rustfs-ecstore 1.0.1

Erasure coding storage backend for RustFS, providing efficient data storage and retrieval with redundancy.
Documentation
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
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
use crate::bucket::metadata_sys::ObjectLockConfigState;
use crate::disk::endpoint::Endpoint;
use crate::disk::format::FormatV3;
use crate::disk::{DiskAPI, DiskOption, DiskStore, WalkDirOptions, new_disk};
use crate::error::Error;
use crate::io_support::rio::HashReader;
use crate::object_api::{BLOCK_SIZE_V2, ObjectLockConfigSnapshot, ObjectOptions, PutObjReader};
use crate::set_disk::SetDisks;
use crate::storage_api_contracts::bucket::{BucketOperations as _, MakeBucketOptions};
use crate::storage_api_contracts::multipart::{CompletePart, MultipartOperations as _};
use crate::storage_api_contracts::object::{ObjectIO as _, ObjectOperations as _};
use crate::storage_api_contracts::range::HTTPRangeSpec;
use crate::store::init_format::save_format_file;
use http::HeaderMap;
use rustfs_filemeta::{MetacacheReader, MetacacheWriter};
use std::io::Cursor;
use std::path::PathBuf;
use std::sync::Arc;
use std::sync::atomic::{AtomicBool, Ordering};
use tokio::fs;
use tokio::io::AsyncReadExt;
use tokio::sync::RwLock;

/// Test-only lock client whose refresh path can be rejected independently of
/// every other lock operation. The observed event is awaitable so lock-loss
/// tests do not depend on sleeps or scheduler timing.
#[derive(Debug)]
pub(crate) struct RefreshLossLockClient {
    inner: rustfs_lock::LocalClient,
    reject_refresh: AtomicBool,
    rejected_refresh: AtomicBool,
    rejected_refresh_notify: tokio::sync::Notify,
}

impl RefreshLossLockClient {
    pub(crate) fn with_manager(manager: Arc<rustfs_lock::GlobalLockManager>) -> Self {
        Self {
            inner: rustfs_lock::LocalClient::with_manager(manager),
            reject_refresh: AtomicBool::new(false),
            rejected_refresh: AtomicBool::new(false),
            rejected_refresh_notify: tokio::sync::Notify::new(),
        }
    }

    pub(crate) fn reject_refreshes(&self) {
        self.reject_refresh.store(true, Ordering::Release);
    }

    pub(crate) fn refreshes_rejected(&self) -> bool {
        self.rejected_refresh.load(Ordering::Acquire)
    }

    pub(crate) async fn wait_for_rejected_refresh(
        &self,
        timeout: std::time::Duration,
    ) -> std::result::Result<(), tokio::time::error::Elapsed> {
        tokio::time::timeout(timeout, async {
            loop {
                let notified = self.rejected_refresh_notify.notified();
                if self.refreshes_rejected() {
                    return;
                }
                notified.await;
            }
        })
        .await
    }
}

#[async_trait::async_trait]
impl rustfs_lock::LockClient for RefreshLossLockClient {
    async fn acquire_lock(&self, request: &rustfs_lock::LockRequest) -> rustfs_lock::Result<rustfs_lock::LockResponse> {
        rustfs_lock::LockClient::acquire_lock(&self.inner, request).await
    }

    async fn release(&self, lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<bool> {
        rustfs_lock::LockClient::release(&self.inner, lock_id).await
    }

    async fn refresh(&self, lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<bool> {
        if self.reject_refresh.load(Ordering::Acquire) {
            self.rejected_refresh.store(true, Ordering::Release);
            self.rejected_refresh_notify.notify_waiters();
            return Ok(false);
        }
        rustfs_lock::LockClient::refresh(&self.inner, lock_id).await
    }

    async fn force_release(&self, lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<bool> {
        rustfs_lock::LockClient::force_release(&self.inner, lock_id).await
    }

    async fn check_status(&self, lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<Option<rustfs_lock::LockInfo>> {
        rustfs_lock::LockClient::check_status(&self.inner, lock_id).await
    }

    async fn list_lock_leases(&self) -> Vec<rustfs_lock::LockLeaseInfo> {
        rustfs_lock::LockClient::list_lock_leases(&self.inner).await
    }

    async fn get_stats(&self) -> rustfs_lock::Result<rustfs_lock::LockStats> {
        rustfs_lock::LockClient::get_stats(&self.inner).await
    }

    async fn close(&self) -> rustfs_lock::Result<()> {
        rustfs_lock::LockClient::close(&self.inner).await
    }

    async fn is_online(&self) -> bool {
        rustfs_lock::LockClient::is_online(&self.inner).await
    }

    async fn is_local(&self) -> bool {
        rustfs_lock::LockClient::is_local(&self.inner).await
    }
}

#[tokio::test]
async fn refresh_loss_lock_client_keeps_rejection_observable_for_late_waiters() {
    let manager = Arc::new(rustfs_lock::GlobalLockManager::Enabled(Arc::new(
        rustfs_lock::FastObjectLockManager::new(),
    )));
    let client = RefreshLossLockClient::with_manager(manager);
    let resource = rustfs_lock::ObjectKey::new("bucket", "object");
    let response = rustfs_lock::LockClient::acquire_lock(
        &client,
        &rustfs_lock::LockRequest::new(resource, rustfs_lock::LockType::Shared, "refresh-loss-harness"),
    )
    .await
    .expect("acquire should reach the inner local client");
    let lock_id = response.lock_info.expect("the inner local client should acquire the lock").id;
    assert_eq!(
        rustfs_lock::LockClient::list_lock_leases(&client).await.len(),
        1,
        "lease diagnostics must remain transparent through the refresh wrapper"
    );

    client.reject_refreshes();
    assert!(
        !rustfs_lock::LockClient::refresh(&client, &lock_id)
            .await
            .expect("refresh should return a response")
    );
    client
        .wait_for_rejected_refresh(std::time::Duration::from_millis(50))
        .await
        .expect("a waiter registered after rejection must still observe the event");
    assert!(client.refreshes_rejected());
    assert!(
        rustfs_lock::LockClient::release(&client, &lock_id)
            .await
            .expect("release should reach the inner local client")
    );
}

/// Returns the backing [`tempfile::TempDir`]s alongside the set so callers keep
/// them alive for the test's duration and the directories are removed on drop.
pub(crate) async fn make_local_set_disks(drive_count: usize, parity_count: usize) -> (Vec<tempfile::TempDir>, Arc<SetDisks>) {
    make_local_set_disks_with_ctx(drive_count, parity_count, crate::runtime::instance::bootstrap_ctx()).await
}

/// Like [`make_local_set_disks`], but binds the set to an explicit instance
/// context. Tests whose assertions depend on ambient runtime state that other
/// tests mutate (e.g. `SetupTypeGuard` flipping the shared setup type to
/// DistErasure from `#[serial]` tests in a different serial group) pass an
/// isolated context here so that state cannot leak into their code path.
pub(crate) async fn make_local_set_disks_with_ctx(
    drive_count: usize,
    parity_count: usize,
    ctx: Arc<crate::runtime::instance::InstanceContext>,
) -> (Vec<tempfile::TempDir>, Arc<SetDisks>) {
    let format = FormatV3::new(1, drive_count);
    let mut dirs = Vec::with_capacity(drive_count);
    let mut endpoints = Vec::with_capacity(drive_count);
    let mut disks = Vec::with_capacity(drive_count);

    for disk_idx in 0..drive_count {
        let dir = tempfile::tempdir().expect("tempdir should be created");
        let mut endpoint =
            Endpoint::try_from(dir.path().to_str().expect("tempdir path should be utf8")).expect("endpoint should parse");
        endpoint.set_pool_index(0);
        endpoint.set_set_index(0);
        endpoint.set_disk_index(disk_idx);

        let disk = new_disk(
            &endpoint,
            &DiskOption {
                cleanup: false,
                health_check: false,
            },
        )
        .await
        .expect("disk should be created");

        let mut disk_format = format.clone();
        disk_format.erasure.this = format.erasure.sets[0][disk_idx];
        save_format_file(&Some(disk.clone()), &Some(disk_format))
            .await
            .expect("format should be saved");

        dirs.push(dir);
        endpoints.push(endpoint);
        disks.push(Some(disk));
    }

    let set_disks = SetDisks::new_with_instance_ctx(
        "ecstore-validation-blackbox".to_string(),
        Arc::new(RwLock::new(disks)),
        drive_count,
        parity_count,
        0,
        0,
        endpoints,
        format,
        Vec::new(),
        ctx,
    )
    .await;

    (dirs, set_disks)
}

async fn first_shard_part_path(disk: &DiskStore, bucket: &str, object: &str) -> PathBuf {
    let object_dir = disk.path().join(bucket).join(object);
    let mut entries = fs::read_dir(&object_dir)
        .await
        .unwrap_or_else(|err| panic!("object data dir should be readable at {object_dir:?}: {err}"));
    while let Some(entry) = entries.next_entry().await.expect("object data dir entry should be readable") {
        if entry
            .file_type()
            .await
            .expect("object data entry type should be readable")
            .is_dir()
        {
            let part_path = entry.path().join("part.1");
            if fs::metadata(&part_path).await.is_ok() {
                return part_path;
            }
        }
    }
    panic!("part.1 shard should exist under {object_dir:?}");
}

async fn shard_part_paths(set_disks: &Arc<SetDisks>, bucket: &str, object: &str) -> Vec<PathBuf> {
    let disks = set_disks.disks.read().await;
    let mut paths = Vec::new();
    for disk in disks.iter().flatten() {
        paths.push(first_shard_part_path(disk, bucket, object).await);
    }
    paths
}

#[tokio::test]
async fn blackbox_put_unknown_actual_size_restores_body_and_records_written_size() {
    let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
    let bucket = "bb-unknown-actual-size";
    let object = "object.bin";
    let payload = (0..(BLOCK_SIZE_V2 + 123))
        .map(|idx| ((idx * 17) % 251) as u8)
        .collect::<Vec<_>>();
    let opts = ObjectOptions {
        no_lock: true,
        ..Default::default()
    };

    set_disks
        .make_bucket(bucket, &MakeBucketOptions::default())
        .await
        .expect("bucket should be created");
    let stream = HashReader::from_stream(Cursor::new(payload.clone()), payload.len() as i64, -1, None, None, false)
        .expect("hash reader should accept unknown actual size");
    let mut reader = PutObjReader::new(stream);
    let written = set_disks
        .put_object(bucket, object, &mut reader, &opts)
        .await
        .expect("object should be written");

    assert_eq!(written.size, payload.len() as i64);
    assert_eq!(written.parts[0].actual_size, payload.len() as i64);

    let mut get_reader = set_disks
        .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
        .await
        .expect("object reader should open");
    let mut restored = Vec::new();
    get_reader
        .stream
        .read_to_end(&mut restored)
        .await
        .expect("object should stream");
    assert_eq!(restored, payload);
}

#[tokio::test]
async fn blackbox_get_restores_body_after_one_shard_file_is_removed() {
    let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
    let bucket = "bb-missing-shard-file";
    let object = "object.bin";
    let payload = (0..(BLOCK_SIZE_V2 + 321))
        .map(|idx| ((idx * 19) % 251) as u8)
        .collect::<Vec<_>>();
    let opts = ObjectOptions {
        no_lock: true,
        ..Default::default()
    };

    set_disks
        .make_bucket(bucket, &MakeBucketOptions::default())
        .await
        .expect("bucket should be created");
    let mut reader = PutObjReader::from_vec(payload.clone());
    set_disks
        .put_object(bucket, object, &mut reader, &opts)
        .await
        .expect("object should be written");

    let disk = {
        let disks = set_disks.disks.read().await;
        disks[0].clone().expect("first disk should exist")
    };
    let shard_path = first_shard_part_path(&disk, bucket, object).await;
    fs::remove_file(&shard_path)
        .await
        .unwrap_or_else(|err| panic!("test shard should be removable at {shard_path:?}: {err}"));

    let mut get_reader = set_disks
        .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
        .await
        .expect("object should remain readable after one shard file is removed");
    let mut restored = Vec::new();
    get_reader
        .stream
        .read_to_end(&mut restored)
        .await
        .expect("degraded object should stream");

    assert_eq!(restored, payload);
}

#[tokio::test]
// Serialized: forces the reader-setup strategy through a process-global env var.
#[serial_test::serial]
async fn blackbox_heal_requests_preserve_repair_scope() {
    use rustfs_heal_contracts::heal_channel::{
        HealAdmissionResult, HealChannelCommand, HealChannelPriority, HealChannelReceiver, HealChannelRequest, HealRequestSource,
    };

    async fn receive_matching_heal(rx: &mut HealChannelReceiver, bucket: &str, object: &str) -> HealChannelRequest {
        tokio::time::timeout(std::time::Duration::from_secs(30), async {
            loop {
                match rx.recv().await.expect("heal channel should stay open") {
                    HealChannelCommand::Start { request, response_tx }
                        if request.bucket == bucket && request.object_prefix.as_deref() == Some(object) =>
                    {
                        let _ = response_tx.send(Ok(HealAdmissionResult::Accepted));
                        break request;
                    }
                    _ => continue,
                }
            }
        })
        .await
        .expect("matching heal request should be submitted")
    }

    // Own the process-global heal channel so the read path's repair submission
    // becomes observable. init_heal_channel() succeeds exactly once per test
    // binary: this must stay the only ecstore unit test that takes the receiver.
    // Submissions from other (non-serial) tests queue in the unbounded channel
    // while this test is setting up, get drained and dropped by the loop below
    // (failing their submitter, which releases their dedup reservation), and
    // fail fast once the receiver drops at test end. Tests that must observe a
    // deterministic channel state serialize under the same serial key.
    let mut heal_rx = rustfs_heal_contracts::heal_channel::init_heal_channel()
        .expect("this must be the only ecstore test that owns the heal channel receiver");

    // Without a durable MRF consumer, partial PUTs fall back to the heal
    // admission channel and wait for its receipt before acknowledging the write.
    // Drive the test receiver alongside the PUT so neither waits on the other.
    let (_put_dirs, put_store) = crate::bucket::metadata_sys::test_support::isolated_store_over_temp_disks().await;
    crate::bucket::metadata_sys::init_bucket_metadata_sys(Arc::clone(&put_store), Vec::new()).await;
    let put_set = Arc::clone(&put_store.pools[0].disk_set[0]);
    assert_eq!(put_set.set_drive_count, 4);
    assert_eq!(put_set.default_parity_count, 2);
    let put_bucket = "bb-put-partial-convergence";
    let put_object = "object.bin";
    put_store
        .make_bucket(put_bucket, &MakeBucketOptions::default())
        .await
        .expect("PUT bucket should be created");
    let put_incarnation = put_store
        .bucket_incarnation_id_from_disk(put_bucket)
        .await
        .expect("partial repair must bind the real persisted bucket generation");
    let offline_disk = {
        let mut disks = put_set.disks.write().await;
        disks[0].take()
    };
    let mut put_reader = PutObjReader::from_vec(vec![0x42; BLOCK_SIZE_V2 + 1024]);
    let (committed, request) = tokio::time::timeout(std::time::Duration::from_secs(30), async {
        tokio::join!(
            async {
                put_set
                    .put_object(
                        put_bucket,
                        put_object,
                        &mut put_reader,
                        &ObjectOptions {
                            no_lock: true,
                            versioned: true,
                            expected_bucket_incarnation_id: Some(put_incarnation),
                            ..Default::default()
                        },
                    )
                    .await
                    .expect("partial ordinary PUT should succeed at write quorum")
            },
            receive_matching_heal(&mut heal_rx, put_bucket, put_object),
        )
    })
    .await
    .expect("partial ordinary PUT and repair admission should complete together");
    let committed_version = committed
        .version_id
        .expect("versioned PUT should return a version id")
        .to_string();

    assert_eq!(request.object_version_id.as_deref(), Some(committed_version.as_str()));
    assert_eq!(request.pool_index, Some(0));
    assert_eq!(request.set_index, Some(0));
    assert_eq!(request.source, HealRequestSource::Mrf);
    assert_eq!(request.expected_bucket_incarnation_id, Some(put_incarnation));

    let duplicate_request = tokio::time::timeout(std::time::Duration::from_millis(100), async {
        loop {
            match heal_rx.recv().await.expect("heal channel should stay open") {
                HealChannelCommand::Start { request, response_tx }
                    if request.bucket == put_bucket && request.object_prefix.as_deref() == Some(put_object) =>
                {
                    let _ = response_tx.send(Ok(HealAdmissionResult::Accepted));
                    break Some(request);
                }
                HealChannelCommand::Start { response_tx, .. } => {
                    let _ = response_tx.send(Ok(HealAdmissionResult::Accepted));
                }
                _ => {}
            }
        }
    })
    .await
    .ok()
    .flatten();
    assert!(duplicate_request.is_none(), "partial ordinary PUT must enqueue exactly one heal request");

    {
        let mut disks = put_set.disks.write().await;
        disks[0] = offline_disk;
    }

    let healthy_bucket = "bb-put-healthy-convergence";
    put_store
        .make_bucket(healthy_bucket, &MakeBucketOptions::default())
        .await
        .expect("healthy PUT bucket should be created");
    let mut healthy_reader = PutObjReader::from_vec(b"healthy".to_vec());
    put_set
        .put_object(
            healthy_bucket,
            put_object,
            &mut healthy_reader,
            &ObjectOptions {
                no_lock: true,
                ..Default::default()
            },
        )
        .await
        .expect("healthy ordinary PUT should succeed");
    let healthy_request = tokio::time::timeout(std::time::Duration::from_millis(100), async {
        loop {
            match heal_rx.recv().await.expect("heal channel should stay open") {
                HealChannelCommand::Start { request, response_tx }
                    if request.bucket == healthy_bucket && request.object_prefix.as_deref() == Some(put_object) =>
                {
                    let _ = response_tx.send(Ok(HealAdmissionResult::Accepted));
                    break Some(request);
                }
                HealChannelCommand::Start { response_tx, .. } => {
                    let _ = response_tx.send(Ok(HealAdmissionResult::Accepted));
                }
                _ => {}
            }
        }
    })
    .await
    .ok()
    .flatten();
    assert!(healthy_request.is_none(), "fully converged ordinary PUT must not enqueue heal");

    // Keep data-blocks-first reader setup explicit for this deterministic
    // repair assertion (see ENV_RUSTFS_GET_DATA_BLOCKS_FIRST_READER_SETUP in
    // set_disk/core/io_primitives.rs): if a caller opts back into all-shards,
    // the corrupt shard's failed open can lose the setup-quorum race, making
    // the enqueue assertion below a coin flip.
    temp_env::async_with_vars([("RUSTFS_GET_DATA_BLOCKS_FIRST_READER_SETUP", Some("true"))], async {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "bb-corrupt-shard-repair";
        let object = "object.bin";
        let payload = (0..(BLOCK_SIZE_V2 + 777))
            .map(|idx| ((idx * 29) % 251) as u8)
            .collect::<Vec<_>>();
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");
        let mut reader = PutObjReader::from_vec(payload.clone());
        set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("object should be written");

        // Corrupt the disk holding DATA shard 1, not blindly disk 0 (which holds
        // a parity shard for this key): a data-blocks-first read must consume the
        // corrupt data shard, making the missing-shard detection deterministic.
        // The write path derives the shard distribution the same way (ops/object.rs).
        let distribution = rustfs_filemeta::FileInfo::new(&format!("{bucket}/{object}"), 2, 2)
            .erasure
            .distribution;
        let corrupt_disk = distribution
            .iter()
            .position(|&shard| shard == 1)
            .expect("distribution should place data shard 1 on a disk");
        let paths = shard_part_paths(&set_disks, bucket, object).await;
        fs::write(&paths[corrupt_disk], b"corrupt shard bytes")
            .await
            .unwrap_or_else(|err| panic!("test shard should be corruptible at {:?}: {err}", paths[corrupt_disk]));

        let mut get_reader = set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
            .expect("object should remain readable after one corrupt shard");
        let mut restored = Vec::new();
        get_reader
            .stream
            .read_to_end(&mut restored)
            .await
            .expect("corrupt-shard object should stream from parity");

        assert_eq!(restored, payload);

        let request = receive_matching_heal(&mut heal_rx, bucket, object).await;

        assert_eq!(request.source, HealRequestSource::ReadRepair);
        assert_eq!(request.object_prefix.as_deref(), Some(object));
        assert_eq!(request.object_version_id, None, "unversioned PUT must submit repair without a version id");
        assert_eq!(request.pool_index, Some(0));
        assert_eq!(request.set_index, Some(0));
        assert_eq!(request.priority, HealChannelPriority::Low);
        assert_eq!(request.recreate_missing, Some(true));

        let mpu_bucket = "bb-mpu-convergence-heal";
        let partial_object = "partial.bin";
        let suspended_object = "suspended.bin";
        let payload = vec![0x5a; 1 << 20];
        let mpu_opts = ObjectOptions {
            no_lock: true,
            versioned: true,
            ..Default::default()
        };

        set_disks
            .make_bucket(mpu_bucket, &MakeBucketOptions::default())
            .await
            .expect("multipart bucket should be created");

        let stage_upload = async |object: &str| {
            let upload = set_disks
                .new_multipart_upload(mpu_bucket, object, &mpu_opts)
                .await
                .expect("multipart upload should be created");
            let mut reader = PutObjReader::new(
                HashReader::from_stream(
                    Cursor::new(payload.clone()),
                    payload.len() as i64,
                    payload.len() as i64,
                    None,
                    None,
                    false,
                )
                .expect("multipart reader should be constructed"),
            );
            let part = set_disks
                .put_object_part(mpu_bucket, object, &upload.upload_id, 1, &mut reader, &mpu_opts)
                .await
                .expect("multipart part should be written");
            (
                upload.upload_id,
                vec![CompletePart {
                    part_num: part.part_num,
                    etag: part.etag,
                    ..Default::default()
                }],
            )
        };

        let (partial_upload_id, partial_parts) = stage_upload(partial_object).await;
        let offline_disk = {
            let mut disks = set_disks.disks.write().await;
            disks[3].take().expect("fourth disk should be online before completion")
        };
        crate::crash_inject::arm(crate::crash_inject::CrashPoint::MultipartAfterCommitBeforePartsCleanup, partial_object);
        let completed = set_disks
            .clone()
            .complete_multipart_upload(mpu_bucket, partial_object, &partial_upload_id, partial_parts, &mpu_opts)
            .await;
        assert!(
            matches!(completed, Err(Error::Unexpected)),
            "partial multipart completion should reach the post-commit crash point, got {completed:?}"
        );
        crate::crash_inject::disarm(crate::crash_inject::CrashPoint::MultipartAfterCommitBeforePartsCleanup, partial_object);

        let request = receive_matching_heal(&mut heal_rx, mpu_bucket, partial_object).await;

        let completed_version_id = request
            .object_version_id
            .clone()
            .expect("versioned multipart convergence heal must bind a version id");
        assert_eq!(request.pool_index, Some(0));
        assert_eq!(request.set_index, Some(0));
        assert_eq!(request.priority, HealChannelPriority::Normal);

        let duplicate = tokio::time::timeout(std::time::Duration::from_millis(250), async {
            loop {
                match heal_rx.recv().await.expect("heal channel should stay open") {
                    HealChannelCommand::Start { request, .. }
                        if request.bucket == mpu_bucket && request.object_prefix.as_deref() == Some(partial_object) =>
                    {
                        break request;
                    }
                    _ => continue,
                }
            }
        })
        .await;
        assert!(duplicate.is_err(), "partial multipart completion must enqueue exactly one heal request");

        {
            let mut disks = set_disks.disks.write().await;
            disks[3] = Some(offline_disk);
        }
        let committed = set_disks
            .get_object_info(
                mpu_bucket,
                partial_object,
                &ObjectOptions {
                    no_lock: true,
                    versioned: true,
                    version_id: Some(completed_version_id.clone()),
                    ..Default::default()
                },
            )
            .await
            .expect("heal-bound multipart version should be committed and addressable");
        assert_eq!(committed.version_id.map(|version_id| version_id.to_string()), Some(completed_version_id));

        let (suspended_upload_id, suspended_parts) = stage_upload(suspended_object).await;
        let offline_disk = {
            let mut disks = set_disks.disks.write().await;
            disks[3]
                .take()
                .expect("fourth disk should be online before suspended completion")
        };
        let suspended_opts = ObjectOptions {
            no_lock: true,
            version_suspended: true,
            ..Default::default()
        };
        let suspended = set_disks
            .clone()
            .complete_multipart_upload(mpu_bucket, suspended_object, &suspended_upload_id, suspended_parts, &suspended_opts)
            .await
            .expect("suspended multipart completion should succeed at write quorum");
        {
            let mut disks = set_disks.disks.write().await;
            disks[3] = Some(offline_disk);
        }
        assert!(
            suspended.version_id.is_some_and(|version_id| version_id.is_nil()),
            "suspended multipart completion should publish the null version"
        );

        let request = receive_matching_heal(&mut heal_rx, mpu_bucket, suspended_object).await;

        let null_version_id = uuid::Uuid::nil().to_string();
        assert_eq!(request.object_version_id.as_deref(), Some(null_version_id.as_str()));
    })
    .await;
}

#[tokio::test]
async fn blackbox_range_read_restores_exact_slice_with_one_offline_disk() {
    let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
    let bucket = "bb-range-offline-disk";
    let object = "object.bin";
    let payload = (0..(BLOCK_SIZE_V2 + 4096))
        .map(|idx| ((idx * 23) % 251) as u8)
        .collect::<Vec<_>>();
    let range_start = 513usize;
    let range_len = 8192usize;
    let opts = ObjectOptions {
        no_lock: true,
        ..Default::default()
    };

    set_disks
        .make_bucket(bucket, &MakeBucketOptions::default())
        .await
        .expect("bucket should be created");
    let mut reader = PutObjReader::from_vec(payload.clone());
    set_disks
        .put_object(bucket, object, &mut reader, &opts)
        .await
        .expect("object should be written");

    {
        let mut disks = set_disks.disks.write().await;
        disks[2] = None;
    }

    let range = HTTPRangeSpec {
        is_suffix_length: false,
        start: range_start as i64,
        end: (range_start + range_len - 1) as i64,
    };
    let mut get_reader = set_disks
        .get_object_reader(bucket, object, Some(range), HeaderMap::new(), &opts)
        .await
        .expect("range reader should open with one offline disk");
    let mut restored = Vec::new();
    get_reader
        .stream
        .read_to_end(&mut restored)
        .await
        .expect("range body should stream");

    assert_eq!(restored, payload[range_start..range_start + range_len]);
}

#[tokio::test]
async fn blackbox_delete_marker_hides_object_body_without_erasing_prior_version_metadata() {
    let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
    let bucket = "bb-delete-marker-read-negative";
    let object = "object.bin";
    let opts = ObjectOptions {
        no_lock: true,
        version_suspended: true,
        object_lock_config_snapshot: Some(Arc::new(ObjectLockConfigSnapshot::new(ObjectLockConfigState::ConfirmedAbsent))),
        ..Default::default()
    };

    set_disks
        .make_bucket(bucket, &MakeBucketOptions::default())
        .await
        .expect("bucket should be created");
    let mut reader = PutObjReader::from_vec(b"body hidden by delete marker".to_vec());
    set_disks
        .put_object(bucket, object, &mut reader, &opts)
        .await
        .expect("object should be written before delete marker");

    let marker = set_disks
        .delete_object(bucket, object, opts.clone())
        .await
        .expect("version suspended delete should create a marker");
    assert!(marker.delete_marker);

    let err = match set_disks
        .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
        .await
    {
        Ok(_) => panic!("delete marker must hide the object body"),
        Err(err) => err,
    };
    assert!(
        matches!(&err, Error::ObjectNotFound(b, o) if b == bucket && o == object),
        "latest delete marker read must map to ObjectNotFound for the requested key, got {err:?}"
    );
}

#[tokio::test]
async fn blackbox_local_disk_walk_dir_emits_metadata_entries_with_prefix_forward_and_limit() {
    let dir = tempfile::tempdir().expect("tempdir should be created");
    let bucket = "bb-local-walk";
    let endpoint = Endpoint::try_from(dir.path().to_str().expect("tempdir path should be utf8")).expect("endpoint should parse");
    let disk = new_disk(
        &endpoint,
        &DiskOption {
            cleanup: false,
            health_check: false,
        },
    )
    .await
    .expect("disk should be created");
    disk.make_volume(bucket).await.expect("bucket volume should be created");

    for object in ["prefix/a", "prefix/b", "prefix/c", "other/d"] {
        disk.write_all(bucket, &format!("{object}/xl.meta"), bytes::Bytes::from_static(b"meta"))
            .await
            .expect("metadata object should be written");
    }

    let (reader, mut writer) = tokio::io::duplex(4096);
    let opts = WalkDirOptions {
        bucket: bucket.to_string(),
        base_dir: "prefix/".to_string(),
        recursive: true,
        forward_to: Some("prefix/b".to_string()),
        limit: 2,
        ..Default::default()
    };

    disk.walk_dir(opts, &mut writer)
        .await
        .expect("walk_dir should stream metadata entries");
    MetacacheWriter::new(&mut writer)
        .close()
        .await
        .expect("metacache stream should close");
    drop(writer);

    let mut reader = MetacacheReader::new(reader);
    let entries = reader.read_all().await.expect("metacache stream should decode");
    let names = entries
        .into_iter()
        .filter(|entry| !entry.metadata.is_empty())
        .map(|entry| entry.name)
        .collect::<Vec<_>>();

    assert_eq!(names, vec!["prefix/b".to_string(), "prefix/c".to_string()]);
}

#[tokio::test]
#[serial_test::serial]
async fn blackbox_issue3031_diag_covers_put_success_cleanup_and_error_summary() {
    temp_env::async_with_vars([("RUSTFS_ISSUE3031_DIAG_ENABLE", Some("true"))], async {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "bb-issue3031-diag";
        let object = "object.bin";
        let first_payload = b"first diagnostic body".to_vec();
        let second_payload = b"second diagnostic body that replaces the first".to_vec();
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");

        let mut first_reader = PutObjReader::from_vec(first_payload);
        set_disks
            .put_object(bucket, object, &mut first_reader, &opts)
            .await
            .expect("first diagnostic write should succeed");

        let mut second_reader = PutObjReader::from_vec(second_payload.clone());
        set_disks
            .put_object(bucket, object, &mut second_reader, &opts)
            .await
            .expect("overwrite should succeed and report cleanup-present diagnostics");

        let mut get_reader = set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
            .expect("overwritten object reader should open");
        let mut restored = Vec::new();
        get_reader
            .stream
            .read_to_end(&mut restored)
            .await
            .expect("overwritten object should stream");
        assert_eq!(restored, second_payload);

        let (_failed_dirs, failed_set) = make_local_set_disks(1, 0).await;
        let failed_bucket = "bb-issue3031-fail";
        failed_set
            .make_bucket(failed_bucket, &MakeBucketOptions::default())
            .await
            .expect("failure bucket should be created");
        {
            let mut disks = failed_set.disks.write().await;
            disks[0] = None;
        }
        let mut failed_reader = PutObjReader::from_vec(b"must fail closed".to_vec());
        let err = failed_set
            .put_object(failed_bucket, "object.bin", &mut failed_reader, &opts)
            .await
            .expect_err("offline only disk must fail writer setup under diagnostics");
        assert!(
            matches!(err, Error::InsufficientWriteQuorum(_, _) | Error::ErasureWriteQuorum),
            "diagnostic failure should still report a write-quorum style error, got {err:?}"
        );
    })
    .await;
}

/// rustfs/backlog#1009: the `put_object` old-size backfill must reproduce, for
/// every PUT shape, exactly what the app layer's pre-PUT `get_object_info`
/// lookup would have observed — that is the invariant that lets the app skip
/// the lookup without changing a single usage-accounting number.
mod old_current_size_backfill {
    use super::*;
    use crate::disk::OldCurrentSize;
    use crate::error::is_err_object_not_found;
    use crate::set_disk::SetDisks;

    /// What the pre-PUT lookup would report right now for `object`.
    async fn prelookup_expectation(set_disks: &Arc<SetDisks>, bucket: &str, object: &str) -> OldCurrentSize {
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };
        match set_disks.get_object_info(bucket, object, &opts).await {
            Ok(object_info) => OldCurrentSize::Present(object_info.size),
            Err(err) => {
                assert!(is_err_object_not_found(&err), "prelookup expectation hit unexpected error: {err:?}");
                OldCurrentSize::Absent
            }
        }
    }

    async fn put_and_backfill(
        set_disks: &Arc<SetDisks>,
        bucket: &str,
        object: &str,
        len: usize,
        opts: &ObjectOptions,
    ) -> Option<OldCurrentSize> {
        let payload = (0..len).map(|idx| ((idx * 31) % 251) as u8).collect::<Vec<_>>();
        let mut reader = PutObjReader::from_vec(payload);
        let (_object_info, backfill) = set_disks
            .put_object_with_old_current_size(bucket, object, &mut reader, opts)
            .await
            .expect("put_object should succeed");
        backfill
    }

    #[tokio::test]
    async fn backfill_matches_prelookup_for_every_put_shape() {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "bb-old-current-size";
        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

        // Fresh key (inline-sized): both the lookup and the backfill say Absent.
        let object = "object.bin";
        assert_eq!(prelookup_expectation(&set_disks, bucket, object).await, OldCurrentSize::Absent);
        assert_eq!(
            put_and_backfill(&set_disks, bucket, object, 1024, &opts).await,
            Some(OldCurrentSize::Absent)
        );

        // Unversioned inline overwrite: the previous live size must surface.
        let expected = prelookup_expectation(&set_disks, bucket, object).await;
        assert_eq!(expected, OldCurrentSize::Present(1024));
        assert_eq!(put_and_backfill(&set_disks, bucket, object, 2048, &opts).await, Some(expected));

        // Non-inline overwrite (payload above the inline block threshold).
        let expected = prelookup_expectation(&set_disks, bucket, object).await;
        assert_eq!(expected, OldCurrentSize::Present(2048));
        assert_eq!(
            put_and_backfill(&set_disks, bucket, object, BLOCK_SIZE_V2 + 123, &opts).await,
            Some(expected)
        );

        // Overwriting a non-inline object reports its size back.
        let expected = prelookup_expectation(&set_disks, bucket, object).await;
        assert_eq!(expected, OldCurrentSize::Present((BLOCK_SIZE_V2 + 123) as i64));
        assert_eq!(put_and_backfill(&set_disks, bucket, object, 64, &opts).await, Some(expected));

        // Versioned writes: a new version over a live latest reports the
        // latest's size, not the incoming version's.
        let versioned = "versioned.bin";
        let first_version_opts = ObjectOptions {
            no_lock: true,
            versioned: true,
            version_id: Some(uuid::Uuid::new_v4().to_string()),
            ..Default::default()
        };
        assert_eq!(
            put_and_backfill(&set_disks, bucket, versioned, 111, &first_version_opts).await,
            Some(OldCurrentSize::Absent)
        );
        let expected = prelookup_expectation(&set_disks, bucket, versioned).await;
        assert_eq!(expected, OldCurrentSize::Present(111));
        let second_version_opts = ObjectOptions {
            no_lock: true,
            versioned: true,
            version_id: Some(uuid::Uuid::new_v4().to_string()),
            ..Default::default()
        };
        assert_eq!(
            put_and_backfill(&set_disks, bucket, versioned, 222, &second_version_opts).await,
            Some(expected)
        );

        // Delete-marker latest: today's lookup returns the marker as
        // Ok(size 0) — NOT not-found — and delete-marker creation never
        // decremented objects_count, so the backfill must reproduce
        // Present(0) to keep versioned accounting identical.
        set_disks
            .delete_object(
                bucket,
                versioned,
                ObjectOptions {
                    no_lock: true,
                    versioned: true,
                    object_lock_config_snapshot: Some(Arc::new(ObjectLockConfigSnapshot::new(
                        ObjectLockConfigState::ConfirmedAbsent,
                    ))),
                    ..Default::default()
                },
            )
            .await
            .expect("versioned delete should create a delete marker");
        let expected = prelookup_expectation(&set_disks, bucket, versioned).await;
        assert_eq!(expected, OldCurrentSize::Present(0));
        let third_version_opts = ObjectOptions {
            no_lock: true,
            versioned: true,
            version_id: Some(uuid::Uuid::new_v4().to_string()),
            ..Default::default()
        };
        assert_eq!(
            put_and_backfill(&set_disks, bucket, versioned, 333, &third_version_opts).await,
            Some(expected)
        );
    }

    /// Sub-quorum divergence through the real set-level fanout: when only 2 of
    /// 4 disks can still decode the destination's xl.meta (write quorum is 3),
    /// the PUT must still commit but the backfill must come back unknown —
    /// never a fabricated `Absent`/`Present` from a below-quorum vote.
    #[tokio::test]
    async fn backfill_is_unknown_below_quorum_but_put_succeeds() {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "bb-old-current-subquorum";
        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

        let object = "object.bin";
        assert_eq!(
            put_and_backfill(&set_disks, bucket, object, 1024, &opts).await,
            Some(OldCurrentSize::Absent)
        );

        // Corrupt the committed xl.meta on two disks: those observations turn
        // unknown, leaving 2 definite votes < write quorum 3.
        {
            let disks = set_disks.disks.read().await;
            for disk in disks.iter().flatten().take(2) {
                let meta_path = disk.path().join(bucket).join(object).join("xl.meta");
                fs::write(&meta_path, b"not-an-xl-meta")
                    .await
                    .expect("xl.meta should be overwritable with garbage");
            }
        }

        let backfill = put_and_backfill(&set_disks, bucket, object, 2048, &opts).await;
        assert_eq!(backfill, None, "a below-quorum vote must surface as unknown");
    }
}