Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!elroy.jpl.nasa.gov!turnkey!orchard.la.locus.com!prodnet.la.locus.com!electra.la.locus.com!brian
From: brian@electra.la.locus.com (Brian D. Horn)
Newsgroups: comp.unix.aix
Subject: Re: Ethernet adapter problems
Message-ID: <18963@oolong.la.locus.com>
Date: 31 Oct 90 15:57:58 GMT
References: <16128@hydra.gatech.EDU>
Sender: news@locus.com
Organization: Locus Computing Corporation, Inglewood, CA
Lines: 16

In article <16128@hydra.gatech.EDU> scott@prism.gatech.EDU (Scott Holt) writes:
>
>We have a number of RS 6000s connected to our campus network -
>a network which sees quite a bit of traffic. From time to time,
>the ethernet interfaces in the machines will simply shut off.

	Note: I haven't seen this on a 6000, but odds are good that what
I have to say follows.

	On EVERY piece of hardware using the Intel 82586 I have found
that the chip gets itself wedged into an "impossible" state under conditions
of heavy traffic, especially broadcast traffic.  I have seen this in
at LEAST 5 different hardware platforms INCLUDING one built by Intel.
The only solution I have seen is having the device driver have a watchdog
timer check for such a condition and pull on the reset line if and when
the condition is detected and restart the I/O.