[Pm-utils] pm-utils and pmu hardware.

[Peter Jones]

On Tue, 2006-05-02 at 11:51 +0200, Stefan Seyfried wrote:
> On Tue, May 02, 2006 at 10:31:43AM +0100, Richard Hughes wrote:
>  
> > Why do we need to check if the pmu device works? It's early, and I've
> > been up for hours so go easy on me :-)
> 
> Well, just open()ing it would probably suffice, but just checking
> for the existence of /dev/pmu ("test -c /dev/pmu"), might not, since
> there could still be people out there using static /dev.

I think that's not right.  If we're on static /dev, then the dev package
should be providing /dev/pmu on appropriate machines.  Likewise, if
you're on a dynamic /dev, udev or initscripts should be
creating /dev/pmu when appropriate.

So if we've got some small pmu shim program (I'll call it "pm-pmu"),
then the interface it presents to the shell should be that we call
"pm-pmu --suspend" or "pm-pmu --hibernate".  If it's not the right thing
to do, it returns an error, else it returns success.  Then our shell
code looks like:

pm-pmu --suspend || echo -n mem > /sys/power/state

And that's it, we're done.  No invoking it twice, no checking to see if
the device is there before we run it.  That stuff is great for a kernel
API, but it makes the shell script senselessly complicated.  If we've
already got to have a C program to do this part, it should just do all
of it on those boxes where it's needed.

-- 
  Peter