Problems with file systems created on 2.6.18.x

Problems with file systems created on 2.6.18.x

[lkml-2006i-ticket]

Hello, on two different boxes with 2.6.18.x kernels I created two file
systems (ReiserFS and JFS). Both failed totally after about 1-2 days of
using them.

Dmesg reported corruption errors/warnings. When I unmounted and tried to
remount - mound was unable to find superblocks.

Fsck for JFS was hopeless.

Fsck for ReiserFS had to recreate superblocks, journal and rebuild tree
(I am doing that now).

So both FS where very badly corrupted.

On both boxes I have other HDs with multiple filesystems (including JFS
and reiserfs) and all is working fine for over year (including long time
for 2.6.18 line).

Before both failures I was playing with swap (swapoff -a swapon -a), I
use encripted swap.

The "old" partitions seem to be 100% ok all the time. It probably is not
 a hardware problem (I runned some low-level tests).

So, perhaps there is a bug connected to some of the following aspects:
- using newly or recently created file system (bug in code that is used
to grow a short, "young" tree)
- problems with swap / encrypted swap

I saved image of 100 mb of the beginning of reiserfs partition after it
failed,  I can sent it (or part of it) if anyone wants to investigate.

Since I use grsecurity patch -
http://forums.grsecurity.net/viewtopic.php?p=6355#6355

-- 
LimCore    C++ Software Architect / Team Lead
---> oo    Linux programs
limcore
software

Re: Problems with file systems created on 2.6.18.x

[Jan Engelhardt]

On Nov 13 2006 08:24, lkml-2006i-ticket@limcore.pl wrote:
>Hello, on two different boxes with 2.6.18.x kernels I created two file
>systems (ReiserFS and JFS). Both failed totally after about 1-2 days of
>using them.

The disk may have gone bad. (Unlikely, though)

>Dmesg reported corruption errors/warnings. When I unmounted and tried to
>remount - mound was unable to find superblocks.
>
>Before both failures I was playing with swap (swapoff -a swapon -a), I
>use encripted swap.
>
>So, perhaps there is a bug connected to some of the following aspects:
>- using newly or recently created file system (bug in code that is used
>to grow a short, "young" tree)
>- problems with swap / encrypted swap
>
>I saved image of 100 mb of the beginning of reiserfs partition after it
>failed,  I can sent it (or part of it) if anyone wants to investigate.
>
>Since I use grsecurity patch -

Recently there seem to be quite a few corruption reports involving 
setups with grsecurity.

If you can reproduce the problem with gr, see if you can reproduce it 
without gr.


	-`J'
--