[PATCH v1] arm64, lib: make ARM64 select ARCH_HAS_CPU_CACHE_INVALIDATE_MEMREGION not GENERIC_CPU_CACHE_MAINTENANCE

[Randy Dunlap]

On 11/20/25 11:29 AM, Conor Dooley wrote:
> On Thu, Nov 20, 2025 at 05:30:46PM +0000, Catalin Marinas wrote:
>> On Wed, Nov 19, 2025 at 07:08:27PM +0000, Conor Dooley wrote:
>>> index 09aec4a1e13f..ac223e627bc5 100644
>>> --- a/lib/Kconfig
>>> +++ b/lib/Kconfig
>>> @@ -544,8 +544,9 @@ config ARCH_HAS_CPU_CACHE_INVALIDATE_MEMREGION
>>>  	bool
>>>  
>>>  config GENERIC_CPU_CACHE_MAINTENANCE
>>> -	bool
>>> -	select ARCH_HAS_CPU_CACHE_INVALIDATE_MEMREGION
>>> +	def_bool y
>>> +	depends on ARCH_HAS_CPU_CACHE_INVALIDATE_MEMREGION
>>> +	depends on ARM64
>>
>> That's what we do if GENERIC_CPU_CACHE_MAINTENANCE depends on some arch
>> code but that's not the case here. GENERIC_CPU_CACHE_MAINTENANCE is an
>> alternative implementation that an arch can select if it does not
>> provide its own. I find the current code without the above patch better.
> 
> Right, I am going to leave it as-is for now then. That's both you and I
> (and presumably Jonathan) finding the current form more natural. Sorry
> Randy!

Thanks. I tried.

I still say that it makes no sense for something generic
(GENERIC_CPU_CACHE_MAINTENANCE in lib/Kconfig) to inform the
config system that the config has something $ARCH-specific
(ARCH_HAS_CPU_CACHE_INVALIDATE_MEMREGION).

>>
>> Maybe what gets confusing here is that the core code uses
>> ARCH_HAS_CPU_CACHE_INVALIDATE_MEMREGION directly. A more involved fix
>> would be something like:
>>
>> config CPU_CACHE_INVALIDATE_MEMREGION
>> 	def_bool y
>> 	depends on ARCH_HAS_CPU_CACHE_INVALIDATE_MEMREGION ||
>> 		GENERIC_CPU_CACHE_MAINTENANCE
>>
>> and then go and change all the uses of
>> ARCH_HAS_CPU_CACHE_INVALIDATE_MEMREGION.
>>
>> Up to you, the current code also works for me.
>>
>> -- 
>> Catalin

-- 
~Randy