I've had this happen twice now, running Andrew's "not quite -mm2" patch.
Symptoms: After about 20-30 minutes uptime, a running gkrellm shows system mode
suddenly shoot up to 99-100%, and the keyboard dies. Oddly enough, a USB mouse
continued working, and the X server was still quite responsive (I was able to
close Firefox by opening a menu with the mouse and selecting 'quit', for
example).
alt-sysrq-foo still worked, but ctl-alt-N to switch virtual consoles didn't.
sysrq-t produced a trace with nothing obviously odd - klogd, syslog, and the
disk were all working.
Nothing interesting in the syslog - no oops, bug, etc..
Another odd data point (I didn't notice if this part happened the first time):
gkrellm reported that link ppp0 had inbound packets on the modem port of a
Xircom ethernet/modem combo card. At the rate of 3.5M/second - a neat trick
for a 56K modem. When I unplugged the RJ-11, gkrellm *kept* reporting the
inbound traffic. When I ejected the card, *then* the ppp0 (and the alleged
inbound packets) stopped - but still sitting at 99% system and no keyboard.
This ring any bells? Any suggestions for instrumentation to help debug this?
On Wednesday 21 September 2005 22:35, [email protected] wrote:
> I've had this happen twice now, running Andrew's "not quite -mm2" patch.
>
> Symptoms: After about 20-30 minutes uptime, a running gkrellm shows system mode
> suddenly shoot up to 99-100%, and the keyboard dies. Oddly enough, a USB mouse
> continued working, and the X server was still quite responsive (I was able to
> close Firefox by opening a menu with the mouse and selecting 'quit', for
> example).
>
> alt-sysrq-foo still worked, but ctl-alt-N to switch virtual consoles didn't.
> sysrq-t produced a trace with nothing obviously odd - klogd, syslog, and the
> disk were all working.
>
> Nothing interesting in the syslog - no oops, bug, etc..
>
> Another odd data point (I didn't notice if this part happened the first time):
> gkrellm reported that link ppp0 had inbound packets on the modem port of a
> Xircom ethernet/modem combo card. At the rate of 3.5M/second - a neat trick
> for a 56K modem. When I unplugged the RJ-11, gkrellm *kept* reporting the
> inbound traffic. When I ejected the card, *then* the ppp0 (and the alleged
> inbound packets) stopped - but still sitting at 99% system and no keyboard.
>
> This ring any bells? Any suggestions for instrumentation to help debug this?
>
I have seen this couple of times when I would eject my prism54 card at
a "bad" time - my card sometimes gets stuck (a known problem with some
prism54 cards) and if I would eject it "too early" I would lose keyboard.
So I wait till it complain:
prism54: Your card/socket may be faulty, or IRQ line too busy :(
and eject and insert it again and all is fine.
--
Dmitry
On Wed, 21 Sep 2005 23:00:17 CDT, Dmitry Torokhov said:
> I have seen this couple of times when I would eject my prism54 card at
> a "bad" time - my card sometimes gets stuck (a known problem with some
> prism54 cards) and if I would eject it "too early" I would lose keyboard.
> So I wait till it complain:
>
> prism54: Your card/socket may be faulty, or IRQ line too busy :(
>
> and eject and insert it again and all is fine.
Hmm.. the wireless card in this is (a) an Orinoco and (b) not ejectable, as
it's a built-in, and (c) wasn't even in use at the time...
i8042 is on IRQ1, and the warp-drive modem is on IRQ11. And the USB that had
the still-working mouse is on IRQ11 too.
Anybody got a better idea than running a cronjob every 5 mins that runs
oprofile for 1 minute, and dumping the stats, and waiting for it to hit again?