[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: ext3 + autofs4 yields kernel oops



On Wed, Aug 22, 2001 at 11:28:16PM -0700, Ben Liblit wrote:
> I recently applied the ext3-2.4-0.9.6-249.gz patch to an otherwise 
> unmodified official-from-Linus 2.4.9 kernel.  I built both ext2 and ext3 
> in as modules, did the "tune2fs -j" incantation, and rebooted.
> 
> Mounting / as ext3 worked fine.  But a few seconds later, my startup 
> scripts tried to launch a squad of automounter daemons, and that didn't 
> work so well.  A stream of kernel oopsen follows, all apparently having 
> something to do with those automount processes. I've attached a copy of 
> those messages below.
> 
> It is also worth mentioning that the oopsen appear when using the 
> ext3-patched kernel even if the no ext3 filesystem is actually being 
> used.  If I change /etc/fstab to request ext2 mounting, the oopsen still 
> appear.  Even if I use tune2fs to actually remove the journal, restoring 
> the filesystem to pure ext2, the oopsen still appear.  I've had no 
> choice but to revert to a non-ext3-patched kernel.  :-(

It sounds like autofs4 probably has some bugs where it's derferencing
freed objects, and it's getting caught by the slab poisoning which the
ext3 patch enables.  Try going to /usr/src/linux/mm/slab.c, and
changing the "#define DEBUG" from 1 back to 0; I suspect that will
make your autofs4 problems go away.

I've cc'ed Jermy Fitzhardinge, the autofs v4 maintainer, so he can get
a heads up.  If your autofs oops messages go away after you turn off
slab cache debugging code, you should definitely let Jermey know.

						- Ted





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]