Re: [GIT PULL] sound updates for 4.21

From: Hans de Goede <hdegoede@redhat.com>
To: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>,
	Takashi Iwai <tiwai@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
	Ingo Molnar <mingo@kernel.org>, Mark Brown <broonie@kernel.org>,
	Liam Girdwood <lgirdwood@gmail.com>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Azat Khuzhin <dohardgopro@gmail.com>
Subject: Re: [GIT PULL] sound updates for 4.21
Date: Mon, 31 Dec 2018 22:02:31 +0100	[thread overview]
Message-ID: <887fbee7-3969-c009-9751-1b82b632eb80@redhat.com> (raw)
In-Reply-To: <f1474b8c-6215-e842-1649-64d9ef17ef7a@linux.intel.com>

Hi,

On 12/31/18 9:10 PM, Pierre-Louis Bossart wrote:
> 
> On 12/31/18 12:15 PM, Takashi Iwai wrote:
>> On Mon, 31 Dec 2018 11:24:41 +0100,
>> Pierre-Louis Bossart wrote:
>>>
>>> On 12/31/18 2:11 AM, Takashi Iwai wrote:
>>>> On Mon, 31 Dec 2018 00:17:58 +0100,
>>>> Pierre-Louis Bossart wrote:
>>>>>> BTW, one thing I'd really like to avoid is to rearrange the probe
>>>>>> procedure of the legacy HDA driver (so that we can get codec_mask
>>>>>> during pci probe() call).  The async probe is the result of the many
>>>>>> struggles with the various and complex configurations.   Moving the
>>>>>> codec probe to the beginning isn't trivial and quite risky to break
>>>>>> something else.
>>>>> Agree, mucking with the probe isn't something we should look into,
>>>>> especially with this Skylake driver being eventually deprecated once
>>>>> SOF is at feature parity. This set of autodetection patches for 4.21
>>>>> was really targeting CFL/WHL+ devices, where the DSP usage is
>>>>> mandatory when directly-attached digital microphones are used. For
>>>>> Skylake and kabylake using the legacy by default is just fine.
>>>> OK, then how about applying the PCI class check only for such ones
>>>> like the patch below?  The macro isn't sexy and can be replaced with
>>>> another way, but you have an idea.
>>> The two patches which added the PCI class checks were supposed to be a
>>> simple bullet-proof way of detecting the DSP presence and solving a
>>> problem of coexistence between two drivers. At this point if we start
>>> adding quirks and still have unclear issues with HDMI support which
>>> isn't different for CFL+, it may be wiser to revert them to let the
>>> 4.21 merge window progress? It's frustrating but I'd rather solve this
>>> problem the right way than with multiple iterations rushed because of
>>> the merge window timing.
>> Fair enough, let's revert them for now.  I'm going to submit the
>> revert patch.
> 
> Thanks Takashi. If everyone who has been impacted by this issue can send me privately the result of the following two commands it'd help us figure out which machines expose the DSP - we may ask for additional information, e.g. NHLT tables. We clearly need to widen the validation scope since there is obviously a disconnect between what 3rd party BIOS expose and the technical consensus within Intel audio teams. Thanks in advance for your time.
> 
> -Pierre

On my Apollo Lake laptop which is also affected by this:

> lspci -s 0:1f.3 -vn

00:0e.0 0403: 8086:5a98 (rev 0b) (prog-if 80)
         Subsystem: 10ec:111a
         Flags: bus master, fast devsel, latency 0, IRQ 128
         Memory at 91218000 (64-bit, non-prefetchable) [size=16K]
         Memory at 91000000 (64-bit, non-prefetchable) [size=1M]
         Capabilities: <access denied>
         Kernel driver in use: snd_hda_intel
         Kernel modules: snd_hda_intel, snd_soc_skl

> more /sys/bus/pci/devices/0000\:00\:1f.3/class (should return 0x040100 or 0x040380, if the value is 0x040300 then the DSP is not exposed)

0x040380

Regards,

Hans