This makes tulip.c handle multiport cards (like the Cogent 964) work
correctly with most BIOSes. I think I've seen a more sophisticated patch
for this problem already on linux-kernel, but that one also didn't made it
into the kernel, so I give it another try.
I think I've also found an important bug in tulip_probe(). It will only
work correctly if there's only one kind of chips in your system (21040 or
21041 or 21140 or ...). This is due a wrong arrangement of the pci_index++
and cno++ loops. They should be exchanged.
Can anyone confirm this behaviour?
Linus, should I prepare a patch for you?