Re: [RFC] drm/msm/dp: Allow attaching a drm_panel

[Doug Anderson]

Hi,

On Wed, Aug 25, 2021 at 6:31 PM Stephen Boyd <swboyd@xxxxxxxxxxxx> wrote:
>
> Quoting Bjorn Andersson (2021-07-26 16:13:51)
> > eDP panels might need some power sequencing and backlight management,
> > so make it possible to associate a drm_panel with a DP instance and
> > prepare and enable the panel accordingly.
> >
> > Signed-off-by: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>
> > ---
> >
> > This solves my immediate problem on my 8cx laptops, of indirectly controlling
> > the backlight during DPMS. But my panel is powered when I boot it and as such I
> > get the hpd interrupt and I don't actually have to deal with a power on
> > sequence - so I'm posting this as an RFC, hoping to get some input on these
> > other aspects.
> >
> > If this is acceptable I'd be happy to write up an accompanying DT binding
> > change that marks port 2 of the DP controller's of_graph as a reference to the
> > attached panel.
>
> dianders@ mentioned creating a connector (and maybe a bridge) for the DP
> connector (not eDP)[1]. I'm not sure that's directly related, but I
> think with the aux bus code the panel isn't managed in the encoder
> driver. Instead the encoder sees a bridge and tries to power it up and
> then query things over the aux bus? It's all a little too fuzzy to me
> right now so I could be spewing nonsense but I think we want to take
> this bridge route if possible.
>
> -Stephen
>
> [1] https://lore.kernel.org/r/CAD=FV=Xd9fizYdxfXYOkpJ_1fZcHp3-ROJ7k4iPg0g0RQ_+A3Q@xxxxxxxxxxxxxx/

The idea of modeling the connector as a bridge is something that makes
sense to me, but it probably makes sense to tackle that separately. We
don't need to block on it.

The whole DP AUX bus can also be tackled separately after the fact. It
really doesn't change things very much--we still have to find/attach
the panel. It just makes the panel probe in a slightly different way
and as a side effect gives the panel access to the DP AUX bus.

-Doug