Re: [PATCH 3/6] clk: qcom: common: ensure runtime PM suspend completes on probe

[Konrad Dybcio]

On 4/8/26 9:26 AM, Taniya Das wrote:
> 
> 
> On 4/7/2026 4:28 PM, Konrad Dybcio wrote:
>> On 4/7/26 11:30 AM, Taniya Das wrote:
>>> When the clock controller is probed with 'use_rpm' enabled, the
>>> runtime PM reference is currently released using pm_runtime_put(),
>>> which may return before the runtime suspend has completed. This
>>> can leave the power domain transition in progress while the probe
>>> path continues or returns.
>>>
>>> Use pm_runtime_put_sync() instead to ensure the runtime PM “put”
>>> completes synchronously during probe, guaranteeing the device is
>>> fully runtime‑suspended before returning.
>>
>> It's not immediately obvious why that's an issue, is there an
>> observable problem when the transition doesn't complete in time?
>>
>> Konrad
> 
> 
> Calling pm_runtime_disable() immediately after pm_runtime_put() prevents
> the runtime suspend from completing, leaving the clock controller active
> and the HW rails in the ON state.

Please mention that you want to do it on a clock controller device
registered through this function, because currently it doesn't seem
to have a functional impact on the existing paths

Konrad