X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=lib%2Flibrte_vhost%2Fvirtio_net.c;h=bd9303c8a9f336757898f42cd0ca9a9a5269c547;hb=6563cf92380a963f283246e1ca9292fdca66cc57;hp=5fdd4172b69ba85d45f6b20a454c6db50bed8b92;hpb=fb3815cc614dc754a96e5933165a2ded66b87164;p=dpdk.git diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c index 5fdd4172b6..bd9303c8a9 100644 --- a/lib/librte_vhost/virtio_net.c +++ b/lib/librte_vhost/virtio_net.c @@ -17,133 +17,399 @@ #include #include #include +#include #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) +{ + return dev->features & (1ULL << VIRTIO_NET_F_MRG_RXBUF); +} + +static __rte_always_inline bool +virtio_net_is_inorder(struct virtio_net *dev) +{ + return dev->features & (1ULL << VIRTIO_F_IN_ORDER); +} + static bool is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring) { return (is_tx ^ (idx & 1)) == 0 && idx < nr_vring; } -static __rte_always_inline struct vring_desc * -alloc_copy_ind_table(struct virtio_net *dev, struct vhost_virtqueue *vq, - struct vring_desc *desc) +static inline void +do_data_copy_enqueue(struct virtio_net *dev, struct vhost_virtqueue *vq) { - struct vring_desc *idesc; - uint64_t src, dst; - uint64_t len, remain = desc->len; - uint64_t desc_addr = desc->addr; - - idesc = rte_malloc(__func__, desc->len, 0); - if (unlikely(!idesc)) - return 0; - - dst = (uint64_t)(uintptr_t)idesc; - - while (remain) { - len = remain; - src = vhost_iova_to_vva(dev, vq, desc_addr, &len, - VHOST_ACCESS_RO); - if (unlikely(!src || !len)) { - rte_free(idesc); - return 0; - } - - rte_memcpy((void *)(uintptr_t)dst, (void *)(uintptr_t)src, len); + struct batch_copy_elem *elem = vq->batch_copy_elems; + uint16_t count = vq->batch_copy_nb_elems; + int i; - remain -= len; - dst += len; - desc_addr += len; + for (i = 0; i < count; i++) { + rte_memcpy(elem[i].dst, elem[i].src, elem[i].len); + vhost_log_cache_write_iova(dev, vq, elem[i].log_addr, + elem[i].len); + PRINT_PACKET(dev, (uintptr_t)elem[i].dst, elem[i].len, 0); } - return idesc; + vq->batch_copy_nb_elems = 0; } -static __rte_always_inline void -free_ind_table(struct vring_desc *idesc) +static inline void +do_data_copy_dequeue(struct vhost_virtqueue *vq) { - rte_free(idesc); + struct batch_copy_elem *elem = vq->batch_copy_elems; + uint16_t count = vq->batch_copy_nb_elems; + int i; + + for (i = 0; i < count; i++) + rte_memcpy(elem[i].dst, elem[i].src, elem[i].len); + + vq->batch_copy_nb_elems = 0; } static __rte_always_inline void -do_flush_shadow_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq, - uint16_t to, uint16_t from, uint16_t size) +do_flush_shadow_used_ring_split(struct virtio_net *dev, + struct vhost_virtqueue *vq, + uint16_t to, uint16_t from, uint16_t size) { rte_memcpy(&vq->used->ring[to], - &vq->shadow_used_ring[from], + &vq->shadow_used_split[from], size * sizeof(struct vring_used_elem)); - vhost_log_used_vring(dev, vq, + vhost_log_cache_used_vring(dev, vq, offsetof(struct vring_used, ring[to]), size * sizeof(struct vring_used_elem)); } static __rte_always_inline void -flush_shadow_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq) +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(dev, vq, used_idx, 0, + 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(dev, vq, used_idx, 0, size); + 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(dev, vq, 0, size, + do_flush_shadow_used_ring_split(dev, vq, 0, size, vq->shadow_used_idx - size); } vq->last_used_idx += vq->shadow_used_idx; - rte_smp_wmb(); + vhost_log_cache_sync(dev, vq); - *(volatile uint16_t *)&vq->used->idx += vq->shadow_used_idx; + __atomic_add_fetch(&vq->used->idx, vq->shadow_used_idx, + __ATOMIC_RELEASE); + vq->shadow_used_idx = 0; vhost_log_used_vring(dev, vq, offsetof(struct vring_used, idx), sizeof(vq->used->idx)); } static __rte_always_inline void -update_shadow_used_ring(struct vhost_virtqueue *vq, - uint16_t desc_idx, uint16_t len) +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) { uint16_t i = vq->shadow_used_idx++; - vq->shadow_used_ring[i].id = desc_idx; - vq->shadow_used_ring[i].len = len; + vq->shadow_used_split[i].id = desc_idx; + vq->shadow_used_split[i].len = len; } -static inline void -do_data_copy_enqueue(struct virtio_net *dev, struct vhost_virtqueue *vq) +static __rte_always_inline void +vhost_flush_enqueue_shadow_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq) { - struct batch_copy_elem *elem = vq->batch_copy_elems; - uint16_t count = vq->batch_copy_nb_elems; int i; + uint16_t used_idx = vq->last_used_idx; + uint16_t head_idx = vq->last_used_idx; + uint16_t head_flags = 0; + + /* Split loop in two to save memory barriers */ + for (i = 0; i < vq->shadow_used_idx; i++) { + vq->desc_packed[used_idx].id = vq->shadow_used_packed[i].id; + vq->desc_packed[used_idx].len = vq->shadow_used_packed[i].len; + + used_idx += vq->shadow_used_packed[i].count; + if (used_idx >= vq->size) + used_idx -= vq->size; + } - for (i = 0; i < count; i++) { - rte_memcpy(elem[i].dst, elem[i].src, elem[i].len); - vhost_log_write(dev, elem[i].log_addr, elem[i].len); - PRINT_PACKET(dev, (uintptr_t)elem[i].dst, elem[i].len, 0); + rte_smp_wmb(); + + for (i = 0; i < vq->shadow_used_idx; i++) { + uint16_t flags; + + if (vq->shadow_used_packed[i].len) + flags = VRING_DESC_F_WRITE; + else + flags = 0; + + if (vq->used_wrap_counter) { + flags |= VRING_DESC_F_USED; + flags |= VRING_DESC_F_AVAIL; + } else { + flags &= ~VRING_DESC_F_USED; + flags &= ~VRING_DESC_F_AVAIL; + } + + if (i > 0) { + vq->desc_packed[vq->last_used_idx].flags = flags; + + vhost_log_cache_used_vring(dev, vq, + vq->last_used_idx * + sizeof(struct vring_packed_desc), + sizeof(struct vring_packed_desc)); + } else { + head_idx = vq->last_used_idx; + head_flags = flags; + } + + vq_inc_last_used_packed(vq, vq->shadow_used_packed[i].count); } + + vq->desc_packed[head_idx].flags = head_flags; + + vhost_log_cache_used_vring(dev, vq, + head_idx * + sizeof(struct vring_packed_desc), + sizeof(struct vring_packed_desc)); + + vq->shadow_used_idx = 0; + vhost_log_cache_sync(dev, vq); } -static inline void -do_data_copy_dequeue(struct vhost_virtqueue *vq) +static __rte_always_inline void +vhost_flush_dequeue_shadow_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq) { - struct batch_copy_elem *elem = vq->batch_copy_elems; - uint16_t count = vq->batch_copy_nb_elems; - int i; + struct vring_used_elem_packed *used_elem = &vq->shadow_used_packed[0]; - for (i = 0; i < count; i++) - rte_memcpy(elem[i].dst, elem[i].src, elem[i].len); + vq->desc_packed[vq->shadow_last_used_idx].id = used_elem->id; + rte_smp_wmb(); + vq->desc_packed[vq->shadow_last_used_idx].flags = used_elem->flags; + + vhost_log_cache_used_vring(dev, vq, vq->shadow_last_used_idx * + sizeof(struct vring_packed_desc), + sizeof(struct vring_packed_desc)); + vq->shadow_used_idx = 0; + vhost_log_cache_sync(dev, vq); +} + +static __rte_always_inline void +vhost_flush_enqueue_batch_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + uint64_t *lens, + uint16_t *ids) +{ + uint16_t i; + uint16_t flags; + + if (vq->shadow_used_idx) { + do_data_copy_enqueue(dev, vq); + vhost_flush_enqueue_shadow_packed(dev, vq); + } + + flags = PACKED_DESC_ENQUEUE_USED_FLAG(vq->used_wrap_counter); + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + vq->desc_packed[vq->last_used_idx + i].id = ids[i]; + vq->desc_packed[vq->last_used_idx + i].len = lens[i]; + } + + rte_smp_wmb(); + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + vq->desc_packed[vq->last_used_idx + i].flags = flags; + + vhost_log_cache_used_vring(dev, vq, vq->last_used_idx * + sizeof(struct vring_packed_desc), + sizeof(struct vring_packed_desc) * + PACKED_BATCH_SIZE); + vhost_log_cache_sync(dev, vq); + + vq_inc_last_used_packed(vq, PACKED_BATCH_SIZE); +} + +static __rte_always_inline void +vhost_shadow_dequeue_batch_packed_inorder(struct vhost_virtqueue *vq, + uint16_t id) +{ + vq->shadow_used_packed[0].id = id; + + if (!vq->shadow_used_idx) { + vq->shadow_last_used_idx = vq->last_used_idx; + vq->shadow_used_packed[0].flags = + PACKED_DESC_DEQUEUE_USED_FLAG(vq->used_wrap_counter); + vq->shadow_used_packed[0].len = 0; + vq->shadow_used_packed[0].count = 1; + vq->shadow_used_idx++; + } + + vq_inc_last_used_packed(vq, PACKED_BATCH_SIZE); +} + +static __rte_always_inline void +vhost_shadow_dequeue_batch_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + uint16_t *ids) +{ + uint16_t flags; + uint16_t i; + uint16_t begin; + + flags = PACKED_DESC_DEQUEUE_USED_FLAG(vq->used_wrap_counter); + + if (!vq->shadow_used_idx) { + vq->shadow_last_used_idx = vq->last_used_idx; + vq->shadow_used_packed[0].id = ids[0]; + vq->shadow_used_packed[0].len = 0; + vq->shadow_used_packed[0].count = 1; + vq->shadow_used_packed[0].flags = flags; + vq->shadow_used_idx++; + begin = 1; + } else + begin = 0; + + vhost_for_each_try_unroll(i, begin, PACKED_BATCH_SIZE) { + vq->desc_packed[vq->last_used_idx + i].id = ids[i]; + vq->desc_packed[vq->last_used_idx + i].len = 0; + } + + rte_smp_wmb(); + vhost_for_each_try_unroll(i, begin, PACKED_BATCH_SIZE) + vq->desc_packed[vq->last_used_idx + i].flags = flags; + + vhost_log_cache_used_vring(dev, vq, vq->last_used_idx * + sizeof(struct vring_packed_desc), + sizeof(struct vring_packed_desc) * + PACKED_BATCH_SIZE); + vhost_log_cache_sync(dev, vq); + + vq_inc_last_used_packed(vq, PACKED_BATCH_SIZE); +} + +static __rte_always_inline void +vhost_shadow_dequeue_single_packed(struct vhost_virtqueue *vq, + uint16_t buf_id, + uint16_t count) +{ + uint16_t flags; + + flags = vq->desc_packed[vq->last_used_idx].flags; + if (vq->used_wrap_counter) { + flags |= VRING_DESC_F_USED; + flags |= VRING_DESC_F_AVAIL; + } else { + flags &= ~VRING_DESC_F_USED; + flags &= ~VRING_DESC_F_AVAIL; + } + + if (!vq->shadow_used_idx) { + vq->shadow_last_used_idx = vq->last_used_idx; + + vq->shadow_used_packed[0].id = buf_id; + vq->shadow_used_packed[0].len = 0; + vq->shadow_used_packed[0].flags = flags; + vq->shadow_used_idx++; + } else { + vq->desc_packed[vq->last_used_idx].id = buf_id; + vq->desc_packed[vq->last_used_idx].len = 0; + vq->desc_packed[vq->last_used_idx].flags = flags; + } + + vq_inc_last_used_packed(vq, count); +} + +static __rte_always_inline void +vhost_shadow_dequeue_single_packed_inorder(struct vhost_virtqueue *vq, + uint16_t buf_id, + uint16_t count) +{ + uint16_t flags; + + vq->shadow_used_packed[0].id = buf_id; + + flags = vq->desc_packed[vq->last_used_idx].flags; + if (vq->used_wrap_counter) { + flags |= VRING_DESC_F_USED; + flags |= VRING_DESC_F_AVAIL; + } else { + flags &= ~VRING_DESC_F_USED; + flags &= ~VRING_DESC_F_AVAIL; + } + + if (!vq->shadow_used_idx) { + vq->shadow_last_used_idx = vq->last_used_idx; + vq->shadow_used_packed[0].len = 0; + vq->shadow_used_packed[0].flags = flags; + vq->shadow_used_idx++; + } + + vq_inc_last_used_packed(vq, count); +} + +static __rte_always_inline void +vhost_shadow_enqueue_single_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + uint32_t len[], + uint16_t id[], + uint16_t count[], + uint16_t num_buffers) +{ + uint16_t i; + for (i = 0; i < num_buffers; i++) { + /* enqueue shadow flush action aligned with batch num */ + if (!vq->shadow_used_idx) + vq->shadow_aligned_idx = vq->last_used_idx & + PACKED_BATCH_MASK; + vq->shadow_used_packed[vq->shadow_used_idx].id = id[i]; + vq->shadow_used_packed[vq->shadow_used_idx].len = len[i]; + vq->shadow_used_packed[vq->shadow_used_idx].count = count[i]; + vq->shadow_aligned_idx += count[i]; + vq->shadow_used_idx++; + } + + if (vq->shadow_aligned_idx >= PACKED_BATCH_SIZE) { + do_data_copy_enqueue(dev, vq); + vhost_flush_enqueue_shadow_packed(dev, vq); + } } /* avoid write operation when necessary, to lessen cache issues */ @@ -152,7 +418,7 @@ do_data_copy_dequeue(struct vhost_virtqueue *vq) (var) = (val); \ } while (0) -static void +static __rte_always_inline void virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr) { uint64_t csum_l4 = m_buf->ol_flags & PKT_TX_L4_MASK; @@ -166,15 +432,15 @@ virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr) switch (csum_l4) { case PKT_TX_TCP_CKSUM: - net_hdr->csum_offset = (offsetof(struct tcp_hdr, + net_hdr->csum_offset = (offsetof(struct rte_tcp_hdr, cksum)); break; case PKT_TX_UDP_CKSUM: - net_hdr->csum_offset = (offsetof(struct udp_hdr, + net_hdr->csum_offset = (offsetof(struct rte_udp_hdr, dgram_cksum)); break; case PKT_TX_SCTP_CKSUM: - net_hdr->csum_offset = (offsetof(struct sctp_hdr, + net_hdr->csum_offset = (offsetof(struct rte_sctp_hdr, cksum)); break; } @@ -186,10 +452,11 @@ virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr) /* IP cksum verification cannot be bypassed, then calculate here */ if (m_buf->ol_flags & PKT_TX_IP_CKSUM) { - struct ipv4_hdr *ipv4_hdr; + struct rte_ipv4_hdr *ipv4_hdr; - ipv4_hdr = rte_pktmbuf_mtod_offset(m_buf, struct ipv4_hdr *, + ipv4_hdr = rte_pktmbuf_mtod_offset(m_buf, struct rte_ipv4_hdr *, m_buf->l2_len); + ipv4_hdr->hdr_checksum = 0; ipv4_hdr->hdr_checksum = rte_ipv4_cksum(ipv4_hdr); } @@ -214,440 +481,338 @@ virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr) } static __rte_always_inline int -copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq, - struct vring_desc *descs, struct rte_mbuf *m, - uint16_t desc_idx, uint32_t size) +map_one_desc(struct virtio_net *dev, struct vhost_virtqueue *vq, + struct buf_vector *buf_vec, uint16_t *vec_idx, + uint64_t desc_iova, uint64_t desc_len, uint8_t perm) { - uint32_t desc_avail, desc_offset; - uint32_t mbuf_avail, mbuf_offset; - uint32_t cpy_len; - uint64_t desc_chunck_len; - struct vring_desc *desc; - uint64_t desc_addr, desc_gaddr; - /* A counter to avoid desc dead loop chain */ - uint16_t nr_desc = 1; - struct batch_copy_elem *batch_copy = vq->batch_copy_elems; - uint16_t copy_nb = vq->batch_copy_nb_elems; - int error = 0; + uint16_t vec_id = *vec_idx; - desc = &descs[desc_idx]; - desc_chunck_len = desc->len; - desc_gaddr = desc->addr; - desc_addr = vhost_iova_to_vva(dev, vq, desc_gaddr, - &desc_chunck_len, VHOST_ACCESS_RW); - /* - * Checking of 'desc_addr' placed outside of 'unlikely' macro to avoid - * performance issue with some versions of gcc (4.8.4 and 5.3.0) which - * otherwise stores offset on the stack instead of in a register. - */ - if (unlikely(desc->len < dev->vhost_hlen) || !desc_addr) { - error = -1; - goto out; - } + while (desc_len) { + uint64_t desc_addr; + uint64_t desc_chunck_len = desc_len; - rte_prefetch0((void *)(uintptr_t)desc_addr); + if (unlikely(vec_id >= BUF_VECTOR_MAX)) + return -1; - if (likely(desc_chunck_len >= dev->vhost_hlen)) { - virtio_enqueue_offload(m, - (struct virtio_net_hdr *)(uintptr_t)desc_addr); - PRINT_PACKET(dev, (uintptr_t)desc_addr, dev->vhost_hlen, 0); - vhost_log_write(dev, desc_gaddr, dev->vhost_hlen); - } else { - struct virtio_net_hdr vnet_hdr; - uint64_t remain = dev->vhost_hlen; - uint64_t len; - uint64_t src = (uint64_t)(uintptr_t)&vnet_hdr, dst; - uint64_t guest_addr = desc_gaddr; - - virtio_enqueue_offload(m, &vnet_hdr); - - while (remain) { - len = remain; - dst = vhost_iova_to_vva(dev, vq, guest_addr, - &len, VHOST_ACCESS_RW); - if (unlikely(!dst || !len)) { - error = -1; - goto out; - } + desc_addr = vhost_iova_to_vva(dev, vq, + desc_iova, + &desc_chunck_len, + perm); + if (unlikely(!desc_addr)) + return -1; - rte_memcpy((void *)(uintptr_t)dst, - (void *)(uintptr_t)src, len); + rte_prefetch0((void *)(uintptr_t)desc_addr); - PRINT_PACKET(dev, (uintptr_t)dst, (uint32_t)len, 0); - vhost_log_write(dev, guest_addr, len); - remain -= len; - guest_addr += len; - dst += len; - } + buf_vec[vec_id].buf_iova = desc_iova; + buf_vec[vec_id].buf_addr = desc_addr; + buf_vec[vec_id].buf_len = desc_chunck_len; + + desc_len -= desc_chunck_len; + desc_iova += desc_chunck_len; + vec_id++; } + *vec_idx = vec_id; - desc_avail = desc->len - dev->vhost_hlen; - if (unlikely(desc_chunck_len < dev->vhost_hlen)) { - desc_chunck_len = desc_avail; - desc_gaddr = desc->addr + dev->vhost_hlen; - desc_addr = vhost_iova_to_vva(dev, - vq, desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RW); - if (unlikely(!desc_addr)) { - error = -1; - goto out; - } + return 0; +} - desc_offset = 0; - } else { - desc_offset = dev->vhost_hlen; - desc_chunck_len -= dev->vhost_hlen; - } +static __rte_always_inline int +fill_vec_buf_split(struct virtio_net *dev, struct vhost_virtqueue *vq, + uint32_t avail_idx, uint16_t *vec_idx, + struct buf_vector *buf_vec, uint16_t *desc_chain_head, + uint32_t *desc_chain_len, uint8_t perm) +{ + uint16_t idx = vq->avail->ring[avail_idx & (vq->size - 1)]; + uint16_t vec_id = *vec_idx; + uint32_t len = 0; + uint64_t dlen; + uint32_t nr_descs = vq->size; + uint32_t cnt = 0; + struct vring_desc *descs = vq->desc; + struct vring_desc *idesc = NULL; - mbuf_avail = rte_pktmbuf_data_len(m); - mbuf_offset = 0; - while (mbuf_avail != 0 || m->next != NULL) { - /* done with current mbuf, fetch next */ - if (mbuf_avail == 0) { - m = m->next; + if (unlikely(idx >= vq->size)) + return -1; - mbuf_offset = 0; - mbuf_avail = rte_pktmbuf_data_len(m); - } + *desc_chain_head = idx; - /* done with current desc buf, fetch next */ - if (desc_avail == 0) { - if ((desc->flags & VRING_DESC_F_NEXT) == 0) { - /* Room in vring buffer is not enough */ - error = -1; - goto out; - } - if (unlikely(desc->next >= size || ++nr_desc > size)) { - error = -1; - goto out; - } + if (vq->desc[idx].flags & VRING_DESC_F_INDIRECT) { + dlen = vq->desc[idx].len; + nr_descs = dlen / sizeof(struct vring_desc); + if (unlikely(nr_descs > vq->size)) + return -1; - desc = &descs[desc->next]; - desc_chunck_len = desc->len; - desc_gaddr = desc->addr; - desc_addr = vhost_iova_to_vva(dev, vq, desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RW); - if (unlikely(!desc_addr)) { - error = -1; - goto out; - } + descs = (struct vring_desc *)(uintptr_t) + vhost_iova_to_vva(dev, vq, vq->desc[idx].addr, + &dlen, + VHOST_ACCESS_RO); + if (unlikely(!descs)) + return -1; - desc_offset = 0; - desc_avail = desc->len; - } else if (unlikely(desc_chunck_len == 0)) { - desc_chunck_len = desc_avail; - desc_gaddr += desc_offset; - desc_addr = vhost_iova_to_vva(dev, - vq, desc_gaddr, - &desc_chunck_len, VHOST_ACCESS_RW); - if (unlikely(!desc_addr)) { - error = -1; - goto out; - } - desc_offset = 0; - } + if (unlikely(dlen < vq->desc[idx].len)) { + /* + * The indirect desc table is not contiguous + * in process VA space, we have to copy it. + */ + idesc = vhost_alloc_copy_ind_table(dev, vq, + vq->desc[idx].addr, vq->desc[idx].len); + if (unlikely(!idesc)) + return -1; - cpy_len = RTE_MIN(desc_chunck_len, mbuf_avail); - if (likely(cpy_len > MAX_BATCH_LEN || copy_nb >= vq->size)) { - rte_memcpy((void *)((uintptr_t)(desc_addr + - desc_offset)), - rte_pktmbuf_mtod_offset(m, void *, mbuf_offset), - cpy_len); - vhost_log_write(dev, desc_gaddr + desc_offset, cpy_len); - PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset), - cpy_len, 0); - } else { - batch_copy[copy_nb].dst = - (void *)((uintptr_t)(desc_addr + desc_offset)); - batch_copy[copy_nb].src = - rte_pktmbuf_mtod_offset(m, void *, mbuf_offset); - batch_copy[copy_nb].log_addr = desc_gaddr + desc_offset; - batch_copy[copy_nb].len = cpy_len; - copy_nb++; + descs = idesc; } - mbuf_avail -= cpy_len; - mbuf_offset += cpy_len; - desc_avail -= cpy_len; - desc_offset += cpy_len; - desc_chunck_len -= cpy_len; + idx = 0; } -out: - vq->batch_copy_nb_elems = copy_nb; + while (1) { + if (unlikely(idx >= nr_descs || cnt++ >= nr_descs)) { + free_ind_table(idesc); + return -1; + } - return error; -} + len += descs[idx].len; -/** - * This function adds buffers to the virtio devices RX virtqueue. Buffers can - * be received from the physical port or from another virtio device. A packet - * count is returned to indicate the number of packets that are successfully - * added to the RX queue. This function works when the mbuf is scattered, but - * it doesn't support the mergeable feature. - */ -static __rte_always_inline uint32_t -virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id, - struct rte_mbuf **pkts, uint32_t count) -{ - struct vhost_virtqueue *vq; - uint16_t avail_idx, free_entries, start_idx; - uint16_t desc_indexes[MAX_PKT_BURST]; - struct vring_desc *descs; - uint16_t used_idx; - uint32_t i, sz; - - VHOST_LOG_DEBUG(VHOST_DATA, "(%d) %s\n", dev->vid, __func__); - if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) { - RTE_LOG(ERR, VHOST_DATA, "(%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 (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)) { - count = 0; - goto out; + if (unlikely(map_one_desc(dev, vq, buf_vec, &vec_id, + descs[idx].addr, descs[idx].len, + perm))) { + free_ind_table(idesc); + return -1; } - } - - avail_idx = *((volatile uint16_t *)&vq->avail->idx); - start_idx = vq->last_used_idx; - free_entries = avail_idx - start_idx; - count = RTE_MIN(count, free_entries); - count = RTE_MIN(count, (uint32_t)MAX_PKT_BURST); - if (count == 0) - goto out; - - VHOST_LOG_DEBUG(VHOST_DATA, "(%d) start_idx %d | end_idx %d\n", - dev->vid, start_idx, start_idx + count); - vq->batch_copy_nb_elems = 0; + if ((descs[idx].flags & VRING_DESC_F_NEXT) == 0) + break; - /* Retrieve all of the desc indexes first to avoid caching issues. */ - rte_prefetch0(&vq->avail->ring[start_idx & (vq->size - 1)]); - for (i = 0; i < count; i++) { - used_idx = (start_idx + i) & (vq->size - 1); - desc_indexes[i] = vq->avail->ring[used_idx]; - vq->used->ring[used_idx].id = desc_indexes[i]; - vq->used->ring[used_idx].len = pkts[i]->pkt_len + - dev->vhost_hlen; - vhost_log_used_vring(dev, vq, - offsetof(struct vring_used, ring[used_idx]), - sizeof(vq->used->ring[used_idx])); + idx = descs[idx].next; } - rte_prefetch0(&vq->desc[desc_indexes[0]]); - for (i = 0; i < count; i++) { - struct vring_desc *idesc = NULL; - uint16_t desc_idx = desc_indexes[i]; - int err; - - if (vq->desc[desc_idx].flags & VRING_DESC_F_INDIRECT) { - uint64_t dlen = vq->desc[desc_idx].len; - descs = (struct vring_desc *)(uintptr_t) - vhost_iova_to_vva(dev, - vq, vq->desc[desc_idx].addr, - &dlen, VHOST_ACCESS_RO); - if (unlikely(!descs)) { - count = i; - break; - } + *desc_chain_len = len; + *vec_idx = vec_id; - if (unlikely(dlen < vq->desc[desc_idx].len)) { - /* - * The indirect desc table is not contiguous - * in process VA space, we have to copy it. - */ - idesc = alloc_copy_ind_table(dev, vq, - &vq->desc[desc_idx]); - if (unlikely(!idesc)) - break; - - descs = idesc; - } + if (unlikely(!!idesc)) + free_ind_table(idesc); - desc_idx = 0; - sz = vq->desc[desc_idx].len / sizeof(*descs); - } else { - descs = vq->desc; - sz = vq->size; - } + return 0; +} - err = copy_mbuf_to_desc(dev, vq, descs, pkts[i], desc_idx, sz); - if (unlikely(err)) { - count = i; - free_ind_table(idesc); - break; - } +/* + * Returns -1 on fail, 0 on success + */ +static inline int +reserve_avail_buf_split(struct virtio_net *dev, struct vhost_virtqueue *vq, + uint32_t size, struct buf_vector *buf_vec, + uint16_t *num_buffers, uint16_t avail_head, + uint16_t *nr_vec) +{ + uint16_t cur_idx; + uint16_t vec_idx = 0; + uint16_t max_tries, tries = 0; - if (i + 1 < count) - rte_prefetch0(&vq->desc[desc_indexes[i+1]]); + uint16_t head_idx = 0; + uint32_t len = 0; - if (unlikely(!!idesc)) - free_ind_table(idesc); - } + *num_buffers = 0; + cur_idx = vq->last_avail_idx; - do_data_copy_enqueue(dev, vq); + if (rxvq_is_mergeable(dev)) + max_tries = vq->size - 1; + else + max_tries = 1; - rte_smp_wmb(); + while (size > 0) { + if (unlikely(cur_idx == avail_head)) + return -1; + /* + * if we tried all available ring items, and still + * can't get enough buf, it means something abnormal + * happened. + */ + if (unlikely(++tries > max_tries)) + return -1; - *(volatile uint16_t *)&vq->used->idx += count; - vq->last_used_idx += count; - vhost_log_used_vring(dev, vq, - offsetof(struct vring_used, idx), - sizeof(vq->used->idx)); + if (unlikely(fill_vec_buf_split(dev, vq, cur_idx, + &vec_idx, buf_vec, + &head_idx, &len, + VHOST_ACCESS_RW) < 0)) + return -1; + len = RTE_MIN(len, size); + update_shadow_used_ring_split(vq, head_idx, len); + size -= len; - vhost_vring_call(dev, vq); -out: - if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) - vhost_user_iotlb_rd_unlock(vq); + cur_idx++; + *num_buffers += 1; + } -out_access_unlock: - rte_spinlock_unlock(&vq->access_lock); + *nr_vec = vec_idx; - return count; + return 0; } static __rte_always_inline int -fill_vec_buf(struct virtio_net *dev, struct vhost_virtqueue *vq, - uint32_t avail_idx, uint32_t *vec_idx, - struct buf_vector *buf_vec, uint16_t *desc_chain_head, - uint16_t *desc_chain_len) +fill_vec_buf_packed_indirect(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct vring_packed_desc *desc, uint16_t *vec_idx, + struct buf_vector *buf_vec, uint32_t *len, uint8_t perm) { - uint16_t idx = vq->avail->ring[avail_idx & (vq->size - 1)]; - uint32_t vec_id = *vec_idx; - uint32_t len = 0; + uint16_t i; + uint32_t nr_descs; + uint16_t vec_id = *vec_idx; uint64_t dlen; - struct vring_desc *descs = vq->desc; - struct vring_desc *idesc = NULL; + struct vring_packed_desc *descs, *idescs = NULL; - *desc_chain_head = idx; + dlen = desc->len; + descs = (struct vring_packed_desc *)(uintptr_t) + vhost_iova_to_vva(dev, vq, desc->addr, &dlen, VHOST_ACCESS_RO); + if (unlikely(!descs)) + return -1; - if (vq->desc[idx].flags & VRING_DESC_F_INDIRECT) { - dlen = vq->desc[idx].len; - descs = (struct vring_desc *)(uintptr_t) - vhost_iova_to_vva(dev, vq, vq->desc[idx].addr, - &dlen, - VHOST_ACCESS_RO); - if (unlikely(!descs)) + if (unlikely(dlen < desc->len)) { + /* + * The indirect desc table is not contiguous + * in process VA space, we have to copy it. + */ + idescs = vhost_alloc_copy_ind_table(dev, + vq, desc->addr, desc->len); + if (unlikely(!idescs)) return -1; - if (unlikely(dlen < vq->desc[idx].len)) { - /* - * The indirect desc table is not contiguous - * in process VA space, we have to copy it. - */ - idesc = alloc_copy_ind_table(dev, vq, &vq->desc[idx]); - if (unlikely(!idesc)) - return -1; - - descs = idesc; - } + descs = idescs; + } - idx = 0; + nr_descs = desc->len / sizeof(struct vring_packed_desc); + if (unlikely(nr_descs >= vq->size)) { + free_ind_table(idescs); + return -1; } - while (1) { - if (unlikely(vec_id >= BUF_VECTOR_MAX || idx >= vq->size)) { - free_ind_table(idesc); + for (i = 0; i < nr_descs; i++) { + if (unlikely(vec_id >= BUF_VECTOR_MAX)) { + free_ind_table(idescs); return -1; } - len += descs[idx].len; - buf_vec[vec_id].buf_addr = descs[idx].addr; - buf_vec[vec_id].buf_len = descs[idx].len; - buf_vec[vec_id].desc_idx = idx; - vec_id++; - - if ((descs[idx].flags & VRING_DESC_F_NEXT) == 0) - break; - - idx = descs[idx].next; + *len += descs[i].len; + if (unlikely(map_one_desc(dev, vq, buf_vec, &vec_id, + descs[i].addr, descs[i].len, + perm))) + return -1; } - - *desc_chain_len = len; *vec_idx = vec_id; - if (unlikely(!!idesc)) - free_ind_table(idesc); + if (unlikely(!!idescs)) + free_ind_table(idescs); return 0; } -/* - * Returns -1 on fail, 0 on success - */ -static inline int -reserve_avail_buf_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq, - uint32_t size, struct buf_vector *buf_vec, - uint16_t *num_buffers, uint16_t avail_head) +static __rte_always_inline int +fill_vec_buf_packed(struct virtio_net *dev, struct vhost_virtqueue *vq, + uint16_t avail_idx, uint16_t *desc_count, + struct buf_vector *buf_vec, uint16_t *vec_idx, + uint16_t *buf_id, uint32_t *len, uint8_t perm) { - uint16_t cur_idx; - uint32_t vec_idx = 0; - uint16_t tries = 0; + bool wrap_counter = vq->avail_wrap_counter; + struct vring_packed_desc *descs = vq->desc_packed; + uint16_t vec_id = *vec_idx; - uint16_t head_idx = 0; - uint16_t len = 0; + if (avail_idx < vq->last_avail_idx) + wrap_counter ^= 1; - *num_buffers = 0; - cur_idx = vq->last_avail_idx; + /* + * Perform a load-acquire barrier in desc_is_avail to + * enforce the ordering between desc flags and desc + * content. + */ + if (unlikely(!desc_is_avail(&descs[avail_idx], wrap_counter))) + return -1; - while (size > 0) { - if (unlikely(cur_idx == avail_head)) + *desc_count = 0; + *len = 0; + + while (1) { + if (unlikely(vec_id >= BUF_VECTOR_MAX)) return -1; - if (unlikely(fill_vec_buf(dev, vq, cur_idx, &vec_idx, buf_vec, - &head_idx, &len) < 0)) + if (unlikely(*desc_count >= vq->size)) return -1; - len = RTE_MIN(len, size); - update_shadow_used_ring(vq, head_idx, len); - size -= len; - cur_idx++; - tries++; - *num_buffers += 1; + *desc_count += 1; + *buf_id = descs[avail_idx].id; - /* - * if we tried all available ring items, and still - * can't get enough buf, it means something abnormal - * happened. - */ - if (unlikely(tries >= vq->size)) - return -1; + if (descs[avail_idx].flags & VRING_DESC_F_INDIRECT) { + if (unlikely(fill_vec_buf_packed_indirect(dev, vq, + &descs[avail_idx], + &vec_id, buf_vec, + len, perm) < 0)) + return -1; + } else { + *len += descs[avail_idx].len; + + if (unlikely(map_one_desc(dev, vq, buf_vec, &vec_id, + descs[avail_idx].addr, + descs[avail_idx].len, + perm))) + return -1; + } + + if ((descs[avail_idx].flags & VRING_DESC_F_NEXT) == 0) + break; + + if (++avail_idx >= vq->size) { + avail_idx -= vq->size; + wrap_counter ^= 1; + } } + *vec_idx = vec_id; + return 0; } +static __rte_noinline void +copy_vnet_hdr_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq, + struct buf_vector *buf_vec, + struct virtio_net_hdr_mrg_rxbuf *hdr) +{ + uint64_t len; + uint64_t remain = dev->vhost_hlen; + uint64_t src = (uint64_t)(uintptr_t)hdr, dst; + uint64_t iova = buf_vec->buf_iova; + + while (remain) { + len = RTE_MIN(remain, + buf_vec->buf_len); + dst = buf_vec->buf_addr; + rte_memcpy((void *)(uintptr_t)dst, + (void *)(uintptr_t)src, + len); + + PRINT_PACKET(dev, (uintptr_t)dst, + (uint32_t)len, 0); + vhost_log_cache_write_iova(dev, vq, + iova, len); + + remain -= len; + iova += len; + src += len; + buf_vec++; + } +} + static __rte_always_inline int -copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq, +copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq, struct rte_mbuf *m, struct buf_vector *buf_vec, - uint16_t num_buffers) + uint16_t nr_vec, uint16_t num_buffers) { uint32_t vec_idx = 0; - uint64_t desc_addr, desc_gaddr; uint32_t mbuf_offset, mbuf_avail; - uint32_t desc_offset, desc_avail; + uint32_t buf_offset, buf_avail; + uint64_t buf_addr, buf_iova, buf_len; uint32_t cpy_len; - uint64_t desc_chunck_len; - uint64_t hdr_addr, hdr_phys_addr; + 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; - uint16_t copy_nb = vq->batch_copy_nb_elems; int error = 0; if (unlikely(m == NULL)) { @@ -655,82 +820,54 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq, goto out; } - desc_chunck_len = buf_vec[vec_idx].buf_len; - desc_gaddr = buf_vec[vec_idx].buf_addr; - desc_addr = vhost_iova_to_vva(dev, vq, - desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RW); - if (buf_vec[vec_idx].buf_len < dev->vhost_hlen || !desc_addr) { + 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 = desc_addr; - if (unlikely(desc_chunck_len < dev->vhost_hlen)) + 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; - hdr_phys_addr = desc_gaddr; - rte_prefetch0((void *)(uintptr_t)hdr_addr); - VHOST_LOG_DEBUG(VHOST_DATA, "(%d) RX: num merge buffers %d\n", + VHOST_LOG_DATA(DEBUG, "(%d) RX: num merge buffers %d\n", dev->vid, num_buffers); - desc_avail = buf_vec[vec_idx].buf_len - dev->vhost_hlen; - if (unlikely(desc_chunck_len < dev->vhost_hlen)) { - desc_chunck_len = desc_avail; - desc_gaddr += dev->vhost_hlen; - desc_addr = vhost_iova_to_vva(dev, vq, - desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RW); - if (unlikely(!desc_addr)) { - error = -1; - goto out; - } - - desc_offset = 0; + 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 { - desc_offset = dev->vhost_hlen; - desc_chunck_len -= dev->vhost_hlen; + 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 desc buf, get the next one */ - if (desc_avail == 0) { + /* done with current buf, get the next one */ + if (buf_avail == 0) { vec_idx++; - desc_chunck_len = buf_vec[vec_idx].buf_len; - desc_gaddr = buf_vec[vec_idx].buf_addr; - desc_addr = - vhost_iova_to_vva(dev, vq, - desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RW); - if (unlikely(!desc_addr)) { + if (unlikely(vec_idx >= nr_vec)) { error = -1; goto out; } - /* Prefetch buffer address. */ - rte_prefetch0((void *)(uintptr_t)desc_addr); - desc_offset = 0; - desc_avail = buf_vec[vec_idx].buf_len; - } else if (unlikely(desc_chunck_len == 0)) { - desc_chunck_len = desc_avail; - desc_gaddr += desc_offset; - desc_addr = vhost_iova_to_vva(dev, vq, - desc_gaddr, - &desc_chunck_len, VHOST_ACCESS_RW); - if (unlikely(!desc_addr)) { - error = -1; - goto out; - } - desc_offset = 0; + 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 */ @@ -743,661 +880,1890 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq, if (hdr_addr) { virtio_enqueue_offload(hdr_mbuf, &hdr->hdr); - ASSIGN_UNLESS_EQUAL(hdr->num_buffers, num_buffers); + if (rxvq_is_mergeable(dev)) + ASSIGN_UNLESS_EQUAL(hdr->num_buffers, + num_buffers); if (unlikely(hdr == &tmp_hdr)) { - uint64_t len; - uint64_t remain = dev->vhost_hlen; - uint64_t src = (uint64_t)(uintptr_t)hdr, dst; - uint64_t guest_addr = hdr_phys_addr; - - while (remain) { - len = remain; - dst = vhost_iova_to_vva(dev, vq, - guest_addr, &len, - VHOST_ACCESS_RW); - if (unlikely(!dst || !len)) { - error = -1; - goto out; - } - - rte_memcpy((void *)(uintptr_t)dst, - (void *)(uintptr_t)src, - len); - - PRINT_PACKET(dev, (uintptr_t)dst, - (uint32_t)len, 0); - vhost_log_write(dev, guest_addr, len); - - remain -= len; - guest_addr += len; - dst += len; - } + copy_vnet_hdr_to_desc(dev, vq, buf_vec, hdr); } else { PRINT_PACKET(dev, (uintptr_t)hdr_addr, dev->vhost_hlen, 0); - vhost_log_write(dev, hdr_phys_addr, + vhost_log_cache_write_iova(dev, vq, + buf_vec[0].buf_iova, dev->vhost_hlen); } hdr_addr = 0; } - cpy_len = RTE_MIN(desc_chunck_len, mbuf_avail); + cpy_len = RTE_MIN(buf_avail, mbuf_avail); - if (likely(cpy_len > MAX_BATCH_LEN || copy_nb >= vq->size)) { - rte_memcpy((void *)((uintptr_t)(desc_addr + - desc_offset)), + if (likely(cpy_len > MAX_BATCH_LEN || + 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); - vhost_log_write(dev, desc_gaddr + desc_offset, cpy_len); - PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset), + vhost_log_cache_write_iova(dev, vq, + buf_iova + buf_offset, + cpy_len); + PRINT_PACKET(dev, (uintptr_t)(buf_addr + buf_offset), cpy_len, 0); } else { - batch_copy[copy_nb].dst = - (void *)((uintptr_t)(desc_addr + desc_offset)); - batch_copy[copy_nb].src = + 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[copy_nb].log_addr = desc_gaddr + desc_offset; - batch_copy[copy_nb].len = cpy_len; - copy_nb++; + 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; - desc_avail -= cpy_len; - desc_offset += cpy_len; - desc_chunck_len -= cpy_len; + buf_avail -= cpy_len; + buf_offset += cpy_len; } out: - vq->batch_copy_nb_elems = copy_nb; return error; } -static __rte_always_inline uint32_t -virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id, - struct rte_mbuf **pkts, uint32_t count) +static __rte_always_inline void +async_fill_vec(struct iovec *v, void *base, size_t len) { - struct vhost_virtqueue *vq; - uint32_t pkt_idx = 0; - uint16_t num_buffers; - struct buf_vector buf_vec[BUF_VECTOR_MAX]; - uint16_t avail_head; + v->iov_base = base; + v->iov_len = len; +} - VHOST_LOG_DEBUG(VHOST_DATA, "(%d) %s\n", dev->vid, __func__); - if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) { - RTE_LOG(ERR, VHOST_DATA, "(%d) %s: invalid virtqueue idx %d.\n", - dev->vid, __func__, queue_id); - return 0; +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; } +} - vq = dev->virtqueue[queue_id]; +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; +} - rte_spinlock_lock(&vq->access_lock); +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; - if (unlikely(vq->enabled == 0)) - goto out_access_unlock; + uint32_t tlen = 0; + int tvec_idx = 0; + void *hpa; - if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) - vhost_user_iotlb_rd_lock(vq); + if (unlikely(m == NULL)) { + error = -1; + goto out; + } - if (unlikely(vq->access_ok == 0)) - if (unlikely(vring_translate(dev, vq) < 0)) - goto out; + cpy_threshold = vq->async_threshold; - count = RTE_MIN((uint32_t)MAX_PKT_BURST, count); - if (count == 0) + 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; + } - vq->batch_copy_nb_elems = 0; + 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; - rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]); + VHOST_LOG_DATA(DEBUG, "(%d) RX: num merge buffers %d\n", + dev->vid, num_buffers); - vq->shadow_used_idx = 0; - avail_head = *((volatile uint16_t *)&vq->avail->idx); - for (pkt_idx = 0; pkt_idx < count; pkt_idx++) { - uint32_t pkt_len = pkts[pkt_idx]->pkt_len + dev->vhost_hlen; + 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; + } - if (unlikely(reserve_avail_buf_mergeable(dev, vq, - pkt_len, buf_vec, &num_buffers, - avail_head) < 0)) { - VHOST_LOG_DEBUG(VHOST_DATA, - "(%d) failed to get enough desc from vring\n", - dev->vid); - vq->shadow_used_idx -= num_buffers; - break; - } + mbuf_avail = rte_pktmbuf_data_len(m); + mbuf_offset = 0; - VHOST_LOG_DEBUG(VHOST_DATA, "(%d) current index %d | end index %d\n", + 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, + struct rte_mbuf *pkt, + struct buf_vector *buf_vec, + uint16_t *nr_descs) +{ + uint16_t nr_vec = 0; + uint16_t avail_idx = vq->last_avail_idx; + uint16_t max_tries, tries = 0; + uint16_t buf_id = 0; + uint32_t len = 0; + uint16_t desc_count; + uint32_t size = pkt->pkt_len + dev->vhost_hlen; + uint16_t num_buffers = 0; + uint32_t buffer_len[vq->size]; + uint16_t buffer_buf_id[vq->size]; + uint16_t buffer_desc_count[vq->size]; + + if (rxvq_is_mergeable(dev)) + max_tries = vq->size - 1; + else + max_tries = 1; + + while (size > 0) { + /* + * if we tried all available ring items, and still + * can't get enough buf, it means something abnormal + * happened. + */ + if (unlikely(++tries > max_tries)) + return -1; + + if (unlikely(fill_vec_buf_packed(dev, vq, + avail_idx, &desc_count, + buf_vec, &nr_vec, + &buf_id, &len, + VHOST_ACCESS_RW) < 0)) + return -1; + + len = RTE_MIN(len, size); + size -= len; + + buffer_len[num_buffers] = len; + buffer_buf_id[num_buffers] = buf_id; + buffer_desc_count[num_buffers] = desc_count; + num_buffers += 1; + + *nr_descs += desc_count; + avail_idx += desc_count; + if (avail_idx >= vq->size) + avail_idx -= vq->size; + } + + if (copy_mbuf_to_desc(dev, vq, pkt, buf_vec, nr_vec, num_buffers) < 0) + return -1; + + vhost_shadow_enqueue_single_packed(dev, vq, buffer_len, buffer_buf_id, + buffer_desc_count, num_buffers); + + return 0; +} + +static __rte_noinline uint32_t +virtio_dev_rx_split(struct virtio_net *dev, struct vhost_virtqueue *vq, + struct rte_mbuf **pkts, uint32_t count) +{ + uint32_t pkt_idx = 0; + uint16_t num_buffers; + struct buf_vector buf_vec[BUF_VECTOR_MAX]; + uint16_t avail_head; + + /* + * The ordering between avail index and + * desc reads needs to be enforced. + */ + avail_head = __atomic_load_n(&vq->avail->idx, __ATOMIC_ACQUIRE); + + 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 (copy_mbuf_to_desc_mergeable(dev, vq, pkts[pkt_idx], - buf_vec, num_buffers) < 0) { + if (copy_mbuf_to_desc(dev, vq, pkts[pkt_idx], + buf_vec, nr_vec, + num_buffers) < 0) { vq->shadow_used_idx -= num_buffers; break; } - - vq->last_avail_idx += num_buffers; + + vq->last_avail_idx += num_buffers; + } + + do_data_copy_enqueue(dev, vq); + + if (likely(vq->shadow_used_idx)) { + flush_shadow_used_ring_split(dev, vq); + vhost_vring_call_split(dev, vq); + } + + return pkt_idx; +} + +static __rte_always_inline int +virtio_dev_rx_batch_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct rte_mbuf **pkts) +{ + bool wrap_counter = vq->avail_wrap_counter; + struct vring_packed_desc *descs = vq->desc_packed; + uint16_t avail_idx = vq->last_avail_idx; + uint64_t desc_addrs[PACKED_BATCH_SIZE]; + struct virtio_net_hdr_mrg_rxbuf *hdrs[PACKED_BATCH_SIZE]; + uint32_t buf_offset = dev->vhost_hlen; + uint64_t lens[PACKED_BATCH_SIZE]; + uint16_t ids[PACKED_BATCH_SIZE]; + uint16_t i; + + if (unlikely(avail_idx & PACKED_BATCH_MASK)) + return -1; + + if (unlikely((avail_idx + PACKED_BATCH_SIZE) > vq->size)) + return -1; + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + if (unlikely(pkts[i]->next != NULL)) + return -1; + if (unlikely(!desc_is_avail(&descs[avail_idx + i], + wrap_counter))) + return -1; + } + + rte_smp_rmb(); + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + lens[i] = descs[avail_idx + i].len; + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + if (unlikely(pkts[i]->pkt_len > (lens[i] - buf_offset))) + return -1; + } + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + desc_addrs[i] = vhost_iova_to_vva(dev, vq, + descs[avail_idx + i].addr, + &lens[i], + 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; + } + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + rte_prefetch0((void *)(uintptr_t)desc_addrs[i]); + hdrs[i] = (struct virtio_net_hdr_mrg_rxbuf *) + (uintptr_t)desc_addrs[i]; + lens[i] = pkts[i]->pkt_len + dev->vhost_hlen; + } + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + virtio_enqueue_offload(pkts[i], &hdrs[i]->hdr); + + vq_inc_last_avail_packed(vq, PACKED_BATCH_SIZE); + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + rte_memcpy((void *)(uintptr_t)(desc_addrs[i] + buf_offset), + rte_pktmbuf_mtod_offset(pkts[i], void *, 0), + pkts[i]->pkt_len); + } + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + vhost_log_cache_write_iova(dev, vq, descs[avail_idx + i].addr, + lens[i]); + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + ids[i] = descs[avail_idx + i].id; + + vhost_flush_enqueue_batch_packed(dev, vq, lens, ids); + + return 0; +} + +static __rte_always_inline int16_t +virtio_dev_rx_single_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct rte_mbuf *pkt) +{ + struct buf_vector buf_vec[BUF_VECTOR_MAX]; + uint16_t nr_descs = 0; + + rte_smp_rmb(); + if (unlikely(vhost_enqueue_single_packed(dev, vq, pkt, buf_vec, + &nr_descs) < 0)) { + VHOST_LOG_DATA(DEBUG, + "(%d) failed to get enough desc from vring\n", + dev->vid); + return -1; + } + + VHOST_LOG_DATA(DEBUG, "(%d) current index %d | end index %d\n", + dev->vid, vq->last_avail_idx, + vq->last_avail_idx + nr_descs); + + vq_inc_last_avail_packed(vq, nr_descs); + + return 0; +} + +static __rte_noinline uint32_t +virtio_dev_rx_packed(struct virtio_net *dev, + struct vhost_virtqueue *__rte_restrict vq, + struct rte_mbuf **__rte_restrict pkts, + uint32_t count) +{ + uint32_t pkt_idx = 0; + uint32_t remained = count; + + do { + rte_prefetch0(&vq->desc_packed[vq->last_avail_idx]); + + if (remained >= PACKED_BATCH_SIZE) { + if (!virtio_dev_rx_batch_packed(dev, vq, + &pkts[pkt_idx])) { + pkt_idx += PACKED_BATCH_SIZE; + remained -= PACKED_BATCH_SIZE; + continue; + } + } + + if (virtio_dev_rx_single_packed(dev, vq, pkts[pkt_idx])) + break; + pkt_idx++; + remained--; + + } while (pkt_idx < count); + + if (vq->shadow_used_idx) { + do_data_copy_enqueue(dev, vq); + vhost_flush_enqueue_shadow_packed(dev, vq); + } + + if (pkt_idx) + vhost_vring_call_packed(dev, vq); + + return pkt_idx; +} + +static __rte_always_inline uint32_t +virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id, + struct rte_mbuf **pkts, uint32_t count) +{ + struct vhost_virtqueue *vq; + uint32_t nb_tx = 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); + + if (unlikely(vq->enabled == 0)) + 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; + + if (vq_is_packed(dev)) + nb_tx = virtio_dev_rx_packed(dev, vq, pkts, count); + else + nb_tx = virtio_dev_rx_split(dev, vq, 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); + + return nb_tx; +} + +uint16_t +rte_vhost_enqueue_burst(int vid, uint16_t queue_id, + struct rte_mbuf **__rte_restrict 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(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) +{ + if (dev->features & + ((1ULL << VIRTIO_NET_F_CSUM) | + (1ULL << VIRTIO_NET_F_HOST_ECN) | + (1ULL << VIRTIO_NET_F_HOST_TSO4) | + (1ULL << VIRTIO_NET_F_HOST_TSO6) | + (1ULL << VIRTIO_NET_F_HOST_UFO))) + return true; + + return false; +} + +static void +parse_ethernet(struct rte_mbuf *m, uint16_t *l4_proto, void **l4_hdr) +{ + struct rte_ipv4_hdr *ipv4_hdr; + struct rte_ipv6_hdr *ipv6_hdr; + void *l3_hdr = NULL; + struct rte_ether_hdr *eth_hdr; + uint16_t ethertype; + + eth_hdr = rte_pktmbuf_mtod(m, struct rte_ether_hdr *); + + m->l2_len = sizeof(struct rte_ether_hdr); + ethertype = rte_be_to_cpu_16(eth_hdr->ether_type); + + if (ethertype == RTE_ETHER_TYPE_VLAN) { + struct rte_vlan_hdr *vlan_hdr = + (struct rte_vlan_hdr *)(eth_hdr + 1); + + m->l2_len += sizeof(struct rte_vlan_hdr); + ethertype = rte_be_to_cpu_16(vlan_hdr->eth_proto); + } + + l3_hdr = (char *)eth_hdr + m->l2_len; + + switch (ethertype) { + case RTE_ETHER_TYPE_IPV4: + ipv4_hdr = l3_hdr; + *l4_proto = ipv4_hdr->next_proto_id; + m->l3_len = (ipv4_hdr->version_ihl & 0x0f) * 4; + *l4_hdr = (char *)l3_hdr + m->l3_len; + m->ol_flags |= PKT_TX_IPV4; + break; + case RTE_ETHER_TYPE_IPV6: + ipv6_hdr = l3_hdr; + *l4_proto = ipv6_hdr->proto; + m->l3_len = sizeof(struct rte_ipv6_hdr); + *l4_hdr = (char *)l3_hdr + m->l3_len; + m->ol_flags |= PKT_TX_IPV6; + break; + default: + m->l3_len = 0; + *l4_proto = 0; + *l4_hdr = NULL; + break; + } +} + +static __rte_always_inline void +vhost_dequeue_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *m) +{ + uint16_t l4_proto = 0; + void *l4_hdr = NULL; + struct rte_tcp_hdr *tcp_hdr = NULL; + + if (hdr->flags == 0 && hdr->gso_type == VIRTIO_NET_HDR_GSO_NONE) + return; + + parse_ethernet(m, &l4_proto, &l4_hdr); + if (hdr->flags == VIRTIO_NET_HDR_F_NEEDS_CSUM) { + if (hdr->csum_start == (m->l2_len + m->l3_len)) { + switch (hdr->csum_offset) { + case (offsetof(struct rte_tcp_hdr, cksum)): + if (l4_proto == IPPROTO_TCP) + m->ol_flags |= PKT_TX_TCP_CKSUM; + break; + case (offsetof(struct rte_udp_hdr, dgram_cksum)): + if (l4_proto == IPPROTO_UDP) + m->ol_flags |= PKT_TX_UDP_CKSUM; + break; + case (offsetof(struct rte_sctp_hdr, cksum)): + if (l4_proto == IPPROTO_SCTP) + m->ol_flags |= PKT_TX_SCTP_CKSUM; + break; + default: + break; + } + } + } + + if (l4_hdr && hdr->gso_type != VIRTIO_NET_HDR_GSO_NONE) { + switch (hdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN) { + case VIRTIO_NET_HDR_GSO_TCPV4: + case VIRTIO_NET_HDR_GSO_TCPV6: + tcp_hdr = l4_hdr; + m->ol_flags |= PKT_TX_TCP_SEG; + m->tso_segsz = hdr->gso_size; + m->l4_len = (tcp_hdr->data_off & 0xf0) >> 2; + break; + case VIRTIO_NET_HDR_GSO_UDP: + m->ol_flags |= PKT_TX_UDP_SEG; + m->tso_segsz = hdr->gso_size; + m->l4_len = sizeof(struct rte_udp_hdr); + break; + default: + VHOST_LOG_DATA(WARNING, + "unsupported gso type %u.\n", hdr->gso_type); + break; + } + } +} + +static __rte_noinline void +copy_vnet_hdr_from_desc(struct virtio_net_hdr *hdr, + struct buf_vector *buf_vec) +{ + uint64_t len; + uint64_t remain = sizeof(struct virtio_net_hdr); + uint64_t src; + uint64_t dst = (uint64_t)(uintptr_t)hdr; + + while (remain) { + len = RTE_MIN(remain, buf_vec->buf_len); + src = buf_vec->buf_addr; + rte_memcpy((void *)(uintptr_t)dst, + (void *)(uintptr_t)src, len); + + remain -= len; + dst += len; + buf_vec++; + } +} + +static __rte_always_inline int +copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq, + struct buf_vector *buf_vec, uint16_t nr_vec, + struct rte_mbuf *m, struct rte_mempool *mbuf_pool) +{ + uint32_t buf_avail, buf_offset; + uint64_t buf_addr, buf_iova, buf_len; + uint32_t mbuf_avail, mbuf_offset; + uint32_t cpy_len; + struct rte_mbuf *cur = m, *prev = m; + struct virtio_net_hdr tmp_hdr; + struct virtio_net_hdr *hdr = NULL; + /* A counter to avoid desc dead loop chain */ + uint16_t vec_idx = 0; + struct batch_copy_elem *batch_copy = vq->batch_copy_elems; + int error = 0; + + 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; + } + + if (virtio_net_with_host_offload(dev)) { + if (unlikely(buf_len < sizeof(struct virtio_net_hdr))) { + /* + * No luck, the virtio-net header doesn't fit + * in a contiguous virtual area. + */ + copy_vnet_hdr_from_desc(&tmp_hdr, buf_vec); + hdr = &tmp_hdr; + } else { + hdr = (struct virtio_net_hdr *)((uintptr_t)buf_addr); + } + } + + /* + * A virtio driver normally uses at least 2 desc buffers + * for Tx: the first for storing the header, and others + * for storing the data. + */ + 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 if (buf_len == dev->vhost_hlen) { + if (unlikely(++vec_idx >= nr_vec)) + 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; + } else { + buf_offset = dev->vhost_hlen; + buf_avail = buf_vec[vec_idx].buf_len - dev->vhost_hlen; + } + + PRINT_PACKET(dev, + (uintptr_t)(buf_addr + buf_offset), + (uint32_t)buf_avail, 0); + + mbuf_offset = 0; + mbuf_avail = m->buf_len - RTE_PKTMBUF_HEADROOM; + while (1) { + uint64_t hpa; + + cpy_len = RTE_MIN(buf_avail, mbuf_avail); + + /* + * A desc buf might across two host physical pages that are + * not continuous. In such case (gpa_to_hpa returns 0), data + * will be copied even though zero copy is enabled. + */ + if (unlikely(dev->dequeue_zero_copy && (hpa = gpa_to_hpa(dev, + buf_iova + buf_offset, cpy_len)))) { + cur->data_len = cpy_len; + cur->data_off = 0; + cur->buf_addr = + (void *)(uintptr_t)(buf_addr + buf_offset); + cur->buf_iova = hpa; + + /* + * In zero copy mode, one mbuf can only reference data + * for one or partial of one desc buff. + */ + mbuf_avail = cpy_len; + } else { + if (likely(cpy_len > MAX_BATCH_LEN || + vq->batch_copy_nb_elems >= vq->size || + (hdr && cur == m))) { + rte_memcpy(rte_pktmbuf_mtod_offset(cur, void *, + mbuf_offset), + (void *)((uintptr_t)(buf_addr + + buf_offset)), + cpy_len); + } else { + batch_copy[vq->batch_copy_nb_elems].dst = + rte_pktmbuf_mtod_offset(cur, void *, + mbuf_offset); + batch_copy[vq->batch_copy_nb_elems].src = + (void *)((uintptr_t)(buf_addr + + 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; + + /* This buf reaches to its end, get the next one */ + if (buf_avail == 0) { + if (++vec_idx >= nr_vec) + break; + + 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; + + PRINT_PACKET(dev, (uintptr_t)buf_addr, + (uint32_t)buf_avail, 0); + } + + /* + * This mbuf reaches to its end, get a new one + * to hold more data. + */ + if (mbuf_avail == 0) { + cur = rte_pktmbuf_alloc(mbuf_pool); + if (unlikely(cur == NULL)) { + VHOST_LOG_DATA(ERR, "Failed to " + "allocate memory for mbuf.\n"); + error = -1; + goto out; + } + if (unlikely(dev->dequeue_zero_copy)) + rte_mbuf_refcnt_update(cur, 1); + + prev->next = cur; + prev->data_len = mbuf_offset; + m->nb_segs += 1; + m->pkt_len += mbuf_offset; + prev = cur; + + mbuf_offset = 0; + mbuf_avail = cur->buf_len - RTE_PKTMBUF_HEADROOM; + } + } + + prev->data_len = mbuf_offset; + m->pkt_len += mbuf_offset; + + if (hdr) + vhost_dequeue_offload(hdr, m); + +out: + + return error; +} + +static __rte_always_inline struct zcopy_mbuf * +get_zmbuf(struct vhost_virtqueue *vq) +{ + uint16_t i; + uint16_t last; + int tries = 0; + + /* search [last_zmbuf_idx, zmbuf_size) */ + i = vq->last_zmbuf_idx; + last = vq->zmbuf_size; + +again: + for (; i < last; i++) { + if (vq->zmbufs[i].in_use == 0) { + vq->last_zmbuf_idx = i + 1; + vq->zmbufs[i].in_use = 1; + return &vq->zmbufs[i]; + } } - do_data_copy_enqueue(dev, vq); - - if (likely(vq->shadow_used_idx)) { - flush_shadow_used_ring(dev, vq); - vhost_vring_call(dev, vq); + tries++; + if (tries == 1) { + /* search [0, last_zmbuf_idx) */ + i = 0; + last = vq->last_zmbuf_idx; + goto again; } -out: - if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) - vhost_user_iotlb_rd_unlock(vq); - -out_access_unlock: - rte_spinlock_unlock(&vq->access_lock); + return NULL; +} - return pkt_idx; +static void +virtio_dev_extbuf_free(void *addr __rte_unused, void *opaque) +{ + rte_free(opaque); } -uint16_t -rte_vhost_enqueue_burst(int vid, uint16_t queue_id, - struct rte_mbuf **pkts, uint16_t count) +static int +virtio_dev_extbuf_alloc(struct rte_mbuf *pkt, uint32_t size) { - struct virtio_net *dev = get_device(vid); + struct rte_mbuf_ext_shared_info *shinfo = NULL; + uint32_t total_len = RTE_PKTMBUF_HEADROOM + size; + uint16_t buf_len; + rte_iova_t iova; + void *buf; + + /* Try to use pkt buffer to store shinfo to reduce the amount of memory + * required, otherwise store shinfo in the new buffer. + */ + if (rte_pktmbuf_tailroom(pkt) >= sizeof(*shinfo)) + shinfo = rte_pktmbuf_mtod(pkt, + struct rte_mbuf_ext_shared_info *); + else { + total_len += sizeof(*shinfo) + sizeof(uintptr_t); + total_len = RTE_ALIGN_CEIL(total_len, sizeof(uintptr_t)); + } - if (!dev) - return 0; + if (unlikely(total_len > UINT16_MAX)) + return -ENOSPC; - if (unlikely(!(dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET))) { - RTE_LOG(ERR, VHOST_DATA, - "(%d) %s: built-in vhost net backend is disabled.\n", - dev->vid, __func__); - return 0; - } + buf_len = total_len; + buf = rte_malloc(NULL, buf_len, RTE_CACHE_LINE_SIZE); + if (unlikely(buf == NULL)) + return -ENOMEM; - if (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF)) - return virtio_dev_merge_rx(dev, queue_id, pkts, count); - else - return virtio_dev_rx(dev, queue_id, pkts, count); -} + /* Initialize shinfo */ + if (shinfo) { + shinfo->free_cb = virtio_dev_extbuf_free; + shinfo->fcb_opaque = buf; + rte_mbuf_ext_refcnt_set(shinfo, 1); + } else { + shinfo = rte_pktmbuf_ext_shinfo_init_helper(buf, &buf_len, + virtio_dev_extbuf_free, buf); + if (unlikely(shinfo == NULL)) { + rte_free(buf); + VHOST_LOG_DATA(ERR, "Failed to init shinfo\n"); + return -1; + } + } -static inline bool -virtio_net_with_host_offload(struct virtio_net *dev) -{ - if (dev->features & - ((1ULL << VIRTIO_NET_F_CSUM) | - (1ULL << VIRTIO_NET_F_HOST_ECN) | - (1ULL << VIRTIO_NET_F_HOST_TSO4) | - (1ULL << VIRTIO_NET_F_HOST_TSO6) | - (1ULL << VIRTIO_NET_F_HOST_UFO))) - return true; + iova = rte_malloc_virt2iova(buf); + rte_pktmbuf_attach_extbuf(pkt, buf, iova, buf_len, shinfo); + rte_pktmbuf_reset_headroom(pkt); - return false; + return 0; } -static void -parse_ethernet(struct rte_mbuf *m, uint16_t *l4_proto, void **l4_hdr) +/* + * Allocate a host supported pktmbuf. + */ +static __rte_always_inline struct rte_mbuf * +virtio_dev_pktmbuf_alloc(struct virtio_net *dev, struct rte_mempool *mp, + uint32_t data_len) { - struct ipv4_hdr *ipv4_hdr; - struct ipv6_hdr *ipv6_hdr; - void *l3_hdr = NULL; - struct ether_hdr *eth_hdr; - uint16_t ethertype; + struct rte_mbuf *pkt = rte_pktmbuf_alloc(mp); - eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *); + if (unlikely(pkt == NULL)) { + VHOST_LOG_DATA(ERR, + "Failed to allocate memory for mbuf.\n"); + return NULL; + } - m->l2_len = sizeof(struct ether_hdr); - ethertype = rte_be_to_cpu_16(eth_hdr->ether_type); + if (rte_pktmbuf_tailroom(pkt) >= data_len) + return pkt; - if (ethertype == ETHER_TYPE_VLAN) { - struct vlan_hdr *vlan_hdr = (struct vlan_hdr *)(eth_hdr + 1); + /* attach an external buffer if supported */ + if (dev->extbuf && !virtio_dev_extbuf_alloc(pkt, data_len)) + return pkt; - m->l2_len += sizeof(struct vlan_hdr); - ethertype = rte_be_to_cpu_16(vlan_hdr->eth_proto); - } + /* check if chained buffers are allowed */ + if (!dev->linearbuf) + return pkt; - l3_hdr = (char *)eth_hdr + m->l2_len; + /* Data doesn't fit into the buffer and the host supports + * only linear buffers + */ + rte_pktmbuf_free(pkt); - switch (ethertype) { - case ETHER_TYPE_IPv4: - ipv4_hdr = l3_hdr; - *l4_proto = ipv4_hdr->next_proto_id; - m->l3_len = (ipv4_hdr->version_ihl & 0x0f) * 4; - *l4_hdr = (char *)l3_hdr + m->l3_len; - m->ol_flags |= PKT_TX_IPV4; - break; - case ETHER_TYPE_IPv6: - ipv6_hdr = l3_hdr; - *l4_proto = ipv6_hdr->proto; - m->l3_len = sizeof(struct ipv6_hdr); - *l4_hdr = (char *)l3_hdr + m->l3_len; - m->ol_flags |= PKT_TX_IPV6; - break; - default: - m->l3_len = 0; - *l4_proto = 0; - *l4_hdr = NULL; - break; - } + return NULL; } -static __rte_always_inline void -vhost_dequeue_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *m) +static __rte_noinline uint16_t +virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count) { - uint16_t l4_proto = 0; - void *l4_hdr = NULL; - struct tcp_hdr *tcp_hdr = NULL; + uint16_t i; + uint16_t free_entries; + uint16_t dropped = 0; + static bool allocerr_warned; - if (hdr->flags == 0 && hdr->gso_type == VIRTIO_NET_HDR_GSO_NONE) - return; + if (unlikely(dev->dequeue_zero_copy)) { + struct zcopy_mbuf *zmbuf, *next; - parse_ethernet(m, &l4_proto, &l4_hdr); - if (hdr->flags == VIRTIO_NET_HDR_F_NEEDS_CSUM) { - if (hdr->csum_start == (m->l2_len + m->l3_len)) { - switch (hdr->csum_offset) { - case (offsetof(struct tcp_hdr, cksum)): - if (l4_proto == IPPROTO_TCP) - m->ol_flags |= PKT_TX_TCP_CKSUM; - break; - case (offsetof(struct udp_hdr, dgram_cksum)): - if (l4_proto == IPPROTO_UDP) - m->ol_flags |= PKT_TX_UDP_CKSUM; - break; - case (offsetof(struct sctp_hdr, cksum)): - if (l4_proto == IPPROTO_SCTP) - m->ol_flags |= PKT_TX_SCTP_CKSUM; - break; - default: - break; + for (zmbuf = TAILQ_FIRST(&vq->zmbuf_list); + zmbuf != NULL; zmbuf = next) { + next = TAILQ_NEXT(zmbuf, next); + + if (mbuf_is_consumed(zmbuf->mbuf)) { + update_shadow_used_ring_split(vq, + zmbuf->desc_idx, 0); + TAILQ_REMOVE(&vq->zmbuf_list, zmbuf, next); + restore_mbuf(zmbuf->mbuf); + rte_pktmbuf_free(zmbuf->mbuf); + put_zmbuf(zmbuf); + vq->nr_zmbuf -= 1; } } + + if (likely(vq->shadow_used_idx)) { + flush_shadow_used_ring_split(dev, vq); + vhost_vring_call_split(dev, vq); + } } - if (l4_hdr && hdr->gso_type != VIRTIO_NET_HDR_GSO_NONE) { - switch (hdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN) { - case VIRTIO_NET_HDR_GSO_TCPV4: - case VIRTIO_NET_HDR_GSO_TCPV6: - tcp_hdr = l4_hdr; - m->ol_flags |= PKT_TX_TCP_SEG; - m->tso_segsz = hdr->gso_size; - m->l4_len = (tcp_hdr->data_off & 0xf0) >> 2; + /* + * The ordering between avail index and + * desc reads needs to be enforced. + */ + free_entries = __atomic_load_n(&vq->avail->idx, __ATOMIC_ACQUIRE) - + vq->last_avail_idx; + if (free_entries == 0) + return 0; + + rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]); + + VHOST_LOG_DATA(DEBUG, "(%d) %s\n", dev->vid, __func__); + + count = RTE_MIN(count, MAX_PKT_BURST); + count = RTE_MIN(count, free_entries); + VHOST_LOG_DATA(DEBUG, "(%d) about to dequeue %u buffers\n", + dev->vid, count); + + for (i = 0; i < count; i++) { + struct buf_vector buf_vec[BUF_VECTOR_MAX]; + uint16_t head_idx; + uint32_t buf_len; + uint16_t nr_vec = 0; + int err; + + if (unlikely(fill_vec_buf_split(dev, vq, + vq->last_avail_idx + i, + &nr_vec, buf_vec, + &head_idx, &buf_len, + VHOST_ACCESS_RO) < 0)) break; - case VIRTIO_NET_HDR_GSO_UDP: - m->ol_flags |= PKT_TX_UDP_SEG; - m->tso_segsz = hdr->gso_size; - m->l4_len = sizeof(struct udp_hdr); + + if (likely(dev->dequeue_zero_copy == 0)) + 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)) { + /* + * 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; - default: - RTE_LOG(WARNING, VHOST_DATA, - "unsupported gso type %u.\n", hdr->gso_type); + } + + 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; } + + if (unlikely(dev->dequeue_zero_copy)) { + struct zcopy_mbuf *zmbuf; + + zmbuf = get_zmbuf(vq); + if (!zmbuf) { + rte_pktmbuf_free(pkts[i]); + dropped += 1; + i++; + break; + } + zmbuf->mbuf = pkts[i]; + zmbuf->desc_idx = head_idx; + + /* + * Pin lock the mbuf; we will check later to see + * whether the mbuf is freed (when we are the last + * user) or not. If that's the case, we then could + * update the used ring safely. + */ + rte_mbuf_refcnt_update(pkts[i], 1); + + vq->nr_zmbuf += 1; + TAILQ_INSERT_TAIL(&vq->zmbuf_list, zmbuf, next); + } } -} + vq->last_avail_idx += i; -static __rte_always_inline void -put_zmbuf(struct zcopy_mbuf *zmbuf) -{ - zmbuf->in_use = 0; + if (likely(dev->dequeue_zero_copy == 0)) { + do_data_copy_dequeue(vq); + if (unlikely(i < count)) + vq->shadow_used_idx = i; + if (likely(vq->shadow_used_idx)) { + flush_shadow_used_ring_split(dev, vq); + vhost_vring_call_split(dev, vq); + } + } + + return (i - dropped); } static __rte_always_inline int -copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq, - struct vring_desc *descs, uint16_t max_desc, - struct rte_mbuf *m, uint16_t desc_idx, - struct rte_mempool *mbuf_pool) +vhost_reserve_avail_batch_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, + struct rte_mbuf **pkts, + uint16_t avail_idx, + uintptr_t *desc_addrs, + uint16_t *ids) { - struct vring_desc *desc; - uint64_t desc_addr, desc_gaddr; - uint32_t desc_avail, desc_offset; - uint32_t mbuf_avail, mbuf_offset; - uint32_t cpy_len; - uint64_t desc_chunck_len; - struct rte_mbuf *cur = m, *prev = m; - struct virtio_net_hdr tmp_hdr; - struct virtio_net_hdr *hdr = NULL; - /* A counter to avoid desc dead loop chain */ - uint32_t nr_desc = 1; - struct batch_copy_elem *batch_copy = vq->batch_copy_elems; - uint16_t copy_nb = vq->batch_copy_nb_elems; - int error = 0; + bool wrap = vq->avail_wrap_counter; + struct vring_packed_desc *descs = vq->desc_packed; + struct virtio_net_hdr *hdr; + uint64_t lens[PACKED_BATCH_SIZE]; + uint64_t buf_lens[PACKED_BATCH_SIZE]; + uint32_t buf_offset = dev->vhost_hlen; + uint16_t flags, i; + + if (unlikely(avail_idx & PACKED_BATCH_MASK)) + return -1; + if (unlikely((avail_idx + PACKED_BATCH_SIZE) > vq->size)) + return -1; + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + flags = descs[avail_idx + i].flags; + if (unlikely((wrap != !!(flags & VRING_DESC_F_AVAIL)) || + (wrap == !!(flags & VRING_DESC_F_USED)) || + (flags & PACKED_DESC_SINGLE_DEQUEUE_FLAG))) + return -1; + } - desc = &descs[desc_idx]; - if (unlikely((desc->len < dev->vhost_hlen)) || - (desc->flags & VRING_DESC_F_INDIRECT)) { - error = -1; - goto out; + rte_smp_rmb(); + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + lens[i] = descs[avail_idx + i].len; + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + desc_addrs[i] = vhost_iova_to_vva(dev, vq, + descs[avail_idx + i].addr, + &lens[i], VHOST_ACCESS_RW); } - desc_chunck_len = desc->len; - desc_gaddr = desc->addr; - desc_addr = vhost_iova_to_vva(dev, - vq, desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RO); - if (unlikely(!desc_addr)) { - error = -1; - goto out; + 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; } - if (virtio_net_with_host_offload(dev)) { - if (unlikely(desc_chunck_len < sizeof(struct virtio_net_hdr))) { - uint64_t len = desc_chunck_len; - uint64_t remain = sizeof(struct virtio_net_hdr); - uint64_t src = desc_addr; - uint64_t dst = (uint64_t)(uintptr_t)&tmp_hdr; - uint64_t guest_addr = desc_gaddr; + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + pkts[i] = virtio_dev_pktmbuf_alloc(dev, mbuf_pool, lens[i]); + if (!pkts[i]) + goto free_buf; + } - /* - * No luck, the virtio-net header doesn't fit - * in a contiguous virtual area. - */ - while (remain) { - len = remain; - src = vhost_iova_to_vva(dev, vq, - guest_addr, &len, - VHOST_ACCESS_RO); - if (unlikely(!src || !len)) { - error = -1; - goto out; - } + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + buf_lens[i] = pkts[i]->buf_len - pkts[i]->data_off; - rte_memcpy((void *)(uintptr_t)dst, - (void *)(uintptr_t)src, len); + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + if (unlikely(buf_lens[i] < (lens[i] - buf_offset))) + goto free_buf; + } - guest_addr += len; - remain -= len; - dst += len; - } + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + pkts[i]->pkt_len = descs[avail_idx + i].len - buf_offset; + pkts[i]->data_len = pkts[i]->pkt_len; + ids[i] = descs[avail_idx + i].id; + } - hdr = &tmp_hdr; - } else { - hdr = (struct virtio_net_hdr *)((uintptr_t)desc_addr); - rte_prefetch0(hdr); + if (virtio_net_with_host_offload(dev)) { + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + hdr = (struct virtio_net_hdr *)(desc_addrs[i]); + vhost_dequeue_offload(hdr, pkts[i]); } } - /* - * A virtio driver normally uses at least 2 desc buffers - * for Tx: the first for storing the header, and others - * for storing the data. - */ - if (likely((desc->len == dev->vhost_hlen) && - (desc->flags & VRING_DESC_F_NEXT) != 0)) { - desc = &descs[desc->next]; - if (unlikely(desc->flags & VRING_DESC_F_INDIRECT)) { - error = -1; - goto out; - } + return 0; - desc_chunck_len = desc->len; - desc_gaddr = desc->addr; - desc_addr = vhost_iova_to_vva(dev, - vq, desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RO); - if (unlikely(!desc_addr)) { - error = -1; - goto out; +free_buf: + for (i = 0; i < PACKED_BATCH_SIZE; i++) + rte_pktmbuf_free(pkts[i]); + + return -1; +} + +static __rte_always_inline int +virtio_dev_tx_batch_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, + struct rte_mbuf **pkts) +{ + uint16_t avail_idx = vq->last_avail_idx; + uint32_t buf_offset = dev->vhost_hlen; + uintptr_t desc_addrs[PACKED_BATCH_SIZE]; + uint16_t ids[PACKED_BATCH_SIZE]; + uint16_t i; + + if (vhost_reserve_avail_batch_packed(dev, vq, mbuf_pool, pkts, + avail_idx, desc_addrs, ids)) + return -1; + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + rte_prefetch0((void *)(uintptr_t)desc_addrs[i]); + + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + rte_memcpy(rte_pktmbuf_mtod_offset(pkts[i], void *, 0), + (void *)(uintptr_t)(desc_addrs[i] + buf_offset), + pkts[i]->pkt_len); + + if (virtio_net_is_inorder(dev)) + vhost_shadow_dequeue_batch_packed_inorder(vq, + ids[PACKED_BATCH_SIZE - 1]); + else + vhost_shadow_dequeue_batch_packed(dev, vq, ids); + + vq_inc_last_avail_packed(vq, PACKED_BATCH_SIZE); + + return 0; +} + +static __rte_always_inline int +vhost_dequeue_single_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, + struct rte_mbuf **pkts, + uint16_t *buf_id, + uint16_t *desc_count) +{ + struct buf_vector buf_vec[BUF_VECTOR_MAX]; + 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, + buf_vec, &nr_vec, + buf_id, &buf_len, + VHOST_ACCESS_RO) < 0)) + return -1; + + *pkts = virtio_dev_pktmbuf_alloc(dev, mbuf_pool, buf_len); + if (unlikely(*pkts == NULL)) { + 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; + } - desc_offset = 0; - desc_avail = desc->len; - nr_desc += 1; - } else { - desc_avail = desc->len - dev->vhost_hlen; - - if (unlikely(desc_chunck_len < dev->vhost_hlen)) { - desc_chunck_len = desc_avail; - desc_gaddr += dev->vhost_hlen; - desc_addr = vhost_iova_to_vva(dev, - vq, desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RO); - if (unlikely(!desc_addr)) { - error = -1; - goto out; - } - - desc_offset = 0; - } else { - desc_offset = dev->vhost_hlen; - desc_chunck_len -= dev->vhost_hlen; + 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; } - rte_prefetch0((void *)(uintptr_t)(desc_addr + desc_offset)); + return 0; +} - PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset), - (uint32_t)desc_chunck_len, 0); +static __rte_always_inline int +virtio_dev_tx_single_packed(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, + struct rte_mbuf **pkts) +{ - mbuf_offset = 0; - mbuf_avail = m->buf_len - RTE_PKTMBUF_HEADROOM; - while (1) { - uint64_t hpa; + uint16_t buf_id, desc_count = 0; + int ret; - cpy_len = RTE_MIN(desc_chunck_len, mbuf_avail); + ret = vhost_dequeue_single_packed(dev, vq, mbuf_pool, pkts, &buf_id, + &desc_count); - /* - * A desc buf might across two host physical pages that are - * not continuous. In such case (gpa_to_hpa returns 0), data - * will be copied even though zero copy is enabled. - */ - if (unlikely(dev->dequeue_zero_copy && (hpa = gpa_to_hpa(dev, - desc_gaddr + desc_offset, cpy_len)))) { - cur->data_len = cpy_len; - cur->data_off = 0; - cur->buf_addr = (void *)(uintptr_t)(desc_addr - + desc_offset); - cur->buf_iova = hpa; + 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); - /* - * In zero copy mode, one mbuf can only reference data - * for one or partial of one desc buff. - */ - mbuf_avail = cpy_len; - } else { - if (likely(cpy_len > MAX_BATCH_LEN || - copy_nb >= vq->size || - (hdr && cur == m) || - desc->len != desc_chunck_len)) { - rte_memcpy(rte_pktmbuf_mtod_offset(cur, void *, - mbuf_offset), - (void *)((uintptr_t)(desc_addr + - desc_offset)), - cpy_len); - } else { - batch_copy[copy_nb].dst = - rte_pktmbuf_mtod_offset(cur, void *, - mbuf_offset); - batch_copy[copy_nb].src = - (void *)((uintptr_t)(desc_addr + - desc_offset)); - batch_copy[copy_nb].len = cpy_len; - copy_nb++; - } - } + vq_inc_last_avail_packed(vq, desc_count); + } - mbuf_avail -= cpy_len; - mbuf_offset += cpy_len; - desc_avail -= cpy_len; - desc_chunck_len -= cpy_len; - desc_offset += cpy_len; + return ret; +} - /* This desc reaches to its end, get the next one */ - if (desc_avail == 0) { - if ((desc->flags & VRING_DESC_F_NEXT) == 0) - break; +static __rte_always_inline int +virtio_dev_tx_batch_packed_zmbuf(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, + struct rte_mbuf **pkts) +{ + struct zcopy_mbuf *zmbufs[PACKED_BATCH_SIZE]; + uintptr_t desc_addrs[PACKED_BATCH_SIZE]; + uint16_t ids[PACKED_BATCH_SIZE]; + uint16_t i; - if (unlikely(desc->next >= max_desc || - ++nr_desc > max_desc)) { - error = -1; - goto out; - } - desc = &descs[desc->next]; - if (unlikely(desc->flags & VRING_DESC_F_INDIRECT)) { - error = -1; - goto out; - } + uint16_t avail_idx = vq->last_avail_idx; - desc_chunck_len = desc->len; - desc_gaddr = desc->addr; - desc_addr = vhost_iova_to_vva(dev, - vq, desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RO); - if (unlikely(!desc_addr)) { - error = -1; - goto out; - } + if (vhost_reserve_avail_batch_packed(dev, vq, mbuf_pool, pkts, + avail_idx, desc_addrs, ids)) + return -1; - rte_prefetch0((void *)(uintptr_t)desc_addr); - - desc_offset = 0; - desc_avail = desc->len; - - PRINT_PACKET(dev, (uintptr_t)desc_addr, - (uint32_t)desc_chunck_len, 0); - } else if (unlikely(desc_chunck_len == 0)) { - desc_chunck_len = desc_avail; - desc_gaddr += desc_offset; - desc_addr = vhost_iova_to_vva(dev, vq, - desc_gaddr, - &desc_chunck_len, - VHOST_ACCESS_RO); - if (unlikely(!desc_addr)) { - error = -1; - goto out; - } - desc_offset = 0; + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + zmbufs[i] = get_zmbuf(vq); - PRINT_PACKET(dev, (uintptr_t)desc_addr, - (uint32_t)desc_chunck_len, 0); - } + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + if (!zmbufs[i]) + goto free_pkt; + } - /* - * This mbuf reaches to its end, get a new one - * to hold more data. - */ - if (mbuf_avail == 0) { - cur = rte_pktmbuf_alloc(mbuf_pool); - if (unlikely(cur == NULL)) { - RTE_LOG(ERR, VHOST_DATA, "Failed to " - "allocate memory for mbuf.\n"); - error = -1; - goto out; - } - if (unlikely(dev->dequeue_zero_copy)) - rte_mbuf_refcnt_update(cur, 1); + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) { + zmbufs[i]->mbuf = pkts[i]; + zmbufs[i]->desc_idx = ids[i]; + zmbufs[i]->desc_count = 1; + } - prev->next = cur; - prev->data_len = mbuf_offset; - m->nb_segs += 1; - m->pkt_len += mbuf_offset; - prev = cur; + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + rte_mbuf_refcnt_update(pkts[i], 1); - mbuf_offset = 0; - mbuf_avail = cur->buf_len - RTE_PKTMBUF_HEADROOM; - } - } + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + TAILQ_INSERT_TAIL(&vq->zmbuf_list, zmbufs[i], next); - prev->data_len = mbuf_offset; - m->pkt_len += mbuf_offset; + vq->nr_zmbuf += PACKED_BATCH_SIZE; + vq_inc_last_avail_packed(vq, PACKED_BATCH_SIZE); - if (hdr) - vhost_dequeue_offload(hdr, m); + return 0; -out: - vq->batch_copy_nb_elems = copy_nb; +free_pkt: + vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) + rte_pktmbuf_free(pkts[i]); - return error; + return -1; } -static __rte_always_inline void -update_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq, - uint32_t used_idx, uint32_t desc_idx) +static __rte_always_inline int +virtio_dev_tx_single_packed_zmbuf(struct virtio_net *dev, + struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, + struct rte_mbuf **pkts) { - vq->used->ring[used_idx].id = desc_idx; - vq->used->ring[used_idx].len = 0; - vhost_log_used_vring(dev, vq, - offsetof(struct vring_used, ring[used_idx]), - sizeof(vq->used->ring[used_idx])); + uint16_t buf_id, desc_count; + struct zcopy_mbuf *zmbuf; + + if (vhost_dequeue_single_packed(dev, vq, mbuf_pool, pkts, &buf_id, + &desc_count)) + return -1; + + zmbuf = get_zmbuf(vq); + if (!zmbuf) { + rte_pktmbuf_free(*pkts); + return -1; + } + zmbuf->mbuf = *pkts; + zmbuf->desc_idx = buf_id; + zmbuf->desc_count = desc_count; + + rte_mbuf_refcnt_update(*pkts, 1); + + vq->nr_zmbuf += 1; + TAILQ_INSERT_TAIL(&vq->zmbuf_list, zmbuf, next); + + vq_inc_last_avail_packed(vq, desc_count); + return 0; } static __rte_always_inline void -update_used_idx(struct virtio_net *dev, struct vhost_virtqueue *vq, - uint32_t count) +free_zmbuf(struct vhost_virtqueue *vq) { - if (unlikely(count == 0)) - return; + struct zcopy_mbuf *next = NULL; + struct zcopy_mbuf *zmbuf; - rte_smp_wmb(); - rte_smp_rmb(); + for (zmbuf = TAILQ_FIRST(&vq->zmbuf_list); + zmbuf != NULL; zmbuf = next) { + next = TAILQ_NEXT(zmbuf, next); - vq->used->idx += count; - vhost_log_used_vring(dev, vq, offsetof(struct vring_used, idx), - sizeof(vq->used->idx)); - vhost_vring_call(dev, vq); + uint16_t last_used_idx = vq->last_used_idx; + + if (mbuf_is_consumed(zmbuf->mbuf)) { + uint16_t flags; + flags = vq->desc_packed[last_used_idx].flags; + if (vq->used_wrap_counter) { + flags |= VRING_DESC_F_USED; + flags |= VRING_DESC_F_AVAIL; + } else { + flags &= ~VRING_DESC_F_USED; + flags &= ~VRING_DESC_F_AVAIL; + } + + vq->desc_packed[last_used_idx].id = zmbuf->desc_idx; + vq->desc_packed[last_used_idx].len = 0; + + rte_smp_wmb(); + vq->desc_packed[last_used_idx].flags = flags; + + vq_inc_last_used_packed(vq, zmbuf->desc_count); + + TAILQ_REMOVE(&vq->zmbuf_list, zmbuf, next); + restore_mbuf(zmbuf->mbuf); + rte_pktmbuf_free(zmbuf->mbuf); + put_zmbuf(zmbuf); + vq->nr_zmbuf -= 1; + } + } } -static __rte_always_inline struct zcopy_mbuf * -get_zmbuf(struct vhost_virtqueue *vq) +static __rte_noinline uint16_t +virtio_dev_tx_packed_zmbuf(struct virtio_net *dev, + struct vhost_virtqueue *__rte_restrict vq, + struct rte_mempool *mbuf_pool, + struct rte_mbuf **__rte_restrict pkts, + uint32_t count) { - uint16_t i; - uint16_t last; - int tries = 0; + uint32_t pkt_idx = 0; + uint32_t remained = count; - /* search [last_zmbuf_idx, zmbuf_size) */ - i = vq->last_zmbuf_idx; - last = vq->zmbuf_size; + free_zmbuf(vq); -again: - for (; i < last; i++) { - if (vq->zmbufs[i].in_use == 0) { - vq->last_zmbuf_idx = i + 1; - vq->zmbufs[i].in_use = 1; - return &vq->zmbufs[i]; + do { + if (remained >= PACKED_BATCH_SIZE) { + if (!virtio_dev_tx_batch_packed_zmbuf(dev, vq, + mbuf_pool, &pkts[pkt_idx])) { + pkt_idx += PACKED_BATCH_SIZE; + remained -= PACKED_BATCH_SIZE; + continue; + } } - } - tries++; - if (tries == 1) { - /* search [0, last_zmbuf_idx) */ - i = 0; - last = vq->last_zmbuf_idx; - goto again; - } + if (virtio_dev_tx_single_packed_zmbuf(dev, vq, mbuf_pool, + &pkts[pkt_idx])) + break; + pkt_idx++; + remained--; - return NULL; -} + } while (remained); -static __rte_always_inline bool -mbuf_is_consumed(struct rte_mbuf *m) -{ - while (m) { - if (rte_mbuf_refcnt_read(m) > 1) - return false; - m = m->next; - } + if (pkt_idx) + vhost_vring_call_packed(dev, vq); - return true; + return pkt_idx; } -static __rte_always_inline void -restore_mbuf(struct rte_mbuf *m) +static __rte_noinline uint16_t +virtio_dev_tx_packed(struct virtio_net *dev, + struct vhost_virtqueue *__rte_restrict vq, + struct rte_mempool *mbuf_pool, + struct rte_mbuf **__rte_restrict pkts, + uint32_t count) { - uint32_t mbuf_size, priv_size; + uint32_t pkt_idx = 0; + uint32_t remained = count; + + do { + rte_prefetch0(&vq->desc_packed[vq->last_avail_idx]); + + if (remained >= PACKED_BATCH_SIZE) { + if (!virtio_dev_tx_batch_packed(dev, vq, mbuf_pool, + &pkts[pkt_idx])) { + pkt_idx += PACKED_BATCH_SIZE; + remained -= PACKED_BATCH_SIZE; + continue; + } + } + + if (virtio_dev_tx_single_packed(dev, vq, mbuf_pool, + &pkts[pkt_idx])) + break; + pkt_idx++; + remained--; - while (m) { - priv_size = rte_pktmbuf_priv_size(m->pool); - mbuf_size = sizeof(struct rte_mbuf) + priv_size; - /* start of buffer is after mbuf structure and priv data */ + } while (remained); - m->buf_addr = (char *)m + mbuf_size; - m->buf_iova = rte_mempool_virt2iova(m) + mbuf_size; - m = m->next; + if (vq->shadow_used_idx) { + do_data_copy_dequeue(vq); + + vhost_flush_dequeue_shadow_packed(dev, vq); + vhost_vring_call_packed(dev, vq); } + + return pkt_idx; } uint16_t @@ -1407,25 +2773,22 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, struct virtio_net *dev; struct rte_mbuf *rarp_mbuf = NULL; struct vhost_virtqueue *vq; - uint32_t desc_indexes[MAX_PKT_BURST]; - uint32_t used_idx; - uint32_t i = 0; - uint16_t free_entries; - uint16_t avail_idx; + int16_t success = 1; dev = get_device(vid); if (!dev) return 0; if (unlikely(!(dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET))) { - RTE_LOG(ERR, VHOST_DATA, + VHOST_LOG_DATA(ERR, "(%d) %s: built-in vhost net backend is disabled.\n", dev->vid, __func__); return 0; } if (unlikely(!is_valid_virt_queue_idx(queue_id, 1, dev->nr_vring))) { - RTE_LOG(ERR, VHOST_DATA, "(%d) %s: invalid virtqueue idx %d.\n", + VHOST_LOG_DATA(ERR, + "(%d) %s: invalid virtqueue idx %d.\n", dev->vid, __func__, queue_id); return 0; } @@ -1435,43 +2798,20 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, if (unlikely(rte_spinlock_trylock(&vq->access_lock) == 0)) return 0; - if (unlikely(vq->enabled == 0)) + if (unlikely(vq->enabled == 0)) { + count = 0; goto out_access_unlock; - - vq->batch_copy_nb_elems = 0; + } 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)) + if (unlikely(vring_translate(dev, vq) < 0)) { + count = 0; goto out; - - if (unlikely(dev->dequeue_zero_copy)) { - struct zcopy_mbuf *zmbuf, *next; - int nr_updated = 0; - - for (zmbuf = TAILQ_FIRST(&vq->zmbuf_list); - zmbuf != NULL; zmbuf = next) { - next = TAILQ_NEXT(zmbuf, next); - - if (mbuf_is_consumed(zmbuf->mbuf)) { - used_idx = vq->last_used_idx++ & (vq->size - 1); - update_used_ring(dev, vq, used_idx, - zmbuf->desc_idx); - nr_updated += 1; - - TAILQ_REMOVE(&vq->zmbuf_list, zmbuf, next); - restore_mbuf(zmbuf->mbuf); - rte_pktmbuf_free(zmbuf->mbuf); - put_zmbuf(zmbuf); - vq->nr_zmbuf -= 1; - } } - update_used_idx(dev, vq, nr_updated); - } - /* * Construct a RARP broadcast packet, and inject it to the "pkts" * array, to looks like that guest actually send such packet. @@ -1480,148 +2820,36 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, * * broadcast_rarp shares a cacheline in the virtio_net structure * with some fields that are accessed during enqueue and - * rte_atomic16_cmpset() causes a write if using cmpxchg. This could - * result in false sharing between enqueue and dequeue. + * __atomic_compare_exchange_n causes a write if performed compare + * and exchange. This could result in false sharing between enqueue + * and dequeue. * * Prevent unnecessary false sharing by reading broadcast_rarp first - * and only performing cmpset if the read indicates it is likely to - * be set. + * and only performing compare and exchange if the read indicates it + * is likely to be set. */ - - if (unlikely(rte_atomic16_read(&dev->broadcast_rarp) && - rte_atomic16_cmpset((volatile uint16_t *) - &dev->broadcast_rarp.cnt, 1, 0))) { + if (unlikely(__atomic_load_n(&dev->broadcast_rarp, __ATOMIC_ACQUIRE) && + __atomic_compare_exchange_n(&dev->broadcast_rarp, + &success, 0, 0, __ATOMIC_RELEASE, __ATOMIC_RELAXED))) { rarp_mbuf = rte_net_make_rarp_packet(mbuf_pool, &dev->mac); if (rarp_mbuf == NULL) { - RTE_LOG(ERR, VHOST_DATA, - "Failed to make RARP packet.\n"); - return 0; + VHOST_LOG_DATA(ERR, "Failed to make RARP packet.\n"); + count = 0; + goto out; } count -= 1; } - free_entries = *((volatile uint16_t *)&vq->avail->idx) - - vq->last_avail_idx; - if (free_entries == 0) - goto out; - - VHOST_LOG_DEBUG(VHOST_DATA, "(%d) %s\n", dev->vid, __func__); - - /* Prefetch available and used ring */ - avail_idx = vq->last_avail_idx & (vq->size - 1); - used_idx = vq->last_used_idx & (vq->size - 1); - rte_prefetch0(&vq->avail->ring[avail_idx]); - rte_prefetch0(&vq->used->ring[used_idx]); - - count = RTE_MIN(count, MAX_PKT_BURST); - count = RTE_MIN(count, free_entries); - VHOST_LOG_DEBUG(VHOST_DATA, "(%d) about to dequeue %u buffers\n", - dev->vid, count); - - /* Retrieve all of the head indexes first to avoid caching issues. */ - for (i = 0; i < count; i++) { - avail_idx = (vq->last_avail_idx + i) & (vq->size - 1); - used_idx = (vq->last_used_idx + i) & (vq->size - 1); - desc_indexes[i] = vq->avail->ring[avail_idx]; - - if (likely(dev->dequeue_zero_copy == 0)) - update_used_ring(dev, vq, used_idx, desc_indexes[i]); - } - - /* Prefetch descriptor index. */ - rte_prefetch0(&vq->desc[desc_indexes[0]]); - for (i = 0; i < count; i++) { - struct vring_desc *desc, *idesc = NULL; - uint16_t sz, idx; - uint64_t dlen; - int err; - - if (likely(i + 1 < count)) - rte_prefetch0(&vq->desc[desc_indexes[i + 1]]); - - if (vq->desc[desc_indexes[i]].flags & VRING_DESC_F_INDIRECT) { - dlen = vq->desc[desc_indexes[i]].len; - desc = (struct vring_desc *)(uintptr_t) - vhost_iova_to_vva(dev, vq, - vq->desc[desc_indexes[i]].addr, - &dlen, - VHOST_ACCESS_RO); - if (unlikely(!desc)) - break; - - if (unlikely(dlen < vq->desc[desc_indexes[i]].len)) { - /* - * The indirect desc table is not contiguous - * in process VA space, we have to copy it. - */ - idesc = alloc_copy_ind_table(dev, vq, - &vq->desc[desc_indexes[i]]); - if (unlikely(!idesc)) - break; - - desc = idesc; - } - - rte_prefetch0(desc); - sz = vq->desc[desc_indexes[i]].len / sizeof(*desc); - idx = 0; - } else { - desc = vq->desc; - sz = vq->size; - idx = desc_indexes[i]; - } - - pkts[i] = rte_pktmbuf_alloc(mbuf_pool); - if (unlikely(pkts[i] == NULL)) { - RTE_LOG(ERR, VHOST_DATA, - "Failed to allocate memory for mbuf.\n"); - free_ind_table(idesc); - break; - } - - err = copy_desc_to_mbuf(dev, vq, desc, sz, pkts[i], idx, - mbuf_pool); - if (unlikely(err)) { - rte_pktmbuf_free(pkts[i]); - free_ind_table(idesc); - break; - } - - if (unlikely(dev->dequeue_zero_copy)) { - struct zcopy_mbuf *zmbuf; - - zmbuf = get_zmbuf(vq); - if (!zmbuf) { - rte_pktmbuf_free(pkts[i]); - free_ind_table(idesc); - break; - } - zmbuf->mbuf = pkts[i]; - zmbuf->desc_idx = desc_indexes[i]; - - /* - * Pin lock the mbuf; we will check later to see - * whether the mbuf is freed (when we are the last - * user) or not. If that's the case, we then could - * update the used ring safely. - */ - rte_mbuf_refcnt_update(pkts[i], 1); - - vq->nr_zmbuf += 1; - TAILQ_INSERT_TAIL(&vq->zmbuf_list, zmbuf, next); - } - - if (unlikely(!!idesc)) - free_ind_table(idesc); - } - vq->last_avail_idx += i; - - if (likely(dev->dequeue_zero_copy == 0)) { - do_data_copy_dequeue(vq); - vq->last_used_idx += i; - update_used_idx(dev, vq, i); - } + if (vq_is_packed(dev)) { + if (unlikely(dev->dequeue_zero_copy)) + count = virtio_dev_tx_packed_zmbuf(dev, vq, mbuf_pool, + pkts, count); + else + count = virtio_dev_tx_packed(dev, vq, mbuf_pool, pkts, + count); + } else + count = virtio_dev_tx_split(dev, vq, mbuf_pool, pkts, count); out: if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) @@ -1635,10 +2863,10 @@ out_access_unlock: * Inject it to the head of "pkts" array, so that switch's mac * learning table will get updated first. */ - memmove(&pkts[1], pkts, i * sizeof(struct rte_mbuf *)); + memmove(&pkts[1], pkts, count * sizeof(struct rte_mbuf *)); pkts[0] = rarp_mbuf; - i += 1; + count += 1; } - return i; + return count; }