dg2 / a750 HDMI Audio doesn't work w/ options snd_hda_codec_hdmi enable_silent_stream=Y

EDIT: My hardware is not working with enable_silent_stream on ANY kernel and works with silent stream disabled on all I have tested.

Running Siduction (based on Debian unstable 'sid'). Testing different kernels from Ubuntu mainline repository. HDMI audio works on 6.2.0 6.2.1 and 6.2.2 but doesn't work on any 6.2.{x>2} or the 6.3 rc's

Alsamixer looks the same, dmesg logs grepped for hda look the same. The device is there but no sound is produced.

Attaching dmesg logs with drm.debug=0xe log_buf_len=4M set from 6.2.2 and 6.2.3

(6.2.2.dmesg.log

6.2.3-dmesg.log

closed

reopened

After reading through closed issue #7353 (closed) I added options snd_hda_codec_hdmi enable_silent_stream=N to a modprobe .conf file.

This makes hdmi audio work on all the kernels I have tested. (6.2.3 6.2.8 6.3.0-rc3).

Looks like there has been a regression since the fix applied to close that issue.

changed title from dg2 / a750 HDMI Audio doesn't work on kernels 6.2.3 and newer to dg2 / a750 HDMI Audio doesn't work w/ options snd_hda_codec_hdmi enable_silent_stream=Y on kernels 6.2.3 and newer

added Community feature: display/audio platform: DG2 labels

@kaivehmanenintel Any comment about the regression ?

Thank you @mosshope for detailed report. There has not been any changes to audio side between 6.2.2 and 6.2.3 so this quite interesting. Changes in audio and i915:

$ git log --pretty=oneline v6.2.2...v6.2.3 -- sound |grep hda
f500c21ae286d754e88137cd822f06e698a9a9f1 ALSA: hda/realtek: Add quirk for HP EliteDesk 800 G6 Tower PC
94f48aed2c9f244d7626416b6b028ed956763f0e ALSA: hda/hdmi: Register with vga_switcheroo on Dual GPU Macbooks
87ca47acc793ec061df95373f1023a23ffe8e886 ALSA: hda: Fix the control element identification for multiple codecs
59473f6b70c470cb6db0fbfe9bb174314810e338 ALSA: hda/ca0132: minor fix for allocation size

$ git log --pretty=oneline v6.2.2...v6.2.3 -- drivers/gpu/drm/i915/
f8c6d7f805855095b79beb510b5746c5890ae6e1 drm/i915: Don't use BAR mappings for ring buffers with LLC
f8354e04e5faac765fef967c24772e358c426cef drm/i915: Don't use stolen memory for ring buffers with LLC
21c7c846279fad6e245d32f343336ae021853b6c drm/i915/quirks: Add inverted backlight quirk for HP 14-r206nv
e7e072cfa50839aefdf1b94e23298b264cbc8c3f drm/i915/mtl: Correct implementation of Wa_18018781329
0873f089822920b603fabf366fcf9aa6df432974 drm/i915: Fix GEN8_MISCCPCTL
34802bbf1fd433d910769551c7a395918dafd19e drm/i915/pvc: Annotate two more workaround/tuning registers as MCR
d483c9a2a4394967412a713a8674ea725fbce1a3 drm/i915/pvc: Implement recommended caching policy
7db51a15a99d954eed093960a39be4255a65866d drm/i915/xehp: Annotate a couple more workaround registers as MCR
76d433431d858522303725a463ae05b689f3fe93 drm/i915/xehp: GAM registers don't need to be re-applied on engine resets
5008c43a7187d2c83949e8d0159aea18f3eccb4f drm/i915/mtl: Add initial gt workarounds

Nothing really pops out here. Let me try to reproduce locally. If this is easy to hit, git bisect should reveal the commit where this starts failing.

@mosshope Hmm, I could not replicate this at least yet. Silent streaming is working both with 6.2.3 and 6.2.3 on the DG2. If you are familiar with the tools, would it be possible for you to do a git bisect on the kernel between 6.2.2 and 6.2.3 to identify the problematic commit? Meanwhile I'll check whether this has been on other machines.

I have the same problem on an A770. Unfortunately I am a bit too busy at the moment to dig into this and properly debug the problem. In two weeks I might have more time available to look into this and maybe do git bisect.

@AndrewAmmerlaan What type receiver (model/make) you are using? HDMI as well?

I have a TV (HKC-43F3) connected with HDMI and a monitor (MSI Optix MAG342CQR) connected with DisplayPort. Both of them don't receive audio unless options snd_hda_codec_hdmi enable_silent_stream=N is set.

changed title from dg2 / a750 HDMI Audio doesn't work w/ options snd_hda_codec_hdmi enable_silent_stream=Y on kernels 6.2.3 and newer to dg2 / a750 HDMI Audio doesn't work w/ options snd_hda_codec_hdmi enable_silent_stream=Y

Not a kernel change issue in the end. Built 6.2.2 and 6.2.1 and found with =Y wasn't working. So I checked what the value was set there on the kernels I was testing before without an override and 6.2.2 had N and 6.2.3 had Y.

Looks like ubuntu mainline repo builds starting with 6.2.3 have CONFIG_SND_HDA_INTEL_HDMI_SILENT_STREAM=Y set because that will be the new config moving forward.

change in config in lunar repository: https://git.launchpad.net/~ubuntu-kernel/ubuntu/+source/linux/+git/lunar/commit/debian.master/config?h=master-next&id=92eaaa79ed5eb0ceb7663b40fd9ab5a6ba487812 bug tracker that inspired the change https://bugs.launchpad.net/ubuntu/+source/linux-oem-6.1/+bug/2009024

changed the description

Is there a way I can try to help get enable_silent_stream working on more hardware by testing against this not working hardware? Seems like it'll be important to have working or fail gracefully if Ubuntu is going to ship it turned on with their first release with a kernel supporting dg2 out of the box.

LMK about enabling debug things or sending me a branch to build for more information.

@khfeng FYI, one problem found with CONFIG_SND_HDA_INTEL_SILENT_STREAM=y (related to our discussion in #8097 (closed) ).

Thanks @mosshope , we'll try to replicate this on some setup. We'd need to figure out what is the delta to #7353 (closed) .. is this specific to A750 or the receiver (you have a LG TV in this case). alsa-info output both in working and non-working cases would help to get started (for alsa-info, see https://manpages.ubuntu.com/manpages/focal/en/man1/alsa-info.sh.1.html ).

Interesting that you mention the silent setting enabled is the default on windows. Did have windows with the a750 for a while and hdmi audio was working

Here's the alsa-info output

working-noHdmiSilent-alsainfo.txt

nosound-withHdmiSilent-alsainfo.txt

For me it is (still) working on Windows.

the specific lg tv is OLED55B6P

I have a few other devices I can try.

• Denon AVR - no sound... device is detected but no sound or video. I'd write a bug here, but it also was happening on windows and even bios doesn't show up.

• I have a little tendak HDMI to vga converter that has a 3.5mm out. That seems to work with either value for enable_silent_stream

• Tomorrow I'll see if I get time to move computer and try it to my desktop monitor.

• Some point next week I will probably be putting together a system with an old 7700k for a friend, so I will try to test that's igpu on the TV, to determine if it's a750+lg tv or snd_hda_intel+lg tv thats having issues.

Thanks @AndrewAmmerlaan @mosshope for quick info. The fault is probably still on A750/770 end, but it will help to speed up debug if this can be pointed down to a certain type/make of receiver. So far in our local tests, we've been not able to repro this yet.

One quick thing to try is to disable runtime-PM but keep silent-stream enabled. E.g by adding following to /etc/modprobe.d/dg2.conf (or any conf file):

# 1h timeout (for testing)
options snd_hda_intel power_save=3600

Also, I'd be interested in snd-hda-codec-intel logs (via dmesg/journalctl). These can be enabled with

options snd_hda_codec dyndbg=+p
options snd_hda_codec_hdmi dyndbg=+p

Again to some file under in /etc/modprobe.d/. Reboot, or module reload ("modprobe -r snd_hda_intel ; modprobe snd_hda_intel") to take these into effect. Logs should have entries like:

[101002.043567] snd_hda_codec_hdmi hdaudioC1D2: HDMI: supports coding type LPCM: channels = 2, rates = 32000 44100 48000, bits = 16 20 24
[101002.043574] snd_hda_codec_hdmi hdaudioC1D2: hdmi: enabling silent stream pin-NID=0xe cvt-NID=0x3
[101002.044228] snd_hda_codec_hdmi hdaudioC1D2: hda_codec_setup_stream: NID=0x3, stream=0xf, channel=15, format=0x11
[101002.044626] snd_hda_codec_hdmi hdaudioC1D2: hda_codec_setup_stream: NID=0x3, stream=0xf, channel=0, format=0x11
[101002.046357] snd_hda_codec_hdmi hdaudioC1D2: HDMI: KAE 1 cvt-NID=0x3

One quick thing to try is to disable runtime-PM but keep silent-stream enabled. E.g by adding following to /etc/modprobe.d/dg2.conf (or any conf file):

With runtime-PM disabled and silent-stream enabled there is again no audio sent to the displays.

Also, I'd be interested in snd-hda-codec-intel logs (via dmesg/journalctl). These can be enabled with

dmesg-withHDMIsilent.log dmesg-withoutHDMIsilent.log journalctl-withHDMIsilent.log journalctl-withoutHDMIsilent.log

Disabling runtime-PM didn't change anything for me either.

I tried adding those to a modprobe file . But I'm not seeing any log entries like the ones you mentioned. Any ideas? debug.conf

Thanks @mosshope and @AndrewAmmerlaan . As runtime-pm-disable did not have effect, that narrows down the problem. Codec programming looks also ok (via alsa-info log). So there is some problem with the switch from fake to real data. I'll try to get this reproduced in some local setup.

For not getting log entries -> you probably don't have dynamic-debug enabled in your kernel (CONFIG_DYNAMIC_DEBUG=y). https://www.kernel.org/doc/html/latest/admin-guide/dynamic-debug-howto.html

@mosshope @AndrewAmmerlaan Here's one patch to try (this on top of 6.2.3 but should apply clean on any recent kernel):

https://github.com/kv2019i/linux/commit/74530f053139525ae2b712231b647c2d2dd7c672

This in theory should have no difference, but it's an alternative flow the hardware supports, so worth a try.

Patch didn't change anything. Heres the dynamic debug enabled log from my attempt.

dmesg_silentY_withPatch.log

Thanks @mosshope . Logs show the patch took effect as expected, but alas no impact.

Here's one more patch to try: https://github.com/kv2019i/linux/commit/7eea8ed43615df0a23eab86271cfdef3861771a2

This will add a drain of the display pipeline just before real streaming starts.

No luck. I built from your branch directly so had both patches. Let me know if you want me to try and reverting the previous patch. dmesg_silentY_withPatch2.log

Thanks @mosshope . This is truly puzzling now. We could try old-style keepalive on your system, patch here: https://github.com/kv2019i/linux/commit/38bc9ca07e358b2321d66e8105e1e76054a3d01d

Thanks, I'll give this a try soon.

Thanks @mosshope . This is truly puzzling now. We could try old-style keepalive on your system, patch here: https://github.com/kv2019i/linux/commit/38bc9ca07e358b2321d66e8105e1e76054a3d01d

This somehow fixes the problem on my end.

For some reason your branch of the kernel did not boot at all on my PC, so I instead applied your patch to the branch of the kernel I was already running.

Mine works with that patch as well.

Thank you @mosshope and @AndrewAmmerlaan for testing. I have not managed to reproduce this on my end, so debug will likely take more time as I need to acquire some affected hardware for debug. As the impact is rather high, I now submitted a patch to revert back to old silent stream (e.g. the patch you tested) to upstream: https://mailman.alsa-project.org/hyperkitty/list/alsa-devel@alsa-project.org/thread/JZNOOAQLDO6GAIMREUDMODYFTW22LV7P/ ... if reviews go through fine, this should land to distro kernels as well.

mentioned in commit nouveau@6ab6f98f

mentioned in issue #8462 (closed)