vhost.c 44.5 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
#include <linux/interval_tree_generic.h>
31 32 33

#include "vhost.h"

34 35 36 37 38
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)");

39 40 41 42
enum {
	VHOST_MEMORY_F_LOG = 0x1,
};

43 44
#define vhost_used_event(vq) ((__virtio16 __user *)&vq->avail->ring[vq->num])
#define vhost_avail_event(vq) ((__virtio16 __user *)&vq->used->ring[vq->num])
45

46 47 48 49
INTERVAL_TREE_DEFINE(struct vhost_umem_node,
		     rb, __u64, __subtree_last,
		     START, LAST, , vhost_umem_interval_tree);

50
#ifdef CONFIG_VHOST_CROSS_ENDIAN_LEGACY
51
static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
52 53 54 55
{
	vq->user_be = !virtio_legacy_is_little_endian();
}

56 57 58 59 60 61 62 63 64 65
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;
}

66 67 68 69 70 71 72 73 74 75 76 77 78 79
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;

80 81 82 83
	if (s.num == VHOST_VRING_BIG_ENDIAN)
		vhost_enable_cross_endian_big(vq);
	else
		vhost_enable_cross_endian_little(vq);
84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111

	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
112
static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
{
}

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 */

134 135 136 137 138
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
139 140 141 142 143 144 145 146 147 148 149 150 151
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);
}

152 153 154 155 156
static void vhost_poll_func(struct file *file, wait_queue_head_t *wqh,
			    poll_table *pt)
{
	struct vhost_poll *poll;

157
	poll = container_of(pt, struct vhost_poll, table);
158 159 160 161 162 163 164
	poll->wqh = wqh;
	add_wait_queue(wqh, &poll->wait);
}

static int vhost_poll_wakeup(wait_queue_t *wait, unsigned mode, int sync,
			     void *key)
{
165 166
	struct vhost_poll *poll = container_of(wait, struct vhost_poll, wait);

167 168 169
	if (!((unsigned long)key & poll->mask))
		return 0;

170
	vhost_poll_queue(poll);
171 172 173
	return 0;
}

174
void vhost_work_init(struct vhost_work *work, vhost_work_fn_t fn)
175
{
Jason Wang's avatar
Jason Wang committed
176
	clear_bit(VHOST_WORK_QUEUED, &work->flags);
177 178 179
	work->fn = fn;
	init_waitqueue_head(&work->done);
}
180
EXPORT_SYMBOL_GPL(vhost_work_init);
181

182
/* Init poll structure */
183 184
void vhost_poll_init(struct vhost_poll *poll, vhost_work_fn_t fn,
		     unsigned long mask, struct vhost_dev *dev)
185 186 187 188
{
	init_waitqueue_func_entry(&poll->wait, vhost_poll_wakeup);
	init_poll_funcptr(&poll->table, vhost_poll_func);
	poll->mask = mask;
189
	poll->dev = dev;
190
	poll->wqh = NULL;
191

192
	vhost_work_init(&poll->work, fn);
193
}
194
EXPORT_SYMBOL_GPL(vhost_poll_init);
195 196 197

/* 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. */
198
int vhost_poll_start(struct vhost_poll *poll, struct file *file)
199 200
{
	unsigned long mask;
201
	int ret = 0;
202

203 204 205
	if (poll->wqh)
		return 0;

206 207 208
	mask = file->f_op->poll(file, &poll->table);
	if (mask)
		vhost_poll_wakeup(&poll->wait, 0, 0, (void *)mask);
209 210 211 212 213 214 215
	if (mask & POLLERR) {
		if (poll->wqh)
			remove_wait_queue(poll->wqh, &poll->wait);
		ret = -EINVAL;
	}

	return ret;
216
}
217
EXPORT_SYMBOL_GPL(vhost_poll_start);
218 219 220 221 222

/* 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)
{
223 224 225 226
	if (poll->wqh) {
		remove_wait_queue(poll->wqh, &poll->wait);
		poll->wqh = NULL;
	}
227
}
228
EXPORT_SYMBOL_GPL(vhost_poll_stop);
229

Jason Wang's avatar
Jason Wang committed
230
void vhost_work_flush(struct vhost_dev *dev, struct vhost_work *work)
231
{
Jason Wang's avatar
Jason Wang committed
232
	struct vhost_flush_struct flush;
233

Jason Wang's avatar
Jason Wang committed
234 235 236
	if (dev->worker) {
		init_completion(&flush.wait_event);
		vhost_work_init(&flush.work, vhost_flush_work);
237

Jason Wang's avatar
Jason Wang committed
238 239 240
		vhost_work_queue(dev, &flush.work);
		wait_for_completion(&flush.wait_event);
	}
241
}
242
EXPORT_SYMBOL_GPL(vhost_work_flush);
243

244 245 246 247 248 249
/* 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);
}
250
EXPORT_SYMBOL_GPL(vhost_poll_flush);
251

252
void vhost_work_queue(struct vhost_dev *dev, struct vhost_work *work)
253
{
Jason Wang's avatar
Jason Wang committed
254 255
	if (!dev->worker)
		return;
256

Jason Wang's avatar
Jason Wang committed
257 258 259 260 261 262
	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);
263 264
		wake_up_process(dev->worker);
	}
265
}
266
EXPORT_SYMBOL_GPL(vhost_work_queue);
267

268 269 270
/* 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
271
	return !llist_empty(&dev->work_list);
272 273 274
}
EXPORT_SYMBOL_GPL(vhost_has_work);

275 276 277 278
void vhost_poll_queue(struct vhost_poll *poll)
{
	vhost_work_queue(poll->dev, &poll->work);
}
279
EXPORT_SYMBOL_GPL(vhost_poll_queue);
280

281 282 283 284 285 286 287 288 289 290
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;
291 292
	vq->signalled_used = 0;
	vq->signalled_used_valid = false;
293 294 295 296
	vq->used_flags = 0;
	vq->log_used = false;
	vq->log_addr = -1ull;
	vq->private_data = NULL;
297
	vq->acked_features = 0;
298 299 300 301 302 303
	vq->log_base = NULL;
	vq->error_ctx = NULL;
	vq->error = NULL;
	vq->kick = NULL;
	vq->call_ctx = NULL;
	vq->call = NULL;
304
	vq->log_ctx = NULL;
305 306
	vhost_reset_is_le(vq);
	vhost_disable_cross_endian(vq);
307
	vq->busyloop_timeout = 0;
308
	vq->umem = NULL;
309 310
}

311 312 313
static int vhost_worker(void *data)
{
	struct vhost_dev *dev = data;
Jason Wang's avatar
Jason Wang committed
314 315
	struct vhost_work *work, *work_next;
	struct llist_node *node;
316
	mm_segment_t oldfs = get_fs();
317

318
	set_fs(USER_DS);
319 320
	use_mm(dev->mm);

321 322 323 324 325 326
	for (;;) {
		/* mb paired w/ kthread_stop */
		set_current_state(TASK_INTERRUPTIBLE);

		if (kthread_should_stop()) {
			__set_current_state(TASK_RUNNING);
327
			break;
328
		}
Jason Wang's avatar
Jason Wang committed
329 330 331 332 333 334 335 336 337 338

		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);
339 340
			__set_current_state(TASK_RUNNING);
			work->fn(work);
341 342
			if (need_resched())
				schedule();
Jason Wang's avatar
Jason Wang committed
343
		}
344
	}
345
	unuse_mm(dev->mm);
346
	set_fs(oldfs);
347
	return 0;
348 349
}

350 351 352 353 354 355 356 357 358 359
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
360 361 362
/* Helper to allocate iovec buffers for all vqs. */
static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
{
363
	struct vhost_virtqueue *vq;
Jason Wang's avatar
Jason Wang committed
364
	int i;
365

Jason Wang's avatar
Jason Wang committed
366
	for (i = 0; i < dev->nvqs; ++i) {
367 368 369 370 371 372
		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
373 374 375
			goto err_nomem;
	}
	return 0;
376

Jason Wang's avatar
Jason Wang committed
377
err_nomem:
378
	for (; i >= 0; --i)
379
		vhost_vq_free_iovecs(dev->vqs[i]);
Jason Wang's avatar
Jason Wang committed
380 381 382 383 384 385
	return -ENOMEM;
}

static void vhost_dev_free_iovecs(struct vhost_dev *dev)
{
	int i;
386

387
	for (i = 0; i < dev->nvqs; ++i)
388
		vhost_vq_free_iovecs(dev->vqs[i]);
Jason Wang's avatar
Jason Wang committed
389 390
}

Zhi Yong Wu's avatar
Zhi Yong Wu committed
391
void vhost_dev_init(struct vhost_dev *dev,
392
		    struct vhost_virtqueue **vqs, int nvqs)
393
{
394
	struct vhost_virtqueue *vq;
395
	int i;
396

397 398 399 400 401
	dev->vqs = vqs;
	dev->nvqs = nvqs;
	mutex_init(&dev->mutex);
	dev->log_ctx = NULL;
	dev->log_file = NULL;
402
	dev->umem = NULL;
403
	dev->mm = NULL;
404
	dev->worker = NULL;
Jason Wang's avatar
Jason Wang committed
405 406
	init_llist_head(&dev->work_list);

407 408

	for (i = 0; i < dev->nvqs; ++i) {
409 410 411 412 413 414 415 416 417 418
		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);
419 420
	}
}
421
EXPORT_SYMBOL_GPL(vhost_dev_init);
422 423 424 425 426 427 428

/* 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;
}
429
EXPORT_SYMBOL_GPL(vhost_dev_check_owner);
430

431
struct vhost_attach_cgroups_struct {
432 433 434
	struct vhost_work work;
	struct task_struct *owner;
	int ret;
435 436 437 438
};

static void vhost_attach_cgroups_work(struct vhost_work *work)
{
439 440 441 442
	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);
443 444 445 446
}

static int vhost_attach_cgroups(struct vhost_dev *dev)
{
447 448 449 450 451 452 453
	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;
454 455
}

456 457 458 459 460
/* Caller should have device mutex */
bool vhost_dev_has_owner(struct vhost_dev *dev)
{
	return dev->mm;
}
461
EXPORT_SYMBOL_GPL(vhost_dev_has_owner);
462

463
/* Caller should have device mutex */
Asias He's avatar
Asias He committed
464
long vhost_dev_set_owner(struct vhost_dev *dev)
465
{
466 467
	struct task_struct *worker;
	int err;
468

469
	/* Is there an owner already? */
470
	if (vhost_dev_has_owner(dev)) {
471 472 473
		err = -EBUSY;
		goto err_mm;
	}
474

475 476
	/* No owner, become one */
	dev->mm = get_task_mm(current);
477 478 479 480 481 482 483
	worker = kthread_create(vhost_worker, dev, "vhost-%d", current->pid);
	if (IS_ERR(worker)) {
		err = PTR_ERR(worker);
		goto err_worker;
	}

	dev->worker = worker;
484 485 486
	wake_up_process(worker);	/* avoid contributing to loadavg */

	err = vhost_attach_cgroups(dev);
487 488
	if (err)
		goto err_cgroup;
489

Jason Wang's avatar
Jason Wang committed
490 491 492 493
	err = vhost_dev_alloc_iovecs(dev);
	if (err)
		goto err_cgroup;

494
	return 0;
495 496
err_cgroup:
	kthread_stop(worker);
497
	dev->worker = NULL;
498 499 500 501 502 503
err_worker:
	if (dev->mm)
		mmput(dev->mm);
	dev->mm = NULL;
err_mm:
	return err;
504
}
505
EXPORT_SYMBOL_GPL(vhost_dev_set_owner);
506

507
static void *vhost_kvzalloc(unsigned long size)
508
{
509 510 511 512 513 514 515 516 517 518
	void *n = kzalloc(size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);

	if (!n)
		n = vzalloc(size);
	return n;
}

struct vhost_umem *vhost_dev_reset_owner_prepare(void)
{
	return vhost_kvzalloc(sizeof(struct vhost_umem));
519
}
520
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner_prepare);
521

522
/* Caller should have device mutex */
523
void vhost_dev_reset_owner(struct vhost_dev *dev, struct vhost_umem *umem)
524
{
525 526
	int i;

527
	vhost_dev_cleanup(dev, true);
528

529
	/* Restore memory to default empty mapping. */
530 531
	INIT_LIST_HEAD(&umem->umem_list);
	dev->umem = umem;
532 533 534 535
	/* We don't need VQ locks below since vhost_dev_cleanup makes sure
	 * VQs aren't running.
	 */
	for (i = 0; i < dev->nvqs; ++i)
536
		dev->vqs[i]->umem = umem;
537
}
538
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner);
539

540
void vhost_dev_stop(struct vhost_dev *dev)
541 542
{
	int i;
543 544

	for (i = 0; i < dev->nvqs; ++i) {
545 546 547
		if (dev->vqs[i]->kick && dev->vqs[i]->handle_kick) {
			vhost_poll_stop(&dev->vqs[i]->poll);
			vhost_poll_flush(&dev->vqs[i]->poll);
548
		}
549 550
	}
}
551
EXPORT_SYMBOL_GPL(vhost_dev_stop);
552

553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
static void vhost_umem_clean(struct vhost_umem *umem)
{
	struct vhost_umem_node *node, *tmp;

	if (!umem)
		return;

	list_for_each_entry_safe(node, tmp, &umem->umem_list, link) {
		vhost_umem_interval_tree_remove(node, &umem->umem_tree);
		list_del(&node->link);
		kvfree(node);
	}
	kvfree(umem);
}

568 569
/* Caller should have device mutex if and only if locked is set */
void vhost_dev_cleanup(struct vhost_dev *dev, bool locked)
570 571
{
	int i;
572

573
	for (i = 0; i < dev->nvqs; ++i) {
574 575 576 577 578 579 580 581 582 583 584
		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]);
585
	}
Jason Wang's avatar
Jason Wang committed
586
	vhost_dev_free_iovecs(dev);
587 588 589 590 591 592 593
	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 */
594 595
	vhost_umem_clean(dev->umem);
	dev->umem = NULL;
Jason Wang's avatar
Jason Wang committed
596
	WARN_ON(!llist_empty(&dev->work_list));
597 598 599 600
	if (dev->worker) {
		kthread_stop(dev->worker);
		dev->worker = NULL;
	}
601 602 603
	if (dev->mm)
		mmput(dev->mm);
	dev->mm = NULL;
604
}
605
EXPORT_SYMBOL_GPL(vhost_dev_cleanup);
606 607 608 609

static int log_access_ok(void __user *log_base, u64 addr, unsigned long sz)
{
	u64 a = addr / VHOST_PAGE_SIZE / 8;
610

611 612 613
	/* Make sure 64 bit math will not overflow. */
	if (a > ULONG_MAX - (unsigned long)log_base ||
	    a + (unsigned long)log_base > ULONG_MAX)
614
		return 0;
615 616 617 618 619 620

	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. */
621
static int vq_memory_access_ok(void __user *log_base, struct vhost_umem *umem,
622 623
			       int log_all)
{
624
	struct vhost_umem_node *node;
625

626
	if (!umem)
Michael S. Tsirkin's avatar
Michael S. Tsirkin committed
627
		return 0;
628

629 630 631 632
	list_for_each_entry(node, &umem->umem_list, link) {
		unsigned long a = node->userspace_addr;

		if (node->size > ULONG_MAX)
633 634
			return 0;
		else if (!access_ok(VERIFY_WRITE, (void __user *)a,
635
				    node->size))
636 637
			return 0;
		else if (log_all && !log_access_ok(log_base,
638 639
						   node->start,
						   node->size))
640 641 642 643 644 645 646
			return 0;
	}
	return 1;
}

/* Can we switch to this memory table? */
/* Caller should have device mutex but not vq mutex */
647
static int memory_access_ok(struct vhost_dev *d, struct vhost_umem *umem,
648 649 650
			    int log_all)
{
	int i;
651

652 653
	for (i = 0; i < d->nvqs; ++i) {
		int ok;
654 655
		bool log;

656
		mutex_lock(&d->vqs[i]->mutex);
657
		log = log_all || vhost_has_feature(d->vqs[i], VHOST_F_LOG_ALL);
658
		/* If ring is inactive, will check when it's enabled. */
659
		if (d->vqs[i]->private_data)
660 661
			ok = vq_memory_access_ok(d->vqs[i]->log_base,
						 umem, log);
662 663
		else
			ok = 1;
664
		mutex_unlock(&d->vqs[i]->mutex);
665 666 667 668 669 670
		if (!ok)
			return 0;
	}
	return 1;
}

671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686
#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);
}

687
static int vq_access_ok(struct vhost_virtqueue *vq, unsigned int num,
688 689 690 691
			struct vring_desc __user *desc,
			struct vring_avail __user *avail,
			struct vring_used __user *used)
{
692
	size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
693 694
	return access_ok(VERIFY_READ, desc, num * sizeof *desc) &&
	       access_ok(VERIFY_READ, avail,
695
			 sizeof *avail + num * sizeof *avail->ring + s) &&
696
	       access_ok(VERIFY_WRITE, used,
697
			sizeof *used + num * sizeof *used->ring + s);
698 699 700 701 702 703
}

/* Can we log writes? */
/* Caller should have device mutex but not vq mutex */
int vhost_log_access_ok(struct vhost_dev *dev)
{
704
	return memory_access_ok(dev, dev->umem, 1);
705
}
706
EXPORT_SYMBOL_GPL(vhost_log_access_ok);
707 708 709

/* Verify access for write logging. */
/* Caller should have vq mutex and device mutex */
710
static int vq_log_access_ok(struct vhost_virtqueue *vq,
711
			    void __user *log_base)
712
{
713
	size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
Arnd Bergmann's avatar
Arnd Bergmann committed
714

715
	return vq_memory_access_ok(log_base, vq->umem,
716
				   vhost_has_feature(vq, VHOST_F_LOG_ALL)) &&
717 718
		(!vq->log_used || log_access_ok(log_base, vq->log_addr,
					sizeof *vq->used +
719
					vq->num * sizeof *vq->used->ring + s));
720 721 722 723 724 725
}

/* Can we start vq? */
/* Caller should have vq mutex and device mutex */
int vhost_vq_access_ok(struct vhost_virtqueue *vq)
{
726 727
	return vq_access_ok(vq, vq->num, vq->desc, vq->avail, vq->used) &&
		vq_log_access_ok(vq, vq->log_base);
728
}
729
EXPORT_SYMBOL_GPL(vhost_vq_access_ok);
730 731 732

static long vhost_set_memory(struct vhost_dev *d, struct vhost_memory __user *m)
{
733 734 735 736
	struct vhost_memory mem, *newmem;
	struct vhost_memory_region *region;
	struct vhost_umem_node *node;
	struct vhost_umem *newumem, *oldumem;
737
	unsigned long size = offsetof(struct vhost_memory, regions);
738
	int i;
739

740 741
	if (copy_from_user(&mem, m, size))
		return -EFAULT;
742 743
	if (mem.padding)
		return -EOPNOTSUPP;
744
	if (mem.nregions > max_mem_regions)
745
		return -E2BIG;
746
	newmem = vhost_kvzalloc(size + mem.nregions * sizeof(*m->regions));
747 748 749 750
	if (!newmem)
		return -ENOMEM;

	memcpy(newmem, &mem, size);
751 752
	if (copy_from_user(newmem->regions, m->regions,
			   mem.nregions * sizeof *m->regions)) {
753
		kvfree(newmem);
754
		return -EFAULT;
755 756
	}

757 758
	newumem = vhost_kvzalloc(sizeof(*newumem));
	if (!newumem) {
759
		kvfree(newmem);
760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778
		return -ENOMEM;
	}

	newumem->umem_tree = RB_ROOT;
	INIT_LIST_HEAD(&newumem->umem_list);

	for (region = newmem->regions;
	     region < newmem->regions + mem.nregions;
	     region++) {
		node = vhost_kvzalloc(sizeof(*node));
		if (!node)
			goto err;
		node->start = region->guest_phys_addr;
		node->size = region->memory_size;
		node->last = node->start + node->size - 1;
		node->userspace_addr = region->userspace_addr;
		INIT_LIST_HEAD(&node->link);
		list_add_tail(&node->link, &newumem->umem_list);
		vhost_umem_interval_tree_insert(node, &newumem->umem_tree);
779
	}
780 781 782 783 784 785

	if (!memory_access_ok(d, newumem, 0))
		goto err;

	oldumem = d->umem;
	d->umem = newumem;
786

787
	/* All memory accesses are done under some VQ mutex. */
788 789
	for (i = 0; i < d->nvqs; ++i) {
		mutex_lock(&d->vqs[i]->mutex);
790
		d->vqs[i]->umem = newumem;
791 792
		mutex_unlock(&d->vqs[i]->mutex);
	}
793 794 795

	kvfree(newmem);
	vhost_umem_clean(oldumem);
796
	return 0;
797 798 799 800 801

err:
	vhost_umem_clean(newumem);
	kvfree(newmem);
	return -EFAULT;
802 803
}

804
long vhost_vring_ioctl(struct vhost_dev *d, int ioctl, void __user *argp)
805
{
806 807
	struct file *eventfp, *filep = NULL;
	bool pollstart = false, pollstop = false;
808 809 810 811 812 813 814 815 816 817 818 819
	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;
820
	if (idx >= d->nvqs)
821 822
		return -ENOBUFS;

823
	vq = d->vqs[idx];
824 825 826 827 828 829 830 831 832 833 834

	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;
		}
835 836
		if (copy_from_user(&s, argp, sizeof s)) {
			r = -EFAULT;
837
			break;
838
		}
839 840 841 842 843 844 845 846 847 848 849 850 851
		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;
		}
852 853
		if (copy_from_user(&s, argp, sizeof s)) {
			r = -EFAULT;
854
			break;
855
		}
856 857 858 859 860 861 862 863 864 865 866
		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;
867 868
		if (copy_to_user(argp, &s, sizeof s))
			r = -EFAULT;
869 870
		break;
	case VHOST_SET_VRING_ADDR:
871 872
		if (copy_from_user(&a, argp, sizeof a)) {
			r = -EFAULT;
873
			break;
874
		}
875 876 877 878 879 880 881 882 883 884 885 886
		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;
		}
887 888 889 890 891 892

		/* 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)) ||
893
		    (a.log_guest_addr & (VRING_USED_ALIGN_SIZE - 1))) {
894 895 896 897 898 899 900 901
			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) {
902
			if (!vq_access_ok(vq, vq->num,
903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
				(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:
927 928
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
929
			break;
930
		}
931
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
932 933 934 935
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
936
		if (eventfp != vq->kick) {
937 938
			pollstop = (filep = vq->kick) != NULL;
			pollstart = (vq->kick = eventfp) != NULL;
939 940 941 942
		} else
			filep = eventfp;
		break;
	case VHOST_SET_VRING_CALL:
943 944
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
945
			break;
946
		}
947
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
948 949 950 951
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
952 953 954 955 956 957 958 959 960 961
		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:
962 963
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
964
			break;
965
		}
966
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
967 968 969 970
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
971 972 973 974 975 976 977 978 979
		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;
980 981 982 983 984 985
	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;
986 987 988 989 990 991 992 993 994 995 996 997 998
	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;
999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
	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)
1012
		r = vhost_poll_start(&vq->poll, vq->kick);
1013 1014 1015 1016 1017 1018 1019

	mutex_unlock(&vq->mutex);

	if (pollstop && vq->handle_kick)
		vhost_poll_flush(&vq->poll);
	return r;
}
1020
EXPORT_SYMBOL_GPL(vhost_vring_ioctl);
1021 1022

/* Caller must have device mutex */
1023
long vhost_dev_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
{
	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:
1047 1048
		if (copy_from_user(&p, argp, sizeof p)) {
			r = -EFAULT;
1049
			break;
1050
		}
1051 1052 1053 1054 1055 1056 1057
		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;
1058
			vq = d->vqs[i];
1059 1060
			mutex_lock(&vq->mutex);
			/* If ring is inactive, will check when it's enabled. */
1061
			if (vq->private_data && !vq_log_access_ok(vq, base))
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
				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;
1079
			d->log_file = eventfp;
1080 1081 1082 1083 1084 1085
			ctx = d->log_ctx;
			d->log_ctx = eventfp ?
				eventfd_ctx_fileget(eventfp) : NULL;
		} else
			filep = eventfp;
		for (i = 0; i < d->nvqs; ++i) {
1086 1087 1088
			mutex_lock(&d->vqs[i]->mutex);
			d->vqs[i]->log_ctx = d->log_ctx;
			mutex_unlock(&d->vqs[i]->mutex);
1089 1090 1091 1092 1093 1094 1095
		}
		if (ctx)
			eventfd_ctx_put(ctx);
		if (filep)
			fput(filep);
		break;
	default:
1096
		r = -ENOIOCTLCMD;
1097 1098 1099 1100 1101
		break;
	}
done:
	return r;
}
1102
EXPORT_SYMBOL_GPL(vhost_dev_ioctl);
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114

/* 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;
1115

1116
	r = get_user_pages_fast(log, 1, 1, &page);
1117
	if (r < 0)
1118
		return r;
1119
	BUG_ON(r != 1);
1120
	base = kmap_atomic(page);
1121
	set_bit(bit, base);
1122
	kunmap_atomic(base);
1123 1124 1125 1126 1127 1128 1129 1130
	set_page_dirty_lock(page);
	put_page(page);
	return 0;
}

static int log_write(void __user *log_base,
		     u64 write_address, u64 write_length)
{
1131
	u64 write_page = write_address / VHOST_PAGE_SIZE;
1132
	int r;
1133

1134 1135
	if (!write_length)
		return 0;
1136
	write_length += write_address % VHOST_PAGE_SIZE;
1137 1138
	for (;;) {
		u64 base = (u64)(unsigned long)log_base;
1139 1140
		u64 log = base + write_page / 8;
		int bit = write_page % 8;
1141 1142 1143 1144 1145 1146 1147 1148
		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;
1149
		write_page += 1;
1150 1151 1152 1153 1154 1155 1156 1157 1158 1159
	}
	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. */
1160
	smp_wmb();
1161 1162 1163 1164 1165 1166
	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;
1167 1168 1169
		if (!len) {
			if (vq->log_ctx)
				eventfd_signal(vq->log_ctx, 1);
1170
			return 0;
1171
		}
1172 1173 1174 1175 1176
	}
	/* Length written exceeds what we have stored. This is a bug. */
	BUG();
	return 0;
}
1177
EXPORT_SYMBOL_GPL(vhost_log_write);
1178

1179 1180 1181
static int vhost_update_used_flags(struct vhost_virtqueue *vq)
{
	void __user *used;
1182 1183
	if (vhost_put_user(vq, cpu_to_vhost16(vq, vq->used_flags),
			   &vq->used->flags) < 0)
1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200
		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)
{
1201 1202
	if (vhost_put_user(vq, cpu_to_vhost16(vq, vq->avail_idx),
			   vhost_avail_event(vq)))