[OpenAFS] client problems

Derrick J Brashear shadow@dementia.org
Wed, 11 Apr 2001 13:21:18 -0400 (EDT)

On Wed, 11 Apr 2001, Steven N. Hirsch wrote:

> > Don't apologize, it's a useful data point. Your earlier message suggested 
> > you were using a mish-mash of patches, and the problem is AFS is at least a 
> > little fragile with respect to kernel structures, so if those patches 
> > modify any of the relevant stuff it can be problematic; It's obviously not 
> > possible to test with every combination of patches, but I will begin 
> > testing with 2.2.19 as soon as I've isolated this SMP problem, and I'm on 
> > the trail of it now.
> Fair enough, thanks!  Thinking about it, Stephen Tweedie's ext3fs patches
> would be my prime suspect.  They affect filesystem code at a very low
> level AFAIK.  I'll give this another try without them applied, using the
> OpenAFS afsd (never heard any opinions on whether my use of the Transarc
> 3.6 afsd with OpenAFS kernel module could have been problematic).

You could probably use the AFS 3.3 afsd with the OpenAFS kernel module, if
you could get it to build. afsd is basically just a small program which
calls into the kernel module to establish the necessary contexts in the
kernel, it's basically not relevant. Anything interesting change-wise
would be to the cache manager and hence to libafs, not to afsd.