On Fri, May 22, 2026 at 03:50:38PM +0300, Ville Syrjälä wrote:
> On Fri, May 22, 2026 at 03:46:50PM +0300, Jani Nikula wrote:
> > On Fri, 22 May 2026, Ville Syrjälä <[email protected]> wrote:
> > > On Fri, May 22, 2026 at 01:29:39PM +0300, Jani Nikula wrote:
> > >> On Fri, 22 May 2026, Imre Deak <[email protected]> wrote:
> > >> > On Fri, May 22, 2026 at 12:36:27AM +0300, Ville Syrjälä wrote:
> > >> >> On Mon, May 18, 2026 at 02:24:22PM +0300, Imre Deak wrote:
> > >> >> > Add helpers to defer and handle link params resets instead of
> > >> >> > open-coding the same. Rename intel_dp_reset_link_params() to
> > >> >> > intel_dp_reset_link_params_force() to align its name with the new
> > >> >> > deferred reset helpers.
> > >> >> > 
> > >> >> > When deferring a reset, return whether a new reset was queued, used 
> > >> >> > by a
> > >> >> > follow-up change.
> > >> >> > 
> > >> >> > Signed-off-by: Imre Deak <[email protected]>
> > >> >> > ---
> > >> >> >  drivers/gpu/drm/i915/display/g4x_dp.c         |  2 +-
> > >> >> >  drivers/gpu/drm/i915/display/intel_ddi.c      |  2 +-
> > >> >> >  drivers/gpu/drm/i915/display/intel_dp.c       | 41 
> > >> >> > +++++++++++++++----
> > >> >> >  drivers/gpu/drm/i915/display/intel_dp.h       |  3 +-
> > >> >> >  .../drm/i915/display/intel_dp_link_training.c |  4 +-
> > >> >> >  5 files changed, 38 insertions(+), 14 deletions(-)
> > >> >> > 
> > >> >> > diff --git a/drivers/gpu/drm/i915/display/g4x_dp.c 
> > >> >> > b/drivers/gpu/drm/i915/display/g4x_dp.c
> > >> >> > index 5ff1cdf4581a5..c20a97e21419b 100644
> > >> >> > --- a/drivers/gpu/drm/i915/display/g4x_dp.c
> > >> >> > +++ b/drivers/gpu/drm/i915/display/g4x_dp.c
> > >> >> > @@ -1265,7 +1265,7 @@ static void intel_dp_encoder_reset(struct 
> > >> >> > drm_encoder *encoder)
> > >> >> >  
> > >> >> >     intel_dp->DP = intel_de_read(display, intel_dp->output_reg);
> > >> >> >  
> > >> >> > -   intel_dp->reset_link_params = true;
> > >> >> > +   intel_dp_reset_link_params_defer(intel_dp);
> > >> >> >     intel_dp_invalidate_source_oui(intel_dp);
> > >> >> >  
> > >> >> >     if (display->platform.valleyview || 
> > >> >> > display->platform.cherryview)
> > >> >> > diff --git a/drivers/gpu/drm/i915/display/intel_ddi.c 
> > >> >> > b/drivers/gpu/drm/i915/display/intel_ddi.c
> > >> >> > index 86520848892e0..77819aaeccb76 100644
> > >> >> > --- a/drivers/gpu/drm/i915/display/intel_ddi.c
> > >> >> > +++ b/drivers/gpu/drm/i915/display/intel_ddi.c
> > >> >> > @@ -4664,7 +4664,7 @@ static void intel_ddi_encoder_reset(struct 
> > >> >> > drm_encoder *encoder)
> > >> >> >     struct intel_dp *intel_dp = 
> > >> >> > enc_to_intel_dp(to_intel_encoder(encoder));
> > >> >> >     struct intel_digital_port *dig_port = 
> > >> >> > enc_to_dig_port(to_intel_encoder(encoder));
> > >> >> >  
> > >> >> > -   intel_dp->reset_link_params = true;
> > >> >> > +   intel_dp_reset_link_params_defer(intel_dp);
> > >> >> >     intel_dp_invalidate_source_oui(intel_dp);
> > >> >> >  
> > >> >> >     intel_pps_encoder_reset(intel_dp);
> > >> >> > diff --git a/drivers/gpu/drm/i915/display/intel_dp.c 
> > >> >> > b/drivers/gpu/drm/i915/display/intel_dp.c
> > >> >> > index 1920d2f026665..13163dd085e91 100644
> > >> >> > --- a/drivers/gpu/drm/i915/display/intel_dp.c
> > >> >> > +++ b/drivers/gpu/drm/i915/display/intel_dp.c
> > >> >> > @@ -3710,7 +3710,11 @@ void intel_dp_set_link_params(struct 
> > >> >> > intel_dp *intel_dp,
> > >> >> >     intel_dp->lane_count = lane_count;
> > >> >> >  }
> > >> >> >  
> > >> >> > -void intel_dp_reset_link_params(struct intel_dp *intel_dp)
> > >> >> > +/*
> > >> >> > + * Reset link params now, preserving any deferred connector
> > >> >> > + * detect-time reset request.
> > >> >> > + */
> > >> >> > +void intel_dp_reset_link_params_force(struct intel_dp *intel_dp)
> > >> >> >  {
> > >> >> >     intel_dp->link.max_lane_count = 
> > >> >> > intel_dp_max_common_lane_count(intel_dp);
> > >> >> >     intel_dp->link.max_rate = intel_dp_max_common_rate(intel_dp);
> > >> >> > @@ -3720,6 +3724,28 @@ void intel_dp_reset_link_params(struct 
> > >> >> > intel_dp *intel_dp)
> > >> >> >     intel_dp->link.seq_train_failures = 0;
> > >> >> >  }
> > >> >> >  
> > >> >> > +/*
> > >> >> > + * Reset link params during the next connector detect.
> > >> >> > + * Return %true if a new reset was queued.
> > >> >> > + */
> > >> >> > +bool intel_dp_reset_link_params_defer(struct intel_dp *intel_dp)
> > >> >> 
> > >> >> I find the intel_dp_reset_link_params_defer() vs.
> > >> >> intel_dp_reset_link_params_force() naming rather confusing.
> > >> >> 
> > >> >> Can't immediately think of a really good name for
> > >> >> intel_dp_reset_link_params_defer() so maybe it's better to not
> > >> >> have a function for it at all (ie. just drop this patch)?
> > >> >
> > >> > The idea was to have an interface to reset the link params directly or
> > >> > in a deferred way, instead of a direct access of the flag.
> > >> >
> > >> > The names are not great yes. I could use what Jouni suggested instead,
> > >> > or if the above argument is not good enough I can also drop this patch.
> > >> 
> > >> I suggest intel_dp_reset_link_params() keeps its name, or gets renamed
> > >> to intel_dp_link_params_reset(). It just does the thing, no "force".
> > >> 
> > >> Then the other two use something like:
> > >> 
> > >> - submit/process
> > >> - queue/process
> > >> - stage/handle
> > >> 
> > >> or some combination i.e. something like:
> > >> 
> > >> intel_dp_link_params_reset()
> > >> intel_dp_link_params_reset_submit()
> > >> intel_dp_link_params_reset_process()
> > >
> > > All of it kinda leaves it a bit unclear how these things
> > > are related, and then one probably has to read through
> > > them to figure out what they're actually doing.
> > >
> > > Hence why I think just setting/checking the flag on the
> > > spot might be the clearest approach. You can immediately
> > > find where it's set vs. checked.
> > 
> > Trouble is if you want to hide that flag in an opaque type in the
> > future.
> 
> Do we want that?

There's a few property of the (active) link like intel_dp::link_rate,
lane_count, link.active, reset_link_params, which all could be tracked
in a separate link specific state. It could also make sense then to move
all the handlers of that state into a separate module, but not sure
about the details for that. For now I think it's better to drop this
patch and reconsider a suitable interface later. Will resend v2 with
that.

> 
> -- 
> Ville Syrjälä
> Intel

Reply via email to