On 30/12/2024 10:10, Peter Griffin wrote:

quoted

Maybe you have here two devices, maybe only one. If it is only one, then
it is not a syscon anymore, IMO.

I was going to suggest modelling PMU_INTR_GEN as its own sycon node,
and then either: -

1) Updating exynos-pmu driver to additionally take a phandle to
pmu-intr-gen syscon, and register the hotplug callbacks.

or

2) Create a new driver named something like exynos-pm or exynos-cpupm
which obtains the PMU regmap and also a phandle to PMU_INTR_GEN
syscon, and register the call backs.

Is there any preference from your side over approach 1 or 2, or maybe
something else entirely?

No preference, choose whatever results in simpler or more readable code.

Option 1 assumes that exynos-pmu on GS101 will drop the "syscon"
compatible, although it still might expose syscon through drivers. Just
the standard binding syscon does not feel fit here.

I don't have the hardware/user manual, so I don't know what PMU_INTR_GEN
really is. GS downstream code has something like PMUCAL, which looks
like separate device.

Best regards,
Krzysztof