From patchwork Wed Jul 30 05:33:32 2014 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Fu, JingguoX" X-Patchwork-Id: 110 Return-Path: Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by dpdk.org (Postfix) with ESMTP id 2D046B363 for ; Wed, 30 Jul 2014 07:32:27 +0200 (CEST) Received: from orsmga001.jf.intel.com ([10.7.209.18]) by orsmga102.jf.intel.com with ESMTP; 29 Jul 2014 22:28:37 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.01,762,1400050800"; d="scan'208";a="551089565" Received: from fmsmsx104.amr.corp.intel.com ([10.19.9.35]) by orsmga001.jf.intel.com with ESMTP; 29 Jul 2014 22:34:20 -0700 Received: from fmsmsx119.amr.corp.intel.com (10.19.9.28) by FMSMSX104.amr.corp.intel.com (10.19.9.35) with Microsoft SMTP Server (TLS) id 14.3.123.3; Tue, 29 Jul 2014 22:34:20 -0700 Received: from shsmsx152.ccr.corp.intel.com (10.239.6.52) by FMSMSX119.amr.corp.intel.com (10.19.9.28) with Microsoft SMTP Server (TLS) id 14.3.123.3; Tue, 29 Jul 2014 22:34:20 -0700 Received: from shsmsx101.ccr.corp.intel.com ([169.254.1.52]) by SHSMSX152.ccr.corp.intel.com ([169.254.6.112]) with mapi id 14.03.0123.003; Wed, 30 Jul 2014 13:33:32 +0800 From: "Fu, JingguoX" To: "dev@dpdk.org" Thread-Topic: [dpdk-dev] [PATCH v2] virtio: Support mergeable buffer in virtio PMD Thread-Index: AQHPp85A6VchzX6YX0a+pnqht5b55pu4H0Cg Date: Wed, 30 Jul 2014 05:33:32 +0000 Message-ID: <6BD6202160B55B409D423293115822625041BE@SHSMSX101.ccr.corp.intel.com> References: <1406268195-24010-1-git-send-email-changchun.ouyang@intel.com> In-Reply-To: <1406268195-24010-1-git-send-email-changchun.ouyang@intel.com> Accept-Language: en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-cr-hashedpuzzle: AExK AQLN AVgW BQ7L CEuX CY2W Cnd2 CzUv DG1v F9TX G0DL HIyT H7Ob IJkd Iw/g JqID; 1; ZABlAHYAQABkAHAAZABrAC4AbwByAGcA; Sosha1_v1; 7; {D60630D7-7611-4042-AF31-49350C2D5567}; agBpAG4AZwBnAHUAbwB4AC4AZgB1AEAAaQBuAHQAZQBsAC4AYwBvAG0A; Wed, 30 Jul 2014 05:33:30 GMT; UgBFADoAIABbAGQAcABkAGsALQBkAGUAdgBdACAAWwBQAEEAVABDAEgAIAB2ADIAXQAgAHYAaQByAHQAaQBvADoAIABTAHUAcABwAG8AcgB0ACAAbQBlAHIAZwBlAGEAYgBsAGUAIABiAHUAZgBmAGUAcgAgAGkAbgAgAHYAaQByAHQAaQBvACAAUABNAEQA x-cr-puzzleid: {D60630D7-7611-4042-AF31-49350C2D5567} x-originating-ip: [10.239.127.40] MIME-Version: 1.0 Subject: Re: [dpdk-dev] [PATCH v2] virtio: Support mergeable buffer in virtio PMD X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 30 Jul 2014 05:32:29 -0000 Tested-by: Jingguo Fu This patch includes 1 file, and has been tested by Intel. Please see information as the following: Host: Fedora 19 x86_64, Linux Kernel 3.9.0, GCC 4.8.2 Intel Xeon CPU E5-2680 v2 @ 2.80GHz NIC: Intel Niantic 82599, Intel i350, Intel 82580 and Intel 82576 Guest: Fedora 16 x86_64, Linux Kernel 3.4.2, GCC 4.6.3 Qemu emulator 1.4.2 We verified Linux legacy regression tests for normal frame, Linux Legacy jumbo frame / scatter gather tests, vhost regression tests Include zero copy and one copy test cases for normal frame. Total case Passed Failed 10 10 0 -----Original Message----- From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ouyang Changchun Sent: Friday, July 25, 2014 2:03 PM To: dev@dpdk.org Subject: [dpdk-dev] [PATCH v2] virtio: Support mergeable buffer in virtio PMD v2 change: - Resolve conflicts wiht the tip code; - And resolve 2 issues: -- fix mbuf leak when discard a uncompleted packet. -- refine pkt.data to point to actual payload data start point. v1 change: This patch supports mergeable buffer feature in DPDK based virtio PMD, which can receive jumbo frame with larger size, like 3K, 4K or even 9K. Signed-off-by: Changchun Ouyang --- lib/librte_pmd_virtio/virtio_ethdev.c | 20 ++-- lib/librte_pmd_virtio/virtio_ethdev.h | 3 + lib/librte_pmd_virtio/virtio_rxtx.c | 206 +++++++++++++++++++++++++++++----- 3 files changed, 194 insertions(+), 35 deletions(-) diff --git a/lib/librte_pmd_virtio/virtio_ethdev.c b/lib/librte_pmd_virtio/virtio_ethdev.c index b9f5529..535d798 100644 --- a/lib/librte_pmd_virtio/virtio_ethdev.c +++ b/lib/librte_pmd_virtio/virtio_ethdev.c @@ -337,7 +337,7 @@ int virtio_dev_queue_setup(struct rte_eth_dev *dev, snprintf(vq_name, sizeof(vq_name), "port%d_tvq%d_hdrzone", dev->data->port_id, queue_idx); vq->virtio_net_hdr_mz = rte_memzone_reserve_aligned(vq_name, - vq_size * sizeof(struct virtio_net_hdr), + vq_size * hw->vtnet_hdr_size, socket_id, 0, CACHE_LINE_SIZE); if (vq->virtio_net_hdr_mz == NULL) { rte_free(vq); @@ -346,7 +346,7 @@ int virtio_dev_queue_setup(struct rte_eth_dev *dev, vq->virtio_net_hdr_mem = vq->virtio_net_hdr_mz->phys_addr; memset(vq->virtio_net_hdr_mz->addr, 0, - vq_size * sizeof(struct virtio_net_hdr)); + vq_size * hw->vtnet_hdr_size); } else if (queue_type == VTNET_CQ) { /* Allocate a page for control vq command, data and status */ snprintf(vq_name, sizeof(vq_name), "port%d_cvq_hdrzone", @@ -571,9 +571,6 @@ virtio_negotiate_features(struct virtio_hw *hw) mask |= VIRTIO_NET_F_GUEST_TSO4 | VIRTIO_NET_F_GUEST_TSO6 | VIRTIO_NET_F_GUEST_ECN; mask |= VTNET_LRO_FEATURES; - /* rx_mbuf should not be in multiple merged segments */ - mask |= VIRTIO_NET_F_MRG_RXBUF; - /* not negotiating INDIRECT descriptor table support */ mask |= VIRTIO_RING_F_INDIRECT_DESC; @@ -746,7 +743,6 @@ eth_virtio_dev_init(__rte_unused struct eth_driver *eth_drv, } eth_dev->dev_ops = &virtio_eth_dev_ops; - eth_dev->rx_pkt_burst = &virtio_recv_pkts; eth_dev->tx_pkt_burst = &virtio_xmit_pkts; if (rte_eal_process_type() == RTE_PROC_SECONDARY) @@ -801,10 +797,13 @@ eth_virtio_dev_init(__rte_unused struct eth_driver *eth_drv, virtio_negotiate_features(hw); /* Setting up rx_header size for the device */ - if (vtpci_with_feature(hw, VIRTIO_NET_F_MRG_RXBUF)) + if (vtpci_with_feature(hw, VIRTIO_NET_F_MRG_RXBUF)) { + eth_dev->rx_pkt_burst = &virtio_recv_mergeable_pkts; hw->vtnet_hdr_size = sizeof(struct virtio_net_hdr_mrg_rxbuf); - else + } else { + eth_dev->rx_pkt_burst = &virtio_recv_pkts; hw->vtnet_hdr_size = sizeof(struct virtio_net_hdr); + } /* Allocate memory for storing MAC addresses */ eth_dev->data->mac_addrs = rte_zmalloc("virtio", ETHER_ADDR_LEN, 0); @@ -1009,7 +1008,7 @@ static void virtio_dev_free_mbufs(struct rte_eth_dev *dev) while ((buf = (struct rte_mbuf *)virtqueue_detatch_unused( dev->data->rx_queues[i])) != NULL) { - rte_pktmbuf_free_seg(buf); + rte_pktmbuf_free(buf); mbuf_num++; } @@ -1028,7 +1027,8 @@ static void virtio_dev_free_mbufs(struct rte_eth_dev *dev) mbuf_num = 0; while ((buf = (struct rte_mbuf *)virtqueue_detatch_unused( dev->data->tx_queues[i])) != NULL) { - rte_pktmbuf_free_seg(buf); + rte_pktmbuf_free(buf); + mbuf_num++; } diff --git a/lib/librte_pmd_virtio/virtio_ethdev.h b/lib/librte_pmd_virtio/virtio_ethdev.h index 858e644..d2e1eed 100644 --- a/lib/librte_pmd_virtio/virtio_ethdev.h +++ b/lib/librte_pmd_virtio/virtio_ethdev.h @@ -104,6 +104,9 @@ int virtio_dev_tx_queue_setup(struct rte_eth_dev *dev, uint16_t tx_queue_id, uint16_t virtio_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts); +uint16_t virtio_recv_mergeable_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, + uint16_t nb_pkts); + uint16_t virtio_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts); diff --git a/lib/librte_pmd_virtio/virtio_rxtx.c b/lib/librte_pmd_virtio/virtio_rxtx.c index fcd8bd1..3d81b34 100644 --- a/lib/librte_pmd_virtio/virtio_rxtx.c +++ b/lib/librte_pmd_virtio/virtio_rxtx.c @@ -146,6 +146,7 @@ static inline int virtqueue_enqueue_recv_refill(struct virtqueue *vq, struct rte_mbuf *cookie) { struct vq_desc_extra *dxp; + struct virtio_hw *hw = vq->hw; struct vring_desc *start_dp; uint16_t needed = 1; uint16_t head_idx, idx; @@ -165,9 +166,11 @@ virtqueue_enqueue_recv_refill(struct virtqueue *vq, struct rte_mbuf *cookie) dxp->ndescs = needed; start_dp = vq->vq_ring.desc; - start_dp[idx].addr = - (uint64_t) (cookie->buf_physaddr + RTE_PKTMBUF_HEADROOM - sizeof(struct virtio_net_hdr)); - start_dp[idx].len = cookie->buf_len - RTE_PKTMBUF_HEADROOM + sizeof(struct virtio_net_hdr); + start_dp[idx].addr = + (uint64_t)(cookie->buf_physaddr + RTE_PKTMBUF_HEADROOM + - hw->vtnet_hdr_size); + start_dp[idx].len = + cookie->buf_len - RTE_PKTMBUF_HEADROOM + hw->vtnet_hdr_size; start_dp[idx].flags = VRING_DESC_F_WRITE; idx = start_dp[idx].next; vq->vq_desc_head_idx = idx; @@ -184,8 +187,10 @@ virtqueue_enqueue_xmit(struct virtqueue *txvq, struct rte_mbuf *cookie) { struct vq_desc_extra *dxp; struct vring_desc *start_dp; - uint16_t needed = 2; + uint16_t seg_num = cookie->pkt.nb_segs; + uint16_t needed = 1 + seg_num; uint16_t head_idx, idx; + uint16_t head_size = txvq->hw->vtnet_hdr_size; if (unlikely(txvq->vq_free_cnt == 0)) return -ENOSPC; @@ -198,19 +203,25 @@ virtqueue_enqueue_xmit(struct virtqueue *txvq, struct rte_mbuf *cookie) idx = head_idx; dxp = &txvq->vq_descx[idx]; if (dxp->cookie != NULL) - rte_pktmbuf_free_seg(dxp->cookie); + rte_pktmbuf_free(dxp->cookie); dxp->cookie = (void *)cookie; dxp->ndescs = needed; start_dp = txvq->vq_ring.desc; - start_dp[idx].addr = - txvq->virtio_net_hdr_mem + idx * sizeof(struct virtio_net_hdr); - start_dp[idx].len = sizeof(struct virtio_net_hdr); + start_dp[idx].addr = + txvq->virtio_net_hdr_mem + idx * head_size; + start_dp[idx].len = (uint32_t)head_size; start_dp[idx].flags = VRING_DESC_F_NEXT; - idx = start_dp[idx].next; - start_dp[idx].addr = RTE_MBUF_DATA_DMA_ADDR(cookie); - start_dp[idx].len = cookie->pkt.data_len; - start_dp[idx].flags = 0; + + for (; ((seg_num > 0) && (cookie != NULL)); seg_num--) { + idx = start_dp[idx].next; + start_dp[idx].addr = RTE_MBUF_DATA_DMA_ADDR(cookie); + start_dp[idx].len = cookie->pkt.data_len; + start_dp[idx].flags = VRING_DESC_F_NEXT; + cookie = cookie->pkt.next; + } + + start_dp[idx].flags &= ~VRING_DESC_F_NEXT; idx = start_dp[idx].next; txvq->vq_desc_head_idx = idx; if (txvq->vq_desc_head_idx == VQ_RING_DESC_CHAIN_END) @@ -284,7 +295,7 @@ virtio_dev_vring_start(struct virtqueue *vq, int queue_type) error = virtqueue_enqueue_recv_refill(vq, m); if (error) { - rte_pktmbuf_free_seg(m); + rte_pktmbuf_free(m); break; } nbufs++; @@ -423,7 +434,7 @@ virtio_discard_rxbuf(struct virtqueue *vq, struct rte_mbuf *m) error = virtqueue_enqueue_recv_refill(vq, m); if (unlikely(error)) { RTE_LOG(ERR, PMD, "cannot requeue discarded mbuf"); - rte_pktmbuf_free_seg(m); + rte_pktmbuf_free(m); } } @@ -471,17 +482,158 @@ virtio_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) rxm->pkt.in_port = rxvq->port_id; rxm->pkt.data = (char *)rxm->buf_addr + RTE_PKTMBUF_HEADROOM; + rxm->pkt.nb_segs = 1; rxm->pkt.next = NULL; - rxm->pkt.pkt_len = (uint32_t)(len[i] - - sizeof(struct virtio_net_hdr)); - rxm->pkt.data_len = (uint16_t)(len[i] - - sizeof(struct virtio_net_hdr)); + rxm->pkt.pkt_len = (uint32_t)(len[i] - hw->vtnet_hdr_size); + rxm->pkt.data_len = (uint16_t)(len[i] - hw->vtnet_hdr_size); VIRTIO_DUMP_PACKET(rxm, rxm->pkt.data_len); rx_pkts[nb_rx++] = rxm; - rxvq->bytes += len[i] - sizeof(struct virtio_net_hdr); + rxvq->bytes += rx_pkts[nb_rx - 1]->pkt.pkt_len; + } + + rxvq->packets += nb_rx; + + /* Allocate new mbuf for the used descriptor */ + error = ENOSPC; + while (likely(!virtqueue_full(rxvq))) { + new_mbuf = rte_rxmbuf_alloc(rxvq->mpool); + if (unlikely(new_mbuf == NULL)) { + struct rte_eth_dev *dev + = &rte_eth_devices[rxvq->port_id]; + dev->data->rx_mbuf_alloc_failed++; + break; + } + error = virtqueue_enqueue_recv_refill(rxvq, new_mbuf); + if (unlikely(error)) { + rte_pktmbuf_free(new_mbuf); + break; + } + nb_enqueued++; + } + + if (likely(nb_enqueued)) { + if (unlikely(virtqueue_kick_prepare(rxvq))) { + virtqueue_notify(rxvq); + PMD_RX_LOG(DEBUG, "Notified\n"); + } + } + + vq_update_avail_idx(rxvq); + + return nb_rx; +} + +uint16_t +virtio_recv_mergeable_pkts(void *rx_queue, + struct rte_mbuf **rx_pkts, + uint16_t nb_pkts) +{ + struct virtqueue *rxvq = rx_queue; + struct virtio_hw *hw = rxvq->hw; + struct rte_mbuf *rxm, *new_mbuf; + uint16_t nb_used, num, nb_rx = 0; + uint32_t len[VIRTIO_MBUF_BURST_SZ]; + struct rte_mbuf *rcv_pkts[VIRTIO_MBUF_BURST_SZ]; + struct rte_mbuf *prev; + int error; + uint32_t i = 0, nb_enqueued = 0; + uint32_t seg_num = 0; + uint16_t extra_idx = 0; + + nb_used = VIRTQUEUE_NUSED(rxvq); + + rmb(); + + if (nb_used == 0) + return 0; + + while (i < nb_used) { + struct virtio_net_hdr_mrg_rxbuf *header; + char *head_ptr; + + if (nb_rx >= nb_pkts) + break; + + num = virtqueue_dequeue_burst_rx(rxvq, rcv_pkts, len, 1); + i += num; + + PMD_RX_LOG(DEBUG, "used:%d dequeue:%d\n", nb_used, num); + PMD_RX_LOG(DEBUG, "packet len:%d\n", len[i]); + + rxm = rcv_pkts[0]; + extra_idx = 0; + + if (unlikely(len[0] + < (uint32_t)hw->vtnet_hdr_size + ETHER_HDR_LEN)) { + PMD_RX_LOG(ERR, "Packet drop\n"); + nb_enqueued++; + virtio_discard_rxbuf(rxvq, rxm); + rxvq->errors++; + continue; + } + + head_ptr = (char *)rxm->pkt.data; + head_ptr -= hw->vtnet_hdr_size; + header = (struct virtio_net_hdr_mrg_rxbuf *)head_ptr; + seg_num = header->num_buffers; + + if (seg_num == 0) + seg_num = 1; + + rxm->pkt.data = (char *)rxm->buf_addr + RTE_PKTMBUF_HEADROOM; + rxm->pkt.nb_segs = seg_num; + rxm->pkt.next = NULL; + rxm->pkt.pkt_len = (uint32_t)(len[0] - hw->vtnet_hdr_size); + rxm->pkt.data_len = (uint16_t)(len[0] - hw->vtnet_hdr_size); + + rxm->pkt.in_port = rxvq->port_id; + rx_pkts[nb_rx] = rxm; + + prev = rxm; + + VIRTIO_DUMP_PACKET(rxm, rxm->pkt.data_len); + + /* + * Get extra segments for current uncompleted packet. + */ + if (VIRTQUEUE_NUSED(rxvq) >= seg_num - 1) { + uint32_t rx_num = virtqueue_dequeue_burst_rx(rxvq, + rcv_pkts, len, seg_num - 1); + i += rx_num; + } else { + PMD_RX_LOG(ERR, "No enough segments for packet.\n"); + nb_enqueued++; + virtio_discard_rxbuf(rxvq, rxm); + rxvq->errors++; + break; + } + + while (extra_idx < seg_num - 1) { + rxm = rcv_pkts[extra_idx]; + + rxm->pkt.data = + (char *)rxm->buf_addr + RTE_PKTMBUF_HEADROOM + - hw->vtnet_hdr_size; + rxm->pkt.next = NULL; + rxm->pkt.pkt_len = (uint32_t)(len[extra_idx]); + rxm->pkt.data_len = (uint16_t)(len[extra_idx]); + + if (prev) + prev->pkt.next = rxm; + + prev = rxm; + rx_pkts[nb_rx]->pkt.pkt_len += rxm->pkt.pkt_len; + + extra_idx++; + + VIRTIO_DUMP_PACKET(rxm, rxm->pkt.data_len); + }; + + rxvq->bytes += rx_pkts[nb_rx]->pkt.pkt_len; + nb_rx++; } rxvq->packets += nb_rx; @@ -498,11 +650,12 @@ virtio_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) } error = virtqueue_enqueue_recv_refill(rxvq, new_mbuf); if (unlikely(error)) { - rte_pktmbuf_free_seg(new_mbuf); + rte_pktmbuf_free(new_mbuf); break; } nb_enqueued++; } + if (likely(nb_enqueued)) { if (unlikely(virtqueue_kick_prepare(rxvq))) { virtqueue_notify(rxvq); @@ -536,12 +689,15 @@ virtio_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) num = (uint16_t)(likely(nb_used < VIRTIO_MBUF_BURST_SZ) ? nb_used : VIRTIO_MBUF_BURST_SZ); while (nb_tx < nb_pkts) { - if (virtqueue_full(txvq) && num) { - virtqueue_dequeue_pkt_tx(txvq); - num--; + int need = tx_pkts[nb_tx]->pkt.nb_segs - txvq->vq_free_cnt; + if ((need > 0) && (num > 0)) { + do { + virtqueue_dequeue_pkt_tx(txvq); + num--; + } while (num > 0); } - if (!virtqueue_full(txvq)) { + if (tx_pkts[nb_tx]->pkt.nb_segs <= txvq->vq_free_cnt) { txm = tx_pkts[nb_tx]; /* Enqueue Packet buffers */ error = virtqueue_enqueue_xmit(txvq, txm); @@ -555,7 +711,7 @@ virtio_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) break; } nb_tx++; - txvq->bytes += txm->pkt.data_len; + txvq->bytes += txm->pkt.pkt_len; } else { PMD_TX_LOG(ERR, "No free tx descriptors to transmit"); break;