Skip to contentSkip to conversation

ALSA: hda: snd_hda_scodec_tas2781_i2c: DSP firmware silently fails to load at cold boot (Lenovo Yoga Pro 9 16IMH9)

BUG Rámon van Raaij 1 participant 1 message May 1 Sound

lore
Share & tools
lore
#0 lore raw
Machine:  Lenovo Yoga Pro 9 16IMH9 (board SSID 83DN)
Kernel:   7.0.3-zen1-1-zen (linux-zen)
Codec:    Realtek ALC287, subsystem ID 17aa:38d6
Amps:     4x TI TAS2781, ACPI HID TIAS2781, I2C bus i2c-4
          (I2C controller: PCI 0000:00:15.2, Intel Designware)

Fixup: ALC287_FIXUP_TAS2781_I2C (applied correctly via HDA_CODEC_QUIRK 17aa:38d6)

PROBLEM
-------
At cold boot the built-in speakers produce ~10-15% of normal volume. The TAS2781 amplifiers are powered and the HDA component bind succeeds:

snd_hda_codec_alc269 ehdaudio0D0: bound i2c-TIAS2781:00

      (ops tas2781_hda_comp_ops [snd_hda_scodec_tas2781_i2c])

However, the DSP firmware (TAS2XXX38D6.bin) is not loaded. The ALSA control "Speaker Force Firmware Load" -- which is only registered after a successful firmware load inside tas2781_hda_comp_bind() -- is absent:

  $ amixer -c sofhdadsp cget name="Speaker Force Firmware Load"
  amixer: Cannot find the given element from control sysdefault:0

No error is logged to dmesg. The driver continues without DSP calibration data, leaving the amplifiers at uncalibrated default gain.

The issue does NOT occur after S3 suspend/resume. After resume, a modprobe -r / modprobe cycle restores full speaker volume.

ROOT CAUSE
----------
The BIOS initialises the TAS2781 amps during POST and leaves them in a hardware state incompatible with the kernel firmware load sequence. The request_firmware_nowait() call in the bind path silently fails when called against hardware in this BIOS-initialised state.

Critical observation: a PCI-level power cycle of the I2C controller (0000:00:15.2) reliably resolves the issue:

echo 1 > /sys/bus/pci/devices/0000:00:15.2/remove sleep 2
echo 1 > /sys/bus/pci/rescan

After this sequence ACPI gates the I2C controller power rail (D3cold), resetting the TAS2781 amps to factory state. On rescan, i2c_designware and snd_hda_scodec_tas2781_i2c re-probe, DSP firmware loads, and "Speaker Force Firmware Load" appears in amixer.

A plain modprobe -r / modprobe cycle WITHOUT the PCI power cycle does NOT fix the issue, confirming this is a hardware state problem, not a module loading or firmware subsystem timing race.

WORKAROUND
----------
The following systemd service (After=sound.target, Before=display-manager.service) reliably restores correct volume on every boot. It must run before PipeWire opens the device; hot-removing the PCI device while WirePlumber has an active ALSA handle causes a SIGSEGV in snd_hctl_elem_get_interface (libasound).

  [Service]
  Type=oneshot
  ExecStartPre=/bin/sleep 2
  ExecStart=/bin/bash -c 'echo 1 > /sys/bus/pci/devices/0000:00:15.2/remove'
  ExecStart=/bin/sleep 2
  ExecStart=/bin/bash -c 'echo 1 > /sys/bus/pci/rescan'
  ExecStart=/bin/bash -c 'modprobe snd_hda_scodec_tas2781_i2c || true'
  ExecStart=/bin/sleep 5
  RemainAfterExit=yes

SUGGESTED FIX
-------------
The driver should detect a failed firmware load in tas2781_hda_comp_bind() and either:

a) Perform a hardware reset sequence on the TAS2781 amps (e.g. via an ACPI power state transition on the I2C controller) before retrying, or

b) Schedule a deferred retry after the firmware subsystem has fully settled, with a fallback hardware reset if the retry also fails.

The fix should live in tas2781_hda_comp_bind() or the request_firmware_nowait() completion callback in snd_hda_scodec_tas2781_i2c.