The problem is that some DDL statements (ALTER TABLE, CREATE
TRIGGER, FLUSH TABLES, ...) when under LOCK TABLES need to
momentarily drop the lock, reopen the table and grab the write
lock again (using reopen_tables). When grabbing the lock again,
reopen_tables doesn't pass a flag to mysql_lock_tables in
order to ignore the impending global read lock, which causes a
assertion because LOCK_open is being hold. Also dropping the
lock must not signal to any threads that the table has been
relinquished (related to the locking/flushing protocol).
The solution is to correct the way the table is reopenned
and the locks grabbed. When reopening the table and under
LOCK TABLES, the table version should be set to 0 so other
threads have to wait for the table. When grabbing the lock,
any other flush should be ignored because it's theoretically
a atomic operation. The chosen solution also fixes a potential
discrepancy between binlog and GRL (global read lock) because
table placeholders were being ignored, now a FLUSH TABLES WITH
READ LOCK will properly for table with open placeholders.
It's also important to mention that this patch doesn't fix
a potential deadlock if one uses two GRLs under LOCK TABLES
concurrently.
cause ROLLBACK of statement", part 1. Review fixes.
Do not send OK/EOF packets to the client until we reached the end of
the current statement.
This is a consolidation, to keep the functionality that is shared by all
SQL statements in one place in the server.
Currently this functionality includes:
- close_thread_tables()
- log_slow_statement().
After this patch and the subsequent patch for Bug#12713, it shall also include:
- ha_autocommit_or_rollback()
- net_end_statement()
- query_cache_end_of_result().
In future it may also include:
- mysql_reset_thd_for_next_command().
The patch for Bug 26379 (Combination of FLUSH TABLE and
REPAIR TABLE corrupts a MERGE table) fixed this bug too.
However it revealed a new bug that crashed the server.
Flushing a merge table at the moment when it is between open
and attach of children crashed the server.
The flushing thread wants to abort locks on the flushed table.
It calls ha_myisammrg::lock_count() and ha_myisammrg::store_lock()
on the TABLE object of the other thread.
Changed ha_myisammrg::lock_count() and ha_myisammrg::store_lock()
to accept non-attached children. ha_myisammrg::lock_count() returns
the number of MyISAM tables in the MERGE table so that the memory
allocation done by get_lock_data() is done correctly, even if the
children become attached before ha_myisammrg::store_lock() is
called. ha_myisammrg::store_lock() will not return any lock if the
children are not attached.
This is however a change in the handler interface. lock_count()
can now return a higher number than store_lock() stores locks.
This is more safe than the reverse implementation would be.
get_lock_data() in the SQL layer is adjusted accordingly. It sets
MYSQL_LOCK::lock_count based on the number of locks returned by
the handler::store_lock() calls, not based on the numbers returned
by the handler::lock_count() calls. The latter are only used for
allocation of memory now.
No test case. The test suite cannot reliably run FLUSH between
lock_count() and store_lock() of another thread. The bug report
contains a program that can repeat the problem with some
probability.
