Making FLUSH TABLES WITH READ LOCK block COMMITs of existing transactions,

in a deadlock-free manner. This splits locking the global read lock in two steps.
This fixes a consequence of this bug, known as:
BUG#4953 'mysqldump --master-data may report incorrect binlog position if using InnoDB'
And a test.


sql/handler.cc:
  making COMMIT wait if FLUSH TABLES WITH READ LOCK happened.
sql/lock.cc:
  an additional stage so that FLUSH TABLES WITH READ LOCK blocks COMMIT:
  make_global_read_lock_block_commit():
  taking the global read lock is TWO steps (2nd step is optional; without
  it, COMMIT of existing transactions will be allowed):
  lock_global_read_lock() THEN make_global_read_lock_block_commit().
sql/mysql_priv.h:
  new argument to wait_if_global_read_lock()
sql/sql_class.h:
  THD::global_read_lock now an uint to reflect the 2 steps of global read lock (does not block COMMIT / does)
sql/sql_db.cc:
  update for new prototype
sql/sql_parse.cc:
  implementing the two steps of global read lock so that FLUSH TABLES WITH READ LOCK can block COMMIT without deadlocking with COMMITs.

1 files changed, 23 insertions, 0 deletions

diff --git a/mysql-test/r/flush_block_commit.result b/mysql-test/r/flush_block_commit.result
new file mode 100644
index 00000000000..3205dd9dad9
--- /dev/null
+++ b/mysql-test/r/flush_block_commit.result
@@ -0,0 +1,23 @@
+drop table if exists t1;
+create table t1 (a int) type=innodb;
+begin;
+insert into t1 values(1);
+flush tables with read lock;
+select * from t1;
+a
+ commit;
+select * from t1;
+a
+unlock tables;
+begin;
+select * from t1 for update;
+a
+1
+begin;
+ select * from t1 for update;
+ flush tables with read lock;
+commit;
+a
+1
+unlock tables;
+drop table t1;