Repository navigation
arm64: dts: apple: t8142-j815: speakers and sound card - #1534
Conversation
J815 has six SN012776 speaker amplifiers where J813 has four. Add the two woofers at 0x3a and 0x3d and name all six as in speakersafetyd's j815.conf. Feed them in ADT speaker-config order, and tell the AOP audio service it has six amplifiers (apple,speaker-amps). The card is apple,j815-macaudio. Enable the AOP audio service and the card. The speakers stay at macaudio's safe level until speakersafetyd holds them. That needs the J815 profile. Depends on the J815 audio series on top of Linux aurora-silicon#1423 (the apple,j815-macaudio and apple,speaker-amps bindings). Tested on J815 (MacBook Air 15-inch, M5) with the equivalent node set in a J815 tree that included the whole J813 board. All six amplifiers probe and each speaker answers on its own sense pair. Capped speaker playback and the microphones work. The headphone jack codec probes; wired headphones have not been tried. Assisted-by: Claude Opus 5.5 Signed-off-by: Nick <nekreshen@yandex.ru>
DjDeveloperr
left a comment
There was a problem hiding this comment.
Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.
Keep draft/dependency hold: this carries the shared J815 DT extraction plus AOP speaker graph and depends on reviewed1423/1533/1532 changes and the capped J815 profile. The profile source may land separately without enabling hardware. Headphone behavior is explicitly untested; do not promote it or generic audio to Working solely when this graph merges. Preserve the existing protection/interlock and no-v2 shipment hold. CI is approved for evaluation; no playback or hardware changes were performed here.
DjDeveloperr
left a comment
There was a problem hiding this comment.
Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.
Updated the stack with normal merges of reviewed DT parent1532@b180163d, audio parent1533@824c7c22 and main955ba62c, preserving the original speaker-graph commit and all parent histories. Final head ea5ab50 is draft and merges cleanly.
The six codecs retain their original order, shared reset151, 250 BCLK ratio and distinct voltage/current sense slots. Both original audio-disable tail overrides are removed as intended by this PR; the graph is not accidentally left disabled by the DT-parent merge. The full four audio/schema blobs equal reviewed1533, including ownership flags, caps and interlocks. Other reviewed DT/schema/shared CPU/OPP/actuator/clocks/ISP memory/N1/alias inputs remain unchanged, and main’s three PMGR/runner paths are preserved.
Twelve schema documents/meta/reference checks, twelve targeted ABI controls, symbolic six-codec graph checks and whitespace checks pass. No kernel/DT build or physical playback was performed here. Keep draft: fresh stack/parent CI, the existing three-object audio validation and author-required J813 DT boot remain gates. Original J815 sweep/playback/microphone evidence predates this integration; headphones remain unqualified. No protection bypass or new v2 shipment is authorized.
Preserve both board configurations and their distinct tweeter HPF grants. Keep main protection and admission logic unchanged.
DjDeveloperr
left a comment
There was a problem hiding this comment.
Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.
The current DT failure is validation, not setup: all137 DTBs built and binding/ADT checks passed. Four new warnings identify the omitted seventh T8142 GPU region; the fifth is the deliberate legacy alias. Fixed the GPU binding in parent1532@9122abe4 and normally merged it into this stack.
Also resolved the newly merged J714 audio overlap against main990e00b7. The final union reconstructs exact main after removing the original J815 config/comment/match, preserving all main admission/interlock logic. J714’s HPF grant stays true, J815’s false and its safe cap unchanged. Actual C config/match GCC/UBSan controls pass, and four mutants compiled then aborted for lost/leaked HPF grants, relaxed J815 cap and wrong J714 match. Sound binding checks pass; GPU controls32/non-target comparisons768 pass. Original six-codec graph and AOP Rust source/schema remain unchanged.
Published head5fac88d61add13bb698148efabc69213d784352a, normal history-preserving merges, both branch refs verified. Fresh CI is running; do not claim the compiled warnings cleared yet. Keep draft: the existing alias/loader-adoption/J813 boot and audio-object gates remain. Old824c whole-C object proof cannot qualify this new union; no new kernel/DT build, playback or hardware test was performed.
DjDeveloperr
left a comment
There was a problem hiding this comment.
Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.
The conflict notification is transient: GitHub currently reports this exact head mergeable. Current main b96b0fc adds the ANE series after this head’s integrated main990; its changed paths do not overlap any of this PR’s 20 changed paths, so no branch rewrite is needed.
Actual head validation is now available: Host job 130813161188 passed. DT job 130813162666 built all 137 base/head DTBs successfully, passed all ten touched bindings (including GPU, AOP audio and macaudio), and passed the 19 coverage-checker tests plus all six board comparisons. Warnings fell 400→391: ten removed, one new. The only new warning is J815 aliases/disp0_piodma, the same loader-compatibility gate tracked in #1532/aura#165. Keep it draft; do not rename/filter that alias before compatible-loader boot/adoption qualification. This is DT/binding validation, not audio object or physical playback qualification. Existing audio safety and hardware limits remain.
Retain current main's ISP binding, including the T8140/Neo sensor-domain order and T8142 eight-domain case. The audio code and sound bindings from aurora-silicon#1533 have already landed and now match main exactly. Preserve all added and removed lines in the remaining 15 DT and binding feature files, including the seven-region T8142 GPU schema and underscore loader alias. Keep compatible-loader qualification/adoption and shared J813/J815 DT and protection qualification as draft gates. Eight binding documents pass metaschema/reference checks. The processed GPU schema passes 32 current controls and 32 main-profile controls. This merge does not claim a compiled DTB, new hardware boot or physical audio. Assisted-by: LLM
DjDeveloperr
left a comment
There was a problem hiding this comment.
Resolved the current-main conflict in c83c100 (normal merge; no history rewrite). Kept main's complete ISP binding, preserving the Neo sensor-domain order and T8142 eight-domain case. The landed #1533 audio code and sound bindings now match main exactly and disappear from the PR diff. All added/removed lines across the remaining 15 DT/binding feature files are unchanged from 5fac88d; the original J815 speaker graph and underscore loader alias remain intact.
Local validation: eight binding documents pass dtschema metaschema/reference checks; the processed GPU memory ABI passes 32 candidate controls and 32 current-main controls. No new kernel/DTB build or hardware run. The completed #1533 audio-object validation is retained; no duplicate audio compilation is needed.
Keep draft. Fresh Host/DT runs 43142783549/43142783350 are queued. The known alias warning is not filtered: Aura #165 needs affected T8142 loader boot qualification and compatible-loader adoption before canonical alias migration. Required J813/shared-DT boot and J815 graph/protection/physical rollout evidence remain outstanding. The original pre-integration hardware report is retained at its stated scope; headphones remain unqualified.
Keep main's complete ISP binding, preserving the T8140/Neo sensor-domain order and T8142 eight-domain case. All added and removed feature lines in the remaining 15 DT and binding files match the preceding parent. The complete bindings directory matches the validated aurora-silicon#1534 integration: eight binding documents and 32 candidate/32 main GPU schema controls. No repeated schema test or new compiled-DTB/hardware result is claimed. J815 audio stays disabled here. Keep the underscore loader alias until compatible-loader qualification/adoption; no warning filtering. Preserve published ancestry and require fresh current-head Host/DT CI. The required J813 shared-DT boot and existing loader/rollout gates remain. No boot, audio playback, image build or protection-policy change. Assisted-by: LLM
Inherit the updated parent1532 J813 factory-address target nodes and its already-published J316s NFC main ancestry. Preserve the complete J815 six-codec sound graph and all other child source unchanged. This normal merge retains both published histories; source checks prove only the J813 N1 fix and the inherited main NFC node differ from the previous child. No audio/protection/loader/module-option change or new hardware qualification is claimed. Require fresh CI for both layers; keep the existing loader-alias and physical rollout limits. Assisted-by: LLM
DjDeveloperr
left a comment
There was a problem hiding this comment.
Normal merged the updated #1532 parent223d7d72 asdf3087d0 in the same atomic publication. This inherits @gig3m's exact J813 N1 factory-address block and the parent's existing J316s NFC ancestry. Relative to the preceding child, those are the only two changed files; J815's complete six-codec sound graph and every other child file remain exact. Relative to the updated parent, the only difference is the original J815 sound-graph DTS.
The J813 first-commit owner-check gate is now satisfied at report6828183980's five-boot/composed-tree scope. The new aliases/PCI nodes have bounded symbolic preservation and J815-topology checks, with no SEP alias, experimental Bluetooth option or audio/protection change. GPU/SEP/default-Bluetooth qualification is not implied.
Fresh CI is required for this head. Keep draft: Aura165's affected-loader qualification/adoption and coordinated canonical-alias migration remain, as do the J815 graph/protection/physical rollout limits. No new audio build, playback, loader installation or physical trial was performed.
Inherit the updated parent/main with aurora-silicon#1527's always-on PMP in the shared PMGR description. Apply the original J815 sound graph without changing its codec slots, resources or protection configuration. The child audio net patch is exact; both old histories are retained. Keep the J813 owner-check evidence at its stated composition scope and retain the loader-alias qualification/adoption and J815 physical gates. No new audio build, playback, loader installation or hardware test. Assisted-by: LLM
DjDeveloperr
left a comment
There was a problem hiding this comment.
The df3087d DT failure is the known alias gate: base/head137 DTBs and make rc=0,402/394 warning lines, one new/nine removed; the sole new warning is disp0_piodma. All seven changed bindings and ADT19/six comparisons passed. No new GPU/N1 failure or unchanged-source retry is inferred.
The actual conflict follows #1527's newly merged PMP retention, overlapping the parent's PMGR layout. Parent e8fb89d keeps that layout and main's complete ps_pmp block, supplier and always-on flag. Child260d5e15 normally merges the updated parent and preserves the exact previous J815 sound net patch; its whole J815 DTS is byte-identical to df3087d. The J813 N1 DTS remains exact in both layers. An explicit three-way trial matches the prepared child tree, and both refs were atomically published without force.
Symbolic parent/child J813/J815 checks confirm the PMP at0x598, SRAM supplier, always-on and correct factory aliases. Fresh hosted CI is required; no new kernel/DTB/audio build or physical/loader/playback action. Historical #1533 objects are not relabeled as a current-main whole-source build.
Keep draft for Aura165 affected-loader qualification/adoption and coordinated canonical-alias migration, plus the existing J815 physical rollout limits. The J813 owner-check gate stays closed at its reported composed scope; it is not requested again.
Inherit the owner-provided J813 ALS node without changing the existing audio graph or SEP profile/policy delta. Preserve both histories. Fresh CI and the independent parent/loader/review gates remain. Assisted-by: LLM (Codex)
Retain the full audio or SEP feature while inheriting the updated parent/main, complete MCC block and semantics-preserving duplicate enum repair. Keep both published histories and all review/loader qualification gates; require fresh CI. Assisted-by: LLM (Codex)
DjDeveloperr
left a comment
There was a problem hiding this comment.
Resolved the main conflict with normal merges: parent1532@0434a61e, audio1534@770097fd and SEP1732@98e509f0. Existing published histories remain; the fork parent/audio refs were updated atomically without force, then the SEP sibling inherited the parent.
The EOF conflict now keeps the complete old J815 board bodies as byte-exact prefixes and adds main’s complete MCC block. This avoids the shared-closing-brace trap in a plain conflict concatenation; the initial incomplete candidate was caught before publication and retained as evidence. Main’s MCC geometry is enabled on J815 and inherited disabled on J813 through the shared extraction. Main RTC/rtc-offset nodes are also carried. PMP retention, N1/ALS, the complete audio graph and the exact two-file SEP/profile feature are preserved. Independent source review found no concrete regression.
Metaschema validation caught a duplicated agx-g17s entry already present in main; removed only its redundant trailing enum value. The allowed set and other conditions are unchanged. Eight exact-tree schema documents pass, as do five T8142 memory cases and560 finite non-target comparisons against current main. Six board/layer symbolic three-way checks and three actual SEP KUnit bodies in host Rust extraction pass. These are source/schema/host checks, not new compiled DTB/kernel/QEMU or hardware qualification.
Main’s intervening673f6184 change touches only J700’s DTS, disjoint from every reviewed/tested input; exact identity permits reusing those results while carrying it. Fresh parent Host43378989747/DT43378989936, audio Host43378989659/DT43378989669 and SEP Host43379009670/SEP43379009612/DT43379009604 are queued. Old green/alias-only results are historical for these new heads.
All drafts and Aura165 affected-loader qualification/adoption, coordinated alias migration, final CI/review and physical qualification limits remain. No warning filter, unchanged-source retry, new hardware/playback, loader/default change or release action occurred.
Carry the current-main prerequisites and @gig3m's tested J813 PMU/MCC/AVE enablement. Preserve the existing audio graph feature delta and all current shared definitions. The board file merged cleanly. Symbolic three-way property checks preserve both board descriptions and the parent enablements. Existing loader-alias and physical/policy qualification gates remain. No hardware, audio playback, XART write or loader/default change occurred. Assisted-by: LLM Signed-off-by: Diljit Singh <hey@djdev.me>
Retain the original feature commits and pull in the current-main GPU binding merge. The J815 board feature file remain byte-identical and four symbolic board/layer comparisons preserve both inputs. Existing parent-first, loader-alias and physical/policy gates remain. No new kernel/DTB/hardware or loader/default operation was performed. Assisted-by: LLM Signed-off-by: Diljit Singh <hey@djdev.me>
|
@DjDeveloperr out of draft. It stays parent-first on #1532, which now has the display-alias fix (beb68d4, merged in here). J815 sound runs under the merged capped speakersafetyd profile (aurora-silicon/speakersafetyd#2) with the UCM from aurora-silicon/alsa-ucm-conf-asahi#14. Please review it alongside #1532, with whatever sound gate you want kept. Assisted-by: Claude Opus 5.5 |
DjDeveloperr
left a comment
There was a problem hiding this comment.
Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr's behalf at f4d8f49e616d6b0b45e1403821edd91f90a7b390.
The audio delta over parent #1532@beb68d4e remains one J815 DTS file, byte-identical to the audio file at 9637b4e9. The parent alias change is retained and J813's property map is unchanged.
I checked the merged source dependencies:
- speakersafetyd #2, merge
813f1374: six speaker names/order, 12 V/I channels, -24 dB woofer / -32 dB tweeter caps, and the Left Woofer 1 delay correction. - UCM #14, merge
96b3206d: six playback channels, jack PCM 0, speaker PCM 1 and the J815 card/control names. - Driver #1533 is merged as
12f3c1bf.
A bounded DTS/profile/UCM cross-check passes for all six codec names, playback order, sense slots/channels, cap/model fields and PCM identifiers. This closes the earlier missing-profile source-availability concern. It does not establish which versions a test image has installed or replace the profile author's hardware measurements.
Ready-for-review status is appropriate. I would retain these landing gates:
- Parent #1532 first, including its canonical-only loader qualification/adoption, and fresh required CI. Host CI passed; DT CI was running when checked.
- The actual deployed audio image uses the merged capped J815 profile and UCM, with the normal kernel speaker lock/protection interface intact. Please pin the installed versions/commits and link any existing protected-playback and protection-withdrawal results. The separately reported raised-ceiling experiments are outside this review.
- Wired-headphone playback/routing remains unqualified in UCM #14's report; please identify any later result. Keep microphone-route and jack results at their reported source/composition scope.
Text-only status/result summaries suffice. No new speaker playback, capture, volume increase, service interruption or hardware operation was performed or requested by this source review.
Carry parent aurora-silicon#1532's strict T6030 compatible-chain schema fix. The J815 audio board file and the net one-file audio delta remain unchanged. Signed-off-by: Diljit Singh <hey@djdev.me> Assisted-by: LLM
|
Carried parent #1532's T6030 ALS binding correction into this stack as normal merge 48916b4, parents The merge changes only the ALS schema relative to the prior audio head. The net audio delta over the new parent remains one J815 board file, with the same |
|
Both checks on The software CI gate is closed. Parent #1532's canonical-only loader qualification/adoption remains pending, along with the requested deployed-protection/version evidence and wired-headphone playback/routing result. No new physical-audio qualification is inferred from these checks. |
The J815 is a T8142 Mac like the J813: same t8152_aop_audio T8142 transport, but six SN012776 amps (LW1, RW1, LT, RT, LW2, RW2) with [V, I] sense per amp, and AID43 tuning with SPP3 protection and no LSID. SPP3 is in units_at_parity, so the J815 qualifies the way the other SPP3 Macs do, and gate.json needs no entry for it. The fixture is wiring.py's reading of the compiled t8142-j815.dtb (aurora-silicon/linux#1534) plus AID43's lane structure, with generic box names and no tuning values. gen.py produces sgv2-j815.conf and 60-sgv2-j815.conf from it. The derived wiring hash, 725f518662e42388, is the one the Mac's own self-check proved from the live device tree and the owner's bundle. On that Mac, protection was then granted per stream. docs/J815-INTEGRATION.md records what was checked on the hardware and what is still open. Assisted-by: Claude Opus 5.5
|
A status update on the remaining gates, from the J815 owner side. Speaker protection: in place and running on the shipped image.
Physical audio with protection active. The ceilings were raised one step per boot, each step only after the previous one's thermal log looked right:
So the thermal governor demonstrably limits a real music load. The speakers are fine after all of it, and they're in daily use since. The jack and microphone cards register as described ( Wired headphones: not tested. We have no wired headphones to test with, so that row stays open, as alsa-ucm-conf-asahi#14 already notes. The jack detection controls are present ( Loader: aura #165 (the canonical T8142 display PIODMA alias) is still open. It's the one gate here we can't close from the J815 side. If it still has to land first, this can wait for it. If the rest looks fine to you, could this go in, with the wired-headphone row recorded as untested? The audio file is unchanged from the head CI passed on ( Assisted-by: Claude Opus 5.5 |
DjDeveloperr
left a comment
There was a problem hiding this comment.
Thanks, Nick. I re-checked this against the current tree, with the speaker rule in mind (nothing audible except through speakerguardd, and limits only ever tightened).
What the PR is now. #1532 merged as 1f38ff8 at your 8e1ffb8, so against its merge base this PR is one file, t8142-j815.dts (+62/-14):
- the two second woofers on
i2c3; - the six-codec order;
apple,speaker-amps = <6>and the J815 chassis name;- removing the two
status = "disabled"overrides.
Speaker safety.
- No kernel limit changes here.
macaudio_j815_cfg(merged with #1533) keepssafe_vol = -20. The SN012776 volumes stay at the floor (max = 1) until a protection daemon unlocks them or speakerguardd grants a lease. With no amp levels from the loader, macaudio disables the speakers. - The slots match the protection side. The order (LW1, RW1, LT, RT, LW2, RW2) and the V/I pairs (vmon below imon: 0/2, 4/6, 8/10, 12/14, 16/18, 20/22) are what speakerguardd's
docs/J815-INTEGRATION.mdand itsj815.jsonfixture describe. That fixture was derived from this DTB, and it records protected playback granted and released per stream on J815 hardware.
Merging. The base has moved a lot in this file since 8e1ffb8 (the lid-angle node, ANE, JPEG). A local three-way merge is clean, and the merged file has exactly one sound node and no leftover disabled override. The last CI run is from 11-15, so I'll update the branch and merge once dt-checks and host-tests are green on the merged head.
These rows come after the merge, with a revert ready:
- one J815 boot on the merged kernel: speakerguardd's self-check derives wiring
725f518662e42388, and protection is granted and released for a stream. Per the project's speaker rule, at 5% sink volume through speakerguardd only; - wired-headphone playback and routing, which UCM #14 leaves untested.
ce7d682
into
aurora-silicon:integration/aurora
|
Merged as ce7d682 (dt-checks and host-tests green on the updated head 804c36b). Post-merge verification as in my review: one J815 boot on the merged kernel where speakerguardd's self-check derives wiring |
Enable the J815's AOP-fed six-speaker audio graph: four woofers and two tweeters, twelve V/I sense channels, the CS42L84 jack and J815 card names. This is a one-board-file follow-up to #1532; the net audio change remains in
t8142-j815.dts.Ready for review at
48916b45e1bef58f449a9afc7ee6238db33abd36(tree5c299ea9ee1aba53bf0f6527efb3f7963c362d8c), stacked on #1532@8e1ffb81. Audio qualification remains pending.Parent #1532 has merged as
1f38ff8f4dc93088f7073c2a16906eb56f74a363. The loader qualification and adoption tracked in Aura #165 and the audio evidence below still gate this follow-up.The current head merges the parent's canonical display alias while preserving the audio board file byte-for-byte from
9637b4e9. J813 remains unchanged. The six codec names/order and sense mapping agree with the merged J815 speakersafetyd profile; the UCM uses six playback channels, jack PCM 0 and speaker PCM 1.Source dependencies:
12f3c1bf.813f1374: J815 profile with -24 dB woofer / -32 dB tweeter caps. Its per-speaker hardware qualification is reported there; raised-ceiling experiments are outside this PR's qualification.96b3206d: J815 speaker/jack and three-channel microphone card descriptions.Validation: the bounded DTS/profile/UCM model matches all six codec identities, playback and V/I ordering, capped model fields and PCM identifiers; parent aliases and J813 properties remain intact. Current-head Host CI 43680098078 passes 321/321 runs. DT CI 43680098049 builds 137 baseline and 137 head DTBs, both make return 0, with zero new warnings (1,482→1,465; 17 removed). All seven changed bindings and ADT's 19 tests/six coverage comparisons pass. Hosted testing used merge
3995f278against integrationfc050503. Source/CI results do not establish physical audio operation; earlier hardware evidence retains its stated version scope.Landing remains gated on compatible-loader qualification and adoption, fresh required checks, the actual image shipping the capped profile/UCM with kernel protection intact, and physical-audio evidence at stated version pins. aurora-silicon/alsa-ucm-conf-asahi#14 explicitly leaves wired-headphone playback untested; any later result should be recorded. Microphone routing and protection-withdrawal evidence retain their own scope.
Original bring-up and owner reports: @m0d16l14n1. Integration/source review on @DjDeveloperr's behalf assisted by Codex. Earlier detailed validation remains in the PR's review timeline.