Re: [RFC] Runtime display PM for VLV/BYT

From: Imre Deak <imre.deak@intel.com>
To: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
Cc: intel-gfx@lists.freedesktop.org,
	Paulo R Zanoni <paulo.r.zanoni@intel.com>
Subject: Re: [RFC] Runtime display PM for VLV/BYT
Date: Tue, 15 Oct 2013 15:16:11 +0300	[thread overview]
Message-ID: <1381839371.26119.12.camel@intelbox> (raw)
In-Reply-To: <20131015080606.GM13047@intel.com>

[-- Attachment #1.1: Type: text/plain, Size: 2030 bytes --]

On Tue, 2013-10-15 at 11:06 +0300, Ville Syrjälä wrote:
> On Mon, Oct 14, 2013 at 04:07:44PM -0700, Jesse Barnes wrote:
> > This set adds bits needed for runtime power support, currently only
> > lightly tested on VLV/BYT:
> >   1) suspend/resume callbacks for different platforms
> >   2) save/restore of display state across a power well toggle
> >   3) get/put of display power well in critical places
> > 
> > The TODO list still has a few items on it, and I'm looking for feedback:
> >   1) sprinkle around some power well WARNs so we can catch things easily
> >   2) add some tests using DPMS and NULL mode sets and comparing power
> >      well state
> >   3) better debugfs support for multiple wells
> >   4) refcount of power well in debugfs (with ref holders?)
> >   5) more testing - I think the load time ref is still busted here and
> >      on HSW
> >   6) convert HSW as well so DPMS will shut things down, not just mode
> >      sets
> > 
> > Thoughts or comments?
> 
> I'd also like to see what Imre cooked up, and then come up with some
> grand unified design. Based on our discussions I think his power well
> abstraction sounded somewhat nicer and more general.

I've pushed what I have so far to:
https://github.com/ideak/linux/commits/powerwells

I've tested this on VLV with VGA output so far and somewhat on HSW. I'd
still have to check the need to do any HW state save/restore and the GFX
clock forcing, afaics Jesse has already code for these in his patchset.

> Also your locking seems to be fubar in places (frobbing with sideband
> while holding a spinlock). I think Imre converted the power wells to
> use a mutex everywhere.

Yea, I solved that by changing power_well->lock to be a mutex.

> Or perhaps we just start with your stuff and Imre rebases his stuff on
> top?

That works for me too. In any case would be nice to get some feedback
especially from Paulo as my changes are mostly about the current power
domain / well handling parts.

--Imre

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx