GA-7VAXP + 2.6.0-testX weirdness

[Peter Lieverdink]

Over the past few weeks I've been trying to play with cryptoloop on my 
relatively recent Athlon acquisition, which is plugged into a Gigabyte 
GA-7VAXP mainboard (KT400-based).

When compiling 2.6.0-test6 and higher (including -mm patches) I get rather 
nasty looking crashes when I attempt to make a filesystem on a cryptoloop 
device. On advice from Andrew Morton I've tried a few different configs 
(build for K7, build for 386, use gcc-3.3, use gcc-2.95 and combinations 
thereof) but I reliably keep experiencing crashes at the same point, 
although depending on the build target I get a different trace.

I have managed to get someone with the same CPU to compile a kernel using 
my config and they did NOT experience the crash, I have no idea what 
mainboard they had, though.

When compiling the same source tree using the same config (attached) and 
same compiler/build tools on a different machine (Celeron 500) the kernel 
does NOT crash when run on the Athlon.

I've spoken to one person online who has the same mainboard and is also 
experiencing odd crashes with 2.6.0-test8.

If you have a GA-7VAXP mainboard, would you be able to check if using 
cryptoloop as per attached output causes your kernel to crash?

Additionally, if anyone who has a look at the output has any suggestion as 
to what may cause this behaviour (hardware/software) and any suggested 
remedies, please let me know. I'd be quite happy to rebuild kernels for a 
day to solve this, if it's a compiler or kernel issue.

- Peter.

Attached:
  .config (config-2.6.0-test8.gz)
  crash output (cryptoloop.txt.gz)

Build info:
  Debian unstable (Sid)
  gcc-3.3.1-2
  libc6-2.3.2-8
  binutils-2.14.90.0.6-5
  make-3.80-4

Attachment: cryptoloop.txt.gz
Description: Binary data

Attachment: config-2.6.0-test8.gz
Description: Binary data