vhost.c 63.1 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
R
Rob Landley 已提交
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 17 18 19 20 21
#include <linux/mm.h>
#include <linux/miscdevice.h>
#include <linux/mutex.h>
#include <linux/poll.h>
#include <linux/file.h>
#include <linux/highmem.h>
22
#include <linux/slab.h>
23
#include <linux/vmalloc.h>
24
#include <linux/kthread.h>
25
#include <linux/cgroup.h>
A
Asias He 已提交
26
#include <linux/module.h>
27
#include <linux/sort.h>
28
#include <linux/sched/mm.h>
29
#include <linux/sched/signal.h>
30
#include <linux/interval_tree_generic.h>
J
Jason Wang 已提交
31
#include <linux/nospec.h>
32
#include <linux/kcov.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)");
J
Jason Wang 已提交
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])
M
Michael S. Tsirkin 已提交
51

52
#ifdef CONFIG_VHOST_CROSS_ENDIAN_LEGACY
G
Greg Kurz 已提交
53
static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
54 55 56 57
{
	vq->user_be = !virtio_legacy_is_little_endian();
}

G
Greg Kurz 已提交
58 59 60 61 62 63 64 65 66 67
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;
}

68 69 70 71 72 73 74 75 76 77 78 79 80 81
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;

G
Greg Kurz 已提交
82 83 84 85
	if (s.num == VHOST_VRING_BIG_ENDIAN)
		vhost_enable_cross_endian_big(vq);
	else
		vhost_enable_cross_endian_little(vq);
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113

	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
G
Greg Kurz 已提交
114
static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
{
}

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)
{
131 132
	vq->is_le = vhost_has_feature(vq, VIRTIO_F_VERSION_1)
		|| virtio_legacy_is_little_endian();
133 134 135
}
#endif /* CONFIG_VHOST_CROSS_ENDIAN_LEGACY */

G
Greg Kurz 已提交
136 137
static void vhost_reset_is_le(struct vhost_virtqueue *vq)
{
138
	vhost_init_is_le(vq);
G
Greg Kurz 已提交
139 140
}

J
Jason Wang 已提交
141 142 143 144 145 146 147 148 149 150 151 152 153
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);
}

154 155 156 157 158
static void vhost_poll_func(struct file *file, wait_queue_head_t *wqh,
			    poll_table *pt)
{
	struct vhost_poll *poll;

K
Krishna Kumar 已提交
159
	poll = container_of(pt, struct vhost_poll, table);
160 161 162 163
	poll->wqh = wqh;
	add_wait_queue(wqh, &poll->wait);
}

164
static int vhost_poll_wakeup(wait_queue_entry_t *wait, unsigned mode, int sync,
165 166
			     void *key)
{
167
	struct vhost_poll *poll = container_of(wait, struct vhost_poll, wait);
168
	struct vhost_work *work = &poll->work;
169

A
Al Viro 已提交
170
	if (!(key_to_poll(key) & poll->mask))
171 172
		return 0;

173 174 175 176 177
	if (!poll->dev->use_worker)
		work->fn(work);
	else
		vhost_poll_queue(poll);

178 179 180
	return 0;
}

181
void vhost_work_init(struct vhost_work *work, vhost_work_fn_t fn)
182
{
J
Jason Wang 已提交
183
	clear_bit(VHOST_WORK_QUEUED, &work->flags);
184 185
	work->fn = fn;
}
A
Asias He 已提交
186
EXPORT_SYMBOL_GPL(vhost_work_init);
187

188
/* Init poll structure */
189
void vhost_poll_init(struct vhost_poll *poll, vhost_work_fn_t fn,
A
Al Viro 已提交
190
		     __poll_t mask, struct vhost_dev *dev)
191 192 193 194
{
	init_waitqueue_func_entry(&poll->wait, vhost_poll_wakeup);
	init_poll_funcptr(&poll->table, vhost_poll_func);
	poll->mask = mask;
195
	poll->dev = dev;
196
	poll->wqh = NULL;
197

198
	vhost_work_init(&poll->work, fn);
199
}
A
Asias He 已提交
200
EXPORT_SYMBOL_GPL(vhost_poll_init);
201 202 203

/* 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. */
204
int vhost_poll_start(struct vhost_poll *poll, struct file *file)
205
{
206
	__poll_t mask;
K
Krishna Kumar 已提交
207

J
Jason Wang 已提交
208 209 210
	if (poll->wqh)
		return 0;

211
	mask = vfs_poll(file, &poll->table);
212
	if (mask)
A
Al Viro 已提交
213
		vhost_poll_wakeup(&poll->wait, 0, 0, poll_to_key(mask));
214
	if (mask & EPOLLERR) {
215
		vhost_poll_stop(poll);
216
		return -EINVAL;
217 218
	}

219
	return 0;
220
}
A
Asias He 已提交
221
EXPORT_SYMBOL_GPL(vhost_poll_start);
222 223 224 225 226

/* 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)
{
227 228 229 230
	if (poll->wqh) {
		remove_wait_queue(poll->wqh, &poll->wait);
		poll->wqh = NULL;
	}
231
}
A
Asias He 已提交
232
EXPORT_SYMBOL_GPL(vhost_poll_stop);
233

J
Jason Wang 已提交
234
void vhost_work_flush(struct vhost_dev *dev, struct vhost_work *work)
235
{
J
Jason Wang 已提交
236
	struct vhost_flush_struct flush;
K
Krishna Kumar 已提交
237

J
Jason Wang 已提交
238 239 240
	if (dev->worker) {
		init_completion(&flush.wait_event);
		vhost_work_init(&flush.work, vhost_flush_work);
241

J
Jason Wang 已提交
242 243 244
		vhost_work_queue(dev, &flush.work);
		wait_for_completion(&flush.wait_event);
	}
245
}
A
Asias He 已提交
246
EXPORT_SYMBOL_GPL(vhost_work_flush);
247

248 249 250 251 252 253
/* 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);
}
A
Asias He 已提交
254
EXPORT_SYMBOL_GPL(vhost_poll_flush);
255

256
void vhost_work_queue(struct vhost_dev *dev, struct vhost_work *work)
257
{
J
Jason Wang 已提交
258 259
	if (!dev->worker)
		return;
260

J
Jason Wang 已提交
261 262 263
	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.
264
		 * test_and_set_bit() implies a memory barrier.
J
Jason Wang 已提交
265 266
		 */
		llist_add(&work->node, &dev->work_list);
267 268
		wake_up_process(dev->worker);
	}
269
}
A
Asias He 已提交
270
EXPORT_SYMBOL_GPL(vhost_work_queue);
271

J
Jason Wang 已提交
272 273 274
/* A lockless hint for busy polling code to exit the loop */
bool vhost_has_work(struct vhost_dev *dev)
{
J
Jason Wang 已提交
275
	return !llist_empty(&dev->work_list);
J
Jason Wang 已提交
276 277 278
}
EXPORT_SYMBOL_GPL(vhost_has_work);

279 280 281 282
void vhost_poll_queue(struct vhost_poll *poll)
{
	vhost_work_queue(poll->dev, &poll->work);
}
A
Asias He 已提交
283
EXPORT_SYMBOL_GPL(vhost_poll_queue);
284

285 286 287 288 289 290 291 292 293 294 295 296
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;

297
	for (i = 0; i < d->nvqs; ++i)
298 299 300
		__vhost_vq_meta_reset(d->vqs[i]);
}

Z
Zhu Lingshan 已提交
301 302 303 304 305 306
static void vhost_vring_call_reset(struct vhost_vring_call *call_ctx)
{
	call_ctx->ctx = NULL;
	memset(&call_ctx->producer, 0x0, sizeof(struct irq_bypass_producer));
}

307 308 309 310 311 312
bool vhost_vq_is_setup(struct vhost_virtqueue *vq)
{
	return vq->avail && vq->desc && vq->used && vhost_vq_access_ok(vq);
}
EXPORT_SYMBOL_GPL(vhost_vq_is_setup);

313 314 315 316 317 318 319 320 321 322
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;
M
Michael S. Tsirkin 已提交
323 324
	vq->signalled_used = 0;
	vq->signalled_used_valid = false;
325 326 327 328
	vq->used_flags = 0;
	vq->log_used = false;
	vq->log_addr = -1ull;
	vq->private_data = NULL;
329
	vq->acked_features = 0;
330
	vq->acked_backend_features = 0;
331 332 333
	vq->log_base = NULL;
	vq->error_ctx = NULL;
	vq->kick = NULL;
334
	vq->log_ctx = NULL;
G
Greg Kurz 已提交
335
	vhost_disable_cross_endian(vq);
L
Laurent Vivier 已提交
336
	vhost_reset_is_le(vq);
J
Jason Wang 已提交
337
	vq->busyloop_timeout = 0;
338
	vq->umem = NULL;
J
Jason Wang 已提交
339
	vq->iotlb = NULL;
Z
Zhu Lingshan 已提交
340
	vhost_vring_call_reset(&vq->call_ctx);
341
	__vhost_vq_meta_reset(vq);
342 343
}

344 345 346
static int vhost_worker(void *data)
{
	struct vhost_dev *dev = data;
J
Jason Wang 已提交
347 348
	struct vhost_work *work, *work_next;
	struct llist_node *node;
349

350
	kthread_use_mm(dev->mm);
M
Michael S. Tsirkin 已提交
351

352 353 354 355 356 357
	for (;;) {
		/* mb paired w/ kthread_stop */
		set_current_state(TASK_INTERRUPTIBLE);

		if (kthread_should_stop()) {
			__set_current_state(TASK_RUNNING);
M
Michael S. Tsirkin 已提交
358
			break;
359
		}
J
Jason Wang 已提交
360 361 362 363 364 365 366 367 368 369

		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);
370
			__set_current_state(TASK_RUNNING);
371
			kcov_remote_start_common(dev->kcov_handle);
372
			work->fn(work);
373
			kcov_remote_stop();
N
Nadav Har'El 已提交
374 375
			if (need_resched())
				schedule();
J
Jason Wang 已提交
376
		}
377
	}
378
	kthread_unuse_mm(dev->mm);
M
Michael S. Tsirkin 已提交
379
	return 0;
380 381
}

382 383 384 385 386 387 388 389 390 391
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;
}

J
Jason Wang 已提交
392 393 394
/* Helper to allocate iovec buffers for all vqs. */
static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
{
A
Asias He 已提交
395
	struct vhost_virtqueue *vq;
J
Jason Wang 已提交
396
	int i;
K
Krishna Kumar 已提交
397

J
Jason Wang 已提交
398
	for (i = 0; i < dev->nvqs; ++i) {
A
Asias He 已提交
399
		vq = dev->vqs[i];
400 401 402
		vq->indirect = kmalloc_array(UIO_MAXIOV,
					     sizeof(*vq->indirect),
					     GFP_KERNEL);
J
Jason Wang 已提交
403
		vq->log = kmalloc_array(dev->iov_limit, sizeof(*vq->log),
404
					GFP_KERNEL);
J
Jason Wang 已提交
405
		vq->heads = kmalloc_array(dev->iov_limit, sizeof(*vq->heads),
406
					  GFP_KERNEL);
A
Asias He 已提交
407
		if (!vq->indirect || !vq->log || !vq->heads)
J
Jason Wang 已提交
408 409 410
			goto err_nomem;
	}
	return 0;
K
Krishna Kumar 已提交
411

J
Jason Wang 已提交
412
err_nomem:
413
	for (; i >= 0; --i)
414
		vhost_vq_free_iovecs(dev->vqs[i]);
J
Jason Wang 已提交
415 416 417 418 419 420
	return -ENOMEM;
}

static void vhost_dev_free_iovecs(struct vhost_dev *dev)
{
	int i;
K
Krishna Kumar 已提交
421

422
	for (i = 0; i < dev->nvqs; ++i)
423
		vhost_vq_free_iovecs(dev->vqs[i]);
J
Jason Wang 已提交
424 425
}

426 427 428 429 430 431 432 433 434 435 436 437 438 439 440
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);

441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466
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;
}

Z
Zhi Yong Wu 已提交
467
void vhost_dev_init(struct vhost_dev *dev,
468
		    struct vhost_virtqueue **vqs, int nvqs,
469
		    int iov_limit, int weight, int byte_weight,
470
		    bool use_worker,
G
Gautam Dawar 已提交
471
		    int (*msg_handler)(struct vhost_dev *dev, u32 asid,
472
				       struct vhost_iotlb_msg *msg))
473
{
A
Asias He 已提交
474
	struct vhost_virtqueue *vq;
475
	int i;
476

477 478 479 480
	dev->vqs = vqs;
	dev->nvqs = nvqs;
	mutex_init(&dev->mutex);
	dev->log_ctx = NULL;
481
	dev->umem = NULL;
J
Jason Wang 已提交
482
	dev->iotlb = NULL;
483
	dev->mm = NULL;
484
	dev->worker = NULL;
J
Jason Wang 已提交
485
	dev->iov_limit = iov_limit;
486 487
	dev->weight = weight;
	dev->byte_weight = byte_weight;
488
	dev->use_worker = use_worker;
489
	dev->msg_handler = msg_handler;
J
Jason Wang 已提交
490
	init_llist_head(&dev->work_list);
J
Jason Wang 已提交
491 492 493 494
	init_waitqueue_head(&dev->wait);
	INIT_LIST_HEAD(&dev->read_list);
	INIT_LIST_HEAD(&dev->pending_list);
	spin_lock_init(&dev->iotlb_lock);
495

496 497

	for (i = 0; i < dev->nvqs; ++i) {
A
Asias He 已提交
498 499 500 501 502 503 504 505 506
		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,
507
					EPOLLIN, dev);
508 509
	}
}
A
Asias He 已提交
510
EXPORT_SYMBOL_GPL(vhost_dev_init);
511 512 513 514 515 516 517

/* 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;
}
A
Asias He 已提交
518
EXPORT_SYMBOL_GPL(vhost_dev_check_owner);
519

520
struct vhost_attach_cgroups_struct {
K
Krishna Kumar 已提交
521 522 523
	struct vhost_work work;
	struct task_struct *owner;
	int ret;
524 525 526 527
};

static void vhost_attach_cgroups_work(struct vhost_work *work)
{
K
Krishna Kumar 已提交
528 529 530 531
	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);
532 533 534 535
}

static int vhost_attach_cgroups(struct vhost_dev *dev)
{
K
Krishna Kumar 已提交
536 537 538 539 540 541 542
	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;
543 544
}

545 546 547 548 549
/* Caller should have device mutex */
bool vhost_dev_has_owner(struct vhost_dev *dev)
{
	return dev->mm;
}
A
Asias He 已提交
550
EXPORT_SYMBOL_GPL(vhost_dev_has_owner);
551

552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581
static void vhost_attach_mm(struct vhost_dev *dev)
{
	/* No owner, become one */
	if (dev->use_worker) {
		dev->mm = get_task_mm(current);
	} else {
		/* vDPA device does not use worker thead, so there's
		 * no need to hold the address space for mm. This help
		 * to avoid deadlock in the case of mmap() which may
		 * held the refcnt of the file and depends on release
		 * method to remove vma.
		 */
		dev->mm = current->mm;
		mmgrab(dev->mm);
	}
}

static void vhost_detach_mm(struct vhost_dev *dev)
{
	if (!dev->mm)
		return;

	if (dev->use_worker)
		mmput(dev->mm);
	else
		mmdrop(dev->mm);

	dev->mm = NULL;
}

582
/* Caller should have device mutex */
A
Asias He 已提交
583
long vhost_dev_set_owner(struct vhost_dev *dev)
584
{
585 586
	struct task_struct *worker;
	int err;
K
Krishna Kumar 已提交
587

588
	/* Is there an owner already? */
589
	if (vhost_dev_has_owner(dev)) {
590 591 592
		err = -EBUSY;
		goto err_mm;
	}
K
Krishna Kumar 已提交
593

594 595
	vhost_attach_mm(dev);

596
	dev->kcov_handle = kcov_common_handle();
597 598 599 600 601 602 603
	if (dev->use_worker) {
		worker = kthread_create(vhost_worker, dev,
					"vhost-%d", current->pid);
		if (IS_ERR(worker)) {
			err = PTR_ERR(worker);
			goto err_worker;
		}
604

605 606
		dev->worker = worker;
		wake_up_process(worker); /* avoid contributing to loadavg */
607

608 609 610 611
		err = vhost_attach_cgroups(dev);
		if (err)
			goto err_cgroup;
	}
612

J
Jason Wang 已提交
613 614 615 616
	err = vhost_dev_alloc_iovecs(dev);
	if (err)
		goto err_cgroup;

617
	return 0;
618
err_cgroup:
619 620 621 622
	if (dev->worker) {
		kthread_stop(dev->worker);
		dev->worker = NULL;
	}
623
err_worker:
624
	vhost_detach_mm(dev);
625
	dev->kcov_handle = 0;
626 627
err_mm:
	return err;
628
}
A
Asias He 已提交
629
EXPORT_SYMBOL_GPL(vhost_dev_set_owner);
630

J
Jason Wang 已提交
631 632 633 634 635 636 637
static struct vhost_iotlb *iotlb_alloc(void)
{
	return vhost_iotlb_alloc(max_iotlb_entries,
				 VHOST_IOTLB_FLAG_RETIRE);
}

struct vhost_iotlb *vhost_dev_reset_owner_prepare(void)
638
{
J
Jason Wang 已提交
639
	return iotlb_alloc();
640
}
A
Asias He 已提交
641
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner_prepare);
642

643
/* Caller should have device mutex */
J
Jason Wang 已提交
644
void vhost_dev_reset_owner(struct vhost_dev *dev, struct vhost_iotlb *umem)
645
{
646 647
	int i;

648
	vhost_dev_cleanup(dev);
649

650
	dev->umem = umem;
651 652 653 654
	/* We don't need VQ locks below since vhost_dev_cleanup makes sure
	 * VQs aren't running.
	 */
	for (i = 0; i < dev->nvqs; ++i)
655
		dev->vqs[i]->umem = umem;
656
}
A
Asias He 已提交
657
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner);
658

659
void vhost_dev_stop(struct vhost_dev *dev)
660 661
{
	int i;
662 663

	for (i = 0; i < dev->nvqs; ++i) {
664 665 666
		if (dev->vqs[i]->kick && dev->vqs[i]->handle_kick) {
			vhost_poll_stop(&dev->vqs[i]->poll);
			vhost_poll_flush(&dev->vqs[i]->poll);
667
		}
668 669
	}
}
A
Asias He 已提交
670
EXPORT_SYMBOL_GPL(vhost_dev_stop);
671

J
Jason Wang 已提交
672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690
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);
}

691
void vhost_dev_cleanup(struct vhost_dev *dev)
692 693
{
	int i;
K
Krishna Kumar 已提交
694

695
	for (i = 0; i < dev->nvqs; ++i) {
696 697 698 699
		if (dev->vqs[i]->error_ctx)
			eventfd_ctx_put(dev->vqs[i]->error_ctx);
		if (dev->vqs[i]->kick)
			fput(dev->vqs[i]->kick);
Z
Zhu Lingshan 已提交
700 701
		if (dev->vqs[i]->call_ctx.ctx)
			eventfd_ctx_put(dev->vqs[i]->call_ctx.ctx);
702
		vhost_vq_reset(dev, dev->vqs[i]);
703
	}
J
Jason Wang 已提交
704
	vhost_dev_free_iovecs(dev);
705 706 707 708
	if (dev->log_ctx)
		eventfd_ctx_put(dev->log_ctx);
	dev->log_ctx = NULL;
	/* No one will access memory at this point */
J
Jason Wang 已提交
709
	vhost_iotlb_free(dev->umem);
710
	dev->umem = NULL;
J
Jason Wang 已提交
711
	vhost_iotlb_free(dev->iotlb);
J
Jason Wang 已提交
712 713
	dev->iotlb = NULL;
	vhost_clear_msg(dev);
714
	wake_up_interruptible_poll(&dev->wait, EPOLLIN | EPOLLRDNORM);
J
Jason Wang 已提交
715
	WARN_ON(!llist_empty(&dev->work_list));
716 717 718
	if (dev->worker) {
		kthread_stop(dev->worker);
		dev->worker = NULL;
719
		dev->kcov_handle = 0;
720
	}
721
	vhost_detach_mm(dev);
722
}
A
Asias He 已提交
723
EXPORT_SYMBOL_GPL(vhost_dev_cleanup);
724

725
static bool log_access_ok(void __user *log_base, u64 addr, unsigned long sz)
726 727
{
	u64 a = addr / VHOST_PAGE_SIZE / 8;
K
Krishna Kumar 已提交
728

729 730 731
	/* Make sure 64 bit math will not overflow. */
	if (a > ULONG_MAX - (unsigned long)log_base ||
	    a + (unsigned long)log_base > ULONG_MAX)
732
		return false;
733

734
	return access_ok(log_base + a,
735 736 737
			 (sz + VHOST_PAGE_SIZE * 8 - 1) / VHOST_PAGE_SIZE / 8);
}

738
/* Make sure 64 bit math will not overflow. */
739 740
static bool vhost_overflow(u64 uaddr, u64 size)
{
741 742 743 744 745 746 747
	if (uaddr > ULONG_MAX || size > ULONG_MAX)
		return true;

	if (!size)
		return false;

	return uaddr > ULONG_MAX - size + 1;
748 749
}

750
/* Caller should have vq mutex and device mutex. */
J
Jason Wang 已提交
751
static bool vq_memory_access_ok(void __user *log_base, struct vhost_iotlb *umem,
752
				int log_all)
753
{
J
Jason Wang 已提交
754
	struct vhost_iotlb_map *map;
755

756
	if (!umem)
757
		return false;
758

J
Jason Wang 已提交
759 760
	list_for_each_entry(map, &umem->list, link) {
		unsigned long a = map->addr;
761

J
Jason Wang 已提交
762
		if (vhost_overflow(map->addr, map->size))
763
			return false;
764 765


J
Jason Wang 已提交
766
		if (!access_ok((void __user *)a, map->size))
767
			return false;
768
		else if (log_all && !log_access_ok(log_base,
J
Jason Wang 已提交
769 770
						   map->start,
						   map->size))
771
			return false;
772
	}
773
	return true;
774 775
}

776 777 778 779
static inline void __user *vhost_vq_meta_fetch(struct vhost_virtqueue *vq,
					       u64 addr, unsigned int size,
					       int type)
{
J
Jason Wang 已提交
780
	const struct vhost_iotlb_map *map = vq->meta_iotlb[type];
781

J
Jason Wang 已提交
782
	if (!map)
783 784
		return NULL;

M
Michael S. Tsirkin 已提交
785
	return (void __user *)(uintptr_t)(map->addr + addr - map->start);
786 787
}

788 789
/* Can we switch to this memory table? */
/* Caller should have device mutex but not vq mutex */
J
Jason Wang 已提交
790
static bool memory_access_ok(struct vhost_dev *d, struct vhost_iotlb *umem,
791
			     int log_all)
792 793
{
	int i;
K
Krishna Kumar 已提交
794

795
	for (i = 0; i < d->nvqs; ++i) {
796
		bool ok;
797 798
		bool log;

799
		mutex_lock(&d->vqs[i]->mutex);
800
		log = log_all || vhost_has_feature(d->vqs[i], VHOST_F_LOG_ALL);
801
		/* If ring is inactive, will check when it's enabled. */
802
		if (d->vqs[i]->private_data)
803 804
			ok = vq_memory_access_ok(d->vqs[i]->log_base,
						 umem, log);
805
		else
806
			ok = true;
807
		mutex_unlock(&d->vqs[i]->mutex);
808
		if (!ok)
809
			return false;
810
	}
811
	return true;
812 813
}

J
Jason Wang 已提交
814 815
static int translate_desc(struct vhost_virtqueue *vq, u64 addr, u32 len,
			  struct iovec iov[], int iov_size, int access);
816

817
static int vhost_copy_to_user(struct vhost_virtqueue *vq, void __user *to,
818 819
			      const void *from, unsigned size)
{
J
Jason Wang 已提交
820
	int ret;
821

J
Jason Wang 已提交
822 823 824 825 826 827 828 829 830
	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;
831 832
		void __user *uaddr = vhost_vq_meta_fetch(vq,
				     (u64)(uintptr_t)to, size,
E
Eric Auger 已提交
833
				     VHOST_ADDR_USED);
834 835 836 837

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

J
Jason Wang 已提交
838 839 840 841 842 843 844 845 846 847 848 849 850
		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;
}
851 852

static int vhost_copy_from_user(struct vhost_virtqueue *vq, void *to,
853
				void __user *from, unsigned size)
854
{
J
Jason Wang 已提交
855 856 857 858 859 860 861 862 863 864
	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.
		 */
865 866 867
		void __user *uaddr = vhost_vq_meta_fetch(vq,
				     (u64)(uintptr_t)from, size,
				     VHOST_ADDR_DESC);
J
Jason Wang 已提交
868
		struct iov_iter f;
869 870 871 872

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

J
Jason Wang 已提交
873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891
		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;
}

892 893 894
static void __user *__vhost_get_user_slow(struct vhost_virtqueue *vq,
					  void __user *addr, unsigned int size,
					  int type)
J
Jason Wang 已提交
895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917
{
	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;
}

918 919 920 921 922 923
/* 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,
M
Michael S. Tsirkin 已提交
924
					    void __user *addr, unsigned int size,
925 926 927 928 929 930 931 932 933 934 935
					    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)		\
J
Jason Wang 已提交
936
({ \
937
	int ret; \
J
Jason Wang 已提交
938 939 940 941
	if (!vq->iotlb) { \
		ret = __put_user(x, ptr); \
	} else { \
		__typeof__(ptr) to = \
942 943
			(__typeof__(ptr)) __vhost_get_user(vq, ptr,	\
					  sizeof(*ptr), VHOST_ADDR_USED); \
J
Jason Wang 已提交
944 945 946 947 948 949 950 951
		if (to != NULL) \
			ret = __put_user(x, to); \
		else \
			ret = -EFAULT;	\
	} \
	ret; \
})

952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979
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);
}

980
#define vhost_get_user(vq, x, ptr, type)		\
J
Jason Wang 已提交
981 982 983 984 985 986
({ \
	int ret; \
	if (!vq->iotlb) { \
		ret = __get_user(x, ptr); \
	} else { \
		__typeof__(ptr) from = \
987 988 989
			(__typeof__(ptr)) __vhost_get_user(vq, ptr, \
							   sizeof(*ptr), \
							   type); \
J
Jason Wang 已提交
990 991 992 993 994 995 996 997
		if (from != NULL) \
			ret = __get_user(x, from); \
		else \
			ret = -EFAULT; \
	} \
	ret; \
})

998 999 1000 1001 1002 1003
#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)

1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
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);
}

1018 1019 1020 1021 1022 1023 1024 1025 1026 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
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));
}

J
Jason Wang 已提交
1055 1056 1057 1058 1059 1060 1061 1062 1063 1064
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 &&
1065
		    msg->iova + msg->size - 1 >= vq_msg->iova &&
J
Jason Wang 已提交
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
		    vq_msg->type == VHOST_IOTLB_MISS) {
			vhost_poll_queue(&node->vq->poll);
			list_del(&node->node);
			kfree(node);
		}
	}

	spin_unlock(&d->iotlb_lock);
}

1076
static bool umem_access_ok(u64 uaddr, u64 size, int access)
J
Jason Wang 已提交
1077 1078 1079
{
	unsigned long a = uaddr;

1080 1081
	/* Make sure 64 bit math will not overflow. */
	if (vhost_overflow(uaddr, size))
1082
		return false;
1083

J
Jason Wang 已提交
1084
	if ((access & VHOST_ACCESS_RO) &&
1085
	    !access_ok((void __user *)a, size))
1086
		return false;
J
Jason Wang 已提交
1087
	if ((access & VHOST_ACCESS_WO) &&
1088
	    !access_ok((void __user *)a, size))
1089 1090
		return false;
	return true;
J
Jason Wang 已提交
1091 1092
}

G
Gautam Dawar 已提交
1093
static int vhost_process_iotlb_msg(struct vhost_dev *dev, u32 asid,
1094
				   struct vhost_iotlb_msg *msg)
J
Jason Wang 已提交
1095 1096 1097
{
	int ret = 0;

G
Gautam Dawar 已提交
1098 1099 1100
	if (asid != 0)
		return -EINVAL;

1101
	mutex_lock(&dev->mutex);
1102
	vhost_dev_lock_vqs(dev);
J
Jason Wang 已提交
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)) {
J
Jason Wang 已提交
1110 1111 1112
			ret = -EFAULT;
			break;
		}
1113
		vhost_vq_meta_reset(dev);
J
Jason Wang 已提交
1114 1115 1116
		if (vhost_iotlb_add_range(dev->iotlb, msg->iova,
					  msg->iova + msg->size - 1,
					  msg->uaddr, msg->perm)) {
J
Jason Wang 已提交
1117 1118 1119 1120 1121 1122
			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);
J
Jason Wang 已提交
1128 1129
		vhost_iotlb_del_range(dev->iotlb, msg->iova,
				      msg->iova + msg->size - 1);
J
Jason Wang 已提交
1130 1131 1132 1133 1134 1135
		break;
	default:
		ret = -EINVAL;
		break;
	}

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

J
Jason Wang 已提交
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;
G
Gautam Dawar 已提交
1147
	u32 asid = 0;
J
Jason Wang 已提交
1148

1149
	ret = copy_from_iter(&type, sizeof(type), from);
1150 1151
	if (ret != sizeof(type)) {
		ret = -EINVAL;
J
Jason Wang 已提交
1152
		goto done;
1153
	}
J
Jason Wang 已提交
1154

1155
	switch (type) {
J
Jason Wang 已提交
1156
	case VHOST_IOTLB_MSG:
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:
G
Gautam Dawar 已提交
1163 1164 1165 1166 1167 1168 1169
		if (vhost_backend_has_feature(dev->vqs[0],
					      VHOST_BACKEND_F_IOTLB_ASID)) {
			ret = copy_from_iter(&asid, sizeof(asid), from);
			if (ret != sizeof(asid)) {
				ret = -EINVAL;
				goto done;
			}
1170
			offset = 0;
G
Gautam Dawar 已提交
1171 1172
		} else
			offset = sizeof(__u32);
J
Jason Wang 已提交
1173 1174 1175
		break;
	default:
		ret = -EINVAL;
1176
		goto done;
J
Jason Wang 已提交
1177 1178
	}

1179 1180
	iov_iter_advance(from, offset);
	ret = copy_from_iter(&msg, sizeof(msg), from);
1181 1182
	if (ret != sizeof(msg)) {
		ret = -EINVAL;
1183
		goto done;
1184
	}
1185

1186 1187 1188 1189 1190
	if (msg.size == 0) {
		ret = -EINVAL;
		goto done;
	}

1191
	if (dev->msg_handler)
G
Gautam Dawar 已提交
1192
		ret = dev->msg_handler(dev, asid, &msg);
1193
	else
G
Gautam Dawar 已提交
1194
		ret = vhost_process_iotlb_msg(dev, asid, &msg);
1195
	if (ret) {
1196 1197 1198 1199 1200 1201
		ret = -EFAULT;
		goto done;
	}

	ret = (type == VHOST_IOTLB_MSG) ? sizeof(struct vhost_msg) :
	      sizeof(struct vhost_msg_v2);
J
Jason Wang 已提交
1202 1203 1204 1205 1206
done:
	return ret;
}
EXPORT_SYMBOL(vhost_chr_write_iter);

1207
__poll_t vhost_chr_poll(struct file *file, struct vhost_dev *dev,
J
Jason Wang 已提交
1208 1209
			    poll_table *wait)
{
1210
	__poll_t mask = 0;
J
Jason Wang 已提交
1211 1212 1213 1214

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

	if (!list_empty(&dev->read_list))
1215
		mask |= EPOLLIN | EPOLLRDNORM;
J
Jason Wang 已提交
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259

	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) {
1260 1261
		struct vhost_iotlb_msg *msg;
		void *start = &node->msg;
J
Jason Wang 已提交
1262

1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
		switch (node->msg.type) {
		case VHOST_IOTLB_MSG:
			size = sizeof(node->msg);
			msg = &node->msg.iotlb;
			break;
		case VHOST_IOTLB_MSG_V2:
			size = sizeof(node->msg_v2);
			msg = &node->msg_v2.iotlb;
			break;
		default:
			BUG();
			break;
		}

		ret = copy_to_iter(start, size, to);
		if (ret != size || msg->type != VHOST_IOTLB_MISS) {
J
Jason Wang 已提交
1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293
			kfree(node);
			return ret;
		}
		vhost_enqueue_msg(dev, &dev->pending_list, node);
	}

	return ret;
}
EXPORT_SYMBOL_GPL(vhost_chr_read_iter);

static int vhost_iotlb_miss(struct vhost_virtqueue *vq, u64 iova, int access)
{
	struct vhost_dev *dev = vq->dev;
	struct vhost_msg_node *node;
	struct vhost_iotlb_msg *msg;
1294
	bool v2 = vhost_backend_has_feature(vq, VHOST_BACKEND_F_IOTLB_MSG_V2);
J
Jason Wang 已提交
1295

1296
	node = vhost_new_msg(vq, v2 ? VHOST_IOTLB_MSG_V2 : VHOST_IOTLB_MSG);
J
Jason Wang 已提交
1297 1298 1299
	if (!node)
		return -ENOMEM;

1300 1301 1302 1303 1304 1305 1306
	if (v2) {
		node->msg_v2.type = VHOST_IOTLB_MSG_V2;
		msg = &node->msg_v2.iotlb;
	} else {
		msg = &node->msg.iotlb;
	}

J
Jason Wang 已提交
1307 1308 1309 1310 1311 1312 1313
	msg->type = VHOST_IOTLB_MISS;
	msg->iova = iova;
	msg->perm = access;

	vhost_enqueue_msg(dev, &dev->read_list, node);

	return 0;
1314 1315
}

1316
static bool vq_access_ok(struct vhost_virtqueue *vq, unsigned int num,
1317 1318 1319
			 vring_desc_t __user *desc,
			 vring_avail_t __user *avail,
			 vring_used_t __user *used)
J
Jason Wang 已提交
1320

1321
{
1322 1323 1324 1325 1326
	/* If an IOTLB device is present, the vring addresses are
	 * GIOVAs. Access validation occurs at prefetch time. */
	if (vq->iotlb)
		return true;

1327 1328 1329
	return access_ok(desc, vhost_get_desc_size(vq, num)) &&
	       access_ok(avail, vhost_get_avail_size(vq, num)) &&
	       access_ok(used, vhost_get_used_size(vq, num));
1330 1331
}

1332
static void vhost_vq_meta_update(struct vhost_virtqueue *vq,
J
Jason Wang 已提交
1333
				 const struct vhost_iotlb_map *map,
1334 1335 1336 1337 1338
				 int type)
{
	int access = (type == VHOST_ADDR_USED) ?
		     VHOST_ACCESS_WO : VHOST_ACCESS_RO;

J
Jason Wang 已提交
1339 1340
	if (likely(map->perm & access))
		vq->meta_iotlb[type] = map;
1341 1342
}

1343 1344
static bool iotlb_access_ok(struct vhost_virtqueue *vq,
			    int access, u64 addr, u64 len, int type)
J
Jason Wang 已提交
1345
{
J
Jason Wang 已提交
1346 1347
	const struct vhost_iotlb_map *map;
	struct vhost_iotlb *umem = vq->iotlb;
1348
	u64 s = 0, size, orig_addr = addr, last = addr + len - 1;
1349 1350 1351

	if (vhost_vq_meta_fetch(vq, addr, len, type))
		return true;
J
Jason Wang 已提交
1352 1353

	while (len > s) {
J
Jason Wang 已提交
1354 1355
		map = vhost_iotlb_itree_first(umem, addr, last);
		if (map == NULL || map->start > addr) {
J
Jason Wang 已提交
1356 1357
			vhost_iotlb_miss(vq, addr, access);
			return false;
J
Jason Wang 已提交
1358
		} else if (!(map->perm & access)) {
J
Jason Wang 已提交
1359 1360 1361 1362 1363 1364
			/* Report the possible access violation by
			 * request another translation from userspace.
			 */
			return false;
		}

J
Jason Wang 已提交
1365
		size = map->size - addr + map->start;
1366 1367

		if (orig_addr == addr && size >= len)
J
Jason Wang 已提交
1368
			vhost_vq_meta_update(vq, map, type);
1369

J
Jason Wang 已提交
1370 1371 1372 1373 1374 1375 1376
		s += size;
		addr += size;
	}

	return true;
}

1377
int vq_meta_prefetch(struct vhost_virtqueue *vq)
J
Jason Wang 已提交
1378 1379 1380
{
	unsigned int num = vq->num;

1381
	if (!vq->iotlb)
J
Jason Wang 已提交
1382 1383
		return 1;

J
Jason Wang 已提交
1384
	return iotlb_access_ok(vq, VHOST_MAP_RO, (u64)(uintptr_t)vq->desc,
1385
			       vhost_get_desc_size(vq, num), VHOST_ADDR_DESC) &&
J
Jason Wang 已提交
1386
	       iotlb_access_ok(vq, VHOST_MAP_RO, (u64)(uintptr_t)vq->avail,
1387
			       vhost_get_avail_size(vq, num),
1388
			       VHOST_ADDR_AVAIL) &&
J
Jason Wang 已提交
1389
	       iotlb_access_ok(vq, VHOST_MAP_WO, (u64)(uintptr_t)vq->used,
1390
			       vhost_get_used_size(vq, num), VHOST_ADDR_USED);
J
Jason Wang 已提交
1391
}
1392
EXPORT_SYMBOL_GPL(vq_meta_prefetch);
J
Jason Wang 已提交
1393

1394 1395
/* Can we log writes? */
/* Caller should have device mutex but not vq mutex */
1396
bool vhost_log_access_ok(struct vhost_dev *dev)
1397
{
1398
	return memory_access_ok(dev, dev->umem, 1);
1399
}
A
Asias He 已提交
1400
EXPORT_SYMBOL_GPL(vhost_log_access_ok);
1401

1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415
static bool vq_log_used_access_ok(struct vhost_virtqueue *vq,
				  void __user *log_base,
				  bool log_used,
				  u64 log_addr)
{
	/* If an IOTLB device is present, log_addr is a GIOVA that
	 * will never be logged by log_used(). */
	if (vq->iotlb)
		return true;

	return !log_used || log_access_ok(log_base, log_addr,
					  vhost_get_used_size(vq, vq->num));
}

1416 1417
/* Verify access for write logging. */
/* Caller should have vq mutex and device mutex */
1418 1419
static bool vq_log_access_ok(struct vhost_virtqueue *vq,
			     void __user *log_base)
1420
{
1421
	return vq_memory_access_ok(log_base, vq->umem,
1422
				   vhost_has_feature(vq, VHOST_F_LOG_ALL)) &&
1423
		vq_log_used_access_ok(vq, log_base, vq->log_used, vq->log_addr);
1424 1425 1426 1427
}

/* Can we start vq? */
/* Caller should have vq mutex and device mutex */
1428
bool vhost_vq_access_ok(struct vhost_virtqueue *vq)
1429
{
1430
	if (!vq_log_access_ok(vq, vq->log_base))
1431
		return false;
1432 1433

	return vq_access_ok(vq, vq->num, vq->desc, vq->avail, vq->used);
1434
}
A
Asias He 已提交
1435
EXPORT_SYMBOL_GPL(vhost_vq_access_ok);
1436 1437 1438

static long vhost_set_memory(struct vhost_dev *d, struct vhost_memory __user *m)
{
1439 1440
	struct vhost_memory mem, *newmem;
	struct vhost_memory_region *region;
J
Jason Wang 已提交
1441
	struct vhost_iotlb *newumem, *oldumem;
1442
	unsigned long size = offsetof(struct vhost_memory, regions);
1443
	int i;
K
Krishna Kumar 已提交
1444

1445 1446
	if (copy_from_user(&mem, m, size))
		return -EFAULT;
1447 1448
	if (mem.padding)
		return -EOPNOTSUPP;
1449
	if (mem.nregions > max_mem_regions)
1450
		return -E2BIG;
M
Matthew Wilcox 已提交
1451 1452
	newmem = kvzalloc(struct_size(newmem, regions, mem.nregions),
			GFP_KERNEL);
1453 1454 1455 1456
	if (!newmem)
		return -ENOMEM;

	memcpy(newmem, &mem, size);
1457
	if (copy_from_user(newmem->regions, m->regions,
1458
			   flex_array_size(newmem, regions, mem.nregions))) {
1459
		kvfree(newmem);
1460
		return -EFAULT;
1461 1462
	}

J
Jason Wang 已提交
1463
	newumem = iotlb_alloc();
1464
	if (!newumem) {
1465
		kvfree(newmem);
1466 1467 1468 1469 1470 1471
		return -ENOMEM;
	}

	for (region = newmem->regions;
	     region < newmem->regions + mem.nregions;
	     region++) {
J
Jason Wang 已提交
1472 1473 1474 1475 1476 1477
		if (vhost_iotlb_add_range(newumem,
					  region->guest_phys_addr,
					  region->guest_phys_addr +
					  region->memory_size - 1,
					  region->userspace_addr,
					  VHOST_MAP_RW))
1478
			goto err;
1479
	}
1480 1481 1482 1483 1484 1485

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

	oldumem = d->umem;
	d->umem = newumem;
1486

1487
	/* All memory accesses are done under some VQ mutex. */
1488 1489
	for (i = 0; i < d->nvqs; ++i) {
		mutex_lock(&d->vqs[i]->mutex);
1490
		d->vqs[i]->umem = newumem;
1491 1492
		mutex_unlock(&d->vqs[i]->mutex);
	}
1493 1494

	kvfree(newmem);
J
Jason Wang 已提交
1495
	vhost_iotlb_free(oldumem);
1496
	return 0;
1497 1498

err:
J
Jason Wang 已提交
1499
	vhost_iotlb_free(newumem);
1500 1501
	kvfree(newmem);
	return -EFAULT;
1502 1503
}

1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
static long vhost_vring_set_num(struct vhost_dev *d,
				struct vhost_virtqueue *vq,
				void __user *argp)
{
	struct vhost_vring_state s;

	/* Resizing ring with an active backend?
	 * You don't want to do that. */
	if (vq->private_data)
		return -EBUSY;

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

	if (!s.num || s.num > 0xffff || (s.num & (s.num - 1)))
		return -EINVAL;
	vq->num = s.num;

	return 0;
}

static long vhost_vring_set_addr(struct vhost_dev *d,
				 struct vhost_virtqueue *vq,
				 void __user *argp)
{
	struct vhost_vring_addr a;

	if (copy_from_user(&a, argp, sizeof a))
		return -EFAULT;
	if (a.flags & ~(0x1 << VHOST_VRING_F_LOG))
		return -EOPNOTSUPP;

	/* 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)
		return -EFAULT;

	/* 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)) ||
	    (a.log_guest_addr & (VRING_USED_ALIGN_SIZE - 1)))
		return -EINVAL;

	/* 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) {
		if (!vq_access_ok(vq, vq->num,
			(void __user *)(unsigned long)a.desc_user_addr,
			(void __user *)(unsigned long)a.avail_user_addr,
			(void __user *)(unsigned long)a.used_user_addr))
			return -EINVAL;

		/* Also validate log access for used ring if enabled. */
1562 1563 1564
		if (!vq_log_used_access_ok(vq, vq->log_base,
				a.flags & (0x1 << VHOST_VRING_F_LOG),
				a.log_guest_addr))
1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
			return -EINVAL;
	}

	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;

	return 0;
}

static long vhost_vring_set_num_addr(struct vhost_dev *d,
				     struct vhost_virtqueue *vq,
				     unsigned int ioctl,
				     void __user *argp)
{
	long r;

	mutex_lock(&vq->mutex);

	switch (ioctl) {
	case VHOST_SET_VRING_NUM:
		r = vhost_vring_set_num(d, vq, argp);
		break;
	case VHOST_SET_VRING_ADDR:
		r = vhost_vring_set_addr(d, vq, argp);
		break;
	default:
		BUG();
	}

	mutex_unlock(&vq->mutex);

	return r;
}
1601
long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
1602
{
1603 1604
	struct file *eventfp, *filep = NULL;
	bool pollstart = false, pollstop = false;
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
	struct eventfd_ctx *ctx = NULL;
	u32 __user *idxp = argp;
	struct vhost_virtqueue *vq;
	struct vhost_vring_state s;
	struct vhost_vring_file f;
	u32 idx;
	long r;

	r = get_user(idx, idxp);
	if (r < 0)
		return r;
1616
	if (idx >= d->nvqs)
1617 1618
		return -ENOBUFS;

J
Jason Wang 已提交
1619
	idx = array_index_nospec(idx, d->nvqs);
1620
	vq = d->vqs[idx];
1621

1622 1623 1624 1625 1626
	if (ioctl == VHOST_SET_VRING_NUM ||
	    ioctl == VHOST_SET_VRING_ADDR) {
		return vhost_vring_set_num_addr(d, vq, ioctl, argp);
	}

1627 1628 1629 1630 1631 1632 1633 1634 1635 1636
	mutex_lock(&vq->mutex);

	switch (ioctl) {
	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;
		}
1637 1638
		if (copy_from_user(&s, argp, sizeof s)) {
			r = -EFAULT;
1639
			break;
1640
		}
1641 1642 1643 1644
		if (s.num > 0xffff) {
			r = -EINVAL;
			break;
		}
1645
		vq->last_avail_idx = s.num;
1646 1647 1648 1649 1650 1651
		/* 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;
1652 1653
		if (copy_to_user(argp, &s, sizeof s))
			r = -EFAULT;
1654 1655
		break;
	case VHOST_SET_VRING_KICK:
1656 1657
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
1658
			break;
1659
		}
1660
		eventfp = f.fd == VHOST_FILE_UNBIND ? NULL : eventfd_fget(f.fd);
1661 1662 1663 1664
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
1665
		if (eventfp != vq->kick) {
1666 1667
			pollstop = (filep = vq->kick) != NULL;
			pollstart = (vq->kick = eventfp) != NULL;
1668 1669 1670 1671
		} else
			filep = eventfp;
		break;
	case VHOST_SET_VRING_CALL:
1672 1673
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
1674
			break;
1675
		}
1676
		ctx = f.fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(f.fd);
1677 1678
		if (IS_ERR(ctx)) {
			r = PTR_ERR(ctx);
1679 1680
			break;
		}
Z
Zhu Lingshan 已提交
1681 1682

		swap(ctx, vq->call_ctx.ctx);
1683 1684
		break;
	case VHOST_SET_VRING_ERR:
1685 1686
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
1687
			break;
1688
		}
1689
		ctx = f.fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(f.fd);
1690 1691
		if (IS_ERR(ctx)) {
			r = PTR_ERR(ctx);
1692 1693
			break;
		}
1694
		swap(ctx, vq->error_ctx);
1695
		break;
1696 1697 1698 1699 1700 1701
	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;
J
Jason Wang 已提交
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714
	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;
1715 1716 1717 1718 1719 1720 1721
	default:
		r = -ENOIOCTLCMD;
	}

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

1722
	if (!IS_ERR_OR_NULL(ctx))
1723 1724 1725 1726 1727
		eventfd_ctx_put(ctx);
	if (filep)
		fput(filep);

	if (pollstart && vq->handle_kick)
1728
		r = vhost_poll_start(&vq->poll, vq->kick);
1729 1730 1731 1732 1733 1734 1735

	mutex_unlock(&vq->mutex);

	if (pollstop && vq->handle_kick)
		vhost_poll_flush(&vq->poll);
	return r;
}
A
Asias He 已提交
1736
EXPORT_SYMBOL_GPL(vhost_vring_ioctl);
1737

J
Jason Wang 已提交
1738 1739
int vhost_init_device_iotlb(struct vhost_dev *d, bool enabled)
{
J
Jason Wang 已提交
1740
	struct vhost_iotlb *niotlb, *oiotlb;
J
Jason Wang 已提交
1741 1742
	int i;

J
Jason Wang 已提交
1743
	niotlb = iotlb_alloc();
J
Jason Wang 已提交
1744 1745 1746 1747 1748 1749 1750
	if (!niotlb)
		return -ENOMEM;

	oiotlb = d->iotlb;
	d->iotlb = niotlb;

	for (i = 0; i < d->nvqs; ++i) {
1751 1752 1753 1754 1755 1756
		struct vhost_virtqueue *vq = d->vqs[i];

		mutex_lock(&vq->mutex);
		vq->iotlb = niotlb;
		__vhost_vq_meta_reset(vq);
		mutex_unlock(&vq->mutex);
J
Jason Wang 已提交
1757 1758
	}

J
Jason Wang 已提交
1759
	vhost_iotlb_free(oiotlb);
J
Jason Wang 已提交
1760 1761 1762 1763 1764

	return 0;
}
EXPORT_SYMBOL_GPL(vhost_init_device_iotlb);

1765
/* Caller must have device mutex */
1766
long vhost_dev_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
1767
{
1768
	struct eventfd_ctx *ctx;
1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788
	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:
1789 1790
		if (copy_from_user(&p, argp, sizeof p)) {
			r = -EFAULT;
1791
			break;
1792
		}
1793 1794 1795 1796 1797 1798 1799
		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;
1800
			vq = d->vqs[i];
1801 1802
			mutex_lock(&vq->mutex);
			/* If ring is inactive, will check when it's enabled. */
1803
			if (vq->private_data && !vq_log_access_ok(vq, base))
1804 1805 1806 1807 1808 1809 1810 1811 1812 1813
				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;
1814
		ctx = fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(fd);
1815 1816
		if (IS_ERR(ctx)) {
			r = PTR_ERR(ctx);
1817 1818
			break;
		}
1819
		swap(ctx, d->log_ctx);
1820
		for (i = 0; i < d->nvqs; ++i) {
1821 1822 1823
			mutex_lock(&d->vqs[i]->mutex);
			d->vqs[i]->log_ctx = d->log_ctx;
			mutex_unlock(&d->vqs[i]->mutex);
1824 1825 1826 1827 1828
		}
		if (ctx)
			eventfd_ctx_put(ctx);
		break;
	default:
1829
		r = -ENOIOCTLCMD;
1830 1831 1832 1833 1834
		break;
	}
done:
	return r;
}
A
Asias He 已提交
1835
EXPORT_SYMBOL_GPL(vhost_dev_ioctl);
1836 1837 1838

/* TODO: This is really inefficient.  We need something like get_user()
 * (instruction directly accesses the data, with an exception table entry
1839
 * returning -EFAULT). See Documentation/x86/exception-tables.rst.
1840 1841 1842 1843 1844 1845 1846 1847
 */
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;
K
Krishna Kumar 已提交
1848

1849
	r = pin_user_pages_fast(log, 1, FOLL_WRITE, &page);
1850
	if (r < 0)
1851
		return r;
1852
	BUG_ON(r != 1);
1853
	base = kmap_atomic(page);
1854
	set_bit(bit, base);
1855
	kunmap_atomic(base);
1856
	unpin_user_pages_dirty_lock(&page, 1, true);
1857 1858 1859 1860 1861 1862
	return 0;
}

static int log_write(void __user *log_base,
		     u64 write_address, u64 write_length)
{
1863
	u64 write_page = write_address / VHOST_PAGE_SIZE;
1864
	int r;
K
Krishna Kumar 已提交
1865

1866 1867
	if (!write_length)
		return 0;
1868
	write_length += write_address % VHOST_PAGE_SIZE;
1869 1870
	for (;;) {
		u64 base = (u64)(unsigned long)log_base;
1871 1872
		u64 log = base + write_page / 8;
		int bit = write_page % 8;
1873 1874 1875 1876 1877 1878 1879 1880
		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;
1881
		write_page += 1;
1882 1883 1884 1885
	}
	return r;
}

J
Jason Wang 已提交
1886 1887
static int log_write_hva(struct vhost_virtqueue *vq, u64 hva, u64 len)
{
J
Jason Wang 已提交
1888 1889
	struct vhost_iotlb *umem = vq->umem;
	struct vhost_iotlb_map *u;
J
Jason Wang 已提交
1890 1891 1892 1893 1894 1895 1896 1897 1898
	u64 start, end, l, min;
	int r;
	bool hit = false;

	while (len) {
		min = len;
		/* More than one GPAs can be mapped into a single HVA. So
		 * iterate all possible umems here to be safe.
		 */
J
Jason Wang 已提交
1899 1900 1901
		list_for_each_entry(u, &umem->list, link) {
			if (u->addr > hva - 1 + len ||
			    u->addr - 1 + u->size < hva)
J
Jason Wang 已提交
1902
				continue;
J
Jason Wang 已提交
1903 1904
			start = max(u->addr, hva);
			end = min(u->addr - 1 + u->size, hva - 1 + len);
J
Jason Wang 已提交
1905 1906
			l = end - start + 1;
			r = log_write(vq->log_base,
J
Jason Wang 已提交
1907
				      u->start + start - u->addr,
J
Jason Wang 已提交
1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
				      l);
			if (r < 0)
				return r;
			hit = true;
			min = min(l, min);
		}

		if (!hit)
			return -EFAULT;

		len -= min;
		hva += min;
	}

	return 0;
}

static int log_used(struct vhost_virtqueue *vq, u64 used_offset, u64 len)
{
L
Li Wang 已提交
1927
	struct iovec *iov = vq->log_iov;
J
Jason Wang 已提交
1928 1929 1930 1931 1932 1933 1934
	int i, ret;

	if (!vq->iotlb)
		return log_write(vq->log_base, vq->log_addr + used_offset, len);

	ret = translate_desc(vq, (uintptr_t)vq->used + used_offset,
			     len, iov, 64, VHOST_ACCESS_WO);
1935
	if (ret < 0)
J
Jason Wang 已提交
1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
		return ret;

	for (i = 0; i < ret; i++) {
		ret = log_write_hva(vq,	(uintptr_t)iov[i].iov_base,
				    iov[i].iov_len);
		if (ret)
			return ret;
	}

	return 0;
}

1948
int vhost_log_write(struct vhost_virtqueue *vq, struct vhost_log *log,
J
Jason Wang 已提交
1949
		    unsigned int log_num, u64 len, struct iovec *iov, int count)
1950 1951 1952 1953
{
	int i, r;

	/* Make sure data written is seen before log. */
1954
	smp_wmb();
J
Jason Wang 已提交
1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965

	if (vq->iotlb) {
		for (i = 0; i < count; i++) {
			r = log_write_hva(vq, (uintptr_t)iov[i].iov_base,
					  iov[i].iov_len);
			if (r < 0)
				return r;
		}
		return 0;
	}

1966 1967 1968 1969 1970 1971
	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;
1972 1973 1974
		if (!len) {
			if (vq->log_ctx)
				eventfd_signal(vq->log_ctx, 1);
1975
			return 0;
1976
		}
1977 1978 1979 1980 1981
	}
	/* Length written exceeds what we have stored. This is a bug. */
	BUG();
	return 0;
}
A
Asias He 已提交
1982
EXPORT_SYMBOL_GPL(vhost_log_write);
1983

1984 1985 1986
static int vhost_update_used_flags(struct vhost_virtqueue *vq)
{
	void __user *used;
1987
	if (vhost_put_used_flags(vq))
1988 1989 1990 1991 1992 1993
		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;
J
Jason Wang 已提交
1994 1995
		log_used(vq, (used - (void __user *)vq->used),
			 sizeof vq->used->flags);
1996 1997 1998 1999 2000 2001 2002 2003
		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)
{
2004
	if (vhost_put_avail_event(vq))
2005 2006 2007 2008 2009 2010 2011
		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);
J
Jason Wang 已提交
2012 2013
		log_used(vq, (used - (void __user *)vq->used),
			 sizeof *vhost_avail_event(vq));
2014 2015 2016 2017 2018 2019
		if (vq->log_ctx)
			eventfd_signal(vq->log_ctx, 1);
	}
	return 0;
}

G
Greg Kurz 已提交
2020
int vhost_vq_init_access(struct vhost_virtqueue *vq)
2021
{
2022
	__virtio16 last_used_idx;
2023
	int r;
2024 2025
	bool is_le = vq->is_le;

2026
	if (!vq->private_data)
2027
		return 0;
2028 2029

	vhost_init_is_le(vq);
2030 2031 2032

	r = vhost_update_used_flags(vq);
	if (r)
2033
		goto err;
2034
	vq->signalled_used_valid = false;
J
Jason Wang 已提交
2035
	if (!vq->iotlb &&
2036
	    !access_ok(&vq->used->idx, sizeof vq->used->idx)) {
2037 2038 2039
		r = -EFAULT;
		goto err;
	}
2040
	r = vhost_get_used_idx(vq, &last_used_idx);
J
Jason Wang 已提交
2041 2042 2043
	if (r) {
		vq_err(vq, "Can't access used idx at %p\n",
		       &vq->used->idx);
2044
		goto err;
J
Jason Wang 已提交
2045
	}
2046
	vq->last_used_idx = vhost16_to_cpu(vq, last_used_idx);
2047
	return 0;
J
Jason Wang 已提交
2048

2049 2050 2051
err:
	vq->is_le = is_le;
	return r;
2052
}
G
Greg Kurz 已提交
2053
EXPORT_SYMBOL_GPL(vhost_vq_init_access);
2054

2055
static int translate_desc(struct vhost_virtqueue *vq, u64 addr, u32 len,
J
Jason Wang 已提交
2056
			  struct iovec iov[], int iov_size, int access)
2057
{
J
Jason Wang 已提交
2058
	const struct vhost_iotlb_map *map;
J
Jason Wang 已提交
2059
	struct vhost_dev *dev = vq->dev;
J
Jason Wang 已提交
2060
	struct vhost_iotlb *umem = dev->iotlb ? dev->iotlb : dev->umem;
2061 2062 2063 2064 2065 2066
	struct iovec *_iov;
	u64 s = 0;
	int ret = 0;

	while ((u64)len > s) {
		u64 size;
2067
		if (unlikely(ret >= iov_size)) {
2068 2069 2070
			ret = -ENOBUFS;
			break;
		}
J
Jason Wang 已提交
2071

J
Jason Wang 已提交
2072 2073
		map = vhost_iotlb_itree_first(umem, addr, addr + len - 1);
		if (map == NULL || map->start > addr) {
J
Jason Wang 已提交
2074 2075 2076 2077 2078 2079
			if (umem != dev->iotlb) {
				ret = -EFAULT;
				break;
			}
			ret = -EAGAIN;
			break;
J
Jason Wang 已提交
2080
		} else if (!(map->perm & access)) {
J
Jason Wang 已提交
2081
			ret = -EPERM;
2082 2083
			break;
		}
J
Jason Wang 已提交
2084

2085
		_iov = iov + ret;
J
Jason Wang 已提交
2086
		size = map->size - addr + map->start;
2087
		_iov->iov_len = min((u64)len - s, size);
2088
		_iov->iov_base = (void __user *)(unsigned long)
J
Jason Wang 已提交
2089
				 (map->addr + addr - map->start);
2090 2091 2092 2093 2094
		s += size;
		addr += size;
		++ret;
	}

J
Jason Wang 已提交
2095 2096
	if (ret == -EAGAIN)
		vhost_iotlb_miss(vq, addr, access);
2097 2098 2099 2100 2101 2102
	return ret;
}

/* Each buffer in the virtqueues is actually a chain of descriptors.  This
 * function returns the next descriptor in the chain,
 * or -1U if we're at the end. */
2103
static unsigned next_desc(struct vhost_virtqueue *vq, struct vring_desc *desc)
2104 2105 2106 2107
{
	unsigned int next;

	/* If this descriptor says it doesn't chain, we're done. */
2108
	if (!(desc->flags & cpu_to_vhost16(vq, VRING_DESC_F_NEXT)))
2109 2110 2111
		return -1U;

	/* Check they're not leading us off end of descriptors. */
2112
	next = vhost16_to_cpu(vq, READ_ONCE(desc->next));
2113 2114 2115
	return next;
}

2116
static int get_indirect(struct vhost_virtqueue *vq,
2117 2118 2119 2120
			struct iovec iov[], unsigned int iov_size,
			unsigned int *out_num, unsigned int *in_num,
			struct vhost_log *log, unsigned int *log_num,
			struct vring_desc *indirect)
2121 2122 2123
{
	struct vring_desc desc;
	unsigned int i = 0, count, found = 0;
2124
	u32 len = vhost32_to_cpu(vq, indirect->len);
2125
	struct iov_iter from;
J
Jason Wang 已提交
2126
	int ret, access;
2127 2128

	/* Sanity check */
2129
	if (unlikely(len % sizeof desc)) {
2130 2131
		vq_err(vq, "Invalid length in indirect descriptor: "
		       "len 0x%llx not multiple of 0x%zx\n",
2132
		       (unsigned long long)len,
2133 2134 2135 2136
		       sizeof desc);
		return -EINVAL;
	}

2137
	ret = translate_desc(vq, vhost64_to_cpu(vq, indirect->addr), len, vq->indirect,
J
Jason Wang 已提交
2138
			     UIO_MAXIOV, VHOST_ACCESS_RO);
2139
	if (unlikely(ret < 0)) {
J
Jason Wang 已提交
2140 2141
		if (ret != -EAGAIN)
			vq_err(vq, "Translation failure %d in indirect.\n", ret);
2142 2143
		return ret;
	}
2144
	iov_iter_init(&from, READ, vq->indirect, ret, len);
2145
	count = len / sizeof desc;
2146 2147
	/* Buffers are chained via a 16 bit next field, so
	 * we can have at most 2^16 of these. */
2148
	if (unlikely(count > USHRT_MAX + 1)) {
2149 2150 2151 2152 2153 2154 2155
		vq_err(vq, "Indirect buffer length too big: %d\n",
		       indirect->len);
		return -E2BIG;
	}

	do {
		unsigned iov_count = *in_num + *out_num;
2156
		if (unlikely(++found > count)) {
2157 2158 2159 2160 2161
			vq_err(vq, "Loop detected: last one at %u "
			       "indirect size %u\n",
			       i, count);
			return -EINVAL;
		}
2162
		if (unlikely(!copy_from_iter_full(&desc, sizeof(desc), &from))) {
2163
			vq_err(vq, "Failed indirect descriptor: idx %d, %zx\n",
2164
			       i, (size_t)vhost64_to_cpu(vq, indirect->addr) + i * sizeof desc);
2165 2166
			return -EINVAL;
		}
2167
		if (unlikely(desc.flags & cpu_to_vhost16(vq, VRING_DESC_F_INDIRECT))) {
2168
			vq_err(vq, "Nested indirect descriptor: idx %d, %zx\n",
2169
			       i, (size_t)vhost64_to_cpu(vq, indirect->addr) + i * sizeof desc);
2170 2171 2172
			return -EINVAL;
		}

J
Jason Wang 已提交
2173 2174 2175 2176 2177
		if (desc.flags & cpu_to_vhost16(vq, VRING_DESC_F_WRITE))
			access = VHOST_ACCESS_WO;
		else
			access = VHOST_ACCESS_RO;

2178 2179
		ret = translate_desc(vq, vhost64_to_cpu(vq, desc.addr),
				     vhost32_to_cpu(vq, desc.len), iov + iov_count,
J
Jason Wang 已提交
2180
				     iov_size - iov_count, access);
2181
		if (unlikely(ret < 0)) {
J
Jason Wang 已提交
2182 2183 2184
			if (ret != -EAGAIN)
				vq_err(vq, "Translation failure %d indirect idx %d\n",
					ret, i);
2185 2186 2187
			return ret;
		}
		/* If this is an input descriptor, increment that count. */
J
Jason Wang 已提交
2188
		if (access == VHOST_ACCESS_WO) {
2189
			*in_num += ret;
Y
yongduan 已提交
2190
			if (unlikely(log && ret)) {
2191 2192
				log[*log_num].addr = vhost64_to_cpu(vq, desc.addr);
				log[*log_num].len = vhost32_to_cpu(vq, desc.len);
2193 2194 2195 2196 2197
				++*log_num;
			}
		} else {
			/* If it's an output descriptor, they're all supposed
			 * to come before any input descriptors. */
2198
			if (unlikely(*in_num)) {
2199 2200 2201 2202 2203 2204
				vq_err(vq, "Indirect descriptor "
				       "has out after in: idx %d\n", i);
				return -EINVAL;
			}
			*out_num += ret;
		}
2205
	} while ((i = next_desc(vq, &desc)) != -1);
2206 2207 2208 2209 2210 2211 2212 2213
	return 0;
}

/* This looks in the virtqueue and for the first available buffer, and converts
 * it to an iovec for convenient access.  Since descriptors consist of some
 * number of output then some number of input descriptors, it's actually two
 * iovecs, but we pack them into one and note how many of each there were.
 *
2214 2215 2216
 * This function returns the descriptor number found, or vq->num (which is
 * never a valid descriptor number) if none was found.  A negative code is
 * returned on error. */
2217
int vhost_get_vq_desc(struct vhost_virtqueue *vq,
2218 2219 2220
		      struct iovec iov[], unsigned int iov_size,
		      unsigned int *out_num, unsigned int *in_num,
		      struct vhost_log *log, unsigned int *log_num)
2221 2222 2223 2224
{
	struct vring_desc desc;
	unsigned int i, head, found = 0;
	u16 last_avail_idx;
2225 2226
	__virtio16 avail_idx;
	__virtio16 ring_head;
J
Jason Wang 已提交
2227
	int ret, access;
2228 2229 2230 2231

	/* Check it isn't doing very strange things with descriptor numbers. */
	last_avail_idx = vq->last_avail_idx;

2232
	if (vq->avail_idx == vq->last_avail_idx) {
2233
		if (unlikely(vhost_get_avail_idx(vq, &avail_idx))) {
2234 2235 2236 2237 2238
			vq_err(vq, "Failed to access avail idx at %p\n",
				&vq->avail->idx);
			return -EFAULT;
		}
		vq->avail_idx = vhost16_to_cpu(vq, avail_idx);
2239

2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250
		if (unlikely((u16)(vq->avail_idx - last_avail_idx) > vq->num)) {
			vq_err(vq, "Guest moved used index from %u to %u",
				last_avail_idx, vq->avail_idx);
			return -EFAULT;
		}

		/* If there's nothing new since last we looked, return
		 * invalid.
		 */
		if (vq->avail_idx == last_avail_idx)
			return vq->num;
2251

2252 2253 2254 2255 2256
		/* Only get avail ring entries after they have been
		 * exposed by guest.
		 */
		smp_rmb();
	}
2257 2258 2259

	/* Grab the next descriptor number they're advertising, and increment
	 * the index we've seen. */
2260
	if (unlikely(vhost_get_avail_head(vq, &ring_head, last_avail_idx))) {
2261 2262 2263
		vq_err(vq, "Failed to read head: idx %d address %p\n",
		       last_avail_idx,
		       &vq->avail->ring[last_avail_idx % vq->num]);
2264
		return -EFAULT;
2265 2266
	}

2267 2268
	head = vhost16_to_cpu(vq, ring_head);

2269
	/* If their number is silly, that's an error. */
2270
	if (unlikely(head >= vq->num)) {
2271 2272
		vq_err(vq, "Guest says index %u > %u is available",
		       head, vq->num);
2273
		return -EINVAL;
2274 2275 2276 2277 2278 2279 2280 2281 2282 2283
	}

	/* When we start there are none of either input nor output. */
	*out_num = *in_num = 0;
	if (unlikely(log))
		*log_num = 0;

	i = head;
	do {
		unsigned iov_count = *in_num + *out_num;
2284
		if (unlikely(i >= vq->num)) {
2285 2286
			vq_err(vq, "Desc index is %u > %u, head = %u",
			       i, vq->num, head);
2287
			return -EINVAL;
2288
		}
2289
		if (unlikely(++found > vq->num)) {
2290 2291 2292
			vq_err(vq, "Loop detected: last one at %u "
			       "vq size %u head %u\n",
			       i, vq->num, head);
2293
			return -EINVAL;
2294
		}
2295
		ret = vhost_get_desc(vq, &desc, i);
2296
		if (unlikely(ret)) {
2297 2298
			vq_err(vq, "Failed to get descriptor: idx %d addr %p\n",
			       i, vq->desc + i);
2299
			return -EFAULT;
2300
		}
2301
		if (desc.flags & cpu_to_vhost16(vq, VRING_DESC_F_INDIRECT)) {
2302
			ret = get_indirect(vq, iov, iov_size,
2303 2304
					   out_num, in_num,
					   log, log_num, &desc);
2305
			if (unlikely(ret < 0)) {
J
Jason Wang 已提交
2306 2307 2308
				if (ret != -EAGAIN)
					vq_err(vq, "Failure detected "
						"in indirect descriptor at idx %d\n", i);
2309
				return ret;
2310 2311 2312 2313
			}
			continue;
		}

J
Jason Wang 已提交
2314 2315 2316 2317
		if (desc.flags & cpu_to_vhost16(vq, VRING_DESC_F_WRITE))
			access = VHOST_ACCESS_WO;
		else
			access = VHOST_ACCESS_RO;
2318 2319
		ret = translate_desc(vq, vhost64_to_cpu(vq, desc.addr),
				     vhost32_to_cpu(vq, desc.len), iov + iov_count,
J
Jason Wang 已提交
2320
				     iov_size - iov_count, access);
2321
		if (unlikely(ret < 0)) {
J
Jason Wang 已提交
2322 2323 2324
			if (ret != -EAGAIN)
				vq_err(vq, "Translation failure %d descriptor idx %d\n",
					ret, i);
2325
			return ret;
2326
		}
J
Jason Wang 已提交
2327
		if (access == VHOST_ACCESS_WO) {
2328 2329 2330
			/* If this is an input descriptor,
			 * increment that count. */
			*in_num += ret;
Y
yongduan 已提交
2331
			if (unlikely(log && ret)) {
2332 2333
				log[*log_num].addr = vhost64_to_cpu(vq, desc.addr);
				log[*log_num].len = vhost32_to_cpu(vq, desc.len);
2334 2335 2336 2337 2338
				++*log_num;
			}
		} else {
			/* If it's an output descriptor, they're all supposed
			 * to come before any input descriptors. */
2339
			if (unlikely(*in_num)) {
2340 2341
				vq_err(vq, "Descriptor has out after in: "
				       "idx %d\n", i);
2342
				return -EINVAL;
2343 2344 2345
			}
			*out_num += ret;
		}
2346
	} while ((i = next_desc(vq, &desc)) != -1);
2347 2348 2349

	/* On success, increment avail index. */
	vq->last_avail_idx++;
M
Michael S. Tsirkin 已提交
2350 2351 2352 2353

	/* Assume notifications from guest are disabled at this point,
	 * if they aren't we would need to update avail_event index. */
	BUG_ON(!(vq->used_flags & VRING_USED_F_NO_NOTIFY));
2354 2355
	return head;
}
A
Asias He 已提交
2356
EXPORT_SYMBOL_GPL(vhost_get_vq_desc);
2357 2358

/* Reverse the effect of vhost_get_vq_desc. Useful for error handling. */
2359
void vhost_discard_vq_desc(struct vhost_virtqueue *vq, int n)
2360
{
2361
	vq->last_avail_idx -= n;
2362
}
A
Asias He 已提交
2363
EXPORT_SYMBOL_GPL(vhost_discard_vq_desc);
2364 2365 2366 2367 2368

/* After we've used one of their buffers, we tell them about it.  We'll then
 * want to notify the guest, using eventfd. */
int vhost_add_used(struct vhost_virtqueue *vq, unsigned int head, int len)
{
2369 2370 2371 2372
	struct vring_used_elem heads = {
		cpu_to_vhost32(vq, head),
		cpu_to_vhost32(vq, len)
	};
2373

2374
	return vhost_add_used_n(vq, &heads, 1);
2375
}
A
Asias He 已提交
2376
EXPORT_SYMBOL_GPL(vhost_add_used);
2377

2378 2379 2380 2381
static int __vhost_add_used_n(struct vhost_virtqueue *vq,
			    struct vring_used_elem *heads,
			    unsigned count)
{
2382
	vring_used_elem_t __user *used;
M
Michael S. Tsirkin 已提交
2383
	u16 old, new;
2384 2385
	int start;

2386
	start = vq->last_used_idx & (vq->num - 1);
2387
	used = vq->used->ring + start;
2388
	if (vhost_put_used(vq, heads, start, count)) {
2389 2390 2391 2392 2393 2394 2395
		vq_err(vq, "Failed to write used");
		return -EFAULT;
	}
	if (unlikely(vq->log_used)) {
		/* Make sure data is seen before log. */
		smp_wmb();
		/* Log used ring entry write. */
J
Jason Wang 已提交
2396 2397
		log_used(vq, ((void __user *)used - (void __user *)vq->used),
			 count * sizeof *used);
2398
	}
M
Michael S. Tsirkin 已提交
2399 2400 2401 2402 2403 2404 2405 2406
	old = vq->last_used_idx;
	new = (vq->last_used_idx += count);
	/* If the driver never bothers to signal in a very long while,
	 * used index might wrap around. If that happens, invalidate
	 * signalled_used index we stored. TODO: make sure driver
	 * signals at least once in 2^16 and remove this. */
	if (unlikely((u16)(new - vq->signalled_used) < (u16)(new - old)))
		vq->signalled_used_valid = false;
2407 2408 2409 2410 2411 2412 2413 2414 2415 2416
	return 0;
}

/* After we've used one of their buffers, we tell them about it.  We'll then
 * want to notify the guest, using eventfd. */
int vhost_add_used_n(struct vhost_virtqueue *vq, struct vring_used_elem *heads,
		     unsigned count)
{
	int start, n, r;

2417
	start = vq->last_used_idx & (vq->num - 1);
2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429
	n = vq->num - start;
	if (n < count) {
		r = __vhost_add_used_n(vq, heads, n);
		if (r < 0)
			return r;
		heads += n;
		count -= n;
	}
	r = __vhost_add_used_n(vq, heads, count);

	/* Make sure buffer is written before we update index. */
	smp_wmb();
2430
	if (vhost_put_used_idx(vq)) {
2431 2432 2433 2434
		vq_err(vq, "Failed to increment used idx");
		return -EFAULT;
	}
	if (unlikely(vq->log_used)) {
2435 2436
		/* Make sure used idx is seen before log. */
		smp_wmb();
2437
		/* Log used index update. */
J
Jason Wang 已提交
2438 2439
		log_used(vq, offsetof(struct vring_used, idx),
			 sizeof vq->used->idx);
2440 2441 2442 2443 2444
		if (vq->log_ctx)
			eventfd_signal(vq->log_ctx, 1);
	}
	return r;
}
A
Asias He 已提交
2445
EXPORT_SYMBOL_GPL(vhost_add_used_n);
2446

M
Michael S. Tsirkin 已提交
2447
static bool vhost_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
2448
{
2449 2450
	__u16 old, new;
	__virtio16 event;
M
Michael S. Tsirkin 已提交
2451
	bool v;
2452 2453 2454 2455
	/* Flush out used index updates. This is paired
	 * with the barrier that the Guest executes when enabling
	 * interrupts. */
	smp_mb();
M
Michael S. Tsirkin 已提交
2456

2457
	if (vhost_has_feature(vq, VIRTIO_F_NOTIFY_ON_EMPTY) &&
M
Michael S. Tsirkin 已提交
2458 2459 2460
	    unlikely(vq->avail_idx == vq->last_avail_idx))
		return true;

2461
	if (!vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX)) {
2462
		__virtio16 flags;
2463
		if (vhost_get_avail_flags(vq, &flags)) {
M
Michael S. Tsirkin 已提交
2464 2465 2466
			vq_err(vq, "Failed to get flags");
			return true;
		}
2467
		return !(flags & cpu_to_vhost16(vq, VRING_AVAIL_F_NO_INTERRUPT));
2468
	}
M
Michael S. Tsirkin 已提交
2469 2470 2471 2472
	old = vq->signalled_used;
	v = vq->signalled_used_valid;
	new = vq->signalled_used = vq->last_used_idx;
	vq->signalled_used_valid = true;
2473

M
Michael S. Tsirkin 已提交
2474 2475
	if (unlikely(!v))
		return true;
2476

2477
	if (vhost_get_used_event(vq, &event)) {
M
Michael S. Tsirkin 已提交
2478 2479 2480
		vq_err(vq, "Failed to get used event idx");
		return true;
	}
2481
	return vring_need_event(vhost16_to_cpu(vq, event), new, old);
M
Michael S. Tsirkin 已提交
2482 2483 2484 2485 2486
}

/* This actually signals the guest, using eventfd. */
void vhost_signal(struct vhost_dev *dev, struct vhost_virtqueue *vq)
{
2487
	/* Signal the Guest tell them we used something up. */
Z
Zhu Lingshan 已提交
2488 2489
	if (vq->call_ctx.ctx && vhost_notify(dev, vq))
		eventfd_signal(vq->call_ctx.ctx, 1);
2490
}
A
Asias He 已提交
2491
EXPORT_SYMBOL_GPL(vhost_signal);
2492 2493 2494 2495 2496 2497 2498 2499 2500

/* And here's the combo meal deal.  Supersize me! */
void vhost_add_used_and_signal(struct vhost_dev *dev,
			       struct vhost_virtqueue *vq,
			       unsigned int head, int len)
{
	vhost_add_used(vq, head, len);
	vhost_signal(dev, vq);
}
A
Asias He 已提交
2501
EXPORT_SYMBOL_GPL(vhost_add_used_and_signal);
2502

2503 2504 2505 2506 2507 2508 2509 2510
/* multi-buffer version of vhost_add_used_and_signal */
void vhost_add_used_and_signal_n(struct vhost_dev *dev,
				 struct vhost_virtqueue *vq,
				 struct vring_used_elem *heads, unsigned count)
{
	vhost_add_used_n(vq, heads, count);
	vhost_signal(dev, vq);
}
A
Asias He 已提交
2511
EXPORT_SYMBOL_GPL(vhost_add_used_and_signal_n);
2512

2513 2514 2515 2516 2517 2518
/* return true if we're sure that avaiable ring is empty */
bool vhost_vq_avail_empty(struct vhost_dev *dev, struct vhost_virtqueue *vq)
{
	__virtio16 avail_idx;
	int r;

2519 2520 2521
	if (vq->avail_idx != vq->last_avail_idx)
		return false;

2522
	r = vhost_get_avail_idx(vq, &avail_idx);
2523
	if (unlikely(r))
2524
		return false;
2525
	vq->avail_idx = vhost16_to_cpu(vq, avail_idx);
2526

2527
	return vq->avail_idx == vq->last_avail_idx;
2528 2529 2530
}
EXPORT_SYMBOL_GPL(vhost_vq_avail_empty);

2531
/* OK, now we need to know about added descriptors. */
M
Michael S. Tsirkin 已提交
2532
bool vhost_enable_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
2533
{
2534
	__virtio16 avail_idx;
2535
	int r;
K
Krishna Kumar 已提交
2536

2537 2538 2539
	if (!(vq->used_flags & VRING_USED_F_NO_NOTIFY))
		return false;
	vq->used_flags &= ~VRING_USED_F_NO_NOTIFY;
2540
	if (!vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX)) {
2541
		r = vhost_update_used_flags(vq);
M
Michael S. Tsirkin 已提交
2542 2543 2544 2545 2546 2547
		if (r) {
			vq_err(vq, "Failed to enable notification at %p: %d\n",
			       &vq->used->flags, r);
			return false;
		}
	} else {
2548
		r = vhost_update_avail_event(vq, vq->avail_idx);
M
Michael S. Tsirkin 已提交
2549 2550 2551 2552 2553 2554
		if (r) {
			vq_err(vq, "Failed to update avail event index at %p: %d\n",
			       vhost_avail_event(vq), r);
			return false;
		}
	}
2555 2556
	/* They could have slipped one in as we were doing that: make
	 * sure it's written, then check again. */
2557
	smp_mb();
2558
	r = vhost_get_avail_idx(vq, &avail_idx);
2559 2560 2561 2562 2563 2564
	if (r) {
		vq_err(vq, "Failed to check avail idx at %p: %d\n",
		       &vq->avail->idx, r);
		return false;
	}

2565
	return vhost16_to_cpu(vq, avail_idx) != vq->avail_idx;
2566
}
A
Asias He 已提交
2567
EXPORT_SYMBOL_GPL(vhost_enable_notify);
2568 2569

/* We don't need to be notified again. */
M
Michael S. Tsirkin 已提交
2570
void vhost_disable_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
2571 2572
{
	int r;
K
Krishna Kumar 已提交
2573

2574 2575 2576
	if (vq->used_flags & VRING_USED_F_NO_NOTIFY)
		return;
	vq->used_flags |= VRING_USED_F_NO_NOTIFY;
2577
	if (!vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX)) {
2578
		r = vhost_update_used_flags(vq);
M
Michael S. Tsirkin 已提交
2579
		if (r)
Y
Yunsheng Lin 已提交
2580
			vq_err(vq, "Failed to disable notification at %p: %d\n",
M
Michael S. Tsirkin 已提交
2581 2582
			       &vq->used->flags, r);
	}
2583
}
A
Asias He 已提交
2584 2585
EXPORT_SYMBOL_GPL(vhost_disable_notify);

J
Jason Wang 已提交
2586 2587 2588 2589 2590 2591
/* Create a new message. */
struct vhost_msg_node *vhost_new_msg(struct vhost_virtqueue *vq, int type)
{
	struct vhost_msg_node *node = kmalloc(sizeof *node, GFP_KERNEL);
	if (!node)
		return NULL;
2592 2593 2594

	/* Make sure all padding within the structure is initialized. */
	memset(&node->msg, 0, sizeof node->msg);
J
Jason Wang 已提交
2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607
	node->vq = vq;
	node->msg.type = type;
	return node;
}
EXPORT_SYMBOL_GPL(vhost_new_msg);

void vhost_enqueue_msg(struct vhost_dev *dev, struct list_head *head,
		       struct vhost_msg_node *node)
{
	spin_lock(&dev->iotlb_lock);
	list_add_tail(&node->node, head);
	spin_unlock(&dev->iotlb_lock);

2608
	wake_up_interruptible_poll(&dev->wait, EPOLLIN | EPOLLRDNORM);
J
Jason Wang 已提交
2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628
}
EXPORT_SYMBOL_GPL(vhost_enqueue_msg);

struct vhost_msg_node *vhost_dequeue_msg(struct vhost_dev *dev,
					 struct list_head *head)
{
	struct vhost_msg_node *node = NULL;

	spin_lock(&dev->iotlb_lock);
	if (!list_empty(head)) {
		node = list_first_entry(head, struct vhost_msg_node,
					node);
		list_del(&node->node);
	}
	spin_unlock(&dev->iotlb_lock);

	return node;
}
EXPORT_SYMBOL_GPL(vhost_dequeue_msg);

2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643
void vhost_set_backend_features(struct vhost_dev *dev, u64 features)
{
	struct vhost_virtqueue *vq;
	int i;

	mutex_lock(&dev->mutex);
	for (i = 0; i < dev->nvqs; ++i) {
		vq = dev->vqs[i];
		mutex_lock(&vq->mutex);
		vq->acked_backend_features = features;
		mutex_unlock(&vq->mutex);
	}
	mutex_unlock(&dev->mutex);
}
EXPORT_SYMBOL_GPL(vhost_set_backend_features);
J
Jason Wang 已提交
2644

A
Asias He 已提交
2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660
static int __init vhost_init(void)
{
	return 0;
}

static void __exit vhost_exit(void)
{
}

module_init(vhost_init);
module_exit(vhost_exit);

MODULE_VERSION("0.0.1");
MODULE_LICENSE("GPL v2");
MODULE_AUTHOR("Michael S. Tsirkin");
MODULE_DESCRIPTION("Host kernel accelerator for virtio");