Files

MechaCat02 ad45873a1b ITERATE-2.V: scheduler priority aging closes 18-day AUDIT-049 wedge

Priority aging in xenia-cpu/scheduler.rs:pick_runnable
(effective_priority = base + age_bonus(now_round - last_run_round),
capped at +31, AGING_ROUNDS_PER_BONUS=1). Strict-priority was parking
priority=0 threads behind CPU-bound priority=15 audio mixer
(sub_824D1328 guest spinwait at PC=0x824d1404 on CPU5). Aging
eventually picks the starved thread, breaking the producer-consumer
cycle that caused 5-tid wedge at PC=0x824ac578 since AUDIT-049 (10 May).

Cascade observed: tid=13 clean exit; events 121K -> 13M (107x); last
host_ns 767ms -> 51,011ms (66x); 8 new threads spawn; VdSwap 1 -> 2.

Complete two-day iterate sequence (2026-05-27 -> 2026-05-28):
- 2.F: VdSwap drain timeout 900ms -> 1ms (xenia-gpu/handle.rs); 876x
       perf win on VdSwap kernel callback
- 2.H: vA0000000 physical heap bucket added (state.rs, exports.rs);
       ctx_ptrs now in 0xA0000000-0xBFFFFFFF range matching canary
- 2.L: Phase-A diff harness categorized [return_value mismatch],
       [status mismatch], [args_resolved.path mismatch] tags
       (tools/diff-events/diff_events.py); closes reading-error #41
       (silent test-harness state leak invalidating trace diffs)
- 2.M: always-on exit-thread-state.json sibling to Phase-A JSONL
       (event_log.rs + xenia-app/main.rs); closes reading-error #42
       (Phase-A blind to blocked-forever waits)
- 2.Q: signal.match kernel instrumentation in NtSetEvent /
       NtReleaseSemaphore / KeSetEvent / KeReleaseSemaphore
       (exports.rs); emits target_handle + waiter_count + waiter_tids
- 2.T: wake.requested kernel instrumentation in wake_eligible_waiters
       (exports.rs); emits target_tid + transition + new_state
- 2.V: scheduler priority aging (xenia-cpu/scheduler.rs) [keystone]

Plus accumulated WIP from earlier May (contention_manifest,
phase_b_snapshot, xam/xaudio enhancements, analysis db, xex loader,
xenia-app main loop, etc.). Audit-runs/ artifacts remain untracked
per project convention.

Tests: 300 xenia-cpu / 227 xenia-kernel / 5 xenia-app / 19 xenia-path
/ 30+ smaller suites -- all PASS, 0 regressions. Determinism preserved
(2x cold runs bit-identical at 13,003,881 events post-2.V).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

2026-05-29 07:27:26 +02:00

2.6 KiB

Raw Blame History

address, classification, confidence, last_audit, aliases

address

classification

confidence

last_audit

aliases

0x82457EF0

thread_proc

high

060

tid=6 thread_proc

sub_82457EF0 — tid=6 thread_proc (worker entry)

Synopsis

Thread procedure for tid=6 in ours. 0 static callers — and that is correct for a thread_proc: it is installed as an entry-point via ExCreateThread somewhere in boot, not invoked via bl. AUDIT-059's "only-caller of sub_82458B90 has 0 callers — fnptr-array only" inference was wrong; the actual activation is thread creation.

Evidence

AUDIT-060 Probe O ours: fires 1× on tid=6 (HW=2, cycle=0, lr=0xbcbcbcbc — thread-entry sentinel).
lr=0xbcbcbcbc is the Xbox 360 / xenia convention for "this is the very first instruction of a thread proc; no return address". This is a diagnostic that distinguishes thread entry from a normal bl fire.
Calls sub_82458B90 at +0x24 (1 callee at this offset).

Activation

Registered as a thread entry-point via ExCreateThread (or similar). The caller of ExCreateThread that installs this entry has not yet been traced — that's the real activation site, and tracing it would close the loop on tid=6's purpose. Once tid=6 starts, the OS scheduler runs sub_82457EF0 from PC 0x82457EF0 with LR=0xbcbcbcbc.

Static graph

Static callers (bl): 0 (correct — see classification).
Callees: bl sub_82458B90 at +0x24 (PC 0x82457F18).
The "indirect call site" that activates this fn is the ExCreateThread invocation, captured at runtime, not in static xrefs.

Audit log

AUDIT-060 (2026-05-12) — identified as tid=6 thread_proc via lr=0xbcbcbcbc thread-entry sentinel + HW=2 + cycle=0 first-fire context. AUDIT-059's static-reachability inference invalidated. [confirmed]
AUDIT-059 (2026-05-11) — flagged as "only-caller of canary signaler A; 0 callers — fnptr-array only". [STATUS: partially correct (0 callers true; fnptr-array WRONG), corrected by AUDIT-060 — it's a thread_proc.]

Open questions

Where is ExCreateThread(entry=sub_82457EF0, ...) called from? Probe the ExCreateThread import thunk in both engines with filtered LR/r3 to find the install site.
What does the thread body do beyond calling sub_82458B90 once? Likely it's a loop that waits on a queue, dequeues work, and signals completion via the bl at +0x24. Disassemble the body.

Cross-references

Thread-body callee: sub_82458B90.
Install site (ExCreateThread caller): not yet identified.
Audits: 059, 060.
Artifacts: audit-runs/audit-060-fnptr-array-bootstrap/ours-phase1.stdout (the lr=0xbcbcbcbc sentinel evidence).

2.6 KiB Raw Blame History Unescape Escape