Dne 11. 03. 21 v 11:40 Amadeusz Sławiński napsal(a):
On 3/11/2021 11:24 AM, Jaroslav Kysela wrote:
Dne 11. 03. 21 v 6:50 Chris Chiu napsal(a):
On Tue, Mar 9, 2021 at 11:29 PM Cezary Rojewski cezary.rojewski@intel.com wrote:
On 2021-03-09 1:19 PM, Chris Chiu wrote:
Hi Guys, We have received reports that on some Kabylake laptops (Acer Swift SF314-54/55 and Lenovo Yoga C930...etc), all sound cards no longer be detected after upgrade to kernel later than 5.8. These laptops have one thing in common, all of them have Realtek audio codec and connect the internal microphone to DMIC of the Intel SST controller either [8086:9d71] or [8086:9dc8]. Please refer to https://bugzilla.kernel.org/show_bug.cgi?id=201251#c246 and https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1915117.
From the dmesg from kernel 5.8, the sound related parts only showas follows but the expected snd_hda_codec_realtek and the snd_soc_skl are not even loaded then. [ 13.357495] snd_hda_intel 0000:00:1f.3: DSP detected with PCI class/subclass/prog-if info 0x040100 [ 13.357500] snd_hda_intel 0000:00:1f.3: Digital mics found on Skylake+ platform, using SST driver
Building the kernel with the CONFIG_SND_SOC_INTEL_KBL removed canload the snd_hda_codec_realtek successfully and the pulseaudio and alsa-utils can detect the sound cards again. The result of bisecting between kernel 5.4 and 5.8 also get similar result, reverting the commit "ALSA: hda: Allow SST driver on SKL and KBL platforms with DMIC" can fix the issue. I tried to generate the required firmware for snd_soc_skl but it did not help. Please refer to what I did in https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1915117/comments/14 and https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1915117/comments/18.
Since the skl_hda_dsp_generic-tplg.bin and dfw_sst.bin are not inthe linux-firmware. The Intel SST support for Skylake family is not yet complete. Can we simply revert the "ALSA: hda: Allow SST driver on SKL and KBL platforms with DMIC" in the current stage and wait for SOF support for Skylake family? Or please suggest a better solution for this. Thanks
Chris
Hello Chris,
Guide: "Linux: HDA+DMIC with skylake driver" [1] should help understanding history behind the problem as well as fixing it.
Upstream skylake driver - snd_soc_skl - is intended to support HDA DSP + DMIC configuration via means of snd_soc_skl_hda_dsp machine board driver. You _may_ switch to legacy HDAudio driver - snd_hda_intel - losing DMIC support in the process. To remove any confusion - for Skylake and Kabylake platforms, snd_soc_skl is your option.
Now, due to above, I doubt any skylake-related topology has ever been upstreamed to linux-firmware as a) most boards are I2S-based, these are used by our clients which we support via separate channel b) hda dsp+dmic support on linux for missing until early 2020.
Topologies for most common skylake driver configurations:
- skl/kbl with i2s rt286
- apl/glk with i2s rt298
- <any> with hda dsp
can be found in alsa-topology-conf [2].
Standard, official tool called 'alsatplg' is capable of compiling these into binary form which, after being transferred to /lib/firmware/ may be consumed by the driver during runtime. I have no problem with providing precompiled binaries to linux-firmware, if that's what community wants.
Regards, Czarek
I think the guild [1] is too complicated for normal users to fix the problem. Given it's not only the internal microphone being affected, it's no sound devices being created at all so no audio functions can work after kernel 5.8.
Is there any potential problem to built-in the "<any> with hda dsp" precompiled binary in linux-firmware?
How do you distribute the SOF firmware? I'm going to include those binary topology files to the SOF firmware package for Fedora. Perhaps, you may follow this.
Wouldn't it make more sense to distribute binaries along with confs from which they are build, which are already installed by alsa-topology package? Similarly Ubuntu could use alsa-topology-conf package...
From the packaging perpective, it would be another package which should be
forcefully installed on x86_64 containing few files (and it's unclear if the list will be extended in future). Also the .conf files are not required for the operation. But it's just my suggestion, other distributions can choose a different approach.
I just created Makefile [1] which may help with the packaging, because the topology .conf -> fw .bin relation is _unclear_ for may of us for those SST drivers. If you know more information (you can fill the rules for other topology files), please, share it with us.
Jaroslav
[1] https://github.com/alsa-project/alsa-topology-conf/blob/master/Makefile