Re: [PATCH] ALSA: hda: intel-dsp-config: Fix Azulle Access 4 quirk detection

[Pierre-Louis Bossart <pierre-louis.bossart@xxxxxxxxxxxxxxx>]

On 6/26/24 15:09, Guenter Roeck wrote:
> On 6/26/24 00:04, Pierre-Louis Bossart wrote:
>>
>>
>> On 6/25/24 23:37, Guenter Roeck wrote:
>>> On Fri, Jun 21, 2024 at 10:35:31AM +0200, Pierre-Louis Bossart wrote:
>>>>
>>>>
>>>> On 6/21/24 08:15, Amadeusz Sławiński wrote:
>>>>> On 6/20/2024 9:27 PM, Allen Ballway wrote:
>>>>>> I filed a bug and after sharing the requested information it looks
>>>>>> like this device won't work on SOF without vendor support. Given
>>>>>> this,
>>>>>> would the original patch returning this device to using HDAudio be
>>>>>> reasonable, or is there an preferred alternative to force this device
>>>>>> into using HDAudio?
>>>>>>
>>>>>
>>>>> And can you share link to the issue on mailing list, so someone
>>>>> reading
>>>>> this thread in the future doesn't have to guess where it is? ;)
>>>>
>>>> https://github.com/thesofproject/linux/issues/4981
>>>>
>>>> I don't know what to do with this configuration.
>>>> We added a quirk to force SOF to be used for ES8336 devices. It worked
>>>> for some, but not for others. Now we have quite a few ES8336-based
>>>> platforms that are broken with zero support from the vendor, with
>>>> obscure I2C/GPIO/clk issues.
>>>> Are we going to tag each one of them and say 'not supported, use
>>>> HDMI only'?
>>>> That's pushing a bit the notion of quirk...It would generate an endless
>>>> stream of patches. The alternative is to do nothing and ask that those
>>>> platforms revert to HDMI audio only with a kernel parameter. That
>>>> latter
>>>> alternative has my vote.
>>>>
>>>
>>> Given that this apparently does not work for many ES8336 devices,
>>> would it make more sense to disable SOF support for those by default
>>> and _enable_ them with a kernel parameter ?
>>
>> Some configurations work, so we would break them.
>>
> 
> Yes, but for others it is a regression, so arguably the change to
> force-enable
> SOF caused a regression and should either be reverted or fixed such that
> all
> previously working configurations still work (even more so since fixing
> the problem one-by-one as affected systems are found is being rejected).

If you disable SOF, all you get with snd-hda-intel is HDMI outputs.
There is no alternative if you want local speakers, mic and jack
support, or extensions based on I2S.

Some of my Intel colleagues have an ES8336-based platform that works for
them, they successfully modified the sof_es8336 machine driver to add
HDMI support, so why would we disable a working config?

I would really err on the side of "if you contributed a working config,
or helped enable a working config, it will be maintained in the kernel".