> Their all PCI, all use tulip.o,
>
> I've tried mixing the nic's around to see if there is order to this, oddly
> regardless of what slot the 10bT nic is in it is always comes up eth2.
> While the other two can be interchanged. Though this motherboard seems
to
> assign eth0 starting on the nic furthest from the io ports. Could this be
> an ordering issue inside the driver? the 10bT probably uses a earlier
> variation than the other two.
>
> Just in case it matters...
> Bering 1.0 rc2
> 2x SMC 1255TX (10/100)
> 1x D-Link 530 CT (10bT) has digital 21041-PB
> Motherboard ASUS TXP4, 32MB ram, P166
ethX to physical card mapping can get to be a bit confusing. Basically,
there are two layers that affect this mapping. The first is the module load
order. If you have cards that use different driver modules, the driver
module loaded first will assign the lower ethernet device numbers.
Once a module is loaded, it is the responsability of *THAT MODULE* to assign
device numbers to the cards that module is talking to. The device ordering
of multiple cards supported by a single driver is *DRIVER DEPENDENT*.
ISA drivers typically assign device numbers in the order you pass the
physical I/O address on the command line, except for some wacky 3-Com cards,
which have a built-in "plug-n-play" type auto-detect feature, which causes
cards to get device numbers assigned in order of MAC address.
PCI drivers typically assign device numbers in the order the cards are found
on the PCI bus, which is motherboard and chipset dependant, but normally
starts at one end of the PCI bus and goes to the other. In your case using
the tulip driver, you're also using cards with multiple PCI device/vendor
ID's. In this case, it would appear the tulip driver is doing something
like:
for each CardID in supported device/vendor ID
FoundCards = Scan PCI bus for CardID
for each Card in FoundCards
assign device ID to Card
done
done
Thus, the two matching cards follow the PCI bus scanning order when getting
assigned ethernet numbers, but the 'odd man out' is always seen last...
Of course for the final word on how multiple cards will be handled, you'll
have to look at the code...
HTH,
Charles Steinkuehler
http://lrp.steinkuehler.net
http://c0wz.steinkuehler.net (lrp.c0wz.com mirror)
