Skip to content

arm64: dts: apple: t8122-j613-j615: share the CPU thermal zones - #2684

Merged
RyanTheTide merged 1 commit into
integration/aurorafrom
j615/t8122-j613-j615-thermal-zones
Nov 22, 2026
Merged

RyanTheTide merged 1 commit into
integration/aurorafrom
j615/t8122-j613-j615-thermal-zones

Conversation

@RyanTheTide

Copy link
Copy Markdown
Contributor

Problem

The CPU thermal zones that keep the fanless M3 Air within limits live only in t8122-j613.dts. The J615 (15-inch) has the same die, the same six CPU die sensors behind smc_hwmon, and no fan, but its DTB has no zones. Nothing throttles it: with all eight cores busy, its die sensors pass 105 C in 150 s.

Change

Move the J613's thermal-zones node, unchanged, into t8122-j613-j615.dtsi, which both boards include. No values change and no driver changes.

Testing

  • t8122-j613.dtb before and after, decompiled and sorted: identical.
  • t8122-j615.dtb gains the same thermal-zones node as the J613 (phandles aside). CHECK_DTBS=y is clean for both boards.
  • The J615 has already been running these exact zones from its boot loader's lab DTB. Its live /sys/firmware/fdt thermal-zones node matches the new DTB's: same trips, sustainable power, contributions and cooling maps. A 10-minute all-core soak with those zones held the P-core zones at 86-91 C, with no critical trip. Eric Juta's overlay tests in drm/asahi: m3: T8122 GPU states 6 to 8 under the thermal limit, with the J615 die zone and CPU thermal zones #2203 show the same: 87.7 C peak over 170 s of eight busy cores, and 91-92 C during a 20-minute -j8 build.

Still to run

Assisted-by: LLM (Claude:Opus 5.5)

Nothing limits the J615's CPU temperature: the CPU thermal zones that
keep the fanless J613 safe are in t8122-j613.dts only. The J615 has the
same die, the same six CPU die sensors behind smc_hwmon and no fan
either; with all eight cores busy its die sensors pass 105 C in 150 s,
and a -j8 kernel build holds 110-114.6 C for 22 minutes.

Move the zones unchanged into t8122-j613-j615.dtsi, which both boards
include. The J613's compiled tree is unchanged (decompiled and sorted,
the old and new t8122-j613.dtb are identical); the J615's gains the
same thermal-zones node.

sustainable-power stays at 6 W. On a J615 with these zones applied as a
boot-time overlay, eight busy cores for 170 s peak at 87.7 C with the
P cores at 3.1-3.6 GHz, and a 20 minute -j8 kernel build holds 91-92 C
with the P cores at 2.9 GHz (peak 97.2 C). Those numbers are Eric
Juta's, from linux#2203.

Link: #2203
Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
@RyanTheTide

Copy link
Copy Markdown
Contributor Author

Looks good; approve once the J615 row below is posted.

The change is a pure move: the thermal-zones node in arch/arm64/boot/dts/apple/t8122-j613-j615.dtsi:203-505 matches what came out of t8122-j613.dts, and the <dt-bindings/thermal/thermal.h> include went with it (t8122-j613-j615.dtsi:15). I checked that the zones resolve the same way on the J615. smc_hwmon and its six CPU die sensors are declared first in t8122.dtsi:1133-1171 with #thermal-sensor-cells = <1>, so indices 0-5 are Te05..Tp0r on both boards. The J615 board file adds no smc_hwmon children that would shift those indices; the shared dtsi pulls in hwmon-laptop.dtsi at line 131, after them. The cpu_e*/cpu_p* cooling phandles come from t8122.dtsi, which both boards include.

HW row: J615, a kernel built from this head with its own t8122-j615.dtb (not the boot loader's lab DTB), default cmdline. Pass if /sys/class/thermal lists the six cpu-*-thermal zones, each bound to both cluster cooling devices; a 10-minute all-core soak holds the zones at or below the 95 C control trip with no critical trip; and the regression run shows 0 AVC denials, 0 failed units and 0 new kernel errors.

Assisted-by: LLM (Claude:Opus 5.5)

@RyanTheTide

Copy link
Copy Markdown
Contributor Author

J615 row at 107c2f8: J615 production default, 7.3.0-aurora-fleet79, which carries this head with the DTB built from the tree. Default command line.

  • /sys/class/thermal: the six zones cpu-e0, cpu-e1 and cpu-p0..p3-thermal are present. Each is bound to both cluster cooling devices. Trips are passive 58 C / passive 95 C / critical 110 C.
  • A 10-minute all-core soak on 8 CPUs (no fan on this Mac): the P zones climbed to about 89 C within 20 s. They were then held at 91-93.7 C by passive throttling, with the P policy going from 4.056 GHz to 2.616-2.808 GHz. The E zones ended at about 85-90 C.
  • One sample reached 95.49 C (cpu-p1, just over the 95 C passive trip) before the throttle pulled it back. No critical trip.
  • Before and after: Enforcing, 0 failed units, 0 AVC, 0 new kernel errors.

@RyanTheTide

Copy link
Copy Markdown
Contributor Author

Merge-train adversarial review (MERGETRAIN-LINUX2) at 107c2f8: NO BLOCKERS.

Code read only, nothing built here. Before the merge it still needs Dj's verdict at this head; the missing rows are listed below.

Full review

#2684 arm64: dts: apple: t8122-j613-j615: share the CPU thermal zones @107c2f88

VERDICT: NO BLOCKERS

What I checked:

  • It is a pure move. Diffing the removed and added lines leaves only the comment block different (t8122-j613-j615.dtsi:203-208 against the old t8122-j613.dts:20-24) plus the dt-bindings/thermal/thermal.h include. Trips are 58/95/110 C, sustainable-power 6000, contributions 1024/3072 and E capped at 3, all unchanged (t8122-j613-j615.dtsi:210-505).
  • The sensors are shared. smc_hwmon with its six CPU die children in index order 0..5 (Te05, Te0P, Tp05, Tp0L, Tp0b, Tp0r) is in t8122.dtsi:1133-1175, which both boards include, so the <&smc_hwmon N> indices mean the same on J615.
  • The base has not touched these files since the merge base (only spi2/spi4 clocks and AVD dma-coherent moved).
  • Gates and params: none. The trip values are in DT, which is an allowed carrier. They are copied from the J700 and J613 rather than read from the J615's ADT or SMC (see nit 1).

Nits:

  1. The safety limits are borrowed: "Trips as on the J700; sustainable-power from J613 (and J615) load tests". This is acceptable as DT, but no trip comes from this machine. The SMC exposes mTPL (t8122.dtsi, apple,thermal-pressure-key), so a follow-up could drive the zones from the machine's own pressure key and not from constants.
  2. The new comment says sustainable-power came "from J613 and J615 load tests". The J615 numbers are Eric Juta's overlay runs (drm/asahi: m3: T8122 GPU states 6 to 8 under the thermal limit, with the J615 die zone and CPU thermal zones #2203), not runs of this DTB. Fine, but the commit message is clearer than the code comment.

Merge-order hazard (not a blocker for this PR): #2203 (open, head b2db791) still carries d30c3bce "t8122-j615: Add CPU thermal zones". That commit adds the same node names and the same labels (cpu_e0_alert, cpu_e0_control, ...) to t8122-j615.dts. If #2203 is merged after this without dropping d30c3bce, dtc fails on duplicate labels. The train must drop that commit from #2203 first.

Missing rows:

  • None of this DTB itself has booted. The J615 evidence is a live /sys/firmware/fdt comparison against the lab bootloader DTB, plus Eric's overlay soaks. That is acceptable for a pure move with a sorted-decompile identity on J613.
  • J615 combined CPU+GPU load (for example a -j8 build plus a GPU benchmark). The zones throttle only the CPUs, and the 110 C critical trip is new for the J615, so before this PR it had no shutdown at all. A GPU-heavy workload could now take the J615 to an orderly critical poweroff at 110 C where before it ran hot. That is arguably correct, but it is unmeasured.
  • 25G83 variants inherit this; there is no row on either.

@DjDeveloperr DjDeveloperr left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Adversarial review at 107c2f88 against pre-pr-checklist.md (§1–§6): 0 code/policy/attribution/CI blockers, and required §5 evidence is complete on this head.

Verification

  • Pure move: Diffing removed lines in arch/arm64/boot/dts/apple/t8122-j613.dts against added lines in arch/arm64/boot/dts/apple/t8122-j613-j615.dtsi confirms all six cpu-*-thermal zones, trips (58 C / 95 C / 110 C), sustainable-power = <6000>, cooling maps (1024 P / 3072 E capped at state 3), and #include <dt-bindings/thermal/thermal.h> moved unchanged (only the comment block at t8122-j613-j615.dtsi:203-208 was updated).
  • Sensor & node safety: smc_hwmon in t8122.dtsi:1134-1176 defines the six CPU die sensors (Te05, Te0P, Tp05, Tp0L, Tp0b, Tp0r) first at indices 0..5 before hwmon-laptop.dtsi is included (t8122-j613-j615.dtsi:130). t8122-j615.dts adds no smc_hwmon children and has no conflicting thermal-zones node.
  • §1–§4, §6: Pass — shared t8122-j613-j615.dtsi (no model gates), default cmdline, DT-carried thermal limits protecting the fanless J615, human-only commit metadata + Signed-off-by:, Link: to #2203 with credit to Eric Juta, PR body Assisted-by:, CI dt-checks and host-tests green.
  • §5 Evidence (complete at 107c2f88): J615 production boot (7.3.0-aurora-fleet79, in-tree DTB, default cmdline) verified all six cpu-*-thermal zones bound to both clusters, 10-minute 8-CPU soak held P-cores at 91-93.7 C (single peak 95.49 C, 0 critical trips), SELinux Enforcing, 0 AVC, 0 failed units, 0 new kernel errors; J613 compiled DTB is identical before/after.

Merge-order note (non-blocking)

  • Open PR #2203 still carries commit d30c3bce adding the same thermal-zones node/labels to t8122-j615.dts; drop that commit from #2203 before merging it to avoid duplicate-label dtc failures.

Assisted-by: LLM (Gemini 4 Neon)

@RyanTheTide
RyanTheTide merged commit 2572862 into integration/aurora Nov 22, 2026
2 checks passed
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