[SRU][OEM-5.6][PATCH 0/1] alsa: should not load sof driver if the internal mic connects to the codec

[Hui Wang]

On 2020/8/6 上午8:33, Hui Wang wrote:
>
> On 2020/8/5 下午11:33, Anthony Wong wrote:
>> Hi Hui,
>>
>> Hui Wang 於 5/8/2020 上午11:02 寫道:
>>> BugLink: https://bugs.launchpad.net/bugs/1890231
>>>
>>> This issue was introduced by #1872916 ("Support Intel Soundwire
>>> in 5.6-OEM Kernel 20.04"), so focal and groovy kernels don't need
>>> this fix.
>>>
>>> [Impact]
>>> We found the sof driver instead of hda driver is enabled on a dell
>>> machine without soundwire and dmic. This will make gnome show dummy
>>> output and users have no audio device to use.
>>>
>>> [Fix]
>>> When we integrated fix for #1872916, we backported 2 patches, these
>>> 2 patches enable the sof driver unconditionally for Dell comet lake
>>> machines. There is an upstream patch to fix it, let the sof driver
>>> enable conditionally on comet lake machines.
>>>
>>> [Test Case]
>>> Boot the kernel with this patch on that dell machine, check the
>>> gnome-control-center, all audio devices exist, playback and recording
>>> work well.
>>>
>>> [Regression Risk]
>>> On the Dell comet lake machines with dmic, the sof driver should
>>> enable. Without this patch, the sof driver will enable unconditionally,
>>> but with this patch, the sof driver will enable conditionally, so it
>>> is possible to introduce a regression that on Dell comet machines,
>>> the sof driver doesn't enable even the machine has dmic
>>
>> Can you confirm if the DMI_PRODUCT_SKUs listed in your patch, or with 
>> the DMIC ACPI check by snd_intel_dsp_check_dmic(), can cover all CML 
>> machines we have enabled? If so, would that eliminate the concern of 
>> regression?
>
> I have tested it on a Dell machine with pci dev id 0x06c8 and 
> DMI_PRODUCT_SKU "0990", it is a soundwire machine, it worked as well 
> as before; and tested it on a Dell machine with pci dev id 0x02c8, it 
> is a dmic machine, and through snd_intel_dsp_check_dmic() the sof 
> driver is enabled, so this machine worked as well as before too.
>
> What I concern is snd_intel_dsp_check_dmic() doesn't work as expected 
> on some Dell CML dmic machines we have enabled, in this situation the 
> sof driver will not be enabled and the dmic will not work. But this 
> possibility is very low since a dmic machine should have a valid dmic 
> definition table in the BIOS.
>
> Because I only have 2 Dell CML machines, it is best to test the patch 
> on other Dell CML machines we have enabled, I already added this 
> comment in the #1890076.

Let me re-organize:

There are 3 groups of Dell CML machines: soundwire,  dmic, hda

without this patch, all machines will enable sof driver unconditionally.

After applying the patch:

The DMI_PRODUCT_SKU will cover all soundwire machines, the ret=true of 
snd_intel_dsp_check_dmic() will cover all dmic machines, and the 
ret=false of snd_intel_dsp_check_dmic() will cover all hda machines. 
What I concern is snd_intel_dsp_check_dmic() will return false on some 
dmic machines we have enabled.

>
> Thanks,
>
> Hui.
>
>
>>
>> Thanks,
>> Anthony
>>
>>>
>>> I have tested this patch on a Dell comet lake dmic machines, it worked
>>> well.
>>>
>>>
>>> Pierre-Louis Bossart (1):
>>>    ALSA: hda: add autodetection for SoundWire
>>>
>>>   sound/hda/intel-dsp-config.c | 23 ++++++++++++++++++-----
>>>   1 file changed, 18 insertions(+), 5 deletions(-)
>>>