From patchwork Mon Jul 8 17:13:11 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Marvin Liu X-Patchwork-Id: 56211 X-Patchwork-Delegate: maxime.coquelin@redhat.com Return-Path: X-Original-To: patchwork@dpdk.org Delivered-To: patchwork@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id 11D411B998; Mon, 8 Jul 2019 11:30:01 +0200 (CEST) Received: from mga11.intel.com (mga11.intel.com [192.55.52.93]) by dpdk.org (Postfix) with ESMTP id A0E2C1B951 for ; Mon, 8 Jul 2019 11:29:53 +0200 (CEST) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga007.jf.intel.com ([10.7.209.58]) by fmsmga102.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 08 Jul 2019 02:29:52 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.63,466,1557212400"; d="scan'208";a="155821656" Received: from npg-dpdk-virtual-marvin-dev.sh.intel.com ([10.67.119.142]) by orsmga007.jf.intel.com with ESMTP; 08 Jul 2019 02:29:50 -0700 From: Marvin Liu To: tiwei.bie@intel.com, maxime.coquelin@redhat.com, dev@dpdk.org Cc: Marvin Liu Date: Tue, 9 Jul 2019 01:13:11 +0800 Message-Id: <20190708171320.38802-5-yong.liu@intel.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20190708171320.38802-1-yong.liu@intel.com> References: <20190708171320.38802-1-yong.liu@intel.com> Subject: [dpdk-dev] [RFC PATCH 04/13] add vhost packed ring fast dequeue function X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Add fast dequeue function just like enqueue function, fast dequeue function will not support chained nor indirect descriptors, normal function will handle that. Signed-off-by: Marvin Liu diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h index f24026acd..329a7658b 100644 --- a/lib/librte_vhost/vhost.h +++ b/lib/librte_vhost/vhost.h @@ -41,6 +41,10 @@ /* Used in fast packed ring functions */ #define PACKED_DESC_PER_CACHELINE (RTE_CACHE_LINE_SIZE / sizeof(struct vring_packed_desc)) + +/* Indicated that normal path will handle */ +#define VIRTIO_DESC_NORMAL_FLAG (VRING_DESC_F_NEXT | VRING_DESC_F_INDIRECT) + /** * Structure contains buffer address, length and descriptor index * from vring to do scatter RX. diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c index 410837122..a62e0feda 100644 --- a/lib/librte_vhost/virtio_net.c +++ b/lib/librte_vhost/virtio_net.c @@ -1613,6 +1613,158 @@ virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq, return i; } +static __rte_always_inline int +vhost_dequeue_fast_packed(struct virtio_net *dev, struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t avail_idx, + uintptr_t *desc_addr, uint16_t *ids) +{ + bool wrap_counter = vq->avail_wrap_counter; + struct vring_packed_desc *descs = vq->desc_packed; + uint64_t len, len1, len2, len3; + uint32_t buf_offset = dev->vhost_hlen; + + // check whether desc is cache aligned + if (unlikely(avail_idx & 0x3)) + return -1; + + // prefetch next cache line + if (unlikely(avail_idx < (vq->size - PACKED_DESC_PER_CACHELINE))) + rte_prefetch0((void *)(uintptr_t)&descs[avail_idx + + PACKED_DESC_PER_CACHELINE]); + else + rte_prefetch0((void *)(uintptr_t)&descs[0]); + + if (unlikely(!desc_is_avail(&descs[avail_idx], wrap_counter)) | + unlikely(!desc_is_avail(&descs[avail_idx + 1], wrap_counter)) | + unlikely(!desc_is_avail(&descs[avail_idx + 2], wrap_counter)) | + unlikely(!desc_is_avail(&descs[avail_idx + 3], wrap_counter))) + return 1; + + if (unlikely((descs[avail_idx].flags & VIRTIO_DESC_NORMAL_FLAG) | + (descs[avail_idx + 1].flags & VIRTIO_DESC_NORMAL_FLAG) | + (descs[avail_idx + 2].flags & VIRTIO_DESC_NORMAL_FLAG) | + (descs[avail_idx + 3].flags & VIRTIO_DESC_NORMAL_FLAG))) + return -1; + + rte_smp_rmb(); + + len = descs[avail_idx].len; + len1 = descs[avail_idx + 1].len; + len2 = descs[avail_idx + 2].len; + len3 = descs[avail_idx + 3].len; + + ids[0] = descs[avail_idx].id; + ids[1] = descs[avail_idx + 1].id; + ids[2] = descs[avail_idx + 2].id; + ids[3] = descs[avail_idx + 3].id; + + desc_addr[0] = vhost_iova_to_vva(dev, vq, + descs[avail_idx].addr, + &len, + VHOST_ACCESS_RW); + + desc_addr[1] = vhost_iova_to_vva(dev, vq, + descs[avail_idx + 1].addr, + &len1, + VHOST_ACCESS_RW); + + desc_addr[2] = vhost_iova_to_vva(dev, vq, + descs[avail_idx + 2].addr, + &len2, + VHOST_ACCESS_RW); + + desc_addr[3] = vhost_iova_to_vva(dev, vq, + descs[avail_idx + 3].addr, + &len3, + VHOST_ACCESS_RW); + + if (unlikely((len != descs[avail_idx].len) | + (len1 != descs[avail_idx + 1].len) | + (len2 != descs[avail_idx + 2].len) | + (len3 != descs[avail_idx + 3].len))) { + return -1; + } + + if (rte_pktmbuf_alloc_bulk(mbuf_pool, pkts ,4)) + return -1; + + if (unlikely(((uint64_t)(pkts[0]->buf_len - pkts[0]->data_off) < + (len + buf_offset)) | + ((uint64_t)(pkts[1]->buf_len - pkts[1]->data_off) < + (len1 + buf_offset)) | + ((uint64_t)(pkts[2]->buf_len - pkts[2]->data_off) < + (len2 + buf_offset)) | + ((uint64_t)(pkts[3]->buf_len - pkts[3]->data_off) < + (len3 + buf_offset)))) { + rte_pktmbuf_free(pkts[0]); + rte_pktmbuf_free(pkts[1]); + rte_pktmbuf_free(pkts[2]); + rte_pktmbuf_free(pkts[3]); + return -1; + } + + pkts[0]->pkt_len = descs[avail_idx].len - buf_offset; + pkts[1]->pkt_len = descs[avail_idx + 1].len - buf_offset; + pkts[2]->pkt_len = descs[avail_idx + 2].len - buf_offset; + pkts[3]->pkt_len = descs[avail_idx + 3].len - buf_offset; + + pkts[0]->data_len = pkts[0]->pkt_len; + pkts[1]->data_len = pkts[1]->pkt_len; + pkts[2]->data_len = pkts[2]->pkt_len; + pkts[3]->data_len = pkts[3]->pkt_len; + + return 0; +} + +static __rte_always_inline int +virtio_dev_tx_fast_packed(struct virtio_net *dev, struct vhost_virtqueue *vq, + struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts) +{ + uint16_t avail_idx = vq->last_avail_idx; + uint32_t buf_offset = dev->vhost_hlen; + uintptr_t desc_addr[4]; + uint16_t ids[4]; + int ret; + struct virtio_net_hdr *hdr, *hdr1, *hdr2, *hdr3; + + ret = vhost_dequeue_fast_packed(dev, vq, mbuf_pool, pkts, avail_idx, + desc_addr, ids); + + if (ret) + return ret; + + rte_memcpy(rte_pktmbuf_mtod_offset(pkts[0], void *, 0), + (void *)(uintptr_t)(desc_addr[0] + buf_offset), + pkts[0]->pkt_len); + rte_memcpy(rte_pktmbuf_mtod_offset(pkts[1], void *, 0), + (void *)(uintptr_t)(desc_addr[1] + buf_offset), + pkts[1]->pkt_len); + rte_memcpy(rte_pktmbuf_mtod_offset(pkts[2], void *, 0), + (void *)(uintptr_t)(desc_addr[2] + buf_offset), + pkts[2]->pkt_len); + rte_memcpy(rte_pktmbuf_mtod_offset(pkts[3], void *, 0), + (void *)(uintptr_t)(desc_addr[3] + buf_offset), + pkts[3]->pkt_len); + + if (virtio_net_with_host_offload(dev)) { + hdr = (struct virtio_net_hdr *)((uintptr_t)desc_addr[0]); + hdr1 = (struct virtio_net_hdr *)((uintptr_t)desc_addr[1]); + hdr2 = (struct virtio_net_hdr *)((uintptr_t)desc_addr[2]); + hdr3 = (struct virtio_net_hdr *)((uintptr_t)desc_addr[3]); + vhost_dequeue_offload(hdr, pkts[0]); + vhost_dequeue_offload(hdr1, pkts[1]); + vhost_dequeue_offload(hdr2, pkts[2]); + vhost_dequeue_offload(hdr3, pkts[3]); + } + + vq->last_avail_idx += PACKED_DESC_PER_CACHELINE; + if (vq->last_avail_idx >= vq->size) { + vq->last_avail_idx -= vq->size; + vq->avail_wrap_counter ^= 1; + } + return 0; +} + static __rte_always_inline int vhost_dequeue_normal_packed(struct virtio_net *dev, struct vhost_virtqueue *vq,