[OpenAFS] Cache file system (ext2) corruption with 1.3.84 and kernel 2.6.11.11

Dr A V Le Blanc Dr A V Le Blanc <LeBlanc@mcc.ac.uk>
Fri, 10 Jun 2005 09:13:01 +0100


On 8 Jun 2005 at 08:55:09 +0100, Dr A V Le Blanc <LeBlanc@mcc.ac.uk> wrote:
> I am running openafs 1.3.84 on Debian sarge machines with kernel
> 2.6.11.11 and openafs-client 1.3.81-3sarge1 from the Debian repository;
> the kernel and openafs.ko modules are compiled with gcc 2.95.4 from
> sarge.
>
> I have been seeing some cache corruption problems for some time,
> and hoped that upgrading to 1.3.84 would solve the problem.
> Unfortunately it doesn't.  I installed this kernel and openafs
> on two machines, did a mkfs of the ext2 file system to be mounted
> at /var/cache/openafs/, rebooted, and waited overnight.  Then I
> logged in and stopped openafs, unmounted the cache filesystem,
> and ran fsck.  Here are the results from two machines; I've
> had many others which are similar.  Can anyone tell me why
> this is happening?  Is there a fix?

As a matter of interest, machines with kernel 2.4.31 and compiled with
gcc 2.95.4 are showing the same problem.  It seems that openafs
1.3.84 corrupts cache partitions pretty reliably.  Machines up for
10 minutes or so don't show this, but all machines using AFS
a reasonable amount seem to have this problem after a few hours.

     -- Owen
     LeBlanc@mcc.ac.uk