You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

739 lines
19 KiB

Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Part of fix for bug#52044 "FLUSH TABLES WITH READ LOCK and FLUSH TABLES <list> WITH READ LOCK are incompatible" to be pushed as separate patch. Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. Also replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it more consistent with other thread state names. Updated test cases and their results according to these changes. Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. mysql-test/r/query_cache.result: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysql-test/suite/sys_vars/r/query_cache_wlock_invalidate_func.result: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/suite/sys_vars/t/query_cache_wlock_invalidate_func.test: Fixed sys_vars.query_cache_wlock_invalidate_func test to not to wait for timeout of wait_condition.inc script. Reverted changes to test which introduced timeout and replaced waiting condition with a more appropriate one. Test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables was added to query_cache.test. mysql-test/t/query_cache.test: Added test coverage for query_cache_wlock_invalidate behavior for implicitly locked tables. mysys/thr_lock.c: Replaced "Table lock" thread state name, which was used by threads waiting on thr_lock.c table level lock, with more elaborate "Waiting for table level lock", to make it consistent with thread state names which are used while waiting for metadata locks and table flush. sql/mdl.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for a metadata lock or table flush, with a set of names which better reflect types of resources being waited for. To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/mdl.h: To implement this: - Adjusted MDL_wait::timed_wait() to take thread state name as parameter. - Introduced method of MDL_key class which allows to get thread state name to be used while waiting for resource corresponding to the key and changed code to use it. Added array translating namespaces to thread state names as part of this change. sql/sql_base.cc: Replaced thread state name "Waiting for table", which was used by threads waiting for table flush, with a more elaborate "Waiting for table flush".
15 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Implementation of simple deadlock detection for metadata locks. This change is supposed to reduce number of ER_LOCK_DEADLOCK errors which occur when multi-statement transaction encounters conflicting metadata lock in cases when waiting is possible. The idea is not to fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting lock go away. To avoid deadlocks we use simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. This patch also fixes bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". The bug was that concurrent execution of UPDATE or MULTI-UPDATE statement as a part of multi-statement transaction that already has used table being updated and ALTER TABLE statement might have resulted of loss of isolation between this transaction and ALTER TABLE statement, which manifested itself as changes performed by ALTER TABLE becoming visible in transaction and wrong binary log order as a consequence. This problem occurred when UPDATE or MULTI-UPDATE's wait in mysql_lock_tables() call was aborted due to metadata lock upgrade performed by concurrent ALTER TABLE. After such abort all metadata locks held by transaction were released but transaction silently continued to be executed as if nothing has happened. We solve this problem by changing our code not to release all locks in such case. Instead we release only locks which were acquired by current statement and then try to reacquire them by restarting open/lock tables process. We piggyback on simple deadlock detector implementation since this change has to be done anyway for it. mysql-test/include/handler.inc: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_innodb.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/handler_myisam.result: After introduction of basic deadlock detector for metadata locks it became necessary to change parts of test for HANDLER statements which covered some of scenarios in which ER_LOCK_DEADLOCK error was detected in absence of real deadlock (with new deadlock detector this no longer happens). Also adjusted test to the fact that HANDLER READ for the table no longer will be blocked by ALTER TABLE for the same table which awaits for metadata lock upgrade (this is due to removal of mysql_lock_abort() from wait_while_table_is_used()). mysql-test/r/mdl_sync.result: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/r/sp-lock.result: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. mysql-test/t/mdl_sync.test: Added test coverage for basic deadlock detection in metadata locking subsystem and for bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER". mysql-test/t/sp-lock.test: Adjusted test coverage for metadata locking for stored routines since after introduction of basic deadlock detector for metadata locks number of scenarios in which ER_LOCK_DEADLOCK error in absence of deadlock has decreased. sql/log_event_old.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/mdl.cc: Changed metadata locking subsystem to support basic deadlock detection with a help of the following simple empiric -- we assume that there is a deadlock if there is a connection which has to wait for a metadata lock which is currently acquired by some connection which is itself waiting to be able to acquire some shared metadata lock. To implement this change: - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by the connection and therefore deadlocks are possible if this connection is going to wait for some metadata lock. To do this added version of MDL_ticket::has_pending_conflicting_lock() method which assumes that its caller already owns LOCK_mdl mutex. - Changed MDL_context::wait_for_locks() to use one of the above methods to check if somebody is waiting for metadata lock owned by this context (and therefore deadlock is possible) and emit ER_LOCK_DEADLOCK error in this case. Also now we mark context of connections waiting inside of this method by setting MDL_context::m_is_waiting_in_mdl member. Thanks to this such connection could be waken up if some other connection starts waiting for one of its metadata locks and so a deadlock can occur. - Adjusted notify_shared_lock() to wake up connections which wait inside MDL_context::wait_for_locks() while holding shared metadata lock. - Changed MDL_ticket::upgrade_shared_lock_to_exclusive() to add temporary ticket for exclusive lock to MDL_lock::waiting queue, so request for metadata lock upgrade can be properly detected by our empiric. Also now this method invokes a callback which forces transactions holding shared metadata lock on the table to call MDL_context:: can_wait_lead_to_deadlock() method even if they don't need any new metadata locks. Thanks to this such transactions can detect deadlocks/ livelocks between MDL and table-level locks. Also reduced timeouts between calls to notify_shared_lock() in MDL_ticket::upgrade_shared_lock_to_exclusive() and MDL_context::acquire_exclusive_locks(). This was necessary to get rid of call to mysql_lock_abort() in wait_while_table_is_used(). (Now we instead rely on notify_shared_lock() timely calling mysql_lock_abort_for_thread() for the table on which lock is being upgraded/acquired). sql/mdl.h: - Added a version of MDL_ticket::has_pending_conflicting_lock() method to be used in situations when caller already has acquired LOCK_mdl mutex. - Added MDL_context::can_wait_lead_to_deadlock()/_impl() methods which allow to find out if there is someone waiting for metadata lock which is held by this connection and thus deadlocks are possible if this connections will start waiting for some metadata lock. - Added MDL_context::m_is_waiting_in_mdl member to mark connections waiting in MDL_context::wait_for_locks() method of metadata locking subsystem. Added getter method for this private member to make it accessible in notify_shared_lock() auxiliary so we can wake-up such connections if they hold shared metadata locks. - Finally, added mysql_abort_transactions_with_shared_lock() callback to be able force transactions which don't need any new metadata locks still call MDL_context::can_wait_lead_to_deadlock() and detect some of deadlocks between metadata locks and table-level locks. sql/mysql_priv.h: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_base.cc: Changed approach to metadata locking for multi-statement transactions. We no longer fail ER_LOCK_DEADLOCK error immediately when we encounter conflicting metadata lock. Instead we release all metadata locks acquired by current statement and start to wait until conflicting locks to go away by calling MDL_context::wait_for_locks() method. To avoid deadlocks the latter implements simple empiric which aborts waiting with ER_LOCK_DEADLOCK error if it turns out that somebody is waiting for metadata locks owned by this transaction. To implement the change described above: - Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. - Changed Open_table_context::request_backoff_action() not to fail with ER_LOCK_DEADLOCK immediately if back-off is requested due to conflicting metadata lock. - Added new argument for close_tables_for_reopen() procedure which allows to specify subset of metadata locks to be released. - Changed open_tables() not to release all metadata locks acquired by current transaction when metadata lock conflict is discovered. Instead we release only locks acquired by current statement. - Changed open_ltable() and open_and_lock_tables_derived() not to emit ER_LOCK_DEADLOCK error when mysql_lock_tables() is aborted in multi-statement transaction when somebody tries to acquire exclusive metadata lock on the table. Instead we release metadata locks acquired by current statement and try to wait until they can be re-acquired. - Adjusted tdc_wait_for_old_versions() to check if there is someone waiting for one of metadata locks held by this connection and run deadlock detection in order to avoid deadlocks in some situations. - Added mysql_abort_transactions_with_shared_lock() callback which allows to force transactions holding shared metadata lock on the table to call MDL_context::can_wait_lead_to_deadlock() even if they don't need any new metadata locks so they can detect potential deadlocks between metadata locking subsystem and table-level locks. - Adjusted wait_while_table_is_used() not to set TABLE::version to 0 as it is now done only when necessary by the above-mentioned callback. Also removed unnecessary call to mysql_lock_abort(). Instead we rely on code performing metadata lock upgrade aborting waits on the table-level lock for this table by calling mysql_lock_abort_for_thread() (invoked by mysql_notify_thread_having_shared_lock()). In future this should allow to reduce number of scenarios in which we produce ER_LOCK_DEADLOCK error even though no real deadlock exists. sql/sql_class.h: Introduced Open_table_context::m_start_of_statement_svp member to store state of metadata locks at the start of the statement. Replaced Open_table_context::m_can_deadlock member with m_has_locks member to reflect the fact that we no longer unconditionally emit ER_LOCK_DEADLOCK error for transaction having some metadata locks when conflicting metadata lock is discovered. sql/sql_insert.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_plist.h: Made I_P_List_iterator<T, B> usable with const lists. sql/sql_show.cc: close_tables_for_reopen() now takes one more argument which specifies at which point it should stop releasing metadata locks acquired by this connection. sql/sql_update.cc: Changed UPDATE and MULTI-UPDATE code not to release all metadata locks when calls to mysql_lock_tables() are aborted. Instead we release only locks which are acquired by this statement and then try to reacquire them by calling open_tables(). This solves bug #46273 "MySQL 5.4.4 new MDL: Bug#989 is not fully fixed in case of ALTER".
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
Apply and review: 3655 Jon Olav Hauglid 2009-10-19 Bug #30977 Concurrent statement using stored function and DROP FUNCTION breaks SBR Bug #48246 assert in close_thread_table Implement a fix for: Bug #41804 purge stored procedure cache causes mysterious hang for many minutes Bug #49972 Crash in prepared statements The problem was that concurrent execution of DML statements that use stored functions and DDL statements that drop/modify the same function might result in incorrect binary log in statement (and mixed) mode and therefore break replication. This patch fixes the problem by introducing metadata locking for stored procedures and functions. This is similar to what is done in Bug#25144 for views. Procedures and functions now are locked using metadata locks until the transaction is either committed or rolled back. This prevents other statements from modifying the procedure/function while it is being executed. This provides commit ordering - guaranteeing serializability across multiple transactions and thus fixes the reported binlog problem. Note that we do not take locks for top-level CALLs. This means that procedures called directly are not protected from changes by simultaneous DDL operations so they are executed at the state they had at the time of the CALL. By not taking locks for top-level CALLs, we still allow transactions to be started inside procedures. This patch also changes stored procedure cache invalidation. Upon a change of cache version, we no longer invalidate the entire cache, but only those routines which we use, only when a statement is executed that uses them. This patch also changes the logic of prepared statement validation. A stored procedure used by a prepared statement is now validated only once a metadata lock has been acquired. A version mismatch causes a flush of the obsolete routine from the cache and statement reprepare. Incompatible changes: 1) ER_LOCK_DEADLOCK is reported for a transaction trying to access a procedure/function that is locked by a DDL operation in another connection. 2) Procedure/function DDL operations are now prohibited in LOCK TABLES mode as exclusive locks must be taken all at once and LOCK TABLES provides no way to specifiy procedures/functions to be locked. Test cases have been added to sp-lock.test and rpl_sp.test. Work on this bug has very much been a team effort and this patch includes and is based on contributions from Davi Arnaut, Dmitry Lenev, Magne Mæhre and Konstantin Osipov. mysql-test/r/ps_ddl.result: Update results (Bug#30977). mysql-test/r/ps_ddl1.result: Update results (Bug#30977). mysql-test/r/sp-error.result: Update results (Bug#30977). mysql-test/r/sp-lock.result: Update results (Bug#30977). mysql-test/suite/rpl/r/rpl_sp.result: Update results (Bug#30977). mysql-test/suite/rpl/t/rpl_sp.test: Add a test case for Bug#30977. mysql-test/t/ps_ddl.test: Update comments. We no longer re-prepare a prepared statement when a stored procedure used in top-level CALL is changed. mysql-test/t/ps_ddl1.test: Modifying stored procedure p1 no longer invalidates prepared statement "call p1" -- we can re-use the prepared statement without invalidation. mysql-test/t/sp-error.test: Use a constant for an error value. mysql-test/t/sp-lock.test: Add test coverage for Bug#30977. sql/lock.cc: Implement lock_routine_name() - a way to acquire an exclusive metadata lock (ex- name-lock) on stored procedure/function. sql/sp.cc: Change semantics of sp_cache_routine() -- now it has an option to make sure that the routine that is cached is up to date (has the latest sp cache version). Add sp_cache_invalidate() to sp_drop_routine(), where it was missing (a bug!). Acquire metadata locks for SP DDL (ALTER/CREATE/DROP). This is the core of the fix for Bug#30977. Since caching and cache invalidation scheme was changed, make sure we don't invalidate the SP cache in the middle of a stored routine execution. At the same time, make sure we don't access stale data due to lack of invalidation. For that, change ALTER FUNCTION/PROCEDURE to not use the cache, and SHOW PROCEDURE CODE/SHOW CREATE PROCEDURE/FUNCTION to always read an up to date version of the routine from the cache. sql/sp.h: Add a helper wrapper around sp_cache_routine(). sql/sp_cache.cc: Implement new sp_cache_version() and sp_cache_flush_obsolete(). Now we flush stale routines individually, rather than all at once. sql/sp_cache.h: Update signatures of sp_cache_version() and sp_cache_flush_obsolete(). sql/sp_head.cc: Add a default initialization of sp_head::m_sp_cache_version. Remove a redundant sp_head::create(). sql/sp_head.h: Add m_sp_cache_version to sp_head class - we now keep track of every routine in the stored procedure cache, rather than of the entire cache. sql/sql_base.cc: Implement prelocking for stored routines. Validate stored routines after they were locked. Flush obsolete routines upon next access, one by one, not all at once (Bug#41804). Style fixes. sql/sql_class.h: Rename a Open_table_context method. sql/sql_parse.cc: Make sure stored procedures DDL commits the active transaction (issues an implicit commit before and after). Remove sp_head::create(), a pure redundancy. Move the semantical check during alter routine inside sp_update_routine() code in order to: - avoid using SP cache during update, it may be obsolete. - speed up and simplify the update procedure. Remove sp_cache_flush_obsolete() calls, we no longer flush the entire cache, ever, stale routines are flushed before next use, one at a time. sql/sql_prepare.cc: Move routine metadata validation to open_and_process_routine(). Fix Bug#49972 (don't swap flags at reprepare). Reset Sroutine_hash_entries in reinit_stmt_before_use(). Remove SP cache invalidation, it's now done by open_tables(). sql/sql_show.cc: Fix a warning: remove an unused label. sql/sql_table.cc: Reset mdl_request.ticket for tickets acquired for routines inlined through a view, in CHECK TABLE statement, to satisfy an MDL assert. sql/sql_update.cc: Move the cleanup of "translation items" to close_tables_for_reopen(), since it's needed in all cases when we back off, not just the back-off in multi-update. This fixes a bug when the server would crash on attempt to back off when opening tables for a statement that uses information_schema tables.
16 years ago
  1. #
  2. # Test coverage for changes performed by the fix
  3. # for Bug#30977 "Concurrent statement using stored function
  4. # and DROP FUNCTION breaks SBR.
  5. #
  6. #
  7. # 1) Verify that the preceding transaction is
  8. # (implicitly) committed before CREATE/ALTER/DROP
  9. # PROCEDURE. Note, that this is already tested
  10. # in implicit_commit.test, but here we use an alternative
  11. # approach.
  12. #
  13. # Start a transaction, create a savepoint,
  14. # then call a DDL operation on a procedure, and then check
  15. # that the savepoint is no longer present.
  16. drop table if exists t1;
  17. drop procedure if exists p1;
  18. drop procedure if exists p2;
  19. drop procedure if exists p3;
  20. drop procedure if exists p4;
  21. drop function if exists f1;
  22. create table t1 (a int);
  23. #
  24. # Test 'CREATE PROCEDURE'.
  25. #
  26. begin;
  27. savepoint sv;
  28. create procedure p1() begin end;
  29. rollback to savepoint sv;
  30. ERROR 42000: SAVEPOINT sv does not exist
  31. #
  32. # Test 'ALTER PROCEDURE'.
  33. #
  34. begin;
  35. savepoint sv;
  36. alter procedure p1 comment 'changed comment';
  37. rollback to savepoint sv;
  38. ERROR 42000: SAVEPOINT sv does not exist
  39. #
  40. # Test 'DROP PROCEDURE'.
  41. #
  42. begin;
  43. savepoint sv;
  44. drop procedure p1;
  45. rollback to savepoint sv;
  46. ERROR 42000: SAVEPOINT sv does not exist
  47. #
  48. # Test 'CREATE FUNCTION'.
  49. #
  50. begin;
  51. savepoint sv;
  52. create function f1() returns int return 1;
  53. rollback to savepoint sv;
  54. ERROR 42000: SAVEPOINT sv does not exist
  55. #
  56. # Test 'ALTER FUNCTION'.
  57. #
  58. begin;
  59. savepoint sv;
  60. alter function f1 comment 'new comment';
  61. rollback to savepoint sv;
  62. ERROR 42000: SAVEPOINT sv does not exist
  63. #
  64. # Test 'DROP FUNCTION'.
  65. #
  66. begin;
  67. savepoint sv;
  68. drop function f1;
  69. rollback to savepoint sv;
  70. ERROR 42000: SAVEPOINT sv does not exist
  71. #
  72. # 2) Verify that procedure DDL operations fail
  73. # under lock tables.
  74. #
  75. # Auxiliary routines to test ALTER.
  76. create procedure p1() begin end;
  77. create function f1() returns int return 1;
  78. lock table t1 write;
  79. create procedure p2() begin end;
  80. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  81. alter procedure p1 comment 'changed comment';
  82. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  83. drop procedure p1;
  84. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  85. create function f2() returns int return 1;
  86. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  87. alter function f1 comment 'changed comment';
  88. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  89. lock table t1 read;
  90. create procedure p2() begin end;
  91. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  92. alter procedure p1 comment 'changed comment';
  93. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  94. drop procedure p1;
  95. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  96. create function f2() returns int return 1;
  97. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  98. alter function f1 comment 'changed comment';
  99. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  100. unlock tables;
  101. #
  102. # Even if we locked a temporary table.
  103. # Todo: this is a restriction we could possibly lift.
  104. #
  105. drop table t1;
  106. create temporary table t1 (a int);
  107. lock table t1 read;
  108. create procedure p2() begin end;
  109. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  110. alter procedure p1 comment 'changed comment';
  111. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  112. drop procedure p1;
  113. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  114. create function f2() returns int return 1;
  115. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  116. alter function f1 comment 'changed comment';
  117. ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
  118. unlock tables;
  119. drop function f1;
  120. drop procedure p1;
  121. drop temporary table t1;
  122. #
  123. # 3) Verify that CREATE/ALTER/DROP routine grab an
  124. # exclusive lock.
  125. #
  126. # For that, start a transaction, use a routine. In a concurrent
  127. # connection, try to drop or alter the routine. It should place
  128. # a pending or exclusive lock and block. In another concurrnet
  129. # connection, try to use the routine.
  130. # That should block on the pending exclusive lock.
  131. #
  132. # Establish helper connections.
  133. #
  134. # Test DROP PROCEDURE.
  135. #
  136. # --> connection default
  137. create procedure p1() begin end;
  138. create function f1() returns int
  139. begin
  140. call p1();
  141. return 1;
  142. end|
  143. begin;
  144. select f1();
  145. f1()
  146. 1
  147. # --> connection con1
  148. # Sending 'drop procedure p1'...
  149. drop procedure p1;
  150. # --> connection con2
  151. # Waiting for 'drop procedure t1' to get blocked on MDL lock...
  152. # Demonstrate that there is a pending exclusive lock.
  153. # Sending 'select f1()'...
  154. select f1();
  155. # --> connection con3
  156. # Waiting for 'select f1()' to get blocked by a pending MDL lock...
  157. # --> connection default
  158. commit;
  159. # --> connection con1
  160. # Reaping 'drop procedure p1'...
  161. # --> connection con2
  162. # Reaping 'select f1()'
  163. ERROR 42000: PROCEDURE test.p1 does not exist
  164. # --> connection default
  165. #
  166. # Test CREATE PROCEDURE.
  167. #
  168. create procedure p1() begin end;
  169. begin;
  170. select f1();
  171. f1()
  172. 1
  173. # --> connection con1
  174. # Sending 'create procedure p1'...
  175. create procedure p1() begin end;
  176. # --> connection con2
  177. # Waiting for 'create procedure t1' to get blocked on MDL lock...
  178. # Demonstrate that there is a pending exclusive lock.
  179. # Sending 'select f1()'...
  180. select f1();
  181. # --> connection con3
  182. # Waiting for 'select f1()' to get blocked by a pending MDL lock...
  183. # --> connection default
  184. commit;
  185. # --> connection con1
  186. # Reaping 'create procedure p1'...
  187. ERROR 42000: PROCEDURE p1 already exists
  188. # --> connection con2
  189. # Reaping 'select f1()'
  190. f1()
  191. 1
  192. #
  193. # Test ALTER PROCEDURE.
  194. #
  195. begin;
  196. select f1();
  197. f1()
  198. 1
  199. # --> connection con1
  200. # Sending 'alter procedure p1'...
  201. alter procedure p1 contains sql;
  202. # --> connection con2
  203. # Waiting for 'alter procedure t1' to get blocked on MDL lock...
  204. # Demonstrate that there is a pending exclusive lock.
  205. # Sending 'select f1()'...
  206. select f1();
  207. # --> connection con3
  208. # Waiting for 'select f1()' to get blocked by a pending MDL lock...
  209. # --> connection default
  210. commit;
  211. # --> connection con1
  212. # Reaping 'alter procedure p1'...
  213. # --> connection con2
  214. # Reaping 'select f1()'
  215. f1()
  216. 1
  217. # --> connection default
  218. #
  219. # Test DROP FUNCTION.
  220. #
  221. begin;
  222. select f1();
  223. f1()
  224. 1
  225. # --> connection con1
  226. # Sending 'drop function f1'...
  227. drop function f1;
  228. # --> connection con2
  229. # Waiting for 'drop function f1' to get blocked on MDL lock...
  230. # Demonstrate that there is a pending exclusive lock.
  231. # Sending 'select f1()'...
  232. select f1();
  233. # --> connection con3
  234. # Waiting for 'select f1()' to get blocked by a pending MDL lock...
  235. # --> connection default
  236. commit;
  237. # --> connection con1
  238. # Reaping 'drop function f1'...
  239. # --> connection con2
  240. # Reaping 'select f1()'
  241. ERROR 42000: FUNCTION test.f1 does not exist
  242. # --> connection default
  243. #
  244. # Test CREATE FUNCTION.
  245. #
  246. create function f1() returns int return 1;
  247. begin;
  248. select f1();
  249. f1()
  250. 1
  251. # --> connection con1
  252. # Sending 'create function f1'...
  253. create function f1() returns int return 2;
  254. # --> connection con2
  255. # Waiting for 'create function f1' to get blocked on MDL lock...
  256. # Demonstrate that there is a pending exclusive lock.
  257. # Sending 'select f1()'...
  258. select f1();
  259. # --> connection con3
  260. # Waiting for 'select f1()' to get blocked by a pending MDL lock...
  261. # --> connection default
  262. commit;
  263. # --> connection con1
  264. # Reaping 'create function f1'...
  265. ERROR 42000: FUNCTION f1 already exists
  266. # --> connection con2
  267. # Reaping 'select f1()'
  268. f1()
  269. 1
  270. # --> connection default
  271. #
  272. # Test ALTER FUNCTION.
  273. #
  274. begin;
  275. select f1();
  276. f1()
  277. 1
  278. # --> connection con1
  279. # Sending 'alter function f1'...
  280. alter function f1 contains sql;
  281. # --> connection con2
  282. # Waiting for 'alter function f1' to get blocked on MDL lock...
  283. # Demonstrate that there is a pending exclusive lock.
  284. # Sending 'select f1()'...
  285. select f1();
  286. # --> connection con3
  287. # Waiting for 'select f1()' to get blocked by a pending MDL lock...
  288. # --> connection default
  289. commit;
  290. # --> connection con1
  291. # Reaping 'alter function f1'...
  292. # --> connection con2
  293. # Reaping 'select f1()'
  294. f1()
  295. 1
  296. # --> connection default
  297. drop function f1;
  298. drop procedure p1;
  299. #
  300. # 4) MDL lock should not be taken for
  301. # unrolled CALL statements.
  302. # The primary goal of metadata locks is a consistent binary log.
  303. # When a call statement is unrolled, it doesn't get to the
  304. # binary log, instead the statements that are contained
  305. # in the procedure body do. This can nest to any level.
  306. #
  307. create procedure p1() begin end;
  308. create procedure p2() begin end;
  309. create table t1 (a int);
  310. create procedure p3()
  311. begin
  312. call p1();
  313. call p1();
  314. call p2();
  315. end|
  316. create procedure p4()
  317. begin
  318. call p1();
  319. call p1();
  320. call p2();
  321. call p2();
  322. call p3();
  323. end|
  324. begin;
  325. select * from t1;
  326. a
  327. savepoint sv;
  328. call p4();
  329. # Prepared statement should not add any locks either.
  330. prepare stmt from "call p4()";
  331. execute stmt;
  332. execute stmt;
  333. # --> connection con1
  334. drop procedure p1;
  335. drop procedure p2;
  336. drop procedure p3;
  337. drop procedure p4;
  338. # --> connection default
  339. # This is to verify there was no implicit commit.
  340. rollback to savepoint sv;
  341. call p4();
  342. ERROR 42000: PROCEDURE test.p4 does not exist
  343. commit;
  344. drop table t1;
  345. #
  346. # 5) Locks should be taken on routines
  347. # used indirectly by views or triggers.
  348. #
  349. #
  350. # A function is used from a trigger.
  351. #
  352. create function f1() returns int return 1;
  353. create table t1 (a int);
  354. create table t2 (a int, b int);
  355. create trigger t1_ai after insert on t1 for each row
  356. insert into t2 (a, b) values (new.a, f1());
  357. begin;
  358. insert into t1 (a) values (1);
  359. # --> connection con1
  360. # Sending 'drop function f1'
  361. drop function f1;
  362. # --> connection con2
  363. # Waiting for 'drop function f1' to get blocked on MDL lock...
  364. # --> connnection default
  365. commit;
  366. # --> connection con1
  367. # Reaping 'drop function f1'...
  368. # --> connection default
  369. #
  370. # A function is used from a view.
  371. #
  372. create function f1() returns int return 1;
  373. create view v1 as select f1() as a;
  374. begin;
  375. select * from v1;
  376. a
  377. 1
  378. # --> connection con1
  379. # Sending 'drop function f1'
  380. drop function f1;
  381. # --> connection con2
  382. # Waiting for 'drop function f1' to get blocked on MDL lock...
  383. # --> connnection default
  384. commit;
  385. # --> connection con1
  386. # Reaping 'drop function f1'...
  387. # --> connection default
  388. #
  389. # A procedure is used from a function.
  390. #
  391. create function f1() returns int
  392. begin
  393. declare v_out int;
  394. call p1(v_out);
  395. return v_out;
  396. end|
  397. create procedure p1(out v_out int) set v_out=3;
  398. begin;
  399. select * from v1;
  400. a
  401. 3
  402. # --> connection con1
  403. # Sending 'drop procedure p1'
  404. drop procedure p1;
  405. # --> connection con2
  406. # Waiting for 'drop procedure p1' to get blocked on MDL lock...
  407. # --> connnection default
  408. commit;
  409. # --> connection con1
  410. # Reaping 'drop procedure p1'...
  411. # --> connection default
  412. #
  413. # Deep nesting: a function is used from a procedure used
  414. # from a function used from a view used in a trigger.
  415. #
  416. create function f2() returns int return 4;
  417. create procedure p1(out v_out int) set v_out=f2();
  418. drop trigger t1_ai;
  419. create trigger t1_ai after insert on t1 for each row
  420. insert into t2 (a, b) values (new.a, (select max(a) from v1));
  421. begin;
  422. insert into t1 (a) values (3);
  423. # --> connection con1
  424. # Sending 'drop function f2'
  425. drop function f2;
  426. # --> connection con2
  427. # Waiting for 'drop function f2' to get blocked on MDL lock...
  428. # --> connnection default
  429. commit;
  430. # --> connection con1
  431. # Reaping 'drop function f2'...
  432. # --> connection default
  433. drop view v1;
  434. drop function f1;
  435. drop procedure p1;
  436. drop table t1, t2;
  437. #
  438. # 6) Check that ER_LOCK_DEADLOCK is reported if
  439. # acquisition of a shared lock fails during a transaction or
  440. # we need to back off to flush the sp cache.
  441. #
  442. # Sic: now this situation does not require a back off since we
  443. # flush the cache on the fly.
  444. #
  445. create function f1() returns int return 7;
  446. create table t1 (a int);
  447. begin;
  448. select * from t1;
  449. a
  450. select f1();
  451. f1()
  452. 7
  453. commit;
  454. drop table t1;
  455. drop function f1;
  456. #
  457. # 7) Demonstrate that under LOCK TABLES we accumulate locks
  458. # on stored routines, and release metadata locks in
  459. # ROLLBACK TO SAVEPOINT. That is done only for those stored
  460. # routines that are not part of LOCK TABLES prelocking list.
  461. # Those stored routines that are part of LOCK TABLES
  462. # prelocking list are implicitly locked when entering
  463. # LOCK TABLES, and ROLLBACK TO SAVEPOINT has no effect on
  464. # them.
  465. #
  466. create function f1() returns varchar(20) return "f1()";
  467. create function f2() returns varchar(20) return "f2()";
  468. create view v1 as select f1() as a;
  469. set @@session.autocommit=0;
  470. lock table v1 read;
  471. select * from v1;
  472. a
  473. f1()
  474. savepoint sv;
  475. select f2();
  476. f2()
  477. f2()
  478. # --> connection con1
  479. # Sending 'drop function f1'...
  480. drop function f1;
  481. # --> connection con2
  482. # Waiting for 'drop function f1' to get blocked on MDL lock...
  483. # Sending 'drop function f2'...
  484. drop function f2;
  485. # --> connection default
  486. # Waiting for 'drop function f2' to get blocked on MDL lock...
  487. rollback to savepoint sv;
  488. # --> connection con2
  489. # Reaping 'drop function f2'...
  490. # --> connection default
  491. unlock tables;
  492. # --> connection con1
  493. # Reaping 'drop function f1'...
  494. # --> connection default
  495. drop function f1;
  496. ERROR 42000: FUNCTION test.f1 does not exist
  497. drop function f2;
  498. ERROR 42000: FUNCTION test.f2 does not exist
  499. drop view v1;
  500. set @@session.autocommit=default;
  501. #
  502. # 8) Check the situation when we're preparing or executing a
  503. # prepared statement, and as part of that try to flush the
  504. # session sp cache. However, one of the procedures that
  505. # needs a flush is in use. Verify that there is no infinite
  506. # reprepare loop and no crash.
  507. #
  508. create function f1() returns int return 1;
  509. #
  510. # We just mention p1() in the body of f2() to make
  511. # sure that p1() metadata is validated when validating
  512. # 'select f2()'.
  513. # Recursion is not allowed in stored functions, so
  514. # an attempt to just invoke p1() from f2() which is in turn
  515. # called from p1() would have given a run-time error.
  516. #
  517. create function f2() returns int
  518. begin
  519. if @var is null then
  520. call p1();
  521. end if;
  522. return 1;
  523. end|
  524. create procedure p1()
  525. begin
  526. select f1() into @var;
  527. execute stmt;
  528. end|
  529. # --> connection con2
  530. prepare stmt from "select f2()";
  531. # --> connection default
  532. begin;
  533. select f1();
  534. f1()
  535. 1
  536. # --> connection con1
  537. # Sending 'alter function f1 ...'...
  538. alter function f1 comment "comment";
  539. # --> connection con2
  540. # Waiting for 'alter function f1 ...' to get blocked on MDL lock...
  541. # Sending 'call p1()'...
  542. call p1();
  543. # Waiting for 'call p1()' to get blocked on MDL lock on f1...
  544. # Let 'alter function f1 ...' go through...
  545. commit;
  546. # --> connection con1
  547. # Reaping 'alter function f1 ...'
  548. # --> connection con2
  549. # Reaping 'call p1()'...
  550. f2()
  551. 1
  552. deallocate prepare stmt;
  553. # --> connection default
  554. drop function f1;
  555. drop function f2;
  556. drop procedure p1;
  557. #
  558. # 9) Check the situation when a stored function is invoked
  559. # from a stored procedure, and recursively invokes the
  560. # stored procedure that is in use. But for the second
  561. # invocation, a cache flush is requested. We can't
  562. # flush the procedure that's in use, and are forced
  563. # to use an old version. It is not a violation of
  564. # consistency, since we unroll top-level calls.
  565. # Just verify the code works.
  566. #
  567. create function f1() returns int return 1;
  568. begin;
  569. select f1();
  570. f1()
  571. 1
  572. # --> connection con1
  573. # Sending 'alter function f1 ...'...
  574. alter function f1 comment "comment";
  575. # --> connection con2
  576. # Waiting for 'alter function f1 ...' to get blocked on MDL lock...
  577. #
  578. # We just mention p1() in the body of f2() to make
  579. # sure that p1() is prelocked for f2().
  580. # Recursion is not allowed in stored functions, so
  581. # an attempt to just invoke p1() from f2() which is in turn
  582. # called from p1() would have given a run-time error.
  583. #
  584. create function f2() returns int
  585. begin
  586. if @var is null then
  587. call p1();
  588. end if;
  589. return 1;
  590. end|
  591. create procedure p1()
  592. begin
  593. select f1() into @var;
  594. select f2() into @var;
  595. end|
  596. # Sending 'call p1()'...
  597. call p1();
  598. # Waiting for 'call p1()' to get blocked on MDL lock on f1...
  599. # Let 'alter function f1 ...' go through...
  600. commit;
  601. # --> connection con1
  602. # Reaping 'alter function f1 ...'
  603. # --> connection con2
  604. # Reaping 'call p1()'...
  605. # --> connection default
  606. drop function f1;
  607. drop function f2;
  608. drop procedure p1;
  609. #
  610. # 10) A select from information_schema.routines now
  611. # flushes the stored routines caches. Test that this
  612. # does not remove from the cache a stored routine
  613. # that is already prelocked.
  614. #
  615. create function f1() returns int return get_lock("30977", 100000);
  616. create function f2() returns int return 2;
  617. create function f3() returns varchar(255)
  618. begin
  619. declare res varchar(255);
  620. declare c cursor for select routine_name from
  621. information_schema.routines where routine_name='f1';
  622. select f1() into @var;
  623. open c;
  624. fetch c into res;
  625. close c;
  626. select f2() into @var;
  627. return res;
  628. end|
  629. # --> connection con1
  630. select get_lock("30977", 0);
  631. get_lock("30977", 0)
  632. 1
  633. # --> connection default
  634. # Sending 'select f3()'...
  635. select f3();
  636. # --> connection con1
  637. # Waiting for 'select f3()' to get blocked on the user level lock...
  638. # Do something to change the cache version.
  639. create function f4() returns int return 4;
  640. drop function f4;
  641. select release_lock("30977");
  642. release_lock("30977")
  643. 1
  644. # --> connection default
  645. # Reaping 'select f3()'...
  646. # Routine 'f2()' should exist and get executed successfully.
  647. f3()
  648. f1
  649. select @var;
  650. @var
  651. 2
  652. drop function f1;
  653. drop function f2;
  654. drop function f3;
  655. # 11) Check the situation when the connection is flushing the
  656. # SP cache which contains a procedure that is being executed.
  657. #
  658. # Function f1() calls p1(). Procedure p1() has a DROP
  659. # VIEW statement, which, we know, invalidates the routines cache.
  660. # During cache flush p1() must not be flushed since it's in
  661. # use.
  662. #
  663. create function f1() returns int
  664. begin
  665. call p1();
  666. return 1;
  667. end|
  668. create procedure p1()
  669. begin
  670. create view v1 as select 1;
  671. drop view v1;
  672. select f1() into @var;
  673. set @exec_count=@exec_count+1;
  674. end|
  675. set @exec_count=0;
  676. call p1();
  677. ERROR HY000: Recursive limit 0 (as set by the max_sp_recursion_depth variable) was exceeded for routine p1
  678. select @exec_count;
  679. @exec_count
  680. 0
  681. set @@session.max_sp_recursion_depth=5;
  682. set @exec_count=0;
  683. call p1();
  684. ERROR HY000: Explicit or implicit commit is not allowed in stored function or trigger.
  685. select @exec_count;
  686. @exec_count
  687. 0
  688. drop procedure p1;
  689. drop function f1;
  690. set @@session.max_sp_recursion_depth=default;
  691. # --> connection con1
  692. # --> connection con2
  693. # --> connection con3
  694. # --> connection default
  695. #
  696. # SHOW CREATE PROCEDURE p1 called from p1, after p1 was altered
  697. #
  698. # We are just covering the existing behaviour with tests. The
  699. # results are not necessarily correct."
  700. #
  701. CREATE PROCEDURE p1()
  702. BEGIN
  703. SELECT get_lock("test", 10);
  704. SHOW CREATE PROCEDURE p1;
  705. END|
  706. # Connection default
  707. SELECT get_lock("test", 10);
  708. get_lock("test", 10)
  709. 1
  710. # Connection 2
  711. # Will halt before executing SHOW CREATE PROCEDURE p1
  712. # Sending:
  713. CALL p1();
  714. # Connection 3
  715. # Alter p1
  716. DROP PROCEDURE p1;
  717. CREATE PROCEDURE p1() BEGIN END;
  718. # Connection default
  719. # Resume CALL p1, now with new p1
  720. SELECT release_lock("test");
  721. release_lock("test")
  722. 1
  723. # Connection 2
  724. # Reaping: CALL p1()
  725. get_lock("test", 10)
  726. 1
  727. Procedure sql_mode Create Procedure character_set_client collation_connection Database Collation
  728. p1 CREATE DEFINER=`root`@`localhost` PROCEDURE `p1`()
  729. BEGIN
  730. SELECT get_lock("test", 10);
  731. SHOW CREATE PROCEDURE p1;
  732. END latin1 latin1_swedish_ci latin1_swedish_ci
  733. # Connection 3
  734. # Connection 2
  735. # Connection default;
  736. DROP PROCEDURE p1;
  737. #
  738. # End of 5.5 tests
  739. #