Re: ksoftirqd uses 99% CPU triggered by network traffic

[Héctor Martín]

> > Interrupt 9 is surely busy, no USB hardware plugged in just in case you're
> > wondering, and normally (while not using eth2) interrupt 9 is rock solid
> > (i.e. I doubt ACPI interrupts at all during normal use unless e.g. the power
> > button is pressed.)
>
> At 60 kirq/s without any network traffic, you may disable acpi then usb
> and eventually poke your nose in the bios setup first. No joke.
That's WITH network traffic, and I get 60k packets/s so it makes sense, 
but then once ksoftirqd starts using up all available cpu, I might be 
able to stop it by stoping the network traffic, but then any traffic 
makes it go nuts anyway. The problem is that after some minutes under 
this high-udp-packet-traffic it starts going nuts, with no prior 
warning. The mouse cursor just gets jerky--nothing gradual here, it just 
starts and then won't stop.

I was talking about IRQ9 because at 60k/s it's understandable that ACPI 
takes some CPU trying to see if it's the intended receiver, but 
ksoftirqd's behaviour is obviously not normal, besides the point that it 
works perfectly por 10-20 minutes at least.

I'll be out for a week at least and I doublt I will be able to answer 
back, sorry for the inconvenience, I should've posted this after acoming 
back. Thanks for the answer though ;)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/