Well, I got only 3 replies to my last query, 2 of them wanting to know
what I've found, so I guess I'm not the only one with the problem.
Richard Jackson verified that the "transmit fifo underflow" messages are
normal.
Since then I've changed my NIC to a new one:
tu1: DECchip 21143: Revision: 3.0
instead of the old:
tu1: DECchip 21140: Revision: 1.2
Now the machine doesn't hang anymore when the NIC dies. I could
ifconfig the device up and down, change speeds etc. I rebooted
the switch. Nothing but a boot helped. So it seems there is somekind
of a bug in the de500xa driver.
So:
- is there any way to re-initialize the NIC?
- any kind of patches around? (the 4.0D megapatch didn't mention
de500-xa)
- any other pointers.... (I'm quite desperate by now...)
One additional note: In my logs named starts whining about buffer space
shortly
after the network dropped:
13-Jun-18:37/daemon.log:Jun 14 04:10:02 tahma named[366]: sysquery:
sendto([192.98.7.241].53): No buffer space available
Could it be that the kernel runs out of bufferspace for the network? In
that case,
which values should I adjust?
Regards, Harald
Received on Mon Jun 15 1998 - 13:39:28 NZST