In article <199810011829.LAA18985@lestat.nas.nasa.gov> you write:
>On Thu, 1 Oct 1998 10:48:45 -0700 (PDT) 
> Phil Nelson <phil@cs.wwu.edu> wrote:
>
> >   I'm just started to use the epic0 interface on one of my machines.  It
> > is only suppored in -current.  After using it for a few minutes, it
> > becomes unusable and the kernel continues to say "epic0: device timeout".
> > 
> > This is with -current built Tuesday, Sept 29.
> > 
> > Is anyone using these successfully?
>
>I have problems in 10mb mode, but it seems to work much better in 100mb
>mode...  I have heard from a few sources that the hardware has some
>annoying bugs (a few of which I try to work around in the driver).
>
>I would very much appreciate debugging help on this one... I am short on
>time to work on it.
>

We've had a number of machines come with this card, and I backported the
driver to 1.3.2 to support it (and along the way turned up a couple bugs...
see bin/5778 on sysinst and port-i386/6226 on a misleading panic message).

The timeout message causes all network activity on the machine to stall
for a few seconds, but otherwise doesn't seem to cause any loss of data
(in a pinch, we were running an nfs server with one of these cards...)

It seems to primarily occur when a new connection is opened to/from the
machine, as I've had no trouble streaming data at high speed, but having
another computer try to open a connection will suddenly lead to a timeout.

I would be willing to do some debugging work on the device timeouts.  I
don't know particularly much about driver internals, but I learn quickly.

jdarrow

--
John Darrow
Computing Services, Wheaton College, Wheaton, IL
John.P.Darrow@wheaton.edu