Lm-Sensors: Run specific commands when temperature goes above/below limits

Question

I have a pretty badly ventilated computer whose temperature reaches 100º C in some occasions. The poor thing can't be ventilated any better ("put a bigger fan" is not a suitable solution). When the CPU reaches 100º C, the machine stops "violently" (just shuts down). That machine is running Ubuntu 10.10 with lm-sensors-3 (the installed package is lm-sensors 1:3.1.2-6)

I know what program is causing the issue (a very demanding media player) I could actually stop it for a while without causing major disruptions when the temperature reaches 98º C and boot it again when it reaches... lets say 90º C.

Is that possible to do something like that directly through lm-sensors or do I have to create my own process that checks lm-sensors periodically and "does its things" depending on the temperature?

Thank you in advance.

Answer 1

It depends on what is the output of sensors. If yours is something like mine:

% sensors
k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +44.0°C  (high = +70.0°C)

then you can use the following script, adapting it accordingly. Besides TEMP_STOP and TEMP_START, you should change the regular expression that filters the line from sensors you want to use. It's the parameter to grep, in the temp function.

#!/bin/bash

TEMP_STOP=98
TEMP_START=90

temp() {
    sensors | grep '^temp1:' | sed -e 's/.*: \+\([+-][0-9.]\+\)°C.*$/0\1/'
}

while true; do
    TEMP=$(temp)
    # keep waiting until temp is too hot
    while [ $(echo "$TEMP < $TEMP_STOP" | bc) = 1 ]; do
        sleep 10
        TEMP=$(temp)
    done

    echo temp $TEMP too hot, stopping.

    # now wait for it to cool down...
    while [ $(echo "$TEMP > $TEMP_START" | bc) = 1 ]; do
        sleep 10
        TEMP=$(temp)
    done

    echo ok now, restarting...
done