Call For Testing: pm-utils-powersave-policy

[Amit Kucheria]

Chase,

I meant to send this a while ago, when I punted over
pm-utils-powersave-policy to you.

1. Why pm-utils-powersave-policy?
We want to get rid of laptop-mode-tools. We want to make pm-utils the
single point of implementing power management features.

2. What was dropped from laptop-mode-tools?
The following scripts were dropped for this first round of testing:
 * bluetooth - The script tries to bring down the bluetooth interface
 and rmmod hci_usb. Very hackish and people might want to use bluetooth on
 battery

 * configuration-file-control - Switching config files based on power
 state, scary!

 * cpufreq - messes with min/max frequencies and throttling states. Really
 unncessary since ondemand should give us optimum power/performance balance.
 If it doesn't that a bug in ondemand that should be fixed.

 * hal-polling - disables CD detection. Will break usecases, but there is no
 other way to detect CDs w/o polling AFAIK.

 * hdparm - tweaks to change idle timeouts and enable write-caching. These
 _could_ be considered if they are known to work for 99% of the HW out there.

 * laptop-mode - changes fs mount options (we default to relatime?), sets
 readahead in block devices (might be worth looking at again) and tweaks fs
 caching options (at the cost of possible data loss!)

 * lcd-brighness - handled by g-p-m

 * start-stop-programs - do we really want to provide a list of programs that
 are stopped in battery mode?

 * syslog-conf - change syslog.conf in battery mode. We use rsyslog now, and
 do buffering for most files before committing to disk

 * usb-autosuspend - breaks many existing usb devices (need to do this on a
 case by case basis). We don't want bugs like 85488

 * video-out - will break presentations! We need to fix the drivers to
 disable the port automatically

 * wireless-ipw - is broken and disabled in kernel now

We can go back and add some more tweaks in from this list, but IMO, we first
need to fix existing kernel drivers to power down HW when idle. The scripts
like bluetooth or video-out wouldn't be necessary.

Regards,
Amit

> I've updated the pm-utils and pm-utils-powersave-policy packages with
> the following changes:
> 
> * pm-utils:
>   - Do not inhibit futher power.d hooks if one fails
>   - Load hook blacklist for power.d hooks
> * pm-utils-powersave-policy:
>   - Fixed broken console-dpms hook
> 
> These changes have been uploaded to my powersave ppa as ~powersave2.
> When they finish building they will be available for testing by
> upgrading the packages using update-manager or apt-get upgrade.
> 
> Again, I appreciate all the testing people can provide. As noted
> earlier, I am interested in ensuring that the scripts work, and of any
> power savings they provide. Please see the original CFT email below
> for instructions on how to test.
> 
> Thanks,
> Chase
> 
> On Thu, Mar 18, 2010 at 10:59 AM, Chase Douglas
> <chase.douglas at ubuntu.com> wrote:
> > Hello all,
> >
> > I've been working on some fixes, additions, and cleanup to the current
> > pm-utils-powersave-policy package that's slated for inclusion in
> > Lucid. The package takes some of the good stuff from the old
> > laptop-mode-tools package and moves it to policy scripts that are
> > invoked by pm-powersave when on battery power. The items currently
> > include:
> >
> > * 10 second audio codec power down (vs no power down)
> > * 1 minute console (tty1) screen powersave timeout
> > * 60 (vs 15) second dirty page writeback timeout
> > * Disable WOL for ethernet devices
> > * Enable SATA link power save mode (when swapping is unlikely)
> > * Enable multi-thread and multi-core power save mode
> >
> > To test these changes, please add my powersave ppa
> > (ppa:chasedouglas/powersave) and upgrade the pm-utils and
> > pm-utils-powersave-policy packages. When you transition from AC to
> > battery power these scripts will fire and log what they did (if
> > anything) to /var/log/pm-powersave.log. You can also test by running
> > 'sudo pm-powersave true' to enable power saving, or 'sudo pm-powersave
> > false' to disable power saving.
> >
> > One easy way to gauge power savings is to install powertop and execute
> > while on battery power:
> >
> > $ sudo sh -c "sleep 30; powertop -d -t 30"
> >
> > This will wait 30 seconds and then retrieve power information for 30
> > seconds. During this time you should not use your laptop as we are
> > trying to get a steady-state result. It wouldn't hurt to run this a
> > few times to ensure the results are consistent. In the output you
> > should find a "Power usage (ACPI estimate)" wattage that describes how
> > much energy is used by the system.
> >
> > I am interested in hearing whether the scripts seem to be doing their
> > job by looking at the pm-powersave.log file and the power usage. A
> > comparison to Karmic power usage would also be helpful, but not
> > necessary.
> >
> > Thanks,
> > Chase
> >
> 
> -- 
> ubuntu-devel mailing list
> ubuntu-devel at lists.ubuntu.com
> Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel

-- 
----------------------------------------------------------------------
Amit Kucheria, Kernel Engineer || amit.kucheria at canonical.com
----------------------------------------------------------------------