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

2.7 KiB

Raw Blame History

address, classification, confidence, last_audit, aliases

address

classification

confidence

last_audit

aliases

0x824F7CD0

normal_callee

high

064

AUDIT-058 caller-ladder fn #3

sub_824F7CD0 — middle of sub_825070F0 activation chain

Synopsis

Normal callee in the linear 4-fn activation chain ending at sub_825070F0. Calls sub_824F7800 at PC 0x824F8314. Has a 4-way computed bctr switch table near its entry (PCs 0x824F7D00..0x824F7D34 — a jump-table dispatch on [r31+0]-1 for values 1..4). AUDIT-064 verified canary fires 1× at ~60s wallclock; ours fires 0×.

Evidence

Disasm prolog at 0x824F7CD0: mflr r12; bl 0x825F0F68; stwu r1, -256(r1); ... — standard normal-callee prolog. NOT MSVC EH-handler shape.
Function size: 1736 bytes / 434 insns. has_eh=False, frame_size=256.
Static caller xref: 1 — bl from PC 0x824F83D4 inside sub_824F8398.
Computed jump-table at 0x824F7D10..0x824F7D24: lis r12, 0x824F; addi r12, r12, 32040; slwi r0, r11, 2; lwzx r0, r12, r0; mtctr r0; bctr — 4-way switch on argument. Targets at 0x824F7D28/2C/30/34/... are jump-table data, NOT call edges.
AUDIT-064 canary 60s probe: fires 1× with lr=0x824F83D8 r3=BE568F00 r4=701CF5B0 r5=701CF658 r6=03A72328 on tid=6. Reproduced bit-identical at 120s and 180s.
AUDIT-064 ours --ctor-probe=0x824F7CD0 -n 500M: 0 fires.

Activation

Direct bl from sub_824F8398+0x3C (PC 0x824F83D4).

Static graph

Static callers (from xrefs.source_func):
- PC 0x824F83D4 inside sub_824F8398.
Callees include sub_824F7800 (PC 0x824F8314), sub_824FD230, sub_824FD240, sub_824FC498, sub_824FCC18, and others.

Audit log

AUDIT-064 (2026-05-12) — disasm confirms normal-callee + 4-way computed jump-table near entry. Canary fires 1× / ours 0×. Single static caller is the actual runtime caller. Chain blocks upstream at the audit-049 wedge (tid=13 thread-join wait on handle 0x12A4). [confirmed]
AUDIT-058 (2026-05-10) — flagged as part of the ladder. [confirmed]

Open questions

The 4-way switch at 0x824F7D10..0x824F7D34: which jump-table entry corresponds to the path that calls sub_824F7800? Disasm shows lwz r11, 0(r31); subi r11, r11, 1; cmplwi cr6, r11, 0x3; bgt cr6, 0x824F80E4 — so input r4 (saved to r31) must be 1..4 to enter switch. Canary's r4 was 0x701CF5B0 (a stack ptr), so the value at [stack] indexes the switch.

Cross-references

Callees: sub_824F7800, sub_824FD230/40, sub_824FC498, sub_824FCC18.
Callers: sub_824F8398+0x3C.
Audits: 058, 064.
Artifacts: audit-runs/audit-064-activation-ladder/canary-{60,120,180}s.log, audit-runs/audit-064-activation-ladder/ours-500M.stdout.

2.7 KiB Raw Blame History Unescape Escape