Re: [PATCH 20/20] arm64: dts: renesas: r8a774e1: Add VIN and CSI-2 nodes

[Geert Uytterhoeven]

Hi Niklas,

On Fri, Aug 7, 2020 at 1:27 PM Niklas Söderlund
<niklas.soderlund@xxxxxxxxxxxx> wrote:
> On 2020-08-06 13:47:58 +0200, Geert Uytterhoeven wrote:
> > On Thu, Aug 6, 2020 at 1:17 PM Lad, Prabhakar
> > <prabhakar.csengg@xxxxxxxxx> wrote:
> > > On Wed, Aug 5, 2020 at 12:19 PM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> > > > On Thu, Jul 16, 2020 at 7:20 PM Lad Prabhakar
> > > > <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx> wrote:
> > > > > Add VIN and CSI-2 nodes to RZ/G2H (R8A774E1) SoC dtsi.
> > > > >
> > > > > Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx>
> > > > > Reviewed-by: Marian-Cristian Rotariu <marian-cristian.rotariu.rb@xxxxxxxxxxxxxx>
> > > >
> > > > Reviewed-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
> > > >
> > > > However, before I queue this in renesas-devel for v5.10, I'd like to
> > > > have some clarification about the issue below.
> > > >
> > > > > --- a/arch/arm64/boot/dts/renesas/r8a774e1.dtsi
> > > > > +++ b/arch/arm64/boot/dts/renesas/r8a774e1.dtsi
> > > >
> > > > > +               vin4: video@e6ef4000 {
> > > > > +                       compatible = "renesas,vin-r8a774e1";
> > > > > +                       reg = <0 0xe6ef4000 0 0x1000>;
> > > > > +                       interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
> > > > > +                       clocks = <&cpg CPG_MOD 807>;
> > > > > +                       power-domains = <&sysc R8A774E1_PD_ALWAYS_ON>;
> > > > > +                       resets = <&cpg 807>;
> > > > > +                       renesas,id = <4>;
> > > > > +                       status = "disabled";
> > > > > +
> > > > > +                       ports {
> > > > > +                               #address-cells = <1>;
> > > > > +                               #size-cells = <0>;
> > > > > +
> > > > > +                               port@1 {
> > > > > +                                       #address-cells = <1>;
> > > > > +                                       #size-cells = <0>;
> > > >
> > > > "make dtbs W=1" says:
> > > >
> > > >     arch/arm64/boot/dts/renesas/r8a774e1.dtsi:1562.12-1572.7: Warning
> > > > (graph_child_address): /soc/video@e6ef4000/ports/port@1: graph node
> > > > has single child node 'endpoint@0', #address-cells/#size-cells are not
> > > > necessary
> > > >
> > > > (same for vin5-7 below)
> > > >
> > > Referring to commit 5e53dbf4edb4d ("arm64: dts: renesas: r8a77990: Fix
> > > VIN endpoint numbering") we definitely need endpoint numbering.
> > > Probably the driver needs to be fixed to handle such cases.
> >
> > > > > +
> > > > > +                                       reg = <1>;
> > > > > +
> > > > > +                                       vin4csi20: endpoint@0 {
> > > > > +                                               reg = <0>;
> > > > > +                                               remote-endpoint = <&csi20vin4>;
> >
> > On R-Car E3, the single endpoint is at address 2, so "make dtbs W=1"doesn't
> > complain. Here it is at address 0.
> >
> > Niklas?
>
> First the R-Car VIN driver makes decisions based on which endpoint is
> described, each endpoint 0-3 represents a different CSI-2 block on the
> other end (0: CSI20, 1: CSI21, 2: CSI40 and 3: CSI41).

That's my understanding, too.

> Then how to handle the warning I'm not sure. I can only really see 2
> options.
>
> 1. Ignore the warning.
> 2. Remove #address-cells, #size-cells and reg properties from port@ if
>    the only endpoint described is endpoint@0.
>
> I would prefers option 2. that is what we do in other cases (for example
> on Gen2 boards that only have a single parallel sensor in some early DTS
> files we don't have the ports node and just describe a single port with
> the same reasoning.
>
> We are not at risk at someone describing a second CSI-2 bock as an
> overlay so I see no real harm in option 2.

Yeah, no overlay possible for on-SoC wiring ;-)

> What are your thoughts Geert?
> You know more about DT then me.

You have too much faith in me ;-)

AFAIK we don't get this warning for e.g. SPI buses, which can have a
single device at address 0, and #{address,size}-cells is mandatory
there. So endpoints (or SPI?) are treated special?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds