aboutsummaryrefslogtreecommitdiff
path: root/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/metaslab.c
blob: c5ce27cb677cf5a81b1f5c6958a0dfd76ef13af7 (plain) (blame)
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
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 */

#include <sys/zfs_context.h>
#include <sys/spa_impl.h>
#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/space_map.h>
#include <sys/metaslab_impl.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>

uint64_t metaslab_aliquot = 512ULL << 10;
uint64_t metaslab_gang_bang = SPA_MAXBLOCKSIZE + 1;	/* force gang blocks */

/*
 * Minimum size which forces the dynamic allocator to change
 * it's allocation strategy.  Once the space map cannot satisfy
 * an allocation of this size then it switches to using more
 * aggressive strategy (i.e search by size rather than offset).
 */
uint64_t metaslab_df_alloc_threshold = SPA_MAXBLOCKSIZE;

/*
 * The minimum free space, in percent, which must be available
 * in a space map to continue allocations in a first-fit fashion.
 * Once the space_map's free space drops below this level we dynamically
 * switch to using best-fit allocations.
 */
int metaslab_df_free_pct = 4;

/*
 * A metaslab is considered "free" if it contains a contiguous
 * segment which is greater than metaslab_min_alloc_size.
 */
uint64_t metaslab_min_alloc_size = DMU_MAX_ACCESS;

/*
 * Max number of space_maps to prefetch.
 */
int metaslab_prefetch_limit = SPA_DVAS_PER_BP;

/*
 * Percentage bonus multiplier for metaslabs that are in the bonus area.
 */
int metaslab_smo_bonus_pct = 150;

/*
 * ==========================================================================
 * Metaslab classes
 * ==========================================================================
 */
metaslab_class_t *
metaslab_class_create(space_map_ops_t *ops)
{
	metaslab_class_t *mc;

	mc = kmem_zalloc(sizeof (metaslab_class_t), KM_SLEEP);

	mc->mc_rotor = NULL;
	mc->mc_ops = ops;

	return (mc);
}

void
metaslab_class_destroy(metaslab_class_t *mc)
{
	metaslab_group_t *mg;

	while ((mg = mc->mc_rotor) != NULL) {
		metaslab_class_remove(mc, mg);
		metaslab_group_destroy(mg);
	}

	kmem_free(mc, sizeof (metaslab_class_t));
}

void
metaslab_class_add(metaslab_class_t *mc, metaslab_group_t *mg)
{
	metaslab_group_t *mgprev, *mgnext;

	ASSERT(mg->mg_class == NULL);

	if ((mgprev = mc->mc_rotor) == NULL) {
		mg->mg_prev = mg;
		mg->mg_next = mg;
	} else {
		mgnext = mgprev->mg_next;
		mg->mg_prev = mgprev;
		mg->mg_next = mgnext;
		mgprev->mg_next = mg;
		mgnext->mg_prev = mg;
	}
	mc->mc_rotor = mg;
	mg->mg_class = mc;
}

void
metaslab_class_remove(metaslab_class_t *mc, metaslab_group_t *mg)
{
	metaslab_group_t *mgprev, *mgnext;

	ASSERT(mg->mg_class == mc);

	mgprev = mg->mg_prev;
	mgnext = mg->mg_next;

	if (mg == mgnext) {
		mc->mc_rotor = NULL;
	} else {
		mc->mc_rotor = mgnext;
		mgprev->mg_next = mgnext;
		mgnext->mg_prev = mgprev;
	}

	mg->mg_prev = NULL;
	mg->mg_next = NULL;
	mg->mg_class = NULL;
}

/*
 * ==========================================================================
 * Metaslab groups
 * ==========================================================================
 */
static int
metaslab_compare(const void *x1, const void *x2)
{
	const metaslab_t *m1 = x1;
	const metaslab_t *m2 = x2;

	if (m1->ms_weight < m2->ms_weight)
		return (1);
	if (m1->ms_weight > m2->ms_weight)
		return (-1);

	/*
	 * If the weights are identical, use the offset to force uniqueness.
	 */
	if (m1->ms_map.sm_start < m2->ms_map.sm_start)
		return (-1);
	if (m1->ms_map.sm_start > m2->ms_map.sm_start)
		return (1);

	ASSERT3P(m1, ==, m2);

	return (0);
}

metaslab_group_t *
metaslab_group_create(metaslab_class_t *mc, vdev_t *vd)
{
	metaslab_group_t *mg;

	mg = kmem_zalloc(sizeof (metaslab_group_t), KM_SLEEP);
	mutex_init(&mg->mg_lock, NULL, MUTEX_DEFAULT, NULL);
	avl_create(&mg->mg_metaslab_tree, metaslab_compare,
	    sizeof (metaslab_t), offsetof(struct metaslab, ms_group_node));
	mg->mg_aliquot = metaslab_aliquot * MAX(1, vd->vdev_children);
	mg->mg_vd = vd;
	metaslab_class_add(mc, mg);

	return (mg);
}

void
metaslab_group_destroy(metaslab_group_t *mg)
{
	avl_destroy(&mg->mg_metaslab_tree);
	mutex_destroy(&mg->mg_lock);
	kmem_free(mg, sizeof (metaslab_group_t));
}

static void
metaslab_group_add(metaslab_group_t *mg, metaslab_t *msp)
{
	mutex_enter(&mg->mg_lock);
	ASSERT(msp->ms_group == NULL);
	msp->ms_group = mg;
	msp->ms_weight = 0;
	avl_add(&mg->mg_metaslab_tree, msp);
	mutex_exit(&mg->mg_lock);
}

static void
metaslab_group_remove(metaslab_group_t *mg, metaslab_t *msp)
{
	mutex_enter(&mg->mg_lock);
	ASSERT(msp->ms_group == mg);
	avl_remove(&mg->mg_metaslab_tree, msp);
	msp->ms_group = NULL;
	mutex_exit(&mg->mg_lock);
}

static void
metaslab_group_sort(metaslab_group_t *mg, metaslab_t *msp, uint64_t weight)
{
	/*
	 * Although in principle the weight can be any value, in
	 * practice we do not use values in the range [1, 510].
	 */
	ASSERT(weight >= SPA_MINBLOCKSIZE-1 || weight == 0);
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	mutex_enter(&mg->mg_lock);
	ASSERT(msp->ms_group == mg);
	avl_remove(&mg->mg_metaslab_tree, msp);
	msp->ms_weight = weight;
	avl_add(&mg->mg_metaslab_tree, msp);
	mutex_exit(&mg->mg_lock);
}

/*
 * ==========================================================================
 * Common allocator routines
 * ==========================================================================
 */
static int
metaslab_segsize_compare(const void *x1, const void *x2)
{
	const space_seg_t *s1 = x1;
	const space_seg_t *s2 = x2;
	uint64_t ss_size1 = s1->ss_end - s1->ss_start;
	uint64_t ss_size2 = s2->ss_end - s2->ss_start;

	if (ss_size1 < ss_size2)
		return (-1);
	if (ss_size1 > ss_size2)
		return (1);

	if (s1->ss_start < s2->ss_start)
		return (-1);
	if (s1->ss_start > s2->ss_start)
		return (1);

	return (0);
}

/*
 * This is a helper function that can be used by the allocator to find
 * a suitable block to allocate. This will search the specified AVL
 * tree looking for a block that matches the specified criteria.
 */
static uint64_t
metaslab_block_picker(avl_tree_t *t, uint64_t *cursor, uint64_t size,
    uint64_t align)
{
	space_seg_t *ss, ssearch;
	avl_index_t where;

	ssearch.ss_start = *cursor;
	ssearch.ss_end = *cursor + size;

	ss = avl_find(t, &ssearch, &where);
	if (ss == NULL)
		ss = avl_nearest(t, where, AVL_AFTER);

	while (ss != NULL) {
		uint64_t offset = P2ROUNDUP(ss->ss_start, align);

		if (offset + size <= ss->ss_end) {
			*cursor = offset + size;
			return (offset);
		}
		ss = AVL_NEXT(t, ss);
	}

	/*
	 * If we know we've searched the whole map (*cursor == 0), give up.
	 * Otherwise, reset the cursor to the beginning and try again.
	 */
	if (*cursor == 0)
		return (-1ULL);

	*cursor = 0;
	return (metaslab_block_picker(t, cursor, size, align));
}

static void
metaslab_pp_load(space_map_t *sm)
{
	space_seg_t *ss;

	ASSERT(sm->sm_ppd == NULL);
	sm->sm_ppd = kmem_zalloc(64 * sizeof (uint64_t), KM_SLEEP);

	sm->sm_pp_root = kmem_alloc(sizeof (avl_tree_t), KM_SLEEP);
	avl_create(sm->sm_pp_root, metaslab_segsize_compare,
	    sizeof (space_seg_t), offsetof(struct space_seg, ss_pp_node));

	for (ss = avl_first(&sm->sm_root); ss; ss = AVL_NEXT(&sm->sm_root, ss))
		avl_add(sm->sm_pp_root, ss);
}

static void
metaslab_pp_unload(space_map_t *sm)
{
	void *cookie = NULL;

	kmem_free(sm->sm_ppd, 64 * sizeof (uint64_t));
	sm->sm_ppd = NULL;

	while (avl_destroy_nodes(sm->sm_pp_root, &cookie) != NULL) {
		/* tear down the tree */
	}

	avl_destroy(sm->sm_pp_root);
	kmem_free(sm->sm_pp_root, sizeof (avl_tree_t));
	sm->sm_pp_root = NULL;
}

/* ARGSUSED */
static void
metaslab_pp_claim(space_map_t *sm, uint64_t start, uint64_t size)
{
	/* No need to update cursor */
}

/* ARGSUSED */
static void
metaslab_pp_free(space_map_t *sm, uint64_t start, uint64_t size)
{
	/* No need to update cursor */
}

/*
 * Return the maximum contiguous segment within the metaslab.
 */
uint64_t
metaslab_pp_maxsize(space_map_t *sm)
{
	avl_tree_t *t = sm->sm_pp_root;
	space_seg_t *ss;

	if (t == NULL || (ss = avl_last(t)) == NULL)
		return (0ULL);

	return (ss->ss_end - ss->ss_start);
}

/*
 * ==========================================================================
 * The first-fit block allocator
 * ==========================================================================
 */
static uint64_t
metaslab_ff_alloc(space_map_t *sm, uint64_t size)
{
	avl_tree_t *t = &sm->sm_root;
	uint64_t align = size & -size;
	uint64_t *cursor = (uint64_t *)sm->sm_ppd + highbit(align) - 1;

	return (metaslab_block_picker(t, cursor, size, align));
}

/* ARGSUSED */
boolean_t
metaslab_ff_fragmented(space_map_t *sm)
{
	return (B_TRUE);
}

static space_map_ops_t metaslab_ff_ops = {
	metaslab_pp_load,
	metaslab_pp_unload,
	metaslab_ff_alloc,
	metaslab_pp_claim,
	metaslab_pp_free,
	metaslab_pp_maxsize,
	metaslab_ff_fragmented
};

/*
 * ==========================================================================
 * Dynamic block allocator -
 * Uses the first fit allocation scheme until space get low and then
 * adjusts to a best fit allocation method. Uses metaslab_df_alloc_threshold
 * and metaslab_df_free_pct to determine when to switch the allocation scheme.
 * ==========================================================================
 */
static uint64_t
metaslab_df_alloc(space_map_t *sm, uint64_t size)
{
	avl_tree_t *t = &sm->sm_root;
	uint64_t align = size & -size;
	uint64_t *cursor = (uint64_t *)sm->sm_ppd + highbit(align) - 1;
	uint64_t max_size = metaslab_pp_maxsize(sm);
	int free_pct = sm->sm_space * 100 / sm->sm_size;

	ASSERT(MUTEX_HELD(sm->sm_lock));
	ASSERT3U(avl_numnodes(&sm->sm_root), ==, avl_numnodes(sm->sm_pp_root));

	if (max_size < size)
		return (-1ULL);

	/*
	 * If we're running low on space switch to using the size
	 * sorted AVL tree (best-fit).
	 */
	if (max_size < metaslab_df_alloc_threshold ||
	    free_pct < metaslab_df_free_pct) {
		t = sm->sm_pp_root;
		*cursor = 0;
	}

	return (metaslab_block_picker(t, cursor, size, 1ULL));
}

static boolean_t
metaslab_df_fragmented(space_map_t *sm)
{
	uint64_t max_size = metaslab_pp_maxsize(sm);
	int free_pct = sm->sm_space * 100 / sm->sm_size;

	if (max_size >= metaslab_df_alloc_threshold &&
	    free_pct >= metaslab_df_free_pct)
		return (B_FALSE);

	return (B_TRUE);
}

static space_map_ops_t metaslab_df_ops = {
	metaslab_pp_load,
	metaslab_pp_unload,
	metaslab_df_alloc,
	metaslab_pp_claim,
	metaslab_pp_free,
	metaslab_pp_maxsize,
	metaslab_df_fragmented
};

/*
 * ==========================================================================
 * Other experimental allocators
 * ==========================================================================
 */
static uint64_t
metaslab_cdf_alloc(space_map_t *sm, uint64_t size)
{
	avl_tree_t *t = &sm->sm_root;
	uint64_t *cursor = (uint64_t *)sm->sm_ppd;
	uint64_t *extent_end = (uint64_t *)sm->sm_ppd + 1;
	uint64_t max_size = metaslab_pp_maxsize(sm);
	uint64_t rsize = size;
	uint64_t offset = 0;

	ASSERT(MUTEX_HELD(sm->sm_lock));
	ASSERT3U(avl_numnodes(&sm->sm_root), ==, avl_numnodes(sm->sm_pp_root));

	if (max_size < size)
		return (-1ULL);

	ASSERT3U(*extent_end, >=, *cursor);

	/*
	 * If we're running low on space switch to using the size
	 * sorted AVL tree (best-fit).
	 */
	if ((*cursor + size) > *extent_end) {

		t = sm->sm_pp_root;
		*cursor = *extent_end = 0;

		if (max_size > 2 * SPA_MAXBLOCKSIZE)
			rsize = MIN(metaslab_min_alloc_size, max_size);
		offset = metaslab_block_picker(t, extent_end, rsize, 1ULL);
		if (offset != -1)
			*cursor = offset + size;
	} else {
		offset = metaslab_block_picker(t, cursor, rsize, 1ULL);
	}
	ASSERT3U(*cursor, <=, *extent_end);
	return (offset);
}

static boolean_t
metaslab_cdf_fragmented(space_map_t *sm)
{
	uint64_t max_size = metaslab_pp_maxsize(sm);

	if (max_size > (metaslab_min_alloc_size * 10))
		return (B_FALSE);
	return (B_TRUE);
}

static space_map_ops_t metaslab_cdf_ops = {
	metaslab_pp_load,
	metaslab_pp_unload,
	metaslab_cdf_alloc,
	metaslab_pp_claim,
	metaslab_pp_free,
	metaslab_pp_maxsize,
	metaslab_cdf_fragmented
};

uint64_t metaslab_ndf_clump_shift = 4;

static uint64_t
metaslab_ndf_alloc(space_map_t *sm, uint64_t size)
{
	avl_tree_t *t = &sm->sm_root;
	avl_index_t where;
	space_seg_t *ss, ssearch;
	uint64_t hbit = highbit(size);
	uint64_t *cursor = (uint64_t *)sm->sm_ppd + hbit - 1;
	uint64_t max_size = metaslab_pp_maxsize(sm);

	ASSERT(MUTEX_HELD(sm->sm_lock));
	ASSERT3U(avl_numnodes(&sm->sm_root), ==, avl_numnodes(sm->sm_pp_root));

	if (max_size < size)
		return (-1ULL);

	ssearch.ss_start = *cursor;
	ssearch.ss_end = *cursor + size;

	ss = avl_find(t, &ssearch, &where);
	if (ss == NULL || (ss->ss_start + size > ss->ss_end)) {
		t = sm->sm_pp_root;

		ssearch.ss_start = 0;
		ssearch.ss_end = MIN(max_size,
		    1ULL << (hbit + metaslab_ndf_clump_shift));
		ss = avl_find(t, &ssearch, &where);
		if (ss == NULL)
			ss = avl_nearest(t, where, AVL_AFTER);
		ASSERT(ss != NULL);
	}

	if (ss != NULL) {
		if (ss->ss_start + size <= ss->ss_end) {
			*cursor = ss->ss_start + size;
			return (ss->ss_start);
		}
	}
	return (-1ULL);
}

static boolean_t
metaslab_ndf_fragmented(space_map_t *sm)
{
	uint64_t max_size = metaslab_pp_maxsize(sm);

	if (max_size > (metaslab_min_alloc_size << metaslab_ndf_clump_shift))
		return (B_FALSE);
	return (B_TRUE);
}


static space_map_ops_t metaslab_ndf_ops = {
	metaslab_pp_load,
	metaslab_pp_unload,
	metaslab_ndf_alloc,
	metaslab_pp_claim,
	metaslab_pp_free,
	metaslab_pp_maxsize,
	metaslab_ndf_fragmented
};

space_map_ops_t *zfs_metaslab_ops = &metaslab_ndf_ops;

/*
 * ==========================================================================
 * Metaslabs
 * ==========================================================================
 */
metaslab_t *
metaslab_init(metaslab_group_t *mg, space_map_obj_t *smo,
	uint64_t start, uint64_t size, uint64_t txg)
{
	vdev_t *vd = mg->mg_vd;
	metaslab_t *msp;

	msp = kmem_zalloc(sizeof (metaslab_t), KM_SLEEP);
	mutex_init(&msp->ms_lock, NULL, MUTEX_DEFAULT, NULL);

	msp->ms_smo_syncing = *smo;

	/*
	 * We create the main space map here, but we don't create the
	 * allocmaps and freemaps until metaslab_sync_done().  This serves
	 * two purposes: it allows metaslab_sync_done() to detect the
	 * addition of new space; and for debugging, it ensures that we'd
	 * data fault on any attempt to use this metaslab before it's ready.
	 */
	space_map_create(&msp->ms_map, start, size,
	    vd->vdev_ashift, &msp->ms_lock);

	metaslab_group_add(mg, msp);

	/*
	 * If we're opening an existing pool (txg == 0) or creating
	 * a new one (txg == TXG_INITIAL), all space is available now.
	 * If we're adding space to an existing pool, the new space
	 * does not become available until after this txg has synced.
	 */
	if (txg <= TXG_INITIAL)
		metaslab_sync_done(msp, 0);

	if (txg != 0) {
		/*
		 * The vdev is dirty, but the metaslab isn't -- it just needs
		 * to have metaslab_sync_done() invoked from vdev_sync_done().
		 * [We could just dirty the metaslab, but that would cause us
		 * to allocate a space map object for it, which is wasteful
		 * and would mess up the locality logic in metaslab_weight().]
		 */
		ASSERT(TXG_CLEAN(txg) == spa_last_synced_txg(vd->vdev_spa));
		vdev_dirty(vd, 0, NULL, txg);
		vdev_dirty(vd, VDD_METASLAB, msp, TXG_CLEAN(txg));
	}

	return (msp);
}

void
metaslab_fini(metaslab_t *msp)
{
	metaslab_group_t *mg = msp->ms_group;
	int t;

	vdev_space_update(mg->mg_vd, -msp->ms_map.sm_size,
	    -msp->ms_smo.smo_alloc, B_TRUE);

	metaslab_group_remove(mg, msp);

	mutex_enter(&msp->ms_lock);

	space_map_unload(&msp->ms_map);
	space_map_destroy(&msp->ms_map);

	for (t = 0; t < TXG_SIZE; t++) {
		space_map_destroy(&msp->ms_allocmap[t]);
		space_map_destroy(&msp->ms_freemap[t]);
	}

	mutex_exit(&msp->ms_lock);
	mutex_destroy(&msp->ms_lock);

	kmem_free(msp, sizeof (metaslab_t));
}

#define	METASLAB_WEIGHT_PRIMARY		(1ULL << 63)
#define	METASLAB_WEIGHT_SECONDARY	(1ULL << 62)
#define	METASLAB_ACTIVE_MASK		\
	(METASLAB_WEIGHT_PRIMARY | METASLAB_WEIGHT_SECONDARY)

static uint64_t
metaslab_weight(metaslab_t *msp)
{
	metaslab_group_t *mg = msp->ms_group;
	space_map_t *sm = &msp->ms_map;
	space_map_obj_t *smo = &msp->ms_smo;
	vdev_t *vd = mg->mg_vd;
	uint64_t weight, space;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	 * The baseline weight is the metaslab's free space.
	 */
	space = sm->sm_size - smo->smo_alloc;
	weight = space;

	/*
	 * Modern disks have uniform bit density and constant angular velocity.
	 * Therefore, the outer recording zones are faster (higher bandwidth)
	 * than the inner zones by the ratio of outer to inner track diameter,
	 * which is typically around 2:1.  We account for this by assigning
	 * higher weight to lower metaslabs (multiplier ranging from 2x to 1x).
	 * In effect, this means that we'll select the metaslab with the most
	 * free bandwidth rather than simply the one with the most free space.
	 */
	weight = 2 * weight -
	    ((sm->sm_start >> vd->vdev_ms_shift) * weight) / vd->vdev_ms_count;
	ASSERT(weight >= space && weight <= 2 * space);

	/*
	 * For locality, assign higher weight to metaslabs which have
	 * a lower offset than what we've already activated.
	 */
	if (sm->sm_start <= mg->mg_bonus_area)
		weight *= (metaslab_smo_bonus_pct / 100);
	ASSERT(weight >= space &&
	    weight <= 2 * (metaslab_smo_bonus_pct / 100) * space);

	if (sm->sm_loaded && !sm->sm_ops->smop_fragmented(sm)) {
		/*
		 * If this metaslab is one we're actively using, adjust its
		 * weight to make it preferable to any inactive metaslab so
		 * we'll polish it off.
		 */
		weight |= (msp->ms_weight & METASLAB_ACTIVE_MASK);
	}
	return (weight);
}

static void
metaslab_prefetch(metaslab_group_t *mg)
{
	spa_t *spa = mg->mg_vd->vdev_spa;
	metaslab_t *msp;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	int m;

	mutex_enter(&mg->mg_lock);

	/*
	 * Prefetch the next potential metaslabs
	 */
	for (msp = avl_first(t), m = 0; msp; msp = AVL_NEXT(t, msp), m++) {
		space_map_t *sm = &msp->ms_map;
		space_map_obj_t *smo = &msp->ms_smo;

		/* If we have reached our prefetch limit then we're done */
		if (m >= metaslab_prefetch_limit)
			break;

		if (!sm->sm_loaded && smo->smo_object != 0) {
			mutex_exit(&mg->mg_lock);
			dmu_prefetch(spa->spa_meta_objset, smo->smo_object,
			    0ULL, smo->smo_objsize);
			mutex_enter(&mg->mg_lock);
		}
	}
	mutex_exit(&mg->mg_lock);
}

static int
metaslab_activate(metaslab_t *msp, uint64_t activation_weight, uint64_t size)
{
	metaslab_group_t *mg = msp->ms_group;
	space_map_t *sm = &msp->ms_map;
	space_map_ops_t *sm_ops = msp->ms_group->mg_class->mc_ops;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0) {
		int error = space_map_load(sm, sm_ops, SM_FREE, &msp->ms_smo,
		    msp->ms_group->mg_vd->vdev_spa->spa_meta_objset);
		if (error) {
			metaslab_group_sort(msp->ms_group, msp, 0);
			return (error);
		}

		/*
		 * Track the bonus area as we activate new metaslabs.
		 */
		if (sm->sm_start > mg->mg_bonus_area) {
			mutex_enter(&mg->mg_lock);
			mg->mg_bonus_area = sm->sm_start;
			mutex_exit(&mg->mg_lock);
		}

		/*
		 * If we were able to load the map then make sure
		 * that this map is still able to satisfy our request.
		 */
		if (msp->ms_weight < size)
			return (ENOSPC);

		metaslab_group_sort(msp->ms_group, msp,
		    msp->ms_weight | activation_weight);
	}
	ASSERT(sm->sm_loaded);
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);

	return (0);
}

static void
metaslab_passivate(metaslab_t *msp, uint64_t size)
{
	/*
	 * If size < SPA_MINBLOCKSIZE, then we will not allocate from
	 * this metaslab again.  In that case, it had better be empty,
	 * or we would be leaving space on the table.
	 */
	ASSERT(size >= SPA_MINBLOCKSIZE || msp->ms_map.sm_space == 0);
	metaslab_group_sort(msp->ms_group, msp, MIN(msp->ms_weight, size));
	ASSERT((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0);
}

/*
 * Write a metaslab to disk in the context of the specified transaction group.
 */
void
metaslab_sync(metaslab_t *msp, uint64_t txg)
{
	vdev_t *vd = msp->ms_group->mg_vd;
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	space_map_t *allocmap = &msp->ms_allocmap[txg & TXG_MASK];
	space_map_t *freemap = &msp->ms_freemap[txg & TXG_MASK];
	space_map_t *freed_map = &msp->ms_freemap[TXG_CLEAN(txg) & TXG_MASK];
	space_map_t *sm = &msp->ms_map;
	space_map_obj_t *smo = &msp->ms_smo_syncing;
	dmu_buf_t *db;
	dmu_tx_t *tx;
	int t;

	tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);

	/*
	 * The only state that can actually be changing concurrently with
	 * metaslab_sync() is the metaslab's ms_map.  No other thread can
	 * be modifying this txg's allocmap, freemap, freed_map, or smo.
	 * Therefore, we only hold ms_lock to satify space_map ASSERTs.
	 * We drop it whenever we call into the DMU, because the DMU
	 * can call down to us (e.g. via zio_free()) at any time.
	 */
	mutex_enter(&msp->ms_lock);

	if (smo->smo_object == 0) {
		ASSERT(smo->smo_objsize == 0);
		ASSERT(smo->smo_alloc == 0);
		mutex_exit(&msp->ms_lock);
		smo->smo_object = dmu_object_alloc(mos,
		    DMU_OT_SPACE_MAP, 1 << SPACE_MAP_BLOCKSHIFT,
		    DMU_OT_SPACE_MAP_HEADER, sizeof (*smo), tx);
		ASSERT(smo->smo_object != 0);
		dmu_write(mos, vd->vdev_ms_array, sizeof (uint64_t) *
		    (sm->sm_start >> vd->vdev_ms_shift),
		    sizeof (uint64_t), &smo->smo_object, tx);
		mutex_enter(&msp->ms_lock);
	}

	space_map_walk(freemap, space_map_add, freed_map);

	if (sm->sm_loaded && spa_sync_pass(spa) == 1 && smo->smo_objsize >=
	    2 * sizeof (uint64_t) * avl_numnodes(&sm->sm_root)) {
		/*
		 * The in-core space map representation is twice as compact
		 * as the on-disk one, so it's time to condense the latter
		 * by generating a pure allocmap from first principles.
		 *
		 * This metaslab is 100% allocated,
		 * minus the content of the in-core map (sm),
		 * minus what's been freed this txg (freed_map),
		 * minus allocations from txgs in the future
		 * (because they haven't been committed yet).
		 */
		space_map_vacate(allocmap, NULL, NULL);
		space_map_vacate(freemap, NULL, NULL);

		space_map_add(allocmap, allocmap->sm_start, allocmap->sm_size);

		space_map_walk(sm, space_map_remove, allocmap);
		space_map_walk(freed_map, space_map_remove, allocmap);

		for (t = 1; t < TXG_CONCURRENT_STATES; t++)
			space_map_walk(&msp->ms_allocmap[(txg + t) & TXG_MASK],
			    space_map_remove, allocmap);

		mutex_exit(&msp->ms_lock);
		space_map_truncate(smo, mos, tx);
		mutex_enter(&msp->ms_lock);
	}

	space_map_sync(allocmap, SM_ALLOC, smo, mos, tx);
	space_map_sync(freemap, SM_FREE, smo, mos, tx);

	mutex_exit(&msp->ms_lock);

	VERIFY(0 == dmu_bonus_hold(mos, smo->smo_object, FTAG, &db));
	dmu_buf_will_dirty(db, tx);
	ASSERT3U(db->db_size, >=, sizeof (*smo));
	bcopy(smo, db->db_data, sizeof (*smo));
	dmu_buf_rele(db, FTAG);

	dmu_tx_commit(tx);
}

/*
 * Called after a transaction group has completely synced to mark
 * all of the metaslab's free space as usable.
 */
void
metaslab_sync_done(metaslab_t *msp, uint64_t txg)
{
	space_map_obj_t *smo = &msp->ms_smo;
	space_map_obj_t *smosync = &msp->ms_smo_syncing;
	space_map_t *sm = &msp->ms_map;
	space_map_t *freed_map = &msp->ms_freemap[TXG_CLEAN(txg) & TXG_MASK];
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	int t;

	mutex_enter(&msp->ms_lock);

	/*
	 * If this metaslab is just becoming available, initialize its
	 * allocmaps and freemaps and add its capacity to the vdev.
	 */
	if (freed_map->sm_size == 0) {
		for (t = 0; t < TXG_SIZE; t++) {
			space_map_create(&msp->ms_allocmap[t], sm->sm_start,
			    sm->sm_size, sm->sm_shift, sm->sm_lock);
			space_map_create(&msp->ms_freemap[t], sm->sm_start,
			    sm->sm_size, sm->sm_shift, sm->sm_lock);
		}
		vdev_space_update(vd, sm->sm_size, 0, B_TRUE);
	}

	vdev_space_update(vd, 0, smosync->smo_alloc - smo->smo_alloc, B_TRUE);

	ASSERT(msp->ms_allocmap[txg & TXG_MASK].sm_space == 0);
	ASSERT(msp->ms_freemap[txg & TXG_MASK].sm_space == 0);

	/*
	 * If there's a space_map_load() in progress, wait for it to complete
	 * so that we have a consistent view of the in-core space map.
	 * Then, add everything we freed in this txg to the map.
	 */
	space_map_load_wait(sm);
	space_map_vacate(freed_map, sm->sm_loaded ? space_map_free : NULL, sm);

	*smo = *smosync;

	/*
	 * If the map is loaded but no longer active, evict it as soon as all
	 * future allocations have synced.  (If we unloaded it now and then
	 * loaded a moment later, the map wouldn't reflect those allocations.)
	 */
	if (sm->sm_loaded && (msp->ms_weight & METASLAB_ACTIVE_MASK) == 0) {
		int evictable = 1;

		for (t = 1; t < TXG_CONCURRENT_STATES; t++)
			if (msp->ms_allocmap[(txg + t) & TXG_MASK].sm_space)
				evictable = 0;

		if (evictable)
			space_map_unload(sm);
	}

	metaslab_group_sort(mg, msp, metaslab_weight(msp));

	mutex_exit(&msp->ms_lock);
}

void
metaslab_sync_reassess(metaslab_group_t *mg)
{
	vdev_t *vd = mg->mg_vd;

	/*
	 * Re-evaluate all metaslabs which have lower offsets than the
	 * bonus area.
	 */
	for (int m = 0; m < vd->vdev_ms_count; m++) {
		metaslab_t *msp = vd->vdev_ms[m];

		if (msp->ms_map.sm_start > mg->mg_bonus_area)
			break;

		mutex_enter(&msp->ms_lock);
		metaslab_group_sort(mg, msp, metaslab_weight(msp));
		mutex_exit(&msp->ms_lock);
	}

	/*
	 * Prefetch the next potential metaslabs
	 */
	metaslab_prefetch(mg);
}

static uint64_t
metaslab_distance(metaslab_t *msp, dva_t *dva)
{
	uint64_t ms_shift = msp->ms_group->mg_vd->vdev_ms_shift;
	uint64_t offset = DVA_GET_OFFSET(dva) >> ms_shift;
	uint64_t start = msp->ms_map.sm_start >> ms_shift;

	if (msp->ms_group->mg_vd->vdev_id != DVA_GET_VDEV(dva))
		return (1ULL << 63);

	if (offset < start)
		return ((start - offset) << ms_shift);
	if (offset > start)
		return ((offset - start) << ms_shift);
	return (0);
}

static uint64_t
metaslab_group_alloc(metaslab_group_t *mg, uint64_t size, uint64_t txg,
    uint64_t min_distance, dva_t *dva, int d)
{
	metaslab_t *msp = NULL;
	uint64_t offset = -1ULL;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	uint64_t activation_weight;
	uint64_t target_distance;
	int i;

	activation_weight = METASLAB_WEIGHT_PRIMARY;
	for (i = 0; i < d; i++) {
		if (DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
			activation_weight = METASLAB_WEIGHT_SECONDARY;
			break;
		}
	}

	for (;;) {
		boolean_t was_active;

		mutex_enter(&mg->mg_lock);
		for (msp = avl_first(t); msp; msp = AVL_NEXT(t, msp)) {
			if (msp->ms_weight < size) {
				mutex_exit(&mg->mg_lock);
				return (-1ULL);
			}

			was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
			if (activation_weight == METASLAB_WEIGHT_PRIMARY)
				break;

			target_distance = min_distance +
			    (msp->ms_smo.smo_alloc ? 0 : min_distance >> 1);

			for (i = 0; i < d; i++)
				if (metaslab_distance(msp, &dva[i]) <
				    target_distance)
					break;
			if (i == d)
				break;
		}
		mutex_exit(&mg->mg_lock);
		if (msp == NULL)
			return (-1ULL);

		mutex_enter(&msp->ms_lock);

		/*
		 * Ensure that the metaslab we have selected is still
		 * capable of handling our request. It's possible that
		 * another thread may have changed the weight while we
		 * were blocked on the metaslab lock.
		 */
		if (msp->ms_weight < size || (was_active &&
		    !(msp->ms_weight & METASLAB_ACTIVE_MASK) &&
		    activation_weight == METASLAB_WEIGHT_PRIMARY)) {
			mutex_exit(&msp->ms_lock);
			continue;
		}

		if ((msp->ms_weight & METASLAB_WEIGHT_SECONDARY) &&
		    activation_weight == METASLAB_WEIGHT_PRIMARY) {
			metaslab_passivate(msp,
			    msp->ms_weight & ~METASLAB_ACTIVE_MASK);
			mutex_exit(&msp->ms_lock);
			continue;
		}

		if (metaslab_activate(msp, activation_weight, size) != 0) {
			mutex_exit(&msp->ms_lock);
			continue;
		}

		if ((offset = space_map_alloc(&msp->ms_map, size)) != -1ULL)
			break;

		metaslab_passivate(msp, space_map_maxsize(&msp->ms_map));

		mutex_exit(&msp->ms_lock);
	}

	if (msp->ms_allocmap[txg & TXG_MASK].sm_space == 0)
		vdev_dirty(mg->mg_vd, VDD_METASLAB, msp, txg);

	space_map_add(&msp->ms_allocmap[txg & TXG_MASK], offset, size);

	mutex_exit(&msp->ms_lock);

	return (offset);
}

/*
 * Allocate a block for the specified i/o.
 */
static int
metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
    dva_t *dva, int d, dva_t *hintdva, uint64_t txg, int flags)
{
	metaslab_group_t *mg, *rotor;
	vdev_t *vd;
	int dshift = 3;
	int all_zero;
	int zio_lock = B_FALSE;
	boolean_t allocatable;
	uint64_t offset = -1ULL;
	uint64_t asize;
	uint64_t distance;

	ASSERT(!DVA_IS_VALID(&dva[d]));

	/*
	 * For testing, make some blocks above a certain size be gang blocks.
	 */
	if (psize >= metaslab_gang_bang && (LBOLT & 3) == 0)
		return (ENOSPC);

	/*
	 * Start at the rotor and loop through all mgs until we find something.
	 * Note that there's no locking on mc_rotor or mc_allocated because
	 * nothing actually breaks if we miss a few updates -- we just won't
	 * allocate quite as evenly.  It all balances out over time.
	 *
	 * If we are doing ditto or log blocks, try to spread them across
	 * consecutive vdevs.  If we're forced to reuse a vdev before we've
	 * allocated all of our ditto blocks, then try and spread them out on
	 * that vdev as much as possible.  If it turns out to not be possible,
	 * gradually lower our standards until anything becomes acceptable.
	 * Also, allocating on consecutive vdevs (as opposed to random vdevs)
	 * gives us hope of containing our fault domains to something we're
	 * able to reason about.  Otherwise, any two top-level vdev failures
	 * will guarantee the loss of data.  With consecutive allocation,
	 * only two adjacent top-level vdev failures will result in data loss.
	 *
	 * If we are doing gang blocks (hintdva is non-NULL), try to keep
	 * ourselves on the same vdev as our gang block header.  That
	 * way, we can hope for locality in vdev_cache, plus it makes our
	 * fault domains something tractable.
	 */
	if (hintdva) {
		vd = vdev_lookup_top(spa, DVA_GET_VDEV(&hintdva[d]));
		if (flags & METASLAB_HINTBP_AVOID)
			mg = vd->vdev_mg->mg_next;
		else
			mg = vd->vdev_mg;
	} else if (d != 0) {
		vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d - 1]));
		mg = vd->vdev_mg->mg_next;
	} else {
		mg = mc->mc_rotor;
	}

	/*
	 * If the hint put us into the wrong class, just follow the rotor.
	 */
	if (mg->mg_class != mc)
		mg = mc->mc_rotor;

	rotor = mg;
top:
	all_zero = B_TRUE;
	do {
		vd = mg->mg_vd;

		/*
		 * Don't allocate from faulted devices.
		 */
		if (zio_lock) {
			spa_config_enter(spa, SCL_ZIO, FTAG, RW_READER);
			allocatable = vdev_allocatable(vd);
			spa_config_exit(spa, SCL_ZIO, FTAG);
		} else {
			allocatable = vdev_allocatable(vd);
		}
		if (!allocatable)
			goto next;

		/*
		 * Avoid writing single-copy data to a failing vdev
		 */
		if ((vd->vdev_stat.vs_write_errors > 0 ||
		    vd->vdev_state < VDEV_STATE_HEALTHY) &&
		    d == 0 && dshift == 3) {
			all_zero = B_FALSE;
			goto next;
		}

		ASSERT(mg->mg_class == mc);

		distance = vd->vdev_asize >> dshift;
		if (distance <= (1ULL << vd->vdev_ms_shift))
			distance = 0;
		else
			all_zero = B_FALSE;

		asize = vdev_psize_to_asize(vd, psize);
		ASSERT(P2PHASE(asize, 1ULL << vd->vdev_ashift) == 0);

		offset = metaslab_group_alloc(mg, asize, txg, distance, dva, d);
		if (offset != -1ULL) {
			/*
			 * If we've just selected this metaslab group,
			 * figure out whether the corresponding vdev is
			 * over- or under-used relative to the pool,
			 * and set an allocation bias to even it out.
			 */
			if (mc->mc_allocated == 0) {
				vdev_stat_t *vs = &vd->vdev_stat;
				uint64_t alloc, space;
				int64_t vu, su;

				alloc = spa_get_alloc(spa);
				space = spa_get_space(spa);

				/*
				 * Determine percent used in units of 0..1024.
				 * (This is just to avoid floating point.)
				 */
				vu = (vs->vs_alloc << 10) / (vs->vs_space + 1);
				su = (alloc << 10) / (space + 1);

				/*
				 * Bias by at most +/- 25% of the aliquot.
				 */
				mg->mg_bias = ((su - vu) *
				    (int64_t)mg->mg_aliquot) / (1024 * 4);
			}

			if (atomic_add_64_nv(&mc->mc_allocated, asize) >=
			    mg->mg_aliquot + mg->mg_bias) {
				mc->mc_rotor = mg->mg_next;
				mc->mc_allocated = 0;
			}

			DVA_SET_VDEV(&dva[d], vd->vdev_id);
			DVA_SET_OFFSET(&dva[d], offset);
			DVA_SET_GANG(&dva[d], !!(flags & METASLAB_GANG_HEADER));
			DVA_SET_ASIZE(&dva[d], asize);

			return (0);
		}
next:
		mc->mc_rotor = mg->mg_next;
		mc->mc_allocated = 0;
	} while ((mg = mg->mg_next) != rotor);

	if (!all_zero) {
		dshift++;
		ASSERT(dshift < 64);
		goto top;
	}

	if (!allocatable && !zio_lock) {
		dshift = 3;
		zio_lock = B_TRUE;
		goto top;
	}

	bzero(&dva[d], sizeof (dva_t));

	return (ENOSPC);
}

/*
 * Free the block represented by DVA in the context of the specified
 * transaction group.
 */
static void
metaslab_free_dva(spa_t *spa, const dva_t *dva, uint64_t txg, boolean_t now)
{
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd;
	metaslab_t *msp;

	ASSERT(DVA_IS_VALID(dva));

	if (txg > spa_freeze_txg(spa))
		return;

	if ((vd = vdev_lookup_top(spa, vdev)) == NULL ||
	    (offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count) {
		cmn_err(CE_WARN, "metaslab_free_dva(): bad DVA %llu:%llu",
		    (u_longlong_t)vdev, (u_longlong_t)offset);
		ASSERT(0);
		return;
	}

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	if (DVA_GET_GANG(dva))
		size = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);

	mutex_enter(&msp->ms_lock);

	if (now) {
		space_map_remove(&msp->ms_allocmap[txg & TXG_MASK],
		    offset, size);
		space_map_free(&msp->ms_map, offset, size);
	} else {
		if (msp->ms_freemap[txg & TXG_MASK].sm_space == 0)
			vdev_dirty(vd, VDD_METASLAB, msp, txg);
		space_map_add(&msp->ms_freemap[txg & TXG_MASK], offset, size);
	}

	mutex_exit(&msp->ms_lock);
}

/*
 * Intent log support: upon opening the pool after a crash, notify the SPA
 * of blocks that the intent log has allocated for immediate write, but
 * which are still considered free by the SPA because the last transaction
 * group didn't commit yet.
 */
static int
metaslab_claim_dva(spa_t *spa, const dva_t *dva, uint64_t txg)
{
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd;
	metaslab_t *msp;
	int error;

	ASSERT(DVA_IS_VALID(dva));

	if ((vd = vdev_lookup_top(spa, vdev)) == NULL ||
	    (offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count)
		return (ENXIO);

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	if (DVA_GET_GANG(dva))
		size = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);

	mutex_enter(&msp->ms_lock);

	error = metaslab_activate(msp, METASLAB_WEIGHT_SECONDARY, 0);
	if (error || txg == 0) {	/* txg == 0 indicates dry run */
		mutex_exit(&msp->ms_lock);
		return (error);
	}

	space_map_claim(&msp->ms_map, offset, size);

	if (spa_writeable(spa)) {	/* don't dirty if we're zdb(1M) */
		if (msp->ms_allocmap[txg & TXG_MASK].sm_space == 0)
			vdev_dirty(vd, VDD_METASLAB, msp, txg);
		space_map_add(&msp->ms_allocmap[txg & TXG_MASK], offset, size);
	}

	mutex_exit(&msp->ms_lock);

	return (0);
}

int
metaslab_alloc(spa_t *spa, metaslab_class_t *mc, uint64_t psize, blkptr_t *bp,
    int ndvas, uint64_t txg, blkptr_t *hintbp, int flags)
{
	dva_t *dva = bp->blk_dva;
	dva_t *hintdva = hintbp->blk_dva;
	int error = 0;

	ASSERT(bp->blk_birth == 0);

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);

	if (mc->mc_rotor == NULL) {	/* no vdevs in this class */
		spa_config_exit(spa, SCL_ALLOC, FTAG);
		return (ENOSPC);
	}

	ASSERT(ndvas > 0 && ndvas <= spa_max_replication(spa));
	ASSERT(BP_GET_NDVAS(bp) == 0);
	ASSERT(hintbp == NULL || ndvas <= BP_GET_NDVAS(hintbp));

	for (int d = 0; d < ndvas; d++) {
		error = metaslab_alloc_dva(spa, mc, psize, dva, d, hintdva,
		    txg, flags);
		if (error) {
			for (d--; d >= 0; d--) {
				metaslab_free_dva(spa, &dva[d], txg, B_TRUE);
				bzero(&dva[d], sizeof (dva_t));
			}
			spa_config_exit(spa, SCL_ALLOC, FTAG);
			return (error);
		}
	}
	ASSERT(error == 0);
	ASSERT(BP_GET_NDVAS(bp) == ndvas);

	spa_config_exit(spa, SCL_ALLOC, FTAG);

	bp->blk_birth = txg;

	return (0);
}

void
metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg, boolean_t now)
{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!now || bp->blk_birth >= spa->spa_syncing_txg);

	spa_config_enter(spa, SCL_FREE, FTAG, RW_READER);

	for (int d = 0; d < ndvas; d++)
		metaslab_free_dva(spa, &dva[d], txg, now);

	spa_config_exit(spa, SCL_FREE, FTAG);
}

int
metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg)
{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	int error = 0;

	ASSERT(!BP_IS_HOLE(bp));

	if (txg != 0) {
		/*
		 * First do a dry run to make sure all DVAs are claimable,
		 * so we don't have to unwind from partial failures below.
		 */
		if ((error = metaslab_claim(spa, bp, 0)) != 0)
			return (error);
	}

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);

	for (int d = 0; d < ndvas; d++)
		if ((error = metaslab_claim_dva(spa, &dva[d], txg)) != 0)
			break;

	spa_config_exit(spa, SCL_ALLOC, FTAG);

	ASSERT(error == 0 || txg == 0);

	return (error);
}