[OpenAFS-devel] OpenAFS client crash on AIX

[Niklas Edmundsson]

Hi all!

We have gotten this crash a couple of times on our AIX 5.3-box. It 
usually happens when I'm harassing the attached storage, which causes 
idle processes such as the openafs client to be paged out.

My theory is that an interrupt happens, which tries to access data 
which is paged out, and the world falls apart. Is this a reasonable 
guess? Has anyone seen this before?

I'm not able to read the system dump with kdb, even though dmp_minimal 
obviously manages to extract a stack trace and put it in the error 
log. I think I'll poke IBM about that.

In the meantime, this is the stack trace:

--------------------------8<---------------------------
Description
Previous system dump information

Detail Data
Crash Code
0000 0300
Crash Stack
000266cc e_clear_wait+100
000266c8 e_clear_wait+fc
0189abb0
040718d0 AfsWaitHack+30
0005c650 clock+174
0005d8fc i_softmod+280
00033af4 sig_errno+fffff9e4


Description
DATA STORAGE INTERRUPT, PROCESSOR

Detail Data
DATA STORAGE INTERRUPT STATUS REGISTER
0000 0000 0000 0000
SEGMENT REGISTER, SEGREG
4000 0000 0000 7FFF
DATA STORAGE INTERRUPT ADDRESS REGISTER
FFFF D000 F100 0800
EXVAL
1195 6018 FFFF FFFF
--------------------------8<---------------------------



/Nikke
-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
  Niklas Edmundsson, Admin @ {acc,hpc2n}.umu.se     |    nikke@hpc2n.umu.se
---------------------------------------------------------------------------
  "Why am I not surprised?" - Iago
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=