ReiserFS: clm-2100: nesting info a different FS

ReiserFS: clm-2100: nesting info a different FS

[Marti Raudsepp]

Hi,

My dmesg output contains a lot of warning messages from reiserfs and
I'm wondering if this is anything important:
ReiserFS: dm-0: warning: clm-2100: nesting info a different FS

I'm running Hardened Gentoo kernel 2.6.16-r11 on AMD64. I've got three
reiserfs filesystems: one for root which is mounted directly from a
RAID1 array (/dev/md/2), and two are LVM volumes from another RAID1
array (/dev/md/3); one of the latter has usrquota enabled. I'm not
really sure which block device "dm-0" refers to. The computer is doing
relatively little disk I/O most of the time.

The comment in fs/reiserfs/journal.c says:
/* we've ended up with a handle from a different filesystem.
 ** save it and restore on journal_end.  This should never
 ** really happen...
 */

The "should never really happen" part made me wonder, is this anything
I should be worried about? Over the last 92 days, my dmesg contains
123 errors like this; 90 of those appeared on 3rd of September, the
rest are many days apart; there doesn't appear to be any correlation
with any messages in other log files. They appear to come in bursts of
spanning 0.4 or less milliseconds, and are sometimes followed by more
bursts a few seconds later.

There have been no apparent problems with the performance of the disks
or system as a result. According to Google searches, earlier kernels
exhibited this problem with quotas enabled, but that bug was fixed
back in 2.6.14. And nothing significant appears to have changed in
fs/reiserfs/journal.c from 2.6.16 till today.

Regards,
Marti Raudsepp

Re: ReiserFS: clm-2100: nesting info a different FS

[Chris Mason]

[-- Attachment #1: Type: text/plain, Size: 1277 bytes --]

On Thu, 6 Sep 2007 17:31:53 +0300
"Marti Raudsepp" <marti@juffo.org> wrote:

> Hi,
> 
> My dmesg output contains a lot of warning messages from reiserfs and
> I'm wondering if this is anything important:
> ReiserFS: dm-0: warning: clm-2100: nesting info a different FS
> 
> I'm running Hardened Gentoo kernel 2.6.16-r11 on AMD64. I've got three
> reiserfs filesystems: one for root which is mounted directly from a
> RAID1 array (/dev/md/2), and two are LVM volumes from another RAID1
> array (/dev/md/3); one of the latter has usrquota enabled. I'm not
> really sure which block device "dm-0" refers to. The computer is doing
> relatively little disk I/O most of the time.
> 
> The comment in fs/reiserfs/journal.c says:
> /* we've ended up with a handle from a different filesystem.
>  ** save it and restore on journal_end.  This should never
>  ** really happen...
>  */
> 
> The "should never really happen" part made me wonder, is this anything
> I should be worried about?

Well, it's really not supposed to happen, and hopefully it indicates
that someone is doing an allocation without GFP_NOFS when they really
should be using it.  A simple stack trace will show us pretty quickly.

Can you try a patch to add a stack trace to this warning?  I've
attached it.

-chris

[-- Attachment #2: warning.patch --]
[-- Type: text/x-patch, Size: 410 bytes --]

diff -r 70b4267862e0 fs/reiserfs/journal.c
--- a/fs/reiserfs/journal.c	Mon Jul 09 00:00:31 2007 +0000
+++ b/fs/reiserfs/journal.c	Mon Sep 10 09:37:39 2007 -0400
@@ -3165,6 +3165,7 @@ int journal_begin(struct reiserfs_transa
 			 */
 			reiserfs_warning(p_s_sb,
 					 "clm-2100: nesting info a different FS");
+			WARN_ON(1);
 			th->t_handle_save = current->journal_info;
 			current->journal_info = th;
 		}