summaryrefslogtreecommitdiff
path: root/fs/bcachefs/buckets.c
blob: 794adde61de10874a08ebcae2aab820bc6a01246 (plain)
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
// SPDX-License-Identifier: GPL-2.0
/*
 * Code for manipulating bucket marks for garbage collection.
 *
 * Copyright 2014 Datera, Inc.
 *
 * Bucket states:
 * - free bucket: mark == 0
 *   The bucket contains no data and will not be read
 *
 * - allocator bucket: owned_by_allocator == 1
 *   The bucket is on a free list, or it is an open bucket
 *
 * - cached bucket: owned_by_allocator == 0 &&
 *                  dirty_sectors == 0 &&
 *                  cached_sectors > 0
 *   The bucket contains data but may be safely discarded as there are
 *   enough replicas of the data on other cache devices, or it has been
 *   written back to the backing device
 *
 * - dirty bucket: owned_by_allocator == 0 &&
 *                 dirty_sectors > 0
 *   The bucket contains data that we must not discard (either only copy,
 *   or one of the 'main copies' for data requiring multiple replicas)
 *
 * - metadata bucket: owned_by_allocator == 0 && is_metadata == 1
 *   This is a btree node, journal or gen/prio bucket
 *
 * Lifecycle:
 *
 * bucket invalidated => bucket on freelist => open bucket =>
 *     [dirty bucket =>] cached bucket => bucket invalidated => ...
 *
 * Note that cache promotion can skip the dirty bucket step, as data
 * is copied from a deeper tier to a shallower tier, onto a cached
 * bucket.
 * Note also that a cached bucket can spontaneously become dirty --
 * see below.
 *
 * Only a traversal of the key space can determine whether a bucket is
 * truly dirty or cached.
 *
 * Transitions:
 *
 * - free => allocator: bucket was invalidated
 * - cached => allocator: bucket was invalidated
 *
 * - allocator => dirty: open bucket was filled up
 * - allocator => cached: open bucket was filled up
 * - allocator => metadata: metadata was allocated
 *
 * - dirty => cached: dirty sectors were copied to a deeper tier
 * - dirty => free: dirty sectors were overwritten or moved (copy gc)
 * - cached => free: cached sectors were overwritten
 *
 * - metadata => free: metadata was freed
 *
 * Oddities:
 * - cached => dirty: a device was removed so formerly replicated data
 *                    is no longer sufficiently replicated
 * - free => cached: cannot happen
 * - free => dirty: cannot happen
 * - free => metadata: cannot happen
 */

#include "bcachefs.h"
#include "alloc_background.h"
#include "bset.h"
#include "btree_gc.h"
#include "btree_update.h"
#include "buckets.h"
#include "ec.h"
#include "error.h"
#include "movinggc.h"
#include "replicas.h"

#include <linux/preempt.h>
#include <trace/events/bcachefs.h>

/*
 * Clear journal_seq_valid for buckets for which it's not needed, to prevent
 * wraparound:
 */
void bch2_bucket_seq_cleanup(struct bch_fs *c)
{
	u64 journal_seq = atomic64_read(&c->journal.seq);
	u16 last_seq_ondisk = c->journal.last_seq_ondisk;
	struct bch_dev *ca;
	struct bucket_array *buckets;
	struct bucket *g;
	struct bucket_mark m;
	unsigned i;

	if (journal_seq - c->last_bucket_seq_cleanup <
	    (1U << (BUCKET_JOURNAL_SEQ_BITS - 2)))
		return;

	c->last_bucket_seq_cleanup = journal_seq;

	for_each_member_device(ca, c, i) {
		down_read(&ca->bucket_lock);
		buckets = bucket_array(ca);

		for_each_bucket(g, buckets) {
			bucket_cmpxchg(g, m, ({
				if (!m.journal_seq_valid ||
				    bucket_needs_journal_commit(m, last_seq_ondisk))
					break;

				m.journal_seq_valid = 0;
			}));
		}
		up_read(&ca->bucket_lock);
	}
}

void bch2_fs_usage_initialize(struct bch_fs *c)
{
	struct bch_fs_usage *usage;
	unsigned i;

	percpu_down_write(&c->mark_lock);
	usage = c->usage_base;

	bch2_fs_usage_acc_to_base(c, 0);
	bch2_fs_usage_acc_to_base(c, 1);

	for (i = 0; i < BCH_REPLICAS_MAX; i++)
		usage->reserved += usage->persistent_reserved[i];

	for (i = 0; i < c->replicas.nr; i++) {
		struct bch_replicas_entry *e =
			cpu_replicas_entry(&c->replicas, i);

		switch (e->data_type) {
		case BCH_DATA_BTREE:
			usage->btree	+= usage->replicas[i];
			break;
		case BCH_DATA_USER:
			usage->data	+= usage->replicas[i];
			break;
		case BCH_DATA_CACHED:
			usage->cached	+= usage->replicas[i];
			break;
		}
	}

	percpu_up_write(&c->mark_lock);
}

void bch2_fs_usage_scratch_put(struct bch_fs *c, struct bch_fs_usage *fs_usage)
{
	if (fs_usage == c->usage_scratch)
		mutex_unlock(&c->usage_scratch_lock);
	else
		kfree(fs_usage);
}

struct bch_fs_usage *bch2_fs_usage_scratch_get(struct bch_fs *c)
{
	struct bch_fs_usage *ret;
	unsigned bytes = fs_usage_u64s(c) * sizeof(u64);

	ret = kzalloc(bytes, GFP_NOWAIT);
	if (ret)
		return ret;

	if (mutex_trylock(&c->usage_scratch_lock))
		goto out_pool;

	ret = kzalloc(bytes, GFP_NOFS);
	if (ret)
		return ret;

	mutex_lock(&c->usage_scratch_lock);
out_pool:
	ret = c->usage_scratch;
	memset(ret, 0, bytes);
	return ret;
}

struct bch_dev_usage bch2_dev_usage_read(struct bch_fs *c, struct bch_dev *ca)
{
	struct bch_dev_usage ret;

	memset(&ret, 0, sizeof(ret));
	acc_u64s_percpu((u64 *) &ret,
			(u64 __percpu *) ca->usage[0],
			sizeof(ret) / sizeof(u64));

	return ret;
}

static inline struct bch_fs_usage *fs_usage_ptr(struct bch_fs *c,
						unsigned journal_seq,
						bool gc)
{
	return this_cpu_ptr(gc
			    ? c->usage_gc
			    : c->usage[journal_seq & 1]);
}

u64 bch2_fs_usage_read_one(struct bch_fs *c, u64 *v)
{
	ssize_t offset = v - (u64 *) c->usage_base;
	unsigned seq;
	u64 ret;

	BUG_ON(offset < 0 || offset >= fs_usage_u64s(c));
	percpu_rwsem_assert_held(&c->mark_lock);

	do {
		seq = read_seqcount_begin(&c->usage_lock);
		ret = *v +
			percpu_u64_get((u64 __percpu *) c->usage[0] + offset) +
			percpu_u64_get((u64 __percpu *) c->usage[1] + offset);
	} while (read_seqcount_retry(&c->usage_lock, seq));

	return ret;
}

struct bch_fs_usage *bch2_fs_usage_read(struct bch_fs *c)
{
	struct bch_fs_usage *ret;
	unsigned seq, v, u64s = fs_usage_u64s(c);
retry:
	ret = kmalloc(u64s * sizeof(u64), GFP_NOFS);
	if (unlikely(!ret))
		return NULL;

	percpu_down_read(&c->mark_lock);

	v = fs_usage_u64s(c);
	if (unlikely(u64s != v)) {
		u64s = v;
		percpu_up_read(&c->mark_lock);
		kfree(ret);
		goto retry;
	}

	do {
		seq = read_seqcount_begin(&c->usage_lock);
		memcpy(ret, c->usage_base, u64s * sizeof(u64));
		acc_u64s_percpu((u64 *) ret, (u64 __percpu *) c->usage[0], u64s);
		acc_u64s_percpu((u64 *) ret, (u64 __percpu *) c->usage[1], u64s);
	} while (read_seqcount_retry(&c->usage_lock, seq));

	return ret;
}

void bch2_fs_usage_acc_to_base(struct bch_fs *c, unsigned idx)
{
	unsigned u64s = fs_usage_u64s(c);

	BUG_ON(idx >= 2);

	write_seqcount_begin(&c->usage_lock);

	acc_u64s_percpu((u64 *) c->usage_base,
			(u64 __percpu *) c->usage[idx], u64s);
	percpu_memset(c->usage[idx], 0, u64s * sizeof(u64));

	write_seqcount_end(&c->usage_lock);
}

void bch2_fs_usage_to_text(struct printbuf *out,
			   struct bch_fs *c,
			   struct bch_fs_usage *fs_usage)
{
	unsigned i;

	pr_buf(out, "capacity:\t\t\t%llu\n", c->capacity);

	pr_buf(out, "hidden:\t\t\t\t%llu\n",
	       fs_usage->hidden);
	pr_buf(out, "data:\t\t\t\t%llu\n",
	       fs_usage->data);
	pr_buf(out, "cached:\t\t\t\t%llu\n",
	       fs_usage->cached);
	pr_buf(out, "reserved:\t\t\t%llu\n",
	       fs_usage->reserved);
	pr_buf(out, "nr_inodes:\t\t\t%llu\n",
	       fs_usage->nr_inodes);
	pr_buf(out, "online reserved:\t\t%llu\n",
	       fs_usage->online_reserved);

	for (i = 0;
	     i < ARRAY_SIZE(fs_usage->persistent_reserved);
	     i++) {
		pr_buf(out, "%u replicas:\n", i + 1);
		pr_buf(out, "\treserved:\t\t%llu\n",
		       fs_usage->persistent_reserved[i]);
	}

	for (i = 0; i < c->replicas.nr; i++) {
		struct bch_replicas_entry *e =
			cpu_replicas_entry(&c->replicas, i);

		pr_buf(out, "\t");
		bch2_replicas_entry_to_text(out, e);
		pr_buf(out, ":\t%llu\n", fs_usage->replicas[i]);
	}
}

#define RESERVE_FACTOR	6

static u64 reserve_factor(u64 r)
{
	return r + (round_up(r, (1 << RESERVE_FACTOR)) >> RESERVE_FACTOR);
}

static u64 avail_factor(u64 r)
{
	return (r << RESERVE_FACTOR) / ((1 << RESERVE_FACTOR) + 1);
}

u64 bch2_fs_sectors_used(struct bch_fs *c, struct bch_fs_usage *fs_usage)
{
	return min(fs_usage->hidden +
		   fs_usage->btree +
		   fs_usage->data +
		   reserve_factor(fs_usage->reserved +
				  fs_usage->online_reserved),
		   c->capacity);
}

static struct bch_fs_usage_short
__bch2_fs_usage_read_short(struct bch_fs *c)
{
	struct bch_fs_usage_short ret;
	u64 data, reserved;

	ret.capacity = c->capacity -
		bch2_fs_usage_read_one(c, &c->usage_base->hidden);

	data		= bch2_fs_usage_read_one(c, &c->usage_base->data) +
		bch2_fs_usage_read_one(c, &c->usage_base->btree);
	reserved	= bch2_fs_usage_read_one(c, &c->usage_base->reserved) +
		bch2_fs_usage_read_one(c, &c->usage_base->online_reserved);

	ret.used	= min(ret.capacity, data + reserve_factor(reserved));
	ret.free	= ret.capacity - ret.used;

	ret.nr_inodes	= bch2_fs_usage_read_one(c, &c->usage_base->nr_inodes);

	return ret;
}

struct bch_fs_usage_short
bch2_fs_usage_read_short(struct bch_fs *c)
{
	struct bch_fs_usage_short ret;

	percpu_down_read(&c->mark_lock);
	ret = __bch2_fs_usage_read_short(c);
	percpu_up_read(&c->mark_lock);

	return ret;
}

static inline int is_unavailable_bucket(struct bucket_mark m)
{
	return !is_available_bucket(m);
}

static inline int is_fragmented_bucket(struct bucket_mark m,
				       struct bch_dev *ca)
{
	if (!m.owned_by_allocator &&
	    m.data_type == BCH_DATA_USER &&
	    bucket_sectors_used(m))
		return max_t(int, 0, (int) ca->mi.bucket_size -
			     bucket_sectors_used(m));
	return 0;
}

static inline enum bch_data_type bucket_type(struct bucket_mark m)
{
	return m.cached_sectors && !m.dirty_sectors
		? BCH_DATA_CACHED
		: m.data_type;
}

static bool bucket_became_unavailable(struct bucket_mark old,
				      struct bucket_mark new)
{
	return is_available_bucket(old) &&
	       !is_available_bucket(new);
}

int bch2_fs_usage_apply(struct bch_fs *c,
			struct bch_fs_usage *fs_usage,
			struct disk_reservation *disk_res,
			unsigned journal_seq)
{
	s64 added = fs_usage->data + fs_usage->reserved;
	s64 should_not_have_added;
	int ret = 0;

	percpu_rwsem_assert_held(&c->mark_lock);

	/*
	 * Not allowed to reduce sectors_available except by getting a
	 * reservation:
	 */
	should_not_have_added = added - (s64) (disk_res ? disk_res->sectors : 0);
	if (WARN_ONCE(should_not_have_added > 0,
		      "disk usage increased without a reservation")) {
		atomic64_sub(should_not_have_added, &c->sectors_available);
		added -= should_not_have_added;
		ret = -1;
	}

	if (added > 0) {
		disk_res->sectors		-= added;
		fs_usage->online_reserved	-= added;
	}

	preempt_disable();
	acc_u64s((u64 *) fs_usage_ptr(c, journal_seq, false),
		 (u64 *) fs_usage, fs_usage_u64s(c));
	preempt_enable();

	return ret;
}

static inline void account_bucket(struct bch_fs_usage *fs_usage,
				  struct bch_dev_usage *dev_usage,
				  enum bch_data_type type,
				  int nr, s64 size)
{
	if (type == BCH_DATA_SB || type == BCH_DATA_JOURNAL)
		fs_usage->hidden	+= size;

	dev_usage->buckets[type]	+= nr;
}

static void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca,
				  struct bch_fs_usage *fs_usage,
				  struct bucket_mark old, struct bucket_mark new,
				  bool gc)
{
	struct bch_dev_usage *dev_usage;

	percpu_rwsem_assert_held(&c->mark_lock);

	bch2_fs_inconsistent_on(old.data_type && new.data_type &&
				old.data_type != new.data_type, c,
		"different types of data in same bucket: %s, %s",
		bch2_data_types[old.data_type],
		bch2_data_types[new.data_type]);

	preempt_disable();
	dev_usage = this_cpu_ptr(ca->usage[gc]);

	if (bucket_type(old))
		account_bucket(fs_usage, dev_usage, bucket_type(old),
			       -1, -ca->mi.bucket_size);

	if (bucket_type(new))
		account_bucket(fs_usage, dev_usage, bucket_type(new),
			       1, ca->mi.bucket_size);

	dev_usage->buckets_alloc +=
		(int) new.owned_by_allocator - (int) old.owned_by_allocator;
	dev_usage->buckets_ec +=
		(int) new.stripe - (int) old.stripe;
	dev_usage->buckets_unavailable +=
		is_unavailable_bucket(new) - is_unavailable_bucket(old);

	dev_usage->sectors[old.data_type] -= old.dirty_sectors;
	dev_usage->sectors[new.data_type] += new.dirty_sectors;
	dev_usage->sectors[BCH_DATA_CACHED] +=
		(int) new.cached_sectors - (int) old.cached_sectors;
	dev_usage->sectors_fragmented +=
		is_fragmented_bucket(new, ca) - is_fragmented_bucket(old, ca);
	preempt_enable();

	if (!is_available_bucket(old) && is_available_bucket(new))
		bch2_wake_allocator(ca);
}

void bch2_dev_usage_from_buckets(struct bch_fs *c)
{
	struct bch_dev *ca;
	struct bucket_mark old = { .v.counter = 0 };
	struct bucket_array *buckets;
	struct bucket *g;
	unsigned i;
	int cpu;

	c->usage_base->hidden = 0;

	for_each_member_device(ca, c, i) {
		for_each_possible_cpu(cpu)
			memset(per_cpu_ptr(ca->usage[0], cpu), 0,
			       sizeof(*ca->usage[0]));

		buckets = bucket_array(ca);

		for_each_bucket(g, buckets)
			bch2_dev_usage_update(c, ca, c->usage_base,
					      old, g->mark, false);
	}
}

#define bucket_data_cmpxchg(c, ca, fs_usage, g, new, expr)	\
({								\
	struct bucket_mark _old = bucket_cmpxchg(g, new, expr);	\
								\
	bch2_dev_usage_update(c, ca, fs_usage, _old, new, gc);	\
	_old;							\
})

static inline void update_replicas(struct bch_fs *c,
				   struct bch_fs_usage *fs_usage,
				   struct bch_replicas_entry *r,
				   s64 sectors)
{
	int idx = bch2_replicas_entry_idx(c, r);

	BUG_ON(idx < 0);
	BUG_ON(!sectors);

	switch (r->data_type) {
	case BCH_DATA_BTREE:
		fs_usage->btree		+= sectors;
		break;
	case BCH_DATA_USER:
		fs_usage->data		+= sectors;
		break;
	case BCH_DATA_CACHED:
		fs_usage->cached	+= sectors;
		break;
	}
	fs_usage->replicas[idx]		+= sectors;
}

static inline void update_cached_sectors(struct bch_fs *c,
					 struct bch_fs_usage *fs_usage,
					 unsigned dev, s64 sectors)
{
	struct bch_replicas_padded r;

	bch2_replicas_entry_cached(&r.e, dev);

	update_replicas(c, fs_usage, &r.e, sectors);
}

static struct replicas_delta_list *
replicas_deltas_realloc(struct btree_trans *trans, unsigned more)
{
	struct replicas_delta_list *d = trans->fs_usage_deltas;
	unsigned new_size = d ? (d->size + more) * 2 : 128;

	if (!d || d->used + more > d->size) {
		d = krealloc(d, sizeof(*d) + new_size, GFP_NOIO|__GFP_ZERO);
		BUG_ON(!d);

		d->size = new_size;
		trans->fs_usage_deltas = d;
	}
	return d;
}

static inline void update_replicas_list(struct btree_trans *trans,
					struct bch_replicas_entry *r,
					s64 sectors)
{
	struct replicas_delta_list *d;
	struct replicas_delta *n;
	unsigned b = replicas_entry_bytes(r) + 8;

	d = replicas_deltas_realloc(trans, b);

	n = (void *) d->d + d->used;
	n->delta = sectors;
	memcpy(&n->r, r, replicas_entry_bytes(r));
	d->used += b;
}

static inline void update_cached_sectors_list(struct btree_trans *trans,
					      unsigned dev, s64 sectors)
{
	struct bch_replicas_padded r;

	bch2_replicas_entry_cached(&r.e, dev);

	update_replicas_list(trans, &r.e, sectors);
}

void bch2_replicas_delta_list_apply(struct bch_fs *c,
				    struct bch_fs_usage *fs_usage,
				    struct replicas_delta_list *r)
{
	struct replicas_delta *d = r->d;
	struct replicas_delta *top = (void *) r->d + r->used;

	acc_u64s((u64 *) fs_usage,
		 (u64 *) &r->fs_usage, sizeof(*fs_usage) / sizeof(u64));

	while (d != top) {
		BUG_ON((void *) d > (void *) top);

		update_replicas(c, fs_usage, &d->r, d->delta);

		d = (void *) d + replicas_entry_bytes(&d->r) + 8;
	}
}

#define do_mark_fn(fn, c, pos, flags, ...)				\
({									\
	int gc, ret = 0;						\
									\
	percpu_rwsem_assert_held(&c->mark_lock);			\
									\
	for (gc = 0; gc < 2 && !ret; gc++)				\
		if (!gc == !(flags & BCH_BUCKET_MARK_GC) ||		\
		    (gc && gc_visited(c, pos)))				\
			ret = fn(c, __VA_ARGS__, gc);			\
	ret;								\
})

static int __bch2_invalidate_bucket(struct bch_fs *c, struct bch_dev *ca,
				    size_t b, struct bucket_mark *ret,
				    bool gc)
{
	struct bch_fs_usage *fs_usage = fs_usage_ptr(c, 0, gc);
	struct bucket *g = __bucket(ca, b, gc);
	struct bucket_mark old, new;

	old = bucket_data_cmpxchg(c, ca, fs_usage, g, new, ({
		BUG_ON(!is_available_bucket(new));

		new.owned_by_allocator	= true;
		new.dirty		= true;
		new.data_type		= 0;
		new.cached_sectors	= 0;
		new.dirty_sectors	= 0;
		new.gen++;
	}));

	if (old.cached_sectors)
		update_cached_sectors(c, fs_usage, ca->dev_idx,
				      -((s64) old.cached_sectors));

	if (!gc)
		*ret = old;
	return 0;
}

void bch2_invalidate_bucket(struct bch_fs *c, struct bch_dev *ca,
			    size_t b, struct bucket_mark *old)
{
	do_mark_fn(__bch2_invalidate_bucket, c, gc_phase(GC_PHASE_START), 0,
		   ca, b, old);

	if (!old->owned_by_allocator && old->cached_sectors)
		trace_invalidate(ca, bucket_to_sector(ca, b),
				 old->cached_sectors);
}

static int __bch2_mark_alloc_bucket(struct bch_fs *c, struct bch_dev *ca,
				    size_t b, bool owned_by_allocator,
				    bool gc)
{
	struct bch_fs_usage *fs_usage = fs_usage_ptr(c, 0, gc);
	struct bucket *g = __bucket(ca, b, gc);
	struct bucket_mark old, new;

	old = bucket_data_cmpxchg(c, ca, fs_usage, g, new, ({
		new.owned_by_allocator	= owned_by_allocator;
	}));

	BUG_ON(!gc &&
	       !owned_by_allocator && !old.owned_by_allocator);

	return 0;
}

void bch2_mark_alloc_bucket(struct bch_fs *c, struct bch_dev *ca,
			    size_t b, bool owned_by_allocator,
			    struct gc_pos pos, unsigned flags)
{
	preempt_disable();

	do_mark_fn(__bch2_mark_alloc_bucket, c, pos, flags,
		   ca, b, owned_by_allocator);

	preempt_enable();
}

static int bch2_mark_alloc(struct bch_fs *c, struct bkey_s_c k,
			   struct bch_fs_usage *fs_usage,
			   u64 journal_seq, unsigned flags)
{
	bool gc = flags & BCH_BUCKET_MARK_GC;
	struct bkey_alloc_unpacked u;
	struct bch_dev *ca;
	struct bucket *g;
	struct bucket_mark old, m;

	/*
	 * alloc btree is read in by bch2_alloc_read, not gc:
	 */
	if ((flags & BCH_BUCKET_MARK_GC) &&
	    !(flags & BCH_BUCKET_MARK_BUCKET_INVALIDATE))
		return 0;

	ca = bch_dev_bkey_exists(c, k.k->p.inode);

	if (k.k->p.offset >= ca->mi.nbuckets)
		return 0;

	g = __bucket(ca, k.k->p.offset, gc);
	u = bch2_alloc_unpack(k);

	old = bucket_cmpxchg(g, m, ({
		m.gen			= u.gen;
		m.data_type		= u.data_type;
		m.dirty_sectors		= u.dirty_sectors;
		m.cached_sectors	= u.cached_sectors;

		if (journal_seq) {
			m.journal_seq_valid	= 1;
			m.journal_seq		= journal_seq;
		}
	}));

	if (!(flags & BCH_BUCKET_MARK_ALLOC_READ))
		bch2_dev_usage_update(c, ca, fs_usage, old, m, gc);

	g->io_time[READ]	= u.read_time;
	g->io_time[WRITE]	= u.write_time;
	g->oldest_gen		= u.oldest_gen;
	g->gen_valid		= 1;

	/*
	 * need to know if we're getting called from the invalidate path or
	 * not:
	 */

	if ((flags & BCH_BUCKET_MARK_BUCKET_INVALIDATE) &&
	    old.cached_sectors) {
		update_cached_sectors(c, fs_usage, ca->dev_idx,
				      -old.cached_sectors);
		trace_invalidate(ca, bucket_to_sector(ca, k.k->p.offset),
				 old.cached_sectors);
	}

	return 0;
}

#define checked_add(a, b)					\
({								\
	unsigned _res = (unsigned) (a) + (b);			\
	bool overflow = _res > U16_MAX;				\
	if (overflow)						\
		_res = U16_MAX;					\
	(a) = _res;						\
	overflow;						\
})

static int __bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
				       size_t b, enum bch_data_type type,
				       unsigned sectors, bool gc)
{
	struct bucket *g = __bucket(ca, b, gc);
	struct bucket_mark old, new;
	bool overflow;

	BUG_ON(type != BCH_DATA_SB &&
	       type != BCH_DATA_JOURNAL);

	old = bucket_cmpxchg(g, new, ({
		new.dirty	= true;
		new.data_type	= type;
		overflow = checked_add(new.dirty_sectors, sectors);
	}));

	bch2_fs_inconsistent_on(overflow, c,
		"bucket sector count overflow: %u + %u > U16_MAX",
		old.dirty_sectors, sectors);

	if (c)
		bch2_dev_usage_update(c, ca, fs_usage_ptr(c, 0, gc),
				      old, new, gc);

	return 0;
}

void bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
			       size_t b, enum bch_data_type type,
			       unsigned sectors, struct gc_pos pos,
			       unsigned flags)
{
	BUG_ON(type != BCH_DATA_SB &&
	       type != BCH_DATA_JOURNAL);

	preempt_disable();

	if (likely(c)) {
		do_mark_fn(__bch2_mark_metadata_bucket, c, pos, flags,
			   ca, b, type, sectors);
	} else {
		__bch2_mark_metadata_bucket(c, ca, b, type, sectors, 0);
	}

	preempt_enable();
}

static s64 ptr_disk_sectors_delta(struct extent_ptr_decoded p,
				  s64 delta)
{
	if (delta > 0) {
		/*
		 * marking a new extent, which _will have size_ @delta
		 *
		 * in the bch2_mark_update -> BCH_EXTENT_OVERLAP_MIDDLE
		 * case, we haven't actually created the key we'll be inserting
		 * yet (for the split) - so we don't want to be using
		 * k->size/crc.live_size here:
		 */
		return __ptr_disk_sectors(p, delta);
	} else {
		BUG_ON(-delta > p.crc.live_size);

		return (s64) __ptr_disk_sectors(p, p.crc.live_size + delta) -
			(s64) ptr_disk_sectors(p);
	}
}

static void bucket_set_stripe(struct bch_fs *c,
			      const struct bch_stripe *v,
			      struct bch_fs_usage *fs_usage,
			      u64 journal_seq,
			      unsigned flags)
{
	bool enabled = !(flags & BCH_BUCKET_MARK_OVERWRITE);
	bool gc = flags & BCH_BUCKET_MARK_GC;
	unsigned i;

	for (i = 0; i < v->nr_blocks; i++) {
		const struct bch_extent_ptr *ptr = v->ptrs + i;
		struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
		struct bucket *g = PTR_BUCKET(ca, ptr, gc);
		struct bucket_mark new, old;

		BUG_ON(ptr_stale(ca, ptr));

		old = bucket_data_cmpxchg(c, ca, fs_usage, g, new, ({
			new.dirty			= true;
			new.stripe			= enabled;
			if (journal_seq) {
				new.journal_seq_valid	= 1;
				new.journal_seq		= journal_seq;
			}
		}));
	}
}

static bool bch2_mark_pointer(struct bch_fs *c,
			      struct extent_ptr_decoded p,
			      s64 sectors, enum bch_data_type data_type,
			      struct bch_fs_usage *fs_usage,
			      u64 journal_seq, unsigned flags)
{
	bool gc = flags & BCH_BUCKET_MARK_GC;
	struct bucket_mark old, new;
	struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
	struct bucket *g = PTR_BUCKET(ca, &p.ptr, gc);
	bool overflow;
	u64 v;

	v = atomic64_read(&g->_mark.v);
	do {
		new.v.counter = old.v.counter = v;

		new.dirty = true;

		/*
		 * Check this after reading bucket mark to guard against
		 * the allocator invalidating a bucket after we've already
		 * checked the gen
		 */
		if (gen_after(new.gen, p.ptr.gen)) {
			BUG_ON(!test_bit(BCH_FS_ALLOC_READ_DONE, &c->flags));
			EBUG_ON(!p.ptr.cached &&
				test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags));
			return true;
		}

		if (!p.ptr.cached)
			overflow = checked_add(new.dirty_sectors, sectors);
		else
			overflow = checked_add(new.cached_sectors, sectors);

		if (!new.dirty_sectors &&
		    !new.cached_sectors) {
			new.data_type	= 0;

			if (journal_seq) {
				new.journal_seq_valid = 1;
				new.journal_seq = journal_seq;
			}
		} else {
			new.data_type = data_type;
		}

		if (flags & BCH_BUCKET_MARK_NOATOMIC) {
			g->_mark = new;
			break;
		}
	} while ((v = atomic64_cmpxchg(&g->_mark.v,
			      old.v.counter,
			      new.v.counter)) != old.v.counter);

	bch2_fs_inconsistent_on(overflow, c,
		"bucket sector count overflow: %u + %lli > U16_MAX",
		!p.ptr.cached
		? old.dirty_sectors
		: old.cached_sectors, sectors);

	bch2_dev_usage_update(c, ca, fs_usage, old, new, gc);

	BUG_ON(!gc && bucket_became_unavailable(old, new));

	return false;
}

static int bch2_mark_stripe_ptr(struct bch_fs *c,
				struct bch_extent_stripe_ptr p,
				enum bch_data_type data_type,
				struct bch_fs_usage *fs_usage,
				s64 sectors, unsigned flags)
{
	bool gc = flags & BCH_BUCKET_MARK_GC;
	struct stripe *m;
	unsigned old, new, nr_data;
	int blocks_nonempty_delta;
	s64 parity_sectors;

	BUG_ON(!sectors);

	m = genradix_ptr(&c->stripes[gc], p.idx);

	spin_lock(&c->ec_stripes_heap_lock);

	if (!m || !m->alive) {
		spin_unlock(&c->ec_stripes_heap_lock);
		bch_err_ratelimited(c, "pointer to nonexistent stripe %llu",
				    (u64) p.idx);
		return -1;
	}

	BUG_ON(m->r.e.data_type != data_type);

	nr_data = m->nr_blocks - m->nr_redundant;

	parity_sectors = DIV_ROUND_UP(abs(sectors) * m->nr_redundant, nr_data);

	if (sectors < 0)
		parity_sectors = -parity_sectors;
	sectors += parity_sectors;

	old = m->block_sectors[p.block];
	m->block_sectors[p.block] += sectors;
	new = m->block_sectors[p.block];

	blocks_nonempty_delta = (int) !!new - (int) !!old;
	if (blocks_nonempty_delta) {
		m->blocks_nonempty += blocks_nonempty_delta;

		if (!gc)
			bch2_stripes_heap_update(c, m, p.idx);
	}

	m->dirty = true;

	spin_unlock(&c->ec_stripes_heap_lock);

	update_replicas(c, fs_usage, &m->r.e, sectors);

	return 0;
}

static int bch2_mark_extent(struct bch_fs *c, struct bkey_s_c k,
			    s64 sectors, enum bch_data_type data_type,
			    struct bch_fs_usage *fs_usage,
			    unsigned journal_seq, unsigned flags)
{
	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
	const union bch_extent_entry *entry;
	struct extent_ptr_decoded p;
	struct bch_replicas_padded r;
	s64 dirty_sectors = 0;
	unsigned i;
	int ret;

	r.e.data_type	= data_type;
	r.e.nr_devs	= 0;
	r.e.nr_required	= 1;

	BUG_ON(!sectors);

	bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
		s64 disk_sectors = data_type == BCH_DATA_BTREE
			? sectors
			: ptr_disk_sectors_delta(p, sectors);
		bool stale = bch2_mark_pointer(c, p, disk_sectors, data_type,
					fs_usage, journal_seq, flags);

		if (p.ptr.cached) {
			if (disk_sectors && !stale)
				update_cached_sectors(c, fs_usage, p.ptr.dev,
						      disk_sectors);
		} else if (!p.ec_nr) {
			dirty_sectors	       += disk_sectors;
			r.e.devs[r.e.nr_devs++]	= p.ptr.dev;
		} else {
			for (i = 0; i < p.ec_nr; i++) {
				ret = bch2_mark_stripe_ptr(c, p.ec[i],
						data_type, fs_usage,
						disk_sectors, flags);
				if (ret)
					return ret;
			}

			r.e.nr_required = 0;
		}
	}

	if (dirty_sectors)
		update_replicas(c, fs_usage, &r.e, dirty_sectors);

	return 0;
}

static int bch2_mark_stripe(struct bch_fs *c, struct bkey_s_c k,
			    struct bch_fs_usage *fs_usage,
			    u64 journal_seq, unsigned flags)
{
	bool gc = flags & BCH_BUCKET_MARK_GC;
	struct bkey_s_c_stripe s = bkey_s_c_to_stripe(k);
	size_t idx = s.k->p.offset;
	struct stripe *m = genradix_ptr(&c->stripes[gc], idx);
	unsigned i;

	spin_lock(&c->ec_stripes_heap_lock);

	if (!m || ((flags & BCH_BUCKET_MARK_OVERWRITE) && !m->alive)) {
		spin_unlock(&c->ec_stripes_heap_lock);
		bch_err_ratelimited(c, "error marking nonexistent stripe %zu",
				    idx);
		return -1;
	}

	if (!(flags & BCH_BUCKET_MARK_OVERWRITE)) {
		m->sectors	= le16_to_cpu(s.v->sectors);
		m->algorithm	= s.v->algorithm;
		m->nr_blocks	= s.v->nr_blocks;
		m->nr_redundant	= s.v->nr_redundant;

		bch2_bkey_to_replicas(&m->r.e, k);

		/*
		 * XXX: account for stripes somehow here
		 */
#if 0
		update_replicas(c, fs_usage, &m->r.e, stripe_sectors);
#endif

		/* gc recalculates these fields: */
		if (!(flags & BCH_BUCKET_MARK_GC)) {
			for (i = 0; i < s.v->nr_blocks; i++) {
				m->block_sectors[i] =
					stripe_blockcount_get(s.v, i);
				m->blocks_nonempty += !!m->block_sectors[i];
			}
		}

		if (!gc)
			bch2_stripes_heap_update(c, m, idx);
		m->alive	= true;
	} else {
		if (!gc)
			bch2_stripes_heap_del(c, m, idx);
		memset(m, 0, sizeof(*m));
	}

	spin_unlock(&c->ec_stripes_heap_lock);

	bucket_set_stripe(c, s.v, fs_usage, 0, flags);
	return 0;
}

int bch2_mark_key_locked(struct bch_fs *c,
		   struct bkey_s_c k, s64 sectors,
		   struct bch_fs_usage *fs_usage,
		   u64 journal_seq, unsigned flags)
{
	int ret = 0;

	preempt_disable();

	if (!fs_usage || (flags & BCH_BUCKET_MARK_GC))
		fs_usage = fs_usage_ptr(c, journal_seq,
					flags & BCH_BUCKET_MARK_GC);

	switch (k.k->type) {
	case KEY_TYPE_alloc:
		ret = bch2_mark_alloc(c, k, fs_usage, journal_seq, flags);
		break;
	case KEY_TYPE_btree_ptr:
		sectors = !(flags & BCH_BUCKET_MARK_OVERWRITE)
			?  c->opts.btree_node_size
			: -c->opts.btree_node_size;

		ret = bch2_mark_extent(c, k, sectors, BCH_DATA_BTREE,
				fs_usage, journal_seq, flags);
		break;
	case KEY_TYPE_extent:
		ret = bch2_mark_extent(c, k, sectors, BCH_DATA_USER,
				fs_usage, journal_seq, flags);
		break;
	case KEY_TYPE_stripe:
		ret = bch2_mark_stripe(c, k, fs_usage, journal_seq, flags);
		break;
	case KEY_TYPE_inode:
		if (!(flags & BCH_BUCKET_MARK_OVERWRITE))
			fs_usage->nr_inodes++;
		else
			fs_usage->nr_inodes--;
		break;
	case KEY_TYPE_reservation: {
		unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas;

		sectors *= replicas;
		replicas = clamp_t(unsigned, replicas, 1,
				   ARRAY_SIZE(fs_usage->persistent_reserved));

		fs_usage->reserved				+= sectors;
		fs_usage->persistent_reserved[replicas - 1]	+= sectors;
		break;
	}
	}

	preempt_enable();

	return ret;
}

int bch2_mark_key(struct bch_fs *c, struct bkey_s_c k,
		  s64 sectors,
		  struct bch_fs_usage *fs_usage,
		  u64 journal_seq, unsigned flags)
{
	int ret;

	percpu_down_read(&c->mark_lock);
	ret = bch2_mark_key_locked(c, k, sectors,
				   fs_usage, journal_seq, flags);
	percpu_up_read(&c->mark_lock);

	return ret;
}

inline int bch2_mark_overwrite(struct btree_trans *trans,
			       struct btree_iter *iter,
			       struct bkey_s_c old,
			       struct bkey_i *new,
			       struct bch_fs_usage *fs_usage,
			       unsigned flags)
{
	struct bch_fs		*c = trans->c;
	struct btree		*b = iter->l[0].b;
	s64			sectors = 0;

	if (btree_node_is_extents(b)
	    ? bkey_cmp(new->k.p, bkey_start_pos(old.k)) <= 0
	    : bkey_cmp(new->k.p, old.k->p))
		return 0;

	if (btree_node_is_extents(b)) {
		switch (bch2_extent_overlap(&new->k, old.k)) {
		case BCH_EXTENT_OVERLAP_ALL:
			sectors = -((s64) old.k->size);
			break;
		case BCH_EXTENT_OVERLAP_BACK:
			sectors = bkey_start_offset(&new->k) -
				old.k->p.offset;
			break;
		case BCH_EXTENT_OVERLAP_FRONT:
			sectors = bkey_start_offset(old.k) -
				new->k.p.offset;
			break;
		case BCH_EXTENT_OVERLAP_MIDDLE:
			sectors = old.k->p.offset - new->k.p.offset;
			BUG_ON(sectors <= 0);

			bch2_mark_key_locked(c, old, sectors,
				fs_usage, trans->journal_res.seq,
				BCH_BUCKET_MARK_INSERT|flags);

			sectors = bkey_start_offset(&new->k) -
				old.k->p.offset;
			break;
		}

		BUG_ON(sectors >= 0);
	}

	return bch2_mark_key_locked(c, old, sectors, fs_usage,
				    trans->journal_res.seq,
				    BCH_BUCKET_MARK_OVERWRITE|flags) ?: 1;
}

int bch2_mark_update(struct btree_trans *trans,
		     struct btree_insert_entry *insert,
		     struct bch_fs_usage *fs_usage,
		     unsigned flags)
{
	struct bch_fs		*c = trans->c;
	struct btree_iter	*iter = insert->iter;
	struct btree		*b = iter->l[0].b;
	struct btree_node_iter	node_iter = iter->l[0].iter;
	struct bkey_packed	*_k;
	int ret = 0;

	if (!btree_node_type_needs_gc(iter->btree_id))
		return 0;

	if (!(trans->flags & BTREE_INSERT_NOMARK_INSERT))
		bch2_mark_key_locked(c, bkey_i_to_s_c(insert->k),
			bpos_min(insert->k->k.p, b->key.k.p).offset -
			bkey_start_offset(&insert->k->k),
			fs_usage, trans->journal_res.seq,
			BCH_BUCKET_MARK_INSERT|flags);

	if (unlikely(trans->flags & BTREE_INSERT_NOMARK_OVERWRITES))
		return 0;

	/*
	 * For non extents, we only mark the new key, not the key being
	 * overwritten - unless we're actually deleting:
	 */
	if ((iter->btree_id == BTREE_ID_ALLOC ||
	     iter->btree_id == BTREE_ID_EC) &&
	    !bkey_deleted(&insert->k->k))
		return 0;

	while ((_k = bch2_btree_node_iter_peek_filter(&node_iter, b,
						      KEY_TYPE_discard))) {
		struct bkey		unpacked;
		struct bkey_s_c		k = bkey_disassemble(b, _k, &unpacked);

		ret = bch2_mark_overwrite(trans, iter, k, insert->k,
					  fs_usage, flags);
		if (ret <= 0)
			break;

		bch2_btree_node_iter_advance(&node_iter, b);
	}

	return ret;
}

void bch2_trans_fs_usage_apply(struct btree_trans *trans,
			       struct bch_fs_usage *fs_usage)
{
	struct bch_fs *c = trans->c;
	struct btree_insert_entry *i;
	static int warned_disk_usage = 0;
	u64 disk_res_sectors = trans->disk_res ? trans->disk_res->sectors : 0;
	char buf[200];

	if (!bch2_fs_usage_apply(c, fs_usage, trans->disk_res,
				 trans->journal_res.seq) ||
	    warned_disk_usage ||
	    xchg(&warned_disk_usage, 1))
		return;

	pr_err("disk usage increased more than %llu sectors reserved", disk_res_sectors);

	trans_for_each_update_iter(trans, i) {
		struct btree_iter	*iter = i->iter;
		struct btree		*b = iter->l[0].b;
		struct btree_node_iter	node_iter = iter->l[0].iter;
		struct bkey_packed	*_k;

		pr_err("while inserting");
		bch2_bkey_val_to_text(&PBUF(buf), c, bkey_i_to_s_c(i->k));
		pr_err("%s", buf);
		pr_err("overlapping with");

		node_iter = iter->l[0].iter;
		while ((_k = bch2_btree_node_iter_peek_filter(&node_iter, b,
							      KEY_TYPE_discard))) {
			struct bkey		unpacked;
			struct bkey_s_c		k;

			k = bkey_disassemble(b, _k, &unpacked);

			if (btree_node_is_extents(b)
			    ? bkey_cmp(i->k->k.p, bkey_start_pos(k.k)) <= 0
			    : bkey_cmp(i->k->k.p, k.k->p))
				break;

			bch2_bkey_val_to_text(&PBUF(buf), c, k);
			pr_err("%s", buf);

			bch2_btree_node_iter_advance(&node_iter, b);
		}
	}
}

/* trans_mark: */

static int trans_get_key(struct btree_trans *trans,
			 enum btree_id btree_id, struct bpos pos,
			 struct btree_insert_entry **insert,
			 struct btree_iter **iter,
			 struct bkey_s_c *k)
{
	unsigned i;
	int ret;

	*insert = NULL;

	for (i = 0; i < trans->nr_updates; i++)
		if (!trans->updates[i].deferred &&
		    trans->updates[i].iter->btree_id == btree_id &&
		    !bkey_cmp(pos, trans->updates[i].iter->pos)) {
			*insert = &trans->updates[i];
			*iter	= (*insert)->iter;
			*k	= bkey_i_to_s_c((*insert)->k);
			return 0;
		}

	*iter = __bch2_trans_get_iter(trans, btree_id, pos,
				   BTREE_ITER_SLOTS|BTREE_ITER_INTENT, 0);
	if (IS_ERR(*iter))
		return PTR_ERR(*iter);

	*k = bch2_btree_iter_peek_slot(*iter);
	ret = bkey_err(*k);
	if (ret)
		bch2_trans_iter_put(trans, *iter);
	return ret;
}

static int trans_update_key(struct btree_trans *trans,
			    struct btree_insert_entry **insert,
			    struct btree_iter *iter,
			    struct bkey_s_c k,
			    unsigned extra_u64s)
{
	struct bkey_i *new_k;

	if (*insert)
		return 0;

	new_k = bch2_trans_kmalloc(trans, bkey_bytes(k.k) +
				   extra_u64s * sizeof(u64));
	if (IS_ERR(new_k))
		return PTR_ERR(new_k);

	*insert = bch2_trans_update(trans, ((struct btree_insert_entry) {
				.iter = iter,
				.k = new_k,
				.triggered = true,
	}));

	bkey_reassemble((*insert)->k, k);
	return 0;
}

static int bch2_trans_mark_pointer(struct btree_trans *trans,
			struct extent_ptr_decoded p,
			s64 sectors, enum bch_data_type data_type)
{
	struct bch_fs *c = trans->c;
	struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
	struct btree_insert_entry *insert;
	struct btree_iter *iter;
	struct bkey_s_c k;
	struct bkey_alloc_unpacked u;
	struct bkey_i_alloc *a;
	bool overflow;
	int ret;

	ret = trans_get_key(trans, BTREE_ID_ALLOC,
			    POS(p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr)),
			    &insert, &iter, &k);
	if (ret)
		return ret;

	if (k.k->type != KEY_TYPE_alloc) {
		bch_err_ratelimited(c, "pointer to nonexistent bucket %u:%zu",
				    p.ptr.dev,
				    PTR_BUCKET_NR(ca, &p.ptr));
		ret = -1;
		goto out;
	}

	u = bch2_alloc_unpack(k);

	if (gen_after(u.gen, p.ptr.gen)) {
		ret = 1;
		goto out;
	}

	if (!p.ptr.cached)
		overflow = checked_add(u.dirty_sectors, sectors);
	else
		overflow = checked_add(u.cached_sectors, sectors);

	u.data_type = u.dirty_sectors || u.cached_sectors
		? data_type : 0;

	bch2_fs_inconsistent_on(overflow, c,
		"bucket sector count overflow: %u + %lli > U16_MAX",
		!p.ptr.cached
		? u.dirty_sectors
		: u.cached_sectors, sectors);

	ret = trans_update_key(trans, &insert, iter, k, 1);
	if (ret)
		goto out;

	a = bkey_alloc_init(insert->k);
	a->k.p = iter->pos;
	bch2_alloc_pack(a, u);
out:
	bch2_trans_iter_put(trans, iter);
	return ret;
}

static int bch2_trans_mark_stripe_ptr(struct btree_trans *trans,
			struct bch_extent_stripe_ptr p,
			s64 sectors, enum bch_data_type data_type)
{
	struct bch_replicas_padded r;
	struct btree_insert_entry *insert;
	struct btree_iter *iter;
	struct bkey_s_c k;
	struct bkey_s_stripe s;
	unsigned nr_data;
	s64 parity_sectors;
	int ret = 0;

	BUG_ON(!sectors);

	ret = trans_get_key(trans, BTREE_ID_EC, POS(0, p.idx),
			    &insert, &iter, &k);
	if (ret)
		return ret;

	if (k.k->type != KEY_TYPE_stripe) {
		bch_err_ratelimited(trans->c,
				    "pointer to nonexistent stripe %llu",
				    (u64) p.idx);
		ret = -1;
		goto out;
	}

	ret = trans_update_key(trans, &insert, iter, k, 1);
	if (ret)
		goto out;

	s = bkey_i_to_s_stripe(insert->k);

	nr_data = s.v->nr_blocks - s.v->nr_redundant;

	parity_sectors = DIV_ROUND_UP(abs(sectors) * s.v->nr_redundant, nr_data);

	if (sectors < 0)
		parity_sectors = -parity_sectors;

	stripe_blockcount_set(s.v, p.block,
		stripe_blockcount_get(s.v, p.block) +
		sectors + parity_sectors);

	bch2_bkey_to_replicas(&r.e, s.s_c);

	update_replicas_list(trans, &r.e, sectors);
out:
	bch2_trans_iter_put(trans, iter);
	return ret;
}

static int bch2_trans_mark_extent(struct btree_trans *trans,
			struct bkey_s_c k,
			s64 sectors, enum bch_data_type data_type)
{
	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
	const union bch_extent_entry *entry;
	struct extent_ptr_decoded p;
	struct bch_replicas_padded r;
	s64 dirty_sectors = 0;
	bool stale;
	unsigned i;
	int ret;

	r.e.data_type	= data_type;
	r.e.nr_devs	= 0;
	r.e.nr_required	= 1;

	BUG_ON(!sectors);

	bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
		s64 disk_sectors = data_type == BCH_DATA_BTREE
			? sectors
			: ptr_disk_sectors_delta(p, sectors);

		ret = bch2_trans_mark_pointer(trans, p, disk_sectors,
					      data_type);
		if (ret < 0)
			return ret;

		stale = ret > 0;

		if (p.ptr.cached) {
			if (disk_sectors && !stale)
				update_cached_sectors_list(trans, p.ptr.dev,
							   disk_sectors);
		} else if (!p.ec_nr) {
			dirty_sectors	       += disk_sectors;
			r.e.devs[r.e.nr_devs++]	= p.ptr.dev;
		} else {
			for (i = 0; i < p.ec_nr; i++) {
				ret = bch2_trans_mark_stripe_ptr(trans, p.ec[i],
						disk_sectors, data_type);
				if (ret)
					return ret;
			}

			r.e.nr_required = 0;
		}
	}

	if (dirty_sectors)
		update_replicas_list(trans, &r.e, dirty_sectors);

	return 0;
}

int bch2_trans_mark_key(struct btree_trans *trans, struct bkey_s_c k,
			s64 sectors, unsigned flags)
{
	struct replicas_delta_list *d;
	struct bch_fs *c = trans->c;

	switch (k.k->type) {
	case KEY_TYPE_btree_ptr:
		sectors = !(flags & BCH_BUCKET_MARK_OVERWRITE)
			?  c->opts.btree_node_size
			: -c->opts.btree_node_size;

		return bch2_trans_mark_extent(trans, k, sectors,
					      BCH_DATA_BTREE);
	case KEY_TYPE_extent:
		return bch2_trans_mark_extent(trans, k, sectors,
					      BCH_DATA_USER);
	case KEY_TYPE_inode:
		d = replicas_deltas_realloc(trans, 0);

		if (!(flags & BCH_BUCKET_MARK_OVERWRITE))
			d->fs_usage.nr_inodes++;
		else
			d->fs_usage.nr_inodes--;
		return 0;
	case KEY_TYPE_reservation: {
		unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas;

		d = replicas_deltas_realloc(trans, 0);

		sectors *= replicas;
		replicas = clamp_t(unsigned, replicas, 1,
				   ARRAY_SIZE(d->fs_usage.persistent_reserved));

		d->fs_usage.reserved				+= sectors;
		d->fs_usage.persistent_reserved[replicas - 1]	+= sectors;
		return 0;
	}
	default:
		return 0;
	}
}

int bch2_trans_mark_update(struct btree_trans *trans,
			   struct btree_insert_entry *insert)
{
	struct btree_iter	*iter = insert->iter;
	struct btree		*b = iter->l[0].b;
	struct btree_node_iter	node_iter = iter->l[0].iter;
	struct bkey_packed	*_k;
	int ret;

	if (!btree_node_type_needs_gc(iter->btree_id))
		return 0;

	ret = bch2_trans_mark_key(trans,
			bkey_i_to_s_c(insert->k),
			bpos_min(insert->k->k.p, b->key.k.p).offset -
			bkey_start_offset(&insert->k->k),
			BCH_BUCKET_MARK_INSERT);
	if (ret)
		return ret;

	while ((_k = bch2_btree_node_iter_peek_filter(&node_iter, b,
						      KEY_TYPE_discard))) {
		struct bkey		unpacked;
		struct bkey_s_c		k;
		s64			sectors = 0;

		k = bkey_disassemble(b, _k, &unpacked);

		if (btree_node_is_extents(b)
		    ? bkey_cmp(insert->k->k.p, bkey_start_pos(k.k)) <= 0
		    : bkey_cmp(insert->k->k.p, k.k->p))
			break;

		if (btree_node_is_extents(b)) {
			switch (bch2_extent_overlap(&insert->k->k, k.k)) {
			case BCH_EXTENT_OVERLAP_ALL:
				sectors = -((s64) k.k->size);
				break;
			case BCH_EXTENT_OVERLAP_BACK:
				sectors = bkey_start_offset(&insert->k->k) -
					k.k->p.offset;
				break;
			case BCH_EXTENT_OVERLAP_FRONT:
				sectors = bkey_start_offset(k.k) -
					insert->k->k.p.offset;
				break;
			case BCH_EXTENT_OVERLAP_MIDDLE:
				sectors = k.k->p.offset - insert->k->k.p.offset;
				BUG_ON(sectors <= 0);

				ret = bch2_trans_mark_key(trans, k, sectors,
						BCH_BUCKET_MARK_INSERT);
				if (ret)
					return ret;

				sectors = bkey_start_offset(&insert->k->k) -
					k.k->p.offset;
				break;
			}

			BUG_ON(sectors >= 0);
		}

		ret = bch2_trans_mark_key(trans, k, sectors,
					  BCH_BUCKET_MARK_OVERWRITE);
		if (ret)
			return ret;

		bch2_btree_node_iter_advance(&node_iter, b);
	}

	return 0;
}

/* Disk reservations: */

static u64 bch2_recalc_sectors_available(struct bch_fs *c)
{
	percpu_u64_set(&c->pcpu->sectors_available, 0);

	return avail_factor(__bch2_fs_usage_read_short(c).free);
}

void __bch2_disk_reservation_put(struct bch_fs *c, struct disk_reservation *res)
{
	percpu_down_read(&c->mark_lock);
	this_cpu_sub(c->usage[0]->online_reserved,
		     res->sectors);
	percpu_up_read(&c->mark_lock);

	res->sectors = 0;
}

#define SECTORS_CACHE	1024

int bch2_disk_reservation_add(struct bch_fs *c, struct disk_reservation *res,
			      unsigned sectors, int flags)
{
	struct bch_fs_pcpu *pcpu;
	u64 old, v, get;
	s64 sectors_available;
	int ret;

	percpu_down_read(&c->mark_lock);
	preempt_disable();
	pcpu = this_cpu_ptr(c->pcpu);

	if (sectors <= pcpu->sectors_available)
		goto out;

	v = atomic64_read(&c->sectors_available);
	do {
		old = v;
		get = min((u64) sectors + SECTORS_CACHE, old);

		if (get < sectors) {
			preempt_enable();
			percpu_up_read(&c->mark_lock);
			goto recalculate;
		}
	} while ((v = atomic64_cmpxchg(&c->sectors_available,
				       old, old - get)) != old);

	pcpu->sectors_available		+= get;

out:
	pcpu->sectors_available		-= sectors;
	this_cpu_add(c->usage[0]->online_reserved, sectors);
	res->sectors			+= sectors;

	preempt_enable();
	percpu_up_read(&c->mark_lock);
	return 0;

recalculate:
	percpu_down_write(&c->mark_lock);

	sectors_available = bch2_recalc_sectors_available(c);

	if (sectors <= sectors_available ||
	    (flags & BCH_DISK_RESERVATION_NOFAIL)) {
		atomic64_set(&c->sectors_available,
			     max_t(s64, 0, sectors_available - sectors));
		this_cpu_add(c->usage[0]->online_reserved, sectors);
		res->sectors			+= sectors;
		ret = 0;
	} else {
		atomic64_set(&c->sectors_available, sectors_available);
		ret = -ENOSPC;
	}

	percpu_up_write(&c->mark_lock);

	return ret;
}

/* Startup/shutdown: */

static void buckets_free_rcu(struct rcu_head *rcu)
{
	struct bucket_array *buckets =
		container_of(rcu, struct bucket_array, rcu);

	kvpfree(buckets,
		sizeof(struct bucket_array) +
		buckets->nbuckets * sizeof(struct bucket));
}

int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
{
	struct bucket_array *buckets = NULL, *old_buckets = NULL;
	unsigned long *buckets_nouse = NULL;
	unsigned long *buckets_written = NULL;
	alloc_fifo	free[RESERVE_NR];
	alloc_fifo	free_inc;
	alloc_heap	alloc_heap;
	copygc_heap	copygc_heap;

	size_t btree_reserve	= DIV_ROUND_UP(BTREE_NODE_RESERVE,
			     ca->mi.bucket_size / c->opts.btree_node_size);
	/* XXX: these should be tunable */
	size_t reserve_none	= max_t(size_t, 1, nbuckets >> 9);
	size_t copygc_reserve	= max_t(size_t, 2, nbuckets >> 7);
	size_t free_inc_nr	= max(max_t(size_t, 1, nbuckets >> 12),
				      btree_reserve * 2);
	bool resize = ca->buckets[0] != NULL,
	     start_copygc = ca->copygc_thread != NULL;
	int ret = -ENOMEM;
	unsigned i;

	memset(&free,		0, sizeof(free));
	memset(&free_inc,	0, sizeof(free_inc));
	memset(&alloc_heap,	0, sizeof(alloc_heap));
	memset(&copygc_heap,	0, sizeof(copygc_heap));

	if (!(buckets		= kvpmalloc(sizeof(struct bucket_array) +
					    nbuckets * sizeof(struct bucket),
					    GFP_KERNEL|__GFP_ZERO)) ||
	    !(buckets_nouse	= kvpmalloc(BITS_TO_LONGS(nbuckets) *
					    sizeof(unsigned long),
					    GFP_KERNEL|__GFP_ZERO)) ||
	    !(buckets_written	= kvpmalloc(BITS_TO_LONGS(nbuckets) *
					    sizeof(unsigned long),
					    GFP_KERNEL|__GFP_ZERO)) ||
	    !init_fifo(&free[RESERVE_BTREE], btree_reserve, GFP_KERNEL) ||
	    !init_fifo(&free[RESERVE_MOVINGGC],
		       copygc_reserve, GFP_KERNEL) ||
	    !init_fifo(&free[RESERVE_NONE], reserve_none, GFP_KERNEL) ||
	    !init_fifo(&free_inc,	free_inc_nr, GFP_KERNEL) ||
	    !init_heap(&alloc_heap,	ALLOC_SCAN_BATCH(ca) << 1, GFP_KERNEL) ||
	    !init_heap(&copygc_heap,	copygc_reserve, GFP_KERNEL))
		goto err;

	buckets->first_bucket	= ca->mi.first_bucket;
	buckets->nbuckets	= nbuckets;

	bch2_copygc_stop(ca);

	if (resize) {
		down_write(&c->gc_lock);
		down_write(&ca->bucket_lock);
		percpu_down_write(&c->mark_lock);
	}

	old_buckets = bucket_array(ca);

	if (resize) {
		size_t n = min(buckets->nbuckets, old_buckets->nbuckets);

		memcpy(buckets->b,
		       old_buckets->b,
		       n * sizeof(struct bucket));
		memcpy(buckets_nouse,
		       ca->buckets_nouse,
		       BITS_TO_LONGS(n) * sizeof(unsigned long));
		memcpy(buckets_written,
		       ca->buckets_written,
		       BITS_TO_LONGS(n) * sizeof(unsigned long));
	}

	rcu_assign_pointer(ca->buckets[0], buckets);
	buckets = old_buckets;

	swap(ca->buckets_nouse, buckets_nouse);
	swap(ca->buckets_written, buckets_written);

	if (resize)
		percpu_up_write(&c->mark_lock);

	spin_lock(&c->freelist_lock);
	for (i = 0; i < RESERVE_NR; i++) {
		fifo_move(&free[i], &ca->free[i]);
		swap(ca->free[i], free[i]);
	}
	fifo_move(&free_inc, &ca->free_inc);
	swap(ca->free_inc, free_inc);
	spin_unlock(&c->freelist_lock);

	/* with gc lock held, alloc_heap can't be in use: */
	swap(ca->alloc_heap, alloc_heap);

	/* and we shut down copygc: */
	swap(ca->copygc_heap, copygc_heap);

	nbuckets = ca->mi.nbuckets;

	if (resize) {
		up_write(&ca->bucket_lock);
		up_write(&c->gc_lock);
	}

	if (start_copygc &&
	    bch2_copygc_start(c, ca))
		bch_err(ca, "error restarting copygc thread");

	ret = 0;
err:
	free_heap(&copygc_heap);
	free_heap(&alloc_heap);
	free_fifo(&free_inc);
	for (i = 0; i < RESERVE_NR; i++)
		free_fifo(&free[i]);
	kvpfree(buckets_nouse,
		BITS_TO_LONGS(nbuckets) * sizeof(unsigned long));
	kvpfree(buckets_written,
		BITS_TO_LONGS(nbuckets) * sizeof(unsigned long));
	if (buckets)
		call_rcu(&old_buckets->rcu, buckets_free_rcu);

	return ret;
}

void bch2_dev_buckets_free(struct bch_dev *ca)
{
	unsigned i;

	free_heap(&ca->copygc_heap);
	free_heap(&ca->alloc_heap);
	free_fifo(&ca->free_inc);
	for (i = 0; i < RESERVE_NR; i++)
		free_fifo(&ca->free[i]);
	kvpfree(ca->buckets_written,
		BITS_TO_LONGS(ca->mi.nbuckets) * sizeof(unsigned long));
	kvpfree(ca->buckets_nouse,
		BITS_TO_LONGS(ca->mi.nbuckets) * sizeof(unsigned long));
	kvpfree(rcu_dereference_protected(ca->buckets[0], 1),
		sizeof(struct bucket_array) +
		ca->mi.nbuckets * sizeof(struct bucket));

	free_percpu(ca->usage[0]);
}

int bch2_dev_buckets_alloc(struct bch_fs *c, struct bch_dev *ca)
{
	if (!(ca->usage[0] = alloc_percpu(struct bch_dev_usage)))
		return -ENOMEM;

	return bch2_dev_buckets_resize(c, ca, ca->mi.nbuckets);;
}