database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-07-30 16:24:05 +03:00
Code Activity

Fix for bug #12739 "Deadlock in multithreaded environment during creating/

Browse Source 
droping trigger on InnoDB table".

Deadlock occured in cases when we were trying to create two triggers for
the same InnoDB table concurrently and both threads were able to reach
close_cached_table() simultaneously. Bugfix implements new approach to
table locking and table cache invalidation during creation/dropping
of trigger.

No testcase is supplied since bug was repeatable only under high concurrency.


sql/mysql_priv.h:
  reopen_name_locked_table():
    Changed function signature to make it more robust against erroneous usage.
sql/sql_base.cc:
  reopen_name_locked_table():
    Changed function signature to make it more robust against erroneous usage.
    Obtaining LOCK_open lock is now resposibility of caller.
    When searching for the table to open we should not prefer connection's current
    database over database which was explicitly specified in TABLE_LIST::db member
    (even if database is not explicitly specified for table in original query
    TABLE_LIST::db will be set properly at parsing stage).
    Fixed behavior of function in cases when error occurs during opening of table.
sql/sql_table.cc:
  prepare_for_restore()/prepare_for_repair():
    We should not prefer connection's current database over database which was
    specified in TABLE_LIST::db member (even if database is not explicitly
    specified for table in original query TABLE_LIST::db will be set properly
    at parsing stage). Fixed behavior in unlikely case when we are unable
    to open table which we are restoring/reparing at the end of preparation
    stage.
sql/sql_trigger.cc:
  mysql_create_or_drop_trigger():
    Now instead of opening and locking table, creating trigger, and then trying
    to invalidate all instances of this table in table cache, we obtain name
    lock on table first (thus ensuring that no other thread has this table
    open), open it, create trigger and then close table therefore releasing lock.
    New approach is more in line with other places where change .frm files
    (i.e. change table meta-data).
    With this change we also get rid of deadlock which occured in cases when we
    were trying to create two triggers for the same InnoDB table concurrently
    and both threads were able to reach close_cached_table() simultaneously.
    (Alternative was to forbid to InnoDB downgrade locks for CREATE/DROP
     TRIGGER statements in one way or another but I think that proposed
     solution is better long term).
This commit is contained in:
unknown
2005-10-17 22:37:24 +04:00
parent 63e7824fc2
commit dd02b98d14
4 changed files with 99 additions and 66 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
sql/mysql_priv.h
View File

@ -757,7 +757,7 @@ bool mysql_create_or_drop_trigger(THD *thd, TABLE_LIST *tables, bool create);
TABLE *open_ltable(THD *thd, TABLE_LIST *table_list, thr_lock_type update);
TABLE *open_table(THD *thd, TABLE_LIST *table_list, MEM_ROOT* mem,
bool *refresh, uint flags);
TABLE *reopen_name_locked_table(THD* thd, TABLE_LIST* table);
bool reopen_name_locked_table(THD* thd, TABLE_LIST* table);
TABLE *find_locked_table(THD *thd, const char *db,const char *table_name);
bool reopen_table(TABLE *table,bool locked);
bool reopen_tables(THD *thd,bool get_locks,bool in_refresh);