SiI680 oops/panic

[Erik Bourget <erik@midmaine.com>]

(Cc to lkml in case anybody else has any intuition).

Hello Andre;

A while ago I spoke with you about my funky SiI680 experiences.  I have since
decided that my Hitachi "DeathStar" 180GXP harddrives were all faulty, and
replaced them with Western Digital..  This was bad.  On the other hand, the
kernel /DID/ oops/panic a few times, and I finally managed to drive to the
datacenter before somebody rebooted it.

And I took a shot with my trusty digital camera.
http://tacos.sus.mcgill.ca/~erik/oops-panic.jpg (24,457b) 
(textual bits have been re-typed below)

Kernel: 2.4.21, SMP (also happened on 2.4.22, also SMP)
(Note that the box has only one CPU, and no 'hyperthreading')

Text from it:

printing eip:
3d3d3d3d
*pde = 00000000
Oops: 0000
CPU:   0
EIP:  0010:[<3d3d3d3d>]  Not tainted
EFLAGS: 00010046
...
Process nfsd (pid: 200)
...
Code: Bad EIP value.
 <0>Kernel panic: Aiee, killing interrupt handler!
In interrupt handler - not syncing

Hardware:
  Dell "650" 1U server, P4 2.4GHz, 512MB, 2x120-GB Hitachi 180GXP DeskStar
  drives in RAID-1 configuration.

Software: vanilla Debian woody, vanilla kernel.org kernels.

Load on the drives was Constant and Extreme.  The machines serve as mail
storage for our ISP's mail system.  The drives were running a constant synch
process that checked a database to see which users had authenticated with the
mail system (via POP3, IMAP, etc - such that they might have deleted mail) and
ran rsync to an identical machine over a 1000mbit link on their directories.
Literally -

while(1) {
         @userlist = changed_directories();
         foreach (@userlist) {
                 do_rsync(localhost:$_, remotehost:$_);
         }
         sleep(30);
}

Can you make heads or tails of this?  My first thought is that some driver
isn't handling faulty hardware in an error-tolerant way.

Thanks for your time;

- Erik Bourget