From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1762827AbYEBCRm (ORCPT ); Thu, 1 May 2008 22:17:42 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1755820AbYEBCRd (ORCPT ); Thu, 1 May 2008 22:17:33 -0400 Received: from mo10.iij4u.or.jp ([210.138.174.78]:41471 "EHLO mo10.iij4u.or.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751564AbYEBCRc (ORCPT ); Thu, 1 May 2008 22:17:32 -0400 Date: Fri, 2 May 2008 11:17:12 +0900 To: jesper.juhl@gmail.com, jens.axboe@oracle.com, James.Bottomley@HansenPartnership.com Cc: linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org Subject: Re: scsi_scan: WARNING: at include/linux/blkdev.h:431 blk_queue_init_tags+0x107/0x120() From: FUJITA Tomonori In-Reply-To: References: Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-Id: <20080502111652T.tomof@acm.org> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 2 May 2008 00:43:30 +0200 (CEST) Jesper Juhl wrote: > Hi, > > For your information; > > I just build and booted the tip of Linus' git tree (HEAD at > 886c35fbcf6fb2eee15687efc2d64d99b6ad9a4a) and got this in dmesg during > bootup : > > ------------[ cut here ]------------ > WARNING: at include/linux/blkdev.h:431 blk_queue_init_tags+0x107/0x120() > Modules linked in: > Pid: 432, comm: scsi_scan_0 Not tainted 2.6.25-07351-g886c35f #1 > [] warn_on_slowpath+0x54/0x70 > [] ? check_object+0xe3/0x1f0 > [] ? init_tag_map+0x59/0xb0 > [] ? mark_held_locks+0x47/0x80 > [] ? __kmalloc+0x6b/0x100 > [] ? trace_hardirqs_on+0xb9/0x150 > [] ? init_tag_map+0x59/0xb0 > [] ? init_tag_map+0x59/0xb0 > [] ? init_tag_map+0x59/0xb0 > [] ? __blk_queue_init_tags+0x24/0x70 > [] ? __blk_queue_init_tags+0x35/0x70 > [] blk_queue_init_tags+0x107/0x120 > [] ahc_platform_set_tags+0x191/0x1d0 > [] ahc_linux_slave_configure+0x103/0x170 > [] scsi_probe_and_add_lun+0x6e5/0x920 > [] __scsi_scan_target+0xe8/0x5d0 > [] ? native_sched_clock+0x87/0xd0 > [] ? get_lock_stats+0x29/0x50 > [] ? put_lock_stats+0xd/0x30 > [] ? mark_held_locks+0x47/0x80 > [] ? mutex_lock_nested+0x1b7/0x2a0 > [] ? trace_hardirqs_on+0xb9/0x150 > [] ? mutex_lock_nested+0x1aa/0x2a0 > [] ? scsi_scan_host_selected+0x3a/0xf0 > [] scsi_scan_channel+0x70/0x90 > [] scsi_scan_host_selected+0xbd/0xf0 > [] ? do_scan_async+0x0/0x140 > [] do_scsi_scan_host+0x78/0x80 > [] do_scan_async+0x14/0x140 > [] ? complete+0x48/0x60 > [] ? do_scan_async+0x0/0x140 > [] kthread+0x42/0x70 > [] ? kthread+0x0/0x70 > [] kernel_thread_helper+0x7/0x1c The problem is that the commit 75ad23b expects that we hold the queue lock for __blk_queue_free_tags, blk_queue_free_tags and blk_queue_init_tags but we haven't. The simple fix is using queue_flag_set/clear_unlocked for them, then it should work as before. However, it would be better to hold the queue lock for blk_queue_free_tags and blk_queue_init_tags (we can hold the queue lock in scsi_activate_tcq and scsi_deactivate_tcq). diff --git a/block/blk-tag.c b/block/blk-tag.c index de64e04..d4e21cd 100644 --- a/block/blk-tag.c +++ b/block/blk-tag.c @@ -70,7 +70,7 @@ void __blk_queue_free_tags(struct request_queue *q) __blk_free_tags(bqt); q->queue_tags = NULL; - queue_flag_clear(QUEUE_FLAG_QUEUED, q); + queue_flag_clear_unlocked(QUEUE_FLAG_QUEUED, q); } /** @@ -98,7 +98,7 @@ EXPORT_SYMBOL(blk_free_tags); **/ void blk_queue_free_tags(struct request_queue *q) { - queue_flag_clear(QUEUE_FLAG_QUEUED, q); + queue_flag_clear_unlocked(QUEUE_FLAG_QUEUED, q); } EXPORT_SYMBOL(blk_queue_free_tags); @@ -197,7 +197,7 @@ int blk_queue_init_tags(struct request_queue *q, int depth, * assign it, all done */ q->queue_tags = tags; - queue_flag_set(QUEUE_FLAG_QUEUED, q); + queue_flag_set_unlocked(QUEUE_FLAG_QUEUED, q); INIT_LIST_HEAD(&q->tag_busy_list); return 0; fail: