From patchwork Wed Oct 18 07:14:02 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Mike Christie X-Patchwork-Id: 10013817 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 1516A60211 for ; Wed, 18 Oct 2017 07:14:28 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 07E38285DB for ; Wed, 18 Oct 2017 07:14:28 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id F0B0D285E5; Wed, 18 Oct 2017 07:14:27 +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 83682285DB for ; Wed, 18 Oct 2017 07:14:27 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1761875AbdJRHO0 (ORCPT ); Wed, 18 Oct 2017 03:14:26 -0400 Received: from mx1.redhat.com ([209.132.183.28]:49374 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758347AbdJRHOY (ORCPT ); Wed, 18 Oct 2017 03:14:24 -0400 Received: from smtp.corp.redhat.com (int-mx05.intmail.prod.int.phx2.redhat.com [10.5.11.15]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 30CB8883A2; Wed, 18 Oct 2017 07:14:24 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mx1.redhat.com 30CB8883A2 Authentication-Results: ext-mx02.extmail.prod.ext.phx2.redhat.com; dmarc=none (p=none dis=none) header.from=redhat.com Authentication-Results: ext-mx02.extmail.prod.ext.phx2.redhat.com; spf=fail smtp.mailfrom=mchristi@redhat.com Received: from rh2.redhat.com (ovpn-120-106.rdu2.redhat.com [10.10.120.106]) by smtp.corp.redhat.com (Postfix) with ESMTP id 745437BA36; Wed, 18 Oct 2017 07:14:23 +0000 (UTC) From: Mike Christie To: lixiubo@cmss.chinamobile.com, target-devel@vger.kernel.org, nab@linux-iscsi.org Cc: Mike Christie Subject: [PATCH 07/17] tcmu: fix dev root_udev_waiter_mutex errors Date: Wed, 18 Oct 2017 02:14:02 -0500 Message-Id: <1508310852-15366-8-git-send-email-mchristi@redhat.com> In-Reply-To: <1508310852-15366-1-git-send-email-mchristi@redhat.com> References: <1508310852-15366-1-git-send-email-mchristi@redhat.com> X-Scanned-By: MIMEDefang 2.79 on 10.5.11.15 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.26]); Wed, 18 Oct 2017 07:14:24 +0000 (UTC) Sender: target-devel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: target-devel@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP This fixes the following bugs: 1. You cannot sleep after calling prepare_to_wait, so you cannot use a mutex. 2. If the unmap thread is running and we have not done the prepare_to_wait when a wake_up(&unmap_wait) is done, then we will miss the wake up. We need to check if a waiter has been added after setting the process state and before calling schedule. Note: there is a similar race with tcmu_irqconrol. One of the next patches will fix that code path. Signed-off-by: Mike Christie --- drivers/target/target_core_user.c | 72 +++++++++++++++++++++++---------------- 1 file changed, 43 insertions(+), 29 deletions(-) diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c index 5ce0034..0c71572 100644 --- a/drivers/target/target_core_user.c +++ b/drivers/target/target_core_user.c @@ -197,17 +197,7 @@ static LIST_HEAD(timed_out_udevs); static DEFINE_MUTEX(root_udev_mutex); static LIST_HEAD(root_udev); -/* - * To avoid dead lock, the mutex locks order should always be: - * - * mutex_lock(&root_udev_waiter_mutex); - * ... - * mutex_lock(&tcmu_dev->cmdr_lock); - * mutex_unlock(&tcmu_dev->cmdr_lock); - * ... - * mutex_unlock(&root_udev_waiter_mutex); - */ -static DEFINE_MUTEX(root_udev_waiter_mutex); +static DEFINE_SPINLOCK(root_udev_waiter_lock); /* The data blocks global pool waiter list */ static LIST_HEAD(root_udev_waiter); @@ -860,19 +850,19 @@ tcmu_queue_cmd_ring(struct tcmu_cmd *tcmu_cmd) prepare_to_wait(&udev->wait_cmdr, &__wait, TASK_INTERRUPTIBLE); pr_debug("sleeping for ring space\n"); - mutex_unlock(&udev->cmdr_lock); - /* * Try to insert the current udev to waiter list and * then wake up the unmap thread */ if (is_waiting_blocks) { - mutex_lock(&root_udev_waiter_mutex); - list_add_tail(&udev->waiter, &root_udev_waiter); - mutex_unlock(&root_udev_waiter_mutex); + spin_lock(&root_udev_waiter_lock); + if (list_empty(&udev->waiter)) + list_add_tail(&udev->waiter, &root_udev_waiter); + spin_unlock(&root_udev_waiter_lock); wake_up(&unmap_wait); } + mutex_unlock(&udev->cmdr_lock); if (udev->cmd_time_out) ret = schedule_timeout( @@ -1170,6 +1160,7 @@ static struct se_device *tcmu_alloc_device(struct se_hba *hba, const char *name) mutex_init(&udev->cmdr_lock); INIT_LIST_HEAD(&udev->timedout_entry); + INIT_LIST_HEAD(&udev->waiter); idr_init(&udev->commands); setup_timer(&udev->timeout, tcmu_device_timedout, @@ -1694,12 +1685,10 @@ static void tcmu_destroy_device(struct se_device *dev) list_del(&udev->node); mutex_unlock(&root_udev_mutex); - mutex_lock(&root_udev_waiter_mutex); - mutex_lock(&udev->cmdr_lock); - if (is_in_waiter_list(udev)) + spin_lock(&root_udev_waiter_lock); + if (!list_empty(&udev->waiter)) list_del(&udev->waiter); - mutex_unlock(&udev->cmdr_lock); - mutex_unlock(&root_udev_waiter_mutex); + spin_unlock(&root_udev_waiter_lock); tcmu_netlink_event(udev, TCMU_CMD_REMOVED_DEVICE, 0, NULL); @@ -2100,29 +2089,43 @@ static uint32_t find_free_blocks(void) static void run_cmdr_queues(uint32_t *free_blocks) { - struct tcmu_dev *udev, *tmp; + struct tcmu_dev *udev, *tmp_udev; + LIST_HEAD(devs); /* * Try to wake up the udevs who are waiting * for the global data pool blocks. */ - mutex_lock(&root_udev_waiter_mutex); - list_for_each_entry_safe(udev, tmp, &root_udev_waiter, waiter) { + spin_lock(&root_udev_waiter_lock); + list_splice_init(&root_udev_waiter, &devs); + + list_for_each_entry_safe(udev, tmp_udev, &devs, waiter) { + list_del_init(&udev->waiter); + spin_unlock(&root_udev_waiter_lock); + mutex_lock(&udev->cmdr_lock); if (udev->waiting_blocks < *free_blocks) { + spin_lock(&root_udev_waiter_lock); + if (list_empty(&udev->waiter)) + list_add_tail(&udev->waiter, &root_udev_waiter); + spin_unlock(&root_udev_waiter_lock); + mutex_unlock(&udev->cmdr_lock); - break; + goto resplice; } *free_blocks -= udev->waiting_blocks; udev->waiting_blocks = 0; mutex_unlock(&udev->cmdr_lock); - list_del(&udev->waiter); - wake_up(&udev->wait_cmdr); } - mutex_unlock(&root_udev_waiter_mutex); + spin_unlock(&root_udev_waiter_lock); + +resplice: + spin_lock(&root_udev_waiter_lock); + list_splice(&devs, &root_udev_waiter); + spin_unlock(&root_udev_waiter_lock); } static void check_timedout_devices(void) @@ -2149,6 +2152,7 @@ static void check_timedout_devices(void) static int unmap_thread_fn(void *data) { uint32_t free_blocks = 0; + bool has_block_waiters; bool has_timed_out_devs; while (!kthread_should_stop()) { @@ -2156,13 +2160,23 @@ static int unmap_thread_fn(void *data) prepare_to_wait(&unmap_wait, &__wait, TASK_INTERRUPTIBLE); + /* + * If we had space left, check if devs were added/readded + * while the lock was dropped. + */ + spin_lock(&root_udev_waiter_lock); + has_block_waiters = true; + if (list_empty(&root_udev_waiter)) + has_block_waiters = false; + spin_unlock(&root_udev_waiter_lock); + spin_lock_bh(&timed_out_udevs_lock); has_timed_out_devs = true; if (list_empty(&timed_out_udevs)) has_timed_out_devs = false; spin_unlock_bh(&timed_out_udevs_lock); - if (!has_timed_out_devs) + if ((!free_blocks || !has_block_waiters) && !has_timed_out_devs) schedule(); finish_wait(&unmap_wait, &__wait);