summaryrefslogtreecommitdiff
path: root/storage/innobase/row/row0umod.c
blob: fae67c95c4306d3243e72503e464f2e981893f04 (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
/*****************************************************************************

Copyright (c) 1997, 2012, Oracle and/or its affiliates. All Rights Reserved.

This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; version 2 of the License.

This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc., 
51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA

*****************************************************************************/

/**************************************************//**
@file row/row0umod.c
Undo modify of a row

Created 2/27/1997 Heikki Tuuri
*******************************************************/

#include "row0umod.h"

#ifdef UNIV_NONINL
#include "row0umod.ic"
#endif

#include "dict0dict.h"
#include "dict0boot.h"
#include "trx0undo.h"
#include "trx0roll.h"
#include "btr0btr.h"
#include "mach0data.h"
#include "row0undo.h"
#include "row0vers.h"
#include "trx0trx.h"
#include "trx0rec.h"
#include "row0row.h"
#include "row0upd.h"
#include "que0que.h"
#include "log0log.h"

/* Considerations on undoing a modify operation.
(1) Undoing a delete marking: all index records should be found. Some of
them may have delete mark already FALSE, if the delete mark operation was
stopped underway, or if the undo operation ended prematurely because of a
system crash.
(2) Undoing an update of a delete unmarked record: the newer version of
an updated secondary index entry should be removed if no prior version
of the clustered index record requires its existence. Otherwise, it should
be delete marked.
(3) Undoing an update of a delete marked record. In this kind of update a
delete marked clustered index record was delete unmarked and possibly also
some of its fields were changed. Now, it is possible that the delete marked
version has become obsolete at the time the undo is started. */

/*************************************************************************
IMPORTANT NOTE: Any operation that generates redo MUST check that there
is enough space in the redo log before for that operation. This is
done by calling log_free_check(). The reason for checking the
availability of the redo log space before the start of the operation is
that we MUST not hold any synchonization objects when performing the
check.
If you make a change in this module make sure that no codepath is
introduced where a call to log_free_check() is bypassed. */

/***********************************************************//**
Undoes a modify in a clustered index record.
@return	DB_SUCCESS, DB_FAIL, or error code: we may run out of file space */
static
ulint
row_undo_mod_clust_low(
/*===================*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr,	/*!< in: query thread */
	mtr_t*		mtr,	/*!< in: mtr; must be committed before
				latching any further pages */
	ulint		mode)	/*!< in: BTR_MODIFY_LEAF or BTR_MODIFY_TREE */
{
	btr_pcur_t*	pcur;
	btr_cur_t*	btr_cur;
	ulint		err;
#ifdef UNIV_DEBUG
	ibool		success;
#endif /* UNIV_DEBUG */

	pcur = &(node->pcur);
	btr_cur = btr_pcur_get_btr_cur(pcur);

#ifdef UNIV_DEBUG
	success =
#endif /* UNIV_DEBUG */
	btr_pcur_restore_position(mode, pcur, mtr);

	ut_ad(success);

	if (mode == BTR_MODIFY_LEAF) {

		err = btr_cur_optimistic_update(BTR_NO_LOCKING_FLAG
						| BTR_NO_UNDO_LOG_FLAG
						| BTR_KEEP_SYS_FLAG,
						btr_cur, node->update,
						node->cmpl_info, thr, mtr);
	} else {
		mem_heap_t*	heap		= NULL;
		big_rec_t*	dummy_big_rec;

		ut_ad(mode == BTR_MODIFY_TREE);

		err = btr_cur_pessimistic_update(
			BTR_NO_LOCKING_FLAG
			| BTR_NO_UNDO_LOG_FLAG
			| BTR_KEEP_SYS_FLAG,
			btr_cur, &heap, &dummy_big_rec, node->update,
			node->cmpl_info, thr, mtr);

		ut_a(!dummy_big_rec);
		if (UNIV_LIKELY_NULL(heap)) {
			mem_heap_free(heap);
		}
	}

	return(err);
}

/***********************************************************//**
Removes a clustered index record after undo if possible.
This is attempted when the record was inserted by updating a
delete-marked record and there no longer exist transactions
that would see the delete-marked record.  In other words, we
roll back the insert by purging the record.
@return	DB_SUCCESS, DB_FAIL, or error code: we may run out of file space */
static
ulint
row_undo_mod_remove_clust_low(
/*==========================*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr,	/*!< in: query thread */
	mtr_t*		mtr,	/*!< in: mtr */
	ulint		mode)	/*!< in: BTR_MODIFY_LEAF or BTR_MODIFY_TREE */
{
	btr_cur_t*	btr_cur;
	ulint		err;

	ut_ad(node->rec_type == TRX_UNDO_UPD_DEL_REC);

	/* Find out if the record has been purged already
	or if we can remove it. */

	if (!btr_pcur_restore_position(mode, &node->pcur, mtr)
	    || row_vers_must_preserve_del_marked(node->new_trx_id, mtr)) {

		return(DB_SUCCESS);
	}

	btr_cur = btr_pcur_get_btr_cur(&node->pcur);

	if (mode == BTR_MODIFY_LEAF) {
		err = btr_cur_optimistic_delete(btr_cur, mtr)
			? DB_SUCCESS
			: DB_FAIL;
	} else {
		ut_ad(mode == BTR_MODIFY_TREE);

		/* This operation is analogous to purge, we can free also
		inherited externally stored fields */

		btr_cur_pessimistic_delete(&err, FALSE, btr_cur,
					   thr_is_recv(thr)
					   ? RB_RECOVERY_PURGE_REC
					   : RB_NONE, mtr);

		/* The delete operation may fail if we have little
		file space left: TODO: easiest to crash the database
		and restart with more file space */
	}

	return(err);
}

/***********************************************************//**
Undoes a modify in a clustered index record. Sets also the node state for the
next round of undo.
@return	DB_SUCCESS or error code: we may run out of file space */
static
ulint
row_undo_mod_clust(
/*===============*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	btr_pcur_t*	pcur;
	mtr_t		mtr;
	ulint		err;

	ut_ad(node && thr);

	log_free_check();

	pcur = &(node->pcur);

	mtr_start(&mtr);

	/* Try optimistic processing of the record, keeping changes within
	the index page */

	err = row_undo_mod_clust_low(node, thr, &mtr, BTR_MODIFY_LEAF);

	if (err != DB_SUCCESS) {
		btr_pcur_commit_specify_mtr(pcur, &mtr);

		/* We may have to modify tree structure: do a pessimistic
		descent down the index tree */

		mtr_start(&mtr);

		err = row_undo_mod_clust_low(node, thr, &mtr, BTR_MODIFY_TREE);
	}

	btr_pcur_commit_specify_mtr(pcur, &mtr);

	if (err == DB_SUCCESS && node->rec_type == TRX_UNDO_UPD_DEL_REC) {

		mtr_start(&mtr);

		err = row_undo_mod_remove_clust_low(node, thr, &mtr,
						    BTR_MODIFY_LEAF);
		if (err != DB_SUCCESS) {
			btr_pcur_commit_specify_mtr(pcur, &mtr);

			/* We may have to modify tree structure: do a
			pessimistic descent down the index tree */

			mtr_start(&mtr);

			err = row_undo_mod_remove_clust_low(node, thr, &mtr,
							    BTR_MODIFY_TREE);
		}

		btr_pcur_commit_specify_mtr(pcur, &mtr);
	}

	node->state = UNDO_NODE_FETCH_NEXT;

	trx_undo_rec_release(node->trx, node->undo_no);

	return(err);
}

/***********************************************************//**
Delete marks or removes a secondary index entry if found.
@return	DB_SUCCESS, DB_FAIL, or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_mod_del_mark_or_remove_sec_low(
/*====================================*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr,	/*!< in: query thread */
	dict_index_t*	index,	/*!< in: index */
	dtuple_t*	entry,	/*!< in: index entry */
	ulint		mode)	/*!< in: latch mode BTR_MODIFY_LEAF or
				BTR_MODIFY_TREE */
{
	btr_pcur_t		pcur;
	btr_cur_t*		btr_cur;
	ibool			success;
	ibool			old_has;
	ulint			err;
	mtr_t			mtr;
	mtr_t			mtr_vers;
	enum row_search_result	search_result;

	log_free_check();
	mtr_start(&mtr);

	btr_cur = btr_pcur_get_btr_cur(&pcur);

	ut_ad(mode == BTR_MODIFY_TREE || mode == BTR_MODIFY_LEAF);

	search_result = row_search_index_entry(index, entry, mode,
					       &pcur, &mtr);

	switch (UNIV_EXPECT(search_result, ROW_FOUND)) {
	case ROW_NOT_FOUND:
		/* In crash recovery, the secondary index record may
		be missing if the UPDATE did not have time to insert
		the secondary index records before the crash.  When we
		are undoing that UPDATE in crash recovery, the record
		may be missing.

		In normal processing, if an update ends in a deadlock
		before it has inserted all updated secondary index
		records, then the undo will not find those records. */

		err = DB_SUCCESS;
		goto func_exit;
	case ROW_FOUND:
		break;
	case ROW_BUFFERED:
	case ROW_NOT_DELETED_REF:
		/* These are invalid outcomes, because the mode passed
		to row_search_index_entry() did not include any of the
		flags BTR_INSERT, BTR_DELETE, or BTR_DELETE_MARK. */
		ut_error;
	}

	/* We should remove the index record if no prior version of the row,
	which cannot be purged yet, requires its existence. If some requires,
	we should delete mark the record. */

	mtr_start(&mtr_vers);

	success = btr_pcur_restore_position(BTR_SEARCH_LEAF, &(node->pcur),
					    &mtr_vers);
	ut_a(success);

	old_has = row_vers_old_has_index_entry(FALSE,
					       btr_pcur_get_rec(&(node->pcur)),
					       &mtr_vers, index, entry);
	if (old_has) {
		err = btr_cur_del_mark_set_sec_rec(BTR_NO_LOCKING_FLAG,
						   btr_cur, TRUE, thr, &mtr);
		ut_ad(err == DB_SUCCESS);
	} else {
		/* Remove the index record */

		if (mode == BTR_MODIFY_LEAF) {
			success = btr_cur_optimistic_delete(btr_cur, &mtr);
			if (success) {
				err = DB_SUCCESS;
			} else {
				err = DB_FAIL;
			}
		} else {
			ut_ad(mode == BTR_MODIFY_TREE);

			/* No need to distinguish RB_RECOVERY_PURGE here,
			because we are deleting a secondary index record:
			the distinction between RB_NORMAL and
			RB_RECOVERY_PURGE only matters when deleting a
			record that contains externally stored
			columns. */
			ut_ad(!dict_index_is_clust(index));
			btr_cur_pessimistic_delete(&err, FALSE, btr_cur,
						   RB_NORMAL, &mtr);

			/* The delete operation may fail if we have little
			file space left: TODO: easiest to crash the database
			and restart with more file space */
		}
	}

	btr_pcur_commit_specify_mtr(&(node->pcur), &mtr_vers);

func_exit:
	btr_pcur_close(&pcur);
	mtr_commit(&mtr);

	return(err);
}

/***********************************************************//**
Delete marks or removes a secondary index entry if found.
NOTE that if we updated the fields of a delete-marked secondary index record
so that alphabetically they stayed the same, e.g., 'abc' -> 'aBc', we cannot
return to the original values because we do not know them. But this should
not cause problems because in row0sel.c, in queries we always retrieve the
clustered index record or an earlier version of it, if the secondary index
record through which we do the search is delete-marked.
@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_mod_del_mark_or_remove_sec(
/*================================*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr,	/*!< in: query thread */
	dict_index_t*	index,	/*!< in: index */
	dtuple_t*	entry)	/*!< in: index entry */
{
	ulint	err;

	err = row_undo_mod_del_mark_or_remove_sec_low(node, thr, index,
						      entry, BTR_MODIFY_LEAF);
	if (err == DB_SUCCESS) {

		return(err);
	}

	err = row_undo_mod_del_mark_or_remove_sec_low(node, thr, index,
						      entry, BTR_MODIFY_TREE);
	return(err);
}

/***********************************************************//**
Delete unmarks a secondary index entry which must be found. It might not be
delete-marked at the moment, but it does not harm to unmark it anyway. We also
need to update the fields of the secondary index record if we updated its
fields but alphabetically they stayed the same, e.g., 'abc' -> 'aBc'.
@return	DB_FAIL or DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_mod_del_unmark_sec_and_undo_update(
/*========================================*/
	ulint		mode,	/*!< in: search mode: BTR_MODIFY_LEAF or
				BTR_MODIFY_TREE */
	que_thr_t*	thr,	/*!< in: query thread */
	dict_index_t*	index,	/*!< in: index */
	const dtuple_t*	entry)	/*!< in: index entry */
{
	mem_heap_t*		heap;
	btr_pcur_t		pcur;
	btr_cur_t*		btr_cur;
	upd_t*			update;
	ulint			err		= DB_SUCCESS;
	big_rec_t*		dummy_big_rec;
	mtr_t			mtr;
	trx_t*			trx		= thr_get_trx(thr);
	enum row_search_result	search_result;

	/* Ignore indexes that are being created. */
	if (UNIV_UNLIKELY(*index->name == TEMP_INDEX_PREFIX)) {

		return(DB_SUCCESS);
	}

	log_free_check();
	mtr_start(&mtr);

	ut_ad(mode == BTR_MODIFY_TREE || mode == BTR_MODIFY_LEAF);

	search_result = row_search_index_entry(index, entry, mode,
					       &pcur, &mtr);

	switch (search_result) {
	case ROW_BUFFERED:
	case ROW_NOT_DELETED_REF:
		/* These are invalid outcomes, because the mode passed
		to row_search_index_entry() did not include any of the
		flags BTR_INSERT, BTR_DELETE, or BTR_DELETE_MARK. */
		ut_error;
	case ROW_NOT_FOUND:
		fputs("InnoDB: error in sec index entry del undo in\n"
		      "InnoDB: ", stderr);
		dict_index_name_print(stderr, trx, index);
		fputs("\n"
		      "InnoDB: tuple ", stderr);
		dtuple_print(stderr, entry);
		fputs("\n"
		      "InnoDB: record ", stderr);
		rec_print(stderr, btr_pcur_get_rec(&pcur), index);
		putc('\n', stderr);
		trx_print(stderr, trx, 0);
		fputs("\n"
		      "InnoDB: Submit a detailed bug report"
		      " to http://bugs.mysql.com\n", stderr);
		ut_ad(0);
		break;
	case ROW_FOUND:
		btr_cur = btr_pcur_get_btr_cur(&pcur);
		err = btr_cur_del_mark_set_sec_rec(BTR_NO_LOCKING_FLAG,
						   btr_cur, FALSE, thr, &mtr);
		ut_a(err == DB_SUCCESS);
		heap = mem_heap_create(100);

		update = row_upd_build_sec_rec_difference_binary(
			index, entry, btr_cur_get_rec(btr_cur), trx, heap);
		if (upd_get_n_fields(update) == 0) {

			/* Do nothing */

		} else if (mode == BTR_MODIFY_LEAF) {
			/* Try an optimistic updating of the record, keeping
			changes within the page */

			err = btr_cur_optimistic_update(
				BTR_KEEP_SYS_FLAG | BTR_NO_LOCKING_FLAG,
				btr_cur, update, 0, thr, &mtr);
			switch (err) {
			case DB_OVERFLOW:
			case DB_UNDERFLOW:
			case DB_ZIP_OVERFLOW:
				err = DB_FAIL;
			}
		} else {
			ut_a(mode == BTR_MODIFY_TREE);
			err = btr_cur_pessimistic_update(
				BTR_KEEP_SYS_FLAG | BTR_NO_LOCKING_FLAG,
				btr_cur, &heap, &dummy_big_rec,
				update, 0, thr, &mtr);
			ut_a(!dummy_big_rec);
		}

		mem_heap_free(heap);
	}

	btr_pcur_close(&pcur);
	mtr_commit(&mtr);

	return(err);
}

/***********************************************************//**
Undoes a modify in secondary indexes when undo record type is UPD_DEL.
@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_mod_upd_del_sec(
/*=====================*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	mem_heap_t*	heap;
	dtuple_t*	entry;
	dict_index_t*	index;
	ulint		err	= DB_SUCCESS;

	ut_ad(node->rec_type == TRX_UNDO_UPD_DEL_REC);
	ut_ad(!node->undo_row);
	heap = mem_heap_create(1024);

	while (node->index != NULL) {

		/* Skip all corrupted secondary index */
		dict_table_skip_corrupt_index(node->index);

		if (!node->index) {
			break;
		}

		index = node->index;

		entry = row_build_index_entry(node->row, node->ext,
					      index, heap);
		if (UNIV_UNLIKELY(!entry)) {
			/* The database must have crashed after
			inserting a clustered index record but before
			writing all the externally stored columns of
			that record.  Because secondary index entries
			are inserted after the clustered index record,
			we may assume that the secondary index record
			does not exist.  However, this situation may
			only occur during the rollback of incomplete
			transactions. */
			ut_a(thr_is_recv(thr));
		} else {
			err = row_undo_mod_del_mark_or_remove_sec(
				node, thr, index, entry);

			if (err != DB_SUCCESS) {

				break;
			}
		}

		mem_heap_empty(heap);

		node->index = dict_table_get_next_index(node->index);
	}

	mem_heap_free(heap);

	return(err);
}

/***********************************************************//**
Undoes a modify in secondary indexes when undo record type is DEL_MARK.
@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_mod_del_mark_sec(
/*======================*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	mem_heap_t*	heap;
	dtuple_t*	entry;
	dict_index_t*	index;
	ulint		err;

	ut_ad(!node->undo_row);

	heap = mem_heap_create(1024);

	while (node->index != NULL) {
		/* Skip all corrupted secondary index */
		dict_table_skip_corrupt_index(node->index);

		if (!node->index) {
			break;
		}

		index = node->index;

		entry = row_build_index_entry(node->row, node->ext,
					      index, heap);
		ut_a(entry);
		err = row_undo_mod_del_unmark_sec_and_undo_update(
			BTR_MODIFY_LEAF, thr, index, entry);
		if (err == DB_FAIL) {
			err = row_undo_mod_del_unmark_sec_and_undo_update(
				BTR_MODIFY_TREE, thr, index, entry);
		}

		if (err != DB_SUCCESS) {

			mem_heap_free(heap);

			return(err);
		}

		node->index = dict_table_get_next_index(node->index);
	}

	mem_heap_free(heap);

	return(DB_SUCCESS);
}

/***********************************************************//**
Undoes a modify in secondary indexes when undo record type is UPD_EXIST.
@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_mod_upd_exist_sec(
/*=======================*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	mem_heap_t*	heap;
	dtuple_t*	entry;
	dict_index_t*	index;
	ulint		err;

	if (node->cmpl_info & UPD_NODE_NO_ORD_CHANGE) {
		/* No change in secondary indexes */

		return(DB_SUCCESS);
	}

	heap = mem_heap_create(1024);

	while (node->index != NULL) {
		/* Skip all corrupted secondary index */
		dict_table_skip_corrupt_index(node->index);

		if (!node->index) {
			break;
		}

		index = node->index;

		if (row_upd_changes_ord_field_binary(node->index, node->update,
						     thr,
						     node->row, node->ext)) {

			/* Build the newest version of the index entry */
			entry = row_build_index_entry(node->row, node->ext,
						      index, heap);
			if (UNIV_UNLIKELY(!entry)) {
				/* The server must have crashed in
				row_upd_clust_rec_by_insert() before
				the updated externally stored columns (BLOBs)
				of the new clustered index entry were
				written. */

				/* The table must be in DYNAMIC or COMPRESSED
				format.  REDUNDANT and COMPACT formats
				store a local 768-byte prefix of each
				externally stored column. */
				ut_a(dict_table_get_format(index->table)
				     >= DICT_TF_FORMAT_ZIP);

				/* This is only legitimate when
				rolling back an incomplete transaction
				after crash recovery. */
				ut_a(thr_get_trx(thr)->is_recovered);

				/* The server must have crashed before
				completing the insert of the new
				clustered index entry and before
				inserting to the secondary indexes.
				Because node->row was not yet written
				to this index, we can ignore it.  But
				we must restore node->undo_row. */
			} else {
				/* NOTE that if we updated the fields of a
				delete-marked secondary index record so that
				alphabetically they stayed the same, e.g.,
				'abc' -> 'aBc', we cannot return to the
				original values because we do not know them.
				But this should not cause problems because
				in row0sel.c, in queries we always retrieve
				the clustered index record or an earlier
				version of it, if the secondary index record
				through which we do the search is
				delete-marked. */

				err = row_undo_mod_del_mark_or_remove_sec(
					node, thr, index, entry);
				if (err != DB_SUCCESS) {
					mem_heap_free(heap);

					return(err);
				}

				mem_heap_empty(heap);
			}

			/* We may have to update the delete mark in the
			secondary index record of the previous version of
			the row. We also need to update the fields of
			the secondary index record if we updated its fields
			but alphabetically they stayed the same, e.g.,
			'abc' -> 'aBc'. */
			entry = row_build_index_entry(node->undo_row,
						      node->undo_ext,
						      index, heap);
			ut_a(entry);

			err = row_undo_mod_del_unmark_sec_and_undo_update(
				BTR_MODIFY_LEAF, thr, index, entry);
			if (err == DB_FAIL) {
				err = row_undo_mod_del_unmark_sec_and_undo_update(
					BTR_MODIFY_TREE, thr, index, entry);
			}

			if (err != DB_SUCCESS) {
				mem_heap_free(heap);

				return(err);
			}
		}

		node->index = dict_table_get_next_index(node->index);
	}

	mem_heap_free(heap);

	return(DB_SUCCESS);
}

/***********************************************************//**
Parses the row reference and other info in a modify undo log record. */
static
void
row_undo_mod_parse_undo_rec(
/*========================*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	dict_index_t*	clust_index;
	byte*		ptr;
	undo_no_t	undo_no;
	table_id_t	table_id;
	trx_id_t	trx_id;
	roll_ptr_t	roll_ptr;
	ulint		info_bits;
	ulint		type;
	ulint		cmpl_info;
	ibool		dummy_extern;
	trx_t*		trx;

	ut_ad(node && thr);
	trx = thr_get_trx(thr);
	ptr = trx_undo_rec_get_pars(node->undo_rec, &type, &cmpl_info,
				    &dummy_extern, &undo_no, &table_id);
	node->rec_type = type;

	node->table = dict_table_get_on_id(table_id, trx);

	/* TODO: other fixes associated with DROP TABLE + rollback in the
	same table by another user */

	if (node->table == NULL) {
		/* Table was dropped */
		return;
	}

	if (node->table->ibd_file_missing) {
		/* We skip undo operations to missing .ibd files */
		node->table = NULL;

		return;
	}

	clust_index = dict_table_get_first_index(node->table);

	ptr = trx_undo_update_rec_get_sys_cols(ptr, &trx_id, &roll_ptr,
					       &info_bits);

	ptr = trx_undo_rec_get_row_ref(ptr, clust_index, &(node->ref),
				       node->heap);

	trx_undo_update_rec_get_update(ptr, clust_index, type, trx_id,
				       roll_ptr, info_bits, trx,
				       node->heap, &(node->update));
	node->new_trx_id = trx_id;
	node->cmpl_info = cmpl_info;
}

/***********************************************************//**
Undoes a modify operation on a row of a table.
@return	DB_SUCCESS or error code */
UNIV_INTERN
ulint
row_undo_mod(
/*=========*/
	undo_node_t*	node,	/*!< in: row undo node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	ulint	err;

	ut_ad(node && thr);
	ut_ad(node->state == UNDO_NODE_MODIFY);

	row_undo_mod_parse_undo_rec(node, thr);

	if (!node->table || !row_undo_search_clust_to_pcur(node)) {
		/* It is already undone, or will be undone by another query
		thread, or table was dropped */

		trx_undo_rec_release(node->trx, node->undo_no);
		node->state = UNDO_NODE_FETCH_NEXT;

		return(DB_SUCCESS);
	}

	node->index = dict_table_get_next_index(
		dict_table_get_first_index(node->table));

	/* Skip all corrupted secondary index */
	dict_table_skip_corrupt_index(node->index);

	if (node->rec_type == TRX_UNDO_UPD_EXIST_REC) {

		err = row_undo_mod_upd_exist_sec(node, thr);

	} else if (node->rec_type == TRX_UNDO_DEL_MARK_REC) {

		err = row_undo_mod_del_mark_sec(node, thr);
	} else {
		ut_ad(node->rec_type == TRX_UNDO_UPD_DEL_REC);
		err = row_undo_mod_upd_del_sec(node, thr);
	}

	if (err != DB_SUCCESS) {

		return(err);
	}

	err = row_undo_mod_clust(node, thr);

	return(err);
}