[tulip] Problems with two tulip based cards

[Simon Hill]

Hello,

I'm having probems with two tulip-based interfaces under RH6.2 (w/2.2.16
patched to fix EV4 FP trap bug) on a UDB. I previously had both working
under RH6.1 (stock kernel), using IIRC something like this:

alias eth0 tulip
alias eth1 tulip
options eth0 -o tulip-0 io=0x8800 irq=10
options eth0 -o tulip-1 io=0x9000 irq=15

/proc/pci shows (summarizing since I can't get to the box over the net):
DEC DC21040 (rev 35) IRQ 10 IO 0x8800 (onboard interface)
DEC DC21041 (rev 33) IRQ 15 IO 0x9000 (Kingston pci card)

Now it appears that the driver (I've tried tulip, old_tulip, and de4x5),
detects both adapters at once, and no longer accepts io or irq settings.
If I try to set them, I get 'invalid parameter parm_io' or 'invalid
parameter parm_irq'. If I let the driver autodetect them, one interface
works fine, the other doesn't. Specifically, either interface will work
when configured for my local network connected to a 10Mb hub, neither
interface will work when configured for my external network and connected
to the LinkSys 10/100 hub (and yes, I've reset the hub, rebooted the
machine, etc). However, when the second adapter is removed, the onboard
one will talk to the LinkSys 10/100 hub just fine.

The characteristics of the 'not working' state are:

ping from another system on the external network -> UDB: works
ping from UDB to other system: doesn't work, summary claims 1 success.

Anyone have any ideas about what the problem might be?

TIA,

Simon Hill (simon@metasystema.org)
http://www.metasystema.org/