muaddib-scanner 2.11.96 → 2.11.97

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "muaddib-scanner",
3
- "version": "2.11.96",
3
+ "version": "2.11.97",
4
4
  "description": "Supply-chain threat detection & response for npm & PyPI/Python",
5
5
  "main": "src/index.js",
6
6
  "bin": {
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "target": "node_modules",
3
- "timestamp": "2026-06-11T13:24:06.648Z",
3
+ "timestamp": "2026-06-11T15:01:04.220Z",
4
4
  "threats": [
5
5
  {
6
6
  "type": "string_mutation_obfuscation",
@@ -19,6 +19,7 @@ const { loadCachedIOCs } = require('../ioc/updater.js');
19
19
  const { scanPackageJson } = require('../scanner/package.js');
20
20
  const { scanShellScripts } = require('../scanner/shell.js');
21
21
  const { buildTrainingRecord } = require('../ml/feature-extractor.js');
22
+ const { appendWorkerMem } = require('./worker-mem.js');
22
23
  const { appendRecord: appendTrainingRecord, relabelRecords } = require('../ml/jsonl-writer.js');
23
24
 
24
25
  // From ./state.js
@@ -426,6 +427,17 @@ function runScanInWorker(extractedDir, timeoutMs, scanContext = null, signal = n
426
427
  const _sc = scanContext || {};
427
428
  _liveWorkers.set(worker, { name: _sc.name, version: _sc.version, ecosystem: _sc.ecosystem });
428
429
 
430
+ // Off-heap attribution (worker-mem.jsonl, gated MUADDIB_WORKER_MEM=1):
431
+ // process RSS around each worker's lifetime. tid captured now — after
432
+ // 'exit' worker.threadId becomes -1.
433
+ const _wmTid = worker.threadId;
434
+ const _wmSpawnedAt = Date.now();
435
+ appendWorkerMem({
436
+ ev: 'spawn', tid: _wmTid,
437
+ name: _sc.name, version: _sc.version, ecosystem: _sc.ecosystem,
438
+ rss: process.memoryUsage().rss
439
+ });
440
+
429
441
  let settled = false;
430
442
  let timer = null;
431
443
  const done = (fn) => {
@@ -462,9 +474,19 @@ function runScanInWorker(extractedDir, timeoutMs, scanContext = null, signal = n
462
474
 
463
475
  worker.on('error', (err) => done(() => reject(err)));
464
476
 
465
- worker.on('exit', (code) => done(() => {
466
- if (code !== 0) reject(new Error(`Worker exited with code ${code}`));
467
- }));
477
+ worker.on('exit', (code) => {
478
+ // 'exit' fires exactly once per worker (even after terminate/error), so
479
+ // it is the one reliable place to close the spawn/exit RSS pair.
480
+ appendWorkerMem({
481
+ ev: 'exit', tid: _wmTid,
482
+ name: _sc.name, version: _sc.version, code,
483
+ durMs: Date.now() - _wmSpawnedAt,
484
+ rss: process.memoryUsage().rss
485
+ });
486
+ done(() => {
487
+ if (code !== 0) reject(new Error(`Worker exited with code ${code}`));
488
+ });
489
+ });
468
490
  });
469
491
  }
470
492
 
@@ -0,0 +1,72 @@
1
+ 'use strict';
2
+
3
+ /**
4
+ * Per-worker memory instrumentation (off-heap RSS attribution, 2026-06).
5
+ *
6
+ * The EMERGENCY breaker fires on process RSS while the heap sits at ~15% —
7
+ * the driver is off-heap (malloc arenas + tarball Buffers) and mem-trend.jsonl
8
+ * only samples the whole process. This module attributes memory to individual
9
+ * scan workers / packages so the worker_threads → child_process decision can
10
+ * be made on data:
11
+ * H1: RSS stays high AFTER workers die → arenas never returned to the OS
12
+ * H2: RSS peaks only WHILE workers live → concurrent in-flight peak
13
+ *
14
+ * Producers:
15
+ * - queue.js (parent): ev:'spawn' / ev:'exit' around each scan worker,
16
+ * with process-wide RSS (delta attributable per package, noisy but
17
+ * aggregable over 24-48h).
18
+ * - scan-worker.js (worker): ev:'sample' every sampleIntervalMs() with the
19
+ * isolate-local heapUsed/external/arrayBuffers (rss there is process-wide).
20
+ *
21
+ * Same hot-path safety rules as spill.js (2026-06-11 prod-freeze lesson):
22
+ * append-only, stat-gated O(1) rotation, never read the file back, never throw.
23
+ * OFF unless MUADDIB_WORKER_MEM=1 (staged rollout, same pattern as
24
+ * MUADDIB_WORKER_MAX_OLD_MB) so tests and CLI runs never touch data/.
25
+ */
26
+
27
+ const fs = require('fs');
28
+ const path = require('path');
29
+
30
+ const DEFAULT_FILE = path.join(__dirname, '..', '..', 'data', 'worker-mem.jsonl');
31
+ const DEFAULT_MAX_MB = 64; // rotate past 64MB (file + .1 = 128MB worst case, ~2.5 days at concurrency 8)
32
+ const DEFAULT_SAMPLE_MS = 10000; // per-worker isolate sample cadence
33
+
34
+ function workerMemEnabled() {
35
+ return process.env.MUADDIB_WORKER_MEM === '1';
36
+ }
37
+
38
+ function workerMemFile() {
39
+ return process.env.MUADDIB_WORKER_MEM_FILE || DEFAULT_FILE;
40
+ }
41
+
42
+ /** 0 = sampling disabled (instrumentation off, or explicit MUADDIB_WORKER_MEM_SAMPLE_MS=0). */
43
+ function sampleIntervalMs() {
44
+ if (!workerMemEnabled()) return 0;
45
+ const v = parseInt(process.env.MUADDIB_WORKER_MEM_SAMPLE_MS, 10);
46
+ if (Number.isFinite(v) && v >= 0) return v;
47
+ return DEFAULT_SAMPLE_MS;
48
+ }
49
+
50
+ /**
51
+ * Append one instrumentation entry (ts stamped here). Bounded resource
52
+ * (CLAUDE.md §2): stat-gated truncate-rotate, no read-back on the hot path.
53
+ * @returns {boolean} true if a line was written
54
+ */
55
+ function appendWorkerMem(entry) {
56
+ if (!workerMemEnabled()) return false;
57
+ try {
58
+ const file = workerMemFile();
59
+ const maxMb = parseInt(process.env.MUADDIB_WORKER_MEM_MAX_MB, 10);
60
+ const maxBytes = (Number.isFinite(maxMb) && maxMb > 0 ? maxMb : DEFAULT_MAX_MB) * 1024 * 1024;
61
+ try {
62
+ const st = fs.statSync(file);
63
+ if (st.size > maxBytes) fs.renameSync(file, file + '.1');
64
+ } catch { /* no file yet — fine */ }
65
+ fs.appendFileSync(file, JSON.stringify({ ts: new Date().toISOString(), ...entry }) + '\n', 'utf8');
66
+ return true;
67
+ } catch { /* instrumentation must never crash the daemon or a worker */
68
+ return false;
69
+ }
70
+ }
71
+
72
+ module.exports = { appendWorkerMem, sampleIntervalMs, workerMemEnabled, workerMemFile };
@@ -12,7 +12,7 @@
12
12
  * parentPort.postMessage({ type: 'error', message: string })
13
13
  */
14
14
 
15
- const { parentPort, workerData } = require('worker_threads');
15
+ const { parentPort, workerData, threadId } = require('worker_threads');
16
16
 
17
17
  if (!parentPort) {
18
18
  // Not running as a worker — exit gracefully
@@ -20,17 +20,38 @@ if (!parentPort) {
20
20
  }
21
21
 
22
22
  const { run } = require('../index.js');
23
+ const { appendWorkerMem, sampleIntervalMs } = require('../monitor/worker-mem.js');
23
24
 
24
25
  (async () => {
26
+ // Off-heap attribution samples (worker-mem.jsonl): heapUsed/external/
27
+ // arrayBuffers are isolate-local here, rss is process-wide. The samples MUST
28
+ // NOT go through parentPort — the parent settles the scan promise on the
29
+ // first message it receives (queue.js done()), so a sample message would
30
+ // hang the scan forever. unref() so the timer never keeps the worker alive.
31
+ const scanContext = workerData.scanContext || {};
32
+ const everyMs = sampleIntervalMs();
33
+ let sampler = null;
34
+ if (everyMs > 0) {
35
+ sampler = setInterval(() => {
36
+ const m = process.memoryUsage();
37
+ appendWorkerMem({
38
+ ev: 'sample', tid: threadId,
39
+ name: scanContext.name, version: scanContext.version,
40
+ heapUsed: m.heapUsed, external: m.external, arrayBuffers: m.arrayBuffers, rss: m.rss
41
+ });
42
+ }, everyMs);
43
+ sampler.unref();
44
+ }
25
45
  try {
26
46
  // scanContext (optional) carries monitor-side info that opt-in scanners need
27
47
  // (e.g. trusted-dep-diff requires package name + version to query the registry).
28
48
  // It is spread INTO the pipeline options, but `_capture: true` always wins so
29
49
  // the worker keeps returning the result object — never prints.
30
- const scanContext = workerData.scanContext || {};
31
50
  const result = await run(workerData.extractedDir, { ...scanContext, _capture: true });
32
51
  parentPort.postMessage({ type: 'result', data: result });
33
52
  } catch (err) {
34
53
  parentPort.postMessage({ type: 'error', message: err.message || String(err) });
54
+ } finally {
55
+ if (sampler) clearInterval(sampler);
35
56
  }
36
57
  })();