[PATCH] dt-bindings: display: msm: dsi: remove address/size cells

[Dmitry Baryshkov]

On Thu, 31 Mar 2022 at 09:05, Vinod Koul <vkoul at kernel.org> wrote:
>
> On 29-03-22, 10:52, Rob Herring wrote:
> > On Tue, Mar 29, 2022 at 12:01:52PM +0530, Vinod Koul wrote:
> > > On 28-03-22, 13:21, Rob Herring wrote:
> > > > On Mon, Mar 28, 2022 at 12:18 PM Krzysztof Kozlowski
> > > > <krzysztof.kozlowski at linaro.org> wrote:
> > > > >
> > > > > On 28/03/2022 19:16, Vinod Koul wrote:
> > > > > > On 28-03-22, 19:43, Dmitry Baryshkov wrote:
> > > > > >> On Mon, 28 Mar 2022 at 18:30, Krzysztof Kozlowski
> > > > > >> <krzysztof.kozlowski at linaro.org> wrote:
> > > > > >>>
> > > > > >>> The DSI node is not a bus and the children do not have unit addresses.
> > > > > >>>
> > > > > >>> Reported-by: Vinod Koul <vkoul at kernel.org>
> > > > > >>> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski at linaro.org>
> > > > > >>
> > > > > >> NAK.
> > > > > >> DSI panels are children of the DSI device tree node with the reg = <0>; address.
> > > > > >> This is the convention used by other platforms too (see e.g.
> > > > > >> arch/arm64/boot/dts/freescale/imx8mq-evk.dts).
> > > > > >
> > > > > > So we should add reg = 0, i will update my dtsi fix
> > > > > >
> > > > >
> > > > > To "ports" node? No. The reg=0 is for children of the bus, so the
> > > > > panels. How to combine both without warnings - ports and panel at 0 - I
> > > > > don't know yet...
> > > >
> > > > I don't think that should case a warning. Or at least it's one we turn off.
> > >
> > > Well in this case I think we might need a fix:
> > > Here is the example quoted in the binding. We have ports{} and then the
> > > two port at 0 and port at 1 underneath.
> >
> > It's the #address-cells/#size-cells under 'ports' that applies to 'port'
> > nodes. As 'ports' has no address (reg) itself, it doesn't need
> > #address-cells/#size-cells in its parent node.
> >
> > >
> > > So it should be okay to drop #address-cells/#size-cells from dsi node
> > > but keep in ports node...
> >
> > Yes.
> >
> > > Thoughts...?
> >
> > But I thought a panel at 0 node was being added? If so then you need to add
> > them back.
>
> I guess we should make this optional, keep it when adding panel at 0 node
> and skip for rest where not applicable..? Dmitry is that fine with you?

This sounds like a workaround. When a panel node is added together
with the '#address-cells' / '#size-cells' properties, we will get
warnings for the 'ports' node.
I'd prefer to leave things to pinpoint that the problem is generic
rather than being specific to several device trees with the DSI panel
nodes.
How do other platforms solve the issue?

In fact we can try shifting to the following dts schema:

dsi at ae940000 {
   compatible = "qcom,mdss-dsi-ctrl";

   ports {
      #adress-cells = <1>;
      #size-cells = <0>;
      port at 0 {
         reg = <0>;
         dsi0_in: endpoint {};
      };
      port at 1 {
         reg = <1>;
         dsi0_out: endpoint {
               remote-endpoint = <&panel_in>;
         };
      };

   /* dsi-bus is a generic part */
   dsi-bus {
      #adress-cells = <1>;
      #size-cells = <0>;
      /* panel at 0 goes to the board file */
      panel at 0 {
          compatible = "vendor,some-panel";
          ports {
             #adress-cells = <1>;
             #size-cells = <0>;
             port at 0 {
               reg = <0>;
                panel_in: endpoint {
                   remote-endpoint = <&dsi0_out>;
                };
             };
        };
   };
};

WDYT?

-- 
With best wishes
Dmitry