machine crashes with kernel panic: fatal exception in interrupt

[Karen Larson <larson@cs.uwp.edu>]

A dual CPU Intel machine was installed with Debian testing on 1/14/05 
with a 2.6.8-1-686-smp kernel.  Security updates were applied over the 
next year as they became available.  Machine was backed up to an 
unpartitioned USB external hard drive at /dev/sdb with a tar command 
(similar to backing up to tape). 

On 1/9/06, in addition to applying security patches, I added the lines 
for sarge main in the sources.list file and upgraded (aptitude upgrade) 
43 packages to turn it into a stable Sarge.  The kernel-image was 
changed to 2.6.8-2-686-smp.  Since that date, the machine crashes while 
backing up.  The backup may run for anywhere from 25 minutes to an hour 
before the machine crashes.  Different USB harddrives have all crashed 
during backups.  Booting back into the old kernel did not make a 
difference -- machine still crashed during backup.

The call trace in the most recent messages displayed to console include:
   scan_async
   echi-work
   echi_irq
   usb__hcd_irq
   handle_IRQ
   do_IRQ
   common_interrupt
   default_idle
   default_idle
   cpu_idle
   printk
   print_cpu_info

earlier ones also had:
   echi_watchdog
   run_timer_softirq
   do_softirq
   smp_apic_timer_interrupt
   apic_timer_interrupt
before the default_idle

Code and final messages are always the same
Code : 89 50 04 89 02 c7 41 04 00 02 20 00 c7 46 38 00 01 10 00 8b

<0> Kernel Panic: fatal exception in interrupt
in interrupt handler - not syncing

Any suggestions for diagnosing and solving this problem would be greatly 
appreciated.  Additional information can be provided.

Thanks,
Karen