lib/vhost/virtio_net.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2010-2016 Intel Corporation
   3  */
   4
   5 #include <stdint.h>
   6 #include <stdbool.h>
   7 #include <linux/virtio_net.h>
   8
   9 #include <rte_mbuf.h>
  10 #include <rte_memcpy.h>
  11 #include <rte_net.h>
  12 #include <rte_ether.h>
  13 #include <rte_ip.h>
  14 #include <rte_vhost.h>
  15 #include <rte_tcp.h>
  16 #include <rte_udp.h>
  17 #include <rte_sctp.h>
  18 #include <rte_arp.h>
  19 #include <rte_spinlock.h>
  20 #include <rte_malloc.h>
  21 #include <rte_vhost_async.h>
  22
  23 #include "iotlb.h"
  24 #include "vhost.h"
  25
  26 #define MAX_BATCH_LEN 256
  27
  28 static  __rte_always_inline bool
  29 rxvq_is_mergeable(struct virtio_net *dev)
  30 {
  31         return dev->features & (1ULL << VIRTIO_NET_F_MRG_RXBUF);
  32 }
  33
  34 static  __rte_always_inline bool
  35 virtio_net_is_inorder(struct virtio_net *dev)
  36 {
  37         return dev->features & (1ULL << VIRTIO_F_IN_ORDER);
  38 }
  39
  40 static bool
  41 is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring)
  42 {
  43         return (is_tx ^ (idx & 1)) == 0 && idx < nr_vring;
  44 }
  45
  46 static inline void
  47 do_data_copy_enqueue(struct virtio_net *dev, struct vhost_virtqueue *vq)
  48 {
  49         struct batch_copy_elem *elem = vq->batch_copy_elems;
  50         uint16_t count = vq->batch_copy_nb_elems;
  51         int i;
  52
  53         for (i = 0; i < count; i++) {
  54                 rte_memcpy(elem[i].dst, elem[i].src, elem[i].len);
  55                 vhost_log_cache_write_iova(dev, vq, elem[i].log_addr,
  56                                            elem[i].len);
  57                 PRINT_PACKET(dev, (uintptr_t)elem[i].dst, elem[i].len, 0);
  58         }
  59
  60         vq->batch_copy_nb_elems = 0;
  61 }
  62
  63 static inline void
  64 do_data_copy_dequeue(struct vhost_virtqueue *vq)
  65 {
  66         struct batch_copy_elem *elem = vq->batch_copy_elems;
  67         uint16_t count = vq->batch_copy_nb_elems;
  68         int i;
  69
  70         for (i = 0; i < count; i++)
  71                 rte_memcpy(elem[i].dst, elem[i].src, elem[i].len);
  72
  73         vq->batch_copy_nb_elems = 0;
  74 }
  75
  76 static __rte_always_inline void
  77 do_flush_shadow_used_ring_split(struct virtio_net *dev,
  78                         struct vhost_virtqueue *vq,
  79                         uint16_t to, uint16_t from, uint16_t size)
  80 {
  81         rte_memcpy(&vq->used->ring[to],
  82                         &vq->shadow_used_split[from],
  83                         size * sizeof(struct vring_used_elem));
  84         vhost_log_cache_used_vring(dev, vq,
  85                         offsetof(struct vring_used, ring[to]),
  86                         size * sizeof(struct vring_used_elem));
  87 }
  88
  89 static __rte_always_inline void
  90 flush_shadow_used_ring_split(struct virtio_net *dev, struct vhost_virtqueue *vq)
  91 {
  92         uint16_t used_idx = vq->last_used_idx & (vq->size - 1);
  93
  94         if (used_idx + vq->shadow_used_idx <= vq->size) {
  95                 do_flush_shadow_used_ring_split(dev, vq, used_idx, 0,
  96                                           vq->shadow_used_idx);
  97         } else {
  98                 uint16_t size;
  99
 100                 /* update used ring interval [used_idx, vq->size] */
 101                 size = vq->size - used_idx;
 102                 do_flush_shadow_used_ring_split(dev, vq, used_idx, 0, size);
 103
 104                 /* update the left half used ring interval [0, left_size] */
 105                 do_flush_shadow_used_ring_split(dev, vq, 0, size,
 106                                           vq->shadow_used_idx - size);
 107         }
 108         vq->last_used_idx += vq->shadow_used_idx;
 109
 110         vhost_log_cache_sync(dev, vq);
 111
 112         __atomic_add_fetch(&vq->used->idx, vq->shadow_used_idx,
 113                            __ATOMIC_RELEASE);
 114         vq->shadow_used_idx = 0;
 115         vhost_log_used_vring(dev, vq, offsetof(struct vring_used, idx),
 116                 sizeof(vq->used->idx));
 117 }
 118
 119 static __rte_always_inline void
 120 update_shadow_used_ring_split(struct vhost_virtqueue *vq,
 121                          uint16_t desc_idx, uint32_t len)
 122 {
 123         uint16_t i = vq->shadow_used_idx++;
 124
 125         vq->shadow_used_split[i].id  = desc_idx;
 126         vq->shadow_used_split[i].len = len;
 127 }
 128
 129 static __rte_always_inline void
 130 vhost_flush_enqueue_shadow_packed(struct virtio_net *dev,
 131                                   struct vhost_virtqueue *vq)
 132 {
 133         int i;
 134         uint16_t used_idx = vq->last_used_idx;
 135         uint16_t head_idx = vq->last_used_idx;
 136         uint16_t head_flags = 0;
 137
 138         /* Split loop in two to save memory barriers */
 139         for (i = 0; i < vq->shadow_used_idx; i++) {
 140                 vq->desc_packed[used_idx].id = vq->shadow_used_packed[i].id;
 141                 vq->desc_packed[used_idx].len = vq->shadow_used_packed[i].len;
 142
 143                 used_idx += vq->shadow_used_packed[i].count;
 144                 if (used_idx >= vq->size)
 145                         used_idx -= vq->size;
 146         }
 147
 148         /* The ordering for storing desc flags needs to be enforced. */
 149         rte_atomic_thread_fence(__ATOMIC_RELEASE);
 150
 151         for (i = 0; i < vq->shadow_used_idx; i++) {
 152                 uint16_t flags;
 153
 154                 if (vq->shadow_used_packed[i].len)
 155                         flags = VRING_DESC_F_WRITE;
 156                 else
 157                         flags = 0;
 158
 159                 if (vq->used_wrap_counter) {
 160                         flags |= VRING_DESC_F_USED;
 161                         flags |= VRING_DESC_F_AVAIL;
 162                 } else {
 163                         flags &= ~VRING_DESC_F_USED;
 164                         flags &= ~VRING_DESC_F_AVAIL;
 165                 }
 166
 167                 if (i > 0) {
 168                         vq->desc_packed[vq->last_used_idx].flags = flags;
 169
 170                         vhost_log_cache_used_vring(dev, vq,
 171                                         vq->last_used_idx *
 172                                         sizeof(struct vring_packed_desc),
 173                                         sizeof(struct vring_packed_desc));
 174                 } else {
 175                         head_idx = vq->last_used_idx;
 176                         head_flags = flags;
 177                 }
 178
 179                 vq_inc_last_used_packed(vq, vq->shadow_used_packed[i].count);
 180         }
 181
 182         vq->desc_packed[head_idx].flags = head_flags;
 183
 184         vhost_log_cache_used_vring(dev, vq,
 185                                 head_idx *
 186                                 sizeof(struct vring_packed_desc),
 187                                 sizeof(struct vring_packed_desc));
 188
 189         vq->shadow_used_idx = 0;
 190         vhost_log_cache_sync(dev, vq);
 191 }
 192
 193 static __rte_always_inline void
 194 vhost_flush_dequeue_shadow_packed(struct virtio_net *dev,
 195                                   struct vhost_virtqueue *vq)
 196 {
 197         struct vring_used_elem_packed *used_elem = &vq->shadow_used_packed[0];
 198
 199         vq->desc_packed[vq->shadow_last_used_idx].id = used_elem->id;
 200         /* desc flags is the synchronization point for virtio packed vring */
 201         __atomic_store_n(&vq->desc_packed[vq->shadow_last_used_idx].flags,
 202                          used_elem->flags, __ATOMIC_RELEASE);
 203
 204         vhost_log_cache_used_vring(dev, vq, vq->shadow_last_used_idx *
 205                                    sizeof(struct vring_packed_desc),
 206                                    sizeof(struct vring_packed_desc));
 207         vq->shadow_used_idx = 0;
 208         vhost_log_cache_sync(dev, vq);
 209 }
 210
 211 static __rte_always_inline void
 212 vhost_flush_enqueue_batch_packed(struct virtio_net *dev,
 213                                  struct vhost_virtqueue *vq,
 214                                  uint64_t *lens,
 215                                  uint16_t *ids)
 216 {
 217         uint16_t i;
 218         uint16_t flags;
 219         uint16_t last_used_idx;
 220         struct vring_packed_desc *desc_base;
 221
 222         last_used_idx = vq->last_used_idx;
 223         desc_base = &vq->desc_packed[last_used_idx];
 224
 225         flags = PACKED_DESC_ENQUEUE_USED_FLAG(vq->used_wrap_counter);
 226
 227         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
 228                 desc_base[i].id = ids[i];
 229                 desc_base[i].len = lens[i];
 230         }
 231
 232         rte_atomic_thread_fence(__ATOMIC_RELEASE);
 233
 234         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
 235                 desc_base[i].flags = flags;
 236         }
 237
 238         vhost_log_cache_used_vring(dev, vq, last_used_idx *
 239                                    sizeof(struct vring_packed_desc),
 240                                    sizeof(struct vring_packed_desc) *
 241                                    PACKED_BATCH_SIZE);
 242         vhost_log_cache_sync(dev, vq);
 243
 244         vq_inc_last_used_packed(vq, PACKED_BATCH_SIZE);
 245 }
 246
 247 static __rte_always_inline void
 248 vhost_shadow_dequeue_batch_packed_inorder(struct vhost_virtqueue *vq,
 249                                           uint16_t id)
 250 {
 251         vq->shadow_used_packed[0].id = id;
 252
 253         if (!vq->shadow_used_idx) {
 254                 vq->shadow_last_used_idx = vq->last_used_idx;
 255                 vq->shadow_used_packed[0].flags =
 256                         PACKED_DESC_DEQUEUE_USED_FLAG(vq->used_wrap_counter);
 257                 vq->shadow_used_packed[0].len = 0;
 258                 vq->shadow_used_packed[0].count = 1;
 259                 vq->shadow_used_idx++;
 260         }
 261
 262         vq_inc_last_used_packed(vq, PACKED_BATCH_SIZE);
 263 }
 264
 265 static __rte_always_inline void
 266 vhost_shadow_dequeue_batch_packed(struct virtio_net *dev,
 267                                   struct vhost_virtqueue *vq,
 268                                   uint16_t *ids)
 269 {
 270         uint16_t flags;
 271         uint16_t i;
 272         uint16_t begin;
 273
 274         flags = PACKED_DESC_DEQUEUE_USED_FLAG(vq->used_wrap_counter);
 275
 276         if (!vq->shadow_used_idx) {
 277                 vq->shadow_last_used_idx = vq->last_used_idx;
 278                 vq->shadow_used_packed[0].id  = ids[0];
 279                 vq->shadow_used_packed[0].len = 0;
 280                 vq->shadow_used_packed[0].count = 1;
 281                 vq->shadow_used_packed[0].flags = flags;
 282                 vq->shadow_used_idx++;
 283                 begin = 1;
 284         } else
 285                 begin = 0;
 286
 287         vhost_for_each_try_unroll(i, begin, PACKED_BATCH_SIZE) {
 288                 vq->desc_packed[vq->last_used_idx + i].id = ids[i];
 289                 vq->desc_packed[vq->last_used_idx + i].len = 0;
 290         }
 291
 292         rte_atomic_thread_fence(__ATOMIC_RELEASE);
 293         vhost_for_each_try_unroll(i, begin, PACKED_BATCH_SIZE)
 294                 vq->desc_packed[vq->last_used_idx + i].flags = flags;
 295
 296         vhost_log_cache_used_vring(dev, vq, vq->last_used_idx *
 297                                    sizeof(struct vring_packed_desc),
 298                                    sizeof(struct vring_packed_desc) *
 299                                    PACKED_BATCH_SIZE);
 300         vhost_log_cache_sync(dev, vq);
 301
 302         vq_inc_last_used_packed(vq, PACKED_BATCH_SIZE);
 303 }
 304
 305 static __rte_always_inline void
 306 vhost_shadow_dequeue_single_packed(struct vhost_virtqueue *vq,
 307                                    uint16_t buf_id,
 308                                    uint16_t count)
 309 {
 310         uint16_t flags;
 311
 312         flags = vq->desc_packed[vq->last_used_idx].flags;
 313         if (vq->used_wrap_counter) {
 314                 flags |= VRING_DESC_F_USED;
 315                 flags |= VRING_DESC_F_AVAIL;
 316         } else {
 317                 flags &= ~VRING_DESC_F_USED;
 318                 flags &= ~VRING_DESC_F_AVAIL;
 319         }
 320
 321         if (!vq->shadow_used_idx) {
 322                 vq->shadow_last_used_idx = vq->last_used_idx;
 323
 324                 vq->shadow_used_packed[0].id  = buf_id;
 325                 vq->shadow_used_packed[0].len = 0;
 326                 vq->shadow_used_packed[0].flags = flags;
 327                 vq->shadow_used_idx++;
 328         } else {
 329                 vq->desc_packed[vq->last_used_idx].id = buf_id;
 330                 vq->desc_packed[vq->last_used_idx].len = 0;
 331                 vq->desc_packed[vq->last_used_idx].flags = flags;
 332         }
 333
 334         vq_inc_last_used_packed(vq, count);
 335 }
 336
 337 static __rte_always_inline void
 338 vhost_shadow_dequeue_single_packed_inorder(struct vhost_virtqueue *vq,
 339                                            uint16_t buf_id,
 340                                            uint16_t count)
 341 {
 342         uint16_t flags;
 343
 344         vq->shadow_used_packed[0].id = buf_id;
 345
 346         flags = vq->desc_packed[vq->last_used_idx].flags;
 347         if (vq->used_wrap_counter) {
 348                 flags |= VRING_DESC_F_USED;
 349                 flags |= VRING_DESC_F_AVAIL;
 350         } else {
 351                 flags &= ~VRING_DESC_F_USED;
 352                 flags &= ~VRING_DESC_F_AVAIL;
 353         }
 354
 355         if (!vq->shadow_used_idx) {
 356                 vq->shadow_last_used_idx = vq->last_used_idx;
 357                 vq->shadow_used_packed[0].len = 0;
 358                 vq->shadow_used_packed[0].flags = flags;
 359                 vq->shadow_used_idx++;
 360         }
 361
 362         vq_inc_last_used_packed(vq, count);
 363 }
 364
 365 static __rte_always_inline void
 366 vhost_shadow_enqueue_packed(struct vhost_virtqueue *vq,
 367                                    uint32_t *len,
 368                                    uint16_t *id,
 369                                    uint16_t *count,
 370                                    uint16_t num_buffers)
 371 {
 372         uint16_t i;
 373
 374         for (i = 0; i < num_buffers; i++) {
 375                 /* enqueue shadow flush action aligned with batch num */
 376                 if (!vq->shadow_used_idx)
 377                         vq->shadow_aligned_idx = vq->last_used_idx &
 378                                 PACKED_BATCH_MASK;
 379                 vq->shadow_used_packed[vq->shadow_used_idx].id  = id[i];
 380                 vq->shadow_used_packed[vq->shadow_used_idx].len = len[i];
 381                 vq->shadow_used_packed[vq->shadow_used_idx].count = count[i];
 382                 vq->shadow_aligned_idx += count[i];
 383                 vq->shadow_used_idx++;
 384         }
 385 }
 386
 387 static __rte_always_inline void
 388 vhost_shadow_enqueue_single_packed(struct virtio_net *dev,
 389                                    struct vhost_virtqueue *vq,
 390                                    uint32_t *len,
 391                                    uint16_t *id,
 392                                    uint16_t *count,
 393                                    uint16_t num_buffers)
 394 {
 395         vhost_shadow_enqueue_packed(vq, len, id, count, num_buffers);
 396
 397         if (vq->shadow_aligned_idx >= PACKED_BATCH_SIZE) {
 398                 do_data_copy_enqueue(dev, vq);
 399                 vhost_flush_enqueue_shadow_packed(dev, vq);
 400         }
 401 }
 402
 403 /* avoid write operation when necessary, to lessen cache issues */
 404 #define ASSIGN_UNLESS_EQUAL(var, val) do {      \
 405         if ((var) != (val))                     \
 406                 (var) = (val);                  \
 407 } while (0)
 408
 409 static __rte_always_inline void
 410 virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr)
 411 {
 412         uint64_t csum_l4 = m_buf->ol_flags & RTE_MBUF_F_TX_L4_MASK;
 413
 414         if (m_buf->ol_flags & RTE_MBUF_F_TX_TCP_SEG)
 415                 csum_l4 |= RTE_MBUF_F_TX_TCP_CKSUM;
 416
 417         if (csum_l4) {
 418                 net_hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
 419                 net_hdr->csum_start = m_buf->l2_len + m_buf->l3_len;
 420
 421                 switch (csum_l4) {
 422                 case RTE_MBUF_F_TX_TCP_CKSUM:
 423                         net_hdr->csum_offset = (offsetof(struct rte_tcp_hdr,
 424                                                 cksum));
 425                         break;
 426                 case RTE_MBUF_F_TX_UDP_CKSUM:
 427                         net_hdr->csum_offset = (offsetof(struct rte_udp_hdr,
 428                                                 dgram_cksum));
 429                         break;
 430                 case RTE_MBUF_F_TX_SCTP_CKSUM:
 431                         net_hdr->csum_offset = (offsetof(struct rte_sctp_hdr,
 432                                                 cksum));
 433                         break;
 434                 }
 435         } else {
 436                 ASSIGN_UNLESS_EQUAL(net_hdr->csum_start, 0);
 437                 ASSIGN_UNLESS_EQUAL(net_hdr->csum_offset, 0);
 438                 ASSIGN_UNLESS_EQUAL(net_hdr->flags, 0);
 439         }
 440
 441         /* IP cksum verification cannot be bypassed, then calculate here */
 442         if (m_buf->ol_flags & RTE_MBUF_F_TX_IP_CKSUM) {
 443                 struct rte_ipv4_hdr *ipv4_hdr;
 444
 445                 ipv4_hdr = rte_pktmbuf_mtod_offset(m_buf, struct rte_ipv4_hdr *,
 446                                                    m_buf->l2_len);
 447                 ipv4_hdr->hdr_checksum = 0;
 448                 ipv4_hdr->hdr_checksum = rte_ipv4_cksum(ipv4_hdr);
 449         }
 450
 451         if (m_buf->ol_flags & RTE_MBUF_F_TX_TCP_SEG) {
 452                 if (m_buf->ol_flags & RTE_MBUF_F_TX_IPV4)
 453                         net_hdr->gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
 454                 else
 455                         net_hdr->gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
 456                 net_hdr->gso_size = m_buf->tso_segsz;
 457                 net_hdr->hdr_len = m_buf->l2_len + m_buf->l3_len
 458                                         + m_buf->l4_len;
 459         } else if (m_buf->ol_flags & RTE_MBUF_F_TX_UDP_SEG) {
 460                 net_hdr->gso_type = VIRTIO_NET_HDR_GSO_UDP;
 461                 net_hdr->gso_size = m_buf->tso_segsz;
 462                 net_hdr->hdr_len = m_buf->l2_len + m_buf->l3_len +
 463                         m_buf->l4_len;
 464         } else {
 465                 ASSIGN_UNLESS_EQUAL(net_hdr->gso_type, 0);
 466                 ASSIGN_UNLESS_EQUAL(net_hdr->gso_size, 0);
 467                 ASSIGN_UNLESS_EQUAL(net_hdr->hdr_len, 0);
 468         }
 469 }
 470
 471 static __rte_always_inline int
 472 map_one_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 473                 struct buf_vector *buf_vec, uint16_t *vec_idx,
 474                 uint64_t desc_iova, uint64_t desc_len, uint8_t perm)
 475 {
 476         uint16_t vec_id = *vec_idx;
 477
 478         while (desc_len) {
 479                 uint64_t desc_addr;
 480                 uint64_t desc_chunck_len = desc_len;
 481
 482                 if (unlikely(vec_id >= BUF_VECTOR_MAX))
 483                         return -1;
 484
 485                 desc_addr = vhost_iova_to_vva(dev, vq,
 486                                 desc_iova,
 487                                 &desc_chunck_len,
 488                                 perm);
 489                 if (unlikely(!desc_addr))
 490                         return -1;
 491
 492                 rte_prefetch0((void *)(uintptr_t)desc_addr);
 493
 494                 buf_vec[vec_id].buf_iova = desc_iova;
 495                 buf_vec[vec_id].buf_addr = desc_addr;
 496                 buf_vec[vec_id].buf_len  = desc_chunck_len;
 497
 498                 desc_len -= desc_chunck_len;
 499                 desc_iova += desc_chunck_len;
 500                 vec_id++;
 501         }
 502         *vec_idx = vec_id;
 503
 504         return 0;
 505 }
 506
 507 static __rte_always_inline int
 508 fill_vec_buf_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
 509                          uint32_t avail_idx, uint16_t *vec_idx,
 510                          struct buf_vector *buf_vec, uint16_t *desc_chain_head,
 511                          uint32_t *desc_chain_len, uint8_t perm)
 512 {
 513         uint16_t idx = vq->avail->ring[avail_idx & (vq->size - 1)];
 514         uint16_t vec_id = *vec_idx;
 515         uint32_t len    = 0;
 516         uint64_t dlen;
 517         uint32_t nr_descs = vq->size;
 518         uint32_t cnt    = 0;
 519         struct vring_desc *descs = vq->desc;
 520         struct vring_desc *idesc = NULL;
 521
 522         if (unlikely(idx >= vq->size))
 523                 return -1;
 524
 525         *desc_chain_head = idx;
 526
 527         if (vq->desc[idx].flags & VRING_DESC_F_INDIRECT) {
 528                 dlen = vq->desc[idx].len;
 529                 nr_descs = dlen / sizeof(struct vring_desc);
 530                 if (unlikely(nr_descs > vq->size))
 531                         return -1;
 532
 533                 descs = (struct vring_desc *)(uintptr_t)
 534                         vhost_iova_to_vva(dev, vq, vq->desc[idx].addr,
 535                                                 &dlen,
 536                                                 VHOST_ACCESS_RO);
 537                 if (unlikely(!descs))
 538                         return -1;
 539
 540                 if (unlikely(dlen < vq->desc[idx].len)) {
 541                         /*
 542                          * The indirect desc table is not contiguous
 543                          * in process VA space, we have to copy it.
 544                          */
 545                         idesc = vhost_alloc_copy_ind_table(dev, vq,
 546                                         vq->desc[idx].addr, vq->desc[idx].len);
 547                         if (unlikely(!idesc))
 548                                 return -1;
 549
 550                         descs = idesc;
 551                 }
 552
 553                 idx = 0;
 554         }
 555
 556         while (1) {
 557                 if (unlikely(idx >= nr_descs || cnt++ >= nr_descs)) {
 558                         free_ind_table(idesc);
 559                         return -1;
 560                 }
 561
 562                 dlen = descs[idx].len;
 563                 len += dlen;
 564
 565                 if (unlikely(map_one_desc(dev, vq, buf_vec, &vec_id,
 566                                                 descs[idx].addr, dlen,
 567                                                 perm))) {
 568                         free_ind_table(idesc);
 569                         return -1;
 570                 }
 571
 572                 if ((descs[idx].flags & VRING_DESC_F_NEXT) == 0)
 573                         break;
 574
 575                 idx = descs[idx].next;
 576         }
 577
 578         *desc_chain_len = len;
 579         *vec_idx = vec_id;
 580
 581         if (unlikely(!!idesc))
 582                 free_ind_table(idesc);
 583
 584         return 0;
 585 }
 586
 587 /*
 588  * Returns -1 on fail, 0 on success
 589  */
 590 static inline int
 591 reserve_avail_buf_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
 592                                 uint32_t size, struct buf_vector *buf_vec,
 593                                 uint16_t *num_buffers, uint16_t avail_head,
 594                                 uint16_t *nr_vec)
 595 {
 596         uint16_t cur_idx;
 597         uint16_t vec_idx = 0;
 598         uint16_t max_tries, tries = 0;
 599
 600         uint16_t head_idx = 0;
 601         uint32_t len = 0;
 602
 603         *num_buffers = 0;
 604         cur_idx  = vq->last_avail_idx;
 605
 606         if (rxvq_is_mergeable(dev))
 607                 max_tries = vq->size - 1;
 608         else
 609                 max_tries = 1;
 610
 611         while (size > 0) {
 612                 if (unlikely(cur_idx == avail_head))
 613                         return -1;
 614                 /*
 615                  * if we tried all available ring items, and still
 616                  * can't get enough buf, it means something abnormal
 617                  * happened.
 618                  */
 619                 if (unlikely(++tries > max_tries))
 620                         return -1;
 621
 622                 if (unlikely(fill_vec_buf_split(dev, vq, cur_idx,
 623                                                 &vec_idx, buf_vec,
 624                                                 &head_idx, &len,
 625                                                 VHOST_ACCESS_RW) < 0))
 626                         return -1;
 627                 len = RTE_MIN(len, size);
 628                 update_shadow_used_ring_split(vq, head_idx, len);
 629                 size -= len;
 630
 631                 cur_idx++;
 632                 *num_buffers += 1;
 633         }
 634
 635         *nr_vec = vec_idx;
 636
 637         return 0;
 638 }
 639
 640 static __rte_always_inline int
 641 fill_vec_buf_packed_indirect(struct virtio_net *dev,
 642                         struct vhost_virtqueue *vq,
 643                         struct vring_packed_desc *desc, uint16_t *vec_idx,
 644                         struct buf_vector *buf_vec, uint32_t *len, uint8_t perm)
 645 {
 646         uint16_t i;
 647         uint32_t nr_descs;
 648         uint16_t vec_id = *vec_idx;
 649         uint64_t dlen;
 650         struct vring_packed_desc *descs, *idescs = NULL;
 651
 652         dlen = desc->len;
 653         descs = (struct vring_packed_desc *)(uintptr_t)
 654                 vhost_iova_to_vva(dev, vq, desc->addr, &dlen, VHOST_ACCESS_RO);
 655         if (unlikely(!descs))
 656                 return -1;
 657
 658         if (unlikely(dlen < desc->len)) {
 659                 /*
 660                  * The indirect desc table is not contiguous
 661                  * in process VA space, we have to copy it.
 662                  */
 663                 idescs = vhost_alloc_copy_ind_table(dev,
 664                                 vq, desc->addr, desc->len);
 665                 if (unlikely(!idescs))
 666                         return -1;
 667
 668                 descs = idescs;
 669         }
 670
 671         nr_descs =  desc->len / sizeof(struct vring_packed_desc);
 672         if (unlikely(nr_descs >= vq->size)) {
 673                 free_ind_table(idescs);
 674                 return -1;
 675         }
 676
 677         for (i = 0; i < nr_descs; i++) {
 678                 if (unlikely(vec_id >= BUF_VECTOR_MAX)) {
 679                         free_ind_table(idescs);
 680                         return -1;
 681                 }
 682
 683                 dlen = descs[i].len;
 684                 *len += dlen;
 685                 if (unlikely(map_one_desc(dev, vq, buf_vec, &vec_id,
 686                                                 descs[i].addr, dlen,
 687                                                 perm)))
 688                         return -1;
 689         }
 690         *vec_idx = vec_id;
 691
 692         if (unlikely(!!idescs))
 693                 free_ind_table(idescs);
 694
 695         return 0;
 696 }
 697
 698 static __rte_always_inline int
 699 fill_vec_buf_packed(struct virtio_net *dev, struct vhost_virtqueue *vq,
 700                                 uint16_t avail_idx, uint16_t *desc_count,
 701                                 struct buf_vector *buf_vec, uint16_t *vec_idx,
 702                                 uint16_t *buf_id, uint32_t *len, uint8_t perm)
 703 {
 704         bool wrap_counter = vq->avail_wrap_counter;
 705         struct vring_packed_desc *descs = vq->desc_packed;
 706         uint16_t vec_id = *vec_idx;
 707         uint64_t dlen;
 708
 709         if (avail_idx < vq->last_avail_idx)
 710                 wrap_counter ^= 1;
 711
 712         /*
 713          * Perform a load-acquire barrier in desc_is_avail to
 714          * enforce the ordering between desc flags and desc
 715          * content.
 716          */
 717         if (unlikely(!desc_is_avail(&descs[avail_idx], wrap_counter)))
 718                 return -1;
 719
 720         *desc_count = 0;
 721         *len = 0;
 722
 723         while (1) {
 724                 if (unlikely(vec_id >= BUF_VECTOR_MAX))
 725                         return -1;
 726
 727                 if (unlikely(*desc_count >= vq->size))
 728                         return -1;
 729
 730                 *desc_count += 1;
 731                 *buf_id = descs[avail_idx].id;
 732
 733                 if (descs[avail_idx].flags & VRING_DESC_F_INDIRECT) {
 734                         if (unlikely(fill_vec_buf_packed_indirect(dev, vq,
 735                                                         &descs[avail_idx],
 736                                                         &vec_id, buf_vec,
 737                                                         len, perm) < 0))
 738                                 return -1;
 739                 } else {
 740                         dlen = descs[avail_idx].len;
 741                         *len += dlen;
 742
 743                         if (unlikely(map_one_desc(dev, vq, buf_vec, &vec_id,
 744                                                         descs[avail_idx].addr,
 745                                                         dlen,
 746                                                         perm)))
 747                                 return -1;
 748                 }
 749
 750                 if ((descs[avail_idx].flags & VRING_DESC_F_NEXT) == 0)
 751                         break;
 752
 753                 if (++avail_idx >= vq->size) {
 754                         avail_idx -= vq->size;
 755                         wrap_counter ^= 1;
 756                 }
 757         }
 758
 759         *vec_idx = vec_id;
 760
 761         return 0;
 762 }
 763
 764 static __rte_noinline void
 765 copy_vnet_hdr_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 766                 struct buf_vector *buf_vec,
 767                 struct virtio_net_hdr_mrg_rxbuf *hdr)
 768 {
 769         uint64_t len;
 770         uint64_t remain = dev->vhost_hlen;
 771         uint64_t src = (uint64_t)(uintptr_t)hdr, dst;
 772         uint64_t iova = buf_vec->buf_iova;
 773
 774         while (remain) {
 775                 len = RTE_MIN(remain,
 776                                 buf_vec->buf_len);
 777                 dst = buf_vec->buf_addr;
 778                 rte_memcpy((void *)(uintptr_t)dst,
 779                                 (void *)(uintptr_t)src,
 780                                 len);
 781
 782                 PRINT_PACKET(dev, (uintptr_t)dst,
 783                                 (uint32_t)len, 0);
 784                 vhost_log_cache_write_iova(dev, vq,
 785                                 iova, len);
 786
 787                 remain -= len;
 788                 iova += len;
 789                 src += len;
 790                 buf_vec++;
 791         }
 792 }
 793
 794 static __rte_always_inline int
 795 async_iter_initialize(struct virtio_net *dev, struct vhost_async *async)
 796 {
 797         struct rte_vhost_iov_iter *iter;
 798
 799         if (unlikely(async->iovec_idx >= VHOST_MAX_ASYNC_VEC)) {
 800                 VHOST_LOG_DATA(ERR, "(%s) no more async iovec available\n", dev->ifname);
 801                 return -1;
 802         }
 803
 804         iter = async->iov_iter + async->iter_idx;
 805         iter->iov = async->iovec + async->iovec_idx;
 806         iter->nr_segs = 0;
 807
 808         return 0;
 809 }
 810
 811 static __rte_always_inline int
 812 async_iter_add_iovec(struct virtio_net *dev, struct vhost_async *async,
 813                 void *src, void *dst, size_t len)
 814 {
 815         struct rte_vhost_iov_iter *iter;
 816         struct rte_vhost_iovec *iovec;
 817
 818         if (unlikely(async->iovec_idx >= VHOST_MAX_ASYNC_VEC)) {
 819                 static bool vhost_max_async_vec_log;
 820
 821                 if (!vhost_max_async_vec_log) {
 822                         VHOST_LOG_DATA(ERR, "(%s) no more async iovec available\n", dev->ifname);
 823                         vhost_max_async_vec_log = true;
 824                 }
 825
 826                 return -1;
 827         }
 828
 829         iter = async->iov_iter + async->iter_idx;
 830         iovec = async->iovec + async->iovec_idx;
 831
 832         iovec->src_addr = src;
 833         iovec->dst_addr = dst;
 834         iovec->len = len;
 835
 836         iter->nr_segs++;
 837         async->iovec_idx++;
 838
 839         return 0;
 840 }
 841
 842 static __rte_always_inline void
 843 async_iter_finalize(struct vhost_async *async)
 844 {
 845         async->iter_idx++;
 846 }
 847
 848 static __rte_always_inline void
 849 async_iter_cancel(struct vhost_async *async)
 850 {
 851         struct rte_vhost_iov_iter *iter;
 852
 853         iter = async->iov_iter + async->iter_idx;
 854         async->iovec_idx -= iter->nr_segs;
 855         iter->nr_segs = 0;
 856         iter->iov = NULL;
 857 }
 858
 859 static __rte_always_inline void
 860 async_iter_reset(struct vhost_async *async)
 861 {
 862         async->iter_idx = 0;
 863         async->iovec_idx = 0;
 864 }
 865
 866 static __rte_always_inline int
 867 async_mbuf_to_desc_seg(struct virtio_net *dev, struct vhost_virtqueue *vq,
 868                 struct rte_mbuf *m, uint32_t mbuf_offset,
 869                 uint64_t buf_iova, uint32_t cpy_len)
 870 {
 871         struct vhost_async *async = vq->async;
 872         uint64_t mapped_len;
 873         uint32_t buf_offset = 0;
 874         void *hpa;
 875
 876         while (cpy_len) {
 877                 hpa = (void *)(uintptr_t)gpa_to_first_hpa(dev,
 878                                 buf_iova + buf_offset, cpy_len, &mapped_len);
 879                 if (unlikely(!hpa)) {
 880                         VHOST_LOG_DATA(ERR, "(%s) %s: failed to get hpa.\n", dev->ifname, __func__);
 881                         return -1;
 882                 }
 883
 884                 if (unlikely(async_iter_add_iovec(dev, async,
 885                                                 (void *)(uintptr_t)rte_pktmbuf_iova_offset(m,
 886                                                         mbuf_offset),
 887                                                 hpa, (size_t)mapped_len)))
 888                         return -1;
 889
 890                 cpy_len -= (uint32_t)mapped_len;
 891                 mbuf_offset += (uint32_t)mapped_len;
 892                 buf_offset += (uint32_t)mapped_len;
 893         }
 894
 895         return 0;
 896 }
 897
 898 static __rte_always_inline void
 899 sync_mbuf_to_desc_seg(struct virtio_net *dev, struct vhost_virtqueue *vq,
 900                 struct rte_mbuf *m, uint32_t mbuf_offset,
 901                 uint64_t buf_addr, uint64_t buf_iova, uint32_t cpy_len)
 902 {
 903         struct batch_copy_elem *batch_copy = vq->batch_copy_elems;
 904
 905         if (likely(cpy_len > MAX_BATCH_LEN || vq->batch_copy_nb_elems >= vq->size)) {
 906                 rte_memcpy((void *)((uintptr_t)(buf_addr)),
 907                                 rte_pktmbuf_mtod_offset(m, void *, mbuf_offset),
 908                                 cpy_len);
 909                 vhost_log_cache_write_iova(dev, vq, buf_iova, cpy_len);
 910                 PRINT_PACKET(dev, (uintptr_t)(buf_addr), cpy_len, 0);
 911         } else {
 912                 batch_copy[vq->batch_copy_nb_elems].dst =
 913                         (void *)((uintptr_t)(buf_addr));
 914                 batch_copy[vq->batch_copy_nb_elems].src =
 915                         rte_pktmbuf_mtod_offset(m, void *, mbuf_offset);
 916                 batch_copy[vq->batch_copy_nb_elems].log_addr = buf_iova;
 917                 batch_copy[vq->batch_copy_nb_elems].len = cpy_len;
 918                 vq->batch_copy_nb_elems++;
 919         }
 920 }
 921
 922 static __rte_always_inline int
 923 mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 924                 struct rte_mbuf *m, struct buf_vector *buf_vec,
 925                 uint16_t nr_vec, uint16_t num_buffers, bool is_async)
 926 {
 927         uint32_t vec_idx = 0;
 928         uint32_t mbuf_offset, mbuf_avail;
 929         uint32_t buf_offset, buf_avail;
 930         uint64_t buf_addr, buf_iova, buf_len;
 931         uint32_t cpy_len;
 932         uint64_t hdr_addr;
 933         struct rte_mbuf *hdr_mbuf;
 934         struct virtio_net_hdr_mrg_rxbuf tmp_hdr, *hdr = NULL;
 935         struct vhost_async *async = vq->async;
 936
 937         if (unlikely(m == NULL))
 938                 return -1;
 939
 940         buf_addr = buf_vec[vec_idx].buf_addr;
 941         buf_iova = buf_vec[vec_idx].buf_iova;
 942         buf_len = buf_vec[vec_idx].buf_len;
 943
 944         if (unlikely(buf_len < dev->vhost_hlen && nr_vec <= 1))
 945                 return -1;
 946
 947         hdr_mbuf = m;
 948         hdr_addr = buf_addr;
 949         if (unlikely(buf_len < dev->vhost_hlen)) {
 950                 memset(&tmp_hdr, 0, sizeof(struct virtio_net_hdr_mrg_rxbuf));
 951                 hdr = &tmp_hdr;
 952         } else
 953                 hdr = (struct virtio_net_hdr_mrg_rxbuf *)(uintptr_t)hdr_addr;
 954
 955         VHOST_LOG_DATA(DEBUG, "(%s) RX: num merge buffers %d\n",
 956                 dev->ifname, num_buffers);
 957
 958         if (unlikely(buf_len < dev->vhost_hlen)) {
 959                 buf_offset = dev->vhost_hlen - buf_len;
 960                 vec_idx++;
 961                 buf_addr = buf_vec[vec_idx].buf_addr;
 962                 buf_iova = buf_vec[vec_idx].buf_iova;
 963                 buf_len = buf_vec[vec_idx].buf_len;
 964                 buf_avail = buf_len - buf_offset;
 965         } else {
 966                 buf_offset = dev->vhost_hlen;
 967                 buf_avail = buf_len - dev->vhost_hlen;
 968         }
 969
 970         mbuf_avail  = rte_pktmbuf_data_len(m);
 971         mbuf_offset = 0;
 972
 973         if (is_async) {
 974                 if (async_iter_initialize(dev, async))
 975                         return -1;
 976         }
 977
 978         while (mbuf_avail != 0 || m->next != NULL) {
 979                 /* done with current buf, get the next one */
 980                 if (buf_avail == 0) {
 981                         vec_idx++;
 982                         if (unlikely(vec_idx >= nr_vec))
 983                                 goto error;
 984
 985                         buf_addr = buf_vec[vec_idx].buf_addr;
 986                         buf_iova = buf_vec[vec_idx].buf_iova;
 987                         buf_len = buf_vec[vec_idx].buf_len;
 988
 989                         buf_offset = 0;
 990                         buf_avail  = buf_len;
 991                 }
 992
 993                 /* done with current mbuf, get the next one */
 994                 if (mbuf_avail == 0) {
 995                         m = m->next;
 996
 997                         mbuf_offset = 0;
 998                         mbuf_avail  = rte_pktmbuf_data_len(m);
 999                 }
1000
1001                 if (hdr_addr) {
1002                         virtio_enqueue_offload(hdr_mbuf, &hdr->hdr);
1003                         if (rxvq_is_mergeable(dev))
1004                                 ASSIGN_UNLESS_EQUAL(hdr->num_buffers,
1005                                                 num_buffers);
1006
1007                         if (unlikely(hdr == &tmp_hdr)) {
1008                                 copy_vnet_hdr_to_desc(dev, vq, buf_vec, hdr);
1009                         } else {
1010                                 PRINT_PACKET(dev, (uintptr_t)hdr_addr,
1011                                                 dev->vhost_hlen, 0);
1012                                 vhost_log_cache_write_iova(dev, vq,
1013                                                 buf_vec[0].buf_iova,
1014                                                 dev->vhost_hlen);
1015                         }
1016
1017                         hdr_addr = 0;
1018                 }
1019
1020                 cpy_len = RTE_MIN(buf_avail, mbuf_avail);
1021
1022                 if (is_async) {
1023                         if (async_mbuf_to_desc_seg(dev, vq, m, mbuf_offset,
1024                                                 buf_iova + buf_offset, cpy_len) < 0)
1025                                 goto error;
1026                 } else {
1027                         sync_mbuf_to_desc_seg(dev, vq, m, mbuf_offset,
1028                                         buf_addr + buf_offset,
1029                                         buf_iova + buf_offset, cpy_len);
1030                 }
1031
1032                 mbuf_avail  -= cpy_len;
1033                 mbuf_offset += cpy_len;
1034                 buf_avail  -= cpy_len;
1035                 buf_offset += cpy_len;
1036         }
1037
1038         if (is_async)
1039                 async_iter_finalize(async);
1040
1041         return 0;
1042 error:
1043         if (is_async)
1044                 async_iter_cancel(async);
1045
1046         return -1;
1047 }
1048
1049 static __rte_always_inline int
1050 vhost_enqueue_single_packed(struct virtio_net *dev,
1051                             struct vhost_virtqueue *vq,
1052                             struct rte_mbuf *pkt,
1053                             struct buf_vector *buf_vec,
1054                             uint16_t *nr_descs)
1055 {
1056         uint16_t nr_vec = 0;
1057         uint16_t avail_idx = vq->last_avail_idx;
1058         uint16_t max_tries, tries = 0;
1059         uint16_t buf_id = 0;
1060         uint32_t len = 0;
1061         uint16_t desc_count;
1062         uint32_t size = pkt->pkt_len + sizeof(struct virtio_net_hdr_mrg_rxbuf);
1063         uint16_t num_buffers = 0;
1064         uint32_t buffer_len[vq->size];
1065         uint16_t buffer_buf_id[vq->size];
1066         uint16_t buffer_desc_count[vq->size];
1067
1068         if (rxvq_is_mergeable(dev))
1069                 max_tries = vq->size - 1;
1070         else
1071                 max_tries = 1;
1072
1073         while (size > 0) {
1074                 /*
1075                  * if we tried all available ring items, and still
1076                  * can't get enough buf, it means something abnormal
1077                  * happened.
1078                  */
1079                 if (unlikely(++tries > max_tries))
1080                         return -1;
1081
1082                 if (unlikely(fill_vec_buf_packed(dev, vq,
1083                                                 avail_idx, &desc_count,
1084                                                 buf_vec, &nr_vec,
1085                                                 &buf_id, &len,
1086                                                 VHOST_ACCESS_RW) < 0))
1087                         return -1;
1088
1089                 len = RTE_MIN(len, size);
1090                 size -= len;
1091
1092                 buffer_len[num_buffers] = len;
1093                 buffer_buf_id[num_buffers] = buf_id;
1094                 buffer_desc_count[num_buffers] = desc_count;
1095                 num_buffers += 1;
1096
1097                 *nr_descs += desc_count;
1098                 avail_idx += desc_count;
1099                 if (avail_idx >= vq->size)
1100                         avail_idx -= vq->size;
1101         }
1102
1103         if (mbuf_to_desc(dev, vq, pkt, buf_vec, nr_vec, num_buffers, false) < 0)
1104                 return -1;
1105
1106         vhost_shadow_enqueue_single_packed(dev, vq, buffer_len, buffer_buf_id,
1107                                            buffer_desc_count, num_buffers);
1108
1109         return 0;
1110 }
1111
1112 static __rte_noinline uint32_t
1113 virtio_dev_rx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
1114         struct rte_mbuf **pkts, uint32_t count)
1115 {
1116         uint32_t pkt_idx = 0;
1117         uint16_t num_buffers;
1118         struct buf_vector buf_vec[BUF_VECTOR_MAX];
1119         uint16_t avail_head;
1120
1121         /*
1122          * The ordering between avail index and
1123          * desc reads needs to be enforced.
1124          */
1125         avail_head = __atomic_load_n(&vq->avail->idx, __ATOMIC_ACQUIRE);
1126
1127         rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]);
1128
1129         for (pkt_idx = 0; pkt_idx < count; pkt_idx++) {
1130                 uint32_t pkt_len = pkts[pkt_idx]->pkt_len + dev->vhost_hlen;
1131                 uint16_t nr_vec = 0;
1132
1133                 if (unlikely(reserve_avail_buf_split(dev, vq,
1134                                                 pkt_len, buf_vec, &num_buffers,
1135                                                 avail_head, &nr_vec) < 0)) {
1136                         VHOST_LOG_DATA(DEBUG,
1137                                 "(%s) failed to get enough desc from vring\n",
1138                                 dev->ifname);
1139                         vq->shadow_used_idx -= num_buffers;
1140                         break;
1141                 }
1142
1143                 VHOST_LOG_DATA(DEBUG, "(%s) current index %d | end index %d\n",
1144                         dev->ifname, vq->last_avail_idx,
1145                         vq->last_avail_idx + num_buffers);
1146
1147                 if (mbuf_to_desc(dev, vq, pkts[pkt_idx], buf_vec, nr_vec,
1148                                         num_buffers, false) < 0) {
1149                         vq->shadow_used_idx -= num_buffers;
1150                         break;
1151                 }
1152
1153                 vq->last_avail_idx += num_buffers;
1154         }
1155
1156         do_data_copy_enqueue(dev, vq);
1157
1158         if (likely(vq->shadow_used_idx)) {
1159                 flush_shadow_used_ring_split(dev, vq);
1160                 vhost_vring_call_split(dev, vq);
1161         }
1162
1163         return pkt_idx;
1164 }
1165
1166 static __rte_always_inline int
1167 virtio_dev_rx_sync_batch_check(struct virtio_net *dev,
1168                            struct vhost_virtqueue *vq,
1169                            struct rte_mbuf **pkts,
1170                            uint64_t *desc_addrs,
1171                            uint64_t *lens)
1172 {
1173         bool wrap_counter = vq->avail_wrap_counter;
1174         struct vring_packed_desc *descs = vq->desc_packed;
1175         uint16_t avail_idx = vq->last_avail_idx;
1176         uint32_t buf_offset = sizeof(struct virtio_net_hdr_mrg_rxbuf);
1177         uint16_t i;
1178
1179         if (unlikely(avail_idx & PACKED_BATCH_MASK))
1180                 return -1;
1181
1182         if (unlikely((avail_idx + PACKED_BATCH_SIZE) > vq->size))
1183                 return -1;
1184
1185         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
1186                 if (unlikely(pkts[i]->next != NULL))
1187                         return -1;
1188                 if (unlikely(!desc_is_avail(&descs[avail_idx + i],
1189                                             wrap_counter)))
1190                         return -1;
1191         }
1192
1193         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
1194                 lens[i] = descs[avail_idx + i].len;
1195
1196         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
1197                 if (unlikely(pkts[i]->pkt_len > (lens[i] - buf_offset)))
1198                         return -1;
1199         }
1200
1201         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
1202                 desc_addrs[i] = vhost_iova_to_vva(dev, vq,
1203                                                   descs[avail_idx + i].addr,
1204                                                   &lens[i],
1205                                                   VHOST_ACCESS_RW);
1206
1207         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
1208                 if (unlikely(!desc_addrs[i]))
1209                         return -1;
1210                 if (unlikely(lens[i] != descs[avail_idx + i].len))
1211                         return -1;
1212         }
1213
1214         return 0;
1215 }
1216
1217 static __rte_always_inline void
1218 virtio_dev_rx_batch_packed_copy(struct virtio_net *dev,
1219                            struct vhost_virtqueue *vq,
1220                            struct rte_mbuf **pkts,
1221                            uint64_t *desc_addrs,
1222                            uint64_t *lens)
1223 {
1224         uint32_t buf_offset = sizeof(struct virtio_net_hdr_mrg_rxbuf);
1225         struct virtio_net_hdr_mrg_rxbuf *hdrs[PACKED_BATCH_SIZE];
1226         struct vring_packed_desc *descs = vq->desc_packed;
1227         uint16_t avail_idx = vq->last_avail_idx;
1228         uint16_t ids[PACKED_BATCH_SIZE];
1229         uint16_t i;
1230
1231         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
1232                 rte_prefetch0((void *)(uintptr_t)desc_addrs[i]);
1233                 hdrs[i] = (struct virtio_net_hdr_mrg_rxbuf *)
1234                                         (uintptr_t)desc_addrs[i];
1235                 lens[i] = pkts[i]->pkt_len +
1236                         sizeof(struct virtio_net_hdr_mrg_rxbuf);
1237         }
1238
1239         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
1240                 virtio_enqueue_offload(pkts[i], &hdrs[i]->hdr);
1241
1242         vq_inc_last_avail_packed(vq, PACKED_BATCH_SIZE);
1243
1244         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
1245                 rte_memcpy((void *)(uintptr_t)(desc_addrs[i] + buf_offset),
1246                            rte_pktmbuf_mtod_offset(pkts[i], void *, 0),
1247                            pkts[i]->pkt_len);
1248         }
1249
1250         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
1251                 vhost_log_cache_write_iova(dev, vq, descs[avail_idx + i].addr,
1252                                            lens[i]);
1253
1254         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
1255                 ids[i] = descs[avail_idx + i].id;
1256
1257         vhost_flush_enqueue_batch_packed(dev, vq, lens, ids);
1258 }
1259
1260 static __rte_always_inline int
1261 virtio_dev_rx_sync_batch_packed(struct virtio_net *dev,
1262                            struct vhost_virtqueue *vq,
1263                            struct rte_mbuf **pkts)
1264 {
1265         uint64_t desc_addrs[PACKED_BATCH_SIZE];
1266         uint64_t lens[PACKED_BATCH_SIZE];
1267
1268         if (virtio_dev_rx_sync_batch_check(dev, vq, pkts, desc_addrs, lens) == -1)
1269                 return -1;
1270
1271         if (vq->shadow_used_idx) {
1272                 do_data_copy_enqueue(dev, vq);
1273                 vhost_flush_enqueue_shadow_packed(dev, vq);
1274         }
1275
1276         virtio_dev_rx_batch_packed_copy(dev, vq, pkts, desc_addrs, lens);
1277
1278         return 0;
1279 }
1280
1281 static __rte_always_inline int16_t
1282 virtio_dev_rx_single_packed(struct virtio_net *dev,
1283                             struct vhost_virtqueue *vq,
1284                             struct rte_mbuf *pkt)
1285 {
1286         struct buf_vector buf_vec[BUF_VECTOR_MAX];
1287         uint16_t nr_descs = 0;
1288
1289         if (unlikely(vhost_enqueue_single_packed(dev, vq, pkt, buf_vec,
1290                                                  &nr_descs) < 0)) {
1291                 VHOST_LOG_DATA(DEBUG, "(%s) failed to get enough desc from vring\n",
1292                                 dev->ifname);
1293                 return -1;
1294         }
1295
1296         VHOST_LOG_DATA(DEBUG, "(%s) current index %d | end index %d\n",
1297                         dev->ifname, vq->last_avail_idx,
1298                         vq->last_avail_idx + nr_descs);
1299
1300         vq_inc_last_avail_packed(vq, nr_descs);
1301
1302         return 0;
1303 }
1304
1305 static __rte_noinline uint32_t
1306 virtio_dev_rx_packed(struct virtio_net *dev,
1307                      struct vhost_virtqueue *__rte_restrict vq,
1308                      struct rte_mbuf **__rte_restrict pkts,
1309                      uint32_t count)
1310 {
1311         uint32_t pkt_idx = 0;
1312
1313         do {
1314                 rte_prefetch0(&vq->desc_packed[vq->last_avail_idx]);
1315
1316                 if (count - pkt_idx >= PACKED_BATCH_SIZE) {
1317                         if (!virtio_dev_rx_sync_batch_packed(dev, vq,
1318                                                         &pkts[pkt_idx])) {
1319                                 pkt_idx += PACKED_BATCH_SIZE;
1320                                 continue;
1321                         }
1322                 }
1323
1324                 if (virtio_dev_rx_single_packed(dev, vq, pkts[pkt_idx]))
1325                         break;
1326                 pkt_idx++;
1327
1328         } while (pkt_idx < count);
1329
1330         if (vq->shadow_used_idx) {
1331                 do_data_copy_enqueue(dev, vq);
1332                 vhost_flush_enqueue_shadow_packed(dev, vq);
1333         }
1334
1335         if (pkt_idx)
1336                 vhost_vring_call_packed(dev, vq);
1337
1338         return pkt_idx;
1339 }
1340
1341 static __rte_always_inline uint32_t
1342 virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id,
1343         struct rte_mbuf **pkts, uint32_t count)
1344 {
1345         struct vhost_virtqueue *vq;
1346         uint32_t nb_tx = 0;
1347
1348         VHOST_LOG_DATA(DEBUG, "(%s) %s\n", dev->ifname, __func__);
1349         if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) {
1350                 VHOST_LOG_DATA(ERR, "(%s) %s: invalid virtqueue idx %d.\n",
1351                         dev->ifname, __func__, queue_id);
1352                 return 0;
1353         }
1354
1355         vq = dev->virtqueue[queue_id];
1356
1357         rte_spinlock_lock(&vq->access_lock);
1358
1359         if (unlikely(!vq->enabled))
1360                 goto out_access_unlock;
1361
1362         if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
1363                 vhost_user_iotlb_rd_lock(vq);
1364
1365         if (unlikely(!vq->access_ok))
1366                 if (unlikely(vring_translate(dev, vq) < 0))
1367                         goto out;
1368
1369         count = RTE_MIN((uint32_t)MAX_PKT_BURST, count);
1370         if (count == 0)
1371                 goto out;
1372
1373         if (vq_is_packed(dev))
1374                 nb_tx = virtio_dev_rx_packed(dev, vq, pkts, count);
1375         else
1376                 nb_tx = virtio_dev_rx_split(dev, vq, pkts, count);
1377
1378 out:
1379         if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
1380                 vhost_user_iotlb_rd_unlock(vq);
1381
1382 out_access_unlock:
1383         rte_spinlock_unlock(&vq->access_lock);
1384
1385         return nb_tx;
1386 }
1387
1388 uint16_t
1389 rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
1390         struct rte_mbuf **__rte_restrict pkts, uint16_t count)
1391 {
1392         struct virtio_net *dev = get_device(vid);
1393
1394         if (!dev)
1395                 return 0;
1396
1397         if (unlikely(!(dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET))) {
1398                 VHOST_LOG_DATA(ERR, "(%s) %s: built-in vhost net backend is disabled.\n",
1399                         dev->ifname, __func__);
1400                 return 0;
1401         }
1402
1403         return virtio_dev_rx(dev, queue_id, pkts, count);
1404 }
1405
1406 static __rte_always_inline uint16_t
1407 async_get_first_inflight_pkt_idx(struct vhost_virtqueue *vq)
1408 {
1409         struct vhost_async *async = vq->async;
1410
1411         if (async->pkts_idx >= async->pkts_inflight_n)
1412                 return async->pkts_idx - async->pkts_inflight_n;
1413         else
1414                 return vq->size - async->pkts_inflight_n + async->pkts_idx;
1415 }
1416
1417 static __rte_always_inline void
1418 store_dma_desc_info_split(struct vring_used_elem *s_ring, struct vring_used_elem *d_ring,
1419                 uint16_t ring_size, uint16_t s_idx, uint16_t d_idx, uint16_t count)
1420 {
1421         size_t elem_size = sizeof(struct vring_used_elem);
1422
1423         if (d_idx + count <= ring_size) {
1424                 rte_memcpy(d_ring + d_idx, s_ring + s_idx, count * elem_size);
1425         } else {
1426                 uint16_t size = ring_size - d_idx;
1427
1428                 rte_memcpy(d_ring + d_idx, s_ring + s_idx, size * elem_size);
1429                 rte_memcpy(d_ring, s_ring + s_idx + size, (count - size) * elem_size);
1430         }
1431 }
1432
1433 static __rte_always_inline void
1434 store_dma_desc_info_packed(struct vring_used_elem_packed *s_ring,
1435                 struct vring_used_elem_packed *d_ring,
1436                 uint16_t ring_size, uint16_t s_idx, uint16_t d_idx, uint16_t count)
1437 {
1438         size_t elem_size = sizeof(struct vring_used_elem_packed);
1439
1440         if (d_idx + count <= ring_size) {
1441                 rte_memcpy(d_ring + d_idx, s_ring + s_idx, count * elem_size);
1442         } else {
1443                 uint16_t size = ring_size - d_idx;
1444
1445                 rte_memcpy(d_ring + d_idx, s_ring + s_idx, size * elem_size);
1446                 rte_memcpy(d_ring, s_ring + s_idx + size, (count - size) * elem_size);
1447         }
1448 }
1449
1450 static __rte_noinline uint32_t
1451 virtio_dev_rx_async_submit_split(struct virtio_net *dev,
1452         struct vhost_virtqueue *vq, uint16_t queue_id,
1453         struct rte_mbuf **pkts, uint32_t count)
1454 {
1455         struct buf_vector buf_vec[BUF_VECTOR_MAX];
1456         uint32_t pkt_idx = 0;
1457         uint16_t num_buffers;
1458         uint16_t avail_head;
1459
1460         struct vhost_async *async = vq->async;
1461         struct async_inflight_info *pkts_info = async->pkts_info;
1462         uint32_t pkt_err = 0;
1463         int32_t n_xfer;
1464         uint16_t slot_idx = 0;
1465
1466         /*
1467          * The ordering between avail index and desc reads need to be enforced.
1468          */
1469         avail_head = __atomic_load_n(&vq->avail->idx, __ATOMIC_ACQUIRE);
1470
1471         rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]);
1472
1473         async_iter_reset(async);
1474
1475         for (pkt_idx = 0; pkt_idx < count; pkt_idx++) {
1476                 uint32_t pkt_len = pkts[pkt_idx]->pkt_len + dev->vhost_hlen;
1477                 uint16_t nr_vec = 0;
1478
1479                 if (unlikely(reserve_avail_buf_split(dev, vq, pkt_len, buf_vec,
1480                                                 &num_buffers, avail_head, &nr_vec) < 0)) {
1481                         VHOST_LOG_DATA(DEBUG, "(%s) failed to get enough desc from vring\n",
1482                                         dev->ifname);
1483                         vq->shadow_used_idx -= num_buffers;
1484                         break;
1485                 }
1486
1487                 VHOST_LOG_DATA(DEBUG, "(%s) current index %d | end index %d\n",
1488                         dev->ifname, vq->last_avail_idx, vq->last_avail_idx + num_buffers);
1489
1490                 if (mbuf_to_desc(dev, vq, pkts[pkt_idx], buf_vec, nr_vec, num_buffers, true) < 0) {
1491                         vq->shadow_used_idx -= num_buffers;
1492                         break;
1493                 }
1494
1495                 slot_idx = (async->pkts_idx + pkt_idx) & (vq->size - 1);
1496                 pkts_info[slot_idx].descs = num_buffers;
1497                 pkts_info[slot_idx].mbuf = pkts[pkt_idx];
1498
1499                 vq->last_avail_idx += num_buffers;
1500         }
1501
1502         if (unlikely(pkt_idx == 0))
1503                 return 0;
1504
1505         n_xfer = async->ops.transfer_data(dev->vid, queue_id, async->iov_iter, 0, pkt_idx);
1506         if (unlikely(n_xfer < 0)) {
1507                 VHOST_LOG_DATA(ERR, "(%s) %s: failed to transfer data for queue id %d.\n",
1508                                 dev->ifname, __func__, queue_id);
1509                 n_xfer = 0;
1510         }
1511
1512         pkt_err = pkt_idx - n_xfer;
1513         if (unlikely(pkt_err)) {
1514                 uint16_t num_descs = 0;
1515
1516                 /* update number of completed packets */
1517                 pkt_idx = n_xfer;
1518
1519                 /* calculate the sum of descriptors to revert */
1520                 while (pkt_err-- > 0) {
1521                         num_descs += pkts_info[slot_idx & (vq->size - 1)].descs;
1522                         slot_idx--;
1523                 }
1524
1525                 /* recover shadow used ring and available ring */
1526                 vq->shadow_used_idx -= num_descs;
1527                 vq->last_avail_idx -= num_descs;
1528         }
1529
1530         /* keep used descriptors */
1531         if (likely(vq->shadow_used_idx)) {
1532                 uint16_t to = async->desc_idx_split & (vq->size - 1);
1533
1534                 store_dma_desc_info_split(vq->shadow_used_split,
1535                                 async->descs_split, vq->size, 0, to,
1536                                 vq->shadow_used_idx);
1537
1538                 async->desc_idx_split += vq->shadow_used_idx;
1539
1540                 async->pkts_idx += pkt_idx;
1541                 if (async->pkts_idx >= vq->size)
1542                         async->pkts_idx -= vq->size;
1543
1544                 async->pkts_inflight_n += pkt_idx;
1545                 vq->shadow_used_idx = 0;
1546         }
1547
1548         return pkt_idx;
1549 }
1550
1551
1552 static __rte_always_inline int
1553 vhost_enqueue_async_packed(struct virtio_net *dev,
1554                             struct vhost_virtqueue *vq,
1555                             struct rte_mbuf *pkt,
1556                             struct buf_vector *buf_vec,
1557                             uint16_t *nr_descs,
1558                             uint16_t *nr_buffers)
1559 {
1560         uint16_t nr_vec = 0;
1561         uint16_t avail_idx = vq->last_avail_idx;
1562         uint16_t max_tries, tries = 0;
1563         uint16_t buf_id = 0;
1564         uint32_t len = 0;
1565         uint16_t desc_count = 0;
1566         uint32_t size = pkt->pkt_len + sizeof(struct virtio_net_hdr_mrg_rxbuf);
1567         uint32_t buffer_len[vq->size];
1568         uint16_t buffer_buf_id[vq->size];
1569         uint16_t buffer_desc_count[vq->size];
1570
1571         if (rxvq_is_mergeable(dev))
1572                 max_tries = vq->size - 1;
1573         else
1574                 max_tries = 1;
1575
1576         while (size > 0) {
1577                 /*
1578                  * if we tried all available ring items, and still
1579                  * can't get enough buf, it means something abnormal
1580                  * happened.
1581                  */
1582                 if (unlikely(++tries > max_tries))
1583                         return -1;
1584
1585                 if (unlikely(fill_vec_buf_packed(dev, vq,
1586                                                 avail_idx, &desc_count,
1587                                                 buf_vec, &nr_vec,
1588                                                 &buf_id, &len,
1589                                                 VHOST_ACCESS_RW) < 0))
1590                         return -1;
1591
1592                 len = RTE_MIN(len, size);
1593                 size -= len;
1594
1595                 buffer_len[*nr_buffers] = len;
1596                 buffer_buf_id[*nr_buffers] = buf_id;
1597                 buffer_desc_count[*nr_buffers] = desc_count;
1598                 *nr_buffers += 1;
1599                 *nr_descs += desc_count;
1600                 avail_idx += desc_count;
1601                 if (avail_idx >= vq->size)
1602                         avail_idx -= vq->size;
1603         }
1604
1605         if (unlikely(mbuf_to_desc(dev, vq, pkt, buf_vec, nr_vec, *nr_buffers, true) < 0))
1606                 return -1;
1607
1608         vhost_shadow_enqueue_packed(vq, buffer_len, buffer_buf_id, buffer_desc_count, *nr_buffers);
1609
1610         return 0;
1611 }
1612
1613 static __rte_always_inline int16_t
1614 virtio_dev_rx_async_packed(struct virtio_net *dev, struct vhost_virtqueue *vq,
1615                             struct rte_mbuf *pkt, uint16_t *nr_descs, uint16_t *nr_buffers)
1616 {
1617         struct buf_vector buf_vec[BUF_VECTOR_MAX];
1618
1619         if (unlikely(vhost_enqueue_async_packed(dev, vq, pkt, buf_vec,
1620                                         nr_descs, nr_buffers) < 0)) {
1621                 VHOST_LOG_DATA(DEBUG, "(%s) failed to get enough desc from vring\n", dev->ifname);
1622                 return -1;
1623         }
1624
1625         VHOST_LOG_DATA(DEBUG, "(%s) current index %d | end index %d\n",
1626                         dev->ifname, vq->last_avail_idx, vq->last_avail_idx + *nr_descs);
1627
1628         return 0;
1629 }
1630
1631 static __rte_always_inline void
1632 dma_error_handler_packed(struct vhost_virtqueue *vq, uint16_t slot_idx,
1633                         uint32_t nr_err, uint32_t *pkt_idx)
1634 {
1635         uint16_t descs_err = 0;
1636         uint16_t buffers_err = 0;
1637         struct async_inflight_info *pkts_info = vq->async->pkts_info;
1638
1639         *pkt_idx -= nr_err;
1640         /* calculate the sum of buffers and descs of DMA-error packets. */
1641         while (nr_err-- > 0) {
1642                 descs_err += pkts_info[slot_idx % vq->size].descs;
1643                 buffers_err += pkts_info[slot_idx % vq->size].nr_buffers;
1644                 slot_idx--;
1645         }
1646
1647         if (vq->last_avail_idx >= descs_err) {
1648                 vq->last_avail_idx -= descs_err;
1649         } else {
1650                 vq->last_avail_idx = vq->last_avail_idx + vq->size - descs_err;
1651                 vq->avail_wrap_counter ^= 1;
1652         }
1653
1654         vq->shadow_used_idx -= buffers_err;
1655 }
1656
1657 static __rte_noinline uint32_t
1658 virtio_dev_rx_async_submit_packed(struct virtio_net *dev,
1659         struct vhost_virtqueue *vq, uint16_t queue_id,
1660         struct rte_mbuf **pkts, uint32_t count)
1661 {
1662         uint32_t pkt_idx = 0;
1663         uint32_t remained = count;
1664         int32_t n_xfer;
1665         uint16_t num_buffers;
1666         uint16_t num_descs;
1667
1668         struct vhost_async *async = vq->async;
1669         struct async_inflight_info *pkts_info = async->pkts_info;
1670         uint32_t pkt_err = 0;
1671         uint16_t slot_idx = 0;
1672
1673         do {
1674                 rte_prefetch0(&vq->desc_packed[vq->last_avail_idx]);
1675
1676                 num_buffers = 0;
1677                 num_descs = 0;
1678                 if (unlikely(virtio_dev_rx_async_packed(dev, vq, pkts[pkt_idx],
1679                                                 &num_descs, &num_buffers) < 0))
1680                         break;
1681
1682                 slot_idx = (async->pkts_idx + pkt_idx) % vq->size;
1683
1684                 pkts_info[slot_idx].descs = num_descs;
1685                 pkts_info[slot_idx].nr_buffers = num_buffers;
1686                 pkts_info[slot_idx].mbuf = pkts[pkt_idx];
1687
1688                 pkt_idx++;
1689                 remained--;
1690                 vq_inc_last_avail_packed(vq, num_descs);
1691         } while (pkt_idx < count);
1692
1693         if (unlikely(pkt_idx == 0))
1694                 return 0;
1695
1696         n_xfer = async->ops.transfer_data(dev->vid, queue_id, async->iov_iter, 0, pkt_idx);
1697         if (unlikely(n_xfer < 0)) {
1698                 VHOST_LOG_DATA(ERR, "(%s) %s: failed to transfer data for queue id %d.\n",
1699                                 dev->ifname, __func__, queue_id);
1700                 n_xfer = 0;
1701         }
1702
1703         pkt_err = pkt_idx - n_xfer;
1704
1705         async_iter_reset(async);
1706
1707         if (unlikely(pkt_err))
1708                 dma_error_handler_packed(vq, slot_idx, pkt_err, &pkt_idx);
1709
1710         if (likely(vq->shadow_used_idx)) {
1711                 /* keep used descriptors. */
1712                 store_dma_desc_info_packed(vq->shadow_used_packed, async->buffers_packed,
1713                                         vq->size, 0, async->buffer_idx_packed,
1714                                         vq->shadow_used_idx);
1715
1716                 async->buffer_idx_packed += vq->shadow_used_idx;
1717                 if (async->buffer_idx_packed >= vq->size)
1718                         async->buffer_idx_packed -= vq->size;
1719
1720                 async->pkts_idx += pkt_idx;
1721                 if (async->pkts_idx >= vq->size)
1722                         async->pkts_idx -= vq->size;
1723
1724                 vq->shadow_used_idx = 0;
1725                 async->pkts_inflight_n += pkt_idx;
1726         }
1727
1728         return pkt_idx;
1729 }
1730
1731 static __rte_always_inline void
1732 write_back_completed_descs_split(struct vhost_virtqueue *vq, uint16_t n_descs)
1733 {
1734         struct vhost_async *async = vq->async;
1735         uint16_t nr_left = n_descs;
1736         uint16_t nr_copy;
1737         uint16_t to, from;
1738
1739         do {
1740                 from = async->last_desc_idx_split & (vq->size - 1);
1741                 nr_copy = nr_left + from <= vq->size ? nr_left : vq->size - from;
1742                 to = vq->last_used_idx & (vq->size - 1);
1743
1744                 if (to + nr_copy <= vq->size) {
1745                         rte_memcpy(&vq->used->ring[to], &async->descs_split[from],
1746                                         nr_copy * sizeof(struct vring_used_elem));
1747                 } else {
1748                         uint16_t size = vq->size - to;
1749
1750                         rte_memcpy(&vq->used->ring[to], &async->descs_split[from],
1751                                         size * sizeof(struct vring_used_elem));
1752                         rte_memcpy(&vq->used->ring[0], &async->descs_split[from + size],
1753                                         (nr_copy - size) * sizeof(struct vring_used_elem));
1754                 }
1755
1756                 async->last_desc_idx_split += nr_copy;
1757                 vq->last_used_idx += nr_copy;
1758                 nr_left -= nr_copy;
1759         } while (nr_left > 0);
1760 }
1761
1762 static __rte_always_inline void
1763 write_back_completed_descs_packed(struct vhost_virtqueue *vq,
1764                                 uint16_t n_buffers)
1765 {
1766         struct vhost_async *async = vq->async;
1767         uint16_t from = async->last_buffer_idx_packed;
1768         uint16_t used_idx = vq->last_used_idx;
1769         uint16_t head_idx = vq->last_used_idx;
1770         uint16_t head_flags = 0;
1771         uint16_t i;
1772
1773         /* Split loop in two to save memory barriers */
1774         for (i = 0; i < n_buffers; i++) {
1775                 vq->desc_packed[used_idx].id = async->buffers_packed[from].id;
1776                 vq->desc_packed[used_idx].len = async->buffers_packed[from].len;
1777
1778                 used_idx += async->buffers_packed[from].count;
1779                 if (used_idx >= vq->size)
1780                         used_idx -= vq->size;
1781
1782                 from++;
1783                 if (from >= vq->size)
1784                         from = 0;
1785         }
1786
1787         /* The ordering for storing desc flags needs to be enforced. */
1788         rte_atomic_thread_fence(__ATOMIC_RELEASE);
1789
1790         from = async->last_buffer_idx_packed;
1791
1792         for (i = 0; i < n_buffers; i++) {
1793                 uint16_t flags;
1794
1795                 if (async->buffers_packed[from].len)
1796                         flags = VRING_DESC_F_WRITE;
1797                 else
1798                         flags = 0;
1799
1800                 if (vq->used_wrap_counter) {
1801                         flags |= VRING_DESC_F_USED;
1802                         flags |= VRING_DESC_F_AVAIL;
1803                 } else {
1804                         flags &= ~VRING_DESC_F_USED;
1805                         flags &= ~VRING_DESC_F_AVAIL;
1806                 }
1807
1808                 if (i > 0) {
1809                         vq->desc_packed[vq->last_used_idx].flags = flags;
1810                 } else {
1811                         head_idx = vq->last_used_idx;
1812                         head_flags = flags;
1813                 }
1814
1815                 vq_inc_last_used_packed(vq, async->buffers_packed[from].count);
1816
1817                 from++;
1818                 if (from == vq->size)
1819                         from = 0;
1820         }
1821
1822         vq->desc_packed[head_idx].flags = head_flags;
1823         async->last_buffer_idx_packed = from;
1824 }
1825
1826 static __rte_always_inline uint16_t
1827 vhost_poll_enqueue_completed(struct virtio_net *dev, uint16_t queue_id,
1828                 struct rte_mbuf **pkts, uint16_t count)
1829 {
1830         struct vhost_virtqueue *vq = dev->virtqueue[queue_id];
1831         struct vhost_async *async = vq->async;
1832         struct async_inflight_info *pkts_info = async->pkts_info;
1833         int32_t n_cpl;
1834         uint16_t n_descs = 0, n_buffers = 0;
1835         uint16_t start_idx, from, i;
1836
1837         n_cpl = async->ops.check_completed_copies(dev->vid, queue_id, 0, count);
1838         if (unlikely(n_cpl < 0)) {
1839                 VHOST_LOG_DATA(ERR, "(%s) %s: failed to check completed copies for queue id %d.\n",
1840                                 dev->ifname, __func__, queue_id);
1841                 return 0;
1842         }
1843
1844         if (n_cpl == 0)
1845                 return 0;
1846
1847         start_idx = async_get_first_inflight_pkt_idx(vq);
1848
1849         for (i = 0; i < n_cpl; i++) {
1850                 from = (start_idx + i) % vq->size;
1851                 /* Only used with packed ring */
1852                 n_buffers += pkts_info[from].nr_buffers;
1853                 /* Only used with split ring */
1854                 n_descs += pkts_info[from].descs;
1855                 pkts[i] = pkts_info[from].mbuf;
1856         }
1857
1858         async->pkts_inflight_n -= n_cpl;
1859
1860         if (likely(vq->enabled && vq->access_ok)) {
1861                 if (vq_is_packed(dev)) {
1862                         write_back_completed_descs_packed(vq, n_buffers);
1863                         vhost_vring_call_packed(dev, vq);
1864                 } else {
1865                         write_back_completed_descs_split(vq, n_descs);
1866                         __atomic_add_fetch(&vq->used->idx, n_descs, __ATOMIC_RELEASE);
1867                         vhost_vring_call_split(dev, vq);
1868                 }
1869         } else {
1870                 if (vq_is_packed(dev)) {
1871                         async->last_buffer_idx_packed += n_buffers;
1872                         if (async->last_buffer_idx_packed >= vq->size)
1873                                 async->last_buffer_idx_packed -= vq->size;
1874                 } else {
1875                         async->last_desc_idx_split += n_descs;
1876                 }
1877         }
1878
1879         return n_cpl;
1880 }
1881
1882 uint16_t
1883 rte_vhost_poll_enqueue_completed(int vid, uint16_t queue_id,
1884                 struct rte_mbuf **pkts, uint16_t count)
1885 {
1886         struct virtio_net *dev = get_device(vid);
1887         struct vhost_virtqueue *vq;
1888         uint16_t n_pkts_cpl = 0;
1889
1890         if (unlikely(!dev))
1891                 return 0;
1892
1893         VHOST_LOG_DATA(DEBUG, "(%s) %s\n", dev->ifname, __func__);
1894         if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) {
1895                 VHOST_LOG_DATA(ERR, "(%s) %s: invalid virtqueue idx %d.\n",
1896                         dev->ifname, __func__, queue_id);
1897                 return 0;
1898         }
1899
1900         vq = dev->virtqueue[queue_id];
1901
1902         if (unlikely(!vq->async)) {
1903                 VHOST_LOG_DATA(ERR, "(%s) %s: async not registered for queue id %d.\n",
1904                         dev->ifname, __func__, queue_id);
1905                 return 0;
1906         }
1907
1908         rte_spinlock_lock(&vq->access_lock);
1909
1910         n_pkts_cpl = vhost_poll_enqueue_completed(dev, queue_id, pkts, count);
1911
1912         rte_spinlock_unlock(&vq->access_lock);
1913
1914         return n_pkts_cpl;
1915 }
1916
1917 uint16_t
1918 rte_vhost_clear_queue_thread_unsafe(int vid, uint16_t queue_id,
1919                 struct rte_mbuf **pkts, uint16_t count)
1920 {
1921         struct virtio_net *dev = get_device(vid);
1922         struct vhost_virtqueue *vq;
1923         uint16_t n_pkts_cpl = 0;
1924
1925         if (!dev)
1926                 return 0;
1927
1928         VHOST_LOG_DATA(DEBUG, "(%s) %s\n", dev->ifname, __func__);
1929         if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) {
1930                 VHOST_LOG_DATA(ERR, "(%s) %s: invalid virtqueue idx %d.\n",
1931                         dev->ifname, __func__, queue_id);
1932                 return 0;
1933         }
1934
1935         vq = dev->virtqueue[queue_id];
1936
1937         if (unlikely(!vq->async)) {
1938                 VHOST_LOG_DATA(ERR, "(%s) %s: async not registered for queue id %d.\n",
1939                         dev->ifname, __func__, queue_id);
1940                 return 0;
1941         }
1942
1943         n_pkts_cpl = vhost_poll_enqueue_completed(dev, queue_id, pkts, count);
1944
1945         return n_pkts_cpl;
1946 }
1947
1948 static __rte_always_inline uint32_t
1949 virtio_dev_rx_async_submit(struct virtio_net *dev, uint16_t queue_id,
1950         struct rte_mbuf **pkts, uint32_t count)
1951 {
1952         struct vhost_virtqueue *vq;
1953         uint32_t nb_tx = 0;
1954
1955         VHOST_LOG_DATA(DEBUG, "(%s) %s\n", dev->ifname, __func__);
1956         if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) {
1957                 VHOST_LOG_DATA(ERR, "(%s) %s: invalid virtqueue idx %d.\n",
1958                         dev->ifname, __func__, queue_id);
1959                 return 0;
1960         }
1961
1962         vq = dev->virtqueue[queue_id];
1963
1964         rte_spinlock_lock(&vq->access_lock);
1965
1966         if (unlikely(!vq->enabled || !vq->async))
1967                 goto out_access_unlock;
1968
1969         if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
1970                 vhost_user_iotlb_rd_lock(vq);
1971
1972         if (unlikely(!vq->access_ok))
1973                 if (unlikely(vring_translate(dev, vq) < 0))
1974                         goto out;
1975
1976         count = RTE_MIN((uint32_t)MAX_PKT_BURST, count);
1977         if (count == 0)
1978                 goto out;
1979
1980         if (vq_is_packed(dev))
1981                 nb_tx = virtio_dev_rx_async_submit_packed(dev, vq, queue_id,
1982                                 pkts, count);
1983         else
1984                 nb_tx = virtio_dev_rx_async_submit_split(dev, vq, queue_id,
1985                                 pkts, count);
1986
1987 out:
1988         if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
1989                 vhost_user_iotlb_rd_unlock(vq);
1990
1991 out_access_unlock:
1992         rte_spinlock_unlock(&vq->access_lock);
1993
1994         return nb_tx;
1995 }
1996
1997 uint16_t
1998 rte_vhost_submit_enqueue_burst(int vid, uint16_t queue_id,
1999                 struct rte_mbuf **pkts, uint16_t count)
2000 {
2001         struct virtio_net *dev = get_device(vid);
2002
2003         if (!dev)
2004                 return 0;
2005
2006         if (unlikely(!(dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET))) {
2007                 VHOST_LOG_DATA(ERR, "(%s) %s: built-in vhost net backend is disabled.\n",
2008                         dev->ifname, __func__);
2009                 return 0;
2010         }
2011
2012         return virtio_dev_rx_async_submit(dev, queue_id, pkts, count);
2013 }
2014
2015 static inline bool
2016 virtio_net_with_host_offload(struct virtio_net *dev)
2017 {
2018         if (dev->features &
2019                         ((1ULL << VIRTIO_NET_F_CSUM) |
2020                          (1ULL << VIRTIO_NET_F_HOST_ECN) |
2021                          (1ULL << VIRTIO_NET_F_HOST_TSO4) |
2022                          (1ULL << VIRTIO_NET_F_HOST_TSO6) |
2023                          (1ULL << VIRTIO_NET_F_HOST_UFO)))
2024                 return true;
2025
2026         return false;
2027 }
2028
2029 static int
2030 parse_headers(struct rte_mbuf *m, uint8_t *l4_proto)
2031 {
2032         struct rte_ipv4_hdr *ipv4_hdr;
2033         struct rte_ipv6_hdr *ipv6_hdr;
2034         struct rte_ether_hdr *eth_hdr;
2035         uint16_t ethertype;
2036         uint16_t data_len = rte_pktmbuf_data_len(m);
2037
2038         if (data_len < sizeof(struct rte_ether_hdr))
2039                 return -EINVAL;
2040
2041         eth_hdr = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
2042
2043         m->l2_len = sizeof(struct rte_ether_hdr);
2044         ethertype = rte_be_to_cpu_16(eth_hdr->ether_type);
2045
2046         if (ethertype == RTE_ETHER_TYPE_VLAN) {
2047                 if (data_len < sizeof(struct rte_ether_hdr) +
2048                                 sizeof(struct rte_vlan_hdr))
2049                         goto error;
2050
2051                 struct rte_vlan_hdr *vlan_hdr =
2052                         (struct rte_vlan_hdr *)(eth_hdr + 1);
2053
2054                 m->l2_len += sizeof(struct rte_vlan_hdr);
2055                 ethertype = rte_be_to_cpu_16(vlan_hdr->eth_proto);
2056         }
2057
2058         switch (ethertype) {
2059         case RTE_ETHER_TYPE_IPV4:
2060                 if (data_len < m->l2_len + sizeof(struct rte_ipv4_hdr))
2061                         goto error;
2062                 ipv4_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
2063                                 m->l2_len);
2064                 m->l3_len = rte_ipv4_hdr_len(ipv4_hdr);
2065                 if (data_len < m->l2_len + m->l3_len)
2066                         goto error;
2067                 m->ol_flags |= RTE_MBUF_F_TX_IPV4;
2068                 *l4_proto = ipv4_hdr->next_proto_id;
2069                 break;
2070         case RTE_ETHER_TYPE_IPV6:
2071                 if (data_len < m->l2_len + sizeof(struct rte_ipv6_hdr))
2072                         goto error;
2073                 ipv6_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
2074                                 m->l2_len);
2075                 m->l3_len = sizeof(struct rte_ipv6_hdr);
2076                 m->ol_flags |= RTE_MBUF_F_TX_IPV6;
2077                 *l4_proto = ipv6_hdr->proto;
2078                 break;
2079         default:
2080                 /* a valid L3 header is needed for further L4 parsing */
2081                 goto error;
2082         }
2083
2084         /* both CSUM and GSO need a valid L4 header */
2085         switch (*l4_proto) {
2086         case IPPROTO_TCP:
2087                 if (data_len < m->l2_len + m->l3_len +
2088                                 sizeof(struct rte_tcp_hdr))
2089                         goto error;
2090                 break;
2091         case IPPROTO_UDP:
2092                 if (data_len < m->l2_len + m->l3_len +
2093                                 sizeof(struct rte_udp_hdr))
2094                         goto error;
2095                 break;
2096         case IPPROTO_SCTP:
2097                 if (data_len < m->l2_len + m->l3_len +
2098                                 sizeof(struct rte_sctp_hdr))
2099                         goto error;
2100                 break;
2101         default:
2102                 goto error;
2103         }
2104
2105         return 0;
2106
2107 error:
2108         m->l2_len = 0;
2109         m->l3_len = 0;
2110         m->ol_flags = 0;
2111         return -EINVAL;
2112 }
2113
2114 static __rte_always_inline void
2115 vhost_dequeue_offload_legacy(struct virtio_net *dev, struct virtio_net_hdr *hdr,
2116                 struct rte_mbuf *m)
2117 {
2118         uint8_t l4_proto = 0;
2119         struct rte_tcp_hdr *tcp_hdr = NULL;
2120         uint16_t tcp_len;
2121         uint16_t data_len = rte_pktmbuf_data_len(m);
2122
2123         if (parse_headers(m, &l4_proto) < 0)
2124                 return;
2125
2126         if (hdr->flags == VIRTIO_NET_HDR_F_NEEDS_CSUM) {
2127                 if (hdr->csum_start == (m->l2_len + m->l3_len)) {
2128                         switch (hdr->csum_offset) {
2129                         case (offsetof(struct rte_tcp_hdr, cksum)):
2130                                 if (l4_proto != IPPROTO_TCP)
2131                                         goto error;
2132                                 m->ol_flags |= RTE_MBUF_F_TX_TCP_CKSUM;
2133                                 break;
2134                         case (offsetof(struct rte_udp_hdr, dgram_cksum)):
2135                                 if (l4_proto != IPPROTO_UDP)
2136                                         goto error;
2137                                 m->ol_flags |= RTE_MBUF_F_TX_UDP_CKSUM;
2138                                 break;
2139                         case (offsetof(struct rte_sctp_hdr, cksum)):
2140                                 if (l4_proto != IPPROTO_SCTP)
2141                                         goto error;
2142                                 m->ol_flags |= RTE_MBUF_F_TX_SCTP_CKSUM;
2143                                 break;
2144                         default:
2145                                 goto error;
2146                         }
2147                 } else {
2148                         goto error;
2149                 }
2150         }
2151
2152         if (hdr->gso_type != VIRTIO_NET_HDR_GSO_NONE) {
2153                 switch (hdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
2154                 case VIRTIO_NET_HDR_GSO_TCPV4:
2155                 case VIRTIO_NET_HDR_GSO_TCPV6:
2156                         if (l4_proto != IPPROTO_TCP)
2157                                 goto error;
2158                         tcp_hdr = rte_pktmbuf_mtod_offset(m,
2159                                         struct rte_tcp_hdr *,
2160                                         m->l2_len + m->l3_len);
2161                         tcp_len = (tcp_hdr->data_off & 0xf0) >> 2;
2162                         if (data_len < m->l2_len + m->l3_len + tcp_len)
2163                                 goto error;
2164                         m->ol_flags |= RTE_MBUF_F_TX_TCP_SEG;
2165                         m->tso_segsz = hdr->gso_size;
2166                         m->l4_len = tcp_len;
2167                         break;
2168                 case VIRTIO_NET_HDR_GSO_UDP:
2169                         if (l4_proto != IPPROTO_UDP)
2170                                 goto error;
2171                         m->ol_flags |= RTE_MBUF_F_TX_UDP_SEG;
2172                         m->tso_segsz = hdr->gso_size;
2173                         m->l4_len = sizeof(struct rte_udp_hdr);
2174                         break;
2175                 default:
2176                         VHOST_LOG_DATA(WARNING, "(%s) unsupported gso type %u.\n",
2177                                         dev->ifname, hdr->gso_type);
2178                         goto error;
2179                 }
2180         }
2181         return;
2182
2183 error:
2184         m->l2_len = 0;
2185         m->l3_len = 0;
2186         m->ol_flags = 0;
2187 }
2188
2189 static __rte_always_inline void
2190 vhost_dequeue_offload(struct virtio_net *dev, struct virtio_net_hdr *hdr,
2191                 struct rte_mbuf *m, bool legacy_ol_flags)
2192 {
2193         struct rte_net_hdr_lens hdr_lens;
2194         int l4_supported = 0;
2195         uint32_t ptype;
2196
2197         if (hdr->flags == 0 && hdr->gso_type == VIRTIO_NET_HDR_GSO_NONE)
2198                 return;
2199
2200         if (legacy_ol_flags) {
2201                 vhost_dequeue_offload_legacy(dev, hdr, m);
2202                 return;
2203         }
2204
2205         m->ol_flags |= RTE_MBUF_F_RX_IP_CKSUM_UNKNOWN;
2206
2207         ptype = rte_net_get_ptype(m, &hdr_lens, RTE_PTYPE_ALL_MASK);
2208         m->packet_type = ptype;
2209         if ((ptype & RTE_PTYPE_L4_MASK) == RTE_PTYPE_L4_TCP ||
2210             (ptype & RTE_PTYPE_L4_MASK) == RTE_PTYPE_L4_UDP ||
2211             (ptype & RTE_PTYPE_L4_MASK) == RTE_PTYPE_L4_SCTP)
2212                 l4_supported = 1;
2213
2214         /* According to Virtio 1.1 spec, the device only needs to look at
2215          * VIRTIO_NET_HDR_F_NEEDS_CSUM in the packet transmission path.
2216          * This differs from the processing incoming packets path where the
2217          * driver could rely on VIRTIO_NET_HDR_F_DATA_VALID flag set by the
2218          * device.
2219          *
2220          * 5.1.6.2.1 Driver Requirements: Packet Transmission
2221          * The driver MUST NOT set the VIRTIO_NET_HDR_F_DATA_VALID and
2222          * VIRTIO_NET_HDR_F_RSC_INFO bits in flags.
2223          *
2224          * 5.1.6.2.2 Device Requirements: Packet Transmission
2225          * The device MUST ignore flag bits that it does not recognize.
2226          */
2227         if (hdr->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
2228                 uint32_t hdrlen;
2229
2230                 hdrlen = hdr_lens.l2_len + hdr_lens.l3_len + hdr_lens.l4_len;
2231                 if (hdr->csum_start <= hdrlen && l4_supported != 0) {
2232                         m->ol_flags |= RTE_MBUF_F_RX_L4_CKSUM_NONE;
2233                 } else {
2234                         /* Unknown proto or tunnel, do sw cksum. We can assume
2235                          * the cksum field is in the first segment since the
2236                          * buffers we provided to the host are large enough.
2237                          * In case of SCTP, this will be wrong since it's a CRC
2238                          * but there's nothing we can do.
2239                          */
2240                         uint16_t csum = 0, off;
2241
2242                         if (rte_raw_cksum_mbuf(m, hdr->csum_start,
2243                                         rte_pktmbuf_pkt_len(m) - hdr->csum_start, &csum) < 0)
2244                                 return;
2245                         if (likely(csum != 0xffff))
2246                                 csum = ~csum;
2247                         off = hdr->csum_offset + hdr->csum_start;
2248                         if (rte_pktmbuf_data_len(m) >= off + 1)
2249                                 *rte_pktmbuf_mtod_offset(m, uint16_t *, off) = csum;
2250                 }
2251         }
2252
2253         if (hdr->gso_type != VIRTIO_NET_HDR_GSO_NONE) {
2254                 if (hdr->gso_size == 0)
2255                         return;
2256
2257                 switch (hdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
2258                 case VIRTIO_NET_HDR_GSO_TCPV4:
2259                 case VIRTIO_NET_HDR_GSO_TCPV6:
2260                         if ((ptype & RTE_PTYPE_L4_MASK) != RTE_PTYPE_L4_TCP)
2261                                 break;
2262                         m->ol_flags |= RTE_MBUF_F_RX_LRO | RTE_MBUF_F_RX_L4_CKSUM_NONE;
2263                         m->tso_segsz = hdr->gso_size;
2264                         break;
2265                 case VIRTIO_NET_HDR_GSO_UDP:
2266                         if ((ptype & RTE_PTYPE_L4_MASK) != RTE_PTYPE_L4_UDP)
2267                                 break;
2268                         m->ol_flags |= RTE_MBUF_F_RX_LRO | RTE_MBUF_F_RX_L4_CKSUM_NONE;
2269                         m->tso_segsz = hdr->gso_size;
2270                         break;
2271                 default:
2272                         break;
2273                 }
2274         }
2275 }
2276
2277 static __rte_noinline void
2278 copy_vnet_hdr_from_desc(struct virtio_net_hdr *hdr,
2279                 struct buf_vector *buf_vec)
2280 {
2281         uint64_t len;
2282         uint64_t remain = sizeof(struct virtio_net_hdr);
2283         uint64_t src;
2284         uint64_t dst = (uint64_t)(uintptr_t)hdr;
2285
2286         while (remain) {
2287                 len = RTE_MIN(remain, buf_vec->buf_len);
2288                 src = buf_vec->buf_addr;
2289                 rte_memcpy((void *)(uintptr_t)dst,
2290                                 (void *)(uintptr_t)src, len);
2291
2292                 remain -= len;
2293                 dst += len;
2294                 buf_vec++;
2295         }
2296 }
2297
2298 static __rte_always_inline int
2299 copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
2300                   struct buf_vector *buf_vec, uint16_t nr_vec,
2301                   struct rte_mbuf *m, struct rte_mempool *mbuf_pool,
2302                   bool legacy_ol_flags)
2303 {
2304         uint32_t buf_avail, buf_offset;
2305         uint64_t buf_addr, buf_len;
2306         uint32_t mbuf_avail, mbuf_offset;
2307         uint32_t cpy_len;
2308         struct rte_mbuf *cur = m, *prev = m;
2309         struct virtio_net_hdr tmp_hdr;
2310         struct virtio_net_hdr *hdr = NULL;
2311         /* A counter to avoid desc dead loop chain */
2312         uint16_t vec_idx = 0;
2313         struct batch_copy_elem *batch_copy = vq->batch_copy_elems;
2314         int error = 0;
2315
2316         buf_addr = buf_vec[vec_idx].buf_addr;
2317         buf_len = buf_vec[vec_idx].buf_len;
2318
2319         if (unlikely(buf_len < dev->vhost_hlen && nr_vec <= 1)) {
2320                 error = -1;
2321                 goto out;
2322         }
2323
2324         if (virtio_net_with_host_offload(dev)) {
2325                 if (unlikely(buf_len < sizeof(struct virtio_net_hdr))) {
2326                         /*
2327                          * No luck, the virtio-net header doesn't fit
2328                          * in a contiguous virtual area.
2329                          */
2330                         copy_vnet_hdr_from_desc(&tmp_hdr, buf_vec);
2331                         hdr = &tmp_hdr;
2332                 } else {
2333                         hdr = (struct virtio_net_hdr *)((uintptr_t)buf_addr);
2334                 }
2335         }
2336
2337         /*
2338          * A virtio driver normally uses at least 2 desc buffers
2339          * for Tx: the first for storing the header, and others
2340          * for storing the data.
2341          */
2342         if (unlikely(buf_len < dev->vhost_hlen)) {
2343                 buf_offset = dev->vhost_hlen - buf_len;
2344                 vec_idx++;
2345                 buf_addr = buf_vec[vec_idx].buf_addr;
2346                 buf_len = buf_vec[vec_idx].buf_len;
2347                 buf_avail  = buf_len - buf_offset;
2348         } else if (buf_len == dev->vhost_hlen) {
2349                 if (unlikely(++vec_idx >= nr_vec))
2350                         goto out;
2351                 buf_addr = buf_vec[vec_idx].buf_addr;
2352                 buf_len = buf_vec[vec_idx].buf_len;
2353
2354                 buf_offset = 0;
2355                 buf_avail = buf_len;
2356         } else {
2357                 buf_offset = dev->vhost_hlen;
2358                 buf_avail = buf_vec[vec_idx].buf_len - dev->vhost_hlen;
2359         }
2360
2361         PRINT_PACKET(dev,
2362                         (uintptr_t)(buf_addr + buf_offset),
2363                         (uint32_t)buf_avail, 0);
2364
2365         mbuf_offset = 0;
2366         mbuf_avail  = m->buf_len - RTE_PKTMBUF_HEADROOM;
2367         while (1) {
2368                 cpy_len = RTE_MIN(buf_avail, mbuf_avail);
2369
2370                 if (likely(cpy_len > MAX_BATCH_LEN ||
2371                                         vq->batch_copy_nb_elems >= vq->size ||
2372                                         (hdr && cur == m))) {
2373                         rte_memcpy(rte_pktmbuf_mtod_offset(cur, void *,
2374                                                 mbuf_offset),
2375                                         (void *)((uintptr_t)(buf_addr +
2376                                                         buf_offset)), cpy_len);
2377                 } else {
2378                         batch_copy[vq->batch_copy_nb_elems].dst =
2379                                 rte_pktmbuf_mtod_offset(cur, void *,
2380                                                 mbuf_offset);
2381                         batch_copy[vq->batch_copy_nb_elems].src =
2382                                 (void *)((uintptr_t)(buf_addr + buf_offset));
2383                         batch_copy[vq->batch_copy_nb_elems].len = cpy_len;
2384                         vq->batch_copy_nb_elems++;
2385                 }
2386
2387                 mbuf_avail  -= cpy_len;
2388                 mbuf_offset += cpy_len;
2389                 buf_avail -= cpy_len;
2390                 buf_offset += cpy_len;
2391
2392                 /* This buf reaches to its end, get the next one */
2393                 if (buf_avail == 0) {
2394                         if (++vec_idx >= nr_vec)
2395                                 break;
2396
2397                         buf_addr = buf_vec[vec_idx].buf_addr;
2398                         buf_len = buf_vec[vec_idx].buf_len;
2399
2400                         buf_offset = 0;
2401                         buf_avail  = buf_len;
2402
2403                         PRINT_PACKET(dev, (uintptr_t)buf_addr,
2404                                         (uint32_t)buf_avail, 0);
2405                 }
2406
2407                 /*
2408                  * This mbuf reaches to its end, get a new one
2409                  * to hold more data.
2410                  */
2411                 if (mbuf_avail == 0) {
2412                         cur = rte_pktmbuf_alloc(mbuf_pool);
2413                         if (unlikely(cur == NULL)) {
2414                                 VHOST_LOG_DATA(ERR, "(%s) failed to allocate memory for mbuf.\n",
2415                                                 dev->ifname);
2416                                 error = -1;
2417                                 goto out;
2418                         }
2419
2420                         prev->next = cur;
2421                         prev->data_len = mbuf_offset;
2422                         m->nb_segs += 1;
2423                         m->pkt_len += mbuf_offset;
2424                         prev = cur;
2425
2426                         mbuf_offset = 0;
2427                         mbuf_avail  = cur->buf_len - RTE_PKTMBUF_HEADROOM;
2428                 }
2429         }
2430
2431         prev->data_len = mbuf_offset;
2432         m->pkt_len    += mbuf_offset;
2433
2434         if (hdr)
2435                 vhost_dequeue_offload(dev, hdr, m, legacy_ol_flags);
2436
2437 out:
2438
2439         return error;
2440 }
2441
2442 static void
2443 virtio_dev_extbuf_free(void *addr __rte_unused, void *opaque)
2444 {
2445         rte_free(opaque);
2446 }
2447
2448 static int
2449 virtio_dev_extbuf_alloc(struct virtio_net *dev, struct rte_mbuf *pkt, uint32_t size)
2450 {
2451         struct rte_mbuf_ext_shared_info *shinfo = NULL;
2452         uint32_t total_len = RTE_PKTMBUF_HEADROOM + size;
2453         uint16_t buf_len;
2454         rte_iova_t iova;
2455         void *buf;
2456
2457         total_len += sizeof(*shinfo) + sizeof(uintptr_t);
2458         total_len = RTE_ALIGN_CEIL(total_len, sizeof(uintptr_t));
2459
2460         if (unlikely(total_len > UINT16_MAX))
2461                 return -ENOSPC;
2462
2463         buf_len = total_len;
2464         buf = rte_malloc(NULL, buf_len, RTE_CACHE_LINE_SIZE);
2465         if (unlikely(buf == NULL))
2466                 return -ENOMEM;
2467
2468         /* Initialize shinfo */
2469         shinfo = rte_pktmbuf_ext_shinfo_init_helper(buf, &buf_len,
2470                                                 virtio_dev_extbuf_free, buf);
2471         if (unlikely(shinfo == NULL)) {
2472                 rte_free(buf);
2473                 VHOST_LOG_DATA(ERR, "(%s) failed to init shinfo\n", dev->ifname);
2474                 return -1;
2475         }
2476
2477         iova = rte_malloc_virt2iova(buf);
2478         rte_pktmbuf_attach_extbuf(pkt, buf, iova, buf_len, shinfo);
2479         rte_pktmbuf_reset_headroom(pkt);
2480
2481         return 0;
2482 }
2483
2484 /*
2485  * Prepare a host supported pktmbuf.
2486  */
2487 static __rte_always_inline int
2488 virtio_dev_pktmbuf_prep(struct virtio_net *dev, struct rte_mbuf *pkt,
2489                          uint32_t data_len)
2490 {
2491         if (rte_pktmbuf_tailroom(pkt) >= data_len)
2492                 return 0;
2493
2494         /* attach an external buffer if supported */
2495         if (dev->extbuf && !virtio_dev_extbuf_alloc(dev, pkt, data_len))
2496                 return 0;
2497
2498         /* check if chained buffers are allowed */
2499         if (!dev->linearbuf)
2500                 return 0;
2501
2502         return -1;
2503 }
2504
2505 __rte_always_inline
2506 static uint16_t
2507 virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
2508         struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count,
2509         bool legacy_ol_flags)
2510 {
2511         uint16_t i;
2512         uint16_t free_entries;
2513         uint16_t dropped = 0;
2514         static bool allocerr_warned;
2515
2516         /*
2517          * The ordering between avail index and
2518          * desc reads needs to be enforced.
2519          */
2520         free_entries = __atomic_load_n(&vq->avail->idx, __ATOMIC_ACQUIRE) -
2521                         vq->last_avail_idx;
2522         if (free_entries == 0)
2523                 return 0;
2524
2525         rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]);
2526
2527         VHOST_LOG_DATA(DEBUG, "(%s) %s\n", dev->ifname, __func__);
2528
2529         count = RTE_MIN(count, MAX_PKT_BURST);
2530         count = RTE_MIN(count, free_entries);
2531         VHOST_LOG_DATA(DEBUG, "(%s) about to dequeue %u buffers\n",
2532                         dev->ifname, count);
2533
2534         if (rte_pktmbuf_alloc_bulk(mbuf_pool, pkts, count))
2535                 return 0;
2536
2537         for (i = 0; i < count; i++) {
2538                 struct buf_vector buf_vec[BUF_VECTOR_MAX];
2539                 uint16_t head_idx;
2540                 uint32_t buf_len;
2541                 uint16_t nr_vec = 0;
2542                 int err;
2543
2544                 if (unlikely(fill_vec_buf_split(dev, vq,
2545                                                 vq->last_avail_idx + i,
2546                                                 &nr_vec, buf_vec,
2547                                                 &head_idx, &buf_len,
2548                                                 VHOST_ACCESS_RO) < 0))
2549                         break;
2550
2551                 update_shadow_used_ring_split(vq, head_idx, 0);
2552
2553                 err = virtio_dev_pktmbuf_prep(dev, pkts[i], buf_len);
2554                 if (unlikely(err)) {
2555                         /*
2556                          * mbuf allocation fails for jumbo packets when external
2557                          * buffer allocation is not allowed and linear buffer
2558                          * is required. Drop this packet.
2559                          */
2560                         if (!allocerr_warned) {
2561                                 VHOST_LOG_DATA(ERR, "(%s) failed mbuf alloc of size %d from %s.\n",
2562                                         dev->ifname, buf_len, mbuf_pool->name);
2563                                 allocerr_warned = true;
2564                         }
2565                         dropped += 1;
2566                         i++;
2567                         break;
2568                 }
2569
2570                 err = copy_desc_to_mbuf(dev, vq, buf_vec, nr_vec, pkts[i],
2571                                 mbuf_pool, legacy_ol_flags);
2572                 if (unlikely(err)) {
2573                         if (!allocerr_warned) {
2574                                 VHOST_LOG_DATA(ERR, "(%s) failed to copy desc to mbuf.\n",
2575                                         dev->ifname);
2576                                 allocerr_warned = true;
2577                         }
2578                         dropped += 1;
2579                         i++;
2580                         break;
2581                 }
2582         }
2583
2584         if (dropped)
2585                 rte_pktmbuf_free_bulk(&pkts[i - 1], count - i + 1);
2586
2587         vq->last_avail_idx += i;
2588
2589         do_data_copy_dequeue(vq);
2590         if (unlikely(i < count))
2591                 vq->shadow_used_idx = i;
2592         if (likely(vq->shadow_used_idx)) {
2593                 flush_shadow_used_ring_split(dev, vq);
2594                 vhost_vring_call_split(dev, vq);
2595         }
2596
2597         return (i - dropped);
2598 }
2599
2600 __rte_noinline
2601 static uint16_t
2602 virtio_dev_tx_split_legacy(struct virtio_net *dev,
2603         struct vhost_virtqueue *vq, struct rte_mempool *mbuf_pool,
2604         struct rte_mbuf **pkts, uint16_t count)
2605 {
2606         return virtio_dev_tx_split(dev, vq, mbuf_pool, pkts, count, true);
2607 }
2608
2609 __rte_noinline
2610 static uint16_t
2611 virtio_dev_tx_split_compliant(struct virtio_net *dev,
2612         struct vhost_virtqueue *vq, struct rte_mempool *mbuf_pool,
2613         struct rte_mbuf **pkts, uint16_t count)
2614 {
2615         return virtio_dev_tx_split(dev, vq, mbuf_pool, pkts, count, false);
2616 }
2617
2618 static __rte_always_inline int
2619 vhost_reserve_avail_batch_packed(struct virtio_net *dev,
2620                                  struct vhost_virtqueue *vq,
2621                                  struct rte_mbuf **pkts,
2622                                  uint16_t avail_idx,
2623                                  uintptr_t *desc_addrs,
2624                                  uint16_t *ids)
2625 {
2626         bool wrap = vq->avail_wrap_counter;
2627         struct vring_packed_desc *descs = vq->desc_packed;
2628         uint64_t lens[PACKED_BATCH_SIZE];
2629         uint64_t buf_lens[PACKED_BATCH_SIZE];
2630         uint32_t buf_offset = sizeof(struct virtio_net_hdr_mrg_rxbuf);
2631         uint16_t flags, i;
2632
2633         if (unlikely(avail_idx & PACKED_BATCH_MASK))
2634                 return -1;
2635         if (unlikely((avail_idx + PACKED_BATCH_SIZE) > vq->size))
2636                 return -1;
2637
2638         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
2639                 flags = descs[avail_idx + i].flags;
2640                 if (unlikely((wrap != !!(flags & VRING_DESC_F_AVAIL)) ||
2641                              (wrap == !!(flags & VRING_DESC_F_USED))  ||
2642                              (flags & PACKED_DESC_SINGLE_DEQUEUE_FLAG)))
2643                         return -1;
2644         }
2645
2646         rte_atomic_thread_fence(__ATOMIC_ACQUIRE);
2647
2648         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
2649                 lens[i] = descs[avail_idx + i].len;
2650
2651         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
2652                 desc_addrs[i] = vhost_iova_to_vva(dev, vq,
2653                                                   descs[avail_idx + i].addr,
2654                                                   &lens[i], VHOST_ACCESS_RW);
2655         }
2656
2657         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
2658                 if (unlikely(!desc_addrs[i]))
2659                         return -1;
2660                 if (unlikely((lens[i] != descs[avail_idx + i].len)))
2661                         return -1;
2662         }
2663
2664         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
2665                 if (virtio_dev_pktmbuf_prep(dev, pkts[i], lens[i]))
2666                         goto err;
2667         }
2668
2669         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
2670                 buf_lens[i] = pkts[i]->buf_len - pkts[i]->data_off;
2671
2672         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
2673                 if (unlikely(buf_lens[i] < (lens[i] - buf_offset)))
2674                         goto err;
2675         }
2676
2677         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
2678                 pkts[i]->pkt_len = lens[i] - buf_offset;
2679                 pkts[i]->data_len = pkts[i]->pkt_len;
2680                 ids[i] = descs[avail_idx + i].id;
2681         }
2682
2683         return 0;
2684
2685 err:
2686         return -1;
2687 }
2688
2689 static __rte_always_inline int
2690 virtio_dev_tx_batch_packed(struct virtio_net *dev,
2691                            struct vhost_virtqueue *vq,
2692                            struct rte_mbuf **pkts,
2693                            bool legacy_ol_flags)
2694 {
2695         uint16_t avail_idx = vq->last_avail_idx;
2696         uint32_t buf_offset = sizeof(struct virtio_net_hdr_mrg_rxbuf);
2697         struct virtio_net_hdr *hdr;
2698         uintptr_t desc_addrs[PACKED_BATCH_SIZE];
2699         uint16_t ids[PACKED_BATCH_SIZE];
2700         uint16_t i;
2701
2702         if (vhost_reserve_avail_batch_packed(dev, vq, pkts, avail_idx,
2703                                              desc_addrs, ids))
2704                 return -1;
2705
2706         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
2707                 rte_prefetch0((void *)(uintptr_t)desc_addrs[i]);
2708
2709         vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
2710                 rte_memcpy(rte_pktmbuf_mtod_offset(pkts[i], void *, 0),
2711                            (void *)(uintptr_t)(desc_addrs[i] + buf_offset),
2712                            pkts[i]->pkt_len);
2713
2714         if (virtio_net_with_host_offload(dev)) {
2715                 vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
2716                         hdr = (struct virtio_net_hdr *)(desc_addrs[i]);
2717                         vhost_dequeue_offload(dev, hdr, pkts[i], legacy_ol_flags);
2718                 }
2719         }
2720
2721         if (virtio_net_is_inorder(dev))
2722                 vhost_shadow_dequeue_batch_packed_inorder(vq,
2723                         ids[PACKED_BATCH_SIZE - 1]);
2724         else
2725                 vhost_shadow_dequeue_batch_packed(dev, vq, ids);
2726
2727         vq_inc_last_avail_packed(vq, PACKED_BATCH_SIZE);
2728
2729         return 0;
2730 }
2731
2732 static __rte_always_inline int
2733 vhost_dequeue_single_packed(struct virtio_net *dev,
2734                             struct vhost_virtqueue *vq,
2735                             struct rte_mempool *mbuf_pool,
2736                             struct rte_mbuf *pkts,
2737                             uint16_t *buf_id,
2738                             uint16_t *desc_count,
2739                             bool legacy_ol_flags)
2740 {
2741         struct buf_vector buf_vec[BUF_VECTOR_MAX];
2742         uint32_t buf_len;
2743         uint16_t nr_vec = 0;
2744         int err;
2745         static bool allocerr_warned;
2746
2747         if (unlikely(fill_vec_buf_packed(dev, vq,
2748                                          vq->last_avail_idx, desc_count,
2749                                          buf_vec, &nr_vec,
2750                                          buf_id, &buf_len,
2751                                          VHOST_ACCESS_RO) < 0))
2752                 return -1;
2753
2754         if (unlikely(virtio_dev_pktmbuf_prep(dev, pkts, buf_len))) {
2755                 if (!allocerr_warned) {
2756                         VHOST_LOG_DATA(ERR, "(%s) failed mbuf alloc of size %d from %s.\n",
2757                                 dev->ifname, buf_len, mbuf_pool->name);
2758                         allocerr_warned = true;
2759                 }
2760                 return -1;
2761         }
2762
2763         err = copy_desc_to_mbuf(dev, vq, buf_vec, nr_vec, pkts,
2764                                 mbuf_pool, legacy_ol_flags);
2765         if (unlikely(err)) {
2766                 if (!allocerr_warned) {
2767                         VHOST_LOG_DATA(ERR, "(%s) failed to copy desc to mbuf.\n",
2768                                 dev->ifname);
2769                         allocerr_warned = true;
2770                 }
2771                 return -1;
2772         }
2773
2774         return 0;
2775 }
2776
2777 static __rte_always_inline int
2778 virtio_dev_tx_single_packed(struct virtio_net *dev,
2779                             struct vhost_virtqueue *vq,
2780                             struct rte_mempool *mbuf_pool,
2781                             struct rte_mbuf *pkts,
2782                             bool legacy_ol_flags)
2783 {
2784
2785         uint16_t buf_id, desc_count = 0;
2786         int ret;
2787
2788         ret = vhost_dequeue_single_packed(dev, vq, mbuf_pool, pkts, &buf_id,
2789                                         &desc_count, legacy_ol_flags);
2790
2791         if (likely(desc_count > 0)) {
2792                 if (virtio_net_is_inorder(dev))
2793                         vhost_shadow_dequeue_single_packed_inorder(vq, buf_id,
2794                                                                    desc_count);
2795                 else
2796                         vhost_shadow_dequeue_single_packed(vq, buf_id,
2797                                         desc_count);
2798
2799                 vq_inc_last_avail_packed(vq, desc_count);
2800         }
2801
2802         return ret;
2803 }
2804
2805 __rte_always_inline
2806 static uint16_t
2807 virtio_dev_tx_packed(struct virtio_net *dev,
2808                      struct vhost_virtqueue *__rte_restrict vq,
2809                      struct rte_mempool *mbuf_pool,
2810                      struct rte_mbuf **__rte_restrict pkts,
2811                      uint32_t count,
2812                      bool legacy_ol_flags)
2813 {
2814         uint32_t pkt_idx = 0;
2815
2816         if (rte_pktmbuf_alloc_bulk(mbuf_pool, pkts, count))
2817                 return 0;
2818
2819         do {
2820                 rte_prefetch0(&vq->desc_packed[vq->last_avail_idx]);
2821
2822                 if (count - pkt_idx >= PACKED_BATCH_SIZE) {
2823                         if (!virtio_dev_tx_batch_packed(dev, vq,
2824                                                         &pkts[pkt_idx],
2825                                                         legacy_ol_flags)) {
2826                                 pkt_idx += PACKED_BATCH_SIZE;
2827                                 continue;
2828                         }
2829                 }
2830
2831                 if (virtio_dev_tx_single_packed(dev, vq, mbuf_pool,
2832                                                 pkts[pkt_idx],
2833                                                 legacy_ol_flags))
2834                         break;
2835                 pkt_idx++;
2836         } while (pkt_idx < count);
2837
2838         if (pkt_idx != count)
2839                 rte_pktmbuf_free_bulk(&pkts[pkt_idx], count - pkt_idx);
2840
2841         if (vq->shadow_used_idx) {
2842                 do_data_copy_dequeue(vq);
2843
2844                 vhost_flush_dequeue_shadow_packed(dev, vq);
2845                 vhost_vring_call_packed(dev, vq);
2846         }
2847
2848         return pkt_idx;
2849 }
2850
2851 __rte_noinline
2852 static uint16_t
2853 virtio_dev_tx_packed_legacy(struct virtio_net *dev,
2854         struct vhost_virtqueue *__rte_restrict vq, struct rte_mempool *mbuf_pool,
2855         struct rte_mbuf **__rte_restrict pkts, uint32_t count)
2856 {
2857         return virtio_dev_tx_packed(dev, vq, mbuf_pool, pkts, count, true);
2858 }
2859
2860 __rte_noinline
2861 static uint16_t
2862 virtio_dev_tx_packed_compliant(struct virtio_net *dev,
2863         struct vhost_virtqueue *__rte_restrict vq, struct rte_mempool *mbuf_pool,
2864         struct rte_mbuf **__rte_restrict pkts, uint32_t count)
2865 {
2866         return virtio_dev_tx_packed(dev, vq, mbuf_pool, pkts, count, false);
2867 }
2868
2869 uint16_t
2870 rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
2871         struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count)
2872 {
2873         struct virtio_net *dev;
2874         struct rte_mbuf *rarp_mbuf = NULL;
2875         struct vhost_virtqueue *vq;
2876         int16_t success = 1;
2877
2878         dev = get_device(vid);
2879         if (!dev)
2880                 return 0;
2881
2882         if (unlikely(!(dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET))) {
2883                 VHOST_LOG_DATA(ERR, "(%s) %s: built-in vhost net backend is disabled.\n",
2884                                 dev->ifname, __func__);
2885                 return 0;
2886         }
2887
2888         if (unlikely(!is_valid_virt_queue_idx(queue_id, 1, dev->nr_vring))) {
2889                 VHOST_LOG_DATA(ERR, "(%s) %s: invalid virtqueue idx %d.\n",
2890                                 dev->ifname, __func__, queue_id);
2891                 return 0;
2892         }
2893
2894         vq = dev->virtqueue[queue_id];
2895
2896         if (unlikely(rte_spinlock_trylock(&vq->access_lock) == 0))
2897                 return 0;
2898
2899         if (unlikely(!vq->enabled)) {
2900                 count = 0;
2901                 goto out_access_unlock;
2902         }
2903
2904         if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
2905                 vhost_user_iotlb_rd_lock(vq);
2906
2907         if (unlikely(!vq->access_ok))
2908                 if (unlikely(vring_translate(dev, vq) < 0)) {
2909                         count = 0;
2910                         goto out;
2911                 }
2912
2913         /*
2914          * Construct a RARP broadcast packet, and inject it to the "pkts"
2915          * array, to looks like that guest actually send such packet.
2916          *
2917          * Check user_send_rarp() for more information.
2918          *
2919          * broadcast_rarp shares a cacheline in the virtio_net structure
2920          * with some fields that are accessed during enqueue and
2921          * __atomic_compare_exchange_n causes a write if performed compare
2922          * and exchange. This could result in false sharing between enqueue
2923          * and dequeue.
2924          *
2925          * Prevent unnecessary false sharing by reading broadcast_rarp first
2926          * and only performing compare and exchange if the read indicates it
2927          * is likely to be set.
2928          */
2929         if (unlikely(__atomic_load_n(&dev->broadcast_rarp, __ATOMIC_ACQUIRE) &&
2930                         __atomic_compare_exchange_n(&dev->broadcast_rarp,
2931                         &success, 0, 0, __ATOMIC_RELEASE, __ATOMIC_RELAXED))) {
2932
2933                 rarp_mbuf = rte_net_make_rarp_packet(mbuf_pool, &dev->mac);
2934                 if (rarp_mbuf == NULL) {
2935                         VHOST_LOG_DATA(ERR, "(%s) failed to make RARP packet.\n", dev->ifname);
2936                         count = 0;
2937                         goto out;
2938                 }
2939                 /*
2940                  * Inject it to the head of "pkts" array, so that switch's mac
2941                  * learning table will get updated first.
2942                  */
2943                 pkts[0] = rarp_mbuf;
2944                 pkts++;
2945                 count -= 1;
2946         }
2947
2948         if (vq_is_packed(dev)) {
2949                 if (dev->flags & VIRTIO_DEV_LEGACY_OL_FLAGS)
2950                         count = virtio_dev_tx_packed_legacy(dev, vq, mbuf_pool, pkts, count);
2951                 else
2952                         count = virtio_dev_tx_packed_compliant(dev, vq, mbuf_pool, pkts, count);
2953         } else {
2954                 if (dev->flags & VIRTIO_DEV_LEGACY_OL_FLAGS)
2955                         count = virtio_dev_tx_split_legacy(dev, vq, mbuf_pool, pkts, count);
2956                 else
2957                         count = virtio_dev_tx_split_compliant(dev, vq, mbuf_pool, pkts, count);
2958         }
2959
2960 out:
2961         if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
2962                 vhost_user_iotlb_rd_unlock(vq);
2963
2964 out_access_unlock:
2965         rte_spinlock_unlock(&vq->access_lock);
2966
2967         if (unlikely(rarp_mbuf != NULL))
2968                 count += 1;
2969
2970         return count;
2971 }