I did some more experiments regarding the
I tried (kernel/pcmcia) 2.2.14/1.3.9 and 2.2.14/1.3.10 with the same results.
I can insert the card and the network config is done.
I can ping & telnet without crashes. I can do a ftp and everything on the control connection works. If I do a download the system crashed without any messages printed anywhere (irq pending and not handled?).
I thought the packet size is triggering the bug. So I did a ping -s 1500. That works long enough to make experiment boring.
Next theory is the packet rate. What happens if new packets arrive while the previous is not transferred off the DEC-chip yet? I thought a flood ping might show this and it did. A ping -f kills the system after a while (60 seconds or so).
Please note that we are talking about a slow computer here. It is a 166MHz Cyrix MediaGX with only 55bogomips (that is 0.055 bogogips :-).
I will do any experiments that are suggested (well, the ones that seem useful - I will not stand on my head and hold my breath for 30 minutes). who is maintaining the tulip_cb module? I suspect the bug there. What can I do if I have a kernel that I know how to crash reproducable but that just enters an endless loop (irq condition not reset?) or locks the bus or whatever. How do I debug this? I would love some oopses, but I can't get any :-(
Many thanks, Rolf