summaryrefslogtreecommitdiffstats
path: root/module/zfs/abd.c
blob: 2d1be9752d4f48d7341d5f4b45f6eb96d685320c (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
/*
 * 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) 2014 by Chunwei Chen. All rights reserved.
 * Copyright (c) 2019 by Delphix. All rights reserved.
 */

/*
 * ARC buffer data (ABD).
 *
 * ABDs are an abstract data structure for the ARC which can use two
 * different ways of storing the underlying data:
 *
 * (a) Linear buffer. In this case, all the data in the ABD is stored in one
 *     contiguous buffer in memory (from a zio_[data_]buf_* kmem cache).
 *
 *         +-------------------+
 *         | ABD (linear)      |
 *         |   abd_flags = ... |
 *         |   abd_size = ...  |     +--------------------------------+
 *         |   abd_buf ------------->| raw buffer of size abd_size    |
 *         +-------------------+     +--------------------------------+
 *              no abd_chunks
 *
 * (b) Scattered buffer. In this case, the data in the ABD is split into
 *     equal-sized chunks (from the abd_chunk_cache kmem_cache), with pointers
 *     to the chunks recorded in an array at the end of the ABD structure.
 *
 *         +-------------------+
 *         | ABD (scattered)   |
 *         |   abd_flags = ... |
 *         |   abd_size = ...  |
 *         |   abd_offset = 0  |                           +-----------+
 *         |   abd_chunks[0] ----------------------------->| chunk 0   |
 *         |   abd_chunks[1] ---------------------+        +-----------+
 *         |   ...             |                  |        +-----------+
 *         |   abd_chunks[N-1] ---------+         +------->| chunk 1   |
 *         +-------------------+        |                  +-----------+
 *                                      |                      ...
 *                                      |                  +-----------+
 *                                      +----------------->| chunk N-1 |
 *                                                         +-----------+
 *
 * In addition to directly allocating a linear or scattered ABD, it is also
 * possible to create an ABD by requesting the "sub-ABD" starting at an offset
 * within an existing ABD. In linear buffers this is simple (set abd_buf of
 * the new ABD to the starting point within the original raw buffer), but
 * scattered ABDs are a little more complex. The new ABD makes a copy of the
 * relevant abd_chunks pointers (but not the underlying data). However, to
 * provide arbitrary rather than only chunk-aligned starting offsets, it also
 * tracks an abd_offset field which represents the starting point of the data
 * within the first chunk in abd_chunks. For both linear and scattered ABDs,
 * creating an offset ABD marks the original ABD as the offset's parent, and the
 * original ABD's abd_children refcount is incremented. This data allows us to
 * ensure the root ABD isn't deleted before its children.
 *
 * Most consumers should never need to know what type of ABD they're using --
 * the ABD public API ensures that it's possible to transparently switch from
 * using a linear ABD to a scattered one when doing so would be beneficial.
 *
 * If you need to use the data within an ABD directly, if you know it's linear
 * (because you allocated it) you can use abd_to_buf() to access the underlying
 * raw buffer. Otherwise, you should use one of the abd_borrow_buf* functions
 * which will allocate a raw buffer if necessary. Use the abd_return_buf*
 * functions to return any raw buffers that are no longer necessary when you're
 * done using them.
 *
 * There are a variety of ABD APIs that implement basic buffer operations:
 * compare, copy, read, write, and fill with zeroes. If you need a custom
 * function which progressively accesses the whole ABD, use the abd_iterate_*
 * functions.
 *
 * As an additional feature, linear and scatter ABD's can be stitched together
 * by using the gang ABD type (abd_alloc_gang_abd()). This allows for
 * multiple ABDs to be viewed as a singular ABD.
 *
 * It is possible to make all ABDs linear by setting zfs_abd_scatter_enabled to
 * B_FALSE.
 */

#include <sys/abd_impl.h>
#include <sys/param.h>
#include <sys/zio.h>
#include <sys/zfs_context.h>
#include <sys/zfs_znode.h>

/* see block comment above for description */
int zfs_abd_scatter_enabled = B_TRUE;

void
abd_verify(abd_t *abd)
{
	ASSERT3U(abd->abd_size, >, 0);
	ASSERT3U(abd->abd_size, <=, SPA_MAXBLOCKSIZE);
	ASSERT3U(abd->abd_flags, ==, abd->abd_flags & (ABD_FLAG_LINEAR |
	    ABD_FLAG_OWNER | ABD_FLAG_META | ABD_FLAG_MULTI_ZONE |
	    ABD_FLAG_MULTI_CHUNK | ABD_FLAG_LINEAR_PAGE | ABD_FLAG_GANG |
	    ABD_FLAG_GANG_FREE | ABD_FLAG_ZEROS | ABD_FLAG_ALLOCD));
#ifdef ZFS_DEBUG
	IMPLY(abd->abd_parent != NULL, !(abd->abd_flags & ABD_FLAG_OWNER));
#endif
	IMPLY(abd->abd_flags & ABD_FLAG_META, abd->abd_flags & ABD_FLAG_OWNER);
	if (abd_is_linear(abd)) {
		ASSERT3P(ABD_LINEAR_BUF(abd), !=, NULL);
	} else if (abd_is_gang(abd)) {
		uint_t child_sizes = 0;
		for (abd_t *cabd = list_head(&ABD_GANG(abd).abd_gang_chain);
		    cabd != NULL;
		    cabd = list_next(&ABD_GANG(abd).abd_gang_chain, cabd)) {
			ASSERT(list_link_active(&cabd->abd_gang_link));
			child_sizes += cabd->abd_size;
			abd_verify(cabd);
		}
		ASSERT3U(abd->abd_size, ==, child_sizes);
	} else {
		abd_verify_scatter(abd);
	}
}

static void
abd_init_struct(abd_t *abd)
{
	list_link_init(&abd->abd_gang_link);
	mutex_init(&abd->abd_mtx, NULL, MUTEX_DEFAULT, NULL);
	abd->abd_flags = 0;
#ifdef ZFS_DEBUG
	zfs_refcount_create(&abd->abd_children);
	abd->abd_parent = NULL;
#endif
	abd->abd_size = 0;
}

static void
abd_fini_struct(abd_t *abd)
{
	mutex_destroy(&abd->abd_mtx);
	ASSERT(!list_link_active(&abd->abd_gang_link));
#ifdef ZFS_DEBUG
	zfs_refcount_destroy(&abd->abd_children);
#endif
}

abd_t *
abd_alloc_struct(size_t size)
{
	abd_t *abd = abd_alloc_struct_impl(size);
	abd_init_struct(abd);
	abd->abd_flags |= ABD_FLAG_ALLOCD;
	return (abd);
}

void
abd_free_struct(abd_t *abd)
{
	abd_fini_struct(abd);
	abd_free_struct_impl(abd);
}

/*
 * Allocate an ABD, along with its own underlying data buffers. Use this if you
 * don't care whether the ABD is linear or not.
 */
abd_t *
abd_alloc(size_t size, boolean_t is_metadata)
{
	if (!zfs_abd_scatter_enabled || abd_size_alloc_linear(size))
		return (abd_alloc_linear(size, is_metadata));

	VERIFY3U(size, <=, SPA_MAXBLOCKSIZE);

	abd_t *abd = abd_alloc_struct(size);
	abd->abd_flags |= ABD_FLAG_OWNER;
	abd->abd_u.abd_scatter.abd_offset = 0;
	abd_alloc_chunks(abd, size);

	if (is_metadata) {
		abd->abd_flags |= ABD_FLAG_META;
	}
	abd->abd_size = size;

	abd_update_scatter_stats(abd, ABDSTAT_INCR);

	return (abd);
}

/*
 * Allocate an ABD that must be linear, along with its own underlying data
 * buffer. Only use this when it would be very annoying to write your ABD
 * consumer with a scattered ABD.
 */
abd_t *
abd_alloc_linear(size_t size, boolean_t is_metadata)
{
	abd_t *abd = abd_alloc_struct(0);

	VERIFY3U(size, <=, SPA_MAXBLOCKSIZE);

	abd->abd_flags |= ABD_FLAG_LINEAR | ABD_FLAG_OWNER;
	if (is_metadata) {
		abd->abd_flags |= ABD_FLAG_META;
	}
	abd->abd_size = size;

	if (is_metadata) {
		ABD_LINEAR_BUF(abd) = zio_buf_alloc(size);
	} else {
		ABD_LINEAR_BUF(abd) = zio_data_buf_alloc(size);
	}

	abd_update_linear_stats(abd, ABDSTAT_INCR);

	return (abd);
}

static void
abd_free_linear(abd_t *abd)
{
	if (abd_is_linear_page(abd)) {
		abd_free_linear_page(abd);
		return;
	}
	if (abd->abd_flags & ABD_FLAG_META) {
		zio_buf_free(ABD_LINEAR_BUF(abd), abd->abd_size);
	} else {
		zio_data_buf_free(ABD_LINEAR_BUF(abd), abd->abd_size);
	}

	abd_update_linear_stats(abd, ABDSTAT_DECR);
}

static void
abd_free_gang(abd_t *abd)
{
	ASSERT(abd_is_gang(abd));
	abd_t *cabd;

	while ((cabd = list_head(&ABD_GANG(abd).abd_gang_chain)) != NULL) {
		/*
		 * We must acquire the child ABDs mutex to ensure that if it
		 * is being added to another gang ABD we will set the link
		 * as inactive when removing it from this gang ABD and before
		 * adding it to the other gang ABD.
		 */
		mutex_enter(&cabd->abd_mtx);
		ASSERT(list_link_active(&cabd->abd_gang_link));
		list_remove(&ABD_GANG(abd).abd_gang_chain, cabd);
		mutex_exit(&cabd->abd_mtx);
		if (cabd->abd_flags & ABD_FLAG_GANG_FREE)
			abd_free(cabd);
	}
	list_destroy(&ABD_GANG(abd).abd_gang_chain);
}

static void
abd_free_scatter(abd_t *abd)
{
	abd_free_chunks(abd);
	abd_update_scatter_stats(abd, ABDSTAT_DECR);
}

/*
 * Free an ABD.  Use with any kind of abd: those created with abd_alloc_*()
 * and abd_get_*(), including abd_get_offset_struct().
 *
 * If the ABD was created with abd_alloc_*(), the underlying data
 * (scatterlist or linear buffer) will also be freed.  (Subject to ownership
 * changes via abd_*_ownership_of_buf().)
 *
 * Unless the ABD was created with abd_get_offset_struct(), the abd_t will
 * also be freed.
 */
void
abd_free(abd_t *abd)
{
	if (abd == NULL)
		return;

	abd_verify(abd);
#ifdef ZFS_DEBUG
	IMPLY(abd->abd_flags & ABD_FLAG_OWNER, abd->abd_parent == NULL);
#endif

	if (abd_is_gang(abd)) {
		abd_free_gang(abd);
	} else if (abd_is_linear(abd)) {
		if (abd->abd_flags & ABD_FLAG_OWNER)
			abd_free_linear(abd);
	} else {
		if (abd->abd_flags & ABD_FLAG_OWNER)
			abd_free_scatter(abd);
	}

#ifdef ZFS_DEBUG
	if (abd->abd_parent != NULL) {
		(void) zfs_refcount_remove_many(&abd->abd_parent->abd_children,
		    abd->abd_size, abd);
	}
#endif

	abd_fini_struct(abd);
	if (abd->abd_flags & ABD_FLAG_ALLOCD)
		abd_free_struct_impl(abd);
}

/*
 * Allocate an ABD of the same format (same metadata flag, same scatterize
 * setting) as another ABD.
 */
abd_t *
abd_alloc_sametype(abd_t *sabd, size_t size)
{
	boolean_t is_metadata = (sabd->abd_flags & ABD_FLAG_META) != 0;
	if (abd_is_linear(sabd) &&
	    !abd_is_linear_page(sabd)) {
		return (abd_alloc_linear(size, is_metadata));
	} else {
		return (abd_alloc(size, is_metadata));
	}
}

/*
 * Create gang ABD that will be the head of a list of ABD's. This is used
 * to "chain" scatter/gather lists together when constructing aggregated
 * IO's. To free this abd, abd_free() must be called.
 */
abd_t *
abd_alloc_gang(void)
{
	abd_t *abd = abd_alloc_struct(0);
	abd->abd_flags |= ABD_FLAG_GANG | ABD_FLAG_OWNER;
	list_create(&ABD_GANG(abd).abd_gang_chain,
	    sizeof (abd_t), offsetof(abd_t, abd_gang_link));
	return (abd);
}

/*
 * Add a child gang ABD to a parent gang ABDs chained list.
 */
static void
abd_gang_add_gang(abd_t *pabd, abd_t *cabd, boolean_t free_on_free)
{
	ASSERT(abd_is_gang(pabd));
	ASSERT(abd_is_gang(cabd));

	if (free_on_free) {
		/*
		 * If the parent is responsible for freeing the child gang
		 * ABD we will just splice the child's children ABD list to
		 * the parent's list and immediately free the child gang ABD
		 * struct. The parent gang ABDs children from the child gang
		 * will retain all the free_on_free settings after being
		 * added to the parents list.
		 */
		pabd->abd_size += cabd->abd_size;
		list_move_tail(&ABD_GANG(pabd).abd_gang_chain,
		    &ABD_GANG(cabd).abd_gang_chain);
		ASSERT(list_is_empty(&ABD_GANG(cabd).abd_gang_chain));
		abd_verify(pabd);
		abd_free(cabd);
	} else {
		for (abd_t *child = list_head(&ABD_GANG(cabd).abd_gang_chain);
		    child != NULL;
		    child = list_next(&ABD_GANG(cabd).abd_gang_chain, child)) {
			/*
			 * We always pass B_FALSE for free_on_free as it is the
			 * original child gang ABDs responsibility to determine
			 * if any of its child ABDs should be free'd on the call
			 * to abd_free().
			 */
			abd_gang_add(pabd, child, B_FALSE);
		}
		abd_verify(pabd);
	}
}

/*
 * Add a child ABD to a gang ABD's chained list.
 */
void
abd_gang_add(abd_t *pabd, abd_t *cabd, boolean_t free_on_free)
{
	ASSERT(abd_is_gang(pabd));
	abd_t *child_abd = NULL;

	/*
	 * If the child being added is a gang ABD, we will add the
	 * child's ABDs to the parent gang ABD. This allows us to account
	 * for the offset correctly in the parent gang ABD.
	 */
	if (abd_is_gang(cabd)) {
		ASSERT(!list_link_active(&cabd->abd_gang_link));
		ASSERT(!list_is_empty(&ABD_GANG(cabd).abd_gang_chain));
		return (abd_gang_add_gang(pabd, cabd, free_on_free));
	}
	ASSERT(!abd_is_gang(cabd));

	/*
	 * In order to verify that an ABD is not already part of
	 * another gang ABD, we must lock the child ABD's abd_mtx
	 * to check its abd_gang_link status. We unlock the abd_mtx
	 * only after it is has been added to a gang ABD, which
	 * will update the abd_gang_link's status. See comment below
	 * for how an ABD can be in multiple gang ABD's simultaneously.
	 */
	mutex_enter(&cabd->abd_mtx);
	if (list_link_active(&cabd->abd_gang_link)) {
		/*
		 * If the child ABD is already part of another
		 * gang ABD then we must allocate a new
		 * ABD to use a separate link. We mark the newly
		 * allocated ABD with ABD_FLAG_GANG_FREE, before
		 * adding it to the gang ABD's list, to make the
		 * gang ABD aware that it is responsible to call
		 * abd_free(). We use abd_get_offset() in order
		 * to just allocate a new ABD but avoid copying the
		 * data over into the newly allocated ABD.
		 *
		 * An ABD may become part of multiple gang ABD's. For
		 * example, when writing ditto bocks, the same ABD
		 * is used to write 2 or 3 locations with 2 or 3
		 * zio_t's. Each of the zio's may be aggregated with
		 * different adjacent zio's. zio aggregation uses gang
		 * zio's, so the single ABD can become part of multiple
		 * gang zio's.
		 *
		 * The ASSERT below is to make sure that if
		 * free_on_free is passed as B_TRUE, the ABD can
		 * not be in multiple gang ABD's. The gang ABD
		 * can not be responsible for cleaning up the child
		 * ABD memory allocation if the ABD can be in
		 * multiple gang ABD's at one time.
		 */
		ASSERT3B(free_on_free, ==, B_FALSE);
		child_abd = abd_get_offset(cabd, 0);
		child_abd->abd_flags |= ABD_FLAG_GANG_FREE;
	} else {
		child_abd = cabd;
		if (free_on_free)
			child_abd->abd_flags |= ABD_FLAG_GANG_FREE;
	}
	ASSERT3P(child_abd, !=, NULL);

	list_insert_tail(&ABD_GANG(pabd).abd_gang_chain, child_abd);
	mutex_exit(&cabd->abd_mtx);
	pabd->abd_size += child_abd->abd_size;
}

/*
 * Locate the ABD for the supplied offset in the gang ABD.
 * Return a new offset relative to the returned ABD.
 */
abd_t *
abd_gang_get_offset(abd_t *abd, size_t *off)
{
	abd_t *cabd;

	ASSERT(abd_is_gang(abd));
	ASSERT3U(*off, <, abd->abd_size);
	for (cabd = list_head(&ABD_GANG(abd).abd_gang_chain); cabd != NULL;
	    cabd = list_next(&ABD_GANG(abd).abd_gang_chain, cabd)) {
		if (*off >= cabd->abd_size)
			*off -= cabd->abd_size;
		else
			return (cabd);
	}
	VERIFY3P(cabd, !=, NULL);
	return (cabd);
}

/*
 * Allocate a new ABD, using the provided struct (if non-NULL, and if
 * circumstances allow - otherwise allocate the struct).  The returned ABD will
 * point to offset off of sabd. It shares the underlying buffer data with sabd.
 * Use abd_free() to free.  sabd must not be freed while any derived ABDs exist.
 */
static abd_t *
abd_get_offset_impl(abd_t *abd, abd_t *sabd, size_t off, size_t size)
{
	abd_verify(sabd);
	ASSERT3U(off + size, <=, sabd->abd_size);

	if (abd_is_linear(sabd)) {
		if (abd == NULL)
			abd = abd_alloc_struct(0);
		/*
		 * Even if this buf is filesystem metadata, we only track that
		 * if we own the underlying data buffer, which is not true in
		 * this case. Therefore, we don't ever use ABD_FLAG_META here.
		 */
		abd->abd_flags |= ABD_FLAG_LINEAR;

		ABD_LINEAR_BUF(abd) = (char *)ABD_LINEAR_BUF(sabd) + off;
	} else if (abd_is_gang(sabd)) {
		size_t left = size;
		if (abd == NULL) {
			abd = abd_alloc_gang();
		} else {
			abd->abd_flags |= ABD_FLAG_GANG;
			list_create(&ABD_GANG(abd).abd_gang_chain,
			    sizeof (abd_t), offsetof(abd_t, abd_gang_link));
		}

		abd->abd_flags &= ~ABD_FLAG_OWNER;
		for (abd_t *cabd = abd_gang_get_offset(sabd, &off);
		    cabd != NULL && left > 0;
		    cabd = list_next(&ABD_GANG(sabd).abd_gang_chain, cabd)) {
			int csize = MIN(left, cabd->abd_size - off);

			abd_t *nabd = abd_get_offset_size(cabd, off, csize);
			abd_gang_add(abd, nabd, B_TRUE);
			left -= csize;
			off = 0;
		}
		ASSERT3U(left, ==, 0);
	} else {
		abd = abd_get_offset_scatter(abd, sabd, off);
	}

	ASSERT3P(abd, !=, NULL);
	abd->abd_size = size;
#ifdef ZFS_DEBUG
	abd->abd_parent = sabd;
	(void) zfs_refcount_add_many(&sabd->abd_children, abd->abd_size, abd);
#endif
	return (abd);
}

/*
 * Like abd_get_offset_size(), but memory for the abd_t is provided by the
 * caller.  Using this routine can improve performance by avoiding the cost
 * of allocating memory for the abd_t struct, and updating the abd stats.
 * Usually, the provided abd is returned, but in some circumstances (FreeBSD,
 * if sabd is scatter and size is more than 2 pages) a new abd_t may need to
 * be allocated.  Therefore callers should be careful to use the returned
 * abd_t*.
 */
abd_t *
abd_get_offset_struct(abd_t *abd, abd_t *sabd, size_t off, size_t size)
{
	abd_t *result;
	abd_init_struct(abd);
	result = abd_get_offset_impl(abd, sabd, off, size);
	if (result != abd)
		abd_fini_struct(abd);
	return (result);
}

abd_t *
abd_get_offset(abd_t *sabd, size_t off)
{
	size_t size = sabd->abd_size > off ? sabd->abd_size - off : 0;
	VERIFY3U(size, >, 0);
	return (abd_get_offset_impl(NULL, sabd, off, size));
}

abd_t *
abd_get_offset_size(abd_t *sabd, size_t off, size_t size)
{
	ASSERT3U(off + size, <=, sabd->abd_size);
	return (abd_get_offset_impl(NULL, sabd, off, size));
}

/*
 * Return a size scatter ABD containing only zeros.
 */
abd_t *
abd_get_zeros(size_t size)
{
	ASSERT3P(abd_zero_scatter, !=, NULL);
	ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
	return (abd_get_offset_size(abd_zero_scatter, 0, size));
}

/*
 * Allocate a linear ABD structure for buf.
 */
abd_t *
abd_get_from_buf(void *buf, size_t size)
{
	abd_t *abd = abd_alloc_struct(0);

	VERIFY3U(size, <=, SPA_MAXBLOCKSIZE);

	/*
	 * Even if this buf is filesystem metadata, we only track that if we
	 * own the underlying data buffer, which is not true in this case.
	 * Therefore, we don't ever use ABD_FLAG_META here.
	 */
	abd->abd_flags |= ABD_FLAG_LINEAR;
	abd->abd_size = size;

	ABD_LINEAR_BUF(abd) = buf;

	return (abd);
}

/*
 * Get the raw buffer associated with a linear ABD.
 */
void *
abd_to_buf(abd_t *abd)
{
	ASSERT(abd_is_linear(abd));
	abd_verify(abd);
	return (ABD_LINEAR_BUF(abd));
}

/*
 * Borrow a raw buffer from an ABD without copying the contents of the ABD
 * into the buffer. If the ABD is scattered, this will allocate a raw buffer
 * whose contents are undefined. To copy over the existing data in the ABD, use
 * abd_borrow_buf_copy() instead.
 */
void *
abd_borrow_buf(abd_t *abd, size_t n)
{
	void *buf;
	abd_verify(abd);
	ASSERT3U(abd->abd_size, >=, n);
	if (abd_is_linear(abd)) {
		buf = abd_to_buf(abd);
	} else {
		buf = zio_buf_alloc(n);
	}
#ifdef ZFS_DEBUG
	(void) zfs_refcount_add_many(&abd->abd_children, n, buf);
#endif
	return (buf);
}

void *
abd_borrow_buf_copy(abd_t *abd, size_t n)
{
	void *buf = abd_borrow_buf(abd, n);
	if (!abd_is_linear(abd)) {
		abd_copy_to_buf(buf, abd, n);
	}
	return (buf);
}

/*
 * Return a borrowed raw buffer to an ABD. If the ABD is scattered, this will
 * not change the contents of the ABD and will ASSERT that you didn't modify
 * the buffer since it was borrowed. If you want any changes you made to buf to
 * be copied back to abd, use abd_return_buf_copy() instead.
 */
void
abd_return_buf(abd_t *abd, void *buf, size_t n)
{
	abd_verify(abd);
	ASSERT3U(abd->abd_size, >=, n);
	if (abd_is_linear(abd)) {
		ASSERT3P(buf, ==, abd_to_buf(abd));
	} else {
		ASSERT0(abd_cmp_buf(abd, buf, n));
		zio_buf_free(buf, n);
	}
#ifdef ZFS_DEBUG
	(void) zfs_refcount_remove_many(&abd->abd_children, n, buf);
#endif
}

void
abd_return_buf_copy(abd_t *abd, void *buf, size_t n)
{
	if (!abd_is_linear(abd)) {
		abd_copy_from_buf(abd, buf, n);
	}
	abd_return_buf(abd, buf, n);
}

void
abd_release_ownership_of_buf(abd_t *abd)
{
	ASSERT(abd_is_linear(abd));
	ASSERT(abd->abd_flags & ABD_FLAG_OWNER);

	/*
	 * abd_free() needs to handle LINEAR_PAGE ABD's specially.
	 * Since that flag does not survive the
	 * abd_release_ownership_of_buf() -> abd_get_from_buf() ->
	 * abd_take_ownership_of_buf() sequence, we don't allow releasing
	 * these "linear but not zio_[data_]buf_alloc()'ed" ABD's.
	 */
	ASSERT(!abd_is_linear_page(abd));

	abd_verify(abd);

	abd->abd_flags &= ~ABD_FLAG_OWNER;
	/* Disable this flag since we no longer own the data buffer */
	abd->abd_flags &= ~ABD_FLAG_META;

	abd_update_linear_stats(abd, ABDSTAT_DECR);
}


/*
 * Give this ABD ownership of the buffer that it's storing. Can only be used on
 * linear ABDs which were allocated via abd_get_from_buf(), or ones allocated
 * with abd_alloc_linear() which subsequently released ownership of their buf
 * with abd_release_ownership_of_buf().
 */
void
abd_take_ownership_of_buf(abd_t *abd, boolean_t is_metadata)
{
	ASSERT(abd_is_linear(abd));
	ASSERT(!(abd->abd_flags & ABD_FLAG_OWNER));
	abd_verify(abd);

	abd->abd_flags |= ABD_FLAG_OWNER;
	if (is_metadata) {
		abd->abd_flags |= ABD_FLAG_META;
	}

	abd_update_linear_stats(abd, ABDSTAT_INCR);
}

/*
 * Initializes an abd_iter based on whether the abd is a gang ABD
 * or just a single ABD.
 */
static inline abd_t *
abd_init_abd_iter(abd_t *abd, struct abd_iter *aiter, size_t off)
{
	abd_t *cabd = NULL;

	if (abd_is_gang(abd)) {
		cabd = abd_gang_get_offset(abd, &off);
		if (cabd) {
			abd_iter_init(aiter, cabd);
			abd_iter_advance(aiter, off);
		}
	} else {
		abd_iter_init(aiter, abd);
		abd_iter_advance(aiter, off);
	}
	return (cabd);
}

/*
 * Advances an abd_iter. We have to be careful with gang ABD as
 * advancing could mean that we are at the end of a particular ABD and
 * must grab the ABD in the gang ABD's list.
 */
static inline abd_t *
abd_advance_abd_iter(abd_t *abd, abd_t *cabd, struct abd_iter *aiter,
    size_t len)
{
	abd_iter_advance(aiter, len);
	if (abd_is_gang(abd) && abd_iter_at_end(aiter)) {
		ASSERT3P(cabd, !=, NULL);
		cabd = list_next(&ABD_GANG(abd).abd_gang_chain, cabd);
		if (cabd) {
			abd_iter_init(aiter, cabd);
			abd_iter_advance(aiter, 0);
		}
	}
	return (cabd);
}

int
abd_iterate_func(abd_t *abd, size_t off, size_t size,
    abd_iter_func_t *func, void *private)
{
	struct abd_iter aiter;
	int ret = 0;

	if (size == 0)
		return (0);

	abd_verify(abd);
	ASSERT3U(off + size, <=, abd->abd_size);

	boolean_t gang = abd_is_gang(abd);
	abd_t *c_abd = abd_init_abd_iter(abd, &aiter, off);

	while (size > 0) {
		/* If we are at the end of the gang ABD we are done */
		if (gang && !c_abd)
			break;

		abd_iter_map(&aiter);

		size_t len = MIN(aiter.iter_mapsize, size);
		ASSERT3U(len, >, 0);

		ret = func(aiter.iter_mapaddr, len, private);

		abd_iter_unmap(&aiter);

		if (ret != 0)
			break;

		size -= len;
		c_abd = abd_advance_abd_iter(abd, c_abd, &aiter, len);
	}

	return (ret);
}

struct buf_arg {
	void *arg_buf;
};

static int
abd_copy_to_buf_off_cb(void *buf, size_t size, void *private)
{
	struct buf_arg *ba_ptr = private;

	(void) memcpy(ba_ptr->arg_buf, buf, size);
	ba_ptr->arg_buf = (char *)ba_ptr->arg_buf + size;

	return (0);
}

/*
 * Copy abd to buf. (off is the offset in abd.)
 */
void
abd_copy_to_buf_off(void *buf, abd_t *abd, size_t off, size_t size)
{
	struct buf_arg ba_ptr = { buf };

	(void) abd_iterate_func(abd, off, size, abd_copy_to_buf_off_cb,
	    &ba_ptr);
}

static int
abd_cmp_buf_off_cb(void *buf, size_t size, void *private)
{
	int ret;
	struct buf_arg *ba_ptr = private;

	ret = memcmp(buf, ba_ptr->arg_buf, size);
	ba_ptr->arg_buf = (char *)ba_ptr->arg_buf + size;

	return (ret);
}

/*
 * Compare the contents of abd to buf. (off is the offset in abd.)
 */
int
abd_cmp_buf_off(abd_t *abd, const void *buf, size_t off, size_t size)
{
	struct buf_arg ba_ptr = { (void *) buf };

	return (abd_iterate_func(abd, off, size, abd_cmp_buf_off_cb, &ba_ptr));
}

static int
abd_copy_from_buf_off_cb(void *buf, size_t size, void *private)
{
	struct buf_arg *ba_ptr = private;

	(void) memcpy(buf, ba_ptr->arg_buf, size);
	ba_ptr->arg_buf = (char *)ba_ptr->arg_buf + size;

	return (0);
}

/*
 * Copy from buf to abd. (off is the offset in abd.)
 */
void
abd_copy_from_buf_off(abd_t *abd, const void *buf, size_t off, size_t size)
{
	struct buf_arg ba_ptr = { (void *) buf };

	(void) abd_iterate_func(abd, off, size, abd_copy_from_buf_off_cb,
	    &ba_ptr);
}

/*ARGSUSED*/
static int
abd_zero_off_cb(void *buf, size_t size, void *private)
{
	(void) memset(buf, 0, size);
	return (0);
}

/*
 * Zero out the abd from a particular offset to the end.
 */
void
abd_zero_off(abd_t *abd, size_t off, size_t size)
{
	(void) abd_iterate_func(abd, off, size, abd_zero_off_cb, NULL);
}

/*
 * Iterate over two ABDs and call func incrementally on the two ABDs' data in
 * equal-sized chunks (passed to func as raw buffers). func could be called many
 * times during this iteration.
 */
int
abd_iterate_func2(abd_t *dabd, abd_t *sabd, size_t doff, size_t soff,
    size_t size, abd_iter_func2_t *func, void *private)
{
	int ret = 0;
	struct abd_iter daiter, saiter;
	boolean_t dabd_is_gang_abd, sabd_is_gang_abd;
	abd_t *c_dabd, *c_sabd;

	if (size == 0)
		return (0);

	abd_verify(dabd);
	abd_verify(sabd);

	ASSERT3U(doff + size, <=, dabd->abd_size);
	ASSERT3U(soff + size, <=, sabd->abd_size);

	dabd_is_gang_abd = abd_is_gang(dabd);
	sabd_is_gang_abd = abd_is_gang(sabd);
	c_dabd = abd_init_abd_iter(dabd, &daiter, doff);
	c_sabd = abd_init_abd_iter(sabd, &saiter, soff);

	while (size > 0) {
		/* if we are at the end of the gang ABD we are done */
		if ((dabd_is_gang_abd && !c_dabd) ||
		    (sabd_is_gang_abd && !c_sabd))
			break;

		abd_iter_map(&daiter);
		abd_iter_map(&saiter);

		size_t dlen = MIN(daiter.iter_mapsize, size);
		size_t slen = MIN(saiter.iter_mapsize, size);
		size_t len = MIN(dlen, slen);
		ASSERT(dlen > 0 || slen > 0);

		ret = func(daiter.iter_mapaddr, saiter.iter_mapaddr, len,
		    private);

		abd_iter_unmap(&saiter);
		abd_iter_unmap(&daiter);

		if (ret != 0)
			break;

		size -= len;
		c_dabd =
		    abd_advance_abd_iter(dabd, c_dabd, &daiter, len);
		c_sabd =
		    abd_advance_abd_iter(sabd, c_sabd, &saiter, len);
	}

	return (ret);
}

/*ARGSUSED*/
static int
abd_copy_off_cb(void *dbuf, void *sbuf, size_t size, void *private)
{
	(void) memcpy(dbuf, sbuf, size);
	return (0);
}

/*
 * Copy from sabd to dabd starting from soff and doff.
 */
void
abd_copy_off(abd_t *dabd, abd_t *sabd, size_t doff, size_t soff, size_t size)
{
	(void) abd_iterate_func2(dabd, sabd, doff, soff, size,
	    abd_copy_off_cb, NULL);
}

/*ARGSUSED*/
static int
abd_cmp_cb(void *bufa, void *bufb, size_t size, void *private)
{
	return (memcmp(bufa, bufb, size));
}

/*
 * Compares the contents of two ABDs.
 */
int
abd_cmp(abd_t *dabd, abd_t *sabd)
{
	ASSERT3U(dabd->abd_size, ==, sabd->abd_size);
	return (abd_iterate_func2(dabd, sabd, 0, 0, dabd->abd_size,
	    abd_cmp_cb, NULL));
}

/*
 * Iterate over code ABDs and a data ABD and call @func_raidz_gen.
 *
 * @cabds          parity ABDs, must have equal size
 * @dabd           data ABD. Can be NULL (in this case @dsize = 0)
 * @func_raidz_gen should be implemented so that its behaviour
 *                 is the same when taking linear and when taking scatter
 */
void
abd_raidz_gen_iterate(abd_t **cabds, abd_t *dabd,
    ssize_t csize, ssize_t dsize, const unsigned parity,
    void (*func_raidz_gen)(void **, const void *, size_t, size_t))
{
	int i;
	ssize_t len, dlen;
	struct abd_iter caiters[3];
	struct abd_iter daiter = {0};
	void *caddrs[3];
	unsigned long flags __maybe_unused = 0;
	abd_t *c_cabds[3];
	abd_t *c_dabd = NULL;
	boolean_t cabds_is_gang_abd[3];
	boolean_t dabd_is_gang_abd = B_FALSE;

	ASSERT3U(parity, <=, 3);

	for (i = 0; i < parity; i++) {
		cabds_is_gang_abd[i] = abd_is_gang(cabds[i]);
		c_cabds[i] = abd_init_abd_iter(cabds[i], &caiters[i], 0);
	}

	if (dabd) {
		dabd_is_gang_abd = abd_is_gang(dabd);
		c_dabd = abd_init_abd_iter(dabd, &daiter, 0);
	}

	ASSERT3S(dsize, >=, 0);

	abd_enter_critical(flags);
	while (csize > 0) {
		/* if we are at the end of the gang ABD we are done */
		if (dabd_is_gang_abd && !c_dabd)
			break;

		for (i = 0; i < parity; i++) {
			/*
			 * If we are at the end of the gang ABD we are
			 * done.
			 */
			if (cabds_is_gang_abd[i] && !c_cabds[i])
				break;
			abd_iter_map(&caiters[i]);
			caddrs[i] = caiters[i].iter_mapaddr;
		}

		len = csize;

		if (dabd && dsize > 0)
			abd_iter_map(&daiter);

		switch (parity) {
			case 3:
				len = MIN(caiters[2].iter_mapsize, len);
				/* falls through */
			case 2:
				len = MIN(caiters[1].iter_mapsize, len);
				/* falls through */
			case 1:
				len = MIN(caiters[0].iter_mapsize, len);
		}

		/* must be progressive */
		ASSERT3S(len, >, 0);

		if (dabd && dsize > 0) {
			/* this needs precise iter.length */
			len = MIN(daiter.iter_mapsize, len);
			dlen = len;
		} else
			dlen = 0;

		/* must be progressive */
		ASSERT3S(len, >, 0);
		/*
		 * The iterated function likely will not do well if each
		 * segment except the last one is not multiple of 512 (raidz).
		 */
		ASSERT3U(((uint64_t)len & 511ULL), ==, 0);

		func_raidz_gen(caddrs, daiter.iter_mapaddr, len, dlen);

		for (i = parity-1; i >= 0; i--) {
			abd_iter_unmap(&caiters[i]);
			c_cabds[i] =
			    abd_advance_abd_iter(cabds[i], c_cabds[i],
			    &caiters[i], len);
		}

		if (dabd && dsize > 0) {
			abd_iter_unmap(&daiter);
			c_dabd =
			    abd_advance_abd_iter(dabd, c_dabd, &daiter,
			    dlen);
			dsize -= dlen;
		}

		csize -= len;

		ASSERT3S(dsize, >=, 0);
		ASSERT3S(csize, >=, 0);
	}
	abd_exit_critical(flags);
}

/*
 * Iterate over code ABDs and data reconstruction target ABDs and call
 * @func_raidz_rec. Function maps at most 6 pages atomically.
 *
 * @cabds           parity ABDs, must have equal size
 * @tabds           rec target ABDs, at most 3
 * @tsize           size of data target columns
 * @func_raidz_rec  expects syndrome data in target columns. Function
 *                  reconstructs data and overwrites target columns.
 */
void
abd_raidz_rec_iterate(abd_t **cabds, abd_t **tabds,
    ssize_t tsize, const unsigned parity,
    void (*func_raidz_rec)(void **t, const size_t tsize, void **c,
    const unsigned *mul),
    const unsigned *mul)
{
	int i;
	ssize_t len;
	struct abd_iter citers[3];
	struct abd_iter xiters[3];
	void *caddrs[3], *xaddrs[3];
	unsigned long flags __maybe_unused = 0;
	boolean_t cabds_is_gang_abd[3];
	boolean_t tabds_is_gang_abd[3];
	abd_t *c_cabds[3];
	abd_t *c_tabds[3];

	ASSERT3U(parity, <=, 3);

	for (i = 0; i < parity; i++) {
		cabds_is_gang_abd[i] = abd_is_gang(cabds[i]);
		tabds_is_gang_abd[i] = abd_is_gang(tabds[i]);
		c_cabds[i] =
		    abd_init_abd_iter(cabds[i], &citers[i], 0);
		c_tabds[i] =
		    abd_init_abd_iter(tabds[i], &xiters[i], 0);
	}

	abd_enter_critical(flags);
	while (tsize > 0) {

		for (i = 0; i < parity; i++) {
			/*
			 * If we are at the end of the gang ABD we
			 * are done.
			 */
			if (cabds_is_gang_abd[i] && !c_cabds[i])
				break;
			if (tabds_is_gang_abd[i] && !c_tabds[i])
				break;
			abd_iter_map(&citers[i]);
			abd_iter_map(&xiters[i]);
			caddrs[i] = citers[i].iter_mapaddr;
			xaddrs[i] = xiters[i].iter_mapaddr;
		}

		len = tsize;
		switch (parity) {
			case 3:
				len = MIN(xiters[2].iter_mapsize, len);
				len = MIN(citers[2].iter_mapsize, len);
				/* falls through */
			case 2:
				len = MIN(xiters[1].iter_mapsize, len);
				len = MIN(citers[1].iter_mapsize, len);
				/* falls through */
			case 1:
				len = MIN(xiters[0].iter_mapsize, len);
				len = MIN(citers[0].iter_mapsize, len);
		}
		/* must be progressive */
		ASSERT3S(len, >, 0);
		/*
		 * The iterated function likely will not do well if each
		 * segment except the last one is not multiple of 512 (raidz).
		 */
		ASSERT3U(((uint64_t)len & 511ULL), ==, 0);

		func_raidz_rec(xaddrs, len, caddrs, mul);

		for (i = parity-1; i >= 0; i--) {
			abd_iter_unmap(&xiters[i]);
			abd_iter_unmap(&citers[i]);
			c_tabds[i] =
			    abd_advance_abd_iter(tabds[i], c_tabds[i],
			    &xiters[i], len);
			c_cabds[i] =
			    abd_advance_abd_iter(cabds[i], c_cabds[i],
			    &citers[i], len);
		}

		tsize -= len;
		ASSERT3S(tsize, >=, 0);
	}
	abd_exit_critical(flags);
}