	232
9329e760 IC	233	gnttab_set_map_op(&map->map_ops[pgnr], pte_maddr,
9329e760 IC	234	GNTMAP_contains_pte \| map->flags,
ab31523c GH	235	map->grants[pgnr].ref,
ab31523c GH	236	map->grants[pgnr].domid);
9329e760 IC	237	gnttab_set_unmap_op(&map->unmap_ops[pgnr], pte_maddr,
9329e760 IC	238	GNTMAP_contains_pte \| map->flags,
ab31523c GH	239	0 /* handle */);
	240	return 0;
	241	}
	242
	243	static int map_grant_pages(struct grant_map *map)
	244	{
	245	int i, err = 0;
	246
	247	pr_debug("map %d+%d\n", map->index, map->count);
a12b4eb3	248	err = gnttab_map_refs(map->map_ops, map->pages, map->count);
ab31523c GH	249	if (err)
	250	return err;
	251
	252	for (i = 0; i < map->count; i++) {
	253	if (map->map_ops[i].status)
	254	err = -EINVAL;
	255	map->unmap_ops[i].handle = map->map_ops[i].handle;
	256	}
	257	return err;
	258	}
	259
	260	static int unmap_grant_pages(struct grant_map *map, int offset, int pages)
	261	{
	262	int i, err = 0;
	263
	264	pr_debug("map %d+%d [%d+%d]\n", map->index, map->count, offset, pages);
a12b4eb3	265	err = gnttab_unmap_refs(map->unmap_ops + offset, map->pages, pages);
ab31523c GH	266	if (err)
	267	return err;
	268
	269	for (i = 0; i < pages; i++) {
	270	if (map->unmap_ops[offset+i].status)
	271	err = -EINVAL;
	272	map->unmap_ops[offset+i].handle = 0;
	273	}
	274	return err;
	275	}
	276
	277	/* ------------------------------------------------------------------ */
	278
	279	static void gntdev_vma_close(struct vm_area_struct *vma)
	280	{
	281	struct grant_map *map = vma->vm_private_data;
	282
	283	pr_debug("close %p\n", vma);
	284	map->is_mapped = 0;
	285	map->vma = NULL;
	286	vma->vm_private_data = NULL;
	287	}
	288
	289	static int gntdev_vma_fault(struct vm_area_struct vma, struct vm_fault vmf)
	290	{
	291	pr_debug("vaddr %p, pgoff %ld (shouldn't happen)\n",
	292	vmf->virtual_address, vmf->pgoff);
	293	vmf->flags = VM_FAULT_ERROR;
	294	return 0;
	295	}
	296
	297	static struct vm_operations_struct gntdev_vmops = {
	298	.close = gntdev_vma_close,
	299	.fault = gntdev_vma_fault,
	300	};
	301
	302	/* ------------------------------------------------------------------ */
	303
	304	static void mn_invl_range_start(struct mmu_notifier *mn,
	305	struct mm_struct *mm,
	306	unsigned long start, unsigned long end)
	307	{
	308	struct gntdev_priv *priv = container_of(mn, struct gntdev_priv, mn);
	309	struct grant_map *map;
	310	unsigned long mstart, mend;
	311	int err;
	312
	313	spin_lock(&priv->lock);
	314	list_for_each_entry(map, &priv->maps, next) {
	315	if (!map->vma)
	316	continue;
	317	if (!map->is_mapped)
	318	continue;
	319	if (map->vma->vm_start >= end)
	320	continue;
	321	if (map->vma->vm_end <= start)
	322	continue;
	323	mstart = max(start, map->vma->vm_start);
	324	mend = min(end, map->vma->vm_end);
	325	pr_debug("map %d+%d (%lx %lx), range %lx %lx, mrange %lx %lx\n",
	326	map->index, map->count,
	327	map->vma->vm_start, map->vma->vm_end,
	328	start, end, mstart, mend);
	329	err = unmap_grant_pages(map,
330	(mstart - map->vma->vm_start) >> PAGE_SHIFT,
331	(mend - mstart) >> PAGE_SHIFT);
332	WARN_ON(err);
333	}
334	spin_unlock(&priv->lock);
335	}
336
337	static void mn_invl_page(struct mmu_notifier *mn,
338	struct mm_struct *mm,
339	unsigned long address)
340	{
341	mn_invl_range_start(mn, mm, address, address + PAGE_SIZE);
342	}
343
344	static void mn_release(struct mmu_notifier *mn,
345	struct mm_struct *mm)
346	{
347	struct gntdev_priv *priv = container_of(mn, struct gntdev_priv, mn);
348	struct grant_map *map;
349	int err;
350
351	spin_lock(&priv->lock);
352	list_for_each_entry(map, &priv->maps, next) {
353	if (!map->vma)
354	continue;
355	pr_debug("map %d+%d (%lx %lx)\n",
356	map->index, map->count,
357	map->vma->vm_start, map->vma->vm_end);
358	err = unmap_grant_pages(map, /* offset */ 0, map->count);
359	WARN_ON(err);
360	}
361	spin_unlock(&priv->lock);
362	}
363
364	struct mmu_notifier_ops gntdev_mmu_ops = {
365	.release = mn_release,
366	.invalidate_page = mn_invl_page,
367	.invalidate_range_start = mn_invl_range_start,
368	};
369
370	/* ------------------------------------------------------------------ */
371
372	static int gntdev_open(struct inode inode, struct file flip)
373	{
374	struct gntdev_priv *priv;
375	int ret = 0;
376
377	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
378	if (!priv)
379	return -ENOMEM;
380
381	INIT_LIST_HEAD(&priv->maps);
382	spin_lock_init(&priv->lock);
ab31523c GH	383
	384	priv->mm = get_task_mm(current);
	385	if (!priv->mm) {
	386	kfree(priv);
	387	return -ENOMEM;
	388	}
	389	priv->mn.ops = &gntdev_mmu_ops;
	390	ret = mmu_notifier_register(&priv->mn, priv->mm);
	391	mmput(priv->mm);
	392
	393	if (ret) {
	394	kfree(priv);
	395	return ret;
	396	}
	397
	398	flip->private_data = priv;
	399	pr_debug("priv %p\n", priv);
	400
	401	return 0;
	402	}
	403
	404	static int gntdev_release(struct inode inode, struct file flip)
	405	{
	406	struct gntdev_priv *priv = flip->private_data;
	407	struct grant_map *map;
	408	int err;
	409
	410	pr_debug("priv %p\n", priv);
	411
	412	spin_lock(&priv->lock);
	413	while (!list_empty(&priv->maps)) {
	414	map = list_entry(priv->maps.next, struct grant_map, next);
	415	err = gntdev_del_map(map);
	416	if (WARN_ON(err))
	417	gntdev_free_map(map);
	418
	419	}
	420	spin_unlock(&priv->lock);
	421
	422	mmu_notifier_unregister(&priv->mn, priv->mm);
	423	kfree(priv);
	424	return 0;
	425	}
	426
	427	static long gntdev_ioctl_map_grant_ref(struct gntdev_priv *priv,
	428	struct ioctl_gntdev_map_grant_ref __user *u)
	429	{
	430	struct ioctl_gntdev_map_grant_ref op;
	431	struct grant_map *map;
	432	int err;
	433
	434	if (copy_from_user(&op, u, sizeof(op)) != 0)
	435	return -EFAULT;
	436	pr_debug("priv %p, add %d\n", priv, op.count);
	437	if (unlikely(op.count <= 0))
	438	return -EINVAL;
ab31523c GH	439
	440	err = -ENOMEM;
	441	map = gntdev_alloc_map(priv, op.count);
	442	if (!map)
	443	return err;
ef91082e	444
ab31523c GH	445	if (copy_from_user(map->grants, &u->refs,
	446	sizeof(map->grants[0]) * op.count) != 0) {
	447	gntdev_free_map(map);
	448	return err;
	449	}
	450
ef91082e DDG	451	if (unlikely(atomic_add_return(op.count, &pages_mapped) > limit)) {
	452	pr_debug("can't map: over limit\n");
	453	gntdev_free_map(map);
	454	return err;
	455	}
	456
ab31523c GH	457	spin_lock(&priv->lock);
	458	gntdev_add_map(priv, map);
	459	op.index = map->index << PAGE_SHIFT;
	460	spin_unlock(&priv->lock);
	461
	462	if (copy_to_user(u, &op, sizeof(op)) != 0) {
	463	spin_lock(&priv->lock);
	464	gntdev_del_map(map);
	465	spin_unlock(&priv->lock);
	466	gntdev_free_map(map);
	467	return err;
	468	}
	469	return 0;
	470	}
	471
	472	static long gntdev_ioctl_unmap_grant_ref(struct gntdev_priv *priv,
	473	struct ioctl_gntdev_unmap_grant_ref __user *u)
	474	{
	475	struct ioctl_gntdev_unmap_grant_ref op;
	476	struct grant_map *map;
	477	int err = -ENOENT;
	478
	479	if (copy_from_user(&op, u, sizeof(op)) != 0)
	480	return -EFAULT;
	481	pr_debug("priv %p, del %d+%d\n", priv, (int)op.index, (int)op.count);
	482
	483	spin_lock(&priv->lock);
	484	map = gntdev_find_map_index(priv, op.index >> PAGE_SHIFT, op.count);
	485	if (map)
	486	err = gntdev_del_map(map);
	487	spin_unlock(&priv->lock);
	488	if (!err)
	489	gntdev_free_map(map);
	490	return err;
	491	}
	492
	493	static long gntdev_ioctl_get_offset_for_vaddr(struct gntdev_priv *priv,
	494	struct ioctl_gntdev_get_offset_for_vaddr __user *u)
	495	{
	496	struct ioctl_gntdev_get_offset_for_vaddr op;
	497	struct grant_map *map;
	498
	499	if (copy_from_user(&op, u, sizeof(op)) != 0)
	500	return -EFAULT;
	501	pr_debug("priv %p, offset for vaddr %lx\n", priv, (unsigned long)op.vaddr);
	502
	503	spin_lock(&priv->lock);
	504	map = gntdev_find_map_vaddr(priv, op.vaddr);
	505	if (map == NULL \|\|
	506	map->vma->vm_start != op.vaddr) {
	507	spin_unlock(&priv->lock);
	508	return -EINVAL;
	509	}
	510	op.offset = map->index << PAGE_SHIFT;
	511	op.count = map->count;
	512	spin_unlock(&priv->lock);
	513
	514	if (copy_to_user(u, &op, sizeof(op)) != 0)
	515	return -EFAULT;
	516	return 0;
	517	}
	518
ab31523c GH	519	static long gntdev_ioctl(struct file *flip,
	520	unsigned int cmd, unsigned long arg)
	521	{
	522	struct gntdev_priv *priv = flip->private_data;
	523	void __user ptr = (void __user )arg;
	524
	525	switch (cmd) {
	526	case IOCTL_GNTDEV_MAP_GRANT_REF:
	527	return gntdev_ioctl_map_grant_ref(priv, ptr);
	528
	529	case IOCTL_GNTDEV_UNMAP_GRANT_REF:
	530	return gntdev_ioctl_unmap_grant_ref(priv, ptr);
	531
	532	case IOCTL_GNTDEV_GET_OFFSET_FOR_VADDR:
	533	return gntdev_ioctl_get_offset_for_vaddr(priv, ptr);
	534
ab31523c GH	535	default:
	536	pr_debug("priv %p, unknown cmd %x\n", priv, cmd);
	537	return -ENOIOCTLCMD;
	538	}
	539
	540	return 0;
	541	}
	542
	543	static int gntdev_mmap(struct file flip, struct vm_area_struct vma)
	544	{
	545	struct gntdev_priv *priv = flip->private_data;
	546	int index = vma->vm_pgoff;
	547	int count = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
	548	struct grant_map *map;
	549	int err = -EINVAL;
	550
	551	if ((vma->vm_flags & VM_WRITE) && !(vma->vm_flags & VM_SHARED))
	552	return -EINVAL;
	553
	554	pr_debug("map %d+%d at %lx (pgoff %lx)\n",
	555	index, count, vma->vm_start, vma->vm_pgoff);
	556
	557	spin_lock(&priv->lock);
	558	map = gntdev_find_map_index(priv, index, count);
	559	if (!map)
	560	goto unlock_out;
	561	if (map->vma)
	562	goto unlock_out;
	563	if (priv->mm != vma->vm_mm) {
	564	printk(KERN_WARNING "Huh? Other mm?\n");
	565	goto unlock_out;
	566	}
	567
	568	vma->vm_ops = &gntdev_vmops;
	569
8d3eaea2	570	vma->vm_flags \|= VM_RESERVED\|VM_DONTCOPY\|VM_DONTEXPAND\|VM_PFNMAP;
ab31523c GH	571
	572	vma->vm_private_data = map;
	573	map->vma = vma;
	574
9329e760	575	map->flags = GNTMAP_host_map \| GNTMAP_application_map;
ab31523c GH	576	if (!(vma->vm_flags & VM_WRITE))
	577	map->flags \|= GNTMAP_readonly;
	578
f0a70c88 DDG	579	spin_unlock(&priv->lock);
f0a70c88 DDG	580
ab31523c GH	581	err = apply_to_page_range(vma->vm_mm, vma->vm_start,
	582	vma->vm_end - vma->vm_start,
	583	find_grant_ptes, map);
	584	if (err) {
	585	printk(KERN_WARNING "find_grant_ptes() failure.\n");
f0a70c88	586	return err;
ab31523c GH	587	}
	588
	589	err = map_grant_pages(map);
	590	if (err) {
	591	printk(KERN_WARNING "map_grant_pages() failure.\n");
f0a70c88	592	return err;
ab31523c	593	}
f0a70c88	594
ab31523c GH	595	map->is_mapped = 1;
ab31523c GH	596
f0a70c88 DDG	597	return 0;
f0a70c88 DDG	598
ab31523c GH	599	unlock_out:
	600	spin_unlock(&priv->lock);
	601	return err;
	602	}
	603
	604	static const struct file_operations gntdev_fops = {
	605	.owner = THIS_MODULE,
	606	.open = gntdev_open,
	607	.release = gntdev_release,
	608	.mmap = gntdev_mmap,
	609	.unlocked_ioctl = gntdev_ioctl
	610	};
	611
	612	static struct miscdevice gntdev_miscdev = {
	613	.minor = MISC_DYNAMIC_MINOR,
	614	.name = "xen/gntdev",
	615	.fops = &gntdev_fops,
	616	};
	617
	618	/* ------------------------------------------------------------------ */
	619
	620	static int __init gntdev_init(void)
	621	{
	622	int err;
	623
	624	if (!xen_domain())
	625	return -ENODEV;
	626
	627	err = misc_register(&gntdev_miscdev);
	628	if (err != 0) {
	629	printk(KERN_ERR "Could not register gntdev device\n");
	630	return err;
	631	}
	632	return 0;
	633	}
	634
	635	static void __exit gntdev_exit(void)
	636	{
	637	misc_deregister(&gntdev_miscdev);
	638	}
	639
	640	module_init(gntdev_init);
	641	module_exit(gntdev_exit);
	642
	643	/* ------------------------------------------------------------------ */