I wrote a program to automatically halt a Linux box when it gets too hot.

Why did I do this ? Well, I have a box running in a very badly ventilated cupboard and it tends to get pretty hot in there. Now when PCs get hot, stuff fails. After losing a hard drive to heat death, I wanted to make sure it did not happen again.

This is a cheap machine made of junk and spare parts which lives under my TV so I can watch the odd DivX movie and muck about with DVB cards, it’s hardly mission-critical so I’m not going to spend loads of cash on a sophisticated cooling system. This script is very simple and works fine.

Prerequisites

To make it work, you need to have lm_sensors installed and configured properly. This can be a little tricky and the procedure varies for different models of motherboard so unfortunately I cannot give you step-by-step instructions. You’ll have to read the manual I’m afraid.

Once that is working, you need to set the limits on each sensor. This is done by editing the file /etc/sensors.conf

Note that whenever you make any changes to /etc/sensors.conf, you must run the following command to commit your changes to the driver:

# sensors -s

If all is good, you should be able to type sensors and see a result like this:

$ sensors

as99127f-i2c-0-2d
Adapter: SMBus Via Pro adapter at e800
VCore:     +1.68 V  (min =  +1.52 V, max =  +1.84 V)
+3.3V:     +3.57 V  (min =  +3.14 V, max =  +3.79 V)
+5V:       +5.08 V  (min =  +4.73 V, max =  +5.24 V)
+12V:     +12.10 V  (min = +10.82 V, max = +13.19 V)
-12V:     -12.46 V  (min = -13.22 V, max = -10.74 V)
-5V:       -6.14 V  (min =  -6.14 V, max =  -3.76 V)
fan1:        0 RPM  (min =    0 RPM, div = 2)
fan2:     5037 RPM  (min = 2836 RPM, div = 2)
fan3:        0 RPM  (min =    0 RPM, div = 2)
M/B Temp:    +32°C  (high =   +60°C, hyst =   +58°C)
CPU Temp:  +46.0°C  (high =   +80°C, hyst =   +78°C)          (beep)
temp3:     -31.5°C  (high =  +122°C, hyst =  +121°C)
vid:      +1.600 V  (VRM Version 9.0)
alarms:
beep_enable:
Sound alarm enabled

My little C program will look at any sensor item whose name has “Temp” in it (“CPU Temp” and “M/B Temp” in the above example). If these items exceed the preset limits, the machine will be cleanly shut down.

You can download the source code here:

sensor-alert.c (2 kB)

Installation Instructions

To install it, first compile it:

# gcc -o sensor-alert sensor-alert.c

Then install it in /usr/sbin

# mv sensor-alert /usr/sbin

And finally, add the following line to /etc/crontab

* * * * * root /usr/sbin/sensor-alert

And there you have it. Hopefully this machine won’t be cooking itself now.

And yes, I know you could probably do it in three lines of PERL but I’m a C programmer so shoot me.