On 2023/4/26 0:56, Conor Dooley wrote:
> On Tue, Apr 25, 2023 at 08:26:35PM +0800, Changhuang Liang wrote:
>> On 2023/4/25 17:35, Conor Dooley wrote:
>>> On Tue, Apr 25, 2023 at 05:18:10PM +0800, Changhuang Liang wrote:
>>>> On 2023/4/25 16:19, Krzysztof Kozlowski wrote:
>>>>> On 25/04/2023 09:57, Changhuang Liang wrote:
>>>>>> Yes, "starfive,jh7110-aon-pmu" is a child-node of "starfive,jh7110-aon-syscon".
>>>>>> In my opinion, "0x17010000" is "aon-syscon" on JH7110 SoC, and this "aon-pmu" is just
>>>>>> a part of "aon-syscon" function, so I think it is inappropriate to make "aon-syscon"
>>>>>> to a power domain controller. I think using the child-node description is closer to
>>>>>> JH7110 SoC.
>>>>>
>>>>> Unfortunately, I do not see the correlation between these, any
>>>>> connection. Why being a child of syscon block would mean that this
>>>>> should no be power domain controller? Really, why? These are two
>>>>> unrelated things.
>>>>
>>>> Let me summarize what has been discussed above.
>>>>
>>>> There has two ways to describe this "starfive,jh7110-aon-syscon"(0x17010000).
>>>> 1. (0x17010000) is power-controller node:
>>>>
>>>> aon_pwrc: power-controller@17010000 {
>>>> compatible = "starfive,jh7110-aon-pmu", "syscon";
>>>> reg = <0x0 0x17010000 0x0 0x1000>;
>>>> #power-domain-cells = <1>;
>>>> };
>>>>
>>>>
>>>> 2. (0x17010000) is syscon node, power-controller is child-node of syscon:
>>>>
>>>> aon_syscon: syscon@17010000 {
>>>> compatible = "starfive,jh7110-aon-syscon", "syscon", "simple-mfd";
>>>> reg = <0x0 0x17010000 0x0 0x1000>;
>>>>
>>>> aon_pwrc: power-controller {
>>>> compatible = "starfive,jh7110-aon-pmu";
>>>> #power-domain-cells = <1>;
>>>> };
>>>> };
>>>
>>> I thought that Rob was suggesting something like this:
>>> aon_syscon: syscon@17010000 {
>>> compatible = "starfive,jh7110-aon-syscon", ...
>>> reg = <0x0 0x17010000 0x0 0x1000>;
>>> #power-domain-cells = <1>;
>>> };
>
>> I see the kernel:
>> https://elixir.bootlin.com/linux/latest/source/arch/arm64/boot/dts/mediatek/mt8167.dtsi
>> this file line 42:
>> it's power-controller also has no meaningful properties.
>> What do you think?
>
> I'm not sure that I follow. It has a bunch of child-nodes does it not,
> each of which is a domain?
>
> I didn't see such domains in your dts patch, they're defined directly in
> the driver instead AFAIU. Assuming I have understood that correctly,
> your situation is different to that mediatek one?
>
> Cheers,
> Conor.
I think there child-nodes just need to operate some clock signals. Maybe
we don't need to discuss other platforms.
If Rob's method is confirmed. I will try it next version.
Maybe like this:
aon_syscon: syscon@17010000 {
compatible = "starfive,jh7110-aon-syscon", "syscon", "starfive,jh7110-aon-pmu";
reg = <0x0 0x17010000 0x0 0x1000>;
#power-domain-cells = <1>;
};
Rob and krzystof:
And I think patch[1][2] need to change. Right?
[1] https://lore.kernel.org/all/20230414024157.53203-6-xingyu.wu@xxxxxxxxxxxxxxxx/
[2] https://lore.kernel.org/all/20230414024157.53203-7-xingyu.wu@xxxxxxxxxxxxxxxx/
