Re: [PATCH v3 5/9] ARM: dts: stm32: add Hardware debug port (HDP) on stm32mp13

[Clement LE GOFFIC <clement.legoffic@foss.st.com>]

On 6/11/25 17:48, Krzysztof Kozlowski wrote:
> On 11/06/2025 16:08, Clement LE GOFFIC wrote:
>> On 6/11/25 08:35, Krzysztof Kozlowski wrote:
>>> On 10/06/2025 15:33, Clement LE GOFFIC wrote:
>>>> On 6/10/25 14:38, Krzysztof Kozlowski wrote:
>>>>> On 10/06/2025 14:02, Clement LE GOFFIC wrote:
>>>>>> On 5/29/25 11:01, Krzysztof Kozlowski wrote:
>>>>>>> On 28/05/2025 14:14, Clement LE GOFFIC wrote:
>>>>>>>>>
>>>>>>>>>> +		};
>>>>>>>>>> +
>>>>>>>>>> +		hdp: pinctrl@5002a000 {
>>>>>>>>>> +			compatible = "st,stm32mp131-hdp";
>>>>>>>>>> +			reg = <0x5002a000 0x400>;
>>>>>>>>>> +			clocks = <&rcc HDP>;
>>>>>>>>>>       			status = "disabled";
>>>>>>>>>
>>>>>>>>> Why are you disabling it? What is missing?
>>>>>>>>
>>>>>>>> Nothing is missing just disabled by default.
>>>>>>>> The node is then enabled when needed in board's dts file.
>>>>>>> Nodes should not be disabled by default if they are complete. That's why
>>>>>>> I asked what is missing. Drop.
>>>>>>
>>>>>> Hi Krzysztof, OK I better understand now.
>>>>>> So yes the 'pinctrl-*' properties which are board dependent are lacking.
>>>>>
>>>>> These are not properties of this node.
>>>>
>>>> Does this mean I should add 'pinctrl-*' properties in bindings yaml file ?
>>>> I don't get it..
>>>
>>> These properties have no meaning here, so the hardware description is
>>> complete. You claim that you miss them thus device is incomplete is just
>>> not correct: these properties do not belong here! They belong to the
>>> board but even there they are totally optional. Why would they be a
>>> required resource?
>>>
>>> To remind: we talk here ONLY about required resources.
>>
>> Yes, 'pinctrl-*' properties belongs to the board and are not required.
>> So nothing is missing.
>>
>> This hdp node in the SoC dtsi file can be enabled by default.
>> But the hdp driver will probe and do nothing because without the
>> 'pinctrl-*' properties from the board files it would not be able to
>> access to any pin.
> 
> 
> Pinctrl has other features in general, including interfaces to userspace
> (as pretty often combined with gpio, although not sure if relevant here).

You're right. Also HDP pinctrl has a GPO feature accessible from userspace.
But by default the HDP is not connected to any pad; it needs the board 
'pinctrl-*' properties to configure the SoC pinctrl and expose HDP on 
the SoC pads.

That's why for me the enabling of the driver should be in the board file 
together with the SoC pinctrl configuration.

The userland cannot change the pinctrl alternate function mux, this is 
statically defined by the devicetree.

> 
>> I consider enabling this driver by default in SoC dtsi file as just
>> increasing the boot time on "every" board.
>> It's the board dts that requires the hdp and provides the 'pinctrl-*'
>> properties to connect the hdp to some SoC pin and then to some signal on
>> the board. For me it's natural to have the status okay only in the board
>> dts file.
> 
> The DTS is not the way to optimize boot processes. It is OS-independent
> hardware description. My BSD system for example uses smart driver which
> avoids probing, but also my user-space needs this device to talk over
> exposed interface, so why choice of Linux probing should affect others?

As I wrote above the HDP will not offer any functionality without the 
'pinctrl-*' properties in the board file.
If you insist, I can enable it in the SoC file but I really don't see 
any reason for that.

Best regards, Clément

> 
> Best regards,
> Krzysztof