From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id DDD9AA034C; Sun, 14 Aug 2022 15:20:21 +0200 (CEST) Received: from [217.70.189.124] (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id C703541611; Sun, 14 Aug 2022 15:20:18 +0200 (CEST) Received: from mga04.intel.com (mga04.intel.com [192.55.52.120]) by mails.dpdk.org (Postfix) with ESMTP id 0C5CA40A7E for ; Sun, 14 Aug 2022 15:20:16 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1660483217; x=1692019217; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=+WXXQzy/cwHD1uzdhqjSUfj8OnAk5qcFoISfgNyi4x0=; b=i7Zactp2srV0jpPxyBqHfo3jwpxhQljQpXthmk7FBOJKRDPov30pdSYi AQDv4peBly5MEJ/XfsTpn1Qpv4JQ0TtJQMsTJoSsnrIou50yUMzFklmSI rzIthEcb444eYWPmeez5gTTJ6ZzEUMdDbvH2m41CeGPdsN55nhEATyKVU suQYG6ji6bWAgEjziovxF/C4o3Q0if/MC8NgwRr15aQ3lMTDjmdUWYAHM /zX+vA/1Xk6gaboU9zfXtGZ4X9ByhlPDfkyU5AWheK819RMLtYFxPBnbi w0pW1piF51wAFc7i3NMw8jY7J0RaUy6/c7jUQ2dAs8ls+p34097ZuYHIn Q==; X-IronPort-AV: E=McAfee;i="6400,9594,10439"; a="290577984" X-IronPort-AV: E=Sophos;i="5.93,236,1654585200"; d="scan'208";a="290577984" Received: from fmsmga008.fm.intel.com ([10.253.24.58]) by fmsmga104.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 14 Aug 2022 06:20:16 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.93,236,1654585200"; d="scan'208";a="666372262" Received: from dpdk_jiangcheng.sh.intel.com ([10.67.118.237]) by fmsmga008.fm.intel.com with ESMTP; 14 Aug 2022 06:20:13 -0700 From: Cheng Jiang To: maxime.coquelin@redhat.com, chenbo.xia@intel.com Cc: dev@dpdk.org, jiayu.hu@intel.com, xuan.ding@intel.com, wenwux.ma@intel.com, yuanx.wang@intel.com, yvonnex.yang@intel.com, Cheng Jiang Subject: [RFC 2/2] vhost: add egress API for port mirroring datapath Date: Sun, 14 Aug 2022 12:49:20 +0000 Message-Id: <20220814124920.14338-3-cheng1.jiang@intel.com> X-Mailer: git-send-email 2.35.1 In-Reply-To: <20220814124920.14338-1-cheng1.jiang@intel.com> References: <20220814124920.14338-1-cheng1.jiang@intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org This patch implements such an egress function on the Vhost lib. When data is received from a front-end, it will also send the data to its mirrorfront-end. Signed-off-by: Cheng Jiang Signed-off-by: Wenwu Ma --- lib/vhost/rte_vhost_async.h | 12 +- lib/vhost/version.map | 3 + lib/vhost/virtio_net.c | 729 ++++++++++++++++++++++++++++++++++++ 3 files changed, 742 insertions(+), 2 deletions(-) diff --git a/lib/vhost/rte_vhost_async.h b/lib/vhost/rte_vhost_async.h index b199af078c..b0772c2db7 100644 --- a/lib/vhost/rte_vhost_async.h +++ b/lib/vhost/rte_vhost_async.h @@ -85,11 +85,12 @@ struct rte_vhost_async_channel_ops { }; /** - * inflight async packet information + * in-flight async packet information */ struct async_inflight_info { struct rte_mbuf *mbuf; - uint16_t descs; /* num of descs inflight */ + struct virtio_net_hdr nethdr; + uint16_t descs; /* num of descs in-flight */ uint16_t nr_buffers; /* num of buffers inflight for packed ring */ }; @@ -268,4 +269,11 @@ __rte_experimental uint16_t rte_vhost_clear_queue_thread_unsafe(int vid, uint16_t queue_id, struct rte_mbuf **pkts, uint16_t count); +__rte_experimental +uint16_t +rte_vhost_async_try_egress_burst(int vid, uint16_t queue_id, + int mirr_vid, uint16_t mirr_queue_id, + struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count, + int *nr_inflight); + #endif /* _RTE_VHOST_ASYNC_H_ */ diff --git a/lib/vhost/version.map b/lib/vhost/version.map index 4c35fa4555..2529943a91 100644 --- a/lib/vhost/version.map +++ b/lib/vhost/version.map @@ -87,4 +87,7 @@ EXPERIMENTAL { rte_vhost_async_channel_register_thread_unsafe; rte_vhost_async_channel_unregister_thread_unsafe; rte_vhost_clear_queue_thread_unsafe; + + # added in 21.11 + rte_vhost_async_try_egress_burst; }; diff --git a/lib/vhost/virtio_net.c b/lib/vhost/virtio_net.c index c9f0bb22e5..92074377b2 100644 --- a/lib/vhost/virtio_net.c +++ b/lib/vhost/virtio_net.c @@ -3827,3 +3827,732 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, return count; } + +static __rte_always_inline uint16_t +async_poll_egress_completed_split(struct virtio_net *dev, + struct vhost_virtqueue *vq, uint16_t queue_id, + struct virtio_net *mirr_dev, struct vhost_virtqueue *mirr_vq, + struct rte_mbuf **pkts, uint16_t count, bool legacy_ol_flags) +{ + uint16_t n_pkts_cpl = 0, n_pkts_put = 0; + uint16_t start_idx, pkt_idx, from; + struct async_inflight_info *pkts_info; + uint16_t mirror_n_pkts_cpl = 0, n_descs = 0; + + pkt_idx = vq->async_pkts_idx & (vq->size - 1); + pkts_info = vq->async_pkts_info; + start_idx = virtio_dev_rx_async_get_info_idx(pkt_idx, vq->size, + vq->async_pkts_inflight_n); + + if (count > vq->async_last_pkts_n) { + int ret; + + ret = vq->async_ops.check_completed_copies(dev->vid, queue_id, + 0, count - vq->async_last_pkts_n); + if (unlikely(ret < 0)) { + VHOST_LOG_DATA(ERR, "(%d) async channel poll error\n", dev->vid); + ret = 0; + } + n_pkts_cpl = ret; + } + + n_pkts_cpl += vq->async_last_pkts_n; + mirror_n_pkts_cpl += mirr_vq->async_last_pkts_n; + if (unlikely(n_pkts_cpl == 0)) { + mirr_vq->async_last_pkts_n = mirror_n_pkts_cpl; + return 0; + } + + n_pkts_put = RTE_MIN(count, n_pkts_cpl); + + for (pkt_idx = 0; pkt_idx < n_pkts_put; pkt_idx++) { + from = (start_idx + pkt_idx) & (vq->size - 1); + pkts[pkt_idx] = pkts_info[from].mbuf; + n_descs += pkts_info[from].descs; + + if (virtio_net_with_host_offload(dev)) + vhost_dequeue_offload(&pkts_info[from].nethdr, + pkts[pkt_idx], legacy_ol_flags); + } + + /* write back completed descs to used ring and update used idx */ + write_back_completed_descs_split(vq, n_pkts_put); + __atomic_add_fetch(&vq->used->idx, n_pkts_put, __ATOMIC_RELEASE); + vhost_vring_call_split(dev, vq); + + vq->async_last_pkts_n = n_pkts_cpl - n_pkts_put; + vq->async_pkts_inflight_n -= n_pkts_put; + + if (likely(mirr_vq->enabled && mirr_vq->access_ok)) { + write_back_completed_descs_split(mirr_vq, n_descs); + + __atomic_add_fetch(&mirr_vq->used->idx, n_descs, + __ATOMIC_RELEASE); + vhost_vring_call_split(mirr_dev, mirr_vq); + } else { + mirr_vq->last_async_desc_idx_split += n_descs; + } + + return n_pkts_put; +} + +static __rte_always_inline void +egress_async_fill_desc(struct rte_vhost_async_desc *desc, + struct rte_vhost_iov_iter *src, struct rte_vhost_iov_iter *dst, + struct rte_vhost_iov_iter *mirror_dst) +{ + desc->src = src; + desc->dst = dst; + desc->mirror_dst = mirror_dst; +} + +static __rte_always_inline int +egress_async_desc_to_mbuf(struct virtio_net *dev, + struct buf_vector *buf_vec, uint16_t nr_vec, + struct rte_mbuf *m, struct rte_mempool *mbuf_pool, + struct iovec *src_iovec, struct iovec *dst_iovec, + struct rte_vhost_iov_iter *src_it, + struct rte_vhost_iov_iter *dst_it, + struct virtio_net *mirr_dev, struct vhost_virtqueue *mirr_vq __rte_unused, + struct buf_vector *mirr_buf_vec, uint16_t mirr_nr_vec, uint16_t mirr_num_buffers, + struct iovec *mirr_dst_iovec, struct rte_vhost_iov_iter *mirr_dst_it, + struct virtio_net_hdr *nethdr, + int nr_iovec) +{ + uint64_t buf_addr, buf_iova; + uint64_t mapped_len; + uint32_t tlen = 0; + uint32_t buf_avail, buf_offset, buf_len; + uint32_t mbuf_avail, mbuf_offset; + uint32_t cpy_len; + /* A counter to avoid desc dead loop chain */ + uint16_t vec_idx = 0; + int tvec_idx = 0; + struct rte_mbuf *cur = m, *prev = m; + struct virtio_net_hdr tmp_hdr; + struct virtio_net_hdr *hdr = NULL; + + int error = 0; + uint64_t mirror_mapped_len; + uint64_t mirror_buf_addr, mirror_buf_iova; + uint32_t mirror_buf_avail, mirror_buf_offset, mirror_buf_len; + uint32_t mirror_vec_idx = 0; + + mirror_buf_addr = mirr_buf_vec[mirror_vec_idx].buf_addr; + mirror_buf_iova = mirr_buf_vec[mirror_vec_idx].buf_iova; + mirror_buf_len = mirr_buf_vec[mirror_vec_idx].buf_len; + + if (unlikely(mirror_buf_len < mirr_dev->vhost_hlen && mirr_nr_vec <= 1)) { + error = -1; + goto out; + } + + struct virtio_net_hdr_mrg_rxbuf *mirror_hdr = NULL; + mirror_hdr = (struct virtio_net_hdr_mrg_rxbuf *)(uintptr_t)mirror_buf_addr; + if (mirror_hdr && rxvq_is_mergeable(mirr_dev)) + ASSIGN_UNLESS_EQUAL(mirror_hdr->num_buffers, mirr_num_buffers); + + if (unlikely(mirror_buf_len < mirr_dev->vhost_hlen)) { + mirror_buf_offset = mirr_dev->vhost_hlen - mirror_buf_len; + mirror_vec_idx++; + mirror_buf_addr = mirr_buf_vec[mirror_vec_idx].buf_addr; + mirror_buf_iova = mirr_buf_vec[mirror_vec_idx].buf_iova; + mirror_buf_len = mirr_buf_vec[mirror_vec_idx].buf_len; + mirror_buf_avail = mirror_buf_len - mirror_buf_offset; + } else { + mirror_buf_offset = mirr_dev->vhost_hlen; + mirror_buf_avail = mirror_buf_len - mirr_dev->vhost_hlen; + } + + buf_addr = buf_vec[vec_idx].buf_addr; + buf_len = buf_vec[vec_idx].buf_len; + buf_iova = buf_vec[vec_idx].buf_iova; + + if (unlikely(buf_len < dev->vhost_hlen && nr_vec <= 1)) + return -1; + + 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)) + 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; + + 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; + int flag = 0; + static uint64_t total = 0; + while (1) { + cpy_len = RTE_MIN(buf_avail, mbuf_avail); + cpy_len = RTE_MIN(cpy_len, mirror_buf_avail); + + while (cpy_len) { + void *hpa = (void *)(uintptr_t)gpa_to_first_hpa(dev, + buf_iova + buf_offset, cpy_len, + &mapped_len); + if (unlikely(!hpa)) { + VHOST_LOG_DATA(ERR, "(%d) %s: failed to get hpa.\n", + dev->vid, __func__); + return -1; + } + + void *mirror_hpa = (void *)(uintptr_t)gpa_to_first_hpa(mirr_dev, + mirror_buf_iova + mirror_buf_offset, cpy_len, + &mirror_mapped_len); + if (unlikely(!mirror_hpa)) { + VHOST_LOG_DATA(ERR, "(%d) %s: failed to get mirror hpa.\n", + mirr_dev->vid, __func__); + error = -1; + goto out; + } + + if (unlikely(tvec_idx >= nr_iovec)) { + VHOST_LOG_DATA(ERR, "iovec is not enough for offloading\n"); + return -1; + } + + if (unlikely(mirror_mapped_len != mapped_len)) { + VHOST_LOG_DATA(ERR, "original mapped len is not equal to mirror len\n"); + return -1; + } + + async_fill_vec(src_iovec + tvec_idx, hpa, (size_t)mapped_len); + async_fill_vec(dst_iovec + tvec_idx, + (void *)(uintptr_t)rte_pktmbuf_iova_offset(cur, mbuf_offset), + (size_t)mapped_len); + async_fill_vec(mirr_dst_iovec + tvec_idx, + mirror_hpa, (size_t)mirror_mapped_len); + + if ((((uintptr_t)rte_pktmbuf_iova_offset(cur, mbuf_offset) & 0xFFF) ^ ((uint64_t)mirror_hpa & 0xFFF)) != 0 && flag == 0) + { + total++; + VHOST_LOG_DATA(ERR, "%lu.........mbuf=%p mirror_hpa=%p.\n", total, (void *)(uintptr_t)rte_pktmbuf_iova_offset(cur, mbuf_offset), mirror_hpa); + } + + tvec_idx++; + 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; + mirror_buf_avail -= (uint32_t)mapped_len; + mirror_buf_offset += (uint32_t)mapped_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); + } + + if (mirror_buf_avail == 0) { + mirror_vec_idx++; + if (unlikely(mirror_vec_idx >= mirr_nr_vec)) { + error = -1; + goto out; + } + + mirror_buf_addr = mirr_buf_vec[mirror_vec_idx].buf_addr; + mirror_buf_iova = mirr_buf_vec[mirror_vec_idx].buf_iova; + mirror_buf_len = mirr_buf_vec[mirror_vec_idx].buf_len; + + mirror_buf_offset = 0; + mirror_buf_avail = mirror_buf_len; + } + + /* + * 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"); + return -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; + } + flag++; + } + + prev->data_len = mbuf_offset; + m->pkt_len += mbuf_offset; + + if (tlen) { + async_fill_iter(src_it, tlen, src_iovec, tvec_idx); + async_fill_iter(dst_it, tlen, dst_iovec, tvec_idx); + async_fill_iter(mirr_dst_it, tlen, mirr_dst_iovec, tvec_idx); + if (hdr) + *nethdr = *hdr; + } +out: + return error; +} + +static __rte_always_inline uint16_t +virtio_mirror_dev_tx_async_split(struct virtio_net *dev, + struct vhost_virtqueue *vq, uint16_t queue_id, + struct virtio_net *mirr_dev, + struct vhost_virtqueue *mirr_vq, uint16_t mirr_queue_id __rte_unused, + struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, + uint16_t count, bool legacy_ol_flags) +{ + static bool allocerr_warned; + bool dropped = false; + uint16_t free_entries; + uint16_t pkt_idx, slot_idx = 0; + uint16_t nr_done_pkts = 0; + uint16_t nr_async_burst = 0; + uint16_t pkt_err = 0; + uint16_t iovec_idx = 0, it_idx = 0; + struct rte_vhost_iov_iter *it_pool = vq->it_pool; + struct iovec *vec_pool = vq->vec_pool; + struct iovec *src_iovec = vec_pool; + struct iovec *dst_iovec = vec_pool + (VHOST_MAX_ASYNC_VEC >> 1); + struct rte_vhost_async_desc tdes[MAX_PKT_BURST]; + struct async_inflight_info *pkts_info = vq->async_pkts_info; + struct rte_mbuf *pkts_prealloc[MAX_PKT_BURST]; + + + struct buf_vector mirror_buf_vec[BUF_VECTOR_MAX]; + uint16_t mirror_num_buffers; + uint16_t mirror_avail_head; + mirror_avail_head = __atomic_load_n(&mirr_vq->avail->idx, __ATOMIC_ACQUIRE); + rte_prefetch0(&mirr_vq->avail->ring[mirr_vq->last_avail_idx & (mirr_vq->size - 1)]); + + struct rte_vhost_iov_iter *mirror_it_pool = mirr_vq->it_pool; + struct iovec *mirror_dst_iovec = mirr_vq->vec_pool; + uint16_t mirror_iovec_idx = 0, mirror_it_idx = 0; + + /** + * 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) + goto out; + + rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]); + + 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); + + if (rte_pktmbuf_alloc_bulk(mbuf_pool, pkts_prealloc, count)) + goto out; + + for (pkt_idx = 0; pkt_idx < count; pkt_idx++) { + uint16_t head_idx = 0; + uint16_t nr_vec = 0; + uint16_t to; + uint32_t buf_len; + int err; + struct buf_vector buf_vec[BUF_VECTOR_MAX]; + struct rte_mbuf *pkt = pkts_prealloc[pkt_idx]; + uint16_t mirror_nr_vec = 0; + + if (unlikely(fill_vec_buf_split(dev, vq, vq->last_avail_idx, + &nr_vec, buf_vec, + &head_idx, &buf_len, + VHOST_ACCESS_RO) < 0)) { + dropped = true; + break; + } + + if (unlikely(reserve_avail_buf_split(mirr_dev, mirr_vq, + buf_len, mirror_buf_vec, &mirror_num_buffers, + mirror_avail_head, &mirror_nr_vec) < 0)) { + VHOST_LOG_DATA(DEBUG, + "(%d) failed to get enough desc from mirror vring\n", + mirr_dev->vid); + mirr_vq->shadow_used_idx -= mirror_num_buffers; + dropped = true; + break; + } + + err = virtio_dev_pktmbuf_prep(dev, pkt, buf_len); + if (unlikely(err)) { + /** + * 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 = true; + break; + } + + slot_idx = (vq->async_pkts_idx + pkt_idx) & (vq->size - 1); + err = egress_async_desc_to_mbuf(dev, buf_vec, nr_vec, pkt, + mbuf_pool, &src_iovec[iovec_idx], + &dst_iovec[iovec_idx], &it_pool[it_idx], + &it_pool[it_idx + 1], + mirr_dev, mirr_vq, + mirror_buf_vec, mirror_nr_vec, mirror_num_buffers, + &mirror_dst_iovec[mirror_iovec_idx], + &mirror_it_pool[mirror_it_idx], + &pkts_info[slot_idx].nethdr, + (VHOST_MAX_ASYNC_VEC >> 1) - iovec_idx); + if (unlikely(err)) { + if (!allocerr_warned) { + VHOST_LOG_DATA(ERR, + "Failed to offload copies to async channel %s.\n", + dev->ifname); + allocerr_warned = true; + } + dropped = true; + break; + } + + egress_async_fill_desc(&tdes[nr_async_burst], + &it_pool[it_idx], &it_pool[it_idx + 1], &mirror_it_pool[mirror_it_idx]); + pkts_info[slot_idx].mbuf = pkt; + nr_async_burst++; + + iovec_idx += it_pool[it_idx].nr_segs; + it_idx += 2; + + mirror_iovec_idx += mirror_it_pool[mirror_it_idx].nr_segs; + mirror_it_idx += 1; + mirr_vq->last_avail_idx += mirror_num_buffers; + pkts_info[slot_idx].descs = mirror_num_buffers; + + /* store used descs */ + to = vq->async_desc_idx_split & (vq->size - 1); + vq->async_descs_split[to].id = head_idx; + vq->async_descs_split[to].len = 0; + vq->async_desc_idx_split++; + + vq->last_avail_idx++; + + if (unlikely(nr_async_burst >= VHOST_ASYNC_BATCH_THRESHOLD)) { + uint16_t nr_pkts; + int32_t ret; + + ret = vq->async_ops.transfer_data(dev->vid, queue_id, + tdes, 0, nr_async_burst, true); + if (unlikely(ret < 0)) { + VHOST_LOG_DATA(ERR, "(%d) async channel submit error\n", dev->vid); + ret = 0; + } + nr_pkts = ret; + + vq->async_pkts_inflight_n += nr_pkts; + it_idx = 0; + iovec_idx = 0; + + mirror_iovec_idx = 0; + mirror_it_idx = 0; + + if (unlikely(nr_pkts < nr_async_burst)) { + pkt_err = nr_async_burst - nr_pkts; + nr_async_burst = 0; + pkt_idx++; + break; + } + nr_async_burst = 0; + } + } + + if (unlikely(dropped)) + rte_pktmbuf_free_bulk(&pkts_prealloc[pkt_idx], count - pkt_idx); + + if (nr_async_burst) { + uint16_t nr_pkts; + int32_t ret; + + ret = vq->async_ops.transfer_data(dev->vid, queue_id, tdes, 0, nr_async_burst, true); + if (unlikely(ret < 0)) { + VHOST_LOG_DATA(ERR, "(%d) async channel submit error\n", dev->vid); + ret = 0; + } + nr_pkts = ret; + + vq->async_pkts_inflight_n += nr_pkts; + + if (unlikely(nr_pkts < nr_async_burst)) + pkt_err = nr_async_burst - nr_pkts; + } + + if (unlikely(pkt_err)) { + uint16_t nr_err_dma = pkt_err; + uint16_t num_descs = 0; + + pkt_idx -= nr_err_dma; + + /** + * recover async channel copy related structures and free pktmbufs + * for error pkts. + */ + vq->async_desc_idx_split -= nr_err_dma; + while (nr_err_dma-- > 0) { + rte_pktmbuf_free(pkts_info[slot_idx & (vq->size - 1)].mbuf); + num_descs += pkts_info[slot_idx & (vq->size - 1)].descs; + slot_idx--; + } + + /* recover available ring */ + vq->last_avail_idx -= pkt_err; + + mirr_vq->shadow_used_idx -= num_descs; + mirr_vq->last_avail_idx -= num_descs; + } + + vq->async_pkts_idx += pkt_idx; + + if (likely(mirr_vq->shadow_used_idx)) { + uint16_t to = mirr_vq->async_desc_idx_split & (mirr_vq->size - 1); + + store_dma_desc_info_split(mirr_vq->shadow_used_split, + mirr_vq->async_descs_split, mirr_vq->size, 0, to, mirr_vq->shadow_used_idx); + + mirr_vq->async_desc_idx_split += mirr_vq->shadow_used_idx; + mirr_vq->async_pkts_idx += pkt_idx; + mirr_vq->async_pkts_inflight_n += pkt_idx; + mirr_vq->shadow_used_idx = 0; + } + +out: + if (vq->async_pkts_inflight_n > 0) { + nr_done_pkts = async_poll_egress_completed_split(dev, vq, + queue_id, + mirr_dev, mirr_vq, + pkts, count, legacy_ol_flags); + } + + return nr_done_pkts; +} + +__rte_noinline +static uint16_t +virtio_mirror_dev_tx_async_split_legacy(struct virtio_net *dev, + struct vhost_virtqueue *vq, uint16_t queue_id, + struct virtio_net *mirr_dev, + struct vhost_virtqueue *mirr_vq, uint16_t mirr_queue_id, + struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, + uint16_t count) +{ + return virtio_mirror_dev_tx_async_split(dev, vq, queue_id, + mirr_dev, mirr_vq, mirr_queue_id, + mbuf_pool, + pkts, count, true); +} + +__rte_noinline +static uint16_t +virtio_mirror_dev_tx_async_split_compliant(struct virtio_net *dev, + struct vhost_virtqueue *vq, uint16_t queue_id, + struct virtio_net *mirr_dev, + struct vhost_virtqueue *mirr_vq, uint16_t mirr_queue_id, + struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, + uint16_t count) +{ + return virtio_mirror_dev_tx_async_split(dev, vq, queue_id, + mirr_dev, mirr_vq, mirr_queue_id, + mbuf_pool, + pkts, count, false); +} + +uint16_t +rte_vhost_async_try_egress_burst(int vid, uint16_t queue_id, + int mirr_vid, uint16_t mirr_queue_id, + struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count, + int *nr_inflight) +{ + struct virtio_net *dev; + struct virtio_net *mirror_dev; + struct rte_mbuf *rarp_mbuf = NULL; + struct vhost_virtqueue *vq; + struct vhost_virtqueue *mirror_vq; + int16_t success = 1; + + *nr_inflight = -1; + + dev = get_device(vid); + mirror_dev = get_device(mirr_vid); + if (!dev || !mirror_dev) + return 0; + + if (unlikely(!(dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET)) + || unlikely(!(mirror_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; + } + + if (unlikely(!is_valid_virt_queue_idx(queue_id, 1, dev->nr_vring)) + || unlikely(!is_valid_virt_queue_idx(mirr_queue_id, 0, mirror_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]; + mirror_vq = mirror_dev->virtqueue[mirr_queue_id]; + + if (unlikely(rte_spinlock_trylock(&vq->access_lock) == 0) + || unlikely(rte_spinlock_trylock(&mirror_vq->access_lock) == 0)) + return 0; + + if (unlikely(vq->enabled == 0) || unlikely(mirror_vq->enabled == 0)) { + count = 0; + goto out_access_unlock; + } + + if (unlikely(!vq->async_registered) || unlikely(!mirror_vq->async_registered)) { + VHOST_LOG_DATA(ERR, "(%d) %s: async not registered for queue id %d.\n", + dev->vid, __func__, queue_id); + count = 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) || unlikely(mirror_vq->access_ok == 0)) + if (unlikely(vring_translate(dev, vq) < 0) + || unlikely(vring_translate(mirror_dev, mirror_vq) < 0)) { + count = 0; + goto out_access_unlock; + } + + /* + * 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 + * __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 compare and exchange if the read indicates it + * is likely to be set. + */ + 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) { + VHOST_LOG_DATA(ERR, "Failed to make RARP packet.\n"); + count = 0; + goto out; + } + count -= 1; + } + +#if 0 + if (unlikely(__atomic_load_n(&mirror_dev->broadcast_rarp, __ATOMIC_ACQUIRE) && + __atomic_compare_exchange_n(&mirror_dev->broadcast_rarp, + &success, 0, 0, __ATOMIC_RELEASE, __ATOMIC_RELAXED))) { + + rarp_mbuf = rte_net_make_rarp_packet(mbuf_pool, &mirror_dev->mac); + if (rarp_mbuf == NULL) { + VHOST_LOG_DATA(ERR, "Failed to make RARP packet.\n"); + count = 0; + goto out; + } + count -= 1; + } +#endif + if (unlikely(vq_is_packed(dev)) || unlikely(vq_is_packed(mirror_dev))) { + VHOST_LOG_DATA(ERR, + "(%d) %s: async dequeue does not support packed ring.\n", + dev->vid, __func__); + return 0; + } + + if (dev->flags & VIRTIO_DEV_LEGACY_OL_FLAGS) + count = virtio_mirror_dev_tx_async_split_legacy(dev, vq, queue_id, + mirror_dev, mirror_vq, mirr_queue_id, + mbuf_pool, pkts, count); + else + count = virtio_mirror_dev_tx_async_split_compliant(dev, vq, queue_id, + mirror_dev, mirror_vq, mirr_queue_id, + mbuf_pool, pkts, count); + +out: + *nr_inflight = vq->async_pkts_inflight_n; + + if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) + vhost_user_iotlb_rd_unlock(vq); + +out_access_unlock: + rte_spinlock_unlock(&vq->access_lock); + rte_spinlock_unlock(&mirror_vq->access_lock); + + if (unlikely(rarp_mbuf != NULL)) { + /* + * Inject it to the head of "pkts" array, so that switch's mac + * learning table will get updated first. + */ + memmove(&pkts[1], pkts, count * sizeof(struct rte_mbuf *)); + pkts[0] = rarp_mbuf; + count += 1; + } + + return count; +} -- 2.35.1