post push fix for bug#40595

Addition of hander function was_semi_consistent_read

mysql-test/r/partition_innodb_semi_consistent.result

+drop table if exists t1;
+set binlog_format=mixed;
+set session transaction isolation level read committed;
+create table t1(a int not null)
+engine=innodb
+DEFAULT CHARSET=latin1
+PARTITION BY RANGE(a)
+(PARTITION p0 VALUES LESS THAN (20),
+PARTITION p1 VALUES LESS THAN MAXVALUE);
+insert into t1 values (1),(2),(3),(4),(5),(6),(7);
+set autocommit=0;
+select * from t1 where a=3 lock in share mode;
+a
+3
+set binlog_format=mixed;
+set session transaction isolation level read committed;
+set autocommit=0;
+update t1 set a=10 where a=5;
+ERROR HY000: Lock wait timeout exceeded; try restarting transaction
+commit;
+update t1 set a=10 where a=5;
+select * from t1 where a=2 for update;
+ERROR HY000: Lock wait timeout exceeded; try restarting transaction
+select * from t1 where a=2 limit 1 for update;
+a
+2
+update t1 set a=11 where a=6;
+update t1 set a=12 where a=2;
+ERROR HY000: Lock wait timeout exceeded; try restarting transaction
+update t1 set a=13 where a=1;
+ERROR HY000: Lock wait timeout exceeded; try restarting transaction
+commit;
+update t1 set a=14 where a=1;
+commit;
+select * from t1;
+a
+10
+11
+14
+2
+3
+4
+7
+drop table t1;
+SET SESSION AUTOCOMMIT = 0;
+SET SESSION TRANSACTION ISOLATION LEVEL READ COMMITTED;
+set binlog_format=mixed;
+# Switch to connection con1
+CREATE TABLE t1 (a INT PRIMARY KEY, b VARCHAR(256))
+ENGINE = InnoDB
+PARTITION BY RANGE (a)
+(PARTITION p0 VALUES LESS THAN (300),
+PARTITION p1 VALUES LESS THAN MAXVALUE);
+INSERT INTO t1 VALUES (1,2);
+# 1. test for locking:
+BEGIN;
+UPDATE t1 SET b = 12 WHERE a = 1;
+affected rows: 1
+info: Rows matched: 1  Changed: 1  Warnings: 0
+SELECT * FROM t1;
+a	b
+1	12
+# Switch to connection con2
+UPDATE t1 SET b = 21 WHERE a = 1;
+ERROR HY000: Lock wait timeout exceeded; try restarting transaction
+# Switch to connection con1
+SELECT * FROM t1;
+a	b
+1	12
+ROLLBACK;
+# 2. test for serialized update:
+CREATE TABLE t2 (a INT);
+TRUNCATE t1;
+INSERT INTO t1 VALUES (1,'init');
+CREATE PROCEDURE p1()
+BEGIN
+UPDATE t1 SET b = CONCAT(b, '+con2')  WHERE a = 1;
+INSERT INTO t2 VALUES ();
+END|
+BEGIN;
+UPDATE t1 SET b = CONCAT(b, '+con1') WHERE a = 1;
+affected rows: 1
+info: Rows matched: 1  Changed: 1  Warnings: 0
+SELECT * FROM t1;
+a	b
+1	init+con1
+# Switch to connection con2
+CALL p1;;
+# Switch to connection con1
+SELECT * FROM t1;
+a	b
+1	init+con1
+COMMIT;
+SELECT * FROM t1;
+a	b
+1	init+con1
+# Switch to connection con2
+SELECT * FROM t1;
+a	b
+1	init+con1+con2
+# Switch to connection con1
+# 3. test for updated key column:
+TRUNCATE t1;
+TRUNCATE t2;
+INSERT INTO t1 VALUES (1,'init');
+BEGIN;
+UPDATE t1 SET a = 2, b = CONCAT(b, '+con1') WHERE a = 1;
+affected rows: 1
+info: Rows matched: 1  Changed: 1  Warnings: 0
+SELECT * FROM t1;
+a	b
+2	init+con1
+# Switch to connection con2
+CALL p1;;
+# Switch to connection con1
+SELECT * FROM t1;
+a	b
+2	init+con1
+COMMIT;
+SELECT * FROM t1;
+a	b
+2	init+con1
+# Switch to connection con2
+SELECT * FROM t1;
+a	b
+2	init+con1
+DROP PROCEDURE p1;
+DROP TABLE t1, t2;

mysql-test/t/partition_innodb_semi_consistent-master.opt

+--innodb_lock_wait_timeout=2

mysql-test/t/partition_innodb_semi_consistent.test

+-- source include/have_partition.inc
+-- source include/not_embedded.inc
+-- source include/have_innodb.inc
+
+--disable_warnings
+drop table if exists t1;
+--enable_warnings
+
+# basic tests of semi-consistent reads
+# for verifying Bug#40595: Non-matching rows not released with READ-COMMITTED
+#               on tables with partitions
+
+connect (a,localhost,root,,);
+connect (b,localhost,root,,);
+connection a;
+set binlog_format=mixed;
+set session transaction isolation level read committed;
+create table t1(a int not null)
+engine=innodb
+DEFAULT CHARSET=latin1
+PARTITION BY RANGE(a)
+(PARTITION p0 VALUES LESS THAN (20),
+ PARTITION p1 VALUES LESS THAN MAXVALUE);
+insert into t1 values (1),(2),(3),(4),(5),(6),(7);
+set autocommit=0;
+# this should lock the entire table
+select * from t1 where a=3 lock in share mode;
+connection b;
+set binlog_format=mixed;
+set session transaction isolation level read committed;
+set autocommit=0;
+-- error ER_LOCK_WAIT_TIMEOUT
+update t1 set a=10 where a=5;
+connection a;
+#DELETE FROM t1 WHERE a=5;
+commit;
+connection b;
+update t1 set a=10 where a=5;
+connection a;
+-- error ER_LOCK_WAIT_TIMEOUT
+select * from t1 where a=2 for update;
+# this should lock the records (1),(2)
+select * from t1 where a=2 limit 1 for update;
+connection b;
+update t1 set a=11 where a=6;
+-- error ER_LOCK_WAIT_TIMEOUT
+update t1 set a=12 where a=2;
+-- error ER_LOCK_WAIT_TIMEOUT
+update t1 set a=13 where a=1;
+connection a;
+commit;
+connection b;
+update t1 set a=14 where a=1;
+commit;
+connection a;
+--sorted_result
+select * from t1;
+drop table t1;
+
+connection default;
+disconnect a;
+disconnect b;
+
+#
+# Bug #31310: Locked rows silently skipped in read-committed isolation level.
+# (This also tests the '*_semi_consistent*' functions in partitioning)
+# Copied from include/mix1.inc
+
+connect (con1,localhost,root,,);
+connect (con2,localhost,root,,);
+SET SESSION AUTOCOMMIT = 0;
+SET SESSION TRANSACTION ISOLATION LEVEL READ COMMITTED;
+set binlog_format=mixed;
+--echo # Switch to connection con1
+connection con1;
+
+eval
+CREATE TABLE t1 (a INT PRIMARY KEY, b VARCHAR(256))
+ENGINE = InnoDB
+PARTITION BY RANGE (a)
+(PARTITION p0 VALUES LESS THAN (300),
+ PARTITION p1 VALUES LESS THAN MAXVALUE);
+INSERT INTO t1 VALUES (1,2);
+
+--echo # 1. test for locking:
+
+BEGIN;
+--enable_info
+UPDATE t1 SET b = 12 WHERE a = 1;
+--disable_info
+SELECT * FROM t1;
+
+--echo # Switch to connection con2
+connection con2;
+
+--enable_info
+--disable_abort_on_error
+--error ER_LOCK_WAIT_TIMEOUT
+UPDATE t1 SET b = 21 WHERE a = 1;
+--disable_info
+
+--echo # Switch to connection con1
+connection con1;
+SELECT * FROM t1;
+ROLLBACK;
+
+--echo # 2. test for serialized update:
+
+CREATE TABLE t2 (a INT);
+
+TRUNCATE t1;
+INSERT INTO t1 VALUES (1,'init');
+
+DELIMITER |;
+CREATE PROCEDURE p1()
+BEGIN
+  UPDATE t1 SET b = CONCAT(b, '+con2')  WHERE a = 1;
+  INSERT INTO t2 VALUES ();
+END|
+DELIMITER ;|
+
+BEGIN;
+--enable_info
+UPDATE t1 SET b = CONCAT(b, '+con1') WHERE a = 1;
+--disable_info
+SELECT * FROM t1;
+
+--echo # Switch to connection con2
+connection con2;
+
+--send CALL p1;
+
+--echo # Switch to connection con1
+connection con1;
+SELECT * FROM t1;
+COMMIT;
+
+let $bug31310 = 1;
+while ($bug31310)
+{
+  let $bug31310= `SELECT 1 - COUNT(*) FROM t2`;
+}
+
+SELECT * FROM t1;
+
+--echo # Switch to connection con2
+connection con2;
+--reap
+SELECT * FROM t1;
+
+--echo # Switch to connection con1
+connection con1;
+
+--echo # 3. test for updated key column:
+
+TRUNCATE t1;
+TRUNCATE t2;
+
+INSERT INTO t1 VALUES (1,'init');
+
+BEGIN;
+--enable_info
+UPDATE t1 SET a = 2, b = CONCAT(b, '+con1') WHERE a = 1;
+--disable_info
+SELECT * FROM t1;
+
+--echo # Switch to connection con2
+connection con2;
+
+--send CALL p1;
+
+--echo # Switch to connection con1
+connection con1;
+SELECT * FROM t1;
+COMMIT;
+
+let $bug31310 = 1;
+while ($bug31310)
+{
+  let $bug31310= `SELECT 1 - COUNT(*) FROM t2`;
+}
+
+SELECT * FROM t1;
+
+--echo # Switch to connection con2
+connection con2;
+SELECT * FROM t1;
+
+connection default;
+disconnect con1;
+disconnect con2;
+DROP PROCEDURE p1;
+DROP TABLE t1, t2;
+

sql/ha_partition.cc

@@ -2818,6 +2818,36 @@ void ha_partition::unlock_row()
   DBUG_VOID_RETURN;
 }
 
+/**
+  Check if semi consistent read was used
+
+  SYNOPSIS
+    was_semi_consistent_read()
+
+  RETURN VALUE
+    TRUE   Previous read was a semi consistent read
+    FALSE  Previous read was not a semi consistent read
+
+  DESCRIPTION
+    See handler.h:
+    In an UPDATE or DELETE, if the row under the cursor was locked by another
+    transaction, and the engine used an optimistic read of the last
+    committed row value under the cursor, then the engine returns 1 from this
+    function. MySQL must NOT try to update this optimistic value. If the
+    optimistic value does not match the WHERE condition, MySQL can decide to
+    skip over this row. Currently only works for InnoDB. This can be used to
+    avoid unnecessary lock waits.
+
+    If this method returns nonzero, it will also signal the storage
+    engine that the next read will be a locking re-read of the row.
+*/
+bool ha_partition::was_semi_consistent_read()
+{
+  DBUG_ENTER("ha_partition::was_semi_consistent_read");
+  DBUG_ASSERT(m_last_part < m_tot_parts &&
+              bitmap_is_set(&(m_part_info->used_partitions), m_last_part));
+  DBUG_RETURN(m_file[m_last_part]->was_semi_consistent_read());
+}
 
 /**
   Use semi consistent read if possible

sql/ha_partition.h

@@ -112,7 +112,7 @@ private:
   uint m_reorged_parts;                  // Number of reorganised parts
   uint m_tot_parts;                      // Total number of partitions;
   uint m_no_locks;                       // For engines like ha_blackhole, which needs no locks
-  uint m_last_part;                      // Last file that we update,write
+  uint m_last_part;                      // Last file that we update,write,read
   int m_lock_type;                       // Remembers type of last
                                          // external_lock
   part_id_range m_part_spec;             // Which parts to scan
@@ -325,6 +325,10 @@ public:
     Call to unlock rows not to be updated in transaction
   */
   virtual void unlock_row();
+  /*
+    Check if semi consistent read
+  */
+  virtual bool was_semi_consistent_read();
   /*
     Call to hint about semi consistent read
   */