M3-AUDIO-02 — corrected PCM notification candidate

Post-deployment update: AUDIO-02 is owner-flashed and now has clean 44.1 kHz physical playback, including the period-notification correction. The ALSA tools are now installed persistently on SD. Use standard aplay/amixer paths; the temporary-tool instructions below record the original artifact handoff and are superseded for this device.

Owner manual deployment only. Built on 2026-09-10 from the startup fix in 137b3f2, with release 6.18.0-y2linux-m3audio02. M3-AUDIO-01 produced audible headphone sound with clicking; clean playback is not yet qualified. This fixes positive ALSA constraint success skipping substream attachment and therefore period notifications. No route, clock, voltage, gain, memory or DT source change.

Flash this image

/home/luca/Dokumente/Code/Y2Linux/out/y2linux-m3-audio-02/BOOTIMG.img

Size: 4816896 bytes. SHA256:

b9dbf6b528e4afbee2a5ffb2fc65df7f6a076564e92960332e24cee3809b5842

1. In the established SP Flash Tool scatter/download-agent setup choose

   Download Only. Deselect everything, then select BOOTIMG only.

2. Select the exact file above. PRELOADER, LK and every other partition remain

   unchecked; never choose Format All or Firmware Upgrade.

3. Owner powers off the Y2, clicks Download and uses the established connection

   sequence. Wait for success, disconnect USB, boot with the existing SD, then attach USB once after startup. The assistant does not flash or reboot.

No SD update is required for this kernel correction. The matching DRM module is embedded in the rescue initramfs and loaded before SD handoff. Existing SD content and SSH keys remain usable. SD is still DEV-01 without permanent ALSA utilities: after boot we restage the reproducible temporary ALSA bundle via SSH. The previously built M3-AUDIO-01 rootfs remains the optional persistent utility update; no rootfs rebuild/image is produced for AUDIO-02. USB reconnect #27 stays deferred. Do not reuse AUDIO-01 BOOTIMG for this correction.

First inspection and controlled playback

ssh -i /home/luca/.ssh/y2linux_ed25519 [email protected]
uname -r
cat /proc/asound/cards
cat /proc/asound/pcm
cat /sys/bus/platform/devices/11220000.audio-controller/state
dmesg

Expected release: 6.18.0-y2linux-m3audio02; card Y2Audio, PCM0 Y2 Headphone Playback, hw:CARD=Y2Audio,DEV=0. Initial format/rates remain stereo S16_LE at 44100/48000 Hz. Once tools are restaged at /tmp/y2-m3-audio-aa3c949 and the owner confirms headphone readiness:

cd /tmp/y2-m3-audio-aa3c949
./amixer -c Y2Audio cset name='PCM Ch Select' 0
./amixer -c Y2Audio sset Master 207,207
./amixer -c Y2Audio sget Master
./amixer -c Y2Audio sset Headphone on
./aplay -D hw:CARD=Y2Audio,DEV=0 --period-size=1024 --buffer-size=8192 audio/headphone-44100.wav
./amixer -c Y2Audio sset Headphone off

Master 207 is -24 dB; the three-second fixture peaks at -30 dBFS with fades. During the supervised test use the retained shell watchdog/cleanup pattern: old SD has no timeout utility. Correlate playback elapsed time, PCM hw_ptr, AFE IRQ/period count and DMA CUR, CON3/CONN0, XRUNs, regulator/ASoC state and dmesg; confirm VGP2 1.8 V and no DAC reset. Require clean owner listening feedback, then short stop/restart/repeated runs and 48 kHz before accepting wired audio. A watchdog-triggered exit0 is not a completed WAV. Speaker/jack/high-resolution paths remain separately unqualified.

Reproduction and validation

Reuse AUDIO-01's isolated kernel output via an independent reflink/copy into a fresh AUDIO-02 directory, and copy its Buildroot target and three rescue helpers. Run the canonical configure.sh (only LOCALVERSION differs), then one incremental make -C /src O=/build/kernel -j12 Image zImage modules through `tools/build/run.py --output out/y2linux-m3-audio-02`. Reuse pinned userspace; do not rebuild Buildroot. Then in that isolated environment, from /project:

python3 -m tools.build.dev_initramfs
sh tools/build/dtb.sh
python3 -m tools.validation.dev_artifacts /build --package
Y2_ARTIFACT_TEST_ROOT=/build python3 -m unittest discover -s tests -p 'test_audio*.py' -v

PASS: four targeted production-code/DT tests; exact ELF/Image/zImage, rescue module ABI, RAM/decompressor/reservation layout, eMMC-disabled DT and BOOTIMG validation. Resolved config equals AUDIO-01 except LOCALVERSION; rescue helper hashes are unchanged. Full build/validation logs retained as compressed evidence. This is offline validation, not physical acceptance of the correction.