Detecting CPU throttling on over temperature

[Daniel O'Connor]

On Wed, 9 Sep 2009, Robert Noland wrote:
> On Wed, 2009-09-09 at 10:20 +0930, Daniel O'Connor wrote:
> > On Wed, 9 Sep 2009, Ian Smith wrote:
> > >  > > Does anyone know if it is possible to determine if this is
> > >  > > the case? ie is there a way to be informed if throttling has
> > >  > > occurred?
> > >
> > > Might be easier to hack powerd.c as an existing pretty
> > > lightweight way of monitoring CPU freq (to log or signal on
> > > detected freq lowered by throttling, say?) even if you don't
> > > need/want it to actually vary freq according to load, eg setting
> > > idle/busy shift factors to 'never/always'?
> >
> > Hmm, that could work.
> >
> > It seems odd to me that there is no direct way the BIOS can notify
> > the OS it's throttling the CPU though.
>
> Some BIOS can and do send an ACPI event when the proc gets hot.  In
> my experience, this was not a good thing though.  The BIOS that I
> remember dealing with this on would continuously send the alarms, so
> while TCC would kick in and throttle the CPU, the event processing
> kept it at 100% utilization until it was powered off to cool.  I have

Ugh!
This system seems to stall for a few seconds and then come back, I 
haven't see any messages about it in dmesg though.

> also been able to determine that TCC had kicked in by looking at the
> cpu frequency via sysctl and comparing that to the max frequency
> reported for the proc.

Yeah, although I couldn't run ps when the CPU was stalled so I'm not 
sure if I'd catch it or not :)

> If the BIOS sent the alarm, but throttled the rate it wouldn't have
> been so bad.  Not that I had any active fan control on that box to do
> anything about it really, but TCC might have actually worked if it
> wasn't flooding the acpi event processor.

Having the BIOS or CPU do it automatically is sensible since it's a time 
critical task.. Some basic notification would be nice though. It 
boggles my mind how difficult it is to do such basic things sometimes..

-- 
Daniel O'Connor software and network engineer
for Genesis Software - http://www.gsoft.com.au
"The nice thing about standards is that there
are so many of them to choose from."
  -- Andrew Tanenbaum
GPG Fingerprint - 5596 B766 97C0 0E94 4347 295E E593 DC20 7B3F CE8C
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 188 bytes
Desc: This is a digitally signed message part.
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20090909/dd86515e/attachment.pgp