Re: [PATCH 0/2] drm/i915/display: reduce clock_gating interface between core and display

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Thu, Jul 02, 2026 at 04:22:43PM +0300, Jani Nikula wrote:
> On Tue, 30 Jun 2026, Luca Coelho <luciano.coelho@intel.com> wrote:
> > As a follow up to my previous clock-gating patches, this series
> > reduces the interface between core and display by combining all the
> > platform-specific exported functions and letting the display code
> > decide what to do for each platform.
> >
> > To do that I created an _init_early(), to hold functions that are
> > called at the beginning of the clock gating initialization and another
> > one to call the later functions, namely _init_late().
> 
> As-is, this feels like forward progress, and I like the reduction of
> interfaces.
> 
> But what I'm really interested in, can we take steps further, and stop
> the back and forth calling of the functions between i915 and display?
> 
> For example, display hsw_disable_pc8() calls intel_clock_gating_init()
> with the comment /* Many display registers don't survive PC8+ */ but
> then it does so much more than just display. And it also calls back to
> display.
> 
> Could i915 core call its own clock gating functions where needed, and
> display call its own clock gating functions where needed, and not call
> from one to the other at all? Separate the two completely, at a level or
> two higher.

Just repeating what I've said several times before; the GT stuff
should be moved out from init_clock_gating() into the
gt/engine/etc. workaround framework. What's left should hopefully
be pure display stuff...

-- 
Ville Syrjälä
Intel