Skip to content

arm64: dts: apple: t6050: Share the camera, add J714c, J716s, K114s, K116c - #2275

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

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

Conversation

@RyanTheTide

@RyanTheTide RyanTheTide commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

Problem

The T6050 camera is described on J714s, J716c and K116s only (K114c is in #1756). J714c, J716s, K114s and K116c have no ISP nodes at all, so the camera cannot come up on them even with a loader that hands it over. The shared description lives in t6050-j716c-isp.dtsi, although nothing in it is J716c specific.

Change

  • Rename t6050-j716c-isp.dtsi to t6050-isp.dtsi (git mv, content unchanged apart from the header comment) and point J714s, J716c and K116s at the new name.
  • Include it from J714c, J716s, K114s and K116c.

Each board's ADT has the same isp0, dart-isp0 and ISP power states: registers, interrupts, DART streams and power-state offsets all match across the eight T6050 laptops. The only per-board value is the isp0 sensor-type. The loader reads it from the ADT and passes it to the driver as apple,platform-id, so the DT has no per-board camera property. Each new include line notes the board's sensor-type in a comment.

Everything stays disabled until the loader has reserved the camera firmware. The loader's T6050 camera handoff is a board x firmware table and fails closed: a board without a row keeps the camera off. So these four boards also need a loader row: board compatible, the sensor-type below and the firmware release. The loader already checks that the ADT /arm-io/isp0 sensor-type matches the row.

K114c is left to #1756. It can switch to t6050-isp.dtsi and keep only its own deltas. #2233 (J714s sizes) touches only J714s's sensor-presets block and merges cleanly with this.

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. No new dtc warnings (0 before, 0 after on every board).
  • J714s, J716c, K116s, K114c, J775c, J775d and J873s are byte-identical.
  • On J714c, J716s, K114s and K116c the ISP, DART and power-state nodes and aliases come out identical to J716c's and K116s's (compared with phandles resolved to paths). The before/after delta, normalised for phandle numbers, is the same on all four.
  • dt-validate: no new messages on J714c, J716s, K114s or K116c. Before and after: J714c has the same one existing pcie@1c70000000 message, the other three have none. apple,t6050-isp and apple,t6050-isp-dart already have bindings.

The driver has no board gate: isp-drv.c binds apple,t6050-isp and takes the platform ID from the DT. The ISP DART's sleep restore is enabled on J714s only (drivers/iommu/apple-dart.c isp_sleep). That does not block probe or capture.

Still to run

Each needs the loader row for the board first. Then: dmesg | grep -i isp shows the T6050 ISP handoff and the driver probe, and v4l2-ctl -d /dev/videoN --stream-mmap --stream-count=900 --stream-to=/dev/null captures 30 s at 1920x1080 with no errors and a steady frame rate. Frames stay on the machine.

  • J714c (MBP 14 M5 Max): sensor-type 0x73. Owner row needed.
  • J716s (MBP 16 M5 Pro): sensor-type 0x76. Owner row needed.
  • K114s (OLED MBP 14 M5 Pro): sensor-type 0xba. Owner row needed.
  • K116c (OLED MBP 16 M5 Max): sensor-type 0xbb. Owner row needed.

The sensor presets are the shared default (as on J716c and K116s). If a board's firmware does not list that sensor config, the probe log says so, and the board gets its own presets block as J714s has.

Per-board values and DTB diffs

ADT isp0 sensor-type per board, the same in 27.0/27.2 and 27.3 beta ADTs: J714c 0x73, J714s 0x74, J716c 0x75, J716s 0x76, K114c 0xb9, K114s 0xba, K116c 0xbb, K116s 0xbc. All boards: cam-connections-scheme 0x10, interrupts 513-516 and 526, the same three reg windows, clock-ids 350/352, the same dart-isp0 nodes, and the same ISP_SYS/ISP_CPU/ISP_FE/DPRX/ISP_SECURE/ISP_BE/ISP_CLR/ISPSENS0/FAB9_SOC power-state entries.

dtdiff:

CHANGED t6050-j714c
IDENTICAL-DTB t6050-j714s
IDENTICAL-DTB t6050-j716c
CHANGED t6050-j716s
IDENTICAL-DTB t6050-j775c
IDENTICAL-DTB t6050-j775d
IDENTICAL-DTB t6050-j873s
IDENTICAL-DTB t6050-k114c
CHANGED t6050-k114s
CHANGED t6050-k116c
IDENTICAL-DTB t6050-k116s

What changes on each of the four boards (phandle numbers normalised; identical on all four):

  • aliases isp, isp-dart0..2 added
  • isp@580000000: the J716c reg/reg-names/interrupt/iommus (streams 0 and 12 on each DART)/eight power-domains/dart-vm-size and the default sensor preset; clocks dropped; still disabled
  • iommu@5828c0000, @5828e0000, @582900000: apple,t6050-isp-dart, DAPF window on the first, remap-streams; still disabled; power-domains ISP_SYS
  • pmgr1: isp_cpu, isp_fe, isp_dprx, isp_secure, isp_be, isp_clr power states added (parent isp-sys); fab9-soc, isp-sys and ispsens0 enabled as on J716c

Assisted-by: LLM (Claude:Opus 5.5)

…K116c

The T6050 camera (ISP, its three DARTs and its power states) was
described for J714s, J716c and K116s only, in t6050-j716c-isp.dtsi.
Nothing in that file is J716c specific: each laptop's ADT has the same
isp0, dart-isp0 and ISP power states (registers, interrupts, streams,
power-state offsets). The only per-board value is the isp0 sensor-type,
and the boot loader reads that from the ADT and passes it on as the
platform ID.

Rename the file to t6050-isp.dtsi and include it from J714c, J716s,
K114s and K116c as well. Each board's include notes its sensor-type
from that board's ADT. Every node stays disabled until the loader has
reserved the camera firmware, so a loader without a row for the board
leaves the camera off.

The J714s, J716c and K116s DTBs are byte-identical. K114c is left out
here: its camera is described in a separate open change. J775c, J775d
and J873s have no camera and are unchanged.

The original J716c description is by Erik Bourget.

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