mariadb/storage/innobase/include/page0page.ic

/******************************************************
Index page routines

(c) 1994-1996 Innobase Oy

Created 2/2/1994 Heikki Tuuri
*******************************************************/

#include "mach0data.h"
#include "rem0cmp.h"
#include "mtr0log.h"

#ifdef UNIV_MATERIALIZE
#undef UNIV_INLINE
#define UNIV_INLINE
#endif

/*****************************************************************
Returns the max trx id field value. */
UNIV_INLINE
dulint
page_get_max_trx_id(
/*================*/
	page_t*	page)	/* in: page */
{
	ut_ad(page);

	return(mach_read_from_8(page + PAGE_HEADER + PAGE_MAX_TRX_ID));
}

/*****************************************************************
Sets the max trx id field value if trx_id is bigger than the previous
value. */
UNIV_INLINE
void
page_update_max_trx_id(
/*===================*/
	page_t*	page,	/* in: page */
	dulint	trx_id)	/* in: transaction id */
{
	ut_ad(page);

	if (ut_dulint_cmp(page_get_max_trx_id(page), trx_id) < 0) {

		page_set_max_trx_id(page, trx_id);
	}
}

/*****************************************************************
Reads the given header field. */
UNIV_INLINE
ulint
page_header_get_field(
/*==================*/
	page_t*	page,	/* in: page */
	ulint	field)	/* in: PAGE_LEVEL, ... */
{
	ut_ad(page);
	ut_ad(field <= PAGE_INDEX_ID);

	return(mach_read_from_2(page + PAGE_HEADER + field));
}

/*****************************************************************
Sets the given header field. */
UNIV_INLINE
void
page_header_set_field(
/*==================*/
	page_t*	page,	/* in: page */
	ulint	field,	/* in: PAGE_LEVEL, ... */
	ulint	val)	/* in: value */
{
	ut_ad(page);
	ut_ad(field <= PAGE_N_RECS);
	ut_ad(field == PAGE_N_HEAP || val < UNIV_PAGE_SIZE);
	ut_ad(field != PAGE_N_HEAP || (val & 0x7fff) < UNIV_PAGE_SIZE);

	mach_write_to_2(page + PAGE_HEADER + field, val);
}

/*****************************************************************
Returns the pointer stored in the given header field. */
UNIV_INLINE
byte*
page_header_get_ptr(
/*================*/
			/* out: pointer or NULL */
	page_t*	page,	/* in: page */
	ulint	field)	/* in: PAGE_FREE, ... */
{
	ulint	offs;

	ut_ad(page);
	ut_ad((field == PAGE_FREE)
		|| (field == PAGE_LAST_INSERT)
		|| (field == PAGE_HEAP_TOP));

	offs = page_header_get_field(page, field);

	ut_ad((field != PAGE_HEAP_TOP) || offs);

	if (offs == 0) {

		return(NULL);
	}

	return(page + offs);
}

/*****************************************************************
Sets the pointer stored in the given header field. */
UNIV_INLINE
void
page_header_set_ptr(
/*================*/
	page_t*	page,	/* in: page */
	ulint	field,	/* in: PAGE_FREE, ... */
	byte*	ptr)	/* in: pointer or NULL*/
{
	ulint	offs;

	ut_ad(page);
	ut_ad((field == PAGE_FREE)
		|| (field == PAGE_LAST_INSERT)
		|| (field == PAGE_HEAP_TOP));

	if (ptr == NULL) {
		offs = 0;
	} else {
		offs = ptr - page;
	}

	ut_ad((field != PAGE_HEAP_TOP) || offs);

	page_header_set_field(page, field, offs);
}

/*****************************************************************
Resets the last insert info field in the page header. Writes to mlog
about this operation. */
UNIV_INLINE
void
page_header_reset_last_insert(
/*==========================*/
	page_t*	page,	/* in: page */
	mtr_t*	mtr)	/* in: mtr */
{
	ut_ad(page && mtr);

	mlog_write_ulint(page + PAGE_HEADER + PAGE_LAST_INSERT, 0,
							MLOG_2BYTES, mtr);
}

/****************************************************************
Determine whether the page is in new-style compact format. */
UNIV_INLINE
ulint
page_is_comp(
/*=========*/
			/* out: nonzero if the page is in compact
			format, zero if it is in old-style format */
	page_t*	page)	/* in: index page */
{
	return(UNIV_EXPECT(page_header_get_field(page, PAGE_N_HEAP) & 0x8000,
		0x8000));
}

/****************************************************************
TRUE if the record is on a page in compact format. */
UNIV_INLINE
ulint
page_rec_is_comp(
/*=============*/
				/* out: nonzero if in compact format */
	const rec_t*	rec)	/* in: record */
{
	return(page_is_comp(ut_align_down((rec_t*) rec, UNIV_PAGE_SIZE)));
}

/****************************************************************
Gets the first record on the page. */
UNIV_INLINE
rec_t*
page_get_infimum_rec(
/*=================*/
			/* out: the first record in record list */
	page_t*	page)	/* in: page which must have record(s) */
{
	ut_ad(page);

	if (page_is_comp(page)) {
		return(page + PAGE_NEW_INFIMUM);
	} else {
		return(page + PAGE_OLD_INFIMUM);
	}
}

/****************************************************************
Gets the last record on the page. */
UNIV_INLINE
rec_t*
page_get_supremum_rec(
/*==================*/
			/* out: the last record in record list */
	page_t*	page)	/* in: page which must have record(s) */
{
	ut_ad(page);

	if (page_is_comp(page)) {
		return(page + PAGE_NEW_SUPREMUM);
	} else {
		return(page + PAGE_OLD_SUPREMUM);
	}
}

/****************************************************************
TRUE if the record is a user record on the page. */
UNIV_INLINE
ibool
page_rec_is_user_rec_low(
/*=====================*/
			/* out: TRUE if a user record */
	ulint	offset)	/* in: record offset on page */
{
	ut_ad(offset >= PAGE_NEW_INFIMUM);
#if PAGE_OLD_INFIMUM < PAGE_NEW_INFIMUM
# error "PAGE_OLD_INFIMUM < PAGE_NEW_INFIMUM"
#endif
#if PAGE_OLD_SUPREMUM < PAGE_NEW_SUPREMUM
# error "PAGE_OLD_SUPREMUM < PAGE_NEW_SUPREMUM"
#endif
#if PAGE_NEW_INFIMUM > PAGE_OLD_SUPREMUM
# error "PAGE_NEW_INFIMUM > PAGE_OLD_SUPREMUM"
#endif
#if PAGE_OLD_INFIMUM > PAGE_NEW_SUPREMUM
# error "PAGE_OLD_INFIMUM > PAGE_NEW_SUPREMUM"
#endif
#if PAGE_NEW_SUPREMUM > PAGE_OLD_SUPREMUM_END
# error "PAGE_NEW_SUPREMUM > PAGE_OLD_SUPREMUM_END"
#endif
#if PAGE_OLD_SUPREMUM > PAGE_NEW_SUPREMUM_END
# error "PAGE_OLD_SUPREMUM > PAGE_NEW_SUPREMUM_END"
#endif
	ut_ad(offset <= UNIV_PAGE_SIZE - PAGE_EMPTY_DIR_START);

	return(UNIV_LIKELY(offset != PAGE_NEW_SUPREMUM)
		&& UNIV_LIKELY(offset != PAGE_NEW_INFIMUM)
		&& UNIV_LIKELY(offset != PAGE_OLD_INFIMUM)
		&& UNIV_LIKELY(offset != PAGE_OLD_SUPREMUM));
}

/****************************************************************
TRUE if the record is the supremum record on a page. */
UNIV_INLINE
ibool
page_rec_is_supremum_low(
/*=====================*/
			/* out: TRUE if the supremum record */
	ulint	offset)	/* in: record offset on page */
{
	ut_ad(offset >= PAGE_NEW_INFIMUM);
	ut_ad(offset <= UNIV_PAGE_SIZE - PAGE_EMPTY_DIR_START);

	return(UNIV_UNLIKELY(offset == PAGE_NEW_SUPREMUM)
		|| UNIV_UNLIKELY(offset == PAGE_OLD_SUPREMUM));
}

/****************************************************************
TRUE if the record is the infimum record on a page. */
UNIV_INLINE
ibool
page_rec_is_infimum_low(
/*====================*/
			/* out: TRUE if the infimum record */
	ulint	offset)	/* in: record offset on page */
{
	ut_ad(offset >= PAGE_NEW_INFIMUM);
	ut_ad(offset <= UNIV_PAGE_SIZE - PAGE_EMPTY_DIR_START);

	return(UNIV_UNLIKELY(offset == PAGE_NEW_INFIMUM)
		|| UNIV_UNLIKELY(offset == PAGE_OLD_INFIMUM));
}

/****************************************************************
TRUE if the record is a user record on the page. */
UNIV_INLINE
ibool
page_rec_is_user_rec(
/*=================*/
				/* out: TRUE if a user record */
	const rec_t*	rec)	/* in: record */
{
	return(page_rec_is_user_rec_low(
			ut_align_offset(rec, UNIV_PAGE_SIZE)));
}

/****************************************************************
TRUE if the record is the supremum record on a page. */
UNIV_INLINE
ibool
page_rec_is_supremum(
/*=================*/
				/* out: TRUE if the supremum record */
	const rec_t*	rec)	/* in: record */
{
	return(page_rec_is_supremum_low(
			ut_align_offset(rec, UNIV_PAGE_SIZE)));
}

/****************************************************************
TRUE if the record is the infimum record on a page. */
UNIV_INLINE
ibool
page_rec_is_infimum(
/*================*/
				/* out: TRUE if the infimum record */
	const rec_t*	rec)	/* in: record */
{
	return(page_rec_is_infimum_low(
			ut_align_offset(rec, UNIV_PAGE_SIZE)));
}

/*****************************************************************
Compares a data tuple to a physical record. Differs from the function
cmp_dtuple_rec_with_match in the way that the record must reside on an
index page, and also page infimum and supremum records can be given in
the parameter rec. These are considered as the negative infinity and
the positive infinity in the alphabetical order. */
UNIV_INLINE
int
page_cmp_dtuple_rec_with_match(
/*===========================*/
				/* out: 1, 0, -1, if dtuple is greater, equal,
				less than rec, respectively, when only the
				common first fields are compared */
	dtuple_t*	dtuple,	/* in: data tuple */
	rec_t*		rec,	/* in: physical record on a page; may also
				be page infimum or supremum, in which case
				matched-parameter values below are not
				affected */
	const ulint*	offsets,/* in: array returned by rec_get_offsets() */
	ulint*		matched_fields, /* in/out: number of already completely
				matched fields; when function returns
				contains the value for current comparison */
	ulint*		matched_bytes) /* in/out: number of already matched
				bytes within the first field not completely
				matched; when function returns contains the
				value for current comparison */
{
	ulint	rec_offset;

	ut_ad(dtuple_check_typed(dtuple));
	ut_ad(rec_offs_validate(rec, NULL, offsets));
	ut_ad(!rec_offs_comp(offsets) == !page_rec_is_comp(rec));

	rec_offset = ut_align_offset(rec, UNIV_PAGE_SIZE);

	if (UNIV_UNLIKELY(rec_offset == PAGE_NEW_INFIMUM)
			|| UNIV_UNLIKELY(rec_offset == PAGE_OLD_INFIMUM)) {
		return(1);
	}
	if (UNIV_UNLIKELY(rec_offset == PAGE_NEW_SUPREMUM)
			|| UNIV_UNLIKELY(rec_offset == PAGE_OLD_SUPREMUM)) {
		return(-1);
	}

	return(cmp_dtuple_rec_with_match(dtuple, rec, offsets,
						matched_fields,
						matched_bytes));
}

/*****************************************************************
Gets the number of user records on page (infimum and supremum records
are not user records). */
UNIV_INLINE
ulint
page_get_n_recs(
/*============*/
			/* out: number of user records */
	page_t*	page)	/* in: index page */
{
	return(page_header_get_field(page, PAGE_N_RECS));
}

/*****************************************************************
Gets the number of dir slots in directory. */
UNIV_INLINE
ulint
page_dir_get_n_slots(
/*=================*/
			/* out: number of slots */
	page_t*	page)	/* in: index page */
{
	return(page_header_get_field(page, PAGE_N_DIR_SLOTS));
}
/*****************************************************************
Sets the number of dir slots in directory. */
UNIV_INLINE
void
page_dir_set_n_slots(
/*=================*/
			/* out: number of slots */
	page_t*	page,	/* in: index page */
	ulint	n_slots)/* in: number of slots */
{
	page_header_set_field(page, PAGE_N_DIR_SLOTS, n_slots);
}

/*****************************************************************
Gets the number of records in the heap. */
UNIV_INLINE
ulint
page_dir_get_n_heap(
/*================*/
			/* out: number of user records */
	page_t*	page)	/* in: index page */
{
	return(page_header_get_field(page, PAGE_N_HEAP) & 0x7fff);
}

/*****************************************************************
Sets the number of records in the heap. */
UNIV_INLINE
void
page_dir_set_n_heap(
/*================*/
	page_t*	page,	/* in: index page */
	ulint	n_heap)	/* in: number of records */
{
	ut_ad(n_heap < 0x8000);

	page_header_set_field(page, PAGE_N_HEAP, n_heap | (0x8000 &
		page_header_get_field(page, PAGE_N_HEAP)));
}

/*****************************************************************
Gets pointer to nth directory slot. */
UNIV_INLINE
page_dir_slot_t*
page_dir_get_nth_slot(
/*==================*/
			/* out: pointer to dir slot */
	page_t*	page,	/* in: index page */
	ulint	n)	/* in: position */
{
	ut_ad(page_dir_get_n_slots(page) > n);

	return(page + UNIV_PAGE_SIZE - PAGE_DIR
					- (n + 1) * PAGE_DIR_SLOT_SIZE);
}

/******************************************************************
Used to check the consistency of a record on a page. */
UNIV_INLINE
ibool
page_rec_check(
/*===========*/
			/* out: TRUE if succeed */
	rec_t*	rec)	/* in: record */
{
	page_t*	page;

	ut_a(rec);

	page = buf_frame_align(rec);

	ut_a(rec <= page_header_get_ptr(page, PAGE_HEAP_TOP));
	ut_a(rec >= page + PAGE_DATA);

	return(TRUE);
}

/*******************************************************************
Gets the record pointed to by a directory slot. */
UNIV_INLINE
rec_t*
page_dir_slot_get_rec(
/*==================*/
					/* out: pointer to record */
	page_dir_slot_t*	slot)	/* in: directory slot */
{
	return(buf_frame_align(slot) + mach_read_from_2(slot));
}

/*******************************************************************
This is used to set the record offset in a directory slot. */
UNIV_INLINE
void
page_dir_slot_set_rec(
/*==================*/
	page_dir_slot_t* slot,	/* in: directory slot */
	rec_t*		 rec)	/* in: record on the page */
{
	ut_ad(page_rec_check(rec));

	mach_write_to_2(slot, ut_align_offset(rec, UNIV_PAGE_SIZE));
}

/*******************************************************************
Gets the number of records owned by a directory slot. */
UNIV_INLINE
ulint
page_dir_slot_get_n_owned(
/*======================*/
					/* out: number of records */
	page_dir_slot_t*	slot)	/* in: page directory slot */
{
	rec_t*	rec	= page_dir_slot_get_rec(slot);
	return(rec_get_n_owned(rec, page_rec_is_comp(rec)));
}

/*******************************************************************
This is used to set the owned records field of a directory slot. */
UNIV_INLINE
void
page_dir_slot_set_n_owned(
/*======================*/
	page_dir_slot_t*	slot,	/* in: directory slot */
	ulint			n)	/* in: number of records owned
					by the slot */
{
	rec_t*	rec	= page_dir_slot_get_rec(slot);
	rec_set_n_owned(rec, page_rec_is_comp(rec), n);
}

/****************************************************************
Calculates the space reserved for directory slots of a given number of
records. The exact value is a fraction number n * PAGE_DIR_SLOT_SIZE /
PAGE_DIR_SLOT_MIN_N_OWNED, and it is rounded upwards to an integer. */
UNIV_INLINE
ulint
page_dir_calc_reserved_space(
/*=========================*/
	ulint	n_recs)		/* in: number of records */
{
	return((PAGE_DIR_SLOT_SIZE * n_recs + PAGE_DIR_SLOT_MIN_N_OWNED - 1)
						/ PAGE_DIR_SLOT_MIN_N_OWNED);
}

/****************************************************************
Gets the pointer to the next record on the page. */
UNIV_INLINE
rec_t*
page_rec_get_next(
/*==============*/
			/* out: pointer to next record */
	rec_t*	rec)	/* in: pointer to record */
{
	ulint	offs;
	page_t*	page;

	ut_ad(page_rec_check(rec));

	page = ut_align_down(rec, UNIV_PAGE_SIZE);

	offs = rec_get_next_offs(rec, page_is_comp(page));

	if (UNIV_UNLIKELY(offs >= UNIV_PAGE_SIZE)) {
		fprintf(stderr,
"InnoDB: Next record offset is nonsensical %lu in record at offset %lu\n"
"InnoDB: rec address %p, first buffer frame %p\n"
"InnoDB: buffer pool high end %p, buf fix count %lu\n",
				(ulong)offs, (ulong)(rec - page),
				rec, buf_pool->frame_zero,
				buf_pool->high_end,
				(ulong)buf_block_align(rec)->buf_fix_count);
		buf_page_print(page);

		ut_error;
	}

	if (UNIV_UNLIKELY(offs == 0)) {

		return(NULL);
	}

	return(page + offs);
}

/****************************************************************
Sets the pointer to the next record on the page. */
UNIV_INLINE
void
page_rec_set_next(
/*==============*/
	rec_t*	rec,	/* in: pointer to record, must not be page supremum */
	rec_t*	next)	/* in: pointer to next record, must not be page
			infimum */
{
	page_t*	page;
	ulint	offs;

	ut_ad(page_rec_check(rec));
	ut_ad(!page_rec_is_supremum(rec));
	page = ut_align_down(rec, UNIV_PAGE_SIZE);

	if (next) {
		ut_ad(!page_rec_is_infimum(next));
		ut_ad(page == ut_align_down(next, UNIV_PAGE_SIZE));
		offs = (ulint) (next - page);
	} else {
		offs = 0;
	}

	rec_set_next_offs(rec, page_is_comp(page), offs);
}

/****************************************************************
Gets the pointer to the previous record. */
UNIV_INLINE
rec_t*
page_rec_get_prev(
/*==============*/
			/* out: pointer to previous record */
	rec_t*	rec)	/* in: pointer to record, must not be page
			infimum */
{
	page_dir_slot_t*	slot;
	ulint			slot_no;
	rec_t*			rec2;
	rec_t*			prev_rec = NULL;
	page_t*			page;

	ut_ad(page_rec_check(rec));

	page = ut_align_down(rec, UNIV_PAGE_SIZE);

	ut_ad(!page_rec_is_infimum(rec));

	slot_no = page_dir_find_owner_slot(rec);

	ut_a(slot_no != 0);

	slot = page_dir_get_nth_slot(page, slot_no - 1);

	rec2 = page_dir_slot_get_rec(slot);

	while (rec != rec2) {
		prev_rec = rec2;
		rec2 = page_rec_get_next(rec2);
	}

	ut_a(prev_rec);

	return(prev_rec);
}

/*******************************************************************
Looks for the record which owns the given record. */
UNIV_INLINE
rec_t*
page_rec_find_owner_rec(
/*====================*/
			/* out: the owner record */
	rec_t*	rec)	/* in: the physical record */
{
	ut_ad(page_rec_check(rec));

	if (page_rec_is_comp(rec)) {
		while (rec_get_n_owned(rec, TRUE) == 0) {
			rec = page_rec_get_next(rec);
		}
	} else {
		while (rec_get_n_owned(rec, FALSE) == 0) {
			rec = page_rec_get_next(rec);
		}
	}

	return(rec);
}

/****************************************************************
Returns the sum of the sizes of the records in the record list, excluding
the infimum and supremum records. */
UNIV_INLINE
ulint
page_get_data_size(
/*===============*/
			/* out: data in bytes */
	page_t*	page)	/* in: index page */
{
	ulint	ret;

	ret = (ulint)(page_header_get_field(page, PAGE_HEAP_TOP)
		- (page_is_comp(page)
			? PAGE_NEW_SUPREMUM_END
			: PAGE_OLD_SUPREMUM_END)
		- page_header_get_field(page, PAGE_GARBAGE));

	ut_ad(ret < UNIV_PAGE_SIZE);

	return(ret);
}

/*****************************************************************
Calculates free space if a page is emptied. */
UNIV_INLINE
ulint
page_get_free_space_of_empty(
/*=========================*/
				/* out: free space */
	ulint	comp)		/* in: nonzero=compact page layout */
{
	if (UNIV_LIKELY(comp)) {
		return((ulint)(UNIV_PAGE_SIZE
		- PAGE_NEW_SUPREMUM_END
		- PAGE_DIR
		- 2 * PAGE_DIR_SLOT_SIZE));
	}

	return((ulint)(UNIV_PAGE_SIZE
		- PAGE_OLD_SUPREMUM_END
		- PAGE_DIR
		- 2 * PAGE_DIR_SLOT_SIZE));
}

/****************************************************************
Each user record on a page, and also the deleted user records in the heap
takes its size plus the fraction of the dir cell size /
PAGE_DIR_SLOT_MIN_N_OWNED bytes for it. If the sum of these exceeds the
value of page_get_free_space_of_empty, the insert is impossible, otherwise
it is allowed. This function returns the maximum combined size of records
which can be inserted on top of the record heap. */
UNIV_INLINE
ulint
page_get_max_insert_size(
/*=====================*/
			/* out: maximum combined size for inserted records */
	page_t*	page,	/* in: index page */
	ulint	n_recs)	/* in: number of records */
{
	ulint	occupied;
	ulint	free_space;

	if (page_is_comp(page)) {
		occupied = page_header_get_field(page, PAGE_HEAP_TOP)
		- PAGE_NEW_SUPREMUM_END + page_dir_calc_reserved_space(
			n_recs + page_dir_get_n_heap(page) - 2);

		free_space = page_get_free_space_of_empty(TRUE);
	} else {
		occupied = page_header_get_field(page, PAGE_HEAP_TOP)
		- PAGE_OLD_SUPREMUM_END + page_dir_calc_reserved_space(
			n_recs + page_dir_get_n_heap(page) - 2);

		free_space = page_get_free_space_of_empty(FALSE);
	}

	/* Above the 'n_recs +' part reserves directory space for the new
	inserted records; the '- 2' excludes page infimum and supremum
	records */

	if (occupied > free_space) {

		return(0);
	}

	return(free_space - occupied);
}

/****************************************************************
Returns the maximum combined size of records which can be inserted on top
of the record heap if a page is first reorganized. */
UNIV_INLINE
ulint
page_get_max_insert_size_after_reorganize(
/*======================================*/
			/* out: maximum combined size for inserted records */
	page_t*	page,	/* in: index page */
	ulint	n_recs)	/* in: number of records */
{
	ulint	occupied;
	ulint	free_space;

	occupied = page_get_data_size(page)
		+ page_dir_calc_reserved_space(n_recs + page_get_n_recs(page));

	free_space = page_get_free_space_of_empty(page_is_comp(page));

	if (occupied > free_space) {

		return(0);
	}

	return(free_space - occupied);
}

/****************************************************************
Puts a record to free list. */
UNIV_INLINE
void
page_mem_free(
/*==========*/
	page_t*		page,	/* in: index page */
	rec_t*		rec,	/* in: pointer to the (origin of) record */
	const ulint*	offsets)/* in: array returned by rec_get_offsets() */
{
	rec_t*		free;
	ulint		garbage;

	ut_ad(rec_offs_validate(rec, NULL, offsets));
	ut_ad(!rec_offs_comp(offsets) == !page_rec_is_comp(rec));
	free = page_header_get_ptr(page, PAGE_FREE);

	page_rec_set_next(rec, free);
	page_header_set_ptr(page, PAGE_FREE, rec);

#if 0	/* It's better not to destroy the user's data. */

	/* Clear the data bytes of the deleted record in order to improve
	the compression ratio of the page and to make it easier to read
	page dumps in corruption reports.  The extra bytes of the record
	cannot be cleared, because page_mem_alloc() needs them in order
	to determine the size of the deleted record. */
	memset(rec, 0, rec_offs_data_size(offsets));
#endif

	garbage = page_header_get_field(page, PAGE_GARBAGE);

	page_header_set_field(page, PAGE_GARBAGE,
				garbage + rec_offs_size(offsets));
}

#ifdef UNIV_MATERIALIZE
#undef UNIV_INLINE
#define UNIV_INLINE	UNIV_INLINE_ORIGINAL
#endif