[Intel-gfx] [PATCH v1] drm/i915: Call intel_edp_init_connector only for eDP.

[Jani Nikula]

On Tue, 11 Feb 2020, "Lisovskiy, Stanislav" <stanislav.lisovskiy at intel.com> wrote:
> On Tue, 2020-02-11 at 15:03 +0200, Jani Nikula wrote:
>> On Tue, 11 Feb 2020, Stanislav Lisovskiy <
>> stanislav.lisovskiy at intel.com> wrote:
>> > I guess it would still be nice to make the code less confusing
>> > and do not call eDP specific function, for non-eDP connectors
>> > just to immediately return true(success) value as a hack.
>> > 
>> > So simply extracted that check out from this function,
>> > that we simply don't call it for non-eDP connectors. Bingo.
>> 
>> Fair enough, I guess...
>> 
>> Though I could be persuaded to take a patch for the reverse, because
>> generally we prefer localized knowledge in the callee than in a
>> potentially irrelevant place in the caller.
>> 
>> Consider the intel_dp_mst_encoder_init() call in the context of this
>> patch. We call it, and the function itself decides whether MST init
>> is
>> relevant or not. But I presume you wouldn't suggest pulling in all
>> the
>> conditions from there one level higher?
>> 
>> Would your opinion on intel_edp_init_connector() be different if the
>> conditions were more than just the single intel_dp_is_edp()? Or if we
>> moved all of eDP support to a separate file?
>
> Well, to me at least intel_edp_init_connector means that we are going
> to init an eDP connector, which already assumes that, we already should
> know that this is an eDP connector :) 
> Otherwise it should have somewhat different name, not saying that this
> is the best philosophy, but generally I would prefer the functions to
> be solely responsible for a single thing so that that init function is
> supposed to do only init, but not also some detection/checking or any 
> other side effects.
>
> So that there is a clear visibility what we are doing, if it's an init
> then we really do only init or if we supposed to detect something
> first, let it be a separate thing..
>
> Also this uses a return value in confusing way, i.e returning "Success"
> status for non-eDP from intel_edp_init_connector seems kind of
> confusing.

Again, how is this different from intel_dp_mst_encoder_init()?

With the *exactly* same rationale you'd end up with this:

        if (HAS_DP_MST(i915) && !intel_dp_is_edp(intel_dp)) &&
            !(INTEL_GEN(i915) < 12 && port == PORT_A) &&
            !(INTEL_GEN(i915) <	11 && port == PORT_E))
                intel_dp_mst_encoder_init(intel_dig_port,
                                          intel_connector->base.base.id);

Surely the information is better localized in a SPOT in the MST code?


BR,
Jani.

>
> Stan
>
>> 
>> BR,
>> Jani.
>> 
>> 
>> > 
>> > Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy at intel.com>
>> > ---
>> >  drivers/gpu/drm/i915/display/intel_dp.c | 13 ++++++-------
>> >  1 file changed, 6 insertions(+), 7 deletions(-)
>> > 
>> > diff --git a/drivers/gpu/drm/i915/display/intel_dp.c
>> > b/drivers/gpu/drm/i915/display/intel_dp.c
>> > index f4dede6253f8..9bd36197a43d 100644
>> > --- a/drivers/gpu/drm/i915/display/intel_dp.c
>> > +++ b/drivers/gpu/drm/i915/display/intel_dp.c
>> > @@ -7370,9 +7370,6 @@ static bool intel_edp_init_connector(struct
>> > intel_dp *intel_dp,
>> >  	intel_wakeref_t wakeref;
>> >  	struct edid *edid;
>> >  
>> > -	if (!intel_dp_is_edp(intel_dp))
>> > -		return true;
>> > -
>> >  	INIT_DELAYED_WORK(&intel_dp->panel_vdd_work,
>> > edp_panel_vdd_work);
>> >  
>> >  	/*
>> > @@ -7590,10 +7587,12 @@ intel_dp_init_connector(struct
>> > intel_digital_port *intel_dig_port,
>> >  	intel_dp_mst_encoder_init(intel_dig_port,
>> >  				  intel_connector->base.base.id);
>> >  
>> > -	if (!intel_edp_init_connector(intel_dp, intel_connector)) {
>> > -		intel_dp_aux_fini(intel_dp);
>> > -		intel_dp_mst_encoder_cleanup(intel_dig_port);
>> > -		goto fail;
>> > +	if (intel_dp_is_edp(intel_dp)) {
>> > +		if (!intel_edp_init_connector(intel_dp,
>> > intel_connector)) {
>> > +			intel_dp_aux_fini(intel_dp);
>> > +			intel_dp_mst_encoder_cleanup(intel_dig_port);
>> > +			goto fail;
>> > +		}
>> >  	}
>> >  
>> >  	intel_dp_add_properties(intel_dp, connector);
>> 
>> 

-- 
Jani Nikula, Intel Open Source Graphics Center