[OpenAFS-devel] Re: System lockup with do_IRQ: stack overflow (SOMEWHAT RESOLVED)

[Alexander Bergolth]

On 02/26/2007 11:13 AM, Axel Thimm wrote:
> On Mon, Feb 26, 2007 at 04:51:54PM +1100, Deon George wrote:
>> I was under the suspicion that it was the atrpm build of the openafs 
>> kernel module that was causing my systems to lock up with do_IRQ: stack 
>> overflow.
>>
>[...]
>> Axel, I dont know why your openafs kmdl is causing my lockup
> 
> The ATrpms packages are rather pristine 1.4.2 sources. Perhaps the
> openafs-1.4.2-rhel4.1.src.rpm package has some fixes? Could you try
> building from source?

I cannot find any related fixes in openafs-1.4.2-rhel4.1.src.rpm.

I guess that the 4k stack is your problem. Maybe you cannot reproduce
the crash using openafs-1.4.2-rhel4.1.src.rpm by some strange chance but
I think you should rather play it safe and try the 8k version.

I've had numerous stack overflows every few weeks on many systems using
different storage layouts (most were using XFS, some with LVM, some
using software raid). From the time when I switched to 8k kernels, all
stack overflows were gone.

Cheers,
--leo