[OpenAFS] AFS hangs on cache initialization at startup

Derrick J Brashear shadow@dementia.org
Fri, 20 Apr 2001 16:09:08 -0400 (EDT)


On Fri, 20 Apr 2001, Michael Durket wrote:

> On s390 Linux kernel 2.2.19 (latest RedHat beta), I've recompiled
> OpenAFS 1.0.3 (using the s390 patches). Mostly, AFS fails to start
> properly - it hangs in the AFS cache scan. On very rare occasions
> it will start.
> 
> I've tried debugging this by putting printf statements in various
> places, but this just moves the problem apparently. Sometimes
> AFS will come up ok, sometimes it hangs in a different place. But,
> it's always hung doing the AFSOP_CACHEINIT call.
> 
> I'm not saying that this is an AFS problem (it might be the beta
> kernel), but I'm looking for some suggestions from more experienced
> AFS debuggers who might have seen similar hangs in cache init on
> Linux or other OSes and have suggestions about how to approach
> debugging the problem.

fstrace setset cm -active
fstrace setset cmlongterm -active
fstrace dump

and then have fun correlating events; if you search google for "fstrace"
you should find some hints on how to use it. i'm aware we need this
documented, and if no one else does anything, it's on my list.

that said i bet it's like the alan cox patches, something's changing in
the inode structure and you're losing because of it.

-D