[PATCH] arm64: make ARCH_BCM4908 select ARCH_BCMBCA

[Rafał Miłecki]

On 14.07.2022 23:25, Kursad Oney wrote:
> On Thu, Jul 14, 2022 at 4:45 PM Florian Fainelli <f.fainelli at gmail.com> wrote:
>>
>> On 7/13/22 23:35, William Zhang wrote:
>>> On 7/13/22 21:58, Rafał Miłecki wrote:
>>>> From: Rafał Miłecki <rafal at milecki.pl>
>>>>
>>>> BCA is a big set / family of devices sharing multiple hardware blocks.
>>>> It covers BCM4908, BCM63xx, BCM68xx devices and more.
>>>>
>>>> Most of drivers that depend on ARCH_BCM4908 should actually depend on
>>>> ARCH_BCMBCA. To make such transition easier, cleaner and breakage-free
>>>> add a proper "select".
>>>>
>>>> Later on - if we decide to keep ARCH_BCM4908 - it may be moved under
>>>> ARCH_BCMBCA menu. Keeping it may be helpful for limited compiling of DTS
>>>> files and "default" Kconfig entires. Or we may just decide to drop it.
>>>>
>>>> Signed-off-by: Rafał Miłecki <rafal at milecki.pl>
>>>> ---
>>>>     arch/arm64/Kconfig.platforms | 1 +
>>>>     1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/arch/arm64/Kconfig.platforms b/arch/arm64/Kconfig.platforms
>>>> index 88ddc2e5b152..7fea6955944e 100644
>>>> --- a/arch/arm64/Kconfig.platforms
>>>> +++ b/arch/arm64/Kconfig.platforms
>>>> @@ -49,6 +49,7 @@ config ARCH_BCM2835
>>>>
>>>>     config ARCH_BCM4908
>>>>         bool "Broadcom BCM4908 family"
>>>> +     select ARCH_BCMBCA
>>>>         select GPIOLIB
>>>>         help
>>>>           This enables support for the Broadcom BCM4906, BCM4908 and
>>> Thank you Rafal for adding this!  Our proposal is to replace with BCMBCA (and I
>>> have a WIP patchset for that). Agree there are pros and cons using
>>> single general config versus individual chip config(we have 15 BCA chip
>>> now in the kernel) as you and Florian were discussing in another
>>> thread. But we feel
>>> single config has more advantage as it can produce the same image to
>>> test all the chips and it is easier to maintain.  That is why we moved BCM63138
>>> ARCH_BCM_63XX to ARCH_BCMBCA.
>>
>> Agreed, I would prefer that we consolidate on the ARCH_BCMBCA Kconfig
>> symbol and remove ARCH_BCM4908. We ought to be able to do that in a
>> single release after v5.20-rc1 contains our pull request.
> 
> You could take Rafal's change, adding a "deprecated / will be removed"
> note to the ARCH_BCM4908 Kconfig text and postpone the removal for a
> cycle or two? Maybe overly cautious but just a thought.

The point of my patch was to make sure that users switching from 5.19 to
5.20 will get ARCH_BCMBCA=y in their config automatically. That's meant
to avoid at least some breakages and keep kernel easily bisectable.

I wrote that *later* we can leave or drop ARCH_BCM4908 to *avoid*
discussing it right now. That should ideally be considered once we see
all patches actually dropping ARCH_BCM4908.

Of couse my intention didn't work and I actually started a blind
discussion now ;)