From patchwork Wed Dec 9 16:02:48 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Maxim Mikityanskiy X-Patchwork-Id: 11961805 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-18.7 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER, INCLUDES_PATCH,MAILING_LIST_MULTI,MSGID_FROM_MTA_HEADER,SPF_HELO_NONE, SPF_PASS,URIBL_BLOCKED,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id AE27AC433FE for ; Wed, 9 Dec 2020 16:03:48 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 7A3702343B for ; Wed, 9 Dec 2020 16:03:48 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1729983AbgLIQDr (ORCPT ); Wed, 9 Dec 2020 11:03:47 -0500 Received: from mail-eopbgr60048.outbound.protection.outlook.com ([40.107.6.48]:36676 "EHLO EUR04-DB3-obe.outbound.protection.outlook.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1728392AbgLIQDr (ORCPT ); Wed, 9 Dec 2020 11:03:47 -0500 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=a0riwAOxWxiDcL9hQAy4Vwa2UtvmaoLqvrtvKE4hBSFnwJrnkiCGj5AHuGZ5d+4o+e4dXls/3uTexFzp6co89nh9seTaQYZkdWZV0S3aJliFdJivOAH+bmtU6hY504A7MyWhOv8RCmTjeoAv2Ny55/j6LwubKwuj0+o3ef2/Z2fKxUN2b1CxLE15y49xZt9pXmffdG4AYAcwr4/CJKuzL4lpj/xJ8D9RABHGAkfBGjWgWHo8cTs1UUy3HR0Sht7R/q54K3oiHOLDU7V5xvui/O8QH7+5/+QvNu+/kOasmnVeA8MRQzvzrZnsQGTYFQugS2eO6h6FmeVj8emm5bjIGg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=nj2/rfizjPUWlZ9QekU751OoWU5HcqkDHSmRwk+E7rI=; b=Xthd5uNO/y4W0RE1A3OM/Vd4k048KsbKfFj2mD2pZCJhQhXyWjBcBjVDMDePZMjhndIdt70xK2Ryslyt72jkZILrXuym7y+K0IdJdlWNANFMF9n7kPWylQ3/XlzOsxnGElwNnAdDqGgYSTvKUSBJBZtes6nRB5x7x8JT4xv4zMYqpeiUaITnGGg1Cp1uuyavswKWFt7PSTlrkZozleIDMtfveXAcPPiZb+NYvlvNVvHhYLzwca+rR9ysqjCPP+RTLA9wO4SS6T1J9EOLyCH8DE1bW8KkhSYrrcUS8u+4ZF4G8/Gdbg8pfpuSaDK9aA/CA24dIf9vbf1zcDLSQv4peQ== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=mellanox.com; dmarc=pass action=none header.from=mellanox.com; dkim=pass header.d=mellanox.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=Mellanox.com; s=selector1; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=nj2/rfizjPUWlZ9QekU751OoWU5HcqkDHSmRwk+E7rI=; b=ewUKwjFdJfFj2gcYZ/ZW4VHWnutmfFnRgM1Xnt/8AwPQtNeJ7RWuN1cCLOgGqvhUd9UmYnzr3EpbnHxPYPUCjVukxTA8U86NxWZF0QvxmCybc9Ozsw4IvdYJfCINpfR8OdLpRbxdCucHV9HRgUZXbvOZvWDqEu12TC1GltfNrho= Authentication-Results: davemloft.net; dkim=none (message not signed) header.d=none;davemloft.net; dmarc=none action=none header.from=mellanox.com; Received: from AM6PR05MB5974.eurprd05.prod.outlook.com (2603:10a6:20b:a7::12) by AM6PR05MB4149.eurprd05.prod.outlook.com (2603:10a6:209:50::21) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.3632.23; Wed, 9 Dec 2020 16:03:01 +0000 Received: from AM6PR05MB5974.eurprd05.prod.outlook.com ([fe80::642a:6259:153:ab88]) by AM6PR05MB5974.eurprd05.prod.outlook.com ([fe80::642a:6259:153:ab88%3]) with mapi id 15.20.3632.024; Wed, 9 Dec 2020 16:03:01 +0000 From: Maxim Mikityanskiy To: "David S. Miller" , Jamal Hadi Salim , Cong Wang , Jiri Pirko Cc: Saeed Mahameed , Jakub Kicinski , Tariq Toukan , Maxim Mikityanskiy , netdev@vger.kernel.org, Maxim Mikityanskiy , Tariq Toukan Subject: [PATCH net-next 1/4] net: sched: Add multi-queue support to sch_tree_lock Date: Wed, 9 Dec 2020 18:02:48 +0200 Message-Id: <20201209160251.19054-2-maximmi@mellanox.com> X-Mailer: git-send-email 2.20.1 In-Reply-To: <20201209160251.19054-1-maximmi@mellanox.com> References: <20201209160251.19054-1-maximmi@mellanox.com> X-Originating-IP: [94.188.199.18] X-ClientProxiedBy: AM0PR04CA0098.eurprd04.prod.outlook.com (2603:10a6:208:be::39) To AM6PR05MB5974.eurprd05.prod.outlook.com (2603:10a6:20b:a7::12) MIME-Version: 1.0 X-MS-Exchange-MessageSentRepresentingType: 1 Received: from dev-l-vrt-208.mtl.labs.mlnx (94.188.199.18) by AM0PR04CA0098.eurprd04.prod.outlook.com (2603:10a6:208:be::39) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.3654.12 via Frontend Transport; Wed, 9 Dec 2020 16:02:36 +0000 X-MS-PublicTrafficType: Email X-MS-Office365-Filtering-HT: Tenant X-MS-Office365-Filtering-Correlation-Id: 925640aa-ee20-43fa-c830-08d89c5be7d1 X-MS-TrafficTypeDiagnostic: AM6PR05MB4149: X-LD-Processed: a652971c-7d2e-4d9b-a6a4-d149256f461b,ExtFwd,ExtAddr X-MS-Exchange-Transport-Forked: True X-Microsoft-Antispam-PRVS: X-MS-Oob-TLC-OOBClassifiers: OLM:7691; X-MS-Exchange-SenderADCheck: 1 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: SL1654Vv+ZQymcAIpPyLAdYpcKkgidMNMpszTLf2iHh1h+7SmI3F7xGe/pSu4+jzOGiMNkQ/v1PZ+27zIwsWsSRHVWIpUiqAtqZY2qjnPK/H33vm/yZIE4cORT+gpOJdNAleGPgCz5CeR6Vn8KapomsuXxxzUlXWxmuMo4vs8ONTejsDhyvd7Iz+yDiUWcpk0IrgoEnu1/kAp5yEOYpaV7tWxv9mcGtbfwyH0SBrnisJtvAd102PPy2A3OB/aIRmFdxC5JWb56BtzQyzUhqj7gfbYvAn/Um9JaCFkuA5nhFZOjskk32VxI2X/TOcyI9oJdMv1WZyaGumaFkraBrKXAUpjlS0YlVmQglgsXqQStunHbBwZhy2h3l6ZWQtvbzN X-Forefront-Antispam-Report: CIP:255.255.255.255;CTRY:;LANG:en;SCL:1;SRV:;IPV:NLI;SFV:NSPM;H:AM6PR05MB5974.eurprd05.prod.outlook.com;PTR:;CAT:NONE;SFS:(4636009)(346002)(366004)(136003)(376002)(5660300002)(16526019)(186003)(26005)(86362001)(110136005)(956004)(2616005)(54906003)(52116002)(8936002)(1076003)(6506007)(6512007)(508600001)(66556008)(34490700003)(66946007)(7416002)(2906002)(66476007)(83380400001)(4326008)(6666004)(6486002)(8676002)(36756003);DIR:OUT;SFP:1101; X-MS-Exchange-AntiSpam-MessageData: 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 X-OriginatorOrg: Mellanox.com X-MS-Exchange-CrossTenant-AuthSource: AM6PR05MB5974.eurprd05.prod.outlook.com X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 09 Dec 2020 16:03:01.7687 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: a652971c-7d2e-4d9b-a6a4-d149256f461b X-MS-Exchange-CrossTenant-Network-Message-Id: 925640aa-ee20-43fa-c830-08d89c5be7d1 X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: Yx6Rrn5ivk8WKRQwqUxlUtqiiUMFPRcBklFZKKzIfnKv2cJNTGvE5f5AvCJALoAGxFyZZzP8IMpsMLpRzPj/Ng== X-MS-Exchange-Transport-CrossTenantHeadersStamped: AM6PR05MB4149 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org The existing qdiscs that set TCQ_F_MQROOT don't use sch_tree_lock. However, hardware-offloaded HTB will start setting this flag while also using sch_tree_lock. The current implementation of sch_tree_lock basically locks on qdisc->dev_queue->qdisc, and it works fine when the tree is attached to some queue. However, it's not the case for MQROOT qdiscs: such a qdisc is the root itself, and its dev_queue just points to queue 0, while not actually being used, because there are real per-queue qdiscs. This patch changes the logic of sch_tree_lock and sch_tree_unlock to lock the qdisc itself if it's the MQROOT. Signed-off-by: Maxim Mikityanskiy Reviewed-by: Tariq Toukan --- include/net/sch_generic.h | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h index 162ed6249e8b..cfac2b0b5cc7 100644 --- a/include/net/sch_generic.h +++ b/include/net/sch_generic.h @@ -563,14 +563,20 @@ static inline struct net_device *qdisc_dev(const struct Qdisc *qdisc) return qdisc->dev_queue->dev; } -static inline void sch_tree_lock(const struct Qdisc *q) +static inline void sch_tree_lock(struct Qdisc *q) { - spin_lock_bh(qdisc_root_sleeping_lock(q)); + if (q->flags & TCQ_F_MQROOT) + spin_lock_bh(qdisc_lock(q)); + else + spin_lock_bh(qdisc_root_sleeping_lock(q)); } -static inline void sch_tree_unlock(const struct Qdisc *q) +static inline void sch_tree_unlock(struct Qdisc *q) { - spin_unlock_bh(qdisc_root_sleeping_lock(q)); + if (q->flags & TCQ_F_MQROOT) + spin_unlock_bh(qdisc_lock(q)); + else + spin_unlock_bh(qdisc_root_sleeping_lock(q)); } extern struct Qdisc noop_qdisc; From patchwork Wed Dec 9 16:02:49 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Maxim Mikityanskiy X-Patchwork-Id: 11961809 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-18.7 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER, INCLUDES_PATCH,MAILING_LIST_MULTI,MSGID_FROM_MTA_HEADER,SPF_HELO_NONE, SPF_PASS,URIBL_BLOCKED,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 70FFDC4361B for ; Wed, 9 Dec 2020 16:04:33 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 3515523BEF for ; Wed, 9 Dec 2020 16:04:33 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1730151AbgLIQEa (ORCPT ); Wed, 9 Dec 2020 11:04:30 -0500 Received: from mail-eopbgr60048.outbound.protection.outlook.com ([40.107.6.48]:36676 "EHLO EUR04-DB3-obe.outbound.protection.outlook.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1728392AbgLIQEO (ORCPT ); Wed, 9 Dec 2020 11:04:14 -0500 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=hqI1APqe0UB8jr9vjrxlEJGnPW+8tGZWbvH7fsG+ON8eIT+8agwioahyIfdxcPRMJ9jPixlJsuh6UpDm0b/257PpTSrd57MIFPDDFBaOC98a20WFHZ2ndEWHdWwmCaRxXakfnlRIFBU1t1ZwBk9Aikt3PSg9C34b7Fki8MEt4eYw+nEFt+D2Pe0CV5ihEvK+JeExB4Rc29NnMtS91UDckkwaankHLn8kMrgmMusRlnQKhqXyAa/lKIMK1r2FiXHzQOnB2SphiNqzfqmnBYGwQPeILv5ng9ygq4Pfqf0D/BAP5yFQ+fCHCWy1rmB6u4JXkAVJhCLI4W38RjmnpKm9kw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=WKm5+Ghvgkuc7BH/ZbeL8gbS1alMdORxkSLw46C5icg=; b=Oj2K4oYcnTUmgPv+4EZJfVqlgFp769hirKWEVfDYOlMwRXLFpTiyR+eyTrkMZtj3YGI2bvJSZtSziH317EOUjvJheXEYhdpn7Bfp4FnLTbDgbVJZE8lhwapWJq122zTo0QN1Qhb3UQiKBFjwyy/FEWRKC3urpu8hH3PoQXXUQIohOz2bBoEunOUXSks/nPFpystA50XeHXWn7bboIq/boOPdiggKPCE4h33jXj6dh3c0daAb29xi12Roj9noY6Q+vQoez+N9SJ3kuvvakLKc3bMOJLnWQ2HCXdz2/YN1YGybGjVDf+KK81/6Ait4dWpB7DaWlMNVb82AsOroxegAgQ== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=mellanox.com; dmarc=pass action=none header.from=mellanox.com; dkim=pass header.d=mellanox.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=Mellanox.com; s=selector1; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=WKm5+Ghvgkuc7BH/ZbeL8gbS1alMdORxkSLw46C5icg=; b=Ry7F77TNxEBaEZ5pt/hum5cg1WXXsodYMrZ/Geh/S/7wo8nENtAOzigp3E5DyO0zF/a7WjiSayBIvGtTbjDy3Aah2FwnkOD7Keldq6ZByv7OQZ8iWGrlYvc32GmekCN5h9tkY2T/KzppGSp4c9ZdcObHbPMmeCozhiuflkksOp8= Authentication-Results: davemloft.net; dkim=none (message not signed) header.d=none;davemloft.net; dmarc=none action=none header.from=mellanox.com; Received: from AM6PR05MB5974.eurprd05.prod.outlook.com (2603:10a6:20b:a7::12) by AM6PR05MB4149.eurprd05.prod.outlook.com (2603:10a6:209:50::21) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.3632.23; Wed, 9 Dec 2020 16:03:04 +0000 Received: from AM6PR05MB5974.eurprd05.prod.outlook.com ([fe80::642a:6259:153:ab88]) by AM6PR05MB5974.eurprd05.prod.outlook.com ([fe80::642a:6259:153:ab88%3]) with mapi id 15.20.3632.024; Wed, 9 Dec 2020 16:03:03 +0000 From: Maxim Mikityanskiy To: "David S. Miller" , Jamal Hadi Salim , Cong Wang , Jiri Pirko Cc: Saeed Mahameed , Jakub Kicinski , Tariq Toukan , Maxim Mikityanskiy , netdev@vger.kernel.org, Maxim Mikityanskiy , Tariq Toukan Subject: [PATCH net-next 2/4] sch_htb: Hierarchical QoS hardware offload Date: Wed, 9 Dec 2020 18:02:49 +0200 Message-Id: <20201209160251.19054-3-maximmi@mellanox.com> X-Mailer: git-send-email 2.20.1 In-Reply-To: <20201209160251.19054-1-maximmi@mellanox.com> References: <20201209160251.19054-1-maximmi@mellanox.com> X-Originating-IP: [94.188.199.18] X-ClientProxiedBy: AM0PR04CA0098.eurprd04.prod.outlook.com (2603:10a6:208:be::39) To AM6PR05MB5974.eurprd05.prod.outlook.com (2603:10a6:20b:a7::12) MIME-Version: 1.0 X-MS-Exchange-MessageSentRepresentingType: 1 Received: from dev-l-vrt-208.mtl.labs.mlnx (94.188.199.18) by AM0PR04CA0098.eurprd04.prod.outlook.com (2603:10a6:208:be::39) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.3654.12 via Frontend Transport; Wed, 9 Dec 2020 16:03:02 +0000 X-MS-PublicTrafficType: Email X-MS-Office365-Filtering-HT: Tenant X-MS-Office365-Filtering-Correlation-Id: 17faab35-a867-4367-01c8-08d89c5be8fb X-MS-TrafficTypeDiagnostic: AM6PR05MB4149: X-LD-Processed: a652971c-7d2e-4d9b-a6a4-d149256f461b,ExtFwd,ExtAddr X-MS-Exchange-Transport-Forked: True X-Microsoft-Antispam-PRVS: X-MS-Oob-TLC-OOBClassifiers: OLM:8882; X-MS-Exchange-SenderADCheck: 1 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 0e8JPYYlDjDpFJjFEtQenpfro54AcD8bWUJ4vLTfMgh7JK+z2zLRb9W3e3WT5Z/i2Km7m1oq/kcJ42WKXVSiK/KwdYYqti7R/qRE40AtqSya3cZULx90qRWG2diccFendkLzUvRf0o93+5lpYlYvhvRXWS9T8Htcd9C+hz1k61yohbsxN9IcdPLngxl0Ytx5ETQYUUZDh8UKRiOl6j1RpSweenfDdqzJyJDMdgVxqJ9yrtkaHK8oTdGp8h+lSCSvkqxr6EMKRKAJJurP1Pt/iu1xC2tMDvMGGzDnRecFW3hsiFmycwYW0mMxW4DwHILKZUPxFHcx9AC9RRzb/uO31z8K0fXcRBOUE8e+GrwGKjodEkXXLvGMMIoE+kqOfuYQ X-Forefront-Antispam-Report: CIP:255.255.255.255;CTRY:;LANG:en;SCL:1;SRV:;IPV:NLI;SFV:NSPM;H:AM6PR05MB5974.eurprd05.prod.outlook.com;PTR:;CAT:NONE;SFS:(4636009)(346002)(366004)(136003)(376002)(30864003)(5660300002)(16526019)(186003)(26005)(86362001)(110136005)(956004)(2616005)(54906003)(52116002)(8936002)(1076003)(6506007)(6512007)(508600001)(66556008)(34490700003)(66946007)(7416002)(2906002)(66476007)(83380400001)(4326008)(6666004)(6486002)(8676002)(36756003);DIR:OUT;SFP:1101; X-MS-Exchange-AntiSpam-MessageData: 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 X-OriginatorOrg: Mellanox.com X-MS-Exchange-CrossTenant-AuthSource: AM6PR05MB5974.eurprd05.prod.outlook.com X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 09 Dec 2020 16:03:03.7276 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: a652971c-7d2e-4d9b-a6a4-d149256f461b X-MS-Exchange-CrossTenant-Network-Message-Id: 17faab35-a867-4367-01c8-08d89c5be8fb X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: tenjwIu4nbABZmFRdriKTnFKAJ11p9o68Dwl6hrwUOjJ3HTjDLrFr32PYqWG4FmXYmydjhmb+Un/EY0DI288iw== X-MS-Exchange-Transport-CrossTenantHeadersStamped: AM6PR05MB4149 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org HTB doesn't scale well because of contention on a single lock, and it also consumes CPU. This patch adds support for offloading HTB to hardware that supports hierarchical rate limiting. This solution addresses two main problems of scaling HTB: 1. Contention by flow classification. Currently the filters are attached to the HTB instance as follows: # tc filter add dev eth0 parent 1:0 protocol ip flower dst_port 80 classid 1:10 It's possible to move classification to clsact egress hook, which is thread-safe and lock-free: # tc filter add dev eth0 egress protocol ip flower dst_port 80 action skbedit priority 1:10 This way classification still happens in software, but the lock contention is eliminated, and it happens before selecting the TX queue, allowing the driver to translate the class to the corresponding hardware queue. Note that this is already compatible with non-offloaded HTB and doesn't require changes to the kernel nor iproute2. 2. Contention by handling packets. HTB is not multi-queue, it attaches to a whole net device, and handling of all packets takes the same lock. When HTB is offloaded, its algorithm is done in hardware. HTB registers itself as a multi-queue qdisc, similarly to mq: HTB is attached to the netdev, and each queue has its own qdisc. The control flow is still done by HTB: it calls the driver via ndo_setup_tc to replicate the hierarchy of classes in the NIC. Leaf classes are presented by hardware queues. The data path works as follows: a packet is classified by clsact, the driver selects a hardware queue according to its class, and the packet is enqueued into this queue's qdisc. Some features of HTB may be not supported by some particular hardware, for example, the maximum number of classes may be limited, the granularity of rate and ceil parameters may be different, etc. - so, the offload is not enabled by default, a new parameter is used to enable it: # tc qdisc replace dev eth0 root handle 1: htb offload Signed-off-by: Maxim Mikityanskiy Reviewed-by: Tariq Toukan --- include/linux/netdevice.h | 1 + include/net/pkt_cls.h | 26 ++ include/uapi/linux/pkt_sched.h | 1 + net/sched/sch_htb.c | 461 +++++++++++++++++++++++++-- tools/include/uapi/linux/pkt_sched.h | 1 + 5 files changed, 462 insertions(+), 28 deletions(-) diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h index 7bf167993c05..6830a8d2dbe9 100644 --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -859,6 +859,7 @@ enum tc_setup_type { TC_SETUP_QDISC_ETS, TC_SETUP_QDISC_TBF, TC_SETUP_QDISC_FIFO, + TC_SETUP_QDISC_HTB, }; /* These structures hold the attributes of bpf state that are being passed diff --git a/include/net/pkt_cls.h b/include/net/pkt_cls.h index 0f2a9c44171c..0204ad3a1091 100644 --- a/include/net/pkt_cls.h +++ b/include/net/pkt_cls.h @@ -783,6 +783,32 @@ struct tc_mq_qopt_offload { }; }; +enum tc_htb_command { + /* Root */ + TC_HTB_CREATE, /* Initialize HTB offload. */ + TC_HTB_DESTROY, /* Destroy HTB offload. */ + /* Classes */ + TC_HTB_LEAF_ALLOC_QUEUE, /* Allocate qid and create leaf. */ + TC_HTB_LEAF_TO_INNER, /* Convert leaf to inner, preserve and return qid, create new leaf. */ + TC_HTB_LEAF_DEL, /* Delete leaf, while siblings remain. */ + TC_HTB_LEAF_DEL_LAST, /* Delete leaf, convert parent to leaf, preserving qid. */ + TC_HTB_NODE_MODIFY, /* Modify parameters of a node. */ + /* Class qdisc */ + TC_HTB_LEAF_QUERY_QUEUE, /* Query qid by classid. */ +}; + +struct tc_htb_qopt_offload { + enum tc_htb_command command; + u16 classid; + u32 parent_classid; + u16 qid; + u16 moved_qid; + u64 rate; + u64 ceil; +}; + +#define TC_HTB_CLASSID_ROOT U32_MAX + enum tc_red_command { TC_RED_REPLACE, TC_RED_DESTROY, diff --git a/include/uapi/linux/pkt_sched.h b/include/uapi/linux/pkt_sched.h index 9e7c2c607845..79a699f106b1 100644 --- a/include/uapi/linux/pkt_sched.h +++ b/include/uapi/linux/pkt_sched.h @@ -434,6 +434,7 @@ enum { TCA_HTB_RATE64, TCA_HTB_CEIL64, TCA_HTB_PAD, + TCA_HTB_OFFLOAD, __TCA_HTB_MAX, }; diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c index cd70dbcbd72f..7aa56a5e1e94 100644 --- a/net/sched/sch_htb.c +++ b/net/sched/sch_htb.c @@ -174,6 +174,11 @@ struct htb_sched { int row_mask[TC_HTB_MAXDEPTH]; struct htb_level hlevel[TC_HTB_MAXDEPTH]; + + struct Qdisc **direct_qdiscs; + unsigned int num_direct_qdiscs; + + bool offload; }; /* find class in global hash table using given handle */ @@ -957,7 +962,7 @@ static void htb_reset(struct Qdisc *sch) if (cl->level) memset(&cl->inner, 0, sizeof(cl->inner)); else { - if (cl->leaf.q) + if (cl->leaf.q && !q->offload) qdisc_reset(cl->leaf.q); } cl->prio_activity = 0; @@ -980,6 +985,7 @@ static const struct nla_policy htb_policy[TCA_HTB_MAX + 1] = { [TCA_HTB_DIRECT_QLEN] = { .type = NLA_U32 }, [TCA_HTB_RATE64] = { .type = NLA_U64 }, [TCA_HTB_CEIL64] = { .type = NLA_U64 }, + [TCA_HTB_OFFLOAD] = { .type = NLA_FLAG }, }; static void htb_work_func(struct work_struct *work) @@ -992,12 +998,27 @@ static void htb_work_func(struct work_struct *work) rcu_read_unlock(); } +static void htb_set_lockdep_class_child(struct Qdisc *q) +{ + static struct lock_class_key child_key; + + lockdep_set_class(qdisc_lock(q), &child_key); +} + +static int htb_offload(struct net_device *dev, struct tc_htb_qopt_offload *opt) +{ + return dev->netdev_ops->ndo_setup_tc(dev, TC_SETUP_QDISC_HTB, opt); +} + static int htb_init(struct Qdisc *sch, struct nlattr *opt, struct netlink_ext_ack *extack) { + struct net_device *dev = qdisc_dev(sch); + struct tc_htb_qopt_offload offload_opt; struct htb_sched *q = qdisc_priv(sch); struct nlattr *tb[TCA_HTB_MAX + 1]; struct tc_htb_glob *gopt; + unsigned int ntx; int err; qdisc_watchdog_init(&q->watchdog, sch); @@ -1022,9 +1043,22 @@ static int htb_init(struct Qdisc *sch, struct nlattr *opt, if (gopt->version != HTB_VER >> 16) return -EINVAL; + q->offload = nla_get_flag(tb[TCA_HTB_OFFLOAD]); + + if (q->offload) { + if (!tc_can_offload(dev) || !dev->netdev_ops->ndo_setup_tc) + return -EOPNOTSUPP; + + q->num_direct_qdiscs = dev->real_num_tx_queues; + q->direct_qdiscs = kcalloc(q->num_direct_qdiscs, + sizeof(*q->direct_qdiscs), GFP_KERNEL); + if (!q->direct_qdiscs) + return -ENOMEM; + } + err = qdisc_class_hash_init(&q->clhash); if (err < 0) - return err; + goto err_free_direct_qdiscs; qdisc_skb_head_init(&q->direct_queue); @@ -1037,7 +1071,104 @@ static int htb_init(struct Qdisc *sch, struct nlattr *opt, q->rate2quantum = 1; q->defcls = gopt->defcls; + if (!q->offload) + return 0; + + for (ntx = 0; ntx < q->num_direct_qdiscs; ntx++) { + struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, ntx); + struct Qdisc *qdisc; + + qdisc = qdisc_create_dflt(dev_queue, &pfifo_qdisc_ops, + TC_H_MAKE(sch->handle, 0), extack); + if (!qdisc) { + err = -ENOMEM; + goto err_free_qdiscs; + } + + htb_set_lockdep_class_child(qdisc); + q->direct_qdiscs[ntx] = qdisc; + qdisc->flags |= TCQ_F_ONETXQUEUE | TCQ_F_NOPARENT; + } + + sch->flags |= TCQ_F_MQROOT; + + offload_opt = (struct tc_htb_qopt_offload) { + .command = TC_HTB_CREATE, + .parent_classid = TC_H_MAJ(sch->handle) >> 16, + .classid = TC_H_MIN(q->defcls), + }; + err = htb_offload(dev, &offload_opt); + if (err) + goto err_free_qdiscs; + return 0; + +err_free_qdiscs: + /* TC_HTB_CREATE call failed, avoid any further calls to the driver. */ + q->offload = false; + + for (ntx = 0; ntx < q->num_direct_qdiscs && q->direct_qdiscs[ntx]; ntx++) + qdisc_put(q->direct_qdiscs[ntx]); + + qdisc_class_hash_destroy(&q->clhash); + /* Prevent use-after-free and double-free when htb_destroy gets called. */ + q->clhash.hash = NULL; + q->clhash.hashsize = 0; + +err_free_direct_qdiscs: + kfree(q->direct_qdiscs); + q->direct_qdiscs = NULL; + return err; +} + +static void htb_attach_offload(struct Qdisc *sch) +{ + struct net_device *dev = qdisc_dev(sch); + struct htb_sched *q = qdisc_priv(sch); + unsigned int ntx; + + for (ntx = 0; ntx < q->num_direct_qdiscs; ntx++) { + struct Qdisc *old, *qdisc = q->direct_qdiscs[ntx]; + + old = dev_graft_qdisc(qdisc->dev_queue, qdisc); + qdisc_put(old); + qdisc_hash_add(qdisc, false); + } + for (ntx = q->num_direct_qdiscs; ntx < dev->num_tx_queues; ntx++) { + struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, ntx); + struct Qdisc *old = dev_graft_qdisc(dev_queue, NULL); + + qdisc_put(old); + } + + kfree(q->direct_qdiscs); + q->direct_qdiscs = NULL; +} + +static void htb_attach_software(struct Qdisc *sch) +{ + struct net_device *dev = qdisc_dev(sch); + unsigned int ntx; + + /* Resemble qdisc_graft behavior. */ + for (ntx = 0; ntx < dev->num_tx_queues; ntx++) { + struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, ntx); + struct Qdisc *old = dev_graft_qdisc(dev_queue, sch); + + qdisc_refcount_inc(sch); + + qdisc_put(old); + } +} + +static void htb_attach(struct Qdisc *sch) +{ + struct htb_sched *q = qdisc_priv(sch); + + if (q->offload) + htb_attach_offload(sch); + else + htb_attach_software(sch); } static int htb_dump(struct Qdisc *sch, struct sk_buff *skb) @@ -1046,6 +1177,11 @@ static int htb_dump(struct Qdisc *sch, struct sk_buff *skb) struct nlattr *nest; struct tc_htb_glob gopt; + if (q->offload) + sch->flags |= TCQ_F_OFFLOADED; + else + sch->flags &= ~TCQ_F_OFFLOADED; + sch->qstats.overlimits = q->overlimits; /* Its safe to not acquire qdisc lock. As we hold RTNL, * no change can happen on the qdisc parameters. @@ -1063,6 +1199,8 @@ static int htb_dump(struct Qdisc *sch, struct sk_buff *skb) if (nla_put(skb, TCA_HTB_INIT, sizeof(gopt), &gopt) || nla_put_u32(skb, TCA_HTB_DIRECT_QLEN, q->direct_qlen)) goto nla_put_failure; + if (q->offload && nla_put_flag(skb, TCA_HTB_OFFLOAD)) + goto nla_put_failure; return nla_nest_end(skb, nest); @@ -1144,19 +1282,95 @@ htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d) return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats)); } +static struct netdev_queue *htb_select_queue(struct Qdisc *sch, struct tcmsg *tcm) +{ + struct net_device *dev = qdisc_dev(sch); + struct tc_htb_qopt_offload offload_opt; + int err; + + offload_opt = (struct tc_htb_qopt_offload) { + .command = TC_HTB_LEAF_QUERY_QUEUE, + .classid = TC_H_MIN(tcm->tcm_parent), + }; + err = htb_offload(dev, &offload_opt); + if (err || offload_opt.qid >= dev->num_tx_queues) + return NULL; + return netdev_get_tx_queue(dev, offload_opt.qid); +} + +static struct Qdisc *htb_graft_helper(struct netdev_queue *dev_queue, struct Qdisc *new_q) +{ + struct net_device *dev = dev_queue->dev; + struct Qdisc *old_q; + + if (dev->flags & IFF_UP) + dev_deactivate(dev); + old_q = dev_graft_qdisc(dev_queue, new_q); + if (new_q) + new_q->flags |= TCQ_F_ONETXQUEUE | TCQ_F_NOPARENT; + if (dev->flags & IFF_UP) + dev_activate(dev); + + return old_q; +} + +static void htb_offload_move_qdisc(struct Qdisc *sch, u16 qid_old, u16 qid_new) +{ + struct netdev_queue *queue_old, *queue_new; + struct net_device *dev = qdisc_dev(sch); + struct Qdisc *qdisc; + + queue_old = netdev_get_tx_queue(dev, qid_old); + queue_new = netdev_get_tx_queue(dev, qid_new); + + if (dev->flags & IFF_UP) + dev_deactivate(dev); + qdisc = dev_graft_qdisc(queue_old, NULL); + qdisc->dev_queue = queue_new; + qdisc = dev_graft_qdisc(queue_new, qdisc); + if (dev->flags & IFF_UP) + dev_activate(dev); + + WARN_ON(!(qdisc->flags & TCQ_F_BUILTIN)); +} + static int htb_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new, struct Qdisc **old, struct netlink_ext_ack *extack) { + struct netdev_queue *dev_queue = sch->dev_queue; struct htb_class *cl = (struct htb_class *)arg; + struct htb_sched *q = qdisc_priv(sch); + struct Qdisc *old_q; if (cl->level) return -EINVAL; - if (new == NULL && - (new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, - cl->common.classid, extack)) == NULL) - return -ENOBUFS; + + if (q->offload) { + dev_queue = new->dev_queue; + WARN_ON(dev_queue != cl->leaf.q->dev_queue); + } + + if (!new) { + new = qdisc_create_dflt(dev_queue, &pfifo_qdisc_ops, + cl->common.classid, extack); + if (!new) + return -ENOBUFS; + } + + if (q->offload) { + htb_set_lockdep_class_child(new); + /* One ref for cl->leaf.q, the other for dev_queue->qdisc. */ + qdisc_refcount_inc(new); + old_q = htb_graft_helper(dev_queue, new); + } *old = qdisc_replace(sch, new, &cl->leaf.q); + + if (q->offload) { + WARN_ON(old_q != *old); + qdisc_put(old_q); + } + return 0; } @@ -1184,9 +1398,10 @@ static inline int htb_parent_last_child(struct htb_class *cl) return 1; } -static void htb_parent_to_leaf(struct htb_sched *q, struct htb_class *cl, +static void htb_parent_to_leaf(struct Qdisc *sch, struct htb_class *cl, struct Qdisc *new_q) { + struct htb_sched *q = qdisc_priv(sch); struct htb_class *parent = cl->parent; WARN_ON(cl->level || !cl->leaf.q || cl->prio_activity); @@ -1204,6 +1419,58 @@ static void htb_parent_to_leaf(struct htb_sched *q, struct htb_class *cl, parent->cmode = HTB_CAN_SEND; } +static void htb_parent_to_leaf_offload(struct Qdisc *sch, struct netdev_queue *dev_queue, + struct Qdisc *new_q) +{ + struct Qdisc *old_q; + + /* One ref for cl->leaf.q, the other for dev_queue->qdisc. */ + qdisc_refcount_inc(new_q); + old_q = htb_graft_helper(dev_queue, new_q); + WARN_ON(!(old_q->flags & TCQ_F_BUILTIN)); +} + +static void htb_destroy_class_offload(struct Qdisc *sch, struct htb_class *cl, + bool last_child, bool destroying) +{ + struct tc_htb_qopt_offload offload_opt; + struct Qdisc *q = cl->leaf.q; + struct Qdisc *old = NULL; + + if (cl->level) + return; + + WARN_ON(!q); + if (!destroying) { + /* On destroy of HTB, two cases are possible: + * 1. q is a normal qdisc, but q->dev_queue has noop qdisc. + * 2. q is a noop qdisc (for nodes that were inner), + * q->dev_queue is noop_netdev_queue. + */ + old = htb_graft_helper(q->dev_queue, NULL); + WARN_ON(!old); + WARN_ON(old != q); + } + + offload_opt = (struct tc_htb_qopt_offload) { + .command = last_child ? TC_HTB_LEAF_DEL_LAST : TC_HTB_LEAF_DEL, + .classid = cl->common.classid, + }; + htb_offload(qdisc_dev(sch), &offload_opt); + + qdisc_put(old); + + if (last_child) + return; + + if (offload_opt.moved_qid != 0) { + if (destroying) + q->dev_queue = netdev_get_tx_queue(qdisc_dev(sch), offload_opt.qid); + else + htb_offload_move_qdisc(sch, offload_opt.moved_qid, offload_opt.qid); + } +} + static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl) { if (!cl->level) { @@ -1217,8 +1484,11 @@ static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl) static void htb_destroy(struct Qdisc *sch) { + struct net_device *dev = qdisc_dev(sch); + struct tc_htb_qopt_offload offload_opt; struct htb_sched *q = qdisc_priv(sch); struct hlist_node *next; + bool nonempty, changed; struct htb_class *cl; unsigned int i; @@ -1237,13 +1507,56 @@ static void htb_destroy(struct Qdisc *sch) cl->block = NULL; } } - for (i = 0; i < q->clhash.hashsize; i++) { - hlist_for_each_entry_safe(cl, next, &q->clhash.hash[i], - common.hnode) - htb_destroy_class(sch, cl); - } + + do { + nonempty = false; + changed = false; + for (i = 0; i < q->clhash.hashsize; i++) { + hlist_for_each_entry_safe(cl, next, &q->clhash.hash[i], + common.hnode) { + bool last_child; + + if (!q->offload) { + htb_destroy_class(sch, cl); + continue; + } + + nonempty = true; + + if (cl->level) + continue; + + changed = true; + + last_child = htb_parent_last_child(cl); + htb_destroy_class_offload(sch, cl, last_child, true); + qdisc_class_hash_remove(&q->clhash, &cl->common); + if (cl->parent) + cl->parent->children--; + if (last_child) + htb_parent_to_leaf(sch, cl, NULL); + htb_destroy_class(sch, cl); + } + } + } while (changed); + WARN_ON(nonempty); + qdisc_class_hash_destroy(&q->clhash); __qdisc_reset_queue(&q->direct_queue); + + if (!q->offload) + return; + + offload_opt = (struct tc_htb_qopt_offload) { + .command = TC_HTB_DESTROY, + }; + htb_offload(dev, &offload_opt); + + if (!q->direct_qdiscs) + return; + for (i = 0; i < q->num_direct_qdiscs && q->direct_qdiscs[i]; i++) + qdisc_put(q->direct_qdiscs[i]); + kfree(q->direct_qdiscs); } static int htb_delete(struct Qdisc *sch, unsigned long arg) @@ -1260,11 +1573,24 @@ static int htb_delete(struct Qdisc *sch, unsigned long arg) if (cl->children || cl->filter_cnt) return -EBUSY; - if (!cl->level && htb_parent_last_child(cl)) { - new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, + if (!cl->level && htb_parent_last_child(cl)) + last_child = 1; + + if (q->offload) + htb_destroy_class_offload(sch, cl, last_child, false); + + if (last_child) { + struct netdev_queue *dev_queue; + + dev_queue = q->offload ? cl->leaf.q->dev_queue : sch->dev_queue; + new_q = qdisc_create_dflt(dev_queue, &pfifo_qdisc_ops, cl->parent->common.classid, NULL); - last_child = 1; + if (q->offload) { + if (new_q) + htb_set_lockdep_class_child(new_q); + htb_parent_to_leaf_offload(sch, dev_queue, new_q); + } } sch_tree_lock(sch); @@ -1285,7 +1611,7 @@ static int htb_delete(struct Qdisc *sch, unsigned long arg) &q->hlevel[cl->level].wait_pq); if (last_child) - htb_parent_to_leaf(q, cl, new_q); + htb_parent_to_leaf(sch, cl, new_q); sch_tree_unlock(sch); @@ -1300,9 +1626,11 @@ static int htb_change_class(struct Qdisc *sch, u32 classid, int err = -EINVAL; struct htb_sched *q = qdisc_priv(sch); struct htb_class *cl = (struct htb_class *)*arg, *parent; + struct tc_htb_qopt_offload offload_opt; struct nlattr *opt = tca[TCA_OPTIONS]; struct nlattr *tb[TCA_HTB_MAX + 1]; struct Qdisc *parent_qdisc = NULL; + struct netdev_queue *dev_queue; struct tc_htb_opt *hopt; u64 rate64, ceil64; int warn = 0; @@ -1335,8 +1663,12 @@ static int htb_change_class(struct Qdisc *sch, u32 classid, qdisc_put_rtab(qdisc_get_rtab(&hopt->ceil, tb[TCA_HTB_CTAB], NULL)); + rate64 = tb[TCA_HTB_RATE64] ? nla_get_u64(tb[TCA_HTB_RATE64]) : 0; + ceil64 = tb[TCA_HTB_CEIL64] ? nla_get_u64(tb[TCA_HTB_CEIL64]) : 0; + if (!cl) { /* new class */ - struct Qdisc *new_q; + struct net_device *dev = qdisc_dev(sch); + struct Qdisc *new_q, *old_q; int prio; struct { struct nlattr nla; @@ -1379,11 +1711,8 @@ static int htb_change_class(struct Qdisc *sch, u32 classid, NULL, qdisc_root_sleeping_running(sch), tca[TCA_RATE] ? : &est.nla); - if (err) { - tcf_block_put(cl->block); - kfree(cl); - goto failure; - } + if (err) + goto err_block_put; } cl->children = 0; @@ -1392,12 +1721,64 @@ static int htb_change_class(struct Qdisc *sch, u32 classid, for (prio = 0; prio < TC_HTB_NUMPRIO; prio++) RB_CLEAR_NODE(&cl->node[prio]); + cl->common.classid = classid; + + /* Make sure nothing interrupts us in between of two ndo_setup_tc calls. */ + ASSERT_RTNL(); + /* create leaf qdisc early because it uses kmalloc(GFP_KERNEL) * so that can't be used inside of sch_tree_lock * -- thanks to Karlis Peisenieks */ - new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, + if (!q->offload) { + dev_queue = sch->dev_queue; + } else if (!(parent && !parent->level)) { + /* Assign a dev_queue to this classid. */ + offload_opt = (struct tc_htb_qopt_offload) { + .command = TC_HTB_LEAF_ALLOC_QUEUE, + .classid = cl->common.classid, + .parent_classid = parent ? + TC_H_MIN(parent->common.classid) : + TC_HTB_CLASSID_ROOT, + .rate = max_t(u64, hopt->rate.rate, rate64), + .ceil = max_t(u64, hopt->ceil.rate, ceil64), + }; + err = htb_offload(dev, &offload_opt); + if (err) { + pr_err("htb: TC_HTB_LEAF_ALLOC_QUEUE failed with err = %d\n", err); + goto err_kill_estimator; + } + dev_queue = netdev_get_tx_queue(dev, offload_opt.qid); + } else { /* First child. */ + dev_queue = parent->leaf.q->dev_queue; + old_q = htb_graft_helper(dev_queue, NULL); + WARN_ON(old_q != parent->leaf.q); + offload_opt = (struct tc_htb_qopt_offload) { + .command = TC_HTB_LEAF_TO_INNER, + .classid = cl->common.classid, + .parent_classid = TC_H_MIN(parent->common.classid), + .rate = max_t(u64, hopt->rate.rate, rate64), + .ceil = max_t(u64, hopt->ceil.rate, ceil64), + }; + err = htb_offload(dev, &offload_opt); + if (err) { + pr_err("htb: TC_HTB_LEAF_TO_INNER failed with err = %d\n", err); + htb_graft_helper(dev_queue, old_q); + goto err_kill_estimator; + } + qdisc_put(old_q); + } + new_q = qdisc_create_dflt(dev_queue, &pfifo_qdisc_ops, classid, NULL); + if (q->offload) { + if (new_q) { + htb_set_lockdep_class_child(new_q); + /* One ref for cl->leaf.q, the other for dev_queue->qdisc. */ + qdisc_refcount_inc(new_q); + } + old_q = htb_graft_helper(dev_queue, new_q); + WARN_ON(!(old_q->flags & TCQ_F_BUILTIN)); /* No qdisc_put needed. */ + } sch_tree_lock(sch); if (parent && !parent->level) { /* turn parent into inner node */ @@ -1415,10 +1796,10 @@ static int htb_change_class(struct Qdisc *sch, u32 classid, : TC_HTB_MAXDEPTH) - 1; memset(&parent->inner, 0, sizeof(parent->inner)); } + /* leaf (we) needs elementary qdisc */ cl->leaf.q = new_q ? new_q : &noop_qdisc; - cl->common.classid = classid; cl->parent = parent; /* set class to be in HTB_CAN_SEND state */ @@ -1444,12 +1825,29 @@ static int htb_change_class(struct Qdisc *sch, u32 classid, if (err) return err; } - sch_tree_lock(sch); - } - rate64 = tb[TCA_HTB_RATE64] ? nla_get_u64(tb[TCA_HTB_RATE64]) : 0; + if (q->offload) { + struct net_device *dev = qdisc_dev(sch); - ceil64 = tb[TCA_HTB_CEIL64] ? nla_get_u64(tb[TCA_HTB_CEIL64]) : 0; + offload_opt = (struct tc_htb_qopt_offload) { + .command = TC_HTB_NODE_MODIFY, + .classid = cl->common.classid, + .rate = max_t(u64, hopt->rate.rate, rate64), + .ceil = max_t(u64, hopt->ceil.rate, ceil64), + }; + err = htb_offload(dev, &offload_opt); + if (err) + /* Estimator was replaced, and rollback may fail + * as well, so we don't try to recover it, and + * the estimator won't work property with the + * offload anyway, because bstats are updated + * only when the stats are queried. + */ + return err; + } + + sch_tree_lock(sch); + } psched_ratecfg_precompute(&cl->rate, &hopt->rate, rate64); psched_ratecfg_precompute(&cl->ceil, &hopt->ceil, ceil64); @@ -1492,6 +1890,11 @@ static int htb_change_class(struct Qdisc *sch, u32 classid, *arg = (unsigned long)cl; return 0; +err_kill_estimator: + gen_kill_estimator(&cl->rate_est); +err_block_put: + tcf_block_put(cl->block); + kfree(cl); failure: return err; } @@ -1557,6 +1960,7 @@ static void htb_walk(struct Qdisc *sch, struct qdisc_walker *arg) } static const struct Qdisc_class_ops htb_class_ops = { + .select_queue = htb_select_queue, .graft = htb_graft, .leaf = htb_leaf, .qlen_notify = htb_qlen_notify, @@ -1579,6 +1983,7 @@ static struct Qdisc_ops htb_qdisc_ops __read_mostly = { .dequeue = htb_dequeue, .peek = qdisc_peek_dequeued, .init = htb_init, + .attach = htb_attach, .reset = htb_reset, .destroy = htb_destroy, .dump = htb_dump, diff --git a/tools/include/uapi/linux/pkt_sched.h b/tools/include/uapi/linux/pkt_sched.h index 0d18b1d1fbbc..5c903abc9fa5 100644 --- a/tools/include/uapi/linux/pkt_sched.h +++ b/tools/include/uapi/linux/pkt_sched.h @@ -414,6 +414,7 @@ enum { TCA_HTB_RATE64, TCA_HTB_CEIL64, TCA_HTB_PAD, + TCA_HTB_OFFLOAD, __TCA_HTB_MAX, }; From patchwork Wed Dec 9 16:02:50 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Maxim Mikityanskiy X-Patchwork-Id: 11961811 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-18.7 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER, INCLUDES_PATCH,MAILING_LIST_MULTI,MSGID_FROM_MTA_HEADER,SPF_HELO_NONE, SPF_PASS,URIBL_BLOCKED,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 6059BC433FE for ; Wed, 9 Dec 2020 16:04:52 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 1F5EA2343B for ; Wed, 9 Dec 2020 16:04:52 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1730969AbgLIQEt (ORCPT ); Wed, 9 Dec 2020 11:04:49 -0500 Received: from mail-eopbgr60048.outbound.protection.outlook.com ([40.107.6.48]:36676 "EHLO EUR04-DB3-obe.outbound.protection.outlook.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1728392AbgLIQEh (ORCPT ); Wed, 9 Dec 2020 11:04:37 -0500 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=SSsU+59XmI+mOR4ibZserrq74hShDwWISRP1m8Ygv21+YdfcSSvovnDWV25ugoVgDAaJRid88BFhcOYyYYJ0TJJRWMX2RzrhFqO3N5to0P/P5eVG5LnHXOh6oFzQRCCGCtQ7je49K4bcfM+o+5mJ39uQitW/xNi6ijSlXXAl46aMYVHMBAGW8QGcMBmm/0sQ+4qAcMK9iCtEYx6bPe1fe5yDIY29jlK4wBonbjpiMCV727rFJanE/bJ2/+rGtb/gQfkbzLLovqpoPYViuvw8zpsFso7oN04AEyEGVJ8zxo+wq7lm9toXkl11wPI37AF8MxW7TxDDHUh/Aag5XswLMw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=cY5SSxJcE/uGT+EfhKaXrSu263O2sEMbUzZdWlb+H2E=; b=a2kmVw2XRa0agZudnAhG5qMXv1JhgQXSlX7ckYdr+UBM4KSOSVZ3Z/NKGzwvz/TA6mU6zzOnHVSkBc15DsUrzLfS93p9Es67UelRCcolPY7d4pyIvRUgAxuJIE1+MEN9tY38Jr0o2oucS1K+aQ3J4u6iD1iiVNBFx0S1K/4mQpa/T5LbKnnDo1Zfr4KmtpQ3X9Dt//hZ/7M2pA6DsdvF3G/7ku4vjqQzrdjv8DEiF70geAWvJnkvzE+Tz2PhVCeYemFVNKFTuEwbUbHuNLVEwf1o+Pj8sH54PU8l9fQseuC3DfDbj7UCIr2RqcnqRFtu2MMBOxxo0iyFOejltle0+Q== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=mellanox.com; dmarc=pass action=none header.from=mellanox.com; dkim=pass header.d=mellanox.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=Mellanox.com; s=selector1; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=cY5SSxJcE/uGT+EfhKaXrSu263O2sEMbUzZdWlb+H2E=; b=XSHO6qPl2/0i9oDkaV9QXkzzCKA+CPLcFd0x48g+i9uENICERQNHnjL/i/SNsGezrbsW3EvGuoWxCzc9IPcaFQ31ILLoMkUGW5s+s/KeiNVOFq5i9WwMQj5JXyG1nKxGadR1wmKbRz9t1S0AvgbMn3BB8SYhLRnI1CmECWcxzZk= Authentication-Results: davemloft.net; dkim=none (message not signed) header.d=none;davemloft.net; dmarc=none action=none header.from=mellanox.com; Received: from AM6PR05MB5974.eurprd05.prod.outlook.com (2603:10a6:20b:a7::12) by AM6PR05MB4149.eurprd05.prod.outlook.com (2603:10a6:209:50::21) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.3632.23; Wed, 9 Dec 2020 16:03:05 +0000 Received: from AM6PR05MB5974.eurprd05.prod.outlook.com ([fe80::642a:6259:153:ab88]) by AM6PR05MB5974.eurprd05.prod.outlook.com ([fe80::642a:6259:153:ab88%3]) with mapi id 15.20.3632.024; Wed, 9 Dec 2020 16:03:05 +0000 From: Maxim Mikityanskiy To: "David S. Miller" , Jamal Hadi Salim , Cong Wang , Jiri Pirko Cc: Saeed Mahameed , Jakub Kicinski , Tariq Toukan , Maxim Mikityanskiy , netdev@vger.kernel.org, Maxim Mikityanskiy , Tariq Toukan Subject: [PATCH net-next 3/4] sch_htb: Stats for offloaded HTB Date: Wed, 9 Dec 2020 18:02:50 +0200 Message-Id: <20201209160251.19054-4-maximmi@mellanox.com> X-Mailer: git-send-email 2.20.1 In-Reply-To: <20201209160251.19054-1-maximmi@mellanox.com> References: <20201209160251.19054-1-maximmi@mellanox.com> X-Originating-IP: [94.188.199.18] X-ClientProxiedBy: AM0PR04CA0098.eurprd04.prod.outlook.com (2603:10a6:208:be::39) To AM6PR05MB5974.eurprd05.prod.outlook.com (2603:10a6:20b:a7::12) MIME-Version: 1.0 X-MS-Exchange-MessageSentRepresentingType: 1 Received: from dev-l-vrt-208.mtl.labs.mlnx (94.188.199.18) by AM0PR04CA0098.eurprd04.prod.outlook.com (2603:10a6:208:be::39) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.3654.12 via Frontend Transport; Wed, 9 Dec 2020 16:03:04 +0000 X-MS-PublicTrafficType: Email X-MS-Office365-Filtering-HT: Tenant X-MS-Office365-Filtering-Correlation-Id: 9a1589d3-f2ca-4207-8b7b-08d89c5bea46 X-MS-TrafficTypeDiagnostic: AM6PR05MB4149: X-LD-Processed: a652971c-7d2e-4d9b-a6a4-d149256f461b,ExtFwd,ExtAddr X-MS-Exchange-Transport-Forked: True X-Microsoft-Antispam-PRVS: X-MS-Oob-TLC-OOBClassifiers: OLM:12; X-MS-Exchange-SenderADCheck: 1 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: TOAs5P13GhuwtEPztUrrXYdp/0jZCGgfZfCx9+R3uoKuuj0Svt9dtsmCp0qLKILyl+EVGa/YoId+7QAQ31RkZplGTD5is2XomZ3BjB9/nPVe5j/zlYz3RvmCgprznrP1LdVPmqusGmO8krqydVmhpprfizXvGMrGdSpJhRW0M4r96zxYar4Ou0iy5l7AyCCdufpOtSfUi89uhQ9TaxCMMmmW951MN1PC03l2e94EGtIM554Y9Xi7/I2z1Hzrifkb6tTzg72+hosTVnVzGCsKkaplZQPdSjk2XhrLRGhoeCz7507Yj/9Ivi1WM6JJAdjUYkSz0PQG8wDjfUr8p6DgnrjPvdkTTG7chc5nKEz8Taekou1tUiYxX2V2xxl65wbd X-Forefront-Antispam-Report: CIP:255.255.255.255;CTRY:;LANG:en;SCL:1;SRV:;IPV:NLI;SFV:NSPM;H:AM6PR05MB5974.eurprd05.prod.outlook.com;PTR:;CAT:NONE;SFS:(4636009)(346002)(366004)(136003)(376002)(5660300002)(16526019)(186003)(26005)(86362001)(110136005)(956004)(2616005)(54906003)(52116002)(8936002)(1076003)(6506007)(6512007)(508600001)(66556008)(34490700003)(66946007)(7416002)(2906002)(66476007)(83380400001)(4326008)(6666004)(6486002)(8676002)(36756003);DIR:OUT;SFP:1101; X-MS-Exchange-AntiSpam-MessageData: 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 X-OriginatorOrg: Mellanox.com X-MS-Exchange-CrossTenant-AuthSource: AM6PR05MB5974.eurprd05.prod.outlook.com X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 09 Dec 2020 16:03:05.6775 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: a652971c-7d2e-4d9b-a6a4-d149256f461b X-MS-Exchange-CrossTenant-Network-Message-Id: 9a1589d3-f2ca-4207-8b7b-08d89c5bea46 X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: LKD2joWX1MVfraeCE6JLhbznXv+CQ7MBuAmFEsiZmvm5MqiIdBB1QksE8phj67S9z+GP20IWihUjFPV6XsJkDA== X-MS-Exchange-Transport-CrossTenantHeadersStamped: AM6PR05MB4149 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org This commit adds support for statistics of offloaded HTB. Bytes and packets counters for leaf and inner nodes are supported, the values are taken from per-queue qdiscs, and the numbers that the user sees should have the same behavior as the software (non-offloaded) HTB. Signed-off-by: Maxim Mikityanskiy Reviewed-by: Tariq Toukan Reported-by: kernel test robot Reported-by: Dan Carpenter --- net/sched/sch_htb.c | 49 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 49 insertions(+) diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c index 7aa56a5e1e94..c2e7efc2af88 100644 --- a/net/sched/sch_htb.c +++ b/net/sched/sch_htb.c @@ -114,6 +114,7 @@ struct htb_class { * Written often fields */ struct gnet_stats_basic_packed bstats; + struct gnet_stats_basic_packed bstats_bias; struct tc_htb_xstats xstats; /* our special stats */ /* token bucket parameters */ @@ -1213,6 +1214,7 @@ static int htb_dump_class(struct Qdisc *sch, unsigned long arg, struct sk_buff *skb, struct tcmsg *tcm) { struct htb_class *cl = (struct htb_class *)arg; + struct htb_sched *q = qdisc_priv(sch); struct nlattr *nest; struct tc_htb_opt opt; @@ -1239,6 +1241,8 @@ static int htb_dump_class(struct Qdisc *sch, unsigned long arg, opt.level = cl->level; if (nla_put(skb, TCA_HTB_PARMS, sizeof(opt), &opt)) goto nla_put_failure; + if (q->offload && nla_put_flag(skb, TCA_HTB_OFFLOAD)) + goto nla_put_failure; if ((cl->rate.rate_bytes_ps >= (1ULL << 32)) && nla_put_u64_64bit(skb, TCA_HTB_RATE64, cl->rate.rate_bytes_ps, TCA_HTB_PAD)) @@ -1255,10 +1259,38 @@ static int htb_dump_class(struct Qdisc *sch, unsigned long arg, return -1; } +static void htb_offload_aggregate_stats(struct htb_sched *q, struct htb_class *cl) +{ + struct htb_class *c; + unsigned int i; + + memset(&cl->bstats, 0, sizeof(cl->bstats)); + + for (i = 0; i < q->clhash.hashsize; i++) { + hlist_for_each_entry(c, &q->clhash.hash[i], common.hnode) { + struct htb_class *p = c; + + while (p && p->level < cl->level) + p = p->parent; + + if (p != cl) + continue; + + cl->bstats.bytes += c->bstats_bias.bytes; + cl->bstats.packets += c->bstats_bias.packets; + if (c->level == 0) { + cl->bstats.bytes += c->leaf.q->bstats.bytes; + cl->bstats.packets += c->leaf.q->bstats.packets; + } + } + } +} + static int htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d) { struct htb_class *cl = (struct htb_class *)arg; + struct htb_sched *q = qdisc_priv(sch); struct gnet_stats_queue qs = { .drops = cl->drops, .overlimits = cl->overlimits, @@ -1273,6 +1305,16 @@ htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d) cl->xstats.ctokens = clamp_t(s64, PSCHED_NS2TICKS(cl->ctokens), INT_MIN, INT_MAX); + if (q->offload) { + if (!cl->level) { + cl->bstats = cl->leaf.q->bstats; + cl->bstats.bytes += cl->bstats_bias.bytes; + cl->bstats.packets += cl->bstats_bias.packets; + } else { + htb_offload_aggregate_stats(q, cl); + } + } + if (gnet_stats_copy_basic(qdisc_root_sleeping_running(sch), d, NULL, &cl->bstats) < 0 || gnet_stats_copy_rate_est(d, &cl->rate_est) < 0 || @@ -1452,6 +1494,11 @@ static void htb_destroy_class_offload(struct Qdisc *sch, struct htb_class *cl, WARN_ON(old != q); } + if (cl->parent) { + cl->parent->bstats_bias.bytes += q->bstats.bytes; + cl->parent->bstats_bias.packets += q->bstats.packets; + } + offload_opt = (struct tc_htb_qopt_offload) { .command = last_child ? TC_HTB_LEAF_DEL_LAST : TC_HTB_LEAF_DEL, .classid = cl->common.classid, @@ -1766,6 +1813,8 @@ static int htb_change_class(struct Qdisc *sch, u32 classid, htb_graft_helper(dev_queue, old_q); goto err_kill_estimator; } + parent->bstats_bias.bytes += old_q->bstats.bytes; + parent->bstats_bias.packets += old_q->bstats.packets; qdisc_put(old_q); } new_q = qdisc_create_dflt(dev_queue, &pfifo_qdisc_ops, From patchwork Wed Dec 9 16:02:51 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Maxim Mikityanskiy X-Patchwork-Id: 11961807 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-18.7 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER, INCLUDES_PATCH,MAILING_LIST_MULTI,MSGID_FROM_MTA_HEADER,SPF_HELO_NONE, SPF_PASS,URIBL_BLOCKED,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id EBBB7C433FE for ; Wed, 9 Dec 2020 16:04:13 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 9456A23BC6 for ; Wed, 9 Dec 2020 16:04:13 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1729957AbgLIQEN (ORCPT ); Wed, 9 Dec 2020 11:04:13 -0500 Received: from mail-eopbgr60044.outbound.protection.outlook.com ([40.107.6.44]:23174 "EHLO EUR04-DB3-obe.outbound.protection.outlook.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1727024AbgLIQEM (ORCPT ); Wed, 9 Dec 2020 11:04:12 -0500 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=DkA43mgMY9QZMWwKY6vxScri5pUJPgsRIX8n7J4owCkuQuNWlbnTVD5G/K09T8EkfnHcm4JwgIsYTox8KEX5CdcOZxdnC0vfVtI6sVsiMUbdvTHvSiM4utJiLqNMFxm3CTfQYbM6l1XxG73wWJIKRZDLNAJSW5sS7H8VF1bbmcgP0Ig4GNizffKo6atuy+50j37q7zoTC3XmFvpW4Jy1Qs8hh8lwcQIgm1o8u5zUOFXfLbvWbMmKM84Vxk8R7TxsnSPleQXd0cxfactGh3nS45xgbUDrj9W8wrazlPfO0ExfEX1szAyv+Rhzn4bM12FATf3n5CAyIBzXykQrQw3sVw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=h0ackAdw5xxhjagVxEO0AjpYs9LAY8IjurhwKDSIrAE=; b=BRQfF+l5UoczuiYkT8S52s0csDgujtI60gVFRyGNYzF+ZTYN7pgH0WbsDitQUWJXSzwaPyp26FOHvhlM4sEhzAf8U14pQeq97nKWzAl5r/3TsKefsQ/P4g99QVhym1WEcd5Ky3oiiMn/9FLgIQBAwgUOKHQlQAaAe4G0P8o1ihyukggtZpmL/9nRNzbC66f60fy7zvyX6dHLdW5uGpx2/PCW9UyGT+V/WEqLBAopwLmcOkPie6j0CtqeFUbdpV3Hvq/ldlUzWHkzQx+Qcxkvx8/96l+ZcH9j9Mh40D3ScI+7bOkW0ssGS7bU9fssvS1gBlGkV+i0QhqnpPtZXmUWCw== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=mellanox.com; dmarc=pass action=none header.from=mellanox.com; dkim=pass header.d=mellanox.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=Mellanox.com; s=selector1; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=h0ackAdw5xxhjagVxEO0AjpYs9LAY8IjurhwKDSIrAE=; b=E4EFWJIQhS1G/+/EwvYpTi0rLu1LANigEHR2984vTHmZYbyujSCUYeFaFSjjU1zxKPi2s6tjhQgJQYAcanQgjuzTqh2HMeOZfe40GSfQP76JqlwRiWPU/uleQsvh5m3Va+G4DUbXf+UnONpKTHx1G1pFhV6SJbAZmnqb9nSN9jU= Authentication-Results: davemloft.net; dkim=none (message not signed) header.d=none;davemloft.net; dmarc=none action=none header.from=mellanox.com; Received: from AM6PR05MB5974.eurprd05.prod.outlook.com (2603:10a6:20b:a7::12) by AM6PR0502MB3734.eurprd05.prod.outlook.com (2603:10a6:209:b::27) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.3632.23; Wed, 9 Dec 2020 16:03:07 +0000 Received: from AM6PR05MB5974.eurprd05.prod.outlook.com ([fe80::642a:6259:153:ab88]) by AM6PR05MB5974.eurprd05.prod.outlook.com ([fe80::642a:6259:153:ab88%3]) with mapi id 15.20.3632.024; Wed, 9 Dec 2020 16:03:07 +0000 From: Maxim Mikityanskiy To: "David S. Miller" , Jamal Hadi Salim , Cong Wang , Jiri Pirko Cc: Saeed Mahameed , Jakub Kicinski , Tariq Toukan , Maxim Mikityanskiy , netdev@vger.kernel.org, Maxim Mikityanskiy , Tariq Toukan Subject: [PATCH net-next 4/4] net/mlx5e: Support HTB offload Date: Wed, 9 Dec 2020 18:02:51 +0200 Message-Id: <20201209160251.19054-5-maximmi@mellanox.com> X-Mailer: git-send-email 2.20.1 In-Reply-To: <20201209160251.19054-1-maximmi@mellanox.com> References: <20201209160251.19054-1-maximmi@mellanox.com> X-Originating-IP: [94.188.199.18] X-ClientProxiedBy: AM0PR04CA0098.eurprd04.prod.outlook.com (2603:10a6:208:be::39) To AM6PR05MB5974.eurprd05.prod.outlook.com (2603:10a6:20b:a7::12) MIME-Version: 1.0 X-MS-Exchange-MessageSentRepresentingType: 1 Received: from dev-l-vrt-208.mtl.labs.mlnx (94.188.199.18) by AM0PR04CA0098.eurprd04.prod.outlook.com (2603:10a6:208:be::39) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.3654.12 via Frontend Transport; Wed, 9 Dec 2020 16:03:06 +0000 X-MS-PublicTrafficType: Email X-MS-Office365-Filtering-HT: Tenant X-MS-Office365-Filtering-Correlation-Id: bec4e137-badc-4d99-3317-08d89c5beb59 X-MS-TrafficTypeDiagnostic: AM6PR0502MB3734: X-LD-Processed: a652971c-7d2e-4d9b-a6a4-d149256f461b,ExtFwd,ExtAddr X-MS-Exchange-Transport-Forked: True X-Microsoft-Antispam-PRVS: X-MS-Oob-TLC-OOBClassifiers: OLM:60; X-MS-Exchange-SenderADCheck: 1 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: gvssCaragsXvtKmUZEJTU8vHa1Y+LOT88gscUInGe/bhzz+KQHF1HoaeJr7iOi6pPKM4pGuWWP3jJcTK4x8Ltroi1wpPgf4UBbhCEfLGoh+SjDM+H6b9mlAj7EFtP27Zy2tNle4AG2LoILSRqbYgMoB3Za/uDR9tUMLjMJV6dEdWLyYgzAOmE+3La3iIwPoMGXXaDekT3mgSlbqwIcnl0NJGoAw2azgXnhjL1gRwdt1rXicAF7JJi8durreJWfYIv63HJnLadg/iA9/sZvpqX1T9Ld0FNe1G6tc9TAqXrHDa/JPV8VTgxuBKLNal9l9tDfgtrO5nnsPZjXLBfhtDNcjKdPipNTkyfjaIbOGPMSlX9V7ylfr9l2IRVad+7DPTUGP5kXyyGpMrCT3+VaO+EPq5iJASiXfj6UdIXY8rTrd9CdvnSskyUJlkK3hH65K7QfhbqXgQAqHwqtaOt9cgRQ== X-Forefront-Antispam-Report: CIP:255.255.255.255;CTRY:;LANG:en;SCL:1;SRV:;IPV:NLI;SFV:NSPM;H:AM6PR05MB5974.eurprd05.prod.outlook.com;PTR:;CAT:NONE;SFS:(4636009)(376002)(366004)(346002)(136003)(54906003)(5660300002)(66946007)(8936002)(6506007)(6486002)(8676002)(66476007)(2906002)(186003)(1076003)(6666004)(52116002)(6512007)(110136005)(30864003)(4326008)(34490700003)(956004)(7416002)(26005)(86362001)(36756003)(508600001)(2616005)(66556008)(16526019)(966005)(83380400001)(579004)(559001);DIR:OUT;SFP:1101; X-MS-Exchange-AntiSpam-MessageData: 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 X-OriginatorOrg: Mellanox.com X-MS-Exchange-CrossTenant-AuthSource: AM6PR05MB5974.eurprd05.prod.outlook.com X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 09 Dec 2020 16:03:07.4475 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: a652971c-7d2e-4d9b-a6a4-d149256f461b X-MS-Exchange-CrossTenant-Network-Message-Id: bec4e137-badc-4d99-3317-08d89c5beb59 X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: pixcnMx5tcRSltw6pI8qoB9fBQSeJeA5L46KPCCci+RVNASUkWu34LRksJlA3cFor4bnA0QvTEt3D681AHkRXg== X-MS-Exchange-Transport-CrossTenantHeadersStamped: AM6PR0502MB3734 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org This commit adds support for HTB offload in the mlx5e driver. Performance: NIC: Mellanox ConnectX-6 Dx CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz (24 cores with HT) 100 Gbit/s line rate, 500 UDP streams @ ~200 Mbit/s each 48 traffic classes, flower used for steering No shaping (rate limits set to 4 Gbit/s per TC) - checking for max throughput. Baseline: 98.7 Gbps, 8.25 Mpps HTB: 6.7 Gbps, 0.56 Mpps HTB offload: 95.6 Gbps, 8.00 Mpps Limitations: 1. 1024 leaf nodes, 3 levels of depth. 2. Granularity for ceil is 1 Mbit/s. Rates are converted to weights, and the bandwidth is split among the siblings according to these weights. Other parameters for classes are not supported. Ethtool statistics support for QoS SQs are also added. The counters are called qos_txN_*, where N is the QoS queue number (starting from 0, the numeration is separate from the normal SQs), and * is the counter name (the counters are the same as for the normal SQs). Signed-off-by: Maxim Mikityanskiy Reviewed-by: Tariq Toukan Reported-by: kernel test robot --- .../net/ethernet/mellanox/mlx5/core/Makefile | 6 +- drivers/net/ethernet/mellanox/mlx5/core/en.h | 27 +- .../ethernet/mellanox/mlx5/core/en/params.h | 2 + .../net/ethernet/mellanox/mlx5/core/en/ptp.c | 2 +- .../net/ethernet/mellanox/mlx5/core/en/qos.c | 935 ++++++++++++++++++ .../net/ethernet/mellanox/mlx5/core/en/qos.h | 39 + .../ethernet/mellanox/mlx5/core/en_ethtool.c | 21 + .../net/ethernet/mellanox/mlx5/core/en_main.c | 168 +++- .../ethernet/mellanox/mlx5/core/en_stats.c | 100 ++ .../ethernet/mellanox/mlx5/core/en_stats.h | 2 + .../net/ethernet/mellanox/mlx5/core/en_tx.c | 47 +- .../net/ethernet/mellanox/mlx5/core/en_txrx.c | 26 + drivers/net/ethernet/mellanox/mlx5/core/qos.c | 85 ++ drivers/net/ethernet/mellanox/mlx5/core/qos.h | 28 + include/linux/mlx5/mlx5_ifc.h | 13 +- 15 files changed, 1451 insertions(+), 50 deletions(-) create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en/qos.c create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en/qos.h create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/qos.c create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/qos.h diff --git a/drivers/net/ethernet/mellanox/mlx5/core/Makefile b/drivers/net/ethernet/mellanox/mlx5/core/Makefile index 77961643d5a9..69797c684d0b 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/Makefile +++ b/drivers/net/ethernet/mellanox/mlx5/core/Makefile @@ -16,7 +16,8 @@ mlx5_core-y := main.o cmd.o debugfs.o fw.o eq.o uar.o pagealloc.o \ transobj.o vport.o sriov.o fs_cmd.o fs_core.o pci_irq.o \ fs_counters.o rl.o lag.o dev.o events.o wq.o lib/gid.o \ lib/devcom.o lib/pci_vsc.o lib/dm.o diag/fs_tracepoint.o \ - diag/fw_tracer.o diag/crdump.o devlink.o diag/rsc_dump.o fw_reset.o + diag/fw_tracer.o diag/crdump.o devlink.o diag/rsc_dump.o \ + fw_reset.o qos.o # # Netdev basic @@ -25,7 +26,8 @@ mlx5_core-$(CONFIG_MLX5_CORE_EN) += en_main.o en_common.o en_fs.o en_ethtool.o \ en_tx.o en_rx.o en_dim.o en_txrx.o en/xdp.o en_stats.o \ en_selftest.o en/port.o en/monitor_stats.o en/health.o \ en/reporter_tx.o en/reporter_rx.o en/params.o en/xsk/pool.o \ - en/xsk/setup.o en/xsk/rx.o en/xsk/tx.o en/devlink.o en/ptp.o + en/xsk/setup.o en/xsk/rx.o en/xsk/tx.o en/devlink.o en/ptp.o \ + en/qos.o # # Netdev extra diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h index a1a81cfeb607..6a469d4d657b 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h @@ -55,6 +55,7 @@ #include "en_stats.h" #include "en/dcbnl.h" #include "en/fs.h" +#include "en/qos.h" #include "lib/hv_vhca.h" extern const struct net_device_ops mlx5e_netdev_ops; @@ -161,6 +162,9 @@ do { \ ##__VA_ARGS__); \ } while (0) +#define mlx5e_state_dereference(priv, p) \ + rcu_dereference_protected((p), lockdep_is_held(&(priv)->state_lock)) + enum mlx5e_rq_group { MLX5E_RQ_GROUP_REGULAR, MLX5E_RQ_GROUP_XSK, @@ -663,11 +667,13 @@ struct mlx5e_channel { struct mlx5e_xdpsq rq_xdpsq; struct mlx5e_txqsq sq[MLX5E_MAX_NUM_TC]; struct mlx5e_icosq icosq; /* internal control operations */ + struct mlx5e_txqsq __rcu * __rcu *qos_sqs; bool xdp; struct napi_struct napi; struct device *pdev; struct net_device *netdev; __be32 mkey_be; + u16 qos_sqs_size; u8 num_tc; u8 lag_port; @@ -756,6 +762,8 @@ struct mlx5e_modify_sq_param { int next_state; int rl_update; int rl_index; + bool qos_update; + u16 qos_queue_group_id; }; #if IS_ENABLED(CONFIG_PCI_HYPERV_INTERFACE) @@ -788,10 +796,20 @@ struct mlx5e_scratchpad { cpumask_var_t cpumask; }; +struct mlx5e_htb { + DECLARE_HASHTABLE(qos_tc2node, order_base_2(MLX5E_QOS_MAX_LEAF_NODES)); + DECLARE_BITMAP(qos_used_qids, MLX5E_QOS_MAX_LEAF_NODES); + struct mlx5e_sq_stats **qos_sq_stats; + u16 max_qos_sqs; + u16 maj_id; + u16 defcls; +}; + struct mlx5e_priv { /* priv data path fields - start */ /* +1 for port ptp ts */ - struct mlx5e_txqsq *txq2sq[(MLX5E_MAX_NUM_CHANNELS + 1) * MLX5E_MAX_NUM_TC]; + struct mlx5e_txqsq *txq2sq[(MLX5E_MAX_NUM_CHANNELS + 1) * MLX5E_MAX_NUM_TC + + MLX5E_QOS_MAX_LEAF_NODES]; int channel_tc2realtxq[MLX5E_MAX_NUM_CHANNELS][MLX5E_MAX_NUM_TC]; int port_ptp_tc2realtxq[MLX5E_MAX_NUM_TC]; #ifdef CONFIG_MLX5_CORE_EN_DCB @@ -859,6 +877,7 @@ struct mlx5e_priv { struct mlx5e_hv_vhca_stats_agent stats_agent; #endif struct mlx5e_scratchpad scratchpad; + struct mlx5e_htb htb; }; struct mlx5e_rx_handlers { @@ -986,6 +1005,7 @@ int mlx5e_safe_switch_channels(struct mlx5e_priv *priv, struct mlx5e_channels *new_chs, mlx5e_fp_preactivate preactivate, void *context); +int mlx5e_update_tx_netdev_queues(struct mlx5e_priv *priv); int mlx5e_num_channels_changed(struct mlx5e_priv *priv); int mlx5e_num_channels_changed_ctx(struct mlx5e_priv *priv, void *context); void mlx5e_activate_priv_channels(struct mlx5e_priv *priv); @@ -1010,6 +1030,9 @@ void mlx5e_deactivate_icosq(struct mlx5e_icosq *icosq); int mlx5e_modify_sq(struct mlx5_core_dev *mdev, u32 sqn, struct mlx5e_modify_sq_param *p); +int mlx5e_open_txqsq(struct mlx5e_channel *c, u32 tisn, int txq_ix, + struct mlx5e_params *params, struct mlx5e_sq_param *param, + struct mlx5e_txqsq *sq, int tc, u16 qos_queue_group_id, u16 qos_qid); void mlx5e_activate_txqsq(struct mlx5e_txqsq *sq); void mlx5e_deactivate_txqsq(struct mlx5e_txqsq *sq); void mlx5e_free_txqsq(struct mlx5e_txqsq *sq); @@ -1020,8 +1043,10 @@ struct mlx5e_create_sq_param; int mlx5e_create_sq_rdy(struct mlx5_core_dev *mdev, struct mlx5e_sq_param *param, struct mlx5e_create_sq_param *csp, + u16 qos_queue_group_id, u32 *sqn); void mlx5e_tx_err_cqe_work(struct work_struct *recover_work); +void mlx5e_close_txqsq(struct mlx5e_txqsq *sq); static inline bool mlx5_tx_swp_supported(struct mlx5_core_dev *mdev) { diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h index 807147d97a0f..ea2cfb04b31a 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h @@ -118,6 +118,8 @@ void mlx5e_build_rq_param(struct mlx5e_priv *priv, struct mlx5e_rq_param *param); void mlx5e_build_sq_param_common(struct mlx5e_priv *priv, struct mlx5e_sq_param *param); +void mlx5e_build_sq_param(struct mlx5e_priv *priv, struct mlx5e_params *params, + struct mlx5e_sq_param *param); void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv, struct mlx5e_params *params, struct mlx5e_xsk_param *xsk, diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c index 351118985a57..b5abef487e7e 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c @@ -261,7 +261,7 @@ static int mlx5e_ptp_open_txqsq(struct mlx5e_port_ptp *c, u32 tisn, csp.min_inline_mode = txqsq->min_inline_mode; csp.ts_cqe_to_dest_cqn = ptpsq->ts_cq.mcq.cqn; - err = mlx5e_create_sq_rdy(c->mdev, sqp, &csp, &txqsq->sqn); + err = mlx5e_create_sq_rdy(c->mdev, sqp, &csp, 0, &txqsq->sqn); if (err) goto err_free_txqsq; diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c b/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c new file mode 100644 index 000000000000..c8c69679e603 --- /dev/null +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c @@ -0,0 +1,935 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* Copyright (c) 2020, Mellanox Technologies inc. All rights reserved. */ + +#include "en.h" +#include "params.h" +#include "../qos.h" + +#define BYTES_IN_MBIT 125000 + +int mlx5e_qos_max_leaf_nodes(struct mlx5_core_dev *mdev) +{ + return min(MLX5E_QOS_MAX_LEAF_NODES, mlx5_qos_max_leaf_nodes(mdev)); +} + +int mlx5e_qos_cur_leaf_nodes(struct mlx5e_priv *priv) +{ + int last = find_last_bit(priv->htb.qos_used_qids, mlx5e_qos_max_leaf_nodes(priv->mdev)); + + return last == mlx5e_qos_max_leaf_nodes(priv->mdev) ? 0 : last + 1; +} + +/* Software representation of the QoS tree (internal to this file) */ + +static int mlx5e_find_unused_qos_qid(struct mlx5e_priv *priv) +{ + int size = mlx5e_qos_max_leaf_nodes(priv->mdev); + int res; + + WARN_ONCE(!mutex_is_locked(&priv->state_lock), "%s: state_lock is not held\n", __func__); + res = find_first_zero_bit(priv->htb.qos_used_qids, size); + + return res == size ? -ENOSPC : res; +} + +struct mlx5e_qos_node { + struct hlist_node hnode; + struct rcu_head rcu; + struct mlx5e_qos_node *parent; + u64 rate; + u32 bw_share; + u32 max_average_bw; + u32 hw_id; + u32 classid; // 16-bit, except root. + u16 qid; +}; + +#define MLX5E_QOS_QID_INNER 0xffff +#define MLX5E_HTB_CLASSID_ROOT 0xffffffff + +static struct mlx5e_qos_node * +mlx5e_sw_node_create_leaf(struct mlx5e_priv *priv, u16 classid, u16 qid, + struct mlx5e_qos_node *parent) +{ + struct mlx5e_qos_node *node; + + node = kzalloc(sizeof(*node), GFP_KERNEL); + if (!node) + return ERR_PTR(-ENOMEM); + + node->parent = parent; + + node->qid = qid; + __set_bit(qid, priv->htb.qos_used_qids); + + node->classid = classid; + hash_add_rcu(priv->htb.qos_tc2node, &node->hnode, classid); + + mlx5e_update_tx_netdev_queues(priv); + + return node; +} + +static struct mlx5e_qos_node *mlx5e_sw_node_create_root(struct mlx5e_priv *priv) +{ + struct mlx5e_qos_node *node; + + node = kzalloc(sizeof(*node), GFP_KERNEL); + if (!node) + return ERR_PTR(-ENOMEM); + + node->qid = MLX5E_QOS_QID_INNER; + node->classid = MLX5E_HTB_CLASSID_ROOT; + hash_add_rcu(priv->htb.qos_tc2node, &node->hnode, node->classid); + + return node; +} + +static struct mlx5e_qos_node *mlx5e_sw_node_find(struct mlx5e_priv *priv, u32 classid) +{ + struct mlx5e_qos_node *node = NULL; + + hash_for_each_possible(priv->htb.qos_tc2node, node, hnode, classid) { + if (node->classid == classid) + break; + } + + return node; +} + +static struct mlx5e_qos_node *mlx5e_sw_node_find_rcu(struct mlx5e_priv *priv, u32 classid) +{ + struct mlx5e_qos_node *node = NULL; + + hash_for_each_possible_rcu(priv->htb.qos_tc2node, node, hnode, classid) { + if (node->classid == classid) + break; + } + + return node; +} + +static void mlx5e_sw_node_delete(struct mlx5e_priv *priv, struct mlx5e_qos_node *node) +{ + hash_del_rcu(&node->hnode); + if (node->qid != MLX5E_QOS_QID_INNER) { + __clear_bit(node->qid, priv->htb.qos_used_qids); + mlx5e_update_tx_netdev_queues(priv); + } + kfree_rcu(node, rcu); +} + +/* TX datapath API */ + +static u16 mlx5e_qid_from_qos(struct mlx5e_channels *chs, u16 qid) +{ + /* These channel params are safe to access from the datapath, because: + * 1. This function is called only after checking priv->htb.maj_id != 0, + * and the number of queues can't change while HTB offload is active. + * 2. When priv->htb.maj_id becomes 0, synchronize_rcu waits for + * mlx5e_select_queue to finish while holding priv->state_lock, + * preventing other code from changing the number of queues. + */ + bool is_ptp = MLX5E_GET_PFLAG(&chs->params, MLX5E_PFLAG_TX_PORT_TS); + + return (chs->params.num_channels + is_ptp) * chs->params.num_tc + qid; +} + +int mlx5e_get_txq_by_classid(struct mlx5e_priv *priv, u16 classid) +{ + struct mlx5e_qos_node *node; + u16 qid; + int res; + + rcu_read_lock(); + + node = mlx5e_sw_node_find_rcu(priv, classid); + if (!node) { + res = -ENOENT; + goto out; + } + qid = READ_ONCE(node->qid); + if (qid == MLX5E_QOS_QID_INNER) { + res = -EINVAL; + goto out; + } + res = mlx5e_qid_from_qos(&priv->channels, qid); + +out: + rcu_read_unlock(); + return res; +} + +static struct mlx5e_txqsq *mlx5e_get_qos_sq(struct mlx5e_priv *priv, int qid) +{ + struct mlx5e_params *params = &priv->channels.params; + struct mlx5e_txqsq **qos_sqs; + struct mlx5e_channel *c; + int ix; + + ix = qid % params->num_channels; + qid /= params->num_channels; + c = priv->channels.c[ix]; + + qos_sqs = mlx5e_state_dereference(priv, c->qos_sqs); + return mlx5e_state_dereference(priv, qos_sqs[qid]); +} + +/* SQ lifecycle */ + +static int mlx5e_open_qos_sq(struct mlx5e_priv *priv, struct mlx5e_channels *chs, + struct mlx5e_qos_node *node) +{ + struct mlx5e_create_cq_param ccp = {}; + struct mlx5e_sq_param param_sq; + struct mlx5e_cq_param param_cq; + struct mlx5e_txqsq **qos_sqs; + int txq_ix, ix, qid, err = 0; + struct mlx5e_params *params; + struct mlx5e_channel *c; + struct mlx5e_txqsq *sq; + + params = &chs->params; + + txq_ix = mlx5e_qid_from_qos(chs, node->qid); + + WARN_ON(node->qid > priv->htb.max_qos_sqs); + if (node->qid == priv->htb.max_qos_sqs) { + struct mlx5e_sq_stats *stats, **stats_list = NULL; + + if (priv->htb.max_qos_sqs == 0) { + stats_list = kvcalloc(mlx5e_qos_max_leaf_nodes(priv->mdev), + sizeof(*stats_list), + GFP_KERNEL); + if (!stats_list) + return -ENOMEM; + } + stats = kzalloc(sizeof(*stats), GFP_KERNEL); + if (!stats) { + kvfree(stats_list); + return -ENOMEM; + } + if (stats_list) + WRITE_ONCE(priv->htb.qos_sq_stats, stats_list); + WRITE_ONCE(priv->htb.qos_sq_stats[node->qid], stats); + /* Order max_qos_sqs increment after writing the array pointer. + * Pairs with smp_load_acquire in en_stats.c. + */ + smp_store_release(&priv->htb.max_qos_sqs, priv->htb.max_qos_sqs + 1); + } + + ix = node->qid % params->num_channels; + qid = node->qid / params->num_channels; + c = chs->c[ix]; + + qos_sqs = mlx5e_state_dereference(priv, c->qos_sqs); + sq = kzalloc(sizeof(*sq), GFP_KERNEL); + + if (!sq) + return -ENOMEM; + + mlx5e_build_create_cq_param(&ccp, c); + + memset(¶m_sq, 0, sizeof(param_sq)); + memset(¶m_cq, 0, sizeof(param_cq)); + mlx5e_build_sq_param(priv, params, ¶m_sq); + mlx5e_build_tx_cq_param(priv, params, ¶m_cq); + err = mlx5e_open_cq(priv, params->tx_cq_moderation, ¶m_cq, &ccp, &sq->cq); + if (err) + goto err_free_sq; + err = mlx5e_open_txqsq(c, priv->tisn[c->lag_port][0], txq_ix, params, + ¶m_sq, sq, 0, node->hw_id, node->qid); + if (err) + goto err_close_cq; + + rcu_assign_pointer(qos_sqs[qid], sq); + + return 0; + +err_close_cq: + mlx5e_close_cq(&sq->cq); +err_free_sq: + kfree(sq); + return err; +} + +static void mlx5e_activate_qos_sq(struct mlx5e_priv *priv, struct mlx5e_qos_node *node) +{ + struct mlx5e_txqsq *sq; + + sq = mlx5e_get_qos_sq(priv, node->qid); + + WRITE_ONCE(priv->txq2sq[mlx5e_qid_from_qos(&priv->channels, node->qid)], sq); + + /* Make the change to txq2sq visible before the queue is started. + * As mlx5e_xmit runs under a spinlock, there is an implicit ACQUIRE, + * which pairs with this barrier. + */ + smp_wmb(); + + qos_dbg(priv->mdev, "Activate QoS SQ qid %hu\n", node->qid); + mlx5e_activate_txqsq(sq); +} + +static void mlx5e_deactivate_qos_sq(struct mlx5e_priv *priv, u16 qid) +{ + struct mlx5e_txqsq *sq; + + sq = mlx5e_get_qos_sq(priv, qid); + if (!sq) /* Handle the case when the SQ failed to open. */ + return; + + qos_dbg(priv->mdev, "Deactivate QoS SQ qid %hu\n", qid); + mlx5e_deactivate_txqsq(sq); + + /* The queue is disabled, no synchronization with datapath is needed. */ + priv->txq2sq[mlx5e_qid_from_qos(&priv->channels, qid)] = NULL; +} + +static void mlx5e_close_qos_sq(struct mlx5e_priv *priv, u16 qid) +{ + struct mlx5e_txqsq **qos_sqs, *sq; + struct mlx5e_params *params; + struct mlx5e_channel *c; + int ix; + + params = &priv->channels.params; + + ix = qid % params->num_channels; + qid /= params->num_channels; + c = priv->channels.c[ix]; + qos_sqs = mlx5e_state_dereference(priv, c->qos_sqs); + sq = rcu_replace_pointer(qos_sqs[qid], NULL, lockdep_is_held(&priv->state_lock)); + if (!sq) /* Handle the case when the SQ failed to open. */ + return; + + synchronize_rcu(); /* Sync with NAPI. */ + + mlx5e_close_txqsq(sq); + mlx5e_close_cq(&sq->cq); + kfree(sq); +} + +void mlx5e_qos_close_queues(struct mlx5e_channel *c) +{ + struct mlx5e_txqsq **qos_sqs; + int i; + + qos_sqs = rcu_replace_pointer(c->qos_sqs, NULL, lockdep_is_held(&c->priv->state_lock)); + if (!qos_sqs) + return; + synchronize_rcu(); /* Sync with NAPI. */ + + for (i = 0; i < c->qos_sqs_size; i++) { + struct mlx5e_txqsq *sq; + + sq = mlx5e_state_dereference(c->priv, qos_sqs[i]); + if (!sq) /* Handle the case when the SQ failed to open. */ + continue; + + mlx5e_close_txqsq(sq); + mlx5e_close_cq(&sq->cq); + kfree(sq); + } + + kvfree(qos_sqs); +} + +static void mlx5e_qos_close_all_queues(struct mlx5e_channels *chs) +{ + int i; + + for (i = 0; i < chs->num; i++) + mlx5e_qos_close_queues(chs->c[i]); +} + +static int mlx5e_qos_alloc_queues(struct mlx5e_priv *priv, struct mlx5e_channels *chs) +{ + u16 qos_sqs_size; + int i; + + qos_sqs_size = DIV_ROUND_UP(mlx5e_qos_max_leaf_nodes(priv->mdev), chs->num); + + for (i = 0; i < chs->num; i++) { + struct mlx5e_txqsq **sqs; + + sqs = kvcalloc(qos_sqs_size, sizeof(struct mlx5e_txqsq *), GFP_KERNEL); + if (!sqs) + goto err_free; + + WRITE_ONCE(chs->c[i]->qos_sqs_size, qos_sqs_size); + smp_wmb(); /* Pairs with mlx5e_napi_poll. */ + rcu_assign_pointer(chs->c[i]->qos_sqs, sqs); + } + + return 0; + +err_free: + while (--i >= 0) { + struct mlx5e_txqsq **sqs; + + sqs = rcu_replace_pointer(chs->c[i]->qos_sqs, NULL, + lockdep_is_held(&priv->state_lock)); + + synchronize_rcu(); /* Sync with NAPI. */ + kvfree(sqs); + } + return -ENOMEM; +} + +int mlx5e_qos_open_queues(struct mlx5e_priv *priv, struct mlx5e_channels *chs) +{ + struct mlx5e_qos_node *node = NULL; + int bkt, err; + + if (!priv->htb.maj_id) + return 0; + + err = mlx5e_qos_alloc_queues(priv, chs); + if (err) + return err; + + hash_for_each(priv->htb.qos_tc2node, bkt, node, hnode) { + if (node->qid == MLX5E_QOS_QID_INNER) + continue; + err = mlx5e_open_qos_sq(priv, chs, node); + if (err) { + mlx5e_qos_close_all_queues(chs); + return err; + } + } + + return 0; +} + +void mlx5e_qos_activate_queues(struct mlx5e_priv *priv) +{ + struct mlx5e_qos_node *node = NULL; + int bkt; + + hash_for_each(priv->htb.qos_tc2node, bkt, node, hnode) { + if (node->qid == MLX5E_QOS_QID_INNER) + continue; + mlx5e_activate_qos_sq(priv, node); + } +} + +void mlx5e_qos_deactivate_queues(struct mlx5e_channel *c) +{ + struct mlx5e_params *params = &c->priv->channels.params; + struct mlx5e_txqsq **qos_sqs; + int i; + + qos_sqs = mlx5e_state_dereference(c->priv, c->qos_sqs); + if (!qos_sqs) + return; + + for (i = 0; i < c->qos_sqs_size; i++) { + u16 qid = params->num_channels * i + c->ix; + struct mlx5e_txqsq *sq; + + sq = mlx5e_state_dereference(c->priv, qos_sqs[i]); + if (!sq) /* Handle the case when the SQ failed to open. */ + continue; + + qos_dbg(c->mdev, "Deactivate QoS SQ qid %hu\n", qid); + mlx5e_deactivate_txqsq(sq); + + /* The queue is disabled, no synchronization with datapath is needed. */ + c->priv->txq2sq[mlx5e_qid_from_qos(&c->priv->channels, qid)] = NULL; + } +} + +static void mlx5e_qos_deactivate_all_queues(struct mlx5e_channels *chs) +{ + int i; + + for (i = 0; i < chs->num; i++) + mlx5e_qos_deactivate_queues(chs->c[i]); +} + +/* HTB API */ + +int mlx5e_htb_root_add(struct mlx5e_priv *priv, u16 htb_maj_id, u16 htb_defcls) +{ + struct mlx5e_qos_node *root; + bool opened; + int err; + + qos_dbg(priv->mdev, "TC_HTB_CREATE handle %04x:, default :%04x\n", htb_maj_id, htb_defcls); + + if (!mlx5_qos_is_supported(priv->mdev)) + return -EOPNOTSUPP; + + opened = test_bit(MLX5E_STATE_OPENED, &priv->state); + if (opened) { + err = mlx5e_qos_alloc_queues(priv, &priv->channels); + if (err) + return err; + } + + root = mlx5e_sw_node_create_root(priv); + if (IS_ERR(root)) { + err = PTR_ERR(root); + goto err_free_queues; + } + + err = mlx5_qos_create_root_node(priv->mdev, &root->hw_id); + if (err) + goto err_sw_node_delete; + + WRITE_ONCE(priv->htb.defcls, htb_defcls); + /* Order maj_id after defcls - pairs with + * mlx5e_select_queue/mlx5e_select_htb_queues. + */ + smp_store_release(&priv->htb.maj_id, htb_maj_id); + + return 0; + +err_sw_node_delete: + mlx5e_sw_node_delete(priv, root); + +err_free_queues: + if (opened) + mlx5e_qos_close_all_queues(&priv->channels); + return err; +} + +int mlx5e_htb_root_del(struct mlx5e_priv *priv) +{ + struct mlx5e_qos_node *root; + int err; + + qos_dbg(priv->mdev, "TC_HTB_DESTROY\n"); + + WRITE_ONCE(priv->htb.maj_id, 0); + synchronize_rcu(); /* Sync with mlx5e_select_htb_queue and TX data path. */ + + root = mlx5e_sw_node_find(priv, MLX5E_HTB_CLASSID_ROOT); + if (!root) { + qos_warn(priv->mdev, "Failed to find the root node in the QoS tree\n"); + return -ENOENT; + } + err = mlx5_qos_destroy_node(priv->mdev, root->hw_id); + if (err) + qos_warn(priv->mdev, "Failed to destroy root node %u, err = %d\n", + root->hw_id, err); + mlx5e_sw_node_delete(priv, root); + + mlx5e_qos_deactivate_all_queues(&priv->channels); + mlx5e_qos_close_all_queues(&priv->channels); + + return err; +} + +static int mlx5e_htb_convert_rate(struct mlx5e_priv *priv, u64 rate, + struct mlx5e_qos_node *parent, u32 *bw_share) +{ + u64 share = 0; + + while (parent->classid != MLX5E_HTB_CLASSID_ROOT && !parent->max_average_bw) + parent = parent->parent; + + if (parent->max_average_bw) + share = div64_u64(div_u64(rate * 100, BYTES_IN_MBIT), + parent->max_average_bw); + else + share = 101; + + *bw_share = share == 0 ? 1 : share > 100 ? 0 : share; + + qos_dbg(priv->mdev, "Convert: rate %llu, parent ceil %llu -> bw_share %u\n", + rate, (u64)parent->max_average_bw * BYTES_IN_MBIT, *bw_share); + + return 0; +} + +static void mlx5e_htb_convert_ceil(struct mlx5e_priv *priv, u64 ceil, u32 *max_average_bw) +{ + *max_average_bw = div_u64(ceil, BYTES_IN_MBIT); + + qos_dbg(priv->mdev, "Convert: ceil %llu -> max_average_bw %u\n", + ceil, *max_average_bw); +} + +int mlx5e_htb_leaf_alloc_queue(struct mlx5e_priv *priv, u16 classid, + u32 parent_classid, u64 rate, u64 ceil) +{ + struct mlx5e_qos_node *node, *parent; + int qid; + int err; + + qos_dbg(priv->mdev, "TC_HTB_LEAF_ALLOC_QUEUE classid %04x, parent %04x, rate %llu, ceil %llu\n", + classid, parent_classid, rate, ceil); + + qid = mlx5e_find_unused_qos_qid(priv); + if (qid < 0) + return qid; + + parent = mlx5e_sw_node_find(priv, parent_classid); + if (!parent) + return -EINVAL; + + node = mlx5e_sw_node_create_leaf(priv, classid, qid, parent); + if (IS_ERR(node)) + return PTR_ERR(node); + + node->rate = rate; + mlx5e_htb_convert_rate(priv, rate, node->parent, &node->bw_share); + mlx5e_htb_convert_ceil(priv, ceil, &node->max_average_bw); + + err = mlx5_qos_create_leaf_node(priv->mdev, node->parent->hw_id, + node->bw_share, node->max_average_bw, + &node->hw_id); + if (err) { + qos_warn(priv->mdev, "Failed to create a leaf node (class %04x), err = %d\n", + classid, err); + mlx5e_sw_node_delete(priv, node); + return err; + } + + if (test_bit(MLX5E_STATE_OPENED, &priv->state)) { + err = mlx5e_open_qos_sq(priv, &priv->channels, node); + if (err) + qos_warn(priv->mdev, "Failed to create a QoS SQ (class %04x), err = %d\n", + classid, err); + else + mlx5e_activate_qos_sq(priv, node); + } + + return mlx5e_qid_from_qos(&priv->channels, node->qid); +} + +int mlx5e_htb_leaf_to_inner(struct mlx5e_priv *priv, u16 classid, u16 child_classid, + u64 rate, u64 ceil) +{ + struct mlx5e_qos_node *node, *child; + int err, tmp_err; + u32 new_hw_id; + u16 qid; + + qos_dbg(priv->mdev, "TC_HTB_LEAF_TO_INNER classid %04x, upcoming child %04x, rate %llu, ceil %llu\n", + classid, child_classid, rate, ceil); + + node = mlx5e_sw_node_find(priv, classid); + if (!node) + return -ENOENT; + + err = mlx5_qos_create_inner_node(priv->mdev, node->parent->hw_id, + node->bw_share, node->max_average_bw, + &new_hw_id); + if (err) { + qos_warn(priv->mdev, "Failed to create an inner node (class %04x), err = %d\n", + classid, err); + return err; + } + + /* Intentionally reuse the qid for the upcoming first child. */ + child = mlx5e_sw_node_create_leaf(priv, child_classid, node->qid, node); + if (IS_ERR(child)) { + err = PTR_ERR(child); + goto err_destroy_hw_node; + } + + child->rate = rate; + mlx5e_htb_convert_rate(priv, rate, node, &child->bw_share); + mlx5e_htb_convert_ceil(priv, ceil, &child->max_average_bw); + + err = mlx5_qos_create_leaf_node(priv->mdev, new_hw_id, child->bw_share, + child->max_average_bw, &child->hw_id); + if (err) { + qos_warn(priv->mdev, "Failed to create a leaf node (class %04x), err = %d\n", + classid, err); + goto err_delete_sw_node; + } + + /* No fail point. */ + + qid = xchg(&node->qid, MLX5E_QOS_QID_INNER); + + if (test_bit(MLX5E_STATE_OPENED, &priv->state)) { + mlx5e_deactivate_qos_sq(priv, qid); + mlx5e_close_qos_sq(priv, qid); + } + + err = mlx5_qos_destroy_node(priv->mdev, node->hw_id); + if (err) /* Not fatal. */ + qos_warn(priv->mdev, "Failed to destroy leaf node %u (class %04x), err = %d\n", + node->hw_id, classid, err); + + node->hw_id = new_hw_id; + + if (test_bit(MLX5E_STATE_OPENED, &priv->state)) { + err = mlx5e_open_qos_sq(priv, &priv->channels, child); + if (err) + qos_warn(priv->mdev, "Failed to create a QoS SQ (class %04x), err = %d\n", + classid, err); + else + mlx5e_activate_qos_sq(priv, child); + } + + return 0; + +err_delete_sw_node: + child->qid = MLX5E_QOS_QID_INNER; + mlx5e_sw_node_delete(priv, child); + +err_destroy_hw_node: + tmp_err = mlx5_qos_destroy_node(priv->mdev, new_hw_id); + if (tmp_err) /* Not fatal. */ + qos_warn(priv->mdev, "Failed to roll back creation of an inner node %u (class %04x), err = %d\n", + new_hw_id, classid, tmp_err); + return err; +} + +static struct mlx5e_qos_node *mlx5e_sw_node_find_by_qid(struct mlx5e_priv *priv, u16 qid) +{ + struct mlx5e_qos_node *node = NULL; + int bkt; + + hash_for_each(priv->htb.qos_tc2node, bkt, node, hnode) + if (node->qid == qid) + break; + + return node; +} + +static void mlx5e_reactivate_qos_sq(struct mlx5e_priv *priv, u16 qid, struct netdev_queue *txq) +{ + qos_dbg(priv->mdev, "Reactivate QoS SQ qid %hu\n", qid); + netdev_tx_reset_queue(txq); + netif_tx_start_queue(txq); +} + +static void mlx5e_reset_qdisc(struct net_device *dev, u16 qid) +{ + struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, qid); + struct Qdisc *qdisc = dev_queue->qdisc_sleeping; + + if (!qdisc) + return; + + spin_lock_bh(qdisc_lock(qdisc)); + qdisc_reset(qdisc); + spin_unlock_bh(qdisc_lock(qdisc)); +} + +int mlx5e_htb_leaf_del(struct mlx5e_priv *priv, u16 classid, u16 *old_qid, u16 *new_qid) +{ + struct mlx5e_qos_node *node; + struct netdev_queue *txq; + u16 qid, moved_qid; + bool opened; + int err; + + qos_dbg(priv->mdev, "TC_HTB_LEAF_DEL classid %04x\n", classid); + + *old_qid = *new_qid = 0; + + node = mlx5e_sw_node_find(priv, classid); + if (!node) + return -ENOENT; + + /* Store qid for reuse. */ + qid = node->qid; + + opened = test_bit(MLX5E_STATE_OPENED, &priv->state); + if (opened) { + txq = netdev_get_tx_queue(priv->netdev, + mlx5e_qid_from_qos(&priv->channels, qid)); + mlx5e_deactivate_qos_sq(priv, qid); + mlx5e_close_qos_sq(priv, qid); + } + + err = mlx5_qos_destroy_node(priv->mdev, node->hw_id); + if (err) /* Not fatal. */ + qos_warn(priv->mdev, "Failed to destroy leaf node %u (class %04x), err = %d\n", + node->hw_id, classid, err); + + mlx5e_sw_node_delete(priv, node); + + moved_qid = mlx5e_qos_cur_leaf_nodes(priv); + + if (moved_qid == 0) { + /* The last QoS SQ was just destroyed. */ + if (opened) + mlx5e_reactivate_qos_sq(priv, qid, txq); + return 0; + } + moved_qid--; + + if (moved_qid < qid) { + /* The highest QoS SQ was just destroyed. */ + WARN(moved_qid != qid - 1, "Gaps in queue numeration: destroyed queue %hu, the highest queue is %hu", + qid, moved_qid); + if (opened) + mlx5e_reactivate_qos_sq(priv, qid, txq); + return 0; + } + + WARN(moved_qid == qid, "Can't move node with qid %hu to itself", qid); + qos_dbg(priv->mdev, "Moving QoS SQ %hu to %hu\n", moved_qid, qid); + + node = mlx5e_sw_node_find_by_qid(priv, moved_qid); + WARN(!node, "Could not find a node with qid %hu to move to queue %hu", + moved_qid, qid); + + /* Stop traffic to the old queue. */ + WRITE_ONCE(node->qid, MLX5E_QOS_QID_INNER); + __clear_bit(moved_qid, priv->htb.qos_used_qids); + + if (opened) { + txq = netdev_get_tx_queue(priv->netdev, + mlx5e_qid_from_qos(&priv->channels, moved_qid)); + mlx5e_deactivate_qos_sq(priv, moved_qid); + mlx5e_close_qos_sq(priv, moved_qid); + } + + /* Prevent packets from the old class from getting into the new one. */ + mlx5e_reset_qdisc(priv->netdev, moved_qid); + + __set_bit(qid, priv->htb.qos_used_qids); + WRITE_ONCE(node->qid, qid); + + if (test_bit(MLX5E_STATE_OPENED, &priv->state)) { + err = mlx5e_open_qos_sq(priv, &priv->channels, node); + if (err) + qos_warn(priv->mdev, "Failed to create a QoS SQ (class %04x) while moving qid %hu to %hu, err = %d\n", + node->classid, moved_qid, qid, err); + else + mlx5e_activate_qos_sq(priv, node); + } + + mlx5e_update_tx_netdev_queues(priv); + if (opened) + mlx5e_reactivate_qos_sq(priv, moved_qid, txq); + + *old_qid = mlx5e_qid_from_qos(&priv->channels, moved_qid); + *new_qid = mlx5e_qid_from_qos(&priv->channels, qid); + return 0; +} + +int mlx5e_htb_leaf_del_last(struct mlx5e_priv *priv, u16 classid) +{ + struct mlx5e_qos_node *node, *parent; + u32 old_hw_id, new_hw_id; + u16 qid; + int err; + + qos_dbg(priv->mdev, "TC_HTB_LEAF_DEL_LAST classid %04x\n", classid); + + node = mlx5e_sw_node_find(priv, classid); + if (!node) + return -ENOENT; + + err = mlx5_qos_create_leaf_node(priv->mdev, node->parent->parent->hw_id, + node->parent->bw_share, + node->parent->max_average_bw, + &new_hw_id); + if (err) { + qos_warn(priv->mdev, "Failed to create a leaf node (class %04x), err = %d\n", + classid, err); + return err; + } + + /* Store qid for reuse and prevent clearing the bit. */ + qid = xchg(&node->qid, MLX5E_QOS_QID_INNER); + + if (test_bit(MLX5E_STATE_OPENED, &priv->state)) { + mlx5e_deactivate_qos_sq(priv, qid); + mlx5e_close_qos_sq(priv, qid); + } + + /* Prevent packets from the old class from getting into the new one. */ + mlx5e_reset_qdisc(priv->netdev, qid); + + err = mlx5_qos_destroy_node(priv->mdev, node->hw_id); + if (err) /* Not fatal. */ + qos_warn(priv->mdev, "Failed to destroy leaf node %u (class %04x), err = %d\n", + node->hw_id, classid, err); + + parent = node->parent; + mlx5e_sw_node_delete(priv, node); + + node = parent; + WRITE_ONCE(node->qid, qid); + old_hw_id = node->hw_id; + node->hw_id = new_hw_id; + + if (test_bit(MLX5E_STATE_OPENED, &priv->state)) { + err = mlx5e_open_qos_sq(priv, &priv->channels, node); + if (err) + qos_warn(priv->mdev, "Failed to create a QoS SQ (class %04x), err = %d\n", + classid, err); + else + mlx5e_activate_qos_sq(priv, node); + } + + err = mlx5_qos_destroy_node(priv->mdev, old_hw_id); + if (err) /* Not fatal. */ + qos_warn(priv->mdev, "Failed to destroy leaf node %u (class %04x), err = %d\n", + node->hw_id, classid, err); + + return 0; +} + +static int mlx5e_qos_update_children(struct mlx5e_priv *priv, struct mlx5e_qos_node *node) +{ + struct mlx5e_qos_node *child; + int err = 0; + int bkt; + + hash_for_each(priv->htb.qos_tc2node, bkt, child, hnode) { + u32 old_bw_share = child->bw_share; + int err_one; + + if (child->parent != node) + continue; + + mlx5e_htb_convert_rate(priv, child->rate, node, &child->bw_share); + if (child->bw_share == old_bw_share) + continue; + + err_one = mlx5_qos_update_node(priv->mdev, child->hw_id, child->bw_share, + child->max_average_bw, child->hw_id); + if (!err) + err = err_one; + } + + return err; +} + +int mlx5e_htb_node_modify(struct mlx5e_priv *priv, u16 classid, u64 rate, u64 ceil) +{ + u32 bw_share, max_average_bw; + struct mlx5e_qos_node *node; + bool ceil_changed = false; + int err; + + qos_dbg(priv->mdev, "TC_HTB_LEAF_MODIFY classid %04x, rate %llu, ceil %llu\n", + classid, rate, ceil); + + node = mlx5e_sw_node_find(priv, classid); + if (!node) + return -ENOENT; + + node->rate = rate; + mlx5e_htb_convert_rate(priv, rate, node->parent, &bw_share); + mlx5e_htb_convert_ceil(priv, ceil, &max_average_bw); + + err = mlx5_qos_update_node(priv->mdev, node->parent->hw_id, bw_share, + max_average_bw, node->hw_id); + if (err) + return err; + + if (max_average_bw != node->max_average_bw) + ceil_changed = true; + + node->bw_share = bw_share; + node->max_average_bw = max_average_bw; + + if (ceil_changed) + err = mlx5e_qos_update_children(priv, node); + + return err; +} diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/qos.h b/drivers/net/ethernet/mellanox/mlx5/core/en/qos.h new file mode 100644 index 000000000000..90d78780d74a --- /dev/null +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/qos.h @@ -0,0 +1,39 @@ +/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */ +/* Copyright (c) 2020, Mellanox Technologies inc. All rights reserved. */ + +#ifndef __MLX5E_EN_QOS_H +#define __MLX5E_EN_QOS_H + +#include + +#define MLX5E_QOS_MAX_LEAF_NODES 1024 + +struct mlx5e_priv; +struct mlx5e_channels; +struct mlx5e_channel; + +int mlx5e_qos_max_leaf_nodes(struct mlx5_core_dev *mdev); +int mlx5e_qos_cur_leaf_nodes(struct mlx5e_priv *priv); + +/* TX datapath API */ +int mlx5e_get_txq_by_classid(struct mlx5e_priv *priv, u16 classid); +struct mlx5e_txqsq *mlx5e_get_sq(struct mlx5e_priv *priv, int qid); + +/* SQ lifecycle */ +int mlx5e_qos_open_queues(struct mlx5e_priv *priv, struct mlx5e_channels *chs); +void mlx5e_qos_activate_queues(struct mlx5e_priv *priv); +void mlx5e_qos_deactivate_queues(struct mlx5e_channel *c); +void mlx5e_qos_close_queues(struct mlx5e_channel *c); + +/* HTB API */ +int mlx5e_htb_root_add(struct mlx5e_priv *priv, u16 htb_maj_id, u16 htb_defcls); +int mlx5e_htb_root_del(struct mlx5e_priv *priv); +int mlx5e_htb_leaf_alloc_queue(struct mlx5e_priv *priv, u16 classid, + u32 parent_classid, u64 rate, u64 ceil); +int mlx5e_htb_leaf_to_inner(struct mlx5e_priv *priv, u16 classid, + u16 child_classid, u64 rate, u64 ceil); +int mlx5e_htb_leaf_del(struct mlx5e_priv *priv, u16 classid, u16 *old_qid, u16 *new_qid); +int mlx5e_htb_leaf_del_last(struct mlx5e_priv *priv, u16 classid); +int mlx5e_htb_node_modify(struct mlx5e_priv *priv, u16 classid, u64 rate, u64 ceil); + +#endif diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c index d9076d543104..2c1fdcca1d02 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c @@ -447,6 +447,17 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv, goto out; } + /* Don't allow changing the number of channels if HTB offload is active, + * because the numeration of the QoS SQs will change, while per-queue + * qdiscs are attached. + */ + if (priv->htb.maj_id) { + err = -EINVAL; + netdev_err(priv->netdev, "%s: HTB offload is active, cannot change the number of channels\n", + __func__); + goto out; + } + new_channels.params = priv->channels.params; new_channels.params.num_channels = count; @@ -1954,6 +1965,16 @@ static int set_pflag_tx_port_ts(struct net_device *netdev, bool enable) if (!MLX5_CAP_GEN(mdev, ts_cqe_to_dest_cqn)) return -EOPNOTSUPP; + /* Don't allow changing the PTP state if HTB offload is active, because + * the numeration of the QoS SQs will change, while per-queue qdiscs are + * attached. + */ + if (priv->htb.maj_id) { + netdev_err(priv->netdev, "%s: HTB offload is active, cannot change the PTP state\n", + __func__); + return -EINVAL; + } + new_channels.params = priv->channels.params; MLX5E_SET_PFLAG(&new_channels.params, MLX5E_PFLAG_TX_PORT_TS, enable); /* No need to verify SQ stop room as diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c index 03831650f655..d70421c13a95 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c @@ -65,6 +65,7 @@ #include "en/devlink.h" #include "lib/mlx5.h" #include "en/ptp.h" +#include "qos.h" bool mlx5e_check_fragmented_striding_rq_cap(struct mlx5_core_dev *mdev) { @@ -1143,7 +1144,6 @@ static int mlx5e_alloc_txqsq(struct mlx5e_channel *c, sq->uar_map = mdev->mlx5e_res.bfreg.map; sq->min_inline_mode = params->tx_min_inline_mode; sq->hw_mtu = MLX5E_SW2HW_MTU(params, params->sw_mtu); - sq->stats = &c->priv->channel_stats[c->ix].sq[tc]; INIT_WORK(&sq->recover_work, mlx5e_tx_err_cqe_work); if (!MLX5_CAP_ETH(mdev, wqe_vlan_insert)) set_bit(MLX5E_SQ_STATE_VLAN_NEED_L2_INLINE, &sq->state); @@ -1233,6 +1233,7 @@ static int mlx5e_create_sq(struct mlx5_core_dev *mdev, int mlx5e_modify_sq(struct mlx5_core_dev *mdev, u32 sqn, struct mlx5e_modify_sq_param *p) { + u64 bitmask = 0; void *in; void *sqc; int inlen; @@ -1248,9 +1249,14 @@ int mlx5e_modify_sq(struct mlx5_core_dev *mdev, u32 sqn, MLX5_SET(modify_sq_in, in, sq_state, p->curr_state); MLX5_SET(sqc, sqc, state, p->next_state); if (p->rl_update && p->next_state == MLX5_SQC_STATE_RDY) { - MLX5_SET64(modify_sq_in, in, modify_bitmask, 1); - MLX5_SET(sqc, sqc, packet_pacing_rate_limit_index, p->rl_index); + bitmask |= 1; + MLX5_SET(sqc, sqc, packet_pacing_rate_limit_index, p->rl_index); } + if (p->qos_update && p->next_state == MLX5_SQC_STATE_RDY) { + bitmask |= 1 << 2; + MLX5_SET(sqc, sqc, qos_queue_group_id, p->qos_queue_group_id); + } + MLX5_SET64(modify_sq_in, in, modify_bitmask, bitmask); err = mlx5_core_modify_sq(mdev, sqn, in); @@ -1267,6 +1273,7 @@ static void mlx5e_destroy_sq(struct mlx5_core_dev *mdev, u32 sqn) int mlx5e_create_sq_rdy(struct mlx5_core_dev *mdev, struct mlx5e_sq_param *param, struct mlx5e_create_sq_param *csp, + u16 qos_queue_group_id, u32 *sqn) { struct mlx5e_modify_sq_param msp = {0}; @@ -1278,6 +1285,10 @@ int mlx5e_create_sq_rdy(struct mlx5_core_dev *mdev, msp.curr_state = MLX5_SQC_STATE_RST; msp.next_state = MLX5_SQC_STATE_RDY; + if (qos_queue_group_id) { + msp.qos_update = true; + msp.qos_queue_group_id = qos_queue_group_id; + } err = mlx5e_modify_sq(mdev, *sqn, &msp); if (err) mlx5e_destroy_sq(mdev, *sqn); @@ -1288,18 +1299,18 @@ int mlx5e_create_sq_rdy(struct mlx5_core_dev *mdev, static int mlx5e_set_sq_maxrate(struct net_device *dev, struct mlx5e_txqsq *sq, u32 rate); -static int mlx5e_open_txqsq(struct mlx5e_channel *c, - u32 tisn, - int txq_ix, - struct mlx5e_params *params, - struct mlx5e_sq_param *param, - struct mlx5e_txqsq *sq, - int tc) +int mlx5e_open_txqsq(struct mlx5e_channel *c, u32 tisn, int txq_ix, + struct mlx5e_params *params, struct mlx5e_sq_param *param, + struct mlx5e_txqsq *sq, int tc, u16 qos_queue_group_id, u16 qos_qid) { struct mlx5e_create_sq_param csp = {}; u32 tx_rate; int err; + if (qos_queue_group_id) + sq->stats = c->priv->htb.qos_sq_stats[qos_qid]; + else + sq->stats = &c->priv->channel_stats[c->ix].sq[tc]; err = mlx5e_alloc_txqsq(c, txq_ix, params, param, sq, tc); if (err) return err; @@ -1309,7 +1320,7 @@ static int mlx5e_open_txqsq(struct mlx5e_channel *c, csp.cqn = sq->cq.mcq.cqn; csp.wq_ctrl = &sq->wq_ctrl; csp.min_inline_mode = sq->min_inline_mode; - err = mlx5e_create_sq_rdy(c->mdev, param, &csp, &sq->sqn); + err = mlx5e_create_sq_rdy(c->mdev, param, &csp, qos_queue_group_id, &sq->sqn); if (err) goto err_free_txqsq; @@ -1366,7 +1377,7 @@ void mlx5e_deactivate_txqsq(struct mlx5e_txqsq *sq) } } -static void mlx5e_close_txqsq(struct mlx5e_txqsq *sq) +void mlx5e_close_txqsq(struct mlx5e_txqsq *sq) { struct mlx5_core_dev *mdev = sq->mdev; struct mlx5_rate_limit rl = {0}; @@ -1403,7 +1414,7 @@ int mlx5e_open_icosq(struct mlx5e_channel *c, struct mlx5e_params *params, csp.cqn = sq->cq.mcq.cqn; csp.wq_ctrl = &sq->wq_ctrl; csp.min_inline_mode = params->tx_min_inline_mode; - err = mlx5e_create_sq_rdy(c->mdev, param, &csp, &sq->sqn); + err = mlx5e_create_sq_rdy(c->mdev, param, &csp, 0, &sq->sqn); if (err) goto err_free_icosq; @@ -1452,7 +1463,7 @@ int mlx5e_open_xdpsq(struct mlx5e_channel *c, struct mlx5e_params *params, csp.wq_ctrl = &sq->wq_ctrl; csp.min_inline_mode = sq->min_inline_mode; set_bit(MLX5E_SQ_STATE_ENABLED, &sq->state); - err = mlx5e_create_sq_rdy(c->mdev, param, &csp, &sq->sqn); + err = mlx5e_create_sq_rdy(c->mdev, param, &csp, 0, &sq->sqn); if (err) goto err_free_xdpsq; @@ -1703,7 +1714,7 @@ static int mlx5e_open_sqs(struct mlx5e_channel *c, int txq_ix = c->ix + tc * params->num_channels; err = mlx5e_open_txqsq(c, c->priv->tisn[c->lag_port][tc], txq_ix, - params, &cparam->txq_sq, &c->sq[tc], tc); + params, &cparam->txq_sq, &c->sq[tc], tc, 0, 0); if (err) goto err_close_sqs; } @@ -2044,6 +2055,7 @@ static void mlx5e_deactivate_channel(struct mlx5e_channel *c) mlx5e_deactivate_icosq(&c->icosq); for (tc = 0; tc < c->num_tc; tc++) mlx5e_deactivate_txqsq(&c->sq[tc]); + mlx5e_qos_deactivate_queues(c); } static void mlx5e_close_channel(struct mlx5e_channel *c) @@ -2051,6 +2063,7 @@ static void mlx5e_close_channel(struct mlx5e_channel *c) if (test_bit(MLX5E_CHANNEL_STATE_XSK, c->state)) mlx5e_close_xsk(c); mlx5e_close_queues(c); + mlx5e_qos_close_queues(c); netif_napi_del(&c->napi); kvfree(c); @@ -2200,9 +2213,8 @@ void mlx5e_build_sq_param_common(struct mlx5e_priv *priv, param->wq.buf_numa_node = dev_to_node(mlx5_core_dma_dev(priv->mdev)); } -static void mlx5e_build_sq_param(struct mlx5e_priv *priv, - struct mlx5e_params *params, - struct mlx5e_sq_param *param) +void mlx5e_build_sq_param(struct mlx5e_priv *priv, struct mlx5e_params *params, + struct mlx5e_sq_param *param) { void *sqc = param->sqc; void *wq = MLX5_ADDR_OF(sqc, sqc, wq); @@ -2381,10 +2393,18 @@ int mlx5e_open_channels(struct mlx5e_priv *priv, goto err_close_channels; } + err = mlx5e_qos_open_queues(priv, chs); + if (err) + goto err_close_ptp; + mlx5e_health_channels_update(priv); kvfree(cparam); return 0; +err_close_ptp: + if (chs->port_ptp) + mlx5e_port_ptp_close(chs->port_ptp); + err_close_channels: for (i--; i >= 0; i--) mlx5e_close_channel(chs->c[i]); @@ -2917,11 +2937,31 @@ static void mlx5e_netdev_set_tcs(struct net_device *netdev, u16 nch, u8 ntc) netdev_set_tc_queue(netdev, tc, nch, 0); } +int mlx5e_update_tx_netdev_queues(struct mlx5e_priv *priv) +{ + int qos_queues, nch, ntc, num_txqs, err; + + qos_queues = mlx5e_qos_cur_leaf_nodes(priv); + + nch = priv->channels.params.num_channels; + ntc = priv->channels.params.num_tc; + num_txqs = nch * ntc + qos_queues; + if (MLX5E_GET_PFLAG(&priv->channels.params, MLX5E_PFLAG_TX_PORT_TS)) + num_txqs += ntc; + + mlx5e_dbg(DRV, priv, "Setting num_txqs %d\n", num_txqs); + err = netif_set_real_num_tx_queues(priv->netdev, num_txqs); + if (err) + netdev_warn(priv->netdev, "netif_set_real_num_tx_queues failed, %d\n", err); + + return err; +} + static int mlx5e_update_netdev_queues(struct mlx5e_priv *priv) { struct net_device *netdev = priv->netdev; - int num_txqs, num_rxqs, nch, ntc; int old_num_txqs, old_ntc; + int num_rxqs, nch, ntc; int err; old_num_txqs = netdev->real_num_tx_queues; @@ -2929,18 +2969,13 @@ static int mlx5e_update_netdev_queues(struct mlx5e_priv *priv) nch = priv->channels.params.num_channels; ntc = priv->channels.params.num_tc; - num_txqs = nch * ntc; - if (MLX5E_GET_PFLAG(&priv->channels.params, MLX5E_PFLAG_TX_PORT_TS)) - num_txqs += ntc; num_rxqs = nch * priv->profile->rq_groups; mlx5e_netdev_set_tcs(netdev, nch, ntc); - err = netif_set_real_num_tx_queues(netdev, num_txqs); - if (err) { - netdev_warn(netdev, "netif_set_real_num_tx_queues failed, %d\n", err); + err = mlx5e_update_tx_netdev_queues(priv); + if (err) goto err_tcs; - } err = netif_set_real_num_rx_queues(netdev, num_rxqs); if (err) { netdev_warn(netdev, "netif_set_real_num_rx_queues failed, %d\n", err); @@ -3044,6 +3079,7 @@ void mlx5e_activate_priv_channels(struct mlx5e_priv *priv) mlx5e_update_num_tc_x_num_ch(priv); mlx5e_build_txq_maps(priv); mlx5e_activate_channels(&priv->channels); + mlx5e_qos_activate_queues(priv); mlx5e_xdp_tx_enable(priv); netif_tx_start_all_queues(priv->netdev); @@ -3609,6 +3645,14 @@ static int mlx5e_setup_tc_mqprio(struct mlx5e_priv *priv, mutex_lock(&priv->state_lock); + /* MQPRIO is another toplevel qdisc that can't be attached + * simultaneously with the offloaded HTB. + */ + if (WARN_ON(priv->htb.maj_id)) { + err = -EINVAL; + goto out; + } + new_channels.params = priv->channels.params; new_channels.params.num_tc = tc ? tc : 1; @@ -3629,12 +3673,49 @@ static int mlx5e_setup_tc_mqprio(struct mlx5e_priv *priv, return err; } +static int mlx5e_setup_tc_htb(struct mlx5e_priv *priv, struct tc_htb_qopt_offload *htb) +{ + int res; + + switch (htb->command) { + case TC_HTB_CREATE: + return mlx5e_htb_root_add(priv, htb->parent_classid, htb->classid); + case TC_HTB_DESTROY: + return mlx5e_htb_root_del(priv); + case TC_HTB_LEAF_ALLOC_QUEUE: + res = mlx5e_htb_leaf_alloc_queue(priv, htb->classid, htb->parent_classid, + htb->rate, htb->ceil); + if (res < 0) + return res; + htb->qid = res; + return 0; + case TC_HTB_LEAF_TO_INNER: + return mlx5e_htb_leaf_to_inner(priv, htb->parent_classid, htb->classid, + htb->rate, htb->ceil); + case TC_HTB_LEAF_DEL: + return mlx5e_htb_leaf_del(priv, htb->classid, &htb->moved_qid, &htb->qid); + case TC_HTB_LEAF_DEL_LAST: + return mlx5e_htb_leaf_del_last(priv, htb->classid); + case TC_HTB_NODE_MODIFY: + return mlx5e_htb_node_modify(priv, htb->classid, htb->rate, htb->ceil); + case TC_HTB_LEAF_QUERY_QUEUE: + res = mlx5e_get_txq_by_classid(priv, htb->classid); + if (res < 0) + return res; + htb->qid = res; + return 0; + default: + return -EOPNOTSUPP; + } +} + static LIST_HEAD(mlx5e_block_cb_list); static int mlx5e_setup_tc(struct net_device *dev, enum tc_setup_type type, void *type_data) { struct mlx5e_priv *priv = netdev_priv(dev); + int err; switch (type) { case TC_SETUP_BLOCK: { @@ -3648,6 +3729,11 @@ static int mlx5e_setup_tc(struct net_device *dev, enum tc_setup_type type, } case TC_SETUP_QDISC_MQPRIO: return mlx5e_setup_tc_mqprio(priv, type_data); + case TC_SETUP_QDISC_HTB: + mutex_lock(&priv->state_lock); + err = mlx5e_setup_tc_htb(priv, type_data); + mutex_unlock(&priv->state_lock); + return err; default: return -EOPNOTSUPP; } @@ -3812,20 +3898,25 @@ static int set_feature_cvlan_filter(struct net_device *netdev, bool enable) return 0; } -#if IS_ENABLED(CONFIG_MLX5_CLS_ACT) -static int set_feature_tc_num_filters(struct net_device *netdev, bool enable) +static int set_feature_hw_tc(struct net_device *netdev, bool enable) { struct mlx5e_priv *priv = netdev_priv(netdev); +#if IS_ENABLED(CONFIG_MLX5_CLS_ACT) if (!enable && mlx5e_tc_num_filters(priv, MLX5_TC_FLAG(NIC_OFFLOAD))) { netdev_err(netdev, "Active offloaded tc filters, can't turn hw_tc_offload off\n"); return -EINVAL; } +#endif + + if (!enable && priv->htb.maj_id) { + netdev_err(netdev, "Active HTB offload, can't turn hw_tc_offload off\n"); + return -EINVAL; + } return 0; } -#endif static int set_feature_rx_all(struct net_device *netdev, bool enable) { @@ -3923,9 +4014,7 @@ int mlx5e_set_features(struct net_device *netdev, netdev_features_t features) err |= MLX5E_HANDLE_FEATURE(NETIF_F_LRO, set_feature_lro); err |= MLX5E_HANDLE_FEATURE(NETIF_F_HW_VLAN_CTAG_FILTER, set_feature_cvlan_filter); -#if IS_ENABLED(CONFIG_MLX5_CLS_ACT) - err |= MLX5E_HANDLE_FEATURE(NETIF_F_HW_TC, set_feature_tc_num_filters); -#endif + err |= MLX5E_HANDLE_FEATURE(NETIF_F_HW_TC, set_feature_hw_tc); err |= MLX5E_HANDLE_FEATURE(NETIF_F_RXALL, set_feature_rx_all); err |= MLX5E_HANDLE_FEATURE(NETIF_F_RXFCS, set_feature_rx_fcs); err |= MLX5E_HANDLE_FEATURE(NETIF_F_HW_VLAN_CTAG_RX, set_feature_rx_vlan); @@ -5033,6 +5122,8 @@ static void mlx5e_build_nic_netdev(struct net_device *netdev) netdev->hw_features |= NETIF_F_NTUPLE; #endif } + if (mlx5_qos_is_supported(mdev)) + netdev->features |= NETIF_F_HW_TC; netdev->features |= NETIF_F_HIGHDMA; netdev->features |= NETIF_F_HW_VLAN_STAG_FILTER; @@ -5338,6 +5429,7 @@ int mlx5e_netdev_init(struct net_device *netdev, return -ENOMEM; mutex_init(&priv->state_lock); + hash_init(priv->htb.qos_tc2node); INIT_WORK(&priv->update_carrier_work, mlx5e_update_carrier_work); INIT_WORK(&priv->set_rx_mode_work, mlx5e_set_rx_mode_work); INIT_WORK(&priv->tx_timeout_work, mlx5e_tx_timeout_work); @@ -5360,8 +5452,14 @@ int mlx5e_netdev_init(struct net_device *netdev, void mlx5e_netdev_cleanup(struct net_device *netdev, struct mlx5e_priv *priv) { + int i; + destroy_workqueue(priv->wq); free_cpumask_var(priv->scratchpad.cpumask); + + for (i = 0; i < priv->htb.max_qos_sqs; i++) + kfree(priv->htb.qos_sq_stats[i]); + kvfree(priv->htb.qos_sq_stats); } struct net_device *mlx5e_create_netdev(struct mlx5_core_dev *mdev, @@ -5371,13 +5469,17 @@ struct net_device *mlx5e_create_netdev(struct mlx5_core_dev *mdev, { struct net_device *netdev; unsigned int ptp_txqs = 0; + int qos_sqs = 0; int err; if (MLX5_CAP_GEN(mdev, ts_cqe_to_dest_cqn)) ptp_txqs = profile->max_tc; + if (mlx5_qos_is_supported(mdev)) + qos_sqs = mlx5e_qos_max_leaf_nodes(mdev); + netdev = alloc_etherdev_mqs(sizeof(struct mlx5e_priv), - nch * profile->max_tc + ptp_txqs, + nch * profile->max_tc + ptp_txqs + qos_sqs, nch * profile->rq_groups); if (!netdev) { mlx5_core_err(mdev, "alloc_etherdev_mqs() failed\n"); diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c index 2cf2042b37c7..92c5b81427b9 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c @@ -420,6 +420,25 @@ static void mlx5e_stats_grp_sw_update_stats_ptp(struct mlx5e_priv *priv, } } +static void mlx5e_stats_grp_sw_update_stats_qos(struct mlx5e_priv *priv, + struct mlx5e_sw_stats *s) +{ + struct mlx5e_sq_stats **stats; + u16 max_qos_sqs; + int i; + + /* Pairs with smp_store_release in mlx5e_open_qos_sq. */ + max_qos_sqs = smp_load_acquire(&priv->htb.max_qos_sqs); + stats = READ_ONCE(priv->htb.qos_sq_stats); + + for (i = 0; i < max_qos_sqs; i++) { + mlx5e_stats_grp_sw_update_stats_sq(s, READ_ONCE(stats[i])); + + /* https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92657 */ + barrier(); + } +} + static MLX5E_DECLARE_STATS_GRP_OP_UPDATE_STATS(sw) { struct mlx5e_sw_stats *s = &priv->stats.sw; @@ -449,6 +468,7 @@ static MLX5E_DECLARE_STATS_GRP_OP_UPDATE_STATS(sw) } } mlx5e_stats_grp_sw_update_stats_ptp(priv, s); + mlx5e_stats_grp_sw_update_stats_qos(priv, s); } static const struct counter_desc q_stats_desc[] = { @@ -1740,6 +1760,41 @@ static const struct counter_desc ptp_cq_stats_desc[] = { { MLX5E_DECLARE_PTP_CQ_STAT(struct mlx5e_ptp_cq_stats, abort_abs_diff_ns) }, }; +static const struct counter_desc qos_sq_stats_desc[] = { + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, packets) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, bytes) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tso_packets) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tso_bytes) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tso_inner_packets) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tso_inner_bytes) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, csum_partial) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, csum_partial_inner) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, added_vlan_packets) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, nop) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, mpwqe_blks) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, mpwqe_pkts) }, +#ifdef CONFIG_MLX5_EN_TLS + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_encrypted_packets) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_encrypted_bytes) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_ctx) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_ooo) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_dump_packets) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_dump_bytes) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_resync_bytes) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_skip_no_sync_data) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_drop_no_sync_data) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, tls_drop_bypass_req) }, +#endif + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, csum_none) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, stopped) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, dropped) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, xmit_more) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, recover) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, cqes) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, wake) }, + { MLX5E_DECLARE_QOS_TX_STAT(struct mlx5e_sq_stats, cqe_err) }, +}; + #define NUM_RQ_STATS ARRAY_SIZE(rq_stats_desc) #define NUM_SQ_STATS ARRAY_SIZE(sq_stats_desc) #define NUM_XDPSQ_STATS ARRAY_SIZE(xdpsq_stats_desc) @@ -1750,6 +1805,49 @@ static const struct counter_desc ptp_cq_stats_desc[] = { #define NUM_PTP_SQ_STATS ARRAY_SIZE(ptp_sq_stats_desc) #define NUM_PTP_CH_STATS ARRAY_SIZE(ptp_ch_stats_desc) #define NUM_PTP_CQ_STATS ARRAY_SIZE(ptp_cq_stats_desc) +#define NUM_QOS_SQ_STATS ARRAY_SIZE(qos_sq_stats_desc) + +static MLX5E_DECLARE_STATS_GRP_OP_NUM_STATS(qos) +{ + /* Pairs with smp_store_release in mlx5e_open_qos_sq. */ + return NUM_QOS_SQ_STATS * smp_load_acquire(&priv->htb.max_qos_sqs); +} + +static MLX5E_DECLARE_STATS_GRP_OP_FILL_STRS(qos) +{ + /* Pairs with smp_store_release in mlx5e_open_qos_sq. */ + u16 max_qos_sqs = smp_load_acquire(&priv->htb.max_qos_sqs); + int i, qid; + + for (qid = 0; qid < max_qos_sqs; qid++) + for (i = 0; i < NUM_QOS_SQ_STATS; i++) + sprintf(data + (idx++) * ETH_GSTRING_LEN, + qos_sq_stats_desc[i].format, qid); + + return idx; +} + +static MLX5E_DECLARE_STATS_GRP_OP_FILL_STATS(qos) +{ + struct mlx5e_sq_stats **stats; + u16 max_qos_sqs; + int i, qid; + + /* Pairs with smp_store_release in mlx5e_open_qos_sq. */ + max_qos_sqs = smp_load_acquire(&priv->htb.max_qos_sqs); + stats = READ_ONCE(priv->htb.qos_sq_stats); + + for (qid = 0; qid < max_qos_sqs; qid++) { + struct mlx5e_sq_stats *s = READ_ONCE(stats[qid]); + + for (i = 0; i < NUM_QOS_SQ_STATS; i++) + data[idx++] = MLX5E_READ_CTR64_CPU(s, qos_sq_stats_desc, i); + } + + return idx; +} + +static MLX5E_DECLARE_STATS_GRP_OP_UPDATE_STATS(qos) { return; } static MLX5E_DECLARE_STATS_GRP_OP_NUM_STATS(ptp) { @@ -1932,6 +2030,7 @@ MLX5E_DEFINE_STATS_GRP(per_port_buff_congest, 0); MLX5E_DEFINE_STATS_GRP(eth_ext, 0); static MLX5E_DEFINE_STATS_GRP(tls, 0); static MLX5E_DEFINE_STATS_GRP(ptp, 0); +static MLX5E_DEFINE_STATS_GRP(qos, 0); /* The stats groups order is opposite to the update_stats() order calls */ mlx5e_stats_grp_t mlx5e_nic_stats_grps[] = { @@ -1955,6 +2054,7 @@ mlx5e_stats_grp_t mlx5e_nic_stats_grps[] = { &MLX5E_STATS_GRP(channels), &MLX5E_STATS_GRP(per_port_buff_congest), &MLX5E_STATS_GRP(ptp), + &MLX5E_STATS_GRP(qos), }; unsigned int mlx5e_nic_stats_grps_num(struct mlx5e_priv *priv) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h index e41fc11f2ce7..93c41312fb03 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h @@ -55,6 +55,8 @@ #define MLX5E_DECLARE_PTP_CH_STAT(type, fld) "ptp_ch_"#fld, offsetof(type, fld) #define MLX5E_DECLARE_PTP_CQ_STAT(type, fld) "ptp_cq%d_"#fld, offsetof(type, fld) +#define MLX5E_DECLARE_QOS_TX_STAT(type, fld) "qos_tx%d_"#fld, offsetof(type, fld) + struct counter_desc { char format[ETH_GSTRING_LEN]; size_t offset; /* Byte offset */ diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c index e47e2a0059d0..a2ab5c36f4c5 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c @@ -106,28 +106,53 @@ static u16 mlx5e_select_ptpsq(struct net_device *dev, struct sk_buff *skb) return priv->port_ptp_tc2realtxq[up]; } +static int mlx5e_select_htb_queue(struct mlx5e_priv *priv, struct sk_buff *skb, + u16 htb_maj_id) +{ + u16 classid; + + if ((TC_H_MAJ(skb->priority) >> 16) == htb_maj_id) + classid = TC_H_MIN(skb->priority); + else + classid = READ_ONCE(priv->htb.defcls); + + if (!classid) + return 0; + + return mlx5e_get_txq_by_classid(priv, classid); +} + u16 mlx5e_select_queue(struct net_device *dev, struct sk_buff *skb, struct net_device *sb_dev) { struct mlx5e_priv *priv = netdev_priv(dev); + int num_tc_x_num_ch; int txq_ix; int up = 0; int ch_ix; - if (unlikely(priv->channels.port_ptp)) { - int num_tc_x_num_ch; + /* Sync with mlx5e_update_num_tc_x_num_ch - avoid refetching. */ + num_tc_x_num_ch = READ_ONCE(priv->num_tc_x_num_ch); + if (unlikely(dev->real_num_tx_queues > num_tc_x_num_ch)) { + /* Order maj_id before defcls - pairs with mlx5e_htb_root_add. */ + u16 htb_maj_id = smp_load_acquire(&priv->htb.maj_id); - if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) && - mlx5e_use_ptpsq(skb)) - return mlx5e_select_ptpsq(dev, skb); + if (unlikely(htb_maj_id)) { + txq_ix = mlx5e_select_htb_queue(priv, skb, htb_maj_id); + if (txq_ix > 0) + return txq_ix; + } - /* Sync with mlx5e_update_num_tc_x_num_ch - avoid refetching. */ - num_tc_x_num_ch = READ_ONCE(priv->num_tc_x_num_ch); + if (unlikely(priv->channels.port_ptp)) + if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) && + mlx5e_use_ptpsq(skb)) + return mlx5e_select_ptpsq(dev, skb); txq_ix = netdev_pick_tx(dev, skb, NULL); - /* Fix netdev_pick_tx() not to choose ptp_channel txqs. + /* Fix netdev_pick_tx() not to choose ptp_channel and HTB txqs. * If they are selected, switch to regular queues. - * Driver to select these queues only at mlx5e_select_ptpsq(). + * Driver to select these queues only at mlx5e_select_ptpsq() + * and mlx5e_select_htb_queue(). */ if (unlikely(txq_ix >= num_tc_x_num_ch)) txq_ix %= num_tc_x_num_ch; @@ -703,6 +728,10 @@ netdev_tx_t mlx5e_xmit(struct sk_buff *skb, struct net_device *dev) u16 pi; sq = priv->txq2sq[skb_get_queue_mapping(skb)]; + if (unlikely(!sq)) { + dev_kfree_skb_any(skb); + return NETDEV_TX_OK; + } /* May send SKBs and WQEs. */ if (unlikely(!mlx5e_accel_tx_begin(dev, sq, skb, &accel))) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c index 1ec3d62f026d..cd1f288cdeeb 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c @@ -121,15 +121,19 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget) struct mlx5e_xdpsq *xsksq = &c->xsksq; struct mlx5e_rq *xskrq = &c->xskrq; struct mlx5e_rq *rq = &c->rq; + struct mlx5e_txqsq **qos_sqs; bool aff_change = false; bool busy_xsk = false; bool busy = false; int work_done = 0; + u16 qos_sqs_size; bool xsk_open; int i; rcu_read_lock(); + qos_sqs = rcu_dereference(c->qos_sqs); + xsk_open = test_bit(MLX5E_CHANNEL_STATE_XSK, c->state); ch_stats->poll++; @@ -137,6 +141,18 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget) for (i = 0; i < c->num_tc; i++) busy |= mlx5e_poll_tx_cq(&c->sq[i].cq, budget); + if (unlikely(qos_sqs)) { + smp_rmb(); /* Pairs with mlx5e_qos_alloc_queues. */ + qos_sqs_size = READ_ONCE(c->qos_sqs_size); + + for (i = 0; i < qos_sqs_size; i++) { + struct mlx5e_txqsq *sq = rcu_dereference(qos_sqs[i]); + + if (sq) + busy |= mlx5e_poll_tx_cq(&sq->cq, budget); + } + } + busy |= mlx5e_poll_xdpsq_cq(&c->xdpsq.cq); if (c->xdp) @@ -190,6 +206,16 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget) mlx5e_handle_tx_dim(&c->sq[i]); mlx5e_cq_arm(&c->sq[i].cq); } + if (unlikely(qos_sqs)) { + for (i = 0; i < qos_sqs_size; i++) { + struct mlx5e_txqsq *sq = rcu_dereference(qos_sqs[i]); + + if (sq) { + mlx5e_handle_tx_dim(sq); + mlx5e_cq_arm(&sq->cq); + } + } + } mlx5e_handle_rx_dim(rq); diff --git a/drivers/net/ethernet/mellanox/mlx5/core/qos.c b/drivers/net/ethernet/mellanox/mlx5/core/qos.c new file mode 100644 index 000000000000..0777be24a307 --- /dev/null +++ b/drivers/net/ethernet/mellanox/mlx5/core/qos.c @@ -0,0 +1,85 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* Copyright (c) 2020, Mellanox Technologies inc. All rights reserved. */ + +#include "qos.h" + +#define MLX5_QOS_DEFAULT_DWRR_UID 0 + +bool mlx5_qos_is_supported(struct mlx5_core_dev *mdev) +{ + if (!MLX5_CAP_GEN(mdev, qos)) + return false; + if (!MLX5_CAP_QOS(mdev, nic_sq_scheduling)) + return false; + if (!MLX5_CAP_QOS(mdev, nic_bw_share)) + return false; + if (!MLX5_CAP_QOS(mdev, nic_rate_limit)) + return false; + return true; +} + +int mlx5_qos_max_leaf_nodes(struct mlx5_core_dev *mdev) +{ + return 1 << MLX5_CAP_QOS(mdev, log_max_qos_nic_queue_group); +} + +int mlx5_qos_create_leaf_node(struct mlx5_core_dev *mdev, u32 parent_id, + u32 bw_share, u32 max_avg_bw, u32 *id) +{ + u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0}; + + MLX5_SET(scheduling_context, sched_ctx, parent_element_id, parent_id); + MLX5_SET(scheduling_context, sched_ctx, element_type, + SCHEDULING_CONTEXT_ELEMENT_TYPE_QUEUE_GROUP); + MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share); + MLX5_SET(scheduling_context, sched_ctx, max_average_bw, max_avg_bw); + + return mlx5_create_scheduling_element_cmd(mdev, SCHEDULING_HIERARCHY_NIC, + sched_ctx, id); +} + +int mlx5_qos_create_inner_node(struct mlx5_core_dev *mdev, u32 parent_id, + u32 bw_share, u32 max_avg_bw, u32 *id) +{ + u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0}; + void *attr; + + MLX5_SET(scheduling_context, sched_ctx, parent_element_id, parent_id); + MLX5_SET(scheduling_context, sched_ctx, element_type, + SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR); + MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share); + MLX5_SET(scheduling_context, sched_ctx, max_average_bw, max_avg_bw); + + attr = MLX5_ADDR_OF(scheduling_context, sched_ctx, element_attributes); + MLX5_SET(tsar_element, attr, tsar_type, TSAR_ELEMENT_TSAR_TYPE_DWRR); + + return mlx5_create_scheduling_element_cmd(mdev, SCHEDULING_HIERARCHY_NIC, + sched_ctx, id); +} + +int mlx5_qos_create_root_node(struct mlx5_core_dev *mdev, u32 *id) +{ + return mlx5_qos_create_inner_node(mdev, MLX5_QOS_DEFAULT_DWRR_UID, 0, 0, id); +} + +int mlx5_qos_update_node(struct mlx5_core_dev *mdev, u32 parent_id, + u32 bw_share, u32 max_avg_bw, u32 id) +{ + u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0}; + u32 bitmask = 0; + + MLX5_SET(scheduling_context, sched_ctx, parent_element_id, parent_id); + MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share); + MLX5_SET(scheduling_context, sched_ctx, max_average_bw, max_avg_bw); + + bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE; + bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW; + + return mlx5_modify_scheduling_element_cmd(mdev, SCHEDULING_HIERARCHY_NIC, + sched_ctx, id, bitmask); +} + +int mlx5_qos_destroy_node(struct mlx5_core_dev *mdev, u32 id) +{ + return mlx5_destroy_scheduling_element_cmd(mdev, SCHEDULING_HIERARCHY_NIC, id); +} diff --git a/drivers/net/ethernet/mellanox/mlx5/core/qos.h b/drivers/net/ethernet/mellanox/mlx5/core/qos.h new file mode 100644 index 000000000000..7af6e5e1df5e --- /dev/null +++ b/drivers/net/ethernet/mellanox/mlx5/core/qos.h @@ -0,0 +1,28 @@ +/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */ +/* Copyright (c) 2020, Mellanox Technologies inc. All rights reserved. */ + +#ifndef __MLX5_QOS_H +#define __MLX5_QOS_H + +#include "mlx5_core.h" + +#define MLX5_DEBUG_QOS_MASK BIT(4) + +#define qos_warn(mdev, fmt, ...) \ + mlx5_core_warn(mdev, "QoS: " fmt, ##__VA_ARGS__) +#define qos_dbg(mdev, fmt, ...) \ + mlx5_core_dbg_mask(mdev, MLX5_DEBUG_QOS_MASK, "QoS: " fmt, ##__VA_ARGS__) + +bool mlx5_qos_is_supported(struct mlx5_core_dev *mdev); +int mlx5_qos_max_leaf_nodes(struct mlx5_core_dev *mdev); + +int mlx5_qos_create_leaf_node(struct mlx5_core_dev *mdev, u32 parent_id, + u32 bw_share, u32 max_avg_bw, u32 *id); +int mlx5_qos_create_inner_node(struct mlx5_core_dev *mdev, u32 parent_id, + u32 bw_share, u32 max_avg_bw, u32 *id); +int mlx5_qos_create_root_node(struct mlx5_core_dev *mdev, u32 *id); +int mlx5_qos_update_node(struct mlx5_core_dev *mdev, u32 parent_id, u32 bw_share, + u32 max_avg_bw, u32 id); +int mlx5_qos_destroy_node(struct mlx5_core_dev *mdev, u32 id); + +#endif diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 0d6e287d614f..58928e3b4e20 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -842,10 +842,13 @@ struct mlx5_ifc_qos_cap_bits { u8 reserved_at_4[0x1]; u8 packet_pacing_burst_bound[0x1]; u8 packet_pacing_typical_size[0x1]; - u8 reserved_at_7[0x4]; + u8 reserved_at_7[0x1]; + u8 nic_sq_scheduling[0x1]; + u8 nic_bw_share[0x1]; + u8 nic_rate_limit[0x1]; u8 packet_pacing_uid[0x1]; - u8 reserved_at_c[0x14]; - + u8 reserved_at_c[0xf]; + u8 log_max_qos_nic_queue_group[0x5]; u8 reserved_at_20[0x20]; u8 packet_pacing_max_rate[0x20]; @@ -3344,7 +3347,7 @@ struct mlx5_ifc_sqc_bits { u8 reserved_at_e0[0x10]; u8 packet_pacing_rate_limit_index[0x10]; u8 tis_lst_sz[0x10]; - u8 reserved_at_110[0x10]; + u8 qos_queue_group_id[0x10]; u8 reserved_at_120[0x40]; @@ -3359,6 +3362,7 @@ enum { SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT = 0x1, SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT_TC = 0x2, SCHEDULING_CONTEXT_ELEMENT_TYPE_PARA_VPORT_TC = 0x3, + SCHEDULING_CONTEXT_ELEMENT_TYPE_QUEUE_GROUP = 0x4, }; enum { @@ -4802,6 +4806,7 @@ struct mlx5_ifc_query_scheduling_element_out_bits { enum { SCHEDULING_HIERARCHY_E_SWITCH = 0x2, + SCHEDULING_HIERARCHY_NIC = 0x3, }; struct mlx5_ifc_query_scheduling_element_in_bits {