Re: (subset) [PATCH v5 00/18] clk: qcom: Add support to attach multiple power domains in cc probe

[Jagadeesh Kona]

On 6/17/2025 6:57 PM, Bjorn Andersson wrote:
> On Mon, Jun 16, 2025 at 12:55:47PM +0530, Jagadeesh Kona wrote:
>>
>>
>> On 6/12/2025 4:22 PM, Krzysztof Kozlowski wrote:
>>> On 12/06/2025 12:03, Jagadeesh Kona wrote:
>>>>
>>>>
>>>> On 6/12/2025 9:30 AM, Bjorn Andersson wrote:
>>>>>
>>>>> On Fri, 30 May 2025 18:50:45 +0530, Jagadeesh Kona wrote:
>>>>>> In recent QCOM chipsets, PLLs require more than one power domain to be
>>>>>> kept ON to configure the PLL. But the current code doesn't enable all
>>>>>> the required power domains while configuring the PLLs, this leads to
>>>>>> functional issues due to suboptimal settings of PLLs.
>>>>>>
>>>>>> To address this, add support for handling runtime power management,
>>>>>> configuring plls and enabling critical clocks from qcom_cc_really_probe.
>>>>>> The clock controller can specify PLLs, critical clocks, and runtime PM
>>>>>> requirements using the descriptor data. The code in qcom_cc_really_probe()
>>>>>> ensures all necessary power domains are enabled before configuring PLLs
>>>>>> or critical clocks.
>>>>>>
>>>>>> [...]
>>>>>
>>>>> Applied, thanks!
>>>>>
>>>>> [01/18] dt-bindings: clock: qcom,sm8450-videocc: Add MXC power domain
>>>>>         commit: 1a42f4d4bb92ea961c58599bac837fb8b377a296
>>>>> [02/18] dt-bindings: clock: qcom,sm8450-camcc: Allow to specify two power domains
>>>>>         commit: a02a8f8cb7f6f54b077a6f9eb74ccd840b472416
>>>>> [03/18] dt-bindings: clock: qcom,sm8450-camcc: Move sc8280xp camcc to sa8775p camcc
>>>>>         commit: 842fa748291553d2f56410034991d0eb36b70900
>>>>> [04/18] clk: qcom: clk-alpha-pll: Add support for common PLL configuration function
>>>>>         commit: 0f698c16358ef300ed28a608368b89a4f6a8623a
>>>>> [05/18] clk: qcom: common: Handle runtime power management in qcom_cc_really_probe
>>>>>         commit: c0b6627369bcfec151ccbd091f9ff1cadb1d40c1
>>>>> [06/18] clk: qcom: common: Add support to configure clk regs in qcom_cc_really_probe
>>>>>         commit: 452ae64997dd1db1fe9bec2e7bd65b33338e7a6b
>>>>> [07/18] clk: qcom: videocc-sm8450: Move PLL & clk configuration to really probe
>>>>>         commit: 512af5bf312efe09698de0870e99c0cec4d13e21
>>>>> [08/18] clk: qcom: videocc-sm8550: Move PLL & clk configuration to really probe
>>>>>         commit: a9dc2cc7279a1967f37192a2f954e7111bfa61b7
>>>>> [09/18] clk: qcom: camcc-sm8450: Move PLL & clk configuration to really probe
>>>>>         commit: eb65d754eb5eaeab7db87ce7e64dab27b7d156d8
>>>>> [10/18] clk: qcom: camcc-sm8550: Move PLL & clk configuration to really probe
>>>>>         commit: adb50c762f3a513a363d91722dbd8d1b4afc5f10
>>>>> [11/18] clk: qcom: camcc-sm8650: Move PLL & clk configuration to really probe
>>>>>         commit: 3f8dd231e60b706fc9395edbf0186b7a0756f45d
>>>>> [12/18] clk: qcom: camcc-x1e80100: Move PLL & clk configuration to really probe
>>>>>         commit: d7eddaf0ed07e79ffdfd20acb2f6f2ca53e7851b
>>>>>
>>>>> Best regards,
>>>>
>>>>
>>>> Hi Bjorn,
>>>>
>>>> Thanks for picking these patches. However, the dt-bindings patches are closely linked with
>>>> the DT patches in this series and needs to be picked together. The dt-bindings changes adds
>>>
>>> DT bindings are the DT patches. What do you mean by DT? DTS? If so, then
>>> you introduce regressions without explaining this at all in cover letter
>>> or patches.
>>>
>>>> multiple power domains support for clock controllers, and without the corresponding DT
>>>> patches, dtbs_check will give warnings.
>>>>
>>>> Can you please help to pick DT patches as well?
>>>
>>> Please read soc maintainer profile explaining how DTS is being organized.
>>>
>>
>> I apologize for not mentioning this details in cover letter. Here the dt-bindings documentation
>> changes(patches 1-3) are only applied and the corresponding DTS changes(patches 13-18) are not
>> yet applied via DTS tree, leading to dtbs_check warnings.
>>
> 
> Mentioning this in the cover letter wouldn't change the fact that the
> binding is changing in an unexpected way.
> 
> As the binding now express that 2 power-domains is required, the driver
> author would be free to expect that the loaded DTB has 2 power-domains
> specified. But the user might still have an older DTB on their system
> (exactly what dtbs_check is complaining about now).
> 
> A quick look makes me think that it's because you removed the maxItems:1
> which means that the properties needs to match completely. Changing this
> to minItems:1 should allow for both cases, I think.
> 
> 
> Can you please send a patch that fixes up the bindings to allow both the
> old and new case?
> 

Thanks Bjorn and Krzysztof for the info and suggestion.

I tried adding minItems as 1 to power-domains and required-opps properties
and dtbs_check is working fine now for both single and multi power-domains cases.

I have posted the below series with this fixes:
https://lore.kernel.org/all/20250618-sm8450-videocc-camcc-bindings-single-pd-fix-v1-0-02e83aeba280@xxxxxxxxxxx/

Thanks,
Jagadeesh

> Regards,
> Bjorn
> 
>> Thanks,
>> Jagadeesh
>>
>>>
>>> Best regards,
>>> Krzysztof