> I, btw, have used the kernel version from the CVS (except Miguel's
> console changes for X) for several days now without major problems
> on a Indy. But I still haven't found where all the bad packets
> that my other machines spit out come from. They're pretty annoying
> as they result in NFS timeouts, sometimes even lockups.
> It's the sgiseeq driver, if it gets hit just right the receiver will
> lock up. I had a work around that would detect the situation and dump
> the rx/tx rings, then reset the card, and it would continue to work.
> Put some timeout in there to do something like that, this will help
> you to track down this bug which I never was able to resolve.
I've seen your workaround on the Indy being triggered but sine I fixed
one of the other not directly networking related bugs everything seems
to work fine. I had the machine up for more than two days and not a
single NFS timeout. Thomas Bogendoerfer reports the same for a M700
using a NE2000. But on the other machines I've tested (P4032, RM200
both using onboard ethernet) this bug strikes very often. The fact
that Thomas can use networking without that problem also eleminates
another byteorder problem as cause.