[Pm-utils] Canceling resume

[Victor Lowther]

On Fri, 2009-11-20 at 13:10 +0000, Enrico Zini wrote:
> On Thu, Nov 19, 2009 at 10:34:44AM -0600, Victor Lowther wrote:
> 
> > >However, upon resume pm-utils will run the various scripts in
> > >/etc/pm in
> > >order. The idea would be to have a hook, say, 80-openmoko, that tells
> > >pm-utils: "forget it, we need to go back to sleep".  And at that point
> > >pm-utils will:
> > >1. stop running further resume scripts
> > >2. start running suspend hooks starting from the one that canceled the
> > >   resume 80-openmoko
> > >3. suspend again
> > 
> > It seems like a reasonable request, nut it is not obvious how to
> > implement it in a general way.
> 
> It looks like I was following the wrong approach. I took some timings,
> and standard pm-utils scripts seem to be a bit too heavyweight for a
> mobile phone:
> 
>   00auto-quirk suspend... result: 0		 0.721047sec
>   00logging suspend... result: 0		 0.800695sec
>   00powersave suspend... result: 0		 1.787498sec
>   49bluetooth suspend... result: 254		 0.551373sec
>   55NetworkManager suspend... result: 254	 0.769849sec
>   55wicd suspend... result: 252			 2.561117sec
>   75modules suspend... result: 254		 1.077796sec
>   90clock suspend... result: 254		 0.808920sec
>   94cpufreq suspend... result: 0		 0.612665sec
>   95led suspend... result: 254		 	 0.106237sec
>   98smart-kernel-video suspend... result: 254	 0.715426sec
>   99video suspend... result: 0			 1.037772sec
>   99video resume... result: 0			 4.003470sec
>   98smart-kernel-video resume... result: 0	 0.755774sec
>   95led resume... result: 254			 0.112618sec
>   94cpufreq resume... result: 0			 0.678049sec
>   90clock resume... result: 254			 0.629024sec
>   75modules resume... result: 0			 0.843291sec
>   55wicd resume... result: 252			 2.202079sec
>   55NetworkManager resume... result: 254	 0.772729sec
>   49bluetooth resume... result: 254		 0.696186sec
>   00powersave resume... result: 0		 2.062235sec
>   00logging resume... result: 0			 0.554852sec
>   00auto-quirk resume... result: 0		 0.732000sec

Woah, shell scripts do not perform well at all on openmoko -- those
times are at least 3 orders of magnitude slower than our usual x86
laptop platform. Which shell are you using as /bin/sh?

> 
> This is not your fault, you're obviously targeting a different platform.
> 
> But I would still want to allow openmoko apps to be installed on my
> desktop and still work, so I'd like to have, on my phone, suspend/resume
> plugins that are compatible with pm-utils.

Neat, I will check it out.

> To that end, I prototyped a "light" version of pm-utils:
> http://git.debian.org/?p=users/enrico/omhacks.git
> 
> It tries to do the same, but it's implemented in C, it's simple minded
> and it does not ship the default heavyweight pm-utils hooks.

Yeah, most of the default hooks that pm-utils ships with are probably
not needed on a celphone. :)

> 
> It also implements "cancel resume", when a hook exits with status 250.
> 
> It's all still a proof of concept, and it hasn't proven the concept yet:
> I'm now trying to port a complicated suspend/resume setup to it to see
> if it copes.
> 
> I seems like a reasonable tradeoff to me to have the normal pm-utils and
> a lighter version that at least share the standard behaviours, so if the
> idea is viable I could be quite happy with it.
> 
> Only, I'd like the standard to stay in sync. I've had to pick an extra
> special exit status to implement "cancel resume", and it might be a good
> idea if we eventually standardise it, so that the proper pm-utils don't
> pick it in the future for something else, forcing all scripts that use
> that feature to be rewritten.

I do not have a problem with standardizing on exit code 250 as cancel
resume.

> Ciao,
> 
> Enrico
> 
> _______________________________________________
> Pm-utils mailing list
> Pm-utils at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/pm-utils