Re: [PATCH] dt-bindings: clock: support NXP i.MX95

From: Krzysztof Kozlowski
Date: Tue Jan 23 2024 - 10:22:14 EST


On 23/01/2024 16:13, Peng Fan wrote:
>> Subject: Re: [PATCH] dt-bindings: clock: support NXP i.MX95
>>
>> On 23/01/2024 14:59, Peng Fan wrote:
>>>> Subject: Re: [PATCH] dt-bindings: clock: support NXP i.MX95
>>>>
>>>> On 23/01/2024 14:08, Peng Fan wrote:
>>>>> Hi Conor, Krzysztof
>>>>>
>>>>> I replied you both here.
>>>>>
>>>>>> Subject: Re: [PATCH] dt-bindings: clock: support NXP i.MX95
>>>>>>
>>>>>> On 21/01/2024 12:46, Peng Fan (OSS) wrote:
>>>>>>> From: Peng Fan <peng.fan@xxxxxxx>
>>>>>>>
>>>>>>> Add i.MX95 clock dt-binding header file
>>>>>>>
>>>>>>> Signed-off-by: Peng Fan <peng.fan@xxxxxxx>
>>>>>>
>>>>>> This should be squashed with respective binding patch.
>>>>>
>>>>> i.MX95 use SCMI firmware, the SCP processor handles
>>>>> clock/power/pinmux/reset/performance and etc.
>>>>>
>>>>> I just add the index that SCMI exports, and Linux/UBoot will use the
>>>>> index and go through SCMI to do real HW configuration.
>>>>>
>>>>> In such case, should I still add HW information in a yaml binding doc?
>>>>> I am not sure what should be added if yes.
>>>>
>>>> Then why do you need it in the binding? I don't see any use of this.
>>>
>>> ok, should I just add the clk index header in same folder as soc dtsi?
>>
>> I don't know yet. How are these being used, except SCMI firmware?
>
> Yeah. The clock module is managed by SCMI firmware, others are
> not able to access because of Trusted Resource Domain controller
> blocked the access from linux or else.
>
> The linux side use such a node, and pass clk index to let SCMI configure.
> scmi_clk: protocol@14 {
> reg = <0x14>;
> #clock-cells = <1>;
> };
>

OK, then my usual recommendation is that this does not go to the binding
and use numbers directly in the DTS. There is no binding here between
Linux drivers and DTS.

Best regards,
Krzysztof