This bug is actually two bugs in one, one of which is CREATE TRIGGER under
LOCK TABLES and the other is CREATE TRIGGER under LOCK TABLES simultaneous
to a FLUSH TABLES WITH READ LOCK (global read lock). Both situations could
lead to a server crash or deadlock.
The first problem arises from the fact that when under LOCK TABLES, if the
table is in the set of locked tables, the table is already open and it doesn't
need to be reopened (not a placeholder). Also in this case, if the table is
not write locked, a exclusive lock can't be acquired because of a possible
deadlock with another thread also holding a (read) lock on the table. The
second issue arises from the fact that one should never wait for a global
read lock if it's holding any locked tables, because the global read lock
is waiting for these tables and this leads to a circular wait deadlock.
The solution for the first case is to check if the table is write locked
and upgraded the write lock to a exclusive lock and fail otherwise for non
write locked tables. Grabbin the exclusive lock in this case also means
to ensure that the table is opened only by the calling thread. The second
issue is partly fixed by not waiting for the global read lock if the thread
is holding any locked tables.
The second issue is only partly addressed in this patch because it turned
out to be much wider and also affects other DDL statements. Reported as
Bug#32395
mysql-test/r/trigger.result:
Add test case result for Bug#23713
mysql-test/r/trigger_notembedded.result:
Add test case result for Bug#23713
mysql-test/t/trigger.test:
Add test case for Bug#23713
mysql-test/t/trigger_notembedded.test:
Add test case for Bug#23713
sql/mysql_priv.h:
Locally export wait_while_table_is_used and name_lock_locked_table
and add flag to mysql_ha_rm_tables to signal that LOCK_open is locked.
sql/sql_base.cc:
Introduce name_lock_locked_table function and match
close_old_data_files function declaration and definition.
sql/sql_handler.cc:
Add flag to mysql_ha_rm_tables to signal that LOCK_open is locked.
sql/sql_rename.cc:
Fix mysql_ha_rm_tables caller.
sql/sql_table.cc:
Export wait_while_table_is_used and assert that LOCK_open is locked
and fix mysql_ha_rm_tables caller.
sql/sql_trigger.cc:
Upgrade write locked tables to a exclusive lock and fail if
the table is not write locked. Also, don't wait for the global
read lock if under LOCK TABLES.
This bug is actually two bugs in one, one of which is CREATE TRIGGER under
LOCK TABLES and the other is CREATE TRIGGER under LOCK TABLES simultaneous
to a FLUSH TABLES WITH READ LOCK (global read lock). Both situations could
lead to a server crash or deadlock.
The first problem arises from the fact that when under LOCK TABLES, if the
table is in the set of locked tables, the table is already open and it doesn't
need to be reopened (not a placeholder). Also in this case, if the table is
not write locked, a exclusive lock can't be acquired because of a possible
deadlock with another thread also holding a (read) lock on the table. The
second issue arises from the fact that one should never wait for a global
read lock if it's holding any locked tables, because the global read lock
is waiting for these tables and this leads to a circular wait deadlock.
The solution for the first case is to check if the table is write locked
and upgraded the write lock to a exclusive lock and fail otherwise for non
write locked tables. Grabbin the exclusive lock in this case also means
to ensure that the table is opened only by the calling thread. The second
issue is partly fixed by not waiting for the global read lock if the thread
is holding any locked tables.
The second issue is only partly addressed in this patch because it turned
out to be much wider and also affects other DDL statements. Reported as
Bug#32395
The problem is that DROP TABLE and other DDL statements failed to
automatically close handlers associated with tables that were marked
for reopen (FLUSH TABLES).
The current implementation fails to properly discard handlers of
dropped tables (that were marked for reopen) because it searches
on the open handler tables list and using the current alias of the
table being dropped. The problem is that it must not use the open
handler tables list to search because the table might have been
closed (marked for reopen) by a flush tables command and also it
must not use the current table alias at all since multiple different
aliases may be associated with a single table. This is specially
visible when a user has two open handlers (using alias) of a same
table and a flush tables command is issued before the table is
dropped (see test case). Scanning the handler table list is also
useless for dropping handlers associated with temporary tables,
because temporary tables are not kept in the THD::handler_tables
list.
The solution is to simple scan the handlers hash table searching
for, and deleting all handlers with matching table names if the
reopen flag is not passed to the flush function, indicating that
the handlers should be deleted. All matching handlers are deleted
even if the associated the table is not open.
mysql-test/include/handler.inc:
Add test case for Bug#31397
mysql-test/r/handler_innodb.result:
Add test case result for Bug#31397
mysql-test/r/handler_myisam.result:
Add test case result for Bug#31397
sql/mysql_priv.h:
Rename flush functions to better match the intent of the caller and
update functions prototypes and remove unused flags.
sql/sql_base.cc:
Rename flush functions to better match the intent of the caller.
sql/sql_class.cc:
Rename the flush functions to better match the intent of the caller.
The hash_free function is moved to the cleanup.
sql/sql_handler.cc:
When dropping tables for a final close, scan the handler's hash table since
the table might not be in the handlers open table list because the table was
marked for reopen or because it's a temporary table.
sql/sql_rename.cc:
Drop handlers associated with tables that are being renamed.
sql/sql_table.cc:
Now that temporary tables are properly removed even when opened
by a SQL HANDLER, enable the assert since this branch can't be taken
outside of SF/trigger/prelocked mode.
The problem is that DROP TABLE and other DDL statements failed to
automatically close handlers associated with tables that were marked
for reopen (FLUSH TABLES).
The current implementation fails to properly discard handlers of
dropped tables (that were marked for reopen) because it searches
on the open handler tables list and using the current alias of the
table being dropped. The problem is that it must not use the open
handler tables list to search because the table might have been
closed (marked for reopen) by a flush tables command and also it
must not use the current table alias at all since multiple different
aliases may be associated with a single table. This is specially
visible when a user has two open handlers (using alias) of a same
table and a flush tables command is issued before the table is
dropped (see test case). Scanning the handler table list is also
useless for dropping handlers associated with temporary tables,
because temporary tables are not kept in the THD::handler_tables
list.
The solution is to simple scan the handlers hash table searching
for, and deleting all handlers with matching table names if the
reopen flag is not passed to the flush function, indicating that
the handlers should be deleted. All matching handlers are deleted
even if the associated the table is not open.
If a stored function that contains a drop temporary table statement
is invoked by a create temporary table of the same name may cause
a server crash. The problem is that when dropping a table no check
is done to ensure that table is not being used by some outer query
(or outer statement), potentially leaving the outer query with a
reference to a stale (freed) table.
The solution is when dropping a temporary table, always check if
the table is being used by some outer statement as a temporary
table can be dropped inside stored procedures.
The check is performed by looking at the TABLE::query_id value for
temporary tables. To simplify this check and to solve a bug related
to handling of temporary tables in prelocked mode, this patch changes
the way in which this member is used to track the fact that table is
used/unused. Now we ensure that TABLE::query_id is zero for unused
temporary tables (which means that all temporary tables which were
used by a statement should be marked as free for reuse after it's
execution has been completed).
mysql-test/include/handler.inc:
Add test case for side effect of Bug#30882
mysql-test/r/handler_innodb.result:
Add test case result for side effect of Bug#30882
mysql-test/r/handler_myisam.result:
Add test case result for side effect of Bug#30882
mysql-test/r/sp-error.result:
Add test case result for Bug#30882
mysql-test/t/sp-error.test:
Add test case for Bug#30882
sql/event_db_repository.cc:
Update close_thread_tables call, no more default values.
sql/mysql_priv.h:
Remove implicit default parameters values of the close_thread_tables
function as no callers are using it.
sql/slave.cc:
Update close_thread_tables call, no more default values
sql/sp_head.cc:
Update close_thread_tables call, no more default values
sql/sql_base.cc:
Changed the approach to distinguishing currently unused temporary tables.
Now we ensure that such tables always have TABLE::query_id set to 0 and
use this fact to perform checks during opening and dropping of temporary
tables. This means that we have to call close_thread_tables() even for
statements which use only temporary tables. To make this call cheaper,
we re-factored close_thread_tables() to not take LOCK_open unless there
are open base tables.
sql/sql_handler.cc:
Properly close temporary tables associated with a handler.
sql/sql_insert.cc:
close_temporary_table is now merged into drop_temporary_table.
sql/sql_parse.cc:
Now the condition doesn't cover all cases because close_thread_tables()
must be called even for statements that use only temporary tables.
sql/sql_table.cc:
Use drop_temporary_table which perform checks to verify if
the table is not being used. Error path problem is due to
a handler tables issue and is going to be addressed in bug
31397.
sql/table.h:
Rename previously unused clear_query_id and document the usage of
query_id and open_by_handler.
If a stored function that contains a drop temporary table statement
is invoked by a create temporary table of the same name may cause
a server crash. The problem is that when dropping a table no check
is done to ensure that table is not being used by some outer query
(or outer statement), potentially leaving the outer query with a
reference to a stale (freed) table.
The solution is when dropping a temporary table, always check if
the table is being used by some outer statement as a temporary
table can be dropped inside stored procedures.
The check is performed by looking at the TABLE::query_id value for
temporary tables. To simplify this check and to solve a bug related
to handling of temporary tables in prelocked mode, this patch changes
the way in which this member is used to track the fact that table is
used/unused. Now we ensure that TABLE::query_id is zero for unused
temporary tables (which means that all temporary tables which were
used by a statement should be marked as free for reuse after it's
execution has been completed).
If mysql_lock_tables fails because the lock was aborted, we need to
reset thd->some_tables_delete, otherwise we might loop indefinitely
because handler's tables are not closed in a standard way, meaning
that close_thread_tables() (which resets some_tables_deleted) is not
used.
This patch fixes sporadical failures of handler_myisam/innodb tests
which were introduced by previous fix for this bug.
sql/sql_handler.cc:
Properly reset thd->some_tables_deleted if mysql_lock_tables
fails for some reason.
If mysql_lock_tables fails because the lock was aborted, we need to
reset thd->some_tables_delete, otherwise we might loop indefinitely
because handler's tables are not closed in a standard way, meaning
that close_thread_tables() (which resets some_tables_deleted) is not
used.
This patch fixes sporadical failures of handler_myisam/innodb tests
which were introduced by previous fix for this bug.
This deadlock occurs when a client issues a HANDLER ... OPEN statement
that tries to open a table that has a pending name-lock on it by another
client that also needs a name-lock on some other table which is already
open and associated to a HANDLER instance owned by the first client.
The deadlock happens because the open_table() function will back-off
and wait until the name-lock goes away, causing a circular wait if some
other name-lock is also pending for one of the open HANDLER tables.
Such situation, for example, can be easily repeated by issuing a RENAME
TABLE command in such a way that the existing table is already open
as a HANDLER table by another client and this client tries to open
a HANDLER to the new table name.
The solution is to allow handler tables with older versions (marked for
flush) to be closed before waiting for the name-lock completion. This is
safe because no other name-lock can be issued between the flush and the
check for pending name-locks.
The test case for this bug is going to be committed into 5.1 because it
requires a test feature only avaiable in 5.1 (wait_condition).
sql/sql_base.cc:
Improve comments in the open_table() function, stating the importance
of the handler tables flushing for the back-off process.
sql/sql_handler.cc:
Allows handler tables flushes when opening new tables in order to avoid
potential deadlocks. Add comments explaining the importance of the flush.
This deadlock occurs when a client issues a HANDLER ... OPEN statement
that tries to open a table that has a pending name-lock on it by another
client that also needs a name-lock on some other table which is already
open and associated to a HANDLER instance owned by the first client.
The deadlock happens because the open_table() function will back-off
and wait until the name-lock goes away, causing a circular wait if some
other name-lock is also pending for one of the open HANDLER tables.
Such situation, for example, can be easily repeated by issuing a RENAME
TABLE command in such a way that the existing table is already open
as a HANDLER table by another client and this client tries to open
a HANDLER to the new table name.
The solution is to allow handler tables with older versions (marked for
flush) to be closed before waiting for the name-lock completion. This is
safe because no other name-lock can be issued between the flush and the
check for pending name-locks.
The test case for this bug is going to be committed into 5.1 because it
requires a test feature only avaiable in 5.1 (wait_condition).
This bug is a symptom of the way handler's tables are managed. The
most different aspect, compared to the conventional behavior, is that
the handler's tables are long lived, meaning that their lifetimes are
not bounded by the duration of the command that opened them. For this
effect the handler code uses its own list (handler_tables instead of
open_tables) to hold open handler tables so that the tables won't be
closed at the end of the command/statement. Besides the handler_tables
list, there is a hash (handler_tables_hash) which is used to associate
handler aliases to tables and to refresh the tables upon demand (flush
tables).
The current implementation doesn't work properly with refreshed tables
-- more precisely when flush commands are issued by other initiators.
This happens because when a handler open or read statement is being
processed, the associated table has to be opened or locked and, for this
matter, the open_tables and handler_tables lists are swapped so that the
new table being opened is inserted into the handler_tables list. But when
opening or locking the table, if the refresh version is different from the
thread refresh version then all used tables in the open_tables list (now
handler_tables) are refreshed. In the "refreshing" process the handler
tables are flushed (closed) without being properly unlinked from the
handler hash.
The current implementation also fails to properly discard handlers of
dropped tables, but this and other problems are going to be addressed
in the fixes for bugs 31397 and 31409.
The chosen approach tries to properly save and restore the table state
so that no table is flushed during the table open and lock operations.
The logic is almost the same as before with the list swapping, but with
a working glue code.
The test case for this bug is going to be committed into 5.1 because it
requires a test feature only avaiable in 5.1 (wait_condition).
sql/sql_handler.cc:
Backup and reset the open_tables and handler_table lists when opening
a new handler table and merge the opened table into the handler_tables
list afterwards. When locking, do the same but possibly close the table
if a refresh is pending.
This bug is a symptom of the way handler's tables are managed. The
most different aspect, compared to the conventional behavior, is that
the handler's tables are long lived, meaning that their lifetimes are
not bounded by the duration of the command that opened them. For this
effect the handler code uses its own list (handler_tables instead of
open_tables) to hold open handler tables so that the tables won't be
closed at the end of the command/statement. Besides the handler_tables
list, there is a hash (handler_tables_hash) which is used to associate
handler aliases to tables and to refresh the tables upon demand (flush
tables).
The current implementation doesn't work properly with refreshed tables
-- more precisely when flush commands are issued by other initiators.
This happens because when a handler open or read statement is being
processed, the associated table has to be opened or locked and, for this
matter, the open_tables and handler_tables lists are swapped so that the
new table being opened is inserted into the handler_tables list. But when
opening or locking the table, if the refresh version is different from the
thread refresh version then all used tables in the open_tables list (now
handler_tables) are refreshed. In the "refreshing" process the handler
tables are flushed (closed) without being properly unlinked from the
handler hash.
The current implementation also fails to properly discard handlers of
dropped tables, but this and other problems are going to be addressed
in the fixes for bugs 31397 and 31409.
The chosen approach tries to properly save and restore the table state
so that no table is flushed during the table open and lock operations.
The logic is almost the same as before with the list swapping, but with
a working glue code.
The test case for this bug is going to be committed into 5.1 because it
requires a test feature only avaiable in 5.1 (wait_condition).
into weblab.(none):/home/marcsql/TREE/mysql-5.1-rt50-merge
client/mysql.cc:
Auto merged
mysql-test/r/query_cache.result:
Auto merged
mysql-test/t/query_cache.test:
Auto merged
sql/item_cmpfunc.h:
Auto merged
sql/sp.cc:
Auto merged
sql/sp_head.cc:
Auto merged
sql/sql_class.h:
Auto merged
sql/sql_db.cc:
Auto merged
sql/sql_handler.cc:
Auto merged
sql/sql_lex.cc:
Auto merged
sql/sql_lex.h:
Auto merged
sql/sql_yacc.yy:
Auto merged
into moksha.local:/Users/davi/mysql/push/bugs/30632-5.1
sql/sql_handler.cc:
Auto merged
mysql-test/include/handler.inc:
Auto merged
mysql-test/r/handler_myisam.result:
Auto merged
If, after the tables are locked, one of the conditions to read from a
HANDLER table is not met, the handler code wrongly jumps to a error path
that won't unlock the tables.
The user-visible effect is that after a error in a handler read command,
all subsequent handler operations on the same table will hang.
The fix is simply to correct the code to jump to the (same) error path that
unlocks the tables.
mysql-test/r/handler.result:
Bug#30632 test case result
mysql-test/t/handler.test:
Bug#30632 test case
sql/sql_handler.cc:
Always unlock the internal and external table level locks if any of the conditions
(including errors) to read from a HANDLER table are not met.
If, after the tables are locked, one of the conditions to read from a
HANDLER table is not met, the handler code wrongly jumps to a error path
that won't unlock the tables.
The user-visible effect is that after a error in a handler read command,
all subsequent handler operations on the same table will hang.
The fix is simply to correct the code to jump to the (same) error path that
unlocks the tables.
into hynda.mysql.fi:/home/my/mysql-5.1-marvel
client/mysqldump.c:
Auto merged
sql/event_db_repository.cc:
Auto merged
sql/event_queue.cc:
Auto merged
sql/field.cc:
Auto merged
sql/item.cc:
Auto merged
sql/item_subselect.cc:
Auto merged
sql/log_event.h:
Auto merged
sql/sp.cc:
Auto merged
sql/sql_class.cc:
Auto merged
sql/sql_db.cc:
Auto merged
sql/sql_handler.cc:
Auto merged
sql/sql_lex.cc:
Auto merged
sql/sql_plugin.cc:
Auto merged
sql/sql_select.cc:
Auto merged
sql/sql_show.cc:
Auto merged
sql/table.h:
Auto merged
sql/sql_yacc.yy:
Manual merge with 5.1 main tree.
into hynda.mysql.fi:/home/my/mysql-5.1-marvel
mysql-test/mysql-test-run.pl:
Auto merged
sql/field.cc:
Auto merged
sql/handler.cc:
Auto merged
sql/item_func.cc:
Auto merged
sql/log.cc:
Auto merged
sql/log_event_old.cc:
Auto merged
sql/mysql_priv.h:
Auto merged
sql/mysqld.cc:
Auto merged
sql/sp_head.cc:
Auto merged
sql/sql_class.cc:
Auto merged
sql/sql_class.h:
Auto merged
sql/sql_handler.cc:
Auto merged
sql/sql_insert.cc:
Auto merged
sql/sql_select.cc:
Auto merged
sql/sql_string.cc:
Auto merged
sql/sql_table.cc:
Auto merged
mysql-test/r/status.result:
Manual merge with main 5.0 tree.
mysql-test/t/status.test:
Manual merge with main 5.0 tree.
sql/log_event.cc:
Manual merge with main 5.0 tree.
sql/log_event.h:
Manual merge with main 5.0 tree.
into weblab.(none):/home/marcsql/TREE/mysql-5.1-rt-merge
sql/field.cc:
Auto merged
sql/field.h:
Auto merged
sql/log_event.h:
Auto merged
sql/rpl_record.cc:
Auto merged
sql/slave.h:
Auto merged
sql/sql_base.cc:
Auto merged
sql/sql_class.cc:
Auto merged
sql/sql_handler.cc:
Auto merged
sql/sql_parse.cc:
Auto merged
sql/sql_select.cc:
Auto merged
mysql_ha_open calls mysql_ha_close on the error path (unsupported) to close the (opened) table before inserting it into the tables hash list handler_tables_hash) but mysql_ha_close only closes tables which are on the hash list, causing the table to be left open and locked.
This change moves the table close logic into a separate function that is always called on the error path of mysql_ha_open or on a normal handler close (mysql_ha_close).
mysql-test/r/lock_multi.result:
Bug#25856 test result
mysql-test/t/lock_multi.test:
Bug#25856 test case
sql/sql_handler.cc:
Move the table close logic into a separate function that is always called on the error path of mysql_ha_open or on a normal handler close.
mysql_ha_open calls mysql_ha_close on the error path (unsupported) to close the (opened) table before inserting it into the tables hash list handler_tables_hash) but mysql_ha_close only closes tables which are on the hash list, causing the table to be left open and locked.
This change moves the table close logic into a separate function that is always called on the error path of mysql_ha_open or on a normal handler close (mysql_ha_close).
mysql_ha_open calls mysql_ha_close on the error path (unsupported) to close the (opened) table before inserting it into the tables hash list handler_tables_hash) but mysql_ha_close only closes tables which are on the hash list, causing the table to be left open and locked.
This change moves the table close logic into a separate function that is always called on the error path of mysql_ha_open or on a normal handler close (mysql_ha_close).
mysql-test/r/handler.result:
Bug#25856 test result
mysql-test/t/handler.test:
Bug#25856 test case
sql/sql_handler.cc:
Move the table close logic into a separate function that is always called on the error path of mysql_ha_open or on a normal handler close
mysql_ha_open calls mysql_ha_close on the error path (unsupported) to close the (opened) table before inserting it into the tables hash list handler_tables_hash) but mysql_ha_close only closes tables which are on the hash list, causing the table to be left open and locked.
This change moves the table close logic into a separate function that is always called on the error path of mysql_ha_open or on a normal handler close (mysql_ha_close).
Faster thr_alarm()
Added 'Opened_files' status variable to track calls to my_open()
Don't give warnings when running mysql_install_db
Added option --source-install to mysql_install_db
I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems
index_read() -> index_read_map()
index_read_idx() -> index_read_idx_map()
index_read_last() -> index_read_last_map()
BUILD/compile-solaris-sparc-forte:
Updated script to current Solaris installations
Now we compile by default for 64 bits
client/mysql.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
client/mysql_upgrade.c:
Fixed compiler warning (on Forte)
client/mysqladmin.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
client/mysqlcheck.c:
Fixed compiler warning (on Forte)
client/mysqldump.c:
Fixed compiler warning (on Forte)
client/mysqlslap.c:
Fixed compiler warning (on Forte)
client/mysqltest.c:
Fixed compiler warning (on Forte)
client/sql_string.cc:
Avoid compiler warnings when using C function pointers in C++
configure.in:
Added detection of mtmalloc and ieeefp.h
extra/replace.c:
Fixed compiler warning (on Forte)
include/m_ctype.h:
Added some typedef's to make it easy to use C function pointers in C++
include/my_sys.h:
Added my_file_total_opened (counter for calls to my_open())
include/myisam.h:
Fixed compiler warning (on Forte)
libmysql/libmysql.c:
Fixed compiler warning (on Forte) by adding casts and change types
libmysql/manager.c:
Fixed compiler warning (on Forte) by adding casts and change types
mysql-test/r/ctype_cp932_binlog_stm.result:
Updated positions
(Needed because we didn't before correctly restore collation_database after running stored procedure
mysys/my_fopen.c:
Count number of opened files
mysys/my_open.c:
Count number of opened files
mysys/my_static.c:
Count number of opened files
mysys/thr_alarm.c:
Optimization to do less alarm() and pthread_sigmask() calls.
Idea is to remember time for next pending alarm and not reschedule a new alarm if it's after the current one.
Before we only did this if there was other pending alarms.
We don't have to use pthread_sigmask() in case of 'USE_ONE_SIGNAL_HAND' as the alarm()
signal will be blocked for the calling thread anyway and no other thread will have the alarm() signal enabled to call process_alarm()
regex/regcomp.c:
Fixed compiler warning (on Forte) by adding casts and change types
scripts/mysql_install_db.sh:
Added option --source-install to allow one to create a mysql database from the source tree without installing MySQL
Don't give (unnecessary) warnings
server-tools/instance-manager/angel.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
server-tools/instance-manager/thread_registry.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
sql/event_db_repository.cc:
index_read() -> index_read_map()
sql/event_queue.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
sql/field.cc:
Fixed compiler warnings about hidden fields
sql/ha_partition.cc:
Fixed compiler warnings about hidden fields
index_read() -> index_read_map()
sql/ha_partition.h:
index_read() -> index_read_map()
sql/handler.cc:
Added PAGE option to row types (to prepare for future)
index_read() -> index_read_map()
sql/handler.h:
Added ROW_TYPE_PAGE (for future)
Added flag to signal if table was to be created transactionally
I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems
index_read() -> index_read_map()
index_read_idx() -> index_read_idx_map()
index_read_last() -> index_read_last_map()
sql/item.cc:
Fixed indentation
Renamed local variable to avoid hiding class variable
sql/item_cmpfunc.cc:
Renamed local variable to avoid hiding class variable
sql/item_cmpfunc.h:
Removed not used variable
sql/item_func.cc:
Renamed local variable to avoid hiding class variable
sql/item_strfunc.cc:
Moved functions from Item_strfunc.cc
sql/item_strfunc.h:
Move functions to item_strfunc.cc
Use C function pointer type to avoid compiler warnings (with Forte)
sql/item_subselect.cc:
index_read() -> index_read_map()
sql/item_xmlfunc.cc:
Renamed local variable to avoid hiding class variable
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
sql/key.cc:
Fixed indentation
sql/log.cc:
Renamed local variable to avoid hiding class variable
sql/log_event.cc:
Removed call to my_time() when creating class instance of Log_event() as this may have static instances.
(One can't call my_time() before my_init())
index_read() -> index_read_map()
Renamed local variable to avoid hiding class variable
sql/log_event_old.cc:
Renamed local variable to avoid hiding class variable
sql/mysql_priv.h:
Made all create_backup_ctx() declarations identical.
This lifted up a bug where wrong create_backup_ctx() was called in some cases.
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
sql/mysqld.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
Fixed indentation
Don't call end_thr_alarm() when calling unireg_abort() as unireg_abort() already calls end_thr_alarm()
Added variable 'Opened_files' (number of calls to my_open() or my_fopen())
Don't print 'loose' warnings when using --bootstrap (to avoid warnings when running mysql_install_db)
Fixed compiler warnings
sql/opt_range.cc:
index_read() -> index_read_map()
sql/opt_sum.cc:
index_read() -> index_read_map()
sql/partition_info.cc:
Renamed local variable to avoid hiding class variable
sql/rpl_filter.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
sql/set_var.cc:
Renamed local variable to avoid hiding class variable
Added 'process_key_cache_t' type to avoid compiler warning (on Forte)
sql/set_var.h:
Added 'process_key_cache_t' type to avoid compiler warning (on Forte)
sql/sp.cc:
More debugging
index_read() -> index_read_map()
sql/sp_cache.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
sql/sp_head.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
Moved 'saved_creation_ctx' higher up to be able to free objects allocated by create_backup_ctx()
sql/sql_acl.cc:
index_read() -> index_read_map()
sql/sql_class.cc:
Renamed local variable to avoid hiding class variable
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
sql/sql_class.h:
Renamed local variable to avoid hiding class variable
sql/sql_db.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
sql/sql_delete.cc:
Renamed local variable to avoid hiding class variable
sql/sql_handler.cc:
index_read() -> index_read_map()
sql/sql_help.cc:
index_read() -> index_read_map()
sql/sql_insert.cc:
index_read() -> index_read_map()
Renamed local variable to avoid hiding class variable
sql/sql_lex.cc:
Renamed local variable to avoid hiding class variable
sql/sql_plugin.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
index_read() -> index_read_map()
Don't give warnings about not used plugins if we are using --warnings=0
sql/sql_select.cc:
index_read() -> index_read_map()
sql-common/client.c:
Fixed compiler warning (on Forte)
sql-common/my_time.c:
Removed never accessed code
Fixed compiler warning (on Forte)
sql/sql_servers.cc:
index_read() -> index_read_map()
sql/sql_show.cc:
Added TRANSACTIONAL to SHOW CREATE
Fixed ROW_TYPE_PAGE
sql/sql_string.cc:
Avoid compiler warnings when using C function pointers in C++
sql/sql_table.cc:
Set create_info->transactional if we used TRANSACTIONAL=1
sql/sql_udf.cc:
index_read() -> index_read_map()
sql/sql_yacc.yy:
Added TRANSACTIONAL=0|1 to CREATE (for future)
Added row type PAGE (was only partionally handled before)
sql/strfunc.cc:
Avoid compiler warnings when using C function pointers in C++
sql/table.cc:
More DBUG statements
Declare all create_backup_ctx() functions identically
Remember if table was created with TRANSACTIONAL flag or not (future safe)
Renamed local variable to avoid hiding class variable
sql/table.h:
Remember if table was created with TRANSACTIONAL=1
sql/tztime.cc:
index_read() -> index_read_map()
sql-common/pack.c:
Fixed compiler warning (on Forte)
storage/archive/archive_reader.c:
Fixed compiler warning (on Forte)
storage/archive/azio.c:
Fixed compiler warning (on Forte)
storage/blackhole/ha_blackhole.cc:
index_read() -> index_read_map()
storage/blackhole/ha_blackhole.h:
index_read() -> index_read_map()
storage/csv/ha_tina.cc:
Declare functions sent to C code with extern "C" to avoid compiler warnings (on Forte)
storage/example/ha_example.cc:
index_read() -> index_read_map()
storage/example/ha_example.h:
index_read() -> index_read_map()
storage/heap/ha_heap.cc:
index_read() -> index_read_map()
storage/heap/ha_heap.h:
index_read() -> index_read_map()
storage/heap/hp_test1.c:
Fixed compiler warning (on Forte)
storage/heap/hp_test2.c:
Fixed compiler warning (on Forte)
storage/myisam/ft_boolean_search.c:
Fixed compiler warning (on Forte)
storage/myisam/ft_nlq_search.c:
Fixed compiler warning (on Forte)
storage/myisam/ft_parser.c:
Fixed compiler warning (on Forte)
storage/myisam/ft_stopwords.c:
Fixed compiler warning (on Forte)
storage/myisam/ha_myisam.cc:
index_read() -> index_read_map()
storage/myisam/ha_myisam.h:
index_read() -> index_read_map()
storage/myisam/mi_check.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_delete.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_dynrec.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_extra.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_key.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_keycache.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_locking.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_log.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_open.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_packrec.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_page.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_rkey.c:
Added comment
storage/myisam/mi_search.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_statrec.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_test1.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_test2.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_test3.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_update.c:
Fixed compiler warning (on Forte)
storage/myisam/mi_write.c:
Fixed compiler warning (on Forte)
storage/myisam/myisamdef.h:
Fixed that file_read/file_write returns type size_t
Changed some functions to use uchar * as argument/return value instead of char*
This fixed some compiler warnings on Forte
storage/myisam/myisamlog.c:
Fixed compiler warning (on Forte)
storage/myisam/myisampack.c:
Fixed compiler warning (on Forte)
storage/myisam/rt_test.c:
Fixed compiler warning (on Forte)
storage/myisam/sort.c:
Fixed compiler warning (on Forte) by adding casts or changing variables to uchar*
storage/myisam/sp_test.c:
Fixed compiler warning (on Forte) by adding casts or changing variables to uchar*
storage/myisammrg/ha_myisammrg.cc:
index_read() -> index_read_map()
storage/myisammrg/ha_myisammrg.h:
index_read() -> index_read_map()
storage/myisammrg/myrg_create.c:
Fixed compiler warning (on Forte) by adding casts or changing variable types
storage/ndb/src/kernel/blocks/dbtup/DbtupExecQuery.cpp:
Tdummy -> align (as in other part of cluster code)
storage/ndb/src/kernel/vm/DynArr256.cpp:
Removed not used variable
storage/ndb/src/ndbapi/Ndb.cpp:
Removed not used variable
strings/strtod.c:
Include ieeefp.h to avoid compiler warning
tests/bug25714.c:
Fixed compiler warning
tests/mysql_client_test.c:
Remove not used variable
Fixed indentation
Removed never reached code
Fixed compiler warning (on Forte) by adding casts or changing variable types
vio/viosocket.c:
Fixed compiler warning (on Forte) by adding casts or changing variable types
Faster thr_alarm()
Added 'Opened_files' status variable to track calls to my_open()
Don't give warnings when running mysql_install_db
Added option --source-install to mysql_install_db
I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems
index_read() -> index_read_map()
index_read_idx() -> index_read_idx_map()
index_read_last() -> index_read_last_map()
into a88-113-38-195.elisa-laajakaista.fi:/home/my/bk/mysql-5.1-marvel
BitKeeper/etc/ignore:
auto-union
client/mysql.cc:
Auto merged
client/mysqldump.c:
Auto merged
client/mysqltest.c:
Auto merged
extra/comp_err.c:
Auto merged
include/decimal.h:
Auto merged
include/my_getopt.h:
Auto merged
include/my_global.h:
Auto merged
include/my_sys.h:
Auto merged
include/mysql.h:
Auto merged
mysys/array.c:
Auto merged
mysys/hash.c:
Auto merged
mysys/typelib.c:
Auto merged
sql/derror.cc:
Auto merged
sql/event_data_objects.cc:
Auto merged
sql/event_queue.cc:
Auto merged
sql/field.cc:
Auto merged
sql/filesort.cc:
Auto merged
sql/ha_ndbcluster.h:
Auto merged
sql/ha_ndbcluster_binlog.cc:
Auto merged
sql/ha_partition.cc:
Auto merged
sql/ha_partition.h:
Auto merged
sql/handler.cc:
Auto merged
sql/handler.h:
Auto merged
sql/item.cc:
Auto merged
sql/item.h:
Auto merged
sql/item_cmpfunc.cc:
Auto merged
sql/item_func.cc:
Auto merged
sql/item_subselect.cc:
Auto merged
sql/item_sum.cc:
Auto merged
sql/item_timefunc.cc:
Auto merged
sql/item_timefunc.h:
Auto merged
sql/log.cc:
Auto merged
sql/log_event.cc:
Auto merged
sql/my_decimal.cc:
Auto merged
sql/my_decimal.h:
Auto merged
sql/mysql_priv.h:
Auto merged
sql/opt_range.cc:
Auto merged
sql/opt_range.h:
Auto merged
sql/opt_sum.cc:
Auto merged
sql/protocol.cc:
Auto merged
sql/protocol.h:
Auto merged
sql/rpl_utility.h:
Auto merged
sql/slave.cc:
Auto merged
sql/sp.cc:
Auto merged
sql/sp_head.cc:
Auto merged
sql/sp_head.h:
Auto merged
sql/sql_cache.cc:
Auto merged
sql/sql_class.cc:
Auto merged
sql/sql_class.h:
Auto merged
sql/sql_connect.cc:
Auto merged
sql/sql_delete.cc:
Auto merged
sql/sql_lex.cc:
Auto merged
sql/sql_lex.h:
Auto merged
sql/sql_load.cc:
Auto merged
sql/sql_parse.cc:
Auto merged
sql/sql_partition.cc:
Auto merged
sql/sql_prepare.cc:
Auto merged
sql/sql_repl.cc:
Auto merged
sql/sql_select.cc:
Auto merged
sql/sql_select.h:
Auto merged
sql/sql_show.cc:
Auto merged
sql/sql_trigger.cc:
Auto merged
sql/sql_union.cc:
Auto merged
sql/sql_update.cc:
Auto merged
sql/sql_view.cc:
Auto merged
sql/structs.h:
Auto merged
sql/table.h:
Auto merged
sql/tztime.cc:
Auto merged
sql/unireg.cc:
Auto merged
storage/example/ha_example.cc:
Auto merged
storage/federated/ha_federated.cc:
Auto merged
storage/heap/ha_heap.cc:
Auto merged
storage/innobase/handler/ha_innodb.h:
Auto merged
storage/myisam/ha_myisam.cc:
Auto merged
storage/myisam/sort.c:
Auto merged
storage/myisammrg/ha_myisammrg.cc:
Auto merged
storage/ndb/tools/restore/consumer_restore.cpp:
Auto merged
strings/decimal.c:
Auto merged
strings/strtod.c:
Auto merged
include/hash.h:
Manual merge with 5.1 main tree.
mysys/my_getopt.c:
Manual merge with 5.1 main tree.
sql/field.h:
Manual merge with 5.1 main tree.
sql/ha_ndbcluster.cc:
Manual merge with 5.1 main tree.
sql/item_cmpfunc.h:
Manual merge with 5.1 main tree.
sql/item_create.cc:
Manual merge with 5.1 main tree.
sql/item_func.h:
Manual merge with 5.1 main tree.
sql/key.cc:
Manual merge with 5.1 main tree.
sql/lock.cc:
Manual merge with 5.1 main tree.
sql/mysqld.cc:
Manual merge with 5.1 main tree.
sql/set_var.cc:
Manual merge with 5.1 main tree.
sql/set_var.h:
Manual merge with 5.1 main tree.
sql/sql_base.cc:
Manual merge with 5.1 main tree.
sql/sql_handler.cc:
Manual merge with 5.1 main tree.
sql/sql_insert.cc:
Manual merge with 5.1 main tree.
sql/sql_plugin.cc:
Manual merge with 5.1 main tree.
sql/sql_table.cc:
Manual merge with 5.1 main tree.
sql/sql_yacc.yy:
Manual merge with 5.1 main tree.
sql/table.cc:
Manual merge with 5.1 main tree.
storage/innobase/handler/ha_innodb.cc:
Manual merge with 5.1 main tree.
storage/ndb/src/mgmsrv/InitConfigFileParser.cpp:
Manual merge with 5.1 main tree.
storage/ndb/tools/restore/restore_main.cpp:
Manual merge with 5.1 main tree.
Bug #20662 "Infinite loop in CREATE TABLE IF NOT EXISTS ... SELECT
with locked tables"
Bug #20903 "Crash when using CREATE TABLE .. SELECT and triggers"
Bug #24738 "CREATE TABLE ... SELECT is not isolated properly"
Bug #24508 "Inconsistent results of CREATE TABLE ... SELECT when
temporary table exists"
Deadlock occured when one tried to execute CREATE TABLE IF NOT
EXISTS ... SELECT statement under LOCK TABLES which held
read lock on target table.
Attempt to execute the same statement for already existing
target table with triggers caused server crashes.
Also concurrent execution of CREATE TABLE ... SELECT statement
and other statements involving target table suffered from
various races (some of which might've led to deadlocks).
Finally, attempt to execute CREATE TABLE ... SELECT in case
when a temporary table with same name was already present
led to the insertion of data into this temporary table and
creation of empty non-temporary table.
All above problems stemmed from the old implementation of CREATE
TABLE ... SELECT in which we created, opened and locked target
table without any special protection in a separate step and not
with the rest of tables used by this statement.
This underminded deadlock-avoidance approach used in server
and created window for races. It also excluded target table
from prelocking causing problems with trigger execution.
The patch solves these problems by implementing new approach to
handling of CREATE TABLE ... SELECT for base tables.
We try to open and lock table to be created at the same time as
the rest of tables used by this statement. If such table does not
exist at this moment we create and place in the table cache special
placeholder for it which prevents its creation or any other usage
by other threads.
We still use old approach for creation of temporary tables.
Note that we have separate fix for 5.0 since there we use slightly
different less intrusive approach.
mysql-test/r/create.result:
Extended test coverage for CREATE TABLE ... SELECT. In particular added
tests for bug #24508 "Inconsistent results of CREATE TABLE ... SELECT
when temporary table exists" and bug #20662 "Infinite loop in CREATE
TABLE IF NOT EXISTS ... SELECT with locked tables".
mysql-test/r/trigger.result:
Added test case for bug #20903 "Crash when using CREATE TABLE .. SELECT
and triggers"
mysql-test/t/create.test:
Extended test coverage for CREATE TABLE ... SELECT. In particular added
tests for bug #24508 "Inconsistent results of CREATE TABLE ... SELECT
when temporary table exists" and bug #20662 "Infinite loop in CREATE
TABLE IF NOT EXISTS ... SELECT with locked tables".
mysql-test/t/trigger.test:
Added test case for bug #20903 "Crash when using CREATE TABLE .. SELECT
and triggers"
sql/lock.cc:
Now for creation of name-lock placeholder lock_table_name() uses
auxiliary function table_cache_insert_placeholder().
sql/mysql_priv.h:
Removed declaration of non-existing build_table_path() routine.
The former mysql_create_table_internal() was renamed to
mysql_create_table_no_lock() and now exposed to other modules to
give them opportunity of creation of tables in cases when name-lock
is already obtained.
reopen_name_locked_table() now has 3rd argument which controls linking
in of table being opened into THD::open_tables (this is useful in
cases when placeholder used for name-locking is already linked into
this list).
Added declaration of auxiliary function table_cache_insert_placeholder()
which is used for creation of table placeholders for name-locking.
Added declaration of lock_table_name_if_not_cached() which can be
used to take an exclusive name-lock on table if there are no records
for it in table cache.
Changed signature of unlink_open_table() function to simplify its use
and make it useful for table placeholders and tables that are only open.
Added auxiliary drop_open_table() routine.
Moved declaration of refresh_version to table.h header to make it
accessible from inline methods of TABLE class.
MYSQL_OPEN_IGNORE_LOCKED_TABLES flag is no longer used. Instead
MYSQL_OPEN_TEMPORARY_ONLY option was added.
sql/sql_base.cc:
Added support for the new approach to the handling of CREATE TABLE
... SELECT for base tables.
Now we try to open and lock table to be created at the same time as
the rest of tables used by this statement. If such table does not
exist at this moment we create and place in the table cache special
placeholder for it which prevents its creation or any other usage
by other threads.
Note significant distinctions of this placeholder from the placeholder
used for normal name-lock: 1) It is treated like open table by other
name-locks so it does not allow name-lock taking operations like DROP
TABLE or RENAME TABLE to proceed. 2) it is linked into THD::open_tables
list and automatically removed during close_thread_tables() call
open_tables():
Implemented logic described above. To do this added
auxiliary check_if_table_exists() function.
Removed support for MYSQL_OPEN_IGNORE_LOCKED_TABLES option
which is no longer used.
Added MYSQL_OPEN_TEMPORARY_ONLY which is used to restrict
search for temporary tables only.
close_cached_tables()/close_thread_table()/reopen_tables()/
close_old_data_files()/table_is_used()/remove_table_from_cache():
Added support for open placeholders (note that we also use them
when we need to re-open tables during flush).
unlink_open_table():
Changed function signature to simplify its use and to make
useful for open placeholders and tables which are only
open and not locked.
Added auxiliary drop_open_table() routine.
reopen_name_locked_table():
Now has 3rd argument which controls linking in of table being
opened into THD::open_tables (this is useful in cases when
placeholder used for name-locking is already linked into
this list).
Added auxiliary table_cache_insert_placeholder() routine which
simplifies creation of placeholders used for name-locking.
Added lock_table_name_if_not_cached() which can be used to take
an exclusive name-lock on table if there are no records for it
in table cache.
sql/sql_handler.cc:
Adjusted mysql_ha_mark_tables_for_reopen() routine to properly
handle placeholders which now can be linked into open tables
list.
sql/sql_insert.cc:
Introduced new approach to handling of base tables in CREATE TABLE
... SELECT statement.
Now we try to open and lock table to be created at the same time as
the rest of tables used by this statement. If such table does not
exist at this moment we create and place in the table cache special
placeholder for it which prevents its creation or any other usage
by other threads. By doing this we avoid races which existed with
previous approach in which we created, opened and locked target in
separate step without any special protection.
This also allows properly calculate prelocking set in cases when
target table already exists and has some on insert triggers.
Note that we don't employ the same approach for temporary tables
(this is okay as such tables are unaffected by other threads).
Changed create_table_from_items() and methods of select_create
class to implement this approach.
sql/sql_parse.cc:
The new approach to handling of CREATE TABLE ... SELECT for
base tables assumes that all tables (including table to be
created) are opened and (or) locked at the same time.
So in cases when we create base table we have to pass to
open_and_lock_tables() table list which includes target table.
sql/sql_prepare.cc:
The new approach to handling of CREATE TABLE ... SELECT for
base tables assumes that all tables (including table to be
created) are opened and (or) locked at the same time.
So in cases when we create base table we have to pass to
open_and_lock_tables() table list which includes target table.
sql/sql_table.cc:
Changed mysql_create_table(), mysql_create_like_table() and
mysql_alter_table() (in rename case) to obtain exclusive name-lock
on the non-temporary table which is going to be created (to which
we going to rename). This ensures that not only destination table
doesn't exist on disk but also that there are no placeholder in
table cache for it (i.e. there is no CREATE TABLE ... SELECT operation
in progress for it). Note that to avoid deadlocks while taking these
name-locks this code assumes that existence of any record for table in
table cache (even name-lock) means that table exists. Altough such
check can lead to false positives these should occur only in case of
highly concurrent DDL operations on the table and should not break
binary logging.
Renamed mysql_create_table_internal() to mysql_create_table_no_lock()
and made it accessible from other files to give them ability to create
table in situation when name-lock is already obtained or not relevant.
Adjusted calls to reopen_name_locked_table(), which now takes
extra argument, which controls linking of open table into
THD::open_tables list.
Removed redundant setting of table's 'version' field before calls
to close_cached_table(). This function will set it to 0 itself
anyway.
sql/sql_trigger.cc:
reopen_name_locked_tables() now has one more argument which controls
linking of opened table into the THD::open_tables list.
sql/sql_yacc.yy:
The new approach to handling of CREATE TABLE ... SELECT statement
for base tables assumes that all tables including table to be
created are open and (or) locked at the same time. Therefore
we need to set correct lock for target table.
sql/table.h:
Moved declaration of refresh_version variable from mysql_priv.h
to make it accessible from inline methods of TABLE class.
Renamed TABLE::locked_by_flush member to open_placeholder since
now it is also used for taking exclusive name-lock and not only
by flush.
Introduced TABLE::is_name_opened() helper method which can be used
to distinguish TABLE instances corresponding to open tables or
placeholders for them from closed instances (e.g. due to their old
version). Also introduced TABLE::needs_reopen_or_name_lock() helper
which allows to check if TABLE instance corresponds to outdated
version of table or to name-lock placeholder.
Introduced TABLE_LIST::create member which marks elements of
table list corresponds to the table to be created.
Adjusted TABLE_LIST::placeholder() method to take into account
name-lock placeholders for tables to be created (this, for example,
allows to properly handle such placeholders in lock_tables()).
Finally, moved currently unused TABLE::open_next/open_prev
members under ifdef NOT_YET.
mysql-test/r/create_select-big.result:
New BitKeeper file ``mysql-test/r/create_select-big.result''
mysql-test/t/create_select-big.test:
New BitKeeper file ``mysql-test/t/create_select-big.test''
Bug #20662 "Infinite loop in CREATE TABLE IF NOT EXISTS ... SELECT
with locked tables"
Bug #20903 "Crash when using CREATE TABLE .. SELECT and triggers"
Bug #24738 "CREATE TABLE ... SELECT is not isolated properly"
Bug #24508 "Inconsistent results of CREATE TABLE ... SELECT when
temporary table exists"
Deadlock occured when one tried to execute CREATE TABLE IF NOT
EXISTS ... SELECT statement under LOCK TABLES which held
read lock on target table.
Attempt to execute the same statement for already existing
target table with triggers caused server crashes.
Also concurrent execution of CREATE TABLE ... SELECT statement
and other statements involving target table suffered from
various races (some of which might've led to deadlocks).
Finally, attempt to execute CREATE TABLE ... SELECT in case
when a temporary table with same name was already present
led to the insertion of data into this temporary table and
creation of empty non-temporary table.
All above problems stemmed from the old implementation of CREATE
TABLE ... SELECT in which we created, opened and locked target
table without any special protection in a separate step and not
with the rest of tables used by this statement.
This underminded deadlock-avoidance approach used in server
and created window for races. It also excluded target table
from prelocking causing problems with trigger execution.
The patch solves these problems by implementing new approach to
handling of CREATE TABLE ... SELECT for base tables.
We try to open and lock table to be created at the same time as
the rest of tables used by this statement. If such table does not
exist at this moment we create and place in the table cache special
placeholder for it which prevents its creation or any other usage
by other threads.
We still use old approach for creation of temporary tables.
Note that we have separate fix for 5.0 since there we use slightly
different less intrusive approach.
Bug #20662 "Infinite loop in CREATE TABLE IF NOT EXISTS ... SELECT
with locked tables"
Bug #20903 "Crash when using CREATE TABLE .. SELECT and triggers"
Bug #24738 "CREATE TABLE ... SELECT is not isolated properly"
Bug #24508 "Inconsistent results of CREATE TABLE ... SELECT when
temporary table exists"
Deadlock occured when one tried to execute CREATE TABLE IF NOT
EXISTS ... SELECT statement under LOCK TABLES which held
read lock on target table.
Attempt to execute the same statement for already existing
target table with triggers caused server crashes.
Also concurrent execution of CREATE TABLE ... SELECT statement
and other statements involving target table suffered from
various races (some of which might've led to deadlocks).
Finally, attempt to execute CREATE TABLE ... SELECT in case
when a temporary table with same name was already present
led to the insertion of data into this temporary table and
creation of empty non-temporary table.
All above problems stemmed from the old implementation of CREATE
TABLE ... SELECT in which we created, opened and locked target
table without any special protection in a separate step and not
with the rest of tables used by this statement.
This underminded deadlock-avoidance approach used in server
and created window for races. It also excluded target table
from prelocking causing problems with trigger execution.
The patch solves these problems by implementing new approach to
handling of CREATE TABLE ... SELECT for base tables.
We try to open and lock table to be created at the same time as
the rest of tables used by this statement. If such table does not
exist at this moment we create and place in the table cache special
placeholder for it which prevents its creation or any other usage
by other threads.
We still use old approach for creation of temporary tables.
Also note that we decided to postpone introduction of some tests
for concurrent behaviour of CREATE TABLE ... SELECT till 5.1.
The main reason for this is absence in 5.0 ability to set @@debug
variable at runtime, which can be circumvented only by using several
test files with individual .opt files. Since the latter is likely
to slowdown test-suite unnecessary we chose not to push this tests
into 5.0, but run them manually for this version and later push
their optimized version into 5.1
mysql-test/r/create.result:
Extended test coverage for CREATE TABLE ... SELECT. In particular added
tests for bug #24508 "Inconsistent results of CREATE TABLE ... SELECT
when temporary table exists" and bug #20662 "Infinite loop in CREATE
TABLE IF NOT EXISTS ... SELECT with locked tables".
mysql-test/r/trigger.result:
Added test case for bug #20903 "Crash when using CREATE TABLE .. SELECT
and triggers"
mysql-test/t/create.test:
Extended test coverage for CREATE TABLE ... SELECT. In particular added
tests for bug #24508 "Inconsistent results of CREATE TABLE ... SELECT
when temporary table exists" and bug #20662 "Infinite loop in CREATE
TABLE IF NOT EXISTS ... SELECT with locked tables".
mysql-test/t/trigger.test:
Added test case for bug #20903 "Crash when using CREATE TABLE .. SELECT
and triggers"
sql/lock.cc:
Now for creation of name-lock placeholder in lock_table_name() we use
auxiliary function table_cache_insert_placeholder().
sql/mysql_priv.h:
Made build_table_path() function available outside of sql_table.cc file.
reopen_name_locked_table() now has 3rd argument which controls linking
in of table being opened into THD::open_tables (this is useful in
cases when placeholder used for name-locking is already linked into
this list).
Added declaration of auxiliary function table_cache_insert_placeholder()
which is used for creation of table placeholders for name-locking.
Added declaration of table_cache_has_open_placeholder() function which
can be used for checking if table cache contains an open placeholder for
the table and if this placeholder was created by another thread.
(This function is needed only in 5.0 where we use it in various versions
of CREATE TABLE in order to protect it from concurrent CREATE TABLE
... SELECT operations for the table. Starting from 5.1 we use different
approach so it is going to be removed there).
Made close_old_data_files() static within sql_base.cc file.
Added auxiliary drop_open_table() routine.
Moved declaration of refresh_version to table.h header to make it
accessible from inline methods of TABLE class.
MYSQL_OPEN_IGNORE_LOCKED_TABLES flag is no longer used. Instead
MYSQL_OPEN_TEMPORARY_ONLY option was added.
sql/sql_base.cc:
Added support for the new approach to the handling of CREATE TABLE
... SELECT for base tables.
Now we try to open and lock table to be created at the same time as
the rest of tables used by this statement. If such table does not
exist at this moment we create and place in the table cache special
placeholder for it which prevents its creation or any other usage
by other threads.
Note significant distinctions of this placeholder from the placeholder
used for normal name-lock: 1) It is treated like open table by other
name-locks so it does not allow name-lock taking operations like DROP
TABLE or RENAME TABLE to proceed. 2) it is linked into THD::open_tables
list and automatically removed during close_thread_tables() call.
open_tables():
Implemented logic described above. To do this added
auxiliary check_if_table_exists() function.
Removed support for MYSQL_OPEN_IGNORE_LOCKED_TABLES option
which is no longer used.
Added MYSQL_OPEN_TEMPORARY_ONLY which is used to restrict
search for temporary tables only.
close_cached_tables()/close_thread_table()/reopen_tables()/
close_old_data_files()/table_is_used()/remove_table_from_cache():
Added support for open placeholders (note that we also use them
when we need to re-open tables during flush).
Added auxiliary drop_open_table() routine.
reopen_name_locked_table():
Now has 3rd argument which controls linking in of table being
opened into THD::open_tables (this is useful in cases when
placeholder used for name-locking is already linked into
this list).
Added auxiliary table_cache_insert_placeholder() routine which
simplifies creation of placeholders used for name-locking.
Added table_cache_has_open_placeholder() function which can be
used for checking if table cache contains an open placeholder for
the table and if this placeholder was created by another thread.
(This function is needed only in 5.0 where we use it in various versions
of CREATE TABLE in order to protect it from concurrent CREATE TABLE
... SELECT operations for the table. Starting from 5.1 we use different
approach so it is going to be removed there).
sql/sql_handler.cc:
Adjusted mysql_ha_mark_tables_for_reopen() routine to properly
handle placeholders which now can be linked into open tables
list.
sql/sql_insert.cc:
Introduced new approach to handling of base tables in CREATE TABLE
... SELECT statement.
Now we try to open and lock table to be created at the same time as
the rest of tables used by this statement. If such table does not
exist at this moment we create and place in the table cache special
placeholder for it which prevents its creation or any other usage
by other threads. By doing this we avoid races which existed with
previous approach in which we created, opened and locked target in
separate step without any special protection.
This also allows properly calculate prelocking set in cases when
target table already exists and has some on insert triggers.
Note that we don't employ the same approach for temporary tables
(this is okay as such tables are unaffected by other threads).
Changed create_table_from_items() and select_create methods to
implement this approach.
sql/sql_parse.cc:
The new approach to handling of CREATE TABLE ... SELECT for
base tables assumes that all tables (including table to be
created) are opened and (or) locked at the same time.
So in cases when we create base table we have to pass to
open_and_lock_tables() table list which includes target table.
sql/sql_prepare.cc:
The new approach to handling of CREATE TABLE ... SELECT for
base tables assumes that all tables (including table to be
created) are opened and (or) locked at the same time.
So in cases when we create base table we have to pass to
open_and_lock_tables() table list which includes target table.
sql/sql_table.cc:
Now mysql_create_table_internal(), mysql_create_like_table() and
mysql_alter_table() not only check that destination table doesn't
exist on disk but also check that there is no create placeholder
in table cache for it (i.e. there is no CREATE TABLE ... SELECT
operation in progress for it). Note that starting from 5.1 we
use different approach in order to to protect CREATE TABLE ... SELECT
from concurrent CREATE TABLE (ALTER TABLE ... RENAME) operations,
the latter simply take name-locks on table before its creation
(on target table name before renaming).
Also made build_table_path() available from other files and
asjusted calls to reopen_name_locked_table(), which now takes
extra argument, which controls linking of open table into
THD::open_tables list.
sql/sql_trigger.cc:
reopen_name_locked_tables() now has one more argument which controls
linking of opened table into the THD::open_tables list.
sql/sql_yacc.yy:
The new approach to handling of CREATE TABLE ... SELECT statement
for base tables assumes that all tables including table to be
created are open and (or) locked at the same time. Therefore
we need to set correct lock for target table.
sql/table.h:
Moved declaration of refresh_version variable from mysql_priv.h
to make it accessible from inline methods of TABLE class.
Renamed TABLE::locked_by_flush member to open_placeholder since
now it is also used for taking exclusive name-lock and not only
by flush.
Introduced TABLE::is_name_opened() helper method which can be used
to distinguish TABLE instances corresponding to open tables or
placeholders for them from closed instances (e.g. due to their old
version). Also introduced TABLE::needs_reopen_or_name_lock() helper
which allows to check if TABLE instance corresponds to outdated
version of table or to name-lock placeholder.
Introduced TABLE_LIST::create member which marks elements of
table list corresponds to the table to be created.
Adjusted TABLE_LIST::placeholder() method to take into account
name-lock placeholders for tables to be created (this, for example,
allows to properly handle such placeholders in lock_tables()).
