Skip to content

arm64: dts: apple: t6050: Add shared USB-C display routes and four boards - #2283

Draft
RyanTheTide wants to merge 1 commit into
integration/aurorafrom
dtfill-t6050/dispext
Draft

RyanTheTide wants to merge 1 commit into
integration/aurorafrom
dtfill-t6050/dispext

Conversation

@RyanTheTide

@RyanTheTide RyanTheTide commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

Problem

USB-C external displays (DisplayPort alt mode through the dcpext engines) are described only on the J714s and the J716c. Every other T6050 board's ADT has the same four external display engines on die 0 (dcpext0..3 with dart-dcpextN and dart-dispextN), the same display crossbar and the same ATC1/ATC2 Type-C PHYs. Their trees leave those engines bare, and their connectors have no DisplayPort endpoint.

Change

  • New t6050-display.dtsi holds the part that is the same on every board:
    • dcpext1..3: firmware index (2/3/4), USB-C output on die 0's DPTX, and no clocks. Their DCP and display DARTs get the DMA range (1 TiB + 3 TiB), and each display DART gets its single register window, as dcpext0's.
    • DisplayPort alt mode routes for dcpext0 and dcpext2 to ATC2 (route@0) and ATC1 (route@1), on crossbar source 0 and 4, with the DP IN controls. The phy and crossbar order is the J714s's.
    • The ATC1 and ATC2 PHYs' DP clock source and DP power states, and both ports' display crossbars.
  • The engines, their mailboxes and DARTs stay disabled. The boot loader hands each engine over when its checks pass, as on the J714s. The Type-C mux lookup passes over a route under a disabled engine, so the PD controllers do not wait for it.
  • Per board, each board only links its connectors' DisplayPort port (port@2) to the four route endpoints:
    • J714c, J716s, K114c and K116s: left-front (ATC1) and right (ATC2).
    • J714c: the right connector's endpoint from t6050-max-atc2.dtsi is replaced.
    • K116s: its own copy of the dcpext1..3 clock and DART range properties is dropped, because the shared file now has them.
  • J714s and J716c keep their own display files. This avoids conflicts with the open PRs that edit t6050-j714s-display.dtsi (arm64: dts: apple: t6050-j714s: the left-back USB-C port (USB, USB4, DisplayPort) #2017, J714s: keep the USB4 dock port's USB host in place through a deep sleep #2072). They can include the shared file once those land.
  • Not in this PR:
    • J775c (via arm64: dts: apple: support M5 Max Mac Studio (J775c) #2102, a full board tree)
    • J775d (the loader path in aura#264 checks dcpext0's route as a whole)
    • J873s (no ATC ports or connectors described yet)
    • K114s and K116c (ATC ports are being added separately)
    • USB4 DP tunnels: the left routers are not described, and acio2 DP IN interrupts are not set on J714c/K114c
    • HDMI

Testing

  • Full kernel tree make dtbs CHECK_DTBS=y (dtschema) for every T6050 DTB on this branch: builds clean, no dtc warnings, and no schema message that integration/aurora does not already have.
  • All 11 T6050 DTBs build with kbuild's cpp and dtc flags. There are no dtc warnings, before or after.
  • J714s, J716c, J775c, J775d, J873s, K114s and K116c are byte-identical. J714c, J716s, K114c and K116s change only in the dcpext, DART, crossbar, ATC PHY, DP power-state and connector nodes listed above.
  • dt-validate on the four changed boards gives no new messages (J714c and K114c keep one existing pciec port0 message). No new compatibles.

Still to run

Each board needs this owner row: a USB-C DisplayPort monitor on the left-front port, then on the right port. On each port, modetest -M apple -c should list a connected DP connector with the monitor's modes, and a mode set (modetest -s) should light the monitor. Unplugging and replugging should report disconnected/connected (hotplug), and the monitor should come back.

The kernel side is ready, but nothing hands these engines over yet. The loader hands external engines over only after an internal panel handoff with an image row that has external engines on:

Per-board values and DTB diffs

Values compared in each board's ADT (J714c, J716s, K114c, K116s, against J714s and J716c):

  • dcpext0..3 and dispext0..3 are present on each board, with crossbar ufp endpoints 0..3 and external index 1..4.
  • dart-dcpextN and dart-dispextN have vm-base 1 TiB and vm-size 3 TiB.
  • atc-phy1 and atc-phy2 are present, with a DP clock source on each.
  • hpm1 is left-front on nub-spmi-a0 0xa, and hpm2 is right on nub-spmi-a1 0xc.
  • Pro boards (J716s, K116s) report the same display count limit as the J714s. Max boards (J714c, K114c) report none, as the J716c.

Changes per board, all of them in the merged tree:

  • dcpext0: phys/mux lists become two ports, and route@1 (ATC1) is added. On J714c, route@0 is enabled.
  • dcpext2: same lists plus route@0 and route@1.
  • dcpext1..3: dcp-index, connector-type and dptx-die. On J714c, J716s and K114c they also get clock-ids.
  • dcpext1..3 and dispext1..3 DARTs: apple,dma-range. The dispext DARTs get a single reg window.
  • atc1_dpxbar gets power-domains and is enabled. atc2_xbar is enabled.
  • ps_atc1_common_dp_0..3 and ps_atc2_common_dp_0..3 are enabled.
  • atc1_phy and atc2_phy: the dp-clock-source reg and dp0..3 power domains are added.
  • The ATC1 and ATC2 connectors each get port@2 with endpoint@0 (dcpext0) and endpoint@1 (dcpext2).

Board drivers that limit, but do not refuse, the new boards in drivers/gpu/drm/apple:

  • dcp.c:471 (dcp_j714s_ext: crash reconnect, deep-sleep refusal, link report)
  • dcp.c:5791 (firmware 27 sleep transport: J716c only)
  • dcp.c:7002 (v27_j714s: A407/A500 method table)
  • dcp-fabric.c:3363 (tunnel pipeline hold: J714s only)
  • dcp-peer-v27.inc:72 (peer limit: J714s only, which the J716s and K116s also need, because they have the same display count limit)

Assisted-by: LLM (Claude:Opus 5.5)

…ards

External displays over USB-C (DisplayPort alt mode) are described only on
the J714s and the J716c. Every other T6050 board's ADT has the same four
external display engines on die 0 (dcpext0..3, with their dart-dcpextN and
dart-dispextN), the same display crossbar and the same ATC1/ATC2 Type-C
PHYs, but their trees leave the engines bare and the connectors without a
DisplayPort endpoint.

Add t6050-display.dtsi with the board-agnostic part:
- dcpext1..3: firmware index, USB-C output on die 0's DPTX, no clocks;
  their DCP and display DARTs' DMA range (vm-base 1 TiB, vm-size 3 TiB) and
  the display DARTs' single register window, as dcpext0's;
- DisplayPort alt mode routes for dcpext0 and dcpext2 to ATC2 (route@0) and
  ATC1 (route@1), on crossbar source 0 and 4, with the DP IN controls, in
  the J714s's phy and crossbar order;
- the ATC1 and ATC2 PHYs' DP clock source and DP power states, and both
  ports' display crossbars.

The engines, their mailboxes and DARTs stay disabled: the boot loader hands
each one over when its checks pass, as on the J714s.

Each board adds only its connectors' DisplayPort endpoints. J714c, J716s,
K114c and K116s include the file and link their left-front (ATC1) and right
(ATC2) connectors. On the J714c the right connector's endpoint from
t6050-max-atc2.dtsi is replaced. The K116s drops its own copy of the
dcpext1..3 clock and DART range properties, now in the shared file. All
values match each board's ADT.

The J714s and J716c keep their own display files and their DTBs are
byte-identical. J714c, J716s, K114c and K116s change; J714s, J716c, J775c,
J775d, J873s, K114s and K116c are byte-identical.

The routes, crossbar indices and PHY values were written by DjDeveloperr,
Isaac Duarte and Erik Bourget for the J714s and J716c.

Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
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.

1 participant