Re: [PATCH v4 2/2] arm64: dts: qcom: Add base QDU1000/QRU1000 IDP DTs

[Krzysztof Kozlowski]

On 15/12/2022 18:56, Melody Olvera wrote:
> 
> 
> On 12/15/2022 12:44 AM, Krzysztof Kozlowski wrote:
>> On 14/12/2022 19:59, Melody Olvera wrote:
>>>>> +			#clock-cells = <0>;
>>>>> +		};
>>>>> +
>>>>> +		sleep_clk: sleep-clk {
>>>>> +			compatible = "fixed-clock";
>>>>> +			clock-frequency = <32000>;
>>>>> +			#clock-cells = <0>;
>>>>> +		};
>>>>> +
>>>>> +		pcie_0_pipe_clk: pcie-0-pipe-clk {
>>>> Afaict these clocks are not referenced anywhere, so please skip them.
>>> Yes, so I included them to be consistent with the bindings. They will be needed later;
>>> should I still remove?
>>>
>> If they are not referenced anywhere, how is it consistent with bindings?
>> Where do the bindings require defining such nodes?
> 
> These bindings here: https://lore.kernel.org/all/20221118181826.28269-2-quic_molvera@xxxxxxxxxxx/
> I believe you commented that we either have these clocks or we don't, correct? I added them to
> the dt since these clocks exist and will be needed later when USB and PCIE nodes are added.
> As Konrad noted, these technically belong in the PHYs, but I was told to put stub fixed
> clocks instead here: https://lore.kernel.org/lkml/2c8c4642-8aee-3da3-7698-5e08b4c5894d@xxxxxxxxxx/
> 
> How is this to be handled? Should I remove the clocks from the dt and the bindings and add them

This was a discussion about clock controller. You either have these
clocks as input of the clock controller or not. If you have - drop
minItems. If you do not have - trim the list to just two items.

What do you actually write in DTS is of course connected with the
bindings but separate problem. I was not referring to what should be or
should not be in DTS. We talked about clock controller and its inputs.

Best regards,
Krzysztof