Re: [PATCH v2 2/4] arm64: dts: qcom: sm8450: remove invalid properties in cluster-sleep nodes

[Krzysztof Kozlowski]

On 24/03/2023 20:57, Bjorn Andersson wrote:
> On Fri, Mar 24, 2023 at 08:27:12PM +0100, Krzysztof Kozlowski wrote:
>> On 24/03/2023 18:45, Bjorn Andersson wrote:
>>> On Fri, Mar 24, 2023 at 10:28:47AM +0100, Neil Armstrong wrote:
>>>> Fixes the following DT bindings check error:
>>>
>>> Is that because idle-state-name and local-timer-stop should not be
>>> defined for domain-idle-states or are you just clearing out the
>>> dtbs_check warning?
>>>
>>> According to cpu-capacity.txt local-timer-stop seems to have been a
>>> property relevant for clusters in the past, was this a mistake in the
>>> binding or did something change when this was moved to
>>> domain-idle-states?
>>
>> I cannot find anything about local-timer-stop in cpu-capacity.txt. Where
>> do you see it?
>>
> 
> Ohh, you're right it's only mentioned in the example.
> 
> But idle-states.yaml documents the property for both cpus and clusters,
> and it's used throughout the examples.
>
> Our cluster states are defined in domanin-idle-states instead of
> idle-state, does this imply that the flag is no longer applicable
> per cluster in this mode of operation?

As you noticed their meaning is interleaving. For example on SC7280 we
use arm,idle-state for cluster. But other Qualcomm platforms rather
define clusters as domain-idle-states and in that case, nothing parses
tgat flag. The flag is only for cpuidle dt_idle_states. For
power-domains it was always ignored.

Funny fact - both cpu/cluster idle-states and power-domain-idle-states
will end up eventually in cpuidle-psci.c...

Best regards,
Krzysztof