fan curves: don't factory-reset via pwmN_enable=3 when reading defaults (fixes asusd start hang + input freeze on TUF FX608JMR)

[aminezzy]

Problem

On the 2025 TUF F16 (FX608JMR, BIOS 306, kernel 6.12), asusd never finishes its first start on a machine with no existing fan_curves.ron. It hangs right after fan discovery while fetching default fan curves, systemd kills it at the start timeout, Restart=on-failure loops it — and the machine's input (mouse cursor) freezes repeatedly while this happens.

asusd[28547]: [INFO  asusd::ctrl_fancurves] Fetching default fan curves
asusd[28547]: [DEBUG rog_profiles] Looking for pwm1_enable
asusd[28547]: [DEBUG rog_profiles] Found pwm1_enable
asusd[28547]: [DEBUG rog_profiles] Looking for pwm2_enable
asusd[28547]: [DEBUG rog_profiles] Found pwm2_enable
asusd[28547]: [DEBUG rog_profiles] Looking for pwm3_enable
systemd[1]: asusd.service: start operation timed out. Terminating.

Root cause

FanCurveProfiles::set_active_curve_to_defaults() writes "3" to pwmN_enable, which makes the kernel re-fetch factory defaults via the fan_curve_get_factory_default WMI call. On this board that call blocks ~30 seconds, stalls the EC (that's the input freeze), and returns EIO. The kernel logs one of these per asusd start attempt, ~30s after the start:

kernel: asus_wmi: fan_curve_get_factory_default (0x00110024) failed: -5

Reads are unaffected — all pwmN_enable and pwmN_auto_pointM_* attributes read instantly with correct values from the driver's probe-time cache. Only the reset write triggers the blocked WMI call.

Fix

Read the curves the driver already reports instead of forcing a firmware re-fetch first. On first run these are the factory defaults anyway. This also protects the D-Bus reset path from wedging the EC.

If dropping the reset globally is undesirable for models where the re-fetch works, I'm happy to rework this as a DMI-based quirk instead. There is likely also a kernel-side/firmware issue here (fan_curve_get_factory_default blocking 30s and stalling the EC), which asusd was only exposing.

Testing

On the affected FX608JMR:

• asusd starts in ~1s (was: timeout loop + input freezes)
• per-profile default curves are read correctly for balanced/performance/quiet
• no asus_wmi errors in the kernel log
• fan curve writes (asusctl fan-curve) still work

Unrelated confirmation while I was testing: the FX608JM entry in aura_support.ron on main is correct for this machine — Static, Breathe, RainbowCycle, Pulse all verified working (single-zone, colours honoured).

🤖 Generated with Claude Code

https://claude.ai/code/session_014NneBinAM7WAFrqBqXZhA9

[scardracs]

You can't expect us to patch it when you are using a old kernel like the 6.12. If you want it merged you MUST provide it against newest stable kernel as clearly stated on the readme:

Due to ongoing driver work, the minimum suggested kernel version is always the latest, as improvements and fixes are continuous.

Support for TDP is tied to the new asus-armoury driver: available mainline since Linux 6.19: everything older is not supported.