[OpenAFS] System still panicing.. no help found yet..

Nickolai Zeldovich kolya@MIT.EDU
Wed, 13 Mar 2002 13:08:08 -0500

> I still have a system which is panicing 2-3 time a day...  SUN says it is 
> in the same place each time
> from the coredump...

This seems very similar to the panics we see on one of our web
servers, though with much less frequency than you describe (maybe
once every few weeks).  Could you also show the stack trace from
the crash dumps?  (E.g., adb -k unix.N vmcore.N, and $c).

> kernel memory allocator: buffer freed to wrong cache
> buffer was allocated from kmem_alloc_16,
> caller attempting free to kmem_alloc_8.
> buffer=300092fd6c0 bufctl=3000bf703c0 cache: kmem_alloc_8

Actually, could you also run something like "300092fd6c0/s"
and "300092fd6c0/4X", to see what's in the link data field?

Given the allocator stack trace you've shown, it seems very
likely to be an AFS problem, though.  Infact, I was suspecting
AFS link data allocation as well, when debugging our webserver,
but didn't have any reliable stack traces to go on further.

-- kolya