> I inserted some debug printfs to see what was happening in the 
> mhzc_alloc_ethernet for loop.  It seems to find a good portion of the 
> address without the 0x80 bit set, but none of those have a 
> successful pcmcia_io_alloc.

Ok, so there must be something else colliding at that "good" adresses it 
tries. Sorry, I don't have the dmesg you send anymore :-(

For any addresses it tries to do a pcmcia_io_alloc: there must be some
other device already mapped at that address - or pcmcica_io_alloc is broken
somehow. If you can hookup kernel gdb (or understand enough assembly to 
follow ddb through this) it might be easiest to set a breakpoint on that
loop and step through pcmcia_io_alloc to find out why it fails.

Or just insert some mor debug printfs in there (it's actually to be found
as function pccbb_pcmcia_io_alloc in file sys/dev/pci/pccbb.c when your
pcmcia/cardbus is behind a PCI<->cardbus/pcmcia bridge). I can't see anything
obvious...


Martin