Christos:

> Our netstat needs fixing :-) since it appears to report the same stats for
> all protocols. So it looks like FreeBSD's sees the errors too, but does not
> report them. What are the errors our driver produces?

The FreeBSD reports the server not responding/is alive again with usually
very little delay (sometimes there are a few seconds, the most being about
10 seconds).

On the NetBSD side I see:

Jul 14 20:08:53 vinland /netbsd: fxp0: device timeout
Jul 14 22:19:17 vinland /netbsd: fxp0: WARNING: SCB timed out!
Jul 14 22:19:29 vinland /netbsd: fxp0: device timeout
Jul 14 22:19:35 vinland /netbsd: fxp0: WARNING: SCB timed out!
Jul 14 22:19:42 vinland /netbsd: fxp0: device timeout
Jul 14 22:19:46 vinland /netbsd: fxp0: WARNING: SCB timed out!
Jul 14 22:19:49 vinland /netbsd: fxp0: WARNING: SCB timed out!
Jul 14 22:19:53 vinland /netbsd: fxp0: device timeout
Jul 14 22:19:57 vinland /netbsd: fxp0: WARNING: SCB timed out!
Jul 14 22:20:00 vinland /netbsd: fxp0: WARNING: SCB timed out!
Jul 14 22:20:04 vinland /netbsd: fxp0: device timeout

As well as a few nfs server not responding/alive (but those are probably
a result of the above).

> Are you sure you are not having some hub/switch problem?

I suppose that is possible.  We're using two cisco Etherswitches with
one interface being a front end, the other a back end (unrouted network).
At the moment the CS20s are linked on their fxp0's with a cross over cable.
This includes the time from the messages above and when both CS20s were
running FreeBSD 5.2.1.  Whats interesting is when they were both running
FreeBSD I did not see any nfs complaints .. so I believe that the time outs
are probably due to the fxp0 resets on the NetBSD side.