row0upd.c 65.1 KB
Newer Older
1 2
/*****************************************************************************

3
Copyright (c) 1996, 2012, Oracle and/or its affiliates. All Rights Reserved.
4 5 6 7 8 9 10 11 12 13

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
14 15
this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
16 17 18 19 20 21 22 23 24 25

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

/**************************************************//**
@file row/row0upd.c
Update of a row

Created 12/27/1996 Heikki Tuuri
*******************************************************/

Marko Mäkelä's avatar
Marko Mäkelä committed
26 27 28 29 30 31 32 33 34
#ifdef __WIN__
/* error LNK2001: unresolved external symbol _debug_sync_C_callback_ptr */
# define DEBUG_SYNC_C(dummy) ((void) 0)
#else
# include "my_global.h" /* HAVE_* */
# include "m_string.h" /* for my_sys.h */
# include "my_sys.h" /* DEBUG_SYNC_C */
#endif

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
#include "row0upd.h"

#ifdef UNIV_NONINL
#include "row0upd.ic"
#endif

#include "dict0dict.h"
#include "trx0undo.h"
#include "rem0rec.h"
#ifndef UNIV_HOTBACKUP
#include "dict0boot.h"
#include "dict0crea.h"
#include "mach0data.h"
#include "btr0btr.h"
#include "btr0cur.h"
#include "que0que.h"
#include "row0ext.h"
#include "row0ins.h"
#include "row0sel.h"
#include "row0row.h"
#include "rem0cmp.h"
#include "lock0lock.h"
#include "log0log.h"
#include "pars0sym.h"
#include "eval0eval.h"
#include "buf0lru.h"


/* What kind of latch and lock can we assume when the control comes to
   -------------------------------------------------------------------
an update node?
--------------
Efficiency of massive updates would require keeping an x-latch on a
clustered index page through many updates, and not setting an explicit
x-lock on clustered index records, as they anyway will get an implicit
x-lock when they are updated. A problem is that the read nodes in the
graph should know that they must keep the latch when passing the control
up to the update node, and not set any record lock on the record which
will be updated. Another problem occurs if the execution is stopped,
as the kernel switches to another query thread, or the transaction must
wait for a lock. Then we should be able to release the latch and, maybe,
acquire an explicit x-lock on the record.
	Because this seems too complicated, we conclude that the less
efficient solution of releasing all the latches when the control is
transferred to another node, and acquiring explicit x-locks, is better. */

/* How is a delete performed? If there is a delete without an
explicit cursor, i.e., a searched delete, there are at least
two different situations:
the implicit select cursor may run on (1) the clustered index or
on (2) a secondary index. The delete is performed by setting
the delete bit in the record and substituting the id of the
deleting transaction for the original trx id, and substituting a
new roll ptr for previous roll ptr. The old trx id and roll ptr
are saved in the undo log record. Thus, no physical changes occur
in the index tree structure at the time of the delete. Only
when the undo log is purged, the index records will be physically
deleted from the index trees.

The query graph executing a searched delete would consist of
a delete node which has as a subtree a select subgraph.
The select subgraph should return a (persistent) cursor
in the clustered index, placed on page which is x-latched.
The delete node should look for all secondary index records for
this clustered index entry and mark them as deleted. When is
the x-latch freed? The most efficient way for performing a
searched delete is obviously to keep the x-latch for several
steps of query graph execution. */

Inaam Rana's avatar
Inaam Rana committed
104
/*************************************************************************
105 106 107 108 109 110 111 112 113
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. */

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
/***********************************************************//**
Checks if an update vector changes some of the first ordering fields of an
index record. This is only used in foreign key checks and we can assume
that index does not contain column prefixes.
@return	TRUE if changes */
static
ibool
row_upd_changes_first_fields_binary(
/*================================*/
	dtuple_t*	entry,	/*!< in: old value of index entry */
	dict_index_t*	index,	/*!< in: index of entry */
	const upd_t*	update,	/*!< in: update vector for the row */
	ulint		n);	/*!< in: how many first fields to check */


/*********************************************************************//**
Checks if index currently is mentioned as a referenced index in a foreign
key constraint.

NOTE that since we do not hold dict_operation_lock when leaving the
function, it may be that the referencing table has been dropped when
we leave this function: this function is only for heuristic use!

@return TRUE if referenced */
static
ibool
row_upd_index_is_referenced(
/*========================*/
	dict_index_t*	index,	/*!< in: index */
	trx_t*		trx)	/*!< in: transaction */
{
	dict_table_t*	table		= index->table;
	dict_foreign_t*	foreign;
	ibool		froze_data_dict	= FALSE;
	ibool		is_referenced	= FALSE;

	if (!UT_LIST_GET_FIRST(table->referenced_list)) {

		return(FALSE);
	}

	if (trx->dict_operation_lock_mode == 0) {
		row_mysql_freeze_data_dictionary(trx);
		froze_data_dict = TRUE;
	}

	foreign = UT_LIST_GET_FIRST(table->referenced_list);

	while (foreign) {
		if (foreign->referenced_index == index) {

			is_referenced = TRUE;
			goto func_exit;
		}

		foreign = UT_LIST_GET_NEXT(referenced_list, foreign);
	}

func_exit:
	if (froze_data_dict) {
		row_mysql_unfreeze_data_dictionary(trx);
	}

	return(is_referenced);
}

/*********************************************************************//**
Checks if possible foreign key constraints hold after a delete of the record
under pcur.

NOTE that this function will temporarily commit mtr and lose the
pcur position!

@return	DB_SUCCESS or an error code */
static
ulint
row_upd_check_references_constraints(
/*=================================*/
	upd_node_t*	node,	/*!< in: row update node */
	btr_pcur_t*	pcur,	/*!< in: cursor positioned on a record; NOTE: the
				cursor position is lost in this function! */
	dict_table_t*	table,	/*!< in: table in question */
	dict_index_t*	index,	/*!< in: index of the cursor */
	ulint*		offsets,/*!< in/out: rec_get_offsets(pcur.rec, index) */
	que_thr_t*	thr,	/*!< in: query thread */
	mtr_t*		mtr)	/*!< in: mtr */
{
	dict_foreign_t*	foreign;
	mem_heap_t*	heap;
	dtuple_t*	entry;
	trx_t*		trx;
	const rec_t*	rec;
	ulint		n_ext;
	ulint		err;
	ibool		got_s_lock	= FALSE;

	if (UT_LIST_GET_FIRST(table->referenced_list) == NULL) {

		return(DB_SUCCESS);
	}

	trx = thr_get_trx(thr);

	rec = btr_pcur_get_rec(pcur);
	ut_ad(rec_offs_validate(rec, index, offsets));

	heap = mem_heap_create(500);

	entry = row_rec_to_index_entry(ROW_COPY_DATA, rec, index, offsets,
				       &n_ext, heap);

	mtr_commit(mtr);

	mtr_start(mtr);

	if (trx->dict_operation_lock_mode == 0) {
		got_s_lock = TRUE;

		row_mysql_freeze_data_dictionary(trx);
	}

	foreign = UT_LIST_GET_FIRST(table->referenced_list);

	while (foreign) {
		/* Note that we may have an update which updates the index
		record, but does NOT update the first fields which are
		referenced in a foreign key constraint. Then the update does
		NOT break the constraint. */

		if (foreign->referenced_index == index
		    && (node->is_delete
			|| row_upd_changes_first_fields_binary(
				entry, index, node->update,
				foreign->n_fields))) {

			if (foreign->foreign_table == NULL) {
				dict_table_get(foreign->foreign_table_name,
					       FALSE);
			}

			if (foreign->foreign_table) {
				mutex_enter(&(dict_sys->mutex));

				(foreign->foreign_table
				 ->n_foreign_key_checks_running)++;

				mutex_exit(&(dict_sys->mutex));
			}

			/* NOTE that if the thread ends up waiting for a lock
			we will release dict_operation_lock temporarily!
			But the counter on the table protects 'foreign' from
			being dropped while the check is running. */

			err = row_ins_check_foreign_constraint(
				FALSE, foreign, table, entry, thr);

			if (foreign->foreign_table) {
				mutex_enter(&(dict_sys->mutex));

				ut_a(foreign->foreign_table
				     ->n_foreign_key_checks_running > 0);

				(foreign->foreign_table
				 ->n_foreign_key_checks_running)--;

				mutex_exit(&(dict_sys->mutex));
			}

			if (err != DB_SUCCESS) {

				goto func_exit;
			}
		}

		foreign = UT_LIST_GET_NEXT(referenced_list, foreign);
	}

	err = DB_SUCCESS;

func_exit:
	if (got_s_lock) {
		row_mysql_unfreeze_data_dictionary(trx);
	}

	mem_heap_free(heap);

	return(err);
}

/*********************************************************************//**
Creates an update node for a query graph.
@return	own: update node */
UNIV_INTERN
upd_node_t*
upd_node_create(
/*============*/
	mem_heap_t*	heap)	/*!< in: mem heap where created */
{
	upd_node_t*	node;

	node = mem_heap_alloc(heap, sizeof(upd_node_t));
	node->common.type = QUE_NODE_UPDATE;

	node->state = UPD_NODE_UPDATE_CLUSTERED;
	node->in_mysql_interface = FALSE;

	node->row = NULL;
	node->ext = NULL;
	node->upd_row = NULL;
	node->upd_ext = NULL;
	node->index = NULL;
	node->update = NULL;

	node->foreign = NULL;
	node->cascade_heap = NULL;
	node->cascade_node = NULL;

	node->select = NULL;

	node->heap = mem_heap_create(128);
	node->magic_n = UPD_NODE_MAGIC_N;

	node->cmpl_info = 0;

	return(node);
}
#endif /* !UNIV_HOTBACKUP */

/*********************************************************************//**
Updates the trx id and roll ptr field in a clustered index record in database
recovery. */
UNIV_INTERN
void
row_upd_rec_sys_fields_in_recovery(
/*===============================*/
	rec_t*		rec,	/*!< in/out: record */
	page_zip_des_t*	page_zip,/*!< in/out: compressed page, or NULL */
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	ulint		pos,	/*!< in: TRX_ID position in rec */
	trx_id_t	trx_id,	/*!< in: transaction id */
	roll_ptr_t	roll_ptr)/*!< in: roll ptr of the undo log record */
{
	ut_ad(rec_offs_validate(rec, NULL, offsets));

	if (UNIV_LIKELY_NULL(page_zip)) {
		page_zip_write_trx_id_and_roll_ptr(
			page_zip, rec, offsets, pos, trx_id, roll_ptr);
	} else {
		byte*	field;
		ulint	len;

		field = rec_get_nth_field(rec, offsets, pos, &len);
		ut_ad(len == DATA_TRX_ID_LEN);
#if DATA_TRX_ID + 1 != DATA_ROLL_PTR
# error "DATA_TRX_ID + 1 != DATA_ROLL_PTR"
#endif
		trx_write_trx_id(field, trx_id);
		trx_write_roll_ptr(field + DATA_TRX_ID_LEN, roll_ptr);
	}
}

#ifndef UNIV_HOTBACKUP
/*********************************************************************//**
Sets the trx id or roll ptr field of a clustered index entry. */
UNIV_INTERN
void
row_upd_index_entry_sys_field(
/*==========================*/
383 384
	dtuple_t*	entry,	/*!< in/out: index entry, where the memory
				buffers for sys fields are already allocated:
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
				the function just copies the new values to
				them */
	dict_index_t*	index,	/*!< in: clustered index */
	ulint		type,	/*!< in: DATA_TRX_ID or DATA_ROLL_PTR */
	dulint		val)	/*!< in: value to write */
{
	dfield_t*	dfield;
	byte*		field;
	ulint		pos;

	ut_ad(dict_index_is_clust(index));

	pos = dict_index_get_sys_col_pos(index, type);

	dfield = dtuple_get_nth_field(entry, pos);
	field = dfield_get_data(dfield);

	if (type == DATA_TRX_ID) {
		trx_write_trx_id(field, val);
	} else {
		ut_ad(type == DATA_ROLL_PTR);
		trx_write_roll_ptr(field, val);
	}
}

/***********************************************************//**
Returns TRUE if row update changes size of some field in index or if some
field to be updated is stored externally in rec or update.
@return TRUE if the update changes the size of some field in index or
the field is external in rec or update */
UNIV_INTERN
ibool
row_upd_changes_field_size_or_external(
/*===================================*/
	dict_index_t*	index,	/*!< in: index */
	const ulint*	offsets,/*!< in: rec_get_offsets(rec, index) */
	const upd_t*	update)	/*!< in: update vector */
{
	const upd_field_t*	upd_field;
	const dfield_t*		new_val;
	ulint			old_len;
	ulint			new_len;
	ulint			n_fields;
	ulint			i;

	ut_ad(rec_offs_validate(NULL, index, offsets));
	n_fields = upd_get_n_fields(update);

	for (i = 0; i < n_fields; i++) {
		upd_field = upd_get_nth_field(update, i);

		new_val = &(upd_field->new_val);
		new_len = dfield_get_len(new_val);

		if (dfield_is_null(new_val) && !rec_offs_comp(offsets)) {
			/* A bug fixed on Dec 31st, 2004: we looked at the
			SQL NULL size from the wrong field! We may backport
			this fix also to 4.0. The merge to 5.0 will be made
			manually immediately after we commit this to 4.1. */

			new_len = dict_col_get_sql_null_size(
				dict_index_get_nth_col(index,
						       upd_field->field_no),
				0);
		}

		old_len = rec_offs_nth_size(offsets, upd_field->field_no);

		if (rec_offs_comp(offsets)
		    && rec_offs_nth_sql_null(offsets,
					     upd_field->field_no)) {
			/* Note that in the compact table format, for a
			variable length field, an SQL NULL will use zero
			bytes in the offset array at the start of the physical
			record, but a zero-length value (empty string) will
			use one byte! Thus, we cannot use update-in-place
			if we update an SQL NULL varchar to an empty string! */

			old_len = UNIV_SQL_NULL;
		}

		if (dfield_is_ext(new_val) || old_len != new_len
		    || rec_offs_nth_extern(offsets, upd_field->field_no)) {

			return(TRUE);
		}
	}

	return(FALSE);
}
#endif /* !UNIV_HOTBACKUP */

/***********************************************************//**
478 479 480 481 482
Replaces the new column values stored in the update vector to the
record given. No field size changes are allowed. This function is
usually invoked on a clustered index. The only use case for a
secondary index is row_ins_sec_index_entry_by_modify() or its
counterpart in ibuf_insert_to_index_page(). */
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
UNIV_INTERN
void
row_upd_rec_in_place(
/*=================*/
	rec_t*		rec,	/*!< in/out: record where replaced */
	dict_index_t*	index,	/*!< in: the index the record belongs to */
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	const upd_t*	update,	/*!< in: update vector */
	page_zip_des_t*	page_zip)/*!< in: compressed page with enough space
				available, or NULL */
{
	const upd_field_t*	upd_field;
	const dfield_t*		new_val;
	ulint			n_fields;
	ulint			i;

	ut_ad(rec_offs_validate(rec, index, offsets));

	if (rec_offs_comp(offsets)) {
		rec_set_info_bits_new(rec, update->info_bits);
	} else {
		rec_set_info_bits_old(rec, update->info_bits);
	}

	n_fields = upd_get_n_fields(update);

	for (i = 0; i < n_fields; i++) {
510 511 512 513 514
#ifdef UNIV_BLOB_DEBUG
		btr_blob_dbg_t	b;
		const byte*	field_ref	= NULL;
#endif /* UNIV_BLOB_DEBUG */

515 516 517 518
		upd_field = upd_get_nth_field(update, i);
		new_val = &(upd_field->new_val);
		ut_ad(!dfield_is_ext(new_val) ==
		      !rec_offs_nth_extern(offsets, upd_field->field_no));
519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
#ifdef UNIV_BLOB_DEBUG
		if (dfield_is_ext(new_val)) {
			ulint	len;
			field_ref = rec_get_nth_field(rec, offsets, i, &len);
			ut_a(len != UNIV_SQL_NULL);
			ut_a(len >= BTR_EXTERN_FIELD_REF_SIZE);
			field_ref += len - BTR_EXTERN_FIELD_REF_SIZE;

			b.ref_page_no = page_get_page_no(page_align(rec));
			b.ref_heap_no = page_rec_get_heap_no(rec);
			b.ref_field_no = i;
			b.blob_page_no = mach_read_from_4(
				field_ref + BTR_EXTERN_PAGE_NO);
			ut_a(b.ref_field_no >= index->n_uniq);
			btr_blob_dbg_rbt_delete(index, &b, "upd_in_place");
		}
#endif /* UNIV_BLOB_DEBUG */
536 537 538 539

		rec_set_nth_field(rec, offsets, upd_field->field_no,
				  dfield_get_data(new_val),
				  dfield_get_len(new_val));
540 541 542 543 544 545 546 547 548 549 550 551 552

#ifdef UNIV_BLOB_DEBUG
		if (dfield_is_ext(new_val)) {
			b.blob_page_no = mach_read_from_4(
				field_ref + BTR_EXTERN_PAGE_NO);
			b.always_owner = b.owner = !(field_ref[BTR_EXTERN_LEN]
						     & BTR_EXTERN_OWNER_FLAG);
			b.del = rec_get_deleted_flag(
				rec, rec_offs_comp(offsets));

			btr_blob_dbg_rbt_insert(index, &b, "upd_in_place");
		}
#endif /* UNIV_BLOB_DEBUG */
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
	}

	if (UNIV_LIKELY_NULL(page_zip)) {
		page_zip_write_rec(page_zip, rec, index, offsets, 0);
	}
}

#ifndef UNIV_HOTBACKUP
/*********************************************************************//**
Writes into the redo log the values of trx id and roll ptr and enough info
to determine their positions within a clustered index record.
@return	new pointer to mlog */
UNIV_INTERN
byte*
row_upd_write_sys_vals_to_log(
/*==========================*/
	dict_index_t*	index,	/*!< in: clustered index */
	trx_t*		trx,	/*!< in: transaction */
	roll_ptr_t	roll_ptr,/*!< in: roll ptr of the undo log record */
	byte*		log_ptr,/*!< pointer to a buffer of size > 20 opened
				in mlog */
	mtr_t*		mtr __attribute__((unused))) /*!< in: mtr */
{
	ut_ad(dict_index_is_clust(index));
	ut_ad(mtr);

	log_ptr += mach_write_compressed(log_ptr,
					 dict_index_get_sys_col_pos(
						 index, DATA_TRX_ID));

	trx_write_roll_ptr(log_ptr, roll_ptr);
	log_ptr += DATA_ROLL_PTR_LEN;

	log_ptr += mach_dulint_write_compressed(log_ptr, trx->id);

	return(log_ptr);
}
#endif /* !UNIV_HOTBACKUP */

/*********************************************************************//**
Parses the log data of system field values.
@return	log data end or NULL */
UNIV_INTERN
byte*
row_upd_parse_sys_vals(
/*===================*/
	byte*		ptr,	/*!< in: buffer */
	byte*		end_ptr,/*!< in: buffer end */
	ulint*		pos,	/*!< out: TRX_ID position in record */
	trx_id_t*	trx_id,	/*!< out: trx id */
	roll_ptr_t*	roll_ptr)/*!< out: roll ptr */
{
	ptr = mach_parse_compressed(ptr, end_ptr, pos);

	if (ptr == NULL) {

		return(NULL);
	}

	if (end_ptr < ptr + DATA_ROLL_PTR_LEN) {

		return(NULL);
	}

	*roll_ptr = trx_read_roll_ptr(ptr);
	ptr += DATA_ROLL_PTR_LEN;

	ptr = mach_dulint_parse_compressed(ptr, end_ptr, trx_id);

	return(ptr);
}

#ifndef UNIV_HOTBACKUP
/***********************************************************//**
Writes to the redo log the new values of the fields occurring in the index. */
UNIV_INTERN
void
row_upd_index_write_log(
/*====================*/
	const upd_t*	update,	/*!< in: update vector */
	byte*		log_ptr,/*!< in: pointer to mlog buffer: must
				contain at least MLOG_BUF_MARGIN bytes
				of free space; the buffer is closed
				within this function */
	mtr_t*		mtr)	/*!< in: mtr into whose log to write */
{
	const upd_field_t*	upd_field;
	const dfield_t*		new_val;
	ulint			len;
	ulint			n_fields;
	byte*			buf_end;
	ulint			i;

	n_fields = upd_get_n_fields(update);

	buf_end = log_ptr + MLOG_BUF_MARGIN;

	mach_write_to_1(log_ptr, update->info_bits);
	log_ptr++;
	log_ptr += mach_write_compressed(log_ptr, n_fields);

	for (i = 0; i < n_fields; i++) {

#if MLOG_BUF_MARGIN <= 30
# error "MLOG_BUF_MARGIN <= 30"
#endif

		if (log_ptr + 30 > buf_end) {
			mlog_close(mtr, log_ptr);

			log_ptr = mlog_open(mtr, MLOG_BUF_MARGIN);
			buf_end = log_ptr + MLOG_BUF_MARGIN;
		}

		upd_field = upd_get_nth_field(update, i);

		new_val = &(upd_field->new_val);

		len = dfield_get_len(new_val);

		log_ptr += mach_write_compressed(log_ptr, upd_field->field_no);
		log_ptr += mach_write_compressed(log_ptr, len);

		if (len != UNIV_SQL_NULL) {
			if (log_ptr + len < buf_end) {
				memcpy(log_ptr, dfield_get_data(new_val), len);

				log_ptr += len;
			} else {
				mlog_close(mtr, log_ptr);

				mlog_catenate_string(mtr,
						     dfield_get_data(new_val),
						     len);

				log_ptr = mlog_open(mtr, MLOG_BUF_MARGIN);
				buf_end = log_ptr + MLOG_BUF_MARGIN;
			}
		}
	}

	mlog_close(mtr, log_ptr);
}
#endif /* !UNIV_HOTBACKUP */

/*********************************************************************//**
Parses the log data written by row_upd_index_write_log.
@return	log data end or NULL */
UNIV_INTERN
byte*
row_upd_index_parse(
/*================*/
	byte*		ptr,	/*!< in: buffer */
	byte*		end_ptr,/*!< in: buffer end */
	mem_heap_t*	heap,	/*!< in: memory heap where update vector is
				built */
	upd_t**		update_out)/*!< out: update vector */
{
	upd_t*		update;
	upd_field_t*	upd_field;
	dfield_t*	new_val;
	ulint		len;
	ulint		n_fields;
	ulint		info_bits;
	ulint		i;

	if (end_ptr < ptr + 1) {

		return(NULL);
	}

	info_bits = mach_read_from_1(ptr);
	ptr++;
	ptr = mach_parse_compressed(ptr, end_ptr, &n_fields);

	if (ptr == NULL) {

		return(NULL);
	}

	update = upd_create(n_fields, heap);
	update->info_bits = info_bits;

	for (i = 0; i < n_fields; i++) {
		ulint	field_no;
		upd_field = upd_get_nth_field(update, i);
		new_val = &(upd_field->new_val);

		ptr = mach_parse_compressed(ptr, end_ptr, &field_no);

		if (ptr == NULL) {

			return(NULL);
		}

		upd_field->field_no = field_no;

		ptr = mach_parse_compressed(ptr, end_ptr, &len);

		if (ptr == NULL) {

			return(NULL);
		}

		if (len != UNIV_SQL_NULL) {

			if (end_ptr < ptr + len) {

				return(NULL);
			}

			dfield_set_data(new_val,
					mem_heap_dup(heap, ptr, len), len);
			ptr += len;
		} else {
			dfield_set_null(new_val);
		}
	}

	*update_out = update;

	return(ptr);
}

#ifndef UNIV_HOTBACKUP
/***************************************************************//**
Builds an update vector from those fields which in a secondary index entry
differ from a record that has the equal ordering fields. NOTE: we compare
the fields as binary strings!
@return	own: update vector of differing fields */
UNIV_INTERN
upd_t*
row_upd_build_sec_rec_difference_binary(
/*====================================*/
	dict_index_t*	index,	/*!< in: index */
	const dtuple_t*	entry,	/*!< in: entry to insert */
	const rec_t*	rec,	/*!< in: secondary index record */
	trx_t*		trx,	/*!< in: transaction */
	mem_heap_t*	heap)	/*!< in: memory heap from which allocated */
{
	upd_field_t*	upd_field;
	const dfield_t*	dfield;
	const byte*	data;
	ulint		len;
	upd_t*		update;
	ulint		n_diff;
	ulint		i;
	ulint		offsets_[REC_OFFS_SMALL_SIZE];
	const ulint*	offsets;
	rec_offs_init(offsets_);

	/* This function is used only for a secondary index */
	ut_a(!dict_index_is_clust(index));

	update = upd_create(dtuple_get_n_fields(entry), heap);

	n_diff = 0;
	offsets = rec_get_offsets(rec, index, offsets_,
				  ULINT_UNDEFINED, &heap);

	for (i = 0; i < dtuple_get_n_fields(entry); i++) {

		data = rec_get_nth_field(rec, offsets, i, &len);

		dfield = dtuple_get_nth_field(entry, i);

		/* NOTE that it may be that len != dfield_get_len(dfield) if we
		are updating in a character set and collation where strings of
		different length can be equal in an alphabetical comparison,
		and also in the case where we have a column prefix index
		and the last characters in the index field are spaces; the
		latter case probably caused the assertion failures reported at
		row0upd.c line 713 in versions 4.0.14 - 4.0.16. */

		/* NOTE: we compare the fields as binary strings!
		(No collation) */

		if (!dfield_data_is_binary_equal(dfield, len, data)) {

			upd_field = upd_get_nth_field(update, n_diff);

			dfield_copy(&(upd_field->new_val), dfield);

			upd_field_set_field_no(upd_field, i, index, trx);

			n_diff++;
		}
	}

	update->n_fields = n_diff;

	return(update);
}

/***************************************************************//**
Builds an update vector from those fields, excluding the roll ptr and
trx id fields, which in an index entry differ from a record that has
the equal ordering fields. NOTE: we compare the fields as binary strings!
@return own: update vector of differing fields, excluding roll ptr and
trx id */
UNIV_INTERN
upd_t*
row_upd_build_difference_binary(
/*============================*/
	dict_index_t*	index,	/*!< in: clustered index */
	const dtuple_t*	entry,	/*!< in: entry to insert */
	const rec_t*	rec,	/*!< in: clustered index record */
	trx_t*		trx,	/*!< in: transaction */
	mem_heap_t*	heap)	/*!< in: memory heap from which allocated */
{
	upd_field_t*	upd_field;
	const dfield_t*	dfield;
	const byte*	data;
	ulint		len;
	upd_t*		update;
	ulint		n_diff;
	ulint		roll_ptr_pos;
	ulint		trx_id_pos;
	ulint		i;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	const ulint*	offsets;
	rec_offs_init(offsets_);

	/* This function is used only for a clustered index */
	ut_a(dict_index_is_clust(index));

	update = upd_create(dtuple_get_n_fields(entry), heap);

	n_diff = 0;

	roll_ptr_pos = dict_index_get_sys_col_pos(index, DATA_ROLL_PTR);
	trx_id_pos = dict_index_get_sys_col_pos(index, DATA_TRX_ID);

	offsets = rec_get_offsets(rec, index, offsets_,
				  ULINT_UNDEFINED, &heap);

	for (i = 0; i < dtuple_get_n_fields(entry); i++) {

		data = rec_get_nth_field(rec, offsets, i, &len);

		dfield = dtuple_get_nth_field(entry, i);

		/* NOTE: we compare the fields as binary strings!
		(No collation) */

		if (i == trx_id_pos || i == roll_ptr_pos) {

			goto skip_compare;
		}

		if (UNIV_UNLIKELY(!dfield_is_ext(dfield)
				  != !rec_offs_nth_extern(offsets, i))
		    || !dfield_data_is_binary_equal(dfield, len, data)) {

			upd_field = upd_get_nth_field(update, n_diff);

			dfield_copy(&(upd_field->new_val), dfield);

			upd_field_set_field_no(upd_field, i, index, trx);

			n_diff++;
		}
skip_compare:
		;
	}

	update->n_fields = n_diff;

	return(update);
}

/***********************************************************//**
Fetch a prefix of an externally stored column.  This is similar
to row_ext_lookup(), but the row_ext_t holds the old values
of the column and must not be poisoned with the new values.
@return	BLOB prefix */
static
byte*
row_upd_ext_fetch(
/*==============*/
	const byte*	data,		/*!< in: 'internally' stored part of the
					field containing also the reference to
					the external part */
	ulint		local_len,	/*!< in: length of data, in bytes */
	ulint		zip_size,	/*!< in: nonzero=compressed BLOB
					page size, zero for uncompressed
					BLOBs */
	ulint*		len,		/*!< in: length of prefix to fetch;
					out: fetched length of the prefix */
	mem_heap_t*	heap)		/*!< in: heap where to allocate */
{
	byte*	buf = mem_heap_alloc(heap, *len);

	*len = btr_copy_externally_stored_field_prefix(buf, *len,
						       zip_size,
						       data, local_len);
	/* We should never update records containing a half-deleted BLOB. */
	ut_a(*len);

	return(buf);
}

/***********************************************************//**
Replaces the new column value stored in the update vector in
the given index entry field. */
static
void
row_upd_index_replace_new_col_val(
/*==============================*/
	dfield_t*		dfield,	/*!< in/out: data field
					of the index entry */
	const dict_field_t*	field,	/*!< in: index field */
	const dict_col_t*	col,	/*!< in: field->col */
	const upd_field_t*	uf,	/*!< in: update field */
	mem_heap_t*		heap,	/*!< in: memory heap for allocating
					and copying the new value */
	ulint			zip_size)/*!< in: compressed page
					 size of the table, or 0 */
{
	ulint		len;
	const byte*	data;

	dfield_copy_data(dfield, &uf->new_val);

	if (dfield_is_null(dfield)) {
		return;
	}

	len = dfield_get_len(dfield);
	data = dfield_get_data(dfield);

	if (field->prefix_len > 0) {
		ibool		fetch_ext = dfield_is_ext(dfield)
			&& len < (ulint) field->prefix_len
			+ BTR_EXTERN_FIELD_REF_SIZE;

		if (fetch_ext) {
			ulint	l = len;

			len = field->prefix_len;

			data = row_upd_ext_fetch(data, l, zip_size,
						 &len, heap);
		}

		len = dtype_get_at_most_n_mbchars(col->prtype,
						  col->mbminlen, col->mbmaxlen,
						  field->prefix_len, len,
						  (const char*) data);

		dfield_set_data(dfield, data, len);

		if (!fetch_ext) {
			dfield_dup(dfield, heap);
		}

		return;
	}

	switch (uf->orig_len) {
		byte*	buf;
	case BTR_EXTERN_FIELD_REF_SIZE:
		/* Restore the original locally stored
		part of the column.  In the undo log,
		InnoDB writes a longer prefix of externally
		stored columns, so that column prefixes
		in secondary indexes can be reconstructed. */
		dfield_set_data(dfield,
				data + len - BTR_EXTERN_FIELD_REF_SIZE,
				BTR_EXTERN_FIELD_REF_SIZE);
		dfield_set_ext(dfield);
		/* fall through */
	case 0:
		dfield_dup(dfield, heap);
		break;
	default:
		/* Reconstruct the original locally
		stored part of the column.  The data
		will have to be copied. */
		ut_a(uf->orig_len > BTR_EXTERN_FIELD_REF_SIZE);
		buf = mem_heap_alloc(heap, uf->orig_len);
		/* Copy the locally stored prefix. */
		memcpy(buf, data,
		       uf->orig_len - BTR_EXTERN_FIELD_REF_SIZE);
		/* Copy the BLOB pointer. */
		memcpy(buf + uf->orig_len - BTR_EXTERN_FIELD_REF_SIZE,
		       data + len - BTR_EXTERN_FIELD_REF_SIZE,
		       BTR_EXTERN_FIELD_REF_SIZE);

		dfield_set_data(dfield, buf, uf->orig_len);
		dfield_set_ext(dfield);
		break;
	}
}

/***********************************************************//**
Replaces the new column values stored in the update vector to the index entry
given. */
UNIV_INTERN
void
row_upd_index_replace_new_col_vals_index_pos(
/*=========================================*/
	dtuple_t*	entry,	/*!< in/out: index entry where replaced;
				the clustered index record must be
				covered by a lock or a page latch to
				prevent deletion (rollback or purge) */
	dict_index_t*	index,	/*!< in: index; NOTE that this may also be a
				non-clustered index */
	const upd_t*	update,	/*!< in: an update vector built for the index so
				that the field number in an upd_field is the
				index position */
	ibool		order_only,
				/*!< in: if TRUE, limit the replacement to
				ordering fields of index; note that this
				does not work for non-clustered indexes. */
	mem_heap_t*	heap)	/*!< in: memory heap for allocating and
				copying the new values */
{
	ulint		i;
	ulint		n_fields;
	const ulint	zip_size	= dict_table_zip_size(index->table);

	ut_ad(index);

	dtuple_set_info_bits(entry, update->info_bits);

	if (order_only) {
		n_fields = dict_index_get_n_unique(index);
	} else {
		n_fields = dict_index_get_n_fields(index);
	}

	for (i = 0; i < n_fields; i++) {
		const dict_field_t*	field;
		const dict_col_t*	col;
		const upd_field_t*	uf;

		field = dict_index_get_nth_field(index, i);
		col = dict_field_get_col(field);
		uf = upd_get_field_by_field_no(update, i);

		if (uf) {
			row_upd_index_replace_new_col_val(
				dtuple_get_nth_field(entry, i),
				field, col, uf, heap, zip_size);
		}
	}
}

/***********************************************************//**
Replaces the new column values stored in the update vector to the index entry
given. */
UNIV_INTERN
void
row_upd_index_replace_new_col_vals(
/*===============================*/
	dtuple_t*	entry,	/*!< in/out: index entry where replaced;
				the clustered index record must be
				covered by a lock or a page latch to
				prevent deletion (rollback or purge) */
	dict_index_t*	index,	/*!< in: index; NOTE that this may also be a
				non-clustered index */
	const upd_t*	update,	/*!< in: an update vector built for the
				CLUSTERED index so that the field number in
				an upd_field is the clustered index position */
	mem_heap_t*	heap)	/*!< in: memory heap for allocating and
				copying the new values */
{
	ulint			i;
	const dict_index_t*	clust_index
		= dict_table_get_first_index(index->table);
	const ulint		zip_size
		= dict_table_zip_size(index->table);

	dtuple_set_info_bits(entry, update->info_bits);

	for (i = 0; i < dict_index_get_n_fields(index); i++) {
		const dict_field_t*	field;
		const dict_col_t*	col;
		const upd_field_t*	uf;

		field = dict_index_get_nth_field(index, i);
		col = dict_field_get_col(field);
		uf = upd_get_field_by_field_no(
			update, dict_col_get_clust_pos(col, clust_index));

		if (uf) {
			row_upd_index_replace_new_col_val(
				dtuple_get_nth_field(entry, i),
				field, col, uf, heap, zip_size);
		}
	}
}

/***********************************************************//**
Replaces the new column values stored in the update vector. */
UNIV_INTERN
void
row_upd_replace(
/*============*/
	dtuple_t*		row,	/*!< in/out: row where replaced,
					indexed by col_no;
					the clustered index record must be
					covered by a lock or a page latch to
					prevent deletion (rollback or purge) */
	row_ext_t**		ext,	/*!< out, own: NULL, or externally
					stored column prefixes */
	const dict_index_t*	index,	/*!< in: clustered index */
	const upd_t*		update,	/*!< in: an update vector built for the
					clustered index */
	mem_heap_t*		heap)	/*!< in: memory heap */
{
	ulint			col_no;
	ulint			i;
	ulint			n_cols;
	ulint			n_ext_cols;
	ulint*			ext_cols;
	const dict_table_t*	table;

	ut_ad(row);
	ut_ad(ext);
	ut_ad(index);
	ut_ad(dict_index_is_clust(index));
	ut_ad(update);
	ut_ad(heap);

	n_cols = dtuple_get_n_fields(row);
	table = index->table;
	ut_ad(n_cols == dict_table_get_n_cols(table));

	ext_cols = mem_heap_alloc(heap, n_cols * sizeof *ext_cols);
	n_ext_cols = 0;

	dtuple_set_info_bits(row, update->info_bits);

	for (col_no = 0; col_no < n_cols; col_no++) {

		const dict_col_t*	col
			= dict_table_get_nth_col(table, col_no);
		const ulint		clust_pos
			= dict_col_get_clust_pos(col, index);
		dfield_t*		dfield;

		if (UNIV_UNLIKELY(clust_pos == ULINT_UNDEFINED)) {

			continue;
		}

		dfield = dtuple_get_nth_field(row, col_no);

		for (i = 0; i < upd_get_n_fields(update); i++) {

			const upd_field_t*	upd_field
				= upd_get_nth_field(update, i);

			if (upd_field->field_no != clust_pos) {

				continue;
			}

			dfield_copy_data(dfield, &upd_field->new_val);
			break;
		}

		if (dfield_is_ext(dfield) && col->ord_part) {
			ext_cols[n_ext_cols++] = col_no;
		}
	}

	if (n_ext_cols) {
		*ext = row_ext_create(n_ext_cols, ext_cols, row,
				      dict_table_zip_size(table), heap);
	} else {
		*ext = NULL;
	}
}

/***********************************************************//**
Checks if an update vector changes an ordering field of an index record.

This function is fast if the update vector is short or the number of ordering
fields in the index is small. Otherwise, this can be quadratic.
NOTE: we compare the fields as binary strings!
@return TRUE if update vector changes an ordering field in the index record */
UNIV_INTERN
ibool
1239 1240 1241 1242 1243 1244 1245 1246 1247
row_upd_changes_ord_field_binary_func(
/*==================================*/
	dict_index_t*	index,	/*!< in: index of the record */
	const upd_t*	update,	/*!< in: update vector for the row; NOTE: the
				field numbers in this MUST be clustered index
				positions! */
#ifdef UNIV_DEBUG
	const que_thr_t*thr,	/*!< in: query thread */
#endif /* UNIV_DEBUG */
1248 1249 1250 1251
	const dtuple_t*	row,	/*!< in: old value of row, or NULL if the
				row and the data values in update are not
				known when this function is called, e.g., at
				compile time */
1252
	const row_ext_t*ext)	/*!< NULL, or prefixes of the externally
1253
				stored columns in the old row */
1254
{
1255 1256 1257
	ulint			n_unique;
	ulint			i;
	const dict_index_t*	clust_index;
1258

1259
	ut_ad(index);
1260 1261 1262 1263
	ut_ad(update);
	ut_ad(thr);
	ut_ad(thr->graph);
	ut_ad(thr->graph->trx);
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273

	n_unique = dict_index_get_n_unique(index);

	clust_index = dict_table_get_first_index(index->table);

	for (i = 0; i < n_unique; i++) {

		const dict_field_t*	ind_field;
		const dict_col_t*	col;
		ulint			col_no;
1274 1275 1276 1277 1278
		const upd_field_t*	upd_field;
		const dfield_t*		dfield;
		dfield_t		dfield_ext;
		ulint			dfield_len;
		const byte*		buf;
1279 1280 1281 1282 1283

		ind_field = dict_index_get_nth_field(index, i);
		col = dict_field_get_col(ind_field);
		col_no = dict_col_get_no(col);

1284 1285
		upd_field = upd_get_field_by_field_no(
			update, dict_col_get_clust_pos(col, clust_index));
1286

1287 1288 1289
		if (upd_field == NULL) {
			continue;
		}
1290

1291 1292 1293 1294
		if (row == NULL) {
			ut_ad(ext == NULL);
			return(TRUE);
		}
1295

1296
		dfield = dtuple_get_nth_field(row, col_no);
1297

1298 1299 1300 1301 1302 1303 1304
		/* This treatment of column prefix indexes is loosely
		based on row_build_index_entry(). */

		if (UNIV_LIKELY(ind_field->prefix_len == 0)
		    || dfield_is_null(dfield)) {
			/* do nothing special */
		} else if (UNIV_LIKELY_NULL(ext)) {
1305 1306 1307 1308
			/* Silence a compiler warning without
			silencing a Valgrind error. */
			dfield_len = 0;
			UNIV_MEM_INVALID(&dfield_len, sizeof dfield_len);
1309 1310 1311 1312 1313 1314 1315
			/* See if the column is stored externally. */
			buf = row_ext_lookup(ext, col_no, &dfield_len);

			ut_ad(col->ord_part);

			if (UNIV_LIKELY_NULL(buf)) {
				if (UNIV_UNLIKELY(buf == field_ref_zero)) {
1316 1317 1318 1319 1320 1321 1322 1323
					/* The externally stored field
					was not written yet. This
					record should only be seen by
					recv_recovery_rollback_active(),
					when the server had crashed before
					storing the field. */
					ut_ad(thr->graph->trx->is_recovered);
					ut_ad(trx_is_recv(thr->graph->trx));
1324 1325 1326 1327
					return(TRUE);
				}

				goto copy_dfield;
1328
			}
1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347
		} else if (dfield_is_ext(dfield)) {
			dfield_len = dfield_get_len(dfield);
			ut_a(dfield_len > BTR_EXTERN_FIELD_REF_SIZE);
			dfield_len -= BTR_EXTERN_FIELD_REF_SIZE;
			ut_a(dict_index_is_clust(index)
			     || ind_field->prefix_len <= dfield_len);
			buf = dfield_get_data(dfield);
copy_dfield:
			ut_a(dfield_len > 0);
			dfield_copy(&dfield_ext, dfield);
			dfield_set_data(&dfield_ext, buf, dfield_len);
			dfield = &dfield_ext;
		}

		if (!dfield_datas_are_binary_equal(
			    dfield, &upd_field->new_val,
			    ind_field->prefix_len)) {

			return(TRUE);
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
		}
	}

	return(FALSE);
}

/***********************************************************//**
Checks if an update vector changes an ordering field of an index record.
NOTE: we compare the fields as binary strings!
@return TRUE if update vector may change an ordering field in an index
record */
UNIV_INTERN
ibool
row_upd_changes_some_index_ord_field_binary(
/*========================================*/
	const dict_table_t*	table,	/*!< in: table */
	const upd_t*		update)	/*!< in: update vector for the row */
{
	upd_field_t*	upd_field;
	dict_index_t*	index;
	ulint		i;

	index = dict_table_get_first_index(table);

	for (i = 0; i < upd_get_n_fields(update); i++) {

		upd_field = upd_get_nth_field(update, i);

		if (dict_field_get_col(dict_index_get_nth_field(
					       index, upd_field->field_no))
		    ->ord_part) {

			return(TRUE);
		}
	}

	return(FALSE);
}

/***********************************************************//**
Checks if an update vector changes some of the first ordering fields of an
index record. This is only used in foreign key checks and we can assume
that index does not contain column prefixes.
@return	TRUE if changes */
static
ibool
row_upd_changes_first_fields_binary(
/*================================*/
	dtuple_t*	entry,	/*!< in: index entry */
	dict_index_t*	index,	/*!< in: index of entry */
	const upd_t*	update,	/*!< in: update vector for the row */
	ulint		n)	/*!< in: how many first fields to check */
{
	ulint		n_upd_fields;
	ulint		i, j;
	dict_index_t*	clust_index;

	ut_ad(update && index);
	ut_ad(n <= dict_index_get_n_fields(index));

	n_upd_fields = upd_get_n_fields(update);
	clust_index = dict_table_get_first_index(index->table);

	for (i = 0; i < n; i++) {

		const dict_field_t*	ind_field;
		const dict_col_t*	col;
		ulint			col_pos;

		ind_field = dict_index_get_nth_field(index, i);
		col = dict_field_get_col(ind_field);
		col_pos = dict_col_get_clust_pos(col, clust_index);

		ut_a(ind_field->prefix_len == 0);

		for (j = 0; j < n_upd_fields; j++) {

			upd_field_t*	upd_field
				= upd_get_nth_field(update, j);

			if (col_pos == upd_field->field_no
			    && !dfield_datas_are_binary_equal(
				    dtuple_get_nth_field(entry, i),
1431
				    &upd_field->new_val, 0)) {
1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502

				return(TRUE);
			}
		}
	}

	return(FALSE);
}

/*********************************************************************//**
Copies the column values from a record. */
UNIV_INLINE
void
row_upd_copy_columns(
/*=================*/
	rec_t*		rec,	/*!< in: record in a clustered index */
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	sym_node_t*	column)	/*!< in: first column in a column list, or
				NULL */
{
	byte*	data;
	ulint	len;

	while (column) {
		data = rec_get_nth_field(rec, offsets,
					 column->field_nos[SYM_CLUST_FIELD_NO],
					 &len);
		eval_node_copy_and_alloc_val(column, data, len);

		column = UT_LIST_GET_NEXT(col_var_list, column);
	}
}

/*********************************************************************//**
Calculates the new values for fields to update. Note that row_upd_copy_columns
must have been called first. */
UNIV_INLINE
void
row_upd_eval_new_vals(
/*==================*/
	upd_t*	update)	/*!< in/out: update vector */
{
	que_node_t*	exp;
	upd_field_t*	upd_field;
	ulint		n_fields;
	ulint		i;

	n_fields = upd_get_n_fields(update);

	for (i = 0; i < n_fields; i++) {
		upd_field = upd_get_nth_field(update, i);

		exp = upd_field->exp;

		eval_exp(exp);

		dfield_copy_data(&(upd_field->new_val), que_node_get_val(exp));
	}
}

/***********************************************************//**
Stores to the heap the row on which the node->pcur is positioned. */
static
void
row_upd_store_row(
/*==============*/
	upd_node_t*	node)	/*!< in: row update node */
{
	dict_index_t*	clust_index;
	rec_t*		rec;
	mem_heap_t*	heap		= NULL;
1503
	row_ext_t**	ext;
1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	const ulint*	offsets;
	rec_offs_init(offsets_);

	ut_ad(node->pcur->latch_mode != BTR_NO_LATCHES);

	if (node->row != NULL) {
		mem_heap_empty(node->heap);
	}

	clust_index = dict_table_get_first_index(node->table);

	rec = btr_pcur_get_rec(node->pcur);

	offsets = rec_get_offsets(rec, clust_index, offsets_,
				  ULINT_UNDEFINED, &heap);
1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533

	if (dict_table_get_format(node->table) >= DICT_TF_FORMAT_ZIP) {
		/* In DYNAMIC or COMPRESSED format, there is no prefix
		of externally stored columns in the clustered index
		record. Build a cache of column prefixes. */
		ext = &node->ext;
	} else {
		/* REDUNDANT and COMPACT formats store a local
		768-byte prefix of each externally stored column.
		No cache is needed. */
		ext = NULL;
		node->ext = NULL;
	}

1534
	node->row = row_build(ROW_COPY_DATA, clust_index, rec, offsets,
1535
			      NULL, ext, node->heap);
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
	if (node->is_delete) {
		node->upd_row = NULL;
		node->upd_ext = NULL;
	} else {
		node->upd_row = dtuple_copy(node->row, node->heap);
		row_upd_replace(node->upd_row, &node->upd_ext,
				clust_index, node->update, node->heap);
	}

	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}
}

/***********************************************************//**
Updates a secondary index entry of a row.
@return DB_SUCCESS if operation successfully completed, else error
code or DB_LOCK_WAIT */
static
ulint
row_upd_sec_index_entry(
/*====================*/
	upd_node_t*	node,	/*!< in: row update node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	ibool		check_ref;
	ibool		found;
	dict_index_t*	index;
	dtuple_t*	entry;
	btr_pcur_t	pcur;
	btr_cur_t*	btr_cur;
	mem_heap_t*	heap;
	rec_t*		rec;
	ulint		err	= DB_SUCCESS;
	mtr_t		mtr;
	trx_t*		trx	= thr_get_trx(thr);

	index = node->index;

	check_ref = row_upd_index_is_referenced(index, trx);

	heap = mem_heap_create(1024);

	/* Build old index entry */
	entry = row_build_index_entry(node->row, node->ext, index, heap);
	ut_a(entry);

	mtr_start(&mtr);

	found = row_search_index_entry(index, entry, BTR_MODIFY_LEAF, &pcur,
				       &mtr);
	btr_cur = btr_pcur_get_btr_cur(&pcur);

	rec = btr_cur_get_rec(btr_cur);

	if (UNIV_UNLIKELY(!found)) {
		fputs("InnoDB: error in sec index entry update 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, rec, index);
		putc('\n', stderr);

		trx_print(stderr, trx, 0);

		fputs("\n"
		      "InnoDB: Submit a detailed bug report"
		      " to http://bugs.mysql.com\n", stderr);
	} else {
		/* Delete mark the old index record; it can already be
		delete marked if we return after a lock wait in
		row_ins_index_entry below */

		if (!rec_get_deleted_flag(rec,
					  dict_table_is_comp(index->table))) {
			err = btr_cur_del_mark_set_sec_rec(0, btr_cur, TRUE,
							   thr, &mtr);
			if (err == DB_SUCCESS && check_ref) {

				ulint*	offsets = rec_get_offsets(
					rec, index, NULL,
					ULINT_UNDEFINED, &heap);
				/* NOTE that the following call loses
				the position of pcur ! */
				err = row_upd_check_references_constraints(
					node, &pcur, index->table,
					index, offsets, thr, &mtr);
			}
		}
	}

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

	if (node->is_delete || err != DB_SUCCESS) {

		goto func_exit;
	}

	/* Build a new index entry */
	entry = row_build_index_entry(node->upd_row, node->upd_ext,
				      index, heap);
	ut_a(entry);

	/* Insert new index entry */
	err = row_ins_index_entry(index, entry, 0, TRUE, thr);

func_exit:
	mem_heap_free(heap);

	return(err);
}

/***********************************************************//**
Updates the secondary index record if it is changed in the row update or
deletes it if this is a delete.
@return DB_SUCCESS if operation successfully completed, else error
code or DB_LOCK_WAIT */
1658
static
1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
ulint
row_upd_sec_step(
/*=============*/
	upd_node_t*	node,	/*!< in: row update node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	ut_ad((node->state == UPD_NODE_UPDATE_ALL_SEC)
	      || (node->state == UPD_NODE_UPDATE_SOME_SEC));
	ut_ad(!dict_index_is_clust(node->index));

	if (node->state == UPD_NODE_UPDATE_ALL_SEC
1670 1671
	    || row_upd_changes_ord_field_binary(node->index, node->update,
						thr, node->row, node->ext)) {
1672 1673 1674 1675 1676 1677
		return(row_upd_sec_index_entry(node, thr));
	}

	return(DB_SUCCESS);
}

1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744
#ifdef UNIV_DEBUG
# define row_upd_clust_rec_by_insert_inherit(rec,offsets,entry,update)	\
	row_upd_clust_rec_by_insert_inherit_func(rec,offsets,entry,update)
#else /* UNIV_DEBUG */
# define row_upd_clust_rec_by_insert_inherit(rec,offsets,entry,update)	\
	row_upd_clust_rec_by_insert_inherit_func(entry,update)
#endif /* UNIV_DEBUG */
/*******************************************************************//**
Mark non-updated off-page columns inherited when the primary key is
updated. We must mark them as inherited in entry, so that they are not
freed in a rollback. A limited version of this function used to be
called btr_cur_mark_dtuple_inherited_extern().
@return TRUE if any columns were inherited */
static __attribute__((warn_unused_result))
ibool
row_upd_clust_rec_by_insert_inherit_func(
/*=====================================*/
#ifdef UNIV_DEBUG
	const rec_t*	rec,	/*!< in: old record, or NULL */
	const ulint*	offsets,/*!< in: rec_get_offsets(rec), or NULL */
#endif /* UNIV_DEBUG */
	dtuple_t*	entry,	/*!< in/out: updated entry to be
				inserted into the clustered index */
	const upd_t*	update)	/*!< in: update vector */
{
	ibool	inherit	= FALSE;
	ulint	i;

	ut_ad(!rec == !offsets);
	ut_ad(!rec || rec_offs_any_extern(offsets));

	for (i = 0; i < dtuple_get_n_fields(entry); i++) {
		dfield_t*	dfield	= dtuple_get_nth_field(entry, i);
		byte*		data;
		ulint		len;

		ut_ad(!offsets
		      || !rec_offs_nth_extern(offsets, i)
		      == !dfield_is_ext(dfield)
		      || upd_get_field_by_field_no(update, i));
		if (!dfield_is_ext(dfield)
		    || upd_get_field_by_field_no(update, i)) {
			continue;
		}

#ifdef UNIV_DEBUG
		if (UNIV_LIKELY(rec != NULL)) {
			const byte* rec_data
				= rec_get_nth_field(rec, offsets, i, &len);
			ut_ad(len == dfield_get_len(dfield));
			ut_ad(len != UNIV_SQL_NULL);
			ut_ad(len >= BTR_EXTERN_FIELD_REF_SIZE);

			rec_data += len - BTR_EXTERN_FIELD_REF_SIZE;

			/* The pointer must not be zero. */
			ut_ad(memcmp(rec_data, field_ref_zero,
				     BTR_EXTERN_FIELD_REF_SIZE));
			/* The BLOB must be owned. */
			ut_ad(!(rec_data[BTR_EXTERN_LEN]
				& BTR_EXTERN_OWNER_FLAG));
		}
#endif /* UNIV_DEBUG */

		len = dfield_get_len(dfield);
		ut_a(len != UNIV_SQL_NULL);
		ut_a(len >= BTR_EXTERN_FIELD_REF_SIZE);
1745 1746
		data = dfield_get_data(dfield);
		data += len - BTR_EXTERN_FIELD_REF_SIZE;
1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762
		/* The pointer must not be zero. */
		ut_a(memcmp(data, field_ref_zero, BTR_EXTERN_FIELD_REF_SIZE));
		/* The BLOB must be owned. */
		ut_a(!(data[BTR_EXTERN_LEN] & BTR_EXTERN_OWNER_FLAG));

		data[BTR_EXTERN_LEN] |= BTR_EXTERN_INHERITED_FLAG;
		/* The BTR_EXTERN_INHERITED_FLAG only matters in
		rollback. Purge will always free the extern fields of
		a delete-marked row. */

		inherit = TRUE;
	}

	return(inherit);
}

1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773
/***********************************************************//**
Marks the clustered index record deleted and inserts the updated version
of the record to the index. This function should be used when the ordering
fields of the clustered index record change. This should be quite rare in
database applications.
@return DB_SUCCESS if operation successfully completed, else error
code or DB_LOCK_WAIT */
static
ulint
row_upd_clust_rec_by_insert(
/*========================*/
1774
	upd_node_t*	node,	/*!< in/out: row update node */
1775 1776 1777 1778
	dict_index_t*	index,	/*!< in: clustered index of the record */
	que_thr_t*	thr,	/*!< in: query thread */
	ibool		check_ref,/*!< in: TRUE if index may be referenced in
				a foreign key constraint */
1779
	mtr_t*		mtr)	/*!< in/out: mtr; gets committed here */
1780
{
1781
	mem_heap_t*	heap;
1782 1783 1784 1785 1786 1787
	btr_pcur_t*	pcur;
	btr_cur_t*	btr_cur;
	trx_t*		trx;
	dict_table_t*	table;
	dtuple_t*	entry;
	ulint		err;
1788 1789 1790
	ibool		change_ownership	= FALSE;
	rec_t*		rec;
	ulint*		offsets			= NULL;
1791 1792 1793 1794 1795 1796 1797 1798 1799

	ut_ad(node);
	ut_ad(dict_index_is_clust(index));

	trx = thr_get_trx(thr);
	table = node->table;
	pcur = node->pcur;
	btr_cur	= btr_pcur_get_btr_cur(pcur);

1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831
	heap = mem_heap_create(1000);

	entry = row_build_index_entry(node->upd_row, node->upd_ext,
				      index, heap);
	ut_a(entry);

	row_upd_index_entry_sys_field(entry, index, DATA_TRX_ID, trx->id);

	switch (node->state) {
	default:
		ut_error;
	case UPD_NODE_INSERT_BLOB:
		/* A lock wait occurred in row_ins_index_entry() in
		the previous invocation of this function. Mark the
		off-page columns in the entry inherited. */

		change_ownership = row_upd_clust_rec_by_insert_inherit(
			NULL, NULL, entry, node->update);
		ut_a(change_ownership);
		/* fall through */
	case UPD_NODE_INSERT_CLUSTERED:
		/* A lock wait occurred in row_ins_index_entry() in
		the previous invocation of this function. */
		break;
	case UPD_NODE_UPDATE_CLUSTERED:
		/* This is the first invocation of the function where
		we update the primary key.  Delete-mark the old record
		in the clustered index and prepare to insert a new entry. */
		rec = btr_cur_get_rec(btr_cur);
		offsets = rec_get_offsets(rec, index, NULL,
					  ULINT_UNDEFINED, &heap);
		ut_ad(page_rec_is_user_rec(rec));
1832

1833 1834 1835
		err = btr_cur_del_mark_set_clust_rec(
			BTR_NO_LOCKING_FLAG, btr_cur_get_block(btr_cur),
			rec, index, offsets, TRUE, thr, mtr);
1836
		if (err != DB_SUCCESS) {
1837
err_exit:
1838
			mtr_commit(mtr);
1839
			mem_heap_free(heap);
1840 1841 1842
			return(err);
		}

1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
		/* If the the new row inherits externally stored
		fields (off-page columns a.k.a. BLOBs) from the
		delete-marked old record, mark them disowned by the
		old record and owned by the new entry. */

		if (rec_offs_any_extern(offsets)) {
			change_ownership = row_upd_clust_rec_by_insert_inherit(
				rec, offsets, entry, node->update);

			if (change_ownership) {
				btr_pcur_store_position(pcur, mtr);
			}
		}
1856 1857 1858 1859 1860 1861 1862

		if (check_ref) {
			/* NOTE that the following call loses
			the position of pcur ! */
			err = row_upd_check_references_constraints(
				node, pcur, table, index, offsets, thr, mtr);
			if (err != DB_SUCCESS) {
1863
				goto err_exit;
1864 1865 1866 1867 1868 1869
			}
		}
	}

	mtr_commit(mtr);

1870 1871 1872 1873 1874 1875
	err = row_ins_index_entry(index, entry,
				  node->upd_ext ? node->upd_ext->n_ext : 0,
				  TRUE, thr);
	node->state = change_ownership
		? UPD_NODE_INSERT_BLOB
		: UPD_NODE_INSERT_CLUSTERED;
1876

1877 1878
	if (err == DB_SUCCESS && change_ownership) {
		/* Mark the non-updated fields disowned by the old record. */
1879

1880 1881 1882 1883 1884 1885 1886
		/* NOTE: this transaction has an x-lock on the record
		and therefore other transactions cannot modify the
		record when we have no latch on the page. In addition,
		we assume that other query threads of the same
		transaction do not modify the record in the meantime.
		Therefore we can assert that the restoration of the
		cursor succeeds. */
1887

1888
		mtr_start(mtr);
1889

1890 1891 1892
		if (!btr_pcur_restore_position(BTR_MODIFY_LEAF, pcur, mtr)) {
			ut_error;
		}
1893

1894 1895 1896 1897
		rec = btr_cur_get_rec(btr_cur);
		offsets = rec_get_offsets(rec, index, offsets,
					  ULINT_UNDEFINED, &heap);
		ut_ad(page_rec_is_user_rec(rec));
1898

1899 1900 1901 1902 1903
		btr_cur_disown_inherited_fields(
			btr_cur_get_page_zip(btr_cur),
			rec, index, offsets, node->update, mtr);

		mtr_commit(mtr);
1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
	}

	mem_heap_free(heap);

	return(err);
}

/***********************************************************//**
Updates a clustered index record of a row when the ordering fields do
not change.
@return DB_SUCCESS if operation successfully completed, else error
code or DB_LOCK_WAIT */
static
ulint
row_upd_clust_rec(
/*==============*/
	upd_node_t*	node,	/*!< in: row update node */
	dict_index_t*	index,	/*!< in: clustered index */
	que_thr_t*	thr,	/*!< in: query thread */
	mtr_t*		mtr)	/*!< in: mtr; gets committed here */
{
	mem_heap_t*	heap	= NULL;
	big_rec_t*	big_rec	= NULL;
	btr_pcur_t*	pcur;
	btr_cur_t*	btr_cur;
	ulint		err;

	ut_ad(node);
	ut_ad(dict_index_is_clust(index));

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

	ut_ad(!rec_get_deleted_flag(btr_pcur_get_rec(pcur),
				    dict_table_is_comp(index->table)));

	/* Try optimistic updating of the record, keeping changes within
	the page; we do not check locks because we assume the x-lock on the
	record to update */

	if (node->cmpl_info & UPD_NODE_NO_SIZE_CHANGE) {
		err = btr_cur_update_in_place(BTR_NO_LOCKING_FLAG,
					      btr_cur, node->update,
					      node->cmpl_info, thr, mtr);
	} else {
		err = btr_cur_optimistic_update(BTR_NO_LOCKING_FLAG,
						btr_cur, node->update,
						node->cmpl_info, thr, mtr);
	}

	mtr_commit(mtr);

	if (UNIV_LIKELY(err == DB_SUCCESS)) {

		return(DB_SUCCESS);
	}

	if (buf_LRU_buf_pool_running_out()) {

		return(DB_LOCK_TABLE_FULL);
	}
	/* We may have to modify the tree structure: do a pessimistic descent
	down the index tree */

	mtr_start(mtr);

	/* NOTE: this transaction has an s-lock or x-lock on the record and
	therefore other transactions cannot modify the record when we have no
	latch on the page. In addition, we assume that other query threads of
	the same transaction do not modify the record in the meantime.
	Therefore we can assert that the restoration of the cursor succeeds. */

	ut_a(btr_pcur_restore_position(BTR_MODIFY_TREE, pcur, mtr));

	ut_ad(!rec_get_deleted_flag(btr_pcur_get_rec(pcur),
				    dict_table_is_comp(index->table)));

1981 1982 1983 1984 1985 1986
	err = btr_cur_pessimistic_update(
		BTR_NO_LOCKING_FLAG | BTR_KEEP_POS_FLAG, btr_cur,
		&heap, &big_rec, node->update, node->cmpl_info, thr, mtr);
	if (big_rec) {
		ulint	offsets_[REC_OFFS_NORMAL_SIZE];
		rec_t*	rec;
1987 1988
		rec_offs_init(offsets_);

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
		ut_a(err == DB_SUCCESS);
		/* Write out the externally stored
		columns while still x-latching
		index->lock and block->lock. Allocate
		pages for big_rec in the mtr that
		modified the B-tree, but be sure to skip
		any pages that were freed in mtr. We will
		write out the big_rec pages before
		committing the B-tree mini-transaction. If
		the system crashes so that crash recovery
		will not replay the mtr_commit(&mtr), the
		big_rec pages will be left orphaned until
		the pages are allocated for something else.

		TODO: If the allocation extends the tablespace, it
		will not be redo logged, in either mini-transaction.
		Tablespace extension should be redo-logged in the
		big_rec mini-transaction, so that recovery will not
		fail when the big_rec was written to the extended
		portion of the file, in case the file was somehow
		truncated in the crash. */
2010

2011
		rec = btr_cur_get_rec(btr_cur);
2012
		DEBUG_SYNC_C("before_row_upd_extern");
2013 2014 2015 2016
		err = btr_store_big_rec_extern_fields(
			index, btr_cur_get_block(btr_cur), rec,
			rec_get_offsets(rec, index, offsets_,
					ULINT_UNDEFINED, &heap),
2017
			big_rec, mtr, BTR_STORE_UPDATE);
2018
		DEBUG_SYNC_C("after_row_upd_extern");
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032
		/* If writing big_rec fails (for example, because of
		DB_OUT_OF_FILE_SPACE), the record will be corrupted.
		Even if we did not update any externally stored
		columns, our update could cause the record to grow so
		that a non-updated column was selected for external
		storage. This non-update would not have been written
		to the undo log, and thus the record cannot be rolled
		back.

		However, because we have not executed mtr_commit(mtr)
		yet, the update will not be replayed in crash
		recovery, and the following assertion failure will
		effectively "roll back" the operation. */
		ut_a(err == DB_SUCCESS);
2033 2034
	}

2035 2036
	mtr_commit(mtr);

2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082
	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}

	if (big_rec) {
		dtuple_big_rec_free(big_rec);
	}

	return(err);
}

/***********************************************************//**
Delete marks a clustered index record.
@return	DB_SUCCESS if operation successfully completed, else error code */
static
ulint
row_upd_del_mark_clust_rec(
/*=======================*/
	upd_node_t*	node,	/*!< in: row update node */
	dict_index_t*	index,	/*!< in: clustered index */
	ulint*		offsets,/*!< in/out: rec_get_offsets() for the
				record under the cursor */
	que_thr_t*	thr,	/*!< in: query thread */
	ibool		check_ref,/*!< in: TRUE if index may be referenced in
				a foreign key constraint */
	mtr_t*		mtr)	/*!< in: mtr; gets committed here */
{
	btr_pcur_t*	pcur;
	btr_cur_t*	btr_cur;
	ulint		err;

	ut_ad(node);
	ut_ad(dict_index_is_clust(index));
	ut_ad(node->is_delete);

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

	/* Store row because we have to build also the secondary index
	entries */

	row_upd_store_row(node);

	/* Mark the clustered index record deleted; we do not have to check
	locks, because we assume that we have an x-lock on the record */

2083 2084 2085
	err = btr_cur_del_mark_set_clust_rec(
		BTR_NO_LOCKING_FLAG, btr_cur_get_block(btr_cur),
		btr_cur_get_rec(btr_cur), index, offsets, TRUE, thr, mtr);
2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
	if (err == DB_SUCCESS && check_ref) {
		/* NOTE that the following call loses the position of pcur ! */

		err = row_upd_check_references_constraints(node,
							   pcur, index->table,
							   index, offsets,
							   thr, mtr);
	}

	mtr_commit(mtr);

	return(err);
}

/***********************************************************//**
Updates the clustered index record.
@return DB_SUCCESS if operation successfully completed, DB_LOCK_WAIT
in case of a lock wait, else error code */
static
ulint
row_upd_clust_step(
/*===============*/
	upd_node_t*	node,	/*!< in: row update node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	dict_index_t*	index;
	btr_pcur_t*	pcur;
	ibool		success;
	ibool		check_ref;
	ulint		err;
	mtr_t*		mtr;
	mtr_t		mtr_buf;
	rec_t*		rec;
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets;
	rec_offs_init(offsets_);

	index = dict_table_get_first_index(node->table);

	check_ref = row_upd_index_is_referenced(index, thr_get_trx(thr));

	pcur = node->pcur;

	/* We have to restore the cursor to its position */
	mtr = &mtr_buf;

	mtr_start(mtr);

	/* If the restoration does not succeed, then the same
	transaction has deleted the record on which the cursor was,
	and that is an SQL error. If the restoration succeeds, it may
	still be that the same transaction has successively deleted
	and inserted a record with the same ordering fields, but in
	that case we know that the transaction has at least an
	implicit x-lock on the record. */

	ut_a(pcur->rel_pos == BTR_PCUR_ON);

	success = btr_pcur_restore_position(BTR_MODIFY_LEAF, pcur, mtr);

	if (!success) {
		err = DB_RECORD_NOT_FOUND;

		mtr_commit(mtr);

		return(err);
	}

	/* If this is a row in SYS_INDEXES table of the data dictionary,
	then we have to free the file segments of the index tree associated
	with the index */

	if (node->is_delete
	    && ut_dulint_cmp(node->table->id, DICT_INDEXES_ID) == 0) {

		dict_drop_index_tree(btr_pcur_get_rec(pcur), mtr);

		mtr_commit(mtr);

		mtr_start(mtr);

		success = btr_pcur_restore_position(BTR_MODIFY_LEAF, pcur,
						    mtr);
		if (!success) {
			err = DB_ERROR;

			mtr_commit(mtr);

			return(err);
		}
	}

	rec = btr_pcur_get_rec(pcur);
	offsets = rec_get_offsets(rec, index, offsets_,
				  ULINT_UNDEFINED, &heap);

	if (!node->has_clust_rec_x_lock) {
		err = lock_clust_rec_modify_check_and_lock(
			0, btr_pcur_get_block(pcur),
			rec, index, offsets, thr);
		if (err != DB_SUCCESS) {
			mtr_commit(mtr);
			goto exit_func;
		}
	}

	/* NOTE: the following function calls will also commit mtr */

	if (node->is_delete) {
		err = row_upd_del_mark_clust_rec(node, index, offsets,
						 thr, check_ref, mtr);
		if (err == DB_SUCCESS) {
			node->state = UPD_NODE_UPDATE_ALL_SEC;
			node->index = dict_table_get_next_index(index);
		}
exit_func:
		if (UNIV_LIKELY_NULL(heap)) {
			mem_heap_free(heap);
		}
		return(err);
	}

	/* If the update is made for MySQL, we already have the update vector
	ready, else we have to do some evaluation: */

	if (UNIV_UNLIKELY(!node->in_mysql_interface)) {
		/* Copy the necessary columns from clust_rec and calculate the
		new values to set */
		row_upd_copy_columns(rec, offsets,
				     UT_LIST_GET_FIRST(node->columns));
		row_upd_eval_new_vals(node->update);
	}

	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}

	if (node->cmpl_info & UPD_NODE_NO_ORD_CHANGE) {

		err = row_upd_clust_rec(node, index, thr, mtr);
		return(err);
	}

	row_upd_store_row(node);

2232 2233
	if (row_upd_changes_ord_field_binary(index, node->update, thr,
					     node->row, node->ext)) {
2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301

		/* Update causes an ordering field (ordering fields within
		the B-tree) of the clustered index record to change: perform
		the update by delete marking and inserting.

		TODO! What to do to the 'Halloween problem', where an update
		moves the record forward in index so that it is again
		updated when the cursor arrives there? Solution: the
		read operation must check the undo record undo number when
		choosing records to update. MySQL solves now the problem
		externally! */

		err = row_upd_clust_rec_by_insert(node, index, thr, check_ref,
						  mtr);
		if (err != DB_SUCCESS) {

			return(err);
		}

		node->state = UPD_NODE_UPDATE_ALL_SEC;
	} else {
		err = row_upd_clust_rec(node, index, thr, mtr);

		if (err != DB_SUCCESS) {

			return(err);
		}

		node->state = UPD_NODE_UPDATE_SOME_SEC;
	}

	node->index = dict_table_get_next_index(index);

	return(err);
}

/***********************************************************//**
Updates the affected index records of a row. When the control is transferred
to this node, we assume that we have a persistent cursor which was on a
record, and the position of the cursor is stored in the cursor.
@return DB_SUCCESS if operation successfully completed, else error
code or DB_LOCK_WAIT */
static
ulint
row_upd(
/*====*/
	upd_node_t*	node,	/*!< in: row update node */
	que_thr_t*	thr)	/*!< in: query thread */
{
	ulint	err	= DB_SUCCESS;

	ut_ad(node && thr);

	if (UNIV_LIKELY(node->in_mysql_interface)) {

		/* We do not get the cmpl_info value from the MySQL
		interpreter: we must calculate it on the fly: */

		if (node->is_delete
		    || row_upd_changes_some_index_ord_field_binary(
			    node->table, node->update)) {
			node->cmpl_info = 0;
		} else {
			node->cmpl_info = UPD_NODE_NO_ORD_CHANGE;
		}
	}

	if (node->state == UPD_NODE_UPDATE_CLUSTERED
2302 2303
	    || node->state == UPD_NODE_INSERT_CLUSTERED
	    || node->state == UPD_NODE_INSERT_BLOB) {
2304

2305
		log_free_check();
2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
		err = row_upd_clust_step(node, thr);

		if (err != DB_SUCCESS) {

			goto function_exit;
		}
	}

	if (!node->is_delete && (node->cmpl_info & UPD_NODE_NO_ORD_CHANGE)) {

		goto function_exit;
	}

	while (node->index != NULL) {
2320 2321

		log_free_check();
2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464
		err = row_upd_sec_step(node, thr);

		if (err != DB_SUCCESS) {

			goto function_exit;
		}

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

function_exit:
	if (err == DB_SUCCESS) {
		/* Do some cleanup */

		if (node->row != NULL) {
			node->row = NULL;
			node->ext = NULL;
			node->upd_row = NULL;
			node->upd_ext = NULL;
			mem_heap_empty(node->heap);
		}

		node->state = UPD_NODE_UPDATE_CLUSTERED;
	}

	return(err);
}

/***********************************************************//**
Updates a row in a table. This is a high-level function used in SQL execution
graphs.
@return	query thread to run next or NULL */
UNIV_INTERN
que_thr_t*
row_upd_step(
/*=========*/
	que_thr_t*	thr)	/*!< in: query thread */
{
	upd_node_t*	node;
	sel_node_t*	sel_node;
	que_node_t*	parent;
	ulint		err		= DB_SUCCESS;
	trx_t*		trx;

	ut_ad(thr);

	trx = thr_get_trx(thr);

	trx_start_if_not_started(trx);

	node = thr->run_node;

	sel_node = node->select;

	parent = que_node_get_parent(node);

	ut_ad(que_node_get_type(node) == QUE_NODE_UPDATE);

	if (thr->prev_node == parent) {
		node->state = UPD_NODE_SET_IX_LOCK;
	}

	if (node->state == UPD_NODE_SET_IX_LOCK) {

		if (!node->has_clust_rec_x_lock) {
			/* It may be that the current session has not yet
			started its transaction, or it has been committed: */

			err = lock_table(0, node->table, LOCK_IX, thr);

			if (err != DB_SUCCESS) {

				goto error_handling;
			}
		}

		node->state = UPD_NODE_UPDATE_CLUSTERED;

		if (node->searched_update) {
			/* Reset the cursor */
			sel_node->state = SEL_NODE_OPEN;

			/* Fetch a row to update */

			thr->run_node = sel_node;

			return(thr);
		}
	}

	/* sel_node is NULL if we are in the MySQL interface */

	if (sel_node && (sel_node->state != SEL_NODE_FETCH)) {

		if (!node->searched_update) {
			/* An explicit cursor should be positioned on a row
			to update */

			ut_error;

			err = DB_ERROR;

			goto error_handling;
		}

		ut_ad(sel_node->state == SEL_NODE_NO_MORE_ROWS);

		/* No more rows to update, or the select node performed the
		updates directly in-place */

		thr->run_node = parent;

		return(thr);
	}

	/* DO THE CHECKS OF THE CONSISTENCY CONSTRAINTS HERE */

	err = row_upd(node, thr);

error_handling:
	trx->error_state = err;

	if (err != DB_SUCCESS) {
		return(NULL);
	}

	/* DO THE TRIGGER ACTIONS HERE */

	if (node->searched_update) {
		/* Fetch next row to update */

		thr->run_node = sel_node;
	} else {
		/* It was an explicit cursor update */

		thr->run_node = parent;
	}

	node->state = UPD_NODE_UPDATE_CLUSTERED;

	return(thr);
}
#endif /* !UNIV_HOTBACKUP */