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 45C55A04FD; Mon, 23 May 2022 07:12:39 +0200 (CEST) Received: from [217.70.189.124] (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id E27054014F; Mon, 23 May 2022 07:12:38 +0200 (CEST) Received: from mga12.intel.com (mga12.intel.com [192.55.52.136]) by mails.dpdk.org (Postfix) with ESMTP id 0DDD640041 for ; Mon, 23 May 2022 07:12:36 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1653282757; x=1684818757; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=9WlEn64x9fu43AIGt/7xvTY4r8tL0/wIOQhscRyEl8k=; b=jMNPutGxi5hkKCCMtzKR4fUYzJ4yUh9oCQD0u+e3UKHOX2uHHNL54GAr dsjNLTuQvKG2RvowSnl055m05MD4KHG/YseeZxeVu5ZpLRPg4e3bzVqkC oUpEo+D9yIZQrpxhfjUweK3G0t2H5CMDcJ5qzh3xvvY2EE7ldaY4iEuOg VvRjvVlomf0RIAUqt7cdJl2Fn7PmmYXU9jtfrOf+yc4mixj1WkVtpiTFx YQMDy30kld4vGGJUXW9rmy6q+vcO5tpQRnUWggrkz+4oORABjLuJqpJe7 m+/G3CnTEn9m7s20tez8GZQsHMxH1bFzbIFXFh/0HtmtrOh6tG3KROEJE Q==; X-IronPort-AV: E=McAfee;i="6400,9594,10355"; a="252979294" X-IronPort-AV: E=Sophos;i="5.91,245,1647327600"; d="scan'208";a="252979294" Received: from orsmga008.jf.intel.com ([10.7.209.65]) by fmsmga106.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 22 May 2022 22:12:36 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.91,245,1647327600"; d="scan'208";a="600445036" Received: from npg-wuwenjun-dpdk-01.sh.intel.com ([10.67.110.181]) by orsmga008.jf.intel.com with ESMTP; 22 May 2022 22:12:33 -0700 From: Wenjun Wu To: dev@dpdk.org, jingjing.wu@intel.com, beilei.xing@intel.com, qi.z.zhang@intel.com, simei.su@intel.com Cc: stephen@networkplumber.org, Wenjun Wu Subject: [PATCH v4] net/iavf: fix race condition for multi-cores Date: Mon, 23 May 2022 12:49:00 +0800 Message-Id: <20220523044900.1437947-1-wenjun1.wu@intel.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20220519011959.613966-1-wenjun1.wu@intel.com> References: <20220519011959.613966-1-wenjun1.wu@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 In multi-cores cases for RX timestamp offload, if packets arrive too fast, aq command to get phc time will be pended. This patch adds spinlock to fix this issue. To avoid phc time being frequently overwritten, move related variables to iavf_rx_queue structure, and each queue will handle timestamp calculation by itself. Fixes: 2a0a62f40af5 ("net/iavf: enable Rx timestamp on Flex Descriptor") Fixes: 33db16136e55 ("net/iavf: improve performance of Rx timestamp offload") Signed-off-by: Wenjun Wu --- v4: rebase to the latest commit. v3: remove restrictions on initialization. v2: move duplicated spinlock initialization. --- drivers/net/iavf/iavf.h | 5 ++-- drivers/net/iavf/iavf_ethdev.c | 11 ++----- drivers/net/iavf/iavf_rxtx.c | 55 ++++++++++++++++++++-------------- drivers/net/iavf/iavf_rxtx.h | 2 ++ drivers/net/iavf/iavf_vchnl.c | 15 ++++++---- 5 files changed, 47 insertions(+), 41 deletions(-) diff --git a/drivers/net/iavf/iavf.h b/drivers/net/iavf/iavf.h index 819510649a..f1c2daa06e 100644 --- a/drivers/net/iavf/iavf.h +++ b/drivers/net/iavf/iavf.h @@ -270,6 +270,7 @@ struct iavf_info { struct rte_eth_dev *eth_dev; uint32_t ptp_caps; + rte_spinlock_t phc_time_aq_lock; }; #define IAVF_MAX_PKT_TYPE 1024 @@ -315,8 +316,6 @@ struct iavf_adapter { bool closed; uint16_t fdir_ref_cnt; struct iavf_devargs devargs; - uint64_t phc_time; - uint64_t hw_time_update; }; /* IAVF_DEV_PRIVATE_TO */ @@ -482,5 +481,5 @@ int iavf_ipsec_crypto_request(struct iavf_adapter *adapter, uint8_t *resp_msg, size_t resp_msg_len); extern const struct rte_tm_ops iavf_tm_ops; int iavf_get_ptp_cap(struct iavf_adapter *adapter); -int iavf_get_phc_time(struct iavf_adapter *adapter); +int iavf_get_phc_time(struct iavf_rx_queue *rxq); #endif /* _IAVF_ETHDEV_H_ */ diff --git a/drivers/net/iavf/iavf_ethdev.c b/drivers/net/iavf/iavf_ethdev.c index a74056f0f1..7a3b37d5e3 100644 --- a/drivers/net/iavf/iavf_ethdev.c +++ b/drivers/net/iavf/iavf_ethdev.c @@ -1039,20 +1039,13 @@ iavf_dev_start(struct rte_eth_dev *dev) iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num, true); + rte_spinlock_init(&vf->phc_time_aq_lock); + if (iavf_start_queues(dev) != 0) { PMD_DRV_LOG(ERR, "enable queues failed"); goto err_mac; } - if (dev->data->dev_conf.rxmode.offloads & - RTE_ETH_RX_OFFLOAD_TIMESTAMP) { - if (iavf_get_phc_time(adapter)) { - PMD_DRV_LOG(ERR, "get physical time failed"); - goto err_mac; - } - adapter->hw_time_update = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); - } - return 0; err_mac: diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c index ff0c98ffc3..bb3d7ea434 100644 --- a/drivers/net/iavf/iavf_rxtx.c +++ b/drivers/net/iavf/iavf_rxtx.c @@ -904,6 +904,15 @@ iavf_dev_rx_queue_start(struct rte_eth_dev *dev, uint16_t rx_queue_id) RTE_ETH_QUEUE_STATE_STARTED; } + if (dev->data->dev_conf.rxmode.offloads & + RTE_ETH_RX_OFFLOAD_TIMESTAMP) { + if (iavf_get_phc_time(rxq)) { + PMD_DRV_LOG(ERR, "get physical time failed"); + return err; + } + rxq->hw_time_update = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); + } + return err; } @@ -1440,6 +1449,7 @@ iavf_recv_pkts_flex_rxd(void *rx_queue, uint64_t dma_addr; uint64_t pkt_flags; const uint32_t *ptype_tbl; + uint64_t ts_ns; nb_rx = 0; nb_hold = 0; @@ -1448,15 +1458,13 @@ iavf_recv_pkts_flex_rxd(void *rx_queue, rx_ring = rxq->rx_ring; ptype_tbl = rxq->vsi->adapter->ptype_tbl; - struct iavf_adapter *ad = rxq->vsi->adapter; - uint64_t ts_ns; - if (rxq->offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP) { uint64_t sw_cur_time = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); - if (sw_cur_time - ad->hw_time_update > 4) { - if (iavf_get_phc_time(ad)) + + if (sw_cur_time - rxq->hw_time_update > 4) { + if (iavf_get_phc_time(rxq)) PMD_DRV_LOG(ERR, "get physical time failed"); - ad->hw_time_update = sw_cur_time; + rxq->hw_time_update = sw_cur_time; } } @@ -1523,11 +1531,11 @@ iavf_recv_pkts_flex_rxd(void *rx_queue, pkt_flags = iavf_flex_rxd_error_to_pkt_flags(rx_stat_err0); if (iavf_timestamp_dynflag > 0) { - ts_ns = iavf_tstamp_convert_32b_64b(ad->phc_time, + ts_ns = iavf_tstamp_convert_32b_64b(rxq->phc_time, rte_le_to_cpu_32(rxd.wb.flex_ts.ts_high)); - ad->phc_time = ts_ns; - ad->hw_time_update = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); + rxq->phc_time = ts_ns; + rxq->hw_time_update = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); *RTE_MBUF_DYNFIELD(rxm, iavf_timestamp_dynfield_offset, @@ -1563,7 +1571,6 @@ iavf_recv_scattered_pkts_flex_rxd(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t rx_stat_err0; uint64_t dma_addr; uint64_t pkt_flags; - struct iavf_adapter *ad = rxq->vsi->adapter; uint64_t ts_ns; volatile union iavf_rx_desc *rx_ring = rxq->rx_ring; @@ -1572,10 +1579,11 @@ iavf_recv_scattered_pkts_flex_rxd(void *rx_queue, struct rte_mbuf **rx_pkts, if (rxq->offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP) { uint64_t sw_cur_time = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); - if (sw_cur_time - ad->hw_time_update > 4) { - if (iavf_get_phc_time(ad)) + + if (sw_cur_time - rxq->hw_time_update > 4) { + if (iavf_get_phc_time(rxq)) PMD_DRV_LOG(ERR, "get physical time failed"); - ad->hw_time_update = sw_cur_time; + rxq->hw_time_update = sw_cur_time; } } @@ -1692,11 +1700,11 @@ iavf_recv_scattered_pkts_flex_rxd(void *rx_queue, struct rte_mbuf **rx_pkts, pkt_flags = iavf_flex_rxd_error_to_pkt_flags(rx_stat_err0); if (iavf_timestamp_dynflag > 0) { - ts_ns = iavf_tstamp_convert_32b_64b(ad->phc_time, + ts_ns = iavf_tstamp_convert_32b_64b(rxq->phc_time, rte_le_to_cpu_32(rxd.wb.flex_ts.ts_high)); - ad->phc_time = ts_ns; - ad->hw_time_update = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); + rxq->phc_time = ts_ns; + rxq->hw_time_update = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); *RTE_MBUF_DYNFIELD(first_seg, iavf_timestamp_dynfield_offset, @@ -1899,7 +1907,6 @@ iavf_rx_scan_hw_ring_flex_rxd(struct iavf_rx_queue *rxq, int32_t nb_staged = 0; uint64_t pkt_flags; const uint32_t *ptype_tbl = rxq->vsi->adapter->ptype_tbl; - struct iavf_adapter *ad = rxq->vsi->adapter; uint64_t ts_ns; rxdp = (volatile union iavf_rx_flex_desc *)&rxq->rx_ring[rxq->rx_tail]; @@ -1913,10 +1920,11 @@ iavf_rx_scan_hw_ring_flex_rxd(struct iavf_rx_queue *rxq, if (rxq->offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP) { uint64_t sw_cur_time = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); - if (sw_cur_time - ad->hw_time_update > 4) { - if (iavf_get_phc_time(ad)) + + if (sw_cur_time - rxq->hw_time_update > 4) { + if (iavf_get_phc_time(rxq)) PMD_DRV_LOG(ERR, "get physical time failed"); - ad->hw_time_update = sw_cur_time; + rxq->hw_time_update = sw_cur_time; } } @@ -1977,11 +1985,12 @@ iavf_rx_scan_hw_ring_flex_rxd(struct iavf_rx_queue *rxq, pkt_flags = iavf_flex_rxd_error_to_pkt_flags(stat_err0); if (iavf_timestamp_dynflag > 0) { - ts_ns = iavf_tstamp_convert_32b_64b(ad->phc_time, + ts_ns = iavf_tstamp_convert_32b_64b(rxq->phc_time, rte_le_to_cpu_32(rxdp[j].wb.flex_ts.ts_high)); - ad->phc_time = ts_ns; - ad->hw_time_update = rte_get_timer_cycles() / (rte_get_timer_hz() / 1000); + rxq->phc_time = ts_ns; + rxq->hw_time_update = rte_get_timer_cycles() / + (rte_get_timer_hz() / 1000); *RTE_MBUF_DYNFIELD(mb, iavf_timestamp_dynfield_offset, diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h index e8362bbd1d..1695e43cd5 100644 --- a/drivers/net/iavf/iavf_rxtx.h +++ b/drivers/net/iavf/iavf_rxtx.h @@ -223,6 +223,8 @@ struct iavf_rx_queue { /* flexible descriptor metadata extraction offload flag */ struct iavf_rx_queue_stats stats; uint64_t offloads; + uint64_t phc_time; + uint64_t hw_time_update; }; struct iavf_tx_entry { diff --git a/drivers/net/iavf/iavf_vchnl.c b/drivers/net/iavf/iavf_vchnl.c index 0520b1045f..21bd1e2193 100644 --- a/drivers/net/iavf/iavf_vchnl.c +++ b/drivers/net/iavf/iavf_vchnl.c @@ -1913,12 +1913,13 @@ iavf_get_ptp_cap(struct iavf_adapter *adapter) } int -iavf_get_phc_time(struct iavf_adapter *adapter) +iavf_get_phc_time(struct iavf_rx_queue *rxq) { + struct iavf_adapter *adapter = rxq->vsi->adapter; struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter); struct virtchnl_phc_time phc_time; struct iavf_cmd_info args; - int err; + int err = 0; args.ops = VIRTCHNL_OP_1588_PTP_GET_TIME; args.in_args = (uint8_t *)&phc_time; @@ -1926,14 +1927,16 @@ iavf_get_phc_time(struct iavf_adapter *adapter) args.out_buffer = vf->aq_resp; args.out_size = IAVF_AQ_BUF_SZ; + rte_spinlock_lock(&vf->phc_time_aq_lock); err = iavf_execute_vf_cmd(adapter, &args, 0); if (err) { PMD_DRV_LOG(ERR, "Failed to execute command of VIRTCHNL_OP_1588_PTP_GET_TIME"); - return err; + goto out; } + rxq->phc_time = ((struct virtchnl_phc_time *)args.out_buffer)->time; - adapter->phc_time = ((struct virtchnl_phc_time *)args.out_buffer)->time; - - return 0; +out: + rte_spinlock_unlock(&vf->phc_time_aq_lock); + return err; } -- 2.25.1