Comment 258 for bug 22336

2.6.27-1-generic from kernel.ubuntu.com made no difference on my notebook.

I tried both with polling disabled and with polling at 2 seconds, as set in /proc/acpi/thermal_zone/THRM/polling_frequency then ran a test with stress, and the temperature just kept climbing until it hit critical, which ACPI detected, shutting down the machine.

The notebook (currently running Hardy, but the problem has been present since Warty):
http://www.durbanet.co.za/colin/mecer-linux/mecer_n223ii_notebook_ubuntu_linux.html

On machines like this, which don't raise an alert via ACPI at any temperature apart from CRITICAL, but which do have a constantly-updated record accessible via ACPI of what the current temperature is, is it not possible for the kernel to do the following:

a. Poll the current temperature at a user-configurable period, with a reasonable default
b. Turn on throttling or whatever else is required if the temperature goes above CRITICAL minus n, where n is a user-configurable value with a reasonable default.
c. To turn off the protective throttling (or whatever) when the temperature drops below beneath CRITICAL minus n minus m, where m is user-configurable with a reasonable default.

In the past, I've tried without success to achieve the above using powersave (which was not part of my Hardy install, so I haven't tried that again recently). I currently keep the machine permanently thorttled by having this line in /etc/rc.local:

/bin/echo -n 800000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq