[RyanTheTide]

J616s has bounded proxy microphone, headphone and individually confirmed six-speaker evidence, but no qualified Linux audio ABI. Add an indexed sound bring-up reference describing each observed route, native amplifier order, the implementation sequence and first-Linux inventory/acceptance requirements.

The CS42L84 driver already sets the microphone-detection latch required by working headset capture. Its J616s AOP/machine integration still needs live validation; T8140/J700 stream geometry, jack codec and MAX98360A feed-forward model are not compatibility evidence. Speaker feedback packing/scales and calibration remain unknown. The paired five-second proxy test timed out during late amplifier shutdown after successful mute/TX restoration; reliable cleanup remains required.

This draft changes documentation only. It adds no device-tree nodes, compatible entries, mixer settings, PCM numbering or speaker enablement. It keeps Asahi speakersafetyd I/V protection and the kernel volume interlock as prerequisites.

All four remaining individual tests pass matching DMA reports, zero residue/DART faults, protected shutdown and exact controller/GPIO restoration. The shared Python I2C observed-status clear is tested directly on both idle speaker buses; this does not establish a cure for the intermittent STOP fault or whole-transfer concurrency safety. Simultaneous all-six playback, continuous audio and calibrated speaker protection remain unqualified.

Evidence: m1n1 bring-up and driver contracts pinned at fd360ef69054e42724b975d7a4c61e0242b6b341. Source review uses current aurora-wip d41b89b, including RyanTheTide's merged audio PR13, HQ prepare PR38, config PR32 and CS42L84 remote PR9. Upstream Asahi asahi-wip is pinned at 94fb23346d522edf53722357c426a3e58030beea.

Validation: git diff --check; standalone docutils RST parse without warnings; changed-path review confirms only the new sound document and its index entry. No kernel build or Linux audio runtime qualification is claimed.

Next code step: a T6040 AOP control-only provider profile with profile-specific bootarg overrides, own EPICv2 service discovery and retained DMA on unconfirmed shutdown. Gate automatic legacy audio-child probe: its 632-byte Rust PDM upload and lpai property301 setter do not match the qualified J616s 626-byte getter/non-writable301 ABI. This is recorded as the concrete provider/child blocker; no driver or device-tree enablement is added.

[RyanTheTide]

Consolidated into PR #231 at a4e7ee8. The sound-card index blob is identical. The J616s document is retained with a focused update replacing the old implementation instructions with the implemented profile-validation and DMA-retention behavior; the remaining document is identical. Closing duplicate documentation tracking preserves this history and does not qualify the gated J616s hardware work in #231.