statement being KILLed".
When statement which was trying to obtain write lock on then table and
which was blocked by existing read lock was killed, concurrent statements
that were trying to obtain read locks on the same table and that were
blocked by the presence of this pending write lock were not woken up and
had to wait until this first read lock goes away.
This problem was caused by the fact that we forgot to wake up threads
which pending requests could have been satisfied after removing lock
request for the killed thread.
The patch solves the problem by waking up those threads in such situation.
Test for this bug will be added to 5.1 only as it has much better
facilities for its implementation. Particularly, by using I_S.PROCESSLIST
and wait_condition.inc script we can wait until thread will be blocked on
certain table lock without relying on unconditional sleep (which usage
increases time needed for test runs and might cause spurious test
failures on slower platforms).
New server (as of 5.1.21) provides new features:
- SHOW CREATE TRIGGER;
- character set information for SHOW TRIGGERS and SHOW CREATE
EVENT | FUNCTION | PROCEDURE statements.
Mysqldump uses these features to generate proper dump.
The bug happened when new mysqldump was used to dump older servers.
The problem was that 5.1.21 new features are not available, so
mysqldump exited with error code or just crashed.
The fix is to detect if mysqldump has ben run against older server
and don't use new 5.1.21 functionality in this case. Certainly,
the dump generated for the older server suffers from the character
set problems fixed by BUG#16291 and the like.
When a table was explicitly locked with LOCK TABLES no associated
tables from any related trigger on the subject table were locked.
As a result of this the user could experience unexpected locking
behavior and statement failures similar to "failed: 1100: Table'xx'
was not locked with LOCK TABLES".
This patch fixes this problem by making sure triggers are
pre-loaded on any statement if the subject table was explicitly
locked with LOCK TABLES.
mysqldump generates view defitions in two stages:
- dump CREATE TABLE statements for the temporary tables. For each view a
temporary table, that has the same structure as the view is created.
- dump DROP TABLE statements for the temporary tables and CREATE VIEW
statements for the view.
This approach is required because views can have dependencies on each other
(a view can use other views). So, they should be created in the particular
order. mysqldump however is not smart enough, so in order to resolve
dependencies it creates temporary tables first of all.
The problem was that mysqldump might have generated incorrect dump for the
temporary table when a view has non-ASCII column name. That happened when
default-character-set is not utf8.
The fix is to:
1. Switch character_set_client for the mysqldump's connection to binary
before issuing SHOW FIELDS statement in order to avoid conversion.
2. Dump switch character_set_client statements to UTF8 and back for
CREATE TABLE statement that is issued to create temporary table.
The problem was a race condition on shutdown -- when IM got shutdown
request while a guarded mysqld is starting. In this case the Guardian
thread tried to stop the mysqld, but might fail if the mysqld hadn't
created pid-file so far. When this happened, the mysqld-monitor thread
didn't stop, so the assert in Thread_registry happened.
The fix is to make several attempts to stop mysqld if it is active.
between perm and temp tables. Review fixes.
The original bug report complains that if we locked a temporary table
with LOCK TABLES statement, we would not leave LOCK TABLES mode
when this temporary table is dropped.
Additionally, the bug was escalated when it was discovered than
when a temporary transactional table that was previously
locked with LOCK TABLES statement was dropped, futher actions with
this table, such as UNLOCK TABLES, would lead to a crash.
The problem originates from incomplete support of transactional temporary
tables. When we added calls to handler::store_lock()/handler::external_lock()
to operations that work with such tables, we only covered the normal
server code flow and did not cover LOCK TABLES mode.
In LOCK TABLES mode, ::external_lock(LOCK) would sometimes be called without
matching ::external_lock(UNLOCK), e.g. when a transactional temporary table
was dropped. Additionally, this table would be left in the list of LOCKed
TABLES.
The patch aims to address this inadequacy. Now, whenever an instance
of 'handler' is destroyed, we assert that it was priorly
external_lock(UNLOCK)-ed. All the places that violate this assert
were fixed.
This patch introduces no changes in behavior -- the discrepancy in
behavior will be fixed when we start calling ::store_lock()/::external_lock()
for all tables, regardless whether they are transactional or not,
temporary or not.
INSERT/DELETE/UPDATE followed by ALTER TABLE within LOCK TABLES
may cause table corruption on Windows.
That happens because ALTER TABLE writes outdated shared state
info into index file.
Fixed by removing obsolete workaround.
Affects MyISAM tables on Windows only.
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
5.1.20 -> 5.1.21 upgrades.
We generate mysql_fix_privilege.sql file, which contains SQL
statements required to upgrade the system database. This script
is generated by concatenation of mysql_system_tables.sql and
mysql_system_tables_fix.sql.
The problem was that
- in order to create general_log and slow_log tables we use
stored programs in mysql_system_tables.sql;
- we upgrade mysql.proc table in mysql_system_tables_fix.sql;
So, if mysql.proc table needs to be upgraded, stored procedures
can not be used in mysql_system_tables.sql.
In other words, in mysql_system_tables.sql stored programs must
not be used because they may be unavailable at this point.
The fix is to use dynamic SQL instead of stored programs.
There is no test case for this bug because our test suite
is not suitable for such test cases. system_mysql_db_fix* test
cases play with the database "test". Here we need to modify
the system database and we can not do that in the test suite.
