On 22.10.2024 12:09 AM, Akhil P Oommen wrote:
> On Mon, Oct 21, 2024 at 11:38:41AM +0200, Konrad Dybcio wrote:
>> On 11.10.2024 10:29 PM, Akhil P Oommen wrote:
>>> ACD a.k.a Adaptive Clock Distribution is a feature which helps to reduce
>>> the power consumption. In some chipsets, it is also a requirement to
>>> support higher GPU frequencies. This patch adds support for GPU ACD by
>>> sending necessary data to GMU and AOSS. The feature support for the
>>> chipset is detected based on devicetree data.
>>>
>>> Signed-off-by: Akhil P Oommen <quic_akhilpo@xxxxxxxxxxx>
>>> ---
>>
>> [...]
>>
>>> +
>>> + /* Initialize qmp node to talk to AOSS */
>>> + gmu->qmp = qmp_get(gmu->dev);
>>> + if (IS_ERR(gmu->qmp)) {
>>> + cmd->enable_by_level = 0;
>>> + return dev_err_probe(gmu->dev, PTR_ERR(gmu->qmp), "Failed to initialize qmp\n");
>>> + }
>>
>> I'm still in favor of keeping qmp_get where it currently is, so that
>> probe can fail/defer faster
>
> Sorry, I somehow missed this email from you until now.
>
> If it fails, then it probably doesn't matter if it is a bit late. But for defer, isn't there
> some optimizations to track the dependency from devicetree data? I am
> not entirely sure!
There's devlink for clocks/supplies etc, it doesn't apply universally
for all phandle references IIUC.
>
> Since qmp node is related to ACD, I felt it is better to:
> 1. Keep all acd probe related code in a single place.
> 2. Be more opportunistic in skipping qmp_get() wherever possible.
>
> But if you still have strong opinion on this, I can move it back in the
> next revision (v3).
I suppose the answer is yes, I have a strong opinion :D
Konrad
