Hi,
I'm trying to run some netperf tests on linux 2.4.0 kernel with smp turned
on
between uniproc pc's as well as netfinity 8 proc and 2 proc boxes.
Distribution is Red Hat 7.0.
I've been experiencing ethernet card lockups in 100Mb mode.
No heavy load - just one connection sending out 4096 sized
packets, for example - and it usually takes only about 3 to 4 mins to
run into the problem.
This happens with all chips and drivers I've tried so far:
- FA311/312 - (natsemi.o)
- DECchip 21140 - (tulip.o)
- AMD 79C970 - (pcnet32.o)
Example: I see this in /var/log/messages: (from the natsemi driver:)
Feb 12 14:10:57 w kernel: NETDEV WATCHDOG: eth1: transmit timed out
Feb 12 14:10:57 w kernel: eth1: Transmit timed out, status 03fbab90,
resetting...
Card times out on a transmit and the watchdog timer complains. However,
the card never resets and fails to get any further interrupts. Bouncing the
interface succeeds in getting things kosher again.
I've seen reports of this problem on the mailing list but no patch or
workaround:
Can anyone suggest something that I've probably missed?
thanks,
Nivedita
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
