Re: [PATCH] ASoC: SOF: amd: fix soundwire dependencies

[Mukunda,Vijendar]

On 20/02/24 13:51, Arnd Bergmann wrote:
> On Tue, Feb 20, 2024, at 08:54, Mukunda,Vijendar wrote:
>> On 20/02/24 12:40, Arnd Bergmann wrote:
>>> On Tue, Feb 20, 2024, at 07:23, Mukunda,Vijendar wrote:
>>>> On 20/02/24 11:43, Arnd Bergmann wrote:
>>>>> On Tue, Feb 20, 2024, at 06:57, Mukunda,Vijendar wrote:
>>>>>> On 19/02/24 15:08, Arnd Bergmann wrote:
>>>>>>> From: Arnd Bergmann <arnd@xxxxxxxx>
>>>>>>> In normal configs, they should all either be built-in or all loadable
>>>>>>> modules anyway, so this simplification does not limit any real usecases.
>>>>>> Tested this patch. SOUNWIRE_AMD flag is not selected by default causing
>>>>>> AMD SOF driver for ACP 6.3 platform is build without enabling SoundWire.
>>>>> Yes, that is what I described. But as SOUNWIRE_AMD is a user visible
>>>>> symbol, there is no problem in expecting users to enable it when they
>>>>> have this hardware, and distros just enable all the drivers anyway.
>>>> Want to set SOUNDWIRE_AMD flag by default, similar to Intel & Qcom
>>>> platforms instead of explicitly enabling the Kconfig option.
>>> Maybe use 'default SND_SOC_SOF_AMD_TOPLEVEL' then?
>> Didn't get your point.
>>
>> Even with the current patch, if we select 'SOUNDWIRE_AMD' flag explicitly
>> AMD ACP63 SOF driver Kconfig option is not visible for user configuration.
>> This results in ACP63 SOF driver won't be built at all.
> I don't understand what you mean here. What I see
> in linux-next both with and without my patch is
>
> config SND_SOC_SOF_AMD_ACP63
>         tristate "SOF support for ACP6.3 platform"
>         depends on SND_SOC_SOF_PCI
>
> so it clearly should be visible as long as SND_SOC_SOF_PCI
> is enabled.
>
> There is still a problem that SND_SOC_SOF_AMD_TOPLEVEL
> can't use my "depends on SOUNDWIRE_AMD || !SOUNDWIRE_AMD"
> trick if SOUNDWIRE_AMD in turn uses
> "default SND_SOC_SOF_AMD_TOPLEVEL", but I don't think you
> meant that, right?
Yes, you are correct.
>
>>> I don't think copying the mistake from the intel driver
>>> is helpful, though I agree it would be nice to be consistent
>>> between them.
>>>
>>> As a general rule, you should not have a Kconfig symbol that
>>> is both user visible and also selected by the drivers that
>>> depend on it.
>>>
>>> To avoid the dependency problems from coming back and keep
>>> the complexity to a minimum, I think there are two logical
>>> ways to handle soundwire:
>>>
>>> a) keep the current drivers/soundwire/Kconfig contents and
>>>    change all the 'select SOUNDWIRE_foo' to 'depends on'.
>> Current patch already using 'depends on SOUNDWIRE_AMD" for
>> SND_SOC_SOF_AMD_SOUNDWIRE Kconfig option.
> Correct, because this is the Kconfig option that actually
> controls whether sound/soc/sof/amd/acp-common.c calls into
> the soundwire-amd module.
>
>> Still we couldn't see SND_SOC_SOF_AMD_ACP63 Kconfig option
>> is enabled.
> I need more information here. Do you have additional
> patches on top of what is in today's linux-next?
> I have it enabled on my build here.
Sorry, it's my bad. My local patches created the problem.
Validated patch on our side. It's working fine.

>
>       Arnd