lib/librte_vhost/vhost_user.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2010-2018 Intel Corporation
   3  */
   4
   5 /* Security model
   6  * --------------
   7  * The vhost-user protocol connection is an external interface, so it must be
   8  * robust against invalid inputs.
   9  *
  10  * This is important because the vhost-user master is only one step removed
  11  * from the guest.  Malicious guests that have escaped will then launch further
  12  * attacks from the vhost-user master.
  13  *
  14  * Even in deployments where guests are trusted, a bug in the vhost-user master
  15  * can still cause invalid messages to be sent.  Such messages must not
  16  * compromise the stability of the DPDK application by causing crashes, memory
  17  * corruption, or other problematic behavior.
  18  *
  19  * Do not assume received VhostUserMsg fields contain sensible values!
  20  */
  21
  22 #include <stdint.h>
  23 #include <stdio.h>
  24 #include <stdlib.h>
  25 #include <string.h>
  26 #include <unistd.h>
  27 #include <sys/mman.h>
  28 #include <sys/types.h>
  29 #include <sys/stat.h>
  30 #include <assert.h>
  31 #ifdef RTE_LIBRTE_VHOST_NUMA
  32 #include <numaif.h>
  33 #endif
  34
  35 #include <rte_common.h>
  36 #include <rte_malloc.h>
  37 #include <rte_log.h>
  38
  39 #include "iotlb.h"
  40 #include "vhost.h"
  41 #include "vhost_user.h"
  42
  43 #define VIRTIO_MIN_MTU 68
  44 #define VIRTIO_MAX_MTU 65535
  45
  46 static const char *vhost_message_str[VHOST_USER_MAX] = {
  47         [VHOST_USER_NONE] = "VHOST_USER_NONE",
  48         [VHOST_USER_GET_FEATURES] = "VHOST_USER_GET_FEATURES",
  49         [VHOST_USER_SET_FEATURES] = "VHOST_USER_SET_FEATURES",
  50         [VHOST_USER_SET_OWNER] = "VHOST_USER_SET_OWNER",
  51         [VHOST_USER_RESET_OWNER] = "VHOST_USER_RESET_OWNER",
  52         [VHOST_USER_SET_MEM_TABLE] = "VHOST_USER_SET_MEM_TABLE",
  53         [VHOST_USER_SET_LOG_BASE] = "VHOST_USER_SET_LOG_BASE",
  54         [VHOST_USER_SET_LOG_FD] = "VHOST_USER_SET_LOG_FD",
  55         [VHOST_USER_SET_VRING_NUM] = "VHOST_USER_SET_VRING_NUM",
  56         [VHOST_USER_SET_VRING_ADDR] = "VHOST_USER_SET_VRING_ADDR",
  57         [VHOST_USER_SET_VRING_BASE] = "VHOST_USER_SET_VRING_BASE",
  58         [VHOST_USER_GET_VRING_BASE] = "VHOST_USER_GET_VRING_BASE",
  59         [VHOST_USER_SET_VRING_KICK] = "VHOST_USER_SET_VRING_KICK",
  60         [VHOST_USER_SET_VRING_CALL] = "VHOST_USER_SET_VRING_CALL",
  61         [VHOST_USER_SET_VRING_ERR]  = "VHOST_USER_SET_VRING_ERR",
  62         [VHOST_USER_GET_PROTOCOL_FEATURES]  = "VHOST_USER_GET_PROTOCOL_FEATURES",
  63         [VHOST_USER_SET_PROTOCOL_FEATURES]  = "VHOST_USER_SET_PROTOCOL_FEATURES",
  64         [VHOST_USER_GET_QUEUE_NUM]  = "VHOST_USER_GET_QUEUE_NUM",
  65         [VHOST_USER_SET_VRING_ENABLE]  = "VHOST_USER_SET_VRING_ENABLE",
  66         [VHOST_USER_SEND_RARP]  = "VHOST_USER_SEND_RARP",
  67         [VHOST_USER_NET_SET_MTU]  = "VHOST_USER_NET_SET_MTU",
  68         [VHOST_USER_SET_SLAVE_REQ_FD]  = "VHOST_USER_SET_SLAVE_REQ_FD",
  69         [VHOST_USER_IOTLB_MSG]  = "VHOST_USER_IOTLB_MSG",
  70         [VHOST_USER_CRYPTO_CREATE_SESS] = "VHOST_USER_CRYPTO_CREATE_SESS",
  71         [VHOST_USER_CRYPTO_CLOSE_SESS] = "VHOST_USER_CRYPTO_CLOSE_SESS",
  72 };
  73
  74 static uint64_t
  75 get_blk_size(int fd)
  76 {
  77         struct stat stat;
  78         int ret;
  79
  80         ret = fstat(fd, &stat);
  81         return ret == -1 ? (uint64_t)-1 : (uint64_t)stat.st_blksize;
  82 }
  83
  84 static void
  85 free_mem_region(struct virtio_net *dev)
  86 {
  87         uint32_t i;
  88         struct rte_vhost_mem_region *reg;
  89
  90         if (!dev || !dev->mem)
  91                 return;
  92
  93         for (i = 0; i < dev->mem->nregions; i++) {
  94                 reg = &dev->mem->regions[i];
  95                 if (reg->host_user_addr) {
  96                         munmap(reg->mmap_addr, reg->mmap_size);
  97                         close(reg->fd);
  98                 }
  99         }
 100 }
 101
 102 void
 103 vhost_backend_cleanup(struct virtio_net *dev)
 104 {
 105         if (dev->mem) {
 106                 free_mem_region(dev);
 107                 rte_free(dev->mem);
 108                 dev->mem = NULL;
 109         }
 110
 111         free(dev->guest_pages);
 112         dev->guest_pages = NULL;
 113
 114         if (dev->log_addr) {
 115                 munmap((void *)(uintptr_t)dev->log_addr, dev->log_size);
 116                 dev->log_addr = 0;
 117         }
 118
 119         if (dev->slave_req_fd >= 0) {
 120                 close(dev->slave_req_fd);
 121                 dev->slave_req_fd = -1;
 122         }
 123 }
 124
 125 /*
 126  * This function just returns success at the moment unless
 127  * the device hasn't been initialised.
 128  */
 129 static int
 130 vhost_user_set_owner(void)
 131 {
 132         return 0;
 133 }
 134
 135 static int
 136 vhost_user_reset_owner(struct virtio_net *dev)
 137 {
 138         struct rte_vdpa_device *vdpa_dev;
 139         int did = -1;
 140
 141         if (dev->flags & VIRTIO_DEV_RUNNING) {
 142                 did = dev->vdpa_dev_id;
 143                 vdpa_dev = rte_vdpa_get_device(did);
 144                 if (vdpa_dev && vdpa_dev->ops->dev_close)
 145                         vdpa_dev->ops->dev_close(dev->vid);
 146                 dev->flags &= ~VIRTIO_DEV_RUNNING;
 147                 dev->notify_ops->destroy_device(dev->vid);
 148         }
 149
 150         cleanup_device(dev, 0);
 151         reset_device(dev);
 152         return 0;
 153 }
 154
 155 /*
 156  * The features that we support are requested.
 157  */
 158 static uint64_t
 159 vhost_user_get_features(struct virtio_net *dev)
 160 {
 161         uint64_t features = 0;
 162
 163         rte_vhost_driver_get_features(dev->ifname, &features);
 164         return features;
 165 }
 166
 167 /*
 168  * The queue number that we support are requested.
 169  */
 170 static uint32_t
 171 vhost_user_get_queue_num(struct virtio_net *dev)
 172 {
 173         uint32_t queue_num = 0;
 174
 175         rte_vhost_driver_get_queue_num(dev->ifname, &queue_num);
 176         return queue_num;
 177 }
 178
 179 /*
 180  * We receive the negotiated features supported by us and the virtio device.
 181  */
 182 static int
 183 vhost_user_set_features(struct virtio_net *dev, uint64_t features)
 184 {
 185         uint64_t vhost_features = 0;
 186         struct rte_vdpa_device *vdpa_dev;
 187         int did = -1;
 188
 189         rte_vhost_driver_get_features(dev->ifname, &vhost_features);
 190         if (features & ~vhost_features) {
 191                 RTE_LOG(ERR, VHOST_CONFIG,
 192                         "(%d) received invalid negotiated features.\n",
 193                         dev->vid);
 194                 return -1;
 195         }
 196
 197         if (dev->flags & VIRTIO_DEV_RUNNING) {
 198                 if (dev->features == features)
 199                         return 0;
 200
 201                 /*
 202                  * Error out if master tries to change features while device is
 203                  * in running state. The exception being VHOST_F_LOG_ALL, which
 204                  * is enabled when the live-migration starts.
 205                  */
 206                 if ((dev->features ^ features) & ~(1ULL << VHOST_F_LOG_ALL)) {
 207                         RTE_LOG(ERR, VHOST_CONFIG,
 208                                 "(%d) features changed while device is running.\n",
 209                                 dev->vid);
 210                         return -1;
 211                 }
 212
 213                 if (dev->notify_ops->features_changed)
 214                         dev->notify_ops->features_changed(dev->vid, features);
 215         }
 216
 217         did = dev->vdpa_dev_id;
 218         vdpa_dev = rte_vdpa_get_device(did);
 219         if (vdpa_dev && vdpa_dev->ops->set_features)
 220                 vdpa_dev->ops->set_features(dev->vid);
 221
 222         dev->features = features;
 223         if (dev->features &
 224                 ((1 << VIRTIO_NET_F_MRG_RXBUF) | (1ULL << VIRTIO_F_VERSION_1))) {
 225                 dev->vhost_hlen = sizeof(struct virtio_net_hdr_mrg_rxbuf);
 226         } else {
 227                 dev->vhost_hlen = sizeof(struct virtio_net_hdr);
 228         }
 229         VHOST_LOG_DEBUG(VHOST_CONFIG,
 230                 "(%d) mergeable RX buffers %s, virtio 1 %s\n",
 231                 dev->vid,
 232                 (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF)) ? "on" : "off",
 233                 (dev->features & (1ULL << VIRTIO_F_VERSION_1)) ? "on" : "off");
 234
 235         if ((dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET) &&
 236             !(dev->features & (1ULL << VIRTIO_NET_F_MQ))) {
 237                 /*
 238                  * Remove all but first queue pair if MQ hasn't been
 239                  * negotiated. This is safe because the device is not
 240                  * running at this stage.
 241                  */
 242                 while (dev->nr_vring > 2) {
 243                         struct vhost_virtqueue *vq;
 244
 245                         vq = dev->virtqueue[--dev->nr_vring];
 246                         if (!vq)
 247                                 continue;
 248
 249                         dev->virtqueue[dev->nr_vring] = NULL;
 250                         cleanup_vq(vq, 1);
 251                         free_vq(vq);
 252                 }
 253         }
 254
 255         return 0;
 256 }
 257
 258 /*
 259  * The virtio device sends us the size of the descriptor ring.
 260  */
 261 static int
 262 vhost_user_set_vring_num(struct virtio_net *dev,
 263                          VhostUserMsg *msg)
 264 {
 265         struct vhost_virtqueue *vq = dev->virtqueue[msg->payload.state.index];
 266
 267         vq->size = msg->payload.state.num;
 268
 269         /* VIRTIO 1.0, 2.4 Virtqueues says:
 270          *
 271          *   Queue Size value is always a power of 2. The maximum Queue Size
 272          *   value is 32768.
 273          */
 274         if ((vq->size & (vq->size - 1)) || vq->size > 32768) {
 275                 RTE_LOG(ERR, VHOST_CONFIG,
 276                         "invalid virtqueue size %u\n", vq->size);
 277                 return -1;
 278         }
 279
 280         if (dev->dequeue_zero_copy) {
 281                 vq->nr_zmbuf = 0;
 282                 vq->last_zmbuf_idx = 0;
 283                 vq->zmbuf_size = vq->size;
 284                 vq->zmbufs = rte_zmalloc(NULL, vq->zmbuf_size *
 285                                          sizeof(struct zcopy_mbuf), 0);
 286                 if (vq->zmbufs == NULL) {
 287                         RTE_LOG(WARNING, VHOST_CONFIG,
 288                                 "failed to allocate mem for zero copy; "
 289                                 "zero copy is force disabled\n");
 290                         dev->dequeue_zero_copy = 0;
 291                 }
 292                 TAILQ_INIT(&vq->zmbuf_list);
 293         }
 294
 295         vq->shadow_used_ring = rte_malloc(NULL,
 296                                 vq->size * sizeof(struct vring_used_elem),
 297                                 RTE_CACHE_LINE_SIZE);
 298         if (!vq->shadow_used_ring) {
 299                 RTE_LOG(ERR, VHOST_CONFIG,
 300                         "failed to allocate memory for shadow used ring.\n");
 301                 return -1;
 302         }
 303
 304         vq->batch_copy_elems = rte_malloc(NULL,
 305                                 vq->size * sizeof(struct batch_copy_elem),
 306                                 RTE_CACHE_LINE_SIZE);
 307         if (!vq->batch_copy_elems) {
 308                 RTE_LOG(ERR, VHOST_CONFIG,
 309                         "failed to allocate memory for batching copy.\n");
 310                 return -1;
 311         }
 312
 313         return 0;
 314 }
 315
 316 /*
 317  * Reallocate virtio_dev and vhost_virtqueue data structure to make them on the
 318  * same numa node as the memory of vring descriptor.
 319  */
 320 #ifdef RTE_LIBRTE_VHOST_NUMA
 321 static struct virtio_net*
 322 numa_realloc(struct virtio_net *dev, int index)
 323 {
 324         int oldnode, newnode;
 325         struct virtio_net *old_dev;
 326         struct vhost_virtqueue *old_vq, *vq;
 327         struct zcopy_mbuf *new_zmbuf;
 328         struct vring_used_elem *new_shadow_used_ring;
 329         struct batch_copy_elem *new_batch_copy_elems;
 330         int ret;
 331
 332         old_dev = dev;
 333         vq = old_vq = dev->virtqueue[index];
 334
 335         ret = get_mempolicy(&newnode, NULL, 0, old_vq->desc,
 336                             MPOL_F_NODE | MPOL_F_ADDR);
 337
 338         /* check if we need to reallocate vq */
 339         ret |= get_mempolicy(&oldnode, NULL, 0, old_vq,
 340                              MPOL_F_NODE | MPOL_F_ADDR);
 341         if (ret) {
 342                 RTE_LOG(ERR, VHOST_CONFIG,
 343                         "Unable to get vq numa information.\n");
 344                 return dev;
 345         }
 346         if (oldnode != newnode) {
 347                 RTE_LOG(INFO, VHOST_CONFIG,
 348                         "reallocate vq from %d to %d node\n", oldnode, newnode);
 349                 vq = rte_malloc_socket(NULL, sizeof(*vq), 0, newnode);
 350                 if (!vq)
 351                         return dev;
 352
 353                 memcpy(vq, old_vq, sizeof(*vq));
 354                 TAILQ_INIT(&vq->zmbuf_list);
 355
 356                 new_zmbuf = rte_malloc_socket(NULL, vq->zmbuf_size *
 357                         sizeof(struct zcopy_mbuf), 0, newnode);
 358                 if (new_zmbuf) {
 359                         rte_free(vq->zmbufs);
 360                         vq->zmbufs = new_zmbuf;
 361                 }
 362
 363                 new_shadow_used_ring = rte_malloc_socket(NULL,
 364                         vq->size * sizeof(struct vring_used_elem),
 365                         RTE_CACHE_LINE_SIZE,
 366                         newnode);
 367                 if (new_shadow_used_ring) {
 368                         rte_free(vq->shadow_used_ring);
 369                         vq->shadow_used_ring = new_shadow_used_ring;
 370                 }
 371
 372                 new_batch_copy_elems = rte_malloc_socket(NULL,
 373                         vq->size * sizeof(struct batch_copy_elem),
 374                         RTE_CACHE_LINE_SIZE,
 375                         newnode);
 376                 if (new_batch_copy_elems) {
 377                         rte_free(vq->batch_copy_elems);
 378                         vq->batch_copy_elems = new_batch_copy_elems;
 379                 }
 380
 381                 rte_free(old_vq);
 382         }
 383
 384         /* check if we need to reallocate dev */
 385         ret = get_mempolicy(&oldnode, NULL, 0, old_dev,
 386                             MPOL_F_NODE | MPOL_F_ADDR);
 387         if (ret) {
 388                 RTE_LOG(ERR, VHOST_CONFIG,
 389                         "Unable to get dev numa information.\n");
 390                 goto out;
 391         }
 392         if (oldnode != newnode) {
 393                 RTE_LOG(INFO, VHOST_CONFIG,
 394                         "reallocate dev from %d to %d node\n",
 395                         oldnode, newnode);
 396                 dev = rte_malloc_socket(NULL, sizeof(*dev), 0, newnode);
 397                 if (!dev) {
 398                         dev = old_dev;
 399                         goto out;
 400                 }
 401
 402                 memcpy(dev, old_dev, sizeof(*dev));
 403                 rte_free(old_dev);
 404         }
 405
 406 out:
 407         dev->virtqueue[index] = vq;
 408         vhost_devices[dev->vid] = dev;
 409
 410         if (old_vq != vq)
 411                 vhost_user_iotlb_init(dev, index);
 412
 413         return dev;
 414 }
 415 #else
 416 static struct virtio_net*
 417 numa_realloc(struct virtio_net *dev, int index __rte_unused)
 418 {
 419         return dev;
 420 }
 421 #endif
 422
 423 /* Converts QEMU virtual address to Vhost virtual address. */
 424 static uint64_t
 425 qva_to_vva(struct virtio_net *dev, uint64_t qva)
 426 {
 427         struct rte_vhost_mem_region *reg;
 428         uint32_t i;
 429
 430         /* Find the region where the address lives. */
 431         for (i = 0; i < dev->mem->nregions; i++) {
 432                 reg = &dev->mem->regions[i];
 433
 434                 if (qva >= reg->guest_user_addr &&
 435                     qva <  reg->guest_user_addr + reg->size) {
 436                         return qva - reg->guest_user_addr +
 437                                reg->host_user_addr;
 438                 }
 439         }
 440
 441         return 0;
 442 }
 443
 444
 445 /*
 446  * Converts ring address to Vhost virtual address.
 447  * If IOMMU is enabled, the ring address is a guest IO virtual address,
 448  * else it is a QEMU virtual address.
 449  */
 450 static uint64_t
 451 ring_addr_to_vva(struct virtio_net *dev, struct vhost_virtqueue *vq,
 452                 uint64_t ra, uint64_t size)
 453 {
 454         if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) {
 455                 uint64_t vva;
 456
 457                 vva = vhost_user_iotlb_cache_find(vq, ra,
 458                                         &size, VHOST_ACCESS_RW);
 459                 if (!vva)
 460                         vhost_user_iotlb_miss(dev, ra, VHOST_ACCESS_RW);
 461
 462                 return vva;
 463         }
 464
 465         return qva_to_vva(dev, ra);
 466 }
 467
 468 static struct virtio_net *
 469 translate_ring_addresses(struct virtio_net *dev, int vq_index)
 470 {
 471         struct vhost_virtqueue *vq = dev->virtqueue[vq_index];
 472         struct vhost_vring_addr *addr = &vq->ring_addrs;
 473
 474         /* The addresses are converted from QEMU virtual to Vhost virtual. */
 475         if (vq->desc && vq->avail && vq->used)
 476                 return dev;
 477
 478         vq->desc = (struct vring_desc *)(uintptr_t)ring_addr_to_vva(dev,
 479                         vq, addr->desc_user_addr, sizeof(struct vring_desc));
 480         if (vq->desc == 0) {
 481                 RTE_LOG(DEBUG, VHOST_CONFIG,
 482                         "(%d) failed to find desc ring address.\n",
 483                         dev->vid);
 484                 return dev;
 485         }
 486
 487         dev = numa_realloc(dev, vq_index);
 488         vq = dev->virtqueue[vq_index];
 489         addr = &vq->ring_addrs;
 490
 491         vq->avail = (struct vring_avail *)(uintptr_t)ring_addr_to_vva(dev,
 492                         vq, addr->avail_user_addr, sizeof(struct vring_avail));
 493         if (vq->avail == 0) {
 494                 RTE_LOG(DEBUG, VHOST_CONFIG,
 495                         "(%d) failed to find avail ring address.\n",
 496                         dev->vid);
 497                 return dev;
 498         }
 499
 500         vq->used = (struct vring_used *)(uintptr_t)ring_addr_to_vva(dev,
 501                         vq, addr->used_user_addr, sizeof(struct vring_used));
 502         if (vq->used == 0) {
 503                 RTE_LOG(DEBUG, VHOST_CONFIG,
 504                         "(%d) failed to find used ring address.\n",
 505                         dev->vid);
 506                 return dev;
 507         }
 508
 509         if (vq->last_used_idx != vq->used->idx) {
 510                 RTE_LOG(WARNING, VHOST_CONFIG,
 511                         "last_used_idx (%u) and vq->used->idx (%u) mismatches; "
 512                         "some packets maybe resent for Tx and dropped for Rx\n",
 513                         vq->last_used_idx, vq->used->idx);
 514                 vq->last_used_idx  = vq->used->idx;
 515                 vq->last_avail_idx = vq->used->idx;
 516         }
 517
 518         vq->log_guest_addr = addr->log_guest_addr;
 519
 520         VHOST_LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address desc: %p\n",
 521                         dev->vid, vq->desc);
 522         VHOST_LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address avail: %p\n",
 523                         dev->vid, vq->avail);
 524         VHOST_LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address used: %p\n",
 525                         dev->vid, vq->used);
 526         VHOST_LOG_DEBUG(VHOST_CONFIG, "(%d) log_guest_addr: %" PRIx64 "\n",
 527                         dev->vid, vq->log_guest_addr);
 528
 529         return dev;
 530 }
 531
 532 /*
 533  * The virtio device sends us the desc, used and avail ring addresses.
 534  * This function then converts these to our address space.
 535  */
 536 static int
 537 vhost_user_set_vring_addr(struct virtio_net **pdev, VhostUserMsg *msg)
 538 {
 539         struct vhost_virtqueue *vq;
 540         struct vhost_vring_addr *addr = &msg->payload.addr;
 541         struct virtio_net *dev = *pdev;
 542
 543         if (dev->mem == NULL)
 544                 return -1;
 545
 546         /* addr->index refers to the queue index. The txq 1, rxq is 0. */
 547         vq = dev->virtqueue[msg->payload.addr.index];
 548
 549         /*
 550          * Rings addresses should not be interpreted as long as the ring is not
 551          * started and enabled
 552          */
 553         memcpy(&vq->ring_addrs, addr, sizeof(*addr));
 554
 555         vring_invalidate(dev, vq);
 556
 557         if (vq->enabled && (dev->features &
 558                                 (1ULL << VHOST_USER_F_PROTOCOL_FEATURES))) {
 559                 dev = translate_ring_addresses(dev, msg->payload.addr.index);
 560                 if (!dev)
 561                         return -1;
 562
 563                 *pdev = dev;
 564         }
 565
 566         return 0;
 567 }
 568
 569 /*
 570  * The virtio device sends us the available ring last used index.
 571  */
 572 static int
 573 vhost_user_set_vring_base(struct virtio_net *dev,
 574                           VhostUserMsg *msg)
 575 {
 576         dev->virtqueue[msg->payload.state.index]->last_used_idx  =
 577                         msg->payload.state.num;
 578         dev->virtqueue[msg->payload.state.index]->last_avail_idx =
 579                         msg->payload.state.num;
 580
 581         return 0;
 582 }
 583
 584 static int
 585 add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,
 586                    uint64_t host_phys_addr, uint64_t size)
 587 {
 588         struct guest_page *page, *last_page;
 589
 590         if (dev->nr_guest_pages == dev->max_guest_pages) {
 591                 dev->max_guest_pages *= 2;
 592                 dev->guest_pages = realloc(dev->guest_pages,
 593                                         dev->max_guest_pages * sizeof(*page));
 594                 if (!dev->guest_pages) {
 595                         RTE_LOG(ERR, VHOST_CONFIG, "cannot realloc guest_pages\n");
 596                         return -1;
 597                 }
 598         }
 599
 600         if (dev->nr_guest_pages > 0) {
 601                 last_page = &dev->guest_pages[dev->nr_guest_pages - 1];
 602                 /* merge if the two pages are continuous */
 603                 if (host_phys_addr == last_page->host_phys_addr +
 604                                       last_page->size) {
 605                         last_page->size += size;
 606                         return 0;
 607                 }
 608         }
 609
 610         page = &dev->guest_pages[dev->nr_guest_pages++];
 611         page->guest_phys_addr = guest_phys_addr;
 612         page->host_phys_addr  = host_phys_addr;
 613         page->size = size;
 614
 615         return 0;
 616 }
 617
 618 static int
 619 add_guest_pages(struct virtio_net *dev, struct rte_vhost_mem_region *reg,
 620                 uint64_t page_size)
 621 {
 622         uint64_t reg_size = reg->size;
 623         uint64_t host_user_addr  = reg->host_user_addr;
 624         uint64_t guest_phys_addr = reg->guest_phys_addr;
 625         uint64_t host_phys_addr;
 626         uint64_t size;
 627
 628         host_phys_addr = rte_mem_virt2iova((void *)(uintptr_t)host_user_addr);
 629         size = page_size - (guest_phys_addr & (page_size - 1));
 630         size = RTE_MIN(size, reg_size);
 631
 632         if (add_one_guest_page(dev, guest_phys_addr, host_phys_addr, size) < 0)
 633                 return -1;
 634
 635         host_user_addr  += size;
 636         guest_phys_addr += size;
 637         reg_size -= size;
 638
 639         while (reg_size > 0) {
 640                 size = RTE_MIN(reg_size, page_size);
 641                 host_phys_addr = rte_mem_virt2iova((void *)(uintptr_t)
 642                                                   host_user_addr);
 643                 if (add_one_guest_page(dev, guest_phys_addr, host_phys_addr,
 644                                 size) < 0)
 645                         return -1;
 646
 647                 host_user_addr  += size;
 648                 guest_phys_addr += size;
 649                 reg_size -= size;
 650         }
 651
 652         return 0;
 653 }
 654
 655 #ifdef RTE_LIBRTE_VHOST_DEBUG
 656 /* TODO: enable it only in debug mode? */
 657 static void
 658 dump_guest_pages(struct virtio_net *dev)
 659 {
 660         uint32_t i;
 661         struct guest_page *page;
 662
 663         for (i = 0; i < dev->nr_guest_pages; i++) {
 664                 page = &dev->guest_pages[i];
 665
 666                 RTE_LOG(INFO, VHOST_CONFIG,
 667                         "guest physical page region %u\n"
 668                         "\t guest_phys_addr: %" PRIx64 "\n"
 669                         "\t host_phys_addr : %" PRIx64 "\n"
 670                         "\t size           : %" PRIx64 "\n",
 671                         i,
 672                         page->guest_phys_addr,
 673                         page->host_phys_addr,
 674                         page->size);
 675         }
 676 }
 677 #else
 678 #define dump_guest_pages(dev)
 679 #endif
 680
 681 static bool
 682 vhost_memory_changed(struct VhostUserMemory *new,
 683                      struct rte_vhost_memory *old)
 684 {
 685         uint32_t i;
 686
 687         if (new->nregions != old->nregions)
 688                 return true;
 689
 690         for (i = 0; i < new->nregions; ++i) {
 691                 VhostUserMemoryRegion *new_r = &new->regions[i];
 692                 struct rte_vhost_mem_region *old_r = &old->regions[i];
 693
 694                 if (new_r->guest_phys_addr != old_r->guest_phys_addr)
 695                         return true;
 696                 if (new_r->memory_size != old_r->size)
 697                         return true;
 698                 if (new_r->userspace_addr != old_r->guest_user_addr)
 699                         return true;
 700         }
 701
 702         return false;
 703 }
 704
 705 static int
 706 vhost_user_set_mem_table(struct virtio_net *dev, struct VhostUserMsg *pmsg)
 707 {
 708         struct VhostUserMemory memory = pmsg->payload.memory;
 709         struct rte_vhost_mem_region *reg;
 710         void *mmap_addr;
 711         uint64_t mmap_size;
 712         uint64_t mmap_offset;
 713         uint64_t alignment;
 714         uint32_t i;
 715         int populate;
 716         int fd;
 717
 718         if (memory.nregions > VHOST_MEMORY_MAX_NREGIONS) {
 719                 RTE_LOG(ERR, VHOST_CONFIG,
 720                         "too many memory regions (%u)\n", memory.nregions);
 721                 return -1;
 722         }
 723
 724         if (dev->mem && !vhost_memory_changed(&memory, dev->mem)) {
 725                 RTE_LOG(INFO, VHOST_CONFIG,
 726                         "(%d) memory regions not changed\n", dev->vid);
 727
 728                 for (i = 0; i < memory.nregions; i++)
 729                         close(pmsg->fds[i]);
 730
 731                 return 0;
 732         }
 733
 734         if (dev->mem) {
 735                 free_mem_region(dev);
 736                 rte_free(dev->mem);
 737                 dev->mem = NULL;
 738         }
 739
 740         dev->nr_guest_pages = 0;
 741         if (!dev->guest_pages) {
 742                 dev->max_guest_pages = 8;
 743                 dev->guest_pages = malloc(dev->max_guest_pages *
 744                                                 sizeof(struct guest_page));
 745                 if (dev->guest_pages == NULL) {
 746                         RTE_LOG(ERR, VHOST_CONFIG,
 747                                 "(%d) failed to allocate memory "
 748                                 "for dev->guest_pages\n",
 749                                 dev->vid);
 750                         return -1;
 751                 }
 752         }
 753
 754         dev->mem = rte_zmalloc("vhost-mem-table", sizeof(struct rte_vhost_memory) +
 755                 sizeof(struct rte_vhost_mem_region) * memory.nregions, 0);
 756         if (dev->mem == NULL) {
 757                 RTE_LOG(ERR, VHOST_CONFIG,
 758                         "(%d) failed to allocate memory for dev->mem\n",
 759                         dev->vid);
 760                 return -1;
 761         }
 762         dev->mem->nregions = memory.nregions;
 763
 764         for (i = 0; i < memory.nregions; i++) {
 765                 fd  = pmsg->fds[i];
 766                 reg = &dev->mem->regions[i];
 767
 768                 reg->guest_phys_addr = memory.regions[i].guest_phys_addr;
 769                 reg->guest_user_addr = memory.regions[i].userspace_addr;
 770                 reg->size            = memory.regions[i].memory_size;
 771                 reg->fd              = fd;
 772
 773                 mmap_offset = memory.regions[i].mmap_offset;
 774
 775                 /* Check for memory_size + mmap_offset overflow */
 776                 if (mmap_offset >= -reg->size) {
 777                         RTE_LOG(ERR, VHOST_CONFIG,
 778                                 "mmap_offset (%#"PRIx64") and memory_size "
 779                                 "(%#"PRIx64") overflow\n",
 780                                 mmap_offset, reg->size);
 781                         goto err_mmap;
 782                 }
 783
 784                 mmap_size = reg->size + mmap_offset;
 785
 786                 /* mmap() without flag of MAP_ANONYMOUS, should be called
 787                  * with length argument aligned with hugepagesz at older
 788                  * longterm version Linux, like 2.6.32 and 3.2.72, or
 789                  * mmap() will fail with EINVAL.
 790                  *
 791                  * to avoid failure, make sure in caller to keep length
 792                  * aligned.
 793                  */
 794                 alignment = get_blk_size(fd);
 795                 if (alignment == (uint64_t)-1) {
 796                         RTE_LOG(ERR, VHOST_CONFIG,
 797                                 "couldn't get hugepage size through fstat\n");
 798                         goto err_mmap;
 799                 }
 800                 mmap_size = RTE_ALIGN_CEIL(mmap_size, alignment);
 801
 802                 populate = (dev->dequeue_zero_copy) ? MAP_POPULATE : 0;
 803                 mmap_addr = mmap(NULL, mmap_size, PROT_READ | PROT_WRITE,
 804                                  MAP_SHARED | populate, fd, 0);
 805
 806                 if (mmap_addr == MAP_FAILED) {
 807                         RTE_LOG(ERR, VHOST_CONFIG,
 808                                 "mmap region %u failed.\n", i);
 809                         goto err_mmap;
 810                 }
 811
 812                 reg->mmap_addr = mmap_addr;
 813                 reg->mmap_size = mmap_size;
 814                 reg->host_user_addr = (uint64_t)(uintptr_t)mmap_addr +
 815                                       mmap_offset;
 816
 817                 if (dev->dequeue_zero_copy)
 818                         if (add_guest_pages(dev, reg, alignment) < 0) {
 819                                 RTE_LOG(ERR, VHOST_CONFIG,
 820                                         "adding guest pages to region %u failed.\n",
 821                                         i);
 822                                 goto err_mmap;
 823                         }
 824
 825                 RTE_LOG(INFO, VHOST_CONFIG,
 826                         "guest memory region %u, size: 0x%" PRIx64 "\n"
 827                         "\t guest physical addr: 0x%" PRIx64 "\n"
 828                         "\t guest virtual  addr: 0x%" PRIx64 "\n"
 829                         "\t host  virtual  addr: 0x%" PRIx64 "\n"
 830                         "\t mmap addr : 0x%" PRIx64 "\n"
 831                         "\t mmap size : 0x%" PRIx64 "\n"
 832                         "\t mmap align: 0x%" PRIx64 "\n"
 833                         "\t mmap off  : 0x%" PRIx64 "\n",
 834                         i, reg->size,
 835                         reg->guest_phys_addr,
 836                         reg->guest_user_addr,
 837                         reg->host_user_addr,
 838                         (uint64_t)(uintptr_t)mmap_addr,
 839                         mmap_size,
 840                         alignment,
 841                         mmap_offset);
 842         }
 843
 844         dump_guest_pages(dev);
 845
 846         return 0;
 847
 848 err_mmap:
 849         free_mem_region(dev);
 850         rte_free(dev->mem);
 851         dev->mem = NULL;
 852         return -1;
 853 }
 854
 855 static int
 856 vq_is_ready(struct vhost_virtqueue *vq)
 857 {
 858         return vq && vq->desc && vq->avail && vq->used &&
 859                vq->kickfd != VIRTIO_UNINITIALIZED_EVENTFD &&
 860                vq->callfd != VIRTIO_UNINITIALIZED_EVENTFD;
 861 }
 862
 863 static int
 864 virtio_is_ready(struct virtio_net *dev)
 865 {
 866         struct vhost_virtqueue *vq;
 867         uint32_t i;
 868
 869         if (dev->nr_vring == 0)
 870                 return 0;
 871
 872         for (i = 0; i < dev->nr_vring; i++) {
 873                 vq = dev->virtqueue[i];
 874
 875                 if (!vq_is_ready(vq))
 876                         return 0;
 877         }
 878
 879         RTE_LOG(INFO, VHOST_CONFIG,
 880                 "virtio is now ready for processing.\n");
 881         return 1;
 882 }
 883
 884 static void
 885 vhost_user_set_vring_call(struct virtio_net *dev, struct VhostUserMsg *pmsg)
 886 {
 887         struct vhost_vring_file file;
 888         struct vhost_virtqueue *vq;
 889
 890         file.index = pmsg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
 891         if (pmsg->payload.u64 & VHOST_USER_VRING_NOFD_MASK)
 892                 file.fd = VIRTIO_INVALID_EVENTFD;
 893         else
 894                 file.fd = pmsg->fds[0];
 895         RTE_LOG(INFO, VHOST_CONFIG,
 896                 "vring call idx:%d file:%d\n", file.index, file.fd);
 897
 898         vq = dev->virtqueue[file.index];
 899         if (vq->callfd >= 0)
 900                 close(vq->callfd);
 901
 902         vq->callfd = file.fd;
 903 }
 904
 905 static void
 906 vhost_user_set_vring_kick(struct virtio_net **pdev, struct VhostUserMsg *pmsg)
 907 {
 908         struct vhost_vring_file file;
 909         struct vhost_virtqueue *vq;
 910         struct virtio_net *dev = *pdev;
 911
 912         file.index = pmsg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
 913         if (pmsg->payload.u64 & VHOST_USER_VRING_NOFD_MASK)
 914                 file.fd = VIRTIO_INVALID_EVENTFD;
 915         else
 916                 file.fd = pmsg->fds[0];
 917         RTE_LOG(INFO, VHOST_CONFIG,
 918                 "vring kick idx:%d file:%d\n", file.index, file.fd);
 919
 920         /* Interpret ring addresses only when ring is started. */
 921         dev = translate_ring_addresses(dev, file.index);
 922         if (!dev)
 923                 return;
 924
 925         *pdev = dev;
 926
 927         vq = dev->virtqueue[file.index];
 928
 929         /*
 930          * When VHOST_USER_F_PROTOCOL_FEATURES is not negotiated,
 931          * the ring starts already enabled. Otherwise, it is enabled via
 932          * the SET_VRING_ENABLE message.
 933          */
 934         if (!(dev->features & (1ULL << VHOST_USER_F_PROTOCOL_FEATURES)))
 935                 vq->enabled = 1;
 936
 937         if (vq->kickfd >= 0)
 938                 close(vq->kickfd);
 939         vq->kickfd = file.fd;
 940 }
 941
 942 static void
 943 free_zmbufs(struct vhost_virtqueue *vq)
 944 {
 945         struct zcopy_mbuf *zmbuf, *next;
 946
 947         for (zmbuf = TAILQ_FIRST(&vq->zmbuf_list);
 948              zmbuf != NULL; zmbuf = next) {
 949                 next = TAILQ_NEXT(zmbuf, next);
 950
 951                 rte_pktmbuf_free(zmbuf->mbuf);
 952                 TAILQ_REMOVE(&vq->zmbuf_list, zmbuf, next);
 953         }
 954
 955         rte_free(vq->zmbufs);
 956 }
 957
 958 /*
 959  * when virtio is stopped, qemu will send us the GET_VRING_BASE message.
 960  */
 961 static int
 962 vhost_user_get_vring_base(struct virtio_net *dev,
 963                           VhostUserMsg *msg)
 964 {
 965         struct vhost_virtqueue *vq = dev->virtqueue[msg->payload.state.index];
 966         struct rte_vdpa_device *vdpa_dev;
 967         int did = -1;
 968
 969         /* We have to stop the queue (virtio) if it is running. */
 970         if (dev->flags & VIRTIO_DEV_RUNNING) {
 971                 did = dev->vdpa_dev_id;
 972                 vdpa_dev = rte_vdpa_get_device(did);
 973                 if (vdpa_dev && vdpa_dev->ops->dev_close)
 974                         vdpa_dev->ops->dev_close(dev->vid);
 975                 dev->flags &= ~VIRTIO_DEV_RUNNING;
 976                 dev->notify_ops->destroy_device(dev->vid);
 977         }
 978
 979         dev->flags &= ~VIRTIO_DEV_READY;
 980         dev->flags &= ~VIRTIO_DEV_VDPA_CONFIGURED;
 981
 982         /* Here we are safe to get the last avail index */
 983         msg->payload.state.num = vq->last_avail_idx;
 984
 985         RTE_LOG(INFO, VHOST_CONFIG,
 986                 "vring base idx:%d file:%d\n", msg->payload.state.index,
 987                 msg->payload.state.num);
 988         /*
 989          * Based on current qemu vhost-user implementation, this message is
 990          * sent and only sent in vhost_vring_stop.
 991          * TODO: cleanup the vring, it isn't usable since here.
 992          */
 993         if (vq->kickfd >= 0)
 994                 close(vq->kickfd);
 995
 996         vq->kickfd = VIRTIO_UNINITIALIZED_EVENTFD;
 997
 998         if (vq->callfd >= 0)
 999                 close(vq->callfd);
1000
1001         vq->callfd = VIRTIO_UNINITIALIZED_EVENTFD;
1002
1003         if (dev->dequeue_zero_copy)
1004                 free_zmbufs(vq);
1005         rte_free(vq->shadow_used_ring);
1006         vq->shadow_used_ring = NULL;
1007
1008         rte_free(vq->batch_copy_elems);
1009         vq->batch_copy_elems = NULL;
1010
1011         return 0;
1012 }
1013
1014 /*
1015  * when virtio queues are ready to work, qemu will send us to
1016  * enable the virtio queue pair.
1017  */
1018 static int
1019 vhost_user_set_vring_enable(struct virtio_net *dev,
1020                             VhostUserMsg *msg)
1021 {
1022         int enable = (int)msg->payload.state.num;
1023         int index = (int)msg->payload.state.index;
1024         struct rte_vdpa_device *vdpa_dev;
1025         int did = -1;
1026
1027         RTE_LOG(INFO, VHOST_CONFIG,
1028                 "set queue enable: %d to qp idx: %d\n",
1029                 enable, index);
1030
1031         did = dev->vdpa_dev_id;
1032         vdpa_dev = rte_vdpa_get_device(did);
1033         if (vdpa_dev && vdpa_dev->ops->set_vring_state)
1034                 vdpa_dev->ops->set_vring_state(dev->vid, index, enable);
1035
1036         if (dev->notify_ops->vring_state_changed)
1037                 dev->notify_ops->vring_state_changed(dev->vid,
1038                                 index, enable);
1039
1040         dev->virtqueue[index]->enabled = enable;
1041
1042         return 0;
1043 }
1044
1045 static void
1046 vhost_user_get_protocol_features(struct virtio_net *dev,
1047                                  struct VhostUserMsg *msg)
1048 {
1049         uint64_t features, protocol_features;
1050
1051         rte_vhost_driver_get_features(dev->ifname, &features);
1052         rte_vhost_driver_get_protocol_features(dev->ifname, &protocol_features);
1053
1054         /*
1055          * REPLY_ACK protocol feature is only mandatory for now
1056          * for IOMMU feature. If IOMMU is explicitly disabled by the
1057          * application, disable also REPLY_ACK feature for older buggy
1058          * Qemu versions (from v2.7.0 to v2.9.0).
1059          */
1060         if (!(features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)))
1061                 protocol_features &= ~(1ULL << VHOST_USER_PROTOCOL_F_REPLY_ACK);
1062
1063         msg->payload.u64 = protocol_features;
1064         msg->size = sizeof(msg->payload.u64);
1065 }
1066
1067 static void
1068 vhost_user_set_protocol_features(struct virtio_net *dev,
1069                                  uint64_t protocol_features)
1070 {
1071         if (protocol_features & ~VHOST_USER_PROTOCOL_FEATURES)
1072                 return;
1073
1074         dev->protocol_features = protocol_features;
1075 }
1076
1077 static int
1078 vhost_user_set_log_base(struct virtio_net *dev, struct VhostUserMsg *msg)
1079 {
1080         int fd = msg->fds[0];
1081         uint64_t size, off;
1082         void *addr;
1083
1084         if (fd < 0) {
1085                 RTE_LOG(ERR, VHOST_CONFIG, "invalid log fd: %d\n", fd);
1086                 return -1;
1087         }
1088
1089         if (msg->size != sizeof(VhostUserLog)) {
1090                 RTE_LOG(ERR, VHOST_CONFIG,
1091                         "invalid log base msg size: %"PRId32" != %d\n",
1092                         msg->size, (int)sizeof(VhostUserLog));
1093                 return -1;
1094         }
1095
1096         size = msg->payload.log.mmap_size;
1097         off  = msg->payload.log.mmap_offset;
1098
1099         /* Don't allow mmap_offset to point outside the mmap region */
1100         if (off > size) {
1101                 RTE_LOG(ERR, VHOST_CONFIG,
1102                         "log offset %#"PRIx64" exceeds log size %#"PRIx64"\n",
1103                         off, size);
1104                 return -1;
1105         }
1106
1107         RTE_LOG(INFO, VHOST_CONFIG,
1108                 "log mmap size: %"PRId64", offset: %"PRId64"\n",
1109                 size, off);
1110
1111         /*
1112          * mmap from 0 to workaround a hugepage mmap bug: mmap will
1113          * fail when offset is not page size aligned.
1114          */
1115         addr = mmap(0, size + off, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
1116         close(fd);
1117         if (addr == MAP_FAILED) {
1118                 RTE_LOG(ERR, VHOST_CONFIG, "mmap log base failed!\n");
1119                 return -1;
1120         }
1121
1122         /*
1123          * Free previously mapped log memory on occasionally
1124          * multiple VHOST_USER_SET_LOG_BASE.
1125          */
1126         if (dev->log_addr) {
1127                 munmap((void *)(uintptr_t)dev->log_addr, dev->log_size);
1128         }
1129         dev->log_addr = (uint64_t)(uintptr_t)addr;
1130         dev->log_base = dev->log_addr + off;
1131         dev->log_size = size;
1132
1133         return 0;
1134 }
1135
1136 /*
1137  * An rarp packet is constructed and broadcasted to notify switches about
1138  * the new location of the migrated VM, so that packets from outside will
1139  * not be lost after migration.
1140  *
1141  * However, we don't actually "send" a rarp packet here, instead, we set
1142  * a flag 'broadcast_rarp' to let rte_vhost_dequeue_burst() inject it.
1143  */
1144 static int
1145 vhost_user_send_rarp(struct virtio_net *dev, struct VhostUserMsg *msg)
1146 {
1147         uint8_t *mac = (uint8_t *)&msg->payload.u64;
1148         struct rte_vdpa_device *vdpa_dev;
1149         int did = -1;
1150
1151         RTE_LOG(DEBUG, VHOST_CONFIG,
1152                 ":: mac: %02x:%02x:%02x:%02x:%02x:%02x\n",
1153                 mac[0], mac[1], mac[2], mac[3], mac[4], mac[5]);
1154         memcpy(dev->mac.addr_bytes, mac, 6);
1155
1156         /*
1157          * Set the flag to inject a RARP broadcast packet at
1158          * rte_vhost_dequeue_burst().
1159          *
1160          * rte_smp_wmb() is for making sure the mac is copied
1161          * before the flag is set.
1162          */
1163         rte_smp_wmb();
1164         rte_atomic16_set(&dev->broadcast_rarp, 1);
1165         did = dev->vdpa_dev_id;
1166         vdpa_dev = rte_vdpa_get_device(did);
1167         if (vdpa_dev && vdpa_dev->ops->migration_done)
1168                 vdpa_dev->ops->migration_done(dev->vid);
1169
1170         return 0;
1171 }
1172
1173 static int
1174 vhost_user_net_set_mtu(struct virtio_net *dev, struct VhostUserMsg *msg)
1175 {
1176         if (msg->payload.u64 < VIRTIO_MIN_MTU ||
1177                         msg->payload.u64 > VIRTIO_MAX_MTU) {
1178                 RTE_LOG(ERR, VHOST_CONFIG, "Invalid MTU size (%"PRIu64")\n",
1179                                 msg->payload.u64);
1180
1181                 return -1;
1182         }
1183
1184         dev->mtu = msg->payload.u64;
1185
1186         return 0;
1187 }
1188
1189 static int
1190 vhost_user_set_req_fd(struct virtio_net *dev, struct VhostUserMsg *msg)
1191 {
1192         int fd = msg->fds[0];
1193
1194         if (fd < 0) {
1195                 RTE_LOG(ERR, VHOST_CONFIG,
1196                                 "Invalid file descriptor for slave channel (%d)\n",
1197                                 fd);
1198                 return -1;
1199         }
1200
1201         dev->slave_req_fd = fd;
1202
1203         return 0;
1204 }
1205
1206 static int
1207 is_vring_iotlb_update(struct vhost_virtqueue *vq, struct vhost_iotlb_msg *imsg)
1208 {
1209         struct vhost_vring_addr *ra;
1210         uint64_t start, end;
1211
1212         start = imsg->iova;
1213         end = start + imsg->size;
1214
1215         ra = &vq->ring_addrs;
1216         if (ra->desc_user_addr >= start && ra->desc_user_addr < end)
1217                 return 1;
1218         if (ra->avail_user_addr >= start && ra->avail_user_addr < end)
1219                 return 1;
1220         if (ra->used_user_addr >= start && ra->used_user_addr < end)
1221                 return 1;
1222
1223         return 0;
1224 }
1225
1226 static int
1227 is_vring_iotlb_invalidate(struct vhost_virtqueue *vq,
1228                                 struct vhost_iotlb_msg *imsg)
1229 {
1230         uint64_t istart, iend, vstart, vend;
1231
1232         istart = imsg->iova;
1233         iend = istart + imsg->size - 1;
1234
1235         vstart = (uintptr_t)vq->desc;
1236         vend = vstart + sizeof(struct vring_desc) * vq->size - 1;
1237         if (vstart <= iend && istart <= vend)
1238                 return 1;
1239
1240         vstart = (uintptr_t)vq->avail;
1241         vend = vstart + sizeof(struct vring_avail);
1242         vend += sizeof(uint16_t) * vq->size - 1;
1243         if (vstart <= iend && istart <= vend)
1244                 return 1;
1245
1246         vstart = (uintptr_t)vq->used;
1247         vend = vstart + sizeof(struct vring_used);
1248         vend += sizeof(struct vring_used_elem) * vq->size - 1;
1249         if (vstart <= iend && istart <= vend)
1250                 return 1;
1251
1252         return 0;
1253 }
1254
1255 static int
1256 vhost_user_iotlb_msg(struct virtio_net **pdev, struct VhostUserMsg *msg)
1257 {
1258         struct virtio_net *dev = *pdev;
1259         struct vhost_iotlb_msg *imsg = &msg->payload.iotlb;
1260         uint16_t i;
1261         uint64_t vva;
1262
1263         switch (imsg->type) {
1264         case VHOST_IOTLB_UPDATE:
1265                 vva = qva_to_vva(dev, imsg->uaddr);
1266                 if (!vva)
1267                         return -1;
1268
1269                 for (i = 0; i < dev->nr_vring; i++) {
1270                         struct vhost_virtqueue *vq = dev->virtqueue[i];
1271
1272                         vhost_user_iotlb_cache_insert(vq, imsg->iova, vva,
1273                                         imsg->size, imsg->perm);
1274
1275                         if (is_vring_iotlb_update(vq, imsg))
1276                                 *pdev = dev = translate_ring_addresses(dev, i);
1277                 }
1278                 break;
1279         case VHOST_IOTLB_INVALIDATE:
1280                 for (i = 0; i < dev->nr_vring; i++) {
1281                         struct vhost_virtqueue *vq = dev->virtqueue[i];
1282
1283                         vhost_user_iotlb_cache_remove(vq, imsg->iova,
1284                                         imsg->size);
1285
1286                         if (is_vring_iotlb_invalidate(vq, imsg))
1287                                 vring_invalidate(dev, vq);
1288                 }
1289                 break;
1290         default:
1291                 RTE_LOG(ERR, VHOST_CONFIG, "Invalid IOTLB message type (%d)\n",
1292                                 imsg->type);
1293                 return -1;
1294         }
1295
1296         return 0;
1297 }
1298
1299 /* return bytes# of read on success or negative val on failure. */
1300 static int
1301 read_vhost_message(int sockfd, struct VhostUserMsg *msg)
1302 {
1303         int ret;
1304
1305         ret = read_fd_message(sockfd, (char *)msg, VHOST_USER_HDR_SIZE,
1306                 msg->fds, VHOST_MEMORY_MAX_NREGIONS);
1307         if (ret <= 0)
1308                 return ret;
1309
1310         if (msg && msg->size) {
1311                 if (msg->size > sizeof(msg->payload)) {
1312                         RTE_LOG(ERR, VHOST_CONFIG,
1313                                 "invalid msg size: %d\n", msg->size);
1314                         return -1;
1315                 }
1316                 ret = read(sockfd, &msg->payload, msg->size);
1317                 if (ret <= 0)
1318                         return ret;
1319                 if (ret != (int)msg->size) {
1320                         RTE_LOG(ERR, VHOST_CONFIG,
1321                                 "read control message failed\n");
1322                         return -1;
1323                 }
1324         }
1325
1326         return ret;
1327 }
1328
1329 static int
1330 send_vhost_message(int sockfd, struct VhostUserMsg *msg, int *fds, int fd_num)
1331 {
1332         if (!msg)
1333                 return 0;
1334
1335         return send_fd_message(sockfd, (char *)msg,
1336                 VHOST_USER_HDR_SIZE + msg->size, fds, fd_num);
1337 }
1338
1339 static int
1340 send_vhost_reply(int sockfd, struct VhostUserMsg *msg)
1341 {
1342         if (!msg)
1343                 return 0;
1344
1345         msg->flags &= ~VHOST_USER_VERSION_MASK;
1346         msg->flags &= ~VHOST_USER_NEED_REPLY;
1347         msg->flags |= VHOST_USER_VERSION;
1348         msg->flags |= VHOST_USER_REPLY_MASK;
1349
1350         return send_vhost_message(sockfd, msg, NULL, 0);
1351 }
1352
1353 /*
1354  * Allocate a queue pair if it hasn't been allocated yet
1355  */
1356 static int
1357 vhost_user_check_and_alloc_queue_pair(struct virtio_net *dev, VhostUserMsg *msg)
1358 {
1359         uint16_t vring_idx;
1360
1361         switch (msg->request.master) {
1362         case VHOST_USER_SET_VRING_KICK:
1363         case VHOST_USER_SET_VRING_CALL:
1364         case VHOST_USER_SET_VRING_ERR:
1365                 vring_idx = msg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
1366                 break;
1367         case VHOST_USER_SET_VRING_NUM:
1368         case VHOST_USER_SET_VRING_BASE:
1369         case VHOST_USER_SET_VRING_ENABLE:
1370                 vring_idx = msg->payload.state.index;
1371                 break;
1372         case VHOST_USER_SET_VRING_ADDR:
1373                 vring_idx = msg->payload.addr.index;
1374                 break;
1375         default:
1376                 return 0;
1377         }
1378
1379         if (vring_idx >= VHOST_MAX_VRING) {
1380                 RTE_LOG(ERR, VHOST_CONFIG,
1381                         "invalid vring index: %u\n", vring_idx);
1382                 return -1;
1383         }
1384
1385         if (dev->virtqueue[vring_idx])
1386                 return 0;
1387
1388         return alloc_vring_queue(dev, vring_idx);
1389 }
1390
1391 static void
1392 vhost_user_lock_all_queue_pairs(struct virtio_net *dev)
1393 {
1394         unsigned int i = 0;
1395         unsigned int vq_num = 0;
1396
1397         while (vq_num < dev->nr_vring) {
1398                 struct vhost_virtqueue *vq = dev->virtqueue[i];
1399
1400                 if (vq) {
1401                         rte_spinlock_lock(&vq->access_lock);
1402                         vq_num++;
1403                 }
1404                 i++;
1405         }
1406 }
1407
1408 static void
1409 vhost_user_unlock_all_queue_pairs(struct virtio_net *dev)
1410 {
1411         unsigned int i = 0;
1412         unsigned int vq_num = 0;
1413
1414         while (vq_num < dev->nr_vring) {
1415                 struct vhost_virtqueue *vq = dev->virtqueue[i];
1416
1417                 if (vq) {
1418                         rte_spinlock_unlock(&vq->access_lock);
1419                         vq_num++;
1420                 }
1421                 i++;
1422         }
1423 }
1424
1425 int
1426 vhost_user_msg_handler(int vid, int fd)
1427 {
1428         struct virtio_net *dev;
1429         struct VhostUserMsg msg;
1430         struct rte_vdpa_device *vdpa_dev;
1431         int did = -1;
1432         int ret;
1433         int unlock_required = 0;
1434         uint32_t skip_master = 0;
1435
1436         dev = get_device(vid);
1437         if (dev == NULL)
1438                 return -1;
1439
1440         if (!dev->notify_ops) {
1441                 dev->notify_ops = vhost_driver_callback_get(dev->ifname);
1442                 if (!dev->notify_ops) {
1443                         RTE_LOG(ERR, VHOST_CONFIG,
1444                                 "failed to get callback ops for driver %s\n",
1445                                 dev->ifname);
1446                         return -1;
1447                 }
1448         }
1449
1450         ret = read_vhost_message(fd, &msg);
1451         if (ret <= 0 || msg.request.master >= VHOST_USER_MAX) {
1452                 if (ret < 0)
1453                         RTE_LOG(ERR, VHOST_CONFIG,
1454                                 "vhost read message failed\n");
1455                 else if (ret == 0)
1456                         RTE_LOG(INFO, VHOST_CONFIG,
1457                                 "vhost peer closed\n");
1458                 else
1459                         RTE_LOG(ERR, VHOST_CONFIG,
1460                                 "vhost read incorrect message\n");
1461
1462                 return -1;
1463         }
1464
1465         ret = 0;
1466         if (msg.request.master != VHOST_USER_IOTLB_MSG)
1467                 RTE_LOG(INFO, VHOST_CONFIG, "read message %s\n",
1468                         vhost_message_str[msg.request.master]);
1469         else
1470                 RTE_LOG(DEBUG, VHOST_CONFIG, "read message %s\n",
1471                         vhost_message_str[msg.request.master]);
1472
1473         ret = vhost_user_check_and_alloc_queue_pair(dev, &msg);
1474         if (ret < 0) {
1475                 RTE_LOG(ERR, VHOST_CONFIG,
1476                         "failed to alloc queue\n");
1477                 return -1;
1478         }
1479
1480         /*
1481          * Note: we don't lock all queues on VHOST_USER_GET_VRING_BASE
1482          * and VHOST_USER_RESET_OWNER, since it is sent when virtio stops
1483          * and device is destroyed. destroy_device waits for queues to be
1484          * inactive, so it is safe. Otherwise taking the access_lock
1485          * would cause a dead lock.
1486          */
1487         switch (msg.request.master) {
1488         case VHOST_USER_SET_FEATURES:
1489         case VHOST_USER_SET_PROTOCOL_FEATURES:
1490         case VHOST_USER_SET_OWNER:
1491         case VHOST_USER_SET_MEM_TABLE:
1492         case VHOST_USER_SET_LOG_BASE:
1493         case VHOST_USER_SET_LOG_FD:
1494         case VHOST_USER_SET_VRING_NUM:
1495         case VHOST_USER_SET_VRING_ADDR:
1496         case VHOST_USER_SET_VRING_BASE:
1497         case VHOST_USER_SET_VRING_KICK:
1498         case VHOST_USER_SET_VRING_CALL:
1499         case VHOST_USER_SET_VRING_ERR:
1500         case VHOST_USER_SET_VRING_ENABLE:
1501         case VHOST_USER_SEND_RARP:
1502         case VHOST_USER_NET_SET_MTU:
1503         case VHOST_USER_SET_SLAVE_REQ_FD:
1504                 vhost_user_lock_all_queue_pairs(dev);
1505                 unlock_required = 1;
1506                 break;
1507         default:
1508                 break;
1509
1510         }
1511
1512         if (dev->extern_ops.pre_msg_handle) {
1513                 uint32_t need_reply;
1514
1515                 ret = (*dev->extern_ops.pre_msg_handle)(dev->vid,
1516                                 (void *)&msg, &need_reply, &skip_master);
1517                 if (ret < 0)
1518                         goto skip_to_reply;
1519
1520                 if (need_reply)
1521                         send_vhost_reply(fd, &msg);
1522
1523                 if (skip_master)
1524                         goto skip_to_post_handle;
1525         }
1526
1527         switch (msg.request.master) {
1528         case VHOST_USER_GET_FEATURES:
1529                 msg.payload.u64 = vhost_user_get_features(dev);
1530                 msg.size = sizeof(msg.payload.u64);
1531                 send_vhost_reply(fd, &msg);
1532                 break;
1533         case VHOST_USER_SET_FEATURES:
1534                 ret = vhost_user_set_features(dev, msg.payload.u64);
1535                 if (ret)
1536                         return -1;
1537                 break;
1538
1539         case VHOST_USER_GET_PROTOCOL_FEATURES:
1540                 vhost_user_get_protocol_features(dev, &msg);
1541                 send_vhost_reply(fd, &msg);
1542                 break;
1543         case VHOST_USER_SET_PROTOCOL_FEATURES:
1544                 vhost_user_set_protocol_features(dev, msg.payload.u64);
1545                 break;
1546
1547         case VHOST_USER_SET_OWNER:
1548                 vhost_user_set_owner();
1549                 break;
1550         case VHOST_USER_RESET_OWNER:
1551                 vhost_user_reset_owner(dev);
1552                 break;
1553
1554         case VHOST_USER_SET_MEM_TABLE:
1555                 ret = vhost_user_set_mem_table(dev, &msg);
1556                 break;
1557
1558         case VHOST_USER_SET_LOG_BASE:
1559                 vhost_user_set_log_base(dev, &msg);
1560
1561                 /* it needs a reply */
1562                 msg.size = sizeof(msg.payload.u64);
1563                 send_vhost_reply(fd, &msg);
1564                 break;
1565         case VHOST_USER_SET_LOG_FD:
1566                 close(msg.fds[0]);
1567                 RTE_LOG(INFO, VHOST_CONFIG, "not implemented.\n");
1568                 break;
1569
1570         case VHOST_USER_SET_VRING_NUM:
1571                 vhost_user_set_vring_num(dev, &msg);
1572                 break;
1573         case VHOST_USER_SET_VRING_ADDR:
1574                 vhost_user_set_vring_addr(&dev, &msg);
1575                 break;
1576         case VHOST_USER_SET_VRING_BASE:
1577                 vhost_user_set_vring_base(dev, &msg);
1578                 break;
1579
1580         case VHOST_USER_GET_VRING_BASE:
1581                 vhost_user_get_vring_base(dev, &msg);
1582                 msg.size = sizeof(msg.payload.state);
1583                 send_vhost_reply(fd, &msg);
1584                 break;
1585
1586         case VHOST_USER_SET_VRING_KICK:
1587                 vhost_user_set_vring_kick(&dev, &msg);
1588                 break;
1589         case VHOST_USER_SET_VRING_CALL:
1590                 vhost_user_set_vring_call(dev, &msg);
1591                 break;
1592
1593         case VHOST_USER_SET_VRING_ERR:
1594                 if (!(msg.payload.u64 & VHOST_USER_VRING_NOFD_MASK))
1595                         close(msg.fds[0]);
1596                 RTE_LOG(INFO, VHOST_CONFIG, "not implemented\n");
1597                 break;
1598
1599         case VHOST_USER_GET_QUEUE_NUM:
1600                 msg.payload.u64 = (uint64_t)vhost_user_get_queue_num(dev);
1601                 msg.size = sizeof(msg.payload.u64);
1602                 send_vhost_reply(fd, &msg);
1603                 break;
1604
1605         case VHOST_USER_SET_VRING_ENABLE:
1606                 vhost_user_set_vring_enable(dev, &msg);
1607                 break;
1608         case VHOST_USER_SEND_RARP:
1609                 vhost_user_send_rarp(dev, &msg);
1610                 break;
1611
1612         case VHOST_USER_NET_SET_MTU:
1613                 ret = vhost_user_net_set_mtu(dev, &msg);
1614                 break;
1615
1616         case VHOST_USER_SET_SLAVE_REQ_FD:
1617                 ret = vhost_user_set_req_fd(dev, &msg);
1618                 break;
1619
1620         case VHOST_USER_IOTLB_MSG:
1621                 ret = vhost_user_iotlb_msg(&dev, &msg);
1622                 break;
1623
1624         default:
1625                 ret = -1;
1626                 break;
1627         }
1628
1629 skip_to_post_handle:
1630         if (dev->extern_ops.post_msg_handle) {
1631                 uint32_t need_reply;
1632
1633                 ret = (*dev->extern_ops.post_msg_handle)(
1634                                 dev->vid, (void *)&msg, &need_reply);
1635                 if (ret < 0)
1636                         goto skip_to_reply;
1637
1638                 if (need_reply)
1639                         send_vhost_reply(fd, &msg);
1640         }
1641
1642 skip_to_reply:
1643         if (unlock_required)
1644                 vhost_user_unlock_all_queue_pairs(dev);
1645
1646         if (msg.flags & VHOST_USER_NEED_REPLY) {
1647                 msg.payload.u64 = !!ret;
1648                 msg.size = sizeof(msg.payload.u64);
1649                 send_vhost_reply(fd, &msg);
1650         }
1651
1652         if (!(dev->flags & VIRTIO_DEV_RUNNING) && virtio_is_ready(dev)) {
1653                 dev->flags |= VIRTIO_DEV_READY;
1654
1655                 if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
1656                         if (dev->dequeue_zero_copy) {
1657                                 RTE_LOG(INFO, VHOST_CONFIG,
1658                                                 "dequeue zero copy is enabled\n");
1659                         }
1660
1661                         if (dev->notify_ops->new_device(dev->vid) == 0)
1662                                 dev->flags |= VIRTIO_DEV_RUNNING;
1663                 }
1664         }
1665
1666         did = dev->vdpa_dev_id;
1667         vdpa_dev = rte_vdpa_get_device(did);
1668         if (vdpa_dev && virtio_is_ready(dev) &&
1669                         !(dev->flags & VIRTIO_DEV_VDPA_CONFIGURED) &&
1670                         msg.request.master == VHOST_USER_SET_VRING_ENABLE) {
1671                 if (vdpa_dev->ops->dev_conf)
1672                         vdpa_dev->ops->dev_conf(dev->vid);
1673                 dev->flags |= VIRTIO_DEV_VDPA_CONFIGURED;
1674         }
1675
1676         return 0;
1677 }
1678
1679 int
1680 vhost_user_iotlb_miss(struct virtio_net *dev, uint64_t iova, uint8_t perm)
1681 {
1682         int ret;
1683         struct VhostUserMsg msg = {
1684                 .request.slave = VHOST_USER_SLAVE_IOTLB_MSG,
1685                 .flags = VHOST_USER_VERSION,
1686                 .size = sizeof(msg.payload.iotlb),
1687                 .payload.iotlb = {
1688                         .iova = iova,
1689                         .perm = perm,
1690                         .type = VHOST_IOTLB_MISS,
1691                 },
1692         };
1693
1694         ret = send_vhost_message(dev->slave_req_fd, &msg, NULL, 0);
1695         if (ret < 0) {
1696                 RTE_LOG(ERR, VHOST_CONFIG,
1697                                 "Failed to send IOTLB miss message (%d)\n",
1698                                 ret);
1699                 return ret;
1700         }
1701
1702         return 0;
1703 }