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
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174
6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414
6415
6416
6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444
6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458
6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472
6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564
6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
6602
6603
6604
6605
6606
6607
6608
6609
6610
6611
6612
6613
6614
6615
6616
6617
6618
6619
6620
6621
6622
6623
6624
6625
6626
6627
6628
6629
6630
6631
6632
6633
6634
6635
6636
6637
6638
6639
6640
6641
6642
6643
6644
6645
6646
6647
6648
6649
6650
6651
6652
6653
6654
6655
6656
6657
6658
6659
6660
6661
6662
6663
6664
6665
6666
6667
6668
6669
6670
6671
6672
6673
6674
6675
6676
6677
6678
6679
6680
6681
6682
6683
6684
6685
6686
6687
6688
6689
6690
6691
6692
6693
6694
6695
6696
6697
6698
6699
6700
6701
6702
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
6728
6729
6730
6731
6732
6733
6734
6735
6736
6737
6738
6739
6740
6741
6742
6743
6744
6745
6746
6747
6748
6749
6750
6751
6752
6753
6754
6755
6756
6757
6758
6759
6760
6761
6762
6763
6764
6765
6766
6767
6768
6769
6770
6771
6772
6773
6774
6775
6776
6777
6778
6779
6780
6781
6782
6783
6784
6785
6786
6787
6788
6789
6790
6791
6792
6793
6794
6795
6796
6797
6798
6799
6800
6801
6802
6803
6804
6805
6806
6807
6808
6809
6810
6811
6812
6813
6814
6815
6816
6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
6833
6834
6835
6836
6837
6838
6839
6840
6841
6842
6843
6844
6845
6846
6847
6848
6849
6850
6851
6852
6853
6854
6855
6856
6857
6858
6859
6860
6861
6862
6863
6864
6865
6866
6867
6868
6869
6870
6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882
6883
6884
6885
6886
6887
6888
6889
6890
6891
6892
6893
6894
6895
6896
6897
6898
6899
6900
6901
6902
6903
6904
6905
6906
6907
6908
6909
6910
6911
6912
6913
6914
6915
6916
6917
6918
6919
6920
6921
6922
6923
6924
6925
6926
6927
6928
6929
6930
6931
6932
6933
6934
6935
6936
6937
6938
6939
6940
6941
6942
6943
6944
6945
6946
6947
6948
6949
6950
6951
6952
6953
6954
6955
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
6971
6972
6973
6974
6975
6976
6977
6978
6979
6980
6981
6982
6983
6984
6985
6986
6987
6988
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
7004
7005
7006
7007
7008
7009
7010
7011
7012
7013
7014
7015
7016
7017
7018
7019
7020
7021
7022
7023
7024
7025
7026
7027
7028
7029
7030
7031
7032
7033
7034
7035
7036
7037
7038
7039
7040
7041
7042
7043
7044
7045
7046
7047
7048
7049
7050
7051
7052
7053
7054
7055
7056
7057
7058
7059
7060
7061
7062
7063
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
7077
7078
7079
7080
7081
7082
7083
7084
7085
7086
7087
7088
7089
7090
7091
7092
7093
7094
7095
7096
7097
7098
7099
7100
7101
7102
7103
7104
7105
7106
7107
7108
7109
7110
7111
7112
7113
7114
7115
7116
7117
7118
7119
7120
7121
7122
7123
7124
7125
7126
7127
7128
7129
7130
7131
7132
7133
7134
7135
7136
7137
7138
7139
7140
7141
7142
7143
7144
7145
7146
7147
7148
7149
7150
7151
7152
7153
7154
7155
7156
7157
7158
7159
7160
7161
7162
7163
7164
7165
7166
7167
7168
7169
7170
7171
7172
7173
7174
7175
7176
7177
7178
7179
7180
7181
7182
7183
7184
7185
7186
7187
7188
7189
7190
7191
7192
7193
7194
7195
7196
7197
7198
7199
7200
7201
7202
7203
7204
7205
7206
7207
7208
7209
7210
7211
7212
7213
7214
7215
7216
7217
7218
7219
7220
7221
7222
7223
7224
7225
7226
7227
7228
7229
7230
7231
7232
7233
7234
7235
7236
7237
7238
7239
7240
7241
7242
7243
7244
7245
7246
7247
7248
7249
7250
7251
7252
7253
7254
7255
7256
7257
7258
7259
7260
7261
7262
7263
7264
7265
7266
7267
7268
7269
7270
7271
7272
7273
7274
7275
7276
7277
7278
7279
7280
7281
7282
7283
7284
7285
7286
7287
7288
7289
7290
7291
7292
7293
7294
7295
7296
7297
7298
7299
7300
7301
7302
7303
7304
7305
7306
7307
7308
7309
7310
7311
7312
7313
7314
7315
7316
7317
7318
7319
7320
7321
7322
7323
7324
7325
7326
7327
7328
7329
7330
7331
7332
7333
7334
7335
7336
7337
7338
7339
7340
7341
7342
7343
7344
7345
7346
7347
7348
7349
7350
7351
7352
7353
7354
7355
7356
7357
7358
7359
7360
7361
7362
7363
7364
7365
7366
7367
7368
7369
7370
7371
7372
7373
7374
7375
7376
7377
7378
7379
7380
7381
7382
7383
7384
7385
7386
7387
7388
7389
7390
7391
7392
7393
7394
7395
7396
7397
7398
7399
7400
7401
7402
7403
7404
7405
7406
7407
7408
7409
7410
7411
7412
7413
7414
7415
7416
7417
7418
7419
7420
7421
7422
7423
7424
7425
7426
7427
7428
7429
7430
7431
7432
7433
7434
7435
7436
7437
7438
7439
7440
7441
7442
7443
7444
7445
7446
7447
7448
7449
7450
7451
7452
7453
7454
7455
7456
7457
7458
7459
7460
7461
7462
7463
7464
7465
7466
7467
7468
7469
7470
7471
7472
7473
7474
7475
7476
7477
7478
7479
7480
7481
7482
7483
7484
7485
7486
7487
7488
7489
7490
7491
7492
7493
7494
7495
7496
7497
7498
7499
7500
7501
7502
7503
7504
7505
7506
7507
7508
7509
7510
7511
7512
7513
7514
7515
7516
7517
7518
7519
7520
7521
7522
7523
7524
7525
7526
7527
7528
7529
7530
7531
7532
7533
7534
7535
7536
7537
7538
7539
7540
7541
7542
7543
7544
7545
7546
7547
7548
7549
7550
7551
7552
7553
7554
7555
7556
7557
7558
7559
7560
7561
7562
7563
7564
7565
7566
7567
7568
7569
7570
7571
7572
7573
7574
7575
7576
7577
7578
7579
7580
7581
7582
7583
7584
7585
7586
7587
7588
7589
7590
7591
7592
7593
7594
7595
7596
7597
7598
7599
7600
7601
7602
7603
7604
7605
7606
7607
7608
7609
7610
7611
7612
7613
7614
7615
7616
7617
7618
7619
7620
7621
7622
7623
7624
7625
7626
7627
7628
7629
7630
7631
7632
7633
7634
7635
7636
7637
7638
7639
7640
7641
7642
7643
7644
7645
7646
7647
7648
7649
7650
7651
7652
7653
7654
7655
7656
7657
7658
7659
7660
7661
7662
7663
7664
7665
7666
7667
7668
7669
7670
7671
7672
7673
7674
7675
7676
7677
7678
7679
7680
7681
7682
7683
7684
7685
7686
7687
7688
7689
7690
7691
7692
7693
7694
7695
7696
7697
7698
7699
7700
7701
7702
7703
7704
7705
7706
7707
7708
7709
7710
7711
7712
7713
7714
7715
7716
7717
7718
7719
7720
7721
7722
7723
7724
7725
7726
7727
7728
7729
7730
7731
7732
7733
7734
7735
7736
7737
7738
7739
7740
7741
7742
7743
7744
7745
7746
7747
7748
7749
7750
7751
7752
7753
7754
7755
7756
7757
7758
7759
7760
7761
7762
7763
7764
7765
7766
7767
7768
7769
7770
7771
7772
7773
7774
7775
7776
7777
7778
7779
7780
7781
7782
7783
7784
7785
7786
7787
7788
7789
7790
7791
7792
7793
7794
7795
7796
7797
7798
7799
7800
7801
7802
7803
7804
7805
7806
7807
7808
7809
7810
7811
7812
7813
7814
7815
7816
7817
7818
7819
7820
7821
7822
7823
7824
7825
7826
7827
7828
7829
7830
7831
7832
7833
7834
7835
7836
7837
7838
7839
7840
7841
7842
7843
7844
7845
7846
7847
7848
7849
7850
7851
7852
7853
7854
7855
7856
7857
7858
7859
7860
7861
7862
7863
7864
7865
7866
7867
7868
7869
7870
7871
7872
7873
7874
7875
7876
7877
7878
7879
7880
7881
7882
7883
7884
7885
7886
7887
7888
7889
7890
7891
7892
7893
7894
7895
7896
7897
7898
7899
7900
7901
7902
7903
7904
7905
7906
7907
7908
7909
7910
7911
7912
7913
7914
7915
7916
7917
7918
7919
7920
7921
7922
7923
7924
7925
7926
7927
7928
7929
7930
7931
7932
7933
7934
7935
7936
7937
7938
7939
7940
7941
7942
7943
7944
7945
7946
7947
7948
7949
7950
7951
7952
7953
7954
7955
7956
7957
7958
7959
7960
7961
7962
7963
7964
7965
7966
7967
7968
7969
7970
7971
7972
7973
7974
7975
7976
7977
7978
7979
7980
7981
7982
7983
7984
7985
7986
7987
7988
7989
7990
7991
7992
7993
7994
7995
7996
7997
7998
7999
8000
8001
8002
8003
8004
8005
8006
8007
8008
8009
8010
8011
8012
8013
8014
8015
8016
8017
8018
8019
8020
8021
8022
8023
8024
8025
8026
8027
8028
8029
8030
8031
8032
8033
8034
8035
8036
8037
8038
8039
8040
8041
8042
8043
8044
8045
8046
8047
8048
8049
8050
8051
8052
8053
8054
8055
8056
8057
8058
8059
8060
8061
8062
8063
8064
8065
8066
8067
8068
8069
8070
8071
8072
8073
8074
8075
8076
8077
8078
8079
8080
8081
8082
8083
8084
8085
8086
8087
8088
8089
8090
8091
8092
8093
8094
8095
8096
8097
8098
8099
8100
8101
8102
8103
8104
8105
8106
8107
8108
8109
8110
8111
8112
8113
8114
8115
8116
8117
8118
8119
8120
8121
8122
8123
8124
8125
8126
8127
8128
8129
8130
8131
8132
8133
8134
8135
8136
8137
8138
8139
8140
8141
8142
8143
8144
8145
8146
8147
8148
8149
8150
8151
8152
8153
8154
8155
8156
8157
8158
8159
8160
8161
8162
8163
8164
8165
8166
8167
8168
8169
8170
8171
8172
8173
8174
8175
8176
8177
8178
8179
8180
8181
8182
8183
8184
8185
8186
8187
8188
8189
8190
8191
8192
8193
8194
8195
8196
8197
8198
8199
8200
8201
8202
8203
8204
8205
8206
8207
8208
8209
8210
8211
8212
8213
8214
8215
8216
8217
8218
8219
8220
8221
8222
8223
8224
8225
8226
8227
8228
8229
8230
8231
8232
8233
8234
8235
8236
8237
8238
8239
8240
8241
8242
8243
8244
8245
8246
8247
8248
8249
8250
8251
8252
8253
8254
8255
8256
8257
8258
8259
8260
8261
8262
8263
8264
8265
8266
8267
8268
8269
8270
8271
8272
8273
8274
8275
8276
8277
8278
8279
8280
8281
8282
8283
8284
8285
8286
8287
8288
8289
8290
8291
8292
8293
8294
8295
8296
8297
8298
8299
8300
8301
8302
8303
8304
8305
8306
8307
8308
8309
8310
8311
8312
8313
8314
8315
8316
8317
8318
8319
8320
8321
8322
8323
8324
8325
8326
8327
8328
8329
8330
8331
8332
8333
8334
8335
8336
8337
8338
8339
8340
8341
8342
8343
8344
8345
8346
8347
8348
8349
8350
8351
8352
8353
8354
8355
8356
8357
8358
8359
8360
8361
8362
8363
8364
8365
8366
8367
8368
8369
8370
8371
8372
8373
8374
8375
8376
8377
8378
8379
8380
8381
8382
8383
8384
8385
8386
8387
8388
8389
8390
8391
8392
8393
8394
8395
8396
8397
8398
8399
8400
8401
8402
8403
8404
8405
8406
8407
8408
8409
8410
8411
8412
8413
8414
8415
8416
8417
8418
8419
8420
8421
8422
8423
8424
8425
8426
8427
8428
8429
8430
8431
8432
8433
8434
8435
8436
8437
8438
8439
8440
8441
8442
8443
8444
8445
8446
8447
8448
8449
8450
8451
8452
8453
8454
8455
8456
8457
8458
8459
8460
8461
8462
8463
8464
8465
8466
8467
8468
8469
8470
8471
8472
8473
8474
8475
8476
8477
8478
8479
8480
8481
8482
8483
8484
8485
8486
8487
8488
8489
8490
8491
8492
8493
8494
8495
8496
8497
8498
8499
8500
8501
8502
8503
8504
8505
8506
8507
8508
8509
8510
8511
8512
8513
8514
8515
8516
8517
8518
8519
8520
8521
8522
8523
8524
8525
8526
8527
8528
8529
8530
8531
8532
8533
8534
8535
8536
8537
8538
8539
8540
8541
8542
8543
8544
8545
8546
8547
8548
8549
8550
8551
8552
8553
8554
8555
8556
8557
8558
8559
8560
8561
8562
8563
8564
8565
8566
8567
8568
8569
8570
8571
8572
8573
8574
8575
8576
8577
8578
8579
8580
8581
8582
8583
8584
8585
8586
8587
8588
8589
8590
8591
8592
8593
8594
8595
8596
8597
8598
8599
8600
8601
8602
8603
8604
8605
8606
8607
8608
8609
8610
8611
8612
8613
8614
8615
8616
8617
8618
8619
8620
8621
8622
8623
8624
8625
8626
8627
8628
8629
8630
8631
8632
8633
8634
8635
8636
8637
8638
8639
8640
8641
8642
8643
8644
8645
8646
8647
8648
8649
8650
8651
8652
8653
8654
8655
8656
8657
8658
8659
8660
8661
8662
8663
8664
8665
8666
8667
8668
8669
8670
8671
8672
8673
8674
8675
8676
8677
8678
8679
8680
8681
8682
8683
8684
8685
8686
8687
8688
8689
8690
8691
8692
8693
8694
8695
8696
8697
8698
8699
8700
8701
8702
8703
8704
8705
8706
8707
8708
8709
8710
8711
8712
8713
8714
8715
8716
8717
8718
8719
8720
8721
8722
8723
8724
8725
8726
8727
8728
8729
8730
8731
8732
8733
8734
8735
8736
8737
8738
8739
8740
8741
8742
8743
8744
8745
8746
8747
8748
8749
8750
8751
8752
8753
8754
8755
8756
8757
8758
8759
8760
8761
8762
8763
8764
8765
8766
8767
8768
8769
8770
8771
8772
8773
8774
8775
8776
8777
8778
8779
8780
8781
8782
8783
8784
8785
8786
8787
8788
8789
8790
8791
8792
8793
8794
8795
8796
8797
8798
8799
8800
8801
8802
8803
8804
8805
8806
8807
8808
8809
8810
8811
8812
8813
8814
8815
8816
8817
8818
8819
8820
8821
8822
8823
8824
8825
8826
8827
8828
8829
8830
8831
8832
8833
8834
8835
8836
8837
8838
8839
8840
8841
8842
8843
8844
8845
8846
8847
8848
8849
8850
8851
8852
8853
8854
8855
8856
8857
8858
8859
8860
8861
8862
8863
8864
8865
8866
8867
8868
8869
8870
8871
8872
8873
8874
8875
8876
8877
8878
8879
8880
8881
8882
8883
8884
8885
8886
8887
8888
8889
8890
8891
8892
8893
8894
8895
8896
8897
8898
8899
8900
8901
8902
8903
8904
8905
8906
8907
8908
8909
8910
8911
8912
8913
8914
8915
8916
8917
8918
8919
8920
8921
8922
8923
8924
8925
8926
8927
8928
8929
8930
8931
8932
8933
8934
8935
8936
8937
8938
8939
8940
8941
8942
8943
8944
8945
8946
8947
8948
8949
8950
8951
8952
8953
8954
8955
8956
8957
8958
8959
8960
8961
8962
8963
8964
8965
8966
8967
8968
8969
8970
8971
8972
8973
8974
8975
8976
8977
8978
8979
8980
8981
8982
8983
8984
8985
8986
8987
8988
8989
8990
8991
8992
8993
8994
8995
8996
8997
8998
8999
9000
9001
9002
9003
9004
9005
9006
9007
9008
9009
9010
9011
9012
9013
9014
9015
9016
9017
9018
9019
9020
9021
9022
9023
9024
9025
9026
9027
9028
9029
9030
9031
9032
9033
9034
9035
9036
9037
9038
9039
9040
9041
9042
9043
9044
9045
9046
9047
9048
9049
9050
9051
9052
9053
9054
9055
9056
9057
9058
9059
9060
9061
9062
9063
9064
9065
9066
9067
9068
9069
9070
9071
9072
9073
9074
9075
9076
9077
9078
9079
9080
9081
9082
9083
9084
9085
9086
9087
9088
9089
9090
9091
9092
9093
9094
9095
9096
9097
9098
9099
9100
9101
9102
9103
9104
9105
9106
9107
9108
9109
9110
9111
9112
9113
9114
9115
9116
9117
9118
9119
9120
9121
9122
9123
9124
9125
9126
9127
9128
9129
9130
9131
9132
9133
9134
9135
9136
9137
9138
9139
9140
9141
9142
9143
9144
9145
9146
9147
9148
9149
9150
9151
9152
9153
9154
9155
9156
9157
9158
9159
9160
9161
9162
9163
9164
9165
9166
9167
9168
9169
9170
9171
9172
9173
9174
9175
9176
9177
9178
9179
9180
9181
9182
9183
9184
9185
9186
9187
9188
9189
9190
9191
9192
9193
9194
9195
9196
9197
9198
9199
9200
9201
9202
9203
9204
9205
9206
9207
9208
9209
9210
9211
9212
9213
9214
9215
9216
9217
9218
9219
9220
9221
9222
9223
9224
9225
9226
9227
9228
9229
9230
9231
9232
9233
9234
9235
9236
9237
9238
9239
9240
9241
9242
9243
9244
9245
9246
9247
9248
9249
9250
9251
9252
9253
9254
9255
9256
9257
9258
9259
9260
9261
9262
9263
9264
9265
9266
9267
9268
9269
9270
9271
9272
9273
9274
9275
9276
9277
9278
9279
9280
9281
9282
9283
9284
9285
9286
9287
9288
9289
9290
9291
9292
9293
9294
9295
9296
9297
9298
9299
9300
9301
9302
9303
9304
9305
9306
9307
9308
9309
9310
9311
9312
9313
9314
9315
9316
9317
9318
9319
9320
9321
9322
9323
9324
9325
9326
9327
9328
9329
9330
9331
9332
9333
9334
9335
9336
9337
9338
9339
9340
9341
9342
9343
9344
9345
9346
9347
9348
9349
9350
9351
9352
9353
9354
9355
9356
9357
9358
9359
9360
9361
9362
9363
9364
9365
9366
9367
9368
9369
9370
9371
9372
9373
9374
9375
9376
9377
9378
9379
9380
9381
9382
9383
9384
9385
9386
9387
9388
9389
9390
9391
9392
9393
9394
9395
9396
9397
9398
9399
9400
9401
9402
9403
9404
9405
9406
9407
9408
9409
9410
9411
9412
9413
9414
9415
9416
9417
9418
9419
9420
9421
9422
9423
9424
9425
9426
9427
9428
9429
9430
9431
9432
9433
9434
9435
9436
9437
9438
9439
9440
9441
9442
9443
9444
9445
9446
9447
9448
9449
9450
9451
9452
9453
9454
9455
9456
9457
9458
9459
9460
9461
9462
9463
9464
9465
9466
9467
9468
9469
9470
9471
9472
9473
9474
9475
9476
9477
9478
9479
9480
9481
9482
9483
9484
9485
9486
9487
9488
9489
9490
9491
9492
9493
9494
9495
9496
9497
9498
9499
9500
9501
9502
9503
9504
9505
9506
9507
9508
9509
9510
9511
9512
9513
9514
9515
9516
9517
9518
9519
9520
9521
9522
9523
9524
9525
9526
9527
9528
9529
9530
9531
9532
9533
9534
9535
9536
9537
9538
9539
9540
9541
9542
9543
9544
9545
9546
9547
9548
9549
9550
9551
9552
9553
9554
9555
9556
9557
9558
9559
9560
9561
9562
9563
9564
9565
9566
9567
9568
9569
9570
9571
9572
9573
9574
9575
9576
9577
9578
9579
9580
9581
9582
9583
9584
9585
9586
9587
9588
9589
9590
9591
9592
9593
9594
9595
9596
9597
9598
9599
9600
9601
9602
9603
9604
9605
9606
9607
9608
9609
9610
9611
9612
9613
9614
9615
9616
9617
9618
9619
9620
9621
9622
9623
9624
9625
9626
9627
9628
9629
9630
9631
9632
9633
9634
9635
9636
9637
9638
9639
9640
9641
9642
9643
9644
9645
9646
9647
9648
9649
9650
9651
9652
9653
9654
9655
9656
9657
9658
9659
9660
9661
9662
9663
9664
9665
9666
9667
9668
9669
9670
9671
9672
9673
9674
9675
9676
9677
9678
9679
9680
9681
9682
9683
9684
9685
9686
9687
9688
9689
9690
9691
9692
9693
9694
9695
9696
9697
9698
9699
9700
9701
9702
9703
9704
9705
9706
9707
9708
9709
9710
9711
9712
9713
9714
9715
9716
9717
9718
9719
9720
9721
9722
9723
9724
9725
9726
9727
9728
9729
9730
9731
9732
9733
9734
9735
9736
9737
9738
9739
9740
9741
9742
9743
9744
9745
9746
9747
9748
9749
9750
9751
9752
9753
9754
9755
9756
9757
9758
9759
9760
9761
9762
9763
9764
9765
9766
9767
9768
9769
9770
9771
9772
9773
9774
9775
9776
9777
9778
9779
9780
9781
9782
9783
9784
9785
9786
9787
9788
9789
9790
9791
9792
9793
9794
9795
9796
9797
9798
9799
9800
9801
9802
9803
9804
9805
9806
9807
9808
9809
9810
9811
9812
9813
9814
9815
9816
9817
9818
9819
9820
9821
9822
9823
9824
9825
9826
9827
9828
9829
9830
9831
9832
9833
9834
9835
9836
9837
9838
9839
9840
9841
9842
9843
9844
9845
9846
9847
9848
9849
9850
9851
9852
9853
9854
9855
9856
9857
9858
9859
9860
9861
9862
9863
9864
9865
9866
9867
9868
9869
9870
9871
9872
9873
9874
9875
9876
9877
9878
9879
9880
9881
9882
9883
9884
9885
9886
9887
9888
9889
9890
9891
9892
9893
9894
9895
9896
9897
9898
9899
9900
9901
9902
9903
9904
9905
9906
9907
9908
9909
9910
9911
9912
9913
9914
9915
9916
9917
9918
9919
9920
9921
9922
9923
9924
9925
9926
9927
9928
9929
9930
9931
9932
9933
9934
9935
9936
9937
9938
9939
9940
9941
9942
9943
9944
9945
9946
9947
9948
9949
9950
9951
9952
9953
9954
9955
9956
9957
9958
9959
9960
9961
9962
9963
9964
9965
9966
9967
9968
9969
9970
9971
9972
9973
9974
9975
9976
9977
9978
9979
9980
9981
9982
9983
9984
9985
9986
9987
9988
9989
9990
9991
9992
9993
9994
9995
9996
9997
9998
9999
10000
10001
10002
10003
10004
10005
10006
10007
10008
10009
10010
10011
10012
10013
10014
10015
10016
10017
10018
10019
10020
10021
10022
10023
10024
10025
10026
10027
10028
10029
10030
10031
10032
10033
10034
10035
10036
10037
10038
10039
10040
10041
10042
10043
10044
10045
10046
10047
10048
10049
10050
10051
10052
10053
10054
10055
10056
10057
10058
10059
10060
10061
10062
10063
10064
10065
10066
10067
10068
10069
10070
10071
10072
10073
10074
10075
10076
10077
10078
10079
10080
10081
10082
10083
10084
10085
10086
10087
10088
10089
10090
10091
10092
10093
10094
10095
10096
10097
10098
10099
10100
10101
10102
10103
10104
10105
10106
10107
10108
10109
10110
10111
10112
10113
10114
10115
10116
10117
10118
10119
10120
10121
10122
10123
10124
10125
10126
10127
10128
10129
10130
10131
10132
10133
10134
10135
10136
10137
10138
10139
10140
10141
10142
10143
10144
10145
10146
10147
10148
10149
10150
10151
10152
10153
10154
10155
10156
10157
10158
10159
10160
10161
10162
10163
10164
10165
10166
10167
10168
10169
10170
10171
10172
10173
10174
10175
10176
10177
10178
10179
10180
10181
10182
10183
10184
10185
10186
10187
10188
10189
10190
10191
10192
10193
10194
10195
10196
10197
10198
10199
10200
10201
10202
10203
10204
10205
10206
10207
10208
10209
10210
10211
10212
10213
10214
10215
10216
10217
10218
10219
10220
10221
10222
10223
10224
10225
10226
10227
10228
10229
10230
10231
10232
10233
10234
10235
10236
10237
10238
10239
10240
10241
10242
10243
10244
10245
10246
10247
10248
10249
10250
10251
10252
10253
10254
10255
10256
10257
10258
10259
10260
10261
10262
10263
10264
10265
10266
10267
10268
10269
10270
10271
10272
10273
10274
10275
10276
10277
10278
10279
10280
10281
10282
10283
10284
10285
10286
10287
10288
10289
10290
10291
10292
10293
10294
10295
10296
10297
10298
10299
10300
10301
10302
10303
10304
10305
10306
10307
10308
10309
10310
10311
10312
10313
10314
10315
10316
10317
10318
10319
10320
10321
10322
10323
10324
10325
10326
10327
10328
10329
10330
10331
10332
10333
10334
10335
10336
10337
10338
10339
10340
10341
10342
10343
10344
10345
10346
10347
10348
10349
10350
10351
10352
10353
10354
10355
10356
10357
10358
10359
10360
10361
10362
10363
10364
10365
10366
10367
10368
10369
10370
10371
10372
10373
10374
10375
10376
10377
10378
10379
10380
10381
10382
10383
10384
10385
10386
10387
10388
10389
10390
10391
10392
10393
10394
10395
10396
10397
10398
10399
10400
10401
10402
10403
10404
10405
10406
10407
10408
10409
10410
10411
10412
10413
10414
10415
10416
10417
10418
10419
10420
10421
10422
10423
10424
10425
10426
10427
10428
10429
10430
10431
10432
10433
10434
10435
10436
10437
10438
10439
10440
10441
10442
10443
10444
10445
10446
10447
10448
10449
10450
10451
10452
10453
10454
10455
10456
10457
10458
10459
10460
10461
10462
10463
10464
10465
10466
10467
10468
10469
10470
10471
10472
10473
10474
10475
10476
10477
10478
10479
10480
10481
10482
10483
10484
10485
10486
10487
10488
10489
10490
10491
10492
10493
10494
10495
10496
10497
10498
10499
10500
10501
10502
10503
10504
10505
10506
10507
10508
10509
10510
10511
10512
10513
10514
10515
10516
10517
10518
10519
10520
10521
10522
10523
10524
10525
10526
10527
10528
10529
10530
10531
10532
10533
10534
10535
10536
10537
10538
10539
10540
10541
10542
10543
10544
10545
10546
10547
10548
10549
10550
10551
10552
10553
10554
10555
10556
10557
10558
10559
10560
10561
10562
10563
10564
10565
10566
10567
10568
10569
10570
10571
10572
10573
10574
10575
10576
10577
10578
10579
10580
10581
10582
10583
10584
10585
10586
10587
10588
10589
10590
10591
10592
10593
10594
10595
10596
10597
10598
10599
10600
10601
10602
10603
10604
10605
10606
10607
10608
10609
10610
10611
10612
10613
10614
10615
10616
10617
10618
10619
10620
10621
10622
10623
10624
10625
10626
10627
10628
10629
10630
10631
10632
10633
10634
10635
10636
10637
10638
10639
10640
10641
10642
10643
10644
10645
10646
10647
10648
10649
10650
10651
10652
10653
10654
10655
10656
10657
10658
10659
10660
10661
10662
10663
10664
10665
10666
10667
10668
10669
10670
10671
10672
10673
10674
10675
10676
10677
10678
10679
10680
10681
10682
10683
10684
10685
10686
10687
10688
10689
10690
10691
10692
10693
10694
10695
10696
10697
10698
10699
10700
10701
10702
10703
10704
10705
10706
10707
10708
10709
10710
10711
10712
10713
10714
10715
10716
10717
10718
10719
10720
10721
10722
10723
10724
10725
10726
10727
10728
10729
10730
10731
10732
10733
10734
10735
10736
10737
10738
10739
10740
10741
10742
10743
10744
10745
10746
10747
10748
10749
10750
10751
10752
10753
10754
10755
10756
10757
10758
10759
10760
10761
10762
10763
10764
10765
10766
10767
10768
10769
10770
10771
10772
10773
10774
10775
10776
10777
10778
10779
10780
10781
10782
10783
10784
10785
10786
10787
10788
10789
10790
10791
10792
10793
10794
10795
10796
10797
10798
10799
10800
10801
10802
10803
10804
10805
10806
10807
10808
10809
10810
10811
10812
10813
10814
10815
10816
10817
10818
10819
10820
10821
10822
10823
10824
10825
10826
10827
10828
10829
10830
10831
10832
10833
10834
10835
10836
10837
10838
10839
10840
10841
10842
10843
10844
10845
10846
10847
10848
10849
10850
10851
10852
10853
10854
10855
10856
10857
10858
10859
10860
10861
10862
10863
10864
10865
10866
10867
10868
10869
10870
10871
10872
10873
10874
10875
10876
10877
10878
10879
10880
10881
10882
10883
10884
10885
10886
10887
10888
10889
10890
10891
10892
10893
10894
10895
10896
10897
10898
10899
10900
10901
10902
10903
10904
10905
10906
10907
10908
10909
10910
10911
10912
10913
10914
10915
10916
10917
10918
10919
10920
10921
10922
10923
10924
10925
10926
10927
10928
10929
10930
10931
10932
10933
10934
10935
10936
10937
10938
10939
10940
10941
10942
10943
10944
10945
10946
10947
10948
10949
10950
10951
10952
10953
10954
10955
10956
10957
10958
10959
10960
10961
10962
10963
10964
10965
10966
10967
10968
10969
10970
10971
10972
10973
10974
10975
10976
10977
10978
10979
10980
10981
10982
10983
10984
10985
10986
10987
10988
10989
10990
10991
10992
10993
10994
10995
10996
10997
10998
10999
11000
11001
11002
11003
11004
11005
11006
11007
11008
11009
11010
11011
11012
11013
11014
11015
11016
11017
11018
11019
11020
11021
11022
11023
11024
11025
11026
11027
11028
11029
11030
11031
11032
11033
11034
11035
11036
11037
11038
11039
11040
11041
11042
11043
11044
11045
11046
11047
11048
11049
11050
11051
11052
11053
11054
11055
11056
11057
11058
11059
11060
11061
11062
11063
11064
11065
11066
11067
11068
11069
11070
11071
11072
11073
11074
11075
11076
11077
11078
11079
11080
11081
11082
11083
11084
11085
11086
11087
11088
11089
11090
11091
11092
11093
11094
11095
11096
11097
11098
11099
11100
11101
11102
11103
11104
11105
11106
11107
11108
11109
11110
11111
11112
11113
11114
11115
11116
11117
11118
11119
11120
11121
11122
11123
11124
11125
11126
11127
11128
11129
11130
11131
11132
11133
11134
11135
11136
11137
11138
11139
11140
11141
11142
11143
11144
11145
11146
11147
11148
11149
11150
11151
11152
11153
11154
11155
11156
11157
11158
11159
11160
11161
11162
11163
11164
11165
11166
11167
11168
11169
11170
11171
11172
11173
11174
11175
11176
11177
11178
11179
11180
11181
11182
11183
11184
11185
11186
11187
11188
11189
11190
11191
11192
11193
11194
11195
11196
11197
11198
11199
11200
11201
11202
11203
11204
11205
11206
11207
11208
11209
11210
11211
11212
11213
11214
11215
11216
11217
11218
11219
11220
11221
11222
11223
11224
11225
11226
11227
11228
11229
11230
11231
11232
11233
11234
11235
11236
11237
11238
11239
11240
11241
11242
11243
11244
11245
11246
11247
11248
11249
11250
11251
11252
11253
11254
11255
11256
11257
11258
11259
11260
11261
11262
11263
11264
11265
11266
11267
11268
11269
11270
11271
11272
11273
11274
11275
11276
11277
11278
11279
11280
11281
11282
11283
11284
11285
11286
11287
11288
11289
11290
11291
11292
11293
11294
11295
11296
11297
11298
11299
11300
11301
11302
11303
11304
11305
11306
11307
11308
11309
11310
11311
11312
11313
11314
11315
11316
11317
11318
11319
11320
11321
11322
11323
11324
11325
11326
11327
11328
11329
11330
11331
11332
11333
11334
11335
11336
11337
11338
11339
11340
11341
11342
11343
11344
11345
11346
11347
11348
11349
11350
11351
11352
11353
11354
11355
11356
11357
11358
11359
11360
11361
11362
11363
11364
11365
11366
11367
11368
11369
11370
11371
11372
11373
11374
11375
11376
11377
11378
11379
11380
11381
11382
11383
11384
11385
11386
11387
11388
11389
11390
11391
11392
11393
11394
11395
11396
11397
11398
11399
11400
11401
11402
11403
11404
11405
11406
11407
11408
11409
11410
11411
11412
11413
11414
11415
11416
11417
11418
11419
11420
11421
11422
11423
11424
11425
11426
11427
11428
11429
11430
11431
11432
11433
11434
11435
11436
11437
11438
11439
11440
11441
11442
11443
11444
11445
11446
11447
11448
11449
11450
11451
11452
11453
11454
11455
11456
11457
11458
11459
11460
11461
11462
11463
11464
11465
11466
11467
11468
11469
11470
11471
11472
11473
11474
11475
11476
11477
11478
11479
11480
11481
11482
11483
11484
11485
11486
11487
11488
11489
11490
11491
11492
11493
11494
11495
11496
11497
11498
11499
11500
11501
11502
11503
11504
11505
11506
11507
11508
11509
11510
11511
11512
11513
11514
11515
11516
11517
11518
11519
11520
11521
11522
11523
11524
11525
11526
11527
11528
11529
11530
11531
11532
11533
11534
11535
11536
11537
11538
11539
11540
11541
11542
11543
11544
11545
11546
11547
11548
11549
11550
11551
11552
11553
11554
11555
11556
11557
11558
11559
11560
11561
11562
11563
11564
11565
11566
11567
11568
11569
11570
11571
11572
11573
11574
11575
11576
11577
11578
11579
11580
11581
11582
11583
11584
11585
11586
11587
11588
11589
11590
11591
11592
11593
11594
11595
11596
11597
11598
11599
11600
11601
11602
11603
11604
11605
11606
11607
11608
11609
11610
11611
11612
11613
11614
11615
11616
11617
11618
11619
11620
11621
11622
11623
11624
11625
11626
11627
11628
11629
11630
11631
11632
11633
11634
11635
11636
11637
11638
11639
11640
11641
11642
11643
11644
11645
11646
11647
11648
11649
11650
11651
11652
11653
11654
11655
11656
11657
11658
11659
11660
11661
11662
11663
11664
11665
11666
11667
11668
11669
11670
11671
11672
11673
11674
11675
11676
11677
11678
11679
11680
11681
11682
11683
11684
11685
11686
11687
11688
11689
11690
11691
11692
11693
11694
11695
11696
11697
11698
11699
11700
11701
11702
11703
11704
11705
11706
11707
11708
11709
11710
11711
11712
11713
11714
11715
11716
11717
11718
11719
11720
11721
11722
11723
11724
11725
11726
11727
11728
11729
11730
11731
11732
11733
11734
11735
11736
11737
11738
11739
11740
11741
11742
11743
11744
11745
11746
11747
11748
11749
11750
11751
11752
11753
11754
11755
11756
11757
11758
11759
11760
11761
11762
11763
11764
11765
11766
11767
11768
11769
11770
11771
11772
11773
11774
11775
11776
11777
11778
11779
11780
11781
11782
11783
11784
11785
11786
11787
11788
11789
11790
11791
11792
11793
11794
11795
11796
11797
11798
11799
11800
11801
11802
11803
11804
11805
11806
11807
11808
11809
11810
11811
11812
11813
11814
11815
11816
11817
11818
11819
11820
11821
11822
11823
11824
11825
11826
11827
11828
11829
11830
11831
11832
11833
11834
11835
11836
11837
11838
11839
11840
11841
11842
11843
11844
11845
11846
11847
11848
11849
11850
11851
11852
11853
11854
11855
11856
11857
11858
11859
11860
11861
11862
11863
11864
11865
11866
11867
11868
11869
11870
11871
11872
11873
11874
11875
11876
11877
11878
11879
11880
11881
11882
11883
11884
11885
11886
11887
11888
11889
11890
11891
11892
11893
11894
11895
11896
11897
11898
11899
11900
11901
11902
11903
11904
11905
11906
11907
11908
11909
11910
11911
11912
11913
11914
11915
11916
11917
11918
11919
11920
11921
11922
11923
11924
11925
11926
11927
11928
11929
11930
11931
11932
11933
11934
11935
11936
11937
11938
11939
11940
11941
11942
11943
11944
11945
11946
11947
11948
11949
11950
11951
11952
11953
11954
11955
11956
11957
11958
11959
11960
11961
11962
11963
11964
11965
11966
11967
11968
11969
11970
11971
11972
11973
11974
11975
11976
11977
11978
11979
11980
11981
11982
11983
11984
11985
11986
11987
11988
11989
11990
11991
11992
11993
11994
11995
11996
11997
11998
11999
12000
12001
12002
12003
12004
12005
12006
12007
12008
12009
12010
12011
12012
12013
12014
12015
12016
12017
12018
12019
12020
12021
12022
12023
12024
12025
12026
12027
12028
12029
12030
12031
12032
12033
12034
12035
12036
12037
12038
12039
12040
12041
12042
12043
12044
12045
12046
12047
12048
12049
12050
12051
12052
12053
12054
12055
12056
12057
12058
12059
12060
12061
12062
12063
12064
12065
12066
12067
12068
12069
12070
12071
12072
12073
12074
12075
12076
12077
12078
12079
12080
12081
12082
12083
12084
12085
12086
12087
12088
12089
12090
12091
12092
12093
12094
12095
12096
12097
12098
12099
12100
12101
12102
12103
12104
12105
12106
12107
12108
12109
12110
12111
12112
12113
12114
12115
12116
12117
12118
12119
12120
12121
12122
12123
12124
12125
12126
12127
12128
12129
12130
12131
12132
12133
12134
12135
12136
12137
12138
12139
12140
12141
12142
12143
12144
12145
12146
12147
12148
12149
12150
12151
12152
12153
12154
12155
12156
12157
12158
12159
12160
12161
12162
12163
12164
12165
12166
12167
12168
12169
12170
12171
12172
12173
12174
12175
12176
12177
12178
12179
12180
12181
12182
12183
12184
12185
12186
12187
12188
12189
12190
12191
12192
12193
12194
12195
12196
12197
12198
12199
12200
12201
12202
12203
12204
12205
12206
12207
12208
12209
12210
12211
12212
12213
12214
12215
12216
12217
12218
12219
12220
12221
12222
12223
12224
12225
12226
12227
12228
12229
12230
12231
12232
12233
12234
12235
12236
12237
12238
12239
12240
12241
12242
12243
12244
12245
12246
12247
12248
12249
12250
12251
12252
12253
12254
12255
12256
12257
12258
12259
12260
12261
12262
12263
12264
12265
12266
12267
12268
12269
12270
12271
12272
12273
12274
12275
12276
12277
12278
12279
12280
12281
12282
12283
12284
12285
12286
12287
12288
12289
12290
12291
12292
12293
12294
12295
12296
12297
12298
12299
12300
12301
12302
12303
12304
12305
12306
12307
12308
12309
12310
12311
12312
12313
12314
12315
12316
12317
12318
12319
12320
12321
12322
12323
12324
12325
12326
12327
12328
12329
12330
12331
12332
12333
12334
12335
12336
12337
12338
12339
12340
12341
12342
12343
12344
12345
12346
12347
12348
12349
12350
12351
12352
12353
12354
12355
12356
12357
12358
12359
12360
12361
12362
12363
12364
12365
12366
12367
12368
12369
12370
12371
12372
12373
12374
12375
12376
12377
12378
12379
12380
12381
12382
12383
12384
12385
12386
12387
12388
12389
12390
12391
12392
12393
12394
12395
12396
12397
12398
12399
12400
12401
12402
12403
12404
12405
12406
12407
12408
12409
12410
12411
12412
12413
12414
12415
12416
12417
12418
12419
12420
12421
12422
12423
12424
12425
12426
12427
12428
12429
12430
12431
12432
12433
12434
12435
12436
12437
12438
12439
12440
12441
12442
12443
12444
12445
12446
12447
12448
12449
12450
12451
12452
12453
12454
12455
12456
12457
12458
12459
12460
12461
12462
12463
12464
12465
12466
12467
12468
12469
12470
12471
12472
12473
12474
12475
12476
12477
12478
12479
12480
12481
12482
12483
12484
12485
12486
12487
12488
12489
12490
12491
12492
12493
12494
12495
12496
12497
12498
12499
12500
12501
12502
12503
12504
12505
12506
12507
12508
12509
12510
12511
12512
12513
12514
12515
12516
12517
12518
12519
12520
12521
12522
12523
12524
12525
12526
12527
12528
12529
12530
12531
12532
12533
12534
12535
12536
12537
12538
12539
12540
12541
12542
12543
12544
12545
12546
12547
12548
12549
12550
12551
12552
12553
12554
12555
12556
12557
12558
12559
12560
12561
12562
12563
12564
12565
12566
12567
12568
12569
12570
12571
12572
12573
12574
12575
12576
12577
12578
12579
12580
12581
12582
12583
12584
12585
12586
12587
12588
12589
12590
12591
12592
12593
12594
12595
12596
12597
12598
12599
12600
12601
12602
12603
12604
12605
12606
12607
12608
12609
12610
12611
12612
12613
12614
12615
12616
12617
12618
12619
12620
12621
12622
12623
12624
12625
12626
12627
12628
12629
12630
12631
12632
12633
12634
12635
12636
12637
12638
12639
12640
12641
12642
12643
12644
12645
12646
12647
12648
12649
12650
12651
12652
12653
12654
12655
12656
12657
12658
12659
12660
12661
12662
12663
12664
12665
12666
12667
12668
12669
12670
12671
12672
12673
12674
12675
12676
12677
12678
12679
12680
12681
12682
12683
12684
12685
12686
12687
12688
12689
12690
12691
12692
12693
12694
12695
12696
12697
12698
12699
12700
12701
12702
12703
12704
12705
12706
12707
12708
12709
12710
12711
12712
12713
12714
12715
12716
12717
12718
12719
12720
12721
12722
12723
12724
12725
12726
12727
12728
12729
12730
12731
12732
12733
12734
12735
12736
12737
12738
12739
12740
12741
12742
12743
12744
12745
12746
12747
12748
12749
12750
12751
12752
12753
12754
12755
12756
12757
12758
12759
12760
12761
12762
12763
12764
12765
12766
12767
12768
12769
12770
12771
12772
12773
12774
12775
12776
12777
12778
12779
12780
12781
12782
12783
12784
12785
12786
12787
12788
12789
12790
12791
12792
12793
12794
12795
12796
12797
12798
12799
12800
12801
12802
12803
12804
12805
12806
12807
12808
12809
12810
12811
12812
12813
12814
12815
12816
12817
12818
12819
12820
12821
12822
12823
12824
12825
12826
12827
12828
12829
12830
12831
12832
12833
12834
12835
12836
12837
12838
12839
12840
12841
12842
12843
12844
12845
12846
12847
12848
12849
12850
12851
12852
12853
12854
12855
12856
12857
12858
12859
12860
12861
12862
12863
12864
12865
12866
12867
12868
12869
12870
12871
12872
12873
12874
12875
12876
12877
12878
12879
12880
12881
12882
12883
12884
12885
12886
12887
12888
12889
12890
12891
12892
12893
12894
12895
12896
12897
12898
12899
12900
12901
12902
12903
12904
12905
12906
12907
12908
12909
12910
12911
12912
12913
12914
12915
12916
12917
12918
12919
12920
12921
12922
12923
12924
12925
12926
12927
12928
12929
12930
12931
12932
12933
12934
12935
12936
12937
12938
12939
12940
12941
12942
12943
12944
12945
12946
12947
12948
12949
12950
12951
12952
12953
12954
12955
12956
12957
12958
12959
12960
12961
12962
12963
12964
12965
12966
12967
12968
12969
12970
12971
12972
12973
12974
12975
12976
12977
12978
12979
12980
12981
12982
12983
12984
12985
12986
12987
12988
12989
12990
12991
12992
12993
12994
12995
12996
12997
12998
12999
13000
13001
13002
13003
13004
13005
13006
13007
13008
13009
13010
13011
13012
13013
13014
13015
13016
13017
13018
13019
13020
13021
13022
13023
13024
13025
13026
13027
13028
13029
13030
13031
13032
13033
13034
13035
13036
13037
13038
13039
13040
13041
13042
13043
13044
13045
13046
13047
13048
13049
13050
13051
13052
13053
13054
13055
13056
13057
13058
13059
13060
13061
13062
13063
13064
13065
13066
13067
13068
13069
13070
13071
13072
13073
13074
13075
13076
13077
13078
13079
13080
13081
13082
13083
13084
13085
13086
13087
13088
13089
13090
13091
13092
13093
13094
13095
13096
13097
13098
13099
13100
13101
13102
13103
13104
13105
13106
13107
13108
13109
13110
13111
13112
13113
13114
13115
13116
13117
13118
13119
13120
13121
13122
13123
13124
13125
13126
13127
13128
13129
13130
13131
13132
13133
13134
13135
13136
13137
13138
13139
13140
13141
13142
13143
13144
13145
13146
13147
13148
13149
13150
13151
13152
13153
13154
13155
13156
13157
13158
13159
13160
13161
13162
13163
13164
13165
13166
13167
13168
13169
13170
13171
13172
13173
13174
13175
13176
13177
13178
13179
13180
13181
13182
13183
13184
13185
13186
13187
13188
13189
13190
13191
13192
13193
13194
13195
13196
13197
13198
13199
13200
13201
13202
13203
13204
13205
13206
13207
13208
13209
13210
13211
13212
13213
13214
13215
13216
13217
13218
13219
13220
13221
13222
13223
13224
13225
13226
13227
13228
13229
13230
13231
13232
13233
13234
13235
13236
13237
13238
13239
13240
13241
13242
13243
13244
13245
13246
13247
13248
13249
13250
13251
13252
13253
13254
13255
13256
13257
13258
13259
13260
13261
13262
13263
13264
13265
13266
13267
13268
13269
13270
13271
13272
13273
13274
13275
13276
13277
13278
13279
13280
13281
13282
13283
13284
13285
13286
13287
13288
13289
13290
13291
13292
13293
13294
13295
13296
13297
13298
13299
13300
13301
13302
13303
13304
13305
13306
13307
13308
13309
13310
13311
13312
13313
13314
13315
13316
13317
13318
13319
13320
13321
13322
13323
13324
13325
13326
13327
13328
13329
13330
13331
13332
13333
13334
13335
13336
13337
13338
13339
13340
13341
13342
13343
13344
13345
13346
13347
13348
13349
13350
13351
13352
13353
13354
13355
13356
13357
13358
13359
13360
13361
13362
13363
13364
13365
13366
13367
13368
13369
13370
13371
13372
13373
13374
13375
13376
13377
13378
13379
13380
13381
13382
13383
13384
13385
13386
13387
13388
13389
13390
13391
13392
13393
13394
13395
13396
13397
13398
13399
13400
13401
13402
13403
13404
13405
13406
13407
13408
13409
13410
13411
13412
13413
13414
13415
13416
13417
13418
13419
13420
13421
13422
13423
13424
13425
13426
13427
13428
13429
13430
13431
13432
13433
13434
13435
13436
13437
13438
13439
13440
13441
13442
13443
13444
13445
13446
13447
13448
13449
13450
13451
13452
13453
13454
13455
13456
13457
13458
13459
13460
13461
13462
13463
13464
13465
13466
13467
13468
13469
13470
13471
13472
13473
13474
13475
13476
13477
13478
13479
13480
13481
13482
13483
13484
13485
13486
13487
13488
13489
13490
13491
13492
13493
13494
13495
13496
13497
13498
13499
13500
13501
13502
13503
13504
13505
13506
13507
13508
13509
13510
13511
13512
13513
13514
13515
13516
13517
13518
13519
13520
13521
13522
13523
13524
13525
13526
13527
13528
13529
13530
13531
13532
13533
13534
13535
13536
13537
13538
13539
13540
13541
13542
13543
13544
13545
13546
13547
13548
13549
13550
13551
13552
13553
13554
13555
13556
13557
13558
13559
13560
13561
13562
13563
13564
13565
13566
13567
13568
13569
13570
13571
13572
13573
13574
13575
13576
13577
13578
13579
13580
13581
13582
13583
13584
13585
13586
13587
13588
13589
13590
13591
13592
13593
13594
13595
13596
13597
13598
13599
13600
13601
13602
13603
13604
13605
13606
13607
13608
13609
13610
13611
13612
13613
13614
13615
13616
13617
13618
13619
13620
13621
13622
13623
13624
13625
13626
13627
13628
13629
13630
13631
13632
13633
13634
13635
13636
13637
13638
13639
13640
13641
13642
13643
13644
13645
13646
13647
13648
13649
13650
13651
13652
13653
13654
13655
13656
13657
13658
13659
13660
13661
13662
13663
13664
13665
13666
13667
13668
13669
13670
13671
13672
13673
13674
13675
13676
13677
13678
13679
13680
13681
13682
13683
13684
13685
13686
13687
13688
13689
13690
13691
13692
13693
13694
13695
13696
13697
13698
13699
13700
13701
13702
13703
13704
13705
13706
13707
13708
13709
13710
13711
13712
13713
13714
13715
13716
13717
13718
13719
13720
13721
13722
13723
13724
13725
13726
13727
13728
13729
13730
13731
13732
13733
13734
13735
13736
13737
13738
13739
13740
13741
13742
13743
13744
13745
13746
13747
13748
13749
13750
13751
13752
13753
13754
13755
13756
13757
13758
13759
13760
13761
13762
13763
13764
13765
13766
13767
13768
13769
13770
13771
13772
13773
13774
13775
13776
13777
13778
13779
13780
13781
13782
13783
13784
13785
13786
13787
13788
13789
13790
13791
13792
13793
13794
13795
13796
13797
13798
13799
13800
13801
13802
13803
13804
13805
13806
13807
13808
13809
13810
13811
13812
13813
13814
13815
13816
13817
13818
13819
13820
13821
13822
13823
13824
13825
13826
13827
13828
13829
13830
13831
13832
13833
13834
13835
13836
13837
13838
13839
13840
13841
13842
13843
13844
13845
13846
13847
13848
13849
13850
13851
13852
13853
13854
13855
13856
13857
13858
13859
13860
13861
13862
13863
13864
13865
13866
13867
13868
13869
13870
13871
13872
13873
13874
13875
13876
13877
13878
13879
13880
13881
13882
13883
13884
13885
13886
13887
13888
13889
13890
13891
13892
13893
13894
13895
13896
13897
13898
13899
13900
13901
13902
13903
13904
13905
13906
13907
13908
13909
13910
13911
13912
13913
13914
13915
13916
13917
13918
13919
13920
13921
13922
13923
13924
13925
13926
13927
13928
13929
13930
13931
13932
13933
13934
13935
13936
13937
13938
13939
13940
13941
13942
13943
13944
13945
13946
13947
13948
13949
13950
13951
13952
13953
13954
13955
13956
13957
13958
13959
13960
13961
13962
13963
13964
13965
13966
13967
13968
13969
13970
13971
13972
13973
13974
13975
13976
13977
13978
13979
13980
13981
13982
13983
13984
13985
13986
13987
13988
13989
13990
13991
13992
13993
13994
13995
13996
13997
13998
13999
14000
14001
14002
14003
14004
14005
14006
14007
14008
14009
14010
14011
14012
14013
14014
14015
14016
14017
14018
14019
14020
14021
14022
14023
14024
14025
14026
14027
14028
14029
14030
14031
14032
14033
14034
14035
14036
14037
14038
14039
14040
14041
14042
14043
14044
14045
14046
14047
14048
14049
14050
14051
14052
14053
14054
14055
14056
14057
14058
14059
14060
14061
14062
14063
14064
14065
14066
14067
14068
14069
14070
14071
14072
14073
14074
14075
14076
14077
14078
14079
14080
14081
14082
14083
14084
14085
14086
14087
14088
14089
14090
14091
14092
14093
14094
14095
14096
14097
14098
14099
14100
14101
14102
14103
14104
14105
14106
14107
14108
14109
14110
14111
14112
14113
14114
14115
14116
14117
14118
14119
14120
14121
14122
14123
14124
14125
14126
14127
14128
14129
14130
14131
14132
14133
14134
14135
14136
14137
14138
14139
14140
14141
14142
14143
14144
14145
14146
14147
14148
14149
14150
14151
14152
14153
14154
14155
14156
14157
14158
14159
14160
14161
14162
14163
14164
14165
14166
14167
14168
14169
14170
14171
14172
14173
14174
14175
14176
14177
14178
14179
14180
14181
14182
14183
14184
14185
14186
14187
14188
14189
14190
14191
14192
14193
14194
14195
14196
14197
14198
14199
14200
14201
14202
14203
14204
14205
14206
14207
14208
14209
14210
14211
14212
14213
14214
14215
14216
14217
14218
14219
14220
14221
14222
14223
14224
14225
14226
14227
14228
14229
14230
14231
14232
14233
14234
14235
14236
14237
14238
14239
14240
14241
14242
14243
14244
14245
14246
14247
14248
14249
14250
14251
14252
14253
14254
14255
14256
14257
14258
14259
14260
14261
14262
14263
14264
14265
14266
14267
14268
14269
14270
14271
14272
14273
14274
14275
14276
14277
14278
14279
14280
14281
14282
14283
14284
14285
14286
14287
14288
14289
14290
14291
14292
14293
14294
14295
14296
14297
14298
14299
14300
14301
14302
14303
14304
14305
14306
14307
14308
14309
14310
14311
14312
14313
14314
14315
14316
14317
14318
14319
14320
14321
14322
14323
14324
14325
14326
14327
14328
14329
14330
14331
14332
14333
14334
14335
14336
14337
14338
14339
14340
14341
14342
14343
14344
14345
14346
14347
14348
14349
14350
14351
14352
14353
14354
14355
14356
14357
14358
14359
14360
14361
14362
14363
14364
14365
14366
14367
14368
14369
14370
14371
14372
14373
14374
14375
14376
14377
14378
14379
14380
14381
14382
14383
14384
14385
14386
14387
14388
14389
14390
14391
14392
14393
14394
14395
14396
14397
14398
14399
14400
14401
14402
14403
14404
14405
14406
14407
14408
14409
14410
14411
14412
14413
14414
14415
14416
14417
14418
14419
14420
14421
14422
14423
14424
14425
14426
14427
14428
14429
14430
14431
14432
14433
14434
14435
14436
14437
14438
14439
14440
14441
14442
14443
14444
14445
14446
14447
14448
14449
14450
14451
14452
14453
14454
14455
14456
14457
14458
14459
14460
14461
14462
14463
14464
14465
14466
14467
14468
14469
14470
14471
14472
14473
14474
14475
14476
14477
14478
14479
14480
14481
14482
14483
14484
14485
14486
14487
14488
14489
14490
14491
14492
14493
14494
14495
14496
14497
14498
14499
14500
14501
14502
14503
14504
14505
14506
14507
14508
14509
14510
14511
14512
14513
14514
14515
14516
14517
14518
14519
14520
14521
14522
14523
14524
14525
14526
14527
14528
14529
14530
14531
14532
14533
14534
14535
14536
14537
14538
14539
14540
14541
14542
14543
14544
14545
14546
14547
14548
14549
14550
14551
14552
14553
14554
14555
14556
14557
14558
14559
14560
14561
14562
14563
14564
14565
14566
14567
14568
14569
14570
14571
14572
14573
14574
14575
14576
14577
14578
14579
14580
14581
14582
14583
14584
14585
14586
14587
14588
14589
14590
14591
14592
14593
14594
14595
14596
14597
14598
14599
14600
14601
14602
14603
14604
14605
14606
14607
14608
14609
14610
14611
14612
14613
14614
14615
14616
14617
14618
14619
14620
14621
14622
14623
14624
14625
14626
14627
14628
14629
14630
14631
14632
14633
14634
14635
14636
14637
14638
14639
14640
14641
14642
14643
14644
14645
14646
14647
14648
14649
14650
14651
14652
14653
14654
14655
14656
14657
14658
14659
14660
14661
14662
14663
14664
14665
14666
14667
14668
14669
14670
14671
14672
14673
14674
14675
14676
14677
14678
14679
14680
14681
14682
14683
14684
14685
14686
14687
14688
14689
14690
14691
14692
14693
14694
14695
14696
14697
14698
14699
14700
14701
14702
14703
14704
14705
14706
14707
14708
14709
14710
14711
14712
14713
14714
14715
14716
14717
14718
14719
14720
14721
14722
14723
14724
14725
14726
14727
14728
14729
14730
14731
14732
14733
14734
14735
14736
14737
14738
14739
14740
14741
14742
14743
14744
14745
14746
14747
14748
14749
14750
14751
14752
14753
14754
14755
14756
14757
14758
14759
14760
14761
14762
14763
14764
14765
14766
14767
14768
14769
14770
14771
14772
14773
14774
14775
14776
14777
14778
14779
14780
14781
14782
14783
14784
14785
14786
14787
14788
14789
14790
14791
14792
14793
14794
14795
14796
14797
14798
14799
14800
14801
14802
14803
14804
14805
14806
14807
14808
14809
14810
14811
14812
14813
14814
14815
14816
14817
14818
14819
14820
14821
14822
14823
14824
14825
14826
14827
14828
14829
14830
14831
14832
14833
14834
14835
14836
14837
14838
14839
14840
14841
14842
14843
14844
14845
14846
14847
14848
14849
14850
14851
14852
14853
14854
14855
14856
14857
14858
14859
14860
14861
14862
14863
14864
14865
14866
14867
14868
14869
14870
14871
14872
14873
14874
14875
14876
14877
14878
14879
14880
14881
14882
14883
14884
14885
14886
14887
14888
14889
14890
14891
14892
14893
14894
14895
14896
14897
14898
14899
14900
14901
14902
14903
14904
14905
14906
14907
14908
14909
14910
14911
14912
14913
14914
14915
14916
14917
14918
14919
14920
14921
14922
14923
14924
14925
14926
14927
14928
14929
14930
14931
14932
14933
14934
14935
14936
14937
14938
14939
14940
14941
14942
14943
14944
14945
14946
14947
14948
14949
14950
14951
14952
14953
14954
14955
14956
14957
14958
14959
14960
14961
14962
14963
14964
14965
14966
14967
14968
14969
14970
14971
14972
14973
14974
14975
14976
14977
14978
14979
14980
14981
14982
14983
14984
14985
14986
14987
14988
14989
14990
14991
14992
14993
14994
14995
14996
14997
14998
14999
15000
15001
15002
15003
15004
15005
15006
15007
15008
15009
15010
15011
15012
15013
15014
15015
15016
15017
15018
15019
15020
15021
15022
15023
15024
15025
15026
15027
15028
15029
15030
15031
15032
15033
15034
15035
15036
15037
15038
15039
15040
15041
15042
15043
15044
15045
15046
15047
15048
15049
15050
15051
15052
15053
15054
15055
15056
15057
15058
15059
15060
15061
15062
15063
15064
15065
15066
15067
15068
15069
15070
15071
15072
15073
15074
15075
15076
15077
15078
15079
15080
15081
15082
15083
15084
15085
15086
15087
15088
15089
15090
15091
15092
15093
15094
15095
15096
15097
15098
15099
15100
15101
15102
15103
15104
15105
15106
15107
15108
15109
15110
15111
15112
15113
15114
15115
15116
15117
15118
15119
15120
15121
15122
15123
15124
15125
15126
15127
15128
15129
15130
15131
15132
15133
15134
15135
15136
15137
15138
15139
15140
15141
15142
15143
15144
15145
15146
15147
15148
15149
15150
15151
15152
15153
15154
15155
15156
15157
15158
15159
15160
15161
15162
15163
15164
15165
15166
15167
15168
15169
15170
15171
15172
15173
15174
15175
15176
15177
15178
15179
15180
15181
15182
15183
15184
15185
15186
15187
15188
15189
15190
15191
15192
15193
15194
15195
15196
15197
15198
15199
15200
15201
15202
15203
15204
15205
15206
15207
15208
15209
15210
15211
15212
15213
15214
15215
15216
15217
15218
15219
15220
15221
15222
15223
15224
15225
15226
15227
15228
15229
15230
15231
15232
15233
15234
15235
15236
15237
15238
15239
15240
15241
15242
15243
15244
15245
15246
15247
15248
15249
15250
15251
15252
15253
15254
15255
15256
15257
15258
15259
15260
15261
15262
15263
15264
15265
15266
15267
15268
15269
15270
15271
15272
15273
15274
15275
15276
15277
15278
15279
15280
15281
15282
15283
15284
15285
15286
15287
15288
15289
15290
15291
15292
15293
15294
15295
15296
15297
15298
15299
15300
15301
15302
15303
15304
15305
15306
15307
15308
15309
15310
15311
15312
15313
15314
15315
15316
15317
15318
15319
15320
15321
15322
15323
15324
15325
15326
15327
15328
15329
15330
15331
15332
15333
15334
15335
15336
15337
15338
15339
15340
15341
15342
15343
15344
15345
15346
15347
15348
15349
15350
15351
15352
15353
15354
15355
15356
15357
15358
15359
15360
15361
15362
15363
15364
15365
15366
15367
15368
15369
15370
15371
15372
15373
15374
15375
15376
15377
15378
15379
15380
15381
15382
15383
15384
15385
15386
15387
15388
15389
15390
15391
15392
15393
15394
15395
15396
15397
15398
15399
15400
15401
15402
15403
15404
15405
15406
15407
15408
15409
15410
15411
15412
15413
15414
15415
15416
15417
15418
15419
15420
15421
15422
15423
15424
15425
15426
15427
15428
15429
15430
15431
15432
15433
15434
15435
15436
15437
15438
15439
15440
15441
15442
15443
15444
15445
15446
15447
15448
15449
15450
15451
15452
15453
15454
15455
15456
15457
15458
15459
15460
15461
15462
15463
15464
15465
15466
15467
15468
15469
15470
15471
15472
15473
15474
15475
15476
15477
15478
15479
15480
15481
15482
15483
15484
15485
15486
15487
15488
15489
15490
15491
15492
15493
15494
15495
15496
15497
15498
15499
15500
15501
15502
15503
15504
15505
15506
15507
15508
15509
15510
15511
15512
15513
15514
15515
15516
15517
15518
15519
15520
15521
15522
15523
15524
15525
15526
15527
15528
15529
15530
15531
15532
15533
15534
15535
15536
15537
15538
15539
15540
15541
15542
15543
15544
15545
15546
15547
15548
15549
15550
15551
15552
15553
15554
15555
15556
15557
15558
15559
15560
15561
15562
15563
15564
15565
15566
15567
15568
15569
15570
15571
15572
15573
15574
15575
15576
15577
15578
15579
15580
15581
15582
15583
15584
15585
15586
15587
15588
15589
15590
15591
15592
15593
15594
15595
15596
15597
15598
15599
15600
15601
15602
15603
15604
15605
15606
15607
15608
15609
15610
15611
15612
15613
15614
15615
15616
15617
15618
15619
15620
15621
15622
15623
15624
15625
15626
15627
15628
15629
15630
15631
15632
15633
15634
15635
15636
15637
15638
15639
15640
15641
15642
15643
15644
15645
15646
15647
15648
15649
15650
15651
15652
15653
15654
15655
15656
15657
15658
15659
15660
15661
15662
15663
15664
15665
15666
15667
15668
15669
15670
15671
15672
15673
15674
15675
15676
15677
15678
15679
15680
15681
15682
15683
15684
15685
15686
15687
15688
15689
15690
15691
15692
15693
15694
15695
15696
15697
15698
15699
15700
15701
15702
15703
15704
15705
15706
15707
15708
15709
15710
15711
15712
15713
15714
15715
15716
15717
15718
15719
15720
15721
15722
15723
15724
15725
15726
15727
15728
15729
15730
15731
15732
15733
15734
15735
15736
15737
15738
15739
15740
15741
15742
15743
15744
15745
15746
15747
15748
15749
15750
15751
15752
15753
15754
15755
15756
15757
15758
15759
15760
15761
15762
15763
15764
15765
15766
15767
15768
15769
15770
15771
15772
15773
15774
15775
15776
15777
15778
15779
15780
15781
15782
15783
15784
15785
15786
15787
15788
15789
15790
15791
15792
15793
15794
15795
15796
15797
15798
15799
15800
15801
15802
15803
15804
15805
15806
15807
15808
15809
15810
15811
15812
15813
15814
15815
15816
15817
15818
15819
15820
15821
15822
15823
15824
15825
15826
15827
15828
15829
15830
15831
15832
15833
15834
15835
15836
15837
15838
15839
15840
15841
15842
15843
15844
15845
15846
15847
15848
15849
15850
15851
15852
15853
15854
15855
15856
15857
15858
15859
15860
15861
15862
15863
15864
15865
15866
15867
15868
15869
15870
15871
15872
15873
15874
15875
15876
15877
15878
15879
15880
15881
15882
15883
15884
15885
15886
15887
15888
15889
15890
15891
15892
15893
15894
15895
15896
15897
15898
15899
15900
15901
15902
15903
15904
15905
15906
15907
15908
15909
15910
15911
15912
15913
15914
15915
15916
15917
15918
15919
15920
15921
15922
15923
15924
15925
15926
15927
15928
15929
15930
15931
15932
15933
15934
15935
15936
15937
15938
15939
15940
15941
15942
15943
15944
15945
15946
15947
15948
15949
15950
15951
15952
15953
15954
15955
15956
15957
15958
15959
15960
15961
15962
15963
15964
15965
15966
15967
15968
15969
15970
15971
15972
15973
15974
15975
15976
15977
15978
15979
15980
15981
15982
15983
15984
15985
15986
15987
15988
15989
15990
15991
15992
15993
15994
15995
15996
15997
15998
15999
16000
16001
16002
16003
16004
16005
16006
16007
16008
16009
16010
16011
16012
16013
16014
16015
16016
16017
16018
16019
16020
16021
16022
16023
16024
16025
16026
16027
16028
16029
16030
16031
16032
16033
16034
16035
16036
// Copyright 2024 RustFS Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

//! `SetDisks` — one erasure set's worth of disks and the object storage logic
//! over them. Historically a single ~19.7k-line God-Object; split into a Core
//! plus borrow-based operation families during backlog#815 (Epic #728).
//!
//! Module layout after the split:
//!
//! - `mod.rs` — the `SetDisks` core struct, its construction, and the shared
//!   inherent state/helpers the operation families borrow.
//! - `ctx` — `SetDisksCtx`, the borrow context that hands operation units
//!   access to the core without cloning `Arc`s (P0, backlog#816).
//! - `core::io_primitives` — the shared low-level read/write/erasure IO
//!   primitives (metadata-fanout quorum, bitrot readers, rename/delete, quorum
//!   helpers) that the operation families call through the core (P5,
//!   backlog#820).
//! - `ops/` — one module per storage-api contract family, each
//!   `impl <Contract> for SetDisks`: `ops::object` (`ObjectIO` +
//!   `ObjectOperations`, the object read/write hot path, P6/backlog#821),
//!   `ops::heal` (`HealOperations`, P1/backlog#817), `ops::multipart`
//!   (`MultipartOperations`, P2/backlog#818), `ops::list` (`ListOperations`)
//!   and `ops::bucket` (`BucketOperations`, P3+P4/backlog#819), `ops::locking`
//!   (`NamespaceLocking` + lock helpers, P7/backlog#822).
//! - `read.rs` — the object-read operation pipeline (`get_object_*`,
//!   `read_version_optimized`) and its metadata cache, kept separate from the
//!   read primitives it drives.
//! - `metadata.rs`, `replication.rs`, `shard_source.rs` — supporting helpers.

use crate::bucket::lifecycle::lifecycle::TRANSITION_COMPLETE;
use crate::bucket::metadata_sys;
use crate::bucket::metadata_sys::ObjectLockConfigState;
use crate::bucket::object_lock::objectlock_sys::{
    check_object_lock_for_deletion_with_state, check_retention_for_modification, replication_delete_may_bypass_governance,
    replication_write_may_pass_worm_gate,
};
#[cfg(test)]
use crate::bucket::replication::ReplicationState;
use crate::bucket::replication::{ReplicateDecision, ReplicationObjectBridge, ReplicationStatusType, VersionPurgeStatusType};
use crate::cluster::rpc::heal_bucket_local_on_disks;
use crate::data_usage::record_compression_total_memory;
use crate::diagnostics::get::{
    GET_CODEC_STREAMING_OBJECT_CLASS_PLAIN_SINGLE_PART, GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_GEOMETRY,
    GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_IDENTITY_MISMATCH,
    GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_MISSING_PAYLOAD,
    GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_MISSING_SHARD, GET_OBJECT_PATH_BODY_CACHE, GET_OBJECT_PATH_CODEC_STREAMING,
    GET_OBJECT_PATH_CODEC_STREAMING_LEGACY_ENGINE, GET_OBJECT_PATH_CODEC_STREAMING_RUSTFS_ENGINE, GET_OBJECT_PATH_DIRECT_MEMORY,
    GET_OBJECT_PATH_EMPTY, GET_OBJECT_PATH_INLINE_DIRECT, GET_OBJECT_PATH_INTERNAL_META, GET_OBJECT_PATH_LEGACY_DUPLEX,
    GET_OBJECT_PATH_REMOTE_TRANSITION, GET_OBJECT_PATH_SET_DISK, GET_STAGE_DECODE, GET_STAGE_EMIT, GET_STAGE_INLINE_PREPARE,
    GET_STAGE_LOCK_ACQUIRE, GET_STAGE_METADATA, GET_STAGE_OBJECT_INFO, GET_STAGE_PATH_DECISION, GET_STAGE_READER_SETUP,
    classify_storage_error, get_stage_timer_if_enabled, record_get_object_pipeline_failure,
    record_get_object_pipeline_failure_for_path, record_get_stage_duration_if_enabled,
};
use crate::diagnostics::object_lock::ObjectLockAttempt;
use crate::disk::error_reduce::{
    BUCKET_OP_IGNORED_ERRS, OBJECT_OP_IGNORED_ERRS, build_write_quorum_failure_summary, count_errs, reduce_read_quorum_errs,
    reduce_write_quorum_errs,
};
#[cfg(test)]
use crate::disk::has_part_err;
use crate::disk::{
    self, CHECK_PART_DISK_NOT_FOUND, CHECK_PART_FILE_CORRUPT, CHECK_PART_FILE_NOT_FOUND, CHECK_PART_SUCCESS, CHECK_PART_UNKNOWN,
    conv_part_err_to_int,
};
use crate::disk::{STORAGE_FORMAT_FILE, count_part_not_success};
use crate::erasure::codec::bridge::{
    CodecStreamingDecodeEngine, GET_CODEC_STREAMING_ENGINE_LEGACY, GET_CODEC_STREAMING_ENGINE_RUSTFS,
};
use crate::erasure::coding;
use crate::error::{Error, Result, is_err_version_not_found};
use crate::error::{GenericError, ObjectApiError, is_err_object_not_found};
use crate::io_support::bitrot::{create_bitrot_reader, create_bitrot_reader_from_bytes, create_bitrot_writer};
use crate::object_api::ObjectOptions;
use crate::object_api::get_object_body_cache_hook;
use crate::object_api::object_api_utils::get_raw_etag;
use crate::runtime::instance::{InstanceContext, bootstrap_ctx};
use crate::runtime::sources as runtime_sources;
#[cfg(test)]
use crate::storage_api_contracts::multipart::MultipartOperations;
use crate::storage_api_contracts::{
    bucket::{BucketInfo, BucketOperations, BucketOptions, DeleteBucketOptions, MakeBucketOptions},
    list::{StorageListObjectVersionsInfo, StorageListObjectsV2Info, StorageObjectInfoOrErr, StorageWalkOptions},
    multipart::{CompletePart, ListMultipartsInfo, ListPartsInfo, MultipartInfo, MultipartUploadResult, PartInfo},
    namespace::NamespaceLocking as _,
    object::{DeleteAccounting, DeletedObject, HTTPPreconditions, ObjectIO as _, ObjectToDelete},
    range::HTTPRangeSpec,
};
use crate::store::utils::is_reserved_or_invalid_bucket;
use crate::{
    bucket::lifecycle::bucket_lifecycle_ops::{LifecycleOps, get_transitioned_object_reader_with_tier_manager, put_restore_opts},
    cache_value::metacache_set::{ListPathRawOptions, list_path_raw},
    config::storageclass,
    disk::{
        CheckPartsResp, DeleteOptions, DiskAPI, DiskInfo, DiskInfoOptions, DiskOption, DiskStore, FileInfoVersions,
        RUSTFS_META_BUCKET, RUSTFS_META_MULTIPART_BUCKET, RUSTFS_META_TMP_BUCKET, ReadMultipleReq, ReadMultipleResp, ReadOptions,
        SnapshotLeaseToken, UpdateMetadataOpts, endpoint::Endpoint, error::DiskError, format::FormatV3,
    },
    error::{StorageError, to_object_err},
    object_api::{GetObjectReader, NamespaceLockFence, ObjectInfo, ObjectLockConfigSnapshot, PutObjReader},
    // event::name::EventName,
    services::event_notification::{EventArgs, send_event},
    store::init_format::{
        formats_match_reference_slots, get_format_erasure_in_quorum, load_format_erasure, load_format_erasure_all,
        save_format_file,
    },
};
use bytes::Bytes;
use bytesize::ByteSize;
use futures::future::join_all;
use http::HeaderMap;
use md5::{Digest as Md5Digest, Md5};
use regex::Regex;
use rustfs_config::MI_B;
use rustfs_filemeta::metadata_keys;
use rustfs_filemeta::{
    FileInfo, FileMeta, FileMetaShallowVersion, MetaCacheEntries, MetaCacheEntry, ObjectPartInfo, RawFileInfo,
    merge_file_meta_versions,
};
use rustfs_heal_contracts::heal_channel::{
    DriveState, HealAdmissionResult, HealChannelPriority, HealItemType, HealOpts, HealRequestSource, HealScanMode,
    send_heal_replacement_disk, send_heal_request_with_admission,
};
use rustfs_io_metrics::{
    record_object_lock_diag_acquire_duration, record_object_lock_diag_enabled, record_object_lock_diag_hold_duration,
    record_object_lock_diag_slow_acquire, record_object_lock_diag_slow_hold,
};
use rustfs_lock::LockClient;
#[cfg(test)]
use rustfs_lock::LockManager;
use rustfs_lock::fast_lock::types::LockResult;
use rustfs_lock::{FastLockGuard, NamespaceLock, NamespaceLockGuard, NamespaceLockWrapper, ObjectKey};
use rustfs_madmin::heal_commands::{HealDriveInfo, HealResultItem, Infos};
use rustfs_object_capacity::capacity_scope::{
    CapacityScope, CapacityScopeDisk, current_dirty_generation, record_capacity_scope, record_global_dirty_scope,
};
use rustfs_s3_client::transition_api::{ObjectReader, ReaderImpl};
use rustfs_s3_types::EventName;
#[cfg(test)]
use rustfs_utils::http::SSEC_ALGORITHM_HEADER;
use rustfs_utils::http::headers::AMZ_OBJECT_TAGGING;
use rustfs_utils::http::headers::{
    CACHE_CONTROL, CONTENT_DISPOSITION, CONTENT_ENCODING, CONTENT_LANGUAGE, CONTENT_TYPE, EXPIRES,
};
use rustfs_utils::http::{
    SUFFIX_ACTUAL_OBJECT_SIZE_CAP, SUFFIX_ACTUAL_SIZE, SUFFIX_BUCKET_INCARNATION_ID, SUFFIX_COMPRESSION, SUFFIX_COMPRESSION_SIZE,
    SUFFIX_REPLICATION_SSEC_CRC, SUFFIX_RESTORE_OPERATION_ID, SUFFIX_RESTORE_WORKER_LOCK, contains_key_str, get_header_map,
    get_str, insert_str, is_object_encryption_marker, remove_header_map,
};
use rustfs_utils::{
    HashAlgorithm,
    crypto::hex,
    path::{SLASH_SEPARATOR, encode_dir_object, encode_dir_object_ref, has_suffix, path_join_buf},
};
use sha2::Sha256;
use std::hash::{BuildHasher, Hash, Hasher};
use std::mem::{self};
use std::pin::Pin;
use std::sync::atomic::{AtomicU64, Ordering};
use std::sync::{Arc, OnceLock};
use std::task::{Context, Poll};
use std::time::{Instant, SystemTime, UNIX_EPOCH};
use std::{
    collections::{HashMap, HashSet},
    io::Cursor,
    path::Path,
    time::Duration,
};
use time::OffsetDateTime;
#[cfg(test)]
use tokio::sync::mpsc;
use tokio::sync::mpsc::Sender;
#[cfg(test)]
use tokio::time::timeout;
use tokio::{
    io::{AsyncRead, AsyncWrite, BufReader, ReadBuf},
    sync::RwLock,
};
use tokio_util::sync::CancellationToken;
use tracing::error;
use tracing::{Instrument, debug, info, warn};
use uuid::Uuid;

pub(super) fn restore_operation_id_from_metadata(metadata: &HashMap<String, String>) -> Result<Option<Uuid>> {
    let Some(value) = rustfs_utils::http::metadata_compat::get_consistent_str(metadata, SUFFIX_RESTORE_OPERATION_ID) else {
        if rustfs_utils::http::metadata_compat::contains_key_str(metadata, SUFFIX_RESTORE_OPERATION_ID) {
            return Err(Error::other("invalid restore operation id metadata".to_string()));
        }
        return Ok(None);
    };
    let id = Uuid::parse_str(value).map_err(|_| Error::other("invalid restore operation id metadata".to_string()))?;
    if id.is_nil() {
        return Err(Error::other("invalid restore operation id metadata".to_string()));
    }
    Ok(Some(id))
}

pub(super) fn require_restore_operation_id(metadata: &HashMap<String, String>, expected: Uuid) -> Result<()> {
    match restore_operation_id_from_metadata(metadata)? {
        Some(actual) if actual == expected => Ok(()),
        _ => Err(Error::other("restore operation id changed before copy-back".to_string())),
    }
}

pub(super) fn restore_commit_operation_id_from_metadata(metadata: &HashMap<String, String>) -> Result<Option<Uuid>> {
    if !metadata.contains_key(metadata_keys::RESTORE) {
        return Ok(None);
    }
    restore_operation_id_from_metadata(metadata)
}

async fn inspect_decommission_tier_free_version_target(
    disk: &DiskStore,
    bucket: &str,
    object: &str,
    source: &FileInfo,
) -> Result<bool> {
    let raw = match disk.read_xl(bucket, object, false).await {
        Ok(raw) => raw,
        Err(DiskError::FileNotFound | DiskError::FileVersionNotFound | DiskError::VolumeNotFound) => return Ok(false),
        Err(err) => return Err(err.into()),
    };
    let meta = FileMeta::load(&raw.buf)?;
    let source_version_id = source.version_id.filter(|version_id| !version_id.is_nil());
    let mut matching_count = 0;
    let mut all_matching_versions_equivalent = true;
    for existing in meta
        .versions
        .iter()
        .filter(|version| version.header.version_id.filter(|version_id| !version_id.is_nil()) == source_version_id)
    {
        matching_count += 1;
        let existing = existing.into_fileinfo(bucket, object, true)?;
        existing.validate_for_metadata_read()?;
        if !existing.tier_free_version() || !crate::store::tiered_data_movement_source_matches(source, &existing)? {
            all_matching_versions_equivalent = false;
        }
    }
    if matching_count == 0 {
        return Ok(false);
    }
    if matching_count == 1 && all_matching_versions_equivalent {
        return Ok(true);
    }

    Err(StorageError::DataMovementOverwriteErr(
        bucket.to_owned(),
        object.to_owned(),
        source_version_id.map(|version_id| version_id.to_string()).unwrap_or_default(),
    ))
}

fn ensure_decommission_tier_free_version_commit_fence(bucket: &str, object: &str, opts: &ObjectOptions) -> Result<()> {
    if opts
        .namespace_lock_fence
        .as_ref()
        .is_some_and(NamespaceLockFence::is_lock_lost)
        || opts
            .bucket_lifecycle_lock_fence
            .as_ref()
            .is_some_and(NamespaceLockFence::is_lock_lost)
    {
        return Err(StorageError::NamespaceLockQuorumUnavailable {
            mode: "decommission_tier_free_version_commit",
            bucket: bucket.to_string(),
            object: object.to_string(),
            required: 1,
            achieved: 0,
        });
    }

    Ok(())
}

impl SetDisks {
    pub(super) async fn require_current_restore_operation_id(
        &self,
        bucket: &str,
        object: &str,
        opts: &ObjectOptions,
        expected: Option<Uuid>,
        mode: &str,
    ) -> Result<()> {
        let Some(expected) = expected else {
            return Ok(());
        };
        let read_opts = ObjectOptions {
            version_id: opts.version_id.clone(),
            versioned: opts.versioned,
            version_suspended: opts.version_suspended,
            no_lock: true,
            ..Default::default()
        };
        let current = self.get_object_fileinfo(bucket, object, &read_opts, true, false).await?;
        restore_operation_id_from_metadata(&current.fi().metadata)?
            .filter(|actual| *actual == expected)
            .ok_or_else(|| Error::other(format!("restore operation id changed before {mode}: expected {expected}")))?;
        Ok(())
    }
}

type ListObjectsV2Info = StorageListObjectsV2Info<ObjectInfo>;
type ListObjectVersionsInfo = StorageListObjectVersionsInfo<ObjectInfo>;
type ObjectInfoOrErr = StorageObjectInfoOrErr<ObjectInfo, Error>;
type WalkOptions = StorageWalkOptions<fn(&FileInfo) -> bool>;
type InlineBitrotReader = coding::BitrotReader<crate::io_support::bitrot::ShardReader>;

const LOG_COMPONENT_ECSTORE: &str = "ecstore";
const LOG_SUBSYSTEM_SET_DISK: &str = "set_disk";
const EVENT_SET_DISK_MULTIPART: &str = "set_disk_multipart";
const COMPLETE_MULTIPART_PART_MISSING: &str = "part_missing";
const COMPLETE_MULTIPART_PART_READ_QUORUM_UNAVAILABLE: &str = "read_quorum_unavailable";
const COMPLETE_MULTIPART_PART_ERROR: &str = "part_error";
const MULTIPART_WRITE_QUORUM_UPLOAD_METADATA: &str = "upload_metadata";
const MULTIPART_WRITE_QUORUM_WRITER_SETUP: &str = "writer_setup";
const MULTIPART_WRITE_QUORUM_RENAME_PART: &str = "rename_part";
const EVENT_SET_DISK_WRITE: &str = "set_disk_write";
const EVENT_SET_DISK_HEAL: &str = "set_disk_heal";
const EVENT_SET_DISK_COMMIT_TAIL_SLOW: &str = "set_disk_commit_tail_slow";
const EVENT_SET_DISK_RENAME_TAIL_DRAIN_FAILED: &str = "set_disk_rename_tail_drain_failed";
const EVENT_SET_DISK_PUT_OBJECT_STAGE_SUMMARY: &str = "set_disk_put_object_stage_summary";
const EVENT_SET_DISK_ORPHAN_PURGE_SKIPPED: &str = "set_disk_orphan_purge_skipped";
const SET_DISK_COMMIT_TAIL_WARN_THRESHOLD_MS: u128 = 5_000;
const ENV_RUSTFS_PUT_LARGE_BATCH_MIN_SIZE_BYTES: &str = "RUSTFS_PUT_LARGE_BATCH_MIN_SIZE_BYTES";
const DEFAULT_RUSTFS_PUT_LARGE_BATCH_MIN_SIZE_BYTES: usize = 64 * 1024 * 1024;
static CACHED_PUT_LARGE_BATCH_MIN_SIZE_BYTES: OnceLock<usize> = OnceLock::new();
const ENV_RUSTFS_MULTIPART_PUT_LARGE_BATCH_MIN_SIZE_BYTES: &str = "RUSTFS_MULTIPART_PUT_LARGE_BATCH_MIN_SIZE_BYTES";
const DEFAULT_RUSTFS_MULTIPART_PUT_LARGE_BATCH_MIN_SIZE_BYTES: usize = 128 * 1024 * 1024;
static CACHED_MULTIPART_PUT_LARGE_BATCH_MIN_SIZE_BYTES: OnceLock<usize> = OnceLock::new();

use crate::io_support::rio::HashReader;

pub const DEFAULT_READ_BUFFER_SIZE: usize = MI_B; // 1 MiB = 1024 * 1024;
pub const MAX_PARTS_COUNT: usize = 10000;
pub(crate) const RUSTFS_MULTIPART_BUCKET_KEY: &str = "x-rustfs-internal-multipart-bucket";
pub(crate) const RUSTFS_MULTIPART_OBJECT_KEY: &str = "x-rustfs-internal-multipart-object";
pub(crate) const DATA_MOVEMENT_MULTIPART_PREFIX: &str = "data-movement";
const ENV_ISSUE3031_DIAG_ENABLE: &str = "RUSTFS_ISSUE3031_DIAG_ENABLE";

/// Validate disk metadata at a boundary that may legitimately return a delete
/// marker. Disk/RPC decode boundaries perform the full collection validation
/// once; repeated quorum passes use the cheap erasure-geometry predicate for
/// payload entries and the canonical marker predicate for pure delete markers.
pub(in crate::set_disk) fn file_info_is_valid_for_metadata(file_info: &FileInfo) -> bool {
    file_info.has_valid_metadata_shape()
}

struct ObjectLockDiagGuard {
    guard: NamespaceLockGuard,
    enabled: bool,
    op: &'static str,
    bucket: Option<String>,
    object: Option<String>,
    owner: Option<String>,
    mode: &'static str,
    acquired_at: Instant,
    _attempt: ObjectLockAttempt,
}

impl ObjectLockDiagGuard {
    fn new(
        guard: NamespaceLockGuard,
        enabled: bool,
        op: &'static str,
        bucket: Option<String>,
        object: Option<String>,
        owner: Option<String>,
        mode: &'static str,
    ) -> Self {
        Self {
            guard,
            enabled,
            op,
            bucket,
            object,
            owner,
            mode,
            acquired_at: Instant::now(),
            _attempt: ObjectLockAttempt::default(),
        }
    }

    fn with_attempt(mut self, attempt: ObjectLockAttempt) -> Self {
        self._attempt = attempt;
        self
    }

    /// Whether the underlying namespace lock's heartbeat has observed a
    /// refresh-quorum loss (backlog#899 Phase 2). Callers fence their commit
    /// point on this so a stale lock holder does not race a double-write.
    fn is_lock_lost(&self) -> bool {
        self.guard.is_lock_lost()
    }
}

impl Drop for ObjectLockDiagGuard {
    fn drop(&mut self) {
        if !self.enabled || self.guard.is_released() {
            return;
        }

        let hold = self.acquired_at.elapsed();
        record_object_lock_diag_hold_duration(self.op, self.mode, hold);
        let threshold = get_object_lock_diag_slow_hold_threshold();
        if hold >= threshold {
            record_object_lock_diag_slow_hold(self.op, self.mode);
            warn!(
                target: "rustfs_ecstore::object_lock_diag",
                op = self.op,
                bucket = %self.bucket.as_deref().unwrap_or_default(),
                object = %self.object.as_deref().unwrap_or_default(),
                mode = self.mode,
                owner = %self.owner.as_deref().unwrap_or_default(),
                hold_ms = hold.as_millis(),
                threshold_ms = threshold.as_millis(),
                "object namespace lock held longer than threshold"
            );
        }
    }
}

struct SetDiskLockGuardedReader {
    inner: Box<dyn AsyncRead + Unpin + Send + Sync>,
    guard: Option<ObjectLockDiagGuard>,
}

impl AsyncRead for SetDiskLockGuardedReader {
    fn poll_read(mut self: Pin<&mut Self>, cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll<std::io::Result<()>> {
        let had_capacity = buf.remaining() > 0;
        let filled_before = buf.filled().len();
        let poll = Pin::new(&mut self.inner).poll_read(cx, buf);
        if had_capacity && matches!(poll, Poll::Ready(Ok(()))) && buf.filled().len() == filled_before {
            self.guard.take();
        }
        poll
    }
}

fn finish_set_disk_read_lock(
    mut reader: GetObjectReader,
    read_lock_guard: Option<ObjectLockDiagGuard>,
    bucket: &str,
    object: &str,
) -> GetObjectReader {
    if reader.buffered_body.is_some() {
        release_materialized_read_lock(bucket, object, read_lock_guard);
        return reader;
    }

    if let Some(guard) = read_lock_guard {
        reader.stream = Box::new(SetDiskLockGuardedReader {
            inner: reader.stream,
            guard: Some(guard),
        });
    }
    reader
}

fn release_materialized_read_lock(bucket: &str, object: &str, read_lock_guard: Option<ObjectLockDiagGuard>) {
    if read_lock_guard.is_some() {
        let lock_id = format!("{}:{}", bucket, object);
        record_lock_release(bucket, object, &lock_id, "read");
        metrics::counter!("rustfs.lock.release.early.total", "type" => "read").increment(1);
    }
    drop(read_lock_guard);
}

pub(crate) fn strip_internal_multipart_metadata(metadata: &mut HashMap<String, String>) {
    metadata.remove(RUSTFS_MULTIPART_BUCKET_KEY);
    metadata.remove(RUSTFS_MULTIPART_OBJECT_KEY);
    rustfs_utils::http::metadata_compat::remove_str(metadata, SUFFIX_BUCKET_INCARNATION_ID);
}

fn should_persist_encryption_original_size(metadata: &HashMap<String, String>) -> bool {
    metadata.keys().any(|key| is_object_encryption_marker(key))
}

/// Per-set memoized capacity dirty scope.
///
/// The disk endpoint/path identity of each slot in a set is immutable for the
/// set's lifetime (heal replaces the [`DiskStore`] instance but keeps the same
/// endpoint and root), so the dirty scope only needs to be built once instead
/// of allocating a `String` per online disk on every successful write
/// (backlog#1315). Slots are filled lazily as disks are observed online; once
/// every slot has contributed, `complete` latches and the hot path returns the
/// shared `Arc` without any scan.
#[derive(Default, Debug)]
struct CapacityScopeCache {
    /// Per-slot resolved disk identity; `None` slots have not been seen online.
    slots: Vec<Option<CapacityScopeDisk>>,
    /// Deduplicated scope covering every slot resolved so far.
    scope: Option<Arc<CapacityScope>>,
    /// Latches once every slot is resolved; the scope is then stable.
    complete: bool,
}

impl CapacityScopeCache {
    /// Whether `disks` contains an online disk whose slot has not been recorded
    /// yet. Read-only, allocation-free (used under the read lock).
    fn has_unresolved_slot(&self, disks: &[Option<DiskStore>]) -> bool {
        disks
            .iter()
            .enumerate()
            .any(|(idx, disk)| disk.is_some() && self.slots.get(idx).is_none_or(|slot| slot.is_none()))
    }
}

impl SetDisks {
    /// Return the memoized dirty scope for this set, resolving any newly-online
    /// slots first. Steady-state writes hit the fast path (a read lock and an
    /// `Arc` clone) with no per-disk `String` allocation.
    fn capacity_scope(&self, disks: &[Option<DiskStore>]) -> Arc<CapacityScope> {
        {
            let cache = self.capacity_scope_cache.read().unwrap_or_else(|p| p.into_inner());
            if let Some(scope) = cache.scope.as_ref()
                && (cache.complete || !cache.has_unresolved_slot(disks))
            {
                return scope.clone();
            }
        }
        self.resolve_capacity_scope(disks)
    }

    /// Slow path: fill newly-observed slots and rebuild the deduplicated scope.
    fn resolve_capacity_scope(&self, disks: &[Option<DiskStore>]) -> Arc<CapacityScope> {
        let mut cache = self.capacity_scope_cache.write().unwrap_or_else(|p| p.into_inner());
        if cache.slots.len() < self.set_drive_count {
            cache.slots.resize(self.set_drive_count, None);
        }

        let mut changed = false;
        for (idx, disk) in disks.iter().enumerate() {
            let Some(slot) = cache.slots.get_mut(idx) else {
                break;
            };
            if slot.is_some() {
                continue;
            }
            if let Some(disk) = disk {
                *slot = Some(CapacityScopeDisk {
                    endpoint: disk.endpoint().to_string(),
                    drive_path: disk.to_string(),
                });
                changed = true;
            }
        }

        if changed || cache.scope.is_none() {
            let mut unique = HashSet::with_capacity(cache.slots.len());
            let mut scoped_disks = Vec::with_capacity(cache.slots.len());
            for slot in cache.slots.iter().flatten() {
                if unique.insert(slot.clone()) {
                    scoped_disks.push(slot.clone());
                }
            }
            cache.complete = !cache.slots.is_empty() && cache.slots.iter().all(|slot| slot.is_some());
            cache.scope = Some(Arc::new(CapacityScope { disks: scoped_disks }));
        }

        cache.scope.clone().unwrap_or_else(|| Arc::new(CapacityScope::default()))
    }

    /// Record the set's dirty scope after a successful write.
    ///
    /// The global dirty registry is only upgraded on the first write of each
    /// registry generation: once this set has marked its disks, subsequent
    /// writes skip the global mutex until a refresh drain advances the
    /// generation, forcing a re-mark (backlog#1315). The scope-token registry
    /// (multipart completion) still records per token so the app-side write
    /// settle can consume it.
    fn record_capacity_scope_if_needed(&self, scope_token: Option<Uuid>, disks: &[Option<DiskStore>]) {
        let scope = self.capacity_scope(disks);
        if scope.disks.is_empty() {
            return;
        }

        let generation = current_dirty_generation();
        if self.capacity_dirty_generation.load(Ordering::Acquire) != generation {
            // First write of this generation (or after a drain): upgrade the
            // global registry and cache the generation observed under its lock.
            let observed = record_global_dirty_scope((*scope).clone());
            self.capacity_dirty_generation.store(observed, Ordering::Release);
        }

        if let Some(token) = scope_token {
            record_capacity_scope(token, (*scope).clone());
        }
    }

    /// Mark healed disks dirty from the (infrequent) heal path.
    ///
    /// Heal passes disks in erasure-distribution order (`shuffle_disks`), not
    /// physical-slot order, so they must not seed the slot-indexed memo — doing
    /// so could record a disk under the wrong slot and drop another from the
    /// steady-state scope. Heal therefore builds an ad-hoc scope from the disks
    /// it actually rewrote and marks the global registry directly. This runs at
    /// heal frequency, so it does not use the per-generation skip fast-path
    /// (backlog#1315).
    fn record_healed_capacity_scope(&self, disks: &[Option<DiskStore>]) {
        let scope = capacity_scope_from_disks(disks);
        if scope.disks.is_empty() {
            return;
        }
        // Do not advance the set's generation marker here: this ad-hoc scope is
        // only the subset of disks heal rewrote, whereas the marker asserts the
        // full set was marked. Leaving the marker untouched keeps the next write
        // free to upgrade the full-set scope if it has not been marked yet.
        let _ = record_global_dirty_scope(scope);
    }
}

/// Build an ad-hoc, deduplicated dirty scope from `disks`. Used where the
/// per-set generation fast path is intentionally bypassed: heal rewrites and
/// an early-ACK tail whose first mark was drained before it completed.
fn capacity_scope_from_disks(disks: &[Option<DiskStore>]) -> CapacityScope {
    let mut unique = HashSet::with_capacity(disks.len());
    let mut scoped_disks = Vec::with_capacity(disks.len());
    for disk in disks.iter().flatten() {
        let scope_disk = CapacityScopeDisk {
            endpoint: disk.endpoint().to_string(),
            drive_path: disk.to_string(),
        };
        if unique.insert(scope_disk.clone()) {
            scoped_disks.push(scope_disk);
        }
    }
    CapacityScope { disks: scoped_disks }
}

/// Get the duplex buffer size from environment variable or use default.
///
/// This function reads `RUSTFS_DUPLEX_BUFFER_SIZE` environment variable
/// to allow runtime configuration of the duplex pipe buffer size.
/// A larger buffer (e.g., 4MB) helps prevent backpressure-related hangs
/// when reading large objects (20-26MB) under high concurrency.
///
/// Default: 4MB (4 * 1024 * 1024 bytes)
/// Get duplex buffer size from environment variable.
///
/// **Deprecated**: Use `adaptive_duplex_buffer_size()` for object-size-aware sizing.
pub fn get_duplex_buffer_size() -> usize {
    static CACHED: OnceLock<usize> = OnceLock::new();
    *CACHED.get_or_init(|| {
        rustfs_utils::get_env_usize(
            rustfs_config::ENV_OBJECT_DUPLEX_BUFFER_SIZE,
            rustfs_config::DEFAULT_OBJECT_DUPLEX_BUFFER_SIZE,
        )
        .max(1)
    })
}

/// Get adaptive duplex buffer size based on object size.
///
/// Smaller objects get smaller buffers to reduce memory waste.
/// Larger objects get larger buffers to prevent backpressure.
fn adaptive_duplex_buffer_size(object_size: i64) -> usize {
    const KB: usize = 1024;
    const MB: usize = 1024 * 1024;
    let target = match object_size {
        0..=131_072 => 64 * KB,             // <= 128KB: 64KB
        131_073..=1_048_576 => 512 * KB,    // <= 1MB: reduce duplex backpressure without a 1MB pipe per request
        1_048_577..=16_777_216 => MB,       // <= 16MB: 1MB
        16_777_217..=268_435_456 => 4 * MB, // <= 256MB: 4MB
        _ => 8 * MB,                        // > 256MB: 8MB
    };
    let object_cap = usize::try_from(object_size).ok().filter(|size| *size > 0).unwrap_or(target);
    target.min(object_cap.max(64 * KB)).min(get_duplex_buffer_size())
}

// ============================================================================
// GET Optimization Configuration
//
// All GET performance optimization flags are consolidated here.
// Each flag uses `OnceLock` for caching — env var changes require process restart.
// Each flag has a corresponding `*_ROLLOUT_PCT` for percentage-based gradual rollout.
// ============================================================================

#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
const DISK_ONLINE_TIMEOUT: Duration = Duration::from_secs(1);
#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
const DISK_HEALTH_CACHE_TTL: Duration = Duration::from_millis(750);
const GET_OBJECT_METADATA_CACHE_TTL: Duration = Duration::from_secs(2); // Increased from 250ms to 2s
const DEFAULT_GET_OBJECT_METADATA_CACHE_MAX_ENTRIES: usize = 4096; // Increased from 1024 to 4096
const ENV_RUSTFS_GET_OBJECT_METADATA_CACHE_MAX_ENTRIES: &str = "RUSTFS_GET_OBJECT_METADATA_CACHE_MAX_ENTRIES";
const GET_OBJECT_METADATA_CACHE_FENCE_SHARDS: u16 = 4096;

// --- Codec Streaming Configuration ---

const ENV_RUSTFS_GET_CODEC_STREAMING_ENABLE: &str = "RUSTFS_GET_CODEC_STREAMING_ENABLE";
// Emergency kill-switch, not the primary enablement knob. The single switch that
// turns codec streaming on/off is `RUSTFS_GET_CODEC_STREAMING_ROLLOUT` (default
// `off`). This flag defaults to `true` and only exists to force-disable the fast
// path regardless of rollout (set to `false`). Body/header compatibility is
// confirmed by the GET codec-streaming parity e2e net + bench A/B (backlog#1183),
// so it no longer gates enablement. See `get_codec_streaming_reader_gate`.
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_ENABLE: bool = true;

const ENV_RUSTFS_GET_CODEC_STREAMING_MIN_SIZE: &str = "RUSTFS_GET_CODEC_STREAMING_MIN_SIZE";
// Meet the direct-memory path at its default ceiling. Codec streaming remains
// rollout-gated and starts where the eager small-object path ends.
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_MIN_SIZE: usize = DEFAULT_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY_THRESHOLD;
const ENV_RUSTFS_GET_CODEC_STREAMING_RUSTFS_MIN_SIZE: &str = "RUSTFS_GET_CODEC_STREAMING_RUSTFS_MIN_SIZE";
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_RUSTFS_MIN_SIZE: usize = DEFAULT_RUSTFS_GET_CODEC_STREAMING_MIN_SIZE;

const ENV_RUSTFS_GET_CODEC_STREAMING_ENGINE: &str = "RUSTFS_GET_CODEC_STREAMING_ENGINE";
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_ENGINE: &str = GET_CODEC_STREAMING_ENGINE_LEGACY;

// Primary single switch for the codec-streaming GET fast path. `off` (default)
// keeps GET on the legacy duplex path; `on` (aliases `full`/`production`, plus
// the legacy `internal`/`benchmark`) opts it in. Combine with
// `..._ROLLOUT_PCT` for a partial (percentage-based) rollout.
const ENV_RUSTFS_GET_CODEC_STREAMING_ROLLOUT: &str = "RUSTFS_GET_CODEC_STREAMING_ROLLOUT";
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_ROLLOUT: &str = "off";

const ENV_RUSTFS_GET_CODEC_STREAMING_BODY_COMPAT_CONFIRMED: &str = "RUSTFS_GET_CODEC_STREAMING_BODY_COMPAT_CONFIRMED";
const ENV_RUSTFS_GET_CODEC_STREAMING_HEADER_COMPAT_CONFIRMED: &str = "RUSTFS_GET_CODEC_STREAMING_HEADER_COMPAT_CONFIRMED";

const ENV_RUSTFS_GET_CODEC_STREAMING_ROLLOUT_PCT: &str = "RUSTFS_GET_CODEC_STREAMING_ROLLOUT_PCT";
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_ROLLOUT_PCT: u32 = 100;

const ENV_RUSTFS_GET_CODEC_STREAMING_MULTIPART_ENABLE: &str = "RUSTFS_GET_CODEC_STREAMING_MULTIPART_ENABLE";
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_MULTIPART_ENABLE: bool = false;

const ENV_RUSTFS_GET_CODEC_STREAMING_MULTIPART_MAX_PARTS: &str = "RUSTFS_GET_CODEC_STREAMING_MULTIPART_MAX_PARTS";
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_MULTIPART_MAX_PARTS: usize = 256;

const ENV_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_ENABLE: &str = "RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_ENABLE";
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_ENABLE: bool = false;
const ENV_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_MAX_SIZE: &str = "RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_MAX_SIZE";
const DEFAULT_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_MAX_SIZE: usize = 512 * 1024;

const ENV_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY: &str = "RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY";
// On by default (rustfs/backlog#1802): a small object whose data shards are
// inlined in xl.meta is reassembled straight from the already-resolved
// metadata, skipping the Erasure reconstruct pipeline. The path has a complete
// fallback — if the inline reassembly returns None, the GET proceeds through
// the normal shard-read pipeline, so a miss is correctness-neutral. Set to
// `false` to force the legacy path (kill switch).
const DEFAULT_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY: bool = true;
const ENV_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY_THRESHOLD: &str = "RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY_THRESHOLD";
const DEFAULT_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY_THRESHOLD: usize = 128 * 1024;
// Bound the in-memory inline decoder independently from the configurable
// storage-class admission policy. This protects its Vec allocation while
// allowing explicitly configured EC layouts to inline objects above 128 KiB.
const INLINE_FAST_PATH_MAX_OBJECT_SIZE: i64 = 1024 * 1024;

// --- Metadata Early-Stop Configuration ---

const ENV_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE: &str = "RUSTFS_GET_METADATA_EARLY_STOP_ENABLE";
// Enabled by default (backlog#872): the early-stop path only engages for
// requests `should_allow_metadata_early_stop` classifies as safe (latest-version
// reads by default, without version_id / healing / free-version needs) and still
// requires a full read-quorum agreement before stopping. Data-read requests add
// a separate inline-shard verifier before cancelling the remaining fanout. Set
// the env var to `false` to fall back to full-wait metadata fanout.
const DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE: bool = true;

#[allow(
    dead_code,
    reason = "percentage-rollout facet of the metadata early-stop switch; its predicate has no caller while the sibling enable flag is live (backlog#1823)"
)]
const ENV_RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT: &str = "RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT";
#[allow(
    dead_code,
    reason = "percentage-rollout facet of the metadata early-stop switch; its predicate has no caller while the sibling enable flag is live (backlog#1823)"
)]
const DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT: u32 = 100;

const ENV_RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE: &str = "RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE";
const DEFAULT_RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE: bool = false;

const ENV_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE: &str = "RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE";
const DEFAULT_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE: bool = true;

// Opt-in non-inline data-read quorum early-stop rollout (backlog#1309). The
// existing metadata fanout still reads data-bearing metadata; this gate only
// permits a safe plain single-part candidate to stop before the full fanout.
// Keep it opt-in until the Linux multi-node slow-tail and small-inline cost
// gates are complete. The environment name is retained for compatibility.
const ENV_RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE: &str = "RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE";
const DEFAULT_RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE: bool = false;

const ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT: &str = "RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT";
const DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT: bool = true;

const ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS: &str = "RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS";
const ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS: &str = "RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS";
const ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET: &str = "RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET";
const ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX: &str = "RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX";

// --- Multipart Reader-Setup Prefetch Configuration (backlog#870) ---

const ENV_RUSTFS_GET_MULTIPART_READER_SETUP_PREFETCH: &str = "RUSTFS_GET_MULTIPART_READER_SETUP_PREFETCH";
const DEFAULT_RUSTFS_GET_MULTIPART_READER_SETUP_PREFETCH: bool = true;

/// Identifies the caller's read contract for policies that are deliberately
/// narrower than the storage API's ordinary GET contract.
///
/// Server-side copy consumes a source reader while a destination writer is
/// applying backpressure.  Its source read must not speculatively open the
/// next multipart part: those extra shard streams can share an internode H2
/// connection with the current part and starve the lockstep decoder.  Keep
/// this context internal so the public `ObjectOptions` and storage traits do
/// not acquire a copy-only field.
#[derive(Clone, Copy, Debug, Default, Eq, PartialEq)]
pub(crate) enum GetObjectReadPolicy {
    #[default]
    Default,
    CopySource,
}

impl GetObjectReadPolicy {
    pub(crate) const fn allows_multipart_setup_prefetch(self) -> bool {
        matches!(self, Self::Default)
    }
}

tokio::task_local! {
    static GET_OBJECT_READ_POLICY: GetObjectReadPolicy;
    static GET_OBJECT_READ_CANCELLATION: tokio_util::sync::CancellationToken;
}

pub(crate) fn get_object_read_policy() -> GetObjectReadPolicy {
    GET_OBJECT_READ_POLICY.try_with(|policy| *policy).unwrap_or_default()
}

pub(crate) async fn with_get_object_read_policy<F>(policy: GetObjectReadPolicy, future: F) -> F::Output
where
    F: std::future::Future,
{
    let decode_policy = match policy {
        GetObjectReadPolicy::Default => crate::erasure::coding::decode::DecodeReadPolicy::Default,
        GetObjectReadPolicy::CopySource => crate::erasure::coding::decode::DecodeReadPolicy::DemandBound,
    };
    crate::erasure::coding::decode::with_decode_read_policy(decode_policy, GET_OBJECT_READ_POLICY.scope(policy, future)).await
}

/// Return the request-owned cancellation token for a copy source, when one is
/// installed. The token is read before the detached legacy producer is spawned;
/// Tokio task-local values do not cross that spawn boundary on their own.
pub(crate) fn get_object_read_cancellation() -> Option<tokio_util::sync::CancellationToken> {
    GET_OBJECT_READ_CANCELLATION.try_with(|token| token.clone()).ok()
}

pub(crate) async fn with_get_object_read_cancellation<F>(
    cancellation: tokio_util::sync::CancellationToken,
    future: F,
) -> F::Output
where
    F: std::future::Future,
{
    GET_OBJECT_READ_CANCELLATION.scope(cancellation, future).await
}

#[cfg(not(test))]
static OBJECT_LOCK_DIAG_ENABLED: OnceLock<bool> = OnceLock::new();

mod core;
#[cfg(test)]
pub(crate) use core::io_primitives::disk_call_counters;
#[cfg(all(test, feature = "test-util"))]
pub(crate) use core::io_primitives::{ENV_RUSTFS_PUT_RENAME_EARLY_ACK_ENABLE, rename_fanout_barrier};
mod ctx;
mod metadata;
mod ops;
pub(crate) use ops::bucket::BucketInfoQuorum;
pub(crate) use ops::heal::AbsenceProofRequest;
#[cfg(test)]
pub(crate) use ops::heal::DanglingDeleteFailure;
pub(crate) use ops::heal::HealedObjectAbsence;

#[cfg(test)]
pub(crate) use ops::hermetic_set_disks_isolated;
#[cfg(test)]
pub(crate) use ops::multipart::NewMultipartUploadCommitObservation;
#[cfg(any(test, feature = "test-util"))]
pub use ops::multipart::{MultipartCommitBarrier, MultipartCommitPause};
#[cfg(test)]
pub(crate) use ops::object::DeleteObjectCommitBarrier;
#[cfg(feature = "test-util")]
pub(crate) use ops::object::TransitionCleanupStoreBarrier as SetDiskTransitionCleanupStoreBarrier;
#[cfg(all(test, feature = "test-util"))]
pub(crate) use ops::object::TransitionUploadedCommitBarrier as SetDiskTransitionUploadedCommitBarrier;
pub(crate) use ops::object::body_cache_plaintext_len;
#[cfg(all(test, feature = "test-util"))]
pub(crate) use ops::object::cleanup_rejected_transition_upload_durably;
#[cfg(any(test, feature = "test-util"))]
pub use ops::object::{PutObjectCommitBarrier, PutObjectCommitPause};
#[cfg(all(test, feature = "test-util"))]
pub(crate) use ops::object::{
    TransitionTransactionKillPoint as SetDiskTransitionTransactionKillPoint,
    TransitionTransactionKillPointBarrier as SetDiskTransitionTransactionKillPointBarrier,
};
#[cfg(all(test, feature = "test-util"))]
pub(crate) use ops::object::{
    TransitionTransactionMutationKind as SetDiskTransitionTransactionMutationKind,
    TransitionTransactionMutationObservation as SetDiskTransitionTransactionMutationObservation,
    TransitionTransactionMutationProbe as SetDiskTransitionTransactionMutationProbe,
};
mod read;
mod replication;
pub(crate) mod shard_source;
#[cfg(all(test, feature = "test-util"))]
mod transition_matrix_tests;

pub use ops::heal_walk::HealWalkVersion;

pub(in crate::set_disk) struct GetObjectFileInfo {
    owned: Option<OwnedGetObjectFileInfo>,
    shared: Option<Arc<GetObjectMetadataCacheEntry>>,
}

struct OwnedGetObjectFileInfo {
    fi: FileInfo,
    parts_metadata: Vec<FileInfo>,
    online_disks: Vec<Option<DiskStore>>,
    late_metadata_fanout_disks: Option<Vec<Option<DiskStore>>>,
}

type OwnedGetObjectFileInfoParts = (FileInfo, Vec<FileInfo>, Vec<Option<DiskStore>>, Option<Vec<Option<DiskStore>>>);

impl GetObjectFileInfo {
    fn owned(fi: FileInfo, parts_metadata: Vec<FileInfo>, online_disks: Vec<Option<DiskStore>>) -> Self {
        Self {
            owned: Some(OwnedGetObjectFileInfo {
                fi,
                parts_metadata,
                online_disks,
                late_metadata_fanout_disks: None,
            }),
            shared: None,
        }
    }

    fn owned_with_late_metadata_fanout(
        fi: FileInfo,
        parts_metadata: Vec<FileInfo>,
        online_disks: Vec<Option<DiskStore>>,
        late_metadata_fanout_disks: Vec<Option<DiskStore>>,
    ) -> Self {
        Self {
            owned: Some(OwnedGetObjectFileInfo {
                fi,
                parts_metadata,
                online_disks,
                late_metadata_fanout_disks: Some(late_metadata_fanout_disks),
            }),
            shared: None,
        }
    }

    fn shared(entry: Arc<GetObjectMetadataCacheEntry>) -> Self {
        Self {
            owned: None,
            shared: Some(entry),
        }
    }

    fn fi(&self) -> &FileInfo {
        match (&self.owned, &self.shared) {
            (Some(snapshot), None) => &snapshot.fi,
            (None, Some(entry)) => &entry.fi,
            _ => unreachable!("GET metadata snapshot representation must be exclusive"),
        }
    }

    fn parts_metadata(&self) -> &[FileInfo] {
        match (&self.owned, &self.shared) {
            (Some(snapshot), None) => &snapshot.parts_metadata,
            (None, Some(entry)) => &entry.parts_metadata,
            _ => unreachable!("GET metadata snapshot representation must be exclusive"),
        }
    }

    fn online_disks(&self) -> &[Option<DiskStore>] {
        match (&self.owned, &self.shared) {
            (Some(snapshot), None) => &snapshot.online_disks,
            (None, Some(entry)) => &entry.online_disks,
            _ => unreachable!("GET metadata snapshot representation must be exclusive"),
        }
    }

    fn has_late_metadata_fanout(&self) -> bool {
        self.owned
            .as_ref()
            .is_some_and(|snapshot| snapshot.late_metadata_fanout_disks.is_some())
    }

    fn into_owned(self) -> (FileInfo, Vec<FileInfo>, Vec<Option<DiskStore>>) {
        let (fi, parts_metadata, online_disks, _) = self.into_owned_with_late_metadata_fanout();
        (fi, parts_metadata, online_disks)
    }

    fn into_owned_with_late_metadata_fanout(self) -> OwnedGetObjectFileInfoParts {
        match (self.owned, self.shared) {
            (Some(snapshot), None) => (
                snapshot.fi,
                snapshot.parts_metadata,
                snapshot.online_disks,
                snapshot.late_metadata_fanout_disks,
            ),
            (None, Some(entry)) => match Arc::try_unwrap(entry) {
                Ok(entry) => (entry.fi, entry.parts_metadata, entry.online_disks, None),
                Err(entry) => (entry.fi.clone(), entry.parts_metadata.clone(), entry.online_disks.clone(), None),
            },
            _ => unreachable!("GET metadata snapshot representation must be exclusive"),
        }
    }

    #[cfg(test)]
    fn has_valid_representation(&self) -> bool {
        self.owned.is_some() ^ self.shared.is_some()
    }

    #[cfg(test)]
    fn shared_entry(&self) -> Option<&Arc<GetObjectMetadataCacheEntry>> {
        self.shared.as_ref()
    }
}

pub(crate) struct PreparedGetObjectMetadata {
    snapshot: GetObjectFileInfo,
    object_info: Option<ObjectInfo>,
}

impl PreparedGetObjectMetadata {
    pub(crate) fn object_info(&self) -> &ObjectInfo {
        self.object_info
            .as_ref()
            .expect("prepared GET metadata must retain its ObjectInfo until consumed")
    }

    pub(crate) fn take_object_info(&mut self) -> ObjectInfo {
        self.object_info
            .take()
            .expect("prepared GET metadata ObjectInfo must be consumed exactly once")
    }

    pub(crate) fn read_semantics_identity(&self) -> [u8; 32] {
        SetDisks::file_info_quorum_hash(self.snapshot.fi())
    }
}

tokio::task_local! {
    static PREPARED_GET_OBJECT_METADATA: std::cell::RefCell<Option<PreparedGetObjectMetadata>>;
}

#[cfg(test)]
tokio::task_local! {
    static GET_OBJECT_INFO_CONVERSIONS: Arc<AtomicU64>;
}

fn build_get_object_info(fi: &FileInfo, bucket: &str, object: &str, versioned: bool) -> ObjectInfo {
    #[cfg(test)]
    let _ = GET_OBJECT_INFO_CONVERSIONS.try_with(|conversions| {
        conversions.fetch_add(1, Ordering::Relaxed);
    });
    ObjectInfo::from_file_info(fi, bucket, object, versioned)
}

fn take_prepared_get_object_metadata() -> Option<PreparedGetObjectMetadata> {
    PREPARED_GET_OBJECT_METADATA
        .try_with(|prepared| prepared.borrow_mut().take())
        .ok()
        .flatten()
}

async fn with_prepared_get_object_metadata<F>(metadata: PreparedGetObjectMetadata, future: F) -> F::Output
where
    F: std::future::Future,
{
    PREPARED_GET_OBJECT_METADATA
        .scope(std::cell::RefCell::new(Some(metadata)), future)
        .await
}

#[cfg(test)]
mod prepared_get_object_metadata_tests {
    use super::*;
    use crate::ecstore_validation_blackbox::make_local_set_disks;
    use crate::object_api::{BLOCK_SIZE_V2, PutObjReader};
    use crate::set_disk::core::io_primitives::{bounded_metadata_fanout_order, disk_call_counters, rename_fanout_barrier};
    use crate::storage_api_contracts::bucket::MakeBucketOptions;
    use crate::test_metrics::CapturingRecorder;
    use http::HeaderMap;
    use tokio::io::AsyncReadExt;

    const READ_VERSION_BARRIER_GUARD: std::time::Duration = std::time::Duration::from_secs(10);

    fn object_with_initial_data_shards(bucket: &str, prefix: &str) -> String {
        object_with_initial_data_shards_for_geometry(bucket, prefix, 4, 2)
    }

    fn object_with_initial_data_shards_for_geometry(bucket: &str, prefix: &str, total_disks: usize, parity: usize) -> String {
        (0..1000)
            .map(|index| format!("{prefix}-{index}.bin"))
            .find(|name| {
                let order = bounded_metadata_fanout_order(bucket, name, total_disks, parity);
                let data = total_disks.saturating_sub(parity);
                let distribution = FileInfo::new(&[bucket, name].join("/"), data, parity).erasure.distribution;
                let mut seen = vec![false; data];
                for disk_index in order.into_iter().take(total_disks.saturating_sub(parity).saturating_add(1)) {
                    if let Some(block_index) = distribution.get(disk_index).copied().filter(|index| *index <= data) {
                        seen[block_index - 1] = true;
                    }
                }
                seen.into_iter().all(|seen| seen)
            })
            .expect("test should find an object whose initial fanout covers both data shards")
    }

    fn bounded_initial_parity_disk_index(bucket: &str, object: &str) -> usize {
        *bounded_metadata_fanout_order(bucket, object, 4, 2)
            .get(2)
            .expect("4-disk test geometry should schedule one parity disk initially")
    }

    #[tokio::test]
    async fn prepared_metadata_is_consumed_exactly_once() {
        let snapshot = GetObjectFileInfo::owned(FileInfo::default(), Vec::new(), Vec::new());
        assert!(snapshot.has_valid_representation());
        assert!(snapshot.shared_entry().is_none());
        let metadata = PreparedGetObjectMetadata {
            snapshot,
            object_info: None,
        };

        with_prepared_get_object_metadata(metadata, async {
            assert!(take_prepared_get_object_metadata().is_some());
            assert!(take_prepared_get_object_metadata().is_none());
        })
        .await;
        assert!(take_prepared_get_object_metadata().is_none());
    }

    #[test]
    fn cache_hit_consumers_release_snapshot_at_legacy_ownership() {
        let fi = FileInfo {
            name: "object".to_owned(),
            ..Default::default()
        };
        let cached = Arc::new(GetObjectMetadataCacheEntry {
            created_at: Instant::now(),
            parts_metadata: vec![fi.clone()],
            fi,
            online_disks: vec![None],
            read_quorum: 0,
        });
        let snapshot = GetObjectFileInfo::shared(Arc::clone(&cached));

        assert!(snapshot.has_valid_representation());
        assert!(std::mem::size_of::<GetObjectFileInfo>() >= std::mem::size_of::<OwnedGetObjectFileInfo>());
        assert!(
            std::mem::size_of::<GetObjectFileInfo>()
                <= std::mem::size_of::<OwnedGetObjectFileInfo>() + 2 * std::mem::size_of::<usize>()
        );
        assert_eq!(Arc::strong_count(&cached), 2, "a cache hit must add one snapshot reference");
        assert_eq!(snapshot.fi().name, "object");
        assert_eq!(snapshot.parts_metadata().len(), 1);
        assert_eq!(snapshot.online_disks().len(), 1);
        assert_eq!(Arc::strong_count(&cached), 2, "borrowing consumers must not clone the snapshot");

        let (owned_fi, owned_parts, disks) = snapshot.into_owned();
        assert_eq!(owned_fi.name, "object");
        assert_eq!(owned_parts.len(), 1);
        assert_eq!(disks.len(), 1);
        assert_eq!(
            Arc::strong_count(&cached),
            1,
            "legacy ownership must release the cache snapshot after cloning its owned inputs"
        );
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn prepared_reader_reuses_metadata_fanout_exactly_once() {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "prepared-metadata-fanout";
        let object = "prepared-metadata-fanout-object.bin";
        let payload = b"prepared-metadata-fanout-payload-".repeat(40_000);
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        let calls = disk_call_counters::observe(object);
        let conversions = Arc::new(AtomicU64::new(0));
        let restored = GET_OBJECT_INFO_CONVERSIONS
            .scope(Arc::clone(&conversions), async {
                let metadata = set_disks
                    .prepare_get_object_metadata(bucket, object, &opts)
                    .await
                    .expect("prepared metadata should resolve");
                let prepared_calls = calls.total(disk_call_counters::KIND_READ_VERSION);
                assert_eq!(prepared_calls, 4, "default prepared GET metadata should keep full data-read fanout");

                let mut reader = set_disks
                    .get_object_reader_with_prepared_metadata(bucket, object, None, HeaderMap::new(), &opts, metadata)
                    .await
                    .expect("prepared body reader should open");
                let mut restored = Vec::new();
                reader
                    .stream
                    .read_to_end(&mut restored)
                    .await
                    .expect("prepared body should stream");
                restored
            })
            .await;

        assert_eq!(restored, payload);
        assert_eq!(
            conversions.load(Ordering::Relaxed),
            1,
            "prepared reader must consume the ObjectInfo built during metadata preparation"
        );
        assert_eq!(
            calls.total(disk_call_counters::KIND_READ_VERSION),
            4,
            "reader construction must consume prepared metadata instead of repeating the fanout"
        );
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn non_inline_data_read_early_stop_uses_quorum_plan() {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "non-inline-read-plan";
        let object = object_with_initial_data_shards(bucket, "non-inline-object");
        let payload = vec![0x5a; 2 * 1024 * 1024];
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        temp_env::async_with_vars(
            [
                ("RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
            ],
            async {
                let calls = disk_call_counters::observe(&object);
                reset_test_get_object_reader_path();
                let mut reader = set_disks
                    .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                    .await
                    .expect("quorum GET reader should open");
                let mut restored = Vec::new();
                reader
                    .stream
                    .read_to_end(&mut restored)
                    .await
                    .expect("quorum GET body should stream");
                assert_eq!(restored, payload);
                assert!(
                    calls.total(disk_call_counters::KIND_READ_VERSION) < 4,
                    "non-inline quorum GET should retain a reconstruction reserve"
                );
            },
        )
        .await;
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn non_inline_two_phase_read_fetches_late_parity_after_two_selected_shards_fail() {
        let (dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "non-inline-read-late-parity";
        let object = object_with_initial_data_shards(bucket, "late-parity-object");
        let payload = vec![0x5a; 1024 * 1024];
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        let order = bounded_metadata_fanout_order(bucket, &object, 4, 2);
        let distribution = FileInfo::new(&[bucket, object.as_str()].join("/"), 2, 2).erasure.distribution;
        assert!(
            order.iter().take(2).all(|disk_index| distribution[*disk_index] <= 2),
            "the two failed selected shards must be data shards"
        );
        assert!(
            distribution[order[3]] > 2,
            "the metadata shard omitted by the plan must be healthy parity"
        );
        for disk_index in order.iter().take(2) {
            let object_dir = dirs[*disk_index].path().join(bucket).join(&object);
            let data_dir = std::fs::read_dir(&object_dir)
                .expect("object directory should be readable")
                .find_map(|entry| {
                    let entry = entry.expect("object directory entry should be readable");
                    entry
                        .file_type()
                        .expect("object directory entry type should be readable")
                        .is_dir()
                        .then(|| entry.path())
                })
                .expect("object data directory should exist");
            let part_path = data_dir.join("part.1");
            let mut shard = std::fs::read(&part_path).expect("selected data shard should be readable before corruption");
            shard[0] ^= 0xff;
            std::fs::write(part_path, shard).expect("selected data shard should be corrupted after metadata was written");
        }

        temp_env::async_with_vars(
            [
                ("RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
            ],
            async {
                // Isolate late body recovery from the independent metadata slow-tail hedge.
                let _hedge_timer = rename_fanout_barrier::arm(&object, 0, rename_fanout_barrier::PHASE_NON_INLINE_HEDGE_TIMER);
                let calls = disk_call_counters::observe(&object);
                let mut reader = set_disks
                    .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                    .await
                    .expect("two-phase GET should recover using late parity");
                let mut restored = Vec::new();
                reader
                    .stream
                    .read_to_end(&mut restored)
                    .await
                    .expect("late parity should restore the exact GET body");
                assert_eq!(restored, payload);
                assert_eq!(calls.total(disk_call_counters::KIND_READ_VERSION), 7);
            },
        )
        .await;
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn non_inline_two_phase_read_fetches_late_parity_when_selected_parts_are_missing() {
        let (dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "non-inline-read-late-parity-missing";
        let object = object_with_initial_data_shards(bucket, "late-parity-missing-object");
        let payload = vec![0x3c; 1024 * 1024];
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        let order = bounded_metadata_fanout_order(bucket, &object, 4, 2);
        for disk_index in order.iter().take(2) {
            let object_dir = dirs[*disk_index].path().join(bucket).join(&object);
            let data_dir = std::fs::read_dir(&object_dir)
                .expect("object directory should be readable")
                .find_map(|entry| {
                    let entry = entry.expect("object directory entry should be readable");
                    entry
                        .file_type()
                        .expect("entry type should be readable")
                        .is_dir()
                        .then(|| entry.path())
                })
                .expect("object data directory should exist");
            std::fs::remove_file(data_dir.join("part.1")).expect("selected data shard should be removed");
        }

        temp_env::async_with_vars(
            [
                ("RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
            ],
            async {
                // Isolate late body recovery from the independent metadata slow-tail hedge.
                let _hedge_timer = rename_fanout_barrier::arm(&object, 0, rename_fanout_barrier::PHASE_NON_INLINE_HEDGE_TIMER);
                let calls = disk_call_counters::observe(&object);
                let mut reader = set_disks
                    .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                    .await
                    .expect("two-phase GET should recover using late parity");
                let mut restored = Vec::new();
                reader
                    .stream
                    .read_to_end(&mut restored)
                    .await
                    .expect("late parity should restore the exact GET body");
                assert_eq!(restored, payload);
                assert_eq!(calls.total(disk_call_counters::KIND_READ_VERSION), 7);
                assert_eq!(
                    calls.for_disk(disk_call_counters::KIND_READ_VERSION, order[3]),
                    1,
                    "the omitted parity disk must only be read by the late metadata refresh"
                );
            },
        )
        .await;
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn four_data_two_parity_two_phase_read_recovers_one_failed_data_shard() {
        let (dirs, set_disks) = make_local_set_disks(6, 2).await;
        let bucket = "four-data-two-parity-late-read";
        let object = object_with_initial_data_shards_for_geometry(bucket, "one-failed-data", 4, 2);
        let payload = vec![0x7a; 1024 * 1024];
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        let order = bounded_metadata_fanout_order(bucket, &object, 6, 2);
        let distribution = FileInfo::new(&[bucket, object.as_str()].join("/"), 4, 2).erasure.distribution;
        let failed_disk = *order
            .iter()
            .take(4)
            .find(|disk_index| distribution[**disk_index] <= 4)
            .expect("initial fanout should include a data shard");
        assert!(
            order.iter().take(4).all(|disk_index| distribution[*disk_index] <= 4),
            "initial fanout should cover all four data shards"
        );
        assert!(distribution[order[5]] > 4, "the final deferred metadata shard should be parity");

        let object_dir = dirs[failed_disk].path().join(bucket).join(&object);
        let data_dir = std::fs::read_dir(&object_dir)
            .expect("object directory should be readable")
            .find_map(|entry| {
                let entry = entry.expect("object directory entry should be readable");
                entry
                    .file_type()
                    .expect("object directory entry type should be readable")
                    .is_dir()
                    .then(|| entry.path())
            })
            .expect("object data directory should exist");
        let part_path = data_dir.join("part.1");
        let mut shard = std::fs::read(&part_path).expect("selected data shard should be readable before corruption");
        shard[0] ^= 0xff;
        std::fs::write(part_path, shard).expect("selected data shard should be corrupted after metadata was written");

        temp_env::async_with_vars(
            [
                ("RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
            ],
            async {
                // Isolate late body recovery from the independent metadata slow-tail hedge.
                let _hedge_timer = rename_fanout_barrier::arm(&object, 0, rename_fanout_barrier::PHASE_NON_INLINE_HEDGE_TIMER);
                let calls = disk_call_counters::observe(&object);
                let mut reader = set_disks
                    .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                    .await
                    .expect("two-phase GET should recover with one failed data shard");
                let mut restored = Vec::new();
                reader
                    .stream
                    .read_to_end(&mut restored)
                    .await
                    .expect("late parity should restore the exact GET body");
                assert_eq!(restored, payload);
                assert_eq!(calls.total(disk_call_counters::KIND_READ_VERSION), 11);
            },
        )
        .await;
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn four_data_two_parity_two_phase_read_rejects_below_read_quorum() {
        let (dirs, set_disks) = make_local_set_disks(6, 2).await;
        let bucket = "four-data-two-parity-quorum-minus-one";
        let object = object_with_initial_data_shards_for_geometry(bucket, "quorum-minus-one", 4, 2);
        let payload = vec![0x4b; 1024 * 1024];
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        let order = bounded_metadata_fanout_order(bucket, &object, 6, 2);
        for disk_index in order.iter().take(3) {
            let object_dir = dirs[*disk_index].path().join(bucket).join(&object);
            let data_dir = std::fs::read_dir(&object_dir)
                .expect("object directory should be readable")
                .find_map(|entry| {
                    let entry = entry.expect("object directory entry should be readable");
                    entry
                        .file_type()
                        .expect("entry type should be readable")
                        .is_dir()
                        .then(|| entry.path())
                })
                .expect("object data directory should exist");
            std::fs::remove_file(data_dir.join("part.1")).expect("selected shard should be removed");
        }

        temp_env::async_with_vars(
            [
                ("RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
            ],
            async {
                let result = set_disks
                    .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                    .await;
                assert!(result.is_err(), "quorum-minus-one read must fail closed without exposing a body");
            },
        )
        .await;
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn non_inline_data_read_early_stop_keeps_reserve_on_unequal_layout() {
        let (_dirs, set_disks) = make_local_set_disks(6, 2).await;
        let bucket = "non-inline-read-reserve";
        let object = object_with_initial_data_shards_for_geometry(bucket, "reserve-object", 6, 2);
        let payload = vec![0x5a; 2 * 1024 * 1024];
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        temp_env::async_with_vars(
            [
                ("RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
            ],
            async {
                // Hold this object's hedge timer so real-disk latency cannot add speculative fanout.
                let _hedge_timer = rename_fanout_barrier::arm(&object, 0, rename_fanout_barrier::PHASE_NON_INLINE_HEDGE_TIMER);
                let calls = disk_call_counters::observe(&object);
                let mut reader = set_disks
                    .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                    .await
                    .expect("quorum GET reader should open");
                let mut restored = Vec::new();
                reader
                    .stream
                    .read_to_end(&mut restored)
                    .await
                    .expect("quorum GET body should stream");
                assert_eq!(restored, payload);
                assert_eq!(
                    calls.total(disk_call_counters::KIND_READ_VERSION),
                    5,
                    "the unequal layout should schedule exactly one reserve beyond its data quorum"
                );
            },
        )
        .await;
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn non_inline_data_read_early_stop_preserves_inline_path() {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "non-inline-read-plan-inline";
        let object = object_with_initial_data_shards(bucket, "inline-object");
        let payload = b"quorum inline payload".repeat(256);
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        temp_env::async_with_vars(
            [
                ("RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
            ],
            async {
                let calls = disk_call_counters::observe(&object);
                let mut reader = set_disks
                    .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                    .await
                    .expect("inline GET reader should open");
                let mut restored = Vec::new();
                reader
                    .stream
                    .read_to_end(&mut restored)
                    .await
                    .expect("inline GET body should stream");
                assert_eq!(restored, payload);
                assert_eq!(
                    test_get_object_reader_path_id(),
                    3,
                    "inline GET should retain the direct inline reader path"
                );
                assert!(calls.total(disk_call_counters::KIND_READ_VERSION) <= 4);
            },
        )
        .await;
    }

    #[test]
    #[serial_test::serial(body_cache_hook)]
    fn non_inline_data_read_early_stop_does_not_add_inline_fanout_on_unequal_layout() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("current-thread runtime should build");
        let bucket = "inline-read-plan-unequal";
        let object = object_with_initial_data_shards_for_geometry(bucket, "inline-object", 6, 2);
        let payload = b"inline quorum payload".repeat(256);
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

        let (_dirs, set_disks) = runtime.block_on(async {
            let (dirs, set_disks) = make_local_set_disks(6, 2).await;
            set_disks
                .make_bucket(bucket, &MakeBucketOptions::default())
                .await
                .expect("bucket should be created");
            let mut put_reader = PutObjReader::from_vec(payload.clone());
            set_disks
                .put_object(bucket, &object, &mut put_reader, &opts)
                .await
                .expect("inline object should be written");
            (dirs, set_disks)
        });

        // disk_call_counters counts tasks that started running, so an
        // early-stop abort races the single-pending inline hedge into a ±1
        // count per read. The fanout lifecycle histogram records the
        // scheduling decision itself and stays deterministic under load.
        let recorder = CapturingRecorder::default();
        let previous_gate = rustfs_io_metrics::get_stage_metrics_enabled();
        rustfs_io_metrics::set_get_stage_metrics_enabled(true);
        metrics::with_local_recorder(&recorder, || {
            runtime.block_on(async {
                for enabled in [false, true] {
                    temp_env::async_with_vars(
                        [
                            (
                                "RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE",
                                Some(if enabled { "true" } else { "false" }),
                            ),
                            ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                            ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
                        ],
                        async {
                            let mut reader = set_disks
                                .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                                .await
                                .expect("inline GET reader should open");
                            let mut restored = Vec::new();
                            reader
                                .stream
                                .read_to_end(&mut restored)
                                .await
                                .expect("inline GET body should stream");
                            assert_eq!(restored, payload);
                        },
                    )
                    .await;
                }
            })
        });
        rustfs_io_metrics::set_get_stage_metrics_enabled(previous_gate);

        let scheduled = recorder.histogram_values(
            "rustfs_io_get_object_metadata_fanout_scheduled",
            &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)],
        );
        assert_eq!(scheduled.len(), 2, "each GET should run exactly one metadata fanout");
        assert_eq!(scheduled[1], scheduled[0], "inline gate must not add reserve fanout");
    }

    #[test]
    #[serial_test::serial(body_cache_hook)]
    fn inline_data_read_early_stop_defaults_return_exact_body() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("current-thread runtime should build");
        let bucket = "inline-data-read-early-stop-reader";
        let object = object_with_initial_data_shards(bucket, "inline-data-read-early-stop-reader-object");
        let payload = b"inline early-stop reader payload".repeat(256);
        let recorder = CapturingRecorder::default();
        let previous_gate = rustfs_io_metrics::get_stage_metrics_enabled();
        rustfs_io_metrics::set_get_stage_metrics_enabled(true);

        let (restored, object_size, calls_total) = metrics::with_local_recorder(&recorder, || {
            runtime.block_on(async {
                let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
                let opts = ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                };

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

                temp_env::async_with_vars(
                    [
                        ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", None::<&str>),
                        ("RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE", None::<&str>),
                        ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", None::<&str>),
                    ],
                    async {
                        let slow_parity_disk = bounded_initial_parity_disk_index(bucket, &object);
                        let barrier =
                            rename_fanout_barrier::arm(&object, slow_parity_disk, rename_fanout_barrier::PHASE_READ_VERSION);
                        let calls = disk_call_counters::observe(&object);
                        let set_disks_for_read = Arc::clone(&set_disks);
                        let opts_for_read = opts.clone();
                        let object_for_read = object.clone();
                        let mut open_reader = tokio::spawn(async move {
                            set_disks_for_read
                                .get_object_reader(bucket, &object_for_read, None, HeaderMap::new(), &opts_for_read)
                                .await
                        });

                        tokio::time::timeout(READ_VERSION_BARRIER_GUARD, barrier.wait_until_paused())
                            .await
                            .expect("default inline GET should pause a slow parity metadata read");
                        let mut reader = tokio::time::timeout(READ_VERSION_BARRIER_GUARD, &mut open_reader)
                            .await
                            .expect("default production inline GET should return before the paused parity metadata response")
                            .expect("inline GET reader task should not panic")
                            .expect("inline GET reader should open");
                        let object_size = reader.object_info.size;
                        let mut restored = Vec::new();
                        reader
                            .stream
                            .read_to_end(&mut restored)
                            .await
                            .expect("inline GET body should stream");

                        (restored, object_size, calls.total(disk_call_counters::KIND_READ_VERSION))
                    },
                )
                .await
            })
        });
        rustfs_io_metrics::set_get_stage_metrics_enabled(previous_gate);

        assert_eq!(object_size, payload.len() as i64);
        assert_eq!(restored, payload);
        assert_eq!(
            calls_total, 4,
            "default production inline GET should schedule the initial bounded quorum plus one hedge"
        );
        assert_eq!(
            recorder.histogram_values(
                "rustfs_io_get_object_metadata_fanout_scheduled",
                &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
            ),
            vec![4.0],
            "default production GET should record all scheduled metadata tasks"
        );
        assert_eq!(
            recorder.histogram_values(
                "rustfs_io_get_object_metadata_fanout_completed",
                &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
            ),
            vec![3.0],
            "default production GET should record only observed metadata responses as completed"
        );
        assert_eq!(
            recorder.histogram_values(
                "rustfs_io_get_object_metadata_fanout_cancelled",
                &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
            ),
            vec![1.0],
            "default production GET should record the aborted slow parity metadata task"
        );
    }

    #[test]
    #[serial_test::serial(body_cache_hook)]
    fn prepared_metadata_uses_full_fanout_even_when_data_read_early_stop_is_enabled() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("current-thread runtime should build");
        let bucket = "prepared-metadata-early-stop-enabled";
        let object = object_with_initial_data_shards(bucket, "prepared-metadata-early-stop-enabled-object");
        let payload = b"prepared metadata early-stop enabled payload".repeat(16);
        let recorder = CapturingRecorder::default();
        let previous_gate = rustfs_io_metrics::get_stage_metrics_enabled();
        rustfs_io_metrics::set_get_stage_metrics_enabled(true);

        let (restored, calls_total) = metrics::with_local_recorder(&recorder, || {
            runtime.block_on(async {
                let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
                let opts = ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                };

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

                temp_env::async_with_vars(
                    [
                        ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", Some("true")),
                        ("RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE", Some("true")),
                        ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", Some("true")),
                    ],
                    async {
                        let calls = disk_call_counters::observe(&object);
                        let metadata = set_disks
                            .prepare_get_object_metadata(bucket, &object, &opts)
                            .await
                            .expect("prepared metadata should resolve");
                        let calls_total = calls.total(disk_call_counters::KIND_READ_VERSION);

                        let mut reader = set_disks
                            .get_object_reader_with_prepared_metadata(bucket, &object, None, HeaderMap::new(), &opts, metadata)
                            .await
                            .expect("prepared body reader should open");
                        let mut restored = Vec::new();
                        reader
                            .stream
                            .read_to_end(&mut restored)
                            .await
                            .expect("prepared body should stream");
                        (restored, calls_total)
                    },
                )
                .await
            })
        });
        rustfs_io_metrics::set_get_stage_metrics_enabled(previous_gate);

        assert_eq!(restored, payload);
        assert_eq!(
            calls_total, 4,
            "prepared metadata must opt out of data-read early-stop until the read shape is known"
        );
        assert_eq!(
            recorder.histogram_values(
                "rustfs_io_get_object_metadata_fanout_scheduled",
                &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
            ),
            vec![4.0],
            "prepared metadata should schedule the full metadata fanout"
        );
        assert_eq!(
            recorder.histogram_values(
                "rustfs_io_get_object_metadata_fanout_completed",
                &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
            ),
            vec![4.0],
            "prepared metadata must wait for every scheduled metadata response"
        );
        assert_eq!(
            recorder.histogram_values(
                "rustfs_io_get_object_metadata_fanout_cancelled",
                &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
            ),
            vec![0.0],
            "prepared metadata must not cancel metadata responses"
        );
    }

    #[test]
    #[serial_test::serial(body_cache_hook)]
    fn data_read_early_stop_request_shapes_full_wait_in_production_reader() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("current-thread runtime should build");
        let bucket = "data-read-early-stop-shape-reader";
        let payload = b"shape-gated inline reader payload".repeat(256);

        for (object_prefix, range, configure_opts, expected_body) in [
            (
                "data-read-early-stop-range-reader-object",
                Some(HTTPRangeSpec {
                    start: 0,
                    end: 3,
                    is_suffix_length: false,
                }),
                None,
                payload[..4].to_vec(),
            ),
            ("data-read-early-stop-part-reader-object", None, Some(1), payload.clone()),
        ] {
            let recorder = CapturingRecorder::default();
            let previous_gate = rustfs_io_metrics::get_stage_metrics_enabled();
            rustfs_io_metrics::set_get_stage_metrics_enabled(true);
            let (restored, calls_total) = metrics::with_local_recorder(&recorder, || {
                runtime.block_on(async {
                    let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
                    let object = object_with_initial_data_shards(bucket, object_prefix);
                    let mut opts = ObjectOptions {
                        no_lock: true,
                        ..Default::default()
                    };
                    opts.part_number = configure_opts;

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

                    temp_env::async_with_vars(
                        [
                            ("RUSTFS_GET_METADATA_EARLY_STOP_ENABLE", None::<&str>),
                            ("RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE", None::<&str>),
                            ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", None::<&str>),
                        ],
                        async {
                            let calls = disk_call_counters::observe(&object);
                            let mut reader = set_disks
                                .get_object_reader(bucket, &object, range, HeaderMap::new(), &opts)
                                .await
                                .expect("shape-gated GET reader should open");
                            let mut restored = Vec::new();
                            reader
                                .stream
                                .read_to_end(&mut restored)
                                .await
                                .expect("shape-gated GET body should stream");
                            (restored, calls.total(disk_call_counters::KIND_READ_VERSION))
                        },
                    )
                    .await
                })
            });
            rustfs_io_metrics::set_get_stage_metrics_enabled(previous_gate);

            assert_eq!(restored, expected_body);
            assert_eq!(calls_total, 4, "shape-gated production GET should keep full metadata fanout");
            assert_eq!(
                recorder.histogram_values(
                    "rustfs_io_get_object_metadata_fanout_scheduled",
                    &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
                ),
                vec![4.0],
                "shape-gated production GET should schedule the full metadata fanout"
            );
            assert_eq!(
                recorder.histogram_values(
                    "rustfs_io_get_object_metadata_fanout_completed",
                    &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
                ),
                vec![4.0],
                "shape-gated production GET must wait for every scheduled metadata response"
            );
            assert_eq!(
                recorder.histogram_values(
                    "rustfs_io_get_object_metadata_fanout_cancelled",
                    &[("path", GET_OBJECT_PATH_LEGACY_DUPLEX)]
                ),
                vec![0.0],
                "shape-gated production GET must not cancel metadata responses"
            );
        }
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn prepared_reader_rebuilds_object_info_when_precomputed_value_is_absent() {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "prepared-object-info-fallback";
        let object = "prepared-object-info-fallback.bin";
        let payload = b"prepared-object-info-fallback-payload".repeat(4_000);
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        let conversions = Arc::new(AtomicU64::new(0));
        let restored = GET_OBJECT_INFO_CONVERSIONS
            .scope(Arc::clone(&conversions), async {
                let mut metadata = set_disks
                    .prepare_get_object_metadata(bucket, object, &opts)
                    .await
                    .expect("prepared metadata should resolve");
                metadata.object_info = None;
                let mut reader = set_disks
                    .get_object_reader_with_prepared_metadata(bucket, object, None, HeaderMap::new(), &opts, metadata)
                    .await
                    .expect("reader should rebuild missing prepared ObjectInfo");
                let mut restored = Vec::new();
                reader
                    .stream
                    .read_to_end(&mut restored)
                    .await
                    .expect("fallback reader should stream");
                restored
            })
            .await;

        assert_eq!(restored, payload);
        assert_eq!(
            conversions.load(Ordering::Relaxed),
            2,
            "missing precomputed ObjectInfo must trigger exactly one structural fallback rebuild"
        );
    }

    #[tokio::test]
    #[serial_test::serial(body_cache_hook)]
    async fn prepared_reader_restores_full_body_with_one_offline_shard() {
        let (_dirs, set_disks) = make_local_set_disks(4, 2).await;
        let bucket = "prepared-reader-offline-shard";
        let object = "prepared-reader-offline-shard-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 put_reader = PutObjReader::from_vec(payload.clone());
        set_disks
            .put_object(bucket, object, &mut put_reader, &opts)
            .await
            .expect("object should be written");
        set_disks.disks.write().await[0] = None;

        let metadata = set_disks
            .prepare_get_object_metadata(bucket, object, &opts)
            .await
            .expect("prepared metadata should tolerate one offline shard");
        let mut reader = set_disks
            .get_object_reader_with_prepared_metadata(bucket, object, None, HeaderMap::new(), &opts, metadata)
            .await
            .expect("prepared body reader should open with one offline shard");
        let mut restored = Vec::new();
        reader
            .stream
            .read_to_end(&mut restored)
            .await
            .expect("degraded prepared body should stream");

        assert_eq!(restored, payload);
    }
}

impl SetDisks {
    #[cfg(test)]
    async fn pause_tiered_metadata_commit(bucket: &str, object: &str) {
        let barrier = TIERED_METADATA_COMMIT_BARRIER
            .get_or_init(|| std::sync::Mutex::new(None))
            .lock()
            .expect("tiered metadata commit barrier should not be poisoned")
            .as_ref()
            .filter(|barrier| barrier.bucket == bucket && barrier.object == object)
            .cloned();
        if let Some(barrier) = barrier {
            barrier.arrived.notify_one();
            barrier.release.notified().await;
        }
    }

    pub(crate) async fn prepare_get_object_metadata(
        &self,
        bucket: &str,
        object: &str,
        opts: &ObjectOptions,
    ) -> Result<PreparedGetObjectMetadata> {
        let snapshot = self.get_object_fileinfo(bucket, object, opts, true, false).await?;
        let object_info = build_get_object_info(snapshot.fi(), bucket, object, opts.versioned || opts.version_suspended);
        Ok(PreparedGetObjectMetadata {
            snapshot,
            object_info: Some(object_info),
        })
    }

    pub(crate) async fn get_object_reader_with_prepared_metadata(
        &self,
        bucket: &str,
        object: &str,
        range: Option<HTTPRangeSpec>,
        headers: HeaderMap,
        opts: &ObjectOptions,
        metadata: PreparedGetObjectMetadata,
    ) -> Result<GetObjectReader> {
        with_prepared_get_object_metadata(metadata, self.get_object_reader(bucket, object, range, headers, opts)).await
    }
}

/// Get lock acquire timeout from environment variable RUSTFS_LOCK_ACQUIRE_TIMEOUT (in seconds)
/// Defaults to 30 seconds if not set or invalid
/// Lock acquisition timeout. Cached: this is consulted on every object
/// lock acquisition and `std::env::var` takes a process-global lock. In test
/// builds the env var is read directly so `temp_env` overrides take effect.
pub fn get_lock_acquire_timeout() -> Duration {
    #[cfg(test)]
    {
        Duration::from_secs(rustfs_utils::get_env_u64(
            rustfs_config::ENV_OBJECT_LOCK_ACQUIRE_TIMEOUT,
            rustfs_config::DEFAULT_OBJECT_LOCK_ACQUIRE_TIMEOUT,
        ))
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<Duration> = OnceLock::new();
        *CACHED.get_or_init(|| {
            Duration::from_secs(rustfs_utils::get_env_u64(
                rustfs_config::ENV_OBJECT_LOCK_ACQUIRE_TIMEOUT,
                rustfs_config::DEFAULT_OBJECT_LOCK_ACQUIRE_TIMEOUT,
            ))
        })
    }
}

fn get_put_object_commit_lock_acquire_timeout_override_ms() -> u64 {
    #[cfg(test)]
    {
        rustfs_utils::get_env_u64(
            rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS,
            rustfs_config::DEFAULT_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<u64> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_u64(
                rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS,
                rustfs_config::DEFAULT_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS,
            )
        })
    }
}

fn get_put_object_commit_lock_acquire_timeout(op: &'static str) -> Duration {
    let default_timeout = get_lock_acquire_timeout();
    if op != "put_object_commit" {
        return default_timeout;
    }

    let timeout_ms = get_put_object_commit_lock_acquire_timeout_override_ms();
    if timeout_ms == 0 {
        default_timeout
    } else {
        Duration::from_millis(timeout_ms)
    }
}

fn put_object_commit_lock_timeout_override_enabled(op: &'static str) -> bool {
    op == "put_object_commit" && get_put_object_commit_lock_acquire_timeout_override_ms() != 0
}

fn put_object_commit_lock_admission_budget_label() -> &'static str {
    match get_put_object_commit_lock_acquire_timeout_override_ms() {
        0 => rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_DISABLED,
        1..=250 => rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_250MS,
        251..=500 => rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_500MS,
        501..=1000 => rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_1000MS,
        _ => rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_GT_1000MS,
    }
}

fn record_put_object_commit_lock_admission(op: &'static str, outcome: &'static str) {
    if op != "put_object_commit" || !rustfs_io_metrics::put_stage_metrics_enabled() {
        return;
    }
    rustfs_io_metrics::record_put_object_commit_lock_admission(put_object_commit_lock_admission_budget_label(), outcome);
}

fn put_object_commit_lock_acquire_error_outcome(op: &'static str, err: &rustfs_lock::error::LockError) -> &'static str {
    if put_object_commit_lock_timeout_override_enabled(op) && matches!(err, rustfs_lock::error::LockError::Timeout { .. }) {
        rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_TIMEOUT_SLOWDOWN
    } else {
        rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_LOCK_ERROR
    }
}

fn resolve_put_object_commit_lock_acquire_result(
    set: &SetDisks,
    op: &'static str,
    bucket: &str,
    object: &str,
    result: std::result::Result<rustfs_lock::namespace::NamespaceLockGuard, rustfs_lock::error::LockError>,
) -> Result<rustfs_lock::namespace::NamespaceLockGuard> {
    match result {
        Ok(guard) => {
            record_put_object_commit_lock_admission(op, rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_ACQUIRED);
            Ok(guard)
        }
        Err(err) => {
            record_put_object_commit_lock_admission(op, put_object_commit_lock_acquire_error_outcome(op, &err));
            Err(map_put_object_commit_lock_acquire_error(set, op, bucket, object, err))
        }
    }
}

fn map_put_object_commit_lock_acquire_error(
    set: &SetDisks,
    op: &'static str,
    bucket: &str,
    object: &str,
    err: rustfs_lock::error::LockError,
) -> StorageError {
    if put_object_commit_lock_timeout_override_enabled(op) && matches!(err, rustfs_lock::error::LockError::Timeout { .. }) {
        StorageError::SlowDown
    } else {
        set.map_namespace_lock_error(bucket, object, "write", err)
    }
}

pub fn is_object_lock_diag_enabled() -> bool {
    let read_enabled = || {
        let enabled = rustfs_utils::get_env_bool(
            rustfs_config::ENV_OBJECT_LOCK_DIAG_ENABLE,
            rustfs_config::DEFAULT_OBJECT_LOCK_DIAG_ENABLE,
        );
        record_object_lock_diag_enabled(enabled);
        enabled
    };
    #[cfg(test)]
    {
        read_enabled()
    }
    #[cfg(not(test))]
    {
        *OBJECT_LOCK_DIAG_ENABLED.get_or_init(read_enabled)
    }
}

pub fn get_object_lock_diag_slow_acquire_threshold() -> Duration {
    Duration::from_millis(rustfs_utils::get_env_u64(
        rustfs_config::ENV_OBJECT_LOCK_DIAG_SLOW_ACQUIRE_MS,
        rustfs_config::DEFAULT_OBJECT_LOCK_DIAG_SLOW_ACQUIRE_MS,
    ))
}

pub fn get_object_lock_diag_slow_hold_threshold() -> Duration {
    Duration::from_millis(rustfs_utils::get_env_u64(
        rustfs_config::ENV_OBJECT_LOCK_DIAG_SLOW_HOLD_MS,
        rustfs_config::DEFAULT_OBJECT_LOCK_DIAG_SLOW_HOLD_MS,
    ))
}

/// Check if lock optimization is enabled.
/// Fully materialized reads release the read lock before returning. Streaming
/// reads retain it until the response body completes or is dropped.
///
/// **Note**: Cached via `OnceLock` in production — env var changes require
/// process restart. In test builds the env var is read directly so that
/// `temp_env` overrides take effect.
pub fn is_lock_optimization_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(
            rustfs_config::ENV_OBJECT_LOCK_OPTIMIZATION_ENABLE,
            rustfs_config::DEFAULT_OBJECT_LOCK_OPTIMIZATION_ENABLE,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(
                rustfs_config::ENV_OBJECT_LOCK_OPTIMIZATION_ENABLE,
                rustfs_config::DEFAULT_OBJECT_LOCK_OPTIMIZATION_ENABLE,
            )
        })
    }
}

/// Check if deadlock detection is enabled.
/// When enabled, lock operations are recorded for deadlock analysis.
///
/// **Note**: Cached via `OnceLock` — env var changes require process restart.
pub fn is_deadlock_detection_enabled() -> bool {
    static CACHED: OnceLock<bool> = OnceLock::new();
    *CACHED.get_or_init(|| {
        rustfs_utils::get_env_bool(
            rustfs_config::ENV_OBJECT_DEADLOCK_DETECTION_ENABLE,
            rustfs_config::DEFAULT_OBJECT_DEADLOCK_DETECTION_ENABLE,
        )
    })
}

// ============================================================================
// GET Optimization Flag Functions
//
// All functions use `OnceLock` for caching. Environment variable changes
// require process restart to take effect.
// ============================================================================

/// Check if multipart codec streaming is enabled.
///
/// When enabled, multipart objects use per-part codec streaming
/// instead of falling back to the legacy duplex path.
fn is_codec_streaming_multipart_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(
            ENV_RUSTFS_GET_CODEC_STREAMING_MULTIPART_ENABLE,
            DEFAULT_RUSTFS_GET_CODEC_STREAMING_MULTIPART_ENABLE,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(
                ENV_RUSTFS_GET_CODEC_STREAMING_MULTIPART_ENABLE,
                DEFAULT_RUSTFS_GET_CODEC_STREAMING_MULTIPART_ENABLE,
            )
        })
    }
}

fn get_codec_streaming_multipart_max_parts() -> usize {
    #[cfg(test)]
    {
        rustfs_utils::get_env_usize(
            ENV_RUSTFS_GET_CODEC_STREAMING_MULTIPART_MAX_PARTS,
            DEFAULT_RUSTFS_GET_CODEC_STREAMING_MULTIPART_MAX_PARTS,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<usize> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_usize(
                ENV_RUSTFS_GET_CODEC_STREAMING_MULTIPART_MAX_PARTS,
                DEFAULT_RUSTFS_GET_CODEC_STREAMING_MULTIPART_MAX_PARTS,
            )
        })
    }
}

/// Check if metadata early-stop is enabled (base flag).
///
/// **Note**: Cached via `OnceLock` in production. In test builds the env var
/// is read directly so that `temp_env` overrides take effect.
fn is_get_metadata_early_stop_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(ENV_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE, DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE)
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(ENV_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE, DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE)
        })
    }
}

/// Check if version-aware early-stop is enabled.
///
/// When enabled, versioned requests can early-stop when the requested
/// version_id reaches quorum across disks.
///
/// **Note**: Cached via `OnceLock` in production. In test builds the env var
/// is read directly so that `temp_env` overrides take effect.
fn is_version_early_stop_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(
            ENV_RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE,
            DEFAULT_RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(
                ENV_RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE,
                DEFAULT_RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE,
            )
        })
    }
}

fn is_get_metadata_data_read_early_stop_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(
            ENV_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE,
            DEFAULT_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(
                ENV_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE,
                DEFAULT_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE,
            )
        })
    }
}

fn is_get_metadata_non_inline_data_read_early_stop_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(
            ENV_RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE,
            DEFAULT_RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(
                ENV_RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE,
                DEFAULT_RUSTFS_GET_METADATA_TWO_PHASE_READ_PLAN_ENABLE,
            )
        })
    }
}

fn is_get_metadata_early_stop_bounded_fanout_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(
            ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT,
            DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(
                ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT,
                DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT,
            )
        })
    }
}

#[derive(Debug)]
struct GetMetadataSlowtailFaultConfig {
    delay: Duration,
    disks: Arc<[usize]>,
    bucket: Option<String>,
    object_prefix: Option<String>,
}

#[derive(Clone, Debug)]
struct GetMetadataSlowtailFaultRequest {
    delay: Duration,
    disks: Arc<[usize]>,
}

impl GetMetadataSlowtailFaultRequest {
    fn delay_for_disk(&self, disk_index: usize) -> Option<Duration> {
        self.disks.contains(&disk_index).then_some(self.delay)
    }
}

fn parse_get_metadata_slowtail_fault_disks(raw: &str) -> Option<Vec<usize>> {
    let mut disks = Vec::new();
    for item in raw.split(',').map(str::trim).filter(|item| !item.is_empty()) {
        let Ok(index) = item.parse::<usize>() else {
            return None;
        };
        if !disks.contains(&index) {
            disks.push(index);
        }
    }
    (!disks.is_empty()).then_some(disks)
}

fn load_get_metadata_slowtail_fault_config() -> Option<GetMetadataSlowtailFaultConfig> {
    let delay_ms = rustfs_utils::get_env_u64(ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS, 0);
    if delay_ms == 0 {
        return None;
    }
    let disks = parse_get_metadata_slowtail_fault_disks(&std::env::var(ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS).ok()?)?;
    let bucket = std::env::var(ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET)
        .ok()
        .filter(|value| !value.is_empty());
    let object_prefix = std::env::var(ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX)
        .ok()
        .filter(|value| !value.is_empty());
    Some(GetMetadataSlowtailFaultConfig {
        delay: Duration::from_millis(delay_ms),
        disks: Arc::from(disks.into_boxed_slice()),
        bucket,
        object_prefix,
    })
}

fn get_metadata_slowtail_fault_request(bucket: &str, object: &str, read_data: bool) -> Option<GetMetadataSlowtailFaultRequest> {
    if !read_data {
        return None;
    }

    #[cfg(test)]
    let config = load_get_metadata_slowtail_fault_config();
    #[cfg(test)]
    let config = config.as_ref()?;
    #[cfg(not(test))]
    let config = ({
        static CACHED: OnceLock<Option<GetMetadataSlowtailFaultConfig>> = OnceLock::new();
        CACHED.get_or_init(load_get_metadata_slowtail_fault_config).as_ref()
    })?;

    if let Some(expected_bucket) = &config.bucket
        && expected_bucket != bucket
    {
        return None;
    }
    if let Some(expected_prefix) = &config.object_prefix
        && !object.starts_with(expected_prefix)
    {
        return None;
    }
    Some(GetMetadataSlowtailFaultRequest {
        delay: config.delay,
        disks: config.disks.clone(),
    })
}

#[cfg(test)]
fn get_metadata_slowtail_fault_delay(bucket: &str, object: &str, disk_index: usize, read_data: bool) -> Option<Duration> {
    get_metadata_slowtail_fault_request(bucket, object, read_data)?.delay_for_disk(disk_index)
}

/// Check if multipart reads prefetch the next part's bitrot reader setup
/// while the current part decodes (backlog#870).
///
/// **Note**: Cached via `OnceLock` in production. In test builds the env var
/// is read directly so that `temp_env` overrides take effect.
fn is_multipart_reader_setup_prefetch_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(
            ENV_RUSTFS_GET_MULTIPART_READER_SETUP_PREFETCH,
            DEFAULT_RUSTFS_GET_MULTIPART_READER_SETUP_PREFETCH,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(
                ENV_RUSTFS_GET_MULTIPART_READER_SETUP_PREFETCH,
                DEFAULT_RUSTFS_GET_MULTIPART_READER_SETUP_PREFETCH,
            )
        })
    }
}

#[allow(
    dead_code,
    reason = "percentage-rollout facet of the metadata early-stop switch; its predicate has no caller while the sibling enable flag is live (backlog#1823)"
)]
fn get_metadata_early_stop_rollout_pct() -> u32 {
    static CACHED: OnceLock<u32> = OnceLock::new();
    *CACHED.get_or_init(|| {
        rustfs_utils::get_env_u32(
            ENV_RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT,
            DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT,
        )
    })
}

// --- Request-Level Decision Functions ---

/// Determine if an optimization should be enabled for a specific request.
///
/// Uses a stable hash of `(bucket, object)` to ensure the same object
/// always gets consistent behavior. This enables percentage-based gradual rollout.
fn is_optimization_enabled_for_request(base_enabled: bool, rollout_pct: u32, bucket: &str, object: &str) -> bool {
    if !base_enabled || rollout_pct == 0 {
        return false;
    }
    if rollout_pct >= 100 {
        return true;
    }

    // Stable hash: same (bucket, object) always produces the same result
    use std::hash::{Hash, Hasher};
    let mut hasher = std::collections::hash_map::DefaultHasher::new();
    bucket.hash(&mut hasher);
    object.hash(&mut hasher);
    let hash = hasher.finish() % 100;

    (hash as u32) < rollout_pct
}
/// Should this specific request use codec streaming?
fn should_use_codec_streaming(config: GetCodecStreamingConfig, bucket: &str, object: &str) -> bool {
    is_optimization_enabled_for_request(config.enabled, config.rollout_pct, bucket, object)
}

pub(in crate::set_disk) fn codec_streaming_rollout_applies(bucket: &str, object: &str) -> bool {
    let config = get_codec_streaming_config();
    config.enabled
        && config.body_compat_confirmed
        && config.header_compat_confirmed
        && config.rollout.is_opted_in()
        && should_use_codec_streaming(config, bucket, object)
}

/// Should this specific request use metadata early-stop?
#[allow(
    dead_code,
    reason = "percentage-rollout facet of the metadata early-stop switch; its predicate has no caller while the sibling enable flag is live (backlog#1823)"
)]
pub fn should_use_metadata_early_stop(bucket: &str, object: &str) -> bool {
    let base = is_get_metadata_early_stop_enabled();
    let pct = get_metadata_early_stop_rollout_pct();
    is_optimization_enabled_for_request(base, pct, bucket, object)
}

fn is_get_codec_streaming_data_blocks_first_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(
            ENV_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_ENABLE,
            DEFAULT_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_ENABLE,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(
                ENV_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_ENABLE,
                DEFAULT_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_ENABLE,
            )
        })
    }
}

fn get_codec_streaming_data_blocks_first_max_size() -> usize {
    #[cfg(test)]
    {
        rustfs_utils::get_env_usize(
            ENV_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_MAX_SIZE,
            DEFAULT_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_MAX_SIZE,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<usize> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_usize(
                ENV_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_MAX_SIZE,
                DEFAULT_RUSTFS_GET_CODEC_STREAMING_DATA_BLOCKS_FIRST_MAX_SIZE,
            )
        })
    }
}

fn get_object_metadata_cache_max_entries() -> usize {
    #[cfg(test)]
    {
        rustfs_utils::get_env_usize(
            ENV_RUSTFS_GET_OBJECT_METADATA_CACHE_MAX_ENTRIES,
            DEFAULT_GET_OBJECT_METADATA_CACHE_MAX_ENTRIES,
        )
        .max(1)
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<usize> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_usize(
                ENV_RUSTFS_GET_OBJECT_METADATA_CACHE_MAX_ENTRIES,
                DEFAULT_GET_OBJECT_METADATA_CACHE_MAX_ENTRIES,
            )
            .max(1)
        })
    }
}

fn is_get_small_object_direct_memory_enabled() -> bool {
    #[cfg(test)]
    {
        rustfs_utils::get_env_bool(ENV_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY, DEFAULT_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY)
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<bool> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_bool(ENV_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY, DEFAULT_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY)
        })
    }
}

fn get_small_object_direct_memory_threshold() -> usize {
    #[cfg(test)]
    {
        rustfs_utils::get_env_usize(
            ENV_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY_THRESHOLD,
            DEFAULT_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY_THRESHOLD,
        )
    }
    #[cfg(not(test))]
    {
        static CACHED: OnceLock<usize> = OnceLock::new();
        *CACHED.get_or_init(|| {
            rustfs_utils::get_env_usize(
                ENV_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY_THRESHOLD,
                DEFAULT_RUSTFS_GET_SMALL_OBJECT_DIRECT_MEMORY_THRESHOLD,
            )
        })
    }
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum GetCodecStreamingEngine {
    Legacy,
    Rustfs,
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
struct GetCodecStreamingConfig {
    enabled: bool,
    rollout: GetCodecStreamingRollout,
    rollout_pct: u32,
    body_compat_confirmed: bool,
    header_compat_confirmed: bool,
    engine: GetCodecStreamingEngine,
    min_size: usize,
}

fn parse_get_codec_streaming_engine(engine: &str) -> GetCodecStreamingEngine {
    match engine.trim() {
        value if value.eq_ignore_ascii_case(GET_CODEC_STREAMING_ENGINE_RUSTFS) => GetCodecStreamingEngine::Rustfs,
        value if value.eq_ignore_ascii_case(GET_CODEC_STREAMING_ENGINE_LEGACY) => GetCodecStreamingEngine::Legacy,
        _ => GetCodecStreamingEngine::Legacy,
    }
}

fn parse_get_codec_streaming_rollout(rollout: &str) -> GetCodecStreamingRollout {
    match rollout.trim() {
        // Clean production token. `internal`/`benchmark` remain accepted aliases
        // for backward compatibility; all three opt the fast path in.
        value
            if value.eq_ignore_ascii_case("on")
                || value.eq_ignore_ascii_case("full")
                || value.eq_ignore_ascii_case("production") =>
        {
            GetCodecStreamingRollout::On
        }
        value if value.eq_ignore_ascii_case("internal") => GetCodecStreamingRollout::Internal,
        value if value.eq_ignore_ascii_case("benchmark") => GetCodecStreamingRollout::Benchmark,
        _ => GetCodecStreamingRollout::Off,
    }
}

fn load_get_codec_streaming_config() -> GetCodecStreamingConfig {
    let engine = parse_get_codec_streaming_engine(&rustfs_utils::get_env_str(
        ENV_RUSTFS_GET_CODEC_STREAMING_ENGINE,
        DEFAULT_RUSTFS_GET_CODEC_STREAMING_ENGINE,
    ));
    let min_size = if std::env::var_os(ENV_RUSTFS_GET_CODEC_STREAMING_MIN_SIZE).is_some() {
        rustfs_utils::get_env_usize(ENV_RUSTFS_GET_CODEC_STREAMING_MIN_SIZE, DEFAULT_RUSTFS_GET_CODEC_STREAMING_MIN_SIZE)
    } else {
        match engine {
            GetCodecStreamingEngine::Rustfs => rustfs_utils::get_env_usize(
                ENV_RUSTFS_GET_CODEC_STREAMING_RUSTFS_MIN_SIZE,
                DEFAULT_RUSTFS_GET_CODEC_STREAMING_RUSTFS_MIN_SIZE,
            ),
            GetCodecStreamingEngine::Legacy => DEFAULT_RUSTFS_GET_CODEC_STREAMING_MIN_SIZE,
        }
    };

    GetCodecStreamingConfig {
        enabled: rustfs_utils::get_env_bool(ENV_RUSTFS_GET_CODEC_STREAMING_ENABLE, DEFAULT_RUSTFS_GET_CODEC_STREAMING_ENABLE),
        rollout: parse_get_codec_streaming_rollout(&rustfs_utils::get_env_str(
            ENV_RUSTFS_GET_CODEC_STREAMING_ROLLOUT,
            DEFAULT_RUSTFS_GET_CODEC_STREAMING_ROLLOUT,
        )),
        rollout_pct: rustfs_utils::get_env_u32(
            ENV_RUSTFS_GET_CODEC_STREAMING_ROLLOUT_PCT,
            DEFAULT_RUSTFS_GET_CODEC_STREAMING_ROLLOUT_PCT,
        ),
        body_compat_confirmed: rustfs_utils::get_env_bool(ENV_RUSTFS_GET_CODEC_STREAMING_BODY_COMPAT_CONFIRMED, true),
        header_compat_confirmed: rustfs_utils::get_env_bool(ENV_RUSTFS_GET_CODEC_STREAMING_HEADER_COMPAT_CONFIRMED, true),
        engine,
        min_size,
    }
}

fn get_codec_streaming_config_cached_core(load: impl FnOnce() -> GetCodecStreamingConfig) -> GetCodecStreamingConfig {
    static CACHED: OnceLock<GetCodecStreamingConfig> = OnceLock::new();
    *CACHED.get_or_init(load)
}

fn get_codec_streaming_config() -> GetCodecStreamingConfig {
    #[cfg(test)]
    {
        load_get_codec_streaming_config()
    }
    #[cfg(not(test))]
    {
        get_codec_streaming_config_cached_core(load_get_codec_streaming_config)
    }
}

/// Shared kill-switch and compatibility guard for all non-duplex GET readers.
/// Size-specific readers may have independent rollout gates, but they must not
/// bypass these emergency controls.
pub(super) fn is_get_codec_streaming_base_enabled() -> bool {
    let config = get_codec_streaming_config();
    config.enabled && config.body_compat_confirmed && config.header_compat_confirmed
}

fn get_codec_streaming_engine() -> GetCodecStreamingEngine {
    get_codec_streaming_config().engine
}

fn build_get_codec_streaming_decode_engine(erasure: coding::Erasure) -> std::io::Result<CodecStreamingDecodeEngine> {
    match get_codec_streaming_engine() {
        GetCodecStreamingEngine::Legacy => Ok(CodecStreamingDecodeEngine::legacy(erasure)),
        GetCodecStreamingEngine::Rustfs => CodecStreamingDecodeEngine::rustfs(&erasure),
    }
}

fn get_codec_streaming_metrics_path() -> &'static str {
    match get_codec_streaming_engine() {
        GetCodecStreamingEngine::Legacy => GET_OBJECT_PATH_CODEC_STREAMING_LEGACY_ENGINE,
        GetCodecStreamingEngine::Rustfs => GET_OBJECT_PATH_CODEC_STREAMING_RUSTFS_ENGINE,
    }
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum GetCodecStreamingDecision {
    Use,
    Fallback(GetCodecStreamingFallbackReason),
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum GetCodecStreamingRollout {
    /// Default. Codec streaming disabled; GET uses the legacy duplex path.
    Off,
    /// Production enablement token (`on`/`full`/`production`).
    On,
    /// Backward-compatible aliases that also opt the fast path in.
    Internal,
    Benchmark,
}

impl GetCodecStreamingRollout {
    const fn is_opted_in(self) -> bool {
        !matches!(self, Self::Off)
    }
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum GetCodecStreamingFallbackReason {
    Disabled,
    RolloutNotOptedIn,
    RolloutPctNotSelected,
    BodyCompatibilityUnconfirmed,
    HeaderCompatibilityUnconfirmed,
    LockOptimizationDisabled,
    Range,
    PartNumber,
    BelowMinSize,
    Encrypted,
    Compressed,
    Remote,
    Multipart,
    InvalidMinSize,
    ReadQuorumNotSafe,
    MultipartPartLimit,
    CopySourceDemandBound,
    InvalidMetadataShape,
}

impl GetCodecStreamingFallbackReason {
    const fn as_str(self) -> &'static str {
        match self {
            Self::Disabled => "disabled",
            Self::RolloutNotOptedIn => "rollout_not_opted_in",
            Self::RolloutPctNotSelected => "rollout_pct_not_selected",
            Self::BodyCompatibilityUnconfirmed => "body_compatibility_unconfirmed",
            Self::HeaderCompatibilityUnconfirmed => "header_compatibility_unconfirmed",
            Self::LockOptimizationDisabled => "lock_optimization_disabled",
            Self::Range => "range",
            Self::PartNumber => "part_number",
            Self::BelowMinSize => "below_min_size",
            Self::Encrypted => "encrypted",
            Self::Compressed => "compressed",
            Self::Remote => "remote",
            Self::Multipart => "multipart",
            Self::InvalidMinSize => "invalid_min_size",
            Self::ReadQuorumNotSafe => "read_quorum_not_safe",
            Self::MultipartPartLimit => "multipart_part_limit",
            Self::CopySourceDemandBound => "copy_source_demand_bound",
            Self::InvalidMetadataShape => "invalid_metadata_shape",
        }
    }
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum GetCodecStreamingObjectClass {
    PlainSinglePart,
    Range,
    Encrypted,
    Compressed,
    Remote,
    Multipart,
}

impl GetCodecStreamingObjectClass {
    const fn as_str(self) -> &'static str {
        match self {
            Self::PlainSinglePart => crate::diagnostics::get::GET_CODEC_STREAMING_OBJECT_CLASS_PLAIN_SINGLE_PART,
            Self::Range => crate::diagnostics::get::GET_CODEC_STREAMING_OBJECT_CLASS_RANGE,
            Self::Encrypted => crate::diagnostics::get::GET_CODEC_STREAMING_OBJECT_CLASS_ENCRYPTED,
            Self::Compressed => crate::diagnostics::get::GET_CODEC_STREAMING_OBJECT_CLASS_COMPRESSED,
            Self::Remote => crate::diagnostics::get::GET_CODEC_STREAMING_OBJECT_CLASS_REMOTE,
            Self::Multipart => crate::diagnostics::get::GET_CODEC_STREAMING_OBJECT_CLASS_MULTIPART,
        }
    }
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
struct GetCodecStreamingGate {
    object_class: GetCodecStreamingObjectClass,
    decision: GetCodecStreamingDecision,
    prefer_data_blocks_first_reader_setup: bool,
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum GetDirectMemoryDecision {
    Use { object_size: usize },
    Fallback(GetDirectMemoryFallbackReason),
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum GetDirectMemoryFallbackReason {
    Disabled,
    ThresholdZero,
    Range,
    PartNumber,
    VersionId,
    InclFreeVersions,
    SkipFreeVersion,
    DataMovement,
    RawDataMovementRead,
    DeleteMarker,
    MetadataOnly,
    VersionOnly,
    Encrypted,
    Compressed,
    Remote,
    ObjectInfoMultipart,
    FileInfoMultipart,
    MetadataShape,
    InvalidSize,
    SizeMismatch,
    AboveThreshold,
}

impl GetDirectMemoryFallbackReason {
    const fn as_str(self) -> &'static str {
        match self {
            Self::Disabled => "disabled",
            Self::ThresholdZero => "threshold_zero",
            Self::Range => "range",
            Self::PartNumber => "part_number",
            Self::VersionId => "version_id",
            Self::InclFreeVersions => "incl_free_versions",
            Self::SkipFreeVersion => "skip_free_version",
            Self::DataMovement => "data_movement",
            Self::RawDataMovementRead => "raw_data_movement_read",
            Self::DeleteMarker => "delete_marker",
            Self::MetadataOnly => "metadata_only",
            Self::VersionOnly => "version_only",
            Self::Encrypted => "encrypted",
            Self::Compressed => "compressed",
            Self::Remote => "remote",
            Self::ObjectInfoMultipart => "object_info_multipart",
            Self::FileInfoMultipart => "file_info_multipart",
            Self::MetadataShape => "metadata_shape",
            Self::InvalidSize => "invalid_size",
            Self::SizeMismatch => "size_mismatch",
            Self::AboveThreshold => "above_threshold",
        }
    }
}

fn record_get_codec_streaming_gate_decision(
    object_class: GetCodecStreamingObjectClass,
    decision: GetCodecStreamingDecision,
    size_bucket: &'static str,
) {
    let (outcome, reason) = match decision {
        GetCodecStreamingDecision::Use => (
            crate::diagnostics::get::GET_CODEC_STREAMING_DECISION_USE,
            crate::diagnostics::get::GET_CODEC_STREAMING_REASON_NONE,
        ),
        GetCodecStreamingDecision::Fallback(reason) => {
            (crate::diagnostics::get::GET_CODEC_STREAMING_DECISION_FALLBACK, reason.as_str())
        }
    };
    let object_class = object_class.as_str();
    rustfs_io_metrics::record_get_object_codec_streaming_decision(outcome, object_class, reason);
    rustfs_io_metrics::record_get_object_codec_streaming_decision_by_size(outcome, object_class, reason, size_bucket);
}

fn record_get_direct_memory_decision(
    object_class: GetCodecStreamingObjectClass,
    decision: GetDirectMemoryDecision,
    size_bucket: &'static str,
) {
    let (outcome, reason) = match decision {
        GetDirectMemoryDecision::Use { .. } => (
            crate::diagnostics::get::GET_DIRECT_MEMORY_DECISION_USE,
            crate::diagnostics::get::GET_DIRECT_MEMORY_REASON_NONE,
        ),
        GetDirectMemoryDecision::Fallback(reason) => {
            (crate::diagnostics::get::GET_DIRECT_MEMORY_DECISION_FALLBACK, reason.as_str())
        }
    };
    rustfs_io_metrics::record_get_object_direct_memory_decision(outcome, object_class.as_str(), reason, size_bucket);
}

fn record_get_object_reader_path_observation(
    path: &'static str,
    object_class: GetCodecStreamingObjectClass,
    size_bucket: &'static str,
) {
    #[cfg(test)]
    LAST_GET_OBJECT_READER_PATH.store(
        match path {
            crate::set_disk::read::GET_OBJECT_PATH_MID_SIZE_STREAMING => 1,
            GET_OBJECT_PATH_DIRECT_MEMORY => 2,
            GET_OBJECT_PATH_INLINE_DIRECT => 3,
            GET_OBJECT_PATH_BODY_CACHE => 4,
            GET_OBJECT_PATH_CODEC_STREAMING => 5,
            GET_OBJECT_PATH_REMOTE_TRANSITION => 6,
            GET_OBJECT_PATH_EMPTY => 7,
            GET_OBJECT_PATH_LEGACY_DUPLEX => 8,
            _ => 255,
        },
        Ordering::Relaxed,
    );
    rustfs_io_metrics::record_get_object_reader_path(path);
    rustfs_io_metrics::record_get_object_reader_path_by_size(path, object_class.as_str(), size_bucket);
}

#[cfg(test)]
static LAST_GET_OBJECT_READER_PATH: AtomicU64 = AtomicU64::new(0);

#[cfg(test)]
pub(crate) fn reset_test_get_object_reader_path() {
    LAST_GET_OBJECT_READER_PATH.store(0, Ordering::Relaxed);
}

#[cfg(test)]
pub(crate) fn test_get_object_reader_selected_mid_size() -> bool {
    LAST_GET_OBJECT_READER_PATH.load(Ordering::Relaxed) == 1
}

#[cfg(test)]
pub(crate) fn test_get_object_reader_path_id() -> u64 {
    LAST_GET_OBJECT_READER_PATH.load(Ordering::Relaxed)
}

fn classify_get_codec_streaming_object_class(
    range: &Option<HTTPRangeSpec>,
    object_info: &ObjectInfo,
    fi: &FileInfo,
) -> GetCodecStreamingObjectClass {
    if range.is_some() {
        return GetCodecStreamingObjectClass::Range;
    }
    if object_info.is_encrypted() {
        return GetCodecStreamingObjectClass::Encrypted;
    }
    if object_info.is_compressed() || fi.is_compressed() {
        return GetCodecStreamingObjectClass::Compressed;
    }
    if object_info.is_remote() {
        return GetCodecStreamingObjectClass::Remote;
    }
    if fi.parts.len() != 1 {
        return GetCodecStreamingObjectClass::Multipart;
    }
    GetCodecStreamingObjectClass::PlainSinglePart
}

#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
fn is_get_small_object_direct_memory_eligible_with_threshold(
    range: &Option<HTTPRangeSpec>,
    object_info: &ObjectInfo,
    fi: &FileInfo,
    opts: &ObjectOptions,
    threshold: usize,
) -> bool {
    matches!(
        get_small_object_direct_memory_decision_with_threshold(range, object_info, fi, opts, true, threshold),
        GetDirectMemoryDecision::Use { .. }
    )
}

fn get_small_object_direct_memory_decision_with_threshold(
    range: &Option<HTTPRangeSpec>,
    object_info: &ObjectInfo,
    fi: &FileInfo,
    opts: &ObjectOptions,
    enabled: bool,
    threshold: usize,
) -> GetDirectMemoryDecision {
    get_small_object_direct_memory_decision_with_threshold_and_plan(
        range,
        object_info,
        fi,
        opts,
        enabled,
        threshold,
        ReadPathPlan::new(object_info, fi),
    )
}

fn get_small_object_direct_memory_decision_with_threshold_and_plan(
    range: &Option<HTTPRangeSpec>,
    object_info: &ObjectInfo,
    fi: &FileInfo,
    opts: &ObjectOptions,
    enabled: bool,
    threshold: usize,
    plan: ReadPathPlan,
) -> GetDirectMemoryDecision {
    if !enabled {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Disabled);
    }
    if threshold == 0 {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::ThresholdZero);
    }
    if range.is_some() {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Range);
    }
    if opts.part_number.is_some() {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::PartNumber);
    }
    if opts.version_id.is_some() {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::VersionId);
    }
    // Bucket-level versioning no longer blocks the inline path (rustfs/backlog#1802):
    // `fi` here is the already-resolved target version, so reassembling its inlined
    // data shards is correct whether the bucket is versioned or not. This direct-memory
    // decision still falls back for an explicit versionId (the `version_id` check above);
    // a delete-marker latest is rejected below.
    if opts.incl_free_versions {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::InclFreeVersions);
    }
    if opts.skip_free_version {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::SkipFreeVersion);
    }
    if opts.data_movement {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::DataMovement);
    }
    if opts.raw_data_movement_read {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::RawDataMovementRead);
    }
    if object_info.delete_marker {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::DeleteMarker);
    }
    if object_info.metadata_only {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::MetadataOnly);
    }
    if object_info.version_only {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::VersionOnly);
    }
    if object_info.is_encrypted() {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Encrypted);
    }
    if object_info.is_compressed() {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Compressed);
    }
    if object_info.is_remote() {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Remote);
    }
    if object_info.parts.len() != 1 {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::ObjectInfoMultipart);
    }
    if fi.parts.len() != 1 {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::FileInfoMultipart);
    }

    if object_info.size != fi.size {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::SizeMismatch);
    }
    let Some(shape) = plan.shape() else {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::MetadataShape);
    };
    let object_size = shape.object_size;
    if object_size == 0 {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::InvalidSize);
    }
    if !plan.is_plain() {
        if object_info.is_encrypted() || fi.metadata.keys().any(|key| is_object_encryption_marker(key)) {
            return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Encrypted);
        }
        if object_info.is_compressed() || fi.is_compressed() {
            return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Compressed);
        }
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Remote);
    }
    if object_size > threshold {
        return GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::AboveThreshold);
    }

    GetDirectMemoryDecision::Use { object_size }
}

#[allow(dead_code, reason = "asserted by this file's gate tests")]
fn get_small_object_direct_memory_decision(
    range: &Option<HTTPRangeSpec>,
    object_info: &ObjectInfo,
    fi: &FileInfo,
    opts: &ObjectOptions,
) -> GetDirectMemoryDecision {
    get_small_object_direct_memory_decision_with_threshold(
        range,
        object_info,
        fi,
        opts,
        is_get_small_object_direct_memory_enabled(),
        get_small_object_direct_memory_threshold(),
    )
}

fn should_prefer_codec_streaming_data_blocks_first_reader_setup(
    object_class: GetCodecStreamingObjectClass,
    object_size: i64,
) -> bool {
    if !is_get_codec_streaming_data_blocks_first_enabled()
        || object_class != GetCodecStreamingObjectClass::PlainSinglePart
        || object_size <= 0
    {
        return false;
    }

    let Ok(object_size) = usize::try_from(object_size) else {
        return false;
    };
    let max_size = get_codec_streaming_data_blocks_first_max_size();
    max_size > 0 && object_size <= max_size
}

#[allow(dead_code, reason = "asserted by this file's gate tests")]
fn get_codec_streaming_reader_gate(
    bucket: &str,
    object: &str,
    part_number: Option<usize>,
    object_class: GetCodecStreamingObjectClass,
    object_info: &ObjectInfo,
    fi: &FileInfo,
    lock_optimization_enabled: bool,
) -> GetCodecStreamingGate {
    get_codec_streaming_reader_gate_with_plan(
        bucket,
        object,
        part_number,
        object_class,
        object_info,
        fi,
        lock_optimization_enabled,
        ReadPathPlan::new(object_info, fi),
    )
}

#[allow(clippy::too_many_arguments, reason = "keeps the hot-path gate inputs explicit")]
fn get_codec_streaming_reader_gate_with_plan(
    bucket: &str,
    object: &str,
    part_number: Option<usize>,
    object_class: GetCodecStreamingObjectClass,
    object_info: &ObjectInfo,
    fi: &FileInfo,
    lock_optimization_enabled: bool,
    plan: ReadPathPlan,
) -> GetCodecStreamingGate {
    let config = get_codec_streaming_config();

    if !config.enabled {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::Disabled),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if matches!(get_object_read_policy(), GetObjectReadPolicy::CopySource) {
        // The codec reader has its own bounded fill worker.  It may still
        // request an additional stripe for a plain single-part object even
        // when multipart setup prefetch is disabled, so copy sources use the
        // legacy demand-bound reader for every object class.
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::CopySourceDemandBound),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if !config.rollout.is_opted_in() {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::RolloutNotOptedIn),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if !should_use_codec_streaming(config, bucket, object) {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::RolloutPctNotSelected),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if !config.body_compat_confirmed {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::BodyCompatibilityUnconfirmed),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if !config.header_compat_confirmed {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::HeaderCompatibilityUnconfirmed),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if object_class == GetCodecStreamingObjectClass::Range {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::Range),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    // A partNumber GET arrives with `range == None`, so it is not caught by the
    // Range class above, yet it still requires a non-zero storage offset/length
    // (synthesized from the part size). The codec-streaming path builds a
    // full-object reader and drops the offset/length returned by
    // `GetObjectReader::new`, so partNumber >= 2 would stream the whole object.
    // Mirror the direct-memory part_number fallback and route these requests back
    // to the legacy duplex path, which applies the offset/length correctly.
    if part_number.is_some() {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::PartNumber),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if !lock_optimization_enabled {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::LockOptimizationDisabled),
            prefer_data_blocks_first_reader_setup: false,
        };
    }

    if object_class == GetCodecStreamingObjectClass::Encrypted {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::Encrypted),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if object_class == GetCodecStreamingObjectClass::Compressed {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::Compressed),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if object_class == GetCodecStreamingObjectClass::Remote {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::Remote),
            prefer_data_blocks_first_reader_setup: false,
        };
    }
    if object_class == GetCodecStreamingObjectClass::Multipart {
        if !is_codec_streaming_multipart_enabled() {
            return GetCodecStreamingGate {
                object_class,
                decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::Multipart),
                prefer_data_blocks_first_reader_setup: false,
            };
        }
        if fi.parts.len() > get_codec_streaming_multipart_max_parts() {
            return GetCodecStreamingGate {
                object_class,
                decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::MultipartPartLimit),
                prefer_data_blocks_first_reader_setup: false,
            };
        }
    }
    if object_class == GetCodecStreamingObjectClass::PlainSinglePart {
        let Some(_shape) = plan.shape() else {
            return GetCodecStreamingGate {
                object_class,
                decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::InvalidMetadataShape),
                prefer_data_blocks_first_reader_setup: false,
            };
        };
        if !plan.is_plain() {
            return GetCodecStreamingGate {
                object_class,
                decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::InvalidMetadataShape),
                prefer_data_blocks_first_reader_setup: false,
            };
        }
    }
    let Ok(min_size) = i64::try_from(config.min_size) else {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::InvalidMinSize),
            prefer_data_blocks_first_reader_setup: false,
        };
    };
    if object_info.size < min_size {
        return GetCodecStreamingGate {
            object_class,
            decision: GetCodecStreamingDecision::Fallback(GetCodecStreamingFallbackReason::BelowMinSize),
            prefer_data_blocks_first_reader_setup: false,
        };
    }

    GetCodecStreamingGate {
        object_class,
        decision: GetCodecStreamingDecision::Use,
        prefer_data_blocks_first_reader_setup: should_prefer_codec_streaming_data_blocks_first_reader_setup(
            object_class,
            object_info.size,
        ),
    }
}

fn is_confirmed_complete_part_missing(err: &str) -> bool {
    err.contains("file not found")
        || err.contains("Specified part could not be found")
        || (err.starts_with("part.") && err.ends_with(" not found"))
}

fn complete_multipart_part_error(part_number: usize, err: &str, bucket: &str, object: &str) -> Error {
    if is_confirmed_complete_part_missing(err) {
        return Error::InvalidPart(part_number, bucket.to_owned(), object.to_owned());
    }

    to_object_err(Error::ErasureReadQuorum, vec![bucket, object])
}

fn complete_multipart_part_error_result(err: &Error) -> &'static str {
    match err {
        Error::InvalidPart(_, _, _) => COMPLETE_MULTIPART_PART_MISSING,
        Error::ErasureReadQuorum | Error::InsufficientReadQuorum(_, _) => COMPLETE_MULTIPART_PART_READ_QUORUM_UNAVAILABLE,
        _ => COMPLETE_MULTIPART_PART_ERROR,
    }
}

/// Record a lock acquisition for deadlock detection.
/// This records detailed lock information for deadlock analysis.
/// Returns the lock_id for later release tracking.
#[inline]
fn record_lock_acquire(bucket: &str, object: &str, lock_type: &str) -> String {
    let lock_id = format!("{}:{}", bucket, object);

    if !is_deadlock_detection_enabled() {
        return lock_id;
    }

    let request_id = format!("get-{}-{}", bucket, object);
    let resource = format!("{}/{}", bucket, object);

    // Log with structured fields for analysis
    debug!(
        request_id = %request_id,
        lock_id = %lock_id,
        lock_type = %lock_type,
        resource = %resource,
        "Lock acquired for deadlock tracking"
    );

    lock_id
}

/// Record a lock release for deadlock detection.
#[inline]
fn record_lock_release(bucket: &str, object: &str, lock_id: &str, lock_type: &str) {
    if !is_deadlock_detection_enabled() {
        return;
    }

    let request_id = format!("get-{}-{}", bucket, object);

    debug!(
        request_id = %request_id,
        lock_id = %lock_id,
        lock_type = %lock_type,
        "Lock released for deadlock tracking"
    );
}

#[derive(Clone, Copy, Debug)]
pub(super) struct MultipartWriteQuorumContext<'a> {
    stage: &'static str,
    bucket: &'a str,
    object: &'a str,
    upload_id: &'a str,
    part_number: Option<usize>,
}

fn log_multipart_write_quorum_failure(
    context: MultipartWriteQuorumContext<'_>,
    errs: &[Option<DiskError>],
    write_quorum: usize,
    returned_error: &DiskError,
) {
    let summary = build_write_quorum_failure_summary(errs, OBJECT_OP_IGNORED_ERRS, write_quorum);
    runtime_sources::record_erasure_write_quorum_failure(context.stage, summary.dominant_error_label);
    warn!(
        target: "rustfs_ecstore::set_disk",
        event = EVENT_SET_DISK_MULTIPART,
        component = LOG_COMPONENT_ECSTORE,
        subsystem = LOG_SUBSYSTEM_SET_DISK,
        op = "upload_part",
        state = "write_quorum_unavailable",
        stage = context.stage,
        bucket = %context.bucket,
        object = %context.object,
        upload_id = %context.upload_id,
        part_number = context.part_number,
        required = summary.required,
        achieved = summary.achieved,
        failed = summary.failed,
        total = summary.total,
        offline_disks = summary.offline_disks,
        retryable_failures = summary.retryable_failures,
        dominant_error = summary.dominant_error_label,
        returned_error = %returned_error,
        "Set disk multipart write quorum unavailable"
    );
}

fn issue3031_diag_enabled() -> bool {
    rustfs_utils::get_env_bool(ENV_ISSUE3031_DIAG_ENABLE, false)
}

#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub(super) struct WriteLayout {
    data_drives: usize,
    parity_drives: usize,
    write_quorum: usize,
}

impl WriteLayout {
    fn from_parity(drive_count: usize, parity_drives: usize) -> Result<Self> {
        let max_parity = drive_count / 2;
        if parity_drives > max_parity {
            return Err(Error::other(format!(
                "write parity {parity_drives} exceeds the maximum {max_parity} for {drive_count} drives"
            )));
        }

        let data_drives = drive_count
            .checked_sub(parity_drives)
            .filter(|&data_drives| data_drives > 0 && parity_drives <= data_drives)
            .ok_or_else(|| Error::other(format!("invalid write layout with {drive_count} drives and parity {parity_drives}")))?;
        let write_quorum = data_drives
            .checked_add(usize::from(data_drives == parity_drives))
            .filter(|&write_quorum| write_quorum <= drive_count)
            .ok_or_else(|| Error::other(format!("invalid write quorum for {drive_count} drives and parity {parity_drives}")))?;

        Ok(Self {
            data_drives,
            parity_drives,
            write_quorum,
        })
    }
}

pub(super) fn resolve_write_layout(
    config: &storageclass::Config,
    pool_index: usize,
    drive_count: usize,
    fallback_parity: usize,
    storage_class: Option<&str>,
    max_parity: bool,
) -> Result<WriteLayout> {
    let configured_parity = if config.is_initialized() {
        config
            .parity_for_pool(storage_class.unwrap_or_default(), pool_index, drive_count)
            .ok_or_else(|| {
                Error::other(format!("storage class layout does not match pool {pool_index} with {drive_count} drives"))
            })?
    } else {
        fallback_parity
    };
    let parity_drives = if max_parity { drive_count / 2 } else { configured_parity };

    WriteLayout::from_parity(drive_count, parity_drives)
}

#[cfg(test)]
mod write_layout_tests {
    use super::{WriteLayout, resolve_write_layout};
    use crate::config::storageclass::{
        CLASS_RRS, CLASS_STANDARD, INLINE_BLOCK_ENV, OPTIMIZE_ENV, RRS, RRS_ENV, STANDARD_ENV, lookup_config_for_pools,
        lookup_config_for_pools_without_env,
    };
    use arc_swap::ArcSwap;
    use rustfs_config::server_config::KVS;
    use std::sync::Arc;

    #[test]
    fn automatic_standard_layout_is_resolved_per_pool() {
        let config = lookup_config_for_pools_without_env(&KVS::new(), &[4, 2])
            .expect("automatic storage class should resolve for both pools");

        assert_eq!(
            resolve_write_layout(&config, 0, 4, 2, None, false).expect("first pool should resolve"),
            WriteLayout {
                data_drives: 2,
                parity_drives: 2,
                write_quorum: 3,
            }
        );
        assert_eq!(
            resolve_write_layout(&config, 1, 2, 1, None, false).expect("second pool should resolve"),
            WriteLayout {
                data_drives: 1,
                parity_drives: 1,
                write_quorum: 2,
            }
        );
    }

    #[test]
    fn reduced_redundancy_layout_allows_single_disk_zero_parity() {
        let config =
            lookup_config_for_pools_without_env(&KVS::new(), &[4, 1]).expect("reduced redundancy should resolve for both pools");

        assert_eq!(
            resolve_write_layout(&config, 0, 4, 2, Some(RRS), false).expect("four-drive RRS pool should resolve"),
            WriteLayout {
                data_drives: 3,
                parity_drives: 1,
                write_quorum: 3,
            }
        );
        assert_eq!(
            resolve_write_layout(&config, 1, 1, 0, Some(RRS), false).expect("single-drive RRS pool should resolve"),
            WriteLayout {
                data_drives: 1,
                parity_drives: 0,
                write_quorum: 1,
            }
        );
    }

    #[test]
    fn write_layout_rejects_unknown_topology_and_invalid_parity() {
        let config = lookup_config_for_pools_without_env(&KVS::new(), &[4, 2]).expect("test storage class should resolve");

        assert!(resolve_write_layout(&config, 2, 2, 1, None, false).is_err());
        assert!(resolve_write_layout(&config, 1, 4, 2, None, false).is_err());
        assert!(WriteLayout::from_parity(4, 3).is_err());
        assert!(WriteLayout::from_parity(0, 0).is_err());

        let mut zero_parity_kvs = KVS::new();
        zero_parity_kvs.insert(CLASS_STANDARD.to_string(), "EC:0".to_string());
        zero_parity_kvs.insert(CLASS_RRS.to_string(), "EC:0".to_string());
        let zero_parity = lookup_config_for_pools_without_env(&zero_parity_kvs, &[4]).expect("zero-parity config should resolve");
        assert_eq!(
            resolve_write_layout(&zero_parity, 0, 4, 2, None, true).expect("max parity should override configured parity"),
            WriteLayout {
                data_drives: 2,
                parity_drives: 2,
                write_quorum: 3,
            }
        );
    }

    #[test]
    fn only_uninitialized_config_falls_back_to_pool_startup_parity() {
        let uninitialized = crate::config::storageclass::Config::default();
        assert_eq!(
            resolve_write_layout(&uninitialized, 99, 2, 0, None, false)
                .expect("uninitialized config should preserve the pool's startup fallback"),
            WriteLayout {
                data_drives: 2,
                parity_drives: 0,
                write_quorum: 2,
            }
        );

        let initialized = lookup_config_for_pools_without_env(&KVS::new(), &[4, 2]).expect("initialized config should resolve");
        assert!(resolve_write_layout(&initialized, 99, 2, 1, None, false).is_err());
        assert!(resolve_write_layout(&initialized, 1, 4, 2, None, false).is_err());
    }

    #[test]
    #[serial_test::serial(storage_class_env)]
    fn held_snapshot_keeps_parity_and_inline_policy_consistent_across_reload() {
        let old = temp_env::with_vars(
            [
                (STANDARD_ENV, Some("")),
                (RRS_ENV, Some("")),
                (OPTIMIZE_ENV, None),
                (INLINE_BLOCK_ENV, Some("1KiB")),
            ],
            || lookup_config_for_pools(&KVS::new(), &[4, 2]),
        )
        .expect("old config should resolve");
        let new = temp_env::with_vars(
            [
                (STANDARD_ENV, Some("EC:1")),
                (RRS_ENV, Some("EC:1")),
                (OPTIMIZE_ENV, None),
                (INLINE_BLOCK_ENV, Some("0B")),
            ],
            || lookup_config_for_pools(&KVS::new(), &[4, 2]),
        )
        .expect("new config should resolve");

        let published = ArcSwap::from_pointee(old);
        let held = published.load_full();
        published.store(Arc::new(new));

        let held_layout = resolve_write_layout(&held, 0, 4, 2, None, false).expect("held snapshot should remain valid");
        assert_eq!(held_layout.parity_drives, 2);
        assert!(held.should_inline(512, held_layout.data_drives, false));

        let current = published.load_full();
        let current_layout = resolve_write_layout(&current, 0, 4, 2, None, false).expect("new snapshot should resolve");
        assert_eq!(current_layout.parity_drives, 1);
        assert!(!current.should_inline(512, current_layout.data_drives, false));
    }
}

fn build_tiered_decommission_file_info(bucket: &str, object: &str, fi: &FileInfo, layout: WriteLayout) -> FileInfo {
    let WriteLayout {
        data_drives,
        parity_drives,
        ..
    } = layout;

    let mut updated = fi.clone();
    updated.erasure = FileInfo::new([bucket, object].join("/").as_str(), data_drives, parity_drives).erasure;

    updated
}

fn resolve_tiered_decommission_write_quorum_result(
    errs: &[Option<DiskError>],
    write_quorum: usize,
    bucket: &str,
    object: &str,
) -> Result<()> {
    if let Some(err) = reduce_write_quorum_errs(errs, OBJECT_OP_IGNORED_ERRS, write_quorum) {
        return Err(to_object_err(err.into(), vec![bucket, object]));
    }

    Ok(())
}

#[derive(Clone, Debug)]
pub struct SetDisks {
    pub locker_owner: String,
    pub disks: Arc<RwLock<Vec<Option<DiskStore>>>>,
    pub set_endpoints: Vec<Endpoint>,
    pub set_drive_count: usize,
    pub default_parity_count: usize,
    pub set_index: usize,
    pub pool_index: usize,
    /// Stable namespace shared by every object lock created for this pool.
    set_lock_namespace: Arc<str>,
    pub format: FormatV3,
    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    disk_health_cache: Arc<RwLock<Vec<Option<DiskHealthEntry>>>>,
    /// Coalesce bounded GET probes of missing remote slots. Shared by set clones;
    /// the timestamp also limits repeated probes while a peer remains absent.
    read_reconnect: Arc<tokio::sync::Mutex<Option<tokio::time::Instant>>>,
    get_object_metadata_cache: moka::future::Cache<GetObjectMetadataCacheKey, Arc<GetObjectMetadataCacheEntry>>,
    get_object_metadata_cache_hash_builder: std::collections::hash_map::RandomState,
    get_object_metadata_cache_generations: Arc<[AtomicU64]>,
    /// GET codecs keyed by every persisted layout dimension that affects
    /// decoding. Clones of a set share the memoized shells.
    erasure_cache: Arc<ErasureCache>,
    pub lockers: Vec<Arc<dyn LockClient>>,
    shared_lockers: Arc<[Arc<dyn LockClient>]>,
    local_lock_manager: Arc<rustfs_lock::GlobalLockManager>,
    /// Per-instance runtime context (Phase 5, backlog#939).
    ///
    /// The leaf of the object-graph ctx plumbing (ECStore → Sets → SetDisks).
    /// Slice 3 sources `local_lock_manager` from this context to give each
    /// instance its own lock namespace.
    ctx: Arc<InstanceContext>,
    /// Memoized capacity dirty scope so successful writes reuse a prebuilt
    /// `Arc` instead of allocating a `String` per online disk (backlog#1315).
    /// `Arc` so clones of a set share one memo.
    capacity_scope_cache: Arc<std::sync::RwLock<CapacityScopeCache>>,
    /// Last global dirty-registry generation this set marked. `u64::MAX` until
    /// the first write; equality with the current generation lets steady-state
    /// writes skip the global registry mutex (backlog#1315). `Arc` so clones of
    /// a set share one generation marker.
    capacity_dirty_generation: Arc<AtomicU64>,
    /// Orphan prefixes whose last purge scan met data that can never be
    /// purged by listing (residue without a committed marker, an in-flight
    /// write), keyed by `bucket/prefix` with the time of that scan. Empty
    /// listings of such a prefix are frequent and the scan reads the whole
    /// subtree, so it is not repeated within `ORPHAN_PURGE_BACKOFF` (#6898).
    orphan_purge_backoff: Arc<std::sync::Mutex<HashMap<String, std::time::Instant>>>,
    #[cfg(test)]
    storage_class_config_override: Arc<std::sync::RwLock<Option<Arc<storageclass::Config>>>>,
    #[cfg(test)]
    rename_tail_heal_capture: Arc<
        std::sync::Mutex<Option<tokio::sync::mpsc::UnboundedSender<rustfs_heal_contracts::heal_channel::HealChannelRequest>>>,
    >,
}

fn marker_purge_receipt_path(
    bucket: &str,
    object: &str,
    version: Uuid,
    purge: &rustfs_common::mrf_channel::MrfDeleteMarkerPurge,
) -> String {
    let object_hash = rustfs_utils::crypto::hex(Sha256::digest(object.as_bytes()));
    let marker_hash = rustfs_utils::crypto::hex(purge.marker_identity);
    format!(
        "marker-purge-receipts/{}/{}/{}-{}-{}.json",
        bucket, purge.bucket_incarnation_id, version, object_hash, marker_hash
    )
}

impl SetDisks {
    /// Persist a quorum DELETE receipt outside the object marker itself. The
    /// receipt is written to the metadata bucket so it survives a node restart
    /// while an inaccessible member still holds the stale marker.
    pub(crate) async fn persist_marker_purge_receipt(
        &self,
        bucket: &str,
        object: &str,
        version: Uuid,
        purge: &rustfs_common::mrf_channel::MrfDeleteMarkerPurge,
    ) -> bool {
        let path = marker_purge_receipt_path(bucket, object, version, purge);
        let api = Arc::new(self.clone());
        crate::config::com::save_config_with_opts(
            api,
            &path,
            b"rustfs-marker-purge-receipt-v1".to_vec(),
            &ObjectOptions {
                max_parity: true,
                no_lock: true,
                ..Default::default()
            },
        )
        .await
        .is_ok()
    }

    pub(crate) async fn has_marker_purge_receipt(
        &self,
        bucket: &str,
        object: &str,
        version: Uuid,
        purge: &rustfs_common::mrf_channel::MrfDeleteMarkerPurge,
    ) -> bool {
        let path = marker_purge_receipt_path(bucket, object, version, purge);
        crate::config::com::read_config_limited_preserve_empty(Arc::new(self.clone()), &path, 128)
            .await
            .is_ok()
    }

    pub(crate) async fn consume_marker_purge_receipt(
        &self,
        bucket: &str,
        object: &str,
        version: Uuid,
        purge: &rustfs_common::mrf_channel::MrfDeleteMarkerPurge,
    ) {
        let path = marker_purge_receipt_path(bucket, object, version, purge);
        let _ = crate::config::com::delete_config_no_lock(Arc::new(self.clone()), &path).await;
    }
}

/// Read every physical copy before selecting a version quorum. A minority
/// legacy record is still evidence and must not disappear behind a majority
/// not-found result. Only an explicit file/volume absence produces `None`;
/// an unreadable disk cannot prove that no conflicting copy exists.
pub(crate) async fn read_legacy_transition_state_metadata_copies(
    set: &SetDisks,
    bucket: &str,
    object: &str,
) -> std::result::Result<Vec<Option<Vec<u8>>>, DiskError> {
    let disk_object = rustfs_utils::path::encode_dir_object(object);
    let disks = set.get_disks_internal().await;
    if disks.is_empty() {
        return Err(DiskError::DiskNotFound);
    }

    // Include inline bytes in the generation: a conditional repair preserves
    // the entire xl.meta, including payloads belonging to other versions.
    let (copies, errs) = SetDisks::read_all_raw_file_info(&disks, bucket, disk_object.as_str(), true).await;
    for err in errs.into_iter().flatten() {
        if !matches!(err, DiskError::FileNotFound | DiskError::FileVersionNotFound | DiskError::VolumeNotFound) {
            return Err(err);
        }
    }

    Ok(copies.into_iter().map(|copy| copy.map(|copy| copy.buf)).collect())
}

// DistributedLock sends the raw ObjectKey to its clients; LockRegistry clones
// each endpoint's canonical Arc, so an exact Arc set identifies the lock domain.
pub(crate) fn same_distributed_lock_domain(left: &[Arc<dyn LockClient>], right: &[Arc<dyn LockClient>]) -> bool {
    left.iter()
        .all(|left_client| right.iter().any(|right_client| Arc::ptr_eq(left_client, right_client)))
        && right
            .iter()
            .all(|right_client| left.iter().any(|left_client| Arc::ptr_eq(left_client, right_client)))
}

const ERASURE_CACHE_MAX_ENTRIES: usize = 32;

#[derive(Clone, Copy, Debug, Eq, Hash, PartialEq)]
struct ErasureCacheKey {
    data_shards: usize,
    parity_shards: usize,
    block_size: usize,
    uses_legacy: bool,
}

struct ErasureCache {
    entries: parking_lot::RwLock<HashMap<ErasureCacheKey, Arc<coding::Erasure>>>,
}

impl std::fmt::Debug for ErasureCache {
    fn fmt(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        formatter
            .debug_struct("ErasureCache")
            .field("entries", &self.entries.read().len())
            .finish()
    }
}

impl ErasureCache {
    fn new() -> Self {
        Self {
            entries: parking_lot::RwLock::new(HashMap::new()),
        }
    }

    fn get_or_try_insert(
        &self,
        key: ErasureCacheKey,
    ) -> std::result::Result<Arc<coding::Erasure>, coding::ErasureConstructionError> {
        if let Some(erasure) = self.entries.read().get(&key) {
            return Ok(Arc::clone(erasure));
        }

        // Serialize first construction for a key so concurrent cold GETs still
        // create exactly one shell. Codec construction never awaits.
        let mut entries = self.entries.write();
        if let Some(erasure) = entries.get(&key) {
            return Ok(Arc::clone(erasure));
        }
        let erasure = Arc::new(coding::Erasure::try_new_with_options(
            key.data_shards,
            key.parity_shards,
            key.block_size,
            key.uses_legacy,
        )?);
        if entries.len() < ERASURE_CACHE_MAX_ENTRIES {
            entries.insert(key, Arc::clone(&erasure));
        }
        Ok(erasure)
    }

    fn get_for_file_info(&self, fi: &FileInfo) -> Result<Arc<coding::Erasure>> {
        self.get_or_try_insert(ErasureCacheKey {
            data_shards: fi.erasure.data_blocks,
            parity_shards: fi.erasure.parity_blocks,
            block_size: fi.erasure.block_size,
            uses_legacy: fi.uses_legacy_checksum,
        })
        .map_err(Error::from)
    }
}

#[cfg(test)]
mod erasure_cache_tests {
    use super::*;

    #[test]
    fn reuses_shells_and_keeps_every_layout_dimension_in_the_key() {
        let cache = ErasureCache::new();
        let base = ErasureCacheKey {
            data_shards: 4,
            parity_shards: 2,
            block_size: 1_048_576,
            uses_legacy: false,
        };
        let first = cache.get_or_try_insert(base).expect("modern shell should construct");
        let reused = cache.get_or_try_insert(base).expect("same modern shell should be cached");
        assert!(Arc::ptr_eq(&first, &reused));

        for distinct in [
            ErasureCacheKey { data_shards: 3, ..base },
            ErasureCacheKey {
                parity_shards: 1,
                ..base
            },
            ErasureCacheKey {
                block_size: 524_288,
                ..base
            },
            ErasureCacheKey {
                uses_legacy: true,
                ..base
            },
        ] {
            let shell = cache.get_or_try_insert(distinct).expect("distinct shell should construct");
            assert!(!Arc::ptr_eq(&first, &shell));
        }
        assert_eq!(cache.entries.read().len(), 5);
    }

    #[test]
    fn does_not_cache_invalid_layouts_or_grow_past_the_bound() {
        let cache = ErasureCache::new();
        let invalid = ErasureCacheKey {
            data_shards: 4,
            parity_shards: 2,
            block_size: 0,
            uses_legacy: false,
        };
        assert!(cache.get_or_try_insert(invalid).is_err());
        assert!(cache.entries.read().is_empty());

        for block_size in 1..=(ERASURE_CACHE_MAX_ENTRIES + 1) {
            cache
                .get_or_try_insert(ErasureCacheKey {
                    data_shards: 4,
                    parity_shards: 2,
                    block_size,
                    uses_legacy: false,
                })
                .expect("bounded cache fixture should construct");
        }
        assert_eq!(cache.entries.read().len(), ERASURE_CACHE_MAX_ENTRIES);
    }
}

#[derive(Clone, Debug, Eq, PartialEq)]
struct GetObjectMetadataCacheKey {
    bucket: Arc<str>,
    object: Arc<str>,
    generation: u64,
    hash: u64,
}

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
struct GetObjectMetadataCacheGeneration {
    index: usize,
    value: u64,
    hash: u64,
}

#[cfg(test)]
struct MetadataCacheInvalidationProbeState {
    bucket: String,
    object: String,
    count: AtomicU64,
}

#[cfg(test)]
struct MetadataCacheInvalidationProbe {
    state: Arc<MetadataCacheInvalidationProbeState>,
}

#[cfg(test)]
static METADATA_CACHE_INVALIDATION_PROBE: std::sync::OnceLock<
    std::sync::Mutex<Option<Arc<MetadataCacheInvalidationProbeState>>>,
> = std::sync::OnceLock::new();

#[cfg(test)]
impl MetadataCacheInvalidationProbe {
    fn install(bucket: &str, object: &str) -> Self {
        let state = Arc::new(MetadataCacheInvalidationProbeState {
            bucket: bucket.to_string(),
            object: object.to_string(),
            count: AtomicU64::new(0),
        });
        let mut slot = METADATA_CACHE_INVALIDATION_PROBE
            .get_or_init(|| std::sync::Mutex::new(None))
            .lock()
            .expect("metadata cache invalidation probe mutex should not poison");
        assert!(
            slot.is_none(),
            "metadata cache invalidation probe must be installed by one test at a time"
        );
        *slot = Some(Arc::clone(&state));
        drop(slot);
        Self { state }
    }

    fn count(&self) -> u64 {
        self.state.count.load(Ordering::Acquire)
    }
}

#[cfg(test)]
impl Drop for MetadataCacheInvalidationProbe {
    fn drop(&mut self) {
        let mut slot = METADATA_CACHE_INVALIDATION_PROBE
            .get_or_init(|| std::sync::Mutex::new(None))
            .lock()
            .expect("metadata cache invalidation probe mutex should not poison");
        if slot.as_ref().is_some_and(|state| Arc::ptr_eq(state, &self.state)) {
            *slot = None;
        }
    }
}

#[cfg(test)]
fn record_metadata_cache_invalidation(bucket: &str, object: &str) {
    let probe = METADATA_CACHE_INVALIDATION_PROBE
        .get_or_init(|| std::sync::Mutex::new(None))
        .lock()
        .expect("metadata cache invalidation probe mutex should not poison")
        .as_ref()
        .filter(|probe| probe.bucket == bucket && probe.object == object)
        .cloned();
    if let Some(probe) = probe {
        probe.count.fetch_add(1, Ordering::AcqRel);
    }
}

impl GetObjectMetadataCacheKey {
    fn new(bucket: &str, object: &str, generation: GetObjectMetadataCacheGeneration) -> Self {
        Self {
            bucket: Arc::from(bucket),
            object: Arc::from(object),
            generation: generation.value,
            hash: generation.hash,
        }
    }
}

impl Hash for GetObjectMetadataCacheKey {
    fn hash<H: Hasher>(&self, state: &mut H) {
        self.hash.hash(state);
        self.generation.hash(state);
    }
}

#[derive(Debug)]
struct GetObjectMetadataCacheEntry {
    #[allow(dead_code)] // Kept for debugging; moka handles TTL internally
    created_at: Instant,
    fi: FileInfo,
    parts_metadata: Vec<FileInfo>,
    online_disks: Vec<Option<DiskStore>>,
    read_quorum: usize,
}

#[derive(Clone, Debug)]
struct DiskHealthEntry {
    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    last_check: Instant,
    online: bool,
}

impl DiskHealthEntry {
    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    fn cached_value(&self) -> Option<bool> {
        if self.last_check.elapsed() <= DISK_HEALTH_CACHE_TTL {
            Some(self.online)
        } else {
            None
        }
    }
}

impl SetDisks {
    pub(in crate::set_disk) async fn persist_delete_marker_purge(
        &self,
        bucket: &str,
        object: &str,
        version: Uuid,
        purge: rustfs_common::mrf_channel::MrfDeleteMarkerPurge,
    ) -> bool {
        use rustfs_common::mrf_channel::{MrfDurableAdmissionError, MrfScope, persist_delete_marker_purge_intent};
        let scope = (|| {
            Ok::<_, MrfDurableAdmissionError>(MrfScope {
                pool_index: u32::try_from(self.pool_index).map_err(|_| MrfDurableAdmissionError::InvalidIdentity)?,
                set_index: u32::try_from(self.set_index).map_err(|_| MrfDurableAdmissionError::InvalidIdentity)?,
            })
        })();
        let result = match scope {
            Ok(scope) => persist_delete_marker_purge_intent(bucket, object, version, scope, purge).await,
            Err(error) => Err(error),
        };
        match result {
            Ok(()) => {
                tracing::trace!(
                    event = EVENT_SET_DISK_HEAL,
                    component = LOG_COMPONENT_ECSTORE,
                    subsystem = LOG_SUBSYSTEM_SET_DISK,
                    state = "delete_marker_purge_durably_admitted",
                    bucket,
                    object,
                    version_id = %version,
                    pool_index = self.pool_index,
                    set_index = self.set_index,
                    "Delete-marker purge responsibility persisted"
                );
                true
            }
            Err(error) => {
                warn!(
                    event = EVENT_SET_DISK_HEAL,
                    component = LOG_COMPONENT_ECSTORE,
                    subsystem = LOG_SUBSYSTEM_SET_DISK,
                    state = "delete_marker_purge_admission_failed",
                    bucket,
                    object,
                    version_id = %version,
                    pool_index = self.pool_index,
                    set_index = self.set_index,
                    error = %error,
                    "Delete-marker purge responsibility could not be persisted"
                );
                false
            }
        }
    }

    pub(in crate::set_disk) async fn persist_partial_write(
        &self,
        bucket: &str,
        object: &str,
        version_id: Option<&str>,
        source_bucket_incarnation_id: Option<Uuid>,
    ) -> bool {
        use rustfs_common::mrf_channel::{
            MrfDurableAdmissionError, MrfScope, mrf_delivery_enabled, persist_partial_write_intent_with_incarnation,
        };

        if !mrf_delivery_enabled() {
            return false;
        }
        let identity = (|| {
            let version = version_id
                .filter(|value| !value.is_empty())
                .map(Uuid::parse_str)
                .transpose()
                .map_err(|_| MrfDurableAdmissionError::InvalidIdentity)?;
            let scope = MrfScope {
                pool_index: u32::try_from(self.pool_index).map_err(|_| MrfDurableAdmissionError::InvalidIdentity)?,
                set_index: u32::try_from(self.set_index).map_err(|_| MrfDurableAdmissionError::InvalidIdentity)?,
            };
            Ok::<_, MrfDurableAdmissionError>((version, scope))
        })();
        let result = match identity {
            Ok((version, scope)) => {
                persist_partial_write_intent_with_incarnation(bucket, object, version, scope, source_bucket_incarnation_id).await
            }
            Err(err) => Err(err),
        };
        match result {
            Ok(()) => {
                tracing::trace!(
                    event = EVENT_SET_DISK_HEAL,
                    component = LOG_COMPONENT_ECSTORE,
                    subsystem = LOG_SUBSYSTEM_SET_DISK,
                    state = "mrf_durably_admitted",
                    bucket,
                    object,
                    version_id,
                    pool_index = self.pool_index,
                    set_index = self.set_index,
                    "Partial write repair responsibility persisted"
                );
                true
            }
            Err(error) => {
                warn!(
                    event = EVENT_SET_DISK_HEAL,
                    component = LOG_COMPONENT_ECSTORE,
                    subsystem = LOG_SUBSYSTEM_SET_DISK,
                    state = "mrf_durable_admission_failed",
                    bucket, object, version_id, pool_index = self.pool_index, set_index = self.set_index,
                    error = %error,
                    "Partial write repair falling back to the heal channel"
                );
                false
            }
        }
    }

    pub(in crate::set_disk) async fn submit_rename_tail_heal(
        &self,
        mut request: rustfs_heal_contracts::heal_channel::HealChannelRequest,
    ) {
        if let Some(object) = request.object_prefix.as_deref()
            && self
                .persist_partial_write(
                    &request.bucket,
                    object,
                    request.object_version_id.as_deref(),
                    request.expected_bucket_incarnation_id,
                )
                .await
        {
            return;
        }
        if request.expected_bucket_incarnation_id.is_none() {
            return;
        }
        request.source = rustfs_heal_contracts::heal_channel::HealRequestSource::Mrf;
        #[cfg(test)]
        {
            let capture = self
                .rename_tail_heal_capture
                .lock()
                .expect("rename tail heal capture mutex should not poison")
                .clone();
            if let Some(capture) = capture {
                let _ = capture.send(request);
                return;
            }
        }

        let _ = rustfs_heal_contracts::heal_channel::send_heal_request(request).await;
    }

    #[cfg(test)]
    pub(in crate::set_disk) fn capture_test_rename_tail_heals(
        &self,
    ) -> tokio::sync::mpsc::UnboundedReceiver<rustfs_heal_contracts::heal_channel::HealChannelRequest> {
        let (capture, requests) = tokio::sync::mpsc::unbounded_channel();
        let mut slot = self
            .rename_tail_heal_capture
            .lock()
            .expect("rename tail heal capture mutex should not poison");
        assert!(slot.is_none(), "only one rename tail heal capture may be installed per set");
        *slot = Some(capture);
        requests
    }

    fn storage_class_config_snapshot(&self) -> Arc<storageclass::Config> {
        #[cfg(test)]
        if let Some(config) = self
            .storage_class_config_override
            .read()
            .expect("test storage class override lock should not be poisoned")
            .as_ref()
        {
            return config.clone();
        }

        runtime_sources::storage_class_config_snapshot()
    }

    #[cfg(test)]
    pub(crate) fn set_test_storage_class_config(&self, config: storageclass::Config) {
        *self
            .storage_class_config_override
            .write()
            .expect("test storage class override lock should not be poisoned") = Some(Arc::new(config));
    }

    fn get_object_metadata_cache_hash(&self, bucket: &str, object: &str) -> u64 {
        let mut hasher = self.get_object_metadata_cache_hash_builder.build_hasher();
        bucket.hash(&mut hasher);
        object.hash(&mut hasher);
        hasher.finish()
    }

    fn get_object_metadata_cache_generation(&self, bucket: &str, object: &str) -> Option<GetObjectMetadataCacheGeneration> {
        let hash = self.get_object_metadata_cache_hash(bucket, object);
        let hash_bytes = hash.to_le_bytes();
        let index = usize::from(u16::from_le_bytes([hash_bytes[0], hash_bytes[1]]) % GET_OBJECT_METADATA_CACHE_FENCE_SHARDS);
        let value = self.get_object_metadata_cache_generations[index].load(Ordering::Acquire);
        (value != u64::MAX).then_some(GetObjectMetadataCacheGeneration { index, value, hash })
    }

    fn is_get_object_metadata_cache_generation_current(&self, generation: GetObjectMetadataCacheGeneration) -> bool {
        self.get_object_metadata_cache_generations[generation.index].load(Ordering::Acquire) == generation.value
    }

    pub(crate) async fn invalidate_get_object_metadata_cache(&self, bucket: &str, object: &str) {
        let hash = self.get_object_metadata_cache_hash(bucket, object);
        let hash_bytes = hash.to_le_bytes();
        let index = usize::from(u16::from_le_bytes([hash_bytes[0], hash_bytes[1]]) % GET_OBJECT_METADATA_CACHE_FENCE_SHARDS);
        let generation = &self.get_object_metadata_cache_generations[index];
        let previous = match generation.try_update(Ordering::AcqRel, Ordering::Acquire, |current| current.checked_add(1)) {
            Ok(previous) | Err(previous) => previous,
        };
        let previous = GetObjectMetadataCacheGeneration {
            index,
            value: previous,
            hash,
        };
        self.get_object_metadata_cache
            .invalidate(&GetObjectMetadataCacheKey::new(bucket, object, previous))
            .await;
        #[cfg(test)]
        record_metadata_cache_invalidation(bucket, object);
    }

    fn invalidate_all_get_object_metadata_cache(&self) {
        for generation in self.get_object_metadata_cache_generations.iter() {
            let _ = generation.try_update(Ordering::AcqRel, Ordering::Acquire, |current| current.checked_add(1));
        }
        self.get_object_metadata_cache.invalidate_all();
    }

    #[inline(always)]
    fn record_put_object_commit_namespace_lock_wait(op: &'static str, acquire_start: Instant) {
        if op != "put_object_commit" || !rustfs_io_metrics::put_stage_metrics_enabled() {
            return;
        }
        rustfs_io_metrics::record_put_object_stage_duration_from(
            rustfs_io_metrics::PUT_STAGE_PUT_OBJECT_COMMIT_NAMESPACE_LOCK_WAIT,
            Some(acquire_start),
        );
    }

    async fn acquire_read_lock_diag(&self, op: &'static str, bucket: &str, object: &str) -> Result<ObjectLockDiagGuard> {
        crate::hp_guard!("SetDisks::acquire_read_lock");
        let diag_enabled = is_object_lock_diag_enabled();
        let ns_lock = self.new_ns_lock(bucket, object).await?;
        let acquire_start = Instant::now();
        let timeout = get_lock_acquire_timeout();
        let mut attempt = ObjectLockAttempt::start(op, bucket, object, None, &ns_lock, "read", timeout);
        let result = ns_lock.get_read_lock(timeout).await;
        attempt.observe(&result);
        let guard = result.map_err(|e| self.map_namespace_lock_error(bucket, object, "read", e))?;
        let owner = diag_enabled.then(|| ns_lock.owner().to_string());
        self.log_object_lock_acquire_if_slow(op, bucket, object, "read", owner.as_deref(), acquire_start.elapsed(), diag_enabled);
        Ok(ObjectLockDiagGuard::new(
            guard,
            diag_enabled,
            op,
            diag_enabled.then(|| bucket.to_string()),
            diag_enabled.then(|| object.to_string()),
            owner,
            "read",
        )
        .with_attempt(attempt))
    }

    pub(crate) async fn read_listing_metadata_after_namespace_barrier(
        &self,
        disks: &[Option<DiskStore>],
        bucket: &str,
        object: &str,
    ) -> Result<(MetaCacheEntries, usize)> {
        // This is only called after the lock-free LIST fast path observes an
        // unresolved metadata generation. Waiting on the ordinary object read
        // lock establishes a publication boundary with concurrent overwrites.
        let stored_object = encode_dir_object_ref(object);
        let _guard = self
            .acquire_read_lock_diag("list_object_reconcile", bucket, stored_object.as_ref())
            .await?;
        let (raw_entries, errors) = Self::read_all_raw_file_info(disks, bucket, stored_object.as_ref(), false).await;
        let confirmed_absent = errors
            .iter()
            .flatten()
            .filter(|err| DiskError::is_err_object_not_found(err) || DiskError::is_err_version_not_found(err))
            .count();
        let entries = raw_entries
            .into_iter()
            .map(|raw| {
                raw.filter(|raw| !raw.buf.is_empty()).map(|raw| MetaCacheEntry {
                    name: object.to_owned(),
                    metadata: raw.buf,
                    cached: None,
                    reusable: false,
                })
            })
            .collect();

        Ok((MetaCacheEntries(entries), confirmed_absent))
    }

    async fn acquire_write_lock_diag(&self, op: &'static str, bucket: &str, object: &str) -> Result<ObjectLockDiagGuard> {
        crate::hp_guard!("SetDisks::acquire_write_lock");
        let diag_enabled = is_object_lock_diag_enabled();
        let ns_lock = self.new_ns_lock(bucket, object).await?;
        let acquire_start = Instant::now();
        let acquire_timeout = get_put_object_commit_lock_acquire_timeout(op);
        let mut attempt = ObjectLockAttempt::start(op, bucket, object, None, &ns_lock, "write", acquire_timeout);
        let result = ns_lock.get_write_lock(acquire_timeout).await;
        attempt.observe(&result);
        let guard = resolve_put_object_commit_lock_acquire_result(self, op, bucket, object, result)?;
        Self::record_put_object_commit_namespace_lock_wait(op, acquire_start);
        let owner = diag_enabled.then(|| ns_lock.owner().to_string());
        self.log_object_lock_acquire_if_slow(
            op,
            bucket,
            object,
            "write",
            owner.as_deref(),
            acquire_start.elapsed(),
            diag_enabled,
        );
        Ok(ObjectLockDiagGuard::new(
            guard,
            diag_enabled,
            op,
            diag_enabled.then(|| bucket.to_string()),
            diag_enabled.then(|| object.to_string()),
            owner,
            "write",
        )
        .with_attempt(attempt))
    }

    #[cfg(any(test, feature = "test-util"))]
    async fn acquire_write_lock_diag_with_pending_hook(
        &self,
        op: &'static str,
        bucket: &str,
        object: &str,
        on_pending: impl FnOnce(),
    ) -> Result<ObjectLockDiagGuard> {
        crate::hp_guard!("SetDisks::acquire_write_lock");
        let diag_enabled = is_object_lock_diag_enabled();
        let ns_lock = self.new_ns_lock(bucket, object).await?;
        let acquire_start = Instant::now();
        let acquire_timeout = get_put_object_commit_lock_acquire_timeout(op);
        let mut attempt = ObjectLockAttempt::start(op, bucket, object, None, &ns_lock, "write", acquire_timeout);
        let acquire = ns_lock.get_write_lock(acquire_timeout);
        tokio::pin!(acquire);
        let mut on_pending = Some(on_pending);
        let result = futures::future::poll_fn(|cx| match std::future::Future::poll(acquire.as_mut(), cx) {
            std::task::Poll::Pending => {
                if let Some(on_pending) = on_pending.take() {
                    on_pending();
                }
                std::task::Poll::Pending
            }
            std::task::Poll::Ready(result) => std::task::Poll::Ready(result),
        })
        .await;
        attempt.observe(&result);
        let guard = resolve_put_object_commit_lock_acquire_result(self, op, bucket, object, result)?;
        Self::record_put_object_commit_namespace_lock_wait(op, acquire_start);
        let owner = diag_enabled.then(|| ns_lock.owner().to_string());
        self.log_object_lock_acquire_if_slow(
            op,
            bucket,
            object,
            "write",
            owner.as_deref(),
            acquire_start.elapsed(),
            diag_enabled,
        );
        Ok(ObjectLockDiagGuard::new(
            guard,
            diag_enabled,
            op,
            diag_enabled.then(|| bucket.to_string()),
            diag_enabled.then(|| object.to_string()),
            owner,
            "write",
        )
        .with_attempt(attempt))
    }

    #[allow(clippy::too_many_arguments)]
    fn log_object_lock_acquire_if_slow(
        &self,
        op: &'static str,
        bucket: &str,
        object: &str,
        mode: &'static str,
        owner: Option<&str>,
        elapsed: Duration,
        diag_enabled: bool,
    ) {
        if !diag_enabled {
            return;
        }

        let threshold = get_object_lock_diag_slow_acquire_threshold();
        record_object_lock_diag_acquire_duration(op, mode, elapsed);
        if elapsed >= threshold {
            record_object_lock_diag_slow_acquire(op, mode);
            warn!(
                target: "rustfs_ecstore::object_lock_diag",
                op,
                bucket,
                object,
                mode,
                owner = owner.unwrap_or_default(),
                acquire_ms = elapsed.as_millis(),
                threshold_ms = threshold.as_millis(),
                "object namespace lock acquisition exceeded threshold"
            );
        }
    }

    #[allow(clippy::too_many_arguments)]
    pub async fn new(
        locker_owner: String,
        disks: Arc<RwLock<Vec<Option<DiskStore>>>>,
        set_drive_count: usize,
        default_parity_count: usize,
        set_index: usize,
        pool_index: usize,
        set_endpoints: Vec<Endpoint>,
        format: FormatV3,
        lockers: Vec<Arc<dyn LockClient>>,
    ) -> Arc<Self> {
        Self::new_with_instance_ctx(
            locker_owner,
            disks,
            set_drive_count,
            default_parity_count,
            set_index,
            pool_index,
            set_endpoints,
            format,
            lockers,
            bootstrap_ctx(),
        )
        .await
    }

    /// Build a set bound to an explicit instance context (Phase 5 follow-up,
    /// backlog#1052). The legacy [`SetDisks::new`] entry adopts the process
    /// bootstrap context; a store constructed around its own context threads it
    /// down here so the whole object graph shares one cell.
    #[allow(clippy::too_many_arguments)]
    pub async fn new_with_instance_ctx(
        locker_owner: String,
        disks: Arc<RwLock<Vec<Option<DiskStore>>>>,
        set_drive_count: usize,
        default_parity_count: usize,
        set_index: usize,
        pool_index: usize,
        set_endpoints: Vec<Endpoint>,
        format: FormatV3,
        lockers: Vec<Arc<dyn LockClient>>,
        instance_ctx: Arc<InstanceContext>,
    ) -> Arc<Self> {
        let ctx = instance_ctx;
        let set_lock_namespace: Arc<str> = format!("pool-{pool_index}").into();
        let shared_lockers = Arc::from(lockers.to_vec());
        Arc::new(SetDisks {
            locker_owner,
            disks,
            set_drive_count,
            default_parity_count,
            set_index,
            pool_index,
            set_lock_namespace,
            format,
            set_endpoints,
            disk_health_cache: Arc::new(RwLock::new(Vec::new())),
            read_reconnect: Arc::new(tokio::sync::Mutex::new(None)),
            get_object_metadata_cache: moka::future::Cache::builder()
                .max_capacity(get_object_metadata_cache_max_entries() as u64)
                .time_to_live(GET_OBJECT_METADATA_CACHE_TTL)
                .build(),
            get_object_metadata_cache_hash_builder: std::collections::hash_map::RandomState::new(),
            get_object_metadata_cache_generations: Arc::from(
                (0..usize::from(GET_OBJECT_METADATA_CACHE_FENCE_SHARDS))
                    .map(|_| AtomicU64::new(0))
                    .collect::<Vec<_>>(),
            ),
            erasure_cache: Arc::new(ErasureCache::new()),
            lockers,
            shared_lockers,
            // Sourced from the instance context so each instance owns its lock
            // namespace (Phase 5 Slice 3). Single-instance: ctx aliases the
            // process lock-manager singleton, so this is unchanged.
            local_lock_manager: ctx.lock_manager(),
            ctx,
            capacity_scope_cache: Arc::new(std::sync::RwLock::new(CapacityScopeCache::default())),
            capacity_dirty_generation: Arc::new(AtomicU64::new(u64::MAX)),
            orphan_purge_backoff: Arc::new(std::sync::Mutex::new(HashMap::new())),
            #[cfg(test)]
            storage_class_config_override: Arc::new(std::sync::RwLock::new(None)),
            #[cfg(test)]
            rename_tail_heal_capture: Arc::new(std::sync::Mutex::new(None)),
        })
    }

    /// This set's per-instance runtime context (Phase 5, backlog#939).
    #[allow(dead_code)] // Read by tests; consumed by later slices.
    pub(crate) fn instance_ctx(&self) -> &Arc<InstanceContext> {
        &self.ctx
    }

    #[cfg(test)]
    pub(crate) fn set_instance_ctx_for_test(&mut self, ctx: Arc<InstanceContext>) {
        self.local_lock_manager = ctx.lock_manager();
        self.ctx = ctx;
    }

    /// Read the persisted bucket identity through this set's metadata owner.
    /// Missing or non-authoritative legacy identities remain errors.
    pub async fn bucket_incarnation_id_from_disk(&self, bucket: &str) -> Result<Uuid> {
        if crate::bucket::utils::is_meta_bucketname(bucket) {
            // Metadata writes can already hold the pool metadata write lock.
            return Err(Error::other("system metadata bucket has no bucket incarnation"));
        }
        metadata_sys::get_bucket_incarnation_id_in(&self.ctx, bucket).await
    }

    /// Admit one short scanner cache publication under this set's instance
    /// movement fence. The caller must hold the returned guard through its
    /// final conditional cache write; no scan-round work belongs under it.
    pub async fn scanner_data_usage_publication_admission_guard(&self) -> Option<(tokio::sync::OwnedRwLockReadGuard<()>, u64)> {
        let operation_gate = self.ctx.data_movement_operation_gate();
        let operation_guard = operation_gate.read_owned().await;
        if self.ctx.scanner_publication_state_allowed() {
            let epoch = self.ctx.data_movement_operation_epoch();
            return Some((operation_guard, epoch));
        }

        // The owner deliberately marks the cached state UNKNOWN after every
        // movement epoch advance. Do not strand remote scanner writers in that
        // state: release this guard before asking the storage owner to refresh
        // its durable movement snapshot, since the owner uses the same gate.
        drop(operation_guard);
        let owner = runtime_sources::object_store_handle().filter(|owner| Arc::ptr_eq(&owner.ctx, &self.ctx))?;
        owner.scanner_data_usage_publication_admission_guard().await
    }

    pub async fn scanner_data_usage_publication_commit_scope(
        &self,
        expected_movement_epoch: u64,
        safe_deadline: tokio::time::Instant,
        remote_lease_tokens: Vec<Uuid>,
    ) -> Option<crate::object_api::ScannerPublicationCommitScope> {
        let (movement_permit, epoch) = self.scanner_data_usage_publication_admission_guard().await?;
        if epoch != expected_movement_epoch {
            return None;
        }
        Some(crate::object_api::ScannerPublicationCommitScope::new_storage_owned(
            epoch,
            safe_deadline,
            remote_lease_tokens,
            movement_permit,
        ))
    }

    pub async fn scanner_data_usage_publication_commit_scope_with_release_flag(
        &self,
        expected_movement_epoch: u64,
        safe_deadline: tokio::time::Instant,
        remote_lease_tokens: Vec<Uuid>,
        lease_release_safe: Arc<std::sync::atomic::AtomicBool>,
    ) -> Option<crate::object_api::ScannerPublicationCommitScope> {
        let (movement_permit, epoch) = self.scanner_data_usage_publication_admission_guard().await?;
        if epoch != expected_movement_epoch {
            return None;
        }
        Some(crate::object_api::ScannerPublicationCommitScope::new_storage_owned_with_release_flag(
            epoch,
            safe_deadline,
            remote_lease_tokens,
            movement_permit,
            lease_release_safe,
        ))
    }

    /// Whether both sets' namespace-lock implementations cover the same object key.
    pub(crate) async fn shares_namespace_lock_domain(&self, other: &Self) -> bool {
        match (self.ctx.is_dist_erasure().await, other.ctx.is_dist_erasure().await) {
            (false, false) => Arc::ptr_eq(&self.local_lock_manager, &other.local_lock_manager),
            (true, true) => {
                self.set_lock_namespace == other.set_lock_namespace && same_distributed_lock_domain(&self.lockers, &other.lockers)
            }
            _ => false,
        }
    }

    /// The lock manager this set actually uses (test-only; Phase 5 Slice 3).
    #[cfg(test)]
    pub(crate) fn local_lock_manager_for_test(&self) -> &Arc<rustfs_lock::GlobalLockManager> {
        &self.local_lock_manager
    }

    // async fn cached_disk_health(&self, index: usize) -> Option<bool> {
    //     let cache = self.disk_health_cache.read().await;
    //     cache
    //         .get(index)
    //         .and_then(|entry| entry.as_ref().and_then(|state| state.cached_value()))
    // }

    // async fn update_disk_health(&self, index: usize, online: bool) {
    //     let mut cache = self.disk_health_cache.write().await;
    //     if cache.len() <= index {
    //         cache.resize(index + 1, None);
    //     }
    //     cache[index] = Some(DiskHealthEntry {
    //         last_check: Instant::now(),
    //         online,
    //     });
    // }

    // async fn is_disk_online_cached(&self, index: usize, disk: &DiskStore) -> bool {
    //     if let Some(online) = self.cached_disk_health(index).await {
    //         return online;
    //     }

    //     let disk_clone = disk.clone();
    //     let online = timeout(DISK_ONLINE_TIMEOUT, async move { disk_clone.is_online().await })
    //         .await
    //         .unwrap_or(false);
    //     self.update_disk_health(index, online).await;
    //     online
    // }

    // async fn filter_online_disks(&self, disks: Vec<Option<DiskStore>>) -> (Vec<Option<DiskStore>>, usize) {
    //     let mut filtered = Vec::with_capacity(disks.len());
    //     let mut online_count = 0;

    //     for (idx, disk) in disks.into_iter().enumerate() {
    //         if let Some(disk_store) = disk {
    //             if self.is_disk_online_cached(idx, &disk_store).await {
    //                 filtered.push(Some(disk_store));
    //                 online_count += 1;
    //             } else {
    //                 filtered.push(None);
    //             }
    //         } else {
    //             filtered.push(None);
    //         }
    //     }

    //     (filtered, online_count)
    // }

    // async fn write_all(disks: &[Option<DiskStore>], bucket: &str, object: &str, buff: Vec<u8>) -> Vec<Option<Error>> {
    //     let mut futures = Vec::with_capacity(disks.len());

    //     let mut errors = Vec::with_capacity(disks.len());

    //     for disk in disks.iter() {
    //         if disk.is_none() {
    //             errors.push(Some(Error::new(DiskError::DiskNotFound)));
    //             continue;
    //         }
    //         let disk = disk.as_ref().unwrap();
    //         futures.push(disk.write_all(bucket, object, buff.clone()));
    //     }

    //     let results = join_all(futures).await;
    //     for result in results {
    //         match result {
    //             Ok(_) => {
    //                 errors.push(None);
    //             }
    //             Err(e) => {
    //                 errors.push(Some(e));
    //             }
    //         }
    //     }
    //     errors
    // }

    // Returns per object readQuorum and writeQuorum
    // readQuorum is the min required disks to read data.
    // writeQuorum is the min required disks to write data.

    // Optimized version using batch processor with quorum support

    // pub async fn walk_dir(&self, opts: &WalkDirOptions) -> (Vec<Option<Vec<MetaCacheEntry>>>, Vec<Option<Error>>) {
    //     let disks = self.disks.read().await;

    //     let disks = disks.clone();
    //     let mut futures = Vec::new();
    //     let mut errs = Vec::new();
    //     let mut ress = Vec::new();

    //     for disk in disks.iter() {
    //         let opts = opts.clone();
    //         futures.push(async move {
    //             if let Some(disk) = disk {
    //                 disk.walk_dir(opts, &mut Writer::NotUse).await
    //             } else {
    //                 Err(DiskError::DiskNotFound)
    //             }
    //         });
    //     }

    //     let results = join_all(futures).await;

    //     for res in results {
    //         match res {
    //             Ok(entries) => {
    //                 ress.push(Some(entries));
    //                 errs.push(None);
    //             }
    //             Err(e) => {
    //                 ress.push(None);
    //                 errs.push(Some(e));
    //             }
    //         }
    //     }

    //     (ress, errs)
    // }

    // async fn remove_object_part(
    //     &self,
    //     bucket: &str,
    //     object: &str,
    //     upload_id: &str,
    //     data_dir: &str,
    //     part_num: usize,
    // ) -> Result<()> {
    //     let upload_id_path = Self::get_upload_id_dir(bucket, object, upload_id);
    //     let disks = self.disks.read().await;

    //     let disks = disks.clone();

    //     let file_path = format!("{}/{}/part.{}", upload_id_path, data_dir, part_num);

    //     let mut futures = Vec::with_capacity(disks.len());
    //     let mut errors = Vec::with_capacity(disks.len());

    //     for disk in disks.iter() {
    //         let file_path = file_path.clone();
    //         let meta_file_path = format!("{}.meta", file_path);

    //         futures.push(async move {
    //             if let Some(disk) = disk {
    //                 disk.delete(RUSTFS_META_MULTIPART_BUCKET, &file_path, DeleteOptions::default())
    //                     .await?;
    //                 disk.delete(RUSTFS_META_MULTIPART_BUCKET, &meta_file_path, DeleteOptions::default())
    //                     .await
    //             } else {
    //                 Err(DiskError::DiskNotFound)
    //             }
    //         });
    //     }

    //     let results = join_all(futures).await;
    //     for result in results {
    //         match result {
    //             Ok(_) => {
    //                 errors.push(None);
    //             }
    //             Err(e) => {
    //                 errors.push(Some(e));
    //             }
    //         }
    //     }

    //     Ok(())
    // }
    // async fn remove_part_meta(&self, bucket: &str, object: &str, upload_id: &str, data_dir: &str, part_num: usize) -> Result<()> {
    //     let upload_id_path = Self::get_upload_id_dir(bucket, object, upload_id);
    //     let disks = self.disks.read().await;

    //     let disks = disks.clone();
    //     // let disks = Self::shuffle_disks(&disks, &fi.erasure.distribution);

    //     let file_path = format!("{}/{}/part.{}.meta", upload_id_path, data_dir, part_num);

    //     let mut futures = Vec::with_capacity(disks.len());
    //     let mut errors = Vec::with_capacity(disks.len());

    //     for disk in disks.iter() {
    //         let file_path = file_path.clone();
    //         futures.push(async move {
    //             if let Some(disk) = disk {
    //                 disk.delete(RUSTFS_META_MULTIPART_BUCKET, &file_path, DeleteOptions::default())
    //                     .await
    //             } else {
    //                 Err(DiskError::DiskNotFound)
    //             }
    //         });
    //     }

    //     let results = join_all(futures).await;
    //     for result in results {
    //         match result {
    //             Ok(_) => {
    //                 errors.push(None);
    //             }
    //             Err(e) => {
    //                 errors.push(Some(e));
    //             }
    //         }
    //     }

    //     Ok(())
    // }
}

fn is_explicit_null_version(version_id: Option<Uuid>) -> bool {
    version_id == Some(Uuid::nil())
}

fn delete_file_info_version_id(version_id: Option<Uuid>) -> Option<Uuid> {
    if is_explicit_null_version(version_id) {
        None
    } else {
        version_id
    }
}

fn object_fits_single_block(object_size: i64, block_size: usize) -> bool {
    match usize::try_from(object_size) {
        Ok(size) => size > 0 && size <= block_size,
        Err(_) => false,
    }
}

/// The common metadata contract consumed by all bounded GET fast paths.
///
/// `ObjectInfo` is assembled from `FileInfo`, but callers may also provide a
/// prepared snapshot or metadata from an older peer.  Never let either copy
/// independently decide that a request is safe: a disagreement must fall
/// back to the regular reader.  The returned size is the only size used for
/// fast-path allocation and reader setup.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub(super) struct ReadPathShape {
    pub(super) object_size: usize,
}

impl ReadPathShape {
    pub(super) fn is_plain(self, object_info: &ObjectInfo, fi: &FileInfo) -> bool {
        !object_info.is_encrypted()
            && !fi.metadata.keys().any(|key| is_object_encryption_marker(key))
            && !object_info.is_compressed()
            && !fi.is_compressed()
            && !object_info.is_remote()
            && !fi.is_remote()
    }
}

/// Request-local metadata decision reused by all small-object GET gates.
///
/// The metadata pair is immutable for the lifetime of `get_object_reader`, so
/// validating it once avoids repeating part-array and transform scans on every
/// fast-path predicate while keeping the trust boundary fail-closed.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
struct ReadPathPlan {
    shape: Option<ReadPathShape>,
    plain: bool,
}

impl ReadPathPlan {
    fn new(object_info: &ObjectInfo, fi: &FileInfo) -> Self {
        let shape = read_path_shape(object_info, fi);
        let plain = shape.is_some_and(|shape| shape.is_plain(object_info, fi));
        Self { shape, plain }
    }

    const fn shape(self) -> Option<ReadPathShape> {
        self.shape
    }

    const fn is_plain(self) -> bool {
        self.plain
    }
}

/// Validate the single-part geometry shared by inline, direct-memory, and
/// bounded mid-size readers.  This is deliberately fail-closed: a stale or
/// mixed-version `ObjectInfo`/`FileInfo` pair must use the legacy path rather
/// than risk allocating or decoding with a mismatched size.
pub(super) fn read_path_shape(object_info: &ObjectInfo, fi: &FileInfo) -> Option<ReadPathShape> {
    if object_info.parts.len() != 1 || fi.parts.len() != 1 || object_info.size < 0 || fi.size < 0 || object_info.size != fi.size {
        return None;
    }

    let object_part = object_info.parts.first()?;
    let file_part = fi.parts.first()?;
    let object_size = usize::try_from(fi.size).ok()?;
    if object_part.number != file_part.number
        || object_part.size != file_part.size
        || object_part.actual_size != file_part.actual_size
        || file_part.size != object_size
        || file_part.actual_size != fi.size
    {
        return None;
    }

    Some(ReadPathShape { object_size })
}

fn should_use_inline_small_fast_path(is_inline_buffer: bool, object_size: i64, block_size: usize) -> bool {
    is_inline_buffer && object_fits_single_block(object_size, block_size)
}

fn should_use_single_block_non_inline_fast_path(is_inline_buffer: bool, object_size: i64, block_size: usize) -> bool {
    !is_inline_buffer && object_fits_single_block(object_size, block_size)
}

#[allow(dead_code, reason = "asserted by this file's gate tests")]
fn should_use_inline_fast_path(
    range: &Option<HTTPRangeSpec>,
    object_info: &ObjectInfo,
    fi: &FileInfo,
    opts: &ObjectOptions,
) -> bool {
    should_use_inline_fast_path_with_plan(range, object_info, fi, opts, ReadPathPlan::new(object_info, fi))
}

fn should_use_inline_fast_path_with_plan(
    range: &Option<HTTPRangeSpec>,
    object_info: &ObjectInfo,
    fi: &FileInfo,
    opts: &ObjectOptions,
    plan: ReadPathPlan,
) -> bool {
    if !object_info.is_inline_fast_path_eligible() || fi.data.is_none() || range.is_some() || opts.part_number.is_some() {
        return false;
    }

    let Some(shape) = plan.shape() else {
        return false;
    };
    // The persisted marker is authoritative for the storage decision, but it
    // is still untrusted metadata at this boundary. Revalidate its geometry so
    // a stale/corrupt marker cannot route an oversized payload into the
    // in-memory decoder. The independent direct-memory reader remains capped
    // at 128 KiB below.
    if !plan.is_plain()
        || shape.object_size > usize::try_from(INLINE_FAST_PATH_MAX_OBJECT_SIZE).expect("inline fast path limit fits usize")
        || fi.erasure.data_blocks == 0
        || fi.erasure.block_size == 0
    {
        return false;
    }
    true
}

enum SmallWritePath {
    Inline,
    SingleBlockNonInline,
    Pipeline,
    PipelineBatchedLarge,
}

impl SmallWritePath {
    fn metric_label(&self) -> &'static str {
        match self {
            SmallWritePath::Inline => "write_inline",
            SmallWritePath::SingleBlockNonInline => "write_single_block_non_inline",
            SmallWritePath::Pipeline => "write_pipeline",
            SmallWritePath::PipelineBatchedLarge => "write_pipeline_batched_large",
        }
    }

    fn multipart_metric_label(&self) -> &'static str {
        match self {
            SmallWritePath::Inline => "multipart_write_inline",
            SmallWritePath::SingleBlockNonInline => "multipart_write_single_block_non_inline",
            SmallWritePath::Pipeline => "multipart_write_pipeline",
            SmallWritePath::PipelineBatchedLarge => "multipart_write_pipeline_batched_large",
        }
    }
}

fn put_large_batch_min_size_bytes() -> usize {
    *CACHED_PUT_LARGE_BATCH_MIN_SIZE_BYTES.get_or_init(|| {
        rustfs_utils::get_env_usize(ENV_RUSTFS_PUT_LARGE_BATCH_MIN_SIZE_BYTES, DEFAULT_RUSTFS_PUT_LARGE_BATCH_MIN_SIZE_BYTES)
    })
}

fn multipart_put_large_batch_min_size_bytes() -> usize {
    *CACHED_MULTIPART_PUT_LARGE_BATCH_MIN_SIZE_BYTES.get_or_init(|| {
        rustfs_utils::get_env_usize(
            ENV_RUSTFS_MULTIPART_PUT_LARGE_BATCH_MIN_SIZE_BYTES,
            DEFAULT_RUSTFS_MULTIPART_PUT_LARGE_BATCH_MIN_SIZE_BYTES,
        )
    })
}

#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
fn classify_small_write_path(is_inline_buffer: bool, object_size: i64, block_size: usize) -> SmallWritePath {
    if should_use_inline_small_fast_path(is_inline_buffer, object_size, block_size) {
        SmallWritePath::Inline
    } else if should_use_single_block_non_inline_fast_path(is_inline_buffer, object_size, block_size) {
        SmallWritePath::SingleBlockNonInline
    } else {
        SmallWritePath::Pipeline
    }
}

fn classify_put_write_path(is_inline_buffer: bool, object_size: i64, block_size: usize) -> SmallWritePath {
    if should_use_inline_small_fast_path(is_inline_buffer, object_size, block_size) {
        return SmallWritePath::Inline;
    }
    if should_use_single_block_non_inline_fast_path(is_inline_buffer, object_size, block_size) {
        return SmallWritePath::SingleBlockNonInline;
    }

    match usize::try_from(object_size) {
        Ok(size) if !is_inline_buffer && size >= put_large_batch_min_size_bytes() => SmallWritePath::PipelineBatchedLarge,
        _ => SmallWritePath::Pipeline,
    }
}

fn classify_multipart_part_write_path(object_size: i64, block_size: usize) -> SmallWritePath {
    if should_use_single_block_non_inline_fast_path(false, object_size, block_size) {
        return SmallWritePath::SingleBlockNonInline;
    }

    match usize::try_from(object_size) {
        Ok(size) if size >= multipart_put_large_batch_min_size_bytes() => SmallWritePath::PipelineBatchedLarge,
        _ => SmallWritePath::Pipeline,
    }
}

fn known_put_object_storage_size(data_size: i64) -> i64 {
    if data_size >= 0 {
        data_size
    } else {
        HashReader::SIZE_PRESERVE_LAYER
    }
}

/// Shard size the inline admission check evaluates for a single PUT.
///
/// A compressed or encrypted stream reports `SIZE_PRESERVE_LAYER` as its
/// stored size because the transformed length is only known after the write.
/// MinIO's `putObject` sizes such objects for inline admission by their
/// plaintext `ActualSize`; without that fallback every transformed object,
/// however small, lands in `part.1` files. A stream with neither size known
/// yields a negative shard size, which `should_inline` rejects.
fn inline_admission_shard_size(erasure: &coding::Erasure, stored_size: i64, actual_size: i64) -> i64 {
    if stored_size >= 0 {
        return erasure.shard_file_size(stored_size);
    }
    if actual_size > 0 {
        return erasure.shard_file_size(actual_size);
    }
    HashReader::SIZE_PRESERVE_LAYER
}

#[allow(clippy::too_many_arguments)]
async fn build_inline_bitrot_readers(
    files: &[FileInfo],
    total_shards: usize,
    bucket: &str,
    object: &str,
    read_length: usize,
    shard_size: usize,
    checksum_algo: &HashAlgorithm,
    skip_verify_bitrot: bool,
) -> disk::error::Result<Vec<Option<InlineBitrotReader>>> {
    let mut readers = Vec::with_capacity(total_shards);
    for file in files.iter().take(total_shards) {
        let reader = if let Some(data) = &file.data {
            create_bitrot_reader_from_bytes(
                Some(data.clone()),
                None,
                bucket,
                object,
                0,
                read_length,
                shard_size,
                checksum_algo.clone(),
                skip_verify_bitrot,
                false,
            )
            .await?
        } else {
            None
        };
        readers.push(reader);
    }
    Ok(readers)
}

#[allow(clippy::too_many_arguments)]
async fn build_inline_bitrot_readers_from_refs(
    files: &[&FileInfo],
    bucket: &str,
    object: &str,
    read_length: usize,
    shard_size: usize,
    checksum_algo: &HashAlgorithm,
    skip_verify_bitrot: bool,
) -> disk::error::Result<Vec<Option<InlineBitrotReader>>> {
    let mut readers = Vec::with_capacity(files.len());
    for file in files {
        let reader = if let Some(data) = &file.data {
            create_bitrot_reader_from_bytes(
                Some(data.clone()),
                None,
                bucket,
                object,
                0,
                read_length,
                shard_size,
                checksum_algo.clone(),
                skip_verify_bitrot,
                false,
            )
            .await?
        } else {
            None
        };
        readers.push(reader);
    }
    Ok(readers)
}

async fn try_read_inline_data_shards_direct(
    readers: &mut [Option<InlineBitrotReader>],
    data_shards: usize,
    read_length: usize,
    object_size: usize,
) -> Option<Bytes> {
    if object_size == 0 || read_length == 0 || readers.len() < data_shards {
        return None;
    }

    let shards_needed = object_size.div_ceil(read_length);
    if shards_needed > data_shards {
        return None;
    }
    let encoded_capacity = read_length.checked_mul(shards_needed)?;
    let mut body = Vec::with_capacity(encoded_capacity);
    for reader in readers.iter_mut().take(shards_needed) {
        let reader = reader.as_mut()?;
        let Ok(read) = reader.read_appending(&mut body, read_length).await else {
            return None;
        };
        if read != read_length {
            return None;
        }

        if body.len() >= object_size {
            let body = Bytes::from(body);
            return Some(if body.len() == object_size {
                body
            } else {
                body.slice(..object_size)
            });
        }
    }

    None
}

fn can_try_inline_data_shards_direct(object_size: usize, block_size: usize) -> bool {
    object_size > 0 && object_size <= block_size
}

fn inline_erasure_shard_size(block_size: usize, data_shards: usize, uses_legacy: bool) -> usize {
    if block_size == 0 || data_shards == 0 {
        return 0;
    }
    if uses_legacy {
        coding::calc_shard_size_legacy(block_size, data_shards)
    } else {
        coding::calc_shard_size(block_size, data_shards)
    }
}

fn inline_erasure_shard_file_size(total_length: usize, block_size: usize, data_shards: usize, uses_legacy: bool) -> usize {
    if total_length == 0 || block_size == 0 || data_shards == 0 {
        return 0;
    }

    let shard_size = inline_erasure_shard_size(block_size, data_shards, uses_legacy);
    let shard_size_fn = if uses_legacy {
        coding::calc_shard_size_legacy
    } else {
        coding::calc_shard_size
    };
    let num_shards = total_length / block_size;
    let last_block_size = total_length % block_size;
    let last_shard_size = shard_size_fn(last_block_size, data_shards);
    num_shards * shard_size + last_shard_size
}

fn inline_erasure_shard_file_offset(
    start_offset: usize,
    length: usize,
    total_length: usize,
    block_size: usize,
    data_shards: usize,
    uses_legacy: bool,
) -> usize {
    if block_size == 0 || data_shards == 0 {
        return 0;
    }

    let shard_size = inline_erasure_shard_size(block_size, data_shards, uses_legacy);
    let shard_file_size = inline_erasure_shard_file_size(total_length, block_size, data_shards, uses_legacy);
    let end_shard = (start_offset + length) / block_size;
    let till_offset = end_shard * shard_size + shard_size;
    till_offset.min(shard_file_size)
}

fn collect_inline_data_shard_fileinfos_by_index<'a>(
    parts_metadata: &'a [FileInfo],
    fi: &FileInfo,
    data_shards: usize,
    disk_is_online: impl FnMut(usize) -> bool,
) -> Option<Vec<&'a FileInfo>> {
    collect_inline_data_shard_fileinfos_by_index_or_reason(parts_metadata, fi, data_shards, disk_is_online).ok()
}

fn collect_inline_data_shard_fileinfos_by_index_or_reason<'a>(
    parts_metadata: &'a [FileInfo],
    fi: &FileInfo,
    data_shards: usize,
    mut disk_is_online: impl FnMut(usize) -> bool,
) -> std::result::Result<Vec<&'a FileInfo>, &'static str> {
    let distribution = &fi.erasure.distribution;
    let mut data_files = vec![None; data_shards];

    for (disk_index, file_info) in parts_metadata.iter().enumerate() {
        if !disk_is_online(disk_index) {
            continue;
        }
        let Some(&block_index) = distribution.get(disk_index) else {
            return Err(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_GEOMETRY);
        };
        if block_index == 0 || block_index > data_shards {
            continue;
        }
        if file_info.name.is_empty() {
            return Err(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_MISSING_SHARD);
        }
        if file_info.erasure.index != block_index {
            return Err(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_IDENTITY_MISMATCH);
        }
        if !file_info.has_valid_erasure_geometry() {
            return Err(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_GEOMETRY);
        }
        if !core::io_primitives::metadata_early_stop_candidate_matches(file_info, fi) {
            return Err(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_IDENTITY_MISMATCH);
        }
        if file_info.data.as_ref().is_none_or(|data| data.is_empty()) {
            return Err(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_MISSING_PAYLOAD);
        }

        data_files[block_index - 1] = Some(file_info);
    }

    data_files
        .into_iter()
        .collect::<Option<Vec<_>>>()
        .ok_or(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_MISSING_SHARD)
}

impl SetDisks {
    async fn acquire_dist_delete_object_locks_batch(
        &self,
        batch: &rustfs_lock::BatchLockRequest,
    ) -> (HashMap<(String, String), String>, HashSet<String>, Vec<Vec<rustfs_lock::LockId>>) {
        let requests: Vec<rustfs_lock::LockRequest> = batch
            .requests
            .iter()
            .map(|req| {
                rustfs_lock::LockRequest::new(req.key.clone(), rustfs_lock::LockType::Exclusive, self.locker_owner.clone())
                    .with_acquire_timeout(get_lock_acquire_timeout())
                    .with_ttl(rustfs_lock::fast_lock::DEFAULT_LOCK_TIMEOUT)
            })
            .collect();

        let write_quorum = if self.lockers.len() > 1 {
            (self.lockers.len() / 2) + 1
        } else {
            1
        };

        let mut lock_ids_by_object: Vec<Vec<(usize, rustfs_lock::LockId)>> = vec![Vec::new(); requests.len()];
        let mut errors_by_object: Vec<Option<String>> = vec![None; requests.len()];
        #[derive(Clone, Copy, Debug, PartialEq, Eq)]
        enum ObjectLockResolution {
            Pending,
            Succeeded,
            Failed,
        }

        let mut resolution_by_object = vec![ObjectLockResolution::Pending; requests.len()];
        let mut pending_clients = self.lockers.len();
        let mut unresolved_objects = requests.len();
        let mut cleanup_lock_ids_by_client = vec![Vec::new(); self.lockers.len()];

        let mut pending = tokio::task::JoinSet::new();
        for (client_idx, client) in self.lockers.iter().cloned().enumerate() {
            let requests = requests.clone();
            pending.spawn(async move { (client_idx, client.acquire_locks_batch(&requests).await) });
        }

        while unresolved_objects > 0 {
            let Some(join_result) = pending.join_next().await else {
                break;
            };
            pending_clients = pending_clients.saturating_sub(1);

            match join_result {
                Ok((client_idx, Ok(responses))) => {
                    for (req_idx, request) in requests.iter().enumerate() {
                        let response = responses.get(req_idx);
                        match resolution_by_object[req_idx] {
                            ObjectLockResolution::Pending => match response {
                                Some(response) if response.success => {
                                    let lock_id = response
                                        .lock_info
                                        .as_ref()
                                        .map(|lock_info| lock_info.id.clone())
                                        .unwrap_or_else(|| request.lock_id.clone());
                                    lock_ids_by_object[req_idx].push((client_idx, lock_id));
                                }
                                Some(response) => {
                                    if errors_by_object[req_idx].is_none() {
                                        errors_by_object[req_idx] = Some(
                                            response
                                                .error
                                                .clone()
                                                .unwrap_or_else(|| "distributed lock acquisition failed".to_string()),
                                        );
                                    }
                                }
                                None => {
                                    if errors_by_object[req_idx].is_none() {
                                        errors_by_object[req_idx] =
                                            Some(format!("client {client_idx} returned incomplete batch lock response"));
                                    }
                                }
                            },
                            ObjectLockResolution::Succeeded | ObjectLockResolution::Failed => {
                                if let Some(response) = response
                                    && response.success
                                {
                                    let lock_id = response
                                        .lock_info
                                        .as_ref()
                                        .map(|lock_info| lock_info.id.clone())
                                        .unwrap_or_else(|| request.lock_id.clone());
                                    cleanup_lock_ids_by_client[client_idx].push(lock_id);
                                }
                            }
                        }
                    }
                }
                Ok((client_idx, Err(err))) => {
                    for (req_idx, error) in errors_by_object.iter_mut().enumerate().take(requests.len()) {
                        if resolution_by_object[req_idx] == ObjectLockResolution::Pending && error.is_none() {
                            *error = Some(format!("client {client_idx} batch lock request failed: {err}"));
                        }
                    }
                }
                Err(err) => {
                    for (req_idx, error) in errors_by_object.iter_mut().enumerate().take(requests.len()) {
                        if resolution_by_object[req_idx] == ObjectLockResolution::Pending && error.is_none() {
                            *error = Some(format!("batch lock task join failed: {err}"));
                        }
                    }
                }
            }

            for req_idx in 0..requests.len() {
                if resolution_by_object[req_idx] != ObjectLockResolution::Pending {
                    continue;
                }

                let success_count = lock_ids_by_object[req_idx].len();
                if success_count >= write_quorum {
                    resolution_by_object[req_idx] = ObjectLockResolution::Succeeded;
                    unresolved_objects -= 1;
                } else if success_count + pending_clients < write_quorum {
                    resolution_by_object[req_idx] = ObjectLockResolution::Failed;
                    unresolved_objects -= 1;
                }
            }
        }

        if issue3031_diag_enabled() {
            let succeeded_count = resolution_by_object
                .iter()
                .filter(|resolution| matches!(resolution, ObjectLockResolution::Succeeded))
                .count();
            let failed_count = resolution_by_object
                .iter()
                .filter(|resolution| matches!(resolution, ObjectLockResolution::Failed))
                .count();
            let pending_count = resolution_by_object
                .iter()
                .filter(|resolution| matches!(resolution, ObjectLockResolution::Pending))
                .count();
            warn!(
                target: "rustfs_ecstore::set_disk",
                request_count = requests.len(),
                locker_count = self.lockers.len(),
                write_quorum,
                succeeded_count,
                failed_count,
                pending_count,
                pending_clients,
                errors_by_object = ?errors_by_object,
                "issue3031_delete_objects_dist_batch_lock_summary"
            );
        }

        if !pending.is_empty() {
            let cleanup_requests = requests.clone();
            let lockers = self.lockers.clone();
            let handle = tokio::spawn(
                async move {
                    let mut late_lock_ids_by_client = vec![Vec::new(); lockers.len()];
                    let mut pending = pending;
                    while let Some(join_result) = pending.join_next().await {
                        match join_result {
                            Ok((client_idx, Ok(responses))) => {
                                for (req_idx, request) in cleanup_requests.iter().enumerate() {
                                    if let Some(response) = responses.get(req_idx)
                                        && response.success
                                    {
                                        let lock_id = response
                                            .lock_info
                                            .as_ref()
                                            .map(|lock_info| lock_info.id.clone())
                                            .unwrap_or_else(|| request.lock_id.clone());
                                        if let Some(client_locks) = late_lock_ids_by_client.get_mut(client_idx) {
                                            client_locks.push(lock_id);
                                        }
                                    }
                                }
                            }
                            Ok((_client_idx, Err(err))) => {
                                warn!("late distributed delete lock batch request failed: {}", err);
                            }
                            Err(err) => {
                                warn!("late distributed delete lock batch task join failed: {}", err);
                            }
                        }
                    }

                    join_all(lockers.iter().cloned().enumerate().filter_map(|(client_idx, client)| {
                        let lock_ids = late_lock_ids_by_client.get(client_idx).cloned().unwrap_or_default();
                        if lock_ids.is_empty() {
                            None
                        } else {
                            Some(async move {
                                if let Err(err) = client.release_locks_batch(&lock_ids).await {
                                    warn!(
                                        client_idx,
                                        lock_count = lock_ids.len(),
                                        "failed to cleanup late distributed delete locks in batch: {}",
                                        err
                                    );
                                }
                            })
                        }
                    }))
                    .await;
                }
                .instrument(tracing::Span::current()),
            );
            drop(handle);
        }

        let mut failed_map = HashMap::new();
        let mut locked_objects = HashSet::new();
        let mut held_lock_ids_by_client = vec![Vec::new(); self.lockers.len()];
        let mut rollback_lock_ids_by_client = vec![Vec::new(); self.lockers.len()];

        for (req_idx, req) in batch.requests.iter().enumerate() {
            let success_count = lock_ids_by_object[req_idx].len();
            match resolution_by_object[req_idx] {
                ObjectLockResolution::Succeeded => {
                    for (client_idx, lock_id) in lock_ids_by_object[req_idx].drain(..) {
                        held_lock_ids_by_client[client_idx].push(lock_id);
                    }
                    locked_objects.insert(req.key.object.as_ref().to_string());
                }
                ObjectLockResolution::Pending | ObjectLockResolution::Failed => {
                    for (client_idx, lock_id) in lock_ids_by_object[req_idx].drain(..) {
                        rollback_lock_ids_by_client[client_idx].push(lock_id);
                    }
                    failed_map.insert(
                        (req.key.bucket.as_ref().to_string(), req.key.object.as_ref().to_string()),
                        errors_by_object[req_idx].clone().unwrap_or_else(|| {
                            format!("failed to acquire distributed delete lock quorum: {success_count}/{write_quorum}")
                        }),
                    );
                }
            }
        }

        for (client_idx, cleanup_ids) in cleanup_lock_ids_by_client.into_iter().enumerate() {
            rollback_lock_ids_by_client[client_idx].extend(cleanup_ids);
        }

        self.release_dist_delete_object_locks_batch(rollback_lock_ids_by_client).await;

        (failed_map, locked_objects, held_lock_ids_by_client)
    }

    async fn release_dist_delete_object_locks_batch(&self, lock_ids_by_client: Vec<Vec<rustfs_lock::LockId>>) {
        join_all(self.lockers.iter().cloned().enumerate().filter_map(|(client_idx, client)| {
            let lock_ids = lock_ids_by_client.get(client_idx).cloned().unwrap_or_default();
            if lock_ids.is_empty() {
                None
            } else {
                Some(async move {
                    if let Err(err) = client.release_locks_batch(&lock_ids).await {
                        warn!(
                            client_idx,
                            lock_count = lock_ids.len(),
                            "failed to release distributed delete locks in batch: {}",
                            err
                        );
                    }
                })
            }
        }))
        .await;
    }
}

impl SetDisks {
    pub(crate) async fn storage_info_snapshot(&self) -> rustfs_madmin::StorageInfo {
        let disks = self.get_disks_internal().await;

        get_storage_info(&disks, &self.set_endpoints).await
    }

    pub(crate) async fn local_storage_info_snapshot(&self) -> rustfs_madmin::StorageInfo {
        let disks = self.get_disks_internal().await;

        let mut local_disks: Vec<Option<DiskStore>> = Vec::new();
        let mut local_endpoints = Vec::new();

        for (i, ep) in self.set_endpoints.iter().enumerate() {
            if ep.is_local {
                local_disks.push(disks[i].clone());
                local_endpoints.push(ep.clone());
            }
        }

        get_storage_info(&local_disks, &local_endpoints).await
    }

    pub(crate) async fn disk_inventory(&self) -> Vec<Option<DiskStore>> {
        self.get_disks_internal().await
    }
}

fn check_object_lock_retention_update(bucket: &str, object: &str, obj_info: &ObjectInfo, opts: &ObjectOptions) -> Result<()> {
    if let Some(retention) = &opts.object_lock_retention
        && check_retention_for_modification(
            &obj_info.user_defined,
            retention
                .mode
                .as_deref()
                .and_then(crate::bucket::object_lock::types::RetentionMode::parse_exact),
            retention.retain_until,
            retention.bypass_governance,
        )
        .is_some()
    {
        return Err(StorageError::PrefixAccessDenied(bucket.to_string(), object.to_string()));
    }

    Ok(())
}

/// Whether the batch-delete path must stat each object under the held lock to
/// run [`check_object_lock_delete`] (the #4297 protection).
///
/// Under S3 semantics retention/legal-hold metadata can only be written to
/// buckets created with Object Lock enabled (`validate_bucket_object_lock_enabled`
/// guards every write surface), and default retention only exists with a bucket
/// lock configuration, so for buckets without Object Lock the per-object stat in
/// `delete_objects` has no consumer and can be skipped (backlog#929 / HP-8).
///
/// Fail closed: when bucket metadata cannot be resolved the check stays on, so
/// object-lock protection is never skipped because of a metadata lookup miss.
#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
pub(crate) fn object_lock_delete_check_required(bucket_meta: Option<&crate::bucket::metadata::BucketMetadata>) -> bool {
    bucket_meta.is_none_or(|meta| meta.object_lock_checks_required())
}

fn restore_expiry_snapshot_matches(obj_info: &ObjectInfo, opts: &ObjectOptions) -> bool {
    let expected = &opts.transition;
    expected.expire_restored
        && expected.status == TRANSITION_COMPLETE
        && obj_info.transitioned_object.status == TRANSITION_COMPLETE
        && !obj_info.transitioned_object.name.is_empty()
        && !obj_info.transitioned_object.tier.is_empty()
        && expected.tier == obj_info.transitioned_object.tier
        && expected.expected_remote_name == obj_info.transitioned_object.name
        && expected.expected_remote_version_id == obj_info.transitioned_object.version_id
        && !expected.etag.is_empty()
        && obj_info.etag.as_deref() == Some(expected.etag.as_str())
        && expected.expected_data_dir.is_some()
        && expected.expected_data_dir == obj_info.data_dir
        && obj_info.restore_expires == Some(expected.restore_expiry)
        && !obj_info.restore_ongoing
        // Deliberately no `restore_expiry <= now` clause. Whether the restored
        // copy is due to expire is the ILM evaluator's decision, already made
        // when it emitted DeleteRestoredAction; re-deriving it here only adds a
        // way for a legitimate action to be rejected. The stale-event risk it
        // looks like it covers is already covered above: a re-restore rewrites
        // `restore_expires`, so a replayed event fails the equality check.
        && match obj_info.version_id {
            Some(version_id) => opts.version_id.as_deref().and_then(|value| Uuid::parse_str(value).ok()) == Some(version_id),
            None => opts.version_id.is_none(),
        }
}

async fn check_object_lock_delete(
    ctx: &InstanceContext,
    bucket: &str,
    object: &str,
    obj_info: &ObjectInfo,
    opts: &ObjectOptions,
) -> Result<()> {
    if crate::bucket::utils::is_meta_bucketname(bucket) {
        return Ok(());
    }
    if opts.transition.expire_restored {
        return restore_expiry_snapshot_matches(obj_info, opts)
            .then_some(())
            .ok_or(StorageError::PreconditionFailed);
    }
    if set_disk_delete_creates_delete_marker(opts) {
        return Ok(());
    }

    // An authorized replicated version purge already passed this gate on the
    // source with the bypass it carried there, so it clears GOVERNANCE
    // retention here without the header; COMPLIANCE and legal hold still
    // block below (see `replication_delete_may_bypass_governance`, #6850).
    let bypass_governance = opts
        .object_lock_delete
        .as_ref()
        .is_some_and(|delete_opts| delete_opts.bypass_governance)
        || replication_delete_may_bypass_governance(opts);
    let blocked = match opts.object_lock_config_snapshot.as_deref() {
        Some(snapshot) => check_object_lock_for_deletion_with_state(snapshot.state(), obj_info, bypass_governance)?.is_some(),
        None => {
            let state = metadata_sys::get_object_lock_config_state_in(ctx, bucket).await?;
            check_object_lock_for_deletion_with_state(&state, obj_info, bypass_governance)?.is_some()
        }
    };
    if blocked {
        return Err(StorageError::PrefixAccessDenied(bucket.to_string(), object.to_string()));
    }

    Ok(())
}

fn ensure_delete_commit_locks_held(
    lock_guard: Option<&ObjectLockDiagGuard>,
    bucket: &str,
    object: &str,
    opts: &ObjectOptions,
) -> Result<()> {
    if lock_guard.is_some_and(ObjectLockDiagGuard::is_lock_lost)
        || opts
            .namespace_lock_fence
            .as_ref()
            .is_some_and(NamespaceLockFence::is_lock_lost)
        || opts
            .bucket_lifecycle_lock_fence
            .as_ref()
            .is_some_and(NamespaceLockFence::is_lock_lost)
    {
        return Err(StorageError::NamespaceLockQuorumUnavailable {
            mode: "delete_object_commit",
            bucket: bucket.to_string(),
            object: object.to_string(),
            required: 1,
            achieved: 0,
        });
    }

    Ok(())
}

fn set_disk_delete_creates_delete_marker(opts: &ObjectOptions) -> bool {
    opts.version_id.is_none() && opts.versioned && !opts.version_suspended
}

fn should_preserve_delete_replication_state(opts: &ObjectOptions) -> bool {
    opts.delete_replication.as_ref().is_some_and(|state| {
        state.replica_status == ReplicationStatusType::Replica
            || (!state.replicate_decision_str.is_empty()
                && (!state.composite_replication_status().is_empty() || !state.composite_version_purge_status().is_empty()))
    }) || opts.version_purge_status() == VersionPurgeStatusType::Complete
}

fn should_force_delete_marker_for_missing_version(opts: &ObjectOptions) -> bool {
    opts.delete_marker || ((opts.versioned || opts.version_suspended) && opts.version_id.is_none() && !opts.data_movement)
}

/// Whether a plain client delete addressed to an explicit version removed an
/// existing delete marker, so the response must carry delete-marker semantics
/// (`x-amz-delete-marker: true` / `DeleteMarker` in `DeleteObjects` entries).
///
/// This is a response-side classification only. It must never feed the
/// storage write shape: a delete request marked `deleted` with the null
/// identity makes `FileMeta::delete_version` re-create the marker it just
/// removed (that is the suspended-bucket "delete mints a marker" write
/// path), which is why `resolve_delete_version_state` cannot simply report
/// `delete_marker` for suspended buckets (issue #6745).
///
/// Purge/replica replication shapes are excluded, mirroring the clauses in
/// `resolve_delete_version_state` that clear `delete_marker` for them.
fn explicit_delete_removed_marker(opts: &ObjectOptions, goi: &ObjectInfo, version_found: bool) -> bool {
    opts.version_id.is_some()
        && version_found
        && goi.delete_marker
        && goi.version_purge_status.is_empty()
        && opts.version_purge_status().is_empty()
        && opts.delete_marker_replication_status().is_empty()
}

fn resolve_delete_version_state(opts: &ObjectOptions, goi: &ObjectInfo, version_found: bool) -> (bool, bool) {
    let mut mark_delete = goi.version_id.is_some() || ((opts.versioned || opts.version_suspended) && opts.version_id.is_none());
    let mut delete_marker = opts.versioned;

    if opts.version_id.is_some() {
        // Decommission/rebalance may recreate a delete marker on a new pool before that
        // exact version exists there, so we must still treat it as a mark-delete write.
        let data_movement_missing_delete_marker = opts.data_movement && opts.delete_marker && !version_found;
        if data_movement_missing_delete_marker {
            mark_delete = true;
            delete_marker = true;
        }

        let delete_marker_version_purge = version_found && goi.delete_marker && !opts.version_purge_status().is_empty();

        if version_found && opts.delete_marker_replication_status() == ReplicationStatusType::Replica {
            mark_delete = false;
        }

        if !data_movement_missing_delete_marker
            && opts.version_purge_status().is_empty()
            && opts.delete_marker_replication_status().is_empty()
        {
            mark_delete = false;
        }

        if opts.version_purge_status() == VersionPurgeStatusType::Complete {
            mark_delete = false;
        }

        let replica_delete_marker_version_purge =
            version_found && goi.delete_marker && opts.delete_marker_replication_status() == ReplicationStatusType::Replica;

        if delete_marker_version_purge {
            mark_delete = false;
        }

        if !version_found && !opts.delete_marker && opts.delete_marker_replication_status() == ReplicationStatusType::Replica {
            delete_marker = false;
        }

        if version_found
            && (!goi.version_purge_status.is_empty()
                || !goi.delete_marker
                || replica_delete_marker_version_purge
                || delete_marker_version_purge)
        {
            delete_marker = false;
        }
    }

    (mark_delete, delete_marker)
}

impl SetDisks {
    /// Publish an internal tier free-version record without changing its
    /// delete-marker shape or remote-tier identity. The caller holds the
    /// source and target object locks; a write quorum is required before the
    /// source cleanup may remove the original record.
    #[tracing::instrument(skip(self, fi, opts))]
    pub(crate) async fn decommission_tier_free_version(
        &self,
        bucket: &str,
        object: &str,
        fi: &FileInfo,
        opts: &ObjectOptions,
    ) -> Result<()> {
        if !fi.deleted || !fi.tier_free_version() {
            return Err(Error::other("decommission tier free-version write requires a free version record"));
        }
        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;

        let write_quorum = self.default_write_quorum();
        if self
            .count_decommission_tier_free_version_equivalents(bucket, object, fi)
            .await?
            >= write_quorum
        {
            ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;
            return Ok(());
        }
        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;

        let disks = self.disks.read().await.clone();
        let futures = disks.into_iter().map(|disk| {
            let file_info = fi.clone();
            async move {
                if let Some(disk) = disk {
                    disk.write_metadata("", bucket, object, file_info).await
                } else {
                    Err(DiskError::DiskNotFound)
                }
            }
        });

        let mut errs = Vec::new();
        for result in join_all(futures).await {
            match result {
                Ok(_) => errs.push(None),
                Err(err) => errs.push(Some(err)),
            }
        }

        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;

        resolve_tiered_decommission_write_quorum_result(&errs, write_quorum, bucket, object)
    }

    async fn count_decommission_tier_free_version_equivalents(&self, bucket: &str, object: &str, fi: &FileInfo) -> Result<usize> {
        // The caller holds the source and target object locks. Inspect every
        // target disk before an idempotent return or metadata fan-out so a
        // sub-quorum conflict cannot be hidden by a successful quorum.
        let disks = self.disks.read().await.clone();
        let preflight = disks.iter().map(|disk| async {
            match disk {
                Some(disk) => inspect_decommission_tier_free_version_target(disk, bucket, object, fi).await,
                None => Ok(false),
            }
        });
        let mut equivalent = 0;
        for result in join_all(preflight).await {
            if result? {
                equivalent += 1;
            }
        }
        Ok(equivalent)
    }

    pub(crate) async fn has_decommission_tier_free_version_write_quorum(
        &self,
        bucket: &str,
        object: &str,
        fi: &FileInfo,
        opts: &ObjectOptions,
    ) -> Result<bool> {
        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;
        let equivalent = self
            .count_decommission_tier_free_version_equivalents(bucket, object, fi)
            .await?;
        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;
        Ok(equivalent >= self.default_write_quorum())
    }

    #[tracing::instrument(skip(self, fi, opts))]
    pub(crate) async fn decommission_tiered_object(
        &self,
        bucket: &str,
        object: &str,
        fi: &FileInfo,
        opts: &ObjectOptions,
    ) -> Result<()> {
        let storage_class_config = self.storage_class_config_snapshot();
        let bucket_lifecycle_guard = if let Some(expected_incarnation_id) = opts.expected_bucket_incarnation_id
            && opts.bucket_lifecycle_lock_fence.is_none()
            && !crate::bucket::utils::is_meta_bucketname(bucket)
        {
            Some(
                metadata_sys::object_store_in(&self.ctx)
                    .await?
                    .acquire_bucket_incarnation_fence(bucket, expected_incarnation_id)
                    .await?,
            )
        } else {
            None
        };
        let _lock_guard = if !opts.no_lock {
            Some(
                self.new_ns_lock(bucket, object)
                    .await?
                    .get_write_lock(get_lock_acquire_timeout())
                    .await
                    .map_err(|e| self.map_namespace_lock_error(bucket, object, "write", e))?,
            )
        } else {
            None
        };

        if opts.http_preconditions.is_some()
            && let Some(err) = self.check_write_precondition(bucket, object, opts).await
        {
            return Err(err);
        }

        let disks = self.disks.read().await.clone();
        let storage_class = opts.user_defined.get(metadata_keys::STORAGE_CLASS).map(String::as_str);
        let layout = resolve_write_layout(
            &storage_class_config,
            self.pool_index,
            disks.len(),
            self.default_parity_count,
            storage_class,
            opts.max_parity,
        )?;
        let fi = build_tiered_decommission_file_info(bucket, object, fi, layout);
        let write_quorum = layout.write_quorum;
        #[cfg(test)]
        Self::pause_tiered_metadata_commit(bucket, object).await;
        if _lock_guard.as_ref().is_some_and(|guard| guard.is_lock_lost())
            || opts
                .namespace_lock_fence
                .as_ref()
                .is_some_and(NamespaceLockFence::is_lock_lost)
            || opts
                .bucket_lifecycle_lock_fence
                .as_ref()
                .is_some_and(NamespaceLockFence::is_lock_lost)
            || bucket_lifecycle_guard.as_ref().is_some_and(|guard| guard.is_lock_lost())
        {
            return Err(StorageError::NamespaceLockQuorumUnavailable {
                mode: "decommission_tiered_object_commit",
                bucket: bucket.to_string(),
                object: object.to_string(),
                required: 1,
                achieved: 0,
            });
        }
        // Rebuilt tiered metadata starts with index zero, but shuffling validates
        // each source slot before assigning the shuffled index below.
        let parts_metadata: Vec<FileInfo> = (0..disks.len())
            .map(|disk_index| {
                let mut part = fi.clone();
                part.erasure.index = fi.erasure.distribution[disk_index];
                part
            })
            .collect();
        let (shuffle_disks, parts_metadata) = Self::shuffle_disks_and_parts_metadata(&disks, &parts_metadata, &fi);

        let mut errs = Vec::with_capacity(shuffle_disks.len());
        let mut futures = Vec::with_capacity(shuffle_disks.len());
        for (index, disk) in shuffle_disks.iter().enumerate() {
            let mut file_info = parts_metadata[index].clone();
            file_info.erasure.index = index + 1;
            futures.push(async move {
                if let Some(disk) = disk {
                    disk.write_metadata("", bucket, object, file_info).await
                } else {
                    Err(DiskError::DiskNotFound)
                }
            });
        }

        for result in join_all(futures).await {
            match result {
                Ok(_) => errs.push(None),
                Err(err) => errs.push(Some(err)),
            }
        }

        resolve_tiered_decommission_write_quorum_result(&errs, write_quorum, bucket, object)
    }
}

#[cfg(test)]
struct TieredMetadataCommitBarrierState {
    bucket: String,
    object: String,
    arrived: tokio::sync::Notify,
    release: tokio::sync::Notify,
}

#[cfg(test)]
pub(crate) struct TieredMetadataCommitBarrier {
    state: Arc<TieredMetadataCommitBarrierState>,
}

#[cfg(test)]
static TIERED_METADATA_COMMIT_BARRIER: std::sync::OnceLock<std::sync::Mutex<Option<Arc<TieredMetadataCommitBarrierState>>>> =
    std::sync::OnceLock::new();

#[cfg(test)]
impl TieredMetadataCommitBarrier {
    pub(crate) fn install(bucket: &str, object: &str) -> Self {
        let state = Arc::new(TieredMetadataCommitBarrierState {
            bucket: bucket.to_string(),
            object: object.to_string(),
            arrived: tokio::sync::Notify::new(),
            release: tokio::sync::Notify::new(),
        });
        let mut slot = TIERED_METADATA_COMMIT_BARRIER
            .get_or_init(|| std::sync::Mutex::new(None))
            .lock()
            .expect("tiered metadata commit barrier should not be poisoned");
        assert!(slot.is_none(), "tiered metadata commit barrier must be unique");
        *slot = Some(Arc::clone(&state));
        Self { state }
    }

    pub(crate) async fn wait_until_paused(&self) {
        tokio::time::timeout(Duration::from_secs(30), self.state.arrived.notified())
            .await
            .expect("tiered metadata write should reach its deterministic commit barrier");
    }

    pub(crate) fn release(&self) {
        self.state.release.notify_one();
    }
}

#[cfg(test)]
impl Drop for TieredMetadataCommitBarrier {
    fn drop(&mut self) {
        self.state.release.notify_one();
        let mut slot = TIERED_METADATA_COMMIT_BARRIER
            .get_or_init(|| std::sync::Mutex::new(None))
            .lock()
            .expect("tiered metadata commit barrier should not be poisoned");
        if slot.as_ref().is_some_and(|state| Arc::ptr_eq(state, &self.state)) {
            *slot = None;
        }
    }
}

#[derive(Debug, PartialEq, Eq)]
struct ObjProps {
    successor_mod_time: Option<OffsetDateTime>,
    num_versions: usize,
}

impl Hash for ObjProps {
    fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
        self.successor_mod_time.hash(state);
        self.num_versions.hash(state);
    }
}

fn is_object_dangling(
    meta_arr: &[FileInfo],
    errs: &[Option<DiskError>],
    data_errs_by_part: &HashMap<usize, Vec<usize>>,
) -> (FileInfo, bool) {
    let (not_found_meta_errs, non_actionable_meta_errs) = dangling_meta_errs_count(errs);

    let (mut not_found_parts_errs, mut non_actionable_parts_errs) = (0, 0);

    data_errs_by_part.iter().for_each(|(_, v)| {
        let (nf, na) = dangling_part_errs_count(v);
        if nf > not_found_parts_errs {
            (not_found_parts_errs, non_actionable_parts_errs) = (nf, na);
        }
    });

    let mut valid_meta = FileInfo::default();

    for fi in meta_arr.iter() {
        if file_info_is_valid_for_metadata(fi) {
            valid_meta = fi.clone();
            break;
        }
    }

    if !file_info_is_valid_for_metadata(&valid_meta) {
        let data_blocks = meta_arr.len().div_ceil(2);
        if not_found_parts_errs > data_blocks {
            return (valid_meta, true);
        }

        return (valid_meta, false);
    }

    if non_actionable_meta_errs > 0 || non_actionable_parts_errs > 0 {
        return (valid_meta, false);
    }

    if valid_meta.is_canonical_delete_marker() {
        let data_blocks = errs.len().div_ceil(2);
        return (valid_meta, not_found_meta_errs > data_blocks);
    }

    if not_found_meta_errs > 0 && not_found_meta_errs > valid_meta.erasure.parity_blocks {
        return (valid_meta, true);
    }

    if !valid_meta.is_remote() && not_found_parts_errs > 0 && not_found_parts_errs > valid_meta.erasure.parity_blocks {
        return (valid_meta, true);
    }

    (valid_meta, false)
}

fn dangling_meta_errs_count(cerrs: &[Option<DiskError>]) -> (usize, usize) {
    let (mut not_found_count, mut non_actionable_count) = (0, 0);
    cerrs.iter().for_each(|err| {
        if let Some(err) = err {
            if err == &DiskError::FileNotFound || err == &DiskError::FileVersionNotFound {
                not_found_count += 1;
            } else {
                non_actionable_count += 1;
            }
        }
    });

    (not_found_count, non_actionable_count)
}

fn dangling_part_errs_count(results: &[usize]) -> (usize, usize) {
    let (mut not_found_count, mut non_actionable_count) = (0, 0);
    results.iter().for_each(|result| {
        if *result == CHECK_PART_SUCCESS {
            // skip
        } else if *result == CHECK_PART_FILE_NOT_FOUND {
            not_found_count += 1;
        } else {
            non_actionable_count += 1;
        }
    });

    (not_found_count, non_actionable_count)
}

fn is_object_dir_dangling(errs: &[Option<DiskError>]) -> bool {
    let mut found = 0;
    let mut not_found = 0;
    let mut found_not_empty = 0;
    let mut other_found = 0;
    errs.iter().for_each(|err| {
        if err.is_none() {
            found += 1;
        } else if let Some(err) = err {
            if err == &DiskError::FileNotFound || err == &DiskError::VolumeNotFound {
                not_found += 1;
            } else if err == &DiskError::VolumeNotEmpty {
                found_not_empty += 1;
            } else {
                other_found += 1;
            }
        }
    });

    found = found + found_not_empty + other_found;
    found < not_found && found > 0
}

fn join_errs(errs: &[Option<DiskError>]) -> String {
    let errs = errs
        .iter()
        .map(|err| {
            if let Some(err) = err {
                return err.to_string();
            }
            "<nil>".to_string()
        })
        .collect::<Vec<_>>();

    errs.join(", ")
}

async fn verify_inline_part_bitrot(meta: &FileInfo) -> disk::error::Result<()> {
    meta.validate_for_metadata_read()?;
    let [part] = meta.parts.as_slice() else {
        return Err(DiskError::FileCorrupt);
    };
    let data = meta.data.as_deref().ok_or(DiskError::FileCorrupt)?;
    let checksum = meta.erasure.get_checksum_info(part.number);
    let algo = match checksum.algorithm {
        HashAlgorithm::HighwayHash256S if meta.uses_legacy_checksum => HashAlgorithm::HighwayHash256SLegacy,
        algo @ (HashAlgorithm::HighwayHash256S | HashAlgorithm::HighwayHash256SLegacy) => algo,
        _ => return Err(DiskError::BitrotHashAlgoInvalid),
    };
    let shard_size = inline_erasure_shard_size(meta.erasure.block_size, meta.erasure.data_blocks, meta.uses_legacy_checksum);
    let part_size =
        inline_erasure_shard_file_size(part.size, meta.erasure.block_size, meta.erasure.data_blocks, meta.uses_legacy_checksum);
    // Check framing arithmetic and the physical buffer length before the shared
    // verifier sizes its scratch buffer from the metadata.
    let encoded_size = part_size
        .div_ceil(shard_size)
        .checked_mul(algo.size())
        .and_then(|hash_size| part_size.checked_add(hash_size))
        .ok_or(DiskError::FileCorrupt)?;
    if encoded_size != data.len() {
        return Err(DiskError::FileCorrupt);
    }
    coding::bitrot_verify(Cursor::new(data), encoded_size, part_size, algo, shard_size)
        .await
        .map_err(|_| DiskError::FileCorrupt)
}

fn validate_deep_scan_results(results: &[usize], expected_parts: usize) -> disk::error::Result<()> {
    if results.len() != expected_parts {
        return Err(DiskError::other(format!(
            "incomplete deep scan result: expected {expected_parts} parts, received {}",
            results.len()
        )));
    }
    for (part, status) in results.iter().enumerate() {
        match *status {
            CHECK_PART_SUCCESS | CHECK_PART_FILE_NOT_FOUND | CHECK_PART_FILE_CORRUPT => {}
            CHECK_PART_DISK_NOT_FOUND => return Err(DiskError::DiskNotFound),
            crate::disk::CHECK_PART_VOLUME_NOT_FOUND => return Err(DiskError::VolumeNotFound),
            _ => {
                return Err(DiskError::other(format!(
                    "incomplete deep scan result: part {part} has unverified status {status}"
                )));
            }
        }
    }
    Ok(())
}

/// disks_with_all_partsv2 is a corrected version based on Go implementation.
/// It sets partsMetadata and onlineDisks when xl.meta is inexistant/corrupted or outdated.
/// It also checks if the status of each part (corrupted, missing, ok) in each drive.
/// Returns (availableDisks, dataErrsByDisk, dataErrsByPart).
#[allow(clippy::too_many_arguments)]
async fn disks_with_all_parts(
    online_disks: &mut [Option<DiskStore>],
    parts_metadata: &mut [FileInfo],
    errs: &[Option<DiskError>],
    latest_meta: &FileInfo,
    filter_by_etag: bool,
    bucket: &str,
    object: &str,
    scan_mode: HealScanMode,
) -> disk::error::Result<(HashMap<usize, Vec<usize>>, HashMap<usize, Vec<usize>>)> {
    let object_name = latest_meta.name.clone();

    // Initialize dataErrsByDisk and dataErrsByPart with 0 (CHECK_PART_UNKNOWN) to match Go
    let mut data_errs_by_disk: HashMap<usize, Vec<usize>> = HashMap::new();
    for i in 0..online_disks.len() {
        data_errs_by_disk.insert(i, vec![CHECK_PART_UNKNOWN; latest_meta.parts.len()]);
    }
    let mut data_errs_by_part: HashMap<usize, Vec<usize>> = HashMap::new();
    for i in 0..latest_meta.parts.len() {
        data_errs_by_part.insert(i, vec![CHECK_PART_UNKNOWN; online_disks.len()]);
    }

    // Check for inconsistent erasure distribution
    let mut inconsistent = 0;
    for (index, meta) in parts_metadata.iter().enumerate() {
        if !file_info_is_valid_for_metadata(meta) {
            // Since for majority of the cases erasure.Index matches with erasure.Distribution we can
            // consider the offline disks as consistent.
            continue;
        }
        if !meta.is_canonical_delete_marker() {
            if meta.erasure.distribution.len() != online_disks.len() {
                // Erasure distribution seems to have lesser
                // number of items than number of online disks.
                inconsistent += 1;
                continue;
            }
            if !meta.erasure.distribution.is_empty()
                && index < meta.erasure.distribution.len()
                && meta.erasure.distribution[index] != meta.erasure.index
            {
                // Mismatch indexes with distribution order
                inconsistent += 1;
            }
        }
    }

    let erasure_distribution_reliable = inconsistent <= parts_metadata.len() / 2;

    // Initialize metaErrs
    let mut meta_errs = Vec::with_capacity(errs.len());
    for _ in 0..errs.len() {
        meta_errs.push(None);
    }

    let online_disks_len = online_disks.len();

    // Process meta errors
    for (index, disk_op) in online_disks.iter_mut().enumerate() {
        if let Some(err) = &errs[index] {
            meta_errs[index] = Some(err.clone());
            continue;
        }

        if disk_op.is_none() {
            meta_errs[index] = Some(DiskError::DiskNotFound);
            continue;
        }

        let meta = &parts_metadata[index];

        let corrupted = if filter_by_etag {
            latest_meta.get_etag() != meta.get_etag()
        } else {
            !meta.mod_time.eq(&latest_meta.mod_time) || !meta.data_dir.eq(&latest_meta.data_dir)
        };

        if corrupted {
            debug!(
                event = EVENT_SET_DISK_HEAL,
                component = LOG_COMPONENT_ECSTORE,
                subsystem = LOG_SUBSYSTEM_SET_DISK,
                bucket,
                object = %object_name,
                disk_index = index,
                state = "metadata_corrupt",
                "Set disk object metadata is corrupt"
            );
            meta_errs[index] = Some(DiskError::FileCorrupt);
            parts_metadata[index] = FileInfo::default();
            *disk_op = None;

            continue;
        }

        if erasure_distribution_reliable {
            if !file_info_is_valid_for_metadata(meta) {
                debug!(
                    event = EVENT_SET_DISK_HEAL,
                    component = LOG_COMPONENT_ECSTORE,
                    subsystem = LOG_SUBSYSTEM_SET_DISK,
                    bucket,
                    object = %object_name,
                    disk_index = index,
                    state = "metadata_invalid",
                    "Set disk object metadata is invalid"
                );
                parts_metadata[index] = FileInfo::default();
                meta_errs[index] = Some(DiskError::FileCorrupt);
                *disk_op = None;
                continue;
            }

            if !meta.is_canonical_delete_marker() && meta.erasure.distribution.len() != online_disks_len {
                // Erasure distribution is not the same as onlineDisks
                // attempt a fix if possible, assuming other entries
                // might have the right erasure distribution.
                debug!(
                    event = EVENT_SET_DISK_HEAL,
                    component = LOG_COMPONENT_ECSTORE,
                    subsystem = LOG_SUBSYSTEM_SET_DISK,
                    bucket,
                    object = %object_name,
                    disk_index = index,
                    state = "erasure_distribution_mismatch",
                    "Set disk erasure distribution mismatched online disks"
                );
                parts_metadata[index] = FileInfo::default();
                meta_errs[index] = Some(DiskError::FileCorrupt);
                *disk_op = None;
                continue;
            }
        }
    }

    // Copy meta errors to part errors
    for (index, err) in meta_errs.iter().enumerate() {
        if err.is_some() {
            let part_err = conv_part_err_to_int(err);
            for p in 0..latest_meta.parts.len() {
                if let Some(vec) = data_errs_by_part.get_mut(&p)
                    && index < vec.len()
                {
                    vec[index] = part_err;
                }
            }
        }
    }

    if scan_mode == HealScanMode::Deep
        && !latest_meta.deleted
        && !latest_meta.is_remote()
        && latest_meta.parts.iter().any(|part| part.integrity.is_some())
    {
        crate::io_support::shard_integrity::verify_deep_parts(
            parts_metadata,
            online_disks,
            latest_meta,
            bucket,
            object,
            &mut data_errs_by_part,
        )
        .await?;
        populate_data_errs_by_disk(&mut data_errs_by_disk, &data_errs_by_part);
        return Ok((data_errs_by_disk, data_errs_by_part));
    }

    // Check data for each disk
    for (index, disk) in online_disks.iter().enumerate() {
        if meta_errs[index].is_some() {
            continue;
        }

        let disk = if let Some(disk) = disk {
            disk
        } else {
            continue;
        };

        let meta = &mut parts_metadata[index];
        if meta.is_canonical_delete_marker() || meta.is_remote() {
            continue;
        }

        // Normal scans only check presence. Deep scans must verify inline bytes
        // before deciding whether reconstruction (and its bitrot readers) is needed.
        if (meta.data.is_some() || meta.size == 0) && !meta.parts.is_empty() {
            let part_status = if scan_mode == HealScanMode::Deep && meta.data.is_some() {
                match verify_inline_part_bitrot(meta).await {
                    Ok(()) => CHECK_PART_SUCCESS,
                    Err(DiskError::FileCorrupt) => CHECK_PART_FILE_CORRUPT,
                    Err(err) => return Err(err),
                }
            } else {
                CHECK_PART_SUCCESS
            };
            if let Some(vec) = data_errs_by_part.get_mut(&0)
                && index < vec.len()
            {
                vec[index] = part_status;
            }
            continue;
        }

        // Verify file or check parts
        let mut verify_resp = CheckPartsResp::default();
        let mut verify_err = None;
        meta.data_dir = latest_meta.data_dir;

        if scan_mode == HealScanMode::Deep {
            // disk has a valid xl.meta but may not have all the
            // parts. This is considered an outdated disk, since
            // it needs healing too.
            match disk.verify_file(bucket, object, meta).await {
                Ok(v) => {
                    validate_deep_scan_results(&v.results, latest_meta.parts.len())?;
                    verify_resp = v;
                }
                Err(err) => {
                    if !matches!(err, DiskError::FileNotFound | DiskError::FileVersionNotFound | DiskError::FileCorrupt) {
                        return Err(err);
                    }
                    debug!(
                        event = EVENT_SET_DISK_HEAL,
                        component = LOG_COMPONENT_ECSTORE,
                        subsystem = LOG_SUBSYSTEM_SET_DISK,
                        bucket,
                        object = %object_name,
                        disk_index = index,
                        state = "verify_failed",
                        error = ?err,
                        "Set disk verify_file failed"
                    );
                    verify_err = Some(err);
                }
            }
        } else {
            match disk.check_parts(bucket, object, meta).await {
                Ok(v) => {
                    verify_resp = v;
                }
                Err(err) => {
                    debug!(
                        event = EVENT_SET_DISK_HEAL,
                        component = LOG_COMPONENT_ECSTORE,
                        subsystem = LOG_SUBSYSTEM_SET_DISK,
                        bucket,
                        object = %object_name,
                        disk_index = index,
                        state = "check_parts_failed",
                        error = ?err,
                        "Set disk check_parts failed"
                    );
                    verify_err = Some(err);
                }
            }
        }

        // Update dataErrsByPart for all parts
        for p in 0..latest_meta.parts.len() {
            if let Some(vec) = data_errs_by_part.get_mut(&p)
                && index < vec.len()
            {
                if verify_err.is_some() {
                    vec[index] = conv_part_err_to_int(&verify_err.clone());
                } else {
                    // Fix: verify_resp.results length is based on meta.parts, not latest_meta.parts
                    // We need to check bounds to avoid panic
                    if p < verify_resp.results.len() {
                        vec[index] = verify_resp.results[p];
                    } else {
                        vec[index] = CHECK_PART_SUCCESS;
                    }
                }
            }
        }
    }

    populate_data_errs_by_disk(&mut data_errs_by_disk, &data_errs_by_part);

    Ok((data_errs_by_disk, data_errs_by_part))
}

fn populate_data_errs_by_disk(
    data_errs_by_disk: &mut HashMap<usize, Vec<usize>>,
    data_errs_by_part: &HashMap<usize, Vec<usize>>,
) {
    for (part_index, part_errs) in data_errs_by_part {
        for (disk_index, part_err) in part_errs.iter().enumerate() {
            if let Some(disk_errs) = data_errs_by_disk.get_mut(&disk_index)
                && *part_index < disk_errs.len()
            {
                disk_errs[*part_index] = *part_err;
            }
        }
    }
}

pub fn should_heal_object_on_disk(
    err: &Option<DiskError>,
    parts_errs: &[usize],
    meta: &FileInfo,
    latest_meta: &FileInfo,
) -> (bool, bool, Option<DiskError>) {
    if let Some(err) = err
        && (err == &DiskError::FileNotFound
            || err == &DiskError::FileVersionNotFound
            || err == &DiskError::FileCorrupt
            || err == &DiskError::VolumeNotFound)
    {
        return (true, true, Some(err.clone()));
    }

    if err.is_some() {
        return (false, false, err.clone());
    }

    // `FileInfo::equals` intentionally compares the erasure payload shape and
    // modification time, but it does not compare the selected version or the
    // delete-marker bit. Heal uses this decision for versioned metadata, so a
    // stale historical version that is still marked latest must be treated as
    // outdated even when those storage-level fields happen to match.
    if !meta.equals(latest_meta) || !heal_metadata_identity_matches(meta, latest_meta) {
        debug!(
            event = EVENT_SET_DISK_HEAL,
            component = LOG_COMPONENT_ECSTORE,
            subsystem = LOG_SUBSYSTEM_SET_DISK,
            object = %meta.name,
            state = "metadata_outdated",
            "Set disk object metadata is outdated"
        );
        return (true, true, Some(DiskError::OutdatedXLMeta));
    }

    if !meta.is_canonical_delete_marker() && !meta.is_remote() {
        if parts_errs.contains(&CHECK_PART_FILE_CORRUPT) {
            return (true, false, Some(DiskError::FileCorrupt));
        }
        if parts_errs.contains(&CHECK_PART_FILE_NOT_FOUND) {
            return (true, false, Some(DiskError::PartMissingOrCorrupt));
        }
    }
    (false, false, None)
}

fn heal_metadata_identity_matches(meta: &FileInfo, latest_meta: &FileInfo) -> bool {
    meta.deleted == latest_meta.deleted
        && meta.version_id.filter(|version| !version.is_nil()) == latest_meta.version_id.filter(|version| !version.is_nil())
}

/// Probe every drive of the set at once. Each live probe is bounded by the
/// drive `disk_info` timeout, and the admin peer probe budget only covers one
/// such timeout; a sequential walk over several stalled drives after a power
/// cut would exceed it and make healthy peers render as unknown (#6488).
async fn get_disks_info(disks: &[Option<DiskStore>], eps: &[Endpoint]) -> Vec<rustfs_madmin::Disk> {
    join_all(disks.iter().zip(eps).map(|(disk, ep)| disk_admin_info(disk.as_ref(), ep))).await
}

async fn disk_admin_info(disk: Option<&DiskStore>, ep: &Endpoint) -> rustfs_madmin::Disk {
    let Some(disk) = disk else {
        return rustfs_madmin::Disk {
            endpoint: ep.to_string(),
            drive_path: ep.get_file_path(),
            local: ep.is_local,
            pool_index: ep.pool_idx,
            set_index: ep.set_idx,
            disk_index: ep.disk_idx,
            runtime_state: None,
            offline_duration_seconds: None,
            state: DiskError::DiskNotFound.to_string(),
            capacity_observation_source: Some("missing".to_owned()),
            capacity_observation_age_seconds: Some(0),
            ..Default::default()
        };
    };

    let runtime_state = disk.runtime_state();
    let offline_duration_seconds = disk.offline_duration_secs();
    let capacity_snapshot = disk.last_capacity_snapshot();
    let cached_disk_id = disk.cached_disk_id().await;
    if !(runtime_state.should_probe_for_admin() || runtime_state == disk::health_state::RuntimeDriveHealthState::Suspect) {
        let mut disk_info = build_runtime_snapshot_disk(ep, runtime_state, offline_duration_seconds, capacity_snapshot);
        disk_info.metrics = disk.metrics_snapshot();
        disk_info.uuid = cached_disk_id.map_or_else(String::new, |id| id.to_string());
        return disk_info;
    }

    match disk
        .disk_info(&DiskInfoOptions {
            metrics: true,
            ..Default::default()
        })
        .await
    {
        Ok(res) => {
            disk.record_capacity_probe(res.total, res.used, res.free);
            rustfs_madmin::Disk {
                endpoint: ep.to_string(),
                local: ep.is_local,
                pool_index: ep.pool_idx,
                set_index: ep.set_idx,
                disk_index: ep.disk_idx,
                state: "ok".to_owned(),

                root_disk: res.root_disk,
                drive_path: res.mount_path.clone(),
                healing: res.healing,
                scanning: res.scanning,
                runtime_state: Some(runtime_state.as_str().to_string()),
                offline_duration_seconds,
                capacity_observation_source: Some("live_probe".to_owned()),
                capacity_observation_age_seconds: Some(0),

                uuid: res.id.map_or_else(|| "".to_string(), |id| id.to_string()),
                major: res.major as u32,
                minor: res.minor as u32,
                model: None,
                total_space: res.total,
                used_space: res.used,
                available_space: res.free,
                physical_device_ids: (!res.physical_device_ids.is_empty()).then_some(res.physical_device_ids.clone()),
                utilization: utilization_percent(res.total, res.used),
                used_inodes: res.used_inodes,
                free_inodes: res.free_inodes,
                metrics: Some(res.metrics),
                ..Default::default()
            }
        }
        Err(err) => {
            let mut disk_info = rustfs_madmin::Disk {
                state: err.to_string(),
                endpoint: ep.to_string(),
                drive_path: ep.get_file_path(),
                local: ep.is_local,
                pool_index: ep.pool_idx,
                set_index: ep.set_idx,
                disk_index: ep.disk_idx,
                runtime_state: Some(runtime_state.as_str().to_string()),
                offline_duration_seconds,
                metrics: disk.metrics_snapshot(),
                uuid: cached_disk_id.map_or_else(String::new, |id| id.to_string()),
                ..Default::default()
            };
            if let Some((total, used, free, _)) = capacity_snapshot {
                disk_info.total_space = total;
                disk_info.used_space = used;
                disk_info.available_space = free;
                disk_info.utilization = utilization_percent(total, used);
                disk_info.capacity_observation_source = Some("snapshot".to_owned());
                disk_info.capacity_observation_age_seconds =
                    capacity_snapshot.map(|(_, _, _, probe_unix_secs)| capacity_snapshot_age_seconds(probe_unix_secs));
            } else {
                disk_info.capacity_observation_source = Some("missing".to_owned());
                disk_info.capacity_observation_age_seconds = Some(0);
            }
            disk_info
        }
    }
}

fn build_runtime_snapshot_disk(
    endpoint: &Endpoint,
    runtime_state: disk::health_state::RuntimeDriveHealthState,
    offline_duration_seconds: Option<u64>,
    capacity_snapshot: Option<(u64, u64, u64, u64)>,
) -> rustfs_madmin::Disk {
    let mut disk = rustfs_madmin::Disk {
        endpoint: endpoint.to_string(),
        drive_path: endpoint.get_file_path(),
        local: endpoint.is_local,
        pool_index: endpoint.pool_idx,
        set_index: endpoint.set_idx,
        disk_index: endpoint.disk_idx,
        state: runtime_state.as_str().to_string(),
        runtime_state: Some(runtime_state.as_str().to_string()),
        offline_duration_seconds,
        ..Default::default()
    };

    if let Some((total, used, free, _)) = capacity_snapshot {
        disk.total_space = total;
        disk.used_space = used;
        disk.available_space = free;
        disk.utilization = utilization_percent(total, used);
        disk.capacity_observation_source = Some("snapshot".to_owned());
        disk.capacity_observation_age_seconds =
            capacity_snapshot.map(|(_, _, _, probe_unix_secs)| capacity_snapshot_age_seconds(probe_unix_secs));
    } else {
        disk.capacity_observation_source = Some("missing".to_owned());
        disk.capacity_observation_age_seconds = Some(0);
    }

    disk
}

fn utilization_percent(total: u64, used: u64) -> f64 {
    if total > 0 {
        used as f64 / total as f64 * 100_f64
    } else {
        0_f64
    }
}

fn capacity_snapshot_age_seconds(probe_unix_secs: u64) -> u64 {
    let now_unix_secs = SystemTime::now()
        .duration_since(UNIX_EPOCH)
        .map(|dur| dur.as_secs())
        .unwrap_or(probe_unix_secs);
    now_unix_secs.saturating_sub(probe_unix_secs)
}
async fn get_storage_info(disks: &[Option<DiskStore>], eps: &[Endpoint]) -> rustfs_madmin::StorageInfo {
    // let mut disks = get_disks_info(disks, eps).await;
    // disks.sort_by(|a, b| a.total_space.cmp(&b.total_space));
    //
    // rustfs_madmin::StorageInfo {
    //     disks,
    //     backend: rustfs_madmin::BackendInfo {
    //         backend_type: rustfs_madmin::BackendByte::Erasure,
    //         ..Default::default()
    //     },
    // }
    let mut disks = get_disks_info(disks, eps).await;
    disks.sort_by_key(|a| a.total_space);

    // Provide minimal backend shape for callers. Do NOT guess parity here since it belongs to higher-level config.
    // Missing/empty standard_sc_data will be handled by capacity fallback logic.
    let drives_per_set = vec![eps.len()];
    let total_sets = vec![1];

    rustfs_madmin::StorageInfo {
        disks,
        backend: rustfs_madmin::BackendInfo {
            backend_type: rustfs_madmin::BackendByte::Erasure,
            drives_per_set,
            total_sets,
            ..Default::default()
        },
        ..Default::default()
    }
}
pub async fn stat_all_dirs(disks: &[Option<DiskStore>], bucket: &str, prefix: &str) -> Vec<Option<DiskError>> {
    let mut futures = Vec::with_capacity(disks.len());
    // Spawn one future per disk slot so the returned vector stays index-aligned with `disks`
    // (and therefore with `set_endpoints`). Offline/None disks must yield DiskNotFound in-place
    // rather than being skipped, otherwise callers that zip `errs` against the full disks array
    // (heal_object_dir) would pair every error with the wrong disk/endpoint whenever any disk is
    // offline — and could `make_volume` on the wrong disk.
    for disk in disks.iter() {
        let disk = disk.clone();
        let bucket = bucket.to_string();
        let prefix = prefix.to_string();
        futures.push(tokio::spawn(async move {
            let Some(disk) = disk else {
                return Some(DiskError::DiskNotFound);
            };
            match disk.list_dir("", &bucket, &prefix, 1).await {
                Ok(entries) => {
                    if !entries.is_empty() {
                        return Some(DiskError::VolumeNotEmpty);
                    }
                    None
                }
                Err(err) => Some(err),
            }
        }));
    }

    let results = join_all(futures).await;

    // Preserve length/index alignment: a panicked probe becomes a corrupt-state error instead of
    // a silently-dropped slot that would re-shift every subsequent index.
    let mut errs = Vec::with_capacity(disks.len());
    for res in results.into_iter() {
        match res {
            Ok(err) => errs.push(err),
            Err(join_err) => errs.push(Some(DiskError::other(join_err.to_string()))),
        }
    }
    errs
}

const GLOBAL_MIN_PART_SIZE: ByteSize = ByteSize::mib(5);
fn is_min_allowed_part_size(size: i64) -> bool {
    size >= GLOBAL_MIN_PART_SIZE.as_u64() as i64
}

fn get_complete_multipart_md5(parts: &[CompletePart]) -> String {
    let mut buf = Vec::new();

    for part in parts.iter() {
        if let Some(etag) = &part.etag {
            if let Ok(etag_bytes) = hex_simd::decode_to_vec(etag.as_bytes()) {
                buf.extend(etag_bytes);
            } else {
                buf.extend(etag.bytes());
            }
        }
    }

    let mut hasher = Md5::new();
    hasher.update(&buf);

    let digest = hasher.finalize();
    let etag_hex = faster_hex::hex_string(digest.as_slice());
    format!("{}-{}", etag_hex, parts.len())
}

fn completed_multipart_object_part(part_num: usize, ext_part: &ObjectPartInfo) -> ObjectPartInfo {
    ObjectPartInfo {
        etag: ext_part.etag.clone(),
        number: part_num,
        size: ext_part.size,
        mod_time: ext_part.mod_time,
        actual_size: ext_part.actual_size,
        index: ext_part.index.clone(),
        checksums: ext_part.checksums.clone(),
        integrity: ext_part.integrity.clone(),
        ..Default::default()
    }
}

fn complete_part_checksum(part: &CompletePart, checksum_type: rustfs_rio::ChecksumType) -> Option<Option<String>> {
    match checksum_type.base() {
        rustfs_rio::ChecksumType::SHA256 => Some(part.checksum_sha256.clone()),
        rustfs_rio::ChecksumType::SHA1 => Some(part.checksum_sha1.clone()),
        rustfs_rio::ChecksumType::CRC32 => Some(part.checksum_crc32.clone()),
        rustfs_rio::ChecksumType::CRC32C => Some(part.checksum_crc32c.clone()),
        rustfs_rio::ChecksumType::CRC64_NVME => Some(part.checksum_crc64nvme.clone()),
        // XXHash3/64/128 and SHA-512 (AWS 2026-04): s3s CompletePart has no typed
        // field to carry a client-supplied per-part value in the
        // CompleteMultipartUpload request, so accept the type with no double-check
        // value (the part was already verified server-side at UploadPart). This
        // mirrors the missing-value path of the five typed algorithms. Reject only
        // genuinely unset/invalid types.
        base if base.is_set() => Some(None),
        _ => None,
    }
}

fn parts_after_marker(part_numbers: &[usize], part_number_marker: usize) -> Option<&[usize]> {
    if part_number_marker == 0 {
        return Some(part_numbers);
    }

    // reduce_quorum_part_numbers returns sorted numbers; the marker need not exist.
    part_numbers.last().filter(|&&last| part_number_marker <= last)?;
    let index = part_numbers.partition_point(|&part_number| part_number <= part_number_marker);
    Some(&part_numbers[index..])
}

pub fn canonicalize_etag(etag: &str) -> String {
    let re = Regex::new("\"*?([^\"]*?)\"*?$").unwrap();
    re.replace_all(etag, "$1").to_string()
}

pub fn e_tag_matches(etag: &str, condition: &str) -> bool {
    if condition.trim() == "*" {
        return true;
    }
    canonicalize_etag(etag) == canonicalize_etag(condition)
}

pub fn should_prevent_write(oi: &ObjectInfo, if_none_match: Option<String>, if_match: Option<String>) -> bool {
    let if_none_match = if_none_match
        .as_deref()
        .map(str::trim)
        .filter(|condition| !condition.is_empty());
    let if_match = if_match.as_deref().map(str::trim).filter(|condition| !condition.is_empty());

    match &oi.etag {
        Some(etag) => {
            if let Some(if_none_match) = if_none_match
                && e_tag_matches(etag, if_none_match)
            {
                return true;
            }
            if let Some(if_match) = if_match
                && !e_tag_matches(etag, if_match)
            {
                return true;
            }
            false
        }
        // If we can't obtain the etag of the object, perevent the write only when we have at least one condition
        None => if_none_match.is_some() || if_match.is_some(),
    }
}

/// Validates if the given storage class is supported
pub fn is_valid_storage_class(storage_class: &str) -> bool {
    storageclass::is_supported_write_class(storage_class)
}

/// Returns true if the storage class is a cold storage tier that requires special handling
#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
pub fn is_cold_storage_class(storage_class: &str) -> bool {
    matches!(
        storage_class,
        storageclass::DEEP_ARCHIVE | storageclass::GLACIER | storageclass::GLACIER_IR
    )
}

/// Returns true if the storage class is an infrequent access tier
#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
pub fn is_infrequent_access_class(storage_class: &str) -> bool {
    matches!(
        storage_class,
        storageclass::ONEZONE_IA | storageclass::STANDARD_IA | storageclass::INTELLIGENT_TIERING
    )
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::bucket::replication::{replication_statuses_map, version_purge_statuses_map};
    use crate::cluster::rpc::{RemoteDisk, TcpHttpInternodeDataTransport};
    use crate::disk::CHECK_PART_UNKNOWN;
    use crate::disk::CHECK_PART_VOLUME_NOT_FOUND;
    use crate::disk::DataDirDeleteStatus;
    use crate::disk::DiskOption;
    use crate::disk::RUSTFS_META_BUCKET;
    use crate::disk::RUSTFS_META_TMP_BUCKET;
    use crate::disk::STORAGE_FORMAT_FILE;
    use crate::disk::STORAGE_FORMAT_FILE_BACKUP;
    use crate::disk::WalkDirOptions;
    use crate::disk::endpoint::Endpoint;
    use crate::disk::error::DiskError;
    use crate::disk::health_state::RuntimeDriveHealthState;
    use crate::disk::new_disk;
    use crate::layout::endpoints::SetupType;
    use crate::object_api::BLOCK_SIZE_V2;
    use crate::object_api::ObjectInfo;
    use crate::set_disk::core::io_primitives::rename_fanout_barrier;
    use crate::set_disk::ops::object::{PutObjectCommitBarrier, PutObjectCommitPause};
    use crate::storage_api_contracts::{
        heal::HealOperations as _, lifecycle::TransitionedObject, list::ListOperations as _, multipart::CompletePart,
        object::ObjectOperations as _,
    };
    use crate::store::init_format::save_format_file;
    use crate::store::list_objects::ListPathOptions;
    use metrics_util::debugging::{DebugValue, DebuggingRecorder};
    use rustfs_filemeta::ErasureInfo;
    use rustfs_filemeta::FileMeta;
    use rustfs_filemeta::MetaCacheEntry;
    use rustfs_lock::client::local::LocalClient;
    use rustfs_lock::{LockError, LockInfo, LockResponse, LockStats};
    use serial_test::serial;
    use std::collections::HashMap;
    use tempfile::TempDir;
    use time::OffsetDateTime;
    use tokio::fs;
    use tokio::io::AsyncReadExt;

    #[tokio::test]
    async fn copy_source_read_policy_is_scoped_and_demand_bound() {
        assert_eq!(get_object_read_policy(), GetObjectReadPolicy::Default);
        assert!(GetObjectReadPolicy::Default.allows_multipart_setup_prefetch());
        assert!(!GetObjectReadPolicy::CopySource.allows_multipart_setup_prefetch());

        with_get_object_read_policy(GetObjectReadPolicy::CopySource, async {
            assert_eq!(get_object_read_policy(), GetObjectReadPolicy::CopySource);
            assert!(!get_object_read_policy().allows_multipart_setup_prefetch());
            assert_eq!(
                crate::erasure::coding::decode::decode_read_policy(),
                crate::erasure::coding::decode::DecodeReadPolicy::DemandBound
            );
        })
        .await;

        assert_eq!(get_object_read_policy(), GetObjectReadPolicy::Default);
        assert_eq!(
            crate::erasure::coding::decode::decode_read_policy(),
            crate::erasure::coding::decode::DecodeReadPolicy::Default
        );
    }

    #[test]
    fn complete_part_error_maps_confirmed_missing_to_invalid_part() {
        for err in ["file not found", "Specified part could not be found", "part.7 not found"] {
            let mapped = complete_multipart_part_error(7, err, "bucket", "object");

            assert!(matches!(
                mapped,
                Error::InvalidPart(7, ref bucket, ref object) if bucket == "bucket" && object == "object"
            ));
            assert_eq!(complete_multipart_part_error_result(&mapped), COMPLETE_MULTIPART_PART_MISSING);
        }
    }

    #[test]
    fn complete_part_error_maps_read_quorum_to_retryable_server_error() {
        let mapped = complete_multipart_part_error(1, "erasure read quorum", "bucket", "object");

        assert!(matches!(
            mapped,
            Error::InsufficientReadQuorum(ref bucket, ref object) if bucket == "bucket" && object == "object"
        ));
        assert_eq!(
            complete_multipart_part_error_result(&mapped),
            COMPLETE_MULTIPART_PART_READ_QUORUM_UNAVAILABLE
        );
    }

    #[test]
    fn complete_part_error_maps_unknown_part_error_to_retryable_server_error() {
        let mapped = complete_multipart_part_error(1, "metadata decode failed", "bucket", "object");

        assert!(matches!(mapped, Error::InsufficientReadQuorum(_, _)));
        assert_ne!(complete_multipart_part_error_result(&mapped), COMPLETE_MULTIPART_PART_MISSING);
    }

    #[derive(Debug, Default)]
    struct FailingClient;

    #[async_trait::async_trait]
    impl LockClient for FailingClient {
        async fn acquire_lock(&self, _request: &rustfs_lock::LockRequest) -> rustfs_lock::Result<LockResponse> {
            Err(LockError::internal("simulated offline client"))
        }

        async fn release(&self, _lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<bool> {
            Ok(false)
        }

        async fn refresh(&self, _lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<bool> {
            Ok(false)
        }

        async fn force_release(&self, _lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<bool> {
            Ok(false)
        }

        async fn check_status(&self, _lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<Option<LockInfo>> {
            Ok(None)
        }

        async fn get_stats(&self) -> rustfs_lock::Result<LockStats> {
            Ok(LockStats::default())
        }

        async fn close(&self) -> rustfs_lock::Result<()> {
            Ok(())
        }

        async fn is_online(&self) -> bool {
            false
        }

        async fn is_local(&self) -> bool {
            false
        }
    }

    #[derive(Debug)]
    struct DelayedBatchClient {
        inner: Arc<dyn LockClient>,
        delay: Duration,
    }

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

        async fn acquire_locks_batch(&self, requests: &[rustfs_lock::LockRequest]) -> rustfs_lock::Result<Vec<LockResponse>> {
            tokio::time::sleep(self.delay).await;
            self.inner.acquire_locks_batch(requests).await
        }

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

        async fn release_locks_batch(&self, lock_ids: &[rustfs_lock::LockId]) -> rustfs_lock::Result<Vec<bool>> {
            self.inner.release_locks_batch(lock_ids).await
        }

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

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

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

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

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

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

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

    async fn make_test_set_disks(lockers: Vec<Arc<dyn LockClient>>) -> Arc<SetDisks> {
        make_test_set_disks_with_ctx(lockers, bootstrap_ctx()).await
    }

    #[tokio::test]
    #[serial_test::serial]
    async fn system_metadata_incarnation_lookup_does_not_reenter_pool_metadata() {
        let (_temp_dirs, store, _other_store) =
            crate::services::rebalance::test_three_pool_stores_with_isolated_node_contexts(None).await;
        let _pool_meta_guard = store.pool_meta.write().await;
        let set = &store.pools[0].disk_set[0];

        for bucket in [RUSTFS_META_BUCKET, RUSTFS_META_TMP_BUCKET, crate::disk::MIGRATING_META_BUCKET] {
            tokio::time::timeout(Duration::from_secs(30), set.bucket_incarnation_id_from_disk(bucket))
                .await
                .expect("system metadata identity lookup must not reacquire the held pool metadata lock")
                .expect_err("system metadata buckets have no user bucket incarnation");
        }

        drop(_pool_meta_guard);
        let bucket = "user-incarnation-boundary";
        let incarnation = Uuid::new_v4();
        crate::bucket::metadata::save_bucket_incarnation(Arc::clone(&store), bucket, incarnation)
            .await
            .expect("persist the user bucket identity through the metadata owner");
        assert_eq!(
            set.bucket_incarnation_id_from_disk(bucket)
                .await
                .expect("user bucket identities must still load from the metadata owner"),
            incarnation,
        );
    }

    async fn make_test_set_disks_with_ctx(
        lockers: Vec<Arc<dyn LockClient>>,
        instance_ctx: Arc<InstanceContext>,
    ) -> Arc<SetDisks> {
        let endpoints = vec![
            Endpoint::try_from("http://127.0.0.1:9000/data").expect("first endpoint should parse"),
            Endpoint::try_from("http://127.0.0.1:9001/data").expect("second endpoint should parse"),
        ];

        SetDisks::new_with_instance_ctx(
            "test-owner".to_string(),
            Arc::new(RwLock::new(vec![None, None])),
            2,
            1,
            0,
            0,
            endpoints,
            FormatV3::new(1, 2),
            lockers,
            instance_ctx,
        )
        .await
    }

    /// Pins the dist-erasure resolution SOURCE for `new_ns_lock` (adversarial
    /// review): the lock strategy must come from the set's own instance
    /// context, never the ambient facade — otherwise another in-process
    /// instance (or a concurrent test's ambient DistErasure window) reroutes
    /// this set's locking onto the wrong strategy.
    #[tokio::test(flavor = "multi_thread")]
    async fn new_ns_lock_resolves_dist_from_set_instance_context() {
        let manager = Arc::new(rustfs_lock::GlobalLockManager::new());
        let locker: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager));

        let dist_ctx = Arc::new(InstanceContext::new());
        dist_ctx.update_erasure_type(SetupType::DistErasure).await;
        let dist_set = make_test_set_disks_with_ctx(vec![locker.clone()], dist_ctx).await;
        let dist_guard = dist_set
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_read_lock(Duration::from_millis(500))
            .await
            .expect("dist read lock should succeed with one healthy locker");
        assert!(
            matches!(dist_guard, NamespaceLockGuard::Standard(_)),
            "a DistErasure instance context must select the distributed lock strategy"
        );
        drop(dist_guard);

        let local_ctx = Arc::new(InstanceContext::new());
        local_ctx.update_erasure_type(SetupType::Erasure).await;
        let local_set = make_test_set_disks_with_ctx(vec![locker], local_ctx).await;
        let local_guard = local_set
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_read_lock(Duration::from_millis(500))
            .await
            .expect("local read lock should succeed");
        assert!(
            matches!(local_guard, NamespaceLockGuard::Fast(_)),
            "a plain-erasure instance context must select the local lock strategy"
        );
    }

    #[tokio::test]
    async fn new_ns_lock_reuses_the_set_namespace_allocation() {
        let ctx = Arc::new(InstanceContext::new());
        ctx.update_erasure_type(SetupType::Erasure).await;
        let set = make_test_set_disks_with_ctx(Vec::new(), ctx).await;

        assert_eq!(&*set.set_lock_namespace, "pool-0");
        let before = Arc::strong_count(&set.set_lock_namespace);
        let lock = set
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created");

        assert_eq!(
            Arc::strong_count(&set.set_lock_namespace),
            before + 1,
            "each lock should share the set namespace instead of formatting a new String"
        );
        drop(lock);
        assert_eq!(Arc::strong_count(&set.set_lock_namespace), before);
    }

    fn put_object_commit_namespace_lock_wait_sample_count(snapshotter: &metrics_util::debugging::Snapshotter) -> usize {
        snapshotter
            .snapshot()
            .into_vec()
            .into_iter()
            .filter(|(composite, _, _, _)| {
                composite.key().name() == "rustfs_s3_put_object_stage_duration_ms"
                    && composite.key().labels().any(|label| {
                        label.key() == "stage"
                            && label.value() == rustfs_io_metrics::PUT_STAGE_PUT_OBJECT_COMMIT_NAMESPACE_LOCK_WAIT
                    })
            })
            .map(|(_, _, _, value)| match value {
                DebugValue::Histogram(samples) => samples.len(),
                _ => 0,
            })
            .sum()
    }

    fn put_object_commit_lock_admission_count(
        rows: &[(
            metrics_util::CompositeKey,
            Option<metrics::Unit>,
            Option<metrics::SharedString>,
            DebugValue,
        )],
        budget: &'static str,
        outcome: &'static str,
    ) -> u64 {
        rows.iter()
            .filter(|(composite, _, _, _)| {
                composite.key().name() == "rustfs_s3_put_object_commit_namespace_lock_admission_total"
                    && composite
                        .key()
                        .labels()
                        .any(|label| label.key() == "budget" && label.value() == budget)
                    && composite
                        .key()
                        .labels()
                        .any(|label| label.key() == "outcome" && label.value() == outcome)
            })
            .map(|(_, _, _, value)| match value {
                DebugValue::Counter(count) => *count,
                _ => 0,
            })
            .sum()
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_admission_budget_labels_are_bounded() {
        let cases = [
            ("0", rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_DISABLED),
            ("250", rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_250MS),
            ("251", rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_500MS),
            ("500", rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_500MS),
            ("501", rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_1000MS),
            ("1000", rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_1000MS),
            ("1001", rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_GT_1000MS),
        ];
        for (timeout_ms, expected) in cases {
            temp_env::with_vars(
                [(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some(timeout_ms))],
                || {
                    assert_eq!(put_object_commit_lock_admission_budget_label(), expected);
                },
            );
        }
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_admission_error_outcomes_are_bounded() {
        let timeout = LockError::timeout("bucket/object", Duration::from_millis(1));
        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("1"))], || {
            assert_eq!(
                put_object_commit_lock_acquire_error_outcome("put_object_commit", &timeout),
                rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_TIMEOUT_SLOWDOWN
            );
            assert_eq!(
                put_object_commit_lock_acquire_error_outcome("complete_multipart_upload_commit", &timeout),
                rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_LOCK_ERROR
            );
        });

        let internal = LockError::internal("simulated lock manager error");
        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("1"))], || {
            assert_eq!(
                put_object_commit_lock_acquire_error_outcome("put_object_commit", &internal),
                rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_LOCK_ERROR
            );
        });
    }

    #[test]
    #[serial]
    fn put_object_commit_namespace_lock_wait_metric_is_wired_to_both_write_lock_paths() {
        let recorder = DebuggingRecorder::new();
        let snapshotter = recorder.snapshotter();
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("test runtime should start");

        metrics::with_local_recorder(&recorder, || {
            runtime.block_on(async {
                let ctx = Arc::new(InstanceContext::new());
                ctx.update_erasure_type(SetupType::Erasure).await;
                let set = make_test_set_disks_with_ctx(Vec::new(), ctx).await;
                let bucket = "bucket";
                let object = "object";

                rustfs_io_metrics::set_put_stage_metrics_enabled(false);
                let guard = set
                    .acquire_write_lock_diag("put_object_commit", bucket, object)
                    .await
                    .expect("disabled metrics acquire should succeed");
                drop(guard);
                assert_eq!(put_object_commit_namespace_lock_wait_sample_count(&snapshotter), 0);

                rustfs_io_metrics::set_put_stage_metrics_enabled(true);
                let guard = set
                    .acquire_write_lock_diag("put_object_commit", bucket, object)
                    .await
                    .expect("normal PUT commit acquire should succeed");
                drop(guard);
                assert_eq!(put_object_commit_namespace_lock_wait_sample_count(&snapshotter), 1);

                let guard = set
                    .acquire_write_lock_diag("complete_multipart_upload_commit", bucket, object)
                    .await
                    .expect("non-PUT commit acquire should succeed");
                drop(guard);
                assert_eq!(put_object_commit_namespace_lock_wait_sample_count(&snapshotter), 0);

                let held_guard = set
                    .acquire_write_lock_diag("put_object_commit", bucket, object)
                    .await
                    .expect("holder acquire should succeed");
                assert_eq!(put_object_commit_namespace_lock_wait_sample_count(&snapshotter), 1);

                let (pending_tx, pending_rx) = tokio::sync::oneshot::channel();
                let pending_acquire =
                    set.acquire_write_lock_diag_with_pending_hook("put_object_commit", bucket, object, move || {
                        let _ = pending_tx.send(());
                    });
                let release_holder = async {
                    pending_rx.await.expect("pending hook should fire");
                    drop(held_guard);
                };
                let (pending_guard, ()) = tokio::join!(pending_acquire, release_holder);
                drop(pending_guard.expect("pending-hook PUT commit acquire should succeed"));
                assert_eq!(put_object_commit_namespace_lock_wait_sample_count(&snapshotter), 1);

                rustfs_io_metrics::set_put_stage_metrics_enabled(false);
            });
        });
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_timeout_override_only_applies_to_put_commit() {
        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("17"))], || {
            assert_eq!(get_put_object_commit_lock_acquire_timeout("put_object_commit"), Duration::from_millis(17));
            assert_eq!(
                get_put_object_commit_lock_acquire_timeout("complete_multipart_upload_commit"),
                get_lock_acquire_timeout()
            );
        });

        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("0"))], || {
            assert_eq!(
                get_put_object_commit_lock_acquire_timeout("put_object_commit"),
                get_lock_acquire_timeout()
            );
        });
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_timeout_override_bounds_contention_wait() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("test runtime should start");

        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("1"))], || {
            runtime.block_on(async {
                let ctx = Arc::new(InstanceContext::new());
                ctx.update_erasure_type(SetupType::Erasure).await;
                let set = make_test_set_disks_with_ctx(Vec::new(), ctx).await;
                let bucket = "bucket";
                let object = "object";

                let held_guard = set
                    .acquire_write_lock_diag("put_object_commit", bucket, object)
                    .await
                    .expect("holder acquire should succeed");
                let started = Instant::now();
                let err = match set.acquire_write_lock_diag("put_object_commit", bucket, object).await {
                    Ok(_) => panic!("contended PUT commit lock should honor the short timeout"),
                    Err(err) => err,
                };
                assert!(
                    started.elapsed() < Duration::from_secs(1),
                    "short PUT commit lock timeout should not wait for the global timeout"
                );
                assert!(matches!(err, StorageError::SlowDown));

                drop(held_guard);
                set.acquire_write_lock_diag("put_object_commit", bucket, object)
                    .await
                    .expect("permit should not leak after timeout");
            });
        });
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_admission_records_acquired_and_timeout() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("test runtime should start");

        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("1"))], || {
            let recorder = DebuggingRecorder::new();
            let snapshotter = recorder.snapshotter();
            metrics::with_local_recorder(&recorder, || {
                rustfs_io_metrics::set_put_stage_metrics_enabled(true);
                runtime.block_on(async {
                    let ctx = Arc::new(InstanceContext::new());
                    ctx.update_erasure_type(SetupType::Erasure).await;
                    let set = make_test_set_disks_with_ctx(Vec::new(), ctx).await;
                    let held_guard = set
                        .acquire_write_lock_diag("put_object_commit", "bucket", "object")
                        .await
                        .expect("holder acquire should succeed");
                    let err = match set.acquire_write_lock_diag("put_object_commit", "bucket", "object").await {
                        Ok(_) => panic!("contended PUT commit acquire should return SlowDown"),
                        Err(err) => err,
                    };
                    assert!(matches!(err, StorageError::SlowDown));
                    drop(held_guard);
                    rustfs_io_metrics::set_put_stage_metrics_enabled(false);
                });
            });

            let rows = snapshotter.snapshot().into_vec();
            assert_eq!(
                put_object_commit_lock_admission_count(
                    &rows,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_250MS,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_ACQUIRED,
                ),
                1
            );
            assert_eq!(
                put_object_commit_lock_admission_count(
                    &rows,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_250MS,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_TIMEOUT_SLOWDOWN,
                ),
                1
            );
        });
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_admission_records_disabled_budget_acquired() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("test runtime should start");

        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("0"))], || {
            let recorder = DebuggingRecorder::new();
            let snapshotter = recorder.snapshotter();
            metrics::with_local_recorder(&recorder, || {
                rustfs_io_metrics::set_put_stage_metrics_enabled(true);
                runtime.block_on(async {
                    let ctx = Arc::new(InstanceContext::new());
                    ctx.update_erasure_type(SetupType::Erasure).await;
                    let set = make_test_set_disks_with_ctx(Vec::new(), ctx).await;
                    let guard = set
                        .acquire_write_lock_diag("put_object_commit", "bucket", "object")
                        .await
                        .expect("PUT commit acquire should succeed with default timeout");
                    drop(guard);
                    rustfs_io_metrics::set_put_stage_metrics_enabled(false);
                });
            });

            let rows = snapshotter.snapshot().into_vec();
            assert_eq!(
                put_object_commit_lock_admission_count(
                    &rows,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_DISABLED,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_ACQUIRED,
                ),
                1
            );
        });
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_admission_skips_non_put_commit_ops() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("test runtime should start");

        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("250"))], || {
            let recorder = DebuggingRecorder::new();
            let snapshotter = recorder.snapshotter();
            metrics::with_local_recorder(&recorder, || {
                rustfs_io_metrics::set_put_stage_metrics_enabled(true);
                runtime.block_on(async {
                    let ctx = Arc::new(InstanceContext::new());
                    ctx.update_erasure_type(SetupType::Erasure).await;
                    let set = make_test_set_disks_with_ctx(Vec::new(), ctx).await;
                    let guard = set
                        .acquire_write_lock_diag("complete_multipart_upload_commit", "bucket", "object")
                        .await
                        .expect("non-PUT commit acquire should succeed");
                    drop(guard);
                    rustfs_io_metrics::set_put_stage_metrics_enabled(false);
                });
            });

            let rows = snapshotter.snapshot().into_vec();
            assert_eq!(
                rows.iter()
                    .filter(|(composite, _, _, _)| {
                        composite.key().name() == "rustfs_s3_put_object_commit_namespace_lock_admission_total"
                    })
                    .count(),
                0
            );
        });
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_admission_records_lock_error() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("test runtime should start");

        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("250"))], || {
            let recorder = DebuggingRecorder::new();
            let snapshotter = recorder.snapshotter();
            metrics::with_local_recorder(&recorder, || {
                rustfs_io_metrics::set_put_stage_metrics_enabled(true);
                runtime.block_on(async {
                    let healthy: Arc<dyn LockClient> =
                        Arc::new(LocalClient::with_manager(Arc::new(rustfs_lock::GlobalLockManager::new())));
                    let failing: Arc<dyn LockClient> = Arc::new(FailingClient);
                    let ctx = Arc::new(InstanceContext::new());
                    ctx.update_erasure_type(SetupType::DistErasure).await;
                    let set = make_test_set_disks_with_ctx(vec![healthy, failing], ctx).await;
                    assert!(
                        set.acquire_write_lock_diag("put_object_commit", "bucket", "object")
                            .await
                            .is_err(),
                        "one healthy locker must not satisfy the PUT commit write quorum"
                    );
                    rustfs_io_metrics::set_put_stage_metrics_enabled(false);
                });
            });

            let rows = snapshotter.snapshot().into_vec();
            assert_eq!(
                put_object_commit_lock_admission_count(
                    &rows,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_250MS,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_LOCK_ERROR,
                ),
                1
            );
            assert_eq!(
                put_object_commit_lock_admission_count(
                    &rows,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_250MS,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_TIMEOUT_SLOWDOWN,
                ),
                0
            );
        });
    }

    #[test]
    #[serial]
    fn put_object_commit_lock_admission_records_pending_hook_acquired() {
        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()
            .expect("test runtime should start");

        temp_env::with_vars([(rustfs_config::ENV_PUT_COMMIT_NAMESPACE_LOCK_ACQUIRE_TIMEOUT_MS, Some("500"))], || {
            let recorder = DebuggingRecorder::new();
            let snapshotter = recorder.snapshotter();
            metrics::with_local_recorder(&recorder, || {
                rustfs_io_metrics::set_put_stage_metrics_enabled(true);
                runtime.block_on(async {
                    let ctx = Arc::new(InstanceContext::new());
                    ctx.update_erasure_type(SetupType::Erasure).await;
                    let set = make_test_set_disks_with_ctx(Vec::new(), ctx).await;
                    let held_guard = set
                        .acquire_write_lock_diag("put_object_commit", "bucket", "object")
                        .await
                        .expect("holder acquire should succeed");
                    let (pending_tx, pending_rx) = tokio::sync::oneshot::channel();
                    let pending_acquire =
                        set.acquire_write_lock_diag_with_pending_hook("put_object_commit", "bucket", "object", move || {
                            let _ = pending_tx.send(());
                        });
                    let release_holder = async {
                        pending_rx.await.expect("pending hook should fire");
                        drop(held_guard);
                    };
                    let (pending_guard, ()) = tokio::join!(pending_acquire, release_holder);
                    drop(pending_guard.expect("pending-hook PUT commit acquire should succeed"));
                    rustfs_io_metrics::set_put_stage_metrics_enabled(false);
                });
            });

            let rows = snapshotter.snapshot().into_vec();
            assert_eq!(
                put_object_commit_lock_admission_count(
                    &rows,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_BUDGET_LE_500MS,
                    rustfs_io_metrics::PUT_COMMIT_LOCK_ADMISSION_OUTCOME_ACQUIRED,
                ),
                2
            );
        });
    }

    #[tokio::test]
    async fn new_ns_lock_shares_clients_without_changing_quorum() {
        let healthy: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(Arc::new(rustfs_lock::GlobalLockManager::new())));
        let failing: Arc<dyn LockClient> = Arc::new(FailingClient);
        let ctx = Arc::new(InstanceContext::new());
        ctx.update_erasure_type(SetupType::DistErasure).await;
        let set = make_test_set_disks_with_ctx(vec![healthy.clone(), failing.clone()], ctx).await;

        assert!(Arc::ptr_eq(&set.lockers[0], &healthy));
        assert!(Arc::ptr_eq(&set.lockers[1], &failing));
        let clients_before = Arc::strong_count(&set.shared_lockers);
        let healthy_before = Arc::strong_count(&healthy);
        let failing_before = Arc::strong_count(&failing);
        let write_lock = set
            .new_ns_lock("bucket", "write-object")
            .await
            .expect("namespace lock should be created");

        assert_eq!(
            Arc::strong_count(&set.shared_lockers),
            clients_before + 1,
            "each object lock should share one client slice allocation"
        );
        assert_eq!(
            Arc::strong_count(&healthy),
            healthy_before,
            "constructing an object lock must not clone each client Arc"
        );
        assert_eq!(
            Arc::strong_count(&failing),
            failing_before,
            "constructing an object lock must not clone each client Arc"
        );

        let write_error = write_lock
            .get_write_lock(Duration::from_millis(500))
            .await
            .expect_err("one healthy client must not satisfy the two-client write quorum");
        assert!(
            matches!(
                write_error,
                LockError::QuorumNotReached {
                    required: 2,
                    achieved: 1
                }
            ),
            "the shared client representation must preserve the exact write quorum result: {write_error}"
        );
        let read_lock = set
            .new_ns_lock("bucket", "read-object")
            .await
            .expect("second namespace lock should be created");
        assert_eq!(Arc::strong_count(&set.shared_lockers), clients_before + 2);
        let read_guard = read_lock
            .get_read_lock(Duration::from_millis(500))
            .await
            .expect("one healthy client should satisfy the two-client read quorum");
        assert!(matches!(read_guard, NamespaceLockGuard::Standard(_)));
    }

    #[tokio::test]
    async fn new_ns_lock_uses_the_current_public_client_domain() {
        let stale_a: Arc<dyn LockClient> = Arc::new(FailingClient);
        let stale_b: Arc<dyn LockClient> = Arc::new(FailingClient);
        let healthy_a: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(Arc::new(rustfs_lock::GlobalLockManager::new())));
        let healthy_b: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(Arc::new(rustfs_lock::GlobalLockManager::new())));
        let ctx = Arc::new(InstanceContext::new());
        ctx.update_erasure_type(SetupType::DistErasure).await;
        let set = make_test_set_disks_with_ctx(vec![stale_a, stale_b], ctx).await;
        let mut set = (*set).clone();
        set.lockers = vec![healthy_a, healthy_b];

        let lock = set
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should use the current public clients");
        let guard = lock
            .get_write_lock(Duration::from_millis(500))
            .await
            .expect("the current healthy clients should satisfy the two-client quorum");
        assert!(matches!(guard, NamespaceLockGuard::Standard(_)));
    }

    struct SetupTypeGuard {
        previous: SetupType,
    }

    impl SetupTypeGuard {
        async fn switch_to(next: SetupType) -> Self {
            let previous = current_setup_type().await;
            runtime_sources::set_setup_type(next).await;
            Self { previous }
        }
    }

    impl Drop for SetupTypeGuard {
        fn drop(&mut self) {
            let previous = self.previous.clone();
            let handle = tokio::runtime::Handle::current();
            tokio::task::block_in_place(|| {
                handle.block_on(async move {
                    runtime_sources::set_setup_type(previous).await;
                });
            });
        }
    }

    async fn current_setup_type() -> SetupType {
        runtime_sources::current_setup_type().await
    }

    async fn make_formatted_local_disk_for_info_test(disk_idx: usize, format: &FormatV3) -> (TempDir, Endpoint, DiskStore) {
        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");

        (dir, endpoint, disk)
    }

    async fn make_remote_disk_for_info_test(disk_idx: usize) -> (Endpoint, DiskStore) {
        let endpoint_url = format!("http://remote-server:9000/data{disk_idx}");
        let mut endpoint = Endpoint::try_from(endpoint_url.as_str()).expect("remote endpoint should parse");
        endpoint.set_pool_index(0);
        endpoint.set_set_index(0);
        endpoint.set_disk_index(disk_idx);
        let remote_disk = RemoteDisk::new(
            &endpoint,
            &DiskOption {
                cleanup: false,
                health_check: false,
            },
            Arc::new(TcpHttpInternodeDataTransport),
        )
        .await
        .expect("remote disk should be created");

        (endpoint, Arc::new(disk::Disk::Remote(Box::new(remote_disk))))
    }

    #[tokio::test]
    async fn test_rename_data_quorum_failure_rolls_back_destination_object() {
        let dir = tempfile::tempdir().expect("tempdir should be created");
        let disk_root = dir.path().join("disk0");
        fs::create_dir_all(&disk_root).await.expect("disk root should be created");
        let endpoint = Endpoint::try_from(disk_root.to_str().expect("disk 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");

        let bucket = "bucket";
        let object = "object";
        let tmp_object = "tmp-object";
        let version_id = Uuid::parse_str("77777777-7777-7777-7777-777777777777").expect("version id should parse");
        let old_data_dir = Uuid::parse_str("88888888-8888-8888-8888-888888888888").expect("old data dir should parse");
        let new_data_dir = Uuid::parse_str("99999999-9999-9999-9999-999999999999").expect("new data dir should parse");

        match disk.make_volume(bucket).await {
            Ok(()) | Err(DiskError::VolumeExists) => {}
            Err(err) => panic!("bucket should be available: {err:?}"),
        }
        match disk.make_volume(RUSTFS_META_TMP_BUCKET).await {
            Ok(()) | Err(DiskError::VolumeExists) => {}
            Err(err) => panic!("tmp bucket should be available: {err:?}"),
        }

        let object_dir = disk_root.join(bucket).join(object);
        fs::create_dir_all(object_dir.join(old_data_dir.to_string()))
            .await
            .expect("old data dir should be created");
        let mut old_fi = FileInfo::new(&format!("{bucket}/{object}"), 1, 1);
        old_fi.name = object.to_string();
        old_fi.version_id = Some(version_id);
        old_fi.data_dir = Some(old_data_dir);
        old_fi.size = 1;
        old_fi.mod_time = Some(OffsetDateTime::now_utc());
        let mut old_meta = FileMeta::default();
        old_meta.add_version(old_fi).expect("old metadata should accept file info");
        let old_meta_buf = old_meta.marshal_msg().expect("old metadata should encode");
        fs::write(object_dir.join(STORAGE_FORMAT_FILE), old_meta_buf.clone())
            .await
            .expect("old metadata should be written");

        let tmp_data_dir = disk_root
            .join(RUSTFS_META_TMP_BUCKET)
            .join(tmp_object)
            .join(new_data_dir.to_string());
        fs::create_dir_all(&tmp_data_dir)
            .await
            .expect("new tmp data dir should be created");
        fs::write(tmp_data_dir.join("part.1"), b"new")
            .await
            .expect("new tmp part should be written");

        let mut new_fi = FileInfo::new(&format!("{bucket}/{object}"), 1, 1);
        new_fi.name = object.to_string();
        new_fi.version_id = Some(version_id);
        new_fi.data_dir = Some(new_data_dir);
        new_fi.size = 1;
        new_fi.mod_time = Some(OffsetDateTime::now_utc());

        let disks = vec![Some(disk), None];
        let file_infos = vec![new_fi.clone(), new_fi];
        let result = SetDisks::rename_data(&disks, RUSTFS_META_TMP_BUCKET, tmp_object, &file_infos, bucket, object, 2).await;

        assert!(result.is_err());
        let restored_meta = fs::read(object_dir.join(STORAGE_FORMAT_FILE))
            .await
            .expect("destination metadata should remain readable");
        assert_eq!(restored_meta, old_meta_buf);
        assert!(!object_dir.join(object).join(STORAGE_FORMAT_FILE).exists());
        assert!(!object_dir.join(new_data_dir.to_string()).exists());
        assert!(
            !object_dir
                .join(old_data_dir.to_string())
                .join(STORAGE_FORMAT_FILE_BACKUP)
                .exists()
        );
    }

    #[tokio::test]
    async fn test_rename_data_inline_quorum_failure_rolls_back_destination_object() {
        let dir = tempfile::tempdir().expect("tempdir should be created");
        let disk_root = dir.path().join("disk0");
        fs::create_dir_all(&disk_root).await.expect("disk root should be created");
        let endpoint = Endpoint::try_from(disk_root.to_str().expect("disk 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");

        let bucket = "bucket";
        let object = "inline-object";
        let tmp_object = "tmp-inline-object";
        let version_id = Uuid::parse_str("aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa").expect("version id should parse");

        match disk.make_volume(bucket).await {
            Ok(()) | Err(DiskError::VolumeExists) => {}
            Err(err) => panic!("bucket should be available: {err:?}"),
        }
        match disk.make_volume(RUSTFS_META_TMP_BUCKET).await {
            Ok(()) | Err(DiskError::VolumeExists) => {}
            Err(err) => panic!("tmp bucket should be available: {err:?}"),
        }

        let object_dir = disk_root.join(bucket).join(object);
        fs::create_dir_all(&object_dir).await.expect("object dir should be created");
        let mut old_fi = FileInfo::new(&format!("{bucket}/{object}"), 1, 1);
        old_fi.name = object.to_string();
        old_fi.version_id = Some(version_id);
        old_fi.data = Some(Bytes::from_static(b"old-inline"));
        old_fi.size = 10;
        old_fi.mod_time = Some(OffsetDateTime::now_utc());
        let mut old_meta = FileMeta::default();
        old_meta.add_version(old_fi).expect("old metadata should accept file info");
        let old_meta_buf = old_meta.marshal_msg().expect("old metadata should encode");
        fs::write(object_dir.join(STORAGE_FORMAT_FILE), old_meta_buf.clone())
            .await
            .expect("old metadata should be written");

        let mut new_fi = FileInfo::new(&format!("{bucket}/{object}"), 1, 1);
        new_fi.name = object.to_string();
        new_fi.version_id = Some(version_id);
        new_fi.data = Some(Bytes::from_static(b"new-inline"));
        new_fi.size = 10;
        new_fi.mod_time = Some(OffsetDateTime::now_utc());

        let disks = vec![Some(disk), None];
        let file_infos = vec![new_fi.clone(), new_fi];
        let result = SetDisks::rename_data(&disks, RUSTFS_META_TMP_BUCKET, tmp_object, &file_infos, bucket, object, 2).await;

        assert!(result.is_err());
        let restored_meta = fs::read(object_dir.join(STORAGE_FORMAT_FILE))
            .await
            .expect("destination metadata should remain readable");
        assert_eq!(restored_meta, old_meta_buf);
    }

    #[test]
    fn disk_health_entry_returns_cached_value_within_ttl() {
        let entry = DiskHealthEntry {
            last_check: Instant::now(),
            online: true,
        };

        assert_eq!(entry.cached_value(), Some(true));
    }

    #[test]
    fn disk_health_entry_expires_after_ttl() {
        let entry = DiskHealthEntry {
            last_check: Instant::now() - (DISK_HEALTH_CACHE_TTL + Duration::from_millis(100)),
            online: true,
        };

        assert!(entry.cached_value().is_none());
    }

    #[test]
    fn test_check_part_constants() {
        // Test that all CHECK_PART constants have expected values
        assert_eq!(CHECK_PART_UNKNOWN, 0);
        assert_eq!(CHECK_PART_SUCCESS, 1);
        assert_eq!(CHECK_PART_FILE_NOT_FOUND, 4); // The actual value is 4, not 2
        assert_eq!(CHECK_PART_VOLUME_NOT_FOUND, 3);
        assert_eq!(CHECK_PART_FILE_CORRUPT, 5);
    }

    #[test]
    fn test_is_min_allowed_part_size() {
        // Test minimum part size validation
        assert!(!is_min_allowed_part_size(0));
        assert!(!is_min_allowed_part_size(1024)); // 1KB - too small
        assert!(!is_min_allowed_part_size(1024 * 1024)); // 1MB - too small
        assert!(is_min_allowed_part_size(5 * 1024 * 1024)); // 5MB - minimum allowed
        assert!(is_min_allowed_part_size(10 * 1024 * 1024)); // 10MB - allowed
        assert!(is_min_allowed_part_size(100 * 1024 * 1024)); // 100MB - allowed
    }

    #[test]
    fn resolve_delete_version_state_clears_delete_marker_for_replica_marker_version_purge() {
        let opts = ObjectOptions {
            versioned: true,
            version_id: Some(Uuid::new_v4().to_string()),
            delete_replication: Some(ReplicationState {
                replica_status: ReplicationStatusType::Replica,
                ..Default::default()
            }),
            ..Default::default()
        };
        let current = ObjectInfo {
            version_id: Some(Uuid::new_v4()),
            delete_marker: true,
            ..Default::default()
        };

        let (mark_delete, delete_marker) = resolve_delete_version_state(&opts, &current, true);

        assert!(!mark_delete);
        assert!(
            !delete_marker,
            "replica purge of an existing delete marker version must remove that version, not preserve delete-marker semantics"
        );
    }

    #[test]
    fn resolve_delete_version_state_keeps_null_marker_removal_write_shape_undeleted() {
        // Removing a null delete marker by explicit version id on a
        // versioning-suspended bucket must NOT mark the write request
        // `deleted`: `FileMeta::delete_version` re-creates a marker for
        // `deleted` requests carrying the null identity (the suspended-bucket
        // "delete mints a marker" path), which would resurrect the marker
        // being removed. The response-side marker semantics come from
        // `explicit_delete_removed_marker` instead (issue #6745).
        let opts = ObjectOptions {
            version_suspended: true,
            version_id: Some(Uuid::nil().to_string()),
            ..Default::default()
        };
        let current = ObjectInfo {
            version_id: Some(Uuid::nil()),
            delete_marker: true,
            ..Default::default()
        };

        let (mark_delete, delete_marker) = resolve_delete_version_state(&opts, &current, true);

        assert!(!mark_delete);
        assert!(!delete_marker, "the storage write for a null-marker removal must stay undeleted");
        assert!(
            explicit_delete_removed_marker(&opts, &current, true),
            "the response must still report delete-marker semantics"
        );
    }

    #[test]
    fn explicit_delete_removed_marker_is_limited_to_plain_marker_removals() {
        let opts = ObjectOptions {
            versioned: true,
            version_id: Some(Uuid::new_v4().to_string()),
            ..Default::default()
        };
        let marker = ObjectInfo {
            version_id: Some(Uuid::new_v4()),
            delete_marker: true,
            ..Default::default()
        };
        assert!(explicit_delete_removed_marker(&opts, &marker, true));
        assert!(
            !explicit_delete_removed_marker(&opts, &marker, false),
            "missing versions are not marker removals"
        );

        let data_version = ObjectInfo {
            version_id: marker.version_id,
            delete_marker: false,
            ..Default::default()
        };
        assert!(!explicit_delete_removed_marker(&opts, &data_version, true));

        let versionless = ObjectOptions {
            versioned: true,
            ..Default::default()
        };
        assert!(
            !explicit_delete_removed_marker(&versionless, &marker, true),
            "marker creation (no version in the request) is not a removal"
        );

        let replica_purge = ObjectOptions {
            versioned: true,
            version_id: Some(Uuid::new_v4().to_string()),
            delete_replication: Some(ReplicationState {
                replica_status: ReplicationStatusType::Replica,
                ..Default::default()
            }),
            ..Default::default()
        };
        assert!(
            !explicit_delete_removed_marker(&replica_purge, &marker, true),
            "replication purge shapes keep their existing response semantics"
        );
    }

    #[test]
    fn resolve_delete_version_state_keeps_delete_marker_for_replica_marker_creation() {
        let opts = ObjectOptions {
            versioned: true,
            version_id: Some(Uuid::new_v4().to_string()),
            delete_marker: true,
            delete_replication: Some(ReplicationState {
                replica_status: ReplicationStatusType::Replica,
                ..Default::default()
            }),
            ..Default::default()
        };

        let (mark_delete, delete_marker) = resolve_delete_version_state(&opts, &ObjectInfo::default(), false);

        assert!(!mark_delete);
        assert!(delete_marker);
    }

    #[test]
    fn resolve_delete_version_state_creates_marker_for_missing_latest_versioned_delete() {
        let opts = ObjectOptions {
            versioned: true,
            ..Default::default()
        };

        let (mark_delete, delete_marker) = resolve_delete_version_state(&opts, &ObjectInfo::default(), false);

        assert!(mark_delete);
        assert!(delete_marker);
    }

    #[test]
    fn resolve_delete_version_state_creates_missing_suspended_data_movement_marker() {
        let opts = ObjectOptions {
            version_suspended: true,
            version_id: Some(Uuid::nil().to_string()),
            data_movement: true,
            delete_marker: true,
            ..Default::default()
        };

        let (mark_delete, delete_marker) = resolve_delete_version_state(&opts, &ObjectInfo::default(), false);

        assert!(mark_delete);
        assert!(delete_marker);
    }

    #[test]
    fn should_force_delete_marker_for_missing_version_rejects_data_movement_latest_delete() {
        let opts = ObjectOptions {
            versioned: true,
            data_movement: true,
            ..Default::default()
        };

        assert!(!should_force_delete_marker_for_missing_version(&opts));
    }

    #[test]
    fn should_force_delete_marker_for_missing_version_allows_explicit_marker_creation() {
        let opts = ObjectOptions {
            versioned: true,
            data_movement: true,
            delete_marker: true,
            ..Default::default()
        };

        assert!(should_force_delete_marker_for_missing_version(&opts));
    }

    #[test]
    fn resolve_delete_version_state_skips_marker_creation_for_replica_purge_when_version_missing() {
        let opts = ObjectOptions {
            versioned: true,
            version_id: Some(Uuid::new_v4().to_string()),
            delete_replication: Some(ReplicationState {
                replica_status: ReplicationStatusType::Replica,
                ..Default::default()
            }),
            ..Default::default()
        };

        let (mark_delete, delete_marker) = resolve_delete_version_state(&opts, &ObjectInfo::default(), false);

        assert!(
            !mark_delete,
            "replica delete-marker purges should not schedule mark-delete writes when the target version is absent"
        );
        assert!(
            !delete_marker,
            "replica delete-marker purges must become no-ops when the marker version has not arrived on the target yet"
        );
    }

    #[test]
    fn should_preserve_delete_replication_state_for_completed_delete_marker_replication_update() {
        let opts = ObjectOptions {
            version_id: Some(Uuid::new_v4().to_string()),
            delete_replication: Some(ReplicationState {
                replicate_decision_str: "target=true;false;target;".to_string(),
                replication_status_internal: Some("target=COMPLETED;".to_string()),
                targets: replication_statuses_map("target=COMPLETED;"),
                ..Default::default()
            }),
            ..Default::default()
        };

        assert!(
            should_preserve_delete_replication_state(&opts),
            "source delete-marker replication status updates must not be re-evaluated as fresh delete replication requests"
        );
    }

    #[test]
    fn should_not_preserve_delete_replication_state_for_new_version_delete_request() {
        let opts = ObjectOptions {
            version_id: Some(Uuid::new_v4().to_string()),
            ..Default::default()
        };

        assert!(
            !should_preserve_delete_replication_state(&opts),
            "fresh versioned deletes still need replication eligibility checks"
        );
    }

    #[test]
    fn resolve_delete_version_state_removes_source_delete_marker_version_during_purge_replication() {
        let opts = ObjectOptions {
            versioned: true,
            version_id: Some(Uuid::new_v4().to_string()),
            delete_replication: Some(ReplicationState {
                version_purge_status_internal: Some("target=PENDING;".to_string()),
                purge_targets: version_purge_statuses_map("target=PENDING;"),
                ..Default::default()
            }),
            ..Default::default()
        };
        let current = ObjectInfo {
            version_id: Some(Uuid::new_v4()),
            delete_marker: true,
            ..Default::default()
        };

        let (mark_delete, delete_marker) = resolve_delete_version_state(&opts, &current, true);

        assert!(
            !mark_delete,
            "source delete-marker version purge should delete the local marker instead of rewriting it with purge metadata"
        );
        assert!(
            !delete_marker,
            "source delete-marker version purge should not leave delete-marker semantics behind locally"
        );
    }

    #[test]
    fn test_get_complete_multipart_md5() {
        // Test MD5 calculation for multipart upload
        let parts = vec![
            CompletePart {
                part_num: 1,
                etag: Some("d41d8cd98f00b204e9800998ecf8427e".to_string()),
                checksum_crc32: None,
                checksum_crc32c: None,
                checksum_sha1: None,
                checksum_sha256: None,
                checksum_crc64nvme: None,
            },
            CompletePart {
                part_num: 2,
                etag: Some("098f6bcd4621d373cade4e832627b4f6".to_string()),
                checksum_crc32: None,
                checksum_crc32c: None,
                checksum_sha1: None,
                checksum_sha256: None,
                checksum_crc64nvme: None,
            },
        ];

        let md5 = get_complete_multipart_md5(&parts);
        assert!(md5.ends_with("-2")); // Should end with part count
        assert!(md5.len() > 10); // Should have reasonable length

        // Test with empty parts
        let empty_parts = vec![];
        let empty_result = get_complete_multipart_md5(&empty_parts);
        assert!(empty_result.ends_with("-0"));

        // Test with single part
        let single_part = vec![CompletePart {
            part_num: 1,
            etag: Some("d41d8cd98f00b204e9800998ecf8427e".to_string()),
            checksum_crc32: None,
            checksum_crc32c: None,
            checksum_sha1: None,
            checksum_sha256: None,
            checksum_crc64nvme: None,
        }];
        let single_result = get_complete_multipart_md5(&single_part);
        assert!(single_result.ends_with("-1"));
    }

    #[test]
    fn test_completed_multipart_object_part_preserves_checksums() {
        let checksums = HashMap::from([
            (rustfs_rio::ChecksumType::CRC32.to_string(), "crc32-value".to_string()),
            (rustfs_rio::ChecksumType::CRC32C.to_string(), "crc32c-value".to_string()),
        ]);
        let ext_part = ObjectPartInfo {
            number: 7,
            etag: "etag-7".to_string(),
            size: 123,
            actual_size: 456,
            mod_time: Some(OffsetDateTime::UNIX_EPOCH),
            index: Some(Bytes::from_static(&[1, 2, 3])),
            checksums: Some(checksums.clone()),
            ..Default::default()
        };

        let completed = completed_multipart_object_part(7, &ext_part);

        assert_eq!(completed.number, 7);
        assert_eq!(completed.etag, ext_part.etag);
        assert_eq!(completed.size, ext_part.size);
        assert_eq!(completed.actual_size, ext_part.actual_size);
        assert_eq!(completed.index, ext_part.index);
        assert_eq!(completed.checksums, Some(checksums));
    }

    #[test]
    fn test_get_upload_id_dir() {
        // Test upload ID directory path generation
        let dir = SetDisks::get_upload_id_dir("bucket", "object", "upload-id");
        // The function returns SHA256 hash of bucket/object + upload_id processing
        assert!(dir.len() > 64); // Should be longer than just SHA256 hash
        assert!(dir.contains("/")); // Should contain path separator

        // Test with base64 encoded upload ID
        let result2 = SetDisks::get_upload_id_dir("bucket", "object", "dXBsb2FkLWlk"); // base64 for "upload-id"
        assert!(!result2.is_empty());
        assert!(result2.len() > 10);
    }

    #[test]
    fn test_get_multipart_sha_dir() {
        // Test multipart SHA directory path generation
        let dir = SetDisks::get_multipart_sha_dir("bucket", "object");
        // The function returns SHA256 hash of "bucket/object"
        assert_eq!(dir.len(), 64); // SHA256 hash length
        assert!(!dir.contains("bucket")); // Should be hash, not original text
        assert!(!dir.contains("object")); // Should be hash, not original text

        // Test with empty strings
        let result2 = SetDisks::get_multipart_sha_dir("", "");
        assert!(!result2.is_empty());
        assert_eq!(result2.len(), 64); // SHA256 hex string length

        // Test that different inputs produce different hashes
        let result3 = SetDisks::get_multipart_sha_dir("bucket1", "object1");
        let result4 = SetDisks::get_multipart_sha_dir("bucket2", "object2");
        assert_ne!(result3, result4);
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn test_new_ns_lock_distributed_read_succeeds_with_two_lockers_one_offline() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::DistErasure).await;

        let manager = Arc::new(rustfs_lock::GlobalLockManager::new());
        let healthy_client: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager));
        let failing_client: Arc<dyn LockClient> = Arc::new(FailingClient);
        let set_disks = make_test_set_disks(vec![healthy_client, failing_client]).await;

        let guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_read_lock(Duration::from_millis(100))
            .await
            .expect("read lock should succeed with one healthy locker");

        match guard {
            NamespaceLockGuard::Standard(_) => {}
            NamespaceLockGuard::Fast(_) => panic!("Expected distributed guard for dist-erasure"),
        }
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn test_new_ns_lock_distributed_write_fails_with_two_lockers_one_offline() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::DistErasure).await;

        let manager = Arc::new(rustfs_lock::GlobalLockManager::new());
        let healthy_client: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager));
        let failing_client: Arc<dyn LockClient> = Arc::new(FailingClient);
        let set_disks = make_test_set_disks(vec![healthy_client, failing_client]).await;

        let err = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_millis(100))
            .await
            .expect_err("write lock should fail with one healthy locker");

        let err_str = err.to_string().to_lowercase();
        assert!(
            err_str.contains("quorum") || err_str.contains("not reached"),
            "expected quorum error, got: {err}"
        );
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn test_new_ns_lock_distributed_write_succeeds_with_three_lockers_one_offline() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::DistErasure).await;

        let manager_a = Arc::new(rustfs_lock::GlobalLockManager::new());
        let manager_b = Arc::new(rustfs_lock::GlobalLockManager::new());
        let healthy_a: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager_a));
        let healthy_b: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager_b));
        let failing_client: Arc<dyn LockClient> = Arc::new(FailingClient);
        let set_disks = make_test_set_disks(vec![healthy_a, failing_client, healthy_b]).await;

        let guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_millis(500))
            .await
            .expect("two healthy lockers should satisfy the three-locker write quorum");

        match guard {
            NamespaceLockGuard::Standard(_) => {}
            NamespaceLockGuard::Fast(_) => panic!("Expected distributed guard for dist-erasure"),
        }
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn namespace_lock_domain_includes_pool_namespace() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::DistErasure).await;

        let first: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(Arc::new(rustfs_lock::GlobalLockManager::new())));
        let second: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(Arc::new(rustfs_lock::GlobalLockManager::new())));
        let lockers = vec![first, second];
        let same_pool_first_set = make_test_set_disks_with_ctx(lockers.clone(), bootstrap_ctx()).await;
        let same_pool_second_set = SetDisks::new_with_instance_ctx(
            "test-owner".to_string(),
            Arc::new(RwLock::new(vec![None, None])),
            2,
            1,
            1,
            0,
            same_pool_first_set.set_endpoints.clone(),
            FormatV3::new(2, 2),
            lockers.clone(),
            bootstrap_ctx(),
        )
        .await;
        let other_pool_set = SetDisks::new_with_instance_ctx(
            "test-owner".to_string(),
            Arc::new(RwLock::new(vec![None, None])),
            2,
            1,
            0,
            1,
            same_pool_first_set.set_endpoints.clone(),
            FormatV3::new(1, 2),
            lockers,
            bootstrap_ctx(),
        )
        .await;

        assert!(
            same_pool_first_set.shares_namespace_lock_domain(&same_pool_second_set).await,
            "sets in the same pool share the object namespace lock domain"
        );
        assert!(
            !same_pool_first_set.shares_namespace_lock_domain(&other_pool_set).await,
            "different pool namespaces must not be deduplicated solely by identical clients"
        );
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn streaming_reader_holds_read_lock_until_eof() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::Erasure).await;
        let set_disks = make_test_set_disks(vec![Arc::new(LocalClient::with_manager(Arc::new(
            rustfs_lock::GlobalLockManager::new(),
        )))])
        .await;

        let read_guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_read_lock(Duration::from_secs(1))
            .await
            .expect("read lock should be acquired");
        let read_guard = ObjectLockDiagGuard::new(
            read_guard,
            false,
            "GetObject",
            Some("bucket".to_owned()),
            Some("object".to_owned()),
            None,
            "read",
        );
        let mut reader = finish_set_disk_read_lock(
            GetObjectReader {
                stream: Box::new(Cursor::new(b"body")),
                object_info: ObjectInfo::default(),
                buffered_body: None,
                body_source: Default::default(),
            },
            Some(read_guard),
            "bucket",
            "object",
        );

        let blocked_write = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_millis(100))
            .await;
        assert!(blocked_write.is_err(), "a stalled response must block an overwrite");

        let mut body = Vec::new();
        reader.stream.read_to_end(&mut body).await.expect("stream should reach EOF");
        assert_eq!(body, b"body");

        let write_guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_secs(1))
            .await;
        assert!(write_guard.is_ok(), "the overwrite should proceed after EOF releases the read lock");
        drop(write_guard);

        let read_guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_read_lock(Duration::from_secs(1))
            .await
            .expect("second read lock should be acquired");
        let reader = finish_set_disk_read_lock(
            GetObjectReader {
                stream: Box::new(tokio::io::empty()),
                object_info: ObjectInfo::default(),
                buffered_body: None,
                body_source: Default::default(),
            },
            Some(ObjectLockDiagGuard::new(
                read_guard,
                false,
                "GetObject",
                Some("bucket".to_owned()),
                Some("object".to_owned()),
                None,
                "read",
            )),
            "bucket",
            "object",
        );
        drop(reader);

        let write_after_drop = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_secs(1))
            .await;
        assert!(write_after_drop.is_ok(), "dropping the response must release the read lock");
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn copy_object_honors_no_lock_when_outer_write_lock_is_held() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::Erasure).await;
        let set_disks = make_test_set_disks(vec![Arc::new(LocalClient::with_manager(Arc::new(
            rustfs_lock::GlobalLockManager::new(),
        )))])
        .await;

        let _outer_guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_secs(1))
            .await
            .expect("outer write lock should be acquired");

        let mut src_info = ObjectInfo {
            metadata_only: true,
            ..Default::default()
        };
        let dst_opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

        let result = timeout(
            Duration::from_secs(1),
            set_disks.copy_object(
                "bucket",
                "object",
                "bucket",
                "object",
                &mut src_info,
                &ObjectOptions::default(),
                &dst_opts,
            ),
        )
        .await
        .expect("no_lock copy path must not wait for the outer lock");

        let err = result.expect_err("empty test disks should fail after bypassing the inner lock");
        assert!(
            !err.to_string().to_ascii_lowercase().contains("lock"),
            "copy_object returned a lock error despite no_lock=true: {err}"
        );
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn copy_object_rejects_metadata_only_cross_key() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::Erasure).await;
        let set_disks = make_test_set_disks(vec![Arc::new(LocalClient::with_manager(Arc::new(
            rustfs_lock::GlobalLockManager::new(),
        )))])
        .await;

        let mut src_info = ObjectInfo {
            metadata_only: true,
            ..Default::default()
        };

        let err = set_disks
            .copy_object(
                "bucket",
                "source",
                "bucket",
                "dest",
                &mut src_info,
                &ObjectOptions::default(),
                &ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await
            .expect_err("metadata-only lower copy is only valid for self-copy updates");

        assert!(matches!(err, StorageError::NotImplemented));
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn delete_object_honors_no_lock_when_outer_write_lock_is_held() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::Erasure).await;
        let set_disks = make_test_set_disks(vec![Arc::new(LocalClient::with_manager(Arc::new(
            rustfs_lock::GlobalLockManager::new(),
        )))])
        .await;

        let _outer_guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_secs(1))
            .await
            .expect("outer write lock should be acquired");

        let result = timeout(
            Duration::from_secs(1),
            set_disks.delete_object(
                "bucket",
                "object",
                ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                },
            ),
        )
        .await
        .expect("no_lock delete path must not wait for the outer lock");

        let err = result.expect_err("empty test disks should fail after bypassing the inner lock");
        assert!(
            !err.to_string().to_ascii_lowercase().contains("lock"),
            "delete_object returned a lock error despite no_lock=true: {err}"
        );
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn delete_prefix_does_not_lock_literal_prefix_key() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::Erasure).await;
        let set_disks = make_test_set_disks(vec![Arc::new(LocalClient::with_manager(Arc::new(
            rustfs_lock::GlobalLockManager::new(),
        )))])
        .await;

        let _outer_guard = set_disks
            .new_ns_lock("bucket", "prefix")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_secs(1))
            .await
            .expect("outer write lock should be acquired");

        let result = timeout(
            Duration::from_secs(1),
            set_disks.delete_object(
                "bucket",
                "prefix",
                ObjectOptions {
                    delete_prefix: true,
                    ..Default::default()
                },
            ),
        )
        .await
        .expect("broad prefix delete must not wait on a literal prefix namespace lock");

        if let Err(err) = result {
            assert!(
                !err.to_string().to_ascii_lowercase().contains("lock"),
                "broad prefix delete returned a lock error: {err}"
            );
        }
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn delete_prefix_object_honors_no_lock_when_outer_write_lock_is_held() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::Erasure).await;
        let set_disks = make_test_set_disks(vec![Arc::new(LocalClient::with_manager(Arc::new(
            rustfs_lock::GlobalLockManager::new(),
        )))])
        .await;

        let _outer_guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_secs(1))
            .await
            .expect("outer write lock should be acquired");

        let result = timeout(
            Duration::from_secs(1),
            set_disks.delete_object(
                "bucket",
                "object",
                ObjectOptions {
                    delete_prefix: true,
                    delete_prefix_object: true,
                    no_lock: true,
                    ..Default::default()
                },
            ),
        )
        .await
        .expect("no_lock exact prefix delete path must not wait for the outer lock");

        if let Err(err) = result {
            assert!(
                !err.to_string().to_ascii_lowercase().contains("lock"),
                "no_lock exact prefix delete returned a lock error: {err}"
            );
        }
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn delete_prefix_object_locks_real_object_key() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::Erasure).await;
        let set_disks = make_test_set_disks(vec![Arc::new(LocalClient::with_manager(Arc::new(
            rustfs_lock::GlobalLockManager::new(),
        )))])
        .await;

        let _outer_guard = set_disks
            .new_ns_lock("bucket", "object")
            .await
            .expect("namespace lock should be created")
            .get_write_lock(Duration::from_secs(1))
            .await
            .expect("outer write lock should be acquired");

        let result = timeout(
            Duration::from_millis(50),
            set_disks.delete_object(
                "bucket",
                "object",
                ObjectOptions {
                    delete_prefix: true,
                    delete_prefix_object: true,
                    ..Default::default()
                },
            ),
        )
        .await;

        assert!(result.is_err(), "exact prefix delete should wait on the real object namespace lock");
    }

    async fn make_single_local_disk() -> (TempDir, DiskStore) {
        let dir = tempfile::tempdir().expect("tempdir should be created");
        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");
        (dir, disk)
    }

    async fn make_set_disks_with(disks: Vec<Option<DiskStore>>) -> Arc<SetDisks> {
        let drive_count = disks.len();
        let endpoints = (0..drive_count)
            .map(|i| Endpoint::try_from(format!("http://127.0.0.1:{}/data", 9000 + i).as_str()).expect("endpoint should parse"))
            .collect::<Vec<_>>();

        SetDisks::new(
            "test-owner".to_string(),
            Arc::new(RwLock::new(disks)),
            drive_count,
            0,
            0,
            0,
            endpoints,
            FormatV3::new(1, drive_count),
            vec![Arc::new(LocalClient::with_manager(Arc::new(
                rustfs_lock::GlobalLockManager::new(),
            )))],
        )
        .await
    }

    // backlog#1315: the memoized dirty scope must be byte-for-byte identical to
    // the ad-hoc scope the previous per-write construction produced, otherwise
    // dirty-disk keys diverge from the disk-cache keys and capacity counts drift.
    #[tokio::test]
    #[serial(capacity_dirty_scope)]
    async fn capacity_scope_memo_matches_adhoc_and_is_reused() {
        use rustfs_object_capacity::capacity_scope::drain_global_dirty_scopes;

        let (dir_a, disk_a) = make_single_local_disk().await;
        let (dir_b, disk_b) = make_single_local_disk().await;
        let disks = vec![Some(disk_a), Some(disk_b)];
        let set = make_set_disks_with(disks.clone()).await;

        let expected = capacity_scope_from_disks(&disks);
        let first = set.capacity_scope(&disks);
        assert_eq!(*first, expected, "memoized scope must equal the ad-hoc per-disk construction bit-for-bit");

        // Second call on a fully-resolved set returns the very same Arc (no new
        // allocation): proving steady-state writes do not rebuild String/HashSet.
        let second = set.capacity_scope(&disks);
        assert!(Arc::ptr_eq(&first, &second), "steady-state scope must reuse the cached Arc");

        let _ = drain_global_dirty_scopes();
        drop((dir_a, dir_b));
    }

    // backlog#1315: the global registry mutex must be upgraded only on the first
    // write of each generation; steady-state writes skip it. Reverting the
    // generation skip makes the upgrade count grow per write and fails this test.
    #[tokio::test]
    #[serial(capacity_dirty_scope)]
    async fn record_capacity_scope_upgrades_registry_once_per_generation() {
        use rustfs_object_capacity::capacity_scope::{drain_global_dirty_scopes, global_dirty_upgrade_count};

        let (dir_a, disk_a) = make_single_local_disk().await;
        let (dir_b, disk_b) = make_single_local_disk().await;
        let disks = vec![Some(disk_a), Some(disk_b)];
        let set = make_set_disks_with(disks.clone()).await;

        // Start from a clean registry generation.
        let _ = drain_global_dirty_scopes();
        let before = global_dirty_upgrade_count();

        // First write of this generation upgrades the registry exactly once.
        set.record_capacity_scope_if_needed(None, &disks);
        assert_eq!(
            global_dirty_upgrade_count(),
            before + 1,
            "first write of a generation must upgrade the global registry"
        );

        // Subsequent writes in the same generation must not touch the mutex.
        for _ in 0..16 {
            set.record_capacity_scope_if_needed(None, &disks);
        }
        assert_eq!(
            global_dirty_upgrade_count(),
            before + 1,
            "steady-state writes must reuse the generation mark, not re-upgrade"
        );

        // A drain advances the generation; the next write must re-mark so the
        // disks it wrote are captured by the following refresh (no lost update).
        let drained = drain_global_dirty_scopes();
        assert_eq!(drained.len(), 2, "both set disks must have been recorded dirty");
        set.record_capacity_scope_if_needed(None, &disks);
        assert_eq!(
            global_dirty_upgrade_count(),
            before + 2,
            "the first write after a drain must re-upgrade the registry"
        );

        let _ = drain_global_dirty_scopes();
        drop((dir_a, dir_b));
    }

    // backlog#1315: an offline slot must not force the per-write slow path, and
    // the resolved scope must still cover every online disk.
    #[tokio::test]
    #[serial(capacity_dirty_scope)]
    async fn capacity_scope_tolerates_offline_slot_without_reallocating() {
        use rustfs_object_capacity::capacity_scope::drain_global_dirty_scopes;

        let (dir_a, disk_a) = make_single_local_disk().await;
        // Slot 1 is permanently offline (None); the set never resolves it.
        let disks = vec![Some(disk_a), None];
        let set = make_set_disks_with(disks.clone()).await;

        let first = set.capacity_scope(&disks);
        assert_eq!(first.disks.len(), 1, "only the online disk contributes to the scope");
        // Even though the set is not "complete" (slot 1 unresolved), repeated
        // writes with the same online set reuse the cached Arc via the
        // no-unresolved-slot fast path.
        let second = set.capacity_scope(&disks);
        assert!(
            Arc::ptr_eq(&first, &second),
            "a stable online subset must reuse the cached Arc even with an offline slot"
        );

        let _ = drain_global_dirty_scopes();
        drop(dir_a);
    }

    // issue #4189: an orphan directory tree (empty dirs, no xl.meta) must be purged.
    #[tokio::test]
    async fn purge_orphan_dir_object_removes_empty_tree() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        fs::create_dir_all(root.join("bucket").join("pfx").join("a").join("b"))
            .await
            .expect("nested empty dir should be created");
        fs::create_dir_all(root.join("bucket").join("pfx").join("c"))
            .await
            .expect("sibling empty dir should be created");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("purge should succeed");

        assert!(purged, "orphan empty tree should be purged");
        assert!(!root.join("bucket").join("pfx").exists(), "prefix directory should be gone");
        assert!(root.join("bucket").exists(), "bucket volume should remain");
    }

    #[tokio::test]
    async fn purge_orphan_dir_object_removes_committed_delete_residue() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let data_dir = Uuid::new_v4();
        let transaction = Uuid::new_v4();
        let residue = root
            .join("bucket")
            .join("pfx")
            .join("nested")
            .join("object")
            .join(data_dir.to_string());
        fs::create_dir_all(&residue)
            .await
            .expect("committed delete residue should be created");
        fs::write(residue.join("part.1"), b"stale")
            .await
            .expect("stale part should be written");
        fs::write(
            residue.join(format!("{}{}", crate::disk::local::DELETE_DATA_DIR_MARKER_PREFIX, transaction)),
            [],
        )
        .await
        .expect("committed delete marker should be written");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("purge should succeed");

        assert!(purged, "committed delete residue should be purgeable");
        assert!(!root.join("bucket").join("pfx").exists(), "prefix directory should be gone");
    }

    #[tokio::test]
    async fn purge_orphan_dir_object_preserves_uncommitted_data_residue() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let residue = root
            .join("bucket")
            .join("pfx")
            .join("object")
            .join(Uuid::new_v4().to_string());
        fs::create_dir_all(&residue)
            .await
            .expect("uncommitted data residue should be created");
        fs::write(residue.join("part.1"), b"possibly live")
            .await
            .expect("data part should be written");
        fs::write(residue.join("delete-data.not-a-uuid"), [])
            .await
            .expect("malformed marker should be written");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("scan should succeed");

        assert!(!purged, "data without a valid committed marker must be preserved");
        assert!(residue.join("part.1").exists(), "possibly live data must remain");
    }

    #[tokio::test]
    async fn purge_orphan_dir_object_preserves_nested_object_below_committed_residue() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let transaction = Uuid::new_v4();
        let residue = root
            .join("bucket")
            .join("pfx")
            .join("object")
            .join(Uuid::new_v4().to_string());
        fs::create_dir_all(&residue)
            .await
            .expect("committed data directory should be created");
        fs::write(residue.join("part.1"), b"stale")
            .await
            .expect("stale part should be written");
        fs::write(
            residue.join(format!("{}{}", crate::disk::local::DELETE_DATA_DIR_MARKER_PREFIX, transaction)),
            [],
        )
        .await
        .expect("committed delete marker should be written");
        let nested_object = residue.join("nested");
        fs::create_dir_all(&nested_object)
            .await
            .expect("nested object directory should be created");
        fs::write(nested_object.join(STORAGE_FORMAT_FILE), b"meta")
            .await
            .expect("nested object metadata should be written");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("scan should succeed");

        assert!(!purged, "nested object metadata must veto committed-residue cleanup");
        assert!(nested_object.join(STORAGE_FORMAT_FILE).exists(), "nested object metadata must remain");
        assert!(residue.join("part.1").exists(), "committed residue must remain when cleanup is vetoed");
    }

    // issue #4189: a prefix that still anchors a real object must be left intact.
    #[tokio::test]
    async fn purge_orphan_dir_object_preserves_prefix_with_object() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let obj_dir = root.join("bucket").join("pfx").join("obj");
        fs::create_dir_all(&obj_dir).await.expect("object dir should be created");
        fs::write(obj_dir.join(STORAGE_FORMAT_FILE), b"meta")
            .await
            .expect("object metadata should be written");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("scan should succeed");

        assert!(!purged, "prefix containing an object must not be purged");
        assert!(obj_dir.join(STORAGE_FORMAT_FILE).exists(), "object metadata must be preserved");
    }

    #[tokio::test]
    async fn purge_orphan_dir_object_missing_returns_false() {
        let (dir, disk) = make_single_local_disk().await;
        fs::create_dir_all(dir.path().join("bucket"))
            .await
            .expect("bucket volume should be created");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "does-not-exist/")
            .await
            .expect("scan should succeed");

        assert!(!purged, "a missing prefix should report nothing to purge");
    }

    #[tokio::test]
    async fn orphan_directory_purge_preserves_tree_when_a_disk_slot_is_offline() {
        let (dir, disk) = make_single_local_disk().await;
        let prefix_dir = dir.path().join("bucket").join("pfx");
        fs::create_dir_all(prefix_dir.join("nested").join("leaf"))
            .await
            .expect("orphan directory tree should be created");

        let set = make_set_disks_with(vec![Some(disk), None]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("an unavailable slot should fail closed without a scan error");

        assert!(!purged, "an incomplete disk scan must not claim the tree is an orphan");
        assert!(prefix_dir.join("nested/leaf").exists(), "online disk contents must remain untouched");

        let bucket_purged = set.purge_orphan_dir_objects_in_bucket("bucket").await;
        assert!(!bucket_purged, "bucket-wide cleanup must fail closed with an unavailable disk slot");
        assert!(
            prefix_dir.join("nested/leaf").exists(),
            "bucket-wide cleanup must preserve the online tree"
        );
    }

    // Cross-disk safety: if any drive still holds object data under the prefix, refuse
    // to purge on every drive so a degraded/healable object is never destroyed.
    #[tokio::test]
    async fn purge_orphan_dir_object_refuses_when_any_disk_has_data() {
        let (dir0, disk0) = make_single_local_disk().await;
        let (dir1, disk1) = make_single_local_disk().await;

        fs::create_dir_all(dir0.path().join("bucket").join("pfx").join("a"))
            .await
            .expect("disk0 empty tree should be created");

        let obj_dir = dir1.path().join("bucket").join("pfx").join("a");
        fs::create_dir_all(&obj_dir)
            .await
            .expect("disk1 object dir should be created");
        fs::write(obj_dir.join(STORAGE_FORMAT_FILE), b"meta")
            .await
            .expect("disk1 object metadata should be written");

        let set = make_set_disks_with(vec![Some(disk0), Some(disk1)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("scan should succeed");

        assert!(!purged, "must not purge when any disk holds object data");
        assert!(
            obj_dir.join(STORAGE_FORMAT_FILE).exists(),
            "object metadata on the healthy disk must be preserved"
        );
        assert!(
            dir0.path().join("bucket").join("pfx").join("a").exists(),
            "empty tree must be left untouched when the purge is aborted"
        );
    }

    async fn write_committed_residue(object_dir: &std::path::Path) -> std::path::PathBuf {
        let residue = object_dir.join(Uuid::new_v4().to_string());
        fs::create_dir_all(&residue)
            .await
            .expect("committed data directory should be created");
        fs::write(residue.join("part.1"), b"stale")
            .await
            .expect("stale part should be written");
        fs::write(
            residue.join(format!("{}{}", crate::disk::local::DELETE_DATA_DIR_MARKER_PREFIX, Uuid::new_v4())),
            [],
        )
        .await
        .expect("committed delete marker should be written");
        residue
    }

    // #6898: residue from a build that never wrote delete markers blocks only
    // its own ancestor chain; a committed sibling subtree is still reclaimed
    // and the call reports the partial purge.
    #[tokio::test]
    async fn purge_orphan_dir_object_reclaims_committed_subtree_beside_blocked_subtree() {
        let (dir0, disk0) = make_single_local_disk().await;
        let (dir1, disk1) = make_single_local_disk().await;

        let committed = write_committed_residue(&dir0.path().join("bucket/pfx/a/obj")).await;
        let unmarked = dir1.path().join("bucket/pfx/b/obj").join(Uuid::new_v4().to_string());
        fs::create_dir_all(&unmarked)
            .await
            .expect("unmarked residue should be created");
        fs::write(unmarked.join("part.1"), b"stale")
            .await
            .expect("stale part should be written");
        fs::create_dir_all(dir0.path().join("bucket/pfx/b/obj"))
            .await
            .expect("disk0 copy of the blocked chain should be created");

        let set = make_set_disks_with(vec![Some(disk0), Some(disk1)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("scan should succeed");

        assert!(purged, "the committed subtree must be reported as purged");
        assert!(!committed.exists(), "committed residue beside a blocked subtree must be reclaimed");
        assert!(!dir0.path().join("bucket/pfx/a").exists(), "the reclaimed subtree's directories must go");
        assert!(unmarked.join("part.1").exists(), "unmarked residue must survive");
        assert!(
            dir0.path().join("bucket/pfx/b/obj").exists(),
            "a directory blocked on another disk must be left alone on this one"
        );
        assert!(dir0.path().join("bucket/pfx").exists() && dir1.path().join("bucket/pfx").exists());
    }

    // A data dir that is committed residue on one disk but still holds an
    // object below it on another disk is blocked on every disk.
    #[tokio::test]
    async fn purge_orphan_dir_object_blocks_committed_files_by_other_disks_data() {
        let (dir0, disk0) = make_single_local_disk().await;
        let (dir1, disk1) = make_single_local_disk().await;

        let data_dir = Uuid::new_v4();
        let committed = dir0.path().join("bucket/pfx/x").join(data_dir.to_string());
        fs::create_dir_all(&committed)
            .await
            .expect("committed data directory should be created");
        fs::write(committed.join("part.1"), b"stale")
            .await
            .expect("stale part should be written");
        fs::write(
            committed.join(format!("{}{}", crate::disk::local::DELETE_DATA_DIR_MARKER_PREFIX, Uuid::new_v4())),
            [],
        )
        .await
        .expect("committed delete marker should be written");
        let nested = dir1.path().join("bucket/pfx/x").join(data_dir.to_string()).join("nested");
        fs::create_dir_all(&nested)
            .await
            .expect("nested object dir should be created");
        fs::write(nested.join(STORAGE_FORMAT_FILE), b"meta")
            .await
            .expect("nested object metadata should be written");

        let set = make_set_disks_with(vec![Some(disk0), Some(disk1)]).await;
        let purged = set
            .purge_orphan_dir_object("bucket", "pfx/")
            .await
            .expect("scan should succeed");

        assert!(!purged);
        assert!(
            committed.join("part.1").exists(),
            "committed files under a dir blocked elsewhere must remain"
        );
        assert!(nested.join(STORAGE_FORMAT_FILE).exists());
    }

    // An unreadable directory anywhere under the prefix aborts the purge on
    // every disk before anything is deleted.
    #[cfg(unix)]
    #[tokio::test]
    async fn purge_orphan_dir_object_aborts_when_a_directory_is_unreadable() {
        use std::os::unix::fs::PermissionsExt;

        let (dir, disk) = make_single_local_disk().await;
        let committed = write_committed_residue(&dir.path().join("bucket/pfx/b/obj")).await;
        let sealed = dir.path().join("bucket/pfx/a");
        fs::create_dir_all(&sealed).await.expect("sealed directory should be created");
        fs::set_permissions(&sealed, std::fs::Permissions::from_mode(0o000))
            .await
            .expect("sealed directory should become unreadable");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let purged = set.purge_orphan_dir_object("bucket", "pfx/").await;
        fs::set_permissions(&sealed, std::fs::Permissions::from_mode(0o755))
            .await
            .expect("sealed directory should be restored");

        assert!(matches!(purged, Ok(false)), "an unreadable directory must abort the purge");
        assert!(committed.join("part.1").exists(), "nothing may be deleted once classification failed");
    }

    // After a scan met unpurgeable data the prefix is not rescanned for a
    // while, even when new committed residue appears under it.
    #[tokio::test]
    async fn purge_orphan_dir_object_backs_off_after_blocked_scan() {
        let (dir, disk) = make_single_local_disk().await;
        let unmarked = dir.path().join("bucket/pfx/b/obj").join(Uuid::new_v4().to_string());
        fs::create_dir_all(&unmarked)
            .await
            .expect("unmarked residue should be created");
        fs::write(unmarked.join("part.1"), b"stale")
            .await
            .expect("stale part should be written");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        assert!(
            !set.purge_orphan_dir_object("bucket", "pfx/")
                .await
                .expect("scan should succeed")
        );

        let committed = write_committed_residue(&dir.path().join("bucket/pfx/a/obj")).await;
        assert!(
            !set.purge_orphan_dir_object("bucket", "pfx/")
                .await
                .expect("scan should succeed"),
            "a prefix in backoff is not rescanned"
        );
        assert!(committed.exists());
        assert!(
            set.purge_orphan_dir_object("bucket", "pfx/a/")
                .await
                .expect("scan should succeed"),
            "backoff is per prefix, a narrower prefix still purges"
        );
        assert!(!committed.exists());

        let committed = write_committed_residue(&dir.path().join("bucket/pfx/a/obj")).await;
        set.clear_orphan_purge_backoff();
        assert!(
            set.purge_orphan_dir_object("bucket", "pfx/")
                .await
                .expect("scan should succeed")
        );
        assert!(!committed.exists());
        assert!(unmarked.join("part.1").exists());
    }

    // Build an `xl.meta` under `object_dir` whose versions reference `data_dirs`
    // (one Object version per data dir, each with its own version id).
    async fn write_object_meta_with_data_dirs(object_dir: &std::path::Path, bucket: &str, object: &str, data_dirs: &[Uuid]) {
        fs::create_dir_all(object_dir).await.expect("object dir should be created");
        let mut meta = FileMeta::default();
        for data_dir in data_dirs {
            let mut fi = FileInfo::new(&format!("{bucket}/{object}"), 1, 1);
            fi.name = object.to_string();
            fi.version_id = Some(Uuid::new_v4());
            fi.data_dir = Some(*data_dir);
            fi.size = 1;
            fi.mod_time = Some(OffsetDateTime::now_utc());
            meta.add_version(fi).expect("metadata should accept file info");
        }
        let buf = meta.marshal_msg().expect("metadata should encode");
        fs::write(object_dir.join(STORAGE_FORMAT_FILE), buf)
            .await
            .expect("metadata should be written");
    }

    // #3231/#3191: a data dir on disk that no version references (a pre-#3510
    // unversioned-overwrite leak) must be reclaimed, leaving the live one intact.
    #[tokio::test]
    async fn reclaim_orphan_data_dirs_removes_unreferenced_dir() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let live = Uuid::new_v4();
        let orphan = Uuid::new_v4();

        let object_dir = root.join("bucket").join("obj");
        write_object_meta_with_data_dirs(&object_dir, "bucket", "obj", &[live]).await;
        fs::create_dir_all(object_dir.join(live.to_string()))
            .await
            .expect("live data dir should be created");
        fs::write(object_dir.join(live.to_string()).join("part.1"), b"live")
            .await
            .expect("live part should be written");
        fs::create_dir_all(object_dir.join(orphan.to_string()))
            .await
            .expect("orphan data dir should be created");
        fs::write(object_dir.join(orphan.to_string()).join("part.1"), b"stale")
            .await
            .expect("orphan part should be written");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let removed = set
            .reclaim_orphan_data_dirs("bucket", "obj")
            .await
            .expect("reclaim should succeed");

        assert_eq!(removed, 1, "exactly the unreferenced data dir should be removed");
        assert!(object_dir.join(live.to_string()).exists(), "referenced data dir must be preserved");
        assert!(!object_dir.join(orphan.to_string()).exists(), "orphaned data dir must be removed");
        assert!(object_dir.join(STORAGE_FORMAT_FILE).exists(), "metadata must be preserved");
    }

    async fn recv_abandoned_parts_trace(
        trace: &mut rustfs_common::trace_bus::TraceSubscription,
        bucket: &str,
        object: &str,
        state: &str,
    ) -> rustfs_common::trace_bus::TraceEvent {
        for _ in 0..32 {
            let event = tokio::time::timeout(std::time::Duration::from_secs(1), trace.recv())
                .await
                .expect("abandoned-parts trace event should arrive")
                .expect("trace bus should stay open");
            if event.kind == rustfs_common::trace_bus::TraceKind::Heal
                && event.func == rustfs_common::trace_bus::TraceFunc::HealCheckAbandonedParts
                && event.bucket.as_deref() == Some(bucket)
                && event.object.as_deref() == Some(object)
                && trace_attr_string(&event, "state").as_deref() == Some(state)
            {
                return (*event).clone();
            }
        }

        panic!("expected abandoned-parts trace state {state} for {bucket}/{object}");
    }

    fn trace_attr_string(event: &rustfs_common::trace_bus::TraceEvent, key: &str) -> Option<String> {
        event.attrs.iter().find_map(|attr| {
            if attr.key != key {
                return None;
            }
            Some(match &attr.value {
                rustfs_common::trace_bus::TraceVal::Bool(value) => value.to_string(),
                rustfs_common::trace_bus::TraceVal::U64(value) => value.to_string(),
                rustfs_common::trace_bus::TraceVal::I64(value) => value.to_string(),
                rustfs_common::trace_bus::TraceVal::Str(value) => value.to_string(),
            })
        })
    }

    #[tokio::test]
    async fn check_abandoned_parts_dry_run_counts_without_deleting() {
        let mut trace = rustfs_common::trace_bus::subscribe_trace_events();
        let (dir, disk) = make_single_local_disk().await;
        let live = Uuid::new_v4();
        let orphan = Uuid::new_v4();

        let object_dir = dir.path().join("bucket").join("obj");
        write_object_meta_with_data_dirs(&object_dir, "bucket", "obj", &[live]).await;
        fs::create_dir_all(object_dir.join(live.to_string()))
            .await
            .expect("live data dir should be created");
        fs::create_dir_all(object_dir.join(orphan.to_string()))
            .await
            .expect("orphan data dir should be created");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        set.check_abandoned_parts(
            "bucket",
            "obj",
            &HealOpts {
                dry_run: true,
                no_lock: true,
                ..Default::default()
            },
        )
        .await
        .expect("dry-run abandoned-parts check should succeed");
        let dry_run_trace = recv_abandoned_parts_trace(&mut trace, "bucket", "obj", "dry_run_matched").await;
        assert_eq!(trace_attr_string(&dry_run_trace, "dry_run").as_deref(), Some("true"));
        assert_eq!(trace_attr_string(&dry_run_trace, "data_dirs").as_deref(), Some("1"));

        assert!(object_dir.join(live.to_string()).exists(), "referenced data dir must be preserved");
        assert!(object_dir.join(orphan.to_string()).exists(), "dry-run must not remove orphaned data dir");

        set.check_abandoned_parts(
            "bucket",
            "obj",
            &HealOpts {
                no_lock: true,
                ..Default::default()
            },
        )
        .await
        .expect("abandoned-parts check should reclaim stale data dir");
        let reclaim_trace = recv_abandoned_parts_trace(&mut trace, "bucket", "obj", "reclaimed").await;
        assert_eq!(trace_attr_string(&reclaim_trace, "dry_run").as_deref(), Some("false"));
        assert_eq!(trace_attr_string(&reclaim_trace, "data_dirs").as_deref(), Some("1"));

        assert!(
            object_dir.join(live.to_string()).exists(),
            "referenced data dir must remain after reclaim"
        );
        assert!(!object_dir.join(orphan.to_string()).exists(), "orphaned data dir must be removed");
    }

    #[tokio::test]
    async fn reclaim_orphan_data_dirs_recovers_deferred_cleanup_after_restart() {
        let (dir, disk) = make_single_local_disk().await;
        let live = Uuid::new_v4();
        let orphan = Uuid::new_v4();
        let object_dir = dir.path().join("bucket").join("obj");
        write_object_meta_with_data_dirs(&object_dir, "bucket", "obj", &[live]).await;
        fs::create_dir_all(object_dir.join(live.to_string()))
            .await
            .expect("live data dir should be created");
        let orphan_path = format!("obj/{orphan}");
        disk.write_all("bucket", &format!("{orphan_path}/part.1"), Bytes::from_static(b"stale"))
            .await
            .expect("orphan part should be written");

        let _token = disk
            .acquire_snapshot_lease("bucket", &orphan_path)
            .await
            .expect("snapshot lease should be acquired");
        assert_eq!(
            disk.delete_data_dir(
                "bucket",
                &orphan_path,
                DeleteOptions {
                    recursive: true,
                    ..Default::default()
                },
            )
            .await
            .expect("cleanup should be deferred"),
            DataDirDeleteStatus::Deferred
        );
        drop(disk);

        let endpoint =
            Endpoint::try_from(dir.path().to_str().expect("tempdir path should be utf8")).expect("endpoint should parse");
        let restarted = new_disk(
            &endpoint,
            &DiskOption {
                cleanup: false,
                health_check: false,
            },
        )
        .await
        .expect("disk should restart");
        let set = make_set_disks_with(vec![Some(restarted)]).await;
        let removed = set
            .reclaim_orphan_data_dirs("bucket", "obj")
            .await
            .expect("restart reclaim should succeed");

        assert_eq!(removed, 1, "the deferred orphan should be reclaimed after restart");
        assert!(object_dir.join(live.to_string()).exists(), "referenced data dir must be preserved");
        assert!(!object_dir.join(orphan.to_string()).exists(), "deferred orphan must be removed");
    }

    // Nothing to reclaim when every physical data dir is still referenced.
    #[tokio::test]
    async fn reclaim_orphan_data_dirs_keeps_referenced_dir() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let live = Uuid::new_v4();

        let object_dir = root.join("bucket").join("obj");
        write_object_meta_with_data_dirs(&object_dir, "bucket", "obj", &[live]).await;
        fs::create_dir_all(object_dir.join(live.to_string()))
            .await
            .expect("live data dir should be created");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let removed = set
            .reclaim_orphan_data_dirs("bucket", "obj")
            .await
            .expect("reclaim should succeed");

        assert_eq!(removed, 0, "no data dir should be removed");
        assert!(object_dir.join(live.to_string()).exists(), "referenced data dir must be preserved");
    }

    // Fail closed: a data dir present without a readable xl.meta is degraded, and
    // must never be removed (a heal has to run first).
    #[tokio::test]
    async fn reclaim_orphan_data_dirs_aborts_when_meta_missing() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let stray = Uuid::new_v4();

        let object_dir = root.join("bucket").join("obj");
        fs::create_dir_all(object_dir.join(stray.to_string()))
            .await
            .expect("data dir should be created");
        fs::write(object_dir.join(stray.to_string()).join("part.1"), b"data")
            .await
            .expect("part should be written");
        fs::write(
            object_dir.join(stray.to_string()).join(format!(
                "{}{}",
                crate::disk::local::RESERVED_DELETE_DATA_DIR_MARKER_PREFIX,
                Uuid::new_v4()
            )),
            [],
        )
        .await
        .expect("pre-commit delete reservation should be written");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let removed = set
            .reclaim_orphan_data_dirs("bucket", "obj")
            .await
            .expect("reclaim should succeed");

        assert_eq!(removed, 0, "must not remove data dirs when metadata is absent");
        assert!(
            object_dir.join(stray.to_string()).exists(),
            "degraded object's data dir must be preserved"
        );
    }

    #[tokio::test]
    async fn reclaim_orphan_data_dirs_recovers_committed_delete_marker_without_meta() {
        let (dir, disk) = make_single_local_disk().await;
        let stale = Uuid::new_v4();
        let transaction = Uuid::new_v4();
        let object_dir = dir.path().join("bucket").join("obj");
        let stale_dir = object_dir.join(stale.to_string());
        fs::create_dir_all(&stale_dir)
            .await
            .expect("committed stale data dir should be created");
        fs::write(stale_dir.join("part.1"), b"stale")
            .await
            .expect("stale part should be written");
        fs::write(
            stale_dir.join(format!("{}{}", crate::disk::local::DELETE_DATA_DIR_MARKER_PREFIX, transaction)),
            [],
        )
        .await
        .expect("committed delete marker should be written");

        let set = make_set_disks_with(vec![Some(disk)]).await;
        let removed = set
            .reclaim_orphan_data_dirs("bucket", "obj")
            .await
            .expect("upgrade reclaim should succeed");

        assert_eq!(removed, 1, "the committed delete residue should be reclaimed");
        assert!(!stale_dir.exists(), "the committed stale data dir should be removed");
    }

    // Cross-replica union: a data dir referenced by ANOTHER disk's xl.meta must be
    // kept even where the local replica does not name it.
    #[tokio::test]
    async fn reclaim_orphan_data_dirs_keeps_dir_referenced_by_other_replica() {
        let (dir0, disk0) = make_single_local_disk().await;
        let (dir1, disk1) = make_single_local_disk().await;
        let shared = Uuid::new_v4();

        // disk0: meta references only a different dir, but physically holds `shared`.
        let other = Uuid::new_v4();
        let obj0 = dir0.path().join("bucket").join("obj");
        write_object_meta_with_data_dirs(&obj0, "bucket", "obj", &[other]).await;
        fs::create_dir_all(obj0.join(other.to_string()))
            .await
            .expect("dir should be created");
        fs::create_dir_all(obj0.join(shared.to_string()))
            .await
            .expect("dir should be created");

        // disk1: meta references `shared`.
        let obj1 = dir1.path().join("bucket").join("obj");
        write_object_meta_with_data_dirs(&obj1, "bucket", "obj", &[shared]).await;
        fs::create_dir_all(obj1.join(shared.to_string()))
            .await
            .expect("dir should be created");

        let set = make_set_disks_with(vec![Some(disk0), Some(disk1)]).await;
        let removed = set
            .reclaim_orphan_data_dirs("bucket", "obj")
            .await
            .expect("reclaim should succeed");

        assert_eq!(removed, 0, "a dir referenced by any replica must be kept");
        assert!(obj0.join(shared.to_string()).exists(), "cross-referenced data dir must survive");
        assert!(obj0.join(other.to_string()).exists(), "locally referenced data dir must survive");
    }

    // backlog#898 A5: old == committed data dir => anti-misdelete guard skips the
    // whole cleanup; the live (just-committed) dir must be left untouched.
    #[tokio::test]
    async fn commit_rename_data_dir_skips_delete_when_old_equals_committed_dir() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let bucket = "bucket";
        let object = "object";
        fs::create_dir_all(root.join(bucket))
            .await
            .expect("bucket volume should be created");

        let same_dir = Uuid::parse_str("55555555-5555-5555-5555-555555555555").expect("dir should parse");
        let data_path = root.join(bucket).join(object).join(same_dir.to_string());
        fs::create_dir_all(&data_path).await.expect("data dir should be created");
        fs::write(data_path.join("part.1"), b"live")
            .await
            .expect("live part should be written");

        let set = make_set_disks_with(vec![Some(disk.clone())]).await;
        let cleanup = set
            .commit_rename_data_dir(&[Some(disk.clone())], bucket, object, &same_dir.to_string(), &same_dir.to_string(), 1)
            .await;

        assert_eq!(cleanup.attempted, 0, "guard must skip: no delete may be issued");
        assert!(cleanup.unreclaimed_disks.is_empty());
        assert!(!cleanup.has_residue());
        assert!(data_path.join("part.1").exists(), "committed data dir must NOT be deleted");
    }

    // backlog#898 A5b: old != committed and old dir exists => normal reclaim; the
    // dereferenced old dir is physically removed and the receipt reports success.
    #[tokio::test]
    async fn commit_rename_data_dir_reclaims_distinct_old_dir() {
        let (dir, disk) = make_single_local_disk().await;
        let root = dir.path();
        let bucket = "bucket";
        let object = "object";
        fs::create_dir_all(root.join(bucket))
            .await
            .expect("bucket volume should be created");

        let old_dir = Uuid::parse_str("11111111-1111-1111-1111-111111111111").expect("old dir should parse");
        let new_dir = Uuid::parse_str("22222222-2222-2222-2222-222222222222").expect("new dir should parse");
        let old_path = root.join(bucket).join(object).join(old_dir.to_string());
        fs::create_dir_all(&old_path).await.expect("old data dir should be created");
        fs::write(old_path.join("part.1"), b"stale")
            .await
            .expect("stale part should be written");

        let set = make_set_disks_with(vec![Some(disk.clone())]).await;
        let cleanup = set
            .commit_rename_data_dir(&[Some(disk.clone())], bucket, object, &old_dir.to_string(), &new_dir.to_string(), 1)
            .await;

        assert_eq!(cleanup.attempted, 1);
        assert_eq!(cleanup.reclaimed, 1);
        assert!(!cleanup.has_residue());
        assert!(!cleanup.below_quorum);
        assert!(!old_path.exists(), "dereferenced old data dir must be physically removed");
    }

    // backlog#898 group B (end-to-end): a real overwrite whose post-commit old
    // data dir cleanup is forced to fail (via the test-only fault seam) must
    // still return 200 with the new ObjectInfo — no false-negative ACK.
    #[tokio::test]
    async fn put_object_overwrite_returns_ok_when_old_data_dir_cleanup_fails() {
        use crate::set_disk::core::io_primitives::cleanup_fault_injection;

        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-cleanup-fault";
        let object = "object-below-quorum-cleanup";

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

        // v1: establishes an old data dir that the overwrite will try to reclaim.
        let mut reader = PutObjReader::from_vec(b"hello".to_vec());
        set_disks
            .put_object(
                bucket,
                object,
                &mut reader,
                &ObjectOptions {
                    no_lock: true,
                    ..ObjectOptions::default()
                },
            )
            .await
            .expect("first write should succeed");

        // Force the old-data-dir cleanup delete to fail on disk 0 during the
        // overwrite. rename_data still commits; only the GC of the dereferenced
        // old dir fails, which must NOT turn the PUT into a 503.
        let _fault = cleanup_fault_injection::fail_cleanup_on(object, &[0]);

        let mut reader = PutObjReader::from_vec(b"goodbye!!".to_vec());
        let oi = set_disks
            .put_object(
                bucket,
                object,
                &mut reader,
                &ObjectOptions {
                    no_lock: true,
                    ..ObjectOptions::default()
                },
            )
            .await
            .expect("overwrite must return Ok even when old-data-dir cleanup fails");

        assert_eq!(oi.size, 9, "returned ObjectInfo must reflect the new committed write");

        // The committed new version must be readable with the new size.
        let read_back = set_disks
            .get_object_info(bucket, object, &ObjectOptions::default())
            .await
            .expect("committed object must be readable after a failed cleanup");
        assert_eq!(read_back.size, 9, "HEAD must observe the new version, not stale metadata");
    }

    // Regression for the inline-overwrite rollback backup leak: #5703 stopped
    // reporting the synthetic rollback dir for recursive post-commit cleanup,
    // which stranded `object/<synthetic>/xl.meta.bkp` after every inline
    // overwrite. The object dir then never emptied, so the s3-tests teardown
    // sequence (delete object, delete bucket) failed with BucketNotEmpty
    // forever. After a committed overwrite the backup must be reclaimed and a
    // subsequent delete must leave nothing behind.
    #[tokio::test]
    async fn inline_overwrite_reclaims_synthetic_rollback_backup() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-inline-rollback-leak";
        let object = "obj";

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

        for body in [b"hello".to_vec(), b"goodbye".to_vec()] {
            let mut reader = PutObjReader::from_vec(body);
            set_disks
                .put_object(
                    bucket,
                    object,
                    &mut reader,
                    &ObjectOptions {
                        no_lock: true,
                        ..ObjectOptions::default()
                    },
                )
                .await
                .expect("inline write should succeed");
        }

        // The committed overwrite must leave only xl.meta in the object dir on
        // every disk; a stranded rollback dir keeps the bucket undeletable.
        for endpoint in &set_disks.set_endpoints {
            let object_dir = std::path::PathBuf::from(endpoint.get_file_path()).join(bucket).join(object);
            let mut entries: Vec<String> = std::fs::read_dir(&object_dir)
                .expect("object dir should exist")
                .map(|entry| entry.expect("entry should read").file_name().to_string_lossy().into_owned())
                .collect();
            entries.sort();
            assert_eq!(
                entries,
                vec![STORAGE_FORMAT_FILE.to_string()],
                "only xl.meta may remain after an inline overwrite in {object_dir:?}"
            );
        }

        // With only xl.meta left, the s3-tests teardown (delete object, delete
        // bucket) empties the dir; the delete paths themselves are covered by
        // their own tests. This harness has no bucket metadata sys, so the
        // full delete_object flow cannot run here.
    }

    // #5703's security property must survive the backup reclamation: the
    // synthetic rollback dir of key K maps to the directory `K/<uuid>`, which
    // can simultaneously be a legitimate child key. Reclaiming the backup must
    // remove exactly the backup file — never the child key's metadata.
    #[tokio::test]
    async fn inline_overwrite_backup_reclaim_spares_child_key_dir() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-inline-rollback-child";
        let object = "obj";
        let synthetic = crate::disk::local::inline_metadata_rollback_dir(Uuid::nil(), &FileMeta::new());
        let child_object = format!("{object}/{synthetic}");

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

        let mut reader = PutObjReader::from_vec(b"child".to_vec());
        set_disks
            .put_object(
                bucket,
                &child_object,
                &mut reader,
                &ObjectOptions {
                    no_lock: true,
                    ..ObjectOptions::default()
                },
            )
            .await
            .expect("child write should succeed");

        // Create then overwrite the parent key: the overwrite writes its
        // rollback backup into the child's directory and must afterwards
        // reclaim only that file.
        for body in [b"first".to_vec(), b"second".to_vec()] {
            let mut reader = PutObjReader::from_vec(body);
            set_disks
                .put_object(
                    bucket,
                    object,
                    &mut reader,
                    &ObjectOptions {
                        no_lock: true,
                        ..ObjectOptions::default()
                    },
                )
                .await
                .expect("parent write should succeed");
        }

        let child_info = set_disks
            .get_object_info(bucket, &child_object, &ObjectOptions::default())
            .await
            .expect("child key must survive the parent's rollback backup reclamation");
        assert_eq!(child_info.size, 5, "child key content must be untouched");

        for endpoint in &set_disks.set_endpoints {
            let child_dir = std::path::PathBuf::from(endpoint.get_file_path())
                .join(bucket)
                .join(object)
                .join(synthetic.to_string());
            let mut entries: Vec<String> = std::fs::read_dir(&child_dir)
                .expect("child object dir should exist")
                .map(|entry| entry.expect("entry should read").file_name().to_string_lossy().into_owned())
                .collect();
            entries.sort();
            assert_eq!(
                entries,
                vec![STORAGE_FORMAT_FILE.to_string()],
                "the child dir must keep its xl.meta and lose only the stray backup in {child_dir:?}"
            );
        }
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn test_acquire_dist_delete_object_locks_batch_succeeds_with_two_healthy_lockers() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::DistErasure).await;

        let manager1 = Arc::new(rustfs_lock::GlobalLockManager::new());
        let manager2 = Arc::new(rustfs_lock::GlobalLockManager::new());
        let client1: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager1.clone()));
        let client2: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager2.clone()));
        let set_disks = make_test_set_disks(vec![client1, client2]).await;

        let batch = rustfs_lock::BatchLockRequest::new(set_disks.locker_owner.as_str())
            .with_all_or_nothing(false)
            .add_write_lock(ObjectKey::new("bucket", "object-a"))
            .add_write_lock(ObjectKey::new("bucket", "object-b"));

        let (failed_map, locked_objects, held_lock_ids_by_client) =
            set_disks.acquire_dist_delete_object_locks_batch(&batch).await;

        assert!(failed_map.is_empty());
        assert_eq!(locked_objects.len(), 2);
        assert!(locked_objects.contains("object-a"));
        assert!(locked_objects.contains("object-b"));
        assert_eq!(held_lock_ids_by_client.iter().map(Vec::len).sum::<usize>(), batch.requests.len() * 2);

        set_disks
            .release_dist_delete_object_locks_batch(held_lock_ids_by_client)
            .await;

        let local_lock_1 = NamespaceLock::with_local_manager("node-1".to_string(), manager1);
        let local_lock_2 = NamespaceLock::with_local_manager("node-2".to_string(), manager2);

        let guard_1 = local_lock_1
            .get_write_lock(ObjectKey::new("bucket", "object-a"), "owner-b", Duration::from_millis(100))
            .await
            .expect("released batch lock should free node 1");
        let guard_2 = local_lock_2
            .get_write_lock(ObjectKey::new("bucket", "object-b"), "owner-b", Duration::from_millis(100))
            .await
            .expect("released batch lock should free node 2");

        drop(guard_1);
        drop(guard_2);
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn test_acquire_dist_delete_object_locks_batch_rolls_back_when_quorum_not_reached() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::DistErasure).await;

        let manager = Arc::new(rustfs_lock::GlobalLockManager::new());
        let healthy_client: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager.clone()));
        let failing_client: Arc<dyn LockClient> = Arc::new(FailingClient);
        let set_disks = make_test_set_disks(vec![healthy_client, failing_client]).await;

        let batch = rustfs_lock::BatchLockRequest::new(set_disks.locker_owner.as_str())
            .with_all_or_nothing(false)
            .add_write_lock(ObjectKey::new("bucket", "object-a"));

        let (failed_map, locked_objects, held_lock_ids_by_client) =
            set_disks.acquire_dist_delete_object_locks_batch(&batch).await;

        assert!(locked_objects.is_empty());
        assert!(failed_map.contains_key(&("bucket".to_string(), "object-a".to_string())));
        assert_eq!(held_lock_ids_by_client.iter().map(Vec::len).sum::<usize>(), 0);

        let local_lock = NamespaceLock::with_local_manager("node-1".to_string(), manager);
        let guard = local_lock
            .get_write_lock(ObjectKey::new("bucket", "object-a"), "owner-b", Duration::from_millis(100))
            .await
            .expect("quorum rollback should release the healthy node lock");

        drop(guard);
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn test_acquire_dist_delete_object_locks_batch_returns_after_quorum_without_waiting_for_slow_lockers() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::DistErasure).await;

        let manager_fast_1 = Arc::new(rustfs_lock::GlobalLockManager::new());
        let manager_fast_2 = Arc::new(rustfs_lock::GlobalLockManager::new());
        let manager_fast_3 = Arc::new(rustfs_lock::GlobalLockManager::new());
        let manager_slow = Arc::new(rustfs_lock::GlobalLockManager::new());

        let client_fast_1: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager_fast_1));
        let client_fast_2: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager_fast_2));
        let client_fast_3: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager_fast_3));
        let client_slow: Arc<dyn LockClient> = Arc::new(DelayedBatchClient {
            inner: Arc::new(LocalClient::with_manager(manager_slow.clone())),
            delay: Duration::from_millis(250),
        });

        let set_disks = make_test_set_disks(vec![client_fast_1, client_fast_2, client_fast_3, client_slow]).await;

        let batch = rustfs_lock::BatchLockRequest::new(set_disks.locker_owner.as_str())
            .with_all_or_nothing(false)
            .add_write_lock(ObjectKey::new("bucket", "object-a"))
            .add_write_lock(ObjectKey::new("bucket", "object-b"));

        let started = Instant::now();
        let (failed_map, locked_objects, held_lock_ids_by_client) =
            set_disks.acquire_dist_delete_object_locks_batch(&batch).await;

        assert!(
            started.elapsed() < Duration::from_millis(150),
            "batch distributed delete locks should return once quorum is satisfied"
        );
        assert!(failed_map.is_empty());
        assert_eq!(locked_objects.len(), 2);

        set_disks
            .release_dist_delete_object_locks_batch(held_lock_ids_by_client)
            .await;

        tokio::time::sleep(Duration::from_millis(350)).await;

        let slow_lock = NamespaceLock::with_local_manager("slow-node".to_string(), manager_slow);
        let guard_a = slow_lock
            .get_write_lock(ObjectKey::new("bucket", "object-a"), "owner-b", Duration::from_millis(100))
            .await
            .expect("late successful batch lock should be cleaned up for object-a");
        let guard_b = slow_lock
            .get_write_lock(ObjectKey::new("bucket", "object-b"), "owner-b", Duration::from_millis(100))
            .await
            .expect("late successful batch lock should be cleaned up for object-b");

        drop(guard_a);
        drop(guard_b);
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn test_acquire_dist_delete_object_locks_batch_fails_early_and_cleans_up_late_successes() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::DistErasure).await;

        let manager_fast = Arc::new(rustfs_lock::GlobalLockManager::new());
        let manager_slow = Arc::new(rustfs_lock::GlobalLockManager::new());

        let client_fast: Arc<dyn LockClient> = Arc::new(LocalClient::with_manager(manager_fast));
        let client_fail_1: Arc<dyn LockClient> = Arc::new(FailingClient);
        let client_fail_2: Arc<dyn LockClient> = Arc::new(FailingClient);
        let client_slow: Arc<dyn LockClient> = Arc::new(DelayedBatchClient {
            inner: Arc::new(LocalClient::with_manager(manager_slow.clone())),
            delay: Duration::from_millis(250),
        });

        let set_disks = make_test_set_disks(vec![client_fast, client_fail_1, client_fail_2, client_slow]).await;
        let batch = rustfs_lock::BatchLockRequest::new(set_disks.locker_owner.as_str())
            .with_all_or_nothing(false)
            .add_write_lock(ObjectKey::new("bucket", "object-a"))
            .add_write_lock(ObjectKey::new("bucket", "object-b"));

        let started = Instant::now();
        let (failed_map, locked_objects, held_lock_ids_by_client) =
            set_disks.acquire_dist_delete_object_locks_batch(&batch).await;

        assert!(
            started.elapsed() < Duration::from_millis(150),
            "batch distributed delete locks should fail as soon as quorum becomes impossible"
        );
        assert!(locked_objects.is_empty());
        assert!(failed_map.contains_key(&("bucket".to_string(), "object-a".to_string())));
        assert!(failed_map.contains_key(&("bucket".to_string(), "object-b".to_string())));
        assert_eq!(held_lock_ids_by_client.iter().map(Vec::len).sum::<usize>(), 0);

        tokio::time::sleep(Duration::from_millis(350)).await;

        let slow_lock = NamespaceLock::with_local_manager("slow-node".to_string(), manager_slow);
        let guard_a = slow_lock
            .get_write_lock(ObjectKey::new("bucket", "object-a"), "owner-b", Duration::from_millis(100))
            .await
            .expect("late successful batch failure cleanup should release object-a");
        let guard_b = slow_lock
            .get_write_lock(ObjectKey::new("bucket", "object-b"), "owner-b", Duration::from_millis(100))
            .await
            .expect("late successful batch failure cleanup should release object-b");

        drop(guard_a);
        drop(guard_b);
    }

    #[test]
    fn test_common_parity() {
        // Test common parity calculation
        // For parities [2, 2, 2, 3] with n=4, default_parity_count=1:
        // - parity=2: read_quorum = 4-2 = 2, occ=3 >= 2, so valid
        // - parity=3: read_quorum = 4-3 = 1, occ=1 >= 1, so valid
        // - max_occ=3 for parity=2, so returns 2
        let parities = vec![2, 2, 2, 3];
        assert_eq!(SetDisks::common_parity(&parities, 1), 2);

        // For parities [1, 2, 3] with n=3, default_parity_count=2:
        // - parity=1: read_quorum = 3-1 = 2, occ=1 < 2, so invalid
        // - parity=2: read_quorum = 3-2 = 1, occ=1 >= 1, so valid
        // - parity=3: read_quorum = 3-3 = 0, occ=1 >= 0, so valid
        // - max_occ=1, both parity=2 and parity=3 have same occurrence
        // - HashMap iteration order is not guaranteed, so result could be either 2 or 3
        let parities = vec![1, 2, 3];
        let result = SetDisks::common_parity(&parities, 2);
        assert!(result == 2 || result == 3); // Either 2 or 3 is valid

        let empty_parities = vec![];
        assert_eq!(SetDisks::common_parity(&empty_parities, 3), -1); // Empty returns -1

        let invalid_parities = vec![-1, -1, -1];
        assert_eq!(SetDisks::common_parity(&invalid_parities, 2), -1); // all invalid

        let single_parity = vec![4];
        assert_eq!(SetDisks::common_parity(&single_parity, 1), 4);

        // Test with -1 values (ignored)
        let parities_with_invalid = vec![-1, 2, 2, -1];
        assert_eq!(SetDisks::common_parity(&parities_with_invalid, 1), 2);
    }

    #[test]
    fn test_common_time() {
        // Test common time calculation
        let now = OffsetDateTime::now_utc();
        let later = now + Duration::from_secs(60);

        let times = vec![Some(now), Some(now), Some(later)];
        assert_eq!(SetDisks::common_time(&times, 2), Some(now));

        let times2 = vec![Some(now), Some(later), Some(later)];
        assert_eq!(SetDisks::common_time(&times2, 2), Some(later));

        let times_with_none = vec![Some(now), None, Some(now)];
        assert_eq!(SetDisks::common_time(&times_with_none, 2), Some(now));

        let times = vec![None, None, None];
        assert_eq!(SetDisks::common_time(&times, 2), None);

        let empty_times = vec![];
        assert_eq!(SetDisks::common_time(&empty_times, 1), None);
    }

    #[test]
    fn test_common_time_and_occurrence() {
        // Test common time with occurrence count
        let now = OffsetDateTime::now_utc();
        let times = vec![Some(now), Some(now), None];
        let (time, count) = SetDisks::common_time_and_occurrence(&times);
        assert_eq!(time, Some(now));
        assert_eq!(count, 2);

        let times = vec![None, None, None];
        let (time, count) = SetDisks::common_time_and_occurrence(&times);
        assert_eq!(time, None);
        assert_eq!(count, 0); // No valid times, so count is 0
    }

    #[test]
    fn test_common_etag() {
        // Test common etag calculation
        let etags = vec![Some("etag1".to_string()), Some("etag1".to_string()), None];
        assert_eq!(SetDisks::common_etag(&etags, 2), Some("etag1".to_string()));

        let etags = vec![None, None, None];
        assert_eq!(SetDisks::common_etag(&etags, 2), None);
    }

    #[test]
    fn test_common_etags() {
        // Test common etags with occurrence count
        let etags = vec![Some("etag1".to_string()), Some("etag1".to_string()), None];
        let (etag, count) = SetDisks::common_etags(&etags);
        assert_eq!(etag, Some("etag1".to_string()));
        assert_eq!(count, 2);
    }

    #[test]
    fn test_list_object_modtimes() {
        // Test extracting modification times from file info
        let now = OffsetDateTime::now_utc();
        let file_info = FileInfo {
            mod_time: Some(now),
            ..Default::default()
        };
        let parts_metadata = vec![file_info];
        let errs = vec![None];

        let modtimes = SetDisks::list_object_modtimes(&parts_metadata, &errs);
        assert_eq!(modtimes.len(), 1);
        assert_eq!(modtimes[0], Some(now));
    }

    #[test]
    fn test_list_object_etags() {
        // Test extracting etags from file info metadata
        let mut metadata = HashMap::new();
        metadata.insert("etag".to_string(), "test-etag".to_string());

        let file_info = FileInfo {
            metadata,
            ..Default::default()
        };
        let parts_metadata = vec![file_info];
        let errs = vec![None];

        let etags = SetDisks::list_object_etags(&parts_metadata, &errs);
        assert_eq!(etags.len(), 1);
        assert_eq!(etags[0], Some("test-etag".to_string()));
    }

    fn quorum_test_fileinfo(mod_time: OffsetDateTime, data_dir: Uuid, part_etag: &str, erasure_index: usize) -> FileInfo {
        let mut metadata = HashMap::new();
        metadata.insert("etag".to_string(), "object-etag".to_string());

        FileInfo {
            name: "bucket/object".to_string(),
            size: 8 * 1024 * 1024,
            mod_time: Some(mod_time),
            data_dir: Some(data_dir),
            metadata,
            parts: vec![ObjectPartInfo {
                etag: part_etag.to_string(),
                number: 1,
                size: 8 * 1024 * 1024,
                actual_size: 8 * 1024 * 1024,
                mod_time: Some(mod_time),
                ..Default::default()
            }],
            erasure: ErasureInfo {
                data_blocks: 2,
                parity_blocks: 2,
                block_size: 4 * 1024 * 1024,
                index: erasure_index,
                distribution: vec![1, 2, 3, 4],
                ..Default::default()
            },
            ..Default::default()
        }
    }

    fn decoded_quorum_test_fileinfo_with_metadata(
        mod_time: OffsetDateTime,
        data_dir: Uuid,
        part_etag: &str,
        erasure_index: usize,
        extra_metadata: &[(&str, &str)],
    ) -> FileInfo {
        let mut fi = quorum_test_fileinfo(mod_time, data_dir, part_etag, erasure_index);
        for (name, value) in extra_metadata {
            fi.metadata.insert((*name).to_string(), (*value).to_string());
        }

        let mut meta = FileMeta::new();
        meta.add_version(fi).expect("test file metadata should accept object version");
        let encoded = meta.marshal_msg().expect("test file metadata should marshal");
        rustfs_filemeta::get_file_info(
            &encoded,
            "bucket",
            "object",
            "",
            rustfs_filemeta::FileInfoOpts {
                data: false,
                include_free_versions: false,
                include_part_checksums: true,
            },
        )
        .expect("test file metadata should decode as file info")
    }

    #[test]
    fn test_find_file_info_in_quorum_uses_part_identity() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let metas = vec![
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 1),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 2),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 3),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-b", 4),
        ];

        let fi = SetDisks::find_file_info_in_quorum(&metas, &Some(mod_time), &None, 3)
            .expect("three matching part identities should reach quorum");

        assert_eq!(fi.parts[0].etag, "part-etag-a");
    }

    #[test]
    fn test_find_file_info_in_quorum_rejects_split_part_identity() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let metas = vec![
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 1),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 2),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-b", 3),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-b", 4),
        ];

        let err = SetDisks::find_file_info_in_quorum(&metas, &Some(mod_time), &None, 3)
            .expect_err("split part identities must not reach write quorum");

        assert_eq!(err, DiskError::ErasureReadQuorum);
    }

    #[test]
    fn test_latest_fileinfo_selection_quorum_requires_write_quorum_when_full_metadata_is_available() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let metas = vec![
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 1),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 2),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-b", 3),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-b", 4),
        ];
        let errs = vec![None, None, None, None];

        let quorum = SetDisks::latest_fileinfo_selection_quorum("", &metas, &errs, 2, 3);

        assert_eq!(quorum, 3);
    }

    #[test]
    fn test_latest_fileinfo_selection_quorum_preserves_read_quorum_for_version_or_degraded_reads() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let metas = vec![
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 1),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 2),
            FileInfo::default(),
            FileInfo::default(),
        ];
        let degraded_errs = vec![None, None, Some(DiskError::DiskNotFound), Some(DiskError::DiskNotFound)];
        let clean_errs = vec![None, None, None, None];

        assert_eq!(SetDisks::latest_fileinfo_selection_quorum("", &metas, &degraded_errs, 2, 3), 2);
        assert_eq!(SetDisks::latest_fileinfo_selection_quorum("version-id", &metas, &clean_errs, 2, 3), 2);
    }

    #[test]
    fn test_latest_fileinfo_selection_quorum_keeps_read_quorum_for_partial_overwrite_with_read_error() {
        let old_mod_time = OffsetDateTime::now_utc();
        let new_mod_time = old_mod_time + time::Duration::seconds(1);
        let old_data_dir = Uuid::new_v4();
        let new_data_dir = Uuid::new_v4();
        let metas = vec![
            quorum_test_fileinfo(old_mod_time, old_data_dir, "part-etag-old", 1),
            quorum_test_fileinfo(old_mod_time, old_data_dir, "part-etag-old", 2),
            quorum_test_fileinfo(new_mod_time, new_data_dir, "part-etag-new", 3),
            FileInfo::default(),
        ];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];

        let quorum = SetDisks::latest_fileinfo_selection_quorum("", &metas, &errs, 2, 3);
        let (online_disks, mod_time, etag) = SetDisks::list_online_disks(&vec![None; metas.len()], &metas, &errs, quorum);
        let fi = SetDisks::pick_valid_fileinfo(&metas, mod_time, etag, quorum)
            .expect("old metadata should remain readable with read quorum");

        assert_eq!(quorum, 2);
        assert_eq!(online_disks.len(), metas.len());
        assert_eq!(fi.data_dir, Some(old_data_dir));
        assert_eq!(fi.parts[0].etag, "part-etag-old");

        let (_, selected, selected_quorum) = SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3)
            .expect("old metadata should remain selectable with read quorum");
        assert_eq!(selected_quorum, 2);
        assert_eq!(selected.data_dir, Some(old_data_dir));
        assert_eq!(selected.parts[0].etag, "part-etag-old");
        assert!(selected.is_latest);
    }

    #[test]
    fn test_latest_fileinfo_selection_rejects_partial_latest_read_quorum_with_read_error() {
        let old_mod_time = OffsetDateTime::now_utc();
        let new_mod_time = old_mod_time + time::Duration::seconds(1);
        let old_data_dir = Uuid::new_v4();
        let new_data_dir = Uuid::new_v4();
        let metas = vec![
            quorum_test_fileinfo(new_mod_time, new_data_dir, "part-etag-new", 1),
            quorum_test_fileinfo(new_mod_time, new_data_dir, "part-etag-new", 2),
            quorum_test_fileinfo(old_mod_time, old_data_dir, "part-etag-old", 3),
            FileInfo::default(),
        ];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];

        let result = SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3);

        assert!(matches!(result, Err(DiskError::ErasureReadQuorum)));
    }

    #[test]
    fn test_latest_fileinfo_selection_preserves_degraded_read_quorum_without_competing_latest() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let mut first = quorum_test_fileinfo(mod_time, data_dir, "part-etag-old", 1);
        rustfs_utils::http::insert_str(
            &mut first.metadata,
            rustfs_utils::http::SUFFIX_PART_CHECKSUMS,
            r#"[[1,[["CRC32C","AAAAAA=="]]]]"#.to_string(),
        );
        let mut second = first.clone();
        second.erasure.index = 2;
        let metas = vec![first, second, FileInfo::default(), FileInfo::default()];
        let errs = vec![None, None, Some(DiskError::DiskNotFound), Some(DiskError::DiskNotFound)];

        let (_, mut selected, selected_quorum) =
            SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3)
                .expect("read quorum should remain enough when no competing latest is visible");

        assert_eq!(selected_quorum, 2);
        assert_eq!(selected.data_dir, Some(data_dir));
        assert_eq!(selected.parts[0].etag, "part-etag-old");
        assert!(selected.parts[0].checksums.is_none());
        SetDisks::hydrate_selected_fileinfo_part_checksums(&mut selected)
            .expect("requested part checksums should hydrate after winner selection");
        assert_eq!(
            selected.parts[0]
                .checksums
                .as_ref()
                .and_then(|checksums| checksums.get("CRC32C"))
                .map(String::as_str),
            Some("AAAAAA==")
        );
    }

    #[test]
    fn test_degraded_fileinfo_selection_rejects_malformed_part_checksum_metadata() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let mut first = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 1);
        rustfs_utils::http::insert_str(&mut first.metadata, rustfs_utils::http::SUFFIX_PART_CHECKSUMS, "not-json".to_string());
        let mut second = first.clone();
        second.erasure.index = 2;
        let metas = vec![first, second, FileInfo::default(), FileInfo::default()];
        let errs = vec![None, None, Some(DiskError::DiskNotFound), Some(DiskError::DiskNotFound)];

        let (_, mut selected, _) = SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3)
            .expect("winner selection should defer sidecar decoding");
        let err = SetDisks::hydrate_selected_fileinfo_part_checksums(&mut selected)
            .expect_err("a malformed degraded winner must fail closed when checksums are requested");

        assert_eq!(err, DiskError::FileCorrupt);
    }

    #[test]
    fn test_pick_valid_fileinfo_rejects_malformed_part_checksum_metadata() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let mut meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 1);
        rustfs_utils::http::insert_str(&mut meta.metadata, rustfs_utils::http::SUFFIX_PART_CHECKSUMS, "not-json".to_string());
        let mut second = meta.clone();
        second.erasure.index = 2;

        let mut selected = SetDisks::pick_valid_fileinfo(&[meta, second], Some(mod_time), None, 2)
            .expect("winner selection should defer sidecar decoding");
        let err = SetDisks::hydrate_selected_fileinfo_part_checksums(&mut selected)
            .expect_err("a malformed winning part-checksum sidecar must fail closed when checksums are requested");

        assert_eq!(err, DiskError::FileCorrupt);
    }

    #[test]
    fn test_part_checksum_hydration_rejects_invalid_algorithm_and_value() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        for encoded in [
            r#"[[1,[["UNKNOWN","AAAAAA=="]]]]"#,
            r#"[[1,[["CRC32C","not-base64"]]]]"#,
            r#"[[1,[["CRC32C","AA=="]]]]"#,
            r#"[[1,[["CRC32C","AAAAAA==-0"]]]]"#,
            r#"[[1,[["CRC32C","AAAAAA==-1"]]]]"#,
            r#"[[1,[["CRC32C","AAAAAA=="],["crc32c","BBBBBB=="]]]]"#,
        ] {
            let mut meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 1);
            rustfs_utils::http::insert_str(&mut meta.metadata, rustfs_utils::http::SUFFIX_PART_CHECKSUMS, encoded.to_string());

            let err = SetDisks::hydrate_selected_fileinfo_part_checksums(&mut meta)
                .expect_err("invalid persisted part checksum metadata must fail closed");
            assert_eq!(err, DiskError::FileCorrupt);
        }
    }

    #[test]
    fn test_latest_fileinfo_selection_ignores_derived_version_stack_drift() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let mut latest_meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 1);
        latest_meta.is_latest = true;
        latest_meta.num_versions = 1;

        let mut stale_stack_meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 2);
        stale_stack_meta.is_latest = false;
        stale_stack_meta.successor_mod_time = Some(mod_time + time::Duration::seconds(1));
        stale_stack_meta.num_versions = 2;

        let mut newer_stack_meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 3);
        newer_stack_meta.is_latest = false;
        newer_stack_meta.successor_mod_time = Some(mod_time + time::Duration::seconds(2));
        newer_stack_meta.num_versions = 3;

        let metas = vec![latest_meta, stale_stack_meta, newer_stack_meta, FileInfo::default()];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];

        let (_, selected, selected_quorum) = SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3)
            .expect("same object version should stay readable despite derived version stack drift");

        assert_eq!(selected_quorum, 3);
        assert_eq!(selected.data_dir, Some(data_dir));
        assert_eq!(selected.parts[0].etag, "part-etag");
        assert_eq!(selected.mod_time, Some(mod_time));
    }

    #[test]
    fn test_latest_fileinfo_selection_uses_successor_mod_time_quorum_for_latest_flag() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let mut stale_stack_meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 1);
        stale_stack_meta.is_latest = false;
        stale_stack_meta.successor_mod_time = Some(mod_time + time::Duration::seconds(1));
        stale_stack_meta.num_versions = 2;

        let mut latest_meta_a = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 2);
        latest_meta_a.is_latest = true;
        latest_meta_a.num_versions = 1;
        let mut latest_meta_b = latest_meta_a.clone();
        latest_meta_b.erasure.index = 3;

        let metas = vec![stale_stack_meta, latest_meta_a, latest_meta_b];

        let selected = SetDisks::find_file_info_in_quorum(&metas, &Some(mod_time), &None, 2)
            .expect("latest flag should be derived from successor mod time quorum");

        assert!(selected.is_latest);
        assert_eq!(selected.successor_mod_time, None);
        assert_eq!(selected.num_versions, 1);
        assert_eq!(selected.mod_time, Some(mod_time));
    }

    #[test]
    fn test_latest_fileinfo_selection_ignores_replication_state_drift() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let replication_status_key = format!(
            "{}{}",
            rustfs_utils::http::RUSTFS_INTERNAL_PREFIX,
            rustfs_utils::http::SUFFIX_REPLICATION_STATUS
        );
        let replication_timestamp_key = format!(
            "{}{}",
            rustfs_utils::http::RUSTFS_INTERNAL_PREFIX,
            rustfs_utils::http::SUFFIX_REPLICATION_TIMESTAMP
        );
        let replication_reset_key = format!(
            "{}{}target-a",
            rustfs_utils::http::RUSTFS_INTERNAL_PREFIX,
            rustfs_utils::http::SUFFIX_REPLICATION_RESET_ARN_PREFIX
        );
        let meta_a = decoded_quorum_test_fileinfo_with_metadata(
            mod_time,
            data_dir,
            "part-etag",
            1,
            &[
                (&replication_status_key, "target-a=COMPLETED;"),
                (&replication_timestamp_key, "2024-01-01T00:00:00Z"),
                (&replication_reset_key, "COMPLETED"),
            ],
        );
        let meta_b = decoded_quorum_test_fileinfo_with_metadata(
            mod_time,
            data_dir,
            "part-etag",
            2,
            &[
                (&replication_status_key, "target-a=PENDING;"),
                (&replication_timestamp_key, "2024-01-01T00:00:01Z"),
                (&replication_reset_key, "PENDING"),
            ],
        );
        let meta_c = decoded_quorum_test_fileinfo_with_metadata(
            mod_time,
            data_dir,
            "part-etag",
            3,
            &[
                (&replication_status_key, "target-a=FAILED;"),
                (&replication_timestamp_key, "2024-01-01T00:00:02Z"),
                (&replication_reset_key, "FAILED"),
            ],
        );
        assert!(meta_a.replication_state_internal.is_some());
        assert_eq!(
            meta_a.metadata.get(metadata_keys::REPLICATION_STATUS).map(String::as_str),
            Some("COMPLETED")
        );

        let metas = vec![meta_a, meta_b, meta_c, FileInfo::default()];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];

        let (_, selected, selected_quorum) = SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3)
            .expect("replication status drift should not split readable object identity");

        assert_eq!(selected_quorum, 3);
        assert_eq!(selected.data_dir, Some(data_dir));
        assert_eq!(selected.parts[0].etag, "part-etag");
    }

    #[test]
    fn test_latest_fileinfo_selection_rejects_same_modtime_metadata_split_without_write_quorum() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let mut old_meta_a = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 1);
        let mut old_meta_b = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 2);
        let mut partial_meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 3);
        old_meta_a.metadata.insert("x-amz-meta-color".to_string(), "blue".to_string());
        old_meta_b.metadata.insert("x-amz-meta-color".to_string(), "blue".to_string());
        partial_meta
            .metadata
            .insert("x-amz-meta-color".to_string(), "red".to_string());
        let metas = vec![old_meta_a, old_meta_b, partial_meta, FileInfo::default()];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];

        let quorum = SetDisks::latest_fileinfo_selection_quorum("", &metas, &errs, 2, 3);
        let result = SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3);

        assert_eq!(quorum, 2);
        assert!(matches!(result, Err(DiskError::ErasureReadQuorum)));
    }

    #[test]
    fn test_latest_fileinfo_selection_rejects_same_modtime_partial_metadata_read_quorum() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let mut old_meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 1);
        let mut partial_meta_a = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 2);
        let mut partial_meta_b = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 3);
        old_meta.metadata.insert("x-amz-meta-color".to_string(), "blue".to_string());
        partial_meta_a
            .metadata
            .insert("x-amz-meta-color".to_string(), "red".to_string());
        partial_meta_b
            .metadata
            .insert("x-amz-meta-color".to_string(), "red".to_string());
        let metas = vec![old_meta, partial_meta_a, partial_meta_b, FileInfo::default()];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];

        let result = SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3);

        assert!(matches!(result, Err(DiskError::ErasureReadQuorum)));
    }

    #[test]
    fn test_latest_fileinfo_selection_rejects_same_modtime_transition_split_without_write_quorum() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let old_meta_a = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 1);
        let old_meta_b = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 2);
        let mut partial_meta = quorum_test_fileinfo(mod_time, data_dir, "part-etag", 3);
        partial_meta.transition_status = TRANSITION_COMPLETE.to_string();
        partial_meta.transition_tier = "WARM".to_string();
        partial_meta.transitioned_objname = "remote/object".to_string();
        partial_meta.transition_version_id = Some(Uuid::new_v4());
        let metas = vec![old_meta_a, old_meta_b, partial_meta, FileInfo::default()];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];

        let quorum = SetDisks::latest_fileinfo_selection_quorum("", &metas, &errs, 2, 3);
        let result = SetDisks::select_valid_fileinfo(&vec![None; metas.len()], &metas, &errs, "", 2, 3);

        assert_eq!(quorum, 2);
        assert!(matches!(result, Err(DiskError::ErasureReadQuorum)));
    }

    #[test]
    fn test_latest_fileinfo_selection_quorum_uses_write_quorum_for_degraded_committed_identity() {
        let mod_time = OffsetDateTime::now_utc();
        let data_dir = Uuid::new_v4();
        let metas = vec![
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 1),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 2),
            quorum_test_fileinfo(mod_time, data_dir, "part-etag-a", 3),
            FileInfo::default(),
        ];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];

        assert_eq!(SetDisks::latest_fileinfo_selection_quorum("", &metas, &errs, 2, 3), 3);
    }

    #[test]
    fn test_list_object_parities() {
        // Test extracting parity counts from file info
        let file_info1 = FileInfo {
            erasure: ErasureInfo {
                data_blocks: 4,
                parity_blocks: 2,
                block_size: 4,
                index: 1,                             // Must be > 0 for is_valid() to return true
                distribution: vec![1, 2, 3, 4, 5, 6], // Must match data_blocks + parity_blocks
                ..Default::default()
            },
            size: 100, // Non-zero size
            deleted: false,
            ..Default::default()
        };
        let file_info2 = FileInfo {
            erasure: ErasureInfo {
                data_blocks: 6,
                parity_blocks: 3,
                block_size: 4,
                index: 1,                                      // Must be > 0 for is_valid() to return true
                distribution: vec![1, 2, 3, 4, 5, 6, 7, 8, 9], // Must match data_blocks + parity_blocks
                ..Default::default()
            },
            size: 200, // Non-zero size
            deleted: false,
            ..Default::default()
        };
        let file_info3 = FileInfo {
            erasure: ErasureInfo {
                data_blocks: 2,
                parity_blocks: 1,
                block_size: 4,
                index: 1,                    // Must be > 0 for is_valid() to return true
                distribution: vec![1, 2, 3], // Must match data_blocks + parity_blocks
                ..Default::default()
            },
            size: 0, // Zero size - function returns half of total shards
            deleted: false,
            ..Default::default()
        };

        let parts_metadata = vec![file_info1, file_info2, file_info3];
        let errs = vec![None, None, None];

        let parities = SetDisks::list_object_parities(&parts_metadata, &errs);
        assert_eq!(parities.len(), 3);
        assert_eq!(parities[0], 2); // parity_blocks from first file
        assert_eq!(parities[1], 3); // parity_blocks from second file
        assert_eq!(parities[2], 1); // half of total shards (3/2 = 1) for zero size file
    }

    #[test]
    fn delete_markers_participate_in_four_disk_metadata_quorum_without_erasure_geometry() {
        let marker = FileInfo {
            name: "bucket/deleted".to_string(),
            deleted: true,
            version_id: Some(Uuid::new_v4()),
            mod_time: Some(OffsetDateTime::now_utc()),
            ..Default::default()
        };
        let parts_metadata = vec![marker; 4];
        let errs = vec![None; 4];

        assert!(parts_metadata.iter().all(file_info_is_valid_for_metadata));
        assert!(parts_metadata.iter().all(|metadata| !metadata.is_valid()));
        assert_eq!(SetDisks::list_object_parities(&parts_metadata, &errs), vec![2; 4]);
        assert_eq!(
            SetDisks::object_quorum_from_meta(&parts_metadata, &errs, 2)
                .expect("four matching delete markers must reach metadata quorum"),
            (2, 3)
        );
    }

    #[test]
    fn metadata_boundary_does_not_relax_non_delete_or_malformed_delete_metadata() {
        assert!(!file_info_is_valid_for_metadata(&FileInfo::default()));

        let mut transitioned = FileInfo::new("bucket/transitioned", 2, 2);
        transitioned.erasure.index = 1;
        transitioned.transition_status = TRANSITION_COMPLETE.to_string();
        assert!(file_info_is_valid_for_metadata(&transitioned));

        transitioned.erasure = ErasureInfo::default();
        assert!(
            !file_info_is_valid_for_metadata(&transitioned),
            "transition state must not relax local erasure validation"
        );

        let mut purge_pending = FileInfo::new("bucket/purge-pending", 2, 2);
        purge_pending.erasure.index = 1;
        purge_pending.deleted = true;
        purge_pending.parts.push(ObjectPartInfo {
            number: 1,
            ..Default::default()
        });
        assert!(
            file_info_is_valid_for_metadata(&purge_pending),
            "purge-pending payload metadata must retain its valid erasure vote"
        );
        assert!(!purge_pending.is_canonical_delete_marker());

        let mut malformed_marker = FileInfo {
            deleted: true,
            ..Default::default()
        };
        malformed_marker.parts = vec![
            ObjectPartInfo {
                number: 1,
                ..Default::default()
            },
            ObjectPartInfo {
                number: 1,
                ..Default::default()
            },
        ];
        assert!(!file_info_is_valid_for_metadata(&malformed_marker));
    }

    #[test]
    fn purge_pending_payload_uses_its_erasure_parity_for_metadata_quorum() {
        let version_id = Uuid::new_v4();
        let mod_time = OffsetDateTime::now_utc();
        let parts_metadata = (1..=6)
            .map(|disk_index| {
                let mut purge_pending = FileInfo::new("bucket/purge-pending", 5, 1);
                purge_pending.name = "bucket/purge-pending".to_string();
                purge_pending.version_id = Some(version_id);
                purge_pending.mod_time = Some(mod_time);
                purge_pending.size = 1;
                purge_pending.deleted = true;
                purge_pending.erasure.index = disk_index;
                purge_pending.add_object_part(1, "part-etag-1".to_string(), 1, None, 1, None, None);
                purge_pending
            })
            .collect::<Vec<_>>();
        let errs = vec![None; 6];

        assert!(parts_metadata.iter().all(file_info_is_valid_for_metadata));
        assert!(parts_metadata.iter().all(|metadata| !metadata.is_canonical_delete_marker()));
        assert_eq!(SetDisks::list_object_parities(&parts_metadata, &errs), vec![1; 6]);
        assert_eq!(
            SetDisks::object_quorum_from_meta(&parts_metadata, &errs, 3)
                .expect("purge-pending payload should retain its EC:1 quorum"),
            (5, 5)
        );
    }

    #[test]
    fn test_conv_part_err_to_int() {
        // Test error conversion to integer codes
        assert_eq!(conv_part_err_to_int(&None), CHECK_PART_SUCCESS);

        let disk_err = DiskError::FileNotFound;
        assert_eq!(conv_part_err_to_int(&Some(disk_err)), CHECK_PART_FILE_NOT_FOUND);

        let other_err = DiskError::other("other error");
        assert_eq!(conv_part_err_to_int(&Some(other_err)), CHECK_PART_UNKNOWN); // Other errors should return UNKNOWN, not SUCCESS
    }

    #[test]
    fn deep_scan_results_accept_only_complete_verified_or_repairable_parts() {
        for (results, expected_parts) in [
            (vec![], 0),
            (vec![CHECK_PART_SUCCESS], 1),
            (vec![CHECK_PART_FILE_NOT_FOUND], 1),
            (vec![CHECK_PART_FILE_CORRUPT], 1),
            (vec![CHECK_PART_SUCCESS, CHECK_PART_FILE_NOT_FOUND, CHECK_PART_FILE_CORRUPT], 3),
        ] {
            validate_deep_scan_results(&results, expected_parts)
                .expect("complete results must allow healthy or repairable parts");
        }
    }

    #[test]
    fn deep_scan_results_reject_unknown_and_incomplete_observations() {
        for (results, expected_parts) in [
            (vec![CHECK_PART_UNKNOWN], 1),
            (vec![usize::MAX], 1),
            (vec![CHECK_PART_SUCCESS, CHECK_PART_UNKNOWN], 2),
            (vec![], 1),
            (vec![CHECK_PART_SUCCESS], 2),
            (vec![CHECK_PART_SUCCESS, CHECK_PART_SUCCESS], 1),
            (vec![CHECK_PART_SUCCESS], 0),
        ] {
            let error =
                validate_deep_scan_results(&results, expected_parts).expect_err("unverified parts must not become healthy");
            assert!(matches!(error, DiskError::Io(_)), "an incomplete observation is not proven corruption");
            assert!(error.to_string().contains("incomplete deep scan result"));
        }
    }

    #[test]
    fn deep_scan_results_preserve_disk_and_volume_failures() {
        for (status, expected_error) in [
            (CHECK_PART_DISK_NOT_FOUND, DiskError::DiskNotFound),
            (CHECK_PART_VOLUME_NOT_FOUND, DiskError::VolumeNotFound),
        ] {
            let error = validate_deep_scan_results(&[CHECK_PART_SUCCESS, status], 2)
                .expect_err("an unavailable part cannot certify a healthy disk");
            assert_eq!(error, expected_error);
        }
    }

    #[test]
    fn test_has_part_err() {
        // Test checking for part errors
        let no_errors = vec![CHECK_PART_SUCCESS, CHECK_PART_SUCCESS];
        assert!(!has_part_err(&no_errors));

        let with_errors = vec![CHECK_PART_SUCCESS, CHECK_PART_FILE_NOT_FOUND];
        assert!(has_part_err(&with_errors));

        let unknown_errors = vec![CHECK_PART_UNKNOWN, CHECK_PART_SUCCESS];
        assert!(has_part_err(&unknown_errors));
    }

    #[test]
    fn test_populate_data_errs_by_disk_uses_disk_index_not_error_code() {
        let mut data_errs_by_disk = HashMap::from([
            (0, vec![CHECK_PART_UNKNOWN, CHECK_PART_UNKNOWN]),
            (1, vec![CHECK_PART_UNKNOWN, CHECK_PART_UNKNOWN]),
            (2, vec![CHECK_PART_UNKNOWN, CHECK_PART_UNKNOWN]),
        ]);
        let data_errs_by_part = HashMap::from([
            (0, vec![CHECK_PART_FILE_NOT_FOUND, CHECK_PART_SUCCESS, CHECK_PART_SUCCESS]),
            (1, vec![CHECK_PART_SUCCESS, CHECK_PART_FILE_CORRUPT, CHECK_PART_SUCCESS]),
        ]);

        populate_data_errs_by_disk(&mut data_errs_by_disk, &data_errs_by_part);

        assert_eq!(data_errs_by_disk.get(&0).unwrap(), &vec![CHECK_PART_FILE_NOT_FOUND, CHECK_PART_SUCCESS]);
        assert_eq!(data_errs_by_disk.get(&1).unwrap(), &vec![CHECK_PART_SUCCESS, CHECK_PART_FILE_CORRUPT]);
        assert_eq!(data_errs_by_disk.get(&2).unwrap(), &vec![CHECK_PART_SUCCESS, CHECK_PART_SUCCESS]);

        let mut data_errs_by_disk = HashMap::from([
            (0, vec![CHECK_PART_UNKNOWN, CHECK_PART_UNKNOWN]),
            (1, vec![CHECK_PART_UNKNOWN, CHECK_PART_UNKNOWN]),
            (2, vec![CHECK_PART_UNKNOWN, CHECK_PART_UNKNOWN]),
            (3, vec![CHECK_PART_UNKNOWN, CHECK_PART_UNKNOWN]),
        ]);
        let data_errs_by_part = HashMap::from([
            (
                0,
                vec![
                    CHECK_PART_FILE_NOT_FOUND,
                    CHECK_PART_SUCCESS,
                    CHECK_PART_SUCCESS,
                    CHECK_PART_SUCCESS,
                ],
            ),
            (
                1,
                vec![
                    CHECK_PART_FILE_CORRUPT,
                    CHECK_PART_SUCCESS,
                    CHECK_PART_SUCCESS,
                    CHECK_PART_SUCCESS,
                ],
            ),
        ]);

        populate_data_errs_by_disk(&mut data_errs_by_disk, &data_errs_by_part);

        assert_eq!(
            data_errs_by_disk.get(&0).unwrap(),
            &vec![CHECK_PART_FILE_NOT_FOUND, CHECK_PART_FILE_CORRUPT]
        );
        assert_eq!(data_errs_by_disk.get(&1).unwrap(), &vec![CHECK_PART_SUCCESS, CHECK_PART_SUCCESS]);
        assert_eq!(data_errs_by_disk.get(&2).unwrap(), &vec![CHECK_PART_SUCCESS, CHECK_PART_SUCCESS]);
        assert_eq!(data_errs_by_disk.get(&3).unwrap(), &vec![CHECK_PART_SUCCESS, CHECK_PART_SUCCESS]);
    }

    #[test]
    fn test_should_heal_object_on_disk() {
        // Test healing decision logic
        let meta = FileInfo::default();
        let latest_meta = FileInfo::default();

        // Test with file not found error
        let err = Some(DiskError::FileNotFound);
        let (should_heal, _, _) = should_heal_object_on_disk(&err, &[], &meta, &latest_meta);
        assert!(should_heal);

        let err = Some(DiskError::VolumeNotFound);
        let (should_heal, is_meta, reason) = should_heal_object_on_disk(&err, &[], &meta, &latest_meta);
        assert!(should_heal);
        assert!(is_meta);
        assert_eq!(reason, Some(DiskError::VolumeNotFound));

        let err = Some(DiskError::FileCorrupt);
        let (should_heal, is_meta, reason) = should_heal_object_on_disk(&err, &[], &meta, &latest_meta);
        assert!(should_heal);
        assert!(is_meta);
        assert_eq!(reason, Some(DiskError::FileCorrupt));

        // Test with no error and no part errors
        let (should_heal, _, _) = should_heal_object_on_disk(&None, &[CHECK_PART_SUCCESS], &meta, &latest_meta);
        assert!(!should_heal);

        // Test with part corruption
        let (should_heal, _, reason) = should_heal_object_on_disk(&None, &[CHECK_PART_FILE_CORRUPT], &meta, &latest_meta);
        assert!(should_heal);
        assert_eq!(reason, Some(DiskError::FileCorrupt));
    }

    #[test]
    fn stale_delete_marker_identity_requires_metadata_heal() {
        let historical_version = Uuid::new_v4();
        let marker_version = Uuid::new_v4();
        let mut stale = FileInfo {
            version_id: Some(historical_version),
            ..FileInfo::default()
        };
        let mut latest = stale.clone();
        stale.deleted = false;
        latest.deleted = true;
        latest.version_id = Some(marker_version);

        // The generic equality helper intentionally considers these records
        // equal when their erasure shape and mod-time match. Heal must still
        // repair the version identity and delete-marker state.
        assert!(stale.equals(&latest));
        let (should_heal, metadata, reason) = should_heal_object_on_disk(&None, &[], &stale, &latest);
        assert!(should_heal);
        assert!(metadata);
        assert_eq!(reason, Some(DiskError::OutdatedXLMeta));
    }

    #[test]
    fn metadata_io_failures_never_authorize_heal_overwrite() {
        let meta = FileInfo::default();
        let io_errors = [
            std::io::Error::new(std::io::ErrorKind::PermissionDenied, "metadata access denied"),
            std::io::Error::other("transient metadata read failure"),
        ];
        let mut errors: Vec<_> = io_errors
            .into_iter()
            .map(|error| DiskError::from(rustfs_filemeta::Error::Io(error)))
            .collect();
        #[cfg(unix)]
        errors.push(DiskError::from(rustfs_filemeta::Error::Io(std::io::Error::from_raw_os_error(libc::EIO))));
        errors.push(DiskError::Timeout);
        for error in errors {
            assert_ne!(error, DiskError::FileCorrupt);
            let (heal, metadata, reason) =
                should_heal_object_on_disk(&Some(error.clone()), &[CHECK_PART_FILE_CORRUPT], &meta, &meta);
            assert!(!heal, "an I/O failure must not authorize overwriting metadata: {error}");
            assert!(!metadata);
            assert_eq!(reason, Some(error));
        }
    }

    #[tokio::test]
    async fn test_get_disks_info_preserves_runtime_state_for_suspect_and_offline_disks() {
        let format = FormatV3::new(1, 3);
        let mut temp_dirs = Vec::new();
        let mut endpoints = Vec::new();
        let mut disks = Vec::new();

        for disk_idx in 0..3 {
            let (dir, endpoint, disk) = make_formatted_local_disk_for_info_test(disk_idx, &format).await;
            temp_dirs.push(dir);
            endpoints.push(endpoint);
            disks.push(Some(disk));
        }

        disks[1]
            .as_ref()
            .expect("disk 1 should exist")
            .force_runtime_state_for_test(RuntimeDriveHealthState::Suspect);
        let offline_disk_id = Uuid::new_v4();
        disks[2]
            .as_ref()
            .expect("disk 2 should exist")
            .set_disk_id_state(Some(offline_disk_id))
            .await
            .expect("offline disk id should be cached");
        disks[2]
            .as_ref()
            .expect("disk 2 should exist")
            .force_runtime_state_for_test(RuntimeDriveHealthState::Offline);

        let info = get_disks_info(&disks, &endpoints).await;
        assert_eq!(info.len(), 3);

        assert_eq!(info[0].state, "ok");
        assert_eq!(info[0].runtime_state.as_deref(), Some("online"));
        assert!(!info[0].drive_path.is_empty(), "online disk should keep immediate disk_info probe");
        assert!(
            info[0]
                .metrics
                .as_ref()
                .and_then(|metrics| metrics.api_calls.get("disk_info"))
                .copied()
                .unwrap_or_default()
                > 0,
            "online disk should expose disk_info operation metrics"
        );

        assert_eq!(info[1].state, "ok");
        assert_eq!(info[1].runtime_state.as_deref(), Some("suspect"));
        assert!(!info[1].drive_path.is_empty(), "suspect disk should still probe for fresher disk info");
        assert!(
            info[1]
                .metrics
                .as_ref()
                .and_then(|metrics| metrics.last_minute.get("disk_info"))
                .map(|action| action.count)
                .unwrap_or_default()
                > 0,
            "suspect disk should expose last-minute disk_info latency"
        );

        assert_eq!(info[2].state, "offline");
        assert_eq!(info[2].runtime_state.as_deref(), Some("offline"));
        assert_eq!(
            info[2].drive_path,
            endpoints[2].get_file_path(),
            "offline disk should keep stable endpoint path"
        );
        assert_eq!(info[2].uuid, offline_disk_id.to_string());
        assert!(
            info[2].metrics.is_some(),
            "offline runtime fallback should preserve disk metrics snapshot"
        );
    }

    #[tokio::test(start_paused = true)]
    async fn test_get_disks_info_probes_drives_concurrently() {
        use crate::disk::disk_store::DISK_INFO_PROBE_DELAY_FOR_TEST;

        let format = FormatV3::new(1, 4);
        let mut temp_dirs = Vec::new();
        let mut endpoints = Vec::new();
        let mut disks = Vec::new();
        for disk_idx in 0..4 {
            let (dir, endpoint, disk) = make_formatted_local_disk_for_info_test(disk_idx, &format).await;
            temp_dirs.push(dir);
            endpoints.push(endpoint);
            disks.push(Some(disk));
        }

        let probe_delay = std::time::Duration::from_secs(2);
        let started = tokio::time::Instant::now();
        let info = DISK_INFO_PROBE_DELAY_FOR_TEST
            .scope(probe_delay, get_disks_info(&disks, &endpoints))
            .await;
        let elapsed = started.elapsed();

        assert_eq!(info.len(), 4);
        assert!(info.iter().all(|disk| disk.state == "ok"), "every drive should still report a live probe");
        assert_eq!(
            info.iter().map(|disk| disk.disk_index).collect::<Vec<_>>(),
            endpoints.iter().map(|ep| ep.disk_idx).collect::<Vec<_>>(),
            "concurrent probes must keep endpoint order"
        );
        assert!(
            elapsed < probe_delay * 2,
            "four stalled drives must cost one probe delay, not four; took {elapsed:?}"
        );
    }

    #[tokio::test]
    async fn test_get_disks_info_preserves_remote_cached_disk_id_when_offline() {
        let (endpoint, disk) = make_remote_disk_for_info_test(0).await;
        let remote_disk_id = Uuid::new_v4();
        disk.set_disk_id_state(Some(remote_disk_id))
            .await
            .expect("remote disk id should be cached");
        disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline);

        let info = get_disks_info(&[Some(disk)], &[endpoint]).await;

        assert_eq!(info.len(), 1);
        assert_eq!(info[0].state, "offline");
        assert_eq!(info[0].runtime_state.as_deref(), Some("offline"));
        assert_eq!(info[0].uuid, remote_disk_id.to_string());
    }

    #[tokio::test]
    async fn test_get_disks_info_preserves_cached_disk_id_after_failed_live_probe() {
        let format = FormatV3::new(1, 1);
        let (temp_dir, endpoint, disk) = make_formatted_local_disk_for_info_test(0, &format).await;
        let cached_disk_id = Uuid::new_v4();
        disk.set_disk_id_state(Some(cached_disk_id))
            .await
            .expect("disk id should be cached before the failed probe");
        disk.force_runtime_state_for_test(RuntimeDriveHealthState::Suspect);

        let info = get_disks_info(&[Some(disk)], &[endpoint]).await;

        assert_eq!(info.len(), 1);
        assert_eq!(info[0].runtime_state.as_deref(), Some("suspect"));
        assert_eq!(info[0].uuid, cached_disk_id.to_string());
        assert_eq!(
            info[0].drive_path,
            temp_dir.path().to_string_lossy(),
            "failed live probe should still keep the endpoint path"
        );
    }

    #[tokio::test]
    async fn test_get_disks_info_uses_capacity_snapshot_for_offline_disk() {
        let format = FormatV3::new(1, 1);
        let (temp_dir, endpoint, disk) = make_formatted_local_disk_for_info_test(0, &format).await;
        disk.record_capacity_probe(100, 40, 60);
        disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline);

        let info = get_disks_info(&[Some(disk)], &[endpoint]).await;
        assert_eq!(info.len(), 1);
        assert_eq!(info[0].state, "offline");
        assert_eq!(info[0].runtime_state.as_deref(), Some("offline"));
        assert_eq!(info[0].capacity_observation_source.as_deref(), Some("snapshot"));
        assert!(info[0].capacity_observation_age_seconds.unwrap_or(u64::MAX) <= 60);
        assert_eq!(info[0].total_space, 100);
        assert_eq!(info[0].used_space, 40);
        assert_eq!(info[0].available_space, 60);
        assert_eq!(info[0].utilization, 40.0);

        drop(temp_dir);
    }

    #[tokio::test]
    async fn list_path_returns_read_quorum_when_runtime_candidates_are_empty() {
        let disk_count = 4;
        let format = FormatV3::new(1, disk_count);
        let mut temp_dirs = Vec::with_capacity(disk_count);
        let mut endpoints = Vec::with_capacity(disk_count);
        let mut disks = Vec::with_capacity(disk_count);

        for disk_idx in 0..disk_count {
            let (dir, endpoint, disk) = make_formatted_local_disk_for_info_test(disk_idx, &format).await;
            temp_dirs.push(dir);
            endpoints.push(endpoint);
            disks.push(Some(disk));
        }

        let set_disks = SetDisks::new(
            "test-owner".to_string(),
            Arc::new(RwLock::new(disks)),
            disk_count,
            disk_count / 2,
            0,
            0,
            endpoints,
            format,
            Vec::new(),
        )
        .await;

        for disk in set_disks.get_disks_internal().await.iter().flatten() {
            disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline);
        }

        let (tx, _rx) = mpsc::channel(1);
        let err = set_disks
            .list_path(
                CancellationToken::new(),
                ListPathOptions {
                    bucket: "bucket".to_string(),
                    recursive: true,
                    ..Default::default()
                },
                tx,
            )
            .await
            .expect_err("empty runtime candidate set should fail before list_path_raw");

        assert_eq!(err, StorageError::ErasureReadQuorum);

        drop(temp_dirs);
    }

    #[tokio::test]
    async fn load_file_info_versions_exact_returns_none_for_explicit_not_found() {
        let format = FormatV3::new(1, 1);
        let (temp_dir, endpoint, disk) = make_formatted_local_disk_for_info_test(0, &format).await;
        let bucket = "bucket";

        disk.make_volume(bucket).await.expect("bucket should be created");

        let set_disks = SetDisks::new(
            "test-owner".to_string(),
            Arc::new(RwLock::new(vec![Some(disk)])),
            1,
            0,
            0,
            0,
            vec![endpoint],
            format,
            Vec::new(),
        )
        .await;

        let versions = set_disks
            .load_file_info_versions_exact(bucket, "missing-object")
            .await
            .expect("explicit object not found should be accepted");

        assert!(versions.is_none());
        drop(temp_dir);
    }

    #[tokio::test]
    async fn load_file_info_versions_exact_rejects_corrupt_metadata() {
        let format = FormatV3::new(1, 1);
        let (temp_dir, endpoint, disk) = make_formatted_local_disk_for_info_test(0, &format).await;
        let bucket = "bucket";
        let object = "object.txt";

        disk.make_volume(bucket).await.expect("bucket should be created");
        let metadata_path = format!("{object}/{STORAGE_FORMAT_FILE}");
        disk.write_all(bucket, &metadata_path, Bytes::from_static(b"not-xl-meta"))
            .await
            .expect("corrupt metadata file should be written");

        let set_disks = SetDisks::new(
            "test-owner".to_string(),
            Arc::new(RwLock::new(vec![Some(disk)])),
            1,
            0,
            0,
            0,
            vec![endpoint],
            format,
            Vec::new(),
        )
        .await;

        let err = set_disks
            .load_file_info_versions_exact(bucket, object)
            .await
            .expect_err("corrupt exact metadata must fail closed");

        assert!(!is_err_object_not_found(&err), "corrupt metadata must not be treated as not found: {err}");
        drop(temp_dir);
    }

    #[tokio::test]
    async fn list_path_still_uses_disk_after_prior_walk_timeout() {
        use std::pin::Pin;
        use std::task::{Context, Poll};
        use tokio::io::AsyncWrite;

        struct PendingWriter;

        impl AsyncWrite for PendingWriter {
            fn poll_write(self: Pin<&mut Self>, _cx: &mut Context<'_>, _buf: &[u8]) -> Poll<std::io::Result<usize>> {
                Poll::Pending
            }

            fn poll_flush(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll<std::io::Result<()>> {
                Poll::Ready(Ok(()))
            }

            fn poll_shutdown(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll<std::io::Result<()>> {
                Poll::Ready(Ok(()))
            }
        }

        let format = FormatV3::new(1, 1);
        let (temp_dir, endpoint, disk) = make_formatted_local_disk_for_info_test(0, &format).await;
        let bucket = "bucket";
        let object = "obj";

        disk.make_volume(bucket).await.expect("bucket should be created");
        let metadata_path = format!("{object}/{STORAGE_FORMAT_FILE}");
        disk.write_all(bucket, &metadata_path, Bytes::from_static(b"not-an-xl-meta"))
            .await
            .expect("metadata file should be created");

        let set_disks = SetDisks::new(
            "test-owner".to_string(),
            Arc::new(RwLock::new(vec![Some(disk.clone())])),
            1,
            0,
            0,
            0,
            vec![endpoint],
            format,
            Vec::new(),
        )
        .await;

        temp_env::async_with_vars(
            [
                (rustfs_config::ENV_DRIVE_WALKDIR_TIMEOUT_SECS, Some("1")),
                (rustfs_config::ENV_DRIVE_WALKDIR_STALL_TIMEOUT_SECS, Some("1")),
            ],
            async {
                let mut writer = PendingWriter;
                let walk_err = disk
                    .walk_dir(
                        WalkDirOptions {
                            bucket: bucket.to_string(),
                            recursive: true,
                            ..Default::default()
                        },
                        &mut writer,
                    )
                    .await
                    .expect_err("walk_dir should time out");
                assert_eq!(walk_err, DiskError::Timeout);
                assert_eq!(disk.runtime_state(), RuntimeDriveHealthState::Online);

                let (tx, mut rx) = mpsc::channel::<MetaCacheEntry>(4);
                set_disks
                    .list_path(
                        CancellationToken::new(),
                        ListPathOptions {
                            bucket: bucket.to_string(),
                            recursive: true,
                            ..Default::default()
                        },
                        tx,
                    )
                    .await
                    .expect("list_path should still succeed after prior walk timeout");

                let entry = rx.recv().await.expect("listing should yield the object entry");
                assert_eq!(entry.name, object);
                assert_eq!(disk.runtime_state(), RuntimeDriveHealthState::Online);
            },
        )
        .await;

        drop(temp_dir);
    }

    #[tokio::test]
    async fn list_path_system_prefix_survives_prior_walk_timeout() {
        use std::pin::Pin;
        use std::task::{Context, Poll};
        use tokio::io::AsyncWrite;

        struct PendingWriter;

        impl AsyncWrite for PendingWriter {
            fn poll_write(self: Pin<&mut Self>, _cx: &mut Context<'_>, _buf: &[u8]) -> Poll<std::io::Result<usize>> {
                Poll::Pending
            }

            fn poll_flush(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll<std::io::Result<()>> {
                Poll::Ready(Ok(()))
            }

            fn poll_shutdown(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll<std::io::Result<()>> {
                Poll::Ready(Ok(()))
            }
        }

        let format = FormatV3::new(1, 1);
        let (temp_dir, endpoint, disk) = make_formatted_local_disk_for_info_test(0, &format).await;
        let object = "config/iam/sts/test/identity.json";

        let metadata_path = format!("{object}/{STORAGE_FORMAT_FILE}");
        disk.write_all(RUSTFS_META_BUCKET, &metadata_path, Bytes::from_static(b"not-an-xl-meta"))
            .await
            .expect("system path metadata file should be created");

        let set_disks = SetDisks::new(
            "test-owner".to_string(),
            Arc::new(RwLock::new(vec![Some(disk.clone())])),
            1,
            0,
            0,
            0,
            vec![endpoint],
            format,
            Vec::new(),
        )
        .await;

        temp_env::async_with_vars(
            [
                (rustfs_config::ENV_DRIVE_WALKDIR_TIMEOUT_SECS, Some("1")),
                (rustfs_config::ENV_DRIVE_WALKDIR_STALL_TIMEOUT_SECS, Some("1")),
            ],
            async {
                let mut writer = PendingWriter;
                let walk_err = disk
                    .walk_dir(
                        WalkDirOptions {
                            bucket: RUSTFS_META_BUCKET.to_string(),
                            base_dir: "config/iam/".to_string(),
                            recursive: true,
                            ..Default::default()
                        },
                        &mut writer,
                    )
                    .await
                    .expect_err("walk_dir should time out");
                assert_eq!(walk_err, DiskError::Timeout);
                assert_eq!(disk.runtime_state(), RuntimeDriveHealthState::Online);

                let (tx, mut rx) = mpsc::channel::<MetaCacheEntry>(4);
                set_disks
                    .list_path(
                        CancellationToken::new(),
                        ListPathOptions {
                            bucket: RUSTFS_META_BUCKET.to_string(),
                            base_dir: "config/iam/".to_string(),
                            recursive: true,
                            ..Default::default()
                        },
                        tx,
                    )
                    .await
                    .expect("system prefix list_path should still succeed after prior walk timeout");

                let entry = rx.recv().await.expect("listing should yield the system-path entry");
                assert_eq!(entry.name, "config/iam/sts/");
                assert!(
                    entry.is_dir(),
                    "system prefix listing should still yield a directory entry after timeout recovery"
                );
                assert_eq!(disk.runtime_state(), RuntimeDriveHealthState::Online);
            },
        )
        .await;

        drop(temp_dir);
    }

    #[test]
    fn test_dangling_meta_errs_count() {
        // Test counting dangling metadata errors
        let errs = vec![None, Some(DiskError::FileNotFound), None];
        let (not_found_count, non_actionable_count) = dangling_meta_errs_count(&errs);
        assert_eq!(not_found_count, 1); // One FileNotFound error
        assert_eq!(non_actionable_count, 0); // No other errors
    }

    #[test]
    fn test_dangling_part_errs_count() {
        // Test counting dangling part errors
        let results = vec![CHECK_PART_SUCCESS, CHECK_PART_FILE_NOT_FOUND, CHECK_PART_SUCCESS];
        let (not_found_count, non_actionable_count) = dangling_part_errs_count(&results);
        assert_eq!(not_found_count, 1); // One FILE_NOT_FOUND error
        assert_eq!(non_actionable_count, 0); // No other errors
    }

    #[test]
    fn test_is_object_dir_dangling() {
        // Test object directory dangling detection
        let errs = vec![Some(DiskError::FileNotFound), Some(DiskError::FileNotFound), None];
        assert!(is_object_dir_dangling(&errs));
        let errs2 = vec![None, None, None];
        assert!(!is_object_dir_dangling(&errs2));

        let errs3 = vec![Some(DiskError::FileCorrupt), Some(DiskError::FileNotFound)];
        assert!(!is_object_dir_dangling(&errs3)); // Mixed errors, not all not found
    }

    #[test]
    fn test_join_errs() {
        // Test joining error messages
        let errs = vec![None, Some(DiskError::other("error1")), Some(DiskError::other("error2"))];
        let joined = join_errs(&errs);
        assert!(joined.contains("<nil>"));
        assert!(joined.contains("io error")); // DiskError::other is rendered as "io error"

        // Test with different error types
        let errs2 = vec![None, Some(DiskError::FileNotFound), Some(DiskError::FileCorrupt)];
        let joined2 = join_errs(&errs2);
        assert!(joined2.contains("<nil>"));
        assert!(joined2.contains("file not found"));
        assert!(joined2.contains("file is corrupted"));
    }

    #[test]
    fn test_reduce_common_data_dir() {
        // Test reducing common data directory
        use uuid::Uuid;

        let uuid1 = Uuid::new_v4();
        let uuid2 = Uuid::new_v4();

        let data_dirs = vec![Some(uuid1), Some(uuid1), Some(uuid2)];
        let result = SetDisks::reduce_common_data_dir(&data_dirs, 2);
        assert_eq!(result, Some(uuid1)); // uuid1 appears twice, meets quorum

        let data_dirs = vec![Some(uuid1), Some(uuid2), None];
        let result = SetDisks::reduce_common_data_dir(&data_dirs, 2);
        assert_eq!(result, None); // No UUID meets quorum of 2

        let data_dirs = vec![Some(uuid1), Some(uuid1), None, None];
        let result = SetDisks::reduce_common_data_dir(&data_dirs, 2);
        assert_eq!(result, Some(uuid1)); // Ignore None votes; uuid1 should still meet quorum
    }

    fn rename_versions_signature(first_key: u8, version_count: usize) -> Vec<u8> {
        let mut signature = vec![0; version_count * 16];
        signature[7] = first_key;
        signature
    }

    // backlog#1321: `classify_rename_convergence` is the single decision that
    // replaced the old `Option<Vec<u8>>::is_some()` heal gate. These are the
    // revert-fails white-box cases behind the issue acceptance matrix — if the
    // decision regressed to "a signature exists => needs heal", the healthy
    // 4/4 and 8/8 cases below would flip from `AllSuccessIdentical` to a
    // heal-worthy variant and fail.
    #[test]
    fn test_classify_rename_convergence_healthy_identical_needs_no_heal() {
        use crate::set_disk::core::io_primitives::RenameConvergence;
        let sig = rename_versions_signature(1, 3);

        // Healthy 4/4: every disk committed with an identical, known signature.
        let disk_versions = vec![Some(sig.clone()), Some(sig.clone()), Some(sig.clone()), Some(sig.clone())];
        let errs = vec![None, None, None, None];
        let convergence = SetDisks::classify_rename_convergence(&disk_versions, &errs);
        assert_eq!(convergence, RenameConvergence::AllSuccessIdentical);
        assert!(!convergence.needs_heal(), "a fully converged healthy MPU must not enqueue heal");

        // Healthy 8/8: same property at a wider set width.
        let disk_versions = vec![Some(sig); 8];
        let errs = vec![None; 8];
        let convergence = SetDisks::classify_rename_convergence(&disk_versions, &errs);
        assert_eq!(convergence, RenameConvergence::AllSuccessIdentical);
        assert!(!convergence.needs_heal());
    }

    #[test]
    fn test_classify_rename_convergence_three_agree_one_diverges_heals() {
        use crate::set_disk::core::io_primitives::RenameConvergence;
        let common = rename_versions_signature(1, 2);
        let odd = rename_versions_signature(2, 2);

        // 3-same, 1-divergent, all committed: reconcile the odd replica.
        let disk_versions = vec![Some(common.clone()), Some(common.clone()), Some(common), Some(odd)];
        let errs = vec![None, None, None, None];
        let convergence = SetDisks::classify_rename_convergence(&disk_versions, &errs);
        assert_eq!(convergence, RenameConvergence::SignatureDivergent);
        assert!(convergence.needs_heal());
    }

    #[test]
    fn test_classify_rename_convergence_failed_or_offline_disk_heals() {
        use crate::set_disk::core::io_primitives::RenameConvergence;
        let sig = rename_versions_signature(1, 2);

        // 1 disk failed/offline while the rest committed identically: past the
        // write-quorum gate this is a `PartialCommit` — a replica is missing.
        let disk_versions = vec![Some(sig.clone()), Some(sig.clone()), Some(sig), None];
        let errs = vec![None, None, None, Some(DiskError::DiskNotFound)];
        let convergence = SetDisks::classify_rename_convergence(&disk_versions, &errs);
        assert_eq!(convergence, RenameConvergence::PartialCommit);
        assert!(convergence.needs_heal());
    }

    #[test]
    fn test_classify_rename_convergence_no_common_quorum_heals() {
        use crate::set_disk::core::io_primitives::RenameConvergence;
        // No signature holds a majority (2/2 split), every disk committed:
        // divergence with no common quorum still reconciles via heal.
        let a = rename_versions_signature(1, 1);
        let b = rename_versions_signature(2, 1);
        let disk_versions = vec![Some(a.clone()), Some(a), Some(b.clone()), Some(b)];
        let errs = vec![None, None, None, None];
        let convergence = SetDisks::classify_rename_convergence(&disk_versions, &errs);
        assert_eq!(convergence, RenameConvergence::SignatureDivergent);
        assert!(convergence.needs_heal());
    }

    #[test]
    fn test_classify_rename_convergence_over_ten_versions_by_success() {
        use crate::set_disk::core::io_primitives::RenameConvergence;
        // >10 versions: every disk deliberately omits the signature (`None`).
        // All committed => `Unknown` => scanner-backstopped, no self-enqueue.
        let disk_versions = vec![None, None, None, None];
        let errs = vec![None, None, None, None];
        let convergence = SetDisks::classify_rename_convergence(&disk_versions, &errs);
        assert_eq!(convergence, RenameConvergence::Unknown);
        assert!(
            !convergence.needs_heal(),
            ">10-version healthy commit relies on the scanner, not self-enqueue"
        );

        // Same >10-version shape but with a failed disk: a failure is
        // conservative-heal regardless of signatures being unavailable.
        let errs = vec![None, None, None, Some(DiskError::FileCorrupt)];
        let convergence = SetDisks::classify_rename_convergence(&disk_versions, &errs);
        assert_eq!(convergence, RenameConvergence::PartialCommit);
        assert!(convergence.needs_heal());
    }

    #[test]
    fn test_classify_rename_convergence_mixed_signed_unsigned_heals() {
        use crate::set_disk::core::io_primitives::RenameConvergence;
        // A committed replica with <=10 versions (signed) alongside one with
        // >10 versions (unsigned) is itself a version-count divergence.
        let sig = rename_versions_signature(1, 2);
        let disk_versions = vec![Some(sig.clone()), Some(sig.clone()), Some(sig), None];
        let errs = vec![None, None, None, None];
        let convergence = SetDisks::classify_rename_convergence(&disk_versions, &errs);
        assert_eq!(convergence, RenameConvergence::SignatureDivergent);
        assert!(convergence.needs_heal());
    }

    #[test]
    fn test_object_quorum_from_meta_returns_not_found_when_all_metadata_is_missing() {
        let errs = vec![
            Some(DiskError::FileNotFound),
            Some(DiskError::VolumeNotFound),
            Some(DiskError::DiskNotFound),
            Some(DiskError::FileNotFound),
        ];

        let err = SetDisks::object_quorum_from_meta(&vec![FileInfo::default(); errs.len()], &errs, 2)
            .expect_err("missing metadata should map to FileNotFound");

        assert_eq!(err, DiskError::FileNotFound);
    }

    #[test]
    fn test_object_quorum_from_meta_preserves_version_not_found() {
        for errs in [
            vec![Some(DiskError::FileVersionNotFound); 4],
            vec![
                Some(DiskError::FileVersionNotFound),
                Some(DiskError::FileNotFound),
                Some(DiskError::FileVersionNotFound),
                Some(DiskError::FileNotFound),
            ],
            vec![
                Some(DiskError::FileVersionNotFound),
                Some(DiskError::VolumeNotFound),
                Some(DiskError::DiskNotFound),
                Some(DiskError::FileVersionNotFound),
            ],
        ] {
            let err = SetDisks::object_quorum_from_meta(&vec![FileInfo::default(); errs.len()], &errs, 2)
                .expect_err("absent version metadata must remain a version miss");
            assert_eq!(err, DiskError::FileVersionNotFound, "disk replies: {errs:?}");
        }
    }

    #[test]
    fn test_object_quorum_from_meta_version_misses_preserve_other_failures() {
        for (errs, expected) in [
            (vec![Some(DiskError::DiskNotFound); 4], DiskError::ErasureReadQuorum),
            (
                vec![
                    Some(DiskError::FileVersionNotFound),
                    Some(DiskError::FileCorrupt),
                    Some(DiskError::DiskNotFound),
                    None,
                ],
                DiskError::ErasureReadQuorum,
            ),
            (
                vec![
                    Some(DiskError::FileVersionNotFound),
                    Some(DiskError::FileAccessDenied),
                    Some(DiskError::FileAccessDenied),
                    Some(DiskError::DiskNotFound),
                ],
                DiskError::FileAccessDenied,
            ),
            (
                vec![
                    Some(DiskError::FileVersionNotFound),
                    Some(DiskError::VolumeNotFound),
                    Some(DiskError::VolumeNotFound),
                    None,
                ],
                DiskError::VolumeNotFound,
            ),
            (
                vec![
                    Some(DiskError::FileVersionNotFound),
                    Some(DiskError::FileVersionNotFound),
                    Some(DiskError::FileCorrupt),
                    None,
                ],
                DiskError::FileVersionNotFound,
            ),
        ] {
            let err = SetDisks::object_quorum_from_meta(&vec![FileInfo::default(); errs.len()], &errs, 2)
                .expect_err("metadata failures must retain quorum reduction semantics");
            assert_eq!(err, expected, "disk replies: {errs:?}");
        }
    }

    #[test]
    fn test_object_quorum_from_meta_preserves_read_quorum_for_mixed_failures() {
        let errs = vec![
            Some(DiskError::FileNotFound),
            Some(DiskError::VolumeNotFound),
            Some(DiskError::FileCorrupt),
            Some(DiskError::DiskNotFound),
        ];

        let err = SetDisks::object_quorum_from_meta(&vec![FileInfo::default(); errs.len()], &errs, 2)
            .expect_err("mixed metadata failures should keep quorum semantics");

        assert_eq!(err, DiskError::ErasureReadQuorum);
    }

    #[test]
    fn test_shuffle_parts_metadata() {
        // Test metadata shuffling
        let metadata = vec![
            FileInfo {
                name: "file1".to_string(),
                ..Default::default()
            },
            FileInfo {
                name: "file2".to_string(),
                ..Default::default()
            },
            FileInfo {
                name: "file3".to_string(),
                ..Default::default()
            },
        ];

        // Distribution uses 1-based indexing
        let distribution = vec![3, 1, 2]; // 1-based shuffle order
        let result = SetDisks::shuffle_parts_metadata(&metadata, &distribution);

        assert_eq!(result.len(), 3);
        assert_eq!(result[0].name, "file2"); // distribution[1] = 1, so metadata[1] goes to index 0
        assert_eq!(result[1].name, "file3"); // distribution[2] = 2, so metadata[2] goes to index 1
        assert_eq!(result[2].name, "file1"); // distribution[0] = 3, so metadata[0] goes to index 2

        // Test with empty distribution
        let empty_distribution = vec![];
        let result2 = SetDisks::shuffle_parts_metadata(&metadata, &empty_distribution);
        assert_eq!(result2.len(), 3);
        assert_eq!(result2[0].name, "file1"); // Should return original order
    }

    #[test]
    fn test_shuffle_disks() {
        // Test disk shuffling
        let disks = vec![None, None, None]; // Mock disks
        let distribution = vec![3, 1, 2]; // 1-based indexing

        let result = SetDisks::shuffle_disks(&disks, &distribution);
        assert_eq!(result.len(), 3);
        // All disks are None, so result should be all None
        assert!(result.iter().all(|d| d.is_none()));

        // Test with empty distribution
        let empty_distribution = vec![];
        let result2 = SetDisks::shuffle_disks(&disks, &empty_distribution);
        assert_eq!(result2.len(), 3);
        assert!(result2.iter().all(|d| d.is_none()));
    }

    #[test]
    fn test_etag_matches() {
        assert!(e_tag_matches("abc", "abc"));
        assert!(e_tag_matches("\"abc\"", "abc"));
        assert!(e_tag_matches("\"abc\"", "*"));
    }

    #[test]
    fn test_build_tiered_decommission_file_info_preserves_transition_metadata() {
        let version_id = Uuid::new_v4();
        let transition_version_id = Uuid::new_v4();
        let original = FileInfo {
            version_id: Some(version_id),
            transition_status: TRANSITION_COMPLETE.to_string(),
            transitioned_objname: "remote/object".to_string(),
            transition_tier: "WARM-TIER".to_string(),
            transition_version_id: Some(transition_version_id),
            erasure: FileInfo::new("old-bucket/old-object", 8, 8).erasure,
            ..Default::default()
        };

        let layout = WriteLayout::from_parity(16, 4).expect("tiered write layout should be valid");
        let updated = build_tiered_decommission_file_info("bucket", "object", &original, layout);

        assert_eq!(updated.version_id, original.version_id);
        assert_eq!(updated.transition_status, original.transition_status);
        assert_eq!(updated.transitioned_objname, original.transitioned_objname);
        assert_eq!(updated.transition_tier, original.transition_tier);
        assert_eq!(updated.transition_version_id, original.transition_version_id);
        assert_eq!(updated.erasure.data_blocks, 12);
        assert_eq!(updated.erasure.parity_blocks, 4);
        assert_eq!(layout.write_quorum, 12);
        assert_ne!(updated.erasure.distribution, original.erasure.distribution);
    }

    #[tokio::test]
    async fn decommission_tier_free_version_preserves_remote_identity() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "free-version-decommission";
        let object = "object.txt";
        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("target bucket should exist before free-version migration");
        let version_id = Uuid::new_v4();
        let mut free_version = FileInfo {
            name: object.to_string(),
            volume: bucket.to_string(),
            version_id: Some(version_id),
            mod_time: Some(time::OffsetDateTime::now_utc()),
            deleted: true,
            transition_tier: "WARM-TIER".to_string(),
            transitioned_objname: "remote/object".to_string(),
            ..Default::default()
        };
        free_version.set_tier_free_version();
        // Decoded free versions always carry the on-disk free-version
        // suffix alongside the in-memory tier marker; mirror that here so
        // the record satisfies delete-marker metadata validation.
        rustfs_utils::http::metadata_compat::insert_str(
            &mut free_version.metadata,
            rustfs_utils::http::metadata_compat::SUFFIX_FREE_VERSION,
            String::new(),
        );

        set_disks
            .decommission_tier_free_version(bucket, object, &free_version, &ObjectOptions::default())
            .await
            .expect("free-version metadata should reach the target quorum");
        set_disks
            .decommission_tier_free_version(bucket, object, &free_version, &ObjectOptions::default())
            .await
            .expect("replaying the same free-version metadata should be idempotent");

        let versions = set_disks
            .load_file_info_versions_exact(bucket, object)
            .await
            .expect("migrated free-version metadata should decode")
            .expect("migrated free-version metadata should exist");
        let migrated = versions
            .versions
            .iter()
            .find(|version| version.version_id == Some(version_id))
            .expect("free version should be present on the target");

        assert_eq!(
            versions
                .versions
                .iter()
                .filter(|version| version.version_id == Some(version_id))
                .count(),
            1
        );
        assert!(migrated.tier_free_version());
        assert_eq!(migrated.transition_tier, "WARM-TIER");
        assert_eq!(migrated.transitioned_objname, "remote/object");
    }

    #[tokio::test]
    async fn decommission_tier_free_version_resume_requires_write_quorum() {
        let set_disks = make_local_bucket_test_set_disks_with_drive_count(4).await;
        let bucket = "free-version-decommission-resume";
        let object = "object.txt";
        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("target bucket should exist before free-version migration");
        let mut free_version = FileInfo {
            name: object.to_string(),
            volume: bucket.to_string(),
            version_id: Some(Uuid::new_v4()),
            mod_time: Some(time::OffsetDateTime::now_utc()),
            deleted: true,
            transition_tier: "WARM-TIER".to_string(),
            transitioned_objname: "remote/object".to_string(),
            ..Default::default()
        };
        free_version.set_tier_free_version();
        // Decoded free versions always carry the on-disk free-version
        // suffix alongside the in-memory tier marker; mirror that here so
        // the record satisfies delete-marker metadata validation.
        rustfs_utils::http::metadata_compat::insert_str(
            &mut free_version.metadata,
            rustfs_utils::http::metadata_compat::SUFFIX_FREE_VERSION,
            String::new(),
        );
        let opts = ObjectOptions::default();

        let disks = set_disks.get_disks_internal().await;
        for disk in disks.iter().take(2).flatten() {
            disk.write_metadata("", bucket, object, free_version.clone())
                .await
                .expect("partial first attempt should leave equivalent metadata");
        }
        assert!(
            !set_disks
                .has_decommission_tier_free_version_write_quorum(bucket, object, &free_version, &opts)
                .await
                .expect("partial target metadata should remain valid"),
            "write-quorum-minus-one must not be accepted as an idempotent migration"
        );

        disks[2]
            .as_ref()
            .expect("third target disk should be online")
            .write_metadata("", bucket, object, free_version.clone())
            .await
            .expect("third equivalent target write should complete quorum");
        assert!(
            set_disks
                .has_decommission_tier_free_version_write_quorum(bucket, object, &free_version, &opts)
                .await
                .expect("write-quorum target metadata should remain valid")
        );
    }

    #[test]
    fn decommission_tier_free_version_commit_rejects_lost_fence() {
        let opts = ObjectOptions {
            namespace_lock_fence: Some(NamespaceLockFence::lost_for_test()),
            ..Default::default()
        };

        let err = ensure_decommission_tier_free_version_commit_fence("bucket", "object", &opts)
            .expect_err("lost target lock must fail the free-version commit");
        assert!(matches!(
            err,
            Error::NamespaceLockQuorumUnavailable {
                mode: "decommission_tier_free_version_commit",
                required: 1,
                achieved: 0,
                ..
            }
        ));
    }

    #[test]
    fn test_resolve_tiered_decommission_write_quorum_result_allows_successful_quorum() {
        let errs = vec![None, None, Some(DiskError::DiskNotFound), None];

        let result = resolve_tiered_decommission_write_quorum_result(&errs, 3, "bucket", "object");

        assert!(result.is_ok());
    }

    #[test]
    fn test_resolve_tiered_decommission_write_quorum_result_wraps_object_context() {
        let errs = vec![
            Some(DiskError::DiskNotFound),
            Some(DiskError::DiskNotFound),
            Some(DiskError::DiskNotFound),
            Some(DiskError::DiskNotFound),
        ];

        let err = resolve_tiered_decommission_write_quorum_result(&errs, 3, "bucket", "object").expect_err("expected error");
        let rendered = err.to_string();

        assert!(rendered.contains("bucket"), "{rendered}");
        assert!(rendered.contains("object"), "{rendered}");
    }

    #[test]
    fn test_check_object_lock_retention_update_blocks_compliance_shorten() {
        let now = OffsetDateTime::now_utc();
        let existing_until = now + Duration::from_secs(60 * 60 * 24 * 60);
        let requested_until = now + Duration::from_secs(60 * 60 * 24);

        let mut user_defined = HashMap::new();
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_MODE.to_string(),
            s3s::dto::ObjectLockRetentionMode::COMPLIANCE.to_string(),
        );
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_RETAIN_UNTIL_DATE.to_string(),
            existing_until.format(&time::format_description::well_known::Rfc3339).unwrap(),
        );

        let obj_info = ObjectInfo {
            user_defined: Arc::new(user_defined),
            ..Default::default()
        };
        let opts = ObjectOptions {
            object_lock_retention: Some(crate::storage_api_contracts::object::ObjectLockRetentionOptions {
                mode: Some(s3s::dto::ObjectLockRetentionMode::COMPLIANCE.to_string()),
                retain_until: Some(requested_until),
                bypass_governance: true,
            }),
            ..Default::default()
        };

        let err = check_object_lock_retention_update("bucket", "object", &obj_info, &opts)
            .expect_err("COMPLIANCE shortening must be blocked");

        assert!(matches!(err, StorageError::PrefixAccessDenied(_, _)));
    }

    #[test]
    fn test_check_object_lock_retention_update_allows_governance_shorten_with_bypass() {
        let now = OffsetDateTime::now_utc();
        let existing_until = now + Duration::from_secs(60 * 60 * 24 * 60);
        let requested_until = now + Duration::from_secs(60 * 60 * 24);

        let mut user_defined = HashMap::new();
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_MODE.to_string(),
            s3s::dto::ObjectLockRetentionMode::GOVERNANCE.to_string(),
        );
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_RETAIN_UNTIL_DATE.to_string(),
            existing_until.format(&time::format_description::well_known::Rfc3339).unwrap(),
        );

        let obj_info = ObjectInfo {
            user_defined: Arc::new(user_defined),
            ..Default::default()
        };
        let opts = ObjectOptions {
            object_lock_retention: Some(crate::storage_api_contracts::object::ObjectLockRetentionOptions {
                mode: Some(s3s::dto::ObjectLockRetentionMode::GOVERNANCE.to_string()),
                retain_until: Some(requested_until),
                bypass_governance: true,
            }),
            ..Default::default()
        };

        check_object_lock_retention_update("bucket", "object", &obj_info, &opts)
            .expect("GOVERNANCE shortening with bypass should remain allowed");
    }

    #[tokio::test]
    async fn test_check_object_lock_delete_blocks_compliance_version_delete() {
        let retain_until = OffsetDateTime::now_utc() + Duration::from_secs(60 * 60 * 24 * 60);
        let mut user_defined = HashMap::new();
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_MODE.to_string(),
            s3s::dto::ObjectLockRetentionMode::COMPLIANCE.to_string(),
        );
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_RETAIN_UNTIL_DATE.to_string(),
            retain_until.format(&time::format_description::well_known::Rfc3339).unwrap(),
        );

        let obj_info = ObjectInfo {
            user_defined: Arc::new(user_defined),
            ..Default::default()
        };
        let opts = ObjectOptions {
            version_id: Some(Uuid::new_v4().to_string()),
            versioned: true,
            object_lock_config_snapshot: Some(Arc::new(ObjectLockConfigSnapshot::new(ObjectLockConfigState::ConfirmedAbsent))),
            ..Default::default()
        };

        let err = check_object_lock_delete(&bootstrap_ctx(), "bucket", "object", &obj_info, &opts)
            .await
            .expect_err("COMPLIANCE retention must block explicit version deletion");

        assert!(matches!(err, StorageError::PrefixAccessDenied(_, _)));
    }

    #[tokio::test]
    async fn test_check_object_lock_delete_allows_retained_restored_copy_expiry() {
        let retain_until = OffsetDateTime::now_utc() + Duration::from_secs(60 * 60 * 24 * 60);
        let restore_expiry = OffsetDateTime::now_utc() - Duration::from_secs(1);
        let version_id = Uuid::new_v4();
        let data_dir = Uuid::new_v4();
        let obj_info = ObjectInfo {
            version_id: Some(version_id),
            data_dir: Some(data_dir),
            etag: Some("etag".to_string()),
            transitioned_object: TransitionedObject {
                name: "remote-object".to_string(),
                tier: "tier".to_string(),
                status: TRANSITION_COMPLETE.to_string(),
                ..Default::default()
            },
            restore_expires: Some(restore_expiry),
            user_defined: Arc::new(HashMap::from([
                (
                    metadata_keys::OBJECT_LOCK_MODE.to_string(),
                    s3s::dto::ObjectLockRetentionMode::COMPLIANCE.to_string(),
                ),
                (
                    metadata_keys::OBJECT_LOCK_RETAIN_UNTIL_DATE.to_string(),
                    retain_until.format(&time::format_description::well_known::Rfc3339).unwrap(),
                ),
            ])),
            ..Default::default()
        };
        let opts = ObjectOptions {
            version_id: Some(version_id.to_string()),
            versioned: true,
            transition: crate::bucket::lifecycle::lifecycle::TransitionOptions {
                status: TRANSITION_COMPLETE.to_string(),
                tier: "tier".to_string(),
                etag: "etag".to_string(),
                expected_data_dir: Some(data_dir),
                expected_remote_name: "remote-object".to_string(),
                restore_expiry,
                expire_restored: true,
                ..Default::default()
            },
            object_lock_config_snapshot: Some(Arc::new(ObjectLockConfigSnapshot::new(ObjectLockConfigState::ConfirmedAbsent))),
            ..Default::default()
        };

        check_object_lock_delete(&bootstrap_ctx(), "bucket", "object", &obj_info, &opts)
            .await
            .expect("restore expiry only strips the local copy and must preserve the retained logical version");
    }

    #[tokio::test]
    async fn test_check_object_lock_delete_rejects_stale_restored_copy_expiry() {
        let restore_expiry = OffsetDateTime::now_utc() - Duration::from_secs(1);
        let data_dir = Uuid::new_v4();
        let obj_info = ObjectInfo {
            data_dir: Some(data_dir),
            etag: Some("etag".to_string()),
            transitioned_object: TransitionedObject {
                name: "remote-object".to_string(),
                tier: "tier".to_string(),
                status: TRANSITION_COMPLETE.to_string(),
                ..Default::default()
            },
            restore_expires: Some(restore_expiry + Duration::from_secs(60)),
            ..Default::default()
        };
        let opts = ObjectOptions {
            versioned: true,
            transition: crate::bucket::lifecycle::lifecycle::TransitionOptions {
                status: TRANSITION_COMPLETE.to_string(),
                tier: "tier".to_string(),
                etag: "etag".to_string(),
                expected_data_dir: Some(data_dir),
                expected_remote_name: "remote-object".to_string(),
                restore_expiry,
                expire_restored: true,
                ..Default::default()
            },
            ..Default::default()
        };

        let err = check_object_lock_delete(&bootstrap_ctx(), "bucket", "object", &obj_info, &opts)
            .await
            .expect_err("a renewed restored copy must reject the stale expiry task");
        assert!(matches!(err, StorageError::PreconditionFailed));
    }

    #[tokio::test]
    async fn test_check_object_lock_delete_allows_versioned_delete_marker_creation() {
        let retain_until = OffsetDateTime::now_utc() + Duration::from_secs(60 * 60 * 24 * 60);
        let mut user_defined = HashMap::new();
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_MODE.to_string(),
            s3s::dto::ObjectLockRetentionMode::COMPLIANCE.to_string(),
        );
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_RETAIN_UNTIL_DATE.to_string(),
            retain_until.format(&time::format_description::well_known::Rfc3339).unwrap(),
        );

        let obj_info = ObjectInfo {
            user_defined: Arc::new(user_defined),
            ..Default::default()
        };
        let opts = ObjectOptions {
            version_id: None,
            versioned: true,
            version_suspended: false,
            ..Default::default()
        };

        check_object_lock_delete(&bootstrap_ctx(), "bucket", "object", &obj_info, &opts)
            .await
            .expect("versioned delete marker creation should not delete the locked version");
    }

    fn governance_retained_obj_info() -> ObjectInfo {
        let retain_until = OffsetDateTime::now_utc() + Duration::from_secs(60 * 60 * 24 * 60);
        let mut user_defined = HashMap::new();
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_MODE.to_string(),
            s3s::dto::ObjectLockRetentionMode::GOVERNANCE.to_string(),
        );
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_RETAIN_UNTIL_DATE.to_string(),
            retain_until.format(&time::format_description::well_known::Rfc3339).unwrap(),
        );
        ObjectInfo {
            user_defined: Arc::new(user_defined),
            ..Default::default()
        }
    }

    fn explicit_version_delete_opts(replication_request: bool) -> ObjectOptions {
        ObjectOptions {
            version_id: Some(Uuid::new_v4().to_string()),
            versioned: true,
            replication_request,
            object_lock_config_snapshot: Some(Arc::new(ObjectLockConfigSnapshot::new(ObjectLockConfigState::ConfirmedAbsent))),
            ..Default::default()
        }
    }

    // Issue #6850: a replicated version purge carries no bypass header, so the
    // GOVERNANCE gate must honor the source's already-judged bypass instead of
    // keeping the sites permanently diverged.
    #[tokio::test]
    async fn test_check_object_lock_delete_allows_replicated_governance_version_purge() {
        let obj_info = governance_retained_obj_info();
        let opts = explicit_version_delete_opts(true);

        check_object_lock_delete(&bootstrap_ctx(), "bucket", "object", &obj_info, &opts)
            .await
            .expect("an authorized replicated version purge must pass GOVERNANCE retention (#6850)");
    }

    #[tokio::test]
    async fn test_check_object_lock_delete_blocks_plain_governance_version_delete_without_bypass() {
        let obj_info = governance_retained_obj_info();
        let opts = explicit_version_delete_opts(false);

        let err = check_object_lock_delete(&bootstrap_ctx(), "bucket", "object", &obj_info, &opts)
            .await
            .expect_err("a plain client delete without the bypass header must stay blocked by GOVERNANCE retention");

        assert!(matches!(err, StorageError::PrefixAccessDenied(_, _)));
    }

    #[tokio::test]
    async fn test_check_object_lock_delete_blocks_replicated_compliance_version_purge() {
        let retain_until = OffsetDateTime::now_utc() + Duration::from_secs(60 * 60 * 24 * 60);
        let mut user_defined = HashMap::new();
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_MODE.to_string(),
            s3s::dto::ObjectLockRetentionMode::COMPLIANCE.to_string(),
        );
        user_defined.insert(
            metadata_keys::OBJECT_LOCK_RETAIN_UNTIL_DATE.to_string(),
            retain_until.format(&time::format_description::well_known::Rfc3339).unwrap(),
        );
        let obj_info = ObjectInfo {
            user_defined: Arc::new(user_defined),
            ..Default::default()
        };
        let opts = explicit_version_delete_opts(true);

        let err = check_object_lock_delete(&bootstrap_ctx(), "bucket", "object", &obj_info, &opts)
            .await
            .expect_err("the source gate can never purge through COMPLIANCE, so a replicated purge fails closed");

        assert!(matches!(err, StorageError::PrefixAccessDenied(_, _)));
    }

    #[tokio::test]
    async fn test_check_object_lock_delete_blocks_replicated_legal_hold_version_purge() {
        let mut user_defined = HashMap::new();
        user_defined.insert(metadata_keys::OBJECT_LOCK_LEGAL_HOLD.to_string(), "ON".to_string());
        let obj_info = ObjectInfo {
            user_defined: Arc::new(user_defined),
            ..Default::default()
        };
        let opts = explicit_version_delete_opts(true);

        let err = check_object_lock_delete(&bootstrap_ctx(), "bucket", "object", &obj_info, &opts)
            .await
            .expect_err("the source gate can never purge through a legal hold, so a replicated purge fails closed");

        assert!(matches!(err, StorageError::PrefixAccessDenied(_, _)));
    }

    // backlog#929 (HP-8): the delete_objects per-object stat is gated on the
    // bucket object-lock configuration. Lock-enabled buckets (either legacy
    // lock_enabled flag or an enabled ObjectLockConfiguration) and unknown
    // metadata must keep the #4297 locked-stat path; only buckets that
    // provably have no Object Lock may skip it.
    #[test]
    fn test_object_lock_delete_check_required_skips_plain_buckets() {
        let bm = crate::bucket::metadata::BucketMetadata::new("plain-bucket");
        assert!(!object_lock_delete_check_required(Some(&bm)));
    }

    #[test]
    fn test_object_lock_delete_check_required_keeps_legacy_lock_enabled_buckets() {
        let mut bm = crate::bucket::metadata::BucketMetadata::new("legacy-lock-bucket");
        bm.lock_enabled = true;
        assert!(object_lock_delete_check_required(Some(&bm)));
    }

    #[test]
    fn test_object_lock_delete_check_required_keeps_object_lock_config_buckets() {
        use s3s::dto::{ObjectLockConfiguration, ObjectLockEnabled};

        let mut bm = crate::bucket::metadata::BucketMetadata::new("lock-config-bucket");
        bm.object_lock_config = Some(ObjectLockConfiguration {
            object_lock_enabled: Some(ObjectLockEnabled::from_static(ObjectLockEnabled::ENABLED)),
            ..Default::default()
        });
        assert!(object_lock_delete_check_required(Some(&bm)));
    }

    #[test]
    fn test_object_lock_delete_check_required_fails_closed_without_metadata() {
        assert!(object_lock_delete_check_required(None));
    }

    /// rustfs/backlog#1734: stored Object Lock bytes that cannot be parsed
    /// mean the lock state is unknown, not absent; the check must stay on.
    #[test]
    fn test_object_lock_delete_check_required_fails_closed_on_unreadable_lock_config() {
        let mut bm = crate::bucket::metadata::BucketMetadata::new("unreadable-lock-bucket");
        bm.object_lock_config_xml = b"<ObjectLockConfiguration>".to_vec();
        assert!(object_lock_delete_check_required(Some(&bm)));
    }

    #[test]
    fn test_should_persist_encryption_original_size_rejects_plain_metadata() {
        let metadata = HashMap::from([("content-type".to_string(), "application/octet-stream".to_string())]);

        assert!(!should_persist_encryption_original_size(&metadata));
    }

    #[test]
    fn test_should_persist_encryption_original_size_accepts_sse_c_metadata() {
        let metadata = HashMap::from([(SSEC_ALGORITHM_HEADER.to_string(), "AES256".to_string())]);

        assert!(should_persist_encryption_original_size(&metadata));
    }

    #[test]
    fn test_should_prevent_write() {
        let oi = ObjectInfo {
            etag: Some("abc".to_string()),
            ..Default::default()
        };
        let if_none_match = Some("abc".to_string());
        let if_match = None;
        assert!(should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = Some("*".to_string());
        let if_match = None;
        assert!(should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = None;
        let if_match = Some("def".to_string());
        assert!(should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = None;
        let if_match = Some("*".to_string());
        assert!(!should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = Some("def".to_string());
        let if_match = None;
        assert!(!should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = Some("def".to_string());
        let if_match = Some("*".to_string());
        assert!(!should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = Some("def".to_string());
        let if_match = Some("\"abc\"".to_string());
        assert!(!should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = Some("*".to_string());
        let if_match = Some("\"abc\"".to_string());
        assert!(should_prevent_write(&oi, if_none_match, if_match));

        let oi = ObjectInfo {
            etag: None,
            ..Default::default()
        };
        let if_none_match = Some("*".to_string());
        let if_match = Some("\"abc\"".to_string());
        assert!(should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = None;
        let if_match = None;
        assert!(!should_prevent_write(&oi, if_none_match, if_match));

        let if_none_match = Some(String::new());
        let if_match = Some(" ".to_string());
        assert!(!should_prevent_write(&oi, if_none_match, if_match));
    }

    #[test]
    fn test_is_valid_storage_class() {
        assert!(is_valid_storage_class(storageclass::STANDARD));
        assert!(is_valid_storage_class(storageclass::RRS));
        assert!(!is_valid_storage_class(storageclass::STANDARD_IA));
        assert!(!is_valid_storage_class("INVALID"));
    }

    #[test]
    fn complete_part_checksum_accepts_missing_value_and_uses_base_type() {
        let missing_checksum_part = CompletePart::default();
        assert_eq!(
            complete_part_checksum(&missing_checksum_part, rustfs_rio::ChecksumType::CRC64_NVME),
            Some(None)
        );

        let full_object_crc32 =
            rustfs_rio::ChecksumType(rustfs_rio::ChecksumType::CRC32.0 | rustfs_rio::ChecksumType::FULL_OBJECT.0);
        let part = CompletePart {
            checksum_crc32: Some("AAAAAA==".to_string()),
            ..Default::default()
        };
        assert_eq!(complete_part_checksum(&part, full_object_crc32), Some(Some("AAAAAA==".to_string())));

        // AWS 2026-04 additional algorithms have no CompletePart field, so they are
        // accepted with no double-check value (verified at UploadPart) — Some(None),
        // NOT the outer None that would fail the multipart completion (#1261).
        for ct in [
            rustfs_rio::ChecksumType::XXHASH3,
            rustfs_rio::ChecksumType::XXHASH64,
            rustfs_rio::ChecksumType::XXHASH128,
            rustfs_rio::ChecksumType::SHA512,
            rustfs_rio::ChecksumType::MD5,
        ] {
            assert_eq!(complete_part_checksum(&CompletePart::default(), ct), Some(None), "{ct:?}");
        }

        // Genuinely unset / invalid types must still be rejected (outer None).
        assert_eq!(complete_part_checksum(&CompletePart::default(), rustfs_rio::ChecksumType::NONE), None);
        assert_eq!(complete_part_checksum(&CompletePart::default(), rustfs_rio::ChecksumType::INVALID), None);
    }

    fn direct_memory_test_metadata(size: i64) -> (ObjectInfo, FileInfo, ObjectOptions) {
        let part_size = usize::try_from(size).expect("test size should fit usize");
        let part = ObjectPartInfo {
            number: 1,
            size: part_size,
            actual_size: size,
            ..Default::default()
        };
        let object_info = ObjectInfo {
            size,
            actual_size: size,
            parts: Arc::new(vec![part]),
            etag: Some("0123456789abcdef0123456789abcdef".to_string()),
            ..Default::default()
        };
        let mut fi = FileInfo::new("bucket/object", 1, 0);
        fi.size = size;
        fi.add_object_part(1, String::new(), part_size, None, size, None, None);
        (object_info, fi, ObjectOptions::default())
    }

    #[test]
    fn small_object_direct_memory_eligibility_is_conservative() {
        let (object_info, fi, opts) = direct_memory_test_metadata(1024);
        assert!(is_get_small_object_direct_memory_eligible_with_threshold(
            &None,
            &object_info,
            &fi,
            &opts,
            128 * 1024
        ));

        assert!(!is_get_small_object_direct_memory_eligible_with_threshold(
            &Some(HTTPRangeSpec {
                start: 0,
                end: 10,
                is_suffix_length: false,
            }),
            &object_info,
            &fi,
            &opts,
            128 * 1024
        ));

        let mut part_opts = opts.clone();
        part_opts.part_number = Some(1);
        assert!(!is_get_small_object_direct_memory_eligible_with_threshold(
            &None,
            &object_info,
            &fi,
            &part_opts,
            128 * 1024
        ));

        // Bucket-level versioning no longer blocks the inline path (rustfs/backlog#1802):
        // a latest-version read on a versioned bucket is eligible.
        let mut versioned_opts = opts.clone();
        versioned_opts.versioned = true;
        assert!(is_get_small_object_direct_memory_eligible_with_threshold(
            &None,
            &object_info,
            &fi,
            &versioned_opts,
            128 * 1024
        ));

        let mut remote = object_info;
        remote.transitioned_object.status = TRANSITION_COMPLETE.to_string();
        remote.transitioned_object.tier = "remote-tier".to_string();
        assert!(!is_get_small_object_direct_memory_eligible_with_threshold(
            &None,
            &remote,
            &fi,
            &opts,
            128 * 1024
        ));
    }

    #[test]
    fn read_path_shape_rejects_mismatched_metadata_copies_and_transforms() {
        let (object_info, fi, _) = direct_memory_test_metadata(1024);
        let shape = read_path_shape(&object_info, &fi).expect("matching metadata should have a valid shape");
        assert_eq!(shape.object_size, 1024);
        assert!(shape.is_plain(&object_info, &fi));

        let mut bad_part_number = object_info.clone();
        Arc::make_mut(&mut bad_part_number.parts)[0].number = 2;
        assert!(read_path_shape(&bad_part_number, &fi).is_none());

        let mut bad_part_size = fi.clone();
        bad_part_size.parts[0].size += 1;
        assert!(read_path_shape(&object_info, &bad_part_size).is_none());

        let mut bad_actual_size = object_info.clone();
        Arc::make_mut(&mut bad_actual_size.parts)[0].actual_size += 1;
        assert!(read_path_shape(&bad_actual_size, &fi).is_none());

        let mut compressed = fi.clone();
        insert_str(&mut compressed.metadata, SUFFIX_COMPRESSION, "zstd".to_string());
        let compressed_shape = read_path_shape(&object_info, &compressed).expect("geometry remains valid");
        assert!(!compressed_shape.is_plain(&object_info, &compressed));

        let mut encrypted = fi;
        encrypted
            .metadata
            .insert("x-amz-server-side-encryption".to_string(), "AES256".to_string());
        assert!(
            !read_path_shape(&object_info, &encrypted)
                .expect("geometry remains valid")
                .is_plain(&object_info, &encrypted)
        );
    }

    #[test]
    fn inline_fast_path_rejects_part_number_requests() {
        let (mut object_info, mut fi, opts) = direct_memory_test_metadata(1024);
        object_info.inlined = true;
        fi.data = Some(Bytes::from(vec![0; 1024]));

        assert!(should_use_inline_fast_path(&None, &object_info, &fi, &opts));

        let mut part_opts = opts;
        part_opts.part_number = Some(1);
        assert!(!should_use_inline_fast_path(&None, &object_info, &fi, &part_opts));
    }

    #[test]
    fn inline_fast_path_allows_layout_specific_ec8_and_ec12_256kib_objects() {
        for (data_blocks, parity_blocks) in [(8, 4), (12, 4)] {
            let object_size = 256 * 1024_i64;
            let mut fi = FileInfo::new("bucket/object", data_blocks, parity_blocks);
            fi.size = object_size;
            fi.data = Some(Bytes::from_static(b"payload"));
            fi.add_object_part(1, String::new(), object_size as usize, None, object_size, None, None);
            let object_info = ObjectInfo {
                size: object_size,
                data_blocks,
                parity_blocks,
                inlined: true,
                parts: Arc::new(fi.parts.clone()),
                ..Default::default()
            };

            assert!(should_use_inline_fast_path(&None, &object_info, &fi, &ObjectOptions::default()));
        }
    }

    #[test]
    fn inline_fast_path_rejects_oversized_or_corrupt_inline_markers() {
        for (data_blocks, parity_blocks) in [(1, 0), (8, 4), (12, 4)] {
            let object_size = 4 * 1024 * 1024_i64;
            let mut fi = FileInfo::new("bucket/object", data_blocks, parity_blocks);
            fi.size = object_size;
            fi.data = Some(Bytes::from_static(b"payload"));
            fi.add_object_part(1, String::new(), object_size as usize, None, object_size, None, None);
            let object_info = ObjectInfo {
                size: object_size,
                data_blocks,
                parity_blocks,
                inlined: true,
                parts: Arc::new(fi.parts.clone()),
                ..Default::default()
            };

            assert!(
                object_info.is_inline_fast_path_eligible(),
                "the marker and object shape alone must not be the final trust boundary"
            );
            assert!(!should_use_inline_fast_path(&None, &object_info, &fi, &ObjectOptions::default()));
        }

        // A malformed erasure geometry must fail closed before shard-size
        // arithmetic, even when an inline marker and payload are present.
        let (mut object_info, mut fi, opts) = direct_memory_test_metadata(4 * 1024 * 1024);
        object_info.inlined = true;
        fi.data = Some(Bytes::from_static(b"payload"));
        fi.erasure.data_blocks = 0;
        assert!(!should_use_inline_fast_path(&None, &object_info, &fi, &opts));

        // A mismatched plain part size must not allow the large object size to
        // reach the inline decoder's `Vec::with_capacity` allocation.
        let (mut object_info, mut fi, opts) = direct_memory_test_metadata(4 * 1024 * 1024);
        object_info.inlined = true;
        fi.data = Some(Bytes::from_static(b"payload"));
        fi.parts[0].actual_size = 0;
        assert!(!should_use_inline_fast_path(&None, &object_info, &fi, &opts));
    }

    #[test]
    fn small_object_direct_memory_decision_reports_bounded_reasons() {
        let (object_info, fi, opts) = direct_memory_test_metadata(1024);

        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &object_info, &fi, &opts, false, 128 * 1024),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Disabled)
        );
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &object_info, &fi, &opts, true, 0),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::ThresholdZero)
        );
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(
                &Some(HTTPRangeSpec {
                    start: 0,
                    end: 10,
                    is_suffix_length: false,
                }),
                &object_info,
                &fi,
                &opts,
                true,
                128 * 1024
            ),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Range)
        );

        // Bucket-level versioning no longer falls back (rustfs/backlog#1802): the
        // latest version on a versioned bucket is served inline like any other.
        let mut versioned_opts = opts.clone();
        versioned_opts.versioned = true;
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &object_info, &fi, &versioned_opts, true, 128 * 1024),
            GetDirectMemoryDecision::Use { object_size: 1024 }
        );

        let mut encrypted = object_info.clone();
        Arc::make_mut(&mut encrypted.user_defined).insert("x-amz-server-side-encryption".to_string(), "AES256".to_string());
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &encrypted, &fi, &opts, true, 128 * 1024),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Encrypted)
        );

        let mut remote = object_info.clone();
        remote.transitioned_object.status = TRANSITION_COMPLETE.to_string();
        remote.transitioned_object.tier = "remote-tier".to_string();
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &remote, &fi, &opts, true, 128 * 1024),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::Remote)
        );

        let mut multipart = object_info.clone();
        multipart.parts = Arc::new(vec![ObjectPartInfo::default(), ObjectPartInfo::default()]);
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &multipart, &fi, &opts, true, 128 * 1024),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::ObjectInfoMultipart)
        );
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &object_info, &fi, &opts, true, 512),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::AboveThreshold)
        );

        let mut size_mismatch = object_info.clone();
        size_mismatch.size += 1;
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &size_mismatch, &fi, &opts, true, 128 * 1024),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::SizeMismatch)
        );
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &object_info, &fi, &opts, true, 128 * 1024),
            GetDirectMemoryDecision::Use { object_size: 1024 }
        );

        let mut corrupt_part = fi;
        corrupt_part.parts[0].actual_size += 1;
        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &object_info, &corrupt_part, &opts, true, 128 * 1024),
            GetDirectMemoryDecision::Fallback(GetDirectMemoryFallbackReason::MetadataShape)
        );
    }

    #[test]
    fn direct_memory_fallback_metric_labels_are_stable() {
        assert_eq!(GetDirectMemoryFallbackReason::Disabled.as_str(), "disabled");
        assert_eq!(GetDirectMemoryFallbackReason::ThresholdZero.as_str(), "threshold_zero");
        assert_eq!(GetDirectMemoryFallbackReason::Range.as_str(), "range");
        assert_eq!(GetDirectMemoryFallbackReason::PartNumber.as_str(), "part_number");
        assert_eq!(GetDirectMemoryFallbackReason::VersionId.as_str(), "version_id");
        assert_eq!(GetDirectMemoryFallbackReason::InclFreeVersions.as_str(), "incl_free_versions");
        assert_eq!(GetDirectMemoryFallbackReason::SkipFreeVersion.as_str(), "skip_free_version");
        assert_eq!(GetDirectMemoryFallbackReason::DataMovement.as_str(), "data_movement");
        assert_eq!(GetDirectMemoryFallbackReason::RawDataMovementRead.as_str(), "raw_data_movement_read");
        assert_eq!(GetDirectMemoryFallbackReason::DeleteMarker.as_str(), "delete_marker");
        assert_eq!(GetDirectMemoryFallbackReason::MetadataOnly.as_str(), "metadata_only");
        assert_eq!(GetDirectMemoryFallbackReason::VersionOnly.as_str(), "version_only");
        assert_eq!(GetDirectMemoryFallbackReason::Encrypted.as_str(), "encrypted");
        assert_eq!(GetDirectMemoryFallbackReason::Compressed.as_str(), "compressed");
        assert_eq!(GetDirectMemoryFallbackReason::Remote.as_str(), "remote");
        assert_eq!(GetDirectMemoryFallbackReason::ObjectInfoMultipart.as_str(), "object_info_multipart");
        assert_eq!(GetDirectMemoryFallbackReason::FileInfoMultipart.as_str(), "file_info_multipart");
        assert_eq!(GetDirectMemoryFallbackReason::InvalidSize.as_str(), "invalid_size");
        assert_eq!(GetDirectMemoryFallbackReason::SizeMismatch.as_str(), "size_mismatch");
        assert_eq!(GetDirectMemoryFallbackReason::AboveThreshold.as_str(), "above_threshold");
    }

    #[test]
    fn small_object_direct_memory_eligibility_respects_threshold_and_shape() {
        let (object_info, fi, opts) = direct_memory_test_metadata(128 * 1024);
        assert!(is_get_small_object_direct_memory_eligible_with_threshold(
            &None,
            &object_info,
            &fi,
            &opts,
            128 * 1024
        ));
        assert!(!is_get_small_object_direct_memory_eligible_with_threshold(
            &None,
            &object_info,
            &fi,
            &opts,
            (128 * 1024) - 1
        ));

        let mut multipart = object_info;
        multipart.parts = Arc::new(vec![ObjectPartInfo::default(), ObjectPartInfo::default()]);
        assert!(!is_get_small_object_direct_memory_eligible_with_threshold(
            &None,
            &multipart,
            &fi,
            &opts,
            128 * 1024
        ));
    }

    async fn inline_bitrot_files_for_payload_with_mode(
        payload: &[u8],
        uses_legacy: bool,
    ) -> (coding::Erasure, Vec<FileInfo>, usize, HashAlgorithm) {
        let erasure = coding::Erasure::new_with_options(4, 2, 1024 * 1024, uses_legacy);
        let read_length = erasure.shard_file_offset(0, payload.len(), payload.len());
        let checksum_algo = if uses_legacy {
            HashAlgorithm::HighwayHash256SLegacy
        } else {
            HashAlgorithm::HighwayHash256S
        };
        let shards = erasure.encode_data(payload).expect("payload should encode");
        let version_id = Some(Uuid::new_v4());
        let data_dir = Some(Uuid::new_v4());
        let mod_time = Some(OffsetDateTime::now_utc());
        let mut files = Vec::with_capacity(shards.len());

        for shard in shards {
            let mut writer = coding::BitrotWriterWrapper::new(
                coding::CustomWriter::new_inline_buffer(),
                erasure.shard_size(),
                checksum_algo.clone(),
            );
            writer.write(&shard).await.expect("inline shard should write");
            writer.shutdown().await.expect("inline writer should shutdown");
            let data = writer.into_inline_data().expect("inline data should be retained");
            let mut file = FileInfo::new("bucket/object", erasure.data_shards, erasure.parity_shards);
            file.volume = "bucket".to_string();
            file.name = "object".to_string();
            file.size = i64::try_from(payload.len()).expect("test payload should fit i64");
            file.is_latest = true;
            file.version_id = version_id;
            file.data_dir = data_dir;
            file.mod_time = mod_time;
            file.metadata.insert("etag".to_string(), "etag-inline".to_string());
            file.add_object_part(1, "part-etag-inline".to_string(), payload.len(), file.mod_time, file.size, None, None);
            file.set_inline_data();
            file.erasure.index = files.len() + 1;
            file.erasure.block_size = erasure.block_size;
            file.uses_legacy_checksum = uses_legacy;
            file.data = Some(Bytes::from(data));
            files.push(file);
        }

        (erasure, files, read_length, checksum_algo)
    }

    async fn inline_bitrot_files_for_payload(payload: &[u8]) -> (coding::Erasure, Vec<FileInfo>, usize, HashAlgorithm) {
        inline_bitrot_files_for_payload_with_mode(payload, false).await
    }

    #[tokio::test]
    async fn deep_heal_inline_bitrot_checks_current_and_legacy_frames() {
        let payload = vec![0x7b; 4113];
        for legacy in [false, true] {
            let (_, files, _, _) = inline_bitrot_files_for_payload_with_mode(&payload, legacy).await;
            for mut meta in files {
                assert!(meta.erasure.checksums.is_empty(), "legacy metadata may omit per-part checksum entries");
                verify_inline_part_bitrot(&meta)
                    .await
                    .expect("healthy data and parity shards should verify");
                let original = meta.data.clone().expect("fixture should retain inline bytes");
                for offset in [0, 32, original.len() - 1] {
                    let mut damaged = original.to_vec();
                    damaged[offset] ^= 1;
                    meta.data = Some(Bytes::from(damaged));
                    assert_eq!(verify_inline_part_bitrot(&meta).await, Err(DiskError::FileCorrupt));
                }
                let mut trailing = original.to_vec();
                trailing.push(0x7b);
                for damaged in [Vec::new(), original[..original.len() - 1].to_vec(), trailing] {
                    meta.data = Some(Bytes::from(damaged));
                    assert_eq!(verify_inline_part_bitrot(&meta).await, Err(DiskError::FileCorrupt));
                }
            }
        }
    }

    #[tokio::test]
    async fn deep_heal_inline_bitrot_checks_empty_objects_and_metadata_bounds() {
        let (_, files, _, _) = inline_bitrot_files_for_payload(b"inline metadata bounds").await;
        let mut empty = files[0].clone();
        empty.size = 0;
        empty.parts[0].size = 0;
        empty.parts[0].actual_size = 0;
        empty.data = Some(Bytes::new());
        verify_inline_part_bitrot(&empty)
            .await
            .expect("empty inline objects have no bitrot frames");
        empty.data = Some(Bytes::from_static(b"unexpected"));
        assert_eq!(verify_inline_part_bitrot(&empty).await, Err(DiskError::FileCorrupt));

        let mut invalid = files[0].clone();
        invalid.erasure.block_size = 0;
        assert_eq!(verify_inline_part_bitrot(&invalid).await, Err(DiskError::FileCorrupt));
        invalid = files[0].clone();
        invalid.parts.push(invalid.parts[0].clone());
        invalid.parts[1].number = 2;
        assert_eq!(verify_inline_part_bitrot(&invalid).await, Err(DiskError::FileCorrupt));

        let mut oversized = files[0].clone();
        oversized.erasure.block_size = 2;
        oversized.size = i64::MAX - 3;
        oversized.parts[0].size = usize::try_from(oversized.size).expect("64-bit metadata size should fit");
        oversized
            .validate_for_metadata_read()
            .expect("logical shard length should fit metadata bounds");
        assert_eq!(verify_inline_part_bitrot(&oversized).await, Err(DiskError::FileCorrupt));

        oversized.erasure.block_size = 1 << 40;
        oversized.size = 1 << 40;
        oversized.parts[0].size = usize::try_from(oversized.size).expect("64-bit metadata size should fit");
        oversized
            .validate_for_metadata_read()
            .expect("large metadata geometry should remain representable");
        assert_eq!(verify_inline_part_bitrot(&oversized).await, Err(DiskError::FileCorrupt));
    }

    #[tokio::test]
    async fn deep_heal_inline_bitrot_accepts_pinned_disk_fixtures() {
        use rustfs_filemeta::test_data::{create_issue_2265_legacy_meta_v2_object_xlmeta, create_issue_2288_legacy_xlmeta};

        for (bytes, size, legacy) in [
            (create_issue_2288_legacy_xlmeta().expect("pinned meta v1 fixture"), 35, false),
            (
                create_issue_2265_legacy_meta_v2_object_xlmeta().expect("pinned meta v2 fixture"),
                707,
                true,
            ),
        ] {
            let file_meta = rustfs_filemeta::FileMeta::load(&bytes).expect("historical xl.meta should decode");
            let mut meta = file_meta
                .into_fileinfo("bucket", "object", "", true, false, true)
                .expect("historical object metadata should decode");
            assert_eq!(meta.size, size);
            assert_eq!(meta.uses_legacy_checksum, legacy);
            assert!(meta.inline_data());
            assert!(meta.data.is_some(), "the production decoder should extract the historical inline value");
            verify_inline_part_bitrot(&meta)
                .await
                .expect("historical inline shard should pass deep verification");
            let mut damaged = meta.data.as_ref().expect("fixture should be inline").to_vec();
            *damaged.last_mut().expect("fixture shard should not be empty") ^= 1;
            meta.data = Some(Bytes::from(damaged));
            assert_eq!(verify_inline_part_bitrot(&meta).await, Err(DiskError::FileCorrupt));
        }
    }

    fn disk_ordered_fileinfos(files: &[FileInfo]) -> Vec<FileInfo> {
        let distribution = &files
            .first()
            .expect("inline data shard fixture should include metadata")
            .erasure
            .distribution;
        distribution
            .iter()
            .map(|block_index| {
                files
                    .get(block_index.checked_sub(1).expect("erasure block indexes are one-based"))
                    .expect("inline data shard fixture should include every distributed shard")
                    .clone()
            })
            .collect()
    }

    fn inline_data_shard_fileinfo(
        data_blocks: usize,
        parity_blocks: usize,
        erasure_index: usize,
        distribution: &[usize],
        data: Option<&'static [u8]>,
    ) -> FileInfo {
        let mut fi = FileInfo::new("object", data_blocks, parity_blocks);
        fi.name = "object".to_string();
        fi.volume = "bucket".to_string();
        fi.size = 4;
        fi.is_latest = true;
        fi.data_dir = Some(Uuid::nil());
        fi.mod_time = Some(OffsetDateTime::UNIX_EPOCH);
        fi.metadata.insert("etag".to_string(), "etag-inline".to_string());
        fi.add_object_part(1, "part-etag-inline".to_string(), 4, fi.mod_time, 4, None, None);
        fi.set_inline_data();
        fi.erasure.index = erasure_index;
        fi.erasure.distribution = distribution.to_vec();
        fi.data = data.map(Bytes::from_static);
        fi
    }

    #[test]
    fn collect_inline_data_shards_by_index_uses_distribution_order() {
        let distribution = vec![3, 1, 5, 2, 4, 6];
        let mut fi = inline_data_shard_fileinfo(4, 2, 1, &distribution, Some(b"x"));
        fi.erasure.index = 1;
        fi.erasure.distribution = distribution.clone();
        let files = vec![
            inline_data_shard_fileinfo(4, 2, 3, &distribution, Some(b"c")),
            inline_data_shard_fileinfo(4, 2, 1, &distribution, Some(b"a")),
            inline_data_shard_fileinfo(4, 2, 5, &distribution, Some(b"p")),
            inline_data_shard_fileinfo(4, 2, 2, &distribution, Some(b"b")),
            inline_data_shard_fileinfo(4, 2, 4, &distribution, Some(b"d")),
            inline_data_shard_fileinfo(4, 2, 6, &distribution, Some(b"q")),
        ];

        let data_files =
            collect_inline_data_shard_fileinfos_by_index(&files, &fi, 4, |_| true).expect("all data shards should be collected");

        assert_eq!(
            data_files
                .iter()
                .map(|file| file.data.as_deref().expect("fixture carries inline bytes"))
                .collect::<Vec<_>>(),
            [b"a".as_slice(), b"b".as_slice(), b"c".as_slice(), b"d".as_slice()]
        );
    }

    #[test]
    fn collect_inline_data_shards_by_index_rejects_missing_data_shard() {
        let distribution = vec![1, 2, 3, 4];
        let mut fi = inline_data_shard_fileinfo(2, 2, 1, &distribution, Some(b"x"));
        fi.erasure.index = 1;
        fi.erasure.distribution = distribution.clone();
        let files = vec![
            inline_data_shard_fileinfo(2, 2, 1, &distribution, Some(b"a")),
            inline_data_shard_fileinfo(2, 2, 2, &distribution, None),
            inline_data_shard_fileinfo(2, 2, 3, &distribution, Some(b"p")),
            inline_data_shard_fileinfo(2, 2, 4, &distribution, Some(b"q")),
        ];

        assert!(collect_inline_data_shard_fileinfos_by_index(&files, &fi, 2, |_| true).is_none());
    }

    #[tokio::test]
    async fn inline_data_shards_direct_read_reassembles_payload() {
        let payload = b"small inline object payload that spans data shards";
        let (erasure, files, read_length, checksum_algo) = inline_bitrot_files_for_payload(payload).await;
        let mut readers = build_inline_bitrot_readers(
            &files,
            erasure.data_shards,
            "bucket",
            "object",
            read_length,
            erasure.shard_size(),
            &checksum_algo,
            false,
        )
        .await
        .expect("inline bitrot readers should build");
        assert_eq!(readers.len(), erasure.data_shards);

        let body = try_read_inline_data_shards_direct(&mut readers, erasure.data_shards, read_length, payload.len())
            .await
            .expect("data shard direct read should succeed");

        assert_eq!(body.as_ref(), payload);
    }

    #[tokio::test]
    async fn inline_data_shards_direct_read_reassembles_legacy_payload_with_padding() {
        let payload = b"legacy inline payload whose size is not divisible by the data shard count";
        let (erasure, files, read_length, checksum_algo) = inline_bitrot_files_for_payload_with_mode(payload, true).await;
        assert_ne!(payload.len() % erasure.data_shards, 0, "test payload must exercise EC padding");
        let mut readers = build_inline_bitrot_readers(
            &files,
            erasure.data_shards,
            "bucket",
            "object",
            read_length,
            erasure.shard_size(),
            &checksum_algo,
            false,
        )
        .await
        .expect("legacy inline bitrot readers should build");

        let body = try_read_inline_data_shards_direct(&mut readers, erasure.data_shards, read_length, payload.len())
            .await
            .expect("legacy data shard direct read should succeed");

        assert_eq!(body.len(), payload.len());
        assert_eq!(body.as_ref(), payload);
    }

    #[tokio::test]
    async fn inline_data_shards_direct_read_rejects_corrupt_shard() {
        let payload = b"small inline object payload that will be corrupted";
        let (erasure, mut files, read_length, checksum_algo) = inline_bitrot_files_for_payload(payload).await;
        let second = files[1].data.as_mut().expect("second shard should exist");
        let mut corrupted = second.to_vec();
        let last = corrupted.last_mut().expect("encoded shard should not be empty");
        *last ^= 0xff;
        *second = Bytes::from(corrupted);

        let mut readers = build_inline_bitrot_readers(
            &files,
            erasure.total_shard_count(),
            "bucket",
            "object",
            read_length,
            erasure.shard_size(),
            &checksum_algo,
            false,
        )
        .await
        .expect("inline bitrot readers should build");

        let body = try_read_inline_data_shards_direct(&mut readers, 4, read_length, payload.len()).await;

        assert!(body.is_none(), "a later corrupt shard must discard the already-appended body prefix");
    }

    #[test]
    fn inline_data_shards_direct_read_requires_single_block() {
        assert!(can_try_inline_data_shards_direct(1024, 1024));
        assert!(!can_try_inline_data_shards_direct(0, 1024));
        assert!(!can_try_inline_data_shards_direct(1025, 1024));
    }

    #[test]
    fn inline_erasure_offset_helpers_match_erasure_methods() {
        for uses_legacy in [false, true] {
            let erasure = coding::Erasure::new_with_options(4, 2, 1024 * 1024, uses_legacy);
            for object_size in [1usize, 1024, 100 * 1024, 1024 * 1024] {
                assert_eq!(
                    inline_erasure_shard_size(erasure.block_size, erasure.data_shards, uses_legacy),
                    erasure.shard_size()
                );
                assert_eq!(
                    inline_erasure_shard_file_offset(
                        0,
                        object_size,
                        object_size,
                        erasure.block_size,
                        erasure.data_shards,
                        uses_legacy,
                    ),
                    erasure.shard_file_offset(0, object_size, object_size)
                );
            }
        }
    }

    #[tokio::test]
    async fn direct_memory_inline_data_shards_direct_read_reassembles_single_block_payload() {
        let tempdir = tempfile::tempdir().expect("tempdir should be created");
        let endpoint =
            Endpoint::try_from(tempdir.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");

        let payload = vec![b'i'; 192 * 1024];
        let (erasure, files, _read_length, _checksum_algo) = inline_bitrot_files_for_payload(&payload).await;
        let fi = files[0].clone();
        let disk_files = disk_ordered_fileinfos(&files);

        let disks = vec![Some(disk); erasure.total_shard_count()];
        let metrics_size_bucket = rustfs_io_metrics::get_object_size_bucket(fi.size);

        let body = SetDisks::try_get_object_direct_data_shards_with_fileinfo(
            "bucket",
            "object",
            Arc::new(ErasureCache::new()),
            &fi,
            &disk_files,
            &disks,
            true,
            GET_CODEC_STREAMING_OBJECT_CLASS_PLAIN_SINGLE_PART,
            metrics_size_bucket,
        )
        .await
        .expect("direct-memory inline data shard read should not fail")
        .expect("inline data shard path should be used");

        assert_eq!(body.as_ref(), payload);
    }

    #[tokio::test]
    async fn direct_memory_versioned_bucket_uses_inline_data_shards_for_latest() {
        let tempdir = tempfile::tempdir().expect("tempdir should be created");
        let endpoint =
            Endpoint::try_from(tempdir.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");

        let payload = vec![b'v'; 64 * 1024];
        let payload_size = i64::try_from(payload.len()).expect("test payload size should fit i64");
        let (erasure, files, _read_length, _checksum_algo) = inline_bitrot_files_for_payload(&payload).await;
        let fi = files[0].clone();
        let disk_files = disk_ordered_fileinfos(&files);

        let mut object_info = ObjectInfo {
            size: payload_size,
            actual_size: payload_size,
            parts: Arc::new(vec![ObjectPartInfo {
                number: 1,
                size: payload.len(),
                actual_size: payload_size,
                ..Default::default()
            }]),
            ..Default::default()
        };
        object_info.inlined = true;
        let opts = ObjectOptions {
            versioned: true,
            ..Default::default()
        };
        let metrics_size_bucket = rustfs_io_metrics::get_object_size_bucket(fi.size);

        assert_eq!(
            get_small_object_direct_memory_decision_with_threshold(&None, &object_info, &fi, &opts, true, 128 * 1024),
            GetDirectMemoryDecision::Use {
                object_size: payload.len()
            }
        );

        let body = SetDisks::try_get_object_direct_data_shards_with_fileinfo(
            "bucket",
            "object",
            Arc::new(ErasureCache::new()),
            &fi,
            &disk_files,
            &vec![Some(disk); erasure.total_shard_count()],
            true,
            GET_CODEC_STREAMING_OBJECT_CLASS_PLAIN_SINGLE_PART,
            metrics_size_bucket,
        )
        .await
        .expect("versioned latest direct-memory read should not fail")
        .expect("versioned latest should use inline data shards");

        assert_eq!(body.as_ref(), payload);
    }

    #[tokio::test]
    async fn direct_memory_data_shards_direct_read_reassembles_single_block_payload() {
        use uuid::Uuid;

        let tempdir = tempfile::tempdir().expect("tempdir should be created");
        let endpoint =
            Endpoint::try_from(tempdir.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");

        let bucket = "bucket";
        let object = "object";
        let payload = vec![b'd'; 192 * 1024];

        disk.make_volume(bucket).await.expect("bucket should be created");

        let mut fi = FileInfo::new(&format!("{bucket}/{object}"), 1, 0);
        let data_dir = Uuid::new_v4();
        fi.data_dir = Some(data_dir);
        fi.size = payload.len() as i64;
        fi.add_object_part(1, String::new(), payload.len(), None, payload.len() as i64, None, None);

        let erasure = coding::Erasure::new_with_options(
            fi.erasure.data_blocks,
            fi.erasure.parity_blocks,
            fi.erasure.block_size,
            fi.uses_legacy_checksum,
        );
        let shard_path = format!("{object}/{data_dir}/part.1");
        let checksum_info = fi.erasure.get_checksum_info(1);

        let mut bitrot_writer = create_bitrot_writer(
            true,
            None,
            bucket,
            &shard_path,
            payload.len() as i64,
            erasure.shard_size(),
            checksum_info.algorithm.clone(),
        )
        .await
        .expect("bitrot writer should be created");

        for chunk in payload.chunks(erasure.shard_size()) {
            bitrot_writer.write(chunk).await.expect("payload chunk should be written");
        }

        let encoded = bitrot_writer.into_inline_data().expect("bitrot encoded data should exist");
        disk.write_all(bucket, &shard_path, Bytes::from(encoded))
            .await
            .expect("encoded shard should be stored");

        let files = vec![fi.clone()];
        let disks = vec![Some(disk)];
        let metrics_size_bucket = rustfs_io_metrics::get_object_size_bucket(fi.size);

        let body = SetDisks::try_get_object_direct_data_shards_with_fileinfo(
            bucket,
            object,
            Arc::new(ErasureCache::new()),
            &fi,
            &files,
            &disks,
            true,
            GET_CODEC_STREAMING_OBJECT_CLASS_PLAIN_SINGLE_PART,
            metrics_size_bucket,
        )
        .await
        .expect("direct-memory data shard read should not fail")
        .expect("single-block data shard path should be used");

        assert_eq!(body.as_ref(), payload);
    }

    #[tokio::test]
    async fn range_reads_use_shard_span_length_for_non_zero_offsets() {
        use tokio::io::AsyncReadExt;
        use uuid::Uuid;

        let tempdir = tempfile::tempdir().expect("tempdir should be created");
        let endpoint =
            Endpoint::try_from(tempdir.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");

        let bucket = "bucket";
        let object = "object";
        let payload = vec![b'x'; 3 * 1024 * 1024 + 1234];
        let range_offset = 2 * 1024 * 1024 + 17;
        let range_length = 512 * 1024;

        disk.make_volume(bucket).await.expect("bucket should be created");

        let mut fi = FileInfo::new(&format!("{bucket}/{object}"), 1, 0);
        let data_dir = Uuid::new_v4();
        fi.data_dir = Some(data_dir);
        fi.size = payload.len() as i64;
        fi.add_object_part(1, String::new(), payload.len(), None, payload.len() as i64, None, None);

        let erasure = coding::Erasure::new_with_options(
            fi.erasure.data_blocks,
            fi.erasure.parity_blocks,
            fi.erasure.block_size,
            fi.uses_legacy_checksum,
        );
        let shard_path = format!("{object}/{data_dir}/part.1");
        let checksum_info = fi.erasure.get_checksum_info(1);

        let mut bitrot_writer = create_bitrot_writer(
            true,
            None,
            bucket,
            &shard_path,
            payload.len() as i64,
            erasure.shard_size(),
            checksum_info.algorithm.clone(),
        )
        .await
        .expect("bitrot writer should be created");

        for chunk in payload.chunks(erasure.shard_size()) {
            bitrot_writer.write(chunk).await.expect("payload chunk should be written");
        }

        let encoded = bitrot_writer.into_inline_data().expect("bitrot encoded data should exist");
        disk.write_all(bucket, &shard_path, Bytes::from(encoded))
            .await
            .expect("encoded shard should be stored");

        let files = vec![fi.clone()];
        let disks = vec![Some(disk.clone())];
        let (mut reader, mut writer) = tokio::io::duplex(range_length * 2);
        let metrics_size_bucket = rustfs_io_metrics::get_object_size_bucket(fi.size);

        let read_task = tokio::spawn(async move {
            SetDisks::get_object_with_fileinfo(
                bucket,
                object,
                Arc::new(ErasureCache::new()),
                range_offset,
                range_length as i64,
                &mut writer,
                fi,
                files,
                &disks,
                0,
                0,
                true,
                false,
                false,
                false,
                GET_OBJECT_PATH_LEGACY_DUPLEX,
                GET_CODEC_STREAMING_OBJECT_CLASS_PLAIN_SINGLE_PART,
                metrics_size_bucket,
            )
            .await
        });

        let mut out = Vec::new();
        reader.read_to_end(&mut out).await.expect("range bytes should be readable");

        read_task
            .await
            .expect("read task should complete")
            .expect("range read should succeed");

        assert_eq!(out, payload[range_offset..range_offset + range_length]);
    }

    #[tokio::test]
    async fn multipart_reads_stream_all_parts_with_setup_prefetch() {
        use tokio::io::AsyncReadExt;
        use uuid::Uuid;

        let tempdir = tempfile::tempdir().expect("tempdir should be created");
        let endpoint =
            Endpoint::try_from(tempdir.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");

        let bucket = "bucket";
        let object = "object";
        // Three parts with distinct fill bytes so cross-part ordering bugs and
        // prefetch boundary mistakes surface as content mismatches
        // (backlog#870 exercises the prefetch hit path for parts 2 and 3).
        let parts: Vec<Vec<u8>> = vec![
            vec![b'a'; 2 * 1024 * 1024 + 111],
            vec![b'b'; 1024 * 1024 + 17],
            vec![b'c'; 3 * 1024 * 1024 + 923],
        ];
        let total_size: usize = parts.iter().map(|part| part.len()).sum();

        disk.make_volume(bucket).await.expect("bucket should be created");

        let mut fi = FileInfo::new(&format!("{bucket}/{object}"), 1, 0);
        let data_dir = Uuid::new_v4();
        fi.data_dir = Some(data_dir);
        fi.size = total_size as i64;
        for (index, part) in parts.iter().enumerate() {
            fi.add_object_part(index + 1, String::new(), part.len(), None, part.len() as i64, None, None);
        }

        let erasure = coding::Erasure::new_with_options(
            fi.erasure.data_blocks,
            fi.erasure.parity_blocks,
            fi.erasure.block_size,
            fi.uses_legacy_checksum,
        );

        for (index, payload) in parts.iter().enumerate() {
            let part_number = index + 1;
            let shard_path = format!("{object}/{data_dir}/part.{part_number}");
            let checksum_info = fi.erasure.get_checksum_info(part_number);

            let mut bitrot_writer = create_bitrot_writer(
                true,
                None,
                bucket,
                &shard_path,
                payload.len() as i64,
                erasure.shard_size(),
                checksum_info.algorithm.clone(),
            )
            .await
            .expect("bitrot writer should be created");

            for chunk in payload.chunks(erasure.shard_size()) {
                bitrot_writer.write(chunk).await.expect("payload chunk should be written");
            }

            let encoded = bitrot_writer.into_inline_data().expect("bitrot encoded data should exist");
            disk.write_all(bucket, &shard_path, Bytes::from(encoded))
                .await
                .expect("encoded shard should be stored");
        }

        let files = vec![fi.clone()];
        let disks = vec![Some(disk.clone())];
        let (mut reader, mut writer) = tokio::io::duplex(64 * 1024);
        let metrics_size_bucket = rustfs_io_metrics::get_object_size_bucket(fi.size);

        let read_task = tokio::spawn(async move {
            SetDisks::get_object_with_fileinfo(
                bucket,
                object,
                Arc::new(ErasureCache::new()),
                0,
                total_size as i64,
                &mut writer,
                fi,
                files,
                &disks,
                0,
                0,
                true,
                false,
                false,
                false,
                GET_OBJECT_PATH_LEGACY_DUPLEX,
                GET_CODEC_STREAMING_OBJECT_CLASS_PLAIN_SINGLE_PART,
                metrics_size_bucket,
            )
            .await
        });

        let mut out = Vec::new();
        reader.read_to_end(&mut out).await.expect("all part bytes should be readable");

        read_task
            .await
            .expect("read task should complete")
            .expect("multipart read should succeed");

        let expected: Vec<u8> = parts.concat();
        assert_eq!(out.len(), expected.len(), "all parts should be streamed");
        assert_eq!(out, expected, "part contents and ordering must survive setup prefetch");
    }

    #[test]
    fn parts_after_marker_uses_marker_position() {
        let part_numbers = (1..=1002).collect::<Vec<_>>();

        let remaining = parts_after_marker(&part_numbers, 1000).expect("marker should exist");

        assert_eq!(remaining, &[1001, 1002]);
    }

    #[test]
    fn parts_after_marker_returns_none_for_missing_marker() {
        let part_numbers = vec![1, 2, 3];

        assert!(parts_after_marker(&part_numbers, 4).is_none());
    }

    #[test]
    fn parts_after_marker_uses_exclusive_numeric_boundary_for_sparse_parts() {
        let part_numbers = [1, 3, 10];
        for (marker, expected) in [
            (0, Some(&part_numbers[..])),
            (1, Some(&part_numbers[1..])),
            (2, Some(&part_numbers[1..])),
            (3, Some(&part_numbers[2..])),
            (9, Some(&part_numbers[2..])),
            (10, Some(&part_numbers[3..])),
            (11, None),
            (usize::MAX, None),
        ] {
            assert_eq!(parts_after_marker(&part_numbers, marker), expected, "marker {marker}");
        }
        assert_eq!(parts_after_marker(&[], 1), None);
    }

    #[test]
    fn delete_file_info_version_id_maps_explicit_null_version_to_stored_null() {
        assert_eq!(delete_file_info_version_id(Some(Uuid::nil())), None);

        let version_id = Uuid::new_v4();
        assert_eq!(delete_file_info_version_id(Some(version_id)), Some(version_id));
        assert_eq!(delete_file_info_version_id(None), None);
    }

    #[test]
    fn put_object_fast_path_selection_prefers_inline_only_when_inline_buffer_and_single_block() {
        assert!(should_use_inline_small_fast_path(true, 1024, 4096));
        assert!(!should_use_single_block_non_inline_fast_path(true, 1024, 4096));
        assert!(matches!(classify_small_write_path(true, 1024, 4096), SmallWritePath::Inline));

        assert!(!should_use_inline_small_fast_path(false, 1024, 4096));
        assert!(should_use_single_block_non_inline_fast_path(false, 1024, 4096));
        assert!(matches!(
            classify_small_write_path(false, 1024, 4096),
            SmallWritePath::SingleBlockNonInline
        ));
    }

    #[test]
    fn put_object_fast_path_selection_rejects_zero_and_multi_block_payloads() {
        assert!(!should_use_inline_small_fast_path(true, 0, 4096));
        assert!(!should_use_single_block_non_inline_fast_path(false, 0, 4096));
        assert!(matches!(classify_small_write_path(true, 0, 4096), SmallWritePath::Pipeline));

        assert!(!should_use_inline_small_fast_path(true, -1, 4096));
        assert!(!should_use_single_block_non_inline_fast_path(false, -1, 4096));
        assert!(matches!(classify_small_write_path(false, -1, 4096), SmallWritePath::Pipeline));

        assert!(!should_use_inline_small_fast_path(true, 8192, 4096));
        assert!(!should_use_single_block_non_inline_fast_path(false, 8192, 4096));
        assert!(matches!(classify_small_write_path(false, 8192, 4096), SmallWritePath::Pipeline));
    }

    #[test]
    fn put_object_large_batch_path_only_applies_to_large_ordinary_puts() {
        assert!(matches!(
            classify_put_write_path(false, 64 * 1024 * 1024, 1024 * 1024),
            SmallWritePath::PipelineBatchedLarge
        ));
        assert!(matches!(
            classify_put_write_path(false, 32 * 1024 * 1024, 1024 * 1024),
            SmallWritePath::Pipeline
        ));
        assert!(matches!(
            classify_put_write_path(false, 31 * 1024 * 1024, 1024 * 1024),
            SmallWritePath::Pipeline
        ));
        assert!(matches!(
            classify_put_write_path(true, 64 * 1024 * 1024, 1024 * 1024),
            SmallWritePath::Pipeline
        ));
    }

    #[test]
    fn put_object_classification_uses_only_known_storage_size() {
        assert_eq!(known_put_object_storage_size(42), 42);
        assert_eq!(
            known_put_object_storage_size(HashReader::SIZE_PRESERVE_LAYER),
            HashReader::SIZE_PRESERVE_LAYER
        );
        assert!(matches!(
            classify_put_write_path(false, known_put_object_storage_size(HashReader::SIZE_PRESERVE_LAYER), 1024 * 1024),
            SmallWritePath::Pipeline
        ));
        assert!(matches!(
            classify_put_write_path(false, known_put_object_storage_size(1024 * 1024), 1024 * 1024),
            SmallWritePath::SingleBlockNonInline
        ));
    }

    #[test]
    fn inline_admission_falls_back_to_actual_size_for_transformed_streams() {
        let erasure = coding::Erasure::new(2, 2, 1024 * 1024);
        let unknown = HashReader::SIZE_PRESERVE_LAYER;

        // A known stored size is authoritative, whatever the plaintext size says.
        assert_eq!(
            inline_admission_shard_size(&erasure, 16 * 1024, 4 * 1024 * 1024),
            erasure.shard_file_size(16 * 1024)
        );
        assert_eq!(inline_admission_shard_size(&erasure, 0, 4 * 1024), 0);

        // A transformed stream is sized by its plaintext length (MinIO parity).
        assert_eq!(
            inline_admission_shard_size(&erasure, unknown, 16 * 1024),
            erasure.shard_file_size(16 * 1024)
        );

        // Neither size known, or an empty transformed stream, cannot be admitted.
        assert!(inline_admission_shard_size(&erasure, unknown, 0) < 0);
        assert!(inline_admission_shard_size(&erasure, unknown, unknown) < 0);
    }

    #[test]
    fn put_object_part_fast_path_selection_matches_single_block_non_inline_rules() {
        assert!(should_use_single_block_non_inline_fast_path(false, 4096, 4096));
        assert!(should_use_single_block_non_inline_fast_path(false, 2048, 4096));
        assert!(!should_use_single_block_non_inline_fast_path(false, 4097, 4096));
        assert!(!should_use_single_block_non_inline_fast_path(false, 0, 4096));
        assert!(matches!(
            classify_small_write_path(false, 4096, 4096),
            SmallWritePath::SingleBlockNonInline
        ));
    }

    #[test]
    fn multipart_put_large_batch_path_only_applies_at_128m_and_above() {
        assert!(matches!(
            classify_multipart_part_write_path(128 * 1024 * 1024, 1024 * 1024),
            SmallWritePath::PipelineBatchedLarge
        ));
        assert!(matches!(
            classify_multipart_part_write_path(64 * 1024 * 1024, 1024 * 1024),
            SmallWritePath::Pipeline
        ));
        assert!(matches!(
            classify_multipart_part_write_path(1024 * 1024, 1024 * 1024),
            SmallWritePath::SingleBlockNonInline
        ));
    }

    #[test]
    fn multipart_write_paths_use_distinct_metric_labels() {
        assert_eq!(SmallWritePath::Pipeline.multipart_metric_label(), "multipart_write_pipeline");
        assert_eq!(
            SmallWritePath::PipelineBatchedLarge.multipart_metric_label(),
            "multipart_write_pipeline_batched_large"
        );
        assert_eq!(
            SmallWritePath::SingleBlockNonInline.multipart_metric_label(),
            "multipart_write_single_block_non_inline"
        );
    }

    #[test]
    fn test_is_cold_storage_class() {
        // Test cold storage classes
        assert!(is_cold_storage_class(storageclass::DEEP_ARCHIVE));
        assert!(is_cold_storage_class(storageclass::GLACIER));
        assert!(is_cold_storage_class(storageclass::GLACIER_IR));

        // Test non-cold storage classes
        assert!(!is_cold_storage_class(storageclass::STANDARD));
        assert!(!is_cold_storage_class(storageclass::RRS));
        assert!(!is_cold_storage_class(storageclass::STANDARD_IA));
        assert!(!is_cold_storage_class(storageclass::EXPRESS_ONEZONE));
    }

    #[test]
    fn test_is_infrequent_access_class() {
        // Test infrequent access classes
        assert!(is_infrequent_access_class(storageclass::ONEZONE_IA));
        assert!(is_infrequent_access_class(storageclass::STANDARD_IA));
        assert!(is_infrequent_access_class(storageclass::INTELLIGENT_TIERING));

        // Test frequent access classes
        assert!(!is_infrequent_access_class(storageclass::STANDARD));
        assert!(!is_infrequent_access_class(storageclass::RRS));
        assert!(!is_infrequent_access_class(storageclass::DEEP_ARCHIVE));
        assert!(!is_infrequent_access_class(storageclass::EXPRESS_ONEZONE));
    }

    // Regression test: `mc cp --storage-class STANDARD` on a tiered object (self-copy) must not
    // return NotImplemented.  When the source object is tiered (transitioned_object.tier is
    // non-empty) the usecase layer in object_usecase.rs intentionally leaves metadata_only=false
    // so that the full copy path is taken.  SetDisks::copy_object must therefore accept a
    // same-bucket/same-key call even when metadata_only=false.
    //
    // Currently this test FAILS because the guard at set_disk.rs:1579 unconditionally rejects
    // !metadata_only with StorageError::NotImplemented.  Once the fix is applied the test will
    // pass (or progress further through the copy path before failing on missing disk data).
    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn copy_object_tiered_self_copy_does_not_return_not_implemented() {
        let _setup_type_guard = SetupTypeGuard::switch_to(SetupType::Erasure).await;
        let set_disks = make_test_set_disks(vec![Arc::new(LocalClient::with_manager(Arc::new(
            rustfs_lock::GlobalLockManager::new(),
        )))])
        .await;

        // Simulate a tiered object: metadata_only is false (set_disk must handle the full copy),
        // and transitioned_object.tier is non-empty (the object lives on a remote tier).
        let mut src_info = ObjectInfo {
            metadata_only: false,
            transitioned_object: TransitionedObject {
                tier: "NEXTCLOUD".to_string(),
                ..Default::default()
            },
            ..Default::default()
        };

        let result = set_disks
            .copy_object(
                "bucket",
                "object",
                "bucket",
                "object",
                &mut src_info,
                &ObjectOptions::default(),
                &ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await;

        // The copy must not be rejected with NotImplemented.  Any other outcome (Ok or a
        // different error such as missing-disk / quorum) is acceptable here.
        if let Err(ref err) = result {
            assert!(
                !matches!(err, StorageError::NotImplemented),
                "tiered self-copy returned NotImplemented — copy_object must handle \
                 metadata_only=false for same-key copies of tiered objects, got: {err}"
            );
        }
    }

    async fn make_local_bucket_test_set_disks() -> Arc<SetDisks> {
        make_local_bucket_test_set_disks_with_drive_count(2).await
    }

    fn assert_exclusive_object_lock_held(set_disks: &SetDisks, bucket: &str, object: &str) {
        let lock = set_disks
            .local_lock_manager_for_test()
            .get_lock_info(&ObjectKey::new(bucket, object))
            .expect("object lock should be visible while rename is paused");
        assert!(matches!(lock.mode, rustfs_lock::LockMode::Exclusive));
        assert_eq!(lock.owner.as_ref(), set_disks.locker_owner.as_str());
    }

    async fn make_local_bucket_test_set_disks_with_drive_count(drive_count: usize) -> Arc<SetDisks> {
        let format = FormatV3::new(1, drive_count);
        let mut endpoints = Vec::new();
        let mut disks = Vec::new();

        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");

            mem::forget(dir);
            endpoints.push(endpoint);
            disks.push(Some(disk));
        }

        let instance_ctx = Arc::new(InstanceContext::new());
        instance_ctx.update_erasure_type(SetupType::Erasure).await;
        let set_disks = SetDisks::new_with_instance_ctx(
            "test-owner".to_string(),
            Arc::new(RwLock::new(disks)),
            drive_count,
            drive_count / 2,
            0,
            0,
            endpoints,
            format,
            Vec::new(),
            instance_ctx,
        )
        .await;
        set_disks.set_test_storage_class_config(
            storageclass::lookup_config_for_pools_without_env(&rustfs_config::server_config::KVS::new(), &[drive_count])
                .expect("test storage class should resolve for the local drive count"),
        );
        set_disks
    }

    async fn make_local_bucket_test_set_disks_with_missing_format() -> Arc<SetDisks> {
        let format = FormatV3::new(1, 2);
        let mut endpoints = Vec::new();
        let mut disks = Vec::new();

        for disk_idx in 0..2 {
            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");

            if disk_idx == 0 {
                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");
            }

            mem::forget(dir);
            endpoints.push(endpoint);
            disks.push(Some(disk));
        }

        SetDisks::new(
            "test-owner".to_string(),
            Arc::new(RwLock::new(disks)),
            2,
            1,
            0,
            0,
            endpoints,
            format,
            Vec::new(),
        )
        .await
    }

    #[tokio::test]
    async fn bucket_operations_round_trip_without_panicking() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-roundtrip";

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

        let info = set_disks
            .get_bucket_info(bucket, &BucketOptions::default())
            .await
            .expect("bucket info should be available");
        assert_eq!(info.name, bucket);

        let buckets = set_disks
            .list_bucket(&BucketOptions::default())
            .await
            .expect("bucket listing should succeed");
        assert!(buckets.iter().any(|entry| entry.name == bucket));

        set_disks
            .delete_bucket(bucket, &DeleteBucketOptions::default())
            .await
            .expect("bucket should be deleted");
    }

    #[tokio::test]
    async fn set_level_listing_trait_methods_use_existing_listing_implementation() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-listing";

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

        let mut reader = PutObjReader::from_vec(b"hello".to_vec());
        set_disks
            .put_object(
                bucket,
                "object",
                &mut reader,
                &ObjectOptions {
                    no_lock: true,
                    ..ObjectOptions::default()
                },
            )
            .await
            .expect("object should be written");

        let list_result = set_disks
            .clone()
            .list_objects_v2(bucket, "", None, None, 1000, false, None, false)
            .await
            .expect("set-level list_objects_v2 should succeed");
        assert_eq!(list_result.objects.len(), 1);
        assert_eq!(list_result.objects[0].name, "object");

        let versions_result = set_disks
            .clone()
            .list_object_versions(bucket, "", None, None, None, 1000)
            .await
            .expect("set-level list_object_versions should succeed");
        assert_eq!(versions_result.objects.len(), 1);
        assert_eq!(versions_result.objects[0].name, "object");

        let (tx, mut rx) = mpsc::channel(4);
        set_disks
            .clone()
            .walk(CancellationToken::new(), bucket, "", tx, WalkOptions::default())
            .await
            .expect("set-level walk should succeed");

        let mut walked_names = Vec::new();
        while let Some(item) = rx.recv().await {
            if let Some(object) = item.item {
                walked_names.push(object.name);
            }
        }
        assert!(walked_names.iter().any(|name| name == "object"));
    }

    #[tokio::test]
    async fn set_level_put_get_delete_restores_large_object_and_fails_closed_after_delete() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-object-roundtrip";
        let object = "nested/object.bin";
        let payload = (0..(BLOCK_SIZE_V2 + 17)).map(|idx| (idx % 251) as u8).collect::<Vec<_>>();
        let opts = ObjectOptions {
            no_lock: 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(payload.clone());
        let written = set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("large object should be written");

        assert_eq!(written.size, payload.len() as i64);
        let mut get_reader = set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
            .expect("large object reader should open");
        let mut restored = Vec::new();
        get_reader
            .stream
            .read_to_end(&mut restored)
            .await
            .expect("large object should stream");
        assert_eq!(restored, payload);

        let deleted = set_disks
            .delete_object(bucket, object, opts.clone())
            .await
            .expect("object delete should succeed");
        assert_eq!(deleted.name, object);

        let err = match set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
        {
            Ok(_) => panic!("deleted object must not be readable"),
            Err(err) => err,
        };
        assert!(
            is_err_object_not_found(&err),
            "deleted object read must fail closed with object-not-found, got {err:?}"
        );
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn streaming_get_blocks_concurrent_overwrite_until_eof() {
        temp_env::async_with_vars([(rustfs_config::ENV_OBJECT_LOCK_OPTIMIZATION_ENABLE, Some("true"))], async {
            let set_disks = make_local_bucket_test_set_disks().await;
            let bucket = "snapshot-streaming-overwrite";
            let object = "object";
            let old_body = vec![0x41; 2 * 1024 * 1024];
            let new_body = vec![0x42; old_body.len()];
            let opts = ObjectOptions::default();

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

            let mut snapshot = set_disks
                .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
                .await
                .expect("snapshot reader should open");
            let overwrite_set = Arc::clone(&set_disks);
            let overwrite_body = new_body.clone();
            let overwrite_opts = opts.clone();
            let mut overwrite = tokio::spawn(async move {
                let mut reader = PutObjReader::from_vec(overwrite_body);
                overwrite_set.put_object(bucket, object, &mut reader, &overwrite_opts).await
            });
            assert!(
                tokio::time::timeout(Duration::from_millis(100), &mut overwrite)
                    .await
                    .is_err(),
                "overwrite must wait while the response body retains the read lock"
            );

            let mut restored = Vec::new();
            snapshot
                .stream
                .read_to_end(&mut restored)
                .await
                .expect("stream should remain readable");
            assert_eq!(restored, old_body);
            tokio::time::timeout(Duration::from_secs(5), overwrite)
                .await
                .expect("overwrite should proceed after EOF")
                .expect("overwrite task should join")
                .expect("overwrite should succeed");

            let mut latest = set_disks
                .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
                .await
                .expect("latest reader should open");
            let mut latest_body = Vec::new();
            latest
                .stream
                .read_to_end(&mut latest_body)
                .await
                .expect("latest object should remain readable");
            assert_eq!(latest_body, new_body);

            let cancelled = set_disks
                .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
                .await
                .expect("cancelled reader should open");
            drop(cancelled);
            let mut replacement = PutObjReader::from_vec(vec![0x43; 2 * 1024 * 1024]);
            tokio::time::timeout(Duration::from_secs(5), set_disks.put_object(bucket, object, &mut replacement, &opts))
                .await
                .expect("reader drop must release its read lock")
                .expect("replacement after cancellation should succeed");
        })
        .await;
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn multipart_streaming_get_blocks_overwrite_across_part_boundary() {
        temp_env::async_with_vars(
            [
                (rustfs_config::ENV_OBJECT_LOCK_OPTIMIZATION_ENABLE, Some("true")),
                (ENV_RUSTFS_GET_MULTIPART_READER_SETUP_PREFETCH, Some("false")),
            ],
            async {
                let set_disks = make_local_bucket_test_set_disks().await;
                let bucket = "snapshot-multipart-overwrite";
                let object = "object";
                let part_size = usize::try_from(GLOBAL_MIN_PART_SIZE.as_u64()).expect("minimum part size should fit usize");
                let first_part = vec![0x41; part_size];
                let second_part = vec![0x42; part_size];
                let replacement = vec![0x43; first_part.len() + second_part.len()];
                let opts = ObjectOptions::default();

                set_disks
                    .make_bucket(bucket, &MakeBucketOptions::default())
                    .await
                    .expect("bucket should be created");
                let upload = set_disks
                    .new_multipart_upload(bucket, object, &opts)
                    .await
                    .expect("multipart upload should be created");
                let mut completed_parts = Vec::with_capacity(2);
                for (part_num, body) in [(1, &first_part), (2, &second_part)] {
                    let mut reader = PutObjReader::from_vec(body.clone());
                    let part = set_disks
                        .put_object_part(bucket, object, &upload.upload_id, part_num, &mut reader, &opts)
                        .await
                        .expect("multipart part should be written");
                    completed_parts.push(CompletePart {
                        part_num,
                        etag: part.etag,
                        ..Default::default()
                    });
                }
                let completed = Arc::clone(&set_disks)
                    .complete_multipart_upload(bucket, object, &upload.upload_id, completed_parts, &opts)
                    .await
                    .expect("multipart upload should complete");
                assert!(completed.is_multipart());

                let mut snapshot = set_disks
                    .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
                    .await
                    .expect("multipart snapshot reader should open");
                let overwrite_set = Arc::clone(&set_disks);
                let overwrite_opts = opts.clone();
                let overwrite_body = replacement.clone();
                let commit_barrier = PutObjectCommitBarrier::install(bucket, object, PutObjectCommitPause::BeforeNamespace);
                let overwrite = tokio::spawn(async move {
                    let mut reader = PutObjReader::from_vec(overwrite_body);
                    overwrite_set.put_object(bucket, object, &mut reader, &overwrite_opts).await
                });
                commit_barrier.wait_until_paused().await;
                commit_barrier.release_and_wait_until_namespace_pending().await;
                assert!(
                    !commit_barrier.namespace_acquired(),
                    "overwrite must wait for the multipart response's read lock"
                );

                let mut restored_first = vec![0; first_part.len()];
                snapshot
                    .stream
                    .read_exact(&mut restored_first)
                    .await
                    .expect("the first multipart part should stream");
                assert_eq!(restored_first, first_part);
                assert!(
                    !commit_barrier.namespace_acquired() && !overwrite.is_finished(),
                    "overwrite must remain blocked at the first/second part boundary"
                );

                let mut restored_second = Vec::new();
                snapshot
                    .stream
                    .read_to_end(&mut restored_second)
                    .await
                    .expect("the second multipart part should stream");
                assert_eq!(restored_second, second_part);
                tokio::time::timeout(Duration::from_secs(5), overwrite)
                    .await
                    .expect("overwrite should proceed after multipart EOF")
                    .expect("overwrite task should join")
                    .expect("overwrite should succeed");

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

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn streaming_get_blocks_concurrent_delete_until_eof() {
        temp_env::async_with_vars([(rustfs_config::ENV_OBJECT_LOCK_OPTIMIZATION_ENABLE, Some("true"))], async {
            let set_disks = make_local_bucket_test_set_disks().await;
            let bucket = "snapshot-streaming-delete";
            let object = "object";
            let body = vec![0x41; 2 * 1024 * 1024];
            let opts = ObjectOptions {
                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(body.clone());
            set_disks
                .put_object(bucket, object, &mut reader, &opts)
                .await
                .expect("object should be written");

            let mut snapshot = set_disks
                .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
                .await
                .expect("snapshot reader should open");
            let delete_set = Arc::clone(&set_disks);
            let delete_opts = opts.clone();
            let mut delete = tokio::spawn(async move { delete_set.delete_object(bucket, object, delete_opts).await });
            assert!(
                tokio::time::timeout(Duration::from_millis(100), &mut delete).await.is_err(),
                "delete must wait while the response body retains the read lock"
            );

            let mut restored = Vec::new();
            snapshot
                .stream
                .read_to_end(&mut restored)
                .await
                .expect("stream should remain readable before delete");
            assert_eq!(restored, body);
            tokio::time::timeout(Duration::from_secs(30), delete)
                .await
                .expect("delete should proceed after EOF")
                .expect("delete task should join")
                .expect("delete should succeed");
            let err = match set_disks
                .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
                .await
            {
                Ok(_) => panic!("a new read must not observe the deleted object"),
                Err(err) => err,
            };
            assert!(is_err_object_not_found(&err));
        })
        .await;
    }

    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn streaming_get_blocks_concurrent_delete_objects_until_eof() {
        temp_env::async_with_vars([(rustfs_config::ENV_OBJECT_LOCK_OPTIMIZATION_ENABLE, Some("true"))], async {
            let set_disks = make_local_bucket_test_set_disks().await;
            let bucket = "snapshot-streaming-delete-objects";
            let object = "object";
            let body = vec![0x41; 2 * 1024 * 1024];
            let opts = ObjectOptions {
                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(body.clone());
            set_disks
                .put_object(bucket, object, &mut reader, &opts)
                .await
                .expect("object should be written");

            let mut snapshot = set_disks
                .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
                .await
                .expect("snapshot reader should open");
            let delete_set = Arc::clone(&set_disks);
            let delete_opts = opts.clone();
            let mut delete = tokio::spawn(async move {
                delete_set
                    .delete_objects(
                        bucket,
                        vec![ObjectToDelete {
                            object_name: object.to_string(),
                            ..Default::default()
                        }],
                        delete_opts,
                    )
                    .await
            });
            assert!(
                tokio::time::timeout(Duration::from_millis(100), &mut delete).await.is_err(),
                "batch delete must wait while the response body retains the read lock"
            );

            let mut restored = Vec::new();
            snapshot
                .stream
                .read_to_end(&mut restored)
                .await
                .expect("stream should remain readable before batch delete");
            assert_eq!(restored, body);
            let (_, errors) = tokio::time::timeout(Duration::from_secs(30), delete)
                .await
                .expect("batch delete should proceed after EOF")
                .expect("batch delete task should join");
            assert!(errors.iter().all(Option::is_none));
        })
        .await;
    }

    #[tokio::test]
    async fn set_level_batched_large_put_get_restores_body() {
        const BATCHED_LARGE_SIZE: usize = 64 * 1024 * 1024;

        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-batched-large-object";
        let object = "large/batched.bin";
        let payload = (0..BATCHED_LARGE_SIZE)
            .map(|idx| ((idx as u64).wrapping_mul(31).wrapping_add(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 mut reader = PutObjReader::from_vec(payload.clone());
        let written = set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("batched-large object should be written");
        assert_eq!(written.size, payload.len() as i64);

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

    #[tokio::test]
    async fn set_level_put_object_fails_closed_when_writer_quorum_is_unavailable() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-put-writer-quorum";
        let object = "object.txt";
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created before disk loss");
        {
            let mut disks = set_disks.disks.write().await;
            disks[1] = None;
        }

        let mut reader = PutObjReader::from_vec(b"quorum guarded body".to_vec());
        let err = set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect_err("missing writer quorum must fail the put");
        assert!(
            matches!(err, Error::ErasureWriteQuorum | Error::InsufficientWriteQuorum(_, _)),
            "expected write quorum failure, got {err:?}"
        );

        let read_err = match set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
        {
            Ok(_) => panic!("failed put must not leave a readable object"),
            Err(err) => err,
        };
        assert!(
            is_err_object_not_found(&read_err) || matches!(read_err, Error::ErasureReadQuorum),
            "failed put must fail closed on read, got {read_err:?}"
        );
    }

    #[tokio::test]
    async fn set_level_put_object_short_reader_fails_closed_across_write_paths() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-put-short-reader";
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };

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

        let cases = [
            ("single-block", b"short body".to_vec(), 32),
            ("pipeline", b"pipeline short body".to_vec(), BLOCK_SIZE_V2 as i64 + 1),
            ("batched-large", b"batched short body".to_vec(), 64 * 1024 * 1024),
        ];

        for (name, payload, declared_size) in cases {
            let object = format!("{name}/object.txt");
            let hash_reader = HashReader::from_stream(Cursor::new(payload), declared_size, declared_size, None, None, false)
                .expect("test reader should be constructed");
            let mut reader = PutObjReader::new(hash_reader);
            let err = set_disks
                .put_object(bucket, &object, &mut reader, &opts)
                .await
                .expect_err("short reader must fail the put");
            let err_text = format!("{err:?}");
            assert!(
                err_text.contains("UnexpectedEof") || err_text.contains("IncompleteBody"),
                "{name} short reader should fail with an EOF/incomplete-body error, got {err:?}"
            );

            let read_err = match set_disks
                .get_object_reader(bucket, &object, None, HeaderMap::new(), &opts)
                .await
            {
                Ok(_) => panic!("{name} short failed put must not leave a readable object"),
                Err(err) => err,
            };
            assert!(
                is_err_object_not_found(&read_err) || matches!(read_err, Error::ErasureReadQuorum),
                "{name} short failed put must fail closed on read, got {read_err:?}"
            );
        }
    }

    #[tokio::test]
    async fn set_level_get_restores_body_when_one_shard_is_missing_after_write() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-read-repair-missing-shard";
        let object = "object.bin";
        let payload = (0..(BLOCK_SIZE_V2 + 211))
            .map(|idx| ((idx * 13) % 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 before shard loss");

        {
            let mut disks = set_disks.disks.write().await;
            disks[1] = None;
        }

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

        assert_eq!(restored, payload);
    }

    #[tokio::test]
    async fn set_level_overwrite_restores_new_body_and_cleans_old_data_dir() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-object-overwrite";
        let object = "nested/object.bin";
        let first_payload = (0..(BLOCK_SIZE_V2 + 31)).map(|idx| (idx % 239) as u8).collect::<Vec<_>>();
        let second_payload = (0..(BLOCK_SIZE_V2 + 97))
            .map(|idx| ((idx * 7) % 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 first_reader = PutObjReader::from_vec(first_payload);
        set_disks
            .put_object(bucket, object, &mut first_reader, &opts)
            .await
            .expect("first object body should be written");

        let mut second_reader = PutObjReader::from_vec(second_payload.clone());
        let second = set_disks
            .put_object(bucket, object, &mut second_reader, &opts)
            .await
            .expect("overwrite body should commit and clean the old data dir");

        assert_eq!(second.size, second_payload.len() as i64);
        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);
    }

    #[tokio::test]
    async fn set_level_write_preconditions_fail_closed_and_allow_matching_etags() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-write-preconditions";
        let object = "object.txt";
        let write_opts = ObjectOptions {
            no_lock: true,
            preserve_etag: Some("conditional-etag".to_string()),
            ..Default::default()
        };

        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");
        let mut reader = PutObjReader::from_vec(b"conditional body".to_vec());
        set_disks
            .put_object(bucket, object, &mut reader, &write_opts)
            .await
            .expect("object should be written with deterministic etag");

        let reject_existing = ObjectOptions {
            http_preconditions: Some(HTTPPreconditions {
                if_none_match: Some("conditional-etag".to_string()),
                ..Default::default()
            }),
            ..Default::default()
        };
        assert!(matches!(
            set_disks.check_write_precondition(bucket, object, &reject_existing).await,
            Some(StorageError::PreconditionFailed)
        ));

        let allow_matching = ObjectOptions {
            http_preconditions: Some(HTTPPreconditions {
                if_match: Some("\"conditional-etag\"".to_string()),
                ..Default::default()
            }),
            ..Default::default()
        };
        assert!(
            set_disks
                .check_write_precondition(bucket, object, &allow_matching)
                .await
                .is_none()
        );

        let missing_if_match = ObjectOptions {
            http_preconditions: Some(HTTPPreconditions {
                if_match: Some("missing-etag".to_string()),
                ..Default::default()
            }),
            ..Default::default()
        };
        assert!(matches!(
            set_disks
                .check_write_precondition(bucket, "missing-object.txt", &missing_if_match)
                .await,
            Some(StorageError::ObjectNotFound(_, _))
        ));
    }

    #[tokio::test]
    async fn conditional_replace_holds_object_lock_through_rename() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-conditional-replace-fence";
        let object = "config/conditional-replace.json";
        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");

        let mut initial_reader = PutObjReader::from_vec(b"initial config".to_vec());
        let initial = set_disks
            .put_object(
                bucket,
                object,
                &mut initial_reader,
                &ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await
            .expect("initial config should be written");
        let initial_etag = initial.etag.expect("initial config should have an ETag");

        let barrier = rename_fanout_barrier::arm(object, 0, rename_fanout_barrier::PHASE_RENAME);
        let writer_store = set_disks.clone();
        let expected_etag = initial_etag.clone();
        let writer = tokio::spawn(async move {
            let mut reader = PutObjReader::from_vec(b"replacement config".to_vec());
            writer_store
                .put_object(
                    bucket,
                    object,
                    &mut reader,
                    &ObjectOptions {
                        preserve_etag: Some("replacement-etag".to_string()),
                        http_preconditions: Some(HTTPPreconditions {
                            if_match: Some(expected_etag),
                            ..Default::default()
                        }),
                        ..Default::default()
                    },
                )
                .await
        });

        tokio::time::timeout(std::time::Duration::from_secs(30), barrier.wait_until_paused())
            .await
            .expect("conditional replace should reach the rename barrier");
        assert_exclusive_object_lock_held(&set_disks, bucket, object);
        barrier.release();
        writer
            .await
            .expect("conditional writer task should finish")
            .expect("matching conditional replace should commit");
        let contender = set_disks
            .new_ns_lock(bucket, object)
            .await
            .expect("contender namespace lock should be created");
        let contender_guard = contender
            .get_write_lock(std::time::Duration::from_secs(30))
            .await
            .expect("contender should acquire after conditional replace commits");
        drop(contender_guard);

        let mut stale_reader = PutObjReader::from_vec(b"stale config".to_vec());
        let err = set_disks
            .put_object(
                bucket,
                object,
                &mut stale_reader,
                &ObjectOptions {
                    http_preconditions: Some(HTTPPreconditions {
                        if_match: Some(initial_etag),
                        ..Default::default()
                    }),
                    ..Default::default()
                },
            )
            .await
            .expect_err("the old ETag must fail after the fenced replacement commits");
        assert_eq!(err, StorageError::PreconditionFailed);
    }

    #[tokio::test]
    async fn listing_metadata_reconcile_encodes_directory_object_and_waits_for_writer() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-list-reconcile-lock";
        let object = "directory/";
        let stored_object = encode_dir_object(object);
        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");
        let mut reader = PutObjReader::from_vec(b"directory marker".to_vec());
        set_disks
            .put_object(bucket, &stored_object, &mut reader, &ObjectOptions::default())
            .await
            .expect("directory object should be written under its encoded key");
        let disks = set_disks.disks.read().await.clone();
        let namespace_lock = set_disks
            .new_ns_lock(bucket, &stored_object)
            .await
            .expect("namespace lock should be created");
        let writer_guard = namespace_lock
            .get_write_lock(std::time::Duration::from_secs(30))
            .await
            .expect("writer lock should be acquired");
        let read = set_disks.read_listing_metadata_after_namespace_barrier(&disks, bucket, object);
        tokio::pin!(read);
        assert!(
            tokio::time::timeout(std::time::Duration::from_millis(50), &mut read)
                .await
                .is_err(),
            "LIST reconciliation must wait for the publishing writer"
        );
        drop(writer_guard);

        let (entries, confirmed_absent) = tokio::time::timeout(std::time::Duration::from_secs(30), read)
            .await
            .expect("LIST reconciliation should resume after writer release")
            .expect("LIST reconciliation should read after the namespace barrier");
        assert_eq!(entries.0.len(), disks.len());
        assert!(
            entries.0.iter().all(|entry| entry
                .as_ref()
                .is_some_and(|entry| entry.name == object && !entry.metadata.is_empty())),
            "LIST reconciliation must read encoded metadata and retain the logical key"
        );
        assert_eq!(confirmed_absent, 0);
    }

    #[tokio::test]
    async fn repeated_body_write_keeps_etag_but_changes_data_dir_generation() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-write-generation";
        let object = "config/write-generation.json";
        let body = b"identical config body".to_vec();
        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");

        let mut first_reader = PutObjReader::from_vec(body.clone());
        let first = set_disks
            .put_object(
                bucket,
                object,
                &mut first_reader,
                &ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await
            .expect("first config body should be written");
        let mut second_reader = PutObjReader::from_vec(body);
        let second = set_disks
            .put_object(
                bucket,
                object,
                &mut second_reader,
                &ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await
            .expect("identical config body should be rewritten");

        assert_eq!(first.etag, second.etag, "content ETag should expose the ABA collision");
        assert_ne!(first.data_dir, second.data_dir, "each committed body write needs a unique generation");
        assert!(first.data_dir.is_some() && second.data_dir.is_some());
    }

    #[tokio::test]
    async fn conditional_create_holds_object_lock_through_rename() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-conditional-create-fence";
        let object = "config/conditional-create.json";
        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");

        let barrier = rename_fanout_barrier::arm(object, 0, rename_fanout_barrier::PHASE_RENAME);
        let writer_store = set_disks.clone();
        let writer = tokio::spawn(async move {
            let mut reader = PutObjReader::from_vec(b"created config".to_vec());
            writer_store
                .put_object(
                    bucket,
                    object,
                    &mut reader,
                    &ObjectOptions {
                        http_preconditions: Some(HTTPPreconditions {
                            if_none_match: Some("*".to_string()),
                            ..Default::default()
                        }),
                        ..Default::default()
                    },
                )
                .await
        });

        tokio::time::timeout(std::time::Duration::from_secs(30), barrier.wait_until_paused())
            .await
            .expect("conditional create should reach the rename barrier");
        assert_exclusive_object_lock_held(&set_disks, bucket, object);
        barrier.release();
        writer
            .await
            .expect("conditional writer task should finish")
            .expect("first conditional create should commit");
        assert!(
            set_disks
                .local_lock_manager_for_test()
                .get_lock_info(&ObjectKey::new(bucket, object))
                .is_none(),
            "conditional create should release the object lock after commit"
        );
        let contender = set_disks
            .new_ns_lock(bucket, object)
            .await
            .expect("contender namespace lock should be created");
        let contender_guard = contender
            .get_write_lock(std::time::Duration::from_secs(30))
            .await
            .expect("contender should acquire after conditional create commits");
        drop(contender_guard);

        let mut duplicate_reader = PutObjReader::from_vec(b"duplicate config".to_vec());
        let err = set_disks
            .put_object(
                bucket,
                object,
                &mut duplicate_reader,
                &ObjectOptions {
                    http_preconditions: Some(HTTPPreconditions {
                        if_none_match: Some("*".to_string()),
                        ..Default::default()
                    }),
                    ..Default::default()
                },
            )
            .await
            .expect_err("a second create-only write must not replace the committed config");
        assert_eq!(err, StorageError::PreconditionFailed);
    }

    #[tokio::test]
    async fn set_level_if_none_match_fails_closed_without_read_quorum() {
        let set_disks = make_local_bucket_test_set_disks_with_drive_count(4).await;
        let bucket = "bucket-write-precondition-quorum";
        let object = "existing-object.txt";

        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created before disk loss");
        let mut reader = PutObjReader::from_vec(b"existing object body".to_vec());
        set_disks
            .put_object(
                bucket,
                object,
                &mut reader,
                &ObjectOptions {
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await
            .expect("object should be written before disk loss");
        {
            let mut disks = set_disks.disks.write().await;
            disks[1..].fill(None);
        }

        let create_only = ObjectOptions {
            http_preconditions: Some(HTTPPreconditions {
                if_none_match: Some("*".to_string()),
                ..Default::default()
            }),
            ..Default::default()
        };
        let result = set_disks.check_write_precondition(bucket, object, &create_only).await;
        assert!(
            matches!(result, Some(StorageError::ErasureReadQuorum | StorageError::InsufficientReadQuorum(_, _))),
            "expected read-quorum failure, got {result:?}"
        );
    }

    #[tokio::test]
    async fn set_level_versioned_delete_marker_hides_object_without_corrupting_version_metadata() {
        let set_disks = make_local_bucket_test_set_disks_with_drive_count(4).await;
        let bucket = "bucket-versioned-delete";
        let object = "object.txt";
        let opts = ObjectOptions {
            no_lock: true,
            versioned: true,
            ..Default::default()
        };

        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");
        let mut reader = PutObjReader::from_vec(b"versioned object body".to_vec());
        let written = set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("versioned object should be written");
        assert!(written.version_id.is_some());

        let marker = set_disks
            .delete_object(bucket, object, opts.clone())
            .await
            .expect("versioned delete should create a marker");

        assert!(marker.delete_marker);
        let marker_version = marker.version_id.expect("versioned delete marker should carry a version id");
        let create_only = ObjectOptions {
            versioned: true,
            data_movement: true,
            http_preconditions: Some(HTTPPreconditions {
                if_none_match: Some("*".to_string()),
                ..Default::default()
            }),
            ..Default::default()
        };
        assert_eq!(
            set_disks.check_write_precondition(bucket, object, &create_only).await,
            Some(StorageError::PreconditionFailed),
            "data movement must not replace a target delete marker"
        );
        let err = match set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
        {
            Ok(_) => panic!("latest delete marker must hide object body"),
            Err(err) => err,
        };
        assert!(
            is_err_object_not_found(&err),
            "latest delete marker read must map to ObjectNotFound, got {err:?}"
        );

        let versioned_opts = ObjectOptions {
            version_id: Some(marker_version.to_string()),
            ..opts.clone()
        };
        let err = match set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &versioned_opts)
            .await
        {
            Ok(_) => panic!("explicit delete marker version must not expose a body"),
            Err(err) => err,
        };
        assert!(
            matches!(err, Error::MethodNotAllowed),
            "explicit delete marker version read must map to MethodNotAllowed, got {err:?}"
        );
    }

    #[tokio::test]
    async fn set_level_metadata_self_copy_preserves_body_and_updates_metadata() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-metadata-copy";
        let object = "object.txt";
        let opts = ObjectOptions {
            no_lock: true,
            ..Default::default()
        };
        let payload = b"metadata copy must not lose committed bytes".to_vec();

        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 source = set_disks
            .get_object_info(bucket, object, &opts)
            .await
            .expect("object info should be readable");
        let mut metadata = (*source.user_defined).clone();
        metadata.insert("x-amz-meta-copy-check".to_string(), "present".to_string());
        source.user_defined = Arc::new(metadata);
        source.metadata_only = true;
        source.etag = Some("metadata-copy-etag".to_string());

        let copied = set_disks
            .copy_object(bucket, object, bucket, object, &mut source, &opts, &opts)
            .await
            .expect("metadata self-copy should succeed");
        assert_eq!(copied.user_defined.get("x-amz-meta-copy-check").map(String::as_str), Some("present"));

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

        let reread = set_disks
            .get_object_info(bucket, object, &opts)
            .await
            .expect("copied object info should be readable");
        assert_eq!(reread.user_defined.get("x-amz-meta-copy-check").map(String::as_str), Some("present"));
    }

    #[tokio::test]
    async fn set_level_empty_object_read_uses_buffered_empty_body_with_locks() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-empty-object";
        let object = "empty.bin";
        let opts = ObjectOptions::default();

        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");
        let mut reader = PutObjReader::from_vec(Vec::new());
        let written = set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("empty object should be written");
        assert_eq!(written.size, 0);

        let mut get_reader = set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
            .expect("empty object reader should open");
        assert_eq!(get_reader.object_info.size, 0);
        assert_eq!(get_reader.buffered_body.as_ref().map(Bytes::len), Some(0));
        let mut restored = Vec::new();
        get_reader
            .stream
            .read_to_end(&mut restored)
            .await
            .expect("empty object should stream");
        assert!(restored.is_empty());

        let info = set_disks
            .get_object_info(bucket, object, &opts)
            .await
            .expect("empty object info should be readable");
        assert_eq!(info.size, 0);
    }

    #[tokio::test]
    async fn set_level_put_object_options_preserve_etag_and_normalize_standard_storage_class() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-put-options";
        let object = "object.txt";
        let mod_time = OffsetDateTime::from_unix_timestamp(1_717_171_717).expect("fixed timestamp should parse");
        let mut user_defined = HashMap::new();
        user_defined.insert(metadata_keys::STORAGE_CLASS.to_string(), storageclass::STANDARD.to_string());
        user_defined.insert(SUFFIX_COMPRESSION.to_string(), "zstd".to_string());
        let mut eval_metadata = HashMap::new();
        eval_metadata.insert("x-amz-meta-evaluated".to_string(), "yes".to_string());
        let opts = ObjectOptions {
            mod_time: Some(mod_time),
            preserve_etag: Some("preserved-etag".to_string()),
            user_defined,
            eval_metadata: Some(eval_metadata),
            ..Default::default()
        };

        set_disks
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("bucket should be created");
        let mut reader = PutObjReader::from_vec(b"option matrix body".to_vec());
        let written = set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("object should be written with option matrix");
        assert_eq!(written.etag.as_deref(), Some("preserved-etag"));
        assert_eq!(written.mod_time, Some(mod_time));
        assert_eq!(written.user_defined.get("x-amz-meta-evaluated").map(String::as_str), Some("yes"));
        assert!(!written.user_defined.contains_key(metadata_keys::STORAGE_CLASS));

        let info = set_disks
            .get_object_info(bucket, object, &opts)
            .await
            .expect("object info should be readable");
        assert_eq!(info.etag.as_deref(), Some("preserved-etag"));
        assert_eq!(info.mod_time, Some(mod_time));
        assert_eq!(info.user_defined.get("x-amz-meta-evaluated").map(String::as_str), Some("yes"));
        assert!(!info.user_defined.contains_key(metadata_keys::STORAGE_CLASS));
    }

    #[tokio::test]
    async fn set_level_put_object_metadata_updates_headers_without_rewriting_body() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-put-metadata";
        let object = "object.txt";
        let payload = b"metadata update must preserve bytes".to_vec();
        let write_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, &write_opts)
            .await
            .expect("object should be written");

        let mut eval_metadata = HashMap::new();
        eval_metadata.insert("x-amz-meta-updated".to_string(), "true".to_string());
        let update_time = OffsetDateTime::from_unix_timestamp(1_717_181_818).expect("fixed timestamp should parse");
        let update_opts = ObjectOptions {
            eval_metadata: Some(eval_metadata),
            mod_time: Some(update_time),
            ..Default::default()
        };
        let updated = set_disks
            .put_object_metadata(bucket, object, &update_opts)
            .await
            .expect("metadata update should succeed");
        assert_eq!(updated.mod_time, Some(update_time));
        assert_eq!(updated.user_defined.get("x-amz-meta-updated").map(String::as_str), Some("true"));

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

    #[tokio::test]
    async fn set_level_copy_object_with_prefetched_reader_restores_body() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-copy-reader";
        let object = "object.txt";
        let payload = b"copy reader body".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 source = ObjectInfo {
            metadata_only: false,
            put_object_reader: Some(PutObjReader::from_vec(payload.clone())),
            ..Default::default()
        };
        let copied = set_disks
            .copy_object(bucket, object, bucket, object, &mut source, &opts, &opts)
            .await
            .expect("copy with prefetched reader should write object data");
        assert_eq!(copied.size, payload.len() as i64);

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

    /// The other half of the suspended-versioning delete contract: a client
    /// that drains such a bucket lists the null delete marker and then purges
    /// it as `?versionId=null`, which is what `nuke_bucket` does before
    /// `DeleteBucket`. That purge must succeed — if it is rejected the marker's
    /// `xl.meta` survives, and `DeleteBucket`'s raw disk scan then reports
    /// `BucketNotEmpty` for a bucket the client has already emptied.
    #[tokio::test]
    async fn set_level_explicit_null_version_delete_purges_the_null_delete_marker() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-null-marker-purge";
        let object = "object.txt";
        // The delete path reads versioned/suspended from the bucket-config
        // snapshot, not from `opts`, so inject a real Suspended config —
        // otherwise `from_file_info` never synthesizes the null version id and
        // the branch under test is not reached.
        let suspended = crate::bucket::replication::DeleteReplicationConfigSnapshot::from_configs_for_test(
            s3s::dto::VersioningConfiguration {
                status: Some(s3s::dto::BucketVersioningStatus::from_static(s3s::dto::BucketVersioningStatus::SUSPENDED)),
                ..Default::default()
            },
            None,
        );
        let opts = ObjectOptions {
            no_lock: true,
            version_suspended: true,
            delete_replication_config_snapshot: Some(Arc::new(suspended)),
            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"suspended version body".to_vec());
        set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("suspended-version object should be written");

        let marker = set_disks
            .delete_object(
                bucket,
                object,
                ObjectOptions {
                    mod_time: Some(OffsetDateTime::now_utc()),
                    ..opts.clone()
                },
            )
            .await
            .expect("version-suspended delete should create a null marker");
        assert!(marker.delete_marker);
        assert_eq!(marker.version_id, Some(Uuid::nil()));

        let stale_target = crate::bucket::lifecycle::bucket_lifecycle_ops::lifecycle_version_delete_target(&marker)
            .expect("the observed null marker should produce an exact lifecycle target");
        let replacement_mod_time =
            marker.mod_time.expect("the first null marker must have a modification time") + time::Duration::seconds(1);

        let mut replacement_reader = PutObjReader::from_vec(b"replacement suspended version body".to_vec());
        set_disks
            .put_object(bucket, object, &mut replacement_reader, &opts)
            .await
            .expect("a suspended-version PUT should replace the first null marker");
        let replacement_marker = set_disks
            .delete_object(
                bucket,
                object,
                ObjectOptions {
                    mod_time: Some(replacement_mod_time),
                    ..opts.clone()
                },
            )
            .await
            .expect("a second suspended-version delete should create a replacement null marker");
        assert!(replacement_marker.delete_marker);
        assert_eq!(replacement_marker.version_id, Some(Uuid::nil()));
        assert_eq!(replacement_marker.mod_time, Some(replacement_mod_time));

        let (_deleted, stale_errors) = set_disks.delete_objects(bucket, vec![stale_target], opts.clone()).await;
        assert!(
            matches!(stale_errors.as_slice(), [Some(StorageError::PreconditionFailed)]),
            "a stale null-marker lifecycle target must not delete its replacement: {stale_errors:?}"
        );

        let (current_marker, _, _, current_error) = set_disks
            .get_object_info_fileinfo_and_quorum(
                bucket,
                object,
                &ObjectOptions {
                    version_id: Some(Uuid::nil().to_string()),
                    version_suspended: true,
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await;
        assert!(matches!(current_error, Some(StorageError::MethodNotAllowed)));
        assert!(current_marker.delete_marker);
        assert_eq!(current_marker.mod_time, Some(replacement_mod_time));

        let (_deleted, errs) = set_disks
            .delete_objects(
                bucket,
                vec![ObjectToDelete {
                    object_name: object.to_string(),
                    version_id: Some(Uuid::nil()),
                    ..Default::default()
                }],
                opts.clone(),
            )
            .await;

        assert!(
            errs.iter().all(Option::is_none),
            "explicit null-version purge of the null delete marker must succeed, got {errs:?}"
        );
    }

    #[tokio::test]
    async fn set_level_version_suspended_delete_creates_null_delete_marker() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-version-suspended-delete";
        let object = "object.txt";
        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"suspended version body".to_vec());
        set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("suspended-version object should be written");

        let marker = set_disks
            .delete_object(bucket, object, opts.clone())
            .await
            .expect("version-suspended delete should create a null marker");
        assert!(marker.delete_marker);
        assert_eq!(marker.version_id, Some(Uuid::nil()));

        let err = match set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
        {
            Ok(_) => panic!("null delete marker must hide object body"),
            Err(err) => err,
        };
        assert!(
            is_err_object_not_found(&err),
            "null delete marker read must map to ObjectNotFound, got {err:?}"
        );
    }

    #[tokio::test]
    async fn set_level_delete_prefix_removes_nested_objects() {
        let set_disks = make_local_bucket_test_set_disks().await;
        let bucket = "bucket-delete-prefix";
        let object = "prefix/object.txt";
        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(b"prefix body".to_vec());
        set_disks
            .put_object(bucket, object, &mut reader, &opts)
            .await
            .expect("prefix object should be written");

        set_disks
            .delete_object(
                bucket,
                "prefix/",
                ObjectOptions {
                    delete_prefix: true,
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await
            .expect("prefix delete should succeed");

        let err = match set_disks
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
        {
            Ok(_) => panic!("object under deleted prefix must not be readable"),
            Err(err) => err,
        };
        assert!(
            is_err_object_not_found(&err),
            "prefix-deleted object must fail closed with object-not-found, got {err:?}"
        );
    }

    #[tokio::test]
    #[serial]
    async fn set_level_delete_objects_batch_removes_live_data_and_tolerates_missing_keys() {
        temp_env::async_with_vars([("RUSTFS_ISSUE3031_DIAG_ENABLE", Some("true"))], async {
            let set_disks = make_local_bucket_test_set_disks().await;
            let bucket = "bucket-delete-objects";
            let existing = "existing.txt";
            let missing = "missing.txt";
            let opts = ObjectOptions {
                no_lock: 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"batch delete body".to_vec());
            set_disks
                .put_object(bucket, existing, &mut reader, &opts)
                .await
                .expect("object should be written before batch delete");

            let (deleted, errors) = set_disks
                .delete_objects(
                    bucket,
                    vec![
                        ObjectToDelete {
                            object_name: existing.to_string(),
                            ..Default::default()
                        },
                        ObjectToDelete {
                            object_name: missing.to_string(),
                            ..Default::default()
                        },
                    ],
                    opts.clone(),
                )
                .await;

            assert_eq!(deleted.len(), 2);
            assert_eq!(errors.len(), 2);
            assert!(errors.iter().all(Option::is_none));
            assert_eq!(deleted[0].object_name, existing);
            assert!(deleted[0].found);
            assert!(!deleted[0].delete_marker);
            assert_eq!(deleted[1].object_name, missing);
            assert!(!deleted[1].found);
            assert!(!deleted[1].delete_marker);

            let err = match set_disks
                .get_object_reader(bucket, existing, None, HeaderMap::new(), &opts)
                .await
            {
                Ok(_) => panic!("batch-deleted object must not be readable as latest"),
                Err(err) => err,
            };
            assert!(
                is_err_object_not_found(&err),
                "batch-deleted object must fail closed with object-not-found, got {err:?}"
            );
        })
        .await;
    }

    #[tokio::test]
    async fn set_level_heal_format_repairs_unformatted_disk() {
        let set_disks = make_local_bucket_test_set_disks_with_missing_format().await;
        let disk = {
            let disks = set_disks.disks.read().await;
            disks[1].clone().expect("second disk should exist")
        };

        let before = load_format_erasure(&disk, true)
            .await
            .expect_err("second disk should start unformatted");
        assert_eq!(before, DiskError::UnformattedDisk);

        let (heal_result, heal_err) = set_disks.heal_format(false).await.expect("heal_format should complete");
        assert!(heal_err.is_none(), "heal_format should repair the local unformatted disk");
        assert_eq!(heal_result.disk_count, 2);
        assert_eq!(heal_result.set_count, 1);
        assert_eq!(heal_result.after.drives[1].state, DriveState::Ok.to_string());

        let repaired = load_format_erasure(&disk, true)
            .await
            .expect("second disk should contain a healed format");
        assert_eq!(repaired.erasure.this, set_disks.format.erasure.sets[0][1]);
    }

    #[tokio::test]
    async fn remaining_unsupported_trait_stubs_return_typed_errors() {
        let set_disks = make_test_set_disks(Vec::new()).await;

        let (heal_result, heal_err) = make_local_bucket_test_set_disks()
            .await
            .heal_format(false)
            .await
            .expect("heal_format should be callable on formatted disks");
        assert!(matches!(heal_err, Some(StorageError::NoHealRequired)));
        assert_eq!(heal_result.disk_count, 2);

        let copy_part_err = set_disks
            .copy_object_part(
                "bucket",
                "src",
                "bucket",
                "dst",
                "upload-id",
                1,
                0,
                1,
                &ObjectInfo::default(),
                &ObjectOptions::default(),
                &ObjectOptions::default(),
            )
            .await
            .expect_err("unsupported copy_object_part should return a typed error");
        assert!(matches!(copy_part_err, StorageError::NotImplemented));

        set_disks
            .check_abandoned_parts(
                "bucket",
                "object",
                &HealOpts {
                    dry_run: true,
                    no_lock: true,
                    ..Default::default()
                },
            )
            .await
            .expect("abandoned-parts check should be callable on empty disk sets");
    }

    #[tokio::test]
    async fn stat_all_dirs_returns_index_aligned_vector_for_offline_disks() {
        // All-offline set: no real disk I/O needed. Isolates the length/index-alignment contract
        // that heal_object_dir depends on when it zips `errs` against the full `disks` array.
        let disks: Vec<Option<DiskStore>> = vec![None, None, None, None];

        let errs = stat_all_dirs(&disks, "bucket", "object").await;

        // Before the fix, offline disks contributed no future and the collected vector had length
        // 0, so any zip against `disks` paired errors with the wrong disk. After the fix each slot
        // is DiskNotFound, index-aligned with `disks`.
        assert_eq!(
            errs.len(),
            disks.len(),
            "stat_all_dirs must return one entry per disk slot to stay index-aligned"
        );
        for err in &errs {
            assert!(
                matches!(err, Some(DiskError::DiskNotFound)),
                "offline (None) disk slot must map to DiskNotFound in-place, got {err:?}"
            );
        }
    }

    #[test]
    fn adaptive_duplex_buffer_size_raises_mid_sized_gets_without_penalizing_tiny_objects() {
        assert_eq!(adaptive_duplex_buffer_size(64 * 1024), 64 * 1024);
        assert_eq!(adaptive_duplex_buffer_size(128 * 1024), 64 * 1024);
        assert_eq!(adaptive_duplex_buffer_size(256 * 1024), 256 * 1024);
        assert_eq!(adaptive_duplex_buffer_size(1024 * 1024), 512 * 1024);
        assert_eq!(adaptive_duplex_buffer_size(2 * 1024 * 1024), 1024 * 1024);
    }
}