ktstr 0.17.0

Test harness for Linux process schedulers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
use super::*;

// ---- Datasec rendering ------------------------------------------
//
// The renderer recognises `BTF_KIND_DATASEC` (the value type
// libbpf assigns to a global-section ARRAY map like `.bss`) and
// walks its `VarSecinfo` entries to render each variable. Before
// this support landed the renderer returned `Unsupported`, so a
// failure dump's `.bss` map showed an opaque hex dump instead of
// `stall=1, crash=0, ...`.
//
// The probe BPF object built by `build.rs` contains a known
// `.bss` Datasec (declared via the `volatile u32
// ktstr_err_exit_detected = 0;` latch, the per-CPU counter
// array `ktstr_pcpu_counters`, the sticky `ktstr_last_trigger_ts`
// / `ktstr_exit_*` snapshot vars, and the `ktstr_miss_log` /
// `ktstr_miss_log_idx` log buffer in `src/bpf/probe.bpf.c`). The
// tests below load that BTF directly via `load_btf_from_path`
// (which falls back to goblin's `.BTF` ELF section parse for
// non-vmlinux files) and exercise the Datasec render path
// against it. Hard-fail on a missing probe.o because build.rs
// always produces it; a silent skip would hide the regression
// the test is designed to catch.

/// Locate the `.bss` Datasec type id in the probe BTF.
/// `resolve_types_by_name(".bss")` returns a list of types named
/// `.bss`; libbpf normally emits exactly one `BTF_KIND_DATASEC`,
/// but the resolver returns Vec<Type> so we scan for the
/// Datasec variant. Returns `(btf, ds_id)` or panics if the
/// build fixture is missing.
fn load_probe_btf_and_bss_id() -> (Btf, u32) {
    let probe_obj = std::path::PathBuf::from(env!("OUT_DIR")).join("probe.o");
    let btf = crate::monitor::btf_offsets::load_btf_from_path(&probe_obj).unwrap_or_else(|e| {
        panic!(
            "load_btf_from_path({}) failed: {e}. \
                 build.rs always produces probe.o; a missing or \
                 unparseable artifact means the build pipeline is \
                 broken.",
            probe_obj.display()
        )
    });
    let ids = btf
        .resolve_ids_by_name(".bss")
        .expect("probe BTF must carry a `.bss` BTF_KIND_DATASEC");
    // Pick the first id that resolves to a Datasec — there
    // should be exactly one, but we don't blow up if libbpf
    // ever emits something else under the `.bss` name.
    for &id in &ids {
        if let Ok(Type::Datasec(_)) = btf.resolve_type_by_id(id) {
            return (btf, id);
        }
    }
    panic!("probe BTF has `.bss` ids {ids:?} but none resolve to BTF_KIND_DATASEC");
}

#[test]
fn render_datasec_emits_struct_with_named_variables() {
    let (btf, bss_id) = load_probe_btf_and_bss_id();
    // Compute the section size by summing each VarSecinfo's
    // (offset + size) — libbpf-emitted Datasecs aren't laid out
    // contiguously (alignment + ordering), so the section's
    // total size is `max(offset + size)` across all entries.
    // Allocate a zeroed buffer of that size so every variable's
    // slice fits.
    let Type::Datasec(ds) = btf.resolve_type_by_id(bss_id).unwrap() else {
        panic!(".bss id did not resolve to Datasec");
    };
    let section_size = ds
        .variables
        .iter()
        .map(|v| v.offset() as usize + v.size())
        .max()
        .expect("`.bss` Datasec must have at least one variable");
    let bytes = vec![0u8; section_size];
    let rendered = render_value(&btf, bss_id, &bytes);

    // Datasec must render as a Struct (not Unsupported, not
    // Truncated — section_size matches the actual section
    // extent so no variable should overrun).
    let RenderedValue::Struct { type_name, members } = rendered else {
        panic!(
            "expected RenderedValue::Struct for Datasec, got something else \
             — Datasec dispatch in render_value_inner must be reachable"
        );
    };
    assert_eq!(
        type_name.as_deref(),
        Some(".bss"),
        "section name must surface as type_name"
    );
    // Variable names: probe.bpf.c declares
    // `ktstr_err_exit_detected` as a writable global. It MUST
    // appear in the rendered .bss members; the freeze
    // coordinator depends on this exact name being resolvable.
    let names: std::collections::HashSet<&str> = members.iter().map(|m| m.name.as_str()).collect();
    assert!(
        names.contains("ktstr_err_exit_detected"),
        "rendered .bss must contain `ktstr_err_exit_detected` \
         (the freeze latch). Found names: {names:?}"
    );
    // Diagnostic counters and snapshots are writable globals →
    // expected in .bss too. Pin every variable name a downstream
    // consumer relies on so a future addition that lands in
    // .bss without renderer coverage surfaces here. The hot
    // counters live in the `ktstr_pcpu_counters` per-CPU array
    // (host-side reader sums across CPUs); sticky snapshot vars
    // remain individual globals because they're written exactly
    // once per error-class exit.
    for required in [
        // Per-CPU diagnostic counter array — replaces the
        // previous N independent globals
        // (ktstr_trigger_count / ktstr_probe_count / etc.).
        "ktstr_pcpu_counters",
        // Sticky timestamp + scheduler-state snapshots written
        // by the tp_btf/sched_ext_exit handler at the first
        // error-class exit.
        "ktstr_last_trigger_ts",
        // SCX_EV_* counter snapshot taken by `scx_bpf_events`
        // at the first error-class exit. Surfaces the
        // system-wide event totals at fault time.
        "ktstr_exit_event_stats",
    ] {
        assert!(
            names.contains(required),
            "rendered .bss must contain `{required}` \
             diagnostic counter. Found names: {names:?}"
        );
    }
    // Each member's `value` must be a concrete renderable
    // type (Uint, Int, Array of int, etc.) — NOT Unsupported.
    // A zero byte buffer can't be Truncated for variables that
    // fit within section_size, so any Truncated result here
    // would indicate a slicing bug.
    for m in &members {
        assert!(
            !matches!(m.value, RenderedValue::Unsupported { .. }),
            "member {:?} rendered as Unsupported: {:?}",
            m.name,
            m.value
        );
        assert!(
            !matches!(m.value, RenderedValue::Truncated { .. }),
            "member {:?} rendered as Truncated despite section_size \
             buffer: {:?}",
            m.name,
            m.value
        );
    }
    // ktstr_err_exit_detected is `volatile u32 = 0;` — must
    // decode to Uint{bits:32, value:0} given a zero buffer.
    let latch = members
        .iter()
        .find(|m| m.name == "ktstr_err_exit_detected")
        .expect("latch member must be present (asserted above)");
    match &latch.value {
        RenderedValue::Uint { bits, value } => {
            assert_eq!(*bits, 32, "latch is u32 (32 bits)");
            assert_eq!(*value, 0, "latch was zeroed in the buffer");
        }
        other => panic!("expected Uint{{32,0}} for latch, got {other:?}"),
    }
}

#[test]
fn render_datasec_truncates_overrunning_variables() {
    // Feed a byte buffer that's too small to cover every
    // variable in the .bss Datasec. Variables whose
    // (offset + size) extends past the buffer must surface as
    // Truncated members, while variables that fit must render
    // normally. The Struct itself is NOT wrapped in Truncated
    // — the section name and the per-variable partial render
    // both stay intact.
    let (btf, bss_id) = load_probe_btf_and_bss_id();
    let Type::Datasec(ds) = btf.resolve_type_by_id(bss_id).unwrap() else {
        panic!(".bss id did not resolve to Datasec");
    };
    // Buffer that holds only the first variable (smallest
    // offset). Variables with higher offsets become Truncated.
    let min_var = ds
        .variables
        .iter()
        .min_by_key(|v| v.offset())
        .expect("`.bss` must have at least one variable");
    let buf_size = (min_var.offset() as usize) + min_var.size();
    let bytes = vec![0u8; buf_size];
    let rendered = render_value(&btf, bss_id, &bytes);

    let RenderedValue::Struct { type_name, members } = rendered else {
        panic!("expected RenderedValue::Struct even with short buffer");
    };
    assert_eq!(type_name.as_deref(), Some(".bss"));
    // At least one member must be Truncated (one of the
    // higher-offset variables). At least one member must be
    // non-Truncated (the variable at min_var's offset).
    let truncated_count = members
        .iter()
        .filter(|m| matches!(m.value, RenderedValue::Truncated { .. }))
        .count();
    let decoded_count = members.len() - truncated_count;
    assert!(
        decoded_count >= 1,
        "at least one member must decode (the variable at the smallest offset, \
         which fits in buf_size={buf_size})"
    );
    // If there is more than one variable in .bss (probe.bpf.c
    // declares several), the short buffer must produce at
    // least one Truncated. A single-variable .bss would have
    // truncated_count == 0, but our probe has multiple — so
    // assert > 0.
    if members.len() > 1 {
        assert!(
            truncated_count >= 1,
            "multi-variable .bss with short buffer must produce >= 1 \
             Truncated member; got 0 from {members:?}"
        );
    }
}

#[test]
fn render_datasec_empty_buffer_yields_struct_with_truncated_members() {
    // Edge case: zero-byte buffer for a non-empty Datasec.
    // Every variable must surface as Truncated rather than
    // crashing the renderer or returning the legacy
    // Unsupported.
    let (btf, bss_id) = load_probe_btf_and_bss_id();
    let rendered = render_value(&btf, bss_id, &[]);
    let RenderedValue::Struct { members, .. } = rendered else {
        panic!("expected Struct render even with empty buffer");
    };
    assert!(!members.is_empty(), "probe `.bss` Datasec is non-empty");
    for m in &members {
        // Every variable should report Truncated{ needed: var
        // size, had: 0, partial: ... }.
        assert!(
            matches!(m.value, RenderedValue::Truncated { had: 0, .. }),
            "member {:?} should be Truncated{{had:0}} for empty buffer, got {:?}",
            m.name,
            m.value
        );
    }
}

// ---- format_cpu_list -------------------------------------------
//
// Range-collapses a sorted CPU id list into a compact string.
// Pin every shape: empty, single, single-range, sparse (gaps),
// multiple ranges + singletons. Collapse rule: a run of >= 2
// consecutive ids renders as `start-end`; a singleton renders as
// just the id.

#[test]
fn format_cpu_list_empty_is_empty_string() {
    assert_eq!(format_cpu_list(&[]), "");
}

#[test]
fn format_cpu_list_single_element() {
    assert_eq!(format_cpu_list(&[5]), "5");
}

#[test]
fn format_cpu_list_contiguous_range() {
    assert_eq!(format_cpu_list(&[0, 1, 2, 3, 4]), "0-4");
}

#[test]
fn format_cpu_list_two_consecutive_collapses_to_range() {
    // The two-element edge: 0,1 must render as "0-1", not "0,1".
    // The end-of-loop flush has its own start==end branch, so a
    // pure-range input exercises the in-loop range emission.
    assert_eq!(format_cpu_list(&[0, 1]), "0-1");
}

#[test]
fn format_cpu_list_gaps_between_ranges() {
    // Mixed: range, singleton, range. Pins the comma-separator
    // and the singleton-formatting path inside the loop.
    assert_eq!(format_cpu_list(&[0, 1, 2, 5, 7, 8, 9]), "0-2,5,7-9");
}

#[test]
fn format_cpu_list_all_singletons() {
    assert_eq!(format_cpu_list(&[0, 2, 4, 6]), "0,2,4,6");
}

#[test]
fn format_cpu_list_first_range_then_singleton() {
    // Trailing singleton — covers the post-loop flush after a
    // mid-list range when the final cpu is alone.
    assert_eq!(format_cpu_list(&[0, 1, 5]), "0-1,5");
}

#[test]
fn format_cpu_list_singleton_then_trailing_range() {
    // Leading singleton followed by a closing range — the
    // post-loop flush emits the trailing range.
    assert_eq!(format_cpu_list(&[0, 3, 4, 5]), "0,3-5");
}

// ---- try_render_cpumask_bits -----------------------------------
//
// Reads u64 LE words from the byte slice and renders set bits as
// a CpuList. The function:
//   - returns None when fewer than 8 bytes are supplied
//     (insufficient for a single u64 word)
//   - returns Some(CpuList { "" }) when bytes are zeroed
//   - extracts bits across multiple words at offset = word*64+bit

#[test]
fn try_render_cpumask_bits_too_short_returns_none() {
    // Strictly less than 8 bytes can't form a u64 word.
    assert!(try_render_cpumask_bits(&[], u32::MAX).is_none());
    assert!(try_render_cpumask_bits(&[0u8; 1], u32::MAX).is_none());
    assert!(try_render_cpumask_bits(&[0u8; 7], u32::MAX).is_none());
}

#[test]
fn try_render_cpumask_bits_all_zero_yields_empty_list() {
    // 8 zeroed bytes: Some(CpuList { cpus: "" }) — the loop sees
    // word=0 and skips it; format_cpu_list of an empty Vec is "".
    let v = try_render_cpumask_bits(&[0u8; 8], u32::MAX);
    match v {
        Some(RenderedValue::CpuList { cpus }) => {
            assert_eq!(cpus, "", "all-zero bytes must produce empty cpu list");
        }
        other => panic!("expected Some(CpuList), got {other:?}"),
    }
}

#[test]
fn try_render_cpumask_bits_single_word_low_bits() {
    // Single word with bits 0,1,2 set → "0-2".
    let bits: u64 = 0b111;
    let bytes = bits.to_le_bytes();
    let v = try_render_cpumask_bits(&bytes, u32::MAX);
    match v {
        Some(RenderedValue::CpuList { cpus }) => assert_eq!(cpus, "0-2"),
        other => panic!("expected CpuList with 0-2, got {other:?}"),
    }
}

#[test]
fn try_render_cpumask_bits_single_word_high_bit() {
    // Bit 63 set → cpu 63.
    let bits: u64 = 1u64 << 63;
    let bytes = bits.to_le_bytes();
    let v = try_render_cpumask_bits(&bytes, u32::MAX);
    match v {
        Some(RenderedValue::CpuList { cpus }) => assert_eq!(cpus, "63"),
        other => panic!("expected CpuList with 63, got {other:?}"),
    }
}

#[test]
fn try_render_cpumask_bits_fully_online_above_128_cpus() {
    // Four all-ones u64 words = CPUs 0-255 fully online. The old
    // `word > 0xFFFF_FFFF && set_cpus.len() > 64` gate bailed at word 2,
    // silently dropping CPUs 128-255; the all-ones-exempting gate (mirror
    // of the kptr path) must render the full set.
    let bytes = [0xFFu8; 32];
    let v = try_render_cpumask_bits(&bytes, 256);
    match v {
        Some(RenderedValue::CpuList { cpus }) => {
            assert_eq!(
                cpus, "0-255",
                "fully-online 256-CPU mask must render all CPUs, got {cpus}"
            );
        }
        other => panic!("expected CpuList 0-255, got {other:?}"),
    }
}

#[test]
fn try_render_cpumask_bits_caps_at_nr_cpu_ids() {
    // 8 CPUs: bits 0..=7 are real, bits 8..=63 are slab padding
    // / freelist garbage. With max_cpus=8, only bits 0..=7
    // should appear in the rendered list even when the word
    // has additional bits set higher up. Pins the fix:
    // an 8-CPU guest must not render bits 64..4035 from a
    // 1024-byte cpumask slab.
    let bits: u64 = 0xFFFF_FFFF_FFFF_FFFF; // every bit set
    let bytes = bits.to_le_bytes();
    let v = try_render_cpumask_bits(&bytes, 8);
    match v {
        Some(RenderedValue::CpuList { cpus }) => {
            assert_eq!(cpus, "0-7", "max_cpus=8 must cap at cpu 7, got {cpus}");
        }
        other => panic!("expected CpuList with 0-7, got {other:?}"),
    }
}

#[test]
fn try_render_cpumask_bits_caps_across_word_boundary() {
    // Two words, all bits set. max_cpus=8 must stop walking
    // immediately in word 0 (cap is partial-word). max_cpus=
    // 64 must walk word 0 fully and stop at the start of word
    // 1 (cap is whole-word).
    let mut bytes = [0u8; 16];
    bytes[0..8].copy_from_slice(&u64::MAX.to_le_bytes());
    bytes[8..16].copy_from_slice(&u64::MAX.to_le_bytes());

    // Partial-word cap inside word 0.
    let v = try_render_cpumask_bits(&bytes, 8);
    match v {
        Some(RenderedValue::CpuList { cpus }) => assert_eq!(cpus, "0-7"),
        other => panic!("expected CpuList 0-7, got {other:?}"),
    }

    // Whole-word cap: 64 means stop at start of word 1.
    // (Word 0 contains bits 0..=63, all 64 of them set.)
    let v = try_render_cpumask_bits(&bytes, 64);
    match v {
        Some(RenderedValue::CpuList { cpus }) => assert_eq!(cpus, "0-63"),
        other => panic!("expected CpuList 0-63, got {other:?}"),
    }
}

#[test]
fn try_render_cpumask_bits_multi_word_offsets() {
    // Two words: word[0] bit 0 → cpu 0, word[1] bit 0 → cpu 64,
    // word[1] bit 1 → cpu 65. Pins the word-index to cpu-id
    // arithmetic (word*64 + bit).
    let mut bytes = [0u8; 16];
    bytes[0..8].copy_from_slice(&1u64.to_le_bytes());
    let w1: u64 = 0b11;
    bytes[8..16].copy_from_slice(&w1.to_le_bytes());
    let v = try_render_cpumask_bits(&bytes, u32::MAX);
    match v {
        Some(RenderedValue::CpuList { cpus }) => assert_eq!(cpus, "0,64-65"),
        other => panic!("expected CpuList with 0,64-65, got {other:?}"),
    }
}

#[test]
fn try_render_cpumask_bits_partial_trailing_bytes_ignored() {
    // 12 bytes (1.5 words) — only the first complete word
    // (8 bytes) parses. The trailing 4 bytes are ignored
    // because n_words = 12/8 = 1.
    let mut bytes = [0u8; 12];
    bytes[0..8].copy_from_slice(&1u64.to_le_bytes());
    // Trailing 4 bytes hold bit 0 (cpu 32 if read as a word)
    // but should NOT be parsed.
    bytes[8] = 0xff;
    let v = try_render_cpumask_bits(&bytes, u32::MAX);
    match v {
        Some(RenderedValue::CpuList { cpus }) => assert_eq!(cpus, "0"),
        other => panic!("expected CpuList with 0, got {other:?}"),
    }
}

/// Garbage cpumask data with a sane nr_cpu_ids cap produces
/// capped output rather than enumerating phantom CPUs from
/// slab-padding / freelist bytes. 16 words (128 bytes) all-FF
/// would otherwise surface 1024 bit-positions; with max_cpus=4
/// only cpus 0..=3 must appear. Pins the backstop: the
/// nr_cpu_ids cap protects the renderer when SLAB_FREELIST_
/// HARDENED XOR-encoding defeats the top-byte heuristic.
#[test]
fn try_render_cpumask_bits_garbage_capped_at_max_cpus() {
    // 128 bytes = 16 u64 words, every bit set. Without the
    // cap this would render 0-1023; with max_cpus=4 the walker
    // stops after bit 3 of word 0.
    let bytes = vec![0xFFu8; 128];
    let v = try_render_cpumask_bits(&bytes, 4);
    match v {
        Some(RenderedValue::CpuList { cpus }) => {
            assert_eq!(
                cpus, "0-3",
                "max_cpus=4 must clip 1024-bit garbage to cpus 0-3, got: {cpus}",
            );
        }
        other => panic!("expected CpuList 0-3, got {other:?}"),
    }
}

/// A multi-word inline cpumask whose trailing word carries the
/// canonical kernel-pointer top byte (`0xff..`) but is NOT all-ones
/// must trigger the per-word plausibility break (`try_render_cpumask_bits`'s
/// `word != u64::MAX && word >> 56 == 0xff`) — that word is slab /
/// pointer garbage, not online CPUs. word0 = 0b11 (cpus 0,1); word1 =
/// 0xff00_0000_0000_0000 (top byte 0xff; bits 120-127 if trusted).
/// max_cpus = 256 is well past word1, so the cap is NOT what stops the
/// walk (that path is pinned by
/// `try_render_cpumask_bits_garbage_capped_at_max_cpus`) — only the
/// 0xff break can drop word1, so the result is "0-1", never
/// "0-1,120-127". The all-ones EXEMPTION side (a fully-online >128-CPU
/// host) is pinned by `try_render_cpumask_bits_fully_online_above_128_cpus`.
#[test]
fn try_render_cpumask_bits_kptr_pattern_word_breaks_walk() {
    let mut bytes = [0u8; 16];
    bytes[0..8].copy_from_slice(&0b11u64.to_le_bytes());
    // Top byte 0xff, rest zero: a kernel-pointer-class word, distinct
    // from all-ones (a legitimately fully-online 64-CPU chunk).
    let w1: u64 = 0xff00_0000_0000_0000;
    bytes[8..16].copy_from_slice(&w1.to_le_bytes());
    let v = try_render_cpumask_bits(&bytes, 256);
    match v {
        Some(RenderedValue::CpuList { cpus }) => assert_eq!(
            cpus, "0-1",
            "0xff-top-byte word must break the walk; only cpus 0,1 from \
             word0 may surface (the cap is well past word1), got: {cpus}",
        ),
        other => panic!("expected CpuList 0-1, got {other:?}"),
    }
}

/// `max_cpus = 0` produces an empty cpu list. The whole-word
/// gate (`word_first_cpu >= max_cpus as u64`) fires immediately
/// at word 0. Defensive: a malformed reader returning
/// `nr_cpu_ids = 0` (not the trait default `u32::MAX`) would
/// otherwise expose the per-bit `cpu >= max_cpus` check to a
/// loop entry; verify both paths converge on empty output.
#[test]
fn try_render_cpumask_bits_max_cpus_zero_yields_empty_list() {
    let bits: u64 = 0xFFFF_FFFF_FFFF_FFFF;
    let bytes = bits.to_le_bytes();
    let v = try_render_cpumask_bits(&bytes, 0);
    match v {
        Some(RenderedValue::CpuList { cpus }) => {
            assert_eq!(cpus, "", "max_cpus=0 must produce empty list, got: {cpus}");
        }
        other => panic!("expected empty CpuList, got {other:?}"),
    }
}

/// Cap matches the actual mask width (max_cpus=64, all 64 bits
/// set in word 0): every bit must surface. A regression that
/// used `>=` for the per-bit cap (clipping the last bit) would
/// produce "0-62" instead of "0-63". Pins the upper-edge
/// off-by-one.
#[test]
fn try_render_cpumask_bits_max_cpus_matches_word_width_keeps_all_bits() {
    let bits: u64 = u64::MAX;
    let bytes = bits.to_le_bytes();
    let v = try_render_cpumask_bits(&bytes, 64);
    match v {
        Some(RenderedValue::CpuList { cpus }) => {
            assert_eq!(
                cpus, "0-63",
                "max_cpus=64 must surface all 64 bits, got: {cpus}",
            );
        }
        other => panic!("expected CpuList 0-63, got {other:?}"),
    }
}

/// `MemReader` trait default `nr_cpu_ids` is `u32::MAX`.
/// Callers that don't override produce no cap, preserving the
/// pre-fix behavior (every set bit reported). A regression
/// that flipped the default to `0` would silently empty every
/// cpumask render.
#[test]
fn mem_reader_default_nr_cpu_ids_is_u32_max() {
    struct DefaultReader;
    impl MemReader for DefaultReader {
        fn read_kva(&self, _: u64, _: usize) -> Option<Vec<u8>> {
            None
        }
    }
    let r = DefaultReader;
    assert_eq!(
        r.nr_cpu_ids(),
        u32::MAX,
        "default nr_cpu_ids must be u32::MAX",
    );
}

/// A custom `MemReader` impl overrides `nr_cpu_ids`. Pin the
/// override path so a regression that ignored the override
/// (always returning the default) is caught.
#[test]
fn mem_reader_custom_nr_cpu_ids_returns_overridden_value() {
    struct CustomReader {
        cpu_count: u32,
    }
    impl MemReader for CustomReader {
        fn read_kva(&self, _: u64, _: usize) -> Option<Vec<u8>> {
            None
        }
        fn nr_cpu_ids(&self) -> u32 {
            self.cpu_count
        }
    }
    let r = CustomReader { cpu_count: 16 };
    assert_eq!(r.nr_cpu_ids(), 16);
}

/// `render_struct` consults `MemReader::nr_cpu_ids` when
/// rendering a cpumask-family struct. With max_cpus=8 supplied
/// by the reader, garbage bits beyond cpu 7 are dropped —
/// the `let max_cpus = mem.map(|m| m.nr_cpu_ids())`
/// in `render_struct` (btf_render.rs) wires the reader value
/// through to `try_render_cpumask_bits`. A regression that
/// passed `u32::MAX` instead would surface phantom cpus 8..63.
#[test]
fn render_value_with_mem_caps_cpumask_at_reader_nr_cpu_ids() {
    let Some(btf) = test_btf() else {
        crate::report::test_skip("test_btf returned None");
        return;
    };
    let Ok(ids) = btf.resolve_ids_by_name("cpumask") else {
        crate::report::test_skip("BTF missing 'cpumask' struct");
        return;
    };
    let Some(&id) = ids.first() else {
        crate::report::test_skip("BTF resolved 'cpumask' to empty id list");
        return;
    };
    // Resolve the underlying struct so we can size the buffer.
    let Some(ty) = peel_modifiers(&btf, id) else {
        crate::report::test_skip("could not peel cpumask modifiers");
        return;
    };
    let size = match type_size(&btf, &ty) {
        Some(n) if n >= 8 => n,
        _ => {
            crate::report::test_skip("cpumask size unresolved or < 8");
            return;
        }
    };
    // Fill the entire cpumask buffer with all-FFs garbage.
    let bytes = vec![0xFFu8; size];
    struct EightCpuReader;
    impl MemReader for EightCpuReader {
        fn read_kva(&self, _: u64, _: usize) -> Option<Vec<u8>> {
            None
        }
        fn nr_cpu_ids(&self) -> u32 {
            8
        }
    }
    let reader = EightCpuReader;
    let v = render_value_with_mem(&btf, id, &bytes, &reader);
    match v {
        RenderedValue::CpuList { cpus } => {
            assert_eq!(
                cpus, "0-7",
                "render_struct must propagate reader.nr_cpu_ids=8 to cpu-list \
                 rendering; got: {cpus}",
            );
        }
        other => panic!("expected CpuList from cpumask render, got {other:?}"),
    }
}

/// Without a `MemReader` (the `render_value` entry point passes
/// `None`), the cpumask renderer falls back to `u32::MAX` cap
/// — every set bit surfaces. Pins the `mem.map(...).unwrap_or(u32::MAX)`
/// fallback in `render_struct` for the no-reader code path.
#[test]
fn render_value_without_mem_uses_u32_max_cap() {
    let Some(btf) = test_btf() else {
        crate::report::test_skip("test_btf returned None");
        return;
    };
    let Ok(ids) = btf.resolve_ids_by_name("cpumask") else {
        crate::report::test_skip("BTF missing 'cpumask' struct");
        return;
    };
    let Some(&id) = ids.first() else {
        crate::report::test_skip("BTF resolved 'cpumask' to empty id list");
        return;
    };
    // 8 bytes (one word) of all-FFs: every bit 0..63 set.
    // No-cap (u32::MAX) means all 64 bits surface as cpus 0-63.
    let bytes = [0xFFu8; 8];
    let v = render_value(&btf, id, &bytes);
    match v {
        RenderedValue::CpuList { cpus } => {
            assert_eq!(
                cpus, "0-63",
                "no-reader cpumask must use u32::MAX cap (all 64 bits), got: {cpus}",
            );
        }
        other => panic!("expected CpuList, got {other:?}"),
    }
}