I just found a fatal flaw in the tulip driver for the LC82C168 (pnic) chipset: The driver assumes that CSR11 is a hardware timer, but the LC82C168 doesn't have that timer :-( Thus no callback, and nic is hung. Jeff, could you include the patch into your tree and forward it to Alan? The patch is against 2.4.1, it fixes all pnic bugs I'm aware of: - missing update of the FastEthernet bit in check_duplex() - do not start the internal autonegotiation if an mii is connected, that prevents successful communication with the mii. - hw timer doesn't exist -- Manfred