I was running into the same problems here; SMP system w/PCI riser card
(HP NetServer LPr), connected to a 10/100 switch. I'd get
"wait_for_command_timeout" errors all the time under moderate network
load. Switching to the e100 driver didn't help in the slightest.
Eventually, I'd experience a complete system lockup.
Replacing the card with a 3c59x-based card put the machine back in
service (I've completely written eepro100s off as a viable cards now),
although I still saw occasional PCI-related issues. Specifically:
Jan 23 10:11:37 x kernel: Uhhuh. NMI received. Dazed and confused, but
trying to continue
Jan 23 10:11:37 x kernel: eth0: Host error, FIFO diagnostic register
0000.
Jan 23 10:11:37 x kernel: eth0: PCI bus error, bus status 80000020
Jan 23 10:11:37 x kernel: You probably have a hardware problem with your
RAM chips
Jan 23 10:11:37 x kernel: eth0: Host error, FIFO diagnostic register
0000.
Jan 23 10:11:37 x kernel: eth0: PCI bus error, bus status 80000020
The last two messages will repeat indefinitely, usually with a hit to
the dist for each pair of log entries (resulting in a very distinctive
drive grinding). Memory problems don't seem to be the issue; with a
fairly extensive run of memtest86, everything came back clean.
Taking a few minutes to try and rectify the situation, I started
shutting down services and manually unloading modules to see what was
causing the problem. Unloading usbcore did the trick:
Jan 26 18:41:24 x kernel: eth0: Host error, FIFO diagnostic register
0000.
Jan 26 18:41:24 x kernel: eth0: PCI bus error, bus status 80000020
Jan 26 18:41:24 x kernel: eth0: Too much work in interrupt, status e003.
Jan 26 18:41:24 x kernel: usb.c: USB disconnect on device 1
Jan 26 18:41:24 x kernel: USB bus 1 deregistered
I've rebooted the machine since then, but have always unloaded usb-uhci
and usbcore after booting. The issue hasn't cropped up again, although
it happened every couple of days previously.
The kernel in question is Red Hat's kernel-smp-2.4.9-21 build.
-- Edward S. Marshall <esm@logic.net> http://esm.logic.net/ ------------------------------------------------------------------------------- [ Felix qui potuit rerum cognoscere causas. ]- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/