[ltp] Critical Temperature Trip Points?

[David Abrahams]

Hi,

Last week I overheated (again) while compiling GCC with -j4 and so I
decided to look at the problem again.  This had appeared in my syslog:

Apr 24 13:42:43 mcbain kernel: [12867.111092] ACPI: Critical trip point
Apr 24 13:42:43 mcbain kernel: [12867.111100] Critical temperature reached (100 C), shutting down.
Apr 24 13:42:43 mcbain kernel: [12867.160581] Critical temperature reached (95 C), shutting down.

So I decided to try some experiments.  I set up sensors-applet to show
me the acpi values for THM0 and THM1, and the ibm-acpi values for CPU,
mPCI, HDD, GPU, BAT0 and BAT1, and I ran a few simultaneous instances of
cpuburn-in to see what happened (just exercising one core with a single
thread didn't heat things up quickly enough).

Result: 

* as far as I can tell THM1 == CPU

* THM0 reached as high as 96 C and THM1 ran at a sustained 103 C with
  no complaints from cpuburn-in and no automatic shutdown from ACPI

* BAT0 runs consistently at 50 C even at idle.  BAT1 reads 21 C at idle,
  but I don't have a second battery in there.

So, my questions:

1. What sensor is ACPI measuring at 100 or 95C that makes it decide to
   shut the computer down?

2. Where does it get those trip points?  Are they arbitrary?  Gleaned
   from manufacturer data?  etc.

3. Is the BAT0 temperature reasonable?  All the fan control scripts I
   can find kick the fan up to maximum when BAT0 reaches 50C.  Perhaps
   not surprisingly to you, IBM-ACPI seems to have the same idea about
   the meaning of my sensors as t60-fancontrold does.  It surprised me a
   little because I assumed t60-fancontrold was mis-labelling BAT0.

Thanks in advance for any input.

-- 
Dave Abrahams
Boost Consulting
http://boost-consulting.com