fixed "HDMI ATI/AMD: no speaker allocation for ELD" on Radeon module

This bug is missing log files that will aid in diagnosing the problem. While running an Ubuntu kernel (not a mainline or third-party kernel) please enter the following command in a terminal window:

apport-collect 1993217

and then change the status of the bug to 'Confirmed'.

If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

4.9 is quite old. Please try mainline kernel v6.0.

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

bug present in vmlinuz-5.15.0-50
I admit I did not try 6.x kernels.

Searching in 4.9, because doing a git-bisect to find the offending commit in that version.

bug present (HDMI audio deactivated, off) in kernel 6.0.1 from https://kernel.ubuntu.com/~kernel-ppa/mainline/v6.0.1/amd64/linux-modules-6.0.1-060001-generic_6.0.1-060001.202210120833_amd64.deb

paul@rabachi:~/colinKernels/new4.9.180/my4.9.180$ git bisect log
git bisect start
# bad: [a5f56b52c878585b12b8bc37f737dcce4a660c64] Linux 4.9.178
git bisect bad a5f56b52c878585b12b8bc37f737dcce4a660c64
# good: [8baec4ebdf084961516f17cadbad14cac082ee4e] Linux 4.9.177
git bisect good 8baec4ebdf084961516f17cadbad14cac082ee4e
# bad: [fef85fb00224cb22f9efec262e9ffe8970e555ee] mm/mincore.c: make mincore() more conservative
git bisect bad fef85fb00224cb22f9efec262e9ffe8970e555ee
# good: [45fe6deee2f2b8889368c5fa5ac832b571c6ea52] sched/x86: Save [ER]FLAGS on context switch
git bisect good 45fe6deee2f2b8889368c5fa5ac832b571c6ea52

Compiling
commit f934bf3f57769282c371c978eb9bbe3930f6ed35 (HEAD)
Author: Wenwen Wang <email address hidden>
Date: Sat Apr 27 01:06:46 2019 -0500

    ALSA: usb-audio: Fix a memory leak bug

    commit cb5173594d50c72b7bfa14113dfc5084b4d2f726 upstream.

    In parse_audio_selector_unit(), the string array 'namelist' is allocated
    through kmalloc_array(), and each string pointer in this array, i.e.,
    'namelist[]', is allocated through kmalloc() in the following for loop.

The commit isn't related to HDMI audio though...

Can you please try
$ git checkout f934bf3f57769282c371c978eb9bbe3930f6ed35 # then compile and test

$ git checkout f934bf3f57769282c371c978eb9bbe3930f6ed35^ # then compile and test

So we are sure it's the offending commit?

Sorry I did not meant f934bf3f57769282c371c978eb9bbe3930f6ed35 was the offending commit.

This is just where I am at testing with git bisect.

There should need about 3 others steps to find the offending commit.

Found the offending commit:
Did not verify this, but I suspect this is the case:
    After this change, the driver will not report Jack_is_plugged event
    through hdmi_present_sense_via_verbs() if monitor_present is 1 and
    eld_valid is 0.

paul@rabachi:~/colinKernels/new4.9.180/my4.9.180$ LANG=en git bisect replay bisect.log
Already on 'master'
Your branch is ahead of 'origin/master' by 4 commits.
  (use "git push" to publish your local commits)
72cd3810e52c606595f9e1a2d442a4cc6e6a3551 is the first bad commit
commit 72cd3810e52c606595f9e1a2d442a4cc6e6a3551
Author: Hui Wang <email address hidden>
Date: Mon May 6 22:09:32 2019 +0800

    ALSA: hda/hdmi - Consider eld_valid when reporting jack event

    commit 7f641e26a6df9269cb25dd7a4b0a91d6586ed441 upstream.

    On the machines with AMD GPU or Nvidia GPU, we often meet this issue:
    after s3, there are 4 HDMI/DP audio devices in the gnome-sound-setting
    even there is no any monitors plugged.

    When this problem happens, we check the /proc/asound/cardX/eld#N.M, we
    will find the monitor_present=1, eld_valid=0.

    The root cause is BIOS or GPU driver makes the PRESENCE valid even no
    monitor plugged, and of course the driver will not get the valid
    eld_data subsequently.

    In this situation, we should not report the jack_plugged event, to do
    so, let us change the function hdmi_present_sense_via_verbs(). In this
    function, it reads the pin_sense via snd_hda_pin_sense(), after
    calling this function, the jack_dirty is 0, and before exiting
    via_verbs(), we change the shadow pin_sense according to both
    monitor_present and eld_valid, then in the snd_hda_jack_report_sync(),
    since the jack_dirty is still 0, it will report jack event according
    to this modified shadow pin_sense.

    After this change, the driver will not report Jack_is_plugged event
    through hdmi_present_sense_via_verbs() if monitor_present is 1 and
    eld_valid is 0.

    Signed-off-by: Hui Wang <email address hidden>
    Cc: <email address hidden>
    Signed-off-by: Takashi Iwai <email address hidden>
    Signed-off-by: Greg Kroah-Hartman <email address hidden>

:040000 040000 0280e65d8aca2561b6ed8ee27383d1cd0e3cc022 ebe57ab9f50fdb2851d074f5d2c95f5e74beb541 M sound
paul@rabachi:~/colinKernels/new4.9.180/my4.9.180$

I believe this is the situation discussed in:
https://gitlab.freedesktop.org/drm/amd/-/issues/1569

Also, I tested master head this morning (showing 6.0.0-rc6, and it was broken.

I had tried a reverse of this commit and previous one, but it was broken too (I probably did not made the right choices when seeing conflicts).

This might be the same problem as:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1834771

attaching enhanced workaround (I added LANG=C) from: https://askubuntu.com/questions/1032811/ubuntu-18-04-no-hdmi-audio-with-amd-radeon-hd-7870/1207919#1207919

briefly tested (work) patch that manually reversed commit 72cd3810e52c606595f9e1a2d442a4cc6e6a3551 applied on HEAD of master on git://git.launchpad.net/~ubuntu-kernel-test/ubuntu/+source/linux/+git/mainline-crack

It would be good to verify if after applying the patch, the old bug that it was fixing come back (that after resuming from sleep, on some hardware, shows 4 (many) entries in preferences/sound/hardware [either because of a bug in TV BIOS or in GPU code]. I did not took time to test that.

So yes, it's a GFX driver bug. The GFX driver should always report correct eld_valid so HDA driver can rely on it.

I think the real problem is there:
https://gitlab.freedesktop.org/drm/amd/-/issues/1569#note_1601169

v.4. of the patch https://gitlab.freedesktop.org/drm/amd/ ... t-v4.patch published yesterday, have been accepted by the radeon maintainer: https://lore.kernel.org/all/CADnq5_N-4t ... gmail.com/ I tested it yesterday on top of 6.0.1-rc2 kernel and it was working fine.

Previous link for the patch may be wrong, here a new one:
https://gitlab.freedesktop.org/agd5f/linux/-/commit/34d84636e5e09521d031863a01f9b9fb22ffa875

The previously accepted patch have received itself a new patch that use a more fine grained mutex.

@Paul_Dufresne, many thanks for driving the fix to this longstanding bug. It's been present for over three years and likely affects millions of users.

Following your work here: https://gitlab.freedesktop.org/drm/amd/-/issues/1569#note_1601169, it appears there's finally going to be a real, permanent correction upstream. Dancing in the streets is unlikely, but a vast number of users will get happy their HDMI audio works again, largely thanks to your efforts.

This should stand as an example of how even a single contributor who is dedicated to the task can make a huge difference.