[garymeyerca]

On the 13-inch M3 MacBook Air (J613), DPMS off and system suspend leave the display backlight lit: the 14.7 IOMFB power-off path only presents a black frame. This patch turns off panel power and restores it when the CRTC is enabled again.

Enable this behavior through the J613 board record. Complete the final black frame before A472(0); wake with A475(1) followed by A472(1) and refresh the colour matrix. Handle the power notifications in the internal callback dispatcher, retain acknowledged state on refusal, and reject swaps while the panel is off.

The firmware sequence and callback layouts follow GehrmannSolutions' J613 clamshell work. Related: 14.x panel brightness control; that broader issue remains open. The PR contains the driver change and one host regression script.

Validation:

• Native appledrm.ko build from this integration/aurora branch against matching installed kernel headers, with LLVM=1 LD=ld.bfd HOSTLD=ld.bfd W=1. Reference and candidate builds report the same warnings: Clang 23.1.3 versus the kernel's 22.1.8, and existing init-to-exit references to dcp_unregister and dcp_audio_unregister. This is a module build, not a full kernel build.
• python3 tools/testing/apple/check-v14-panel-power.py passes with ASan/UBSan and compiler warnings denied. It executes the production power functions and the internal callback-dispatch prefix: repeated transitions, idempotence, wire layouts, firmware refusal, transport failure, callback shapes and board gating. Negative controls for black-only blanking and callbacks placed only in the external handler fail.
• Hardware-tested as a backport to source c46d7dcdb8712ea9aeb0ce47b439ffae76371e55, whose affected driver files match this PR's base. Reboot verification matched the running module build ID to the tested binary. DPMS off/on and two s2idle cycles passed. The operator confirmed a fully dark backlight during sleep and normal display recovery. No DCP power, callback, link or flip errors appeared in the successful tests.
• git diff --check and strict checkpatch with --no-signoff --ignore FILE_PATH_CHANGES pass. The existing tools/testing/apple/ MAINTAINERS entry covers the new test.

Test machine: J613/T8122, Omarchy on Arch Linux ARM; OS firmware 14.7, system firmware 27.0; Aurora release aurora-2026.10.17.20, kernel 7.1.12-2-20261017.20-sep-ARCH (linux-aurora 7.1.12.aurora2-20261017.20); m1n1 1.6.1.aurora17-5, U-Boot 2026.07.asahi2-4, Hyprland 0.56.2-7, Mesa 26.2.5, with asahi.t8122_start=1. The installed test system also carries local camera and platform changes. External displays, other boards, and lock-screen security were not qualified by these tests.

Draft pending human review and DCO sign-off, per the repository's coding-assistant guidance. The commit includes Assisted-by: Codex:GPT-6 and no agent-added sign-off.

[RyanTheTide]

What's needed before this can merge:

1. Commit 9cd46b2 has no Signed-off-by. Please add yours (git commit --amend -s) and push.
2. Rebase onto the current integration/aurora. It conflicts in drivers/gpu/drm/apple/iomfb_v14_7.c with 2984ba9 ("drm/apple: enable the J613 panel brightness path") and the later 14.7 changes:
   • dcp_v14_board_j613 now sets .panel_nits = true; keep it next to .panel_power.
   • dcp_v14_swap() now takes v14->lock differently; put the panel_off check inside the current locked section.
   • iomfb_v14_7_poweron() now calls dcp_backlight_dpms(dcp, true) when dcp_v14_bl_dpms_applies() is true. The panel must be powered (A475(1), A472(1)) before that restore runs.
   • iomfb_v14_7_poweroff() now blanks through dcp_v14_blank(), not dcp_v14_present(..., NULL, ...). The A472(0) step has to follow that path and only run once the black frame has completed.
3. Testing: the hardware result in the description comes from a 7.1.12 backport. After the rebase, please post a J613 (14.7 firmware) result from the PR head: DPMS off/on, two s2idle cycles with the backlight dark, brightness restored after wake, and check-v14-panel-power.py passing. The non-J613 14.7 boards keep panel_power unset; one of them should be checked to confirm DPMS still only blanks there.