[BUG] ADL / ThinkPad T14 Gen 3 (21AH): playback position stuck at 0 - audio audible, all video playback frozen (IPC3 & IPC4, fw 2023.12 & 2025.12)

[VirginiaBalseiro]

Describe the bug
On a Lenovo ThinkPad T14 Gen 3 (Intel, Alder Lake, model 21AHCTO1WW) running Ubuntu 24.04, whenever the SOF driver is active (dsp_driver=3), audio playback is audible but the reported playback position never advances past 0. Because video players sync video to the audio clock, all video playback system-wide freezes on the first frame (browser, mpv, everything) while the audio track plays normally. Progress bars stay at 0:00.

The internal DMIC works correctly under SOF. The bug is only in playback position reporting.

With the legacy HDA driver (snd-intel-dspcfg dsp_driver=1) everything works except the internal DMIC (expected, since it requires the DSP path).

To Reproduce
Boot with SOF active (default on this hardware, or options snd-intel-dspcfg dsp_driver=3).
Play any file with an audio track, e.g. mpv -v test.mp4.
Audio is audible; video frame and OSD timeline never advance.

mpv verbose log shows the status line frozen for the entire duration, yet playback "completes":

[statusline] AV: 00:00:00 / 00:00:10 (0%) A-V:  0.013   (repeated for the whole file)
[cplayer] EOF code: 5
[cplayer] finished playback, success (reason 3)

Occasionally, immediately after restarting the PipeWire stack (systemctl --user restart wireplumber pipewire pipewire-pulse), a freshly started stream advances briefly (~1 s) before the position stops again.

Reproduction Rate
100% on every player and browser tested.

Expected behavior
Playback position advances normally under SOF, as it does under the legacy HDA driver, so A/V sync works.

Environment

1. Branch/commit:
   • Kernel: distro kernels, not git builds — Ubuntu 6.8.0-136-generic and 7.0.0-28-generic (HWE); bug identical on both
   • SOF: release binaries, not git builds — firmware-sof-signed 2023.12.1-1ubuntu1.11 (Ubuntu) and sof-bin-2025.12.2 release tarball (IPC3 v2.2-stable line and IPC4 v2.12 line both tested); bug identical on all
2. Topology: sof-hda-generic-2ch.tplg
3. Platform: Intel Alder Lake (ADL-P), driver binds as sof-audio-pci-intel-tgl — Lenovo ThinkPad T14 Gen 3 (21AHCTO1WW), ALC257 + DMIC

Screenshots or console output
Attached full logs, collected while SOF was active and the bug reproducing:

alsa-info.txt

sof-dmesg.txt

sof-logger trace not available (distro install); happy to collect additional logs on request.

Workarounds tried (none fixed it)

api.alsa.disable-mmap = true + api.alsa.headroom = 1024 (WirePlumber rule)
Pinning the sink rate: audio.rate = 44100, audio.allowed-rates = [ 44100 ]; also pw-metadata -n settings 0 clock.force-rate 44100 (bug occurs at both 44.1 kHz and 48 kHz)
session.suspend-timeout-seconds = 0 (no device suspend)
snd_hda_intel power_save=0
Full PipeWire/WirePlumber restarts; fully restarted client apps

Only effective workaround: options snd-intel-dspcfg dsp_driver=1 (legacy HDA) — restores video/audio sync but loses the internal DMIC.

[lgirdwood]

@VirginiaBalseiro I assume this is the case for all endpoints like speakers, headphones and HDMI ?
@ujfalusi fyi

[ujfalusi]

@VirginiaBalseiro , I have not seen such issue so far.
I have installed Ubuntu 24.04.4 (defaults to 7.0.0-28-generic kernel) to a TGL-H HDA laptop - same DSP arch as ADL.
playing a video with mpv shows correct progress and video is playing correctly as well.
Tested with IPC3 and IPC4 as well.

The PCM position reporting is identical among SOF IPC3, IPC4 and legacy HDA, I'm not sure what might cause the issue you are seeing.

Do you have pauvucontrol running?
What PCMs are open? for pcm in /proc/asound/card*/pcm*; do echo ${pcm}; cat ${pcm}/sub0/status; done

Few things to check:
Use the audio hardware directly: mpv --audio-device=alsa/hw:0,0 test.mp4
Try the developer 'DSPless' mode: options snd_sof sof_debug=0x8000 DMIC will not work, pipewire might get confused as it will fail with DMIC probing, but the analog HDA endpoints should work via hw:0,0

Note: The position reporting is identical on all Intel platforms regardless of DSP architecture, the same registers are used on the Linux side of the HDA controller to read the DMA position.