Ext3 errors with 2.4.18

Ext3 errors with 2.4.18

[Udo A. Steinberg]

Hi,

With Linux 2.4.18, I'm getting multiple of the following error:

EXT3-fs error (device ide0(3,2)): ext3_readdir: bad entry in directory #1966094:
rec_len % 4 != 0 - offset=0, inode=3180611420, rec_len=53134, name_len=138

Can someone decipher this?

Regards,
-Udo.

Re: Ext3 errors with 2.4.18

[Mike Fedyk]

On Mon, May 06, 2002 at 06:25:30PM +0200, Udo A. Steinberg wrote:
> 
> Hi,
> 
> With Linux 2.4.18, I'm getting multiple of the following error:
> 
> EXT3-fs error (device ide0(3,2)): ext3_readdir: bad entry in directory #1966094:
> rec_len % 4 != 0 - offset=0, inode=3180611420, rec_len=53134, name_len=138
> 
> Can someone decipher this?

Can you post the surrounding lines from dmesg?  Many times there is another
error that tells much more than the one you just posted...

Re: Ext3 errors with 2.4.18

[Udo A. Steinberg]

Mike Fedyk wrote:
> 
> Can you post the surrounding lines from dmesg?  Many times there is another
> error that tells much more than the one you just posted...

There are no surrounding lines related to this error.

-Udo.

Re: Ext3 errors with 2.4.18

[Andrew Morton]

"Udo A. Steinberg" wrote:
> 
> Hi,
> 
> With Linux 2.4.18, I'm getting multiple of the following error:
> 
> EXT3-fs error (device ide0(3,2)): ext3_readdir: bad entry in directory #1966094:
> rec_len % 4 != 0 - offset=0, inode=3180611420, rec_len=53134, name_len=138
> 
> Can someone decipher this?
> 

That's random junk.  Heaven knows.  It could have come
from anywhere in the kernel.  Including ext3, of course.

When you see this sort of thing you should immediately
take the machine down, because the corruption could be
only in-memory. The longer the machine stays up, the
better the chance that the corruption will go to disk.

And with ext3, the best way to take the machine down
is to pull the power plug.  Normal shutdown tools will
sync the disks, which you don't want to happen.

Then reboot with `init=/bin/sh' and force a fsck against
all filesystems.

-

Re: Ext3 errors with 2.4.18

[Oliver Feiler]

On Monday 06 May 2002 18:25, Udo A. Steinberg wrote:
> Hi,
>
> With Linux 2.4.18, I'm getting multiple of the following error:
>
> EXT3-fs error (device ide0(3,2)): ext3_readdir: bad entry in directory
> #1966094: rec_len % 4 != 0 - offset=0, inode=3180611420, rec_len=53134,
> name_len=138

Hi,

I experienced the same problem with ext3 + 2.4.18 on a RAID-1. Out of nowhere 
the following error popped up in the syslog, no other surrounding error 
messages. The fs was mounted read-only automatically. After reboot and fsck 
there were /a lot/ of errors on the disk. Virtually all errors fsck knows I 
think. :)  After fsck ran multiple times on the disk, lost+found was filled 
with stuff from all accross the disk, but other than that the fs seems to 
have survived it.

May 14 01:24:41 kianga kernel: EXT3-fs error (device md(9,0)): ext3_readdir: 
bad entry in directory
#2142115: rec_len %% 4 != 0 - offset=0, inode=4186379891, rec_len=16755, 
name_len=219
May 14 01:24:51 kianga last message repeated 2 times

Regards,
Oliver

-- 
Oliver Feiler  <kiza@(gmx.net|lionking.org|claws-and-paws.com|spot.dtip.de)> /
http://www.lionking.org/~kiza/      (http://spot.dtip.de/)                  /
___________________________________________________________________________/

Re: Ext3 errors with 2.4.18

[Andreas Dilger]

On May 14, 2002  12:38 +0200, Oliver Feiler wrote:
> On Monday 06 May 2002 18:25, Udo A. Steinberg wrote:
> > Hi,
> >
> > With Linux 2.4.18, I'm getting multiple of the following error:
> >
> > EXT3-fs error (device ide0(3,2)): ext3_readdir: bad entry in directory
> > #1966094: rec_len % 4 != 0 - offset=0, inode=3180611420, rec_len=53134,
> > name_len=138
> 
> Hi,
> 
> I experienced the same problem with ext3 + 2.4.18 on a RAID-1. Out of nowhere 
> the following error popped up in the syslog, no other surrounding error 
> messages. The fs was mounted read-only automatically. After reboot and fsck 
> there were /a lot/ of errors on the disk. Virtually all errors fsck knows I 
> think. :)  After fsck ran multiple times on the disk, lost+found was filled 
> with stuff from all accross the disk, but other than that the fs seems to 
> have survived it.

There have been several reports of strange ext3 errors when running on
MD RAID.  I don't know if anyone is looking into this yet.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Re: Ext3 errors with 2.4.18

[Neil Conway]

Your errors are just like the ones I experienced due to an IDE bug.

Are you: 
a) using a hard-disk on the same cable as a CD/DVD

b) seeing any "hdX: status error: status=0x58 {DriveReady SeekComplete
DataRequest}" errors in your log

c) seeing (b) in close proximity to ATAPI module init messages like:
"hdb: ATAPI 24X DVD-ROM drive, 128kB Cache, UDMA(33)"

?

If yes to all/most of the above, you may be suffering from the same
bug.  If not, your inodes in RAM are probably being trashed by a
different bug :-))

Neil

Re: Ext3 errors with 2.4.18

[Oliver Feiler]

On Tuesday 14 May 2002 12:59, Neil Conway wrote:
> Your errors are just like the ones I experienced due to an IDE bug.

It's a slightly older P3B-F board with Intel IDE. It's not my system, but it 
has been running fine as long as I can remember.

>
> Are you:
> a) using a hard-disk on the same cable as a CD/DVD

The IDE disk layout is:

IDE pri: HD + HD
IDE sec: CD + HD

The RAID is constructed from IDE primary master + secondary slave. (40 + 30 GB 
disks)

>
> b) seeing any "hdX: status error: status=0x58 {DriveReady SeekComplete
> DataRequest}" errors in your log
>
> c) seeing (b) in close proximity to ATAPI module init messages like:
> "hdb: ATAPI 24X DVD-ROM drive, 128kB Cache, UDMA(33)"
>
> ?
>
> If yes to all/most of the above, you may be suffering from the same
> bug.

b) and c) are no, no.

>  If not, your inodes in RAM are probably being trashed by a
> different bug :-))

Hm, wherever. ;)  The problem manifested itself right after a reboot. Postfix 
complained at boot about dirs in spool/postfix not owned by postfix and at 
some point the fs was remounted ro. I dunno where to look, maybe it's indeed 
MD RAID + ext3 like Andreas pointed out?

Bye,
Oliver

-- 
Oliver Feiler  <kiza@(gmx.net|lionking.org|claws-and-paws.com|spot.dtip.de)> /
http://www.lionking.org/~kiza/      (http://spot.dtip.de/)                  /
___________________________________________________________________________/