server to crash".
Crash caused by assertion failure happened when one ran SHOW OPEN TABLES
while concurrently doing DROP TABLE (or RENAME TABLE, CREATE TABLE LIKE
or any other command that takes name-lock) in other connection.
For non-debug version of server problem exposed itself as wrong output
of SHOW OPEN TABLES statement (it was missing name-locked tables).
Finally in 5.1 both debug and non-debug versions simply crashed in
this situation due to NULL-pointer dereference.
This problem was caused by the fact that table placeholders which were
added to table cache in order to obtain name-lock had TABLE_SHARE::table_name
set to 0. Therefore they broke assumption that this member is non-0 for
all tables in table cache which was checked by assert in list_open_tables()
(in 5.1 this function simply relies on it).
The fix simply sets this member for such placeholders to appropriate value
making this assumption true again.
This patch also includes test for similar bug 12212 "Crash that happens
during removing of database name from cache" reappeared in 5.1 as bug 19403.
Addendum fixes after changing the condition variable
for the global read lock.
The stress test suite revealed some deadlocks. Some were
related to the new condition variable (COND_global_read_lock)
and some were general problems with the global read lock.
It is now necessary to signal COND_global_read_lock whenever
COND_refresh is signalled.
We need to wait for the release of a global read lock if one
is set before every operation that requires a write lock.
But we must not wait if we have locked tables by LOCK TABLES.
After setting a global read lock a thread waits until all
write locks are released.
Replaced COND_refresh with COND_global_read_lock becasue of a bug in NTPL threads when using different mutexes as arguments to pthread_cond_wait()
The original code caused a hang in FLUSH TABLES WITH READ LOCK in some circumstances because pthread_cond_broadcast() was not delivered to other threads.
This fixes:
Bug#16986: Deadlock condition with MyISAM tables
Bug#20048: FLUSH TABLES WITH READ LOCK causes a deadlock
After a locking error the open table(s) were not fully
cleaned up for reuse. But they were put into the open table
cache even before the lock was tried. The next statement
reused the table(s) with a wrong lock type set up. This
tricked MyISAM into believing that it don't need to update
the table statistics. Hence CHECK TABLE reported a mismatch
of record count and table size.
Fortunately nothing worse has been detected yet. The effect
of the test case was that the insert worked on a read locked
table. (!)
I added a new function that clears the lock type from all
tables that were prepared for a lock. I call this function
when a lock failes.
No test case. One test would add 50 seconds to the
test suite. Another test requires file mode modifications.
I added a test script to the bug report. It contains three
cases for failing locks. All could reproduce a table
corruption. All are fixed by this patch.
This bug was not lock timeout specific.
Additional fix for INSERT DELAYED with subselect.
Originally detected in 5.1, but 5.0 could also be affected.
The user thread creates a dummy table object,
which is not included in the lock. The 'real' table is
opened and locked by the 'delayed' system thread.
The dummy object is now marked as not locked and this is
tested in mysql_lock_have_duplicate().
when high concurrency": remove HASH::current_record and make it
an external search parameter, so that it can not be the cause of a
race condition under high concurrent load.
The bug was in a race condition in table_hash_search,
when column_priv_hash.current_record was overwritten simultaneously
by multiple threads, causing the search for a suitable grant record
to fail.
No test case as the bug is repeatable only under concurrent load.
Problem #1: INSERT...SELECT, Version for 5.0.
Extended the unique table check by a check of lock data.
Merge sub-tables cannot be detected by doing name checks only.
Problem #1: INSERT...SELECT, Version for 4.1.
INSERT ... SELECT with the same table on both sides (hidden
below a MERGE table) does now work by buffering the select result.
The duplicate detection works now after open_and_lock_tables()
on the locks.
I did not find a test case that failed without the change in
sql_update.cc. I made the change anyway as it should in theory
fix a possible MERGE table problem with multi-table update.
Problem #1: INSERT...SELECT
INSERT ... SELECT with the same table on both sides (hidden
below a MERGE table) does now work by buffering the select result.
The duplicate detection works now after open_and_lock_tables()
on the locks.
I did not find a test case that failed without the change in
sql_update.cc. I made the change anyway as it should in theory
fix a possible MERGE table problem with multi-table update.
Version for 5.0.
It fixes three problems:
1. The cause of the bug was that we did not check the table version for
the HANDLER ... READ commands. We did not notice when a table was
replaced by a new one. This can happen during ALTER TABLE, REPAIR
TABLE, and OPTIMIZE TABLE (there might be more cases). I call the fix
for this problem "the primary bug fix".
2. mysql_ha_flush() was not always called with a locked LOCK_open.
Though the function comment clearly said it must.
I changed the code so that the locking is done when required. I call
the fix for this problem "the secondary fix".
3. In 5.0 (not in 4.1 or 4.0) DROP TABLE had a possible deadlock flaw in
concur with FLUSH TABLES WITH READ LOCK. I call the fix for this
problem "the 5.0 addendum fix".
This bug occurs when some trigger for table used by DML statement is created
or changed while statement was waiting in lock_tables(). In this situation
prelocking set which we have calculated becames invalid which can easily lead
to errors and even in some cases to crashes.
With proposed patch we no longer silently reopen tables in lock_tables(),
instead caller of lock_tables() becomes responsible for reopening tables and
recalculation of prelocking set.
Bug was introduced by cset 1.1659.14.1. Before it server was silently
ignoring that lock can't be acquired because it already acquired.
This patch makes make_global_read_lock_block_commit() return without error
if lock already acquired.
Fixed portability problem with bool in C programs
Moved close_thread_tables out from LOCK_thread_count mutex (safety fix)
my_sleep() -> pthread_cond_timedwait()
The idea of the patch
is that every cursor gets its own lock id for table level locking.
Thus cursors are protected from updates performed within the same
connection. Additionally a list of transient (must be closed at
commit) cursors is maintained and all transient cursors are closed
when necessary. Lastly, this patch adds support for deadlock
timeouts to TLL locking when using cursors.
+ post-review fixes.
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".
1.) Added a new option to mysql_lock_tables() for ignoring FLUSH TABLES.
Used the new option in create_table_from_items().
It is necessary to prevent the SELECT table from being reopend.
It would get new storage assigned for its fields, while the
SELECT part of the command would still use the old (freed) storage.
2.) Protected the CREATE TABLE and CREATE TABLE ... SELECT commands
against a global read lock. This prevents a deadlock in
CREATE TABLE ... SELECT in conjunction with FLUSH TABLES WITH READ LOCK
and avoids the creation of new tables during a global read lock.
3.) Replaced set_protect_against_global_read_lock() and
unset_protect_against_global_read_lock() by
wait_if_global_read_lock() and start_waiting_global_read_lock()
in the INSERT DELAYED handling.
Added protection against global read lock while creating and
initializing a delayed insert handler.
Allowed to ignore a global read lock when locking the table
inside the delayed insert handler.
Added some minor improvements.
(otherwise a deadlock when ALTER writes to
binlog holding LOCK_open, it causes binlog rotation,
binlog waits for prepared transactions to commit, and commit
needs LOCK_open to check for global read lock)
start_waiting_global_read_lock() should wake up all those who are waiting
for protect_against_global_read_lock to go down to 0: those registered in waiting_for_read_lock
AND those registered in global_read_lock_blocks_commit.
Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon)
Created Field::make_field() and made Field_num::make_field() to call this
Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead
Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer.
Renamed TABLE_SHARE->real_name to table_name
Renamed TABLE->table_name to alias
Renamed TABLE_LIST->real_name to table_name
