Xen and crashdumps

Xen and crashdumps

[Goncalo Gomes]

I was wondering how do you usually analyze post-mortem dumps in atypical 
deployments of Xen with Linux as dom0 such as in Citrix XenServer, where 
the hypervisor is 64bit and the dom0 is 32bits (or the other way around 
- if architecturally possible - for that matter)?

If you have any thoughts on how to approach these scenarios, I would 
appreciate if you could shed some light. The crash utility and gdb 
doesn't seem to support these combinations out-of-the-box and the 
minidump output of the kdump tool in XenServer is only useful to 
find/understand the place in the text section where the softlockup is 
being triggered from, but not sufficient to understand why an IO region 
is faulting on access at some point during the driver execution when the 
driver itself only releases this region during cleanup (clearly not the 
case here, as it happens during intensive IO and the first few dozens of 
IOs actually go through). Analyzing the in memory kernel/xen/driver 
structures seems to be the next plausible step to further 
analyze/understand the issue that I can think of. Your thoughts on how 
to analyze the 64bit vmcore using a 32bit vmlinux are much appreciated, 
though.

Thanks,

Goncalo.