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 1478BA0524; Thu, 4 Feb 2021 17:29:04 +0100 (CET) Received: from [217.70.189.124] (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id CCE6D2406B6; Thu, 4 Feb 2021 17:29:03 +0100 (CET) Received: from mga12.intel.com (mga12.intel.com [192.55.52.136]) by mails.dpdk.org (Postfix) with ESMTP id 4CCBD2406A0; Thu, 4 Feb 2021 17:29:02 +0100 (CET) IronPort-SDR: /pTGxeYG+6OnGoxz+STOIyY+yeGFXsxTBmBsZPvZQW2n3HF6Zj3CzXjdvhj0ACcwv3o7QGPVOG Rne7BdqHw9UA== X-IronPort-AV: E=McAfee;i="6000,8403,9885"; a="160437705" X-IronPort-AV: E=Sophos;i="5.79,401,1602572400"; d="scan'208";a="160437705" Received: from orsmga008.jf.intel.com ([10.7.209.65]) by fmsmga106.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 04 Feb 2021 08:28:45 -0800 IronPort-SDR: AXBiI0ZZSEQultnVXIgD0kKDu5GuZbB/3f5vmLEN0mYxO8oujJkMsEAOA8sQawI5ao9u8onp+w M+joDggb+cPA== X-IronPort-AV: E=Sophos;i="5.79,401,1602572400"; d="scan'208";a="393262592" Received: from fyigit-mobl1.ger.corp.intel.com (HELO [10.213.211.210]) ([10.213.211.210]) by orsmga008-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 04 Feb 2021 08:28:44 -0800 To: "Ferriter, Cian" Cc: "dev@dpdk.org" , "stable@dpdk.org" References: <20210203154920.2449179-1-ferruh.yigit@intel.com> From: Ferruh Yigit Message-ID: <57f7221e-b90a-5f06-d732-fa6eb8465165@intel.com> Date: Thu, 4 Feb 2021 16:28:41 +0000 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=utf-8; format=flowed Content-Language: en-US Content-Transfer-Encoding: 8bit Subject: Re: [dpdk-dev] [PATCH] net/pcap: fix infinite Rx with large files 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 Sender: "dev" On 2/4/2021 4:03 PM, Ferriter, Cian wrote: > Hi Ferruh, > > This fixes the issue I was seeing. Now an error is reported, rather than silent failure. > > I have one piece of feedback about the particular error message below inline which you can take or leave, I'm happy for you to upstream this fix either way. > > Acked-by: Cian Ferriter > >> -----Original Message----- >> From: Yigit, Ferruh >> Sent: Wednesday 3 February 2021 15:49 >> To: Ferriter, Cian >> Cc: Yigit, Ferruh ; dev@dpdk.org; stable@dpdk.org >> Subject: [PATCH] net/pcap: fix infinite Rx with large files >> >> Packet forwarding is not working when infinite Rx feature is used with >> large .pcap files that has high number of packets. >> >> The problem is number of allocated mbufs are less than the infinite Rx >> ring size, and all mbufs consumed to fill the ring, so there is no mbuf >> left for forwarding. >> >> Current logic can not detect that infinite Rx ring is not filled >> completely and no more mbufs left, and setup continues which leads >> silent fail on packet forwarding. >> >> There isn't much can be done when there is not enough mbuf for the given >> .pcap file, so additional checks added to detect the case and fail >> explicitly with an error log. >> >> Bugzilla ID: 595 >> Fixes: a3f5252e5cbd ("net/pcap: enable infinitely Rx a pcap file") >> Cc: stable@dpdk.org >> >> Reported-by: Cian Ferriter >> Signed-off-by: Ferruh Yigit >> --- >> drivers/net/pcap/rte_eth_pcap.c | 40 ++++++++++++++++++++------------- >> 1 file changed, 25 insertions(+), 15 deletions(-) >> >> diff --git a/drivers/net/pcap/rte_eth_pcap.c >> b/drivers/net/pcap/rte_eth_pcap.c >> index ff02ade70d1a..98f80368ca1d 100644 >> --- a/drivers/net/pcap/rte_eth_pcap.c >> +++ b/drivers/net/pcap/rte_eth_pcap.c >> @@ -735,6 +735,17 @@ eth_stats_reset(struct rte_eth_dev *dev) >> return 0; >> } >> >> +static inline void >> +infinite_rx_ring_free(struct rte_ring *pkts) >> +{ >> +struct rte_mbuf *bufs; >> + >> +while (!rte_ring_dequeue(pkts, (void **)&bufs)) >> +rte_pktmbuf_free(bufs); >> + >> +rte_ring_free(pkts); >> +} >> + >> static int >> eth_dev_close(struct rte_eth_dev *dev) >> { >> @@ -753,7 +764,6 @@ eth_dev_close(struct rte_eth_dev *dev) >> if (internals->infinite_rx) { >> for (i = 0; i < dev->data->nb_rx_queues; i++) { >> struct pcap_rx_queue *pcap_q = &internals- >>> rx_queue[i]; >> -struct rte_mbuf *pcap_buf; >> >> /* >> * 'pcap_q->pkts' can be NULL if 'eth_dev_close()' >> @@ -762,11 +772,7 @@ eth_dev_close(struct rte_eth_dev *dev) >> if (pcap_q->pkts == NULL) >> continue; >> >> -while (!rte_ring_dequeue(pcap_q->pkts, >> -(void **)&pcap_buf)) >> -rte_pktmbuf_free(pcap_buf); >> - >> -rte_ring_free(pcap_q->pkts); >> +infinite_rx_ring_free(pcap_q->pkts); >> } >> } >> >> @@ -835,21 +841,25 @@ eth_rx_queue_setup(struct rte_eth_dev *dev, >> while (eth_pcap_rx(pcap_q, bufs, 1)) { >> /* Check for multiseg mbufs. */ >> if (bufs[0]->nb_segs != 1) { >> -rte_pktmbuf_free(*bufs); >> - >> -while (!rte_ring_dequeue(pcap_q->pkts, >> -(void **)bufs)) >> -rte_pktmbuf_free(*bufs); >> - >> -rte_ring_free(pcap_q->pkts); >> -PMD_LOG(ERR, "Multiseg mbufs are not >> supported in infinite_rx " >> -"mode."); >> +infinite_rx_ring_free(pcap_q->pkts); >> +PMD_LOG(ERR, >> +"Multiseg mbufs are not supported in >> infinite_rx mode."); >> return -EINVAL; >> } >> >> rte_ring_enqueue_bulk(pcap_q->pkts, >> (void * const *)bufs, 1, NULL); >> } >> + >> +if (rte_ring_count(pcap_q->pkts) < pcap_pkt_count) { >> +infinite_rx_ring_free(pcap_q->pkts); >> +PMD_LOG(ERR, >> +"Not enough mbuf to fill the infinite_rx ring. " >> +"At least %" PRIu64 " mbufs per queue is >> required to fill the ring", >> +pcap_pkt_count); > > [Cian Ferriter] > So we can say that the issue is either too many packets in the PCAP or too few mbufs for the ring. What can the user do about this? > They can use a PCAP with less packets. > Can they change how many mbufs are available by passing more memory or any other method? > > Should be mention these remedies, or is this outside the scope for an error message? User can change the number of allocated mbuf easily, like this is done in the testpmd via '--total-num-mbufs=N' command in testpmd. Assuming user would like to use the bigger .pcap file, I go with too few mbufs for the ring. But "infinite_rx ring" can be implementation detail, is following more clear message: "Not enough mbufs to accommodate packets in pcap file. At least %" PRIu64 " mbufs per queue is required." > > As I mentioned, I'm happy for you to upstream either way. > >> +return -EINVAL; >> +} >> + >> /* >> * Reset the stats for this queue since eth_pcap_rx calls >> above >> * didn't result in the application receiving packets. >> -- >> 2.29.2 >