> If I remove the card, my machine freezes instantly. This is due to a
> stupid dev->poll function of the tulip driver.
> drivers/net/tulip/interrupt.c:tulip_poll() gets stuck in an endless loop
> in interrupt context if the hardware returns 0xffffffff on certain reads.
> But this is exactly what happens if you remove a pci device.
> My patch replaces the deadlock with something resembling a livelock. At
> least SysRq-S works now because we leave the poll function after some time.
Could you explicitely check for read returning 0xffffffff?
64 bytes from 220.127.116.11: icmp_seq=28 ttl=51 time=448769.1 ms