[tulip] Osicom 123400-21 40port card

Donald Becker becker@scyld.com
Wed Jul 31 14:17:01 2002

On Wed, 31 Jul 2002, Jason Marshall wrote:

> > The problem is that the driver uses the missing EEPROM on ports 1-3 to
> > trigger a work-around for a common BIOS bug: the BIOS does not correctly
> > record which devices are using which IRQs.

> Okay...  I confess that I'm not sure what that really means, but are you
> suggesting that putting the reverse-probe fix in place might do the
> trick?

No.  The IRQ fixup is triggered by subsequent ports not having an
It will take a minor source code change to trigger the fix-up with your

This was supposed to be handled in a generic way by having special "PCI
fixups" implemented in the 2.4 kernel.  That doesn't seem to have
happened, or it doesn't work in this case...

> > Have you tested all four ports?  If they all work, you don't have the
> > bug.
> You caught me.  I've assigned IP's to each port, but haven't even plugged
> them into the switch.  My bad *8-)

Note that a 21140 with a SYM transceiver cannot do autonegotiation.
This is expected from a 1996-era board
I'll repeat that:
       a 21140 with a SYM transceiver cannot do autonegotiation.

> is lacking; this level of service would easily cost $10,000/yr from Sun or
> HP!  Canadian dollars, but the bigger numbers look more impressive *8-)

You can purchase Scyld support through "the new HP", as well as
preconfigured clusters with the Scyld system installed.

Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993