X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=lib%2Flibrte_vhost%2Fvirtio_net.c;h=b75c93cf184e568144a3e4b7d4a877dd3628997b;hb=2ac67c32837a3e47eed1a6610edd80d6a2754488;hp=812e5d33622bcb0c78606ebbe28e704744706b86;hpb=f46f6551436a5a7c8509a4cbade1a7fda39905b3;p=dpdk.git diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c index 812e5d3362..b75c93cf18 100644 --- a/lib/librte_vhost/virtio_net.c +++ b/lib/librte_vhost/virtio_net.c @@ -39,66 +39,119 @@ #include #include #include -#include +#include #include #include #include #include +#include "iotlb.h" #include "vhost.h" #define MAX_PKT_BURST 32 -#define VHOST_LOG_PAGE 4096 -static inline void __attribute__((always_inline)) -vhost_log_page(uint8_t *log_base, uint64_t page) +#define MAX_BATCH_LEN 256 + +static bool +is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring) { - log_base[page / 8] |= 1 << (page % 8); + return (is_tx ^ (idx & 1)) == 0 && idx < nr_vring; } -static inline void __attribute__((always_inline)) -vhost_log_write(struct virtio_net *dev, uint64_t addr, uint64_t len) +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) { - uint64_t page; + rte_memcpy(&vq->used->ring[to], + &vq->shadow_used_ring[from], + size * sizeof(struct vring_used_elem)); + vhost_log_used_vring(dev, vq, + offsetof(struct vring_used, ring[to]), + size * sizeof(struct vring_used_elem)); +} - if (likely(((dev->features & (1ULL << VHOST_F_LOG_ALL)) == 0) || - !dev->log_base || !len)) - return; +static __rte_always_inline void +flush_shadow_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq) +{ + uint16_t used_idx = vq->last_used_idx & (vq->size - 1); - if (unlikely(dev->log_size <= ((addr + len - 1) / VHOST_LOG_PAGE / 8))) - return; + if (used_idx + vq->shadow_used_idx <= vq->size) { + do_flush_shadow_used_ring(dev, vq, used_idx, 0, + vq->shadow_used_idx); + } else { + uint16_t size; - /* To make sure guest memory updates are committed before logging */ - rte_smp_wmb(); + /* update used ring interval [used_idx, vq->size] */ + size = vq->size - used_idx; + do_flush_shadow_used_ring(dev, vq, used_idx, 0, size); - page = addr / VHOST_LOG_PAGE; - while (page * VHOST_LOG_PAGE < addr + len) { - vhost_log_page((uint8_t *)(uintptr_t)dev->log_base, page); - page += 1; + /* update the left half used ring interval [0, left_size] */ + do_flush_shadow_used_ring(dev, vq, 0, size, + vq->shadow_used_idx - size); } + vq->last_used_idx += vq->shadow_used_idx; + + rte_smp_wmb(); + + *(volatile uint16_t *)&vq->used->idx += vq->shadow_used_idx; + vhost_log_used_vring(dev, vq, offsetof(struct vring_used, idx), + sizeof(vq->used->idx)); } -static inline void __attribute__((always_inline)) -vhost_log_used_vring(struct virtio_net *dev, struct vhost_virtqueue *vq, - uint64_t offset, uint64_t len) +static __rte_always_inline void +update_shadow_used_ring(struct vhost_virtqueue *vq, + uint16_t desc_idx, uint16_t len) { - vhost_log_write(dev, vq->log_guest_addr + offset, len); + uint16_t i = vq->shadow_used_idx++; + + vq->shadow_used_ring[i].id = desc_idx; + vq->shadow_used_ring[i].len = len; } -static bool -is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t qp_nb) +static inline void +do_data_copy_enqueue(struct virtio_net *dev, struct vhost_virtqueue *vq) { - return (is_tx ^ (idx & 1)) == 0 && idx < qp_nb * VIRTIO_QNUM; + 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); + vhost_log_write(dev, elem[i].log_addr, elem[i].len); + PRINT_PACKET(dev, (uintptr_t)elem[i].dst, elem[i].len, 0); + } } +static inline void +do_data_copy_dequeue(struct vhost_virtqueue *vq) +{ + 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); +} + +/* avoid write operation when necessary, to lessen cache issues */ +#define ASSIGN_UNLESS_EQUAL(var, val) do { \ + if ((var) != (val)) \ + (var) = (val); \ +} while (0) + static void virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr) { - if (m_buf->ol_flags & PKT_TX_L4_MASK) { + uint64_t csum_l4 = m_buf->ol_flags & PKT_TX_L4_MASK; + + if (m_buf->ol_flags & PKT_TX_TCP_SEG) + csum_l4 |= PKT_TX_TCP_CKSUM; + + if (csum_l4) { net_hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; net_hdr->csum_start = m_buf->l2_len + m_buf->l3_len; - switch (m_buf->ol_flags & PKT_TX_L4_MASK) { + switch (csum_l4) { case PKT_TX_TCP_CKSUM: net_hdr->csum_offset = (offsetof(struct tcp_hdr, cksum)); @@ -112,6 +165,19 @@ virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr) cksum)); break; } + } else { + ASSIGN_UNLESS_EQUAL(net_hdr->csum_start, 0); + ASSIGN_UNLESS_EQUAL(net_hdr->csum_offset, 0); + ASSIGN_UNLESS_EQUAL(net_hdr->flags, 0); + } + + /* IP cksum verification cannot be bypassed, then calculate here */ + if (m_buf->ol_flags & PKT_TX_IP_CKSUM) { + struct ipv4_hdr *ipv4_hdr; + + ipv4_hdr = rte_pktmbuf_mtod_offset(m_buf, struct ipv4_hdr *, + m_buf->l2_len); + ipv4_hdr->hdr_checksum = rte_ipv4_cksum(ipv4_hdr); } if (m_buf->ol_flags & PKT_TX_TCP_SEG) { @@ -122,44 +188,45 @@ virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr) net_hdr->gso_size = m_buf->tso_segsz; net_hdr->hdr_len = m_buf->l2_len + m_buf->l3_len + m_buf->l4_len; + } else { + ASSIGN_UNLESS_EQUAL(net_hdr->gso_type, 0); + ASSIGN_UNLESS_EQUAL(net_hdr->gso_size, 0); + ASSIGN_UNLESS_EQUAL(net_hdr->hdr_len, 0); } } -static inline void -copy_virtio_net_hdr(struct virtio_net *dev, uint64_t desc_addr, - struct virtio_net_hdr_mrg_rxbuf hdr) -{ - if (dev->vhost_hlen == sizeof(struct virtio_net_hdr_mrg_rxbuf)) - *(struct virtio_net_hdr_mrg_rxbuf *)(uintptr_t)desc_addr = hdr; - else - *(struct virtio_net_hdr *)(uintptr_t)desc_addr = hdr.hdr; -} - -static inline int __attribute__((always_inline)) +static __rte_always_inline int copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq, - struct rte_mbuf *m, uint16_t desc_idx) + struct vring_desc *descs, struct rte_mbuf *m, + uint16_t desc_idx, uint32_t size) { uint32_t desc_avail, desc_offset; uint32_t mbuf_avail, mbuf_offset; uint32_t cpy_len; struct vring_desc *desc; uint64_t desc_addr; - struct virtio_net_hdr_mrg_rxbuf virtio_hdr = {{0, 0, 0, 0, 0, 0}, 0}; + /* 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; - desc = &vq->desc[desc_idx]; - desc_addr = gpa_to_vva(dev, desc->addr); + desc = &descs[desc_idx]; + desc_addr = vhost_iova_to_vva(dev, vq, desc->addr, + desc->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) - return -1; + if (unlikely(desc->len < dev->vhost_hlen) || !desc_addr) { + error = -1; + goto out; + } rte_prefetch0((void *)(uintptr_t)desc_addr); - virtio_enqueue_offload(m, &virtio_hdr.hdr); - copy_virtio_net_hdr(dev, desc_addr, virtio_hdr); + virtio_enqueue_offload(m, (struct virtio_net_hdr *)(uintptr_t)desc_addr); vhost_log_write(dev, desc->addr, dev->vhost_hlen); PRINT_PACKET(dev, (uintptr_t)desc_addr, dev->vhost_hlen, 0); @@ -181,27 +248,45 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq, if (desc_avail == 0) { if ((desc->flags & VRING_DESC_F_NEXT) == 0) { /* Room in vring buffer is not enough */ - return -1; + error = -1; + goto out; + } + if (unlikely(desc->next >= size || ++nr_desc > size)) { + error = -1; + goto out; } - if (unlikely(desc->next >= vq->size)) - return -1; - desc = &vq->desc[desc->next]; - desc_addr = gpa_to_vva(dev, desc->addr); - if (unlikely(!desc_addr)) - return -1; + desc = &descs[desc->next]; + desc_addr = vhost_iova_to_vva(dev, vq, desc->addr, + desc->len, + VHOST_ACCESS_RW); + if (unlikely(!desc_addr)) { + error = -1; + goto out; + } desc_offset = 0; desc_avail = desc->len; } cpy_len = RTE_MIN(desc_avail, mbuf_avail); - rte_memcpy((void *)((uintptr_t)(desc_addr + desc_offset)), - rte_pktmbuf_mtod_offset(m, void *, mbuf_offset), - cpy_len); - vhost_log_write(dev, desc->addr + desc_offset, cpy_len); - PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset), - cpy_len, 0); + 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->addr + 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->addr + desc_offset; + batch_copy[copy_nb].len = cpy_len; + copy_nb++; + } mbuf_avail -= cpy_len; mbuf_offset += cpy_len; @@ -209,28 +294,32 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq, desc_offset += cpy_len; } - return 0; +out: + vq->batch_copy_nb_elems = copy_nb; + + return error; } /** * 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 succesfully + * 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 inline uint32_t __attribute__((always_inline)) +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; + uint32_t i, sz; LOG_DEBUG(VHOST_DATA, "(%d) %s\n", dev->vid, __func__); - if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->virt_qp_nb))) { + 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; @@ -240,17 +329,29 @@ virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id, if (unlikely(vq->enabled == 0)) return 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)) { + count = 0; + goto out; + } + } + 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) - return 0; + goto out; 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; + /* 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++) { @@ -269,19 +370,36 @@ virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id, uint16_t desc_idx = desc_indexes[i]; int err; - err = copy_mbuf_to_desc(dev, vq, pkts[i], desc_idx); + if (vq->desc[desc_idx].flags & VRING_DESC_F_INDIRECT) { + descs = (struct vring_desc *)(uintptr_t) + vhost_iova_to_vva(dev, + vq, vq->desc[desc_idx].addr, + vq->desc[desc_idx].len, + VHOST_ACCESS_RO); + if (unlikely(!descs)) { + count = i; + break; + } + + desc_idx = 0; + sz = vq->desc[desc_idx].len / sizeof(*descs); + } else { + descs = vq->desc; + sz = vq->size; + } + + err = copy_mbuf_to_desc(dev, vq, descs, pkts[i], desc_idx, sz); if (unlikely(err)) { - used_idx = (start_idx + i) & (vq->size - 1); - vq->used->ring[used_idx].len = dev->vhost_hlen; - vhost_log_used_vring(dev, vq, - offsetof(struct vring_used, ring[used_idx]), - sizeof(vq->used->ring[used_idx])); + count = i; + break; } if (i + 1 < count) rte_prefetch0(&vq->desc[desc_indexes[i+1]]); } + do_data_copy_enqueue(dev, vq); + rte_smp_wmb(); *(volatile uint16_t *)&vq->used->idx += count; @@ -297,36 +415,55 @@ virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id, if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT) && (vq->callfd >= 0)) eventfd_write(vq->callfd, (eventfd_t)1); +out: + if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) + vhost_user_iotlb_rd_unlock(vq); + return count; } -static inline int -fill_vec_buf(struct vhost_virtqueue *vq, uint32_t avail_idx, - uint32_t *allocated, uint32_t *vec_idx, - struct buf_vector *buf_vec) +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) { uint16_t idx = vq->avail->ring[avail_idx & (vq->size - 1)]; uint32_t vec_id = *vec_idx; - uint32_t len = *allocated; + uint32_t len = 0; + struct vring_desc *descs = vq->desc; + + *desc_chain_head = idx; + + if (vq->desc[idx].flags & VRING_DESC_F_INDIRECT) { + descs = (struct vring_desc *)(uintptr_t) + vhost_iova_to_vva(dev, vq, vq->desc[idx].addr, + vq->desc[idx].len, + VHOST_ACCESS_RO); + if (unlikely(!descs)) + return -1; + + idx = 0; + } while (1) { if (unlikely(vec_id >= BUF_VECTOR_MAX || idx >= vq->size)) return -1; - len += vq->desc[idx].len; - buf_vec[vec_id].buf_addr = vq->desc[idx].addr; - buf_vec[vec_id].buf_len = vq->desc[idx].len; + 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 ((vq->desc[idx].flags & VRING_DESC_F_NEXT) == 0) + if ((descs[idx].flags & VRING_DESC_F_NEXT) == 0) break; - idx = vq->desc[idx].next; + idx = descs[idx].next; } - *allocated = len; - *vec_idx = vec_id; + *desc_chain_len = len; + *vec_idx = vec_id; return 0; } @@ -335,31 +472,34 @@ fill_vec_buf(struct vhost_virtqueue *vq, uint32_t avail_idx, * Returns -1 on fail, 0 on success */ static inline int -reserve_avail_buf_mergeable(struct vhost_virtqueue *vq, uint32_t size, - uint16_t *end, struct buf_vector *buf_vec) +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) { uint16_t cur_idx; - uint16_t avail_idx; - uint32_t allocated = 0; uint32_t vec_idx = 0; uint16_t tries = 0; - cur_idx = vq->last_used_idx; + uint16_t head_idx = 0; + uint16_t len = 0; - while (1) { - avail_idx = *((volatile uint16_t *)&vq->avail->idx); - if (unlikely(cur_idx == avail_idx)) + *num_buffers = 0; + cur_idx = vq->last_avail_idx; + + while (size > 0) { + if (unlikely(cur_idx == avail_head)) return -1; - if (unlikely(fill_vec_buf(vq, cur_idx, &allocated, - &vec_idx, buf_vec) < 0)) + if (unlikely(fill_vec_buf(dev, vq, cur_idx, &vec_idx, buf_vec, + &head_idx, &len) < 0)) return -1; + len = RTE_MIN(len, size); + update_shadow_used_ring(vq, head_idx, len); + size -= len; cur_idx++; tries++; - - if (allocated >= size) - break; + *num_buffers += 1; /* * if we tried all available ring items, and still @@ -370,78 +510,65 @@ reserve_avail_buf_mergeable(struct vhost_virtqueue *vq, uint32_t size, return -1; } - *end = cur_idx; return 0; } -static inline uint32_t __attribute__((always_inline)) +static __rte_always_inline int copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq, - uint16_t end_idx, struct rte_mbuf *m, - struct buf_vector *buf_vec) + struct rte_mbuf *m, struct buf_vector *buf_vec, + uint16_t num_buffers) { - struct virtio_net_hdr_mrg_rxbuf virtio_hdr = {{0, 0, 0, 0, 0, 0}, 0}; uint32_t vec_idx = 0; - uint16_t start_idx = vq->last_used_idx; - uint16_t cur_idx = start_idx; uint64_t desc_addr; - uint32_t desc_chain_head; - uint32_t desc_chain_len; uint32_t mbuf_offset, mbuf_avail; uint32_t desc_offset, desc_avail; uint32_t cpy_len; - uint16_t desc_idx, used_idx; - - if (unlikely(m == NULL)) - return 0; - - LOG_DEBUG(VHOST_DATA, "(%d) current index %d | end index %d\n", - dev->vid, cur_idx, end_idx); + uint64_t hdr_addr, hdr_phys_addr; + struct rte_mbuf *hdr_mbuf; + struct batch_copy_elem *batch_copy = vq->batch_copy_elems; + uint16_t copy_nb = vq->batch_copy_nb_elems; + int error = 0; + + if (unlikely(m == NULL)) { + error = -1; + goto out; + } - desc_addr = gpa_to_vva(dev, buf_vec[vec_idx].buf_addr); - if (buf_vec[vec_idx].buf_len < dev->vhost_hlen || !desc_addr) - return 0; + desc_addr = vhost_iova_to_vva(dev, vq, buf_vec[vec_idx].buf_addr, + buf_vec[vec_idx].buf_len, + VHOST_ACCESS_RW); + if (buf_vec[vec_idx].buf_len < dev->vhost_hlen || !desc_addr) { + error = -1; + goto out; + } - rte_prefetch0((void *)(uintptr_t)desc_addr); + hdr_mbuf = m; + hdr_addr = desc_addr; + hdr_phys_addr = buf_vec[vec_idx].buf_addr; + rte_prefetch0((void *)(uintptr_t)hdr_addr); - virtio_hdr.num_buffers = end_idx - start_idx; LOG_DEBUG(VHOST_DATA, "(%d) RX: num merge buffers %d\n", - dev->vid, virtio_hdr.num_buffers); - - virtio_enqueue_offload(m, &virtio_hdr.hdr); - copy_virtio_net_hdr(dev, desc_addr, virtio_hdr); - vhost_log_write(dev, buf_vec[vec_idx].buf_addr, dev->vhost_hlen); - PRINT_PACKET(dev, (uintptr_t)desc_addr, dev->vhost_hlen, 0); + dev->vid, num_buffers); desc_avail = buf_vec[vec_idx].buf_len - dev->vhost_hlen; desc_offset = dev->vhost_hlen; - desc_chain_head = buf_vec[vec_idx].desc_idx; - desc_chain_len = desc_offset; 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) { - desc_idx = buf_vec[vec_idx].desc_idx; vec_idx++; - - if (!(vq->desc[desc_idx].flags & VRING_DESC_F_NEXT)) { - /* Update used ring with desc information */ - used_idx = cur_idx++ & (vq->size - 1); - vq->used->ring[used_idx].id = desc_chain_head; - vq->used->ring[used_idx].len = desc_chain_len; - vhost_log_used_vring(dev, vq, - offsetof(struct vring_used, - ring[used_idx]), - sizeof(vq->used->ring[used_idx])); - desc_chain_head = buf_vec[vec_idx].desc_idx; - desc_chain_len = 0; + desc_addr = + vhost_iova_to_vva(dev, vq, + buf_vec[vec_idx].buf_addr, + buf_vec[vec_idx].buf_len, + VHOST_ACCESS_RW); + if (unlikely(!desc_addr)) { + error = -1; + goto out; } - desc_addr = gpa_to_vva(dev, buf_vec[vec_idx].buf_addr); - if (unlikely(!desc_addr)) - return 0; - /* Prefetch buffer address. */ rte_prefetch0((void *)(uintptr_t)desc_addr); desc_offset = 0; @@ -456,43 +583,68 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq, mbuf_avail = rte_pktmbuf_data_len(m); } + if (hdr_addr) { + struct virtio_net_hdr_mrg_rxbuf *hdr; + + hdr = (struct virtio_net_hdr_mrg_rxbuf *)(uintptr_t) + hdr_addr; + virtio_enqueue_offload(hdr_mbuf, &hdr->hdr); + ASSIGN_UNLESS_EQUAL(hdr->num_buffers, num_buffers); + + vhost_log_write(dev, hdr_phys_addr, dev->vhost_hlen); + PRINT_PACKET(dev, (uintptr_t)hdr_addr, + dev->vhost_hlen, 0); + + hdr_addr = 0; + } + cpy_len = RTE_MIN(desc_avail, mbuf_avail); - rte_memcpy((void *)((uintptr_t)(desc_addr + desc_offset)), - rte_pktmbuf_mtod_offset(m, void *, mbuf_offset), - cpy_len); - vhost_log_write(dev, buf_vec[vec_idx].buf_addr + desc_offset, - cpy_len); - PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset), - cpy_len, 0); + + 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, + buf_vec[vec_idx].buf_addr + 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 = + buf_vec[vec_idx].buf_addr + desc_offset; + batch_copy[copy_nb].len = cpy_len; + copy_nb++; + } mbuf_avail -= cpy_len; mbuf_offset += cpy_len; desc_avail -= cpy_len; desc_offset += cpy_len; - desc_chain_len += cpy_len; } - used_idx = cur_idx & (vq->size - 1); - vq->used->ring[used_idx].id = desc_chain_head; - vq->used->ring[used_idx].len = desc_chain_len; - vhost_log_used_vring(dev, vq, - offsetof(struct vring_used, ring[used_idx]), - sizeof(vq->used->ring[used_idx])); +out: + vq->batch_copy_nb_elems = copy_nb; - return end_idx - start_idx; + return error; } -static inline uint32_t __attribute__((always_inline)) +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) { struct vhost_virtqueue *vq; - uint32_t pkt_idx = 0, nr_used = 0; - uint16_t end; + uint32_t pkt_idx = 0; + uint16_t num_buffers; struct buf_vector buf_vec[BUF_VECTOR_MAX]; + uint16_t avail_head; LOG_DEBUG(VHOST_DATA, "(%d) %s\n", dev->vid, __func__); - if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->virt_qp_nb))) { + 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; @@ -502,32 +654,54 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id, if (unlikely(vq->enabled == 0)) return 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)) + goto out; + count = RTE_MIN((uint32_t)MAX_PKT_BURST, count); if (count == 0) - return 0; + goto out; + + vq->batch_copy_nb_elems = 0; + rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]); + + 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(reserve_avail_buf_mergeable(vq, pkt_len, - &end, buf_vec) < 0)) { + if (unlikely(reserve_avail_buf_mergeable(dev, vq, + pkt_len, buf_vec, &num_buffers, + avail_head) < 0)) { LOG_DEBUG(VHOST_DATA, "(%d) failed to get enough desc from vring\n", dev->vid); + vq->shadow_used_idx -= num_buffers; break; } - nr_used = copy_mbuf_to_desc_mergeable(dev, vq, end, - pkts[pkt_idx], buf_vec); - rte_smp_wmb(); + LOG_DEBUG(VHOST_DATA, "(%d) current index %d | end index %d\n", + dev->vid, vq->last_avail_idx, + vq->last_avail_idx + num_buffers); - *(volatile uint16_t *)&vq->used->idx += nr_used; - vhost_log_used_vring(dev, vq, offsetof(struct vring_used, idx), - sizeof(vq->used->idx)); - vq->last_used_idx += nr_used; + if (copy_mbuf_to_desc_mergeable(dev, vq, pkts[pkt_idx], + buf_vec, num_buffers) < 0) { + vq->shadow_used_idx -= num_buffers; + break; + } + + vq->last_avail_idx += num_buffers; } - if (likely(pkt_idx)) { + do_data_copy_enqueue(dev, vq); + + if (likely(vq->shadow_used_idx)) { + flush_shadow_used_ring(dev, vq); + /* flush used->idx update before we read avail->flags. */ rte_mb(); @@ -537,6 +711,10 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id, eventfd_write(vq->callfd, (eventfd_t)1); } +out: + if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) + vhost_user_iotlb_rd_unlock(vq); + return pkt_idx; } @@ -555,6 +733,20 @@ rte_vhost_enqueue_burst(int vid, uint16_t queue_id, return virtio_dev_rx(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) { @@ -580,14 +772,14 @@ parse_ethernet(struct rte_mbuf *m, uint16_t *l4_proto, void **l4_hdr) switch (ethertype) { case ETHER_TYPE_IPv4: - ipv4_hdr = (struct ipv4_hdr *)l3_hdr; + 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 = (struct ipv6_hdr *)l3_hdr; + ipv6_hdr = l3_hdr; *l4_proto = ipv6_hdr->proto; m->l3_len = sizeof(struct ipv6_hdr); *l4_hdr = (char *)l3_hdr + m->l3_len; @@ -596,17 +788,21 @@ parse_ethernet(struct rte_mbuf *m, uint16_t *l4_proto, void **l4_hdr) default: m->l3_len = 0; *l4_proto = 0; + *l4_hdr = NULL; break; } } -static inline void __attribute__((always_inline)) +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 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)) { @@ -629,11 +825,11 @@ vhost_dequeue_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *m) } } - if (hdr->gso_type != VIRTIO_NET_HDR_GSO_NONE) { + 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 = (struct tcp_hdr *)l4_hdr; + 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; @@ -685,15 +881,16 @@ make_rarp_packet(struct rte_mbuf *rarp_mbuf, const struct ether_addr *mac) return 0; } -static inline void __attribute__((always_inline)) +static __rte_always_inline void put_zmbuf(struct zcopy_mbuf *zmbuf) { zmbuf->in_use = 0; } -static inline int __attribute__((always_inline)) -copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs, - uint16_t max_desc, struct rte_mbuf *m, uint16_t desc_idx, +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) { struct vring_desc *desc; @@ -702,21 +899,33 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs, uint32_t mbuf_avail, mbuf_offset; uint32_t cpy_len; struct rte_mbuf *cur = m, *prev = m; - struct virtio_net_hdr *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; desc = &descs[desc_idx]; if (unlikely((desc->len < dev->vhost_hlen)) || - (desc->flags & VRING_DESC_F_INDIRECT)) - return -1; + (desc->flags & VRING_DESC_F_INDIRECT)) { + error = -1; + goto out; + } - desc_addr = gpa_to_vva(dev, desc->addr); - if (unlikely(!desc_addr)) - return -1; + desc_addr = vhost_iova_to_vva(dev, + vq, desc->addr, + desc->len, + VHOST_ACCESS_RO); + if (unlikely(!desc_addr)) { + error = -1; + goto out; + } - hdr = (struct virtio_net_hdr *)((uintptr_t)desc_addr); - rte_prefetch0(hdr); + if (virtio_net_with_host_offload(dev)) { + hdr = (struct virtio_net_hdr *)((uintptr_t)desc_addr); + rte_prefetch0(hdr); + } /* * A virtio driver normally uses at least 2 desc buffers @@ -726,25 +935,32 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs, 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)) - return -1; - - desc_addr = gpa_to_vva(dev, desc->addr); - if (unlikely(!desc_addr)) - return -1; + if (unlikely(desc->flags & VRING_DESC_F_INDIRECT)) { + error = -1; + goto out; + } - rte_prefetch0((void *)(uintptr_t)desc_addr); + desc_addr = vhost_iova_to_vva(dev, + vq, desc->addr, + desc->len, + VHOST_ACCESS_RO); + if (unlikely(!desc_addr)) { + error = -1; + goto out; + } desc_offset = 0; desc_avail = desc->len; nr_desc += 1; - - PRINT_PACKET(dev, (uintptr_t)desc_addr, desc->len, 0); } else { desc_avail = desc->len - dev->vhost_hlen; desc_offset = dev->vhost_hlen; } + rte_prefetch0((void *)(uintptr_t)(desc_addr + desc_offset)); + + PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset), desc_avail, 0); + mbuf_offset = 0; mbuf_avail = m->buf_len - RTE_PKTMBUF_HEADROOM; while (1) { @@ -770,10 +986,23 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs, */ mbuf_avail = cpy_len; } else { - rte_memcpy(rte_pktmbuf_mtod_offset(cur, void *, - mbuf_offset), - (void *)((uintptr_t)(desc_addr + desc_offset)), - cpy_len); + if (likely(cpy_len > MAX_BATCH_LEN || + copy_nb >= vq->size)) { + 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++; + } } mbuf_avail -= cpy_len; @@ -787,15 +1016,24 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs, break; if (unlikely(desc->next >= max_desc || - ++nr_desc > max_desc)) - return -1; + ++nr_desc > max_desc)) { + error = -1; + goto out; + } desc = &descs[desc->next]; - if (unlikely(desc->flags & VRING_DESC_F_INDIRECT)) - return -1; + if (unlikely(desc->flags & VRING_DESC_F_INDIRECT)) { + error = -1; + goto out; + } - desc_addr = gpa_to_vva(dev, desc->addr); - if (unlikely(!desc_addr)) - return -1; + desc_addr = vhost_iova_to_vva(dev, + vq, desc->addr, + desc->len, + VHOST_ACCESS_RO); + if (unlikely(!desc_addr)) { + error = -1; + goto out; + } rte_prefetch0((void *)(uintptr_t)desc_addr); @@ -814,8 +1052,11 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs, if (unlikely(cur == NULL)) { RTE_LOG(ERR, VHOST_DATA, "Failed to " "allocate memory for mbuf.\n"); - return -1; + error = -1; + goto out; } + if (unlikely(dev->dequeue_zero_copy)) + rte_mbuf_refcnt_update(cur, 1); prev->next = cur; prev->data_len = mbuf_offset; @@ -831,13 +1072,16 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs, prev->data_len = mbuf_offset; m->pkt_len += mbuf_offset; - if (hdr->flags != 0 || hdr->gso_type != VIRTIO_NET_HDR_GSO_NONE) + if (hdr) vhost_dequeue_offload(hdr, m); - return 0; +out: + vq->batch_copy_nb_elems = copy_nb; + + return error; } -static inline void __attribute__((always_inline)) +static __rte_always_inline void update_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq, uint32_t used_idx, uint32_t desc_idx) { @@ -848,7 +1092,7 @@ update_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq, sizeof(vq->used->ring[used_idx])); } -static inline void __attribute__((always_inline)) +static __rte_always_inline void update_used_idx(struct virtio_net *dev, struct vhost_virtqueue *vq, uint32_t count) { @@ -868,7 +1112,7 @@ update_used_idx(struct virtio_net *dev, struct vhost_virtqueue *vq, eventfd_write(vq->callfd, (eventfd_t)1); } -static inline struct zcopy_mbuf *__attribute__((always_inline)) +static __rte_always_inline struct zcopy_mbuf * get_zmbuf(struct vhost_virtqueue *vq) { uint16_t i; @@ -899,7 +1143,7 @@ again: return NULL; } -static inline bool __attribute__((always_inline)) +static __rte_always_inline bool mbuf_is_consumed(struct rte_mbuf *m) { while (m) { @@ -928,7 +1172,7 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, if (!dev) return 0; - if (unlikely(!is_valid_virt_queue_idx(queue_id, 1, dev->virt_qp_nb))) { + 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", dev->vid, __func__, queue_id); return 0; @@ -938,6 +1182,15 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, if (unlikely(vq->enabled == 0)) return 0; + 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)) + goto out; + if (unlikely(dev->dequeue_zero_copy)) { struct zcopy_mbuf *zmbuf, *next; int nr_updated = 0; @@ -967,9 +1220,21 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, * array, to looks like that guest actually send such packet. * * Check user_send_rarp() for more information. + * + * 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. + * + * Prevent unnecessary false sharing by reading broadcast_rarp first + * and only performing cmpset if the read indicates it is likely to + * be set. */ - if (unlikely(rte_atomic16_cmpset((volatile uint16_t *) - &dev->broadcast_rarp.cnt, 1, 0))) { + + if (unlikely(rte_atomic16_read(&dev->broadcast_rarp) && + rte_atomic16_cmpset((volatile uint16_t *) + &dev->broadcast_rarp.cnt, 1, 0))) { + rarp_mbuf = rte_pktmbuf_alloc(mbuf_pool); if (rarp_mbuf == NULL) { RTE_LOG(ERR, VHOST_DATA, @@ -1024,8 +1289,11 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, rte_prefetch0(&vq->desc[desc_indexes[i + 1]]); if (vq->desc[desc_indexes[i]].flags & VRING_DESC_F_INDIRECT) { - desc = (struct vring_desc *)(uintptr_t)gpa_to_vva(dev, - vq->desc[desc_indexes[i]].addr); + desc = (struct vring_desc *)(uintptr_t) + vhost_iova_to_vva(dev, vq, + vq->desc[desc_indexes[i]].addr, + sizeof(*desc), + VHOST_ACCESS_RO); if (unlikely(!desc)) break; @@ -1045,7 +1313,8 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, break; } - err = copy_desc_to_mbuf(dev, desc, sz, pkts[i], idx, mbuf_pool); + err = copy_desc_to_mbuf(dev, vq, desc, sz, pkts[i], idx, + mbuf_pool); if (unlikely(err)) { rte_pktmbuf_free(pkts[i]); break; @@ -1077,11 +1346,15 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, 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); } out: + if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) + vhost_user_iotlb_rd_unlock(vq); + if (unlikely(rarp_mbuf != NULL)) { /* * Inject it to the head of "pkts" array, so that switch's mac