[ltp] fan stuck on high - temps appear fine

[John Li]

On Thu, Oct 30, 2008 at 08:50:19AM -0200, Henrique de Moraes Holschuh wrote:
> On Thu, 30 Oct 2008, John Li wrote:
> 
> > $ cat /proc/acpi/ibm/thermal
> > temperatures: 31 41 -128 31 26 -128 24 -128 40 35 -128 -128 -128 -128 -128 -128
> 
> Sane, no reason at all for a speedy fan.

Okay, I've noticed some more things over the past few days. Maybe
they're helpful?


It seems like the fan speeds up a notch (or more, not sure) when the
2nd or 9th reading gets to 44 or 45. I'm not sure which one really
triggers it; maybe either. I see both of these are listed as
model-dependent values in the thinkpad-acpi README; the Thinkwiki page
for tracking sensor locations doesn't list recent models, it seems.


Also, I ran yes(1) until my fan revved up. Temperature readings were:
42 40 -128 42 27 -128 24 -128 40 39 -128 -128 -128 -128 -128 -128

Then I immediately killed yes. After a few seconds, the fan slowed
down. Readings were:
39 40 -128 40 27 -128 24 -128 40 39 -128 -128 -128 -128 -128 -128

In this case, the 2nd and 9th readings were sane in both cases,
whereas the 1st and 4th readings changed.


And finally, I've noticed that sometimes when I unplug the laptop and
run off the battery, the fan slows down. I just discovered this, so
I'm not sure what the pattern is exactly. Usually the temperatures
rise a bit and the fan kicks back in, though. dmesg just shows
messages about "CPU[01] attaching NULL sched-domain".


I would have imagined the BIOS controlled the fan directly - what sort
of bug would cause the fan to not slow down, even at sane
temperatures?


Thanks,
John