Panics under kernel 2.4.0-test8

[Craig McPherson <craig@bsu-hog.org>]

(I know I should probably ask this on one of the kernel mailing lists,
but you folks are nicer, and smarter.  Please CC me if you reply to
this, because the list server won't seem to let me resubscribe right
now.)

I have one old Debian box running up-to-date Woody with kernel 2.4.0-test6.
I have a new Debian box that I set up two days ago, that's going to
replace the old box, also running up-to-date Wooody, but I have a problem
with it.

For the rest of the e-mail, I'll call the systems OLD and NEW.
OLD has a rtl8139 network card on a 100BaseTX network.  NEW also
has a rtl8139 network card (the model name is "SMC EZ-CARD" --
I've learned to avoid anything with "EZ" in the name, but I didn't
have a choice in this case); actually, it has two rtl8139 network
cards (one for the DSL interface), but the problem existed before
I put the second card in.

Running Debian's stock 2.2.16 kernel on NEW, everything works fine.
However, under 2.4.0-test8, I get regular kernel panics and reboots
whenever there's heavy traffic on the local network.  The following
things will all cause NEW to reboot:

1.  Ping-flooding any machine on the network from NEW
2.  Ping-flooding NEW from any other machine on the net
3.  Telnetting from NEW to OLD and doing anything that produces a lot
of text, like "find / *"
4.  Moving a large file between OLD and NEW via FTP or SMB.

1-3 will crash the machine in seconds, 4 is a toss-up -- if the file
transfer takes more than 5 seconds, NEW dies before it comples; with
small files it varies.  Even a simple continuous ping involving NEW
(in either direction) will crash it eventually, although it might
take hours.

I'd like to say again that everything works fine under 2.2.16.  I
ping-flooded it to hades and back, and it didn't miss a beat.  I've
recompiled 2.4.0-test8 numerous times with numerous options, but
I still haven't had any luck.  It always crashes.

Here's the kernel panic message (sometimes it just reboots instead
of showing me this, though):

Oops!
(a whole lot of hex dump stuff)
Aieeee, killing interrupt handler!
Kernel panic: attempted to kill the idle task.
Interrupt handler - not syncing.

Here's my kernel configuration:

http://www.bsu-hog.org/config
(If you can't access that site... well, she's crashed again)

I know that the real "solution" would be to go back to 2.2.X, or an
earlier version of 2.4.0-test, but there are several reasons why
that would not be convenient:  my firewall script is written in
iptables format, and it would take a lot of work to convert it back
to ipchains, every partition except my root partition is using
ReiserFS and I'd have to backup and re-create them (going back to
an earlier kernel version would require a ReiserFS downgrade as well),
I just got a USB UPS I want to play with, I haven't checked to see
if 2.2.X supports my UDMA chipset, and I'm using devfs and am not sure
if things would break if I stopped using it.

Thanks for any insight.  (Don't forget to CC me, please)

-- 
Craig McPherson
Network Admin, Baptist Student Union
Fayetteville, Arkansas
craig@bsu-hog.org