On Mon, Feb 07, 2005 at 01:22:45PM -0600, Nicholas Haggin wrote:
> 	Scenario: laptop (IBM TP R40, Celeron 2 GHz, 1 GB RAM, Gentoo Linux,
> kernel 2.6.10 with the Gentoo patchset, Coda compiled statically into
> the kernel, venus 6.0.8) is left running for at least 1 day, perhaps
> longer. At initial startup I clog-ed in and brought X up. Upon returning
> the system seems unstable (Mozilla segfaults on startup, for instance),

Is mozilla using anything from /coda? I'm running the vanilla kernel.org
2.6.10 kernel with Coda built as a loadable module and haven't seen such
problems. You mention 'returning the system', is that return from
hibernation, hardware suspend, swsusp or something similar, or just
returning from standby?

I browsed through the gentoo archive and found 'genpatches-2.6-10.08'
(both base and extra) which seem to be the "gentoo patchset" you
mentioned. It doesn't look like any of the patches in those tarballs
touch anything Coda related.

One thing you could check is if uslocate is trashing through /coda on
it's nightly run, that could tickle some bugs in the venus cache
replacement. But the effect would be that venus dies and accesses to
files in /coda return EIO, it definitely shouldn't destabilize the
system in the way you describe or give any kernel oopses.

> and I decide to restart. During Venus shutdown an oops is generated in
> the Coda kernel code; sometimes after that there is also a panic and
> sometimes not.
> 	Has anyone been experiencing similar symptoms? If the oops backtrace
> would help, how might I obtain one (given that the system is in the
> process of shutting down and the gettys have been killed)?

I am definitely interested in that kernel oops. You could use
'ctrl-alt-f1' to switch to a console, log in as root and kill just the
venus process. Maybe that will trigger the oops without completely
shutting down the system.

Jan