crojewsk-intel
commented
10 months ago Hello @Gnoenk,
Thanks for the detailed report. That's quite a lot of effort you did there!
FYI the "error" messages you see when converting from UCM to BIN are info-messages, really. That is, alsatplg has no knowledge of what "lies ahead" when the dynamic approach is chosen. By dynamic I mean when the FE DAIs are initialized dynamically thanks to the topology plus CPUs are provided by the separate component driver which in this case represents SKL-H. Because it cannot know for sure, it dumps the messages for the user to interpret. Our interpretation: ignore.
In regard to Question 1:
Nov 12 15:10:51 kernel: avs_dmic avs_dmic: ASoC: no complete control IO handler for DMIC Volume type (g,p,i) 257:257:1 at 0x1990
Nov 12 15:10:51 kernel: avs_dmic avs_dmic: ASoC: failed to load widget DMIC FE
Nov 12 15:10:51 kernel: avs_dmic avs_dmic: ASoC: topology: could not load header: -22
Nov 12 15:10:51 kernel: snd_soc_avs 0000:00:1f.3: load topology "intel/avs/dmic-tplg.bin" failed: -22
Nov 12 15:10:51 kernel: snd_soc_avs 0000:00:1f.3: ASoC: error at snd_soc_component_probe on dmic-platform: -22your kernel is missing the peakvolume module support. The kernel driver encountered unknown/invalid (from its perspective) token so it dumped an error, ignored the file and continued on - HDAudio can be utilized like nothing has happened though. Peakvolume supported is long merged [1] into the Linus' tree, guess your base is simply older and has no support for that feature.
In regard to Question 2:
SKL-H devices usually default to non-DSP HDAudio stack - the DSP capabilities may be even fused out there. Without:
static const struct avs_acpi_boards i2s_boards[] = {
AVS_MACH_ENTRY(HDA_SKL, avs_skl_i2s_machines),
(...)you "only" lose I2S boards i.e.: sound cards where data is streamed over I2S interface. HDAudio/DMIC configurations are left intact. As I'm not aware of any SKL-H + I2C codec configurations available on the market, I believe the entry is not required.
Last but not least, Question 3:
That's yet another problem that is related to the older kernel base. shutdown() procedure was added to the driver later on [2] and wasn't part of the initial patchset.
[1]: https://lore.kernel.org/all/20221214185500.3896902-1-amadeuszx.slawinski@linux.intel.com/ [2]: https://lore.kernel.org/all/20230113190310.1451693-2-amadeuszx.slawinski@linux.intel.com/
Gnoenk
Dear @crojewsk-intel it is me again.
After sometime I could continue to work on my computer trying to enable DMIC functionality on my ACER Aspire-VN7-792G with a Skylake-H (ID: 0xa170) running under Linux Mint 21.1 now with Linux Kernel 6.2.
According to your advice in our conversation under https://gist.github.com/crojewsk/4e6382bfb0dbfaaf60513174211f29cb I have updated my Linux to Kernel 6.20 under MINT. My understanding was that the new AVS driver, which are now part of Kernel 6.x should then be part of the system and get activated automatically. As this did not happen I tried to force the activation by setting
snd_intel_dspcfg.dsp_driver=4in/etc/modprobe.d/alsa-base.conf. Result: I had no sound at all.Finding this issue https://github.com/thesofproject/avs-topology-xml/issues/5 in GitHub my next assumption was that despite the Kernel update performed the
linux-firmwarepackage was not up to date. Thus I manually downloaded the/lib/intel/avs/skl/*firmware from https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/ and installed it in/lib/firmware.Next step was to manually convert and install all the topology files from https://github.com/thesofproject/avs-topology-xml according to your instructions given there.
So I installed
alsa-utilsforalsatplgand builtavstplgmanually from https://github.com/thesofproject/avsdk/commit/999b25e3f0d7d7d7dbe4eb2496953d08779c2bff.Step 1: Conversion from XML to UCM by using
avstplgworked fine. Note: I had to use DOTNET 6 as I encountered error messages while using DOTNET 7 first.Step 2: Conversion from UCM to BIN by using
alsatplgshowed some messages. Should I be using a newer version ofalsatplgthan the one provided by MINT?Error Messages
Then I also faced the issue with the missing
hda-10ec0255-tplg.binfile mentioned in https://github.com/thesofproject/avs-topology-xml/issues/20 thus I created a symlink according to your advice.This is shown by the
treecommandNow I had completed all steps described but still no succes.
The command
results in:
Output
Question 1: as the log reports that
dmic-tplg.binfailed to load - does this means that the file conversion to bin failed?In parallel I searched for further errors.
In our former conversation you have written, that Skylake-H with ID 0xa170 has been added to the AVS driver. I did some research and could find that this ID got mentioned in your patch for the
core.cfile.https://lore.kernel.org/alsa-devel/20221010121955.718168-8-cezary.rojewski@intel.com/
Here I could read that you have added 0xa170 to
avs_ids(the list of supported boards?) incore.c.This file got modified a bit and the ID got moved into https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/linux/pci_ids.h and gets refered to by the Symbol
HDA_SKL.https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/sound/soc/intel/avs/core.c I can now find in row 757 a similar section using the symbols but no device ids.
So
core.cappears to be complete.The descripton page of https://github.com/thesofproject/avs-topology-xml has this note:
Thus I checked
board_selection.cin https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/sound/soc/intel/avs/board_selection.c and I find in row 293 only a line with the symbolHDA_SKL_LPwhich targets the low power version only.Question2: Does this still prevents SKL-H platforms from being included?
(Working Note: for the sake of completeness I have crosschecked the linux kernel source files under 6.2 and current main. Only difference is the shift to using symbols and addin the PCI-ID into the header file pci_ids.h).
Question 3: Additionally when shutting down the machine with the activated AVS-Driver I experienced issues at the end of the shutdown process. These disappeared when deactiving the option in
alsa-base.config.Grateful for your assistance in resolving this issue. Have a happy New Year!