Re: [PATCH] ACPI: processor: thermal: Do not leave a stale limit when not throttling

[Hugo Sales]

I ran into the same bug on an ASUS ROG Flow X13 (Ryzen 9 5980HS,
amd-pstate active), whose battery profile turns boost off. I've been
carrying a similar fix since 2026-10-17 and found this patch while
reviewing mine before sending it. After boost came back, every policy
stayed at 3.3 GHz instead of 4.85 GHz until the Processor cooling
devices were moved to 1 and back to 0. With the fix, all 16 policies
return to 4.85 GHz. My copy reads reduction_step(cpu) once into a
local, which I'd suggest if you respin, but it is the same change.

Two additions for the commit message, if you think they help: the
cpufreq core lowers cpuinfo.max_freq in cpufreq_boost_set_sw() as well,
so acpi-cpufreq is affected too, and the thermal core writes state 0 at
zone registration, not only on resume.

Steps above 0 still keep the boost state of when they were set. I'm
looking at recomputing the limit when boost changes and will send that
separately as an RFC series.

Adding linux-pm, which this didn't go to.

Tested-by: Hugo Sales <hugo@xxxxxxx>