On Fri, Aug 16, 2002 at 05:31:40AM -0500, Stephen Lord wrote:
> On Fri, 2002-08-16 at 03:33, Ethan Benson wrote:
> >
> > i reported a deadlocked mount -o remount,ro process a couple monthes
> > ago and got no response.
> >
> > now today i have had a ordinary umount (not remount) go into an
> > unkillable D state permanently.
>
>
> I have seen a hang in unmount, usually unmounting my root on my
> laptop, but failed to reproduce it anywhere else. I also failed
> to reproduce it without having the binary kernel module I have
> to use to to get through SGI's firewall loaded. This last one
> made me think it was actually something that module was doing.
>
> Maybe its time to dig a little more.
my suggestion on reproducing would be to write as much data as
possible and then try and umount (or remount ro) as quickly as
possible, even try to umount at the same time as some operation (my
remount ro hang occured when the remount was happening at the same
moment some rm commands were running)
im no kernel hacker but it just seems like its some sort of race when
xfs or whatever is commiting things to the fs in preperation for
umount.
--
Ethan Benson
http://www.alaska.net/~erbenson/
pgpRurcNXEQ5i.pgp
Description: PGP signature
|