On Fri, 24 Sep 1999, Matthias Drochner wrote:

> 
> blackbox@openface.ca said:
> > I can't seem to figure out how to track down the cause of this
> > problem, or how to get rid of it. 
> 
> Did you succeed in increasing the receive ring?

Modifying source parameters on critical machines scares me, especially when
preceeded by the disclaimer "who knows how this will affect the rest of the
system...".  :)  I haven't tried it yet.  I still can though, if you think it
will help.  The machine isn't yet in critical operation.

> What about the interrupt sharing? Can you try
> to rearrange things so that the card gets an own
> interrupt (I know it's a PC...), or at least give
> some data points?

I'll try a few reboots tonight to see if I can make any such determinations.

> Does it happen on all 3com cards on on one only?

It happens on my "external" card, regardless of which card that is.  I have
three cards, ex0, ex1, ex2.  ex2 is my "internal" card, ex1 is my "external"
card.  I've tried on ex0, and ex1 only so far.

I just realized that the ethernet cable in question is a crossover cable that
goes straight into a Cisco 7206.  Maybe that has something to do with it.  
The problem seems to be on transmit (ex1) only.  The ex2 card is receiving
only, and the ex1 card is transmitting only.  I have an asymmetric routing
situation where I'm only sending outbound traffic to the 7206 and the incoming
stuff is coming into my network on entirely different equipment/connections.

Hmm.

Must investigate further.

Erik

> > When it does happen, everything freezes for about
> > 5 seconds preceeding the message (I guess thats the Ethernet card
> > reinitializing?)
> 
> Strange that this takes so long. Perhaps the hardware
> hangs up completely.
> 
> > Maybe I should just switch to other cards?  Which cards are easily available
> > and known to be good?  I've heard good things about the fxp/EtherExpress Pro
> > cards.
> 
> I can't give a serious advice, but I'm using boards
> with the Intel chip in 6 or 8 boxes and didn't have
> problems so far. The DEC/Tulip chips are potentially
> better (a bit faster and well documented), but the
> '143 is not well supported by NetBSD, and clone chips
> seem to be of varying quality, su you'd have to watch
> out what you buy.
> But as I said, I don't have problems with 3c90xb cards
> either, even in data acquisition applications.