Index page routines. More...

#include "my_dbug.h"
#include "btr0btr.h"
#include "buf0buf.h"
#include "ibuf0ibuf.h"
#include "page0cur.h"
#include "page0page.h"
#include "page0zip.h"
#include "btr0sea.h"
#include "fut0lst.h"
#include "lock0lock.h"
#include "srv0srv.h"

Functions
ulint	page_dir_find_owner_slot (const rec_t *rec)
	Looks for the directory slot which owns the given record. More...

static bool	page_dir_slot_check (const page_dir_slot_t *slot)
	Used to check the consistency of a directory slot. More...

void	page_set_max_trx_id (buf_block_t block, page_zip_des_t page_zip, trx_id_t trx_id, mtr_t *mtr)
	Sets the max trx id field value. More...

byte *	page_mem_alloc_heap (page_t page, page_zip_des_t page_zip, ulint need, ulint *heap_no)
	Allocates a block of memory from the heap of an index page. More...

static void	page_create_write_log (buf_frame_t frame, mtr_t mtr, bool comp, page_type_t page_type)
	Writes a log record of page creation. More...

static page_t *	page_create_low (buf_block_t *block, ulint comp, page_type_t page_type)
	The index page creation function. More...

void	page_parse_create (buf_block_t *block, ulint comp, page_type_t page_type)
	Parses a redo log record of creating a page. More...

page_t *	page_create (buf_block_t block, mtr_t mtr, ulint comp, page_type_t page_type)
	Create an uncompressed B-tree or R-tree or SDI index page. More...

page_t *	page_create_zip (buf_block_t block, dict_index_t index, ulint level, trx_id_t max_trx_id, mtr_t *mtr, page_type_t page_type)
	Create a compressed B-tree index page. More...

void	page_create_empty (buf_block_t block, dict_index_t index, mtr_t *mtr)
	Empty a previously created B-tree index page. More...

void	page_copy_rec_list_end_no_locks (buf_block_t new_block, buf_block_t block, rec_t rec, dict_index_t index, mtr_t *mtr)
	Differs from page_copy_rec_list_end, because this function does not touch the lock table and max trx id on page or compress the page. More...

rec_t *	page_copy_rec_list_end (buf_block_t new_block, buf_block_t block, rec_t rec, dict_index_t index, mtr_t *mtr)
	Copies records from page to new_page, from a given record onward, including that record. More...

rec_t *	page_copy_rec_list_start (buf_block_t new_block, buf_block_t block, rec_t rec, dict_index_t index, mtr_t *mtr)
	Copies records from page to new_page, up to the given record, NOT including that record. More...

static void	page_delete_rec_list_write_log (rec_t rec, dict_index_t index, mlog_id_t type, mtr_t *mtr)
	Writes a log record of a record list end or start deletion. More...

byte *	page_parse_delete_rec_list (mlog_id_t type, byte ptr, byte end_ptr, buf_block_t block, dict_index_t index, mtr_t *mtr)
	Parses a log record of a record list end or start deletion. More...

void	page_delete_rec_list_end (rec_t rec, buf_block_t block, dict_index_t index, ulint n_recs, ulint size, mtr_t mtr)
	Deletes records from a page from a given record onward, including that record. More...

void	page_delete_rec_list_start (rec_t rec, buf_block_t block, dict_index_t index, mtr_t mtr)
	Deletes records from page, up to the given record, NOT including that record. More...

bool	page_move_rec_list_end (buf_block_t new_block, buf_block_t block, rec_t split_rec, dict_index_t index, mtr_t *mtr)
	Moves record list end to another page. More...

bool	page_move_rec_list_start (buf_block_t new_block, buf_block_t block, rec_t split_rec, dict_index_t index, mtr_t *mtr)
	Moves record list start to another page. More...

static void	page_dir_delete_slot (page_t page, page_zip_des_t page_zip, ulint slot_no)
	Used to delete n slots from the directory. More...

static void	page_dir_add_slot (page_t page, page_zip_des_t page_zip, ulint start)
	Used to add n slots to the directory. More...

void	page_dir_split_slot (page_t page, page_zip_des_t page_zip, ulint slot_no)
	Splits a directory slot which owns too many records. More...

void	page_dir_balance_slot (page_t page, page_zip_des_t page_zip, ulint slot_no)
	Tries to balance the given directory slot with too few records with the upper neighbor, so that there are at least the minimum number of records owned by the slot; this may result in the merging of two slots. More...

const rec_t *	page_rec_get_nth_const (const page_t *page, ulint nth)
	Returns the nth record of the record list. More...

ulint	page_rec_get_n_recs_before (const rec_t *rec)
	Returns the number of records before the given record in chain. More...

void	page_rec_print (const rec_t rec, const ulint offsets)
	Prints record contents including the data relevant only in the index page context. More...

bool	page_rec_validate (const rec_t rec, const ulint offsets)
	The following is used to validate a record on a page. More...

void	page_check_dir (const page_t *page)
	Checks that the first directory slot points to the infimum record and the last to the supremum. More...

bool	page_simple_validate_old (const page_t *page)
	This function checks the consistency of an index page when we do not know the index. More...

bool	page_simple_validate_new (const page_t *page)
	This function checks the consistency of an index page when we do not know the index. More...

bool	page_is_spatial_non_leaf (const rec_t rec, dict_index_t index)
	This function checks if the page in which record is present is a non-leaf node of a spatial index. More...

bool	page_validate (const page_t page, dict_index_t index)
	This function checks the consistency of an index page. More...

const rec_t *	page_find_rec_with_heap_no (const page_t *page, ulint heap_no)
	Looks in the page record list for a record with the given heap number. More...

bool	page_delete_rec (const dict_index_t index, page_cur_t pcur, const ulint *offsets)
	Removes the record from a leaf page. More...

const rec_t *	page_find_rec_last_not_deleted (const page_t *page)
	Get the last non-delete-marked record on a page. More...

void	page_warn_strict_checksum (srv_checksum_algorithm_t curr_algo, srv_checksum_algorithm_t page_checksum, const page_id_t &page_id)
	Issue a warning when the checksum that is stored in the page is valid, but different than the global setting innodb_checksum_algorithm. More...

Variables
static const byte	infimum_supremum_redundant []
	The page infimum and supremum of an empty page in ROW_FORMAT=REDUNDANT. More...

static const byte	infimum_supremum_compact []
	The page infimum and supremum of an empty page in ROW_FORMAT=COMPACT. More...

Detailed Description

Index page routines.

Created 2/2/1994 Heikki Tuuri

Function Documentation

◆ page_check_dir()

void page_check_dir ( const page_t * page )

Checks that the first directory slot points to the infimum record and the last to the supremum.

This function is intended to track if the bug fixed in 4.0.14 has caused corruption to users' databases.

Parameters

page	in: index page

◆ page_copy_rec_list_end()

rec_t * page_copy_rec_list_end	(	buf_block_t *	new_block,
		buf_block_t *	block,
		rec_t *	rec,
		dict_index_t *	index,
		mtr_t *	mtr
	)

Copies records from page to new_page, from a given record onward, including that record.

Copies records from page to new_page, from the given record onward, including that record.

Infimum and supremum records are not copied. The records are copied to the start of the record list on new_page.

IMPORTANT: The caller will have to update IBUF_BITMAP_FREE if new_block is a compressed leaf page in a secondary index. This has to be done either within the same mini-transaction, or by invoking ibuf_reset_free_bits() before mtr_commit().

Returns: pointer to the original successor of the infimum record on new_page, or NULL on zip overflow (new_block will be decompressed)

Parameters

new_block	in/out: index page to copy to
block	in: index page containing rec
rec	in: record on page
index	in: record descriptor
mtr	in: mtr

◆ page_copy_rec_list_end_no_locks()

void page_copy_rec_list_end_no_locks	(	buf_block_t *	new_block,
		buf_block_t *	block,
		rec_t *	rec,
		dict_index_t *	index,
		mtr_t *	mtr
	)

Differs from page_copy_rec_list_end, because this function does not touch the lock table and max trx id on page or compress the page.

IMPORTANT: The caller will have to update IBUF_BITMAP_FREE if new_block is a compressed leaf page in a secondary index. This has to be done either within the same mini-transaction, or by invoking ibuf_reset_free_bits() before mtr_commit().

Parameters

new_block	in: index page to copy to
block	in: index page of rec
rec	in: record on page
index	in: record descriptor
mtr	in: mtr

◆ page_copy_rec_list_start()

rec_t * page_copy_rec_list_start	(	buf_block_t *	new_block,
		buf_block_t *	block,
		rec_t *	rec,
		dict_index_t *	index,
		mtr_t *	mtr
	)

Copies records from page to new_page, up to the given record, NOT including that record.

Infimum and supremum records are not copied. The records are copied to the end of the record list on new_page.

IMPORTANT: The caller will have to update IBUF_BITMAP_FREE if new_block is a compressed leaf page in a secondary index. This has to be done either within the same mini-transaction, or by invoking ibuf_reset_free_bits() before mtr_commit().

Returns: pointer to the original predecessor of the supremum record on new_page, or NULL on zip overflow (new_block will be decompressed)

Parameters

new_block	in/out: index page to copy to
block	in: index page containing rec
rec	in: record on page
index	in: record descriptor
mtr	in: mtr

◆ page_create()

page_t * page_create	(	buf_block_t *	block,
		mtr_t *	mtr,
		ulint	comp,
		page_type_t	page_type
	)

Create an uncompressed B-tree or R-tree or SDI index page.

Parameters

[in]	block	A buffer block where the page is created
[in]	mtr	Mini-transaction handle
[in]	comp	nonzero=compact page format
[in]	page_type	Page type

Returns: pointer to the page

◆ page_create_empty()

void page_create_empty	(	buf_block_t *	block,
		dict_index_t *	index,
		mtr_t *	mtr
	)

Empty a previously created B-tree index page.

Parameters

[in,out]	block	B-tree block
[in]	index	The index of the page
[in,out]	mtr	Mini-transaction

◆ page_create_low()

static page_t * page_create_low	(	buf_block_t *	block,
		ulint	comp,
		page_type_t	page_type
	)

static

The index page creation function.

Parameters

[in,out]	block	a buffer block where the page is created
[in]	comp	nonzero=compact page format
[in]	page_type	page type

Returns: pointer to the page

◆ page_create_write_log()

static void page_create_write_log	(	buf_frame_t *	frame,
		mtr_t *	mtr,
		bool	comp,
		page_type_t	page_type
	)

inlinestatic

Writes a log record of page creation.

Parameters

[in]	frame	A buffer frame where the page is created
[in]	mtr	Mini-transaction handle
[in]	comp	true=compact page format
[in]	page_type	Page type

◆ page_create_zip()

page_t * page_create_zip	(	buf_block_t *	block,
		dict_index_t *	index,
		ulint	level,
		trx_id_t	max_trx_id,
		mtr_t *	mtr,
		page_type_t	page_type
	)

Create a compressed B-tree index page.

Parameters

[in,out]	block	Buffer frame where the page is created
[in]	index	Index of the page, or NULL when applying TRUNCATE log record during recovery
[in]	level	The B-tree level of the page
[in]	max_trx_id	PAGE_MAX_TRX_ID
[in]	mtr	Mini-transaction handle
[in]	page_type	Page type to be created. Only FIL_PAGE_INDEX, FIL_PAGE_RTREE, FIL_PAGE_SDI allowed

Returns: pointer to the page

◆ page_delete_rec()

bool page_delete_rec	(	const dict_index_t *	index,
		page_cur_t *	pcur,
		const ulint *	offsets
	)

Removes the record from a leaf page.

This function does not log any changes. It is used by the IMPORT tablespace functions.

Returns: true if success, i.e., the page did not become too empty

Parameters

[in]	index	The index that the record belongs to.
[in,out]	pcur	Page cursor on record to delete.
[in]	offsets	Offsets for record.

◆ page_delete_rec_list_end()

void page_delete_rec_list_end	(	rec_t *	rec,
		buf_block_t *	block,
		dict_index_t *	index,
		ulint	n_recs,
		ulint	size,
		mtr_t *	mtr
	)

Deletes records from a page from a given record onward, including that record.

The infimum and supremum records are not deleted.

Parameters

rec	in: pointer to record on page
block	in: buffer block of the page
index	in: record descriptor
n_recs	in: number of records to delete, or ULINT_UNDEFINED if not known
size	in: the sum of the sizes of the records in the end of the chain to delete, or ULINT_UNDEFINED if not known
mtr	in: mtr

◆ page_delete_rec_list_start()

void page_delete_rec_list_start	(	rec_t *	rec,
		buf_block_t *	block,
		dict_index_t *	index,
		mtr_t *	mtr
	)

Deletes records from page, up to the given record, NOT including that record.

Infimum and supremum records are not deleted.

Parameters

rec	in: record on page
block	in: buffer block of the page
index	in: record descriptor
mtr	in: mtr

◆ page_delete_rec_list_write_log()

static void page_delete_rec_list_write_log	(	rec_t *	rec,
		dict_index_t *	index,
		mlog_id_t	type,
		mtr_t *	mtr
	)

inlinestatic

Writes a log record of a record list end or start deletion.

Parameters

rec	in: record on page
index	in: record descriptor
type	in: operation type: MLOG_LIST_END_DELETE, ...
mtr	in: mtr

◆ page_dir_add_slot()

static void page_dir_add_slot	(	page_t *	page,
		page_zip_des_t *	page_zip,
		ulint	start
	)

inlinestatic

Used to add n slots to the directory.

Does not set the record pointers in the added slots or update n_owned values: this is the responsibility of the caller.

Parameters

page	in/out: the index page
page_zip	in/out: comprssed page, or NULL
start	in: the slot above which the new slots are added

◆ page_dir_balance_slot()

void page_dir_balance_slot	(	page_t *	page,
		page_zip_des_t *	page_zip,
		ulint	slot_no
	)

Tries to balance the given directory slot with too few records with the upper neighbor, so that there are at least the minimum number of records owned by the slot; this may result in the merging of two slots.

Parameters

[in,out]	page	Index page
[in,out]	page_zip	Compressed page, or null
[in]	slot_no	The directory slot

◆ page_dir_delete_slot()

static void page_dir_delete_slot	(	page_t *	page,
		page_zip_des_t *	page_zip,
		ulint	slot_no
	)

inlinestatic

Used to delete n slots from the directory.

This function updates also n_owned fields in the records, so that the first slot after the deleted ones inherits the records of the deleted slots.

Parameters

page	in/out: the index page
page_zip	in/out: compressed page, or NULL
slot_no	in: slot to be deleted

◆ page_dir_find_owner_slot()

ulint page_dir_find_owner_slot ( const rec_t * rec )

Looks for the directory slot which owns the given record.

Returns: the directory slot number

Parameters

rec	in: the physical record

◆ page_dir_slot_check()

static bool page_dir_slot_check ( const page_dir_slot_t * slot )

static

Used to check the consistency of a directory slot.

Returns: true if succeed

Parameters

slot in: slot

◆ page_dir_split_slot()

void page_dir_split_slot	(	page_t *	page,
		page_zip_des_t *	page_zip,
		ulint	slot_no
	)

Splits a directory slot which owns too many records.

Parameters

[in,out]	page	Index page
[in,out]	page_zip	Compressed page whose uncompressed part will be written, or null
[in]	slot_no	The directory slot

◆ page_find_rec_last_not_deleted()

const rec_t * page_find_rec_last_not_deleted ( const page_t * page )

Get the last non-delete-marked record on a page.

Parameters

[in] page index tree leaf page

Returns: the last record, not delete-marked

Return values

infimum record if all records are delete-marked

◆ page_find_rec_with_heap_no()

const rec_t * page_find_rec_with_heap_no	(	const page_t *	page,
		ulint	heap_no
	)

Looks in the page record list for a record with the given heap number.

Returns: record, NULL if not found

Parameters

page	in: index page
heap_no	in: heap number

◆ page_is_spatial_non_leaf()

bool page_is_spatial_non_leaf	(	const rec_t *	rec,
		dict_index_t *	index
	)

This function checks if the page in which record is present is a non-leaf node of a spatial index.

param[in] rec Btree record param[in] index index

Returns: true if ok

◆ page_mem_alloc_heap()

byte * page_mem_alloc_heap	(	page_t *	page,
		page_zip_des_t *	page_zip,
		ulint	need,
		ulint *	heap_no
	)

Allocates a block of memory from the heap of an index page.

Returns: pointer to start of allocated buffer, or NULL if allocation fails

Parameters

page	in/out: index page
page_zip	in/out: compressed page with enough space available for inserting the record, or NULL
need	in: total number of bytes needed
heap_no	out: this contains the heap number of the allocated record if allocation succeeds

◆ page_move_rec_list_end()

bool page_move_rec_list_end	(	buf_block_t *	new_block,
		buf_block_t *	block,
		rec_t *	split_rec,
		dict_index_t *	index,
		mtr_t *	mtr
	)

Moves record list end to another page.

Moved records include split_rec.

IMPORTANT: The caller will have to update IBUF_BITMAP_FREE if new_block is a compressed leaf page in a secondary index. This has to be done either within the same mini-transaction, or by invoking ibuf_reset_free_bits() before mtr_commit().

Returns: true on success; false on compression failure (new_block will be decompressed)

Parameters

new_block	in/out: index page where to move
block	in: index page from where to move
split_rec	in: first record to move
index	in: record descriptor
mtr	in: mtr

◆ page_move_rec_list_start()

bool page_move_rec_list_start	(	buf_block_t *	new_block,
		buf_block_t *	block,
		rec_t *	split_rec,
		dict_index_t *	index,
		mtr_t *	mtr
	)

Moves record list start to another page.

Moved records do not include split_rec.

IMPORTANT: The caller will have to update IBUF_BITMAP_FREE if new_block is a compressed leaf page in a secondary index. This has to be done either within the same mini-transaction, or by invoking ibuf_reset_free_bits() before mtr_commit().

Returns: true on success; false on compression failure

Parameters

new_block	in/out: index page where to move
block	in/out: page containing split_rec
split_rec	in: first record not to move
index	in: record descriptor
mtr	in: mtr

◆ page_parse_create()

void page_parse_create	(	buf_block_t *	block,
		ulint	comp,
		page_type_t	page_type
	)

Parses a redo log record of creating a page.

Parameters

[in,out]	block	buffer block, or NULL
[in]	comp	nonzero=compact page format
[in]	page_type	page type (FIL_PAGE_INDEX, FIL_PAGE_RTREE or FIL_PAGE_SDI)

◆ page_parse_delete_rec_list()

byte * page_parse_delete_rec_list	(	mlog_id_t	type,
		byte *	ptr,
		byte *	end_ptr,
		buf_block_t *	block,
		dict_index_t *	index,
		mtr_t *	mtr
	)

Parses a log record of a record list end or start deletion.

Returns: end of log record or NULL

Parameters

type	in: MLOG_LIST_END_DELETE, MLOG_LIST_START_DELETE, MLOG_COMP_LIST_END_DELETE or MLOG_COMP_LIST_START_DELETE
ptr	in: buffer
end_ptr	in: buffer end
block	in/out: buffer block or NULL
index	in: record descriptor
mtr	in: mtr or NULL

◆ page_rec_get_n_recs_before()

ulint page_rec_get_n_recs_before ( const rec_t * rec )

Returns the number of records before the given record in chain.

The number includes infimum and supremum records.

Returns: number of records

Parameters

rec	in: the physical record

◆ page_rec_get_nth_const()

const rec_t * page_rec_get_nth_const	(	const page_t *	page,
		ulint	nth
	)

Returns the nth record of the record list.

This is the inverse function of page_rec_get_n_recs_before().

Returns: nth record

Parameters

page	in: page
nth	in: nth record

◆ page_rec_print()

void page_rec_print	(	const rec_t *	rec,
		const ulint *	offsets
	)

Prints record contents including the data relevant only in the index page context.

Parameters

[in]	rec	Physical record
[in]	offsets	Record descriptor

◆ page_rec_validate()

bool page_rec_validate	(	const rec_t *	rec,
		const ulint *	offsets
	)

The following is used to validate a record on a page.

This function differs from rec_validate as it can also check the n_owned field and the heap_no field.

Returns: true if ok

Parameters

rec	in: physical record
offsets	in: array returned by rec_get_offsets()

◆ page_set_max_trx_id()

void page_set_max_trx_id	(	buf_block_t *	block,
		page_zip_des_t *	page_zip,
		trx_id_t	trx_id,
		mtr_t *	mtr
	)

Sets the max trx id field value.

Parameters

[in,out]	block	Page
[in,out]	page_zip	Compressed page, or NULL
[in]	trx_id	Transaction id
[in,out]	mtr	Mini-transaction, or NULL

◆ page_simple_validate_new()

bool page_simple_validate_new ( const page_t * page )

This function checks the consistency of an index page when we do not know the index.

This is also resilient so that this should never crash even if the page is total garbage.

Returns: true if ok

Parameters

page	in: index page in ROW_FORMAT!=REDUNDANT

◆ page_simple_validate_old()

bool page_simple_validate_old ( const page_t * page )

This function checks the consistency of an index page when we do not know the index.

This is also resilient so that this should never crash even if the page is total garbage.

Returns: true if ok

Parameters

page	in: index page in ROW_FORMAT=REDUNDANT

◆ page_validate()

bool page_validate	(	const page_t *	page,
		dict_index_t *	index
	)

This function checks the consistency of an index page.

Parameters

[in]	page	index page
[in]	index	data dictionary index containing the page record type definition

Returns: true if ok

◆ page_warn_strict_checksum()

void page_warn_strict_checksum	(	srv_checksum_algorithm_t	curr_algo,
		srv_checksum_algorithm_t	page_checksum,
		const page_id_t &	page_id
	)

Issue a warning when the checksum that is stored in the page is valid, but different than the global setting innodb_checksum_algorithm.

Parameters

[in]	curr_algo	current checksum algorithm
[in]	page_checksum	page valid checksum
[in]	page_id	page identifier

Variable Documentation

◆ infimum_supremum_compact

const byte infimum_supremum_compact[]

static

Initial value:

= {
    
    0x01 , 0x00, 0x02 , 0x00,
    0x0d , 'i', 'n', 'f', 'i', 'm', 'u', 'm', 0,
    
    0x01 , 0x00, 0x0b , 0x00,
    0x00 , 's', 'u', 'p', 'r', 'e', 'm', 'u', 'm'}

The page infimum and supremum of an empty page in ROW_FORMAT=COMPACT.

◆ infimum_supremum_redundant

const byte infimum_supremum_redundant[]

static

Initial value:

= {
    
    0x08 , 0x01 , 0x00, 0x00 ,
    0x03 , 0x00, 0x74 ,
    'i', 'n', 'f', 'i', 'm', 'u', 'm', 0,
    
    0x09 , 0x01 , 0x00, 0x08 ,
    0x03 , 0x00, 0x00 ,
    's', 'u', 'p', 'r', 'e', 'm', 'u', 'm', 0}

The page infimum and supremum of an empty page in ROW_FORMAT=REDUNDANT.

Functions

Variables

Detailed Description

Function Documentation

◆ page_check_dir()

◆ page_copy_rec_list_end()

◆ page_copy_rec_list_end_no_locks()

◆ page_copy_rec_list_start()

◆ page_create()

◆ page_create_empty()

◆ page_create_low()

◆ page_create_write_log()

◆ page_create_zip()

◆ page_delete_rec()

◆ page_delete_rec_list_end()

◆ page_delete_rec_list_start()

◆ page_delete_rec_list_write_log()

◆ page_dir_add_slot()

◆ page_dir_balance_slot()

◆ page_dir_delete_slot()

◆ page_dir_find_owner_slot()

◆ page_dir_slot_check()

◆ page_dir_split_slot()

◆ page_find_rec_last_not_deleted()

◆ page_find_rec_with_heap_no()

◆ page_is_spatial_non_leaf()

◆ page_mem_alloc_heap()

◆ page_move_rec_list_end()

◆ page_move_rec_list_start()

◆ page_parse_create()

◆ page_parse_delete_rec_list()

◆ page_rec_get_n_recs_before()

◆ page_rec_get_nth_const()

◆ page_rec_print()

◆ page_rec_validate()

◆ page_set_max_trx_id()

◆ page_simple_validate_new()

◆ page_simple_validate_old()

◆ page_validate()

◆ page_warn_strict_checksum()

Variable Documentation

◆ infimum_supremum_compact

◆ infimum_supremum_redundant