Re: [PATCH 11/11] arm64: defconfig: Enable Apple Silicon drivers

[Robin Murphy <robin.murphy@arm.com>]

On 2025-06-21 5:01 pm, Sven Peter wrote:
> On 13.06.25 18:50, Robin Murphy wrote:
>> On 2025-06-12 10:11 pm, Sven Peter wrote:
>>> Enable drivers for hardware present on Apple Silicon machines.
>>> The power domain and interrupt driver should be built-it since these are
>>> critical for the system to boot, the rest can be build as modules.
>>
>> Nit: I'd be tempted to put this patch first, just in case anyone 
>> bisecting with "make defconfig" in their process lands in the middle 
>> and suddenly loses some drivers (although arguably them going from 
>> "=y" to "=m" could still be a surprise, but at least a bit less so).
> 
> Ah, that's a good point that I hadn't even thought about.
> Now that most of these have already been merged into different trees 
> that ship has sailed though.

Yeah, like I say it's only minor and rather contrived anyway. Not a big 
worry.

>> [...]
>>> @@ -1504,6 +1520,7 @@ CONFIG_ARCH_TEGRA_194_SOC=y
>>>   CONFIG_ARCH_TEGRA_234_SOC=y
>>>   CONFIG_TI_PRUSS=m
>>>   CONFIG_OWL_PM_DOMAINS=y
>>> +CONFIG_APPLE_PMGR_PWRSTATE=y
>>
>> If this is critical for any Apple platform to work then it would 
>> probably make sense to explicitly select it from ARCH_APPLE, as is 
>> done for APPLE_AIC...
> 
> 
> Documentation/kbuild/kconfig-language.rst:
> 
>    select should be used with care. select will force a symbol to a value
>    without visiting the dependencies. By abusing select you are able to
>    select a symbol FOO even if FOO depends on BAR that is not set. In
>    general use select only for non-visible symbols (no prompts anywhere)
>    and for symbols with no dependencies. That will limit the usefulness
>    but on the other hand avoid the illegal configurations all over.
> 
> 
> That's probably fine for APPLE_AIC which only depends on ARM64 (and 
> ARCH_APPLE) which is guaranteed to be set when ARCH_APPLE is set anyway.
> APPLE_PMGR_PWRSTATE also has an additional dependency on PM so it should 
> probably remain in defconfig and not use select.

Sorry if the implication wasn't clear, but I did mean "select it" in the 
sense of "select it and all of its necessary dependencies in a manner 
that works correctly". The same argument applies all the way down - if a 
!PM config will not be able to boot on ARCH_APPLE platforms due to 
forcibly deselecting APPLE_PMGR_PWRSTATE, then there is little point in 
permitting that combination. Which of course the user would already have 
to disable at least ARCH_MXC, ARCH_ROCKCHIP, ARCH_TEGRA and 
ARCH_VEXPRESS to achieve in the first place ;)

Thanks,
Robin.