Re: [PATCH] drm/bridge: ti-sn65dsi86: Implement lane reordering + polarity

From: Laurent Pinchart
Date: Wed May 06 2020 - 11:56:57 EST


Hi Doug,

On Tue, May 05, 2020 at 05:18:48PM -0700, Doug Anderson wrote:
> On Tue, May 5, 2020 at 2:24 PM Doug Anderson wrote:
> > On Tue, May 5, 2020 at 2:14 PM Laurent Pinchart wrote:
> > >
> > > > I'll add this documentation into the comments of the yaml, but I'm not
> > > > going to try to implement enforcement at the yaml level.
> > >
> > > Why not ? :-)
> >
> > Because trying to describe anything in the yaml bindings that doesn't
> > fit in the exact pattern of things that the yaml bindings are designed
> > to check is like constructing the empire state building with only
> > toothpicks.
> >
> > If you want to suggest some syntax that would actually make this
> > doable without blowing out the yaml bindings then I'm happy to add it.
> > Me being naive would assume that we'd need to do an exhaustive list of
> > the OK combinations. That would be fine for the 1-land and 2-lane
> > cases, but for 4 lanes that means adding 256 entries to the bindings.
> >
> > I think the correct way to do this would require adding code in the
> > <https://github.com/devicetree-org/dt-schema> project but that's
> > really only done for generic subsystem-level concepts and not for a
> > single driver.
>
> OK. Looked at your review of the .yaml and the "uniqueItems" is
> probably the bit I didn't think of. With that I can limit this but
> it's still a little awkward. I still haven't figured out how to force
> data-lanes and lane-polarities to have the same number of items, too.
> I'll add this as an add-on patch to my v2 and folks can decide if they
> like it or hate it.

Thanks for looking into it. Looks good to me. Regarding the same number
of items I would assume it should be possible, I would be surprised if
the schemas allowed a different number of items for clocks and
clock-names for instance, but maybe that's not implemented yet. In any
case, no big deal.

> # See ../../media/video-interface.txt for details.
> data-lanes:
> oneOf:
> - minItems: 1
> maxItems: 1
> uniqueItems: true
> items:
> enum:
> - 0
> - 1
> description:
> If you have 1 logical lane it can go to either physical
> port 0 or port 1. Port 0 is suggested.
>
> - minItems: 2
> maxItems: 2
> uniqueItems: true
> items:
> enum:
> - 0
> - 1
> description:
> If you have 2 logical lanes they can be reordered on
> physical ports 0 and 1.
>
> - minItems: 4
> maxItems: 4
> uniqueItems: true
> items:
> enum:
> - 0
> - 1
> - 2
> - 3
> description:
> If you have 4 logical lanes they can be reordered on
> in any way.

--
Regards,

Laurent Pinchart