Re: [PATCH v2 2/4] soc: qcom: rpmh: Add support to read back resource settings

[Konrad Dybcio]

On 11/21/25 9:41 AM, Maulik Shah (mkshah) wrote:
> 
> 
> On 11/17/2025 6:04 PM, Konrad Dybcio wrote:
>> On 11/17/25 9:26 AM, Maulik Shah (mkshah) wrote:
>>>
>>>
>>> On 11/12/2025 4:56 PM, Konrad Dybcio wrote:
>>>> On 10/21/25 11:08 PM, Kamal Wadhwa wrote:
>>>>> From: Maulik Shah <maulik.shah@xxxxxxxxxxxxxxxx>
>>>>>
>>>>> All rpmh_*() APIs so far have supported placing votes for various
>>>>> resource settings but the H/W also have option to read resource
>>>>> settings.
>>>>>
>>>>> This change adds a new rpmh_read() API to allow clients
>>>>> to read back resource setting from H/W. This will be useful for
>>>>> clients like regulators, which currently don't have a way to know
>>>>> the settings applied during bootloader stage.
>>>>>
>>>>> Link: https://lore.kernel.org/r/20250623-add-rpmh-read-support-v1-1-ae583d260195@xxxxxxxxxxxxxxxx
>>>>> Signed-off-by: Maulik Shah <maulik.shah@xxxxxxxxxxxxxxxx>
>>>>> Signed-off-by: Kamal Wadhwa <kamal.wadhwa@xxxxxxxxxxxxxxxx>
>>>>> ---
>>>>
>>>> [...]
>>>>
>>>>> +int rpmh_read(const struct device *dev, struct tcs_cmd *cmd)
>>>>> +{
>>>>> +	DECLARE_COMPLETION_ONSTACK(compl);
>>>>> +	DEFINE_RPMH_MSG_ONSTACK(dev, RPMH_ACTIVE_ONLY_STATE, &compl, rpm_msg);
>>>>> +	int ret;
>>>>> +
>>>>> +	ret = __fill_rpmh_msg(&rpm_msg, RPMH_ACTIVE_ONLY_STATE, cmd, 1, true);
>>>>> +	if (ret)
>>>>> +		return ret;
>>>>> +
>>>>> +	ret = __rpmh_write(dev, RPMH_ACTIVE_ONLY_STATE, &rpm_msg);
>>>>
>>>> Is there a reason for making this ACTIVE_ONLY?
>>>
>>> Yes, using ACTIVE_ONLY makes the read request place immediately to read back the current resource setting.
>>> Sleep/Wake are H/W based trigger and are not applicable for this API.
>>
>> Huh? So if I send a read request with e.g. SLEEP_STATE, it would only
>> get fulfilled upon an active->sleep transition?
>>
> 
> Read requests will get fulfilled immediately with the return of the current resource setting,
> there is no separate active/sleep/wake vote values that can be read, put it other way the
> rpmh_read() API do not take any "enum rpmh_state state" argument like various rpmh_write_*() APIs.

I see, thanks

Konrad