Re: [PATCH 2/2] softirq: Treat an interrupted idle task as idle in tick_irq_exit()

From: Radu Toader

Date: Fri Nov 06 2026 - 21:16:19 EST


Both points are right and both are changelog-only, as the report says.
Fixed in v2.

On the first: PF_IDLE is indeed set by cpu_startup_entry() and not by
init_idle(), which I checked - init_idle() sets PF_KTHREAD and
PF_NO_SETAFFINITY only. So the gates overlap rather than one containing
the other, and "strictly wider" is wrong. v2 drops it, names the state
that is no longer covered, and says why nothing is lost there:
TS_FLAG_INIDLE is only set inside do_idle(), so tick_nohz_irq_exit() had
nothing to do before cpu_startup_entry(), and nohz_full CPUs stay
covered by tick_nohz_full_cpu(cpu).

On the second: injected idle is now named among the newly covered
states, with play_idle_precise() cited and the reach beyond sched_ext
spelled out - intel_powerclamp and cpuidle_cooling via
drivers/powercap/idle_inject.c. Peter Zijlstra raised the same point and
has a reply of his own in the thread; the gap in the changelog was real
and is closed.

No code change in v2.

Disclosure: this reply was written by Claude Opus 5 (Claude Code) and
sent on my behalf, in the same session as the patch; the patch changelog
carries the same note.