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

[Deon George]

Axel Thimm wrote:
> 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?
>   
Maybe sombody can comment about the src.rpms? From what I can see, the 
rhel4 src.rpm does have a couple of patches - one to rename the kernel 
module from libafs to openafs, and some (5) labeled as Krb5 migration 
kit patches - I dont know if they are involved with the kernel module?

I woundnt have thought that the source tar file would be different to 
the pristine release?

> But I don't think a rebuild would yield different results.
>   
Maybe we are using different compile options (or something)? Sorry, I'm 
not a C/C++ developer (or whatever the language this is written in) - I 
know how to hack things to work, but beyond that its a different world :)

I've just found the patches in RT about RHEL5 beta2 not compiling, and 
I've used them, got it compiled and now I'm testing it... So far it is 
looking good :) So here's what I've done:

Taken the 1.4.2.rhel4.1.src.rpm, swapped out the 1.4.2 source for the 
1.4.3rc2 source.
Added the patches from RT
#54916    1.4.3rc2 on RHEL5b2: tasklist_lock test does not work
#54922    1.4.3rc2: src/rx/LINUX/rx_knet.c still using tasklist_lock

And compiled using "rpmbuild -ba openafs.spec --defined 'build_modules 1'"

Bugger, before finishing this message, I just noticed my machine just 
paniced - it was looking like it was going to run fine - and no sysreq 
working :(

...deon