2 eepros + erros

Gregory R. Warnes warnes@biostat.washington.edu
Tue Feb 2 16:54:46 1999


OK, this change removed the log messages.  However, I still get recieved
lockups.  Now they do not get reported in the logs!

Does anyone know how to force my two eepro's to use different interrupts?
I'm wondering if this would help.

-Greg


On Tue, 26 Jan 1999, Osma Ahvenlampi wrote:

> "Gregory R. Warnes" wrote:
> > 
> > I've got 2 eepro100 cards in my Dual P-II.  Onne is connected to the
> > outside world by a 100MBS switch.  The other is connected to the rest of
> > the beowulf via  another 100MBS switch. I'm getting rather consistent
> > problems with the card to the external world failing.  My logs contan the
> > following messages, repeated ad nausium:
> > 
> > Jan 24 04:02:08 queenbee kernel: eth0: Trying to restart the transmitter...
> > Jan 24 04:02:08 queenbee kernel: eth1: Transmit timed out: status 7048 0000 at
> > 994989/995004 command 000ca000.
> 
> you're running into the multicast filter bug. Add 
> 
> options eepro100 multicast_filter_limit=3
> 
> to your /etc/conf.modules to cure it.
> 
> > Oth thing that consudes me is that one message says eth0 and the other
> > eth1.  Is the driver getting confused?
> 
> That is curious. The shared interrupt and this inconsistency might also
> mean that the above multicast filter limit would not be enough, and the
> driver screws up if it's in the middle of a reconfigure operation on one
> card when it receives a packet on the other. The problem certainly seems
> like a driver re-entrancy bug on systems with one interface.
> 
> -- osma
>