Skip to content

arm64: dts: apple: t6050-k114c: describe the internal panel handoff - #2183

Draft
joshuaswarren wants to merge 2 commits into
aurora-silicon:integration/aurorafrom
joshuaswarren:w7j/k114c-panel-dts
Draft

joshuaswarren wants to merge 2 commits into
aurora-silicon:integration/aurorafrom
joshuaswarren:w7j/k114c-panel-dts

Conversation

@joshuaswarren

@joshuaswarren joshuaswarren commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

[Joshua's agent] Preserved commit from our October/November K114c work, opened as draft per a patch-id audit: this commit is neither in integration/aurora nor on any PR head. It is the third revision of the series (the first two are on the fork as w7k/k114c-display and -r2 and are superseded by this one).

The commit adds the internal panel half of the J714s handoff description to the K114c: the display subsystem, the dcp and disp0 aliases the boot loader looks for, the DART DMA windows, and the panel node. Everything stays disabled until the loader enables it, as on J714s; without a loader handoff the boot framebuffer stays on simpledrm. The panel is the tandem OLED (no I2C backlight controller), so it takes the generic panel node and its brightness is a DCP request; firmware 27 needs three values the ADT does not give the driver, so the node carries them. Built with W=1 and checked against the dcp binding schema; not booted.

Based on integration/aurora as of 2026-11-11; needs a rebase before review. Draft because it is not booted on the current base.

Unvoiced; checked by hand.

Rebased onto integration/aurora tip (59b02d5, 2026-11-17) as e8ddbc1: clean auto-merge, no conflicts. Built at that head: apple/t6050-k114c.dtb compiles (LLVM), and the schema check (CHECK_DTBS=y) reports no new violations from this commit; the one finding it prints (pcie@1c70000000 port0 vs apple,pcie.yaml) is identical on unmodified integration/aurora and predates this series. Never booted: no boot, no handoff run, no panel probe; the M5 Max is down tonight, so boot proof waits.

Add the internal panel half of the J714s handoff description to the K114c:
the display subsystem, the dcp and disp0 aliases the boot loader looks for,
the DART DMA windows and the panel node. Everything stays disabled until the
loader enables it, as on J714s; without a loader handoff the boot
framebuffer stays on simpledrm.

The DART windows are those of the K114c ADT (dart-dcp and dart-disp0:
vm-base 1 TiB, vm-size 3 TiB). The panel is the tandem OLED of the K114
(ADT oled-display) and has no I2C backlight controller (ADT backlight node:
dcp-brightness-node), so it takes the generic panel node, not the mini-LED
one, and its brightness is a DCP request.

Firmware 27 needs three values the ADT does not give the driver, so the
node carries them: an empty apple,clock-ids (the ADT dcp lists no clocks),
the 302 x 196 mm size of the 14.2-inch panel (ADT disp0 dot pitch 254.6 ppi
at 3024 x 1964), and apple,max-brightness of 1000 (ADT backlight node
user-accessible-max-nits, 0x03e80000 in 16.16). Without the size the driver
sees no panel and registers no backlight; without the maximum it drops every
present.

Built with W=1 and checked against the dcp binding schema; not booted.

Signed-off-by: Joshua Warren <816217+joshuaswarren@users.noreply.github.com>
@RyanTheTide

Copy link
Copy Markdown
Contributor

Thanks for rebasing. Keeping every node disabled until the loader enables it is the right shape, the values are quoted from the ADT, and the checks are green.

What's missing is a boot at e8ddbc1 on the K114c, with two things to show:

  • With a loader that does the handoff: dcp and disp0 probe, the panel comes up at its native mode, width-mm/height-mm reach the connector, and brightness follows a 0 to max sweep.
  • With a loader that does not: the nodes stay disabled and simpledrm still drives the screen, so nothing regresses.

One small check while you're there: the dtsi comment says 3024x1964, but your other K114c logs show eDP at 3024x1986. Please make the comment match the panel.

We have no T6050 machine, so this row has to come from you or another M5 owner. #2009 depends on this panel node.

DCP reports 3024x1986. Keep the inherited J714s panel dimensions
and describe their source separately from the reported mode.

The original and updated preprocessed DTS have identical hashes.

Signed-off-by: Joshua Warren <816217+joshuaswarren@users.noreply.github.com>
@joshuaswarren

Copy link
Copy Markdown
Contributor Author

Note

🤖 OpenAI Codex responding on behalf of joshuaswarren

7a173e51860c corrects the comment to match the DCP's reported 3024x1986 mode. The panel dimensions still come from the J714s. The preprocessed DTS is byte-identical before and after the edit.

The draft stays open for both loader tests. We have not booted this head with or without the DCP handoff.

@RyanTheTide

Copy link
Copy Markdown
Contributor

Thanks, 7a173e5 fixes the comment, and a byte-identical DTB is what we'd expect. Since width-mm/height-mm are borrowed from the J714s, please check them against the panel during the boot. The two loader rows (with and without the handoff) are still what's missing.

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

Note

🤖 GPT-6.2 Sol responding on behalf of Joshua Warren

The exact 7a173e51860c build completed Image, modules, and DTBs. Both CI checks passed. Neither loader arm has booted, and daily2 does not cover them. With handoff enabled, we will check DCP and disp0 probe, native mode, and brightness from zero to max. We will measure the lit area and compare its width and height with the connector values. With handoff disabled, we will check that DCP and disp0 stay disabled and simpledrm drives the panel. Both runs still need a K114c window.

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