gntdev.c 27.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
/******************************************************************************
 * 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

22 23
#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt

24 25 26 27 28 29 30 31 32 33 34
#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>
35
#include <linux/sched/mm.h>
36 37
#include <linux/spinlock.h>
#include <linux/slab.h>
38
#include <linux/highmem.h>
39
#include <linux/refcount.h>
40 41 42

#include <xen/xen.h>
#include <xen/grant_table.h>
43
#include <xen/balloon.h>
44
#include <xen/gntdev.h>
45
#include <xen/events.h>
46
#include <xen/page.h>
47 48 49 50 51 52 53 54
#include <asm/xen/hypervisor.h>
#include <asm/xen/hypercall.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");

55
static int limit = 1024*1024;
56
module_param(limit, int, 0644);
57 58 59 60
MODULE_PARM_DESC(limit, "Maximum number of grants that may be mapped by "
		"the gntdev device");

static atomic_t pages_mapped = ATOMIC_INIT(0);
61

62
static int use_ptemod;
63
#define populate_freeable_maps use_ptemod
64

65
struct gntdev_priv {
66
	/* maps with visible offsets in the file descriptor */
67
	struct list_head maps;
68 69 70 71
	/* 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 */
72
	struct mutex lock;
73 74 75 76
	struct mm_struct *mm;
	struct mmu_notifier mn;
};

77 78 79 80 81 82 83
struct unmap_notify {
	int flags;
	/* Address relative to the start of the grant_map */
	int addr;
	int event;
};

84 85 86 87 88 89
struct grant_map {
	struct list_head next;
	struct vm_area_struct *vma;
	int index;
	int count;
	int flags;
90
	refcount_t users;
91
	struct unmap_notify notify;
92 93 94
	struct ioctl_gntdev_grant_ref *grants;
	struct gnttab_map_grant_ref   *map_ops;
	struct gnttab_unmap_grant_ref *unmap_ops;
95
	struct gnttab_map_grant_ref   *kmap_ops;
96
	struct gnttab_unmap_grant_ref *kunmap_ops;
97
	struct page **pages;
98
	unsigned long pages_vm_start;
99 100
};

101 102
static int unmap_grant_pages(struct grant_map *map, int offset, int pages);

103 104 105 106 107 108 109 110
/* ------------------------------------------------------------------ */

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

111
	pr_debug("%s: maps list (priv %p)\n", __func__, priv);
112 113 114 115 116 117 118
	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
}

119 120 121 122 123 124
static void gntdev_free_map(struct grant_map *map)
{
	if (map == NULL)
		return;

	if (map->pages)
125
		gnttab_free_pages(map->count, map->pages);
126 127 128 129 130
	kfree(map->pages);
	kfree(map->grants);
	kfree(map->map_ops);
	kfree(map->unmap_ops);
	kfree(map->kmap_ops);
131
	kfree(map->kunmap_ops);
132 133 134
	kfree(map);
}

135 136 137
static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count)
{
	struct grant_map *add;
138
	int i;
139 140 141 142 143

	add = kzalloc(sizeof(struct grant_map), GFP_KERNEL);
	if (NULL == add)
		return NULL;

144 145 146 147
	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);
148
	add->kunmap_ops = kcalloc(count, sizeof(add->kunmap_ops[0]), GFP_KERNEL);
149
	add->pages     = kcalloc(count, sizeof(add->pages[0]), GFP_KERNEL);
150 151 152
	if (NULL == add->grants    ||
	    NULL == add->map_ops   ||
	    NULL == add->unmap_ops ||
153
	    NULL == add->kmap_ops  ||
154
	    NULL == add->kunmap_ops ||
155
	    NULL == add->pages)
156 157
		goto err;

158
	if (gnttab_alloc_pages(count, add->pages))
159 160
		goto err;

161
	for (i = 0; i < count; i++) {
162 163
		add->map_ops[i].handle = -1;
		add->unmap_ops[i].handle = -1;
164
		add->kmap_ops[i].handle = -1;
165
		add->kunmap_ops[i].handle = -1;
166 167
	}

168 169
	add->index = 0;
	add->count = count;
170
	refcount_set(&add->users, 1);
171 172 173 174

	return add;

err:
175
	gntdev_free_map(add);
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203
	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;
204
		if (count && map->count != count)
205 206 207 208 209 210
			continue;
		return map;
	}
	return NULL;
}

211
static void gntdev_put_map(struct gntdev_priv *priv, struct grant_map *map)
212 213 214
{
	if (!map)
		return;
215

216
	if (!refcount_dec_and_test(&map->users))
217 218 219 220
		return;

	atomic_sub(map->count, &pages_mapped);

221
	if (map->notify.flags & UNMAP_NOTIFY_SEND_EVENT) {
222
		notify_remote_via_evtchn(map->notify.event);
223 224
		evtchn_put(map->notify.event);
	}
225

226
	if (populate_freeable_maps && priv) {
227
		mutex_lock(&priv->lock);
228
		list_del(&map->next);
229
		mutex_unlock(&priv->lock);
230 231
	}

232 233 234
	if (map->pages && !use_ptemod)
		unmap_grant_pages(map, 0, map->count);
	gntdev_free_map(map);
235 236 237 238 239 240 241 242 243
}

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

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;
244
	int flags = map->flags | GNTMAP_application_map | GNTMAP_contains_pte;
245 246 247
	u64 pte_maddr;

	BUG_ON(pgnr >= map->count);
248 249
	pte_maddr = arbitrary_virt_to_machine(pte).maddr;

250 251 252 253 254 255 256 257
	/*
	 * Set the PTE as special to force get_user_pages_fast() fall
	 * back to the slow path.  If this is not supported as part of
	 * the grant map, it will be done afterwards.
	 */
	if (xen_feature(XENFEAT_gnttab_map_avail_bits))
		flags |= (1 << _GNTMAP_guest_avail0);

258
	gnttab_set_map_op(&map->map_ops[pgnr], pte_maddr, flags,
259 260
			  map->grants[pgnr].ref,
			  map->grants[pgnr].domid);
261
	gnttab_set_unmap_op(&map->unmap_ops[pgnr], pte_maddr, flags,
262
			    -1 /* handle */);
263 264 265
	return 0;
}

266 267 268 269 270 271 272 273 274
#ifdef CONFIG_X86
static int set_grant_ptes_as_special(pte_t *pte, pgtable_t token,
				     unsigned long addr, void *data)
{
	set_pte_at(current->mm, addr, pte, pte_mkspecial(*pte));
	return 0;
}
#endif

275 276 277
static int map_grant_pages(struct grant_map *map)
{
	int i, err = 0;
278 279

	if (!use_ptemod) {
280
		/* Note: it could already be mapped */
281
		if (map->map_ops[0].handle != -1)
282
			return 0;
283
		for (i = 0; i < map->count; i++) {
284
			unsigned long addr = (unsigned long)
285 286 287 288 289
				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,
290
				map->flags, -1 /* handle */);
291
		}
292 293 294 295 296 297 298 299 300 301 302 303
	} 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]));

304 305
			gnttab_set_map_op(&map->kmap_ops[i], address,
				map->flags | GNTMAP_host_map,
306 307
				map->grants[i].ref,
				map->grants[i].domid);
308 309
			gnttab_set_unmap_op(&map->kunmap_ops[i], address,
				map->flags | GNTMAP_host_map, -1);
310
		}
311
	}
312 313

	pr_debug("map %d+%d\n", map->index, map->count);
314 315
	err = gnttab_map_refs(map->map_ops, use_ptemod ? map->kmap_ops : NULL,
			map->pages, map->count);
316 317 318 319
	if (err)
		return err;

	for (i = 0; i < map->count; i++) {
320
		if (map->map_ops[i].status) {
321
			err = -EINVAL;
322
			continue;
323
		}
324 325 326 327

		map->unmap_ops[i].handle = map->map_ops[i].handle;
		if (use_ptemod)
			map->kunmap_ops[i].handle = map->kmap_ops[i].handle;
328 329 330 331
	}
	return err;
}

332
static int __unmap_grant_pages(struct grant_map *map, int offset, int pages)
333 334
{
	int i, err = 0;
335
	struct gntab_unmap_queue_data unmap_data;
336

337 338
	if (map->notify.flags & UNMAP_NOTIFY_CLEAR_BYTE) {
		int pgno = (map->notify.addr >> PAGE_SHIFT);
339 340 341
		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]));
342 343 344 345 346
			tmp[map->notify.addr & (PAGE_SIZE-1)] = 0;
			map->notify.flags &= ~UNMAP_NOTIFY_CLEAR_BYTE;
		}
	}

347 348 349 350 351
	unmap_data.unmap_ops = map->unmap_ops + offset;
	unmap_data.kunmap_ops = use_ptemod ? map->kunmap_ops + offset : NULL;
	unmap_data.pages = map->pages + offset;
	unmap_data.count = pages;

352 353 354
	err = gnttab_unmap_refs_sync(&unmap_data);
	if (err)
		return err;
355 356 357 358

	for (i = 0; i < pages; i++) {
		if (map->unmap_ops[offset+i].status)
			err = -EINVAL;
359 360 361 362
		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;
363 364 365 366
	}
	return err;
}

367 368 369 370 371 372 373 374 375 376
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) {
377
		while (pages && map->unmap_ops[offset].handle == -1) {
378 379 380 381 382
			offset++;
			pages--;
		}
		range = 0;
		while (range < pages) {
383
			if (map->unmap_ops[offset+range].handle == -1)
384 385 386 387 388 389 390 391 392 393 394
				break;
			range++;
		}
		err = __unmap_grant_pages(map, offset, range);
		offset += range;
		pages -= range;
	}

	return err;
}

395 396
/* ------------------------------------------------------------------ */

397 398 399 400 401
static void gntdev_vma_open(struct vm_area_struct *vma)
{
	struct grant_map *map = vma->vm_private_data;

	pr_debug("gntdev_vma_open %p\n", vma);
402
	refcount_inc(&map->users);
403 404
}

405 406 407
static void gntdev_vma_close(struct vm_area_struct *vma)
{
	struct grant_map *map = vma->vm_private_data;
408 409
	struct file *file = vma->vm_file;
	struct gntdev_priv *priv = file->private_data;
410

411
	pr_debug("gntdev_vma_close %p\n", vma);
412 413 414 415 416 417 418 419
	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.
		 */
420
		mutex_lock(&priv->lock);
421
		map->vma = NULL;
422
		mutex_unlock(&priv->lock);
423
	}
424
	vma->vm_private_data = NULL;
425
	gntdev_put_map(priv, map);
426 427
}

428 429 430 431 432 433 434 435
static struct page *gntdev_vma_find_special_page(struct vm_area_struct *vma,
						 unsigned long addr)
{
	struct grant_map *map = vma->vm_private_data;

	return map->pages[(addr - map->pages_vm_start) >> PAGE_SHIFT];
}

436
static const struct vm_operations_struct gntdev_vmops = {
437
	.open = gntdev_vma_open,
438
	.close = gntdev_vma_close,
439
	.find_special_page = gntdev_vma_find_special_page,
440 441 442 443
};

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

444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
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);
}

468 469 470 471 472 473 474
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;

475
	mutex_lock(&priv->lock);
476
	list_for_each_entry(map, &priv->maps, next) {
477 478 479 480
		unmap_if_in_range(map, start, end);
	}
	list_for_each_entry(map, &priv->freeable_maps, next) {
		unmap_if_in_range(map, start, end);
481
	}
482
	mutex_unlock(&priv->lock);
483 484 485 486 487 488 489 490 491
}

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;

492
	mutex_lock(&priv->lock);
493 494 495 496 497 498 499 500 501
	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);
	}
502 503 504 505 506 507 508 509 510
	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);
	}
511
	mutex_unlock(&priv->lock);
512 513
}

514
static const struct mmu_notifier_ops gntdev_mmu_ops = {
515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530
	.release                = mn_release,
	.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);
531
	INIT_LIST_HEAD(&priv->freeable_maps);
532
	mutex_init(&priv->lock);
533

534 535 536 537 538 539 540 541 542
	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);
543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562
	}

	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;

	pr_debug("priv %p\n", priv);

563
	mutex_lock(&priv->lock);
564 565
	while (!list_empty(&priv->maps)) {
		map = list_entry(priv->maps.next, struct grant_map, next);
566
		list_del(&map->next);
567
		gntdev_put_map(NULL /* already removed */, map);
568
	}
569
	WARN_ON(!list_empty(&priv->freeable_maps));
570
	mutex_unlock(&priv->lock);
571

572 573
	if (use_ptemod)
		mmu_notifier_unregister(&priv->mn, priv->mm);
574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
	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;
595

596 597
	if (unlikely(atomic_add_return(op.count, &pages_mapped) > limit)) {
		pr_debug("can't map: over limit\n");
598
		gntdev_put_map(NULL, map);
599 600 601
		return err;
	}

602 603
	if (copy_from_user(map->grants, &u->refs,
			   sizeof(map->grants[0]) * op.count) != 0) {
604 605
		gntdev_put_map(NULL, map);
		return -EFAULT;
606 607
	}

608
	mutex_lock(&priv->lock);
609 610
	gntdev_add_map(priv, map);
	op.index = map->index << PAGE_SHIFT;
611
	mutex_unlock(&priv->lock);
612

613 614 615
	if (copy_to_user(u, &op, sizeof(op)) != 0)
		return -EFAULT;

616 617 618 619 620 621 622 623 624 625 626 627 628 629
	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);

630
	mutex_lock(&priv->lock);
631
	map = gntdev_find_map_index(priv, op.index >> PAGE_SHIFT, op.count);
632 633
	if (map) {
		list_del(&map->next);
634 635
		if (populate_freeable_maps)
			list_add_tail(&map->next, &priv->freeable_maps);
636 637
		err = 0;
	}
638
	mutex_unlock(&priv->lock);
639
	if (map)
640
		gntdev_put_map(priv, map);
641 642 643 644 645 646 647
	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;
648
	struct vm_area_struct *vma;
649
	struct grant_map *map;
650
	int rv = -EINVAL;
651 652 653 654 655

	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);

656
	down_read(&current->mm->mmap_sem);
657 658
	vma = find_vma(current->mm, op.vaddr);
	if (!vma || vma->vm_ops != &gntdev_vmops)
659
		goto out_unlock;
660 661 662

	map = vma->vm_private_data;
	if (!map)
663
		goto out_unlock;
664

665 666
	op.offset = map->index << PAGE_SHIFT;
	op.count = map->count;
667
	rv = 0;
668

669 670 671 672
 out_unlock:
	up_read(&current->mm->mmap_sem);

	if (rv == 0 && copy_to_user(u, &op, sizeof(op)) != 0)
673
		return -EFAULT;
674
	return rv;
675 676
}

677 678 679 680 681
static long gntdev_ioctl_notify(struct gntdev_priv *priv, void __user *u)
{
	struct ioctl_gntdev_unmap_notify op;
	struct grant_map *map;
	int rc;
682 683
	int out_flags;
	unsigned int out_event;
684 685 686 687 688 689 690

	if (copy_from_user(&op, u, sizeof(op)))
		return -EFAULT;

	if (op.action & ~(UNMAP_NOTIFY_CLEAR_BYTE|UNMAP_NOTIFY_SEND_EVENT))
		return -EINVAL;

691 692 693 694 695 696 697 698 699 700 701 702 703 704 705
	/* 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;

706
	mutex_lock(&priv->lock);
707 708 709 710 711 712 713 714 715 716 717

	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:
718 719 720 721 722 723
	if ((op.action & UNMAP_NOTIFY_CLEAR_BYTE) &&
			(map->flags & GNTMAP_readonly)) {
		rc = -EINVAL;
		goto unlock_out;
	}

724 725 726
	out_flags = map->notify.flags;
	out_event = map->notify.event;

727 728 729
	map->notify.flags = op.action;
	map->notify.addr = op.index - (map->index << PAGE_SHIFT);
	map->notify.event = op.event_channel_port;
730

731
	rc = 0;
732

733
 unlock_out:
734
	mutex_unlock(&priv->lock);
735 736 737 738 739

	/* 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);

740 741 742
	return rc;
}

743
#define GNTDEV_COPY_BATCH 16
744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942

struct gntdev_copy_batch {
	struct gnttab_copy ops[GNTDEV_COPY_BATCH];
	struct page *pages[GNTDEV_COPY_BATCH];
	s16 __user *status[GNTDEV_COPY_BATCH];
	unsigned int nr_ops;
	unsigned int nr_pages;
};

static int gntdev_get_page(struct gntdev_copy_batch *batch, void __user *virt,
			   bool writeable, unsigned long *gfn)
{
	unsigned long addr = (unsigned long)virt;
	struct page *page;
	unsigned long xen_pfn;
	int ret;

	ret = get_user_pages_fast(addr, 1, writeable, &page);
	if (ret < 0)
		return ret;

	batch->pages[batch->nr_pages++] = page;

	xen_pfn = page_to_xen_pfn(page) + XEN_PFN_DOWN(addr & ~PAGE_MASK);
	*gfn = pfn_to_gfn(xen_pfn);

	return 0;
}

static void gntdev_put_pages(struct gntdev_copy_batch *batch)
{
	unsigned int i;

	for (i = 0; i < batch->nr_pages; i++)
		put_page(batch->pages[i]);
	batch->nr_pages = 0;
}

static int gntdev_copy(struct gntdev_copy_batch *batch)
{
	unsigned int i;

	gnttab_batch_copy(batch->ops, batch->nr_ops);
	gntdev_put_pages(batch);

	/*
	 * For each completed op, update the status if the op failed
	 * and all previous ops for the segment were successful.
	 */
	for (i = 0; i < batch->nr_ops; i++) {
		s16 status = batch->ops[i].status;
		s16 old_status;

		if (status == GNTST_okay)
			continue;

		if (__get_user(old_status, batch->status[i]))
			return -EFAULT;

		if (old_status != GNTST_okay)
			continue;

		if (__put_user(status, batch->status[i]))
			return -EFAULT;
	}

	batch->nr_ops = 0;
	return 0;
}

static int gntdev_grant_copy_seg(struct gntdev_copy_batch *batch,
				 struct gntdev_grant_copy_segment *seg,
				 s16 __user *status)
{
	uint16_t copied = 0;

	/*
	 * Disallow local -> local copies since there is only space in
	 * batch->pages for one page per-op and this would be a very
	 * expensive memcpy().
	 */
	if (!(seg->flags & (GNTCOPY_source_gref | GNTCOPY_dest_gref)))
		return -EINVAL;

	/* Can't cross page if source/dest is a grant ref. */
	if (seg->flags & GNTCOPY_source_gref) {
		if (seg->source.foreign.offset + seg->len > XEN_PAGE_SIZE)
			return -EINVAL;
	}
	if (seg->flags & GNTCOPY_dest_gref) {
		if (seg->dest.foreign.offset + seg->len > XEN_PAGE_SIZE)
			return -EINVAL;
	}

	if (put_user(GNTST_okay, status))
		return -EFAULT;

	while (copied < seg->len) {
		struct gnttab_copy *op;
		void __user *virt;
		size_t len, off;
		unsigned long gfn;
		int ret;

		if (batch->nr_ops >= GNTDEV_COPY_BATCH) {
			ret = gntdev_copy(batch);
			if (ret < 0)
				return ret;
		}

		len = seg->len - copied;

		op = &batch->ops[batch->nr_ops];
		op->flags = 0;

		if (seg->flags & GNTCOPY_source_gref) {
			op->source.u.ref = seg->source.foreign.ref;
			op->source.domid = seg->source.foreign.domid;
			op->source.offset = seg->source.foreign.offset + copied;
			op->flags |= GNTCOPY_source_gref;
		} else {
			virt = seg->source.virt + copied;
			off = (unsigned long)virt & ~XEN_PAGE_MASK;
			len = min(len, (size_t)XEN_PAGE_SIZE - off);

			ret = gntdev_get_page(batch, virt, false, &gfn);
			if (ret < 0)
				return ret;

			op->source.u.gmfn = gfn;
			op->source.domid = DOMID_SELF;
			op->source.offset = off;
		}

		if (seg->flags & GNTCOPY_dest_gref) {
			op->dest.u.ref = seg->dest.foreign.ref;
			op->dest.domid = seg->dest.foreign.domid;
			op->dest.offset = seg->dest.foreign.offset + copied;
			op->flags |= GNTCOPY_dest_gref;
		} else {
			virt = seg->dest.virt + copied;
			off = (unsigned long)virt & ~XEN_PAGE_MASK;
			len = min(len, (size_t)XEN_PAGE_SIZE - off);

			ret = gntdev_get_page(batch, virt, true, &gfn);
			if (ret < 0)
				return ret;

			op->dest.u.gmfn = gfn;
			op->dest.domid = DOMID_SELF;
			op->dest.offset = off;
		}

		op->len = len;
		copied += len;

		batch->status[batch->nr_ops] = status;
		batch->nr_ops++;
	}

	return 0;
}

static long gntdev_ioctl_grant_copy(struct gntdev_priv *priv, void __user *u)
{
	struct ioctl_gntdev_grant_copy copy;
	struct gntdev_copy_batch batch;
	unsigned int i;
	int ret = 0;

	if (copy_from_user(&copy, u, sizeof(copy)))
		return -EFAULT;

	batch.nr_ops = 0;
	batch.nr_pages = 0;

	for (i = 0; i < copy.count; i++) {
		struct gntdev_grant_copy_segment seg;

		if (copy_from_user(&seg, &copy.segments[i], sizeof(seg))) {
			ret = -EFAULT;
			goto out;
		}

		ret = gntdev_grant_copy_seg(&batch, &seg, &copy.segments[i].status);
		if (ret < 0)
			goto out;

		cond_resched();
	}
	if (batch.nr_ops)
		ret = gntdev_copy(&batch);
	return ret;

  out:
	gntdev_put_pages(&batch);
	return ret;
}

943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958
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);

959 960 961
	case IOCTL_GNTDEV_SET_UNMAP_NOTIFY:
		return gntdev_ioctl_notify(priv, ptr);

962 963 964
	case IOCTL_GNTDEV_GRANT_COPY:
		return gntdev_ioctl_grant_copy(priv, ptr);

965 966 967 968 969 970 971 972 973 974 975 976
	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;
977
	int count = vma_pages(vma);
978
	struct grant_map *map;
979
	int i, err = -EINVAL;
980 981 982 983 984 985 986

	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);

987
	mutex_lock(&priv->lock);
988 989 990
	map = gntdev_find_map_index(priv, index, count);
	if (!map)
		goto unlock_out;
991
	if (use_ptemod && map->vma)
992
		goto unlock_out;
993
	if (use_ptemod && priv->mm != vma->vm_mm) {
994
		pr_warn("Huh? Other mm?\n");
995 996 997
		goto unlock_out;
	}

998
	refcount_inc(&map->users);
999

1000 1001
	vma->vm_ops = &gntdev_vmops;

1002
	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP | VM_MIXEDMAP;
1003 1004

	if (use_ptemod)
1005
		vma->vm_flags |= VM_DONTCOPY;
1006 1007 1008

	vma->vm_private_data = map;

1009 1010 1011
	if (use_ptemod)
		map->vma = vma;

1012 1013 1014
	if (map->flags) {
		if ((vma->vm_flags & VM_WRITE) &&
				(map->flags & GNTMAP_readonly))
1015
			goto out_unlock_put;
1016 1017 1018 1019 1020
	} else {
		map->flags = GNTMAP_host_map;
		if (!(vma->vm_flags & VM_WRITE))
			map->flags |= GNTMAP_readonly;
	}
1021

1022
	mutex_unlock(&priv->lock);
1023

1024
	if (use_ptemod) {
1025
		map->pages_vm_start = vma->vm_start;
1026 1027 1028 1029
		err = apply_to_page_range(vma->vm_mm, vma->vm_start,
					  vma->vm_end - vma->vm_start,
					  find_grant_ptes, map);
		if (err) {
1030
			pr_warn("find_grant_ptes() failure.\n");
1031
			goto out_put_map;
1032
		}
1033 1034 1035
	}

	err = map_grant_pages(map);
1036 1037
	if (err)
		goto out_put_map;
1038

1039 1040 1041 1042 1043
	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)
1044
				goto out_put_map;
1045
		}
1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
	} else {
#ifdef CONFIG_X86
		/*
		 * If the PTEs were not made special by the grant map
		 * hypercall, do so here.
		 *
		 * This is racy since the mapping is already visible
		 * to userspace but userspace should be well-behaved
		 * enough to not touch it until the mmap() call
		 * returns.
		 */
		if (!xen_feature(XENFEAT_gnttab_map_avail_bits)) {
			apply_to_page_range(vma->vm_mm, vma->vm_start,
					    vma->vm_end - vma->vm_start,
					    set_grant_ptes_as_special, NULL);
		}
#endif
1063 1064
	}

1065 1066
	return 0;

1067
unlock_out:
1068
	mutex_unlock(&priv->lock);
1069
	return err;
1070

1071
out_unlock_put:
1072
	mutex_unlock(&priv->lock);
1073
out_put_map:
1074
	if (use_ptemod) {
1075
		map->vma = NULL;
1076 1077
		unmap_grant_pages(map, 0, map->count);
	}
1078
	gntdev_put_map(priv, map);
1079
	return err;
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
}

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;

1105
	use_ptemod = !xen_feature(XENFEAT_auto_translated_physmap);
1106

1107 1108
	err = misc_register(&gntdev_miscdev);
	if (err != 0) {
1109
		pr_err("Could not register gntdev device\n");
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
		return err;
	}
	return 0;
}

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

module_init(gntdev_init);
module_exit(gntdev_exit);

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