Skip to content

arm64: dts: apple: t6050: share the K boards' N1 and USB-C ports, add K114s and K116c - #2280

Draft
RyanTheTide wants to merge 3 commits into
integration/aurorafrom
dtfill-t6050/k-io
Draft

RyanTheTide wants to merge 3 commits into
integration/aurorafrom
dtfill-t6050/k-io

Conversation

@RyanTheTide

@RyanTheTide RyanTheTide commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

Problem

The K114s (14-inch M5 Pro) and K116c (16-inch M5 Max) have no USB, no Wi-Fi/Bluetooth and no PCIe. t6050-k114-k116.dtsi disables ATC1, nub-spmi-a0 and the PCIe host, and the two board files add nothing back. The K114c and K116s describe the same N1 port and USB-C ports, each in its own copy.

Change

The four boards' ADTs have the same apcie0 port 0 (registers, IRQ 1523, PERST# GPIO 80, the N1 on dart-apcie0 streams 1-3), the same PD controllers (hpm1 at 0xa on nub-spmi-a0, hpm2 at 0xc on nub-spmi-a1, SN201202x) and the same atc-phy1/2, usb-drd1/2 and dart-usb1/2 nodes. Only phandles differ.

  • t6050-k114-k116.dtsi: PCIe port 0 with the N1 (control, Wi-Fi, Bluetooth), the wifi0/bluetooth0 aliases, and PERST# pin 80 exposed. The board file defines T6050_K_N1_FIRMWARE and T6050_K_BOARD_ID before the include. The platform ID is (board-id << 16) | 0x6050.
  • New t6050-k114-k116-usb.dtsi: ATC1 (left-front) and ATC2 (right). It has the PD controllers with their connector graphs, the combo PHYs and mailboxes, the DWC3s, their DARTs, the ATC2 power states and the atcphy1/atcphy2 aliases. A board includes it after t6050-peripherals.dtsi, which defines nub-spmi-a1 for these boards.
  • K114c keeps only its deltas: the nub-spmi-a0 fault-queue window, the MagSafe PD controller (hpm5) and the ATC2 router.
  • K116s keeps only its deltas: the SD reader on port 1 and the single nub-spmi-a1 register window.
  • K114s gets board-id 0x14 and K116c gets board-id 0x06. Each includes both files. Like K116s, these two boards do not describe hpm5 yet.

The shared blocks are Joshua Warren's K114c and Jens Axboe's K116s descriptions.

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.
  • DTB build (kbuild cpp + dtc flags) for all 11 T6050 boards: no warnings.
  • J714c, J714s, J716c, J716s, J775c, J775d, J873s, K114c and K116s are byte-identical before and after.
  • K114s and K116c change. Their PCIe host and port 0 match the K114c's except for the firmware name and platform ID. Their USB nodes match the K116s's.
  • dt-validate: no new messages on K114s, K116c, K114c or K116s.
  • Not booted.

Still to run

Per-board values and DTB diffs
Board board-id N1 platform ID firmware-name
K114c 0x04 0x00046050 apple/centauri/k114c-ftab.bin
K114s 0x14 0x00146050 apple/centauri/k114s-ftab.bin
K116c 0x06 0x00066050 apple/centauri/k116c-ftab.bin
K116s 0x16 0x00166050 apple/centauri/k116s-ftab.bin

K114s and K116c DTB changes (both boards the same apart from the values above):

  • aliases: wifi0, bluetooth0, atcphy1, atcphy2
  • pcie@1cb0000000: status okay; reg, reg-names, interrupts and interrupt-names for port 0; pci@0,0 with control@0,0, wifi@0,1, bluetooth@0,2
  • pinctrl gpio-reserved-ranges <0 80>, <81 140> (was <0 221>)
  • nub-spmi-a0: okay, usb-pd@a (USB-C Left-front)
  • nub-spmi-a1: okay, interrupts removed, usb-pd@c (USB-C Right)
  • ATC1 and ATC2: PHY mailbox, PHY (with graph port), DWC3 (with graph port), both DARTs, ATC2 power states okay

Assisted-by: LLM (Claude:Opus 5.5)

… K114s and K116c

The K114s and K116c had no USB, no Wi-Fi/Bluetooth and no PCIe: the
shared t6050-k114-k116.dtsi disabled ATC1, nub-spmi-a0 and the PCIe host,
and only the K114c and K116s described their N1 port and USB-C ports,
each in its own copy.

The boards' ADTs have the same apcie0 port 0 (registers, IRQ 1523, PERST#
GPIO 80, N1 on dart-apcie0 streams 1-3), the same PD controllers (hpm1 at
0xa on nub-spmi-a0, hpm2 at 0xc on nub-spmi-a1, SN201202x) and the same
atc-phy, usb-drd and dart-usb nodes on all four boards. So:

- t6050-k114-k116.dtsi now describes PCIe port 0 with the N1, its
  wifi0/bluetooth0 aliases and PERST# pin 80. The board file defines the
  N1's firmware name and its board-id before including it; the platform
  ID is board-id << 16 | 0x6050.
- The new t6050-k114-k116-usb.dtsi describes ATC1 (left-front) and ATC2
  (right): PD controllers, combo PHYs, DWC3s, DARTs and the atcphy
  aliases. Boards include it after t6050-peripherals.dtsi, which defines
  nub-spmi-a1 for them.
- K114c and K116s keep only their deltas: K114c the nub-spmi-a0 fault
  queue window, the MagSafe PD controller and the ATC2 router; K116s the
  SD reader on port 1 and the single nub-spmi-a1 window.
- K114s (board-id 0x14) and K116c (board-id 0x06) include both.

The K114c and K116s DTBs are byte-identical before and after, as are
the other T6050 DTBs. The shared blocks are Joshua Warren's K114c and
Jens Axboe's K116s descriptions.

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

Copy link
Copy Markdown
Contributor Author

Blockers: 1.

The move itself checks out at the head: K114c and K116s are byte-identical, K114s and K116c get the K114c PCIe port 0 (only firmware-name and platform ID differ) and the K116s ATC1/ATC2 nodes, and the only newly exposed pin is 80 (112 stays reserved).

  1. The PR no longer merges. integration/aurora gained 18cfec4 (arm64: dts: apple: t6050-k114c: boot NOR and read-only nvram partition #2563, K114c boot NOR and its read-only nvram partition), which adds nvram = &nvram; to the same K114c aliases block this PR deletes from t6050-k114c.dts. Rebase, keep the nvram alias as a K114c delta next to the shared wifi0/bluetooth0, and re-run the byte-identical check for K114c and K116s on the rebased head. dt-checks was cancelled on the current head, so it needs a green run after the rebase too.
  2. Heads-up, not a blocker: arm64: dts: apple: t6050: describe the headphone jack codec on every board #2277 rewrites the same &pinctrl_ap list in t6050-k114-k116.dtsi (this PR line 126 <0 80>, <81 140>, arm64: dts: apple: t6050: describe the headphone jack codec on every board #2277 <0 134>, <135 4>, <140 81>). Whichever lands second must expose 80, 134 and 139: <0 80>, <81 53>, <135 4>, <140 81>. The two branches conflict textually there, so it will not merge silently, but the obvious resolution of taking either side drops a pin.

HW row: K114s on the rebased head: USB 3 device enumerates on left-front and right in both orientations, iw dev wlan0 link associated, bluetoothctl list shows one controller, 2x s2idle with Wi-Fi back after each; K116c the same once an owner is found; K114c and K116s boot regression (Wi-Fi, both USB-C ports); 0 AVC, 0 failed units, no new kernel errors.

Assisted-by: LLM (Claude:Opus 5.5)

t6050-k114c.dts: kept this branch's move of the N1 port and its aliases into t6050-k114-k116.dtsi, kept integration's nvram alias in a K114c aliases block.

Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
…e PERST# pin

The K board headphone jack change opens pins 134 and 139 (the CS42L84's
interrupt and reset) in the shared K114/K116 pinctrl_ap list, and this branch
opens pin 80 for PCIe port 0. Use the union, <0 80>, <81 53>, <135 4>,
<140 81>, in both so they land in either order. Both pins read cleanly; only
pin 112 takes an SError.

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

Copy link
Copy Markdown
Contributor Author

Merged integration/aurora (f5f023b); conflicts in t6050-k114c.dts resolved (K114c keeps integration's nvram alias; shared K114/K116 pinctrl_ap list set to the union with #2277: <0 80>, <81 53>, <135 4>, <140 81>), all eleven T6050 board DTBs build with unchanged dtc warnings; K114c, K116s and the other boards decompile identical to integration.

@RyanTheTide

Copy link
Copy Markdown
Contributor Author

The rebase is in, but dt-checks now fails: the DART power-domain check flags dart_apcie0 (iommu@410000000) on K114s and K116c as having no power-domains. Still needed after that: a K114s row on this head (USB 3 on both ports in both orientations, Wi-Fi and Bluetooth up, Wi-Fi back after 2x s2idle) and K114c/K116s boot regression rows.

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