mariadb/mysql-test/t/lock_sync.test

1029 lines
33 KiB
Text
Raw Normal View History

Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
#
# Locking related tests which use DEBUG_SYNC facility.
#
--source include/have_debug_sync.inc
# We need InnoDB to be able use TL_WRITE_ALLOW_WRITE type of locks in our tests.
--source include/have_innodb.inc
Committing on behalf or Dmitry Lenev: Fix for bug #46947 "Embedded SELECT without FOR UPDATE is causing a lock", with after-review fixes. 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 tries to solve this problem by rethinking our approach to how we handle locking for SELECT and subqueries. Now we always set TL_READ_DEFAULT lock type for all cases when we read data. When at open_tables() time this lock is interpreted as TL_READ_NO_INSERT or TL_READ depending on whether this statement as a whole or call to function which uses particular table should be written to the binary log or not (if yes then statement should be properly serialized with concurrent statements and stronger lock should be acquired). 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 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. On of the patches for 5.0 broke this behaviour (which was not documented or tested), and started to use locking reads fora all subqueries in SELECT ... FOR UPDATE/IN SHARE MODE. This patch restored 4.1 behaviour. mysql-test/include/check_concurrent_insert.inc: Added auxiliary script which allows to check if statement reading table allows concurrent inserts in it. mysql-test/include/check_no_concurrent_insert.inc: Added auxiliary script which allows to check that statement reading table doesn't allow concurrent inserts in it. mysql-test/include/check_no_row_lock.inc: Added auxiliary script which allows to check if statement reading table doesn't take locks on its rows. mysql-test/include/check_shared_row_lock.inc: Added auxiliary script which allows to check if statement reading table takes shared locks on some of its rows. mysql-test/r/bug39022.result: After bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock' was fixed test case for bug 39022 has to be adjusted in order to trigger execution path on which original problem was encountered. mysql-test/r/innodb_mysql_lock2.result: Added coverage for handling of locking in various cases when we read data from InnoDB tables (includes test case for bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock'). mysql-test/r/lock_sync.result: Added coverage for handling of locking in various cases when we read data from MyISAM tables. mysql-test/t/bug39022.test: After bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock' was fixed test case for bug 39022 has to be adjusted in order to trigger execution path on which original problem was encountered. mysql-test/t/innodb_mysql_lock2.test: Added coverage for handling of locking in various cases when we read data from InnoDB tables (includes test case for bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock'). mysql-test/t/lock_sync.test: Added coverage for handling of locking in various cases when we read data from MyISAM tables. sql/log_event.cc: Since LEX::lock_option member was removed we no longer can rely on its value in Load_log_event::print_query() to determine that log event correponds to LOAD DATA CONCURRENT statement (this was not correct in all situations anyway). A new Load_log_event's member was introduced as a replacement. It is initialized at event object construction time and explicitly indicates whether LOAD DATA was concurrent. sql/log_event.h: Since LEX::lock_option member was removed we no longer can rely on its value in Load_log_event::print_query() to determine that log event correponds to LOAD DATA CONCURRENT statement (this was not correct in all situations anyway). A new Load_log_event's member was introduced as a replacement. It is initialized at event object construction time and explicitly indicates whether LOAD DATA was concurrent. sql/sp_head.cc: sp_head::reset_lex(): Before parsing substatement reset part of parser state which needs this (e.g. set Yacc_state::m_lock_type to default value). sql/sql_acl.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed (for example, when we are logging statement to the binary log while having Query_tables_list reset and backed up). sql/sql_base.cc: Changed read_lock_type_for_table() to return a weak TL_READ type of lock in cases when we are executing statement which won't update tables directly and table doesn't belong to statement's prelocking list and thus can't be used by a stored function. It is OK to do so since in this case table won't be used by statement or function call which will be written to the binary log, so serializability requirements for it can be relaxed. One of results from this change is that SELECTs on InnoDB tables no longer takes shared row locks for tables which are used in subqueries (i.e. bug #46947 is fixed). Another result is that for similar SELECTs on MyISAM tables concurrent inserts are allowed. In order to implement this change signature of read_lock_type_for_table() function was changed to take pointers to Query_tables_list and TABLE_LIST objects. sql/sql_base.h: - Function read_lock_type_for_table() now takes pointers to Query_tables_list and TABLE_LIST elements as its arguments since to correctly determine lock type it needs to know what statement is being performed and whether table element for which lock type to be determined belongs to prelocking list. sql/sql_lex.cc: - Removed LEX::lock_option and st_select_lex::lock_option members. Places in parser that were using them now use Yacc_state::m_lock_type instead. - To emphasize that LEX::sql_command member is used during process of opening and locking of tables it was moved to Query_tables_list class. It is now reset by Query_tables_list::reset_query_tables_list() method. sql/sql_lex.h: - Removed st_select_lex::lock_option member as there is no real need for per-SELECT lock type (HIGH_PRIORITY option should apply to the whole statement. FOR UPDATE/LOCK IN SHARE MODE clauses can be handled without this member). The main effect which was achieved by introduction of this member, i.e. using TL_READ_DEFAULT lock type for subqueries, is now achieved by setting LEX::lock_option (or rather its replacement - Yacc_state::m_lock_type) to TL_READ_DEFAULT in almost all cases. - To emphasize that LEX::sql_command member is used during process of opening and locking of tables it was moved to Query_tables_list class. - Replaced LEX::lock_option with Yacc_state::m_lock_type in order to emphasize that this value is relevant only during parsing. Unlike for LEX::lock_option the default value for Yacc_state::m_lock_type is TL_READ_DEFAULT. Note that for cases when it is OK to take a "weak" read lock (e.g. simple SELECT) this lock type will be converted to TL_READ at open_tables() time. So this change won't cause negative change in behavior for such statements. OTOH this change ensures that, for example, for SELECTs which are used in stored functions TL_READ_NO_INSERT lock is taken when necessary and as result calls to such stored functions can be written to the binary log with correct serialization. sql/sql_load.cc: Load_log_event constructor now requires a parameter that indicates whether LOAD DATA is concurrent. sql/sql_parse.cc: LEX::lock_option was replaced with Yacc_state::m_lock_type. And instead of resetting the latter implicitly in mysql_init_multi_delete() we do it explicitly in the places in parser which call this function. sql/sql_priv.h: - To be able more easily distinguish high-priority SELECTs in st_select_lex::print() method added flag for HIGH_PRIORITY option. sql/sql_select.cc: Changed code not to rely on LEX::lock_option to determine that it is high-priority SELECT. It was replaced with Yacc_state::m_lock_type which is accessible only at parse time. So instead of LEX::lock_option we now rely on a newly introduced flag for st_select_lex::options - SELECT_HIGH_PRIORITY. sql/sql_show.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed. sql/sql_table.cc: Since LEX::reset_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore value of this member when this method is called by mysql_admin_table(), to make this code safe for re-execution. sql/sql_trigger.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed (for example, when we are logging statement to the binary log while having Query_tables_list reset and backed up). sql/sql_update.cc: Function read_lock_type_for_table() now takes pointers to Query_tables_list and TABLE_LIST elements as its arguments since to correctly determine lock type it needs to know what statement is being performed and whether table element for which lock type to be determined belongs to prelocking list. sql/sql_yacc.yy: - Removed st_select_lex::lock_option member as there is no real need for per-SELECT lock type (HIGH_PRIORITY option should apply to the whole statement. FOR UPDATE/LOCK IN SHARE MODE clauses can be handled without this member). The main effect which was achieved by introduction of this member, i.e. using TL_READ_DEFAULT lock type for subqueries, is now achieved by setting LEX::lock_option (or rather its replacement - Yacc_state::m_lock_type) to TL_READ_DEFAULT in almost all cases. - Replaced LEX::lock_option with Yacc_state::m_lock_type in order to emphasize that this value is relevant only during parsing. Unlike for LEX::lock_option the default value for Yacc_state::m_lock_type is TL_READ_DEFAULT. Note that for cases when it is OK to take a "weak" read lock (e.g. simple SELECT) this lock type will be converted to TL_READ at open_tables() time. So this change won't cause negative change in behavior for such statements. OTOH this change ensures that, for example, for SELECTs which are used in stored functions TL_READ_NO_INSERT lock is taken when necessary and as result calls to such stored functions can be written to the binary log with correct serialization. - To be able more easily distinguish high-priority SELECTs in st_select_lex::print() method we now use new flag in st_select_lex::options bit-field.
2010-04-28 12:04:11 +02:00
# This test requires statement/mixed mode binary logging.
# Row-based mode puts weaker serializability requirements
# so weaker locks are acquired for it.
--source include/have_binlog_format_mixed_or_statement.inc
# Until bug#41971 'Thread state on embedded server is always "Writing to net"'
# is fixed this test can't be run on embedded version of server.
--source include/not_embedded.inc
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
# Save the initial number of concurrent sessions.
--source include/count_sessions.inc
Committing on behalf or Dmitry Lenev: Fix for bug #46947 "Embedded SELECT without FOR UPDATE is causing a lock", with after-review fixes. 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 tries to solve this problem by rethinking our approach to how we handle locking for SELECT and subqueries. Now we always set TL_READ_DEFAULT lock type for all cases when we read data. When at open_tables() time this lock is interpreted as TL_READ_NO_INSERT or TL_READ depending on whether this statement as a whole or call to function which uses particular table should be written to the binary log or not (if yes then statement should be properly serialized with concurrent statements and stronger lock should be acquired). 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 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. On of the patches for 5.0 broke this behaviour (which was not documented or tested), and started to use locking reads fora all subqueries in SELECT ... FOR UPDATE/IN SHARE MODE. This patch restored 4.1 behaviour. mysql-test/include/check_concurrent_insert.inc: Added auxiliary script which allows to check if statement reading table allows concurrent inserts in it. mysql-test/include/check_no_concurrent_insert.inc: Added auxiliary script which allows to check that statement reading table doesn't allow concurrent inserts in it. mysql-test/include/check_no_row_lock.inc: Added auxiliary script which allows to check if statement reading table doesn't take locks on its rows. mysql-test/include/check_shared_row_lock.inc: Added auxiliary script which allows to check if statement reading table takes shared locks on some of its rows. mysql-test/r/bug39022.result: After bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock' was fixed test case for bug 39022 has to be adjusted in order to trigger execution path on which original problem was encountered. mysql-test/r/innodb_mysql_lock2.result: Added coverage for handling of locking in various cases when we read data from InnoDB tables (includes test case for bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock'). mysql-test/r/lock_sync.result: Added coverage for handling of locking in various cases when we read data from MyISAM tables. mysql-test/t/bug39022.test: After bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock' was fixed test case for bug 39022 has to be adjusted in order to trigger execution path on which original problem was encountered. mysql-test/t/innodb_mysql_lock2.test: Added coverage for handling of locking in various cases when we read data from InnoDB tables (includes test case for bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock'). mysql-test/t/lock_sync.test: Added coverage for handling of locking in various cases when we read data from MyISAM tables. sql/log_event.cc: Since LEX::lock_option member was removed we no longer can rely on its value in Load_log_event::print_query() to determine that log event correponds to LOAD DATA CONCURRENT statement (this was not correct in all situations anyway). A new Load_log_event's member was introduced as a replacement. It is initialized at event object construction time and explicitly indicates whether LOAD DATA was concurrent. sql/log_event.h: Since LEX::lock_option member was removed we no longer can rely on its value in Load_log_event::print_query() to determine that log event correponds to LOAD DATA CONCURRENT statement (this was not correct in all situations anyway). A new Load_log_event's member was introduced as a replacement. It is initialized at event object construction time and explicitly indicates whether LOAD DATA was concurrent. sql/sp_head.cc: sp_head::reset_lex(): Before parsing substatement reset part of parser state which needs this (e.g. set Yacc_state::m_lock_type to default value). sql/sql_acl.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed (for example, when we are logging statement to the binary log while having Query_tables_list reset and backed up). sql/sql_base.cc: Changed read_lock_type_for_table() to return a weak TL_READ type of lock in cases when we are executing statement which won't update tables directly and table doesn't belong to statement's prelocking list and thus can't be used by a stored function. It is OK to do so since in this case table won't be used by statement or function call which will be written to the binary log, so serializability requirements for it can be relaxed. One of results from this change is that SELECTs on InnoDB tables no longer takes shared row locks for tables which are used in subqueries (i.e. bug #46947 is fixed). Another result is that for similar SELECTs on MyISAM tables concurrent inserts are allowed. In order to implement this change signature of read_lock_type_for_table() function was changed to take pointers to Query_tables_list and TABLE_LIST objects. sql/sql_base.h: - Function read_lock_type_for_table() now takes pointers to Query_tables_list and TABLE_LIST elements as its arguments since to correctly determine lock type it needs to know what statement is being performed and whether table element for which lock type to be determined belongs to prelocking list. sql/sql_lex.cc: - Removed LEX::lock_option and st_select_lex::lock_option members. Places in parser that were using them now use Yacc_state::m_lock_type instead. - To emphasize that LEX::sql_command member is used during process of opening and locking of tables it was moved to Query_tables_list class. It is now reset by Query_tables_list::reset_query_tables_list() method. sql/sql_lex.h: - Removed st_select_lex::lock_option member as there is no real need for per-SELECT lock type (HIGH_PRIORITY option should apply to the whole statement. FOR UPDATE/LOCK IN SHARE MODE clauses can be handled without this member). The main effect which was achieved by introduction of this member, i.e. using TL_READ_DEFAULT lock type for subqueries, is now achieved by setting LEX::lock_option (or rather its replacement - Yacc_state::m_lock_type) to TL_READ_DEFAULT in almost all cases. - To emphasize that LEX::sql_command member is used during process of opening and locking of tables it was moved to Query_tables_list class. - Replaced LEX::lock_option with Yacc_state::m_lock_type in order to emphasize that this value is relevant only during parsing. Unlike for LEX::lock_option the default value for Yacc_state::m_lock_type is TL_READ_DEFAULT. Note that for cases when it is OK to take a "weak" read lock (e.g. simple SELECT) this lock type will be converted to TL_READ at open_tables() time. So this change won't cause negative change in behavior for such statements. OTOH this change ensures that, for example, for SELECTs which are used in stored functions TL_READ_NO_INSERT lock is taken when necessary and as result calls to such stored functions can be written to the binary log with correct serialization. sql/sql_load.cc: Load_log_event constructor now requires a parameter that indicates whether LOAD DATA is concurrent. sql/sql_parse.cc: LEX::lock_option was replaced with Yacc_state::m_lock_type. And instead of resetting the latter implicitly in mysql_init_multi_delete() we do it explicitly in the places in parser which call this function. sql/sql_priv.h: - To be able more easily distinguish high-priority SELECTs in st_select_lex::print() method added flag for HIGH_PRIORITY option. sql/sql_select.cc: Changed code not to rely on LEX::lock_option to determine that it is high-priority SELECT. It was replaced with Yacc_state::m_lock_type which is accessible only at parse time. So instead of LEX::lock_option we now rely on a newly introduced flag for st_select_lex::options - SELECT_HIGH_PRIORITY. sql/sql_show.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed. sql/sql_table.cc: Since LEX::reset_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore value of this member when this method is called by mysql_admin_table(), to make this code safe for re-execution. sql/sql_trigger.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed (for example, when we are logging statement to the binary log while having Query_tables_list reset and backed up). sql/sql_update.cc: Function read_lock_type_for_table() now takes pointers to Query_tables_list and TABLE_LIST elements as its arguments since to correctly determine lock type it needs to know what statement is being performed and whether table element for which lock type to be determined belongs to prelocking list. sql/sql_yacc.yy: - Removed st_select_lex::lock_option member as there is no real need for per-SELECT lock type (HIGH_PRIORITY option should apply to the whole statement. FOR UPDATE/LOCK IN SHARE MODE clauses can be handled without this member). The main effect which was achieved by introduction of this member, i.e. using TL_READ_DEFAULT lock type for subqueries, is now achieved by setting LEX::lock_option (or rather its replacement - Yacc_state::m_lock_type) to TL_READ_DEFAULT in almost all cases. - Replaced LEX::lock_option with Yacc_state::m_lock_type in order to emphasize that this value is relevant only during parsing. Unlike for LEX::lock_option the default value for Yacc_state::m_lock_type is TL_READ_DEFAULT. Note that for cases when it is OK to take a "weak" read lock (e.g. simple SELECT) this lock type will be converted to TL_READ at open_tables() time. So this change won't cause negative change in behavior for such statements. OTOH this change ensures that, for example, for SELECTs which are used in stored functions TL_READ_NO_INSERT lock is taken when necessary and as result calls to such stored functions can be written to the binary log with correct serialization. - To be able more easily distinguish high-priority SELECTs in st_select_lex::print() method we now use new flag in st_select_lex::options bit-field.
2010-04-28 12:04:11 +02:00
--echo #
--echo # Test how we handle locking in various cases when
--echo # we read data from MyISAM tables.
--echo #
--echo # In this test we mostly check that the SQL-layer correctly
--echo # determines the type of thr_lock.c lock for a table being
--echo # read.
--echo # I.e. that it disallows concurrent inserts when the statement
--echo # is going to be written to the binary log and therefore
--echo # should be serialized, and allows concurrent inserts when
--echo # such serialization is not necessary (e.g. when
--echo # the statement is not written to binary log).
--echo #
--echo # Force concurrent inserts to be performed even if the table
--echo # has gaps. This allows to simplify clean up in scripts
--echo # used below (instead of backing up table being inserted
--echo # into and then restoring it from backup at the end of the
--echo # script we can simply delete rows which were inserted).
set @old_concurrent_insert= @@global.concurrent_insert;
set @@global.concurrent_insert= 2;
select @@global.concurrent_insert;
--echo # Prepare playground by creating tables, views,
--echo # routines and triggers used in tests.
connect (con1, localhost, root,,);
connect (con2, localhost, root,,);
connection default;
--disable_warnings
drop table if exists t0, t1, t2, t3, t4, t5;
drop view if exists v1, v2;
drop procedure if exists p1;
drop procedure if exists p2;
drop function if exists f1;
drop function if exists f2;
drop function if exists f3;
drop function if exists f4;
drop function if exists f5;
drop function if exists f6;
drop function if exists f7;
drop function if exists f8;
drop function if exists f9;
drop function if exists f10;
drop function if exists f11;
drop function if exists f12;
drop function if exists f13;
drop function if exists f14;
drop function if exists f15;
--enable_warnings
create table t1 (i int primary key);
insert into t1 values (1), (2), (3), (4), (5);
create table t2 (j int primary key);
insert into t2 values (1), (2), (3), (4), (5);
create table t3 (k int primary key);
insert into t3 values (1), (2), (3);
create table t4 (l int primary key);
insert into t4 values (1);
create table t5 (l int primary key);
insert into t5 values (1);
create view v1 as select i from t1;
create view v2 as select j from t2 where j in (select i from t1);
create procedure p1(k int) insert into t2 values (k);
delimiter |;
create function f1() returns int
begin
declare j int;
select i from t1 where i = 1 into j;
return j;
end|
create function f2() returns int
begin
declare k int;
select i from t1 where i = 1 into k;
insert into t2 values (k + 5);
return 0;
end|
create function f3() returns int
begin
return (select i from t1 where i = 3);
end|
create function f4() returns int
begin
if (select i from t1 where i = 3) then
return 1;
else
return 0;
end if;
end|
create function f5() returns int
begin
insert into t2 values ((select i from t1 where i = 1) + 5);
return 0;
end|
create function f6() returns int
begin
declare k int;
select i from v1 where i = 1 into k;
return k;
end|
create function f7() returns int
begin
declare k int;
select j from v2 where j = 1 into k;
return k;
end|
create function f8() returns int
begin
declare k int;
select i from v1 where i = 1 into k;
insert into t2 values (k+5);
return k;
end|
create function f9() returns int
begin
update v2 set j=j+10 where j=1;
return 1;
end|
create function f10() returns int
begin
return f1();
end|
create function f11() returns int
begin
declare k int;
set k= f1();
insert into t2 values (k+5);
return k;
end|
create function f12(p int) returns int
begin
insert into t2 values (p);
return p;
end|
create function f13(p int) returns int
begin
return p;
end|
create procedure p2(inout p int)
begin
select i from t1 where i = 1 into p;
end|
create function f14() returns int
begin
declare k int;
call p2(k);
insert into t2 values (k+5);
return k;
end|
create function f15() returns int
begin
declare k int;
call p2(k);
return k;
end|
create trigger t4_bi before insert on t4 for each row
begin
declare k int;
select i from t1 where i=1 into k;
set new.l= k+1;
end|
create trigger t4_bu before update on t4 for each row
begin
if (select i from t1 where i=1) then
set new.l= 2;
end if;
end|
create trigger t4_bd before delete on t4 for each row
begin
if !(select i from v1 where i=1) then
signal sqlstate '45000';
end if;
end|
create trigger t5_bi before insert on t5 for each row
begin
set new.l= f1()+1;
end|
create trigger t5_bu before update on t5 for each row
begin
declare j int;
call p2(j);
set new.l= j + 1;
end|
delimiter ;|
--echo #
--echo # Set common variables to be used by the scripts
--echo # called below.
--echo #
let $con_aux1= con1;
let $con_aux2= con2;
let $table= t1;
--echo # Switch to connection 'con1'.
connection con1;
--echo # Cache all functions used in the tests below so statements
--echo # calling them won't need to open and lock mysql.proc table
--echo # and we can assume that each statement locks its tables
--echo # once during its execution.
--disable_result_log
show create procedure p1;
show create procedure p2;
show create function f1;
show create function f2;
show create function f3;
show create function f4;
show create function f5;
show create function f6;
show create function f7;
show create function f8;
show create function f9;
show create function f10;
show create function f11;
show create function f12;
show create function f13;
show create function f14;
show create function f15;
--enable_result_log
--echo # Switch back to connection 'default'.
connection default;
--echo #
--echo # 1. Statements that read tables and do not use subqueries.
--echo #
--echo #
--echo # 1.1 Simple SELECT statement.
--echo #
--echo # No locks are necessary as this statement won't be written
--echo # to the binary log and thanks to how MyISAM works SELECT
--echo # will see version of the table prior to concurrent insert.
let $statement= select * from t1;
let $restore_table= ;
--source include/check_concurrent_insert.inc
--echo #
--echo # 1.2 Multi-UPDATE statement.
--echo #
--echo # Has to take shared locks on rows in the table being read as this
--echo # statement will be written to the binary log and therefore should
--echo # be serialized with concurrent statements.
let $statement= update t2, t1 set j= j - 1 where i = j;
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 1.3 Multi-DELETE statement.
--echo #
--echo # The above is true for this statement as well.
let $statement= delete t2 from t1, t2 where i = j;
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 1.4 DESCRIBE statement.
--echo #
--echo # This statement does not really read data from the
--echo # target table and thus does not take any lock on it.
--echo # We check this for completeness of coverage.
lock table t1 write;
--echo # Switching to connection 'con1'.
connection con1;
--echo # This statement should not be blocked.
--disable_result_log
describe t1;
--enable_result_log
--echo # Switching to connection 'default'.
connection default;
unlock tables;
--echo #
--echo # 1.5 SHOW statements.
--echo #
--echo # The above is true for SHOW statements as well.
lock table t1 write;
--echo # Switching to connection 'con1'.
connection con1;
--echo # These statements should not be blocked.
# The below test for SHOW CREATE TABLE is disabled until bug 52593
# "SHOW CREATE TABLE is blocked if table is locked for write by another
# connection" is fixed.
--disable_parsing
show create table t1;
--enable_parsing
--disable_result_log
show keys from t1;
--enable_result_log
--echo # Switching to connection 'default'.
connection default;
unlock tables;
--echo #
--echo # 2. Statements which read tables through subqueries.
--echo #
--echo #
--echo # 2.1 CALL with a subquery.
--echo #
--echo # A strong lock is not necessary as this statement is not
--echo # written to the binary log as a whole (it is written
--echo # statement-by-statement).
let $statement= call p1((select i + 5 from t1 where i = 1));
let $restore_table= t2;
--source include/check_concurrent_insert.inc
--echo #
--echo # 2.2 CREATE TABLE with a subquery.
--echo #
--echo # Has to take a strong lock on the table being read as
--echo # this statement is written to the binary log and therefore
--echo # should be serialized with concurrent statements.
let $statement= create table t0 select * from t1;
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
drop table t0;
let $statement= create table t0 select j from t2 where j in (select i from t1);
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
drop table t0;
--echo #
--echo # 2.3 DELETE with a subquery.
--echo #
--echo # The above is true for this statement as well.
let $statement= delete from t2 where j in (select i from t1);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 2.4 MULTI-DELETE with a subquery.
--echo #
--echo # Same is true for this statement as well.
let $statement= delete t2 from t3, t2 where k = j and j in (select i from t1);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 2.5 DO with a subquery.
--echo #
--echo # A strong lock is not necessary as it is not logged.
let $statement= do (select i from t1 where i = 1);
let $restore_table= ;
--source include/check_concurrent_insert.inc
--echo #
--echo # 2.6 INSERT with a subquery.
--echo #
--echo # Has to take a strong lock on the table being read as
--echo # this statement is written to the binary log and therefore
--echo # should be serialized with concurrent inserts.
let $statement= insert into t2 select i+5 from t1;
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= insert into t2 values ((select i+5 from t1 where i = 4));
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 2.7 LOAD DATA with a subquery.
--echo #
--echo # The above is true for this statement as well.
let $statement= load data infile '../../std_data/rpl_loaddata.dat' into table t2 (@a, @b) set j= @b + (select i from t1 where i = 1);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 2.8 REPLACE with a subquery.
--echo #
--echo # Same is true for this statement as well.
let $statement= replace into t2 select i+5 from t1;
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= replace into t2 values ((select i+5 from t1 where i = 4));
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 2.9 SELECT with a subquery.
--echo #
--echo # Strong locks are not necessary as this statement is not written
--echo # to the binary log and thanks to how MyISAM works this statement
--echo # sees a version of the table prior to the concurrent insert.
let $statement= select * from t2 where j in (select i from t1);
let $restore_table= ;
--source include/check_concurrent_insert.inc
--echo #
--echo # 2.10 SET with a subquery.
--echo #
--echo # The same is true for this statement as well.
let $statement= set @a:= (select i from t1 where i = 1);
let $restore_table= ;
--source include/check_concurrent_insert.inc
--echo #
--echo # 2.11 SHOW with a subquery.
--echo #
--echo # And for this statement too.
let $statement= show tables from test where Tables_in_test = 't2' and (select i from t1 where i = 1);
let $restore_table= ;
--source include/check_concurrent_insert.inc
let $statement= show columns from t2 where (select i from t1 where i = 1);
let $restore_table= ;
--source include/check_concurrent_insert.inc
--echo #
--echo # 2.12 UPDATE with a subquery.
--echo #
--echo # Has to take a strong lock on the table being read as
--echo # this statement is written to the binary log and therefore
--echo # should be serialized with concurrent inserts.
let $statement= update t2 set j= j-10 where j in (select i from t1);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 2.13 MULTI-UPDATE with a subquery.
--echo #
--echo # Same is true for this statement as well.
let $statement= update t2, t3 set j= j -10 where j=k and j in (select i from t1);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 3. Statements which read tables through a view.
--echo #
--echo #
--echo # 3.1 SELECT statement which uses some table through a view.
--echo #
--echo # Since this statement is not written to the binary log and
--echo # an old version of the table is accessible thanks to how MyISAM
--echo # handles concurrent insert, no locking is necessary.
let $statement= select * from v1;
let $restore_table= ;
--source include/check_concurrent_insert.inc
let $statement= select * from v2;
let $restore_table= ;
--source include/check_concurrent_insert.inc
let $statement= select * from t2 where j in (select i from v1);
let $restore_table= ;
--source include/check_concurrent_insert.inc
let $statement= select * from t3 where k in (select j from v2);
let $restore_table= ;
--source include/check_concurrent_insert.inc
--echo #
--echo # 3.2 Statements which modify a table and use views.
--echo #
--echo # Since such statements are going to be written to the binary
--echo # log they need to be serialized against concurrent statements
--echo # and therefore should take strong locks on the data read.
let $statement= update t2 set j= j-10 where j in (select i from v1);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= update t3 set k= k-10 where k in (select j from v2);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= update t2, v1 set j= j-10 where j = i;
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= update v2 set j= j-10 where j = 3;
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4. Statements which read tables through stored functions.
--echo #
--echo #
--echo # 4.1 SELECT/SET with a stored function which does not
--echo # modify data and uses SELECT in its turn.
--echo #
--echo # In theory there is no need to take strong locks on the table
--echo # being selected from in SF as the call to such function
--echo # won't get into the binary log. In practice, however, we
--echo # discover that fact too late in the process to be able to
--echo # affect the decision what locks should be taken.
--echo # Hence, strong locks are taken in this case.
let $statement= select f1();
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
let $statement= set @a:= f1();
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.2 INSERT (or other statement which modifies data) with
--echo # a stored function which does not modify data and uses
--echo # SELECT.
--echo #
--echo # Since such statement is written to the binary log it should
--echo # be serialized with concurrent statements affecting the data
--echo # it uses. Therefore it should take strong lock on the data
--echo # it reads.
let $statement= insert into t2 values (f1() + 5);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.3 SELECT/SET with a stored function which
--echo # reads and modifies data.
--echo #
--echo # Since a call to such function is written to the binary log,
--echo # it should be serialized with concurrent statements affecting
--echo # the data it uses. Hence, a strong lock on the data read
--echo # should be taken.
let $statement= select f2();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= set @a:= f2();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.4. SELECT/SET with a stored function which does not
--echo # modify data and reads a table through subselect
--echo # in a control construct.
--echo #
--echo # Again, in theory a call to this function won't get to the
--echo # binary log and thus no strong lock is needed. But in practice
--echo # we don't detect this fact early enough (get_lock_type_for_table())
--echo # to avoid taking a strong lock.
let $statement= select f3();
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
let $statement= set @a:= f3();
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
let $statement= select f4();
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
let $statement= set @a:= f4();
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.5. INSERT (or other statement which modifies data) with
--echo # a stored function which does not modify data and reads
--echo # the table through a subselect in one of its control
--echo # constructs.
--echo #
--echo # Since such statement is written to the binary log it should
--echo # be serialized with concurrent statements affecting data it
--echo # uses. Therefore it should take a strong lock on the data
--echo # it reads.
let $statement= insert into t2 values (f3() + 5);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= insert into t2 values (f4() + 6);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.6 SELECT/SET which uses a stored function with
--echo # DML which reads a table via a subquery.
--echo #
--echo # Since call to such function is written to the binary log
--echo # it should be serialized with concurrent statements.
--echo # Hence reads should take a strong lock.
let $statement= select f5();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= set @a:= f5();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.7 SELECT/SET which uses a stored function which
--echo # doesn't modify data and reads tables through
--echo # a view.
--echo #
--echo # Once again, in theory, calls to such functions won't
--echo # get into the binary log and thus don't need strong
--echo # locks. But in practice this fact is discovered
--echo # too late to have any effect.
let $statement= select f6();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= set @a:= f6();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= select f7();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= set @a:= f7();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.8 INSERT which uses stored function which
--echo # doesn't modify data and reads a table
--echo # through a view.
--echo #
--echo # Since such statement is written to the binary log and
--echo # should be serialized with concurrent statements affecting
--echo # the data it uses. Therefore it should take a strong lock on
--echo # the table it reads.
let $statement= insert into t3 values (f6() + 5);
let $restore_table= t3;
--source include/check_no_concurrent_insert.inc
let $statement= insert into t3 values (f7() + 5);
let $restore_table= t3;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.9 SELECT which uses a stored function which
--echo # modifies data and reads tables through a view.
--echo #
--echo # Since a call to such function is written to the binary log
--echo # it should be serialized with concurrent statements.
--echo # Hence, reads should take strong locks.
let $statement= select f8();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
let $statement= select f9();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.10 SELECT which uses a stored function which doesn't modify
--echo # data and reads a table indirectly, by calling another
--echo # function.
--echo #
--echo # In theory, calls to such functions won't get into the binary
--echo # log and thus don't need to acquire strong locks. But in practice
--echo # this fact is discovered too late to have any effect.
let $statement= select f10();
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.11 INSERT which uses a stored function which doesn't modify
--echo # data and reads a table indirectly, by calling another
--echo # function.
--echo #
--echo # Since such statement is written to the binary log, it should
--echo # be serialized with concurrent statements affecting the data it
--echo # uses. Therefore it should take strong locks on data it reads.
let $statement= insert into t2 values (f10() + 5);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.12 SELECT which uses a stored function which modifies
--echo # data and reads a table indirectly, by calling another
--echo # function.
--echo #
--echo # Since a call to such function is written to the binary log
--echo # it should be serialized from concurrent statements.
--echo # Hence, read should take a strong lock.
let $statement= select f11();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 4.13 SELECT that reads a table through a subquery passed
--echo # as a parameter to a stored function which modifies
--echo # data.
--echo #
--echo # Even though a call to this function is written to the
--echo # binary log, values of its parameters are written as literals.
--echo # So there is no need to acquire strong locks for tables used in
--echo # the subquery.
let $statement= select f12((select i+10 from t1 where i=1));
let $restore_table= t2;
--source include/check_concurrent_insert.inc
--echo #
--echo # 4.14 INSERT that reads a table via a subquery passed
--echo # as a parameter to a stored function which doesn't
--echo # modify data.
--echo #
--echo # Since this statement is written to the binary log it should
--echo # be serialized with concurrent statements affecting the data it
--echo # uses. Therefore it should take strong locks on the data it reads.
let $statement= insert into t2 values (f13((select i+10 from t1 where i=1)));
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 5. Statements that read tables through stored procedures.
--echo #
--echo #
--echo # 5.1 CALL statement which reads a table via SELECT.
--echo #
--echo # Since neither this statement nor its components are
--echo # written to the binary log, there is no need to take
--echo # strong locks on the data it reads.
let $statement= call p2(@a);
let $restore_table= ;
--source include/check_concurrent_insert.inc
--echo #
--echo # 5.2 Function that modifies data and uses CALL,
Committing on behalf or Dmitry Lenev: Fix for bug #46947 "Embedded SELECT without FOR UPDATE is causing a lock", with after-review fixes. 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 tries to solve this problem by rethinking our approach to how we handle locking for SELECT and subqueries. Now we always set TL_READ_DEFAULT lock type for all cases when we read data. When at open_tables() time this lock is interpreted as TL_READ_NO_INSERT or TL_READ depending on whether this statement as a whole or call to function which uses particular table should be written to the binary log or not (if yes then statement should be properly serialized with concurrent statements and stronger lock should be acquired). 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 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. On of the patches for 5.0 broke this behaviour (which was not documented or tested), and started to use locking reads fora all subqueries in SELECT ... FOR UPDATE/IN SHARE MODE. This patch restored 4.1 behaviour. mysql-test/include/check_concurrent_insert.inc: Added auxiliary script which allows to check if statement reading table allows concurrent inserts in it. mysql-test/include/check_no_concurrent_insert.inc: Added auxiliary script which allows to check that statement reading table doesn't allow concurrent inserts in it. mysql-test/include/check_no_row_lock.inc: Added auxiliary script which allows to check if statement reading table doesn't take locks on its rows. mysql-test/include/check_shared_row_lock.inc: Added auxiliary script which allows to check if statement reading table takes shared locks on some of its rows. mysql-test/r/bug39022.result: After bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock' was fixed test case for bug 39022 has to be adjusted in order to trigger execution path on which original problem was encountered. mysql-test/r/innodb_mysql_lock2.result: Added coverage for handling of locking in various cases when we read data from InnoDB tables (includes test case for bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock'). mysql-test/r/lock_sync.result: Added coverage for handling of locking in various cases when we read data from MyISAM tables. mysql-test/t/bug39022.test: After bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock' was fixed test case for bug 39022 has to be adjusted in order to trigger execution path on which original problem was encountered. mysql-test/t/innodb_mysql_lock2.test: Added coverage for handling of locking in various cases when we read data from InnoDB tables (includes test case for bug #46947 'Embedded SELECT without FOR UPDATE is causing a lock'). mysql-test/t/lock_sync.test: Added coverage for handling of locking in various cases when we read data from MyISAM tables. sql/log_event.cc: Since LEX::lock_option member was removed we no longer can rely on its value in Load_log_event::print_query() to determine that log event correponds to LOAD DATA CONCURRENT statement (this was not correct in all situations anyway). A new Load_log_event's member was introduced as a replacement. It is initialized at event object construction time and explicitly indicates whether LOAD DATA was concurrent. sql/log_event.h: Since LEX::lock_option member was removed we no longer can rely on its value in Load_log_event::print_query() to determine that log event correponds to LOAD DATA CONCURRENT statement (this was not correct in all situations anyway). A new Load_log_event's member was introduced as a replacement. It is initialized at event object construction time and explicitly indicates whether LOAD DATA was concurrent. sql/sp_head.cc: sp_head::reset_lex(): Before parsing substatement reset part of parser state which needs this (e.g. set Yacc_state::m_lock_type to default value). sql/sql_acl.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed (for example, when we are logging statement to the binary log while having Query_tables_list reset and backed up). sql/sql_base.cc: Changed read_lock_type_for_table() to return a weak TL_READ type of lock in cases when we are executing statement which won't update tables directly and table doesn't belong to statement's prelocking list and thus can't be used by a stored function. It is OK to do so since in this case table won't be used by statement or function call which will be written to the binary log, so serializability requirements for it can be relaxed. One of results from this change is that SELECTs on InnoDB tables no longer takes shared row locks for tables which are used in subqueries (i.e. bug #46947 is fixed). Another result is that for similar SELECTs on MyISAM tables concurrent inserts are allowed. In order to implement this change signature of read_lock_type_for_table() function was changed to take pointers to Query_tables_list and TABLE_LIST objects. sql/sql_base.h: - Function read_lock_type_for_table() now takes pointers to Query_tables_list and TABLE_LIST elements as its arguments since to correctly determine lock type it needs to know what statement is being performed and whether table element for which lock type to be determined belongs to prelocking list. sql/sql_lex.cc: - Removed LEX::lock_option and st_select_lex::lock_option members. Places in parser that were using them now use Yacc_state::m_lock_type instead. - To emphasize that LEX::sql_command member is used during process of opening and locking of tables it was moved to Query_tables_list class. It is now reset by Query_tables_list::reset_query_tables_list() method. sql/sql_lex.h: - Removed st_select_lex::lock_option member as there is no real need for per-SELECT lock type (HIGH_PRIORITY option should apply to the whole statement. FOR UPDATE/LOCK IN SHARE MODE clauses can be handled without this member). The main effect which was achieved by introduction of this member, i.e. using TL_READ_DEFAULT lock type for subqueries, is now achieved by setting LEX::lock_option (or rather its replacement - Yacc_state::m_lock_type) to TL_READ_DEFAULT in almost all cases. - To emphasize that LEX::sql_command member is used during process of opening and locking of tables it was moved to Query_tables_list class. - Replaced LEX::lock_option with Yacc_state::m_lock_type in order to emphasize that this value is relevant only during parsing. Unlike for LEX::lock_option the default value for Yacc_state::m_lock_type is TL_READ_DEFAULT. Note that for cases when it is OK to take a "weak" read lock (e.g. simple SELECT) this lock type will be converted to TL_READ at open_tables() time. So this change won't cause negative change in behavior for such statements. OTOH this change ensures that, for example, for SELECTs which are used in stored functions TL_READ_NO_INSERT lock is taken when necessary and as result calls to such stored functions can be written to the binary log with correct serialization. sql/sql_load.cc: Load_log_event constructor now requires a parameter that indicates whether LOAD DATA is concurrent. sql/sql_parse.cc: LEX::lock_option was replaced with Yacc_state::m_lock_type. And instead of resetting the latter implicitly in mysql_init_multi_delete() we do it explicitly in the places in parser which call this function. sql/sql_priv.h: - To be able more easily distinguish high-priority SELECTs in st_select_lex::print() method added flag for HIGH_PRIORITY option. sql/sql_select.cc: Changed code not to rely on LEX::lock_option to determine that it is high-priority SELECT. It was replaced with Yacc_state::m_lock_type which is accessible only at parse time. So instead of LEX::lock_option we now rely on a newly introduced flag for st_select_lex::options - SELECT_HIGH_PRIORITY. sql/sql_show.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed. sql/sql_table.cc: Since LEX::reset_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore value of this member when this method is called by mysql_admin_table(), to make this code safe for re-execution. sql/sql_trigger.cc: Since LEX::reset_n_backup_query_tables_list() now also resets LEX::sql_command member (as it became part of Query_tables_list class) we have to restore it in cases when while working with proxy Query_table_list we assume that LEX::sql_command still corresponds to original SQL command being executed (for example, when we are logging statement to the binary log while having Query_tables_list reset and backed up). sql/sql_update.cc: Function read_lock_type_for_table() now takes pointers to Query_tables_list and TABLE_LIST elements as its arguments since to correctly determine lock type it needs to know what statement is being performed and whether table element for which lock type to be determined belongs to prelocking list. sql/sql_yacc.yy: - Removed st_select_lex::lock_option member as there is no real need for per-SELECT lock type (HIGH_PRIORITY option should apply to the whole statement. FOR UPDATE/LOCK IN SHARE MODE clauses can be handled without this member). The main effect which was achieved by introduction of this member, i.e. using TL_READ_DEFAULT lock type for subqueries, is now achieved by setting LEX::lock_option (or rather its replacement - Yacc_state::m_lock_type) to TL_READ_DEFAULT in almost all cases. - Replaced LEX::lock_option with Yacc_state::m_lock_type in order to emphasize that this value is relevant only during parsing. Unlike for LEX::lock_option the default value for Yacc_state::m_lock_type is TL_READ_DEFAULT. Note that for cases when it is OK to take a "weak" read lock (e.g. simple SELECT) this lock type will be converted to TL_READ at open_tables() time. So this change won't cause negative change in behavior for such statements. OTOH this change ensures that, for example, for SELECTs which are used in stored functions TL_READ_NO_INSERT lock is taken when necessary and as result calls to such stored functions can be written to the binary log with correct serialization. - To be able more easily distinguish high-priority SELECTs in st_select_lex::print() method we now use new flag in st_select_lex::options bit-field.
2010-04-28 12:04:11 +02:00
--echo # which reads a table through SELECT.
--echo #
--echo # Since a call to such function is written to the binary
--echo # log, it should be serialized with concurrent statements.
--echo # Hence, in this case reads should take strong locks on data.
let $statement= select f14();
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 5.3 SELECT that calls a function that doesn't modify data and
--echo # uses a CALL statement that reads a table via SELECT.
--echo #
--echo # In theory, calls to such functions won't get into the binary
--echo # log and thus don't need to acquire strong locks. But in practice
--echo # this fact is discovered too late to have any effect.
let $statement= select f15();
let $restore_table= ;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 5.4 INSERT which calls function which doesn't modify data and
--echo # uses CALL statement which reads table through SELECT.
--echo #
--echo # Since such statement is written to the binary log it should
--echo # be serialized with concurrent statements affecting data it
--echo # uses. Therefore it should take strong locks on data it reads.
let $statement= insert into t2 values (f15()+5);
let $restore_table= t2;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 6. Statements that use triggers.
--echo #
--echo #
--echo # 6.1 Statement invoking a trigger that reads table via SELECT.
--echo #
--echo # Since this statement is written to the binary log it should
--echo # be serialized with concurrent statements affecting the data
--echo # it uses. Therefore, it should take strong locks on the data
--echo # it reads.
let $statement= insert into t4 values (2);
let $restore_table= t4;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 6.2 Statement invoking a trigger that reads table through
--echo # a subquery in a control construct.
--echo #
--echo # The above is true for this statement as well.
let $statement= update t4 set l= 2 where l = 1;
let $restore_table= t4;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 6.3 Statement invoking a trigger that reads a table through
--echo # a view.
--echo #
--echo # And for this statement.
let $statement= delete from t4 where l = 1;
let $restore_table= t4;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 6.4 Statement invoking a trigger that reads a table through
--echo # a stored function.
--echo #
--echo # And for this statement.
let $statement= insert into t5 values (2);
let $restore_table= t5;
--source include/check_no_concurrent_insert.inc
--echo #
--echo # 6.5 Statement invoking a trigger that reads a table through
--echo # stored procedure.
--echo #
--echo # And for this statement.
let $statement= update t5 set l= 2 where l = 1;
let $restore_table= t5;
--source include/check_no_concurrent_insert.inc
--echo # Clean-up.
drop function f1;
drop function f2;
drop function f3;
drop function f4;
drop function f5;
drop function f6;
drop function f7;
drop function f8;
drop function f9;
drop function f10;
drop function f11;
drop function f12;
drop function f13;
drop function f14;
drop function f15;
drop view v1, v2;
drop procedure p1;
drop procedure p2;
drop table t1, t2, t3, t4, t5;
disconnect con1;
disconnect con2;
set @@global.concurrent_insert= @old_concurrent_insert;
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
--echo #
--echo # Test for bug #45143 "All connections hang on concurrent ALTER TABLE".
--echo #
--echo # Concurrent execution of statements which required weak write lock
--echo # (TL_WRITE_ALLOW_WRITE) on several instances of the same table and
--echo # statements which tried to acquire stronger write lock (TL_WRITE,
--echo # TL_WRITE_ALLOW_READ) on this table might have led to deadlock.
2010-02-04 23:08:08 +01:00
#
# Suppress warnings for INSERTs that use get_lock().
#
disable_query_log;
BUG#50670: Slave stops with error code 1644 Clarified error messages related to unsafe statements: - avoid the internal technical term "row injection" - use 'binary log' instead of 'binlog' - avoid the word 'unsafeness' mysql-test/extra/binlog_tests/blackhole.test: updated suppression pattern mysql-test/extra/binlog_tests/mix_innodb_myisam_binlog.test: updated suppression pattern mysql-test/extra/rpl_tests/rpl_binlog_max_cache_size.test: updated suppression pattern mysql-test/extra/rpl_tests/rpl_insert_id.test: updated suppression pattern mysql-test/extra/rpl_tests/rpl_loaddata.test: updated suppression pattern mysql-test/extra/rpl_tests/rpl_mixing_engines.test: updated suppression pattern mysql-test/extra/rpl_tests/rpl_start_stop_slave.test: updated suppression pattern mysql-test/extra/rpl_tests/rpl_stm_000001.test: updated suppression pattern mysql-test/extra/rpl_tests/rpl_stop_middle_group.test: updated suppression pattern mysql-test/r/archive.result: updated result file mysql-test/r/commit_1innodb.result: updated result file mysql-test/r/ctype_cp932_binlog_stm.result: updated result file mysql-test/r/partition_innodb_stmt.result: updated result file mysql-test/suite/binlog/r/binlog_innodb.result: updated result file mysql-test/suite/binlog/r/binlog_killed.result: updated result file mysql-test/suite/binlog/r/binlog_multi_engine.result: updated result file mysql-test/suite/binlog/r/binlog_row_mix_innodb_myisam.result: updated result file mysql-test/suite/binlog/r/binlog_statement_insert_delayed.result: updated result file mysql-test/suite/binlog/r/binlog_stm_blackhole.result: updated result file mysql-test/suite/binlog/r/binlog_stm_do_db.result: updated result file mysql-test/suite/binlog/r/binlog_stm_mix_innodb_myisam.result: updated result file mysql-test/suite/binlog/r/binlog_stm_ps.result: updated result file mysql-test/suite/binlog/r/binlog_stm_row.result: updated result file mysql-test/suite/binlog/r/binlog_stm_unsafe_warning.result: updated result file mysql-test/suite/binlog/r/binlog_unsafe.result: updated result file mysql-test/suite/binlog/t/binlog_killed.test: updated suppression pattern mysql-test/suite/binlog/t/binlog_multi_engine.test: updated suppression pattern mysql-test/suite/binlog/t/binlog_statement_insert_delayed.test: updated suppression pattern mysql-test/suite/binlog/t/binlog_stm_ps.test: updated suppression pattern mysql-test/suite/binlog/t/binlog_stm_row.test: updated suppression pattern mysql-test/suite/binlog/t/binlog_stm_unsafe_warning.test: updated suppression pattern mysql-test/suite/binlog/t/binlog_tmp_table.test: updated suppression pattern mysql-test/suite/binlog/t/binlog_unsafe.test: updated suppression pattern mysql-test/suite/ndb/r/ndb_binlog_format.result: updated result file mysql-test/suite/ndb/t/ndb_binlog_format.test: updated suppression pattern mysql-test/suite/perfschema/r/binlog_stmt.result: updated result file mysql-test/suite/perfschema/t/binlog_stmt.test: updated suppression pattern mysql-test/suite/rpl/r/rpl_begin_commit_rollback.result: updated result file mysql-test/suite/rpl/r/rpl_blackhole.result: updated result file mysql-test/suite/rpl/r/rpl_concurrency_error.result: updated result file mysql-test/suite/rpl/r/rpl_get_lock.result: updated result file mysql-test/suite/rpl/r/rpl_insert_id.result: updated result file mysql-test/suite/rpl/r/rpl_insert_ignore.result: updated result file mysql-test/suite/rpl/r/rpl_misc_functions.result: updated result file mysql-test/suite/rpl/r/rpl_mixed_binlog_max_cache_size.result: updated result file mysql-test/suite/rpl/r/rpl_mixed_mixing_engines.result: updated result file mysql-test/suite/rpl/r/rpl_mysql_upgrade.result: updated result file mysql-test/suite/rpl/r/rpl_non_direct_mixed_mixing_engines.result: updated result file mysql-test/suite/rpl/r/rpl_non_direct_row_mixing_engines.result: updated result file mysql-test/suite/rpl/r/rpl_non_direct_stm_mixing_engines.result: updated result file mysql-test/suite/rpl/r/rpl_nondeterministic_functions.result: updated result file mysql-test/suite/rpl/r/rpl_read_only.result: updated result file mysql-test/suite/rpl/r/rpl_row_binlog_max_cache_size.result: updated result file mysql-test/suite/rpl/r/rpl_row_inexist_tbl.result: updated result file mysql-test/suite/rpl/r/rpl_row_mixing_engines.result: updated result file mysql-test/suite/rpl/r/rpl_slow_query_log.result: updated result file mysql-test/suite/rpl/r/rpl_sp.result: updated result file mysql-test/suite/rpl/r/rpl_stm_000001.result: updated result file mysql-test/suite/rpl/r/rpl_stm_auto_increment_bug33029.result: updated result file mysql-test/suite/rpl/r/rpl_stm_binlog_max_cache_size.result: updated result file mysql-test/suite/rpl/r/rpl_stm_loadfile.result: updated result file mysql-test/suite/rpl/r/rpl_stm_mixing_engines.result: updated result file mysql-test/suite/rpl/r/rpl_stm_start_stop_slave.result: updated result file mysql-test/suite/rpl/r/rpl_stm_stop_middle_group.result: updated result file mysql-test/suite/rpl/r/rpl_temp_temporary.result: updated result file mysql-test/suite/rpl/r/rpl_variables_stm.result: updated result file mysql-test/suite/rpl/t/rpl000013.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_begin_commit_rollback.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_blackhole.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_concurrency_error.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_err_ignoredtable.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_get_lock.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_insert.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_insert_id.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_insert_ignore.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_invoked_features.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_misc_functions.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_mixed_bit_pk.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_mysql_upgrade.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_nondeterministic_functions.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_optimize.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_read_only.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_semi_sync.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_semi_sync_event.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_session_var.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_slow_query_log.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_sp.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_stm_auto_increment_bug33029.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_stm_found_rows.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_stm_insert_delayed.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_stm_loadfile.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_temp_table.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_temp_temporary.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_temporary.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_timezone.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_trigger.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_udf.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_user_variables.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_variables_stm.test: updated suppression pattern mysql-test/suite/rpl/t/rpl_view_multi.test: updated suppression pattern mysql-test/suite/rpl_ndb/r/rpl_ndb_binlog_format_errors.result: updated result file mysql-test/suite/rpl_ndb/t/rpl_ndb_binlog_format_errors.test: updated suppression pattern mysql-test/suite/rpl_ndb/t/rpl_ndb_stm_innodb.test: updated suppression pattern mysql-test/suite/sys_vars/t/rpl_init_slave_func.test: updated suppression pattern mysql-test/t/archive.test: updated suppression pattern mysql-test/t/commit_1innodb.test: updated suppression pattern mysql-test/t/create_select_tmp.test: updated suppression pattern mysql-test/t/ctype_cp932_binlog_stm.test: updated suppression pattern mysql-test/t/lock_sync.test: updated suppression pattern mysql-test/t/mysqlbinlog.test: updated suppression pattern mysql-test/t/mysqldump.test: updated suppression pattern mysql-test/t/sp_trans.test: updated suppression pattern sql/log_event.cc: Clarified error message. sql/share/errmsg-utf8.txt: Clarified error messages.
2010-04-28 14:47:49 +02:00
call mtr.add_suppression("Unsafe statement written to the binary log using statement format since BINLOG_FORMAT = STATEMENT");
2010-02-04 23:08:08 +01:00
enable_query_log;
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
--disable_warnings
drop table if exists t1;
Implement new type-of-operation-aware metadata locks. Add a wait-for graph based deadlock detector to the MDL subsystem. Fixes bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock" and bug #37346 "innodb does not detect deadlock between update and alter table". The first bug manifested itself as an unwarranted abort of a transaction with ER_LOCK_DEADLOCK error by a concurrent ALTER statement, when this transaction tried to repeat use of a table, which it has already used in a similar fashion before ALTER started. The second bug showed up as a deadlock between table-level locks and InnoDB row locks, which was "detected" only after innodb_lock_wait_timeout timeout. A transaction would start using the table and modify a few rows. Then ALTER TABLE would come in, and start copying rows into a temporary table. Eventually it would stumble on the modified records and get blocked on a row lock. The first transaction would try to do more updates, and get blocked on thr_lock.c lock. This situation of circular wait would only get resolved by a timeout. Both these bugs stemmed from inadequate solutions to the problem of deadlocks occurring between different locking subsystems. In the first case we tried to avoid deadlocks between metadata locking and table-level locking subsystems, when upgrading shared metadata lock to exclusive one. Transactions holding the shared lock on the table and waiting for some table-level lock used to be aborted too aggressively. We also allowed ALTER TABLE to start in presence of transactions that modify the subject table. ALTER TABLE acquires TL_WRITE_ALLOW_READ lock at start, and that block all writes against the table (naturally, we don't want any writes to be lost when switching the old and the new table). TL_WRITE_ALLOW_READ lock, in turn, would block the started transaction on thr_lock.c lock, should they do more updates. This, again, lead to the need to abort such transactions. The second bug occurred simply because we didn't have any mechanism to detect deadlocks between the table-level locks in thr_lock.c and row-level locks in InnoDB, other than innodb_lock_wait_timeout. This patch solves both these problems by moving lock conflicts which are causing these deadlocks into the metadata locking subsystem, thus making it possible to avoid or detect such deadlocks inside MDL. To do this we introduce new type-of-operation-aware metadata locks, which allow MDL subsystem to know not only the fact that transaction has used or is going to use some object but also what kind of operation it has carried out or going to carry out on the object. This, along with the addition of a special kind of upgradable metadata lock, allows ALTER TABLE to wait until all transactions which has updated the table to go away. This solves the second issue. Another special type of upgradable metadata lock is acquired by LOCK TABLE WRITE. This second lock type allows to solve the first issue, since abortion of table-level locks in event of DDL under LOCK TABLES becomes also unnecessary. Below follows the list of incompatible changes introduced by this patch: - From now on, ALTER TABLE and CREATE/DROP TRIGGER SQL (i.e. those statements that acquire TL_WRITE_ALLOW_READ lock) wait for all transactions which has *updated* the table to complete. - From now on, LOCK TABLES ... WRITE, REPAIR/OPTIMIZE TABLE (i.e. all statements which acquire TL_WRITE table-level lock) wait for all transaction which *updated or read* from the table to complete. As a consequence, innodb_table_locks=0 option no longer applies to LOCK TABLES ... WRITE. - DROP DATABASE, DROP TABLE, RENAME TABLE no longer abort statements or transactions which use tables being dropped or renamed, and instead wait for these transactions to complete. - Since LOCK TABLES WRITE now takes a special metadata lock, not compatible with with reads or writes against the subject table and transaction-wide, thr_lock.c deadlock avoidance algorithm that used to ensure absence of deadlocks between LOCK TABLES WRITE and other statements is no longer sufficient, even for MyISAM. The wait-for graph based deadlock detector of MDL subsystem may sometimes be necessary and is involved. This may lead to ER_LOCK_DEADLOCK error produced for multi-statement transactions even if these only use MyISAM: session 1: session 2: begin; update t1 ... lock table t2 write, t1 write; -- gets a lock on t2, blocks on t1 update t2 ... (ER_LOCK_DEADLOCK) - Finally, support of LOW_PRIORITY option for LOCK TABLES ... WRITE was abandoned. LOCK TABLE ... LOW_PRIORITY WRITE from now on has the same priority as the usual LOCK TABLE ... WRITE. SELECT HIGH PRIORITY no longer trumps LOCK TABLE ... WRITE in the wait queue. - We do not take upgradable metadata locks on implicitly locked tables. So if one has, say, a view v1 that uses table t1, and issues: LOCK TABLE v1 WRITE; FLUSH TABLE t1; -- (or just 'FLUSH TABLES'), an error is produced. In order to be able to perform DDL on a table under LOCK TABLES, the table must be locked explicitly in the LOCK TABLES list. mysql-test/include/handler.inc: Adjusted test case to trigger an execution path on which bug 41110 "crash with handler command when used concurrently with alter table" and bug 41112 "crash in mysql_ha_close_table/get_lock_data with alter table" were originally discovered. Left old test case which no longer triggers this execution path for the sake of coverage. Added test coverage for HANDLER SQL statements and type-aware metadata locks. Added a test for the global shared lock and HANDLER SQL. Updated tests to take into account that the old simple deadlock detection heuristics was replaced with a graph-based deadlock detector. mysql-test/r/debug_sync.result: Updated results (see debug_sync.test). mysql-test/r/handler_innodb.result: Updated results (see handler.inc test). mysql-test/r/handler_myisam.result: Updated results (see handler.inc test). mysql-test/r/innodb-lock.result: Updated results (see innodb-lock.test). mysql-test/r/innodb_mysql_lock.result: Updated results (see innodb_mysql_lock.test). mysql-test/r/lock.result: Updated results (see lock.test). mysql-test/r/lock_multi.result: Updated results (see lock_multi.test). mysql-test/r/lock_sync.result: Updated results (see lock_sync.test). mysql-test/r/mdl_sync.result: Updated results (see mdl_sync.test). mysql-test/r/sp-threads.result: SHOW PROCESSLIST output has changed due to the fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/r/truncate_coverage.result: Updated results (see truncate_coverage.test). mysql-test/suite/funcs_1/datadict/processlist_val.inc: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/funcs_1/r/processlist_val_no_prot.result: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/rpl/t/rpl_sp.test: Updated to a new SHOW PROCESSLIST state name. mysql-test/t/debug_sync.test: Use LOCK TABLES READ instead of LOCK TABLES WRITE as the latter no longer allows to trigger execution path involving waiting on thr_lock.c lock and therefore reaching debug sync-point covered by this test. mysql-test/t/innodb-lock.test: Adjusted test case to the fact that innodb_table_locks=0 option is no longer supported, since LOCK TABLES WRITE handles all its conflicts within MDL subsystem. mysql-test/t/innodb_mysql_lock.test: Added test for bug #37346 "innodb does not detect deadlock between update and alter table". mysql-test/t/lock.test: Added test coverage which checks the fact that we no longer support DDL under LOCK TABLES on tables which were locked implicitly. Adjusted existing test cases accordingly. mysql-test/t/lock_multi.test: Added test for bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock". Adjusted other test cases to take into account the fact that waiting for LOCK TABLES ... WRITE now happens within MDL subsystem. mysql-test/t/lock_sync.test: Since LOCK TABLES ... WRITE now takes SNRW metadata lock for tables locked explicitly we have to implicitly lock InnoDB tables (through view) to trigger the table-level lock conflict between TL_WRITE and TL_WRITE_ALLOW_WRITE. mysql-test/t/mdl_sync.test: Added basic test coverage for type-of-operation-aware metadata locks. Also covered with tests some use cases involving HANDLER statements in which a deadlock could arise. Adjusted existing tests to take type-of-operation-aware MDL into account. mysql-test/t/multi_update.test: Update to a new SHOW PROCESSLIST state name. mysql-test/t/truncate_coverage.test: Adjusted test case after making LOCK TABLES WRITE to wait until transactions that use the table to be locked are completed. Updated to the changed name of DEBUG_SYNC point. sql/handler.cc: Global read lock functionality has been moved into a class. sql/lock.cc: Global read lock functionality has been moved into a class. Updated code to use the new MDL API. sql/mdl.cc: Introduced new type-of-operation aware metadata locks. To do this: - Changed MDL_lock to use one list for waiting requests and one list for granted requests. For each list, added a bitmap that holds information what lock types a list contains. Added a helper class MDL_lock::List to manipulate with granted and waited lists while keeping the bitmaps in sync with list contents. - Changed lock-compatibility functions to use bitmaps that define compatibility. - Introduced a graph based deadlock detector inspired by waiting_threads.c from Maria implementation. - Now that we have a deadlock detector, and no longer have a global lock to protect individual lock objects, but rather use an rw lock per object, removed redundant code for upgrade, and the global read lock. Changed the MDL API to no longer require the caller to acquire the global intention exclusive lock by means of a separate method. Removed a few more methods that became redundant. - Removed deadlock detection heuristic, it has been made obsolete by the deadlock detector. - With operation-type-aware metadata locks, MDL subsystem has become aware of potential conflicts between DDL and open transactions. This made it possible to remove calls to mysql_abort_transactions_with_shared_lock() from acquisition paths for exclusive lock and lock upgrade. Now we can simply wait for these transactions to complete without fear of deadlock. Function mysql_lock_abort() has also become unnecessary for all conflicting cases except when a DDL conflicts with a connection that has an open HANDLER. sql/mdl.h: Introduced new type-of-operation aware metadata locks. Introduced a graph based deadlock detector and supporting methods. Added comments. God rid of redundant API calls. Renamed m_lt_or_ha_sentinel to m_trans_sentinel, since now it guards the global read lock as well as LOCK TABLES and HANDLER locks. sql/mysql_priv.h: Moved the global read lock functionality into a class. Added MYSQL_OPEN_FORCE_SHARED_MDL flag which forces open_tables() to take MDL_SHARED on tables instead of metadata locks specified in the parser. We use this to allow PREPARE run concurrently in presence of LOCK TABLES ... WRITE. Added signature for find_table_for_mdl_ugprade(). sql/set_var.cc: Global read lock functionality has been moved into a class. sql/sp_head.cc: When creating TABLE_LIST elements for prelocking or system tables set the type of request for metadata lock according to the operation that will be performed on the table. sql/sql_base.cc: - Updated code to use the new MDL API. - In order to avoid locks starvation we take upgradable locks all at once. As result implicitly locked tables no longer get an upgradable lock. Consequently DDL and FLUSH TABLES for such tables is prohibited. find_write_locked_table() was replaced by find_table_for_mdl_upgrade() function. open_table() was adjusted to return TABLE instance with upgradable ticket when necessary. - We no longer wait for all locks on OT_WAIT back off action -- only on the lock that caused the wait conflict. Moreover, now we distinguish cases when we have to wait due to conflict in MDL and old version of table in TDC. - Upate mysql_notify_threads_having_share_locks() to only abort thr_lock.c waits of threads that have open HANDLERs, since lock conflicts with only these threads now can lead to deadlocks not detectable by the MDL deadlock detector. - Remove mysql_abort_transactions_with_shared_locks() which is no longer needed. sql/sql_class.cc: Global read lock functionality has been moved into a class. Re-arranged code in THD::cleanup() to simplify assert. sql/sql_class.h: Introduced class to incapsulate global read lock functionality. Now sentinel in MDL subsystem guards the global read lock as well as LOCK TABLES and HANDLER locks. Adjusted code accordingly. sql/sql_db.cc: Global read lock functionality has been moved into a class. sql/sql_delete.cc: We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result TRUNCATE TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_handler.cc: Inform MDL_context about presence of open HANDLERs. Since HANLDERs break MDL protocol by acquiring table-level lock while holding only S metadata lock on a table MDL subsystem should take special care about such contexts (Now this is the only case when mysql_lock_abort() is used). sql/sql_parse.cc: Global read lock functionality has been moved into a class. Do not take upgradable metadata locks when opening tables for CREATE TABLE SELECT as it is not necessary and limits concurrency. When initializing TABLE_LIST objects before adding them to the table list set the type of request for metadata lock according to the operation that will be performed on the table. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result FLUSH TABLES is no longer allowed for such tables. sql/sql_prepare.cc: Use MYSQL_OPEN_FORCE_SHARED_MDL flag when opening tables during PREPARE. This allows PREPARE to run concurrently in presence of LOCK TABLES ... WRITE. sql/sql_rename.cc: Global read lock functionality has been moved into a class. sql/sql_show.cc: Updated code to use the new MDL API. sql/sql_table.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result DROP TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_trigger.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result CREATE/DROP TRIGGER is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_view.cc: Global read lock functionality has been moved into a class. Fixed results of wrong merge that led to misuse of GLR API. CREATE VIEW statement is not a commit statement. sql/table.cc: When resetting TABLE_LIST objects for PS or SP re-execution set the type of request for metadata lock according to the operation that will be performed on the table. Do the same in auxiliary function initializing metadata lock requests in a table list. sql/table.h: When initializing TABLE_LIST objects set the type of request for metadata lock according to the operation that will be performed on the table. sql/transaction.cc: Global read lock functionality has been moved into a class.
2010-02-01 12:43:06 +01:00
drop view if exists v1;
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
--enable_warnings
--echo # Create auxiliary connections used through the test.
connect (con_bug45143_1,localhost,root,,test,,);
connect (con_bug45143_3,localhost,root,,test,,);
connect (con_bug45143_2,localhost,root,,test,,);
connection default;
--echo # Reset DEBUG_SYNC facility before using it.
set debug_sync= 'RESET';
--echo # Turn off logging so calls to locking subsystem performed
--echo # for general_log table won't interfere with our test.
set @old_general_log = @@global.general_log;
set @@global.general_log= OFF;
create table t1 (i int) engine=InnoDB;
Implement new type-of-operation-aware metadata locks. Add a wait-for graph based deadlock detector to the MDL subsystem. Fixes bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock" and bug #37346 "innodb does not detect deadlock between update and alter table". The first bug manifested itself as an unwarranted abort of a transaction with ER_LOCK_DEADLOCK error by a concurrent ALTER statement, when this transaction tried to repeat use of a table, which it has already used in a similar fashion before ALTER started. The second bug showed up as a deadlock between table-level locks and InnoDB row locks, which was "detected" only after innodb_lock_wait_timeout timeout. A transaction would start using the table and modify a few rows. Then ALTER TABLE would come in, and start copying rows into a temporary table. Eventually it would stumble on the modified records and get blocked on a row lock. The first transaction would try to do more updates, and get blocked on thr_lock.c lock. This situation of circular wait would only get resolved by a timeout. Both these bugs stemmed from inadequate solutions to the problem of deadlocks occurring between different locking subsystems. In the first case we tried to avoid deadlocks between metadata locking and table-level locking subsystems, when upgrading shared metadata lock to exclusive one. Transactions holding the shared lock on the table and waiting for some table-level lock used to be aborted too aggressively. We also allowed ALTER TABLE to start in presence of transactions that modify the subject table. ALTER TABLE acquires TL_WRITE_ALLOW_READ lock at start, and that block all writes against the table (naturally, we don't want any writes to be lost when switching the old and the new table). TL_WRITE_ALLOW_READ lock, in turn, would block the started transaction on thr_lock.c lock, should they do more updates. This, again, lead to the need to abort such transactions. The second bug occurred simply because we didn't have any mechanism to detect deadlocks between the table-level locks in thr_lock.c and row-level locks in InnoDB, other than innodb_lock_wait_timeout. This patch solves both these problems by moving lock conflicts which are causing these deadlocks into the metadata locking subsystem, thus making it possible to avoid or detect such deadlocks inside MDL. To do this we introduce new type-of-operation-aware metadata locks, which allow MDL subsystem to know not only the fact that transaction has used or is going to use some object but also what kind of operation it has carried out or going to carry out on the object. This, along with the addition of a special kind of upgradable metadata lock, allows ALTER TABLE to wait until all transactions which has updated the table to go away. This solves the second issue. Another special type of upgradable metadata lock is acquired by LOCK TABLE WRITE. This second lock type allows to solve the first issue, since abortion of table-level locks in event of DDL under LOCK TABLES becomes also unnecessary. Below follows the list of incompatible changes introduced by this patch: - From now on, ALTER TABLE and CREATE/DROP TRIGGER SQL (i.e. those statements that acquire TL_WRITE_ALLOW_READ lock) wait for all transactions which has *updated* the table to complete. - From now on, LOCK TABLES ... WRITE, REPAIR/OPTIMIZE TABLE (i.e. all statements which acquire TL_WRITE table-level lock) wait for all transaction which *updated or read* from the table to complete. As a consequence, innodb_table_locks=0 option no longer applies to LOCK TABLES ... WRITE. - DROP DATABASE, DROP TABLE, RENAME TABLE no longer abort statements or transactions which use tables being dropped or renamed, and instead wait for these transactions to complete. - Since LOCK TABLES WRITE now takes a special metadata lock, not compatible with with reads or writes against the subject table and transaction-wide, thr_lock.c deadlock avoidance algorithm that used to ensure absence of deadlocks between LOCK TABLES WRITE and other statements is no longer sufficient, even for MyISAM. The wait-for graph based deadlock detector of MDL subsystem may sometimes be necessary and is involved. This may lead to ER_LOCK_DEADLOCK error produced for multi-statement transactions even if these only use MyISAM: session 1: session 2: begin; update t1 ... lock table t2 write, t1 write; -- gets a lock on t2, blocks on t1 update t2 ... (ER_LOCK_DEADLOCK) - Finally, support of LOW_PRIORITY option for LOCK TABLES ... WRITE was abandoned. LOCK TABLE ... LOW_PRIORITY WRITE from now on has the same priority as the usual LOCK TABLE ... WRITE. SELECT HIGH PRIORITY no longer trumps LOCK TABLE ... WRITE in the wait queue. - We do not take upgradable metadata locks on implicitly locked tables. So if one has, say, a view v1 that uses table t1, and issues: LOCK TABLE v1 WRITE; FLUSH TABLE t1; -- (or just 'FLUSH TABLES'), an error is produced. In order to be able to perform DDL on a table under LOCK TABLES, the table must be locked explicitly in the LOCK TABLES list. mysql-test/include/handler.inc: Adjusted test case to trigger an execution path on which bug 41110 "crash with handler command when used concurrently with alter table" and bug 41112 "crash in mysql_ha_close_table/get_lock_data with alter table" were originally discovered. Left old test case which no longer triggers this execution path for the sake of coverage. Added test coverage for HANDLER SQL statements and type-aware metadata locks. Added a test for the global shared lock and HANDLER SQL. Updated tests to take into account that the old simple deadlock detection heuristics was replaced with a graph-based deadlock detector. mysql-test/r/debug_sync.result: Updated results (see debug_sync.test). mysql-test/r/handler_innodb.result: Updated results (see handler.inc test). mysql-test/r/handler_myisam.result: Updated results (see handler.inc test). mysql-test/r/innodb-lock.result: Updated results (see innodb-lock.test). mysql-test/r/innodb_mysql_lock.result: Updated results (see innodb_mysql_lock.test). mysql-test/r/lock.result: Updated results (see lock.test). mysql-test/r/lock_multi.result: Updated results (see lock_multi.test). mysql-test/r/lock_sync.result: Updated results (see lock_sync.test). mysql-test/r/mdl_sync.result: Updated results (see mdl_sync.test). mysql-test/r/sp-threads.result: SHOW PROCESSLIST output has changed due to the fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/r/truncate_coverage.result: Updated results (see truncate_coverage.test). mysql-test/suite/funcs_1/datadict/processlist_val.inc: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/funcs_1/r/processlist_val_no_prot.result: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/rpl/t/rpl_sp.test: Updated to a new SHOW PROCESSLIST state name. mysql-test/t/debug_sync.test: Use LOCK TABLES READ instead of LOCK TABLES WRITE as the latter no longer allows to trigger execution path involving waiting on thr_lock.c lock and therefore reaching debug sync-point covered by this test. mysql-test/t/innodb-lock.test: Adjusted test case to the fact that innodb_table_locks=0 option is no longer supported, since LOCK TABLES WRITE handles all its conflicts within MDL subsystem. mysql-test/t/innodb_mysql_lock.test: Added test for bug #37346 "innodb does not detect deadlock between update and alter table". mysql-test/t/lock.test: Added test coverage which checks the fact that we no longer support DDL under LOCK TABLES on tables which were locked implicitly. Adjusted existing test cases accordingly. mysql-test/t/lock_multi.test: Added test for bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock". Adjusted other test cases to take into account the fact that waiting for LOCK TABLES ... WRITE now happens within MDL subsystem. mysql-test/t/lock_sync.test: Since LOCK TABLES ... WRITE now takes SNRW metadata lock for tables locked explicitly we have to implicitly lock InnoDB tables (through view) to trigger the table-level lock conflict between TL_WRITE and TL_WRITE_ALLOW_WRITE. mysql-test/t/mdl_sync.test: Added basic test coverage for type-of-operation-aware metadata locks. Also covered with tests some use cases involving HANDLER statements in which a deadlock could arise. Adjusted existing tests to take type-of-operation-aware MDL into account. mysql-test/t/multi_update.test: Update to a new SHOW PROCESSLIST state name. mysql-test/t/truncate_coverage.test: Adjusted test case after making LOCK TABLES WRITE to wait until transactions that use the table to be locked are completed. Updated to the changed name of DEBUG_SYNC point. sql/handler.cc: Global read lock functionality has been moved into a class. sql/lock.cc: Global read lock functionality has been moved into a class. Updated code to use the new MDL API. sql/mdl.cc: Introduced new type-of-operation aware metadata locks. To do this: - Changed MDL_lock to use one list for waiting requests and one list for granted requests. For each list, added a bitmap that holds information what lock types a list contains. Added a helper class MDL_lock::List to manipulate with granted and waited lists while keeping the bitmaps in sync with list contents. - Changed lock-compatibility functions to use bitmaps that define compatibility. - Introduced a graph based deadlock detector inspired by waiting_threads.c from Maria implementation. - Now that we have a deadlock detector, and no longer have a global lock to protect individual lock objects, but rather use an rw lock per object, removed redundant code for upgrade, and the global read lock. Changed the MDL API to no longer require the caller to acquire the global intention exclusive lock by means of a separate method. Removed a few more methods that became redundant. - Removed deadlock detection heuristic, it has been made obsolete by the deadlock detector. - With operation-type-aware metadata locks, MDL subsystem has become aware of potential conflicts between DDL and open transactions. This made it possible to remove calls to mysql_abort_transactions_with_shared_lock() from acquisition paths for exclusive lock and lock upgrade. Now we can simply wait for these transactions to complete without fear of deadlock. Function mysql_lock_abort() has also become unnecessary for all conflicting cases except when a DDL conflicts with a connection that has an open HANDLER. sql/mdl.h: Introduced new type-of-operation aware metadata locks. Introduced a graph based deadlock detector and supporting methods. Added comments. God rid of redundant API calls. Renamed m_lt_or_ha_sentinel to m_trans_sentinel, since now it guards the global read lock as well as LOCK TABLES and HANDLER locks. sql/mysql_priv.h: Moved the global read lock functionality into a class. Added MYSQL_OPEN_FORCE_SHARED_MDL flag which forces open_tables() to take MDL_SHARED on tables instead of metadata locks specified in the parser. We use this to allow PREPARE run concurrently in presence of LOCK TABLES ... WRITE. Added signature for find_table_for_mdl_ugprade(). sql/set_var.cc: Global read lock functionality has been moved into a class. sql/sp_head.cc: When creating TABLE_LIST elements for prelocking or system tables set the type of request for metadata lock according to the operation that will be performed on the table. sql/sql_base.cc: - Updated code to use the new MDL API. - In order to avoid locks starvation we take upgradable locks all at once. As result implicitly locked tables no longer get an upgradable lock. Consequently DDL and FLUSH TABLES for such tables is prohibited. find_write_locked_table() was replaced by find_table_for_mdl_upgrade() function. open_table() was adjusted to return TABLE instance with upgradable ticket when necessary. - We no longer wait for all locks on OT_WAIT back off action -- only on the lock that caused the wait conflict. Moreover, now we distinguish cases when we have to wait due to conflict in MDL and old version of table in TDC. - Upate mysql_notify_threads_having_share_locks() to only abort thr_lock.c waits of threads that have open HANDLERs, since lock conflicts with only these threads now can lead to deadlocks not detectable by the MDL deadlock detector. - Remove mysql_abort_transactions_with_shared_locks() which is no longer needed. sql/sql_class.cc: Global read lock functionality has been moved into a class. Re-arranged code in THD::cleanup() to simplify assert. sql/sql_class.h: Introduced class to incapsulate global read lock functionality. Now sentinel in MDL subsystem guards the global read lock as well as LOCK TABLES and HANDLER locks. Adjusted code accordingly. sql/sql_db.cc: Global read lock functionality has been moved into a class. sql/sql_delete.cc: We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result TRUNCATE TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_handler.cc: Inform MDL_context about presence of open HANDLERs. Since HANLDERs break MDL protocol by acquiring table-level lock while holding only S metadata lock on a table MDL subsystem should take special care about such contexts (Now this is the only case when mysql_lock_abort() is used). sql/sql_parse.cc: Global read lock functionality has been moved into a class. Do not take upgradable metadata locks when opening tables for CREATE TABLE SELECT as it is not necessary and limits concurrency. When initializing TABLE_LIST objects before adding them to the table list set the type of request for metadata lock according to the operation that will be performed on the table. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result FLUSH TABLES is no longer allowed for such tables. sql/sql_prepare.cc: Use MYSQL_OPEN_FORCE_SHARED_MDL flag when opening tables during PREPARE. This allows PREPARE to run concurrently in presence of LOCK TABLES ... WRITE. sql/sql_rename.cc: Global read lock functionality has been moved into a class. sql/sql_show.cc: Updated code to use the new MDL API. sql/sql_table.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result DROP TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_trigger.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result CREATE/DROP TRIGGER is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_view.cc: Global read lock functionality has been moved into a class. Fixed results of wrong merge that led to misuse of GLR API. CREATE VIEW statement is not a commit statement. sql/table.cc: When resetting TABLE_LIST objects for PS or SP re-execution set the type of request for metadata lock according to the operation that will be performed on the table. Do the same in auxiliary function initializing metadata lock requests in a table list. sql/table.h: When initializing TABLE_LIST objects set the type of request for metadata lock according to the operation that will be performed on the table. sql/transaction.cc: Global read lock functionality has been moved into a class.
2010-02-01 12:43:06 +01:00
--echo # We have to use view in order to make LOCK TABLES avoid
--echo # acquiring SNRW metadata lock on table.
create view v1 as select * from t1;
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
insert into t1 values (1);
--echo # Prepare user lock which will be used for resuming execution of
--echo # the first statement after it acquires TL_WRITE_ALLOW_WRITE lock.
select get_lock("lock_bug45143_wait", 0);
--echo # Switch to connection 'con_bug45143_1'.
connection con_bug45143_1;
--echo # Sending:
--send insert into t1 values (get_lock("lock_bug45143_wait", 100));
--echo # Switch to connection 'con_bug45143_2'.
connection con_bug45143_2;
--echo # Wait until the above INSERT takes TL_WRITE_ALLOW_WRITE lock on 't1'
--echo # and then gets blocked on user lock 'lock_bug45143_wait'.
let $wait_condition= select count(*)= 1 from information_schema.processlist
where state= 'User lock' and
info='insert into t1 values (get_lock("lock_bug45143_wait", 100))';
--source include/wait_condition.inc
--echo # Ensure that upcoming SELECT waits after acquiring TL_WRITE_ALLOW_WRITE
--echo # lock for the first instance of 't1'.
set debug_sync='thr_multi_lock_after_thr_lock SIGNAL parked WAIT_FOR go';
--echo # Sending:
--send select count(*) > 0 from t1 as a, t1 as b for update;
--echo # Switch to connection 'con_bug45143_3'.
connection con_bug45143_3;
--echo # Wait until the above SELECT ... FOR UPDATE is blocked after
--echo # acquiring lock for the the first instance of 't1'.
set debug_sync= 'now WAIT_FOR parked';
--echo # Send LOCK TABLE statement which will try to get TL_WRITE lock on 't1':
Implement new type-of-operation-aware metadata locks. Add a wait-for graph based deadlock detector to the MDL subsystem. Fixes bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock" and bug #37346 "innodb does not detect deadlock between update and alter table". The first bug manifested itself as an unwarranted abort of a transaction with ER_LOCK_DEADLOCK error by a concurrent ALTER statement, when this transaction tried to repeat use of a table, which it has already used in a similar fashion before ALTER started. The second bug showed up as a deadlock between table-level locks and InnoDB row locks, which was "detected" only after innodb_lock_wait_timeout timeout. A transaction would start using the table and modify a few rows. Then ALTER TABLE would come in, and start copying rows into a temporary table. Eventually it would stumble on the modified records and get blocked on a row lock. The first transaction would try to do more updates, and get blocked on thr_lock.c lock. This situation of circular wait would only get resolved by a timeout. Both these bugs stemmed from inadequate solutions to the problem of deadlocks occurring between different locking subsystems. In the first case we tried to avoid deadlocks between metadata locking and table-level locking subsystems, when upgrading shared metadata lock to exclusive one. Transactions holding the shared lock on the table and waiting for some table-level lock used to be aborted too aggressively. We also allowed ALTER TABLE to start in presence of transactions that modify the subject table. ALTER TABLE acquires TL_WRITE_ALLOW_READ lock at start, and that block all writes against the table (naturally, we don't want any writes to be lost when switching the old and the new table). TL_WRITE_ALLOW_READ lock, in turn, would block the started transaction on thr_lock.c lock, should they do more updates. This, again, lead to the need to abort such transactions. The second bug occurred simply because we didn't have any mechanism to detect deadlocks between the table-level locks in thr_lock.c and row-level locks in InnoDB, other than innodb_lock_wait_timeout. This patch solves both these problems by moving lock conflicts which are causing these deadlocks into the metadata locking subsystem, thus making it possible to avoid or detect such deadlocks inside MDL. To do this we introduce new type-of-operation-aware metadata locks, which allow MDL subsystem to know not only the fact that transaction has used or is going to use some object but also what kind of operation it has carried out or going to carry out on the object. This, along with the addition of a special kind of upgradable metadata lock, allows ALTER TABLE to wait until all transactions which has updated the table to go away. This solves the second issue. Another special type of upgradable metadata lock is acquired by LOCK TABLE WRITE. This second lock type allows to solve the first issue, since abortion of table-level locks in event of DDL under LOCK TABLES becomes also unnecessary. Below follows the list of incompatible changes introduced by this patch: - From now on, ALTER TABLE and CREATE/DROP TRIGGER SQL (i.e. those statements that acquire TL_WRITE_ALLOW_READ lock) wait for all transactions which has *updated* the table to complete. - From now on, LOCK TABLES ... WRITE, REPAIR/OPTIMIZE TABLE (i.e. all statements which acquire TL_WRITE table-level lock) wait for all transaction which *updated or read* from the table to complete. As a consequence, innodb_table_locks=0 option no longer applies to LOCK TABLES ... WRITE. - DROP DATABASE, DROP TABLE, RENAME TABLE no longer abort statements or transactions which use tables being dropped or renamed, and instead wait for these transactions to complete. - Since LOCK TABLES WRITE now takes a special metadata lock, not compatible with with reads or writes against the subject table and transaction-wide, thr_lock.c deadlock avoidance algorithm that used to ensure absence of deadlocks between LOCK TABLES WRITE and other statements is no longer sufficient, even for MyISAM. The wait-for graph based deadlock detector of MDL subsystem may sometimes be necessary and is involved. This may lead to ER_LOCK_DEADLOCK error produced for multi-statement transactions even if these only use MyISAM: session 1: session 2: begin; update t1 ... lock table t2 write, t1 write; -- gets a lock on t2, blocks on t1 update t2 ... (ER_LOCK_DEADLOCK) - Finally, support of LOW_PRIORITY option for LOCK TABLES ... WRITE was abandoned. LOCK TABLE ... LOW_PRIORITY WRITE from now on has the same priority as the usual LOCK TABLE ... WRITE. SELECT HIGH PRIORITY no longer trumps LOCK TABLE ... WRITE in the wait queue. - We do not take upgradable metadata locks on implicitly locked tables. So if one has, say, a view v1 that uses table t1, and issues: LOCK TABLE v1 WRITE; FLUSH TABLE t1; -- (or just 'FLUSH TABLES'), an error is produced. In order to be able to perform DDL on a table under LOCK TABLES, the table must be locked explicitly in the LOCK TABLES list. mysql-test/include/handler.inc: Adjusted test case to trigger an execution path on which bug 41110 "crash with handler command when used concurrently with alter table" and bug 41112 "crash in mysql_ha_close_table/get_lock_data with alter table" were originally discovered. Left old test case which no longer triggers this execution path for the sake of coverage. Added test coverage for HANDLER SQL statements and type-aware metadata locks. Added a test for the global shared lock and HANDLER SQL. Updated tests to take into account that the old simple deadlock detection heuristics was replaced with a graph-based deadlock detector. mysql-test/r/debug_sync.result: Updated results (see debug_sync.test). mysql-test/r/handler_innodb.result: Updated results (see handler.inc test). mysql-test/r/handler_myisam.result: Updated results (see handler.inc test). mysql-test/r/innodb-lock.result: Updated results (see innodb-lock.test). mysql-test/r/innodb_mysql_lock.result: Updated results (see innodb_mysql_lock.test). mysql-test/r/lock.result: Updated results (see lock.test). mysql-test/r/lock_multi.result: Updated results (see lock_multi.test). mysql-test/r/lock_sync.result: Updated results (see lock_sync.test). mysql-test/r/mdl_sync.result: Updated results (see mdl_sync.test). mysql-test/r/sp-threads.result: SHOW PROCESSLIST output has changed due to the fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/r/truncate_coverage.result: Updated results (see truncate_coverage.test). mysql-test/suite/funcs_1/datadict/processlist_val.inc: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/funcs_1/r/processlist_val_no_prot.result: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/rpl/t/rpl_sp.test: Updated to a new SHOW PROCESSLIST state name. mysql-test/t/debug_sync.test: Use LOCK TABLES READ instead of LOCK TABLES WRITE as the latter no longer allows to trigger execution path involving waiting on thr_lock.c lock and therefore reaching debug sync-point covered by this test. mysql-test/t/innodb-lock.test: Adjusted test case to the fact that innodb_table_locks=0 option is no longer supported, since LOCK TABLES WRITE handles all its conflicts within MDL subsystem. mysql-test/t/innodb_mysql_lock.test: Added test for bug #37346 "innodb does not detect deadlock between update and alter table". mysql-test/t/lock.test: Added test coverage which checks the fact that we no longer support DDL under LOCK TABLES on tables which were locked implicitly. Adjusted existing test cases accordingly. mysql-test/t/lock_multi.test: Added test for bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock". Adjusted other test cases to take into account the fact that waiting for LOCK TABLES ... WRITE now happens within MDL subsystem. mysql-test/t/lock_sync.test: Since LOCK TABLES ... WRITE now takes SNRW metadata lock for tables locked explicitly we have to implicitly lock InnoDB tables (through view) to trigger the table-level lock conflict between TL_WRITE and TL_WRITE_ALLOW_WRITE. mysql-test/t/mdl_sync.test: Added basic test coverage for type-of-operation-aware metadata locks. Also covered with tests some use cases involving HANDLER statements in which a deadlock could arise. Adjusted existing tests to take type-of-operation-aware MDL into account. mysql-test/t/multi_update.test: Update to a new SHOW PROCESSLIST state name. mysql-test/t/truncate_coverage.test: Adjusted test case after making LOCK TABLES WRITE to wait until transactions that use the table to be locked are completed. Updated to the changed name of DEBUG_SYNC point. sql/handler.cc: Global read lock functionality has been moved into a class. sql/lock.cc: Global read lock functionality has been moved into a class. Updated code to use the new MDL API. sql/mdl.cc: Introduced new type-of-operation aware metadata locks. To do this: - Changed MDL_lock to use one list for waiting requests and one list for granted requests. For each list, added a bitmap that holds information what lock types a list contains. Added a helper class MDL_lock::List to manipulate with granted and waited lists while keeping the bitmaps in sync with list contents. - Changed lock-compatibility functions to use bitmaps that define compatibility. - Introduced a graph based deadlock detector inspired by waiting_threads.c from Maria implementation. - Now that we have a deadlock detector, and no longer have a global lock to protect individual lock objects, but rather use an rw lock per object, removed redundant code for upgrade, and the global read lock. Changed the MDL API to no longer require the caller to acquire the global intention exclusive lock by means of a separate method. Removed a few more methods that became redundant. - Removed deadlock detection heuristic, it has been made obsolete by the deadlock detector. - With operation-type-aware metadata locks, MDL subsystem has become aware of potential conflicts between DDL and open transactions. This made it possible to remove calls to mysql_abort_transactions_with_shared_lock() from acquisition paths for exclusive lock and lock upgrade. Now we can simply wait for these transactions to complete without fear of deadlock. Function mysql_lock_abort() has also become unnecessary for all conflicting cases except when a DDL conflicts with a connection that has an open HANDLER. sql/mdl.h: Introduced new type-of-operation aware metadata locks. Introduced a graph based deadlock detector and supporting methods. Added comments. God rid of redundant API calls. Renamed m_lt_or_ha_sentinel to m_trans_sentinel, since now it guards the global read lock as well as LOCK TABLES and HANDLER locks. sql/mysql_priv.h: Moved the global read lock functionality into a class. Added MYSQL_OPEN_FORCE_SHARED_MDL flag which forces open_tables() to take MDL_SHARED on tables instead of metadata locks specified in the parser. We use this to allow PREPARE run concurrently in presence of LOCK TABLES ... WRITE. Added signature for find_table_for_mdl_ugprade(). sql/set_var.cc: Global read lock functionality has been moved into a class. sql/sp_head.cc: When creating TABLE_LIST elements for prelocking or system tables set the type of request for metadata lock according to the operation that will be performed on the table. sql/sql_base.cc: - Updated code to use the new MDL API. - In order to avoid locks starvation we take upgradable locks all at once. As result implicitly locked tables no longer get an upgradable lock. Consequently DDL and FLUSH TABLES for such tables is prohibited. find_write_locked_table() was replaced by find_table_for_mdl_upgrade() function. open_table() was adjusted to return TABLE instance with upgradable ticket when necessary. - We no longer wait for all locks on OT_WAIT back off action -- only on the lock that caused the wait conflict. Moreover, now we distinguish cases when we have to wait due to conflict in MDL and old version of table in TDC. - Upate mysql_notify_threads_having_share_locks() to only abort thr_lock.c waits of threads that have open HANDLERs, since lock conflicts with only these threads now can lead to deadlocks not detectable by the MDL deadlock detector. - Remove mysql_abort_transactions_with_shared_locks() which is no longer needed. sql/sql_class.cc: Global read lock functionality has been moved into a class. Re-arranged code in THD::cleanup() to simplify assert. sql/sql_class.h: Introduced class to incapsulate global read lock functionality. Now sentinel in MDL subsystem guards the global read lock as well as LOCK TABLES and HANDLER locks. Adjusted code accordingly. sql/sql_db.cc: Global read lock functionality has been moved into a class. sql/sql_delete.cc: We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result TRUNCATE TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_handler.cc: Inform MDL_context about presence of open HANDLERs. Since HANLDERs break MDL protocol by acquiring table-level lock while holding only S metadata lock on a table MDL subsystem should take special care about such contexts (Now this is the only case when mysql_lock_abort() is used). sql/sql_parse.cc: Global read lock functionality has been moved into a class. Do not take upgradable metadata locks when opening tables for CREATE TABLE SELECT as it is not necessary and limits concurrency. When initializing TABLE_LIST objects before adding them to the table list set the type of request for metadata lock according to the operation that will be performed on the table. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result FLUSH TABLES is no longer allowed for such tables. sql/sql_prepare.cc: Use MYSQL_OPEN_FORCE_SHARED_MDL flag when opening tables during PREPARE. This allows PREPARE to run concurrently in presence of LOCK TABLES ... WRITE. sql/sql_rename.cc: Global read lock functionality has been moved into a class. sql/sql_show.cc: Updated code to use the new MDL API. sql/sql_table.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result DROP TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_trigger.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result CREATE/DROP TRIGGER is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_view.cc: Global read lock functionality has been moved into a class. Fixed results of wrong merge that led to misuse of GLR API. CREATE VIEW statement is not a commit statement. sql/table.cc: When resetting TABLE_LIST objects for PS or SP re-execution set the type of request for metadata lock according to the operation that will be performed on the table. Do the same in auxiliary function initializing metadata lock requests in a table list. sql/table.h: When initializing TABLE_LIST objects set the type of request for metadata lock according to the operation that will be performed on the table. sql/transaction.cc: Global read lock functionality has been moved into a class.
2010-02-01 12:43:06 +01:00
--send lock table v1 write;
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
--echo # Switch to connection 'default'.
connection default;
--echo # Wait until this LOCK TABLES statement starts waiting for table lock.
let $wait_condition= select count(*)= 1 from information_schema.processlist
Backport of: ------------------------------------------------------------ revno: 3035.4.1 committer: Davi Arnaut <Davi.Arnaut@Sun.COM> branch nick: 39897-6.0 timestamp: Thu 2009-01-15 12:17:57 -0200 message: Bug#39897: lock_multi fails in pushbuild: timeout waiting for processlist The problem is that relying on the "Table lock" thread state in its current position to detect that a thread is waiting on a lock is race prone. The "Table lock" state change happens before the thread actually tries to grab a lock on a table. The solution is to move the "Table lock" state so that its set only when a thread is actually going to wait for a lock. The state change happens after the thread fails to grab the lock (because it is owned by other thread) and proceeds to wait on a condition. This is considered part of work related to WL#4284 "Transactional DDL locking" Warning: this patch contains an incompatible change. When waiting on a lock in thr_lock.c, the server used to display "Locked" processlist state. After this patch, the state is "Table lock". The new state was actually intended to be display since year 2002, when Monty added it. But up until removal of thd->locked boolean member, this state was ignored by SHOW PROCESSLIST code. mysql-test/r/lock_multi.result: A style fix. mysql-test/r/sp-threads.result: Changed output of SHOW PROCESSLIST (new wait state). mysql-test/t/lock_multi.test: Use a more accurate state description when waiting inside thr_lock.c. mysql-test/t/lock_sync.test: Use a more accurate state description when waiting inside thr_lock.c. mysql-test/t/multi_update.test: Use a more accurate state description when waiting inside thr_lock.c. mysql-test/t/query_cache_28249.test: Use a more accurate state description when waiting inside thr_lock.c. mysql-test/t/sp_notembedded.test: Use a more accurate state description when waiting inside thr_lock.c. mysql-test/t/status.test: Use a more accurate state description when waiting inside thr_lock.c. mysys/thr_lock.c: Update thread state while waiting for a table lock. sql/lock.cc: State change was moved inside thr_lock.c.
2009-12-03 21:08:27 +01:00
where state= 'Table lock' and
Implement new type-of-operation-aware metadata locks. Add a wait-for graph based deadlock detector to the MDL subsystem. Fixes bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock" and bug #37346 "innodb does not detect deadlock between update and alter table". The first bug manifested itself as an unwarranted abort of a transaction with ER_LOCK_DEADLOCK error by a concurrent ALTER statement, when this transaction tried to repeat use of a table, which it has already used in a similar fashion before ALTER started. The second bug showed up as a deadlock between table-level locks and InnoDB row locks, which was "detected" only after innodb_lock_wait_timeout timeout. A transaction would start using the table and modify a few rows. Then ALTER TABLE would come in, and start copying rows into a temporary table. Eventually it would stumble on the modified records and get blocked on a row lock. The first transaction would try to do more updates, and get blocked on thr_lock.c lock. This situation of circular wait would only get resolved by a timeout. Both these bugs stemmed from inadequate solutions to the problem of deadlocks occurring between different locking subsystems. In the first case we tried to avoid deadlocks between metadata locking and table-level locking subsystems, when upgrading shared metadata lock to exclusive one. Transactions holding the shared lock on the table and waiting for some table-level lock used to be aborted too aggressively. We also allowed ALTER TABLE to start in presence of transactions that modify the subject table. ALTER TABLE acquires TL_WRITE_ALLOW_READ lock at start, and that block all writes against the table (naturally, we don't want any writes to be lost when switching the old and the new table). TL_WRITE_ALLOW_READ lock, in turn, would block the started transaction on thr_lock.c lock, should they do more updates. This, again, lead to the need to abort such transactions. The second bug occurred simply because we didn't have any mechanism to detect deadlocks between the table-level locks in thr_lock.c and row-level locks in InnoDB, other than innodb_lock_wait_timeout. This patch solves both these problems by moving lock conflicts which are causing these deadlocks into the metadata locking subsystem, thus making it possible to avoid or detect such deadlocks inside MDL. To do this we introduce new type-of-operation-aware metadata locks, which allow MDL subsystem to know not only the fact that transaction has used or is going to use some object but also what kind of operation it has carried out or going to carry out on the object. This, along with the addition of a special kind of upgradable metadata lock, allows ALTER TABLE to wait until all transactions which has updated the table to go away. This solves the second issue. Another special type of upgradable metadata lock is acquired by LOCK TABLE WRITE. This second lock type allows to solve the first issue, since abortion of table-level locks in event of DDL under LOCK TABLES becomes also unnecessary. Below follows the list of incompatible changes introduced by this patch: - From now on, ALTER TABLE and CREATE/DROP TRIGGER SQL (i.e. those statements that acquire TL_WRITE_ALLOW_READ lock) wait for all transactions which has *updated* the table to complete. - From now on, LOCK TABLES ... WRITE, REPAIR/OPTIMIZE TABLE (i.e. all statements which acquire TL_WRITE table-level lock) wait for all transaction which *updated or read* from the table to complete. As a consequence, innodb_table_locks=0 option no longer applies to LOCK TABLES ... WRITE. - DROP DATABASE, DROP TABLE, RENAME TABLE no longer abort statements or transactions which use tables being dropped or renamed, and instead wait for these transactions to complete. - Since LOCK TABLES WRITE now takes a special metadata lock, not compatible with with reads or writes against the subject table and transaction-wide, thr_lock.c deadlock avoidance algorithm that used to ensure absence of deadlocks between LOCK TABLES WRITE and other statements is no longer sufficient, even for MyISAM. The wait-for graph based deadlock detector of MDL subsystem may sometimes be necessary and is involved. This may lead to ER_LOCK_DEADLOCK error produced for multi-statement transactions even if these only use MyISAM: session 1: session 2: begin; update t1 ... lock table t2 write, t1 write; -- gets a lock on t2, blocks on t1 update t2 ... (ER_LOCK_DEADLOCK) - Finally, support of LOW_PRIORITY option for LOCK TABLES ... WRITE was abandoned. LOCK TABLE ... LOW_PRIORITY WRITE from now on has the same priority as the usual LOCK TABLE ... WRITE. SELECT HIGH PRIORITY no longer trumps LOCK TABLE ... WRITE in the wait queue. - We do not take upgradable metadata locks on implicitly locked tables. So if one has, say, a view v1 that uses table t1, and issues: LOCK TABLE v1 WRITE; FLUSH TABLE t1; -- (or just 'FLUSH TABLES'), an error is produced. In order to be able to perform DDL on a table under LOCK TABLES, the table must be locked explicitly in the LOCK TABLES list. mysql-test/include/handler.inc: Adjusted test case to trigger an execution path on which bug 41110 "crash with handler command when used concurrently with alter table" and bug 41112 "crash in mysql_ha_close_table/get_lock_data with alter table" were originally discovered. Left old test case which no longer triggers this execution path for the sake of coverage. Added test coverage for HANDLER SQL statements and type-aware metadata locks. Added a test for the global shared lock and HANDLER SQL. Updated tests to take into account that the old simple deadlock detection heuristics was replaced with a graph-based deadlock detector. mysql-test/r/debug_sync.result: Updated results (see debug_sync.test). mysql-test/r/handler_innodb.result: Updated results (see handler.inc test). mysql-test/r/handler_myisam.result: Updated results (see handler.inc test). mysql-test/r/innodb-lock.result: Updated results (see innodb-lock.test). mysql-test/r/innodb_mysql_lock.result: Updated results (see innodb_mysql_lock.test). mysql-test/r/lock.result: Updated results (see lock.test). mysql-test/r/lock_multi.result: Updated results (see lock_multi.test). mysql-test/r/lock_sync.result: Updated results (see lock_sync.test). mysql-test/r/mdl_sync.result: Updated results (see mdl_sync.test). mysql-test/r/sp-threads.result: SHOW PROCESSLIST output has changed due to the fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/r/truncate_coverage.result: Updated results (see truncate_coverage.test). mysql-test/suite/funcs_1/datadict/processlist_val.inc: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/funcs_1/r/processlist_val_no_prot.result: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/rpl/t/rpl_sp.test: Updated to a new SHOW PROCESSLIST state name. mysql-test/t/debug_sync.test: Use LOCK TABLES READ instead of LOCK TABLES WRITE as the latter no longer allows to trigger execution path involving waiting on thr_lock.c lock and therefore reaching debug sync-point covered by this test. mysql-test/t/innodb-lock.test: Adjusted test case to the fact that innodb_table_locks=0 option is no longer supported, since LOCK TABLES WRITE handles all its conflicts within MDL subsystem. mysql-test/t/innodb_mysql_lock.test: Added test for bug #37346 "innodb does not detect deadlock between update and alter table". mysql-test/t/lock.test: Added test coverage which checks the fact that we no longer support DDL under LOCK TABLES on tables which were locked implicitly. Adjusted existing test cases accordingly. mysql-test/t/lock_multi.test: Added test for bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock". Adjusted other test cases to take into account the fact that waiting for LOCK TABLES ... WRITE now happens within MDL subsystem. mysql-test/t/lock_sync.test: Since LOCK TABLES ... WRITE now takes SNRW metadata lock for tables locked explicitly we have to implicitly lock InnoDB tables (through view) to trigger the table-level lock conflict between TL_WRITE and TL_WRITE_ALLOW_WRITE. mysql-test/t/mdl_sync.test: Added basic test coverage for type-of-operation-aware metadata locks. Also covered with tests some use cases involving HANDLER statements in which a deadlock could arise. Adjusted existing tests to take type-of-operation-aware MDL into account. mysql-test/t/multi_update.test: Update to a new SHOW PROCESSLIST state name. mysql-test/t/truncate_coverage.test: Adjusted test case after making LOCK TABLES WRITE to wait until transactions that use the table to be locked are completed. Updated to the changed name of DEBUG_SYNC point. sql/handler.cc: Global read lock functionality has been moved into a class. sql/lock.cc: Global read lock functionality has been moved into a class. Updated code to use the new MDL API. sql/mdl.cc: Introduced new type-of-operation aware metadata locks. To do this: - Changed MDL_lock to use one list for waiting requests and one list for granted requests. For each list, added a bitmap that holds information what lock types a list contains. Added a helper class MDL_lock::List to manipulate with granted and waited lists while keeping the bitmaps in sync with list contents. - Changed lock-compatibility functions to use bitmaps that define compatibility. - Introduced a graph based deadlock detector inspired by waiting_threads.c from Maria implementation. - Now that we have a deadlock detector, and no longer have a global lock to protect individual lock objects, but rather use an rw lock per object, removed redundant code for upgrade, and the global read lock. Changed the MDL API to no longer require the caller to acquire the global intention exclusive lock by means of a separate method. Removed a few more methods that became redundant. - Removed deadlock detection heuristic, it has been made obsolete by the deadlock detector. - With operation-type-aware metadata locks, MDL subsystem has become aware of potential conflicts between DDL and open transactions. This made it possible to remove calls to mysql_abort_transactions_with_shared_lock() from acquisition paths for exclusive lock and lock upgrade. Now we can simply wait for these transactions to complete without fear of deadlock. Function mysql_lock_abort() has also become unnecessary for all conflicting cases except when a DDL conflicts with a connection that has an open HANDLER. sql/mdl.h: Introduced new type-of-operation aware metadata locks. Introduced a graph based deadlock detector and supporting methods. Added comments. God rid of redundant API calls. Renamed m_lt_or_ha_sentinel to m_trans_sentinel, since now it guards the global read lock as well as LOCK TABLES and HANDLER locks. sql/mysql_priv.h: Moved the global read lock functionality into a class. Added MYSQL_OPEN_FORCE_SHARED_MDL flag which forces open_tables() to take MDL_SHARED on tables instead of metadata locks specified in the parser. We use this to allow PREPARE run concurrently in presence of LOCK TABLES ... WRITE. Added signature for find_table_for_mdl_ugprade(). sql/set_var.cc: Global read lock functionality has been moved into a class. sql/sp_head.cc: When creating TABLE_LIST elements for prelocking or system tables set the type of request for metadata lock according to the operation that will be performed on the table. sql/sql_base.cc: - Updated code to use the new MDL API. - In order to avoid locks starvation we take upgradable locks all at once. As result implicitly locked tables no longer get an upgradable lock. Consequently DDL and FLUSH TABLES for such tables is prohibited. find_write_locked_table() was replaced by find_table_for_mdl_upgrade() function. open_table() was adjusted to return TABLE instance with upgradable ticket when necessary. - We no longer wait for all locks on OT_WAIT back off action -- only on the lock that caused the wait conflict. Moreover, now we distinguish cases when we have to wait due to conflict in MDL and old version of table in TDC. - Upate mysql_notify_threads_having_share_locks() to only abort thr_lock.c waits of threads that have open HANDLERs, since lock conflicts with only these threads now can lead to deadlocks not detectable by the MDL deadlock detector. - Remove mysql_abort_transactions_with_shared_locks() which is no longer needed. sql/sql_class.cc: Global read lock functionality has been moved into a class. Re-arranged code in THD::cleanup() to simplify assert. sql/sql_class.h: Introduced class to incapsulate global read lock functionality. Now sentinel in MDL subsystem guards the global read lock as well as LOCK TABLES and HANDLER locks. Adjusted code accordingly. sql/sql_db.cc: Global read lock functionality has been moved into a class. sql/sql_delete.cc: We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result TRUNCATE TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_handler.cc: Inform MDL_context about presence of open HANDLERs. Since HANLDERs break MDL protocol by acquiring table-level lock while holding only S metadata lock on a table MDL subsystem should take special care about such contexts (Now this is the only case when mysql_lock_abort() is used). sql/sql_parse.cc: Global read lock functionality has been moved into a class. Do not take upgradable metadata locks when opening tables for CREATE TABLE SELECT as it is not necessary and limits concurrency. When initializing TABLE_LIST objects before adding them to the table list set the type of request for metadata lock according to the operation that will be performed on the table. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result FLUSH TABLES is no longer allowed for such tables. sql/sql_prepare.cc: Use MYSQL_OPEN_FORCE_SHARED_MDL flag when opening tables during PREPARE. This allows PREPARE to run concurrently in presence of LOCK TABLES ... WRITE. sql/sql_rename.cc: Global read lock functionality has been moved into a class. sql/sql_show.cc: Updated code to use the new MDL API. sql/sql_table.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result DROP TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_trigger.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result CREATE/DROP TRIGGER is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_view.cc: Global read lock functionality has been moved into a class. Fixed results of wrong merge that led to misuse of GLR API. CREATE VIEW statement is not a commit statement. sql/table.cc: When resetting TABLE_LIST objects for PS or SP re-execution set the type of request for metadata lock according to the operation that will be performed on the table. Do the same in auxiliary function initializing metadata lock requests in a table list. sql/table.h: When initializing TABLE_LIST objects set the type of request for metadata lock according to the operation that will be performed on the table. sql/transaction.cc: Global read lock functionality has been moved into a class.
2010-02-01 12:43:06 +01:00
info='lock table v1 write';
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
--source include/wait_condition.inc
--echo # Allow SELECT ... FOR UPDATE to resume.
--echo # Since it already has TL_WRITE_ALLOW_WRITE lock on the first instance
--echo # of 't1' it should be able to get lock on the second instance without
--echo # waiting, even although there is another thread which has such lock
--echo # on this table and also there is a thread waiting for a TL_WRITE on it.
set debug_sync= 'now SIGNAL go';
--echo # Switch to connection 'con_bug45143_2'.
connection con_bug45143_2;
--echo # Reap SELECT ... FOR UPDATE
--reap
--echo # Switch to connection 'default'.
connection default;
--echo # Resume execution of the INSERT statement.
select release_lock("lock_bug45143_wait");
--echo # Switch to connection 'con_bug45143_1'.
connection con_bug45143_1;
--echo # Reap INSERT statement.
--echo # In Statement and Mixed replication mode we get here "Unsafe
--echo # for binlog" warnings. In row mode there are no warnings.
--echo # Hide the discrepancy.
--disable_warnings
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
--reap
--enable_warnings
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
--echo # Switch to connection 'con_bug45143_3'.
connection con_bug45143_3;
--echo # Reap LOCK TABLES statement.
--reap
unlock tables;
--echo # Switch to connection 'default'.
connection default;
--echo # Do clean-up.
disconnect con_bug45143_1;
disconnect con_bug45143_2;
disconnect con_bug45143_3;
set debug_sync= 'RESET';
set @@global.general_log= @old_general_log;
Implement new type-of-operation-aware metadata locks. Add a wait-for graph based deadlock detector to the MDL subsystem. Fixes bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock" and bug #37346 "innodb does not detect deadlock between update and alter table". The first bug manifested itself as an unwarranted abort of a transaction with ER_LOCK_DEADLOCK error by a concurrent ALTER statement, when this transaction tried to repeat use of a table, which it has already used in a similar fashion before ALTER started. The second bug showed up as a deadlock between table-level locks and InnoDB row locks, which was "detected" only after innodb_lock_wait_timeout timeout. A transaction would start using the table and modify a few rows. Then ALTER TABLE would come in, and start copying rows into a temporary table. Eventually it would stumble on the modified records and get blocked on a row lock. The first transaction would try to do more updates, and get blocked on thr_lock.c lock. This situation of circular wait would only get resolved by a timeout. Both these bugs stemmed from inadequate solutions to the problem of deadlocks occurring between different locking subsystems. In the first case we tried to avoid deadlocks between metadata locking and table-level locking subsystems, when upgrading shared metadata lock to exclusive one. Transactions holding the shared lock on the table and waiting for some table-level lock used to be aborted too aggressively. We also allowed ALTER TABLE to start in presence of transactions that modify the subject table. ALTER TABLE acquires TL_WRITE_ALLOW_READ lock at start, and that block all writes against the table (naturally, we don't want any writes to be lost when switching the old and the new table). TL_WRITE_ALLOW_READ lock, in turn, would block the started transaction on thr_lock.c lock, should they do more updates. This, again, lead to the need to abort such transactions. The second bug occurred simply because we didn't have any mechanism to detect deadlocks between the table-level locks in thr_lock.c and row-level locks in InnoDB, other than innodb_lock_wait_timeout. This patch solves both these problems by moving lock conflicts which are causing these deadlocks into the metadata locking subsystem, thus making it possible to avoid or detect such deadlocks inside MDL. To do this we introduce new type-of-operation-aware metadata locks, which allow MDL subsystem to know not only the fact that transaction has used or is going to use some object but also what kind of operation it has carried out or going to carry out on the object. This, along with the addition of a special kind of upgradable metadata lock, allows ALTER TABLE to wait until all transactions which has updated the table to go away. This solves the second issue. Another special type of upgradable metadata lock is acquired by LOCK TABLE WRITE. This second lock type allows to solve the first issue, since abortion of table-level locks in event of DDL under LOCK TABLES becomes also unnecessary. Below follows the list of incompatible changes introduced by this patch: - From now on, ALTER TABLE and CREATE/DROP TRIGGER SQL (i.e. those statements that acquire TL_WRITE_ALLOW_READ lock) wait for all transactions which has *updated* the table to complete. - From now on, LOCK TABLES ... WRITE, REPAIR/OPTIMIZE TABLE (i.e. all statements which acquire TL_WRITE table-level lock) wait for all transaction which *updated or read* from the table to complete. As a consequence, innodb_table_locks=0 option no longer applies to LOCK TABLES ... WRITE. - DROP DATABASE, DROP TABLE, RENAME TABLE no longer abort statements or transactions which use tables being dropped or renamed, and instead wait for these transactions to complete. - Since LOCK TABLES WRITE now takes a special metadata lock, not compatible with with reads or writes against the subject table and transaction-wide, thr_lock.c deadlock avoidance algorithm that used to ensure absence of deadlocks between LOCK TABLES WRITE and other statements is no longer sufficient, even for MyISAM. The wait-for graph based deadlock detector of MDL subsystem may sometimes be necessary and is involved. This may lead to ER_LOCK_DEADLOCK error produced for multi-statement transactions even if these only use MyISAM: session 1: session 2: begin; update t1 ... lock table t2 write, t1 write; -- gets a lock on t2, blocks on t1 update t2 ... (ER_LOCK_DEADLOCK) - Finally, support of LOW_PRIORITY option for LOCK TABLES ... WRITE was abandoned. LOCK TABLE ... LOW_PRIORITY WRITE from now on has the same priority as the usual LOCK TABLE ... WRITE. SELECT HIGH PRIORITY no longer trumps LOCK TABLE ... WRITE in the wait queue. - We do not take upgradable metadata locks on implicitly locked tables. So if one has, say, a view v1 that uses table t1, and issues: LOCK TABLE v1 WRITE; FLUSH TABLE t1; -- (or just 'FLUSH TABLES'), an error is produced. In order to be able to perform DDL on a table under LOCK TABLES, the table must be locked explicitly in the LOCK TABLES list. mysql-test/include/handler.inc: Adjusted test case to trigger an execution path on which bug 41110 "crash with handler command when used concurrently with alter table" and bug 41112 "crash in mysql_ha_close_table/get_lock_data with alter table" were originally discovered. Left old test case which no longer triggers this execution path for the sake of coverage. Added test coverage for HANDLER SQL statements and type-aware metadata locks. Added a test for the global shared lock and HANDLER SQL. Updated tests to take into account that the old simple deadlock detection heuristics was replaced with a graph-based deadlock detector. mysql-test/r/debug_sync.result: Updated results (see debug_sync.test). mysql-test/r/handler_innodb.result: Updated results (see handler.inc test). mysql-test/r/handler_myisam.result: Updated results (see handler.inc test). mysql-test/r/innodb-lock.result: Updated results (see innodb-lock.test). mysql-test/r/innodb_mysql_lock.result: Updated results (see innodb_mysql_lock.test). mysql-test/r/lock.result: Updated results (see lock.test). mysql-test/r/lock_multi.result: Updated results (see lock_multi.test). mysql-test/r/lock_sync.result: Updated results (see lock_sync.test). mysql-test/r/mdl_sync.result: Updated results (see mdl_sync.test). mysql-test/r/sp-threads.result: SHOW PROCESSLIST output has changed due to the fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/r/truncate_coverage.result: Updated results (see truncate_coverage.test). mysql-test/suite/funcs_1/datadict/processlist_val.inc: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/funcs_1/r/processlist_val_no_prot.result: SELECT FROM I_S.PROCESSLIST output has changed due to fact that waiting for LOCK TABLES WRITE now happens within metadata locking subsystem. mysql-test/suite/rpl/t/rpl_sp.test: Updated to a new SHOW PROCESSLIST state name. mysql-test/t/debug_sync.test: Use LOCK TABLES READ instead of LOCK TABLES WRITE as the latter no longer allows to trigger execution path involving waiting on thr_lock.c lock and therefore reaching debug sync-point covered by this test. mysql-test/t/innodb-lock.test: Adjusted test case to the fact that innodb_table_locks=0 option is no longer supported, since LOCK TABLES WRITE handles all its conflicts within MDL subsystem. mysql-test/t/innodb_mysql_lock.test: Added test for bug #37346 "innodb does not detect deadlock between update and alter table". mysql-test/t/lock.test: Added test coverage which checks the fact that we no longer support DDL under LOCK TABLES on tables which were locked implicitly. Adjusted existing test cases accordingly. mysql-test/t/lock_multi.test: Added test for bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock". Adjusted other test cases to take into account the fact that waiting for LOCK TABLES ... WRITE now happens within MDL subsystem. mysql-test/t/lock_sync.test: Since LOCK TABLES ... WRITE now takes SNRW metadata lock for tables locked explicitly we have to implicitly lock InnoDB tables (through view) to trigger the table-level lock conflict between TL_WRITE and TL_WRITE_ALLOW_WRITE. mysql-test/t/mdl_sync.test: Added basic test coverage for type-of-operation-aware metadata locks. Also covered with tests some use cases involving HANDLER statements in which a deadlock could arise. Adjusted existing tests to take type-of-operation-aware MDL into account. mysql-test/t/multi_update.test: Update to a new SHOW PROCESSLIST state name. mysql-test/t/truncate_coverage.test: Adjusted test case after making LOCK TABLES WRITE to wait until transactions that use the table to be locked are completed. Updated to the changed name of DEBUG_SYNC point. sql/handler.cc: Global read lock functionality has been moved into a class. sql/lock.cc: Global read lock functionality has been moved into a class. Updated code to use the new MDL API. sql/mdl.cc: Introduced new type-of-operation aware metadata locks. To do this: - Changed MDL_lock to use one list for waiting requests and one list for granted requests. For each list, added a bitmap that holds information what lock types a list contains. Added a helper class MDL_lock::List to manipulate with granted and waited lists while keeping the bitmaps in sync with list contents. - Changed lock-compatibility functions to use bitmaps that define compatibility. - Introduced a graph based deadlock detector inspired by waiting_threads.c from Maria implementation. - Now that we have a deadlock detector, and no longer have a global lock to protect individual lock objects, but rather use an rw lock per object, removed redundant code for upgrade, and the global read lock. Changed the MDL API to no longer require the caller to acquire the global intention exclusive lock by means of a separate method. Removed a few more methods that became redundant. - Removed deadlock detection heuristic, it has been made obsolete by the deadlock detector. - With operation-type-aware metadata locks, MDL subsystem has become aware of potential conflicts between DDL and open transactions. This made it possible to remove calls to mysql_abort_transactions_with_shared_lock() from acquisition paths for exclusive lock and lock upgrade. Now we can simply wait for these transactions to complete without fear of deadlock. Function mysql_lock_abort() has also become unnecessary for all conflicting cases except when a DDL conflicts with a connection that has an open HANDLER. sql/mdl.h: Introduced new type-of-operation aware metadata locks. Introduced a graph based deadlock detector and supporting methods. Added comments. God rid of redundant API calls. Renamed m_lt_or_ha_sentinel to m_trans_sentinel, since now it guards the global read lock as well as LOCK TABLES and HANDLER locks. sql/mysql_priv.h: Moved the global read lock functionality into a class. Added MYSQL_OPEN_FORCE_SHARED_MDL flag which forces open_tables() to take MDL_SHARED on tables instead of metadata locks specified in the parser. We use this to allow PREPARE run concurrently in presence of LOCK TABLES ... WRITE. Added signature for find_table_for_mdl_ugprade(). sql/set_var.cc: Global read lock functionality has been moved into a class. sql/sp_head.cc: When creating TABLE_LIST elements for prelocking or system tables set the type of request for metadata lock according to the operation that will be performed on the table. sql/sql_base.cc: - Updated code to use the new MDL API. - In order to avoid locks starvation we take upgradable locks all at once. As result implicitly locked tables no longer get an upgradable lock. Consequently DDL and FLUSH TABLES for such tables is prohibited. find_write_locked_table() was replaced by find_table_for_mdl_upgrade() function. open_table() was adjusted to return TABLE instance with upgradable ticket when necessary. - We no longer wait for all locks on OT_WAIT back off action -- only on the lock that caused the wait conflict. Moreover, now we distinguish cases when we have to wait due to conflict in MDL and old version of table in TDC. - Upate mysql_notify_threads_having_share_locks() to only abort thr_lock.c waits of threads that have open HANDLERs, since lock conflicts with only these threads now can lead to deadlocks not detectable by the MDL deadlock detector. - Remove mysql_abort_transactions_with_shared_locks() which is no longer needed. sql/sql_class.cc: Global read lock functionality has been moved into a class. Re-arranged code in THD::cleanup() to simplify assert. sql/sql_class.h: Introduced class to incapsulate global read lock functionality. Now sentinel in MDL subsystem guards the global read lock as well as LOCK TABLES and HANDLER locks. Adjusted code accordingly. sql/sql_db.cc: Global read lock functionality has been moved into a class. sql/sql_delete.cc: We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result TRUNCATE TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_handler.cc: Inform MDL_context about presence of open HANDLERs. Since HANLDERs break MDL protocol by acquiring table-level lock while holding only S metadata lock on a table MDL subsystem should take special care about such contexts (Now this is the only case when mysql_lock_abort() is used). sql/sql_parse.cc: Global read lock functionality has been moved into a class. Do not take upgradable metadata locks when opening tables for CREATE TABLE SELECT as it is not necessary and limits concurrency. When initializing TABLE_LIST objects before adding them to the table list set the type of request for metadata lock according to the operation that will be performed on the table. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result FLUSH TABLES is no longer allowed for such tables. sql/sql_prepare.cc: Use MYSQL_OPEN_FORCE_SHARED_MDL flag when opening tables during PREPARE. This allows PREPARE to run concurrently in presence of LOCK TABLES ... WRITE. sql/sql_rename.cc: Global read lock functionality has been moved into a class. sql/sql_show.cc: Updated code to use the new MDL API. sql/sql_table.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result DROP TABLE is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_trigger.cc: Global read lock functionality has been moved into a class. We no longer acquire upgradable metadata locks on tables which are locked by LOCK TABLES implicitly. As result CREATE/DROP TRIGGER is no longer allowed for such tables. Updated code to use the new MDL API. sql/sql_view.cc: Global read lock functionality has been moved into a class. Fixed results of wrong merge that led to misuse of GLR API. CREATE VIEW statement is not a commit statement. sql/table.cc: When resetting TABLE_LIST objects for PS or SP re-execution set the type of request for metadata lock according to the operation that will be performed on the table. Do the same in auxiliary function initializing metadata lock requests in a table list. sql/table.h: When initializing TABLE_LIST objects set the type of request for metadata lock according to the operation that will be performed on the table. sql/transaction.cc: Global read lock functionality has been moved into a class.
2010-02-01 12:43:06 +01:00
drop view v1;
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
drop table t1;
--echo #
--echo # Bug#50821 Deadlock between LOCK TABLES and ALTER TABLE
--echo #
--disable_warnings
DROP TABLE IF EXISTS t1, t2;
--enable_warnings
CREATE TABLE t1(id INT);
CREATE TABLE t2(id INT);
--echo # Connection con2
connect (con2, localhost, root);
START TRANSACTION;
SELECT * FROM t1;
--echo # Connection default
connection default;
--echo # Sending:
--send ALTER TABLE t1 ADD COLUMN j INT
--echo # Connection con2
connection con2;
let $wait_condition=
SELECT COUNT(*) = 1 FROM information_schema.processlist
WHERE state = "Waiting for table"
AND info = "ALTER TABLE t1 ADD COLUMN j INT";
--source include/wait_condition.inc
--echo # This used to cause a deadlock.
INSERT INTO t2 SELECT * FROM t1;
COMMIT;
--echo # Connection default
connection default;
--echo # Reaping ALTER TABLE t1 ADD COLUMN j INT
--reap
DROP TABLE t1, t2;
disconnect con2;
--echo #
--echo # Bug#51391 Deadlock involving events during rqg_info_schema test
--echo #
CREATE EVENT e1 ON SCHEDULE EVERY 5 HOUR DO SELECT 1;
CREATE EVENT e2 ON SCHEDULE EVERY 5 HOUR DO SELECT 2;
--echo # Connection con1
connect(con1, localhost, root);
SET DEBUG_SYNC="before_lock_tables_takes_lock SIGNAL drop WAIT_FOR query";
--echo # Sending:
--send DROP EVENT e1;
--echo # Connection default
connection default;
SET DEBUG_SYNC="now WAIT_FOR drop";
SELECT name FROM mysql.event, INFORMATION_SCHEMA.GLOBAL_VARIABLES
WHERE definer = VARIABLE_VALUE;
SET DEBUG_SYNC="now SIGNAL query";
--echo # Connection con1
connection con1;
--echo # Reaping: DROP EVENT t1
--reap
disconnect con1;
--source include/wait_until_disconnected.inc
--echo # Connection default
connection default;
DROP EVENT e2;
SET DEBUG_SYNC="RESET";
Fix for bug #45143 "All connections hang on concurrent ALTER TABLE". Concurrent execution of statements which require non-table-level write locks on several instances of the same table (such as SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML statement which invokes trigger which tries to update same InnoDB table directly and through stored function) and statements which required table-level locks on this table (e.g. LOCK TABLE ... WRITE, ALTER TABLE, ...) might have resulted in a deadlock. The problem occured when a thread tried to acquire write lock (TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table and we failed to detect that this thread already had another instance of write lock on it (so in fact we were trying to acquire recursive lock) because there was also another thread holding write lock on the table (also TL_WRITE_ALLOW_WRITE). When the latter thread released its lock neither the first thread nor the thread trying to acquire TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write locked by the first thread) so we ended up with a deadlock. This patch solves this problem by ensuring that thread which already has write lock on the table won't wait when it tries to acquire second write lock on the same table. mysql-test/r/lock_sync.result: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysql-test/t/lock_sync.test: Added test case for bug #45143 "All connections hang on concurrent ALTER TABLE". mysys/thr_lock.c: Ensured that thread can acquire write lock on the table without waiting if it already has write lock on it even if there are other threads holding write locks on this table (this is normal situation for, e.g., TL_WRITE_ALLOW_WRITE type of lock). Adjusted comments to better explain why it is OK to do so and added asserts to prevent introduction of scenarios in which this can cause problems.
2009-10-26 20:38:03 +01:00
# Check that all connections opened by test cases in this file are really
# gone so execution of other tests won't be affected by their presence.
--source include/wait_until_count_sessions.inc