Re: [PATCH v9 4/5] arm64: dts: ti: k3-j784s4-evm: Enable DisplayPort-0

[Andrew Davis <afd@ti.com>]

On 8/7/23 1:29 PM, Aradhya Bhatia wrote:
> 
> 
> On 07-Aug-23 21:19, Andrew Davis wrote:
>> On 8/7/23 7:56 AM, Aradhya Bhatia wrote:
>>> Hi Jayesh,
>>>
>>> On 07-Aug-23 17:54, Jayesh Choudhary wrote:
>>>> Hello Aradhya,
>>>>
>>>> Thank you for the review.
>>>>
>>>> On 05/08/23 00:52, Aradhya Bhatia wrote:
>>>>> Hi Jayesh,
>>>>>
>>>>>
>>>>> On 03-Aug-23 13:34, Jayesh Choudhary wrote:
>>>>>> From: Rahul T R <r-ravikumar@ti.com>
>>>>>>
>>>>>> Enable display for J784S4 EVM.
>>>>>>
>>>>>> Add assigned clocks for DSS, DT node for DisplayPort PHY and pinmux
>>>>>> for
>>>>>> DP HPD. Add the clock frequency for serdes_refclk.
>>>>>>
>>>>>> Add the endpoint nodes to describe connection from:
>>>>>> DSS => MHDP => DisplayPort connector.
>>>>>>
>>>>>> Also add the GPIO expander-4 node and pinmux for main_i2c4 which is
>>>>>> required for controlling DP power. Set status for all required nodes
>>>>>> for DP-0 as "okay".
>>>>>>
>>>>>> Signed-off-by: Rahul T R <r-ravikumar@ti.com>
>>>>>> [j-choudhary@ti.com: move all the changes together to enable DP-0 in
>>>>>> EVM]
>>>>>> Signed-off-by: Jayesh Choudhary <j-choudhary@ti.com>
>>>>>> ---
>>>>>>     arch/arm64/boot/dts/ti/k3-j784s4-evm.dts | 119
>>>>>> +++++++++++++++++++++++
>>>>>>     1 file changed, 119 insertions(+)
>>>>
>>>> [...]
>>>>
>>>>>> +        reg = <0>;
>>>>>> +        cdns,num-lanes = <4>;
>>>>>> +        #phy-cells = <0>;
>>>>>> +        cdns,phy-type = <PHY_TYPE_DP>;
>>>>>> +        resets = <&serdes_wiz4 1>, <&serdes_wiz4 2>,
>>>>>> +             <&serdes_wiz4 3>, <&serdes_wiz4 4>;
>>>>>> +    };
>>>>>> +};
>>>>>> +
>>>>>> +&mhdp {
>>>>>> +    status = "okay";
>>>>>> +    pinctrl-names = "default";
>>>>>> +    pinctrl-0 = <&dp0_pins_default>;
>>>>>> +    phys = <&serdes4_dp_link>;
>>>>>> +    phy-names = "dpphy";
>>>>>> +};
>>>>>> +
>>>>>> +&dss_ports {
>>>>>> +    port {
>>>>>
>>>>> Port index has not been added here. Since this port outputs to MHDP
>>>>> bridge, this should be "port@0", and a "reg = <0>;" property should be
>>>>> added below (along with the address and size cells properties).
>>>>>
>>>>> I suppose this works functionally in this case, because the port gets
>>>>> defaulted to "0" by the driver. But in future, when we add support for
>>>>> other dss output(s) on j784s4-evm, the driver will need indices to
>>>>> distinguish among them.
>>>>>
>>>>
>>>> Okay. It makes sense.
>>>> Just one thing here. Adding reg here would require it to have #address-
>>>> cells and #size-cell but since we have only single child port that too
>>>> at reg=<0>, it would throw dtbs_check warning:
>>>>
>>>> arch/arm64/boot/dts/ti/k3-j784s4-main.dtsi:1828.20-1831.5: Warning
>>>> (graph_child_address): /bus@100000/dss@4a00000/ports: graph node has
>>>> single child node 'port@0', #address-cells/#size-cells are not necessary
>>>>     also defined at arch/arm64/boot/dts/ti/k3-j784s4-evm.dts:911.12-919.3
>>>>
>>>
>>> Okay! Was not aware about this. I still think "port@0" should be
>>> specified instead of just "port" and the warning should be ignored, if
>>> possible.
>>>
>>
>> Do not ignore new DT check warnings, if you go with "port@0" (which you
>> need to do as the "ti,j721e-dss" binding requires it) you must also add
>> the #address-cells/#size-cells.
>>
> 
> The warning that Jayesh mentioned above comes when "port@0" is
> mentioned, *along-with* the #address-cells/#size-cells properties.
> Essentially, it wants us to not use "port@0" when only single port is
> being added whose reg values is 0.
> 
> This warning does not come when only a single port other than 0,
> "port@1" for e.g., is being used. That's the warning, that should get
> ignored, if possible.
> 

Ah, I see now.

Almost seems like a bug in dtc checks, but checking the code it
looks deliberate, although I cannot see why..

Rob,

Could you provide some guidance on why graph nodes are handled
this way? Seems this is valid:

ports {
	#address-cells = <1>;
	#size-cells = <0>;

	port@1 {
		reg = <1>;
	};
}

but this is not:

ports {
	#address-cells = <1>;
	#size-cells = <0>;

	port@0 {
		reg = <0>;
	};
};

I'm guessing we allow port 0 to not be numbered if it is the only
one for legacy convenience, but *forcing* it to not be numbered
when it would otherwise be more consistent seems overly strict.

Andrew

> However, just mentioning "port@0", without the #address-cells/
> #size-cells, would be plain wrong.
> 
> Regards
> Aradhya
> 
>>
>>> If there were only a "port@1" child node, this warning would not have
>>> come up, and I believe "port@0" should be treated just the same.
>>>
>>> Moreover, while we can add these properties at a later stage as an
>>> incremental patch, adding the size and address cells in the dtsi would
>>> affect other platform dts files as well, that use this SoC.
>>>
>>> For e.g., the patch 5/5 of this series, on AM69-SK will still require
>>> the size and address cells for its ports. The clean up then will be that
>>> much more, when adding those incremental patches.
>>>
>>> Anyway, I will let Nishanth and Vignesh take the final call on this.
>>>
>>> Regards
>>> Aradhya
>>>
>>>>
>>>>>> +        dpi0_out: endpoint {
>>>>>> +            remote-endpoint = <&dp0_in>;
>>>>
>>>>
>>>> [...]
>>>
>