From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from NAM01-BN3-obe.outbound.protection.outlook.com (mail-bn3nam01on0058.outbound.protection.outlook.com [104.47.33.58]) by dpdk.org (Postfix) with ESMTP id 6910B1B13B for ; Thu, 20 Sep 2018 00:23:14 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=CAVIUMNETWORKS.onmicrosoft.com; s=selector1-cavium-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=mAk+3q6Z/6WeTIg6Nys4icAbhP/3AQpQB+LTMq6R+Ck=; b=hbLKicsPm9jiR41fU5iF42CZ3QGD30fRvqbEGWa0yBCm897S6jyay0ZZ9kFDn3XKSGqEUlBGs4swSHV5s0lf9+OZ5HOeOve0Zz2wqTRQl9veF6AGcWFcYbHv/osQRZ30qYZqU8hgtSxB5JmqjIUHxIglvTcwGszlbRE3qI22zaM= Received: from localhost.localdomain (115.113.156.3) by BYAPR07MB4965.namprd07.prod.outlook.com (2603:10b6:a03:5b::14) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.1143.18; Wed, 19 Sep 2018 22:23:10 +0000 From: Pavan Nikhilesh To: jerin.jacob@caviumnetworks.com, nikhil.rao@intel.com, anoob.joseph@caviumnetworks.com Cc: dev@dpdk.org, Pavan Nikhilesh Date: Thu, 20 Sep 2018 03:52:34 +0530 Message-Id: <20180919222235.6239-3-pbhagavatula@caviumnetworks.com> X-Mailer: git-send-email 2.18.0 In-Reply-To: <20180919222235.6239-1-pbhagavatula@caviumnetworks.com> References: <20180904141223.24216-1-pbhagavatula@caviumnetworks.com> <20180919222235.6239-1-pbhagavatula@caviumnetworks.com> MIME-Version: 1.0 Content-Type: text/plain X-Originating-IP: [115.113.156.3] X-ClientProxiedBy: MA1PR0101CA0003.INDPRD01.PROD.OUTLOOK.COM (2603:1096:a00:21::13) To BYAPR07MB4965.namprd07.prod.outlook.com (2603:10b6:a03:5b::14) X-MS-PublicTrafficType: Email X-MS-Office365-Filtering-Correlation-Id: f8f354a8-3dc1-456d-d221-08d61e7e7c9e X-Microsoft-Antispam: BCL:0; PCL:0; RULEID:(7020095)(4652040)(8989299)(5600074)(711020)(2017052603328)(7153060)(7193020); SRVR:BYAPR07MB4965; X-Microsoft-Exchange-Diagnostics: 1; BYAPR07MB4965; 3:2K+wF4qc0Jg/oVGnL5mOFBk8Om29P2ZH2HlakoMdI9nU2irmcHPz5sNTB6ypzkIJEteKcqbpbQVJvPTveo5wD8X3hBcH1ortVrovMAd6camMWJa6iFZXfQrsg0jOav6AqlDaLE8v5QYpgWx1iUYhoCafh4SayDTV8KXrMT3zoifjR9H9TKnZByQcZ2G+43RGXKDZVrp0GgdDZcIu45zQNVY8x0uOL0/1twc5+QSnyrrAnd0srgN6MmKs2Fxsz81M; 25:VELPmH4sm9gcKPwPSDHZLv/Kqu2PyiaCk9gc2KNbNZoQwc1uMyvCj14NmeuAKYqDziWxgghxx24ELCRsh3+eGBUN6U9vdlRMgIjoIqxl1uorZ5ZxdOX4xEgySVsIgUOf8e0z90UFwpRpg1qSjCBjulXbj8P4xvOKs6zKFFNY7aqp66mRL9/jZo1FTC3rOWIW3u1y4QyHzZfmG2FUc9ALDo51jCafZOxvypUYfKz+4U11Jkufd6NzmDrbiewOkrtOq5JZh27I6H5wShjvIx5cJA/+WQFxy6ooQzT0LkqyKos9vfRoU/hcQ21gGd8VqdpawB+nSqe+fya8YOVSxmmHDQ==; 31:2GIxd2szZodh8Czu4ZAkaaNfy9iNQSDgweL3oTMeShmVXLo/NTKSKiZTKiobbSWvm/pXgJ+rnPW1ZbZgN+uuipmEwdY69Tt0QrZogoSXFU/XPYcZZ+y8ha6/zdUOTNniVsw5MVpKCy3Rgod/9ZiZc1Ebhhwc0WKrfosnqTbip0u75qmZodSnMPbJ/eQ/r2mcstYiKuTk94iwYKuQmvr/41Z/Fhtky8pip82XCbUw7qA= X-MS-TrafficTypeDiagnostic: BYAPR07MB4965: X-Microsoft-Exchange-Diagnostics: 1; BYAPR07MB4965; 20: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; 4:oNb1787t0JE70WUIKjcb4SweBnjTqT3miAOxO/o7V4Yp+3/OO0XzYZjvi8SQ2Wub2VvEXW8bWAblJJbvWUPakkTcIeFWP4ZAdvTi3laQ+fbbf4SiGjKhPXCMLsLKKteM0yArZCqS9r53ffMI6+VWk1ICAT6tya3uWJmNKLAiVdpop8fUHt9Yj/Gv14huQt6vlYsmBo55fP62Arn80Wc8GC3rDj2tRW9YxrTYzF3OTFCKl99hcD5tE4e+5ka7VV/ck+Ugh/3w8yCYK5GyA/02f9ZeKrH34p8UkVpuvFoRxZr7Bt91BDBX5sAtZe1Q5S43 X-Microsoft-Antispam-PRVS: X-Exchange-Antispam-Report-Test: UriScan:(21532816269658); X-MS-Exchange-SenderADCheck: 1 X-Exchange-Antispam-Report-CFA-Test: BCL:0; PCL:0; RULEID:(8211001083)(6040522)(2401047)(5005006)(8121501046)(93006095)(3002001)(10201501046)(3231355)(944501410)(52105095)(149027)(150027)(6041310)(20161123562045)(201703131423095)(201702281528075)(20161123555045)(201703061421075)(201703061406153)(20161123564045)(20161123560045)(20161123558120)(201708071742011)(7699051)(76991041); SRVR:BYAPR07MB4965; BCL:0; PCL:0; RULEID:; SRVR:BYAPR07MB4965; X-Forefront-PRVS: 0800C0C167 X-Forefront-Antispam-Report: SFV:NSPM; SFS:(10009020)(6069001)(39860400002)(346002)(376002)(136003)(396003)(366004)(189003)(199004)(66066001)(6636002)(575784001)(53946003)(6512007)(81166006)(81156014)(6666003)(53936002)(48376002)(4326008)(1857600001)(316002)(50226002)(486006)(36756003)(8936002)(446003)(11346002)(16526019)(26005)(476003)(956004)(8676002)(2616005)(2906002)(14444005)(42882007)(6346003)(5660300001)(68736007)(16586007)(97736004)(47776003)(7736002)(107886003)(6486002)(72206003)(478600001)(52116002)(105586002)(50466002)(51416003)(1076002)(3846002)(6506007)(6116002)(69590400006)(386003)(25786009)(305945005)(106356001)(76176011)(42262002)(579004); DIR:OUT; SFP:1101; SCL:1; SRVR:BYAPR07MB4965; H:localhost.localdomain; FPR:; SPF:None; LANG:en; PTR:InfoNoRecords; MX:1; A:1; Received-SPF: None (protection.outlook.com: cavium.com does not designate permitted sender hosts) Authentication-Results: spf=none (sender IP is ) smtp.mailfrom=Pavan.Bhagavatula@cavium.com; X-Microsoft-Exchange-Diagnostics: =?us-ascii?Q?1; BYAPR07MB4965; 23:i4bXtuI5dNob7+yiVSyzy9xsIPucYvrmIU9W3avXc?= =?us-ascii?Q?5DtYqt82Skpk1M5DEt8yKD3vx8T9y9KRTwTnwHz9tN2ZlErJu1u8frNLGANg?= =?us-ascii?Q?Y5ynCkhz2NH7ATuY+AW+nqNXZI+/z6+W7lYHug8BP0OOcsTQETFn+h6pgw1A?= =?us-ascii?Q?u3BEbtrOVdx49CdE8JGxVKKwEE/lOiDT+Vbagcf1eIrPkJRBWbo30wLCZz4C?= =?us-ascii?Q?pTgZuIhBU3v1WCBoF1XPHbSXKJhQfn9W4qNhv6bbqB85aB9YWkeZn9JaVipv?= =?us-ascii?Q?NiK0P6mWC/hmUebzZGnG7zKg1ZjevUKFCfVZauTPCpldQPuvD9mRy/CpshXa?= =?us-ascii?Q?vGeB22jNfaq4J5GXtXhFaI23It4QGP5JcaM/zCgDHWGPyiIwDZrcgELXZesI?= =?us-ascii?Q?ZfuRnHq3J+iRwguSoQm3+6doGGEQjPJ9UyLJ8x/Jo1zB+aIADRmOgErRp05W?= =?us-ascii?Q?HIpJQLheapuMDVgVSRAg5wKn52APbhf1ShhRuqBWNV2k8JPJeP70lPIeDilH?= =?us-ascii?Q?Bikg1SIG1E2bPcCdoeRQE9Sl3/xsjRmG3tPImfmTtENomg443BVE8a7TrSI3?= =?us-ascii?Q?U1UCdo9cqLxxJSd3NVbfDikFC7qyXNDeOdKJBbm9FXloY/HR2Tab8kGjPWum?= =?us-ascii?Q?mZVBfKHlQWA6JcRbvd/7fSQ9yUsXRevRvGvg0kzR+bOu9/3RHTcKvz453jLQ?= =?us-ascii?Q?8KHmoEgaHi4YTwVGFKkPu4r2qB3YPLQMvuTiU2vMLDNFD+TbXPnAeaGjcDI6?= =?us-ascii?Q?HXQpxQQJPlOu8Vliq1Jx1pROQsBatIUFJY/dfLoNBw2U/HS9HUUPNZK4b9JN?= =?us-ascii?Q?V8IOnfOenoDDJ4twxQIznoy78wVqM5VgpMRqKnrHM4tXUPtanLFXoy0vcA6r?= =?us-ascii?Q?zjstB0kcUkRaYlkBefuJlHC126vSnpiUN0oUpDL4nNO6EJYmYenEo77DlAZG?= =?us-ascii?Q?0VBMqAw0uSqi716orS30QaCoQav5Hb3PHaxiqsqxK1ZGltYINSbg3jkfzdj7?= =?us-ascii?Q?+P+faqQUwH8nJ5YdVZCA5juL6u8LsugRLfZ+8fcibyhTeFUGOYM2z3qXfJYR?= =?us-ascii?Q?PjCNVNhRGzywEjPUQ7Fgt1HIK5RcDJnz8TDD+oRFBTnp++D7tcayFi6Atpz3?= =?us-ascii?Q?bPv+Ij34MqlI/073Axo00FTdiakOm5WaKebG7+zKE47Jn/+6AlRLcV6EM4SW?= =?us-ascii?Q?3eZXuB96by6Vg8EhA74wqsfI46OlQGwQfIWIZWtm4czcqA0bJpMBqivnLSyF?= =?us-ascii?Q?ni4CdcB+gdvwjllgYi+ePhYU2r/W5EOPsNJY78TKf3eDv3gUMEDi34hvzT9B?= =?us-ascii?Q?FMcCFU4PM4qpxJzhCtiwkXfNuGtQ7FDUMAnXgn+tMr+QkxP2fXEzTAdVPgtY?= =?us-ascii?Q?UnvhpHSwuSMjSmnC7uAKJvmGtFpPS4urRhg6LehnNMRHKkc?= X-Microsoft-Antispam-Message-Info: ujRXT7Im9uIhqtnQjwhVuzqU6+XoevAkLyNp72+Faw4N/041KbJlHyo97aOyRtmt750IxP00F0t3r0xlWguqOzGsBfZwyeQVmFqt/0QBbzYXIlPt5Xu5gZTZYUZKWdGNmUcgFG6J0oQnVwEujjzDP1qq47Sr4fIUqPdOleHvjV5wuUgWIaDKC6XN5+Htx+D9OEgOeXPuRNIXD8NbAie6yxi6hJKArSFHtwBpH9XHgO9n8gvYeqjtr84FAKOUmwRGan5XRgrcZBhLDee3Wq2p7MaUyqK6/r6uDJNySL6dGy5FEDKMrnWQU8brliHEEvLWGmhd40TN5pSCxUelpfrI2IqUAjgzBi0GnOJojCvLLXg= X-Microsoft-Exchange-Diagnostics: 1; BYAPR07MB4965; 6:3KwkdCRaD75Srazw0j8l1pDH5B/5IQSBSTxHbPl9W9bCg2sU0Fk1a1PWHUtz2mXJns6JJrKiQrzAZGy8sdSJiZfaOkgVwEnFRIEPfHXcfDNPu3sZqgDIdVIl5qRZRK5TVGywQGcIS3jpwykjPD3WoTGW3cy8nHtsaxguJvTDZgiKoXmmQkPWWeB5uYk4qjVbbQDYXGxITBtN0pwDA0hqGVBKJGl/sVnR49EV0KQfjBvGyzkKy+eiYru/dv9K+zCiIbOry348Ld4wM6kGPJ6zVfJ0DAV/f4nNnM6CKUUhPIdxgWBMLYd0uG4us6r0vsKTDNatR0UiByVLqP9F5EfnFWg7yjvi6yGeza/EmFeKWaMLWf9B+jhVtER05ustPBUsdp2gS80Vm6jUlG+d5IUBGSuwVH9MqiFjLVbHMQHijfcdwMtQ+Xyzzp+cav/UDZL5tRp/SbM4hjrmfjJsSY0ZEg==; 5:LS9UwoMcvnQJmInr8McL3sFDoHjgzGD30Df/fx1bpcYEBt8FT6asEuohlpckCU+9jNlSGiVI7D/RFoRV45gZZc7dPX9moA2IkuS8GVrTEGPUV/BpqqL2bhlor/6f+Sg6FAoqiA0qVSu59SFWL8bOdfKrF2/jzrO+18gGCQY3NHg=; 7:oXlIRxZrjGJqh0CqjznwdWnbCFgqdIw99/gbT+mNx5+lZ/XBt7SbGU83B990MpwtkBAZ/vUDbRNoDHxAjfw951CXyOLWxXP5o7XBSuvlOgG+2vfKhMiFj8URDzJD1Tb15Lb836CNOlhuhD7jOhvztW8FbCM/OXWeY01DVvfCq1Nf1ZLRM/oA1+1UGwrT5WK+faDRyDY7N/lbx2SM4vg3ZxBrt472D0c6ofjdxN1qC8BFsrkCJ3uCyAENBB6WV9t1 SpamDiagnosticOutput: 1:99 SpamDiagnosticMetadata: NSPM X-OriginatorOrg: caviumnetworks.com X-MS-Exchange-CrossTenant-OriginalArrivalTime: 19 Sep 2018 22:23:10.8124 (UTC) X-MS-Exchange-CrossTenant-Network-Message-Id: f8f354a8-3dc1-456d-d221-08d61e7e7c9e X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 711e4ccf-2e9b-4bcf-a551-4094005b6194 X-MS-Exchange-Transport-CrossTenantHeadersStamped: BYAPR07MB4965 Subject: [dpdk-dev] [PATCH v3 3/4] app/test-eventdev: add Tx adapter support 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: Wed, 19 Sep 2018 22:23:15 -0000 Convert existing Tx service based pipeline to Tx adapter based APIs and simplify worker functions. Signed-off-by: Pavan Nikhilesh --- app/test-eventdev/test_pipeline_atq.c | 269 ++++++++++++----------- app/test-eventdev/test_pipeline_common.c | 202 +++++------------ app/test-eventdev/test_pipeline_common.h | 62 +++--- app/test-eventdev/test_pipeline_queue.c | 241 ++++++++++---------- 4 files changed, 363 insertions(+), 411 deletions(-) diff --git a/app/test-eventdev/test_pipeline_atq.c b/app/test-eventdev/test_pipeline_atq.c index f0b2f9015..01af298f3 100644 --- a/app/test-eventdev/test_pipeline_atq.c +++ b/app/test-eventdev/test_pipeline_atq.c @@ -15,7 +15,7 @@ pipeline_atq_nb_event_queues(struct evt_options *opt) return rte_eth_dev_count_avail(); } -static int +static __rte_noinline int pipeline_atq_worker_single_stage_tx(void *arg) { PIPELINE_WORKER_SINGLE_STAGE_INIT; @@ -28,23 +28,18 @@ pipeline_atq_worker_single_stage_tx(void *arg) continue; } - if (ev.sched_type == RTE_SCHED_TYPE_ATOMIC) { - pipeline_tx_pkt(ev.mbuf); - w->processed_pkts++; - continue; - } - pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC); - pipeline_event_enqueue(dev, port, &ev); + pipeline_event_tx(dev, port, &ev); + w->processed_pkts++; } return 0; } -static int +static __rte_noinline int pipeline_atq_worker_single_stage_fwd(void *arg) { PIPELINE_WORKER_SINGLE_STAGE_INIT; - const uint8_t tx_queue = t->tx_service.queue_id; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t event = rte_event_dequeue_burst(dev, port, &ev, 1, 0); @@ -54,16 +49,16 @@ pipeline_atq_worker_single_stage_fwd(void *arg) continue; } - w->processed_pkts++; - ev.queue_id = tx_queue; + ev.queue_id = tx_queue[ev.mbuf->port]; pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC); pipeline_event_enqueue(dev, port, &ev); + w->processed_pkts++; } return 0; } -static int +static __rte_noinline int pipeline_atq_worker_single_stage_burst_tx(void *arg) { PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT; @@ -79,27 +74,21 @@ pipeline_atq_worker_single_stage_burst_tx(void *arg) for (i = 0; i < nb_rx; i++) { rte_prefetch0(ev[i + 1].mbuf); - if (ev[i].sched_type == RTE_SCHED_TYPE_ATOMIC) { - - pipeline_tx_pkt(ev[i].mbuf); - ev[i].op = RTE_EVENT_OP_RELEASE; - w->processed_pkts++; - } else - pipeline_fwd_event(&ev[i], - RTE_SCHED_TYPE_ATOMIC); + rte_event_eth_tx_adapter_txq_set(ev[i].mbuf, 0); } - pipeline_event_enqueue_burst(dev, port, ev, nb_rx); + pipeline_event_tx_burst(dev, port, ev, nb_rx); + w->processed_pkts += nb_rx; } return 0; } -static int +static __rte_noinline int pipeline_atq_worker_single_stage_burst_fwd(void *arg) { PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT; - const uint8_t tx_queue = t->tx_service.queue_id; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t nb_rx = rte_event_dequeue_burst(dev, port, ev, @@ -112,23 +101,22 @@ pipeline_atq_worker_single_stage_burst_fwd(void *arg) for (i = 0; i < nb_rx; i++) { rte_prefetch0(ev[i + 1].mbuf); - ev[i].queue_id = tx_queue; + rte_event_eth_tx_adapter_txq_set(ev[i].mbuf, 0); + ev[i].queue_id = tx_queue[ev[i].mbuf->port]; pipeline_fwd_event(&ev[i], RTE_SCHED_TYPE_ATOMIC); - w->processed_pkts++; } pipeline_event_enqueue_burst(dev, port, ev, nb_rx); + w->processed_pkts += nb_rx; } return 0; } -static int +static __rte_noinline int pipeline_atq_worker_multi_stage_tx(void *arg) { PIPELINE_WORKER_MULTI_STAGE_INIT; - const uint8_t nb_stages = t->opt->nb_stages; - while (t->done == false) { uint16_t event = rte_event_dequeue_burst(dev, port, &ev, 1, 0); @@ -141,29 +129,24 @@ pipeline_atq_worker_multi_stage_tx(void *arg) cq_id = ev.sub_event_type % nb_stages; if (cq_id == last_queue) { - if (ev.sched_type == RTE_SCHED_TYPE_ATOMIC) { - - pipeline_tx_pkt(ev.mbuf); - w->processed_pkts++; - continue; - } - pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC); - } else { - ev.sub_event_type++; - pipeline_fwd_event(&ev, sched_type_list[cq_id]); + pipeline_event_tx(dev, port, &ev); + w->processed_pkts++; + continue; } + ev.sub_event_type++; + pipeline_fwd_event(&ev, sched_type_list[cq_id]); pipeline_event_enqueue(dev, port, &ev); } + return 0; } -static int +static __rte_noinline int pipeline_atq_worker_multi_stage_fwd(void *arg) { PIPELINE_WORKER_MULTI_STAGE_INIT; - const uint8_t nb_stages = t->opt->nb_stages; - const uint8_t tx_queue = t->tx_service.queue_id; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t event = rte_event_dequeue_burst(dev, port, &ev, 1, 0); @@ -176,9 +159,9 @@ pipeline_atq_worker_multi_stage_fwd(void *arg) cq_id = ev.sub_event_type % nb_stages; if (cq_id == last_queue) { - w->processed_pkts++; - ev.queue_id = tx_queue; + ev.queue_id = tx_queue[ev.mbuf->port]; pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC); + w->processed_pkts++; } else { ev.sub_event_type++; pipeline_fwd_event(&ev, sched_type_list[cq_id]); @@ -186,14 +169,14 @@ pipeline_atq_worker_multi_stage_fwd(void *arg) pipeline_event_enqueue(dev, port, &ev); } + return 0; } -static int +static __rte_noinline int pipeline_atq_worker_multi_stage_burst_tx(void *arg) { PIPELINE_WORKER_MULTI_STAGE_BURST_INIT; - const uint8_t nb_stages = t->opt->nb_stages; while (t->done == false) { uint16_t nb_rx = rte_event_dequeue_burst(dev, port, ev, @@ -209,34 +192,27 @@ pipeline_atq_worker_multi_stage_burst_tx(void *arg) cq_id = ev[i].sub_event_type % nb_stages; if (cq_id == last_queue) { - if (ev[i].sched_type == RTE_SCHED_TYPE_ATOMIC) { - - pipeline_tx_pkt(ev[i].mbuf); - ev[i].op = RTE_EVENT_OP_RELEASE; - w->processed_pkts++; - continue; - } - - pipeline_fwd_event(&ev[i], - RTE_SCHED_TYPE_ATOMIC); - } else { - ev[i].sub_event_type++; - pipeline_fwd_event(&ev[i], - sched_type_list[cq_id]); + pipeline_event_tx(dev, port, &ev[i]); + ev[i].op = RTE_EVENT_OP_RELEASE; + w->processed_pkts++; + continue; } + + ev[i].sub_event_type++; + pipeline_fwd_event(&ev[i], sched_type_list[cq_id]); } pipeline_event_enqueue_burst(dev, port, ev, nb_rx); } + return 0; } -static int +static __rte_noinline int pipeline_atq_worker_multi_stage_burst_fwd(void *arg) { PIPELINE_WORKER_MULTI_STAGE_BURST_INIT; - const uint8_t nb_stages = t->opt->nb_stages; - const uint8_t tx_queue = t->tx_service.queue_id; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t nb_rx = rte_event_dequeue_burst(dev, port, ev, @@ -253,7 +229,7 @@ pipeline_atq_worker_multi_stage_burst_fwd(void *arg) if (cq_id == last_queue) { w->processed_pkts++; - ev[i].queue_id = tx_queue; + ev[i].queue_id = tx_queue[ev[i].mbuf->port]; pipeline_fwd_event(&ev[i], RTE_SCHED_TYPE_ATOMIC); } else { @@ -265,6 +241,7 @@ pipeline_atq_worker_multi_stage_burst_fwd(void *arg) pipeline_event_enqueue_burst(dev, port, ev, nb_rx); } + return 0; } @@ -274,39 +251,36 @@ worker_wrapper(void *arg) struct worker_data *w = arg; struct evt_options *opt = w->t->opt; const bool burst = evt_has_burst_mode(w->dev_id); - const bool mt_safe = !w->t->mt_unsafe; + const bool internal_port = w->t->internal_port; const uint8_t nb_stages = opt->nb_stages; RTE_SET_USED(opt); if (nb_stages == 1) { - if (!burst && mt_safe) + if (!burst && internal_port) return pipeline_atq_worker_single_stage_tx(arg); - else if (!burst && !mt_safe) + else if (!burst && !internal_port) return pipeline_atq_worker_single_stage_fwd(arg); - else if (burst && mt_safe) + else if (burst && internal_port) return pipeline_atq_worker_single_stage_burst_tx(arg); - else if (burst && !mt_safe) + else if (burst && !internal_port) return pipeline_atq_worker_single_stage_burst_fwd(arg); } else { - if (!burst && mt_safe) + if (!burst && internal_port) return pipeline_atq_worker_multi_stage_tx(arg); - else if (!burst && !mt_safe) + else if (!burst && !internal_port) return pipeline_atq_worker_multi_stage_fwd(arg); - if (burst && mt_safe) + if (burst && internal_port) return pipeline_atq_worker_multi_stage_burst_tx(arg); - else if (burst && !mt_safe) + else if (burst && !internal_port) return pipeline_atq_worker_multi_stage_burst_fwd(arg); } + rte_panic("invalid worker\n"); } static int pipeline_atq_launch_lcores(struct evt_test *test, struct evt_options *opt) { - struct test_pipeline *t = evt_test_priv(test); - - if (t->mt_unsafe) - rte_service_component_runstate_set(t->tx_service.service_id, 1); return pipeline_launch_lcores(test, opt, worker_wrapper); } @@ -317,34 +291,36 @@ pipeline_atq_eventdev_setup(struct evt_test *test, struct evt_options *opt) int nb_ports; int nb_queues; uint8_t queue; - struct rte_event_dev_info info; - struct test_pipeline *t = evt_test_priv(test); - uint8_t tx_evqueue_id = 0; + uint8_t tx_evqueue_id[RTE_MAX_ETHPORTS] = {0}; uint8_t queue_arr[RTE_EVENT_MAX_QUEUES_PER_DEV]; uint8_t nb_worker_queues = 0; + uint8_t tx_evport_id = 0; + uint16_t prod = 0; + struct rte_event_dev_info info; + struct test_pipeline *t = evt_test_priv(test); nb_ports = evt_nr_active_lcores(opt->wlcores); nb_queues = rte_eth_dev_count_avail(); - /* One extra port and queueu for Tx service */ - if (t->mt_unsafe) { - tx_evqueue_id = nb_queues; - nb_ports++; - nb_queues++; + /* One queue for Tx adapter per port */ + if (!t->internal_port) { + RTE_ETH_FOREACH_DEV(prod) { + tx_evqueue_id[prod] = nb_queues; + nb_queues++; + } } - rte_event_dev_info_get(opt->dev_id, &info); const struct rte_event_dev_config config = { - .nb_event_queues = nb_queues, - .nb_event_ports = nb_ports, - .nb_events_limit = info.max_num_events, - .nb_event_queue_flows = opt->nb_flows, - .nb_event_port_dequeue_depth = - info.max_event_port_dequeue_depth, - .nb_event_port_enqueue_depth = - info.max_event_port_enqueue_depth, + .nb_event_queues = nb_queues, + .nb_event_ports = nb_ports, + .nb_events_limit = info.max_num_events, + .nb_event_queue_flows = opt->nb_flows, + .nb_event_port_dequeue_depth = + info.max_event_port_dequeue_depth, + .nb_event_port_enqueue_depth = + info.max_event_port_enqueue_depth, }; ret = rte_event_dev_configure(opt->dev_id, &config); if (ret) { @@ -353,21 +329,23 @@ pipeline_atq_eventdev_setup(struct evt_test *test, struct evt_options *opt) } struct rte_event_queue_conf q_conf = { - .priority = RTE_EVENT_DEV_PRIORITY_NORMAL, - .nb_atomic_flows = opt->nb_flows, - .nb_atomic_order_sequences = opt->nb_flows, + .priority = RTE_EVENT_DEV_PRIORITY_NORMAL, + .nb_atomic_flows = opt->nb_flows, + .nb_atomic_order_sequences = opt->nb_flows, }; /* queue configurations */ for (queue = 0; queue < nb_queues; queue++) { q_conf.event_queue_cfg = RTE_EVENT_QUEUE_CFG_ALL_TYPES; - if (t->mt_unsafe) { - if (queue == tx_evqueue_id) { - q_conf.event_queue_cfg = - RTE_EVENT_QUEUE_CFG_SINGLE_LINK; - } else { - queue_arr[nb_worker_queues] = queue; - nb_worker_queues++; + if (!t->internal_port) { + RTE_ETH_FOREACH_DEV(prod) { + if (queue == tx_evqueue_id[prod]) { + q_conf.event_queue_cfg = + RTE_EVENT_QUEUE_CFG_SINGLE_LINK; + } else { + queue_arr[nb_worker_queues] = queue; + nb_worker_queues++; + } } } @@ -383,20 +361,15 @@ pipeline_atq_eventdev_setup(struct evt_test *test, struct evt_options *opt) /* port configuration */ const struct rte_event_port_conf p_conf = { - .dequeue_depth = opt->wkr_deq_dep, - .enqueue_depth = info.max_event_port_dequeue_depth, - .new_event_threshold = info.max_num_events, + .dequeue_depth = opt->wkr_deq_dep, + .enqueue_depth = info.max_event_port_dequeue_depth, + .new_event_threshold = info.max_num_events, }; - if (t->mt_unsafe) { + if (!t->internal_port) ret = pipeline_event_port_setup(test, opt, queue_arr, nb_worker_queues, p_conf); - if (ret) - return ret; - - ret = pipeline_event_tx_service_setup(test, opt, tx_evqueue_id, - nb_ports - 1, p_conf); - } else + else ret = pipeline_event_port_setup(test, opt, NULL, nb_queues, p_conf); @@ -408,30 +381,32 @@ pipeline_atq_eventdev_setup(struct evt_test *test, struct evt_options *opt) * * eth_dev_count = 2, nb_stages = 2, atq mode * - * Multi thread safe : + * eth0, eth1 have Internal port capability : * queues = 2 * stride = 1 * * event queue pipelines: - * eth0 -> q0 ->tx - * eth1 -> q1 ->tx + * eth0 -> q0 ->Tx + * eth1 -> q1 ->Tx * * q0, q1 are configured as ATQ so, all the different stages can * be enqueued on the same queue. * - * Multi thread unsafe : - * queues = 3 + * eth0, eth1 use Tx adapters service core : + * queues = 4 * stride = 1 * * event queue pipelines: - * eth0 -> q0 - * } (q3->tx) Tx service - * eth1 -> q1 + * eth0 -> q0 -> q2 -> Tx + * eth1 -> q1 -> q3 -> Tx * - * q0,q1 are configured as stated above. - * q3 configured as SINGLE_LINK|ATOMIC. + * q0, q1 are configured as stated above. + * q2, q3 configured as SINGLE_LINK. */ ret = pipeline_event_rx_adapter_setup(opt, 1, p_conf); + if (ret) + return ret; + ret = pipeline_event_tx_adapter_setup(opt, p_conf); if (ret) return ret; @@ -445,12 +420,58 @@ pipeline_atq_eventdev_setup(struct evt_test *test, struct evt_options *opt) } } + /* Connect the tx_evqueue_id to the Tx adapter port */ + if (!t->internal_port) { + RTE_ETH_FOREACH_DEV(prod) { + ret = rte_event_eth_tx_adapter_event_port_get(prod, + &tx_evport_id); + if (ret) { + evt_err("Unable to get Tx adapter[%d]", prod); + return ret; + } + + if (rte_event_port_link(opt->dev_id, tx_evport_id, + &tx_evqueue_id[prod], + NULL, 1) != 1) { + evt_err("Unable to link Tx adptr[%d] evprt[%d]", + prod, tx_evport_id); + return ret; + } + } + } + + RTE_ETH_FOREACH_DEV(prod) { + ret = rte_eth_dev_start(prod); + if (ret) { + evt_err("Ethernet dev [%d] failed to start." + " Using synthetic producer", prod); + return ret; + } + } + ret = rte_event_dev_start(opt->dev_id); if (ret) { evt_err("failed to start eventdev %d", opt->dev_id); return ret; } + RTE_ETH_FOREACH_DEV(prod) { + ret = rte_event_eth_rx_adapter_start(prod); + if (ret) { + evt_err("Rx adapter[%d] start failed", prod); + return ret; + } + + ret = rte_event_eth_tx_adapter_start(prod); + if (ret) { + evt_err("Tx adapter[%d] start failed", prod); + return ret; + } + } + + memcpy(t->tx_evqueue_id, tx_evqueue_id, sizeof(uint8_t) * + RTE_MAX_ETHPORTS); + return 0; } diff --git a/app/test-eventdev/test_pipeline_common.c b/app/test-eventdev/test_pipeline_common.c index cdf9010b4..398b0010e 100644 --- a/app/test-eventdev/test_pipeline_common.c +++ b/app/test-eventdev/test_pipeline_common.c @@ -5,58 +5,6 @@ #include "test_pipeline_common.h" -static int32_t -pipeline_event_tx_burst_service_func(void *args) -{ - - int i; - struct tx_service_data *tx = args; - const uint8_t dev = tx->dev_id; - const uint8_t port = tx->port_id; - struct rte_event ev[BURST_SIZE + 1]; - - uint16_t nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0); - - if (!nb_rx) { - for (i = 0; i < tx->nb_ethports; i++) - rte_eth_tx_buffer_flush(i, 0, tx->tx_buf[i]); - return 0; - } - - for (i = 0; i < nb_rx; i++) { - struct rte_mbuf *m = ev[i].mbuf; - rte_eth_tx_buffer(m->port, 0, tx->tx_buf[m->port], m); - } - tx->processed_pkts += nb_rx; - - return 0; -} - -static int32_t -pipeline_event_tx_service_func(void *args) -{ - - int i; - struct tx_service_data *tx = args; - const uint8_t dev = tx->dev_id; - const uint8_t port = tx->port_id; - struct rte_event ev; - - uint16_t nb_rx = rte_event_dequeue_burst(dev, port, &ev, 1, 0); - - if (!nb_rx) { - for (i = 0; i < tx->nb_ethports; i++) - rte_eth_tx_buffer_flush(i, 0, tx->tx_buf[i]); - return 0; - } - - struct rte_mbuf *m = ev.mbuf; - rte_eth_tx_buffer(m->port, 0, tx->tx_buf[m->port], m); - tx->processed_pkts++; - - return 0; -} - int pipeline_test_result(struct evt_test *test, struct evt_options *opt) { @@ -97,11 +45,8 @@ processed_pkts(struct test_pipeline *t) uint64_t total = 0; rte_smp_rmb(); - if (t->mt_unsafe) - total = t->tx_service.processed_pkts; - else - for (i = 0; i < t->nb_workers; i++) - total += t->worker[i].processed_pkts; + for (i = 0; i < t->nb_workers; i++) + total += t->worker[i].processed_pkts; return total; } @@ -215,7 +160,6 @@ pipeline_ethdev_setup(struct evt_test *test, struct evt_options *opt) { uint16_t i; uint8_t nb_queues = 1; - uint8_t mt_state = 0; struct test_pipeline *t = evt_test_priv(test); struct rte_eth_rxconf rx_conf; struct rte_eth_conf port_conf = { @@ -237,13 +181,17 @@ pipeline_ethdev_setup(struct evt_test *test, struct evt_options *opt) return -ENODEV; } + t->internal_port = 1; RTE_ETH_FOREACH_DEV(i) { struct rte_eth_dev_info dev_info; struct rte_eth_conf local_port_conf = port_conf; + uint32_t caps = 0; + + rte_event_eth_tx_adapter_caps_get(opt->dev_id, i, &caps); + if (!(caps & RTE_EVENT_ETH_TX_ADAPTER_CAP_INTERNAL_PORT)) + t->internal_port = 0; rte_eth_dev_info_get(i, &dev_info); - mt_state = !(dev_info.tx_offload_capa & - DEV_TX_OFFLOAD_MT_LOCKFREE); rx_conf = dev_info.default_rxconf; rx_conf.offloads = port_conf.rxmode.offloads; @@ -278,11 +226,6 @@ pipeline_ethdev_setup(struct evt_test *test, struct evt_options *opt) return -EINVAL; } - t->mt_unsafe |= mt_state; - t->tx_service.tx_buf[i] = - rte_malloc(NULL, RTE_ETH_TX_BUFFER_SIZE(BURST_SIZE), 0); - if (t->tx_service.tx_buf[i] == NULL) - rte_panic("Unable to allocate Tx buffer memory."); rte_eth_promiscuous_enable(i); } @@ -294,7 +237,6 @@ pipeline_event_port_setup(struct evt_test *test, struct evt_options *opt, uint8_t *queue_arr, uint8_t nb_queues, const struct rte_event_port_conf p_conf) { - int i; int ret; uint8_t port; struct test_pipeline *t = evt_test_priv(test); @@ -315,23 +257,15 @@ pipeline_event_port_setup(struct evt_test *test, struct evt_options *opt, return ret; } - if (queue_arr == NULL) { - if (rte_event_port_link(opt->dev_id, port, NULL, NULL, - 0) != nb_queues) - goto link_fail; - } else { - for (i = 0; i < nb_queues; i++) { - if (rte_event_port_link(opt->dev_id, port, - &queue_arr[i], NULL, 1) != 1) - goto link_fail; - } - } + if (rte_event_port_link(opt->dev_id, port, queue_arr, NULL, + nb_queues) != nb_queues) + goto link_fail; } return 0; link_fail: - evt_err("failed to link all queues to port %d", port); + evt_err("failed to link queues to port %d", port); return -EINVAL; } @@ -384,79 +318,64 @@ pipeline_event_rx_adapter_setup(struct evt_options *opt, uint8_t stride, } } - ret = rte_eth_dev_start(prod); - if (ret) { - evt_err("Ethernet dev [%d] failed to start." - " Using synthetic producer", prod); - return ret; - } - - ret = rte_event_eth_rx_adapter_start(prod); - if (ret) { - evt_err("Rx adapter[%d] start failed", prod); - return ret; - } - evt_info("Port[%d] using Rx adapter[%d] started", prod, prod); + evt_info("Port[%d] using Rx adapter[%d] configured", prod, + prod); } return ret; } int -pipeline_event_tx_service_setup(struct evt_test *test, struct evt_options *opt, - uint8_t tx_queue_id, uint8_t tx_port_id, - const struct rte_event_port_conf p_conf) +pipeline_event_tx_adapter_setup(struct evt_options *opt, + struct rte_event_port_conf port_conf) { - int ret; - struct rte_service_spec serv; - struct test_pipeline *t = evt_test_priv(test); - struct tx_service_data *tx = &t->tx_service; + int ret = 0; + uint16_t consm = 0; - ret = rte_event_port_setup(opt->dev_id, tx_port_id, &p_conf); - if (ret) { - evt_err("failed to setup port %d", tx_port_id); - return ret; - } + RTE_ETH_FOREACH_DEV(consm) { + uint32_t cap; - if (rte_event_port_link(opt->dev_id, tx_port_id, &tx_queue_id, - NULL, 1) != 1) { - evt_err("failed to link queues to port %d", tx_port_id); - return -EINVAL; - } + ret = rte_event_eth_tx_adapter_caps_get(opt->dev_id, + consm, &cap); + if (ret) { + evt_err("failed to get event tx adapter[%d] caps", + consm); + return ret; + } - tx->dev_id = opt->dev_id; - tx->queue_id = tx_queue_id; - tx->port_id = tx_port_id; - tx->nb_ethports = rte_eth_dev_count_avail(); - tx->t = t; - - /* Register Tx service */ - memset(&serv, 0, sizeof(struct rte_service_spec)); - snprintf(serv.name, sizeof(serv.name), "Tx_service"); - - if (evt_has_burst_mode(opt->dev_id)) - serv.callback = pipeline_event_tx_burst_service_func; - else - serv.callback = pipeline_event_tx_service_func; - - serv.callback_userdata = (void *)tx; - ret = rte_service_component_register(&serv, &tx->service_id); - if (ret) { - evt_err("failed to register Tx service"); - return ret; - } + ret = rte_event_eth_tx_adapter_create(consm, opt->dev_id, + &port_conf); + if (ret) { + evt_err("failed to create tx adapter[%d]", consm); + return ret; + } - ret = evt_service_setup(tx->service_id); - if (ret) { - evt_err("Failed to setup service core for Tx service\n"); - return ret; - } + ret = rte_event_eth_tx_adapter_queue_add(consm, consm, -1); + if (ret) { + evt_err("failed to add tx queues to adapter[%d]", + consm); + return ret; + } - rte_service_runstate_set(tx->service_id, 1); + if (!(cap & RTE_EVENT_ETH_TX_ADAPTER_CAP_INTERNAL_PORT)) { + uint32_t service_id; - return 0; -} + rte_event_eth_tx_adapter_service_id_get(consm, + &service_id); + ret = evt_service_setup(service_id); + if (ret) { + evt_err("Failed to setup service core" + " for Tx adapter\n"); + return ret; + } + } + + evt_info("Port[%d] using Tx adapter[%d] Configured", consm, + consm); + } + return ret; +} void pipeline_ethdev_destroy(struct evt_test *test, struct evt_options *opt) @@ -464,16 +383,10 @@ pipeline_ethdev_destroy(struct evt_test *test, struct evt_options *opt) uint16_t i; RTE_SET_USED(test); RTE_SET_USED(opt); - struct test_pipeline *t = evt_test_priv(test); - - if (t->mt_unsafe) { - rte_service_component_runstate_set(t->tx_service.service_id, 0); - rte_service_runstate_set(t->tx_service.service_id, 0); - rte_service_component_unregister(t->tx_service.service_id); - } RTE_ETH_FOREACH_DEV(i) { rte_event_eth_rx_adapter_stop(i); + rte_event_eth_tx_adapter_stop(i); rte_eth_dev_stop(i); } } @@ -483,7 +396,6 @@ pipeline_eventdev_destroy(struct evt_test *test, struct evt_options *opt) { RTE_SET_USED(test); - rte_event_dev_stop(opt->dev_id); rte_event_dev_close(opt->dev_id); } diff --git a/app/test-eventdev/test_pipeline_common.h b/app/test-eventdev/test_pipeline_common.h index 9cd6b905b..0440b9e29 100644 --- a/app/test-eventdev/test_pipeline_common.h +++ b/app/test-eventdev/test_pipeline_common.h @@ -14,6 +14,7 @@ #include #include #include +#include #include #include #include @@ -35,30 +36,19 @@ struct worker_data { struct test_pipeline *t; } __rte_cache_aligned; -struct tx_service_data { - uint8_t dev_id; - uint8_t queue_id; - uint8_t port_id; - uint32_t service_id; - uint64_t processed_pkts; - uint16_t nb_ethports; - struct rte_eth_dev_tx_buffer *tx_buf[RTE_MAX_ETHPORTS]; - struct test_pipeline *t; -} __rte_cache_aligned; - struct test_pipeline { /* Don't change the offset of "done". Signal handler use this memory * to terminate all lcores work. */ int done; uint8_t nb_workers; - uint8_t mt_unsafe; + uint8_t internal_port; + uint8_t tx_evqueue_id[RTE_MAX_ETHPORTS]; enum evt_test_result result; uint32_t nb_flows; uint64_t outstand_pkts; struct rte_mempool *pool; struct worker_data worker[EVT_MAX_PORTS]; - struct tx_service_data tx_service; struct evt_options *opt; uint8_t sched_type_list[EVT_MAX_STAGES] __rte_cache_aligned; } __rte_cache_aligned; @@ -70,7 +60,7 @@ struct test_pipeline { struct test_pipeline *t = w->t; \ const uint8_t dev = w->dev_id; \ const uint8_t port = w->port_id; \ - struct rte_event ev + struct rte_event ev __rte_cache_aligned #define PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT \ int i; \ @@ -78,7 +68,7 @@ struct test_pipeline { struct test_pipeline *t = w->t; \ const uint8_t dev = w->dev_id; \ const uint8_t port = w->port_id; \ - struct rte_event ev[BURST_SIZE + 1] + struct rte_event ev[BURST_SIZE + 1] __rte_cache_aligned #define PIPELINE_WORKER_MULTI_STAGE_INIT \ struct worker_data *w = arg; \ @@ -88,10 +78,11 @@ struct test_pipeline { const uint8_t port = w->port_id; \ const uint8_t last_queue = t->opt->nb_stages - 1; \ uint8_t *const sched_type_list = &t->sched_type_list[0]; \ - struct rte_event ev + const uint8_t nb_stages = t->opt->nb_stages + 1; \ + struct rte_event ev __rte_cache_aligned #define PIPELINE_WORKER_MULTI_STAGE_BURST_INIT \ - int i; \ + int i; \ struct worker_data *w = arg; \ struct test_pipeline *t = w->t; \ uint8_t cq_id; \ @@ -99,7 +90,8 @@ struct test_pipeline { const uint8_t port = w->port_id; \ const uint8_t last_queue = t->opt->nb_stages - 1; \ uint8_t *const sched_type_list = &t->sched_type_list[0]; \ - struct rte_event ev[BURST_SIZE + 1] + const uint8_t nb_stages = t->opt->nb_stages + 1; \ + struct rte_event ev[BURST_SIZE + 1] __rte_cache_aligned static __rte_always_inline void pipeline_fwd_event(struct rte_event *ev, uint8_t sched) @@ -109,6 +101,28 @@ pipeline_fwd_event(struct rte_event *ev, uint8_t sched) ev->sched_type = sched; } +static __rte_always_inline void +pipeline_event_tx(const uint8_t dev, const uint8_t port, + struct rte_event * const ev) +{ + rte_event_eth_tx_adapter_txq_set(ev->mbuf, 0); + while (!rte_event_eth_tx_adapter_enqueue(dev, port, ev, 1)) + rte_pause(); +} + +static __rte_always_inline void +pipeline_event_tx_burst(const uint8_t dev, const uint8_t port, + struct rte_event *ev, const uint16_t nb_rx) +{ + uint16_t enq; + + enq = rte_event_eth_tx_adapter_enqueue(dev, port, ev, nb_rx); + while (enq < nb_rx) { + enq += rte_event_eth_tx_adapter_enqueue(dev, port, + ev + enq, nb_rx - enq); + } +} + static __rte_always_inline void pipeline_event_enqueue(const uint8_t dev, const uint8_t port, struct rte_event *ev) @@ -130,13 +144,6 @@ pipeline_event_enqueue_burst(const uint8_t dev, const uint8_t port, } } -static __rte_always_inline void -pipeline_tx_pkt(struct rte_mbuf *mbuf) -{ - while (rte_eth_tx_burst(mbuf->port, 0, &mbuf, 1) != 1) - rte_pause(); -} - static inline int pipeline_nb_event_ports(struct evt_options *opt) { @@ -149,9 +156,8 @@ int pipeline_test_setup(struct evt_test *test, struct evt_options *opt); int pipeline_ethdev_setup(struct evt_test *test, struct evt_options *opt); int pipeline_event_rx_adapter_setup(struct evt_options *opt, uint8_t stride, struct rte_event_port_conf prod_conf); -int pipeline_event_tx_service_setup(struct evt_test *test, - struct evt_options *opt, uint8_t tx_queue_id, - uint8_t tx_port_id, const struct rte_event_port_conf p_conf); +int pipeline_event_tx_adapter_setup(struct evt_options *opt, + struct rte_event_port_conf prod_conf); int pipeline_mempool_setup(struct evt_test *test, struct evt_options *opt); int pipeline_event_port_setup(struct evt_test *test, struct evt_options *opt, uint8_t *queue_arr, uint8_t nb_queues, diff --git a/app/test-eventdev/test_pipeline_queue.c b/app/test-eventdev/test_pipeline_queue.c index 2e0d93d99..1156c2555 100644 --- a/app/test-eventdev/test_pipeline_queue.c +++ b/app/test-eventdev/test_pipeline_queue.c @@ -15,7 +15,7 @@ pipeline_queue_nb_event_queues(struct evt_options *opt) return (eth_count * opt->nb_stages) + eth_count; } -static int +static __rte_noinline int pipeline_queue_worker_single_stage_tx(void *arg) { PIPELINE_WORKER_SINGLE_STAGE_INIT; @@ -29,7 +29,7 @@ pipeline_queue_worker_single_stage_tx(void *arg) } if (ev.sched_type == RTE_SCHED_TYPE_ATOMIC) { - pipeline_tx_pkt(ev.mbuf); + pipeline_event_tx(dev, port, &ev); w->processed_pkts++; } else { ev.queue_id++; @@ -41,11 +41,11 @@ pipeline_queue_worker_single_stage_tx(void *arg) return 0; } -static int +static __rte_noinline int pipeline_queue_worker_single_stage_fwd(void *arg) { PIPELINE_WORKER_SINGLE_STAGE_INIT; - const uint8_t tx_queue = t->tx_service.queue_id; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t event = rte_event_dequeue_burst(dev, port, &ev, 1, 0); @@ -55,7 +55,8 @@ pipeline_queue_worker_single_stage_fwd(void *arg) continue; } - ev.queue_id = tx_queue; + ev.queue_id = tx_queue[ev.mbuf->port]; + rte_event_eth_tx_adapter_txq_set(ev.mbuf, 0); pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC); pipeline_event_enqueue(dev, port, &ev); w->processed_pkts++; @@ -64,7 +65,7 @@ pipeline_queue_worker_single_stage_fwd(void *arg) return 0; } -static int +static __rte_noinline int pipeline_queue_worker_single_stage_burst_tx(void *arg) { PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT; @@ -81,8 +82,7 @@ pipeline_queue_worker_single_stage_burst_tx(void *arg) for (i = 0; i < nb_rx; i++) { rte_prefetch0(ev[i + 1].mbuf); if (ev[i].sched_type == RTE_SCHED_TYPE_ATOMIC) { - - pipeline_tx_pkt(ev[i].mbuf); + pipeline_event_tx(dev, port, &ev[i]); ev[i].op = RTE_EVENT_OP_RELEASE; w->processed_pkts++; } else { @@ -98,11 +98,11 @@ pipeline_queue_worker_single_stage_burst_tx(void *arg) return 0; } -static int +static __rte_noinline int pipeline_queue_worker_single_stage_burst_fwd(void *arg) { PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT; - const uint8_t tx_queue = t->tx_service.queue_id; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t nb_rx = rte_event_dequeue_burst(dev, port, ev, @@ -115,23 +115,24 @@ pipeline_queue_worker_single_stage_burst_fwd(void *arg) for (i = 0; i < nb_rx; i++) { rte_prefetch0(ev[i + 1].mbuf); - ev[i].queue_id = tx_queue; + ev[i].queue_id = tx_queue[ev[i].mbuf->port]; + rte_event_eth_tx_adapter_txq_set(ev[i].mbuf, 0); pipeline_fwd_event(&ev[i], RTE_SCHED_TYPE_ATOMIC); - w->processed_pkts++; } pipeline_event_enqueue_burst(dev, port, ev, nb_rx); + w->processed_pkts += nb_rx; } return 0; } -static int +static __rte_noinline int pipeline_queue_worker_multi_stage_tx(void *arg) { PIPELINE_WORKER_MULTI_STAGE_INIT; - const uint8_t nb_stages = t->opt->nb_stages + 1; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t event = rte_event_dequeue_burst(dev, port, &ev, 1, 0); @@ -143,31 +144,27 @@ pipeline_queue_worker_multi_stage_tx(void *arg) cq_id = ev.queue_id % nb_stages; - if (cq_id >= last_queue) { - if (ev.sched_type == RTE_SCHED_TYPE_ATOMIC) { - - pipeline_tx_pkt(ev.mbuf); - w->processed_pkts++; - continue; - } - ev.queue_id += (cq_id == last_queue) ? 1 : 0; - pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC); - } else { - ev.queue_id++; - pipeline_fwd_event(&ev, sched_type_list[cq_id]); + if (ev.queue_id == tx_queue[ev.mbuf->port]) { + pipeline_event_tx(dev, port, &ev); + w->processed_pkts++; + continue; } + ev.queue_id++; + pipeline_fwd_event(&ev, cq_id != last_queue ? + sched_type_list[cq_id] : + RTE_SCHED_TYPE_ATOMIC); pipeline_event_enqueue(dev, port, &ev); } + return 0; } -static int +static __rte_noinline int pipeline_queue_worker_multi_stage_fwd(void *arg) { PIPELINE_WORKER_MULTI_STAGE_INIT; - const uint8_t nb_stages = t->opt->nb_stages + 1; - const uint8_t tx_queue = t->tx_service.queue_id; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t event = rte_event_dequeue_burst(dev, port, &ev, 1, 0); @@ -180,7 +177,8 @@ pipeline_queue_worker_multi_stage_fwd(void *arg) cq_id = ev.queue_id % nb_stages; if (cq_id == last_queue) { - ev.queue_id = tx_queue; + ev.queue_id = tx_queue[ev.mbuf->port]; + rte_event_eth_tx_adapter_txq_set(ev.mbuf, 0); pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC); w->processed_pkts++; } else { @@ -190,14 +188,15 @@ pipeline_queue_worker_multi_stage_fwd(void *arg) pipeline_event_enqueue(dev, port, &ev); } + return 0; } -static int +static __rte_noinline int pipeline_queue_worker_multi_stage_burst_tx(void *arg) { PIPELINE_WORKER_MULTI_STAGE_BURST_INIT; - const uint8_t nb_stages = t->opt->nb_stages + 1; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t nb_rx = rte_event_dequeue_burst(dev, port, ev, @@ -212,37 +211,30 @@ pipeline_queue_worker_multi_stage_burst_tx(void *arg) rte_prefetch0(ev[i + 1].mbuf); cq_id = ev[i].queue_id % nb_stages; - if (cq_id >= last_queue) { - if (ev[i].sched_type == RTE_SCHED_TYPE_ATOMIC) { - - pipeline_tx_pkt(ev[i].mbuf); - ev[i].op = RTE_EVENT_OP_RELEASE; - w->processed_pkts++; - continue; - } - - ev[i].queue_id += (cq_id == last_queue) ? 1 : 0; - pipeline_fwd_event(&ev[i], - RTE_SCHED_TYPE_ATOMIC); - } else { - ev[i].queue_id++; - pipeline_fwd_event(&ev[i], - sched_type_list[cq_id]); + if (ev[i].queue_id == tx_queue[ev[i].mbuf->port]) { + pipeline_event_tx(dev, port, &ev[i]); + ev[i].op = RTE_EVENT_OP_RELEASE; + w->processed_pkts++; + continue; } + ev[i].queue_id++; + pipeline_fwd_event(&ev[i], cq_id != last_queue ? + sched_type_list[cq_id] : + RTE_SCHED_TYPE_ATOMIC); } pipeline_event_enqueue_burst(dev, port, ev, nb_rx); } + return 0; } -static int +static __rte_noinline int pipeline_queue_worker_multi_stage_burst_fwd(void *arg) { PIPELINE_WORKER_MULTI_STAGE_BURST_INIT; - const uint8_t nb_stages = t->opt->nb_stages + 1; - const uint8_t tx_queue = t->tx_service.queue_id; + const uint8_t *tx_queue = t->tx_evqueue_id; while (t->done == false) { uint16_t nb_rx = rte_event_dequeue_burst(dev, port, ev, @@ -258,7 +250,8 @@ pipeline_queue_worker_multi_stage_burst_fwd(void *arg) cq_id = ev[i].queue_id % nb_stages; if (cq_id == last_queue) { - ev[i].queue_id = tx_queue; + ev[i].queue_id = tx_queue[ev[i].mbuf->port]; + rte_event_eth_tx_adapter_txq_set(ev[i].mbuf, 0); pipeline_fwd_event(&ev[i], RTE_SCHED_TYPE_ATOMIC); w->processed_pkts++; @@ -271,6 +264,7 @@ pipeline_queue_worker_multi_stage_burst_fwd(void *arg) pipeline_event_enqueue_burst(dev, port, ev, nb_rx); } + return 0; } @@ -280,28 +274,28 @@ worker_wrapper(void *arg) struct worker_data *w = arg; struct evt_options *opt = w->t->opt; const bool burst = evt_has_burst_mode(w->dev_id); - const bool mt_safe = !w->t->mt_unsafe; + const bool internal_port = w->t->internal_port; const uint8_t nb_stages = opt->nb_stages; RTE_SET_USED(opt); if (nb_stages == 1) { - if (!burst && mt_safe) + if (!burst && internal_port) return pipeline_queue_worker_single_stage_tx(arg); - else if (!burst && !mt_safe) + else if (!burst && !internal_port) return pipeline_queue_worker_single_stage_fwd(arg); - else if (burst && mt_safe) + else if (burst && internal_port) return pipeline_queue_worker_single_stage_burst_tx(arg); - else if (burst && !mt_safe) + else if (burst && !internal_port) return pipeline_queue_worker_single_stage_burst_fwd( arg); } else { - if (!burst && mt_safe) + if (!burst && internal_port) return pipeline_queue_worker_multi_stage_tx(arg); - else if (!burst && !mt_safe) + else if (!burst && !internal_port) return pipeline_queue_worker_multi_stage_fwd(arg); - else if (burst && mt_safe) + else if (burst && internal_port) return pipeline_queue_worker_multi_stage_burst_tx(arg); - else if (burst && !mt_safe) + else if (burst && !internal_port) return pipeline_queue_worker_multi_stage_burst_fwd(arg); } @@ -311,10 +305,6 @@ worker_wrapper(void *arg) static int pipeline_queue_launch_lcores(struct evt_test *test, struct evt_options *opt) { - struct test_pipeline *t = evt_test_priv(test); - - if (t->mt_unsafe) - rte_service_component_runstate_set(t->tx_service.service_id, 1); return pipeline_launch_lcores(test, opt, worker_wrapper); } @@ -326,25 +316,21 @@ pipeline_queue_eventdev_setup(struct evt_test *test, struct evt_options *opt) int nb_queues; int nb_stages = opt->nb_stages; uint8_t queue; - struct rte_event_dev_info info; - struct test_pipeline *t = evt_test_priv(test); - uint8_t tx_evqueue_id = 0; + uint8_t tx_evport_id = 0; + uint8_t tx_evqueue_id[RTE_MAX_ETHPORTS] = {0}; uint8_t queue_arr[RTE_EVENT_MAX_QUEUES_PER_DEV]; uint8_t nb_worker_queues = 0; + uint16_t prod = 0; + struct rte_event_dev_info info; + struct test_pipeline *t = evt_test_priv(test); nb_ports = evt_nr_active_lcores(opt->wlcores); nb_queues = rte_eth_dev_count_avail() * (nb_stages); - /* Extra port for Tx service. */ - if (t->mt_unsafe) { - tx_evqueue_id = nb_queues; - nb_ports++; - nb_queues++; - } else - nb_queues += rte_eth_dev_count_avail(); + /* One queue for Tx adapter per port */ + nb_queues += rte_eth_dev_count_avail(); rte_event_dev_info_get(opt->dev_id, &info); - const struct rte_event_dev_config config = { .nb_event_queues = nb_queues, .nb_event_ports = nb_ports, @@ -370,24 +356,19 @@ pipeline_queue_eventdev_setup(struct evt_test *test, struct evt_options *opt) for (queue = 0; queue < nb_queues; queue++) { uint8_t slot; - if (!t->mt_unsafe) { - slot = queue % (nb_stages + 1); - q_conf.schedule_type = slot == nb_stages ? - RTE_SCHED_TYPE_ATOMIC : - opt->sched_type_list[slot]; - } else { - slot = queue % nb_stages; - - if (queue == tx_evqueue_id) { - q_conf.schedule_type = RTE_SCHED_TYPE_ATOMIC; + q_conf.event_queue_cfg = 0; + slot = queue % (nb_stages + 1); + if (slot == nb_stages) { + q_conf.schedule_type = RTE_SCHED_TYPE_ATOMIC; + if (!t->internal_port) { q_conf.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK; - } else { - q_conf.schedule_type = - opt->sched_type_list[slot]; - queue_arr[nb_worker_queues] = queue; - nb_worker_queues++; } + tx_evqueue_id[prod++] = queue; + } else { + q_conf.schedule_type = opt->sched_type_list[slot]; + queue_arr[nb_worker_queues] = queue; + nb_worker_queues++; } ret = rte_event_queue_setup(opt->dev_id, queue, &q_conf); @@ -407,19 +388,11 @@ pipeline_queue_eventdev_setup(struct evt_test *test, struct evt_options *opt) .new_event_threshold = info.max_num_events, }; - /* - * If tx is multi thread safe then allow workers to do Tx else use Tx - * service to Tx packets. - */ - if (t->mt_unsafe) { + if (!t->internal_port) { ret = pipeline_event_port_setup(test, opt, queue_arr, nb_worker_queues, p_conf); if (ret) return ret; - - ret = pipeline_event_tx_service_setup(test, opt, tx_evqueue_id, - nb_ports - 1, p_conf); - } else ret = pipeline_event_port_setup(test, opt, NULL, nb_queues, p_conf); @@ -431,7 +404,6 @@ pipeline_queue_eventdev_setup(struct evt_test *test, struct evt_options *opt) * * eth_dev_count = 2, nb_stages = 2. * - * Multi thread safe : * queues = 6 * stride = 3 * @@ -439,21 +411,14 @@ pipeline_queue_eventdev_setup(struct evt_test *test, struct evt_options *opt) * eth0 -> q0 -> q1 -> (q2->tx) * eth1 -> q3 -> q4 -> (q5->tx) * - * q2, q5 configured as ATOMIC - * - * Multi thread unsafe : - * queues = 5 - * stride = 2 + * q2, q5 configured as ATOMIC | SINGLE_LINK * - * event queue pipelines: - * eth0 -> q0 -> q1 - * } (q4->tx) Tx service - * eth1 -> q2 -> q3 - * - * q4 configured as SINGLE_LINK|ATOMIC */ - ret = pipeline_event_rx_adapter_setup(opt, - t->mt_unsafe ? nb_stages : nb_stages + 1, p_conf); + ret = pipeline_event_rx_adapter_setup(opt, nb_stages + 1, p_conf); + if (ret) + return ret; + + ret = pipeline_event_tx_adapter_setup(opt, p_conf); if (ret) return ret; @@ -467,12 +432,60 @@ pipeline_queue_eventdev_setup(struct evt_test *test, struct evt_options *opt) } } + /* Connect the tx_evqueue_id to the Tx adapter port */ + if (!t->internal_port) { + RTE_ETH_FOREACH_DEV(prod) { + ret = rte_event_eth_tx_adapter_event_port_get(prod, + &tx_evport_id); + if (ret) { + evt_err("Unable to get Tx adptr[%d] evprt[%d]", + prod, tx_evport_id); + return ret; + } + + if (rte_event_port_link(opt->dev_id, tx_evport_id, + &tx_evqueue_id[prod], + NULL, 1) != 1) { + evt_err("Unable to link Tx adptr[%d] evprt[%d]", + prod, tx_evport_id); + return ret; + } + } + } + + RTE_ETH_FOREACH_DEV(prod) { + ret = rte_eth_dev_start(prod); + if (ret) { + evt_err("Ethernet dev [%d] failed to start." + " Using synthetic producer", prod); + return ret; + } + + } + ret = rte_event_dev_start(opt->dev_id); if (ret) { evt_err("failed to start eventdev %d", opt->dev_id); return ret; } + RTE_ETH_FOREACH_DEV(prod) { + ret = rte_event_eth_rx_adapter_start(prod); + if (ret) { + evt_err("Rx adapter[%d] start failed", prod); + return ret; + } + + ret = rte_event_eth_tx_adapter_start(prod); + if (ret) { + evt_err("Tx adapter[%d] start failed", prod); + return ret; + } + } + + memcpy(t->tx_evqueue_id, tx_evqueue_id, sizeof(uint8_t) * + RTE_MAX_ETHPORTS); + return 0; } -- 2.19.0