Re: [PATCH 1/7] dt-bindings: interrupt-controller: mpm: Document power-domains property

On 15-Jul-26 2:24 PM, Marc Zyngier wrote:
> On Wed, 15 Jul 2026 07:38:59 +0100,
> Marek Szyprowski <[email protected]> wrote:
>>
>> On 13.07.2026 17:11, Marc Zyngier wrote:
>>> On Mon, 13 Jul 2026 11:25:41 +0100,
>>> Sneh Mankad <[email protected]> wrote:
>>>> Remove #power-domain-cells property and add power-domains property for
>>>> MPM device.
>>>>
>>>> Signed-off-by: Sneh Mankad <[email protected]>
>>>> ---
>>>>  .../devicetree/bindings/interrupt-controller/qcom,mpm.yaml          | 6 +++---
>>>>  1 file changed, 3 insertions(+), 3 deletions(-)
>>>>
>>>> diff --git a/Documentation/devicetree/bindings/interrupt-controller/qcom,mpm.yaml b/Documentation/devicetree/bindings/interrupt-controller/qcom,mpm.yaml
>>>> index ebb40c48950ab3a8fc86f5708acfc33c33d68993..3f9645fbc3c20633077aaa589e5d5a43928dab51 100644
>>>> --- a/Documentation/devicetree/bindings/interrupt-controller/qcom,mpm.yaml
>>>> +++ b/Documentation/devicetree/bindings/interrupt-controller/qcom,mpm.yaml
>>>> @@ -68,8 +68,8 @@ properties:
>>>>          - description: MPM pin number
>>>>          - description: GIC SPI number for the MPM pin
>>>>  
>>>> -  '#power-domain-cells':
>>>> -    const: 0
>>>> +  power-domains:
>>>> +    maxItems: 1
>>> What makes you think it is OK to change an existing binding in an
>>> incompatible way?
>>
>>
>> Ulf suggests that this has been modeled upside down in the initial
>> submission:
>>
>> https://lore.kernel.org/all/CAPx+jO9d1qH12mxg-n1rkbp6Xd__sdrSMeoc7CPELE+jgxRYHA@mail.gmail.com/
> 
> Sure. We have collectively fucked up a lot of bindings over the years,
> mostly because of the lack of accurate documentation and people not
> giving a damn about the correctness of these descriptions. I'm glad
> that Ulf is now able to make it right.
> 
> But that is not a license to invalidate DTs retrospectively. We don't
> update DTs in lockstep with the kernel (I *never* do), and something
> that existed must be supported in the long run.
> 
> Properties can be deprecated, but:
> 
> - they must be kept in the binding
> 
> - drivers must still honor them if they are present in the DT

Currently only two chipsets have #power-domain-cells property mentioned in MPM DT node,
agatti and sm6375.

Agatti - does not execute CPU cluster LPMs as of now, so not handling the #power-domain-cells in
driver will have no effect in functionality.

sm6375 - Being an old chipset I am not aware of its status. 

If the driver honors the #power-domain-cells property, even the CPU level idle states will fail to
probe on these devices, due to the reasoning provided in cover letter.

With these patches, the CPU level idle states will function properly, so will cluster level
idle states if enabled.
In addition to that, the fallback mechanism of registering for CPU PM notification when power-domains
property is not mentioned in DT node will take care of the RPM notification when cluster idle state
is entered.

As suggested by Konrad, I will add a brief explanation on the need for this bindings change in this patch.

Thanks,
Sneh