btr0sea.c 48.3 KB
Newer Older
1 2
/*****************************************************************************

3
Copyright (c) 1996, 2011, Oracle and/or its affiliates. All Rights Reserved.
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
Copyright (c) 2008, Google Inc.

Portions of this file contain modifications contributed and copyrighted by
Google, Inc. Those modifications are gratefully acknowledged and are described
briefly in the InnoDB documentation. The contributions by Google are
incorporated with their permission, and subject to the conditions contained in
the file COPYING.Google.

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., 59 Temple
Place, Suite 330, Boston, MA 02111-1307 USA

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

/********************************************************************//**
@file btr/btr0sea.c
The index tree adaptive search

Created 2/17/1996 Heikki Tuuri
*************************************************************************/

#include "btr0sea.h"
#ifdef UNIV_NONINL
#include "btr0sea.ic"
#endif

#include "buf0buf.h"
#include "page0page.h"
#include "page0cur.h"
#include "btr0cur.h"
#include "btr0pcur.h"
#include "btr0btr.h"
#include "ha0ha.h"

/** Flag: has the search system been enabled?
47
Protected by btr_search_latch. */
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
UNIV_INTERN char		btr_search_enabled	= TRUE;

/** A dummy variable to fool the compiler */
UNIV_INTERN ulint		btr_search_this_is_zero = 0;

#ifdef UNIV_SEARCH_PERF_STAT
/** Number of successful adaptive hash index lookups */
UNIV_INTERN ulint		btr_search_n_succ	= 0;
/** Number of failed adaptive hash index lookups */
UNIV_INTERN ulint		btr_search_n_hash_fail	= 0;
#endif /* UNIV_SEARCH_PERF_STAT */

/** padding to prevent other memory update
hotspots from residing on the same memory
cache line as btr_search_latch */
UNIV_INTERN byte		btr_sea_pad1[64];

/** The latch protecting the adaptive search system: this latch protects the
(1) positions of records on those pages where a hash index has been built.
NOTE: It does not protect values of non-ordering fields within a record from
being updated in-place! We can use fact (1) to perform unique searches to
indexes. */

/* We will allocate the latch from dynamic memory to get it to the
same DRAM page as other hotspot semaphores */
UNIV_INTERN rw_lock_t*		btr_search_latch_temp;

/** padding to prevent other memory update hotspots from residing on
the same memory cache line */
UNIV_INTERN byte		btr_sea_pad2[64];

/** The adaptive hash index */
UNIV_INTERN btr_search_sys_t*	btr_search_sys;

/** If the number of records on the page divided by this parameter
would have been successfully accessed using a hash index, the index
is then built on the page, assuming the global limit has been reached */
#define BTR_SEARCH_PAGE_BUILD_LIMIT	16

/** The global limit for consecutive potentially successful hash searches,
before hash index building is started */
#define BTR_SEARCH_BUILD_LIMIT		100

/********************************************************************//**
Builds a hash index on a page with the given parameters. If the page already
has a hash index with different parameters, the old hash index is removed.
If index is non-NULL, this function checks if n_fields and n_bytes are
sensible values, and does not build a hash index if not. */
static
void
btr_search_build_page_hash_index(
/*=============================*/
	dict_index_t*	index,	/*!< in: index for which to build, or NULL if
				not known */
	buf_block_t*	block,	/*!< in: index page, s- or x-latched */
	ulint		n_fields,/*!< in: hash this many full fields */
	ulint		n_bytes,/*!< in: hash this many bytes from the next
				field */
	ibool		left_side);/*!< in: hash for searches from left side? */

/*****************************************************************//**
This function should be called before reserving any btr search mutex, if
the intended operation might add nodes to the search system hash table.
Because of the latching order, once we have reserved the btr search system
latch, we cannot allocate a free frame from the buffer pool. Checks that
there is a free buffer frame allocated for hash table heap in the btr search
system. If not, allocates a free frames for the heap. This check makes it
probable that, when have reserved the btr search system latch and we need to
allocate a new node to the hash table, it will succeed. However, the check
will not guarantee success. */
static
void
btr_search_check_free_space_in_heap(void)
/*=====================================*/
{
	hash_table_t*	table;
	mem_heap_t*	heap;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_SHARED));
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

	table = btr_search_sys->hash_index;

	heap = table->heap;

	/* Note that we peek the value of heap->free_block without reserving
	the latch: this is ok, because we will not guarantee that there will
	be enough free space in the hash table. */

	if (heap->free_block == NULL) {
140
		buf_block_t*	block = buf_block_alloc();
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

		rw_lock_x_lock(&btr_search_latch);

		if (heap->free_block == NULL) {
			heap->free_block = block;
		} else {
			buf_block_free(block);
		}

		rw_lock_x_unlock(&btr_search_latch);
	}
}

/*****************************************************************//**
Creates and initializes the adaptive search system at a database start. */
UNIV_INTERN
void
btr_search_sys_create(
/*==================*/
	ulint	hash_size)	/*!< in: hash index hash table size */
{
	/* We allocate the search latch from dynamic memory:
	see above at the global variable definition */

	btr_search_latch_temp = mem_alloc(sizeof(rw_lock_t));

	rw_lock_create(&btr_search_latch, SYNC_SEARCH_SYS);

	btr_search_sys = mem_alloc(sizeof(btr_search_sys_t));

	btr_search_sys->hash_index = ha_create(hash_size, 0, 0);
}

174 175 176 177 178 179 180
/*****************************************************************//**
Frees the adaptive search system at a database shutdown. */
UNIV_INTERN
void
btr_search_sys_free(void)
/*=====================*/
{
181
	rw_lock_free(&btr_search_latch);
182 183 184 185 186 187 188 189
	mem_free(btr_search_latch_temp);
	btr_search_latch_temp = NULL;
	mem_heap_free(btr_search_sys->hash_index->heap);
	hash_table_free(btr_search_sys->hash_index);
	mem_free(btr_search_sys);
	btr_search_sys = NULL;
}

190 191 192 193 194 195 196
/********************************************************************//**
Disable the adaptive hash search system and empty the index. */
UNIV_INTERN
void
btr_search_disable(void)
/*====================*/
{
197 198 199
	dict_table_t*	table;

	mutex_enter(&dict_sys->mutex);
200 201 202 203
	rw_lock_x_lock(&btr_search_latch);

	btr_search_enabled = FALSE;

204 205 206 207 208 209 210 211 212 213 214 215 216 217 218
	/* Clear the index->search_info->ref_count of every index in
	the data dictionary cache. */
	for (table = UT_LIST_GET_FIRST(dict_sys->table_LRU); table;
	     table = UT_LIST_GET_NEXT(table_LRU, table)) {

		dict_index_t*	index;

		for (index = dict_table_get_first_index(table); index;
		     index = dict_table_get_next_index(index)) {

			index->search_info->ref_count = 0;
		}
	}

	mutex_exit(&dict_sys->mutex);
219

220 221
	/* Set all block->index = NULL. */
	buf_pool_clear_hash_index();
222

223 224 225
	/* Clear the adaptive hash index. */
	hash_table_clear(btr_search_sys->hash_index);
	mem_heap_empty(btr_search_sys->hash_index->heap);
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

	rw_lock_x_unlock(&btr_search_latch);
}

/********************************************************************//**
Enable the adaptive hash search system. */
UNIV_INTERN
void
btr_search_enable(void)
/*====================*/
{
	rw_lock_x_lock(&btr_search_latch);

	btr_search_enabled = TRUE;

	rw_lock_x_unlock(&btr_search_latch);
}

/*****************************************************************//**
Creates and initializes a search info struct.
@return	own: search info struct */
UNIV_INTERN
btr_search_t*
btr_search_info_create(
/*===================*/
	mem_heap_t*	heap)	/*!< in: heap where created */
{
	btr_search_t*	info;

	info = mem_heap_alloc(heap, sizeof(btr_search_t));

#ifdef UNIV_DEBUG
	info->magic_n = BTR_SEARCH_MAGIC_N;
#endif /* UNIV_DEBUG */

	info->ref_count = 0;
	info->root_guess = NULL;

	info->hash_analysis = 0;
	info->n_hash_potential = 0;

	info->last_hash_succ = FALSE;

#ifdef UNIV_SEARCH_PERF_STAT
	info->n_hash_succ = 0;
	info->n_hash_fail = 0;
	info->n_patt_succ = 0;
	info->n_searches = 0;
#endif /* UNIV_SEARCH_PERF_STAT */

	/* Set some sensible values */
	info->n_fields = 1;
	info->n_bytes = 0;

	info->left_side = TRUE;

	return(info);
}

/*****************************************************************//**
Returns the value of ref_count. The value is protected by
btr_search_latch.
@return	ref_count value. */
UNIV_INTERN
ulint
btr_search_info_get_ref_count(
/*==========================*/
	btr_search_t*   info)	/*!< in: search info. */
{
	ulint ret;

	ut_ad(info);

#ifdef UNIV_SYNC_DEBUG
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_SHARED));
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

	rw_lock_s_lock(&btr_search_latch);
	ret = info->ref_count;
	rw_lock_s_unlock(&btr_search_latch);

	return(ret);
}

/*********************************************************************//**
Updates the search info of an index about hash successes. NOTE that info
is NOT protected by any semaphore, to save CPU time! Do not assume its fields
are consistent. */
static
void
btr_search_info_update_hash(
/*========================*/
	btr_search_t*	info,	/*!< in/out: search info */
	btr_cur_t*	cursor)	/*!< in: cursor which was just positioned */
{
	dict_index_t*	index;
	ulint		n_unique;
	int		cmp;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_SHARED));
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

	index = cursor->index;

	if (dict_index_is_ibuf(index)) {
		/* So many deletes are performed on an insert buffer tree
		that we do not consider a hash index useful on it: */

		return;
	}

	n_unique = dict_index_get_n_unique_in_tree(index);

	if (info->n_hash_potential == 0) {

		goto set_new_recomm;
	}

	/* Test if the search would have succeeded using the recommended
	hash prefix */

	if (info->n_fields >= n_unique && cursor->up_match >= n_unique) {
increment_potential:
		info->n_hash_potential++;

		return;
	}

	cmp = ut_pair_cmp(info->n_fields, info->n_bytes,
			  cursor->low_match, cursor->low_bytes);

	if (info->left_side ? cmp <= 0 : cmp > 0) {

		goto set_new_recomm;
	}

	cmp = ut_pair_cmp(info->n_fields, info->n_bytes,
			  cursor->up_match, cursor->up_bytes);

	if (info->left_side ? cmp <= 0 : cmp > 0) {

		goto increment_potential;
	}

set_new_recomm:
	/* We have to set a new recommendation; skip the hash analysis
	for a while to avoid unnecessary CPU time usage when there is no
	chance for success */

	info->hash_analysis = 0;

	cmp = ut_pair_cmp(cursor->up_match, cursor->up_bytes,
			  cursor->low_match, cursor->low_bytes);
	if (cmp == 0) {
		info->n_hash_potential = 0;

		/* For extra safety, we set some sensible values here */

		info->n_fields = 1;
		info->n_bytes = 0;

		info->left_side = TRUE;

	} else if (cmp > 0) {
		info->n_hash_potential = 1;

		if (cursor->up_match >= n_unique) {

			info->n_fields = n_unique;
			info->n_bytes = 0;

		} else if (cursor->low_match < cursor->up_match) {

			info->n_fields = cursor->low_match + 1;
			info->n_bytes = 0;
		} else {
			info->n_fields = cursor->low_match;
			info->n_bytes = cursor->low_bytes + 1;
		}

		info->left_side = TRUE;
	} else {
		info->n_hash_potential = 1;

		if (cursor->low_match >= n_unique) {

			info->n_fields = n_unique;
			info->n_bytes = 0;

		} else if (cursor->low_match > cursor->up_match) {

			info->n_fields = cursor->up_match + 1;
			info->n_bytes = 0;
		} else {
			info->n_fields = cursor->up_match;
			info->n_bytes = cursor->up_bytes + 1;
		}

		info->left_side = FALSE;
	}
}

/*********************************************************************//**
Updates the block search info on hash successes. NOTE that info and
block->n_hash_helps, n_fields, n_bytes, side are NOT protected by any
semaphore, to save CPU time! Do not assume the fields are consistent.
@return	TRUE if building a (new) hash index on the block is recommended */
static
ibool
btr_search_update_block_hash_info(
/*==============================*/
	btr_search_t*	info,	/*!< in: search info */
	buf_block_t*	block,	/*!< in: buffer block */
	btr_cur_t*	cursor __attribute__((unused)))
				/*!< in: cursor */
{
#ifdef UNIV_SYNC_DEBUG
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_SHARED));
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_EX));
	ut_ad(rw_lock_own(&block->lock, RW_LOCK_SHARED)
	      || rw_lock_own(&block->lock, RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */
	ut_ad(cursor);

	info->last_hash_succ = FALSE;

	ut_a(buf_block_state_valid(block));
	ut_ad(info->magic_n == BTR_SEARCH_MAGIC_N);

	if ((block->n_hash_helps > 0)
	    && (info->n_hash_potential > 0)
	    && (block->n_fields == info->n_fields)
	    && (block->n_bytes == info->n_bytes)
	    && (block->left_side == info->left_side)) {

464
		if ((block->index)
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
		    && (block->curr_n_fields == info->n_fields)
		    && (block->curr_n_bytes == info->n_bytes)
		    && (block->curr_left_side == info->left_side)) {

			/* The search would presumably have succeeded using
			the hash index */

			info->last_hash_succ = TRUE;
		}

		block->n_hash_helps++;
	} else {
		block->n_hash_helps = 1;
		block->n_fields = info->n_fields;
		block->n_bytes = info->n_bytes;
		block->left_side = info->left_side;
	}

#ifdef UNIV_DEBUG
	if (cursor->index->table->does_not_fit_in_memory) {
		block->n_hash_helps = 0;
	}
#endif /* UNIV_DEBUG */

	if ((block->n_hash_helps > page_get_n_recs(block->frame)
	     / BTR_SEARCH_PAGE_BUILD_LIMIT)
	    && (info->n_hash_potential >= BTR_SEARCH_BUILD_LIMIT)) {

493
		if ((!block->index)
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
		    || (block->n_hash_helps
			> 2 * page_get_n_recs(block->frame))
		    || (block->n_fields != block->curr_n_fields)
		    || (block->n_bytes != block->curr_n_bytes)
		    || (block->left_side != block->curr_left_side)) {

			/* Build a new hash index on the page */

			return(TRUE);
		}
	}

	return(FALSE);
}

/*********************************************************************//**
Updates a hash node reference when it has been unsuccessfully used in a
search which could have succeeded with the used hash parameters. This can
happen because when building a hash index for a page, we do not check
what happens at page boundaries, and therefore there can be misleading
hash nodes. Also, collisions in the fold value can lead to misleading
references. This function lazily fixes these imperfections in the hash
index. */
static
void
btr_search_update_hash_ref(
/*=======================*/
	btr_search_t*	info,	/*!< in: search info */
	buf_block_t*	block,	/*!< in: buffer block where cursor positioned */
	btr_cur_t*	cursor)	/*!< in: cursor */
{
525 526
	dict_index_t*	index;
	ulint		fold;
527
	rec_t*		rec;
528 529 530 531 532 533 534 535 536 537

	ut_ad(cursor->flag == BTR_CUR_HASH_FAIL);
#ifdef UNIV_SYNC_DEBUG
	ut_ad(rw_lock_own(&btr_search_latch, RW_LOCK_EX));
	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_SHARED)
	      || rw_lock_own(&(block->lock), RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */
	ut_ad(page_align(btr_cur_get_rec(cursor))
	      == buf_block_get_frame(block));

538 539 540
	index = block->index;

	if (!index) {
541 542 543 544

		return;
	}

545 546
	ut_a(index == cursor->index);
	ut_a(!dict_index_is_ibuf(index));
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563

	if ((info->n_hash_potential > 0)
	    && (block->curr_n_fields == info->n_fields)
	    && (block->curr_n_bytes == info->n_bytes)
	    && (block->curr_left_side == info->left_side)) {
		mem_heap_t*	heap		= NULL;
		ulint		offsets_[REC_OFFS_NORMAL_SIZE];
		rec_offs_init(offsets_);

		rec = btr_cur_get_rec(cursor);

		if (!page_rec_is_user_rec(rec)) {

			return;
		}

		fold = rec_fold(rec,
564
				rec_get_offsets(rec, index, offsets_,
565 566
						ULINT_UNDEFINED, &heap),
				block->curr_n_fields,
567
				block->curr_n_bytes, index->id);
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
		if (UNIV_LIKELY_NULL(heap)) {
			mem_heap_free(heap);
		}
#ifdef UNIV_SYNC_DEBUG
		ut_ad(rw_lock_own(&btr_search_latch, RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

		ha_insert_for_fold(btr_search_sys->hash_index, fold,
				   block, rec);
	}
}

/*********************************************************************//**
Updates the search info. */
UNIV_INTERN
void
btr_search_info_update_slow(
/*========================*/
	btr_search_t*	info,	/*!< in/out: search info */
	btr_cur_t*	cursor)	/*!< in: cursor which was just positioned */
{
	buf_block_t*	block;
	ibool		build_index;
	ulint*		params;
	ulint*		params2;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_SHARED));
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

	block = btr_cur_get_block(cursor);

	/* NOTE that the following two function calls do NOT protect
	info or block->n_fields etc. with any semaphore, to save CPU time!
	We cannot assume the fields are consistent when we return from
	those functions! */

	btr_search_info_update_hash(info, cursor);

	build_index = btr_search_update_block_hash_info(info, block, cursor);

	if (build_index || (cursor->flag == BTR_CUR_HASH_FAIL)) {

		btr_search_check_free_space_in_heap();
	}

	if (cursor->flag == BTR_CUR_HASH_FAIL) {
		/* Update the hash node reference, if appropriate */

#ifdef UNIV_SEARCH_PERF_STAT
		btr_search_n_hash_fail++;
#endif /* UNIV_SEARCH_PERF_STAT */

		rw_lock_x_lock(&btr_search_latch);

		btr_search_update_hash_ref(info, block, cursor);

		rw_lock_x_unlock(&btr_search_latch);
	}

	if (build_index) {
		/* Note that since we did not protect block->n_fields etc.
		with any semaphore, the values can be inconsistent. We have
		to check inside the function call that they make sense. We
		also malloc an array and store the values there to make sure
		the compiler does not let the function call parameters change
		inside the called function. It might be that the compiler
		would optimize the call just to pass pointers to block. */

		params = mem_alloc(3 * sizeof(ulint));
		params[0] = block->n_fields;
		params[1] = block->n_bytes;
		params[2] = block->left_side;

		/* Make sure the compiler cannot deduce the values and do
		optimizations */

		params2 = params + btr_search_this_is_zero;

		btr_search_build_page_hash_index(cursor->index,
						 block,
						 params2[0],
						 params2[1],
						 params2[2]);
		mem_free(params);
	}
}

/******************************************************************//**
Checks if a guessed position for a tree cursor is right. Note that if
mode is PAGE_CUR_LE, which is used in inserts, and the function returns
TRUE, then cursor->up_match and cursor->low_match both have sensible values.
@return	TRUE if success */
static
ibool
btr_search_check_guess(
/*===================*/
	btr_cur_t*	cursor,	/*!< in: guessed cursor position */
	ibool		can_only_compare_to_cursor_rec,
				/*!< in: if we do not have a latch on the page
				of cursor, but only a latch on
				btr_search_latch, then ONLY the columns
				of the record UNDER the cursor are
				protected, not the next or previous record
				in the chain: we cannot look at the next or
				previous record to check our guess! */
	const dtuple_t*	tuple,	/*!< in: data tuple */
	ulint		mode,	/*!< in: PAGE_CUR_L, PAGE_CUR_LE, PAGE_CUR_G,
				or PAGE_CUR_GE */
	mtr_t*		mtr)	/*!< in: mtr */
{
	rec_t*		rec;
	ulint		n_unique;
	ulint		match;
	ulint		bytes;
	int		cmp;
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets		= offsets_;
	ibool		success		= FALSE;
	rec_offs_init(offsets_);

	n_unique = dict_index_get_n_unique_in_tree(cursor->index);

	rec = btr_cur_get_rec(cursor);

	ut_ad(page_rec_is_user_rec(rec));

	match = 0;
	bytes = 0;

	offsets = rec_get_offsets(rec, cursor->index, offsets,
				  n_unique, &heap);
	cmp = page_cmp_dtuple_rec_with_match(tuple, rec,
					     offsets, &match, &bytes);

	if (mode == PAGE_CUR_GE) {
		if (cmp == 1) {
			goto exit_func;
		}

		cursor->up_match = match;

		if (match >= n_unique) {
			success = TRUE;
			goto exit_func;
		}
	} else if (mode == PAGE_CUR_LE) {
		if (cmp == -1) {
			goto exit_func;
		}

		cursor->low_match = match;

	} else if (mode == PAGE_CUR_G) {
		if (cmp != -1) {
			goto exit_func;
		}
	} else if (mode == PAGE_CUR_L) {
		if (cmp != 1) {
			goto exit_func;
		}
	}

	if (can_only_compare_to_cursor_rec) {
		/* Since we could not determine if our guess is right just by
		looking at the record under the cursor, return FALSE */
		goto exit_func;
	}

	match = 0;
	bytes = 0;

	if ((mode == PAGE_CUR_G) || (mode == PAGE_CUR_GE)) {
		rec_t*	prev_rec;

		ut_ad(!page_rec_is_infimum(rec));

		prev_rec = page_rec_get_prev(rec);

		if (page_rec_is_infimum(prev_rec)) {
			success = btr_page_get_prev(page_align(prev_rec), mtr)
				== FIL_NULL;

			goto exit_func;
		}

		offsets = rec_get_offsets(prev_rec, cursor->index, offsets,
					  n_unique, &heap);
		cmp = page_cmp_dtuple_rec_with_match(tuple, prev_rec,
						     offsets, &match, &bytes);
		if (mode == PAGE_CUR_GE) {
			success = cmp == 1;
		} else {
			success = cmp != -1;
		}

		goto exit_func;
	} else {
		rec_t*	next_rec;

		ut_ad(!page_rec_is_supremum(rec));

		next_rec = page_rec_get_next(rec);

		if (page_rec_is_supremum(next_rec)) {
			if (btr_page_get_next(page_align(next_rec), mtr)
			    == FIL_NULL) {

				cursor->up_match = 0;
				success = TRUE;
			}

			goto exit_func;
		}

		offsets = rec_get_offsets(next_rec, cursor->index, offsets,
					  n_unique, &heap);
		cmp = page_cmp_dtuple_rec_with_match(tuple, next_rec,
						     offsets, &match, &bytes);
		if (mode == PAGE_CUR_LE) {
			success = cmp == -1;
			cursor->up_match = match;
		} else {
			success = cmp != 1;
		}
	}
exit_func:
	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}
	return(success);
}

/******************************************************************//**
Tries to guess the right search position based on the hash search info
of the index. Note that if mode is PAGE_CUR_LE, which is used in inserts,
and the function returns TRUE, then cursor->up_match and cursor->low_match
both have sensible values.
@return	TRUE if succeeded */
UNIV_INTERN
ibool
btr_search_guess_on_hash(
/*=====================*/
	dict_index_t*	index,		/*!< in: index */
	btr_search_t*	info,		/*!< in: index search info */
	const dtuple_t*	tuple,		/*!< in: logical record */
	ulint		mode,		/*!< in: PAGE_CUR_L, ... */
	ulint		latch_mode,	/*!< in: BTR_SEARCH_LEAF, ...;
					NOTE that only if has_search_latch
					is 0, we will have a latch set on
					the cursor page, otherwise we assume
					the caller uses his search latch
					to protect the record! */
	btr_cur_t*	cursor,		/*!< out: tree cursor */
	ulint		has_search_latch,/*!< in: latch mode the caller
					currently has on btr_search_latch:
					RW_S_LATCH, RW_X_LATCH, or 0 */
	mtr_t*		mtr)		/*!< in: mtr */
{
	buf_block_t*	block;
830
	rec_t*		rec;
831 832 833 834 835 836 837
	ulint		fold;
	dulint		index_id;
#ifdef notdefined
	btr_cur_t	cursor2;
	btr_pcur_t	pcur;
#endif
	ut_ad(index && info && tuple && cursor && mtr);
838
	ut_ad(!dict_index_is_ibuf(index));
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
	ut_ad((latch_mode == BTR_SEARCH_LEAF)
	      || (latch_mode == BTR_MODIFY_LEAF));

	/* Note that, for efficiency, the struct info may not be protected by
	any latch here! */

	if (UNIV_UNLIKELY(info->n_hash_potential == 0)) {

		return(FALSE);
	}

	cursor->n_fields = info->n_fields;
	cursor->n_bytes = info->n_bytes;

	if (UNIV_UNLIKELY(dtuple_get_n_fields(tuple)
			  < cursor->n_fields + (cursor->n_bytes > 0))) {

		return(FALSE);
	}

	index_id = index->id;

#ifdef UNIV_SEARCH_PERF_STAT
	info->n_hash_succ++;
#endif
	fold = dtuple_fold(tuple, cursor->n_fields, cursor->n_bytes, index_id);

	cursor->fold = fold;
	cursor->flag = BTR_CUR_HASH;

	if (UNIV_LIKELY(!has_search_latch)) {
		rw_lock_s_lock(&btr_search_latch);

		if (UNIV_UNLIKELY(!btr_search_enabled)) {
			goto failure_unlock;
		}
	}

	ut_ad(rw_lock_get_writer(&btr_search_latch) != RW_LOCK_EX);
	ut_ad(rw_lock_get_reader_count(&btr_search_latch) > 0);

	rec = ha_search_and_get_data(btr_search_sys->hash_index, fold);

	if (UNIV_UNLIKELY(!rec)) {
		goto failure_unlock;
	}

	block = buf_block_align(rec);

	if (UNIV_LIKELY(!has_search_latch)) {

		if (UNIV_UNLIKELY(
			    !buf_page_get_known_nowait(latch_mode, block,
						       BUF_MAKE_YOUNG,
						       __FILE__, __LINE__,
						       mtr))) {
			goto failure_unlock;
		}

		rw_lock_s_unlock(&btr_search_latch);

		buf_block_dbg_add_level(block, SYNC_TREE_NODE_FROM_HASH);
	}

	if (UNIV_UNLIKELY(buf_block_get_state(block) != BUF_BLOCK_FILE_PAGE)) {
		ut_ad(buf_block_get_state(block) == BUF_BLOCK_REMOVE_HASH);

		if (UNIV_LIKELY(!has_search_latch)) {

			btr_leaf_page_release(block, latch_mode, mtr);
		}

		goto failure;
	}

	ut_ad(page_rec_is_user_rec(rec));

916
	btr_cur_position(index, rec, block, cursor);
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

	/* Check the validity of the guess within the page */

	/* If we only have the latch on btr_search_latch, not on the
	page, it only protects the columns of the record the cursor
	is positioned on. We cannot look at the next of the previous
	record to determine if our guess for the cursor position is
	right. */
	if (UNIV_EXPECT
	    (ut_dulint_cmp(index_id, btr_page_get_index_id(block->frame)), 0)
	    || !btr_search_check_guess(cursor,
				       has_search_latch,
				       tuple, mode, mtr)) {
		if (UNIV_LIKELY(!has_search_latch)) {
			btr_leaf_page_release(block, latch_mode, mtr);
		}

		goto failure;
	}

	if (UNIV_LIKELY(info->n_hash_potential < BTR_SEARCH_BUILD_LIMIT + 5)) {

		info->n_hash_potential++;
	}

#ifdef notdefined
	/* These lines of code can be used in a debug version to check
	the correctness of the searched cursor position: */

	info->last_hash_succ = FALSE;

	/* Currently, does not work if the following fails: */
	ut_ad(!has_search_latch);

	btr_leaf_page_release(block, latch_mode, mtr);

	btr_cur_search_to_nth_level(index, 0, tuple, mode, latch_mode,
				    &cursor2, 0, mtr);
	if (mode == PAGE_CUR_GE
	    && page_rec_is_supremum(btr_cur_get_rec(&cursor2))) {

		/* If mode is PAGE_CUR_GE, then the binary search
		in the index tree may actually take us to the supremum
		of the previous page */

		info->last_hash_succ = FALSE;

		btr_pcur_open_on_user_rec(index, tuple, mode, latch_mode,
					  &pcur, mtr);
		ut_ad(btr_pcur_get_rec(&pcur) == btr_cur_get_rec(cursor));
	} else {
		ut_ad(btr_cur_get_rec(&cursor2) == btr_cur_get_rec(cursor));
	}

	/* NOTE that it is theoretically possible that the above assertions
	fail if the page of the cursor gets removed from the buffer pool
	meanwhile! Thus it might not be a bug. */
#endif
	info->last_hash_succ = TRUE;

#ifdef UNIV_SEARCH_PERF_STAT
	btr_search_n_succ++;
#endif
	if (UNIV_LIKELY(!has_search_latch)
	    && buf_page_peek_if_too_old(&block->page)) {

		buf_page_make_young(&block->page);
	}

	/* Increment the page get statistics though we did not really
	fix the page: for user info only */

989
	buf_pool->stat.n_page_gets++;
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

	return(TRUE);

	/*-------------------------------------------*/
failure_unlock:
	if (UNIV_LIKELY(!has_search_latch)) {
		rw_lock_s_unlock(&btr_search_latch);
	}
failure:
	cursor->flag = BTR_CUR_HASH_FAIL;

#ifdef UNIV_SEARCH_PERF_STAT
	info->n_hash_fail++;

	if (info->n_hash_succ > 0) {
		info->n_hash_succ--;
	}
#endif
	info->last_hash_succ = FALSE;

	return(FALSE);
}

/********************************************************************//**
Drops a page hash index. */
UNIV_INTERN
void
btr_search_drop_page_hash_index(
/*============================*/
	buf_block_t*	block)	/*!< in: block containing index page,
				s- or x-latched, or an index page
				for which we know that
				block->buf_fix_count == 0 */
{
	hash_table_t*		table;
	ulint			n_fields;
	ulint			n_bytes;
	const page_t*		page;
	const rec_t*		rec;
	ulint			fold;
	ulint			prev_fold;
	dulint			index_id;
	ulint			n_cached;
	ulint			n_recs;
	ulint*			folds;
	ulint			i;
	mem_heap_t*		heap;
	const dict_index_t*	index;
	ulint*			offsets;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_SHARED));
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

retry:
	rw_lock_s_lock(&btr_search_latch);
1047
	index = block->index;
1048

1049
	if (UNIV_LIKELY(!index)) {
1050 1051 1052 1053 1054 1055

		rw_lock_s_unlock(&btr_search_latch);

		return;
	}

1056
	ut_a(!dict_index_is_ibuf(index));
1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
	table = btr_search_sys->hash_index;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_SHARED)
	      || rw_lock_own(&(block->lock), RW_LOCK_EX)
	      || (block->page.buf_fix_count == 0));
#endif /* UNIV_SYNC_DEBUG */

	n_fields = block->curr_n_fields;
	n_bytes = block->curr_n_bytes;

	/* NOTE: The fields of block must not be accessed after
	releasing btr_search_latch, as the index page might only
	be s-latched! */

	rw_lock_s_unlock(&btr_search_latch);

	ut_a(n_fields + n_bytes > 0);

1076
	page = block->frame;
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
	n_recs = page_get_n_recs(page);

	/* Calculate and cache fold values into an array for fast deletion
	from the hash index */

	folds = mem_alloc(n_recs * sizeof(ulint));

	n_cached = 0;

	rec = page_get_infimum_rec(page);
	rec = page_rec_get_next_low(rec, page_is_comp(page));

	index_id = btr_page_get_index_id(page);

	ut_a(0 == ut_dulint_cmp(index_id, index->id));

	prev_fold = 0;

	heap = NULL;
	offsets = NULL;

	while (!page_rec_is_supremum(rec)) {
		offsets = rec_get_offsets(rec, index, offsets,
					  n_fields + (n_bytes > 0), &heap);
		ut_a(rec_offs_n_fields(offsets) == n_fields + (n_bytes > 0));
		fold = rec_fold(rec, offsets, n_fields, n_bytes, index_id);

		if (fold == prev_fold && prev_fold != 0) {

			goto next_rec;
		}

		/* Remove all hash nodes pointing to this page from the
		hash chain */

		folds[n_cached] = fold;
		n_cached++;
next_rec:
		rec = page_rec_get_next_low(rec, page_rec_is_comp(rec));
		prev_fold = fold;
	}

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

	rw_lock_x_lock(&btr_search_latch);

1125
	if (UNIV_UNLIKELY(!block->index)) {
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
		/* Someone else has meanwhile dropped the hash index */

		goto cleanup;
	}

	ut_a(block->index == index);

	if (UNIV_UNLIKELY(block->curr_n_fields != n_fields)
	    || UNIV_UNLIKELY(block->curr_n_bytes != n_bytes)) {

		/* Someone else has meanwhile built a new hash index on the
		page, with different parameters */

		rw_lock_x_unlock(&btr_search_latch);

		mem_free(folds);
		goto retry;
	}

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

		ha_remove_all_nodes_to_page(table, folds[i], page);
	}

	ut_a(index->search_info->ref_count > 0);
	index->search_info->ref_count--;

	block->index = NULL;
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
cleanup:
#if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
	if (UNIV_UNLIKELY(block->n_pointers)) {
		/* Corruption */
		ut_print_timestamp(stderr);
		fprintf(stderr,
			"  InnoDB: Corruption of adaptive hash index."
			" After dropping\n"
			"InnoDB: the hash index to a page of %s,"
			" still %lu hash nodes remain.\n",
			index->name, (ulong) block->n_pointers);
		rw_lock_x_unlock(&btr_search_latch);

		btr_search_validate();
	} else {
		rw_lock_x_unlock(&btr_search_latch);
	}
#else /* UNIV_AHI_DEBUG || UNIV_DEBUG */
	rw_lock_x_unlock(&btr_search_latch);
#endif /* UNIV_AHI_DEBUG || UNIV_DEBUG */

	mem_free(folds);
}

/********************************************************************//**
1180 1181
Drops a possible page hash index when a page is evicted from the buffer pool
or freed in a file segment. */
1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195
UNIV_INTERN
void
btr_search_drop_page_hash_when_freed(
/*=================================*/
	ulint	space,		/*!< in: space id */
	ulint	zip_size,	/*!< in: compressed page size in bytes
				or 0 for uncompressed pages */
	ulint	page_no)	/*!< in: page number */
{
	buf_block_t*	block;
	mtr_t		mtr;

	mtr_start(&mtr);

1196 1197 1198 1199 1200
	/* If the caller has a latch on the page, then the caller must
	have a x-latch on the page and it must have already dropped
	the hash index for the page. Because of the x-latch that we
	are possibly holding, we cannot s-latch the page, but must
	(recursively) x-latch it, even though we are only reading. */
1201

1202
	block = buf_page_get_gen(space, zip_size, page_no, RW_X_LATCH, NULL,
1203 1204
				 BUF_PEEK_IF_IN_POOL, __FILE__, __LINE__,
				 &mtr);
1205

1206
	if (block && block->index) {
1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258

		buf_block_dbg_add_level(block, SYNC_TREE_NODE_FROM_HASH);

		btr_search_drop_page_hash_index(block);
	}

	mtr_commit(&mtr);
}

/********************************************************************//**
Builds a hash index on a page with the given parameters. If the page already
has a hash index with different parameters, the old hash index is removed.
If index is non-NULL, this function checks if n_fields and n_bytes are
sensible values, and does not build a hash index if not. */
static
void
btr_search_build_page_hash_index(
/*=============================*/
	dict_index_t*	index,	/*!< in: index for which to build */
	buf_block_t*	block,	/*!< in: index page, s- or x-latched */
	ulint		n_fields,/*!< in: hash this many full fields */
	ulint		n_bytes,/*!< in: hash this many bytes from the next
				field */
	ibool		left_side)/*!< in: hash for searches from left side? */
{
	hash_table_t*	table;
	page_t*		page;
	rec_t*		rec;
	rec_t*		next_rec;
	ulint		fold;
	ulint		next_fold;
	ulint		n_cached;
	ulint		n_recs;
	ulint*		folds;
	rec_t**		recs;
	ulint		i;
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets		= offsets_;
	rec_offs_init(offsets_);

	ut_ad(index);
	ut_a(!dict_index_is_ibuf(index));

#ifdef UNIV_SYNC_DEBUG
	ut_ad(!rw_lock_own(&btr_search_latch, RW_LOCK_EX));
	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_SHARED)
	      || rw_lock_own(&(block->lock), RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

	rw_lock_s_lock(&btr_search_latch);

1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
	if (!btr_search_enabled) {
		rw_lock_s_unlock(&btr_search_latch);
		return;
	}

	table = btr_search_sys->hash_index;
	page = buf_block_get_frame(block);

	if (block->index && ((block->curr_n_fields != n_fields)
			     || (block->curr_n_bytes != n_bytes)
			     || (block->curr_left_side != left_side))) {
1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305

		rw_lock_s_unlock(&btr_search_latch);

		btr_search_drop_page_hash_index(block);
	} else {
		rw_lock_s_unlock(&btr_search_latch);
	}

	n_recs = page_get_n_recs(page);

	if (n_recs == 0) {

		return;
	}

	/* Check that the values for hash index build are sensible */

	if (n_fields + n_bytes == 0) {

		return;
	}

	if (dict_index_get_n_unique_in_tree(index) < n_fields
	    || (dict_index_get_n_unique_in_tree(index) == n_fields
		&& n_bytes > 0)) {
		return;
	}

	/* Calculate and cache fold values and corresponding records into
	an array for fast insertion to the hash index */

	folds = mem_alloc(n_recs * sizeof(ulint));
	recs = mem_alloc(n_recs * sizeof(rec_t*));

	n_cached = 0;

1306
	ut_a(UT_DULINT_EQ(index->id, btr_page_get_index_id(page)));
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320

	rec = page_rec_get_next(page_get_infimum_rec(page));

	offsets = rec_get_offsets(rec, index, offsets,
				  n_fields + (n_bytes > 0), &heap);

	if (!page_rec_is_supremum(rec)) {
		ut_a(n_fields <= rec_offs_n_fields(offsets));

		if (n_bytes > 0) {
			ut_a(n_fields < rec_offs_n_fields(offsets));
		}
	}

1321
	fold = rec_fold(rec, offsets, n_fields, n_bytes, index->id);
1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347

	if (left_side) {

		folds[n_cached] = fold;
		recs[n_cached] = rec;
		n_cached++;
	}

	for (;;) {
		next_rec = page_rec_get_next(rec);

		if (page_rec_is_supremum(next_rec)) {

			if (!left_side) {

				folds[n_cached] = fold;
				recs[n_cached] = rec;
				n_cached++;
			}

			break;
		}

		offsets = rec_get_offsets(next_rec, index, offsets,
					  n_fields + (n_bytes > 0), &heap);
		next_fold = rec_fold(next_rec, offsets, n_fields,
1348
				     n_bytes, index->id);
1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372

		if (fold != next_fold) {
			/* Insert an entry into the hash index */

			if (left_side) {

				folds[n_cached] = next_fold;
				recs[n_cached] = next_rec;
				n_cached++;
			} else {
				folds[n_cached] = fold;
				recs[n_cached] = rec;
				n_cached++;
			}
		}

		rec = next_rec;
		fold = next_fold;
	}

	btr_search_check_free_space_in_heap();

	rw_lock_x_lock(&btr_search_latch);

1373
	if (UNIV_UNLIKELY(!btr_search_enabled)) {
1374 1375 1376
		goto exit_func;
	}

1377 1378 1379
	if (block->index && ((block->curr_n_fields != n_fields)
			     || (block->curr_n_bytes != n_bytes)
			     || (block->curr_left_side != left_side))) {
1380 1381 1382 1383 1384 1385 1386 1387
		goto exit_func;
	}

	/* This counter is decremented every time we drop page
	hash index entries and is incremented here. Since we can
	rebuild hash index for a page that is already hashed, we
	have to take care not to increment the counter in that
	case. */
1388
	if (!block->index) {
1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
		index->search_info->ref_count++;
	}

	block->n_hash_helps = 0;

	block->curr_n_fields = n_fields;
	block->curr_n_bytes = n_bytes;
	block->curr_left_side = left_side;
	block->index = index;

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

		ha_insert_for_fold(table, folds[i], block, recs[i]);
	}

exit_func:
	rw_lock_x_unlock(&btr_search_latch);

	mem_free(folds);
	mem_free(recs);
	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}
}

/********************************************************************//**
Moves or deletes hash entries for moved records. If new_page is already hashed,
then the hash index for page, if any, is dropped. If new_page is not hashed,
and page is hashed, then a new hash index is built to new_page with the same
parameters as page (this often happens when a page is split). */
UNIV_INTERN
void
btr_search_move_or_delete_hash_entries(
/*===================================*/
	buf_block_t*	new_block,	/*!< in: records are copied
					to this page */
	buf_block_t*	block,		/*!< in: index page from which
					records were copied, and the
					copied records will be deleted
					from this page */
	dict_index_t*	index)		/*!< in: record descriptor */
{
	ulint	n_fields;
	ulint	n_bytes;
	ibool	left_side;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_EX));
	ut_ad(rw_lock_own(&(new_block->lock), RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

	rw_lock_s_lock(&btr_search_latch);

1442 1443 1444 1445 1446 1447
	ut_a(!new_block->index || new_block->index == index);
	ut_a(!block->index || block->index == index);
	ut_a(!(new_block->index || block->index)
	     || !dict_index_is_ibuf(index));

	if (new_block->index) {
1448 1449 1450 1451 1452 1453 1454 1455

		rw_lock_s_unlock(&btr_search_latch);

		btr_search_drop_page_hash_index(block);

		return;
	}

1456
	if (block->index) {
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

		n_fields = block->curr_n_fields;
		n_bytes = block->curr_n_bytes;
		left_side = block->curr_left_side;

		new_block->n_fields = block->curr_n_fields;
		new_block->n_bytes = block->curr_n_bytes;
		new_block->left_side = left_side;

		rw_lock_s_unlock(&btr_search_latch);

		ut_a(n_fields + n_bytes > 0);

		btr_search_build_page_hash_index(index, new_block, n_fields,
						 n_bytes, left_side);
		ut_ad(n_fields == block->curr_n_fields);
		ut_ad(n_bytes == block->curr_n_bytes);
		ut_ad(left_side == block->curr_left_side);
		return;
	}

	rw_lock_s_unlock(&btr_search_latch);
}

/********************************************************************//**
Updates the page hash index when a single record is deleted from a page. */
UNIV_INTERN
void
btr_search_update_hash_on_delete(
/*=============================*/
	btr_cur_t*	cursor)	/*!< in: cursor which was positioned on the
				record to delete using btr_cur_search_...,
				the record is not yet deleted */
{
	hash_table_t*	table;
	buf_block_t*	block;
1493
	const rec_t*	rec;
1494
	ulint		fold;
1495
	dict_index_t*	index;
1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	mem_heap_t*	heap		= NULL;
	rec_offs_init(offsets_);

	block = btr_cur_get_block(cursor);

#ifdef UNIV_SYNC_DEBUG
	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

1506 1507 1508
	index = block->index;

	if (!index) {
1509 1510 1511 1512

		return;
	}

1513
	ut_a(index == cursor->index);
1514
	ut_a(block->curr_n_fields + block->curr_n_bytes > 0);
1515
	ut_a(!dict_index_is_ibuf(index));
1516 1517 1518

	table = btr_search_sys->hash_index;

1519 1520 1521
	rec = btr_cur_get_rec(cursor);

	fold = rec_fold(rec, rec_get_offsets(rec, index, offsets_,
1522
					     ULINT_UNDEFINED, &heap),
1523
			block->curr_n_fields, block->curr_n_bytes, index->id);
1524 1525 1526
	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}
1527

1528 1529
	rw_lock_x_lock(&btr_search_latch);

1530 1531 1532 1533 1534
	if (block->index) {
		ut_a(block->index == index);

		ha_search_and_delete_if_found(table, fold, rec);
	}
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551

	rw_lock_x_unlock(&btr_search_latch);
}

/********************************************************************//**
Updates the page hash index when a single record is inserted on a page. */
UNIV_INTERN
void
btr_search_update_hash_node_on_insert(
/*==================================*/
	btr_cur_t*	cursor)	/*!< in: cursor which was positioned to the
				place to insert using btr_cur_search_...,
				and the new record has been inserted next
				to the cursor */
{
	hash_table_t*	table;
	buf_block_t*	block;
1552
	dict_index_t*	index;
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562
	rec_t*		rec;

	rec = btr_cur_get_rec(cursor);

	block = btr_cur_get_block(cursor);

#ifdef UNIV_SYNC_DEBUG
	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

1563 1564 1565
	index = block->index;

	if (!index) {
1566 1567 1568 1569

		return;
	}

1570 1571
	ut_a(cursor->index == index);
	ut_a(!dict_index_is_ibuf(index));
1572 1573 1574

	rw_lock_x_lock(&btr_search_latch);

1575 1576 1577 1578 1579 1580 1581
	if (!block->index) {

		goto func_exit;
	}

	ut_a(block->index == index);

1582 1583 1584 1585 1586 1587 1588 1589 1590 1591
	if ((cursor->flag == BTR_CUR_HASH)
	    && (cursor->n_fields == block->curr_n_fields)
	    && (cursor->n_bytes == block->curr_n_bytes)
	    && !block->curr_left_side) {

		table = btr_search_sys->hash_index;

		ha_search_and_update_if_found(table, cursor->fold, rec,
					      block, page_rec_get_next(rec));

1592
func_exit:
1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
		rw_lock_x_unlock(&btr_search_latch);
	} else {
		rw_lock_x_unlock(&btr_search_latch);

		btr_search_update_hash_on_insert(cursor);
	}
}

/********************************************************************//**
Updates the page hash index when a single record is inserted on a page. */
UNIV_INTERN
void
btr_search_update_hash_on_insert(
/*=============================*/
	btr_cur_t*	cursor)	/*!< in: cursor which was positioned to the
				place to insert using btr_cur_search_...,
				and the new record has been inserted next
				to the cursor */
{
	hash_table_t*	table;
	buf_block_t*	block;
1614
	dict_index_t*	index;
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
	rec_t*		rec;
	rec_t*		ins_rec;
	rec_t*		next_rec;
	ulint		fold;
	ulint		ins_fold;
	ulint		next_fold = 0; /* remove warning (??? bug ???) */
	ulint		n_fields;
	ulint		n_bytes;
	ibool		left_side;
	ibool		locked		= FALSE;
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets		= offsets_;
	rec_offs_init(offsets_);

	table = btr_search_sys->hash_index;

	btr_search_check_free_space_in_heap();

	rec = btr_cur_get_rec(cursor);

	block = btr_cur_get_block(cursor);

#ifdef UNIV_SYNC_DEBUG
	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_EX));
#endif /* UNIV_SYNC_DEBUG */

1642 1643 1644
	index = block->index;

	if (!index) {
1645 1646 1647 1648

		return;
	}

1649 1650
	ut_a(index == cursor->index);
	ut_a(!dict_index_is_ibuf(index));
1651 1652 1653 1654 1655 1656 1657 1658

	n_fields = block->curr_n_fields;
	n_bytes = block->curr_n_bytes;
	left_side = block->curr_left_side;

	ins_rec = page_rec_get_next(rec);
	next_rec = page_rec_get_next(ins_rec);

1659
	offsets = rec_get_offsets(ins_rec, index, offsets,
1660
				  ULINT_UNDEFINED, &heap);
1661
	ins_fold = rec_fold(ins_rec, offsets, n_fields, n_bytes, index->id);
1662 1663

	if (!page_rec_is_supremum(next_rec)) {
1664
		offsets = rec_get_offsets(next_rec, index, offsets,
1665 1666
					  n_fields + (n_bytes > 0), &heap);
		next_fold = rec_fold(next_rec, offsets, n_fields,
1667
				     n_bytes, index->id);
1668 1669 1670
	}

	if (!page_rec_is_infimum(rec)) {
1671
		offsets = rec_get_offsets(rec, index, offsets,
1672
					  n_fields + (n_bytes > 0), &heap);
1673
		fold = rec_fold(rec, offsets, n_fields, n_bytes, index->id);
1674 1675 1676 1677 1678 1679 1680
	} else {
		if (left_side) {

			rw_lock_x_lock(&btr_search_latch);

			locked = TRUE;

1681 1682 1683 1684
			if (!btr_search_enabled) {
				goto function_exit;
			}

1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
			ha_insert_for_fold(table, ins_fold, block, ins_rec);
		}

		goto check_next_rec;
	}

	if (fold != ins_fold) {

		if (!locked) {

			rw_lock_x_lock(&btr_search_latch);

			locked = TRUE;
1698 1699 1700 1701

			if (!btr_search_enabled) {
				goto function_exit;
			}
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719
		}

		if (!left_side) {
			ha_insert_for_fold(table, fold, block, rec);
		} else {
			ha_insert_for_fold(table, ins_fold, block, ins_rec);
		}
	}

check_next_rec:
	if (page_rec_is_supremum(next_rec)) {

		if (!left_side) {

			if (!locked) {
				rw_lock_x_lock(&btr_search_latch);

				locked = TRUE;
1720 1721 1722 1723

				if (!btr_search_enabled) {
					goto function_exit;
				}
1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738
			}

			ha_insert_for_fold(table, ins_fold, block, ins_rec);
		}

		goto function_exit;
	}

	if (ins_fold != next_fold) {

		if (!locked) {

			rw_lock_x_lock(&btr_search_latch);

			locked = TRUE;
1739 1740 1741 1742

			if (!btr_search_enabled) {
				goto function_exit;
			}
1743 1744 1745 1746 1747 1748 1749
		}

		if (!left_side) {

			ha_insert_for_fold(table, ins_fold, block, ins_rec);
			/*
			fputs("Hash insert for ", stderr);
1750
			dict_index_name_print(stderr, index);
1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766
			fprintf(stderr, " fold %lu\n", ins_fold);
			*/
		} else {
			ha_insert_for_fold(table, next_fold, block, next_rec);
		}
	}

function_exit:
	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}
	if (locked) {
		rw_lock_x_unlock(&btr_search_latch);
	}
}

1767
#if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853
/********************************************************************//**
Validates the search system.
@return	TRUE if ok */
UNIV_INTERN
ibool
btr_search_validate(void)
/*=====================*/
{
	ha_node_t*	node;
	ulint		n_page_dumps	= 0;
	ibool		ok		= TRUE;
	ulint		i;
	ulint		cell_count;
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets		= offsets_;

	/* How many cells to check before temporarily releasing
	btr_search_latch. */
	ulint		chunk_size = 10000;

	rec_offs_init(offsets_);

	rw_lock_x_lock(&btr_search_latch);
	buf_pool_mutex_enter();

	cell_count = hash_get_n_cells(btr_search_sys->hash_index);

	for (i = 0; i < cell_count; i++) {
		/* We release btr_search_latch every once in a while to
		give other queries a chance to run. */
		if ((i != 0) && ((i % chunk_size) == 0)) {
			buf_pool_mutex_exit();
			rw_lock_x_unlock(&btr_search_latch);
			os_thread_yield();
			rw_lock_x_lock(&btr_search_latch);
			buf_pool_mutex_enter();
		}

		node = hash_get_nth_cell(btr_search_sys->hash_index, i)->node;

		for (; node != NULL; node = node->next) {
			const buf_block_t*	block
				= buf_block_align(node->data);
			const buf_block_t*	hash_block;

			if (UNIV_LIKELY(buf_block_get_state(block)
					== BUF_BLOCK_FILE_PAGE)) {

				/* The space and offset are only valid
				for file blocks.  It is possible that
				the block is being freed
				(BUF_BLOCK_REMOVE_HASH, see the
				assertion and the comment below) */
				hash_block = buf_block_hash_get(
					buf_block_get_space(block),
					buf_block_get_page_no(block));
			} else {
				hash_block = NULL;
			}

			if (hash_block) {
				ut_a(hash_block == block);
			} else {
				/* When a block is being freed,
				buf_LRU_search_and_free_block() first
				removes the block from
				buf_pool->page_hash by calling
				buf_LRU_block_remove_hashed_page().
				After that, it invokes
				btr_search_drop_page_hash_index() to
				remove the block from
				btr_search_sys->hash_index. */

				ut_a(buf_block_get_state(block)
				     == BUF_BLOCK_REMOVE_HASH);
			}

			ut_a(!dict_index_is_ibuf(block->index));

			offsets = rec_get_offsets((const rec_t*) node->data,
						  block->index, offsets,
						  block->curr_n_fields
						  + (block->curr_n_bytes > 0),
						  &heap);

1854
			if (!block->index || node->fold
1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888
			    != rec_fold((rec_t*)(node->data),
					offsets,
					block->curr_n_fields,
					block->curr_n_bytes,
					btr_page_get_index_id(block->frame))) {
				const page_t*	page = block->frame;

				ok = FALSE;
				ut_print_timestamp(stderr);

				fprintf(stderr,
					"  InnoDB: Error in an adaptive hash"
					" index pointer to page %lu\n"
					"InnoDB: ptr mem address %p"
					" index id %lu %lu,"
					" node fold %lu, rec fold %lu\n",
					(ulong) page_get_page_no(page),
					node->data,
					(ulong) ut_dulint_get_high(
						btr_page_get_index_id(page)),
					(ulong) ut_dulint_get_low(
						btr_page_get_index_id(page)),
					(ulong) node->fold,
					(ulong) rec_fold((rec_t*)(node->data),
							 offsets,
							 block->curr_n_fields,
							 block->curr_n_bytes,
							 btr_page_get_index_id(
								 page)));

				fputs("InnoDB: Record ", stderr);
				rec_print_new(stderr, (rec_t*)node->data,
					      offsets);
				fprintf(stderr, "\nInnoDB: on that page."
1889
					" Page mem address %p, is hashed %p,"
1890 1891
					" n fields %lu, n bytes %lu\n"
					"InnoDB: side %lu\n",
1892
					(void*) page, (void*) block->index,
1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 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
					(ulong) block->curr_n_fields,
					(ulong) block->curr_n_bytes,
					(ulong) block->curr_left_side);

				if (n_page_dumps < 20) {
					buf_page_print(page, 0);
					n_page_dumps++;
				}
			}
		}
	}

	for (i = 0; i < cell_count; i += chunk_size) {
		ulint end_index = ut_min(i + chunk_size - 1, cell_count - 1);

		/* We release btr_search_latch every once in a while to
		give other queries a chance to run. */
		if (i != 0) {
			buf_pool_mutex_exit();
			rw_lock_x_unlock(&btr_search_latch);
			os_thread_yield();
			rw_lock_x_lock(&btr_search_latch);
			buf_pool_mutex_enter();
		}

		if (!ha_validate(btr_search_sys->hash_index, i, end_index)) {
			ok = FALSE;
		}
	}

	buf_pool_mutex_exit();
	rw_lock_x_unlock(&btr_search_latch);
	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}

	return(ok);
}
1931
#endif /* defined UNIV_AHI_DEBUG || defined UNIV_DEBUG */