database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-07-29 05:21:33 +03:00
Code Activity

Implementation of Monty's idea: Now we can open mysql.proc table for lookup

Browse Source 
of stored routines definitions even if we already have some tables open and
locked. To avoid deadlocks in this case we have to put certain restrictions
on locking of mysql.proc table.

This allows to use stored routines safely under LOCK TABLES without explicitly
mentioning mysql.proc in the list of locked tables. It also fixes bug #11554
"Server crashes on statement indirectly using non-cached function".


mysql-test/r/sp-error.result:
  Added test which checks that now we can read stored routines definitions
  under LOCK TABLES even if we have not locked mysql.proc explicitly. Also
  added check for restrictions which this ability puts on mysql.proc locking.
  Updated test for bug #9566 to correspond this new situation.
mysql-test/r/sp-threads.result:
  Added test for bug #11554 "Server crashes on statement indirectly using
  non-cached function".
mysql-test/t/sp-error.test:
  Added test which checks that now we can read stored routines definitions
  under LOCK TABLES even if we have not locked mysql.proc explicitly. Also
  added check for restrictions which this ability puts on mysql.proc locking.
  Updated test for bug #9566 to correspond this new situation.
mysql-test/t/sp-threads.test:
  Added test for bug #11554 "Server crashes on statement indirectly using
  non-cached function".
sql/lock.cc:
  get_lock_data():
   To be able to open and lock for reading system tables like 'mysql.proc',
   when we already have some tables opened and locked, and avoid deadlocks
   we have to disallow write-locking of these tables with any other tables.
sql/mysql_priv.h:
  open_table() has new parameter which allows to open table even if some-one
  has done a flush or holding namelock on it.
sql/share/errmsg.txt:
  Added error message saying that one cannot write-lock some of system tables
  with any other tables.
sql/sp.cc:
  open_proc_table_for_read()/close_proc_table():
    Added functions to be able open and close mysql.proc table when we already
    have some tables open and locked.
  open_proc_table_for_update():
    Added function to simplify opening of mysql.proc for updates.
  db_find_routine_aux()/db_find_routine()/db_update_routine()/...
    Moved responsibility for opening mysql.proc table from db_find_routine_aux()
    one level up, since this level knows better which type of table access for
    reading of for update it needs.
  sp_function_exists():
    Removed unused function.
sql/sp.h:
  sp_function_exists():
    Removed unused function.
sql/sql_base.cc:
  open_table():
    Added new parameter which allows to open table even if some-one has done a
    flush or holding namelock on it.
  open_unireg_entry():
    Mark 'mysql.proc' as a system table which has special restrictions on its
    locking, but thanks to them can be open and locked even if we already have
    some open and locked.
sql/sql_class.cc:
  Moved THD members holding information about open and locked tables to separate
  Open_tables_state class to be able to save/restore this state easier.
  Added THD::push_open_tables_state()/pop_open_tables_state() methods for
  saving/restoring this state.
sql/sql_class.h:
  Moved THD members holding information about open and locked tables to separate
  Open_tables_state class to be able to save/restore this state easier.
  Added THD::push_open_tables_state()/pop_open_tables_state() methods for
  saving/restoring this state.
sql/sql_lex.cc:
  Removed LEX::proc_table member which was not really used.
sql/sql_lex.h:
  Removed LEX::proc_table member which was not really used.
sql/sql_table.cc:
  open_table() has new parameter which allows to open table even if some-one
  has done a flush or holding namelock on it.
sql/table.h:
  Added TABLE_SHARE::system_table indicating that this table is system table
  like 'mysql.proc' and we want to be able to open and read-lock it even when
  we already have some tables open and locked (and because of this we have
  to put some restrictions on write locking it).
This commit is contained in:
unknown
2005-07-13 13:48:13 +04:00
parent 98a3bae33d
commit 11f9e513d7
16 changed files with 459 additions and 202 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
mysql-test/t/sp-error.test
View File

@ -386,6 +386,29 @@ call p(42, @tmp_y, 43)|
drop procedure p|
#
# Let us test that we can access mysql.proc table for routines
# definitions lookup without locking it explicitly.
#
create procedure p() begin end|
lock table t1 read|
# This should succeed
call p()|
unlock tables|
drop procedure p|
# Let us check restrictions which this ability puts on mysql.proc locking.
--error ER_WRONG_LOCK_OF_SYSTEM_TABLE
lock tables t1 read, mysql.proc write|
--error ER_WRONG_LOCK_OF_SYSTEM_TABLE
lock tables mysql.proc write, mysql.user write|
# Locking for read should be OK
lock tables t1 read, mysql.proc read|
unlock tables|
# You also should be able lock only mysql.proc for write
lock tables mysql.proc write|
unlock tables|
#
# BUG#1965
#
@ -676,9 +699,7 @@ create procedure bug6600()
# BUG#9566: explicit LOCK TABLE and store procedures result in illegal state
#
# We should not think that mysql.proc table does not exist if we are unable
# to open it under LOCK TABLE or in prelocked mode. Probably this test
# should be removed when Monty will allow access to mysql.proc without
# locking it.
# to open it under LOCK TABLE or in prelocked mode.
#
--disable_warnings
drop procedure if exists bug9566|
@ -688,9 +709,11 @@ begin
select * from t1;
end|
lock table t1 read|
# This should fail because we forgot to lock mysql.proc table explicitly
# This should fail since we forgot to lock mysql.proc for writing
# explicitly, and we can't open mysql.proc for _writing_ if there
# are locked tables.
--error 1100
call bug9566()|
alter procedure bug9566 comment 'Some comment'|
unlock tables|
# This should succeed
drop procedure bug9566|