Commit 78c6a8ca authored by Dmitry Lenev's avatar Dmitry Lenev

A 5.1-only version of fix for bug #46947 "Embedded SELECT

without FOR UPDATE is causing a lock".

SELECT statements with subqueries referencing InnoDB tables
were acquiring shared locks on rows in these tables when they
were executed in REPEATABLE-READ mode and with statement or
mixed mode binary logging turned on.

This was a regression which were introduced when fixing
bug 39843.

The problem was that for tables belonging to subqueries
parser set TL_READ_DEFAULT as a lock type. In cases when
statement/mixed binary logging at open_tables() time this
type of lock was converted to TL_READ_NO_INSERT lock at
open_tables() time and caused InnoDB engine to acquire
shared locks on reads from these tables. Although in some
cases such behavior was correct (e.g. for subqueries in
DELETE) in case of SELECT it has caused unnecessary locking.

This patch implements minimal version of the fix for the
specific problem described in the bug-report which supposed
to be not too risky for pushing into 5.1 tree.
The 5.5 tree already contains a more appropriate solution
which also addresses other related issues like bug 53921
"Wrong locks for SELECTs used stored functions may lead
to broken SBR".

This patch tries to solve the problem by ensuring that
TL_READ_DEFAULT lock which is set in the parser for
tables participating in subqueries at open_tables()
time is interpreted as TL_READ_NO_INSERT or TL_READ.
TL_READ is used only if we know that this is a SELECT
and that this particular table is not used by a stored
function.

Test coverage is added for both InnoDB and MyISAM.

This patch introduces an "incompatible" change in locking
scheme for subqueries used in SELECT ... FOR UPDATE and
SELECT .. IN SHARE MODE.

In 4.1 (as well as in 5.0 and 5.1 before fix for bug 39843)
the server would use a snapshot InnoDB read for subqueries
in SELECT FOR UPDATE and SELECT .. IN SHARE MODE statements,
regardless of whether the binary log is on or off.

If the user required a different type of read (i.e. locking
read), he/she could request so explicitly by providing FOR
UPDATE/IN SHARE MODE clause for each individual subquery.

The patch for bug 39843 broke this behaviour (which was not
documented or tested), and started to use locking reads for
all subqueries in SELECT ... FOR UPDATE/IN SHARE MODE.
This patch restores 4.1 behaviour.

This patch should be mostly null-merged into 5.5 tree.
parent fa3570f9
#
# SUMMARY
# Check if statement reading table '$table' allows concurrent
# inserts in it.
#
# PARAMETERS
# $table Table in which concurrent inserts should be allowed.
# $con_aux1 Name of the first auxiliary connection to be used by this
# script.
# $con_aux2 Name of the second auxiliary connection to be used by this
# script.
# $statement Statement to be checked.
# $restore_table Table which might be modified by statement to be checked
# and thus needs backing up before its execution and
# restoring after it (can be empty).
#
# EXAMPLE
# lock_sync.test
#
--disable_result_log
--disable_query_log
# Reset DEBUG_SYNC facility for safety.
set debug_sync= "RESET";
if (`SELECT '$restore_table' <> ''`)
{
--eval create temporary table t_backup select * from $restore_table;
}
connection $con_aux1;
set debug_sync='after_lock_tables_takes_lock SIGNAL parked WAIT_FOR go';
--send_eval $statement;
connection $con_aux2;
set debug_sync='now WAIT_FOR parked';
--send_eval insert into $table (i) values (0);
--enable_result_log
--enable_query_log
connection default;
# Wait until concurrent insert is successfully executed while
# statement being checked has its tables locked.
# We use wait_condition.inc instead of simply reaping
# concurrent insert here in order to avoid deadlocks if test
# fails and to time out gracefully instead.
let $wait_condition=
select count(*) = 0 from information_schema.processlist
where info = "insert into $table (i) values (0)";
--source include/wait_condition.inc
--disable_result_log
--disable_query_log
if ($success)
{
# Apparently concurrent insert was successfully executed.
# To be safe against wait_condition.inc succeeding due to
# races let us first reap concurrent insert to ensure that
# it has really been successfully executed.
connection $con_aux2;
--reap
connection default;
set debug_sync= 'now SIGNAL go';
connection $con_aux1;
--reap
connection default;
--echo Success: '$statement' allows concurrent inserts into '$table'.
}
if (!$success)
{
# Waiting has timed out. Apparently concurrent insert was blocked.
# So to be able to continue we need to end our statement first.
set debug_sync= 'now SIGNAL go';
connection $con_aux1;
--reap
connection $con_aux2;
--reap
connection default;
--echo Error: '$statement' doesn't allow concurrent inserts into '$table'!
}
--eval delete from $table where i = 0;
if (`SELECT '$restore_table' <> ''`)
{
--eval truncate table $restore_table;
--eval insert into $restore_table select * from t_backup;
drop temporary table t_backup;
}
# Clean-up. Reset DEBUG_SYNC facility after use.
set debug_sync= "RESET";
--enable_result_log
--enable_query_log
#
# SUMMARY
# Check that statement reading table '$table' doesn't allow concurrent
# inserts in it.
#
# PARAMETERS
# $table Table in which concurrent inserts should be disallowed.
# $con_aux1 Name of the first auxiliary connection to be used by this
# script.
# $con_aux2 Name of the second auxiliary connection to be used by this
# script.
# $statement Statement to be checked.
# $restore_table Table which might be modified by statement to be checked
# and thus needs backing up before its execution and
# restoring after it (can be empty).
#
# EXAMPLE
# lock_sync.test
#
--disable_result_log
--disable_query_log
# Reset DEBUG_SYNC facility for safety.
set debug_sync= "RESET";
if (`SELECT '$restore_table' <> ''`)
{
--eval create temporary table t_backup select * from $restore_table;
}
connection $con_aux1;
set debug_sync='after_lock_tables_takes_lock SIGNAL parked WAIT_FOR go';
--send_eval $statement;
connection $con_aux2;
set debug_sync='now WAIT_FOR parked';
--send_eval insert into $table (i) values (0);
--enable_result_log
--enable_query_log
connection default;
# Wait until concurrent insert is successfully blocked because
# of our statement.
let $wait_condition=
select count(*) = 1 from information_schema.processlist
where state = "Locked" and info = "insert into $table (i) values (0)";
--source include/wait_condition.inc
--disable_result_log
--disable_query_log
set debug_sync= 'now SIGNAL go';
connection $con_aux1;
--reap
connection $con_aux2;
--reap
connection default;
if ($success)
{
--echo Success: '$statement' doesn't allow concurrent inserts into '$table'.
}
if (!$success)
{
--echo Error: '$statement' allows concurrent inserts into '$table'!
}
--eval delete from $table where i = 0;
if (`SELECT '$restore_table' <> ''`)
{
--eval truncate table $restore_table;
--eval insert into $restore_table select * from t_backup;
drop temporary table t_backup;
}
# Clean-up. Reset DEBUG_SYNC facility after use.
set debug_sync= "RESET";
--enable_result_log
--enable_query_log
#
# SUMMARY
# Check if statement affecting or reading table '$table' doesn't
# take any kind of locks on its rows.
#
# PARAMETERS
# $table Table for which presence of row locks should be checked.
# $con_aux Name of auxiliary connection to be used by this script.
# $statement Statement to be checked.
#
# EXAMPLE
# innodb_mysql_lock2.test
#
--disable_result_log
--disable_query_log
connection default;
begin;
--eval select * from $table for update;
connection $con_aux;
begin;
--send_eval $statement;
--enable_result_log
--enable_query_log
connection default;
# Wait until statement is successfully executed while
# all rows in table are X-locked. This means that it
# does not acquire any row locks.
# We use wait_condition.inc instead of simply reaping
# statement here in order to avoid deadlocks if test
# fails and to time out gracefully instead.
let $wait_condition=
select count(*) = 0 from information_schema.processlist
where info = "$statement";
--source include/wait_condition.inc
--disable_result_log
--disable_query_log
if ($success)
{
# Apparently statement was successfully executed and thus it
# has not required any row locks.
# To be safe against wait_condition.inc succeeding due to
# races let us first reap the statement being checked to
# ensure that it has been successfully executed.
connection $con_aux;
--reap
rollback;
connection default;
rollback;
--echo Success: '$statement' doesn't take row locks on '$table'.
}
if (!$success)
{
# Waiting has timed out. Apparently statement was blocked on
# some row lock. So to be able to continue we need to unlock
# rows first.
rollback;
connection $con_aux;
--reap
rollback;
connection default;
--echo Error: '$statement' takes some row locks on '$table'!
}
--enable_result_log
--enable_query_log
#
# SUMMARY
# Check if statement reading table '$table' takes shared locks
# on some of its rows.
#
# PARAMETERS
# $table Table for which presence of row locks should be checked.
# $con_aux Name of auxiliary connection to be used by this script.
# $statement Statement to be checked.
# $wait_statement Sub-statement which is supposed to acquire locks (should
# be the same as $statement for ordinary statements).
#
# EXAMPLE
# innodb_mysql_lock2.test
#
--disable_result_log
--disable_query_log
connection default;
begin;
--eval select * from $table for update;
connection $con_aux;
begin;
--send_eval $statement;
--enable_result_log
--enable_query_log
connection default;
# Wait until statement is successfully blocked because
# all rows in table are X-locked. This means that at
# least it acquires S-locks on some of rows.
let $wait_condition=
select count(*) = 1 from information_schema.processlist
where state in ("Sending data","statistics", "preparing") and
info = "$wait_statement";
--source include/wait_condition.inc
--disable_result_log
--disable_query_log
rollback;
connection $con_aux;
--reap
rollback;
connection default;
--enable_result_log
--enable_query_log
if ($success)
{
--echo Success: '$statement' takes shared row locks on '$table'.
}
if (!$success)
{
--echo Error: '$statement' hasn't taken shared row locks on '$table'!
}
......@@ -12,7 +12,7 @@ INSERT INTO t2 VALUES (0),(1),(2),(3),(4),(5),(6),(7),(8),(9),(10),
START TRANSACTION;
# in thread2
REPLACE INTO t2 VALUES (-17);
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d);
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d) LOCK IN SHARE MODE;
d
# in thread1
REPLACE INTO t1(a,b) VALUES (67,20);
......@@ -21,10 +21,10 @@ COMMIT;
START TRANSACTION;
REPLACE INTO t1(a,b) VALUES (65,-50);
REPLACE INTO t2 VALUES (-91);
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d);
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d) LOCK IN SHARE MODE;
# in thread1
# should not crash
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d);
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d) LOCK IN SHARE MODE;
ERROR 40001: Deadlock found when trying to get lock; try restarting transaction
# in thread2
d
......
This diff is collapsed.
This diff is collapsed.
......@@ -24,7 +24,7 @@ START TRANSACTION;
connection thread2;
--echo # in thread2
REPLACE INTO t2 VALUES (-17);
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d);
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d) LOCK IN SHARE MODE;
connection thread1;
--echo # in thread1
......@@ -37,14 +37,14 @@ START TRANSACTION;
REPLACE INTO t1(a,b) VALUES (65,-50);
REPLACE INTO t2 VALUES (-91);
send;
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d); #waits
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d) LOCK IN SHARE MODE; #waits
connection thread1;
--echo # in thread1
--echo # should not crash
--error ER_LOCK_DEADLOCK
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d); #crashes
SELECT d FROM t2,t1 WHERE d=(SELECT MAX(a) FROM t1 WHERE t1.a > t2.d) LOCK IN SHARE MODE; #crashes
connection thread2;
--echo # in thread2
......
This diff is collapsed.
This diff is collapsed.
......@@ -1288,7 +1288,8 @@ bool fix_merge_after_open(TABLE_LIST *old_child_list, TABLE_LIST **old_last,
TABLE_LIST *new_child_list, TABLE_LIST **new_last);
bool reopen_table(TABLE *table);
bool reopen_tables(THD *thd,bool get_locks,bool in_refresh);
thr_lock_type read_lock_type_for_table(THD *thd, TABLE *table);
thr_lock_type read_lock_type_for_table(THD *thd, LEX *lex,
TABLE_LIST *table_list);
void close_data_files_and_morph_locks(THD *thd, const char *db,
const char *table_name);
void close_handle_and_leave_table_as_lock(TABLE *table);
......
......@@ -4418,7 +4418,8 @@ bool fix_merge_after_open(TABLE_LIST *old_child_list, TABLE_LIST **old_last,
Return a appropriate read lock type given a table object.
@param thd Thread context
@param table TABLE object for table to be locked
@param lex LEX for the current statement.
@param table_list Table list element for table to be locked.
@remark Due to a statement-based replication limitation, statements such as
INSERT INTO .. SELECT FROM .. and CREATE TABLE .. SELECT FROM need
......@@ -4427,19 +4428,32 @@ bool fix_merge_after_open(TABLE_LIST *old_child_list, TABLE_LIST **old_last,
source table. If such a statement gets applied on the slave before
the INSERT .. SELECT statement finishes, data on the master could
differ from data on the slave and end-up with a discrepancy between
the binary log and table state. Furthermore, this does not apply to
I_S and log tables as it's always unsafe to replicate such tables
under statement-based replication as the table on the slave might
contain other data (ie: general_log is enabled on the slave). The
statement will be marked as unsafe for SBR in decide_logging_format().
the binary log and table state.
This also applies to SELECT/SET/DO statements which use stored
functions. Calls to such functions are going to be logged as a
whole and thus should be serialized against concurrent changes
to tables used by those functions. This can be avoided if functions
only read data but doing so requires more complex analysis than it
is done now (unfortunately, due to bug #53921 "Wrong locks for
SELECTs used stored functions may lead to broken SBR" this rule
is not followed in cases when stored function or trigger use
simple SELECT and not a subselect in their body).
Furthermore, this does not apply to I_S and log tables as it's
always unsafe to replicate such tables under statement-based
replication as the table on the slave might contain other data
(ie: general_log is enabled on the slave). The statement will
be marked as unsafe for SBR in decide_logging_format().
*/
thr_lock_type read_lock_type_for_table(THD *thd, TABLE *table)
thr_lock_type read_lock_type_for_table(THD *thd, LEX *lex,
TABLE_LIST *table_list)
{
bool log_on= mysql_bin_log.is_open() && (thd->options & OPTION_BIN_LOG);
ulong binlog_format= thd->variables.binlog_format;
if ((log_on == FALSE) || (binlog_format == BINLOG_FORMAT_ROW) ||
(table->s->table_category == TABLE_CATEGORY_PERFORMANCE))
(table_list->table->s->table_category == TABLE_CATEGORY_PERFORMANCE) ||
(lex->sql_command == SQLCOM_SELECT &&
! table_list->prelocking_placeholder))
return TL_READ;
else
return TL_READ_NO_INSERT;
......@@ -4735,7 +4749,7 @@ int open_tables(THD *thd, TABLE_LIST **start, uint *counter, uint flags)
tables->table->reginfo.lock_type= thd->update_lock_default;
else if (tables->lock_type == TL_READ_DEFAULT)
tables->table->reginfo.lock_type=
read_lock_type_for_table(thd, tables->table);
read_lock_type_for_table(thd, thd->lex, tables);
else
tables->table->reginfo.lock_type= tables->lock_type;
}
......@@ -5389,6 +5403,8 @@ int lock_tables(THD *thd, TABLE_LIST *tables, uint count, bool *need_reopen)
DBUG_RETURN(-1);
}
DEBUG_SYNC(thd, "after_lock_tables_takes_lock");
if (thd->lex->requires_prelocking() &&
thd->lex->sql_command != SQLCOM_LOCK_TABLES)
{
......
......@@ -1053,7 +1053,7 @@ int mysql_multi_update_prepare(THD *thd)
correct order of statements. Otherwise, we use a TL_READ lock to
improve performance.
*/
tl->lock_type= read_lock_type_for_table(thd, table);
tl->lock_type= read_lock_type_for_table(thd, lex, tl);
tl->updating= 0;
/* Update TABLE::lock_type accordingly. */
if (!tl->placeholder() && !using_lock_tables)
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment