Second Ethernet interface

[mikebw at bilow.bilow.uu.ids.net (Mike Bilow)]

Chuck Young wrote in a message to Mike Bilow:

 CY> I apologize if this has been discussed, but are the two
 CY> ethernet cards the SAME.  Are they both 3C509's?

 CY> If so, I have done it by using the config utility in DOS
 CY> realmode (F8 as Mike pointed out) and assigning them each a
 CY> unique I/O and IRQ (to the system as well as the cards
 CY> themselves) and writing the info down.  Then, boot the
 CY> machine into Linux and make sure it sees one of the cards
 CY> and calls it eth0.  Then append the parameters calling the
 CY> second card eth1 to lilo.conf, re-install Lilo and reboot
 CY> into linux.  When the kernel gets the new parameters, it
 CY> loads the devices and you see the messages on the console
 CY> during boot-up.  With two cards, you have to pass
 CY> parameters.   

 CY> It worked well for me...

Well, yes and no.  First, if the drivers are modularized, then you can pass
arguments to them at load time rather than at boot time.  If the drivers are
embedded in the kernel, you do have to go through Lilo.  I recommend using the
drivers in modularized format.

Second, 3Com cards have the interesting property that their configuration
registers are always conveniently accessible.  As a result, they can be probed
by a driver much more reliably than, say, and NE2000 clone card.  In fact,
there is no safe way to probe for an NE2000, and you always risk crashing the
whole system.  With 3Com cards, you have much more reliable detection.

If you have all 3Com cards, then they should be automatically detected by the
driver and should be assigned device labels ("eth0", "eth1", etc.) in the order
in which the probe is carried out among base addresses.  Each card must have a
unique base address and IRQ, of course.
 
-- Mike