X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=lib%2Flibrte_vhost%2Fvirtio_net.c;h=82d5d9e1746df083fad84cb51bc9c9cc2fe2e39c;hb=611994fc1bacc86decb2d8cd8ff6f7b0b1a883bc;hp=8a151afd946781dca87cf694085ced90f4b63783;hpb=a277c715987647ecd4ec80d18c5cc95c28d9dfd3;p=dpdk.git diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c index 8a151afd94..82d5d9e174 100644 --- a/lib/librte_vhost/virtio_net.c +++ b/lib/librte_vhost/virtio_net.c @@ -1,34 +1,5 @@ -/*- - * BSD LICENSE - * - * Copyright(c) 2010-2016 Intel Corporation. All rights reserved. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted provided that the following conditions - * are met: - * - * * Redistributions of source code must retain the above copyright - * notice, this list of conditions and the following disclaimer. - * * Redistributions in binary form must reproduce the above copyright - * notice, this list of conditions and the following disclaimer in - * the documentation and/or other materials provided with the - * distribution. - * * Neither the name of Intel Corporation nor the names of its - * contributors may be used to endorse or promote products derived - * from this software without specific prior written permission. - * - * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS - * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT - * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR - * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT - * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, - * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT - * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, - * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY - * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE - * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2010-2016 Intel Corporation */ #include @@ -39,66 +10,169 @@ #include #include #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 __rte_always_inline bool +rxvq_is_mergeable(struct virtio_net *dev) { - log_base[page / 8] |= 1 << (page % 8); + return dev->features & (1ULL << VIRTIO_NET_F_MRG_RXBUF); } -static inline void __attribute__((always_inline)) -vhost_log_write(struct virtio_net *dev, uint64_t addr, uint64_t len) +static bool +is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring) { - uint64_t page; + return (is_tx ^ (idx & 1)) == 0 && idx < nr_vring; +} - if (likely(((dev->features & (1ULL << VHOST_F_LOG_ALL)) == 0) || - !dev->log_base || !len)) - return; +static __rte_always_inline struct vring_desc * +alloc_copy_ind_table(struct virtio_net *dev, struct vhost_virtqueue *vq, + struct vring_desc *desc) +{ + struct vring_desc *idesc; + uint64_t src, dst; + uint64_t len, remain = desc->len; + uint64_t desc_addr = desc->addr; - if (unlikely(dev->log_size <= ((addr + len - 1) / VHOST_LOG_PAGE / 8))) - return; + idesc = rte_malloc(__func__, desc->len, 0); + if (unlikely(!idesc)) + return 0; - /* To make sure guest memory updates are committed before logging */ - rte_smp_wmb(); + dst = (uint64_t)(uintptr_t)idesc; - 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; + 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); + + remain -= len; + dst += len; + desc_addr += len; } + + return idesc; } -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 +free_ind_table(struct vring_desc *idesc) { - vhost_log_write(dev, vq->log_guest_addr + offset, len); + rte_free(idesc); } -static bool -is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t qp_nb) +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) { - return (is_tx ^ (idx & 1)) == 0 && idx < qp_nb * VIRTIO_QNUM; + rte_memcpy(&vq->used->ring[to], + &vq->shadow_used_ring[from], + size * sizeof(struct vring_used_elem)); + vhost_log_cache_used_vring(dev, vq, + offsetof(struct vring_used, ring[to]), + size * sizeof(struct vring_used_elem)); } -static void +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 (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; + + /* update used ring interval [used_idx, vq->size] */ + size = vq->size - used_idx; + do_flush_shadow_used_ring(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, + 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; + 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) +{ + uint16_t i = vq->shadow_used_idx++; + + vq->shadow_used_ring[i].id = desc_idx; + vq->shadow_used_ring[i].len = len; +} + +static inline void +do_data_copy_enqueue(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; + + for (i = 0; i < count; i++) { + rte_memcpy(elem[i].dst, elem[i].src, elem[i].len); + vhost_log_cache_write(dev, vq, 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 __rte_always_inline 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 +186,19 @@ 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) { @@ -122,211 +209,107 @@ 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 if (m_buf->ol_flags & PKT_TX_UDP_SEG) { + net_hdr->gso_type = VIRTIO_NET_HDR_GSO_UDP; + 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) +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, uint8_t perm) { - 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)) -copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq, - struct rte_mbuf *m, uint16_t desc_idx) -{ - 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}; - - desc = &vq->desc[desc_idx]; - desc_addr = gpa_to_vva(dev, desc->addr); - /* - * 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; - - rte_prefetch0((void *)(uintptr_t)desc_addr); - - virtio_enqueue_offload(m, &virtio_hdr.hdr); - copy_virtio_net_hdr(dev, desc_addr, virtio_hdr); - vhost_log_write(dev, desc->addr, dev->vhost_hlen); - PRINT_PACKET(dev, (uintptr_t)desc_addr, dev->vhost_hlen, 0); - - desc_offset = dev->vhost_hlen; - desc_avail = desc->len - dev->vhost_hlen; - - 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; - - mbuf_offset = 0; - mbuf_avail = rte_pktmbuf_data_len(m); - } - - /* 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 */ - return -1; - } - if (unlikely(desc->next >= vq->size)) - return -1; + uint16_t idx = vq->avail->ring[avail_idx & (vq->size - 1)]; + uint32_t vec_id = *vec_idx; + uint32_t len = 0; + uint64_t dlen, desc_avail, desc_iova; + struct vring_desc *descs = vq->desc; + struct vring_desc *idesc = NULL; + + *desc_chain_head = idx; + + 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)) + return -1; - desc = &vq->desc[desc->next]; - desc_addr = gpa_to_vva(dev, desc->addr); - if (unlikely(!desc_addr)) + 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; - desc_offset = 0; - desc_avail = desc->len; + descs = idesc; } - 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); - - mbuf_avail -= cpy_len; - mbuf_offset += cpy_len; - desc_avail -= cpy_len; - desc_offset += cpy_len; - } - - return 0; -} - -/** - * 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 - * 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)) -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]; - uint16_t used_idx; - uint32_t i; - - LOG_DEBUG(VHOST_DATA, "(%d) %s\n", dev->vid, __func__); - if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->virt_qp_nb))) { - RTE_LOG(ERR, VHOST_DATA, "(%d) %s: invalid virtqueue idx %d.\n", - dev->vid, __func__, queue_id); - return 0; - } - - vq = dev->virtqueue[queue_id]; - if (unlikely(vq->enabled == 0)) - return 0; - - 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; - - LOG_DEBUG(VHOST_DATA, "(%d) start_idx %d | end_idx %d\n", - dev->vid, start_idx, start_idx + count); - - /* 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 = 0; } - rte_prefetch0(&vq->desc[desc_indexes[0]]); - for (i = 0; i < count; i++) { - uint16_t desc_idx = desc_indexes[i]; - int err; - - err = copy_mbuf_to_desc(dev, vq, pkts[i], desc_idx); - 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])); + while (1) { + if (unlikely(idx >= vq->size)) { + free_ind_table(idesc); + return -1; } - if (i + 1 < count) - rte_prefetch0(&vq->desc[desc_indexes[i+1]]); - } - rte_smp_wmb(); + len += descs[idx].len; + desc_avail = descs[idx].len; + desc_iova = descs[idx].addr; - *(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)); - - /* flush used->idx update before we read avail->flags. */ - rte_mb(); + while (desc_avail) { + uint64_t desc_addr; + uint64_t desc_chunck_len = desc_avail; - /* Kick the guest if necessary. */ - if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT) - && (vq->callfd >= 0)) - eventfd_write(vq->callfd, (eventfd_t)1); - return count; -} + if (unlikely(vec_id >= BUF_VECTOR_MAX)) { + free_ind_table(idesc); + return -1; + } -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) -{ - uint16_t idx = vq->avail->ring[avail_idx & (vq->size - 1)]; - uint32_t vec_id = *vec_idx; - uint32_t len = *allocated; + desc_addr = vhost_iova_to_vva(dev, vq, + desc_iova, + &desc_chunck_len, + perm); + if (unlikely(!desc_addr)) { + free_ind_table(idesc); + return -1; + } - while (1) { - if (unlikely(vec_id >= BUF_VECTOR_MAX || idx >= vq->size)) - return -1; + 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; + buf_vec[vec_id].desc_idx = idx; - 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; - buf_vec[vec_id].desc_idx = idx; - vec_id++; + desc_avail -= desc_chunck_len; + desc_iova += desc_chunck_len; + 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; + + if (unlikely(!!idesc)) + free_ind_table(idesc); return 0; } @@ -335,111 +318,132 @@ 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(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 avail_idx; - uint32_t allocated = 0; uint32_t vec_idx = 0; - uint16_t tries = 0; + uint16_t max_tries, 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; + + if (rxvq_is_mergeable(dev)) + max_tries = vq->size; + else + max_tries = 1; + + 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, + VHOST_ACCESS_RW) < 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 * can't get enough buf, it means something abnormal * happened. */ - if (unlikely(tries >= vq->size)) + if (unlikely(tries > max_tries)) return -1; } - *end = cur_idx; + *nr_vec = vec_idx; + return 0; } -static inline uint32_t __attribute__((always_inline)) -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) +static __rte_always_inline int +copy_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 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 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; - 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); - - 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; + 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; + + if (unlikely(m == NULL)) { + error = -1; + goto out; + } - rte_prefetch0((void *)(uintptr_t)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; - 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); + if (nr_vec > 1) + rte_prefetch0((void *)(uintptr_t)buf_vec[1].buf_addr); - 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); + if (unlikely(buf_len < dev->vhost_hlen && nr_vec <= 1)) { + error = -1; + goto out; + } - desc_avail = buf_vec[vec_idx].buf_len - dev->vhost_hlen; - desc_offset = dev->vhost_hlen; + hdr_mbuf = m; + hdr_addr = buf_addr; + if (unlikely(buf_len < dev->vhost_hlen)) + hdr = &tmp_hdr; + else + hdr = (struct virtio_net_hdr_mrg_rxbuf *)(uintptr_t)hdr_addr; + + VHOST_LOG_DEBUG(VHOST_DATA, "(%d) RX: num merge buffers %d\n", + dev->vid, num_buffers); + + if (unlikely(buf_len < dev->vhost_hlen)) { + buf_offset = dev->vhost_hlen - buf_len; + vec_idx++; + buf_addr = buf_vec[vec_idx].buf_addr; + buf_iova = buf_vec[vec_idx].buf_iova; + buf_len = buf_vec[vec_idx].buf_len; + buf_avail = buf_len - buf_offset; + } else { + buf_offset = dev->vhost_hlen; + buf_avail = buf_len - dev->vhost_hlen; + } mbuf_avail = rte_pktmbuf_data_len(m); mbuf_offset = 0; while (mbuf_avail != 0 || m->next != NULL) { - /* done with current desc buf, get the next one */ - if (desc_avail == 0) { - desc_idx = buf_vec[vec_idx].desc_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_idx; - vq->used->ring[used_idx].len = desc_offset; - vhost_log_used_vring(dev, vq, - offsetof(struct vring_used, - ring[used_idx]), - sizeof(vq->used->ring[used_idx])); + /* done with current buf, get the next one */ + if (buf_avail == 0) { + vec_idx++; + if (unlikely(vec_idx >= nr_vec)) { + error = -1; + goto out; } - vec_idx++; - 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; - desc_avail = buf_vec[vec_idx].buf_len; + buf_addr = buf_vec[vec_idx].buf_addr; + buf_iova = buf_vec[vec_idx].buf_iova; + buf_len = buf_vec[vec_idx].buf_len; + + /* Prefetch next buffer address. */ + if (vec_idx + 1 < nr_vec) + rte_prefetch0((void *)(uintptr_t) + buf_vec[vec_idx + 1].buf_addr); + buf_offset = 0; + buf_avail = buf_len; } /* done with current mbuf, get the next one */ @@ -450,86 +454,165 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq, mbuf_avail = rte_pktmbuf_data_len(m); } - 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 (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)) { + uint64_t len; + uint64_t remain = dev->vhost_hlen; + uint64_t src = (uint64_t)(uintptr_t)hdr, dst; + uint64_t iova = buf_vec[0].buf_iova; + uint16_t hdr_vec_idx = 0; + + while (remain) { + len = remain; + dst = buf_vec[hdr_vec_idx].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(dev, vq, + iova, len); + + remain -= len; + iova += len; + src += len; + hdr_vec_idx++; + } + } else { + PRINT_PACKET(dev, (uintptr_t)hdr_addr, + dev->vhost_hlen, 0); + vhost_log_cache_write(dev, vq, + buf_vec[0].buf_iova, + dev->vhost_hlen); + } + + hdr_addr = 0; + } + + cpy_len = RTE_MIN(buf_len, mbuf_avail); + + 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_cache_write(dev, vq, buf_iova + buf_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; - desc_avail -= cpy_len; - desc_offset += cpy_len; + buf_avail -= cpy_len; + buf_offset += cpy_len; } - used_idx = cur_idx & (vq->size - 1); - vq->used->ring[used_idx].id = buf_vec[vec_idx].desc_idx; - vq->used->ring[used_idx].len = desc_offset; - vhost_log_used_vring(dev, vq, - offsetof(struct vring_used, ring[used_idx]), - sizeof(vq->used->ring[used_idx])); +out: - return end_idx - start_idx; + return error; } -static inline uint32_t __attribute__((always_inline)) -virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id, +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 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))) { + 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)) - return 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) - 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; + uint16_t nr_vec = 0; - if (unlikely(reserve_avail_buf_mergeable(vq, pkt_len, - &end, buf_vec) < 0)) { - LOG_DEBUG(VHOST_DATA, + if (unlikely(reserve_avail_buf(dev, vq, + pkt_len, buf_vec, &num_buffers, + avail_head, &nr_vec) < 0)) { + VHOST_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(); + rte_prefetch0((void *)(uintptr_t)buf_vec[0].buf_addr); - *(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; + VHOST_LOG_DEBUG(VHOST_DATA, "(%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(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; } - if (likely(pkt_idx)) { - /* flush used->idx update before we read avail->flags. */ - rte_mb(); + do_data_copy_enqueue(dev, vq); - /* Kick the guest if necessary. */ - if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT) - && (vq->callfd >= 0)) - eventfd_write(vq->callfd, (eventfd_t)1); + if (likely(vq->shadow_used_idx)) { + flush_shadow_used_ring(dev, vq); + vhost_vring_call(dev, vq); } +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 pkt_idx; } @@ -542,10 +625,28 @@ rte_vhost_enqueue_burst(int vid, uint16_t queue_id, if (!dev) return 0; - 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); + 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; + } + + 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 @@ -573,14 +674,14 @@ 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; @@ -589,17 +690,21 @@ 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)) { @@ -622,15 +727,20 @@ 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; 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); + break; default: RTE_LOG(WARNING, VHOST_DATA, "unsupported gso type %u.\n", hdr->gso_type); @@ -639,129 +749,178 @@ vhost_dequeue_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *m) } } -#define RARP_PKT_SIZE 64 - -static int -make_rarp_packet(struct rte_mbuf *rarp_mbuf, const struct ether_addr *mac) +static __rte_always_inline void +put_zmbuf(struct zcopy_mbuf *zmbuf) { - struct ether_hdr *eth_hdr; - struct arp_hdr *rarp; - - if (rarp_mbuf->buf_len < 64) { - RTE_LOG(WARNING, VHOST_DATA, - "failed to make RARP; mbuf size too small %u (< %d)\n", - rarp_mbuf->buf_len, RARP_PKT_SIZE); - return -1; - } - - /* Ethernet header. */ - eth_hdr = rte_pktmbuf_mtod_offset(rarp_mbuf, struct ether_hdr *, 0); - memset(eth_hdr->d_addr.addr_bytes, 0xff, ETHER_ADDR_LEN); - ether_addr_copy(mac, ð_hdr->s_addr); - eth_hdr->ether_type = htons(ETHER_TYPE_RARP); - - /* RARP header. */ - rarp = (struct arp_hdr *)(eth_hdr + 1); - rarp->arp_hrd = htons(ARP_HRD_ETHER); - rarp->arp_pro = htons(ETHER_TYPE_IPv4); - rarp->arp_hln = ETHER_ADDR_LEN; - rarp->arp_pln = 4; - rarp->arp_op = htons(ARP_OP_REVREQUEST); - - ether_addr_copy(mac, &rarp->arp_data.arp_sha); - ether_addr_copy(mac, &rarp->arp_data.arp_tha); - memset(&rarp->arp_data.arp_sip, 0x00, 4); - memset(&rarp->arp_data.arp_tip, 0x00, 4); - - rarp_mbuf->pkt_len = rarp_mbuf->data_len = RARP_PKT_SIZE; - - return 0; + zmbuf->in_use = 0; } -static inline int __attribute__((always_inline)) +static __rte_always_inline int copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq, - struct rte_mbuf *m, uint16_t desc_idx, - struct rte_mempool *mbuf_pool) + struct buf_vector *buf_vec, uint16_t nr_vec, + struct rte_mbuf *m, struct rte_mempool *mbuf_pool) { - struct vring_desc *desc; - uint64_t desc_addr; - uint32_t desc_avail, desc_offset; + 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 *hdr; + 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; + uint16_t vec_idx = 0; + struct batch_copy_elem *batch_copy = vq->batch_copy_elems; + int error = 0; - desc = &vq->desc[desc_idx]; - if (unlikely(desc->len < dev->vhost_hlen)) - return -1; + buf_addr = buf_vec[vec_idx].buf_addr; + buf_iova = buf_vec[vec_idx].buf_iova; + buf_len = buf_vec[vec_idx].buf_len; - desc_addr = gpa_to_vva(dev, desc->addr); - if (unlikely(!desc_addr)) - return -1; + if (unlikely(buf_len < dev->vhost_hlen && nr_vec <= 1)) { + error = -1; + goto out; + } - hdr = (struct virtio_net_hdr *)((uintptr_t)desc_addr); - rte_prefetch0(hdr); + if (likely(nr_vec > 1)) + rte_prefetch0((void *)(uintptr_t)buf_vec[1].buf_addr); + + if (virtio_net_with_host_offload(dev)) { + if (unlikely(buf_len < sizeof(struct virtio_net_hdr))) { + uint64_t len; + uint64_t remain = sizeof(struct virtio_net_hdr); + uint64_t src; + uint64_t dst = (uint64_t)(uintptr_t)&tmp_hdr; + uint16_t hdr_vec_idx = 0; + + /* + * No luck, the virtio-net header doesn't fit + * in a contiguous virtual area. + */ + while (remain) { + len = remain; + src = buf_vec[hdr_vec_idx].buf_addr; + rte_memcpy((void *)(uintptr_t)dst, + (void *)(uintptr_t)src, len); + + remain -= len; + dst += len; + hdr_vec_idx++; + } + + hdr = &tmp_hdr; + } else { + hdr = (struct virtio_net_hdr *)((uintptr_t)buf_addr); + rte_prefetch0(hdr); + } + } /* * 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 = &vq->desc[desc->next]; - - desc_addr = gpa_to_vva(dev, desc->addr); - if (unlikely(!desc_addr)) - return -1; - - rte_prefetch0((void *)(uintptr_t)desc_addr); - - desc_offset = 0; - desc_avail = desc->len; - nr_desc += 1; - - PRINT_PACKET(dev, (uintptr_t)desc_addr, desc->len, 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 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 { - desc_avail = desc->len - dev->vhost_hlen; - desc_offset = dev->vhost_hlen; + buf_offset = dev->vhost_hlen; + buf_avail = buf_vec[vec_idx].buf_len - dev->vhost_hlen; } + rte_prefetch0((void *)(uintptr_t) + (buf_addr + buf_offset)); + + 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) { - cpy_len = RTE_MIN(desc_avail, mbuf_avail); - rte_memcpy(rte_pktmbuf_mtod_offset(cur, void *, mbuf_offset), - (void *)((uintptr_t)(desc_addr + desc_offset)), - cpy_len); + 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; - desc_avail -= cpy_len; - desc_offset += cpy_len; + buf_avail -= cpy_len; + buf_offset += cpy_len; - /* This desc reaches to its end, get the next one */ - if (desc_avail == 0) { - if ((desc->flags & VRING_DESC_F_NEXT) == 0) + /* This buf reaches to its end, get the next one */ + if (buf_avail == 0) { + if (++vec_idx >= nr_vec) break; - if (unlikely(desc->next >= vq->size || - ++nr_desc > vq->size)) - return -1; - desc = &vq->desc[desc->next]; - - desc_addr = gpa_to_vva(dev, desc->addr); - if (unlikely(!desc_addr)) - return -1; + buf_addr = buf_vec[vec_idx].buf_addr; + buf_iova = buf_vec[vec_idx].buf_iova; + buf_len = buf_vec[vec_idx].buf_len; - rte_prefetch0((void *)(uintptr_t)desc_addr); + /* + * Prefecth desc n + 1 buffer while + * desc n buffer is processed. + */ + if (vec_idx + 1 < nr_vec) + rte_prefetch0((void *)(uintptr_t) + buf_vec[vec_idx + 1].buf_addr); - desc_offset = 0; - desc_avail = desc->len; + buf_offset = 0; + buf_avail = buf_len; - PRINT_PACKET(dev, (uintptr_t)desc_addr, desc->len, 0); + PRINT_PACKET(dev, (uintptr_t)buf_addr, + (uint32_t)buf_avail, 0); } /* @@ -773,8 +932,11 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq, 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; @@ -790,10 +952,71 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq, 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: + + 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]; + } + } + + tries++; + if (tries == 1) { + /* search [0, last_zmbuf_idx) */ + i = 0; + last = vq->last_zmbuf_idx; + goto again; + } + + return NULL; +} + +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; + } + + return true; +} + +static __rte_always_inline void +restore_mbuf(struct rte_mbuf *m) +{ + uint32_t mbuf_size, priv_size; + + 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 */ + + m->buf_addr = (char *)m + mbuf_size; + m->buf_iova = rte_mempool_virt2iova(m) + mbuf_size; + m = m->next; + } } uint16_t @@ -803,85 +1026,129 @@ 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; dev = get_device(vid); if (!dev) return 0; - if (unlikely(!is_valid_virt_queue_idx(queue_id, 1, dev->virt_qp_nb))) { + 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; + } + + 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; } vq = dev->virtqueue[queue_id]; - if (unlikely(vq->enabled == 0)) + + if (unlikely(rte_spinlock_trylock(&vq->access_lock) == 0)) return 0; + if (unlikely(vq->enabled == 0)) + goto out_access_unlock; + + vq->batch_copy_nb_elems = 0; + vq->shadow_used_idx = 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; + + 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(vq, zmbuf->desc_idx, 0); + 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; + } + } + + flush_shadow_used_ring(dev, vq); + vhost_vring_call(dev, vq); + vq->shadow_used_idx = 0; + } + + rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]); + /* * Construct a RARP broadcast packet, and inject it to the "pkts" * 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))) { - rarp_mbuf = rte_pktmbuf_alloc(mbuf_pool); + + if (unlikely(rte_atomic16_read(&dev->broadcast_rarp) && + rte_atomic16_cmpset((volatile uint16_t *) + &dev->broadcast_rarp.cnt, 1, 0))) { + + rarp_mbuf = rte_net_make_rarp_packet(mbuf_pool, &dev->mac); if (rarp_mbuf == NULL) { RTE_LOG(ERR, VHOST_DATA, - "Failed to allocate memory for mbuf.\n"); + "Failed to make RARP packet.\n"); return 0; } - - if (make_rarp_packet(rarp_mbuf, &dev->mac)) { - rte_pktmbuf_free(rarp_mbuf); - rarp_mbuf = NULL; - } else { - count -= 1; - } + count -= 1; } - avail_idx = *((volatile uint16_t *)&vq->avail->idx); - free_entries = avail_idx - vq->last_used_idx; + free_entries = *((volatile uint16_t *)&vq->avail->idx) - + vq->last_avail_idx; if (free_entries == 0) goto out; - LOG_DEBUG(VHOST_DATA, "(%d) %s\n", dev->vid, __func__); - - /* Prefetch available ring to retrieve head indexes. */ - used_idx = vq->last_used_idx & (vq->size - 1); - rte_prefetch0(&vq->avail->ring[used_idx]); - rte_prefetch0(&vq->used->ring[used_idx]); + VHOST_LOG_DEBUG(VHOST_DATA, "(%d) %s\n", dev->vid, __func__); count = RTE_MIN(count, MAX_PKT_BURST); count = RTE_MIN(count, free_entries); - LOG_DEBUG(VHOST_DATA, "(%d) about to dequeue %u buffers\n", + 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++) { - used_idx = (vq->last_used_idx + i) & (vq->size - 1); - desc_indexes[i] = vq->avail->ring[used_idx]; + struct buf_vector buf_vec[BUF_VECTOR_MAX]; + uint16_t head_idx, dummy_len; + uint32_t nr_vec = 0; + int err; - vq->used->ring[used_idx].id = desc_indexes[i]; - 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])); - } + if (unlikely(fill_vec_buf(dev, vq, + vq->last_avail_idx + i, + &nr_vec, buf_vec, + &head_idx, &dummy_len, + VHOST_ACCESS_RO) < 0)) + break; - /* Prefetch descriptor index. */ - rte_prefetch0(&vq->desc[desc_indexes[0]]); - for (i = 0; i < count; i++) { - int err; + if (likely(dev->dequeue_zero_copy == 0)) + update_shadow_used_ring(vq, head_idx, 0); - if (likely(i + 1 < count)) - rte_prefetch0(&vq->desc[desc_indexes[i + 1]]); + rte_prefetch0((void *)(uintptr_t)buf_vec[0].buf_addr); pkts[i] = rte_pktmbuf_alloc(mbuf_pool); if (unlikely(pkts[i] == NULL)) { @@ -889,27 +1156,54 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, "Failed to allocate memory for mbuf.\n"); break; } - err = copy_desc_to_mbuf(dev, vq, pkts[i], desc_indexes[i], - mbuf_pool); + + err = copy_desc_to_mbuf(dev, vq, buf_vec, nr_vec, pkts[i], + mbuf_pool); if (unlikely(err)) { rte_pktmbuf_free(pkts[i]); break; } - } - rte_smp_wmb(); - rte_smp_rmb(); - vq->used->idx += i; - vq->last_used_idx += i; - vhost_log_used_vring(dev, vq, offsetof(struct vring_used, idx), - sizeof(vq->used->idx)); + if (unlikely(dev->dequeue_zero_copy)) { + struct zcopy_mbuf *zmbuf; - /* Kick guest if required. */ - if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT) - && (vq->callfd >= 0)) - eventfd_write(vq->callfd, (eventfd_t)1); + zmbuf = get_zmbuf(vq); + if (!zmbuf) { + rte_pktmbuf_free(pkts[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; + + if (likely(dev->dequeue_zero_copy == 0)) { + do_data_copy_dequeue(vq); + if (unlikely(i < count)) + vq->shadow_used_idx = i; + flush_shadow_used_ring(dev, vq); + vhost_vring_call(dev, vq); + } 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 (unlikely(rarp_mbuf != NULL)) { /* * Inject it to the head of "pkts" array, so that switch's mac