acpi_task_0 consuming cpu resource
Robert Noland
rnoland at FreeBSD.org
Fri Aug 7 16:01:14 UTC 2009
On Fri, 2009-08-07 at 18:27 +0800, C. C. Tang wrote:
> > Not sure then... I've got that board swapped out for a VIA board right
> > now. It tends to be noisy enough in my office with all the other fans
> > running that I really can't hear that one. I did have the smart fan
> > control enabled on both the case fan and cpu fan. It also periodically
> > rebuilds lots of ports or kernels with -j 5, so if it was going to heat
> > up, that should do it... If I can dig up a spare case and power, I'll
> > hook it back up and see if I can break it. Otherwise it will have to
> > wait until I get done with the VIA driver... I haven't really had any
> > issues that I can think of with that board, other than it is a bit slow
> > for compiling. It runs amd64 just fine.
>
> Really thanks for your reply.
> I will keeping watching it to see what situation in which the problem
> will happen.
> I am running on i386 version and I think the compile speed is OK for me.
> Anyway it is not a very fast CPU :)
>
> > Just remembering an issue that I had with some old p4 boxes though. On
> > those boxes, when they got hot, they would trigger an acpi message that
> > was picked up by devd. There was no way to throttle the messages, so it
> > would work the cpu as hard as it could, processing the messages. The
> > only solution at that point was to power it off for a few minutes while
> > it cooled off and not work it too hard. That sounds like what you are
> > seeing. Are you seeing messages in syslog?
>
> I have just done a
> # stress --cpu 1 --io 1 --vm 1 -d 1
> for several hours on the machine.
> Although it is not quite stressful, I think it is enough for it to reach
> 100% load.
> But I didn't notice any such warning in the system log and the
> acpi_task_0 problem didn't appear...
>
> By the way I cannot read any temperature from acpi or smbus except
> coretemp. (may be I loaded wrong driver?)
> So I am not sure how hot did the system got.
Yeah, IIRC coretemp is all I have gotten as well, but that is the
important one.
robert.
> Thanks,
> C.C.
>
--
Robert Noland <rnoland at FreeBSD.org>
FreeBSD
More information about the freebsd-acpi
mailing list