[PATCH 11/13] imx8mm-mx8menlo: Drop SPL_BOARD_INIT

[Marek Vasut]

On 2/16/23 15:41, Tom Rini wrote:
> On Thu, Feb 16, 2023 at 03:21:43PM +0100, Marek Vasut wrote:
>> On 2/16/23 15:13, Tom Rini wrote:
>>> On Thu, Feb 16, 2023 at 03:04:35PM +0100, Marek Vasut wrote:
>>>> On 2/16/23 04:36, Tom Rini wrote:
>>>>> On this platform spl_board_init is a call to arch_misc_init which is a
>>>>> no-op, so drop the CONFIG options.
>>>>>
>>>>> Cc: Marek Vasut <marex at denx.de>
>>>>> Cc: Olaf Mandel <o.mandel at menlosystems.com>
>>>>
>>>> btw. put those under --- next time, that way they don't end up in commit
>>>> message.
>>>
>>> Er, did the kernel change expected behavior here?
>>
>> Er ... wasn't that the case for like a year now ?
> 
> Is that a yes then?

I got repeated flak for sticking a wall of Cc: into the commit message 
recently, so I guess that's a yes .

>>>>> Signed-off-by: Tom Rini <trini at konsulko.com>
>>>>
>>>> If CAAM is enabled, ARCH_MISC_INIT brings up the CAAM , and this is needed
>>>> in SPL for U-Boot authentication using HABv4 . I believe that is why Verdin
>>>> spl.c calls it.
>>>
>>> Then I'll put doing a follow-up on SPL_BOARD_INIT (which is another
>>> option that shouldn't be directly asked, but select'd when used) for
>>> that case. It fails to build in this series because CAAM isn't enabled
>>> so there's no arch_misc_init.
>>
>> Just call the arch_board_init unconditionally, the CAAM inside of it is
>> already conditional, so the compiler should inline the result if CAAM is
>> disabled.
> 
> It doesn't, and only maybe does with LTO. But we also shouldn't be
> enabling unused hooks. It sounds like imx8m should follow the other
> platforms that have an spl_board_init under arch/ ?

spl_BOARD_init shouldn't be in arch in the first place, but I think what 
needs to be done here in the long run is, set DM_FLAG_PROBE_AFTER_BIND 
on CAAM in SPL if CAAM is enabled. That can be done somewhere in 
arch/arm/mach-imx/imx8m early boot code. And then let DM bring the CAAM 
up. I think that's the way to go with cleaning up the CAAM and 
spl_board_init part, without breaking support for HABv4.