Skip to content

arm64: dts: apple: t6050-j775c: give the AOP its setup mailbox (stacked on #2102) - #2542

Closed
satchlj wants to merge 1 commit into
feat/j775c-rebased-20261116from
satchlj/j775c-aop-setup-mailbox-20261120
Closed

satchlj wants to merge 1 commit into
feat/j775c-rebased-20261116from
satchlj/j775c-aop-setup-mailbox-20261120

Conversation

@satchlj

@satchlj satchlj commented Nov 20, 2026 •

Copy link
Copy Markdown
Member

Claude Opus 5.5 on @satchlj's behalf.

Stacked on #2102 (the J775c device tree); review the last commit. arm64: dts: apple: t6050-j775c: give the AOP its setup mailbox.

Problem. On lobata's cand-4 (integration/aurora e0ee46a1 + 80 commits, including #2102's J775c DT), apple_aop 290c00000.aop: probe with driver apple_aop failed with error -61, at boot and again on a manual bind. Without the AOP there are no AOP services and no J775c sound card.

  • The tip's aop.rs matches apple,t6050-aop with HW_CFG_T6050, which opens a mailbox named setup.
  • This file's AOP had mbox-names = "mbox" only.
  • Everything else was in place: Aura's T6050 handoff enabled the AOP, its mailbox, DART and ADMAC; the DART, mailbox and ADMAC were bound; both fw_devlink supplier links were available.

Change.

  • The file already described this mailbox, but as mailbox@4d067c000 (apple,t6050-secure-mailbox), and nothing referenced it.
  • J775c's ADT has it at 0x2d067c000 (aop-exclave-mailbox, size 0x4000, interrupts 0x249 and 0x246). That is where t6050-peripherals.dtsi puts aop_setup_mbox for the laptops.
  • This commit describes it as that dtsi does and adds it to the AOP: mboxes = <&aop_mbox>, <&aop_setup_mbox>; mbox-names = "mbox", "setup";.
  • Its iommus = <&aop_dart 1> adds dart-aop stream 1, so Aura's handoff also checks that stream's locked table.

All values are from J775c's ADT or the tip's T6050 DT (interfacing data).

Testing

  • dtc compiles the file. The AOP's mboxes holds two phandles, the second being mailbox@2d067c000, and the 4d067c000 node is gone.
  • Untested on hardware. It passes only once a J775c boot shows apple_aop binding and Aura's handoff accepting stream 1. That boot is lobata's next candidate, and I'll post the kernel identity and the result here.

Assisted-by: LLM (Anthropic:Claude Opus 5.5)

On T6050, apple-aop opens the AOP's setup port by name ("setup"), as
t6050-die0.dtsi describes for the laptops. The J775c file gave the AOP
only its ASC mailbox, so the probe fails with -ENODATA and J775c gets no
AOP services, its audio included.

The file already had the mailbox, but as mailbox@4d067c000, an
"apple,t6050-secure-mailbox" that nothing referenced. The ADT has it
at 0x2d067c000 (aop-exclave-mailbox, size 0x4000, interrupts 0x249 and
0x246), where t6050-peripherals.dtsi puts aop_setup_mbox. Describe it
as that dtsi does, on dart-aop stream 1, and add it to the AOP's
mboxes.

Signed-off-by: Satya Benson <satya@benson.earth>
@satchlj

satchlj commented Nov 20, 2026

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf: J775c row on this head: the DT part works, and the probe now fails one step later. It times out on the setup port instead of failing the lookup.

Kernel: lobata's cand-5, d2f8af83bf98 = integration/aurora 09c7ad0eb4d6 (0 behind at 09:00Z) + 82 commits. It includes #2102's J775c DT, this commit, and a local diagnostic print in pcie-apple's s2idle check, unrelated to the AOP. Booted 2026-11-20 09:05Z.

Aura's T6050 handoff now accepts stream 1 and enables the mailbox:

AOP: T6050: dart-aop SID 1: TCR 0x9 TTBR 0x10fffba1 (root 0x10fffba0000), ADT l1-tt 0x10fffba0000, tables [0x10fffba0000, 0x10fffbc0000): ok
AOP: T6050: enabled /soc/mailbox@2d067c000
AOP: T6050: handed off (5 nodes, dart-aop streams 0x103)

Kernel:

  • 2d067c000.mailbox is bound to apple-mailbox, in IOMMU group 9.
  • The AOP no longer fails with -61:
apple_aop 290c00000.aop: RTKit: Initializing (protocol version 12)
apple_aop 290c00000.aop: setup port: endpoint map incomplete before the AP power request
apple_aop 290c00000.aop: setup port: boot incomplete (AP power on: false, endpoints: 0/0): ETIMEDOUT
apple_aop 290c00000.aop: probe with driver apple_aop failed with error -110

So nothing answers on the setup port: 0 endpoints, no AP power ack.

Facts that may matter (J775c ADT, interfacing data): the node is aop-exclave-mailbox@D067C000, compatible iop,secure-rtbuddy-proxy, with exclave-assigned and role AOP-EXCLAVE. On this desktop, the setup port may not be served to the AP the way the laptops' is. lobata's default kernel binds this AOP through the apple,t8152-aop fallback, which uses no setup port, and its AOP services and sound card work.

Status: this PR is correct as a description of the hardware, but on its own it does not make J775c's AOP probe. Options I see:

  1. If J775c's AOP firmware runs without the setup port, aop.rs could treat the setup port as optional, selected by whether the DT gives the AOP a setup mailbox, not by board. J775c would then drop the setup entry. That's a driver question for whoever owns HW_CFG_T6050.
  2. The setup port needs a step on J775c that the laptops don't. I can't tell which without more tracing.

I'll keep this PR open until that is settled, and won't claim a working AOP from it.

🤖 Generated with Claude Code

@satchlj

satchlj commented Nov 20, 2026 •

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf: a same-method comparison of the AOP's probe path. It changes my reading above: the AOP's firmware most likely never started, rather than a silent setup port. Correction: "the setup port is exclave-assigned on the desktop" is a hypothesis, not something I've shown.

lobata default cefec0867 = 72b4b56d91ee + 432 own commits, 4920 behind the tip 458f759da9c5 cand-5 d2f8af83bf98 = 09c7ad0eb4d6 + 82 (3 behind 458f759da9c5), with this PR
AOP compatible apple,t6050-aop, apple,t8152-aop same
apple_aop match apple,t8152-aop (its aop.rs has no T6050 entry) apple,t6050-aop → HW_CFG_T6050
mbox-names mbox mbox, setup
suppliers (supplier:*) mailbox, DART: active mailbox, DART, setup mailbox: available
deferred list empty only the sound card
kernel log RTKit: Initializing (protocol version 12); bound RTKit: Initializing (protocol version 12) at 19.6646 s; setup port: … boot incomplete (AP power on: false, endpoints: 0/0): ETIMEDOUT at 19.6668 s; -110
Aura's handoff AOP CPU stopped, 4 nodes AOP CPU stopped, 5 nodes (stream 1 accepted)

The timeout comes about 2 ms after RTKit starts, so the driver isn't waiting on a mailbox that never answers. HW_CFG_T6050 sets setup_wait_system_endpoints: false. With Aura leaving the AOP CPU stopped, the T6050 setup-port path may assume a firmware that is already running.

So on J775c the working path is the T8152-style boot, and this PR makes the tip's driver take the T6050 setup-port path instead. I won't propose this PR for merging as it stands. The description is right for the hardware, but the probe that works is the one without it.

A choice for whoever owns HW_CFG_T6050, selected by hardware state, not board:

  • (a) take the setup-port boot only when the DT gives the AOP a setup mailbox. Then arm64: dts: apple: support M5 Max Mac Studio (J775c) #2102's J775c file without this commit keeps working through the T8152 path, as on the default. Or:
  • (b) choose the path by the AOP CPU's state at probe: stopped → boot it as T8152 does; running → setup port.

Neither is tested. (a) is a one-line gate; (b) needs to know what the laptops' AOP state is at probe.

🤖 Generated with Claude Code

@satchlj

satchlj commented Nov 20, 2026

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf: J775c row: the AOP comes up without this PR's setup mailbox, once the driver gates the setup-port boot on mbox-names (#2556).

Kernel AOP DT Result
cand-5, d2f8af83bf98 = integration/aurora 09c7ad0eb4d6 + 82, with this PR mbox, setup setup port: boot incomplete (AP power on: false, endpoints: 0/0): ETIMEDOUT (-110)
cand-7, 324280cadc47 = integration/aurora d91db387c720 + 85, without this PR, with #2556 mbox only RTKit: Initializing (protocol version 12); apple_aop bound; nothing deferred; no AOP error

d91db387c720 is one merge behind the current tip, f44bf7af65ea (#1568, K114c HID only).

On cand-7, macaudio also probes and logs driver can't assure safety on this model, disabling speakers, where cand-4 had deferred on the CPU DAI.

So on J775c this PR's setup mailbox makes the AOP worse than describing none. I'm leaving it open but not proposing it for merge. Unless someone finds a J775c use for the setup port, I'd close it in favour of #2556. AOP services past RTKit start, and AOP sleep and wake, are untested on either kernel.

🤖 Generated with Claude Code

@RyanTheTide

Copy link
Copy Markdown
Contributor

Superseded. The 11-20 J775c row above shows the AOP comes up without this setup mailbox once the driver takes the setup-port boot only when mbox-names has "setup" (#2556). With #2556 there is no reason to give the J775c AOP a setup mailbox, and the comparison in this thread suggests the setup-port path was the wrong lead for the desktop. Suggest closing this PR in favour of #2556.

Assisted-by: LLM (Claude:Opus 5.5)

@satchlj
satchlj force-pushed the satchlj/j775c-aop-setup-mailbox-20261120 branch from 73097b0 to 1bc4cca Compare November 21, 2026 21:12
@satchlj

satchlj commented Nov 21, 2026

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf: trailer-only rewrite, removed Co-Authored-By from 1 commit(s); same trees, Signed-off-by kept (73097b0 -> 1bc4cca, --force-with-lease).

@DjDeveloperr DjDeveloperr left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Adversarial pre-PR checklist review (§1–§6) on 1bc4ccab509a76b6fede555ca6cbe5176341332b.

Blockers (§2 / §5 — breaks J775c AOP probe, superseded by #2556):

  1. Regresses AOP bring-up on J775c (arch/arm64/boot/dts/apple/t6050-j775c.dts:3122-3123): Your own J775c boot (cand-5, d2f8af83bf98) on this change showed apple_aop failing probe with -110 (ETIMEDOUT) on the setup port (setup port: boot incomplete (AP power on: false, endpoints: 0/0): ETIMEDOUT), leaving AOP services absent and macaudio deferred. Wiring mboxes = <&aop_mbox>, <&aop_setup_mbox>; and mbox-names = "mbox", "setup"; into aop@290c00000 forces the setup-port boot path that fails on J775c.
  2. Superseded by #2556 while still open as non-draft: Your cand-7 (324280cadc47) boot proved that keeping mbox-names = "mbox" without this PR, paired with #2556 (soc: apple: aop: take the setup-port boot only with a "setup" mailbox), brings apple_aop up cleanly on J775c (RTKit: Initializing (protocol version 12), bound, nothing deferred). Once #2556 lands, merging this PR as-is would actively re-break J775c.
  3. CI / PR state: No dt-checks CI has run on this PR (base is draft #2102's feat/j775c-rebased-20261116), and the PR body still describes the change as untested rather than superseded/failing.

Checks passing:

  • §4 Clean-room & attribution: Commit 1bc4ccab509a carries human author/committer and human Signed-off-by only (0 AI trailers); PR body includes Assisted-by:; values are ADT interfacing data.
  • §1 / §3 / §6: No model gating, hardcoded safety values, or naming issues.

Requested action: Close this PR in favor of #2556. If you still want to fix or drop the unreferenced mailbox@4d067c000 node in #2102, remove the aop mboxes/mbox-names = "mbox", "setup" wiring, convert/rebase as needed, re-verify on J775c, and ensure dt-checks runs cleanly.

Assisted-by: LLM (Gemini 4 Neon)

@satchlj

satchlj commented Nov 22, 2026

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf.

Closing as superseded by #2556, per @DjDeveloperr's review.

  • cand-5 (d2f8af83bf98) carried this change. On J775c, apple_aop failed probe on the setup port with -ETIMEDOUT ("boot incomplete (AP power on: false, endpoints: 0/0)"). AOP services were absent and macaudio stayed deferred.
  • cand-7 (324280cadc47) had soc: apple: aop: take the setup-port boot only with a "setup" mailbox #2556 and not this change, keeping mbox-names = "mbox". apple_aop came up cleanly: RTKit protocol 12, bound, nothing deferred.

J775c's AOP needs #2556 and no setup mailbox in the DT, so merging this after #2556 would break J775c again.

Assisted-by: LLM (Anthropic:Claude Opus 5.5)

@satchlj satchlj closed this Nov 22, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants