[BUG] ASoC/AMD ACP70/RT721: ASUS TUF Gaming A14 FA401EA speakers silent and headphone jack buzzes on playback

[Cristian Timohi]

****
Resending as plain text, as the previous message was rejected by the
vger.kernel.org mailing list due to an HTML MIME part.
Apologies to recipients who received the previous copy directly.
***

Hello,

I'm reporting an audio issue on an ASUS TUF Gaming A14 FA401EA
(Ryzen AI MAX+ 392) using AMD ACP70 SoundWire and a Realtek RT721
SDCA codec.

I have reproduced the issue on a clean current vanilla 7.3-rc1
kernel, so this does not appear to be Fedora-, PipeWire-, or
desktop-specific.

Current clean test environment
==============================

Distribution:
    Fedora 45

Kernel:
    7.3.0-0.rc1.260901.786262be.417.vanilla.fc45.x86_64

snd_acp_sdw_legacy_mach quirk:
    -1

There are no remaining custom snd_acp_sdw_legacy_mach/modprobe
overrides in the current reproduction environment.

Hardware
========

Laptop:
    ASUS TUF Gaming A14 FA401EA

DMI product:
    TUF Gaming A14 FA401EA

DMI board:
    FA401EA

BIOS:
    FA401EA.304

CPU/APU:
    AMD Ryzen AI MAX+ 392

Audio:
    AMD ACP70 SoundWire
    Realtek RT721 SDCA

The SoundWire bus exposes one peripheral:

    sdw:0:1:025d:0721:01: Attached

I do not see TAS2783 devices enumerated on this unit.

Known-good cases
================

The laptop's built-in speakers worked correctly under the
factory-installed Windows 11.

USB audio also works correctly under Linux. I tested Corsair
Virtuoso USB headphones and normal audio playback works.

Therefore the issue appears specific to the internal
AMD ACP70 -> SoundWire -> RT721 audio path.

Actual behavior
===============

Internal speakers:
    Completely silent.

3.5 mm headphone jack:
    No intelligible audio is produced.
    When the analog audio path becomes active, the headphones produce
    a continuous buzzing noise. The buzz stops when the playback/path
    is deactivated.

Opening the desktop sound control panel can also trigger the buzzing
while the analog audio path is active.

ALSA playback devices
=====================

The amd-soundwire card exposes:

    device 0:
    SDW1-PIN0-PLAYBACK-SimpleJack rt721-sdca-aif1-0

    device 2:
    SDW1-PIN1-PLAYBACK-SmartAmp rt721-sdca-aif2-2

Direct ALSA reproduction
========================

Internal speakers:

    speaker-test -D hw:CARD=amdsoundwire,DEV=2 \
        -c 2 -r 48000 -F S16_LE -t wav

The PCM stream starts normally without an ALSA error, but there is
no physical speaker output.

Analog headphones:

    aplay -D plughw:CARD=amdsoundwire,DEV=0 \
        /usr/share/sounds/alsa/Front_Left.wav

No speech is heard. The headphone output only buzzes while the
audio path is active.

Zero-PCM test
=============

I also sent only zero-valued PCM samples:

    timeout 10s bash -c '
    dd if=/dev/zero bs=192000 2>/dev/null |
    aplay -D hw:CARD=amdsoundwire,DEV=0 \
          -t raw -f S16_LE -r 48000 -c 2
    '

The headphone jack still buzzes while the zero-PCM stream is active.

Therefore the buzz does not appear to represent corrupted application
audio; merely activating playback is sufficient to produce it.

Mixer / UCM
===========

Speaker:

    Speaker Switch = on
    rt721 FU06 Playback Volume = 87,87

Headphones:

    Headphone Switch = on
    rt721 FU05 Playback Volume = 87,87

I also explicitly enabled the corresponding Speaker and Headphones
UCM devices. This did not change the result.

ASoC DAPM during speaker playback
=================================

I inspected debugfs while speaker-test on the SmartAmp PCM was
actively running.

All relevant widgets were ON:

    Speaker: On
    rt721 DP3 Speaker Playback: On
        stream DP3 Speaker Playback active
    rt721 DP3RX: On
    rt721 FU 21: On
    rt721 FU 23: On
    rt721 PDE 41: On
    rt721 SPK: On

The DAPM graph therefore appears to activate the entire expected
speaker path:

    DP3 Speaker Playback
      -> DP3RX
      -> FU 21 / FU 23
      -> SPK
      -> Speaker

but the physical speakers remain silent.

Kernel messages
===============

RT721 attaches successfully.

I see:

    rt721-sdca sdw:0:1:025d:0721:01:
    Driver needs fixing: Unsorted reg_defaults, sorting the copy

On an earlier 7.2.x boot I also observed:

    soundwire sdw:0:1:025d:0721:01:
    PARITY error detected before INT mask is enabled

I did not observe additional SoundWire/RT721 errors when starting
or stopping playback.

Kernels tested
==============

The same issue was reproduced on:

    Fedora 44:
    7.2.2-300.fc44.x86_64

    Fedora 45:
    7.2.2-300.fc45.x86_64

    Vanilla mainline:
    7.3.0-0.rc1.260901.786262be.417.vanilla.fc45.x86_64

    CachyOS kernel:
    7.2.2-cachyos1.fc45.x86_64

The current reproduction is on the clean vanilla 7.3-rc1 kernel.

Previous experiments
====================

Before testing clean mainline, I also built a Fedora 7.2.2 kernel
locally and experimentally changed the RT721 speaker endpoint in:

    sound/soc/amd/acp/amd-acp70-acpi-match.c

from:

    .aggregated = 0
    .group_id = 0

to:

    .aggregated = 1
    .group_id = 1

This did not change the behavior.

I also tested that custom kernel with:

    options snd_acp_sdw_legacy_mach quirk=32768

The kernel reported:

    amd_sdw amd_sdw: Overriding quirk 0x1 => 0x8000

but there was still no speaker output.

Those experimental overrides have been removed. The current vanilla
7.3-rc1 reproduction uses quirk=-1 and no custom modprobe options.

Question
========

Could this FA401EA variant require a different RT721-only topology,
ACP70 configuration, SoundWire setup, or codec initialization quirk?

The fact that all-zero PCM causes the analog headphone output to buzz
while the stream is active may be particularly relevant.

I'm happy to test patches or provide register/debug output.

I can also provide full alsa-info, kernel journal, lspci output,
SoundWire sysfs state, mixer state, and DAPM debugfs output.

Thank you.
Cristian