Re: xlog_clear_stale_blocks crashes in line 1242 (suse 9)

[Klaus Ridder <mails.2003__contact-me-at__klaus@xxxxxxxxxx>]

Eric Sandeen wrote:

> On Sun, 2003-11-30 at 04:38, Klaus Ridder wrote:
>
>  
>
> > I am not on the XFS list (and wil not subscribe), so it would be kind of 
> > you to send me a cc when you figured out the problem or need further 
> > information.
> >    
>
> Detailed steps on how to reproduce this problem would be most helpful;
> I'm not certain what steps you took leading up to your problem.
>
> Filing this information in a Bugzilla bug would also be very helpful, if
> possible.  http://oss.sgi.com/bugzilla/
>
> Thanks,
> -Eric
>
>  
1. Installed SUSE 9
2. Created encrypted Raid 5 on 8 IDE disks a 160 GB on 2 IDE Controllers
3. mounted the file system, copied a file on it
4. rebooted the system
5. trying to mount the file system --> error above.

I figured out now that - for whatever reason - the blowfish module was 
not loaded, so I guess that
a) XFS got an encrypted stream instead of the decrypted data
b) maybe the controllers got mixed up so they got junk data as well.

I guess XFS got some sort of totally junk data and didn't handle it 
proberly somehow.
He seems to have scanned for someting for approx. 20 seconds before he 
mentioned the error.
In fact, the first - say - 5 times I tried to mount the drive he didn't 
give the error; but after that the error was exactly reproducable.

I formatted the array with ext3 now so I'm sorry I cannot run any 
further tests as the array is already used in production.
Sorry for the not very detailed problem description.
Probably you can simply catch such errors so that XFS doen't throw an 
error in the mentioned c file with a dump, which looks quite ... well, 
say ... scary ;)

I'm a quit short in time now, so as I don't have a bugzilla account feel 
free to put my mails in there if you want.
It's not urgent for me as I won't use XFS for now.

Regards,
Klaus