mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-02-05 13:22:17 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions
mariadb/storage/innobase/include/row0row.h
Marko Mäkelä b8a6719889 MDEV-26642/MDEV-26643/MDEV-32898 Implement innodb_snapshot_isolation 
https://jepsen.io/analyses/mysql-8.0.34 highlights that the
transaction isolation levels in the InnoDB storage engine do not
correspond to any widely accepted definitions, such as
"Generalized Isolation Level Definitions"
https://pmg.csail.mit.edu/papers/icde00.pdf
(PL-1 = READ UNCOMMITTED, PL-2 = READ COMMITTED, PL-2.99 = REPEATABLE READ,
PL-3 = SERIALIZABLE).
Only READ UNCOMMITTED in InnoDB seems to match the above definition.

The issue is that InnoDB does not detect write/write conflicts
(Section 4.4.3, Definition 6) in the above.

It appears that as soon as we implement write/write conflict detection
(SET SESSION innodb_snapshot_isolation=ON), the default isolation level
(SET TRANSACTION ISOLATION LEVEL REPEATABLE READ) will become
Snapshot Isolation (similar to Postgres), as defined in Section 4.2 of
"A Critique of ANSI SQL Isolation Levels", MSR-TR-95-51, June 1995
https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/tr-95-51.pdf

Locking reads inside InnoDB used to read the latest committed version,
ignoring what should actually be visible to the transaction.
The added test innodb.lock_isolation illustrates this. The statement
	UPDATE t SET a=3 WHERE b=2;
is executed in a transaction that was started before a read view or
a snapshot of the current transaction was created, and committed before
the current transaction attempts to execute
	UPDATE t SET b=3;
If SET innodb_snapshot_isolation=ON is in effect when the second
transaction was started, the second transaction will be aborted with
the error ER_CHECKREAD. By default (innodb_snapshot_isolation=OFF),
the second transaction would execute inconsistently, displaying an
incorrect SELECT COUNT(*) FROM t in its read view.

If innodb_snapshot_isolation=ON, if an attempt to acquire a lock on a
record that does not exist in the current read view is made, an error
DB_RECORD_CHANGED (HA_ERR_RECORD_CHANGED, ER_CHECKREAD) will
be raised. This error will be treated in the same way as a deadlock:
the transaction will be rolled back.

lock_clust_rec_read_check_and_lock(): If the current transaction has
a read view where the record is not visible and
innodb_snapshot_isolation=ON, fail before trying to acquire the lock.

row_sel_build_committed_vers_for_mysql(): If innodb_snapshot_isolation=ON,
disable the "semi-consistent read" logic that had been implemented by
myself on the directions of Heikki Tuuri in order to address
https://bugs.mysql.com/bug.php?id=3300 that was motivated by a customer
wanting UPDATE to skip locked rows that do not match the WHERE condition.
It looks like my changes were included in the MySQL 5.1.5
commit ad126d90e019f223470e73e1b2b528f9007c4532; at that time, employees
of Innobase Oy (a recent acquisition of Oracle) had lost write access to
the repository.

The only reason why we set innodb_snapshot_isolation=OFF by default is
backward compatibility with applications, such as the one that motivated
the implementation of "semi-consistent read" back in 2005. In a later
major release, we can default to innodb_snapshot_isolation=ON.

Thanks to Peter Alvaro, Kyle Kingsbury and Alexey Gotsman for their work
on https://github.com/jepsen-io/ and to Kyle and Alexey for explanations
and some testing of this fix.

Thanks to Vladislav Lesin for the initial test for MDEV-26643,
as well as reviewing these changes.
2024-03-20 09:48:03 +02:00

437 lines
16 KiB
C
Raw Permalink Blame History

/*****************************************************************************
Copyright (c) 1996, 2016, Oracle and/or its affiliates. All Rights Reserved.
Copyright (c) 2016, 2022, MariaDB Corporation.
This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1335 USA
*****************************************************************************/
/**************************************************//**
@file include/row0row.h
General row routines
Created 4/20/1996 Heikki Tuuri
*******************************************************/
#ifndef row0row_h
#define row0row_h
#include "que0types.h"
#include "ibuf0ibuf.h"
#include "trx0types.h"
#include "mtr0mtr.h"
#include "rem0types.h"
#include "row0types.h"
#include "btr0types.h"
/*********************************************************************//**
Gets the offset of the DB_TRX_ID field, in bytes relative to the origin of
a clustered index record.
@return offset of DATA_TRX_ID */
UNIV_INLINE
ulint
row_get_trx_id_offset(
/*==================*/
const dict_index_t* index, /*!< in: clustered index */
const rec_offs* offsets)/*!< in: record offsets */
MY_ATTRIBUTE((nonnull, warn_unused_result));
/*********************************************************************//**
Reads the trx id field from a clustered index record.
@return value of the field */
UNIV_INLINE
trx_id_t
row_get_rec_trx_id(
/*===============*/
const rec_t* rec, /*!< in: record */
const dict_index_t* index, /*!< in: clustered index */
const rec_offs* offsets)/*!< in: rec_get_offsets(rec, index) */
MY_ATTRIBUTE((nonnull, warn_unused_result));
/*********************************************************************//**
Reads the roll pointer field from a clustered index record.
@return value of the field */
UNIV_INLINE
roll_ptr_t
row_get_rec_roll_ptr(
/*=================*/
const rec_t* rec, /*!< in: record */
const dict_index_t* index, /*!< in: clustered index */
const rec_offs* offsets)/*!< in: rec_get_offsets(rec, index) */
MY_ATTRIBUTE((nonnull, warn_unused_result));
/* Flags for row build type. */
#define ROW_BUILD_NORMAL 0 /*!< build index row */
#define ROW_BUILD_FOR_PURGE 1 /*!< build row for purge. */
#define ROW_BUILD_FOR_UNDO 2 /*!< build row for undo. */
#define ROW_BUILD_FOR_INSERT 3 /*!< build row for insert. */
/*****************************************************************//**
When an insert or purge to a table is performed, this function builds
the entry to be inserted into or purged from an index on the table.
@return index entry which should be inserted or purged
@retval NULL if the externally stored columns in the clustered index record
are unavailable and ext != NULL, or row is missing some needed columns. */
dtuple_t*
row_build_index_entry_low(
/*======================*/
const dtuple_t* row, /*!< in: row which should be
inserted or purged */
const row_ext_t* ext, /*!< in: externally stored column
prefixes, or NULL */
const dict_index_t* index, /*!< in: index on the table */
mem_heap_t* heap, /*!< in,out: memory heap from which
the memory for the index entry
is allocated */
ulint flag) /*!< in: ROW_BUILD_NORMAL,
ROW_BUILD_FOR_PURGE
or ROW_BUILD_FOR_UNDO */
MY_ATTRIBUTE((warn_unused_result, nonnull(1,3,4)));
/*****************************************************************//**
When an insert or purge to a table is performed, this function builds
the entry to be inserted into or purged from an index on the table.
@return index entry which should be inserted or purged, or NULL if the
externally stored columns in the clustered index record are
unavailable and ext != NULL */
UNIV_INLINE
dtuple_t*
row_build_index_entry(
/*==================*/
const dtuple_t* row, /*!< in: row which should be
inserted or purged */
const row_ext_t* ext, /*!< in: externally stored column
prefixes, or NULL */
const dict_index_t* index, /*!< in: index on the table */
mem_heap_t* heap) /*!< in,out: memory heap from which
the memory for the index entry
is allocated */
MY_ATTRIBUTE((warn_unused_result, nonnull(1,3,4)));
/*******************************************************************//**
An inverse function to row_build_index_entry. Builds a row from a
record in a clustered index.
@return own: row built; see the NOTE below! */
dtuple_t*
row_build(
/*======*/
ulint type, /*!< in: ROW_COPY_POINTERS or
ROW_COPY_DATA; the latter
copies also the data fields to
heap while the first only
places pointers to data fields
on the index page, and thus is
more efficient */
const dict_index_t* index, /*!< in: clustered index */
const rec_t* rec, /*!< in: record in the clustered
index; NOTE: in the case
ROW_COPY_POINTERS the data
fields in the row will point
directly into this record,
therefore, the buffer page of
this record must be at least
s-latched and the latch held
as long as the row dtuple is used! */
const rec_offs* offsets,/*!< in: rec_get_offsets(rec,index)
or NULL, in which case this function
will invoke rec_get_offsets() */
const dict_table_t* col_table,
/*!< in: table, to check which
externally stored columns
occur in the ordering columns
of an index, or NULL if
index->table should be
consulted instead; the user
columns in this table should be
the same columns as in index->table */
const dtuple_t* defaults,
/*!< in: default values of
added, changed columns, or NULL */
const ulint* col_map,/*!< in: mapping of old column
numbers to new ones, or NULL */
row_ext_t** ext, /*!< out, own: cache of
externally stored column
prefixes, or NULL */
mem_heap_t* heap); /*!< in: memory heap from which
the memory needed is allocated */
/** An inverse function to row_build_index_entry. Builds a row from a
record in a clustered index, with possible indexing on ongoing
addition of new virtual columns.
@param[in] type ROW_COPY_POINTERS or ROW_COPY_DATA;
@param[in] index clustered index
@param[in] rec record in the clustered index
@param[in] offsets rec_get_offsets(rec,index) or NULL
@param[in] col_table table, to check which
externally stored columns
occur in the ordering columns
of an index, or NULL if
index->table should be
consulted instead
@param[in] defaults default values of added, changed columns, or NULL
@param[in] add_v new virtual columns added
along with new indexes
@param[in] col_map mapping of old column
numbers to new ones, or NULL
@param[in] ext cache of externally stored column
prefixes, or NULL
@param[in] heap memory heap from which
the memory needed is allocated
@return own: row built */
dtuple_t*
row_build_w_add_vcol(
ulint type,
const dict_index_t* index,
const rec_t* rec,
const rec_offs* offsets,
const dict_table_t* col_table,
const dtuple_t* defaults,
const dict_add_v_col_t* add_v,
const ulint* col_map,
row_ext_t** ext,
mem_heap_t* heap);
/*******************************************************************//**
Converts an index record to a typed data tuple.
@return index entry built; does not set info_bits, and the data fields
in the entry will point directly to rec */
dtuple_t*
row_rec_to_index_entry_low(
/*=======================*/
const rec_t* rec, /*!< in: record in the index */
const dict_index_t* index, /*!< in: index */
const rec_offs* offsets,/*!< in: rec_get_offsets(rec, index) */
mem_heap_t* heap) /*!< in: memory heap from which
the memory needed is allocated */
MY_ATTRIBUTE((warn_unused_result));
/*******************************************************************//**
Converts an index record to a typed data tuple. NOTE that externally
stored (often big) fields are NOT copied to heap.
@return own: index entry built */
dtuple_t*
row_rec_to_index_entry(
/*===================*/
const rec_t* rec, /*!< in: record in the index */
const dict_index_t* index, /*!< in: index */
const rec_offs* offsets,/*!< in/out: rec_get_offsets(rec) */
mem_heap_t* heap) /*!< in: memory heap from which
the memory needed is allocated */
MY_ATTRIBUTE((warn_unused_result));
/** Convert a metadata record to a data tuple.
@param[in] rec metadata record
@param[in] index clustered index after instant ALTER TABLE
@param[in] offsets rec_get_offsets(rec)
@param[in,out] heap memory heap for allocations
@param[in] info_bits the info_bits after an update
@param[in] pad whether to pad to index->n_fields */
dtuple_t*
row_metadata_to_tuple(
const rec_t* rec,
const dict_index_t* index,
const rec_offs* offsets,
mem_heap_t* heap,
ulint info_bits,
bool pad)
MY_ATTRIBUTE((nonnull,warn_unused_result));
/*******************************************************************//**
Builds from a secondary index record a row reference with which we can
search the clustered index record.
@return own: row reference built; see the NOTE below! */
dtuple_t*
row_build_row_ref(
/*==============*/
ulint type, /*!< in: ROW_COPY_DATA, or ROW_COPY_POINTERS:
the former copies also the data fields to
heap, whereas the latter only places pointers
to data fields on the index page */
dict_index_t* index, /*!< in: secondary index */
const rec_t* rec, /*!< in: record in the index;
NOTE: in the case ROW_COPY_POINTERS
the data fields in the row will point
directly into this record, therefore,
the buffer page of this record must be
at least s-latched and the latch held
as long as the row reference is used! */
mem_heap_t* heap) /*!< in: memory heap from which the memory
needed is allocated */
MY_ATTRIBUTE((warn_unused_result));
/*******************************************************************//**
Builds from a secondary index record a row reference with which we can
search the clustered index record. */
void
row_build_row_ref_in_tuple(
/*=======================*/
dtuple_t* ref, /*!< in/out: row reference built;
see the NOTE below! */
const rec_t* rec, /*!< in: record in the index;
NOTE: the data fields in ref
will point directly into this
record, therefore, the buffer
page of this record must be at
least s-latched and the latch
held as long as the row
reference is used! */
const dict_index_t* index, /*!< in: secondary index */
rec_offs* offsets)/*!< in: rec_get_offsets(rec, index)
or NULL */
MY_ATTRIBUTE((nonnull(1,2,3)));
/*******************************************************************//**
Builds from a secondary index record a row reference with which we can
search the clustered index record. */
UNIV_INLINE
void
row_build_row_ref_fast(
/*===================*/
dtuple_t* ref, /*!< in/out: typed data tuple where the
reference is built */
const ulint* map, /*!< in: array of field numbers in rec
telling how ref should be built from
the fields of rec */
const rec_t* rec, /*!< in: secondary index record;
must be preserved while ref is used, as we do
not copy field values to heap */
const rec_offs* offsets);/*!< in: array returned by rec_get_offsets() */
/***************************************************************//**
Searches the clustered index record for a row, if we have the row
reference.
@return true if found */
bool
row_search_on_row_ref(
/*==================*/
btr_pcur_t* pcur, /*!< out: persistent cursor, which must
be closed by the caller */
btr_latch_mode mode, /*!< in: BTR_MODIFY_LEAF, ... */
const dict_table_t* table, /*!< in: table */
const dtuple_t* ref, /*!< in: row reference */
mtr_t* mtr) /*!< in/out: mtr */
MY_ATTRIBUTE((nonnull, warn_unused_result));
/*********************************************************************//**
Fetches the clustered index record for a secondary index record. The latches
on the secondary index record are preserved.
@return record or NULL, if no record found */
rec_t*
row_get_clust_rec(
/*==============*/
btr_latch_mode mode, /*!< in: BTR_MODIFY_LEAF, ... */
const rec_t* rec, /*!< in: record in a secondary index */
dict_index_t* index, /*!< in: secondary index */
dict_index_t** clust_index,/*!< out: clustered index */
mtr_t* mtr) /*!< in: mtr */
MY_ATTRIBUTE((nonnull, warn_unused_result));
/** Parse the integer data from specified data, which could be
DATA_INT, DATA_FLOAT or DATA_DOUBLE. If the value is less than 0
and the type is not unsigned then we reset the value to 0
@param[in] data data to read
@param[in] len length of data
@param[in] mtype mtype of data
@param[in] unsigned_type if the data is unsigned
@return the integer value from the data */
inline
ib_uint64_t
row_parse_int(
const byte* data,
ulint len,
ulint mtype,
bool unsigned_type);
/** Result of row_search_index_entry */
enum row_search_result {
ROW_FOUND = 0, /*!< the record was found */
ROW_NOT_FOUND, /*!< record not found */
ROW_BUFFERED, /*!< one of BTR_INSERT, BTR_DELETE, or
BTR_DELETE_MARK was specified, the
secondary index leaf page was not in
the buffer pool, and the operation was
enqueued in the insert/delete buffer */
ROW_NOT_DELETED_REF /*!< BTR_DELETE was specified, and
row_purge_poss_sec() failed */
};
/***************************************************************//**
Searches an index record.
@return whether the record was found or buffered */
enum row_search_result
row_search_index_entry(
/*===================*/
const dtuple_t* entry, /*!< in: index entry */
btr_latch_mode mode, /*!< in: BTR_MODIFY_LEAF, ... */
btr_pcur_t* pcur, /*!< in/out: persistent cursor, which must
be closed by the caller */
mtr_t* mtr) /*!< in: mtr */
MY_ATTRIBUTE((nonnull, warn_unused_result));
/** Get the byte offset of the DB_TRX_ID column
@param[in] rec clustered index record
@param[in] index clustered index
@return the byte offset of DB_TRX_ID, from the start of rec */
ulint row_trx_id_offset(const rec_t* rec, const dict_index_t* index);
#define ROW_COPY_DATA 1
#define ROW_COPY_POINTERS 2
/* The allowed latching order of index records is the following:
(1) a secondary index record ->
(2) the clustered index record ->
(3) rollback segment data for the clustered index record. */
/*******************************************************************//**
Formats the raw data in "data" (in InnoDB on-disk format) using
"dict_field" and writes the result to "buf".
Not more than "buf_size" bytes are written to "buf".
The result is always NUL-terminated (provided buf_size is positive) and the
number of bytes that were written to "buf" is returned (including the
terminating NUL).
@return number of bytes that were written */
ulint
row_raw_format(
/*===========*/
const char* data, /*!< in: raw data */
ulint data_len, /*!< in: raw data length
in bytes */
const dict_field_t* dict_field, /*!< in: index field */
char* buf, /*!< out: output buffer */
ulint buf_size) /*!< in: output buffer size
in bytes */
MY_ATTRIBUTE((nonnull, warn_unused_result));
/** Prepare to start a mini-transaction to modify an index.
@param[in,out] mtr mini-transaction
@param[in,out] index possibly secondary index
@param[in] pessimistic whether this is a pessimistic operation */
inline
void
row_mtr_start(mtr_t* mtr, dict_index_t* index, bool pessimistic)
{
mtr->start();
switch (index->table->space_id) {
case IBUF_SPACE_ID:
if (pessimistic
&& !(index->type & (DICT_UNIQUE | DICT_SPATIAL))) {
ibuf_free_excess_pages();
}
break;
case SRV_TMP_SPACE_ID:
mtr->set_log_mode(MTR_LOG_NO_REDO);
break;
default:
index->set_modified(*mtr);
break;
}
log_free_check();
}
#include "row0row.inl"
#endif
Reference in a new issue View git blame Copy permalink