Re: [PATCH] reset: Remove ARCH_HAS_RESET_CONTROLLER Kconfig symbol

[Andrew Davis <afd@ti.com>]

On 11/13/26 7:27 AM, Arnd Bergmann wrote:
> On Thu, Nov 12, 2026, at 22:42, Andrew Davis wrote:
>> Many ARCH_HAS_* symbols change some behavior based on what an ARCH
>> supports, but ARCH_HAS_RESET_CONTROLLER is used only to set a default
>> for RESET_CONTROLLER. Those ARCH symbols that select this can skip the
>> middleman and directly select RESET_CONTROLLER, and many already do.
>>
>> Replace all selection of this symbol with a direct selection of
>> RESET_CONTROLLER when it is not already in the selection list. In the
>> latter case when RESET_CONTROLLER is also already selected we can simply
>> remove the ARCH_HAS_RESET_CONTROLLER without changing existing behavior.
> 
> This is fine
> 
>> Some ARCHs "select" this symbol when they actually "depend on" the
>> RESET_CONTROLLER framework to build, in those cases the indirect select
>> could lead to broken randconfig builds. In cases when it is not needed
>> for building the ARCH code, these selections can be switched to the weaker
>> "imply" to allow it deselected for minimal builds without reset support.
>> Additional checking will be needed to find such cases.
> 
> Please don't recommend the use of 'imply', this is rarely what you
> want and everyone misunderstands what it does.
> 

Fair enough, I'll drop this suggestion for v2.

> If you want to have a default, please add another symbol and use
> that as the default value, such as
> 
> config ARCH_WANT_RESET_CONTROLLER
>        bool
> 

Could work, but in this case most of these ARCH don't actually "want"
the reset controller, at least not more than they want any other
random subsystem for which they have drivers like GPIO, NET, etc..

In a few cases there is an actual hard dependency as some ARCH code
calls into the reset controller framework, and in those cases directly
selecting RESET_CONTROLLER seems more correct than going through some
a "ARCH_WANT" symbol.

Andrew

> 
> 
>         Arnd