Skip to content

arm64: dts: apple: t6050-k114-k116: report the built-in panel as eDP (draft, unbooted) - #2009

Draft
joshuaswarren wants to merge 1 commit into
aurora-silicon:integration/aurorafrom
joshuaswarren:dhp4/t6050-k114-internal-panel
Draft

joshuaswarren wants to merge 1 commit into
aurora-silicon:integration/aurorafrom
joshuaswarren:dhp4/t6050-k114-internal-panel

Conversation

@joshuaswarren

@joshuaswarren joshuaswarren commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

Note

🤖 OpenAI Codex responding on behalf of joshuaswarren

Without a DCP handoff, the K114c panel appears as card0-Unknown-1. Omarchy counts it as an external display, so it keeps a lid-switch inhibitor and the laptop stays awake when the lid closes.

The four K114 and K116 boards now link the boot framebuffer to their existing panel binding. The phandle resolves to one panel per DCP. The loader still controls when the DCP becomes active. This is the device-tree half; land the simpledrm change in #1688 first. Related: #2183.

At 138590e64155, all four DTBs compiled and their phandles resolved. The four framebuffer schema warnings are gone and the DCP schema has no new warnings.

Keep in draft until one K114c boot with both PRs and no DCP handoff shows card0-eDP-1, no handle-lid-switch inhibitor, and suspend and resume in the journal after a lid cycle. No such boot has run on this head.

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

GREEN on 503b7bb
Model: zai/glm-5.3

One line, one file, one commit on the current integration/aurora - the diff is exactly the claimed apple,internal-panel on the shared K114/K116 framebuffer node, and the merge-order paragraph gives the correct reasons in both directions (without the binding, CHECK_DTBS warns on all four K-series dtbs; without the driver change the property does nothing). The Validated-on section is the honest scope: dtc builds of both dtbs, a dtb-to-dts round-trip proving the property landed in the K114c dtb, no CHECK_DTBS here with the earlier 1688 one attributed to another machine, not booted, and the lid test named as pending on both PRs together. The reach answer is the right no.

Model: zai/glm-5.3 (live from lane_models.py this turn).

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

GREEN on 503b7bb
Model: anthropic/claude-sonnet

Independent review. No blocking items. This is a draft and I did not boot it. The PR text says so plainly, in the title and in the body.

What I checked:

  • One line, one file. The diff adds apple,internal-panel; to the framebuffer node in arch/arm64/boot/dts/apple/t6050-k114-k116.dtsi and nothing else.
  • Merge order. The body says it lands after drm/sysfb: report eDP for framebuffer panels #1688, and why: drm/sysfb: report eDP for framebuffer panels #1688 holds the driver change and the binding, and without the binding CHECK_DTBS warns on these nodes. I checked that without the binding the schema rejects the property on a framebuffer node (my review of drm/sysfb: report eDP for framebuffer panels #1688).
  • The dtbs build. I built t6050-k114c, k114s, k116c and k116s with cpp and dtc at this head. All four build, and dtc -I dtb -O dts shows apple,internal-panel on framebuffer@0 in each. The same boards built without the line have no match. The dtc warnings are the same before and after for the K114c (none from this change).
  • Driver build. simpledrm.o builds with W=1 and no warnings. I checked that on the drm/sysfb: report eDP for framebuffer panels #1688 head, since this PR has no driver change.

Non-blocking notes:

  1. It covers four boards, not two. The shared t6050-k114-k116.dtsi node is included by the K114c, K114s, K116c and K116s dts files, so all four get the property. The text names the K114 and K116 boards, which is right, but only the K114c was ever seen with the problem.
  2. A lane name sits in public text. The "Not run" paragraph ends with a pane name. Please drop it, or say "in a slot with the lab owner".
  3. No schema run here. The body says CHECK_DTBS was not run on this PR and points at the earlier numbers on drm/sysfb: report eDP for framebuffer panels #1688. That is accurate.

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

The four dt-checks warnings on this PR (apple,internal-panel not allowed on the apple,simple-framebuffer node of the K114c, K114s, K116c and K116s) are expected. #1688 adds the property to the simple-framebuffer binding, and this PR is stacked on it. The warnings clear once #1688 is in the base. No change needed here.

@joshuaswarren
joshuaswarren force-pushed the dhp4/t6050-k114-internal-panel branch from 503b7bb to b8c457c Compare November 18, 2026 01:12
@RyanTheTide

Copy link
Copy Markdown
Contributor

Thanks for splitting this out. On #1688 you agreed to switch simpledrm to report eDP when the framebuffer node has the existing panel phandle, and to drop the apple,internal-panel binding. That changes this PR as well: the line at t6050-k114-k116.dtsi:50 should become panel = <&panel>; on the shared framebuffer node, pointing at a panel node with width-mm/height-mm (the K114c one in #2183, or a shared K114/K116 node). That also clears the four dt-checks failures here, because panel is already in the binding.

What's missing after the rework:

We can't supply K114c rows ourselves.

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

Note

🤖 OpenAI Codex responding on behalf of joshuaswarren

138590e replaces apple,internal-panel with panel = <&panel> in the shared framebuffer. The four boards link to one panel per DCP. K114 uses 302 by 196 mm from the #2183 panel description. K116 uses 345 by 223 mm from the existing K116s node, which keeps its compatible. DCP stays disabled until the loader enables it.

The four old DTBs each reported the vendor-property warning. All four new DTBs compile and resolve the panel phandle. The framebuffer schema reports no diagnostics, and the DCP schema reports no new ones.

The simpledrm rework in #1688 must land first. On K114c we still owe the shared no-DCP eDP, inhibitor and lid suspend/resume check. This stays in draft.

@RyanTheTide

Copy link
Copy Markdown
Contributor

Thanks, 138590e uses panel = <&panel> and clears the vendor property. A few points:

  1. Size: 302 x 196 mm is the J714s value for a 3024x1964 panel. The K114c panel is 3024x1986, and HID: apple-mtp-kiwi: the K114c touch screen (needs #1568) #1605 computes 302 x 199 mm from 254 ppi. Please use 199 here. If the K114s panel is the same, it should be 199 as well.
  2. The panel node and its sizes are added to the SoC file (t6050-peripherals.dtsi:2692-2703) under board #ifs. Board data fits better in t6050-k114-k116.dtsi (node and label) with width-mm/height-mm in each board .dts.
  3. arm64: dts: apple: t6050-k114c: describe the internal panel handoff #2183 also adds a K114c &dcp { panel { ... 302 x 196 } }. Please keep a single definition.
  4. Squash b8c457c and 138590e so the vendor property never lands.

As before, the reworked #1688 goes first, and then one K114c row: card0-eDP-1, no handle-lid-switch inhibitor, and lid suspend and resume in the journal.

Link the shared boot framebuffer to one panel under DCP. Keep the
panel node with the board family and set its dimensions per board.
Use 302 by 199 mm for the 14-inch panels based on the K114c mode
and 254 ppi. Keep the existing 16-inch size and K116s compatible.

All four DTBs compile and resolve the framebuffer panel phandle.
Both the framebuffer and DCP stay disabled for the loader.

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

The single-commit candidate puts the shared panel in t6050-k114-k116.dtsi and its size in each board file. It never adds the vendor property. The 302 by 199 mm value comes from the K114c mode and 254 ppi in #1605. It is an estimate on K114s.

All four DTBs compiled and resolved the framebuffer phandle to the DCP panel. The framebuffer and DCP both stay disabled for the loader. K116s keeps its panel compatible.

The #2183 loader test keeps its pinned kernel. After those runs, #2183 can put its brightness property on the shared panel. The K114c no-handoff eDP, lid inhibitor and suspend/resume check is still open.

@RyanTheTide

Copy link
Copy Markdown
Contributor

Thanks, de86dd3 covers all four points: one panel node under &dcp in t6050-k114-k116.dtsi, the sizes in each board file, 302 x 199 mm for K114, nothing added to t6050-peripherals.dtsi, and no vendor property in the history. Leaving the #2183 brightness property for a later change to this shared node is fine.

The PR head is still 138590e, so please push the candidate to the branch so dt-checks run on it. Moving the t6050-peripherals.dtsi include into the family file also changes where it is included in all four boards, so a short note that the four DTBs differ only by the panel and the phandle (for example dtx_diff old vs new) would help the review.

After that, the order is the same: #1688 lands first, then the K114c row (card0-eDP-1, no handle-lid-switch inhibitor, lid suspend and resume in the journal).

@joshuaswarren
joshuaswarren force-pushed the dhp4/t6050-k114-internal-panel branch from 138590e to de86dd3 Compare November 20, 2026 05:42
@joshuaswarren

Copy link
Copy Markdown
Contributor Author

Note

🤖 OpenAI Codex responding on behalf of joshuaswarren

de86dd32b8b9 is now on this branch. I compiled all four DTBs at 138590e64155 and de86dd32b8b9 and compared their sorted trees. Only the K114c and K114s panel height changed, from 196 to 199 mm. The K116c and K116s trees and all four framebuffer panel phandles stayed the same.

The PR stays draft. The K114c lid test still needs #1688 and this change together.

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