vhost.c 43.6 KB
Newer Older
1 2 3 4 5 6
/* Copyright (C) 2009 Red Hat, Inc.
 * Copyright (C) 2006 Rusty Russell IBM Corporation
 *
 * Author: Michael S. Tsirkin <mst@redhat.com>
 *
 * Inspiration, some code, and most witty comments come from
Rob Landley's avatar
Rob Landley committed
7
 * Documentation/virtual/lguest/lguest.c, by Rusty Russell
8 9 10 11 12 13 14 15
 *
 * This work is licensed under the terms of the GNU GPL, version 2.
 *
 * Generic code for virtio server in host kernel.
 */

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

#include "vhost.h"

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

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

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

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

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

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

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

	if (vq->private_data)
		return -EBUSY;

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

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

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

	return 0;
}

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

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

	return 0;
}

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

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

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

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

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

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

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

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

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

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

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

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

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

169
void vhost_work_init(struct vhost_work *work, vhost_work_fn_t fn)
170 171 172 173 174
{
	INIT_LIST_HEAD(&work->node);
	work->fn = fn;
	init_waitqueue_head(&work->done);
}
175
EXPORT_SYMBOL_GPL(vhost_work_init);
176

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

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

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

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

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

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

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

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

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

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

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

247
void vhost_work_queue(struct vhost_dev *dev, struct vhost_work *work)
248
{
249 250 251 252 253
	unsigned long flags;

	spin_lock_irqsave(&dev->work_lock, flags);
	if (list_empty(&work->node)) {
		list_add_tail(&work->node, &dev->work_list);
254
		spin_unlock_irqrestore(&dev->work_lock, flags);
255
		wake_up_process(dev->worker);
256 257
	} else {
		spin_unlock_irqrestore(&dev->work_lock, flags);
258
	}
259
}
260
EXPORT_SYMBOL_GPL(vhost_work_queue);
261

262 263 264 265 266 267 268
/* A lockless hint for busy polling code to exit the loop */
bool vhost_has_work(struct vhost_dev *dev)
{
	return !list_empty(&dev->work_list);
}
EXPORT_SYMBOL_GPL(vhost_has_work);

269 270 271 272
void vhost_poll_queue(struct vhost_poll *poll)
{
	vhost_work_queue(poll->dev, &poll->work);
}
273
EXPORT_SYMBOL_GPL(vhost_poll_queue);
274

275 276 277 278 279 280 281 282 283 284
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;
285 286
	vq->signalled_used = 0;
	vq->signalled_used_valid = false;
287 288 289 290
	vq->used_flags = 0;
	vq->log_used = false;
	vq->log_addr = -1ull;
	vq->private_data = NULL;
291
	vq->acked_features = 0;
292 293 294 295 296 297
	vq->log_base = NULL;
	vq->error_ctx = NULL;
	vq->error = NULL;
	vq->kick = NULL;
	vq->call_ctx = NULL;
	vq->call = NULL;
298
	vq->log_ctx = NULL;
299
	vq->memory = NULL;
300 301
	vhost_reset_is_le(vq);
	vhost_disable_cross_endian(vq);
302
	vq->busyloop_timeout = 0;
303 304
}

305 306 307 308
static int vhost_worker(void *data)
{
	struct vhost_dev *dev = data;
	struct vhost_work *work = NULL;
309
	mm_segment_t oldfs = get_fs();
310

311
	set_fs(USER_DS);
312 313
	use_mm(dev->mm);

314 315 316 317 318 319 320 321 322
	for (;;) {
		/* mb paired w/ kthread_stop */
		set_current_state(TASK_INTERRUPTIBLE);

		spin_lock_irq(&dev->work_lock);

		if (kthread_should_stop()) {
			spin_unlock_irq(&dev->work_lock);
			__set_current_state(TASK_RUNNING);
323
			break;
324 325 326 327 328 329 330 331 332 333 334 335
		}
		if (!list_empty(&dev->work_list)) {
			work = list_first_entry(&dev->work_list,
						struct vhost_work, node);
			list_del_init(&work->node);
		} else
			work = NULL;
		spin_unlock_irq(&dev->work_lock);

		if (work) {
			__set_current_state(TASK_RUNNING);
			work->fn(work);
336 337
			if (need_resched())
				schedule();
338 339 340 341
		} else
			schedule();

	}
342
	unuse_mm(dev->mm);
343
	set_fs(oldfs);
344
	return 0;
345 346
}

347 348 349 350 351 352 353 354 355 356
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
357 358 359
/* Helper to allocate iovec buffers for all vqs. */
static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
{
360
	struct vhost_virtqueue *vq;
Jason Wang's avatar
Jason Wang committed
361
	int i;
362

Jason Wang's avatar
Jason Wang committed
363
	for (i = 0; i < dev->nvqs; ++i) {
364 365 366 367 368 369
		vq = dev->vqs[i];
		vq->indirect = kmalloc(sizeof *vq->indirect * UIO_MAXIOV,
				       GFP_KERNEL);
		vq->log = kmalloc(sizeof *vq->log * UIO_MAXIOV, GFP_KERNEL);
		vq->heads = kmalloc(sizeof *vq->heads * UIO_MAXIOV, GFP_KERNEL);
		if (!vq->indirect || !vq->log || !vq->heads)
Jason Wang's avatar
Jason Wang committed
370 371 372
			goto err_nomem;
	}
	return 0;
373

Jason Wang's avatar
Jason Wang committed
374
err_nomem:
375
	for (; i >= 0; --i)
376
		vhost_vq_free_iovecs(dev->vqs[i]);
Jason Wang's avatar
Jason Wang committed
377 378 379 380 381 382
	return -ENOMEM;
}

static void vhost_dev_free_iovecs(struct vhost_dev *dev)
{
	int i;
383

384
	for (i = 0; i < dev->nvqs; ++i)
385
		vhost_vq_free_iovecs(dev->vqs[i]);
Jason Wang's avatar
Jason Wang committed
386 387
}

Zhi Yong Wu's avatar
Zhi Yong Wu committed
388
void vhost_dev_init(struct vhost_dev *dev,
389
		    struct vhost_virtqueue **vqs, int nvqs)
390
{
391
	struct vhost_virtqueue *vq;
392
	int i;
393

394 395 396 397 398 399 400
	dev->vqs = vqs;
	dev->nvqs = nvqs;
	mutex_init(&dev->mutex);
	dev->log_ctx = NULL;
	dev->log_file = NULL;
	dev->memory = NULL;
	dev->mm = NULL;
401 402 403
	spin_lock_init(&dev->work_lock);
	INIT_LIST_HEAD(&dev->work_list);
	dev->worker = NULL;
404 405

	for (i = 0; i < dev->nvqs; ++i) {
406 407 408 409 410 411 412 413 414 415
		vq = dev->vqs[i];
		vq->log = NULL;
		vq->indirect = NULL;
		vq->heads = NULL;
		vq->dev = dev;
		mutex_init(&vq->mutex);
		vhost_vq_reset(dev, vq);
		if (vq->handle_kick)
			vhost_poll_init(&vq->poll, vq->handle_kick,
					POLLIN, dev);
416 417
	}
}
418
EXPORT_SYMBOL_GPL(vhost_dev_init);
419 420 421 422 423 424 425

/* 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;
}
426
EXPORT_SYMBOL_GPL(vhost_dev_check_owner);
427

428
struct vhost_attach_cgroups_struct {
429 430 431
	struct vhost_work work;
	struct task_struct *owner;
	int ret;
432 433 434 435
};

static void vhost_attach_cgroups_work(struct vhost_work *work)
{
436 437 438 439
	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);
440 441 442 443
}

static int vhost_attach_cgroups(struct vhost_dev *dev)
{
444 445 446 447 448 449 450
	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;
451 452
}

453 454 455 456 457
/* Caller should have device mutex */
bool vhost_dev_has_owner(struct vhost_dev *dev)
{
	return dev->mm;
}
458
EXPORT_SYMBOL_GPL(vhost_dev_has_owner);
459

460
/* Caller should have device mutex */
Asias He's avatar
Asias He committed
461
long vhost_dev_set_owner(struct vhost_dev *dev)
462
{
463 464
	struct task_struct *worker;
	int err;
465

466
	/* Is there an owner already? */
467
	if (vhost_dev_has_owner(dev)) {
468 469 470
		err = -EBUSY;
		goto err_mm;
	}
471

472 473
	/* No owner, become one */
	dev->mm = get_task_mm(current);
474 475 476 477 478 479 480
	worker = kthread_create(vhost_worker, dev, "vhost-%d", current->pid);
	if (IS_ERR(worker)) {
		err = PTR_ERR(worker);
		goto err_worker;
	}

	dev->worker = worker;
481 482 483
	wake_up_process(worker);	/* avoid contributing to loadavg */

	err = vhost_attach_cgroups(dev);
484 485
	if (err)
		goto err_cgroup;
486

Jason Wang's avatar
Jason Wang committed
487 488 489 490
	err = vhost_dev_alloc_iovecs(dev);
	if (err)
		goto err_cgroup;

491
	return 0;
492 493
err_cgroup:
	kthread_stop(worker);
494
	dev->worker = NULL;
495 496 497 498 499 500
err_worker:
	if (dev->mm)
		mmput(dev->mm);
	dev->mm = NULL;
err_mm:
	return err;
501
}
502
EXPORT_SYMBOL_GPL(vhost_dev_set_owner);
503

504
struct vhost_memory *vhost_dev_reset_owner_prepare(void)
505
{
506 507
	return kmalloc(offsetof(struct vhost_memory, regions), GFP_KERNEL);
}
508
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner_prepare);
509

510 511 512
/* Caller should have device mutex */
void vhost_dev_reset_owner(struct vhost_dev *dev, struct vhost_memory *memory)
{
513 514
	int i;

515
	vhost_dev_cleanup(dev, true);
516

517
	/* Restore memory to default empty mapping. */
518
	memory->nregions = 0;
519 520 521 522 523 524
	dev->memory = memory;
	/* We don't need VQ locks below since vhost_dev_cleanup makes sure
	 * VQs aren't running.
	 */
	for (i = 0; i < dev->nvqs; ++i)
		dev->vqs[i]->memory = memory;
525
}
526
EXPORT_SYMBOL_GPL(vhost_dev_reset_owner);
527

528
void vhost_dev_stop(struct vhost_dev *dev)
529 530
{
	int i;
531 532

	for (i = 0; i < dev->nvqs; ++i) {
533 534 535
		if (dev->vqs[i]->kick && dev->vqs[i]->handle_kick) {
			vhost_poll_stop(&dev->vqs[i]->poll);
			vhost_poll_flush(&dev->vqs[i]->poll);
536
		}
537 538
	}
}
539
EXPORT_SYMBOL_GPL(vhost_dev_stop);
540

541 542
/* Caller should have device mutex if and only if locked is set */
void vhost_dev_cleanup(struct vhost_dev *dev, bool locked)
543 544
{
	int i;
545

546
	for (i = 0; i < dev->nvqs; ++i) {
547 548 549 550 551 552 553 554 555 556 557
		if (dev->vqs[i]->error_ctx)
			eventfd_ctx_put(dev->vqs[i]->error_ctx);
		if (dev->vqs[i]->error)
			fput(dev->vqs[i]->error);
		if (dev->vqs[i]->kick)
			fput(dev->vqs[i]->kick);
		if (dev->vqs[i]->call_ctx)
			eventfd_ctx_put(dev->vqs[i]->call_ctx);
		if (dev->vqs[i]->call)
			fput(dev->vqs[i]->call);
		vhost_vq_reset(dev, dev->vqs[i]);
558
	}
Jason Wang's avatar
Jason Wang committed
559
	vhost_dev_free_iovecs(dev);
560 561 562 563 564 565 566
	if (dev->log_ctx)
		eventfd_ctx_put(dev->log_ctx);
	dev->log_ctx = NULL;
	if (dev->log_file)
		fput(dev->log_file);
	dev->log_file = NULL;
	/* No one will access memory at this point */
567
	kvfree(dev->memory);
568
	dev->memory = NULL;
569
	WARN_ON(!list_empty(&dev->work_list));
570 571 572 573
	if (dev->worker) {
		kthread_stop(dev->worker);
		dev->worker = NULL;
	}
574 575 576
	if (dev->mm)
		mmput(dev->mm);
	dev->mm = NULL;
577
}
578
EXPORT_SYMBOL_GPL(vhost_dev_cleanup);
579 580 581 582

static int log_access_ok(void __user *log_base, u64 addr, unsigned long sz)
{
	u64 a = addr / VHOST_PAGE_SIZE / 8;
583

584 585 586
	/* Make sure 64 bit math will not overflow. */
	if (a > ULONG_MAX - (unsigned long)log_base ||
	    a + (unsigned long)log_base > ULONG_MAX)
587
		return 0;
588 589 590 591 592 593 594 595 596 597

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

/* Caller should have vq mutex and device mutex. */
static int vq_memory_access_ok(void __user *log_base, struct vhost_memory *mem,
			       int log_all)
{
	int i;
598

Michael S. Tsirkin's avatar
Michael S. Tsirkin committed
599 600
	if (!mem)
		return 0;
601

602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623
	for (i = 0; i < mem->nregions; ++i) {
		struct vhost_memory_region *m = mem->regions + i;
		unsigned long a = m->userspace_addr;
		if (m->memory_size > ULONG_MAX)
			return 0;
		else if (!access_ok(VERIFY_WRITE, (void __user *)a,
				    m->memory_size))
			return 0;
		else if (log_all && !log_access_ok(log_base,
						   m->guest_phys_addr,
						   m->memory_size))
			return 0;
	}
	return 1;
}

/* Can we switch to this memory table? */
/* Caller should have device mutex but not vq mutex */
static int memory_access_ok(struct vhost_dev *d, struct vhost_memory *mem,
			    int log_all)
{
	int i;
624

625 626
	for (i = 0; i < d->nvqs; ++i) {
		int ok;
627 628
		bool log;

629
		mutex_lock(&d->vqs[i]->mutex);
630
		log = log_all || vhost_has_feature(d->vqs[i], VHOST_F_LOG_ALL);
631
		/* If ring is inactive, will check when it's enabled. */
632
		if (d->vqs[i]->private_data)
633
			ok = vq_memory_access_ok(d->vqs[i]->log_base, mem, log);
634 635
		else
			ok = 1;
636
		mutex_unlock(&d->vqs[i]->mutex);
637 638 639 640 641 642
		if (!ok)
			return 0;
	}
	return 1;
}

643
static int vq_access_ok(struct vhost_virtqueue *vq, unsigned int num,
644 645 646 647
			struct vring_desc __user *desc,
			struct vring_avail __user *avail,
			struct vring_used __user *used)
{
648
	size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
649 650
	return access_ok(VERIFY_READ, desc, num * sizeof *desc) &&
	       access_ok(VERIFY_READ, avail,
651
			 sizeof *avail + num * sizeof *avail->ring + s) &&
652
	       access_ok(VERIFY_WRITE, used,
653
			sizeof *used + num * sizeof *used->ring + s);
654 655 656 657 658 659
}

/* Can we log writes? */
/* Caller should have device mutex but not vq mutex */
int vhost_log_access_ok(struct vhost_dev *dev)
{
660
	return memory_access_ok(dev, dev->memory, 1);
661
}
662
EXPORT_SYMBOL_GPL(vhost_log_access_ok);
663 664 665

/* Verify access for write logging. */
/* Caller should have vq mutex and device mutex */
666
static int vq_log_access_ok(struct vhost_virtqueue *vq,
667
			    void __user *log_base)
668
{
669
	size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
Arnd Bergmann's avatar
Arnd Bergmann committed
670

671
	return vq_memory_access_ok(log_base, vq->memory,
672
				   vhost_has_feature(vq, VHOST_F_LOG_ALL)) &&
673 674
		(!vq->log_used || log_access_ok(log_base, vq->log_addr,
					sizeof *vq->used +
675
					vq->num * sizeof *vq->used->ring + s));
676 677 678 679 680 681
}

/* Can we start vq? */
/* Caller should have vq mutex and device mutex */
int vhost_vq_access_ok(struct vhost_virtqueue *vq)
{
682 683
	return vq_access_ok(vq, vq->num, vq->desc, vq->avail, vq->used) &&
		vq_log_access_ok(vq, vq->log_base);
684
}
685
EXPORT_SYMBOL_GPL(vhost_vq_access_ok);
686

687 688 689 690 691 692 693 694 695 696
static int vhost_memory_reg_sort_cmp(const void *p1, const void *p2)
{
	const struct vhost_memory_region *r1 = p1, *r2 = p2;
	if (r1->guest_phys_addr < r2->guest_phys_addr)
		return 1;
	if (r1->guest_phys_addr > r2->guest_phys_addr)
		return -1;
	return 0;
}

697 698 699 700
static void *vhost_kvzalloc(unsigned long size)
{
	void *n = kzalloc(size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);

701
	if (!n)
702 703 704 705
		n = vzalloc(size);
	return n;
}

706 707 708 709
static long vhost_set_memory(struct vhost_dev *d, struct vhost_memory __user *m)
{
	struct vhost_memory mem, *newmem, *oldmem;
	unsigned long size = offsetof(struct vhost_memory, regions);
710
	int i;
711

712 713
	if (copy_from_user(&mem, m, size))
		return -EFAULT;
714 715
	if (mem.padding)
		return -EOPNOTSUPP;
716
	if (mem.nregions > max_mem_regions)
717
		return -E2BIG;
718
	newmem = vhost_kvzalloc(size + mem.nregions * sizeof(*m->regions));
719 720 721 722
	if (!newmem)
		return -ENOMEM;

	memcpy(newmem, &mem, size);
723 724
	if (copy_from_user(newmem->regions, m->regions,
			   mem.nregions * sizeof *m->regions)) {
725
		kvfree(newmem);
726
		return -EFAULT;
727
	}
728 729
	sort(newmem->regions, newmem->nregions, sizeof(*newmem->regions),
		vhost_memory_reg_sort_cmp, NULL);
730

731
	if (!memory_access_ok(d, newmem, 0)) {
732
		kvfree(newmem);
733
		return -EFAULT;
734
	}
735 736
	oldmem = d->memory;
	d->memory = newmem;
737

738
	/* All memory accesses are done under some VQ mutex. */
739 740
	for (i = 0; i < d->nvqs; ++i) {
		mutex_lock(&d->vqs[i]->mutex);
741
		d->vqs[i]->memory = newmem;
742 743
		mutex_unlock(&d->vqs[i]->mutex);
	}
744
	kvfree(oldmem);
745 746 747
	return 0;
}

748
long vhost_vring_ioctl(struct vhost_dev *d, int ioctl, void __user *argp)
749
{
750 751
	struct file *eventfp, *filep = NULL;
	bool pollstart = false, pollstop = false;
752 753 754 755 756 757 758 759 760 761 762 763
	struct eventfd_ctx *ctx = NULL;
	u32 __user *idxp = argp;
	struct vhost_virtqueue *vq;
	struct vhost_vring_state s;
	struct vhost_vring_file f;
	struct vhost_vring_addr a;
	u32 idx;
	long r;

	r = get_user(idx, idxp);
	if (r < 0)
		return r;
764
	if (idx >= d->nvqs)
765 766
		return -ENOBUFS;

767
	vq = d->vqs[idx];
768 769 770 771 772 773 774 775 776 777 778

	mutex_lock(&vq->mutex);

	switch (ioctl) {
	case VHOST_SET_VRING_NUM:
		/* Resizing ring with an active backend?
		 * You don't want to do that. */
		if (vq->private_data) {
			r = -EBUSY;
			break;
		}
779 780
		if (copy_from_user(&s, argp, sizeof s)) {
			r = -EFAULT;
781
			break;
782
		}
783 784 785 786 787 788 789 790 791 792 793 794 795
		if (!s.num || s.num > 0xffff || (s.num & (s.num - 1))) {
			r = -EINVAL;
			break;
		}
		vq->num = s.num;
		break;
	case VHOST_SET_VRING_BASE:
		/* Moving base with an active backend?
		 * You don't want to do that. */
		if (vq->private_data) {
			r = -EBUSY;
			break;
		}
796 797
		if (copy_from_user(&s, argp, sizeof s)) {
			r = -EFAULT;
798
			break;
799
		}
800 801 802 803 804 805 806 807 808 809 810
		if (s.num > 0xffff) {
			r = -EINVAL;
			break;
		}
		vq->last_avail_idx = s.num;
		/* Forget the cached index value. */
		vq->avail_idx = vq->last_avail_idx;
		break;
	case VHOST_GET_VRING_BASE:
		s.index = idx;
		s.num = vq->last_avail_idx;
811 812
		if (copy_to_user(argp, &s, sizeof s))
			r = -EFAULT;
813 814
		break;
	case VHOST_SET_VRING_ADDR:
815 816
		if (copy_from_user(&a, argp, sizeof a)) {
			r = -EFAULT;
817
			break;
818
		}
819 820 821 822 823 824 825 826 827 828 829 830
		if (a.flags & ~(0x1 << VHOST_VRING_F_LOG)) {
			r = -EOPNOTSUPP;
			break;
		}
		/* For 32bit, verify that the top 32bits of the user
		   data are set to zero. */
		if ((u64)(unsigned long)a.desc_user_addr != a.desc_user_addr ||
		    (u64)(unsigned long)a.used_user_addr != a.used_user_addr ||
		    (u64)(unsigned long)a.avail_user_addr != a.avail_user_addr) {
			r = -EFAULT;
			break;
		}
831 832 833 834 835 836

		/* 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)) ||
837
		    (a.log_guest_addr & (VRING_USED_ALIGN_SIZE - 1))) {
838 839 840 841 842 843 844 845
			r = -EINVAL;
			break;
		}

		/* We only verify access here if backend is configured.
		 * If it is not, we don't as size might not have been setup.
		 * We will verify when backend is configured. */
		if (vq->private_data) {
846
			if (!vq_access_ok(vq, vq->num,
847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870
				(void __user *)(unsigned long)a.desc_user_addr,
				(void __user *)(unsigned long)a.avail_user_addr,
				(void __user *)(unsigned long)a.used_user_addr)) {
				r = -EINVAL;
				break;
			}

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

		vq->log_used = !!(a.flags & (0x1 << VHOST_VRING_F_LOG));
		vq->desc = (void __user *)(unsigned long)a.desc_user_addr;
		vq->avail = (void __user *)(unsigned long)a.avail_user_addr;
		vq->log_addr = a.log_guest_addr;
		vq->used = (void __user *)(unsigned long)a.used_user_addr;
		break;
	case VHOST_SET_VRING_KICK:
871 872
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
873
			break;
874
		}
875
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
876 877 878 879
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
880
		if (eventfp != vq->kick) {
881 882
			pollstop = (filep = vq->kick) != NULL;
			pollstart = (vq->kick = eventfp) != NULL;
883 884 885 886
		} else
			filep = eventfp;
		break;
	case VHOST_SET_VRING_CALL:
887 888
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
889
			break;
890
		}
891
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
892 893 894 895
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
896 897 898 899 900 901 902 903 904 905
		if (eventfp != vq->call) {
			filep = vq->call;
			ctx = vq->call_ctx;
			vq->call = eventfp;
			vq->call_ctx = eventfp ?
				eventfd_ctx_fileget(eventfp) : NULL;
		} else
			filep = eventfp;
		break;
	case VHOST_SET_VRING_ERR:
906 907
		if (copy_from_user(&f, argp, sizeof f)) {
			r = -EFAULT;
908
			break;
909
		}
910
		eventfp = f.fd == -1 ? NULL : eventfd_fget(f.fd);
911 912 913 914
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
915 916 917 918 919 920 921 922 923
		if (eventfp != vq->error) {
			filep = vq->error;
			vq->error = eventfp;
			ctx = vq->error_ctx;
			vq->error_ctx = eventfp ?
				eventfd_ctx_fileget(eventfp) : NULL;
		} else
			filep = eventfp;
		break;
924 925 926 927 928 929
	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;
930 931 932 933 934 935 936 937 938 939 940 941 942
	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;
943 944 945 946 947 948 949 950 951 952 953 954 955
	default:
		r = -ENOIOCTLCMD;
	}

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

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

	if (pollstart && vq->handle_kick)
956
		r = vhost_poll_start(&vq->poll, vq->kick);
957 958 959 960 961 962 963

	mutex_unlock(&vq->mutex);

	if (pollstop && vq->handle_kick)
		vhost_poll_flush(&vq->poll);
	return r;
}
964
EXPORT_SYMBOL_GPL(vhost_vring_ioctl);
965 966

/* Caller must have device mutex */
967
long vhost_dev_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990
{
	struct file *eventfp, *filep = NULL;
	struct eventfd_ctx *ctx = NULL;
	u64 p;
	long r;
	int i, fd;

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

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

	switch (ioctl) {
	case VHOST_SET_MEM_TABLE:
		r = vhost_set_memory(d, argp);
		break;
	case VHOST_SET_LOG_BASE:
991 992
		if (copy_from_user(&p, argp, sizeof p)) {
			r = -EFAULT;
993
			break;
994
		}
995 996 997 998 999 1000 1001
		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;
1002
			vq = d->vqs[i];
1003 1004
			mutex_lock(&vq->mutex);
			/* If ring is inactive, will check when it's enabled. */
1005
			if (vq->private_data && !vq_log_access_ok(vq, base))
1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
				r = -EFAULT;
			else
				vq->log_base = base;
			mutex_unlock(&vq->mutex);
		}
		break;
	case VHOST_SET_LOG_FD:
		r = get_user(fd, (int __user *)argp);
		if (r < 0)
			break;
		eventfp = fd == -1 ? NULL : eventfd_fget(fd);
		if (IS_ERR(eventfp)) {
			r = PTR_ERR(eventfp);
			break;
		}
		if (eventfp != d->log_file) {
			filep = d->log_file;
1023
			d->log_file = eventfp;
1024 1025 1026 1027 1028 1029
			ctx = d->log_ctx;
			d->log_ctx = eventfp ?
				eventfd_ctx_fileget(eventfp) : NULL;
		} else
			filep = eventfp;
		for (i = 0; i < d->nvqs; ++i) {
1030 1031 1032
			mutex_lock(&d->vqs[i]->mutex);
			d->vqs[i]->log_ctx = d->log_ctx;
			mutex_unlock(&d->vqs[i]->mutex);
1033 1034 1035 1036 1037 1038 1039
		}
		if (ctx)
			eventfd_ctx_put(ctx);
		if (filep)
			fput(filep);
		break;
	default:
1040
		r = -ENOIOCTLCMD;
1041 1042 1043 1044 1045
		break;
	}
done:
	return r;
}
1046
EXPORT_SYMBOL_GPL(vhost_dev_ioctl);
1047 1048 1049 1050

static const struct vhost_memory_region *find_region(struct vhost_memory *mem,
						     __u64 addr, __u32 len)
{
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
	const struct vhost_memory_region *reg;
	int start = 0, end = mem->nregions;

	while (start < end) {
		int slot = start + (end - start) / 2;
		reg = mem->regions + slot;
		if (addr >= reg->guest_phys_addr)
			end = slot;
		else
			start = slot + 1;
1061
	}
1062 1063 1064 1065 1066

	reg = mem->regions + start;
	if (addr >= reg->guest_phys_addr &&
		reg->guest_phys_addr + reg->memory_size > addr)
		return reg;
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
	return NULL;
}

/* TODO: This is really inefficient.  We need something like get_user()
 * (instruction directly accesses the data, with an exception table entry
 * returning -EFAULT). See Documentation/x86/exception-tables.txt.
 */
static int set_bit_to_user(int nr, void __user *addr)
{
	unsigned long log = (unsigned long)addr;
	struct page *page;
	void *base;
	int bit = nr + (log % PAGE_SIZE) * 8;
	int r;
1081

1082
	r = get_user_pages_fast(log, 1, 1, &page);
1083
	if (r < 0)
1084
		return r;
1085
	BUG_ON(r != 1);
1086
	base = kmap_atomic(page);
1087
	set_bit(bit, base);
1088
	kunmap_atomic(base);
1089 1090 1091 1092 1093 1094 1095 1096
	set_page_dirty_lock(page);
	put_page(page);
	return 0;
}

static int log_write(void __user *log_base,
		     u64 write_address, u64 write_length)
{
1097
	u64 write_page = write_address / VHOST_PAGE_SIZE;
1098
	int r;
1099

1100 1101
	if (!write_length)
		return 0;
1102
	write_length += write_address % VHOST_PAGE_SIZE;
1103 1104
	for (;;) {
		u64 base = (u64)(unsigned long)log_base;
1105 1106
		u64 log = base + write_page / 8;
		int bit = write_page % 8;
1107 1108 1109 1110 1111 1112 1113 1114
		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;
1115
		write_page += 1;
1116 1117 1118 1119 1120 1121 1122 1123 1124 1125
	}
	return r;
}

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

	/* Make sure data written is seen before log. */
1126
	smp_wmb();
1127 1128 1129 1130 1131 1132
	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;
1133 1134 1135
		if (!len) {
			if (vq->log_ctx)
				eventfd_signal(vq->log_ctx, 1);
1136
			return 0;
1137
		}
1138 1139 1140 1141 1142
	}
	/* Length written exceeds what we have stored. This is a bug. */
	BUG();
	return 0;
}
1143
EXPORT_SYMBOL_GPL(vhost_log_write);
1144

1145 1146 1147
static int vhost_update_used_flags(struct vhost_virtqueue *vq)
{
	void __user *used;
1148
	if (__put_user(cpu_to_vhost16(vq, vq->used_flags), &vq->used->flags) < 0)
1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165
		return -EFAULT;
	if (unlikely(vq->log_used)) {
		/* Make sure the flag is seen before log. */
		smp_wmb();
		/* Log used flag write. */
		used = &vq->used->flags;
		log_write(vq->log_base, vq->log_addr +
			  (used - (void __user *)vq->used),
			  sizeof vq->used->flags);
		if (vq->log_ctx)
			eventfd_signal(vq->log_ctx, 1);
	}
	return 0;
}

static int vhost_update_avail_event(struct vhost_virtqueue *vq, u16 avail_event)
{
1166
	if (__put_user(cpu_to_vhost16(vq, vq->avail_idx), vhost_avail_event(vq)))
1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182
		return -EFAULT;
	if (unlikely(vq->log_used)) {
		void __user *used;
		/* Make sure the event is seen before log. */
		smp_wmb();
		/* Log avail event write */
		used = vhost_avail_event(vq);
		log_write(vq->log_base, vq->log_addr +
			  (used - (void __user *)vq->used),
			  sizeof *vhost_avail_event(vq));
		if (vq->log_ctx)
			eventfd_signal(vq->log_ctx, 1);
	}
	return 0;
}

Greg Kurz's avatar
Greg Kurz committed
1183
int vhost_vq_init_access(struct vhost_virtqueue *vq)
1184
{
1185
	__virtio16 last_used_idx;
1186
	int r;
1187 1188
	bool is_le = vq->is_le;

1189
	if (!vq->private_data) {
1190
		vhost_reset_is_le(vq);
1191
		return 0;
1192 1193 1194
	}

	vhost_init_is_le(vq);
1195 1196 1197

	r = vhost_update_used_flags(vq);
	if (r)
1198
		goto err;
1199
	vq->signalled_used_valid = false;
1200 1201 1202 1203
	if (!access_ok(VERIFY_READ, &vq->used->idx, sizeof vq->used->idx)) {
		r = -EFAULT;
		goto err;
	}
1204 1205
	r = __get_user(last_used_idx, &vq->used->idx);
	if (r)
1206
		goto err;
1207
	vq->last_used_idx = vhost16_to_cpu(vq, last_used_idx);
1208
	return 0;