reiserfs problem

reiserfs problem

[Patrick Mackinlay]

When 2.4.1 was released I reported a kernel oops with reiserfs, I got no
response.
The problem still persists with kernel 2.4.2. I am using mkreiserfs v2
(3.6.25) and tried on different ide hards disks as well as on a software
raid 0 partition, I also tried with and without CONFIG_REISERFS_CHECK
defined, the oops always happens.

My system is a dual pentium 500, 256M memory, IDE:
IDE interface: Intel Corporation 82371AB PIIX4 IDE (rev 01)
The system has been running flawlessly with 2.2 kernels for months.

This is how I caused the first oops:
Created 2Gig reiserfs partition on /dev/md0 (raid0)
mount /dev/md0 /mnt/md0
Mount took 3-5 seconds
cp -ax / /mnt/md0
Only the /bin directory (not the contents!) was copied and then cp core
dumped. syslog reports an oops, however the system is still usable but
mount/umount/sync is not possible on any filesystems.

kernel oops report:
>>EIP; d2cf9db8 <[reiserfs]create_virtual_node+298/490>   <=====
Trace; d2cfc053 <[reiserfs]dc_check_balance_leaf+53/150>
Trace; d2cfc975 <[reiserfs]fix_nodes+115/450>
Trace; d2d062ac <[reiserfs]reiserfs_cut_from_item+1fc/430>
Trace; d2d152f0 <[reiserfs]reiserfs_mounted_fs_count+0/4>
Trace; d2d0686e <[reiserfs]reiserfs_do_truncate+32e/470>
Trace; d2d18044 <.bss.end+1165/???
Trace; d2cf8567 <[reiserfs]reiserfs_truncate_file+b7/1a0>
Trace; d2d0fa24 <[reiserfs].rodata.start+11e4/678e>
Trace; d2d141af <[reiserfs].rodata.start+596f/678e>
Trace; d2cf9382 <[reiserfs]reiserfs_file_release+1d2/380>
Trace; d2cf94cf <[reiserfs]reiserfs_file_release+31f/380>
Trace; d2d0fb8c <[reiserfs].rodata.start+134c/678e>
Trace; d2d141af <[reiserfs].rodata.start+596f/678e>
Trace; c0128230 <file_read_actor+0/60>
Trace; c0135e47 <fput+37/e0>
Trace; c0134d56 <filp_close+a6/b0>
Trace; c0134dbb <sys_close+5b/70>
Trace; c0109127 <system_call+33/38>
Code;  d2cf9db8 <[reiserfs]create_virtual_node+298/490>
00000000 <_EIP>:
Code;  d2cf9db8 <[reiserfs]create_virtual_node+298/490>   <=====
   0:   8b 40 14                  mov    0x14(%eax),%eax   <=====
Code;  d2cf9dbb <[reiserfs]create_virtual_node+29b/490>
   3:   ff d0                     call   *%eax
Code;  d2cf9dbd <[reiserfs]create_virtual_node+29d/490>
   5:   89 c2                     mov    %eax,%edx
Code;  d2cf9dbf <[reiserfs]create_virtual_node+29f/490>
   7:   8b 06                     mov    (%esi),%eax
Code;  d2cf9dc1 <[reiserfs]create_virtual_node+2a1/490>
   9:   83 c4 10                  add    $0x10,%esp
Code;  d2cf9dc4 <[reiserfs]create_virtual_node+2a4/490>
   c:   01 c2                     add    %eax,%edx
Code;  d2cf9dc6 <[reiserfs]create_virtual_node+2a6/490>
   e:   89 16                     mov    %edx,(%esi)
Code;  d2cf9dc8 <[reiserfs]create_virtual_node+2a8/490>
  10:   8b 83 8c 01 00 00         mov    0x18c(%ebx),%eax


-- 
Patrick Mackinlay                                patrick@spacesurfer.com
ICQ: 59277981                                        tel: +44 7050699851
                                                     fax: +44 7050699852
SpaceSurfer Limited                          http://www.spacesurfer.com/

Re: reiserfs problem

[Chris Mason]

On Friday, February 23, 2001 05:03:40 PM +0000 Patrick Mackinlay
<patrick@spacesurfer.com> wrote:

> When 2.4.1 was released I reported a kernel oops with reiserfs, I got no
> response.

Hmmm, don't seem to have any other reiserfs mail from you.  Sorry I missed
it.

[ ...]
> 
> kernel oops report:
>>> EIP; d2cf9db8 <[reiserfs]create_virtual_node+298/490>   <=====

We've had a half dozen or so reports so far of this, almost every time it
has been resolved as a compiler issue, usually an unpatched gcc 2.96.  One
user was using 2.95.3, but did not report back if switching to 2.95.2 or
egcs fixed the oops.

Anyway, which compiler did you use to compile the kernel?

-chris

Re: reiserfs problem

[Patrick Mackinlay]

Chris Mason wrote:
> 
> On Friday, February 23, 2001 05:03:40 PM +0000 Patrick Mackinlay
> <patrick@spacesurfer.com> wrote:
> 
> > When 2.4.1 was released I reported a kernel oops with reiserfs, I got no
> > response.
> 
> Hmmm, don't seem to have any other reiserfs mail from you.  Sorry I missed
> it.
> 
> [ ...]
> >
> > kernel oops report:
> >>> EIP; d2cf9db8 <[reiserfs]create_virtual_node+298/490>   <=====
> 
> We've had a half dozen or so reports so far of this, almost every time it
> has been resolved as a compiler issue, usually an unpatched gcc 2.96.  One
> user was using 2.95.3, but did not report back if switching to 2.95.2 or
> egcs fixed the oops.
> 
> Anyway, which compiler did you use to compile the kernel?
> 
> -chris

I didn't want to join the kernel mailing list (too much traffic) so I
sent it to the email address suggested on the reiserfs home page. I
guess it didn't get through.

It seems I am in fact using a buggy compiler (damn RedHat 7, forgot I
had upgraded on this machine) I will try re-compiling with another
compiler and let you know if I still have problems.

regards,
Patrick

-- 
Patrick Mackinlay                                patrick@spacesurfer.com
ICQ: 59277981                                        tel: +44 7050699851
                                                     fax: +44 7050699852
SpaceSurfer Limited                          http://www.spacesurfer.com/