Linux's DRM Power Saving Policy Gets Reverted For Now

[phoronix]

Phoronix: Linux's DRM Power Saving Policy Gets Reverted For Now

Submitted for DRM-Next last week with intentions of getting it into the Linux 6.12 kernel was a new DRM "power saving policy" property. The intent was for this new monitor/display connector property to indicate whether power saving features should be used that could compromise the experience intended by the desktop compositor. But one week later this property is now set to be removed as it's been deemed immature...

https://www.phoronix.com/news/DRM-Revert-Power-Saving-Prop

[ahrs]

What was the misunderstanding? Is that clarified anywhere? It was mentioned in the previous article that KWin already had an implementation making use of this and ready to go.

[emansom]

Still waiting on a Mesa GPU governor implementation.

Having to raise min clocks to 60% using CoreCtrl to get a somewhat decent desktop experience is ridicilous.

And no, Triple Buffering doesn't fix this either. (but it does help on low end/integrated)

[mxan]

What was the misunderstanding? Is that clarified anywhere? It was mentioned in the previous article that KWin already had an implementation making use of this and ready to go.

You can thank GNOME devs for it, they didn't like it, as usual

https://gitlab.gnome.org/GNOME/mutter/-/issues/3589

[IverCoder]

You can thank GNOME devs for it, they didn't like it, as usual

https://gitlab.gnome.org/GNOME/mutter/-/issues/3589

What does that have to do with Linux reverting this?

[ahrs]

What does that have to do with Linux reverting this?

It seems they want to shoehorn it into power-profiles-daemon instead. KDE still wants a solution in the kernel:

https://gitlab.freedesktop.org/upower/power-profiles-daemon/-/issues/159#note_2511705

That doesn't answer why it was removed from the kernel though, although Linux has always preferred things being done from user-space if they can do.

[archkde]

You can thank GNOME devs for it, they didn't like it, as usual

https://gitlab.gnome.org/GNOME/mutter/-/issues/3589

Really, the GNOME people don't like the sysfs thing bypassing the compositor in the first place, and think it should be removed instead of giving compositors yet another knob that they have to toggle to be in control again. IIRC the KDE people's opinion is actually the same.

[IverCoder]

It seems they want to shoehorn it into power-profiles-daemon instead. KDE still wants a solution in the kernel:

https://gitlab.freedesktop.org/upower/power-profiles-daemon/-/issues/159#note_2511705

That doesn't answer why it was removed from the kernel though, although Linux has always preferred things being done from user-space if they can do.

This thing has no business being in the kernel. This should be managed by `tuned` or `power-profiles-daemon` instead. I have no idea what the KDE developers has eaten that they support shoving this into the kernel.

[ahrs]

I have no idea what the KDE developers has eaten that they support shoving this into the kernel.

They want to avoid modesets which are annoying and cause flickering or the screen to turn off and back on.

The only argument for shoving this in PPD seems to be "I want this to work on muh legacy X11 session", "compositors other than Mutter and KWin and some Wayfire thing are too slow to update and only have a release every half a century. We need a stopgap!"