Spurious Parity Panics w/SCSI under ISC 2.2

[Brad Karp]

I recently upgraded my system from using an old MFM ("standard AT," in ISC's
lingo) hard disk controller and hard disk to using a SCSI controller and hard
disk.

My SCSI controller is the Adaptec AHA-1542B bus mastering controller, and I
have a Maxtor XT-8760S SCSI drive and an Archive 2150s tape drive attached
to it. The controller has SCSI ID 7, the Maxtor has SCSI ID 0, and the
Archive has SCSI ID 2.

I successfully reinstalled ISC UNIX 2.2 on my system, and then took care to
configure the HPDD appropriately to match my AHA-1542B configuration (Int 11,
one Archive tape drive, no secondary controllers, primary SCSI controller).

My AHA-1542B is jumpered to use DMA REQ 5 and DMA ACK 5, and IRQ 11. I was
very happy with the system; the drives and controller are wondrous quick, and
having so much space was almost intoxicating. :-)

Unfortunately, my joy was short-lived. I've discovered that every two or three
days, while in the midst of intense disk activity (e.g. running a du on a large
subdir tree, doing rm *.o in a directory containing lots of .o files), my
system halts with a message similar to the following from the kernel on the
console:

FATAL:
PANIC: Parity error on motherboard, address <blah blah>.
Trying to dump 1776 pages ....

Press any key to reboot.

Needless to say, such behavior from my system is unacceptable. I suspect foul
play from my disk subsystem because the memory address given by this message
is different every time, my motherboard tests its own memory to be OK when my
machine powers on, and because the only candidate left seems to be the DMA
that my AHA-1542B must do during heavy disk activity.

My system contains the following other devices, _none_ of which do DMA to the
best of my knowledge: Paradise VGA 1024, Logitech Bus Mouse adapter, 2 serial/
1 parallel port adapter. I have 7 MB of memory, and my motherboard is a
Hauppauge Computer Works 386-AT/20.

I noticed a note in the release notes for ISC UNIX 2.2 that there have been
reports of kernel panics during heavy disk activity with the SCSI drivers. Has
anyone else out there had an experience like this one? Is there a fix?
Thanks.

Replies by email preferred; I will summarize.
-- 
Brad Karp, (203) 436-3060 (voice)          | The views expressed in the text
my 386: karp%softshop.uucp at cs.yale.edu     | you have just perused are not my
via Yale CS Dept: karp-brad at cs.yale.edu    | own; rather, they are those of the 
Box 2443 Yale Station, New Haven, CT 06520 | heavenly muse who sings in me. -JM