Ext3 users - Jan 2002 - Re: OOPS: kernel BUG at transaction.c:1857 on 2.4.17 while rm'ing 700mb file on ext3 partition.

Hi,

On Thu, Jan 24, 2002 at 04:54:34PM +0100, frode wrote:
> 
> I got the following error while rm'ing a 700mb file from an ext3
partition:
> 
> Assertion failure in journal_unmap_buffer() at transaction.c:1857:
> "transaction == journal->j_running_transaction"
Hmm --- this is not one I think I've ever seen before.
> >>EIP; c015ea1a <journal_unmap_buffer+fa/1b0>   <====>
Trace; c015eb6e <journal_flushpage+9e/140>
> Trace; c0156ae2 <ext3_flushpage+22/30>
> Trace; c0125738 <do_flushpage+18/30>
> Trace; c0125762 <truncate_complete_page+12/50>
> Trace; c01258c6 <truncate_list_pages+126/190>
> Trace; c0125970 <truncate_inode_pages+40/70>
> Trace; c014485e <iput+ae/200>
> Trace; c0142e4c <d_delete+4c/70>
> Trace; c013c69c <vfs_unlink+13c/170>
> Trace; c013c778 <sys_unlink+a8/120>
> Trace; c0106e8a <system_call+32/38>
Well, that's a straight forward trace, and looks perfectly normal for
a delete operation.  The buffer_head is locked at this point, and the
transaction itself is pinned, so I can't see any way to have an
unrecognised transaction here.
> I use the 'mem=nopentium' option on the lilo prompt while booting,
hoping to
> reduce the rather large amount of oopses I have had recently, as I read 
> something about AMD Athlons and AGP causing troubles.
Those problems included AGP cache coherency problems, but I didn't see
any mention of other instabilities as a result.  Also,
> NVRM: loading NVIDIA NVdriver Kernel Module  1.0.2313  Tue Nov 27 12:01:24
PST 2001
with this driver loaded we really can't make any guarantees about your
system stability at all.  If you manage to eliminate other oopses and
still get the ext3 one, even without the NVidia driver loaded, then
there would be a much better change of debugging things, but right now
it sounds like a hardware problem.

Cheers,
 Stephen

Stephen C. Tweedie wrote:
> On Thu, Jan 24, 2002 at 04:54:34PM +0100, frode wrote:
>>I got the following error while rm'ing a 700mb file from an ext3
partition:
>>Assertion failure in journal_unmap_buffer() at transaction.c:1857:
>>"transaction == journal->j_running_transaction"
> Hmm --- this is not one I think I've ever seen before.
[oops trace snipped]
>>NVRM: loading NVIDIA NVdriver Kernel Module  1.0.2313  Tue Nov 27
12:01:24 PST 2001
> with this driver loaded we really can't make any guarantees about your
> system stability at all.  If you manage to eliminate other oopses and
> still get the ext3 one, even without the NVidia driver loaded, then
> there would be a much better change of debugging things, but right now
> it sounds like a hardware problem.
OK, I rebooted and gzip'ed the NVdriver in /lib/modules... to make sure the 
module doesn't load (lsmod now says my kernel isn't tainted). I'll
try using the
plain 'nv' driver shipped with XFree instead for a while. I tried making
another
700mb iso image and fool around with it (loopback mount it, umount it, then rm 
it) but couldn't trigger anything - but I just spent five minutes trying.

As I mentioned I have had quite a few oopses lately, most of them regarding 
paging etc. (but I'm no kernel expert). See for example
http://marc.theaimsgroup.com/?l=linux-kernel&m=101096234600708&w=2
and
http://marc.theaimsgroup.com/?l=linux-kernel&m=101128528029736&w=2

I'm running linux on an old p100 as well but don't see any problems, so
as you
say I suspected a hardware problem. I ran MemTest86 for about half an hour 
without any errors (but of course there's plenty of other things that may be
wrong).

Do you have any suggestions on other ways I could try to put my hardware 
stability on trial, or try to reproduce the bug (to see if it occurs on a 
non-tainted kernel)?

  - Frode