Re: [PATCH v3 0/8] Add dual-LVDS panel support to EK874

[Rob Herring <robh+dt@kernel.org>]

On Wed, Aug 28, 2019 at 1:36 PM Fabrizio Castro
<fabrizio.castro@bp.renesas.com> wrote:
>
> Dear All,
>
> this series adds support for dual-LVDS panel IDK-2121WR
> from Advantech:
> https://buy.advantech.eu/Displays/Embedded-LCD-Kits-High-Brightness/model-IDK-2121WR-K2FHA2E.htm
>
> V3 approaches the problem in a completely different way, we now
> have two new properties to mark the ports in the DT as receiving
> even pixels and odd pixels: dual-lvds-even-pixels and dual-lvds-odd-pixels,
> which means device drivers should not use bridge specific or panel
> specific dual_link flags. Also, in this case the DT describes the
> connection fully.
>
> In order for the solution to be generic, I have exported a new helper
> (drm_of_lvds_get_dual_link_configuration) to walk the device tree,
> and figure out if the connection is dual-LVDS. The same helper gives
> information about the configuration of the connection. If Px is connected
> to a port expecting even pixels and Py is connected to a port expecting
> odd pixels, then the helper returns DRM_LVDS_DUAL_LINK_EVEN_ODD_PIXELS
> (like in the example below), otherwise it returns
> DRM_LVDS_DUAL_LINK_ODD_EVEN_PIXELS.
>
>
>  --------            dual-lvds-even-pixels  --------
> |        |----                         ----|        |
> |        | Px |---------------------->| Pn |        |
> |        |----                         ----|        |
> | SOURCE |           dual-lvds-odd-pixels  |  SINK  |
> |        |----                         ----|        |
> |        | Py |---------------------->| Pm |        |
> |        |----                         ----|        |
>  --------                                   --------
>
> The device driver for the encoder then will work out if with the current
> wiring the pixels need swapping or not.
>
> The same solution works for both panels and bridges.
>
> Since the DT describes the connection fully, driver
> drivers/gpu/drm/panel/panel-lvds.c works out-of-the-box, no changes
> required, however, this implementation opens up a problem with the
> dt-bindings.
> Driver drivers/gpu/drm/panel/panel-lvds.c can still be pleased by
> a port node, but also by a ports node.
> I have created Documentation/devicetree/bindings/display/bus-timings/lvds.yaml
> with the idea of including it from panels and bridges dt-bindings
> supporting dual-LVDS (and of course the dt-bindings for the specific
> devices should say which port should be marked as what), but file
> Documentation/devicetree/bindings/display/panel/lvds.yaml formally
> requires property "port", while with this implementation it should require
> OneOf "port" and "ports", and unfortunately I can't seem to find a neat way
> aroud that, other than creating a new compatible string

Just add 'ports' and drop 'port' from being required in the common
binding. Then it is up to the panel specific bindings to define which
one is required. Or we just leave it to allow either form which the
graph code can handle.

We could have this in the common binding:

oneOf:
 - required: [ports]
 - required: [port]

Rob