From patchwork Sat Sep 24 23:21:36 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Adit Ranadive X-Patchwork-Id: 674453 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 3shRCB22slz9s5w for ; Sun, 25 Sep 2016 09:24:38 +1000 (AEST) Authentication-Results: ozlabs.org; dkim=pass (1024-bit key; unprotected) header.d=onevmw.onmicrosoft.com header.i=@onevmw.onmicrosoft.com header.b=sd69rnNq; dkim-atps=neutral Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S965639AbcIXXYd (ORCPT ); Sat, 24 Sep 2016 19:24:33 -0400 Received: from ex13-edg-ou-001.vmware.com ([208.91.0.189]:24765 "EHLO EX13-EDG-OU-001.vmware.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S935566AbcIXXYK (ORCPT ); Sat, 24 Sep 2016 19:24:10 -0400 Received: from sc9-mailhost3.vmware.com (10.113.161.73) by EX13-EDG-OU-001.vmware.com (10.113.208.155) with Microsoft SMTP Server id 15.0.1156.6; Sat, 24 Sep 2016 16:23:05 -0700 Received: from EX13-CAS-005.vmware.com (smtp-inbound.vmware.com [10.113.191.55]) by sc9-mailhost3.vmware.com (Postfix) with ESMTP id 18E2C404B5; Sat, 24 Sep 2016 16:23:32 -0700 (PDT) Received: from EX13-MBX-TERM.vmware.com (10.113.191.143) by EX13-MBX-026.vmware.com (10.113.191.46) with Microsoft SMTP Server (TLS) id 15.0.1156.6; Sat, 24 Sep 2016 16:22:32 -0700 Received: from EX13-CAS-001.vmware.com (10.113.191.51) by EX13-MBX-TERM.vmware.com (10.113.191.143) with Microsoft SMTP Server (TLS) id 15.0.1156.6; Sat, 24 Sep 2016 16:22:32 -0700 Received: from NAM03-CO1-obe.outbound.protection.outlook.com (10.113.170.11) by EX13-CAS-001.vmware.com (10.113.191.51) with Microsoft SMTP Server (TLS) id 15.0.1156.6 via Frontend Transport; Sat, 24 Sep 2016 16:22:32 -0700 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=onevmw.onmicrosoft.com; s=selector1-vmware-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version; bh=T8qk1yzNqzMU3upQR6/s5UkxjGz1duqXEM28kkEOU6Y=; b=sd69rnNqBpjbl6Dsa7ycLw4J4jaw1hX0DSnoKehDMxqZYvSq1SQdCm/mS8y2qhdD40CVtFLfH/zaP8GQMyKuY/jxthdHxWDxhbnODPOM1b0t3siZ/DI8VZtClen+KmcYJHFrs0EXzTerNJBdG2JPxoY/QMPVXqz44C2JWVaa05U= Authentication-Results: spf=none (sender IP is ) smtp.mailfrom=aditr@vmware.com; Received: from promb-2s-dhcp95-136.eng.vmware.com (208.91.1.34) by CO2PR0501MB837.namprd05.prod.outlook.com (10.141.244.147) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA384_P384) id 15.1.639.2; Sat, 24 Sep 2016 23:22:26 +0000 From: Adit Ranadive To: , , CC: Adit Ranadive , , , , , , Subject: [PATCH v5 12/16] IB/pvrdma: Add Queue Pair support Date: Sat, 24 Sep 2016 16:21:36 -0700 Message-ID: <671b72d9164f7483d63c2c927a04ac39998ccb3f.1474759181.git.aditr@vmware.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: References: MIME-Version: 1.0 X-Originating-IP: [208.91.1.34] X-ClientProxiedBy: CY1PR18CA0027.namprd18.prod.outlook.com (10.163.31.37) To CO2PR0501MB837.namprd05.prod.outlook.com (10.141.244.147) X-MS-Office365-Filtering-Correlation-Id: 05f2a23e-9784-46b8-e3c1-08d3e4d1a625 X-Microsoft-Exchange-Diagnostics: 1; CO2PR0501MB837; 2:sGRPh12iLG+FeWhPdoOd6HVx4Efq+XTTjCwazABVADVG+OAXBrJaotwNkYtO2p8c4uIor+UkfhQxdmSbME/Xp1P6dXyBVTlUSrUGcN9ldGDgxvKlEKl2hT1QkKXct1PTJnYQaTUKiYBVgkHY6tanXWXpmd0JPqDIWV0YV26agyihSxEU2sRWjPdLAjyzxTDZ; 3:JSniiRIogrO1Ywb2WR1teBQvB8Wju0WuC1L+s4+NI1wwL2JNmpXJ2PdLejS1o5smT4UA+ABcpZh3GpSPGCXqGt+KgKCag+3qUeHp8IlzoOQWmiiHWrGsTsdxN8iL75N1 X-Microsoft-Antispam: UriScan:;BCL:0;PCL:0;RULEID:;SRVR:CO2PR0501MB837; X-Microsoft-Exchange-Diagnostics: 1; CO2PR0501MB837; 25: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 X-LD-Processed: b39138ca-3cee-4b4a-a4d6-cd83d9dd62f0,ExtAddr X-Microsoft-Exchange-Diagnostics: 1; CO2PR0501MB837; 31:0NXsd4f9hZiHIzcrCNEd01Vxqtvu0zin54wbkDHul36tvmtK0VQNc7wQqqLVvvassUW0IVnxnlvW2f0stIflEsoGB6xE9Jig4yZuiiTLrEWH9bgfMBCKXU5hTyueqggkKwCJVqzEo2o6uDbDfhqATROURbHwm+gKxy4v5mLqJgaRtS5s4qiMw1rzPzImOu2P3Wf6GsswQxptbkHzR44XrK64thBcH/ce0F8Lzgg1goQ=; 20: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 X-Microsoft-Antispam-PRVS: X-Exchange-Antispam-Report-Test: UriScan:(250305191791016)(61668805478150)(22074186197030)(146099531331640); X-Exchange-Antispam-Report-CFA-Test: BCL:0; PCL:0; RULEID:(6040176)(601004)(2401047)(8121501046)(5005006)(3002001)(10201501046); SRVR:CO2PR0501MB837; BCL:0; PCL:0; RULEID:; SRVR:CO2PR0501MB837; X-Microsoft-Exchange-Diagnostics: 1; CO2PR0501MB837; 4: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 X-Forefront-PRVS: 0075CB064E X-Forefront-Antispam-Report: SFV:NSPM; SFS:(10009020)(4630300001)(6009001)(7916002)(189002)(199003)(107886002)(77096005)(97736004)(4326007)(106356001)(105586002)(15975445007)(76176999)(586003)(6862003)(118296001)(50986999)(2950100002)(33646002)(6116002)(5001770100001)(19580395003)(5660300001)(19580405001)(92566002)(2906002)(3846002)(47776003)(189998001)(50466002)(86362001)(48376002)(2201001)(81156014)(81166006)(101416001)(8676002)(50226002)(5003940100001)(36756003)(7736002)(7846002)(305945005)(229853001)(42186005)(68736007)(66066001)(4001430100002)(7099028)(2004002); DIR:OUT; SFP:1101; SCL:1; SRVR:CO2PR0501MB837; H:promb-2s-dhcp95-136.eng.vmware.com; FPR:; SPF:None; PTR:InfoNoRecords; MX:1; A:1; LANG:en; Received-SPF: None (EX13-EDG-OU-001.vmware.com: aditr@vmware.com does not designate permitted sender hosts) Received-SPF: None (protection.outlook.com: vmware.com does not designate permitted sender hosts) X-Microsoft-Exchange-Diagnostics: =?us-ascii?Q?1; CO2PR0501MB837; 23:eD3tw1Ai8n4Ckgw4/08SUh8rDmMigMSKb0o2GGSt?= =?us-ascii?Q?JHlXcoNpxhcJxSA0aSm5zJRMcNLiOCMxBg4yfBKvZ3LuxCIWtLifAjlWWS9R?= =?us-ascii?Q?WBEMxutbo+pe/Q/5BP9bGJPLSOT744NUT+Nj25dUT51Fo42B61aL357mbJNK?= =?us-ascii?Q?zE7E0GUywHM23zHbO84oGZJro7IcIwWqSOyedSLp0dule6mgEe93jg9uWeKR?= =?us-ascii?Q?sXdr+Aehdkcn4oEHoDN23NOzXa6WOg0fh5Q3t2pcJ5FRDPPYKSMRwVyRhvQA?= =?us-ascii?Q?hlnirJYQH0zRetmE7MJ46Qx1/4JYO2MiEJtnboo8w2Za0BmMOQdS7djiIj8U?= =?us-ascii?Q?JpEFrWaqNk7pSMlNEk69fV/sXQXC5PmG+EkNBRQl1Cvl7NjWMRWD9EzGQjGu?= =?us-ascii?Q?CWnCWKfVUTgWhy9RLDKXC0kx6fu5vfLm3cOUR4B2KfD/vLIHqRJfWucPkWMY?= =?us-ascii?Q?wfMD1qZBX0Ps6DHQpJbY2aJbDAkUUWKY0c8lOt6kZ0BGmDaPwRjY7EOs1Kg8?= =?us-ascii?Q?OQqTIiHWp6tG9sEZiKQi18dJrLTlFdjVMWCkWCz9B+CMuL1QCGi+7B5KiBwu?= =?us-ascii?Q?3NY5naUxrUA8PfqwJmH80SqK8WxmdOrNs5zmZBn+jKfOwL7DN7M9bMEH3lWt?= =?us-ascii?Q?XkYeuHsP2fTvVu/X9w4zcjzBHzIK1Vx3Pn+tq6GiqaAWI2gNNosvwL0F60/l?= =?us-ascii?Q?N2QICM+qahqORlyJS7Spz5kDtKCNX70CyBmyvi4bB3yb8A+WndpgnfvcdejO?= =?us-ascii?Q?zP6RMgFn/NdjdnBRiWYuFvQlgRL9LtJLNsvJQU42RG4KqMtY3fkCG7N+oAwP?= =?us-ascii?Q?xweZRg+Lwv509UNBZELRA24Npgw69O8WnR4JBfsmRLHM/7ydXibSyfDw7mIR?= =?us-ascii?Q?Tjs5/9I8OuqHgCkTfhDlbVhghD8Aem6NIellB/MwOdFFH0DEzSBL8xPMinLp?= =?us-ascii?Q?ZCj1tOccZGTj94+ncykgkd7V5bzH1ID9E3PpuneI1dE2cxVkRusquuR12hke?= =?us-ascii?Q?tDFWhHt4sM/DlEr9o7CI543hii/V9T70y6etnGQ9E8zoYSnXC4l9IfM1+g8M?= =?us-ascii?Q?jLmW3ILEqkYkzlPeochKJLtVbQ3LYp1yHvsCbpv8Tu0L4hkCsE8iBJxmLF6t?= =?us-ascii?Q?QZGYZkZORRHKPwbsQuyS2X5YHTcBDFGxXg+HUgh4CmlX3xq6rxUdDfYkYEzK?= =?us-ascii?Q?wj0jr+bmmYshLjxSzeuGnrRSy1k6tMEOyQEjjxJVEpzi0aOJaRWg26Y2Yg?= =?us-ascii?Q?=3D=3D?= X-Microsoft-Exchange-Diagnostics: 1; CO2PR0501MB837; 6:c9xxGLEdwc/gAUIz1Gx0TrV3wvpip/drYi9GibWJu05uCQLatLuIHKTRxGAdulRjSwuoxyYWvRB2XZZr/HjXEdEINBo61WZebEbRju+LfIJVm0rToYrEncs+T6Mrqk57/KJHAg7T0Q0/PWDQLv32niEm066sTfvtwLTmbKYnen1LbUkj0C5MnLFGcKqTtgIAq6YOUN6YfseQLq6sJB8JpT4k4MyG6IsMl5Db/GOKtOvoDigcFODZIA01bY19+JVcWIcUBI+q83b22HqMsWTfeBoqb5/jcsAQTXyTNJPo5YI=; 5:QWLzLaRgzEwDzcBNZiw9xF4LShMevVjMyDzezVcbOB/j5qOSuaUa19WNiK6J2FKks6H539jD/4mdiHTo/7wrKASSgUqStmCj+lmBi08ZO2oNCqj9oxOEKa2HD6awRj6J3bQ/jDou8WsKSXZuOm8pXw==; 24:VJSdZNt6BsAi7pCbhYvTE3onxdYTi3V7h8ikUVjGTUchh1LbVThxoRcvK38T5/6Sd2y4ITPoRTn6QmOYeCgiWLIUt1qDRWMAtSQIt/r2bJs=; 7:QOAcdfaBzUOf7TFc5ZAvioltX1ZE6tgWJuTD1vmM5pTwy2iDHd99i9wYW8GeHj7FucAlgsTMlvpWhAXcY8wql1QgAWIJT2jlnbwz+V9GSIw+k2oeSnwcaPJlYKj2ctbBWjxCn/zBL244O7rdEPtllPh2CIQZFVblXNNaFN+oZiPCKlAb2JUopLKyH3RtRM3lw67wZ0CQYT6U7WgmDOdDeA/PB3PJqzdzOW3GyXO8p6Gyo7FWEl50LiCb26aNJko1fkVQLCCOOzXIa8arZqizjJ5Uvo/dJkxYhAquAjBOiTgFqQD2qPY3fzAaaQL7QXyO SpamDiagnosticOutput: 1:99 SpamDiagnosticMetadata: NSPM X-Microsoft-Exchange-Diagnostics: 1; CO2PR0501MB837; 20:76gvTOHH9xQAXlCrdg8EuqeqsIokTPZKiWb1ae2A+R2O7DoGzkmnjbMFUG9ohucL3vW/c7qXdnPxZX43gNeEDN91euMTUrD8yLsGD81TTQoIS9CTNx6KbKhROun7m9AO4qdH6IlOaF9MO7nh//YWN81brUylxwRNPmY3+JuCidI= X-MS-Exchange-CrossTenant-OriginalArrivalTime: 24 Sep 2016 23:22:26.3850 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-Transport-CrossTenantHeadersStamped: CO2PR0501MB837 Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org This patch adds the ability to create, modify, query and destroy QPs. The PVRDMA device supports RC, UD and GSI QPs. Reviewed-by: Yuval Shaia Reviewed-by: Jorgen Hansen Reviewed-by: George Zhang Reviewed-by: Aditya Sarwade Reviewed-by: Bryan Tan Signed-off-by: Adit Ranadive --- Changes v4->v5: - Updated include for headers in UAPI folder. - Update to pvrdma_cmd_post for creating/destroying/querying/modifying QPs. - Use the pvrdma_sge struct when posting WRs/allocating QP memory. - Removed two set but unused variables. Changes v3->v4: - Removed an unnecessary switch case. - Unified the returns in pvrdma_create_qp to use one exit point. - Renamed pvrdma_flush_cqe to _pvrdma_flush_cqe since we need a lock to be held when calling this. - Updated to use wrapper for UAR write for QP. - Updated conversion function to func_name(dst, src) format. - Renamed max_gs to max_sg. - Renamed cap variable to req_cap in pvrdma_set_sq/rq_size. - Changed dev_warn to dev_warn_ratelimited in pvrdma_post_send/recv. - Added nesting locking for flushing CQs when destroying/resetting a QP. - Added missing ret value. Changes v2->v3: - Removed boolean in pvrdma_cmd_post. --- drivers/infiniband/hw/pvrdma/pvrdma_qp.c | 973 +++++++++++++++++++++++++++++++ 1 file changed, 973 insertions(+) create mode 100644 drivers/infiniband/hw/pvrdma/pvrdma_qp.c diff --git a/drivers/infiniband/hw/pvrdma/pvrdma_qp.c b/drivers/infiniband/hw/pvrdma/pvrdma_qp.c new file mode 100644 index 0000000..52c689d --- /dev/null +++ b/drivers/infiniband/hw/pvrdma/pvrdma_qp.c @@ -0,0 +1,973 @@ +/* + * Copyright (c) 2012-2016 VMware, Inc. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of EITHER the GNU General Public License + * version 2 as published by the Free Software Foundation or the BSD + * 2-Clause License. This program is distributed in the hope that it + * will be useful, but WITHOUT ANY WARRANTY; WITHOUT EVEN THE IMPLIED + * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. + * See the GNU General Public License version 2 for more details at + * http://www.gnu.org/licenses/old-licenses/gpl-2.0.en.html. + * + * You should have received a copy of the GNU General Public License + * along with this program available in the file COPYING in the main + * directory of this source tree. + * + * The BSD 2-Clause License + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS + * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE + * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, + * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR + * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, + * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED + * OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include +#include +#include +#include +#include +#include +#include + +#include "pvrdma.h" + +static inline void get_cqs(struct pvrdma_qp *qp, struct pvrdma_cq **send_cq, + struct pvrdma_cq **recv_cq) +{ + *send_cq = to_vcq(qp->ibqp.send_cq); + *recv_cq = to_vcq(qp->ibqp.recv_cq); +} + +static void pvrdma_lock_cqs(struct pvrdma_cq *scq, struct pvrdma_cq *rcq, + unsigned long *scq_flags, + unsigned long *rcq_flags) + __acquires(scq->cq_lock) __acquires(rcq->cq_lock) +{ + if (scq == rcq) { + spin_lock_irqsave(&scq->cq_lock, *scq_flags); + __acquire(rcq->cq_lock); + } else if (scq->cq_handle < rcq->cq_handle) { + spin_lock_irqsave(&scq->cq_lock, *scq_flags); + spin_lock_irqsave_nested(&rcq->cq_lock, *rcq_flags, + SINGLE_DEPTH_NESTING); + } else { + spin_lock_irqsave(&rcq->cq_lock, *rcq_flags); + spin_lock_irqsave_nested(&scq->cq_lock, *scq_flags, + SINGLE_DEPTH_NESTING); + } +} + +static void pvrdma_unlock_cqs(struct pvrdma_cq *scq, struct pvrdma_cq *rcq, + unsigned long *scq_flags, + unsigned long *rcq_flags) + __releases(scq->cq_lock) __releases(rcq->cq_lock) +{ + if (scq == rcq) { + __release(rcq->cq_lock); + spin_unlock_irqrestore(&scq->cq_lock, *scq_flags); + } else if (scq->cq_handle < rcq->cq_handle) { + spin_unlock_irqrestore(&rcq->cq_lock, *rcq_flags); + spin_unlock_irqrestore(&scq->cq_lock, *scq_flags); + } else { + spin_unlock_irqrestore(&scq->cq_lock, *scq_flags); + spin_unlock_irqrestore(&rcq->cq_lock, *rcq_flags); + } +} + +static void pvrdma_reset_qp(struct pvrdma_qp *qp) +{ + struct pvrdma_cq *scq, *rcq; + unsigned long scq_flags, rcq_flags; + + /* Clean up cqes */ + get_cqs(qp, &scq, &rcq); + pvrdma_lock_cqs(scq, rcq, &scq_flags, &rcq_flags); + + _pvrdma_flush_cqe(qp, scq); + if (scq != rcq) + _pvrdma_flush_cqe(qp, rcq); + + pvrdma_unlock_cqs(scq, rcq, &scq_flags, &rcq_flags); + + /* + * Reset queuepair. The checks are because usermode queuepairs won't + * have kernel ringstates. + */ + if (qp->rq.ring) { + atomic_set(&qp->rq.ring->cons_head, 0); + atomic_set(&qp->rq.ring->prod_tail, 0); + } + if (qp->sq.ring) { + atomic_set(&qp->sq.ring->cons_head, 0); + atomic_set(&qp->sq.ring->prod_tail, 0); + } +} + +static int pvrdma_set_rq_size(struct pvrdma_dev *dev, + struct ib_qp_cap *req_cap, + struct pvrdma_qp *qp) +{ + if (req_cap->max_recv_wr > dev->dsr->caps.max_qp_wr || + req_cap->max_recv_sge > dev->dsr->caps.max_sge) { + dev_warn(&dev->pdev->dev, "recv queue size invalid\n"); + return -EINVAL; + } + + qp->rq.wqe_cnt = roundup_pow_of_two(max(1U, req_cap->max_recv_wr)); + qp->rq.max_sg = roundup_pow_of_two(max(1U, req_cap->max_recv_sge)); + + /* Write back */ + req_cap->max_recv_wr = qp->rq.wqe_cnt; + req_cap->max_recv_sge = qp->rq.max_sg; + + qp->rq.wqe_size = roundup_pow_of_two(sizeof(struct pvrdma_rq_wqe_hdr) + + sizeof(struct pvrdma_sge) * + qp->rq.max_sg); + qp->npages_recv = (qp->rq.wqe_cnt * qp->rq.wqe_size + PAGE_SIZE - 1) / + PAGE_SIZE; + + return 0; +} + +static int pvrdma_set_sq_size(struct pvrdma_dev *dev, struct ib_qp_cap *req_cap, + enum ib_qp_type type, struct pvrdma_qp *qp) +{ + if (req_cap->max_send_wr > dev->dsr->caps.max_qp_wr || + req_cap->max_send_sge > dev->dsr->caps.max_sge) { + dev_warn(&dev->pdev->dev, "send queue size invalid\n"); + return -EINVAL; + } + + qp->sq.wqe_cnt = roundup_pow_of_two(max(1U, req_cap->max_send_wr)); + qp->sq.max_sg = roundup_pow_of_two(max(1U, req_cap->max_send_sge)); + + /* Write back */ + req_cap->max_send_wr = qp->sq.wqe_cnt; + req_cap->max_send_sge = qp->sq.max_sg; + + qp->sq.wqe_size = roundup_pow_of_two(sizeof(struct pvrdma_sq_wqe_hdr) + + sizeof(struct pvrdma_sge) * + qp->sq.max_sg); + /* Note: one extra page for the header. */ + qp->npages_send = 1 + (qp->sq.wqe_cnt * qp->sq.wqe_size + + PAGE_SIZE - 1) / PAGE_SIZE; + + return 0; +} + +/** + * pvrdma_create_qp - create queue pair + * @pd: protection domain + * @init_attr: queue pair attributes + * @udata: user data + * + * @return: the ib_qp pointer on success, otherwise returns an errno. + */ +struct ib_qp *pvrdma_create_qp(struct ib_pd *pd, + struct ib_qp_init_attr *init_attr, + struct ib_udata *udata) +{ + struct pvrdma_qp *qp = NULL; + struct pvrdma_dev *dev = to_vdev(pd->device); + union pvrdma_cmd_req req; + union pvrdma_cmd_resp rsp; + struct pvrdma_cmd_create_qp *cmd = &req.create_qp; + struct pvrdma_cmd_create_qp_resp *resp = &rsp.create_qp_resp; + struct pvrdma_create_qp ucmd; + unsigned long flags; + int ret; + + if (init_attr->create_flags) { + dev_warn(&dev->pdev->dev, + "invalid create queuepair flags %#x\n", + init_attr->create_flags); + return ERR_PTR(-EINVAL); + } + + if (init_attr->qp_type != IB_QPT_RC && + init_attr->qp_type != IB_QPT_UD && + init_attr->qp_type != IB_QPT_GSI) { + dev_warn(&dev->pdev->dev, "queuepair type %d not supported\n", + init_attr->qp_type); + return ERR_PTR(-EINVAL); + } + + if (!atomic_add_unless(&dev->num_qps, 1, dev->dsr->caps.max_qp)) + return ERR_PTR(-ENOMEM); + + switch (init_attr->qp_type) { + case IB_QPT_GSI: + if (init_attr->port_num == 0 || + init_attr->port_num > pd->device->phys_port_cnt || + udata) { + dev_warn(&dev->pdev->dev, "invalid queuepair attrs\n"); + ret = -EINVAL; + goto err_qp; + } + /* fall through */ + case IB_QPT_RC: + case IB_QPT_UD: + qp = kzalloc(sizeof(*qp), GFP_KERNEL); + if (!qp) { + ret = -ENOMEM; + goto err_qp; + } + + spin_lock_init(&qp->sq.lock); + spin_lock_init(&qp->rq.lock); + mutex_init(&qp->mutex); + atomic_set(&qp->refcnt, 1); + init_waitqueue_head(&qp->wait); + + qp->state = IB_QPS_RESET; + + if (pd->uobject && udata) { + dev_dbg(&dev->pdev->dev, + "create queuepair from user space\n"); + + if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) { + ret = -EFAULT; + goto err_qp; + } + + /* set qp->sq.wqe_cnt, shift, buf_size.. */ + qp->rumem = ib_umem_get(pd->uobject->context, + ucmd.rbuf_addr, + ucmd.rbuf_size, 0, 0); + if (IS_ERR(qp->rumem)) { + ret = PTR_ERR(qp->rumem); + goto err_qp; + } + + qp->sumem = ib_umem_get(pd->uobject->context, + ucmd.sbuf_addr, + ucmd.sbuf_size, 0, 0); + if (IS_ERR(qp->sumem)) { + ib_umem_release(qp->rumem); + ret = PTR_ERR(qp->sumem); + goto err_qp; + } + + qp->npages_send = ib_umem_page_count(qp->sumem); + qp->npages_recv = ib_umem_page_count(qp->rumem); + qp->npages = qp->npages_send + qp->npages_recv; + } else { + qp->is_kernel = true; + + ret = pvrdma_set_sq_size(to_vdev(pd->device), + &init_attr->cap, + init_attr->qp_type, qp); + if (ret) + goto err_qp; + + ret = pvrdma_set_rq_size(to_vdev(pd->device), + &init_attr->cap, qp); + if (ret) + goto err_qp; + + qp->npages = qp->npages_send + qp->npages_recv; + + /* Skip header page. */ + qp->sq.offset = PAGE_SIZE; + + /* Recv queue pages are after send pages. */ + qp->rq.offset = qp->npages_send * PAGE_SIZE; + } + + if (qp->npages < 0 || qp->npages > PVRDMA_PAGE_DIR_MAX_PAGES) { + dev_warn(&dev->pdev->dev, + "overflow pages in queuepair\n"); + ret = -EINVAL; + goto err_umem; + } + + ret = pvrdma_page_dir_init(dev, &qp->pdir, qp->npages, + qp->is_kernel); + if (ret) { + dev_warn(&dev->pdev->dev, + "could not allocate page directory\n"); + goto err_umem; + } + + if (!qp->is_kernel) { + pvrdma_page_dir_insert_umem(&qp->pdir, qp->sumem, 0); + pvrdma_page_dir_insert_umem(&qp->pdir, qp->rumem, + qp->npages_send); + } else { + /* Ring state is always the first page. */ + qp->sq.ring = qp->pdir.pages[0]; + qp->rq.ring = &qp->sq.ring[1]; + } + break; + default: + ret = -EINVAL; + goto err_qp; + } + + /* Not supported */ + init_attr->cap.max_inline_data = 0; + + memset(cmd, 0, sizeof(*cmd)); + cmd->hdr.cmd = PVRDMA_CMD_CREATE_QP; + cmd->pd_handle = to_vpd(pd)->pd_handle; + cmd->send_cq_handle = to_vcq(init_attr->send_cq)->cq_handle; + cmd->recv_cq_handle = to_vcq(init_attr->recv_cq)->cq_handle; + cmd->max_send_wr = init_attr->cap.max_send_wr; + cmd->max_recv_wr = init_attr->cap.max_recv_wr; + cmd->max_send_sge = init_attr->cap.max_send_sge; + cmd->max_recv_sge = init_attr->cap.max_recv_sge; + cmd->max_inline_data = init_attr->cap.max_inline_data; + cmd->sq_sig_all = (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR) ? 1 : 0; + cmd->qp_type = ib_qp_type_to_pvrdma(init_attr->qp_type); + cmd->access_flags = IB_ACCESS_LOCAL_WRITE; + cmd->total_chunks = qp->npages; + cmd->send_chunks = qp->npages_send - 1; + cmd->pdir_dma = qp->pdir.dir_dma; + + dev_dbg(&dev->pdev->dev, "create queuepair with %d, %d, %d, %d\n", + cmd->max_send_wr, cmd->max_recv_wr, cmd->max_send_sge, + cmd->max_recv_sge); + + ret = pvrdma_cmd_post(dev, &req, &rsp, PVRDMA_CMD_CREATE_QP_RESP); + if (ret < 0) { + dev_warn(&dev->pdev->dev, + "could not create queuepair, error: %d\n", ret); + goto err_pdir; + } + + /* max_send_wr/_recv_wr/_send_sge/_recv_sge/_inline_data */ + qp->qp_handle = resp->qpn; + qp->port = init_attr->port_num; + qp->ibqp.qp_num = resp->qpn; + spin_lock_irqsave(&dev->qp_tbl_lock, flags); + dev->qp_tbl[qp->qp_handle % dev->dsr->caps.max_qp] = qp; + spin_unlock_irqrestore(&dev->qp_tbl_lock, flags); + + return &qp->ibqp; + +err_pdir: + pvrdma_page_dir_cleanup(dev, &qp->pdir); +err_umem: + if (pd->uobject && udata) { + if (qp->rumem) + ib_umem_release(qp->rumem); + if (qp->sumem) + ib_umem_release(qp->sumem); + } +err_qp: + kfree(qp); + atomic_dec(&dev->num_qps); + + return ERR_PTR(ret); +} + +static void pvrdma_free_qp(struct pvrdma_qp *qp) +{ + struct pvrdma_dev *dev = to_vdev(qp->ibqp.device); + struct pvrdma_cq *scq; + struct pvrdma_cq *rcq; + unsigned long flags, scq_flags, rcq_flags; + + /* In case cq is polling */ + get_cqs(qp, &scq, &rcq); + pvrdma_lock_cqs(scq, rcq, &scq_flags, &rcq_flags); + + _pvrdma_flush_cqe(qp, scq); + if (scq != rcq) + _pvrdma_flush_cqe(qp, rcq); + + spin_lock_irqsave(&dev->qp_tbl_lock, flags); + dev->qp_tbl[qp->qp_handle] = NULL; + spin_unlock_irqrestore(&dev->qp_tbl_lock, flags); + + pvrdma_unlock_cqs(scq, rcq, &scq_flags, &rcq_flags); + + atomic_dec(&qp->refcnt); + wait_event(qp->wait, !atomic_read(&qp->refcnt)); + + pvrdma_page_dir_cleanup(dev, &qp->pdir); + + kfree(qp); + + atomic_dec(&dev->num_qps); +} + +/** + * pvrdma_destroy_qp - destroy a queue pair + * @qp: the queue pair to destroy + * + * @return: 0 on success. + */ +int pvrdma_destroy_qp(struct ib_qp *qp) +{ + struct pvrdma_qp *vqp = to_vqp(qp); + union pvrdma_cmd_req req; + struct pvrdma_cmd_destroy_qp *cmd = &req.destroy_qp; + int ret; + + memset(cmd, 0, sizeof(*cmd)); + cmd->hdr.cmd = PVRDMA_CMD_DESTROY_QP; + cmd->qp_handle = vqp->qp_handle; + + ret = pvrdma_cmd_post(to_vdev(qp->device), &req, NULL, 0); + if (ret < 0) + dev_warn(&to_vdev(qp->device)->pdev->dev, + "destroy queuepair failed, error: %d\n", ret); + + pvrdma_free_qp(vqp); + + return 0; +} + +/** + * pvrdma_modify_qp - modify queue pair attributes + * @ibqp: the queue pair + * @attr: the new queue pair's attributes + * @attr_mask: attributes mask + * @udata: user data + * + * @returns 0 on success, otherwise returns an errno. + */ +int pvrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, + int attr_mask, struct ib_udata *udata) +{ + struct pvrdma_dev *dev = to_vdev(ibqp->device); + struct pvrdma_qp *qp = to_vqp(ibqp); + union pvrdma_cmd_req req; + union pvrdma_cmd_resp rsp; + struct pvrdma_cmd_modify_qp *cmd = &req.modify_qp; + int cur_state, next_state; + int ret; + + /* Sanity checking. Should need lock here */ + mutex_lock(&qp->mutex); + cur_state = (attr_mask & IB_QP_CUR_STATE) ? attr->cur_qp_state : + qp->state; + next_state = (attr_mask & IB_QP_STATE) ? attr->qp_state : cur_state; + + if (!ib_modify_qp_is_ok(cur_state, next_state, ibqp->qp_type, + attr_mask, IB_LINK_LAYER_ETHERNET)) { + ret = -EINVAL; + goto out; + } + + if (attr_mask & IB_QP_PORT) { + if (attr->port_num == 0 || + attr->port_num > ibqp->device->phys_port_cnt) { + ret = -EINVAL; + goto out; + } + } + + if (attr_mask & IB_QP_MIN_RNR_TIMER) { + if (attr->min_rnr_timer > 31) { + ret = -EINVAL; + goto out; + } + } + + if (attr_mask & IB_QP_PKEY_INDEX) { + if (attr->pkey_index >= dev->dsr->caps.max_pkeys) { + ret = -EINVAL; + goto out; + } + } + + if (attr_mask & IB_QP_QKEY) + qp->qkey = attr->qkey; + + if (cur_state == next_state && cur_state == IB_QPS_RESET) { + ret = 0; + goto out; + } + + qp->state = next_state; + memset(cmd, 0, sizeof(*cmd)); + cmd->hdr.cmd = PVRDMA_CMD_MODIFY_QP; + cmd->qp_handle = qp->qp_handle; + cmd->attr_mask = ib_qp_attr_mask_to_pvrdma(attr_mask); + cmd->attrs.qp_state = ib_qp_state_to_pvrdma(attr->qp_state); + cmd->attrs.cur_qp_state = + ib_qp_state_to_pvrdma(attr->cur_qp_state); + cmd->attrs.path_mtu = ib_mtu_to_pvrdma(attr->path_mtu); + cmd->attrs.path_mig_state = + ib_mig_state_to_pvrdma(attr->path_mig_state); + cmd->attrs.qkey = attr->qkey; + cmd->attrs.rq_psn = attr->rq_psn; + cmd->attrs.sq_psn = attr->sq_psn; + cmd->attrs.dest_qp_num = attr->dest_qp_num; + cmd->attrs.qp_access_flags = + ib_access_flags_to_pvrdma(attr->qp_access_flags); + cmd->attrs.pkey_index = attr->pkey_index; + cmd->attrs.alt_pkey_index = attr->alt_pkey_index; + cmd->attrs.en_sqd_async_notify = attr->en_sqd_async_notify; + cmd->attrs.sq_draining = attr->sq_draining; + cmd->attrs.max_rd_atomic = attr->max_rd_atomic; + cmd->attrs.max_dest_rd_atomic = attr->max_dest_rd_atomic; + cmd->attrs.min_rnr_timer = attr->min_rnr_timer; + cmd->attrs.port_num = attr->port_num; + cmd->attrs.timeout = attr->timeout; + cmd->attrs.retry_cnt = attr->retry_cnt; + cmd->attrs.rnr_retry = attr->rnr_retry; + cmd->attrs.alt_port_num = attr->alt_port_num; + cmd->attrs.alt_timeout = attr->alt_timeout; + ib_qp_cap_to_pvrdma(&cmd->attrs.cap, &attr->cap); + ib_ah_attr_to_pvrdma(&cmd->attrs.ah_attr, &attr->ah_attr); + ib_ah_attr_to_pvrdma(&cmd->attrs.alt_ah_attr, &attr->alt_ah_attr); + + ret = pvrdma_cmd_post(dev, &req, &rsp, PVRDMA_CMD_MODIFY_QP_RESP); + if (ret < 0) { + dev_warn(&dev->pdev->dev, + "could not modify queuepair, error: %d\n", ret); + } else if (rsp.hdr.err > 0) { + dev_warn(&dev->pdev->dev, + "cannot modify queuepair, error: %d\n", rsp.hdr.err); + ret = -EINVAL; + } + + if (ret == 0 && next_state == IB_QPS_RESET) + pvrdma_reset_qp(qp); + +out: + mutex_unlock(&qp->mutex); + + return ret; +} + +static inline void *get_sq_wqe(struct pvrdma_qp *qp, int n) +{ + return pvrdma_page_dir_get_ptr(&qp->pdir, + qp->sq.offset + n * qp->sq.wqe_size); +} + +static inline void *get_rq_wqe(struct pvrdma_qp *qp, int n) +{ + return pvrdma_page_dir_get_ptr(&qp->pdir, + qp->rq.offset + n * qp->rq.wqe_size); +} + +static int set_reg_seg(struct pvrdma_sq_wqe_hdr *wqe_hdr, struct ib_reg_wr *wr) +{ + struct pvrdma_user_mr *mr = to_vmr(wr->mr); + + wqe_hdr->wr.fast_reg.iova_start = mr->ibmr.iova; + wqe_hdr->wr.fast_reg.pl_pdir_dma = mr->pdir.dir_dma; + wqe_hdr->wr.fast_reg.page_shift = mr->page_shift; + wqe_hdr->wr.fast_reg.page_list_len = mr->npages; + wqe_hdr->wr.fast_reg.length = mr->ibmr.length; + wqe_hdr->wr.fast_reg.access_flags = wr->access; + wqe_hdr->wr.fast_reg.rkey = wr->key; + + return pvrdma_page_dir_insert_page_list(&mr->pdir, mr->pages, + mr->npages); +} + +/** + * pvrdma_post_send - post send work request entries on a QP + * @ibqp: the QP + * @wr: work request list to post + * @bad_wr: the first bad WR returned + * + * @return: 0 on success, otherwise errno returned. + */ +int pvrdma_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, + struct ib_send_wr **bad_wr) +{ + struct pvrdma_qp *qp = to_vqp(ibqp); + struct pvrdma_dev *dev = to_vdev(ibqp->device); + unsigned long flags; + struct pvrdma_sq_wqe_hdr *wqe_hdr; + struct pvrdma_sge *sge; + int i, index; + int nreq; + int ret; + + /* + * In states lower than RTS, we can fail immediately. In other states, + * just post and let the device figure it out. + */ + if (qp->state < IB_QPS_RTS) { + *bad_wr = wr; + return -EINVAL; + } + + spin_lock_irqsave(&qp->sq.lock, flags); + + index = pvrdma_idx(&qp->sq.ring->prod_tail, qp->sq.wqe_cnt); + for (nreq = 0; wr; nreq++, wr = wr->next) { + unsigned int tail; + + if (unlikely(!pvrdma_idx_ring_has_space( + qp->sq.ring, qp->sq.wqe_cnt, &tail))) { + dev_warn_ratelimited(&dev->pdev->dev, + "send queue is full\n"); + *bad_wr = wr; + ret = -ENOMEM; + goto out; + } + + if (unlikely(wr->num_sge > qp->sq.max_sg || wr->num_sge < 0)) { + dev_warn_ratelimited(&dev->pdev->dev, + "send SGE overflow\n"); + *bad_wr = wr; + ret = -EINVAL; + goto out; + } + + if (unlikely(wr->opcode < 0)) { + dev_warn_ratelimited(&dev->pdev->dev, + "invalid send opcode\n"); + *bad_wr = wr; + ret = -EINVAL; + goto out; + } + + /* + * Only support UD, RC. + * Need to check opcode table for thorough checking. + * opcode _UD _UC _RC + * _SEND x x x + * _SEND_WITH_IMM x x x + * _RDMA_WRITE x x + * _RDMA_WRITE_WITH_IMM x x + * _LOCAL_INV x x + * _SEND_WITH_INV x x + * _RDMA_READ x + * _ATOMIC_CMP_AND_SWP x + * _ATOMIC_FETCH_AND_ADD x + * _MASK_ATOMIC_CMP_AND_SWP x + * _MASK_ATOMIC_FETCH_AND_ADD x + * _REG_MR x + * + */ + if (qp->ibqp.qp_type != IB_QPT_UD && + qp->ibqp.qp_type != IB_QPT_RC && + wr->opcode != IB_WR_SEND) { + dev_warn_ratelimited(&dev->pdev->dev, + "unsupported queuepair type\n"); + *bad_wr = wr; + ret = -EINVAL; + goto out; + } else if (qp->ibqp.qp_type == IB_QPT_UD || + qp->ibqp.qp_type == IB_QPT_GSI) { + if (wr->opcode != IB_WR_SEND && + wr->opcode != IB_WR_SEND_WITH_IMM) { + dev_warn_ratelimited(&dev->pdev->dev, + "invalid send opcode\n"); + *bad_wr = wr; + ret = -EINVAL; + goto out; + } + } + + wqe_hdr = (struct pvrdma_sq_wqe_hdr *)get_sq_wqe(qp, index); + memset(wqe_hdr, 0, sizeof(*wqe_hdr)); + wqe_hdr->wr_id = wr->wr_id; + wqe_hdr->num_sge = wr->num_sge; + wqe_hdr->opcode = ib_wr_opcode_to_pvrdma(wr->opcode); + wqe_hdr->send_flags = ib_send_flags_to_pvrdma(wr->send_flags); + if (wr->opcode == IB_WR_SEND_WITH_IMM || + wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM) + wqe_hdr->ex.imm_data = wr->ex.imm_data; + + switch (qp->ibqp.qp_type) { + case IB_QPT_GSI: + case IB_QPT_UD: + if (unlikely(!ud_wr(wr)->ah)) { + dev_warn_ratelimited(&dev->pdev->dev, + "invalid address handle\n"); + *bad_wr = wr; + ret = -EINVAL; + goto out; + } + + /* + * Use qkey from qp context if high order bit set, + * otherwise from work request. + */ + wqe_hdr->wr.ud.remote_qpn = ud_wr(wr)->remote_qpn; + wqe_hdr->wr.ud.remote_qkey = + ud_wr(wr)->remote_qkey & 0x80000000 ? + qp->qkey : ud_wr(wr)->remote_qkey; + wqe_hdr->wr.ud.av = to_vah(ud_wr(wr)->ah)->av; + + break; + case IB_QPT_RC: + switch (wr->opcode) { + case IB_WR_RDMA_READ: + case IB_WR_RDMA_WRITE: + case IB_WR_RDMA_WRITE_WITH_IMM: + wqe_hdr->wr.rdma.remote_addr = + rdma_wr(wr)->remote_addr; + wqe_hdr->wr.rdma.rkey = rdma_wr(wr)->rkey; + break; + case IB_WR_LOCAL_INV: + case IB_WR_SEND_WITH_INV: + wqe_hdr->ex.invalidate_rkey = + wr->ex.invalidate_rkey; + break; + case IB_WR_ATOMIC_CMP_AND_SWP: + case IB_WR_ATOMIC_FETCH_AND_ADD: + wqe_hdr->wr.atomic.remote_addr = + atomic_wr(wr)->remote_addr; + wqe_hdr->wr.atomic.rkey = atomic_wr(wr)->rkey; + wqe_hdr->wr.atomic.compare_add = + atomic_wr(wr)->compare_add; + if (wr->opcode == IB_WR_ATOMIC_CMP_AND_SWP) + wqe_hdr->wr.atomic.swap = + atomic_wr(wr)->swap; + break; + case IB_WR_REG_MR: + ret = set_reg_seg(wqe_hdr, reg_wr(wr)); + if (ret < 0) { + dev_warn_ratelimited(&dev->pdev->dev, + "Failed to set fast register work request\n"); + *bad_wr = wr; + goto out; + } + break; + default: + break; + } + + break; + default: + dev_warn_ratelimited(&dev->pdev->dev, + "invalid queuepair type\n"); + ret = -EINVAL; + *bad_wr = wr; + goto out; + } + + sge = (struct pvrdma_sge *)(wqe_hdr + 1); + for (i = 0; i < wr->num_sge; i++) { + /* Need to check wqe_size 0 or max size */ + sge->addr = wr->sg_list[i].addr; + sge->length = wr->sg_list[i].length; + sge->lkey = wr->sg_list[i].lkey; + sge++; + } + + /* Make sure wqe is written before index update */ + smp_wmb(); + + index++; + if (unlikely(index >= qp->sq.wqe_cnt)) + index = 0; + /* Update shared sq ring */ + pvrdma_idx_ring_inc(&qp->sq.ring->prod_tail, + qp->sq.wqe_cnt); + } + + ret = 0; + +out: + spin_unlock_irqrestore(&qp->sq.lock, flags); + + if (!ret) + pvrdma_write_uar_qp(dev, PVRDMA_UAR_QP_SEND | qp->qp_handle); + + return ret; +} + +/** + * pvrdma_post_receive - post receive work request entries on a QP + * @ibqp: the QP + * @wr: the work request list to post + * @bad_wr: the first bad WR returned + * + * @return: 0 on success, otherwise errno returned. + */ +int pvrdma_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr, + struct ib_recv_wr **bad_wr) +{ + struct pvrdma_dev *dev = to_vdev(ibqp->device); + unsigned long flags; + struct pvrdma_qp *qp = to_vqp(ibqp); + struct pvrdma_rq_wqe_hdr *wqe_hdr; + struct pvrdma_sge *sge; + int index, nreq; + int ret = 0; + int i; + + /* + * In the RESET state, we can fail immediately. For other states, + * just post and let the device figure it out. + */ + if (qp->state == IB_QPS_RESET) { + *bad_wr = wr; + return -EINVAL; + } + + spin_lock_irqsave(&qp->rq.lock, flags); + + index = pvrdma_idx(&qp->rq.ring->prod_tail, qp->rq.wqe_cnt); + for (nreq = 0; wr; nreq++, wr = wr->next) { + unsigned int tail; + + if (unlikely(wr->num_sge > qp->rq.max_sg || + wr->num_sge < 0)) { + ret = -EINVAL; + *bad_wr = wr; + dev_warn_ratelimited(&dev->pdev->dev, + "recv SGE overflow\n"); + goto out; + } + + if (unlikely(!pvrdma_idx_ring_has_space( + qp->rq.ring, qp->rq.wqe_cnt, &tail))) { + ret = -ENOMEM; + *bad_wr = wr; + dev_warn_ratelimited(&dev->pdev->dev, + "recv queue full\n"); + goto out; + } + + wqe_hdr = (struct pvrdma_rq_wqe_hdr *)get_rq_wqe(qp, index); + wqe_hdr->wr_id = wr->wr_id; + wqe_hdr->num_sge = wr->num_sge; + wqe_hdr->total_len = 0; + + sge = (struct pvrdma_sge *)(wqe_hdr + 1); + for (i = 0; i < wr->num_sge; i++) { + sge->addr = wr->sg_list[i].addr; + sge->length = wr->sg_list[i].length; + sge->lkey = wr->sg_list[i].lkey; + sge++; + } + + /* Make sure wqe is written before index update */ + smp_wmb(); + + index++; + if (unlikely(index >= qp->rq.wqe_cnt)) + index = 0; + /* Update shared rq ring */ + pvrdma_idx_ring_inc(&qp->rq.ring->prod_tail, + qp->rq.wqe_cnt); + } + + spin_unlock_irqrestore(&qp->rq.lock, flags); + + pvrdma_write_uar_qp(dev, PVRDMA_UAR_QP_RECV | qp->qp_handle); + + return ret; + +out: + spin_unlock_irqrestore(&qp->rq.lock, flags); + + return ret; +} + +/** + * pvrdma_query_qp - query a queue pair's attributes + * @ibqp: the queue pair to query + * @attr: the queue pair's attributes + * @attr_mask: attributes mask + * @init_attr: initial queue pair attributes + * + * @returns 0 on success, otherwise returns an errno. + */ +int pvrdma_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, + int attr_mask, struct ib_qp_init_attr *init_attr) +{ + struct pvrdma_dev *dev = to_vdev(ibqp->device); + struct pvrdma_qp *qp = to_vqp(ibqp); + union pvrdma_cmd_req req; + union pvrdma_cmd_resp rsp; + struct pvrdma_cmd_query_qp *cmd = &req.query_qp; + struct pvrdma_cmd_query_qp_resp *resp = &rsp.query_qp_resp; + int ret = 0; + + mutex_lock(&qp->mutex); + + if (qp->state == IB_QPS_RESET) { + attr->qp_state = IB_QPS_RESET; + goto out; + } + + memset(cmd, 0, sizeof(*cmd)); + cmd->hdr.cmd = PVRDMA_CMD_QUERY_QP; + cmd->qp_handle = qp->qp_handle; + cmd->attr_mask = ib_qp_attr_mask_to_pvrdma(attr_mask); + + ret = pvrdma_cmd_post(dev, &req, &rsp, PVRDMA_CMD_QUERY_QP_RESP); + if (ret < 0) { + dev_warn(&dev->pdev->dev, + "could not query queuepair, error: %d\n", ret); + goto out; + } + + attr->qp_state = pvrdma_qp_state_to_ib(resp->attrs.qp_state); + attr->cur_qp_state = + pvrdma_qp_state_to_ib(resp->attrs.cur_qp_state); + attr->path_mtu = pvrdma_mtu_to_ib(resp->attrs.path_mtu); + attr->path_mig_state = + pvrdma_mig_state_to_ib(resp->attrs.path_mig_state); + attr->qkey = resp->attrs.qkey; + attr->rq_psn = resp->attrs.rq_psn; + attr->sq_psn = resp->attrs.sq_psn; + attr->dest_qp_num = resp->attrs.dest_qp_num; + attr->qp_access_flags = + pvrdma_access_flags_to_ib(resp->attrs.qp_access_flags); + attr->pkey_index = resp->attrs.pkey_index; + attr->alt_pkey_index = resp->attrs.alt_pkey_index; + attr->en_sqd_async_notify = resp->attrs.en_sqd_async_notify; + attr->sq_draining = resp->attrs.sq_draining; + attr->max_rd_atomic = resp->attrs.max_rd_atomic; + attr->max_dest_rd_atomic = resp->attrs.max_dest_rd_atomic; + attr->min_rnr_timer = resp->attrs.min_rnr_timer; + attr->port_num = resp->attrs.port_num; + attr->timeout = resp->attrs.timeout; + attr->retry_cnt = resp->attrs.retry_cnt; + attr->rnr_retry = resp->attrs.rnr_retry; + attr->alt_port_num = resp->attrs.alt_port_num; + attr->alt_timeout = resp->attrs.alt_timeout; + pvrdma_qp_cap_to_ib(&attr->cap, &resp->attrs.cap); + pvrdma_ah_attr_to_ib(&attr->ah_attr, &resp->attrs.ah_attr); + pvrdma_ah_attr_to_ib(&attr->alt_ah_attr, &resp->attrs.alt_ah_attr); + + qp->state = attr->qp_state; + + ret = 0; + +out: + attr->cur_qp_state = attr->qp_state; + + init_attr->event_handler = qp->ibqp.event_handler; + init_attr->qp_context = qp->ibqp.qp_context; + init_attr->send_cq = qp->ibqp.send_cq; + init_attr->recv_cq = qp->ibqp.recv_cq; + init_attr->srq = qp->ibqp.srq; + init_attr->xrcd = NULL; + init_attr->cap = attr->cap; + init_attr->sq_sig_type = 0; + init_attr->qp_type = qp->ibqp.qp_type; + init_attr->create_flags = 0; + init_attr->port_num = qp->port; + + mutex_unlock(&qp->mutex); + return ret; +}