From patchwork Tue Feb 28 00:58:36 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chris Leech X-Patchwork-Id: 9594515 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id 7EE8560453 for ; Tue, 28 Feb 2017 01:01:14 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 6722B284CB for ; Tue, 28 Feb 2017 01:01:14 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 5BB55284D1; Tue, 28 Feb 2017 01:01:14 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.9 required=2.0 tests=BAYES_00,RCVD_IN_DNSWL_HI autolearn=ham version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id BCB14284CB for ; Tue, 28 Feb 2017 01:01:13 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751736AbdB1BBL (ORCPT ); Mon, 27 Feb 2017 20:01:11 -0500 Received: from mx1.redhat.com ([209.132.183.28]:46600 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751372AbdB1BBJ (ORCPT ); Mon, 27 Feb 2017 20:01:09 -0500 Received: from int-mx13.intmail.prod.int.phx2.redhat.com (int-mx13.intmail.prod.int.phx2.redhat.com [10.5.11.26]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 330EE6AAC2; Tue, 28 Feb 2017 01:00:29 +0000 (UTC) Received: from straylight.hirudinean.org.com (ovpn-120-182.rdu2.redhat.com [10.10.120.182] (may be forged)) by int-mx13.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id v1S10QPq009176; Mon, 27 Feb 2017 20:00:27 -0500 From: Chris Leech To: linux-scsi@vger.kernel.org, open-iscsi@googlegroups.com, martin.petersen@oracle.com, jejb@linux.vnet.ibm.com, lduncan@suse.com Cc: Chris Leech , ilkka@kapsi.fi, gpiccoli@linux.vnet.ibm.com, "# v3 . 15+" Subject: [PATCH v2] libiscsi: add lock around task lists to fix list corruption regression Date: Mon, 27 Feb 2017 16:58:36 -0800 Message-Id: <20170228005836.12444-1-cleech@redhat.com> In-Reply-To: <20170221173536.32024-1-cleech@redhat.com> References: <20170221173536.32024-1-cleech@redhat.com> X-Scanned-By: MIMEDefang 2.68 on 10.5.11.26 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.26]); Tue, 28 Feb 2017 01:00:29 +0000 (UTC) Sender: linux-scsi-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-scsi@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP There's a rather long standing regression from the commit "libiscsi: Reduce locking contention in fast path" Depending on iSCSI target behavior, it's possible to hit the case in iscsi_complete_task where the task is still on a pending list (!list_empty(&task->running)). When that happens the task is removed from the list while holding the session back_lock, but other task list modification occur under the frwd_lock. That leads to linked list corruption and eventually a panicked system. Rather than back out the session lock split entirely, in order to try and keep some of the performance gains this patch adds another lock to maintain the task lists integrity. Major enterprise supported kernels have been backing out the lock split for while now, thanks to the efforts at IBM where a lab setup has the most reliable reproducer I've seen on this issue. This patch has been tested there successfully. v2: changed WARN_ONCE to pr_debug_once Signed-off-by: Chris Leech Fixes: 659743b02c41 ("[SCSI] libiscsi: Reduce locking contention in fast path") Reported-by: Prashantha Subbarao Reviewed-by: Guilherme G. Piccoli Cc: # v3.15+ --- drivers/scsi/libiscsi.c | 26 +++++++++++++++++++++++++- include/scsi/libiscsi.h | 1 + 2 files changed, 26 insertions(+), 1 deletion(-) diff --git a/drivers/scsi/libiscsi.c b/drivers/scsi/libiscsi.c index 834d1212b6d5..3fca34a675af 100644 --- a/drivers/scsi/libiscsi.c +++ b/drivers/scsi/libiscsi.c @@ -560,8 +560,12 @@ static void iscsi_complete_task(struct iscsi_task *task, int state) WARN_ON_ONCE(task->state == ISCSI_TASK_FREE); task->state = state; - if (!list_empty(&task->running)) + spin_lock_bh(&conn->taskqueuelock); + if (!list_empty(&task->running)) { + pr_debug_once("%s while task on list", __func__); list_del_init(&task->running); + } + spin_unlock_bh(&conn->taskqueuelock); if (conn->task == task) conn->task = NULL; @@ -783,7 +787,9 @@ __iscsi_conn_send_pdu(struct iscsi_conn *conn, struct iscsi_hdr *hdr, if (session->tt->xmit_task(task)) goto free_task; } else { + spin_lock_bh(&conn->taskqueuelock); list_add_tail(&task->running, &conn->mgmtqueue); + spin_unlock_bh(&conn->taskqueuelock); iscsi_conn_queue_work(conn); } @@ -1474,8 +1480,10 @@ void iscsi_requeue_task(struct iscsi_task *task) * this may be on the requeue list already if the xmit_task callout * is handling the r2ts while we are adding new ones */ + spin_lock_bh(&conn->taskqueuelock); if (list_empty(&task->running)) list_add_tail(&task->running, &conn->requeue); + spin_unlock_bh(&conn->taskqueuelock); iscsi_conn_queue_work(conn); } EXPORT_SYMBOL_GPL(iscsi_requeue_task); @@ -1512,22 +1520,26 @@ static int iscsi_data_xmit(struct iscsi_conn *conn) * only have one nop-out as a ping from us and targets should not * overflow us with nop-ins */ + spin_lock_bh(&conn->taskqueuelock); check_mgmt: while (!list_empty(&conn->mgmtqueue)) { conn->task = list_entry(conn->mgmtqueue.next, struct iscsi_task, running); list_del_init(&conn->task->running); + spin_unlock_bh(&conn->taskqueuelock); if (iscsi_prep_mgmt_task(conn, conn->task)) { /* regular RX path uses back_lock */ spin_lock_bh(&conn->session->back_lock); __iscsi_put_task(conn->task); spin_unlock_bh(&conn->session->back_lock); conn->task = NULL; + spin_lock_bh(&conn->taskqueuelock); continue; } rc = iscsi_xmit_task(conn); if (rc) goto done; + spin_lock_bh(&conn->taskqueuelock); } /* process pending command queue */ @@ -1535,19 +1547,24 @@ static int iscsi_data_xmit(struct iscsi_conn *conn) conn->task = list_entry(conn->cmdqueue.next, struct iscsi_task, running); list_del_init(&conn->task->running); + spin_unlock_bh(&conn->taskqueuelock); if (conn->session->state == ISCSI_STATE_LOGGING_OUT) { fail_scsi_task(conn->task, DID_IMM_RETRY); + spin_lock_bh(&conn->taskqueuelock); continue; } rc = iscsi_prep_scsi_cmd_pdu(conn->task); if (rc) { if (rc == -ENOMEM || rc == -EACCES) { + spin_lock_bh(&conn->taskqueuelock); list_add_tail(&conn->task->running, &conn->cmdqueue); conn->task = NULL; + spin_unlock_bh(&conn->taskqueuelock); goto done; } else fail_scsi_task(conn->task, DID_ABORT); + spin_lock_bh(&conn->taskqueuelock); continue; } rc = iscsi_xmit_task(conn); @@ -1558,6 +1575,7 @@ static int iscsi_data_xmit(struct iscsi_conn *conn) * we need to check the mgmt queue for nops that need to * be sent to aviod starvation */ + spin_lock_bh(&conn->taskqueuelock); if (!list_empty(&conn->mgmtqueue)) goto check_mgmt; } @@ -1577,12 +1595,15 @@ static int iscsi_data_xmit(struct iscsi_conn *conn) conn->task = task; list_del_init(&conn->task->running); conn->task->state = ISCSI_TASK_RUNNING; + spin_unlock_bh(&conn->taskqueuelock); rc = iscsi_xmit_task(conn); if (rc) goto done; + spin_lock_bh(&conn->taskqueuelock); if (!list_empty(&conn->mgmtqueue)) goto check_mgmt; } + spin_unlock_bh(&conn->taskqueuelock); spin_unlock_bh(&conn->session->frwd_lock); return -ENODATA; @@ -1738,7 +1759,9 @@ int iscsi_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *sc) goto prepd_reject; } } else { + spin_lock_bh(&conn->taskqueuelock); list_add_tail(&task->running, &conn->cmdqueue); + spin_unlock_bh(&conn->taskqueuelock); iscsi_conn_queue_work(conn); } @@ -2896,6 +2919,7 @@ iscsi_conn_setup(struct iscsi_cls_session *cls_session, int dd_size, INIT_LIST_HEAD(&conn->mgmtqueue); INIT_LIST_HEAD(&conn->cmdqueue); INIT_LIST_HEAD(&conn->requeue); + spin_lock_init(&conn->taskqueuelock); INIT_WORK(&conn->xmitwork, iscsi_xmitworker); /* allocate login_task used for the login/text sequences */ diff --git a/include/scsi/libiscsi.h b/include/scsi/libiscsi.h index b0e275de6dec..583875ea136a 100644 --- a/include/scsi/libiscsi.h +++ b/include/scsi/libiscsi.h @@ -196,6 +196,7 @@ struct iscsi_conn { struct iscsi_task *task; /* xmit task in progress */ /* xmit */ + spinlock_t taskqueuelock; /* protects the next three lists */ struct list_head mgmtqueue; /* mgmt (control) xmit queue */ struct list_head cmdqueue; /* data-path cmd queue */ struct list_head requeue; /* tasks needing another run */