vhost.c 62.6 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
2 3 4 5 6 7
/* 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
8
 * Documentation/virtual/lguest/lguest.c, by Rusty Russell
9 10 11 12 13 14
 *
 * Generic code for virtio server in host kernel.
 */

#include <linux/eventfd.h>
#include <linux/vhost.h>
15
#include <linux/uio.h>
16
#include <linux/mm.h>
17
#include <linux/mmu_context.h>
18 19 20 21 22
#include <linux/miscdevice.h>
#include <linux/mutex.h>
#include <linux/poll.h>
#include <linux/file.h>
#include <linux/highmem.h>
23
#include <linux/slab.h>
24
#include <linux/vmalloc.h>
25
#include <linux/kthread.h>
26
#include <linux/cgroup.h>
27
#include <linux/module.h>
28
#include <linux/sort.h>
29
#include <linux/sched/mm.h>
30
#include <linux/sched/signal.h>
31
#include <linux/interval_tree_generic.h>
32
#include <linux/nospec.h>
33 34 35

#include "vhost.h"

36 37 38 39
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)");
Jason Wang's avatar
Jason Wang committed
40 41 42 43
static int max_iotlb_entries = 2048;
module_param(max_iotlb_entries, int, 0444);
MODULE_PARM_DESC(max_iotlb_entries,
	"Maximum number of iotlb entries. (default: 2048)");
44

45 46 47 48
enum {
	VHOST_MEMORY_F_LOG = 0x1,
};

49 50
#define vhost_used_event(vq) ((__virtio16 __user *)&vq->avail->ring[vq->num])
#define vhost_avail_event(vq) ((__virtio16 __user *)&vq->used->ring[vq->num])
51

52 53
INTERVAL_TREE_DEFINE(struct vhost_umem_node,
		     rb, __u64, __subtree_last,
54
		     START, LAST, static inline, vhost_umem_interval_tree);
55

56
#ifdef CONFIG_VHOST_CROSS_ENDIAN_LEGACY
57
static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
58 59 60 61
{
	vq->user_be = !virtio_legacy_is_little_endian();
}

62 63 64 65 66 67 68 69 70 71
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;
}

72 73 74 75 76 77 78 79 80 81 82 83 84 85
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;

86 87 88 89
	if (s.num == VHOST_VRING_BIG_ENDIAN)
		vhost_enable_cross_endian_big(vq);
	else
		vhost_enable_cross_endian_little(vq);
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117

	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
118
static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134
{
}

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)
{
135 136
	vq->is_le = vhost_has_feature(vq, VIRTIO_F_VERSION_1)
		|| virtio_legacy_is_little_endian();
137 138 139
}
#endif /* CONFIG_VHOST_CROSS_ENDIAN_LEGACY */

140 141
static void vhost_reset_is_le(struct vhost_virtqueue *vq)
{
142
	vhost_init_is_le(vq);
143 144
}

Jason Wang's avatar
Jason Wang committed
145 146 147 148 149 150 151 152 153 154 155 156 157
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);
}

158 159 160 161 162
static void vhost_poll_func(struct file *file, wait_queue_head_t *wqh,
			    poll_table *pt)
{
	struct vhost_poll *poll;

163
	poll = container_of(pt, struct vhost_poll, table);
164 165 166 167
	poll->wqh = wqh;
	add_wait_queue(wqh, &poll->wait);
}

168
static int vhost_poll_wakeup(wait_queue_entry_t *wait, unsigned mode, int sync,
169 170
			     void *key)
{
171 172
	struct vhost_poll *poll = container_of(wait, struct vhost_poll, wait);

Al Viro's avatar
Al Viro committed
173
	if (!(key_to_poll(key) & poll->mask))
174 175
		return 0;

176
	vhost_poll_queue(poll);
177 178 179
	return 0;
}

180
void vhost_work_init(struct vhost_work *work, vhost_work_fn_t fn)
181
{
Jason Wang's avatar
Jason Wang committed
182
	clear_bit(VHOST_WORK_QUEUED, &work->flags);
183 184
	work->fn = fn;
}
185
EXPORT_SYMBOL_GPL(vhost_work_init);
186

187
/* Init poll structure */
188
void vhost_poll_init(struct vhost_poll *poll, vhost_work_fn_t fn,
Al Viro's avatar
Al Viro committed
189
		     __poll_t mask, struct vhost_dev *dev)
190 191 192 193
{
	init_waitqueue_func_entry(&poll->wait, vhost_poll_wakeup);
	init_poll_funcptr(&poll->table, vhost_poll_func);
	poll->mask = mask;
194
	poll->dev = dev;
195
	poll->wqh = NULL;
196

197
	vhost_work_init(&poll->work, fn);
198
}
199
EXPORT_SYMBOL_GPL(vhost_poll_init);
200 201 202

/* 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. */
203
int vhost_poll_start(struct vhost_poll *poll, struct file *file)
204
{
205
	__poll_t mask;
206

207 208 209
	if (poll->wqh)
		return 0;

210
	mask = vfs_poll(file, &poll->table);
211
	if (mask)
Al Viro's avatar
Al Viro committed
212
		vhost_poll_wakeup(&poll->wait, 0, 0, poll_to_key(mask));
213
	if (mask & EPOLLERR) {
214
		vhost_poll_stop(poll);
215
		return -EINVAL;
216 217
	}

218
	return 0;
219
}
220
EXPORT_SYMBOL_GPL(vhost_poll_start);
221 222 223 224 225

/* 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)
{
226 227 228 229
	if (poll->wqh) {
		remove_wait_queue(poll->wqh, &poll->wait);
		poll->wqh = NULL;
	}
230
}
231
EXPORT_SYMBOL_GPL(vhost_poll_stop);
232

Jason Wang's avatar
Jason Wang committed
233
void vhost_work_flush(struct vhost_dev *dev, struct vhost_work *work)
234
{
Jason Wang's avatar
Jason Wang committed
235
	struct vhost_flush_struct flush;
236

Jason Wang's avatar
Jason Wang committed
237 238 239
	if (dev->worker) {
		init_completion(&flush.wait_event);
		vhost_work_init(&flush.work, vhost_flush_work);
240

Jason Wang's avatar
Jason Wang committed
241 242 243
		vhost_work_queue(dev, &flush.work);
		wait_for_completion(&flush.wait_event);
	}
244
}
245
EXPORT_SYMBOL_GPL(vhost_work_flush);
246

247 248 249 250 251 252
/* 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);
}
253
EXPORT_SYMBOL_GPL(vhost_poll_flush);
254

255
void vhost_work_queue(struct vhost_dev *dev, struct vhost_work *work)
256
{
Jason Wang's avatar
Jason Wang committed
257 258
	if (!dev->worker)
		return;
259

Jason Wang's avatar
Jason Wang committed
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.
263
		 * test_and_set_bit() implies a memory barrier.
Jason Wang's avatar
Jason Wang committed
264 265
		 */
		llist_add(&work->node, &dev->work_list);
266 267
		wake_up_process(dev->worker);
	}
268
}
269
EXPORT_SYMBOL_GPL(vhost_work_queue);
270

271 272 273
/* 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
274
	return !llist_empty(&dev->work_list);
275 276 277
}
EXPORT_SYMBOL_GPL(vhost_has_work);

278 279 280 281
void vhost_poll_queue(struct vhost_poll *poll)
{
	vhost_work_queue(poll->dev, &poll->work);
}
282
EXPORT_SYMBOL_GPL(vhost_poll_queue);
283

284 285 286 287 288 289 290 291 292 293 294 295
static void __vhost_vq_meta_reset(struct vhost_virtqueue *vq)
{
	int j;

	for (j = 0; j < VHOST_NUM_ADDRS; j++)
		vq->meta_iotlb[j] = NULL;
}

static void vhost_vq_meta_reset(struct vhost_dev *d)
{
	int i;

296
	for (i = 0; i < d->nvqs; ++i)
297 298 299
		__vhost_vq_meta_reset(d->vqs[i]);
}

300 301 302 303 304 305 306 307 308 309
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;
310 311
	vq->signalled_used = 0;
	vq->signalled_used_valid = false;
312 313 314 315
	vq->used_flags = 0;
	vq->log_used = false;
	vq->log_addr = -1ull;
	vq->private_data = NULL;
316
	vq->acked_features = 0;
317
	vq->acked_backend_features = 0;
318 319 320 321
	vq->log_base = NULL;
	vq->error_ctx = NULL;
	vq->kick = NULL;
	vq->call_ctx = NULL;
322
	vq->log_ctx = NULL;
323 324
	vhost_reset_is_le(vq);
	vhost_disable_cross_endian(vq);
325
	vq->busyloop_timeout = 0;
326
	vq->umem = NULL;
Jason Wang's avatar
Jason Wang committed
327
	vq->iotlb = NULL;
328
	__vhost_vq_meta_reset(vq);
329 330
}

331 332 333
static int vhost_worker(void *data)
{
	struct vhost_dev *dev = data;
Jason Wang's avatar
Jason Wang committed
334 335
	struct vhost_work *work, *work_next;
	struct llist_node *node;
336
	mm_segment_t oldfs = get_fs();
337

338
	set_fs(USER_DS);
339 340
	use_mm(dev->mm);

341 342 343 344 345 346
	for (;;) {
		/* mb paired w/ kthread_stop */
		set_current_state(TASK_INTERRUPTIBLE);

		if (kthread_should_stop()) {
			__set_current_state(TASK_RUNNING);
347
			break;
348
		}
Jason Wang's avatar
Jason Wang committed
349 350 351 352 353 354 355 356 357 358

		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);
359 360
			__set_current_state(TASK_RUNNING);
			work->fn(work);
361 362
			if (need_resched())
				schedule();
Jason Wang's avatar
Jason Wang committed
363
		}
364
	}
365
	unuse_mm(dev->mm);
366
	set_fs(oldfs);
367
	return 0;
368 369
}

370 371 372 373 374 375 376 377 378 379
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
380 381 382
/* Helper to allocate iovec buffers for all vqs. */
static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
{
383
	struct vhost_virtqueue *vq;
Jason Wang's avatar
Jason Wang committed
384
	int i;
385

Jason Wang's avatar
Jason Wang committed
386
	for (i = 0; i < dev->nvqs; ++i) {
387
		vq = dev->vqs[i];
388 389 390
		vq->indirect = kmalloc_array(UIO_MAXIOV,
					     sizeof(*vq->indirect),
					     GFP_KERNEL);
Jason Wang's avatar
Jason Wang committed
391
		vq->log = kmalloc_array(dev->iov_limit, sizeof(*vq->log),
392
					GFP_KERNEL);
Jason Wang's avatar
Jason Wang committed
393
		vq->heads = kmalloc_array(dev->iov_limit, sizeof(*vq->heads),
394
					  GFP_KERNEL);
395
		if (!vq->indirect || !vq->log || !vq->heads)
Jason Wang's avatar
Jason Wang committed
396 397 398
			goto err_nomem;
	}
	return 0;
399

Jason Wang's avatar
Jason Wang committed
400
err_nomem:
401
	for (; i >= 0; --i)
402
		vhost_vq_free_iovecs(dev->vqs[i]);
Jason Wang's avatar
Jason Wang committed
403 404 405 406 407 408
	return -ENOMEM;
}

static void vhost_dev_free_iovecs(struct vhost_dev *dev)
{
	int i;
409

410
	for (i = 0; i < dev->nvqs; ++i)
411
		vhost_vq_free_iovecs(dev->vqs[i]);
Jason Wang's avatar
Jason Wang committed
412 413
}

414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
bool vhost_exceeds_weight(struct vhost_virtqueue *vq,
			  int pkts, int total_len)
{
	struct vhost_dev *dev = vq->dev;

	if ((dev->byte_weight && total_len >= dev->byte_weight) ||
	    pkts >= dev->weight) {
		vhost_poll_queue(&vq->poll);
		return true;
	}

	return false;
}
EXPORT_SYMBOL_GPL(vhost_exceeds_weight);

429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
static size_t vhost_get_avail_size(struct vhost_virtqueue *vq,
				   unsigned int num)
{
	size_t event __maybe_unused =
	       vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;

	return sizeof(*vq->avail) +
	       sizeof(*vq->avail->ring) * num + event;
}

static size_t vhost_get_used_size(struct vhost_virtqueue *vq,
				  unsigned int num)
{
	size_t event __maybe_unused =
	       vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;

	return sizeof(*vq->used) +
	       sizeof(*vq->used->ring) * num + event;
}

static size_t vhost_get_desc_size(struct vhost_virtqueue *vq,
				  unsigned int num)
{
	return sizeof(*vq->desc) * num;
}

Zhi Yong Wu's avatar
Zhi Yong Wu committed
455
void vhost_dev_init(struct vhost_dev *dev,
456 457
		    struct vhost_virtqueue **vqs, int nvqs,
		    int iov_limit, int weight, int byte_weight)
458
{
459
	struct vhost_virtqueue *vq;
460
	int i;
461

462 463 464 465
	dev->vqs = vqs;
	dev->nvqs = nvqs;
	mutex_init(&dev->mutex);
	dev->log_ctx = NULL;
466
	dev->umem = NULL;
Jason Wang's avatar
Jason Wang committed
467
	dev->iotlb = NULL;
468
	dev->mm = NULL;
469
	dev->worker = NULL;
Jason Wang's avatar
Jason Wang committed
470
	dev->iov_limit = iov_limit;
471 472
	dev->weight = weight;
	dev->byte_weight = byte_weight;
Jason Wang's avatar
Jason Wang committed
473
	init_llist_head(&dev->work_list);
Jason Wang's avatar
Jason Wang committed
474 475 476 477
	init_waitqueue_head(&dev->wait);
	INIT_LIST_HEAD(&dev->read_list);
	INIT_LIST_HEAD(&dev->pending_list);
	spin_lock_init(&dev->iotlb_lock);
478

479 480

	for (i = 0; i < dev->nvqs; ++i) {
481 482 483 484 485 486 487 488 489
		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,
490
					EPOLLIN, dev);
491 492
	}
}
493
EXPORT_SYMBOL_GPL(vhost_dev_init);
494 495 496 497 498 499 500

/* 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;
}
501
EXPORT_SYMBOL_GPL(vhost_dev_check_owner);
502

503
struct vhost_attach_cgroups_struct {
504 505 506
	struct vhost_work work;
	struct task_struct *owner;
	int ret;
507 508 509 510
};

static void vhost_attach_cgroups_work(struct vhost_work *work)
{
511 512 513 514
	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);
515 516 517 518
}

static int vhost_attach_cgroups(struct vhost_dev *dev)
{
519 520 521 522 523 524 525
	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;
526 527
}

528 529 530 531 532
/* Caller should have device mutex */
bool vhost_dev_has_owner(struct vhost_dev *dev)
{
	return dev->mm;
}
533
EXPORT_SYMBOL_GPL(vhost_dev_has_owner);
534

535
/* Caller should have device mutex */
Asias He's avatar
Asias He committed
536
long vhost_dev_set_owner(struct vhost_dev *dev)
537
{
538 539
	struct task_struct *worker;
	int err;
540

541
	/* Is there an owner already? */
542
	if (vhost_dev_has_owner(dev)) {
543 544 545
		err = -EBUSY;
		goto err_mm;
	}
546

547 548
	/* No owner, become one */
	dev->mm = get_task_mm(current);
549 550 551 552 553 554 555
	worker = kthread_create(vhost_worker, dev, "vhost-%d", current->pid);
	if (IS_ERR(worker)) {
		err = PTR_ERR(worker);
		goto err_worker;
	}

	dev->worker = worker;
556 557 558
	wake_up_process(worker);	/* avoid contributing to loadavg */

	err = vhost_attach_cgroups(dev);
559 560
	if (err)
		goto err_cgroup;
561

Jason Wang's avatar
Jason Wang committed
562 563 564 565
	err = vhost_dev_alloc_iovecs(dev);
	if (err)
		goto err_cgroup;

566
	return 0;
567 568
err_cgroup:
	kthread_stop(worker);
569
	dev->worker = NULL;
570 571 572 573 574 575
err_worker:
	if (dev->mm)
		mmput(dev->mm);
	dev->mm = NULL;
err_mm:
	return err;
576
}
577
EXPORT_SYMBOL_GPL(vhost_dev_set_owner);
578

579 580
struct vhost_umem *vhost_dev_reset_owner_prepare(void)
{
581
	return kvzalloc(sizeof(struct vhost_umem), GFP_KERNEL);
582
}
583
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner_prepare);
584

585
/* Caller should have device mutex */
586
void vhost_dev_reset_owner(struct vhost_dev *dev, struct vhost_umem *umem)
587
{
588 589
	int i;

590
	vhost_dev_cleanup(dev);
591

592
	/* Restore memory to default empty mapping. */
593 594
	INIT_LIST_HEAD(&umem->umem_list);
	dev->umem = umem;
595 596 597 598
	/* We don't need VQ locks below since vhost_dev_cleanup makes sure
	 * VQs aren't running.
	 */
	for (i = 0; i < dev->nvqs; ++i)
599
		dev->vqs[i]->umem = umem;
600
}
601
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner);
602

603
void vhost_dev_stop(struct vhost_dev *dev)
604 605
{
	int i;
606 607

	for (i = 0; i < dev->nvqs; ++i) {
608 609 610
		if (dev->vqs[i]->kick && dev->vqs[i]->handle_kick) {
			vhost_poll_stop(&dev->vqs[i]->poll);
			vhost_poll_flush(&dev->vqs[i]->poll);
611
		}
612 613
	}
}
614
EXPORT_SYMBOL_GPL(vhost_dev_stop);
615

Jason Wang's avatar
Jason Wang committed
616 617 618 619 620 621 622 623 624
static void vhost_umem_free(struct vhost_umem *umem,
			    struct vhost_umem_node *node)
{
	vhost_umem_interval_tree_remove(node, &umem->umem_tree);
	list_del(&node->link);
	kfree(node);
	umem->numem--;
}

625 626 627 628 629 630 631
static void vhost_umem_clean(struct vhost_umem *umem)
{
	struct vhost_umem_node *node, *tmp;

	if (!umem)
		return;

Jason Wang's avatar
Jason Wang committed
632 633 634
	list_for_each_entry_safe(node, tmp, &umem->umem_list, link)
		vhost_umem_free(umem, node);

635 636 637
	kvfree(umem);
}

Jason Wang's avatar
Jason Wang committed
638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
static void vhost_clear_msg(struct vhost_dev *dev)
{
	struct vhost_msg_node *node, *n;

	spin_lock(&dev->iotlb_lock);

	list_for_each_entry_safe(node, n, &dev->read_list, node) {
		list_del(&node->node);
		kfree(node);
	}

	list_for_each_entry_safe(node, n, &dev->pending_list, node) {
		list_del(&node->node);
		kfree(node);
	}

	spin_unlock(&dev->iotlb_lock);
}

657
void vhost_dev_cleanup(struct vhost_dev *dev)
658 659
{
	int i;
660

661
	for (i = 0; i < dev->nvqs; ++i) {
662 663 664 665 666 667 668
		if (dev->vqs[i]->error_ctx)
			eventfd_ctx_put(dev->vqs[i]->error_ctx);
		if (dev->vqs[i]->kick)
			fput(dev->vqs[i]->kick);
		if (dev->vqs[i]->call_ctx)
			eventfd_ctx_put(dev->vqs[i]->call_ctx);
		vhost_vq_reset(dev, dev->vqs[i]);
669
	}
Jason Wang's avatar
Jason Wang committed
670
	vhost_dev_free_iovecs(dev);
671 672 673 674
	if (dev->log_ctx)
		eventfd_ctx_put(dev->log_ctx);
	dev->log_ctx = NULL;
	/* No one will access memory at this point */
675 676
	vhost_umem_clean(dev->umem);
	dev->umem = NULL;
Jason Wang's avatar
Jason Wang committed
677 678 679
	vhost_umem_clean(dev->iotlb);
	dev->iotlb = NULL;
	vhost_clear_msg(dev);
680
	wake_up_interruptible_poll(&dev->wait, EPOLLIN | EPOLLRDNORM);
Jason Wang's avatar
Jason Wang committed
681
	WARN_ON(!llist_empty(&dev->work_list));
682 683 684 685
	if (dev->worker) {
		kthread_stop(dev->worker);
		dev->worker = NULL;
	}
686
	if (dev->mm)
687 688
		mmput(dev->mm);
	dev->mm = NULL;
689
}
690
EXPORT_SYMBOL_GPL(vhost_dev_cleanup);
691

692
static bool log_access_ok(void __user *log_base, u64 addr, unsigned long sz)
693 694
{
	u64 a = addr / VHOST_PAGE_SIZE / 8;
695

696 697 698
	/* Make sure 64 bit math will not overflow. */
	if (a > ULONG_MAX - (unsigned long)log_base ||
	    a + (unsigned long)log_base > ULONG_MAX)
699
		return false;
700

701
	return access_ok(log_base + a,
702 703 704
			 (sz + VHOST_PAGE_SIZE * 8 - 1) / VHOST_PAGE_SIZE / 8);
}

705 706 707 708 709 710
static bool vhost_overflow(u64 uaddr, u64 size)
{
	/* Make sure 64 bit math will not overflow. */
	return uaddr > ULONG_MAX || size > ULONG_MAX || uaddr > ULONG_MAX - size;
}

711
/* Caller should have vq mutex and device mutex. */
712 713
static bool vq_memory_access_ok(void __user *log_base, struct vhost_umem *umem,
				int log_all)
714
{
715
	struct vhost_umem_node *node;
716

717
	if (!umem)
718
		return false;
719

720 721 722
	list_for_each_entry(node, &umem->umem_list, link) {
		unsigned long a = node->userspace_addr;

723
		if (vhost_overflow(node->userspace_addr, node->size))
724
			return false;
725 726


727
		if (!access_ok((void __user *)a,
728
				    node->size))
729
			return false;
730
		else if (log_all && !log_access_ok(log_base,
731 732
						   node->start,
						   node->size))
733
			return false;
734
	}
735
	return true;
736 737
}

738 739 740 741 742 743 744 745 746 747 748 749
static inline void __user *vhost_vq_meta_fetch(struct vhost_virtqueue *vq,
					       u64 addr, unsigned int size,
					       int type)
{
	const struct vhost_umem_node *node = vq->meta_iotlb[type];

	if (!node)
		return NULL;

	return (void *)(uintptr_t)(node->userspace_addr + addr - node->start);
}

750 751
/* Can we switch to this memory table? */
/* Caller should have device mutex but not vq mutex */
752 753
static bool memory_access_ok(struct vhost_dev *d, struct vhost_umem *umem,
			     int log_all)
754 755
{
	int i;
756

757
	for (i = 0; i < d->nvqs; ++i) {
758
		bool ok;
759 760
		bool log;

761
		mutex_lock(&d->vqs[i]->mutex);
762
		log = log_all || vhost_has_feature(d->vqs[i], VHOST_F_LOG_ALL);
763
		/* If ring is inactive, will check when it's enabled. */
764
		if (d->vqs[i]->private_data)
765 766
			ok = vq_memory_access_ok(d->vqs[i]->log_base,
						 umem, log);
767
		else
768
			ok = true;
769
		mutex_unlock(&d->vqs[i]->mutex);
770
		if (!ok)
771
			return false;
772
	}
773
	return true;
774 775
}

Jason Wang's avatar
Jason Wang committed
776 777
static int translate_desc(struct vhost_virtqueue *vq, u64 addr, u32 len,
			  struct iovec iov[], int iov_size, int access);
778

779
static int vhost_copy_to_user(struct vhost_virtqueue *vq, void __user *to,
780 781
			      const void *from, unsigned size)
{
Jason Wang's avatar
Jason Wang committed
782
	int ret;
783

Jason Wang's avatar
Jason Wang committed
784 785 786 787 788 789 790 791 792
	if (!vq->iotlb)
		return __copy_to_user(to, from, size);
	else {
		/* This function should be called after iotlb
		 * prefetch, which means we're sure that all vq
		 * could be access through iotlb. So -EAGAIN should
		 * not happen in this case.
		 */
		struct iov_iter t;
793 794
		void __user *uaddr = vhost_vq_meta_fetch(vq,
				     (u64)(uintptr_t)to, size,
Eric Auger's avatar
Eric Auger committed
795
				     VHOST_ADDR_USED);
796 797 798 799

		if (uaddr)
			return __copy_to_user(uaddr, from, size);

Jason Wang's avatar
Jason Wang committed
800 801 802 803 804 805 806 807 808 809 810 811 812
		ret = translate_desc(vq, (u64)(uintptr_t)to, size, vq->iotlb_iov,
				     ARRAY_SIZE(vq->iotlb_iov),
				     VHOST_ACCESS_WO);
		if (ret < 0)
			goto out;
		iov_iter_init(&t, WRITE, vq->iotlb_iov, ret, size);
		ret = copy_to_iter(from, size, &t);
		if (ret == size)
			ret = 0;
	}
out:
	return ret;
}
813 814

static int vhost_copy_from_user(struct vhost_virtqueue *vq, void *to,
815
				void __user *from, unsigned size)
816
{
Jason Wang's avatar
Jason Wang committed
817 818 819 820 821 822 823 824 825 826
	int ret;

	if (!vq->iotlb)
		return __copy_from_user(to, from, size);
	else {
		/* This function should be called after iotlb
		 * prefetch, which means we're sure that vq
		 * could be access through iotlb. So -EAGAIN should
		 * not happen in this case.
		 */
827 828 829
		void __user *uaddr = vhost_vq_meta_fetch(vq,
				     (u64)(uintptr_t)from, size,
				     VHOST_ADDR_DESC);
Jason Wang's avatar
Jason Wang committed
830
		struct iov_iter f;
831 832 833 834

		if (uaddr)
			return __copy_from_user(to, uaddr, size);

Jason Wang's avatar
Jason Wang committed
835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853
		ret = translate_desc(vq, (u64)(uintptr_t)from, size, vq->iotlb_iov,
				     ARRAY_SIZE(vq->iotlb_iov),
				     VHOST_ACCESS_RO);
		if (ret < 0) {
			vq_err(vq, "IOTLB translation failure: uaddr "
			       "%p size 0x%llx\n", from,
			       (unsigned long long) size);
			goto out;
		}
		iov_iter_init(&f, READ, vq->iotlb_iov, ret, size);
		ret = copy_from_iter(to, size, &f);
		if (ret == size)
			ret = 0;
	}

out:
	return ret;
}

854 855 856
static void __user *__vhost_get_user_slow(struct vhost_virtqueue *vq,
					  void __user *addr, unsigned int size,
					  int type)
Jason Wang's avatar
Jason Wang committed
857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879
{
	int ret;

	ret = translate_desc(vq, (u64)(uintptr_t)addr, size, vq->iotlb_iov,
			     ARRAY_SIZE(vq->iotlb_iov),
			     VHOST_ACCESS_RO);
	if (ret < 0) {
		vq_err(vq, "IOTLB translation failure: uaddr "
			"%p size 0x%llx\n", addr,
			(unsigned long long) size);
		return NULL;
	}

	if (ret != 1 || vq->iotlb_iov[0].iov_len != size) {
		vq_err(vq, "Non atomic userspace memory access: uaddr "
			"%p size 0x%llx\n", addr,
			(unsigned long long) size);
		return NULL;
	}

	return vq->iotlb_iov[0].iov_base;
}

880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897
/* This function should be called after iotlb
 * prefetch, which means we're sure that vq
 * could be access through iotlb. So -EAGAIN should
 * not happen in this case.
 */
static inline void __user *__vhost_get_user(struct vhost_virtqueue *vq,
					    void *addr, unsigned int size,
					    int type)
{
	void __user *uaddr = vhost_vq_meta_fetch(vq,
			     (u64)(uintptr_t)addr, size, type);
	if (uaddr)
		return uaddr;

	return __vhost_get_user_slow(vq, addr, size, type);
}

#define vhost_put_user(vq, x, ptr)		\
Jason Wang's avatar
Jason Wang committed
898 899 900 901 902 903
({ \
	int ret = -EFAULT; \
	if (!vq->iotlb) { \
		ret = __put_user(x, ptr); \
	} else { \
		__typeof__(ptr) to = \
904 905
			(__typeof__(ptr)) __vhost_get_user(vq, ptr,	\
					  sizeof(*ptr), VHOST_ADDR_USED); \
Jason Wang's avatar
Jason Wang committed
906 907 908 909 910 911 912 913
		if (to != NULL) \
			ret = __put_user(x, to); \
		else \
			ret = -EFAULT;	\
	} \
	ret; \
})

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
static inline int vhost_put_avail_event(struct vhost_virtqueue *vq)
{
	return vhost_put_user(vq, cpu_to_vhost16(vq, vq->avail_idx),
			      vhost_avail_event(vq));
}

static inline int vhost_put_used(struct vhost_virtqueue *vq,
				 struct vring_used_elem *head, int idx,
				 int count)
{
	return vhost_copy_to_user(vq, vq->used->ring + idx, head,
				  count * sizeof(*head));
}

static inline int vhost_put_used_flags(struct vhost_virtqueue *vq)

{
	return vhost_put_user(vq, cpu_to_vhost16(vq, vq->used_flags),
			      &vq->used->flags);
}

static inline int vhost_put_used_idx(struct vhost_virtqueue *vq)

{
	return vhost_put_user(vq, cpu_to_vhost16(vq, vq->last_used_idx),
			      &vq->used->idx);
}

942
#define vhost_get_user(vq, x, ptr, type)		\
Jason Wang's avatar
Jason Wang committed
943 944 945 946 947 948
({ \
	int ret; \
	if (!vq->iotlb) { \
		ret = __get_user(x, ptr); \
	} else { \
		__typeof__(ptr) from = \
949 950 951
			(__typeof__(ptr)) __vhost_get_user(vq, ptr, \
							   sizeof(*ptr), \
							   type); \
Jason Wang's avatar
Jason Wang committed
952 953 954 955 956 957 958 959
		if (from != NULL) \
			ret = __get_user(x, from); \
		else \
			ret = -EFAULT; \
	} \
	ret; \
})

960 961 962 963 964 965
#define vhost_get_avail(vq, x, ptr) \
	vhost_get_user(vq, x, ptr, VHOST_ADDR_AVAIL)

#define vhost_get_used(vq, x, ptr) \
	vhost_get_user(vq, x, ptr, VHOST_ADDR_USED)

966 967 968 969 970 971 972 973 974 975 976 977 978 979
static void vhost_dev_lock_vqs(struct vhost_dev *d)
{
	int i = 0;
	for (i = 0; i < d->nvqs; ++i)
		mutex_lock_nested(&d->vqs[i]->mutex, i);
}

static void vhost_dev_unlock_vqs(struct vhost_dev *d)
{
	int i = 0;
	for (i = 0; i < d->nvqs; ++i)
		mutex_unlock(&d->vqs[i]->mutex);
}

980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016
static inline int vhost_get_avail_idx(struct vhost_virtqueue *vq,
				      __virtio16 *idx)
{
	return vhost_get_avail(vq, *idx, &vq->avail->idx);
}

static inline int vhost_get_avail_head(struct vhost_virtqueue *vq,
				       __virtio16 *head, int idx)
{
	return vhost_get_avail(vq, *head,
			       &vq->avail->ring[idx & (vq->num - 1)]);
}

static inline int vhost_get_avail_flags(struct vhost_virtqueue *vq,
					__virtio16 *flags)
{
	return vhost_get_avail(vq, *flags, &vq->avail->flags);
}

static inline int vhost_get_used_event(struct vhost_virtqueue *vq,
				       __virtio16 *event)
{
	return vhost_get_avail(vq, *event, vhost_used_event(vq));
}

static inline int vhost_get_used_idx(struct vhost_virtqueue *vq,
				     __virtio16 *idx)
{
	return vhost_get_used(vq, *idx, &vq->used->idx);
}

static inline int vhost_get_desc(struct vhost_virtqueue *vq,
				 struct vring_desc *desc, int idx)
{
	return vhost_copy_from_user(vq, desc, vq->desc + idx, sizeof(*desc));
}

Jason Wang's avatar
Jason Wang committed
1017 1018 1019 1020
static int vhost_new_umem_range(struct vhost_umem *umem,
				u64 start, u64 size, u64 end,
				u64 userspace_addr, int perm)
{
1021
	struct vhost_umem_node *tmp, *node;
Jason Wang's avatar
Jason Wang committed
1022

1023 1024 1025 1026
	if (!size)
		return -EFAULT;

	node = kmalloc(sizeof(*node), GFP_ATOMIC);
Jason Wang's avatar
Jason Wang committed
1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
	if (!node)
		return -ENOMEM;

	if (umem->numem == max_iotlb_entries) {
		tmp = list_first_entry(&umem->umem_list, typeof(*tmp), link);
		vhost_umem_free(umem, tmp);
	}

	node->start = start;
	node->size = size;
	node->last = end;
	node->userspace_addr = userspace_addr;
	node->perm = perm;
	INIT_LIST_HEAD(&node->link);
	list_add_tail(&node->link, &umem->umem_list);
	vhost_umem_interval_tree_insert(node, &umem->umem_tree);
	umem->numem++;

	return 0;
}

static void vhost_del_umem_range(struct vhost_umem *umem,
				 u64 start, u64 end)
{
	struct vhost_umem_node *node;

	while ((node = vhost_umem_interval_tree_iter_first(&umem->umem_tree,
							   start, end)))
		vhost_umem_free(umem, node);
}

static void vhost_iotlb_notify_vq(struct vhost_dev *d,
				  struct vhost_iotlb_msg *msg)
{
	struct vhost_msg_node *node, *n;

	spin_lock(&d->iotlb_lock);

	list_for_each_entry_safe(node, n, &d->pending_list, node) {
		struct vhost_iotlb_msg *vq_msg = &node->msg.iotlb;
		if (msg->iova <= vq_msg->iova &&
1068
		    msg->iova + msg->size - 1 >= vq_msg->iova &&
Jason Wang's avatar
Jason Wang committed
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
		    vq_msg->type == VHOST_IOTLB_MISS) {
			vhost_poll_queue(&node->vq->poll);
			list_del(&node->node);
			kfree(node);
		}
	}

	spin_unlock(&d->iotlb_lock);
}

1079
static bool umem_access_ok(u64 uaddr, u64 size, int access)
Jason Wang's avatar
Jason Wang committed
1080 1081 1082
{
	unsigned long a = uaddr;

1083 1084
	/* Make sure 64 bit math will not overflow. */
	if (vhost_overflow(uaddr, size))
1085
		return false;
1086

Jason Wang's avatar
Jason Wang committed
1087
	if ((access & VHOST_ACCESS_RO) &&
1088
	    !access_ok((void __user *)a, size))
1089
		return false;
Jason Wang's avatar
Jason Wang committed
1090
	if ((access & VHOST_ACCESS_WO) &&
1091
	    !access_ok((void __user *)a, size))
1092 1093
		return false;
	return true;
Jason Wang's avatar
Jason Wang committed
1094 1095
}

1096 1097
static int vhost_process_iotlb_msg(struct vhost_dev *dev,
				   struct vhost_iotlb_msg *msg)
Jason Wang's avatar
Jason Wang committed
1098 1099 1100
{
	int ret = 0;

1101
	mutex_lock(&dev->mutex);
1102
	vhost_dev_lock_vqs(dev);
Jason Wang's avatar
Jason Wang committed
1103 1104 1105 1106 1107 1108
	switch (msg->type) {
	case VHOST_IOTLB_UPDATE:
		if (!dev->iotlb) {
			ret = -EFAULT;
			break;
		}
1109
		if (!umem_access_ok(msg->uaddr, msg->size, msg->perm)) {
Jason Wang's avatar
Jason Wang committed
1110 1111 1112
			ret = -EFAULT;
			break;
		}
1113
		vhost_vq_meta_reset(dev);
Jason Wang's avatar
Jason Wang committed
1114 1115 1116 1117 1118 1119 1120 1121 1122
		if (vhost_new_umem_range(dev->iotlb, msg->iova, msg->size,
					 msg->iova + msg->size - 1,
					 msg->uaddr, msg->perm)) {
			ret = -ENOMEM;
			break;
		}
		vhost_iotlb_notify_vq(dev, msg);
		break;
	case VHOST_IOTLB_INVALIDATE:
1123 1124 1125 1126
		if (!dev->iotlb) {
			ret = -EFAULT;
			break;
		}
1127
		vhost_vq_meta_reset(dev);
Jason Wang's avatar
Jason Wang committed
1128 1129 1130 1131 1132 1133 1134 1135
		vhost_del_umem_range(dev->iotlb, msg->iova,
				     msg->iova + msg->size - 1);
		break;
	default:
		ret = -EINVAL;
		break;
	}

1136
	vhost_dev_unlock_vqs(dev);
1137 1138
	mutex_unlock(&dev->mutex);

Jason Wang's avatar
Jason Wang committed
1139 1140 1141 1142 1143
	return ret;
}
ssize_t vhost_chr_write_iter(struct vhost_dev *dev,
			     struct iov_iter *from)
{
1144 1145 1146
	struct vhost_iotlb_msg msg;
	size_t offset;
	int type, ret;
Jason Wang's avatar
Jason Wang committed
1147

1148
	ret = copy_from_iter(&type, sizeof(type), from);
1149 1150
	if (ret != sizeof(type)) {
		ret = -EINVAL;
Jason Wang's avatar
Jason Wang committed
1151
		goto done;
1152
	}
Jason Wang's avatar
Jason Wang committed
1153

1154
	switch (type) {
Jason Wang's avatar
Jason Wang committed
1155
	case VHOST_IOTLB_MSG:
1156 1157 1158 1159 1160 1161 1162
		/* There maybe a hole after type for V1 message type,
		 * so skip it here.
		 */
		offset = offsetof(struct vhost_msg, iotlb) - sizeof(int);
		break;
	case VHOST_IOTLB_MSG_V2:
		offset = sizeof(__u32);
Jason Wang's avatar
Jason Wang committed
1163 1164 1165
		break;
	default:
		ret = -EINVAL;
1166
		goto done;
Jason Wang's avatar
Jason Wang committed
1167 1168
	}

1169 1170
	iov_iter_advance(from, offset);
	ret = copy_from_iter(&msg, sizeof(msg), from);
1171 1172
	if (ret != sizeof(msg)) {
		ret = -EINVAL;
1173
		goto done;
1174
	}
1175 1176 1177 1178 1179 1180 1181
	if (vhost_process_iotlb_msg(dev, &msg)) {
		ret = -EFAULT;
		goto done;
	}

	ret = (type == VHOST_IOTLB_MSG) ? sizeof(struct vhost_msg) :
	      sizeof(struct vhost_msg_v2);
Jason Wang's avatar
Jason Wang committed
1182 1183 1184 1185 1186
done:
	return ret;
}
EXPORT_SYMBOL(vhost_chr_write_iter);

1187
__poll_t vhost_chr_poll(struct file *file, struct vhost_dev *dev,
Jason Wang's avatar
Jason Wang committed
1188 1189
			    poll_table *wait)
{
1190
	__poll_t mask = 0;
Jason Wang's avatar
Jason Wang committed
1191 1192 1193 1194

	poll_wait(file, &dev->wait, wait);

	if (!list_empty(&dev->read_list))
1195
		mask |= EPOLLIN | EPOLLRDNORM;
Jason Wang's avatar
Jason Wang committed
1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239

	return mask;
}
EXPORT_SYMBOL(vhost_chr_poll);

ssize_t vhost_chr_read_iter(struct vhost_dev *dev, struct iov_iter *to,
			    int noblock)
{
	DEFINE_WAIT(wait);
	struct vhost_msg_node *node;
	ssize_t ret = 0;
	unsigned size = sizeof(struct vhost_msg);

	if (iov_iter_count(to) < size)
		return 0;

	while (1) {
		if (!noblock)
			prepare_to_wait(&dev->wait, &wait,
					TASK_INTERRUPTIBLE);

		node = vhost_dequeue_msg(dev, &dev->read_list);
		if (node)
			break;
		if (noblock) {
			ret = -EAGAIN;
			break;
		}
		if (signal_pending(current)) {
			ret = -ERESTARTSYS;
			break;
		}
		if (!dev->iotlb) {
			ret = -EBADFD;
			break;
		}

		schedule();
	}

	if (!noblock)
		finish_wait(&dev->wait, &wait);

	if (node) {
1240 1241
		struct vhost_iotlb_msg *msg;
		void *start = &node->msg;