vhost: fix async copy on multi-page buffers
[dpdk.git] / lib / librte_vhost / virtio_net.c
index fe5cac0..bd9303c 100644 (file)
 #include <rte_arp.h>
 #include <rte_spinlock.h>
 #include <rte_malloc.h>
+#include <rte_vhost_async.h>
 
 #include "iotlb.h"
 #include "vhost.h"
 
-#define MAX_PKT_BURST 32
-
 #define MAX_BATCH_LEN 256
 
+#define VHOST_ASYNC_BATCH_THRESHOLD 32
+
 static  __rte_always_inline bool
 rxvq_is_mergeable(struct virtio_net *dev)
 {
@@ -116,6 +117,31 @@ flush_shadow_used_ring_split(struct virtio_net *dev, struct vhost_virtqueue *vq)
                sizeof(vq->used->idx));
 }
 
+static __rte_always_inline void
+async_flush_shadow_used_ring_split(struct virtio_net *dev,
+       struct vhost_virtqueue *vq)
+{
+       uint16_t used_idx = vq->last_used_idx & (vq->size - 1);
+
+       if (used_idx + vq->shadow_used_idx <= vq->size) {
+               do_flush_shadow_used_ring_split(dev, vq, used_idx, 0,
+                                         vq->shadow_used_idx);
+       } else {
+               uint16_t size;
+
+               /* update used ring interval [used_idx, vq->size] */
+               size = vq->size - used_idx;
+               do_flush_shadow_used_ring_split(dev, vq, used_idx, 0, size);
+
+               /* update the left half used ring interval [0, left_size] */
+               do_flush_shadow_used_ring_split(dev, vq, 0, size,
+                                         vq->shadow_used_idx - size);
+       }
+
+       vq->last_used_idx += vq->shadow_used_idx;
+       vq->shadow_used_idx = 0;
+}
+
 static __rte_always_inline void
 update_shadow_used_ring_split(struct vhost_virtqueue *vq,
                         uint16_t desc_idx, uint32_t len)
@@ -905,6 +931,209 @@ out:
        return error;
 }
 
+static __rte_always_inline void
+async_fill_vec(struct iovec *v, void *base, size_t len)
+{
+       v->iov_base = base;
+       v->iov_len = len;
+}
+
+static __rte_always_inline void
+async_fill_iter(struct rte_vhost_iov_iter *it, size_t count,
+       struct iovec *vec, unsigned long nr_seg)
+{
+       it->offset = 0;
+       it->count = count;
+
+       if (count) {
+               it->iov = vec;
+               it->nr_segs = nr_seg;
+       } else {
+               it->iov = 0;
+               it->nr_segs = 0;
+       }
+}
+
+static __rte_always_inline void
+async_fill_desc(struct rte_vhost_async_desc *desc,
+       struct rte_vhost_iov_iter *src, struct rte_vhost_iov_iter *dst)
+{
+       desc->src = src;
+       desc->dst = dst;
+}
+
+static __rte_always_inline int
+async_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
+                       struct rte_mbuf *m, struct buf_vector *buf_vec,
+                       uint16_t nr_vec, uint16_t num_buffers,
+                       struct iovec *src_iovec, struct iovec *dst_iovec,
+                       struct rte_vhost_iov_iter *src_it,
+                       struct rte_vhost_iov_iter *dst_it)
+{
+       uint32_t vec_idx = 0;
+       uint32_t mbuf_offset, mbuf_avail;
+       uint32_t buf_offset, buf_avail;
+       uint64_t buf_addr, buf_iova, buf_len;
+       uint32_t cpy_len, cpy_threshold;
+       uint64_t hdr_addr;
+       struct rte_mbuf *hdr_mbuf;
+       struct batch_copy_elem *batch_copy = vq->batch_copy_elems;
+       struct virtio_net_hdr_mrg_rxbuf tmp_hdr, *hdr = NULL;
+       int error = 0;
+       uint64_t mapped_len;
+
+       uint32_t tlen = 0;
+       int tvec_idx = 0;
+       void *hpa;
+
+       if (unlikely(m == NULL)) {
+               error = -1;
+               goto out;
+       }
+
+       cpy_threshold = vq->async_threshold;
+
+       buf_addr = buf_vec[vec_idx].buf_addr;
+       buf_iova = buf_vec[vec_idx].buf_iova;
+       buf_len = buf_vec[vec_idx].buf_len;
+
+       if (unlikely(buf_len < dev->vhost_hlen && nr_vec <= 1)) {
+               error = -1;
+               goto out;
+       }
+
+       hdr_mbuf = m;
+       hdr_addr = buf_addr;
+       if (unlikely(buf_len < dev->vhost_hlen))
+               hdr = &tmp_hdr;
+       else
+               hdr = (struct virtio_net_hdr_mrg_rxbuf *)(uintptr_t)hdr_addr;
+
+       VHOST_LOG_DATA(DEBUG, "(%d) RX: num merge buffers %d\n",
+               dev->vid, num_buffers);
+
+       if (unlikely(buf_len < dev->vhost_hlen)) {
+               buf_offset = dev->vhost_hlen - buf_len;
+               vec_idx++;
+               buf_addr = buf_vec[vec_idx].buf_addr;
+               buf_iova = buf_vec[vec_idx].buf_iova;
+               buf_len = buf_vec[vec_idx].buf_len;
+               buf_avail = buf_len - buf_offset;
+       } else {
+               buf_offset = dev->vhost_hlen;
+               buf_avail = buf_len - dev->vhost_hlen;
+       }
+
+       mbuf_avail  = rte_pktmbuf_data_len(m);
+       mbuf_offset = 0;
+
+       while (mbuf_avail != 0 || m->next != NULL) {
+               /* done with current buf, get the next one */
+               if (buf_avail == 0) {
+                       vec_idx++;
+                       if (unlikely(vec_idx >= nr_vec)) {
+                               error = -1;
+                               goto out;
+                       }
+
+                       buf_addr = buf_vec[vec_idx].buf_addr;
+                       buf_iova = buf_vec[vec_idx].buf_iova;
+                       buf_len = buf_vec[vec_idx].buf_len;
+
+                       buf_offset = 0;
+                       buf_avail  = buf_len;
+               }
+
+               /* done with current mbuf, get the next one */
+               if (mbuf_avail == 0) {
+                       m = m->next;
+
+                       mbuf_offset = 0;
+                       mbuf_avail  = rte_pktmbuf_data_len(m);
+               }
+
+               if (hdr_addr) {
+                       virtio_enqueue_offload(hdr_mbuf, &hdr->hdr);
+                       if (rxvq_is_mergeable(dev))
+                               ASSIGN_UNLESS_EQUAL(hdr->num_buffers,
+                                               num_buffers);
+
+                       if (unlikely(hdr == &tmp_hdr)) {
+                               copy_vnet_hdr_to_desc(dev, vq, buf_vec, hdr);
+                       } else {
+                               PRINT_PACKET(dev, (uintptr_t)hdr_addr,
+                                               dev->vhost_hlen, 0);
+                               vhost_log_cache_write_iova(dev, vq,
+                                               buf_vec[0].buf_iova,
+                                               dev->vhost_hlen);
+                       }
+
+                       hdr_addr = 0;
+               }
+
+               cpy_len = RTE_MIN(buf_avail, mbuf_avail);
+
+               while (unlikely(cpy_len && cpy_len >= cpy_threshold)) {
+                       hpa = (void *)(uintptr_t)gpa_to_first_hpa(dev,
+                                       buf_iova + buf_offset,
+                                       cpy_len, &mapped_len);
+
+                       if (unlikely(!hpa || mapped_len < cpy_threshold))
+                               break;
+
+                       async_fill_vec(src_iovec + tvec_idx,
+                               (void *)(uintptr_t)rte_pktmbuf_iova_offset(m,
+                               mbuf_offset), (size_t)mapped_len);
+
+                       async_fill_vec(dst_iovec + tvec_idx,
+                                       hpa, (size_t)mapped_len);
+
+                       tlen += (uint32_t)mapped_len;
+                       cpy_len -= (uint32_t)mapped_len;
+                       mbuf_avail  -= (uint32_t)mapped_len;
+                       mbuf_offset += (uint32_t)mapped_len;
+                       buf_avail  -= (uint32_t)mapped_len;
+                       buf_offset += (uint32_t)mapped_len;
+                       tvec_idx++;
+               }
+
+               if (likely(cpy_len)) {
+                       if (unlikely(vq->batch_copy_nb_elems >= vq->size)) {
+                               rte_memcpy(
+                               (void *)((uintptr_t)(buf_addr + buf_offset)),
+                               rte_pktmbuf_mtod_offset(m, void *, mbuf_offset),
+                               cpy_len);
+
+                               PRINT_PACKET(dev,
+                                       (uintptr_t)(buf_addr + buf_offset),
+                                       cpy_len, 0);
+                       } else {
+                               batch_copy[vq->batch_copy_nb_elems].dst =
+                               (void *)((uintptr_t)(buf_addr + buf_offset));
+                               batch_copy[vq->batch_copy_nb_elems].src =
+                               rte_pktmbuf_mtod_offset(m, void *, mbuf_offset);
+                               batch_copy[vq->batch_copy_nb_elems].log_addr =
+                                       buf_iova + buf_offset;
+                               batch_copy[vq->batch_copy_nb_elems].len =
+                                       cpy_len;
+                               vq->batch_copy_nb_elems++;
+                       }
+
+                       mbuf_avail  -= cpy_len;
+                       mbuf_offset += cpy_len;
+                       buf_avail  -= cpy_len;
+                       buf_offset += cpy_len;
+               }
+
+       }
+
+out:
+       async_fill_iter(src_it, tlen, src_iovec, tvec_idx);
+       async_fill_iter(dst_it, tlen, dst_iovec, tvec_idx);
+
+       return error;
+}
+
 static __rte_always_inline int
 vhost_enqueue_single_packed(struct virtio_net *dev,
                            struct vhost_virtqueue *vq,
@@ -1069,6 +1298,8 @@ virtio_dev_rx_batch_packed(struct virtio_net *dev,
                                                  VHOST_ACCESS_RW);
 
        vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
+               if (unlikely(!desc_addrs[i]))
+                       return -1;
                if (unlikely(lens[i] != descs[avail_idx + i].len))
                        return -1;
        }
@@ -1131,8 +1362,8 @@ virtio_dev_rx_single_packed(struct virtio_net *dev,
 
 static __rte_noinline uint32_t
 virtio_dev_rx_packed(struct virtio_net *dev,
-                    struct vhost_virtqueue *vq,
-                    struct rte_mbuf **pkts,
+                    struct vhost_virtqueue *__rte_restrict vq,
+                    struct rte_mbuf **__rte_restrict pkts,
                     uint32_t count)
 {
        uint32_t pkt_idx = 0;
@@ -1217,7 +1448,7 @@ out_access_unlock:
 
 uint16_t
 rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
-       struct rte_mbuf **pkts, uint16_t count)
+       struct rte_mbuf **__rte_restrict pkts, uint16_t count)
 {
        struct virtio_net *dev = get_device(vid);
 
@@ -1234,6 +1465,345 @@ rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
        return virtio_dev_rx(dev, queue_id, pkts, count);
 }
 
+static __rte_always_inline uint16_t
+virtio_dev_rx_async_get_info_idx(uint16_t pkts_idx,
+       uint16_t vq_size, uint16_t n_inflight)
+{
+       return pkts_idx > n_inflight ? (pkts_idx - n_inflight) :
+               (vq_size - n_inflight + pkts_idx) & (vq_size - 1);
+}
+
+static __rte_always_inline void
+virtio_dev_rx_async_submit_split_err(struct virtio_net *dev,
+       struct vhost_virtqueue *vq, uint16_t queue_id,
+       uint16_t last_idx, uint16_t shadow_idx)
+{
+       uint16_t start_idx, pkts_idx, vq_size;
+       uint64_t *async_pending_info;
+
+       pkts_idx = vq->async_pkts_idx;
+       async_pending_info = vq->async_pending_info;
+       vq_size = vq->size;
+       start_idx = virtio_dev_rx_async_get_info_idx(pkts_idx,
+               vq_size, vq->async_pkts_inflight_n);
+
+       while (likely((start_idx & (vq_size - 1)) != pkts_idx)) {
+               uint64_t n_seg =
+                       async_pending_info[(start_idx) & (vq_size - 1)] >>
+                       ASYNC_PENDING_INFO_N_SFT;
+
+               while (n_seg)
+                       n_seg -= vq->async_ops.check_completed_copies(dev->vid,
+                               queue_id, 0, 1);
+       }
+
+       vq->async_pkts_inflight_n = 0;
+       vq->batch_copy_nb_elems = 0;
+
+       vq->shadow_used_idx = shadow_idx;
+       vq->last_avail_idx = last_idx;
+}
+
+static __rte_noinline uint32_t
+virtio_dev_rx_async_submit_split(struct virtio_net *dev,
+       struct vhost_virtqueue *vq, uint16_t queue_id,
+       struct rte_mbuf **pkts, uint32_t count)
+{
+       uint32_t pkt_idx = 0, pkt_burst_idx = 0;
+       uint16_t num_buffers;
+       struct buf_vector buf_vec[BUF_VECTOR_MAX];
+       uint16_t avail_head, last_idx, shadow_idx;
+
+       struct rte_vhost_iov_iter *it_pool = vq->it_pool;
+       struct iovec *vec_pool = vq->vec_pool;
+       struct rte_vhost_async_desc tdes[MAX_PKT_BURST];
+       struct iovec *src_iovec = vec_pool;
+       struct iovec *dst_iovec = vec_pool + (VHOST_MAX_ASYNC_VEC >> 1);
+       struct rte_vhost_iov_iter *src_it = it_pool;
+       struct rte_vhost_iov_iter *dst_it = it_pool + 1;
+       uint16_t n_free_slot, slot_idx;
+       int n_pkts = 0;
+
+       avail_head = __atomic_load_n(&vq->avail->idx, __ATOMIC_ACQUIRE);
+       last_idx = vq->last_avail_idx;
+       shadow_idx = vq->shadow_used_idx;
+
+       /*
+        * The ordering between avail index and
+        * desc reads needs to be enforced.
+        */
+       rte_smp_rmb();
+
+       rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]);
+
+       for (pkt_idx = 0; pkt_idx < count; pkt_idx++) {
+               uint32_t pkt_len = pkts[pkt_idx]->pkt_len + dev->vhost_hlen;
+               uint16_t nr_vec = 0;
+
+               if (unlikely(reserve_avail_buf_split(dev, vq,
+                                               pkt_len, buf_vec, &num_buffers,
+                                               avail_head, &nr_vec) < 0)) {
+                       VHOST_LOG_DATA(DEBUG,
+                               "(%d) failed to get enough desc from vring\n",
+                               dev->vid);
+                       vq->shadow_used_idx -= num_buffers;
+                       break;
+               }
+
+               VHOST_LOG_DATA(DEBUG, "(%d) current index %d | end index %d\n",
+                       dev->vid, vq->last_avail_idx,
+                       vq->last_avail_idx + num_buffers);
+
+               if (async_mbuf_to_desc(dev, vq, pkts[pkt_idx],
+                               buf_vec, nr_vec, num_buffers,
+                               src_iovec, dst_iovec, src_it, dst_it) < 0) {
+                       vq->shadow_used_idx -= num_buffers;
+                       break;
+               }
+
+               slot_idx = (vq->async_pkts_idx + pkt_idx) & (vq->size - 1);
+               if (src_it->count) {
+                       async_fill_desc(&tdes[pkt_burst_idx], src_it, dst_it);
+                       pkt_burst_idx++;
+                       vq->async_pending_info[slot_idx] =
+                               num_buffers | (src_it->nr_segs << 16);
+                       src_iovec += src_it->nr_segs;
+                       dst_iovec += dst_it->nr_segs;
+                       src_it += 2;
+                       dst_it += 2;
+               } else {
+                       vq->async_pending_info[slot_idx] = num_buffers;
+                       vq->async_pkts_inflight_n++;
+               }
+
+               vq->last_avail_idx += num_buffers;
+
+               if (pkt_burst_idx >= VHOST_ASYNC_BATCH_THRESHOLD ||
+                               (pkt_idx == count - 1 && pkt_burst_idx)) {
+                       n_pkts = vq->async_ops.transfer_data(dev->vid,
+                                       queue_id, tdes, 0, pkt_burst_idx);
+                       src_iovec = vec_pool;
+                       dst_iovec = vec_pool + (VHOST_MAX_ASYNC_VEC >> 1);
+                       src_it = it_pool;
+                       dst_it = it_pool + 1;
+
+                       if (unlikely(n_pkts < (int)pkt_burst_idx)) {
+                               vq->async_pkts_inflight_n +=
+                                       n_pkts > 0 ? n_pkts : 0;
+                               virtio_dev_rx_async_submit_split_err(dev,
+                                       vq, queue_id, last_idx, shadow_idx);
+                               return 0;
+                       }
+
+                       pkt_burst_idx = 0;
+                       vq->async_pkts_inflight_n += n_pkts;
+               }
+       }
+
+       if (pkt_burst_idx) {
+               n_pkts = vq->async_ops.transfer_data(dev->vid,
+                               queue_id, tdes, 0, pkt_burst_idx);
+               if (unlikely(n_pkts < (int)pkt_burst_idx)) {
+                       vq->async_pkts_inflight_n += n_pkts > 0 ? n_pkts : 0;
+                       virtio_dev_rx_async_submit_split_err(dev, vq, queue_id,
+                               last_idx, shadow_idx);
+                       return 0;
+               }
+
+               vq->async_pkts_inflight_n += n_pkts;
+       }
+
+       do_data_copy_enqueue(dev, vq);
+
+       n_free_slot = vq->size - vq->async_pkts_idx;
+       if (n_free_slot > pkt_idx) {
+               rte_memcpy(&vq->async_pkts_pending[vq->async_pkts_idx],
+                       pkts, pkt_idx * sizeof(uintptr_t));
+               vq->async_pkts_idx += pkt_idx;
+       } else {
+               rte_memcpy(&vq->async_pkts_pending[vq->async_pkts_idx],
+                       pkts, n_free_slot * sizeof(uintptr_t));
+               rte_memcpy(&vq->async_pkts_pending[0],
+                       &pkts[n_free_slot],
+                       (pkt_idx - n_free_slot) * sizeof(uintptr_t));
+               vq->async_pkts_idx = pkt_idx - n_free_slot;
+       }
+
+       if (likely(vq->shadow_used_idx))
+               async_flush_shadow_used_ring_split(dev, vq);
+
+       return pkt_idx;
+}
+
+uint16_t rte_vhost_poll_enqueue_completed(int vid, uint16_t queue_id,
+               struct rte_mbuf **pkts, uint16_t count)
+{
+       struct virtio_net *dev = get_device(vid);
+       struct vhost_virtqueue *vq;
+       uint16_t n_segs_cpl, n_pkts_put = 0, n_descs = 0;
+       uint16_t start_idx, pkts_idx, vq_size;
+       uint16_t n_inflight;
+       uint64_t *async_pending_info;
+
+       if (!dev)
+               return 0;
+
+       VHOST_LOG_DATA(DEBUG, "(%d) %s\n", dev->vid, __func__);
+       if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) {
+               VHOST_LOG_DATA(ERR, "(%d) %s: invalid virtqueue idx %d.\n",
+                       dev->vid, __func__, queue_id);
+               return 0;
+       }
+
+       vq = dev->virtqueue[queue_id];
+
+       rte_spinlock_lock(&vq->access_lock);
+
+       n_inflight = vq->async_pkts_inflight_n;
+       pkts_idx = vq->async_pkts_idx;
+       async_pending_info = vq->async_pending_info;
+       vq_size = vq->size;
+       start_idx = virtio_dev_rx_async_get_info_idx(pkts_idx,
+               vq_size, vq->async_pkts_inflight_n);
+
+       n_segs_cpl = vq->async_ops.check_completed_copies(vid, queue_id,
+               0, ASYNC_MAX_POLL_SEG - vq->async_last_seg_n) +
+               vq->async_last_seg_n;
+
+       rte_smp_wmb();
+
+       while (likely((n_pkts_put < count) && n_inflight)) {
+               uint64_t info = async_pending_info[
+                       (start_idx + n_pkts_put) & (vq_size - 1)];
+               uint64_t n_segs;
+               n_pkts_put++;
+               n_inflight--;
+               n_descs += info & ASYNC_PENDING_INFO_N_MSK;
+               n_segs = info >> ASYNC_PENDING_INFO_N_SFT;
+
+               if (n_segs) {
+                       if (unlikely(n_segs_cpl < n_segs)) {
+                               n_pkts_put--;
+                               n_inflight++;
+                               n_descs -= info & ASYNC_PENDING_INFO_N_MSK;
+                               if (n_segs_cpl) {
+                                       async_pending_info[
+                                               (start_idx + n_pkts_put) &
+                                               (vq_size - 1)] =
+                                       ((n_segs - n_segs_cpl) <<
+                                        ASYNC_PENDING_INFO_N_SFT) |
+                                       (info & ASYNC_PENDING_INFO_N_MSK);
+                                       n_segs_cpl = 0;
+                               }
+                               break;
+                       }
+                       n_segs_cpl -= n_segs;
+               }
+       }
+
+       vq->async_last_seg_n = n_segs_cpl;
+
+       if (n_pkts_put) {
+               vq->async_pkts_inflight_n = n_inflight;
+               if (likely(vq->enabled && vq->access_ok)) {
+                       __atomic_add_fetch(&vq->used->idx,
+                                       n_descs, __ATOMIC_RELEASE);
+                       vhost_vring_call_split(dev, vq);
+               }
+       }
+
+       if (start_idx + n_pkts_put <= vq_size) {
+               rte_memcpy(pkts, &vq->async_pkts_pending[start_idx],
+                       n_pkts_put * sizeof(uintptr_t));
+       } else {
+               rte_memcpy(pkts, &vq->async_pkts_pending[start_idx],
+                       (vq_size - start_idx) * sizeof(uintptr_t));
+               rte_memcpy(&pkts[vq_size - start_idx], vq->async_pkts_pending,
+                       (n_pkts_put - vq_size + start_idx) * sizeof(uintptr_t));
+       }
+
+       rte_spinlock_unlock(&vq->access_lock);
+
+       return n_pkts_put;
+}
+
+static __rte_always_inline uint32_t
+virtio_dev_rx_async_submit(struct virtio_net *dev, uint16_t queue_id,
+       struct rte_mbuf **pkts, uint32_t count)
+{
+       struct vhost_virtqueue *vq;
+       uint32_t nb_tx = 0;
+       bool drawback = false;
+
+       VHOST_LOG_DATA(DEBUG, "(%d) %s\n", dev->vid, __func__);
+       if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) {
+               VHOST_LOG_DATA(ERR, "(%d) %s: invalid virtqueue idx %d.\n",
+                       dev->vid, __func__, queue_id);
+               return 0;
+       }
+
+       vq = dev->virtqueue[queue_id];
+
+       rte_spinlock_lock(&vq->access_lock);
+
+       if (unlikely(vq->enabled == 0))
+               goto out_access_unlock;
+
+       if (unlikely(!vq->async_registered)) {
+               drawback = true;
+               goto out_access_unlock;
+       }
+
+       if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
+               vhost_user_iotlb_rd_lock(vq);
+
+       if (unlikely(vq->access_ok == 0))
+               if (unlikely(vring_translate(dev, vq) < 0))
+                       goto out;
+
+       count = RTE_MIN((uint32_t)MAX_PKT_BURST, count);
+       if (count == 0)
+               goto out;
+
+       /* TODO: packed queue not implemented */
+       if (vq_is_packed(dev))
+               nb_tx = 0;
+       else
+               nb_tx = virtio_dev_rx_async_submit_split(dev,
+                               vq, queue_id, pkts, count);
+
+out:
+       if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
+               vhost_user_iotlb_rd_unlock(vq);
+
+out_access_unlock:
+       rte_spinlock_unlock(&vq->access_lock);
+
+       if (drawback)
+               return rte_vhost_enqueue_burst(dev->vid, queue_id, pkts, count);
+
+       return nb_tx;
+}
+
+uint16_t
+rte_vhost_submit_enqueue_burst(int vid, uint16_t queue_id,
+               struct rte_mbuf **pkts, uint16_t count)
+{
+       struct virtio_net *dev = get_device(vid);
+
+       if (!dev)
+               return 0;
+
+       if (unlikely(!(dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET))) {
+               VHOST_LOG_DATA(ERR,
+                       "(%d) %s: built-in vhost net backend is disabled.\n",
+                       dev->vid, __func__);
+               return 0;
+       }
+
+       return virtio_dev_rx_async_submit(dev, queue_id, pkts, count);
+}
+
 static inline bool
 virtio_net_with_host_offload(struct virtio_net *dev)
 {
@@ -1671,6 +2241,8 @@ virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
 {
        uint16_t i;
        uint16_t free_entries;
+       uint16_t dropped = 0;
+       static bool allocerr_warned;
 
        if (unlikely(dev->dequeue_zero_copy)) {
                struct zcopy_mbuf *zmbuf, *next;
@@ -1732,13 +2304,35 @@ virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
                        update_shadow_used_ring_split(vq, head_idx, 0);
 
                pkts[i] = virtio_dev_pktmbuf_alloc(dev, mbuf_pool, buf_len);
-               if (unlikely(pkts[i] == NULL))
+               if (unlikely(pkts[i] == NULL)) {
+                       /*
+                        * mbuf allocation fails for jumbo packets when external
+                        * buffer allocation is not allowed and linear buffer
+                        * is required. Drop this packet.
+                        */
+                       if (!allocerr_warned) {
+                               VHOST_LOG_DATA(ERR,
+                                       "Failed mbuf alloc of size %d from %s on %s.\n",
+                                       buf_len, mbuf_pool->name, dev->ifname);
+                               allocerr_warned = true;
+                       }
+                       dropped += 1;
+                       i++;
                        break;
+               }
 
                err = copy_desc_to_mbuf(dev, vq, buf_vec, nr_vec, pkts[i],
                                mbuf_pool);
                if (unlikely(err)) {
                        rte_pktmbuf_free(pkts[i]);
+                       if (!allocerr_warned) {
+                               VHOST_LOG_DATA(ERR,
+                                       "Failed to copy desc to mbuf on %s.\n",
+                                       dev->ifname);
+                               allocerr_warned = true;
+                       }
+                       dropped += 1;
+                       i++;
                        break;
                }
 
@@ -1748,6 +2342,8 @@ virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
                        zmbuf = get_zmbuf(vq);
                        if (!zmbuf) {
                                rte_pktmbuf_free(pkts[i]);
+                               dropped += 1;
+                               i++;
                                break;
                        }
                        zmbuf->mbuf = pkts[i];
@@ -1777,7 +2373,7 @@ virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
                }
        }
 
-       return i;
+       return (i - dropped);
 }
 
 static __rte_always_inline int
@@ -1822,6 +2418,8 @@ vhost_reserve_avail_batch_packed(struct virtio_net *dev,
        }
 
        vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
+               if (unlikely(!desc_addrs[i]))
+                       return -1;
                if (unlikely((lens[i] != descs[avail_idx + i].len)))
                        return -1;
        }
@@ -1909,6 +2507,7 @@ vhost_dequeue_single_packed(struct virtio_net *dev,
        uint32_t buf_len;
        uint16_t nr_vec = 0;
        int err;
+       static bool allocerr_warned;
 
        if (unlikely(fill_vec_buf_packed(dev, vq,
                                         vq->last_avail_idx, desc_count,
@@ -1919,14 +2518,24 @@ vhost_dequeue_single_packed(struct virtio_net *dev,
 
        *pkts = virtio_dev_pktmbuf_alloc(dev, mbuf_pool, buf_len);
        if (unlikely(*pkts == NULL)) {
-               VHOST_LOG_DATA(ERR,
-                       "Failed to allocate memory for mbuf.\n");
+               if (!allocerr_warned) {
+                       VHOST_LOG_DATA(ERR,
+                               "Failed mbuf alloc of size %d from %s on %s.\n",
+                               buf_len, mbuf_pool->name, dev->ifname);
+                       allocerr_warned = true;
+               }
                return -1;
        }
 
        err = copy_desc_to_mbuf(dev, vq, buf_vec, nr_vec, *pkts,
                                mbuf_pool);
        if (unlikely(err)) {
+               if (!allocerr_warned) {
+                       VHOST_LOG_DATA(ERR,
+                               "Failed to copy desc to mbuf on %s.\n",
+                               dev->ifname);
+                       allocerr_warned = true;
+               }
                rte_pktmbuf_free(*pkts);
                return -1;
        }
@@ -1941,21 +2550,24 @@ virtio_dev_tx_single_packed(struct virtio_net *dev,
                            struct rte_mbuf **pkts)
 {
 
-       uint16_t buf_id, desc_count;
+       uint16_t buf_id, desc_count = 0;
+       int ret;
 
-       if (vhost_dequeue_single_packed(dev, vq, mbuf_pool, pkts, &buf_id,
-                                       &desc_count))
-               return -1;
+       ret = vhost_dequeue_single_packed(dev, vq, mbuf_pool, pkts, &buf_id,
+                                       &desc_count);
 
-       if (virtio_net_is_inorder(dev))
-               vhost_shadow_dequeue_single_packed_inorder(vq, buf_id,
-                                                          desc_count);
-       else
-               vhost_shadow_dequeue_single_packed(vq, buf_id, desc_count);
+       if (likely(desc_count > 0)) {
+               if (virtio_net_is_inorder(dev))
+                       vhost_shadow_dequeue_single_packed_inorder(vq, buf_id,
+                                                                  desc_count);
+               else
+                       vhost_shadow_dequeue_single_packed(vq, buf_id,
+                                       desc_count);
 
-       vq_inc_last_avail_packed(vq, desc_count);
+               vq_inc_last_avail_packed(vq, desc_count);
+       }
 
-       return 0;
+       return ret;
 }
 
 static __rte_always_inline int
@@ -2080,9 +2692,9 @@ free_zmbuf(struct vhost_virtqueue *vq)
 
 static __rte_noinline uint16_t
 virtio_dev_tx_packed_zmbuf(struct virtio_net *dev,
-                          struct vhost_virtqueue *vq,
+                          struct vhost_virtqueue *__rte_restrict vq,
                           struct rte_mempool *mbuf_pool,
-                          struct rte_mbuf **pkts,
+                          struct rte_mbuf **__rte_restrict pkts,
                           uint32_t count)
 {
        uint32_t pkt_idx = 0;
@@ -2116,9 +2728,9 @@ virtio_dev_tx_packed_zmbuf(struct virtio_net *dev,
 
 static __rte_noinline uint16_t
 virtio_dev_tx_packed(struct virtio_net *dev,
-                    struct vhost_virtqueue *vq,
+                    struct vhost_virtqueue *__rte_restrict vq,
                     struct rte_mempool *mbuf_pool,
-                    struct rte_mbuf **pkts,
+                    struct rte_mbuf **__rte_restrict pkts,
                     uint32_t count)
 {
        uint32_t pkt_idx = 0;