kernel oops 2.6.10

[Tomas Olsson]

Andrew Walters <awalter1 at nd.edu> writes:

> I posted both the syslog for the time of the crash with nnpfsdebug and
> the objdump on my afs space:
> 
Ok, so it seems arlad does an fhopen pioctl, but iget() doesn't find the
inode and so tries to create a new one. Unfortunately, the device's
superblock has a null read_inode method so we crash. This is fishy; arlad
should only ask for valid and existing inodes. It does look like the inode
and device are very similar to the working ones earlier in the log, so they
aren't totally bogus.

Right now I don't understand how this could be unless the cache fs is
behaving strangely, but I may be missing something. Does this happen if you
run startarla by hand? Are there matching fhgets earlier in the log? Do you
have a desktop.ini in your afs root?

If you have a separate cache partition, does this happen if you use plain
ext2 on it?

Maybe I'll be able to find some time to get myself an ext3 partition and
play around a bit in a few days.

/t