feat(sched): scheduler perf/stability baseline recorder (ADR-0022 prereq 2, PR-1)

[NickFlach]

What

PR-1 of ADR-0022 prerequisite 2 — a scheduler perf/stability baseline so a future scheduler-constant change (the deferred COM2-latency fix) is measured, not blind (ADR-0016). Measurement-only: no timer HZ / SCHED_QUANTUM_TICKS / service-model change.

The design panel earned its keep

An adversarial design panel refuted the obvious metric: aggregate context switches are dominated by voluntary yield()s — every long-lived citizen busy-yields with a non-blocking recv, resetting quantum_counter thousands of times per 10 ms tick, so the timer-preemption path is nearly dead at rest. switch_count/tick is therefore both quantum-insensitive (a gate on it is vacuous) and host-throughput dependent (CI-flaky). The fix is a dedicated preempt_count incremented only at the timer-quantum-expiry reschedule in scheduler_tick (never on SYS_YIELD) — 1/quantum-paced by construction.

Instrumentation (minimal, measurement-only)

• scheduler.c: preempt_count + scheduler_get_preempts(), and a read-only scheduler_get_fairness() (max reschedule gap, run-count spread, runnable count).
• process.c: the dead now = 0 /* TODO */ stub in process_switch_to goes live (timer_get_ticks()), plus a per-PCB sched_picks counter — so last_scheduled/runtime_* stop being always-zero.
• SYSINFO_SCHED sub-op (no new syscall → no ADR-0020 golden churn): one atomic line SCHED: switches=.. preempt=.. ticks=.. maxgap=.. spread=.. runnable=.., driven by a qsh sched command (argument-free) and QosVM.sched().

Report-only gate

make ci-smoke-sched (CI job scheduler-baseline, not release-gating) runs scripts/test_qos_sched.py in report-only mode: it records the tick-normalized baseline and asserts only a non-vacuous liveness floor (≥100 switches and ≥12 runnable over the window), so it reddens on a dead/wedged scheduler but arms no perf band yet.

Calibration selected PR-2's gated scalar

Empirical WSL calibration — preempt-per-1000-guest-ticks tracks 1/quantum almost exactly, host-invariantly:

SCHED_QUANTUM_TICKS	preempt/1000t (idle, load)	switch/1000t
5 (default)	199.5–200.0	~520
20	49.9, 49.8	~131
80	harness times out (system too sluggish to fill the tick window)

A 4× quantum change moves it 4× with <0.5 % variance. The follow-up arms a floor of ~100 (= q5-median × 0.5), revert-confirmed by a SCHED_QUANTUM_TICKS bump to 20 (q=80 grinds the harness itself, so 20 is the practical lever).

Verification

• Build clean under -Werror; clang-format / cppcheck / check-api-consistency.sh / YAML all green.
• make ci-smoke-sched passes locally (idle P=199.9, load P=200, liveness floor asserted).

🤖 Generated with Claude Code