drivers/vhost/vhost.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /* Copyright (C) 2009 Red Hat, Inc.
   3  * Copyright (C) 2006 Rusty Russell IBM Corporation
   4  *
   5  * Author: Michael S. Tsirkin <mst@redhat.com>
   6  *
   7  * Inspiration, some code, and most witty comments come from
   8  * Documentation/virtual/lguest/lguest.c, by Rusty Russell
   9  *
  10  * Generic code for virtio server in host kernel.
  11  */
  12
  13 #include <linux/eventfd.h>
  14 #include <linux/vhost.h>
  15 #include <linux/uio.h>
  16 #include <linux/mm.h>
  17 #include <linux/miscdevice.h>
  18 #include <linux/mutex.h>
  19 #include <linux/poll.h>
  20 #include <linux/file.h>
  21 #include <linux/highmem.h>
  22 #include <linux/slab.h>
  23 #include <linux/vmalloc.h>
  24 #include <linux/kthread.h>
  25 #include <linux/cgroup.h>
  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>
  31 #include <linux/nospec.h>
  32 #include <linux/kcov.h>
  33
  34 #include "vhost.h"
  35
  36 static ushort max_mem_regions = 64;
  37 module_param(max_mem_regions, ushort, 0444);
  38 MODULE_PARM_DESC(max_mem_regions,
  39         "Maximum number of memory regions in memory map. (default: 64)");
  40 static int max_iotlb_entries = 2048;
  41 module_param(max_iotlb_entries, int, 0444);
  42 MODULE_PARM_DESC(max_iotlb_entries,
  43         "Maximum number of iotlb entries. (default: 2048)");
  44
  45 enum {
  46         VHOST_MEMORY_F_LOG = 0x1,
  47 };
  48
  49 #define vhost_used_event(vq) ((__virtio16 __user *)&vq->avail->ring[vq->num])
  50 #define vhost_avail_event(vq) ((__virtio16 __user *)&vq->used->ring[vq->num])
  51
  52 #ifdef CONFIG_VHOST_CROSS_ENDIAN_LEGACY
  53 static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
  54 {
  55         vq->user_be = !virtio_legacy_is_little_endian();
  56 }
  57
  58 static void vhost_enable_cross_endian_big(struct vhost_virtqueue *vq)
  59 {
  60         vq->user_be = true;
  61 }
  62
  63 static void vhost_enable_cross_endian_little(struct vhost_virtqueue *vq)
  64 {
  65         vq->user_be = false;
  66 }
  67
  68 static long vhost_set_vring_endian(struct vhost_virtqueue *vq, int __user *argp)
  69 {
  70         struct vhost_vring_state s;
  71
  72         if (vq->private_data)
  73                 return -EBUSY;
  74
  75         if (copy_from_user(&s, argp, sizeof(s)))
  76                 return -EFAULT;
  77
  78         if (s.num != VHOST_VRING_LITTLE_ENDIAN &&
  79             s.num != VHOST_VRING_BIG_ENDIAN)
  80                 return -EINVAL;
  81
  82         if (s.num == VHOST_VRING_BIG_ENDIAN)
  83                 vhost_enable_cross_endian_big(vq);
  84         else
  85                 vhost_enable_cross_endian_little(vq);
  86
  87         return 0;
  88 }
  89
  90 static long vhost_get_vring_endian(struct vhost_virtqueue *vq, u32 idx,
  91                                    int __user *argp)
  92 {
  93         struct vhost_vring_state s = {
  94                 .index = idx,
  95                 .num = vq->user_be
  96         };
  97
  98         if (copy_to_user(argp, &s, sizeof(s)))
  99                 return -EFAULT;
 100
 101         return 0;
 102 }
 103
 104 static void vhost_init_is_le(struct vhost_virtqueue *vq)
 105 {
 106         /* Note for legacy virtio: user_be is initialized at reset time
 107          * according to the host endianness. If userspace does not set an
 108          * explicit endianness, the default behavior is native endian, as
 109          * expected by legacy virtio.
 110          */
 111         vq->is_le = vhost_has_feature(vq, VIRTIO_F_VERSION_1) || !vq->user_be;
 112 }
 113 #else
 114 static void vhost_disable_cross_endian(struct vhost_virtqueue *vq)
 115 {
 116 }
 117
 118 static long vhost_set_vring_endian(struct vhost_virtqueue *vq, int __user *argp)
 119 {
 120         return -ENOIOCTLCMD;
 121 }
 122
 123 static long vhost_get_vring_endian(struct vhost_virtqueue *vq, u32 idx,
 124                                    int __user *argp)
 125 {
 126         return -ENOIOCTLCMD;
 127 }
 128
 129 static void vhost_init_is_le(struct vhost_virtqueue *vq)
 130 {
 131         vq->is_le = vhost_has_feature(vq, VIRTIO_F_VERSION_1)
 132                 || virtio_legacy_is_little_endian();
 133 }
 134 #endif /* CONFIG_VHOST_CROSS_ENDIAN_LEGACY */
 135
 136 static void vhost_reset_is_le(struct vhost_virtqueue *vq)
 137 {
 138         vhost_init_is_le(vq);
 139 }
 140
 141 struct vhost_flush_struct {
 142         struct vhost_work work;
 143         struct completion wait_event;
 144 };
 145
 146 static void vhost_flush_work(struct vhost_work *work)
 147 {
 148         struct vhost_flush_struct *s;
 149
 150         s = container_of(work, struct vhost_flush_struct, work);
 151         complete(&s->wait_event);
 152 }
 153
 154 static void vhost_poll_func(struct file *file, wait_queue_head_t *wqh,
 155                             poll_table *pt)
 156 {
 157         struct vhost_poll *poll;
 158
 159         poll = container_of(pt, struct vhost_poll, table);
 160         poll->wqh = wqh;
 161         add_wait_queue(wqh, &poll->wait);
 162 }
 163
 164 static int vhost_poll_wakeup(wait_queue_entry_t *wait, unsigned mode, int sync,
 165                              void *key)
 166 {
 167         struct vhost_poll *poll = container_of(wait, struct vhost_poll, wait);
 168         struct vhost_work *work = &poll->work;
 169
 170         if (!(key_to_poll(key) & poll->mask))
 171                 return 0;
 172
 173         if (!poll->dev->use_worker)
 174                 work->fn(work);
 175         else
 176                 vhost_poll_queue(poll);
 177
 178         return 0;
 179 }
 180
 181 void vhost_work_init(struct vhost_work *work, vhost_work_fn_t fn)
 182 {
 183         clear_bit(VHOST_WORK_QUEUED, &work->flags);
 184         work->fn = fn;
 185 }
 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,
 190                      __poll_t mask, struct vhost_dev *dev)
 191 {
 192         init_waitqueue_func_entry(&poll->wait, vhost_poll_wakeup);
 193         init_poll_funcptr(&poll->table, vhost_poll_func);
 194         poll->mask = mask;
 195         poll->dev = dev;
 196         poll->wqh = NULL;
 197
 198         vhost_work_init(&poll->work, fn);
 199 }
 200 EXPORT_SYMBOL_GPL(vhost_poll_init);
 201
 202 /* Start polling a file. We add ourselves to file's wait queue. The caller must
 203  * 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;
 207
 208         if (poll->wqh)
 209                 return 0;
 210
 211         mask = vfs_poll(file, &poll->table);
 212         if (mask)
 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 }
 221 EXPORT_SYMBOL_GPL(vhost_poll_start);
 222
 223 /* Stop polling a file. After this function returns, it becomes safe to drop the
 224  * file reference. You must also flush afterwards. */
 225 void vhost_poll_stop(struct vhost_poll *poll)
 226 {
 227         if (poll->wqh) {
 228                 remove_wait_queue(poll->wqh, &poll->wait);
 229                 poll->wqh = NULL;
 230         }
 231 }
 232 EXPORT_SYMBOL_GPL(vhost_poll_stop);
 233
 234 void vhost_dev_flush(struct vhost_dev *dev)
 235 {
 236         struct vhost_flush_struct flush;
 237
 238         if (dev->worker) {
 239                 init_completion(&flush.wait_event);
 240                 vhost_work_init(&flush.work, vhost_flush_work);
 241
 242                 vhost_work_queue(dev, &flush.work);
 243                 wait_for_completion(&flush.wait_event);
 244         }
 245 }
 246 EXPORT_SYMBOL_GPL(vhost_dev_flush);
 247
 248 void vhost_work_queue(struct vhost_dev *dev, struct vhost_work *work)
 249 {
 250         if (!dev->worker)
 251                 return;
 252
 253         if (!test_and_set_bit(VHOST_WORK_QUEUED, &work->flags)) {
 254                 /* We can only add the work to the list after we're
 255                  * sure it was not in the list.
 256                  * test_and_set_bit() implies a memory barrier.
 257                  */
 258                 llist_add(&work->node, &dev->work_list);
 259                 wake_up_process(dev->worker);
 260         }
 261 }
 262 EXPORT_SYMBOL_GPL(vhost_work_queue);
 263
 264 /* A lockless hint for busy polling code to exit the loop */
 265 bool vhost_has_work(struct vhost_dev *dev)
 266 {
 267         return !llist_empty(&dev->work_list);
 268 }
 269 EXPORT_SYMBOL_GPL(vhost_has_work);
 270
 271 void vhost_poll_queue(struct vhost_poll *poll)
 272 {
 273         vhost_work_queue(poll->dev, &poll->work);
 274 }
 275 EXPORT_SYMBOL_GPL(vhost_poll_queue);
 276
 277 static void __vhost_vq_meta_reset(struct vhost_virtqueue *vq)
 278 {
 279         int j;
 280
 281         for (j = 0; j < VHOST_NUM_ADDRS; j++)
 282                 vq->meta_iotlb[j] = NULL;
 283 }
 284
 285 static void vhost_vq_meta_reset(struct vhost_dev *d)
 286 {
 287         int i;
 288
 289         for (i = 0; i < d->nvqs; ++i)
 290                 __vhost_vq_meta_reset(d->vqs[i]);
 291 }
 292
 293 static void vhost_vring_call_reset(struct vhost_vring_call *call_ctx)
 294 {
 295         call_ctx->ctx = NULL;
 296         memset(&call_ctx->producer, 0x0, sizeof(struct irq_bypass_producer));
 297 }
 298
 299 bool vhost_vq_is_setup(struct vhost_virtqueue *vq)
 300 {
 301         return vq->avail && vq->desc && vq->used && vhost_vq_access_ok(vq);
 302 }
 303 EXPORT_SYMBOL_GPL(vhost_vq_is_setup);
 304
 305 static void vhost_vq_reset(struct vhost_dev *dev,
 306                            struct vhost_virtqueue *vq)
 307 {
 308         vq->num = 1;
 309         vq->desc = NULL;
 310         vq->avail = NULL;
 311         vq->used = NULL;
 312         vq->last_avail_idx = 0;
 313         vq->avail_idx = 0;
 314         vq->last_used_idx = 0;
 315         vq->signalled_used = 0;
 316         vq->signalled_used_valid = false;
 317         vq->used_flags = 0;
 318         vq->log_used = false;
 319         vq->log_addr = -1ull;
 320         vq->private_data = NULL;
 321         vq->acked_features = 0;
 322         vq->acked_backend_features = 0;
 323         vq->log_base = NULL;
 324         vq->error_ctx = NULL;
 325         vq->kick = NULL;
 326         vq->log_ctx = NULL;
 327         vhost_disable_cross_endian(vq);
 328         vhost_reset_is_le(vq);
 329         vq->busyloop_timeout = 0;
 330         vq->umem = NULL;
 331         vq->iotlb = NULL;
 332         vhost_vring_call_reset(&vq->call_ctx);
 333         __vhost_vq_meta_reset(vq);
 334 }
 335
 336 static int vhost_worker(void *data)
 337 {
 338         struct vhost_dev *dev = data;
 339         struct vhost_work *work, *work_next;
 340         struct llist_node *node;
 341
 342         kthread_use_mm(dev->mm);
 343
 344         for (;;) {
 345                 /* mb paired w/ kthread_stop */
 346                 set_current_state(TASK_INTERRUPTIBLE);
 347
 348                 if (kthread_should_stop()) {
 349                         __set_current_state(TASK_RUNNING);
 350                         break;
 351                 }
 352
 353                 node = llist_del_all(&dev->work_list);
 354                 if (!node)
 355                         schedule();
 356
 357                 node = llist_reverse_order(node);
 358                 /* make sure flag is seen after deletion */
 359                 smp_wmb();
 360                 llist_for_each_entry_safe(work, work_next, node, node) {
 361                         clear_bit(VHOST_WORK_QUEUED, &work->flags);
 362                         __set_current_state(TASK_RUNNING);
 363                         kcov_remote_start_common(dev->kcov_handle);
 364                         work->fn(work);
 365                         kcov_remote_stop();
 366                         if (need_resched())
 367                                 schedule();
 368                 }
 369         }
 370         kthread_unuse_mm(dev->mm);
 371         return 0;
 372 }
 373
 374 static void vhost_vq_free_iovecs(struct vhost_virtqueue *vq)
 375 {
 376         kfree(vq->indirect);
 377         vq->indirect = NULL;
 378         kfree(vq->log);
 379         vq->log = NULL;
 380         kfree(vq->heads);
 381         vq->heads = NULL;
 382 }
 383
 384 /* Helper to allocate iovec buffers for all vqs. */
 385 static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
 386 {
 387         struct vhost_virtqueue *vq;
 388         int i;
 389
 390         for (i = 0; i < dev->nvqs; ++i) {
 391                 vq = dev->vqs[i];
 392                 vq->indirect = kmalloc_array(UIO_MAXIOV,
 393                                              sizeof(*vq->indirect),
 394                                              GFP_KERNEL);
 395                 vq->log = kmalloc_array(dev->iov_limit, sizeof(*vq->log),
 396                                         GFP_KERNEL);
 397                 vq->heads = kmalloc_array(dev->iov_limit, sizeof(*vq->heads),
 398                                           GFP_KERNEL);
 399                 if (!vq->indirect || !vq->log || !vq->heads)
 400                         goto err_nomem;
 401         }
 402         return 0;
 403
 404 err_nomem:
 405         for (; i >= 0; --i)
 406                 vhost_vq_free_iovecs(dev->vqs[i]);
 407         return -ENOMEM;
 408 }
 409
 410 static void vhost_dev_free_iovecs(struct vhost_dev *dev)
 411 {
 412         int i;
 413
 414         for (i = 0; i < dev->nvqs; ++i)
 415                 vhost_vq_free_iovecs(dev->vqs[i]);
 416 }
 417
 418 bool vhost_exceeds_weight(struct vhost_virtqueue *vq,
 419                           int pkts, int total_len)
 420 {
 421         struct vhost_dev *dev = vq->dev;
 422
 423         if ((dev->byte_weight && total_len >= dev->byte_weight) ||
 424             pkts >= dev->weight) {
 425                 vhost_poll_queue(&vq->poll);
 426                 return true;
 427         }
 428
 429         return false;
 430 }
 431 EXPORT_SYMBOL_GPL(vhost_exceeds_weight);
 432
 433 static size_t vhost_get_avail_size(struct vhost_virtqueue *vq,
 434                                    unsigned int num)
 435 {
 436         size_t event __maybe_unused =
 437                vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
 438
 439         return sizeof(*vq->avail) +
 440                sizeof(*vq->avail->ring) * num + event;
 441 }
 442
 443 static size_t vhost_get_used_size(struct vhost_virtqueue *vq,
 444                                   unsigned int num)
 445 {
 446         size_t event __maybe_unused =
 447                vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
 448
 449         return sizeof(*vq->used) +
 450                sizeof(*vq->used->ring) * num + event;
 451 }
 452
 453 static size_t vhost_get_desc_size(struct vhost_virtqueue *vq,
 454                                   unsigned int num)
 455 {
 456         return sizeof(*vq->desc) * num;
 457 }
 458
 459 void vhost_dev_init(struct vhost_dev *dev,
 460                     struct vhost_virtqueue **vqs, int nvqs,
 461                     int iov_limit, int weight, int byte_weight,
 462                     bool use_worker,
 463                     int (*msg_handler)(struct vhost_dev *dev, u32 asid,
 464                                        struct vhost_iotlb_msg *msg))
 465 {
 466         struct vhost_virtqueue *vq;
 467         int i;
 468
 469         dev->vqs = vqs;
 470         dev->nvqs = nvqs;
 471         mutex_init(&dev->mutex);
 472         dev->log_ctx = NULL;
 473         dev->umem = NULL;
 474         dev->iotlb = NULL;
 475         dev->mm = NULL;
 476         dev->worker = NULL;
 477         dev->iov_limit = iov_limit;
 478         dev->weight = weight;
 479         dev->byte_weight = byte_weight;
 480         dev->use_worker = use_worker;
 481         dev->msg_handler = msg_handler;
 482         init_llist_head(&dev->work_list);
 483         init_waitqueue_head(&dev->wait);
 484         INIT_LIST_HEAD(&dev->read_list);
 485         INIT_LIST_HEAD(&dev->pending_list);
 486         spin_lock_init(&dev->iotlb_lock);
 487
 488
 489         for (i = 0; i < dev->nvqs; ++i) {
 490                 vq = dev->vqs[i];
 491                 vq->log = NULL;
 492                 vq->indirect = NULL;
 493                 vq->heads = NULL;
 494                 vq->dev = dev;
 495                 mutex_init(&vq->mutex);
 496                 vhost_vq_reset(dev, vq);
 497                 if (vq->handle_kick)
 498                         vhost_poll_init(&vq->poll, vq->handle_kick,
 499                                         EPOLLIN, dev);
 500         }
 501 }
 502 EXPORT_SYMBOL_GPL(vhost_dev_init);
 503
 504 /* Caller should have device mutex */
 505 long vhost_dev_check_owner(struct vhost_dev *dev)
 506 {
 507         /* Are you the owner? If not, I don't think you mean to do that */
 508         return dev->mm == current->mm ? 0 : -EPERM;
 509 }
 510 EXPORT_SYMBOL_GPL(vhost_dev_check_owner);
 511
 512 struct vhost_attach_cgroups_struct {
 513         struct vhost_work work;
 514         struct task_struct *owner;
 515         int ret;
 516 };
 517
 518 static void vhost_attach_cgroups_work(struct vhost_work *work)
 519 {
 520         struct vhost_attach_cgroups_struct *s;
 521
 522         s = container_of(work, struct vhost_attach_cgroups_struct, work);
 523         s->ret = cgroup_attach_task_all(s->owner, current);
 524 }
 525
 526 static int vhost_attach_cgroups(struct vhost_dev *dev)
 527 {
 528         struct vhost_attach_cgroups_struct attach;
 529
 530         attach.owner = current;
 531         vhost_work_init(&attach.work, vhost_attach_cgroups_work);
 532         vhost_work_queue(dev, &attach.work);
 533         vhost_dev_flush(dev);
 534         return attach.ret;
 535 }
 536
 537 /* Caller should have device mutex */
 538 bool vhost_dev_has_owner(struct vhost_dev *dev)
 539 {
 540         return dev->mm;
 541 }
 542 EXPORT_SYMBOL_GPL(vhost_dev_has_owner);
 543
 544 static void vhost_attach_mm(struct vhost_dev *dev)
 545 {
 546         /* No owner, become one */
 547         if (dev->use_worker) {
 548                 dev->mm = get_task_mm(current);
 549         } else {
 550                 /* vDPA device does not use worker thead, so there's
 551                  * no need to hold the address space for mm. This help
 552                  * to avoid deadlock in the case of mmap() which may
 553                  * held the refcnt of the file and depends on release
 554                  * method to remove vma.
 555                  */
 556                 dev->mm = current->mm;
 557                 mmgrab(dev->mm);
 558         }
 559 }
 560
 561 static void vhost_detach_mm(struct vhost_dev *dev)
 562 {
 563         if (!dev->mm)
 564                 return;
 565
 566         if (dev->use_worker)
 567                 mmput(dev->mm);
 568         else
 569                 mmdrop(dev->mm);
 570
 571         dev->mm = NULL;
 572 }
 573
 574 /* Caller should have device mutex */
 575 long vhost_dev_set_owner(struct vhost_dev *dev)
 576 {
 577         struct task_struct *worker;
 578         int err;
 579
 580         /* Is there an owner already? */
 581         if (vhost_dev_has_owner(dev)) {
 582                 err = -EBUSY;
 583                 goto err_mm;
 584         }
 585
 586         vhost_attach_mm(dev);
 587
 588         dev->kcov_handle = kcov_common_handle();
 589         if (dev->use_worker) {
 590                 worker = kthread_create(vhost_worker, dev,
 591                                         "vhost-%d", current->pid);
 592                 if (IS_ERR(worker)) {
 593                         err = PTR_ERR(worker);
 594                         goto err_worker;
 595                 }
 596
 597                 dev->worker = worker;
 598                 wake_up_process(worker); /* avoid contributing to loadavg */
 599
 600                 err = vhost_attach_cgroups(dev);
 601                 if (err)
 602                         goto err_cgroup;
 603         }
 604
 605         err = vhost_dev_alloc_iovecs(dev);
 606         if (err)
 607                 goto err_cgroup;
 608
 609         return 0;
 610 err_cgroup:
 611         if (dev->worker) {
 612                 kthread_stop(dev->worker);
 613                 dev->worker = NULL;
 614         }
 615 err_worker:
 616         vhost_detach_mm(dev);
 617         dev->kcov_handle = 0;
 618 err_mm:
 619         return err;
 620 }
 621 EXPORT_SYMBOL_GPL(vhost_dev_set_owner);
 622
 623 static struct vhost_iotlb *iotlb_alloc(void)
 624 {
 625         return vhost_iotlb_alloc(max_iotlb_entries,
 626                                  VHOST_IOTLB_FLAG_RETIRE);
 627 }
 628
 629 struct vhost_iotlb *vhost_dev_reset_owner_prepare(void)
 630 {
 631         return iotlb_alloc();
 632 }
 633 EXPORT_SYMBOL_GPL(vhost_dev_reset_owner_prepare);
 634
 635 /* Caller should have device mutex */
 636 void vhost_dev_reset_owner(struct vhost_dev *dev, struct vhost_iotlb *umem)
 637 {
 638         int i;
 639
 640         vhost_dev_cleanup(dev);
 641
 642         dev->umem = umem;
 643         /* We don't need VQ locks below since vhost_dev_cleanup makes sure
 644          * VQs aren't running.
 645          */
 646         for (i = 0; i < dev->nvqs; ++i)
 647                 dev->vqs[i]->umem = umem;
 648 }
 649 EXPORT_SYMBOL_GPL(vhost_dev_reset_owner);
 650
 651 void vhost_dev_stop(struct vhost_dev *dev)
 652 {
 653         int i;
 654
 655         for (i = 0; i < dev->nvqs; ++i) {
 656                 if (dev->vqs[i]->kick && dev->vqs[i]->handle_kick)
 657                         vhost_poll_stop(&dev->vqs[i]->poll);
 658         }
 659
 660         vhost_dev_flush(dev);
 661 }
 662 EXPORT_SYMBOL_GPL(vhost_dev_stop);
 663
 664 static void vhost_clear_msg(struct vhost_dev *dev)
 665 {
 666         struct vhost_msg_node *node, *n;
 667
 668         spin_lock(&dev->iotlb_lock);
 669
 670         list_for_each_entry_safe(node, n, &dev->read_list, node) {
 671                 list_del(&node->node);
 672                 kfree(node);
 673         }
 674
 675         list_for_each_entry_safe(node, n, &dev->pending_list, node) {
 676                 list_del(&node->node);
 677                 kfree(node);
 678         }
 679
 680         spin_unlock(&dev->iotlb_lock);
 681 }
 682
 683 void vhost_dev_cleanup(struct vhost_dev *dev)
 684 {
 685         int i;
 686
 687         for (i = 0; i < dev->nvqs; ++i) {
 688                 if (dev->vqs[i]->error_ctx)
 689                         eventfd_ctx_put(dev->vqs[i]->error_ctx);
 690                 if (dev->vqs[i]->kick)
 691                         fput(dev->vqs[i]->kick);
 692                 if (dev->vqs[i]->call_ctx.ctx)
 693                         eventfd_ctx_put(dev->vqs[i]->call_ctx.ctx);
 694                 vhost_vq_reset(dev, dev->vqs[i]);
 695         }
 696         vhost_dev_free_iovecs(dev);
 697         if (dev->log_ctx)
 698                 eventfd_ctx_put(dev->log_ctx);
 699         dev->log_ctx = NULL;
 700         /* No one will access memory at this point */
 701         vhost_iotlb_free(dev->umem);
 702         dev->umem = NULL;
 703         vhost_iotlb_free(dev->iotlb);
 704         dev->iotlb = NULL;
 705         vhost_clear_msg(dev);
 706         wake_up_interruptible_poll(&dev->wait, EPOLLIN | EPOLLRDNORM);
 707         WARN_ON(!llist_empty(&dev->work_list));
 708         if (dev->worker) {
 709                 kthread_stop(dev->worker);
 710                 dev->worker = NULL;
 711                 dev->kcov_handle = 0;
 712         }
 713         vhost_detach_mm(dev);
 714 }
 715 EXPORT_SYMBOL_GPL(vhost_dev_cleanup);
 716
 717 static bool log_access_ok(void __user *log_base, u64 addr, unsigned long sz)
 718 {
 719         u64 a = addr / VHOST_PAGE_SIZE / 8;
 720
 721         /* Make sure 64 bit math will not overflow. */
 722         if (a > ULONG_MAX - (unsigned long)log_base ||
 723             a + (unsigned long)log_base > ULONG_MAX)
 724                 return false;
 725
 726         return access_ok(log_base + a,
 727                          (sz + VHOST_PAGE_SIZE * 8 - 1) / VHOST_PAGE_SIZE / 8);
 728 }
 729
 730 /* Make sure 64 bit math will not overflow. */
 731 static bool vhost_overflow(u64 uaddr, u64 size)
 732 {
 733         if (uaddr > ULONG_MAX || size > ULONG_MAX)
 734                 return true;
 735
 736         if (!size)
 737                 return false;
 738
 739         return uaddr > ULONG_MAX - size + 1;
 740 }
 741
 742 /* Caller should have vq mutex and device mutex. */
 743 static bool vq_memory_access_ok(void __user *log_base, struct vhost_iotlb *umem,
 744                                 int log_all)
 745 {
 746         struct vhost_iotlb_map *map;
 747
 748         if (!umem)
 749                 return false;
 750
 751         list_for_each_entry(map, &umem->list, link) {
 752                 unsigned long a = map->addr;
 753
 754                 if (vhost_overflow(map->addr, map->size))
 755                         return false;
 756
 757
 758                 if (!access_ok((void __user *)a, map->size))
 759                         return false;
 760                 else if (log_all && !log_access_ok(log_base,
 761                                                    map->start,
 762                                                    map->size))
 763                         return false;
 764         }
 765         return true;
 766 }
 767
 768 static inline void __user *vhost_vq_meta_fetch(struct vhost_virtqueue *vq,
 769                                                u64 addr, unsigned int size,
 770                                                int type)
 771 {
 772         const struct vhost_iotlb_map *map = vq->meta_iotlb[type];
 773
 774         if (!map)
 775                 return NULL;
 776
 777         return (void __user *)(uintptr_t)(map->addr + addr - map->start);
 778 }
 779
 780 /* Can we switch to this memory table? */
 781 /* Caller should have device mutex but not vq mutex */
 782 static bool memory_access_ok(struct vhost_dev *d, struct vhost_iotlb *umem,
 783                              int log_all)
 784 {
 785         int i;
 786
 787         for (i = 0; i < d->nvqs; ++i) {
 788                 bool ok;
 789                 bool log;
 790
 791                 mutex_lock(&d->vqs[i]->mutex);
 792                 log = log_all || vhost_has_feature(d->vqs[i], VHOST_F_LOG_ALL);
 793                 /* If ring is inactive, will check when it's enabled. */
 794                 if (d->vqs[i]->private_data)
 795                         ok = vq_memory_access_ok(d->vqs[i]->log_base,
 796                                                  umem, log);
 797                 else
 798                         ok = true;
 799                 mutex_unlock(&d->vqs[i]->mutex);
 800                 if (!ok)
 801                         return false;
 802         }
 803         return true;
 804 }
 805
 806 static int translate_desc(struct vhost_virtqueue *vq, u64 addr, u32 len,
 807                           struct iovec iov[], int iov_size, int access);
 808
 809 static int vhost_copy_to_user(struct vhost_virtqueue *vq, void __user *to,
 810                               const void *from, unsigned size)
 811 {
 812         int ret;
 813
 814         if (!vq->iotlb)
 815                 return __copy_to_user(to, from, size);
 816         else {
 817                 /* This function should be called after iotlb
 818                  * prefetch, which means we're sure that all vq
 819                  * could be access through iotlb. So -EAGAIN should
 820                  * not happen in this case.
 821                  */
 822                 struct iov_iter t;
 823                 void __user *uaddr = vhost_vq_meta_fetch(vq,
 824                                      (u64)(uintptr_t)to, size,
 825                                      VHOST_ADDR_USED);
 826
 827                 if (uaddr)
 828                         return __copy_to_user(uaddr, from, size);
 829
 830                 ret = translate_desc(vq, (u64)(uintptr_t)to, size, vq->iotlb_iov,
 831                                      ARRAY_SIZE(vq->iotlb_iov),
 832                                      VHOST_ACCESS_WO);
 833                 if (ret < 0)
 834                         goto out;
 835                 iov_iter_init(&t, ITER_DEST, vq->iotlb_iov, ret, size);
 836                 ret = copy_to_iter(from, size, &t);
 837                 if (ret == size)
 838                         ret = 0;
 839         }
 840 out:
 841         return ret;
 842 }
 843
 844 static int vhost_copy_from_user(struct vhost_virtqueue *vq, void *to,
 845                                 void __user *from, unsigned size)
 846 {
 847         int ret;
 848
 849         if (!vq->iotlb)
 850                 return __copy_from_user(to, from, size);
 851         else {
 852                 /* This function should be called after iotlb
 853                  * prefetch, which means we're sure that vq
 854                  * could be access through iotlb. So -EAGAIN should
 855                  * not happen in this case.
 856                  */
 857                 void __user *uaddr = vhost_vq_meta_fetch(vq,
 858                                      (u64)(uintptr_t)from, size,
 859                                      VHOST_ADDR_DESC);
 860                 struct iov_iter f;
 861
 862                 if (uaddr)
 863                         return __copy_from_user(to, uaddr, size);
 864
 865                 ret = translate_desc(vq, (u64)(uintptr_t)from, size, vq->iotlb_iov,
 866                                      ARRAY_SIZE(vq->iotlb_iov),
 867                                      VHOST_ACCESS_RO);
 868                 if (ret < 0) {
 869                         vq_err(vq, "IOTLB translation failure: uaddr "
 870                                "%p size 0x%llx\n", from,
 871                                (unsigned long long) size);
 872                         goto out;
 873                 }
 874                 iov_iter_init(&f, ITER_SOURCE, vq->iotlb_iov, ret, size);
 875                 ret = copy_from_iter(to, size, &f);
 876                 if (ret == size)
 877                         ret = 0;
 878         }
 879
 880 out:
 881         return ret;
 882 }
 883
 884 static void __user *__vhost_get_user_slow(struct vhost_virtqueue *vq,
 885                                           void __user *addr, unsigned int size,
 886                                           int type)
 887 {
 888         int ret;
 889
 890         ret = translate_desc(vq, (u64)(uintptr_t)addr, size, vq->iotlb_iov,
 891                              ARRAY_SIZE(vq->iotlb_iov),
 892                              VHOST_ACCESS_RO);
 893         if (ret < 0) {
 894                 vq_err(vq, "IOTLB translation failure: uaddr "
 895                         "%p size 0x%llx\n", addr,
 896                         (unsigned long long) size);
 897                 return NULL;
 898         }
 899
 900         if (ret != 1 || vq->iotlb_iov[0].iov_len != size) {
 901                 vq_err(vq, "Non atomic userspace memory access: uaddr "
 902                         "%p size 0x%llx\n", addr,
 903                         (unsigned long long) size);
 904                 return NULL;
 905         }
 906
 907         return vq->iotlb_iov[0].iov_base;
 908 }
 909
 910 /* This function should be called after iotlb
 911  * prefetch, which means we're sure that vq
 912  * could be access through iotlb. So -EAGAIN should
 913  * not happen in this case.
 914  */
 915 static inline void __user *__vhost_get_user(struct vhost_virtqueue *vq,
 916                                             void __user *addr, unsigned int size,
 917                                             int type)
 918 {
 919         void __user *uaddr = vhost_vq_meta_fetch(vq,
 920                              (u64)(uintptr_t)addr, size, type);
 921         if (uaddr)
 922                 return uaddr;
 923
 924         return __vhost_get_user_slow(vq, addr, size, type);
 925 }
 926
 927 #define vhost_put_user(vq, x, ptr)              \
 928 ({ \
 929         int ret; \
 930         if (!vq->iotlb) { \
 931                 ret = __put_user(x, ptr); \
 932         } else { \
 933                 __typeof__(ptr) to = \
 934                         (__typeof__(ptr)) __vhost_get_user(vq, ptr,     \
 935                                           sizeof(*ptr), VHOST_ADDR_USED); \
 936                 if (to != NULL) \
 937                         ret = __put_user(x, to); \
 938                 else \
 939                         ret = -EFAULT;  \
 940         } \
 941         ret; \
 942 })
 943
 944 static inline int vhost_put_avail_event(struct vhost_virtqueue *vq)
 945 {
 946         return vhost_put_user(vq, cpu_to_vhost16(vq, vq->avail_idx),
 947                               vhost_avail_event(vq));
 948 }
 949
 950 static inline int vhost_put_used(struct vhost_virtqueue *vq,
 951                                  struct vring_used_elem *head, int idx,
 952                                  int count)
 953 {
 954         return vhost_copy_to_user(vq, vq->used->ring + idx, head,
 955                                   count * sizeof(*head));
 956 }
 957
 958 static inline int vhost_put_used_flags(struct vhost_virtqueue *vq)
 959
 960 {
 961         return vhost_put_user(vq, cpu_to_vhost16(vq, vq->used_flags),
 962                               &vq->used->flags);
 963 }
 964
 965 static inline int vhost_put_used_idx(struct vhost_virtqueue *vq)
 966
 967 {
 968         return vhost_put_user(vq, cpu_to_vhost16(vq, vq->last_used_idx),
 969                               &vq->used->idx);
 970 }
 971
 972 #define vhost_get_user(vq, x, ptr, type)                \
 973 ({ \
 974         int ret; \
 975         if (!vq->iotlb) { \
 976                 ret = __get_user(x, ptr); \
 977         } else { \
 978                 __typeof__(ptr) from = \
 979                         (__typeof__(ptr)) __vhost_get_user(vq, ptr, \
 980                                                            sizeof(*ptr), \
 981                                                            type); \
 982                 if (from != NULL) \
 983                         ret = __get_user(x, from); \
 984                 else \
 985                         ret = -EFAULT; \
 986         } \
 987         ret; \
 988 })
 989
 990 #define vhost_get_avail(vq, x, ptr) \
 991         vhost_get_user(vq, x, ptr, VHOST_ADDR_AVAIL)
 992
 993 #define vhost_get_used(vq, x, ptr) \
 994         vhost_get_user(vq, x, ptr, VHOST_ADDR_USED)
 995
 996 static void vhost_dev_lock_vqs(struct vhost_dev *d)
 997 {
 998         int i = 0;
 999         for (i = 0; i < d->nvqs; ++i)
1000                 mutex_lock_nested(&d->vqs[i]->mutex, i);
1001 }
1002
1003 static void vhost_dev_unlock_vqs(struct vhost_dev *d)
1004 {
1005         int i = 0;
1006         for (i = 0; i < d->nvqs; ++i)
1007                 mutex_unlock(&d->vqs[i]->mutex);
1008 }
1009
1010 static inline int vhost_get_avail_idx(struct vhost_virtqueue *vq,
1011                                       __virtio16 *idx)
1012 {
1013         return vhost_get_avail(vq, *idx, &vq->avail->idx);
1014 }
1015
1016 static inline int vhost_get_avail_head(struct vhost_virtqueue *vq,
1017                                        __virtio16 *head, int idx)
1018 {
1019         return vhost_get_avail(vq, *head,
1020                                &vq->avail->ring[idx & (vq->num - 1)]);
1021 }
1022
1023 static inline int vhost_get_avail_flags(struct vhost_virtqueue *vq,
1024                                         __virtio16 *flags)
1025 {
1026         return vhost_get_avail(vq, *flags, &vq->avail->flags);
1027 }
1028
1029 static inline int vhost_get_used_event(struct vhost_virtqueue *vq,
1030                                        __virtio16 *event)
1031 {
1032         return vhost_get_avail(vq, *event, vhost_used_event(vq));
1033 }
1034
1035 static inline int vhost_get_used_idx(struct vhost_virtqueue *vq,
1036                                      __virtio16 *idx)
1037 {
1038         return vhost_get_used(vq, *idx, &vq->used->idx);
1039 }
1040
1041 static inline int vhost_get_desc(struct vhost_virtqueue *vq,
1042                                  struct vring_desc *desc, int idx)
1043 {
1044         return vhost_copy_from_user(vq, desc, vq->desc + idx, sizeof(*desc));
1045 }
1046
1047 static void vhost_iotlb_notify_vq(struct vhost_dev *d,
1048                                   struct vhost_iotlb_msg *msg)
1049 {
1050         struct vhost_msg_node *node, *n;
1051
1052         spin_lock(&d->iotlb_lock);
1053
1054         list_for_each_entry_safe(node, n, &d->pending_list, node) {
1055                 struct vhost_iotlb_msg *vq_msg = &node->msg.iotlb;
1056                 if (msg->iova <= vq_msg->iova &&
1057                     msg->iova + msg->size - 1 >= vq_msg->iova &&
1058                     vq_msg->type == VHOST_IOTLB_MISS) {
1059                         vhost_poll_queue(&node->vq->poll);
1060                         list_del(&node->node);
1061                         kfree(node);
1062                 }
1063         }
1064
1065         spin_unlock(&d->iotlb_lock);
1066 }
1067
1068 static bool umem_access_ok(u64 uaddr, u64 size, int access)
1069 {
1070         unsigned long a = uaddr;
1071
1072         /* Make sure 64 bit math will not overflow. */
1073         if (vhost_overflow(uaddr, size))
1074                 return false;
1075
1076         if ((access & VHOST_ACCESS_RO) &&
1077             !access_ok((void __user *)a, size))
1078                 return false;
1079         if ((access & VHOST_ACCESS_WO) &&
1080             !access_ok((void __user *)a, size))
1081                 return false;
1082         return true;
1083 }
1084
1085 static int vhost_process_iotlb_msg(struct vhost_dev *dev, u32 asid,
1086                                    struct vhost_iotlb_msg *msg)
1087 {
1088         int ret = 0;
1089
1090         if (asid != 0)
1091                 return -EINVAL;
1092
1093         mutex_lock(&dev->mutex);
1094         vhost_dev_lock_vqs(dev);
1095         switch (msg->type) {
1096         case VHOST_IOTLB_UPDATE:
1097                 if (!dev->iotlb) {
1098                         ret = -EFAULT;
1099                         break;
1100                 }
1101                 if (!umem_access_ok(msg->uaddr, msg->size, msg->perm)) {
1102                         ret = -EFAULT;
1103                         break;
1104                 }
1105                 vhost_vq_meta_reset(dev);
1106                 if (vhost_iotlb_add_range(dev->iotlb, msg->iova,
1107                                           msg->iova + msg->size - 1,
1108                                           msg->uaddr, msg->perm)) {
1109                         ret = -ENOMEM;
1110                         break;
1111                 }
1112                 vhost_iotlb_notify_vq(dev, msg);
1113                 break;
1114         case VHOST_IOTLB_INVALIDATE:
1115                 if (!dev->iotlb) {
1116                         ret = -EFAULT;
1117                         break;
1118                 }
1119                 vhost_vq_meta_reset(dev);
1120                 vhost_iotlb_del_range(dev->iotlb, msg->iova,
1121                                       msg->iova + msg->size - 1);
1122                 break;
1123         default:
1124                 ret = -EINVAL;
1125                 break;
1126         }
1127
1128         vhost_dev_unlock_vqs(dev);
1129         mutex_unlock(&dev->mutex);
1130
1131         return ret;
1132 }
1133 ssize_t vhost_chr_write_iter(struct vhost_dev *dev,
1134                              struct iov_iter *from)
1135 {
1136         struct vhost_iotlb_msg msg;
1137         size_t offset;
1138         int type, ret;
1139         u32 asid = 0;
1140
1141         ret = copy_from_iter(&type, sizeof(type), from);
1142         if (ret != sizeof(type)) {
1143                 ret = -EINVAL;
1144                 goto done;
1145         }
1146
1147         switch (type) {
1148         case VHOST_IOTLB_MSG:
1149                 /* There maybe a hole after type for V1 message type,
1150                  * so skip it here.
1151                  */
1152                 offset = offsetof(struct vhost_msg, iotlb) - sizeof(int);
1153                 break;
1154         case VHOST_IOTLB_MSG_V2:
1155                 if (vhost_backend_has_feature(dev->vqs[0],
1156                                               VHOST_BACKEND_F_IOTLB_ASID)) {
1157                         ret = copy_from_iter(&asid, sizeof(asid), from);
1158                         if (ret != sizeof(asid)) {
1159                                 ret = -EINVAL;
1160                                 goto done;
1161                         }
1162                         offset = 0;
1163                 } else
1164                         offset = sizeof(__u32);
1165                 break;
1166         default:
1167                 ret = -EINVAL;
1168                 goto done;
1169         }
1170
1171         iov_iter_advance(from, offset);
1172         ret = copy_from_iter(&msg, sizeof(msg), from);
1173         if (ret != sizeof(msg)) {
1174                 ret = -EINVAL;
1175                 goto done;
1176         }
1177
1178         if ((msg.type == VHOST_IOTLB_UPDATE ||
1179              msg.type == VHOST_IOTLB_INVALIDATE) &&
1180              msg.size == 0) {
1181                 ret = -EINVAL;
1182                 goto done;
1183         }
1184
1185         if (dev->msg_handler)
1186                 ret = dev->msg_handler(dev, asid, &msg);
1187         else
1188                 ret = vhost_process_iotlb_msg(dev, asid, &msg);
1189         if (ret) {
1190                 ret = -EFAULT;
1191                 goto done;
1192         }
1193
1194         ret = (type == VHOST_IOTLB_MSG) ? sizeof(struct vhost_msg) :
1195               sizeof(struct vhost_msg_v2);
1196 done:
1197         return ret;
1198 }
1199 EXPORT_SYMBOL(vhost_chr_write_iter);
1200
1201 __poll_t vhost_chr_poll(struct file *file, struct vhost_dev *dev,
1202                             poll_table *wait)
1203 {
1204         __poll_t mask = 0;
1205
1206         poll_wait(file, &dev->wait, wait);
1207
1208         if (!list_empty(&dev->read_list))
1209                 mask |= EPOLLIN | EPOLLRDNORM;
1210
1211         return mask;
1212 }
1213 EXPORT_SYMBOL(vhost_chr_poll);
1214
1215 ssize_t vhost_chr_read_iter(struct vhost_dev *dev, struct iov_iter *to,
1216                             int noblock)
1217 {
1218         DEFINE_WAIT(wait);
1219         struct vhost_msg_node *node;
1220         ssize_t ret = 0;
1221         unsigned size = sizeof(struct vhost_msg);
1222
1223         if (iov_iter_count(to) < size)
1224                 return 0;
1225
1226         while (1) {
1227                 if (!noblock)
1228                         prepare_to_wait(&dev->wait, &wait,
1229                                         TASK_INTERRUPTIBLE);
1230
1231                 node = vhost_dequeue_msg(dev, &dev->read_list);
1232                 if (node)
1233                         break;
1234                 if (noblock) {
1235                         ret = -EAGAIN;
1236                         break;
1237                 }
1238                 if (signal_pending(current)) {
1239                         ret = -ERESTARTSYS;
1240                         break;
1241                 }
1242                 if (!dev->iotlb) {
1243                         ret = -EBADFD;
1244                         break;
1245                 }
1246
1247                 schedule();
1248         }
1249
1250         if (!noblock)
1251                 finish_wait(&dev->wait, &wait);
1252
1253         if (node) {
1254                 struct vhost_iotlb_msg *msg;
1255                 void *start = &node->msg;
1256
1257                 switch (node->msg.type) {
1258                 case VHOST_IOTLB_MSG:
1259                         size = sizeof(node->msg);
1260                         msg = &node->msg.iotlb;
1261                         break;
1262                 case VHOST_IOTLB_MSG_V2:
1263                         size = sizeof(node->msg_v2);
1264                         msg = &node->msg_v2.iotlb;
1265                         break;
1266                 default:
1267                         BUG();
1268                         break;
1269                 }
1270
1271                 ret = copy_to_iter(start, size, to);
1272                 if (ret != size || msg->type != VHOST_IOTLB_MISS) {
1273                         kfree(node);
1274                         return ret;
1275                 }
1276                 vhost_enqueue_msg(dev, &dev->pending_list, node);
1277         }
1278
1279         return ret;
1280 }
1281 EXPORT_SYMBOL_GPL(vhost_chr_read_iter);
1282
1283 static int vhost_iotlb_miss(struct vhost_virtqueue *vq, u64 iova, int access)
1284 {
1285         struct vhost_dev *dev = vq->dev;
1286         struct vhost_msg_node *node;
1287         struct vhost_iotlb_msg *msg;
1288         bool v2 = vhost_backend_has_feature(vq, VHOST_BACKEND_F_IOTLB_MSG_V2);
1289
1290         node = vhost_new_msg(vq, v2 ? VHOST_IOTLB_MSG_V2 : VHOST_IOTLB_MSG);
1291         if (!node)
1292                 return -ENOMEM;
1293
1294         if (v2) {
1295                 node->msg_v2.type = VHOST_IOTLB_MSG_V2;
1296                 msg = &node->msg_v2.iotlb;
1297         } else {
1298                 msg = &node->msg.iotlb;
1299         }
1300
1301         msg->type = VHOST_IOTLB_MISS;
1302         msg->iova = iova;
1303         msg->perm = access;
1304
1305         vhost_enqueue_msg(dev, &dev->read_list, node);
1306
1307         return 0;
1308 }
1309
1310 static bool vq_access_ok(struct vhost_virtqueue *vq, unsigned int num,
1311                          vring_desc_t __user *desc,
1312                          vring_avail_t __user *avail,
1313                          vring_used_t __user *used)
1314
1315 {
1316         /* If an IOTLB device is present, the vring addresses are
1317          * GIOVAs. Access validation occurs at prefetch time. */
1318         if (vq->iotlb)
1319                 return true;
1320
1321         return access_ok(desc, vhost_get_desc_size(vq, num)) &&
1322                access_ok(avail, vhost_get_avail_size(vq, num)) &&
1323                access_ok(used, vhost_get_used_size(vq, num));
1324 }
1325
1326 static void vhost_vq_meta_update(struct vhost_virtqueue *vq,
1327                                  const struct vhost_iotlb_map *map,
1328                                  int type)
1329 {
1330         int access = (type == VHOST_ADDR_USED) ?
1331                      VHOST_ACCESS_WO : VHOST_ACCESS_RO;
1332
1333         if (likely(map->perm & access))
1334                 vq->meta_iotlb[type] = map;
1335 }
1336
1337 static bool iotlb_access_ok(struct vhost_virtqueue *vq,
1338                             int access, u64 addr, u64 len, int type)
1339 {
1340         const struct vhost_iotlb_map *map;
1341         struct vhost_iotlb *umem = vq->iotlb;
1342         u64 s = 0, size, orig_addr = addr, last = addr + len - 1;
1343
1344         if (vhost_vq_meta_fetch(vq, addr, len, type))
1345                 return true;
1346
1347         while (len > s) {
1348                 map = vhost_iotlb_itree_first(umem, addr, last);
1349                 if (map == NULL || map->start > addr) {
1350                         vhost_iotlb_miss(vq, addr, access);
1351                         return false;
1352                 } else if (!(map->perm & access)) {
1353                         /* Report the possible access violation by
1354                          * request another translation from userspace.
1355                          */
1356                         return false;
1357                 }
1358
1359                 size = map->size - addr + map->start;
1360
1361                 if (orig_addr == addr && size >= len)
1362                         vhost_vq_meta_update(vq, map, type);
1363
1364                 s += size;
1365                 addr += size;
1366         }
1367
1368         return true;
1369 }
1370
1371 int vq_meta_prefetch(struct vhost_virtqueue *vq)
1372 {
1373         unsigned int num = vq->num;
1374
1375         if (!vq->iotlb)
1376                 return 1;
1377
1378         return iotlb_access_ok(vq, VHOST_MAP_RO, (u64)(uintptr_t)vq->desc,
1379                                vhost_get_desc_size(vq, num), VHOST_ADDR_DESC) &&
1380                iotlb_access_ok(vq, VHOST_MAP_RO, (u64)(uintptr_t)vq->avail,
1381                                vhost_get_avail_size(vq, num),
1382                                VHOST_ADDR_AVAIL) &&
1383                iotlb_access_ok(vq, VHOST_MAP_WO, (u64)(uintptr_t)vq->used,
1384                                vhost_get_used_size(vq, num), VHOST_ADDR_USED);
1385 }
1386 EXPORT_SYMBOL_GPL(vq_meta_prefetch);
1387
1388 /* Can we log writes? */
1389 /* Caller should have device mutex but not vq mutex */
1390 bool vhost_log_access_ok(struct vhost_dev *dev)
1391 {
1392         return memory_access_ok(dev, dev->umem, 1);
1393 }
1394 EXPORT_SYMBOL_GPL(vhost_log_access_ok);
1395
1396 static bool vq_log_used_access_ok(struct vhost_virtqueue *vq,
1397                                   void __user *log_base,
1398                                   bool log_used,
1399                                   u64 log_addr)
1400 {
1401         /* If an IOTLB device is present, log_addr is a GIOVA that
1402          * will never be logged by log_used(). */
1403         if (vq->iotlb)
1404                 return true;
1405
1406         return !log_used || log_access_ok(log_base, log_addr,
1407                                           vhost_get_used_size(vq, vq->num));
1408 }
1409
1410 /* Verify access for write logging. */
1411 /* Caller should have vq mutex and device mutex */
1412 static bool vq_log_access_ok(struct vhost_virtqueue *vq,
1413                              void __user *log_base)
1414 {
1415         return vq_memory_access_ok(log_base, vq->umem,
1416                                    vhost_has_feature(vq, VHOST_F_LOG_ALL)) &&
1417                 vq_log_used_access_ok(vq, log_base, vq->log_used, vq->log_addr);
1418 }
1419
1420 /* Can we start vq? */
1421 /* Caller should have vq mutex and device mutex */
1422 bool vhost_vq_access_ok(struct vhost_virtqueue *vq)
1423 {
1424         if (!vq_log_access_ok(vq, vq->log_base))
1425                 return false;
1426
1427         return vq_access_ok(vq, vq->num, vq->desc, vq->avail, vq->used);
1428 }
1429 EXPORT_SYMBOL_GPL(vhost_vq_access_ok);
1430
1431 static long vhost_set_memory(struct vhost_dev *d, struct vhost_memory __user *m)
1432 {
1433         struct vhost_memory mem, *newmem;
1434         struct vhost_memory_region *region;
1435         struct vhost_iotlb *newumem, *oldumem;
1436         unsigned long size = offsetof(struct vhost_memory, regions);
1437         int i;
1438
1439         if (copy_from_user(&mem, m, size))
1440                 return -EFAULT;
1441         if (mem.padding)
1442                 return -EOPNOTSUPP;
1443         if (mem.nregions > max_mem_regions)
1444                 return -E2BIG;
1445         newmem = kvzalloc(struct_size(newmem, regions, mem.nregions),
1446                         GFP_KERNEL);
1447         if (!newmem)
1448                 return -ENOMEM;
1449
1450         memcpy(newmem, &mem, size);
1451         if (copy_from_user(newmem->regions, m->regions,
1452                            flex_array_size(newmem, regions, mem.nregions))) {
1453                 kvfree(newmem);
1454                 return -EFAULT;
1455         }
1456
1457         newumem = iotlb_alloc();
1458         if (!newumem) {
1459                 kvfree(newmem);
1460                 return -ENOMEM;
1461         }
1462
1463         for (region = newmem->regions;
1464              region < newmem->regions + mem.nregions;
1465              region++) {
1466                 if (vhost_iotlb_add_range(newumem,
1467                                           region->guest_phys_addr,
1468                                           region->guest_phys_addr +
1469                                           region->memory_size - 1,
1470                                           region->userspace_addr,
1471                                           VHOST_MAP_RW))
1472                         goto err;
1473         }
1474
1475         if (!memory_access_ok(d, newumem, 0))
1476                 goto err;
1477
1478         oldumem = d->umem;
1479         d->umem = newumem;
1480
1481         /* All memory accesses are done under some VQ mutex. */
1482         for (i = 0; i < d->nvqs; ++i) {
1483                 mutex_lock(&d->vqs[i]->mutex);
1484                 d->vqs[i]->umem = newumem;
1485                 mutex_unlock(&d->vqs[i]->mutex);
1486         }
1487
1488         kvfree(newmem);
1489         vhost_iotlb_free(oldumem);
1490         return 0;
1491
1492 err:
1493         vhost_iotlb_free(newumem);
1494         kvfree(newmem);
1495         return -EFAULT;
1496 }
1497
1498 static long vhost_vring_set_num(struct vhost_dev *d,
1499                                 struct vhost_virtqueue *vq,
1500                                 void __user *argp)
1501 {
1502         struct vhost_vring_state s;
1503
1504         /* Resizing ring with an active backend?
1505          * You don't want to do that. */
1506         if (vq->private_data)
1507                 return -EBUSY;
1508
1509         if (copy_from_user(&s, argp, sizeof s))
1510                 return -EFAULT;
1511
1512         if (!s.num || s.num > 0xffff || (s.num & (s.num - 1)))
1513                 return -EINVAL;
1514         vq->num = s.num;
1515
1516         return 0;
1517 }
1518
1519 static long vhost_vring_set_addr(struct vhost_dev *d,
1520                                  struct vhost_virtqueue *vq,
1521                                  void __user *argp)
1522 {
1523         struct vhost_vring_addr a;
1524
1525         if (copy_from_user(&a, argp, sizeof a))
1526                 return -EFAULT;
1527         if (a.flags & ~(0x1 << VHOST_VRING_F_LOG))
1528                 return -EOPNOTSUPP;
1529
1530         /* For 32bit, verify that the top 32bits of the user
1531            data are set to zero. */
1532         if ((u64)(unsigned long)a.desc_user_addr != a.desc_user_addr ||
1533             (u64)(unsigned long)a.used_user_addr != a.used_user_addr ||
1534             (u64)(unsigned long)a.avail_user_addr != a.avail_user_addr)
1535                 return -EFAULT;
1536
1537         /* Make sure it's safe to cast pointers to vring types. */
1538         BUILD_BUG_ON(__alignof__ *vq->avail > VRING_AVAIL_ALIGN_SIZE);
1539         BUILD_BUG_ON(__alignof__ *vq->used > VRING_USED_ALIGN_SIZE);
1540         if ((a.avail_user_addr & (VRING_AVAIL_ALIGN_SIZE - 1)) ||
1541             (a.used_user_addr & (VRING_USED_ALIGN_SIZE - 1)) ||
1542             (a.log_guest_addr & (VRING_USED_ALIGN_SIZE - 1)))
1543                 return -EINVAL;
1544
1545         /* We only verify access here if backend is configured.
1546          * If it is not, we don't as size might not have been setup.
1547          * We will verify when backend is configured. */
1548         if (vq->private_data) {
1549                 if (!vq_access_ok(vq, vq->num,
1550                         (void __user *)(unsigned long)a.desc_user_addr,
1551                         (void __user *)(unsigned long)a.avail_user_addr,
1552                         (void __user *)(unsigned long)a.used_user_addr))
1553                         return -EINVAL;
1554
1555                 /* Also validate log access for used ring if enabled. */
1556                 if (!vq_log_used_access_ok(vq, vq->log_base,
1557                                 a.flags & (0x1 << VHOST_VRING_F_LOG),
1558                                 a.log_guest_addr))
1559                         return -EINVAL;
1560         }
1561
1562         vq->log_used = !!(a.flags & (0x1 << VHOST_VRING_F_LOG));
1563         vq->desc = (void __user *)(unsigned long)a.desc_user_addr;
1564         vq->avail = (void __user *)(unsigned long)a.avail_user_addr;
1565         vq->log_addr = a.log_guest_addr;
1566         vq->used = (void __user *)(unsigned long)a.used_user_addr;
1567
1568         return 0;
1569 }
1570
1571 static long vhost_vring_set_num_addr(struct vhost_dev *d,
1572                                      struct vhost_virtqueue *vq,
1573                                      unsigned int ioctl,
1574                                      void __user *argp)
1575 {
1576         long r;
1577
1578         mutex_lock(&vq->mutex);
1579
1580         switch (ioctl) {
1581         case VHOST_SET_VRING_NUM:
1582                 r = vhost_vring_set_num(d, vq, argp);
1583                 break;
1584         case VHOST_SET_VRING_ADDR:
1585                 r = vhost_vring_set_addr(d, vq, argp);
1586                 break;
1587         default:
1588                 BUG();
1589         }
1590
1591         mutex_unlock(&vq->mutex);
1592
1593         return r;
1594 }
1595 long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
1596 {
1597         struct file *eventfp, *filep = NULL;
1598         bool pollstart = false, pollstop = false;
1599         struct eventfd_ctx *ctx = NULL;
1600         u32 __user *idxp = argp;
1601         struct vhost_virtqueue *vq;
1602         struct vhost_vring_state s;
1603         struct vhost_vring_file f;
1604         u32 idx;
1605         long r;
1606
1607         r = get_user(idx, idxp);
1608         if (r < 0)
1609                 return r;
1610         if (idx >= d->nvqs)
1611                 return -ENOBUFS;
1612
1613         idx = array_index_nospec(idx, d->nvqs);
1614         vq = d->vqs[idx];
1615
1616         if (ioctl == VHOST_SET_VRING_NUM ||
1617             ioctl == VHOST_SET_VRING_ADDR) {
1618                 return vhost_vring_set_num_addr(d, vq, ioctl, argp);
1619         }
1620
1621         mutex_lock(&vq->mutex);
1622
1623         switch (ioctl) {
1624         case VHOST_SET_VRING_BASE:
1625                 /* Moving base with an active backend?
1626                  * You don't want to do that. */
1627                 if (vq->private_data) {
1628                         r = -EBUSY;
1629                         break;
1630                 }
1631                 if (copy_from_user(&s, argp, sizeof s)) {
1632                         r = -EFAULT;
1633                         break;
1634                 }
1635                 if (s.num > 0xffff) {
1636                         r = -EINVAL;
1637                         break;
1638                 }
1639                 vq->last_avail_idx = s.num;
1640                 /* Forget the cached index value. */
1641                 vq->avail_idx = vq->last_avail_idx;
1642                 break;
1643         case VHOST_GET_VRING_BASE:
1644                 s.index = idx;
1645                 s.num = vq->last_avail_idx;
1646                 if (copy_to_user(argp, &s, sizeof s))
1647                         r = -EFAULT;
1648                 break;
1649         case VHOST_SET_VRING_KICK:
1650                 if (copy_from_user(&f, argp, sizeof f)) {
1651                         r = -EFAULT;
1652                         break;
1653                 }
1654                 eventfp = f.fd == VHOST_FILE_UNBIND ? NULL : eventfd_fget(f.fd);
1655                 if (IS_ERR(eventfp)) {
1656                         r = PTR_ERR(eventfp);
1657                         break;
1658                 }
1659                 if (eventfp != vq->kick) {
1660                         pollstop = (filep = vq->kick) != NULL;
1661                         pollstart = (vq->kick = eventfp) != NULL;
1662                 } else
1663                         filep = eventfp;
1664                 break;
1665         case VHOST_SET_VRING_CALL:
1666                 if (copy_from_user(&f, argp, sizeof f)) {
1667                         r = -EFAULT;
1668                         break;
1669                 }
1670                 ctx = f.fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(f.fd);
1671                 if (IS_ERR(ctx)) {
1672                         r = PTR_ERR(ctx);
1673                         break;
1674                 }
1675
1676                 swap(ctx, vq->call_ctx.ctx);
1677                 break;
1678         case VHOST_SET_VRING_ERR:
1679                 if (copy_from_user(&f, argp, sizeof f)) {
1680                         r = -EFAULT;
1681                         break;
1682                 }
1683                 ctx = f.fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(f.fd);
1684                 if (IS_ERR(ctx)) {
1685                         r = PTR_ERR(ctx);
1686                         break;
1687                 }
1688                 swap(ctx, vq->error_ctx);
1689                 break;
1690         case VHOST_SET_VRING_ENDIAN:
1691                 r = vhost_set_vring_endian(vq, argp);
1692                 break;
1693         case VHOST_GET_VRING_ENDIAN:
1694                 r = vhost_get_vring_endian(vq, idx, argp);
1695                 break;
1696         case VHOST_SET_VRING_BUSYLOOP_TIMEOUT:
1697                 if (copy_from_user(&s, argp, sizeof(s))) {
1698                         r = -EFAULT;
1699                         break;
1700                 }
1701                 vq->busyloop_timeout = s.num;
1702                 break;
1703         case VHOST_GET_VRING_BUSYLOOP_TIMEOUT:
1704                 s.index = idx;
1705                 s.num = vq->busyloop_timeout;
1706                 if (copy_to_user(argp, &s, sizeof(s)))
1707                         r = -EFAULT;
1708                 break;
1709         default:
1710                 r = -ENOIOCTLCMD;
1711         }
1712
1713         if (pollstop && vq->handle_kick)
1714                 vhost_poll_stop(&vq->poll);
1715
1716         if (!IS_ERR_OR_NULL(ctx))
1717                 eventfd_ctx_put(ctx);
1718         if (filep)
1719                 fput(filep);
1720
1721         if (pollstart && vq->handle_kick)
1722                 r = vhost_poll_start(&vq->poll, vq->kick);
1723
1724         mutex_unlock(&vq->mutex);
1725
1726         if (pollstop && vq->handle_kick)
1727                 vhost_dev_flush(vq->poll.dev);
1728         return r;
1729 }
1730 EXPORT_SYMBOL_GPL(vhost_vring_ioctl);
1731
1732 int vhost_init_device_iotlb(struct vhost_dev *d, bool enabled)
1733 {
1734         struct vhost_iotlb *niotlb, *oiotlb;
1735         int i;
1736
1737         niotlb = iotlb_alloc();
1738         if (!niotlb)
1739                 return -ENOMEM;
1740
1741         oiotlb = d->iotlb;
1742         d->iotlb = niotlb;
1743
1744         for (i = 0; i < d->nvqs; ++i) {
1745                 struct vhost_virtqueue *vq = d->vqs[i];
1746
1747                 mutex_lock(&vq->mutex);
1748                 vq->iotlb = niotlb;
1749                 __vhost_vq_meta_reset(vq);
1750                 mutex_unlock(&vq->mutex);
1751         }
1752
1753         vhost_iotlb_free(oiotlb);
1754
1755         return 0;
1756 }
1757 EXPORT_SYMBOL_GPL(vhost_init_device_iotlb);
1758
1759 /* Caller must have device mutex */
1760 long vhost_dev_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
1761 {
1762         struct eventfd_ctx *ctx;
1763         u64 p;
1764         long r;
1765         int i, fd;
1766
1767         /* If you are not the owner, you can become one */
1768         if (ioctl == VHOST_SET_OWNER) {
1769                 r = vhost_dev_set_owner(d);
1770                 goto done;
1771         }
1772
1773         /* You must be the owner to do anything else */
1774         r = vhost_dev_check_owner(d);
1775         if (r)
1776                 goto done;
1777
1778         switch (ioctl) {
1779         case VHOST_SET_MEM_TABLE:
1780                 r = vhost_set_memory(d, argp);
1781                 break;
1782         case VHOST_SET_LOG_BASE:
1783                 if (copy_from_user(&p, argp, sizeof p)) {
1784                         r = -EFAULT;
1785                         break;
1786                 }
1787                 if ((u64)(unsigned long)p != p) {
1788                         r = -EFAULT;
1789                         break;
1790                 }
1791                 for (i = 0; i < d->nvqs; ++i) {
1792                         struct vhost_virtqueue *vq;
1793                         void __user *base = (void __user *)(unsigned long)p;
1794                         vq = d->vqs[i];
1795                         mutex_lock(&vq->mutex);
1796                         /* If ring is inactive, will check when it's enabled. */
1797                         if (vq->private_data && !vq_log_access_ok(vq, base))
1798                                 r = -EFAULT;
1799                         else
1800                                 vq->log_base = base;
1801                         mutex_unlock(&vq->mutex);
1802                 }
1803                 break;
1804         case VHOST_SET_LOG_FD:
1805                 r = get_user(fd, (int __user *)argp);
1806                 if (r < 0)
1807                         break;
1808                 ctx = fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(fd);
1809                 if (IS_ERR(ctx)) {
1810                         r = PTR_ERR(ctx);
1811                         break;
1812                 }
1813                 swap(ctx, d->log_ctx);
1814                 for (i = 0; i < d->nvqs; ++i) {
1815                         mutex_lock(&d->vqs[i]->mutex);
1816                         d->vqs[i]->log_ctx = d->log_ctx;
1817                         mutex_unlock(&d->vqs[i]->mutex);
1818                 }
1819                 if (ctx)
1820                         eventfd_ctx_put(ctx);
1821                 break;
1822         default:
1823                 r = -ENOIOCTLCMD;
1824                 break;
1825         }
1826 done:
1827         return r;
1828 }
1829 EXPORT_SYMBOL_GPL(vhost_dev_ioctl);
1830
1831 /* TODO: This is really inefficient.  We need something like get_user()
1832  * (instruction directly accesses the data, with an exception table entry
1833  * returning -EFAULT). See Documentation/x86/exception-tables.rst.
1834  */
1835 static int set_bit_to_user(int nr, void __user *addr)
1836 {
1837         unsigned long log = (unsigned long)addr;
1838         struct page *page;
1839         void *base;
1840         int bit = nr + (log % PAGE_SIZE) * 8;
1841         int r;
1842
1843         r = pin_user_pages_fast(log, 1, FOLL_WRITE, &page);
1844         if (r < 0)
1845                 return r;
1846         BUG_ON(r != 1);
1847         base = kmap_atomic(page);
1848         set_bit(bit, base);
1849         kunmap_atomic(base);
1850         unpin_user_pages_dirty_lock(&page, 1, true);
1851         return 0;
1852 }
1853
1854 static int log_write(void __user *log_base,
1855                      u64 write_address, u64 write_length)
1856 {
1857         u64 write_page = write_address / VHOST_PAGE_SIZE;
1858         int r;
1859
1860         if (!write_length)
1861                 return 0;
1862         write_length += write_address % VHOST_PAGE_SIZE;
1863         for (;;) {
1864                 u64 base = (u64)(unsigned long)log_base;
1865                 u64 log = base + write_page / 8;
1866                 int bit = write_page % 8;
1867                 if ((u64)(unsigned long)log != log)
1868                         return -EFAULT;
1869                 r = set_bit_to_user(bit, (void __user *)(unsigned long)log);
1870                 if (r < 0)
1871                         return r;
1872                 if (write_length <= VHOST_PAGE_SIZE)
1873                         break;
1874                 write_length -= VHOST_PAGE_SIZE;
1875                 write_page += 1;
1876         }
1877         return r;
1878 }
1879
1880 static int log_write_hva(struct vhost_virtqueue *vq, u64 hva, u64 len)
1881 {
1882         struct vhost_iotlb *umem = vq->umem;
1883         struct vhost_iotlb_map *u;
1884         u64 start, end, l, min;
1885         int r;
1886         bool hit = false;
1887
1888         while (len) {
1889                 min = len;
1890                 /* More than one GPAs can be mapped into a single HVA. So
1891                  * iterate all possible umems here to be safe.
1892                  */
1893                 list_for_each_entry(u, &umem->list, link) {
1894                         if (u->addr > hva - 1 + len ||
1895                             u->addr - 1 + u->size < hva)
1896                                 continue;
1897                         start = max(u->addr, hva);
1898                         end = min(u->addr - 1 + u->size, hva - 1 + len);
1899                         l = end - start + 1;
1900                         r = log_write(vq->log_base,
1901                                       u->start + start - u->addr,
1902                                       l);
1903                         if (r < 0)
1904                                 return r;
1905                         hit = true;
1906                         min = min(l, min);
1907                 }
1908
1909                 if (!hit)
1910                         return -EFAULT;
1911
1912                 len -= min;
1913                 hva += min;
1914         }
1915
1916         return 0;
1917 }
1918
1919 static int log_used(struct vhost_virtqueue *vq, u64 used_offset, u64 len)
1920 {
1921         struct iovec *iov = vq->log_iov;
1922         int i, ret;
1923
1924         if (!vq->iotlb)
1925                 return log_write(vq->log_base, vq->log_addr + used_offset, len);
1926
1927         ret = translate_desc(vq, (uintptr_t)vq->used + used_offset,
1928                              len, iov, 64, VHOST_ACCESS_WO);
1929         if (ret < 0)
1930                 return ret;
1931
1932         for (i = 0; i < ret; i++) {
1933                 ret = log_write_hva(vq, (uintptr_t)iov[i].iov_base,
1934                                     iov[i].iov_len);
1935                 if (ret)
1936                         return ret;
1937         }
1938
1939         return 0;
1940 }
1941
1942 int vhost_log_write(struct vhost_virtqueue *vq, struct vhost_log *log,
1943                     unsigned int log_num, u64 len, struct iovec *iov, int count)
1944 {
1945         int i, r;
1946
1947         /* Make sure data written is seen before log. */
1948         smp_wmb();
1949
1950         if (vq->iotlb) {
1951                 for (i = 0; i < count; i++) {
1952                         r = log_write_hva(vq, (uintptr_t)iov[i].iov_base,
1953                                           iov[i].iov_len);
1954                         if (r < 0)
1955                                 return r;
1956                 }
1957                 return 0;
1958         }
1959
1960         for (i = 0; i < log_num; ++i) {
1961                 u64 l = min(log[i].len, len);
1962                 r = log_write(vq->log_base, log[i].addr, l);
1963                 if (r < 0)
1964                         return r;
1965                 len -= l;
1966                 if (!len) {
1967                         if (vq->log_ctx)
1968                                 eventfd_signal(vq->log_ctx, 1);
1969                         return 0;
1970                 }
1971         }
1972         /* Length written exceeds what we have stored. This is a bug. */
1973         BUG();
1974         return 0;
1975 }
1976 EXPORT_SYMBOL_GPL(vhost_log_write);
1977
1978 static int vhost_update_used_flags(struct vhost_virtqueue *vq)
1979 {
1980         void __user *used;
1981         if (vhost_put_used_flags(vq))
1982                 return -EFAULT;
1983         if (unlikely(vq->log_used)) {
1984                 /* Make sure the flag is seen before log. */
1985                 smp_wmb();
1986                 /* Log used flag write. */
1987                 used = &vq->used->flags;
1988                 log_used(vq, (used - (void __user *)vq->used),
1989                          sizeof vq->used->flags);
1990                 if (vq->log_ctx)
1991                         eventfd_signal(vq->log_ctx, 1);
1992         }
1993         return 0;
1994 }
1995
1996 static int vhost_update_avail_event(struct vhost_virtqueue *vq)
1997 {
1998         if (vhost_put_avail_event(vq))
1999                 return -EFAULT;
2000         if (unlikely(vq->log_used)) {
2001                 void __user *used;
2002                 /* Make sure the event is seen before log. */
2003                 smp_wmb();
2004                 /* Log avail event write */
2005                 used = vhost_avail_event(vq);
2006                 log_used(vq, (used - (void __user *)vq->used),
2007                          sizeof *vhost_avail_event(vq));
2008                 if (vq->log_ctx)
2009                         eventfd_signal(vq->log_ctx, 1);
2010         }
2011         return 0;
2012 }
2013
2014 int vhost_vq_init_access(struct vhost_virtqueue *vq)
2015 {
2016         __virtio16 last_used_idx;
2017         int r;
2018         bool is_le = vq->is_le;
2019
2020         if (!vq->private_data)
2021                 return 0;
2022
2023         vhost_init_is_le(vq);
2024
2025         r = vhost_update_used_flags(vq);
2026         if (r)
2027                 goto err;
2028         vq->signalled_used_valid = false;
2029         if (!vq->iotlb &&
2030             !access_ok(&vq->used->idx, sizeof vq->used->idx)) {
2031                 r = -EFAULT;
2032                 goto err;
2033         }
2034         r = vhost_get_used_idx(vq, &last_used_idx);
2035         if (r) {
2036                 vq_err(vq, "Can't access used idx at %p\n",
2037                        &vq->used->idx);
2038                 goto err;
2039         }
2040         vq->last_used_idx = vhost16_to_cpu(vq, last_used_idx);
2041         return 0;
2042
2043 err:
2044         vq->is_le = is_le;
2045         return r;
2046 }
2047 EXPORT_SYMBOL_GPL(vhost_vq_init_access);
2048
2049 static int translate_desc(struct vhost_virtqueue *vq, u64 addr, u32 len,
2050                           struct iovec iov[], int iov_size, int access)
2051 {
2052         const struct vhost_iotlb_map *map;
2053         struct vhost_dev *dev = vq->dev;
2054         struct vhost_iotlb *umem = dev->iotlb ? dev->iotlb : dev->umem;
2055         struct iovec *_iov;
2056         u64 s = 0;
2057         int ret = 0;
2058
2059         while ((u64)len > s) {
2060                 u64 size;
2061                 if (unlikely(ret >= iov_size)) {
2062                         ret = -ENOBUFS;
2063                         break;
2064                 }
2065
2066                 map = vhost_iotlb_itree_first(umem, addr, addr + len - 1);
2067                 if (map == NULL || map->start > addr) {
2068                         if (umem != dev->iotlb) {
2069                                 ret = -EFAULT;
2070                                 break;
2071                         }
2072                         ret = -EAGAIN;
2073                         break;
2074                 } else if (!(map->perm & access)) {
2075                         ret = -EPERM;
2076                         break;
2077                 }
2078
2079                 _iov = iov + ret;
2080                 size = map->size - addr + map->start;
2081                 _iov->iov_len = min((u64)len - s, size);
2082                 _iov->iov_base = (void __user *)(unsigned long)
2083                                  (map->addr + addr - map->start);
2084                 s += size;
2085                 addr += size;
2086                 ++ret;
2087         }
2088
2089         if (ret == -EAGAIN)
2090                 vhost_iotlb_miss(vq, addr, access);
2091         return ret;
2092 }
2093
2094 /* Each buffer in the virtqueues is actually a chain of descriptors.  This
2095  * function returns the next descriptor in the chain,
2096  * or -1U if we're at the end. */
2097 static unsigned next_desc(struct vhost_virtqueue *vq, struct vring_desc *desc)
2098 {
2099         unsigned int next;
2100
2101         /* If this descriptor says it doesn't chain, we're done. */
2102         if (!(desc->flags & cpu_to_vhost16(vq, VRING_DESC_F_NEXT)))
2103                 return -1U;
2104
2105         /* Check they're not leading us off end of descriptors. */
2106         next = vhost16_to_cpu(vq, READ_ONCE(desc->next));
2107         return next;
2108 }
2109
2110 static int get_indirect(struct vhost_virtqueue *vq,
2111                         struct iovec iov[], unsigned int iov_size,
2112                         unsigned int *out_num, unsigned int *in_num,
2113                         struct vhost_log *log, unsigned int *log_num,
2114                         struct vring_desc *indirect)
2115 {
2116         struct vring_desc desc;
2117         unsigned int i = 0, count, found = 0;
2118         u32 len = vhost32_to_cpu(vq, indirect->len);
2119         struct iov_iter from;
2120         int ret, access;
2121
2122         /* Sanity check */
2123         if (unlikely(len % sizeof desc)) {
2124                 vq_err(vq, "Invalid length in indirect descriptor: "
2125                        "len 0x%llx not multiple of 0x%zx\n",
2126                        (unsigned long long)len,
2127                        sizeof desc);
2128                 return -EINVAL;
2129         }
2130
2131         ret = translate_desc(vq, vhost64_to_cpu(vq, indirect->addr), len, vq->indirect,
2132                              UIO_MAXIOV, VHOST_ACCESS_RO);
2133         if (unlikely(ret < 0)) {
2134                 if (ret != -EAGAIN)
2135                         vq_err(vq, "Translation failure %d in indirect.\n", ret);
2136                 return ret;
2137         }
2138         iov_iter_init(&from, ITER_SOURCE, vq->indirect, ret, len);
2139         count = len / sizeof desc;
2140         /* Buffers are chained via a 16 bit next field, so
2141          * we can have at most 2^16 of these. */
2142         if (unlikely(count > USHRT_MAX + 1)) {
2143                 vq_err(vq, "Indirect buffer length too big: %d\n",
2144                        indirect->len);
2145                 return -E2BIG;
2146         }
2147
2148         do {
2149                 unsigned iov_count = *in_num + *out_num;
2150                 if (unlikely(++found > count)) {
2151                         vq_err(vq, "Loop detected: last one at %u "
2152                                "indirect size %u\n",
2153                                i, count);
2154                         return -EINVAL;
2155                 }
2156                 if (unlikely(!copy_from_iter_full(&desc, sizeof(desc), &from))) {
2157                         vq_err(vq, "Failed indirect descriptor: idx %d, %zx\n",
2158                                i, (size_t)vhost64_to_cpu(vq, indirect->addr) + i * sizeof desc);
2159                         return -EINVAL;
2160                 }
2161                 if (unlikely(desc.flags & cpu_to_vhost16(vq, VRING_DESC_F_INDIRECT))) {
2162                         vq_err(vq, "Nested indirect descriptor: idx %d, %zx\n",
2163                                i, (size_t)vhost64_to_cpu(vq, indirect->addr) + i * sizeof desc);
2164                         return -EINVAL;
2165                 }
2166
2167                 if (desc.flags & cpu_to_vhost16(vq, VRING_DESC_F_WRITE))
2168                         access = VHOST_ACCESS_WO;
2169                 else
2170                         access = VHOST_ACCESS_RO;
2171
2172                 ret = translate_desc(vq, vhost64_to_cpu(vq, desc.addr),
2173                                      vhost32_to_cpu(vq, desc.len), iov + iov_count,
2174                                      iov_size - iov_count, access);
2175                 if (unlikely(ret < 0)) {
2176                         if (ret != -EAGAIN)
2177                                 vq_err(vq, "Translation failure %d indirect idx %d\n",
2178                                         ret, i);
2179                         return ret;
2180                 }
2181                 /* If this is an input descriptor, increment that count. */
2182                 if (access == VHOST_ACCESS_WO) {
2183                         *in_num += ret;
2184                         if (unlikely(log && ret)) {
2185                                 log[*log_num].addr = vhost64_to_cpu(vq, desc.addr);
2186                                 log[*log_num].len = vhost32_to_cpu(vq, desc.len);
2187                                 ++*log_num;
2188                         }
2189                 } else {
2190                         /* If it's an output descriptor, they're all supposed
2191                          * to come before any input descriptors. */
2192                         if (unlikely(*in_num)) {
2193                                 vq_err(vq, "Indirect descriptor "
2194                                        "has out after in: idx %d\n", i);
2195                                 return -EINVAL;
2196                         }
2197                         *out_num += ret;
2198                 }
2199         } while ((i = next_desc(vq, &desc)) != -1);
2200         return 0;
2201 }
2202
2203 /* This looks in the virtqueue and for the first available buffer, and converts
2204  * it to an iovec for convenient access.  Since descriptors consist of some
2205  * number of output then some number of input descriptors, it's actually two
2206  * iovecs, but we pack them into one and note how many of each there were.
2207  *
2208  * This function returns the descriptor number found, or vq->num (which is
2209  * never a valid descriptor number) if none was found.  A negative code is
2210  * returned on error. */
2211 int vhost_get_vq_desc(struct vhost_virtqueue *vq,
2212                       struct iovec iov[], unsigned int iov_size,
2213                       unsigned int *out_num, unsigned int *in_num,
2214                       struct vhost_log *log, unsigned int *log_num)
2215 {
2216         struct vring_desc desc;
2217         unsigned int i, head, found = 0;
2218         u16 last_avail_idx;
2219         __virtio16 avail_idx;
2220         __virtio16 ring_head;
2221         int ret, access;
2222
2223         /* Check it isn't doing very strange things with descriptor numbers. */
2224         last_avail_idx = vq->last_avail_idx;
2225
2226         if (vq->avail_idx == vq->last_avail_idx) {
2227                 if (unlikely(vhost_get_avail_idx(vq, &avail_idx))) {
2228                         vq_err(vq, "Failed to access avail idx at %p\n",
2229                                 &vq->avail->idx);
2230                         return -EFAULT;
2231                 }
2232                 vq->avail_idx = vhost16_to_cpu(vq, avail_idx);
2233
2234                 if (unlikely((u16)(vq->avail_idx - last_avail_idx) > vq->num)) {
2235                         vq_err(vq, "Guest moved used index from %u to %u",
2236                                 last_avail_idx, vq->avail_idx);
2237                         return -EFAULT;
2238                 }
2239
2240                 /* If there's nothing new since last we looked, return
2241                  * invalid.
2242                  */
2243                 if (vq->avail_idx == last_avail_idx)
2244                         return vq->num;
2245
2246                 /* Only get avail ring entries after they have been
2247                  * exposed by guest.
2248                  */
2249                 smp_rmb();
2250         }
2251
2252         /* Grab the next descriptor number they're advertising, and increment
2253          * the index we've seen. */
2254         if (unlikely(vhost_get_avail_head(vq, &ring_head, last_avail_idx))) {
2255                 vq_err(vq, "Failed to read head: idx %d address %p\n",
2256                        last_avail_idx,
2257                        &vq->avail->ring[last_avail_idx % vq->num]);
2258                 return -EFAULT;
2259         }
2260
2261         head = vhost16_to_cpu(vq, ring_head);
2262
2263         /* If their number is silly, that's an error. */
2264         if (unlikely(head >= vq->num)) {
2265                 vq_err(vq, "Guest says index %u > %u is available",
2266                        head, vq->num);
2267                 return -EINVAL;
2268         }
2269
2270         /* When we start there are none of either input nor output. */
2271         *out_num = *in_num = 0;
2272         if (unlikely(log))
2273                 *log_num = 0;
2274
2275         i = head;
2276         do {
2277                 unsigned iov_count = *in_num + *out_num;
2278                 if (unlikely(i >= vq->num)) {
2279                         vq_err(vq, "Desc index is %u > %u, head = %u",
2280                                i, vq->num, head);
2281                         return -EINVAL;
2282                 }
2283                 if (unlikely(++found > vq->num)) {
2284                         vq_err(vq, "Loop detected: last one at %u "
2285                                "vq size %u head %u\n",
2286                                i, vq->num, head);
2287                         return -EINVAL;
2288                 }
2289                 ret = vhost_get_desc(vq, &desc, i);
2290                 if (unlikely(ret)) {
2291                         vq_err(vq, "Failed to get descriptor: idx %d addr %p\n",
2292                                i, vq->desc + i);
2293                         return -EFAULT;
2294                 }
2295                 if (desc.flags & cpu_to_vhost16(vq, VRING_DESC_F_INDIRECT)) {
2296                         ret = get_indirect(vq, iov, iov_size,
2297                                            out_num, in_num,
2298                                            log, log_num, &desc);
2299                         if (unlikely(ret < 0)) {
2300                                 if (ret != -EAGAIN)
2301                                         vq_err(vq, "Failure detected "
2302                                                 "in indirect descriptor at idx %d\n", i);
2303                                 return ret;
2304                         }
2305                         continue;
2306                 }
2307
2308                 if (desc.flags & cpu_to_vhost16(vq, VRING_DESC_F_WRITE))
2309                         access = VHOST_ACCESS_WO;
2310                 else
2311                         access = VHOST_ACCESS_RO;
2312                 ret = translate_desc(vq, vhost64_to_cpu(vq, desc.addr),
2313                                      vhost32_to_cpu(vq, desc.len), iov + iov_count,
2314                                      iov_size - iov_count, access);
2315                 if (unlikely(ret < 0)) {
2316                         if (ret != -EAGAIN)
2317                                 vq_err(vq, "Translation failure %d descriptor idx %d\n",
2318                                         ret, i);
2319                         return ret;
2320                 }
2321                 if (access == VHOST_ACCESS_WO) {
2322                         /* If this is an input descriptor,
2323                          * increment that count. */
2324                         *in_num += ret;
2325                         if (unlikely(log && ret)) {
2326                                 log[*log_num].addr = vhost64_to_cpu(vq, desc.addr);
2327                                 log[*log_num].len = vhost32_to_cpu(vq, desc.len);
2328                                 ++*log_num;
2329                         }
2330                 } else {
2331                         /* If it's an output descriptor, they're all supposed
2332                          * to come before any input descriptors. */
2333                         if (unlikely(*in_num)) {
2334                                 vq_err(vq, "Descriptor has out after in: "
2335                                        "idx %d\n", i);
2336                                 return -EINVAL;
2337                         }
2338                         *out_num += ret;
2339                 }
2340         } while ((i = next_desc(vq, &desc)) != -1);
2341
2342         /* On success, increment avail index. */
2343         vq->last_avail_idx++;
2344
2345         /* Assume notifications from guest are disabled at this point,
2346          * if they aren't we would need to update avail_event index. */
2347         BUG_ON(!(vq->used_flags & VRING_USED_F_NO_NOTIFY));
2348         return head;
2349 }
2350 EXPORT_SYMBOL_GPL(vhost_get_vq_desc);
2351
2352 /* Reverse the effect of vhost_get_vq_desc. Useful for error handling. */
2353 void vhost_discard_vq_desc(struct vhost_virtqueue *vq, int n)
2354 {
2355         vq->last_avail_idx -= n;
2356 }
2357 EXPORT_SYMBOL_GPL(vhost_discard_vq_desc);
2358
2359 /* After we've used one of their buffers, we tell them about it.  We'll then
2360  * want to notify the guest, using eventfd. */
2361 int vhost_add_used(struct vhost_virtqueue *vq, unsigned int head, int len)
2362 {
2363         struct vring_used_elem heads = {
2364                 cpu_to_vhost32(vq, head),
2365                 cpu_to_vhost32(vq, len)
2366         };
2367
2368         return vhost_add_used_n(vq, &heads, 1);
2369 }
2370 EXPORT_SYMBOL_GPL(vhost_add_used);
2371
2372 static int __vhost_add_used_n(struct vhost_virtqueue *vq,
2373                             struct vring_used_elem *heads,
2374                             unsigned count)
2375 {
2376         vring_used_elem_t __user *used;
2377         u16 old, new;
2378         int start;
2379
2380         start = vq->last_used_idx & (vq->num - 1);
2381         used = vq->used->ring + start;
2382         if (vhost_put_used(vq, heads, start, count)) {
2383                 vq_err(vq, "Failed to write used");
2384                 return -EFAULT;
2385         }
2386         if (unlikely(vq->log_used)) {
2387                 /* Make sure data is seen before log. */
2388                 smp_wmb();
2389                 /* Log used ring entry write. */
2390                 log_used(vq, ((void __user *)used - (void __user *)vq->used),
2391                          count * sizeof *used);
2392         }
2393         old = vq->last_used_idx;
2394         new = (vq->last_used_idx += count);
2395         /* If the driver never bothers to signal in a very long while,
2396          * used index might wrap around. If that happens, invalidate
2397          * signalled_used index we stored. TODO: make sure driver
2398          * signals at least once in 2^16 and remove this. */
2399         if (unlikely((u16)(new - vq->signalled_used) < (u16)(new - old)))
2400                 vq->signalled_used_valid = false;
2401         return 0;
2402 }
2403
2404 /* After we've used one of their buffers, we tell them about it.  We'll then
2405  * want to notify the guest, using eventfd. */
2406 int vhost_add_used_n(struct vhost_virtqueue *vq, struct vring_used_elem *heads,
2407                      unsigned count)
2408 {
2409         int start, n, r;
2410
2411         start = vq->last_used_idx & (vq->num - 1);
2412         n = vq->num - start;
2413         if (n < count) {
2414                 r = __vhost_add_used_n(vq, heads, n);
2415                 if (r < 0)
2416                         return r;
2417                 heads += n;
2418                 count -= n;
2419         }
2420         r = __vhost_add_used_n(vq, heads, count);
2421
2422         /* Make sure buffer is written before we update index. */
2423         smp_wmb();
2424         if (vhost_put_used_idx(vq)) {
2425                 vq_err(vq, "Failed to increment used idx");
2426                 return -EFAULT;
2427         }
2428         if (unlikely(vq->log_used)) {
2429                 /* Make sure used idx is seen before log. */
2430                 smp_wmb();
2431                 /* Log used index update. */
2432                 log_used(vq, offsetof(struct vring_used, idx),
2433                          sizeof vq->used->idx);
2434                 if (vq->log_ctx)
2435                         eventfd_signal(vq->log_ctx, 1);
2436         }
2437         return r;
2438 }
2439 EXPORT_SYMBOL_GPL(vhost_add_used_n);
2440
2441 static bool vhost_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
2442 {
2443         __u16 old, new;
2444         __virtio16 event;
2445         bool v;
2446         /* Flush out used index updates. This is paired
2447          * with the barrier that the Guest executes when enabling
2448          * interrupts. */
2449         smp_mb();
2450
2451         if (vhost_has_feature(vq, VIRTIO_F_NOTIFY_ON_EMPTY) &&
2452             unlikely(vq->avail_idx == vq->last_avail_idx))
2453                 return true;
2454
2455         if (!vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX)) {
2456                 __virtio16 flags;
2457                 if (vhost_get_avail_flags(vq, &flags)) {
2458                         vq_err(vq, "Failed to get flags");
2459                         return true;
2460                 }
2461                 return !(flags & cpu_to_vhost16(vq, VRING_AVAIL_F_NO_INTERRUPT));
2462         }
2463         old = vq->signalled_used;
2464         v = vq->signalled_used_valid;
2465         new = vq->signalled_used = vq->last_used_idx;
2466         vq->signalled_used_valid = true;
2467
2468         if (unlikely(!v))
2469                 return true;
2470
2471         if (vhost_get_used_event(vq, &event)) {
2472                 vq_err(vq, "Failed to get used event idx");
2473                 return true;
2474         }
2475         return vring_need_event(vhost16_to_cpu(vq, event), new, old);
2476 }
2477
2478 /* This actually signals the guest, using eventfd. */
2479 void vhost_signal(struct vhost_dev *dev, struct vhost_virtqueue *vq)
2480 {
2481         /* Signal the Guest tell them we used something up. */
2482         if (vq->call_ctx.ctx && vhost_notify(dev, vq))
2483                 eventfd_signal(vq->call_ctx.ctx, 1);
2484 }
2485 EXPORT_SYMBOL_GPL(vhost_signal);
2486
2487 /* And here's the combo meal deal.  Supersize me! */
2488 void vhost_add_used_and_signal(struct vhost_dev *dev,
2489                                struct vhost_virtqueue *vq,
2490                                unsigned int head, int len)
2491 {
2492         vhost_add_used(vq, head, len);
2493         vhost_signal(dev, vq);
2494 }
2495 EXPORT_SYMBOL_GPL(vhost_add_used_and_signal);
2496
2497 /* multi-buffer version of vhost_add_used_and_signal */
2498 void vhost_add_used_and_signal_n(struct vhost_dev *dev,
2499                                  struct vhost_virtqueue *vq,
2500                                  struct vring_used_elem *heads, unsigned count)
2501 {
2502         vhost_add_used_n(vq, heads, count);
2503         vhost_signal(dev, vq);
2504 }
2505 EXPORT_SYMBOL_GPL(vhost_add_used_and_signal_n);
2506
2507 /* return true if we're sure that avaiable ring is empty */
2508 bool vhost_vq_avail_empty(struct vhost_dev *dev, struct vhost_virtqueue *vq)
2509 {
2510         __virtio16 avail_idx;
2511         int r;
2512
2513         if (vq->avail_idx != vq->last_avail_idx)
2514                 return false;
2515
2516         r = vhost_get_avail_idx(vq, &avail_idx);
2517         if (unlikely(r))
2518                 return false;
2519         vq->avail_idx = vhost16_to_cpu(vq, avail_idx);
2520
2521         return vq->avail_idx == vq->last_avail_idx;
2522 }
2523 EXPORT_SYMBOL_GPL(vhost_vq_avail_empty);
2524
2525 /* OK, now we need to know about added descriptors. */
2526 bool vhost_enable_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
2527 {
2528         __virtio16 avail_idx;
2529         int r;
2530
2531         if (!(vq->used_flags & VRING_USED_F_NO_NOTIFY))
2532                 return false;
2533         vq->used_flags &= ~VRING_USED_F_NO_NOTIFY;
2534         if (!vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX)) {
2535                 r = vhost_update_used_flags(vq);
2536                 if (r) {
2537                         vq_err(vq, "Failed to enable notification at %p: %d\n",
2538                                &vq->used->flags, r);
2539                         return false;
2540                 }
2541         } else {
2542                 r = vhost_update_avail_event(vq);
2543                 if (r) {
2544                         vq_err(vq, "Failed to update avail event index at %p: %d\n",
2545                                vhost_avail_event(vq), r);
2546                         return false;
2547                 }
2548         }
2549         /* They could have slipped one in as we were doing that: make
2550          * sure it's written, then check again. */
2551         smp_mb();
2552         r = vhost_get_avail_idx(vq, &avail_idx);
2553         if (r) {
2554                 vq_err(vq, "Failed to check avail idx at %p: %d\n",
2555                        &vq->avail->idx, r);
2556                 return false;
2557         }
2558         vq->avail_idx = vhost16_to_cpu(vq, avail_idx);
2559
2560         return vq->avail_idx != vq->last_avail_idx;
2561 }
2562 EXPORT_SYMBOL_GPL(vhost_enable_notify);
2563
2564 /* We don't need to be notified again. */
2565 void vhost_disable_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
2566 {
2567         int r;
2568
2569         if (vq->used_flags & VRING_USED_F_NO_NOTIFY)
2570                 return;
2571         vq->used_flags |= VRING_USED_F_NO_NOTIFY;
2572         if (!vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX)) {
2573                 r = vhost_update_used_flags(vq);
2574                 if (r)
2575                         vq_err(vq, "Failed to disable notification at %p: %d\n",
2576                                &vq->used->flags, r);
2577         }
2578 }
2579 EXPORT_SYMBOL_GPL(vhost_disable_notify);
2580
2581 /* Create a new message. */
2582 struct vhost_msg_node *vhost_new_msg(struct vhost_virtqueue *vq, int type)
2583 {
2584         struct vhost_msg_node *node = kmalloc(sizeof *node, GFP_KERNEL);
2585         if (!node)
2586                 return NULL;
2587
2588         /* Make sure all padding within the structure is initialized. */
2589         memset(&node->msg, 0, sizeof node->msg);
2590         node->vq = vq;
2591         node->msg.type = type;
2592         return node;
2593 }
2594 EXPORT_SYMBOL_GPL(vhost_new_msg);
2595
2596 void vhost_enqueue_msg(struct vhost_dev *dev, struct list_head *head,
2597                        struct vhost_msg_node *node)
2598 {
2599         spin_lock(&dev->iotlb_lock);
2600         list_add_tail(&node->node, head);
2601         spin_unlock(&dev->iotlb_lock);
2602
2603         wake_up_interruptible_poll(&dev->wait, EPOLLIN | EPOLLRDNORM);
2604 }
2605 EXPORT_SYMBOL_GPL(vhost_enqueue_msg);
2606
2607 struct vhost_msg_node *vhost_dequeue_msg(struct vhost_dev *dev,
2608                                          struct list_head *head)
2609 {
2610         struct vhost_msg_node *node = NULL;
2611
2612         spin_lock(&dev->iotlb_lock);
2613         if (!list_empty(head)) {
2614                 node = list_first_entry(head, struct vhost_msg_node,
2615                                         node);
2616                 list_del(&node->node);
2617         }
2618         spin_unlock(&dev->iotlb_lock);
2619
2620         return node;
2621 }
2622 EXPORT_SYMBOL_GPL(vhost_dequeue_msg);
2623
2624 void vhost_set_backend_features(struct vhost_dev *dev, u64 features)
2625 {
2626         struct vhost_virtqueue *vq;
2627         int i;
2628
2629         mutex_lock(&dev->mutex);
2630         for (i = 0; i < dev->nvqs; ++i) {
2631                 vq = dev->vqs[i];
2632                 mutex_lock(&vq->mutex);
2633                 vq->acked_backend_features = features;
2634                 mutex_unlock(&vq->mutex);
2635         }
2636         mutex_unlock(&dev->mutex);
2637 }
2638 EXPORT_SYMBOL_GPL(vhost_set_backend_features);
2639
2640 static int __init vhost_init(void)
2641 {
2642         return 0;
2643 }
2644
2645 static void __exit vhost_exit(void)
2646 {
2647 }
2648
2649 module_init(vhost_init);
2650 module_exit(vhost_exit);
2651
2652 MODULE_VERSION("0.0.1");
2653 MODULE_LICENSE("GPL v2");
2654 MODULE_AUTHOR("Michael S. Tsirkin");
2655 MODULE_DESCRIPTION("Host kernel accelerator for virtio");