Repository navigation
arm64: dts: apple: t6050: Share the camera, add J714c, J716s, K114s, K116c - #2275
RyanTheTide wants to merge 1 commit into
Conversation
…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>
|
No blockers. On current integration/aurora, J714s, J716c, K116s and the other non-target boards are byte-identical after the rename. J714c, J716s, K114s and K116c gain the same ISP, ISP DART and ISP power-state nodes, all still disabled until a loader row hands the camera over. dtc warning counts are unchanged and the ADT coverage check passes. dt-checks failed on the runner's apt mirror; rerun it. Can leave draft.
HW row: per board, once its loader row exists: ISP handoff and probe in dmesg, a 900-frame 1080p capture with no errors, then one s2idle and the same capture again; 0 DART faults, 0 AVC, 0 failed units, no new kernel errors. Assisted-by: LLM (Claude:Opus 5.5) |
|
Adversarial pass at 63f9b86: no blockers, one wording point with a power consequence.
Still owed, per board once its loader row exists (J714c, J716s, K114s, K116c): ISP handoff and probe in dmesg, a 900-frame 1080p capture with no errors, and the idle power with and without the camera handoff, to show what the always-on ISP_SYS costs. Assisted-by: LLM (Claude:Opus 5.5) |
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
t6050-j716c-isp.dtsitot6050-isp.dtsi(git mv, content unchanged apart from the header comment) and point J714s, J716c and K116s at the new name.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 asapple,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/isp0sensor-typematches the row.K114c is left to #1756. It can switch to
t6050-isp.dtsiand keep only its own deltas. #2233 (J714s sizes) touches only J714s's sensor-presets block and merges cleanly with this.Testing
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.apple,t6050-ispandapple,t6050-isp-dartalready have bindings.The driver has no board gate:
isp-drv.cbindsapple,t6050-ispand takes the platform ID from the DT. The ISP DART's sleep restore is enabled on J714s only (drivers/iommu/apple-dart.cisp_sleep). That does not block probe or capture.Still to run
Each needs the loader row for the board first. Then:
dmesg | grep -i ispshows the T6050 ISP handoff and the driver probe, andv4l2-ctl -d /dev/videoN --stream-mmap --stream-count=900 --stream-to=/dev/nullcaptures 30 s at 1920x1080 with no errors and a steady frame rate. Frames stay on the machine.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:
What changes on each of the four boards (phandle numbers normalised; identical on all four):
apple,t6050-isp-dart, DAPF window on the first, remap-streams; still disabled; power-domains ISP_SYSAssisted-by: LLM (Claude:Opus 5.5)