On Sat, 5 Oct 1996, Jonathan Stone wrote:

> No, *not* everyone.  I have never, ever, ever seen anything like
> this,  with several 3c595s running on 100-TX.

Have you tried them at 10 Mbps? I have two 3c595s, and I see this
at 10 Mbps. I've not tried it at 100 Mbps, though I will soon.

Also, how fast are your machines? This possibly has something to
do with the queue of packets to send filling up. A 486 has enough
CPU power to overrun a 10 Mbps link, but in my tests with netperf
my P133 hardly reached 100 Mbps on the loopback interface, so it
could be that your machines are not pushing the cards hard enough
to exhibit the problem in 100 Mbps mode.

There was a recent patch to the FreeBSD driver to fix a problem
that sounds similar to ours; I'm looking into this now, and will
be playing with it once I get another machine upgraded to 1.2.
(Unfortunately, that may not be this week-end; the machine is at
the office, and I'm at home with a horrible cold.)

> 	* Is port-i386 really the right place for bugs with PCI boards?

I think it should probably be in -current, if anyone is using these
on other architectures. I follow that list, so I'll be just as
happy if others want to redirect it there.

cjs

Curt Sampson    curt@portal.ca		Info at http://www.portal.ca/
Internet Portal Services, Inc.	
Vancouver, BC   (604) 257-9400		De gustibus, aut bene aut nihil.