Re: [BUG] Lunar Lake: 400 MHz after s2idle; PL1 Tau affects reproduction and recovery

[Scrimas <isma.ph@proton.me>]

On Mon, 7 Sep 2026 12:00:37 -0400, Sandor Nagy wrote:
> My ThinkPad X1 Carbon Gen 13 intermittently resumes from long s2idle
> suspends with all eight CPUs stuck around 400 MHz.

Same symptom here on a different OEM and a different SKU, so this does
not look Lenovo-specific:

  Samsung Galaxy Book5 Pro 360 (NP960QHA)
  Intel Core Ultra 7 256V (Lunar Lake), BIOS P17ALY.390.260616.03
  Kernel 7.2.9 (CachyOS), intel_pstate active, HWP, s2idle only
  platform_profile=performance, EPP=performance, on battery
  PL1 34 W (MSR and MMIO), Tau 27983872 us (same value as your X1C13)
  thermald not running

I log 60 s of turbostat and the xe GT throttle reasons after every
resume. Over 2026-10-02..05: 13 resumes after sleeps of 1.5 h or less
had no 400 MHz samples at all. The one resume after an 8.9 h s2idle
had all cores at 400 MHz for the first ~9 s of the log (the log
starts ~1 s after "PM: suspend exit"):

  Busy%  Bzy_MHz  PkgTmp  GFXMHz  PkgWatt
  16.37   400     25      400     2.38
  73.94   400     26      400     3.23
  49.75   400     27      400     2.87
  ...    (6 more samples at 400 MHz, 2.8-3.2 W)
  40.94   480     28      400     3.04
  37.83  3557     39      400    10.57

During exactly those seconds, gt0/freq0/throttle/reasons reported
"pl1", then "none". So the package was being power-limited while
drawing ~3 W against a 34 W PL1, at 25-28 C. That matches your
energy-averaging hypothesis.

Ruled out on this machine:
- ACPI _PPC: booted with processor.ignore_ppc=1, and scaling_max_freq
  stayed at 4.7 GHz during the episode.
- Thermal: package_throttle_count did not change during the window.

I have not captured MSR 0x19C bits 10-11 yet.

Thanks,
Scrimas