ISC 2.2.1 Ethernet/networking induced panics?

[Rich Brennan]

I have an AMI 486/33 system (8 MHz bus) running ISC 2.2.1 that seems to panic
whenever I talk to it using its WD8003E card. The panic doesn't always
happen immediately, but usually within an hour or so. I'm pretty sure the
basic hardware is solid, as I ran the system for a week with 4 "makes" going
and a couple of "ls -lR"s and "find"s going to really load the system. During
the testing, I was at run level 3 (networking up), but no ethernet activity.

However, when I tried to use the PC-Interface remote drive from a DOS
machine to add some more load, the system paniced (trap D) within 15 minutes.
I rebooted, but left the system idle, and tried the PC-Interface drive again.
I paniced in 30 minutes.

Thinking it might be PC-Interface (if it has a kernel portion) I tried
plain old TCP/IP. I telneted into my machine, and did an endless "ls -lR"
loop. The machine stayed up for about an hour until I did a "netstat -i",
whereupon it promptly paniced (trap E).

The trap address is never consistent. In fact I once trapped in the kernel's
bss area. I usually get a trap D or trap E.

I believe my system is up-to-date, including the NT00001 network drivers.
I'm not sure where to look for trouble. I believe the WD8003E uses a shared
memory area to communicate with the O/S, so there should be no DMA glitches
to hammer me. I am running with a 1542B, but in the week long load test,
there were no problems there. This same problem occurs on an AMI 486/25
motherboard (using the same controllers/drives, etc), so unless its a
generic AMI problem (which I kind of doubt), I don't think it's my mother-
board.


Any suggestions? WD8003 swapout? 1542B swapout? Buy a RISC box?



Rich
-- 
brennan at merk.com	...!uunet!merk!brennan		Rich Brennan