Crashes due to incorrect EPP handling

[NotWadeGrimridge]

I'm running c186c9a with the default config on Fedora 43, with tuned, tuned-ppd, and thermald disabled. My CPU: i7-8550U.

watt[4472]: [ERROR watt] failed to apply delta to CPU 0
watt[4472]: [ERROR watt] cause: this probably means that CPU 0 doesn't exist or doesn't support changing EPP
watt[4472]: [ERROR watt] cause: failed to write 'balance_performance' to '/sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference'
watt[4472]: [ERROR watt] cause: device or resource busy (os error 16)
systemd[1]: watt.service: Main process exited, code=exited, status=1/FAILURE
systemd[1]: watt.service: Failed with result 'exit-code'.

This happens because cpu0's governor is set to performance. With intel_pstate in active mode, cpufreq governors work completely different. Only two are available. powersave becomes a dynamic governor, and performance locks EPP to 0 and writes are blocked. Relevant code in the kernel: intel_pstate.c#L768 (docs). The fix is simple: if EPP/EPB support is available, set the governor to powersave before touching the EPP/EPB sysfs knobs.

The daemon doesn't set the governor uniformly either:

$ grep . /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor:performance
/sys/devices/system/cpu/cpu1/cpufreq/scaling_governor:powersave
/sys/devices/system/cpu/cpu2/cpufreq/scaling_governor:powersave
/sys/devices/system/cpu/cpu3/cpufreq/scaling_governor:performance
/sys/devices/system/cpu/cpu4/cpufreq/scaling_governor:performance
/sys/devices/system/cpu/cpu5/cpufreq/scaling_governor:powersave
/sys/devices/system/cpu/cpu6/cpufreq/scaling_governor:performance
/sys/devices/system/cpu/cpu7/cpufreq/scaling_governor:performance

This leads to:

$ grep . /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference
/sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference:performance
/sys/devices/system/cpu/cpu1/cpufreq/energy_performance_preference:balance_performance
/sys/devices/system/cpu/cpu2/cpufreq/energy_performance_preference:balance_performance
/sys/devices/system/cpu/cpu3/cpufreq/energy_performance_preference:performance
/sys/devices/system/cpu/cpu4/cpufreq/energy_performance_preference:performance
/sys/devices/system/cpu/cpu5/cpufreq/energy_performance_preference:balance_performance
/sys/devices/system/cpu/cpu6/cpufreq/energy_performance_preference:performance
/sys/devices/system/cpu/cpu7/cpufreq/energy_performance_preference:performance

And the daemon crashes when it hits the first write that fails. This inconsistency is probably caused by some other problem(?)

cc: @RGBCube

[NotAShelf]

This also explains the issue @faukah has been running into I believe. I haven't been able to reproduce the issue because I'm on AMD.

[RGBCube]

That makes sense, currently I've made it crash and burn on the first failure to squash out these bugs. Addressing this soon, eventually I'll make it print error logs instead of exit