Re: pownernowd run script

[Gerrit Pape <pape@smarden.org>]

On Mon, Jul 03, 2006 at 10:32:26AM +0200, xavier dutoit wrote:
> I've written a run script for powernowd (tested on a sarge with
> standard package):
> 
> #! /bin/sh
> modprobe cpufreq-userspace
> if [ -f /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor ]
>  then
>    exec /usr/sbin/powernowd -d
>  else
>    echo "required sysfs objects not found!"
>    echo -e "\tRead /usr/share/doc/powernowd/README.Debian for more
> information."
> fi
> 
> 
> I wasn't sure about how to deal with error messages if the proc can't
> change its frequency. I kept what the init.d script did.

If the scaling_governor file does not exist, runit will retry to start
the service every second.  That's good for temporary errors but not for
permanent ones.  I'm not sure whether this one is temporary.
Additionally modprobe might fail, so, if the absence of scaling_governor
is a permanent error, I'd suggest:

  #!/bin/sh
  set -e  # barf if modprobe fails
  modprobe cpufreq-userspace
  test ! -f /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor ||
    exec /usr/sbin/powernowd -d

  echo "required sysfs objects not found!"
  echo "Read /usr/share/doc/powernowd/README.Debian for more information."
  sv down "$(pwd)"

And note where the echo output goes to.  If this service has a log
service, the warning goes to the log, if not, it goes to the console.
If you redirect the echo output to stderr, the warning would end up in
the readproctitle log.

HTH, Gerrit.