From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by dpdk.org (Postfix) with ESMTP id DCC491B27C for ; Tue, 31 Oct 2017 02:07:22 +0100 (CET) Received: from orsmga005.jf.intel.com ([10.7.209.41]) by orsmga102.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 30 Oct 2017 18:07:22 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.44,321,1505804400"; d="scan'208";a="167110263" Received: from fyigit-mobl1.ger.corp.intel.com (HELO [10.241.225.254]) ([10.241.225.254]) by orsmga005.jf.intel.com with ESMTP; 30 Oct 2017 18:07:21 -0700 References: <1470084176-79932-1-git-send-email-rsanford@akamai.com> <1470084176-79932-2-git-send-email-rsanford@akamai.com> <9BB6961774997848B5B42BEC655768F810DAC7D9@SHSMSX103.ccr.corp.intel.com> To: Robert Sanford Cc: Jingjing Wu , declan.doherty@intel.com, pablo.de.lara.guarch@intel.com, olivier.matz@6wind.com, DPDK From: Ferruh Yigit Message-ID: Date: Mon, 30 Oct 2017 18:07:20 -0700 User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:52.0) Gecko/20100101 Thunderbird/52.4.0 MIME-Version: 1.0 In-Reply-To: <9BB6961774997848B5B42BEC655768F810DAC7D9@SHSMSX103.ccr.corp.intel.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 8bit Subject: Re: [dpdk-dev] [PATCH 1/4] testpmd: fix LACP ports to work with idle links 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: , X-List-Received-Date: Tue, 31 Oct 2017 01:07:23 -0000 On 6/21/2017 6:25 PM, jingjing.wu at intel.com (Wu, Jingjing) wrote: >> -----Original Message----- >> From: rsanford2 at gmail.com >> Sent: Tuesday, August 2, 2016 4:43 AM >> Subject: [dpdk-dev] [PATCH 1/4] testpmd: fix LACP ports to work with idle links >> >> Problem: When there is little or no TX traffic on an LACP port (bonding mode 4), >> we don't call its tx burst function in a timely manner, and thus we don't >> transmit LACPDUs when we should. >> >> Solution: Add and maintain an "lacp_master" flag in rte_port struct. >> In the main packet forwarding loop, if port is an LACP master, in 1 out of N >> loops force an empty call to the tx burst API. >> >> Signed-off-by: Robert Sanford >> --- >> app/test-pmd/cmdline.c | 9 +++++++++ >> app/test-pmd/testpmd.c | 37 +++++++++++++++++++++++++++++++++++++ >> app/test-pmd/testpmd.h | 4 ++++ >> 3 files changed, 50 insertions(+), 0 deletions(-) >> >> diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c index >> f90befc..2a629ee 100644 >> --- a/app/test-pmd/cmdline.c >> +++ b/app/test-pmd/cmdline.c >> @@ -3975,6 +3975,10 @@ static void cmd_set_bonding_mode_parsed(void >> *parsed_result, >> /* Set the bonding mode for the relevant port. */ >> if (0 != rte_eth_bond_mode_set(port_id, res->value)) >> printf("\t Failed to set bonding mode for port = %d.\n", port_id); >> + else if (res->value == BONDING_MODE_8023AD) >> + set_port_lacp_master(port_id); >> + else >> + clear_port_lacp_master(port_id); >> } >> >> cmdline_parse_token_string_t cmd_setbonding_mode_set = @@ -4408,6 >> +4412,11 @@ static void cmd_create_bonded_device_parsed(void >> *parsed_result, >> reconfig(port_id, res->socket); >> rte_eth_promiscuous_enable(port_id); >> ports[port_id].enabled = 1; >> + >> + if (res->mode == BONDING_MODE_8023AD) >> + set_port_lacp_master(port_id); >> + else >> + clear_port_lacp_master(port_id); >> } >> >> } >> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index >> 1428974..806667e 100644 >> --- a/app/test-pmd/testpmd.c >> +++ b/app/test-pmd/testpmd.c >> @@ -920,12 +920,28 @@ run_pkt_fwd_on_lcore(struct fwd_lcore *fc, >> packet_fwd_t pkt_fwd) >> struct fwd_stream **fsm; >> streamid_t nb_fs; >> streamid_t sm_id; >> + unsigned int loop_count = 0; >> >> fsm = &fwd_streams[fc->stream_idx]; >> nb_fs = fc->stream_nb; >> do { >> for (sm_id = 0; sm_id < nb_fs; sm_id++) >> (*pkt_fwd)(fsm[sm_id]); >> + >> + /* >> + * Per rte_eth_bond.h, we must invoke LACP master's tx >> + * burst function at least once every 100 ms. >> + */ >> + loop_count++; >> + if (likely(loop_count % 1024 != 0)) >> + continue; >> + for (sm_id = 0; sm_id < nb_fs; sm_id++) { >> + struct fwd_stream *fs = fsm[sm_id]; >> + >> + if (port_is_lacp_master(fs->tx_port)) >> + rte_eth_tx_burst(fs->tx_port, fs->tx_queue, >> + NULL, 0); >> + } > By add tx_burst here will change default logic. Some fwd engine are usually used to measure > Performance . Additional cycle in fast path may impact that. It would be better to move additional > sending to your fwd engine. Hi Robert, This patchset is older than a year and not responded back to latest comments. Is this still a valid patch? Do you have any objection to remove this from patchwork? Thanks, ferruh