NetBSD Problem Report #41078

From www@NetBSD.org  Thu Mar 26 16:01:18 2009
Return-Path: <www@NetBSD.org>
Received: from mail.netbsd.org (mail.netbsd.org [204.152.190.11])
	by www.NetBSD.org (Postfix) with ESMTP id 41ED763C133
	for <gnats-bugs@gnats.netbsd.org>; Thu, 26 Mar 2009 16:01:18 +0000 (UTC)
Message-Id: <20090326160117.DB22063B8DE@www.NetBSD.org>
Date: Thu, 26 Mar 2009 16:01:17 +0000 (UTC)
From: Christoph_Egger@gmx.de
Reply-To: Christoph_Egger@gmx.de
To: gnats-bugs@NetBSD.org
Subject: mounting ext2fs via vnd causes LOCKDEBUG panic
X-Send-Pr-Version: www-1.0

>Number:         41078
>Category:       kern
>Synopsis:       mounting ext2fs via vnd causes LOCKDEBUG panic
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    haad
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Mar 26 16:05:00 +0000 2009
>Closed-Date:    Wed Dec 30 00:14:17 +0000 2009
>Last-Modified:  Wed Dec 30 00:14:17 +0000 2009
>Originator:     Christoph Egger
>Release:        NetBSD/amd64 5.99.8
>Organization:
>Environment:
>Description:

Lockdebug panic happens when mounting ext2fs via vnd.

Reader / writer lock error: lockdebug_unlocked: no shared holds on this lock

lock address : 0xffff80004eea4d20 type     :     sleep/adaptive
initialized  : 0xffffffff80711166
shared holds :                  0 exclusive:                  0
shares wanted:                  0 exclusive:                  0
current cpu  :                  0 last held:                  0
current lwp  : 0xffff80004f4d97e0 last held: 000000000000000000
last locked  : 0xffffffff8070ea52 unlocked : 0xffffffff8070ead8
owner/count  : 000000000000000000 flags    : 0x0000000000000008

Turnstile chain at 0xffffffff80ce89e0.
=> No active turnstile for this lock.

panic: LOCKDEBUG
fatal breakpoint trap in supervisor mode
trap type 1 code 0 rip ffffffff802388c5 cs 8 rflags 246 cr2  7f7ffd96ecb0 cpl 0
rsp ffff80004f4fe650
Stopped in pid 435.1 (mount_ext2fs) at  netbsd:breakpoint+0x5:  leave
db{0}> bt
breakpoint() at netbsd:breakpoint+0x5
panic() at netbsd:panic+0x289
lockdebug_abort1() at netbsd:lockdebug_abort1+0xd3
rw_exit() at netbsd:rw_exit+0xe9
vlockmgr() at netbsd:vlockmgr+0xd8
VOP_UNLOCK() at netbsd:VOP_UNLOCK+0x64
spec_open() at netbsd:spec_open+0x340
VOP_OPEN() at netbsd:VOP_OPEN+0x62
ext2fs_mount() at netbsd:ext2fs_mount+0x25a
VFS_MOUNT() at netbsd:VFS_MOUNT+0x44
do_sys_mount() at netbsd:do_sys_mount+0x62d
sys___mount50() at netbsd:sys___mount50+0x33
syscall() at netbsd:syscall+0xb6

>How-To-Repeat:

vndconfig -c vnd0 linux.img
mount -t ext2fs /dev/vnd0a /mnt

>Fix:

>Release-Note:

>Audit-Trail:
From: Izumi Tsutsui <tsutsui@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/41078 CVS commit: src/sys/ufs/ext2fs
Date: Fri, 11 Sep 2009 15:59:07 +0000

 Module Name:	src
 Committed By:	tsutsui
 Date:		Fri Sep 11 15:59:07 UTC 2009

 Modified Files:
 	src/sys/ufs/ext2fs: ext2fs_vfsops.c

 Log Message:
 Fix botch around argument check in ext2fs_mount(). Taken from ffs_vfsops.c.

 Fixes LOCKDEBUG panic which is the same one mentioned in PR kern/41078
 on trying to mount_ext2fs against a raw device, while that panic
 seems to have another route cause around module_autoload() in
 sys/miscfs/specfs/spec_vnops.c:spec_open().


 To generate a diff of this commit:
 cvs rdiff -u -r1.144 -r1.145 src/sys/ufs/ext2fs/ext2fs_vfsops.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

From: Izumi Tsutsui <tsutsui@ceres.dti.ne.jp>
To: gnats-bugs@NetBSD.org
Cc: kern-bug-people@NetBSD.org, gnats-admin@NetBSD.org, netbsd-bugs@NetBSD.org,
        tsutsui@ceres.dti.ne.jp
Subject: Re: kern/41078: mounting ext2fs via vnd causes LOCKDEBUG panic
Date: Sat, 12 Sep 2009 02:01:44 +0900

 > >Synopsis:       mounting ext2fs via vnd causes LOCKDEBUG panic
  :
 > panic() at netbsd:panic+0x289
 > lockdebug_abort1() at netbsd:lockdebug_abort1+0xd3
 > rw_exit() at netbsd:rw_exit+0xe9
 > vlockmgr() at netbsd:vlockmgr+0xd8
 > VOP_UNLOCK() at netbsd:VOP_UNLOCK+0x64
 > spec_open() at netbsd:spec_open+0x340
 > VOP_OPEN() at netbsd:VOP_OPEN+0x62
 > ext2fs_mount() at netbsd:ext2fs_mount+0x25a
 > VFS_MOUNT() at netbsd:VFS_MOUNT+0x44
 > do_sys_mount() at netbsd:do_sys_mount+0x62d
 > sys___mount50() at netbsd:sys___mount50+0x33
 > syscall() at netbsd:syscall+0xb6

 This is not ext2fs or vnd(4) specific at all.

 "mount_ffs /dev/wd7i /mnt" (where no wd7 disk)
 on a LOCKDEBUG kernel is enough to cause this panic.

 spec_open() in spec_vnops.c tries to call module_autoload()
 if bdev_open() returns ENXIO (== device not configured),
 but bdev_open() could return ENXIO even if the driver is
 already loaded or builtin. (due to no device, no vnconfig(8) etc.)

 The following patch seems to fix the panic and
 vnd.kmod without builtin vnd still works properly,
 but I wonder if it's worth to have some helper functions
 in kern/subr_devsw.c for module_autoload() calls on demand.

 ---
 Index: spec_vnops.c
 ===================================================================
 RCS file: /cvsroot/src/sys/miscfs/specfs/spec_vnops.c,v
 retrieving revision 1.124
 diff -u -r1.124 spec_vnops.c
 --- spec_vnops.c	25 Apr 2009 15:06:32 -0000	1.124
 +++ spec_vnops.c	11 Sep 2009 16:28:28 -0000
 @@ -405,11 +405,20 @@
  			vp->v_vflag |= VV_ISTTY;
  		VOP_UNLOCK(vp, 0);
  		do {
 +			const struct cdevsw *cdev;
 +
  			gen = module_gen;
  			error = cdev_open(dev, ap->a_mode, S_IFCHR, l);
  			if (error != ENXIO)
  				break;

 +			/* Check if we already have a valid driver */
 +			mutex_enter(&device_lock);
 +			cdev = cdevsw_lookup(dev);
 +			mutex_exit(&device_lock);
 +			if (cdev != NULL)
 +				break;
 +
  			/* Get device name from devsw_conv array */
  			if ((name = cdevsw_getname(major(dev))) == NULL)
  				break;
 @@ -447,11 +456,20 @@
  		sd->sd_bdevvp = vp;
  		mutex_exit(&device_lock);
  		do {
 +			const struct bdevsw *bdev;
 +
  			gen = module_gen;
  			error = bdev_open(dev, ap->a_mode, S_IFBLK, l);
  			if (error != ENXIO)
  				break;

 +			/* Check if we already have a valid driver */
 +			mutex_enter(&device_lock);
 +			bdev = bdevsw_lookup(dev);
 +			mutex_exit(&device_lock);
 +			if (bdev != NULL)
 +				break;
 +
  			/* Get device name from devsw_conv array */
  			if ((name = bdevsw_getname(major(dev))) == NULL)
  				break;

 ---
 Izumi Tsutsui

From: Izumi Tsutsui <tsutsui@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/41078 CVS commit: src/sys/miscfs/specfs
Date: Sun, 4 Oct 2009 06:23:58 +0000

 Module Name:	src
 Committed By:	tsutsui
 Date:		Sun Oct  4 06:23:58 UTC 2009

 Modified Files:
 	src/sys/miscfs/specfs: spec_vnops.c

 Log Message:
 Put workaround fix for LOCKDEBUG panic mentioned in PR kern/41078:
  Don't try to load a driver module if the driver is already exist but just
  not attached. [bc]dev_open() could return ENXIO even if the driver exists.

 XXX: Maybe this should be handled by helper functions for
 XXX: module_autoload() calls on demand.


 To generate a diff of this commit:
 cvs rdiff -u -r1.124 -r1.125 src/sys/miscfs/specfs/spec_vnops.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

Responsible-Changed-From-To: kern-bug-people->haad
Responsible-Changed-By: tsutsui@NetBSD.org
Responsible-Changed-When: Sun, 04 Oct 2009 15:33:41 +0900
Responsible-Changed-Why:
Assign to a person who implemented it.


From: "Christoph Egger" <Christoph_Egger@gmx.de>
To: gnats-bugs@NetBSD.org, tsutsui@NetBSD.org, gnats-admin@netbsd.org,
 netbsd-bugs@netbsd.org, kern-bug-people@netbsd.org, haad@NetBSD.org
Cc: 
Subject: Re: kern/41078 (mounting ext2fs via vnd causes LOCKDEBUG panic)
Date: Mon, 05 Oct 2009 09:17:34 +0200

 I retested and I can no longer reproduce the LOCKDEBUG panic.

 I can mount ext2fs via vnd, some simple read/write accesses
 with ls, cat and echo "bla" >>testfile
 and unmounting the image work well.

 Christoph

State-Changed-From-To: open->closed
State-Changed-By: haad@NetBSD.org
State-Changed-When: Wed, 30 Dec 2009 00:14:17 +0000
State-Changed-Why:
User can't reproduce this panic anymore and there were some fixes commited to this code some time ago.


>Unformatted:

Home	PR Database Search