Debugging kernel lock-ups during network activity

[Art Haas]

Hi.

I've got Rawhide running on an SMP PentiunIII machine, and up until the
post 2.6.17 kernel things had been very stable on the machine. The
development kernels leading to 2.6.18 have not been nearly as reliable
as many others on the list have observed. In my particular case,
the machine continues to lock up during network activity as that done
for 'yum update' or browsing with Firefox.

What I'm vexed about is how to try and debug the problem I'm seeing. A
bugzilla report saying 'Kernel locks up during network activity' is just
to vague to be of any use. Also, the lock ups are not consistent - I can
do somethings like 'git pull' or updates with cvs/svn and more often
than not the machine hums along. But then that one event happens and
things lock up solid requiring a reboot. The 'yum update' I try and do
daily is almost always guaranteed to cause a lockup at least one out of
two times I try to get the latest packages.

In an effort to try and get the latest Rawhide packages, this morning I
booted off an FC5 install disc with 'linux rescue', and with a bit of
chroot magic I was able to get many of the updates available this
morning with some exceptions due to dependency issues which I'm sure
will be resolved shortly. I'd also run 'memtest86' on the machine for a
day or two and it ran without errors.

Given my description above, just what sort of steps can people recommend
so that I can try and narrow down the problem? I'd like to try and dig
in a bit so that I can possibly fix things or at least make a useful bug
report. I'd hoped that the post 2.6.18-rc2 kernels would have fixed
things as a variety of networking bugs have been patched but that has
not happened. Suggestions welcomed.

Art Haas
-- 
Man once surrendering his reason, has no remaining guard against absurdities
the most monstrous, and like a ship without rudder, is the sport of every wind.

-Thomas Jefferson to James Smith, 1822