[m0d16l14n1]

Add the T8142 six-amplifier speaker shape selected by an explicit apple,speaker-amps = <6> property, its J815 macaudio configuration and the matching sound bindings. Existing four-amp defaults and T6050 ownership remain unchanged; unknown shape/config combinations are rejected.

The J815 configuration retains the -20 dB safe cap, existing ADT level/interlock checks and no tweeter-HPF grant. Before microphone-card registration, the shared output service is initialized to standby; this starts no DMA and unmutes no codec. Current J714 configuration, HPF grant and admission/protection logic are preserved. The speakersafetyd J815 profile has merged separately; that is not DT or whole-system rollout qualification.

Current source ab0a715b10f4caa4228297d2c1e0f76455984941 passed both actual hosted checks: Host 43111512499 and DT 43111512482. Exact-source local validation also compiled all three AArch64 objects with zero warnings/errors: t8152_aop_audio.o (267584 bytes), aop.o (808760 bytes) and macaudio.o (470824 bytes). Required Rust/module configuration, source/tree/blob seals and bindgen checksum passed. The parent independently verified object sizes, SHA256s, ELF architecture and clean source. Toolchain: Clang 23.1.1, Rust 1.99.0, bindgen 0.73.2. This is object compilation, not full module/kernel linkage or a new hardware test.

Actual config/match GCC/UBSan controls, compiled ownership/HPF negative controls and strict sound-binding checks retain their earlier source scope. Current main has no subsequent changes in these four feature files or their AOP/audio interfaces. The normal main merge retains original authorship and the independently reviewed J714/J815 union.

This source-only layer does not enable the J815 board graph. #1532/#1534 DT, loader and physical pairing/qualification gates remain separate. Original reported J815 playback/sense/microphone results below predate this integration and are not relabeled as new-head hardware results. No new playback, recording, loader/default or image operation was performed.

Original feature description and hardware report:

• Acelogic's "initialize J813 output standby before microphone registration" (cherry-picked, authorship kept)
• dt-bindings: macaudio: J815
• dt-bindings: t8152-aop-audio: apple,speaker-amps
• t8152-aop-audio: the six-amplifier T8142 speaker stream (4 woofers + 2 tweeters, one V/I sense pair each); each amp's level comes from the ADT apple,speaker-config as in J813: AOP sound on integration/aurora (speakers, jack, microphones) #1423
• macaudio: J815

Tested on J815 with speakersafetyd (J815 profile: aurora-silicon/speakersafetyd#2): a low-level sweep put every speaker on its expected sense pair (R 3.4–4.1 ohm); sustained music playback with the thermal model logging; the three-mic array records through the UCM/DSP path. Happy to fold this into #1423 instead if that is easier.

Part of the J815 (MacBook Air 15-inch, M5, T8142) bring-up: board results are summarised in discussion #135. Tested on a J815 with a macOS 27.0 stub (firmware 27.0) through aura Stage 2.

Assisted-by: Claude Opus 5.5

[DjDeveloperr]

Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.

Keep draft until its foundation/dependencies and safety gates are complete. Source review found that1423's mic PCM lacked the worker/DMA synchronization already present in this child; I am carrying that narrow flush/sync_stop subset into1423 first so the foundation does not briefly ship without it. This does not approve a six-amp board graph, higher speaker caps or unqualified playback. The capped speakersafetyd J815 profile source is reviewed/merged separately; that alone does not clear kernel/DT/Asahi DSP pairing or headphone/microphone qualification. Contributor CI approvals are now unblocked for actual results.

[DjDeveloperr]

Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.

The original head’s five checks pass and #1423 has merged through its owning M5 lane. Integrated current main925131cc on this child branch as824c7c22 without changing #1423. The net change against that main is four files: two bindings, the J815 macaudio table/match, and the six-amp Rust configuration/property/standby changes.

The clean textual merge exposed an actual Rust initializer error: main added ap_mca/ap_leap, but HW_CFG_T8142_6AMP omitted them. A production-config extraction compiled the actual struct and all four initializers and reproduced E0063 before the fix. The initializer now retains the existing T8142 true/true AP ownership; fixed config/selector tests pass, and compiled ownership mutations fail runtime assertions. Existing four-amp defaults and T6050 false/false ownership are unchanged.

Source review preserves #1423’s mic worker flush, PCM/component sync_stop, current AOP ring/API contracts and all existing macaudio interlock/ADT-level/cap handling. The six-amp mode is selected only for the T8142 config with explicit4/6 counts; unknown shapes are rejected. J815 safe_vol remains-20. COUT initialization before microphone registration is standby only, before userspace exposure; it adds no DMA start or codec unmute. Both sound schemas pass strict document/reference checks. These are source/config checks, not complete hardware qualification.

Keep draft. Fresh current-head CI and actual AArch64 t8152_aop_audio.o, aop.o and macaudio.o validation are pending. I queued exactly one pinned-source build under the owner’s shared kbuild lock, nice19/ionice-c3/-j2, with disk-floor guards; it must not run alongside the current lock holder. SEP Rust/KUnit checks are not substituted for the audio-object build. The original J815 protection/sense/microphone report retains its old source scope; no new playback, recording, enrollment, loader/default or hardware operation was requested. J815 DT1534/1532 rollout gates remain separate.

[DjDeveloperr]

Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.

Both824c7c22 Host/DT checks passed, but main’s new J714 audio support introduced a real source conflict. Normally merged main990e00b7 as ab0a715 using the independently reviewed union already prepared for1534: whole main C/admission/protection logic plus the original J815 config/comment/match. J714’s HPF grant stays true and J815’s false with its safe cap unchanged. AOP Rust and its schema are unchanged from824c; four audio/binding files remain above main.

Actual C config/match GCC/UBSan controls and four compiled runtime mutants pass; strict sound-binding controls pass. Kernel-code directory IDs for drivers/sound/rust/include/scripts exactly match1534@5fac88d6, so a single verified build of these common inputs can support their shared audio source without duplicate compilation. This is source identity, not an object-build pass.

Keep draft for new-head CI and exact AArch64 audio/AOP/macaudio object validation. The previous manager exited before building, and only its orphaned lock waiter was retired; the keeper’s active lock/build was not touched. One new bounded validation task will queue normally with nice19/ionice-c3/-j2 and8GiB reserve-floor rules. Existing DT/physical/protection limits remain; no playback, recording, loader/default or other-build operation was performed.

[DjDeveloperr]

Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.

Current-head Host and DT checks passed, as already recorded. The separate three-object validation did not reach source verification, configuration or compilation: after acquiring the shared build lock, its storage guard found 13,307,084,800 bytes free against the required 13,958,643,712 bytes (8 GiB free floor plus a 5 GiB initial peak estimate). It stopped before writing source/build data and released its own lock.

This is a local validation/storage blocker, not a compiler diagnostic or a failed public CI check. No object, kernel, module or hardware pass is claimed. Current source remains ab0a715; retain draft status until actual object validation and final review. No automatic retry, other-lane interruption, reserve use or deletion of source/build/final/rollback evidence occurred. The existing CI and physical/audio-protection scopes remain unchanged.

[DjDeveloperr]

Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.

Final source approval for ab0a715. The remaining software gates are closed: both actual Host/DT checks passed, and the pinned-source validation compiled t8152_aop_audio.o, aop.o and macaudio.o as AArch64 objects with zero warnings/errors. Source/tree/blob seals, required Rust/module config and bindgen checksum passed; I independently verified object bytes/SHA256/ELF proofs and clean source. The shared lock was released, no worker/queue/lease remains, disk floor and emergency reserve were preserved. No repeat build is needed.

The four-file source diff preserves main’s complete J714/admission/protection code and the original capped J815 configuration, separate match and no HPF grant. T8142 alone accepts the explicit four/six amplifier selector, preserving existing AP ownership and other-SoC defaults. COUT setup before microphone registration is standby only, with no DMA start or codec unmute. Previously reviewed config/selector/schema and compiled ownership/HPF controls apply unchanged; current main has no subsequent feature-file or relevant AOP/audio-interface changes. The original J815 sense/playback/microphone report retains its stated earlier-source scope.

This approves the source layer, without claiming full module/kernel linkage or new physical audio qualification. J815 board DT enablement and #1532/#1534 loader/physical rollout gates remain separate; no speaker, recording, loader/default or image operation was requested. Draft status can now be removed for this source merge.