Skip to content

kboot: give the J504 the T8122 display and GPU defaults - #25

Merged
maralcbr merged 1 commit into
omacom:integration/aurorafrom
scottjones:j504/bringup
Nov 7, 2026
Merged

maralcbr merged 1 commit into
omacom:integration/aurorafrom
scottjones:j504/bringup

Conversation

@scottjones

Copy link
Copy Markdown
Collaborator

Summary

Give the 14" M3 MacBook Pro (J504) the same T8122 display and GPU defaults as the M3 MacBook Airs. Its ADT names the Airs' DCP (90F849E1-…), GPU (DF697F05-…) and PMP (3B18C886-…) images, and its iBoot is the Airs' 14.x one.

 apple,j613 / apple,j615 (+ apple,t8122)
+apple,j504 (+ apple,t8122)
   defaults: t8122-dcp, -gpu, -gpu-power-standin, -gpu-fuse-leakage
   PMP values and facts, DCP handoff board list, leakage fuse export, powered identity probe

The J504 also takes the Airs' stand-in GPU power model and leakage-fuse decode, both worked out on the J613 and not measured on a J504. On Jordan's Mac the fuse decoded to field A 0x321, field B 0x038, leakage 1604. m1n1 logs the model as "not measured; does not limit clocks": it sets power targets, not clocks. Every check behind the switches still runs: the firmware UUIDs, the powered identity (8 or 10 cores), reservations and resources. Owner opt-outs and configuration-loss handling are unchanged. The M3 iMacs (J433/J434) and the 26.x (25G83) paths are untouched.

Paired kernel change: omacom/linux-aurora#388. Ship order: this must not ship before that kernel. With an older kernel a J504 gets the handoff but has no board record, and stays on the boot framebuffer. The m1n1-aurora package bump will carry a floor on the linux-aurora release that has the kernel change.

Evidence

Hardware: Jordan Goodman's J504 (jordan-m3) and our M3 Air (m3air).

  • Before (stock aurora22, j1105-174533-6072): "DCP: T8122: not armed (no chosen.asahi,t8122-dcp=1): no display handoff". No GPU setup.
  • After (j1105-183940-5ec1, no config lines):
    • stage2 v1.6.1-466-gba5e99ca armed t8122-dcp, gpu, standin and fuse-leakage, and passed the PMP values.
    • Powered identity PASS: 0x07022000, counts 0x0011010a, core-mask 0x3ff.
    • With the paired kernel: native display at 3024x1964@120 and the GPU firmware running.
  • M3 Air: booted this stage 2 on the Air (j1105-182656-8f3c) and restored (j1105-183020-93d1).
  • Host tests:
    • The defaults/opt-out suites pass 169 tests (1 expected xfail).
    • The display-handoff, PMP, identity and fuse tests pass, with J504 cases and J433 refusal controls.
    • Dropping J504 from any of the six gates fails a test, and so does adding J433.
    • run-kboot-gpu-g15.sh has 12 t8122-25-* and t8122-j613-standin-* failures, identical on the 121fcbf base. It stops at the first of them, so run the new J504 case directly: G15_SCENARIOS=t8122-j504-fuse-export sh tests/run-kboot-gpu-g15.sh.
  • Hardware-tested build: ba5e99ca. This head differs only by a comment.

Merge Danger

Door: two-way

Blast Radius: J504 only

Nothing changes for any other board. On a J504 the defaults can be turned off with chosen.asahi,t8122-dcp=0 or chosen.asahi,t8122-gpu=0. The GPU runs at the Airs' conservative cap (state 2, 618 MHz).

🤖 Generated with Claude Code

The 14" M3 MacBook Pro (J504) is a T8122 Mac running the same 14.x
images as the M3 MacBook Airs: its ADT names DCP image 90F849E1-...,
GPU image DF697F05-... and PMP image 3B18C886-..., and its OS firmware
iBoot is the Airs' iBoot-10151.140.19.700.2. Today it gets no display
or GPU handoff and runs on the boot framebuffer.

Admit exactly apple,j504 on apple,t8122 wherever the Air recipe is
gated: the default switches, the PMP values and facts, the DCP
handoff's board list, the leakage fuse export and the powered identity
probe. Every check behind the switches still runs (firmware UUIDs,
powered identity with 8 or 10 cores, reservations, resources), and
owner opt-outs and configuration-loss handling are unchanged. The
J433/J434 iMacs and the 25G83 (26.x) paths are untouched.

Host tests cover J504 in the default/opt-out and configuration-loss
suites, the display handoff scenarios, the PMP admission, the fuse
export and the powered identity, with J433 refusal controls. Dropping
J504 from any gate, or adding J433, fails a test.

Assisted-by: Claude:claude-opus-5-5
@greptile-apps

greptile-apps Bot commented Nov 6, 2026 •

Copy link
Copy Markdown

RetriggerView in GreptileConfidence Score: 5/5

[Medium impact] The PR appears safe to merge with the stated paired-kernel ship order.

Summary

Adds J504 to the existing T8122 display and GPU defaults.

  • The J504 gets the Airs’ T8122 display handoff and boot defaults.
  • The J504 can use the T8122 GPU fuse and identity checks.

Acknowledged limits from scottjones:

  • J504 uses the Airs’ unmeasured stand-in power model and fuse decode intentionally.
  • Shipping before the paired kernel leaves J504 on the boot framebuffer; the planned package dependency will enforce ship order.
  • The reported 12 GPU host-test failures also occur on the base and are outside this change.

Reviews (1) · Last reviewed commit: "kboot: give the J504 the T8122 display a..." · Reviewed by Greptile

@scottjones

Copy link
Copy Markdown
Collaborator Author

Hardware results from our Mac lab, m1n1 built from 9b662fc (stage2 9b662fc in the boot logs), with the kernel from linux-aurora#388:

  • J504 (14" M3 MacBook Pro), j1107-134052-6e88: eDP-1 3024x1964, G15G firmware ready at defaults, backlight set and read back, Vulkan 782/0 bad, 0 DART faults, 0 regressed vs stock.
  • J613 (13" M3 Air), j1107-132933-b912: eDP-1 2560x1664, G15G firmware ready with no flags, GUI smoke passed, 0 DART faults, 0 regressed.
  • Each Mac's boot.bin was checked as stock before, swapped to the candidate, and restored to stock after, with the restore verified.

Also in the combined trial (m1n1 #23 + #24 + #25, linux-aurora AsahiLinux#363 + AsahiLinux#372 + AsahiLinux#388 + AsahiLinux#400, stage2 v1.6.1-473-gd2b8d44b): M3 Air and J504 Vulkan 780/0 bad each, 0 DART faults across the four Macs, no regressions vs stock apart from audio.microphone-signal on the J514M, which fails on stock too.

@scottjones
scottjones removed the request for review from chriskearney November 7, 2026 17:31

@maralcbr maralcbr left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

Reviewed: J504 added to the T8122 display/GPU/PMP/fuse/identity gates; board checks exact, J433 still refused, no change for other boards. Linux CI green; the macOS pytest SIGABRTs also fail on the integration/aurora base. Merging after linux-aurora#388 per the ship order.

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.

2 participants