Re: [PATCH 3/4] ARM: dts: rockchip: add display nodes for rk322x

[Heiko Stuebner]

Stephen,

Am Freitag, 14. Juni 2019, 22:36:09 CEST schrieb Stephen Boyd:
> Quoting Heiko Stuebner (2019-06-14 12:33:12)
> > Am Freitag, 14. Juni 2019, 20:32:35 CEST schrieb Justin Swartz:
> > > On 2019-06-14 19:45, Stephen Boyd wrote:
> > > >> diff --git a/arch/arm/boot/dts/rk322x.dtsi 
> > > >> b/arch/arm/boot/dts/rk322x.dtsi
> > > >> index da102fff96a2..148f9b5157ea 100644
> > > >> --- a/arch/arm/boot/dts/rk322x.dtsi
> > > >> +++ b/arch/arm/boot/dts/rk322x.dtsi
> > > >> @@ -143,6 +143,11 @@
> > > >> #clock-cells = <0>;
> > > >> };
> > > >> 
> > > >> +       display_subsystem: display-subsystem {
> > > >> +               compatible = "rockchip,display-subsystem";
> > > >> +               ports = <&vop_out>;
> > > >> +       };
> > > >> +
> > > > 
> > > > What is this? It doesn't have a reg property so it looks like a virtual
> > > > device. Why is it in DT?
> > > 
> > > This is a virtual device.
> > > 
> > > I assumed it would be acceptable to it find in a device tree due to 
> > > binding documentation, 
> > > "Documentation/devicetree/bindings/display/rockchip/rockchip-drm.txt, 
> > > which states:
> > > 
> > > <quote>
> > > The Rockchip DRM master device is a virtual device needed to list all
> > > vop devices or other display interface nodes that comprise the
> > > graphics subsystem.
> > > </quote>
> > > 
> > > Without the "display_subsystem" device node, the HDMI PHY and 
> > > rockchipdrmfb frame buffer device are not initialized.
> > > 
> > > Perhaps I should have included this in my commit message? :)
> > 
> > As Justin said, that is very much common as the root of the components
> > that make up the drm device and pretty much common in a lot of devicetrees
> > for the last 5 years and longer ;-) .
> > 
> > Also gpio-keys also don't have a reg property ;-) .
> > 
> 
> Do you have a SoC node? If so, this virtual device should live in the
> root, away from the nodes that have reg properties and are thus in the
> SoC node.

no we don't have any soc node and that display-subsystem is really the
same on _all_ currently supported Rockchip socs ... has gone through
dt-reviews numerous times for multiple socs, so I'm somewhat
confident that we're not doing something terrible wrong.

Heiko