vhost.c 44.1 KB
Newer Older
1 2 3 4 5 6
/* Copyright (C) 2009 Red Hat, Inc.
 * Copyright (C) 2006 Rusty Russell IBM Corporation
 *
 * Author: Michael S. Tsirkin <mst@redhat.com>
 *
 * Inspiration, some code, and most witty comments come from
Rob Landley's avatar
Rob Landley committed
7
 * Documentation/virtual/lguest/lguest.c, by Rusty Russell
8 9 10 11 12 13 14 15
 *
 * This work is licensed under the terms of the GNU GPL, version 2.
 *
 * Generic code for virtio server in host kernel.
 */

#include <linux/eventfd.h>
#include <linux/vhost.h>
16
#include <linux/uio.h>
17
#include <linux/mm.h>
18
#include <linux/mmu_context.h>
19 20 21 22 23
#include <linux/miscdevice.h>
#include <linux/mutex.h>
#include <linux/poll.h>
#include <linux/file.h>
#include <linux/highmem.h>
24
#include <linux/slab.h>
25
#include <linux/vmalloc.h>
26
#include <linux/kthread.h>
27
#include <linux/cgroup.h>
28
#include <linux/module.h>
29
#include <linux/sort.h>
30 31 32

#include "vhost.h"

33 34 35 36 37
static ushort max_mem_regions = 64;
module_param(max_mem_regions, ushort, 0444);
MODULE_PARM_DESC(max_mem_regions,
	"Maximum number of memory regions in memory map. (default: 64)");

38 39 40 41
enum {
	VHOST_MEMORY_F_LOG = 0x1,
};

42 43
#define vhost_used_event(vq) ((__virtio16 __user *)&vq->avail->ring[vq->num])
#define vhost_avail_event(vq) ((__virtio16 __user *)&vq->used->ring[vq->num])
44

45
#ifdef CONFIG_VHOST_CROSS_ENDIAN_LEGACY
46
static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
47 48 49 50
{
	vq->user_be = !virtio_legacy_is_little_endian();
}

51 52 53 54 55 56 57 58 59 60
static void vhost_enable_cross_endian_big(struct vhost_virtqueue *vq)
{
	vq->user_be = true;
}

static void vhost_enable_cross_endian_little(struct vhost_virtqueue *vq)
{
	vq->user_be = false;
}

61 62 63 64 65 66 67 68 69 70 71 72 73 74
static long vhost_set_vring_endian(struct vhost_virtqueue *vq, int __user *argp)
{
	struct vhost_vring_state s;

	if (vq->private_data)
		return -EBUSY;

	if (copy_from_user(&s, argp, sizeof(s)))
		return -EFAULT;

	if (s.num != VHOST_VRING_LITTLE_ENDIAN &&
	    s.num != VHOST_VRING_BIG_ENDIAN)
		return -EINVAL;

75 76 77 78
	if (s.num == VHOST_VRING_BIG_ENDIAN)
		vhost_enable_cross_endian_big(vq);
	else
		vhost_enable_cross_endian_little(vq);
79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106

	return 0;
}

static long vhost_get_vring_endian(struct vhost_virtqueue *vq, u32 idx,
				   int __user *argp)
{
	struct vhost_vring_state s = {
		.index = idx,
		.num = vq->user_be
	};

	if (copy_to_user(argp, &s, sizeof(s)))
		return -EFAULT;

	return 0;
}

static void vhost_init_is_le(struct vhost_virtqueue *vq)
{
	/* Note for legacy virtio: user_be is initialized at reset time
	 * according to the host endianness. If userspace does not set an
	 * explicit endianness, the default behavior is native endian, as
	 * expected by legacy virtio.
	 */
	vq->is_le = vhost_has_feature(vq, VIRTIO_F_VERSION_1) || !vq->user_be;
}
#else
107
static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
{
}

static long vhost_set_vring_endian(struct vhost_virtqueue *vq, int __user *argp)
{
	return -ENOIOCTLCMD;
}

static long vhost_get_vring_endian(struct vhost_virtqueue *vq, u32 idx,
				   int __user *argp)
{
	return -ENOIOCTLCMD;
}

static void vhost_init_is_le(struct vhost_virtqueue *vq)
{
	if (vhost_has_feature(vq, VIRTIO_F_VERSION_1))
		vq->is_le = true;
}
#endif /* CONFIG_VHOST_CROSS_ENDIAN_LEGACY */

129 130 131 132 133
static void vhost_reset_is_le(struct vhost_virtqueue *vq)
{
	vq->is_le = virtio_legacy_is_little_endian();
}

Jason Wang's avatar
Jason Wang committed
134 135 136 137 138 139 140 141 142 143 144 145 146
struct vhost_flush_struct {
	struct vhost_work work;
	struct completion wait_event;
};

static void vhost_flush_work(struct vhost_work *work)
{
	struct vhost_flush_struct *s;

	s = container_of(work, struct vhost_flush_struct, work);
	complete(&s->wait_event);
}

147 148 149 150 151
static void vhost_poll_func(struct file *file, wait_queue_head_t *wqh,
			    poll_table *pt)
{
	struct vhost_poll *poll;

152
	poll = container_of(pt, struct vhost_poll, table);
153 154 155 156 157 158 159
	poll->wqh = wqh;
	add_wait_queue(wqh, &poll->wait);
}

static int vhost_poll_wakeup(wait_queue_t *wait, unsigned mode, int sync,
			     void *key)
{
160 161
	struct vhost_poll *poll = container_of(wait, struct vhost_poll, wait);

162 163 164
	if (!((unsigned long)key & poll->mask))
		return 0;

165
	vhost_poll_queue(poll);
166 167 168
	return 0;
}

169
void vhost_work_init(struct vhost_work *work, vhost_work_fn_t fn)
170
{
Jason Wang's avatar
Jason Wang committed
171
	clear_bit(VHOST_WORK_QUEUED, &work->flags);
172 173 174
	work->fn = fn;
	init_waitqueue_head(&work->done);
}
175
EXPORT_SYMBOL_GPL(vhost_work_init);
176

177
/* Init poll structure */
178 179
void vhost_poll_init(struct vhost_poll *poll, vhost_work_fn_t fn,
		     unsigned long mask, struct vhost_dev *dev)
180 181 182 183
{
	init_waitqueue_func_entry(&poll->wait, vhost_poll_wakeup);
	init_poll_funcptr(&poll->table, vhost_poll_func);
	poll->mask = mask;
184
	poll->dev = dev;
185
	poll->wqh = NULL;
186

187
	vhost_work_init(&poll->work, fn);
188
}
189
EXPORT_SYMBOL_GPL(vhost_poll_init);
190 191 192

/* Start polling a file. We add ourselves to file's wait queue. The caller must
 * keep a reference to a file until after vhost_poll_stop is called. */
193
int vhost_poll_start(struct vhost_poll *poll, struct file *file)
194 195
{
	unsigned long mask;
196
	int ret = 0;
197

198 199 200
	if (poll->wqh)
		return 0;

201 202 203
	mask = file->f_op->poll(file, &poll->table);
	if (mask)
		vhost_poll_wakeup(&poll->wait, 0, 0, (void *)mask);
204 205 206 207 208 209 210
	if (mask & POLLERR) {
		if (poll->wqh)
			remove_wait_queue(poll->wqh, &poll->wait);
		ret = -EINVAL;
	}

	return ret;
211
}
212
EXPORT_SYMBOL_GPL(vhost_poll_start);
213 214 215 216 217

/* Stop polling a file. After this function returns, it becomes safe to drop the
 * file reference. You must also flush afterwards. */
void vhost_poll_stop(struct vhost_poll *poll)
{
218 219 220 221
	if (poll->wqh) {
		remove_wait_queue(poll->wqh, &poll->wait);
		poll->wqh = NULL;
	}
222
}
223
EXPORT_SYMBOL_GPL(vhost_poll_stop);
224

Jason Wang's avatar
Jason Wang committed
225
void vhost_work_flush(struct vhost_dev *dev, struct vhost_work *work)
226
{
Jason Wang's avatar
Jason Wang committed
227
	struct vhost_flush_struct flush;
228

Jason Wang's avatar
Jason Wang committed
229 230 231
	if (dev->worker) {
		init_completion(&flush.wait_event);
		vhost_work_init(&flush.work, vhost_flush_work);
232

Jason Wang's avatar
Jason Wang committed
233 234 235
		vhost_work_queue(dev, &flush.work);
		wait_for_completion(&flush.wait_event);
	}
236
}
237
EXPORT_SYMBOL_GPL(vhost_work_flush);
238

239 240 241 242 243 244
/* Flush any work that has been scheduled. When calling this, don't hold any
 * locks that are also used by the callback. */
void vhost_poll_flush(struct vhost_poll *poll)
{
	vhost_work_flush(poll->dev, &poll->work);
}
245
EXPORT_SYMBOL_GPL(vhost_poll_flush);
246

247
void vhost_work_queue(struct vhost_dev *dev, struct vhost_work *work)
248
{
Jason Wang's avatar
Jason Wang committed
249 250
	if (!dev->worker)
		return;
251

Jason Wang's avatar
Jason Wang committed
252 253 254 255 256 257
	if (!test_and_set_bit(VHOST_WORK_QUEUED, &work->flags)) {
		/* We can only add the work to the list after we're
		 * sure it was not in the list.
		 */
		smp_mb();
		llist_add(&work->node, &dev->work_list);
258 259
		wake_up_process(dev->worker);
	}
260
}
261
EXPORT_SYMBOL_GPL(vhost_work_queue);
262

263 264 265
/* A lockless hint for busy polling code to exit the loop */
bool vhost_has_work(struct vhost_dev *dev)
{
Jason Wang's avatar
Jason Wang committed
266
	return !llist_empty(&dev->work_list);
267 268 269
}
EXPORT_SYMBOL_GPL(vhost_has_work);

270 271 272 273
void vhost_poll_queue(struct vhost_poll *poll)
{
	vhost_work_queue(poll->dev, &poll->work);
}
274
EXPORT_SYMBOL_GPL(vhost_poll_queue);
275

276 277 278 279 280 281 282 283 284 285
static void vhost_vq_reset(struct vhost_dev *dev,
			   struct vhost_virtqueue *vq)
{
	vq->num = 1;
	vq->desc = NULL;
	vq->avail = NULL;
	vq->used = NULL;
	vq->last_avail_idx = 0;
	vq->avail_idx = 0;
	vq->last_used_idx = 0;
286 287
	vq->signalled_used = 0;
	vq->signalled_used_valid = false;
288 289 290 291
	vq->used_flags = 0;
	vq->log_used = false;
	vq->log_addr = -1ull;
	vq->private_data = NULL;
292
	vq->acked_features = 0;
293 294 295 296 297 298
	vq->log_base = NULL;
	vq->error_ctx = NULL;
	vq->error = NULL;
	vq->kick = NULL;
	vq->call_ctx = NULL;
	vq->call = NULL;
299
	vq->log_ctx = NULL;
300
	vq->memory = NULL;
301 302
	vhost_reset_is_le(vq);
	vhost_disable_cross_endian(vq);
303
	vq->busyloop_timeout = 0;
304 305
}

306 307 308
static int vhost_worker(void *data)
{
	struct vhost_dev *dev = data;
Jason Wang's avatar
Jason Wang committed
309 310
	struct vhost_work *work, *work_next;
	struct llist_node *node;
311
	mm_segment_t oldfs = get_fs();
312

313
	set_fs(USER_DS);
314 315
	use_mm(dev->mm);

316 317 318 319 320 321
	for (;;) {
		/* mb paired w/ kthread_stop */
		set_current_state(TASK_INTERRUPTIBLE);

		if (kthread_should_stop()) {
			__set_current_state(TASK_RUNNING);
322
			break;
323
		}
Jason Wang's avatar
Jason Wang committed
324 325 326 327 328 329 330 331 332 333

		node = llist_del_all(&dev->work_list);
		if (!node)
			schedule();

		node = llist_reverse_order(node);
		/* make sure flag is seen after deletion */
		smp_wmb();
		llist_for_each_entry_safe(work, work_next, node, node) {
			clear_bit(VHOST_WORK_QUEUED, &work->flags);
334 335
			__set_current_state(TASK_RUNNING);
			work->fn(work);
336 337
			if (need_resched())
				schedule();
Jason Wang's avatar
Jason Wang committed
338
		}
339
	}
340
	unuse_mm(dev->mm);
341
	set_fs(oldfs);
342
	return 0;
343 344
}

345 346 347 348 349 350 351 352 353 354
static void vhost_vq_free_iovecs(struct vhost_virtqueue *vq)
{
	kfree(vq->indirect);
	vq->indirect = NULL;
	kfree(vq->log);
	vq->log = NULL;
	kfree(vq->heads);
	vq->heads = NULL;
}

Jason Wang's avatar
Jason Wang committed
355 356 357
/* Helper to allocate iovec buffers for all vqs. */
static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
{
358
	struct vhost_virtqueue *vq;
Jason Wang's avatar
Jason Wang committed
359
	int i;
360

Jason Wang's avatar
Jason Wang committed
361
	for (i = 0; i < dev->nvqs; ++i) {
362 363 364 365 366 367
		vq = dev->vqs[i];
		vq->indirect = kmalloc(sizeof *vq->indirect * UIO_MAXIOV,
				       GFP_KERNEL);
		vq->log = kmalloc(sizeof *vq->log * UIO_MAXIOV, GFP_KERNEL);
		vq->heads = kmalloc(sizeof *vq->heads * UIO_MAXIOV, GFP_KERNEL);
		if (!vq->indirect || !vq->log || !vq->heads)
Jason Wang's avatar
Jason Wang committed
368 369 370
			goto err_nomem;
	}
	return 0;
371

Jason Wang's avatar
Jason Wang committed
372
err_nomem:
373
	for (; i >= 0; --i)
374
		vhost_vq_free_iovecs(dev->vqs[i]);
Jason Wang's avatar
Jason Wang committed
375 376 377 378 379 380
	return -ENOMEM;
}

static void vhost_dev_free_iovecs(struct vhost_dev *dev)
{
	int i;
381

382
	for (i = 0; i < dev->nvqs; ++i)
383
		vhost_vq_free_iovecs(dev->vqs[i]);
Jason Wang's avatar
Jason Wang committed
384 385
}

Zhi Yong Wu's avatar
Zhi Yong Wu committed
386
void vhost_dev_init(struct vhost_dev *dev,
387
		    struct vhost_virtqueue **vqs, int nvqs)
388
{
389
	struct vhost_virtqueue *vq;
390
	int i;
391

392 393 394 395 396 397 398
	dev->vqs = vqs;
	dev->nvqs = nvqs;
	mutex_init(&dev->mutex);
	dev->log_ctx = NULL;
	dev->log_file = NULL;
	dev->memory = NULL;
	dev->mm = NULL;
399
	dev->worker = NULL;
Jason Wang's avatar
Jason Wang committed
400 401
	init_llist_head(&dev->work_list);

402 403

	for (i = 0; i < dev->nvqs; ++i) {
404 405 406 407 408 409 410 411 412 413
		vq = dev->vqs[i];
		vq->log = NULL;
		vq->indirect = NULL;
		vq->heads = NULL;
		vq->dev = dev;
		mutex_init(&vq->mutex);
		vhost_vq_reset(dev, vq);
		if (vq->handle_kick)
			vhost_poll_init(&vq->poll, vq->handle_kick,
					POLLIN, dev);
414 415
	}
}
416
EXPORT_SYMBOL_GPL(vhost_dev_init);
417 418 419 420 421 422 423

/* Caller should have device mutex */
long vhost_dev_check_owner(struct vhost_dev *dev)
{
	/* Are you the owner? If not, I don't think you mean to do that */
	return dev->mm == current->mm ? 0 : -EPERM;
}
424
EXPORT_SYMBOL_GPL(vhost_dev_check_owner);
425

426
struct vhost_attach_cgroups_struct {
427 428 429
	struct vhost_work work;
	struct task_struct *owner;
	int ret;
430 431 432 433
};

static void vhost_attach_cgroups_work(struct vhost_work *work)
{
434 435 436 437
	struct vhost_attach_cgroups_struct *s;

	s = container_of(work, struct vhost_attach_cgroups_struct, work);
	s->ret = cgroup_attach_task_all(s->owner, current);
438 439 440 441
}

static int vhost_attach_cgroups(struct vhost_dev *dev)
{
442 443 444 445 446 447 448
	struct vhost_attach_cgroups_struct attach;

	attach.owner = current;
	vhost_work_init(&attach.work, vhost_attach_cgroups_work);
	vhost_work_queue(dev, &attach.work);
	vhost_work_flush(dev, &attach.work);
	return attach.ret;
449 450
}

451 452 453 454 455
/* Caller should have device mutex */
bool vhost_dev_has_owner(struct vhost_dev *dev)
{
	return dev->mm;
}
456
EXPORT_SYMBOL_GPL(vhost_dev_has_owner);
457

458
/* Caller should have device mutex */
Asias He's avatar
Asias He committed
459
long vhost_dev_set_owner(struct vhost_dev *dev)
460
{
461 462
	struct task_struct *worker;
	int err;
463

464
	/* Is there an owner already? */
465
	if (vhost_dev_has_owner(dev)) {
466 467 468
		err = -EBUSY;
		goto err_mm;
	}
469

470 471
	/* No owner, become one */
	dev->mm = get_task_mm(current);
472 473 474 475 476 477 478
	worker = kthread_create(vhost_worker, dev, "vhost-%d", current->pid);
	if (IS_ERR(worker)) {
		err = PTR_ERR(worker);
		goto err_worker;
	}

	dev->worker = worker;
479 480 481
	wake_up_process(worker);	/* avoid contributing to loadavg */

	err = vhost_attach_cgroups(dev);
482 483
	if (err)
		goto err_cgroup;
484

Jason Wang's avatar
Jason Wang committed
485 486 487 488
	err = vhost_dev_alloc_iovecs(dev);
	if (err)
		goto err_cgroup;

489
	return 0;
490 491
err_cgroup:
	kthread_stop(worker);
492
	dev->worker = NULL;
493 494 495 496 497 498
err_worker:
	if (dev->mm)
		mmput(dev->mm);
	dev->mm = NULL;
err_mm:
	return err;
499
}
500
EXPORT_SYMBOL_GPL(vhost_dev_set_owner);
501

502
struct vhost_memory *vhost_dev_reset_owner_prepare(void)
503
{
504 505
	return kmalloc(offsetof(struct vhost_memory, regions), GFP_KERNEL);
}
506
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner_prepare);
507

508 509 510
/* Caller should have device mutex */
void vhost_dev_reset_owner(struct vhost_dev *dev, struct vhost_memory *memory)
{
511 512
	int i;

513
	vhost_dev_cleanup(dev, true);
514

515
	/* Restore memory to default empty mapping. */
516
	memory->nregions = 0;
517 518 519 520 521 522
	dev->memory = memory;
	/* We don't need VQ locks below since vhost_dev_cleanup makes sure
	 * VQs aren't running.
	 */
	for (i = 0; i < dev->nvqs; ++i)
		dev->vqs[i]->memory = memory;
523
}
524
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner);
525

526
void vhost_dev_stop(struct vhost_dev *dev)
527 528
{
	int i;
529 530

	for (i = 0; i < dev->nvqs; ++i) {
531 532 533
		if (dev->vqs[i]->kick && dev->vqs[i]->handle_kick) {
			vhost_poll_stop(&dev->vqs[i]->poll);
			vhost_poll_flush(&dev->vqs[i]->poll);
534
		}
535 536
	}
}
537
EXPORT_SYMBOL_GPL(vhost_dev_stop);
538

539 540
/* Caller should have device mutex if and only if locked is set */
void vhost_dev_cleanup(struct vhost_dev *dev, bool locked)
541 542
{
	int i;
543

544
	for (i = 0; i < dev->nvqs; ++i) {
545 546 547 548 549 550 551 552 553 554 555
		if (dev->vqs[i]->error_ctx)
			eventfd_ctx_put(dev->vqs[i]->error_ctx);
		if (dev->vqs[i]->error)
			fput(dev->vqs[i]->error);
		if (dev->vqs[i]->kick)
			fput(dev->vqs[i]->kick);
		if (dev->vqs[i]->call_ctx)
			eventfd_ctx_put(dev->vqs[i]->call_ctx);
		if (dev->vqs[i]->call)
			fput(dev->vqs[i]->call);
		vhost_vq_reset(dev, dev->vqs[i]);
556
	}
Jason Wang's avatar
Jason Wang committed
557
	vhost_dev_free_iovecs(dev);
558 559 560 561 562 563 564
	if (dev->log_ctx)
		eventfd_ctx_put(dev->log_ctx);
	dev->log_ctx = NULL;
	if (dev->log_file)
		fput(dev->log_file);
	dev->log_file = NULL;
	/* No one will access memory at this point */
565
	kvfree(dev->memory);
566
	dev->memory = NULL;
Jason Wang's avatar
Jason Wang committed
567
	WARN_ON(!llist_empty(&dev->work_list));
568 569 570 571
	if (dev->worker) {
		kthread_stop(dev->worker);
		dev->worker = NULL;
	}
572 573 574
	if (dev->mm)
		mmput(dev->mm);
	dev->mm = NULL;
575
}
576
EXPORT_SYMBOL_GPL(vhost_dev_cleanup);
577 578 579 580

static int log_access_ok(void __user *log_base, u64 addr, unsigned long sz)
{
	u64 a = addr / VHOST_PAGE_SIZE / 8;
581

582 583 584
	/* Make sure 64 bit math will not overflow. */
	if (a > ULONG_MAX - (unsigned long)log_base ||
	    a + (unsigned long)log_base > ULONG_MAX)
585
		return 0;
586 587 588 589 590 591 592 593 594 595

	return access_ok(VERIFY_WRITE, log_base + a,
			 (sz + VHOST_PAGE_SIZE * 8 - 1) / VHOST_PAGE_SIZE / 8);
}

/* Caller should have vq mutex and device mutex. */
static int vq_memory_access_ok(void __user *log_base, struct vhost_memory *mem,
			       int log_all)
{
	int i;
596

Michael S. Tsirkin's avatar
Michael S. Tsirkin committed
597 598
	if (!mem)
		return 0;
599

600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621
	for (i = 0; i < mem->nregions; ++i) {
		struct vhost_memory_region *m = mem->regions + i;
		unsigned long a = m->userspace_addr;
		if (m->memory_size > ULONG_MAX)
			return 0;
		else if (!access_ok(VERIFY_WRITE, (void __user *)a,
				    m->memory_size))
			return 0;
		else if (log_all && !log_access_ok(log_base,
						   m->guest_phys_addr,
						   m->memory_size))
			return 0;
	}
	return 1;
}

/* Can we switch to this memory table? */
/* Caller should have device mutex but not vq mutex */
static int memory_access_ok(struct vhost_dev *d, struct vhost_memory *mem,
			    int log_all)
{
	int i;
622

623 624
	for (i = 0; i < d->nvqs; ++i) {
		int ok;
625 626
		bool log;

627
		mutex_lock(&d->vqs[i]->mutex);
628
		log = log_all || vhost_has_feature(d->vqs[i], VHOST_F_LOG_ALL);
629
		/* If ring is inactive, will check when it's enabled. */
630
		if (d->vqs[i]->private_data)
631
			ok = vq_memory_access_ok(d->vqs[i]->log_base, mem, log);
632 633
		else
			ok = 1;
634
		mutex_unlock(&d->vqs[i]->mutex);
635 636 637 638 639 640
		if (!ok)
			return 0;
	}
	return 1;
}

641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
#define vhost_put_user(vq, x, ptr)  __put_user(x, ptr)

static int vhost_copy_to_user(struct vhost_virtqueue *vq, void *to,
			      const void *from, unsigned size)
{
	return __copy_to_user(to, from, size);
}

#define vhost_get_user(vq, x, ptr) __get_user(x, ptr)

static int vhost_copy_from_user(struct vhost_virtqueue *vq, void *to,
				void *from, unsigned size)
{
	return __copy_from_user(to, from, size);
}

657
static int vq_access_ok(struct vhost_virtqueue *vq, unsigned int num,
658 659 660 661
			struct vring_desc __user *desc,
			struct vring_avail __user *avail,
			struct vring_used __user *used)
{
662
	size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
663 664
	return access_ok(VERIFY_READ, desc, num * sizeof *desc) &&
	       access_ok(VERIFY_READ, avail,
665
			 sizeof *avail + num * sizeof *avail->ring + s) &&
666
	       access_ok(VERIFY_WRITE, used,
667
			sizeof *used + num * sizeof *used->ring + s);
668 669 670 671 672 673
}

/* Can we log writes? */
/* Caller should have device mutex but not vq mutex */
int vhost_log_access_ok(struct vhost_dev *dev)
{
674
	return memory_access_ok(dev, dev->memory, 1);
675
}
676
EXPORT_SYMBOL_GPL(vhost_log_access_ok);
677 678 679

/* Verify access for write logging. */
/* Caller should have vq mutex and device mutex */
680
static int vq_log_access_ok(struct vhost_virtqueue *vq,
681
			    void __user *log_base)
682
{
683
	size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
Arnd Bergmann's avatar
Arnd Bergmann committed
684

685
	return vq_memory_access_ok(log_base, vq->memory,
686
				   vhost_has_feature(vq, VHOST_F_LOG_ALL)) &&
687 688
		(!vq->log_used || log_access_ok(log_base, vq->log_addr,
					sizeof *vq->used +
689
					vq->num * sizeof *vq->used->ring + s));
690 691 692 693 694 695
}

/* Can we start vq? */
/* Caller should have vq mutex and device mutex */
int vhost_vq_access_ok(struct vhost_virtqueue *vq)
{
696 697
	return vq_access_ok(vq, vq->num, vq->desc, vq->avail, vq->used) &&
		vq_log_access_ok(vq, vq->log_base);
698
}
699
EXPORT_SYMBOL_GPL(vhost_vq_access_ok);
700

701 702 703 704 705 706 707 708 709 710
static int vhost_memory_reg_sort_cmp(const void *p1, const void *p2)
{
	const struct vhost_memory_region *r1 = p1, *r2 = p2;
	if (r1->guest_phys_addr < r2->guest_phys_addr)
		return 1;
	if (r1->guest_phys_addr > r2->guest_phys_addr)
		return -1;
	return 0;
}

711 712 713 714
static void *vhost_kvzalloc(unsigned long size)
{
	void *n = kzalloc(size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);

715
	if (!n)
716 717 718 719
		n = vzalloc(size);
	return n;
}

720 721 722 723
static long vhost_set_memory(struct vhost_dev *d, struct vhost_memory __user *m)
{
	struct vhost_memory mem, *newmem, *oldmem;
	unsigned long size = offsetof(struct vhost_memory, regions);
724
	int i;
725

726 727
	if (copy_from_user(&mem, m, size))
		return -EFAULT;
728 729
	if (mem.padding)
		return -EOPNOTSUPP;
730
	if (mem.nregions > max_mem_regions)
731
		return -E2BIG;
732
	newmem = vhost_kvzalloc(size + mem.nregions * sizeof(*m->regions));
733 734 735 736
	if (!newmem)
		return -ENOMEM;

	memcpy(newmem, &mem, size);
737 738
	if (copy_from_user(newmem->regions, m->regions,
			   mem.nregions * sizeof *m->regions)) {
739
		kvfree(newmem);
740
		return -EFAULT;
741
	}
742 743
	sort(newmem->regions, newmem->nregions, sizeof(*newmem->regions),
		vhost_memory_reg_sort_cmp, NULL);
744

745
	if (!memory_access_ok(d, newmem, 0)) {
746
		kvfree(newmem);
747
		return -EFAULT;
748
	}
749 750
	oldmem = d->memory;
	d->memory = newmem;
751

752
	/* All memory accesses are done under some VQ mutex. */
753 754
	for (i = 0; i < d->nvqs; ++i) {
		mutex_lock(&d->vqs[i]->mutex);
755
		d->vqs[i]->memory = newmem;
756 757
		mutex_unlock(&d->vqs[i]->mutex);
	}
758
	kvfree(oldmem);
759 760 761
	return 0;
}

762
long vhost_vring_ioctl(struct vhost_dev *d, int ioctl, void __user *argp)
763
{
764 765
	struct file *eventfp, *filep = NULL;
	bool pollstart = false, pollstop = false;
766 767 768 769 770 771 772 773 774 775 776 777
	struct eventfd_ctx *ctx = NULL;
	u32 __user *idxp = argp;
	struct vhost_virtqueue *vq;
	struct vhost_vring_state s;
	struct vhost_vring_file f;
	struct vhost_vring_addr a;
	u32 idx;
	long r;

	r = get_user(idx, idxp);
	if (r < 0)
		return r;
778
	if (idx >= d->nvqs)
779 780
		return -ENOBUFS;

781
	vq = d->vqs[idx];
782 783 784 785 786 787 788 789 790 791 792

	mutex_lock(&vq->mutex);

	switch (ioctl) {
	case VHOST_SET_VRING_NUM:
		/* Resizing ring with an active backend?
		 * You don't want to do that. */
		if (vq->private_data) {
			r = -EBUSY;
			break;
		}
793 794
		if (copy_from_user(&s, argp, sizeof s)) {
			r = -EFAULT;
795
			break;
796
		}
797 798 799 800 801 802 803 804 805 806 807 808 809
		if (!s.num || s.num > 0xffff || (s.num & (s.num - 1))) {
			r = -EINVAL;
			break;
		}
		vq->num = s.num;
		break;
	case VHOST_SET_VRING_BASE:
		/* Moving base with an active backend?
		 * You don't want to do that. */
		if (vq->private_data) {
			r = -EBUSY;
			break;
		}
810 811
		if (copy_from_user(&s, argp, sizeof s)) {
			r = -EFAULT;
812
			break;
813
		}
814 815 816 817 818 819 820 821 822 823 824
		if (s.num > 0xffff) {
			r = -EINVAL;
			break;
		}
		vq->last_avail_idx = s.num;
		/* Forget the cached index value. */
		vq->avail_idx = vq->last_avail_idx;
		break;
	case VHOST_GET_VRING_BASE:
		s.index = idx;
		s.num = vq->last_avail_idx;
825 826
		if (copy_to_user(argp, &s, sizeof s))
			r = -EFAULT;
827 828
		break;
	case VHOST_SET_VRING_ADDR:
829 830
		if (copy_from_user(&a, argp, sizeof a)) {
			r = -EFAULT;
831
			break;
832
		}
833 834 835 836 837 838 839 840 841 842 843 844
		if (a.flags & ~(0x1 << VHOST_VRING_F_LOG)) {
			r = -EOPNOTSUPP;
			break;
		}
		/* For 32bit, verify that the top 32bits of the user
		   data are set to zero. */
		if ((u64)(unsigned long)a.desc_user_addr != a.desc_user_addr ||
		    (u64)(unsigned long)a.used_user_addr != a.used_user_addr ||
		    (u64)(unsigned long)a.avail_user_addr != a.avail_user_addr) {
			r = -EFAULT;
			break;
		}
845 846 847 848 849 850

		/* Make sure it's safe to cast pointers to vring types. */
		BUILD_BUG_ON(__alignof__ *vq->avail > VRING_AVAIL_ALIGN_SIZE);
		BUILD_BUG_ON(__alignof__ *vq->used > VRING_USED_ALIGN_SIZE);
		if ((a.avail_user_addr & (VRING_AVAIL_ALIGN_SIZE - 1)) ||
		    (a.used_user_addr & (VRING_USED_ALIGN_SIZE - 1)) ||
851
		    (a.log_guest_addr & (VRING_USED_ALIGN_SIZE - 1))) {
852 853 854 855 856 857 858 859
			r = -EINVAL;
			break;
		}

		/* We only verify access here if backend is configured.
		 * If it is not, we don't as size might not have been setup.
		 * We will verify when backend is configured. */
		if (vq->private_data) {
860
			if (!vq_access_ok(vq, vq->num,
861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884
				(void __user *)(unsigned long)a.desc_user_addr,
				(void __user *)(unsigned long)a.avail_user_addr,
				(void __user *)(unsigned long)a.used_user_addr)) {
				r = -EINVAL;
				break;
			}

			/* Also validate log access for used ring if enabled. */
			if ((a.flags & (0x1 << VHOST_VRING_F_LOG)) &&
			    !log_access_ok(vq->log_base, a.log_guest_addr,
					   sizeof *vq->used +
					   vq->num * sizeof *vq->used->ring)) {
				r = -EINVAL;
				break;
			}
		}

		vq->log_used = !!(a.flags & (0x1 << VHOST_VRING_F_LOG));
		vq->desc = (void __user *)(unsigned long)a.desc_user_addr;
		vq->avail = (void __user *)(unsigned long)a.avail_user_addr;
		vq->log_addr = a.log_guest_addr;
		vq->used = (void __user *)(unsigned long)a.used_user_addr;
		break;
	case VHOST_SET_VRING_KICK:
885 886
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
887
			break;
888
		}
889
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
890 891 892 893
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
894
		if (eventfp != vq->kick) {
895 896
			pollstop = (filep = vq->kick) != NULL;
			pollstart = (vq->kick = eventfp) != NULL;
897 898 899 900
		} else
			filep = eventfp;
		break;
	case VHOST_SET_VRING_CALL:
901 902
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
903
			break;
904
		}
905
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
906 907 908 909
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
910 911 912 913 914 915 916 917 918 919
		if (eventfp != vq->call) {
			filep = vq->call;
			ctx = vq->call_ctx;
			vq->call = eventfp;
			vq->call_ctx = eventfp ?
				eventfd_ctx_fileget(eventfp) : NULL;
		} else
			filep = eventfp;
		break;
	case VHOST_SET_VRING_ERR:
920 921
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
922
			break;
923
		}
924
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
925 926 927 928
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
929 930 931 932 933 934 935 936 937
		if (eventfp != vq->error) {
			filep = vq->error;
			vq->error = eventfp;
			ctx = vq->error_ctx;
			vq->error_ctx = eventfp ?
				eventfd_ctx_fileget(eventfp) : NULL;
		} else
			filep = eventfp;
		break;
938 939 940 941 942 943
	case VHOST_SET_VRING_ENDIAN:
		r = vhost_set_vring_endian(vq, argp);
		break;
	case VHOST_GET_VRING_ENDIAN:
		r = vhost_get_vring_endian(vq, idx, argp);
		break;
944 945 946 947 948 949 950 951 952 953 954 955 956
	case VHOST_SET_VRING_BUSYLOOP_TIMEOUT:
		if (copy_from_user(&s, argp, sizeof(s))) {
			r = -EFAULT;
			break;
		}
		vq->busyloop_timeout = s.num;
		break;
	case VHOST_GET_VRING_BUSYLOOP_TIMEOUT:
		s.index = idx;
		s.num = vq->busyloop_timeout;
		if (copy_to_user(argp, &s, sizeof(s)))
			r = -EFAULT;
		break;
957 958 959 960 961 962 963 964 965 966 967 968 969
	default:
		r = -ENOIOCTLCMD;
	}

	if (pollstop && vq->handle_kick)
		vhost_poll_stop(&vq->poll);

	if (ctx)
		eventfd_ctx_put(ctx);
	if (filep)
		fput(filep);

	if (pollstart && vq->handle_kick)
970
		r = vhost_poll_start(&vq->poll, vq->kick);
971 972 973 974 975 976 977

	mutex_unlock(&vq->mutex);

	if (pollstop && vq->handle_kick)
		vhost_poll_flush(&vq->poll);
	return r;
}
978
EXPORT_SYMBOL_GPL(vhost_vring_ioctl);
979 980

/* Caller must have device mutex */
981
long vhost_dev_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
{
	struct file *eventfp, *filep = NULL;
	struct eventfd_ctx *ctx = NULL;
	u64 p;
	long r;
	int i, fd;

	/* If you are not the owner, you can become one */
	if (ioctl == VHOST_SET_OWNER) {
		r = vhost_dev_set_owner(d);
		goto done;
	}

	/* You must be the owner to do anything else */
	r = vhost_dev_check_owner(d);
	if (r)
		goto done;

	switch (ioctl) {
	case VHOST_SET_MEM_TABLE:
		r = vhost_set_memory(d, argp);
		break;
	case VHOST_SET_LOG_BASE:
1005 1006
		if (copy_from_user(&p, argp, sizeof p)) {
			r = -EFAULT;
1007
			break;
1008
		}
1009 1010 1011 1012 1013 1014 1015
		if ((u64)(unsigned long)p != p) {
			r = -EFAULT;
			break;
		}
		for (i = 0; i < d->nvqs; ++i) {
			struct vhost_virtqueue *vq;
			void __user *base = (void __user *)(unsigned long)p;
1016
			vq = d->vqs[i];
1017 1018
			mutex_lock(&vq->mutex);
			/* If ring is inactive, will check when it's enabled. */
1019
			if (vq->private_data && !vq_log_access_ok(vq, base))
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
				r = -EFAULT;
			else
				vq->log_base = base;
			mutex_unlock(&vq->mutex);
		}
		break;
	case VHOST_SET_LOG_FD:
		r = get_user(fd, (int __user *)argp);
		if (r < 0)
			break;
		eventfp = fd == -1 ? NULL : eventfd_fget(fd);
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
		if (eventfp != d->log_file) {
			filep = d->log_file;
1037
			d->log_file = eventfp;
1038 1039 1040 1041 1042 1043
			ctx = d->log_ctx;
			d->log_ctx = eventfp ?
				eventfd_ctx_fileget(eventfp) : NULL;
		} else
			filep = eventfp;
		for (i = 0; i < d->nvqs; ++i) {
1044 1045 1046
			mutex_lock(&d->vqs[i]->mutex);
			d->vqs[i]->log_ctx = d->log_ctx;
			mutex_unlock(&d->vqs[i]->mutex);
1047 1048 1049 1050 1051 1052 1053
		}
		if (ctx)
			eventfd_ctx_put(ctx);
		if (filep)
			fput(filep);
		break;
	default:
1054
		r = -ENOIOCTLCMD;
1055 1056 1057 1058 1059
		break;
	}
done:
	return r;
}
1060
EXPORT_SYMBOL_GPL(vhost_dev_ioctl);
1061 1062 1063 1064

static const struct vhost_memory_region *find_region(struct vhost_memory *mem,
						     __u64 addr, __u32 len)
{
1065 1066 1067 1068 1069 1070 1071 1072 1073 1074
	const struct vhost_memory_region *reg;
	int start = 0, end = mem->nregions;

	while (start < end) {
		int slot = start + (end - start) / 2;
		reg = mem->regions + slot;
		if (addr >= reg->guest_phys_addr)
			end = slot;
		else
			start = slot + 1;
1075
	}
1076 1077 1078 1079 1080

	reg = mem->regions + start;
	if (addr >= reg->guest_phys_addr &&
		reg->guest_phys_addr + reg->memory_size > addr)
		return reg;
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
	return NULL;
}

/* TODO: This is really inefficient.  We need something like get_user()
 * (instruction directly accesses the data, with an exception table entry
 * returning -EFAULT). See Documentation/x86/exception-tables.txt.
 */
static int set_bit_to_user(int nr, void __user *addr)
{
	unsigned long log = (unsigned long)addr;
	struct page *page;
	void *base;
	int bit = nr + (log % PAGE_SIZE) * 8;
	int r;
1095

1096
	r = get_user_pages_fast(log, 1, 1, &page);
1097
	if (r < 0)
1098
		return r;
1099
	BUG_ON(r != 1);
1100
	base = kmap_atomic(page);
1101
	set_bit(bit, base);
1102
	kunmap_atomic(base);
1103 1104 1105 1106 1107 1108 1109 1110
	set_page_dirty_lock(page);
	put_page(page);
	return 0;
}

static int log_write(void __user *log_base,
		     u64 write_address, u64 write_length)
{
1111
	u64 write_page = write_address / VHOST_PAGE_SIZE;
1112
	int r;
1113

1114 1115
	if (!write_length)
		return 0;
1116
	write_length += write_address % VHOST_PAGE_SIZE;
1117 1118
	for (;;) {
		u64 base = (u64)(unsigned long)log_base;
1119 1120
		u64 log = base + write_page / 8;
		int bit = write_page % 8;
1121 1122 1123 1124 1125 1126 1127 1128
		if ((u64)(unsigned long)log != log)
			return -EFAULT;
		r = set_bit_to_user(bit, (void __user *)(unsigned long)log);
		if (r < 0)
			return r;
		if (write_length <= VHOST_PAGE_SIZE)
			break;
		write_length -= VHOST_PAGE_SIZE;
1129
		write_page += 1;
1130 1131 1132 1133 1134 1135 1136 1137 1138 1139
	}
	return r;
}

int vhost_log_write(struct vhost_virtqueue *vq, struct vhost_log *log,
		    unsigned int log_num, u64 len)
{
	int i, r;

	/* Make sure data written is seen before log. */
1140
	smp_wmb();
1141 1142 1143 1144 1145 1146
	for (i = 0; i < log_num; ++i) {
		u64 l = min(log[i].len, len);
		r = log_write(vq->log_base, log[i].addr, l);
		if (r < 0)
			return r;
		len -= l;
1147 1148 1149
		if (!len) {
			if (vq->log_ctx)
				eventfd_signal(vq->log_ctx, 1);
1150
			return 0;
1151
		}
1152 1153 1154 1155 1156
	}
	/* Length written exceeds what we have stored. This is a bug. */
	BUG();
	return 0;
}
1157
EXPORT_SYMBOL_GPL(vhost_log_write);
1158

1159 1160 1161
static int vhost_update_used_flags(struct vhost_virtqueue *vq)
{
	void __user *used;
1162 1163
	if (vhost_put_user(vq, cpu_to_vhost16(vq, vq->used_flags),
			   &vq->used->flags) < 0)
1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180
		return -EFAULT;
	if (unlikely(vq->log_used)) {
		/* Make sure the flag is seen before log. */
		smp_wmb();
		/* Log used flag write. */
		used = &vq->used->flags;
		log_write(vq->log_base, vq->log_addr +
			  (used - (void __user *)vq->used),
			  sizeof vq->used->flags);
		if (vq->log_ctx)
			eventfd_signal(vq->log_ctx, 1);
	}
	return 0;
}

static int vhost_update_avail_event(struct vhost_virtqueue *vq, u16 avail_event)
{
1181 1182
	if (vhost_put_user(vq, cpu_to_vhost16(vq, vq->avail_idx),
			   vhost_avail_event(vq)))
1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
		return -EFAULT;
	if (unlikely(vq->log_used)) {
		void __user *used;
		/* Make sure the event is seen before log. */
		smp_wmb();
		/* Log avail event write */
		used = vhost_avail_event(vq);
		log_write(vq->log_base, vq->log_addr +
			  (used - (void __user *)vq->used),
			  sizeof *vhost_avail_event(vq));
		if (vq->log_ctx)
			eventfd_signal(vq->log_ctx, 1);
	}
	return 0;
}

Greg Kurz's avatar
Greg Kurz committed
1199
int vhost_vq_init_access(struct vhost_virtqueue *vq)
1200
{
1201
	__virtio16 last_used_idx;
1202
	int r;
1203 1204
	bool is_le = vq->is_le;

1205
	if (!vq->private_data) {
1206
		vhost_reset_is_le(vq);
1207
		return 0;
1208 1209 1210
	}

	vhost_init_is_le(vq);