On Tue, Aug 28, 2012 at 08:16:54AM +0100, Chris Wilson wrote: > On Tue, 28 Aug 2012 09:53:36 +0300, Jani Nikula <jani.nikula at intel.com> wrote: > > From: Daniel Vetter <daniel.vetter at ffwll.ch> > > > > This is a prep patch to stop drm/i915 from changing the LBPC registers > > itself - but we still need to properly save/restore it on > > suspend/resume. > > My presumption was that there were BIOSes out there that only adjusted > the LBPC values, and so any device setting that flag would likely only > with it through the BIOS hotkeys. I couldn't see any other rational > reason for the hardware maintaining multiple interfaces... I've run a hackish version of this through a few bug reports and they reported that there's no more black screen at boot-up, but also that the backlight doesn't work. My conclusion was: - there are machines that are only controlled through the lbpc reg - some of those use lbpc in an inverted sense - we have now idea how to figure out the sense of lbpc Hence my thinking is that we should avoid to touch lbpc (but just restore it across suspend/resume since some machines need that) and rely on any firmware madness to properly adjust the backlight. I also realize that this will regress machines without working firmware, but for which the current lbpc-adjusting mess seems to do the right thing. But if the only regression this causes is some non-working backlight adjusting I'll ignore that, since it looks like we fundamentally can't do the right thing with lbpc, and this series here should kill a few black screen bugs. Cheers, Daniel -- Daniel Vetter Mail: daniel at ffwll.ch Mobile: +41 (0)79 365 57 48
