From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751889Ab1JQLoM (ORCPT ); Mon, 17 Oct 2011 07:44:12 -0400 Received: from 87-104-106-3-dynamic-customer.profibernet.dk ([87.104.106.3]:38652 "EHLO kernel.dk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750850Ab1JQLoK (ORCPT ); Mon, 17 Oct 2011 07:44:10 -0400 Message-ID: <4E9C1509.6030103@kernel.dk> Date: Mon, 17 Oct 2011 13:44:09 +0200 From: Jens Axboe MIME-Version: 1.0 To: Tejun Heo CC: stable@kernel.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH] block: make gendisk hold a reference to its queue References: <20111017024329.GA23507@google.com> In-Reply-To: <20111017024329.GA23507@google.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 2011-10-17 04:43, Tejun Heo wrote: > The following command sequence triggers an oops. > > # mount /dev/sdb1 /mnt > # echo 1 > /sys/class/scsi_device/0\:0\:1\:0/device/delete > # umount /mnt > > general protection fault: 0000 [#1] PREEMPT SMP > CPU 2 > Modules linked in: > > Pid: 791, comm: umount Not tainted 3.1.0-rc3-work+ #8 Bochs Bochs > RIP: 0010:[] [] __lock_acquire+0x389/0x1d60 > ... > Call Trace: > [] lock_acquire+0x95/0x140 > [] _raw_spin_lock+0x3b/0x50 > [] bdi_lock_two+0x5c/0x70 > [] bdev_inode_switch_bdi+0x4c/0xf0 > [] __blkdev_put+0x11b/0x1d0 > [] __blkdev_put+0x160/0x1d0 > [] blkdev_put+0x5f/0x190 > [] kill_block_super+0x4d/0x80 > [] deactivate_locked_super+0x45/0x70 > [] deactivate_super+0x4a/0x70 > [] mntput_no_expire+0xed/0x130 > [] sys_umount+0x7e/0x3a0 > [] system_call_fastpath+0x16/0x1b > > This is because bdev holds on to disk but disk doesn't pin the > associated queue. If a SCSI device is removed while the device is > still open, the sdev puts the base reference to the queue on release. > When the bdev is finally released, the associated queue is already > gone along with the bdi and bdev_inode_switch_bdi() ends up > dereferencing already freed bdi. > > Even if it were not for this bug, disk not holding onto the associated > queue is very unusual and error-prone. > > Fix it by making add_disk() take an extra reference to its queue and > put it on disk_release() and ensuring that disk and its fops owner are > put in that order after all accesses to the disk and queue are > complete. Thanks, applied. -- Jens Axboe