From patchwork Tue Aug 8 17:06:12 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Bodireddy, Bhanuprakash" X-Patchwork-Id: 799368 X-Patchwork-Delegate: dlu998@gmail.com Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Authentication-Results: ozlabs.org; spf=pass (mailfrom) smtp.mailfrom=openvswitch.org (client-ip=140.211.169.12; helo=mail.linuxfoundation.org; envelope-from=ovs-dev-bounces@openvswitch.org; receiver=) Received: from mail.linuxfoundation.org (mail.linuxfoundation.org [140.211.169.12]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ozlabs.org (Postfix) with ESMTPS id 3xRh1X2QD3z9s7C for ; Wed, 9 Aug 2017 03:18:08 +1000 (AEST) Received: from mail.linux-foundation.org (localhost [127.0.0.1]) by mail.linuxfoundation.org (Postfix) with ESMTP id E5C36AB6; Tue, 8 Aug 2017 17:17:37 +0000 (UTC) X-Original-To: dev@openvswitch.org Delivered-To: ovs-dev@mail.linuxfoundation.org Received: from smtp1.linuxfoundation.org (smtp1.linux-foundation.org [172.17.192.35]) by mail.linuxfoundation.org (Postfix) with ESMTPS id 461BDA7F for ; Tue, 8 Aug 2017 17:17:35 +0000 (UTC) X-Greylist: domain auto-whitelisted by SQLgrey-1.7.6 Received: from mga04.intel.com (mga04.intel.com [192.55.52.120]) by smtp1.linuxfoundation.org (Postfix) with ESMTPS id 54D32444 for ; Tue, 8 Aug 2017 17:15:27 +0000 (UTC) Received: from fmsmga004.fm.intel.com ([10.253.24.48]) by fmsmga104.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 08 Aug 2017 10:15:26 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.41,344,1498546800"; d="scan'208";a="297454186" Received: from silpixa00393942.ir.intel.com (HELO silpixa00393942.ger.corp.intel.com) ([10.237.223.42]) by fmsmga004.fm.intel.com with ESMTP; 08 Aug 2017 10:15:25 -0700 From: Bhanuprakash Bodireddy To: dev@openvswitch.org Date: Tue, 8 Aug 2017 18:06:12 +0100 Message-Id: <1502211976-76937-2-git-send-email-bhanuprakash.bodireddy@intel.com> X-Mailer: git-send-email 2.4.11 In-Reply-To: <1502211976-76937-1-git-send-email-bhanuprakash.bodireddy@intel.com> References: <1502211976-76937-1-git-send-email-bhanuprakash.bodireddy@intel.com> X-Spam-Status: No, score=-4.2 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_MED, RP_MATCHES_RCVD autolearn=unavailable version=3.3.1 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on smtp1.linux-foundation.org Subject: [ovs-dev] [PATCH v4 1/5] netdev: Add netdev_txq_flush function. X-BeenThere: ovs-dev@openvswitch.org X-Mailman-Version: 2.1.12 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , MIME-Version: 1.0 Sender: ovs-dev-bounces@openvswitch.org Errors-To: ovs-dev-bounces@openvswitch.org Add netdev_txq_flush(), that flush packets on a queue. This is needed to transmit packets on the intermediate queue. This commit also implements netdev_dpdk_txq_flush() function. If there are any packets waiting in the queue, they are transmitted instantly using the rte_eth_tx_burst function. In XPS enabled case, lock is taken on the tx queue before flushing the queue. Signed-off-by: Bhanuprakash Bodireddy Signed-off-by: Antonio Fischetti Co-authored-by: Antonio Fischetti Signed-off-by: Markus Magnusson Co-authored-by: Markus Magnusson Acked-by: Eelco Chaudron --- lib/netdev-bsd.c | 1 + lib/netdev-dpdk.c | 52 ++++++++++++++++++++++++++++++++++++++++++++++----- lib/netdev-dummy.c | 1 + lib/netdev-linux.c | 1 + lib/netdev-provider.h | 8 ++++++++ lib/netdev-vport.c | 2 +- lib/netdev.c | 9 +++++++++ lib/netdev.h | 1 + 8 files changed, 69 insertions(+), 6 deletions(-) diff --git a/lib/netdev-bsd.c b/lib/netdev-bsd.c index 8a4cdb3..75483ad 100644 --- a/lib/netdev-bsd.c +++ b/lib/netdev-bsd.c @@ -1546,6 +1546,7 @@ netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off, netdev_bsd_rxq_recv, \ netdev_bsd_rxq_wait, \ netdev_bsd_rxq_drain, \ + NULL, \ \ NO_OFFLOAD_API \ } diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c index 1d82bca..50d6b29 100644 --- a/lib/netdev-dpdk.c +++ b/lib/netdev-dpdk.c @@ -313,6 +313,11 @@ struct dpdk_mp { struct ovs_list list_node OVS_GUARDED_BY(dpdk_mp_mutex); }; +/* Queue 'INTERIM_QUEUE_BURST_THRESHOLD' packets before transmitting. + * Defaults to 'NETDEV_MAX_BURST'(32) packets. + */ +#define INTERIM_QUEUE_BURST_THRESHOLD NETDEV_MAX_BURST + /* There should be one 'struct dpdk_tx_queue' created for * each cpu core. */ struct dpdk_tx_queue { @@ -322,6 +327,12 @@ struct dpdk_tx_queue { * pmd threads (see 'concurrent_txq'). */ int map; /* Mapping of configured vhost-user queues * to enabled by guest. */ + int dpdk_pkt_cnt; /* Number of buffered packets waiting to + be sent on DPDK tx queue. */ + struct rte_mbuf *dpdk_burst_pkts[INTERIM_QUEUE_BURST_THRESHOLD]; + /* Intermediate queue where packets can + * be buffered to amortize the cost of MMIO + * writes. */ }; /* dpdk has no way to remove dpdk ring ethernet devices @@ -1931,6 +1942,32 @@ netdev_dpdk_send__(struct netdev_dpdk *dev, int qid, } } +/* Flush tx queues. + * This is done periodically to empty the intermediate queue in case of + * fewer packets (< INTERIM_QUEUE_BURST_THRESHOLD) buffered in the queue. + */ +static int +netdev_dpdk_txq_flush(struct netdev *netdev, int qid , bool concurrent_txq) +{ + struct netdev_dpdk *dev = netdev_dpdk_cast(netdev); + struct dpdk_tx_queue *txq = &dev->tx_q[qid]; + + if (OVS_LIKELY(txq->dpdk_pkt_cnt)) { + if (OVS_UNLIKELY(concurrent_txq)) { + qid = qid % dev->up.n_txq; + rte_spinlock_lock(&dev->tx_q[qid].tx_lock); + } + + netdev_dpdk_eth_tx_burst(dev, qid, txq->dpdk_burst_pkts, + txq->dpdk_pkt_cnt); + + if (OVS_UNLIKELY(concurrent_txq)) { + rte_spinlock_unlock(&dev->tx_q[qid].tx_lock); + } + } + return 0; +} + static int netdev_dpdk_eth_send(struct netdev *netdev, int qid, struct dp_packet_batch *batch, bool may_steal, @@ -3313,7 +3350,7 @@ unlock: SET_CONFIG, SET_TX_MULTIQ, SEND, \ GET_CARRIER, GET_STATS, \ GET_FEATURES, GET_STATUS, \ - RECONFIGURE, RXQ_RECV) \ + RECONFIGURE, RXQ_RECV, TXQ_FLUSH) \ { \ NAME, \ true, /* is_pmd */ \ @@ -3381,6 +3418,7 @@ unlock: RXQ_RECV, \ NULL, /* rx_wait */ \ NULL, /* rxq_drain */ \ + TXQ_FLUSH, /* txq_flush */ \ NO_OFFLOAD_API \ } @@ -3398,7 +3436,8 @@ static const struct netdev_class dpdk_class = netdev_dpdk_get_features, netdev_dpdk_get_status, netdev_dpdk_reconfigure, - netdev_dpdk_rxq_recv); + netdev_dpdk_rxq_recv, + netdev_dpdk_txq_flush); static const struct netdev_class dpdk_ring_class = NETDEV_DPDK_CLASS( @@ -3414,7 +3453,8 @@ static const struct netdev_class dpdk_ring_class = netdev_dpdk_get_features, netdev_dpdk_get_status, netdev_dpdk_reconfigure, - netdev_dpdk_rxq_recv); + netdev_dpdk_rxq_recv, + NULL); static const struct netdev_class dpdk_vhost_class = NETDEV_DPDK_CLASS( @@ -3430,7 +3470,8 @@ static const struct netdev_class dpdk_vhost_class = NULL, NULL, netdev_dpdk_vhost_reconfigure, - netdev_dpdk_vhost_rxq_recv); + netdev_dpdk_vhost_rxq_recv, + NULL); static const struct netdev_class dpdk_vhost_client_class = NETDEV_DPDK_CLASS( "dpdkvhostuserclient", @@ -3445,7 +3486,8 @@ static const struct netdev_class dpdk_vhost_client_class = NULL, NULL, netdev_dpdk_vhost_client_reconfigure, - netdev_dpdk_vhost_rxq_recv); + netdev_dpdk_vhost_rxq_recv, + NULL); void netdev_dpdk_register(void) diff --git a/lib/netdev-dummy.c b/lib/netdev-dummy.c index 752f157..86ec902 100644 --- a/lib/netdev-dummy.c +++ b/lib/netdev-dummy.c @@ -1413,6 +1413,7 @@ netdev_dummy_update_flags(struct netdev *netdev_, netdev_dummy_rxq_recv, \ netdev_dummy_rxq_wait, \ netdev_dummy_rxq_drain, \ + NULL, \ \ NO_OFFLOAD_API \ } diff --git a/lib/netdev-linux.c b/lib/netdev-linux.c index 98820ed..992f887 100644 --- a/lib/netdev-linux.c +++ b/lib/netdev-linux.c @@ -2888,6 +2888,7 @@ netdev_linux_update_flags(struct netdev *netdev_, enum netdev_flags off, netdev_linux_rxq_recv, \ netdev_linux_rxq_wait, \ netdev_linux_rxq_drain, \ + NULL, \ \ FLOW_OFFLOAD_API \ } diff --git a/lib/netdev-provider.h b/lib/netdev-provider.h index b3c57d5..9c47fdf 100644 --- a/lib/netdev-provider.h +++ b/lib/netdev-provider.h @@ -347,6 +347,11 @@ struct netdev_class { * If the function returns a non-zero value, some of the packets might have * been sent anyway. * + * Some netdev provider - like in case of 'dpdk' - may buffer the batch + * of packets into an intermediate queue. Buffered packets shall be + * transmitted when the packet count exceeds a threshold (or) by the + * periodic call to the flush function. + * * If 'may_steal' is false, the caller retains ownership of all the * packets. If 'may_steal' is true, the caller transfers ownership of all * the packets to the network device, regardless of success. @@ -788,6 +793,9 @@ struct netdev_class { /* Discards all packets waiting to be received from 'rx'. */ int (*rxq_drain)(struct netdev_rxq *rx); + /* Flush all packets waiting to be sent on 'qid' queue. */ + int (*txq_flush)(struct netdev *netdev, int qid, bool concurrent_txq); + /* ## -------------------------------- ## */ /* ## netdev flow offloading functions ## */ /* ## -------------------------------- ## */ diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c index 64a3ba3..3c5eacf 100644 --- a/lib/netdev-vport.c +++ b/lib/netdev-vport.c @@ -951,10 +951,10 @@ netdev_vport_get_ifindex(const struct netdev *netdev_) NULL, /* rx_recv */ \ NULL, /* rx_wait */ \ NULL, /* rx_drain */ \ + NULL, /* tx_flush */ \ \ NETDEV_FLOW_OFFLOAD_API - #define TUNNEL_CLASS(NAME, DPIF_PORT, BUILD_HEADER, PUSH_HEADER, POP_HEADER, \ GET_IFINDEX) \ { DPIF_PORT, \ diff --git a/lib/netdev.c b/lib/netdev.c index 7e9896b..8250396 100644 --- a/lib/netdev.c +++ b/lib/netdev.c @@ -724,6 +724,15 @@ netdev_rxq_drain(struct netdev_rxq *rx) : 0); } +/* Flush packets on the 'qid' queue. */ +int +netdev_txq_flush(struct netdev *netdev, int qid, bool netdev_txq_flush) +{ + return (netdev->netdev_class->txq_flush + ? netdev->netdev_class->txq_flush(netdev, qid, netdev_txq_flush) + : EOPNOTSUPP); +} + /* Configures the number of tx queues of 'netdev'. Returns 0 if successful, * otherwise a positive errno value. * diff --git a/lib/netdev.h b/lib/netdev.h index f8482f7..328a158 100644 --- a/lib/netdev.h +++ b/lib/netdev.h @@ -183,6 +183,7 @@ int netdev_rxq_drain(struct netdev_rxq *); int netdev_send(struct netdev *, int qid, struct dp_packet_batch *, bool may_steal, bool concurrent_txq); void netdev_send_wait(struct netdev *, int qid); +int netdev_txq_flush(struct netdev *, int qid, bool concurrent_txq); /* Flow offloading. */ struct offload_info {