[vortex] One more 3c905 problem :(

[Donald Becker]

On Fri, 15 Nov 2002, David Nordenberg wrote:

> I guess my problem is related with this
> http://www.scyld.com/pipermail/vortex/2002-November/002013.html.

Nope, this is very likely unrelated.

> I have my roter/firewall, it has eth0 and that is a 3c509b and a eth1 as
> internal interface and that is a 3c905C-TX-M. It has worked fine for a long
> time with kernel 2.4.19 and then 2.4.20pre7 and for the last weeks,
> 2.4.20pre11. It have been connected to a old 10mbit hub.
...
> Today I invested in
> a new "FSW-2108TX" switch from "level one". I connected the network cable to
> the switch and click, the 100mbit and full duplex LEDs lights up on the
> switch. I'm very happy for my new switch :) After a few hours I noticed my
> internet connection stopped working and I found the link LEDs on the switch
> black :( I tried to revive the link with "ifconfig down" and start it again
> but now it is dead. I rebooted and the LED is still black. I turned the
> computer off for a few seconds and then back on, now the link LED is on
> again and everything worked normal... a few hours later... no internet again
...
> First death of eth1: http://217.210.183.83/eth1_dies.log
...
> I hardly believe that this is enough info for you to fix the problem 

This report makes it very clear what has happened: the NIC has stopped
responding.

My guess:
Your card is overheating or overloading the power supply
because of the added workload of 100baseTx, and shutting down.  You are
reviving it by power cycling the machines.

If it still responds to PCI configuration space accesses, you can try to
revive it in software by using 'pci-config' to put it into D3-warm power
down state and then bringing it back to D0 state.   But the real
solution is to improve the power supply or cooling.

-- 
Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Scyld Beowulf cluster system
Annapolis MD 21403			410-990-9993