Skip to content

arm64: dts: apple: t6050-j714s: let userspace set the fan speeds - #1944

Merged
DjDeveloperr merged 1 commit into
aurora-silicon:integration/aurorafrom
Isaac-Duarte:j714s/fan-manual-control
Nov 16, 2026
Merged

DjDeveloperr merged 1 commit into
aurora-silicon:integration/aurorafrom
Isaac-Duarte:j714s/fan-manual-control

Conversation

@Isaac-Duarte

@Isaac-Duarte Isaac-Duarte commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

Mark both J714s (M5 Pro) fans apple,fan-manual-control, as the M4 Pro/Max MacBook Pros and the J704 already are, so fanX_target is writable without the unsafe fan_control module parameter. This is for userspace fan control, such as a custom fan curve.

Nothing changes until userspace writes a target: the SMC keeps driving the fans, and writing 0 hands a fan back to the firmware. This stays within #670's policy (host fan writes only for explicit user targets). The M5 GPU governor already sets these fans through the same manual mode; a userspace target takes a fan over from it until 0 is written.

Testing

On the J714s (7.3.0-36.n1usb), booted with its installed device tree plus this property and fan_control=N:

  • both fanX_target are 0644 and the property is present on both fan nodes in /proc/device-tree;
  • the kernel is not tainted (bit 6 clear);
  • fan 1 set to 4000 rpm ran at 3996 rpm, and after writing 0 it returned to firmware control;
  • no macsmc-hwmon warnings in the kernel log;
  • with arm64: dts: apple: t6050-j714s: Enforce the PMP's fast-die effort #1950's fast-die limit, a userspace fan curve drove both fans through one sustained-load round (7833/7811 rpm) and then handed both back: they followed the firmware curve again, idling at 0 rpm and ramping under the next load to 3816/4114, then 4952/5347 rpm.

Even with #1950, the SMC's fan ceiling (fRMX) stays at 4950 rpm (the fans' maximum is 7826 rpm): under firmware control the post-#1950 run peaked at 4973 rpm on fan 1 and 5410 rpm on fan 2, with TCMb at 110.6 C. Only a host target can use the top third of the range.

Earlier revisions of this PR carried benchmarks taken before #1950; they were dropped because #1950 fixed that slowdown.

J716s and J775D, which share the T6050 SMC, are left as they are until their fans are tested the same way.

@Isaac-Duarte
Isaac-Duarte force-pushed the j714s/fan-manual-control branch from 36df945 to 7952426 Compare November 15, 2026 02:32
Mark both J714s fans apple,fan-manual-control, as the M4 Pro/Max
MacBook Pros and the J704 already are, so fanX_target is writable
without the unsafe fan_control module parameter, for userspace fan
control such as a custom fan curve.

The SMC still drives the fans until userspace writes a target, and
writing 0 hands a fan back to the firmware. The M5 GPU governor already
sets these fans through the same manual mode; a userspace target takes
a fan over from it until 0 is written.

J716s and J775D are left as they are until their fans are tested the
same way.

Tested on J714s (7.3.0-36.n1usb) with its installed device tree plus
this property and fan_control=N: both fanX_target are 0644, the kernel
is not tainted, fan 1 set to 4000 rpm ran at 3996 rpm and returned to
firmware control after 0 was written, with no macsmc-hwmon warnings.
Both fans are handed back the same way: after a userspace curve that
had them at 7833 and 7811 rpm let go, both followed the firmware curve
again, idling at 0 rpm and ramping under the next load to 3816/4114,
then 4952/5347 rpm.

Signed-off-by: Isaac Duarte <31712685+Isaac-Duarte@users.noreply.github.com>
@Isaac-Duarte
Isaac-Duarte force-pushed the j714s/fan-manual-control branch from 7952426 to d71ab67 Compare November 15, 2026 15:00
@Isaac-Duarte
Isaac-Duarte marked this pull request as ready for review November 15, 2026 15:00
@TripleU613

Copy link
Copy Markdown
Contributor

Test report from a second J714s (MacBook Pro 14", M5 Pro), in case it helps review.

  • Setup: 27.2 stub (26B123); kernel integration/aurora a5ebec5 plus this PR's commit unchanged (betizzel/omarchy-m5 linux-aurora v272-8 through v272-11); since 2026-11-15 also a from-scratch install of that image.
  • Result: fan1_target is writable with this change. Writing 3000 brought fan 1 to 2989 rpm; writing 0 handed it back to the firmware (0 rpm at idle). Without a userspace write the SMC keeps driving the fans as before. No errors in the kernel log.
  • Not tried here: a sustained-load comparison like the one in your description, and fan 2 on its own.

Thanks for this one.

🤖 Generated with Claude Code

@Isaac-Duarte
Isaac-Duarte force-pushed the j714s/fan-manual-control branch from d71ab67 to 7a479d4 Compare November 15, 2026 20:29
@DjDeveloperr
DjDeveloperr merged commit 4db2f2f into aurora-silicon:integration/aurora Nov 16, 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.

3 participants