Merge 10.5 into 10.6bb-10.6-merge-galera

20 files changed, 496 insertions, 54 deletions

diff --git a/mysql-test/main/join.result b/mysql-test/main/join.result
index 11b7ecad3ee..942ee96fc32 100644
--- a/mysql-test/main/join.result
+++ b/mysql-test/main/join.result
@@ -3407,3 +3407,20 @@ id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
 drop table t1,t2,t3;
 drop table t1000,t10,t03;
 # End of 10.3 tests
+#
+# MDEV-30080 Wrong result with LEFT JOINs involving constant tables
+#
+CREATE TABLE t1 (a INT) ENGINE=MyISAM;
+INSERT INTO t1 VALUES (1);
+CREATE TABLE t2 (b INT) ENGINE=MyISAM;
+INSERT INTO t2 VALUES (1),(1);
+CREATE TABLE t3 (c INT PRIMARY KEY) ENGINE=MyISAM;
+SELECT * FROM t1 LEFT JOIN (t2 LEFT JOIN t3 ON t2.b = t3.c) ON t1.a = t2.b;
+a	b	c
+1	1	NULL
+1	1	NULL
+SELECT COUNT(*) FROM t1 LEFT JOIN (t2 LEFT JOIN t3 ON t2.b = t3.c) ON t1.a = t2.b;
+COUNT(*)
+2
+DROP TABLE t1, t2, t3;
+# End of 10.5 tests
diff --git a/mysql-test/main/join.test b/mysql-test/main/join.test
index b99f05f7c88..c8bd2886b30 100644
--- a/mysql-test/main/join.test
+++ b/mysql-test/main/join.test
@@ -1820,3 +1820,18 @@ drop table t1,t2,t3;
 drop table t1000,t10,t03;
 
 --echo # End of 10.3 tests
+
+--echo #
+--echo # MDEV-30080 Wrong result with LEFT JOINs involving constant tables
+--echo #
+
+CREATE TABLE t1 (a INT) ENGINE=MyISAM;
+INSERT INTO t1 VALUES (1);
+CREATE TABLE t2 (b INT) ENGINE=MyISAM;
+INSERT INTO t2 VALUES (1),(1);
+CREATE TABLE t3 (c INT PRIMARY KEY) ENGINE=MyISAM;
+SELECT * FROM t1 LEFT JOIN (t2 LEFT JOIN t3 ON t2.b = t3.c) ON t1.a = t2.b;
+SELECT COUNT(*) FROM t1 LEFT JOIN (t2 LEFT JOIN t3 ON t2.b = t3.c) ON t1.a = t2.b;
+DROP TABLE t1, t2, t3;
+
+--echo # End of 10.5 tests
diff --git a/mysql-test/suite/federated/federatedx.result b/mysql-test/suite/federated/federatedx.result
index c18665e4d99..49deff81c4c 100644
--- a/mysql-test/suite/federated/federatedx.result
+++ b/mysql-test/suite/federated/federatedx.result
@@ -2325,6 +2325,38 @@ DROP TABLE federated.t1;
 connection slave;
 DROP TABLE federated.t1;
 connection default;
+#
+# MDEV-30395 Wrong result with semijoin and Federated as outer table
+#
+create server s foreign data wrapper mysql options (host "127.0.0.1", database "test", user "root", port MASTER_PORT);
+CREATE TABLE t1 (a INT);
+INSERT INTO t1 VALUES (3),(2),(3);
+CREATE TABLE t2 (pk INT PRIMARY KEY);
+INSERT INTO t2 VALUES (1),(2),(3),(4);
+set @save_optimizer_switch=@@optimizer_switch;
+set optimizer_switch="materialization=off";
+CREATE TABLE t2_fed ENGINE=FEDERATED CONNECTION='s/t2';
+explain SELECT * FROM t2_fed WHERE pk IN ( SELECT a FROM t1 );
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	PRIMARY	t2_fed	ALL	NULL	NULL	NULL	NULL	4	Using where
+2	DEPENDENT SUBQUERY	t1	ALL	NULL	NULL	NULL	NULL	3	Using where
+SELECT * FROM t2_fed WHERE pk IN ( SELECT a FROM t1 );
+pk
+2
+3
+SET optimizer_switch='semijoin=off';
+explain SELECT * FROM t2_fed WHERE pk IN ( SELECT a FROM t1 );
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	PRIMARY	t2_fed	ALL	NULL	NULL	NULL	NULL	4	Using where
+2	DEPENDENT SUBQUERY	t1	ALL	NULL	NULL	NULL	NULL	3	Using where
+SELECT * FROM t2_fed WHERE pk IN ( SELECT a FROM t1 );
+pk
+2
+3
+DROP TABLE t2_fed, t1, t2;
+set @@optimizer_switch=@save_optimizer_switch;
+DROP SERVER s;
+# End of 10.5 tests
 connection master;
 DROP TABLE IF EXISTS federated.t1;
 DROP DATABASE IF EXISTS federated;
diff --git a/mysql-test/suite/federated/federatedx.test b/mysql-test/suite/federated/federatedx.test
index 51d34298626..7e5a335b786 100644
--- a/mysql-test/suite/federated/federatedx.test
+++ b/mysql-test/suite/federated/federatedx.test
@@ -2060,4 +2060,34 @@ connection slave;
 DROP TABLE federated.t1;
 connection default;
 
+--echo #
+--echo # MDEV-30395 Wrong result with semijoin and Federated as outer table
+--echo #
+
+
+--replace_result $MASTER_MYPORT MASTER_PORT
+eval create server s foreign data wrapper mysql options (host "127.0.0.1", database "test", user "root", port $MASTER_MYPORT);
+
+CREATE TABLE t1 (a INT);
+INSERT INTO t1 VALUES (3),(2),(3);
+CREATE TABLE t2 (pk INT PRIMARY KEY);
+INSERT INTO t2 VALUES (1),(2),(3),(4);
+
+set @save_optimizer_switch=@@optimizer_switch;
+set optimizer_switch="materialization=off";
+
+CREATE TABLE t2_fed ENGINE=FEDERATED CONNECTION='s/t2';
+explain SELECT * FROM t2_fed WHERE pk IN ( SELECT a FROM t1 );
+SELECT * FROM t2_fed WHERE pk IN ( SELECT a FROM t1 );
+SET optimizer_switch='semijoin=off';
+explain SELECT * FROM t2_fed WHERE pk IN ( SELECT a FROM t1 );
+SELECT * FROM t2_fed WHERE pk IN ( SELECT a FROM t1 );
+
+DROP TABLE t2_fed, t1, t2;
+set @@optimizer_switch=@save_optimizer_switch;
+
+DROP SERVER s;
+
+--echo # End of 10.5 tests
+
 source include/federated_cleanup.inc;
diff --git a/mysql-test/suite/galera/r/galera_MDEV-29512.result b/mysql-test/suite/galera/r/galera_MDEV-29512.result
new file mode 100644
index 00000000000..aaf24df920e
--- /dev/null
+++ b/mysql-test/suite/galera/r/galera_MDEV-29512.result
@@ -0,0 +1,40 @@
+connection node_2;
+connection node_1;
+CREATE TABLE t1 (f1 INTEGER PRIMARY KEY, f2 int, f3 varchar(2000));
+INSERT INTO t1 VALUES (1, 0, REPEAT('1234567890', 200));
+INSERT INTO t1 VALUES (3, 3, REPEAT('1234567890', 200));
+SET SESSION wsrep_sync_wait=0;
+SET GLOBAL DEBUG_DBUG = "d,sync.wsrep_apply_cb";
+connect node_1a, 127.0.0.1, root, , test, $NODE_MYPORT_1;
+connection node_1a;
+SET SESSION wsrep_sync_wait=0;
+connection node_1;
+begin;
+select f1,f2 from t1;
+f1	f2
+1	0
+3	3
+connection node_2;
+UPDATE t1 SET f2=2 WHERE f1=3;
+connection node_1a;
+SET SESSION DEBUG_SYNC = "now WAIT_FOR sync.wsrep_apply_cb_reached";
+connection node_1;
+UPDATE t1 SET f2=1 WHERE f1=3;
+SET GLOBAL wsrep_provider_options = 'dbug=d,commit_monitor_master_enter_sync';
+COMMIT;
+connection node_1a;
+SET SESSION wsrep_on = 0;
+SET SESSION wsrep_on = 1;
+SET GLOBAL wsrep_provider_options = 'dbug=';
+SET GLOBAL wsrep_provider_options = 'signal=commit_monitor_master_enter_sync';
+SET GLOBAL DEBUG_DBUG = "";
+SET DEBUG_SYNC = "now SIGNAL signal.wsrep_apply_cb";
+SET GLOBAL debug_dbug = NULL;
+SET debug_sync='RESET';
+connection node_1;
+ERROR 40001: Deadlock found when trying to get lock; try restarting transaction
+select f1,f2 from t1;
+f1	f2
+1	0
+3	2
+DROP TABLE t1;
diff --git a/mysql-test/suite/galera/r/galera_savepoint_replay.result b/mysql-test/suite/galera/r/galera_savepoint_replay.result
new file mode 100644
index 00000000000..afea5f82e3c
--- /dev/null
+++ b/mysql-test/suite/galera/r/galera_savepoint_replay.result
@@ -0,0 +1,53 @@
+connection node_2;
+connection node_1;
+CREATE TABLE t1 (f1 INTEGER PRIMARY KEY, f2 CHAR(1));
+INSERT INTO t1 VALUES (1, 'a');
+INSERT INTO t1 VALUES (2, 'a');
+connection node_1;
+SET AUTOCOMMIT=ON;
+START TRANSACTION;
+UPDATE t1 SET f2 = 'b' WHERE f1 = 1;
+SELECT * FROM t1 WHERE f1 = 2 FOR UPDATE;
+f1	f2
+2	a
+SAVEPOINT my_sp;
+connect node_1a, 127.0.0.1, root, , test, $NODE_MYPORT_1;
+SET SESSION wsrep_sync_wait=0;
+SET GLOBAL wsrep_provider_options = 'dbug=d,apply_monitor_slave_enter_sync';
+connection node_2;
+UPDATE t1 SET f2 = 'c' WHERE f1 = 2;
+connection node_1a;
+SET SESSION wsrep_on = 0;
+SET SESSION wsrep_on = 1;
+SET GLOBAL wsrep_provider_options = 'dbug=';
+SET GLOBAL wsrep_provider_options = 'dbug=d,commit_monitor_master_enter_sync';
+connection node_1;
+COMMIT;
+connection node_1a;
+SET SESSION wsrep_on = 0;
+SET SESSION wsrep_on = 1;
+SET GLOBAL wsrep_provider_options = 'dbug=';
+SET GLOBAL wsrep_provider_options = 'dbug=d,abort_trx_end';
+SET GLOBAL wsrep_provider_options = 'signal=apply_monitor_slave_enter_sync';
+SET SESSION wsrep_on = 0;
+SET SESSION wsrep_on = 1;
+SET GLOBAL wsrep_provider_options = 'dbug=';
+SET GLOBAL wsrep_provider_options = 'signal=abort_trx_end';
+SET GLOBAL wsrep_provider_options = 'signal=commit_monitor_master_enter_sync';
+connection node_1;
+SELECT COUNT(*) = 1 FROM t1 WHERE f2 = 'b';
+COUNT(*) = 1
+1
+SELECT COUNT(*) = 1 FROM t1 WHERE f2 = 'c';
+COUNT(*) = 1
+1
+wsrep_local_replays
+1
+connection node_2;
+SELECT COUNT(*) = 1 FROM t1 WHERE f2 = 'b';
+COUNT(*) = 1
+1
+SELECT COUNT(*) = 1 FROM t1 WHERE f2 = 'c';
+COUNT(*) = 1
+1
+DROP TABLE t1;
diff --git a/mysql-test/suite/galera/t/galera_MDEV-29512.cnf b/mysql-test/suite/galera/t/galera_MDEV-29512.cnf
new file mode 100644
index 00000000000..bf8e0c37984
--- /dev/null
+++ b/mysql-test/suite/galera/t/galera_MDEV-29512.cnf
@@ -0,0 +1,15 @@
+!include ../galera_2nodes.cnf
+
+[mysqld]
+log-bin
+log-slave-updates
+
+[mysqld.1]
+log_bin
+log_slave_updates
+max-binlog-size=4096
+expire-logs-days=1
+
+
+[mysqld.2]
+
diff --git a/mysql-test/suite/galera/t/galera_MDEV-29512.test b/mysql-test/suite/galera/t/galera_MDEV-29512.test
new file mode 100644
index 00000000000..ffcef792f85
--- /dev/null
+++ b/mysql-test/suite/galera/t/galera_MDEV-29512.test
@@ -0,0 +1,91 @@
+#
+# This test is for reproducing the issue in:
+# https://jira.mariadb.org/browse/MDEV-29512
+#
+# The hanging in MDEV-29512 happens when binlog purging is attempted, and there is
+# one local BF aborted transaction waiting for commit monitor.
+#
+# The test will launch two node cluster and enable binlogging with expire log days,
+# to force binlog purging to happen.
+# A local transaction is executed so that will become BF abort victim, and has advanced
+# to replication stage waiting for commit monitor for final cleanup (to mark position in innodb)
+# after that, applier is released to complete the BF abort and due to binlog configuration,
+# starting the binlog purging. This is where the hanging would occur, if code is buggy
+#
+--source include/galera_cluster.inc
+--source include/have_innodb.inc
+--source include/have_debug_sync.inc
+--source include/galera_have_debug_sync.inc
+
+#
+# binlog size is limited to 4096 bytes, we will create enough events to
+# cause binlog rotation
+#
+CREATE TABLE t1 (f1 INTEGER PRIMARY KEY, f2 int, f3 varchar(2000));
+INSERT INTO t1 VALUES (1, 0, REPEAT('1234567890', 200));
+INSERT INTO t1 VALUES (3, 3, REPEAT('1234567890', 200));
+
+SET SESSION wsrep_sync_wait=0;
+
+# set sync point for replication applier
+SET GLOBAL DEBUG_DBUG = "d,sync.wsrep_apply_cb";
+
+# Control connection to manage sync points for appliers
+--connect node_1a, 127.0.0.1, root, , test, $NODE_MYPORT_1
+--connection node_1a
+SET SESSION wsrep_sync_wait=0;
+
+# starting local transaction, only select so far,
+# write will happen later and this will be ordered after the transaction in node_2
+--connection node_1
+begin;
+select f1,f2 from t1;
+
+# send from node 2 an UPDATE transaction,  which will BF abort the transaction in node_1
+--connection node_2
+--let $wait_condition=select count(*)=2 from t1
+--source include/wait_condition.inc
+
+UPDATE t1 SET f2=2 WHERE f1=3;
+
+--connection node_1a
+# wait to see the UPDATE from node_2 in apply_cb sync point
+SET SESSION DEBUG_SYNC = "now WAIT_FOR sync.wsrep_apply_cb_reached";
+
+--connection node_1
+# now issuing conflicting update
+UPDATE t1 SET f2=1 WHERE f1=3;
+
+# Block the local commit, send final COMMIT and wait until it gets blocked
+--let $galera_sync_point = commit_monitor_master_enter_sync
+--source include/galera_set_sync_point.inc
+--send COMMIT
+
+--connection node_1a
+# wait for the local commit to enter in commit monitor wait state
+--let $galera_sync_point = commit_monitor_master_enter_sync
+--source include/galera_wait_sync_point.inc
+--source include/galera_clear_sync_point.inc
+
+# release the local transaction to continue with commit
+--let $galera_sync_point = commit_monitor_master_enter_sync
+--source include/galera_signal_sync_point.inc
+
+# and now release the applier, it should force local trx to abort
+SET GLOBAL DEBUG_DBUG = "";
+SET DEBUG_SYNC = "now SIGNAL signal.wsrep_apply_cb";
+SET GLOBAL debug_dbug = NULL;
+SET debug_sync='RESET';
+
+--connection node_1
+--error ER_LOCK_DEADLOCK
+--reap
+
+# wait until applying is complete
+--let $wait_condition = SELECT COUNT(*)=1 FROM t1 WHERE f2=2
+--source include/wait_condition.inc
+
+# final read to verify what we got
+select f1,f2 from t1;
+
+DROP TABLE t1;
diff --git a/mysql-test/suite/galera/t/galera_savepoint_replay.test b/mysql-test/suite/galera/t/galera_savepoint_replay.test
new file mode 100644
index 00000000000..cff26f4a94f
--- /dev/null
+++ b/mysql-test/suite/galera/t/galera_savepoint_replay.test
@@ -0,0 +1,86 @@
+#
+# This test tests replaying a transaction with savepoint
+#
+
+--source include/galera_cluster.inc
+--source include/have_innodb.inc
+--source include/have_debug_sync.inc
+--source include/galera_have_debug_sync.inc
+
+--let $wsrep_local_replays_old = `SELECT VARIABLE_VALUE FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_local_replays'`
+
+CREATE TABLE t1 (f1 INTEGER PRIMARY KEY, f2 CHAR(1));
+INSERT INTO t1 VALUES (1, 'a');
+INSERT INTO t1 VALUES (2, 'a');
+
+--connection node_1
+SET AUTOCOMMIT=ON;
+START TRANSACTION;
+
+UPDATE t1 SET f2 = 'b' WHERE f1 = 1;
+SELECT * FROM t1 WHERE f1 = 2 FOR UPDATE;
+SAVEPOINT my_sp;
+
+# Block the applier on node #1 and issue a conflicting update on node #2
+--connect node_1a, 127.0.0.1, root, , test, $NODE_MYPORT_1
+SET SESSION wsrep_sync_wait=0;
+--let $galera_sync_point = apply_monitor_slave_enter_sync
+--source include/galera_set_sync_point.inc
+
+--connection node_2
+UPDATE t1 SET f2 = 'c' WHERE f1 = 2;
+
+--connection node_1a
+--source include/galera_wait_sync_point.inc
+--source include/galera_clear_sync_point.inc
+
+# Block the commit, send the COMMIT and wait until it gets blocked
+
+--let $galera_sync_point = commit_monitor_master_enter_sync
+--source include/galera_set_sync_point.inc
+
+--connection node_1
+--send COMMIT
+
+--connection node_1a
+
+--let $galera_sync_point = apply_monitor_slave_enter_sync commit_monitor_master_enter_sync
+--source include/galera_wait_sync_point.inc
+--source include/galera_clear_sync_point.inc
+
+# Let the conflicting UPDATE proceed and wait until it hits abort_trx_end.
+# The victim transaction still sits in commit_monitor_master_sync_point.
+
+--let $galera_sync_point = abort_trx_end
+--source include/galera_set_sync_point.inc
+--let $galera_sync_point = apply_monitor_slave_enter_sync
+--source include/galera_signal_sync_point.inc
+--let $galera_sync_point = abort_trx_end commit_monitor_master_enter_sync
+--source include/galera_wait_sync_point.inc
+
+# Let the transactions proceed
+--source include/galera_clear_sync_point.inc
+--let $galera_sync_point = abort_trx_end
+--source include/galera_signal_sync_point.inc
+--let $galera_sync_point = commit_monitor_master_enter_sync
+--source include/galera_signal_sync_point.inc
+
+# Commit succeeds
+--connection node_1
+--reap
+
+SELECT COUNT(*) = 1 FROM t1 WHERE f2 = 'b';
+SELECT COUNT(*) = 1 FROM t1 WHERE f2 = 'c';
+
+# wsrep_local_replays has increased by 1
+--let $wsrep_local_replays_new = `SELECT VARIABLE_VALUE FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_local_replays'`
+--disable_query_log
+--eval SELECT $wsrep_local_replays_new - $wsrep_local_replays_old = 1 AS wsrep_local_replays;
+--enable_query_log
+
+--connection node_2
+SELECT COUNT(*) = 1 FROM t1 WHERE f2 = 'b';
+SELECT COUNT(*) = 1 FROM t1 WHERE f2 = 'c';
+
+DROP TABLE t1;
+
diff --git a/sql/handler.h b/sql/handler.h
index de08e2c2137..f4d94924bda 100644
--- a/sql/handler.h
+++ b/sql/handler.h
@@ -357,9 +357,9 @@ enum chf_create_flags {
   Rowid's are not comparable. This is set if the rowid is unique to the
   current open handler, like it is with federated where the rowid is a
   pointer to a local result set buffer. The effect of having this set is
-  that the optimizer will not consirer the following optimizations for
+  that the optimizer will not consider the following optimizations for
   the table:
-  ror scans or filtering
+  ror scans, filtering or duplicate weedout
 */
 #define HA_NON_COMPARABLE_ROWID (1ULL << 60)
 
diff --git a/sql/opt_subselect.cc b/sql/opt_subselect.cc
index b971c96cda2..aa61d2c4605 100644
--- a/sql/opt_subselect.cc
+++ b/sql/opt_subselect.cc
@@ -664,6 +664,17 @@ int check_and_do_in_subquery_rewrites(JOIN *join)
         DBUG_RETURN(-1);
       }
     }
+    /* Check if any table is not supporting comparable rowids */
+    {
+      List_iterator_fast<TABLE_LIST> li(select_lex->outer_select()->leaf_tables);
+      TABLE_LIST *tbl;
+      while ((tbl = li++))
+      {
+        TABLE *table= tbl->table;
+        if (table && table->file->ha_table_flags() & HA_NON_COMPARABLE_ROWID)
+          join->not_usable_rowid_map|= table->map;
+      }
+    }
 
     DBUG_PRINT("info", ("Checking if subq can be converted to semi-join"));
     /*
@@ -683,8 +694,11 @@ int check_and_do_in_subquery_rewrites(JOIN *join)
         9. Parent select is not a table-less select
         10. Neither parent nor child select have STRAIGHT_JOIN option.
         11. It is first optimisation (the subquery could be moved from ON
-        clause during first optimisation and then be considered for SJ
-        on the second when it is too late)
+            clause during first optimisation and then be considered for SJ
+            on the second when it is too late)
+        12. All tables supports comparable rowids.
+            This is needed for DuplicateWeedout strategy to work (which
+            is the catch-all semi-join strategy so it must be applicable).
     */
     if (optimizer_flag(thd, OPTIMIZER_SWITCH_SEMIJOIN) &&
         in_subs &&                                                    // 1
@@ -699,7 +713,8 @@ int check_and_do_in_subquery_rewrites(JOIN *join)
         !((join->select_options |                                     // 10
            select_lex->outer_select()->join->select_options)          // 10
           & SELECT_STRAIGHT_JOIN) &&                                  // 10
-        select_lex->first_cond_optimization)                          // 11
+        select_lex->first_cond_optimization &&                        // 11
+        join->not_usable_rowid_map == 0)                              // 12
     {
       DBUG_PRINT("info", ("Subquery is semi-join conversion candidate"));
 
@@ -3556,6 +3571,9 @@ bool Duplicate_weedout_picker::check_qep(JOIN *join,
       }
       else
       {
+        /* Ensure that table supports comparable rowids */
+        DBUG_ASSERT(!(p->table->table->file->ha_table_flags() & HA_NON_COMPARABLE_ROWID));
+
         sj_outer_fanout= COST_MULT(sj_outer_fanout, p->records_read);
         temptable_rec_size += p->table->table->file->ref_length;
       }
diff --git a/sql/service_wsrep.cc b/sql/service_wsrep.cc
index 43183ff7595..b8c6d842301 100644
--- a/sql/service_wsrep.cc
+++ b/sql/service_wsrep.cc
@@ -1,4 +1,4 @@
-/* Copyright 2018 Codership Oy <info@codership.com>
+/* Copyright 2018-2023 Codership Oy <info@codership.com>
 
    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
@@ -253,7 +253,9 @@ extern "C" my_bool wsrep_thd_skip_locking(const THD *thd)
 
 extern "C" my_bool wsrep_thd_order_before(const THD *left, const THD *right)
 {
-  if (wsrep_thd_trx_seqno(left) < wsrep_thd_trx_seqno(right)) {
+  if (wsrep_thd_is_BF(left, false) &&
+      wsrep_thd_is_BF(right, false) &&
+      wsrep_thd_trx_seqno(left) < wsrep_thd_trx_seqno(right)) {
     WSREP_DEBUG("BF conflict, order: %lld %lld\n",
                 (long long)wsrep_thd_trx_seqno(left),
                 (long long)wsrep_thd_trx_seqno(right));
@@ -369,13 +371,20 @@ extern "C" ulong wsrep_OSU_method_get(const MYSQL_THD thd)
 
 extern "C" bool wsrep_thd_set_wsrep_aborter(THD *bf_thd, THD *victim_thd)
 {
-  WSREP_DEBUG("wsrep_thd_set_wsrep_aborter called");
   mysql_mutex_assert_owner(&victim_thd->LOCK_thd_data);
+  if (!bf_thd)
+  {
+    victim_thd->wsrep_aborter= 0;
+    WSREP_DEBUG("wsrep_thd_set_wsrep_aborter resetting wsrep_aborter");
+    return false;
+  }
   if (victim_thd->wsrep_aborter && victim_thd->wsrep_aborter != bf_thd->thread_id)
   {
     return true;
   }
-  victim_thd->wsrep_aborter = bf_thd->thread_id;
+  victim_thd->wsrep_aborter= bf_thd->thread_id;
+  WSREP_DEBUG("wsrep_thd_set_wsrep_aborter setting wsrep_aborter %u",
+              victim_thd->wsrep_aborter);
   return false;
 }
 
diff --git a/sql/sql_class.cc b/sql/sql_class.cc
index 6276b00b939..fa3c0150716 100644
--- a/sql/sql_class.cc
+++ b/sql/sql_class.cc
@@ -5462,7 +5462,6 @@ thd_need_ordering_with(const MYSQL_THD thd, const MYSQL_THD other_thd)
      between high priority wsrep threads.
      Note that wsrep_thd_is_BF() doesn't take LOCK_thd_data for either thd,
      the caller should guarantee that the BF state won't change.
-     (e.g. InnoDB does it by keeping lock_sys.mutex locked)
   */
   if (WSREP_ON && wsrep_thd_is_BF(thd, false) &&
       wsrep_thd_is_BF(other_thd, false))
diff --git a/sql/sql_select.cc b/sql/sql_select.cc
index 3c2be0b1c8b..c6253614889 100644
--- a/sql/sql_select.cc
+++ b/sql/sql_select.cc
@@ -2476,7 +2476,7 @@ JOIN::optimize_inner()
   /*
     We have to remove constants and duplicates from group_list before
     calling make_join_statistics() as this may call get_best_group_min_max()
-    which needs a simplfied group_list.
+    which needs a simplified group_list.
   */
   if (group_list && table_count == 1)
   {
@@ -5864,7 +5864,7 @@ make_join_statistics(JOIN *join, List<TABLE_LIST> &tables_list,
             caller to abort with a zero row result.
           */
           TABLE_LIST *emb= s->table->pos_in_table_list->embedding;
-          if (emb && !emb->sj_on_expr)
+          if (emb && !emb->sj_on_expr && !*s->on_expr_ref)
           {
             /* Mark all tables in a multi-table join nest as const */
             mark_join_nest_as_const(join, emb, &found_const_table_map,
diff --git a/sql/sql_select.h b/sql/sql_select.h
index ba38cdade55..15d841dfd6a 100644
--- a/sql/sql_select.h
+++ b/sql/sql_select.h
@@ -1254,6 +1254,8 @@ public:
   table_map outer_join;
   /* Bitmap of tables used in the select list items */
   table_map select_list_used_tables;
+  /* Tables that has HA_NON_COMPARABLE_ROWID (does not support rowid) set */
+  table_map not_usable_rowid_map;
   ha_rows  send_records,found_records,join_examined_rows, accepted_rows;
 
   /*
diff --git a/sql/wsrep_high_priority_service.cc b/sql/wsrep_high_priority_service.cc
index 708eb552866..93d4738212d 100644
--- a/sql/wsrep_high_priority_service.cc
+++ b/sql/wsrep_high_priority_service.cc
@@ -1,4 +1,4 @@
-/* Copyright 2018-2021 Codership Oy <info@codership.com>
+/* Copyright 2018-2023 Codership Oy <info@codership.com>
 
    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
@@ -639,6 +639,9 @@ Wsrep_replayer_service::Wsrep_replayer_service(THD* replayer_thd, THD* orig_thd)
      transactional locks */
   DBUG_ASSERT(!orig_thd->mdl_context.has_transactional_locks());
 
+  replayer_thd->system_thread_info.rpl_sql_info=
+    new rpl_sql_thread_info(replayer_thd->wsrep_rgi->rli->mi->rpl_filter);
+
   /* Make a shadow copy of diagnostics area and reset */
   m_da_shadow.status= orig_thd->get_stmt_da()->status();
   if (m_da_shadow.status == Diagnostics_area::DA_OK)
@@ -677,35 +680,35 @@ Wsrep_replayer_service::Wsrep_replayer_service(THD* replayer_thd, THD* orig_thd)
 
 Wsrep_replayer_service::~Wsrep_replayer_service()
 {
-  THD* replayer_thd= m_thd;
-  THD* orig_thd= m_orig_thd;
-
   /* Switch execution context back to original. */
-  wsrep_after_apply(replayer_thd);
-  wsrep_after_command_ignore_result(replayer_thd);
-  wsrep_close(replayer_thd);
-  wsrep_reset_threadvars(replayer_thd);
-  wsrep_store_threadvars(orig_thd);
+  wsrep_after_apply(m_thd);
+  wsrep_after_command_ignore_result(m_thd);
+  wsrep_close(m_thd);
+  wsrep_reset_threadvars(m_thd);
+  wsrep_store_threadvars(m_orig_thd);
 
-  DBUG_ASSERT(!orig_thd->get_stmt_da()->is_sent());
-  DBUG_ASSERT(!orig_thd->get_stmt_da()->is_set());
+  DBUG_ASSERT(!m_orig_thd->get_stmt_da()->is_sent());
+  DBUG_ASSERT(!m_orig_thd->get_stmt_da()->is_set());
+
+  delete m_thd->system_thread_info.rpl_sql_info;
+  m_thd->system_thread_info.rpl_sql_info= nullptr;
 
   if (m_replay_status == wsrep::provider::success)
   {
-    DBUG_ASSERT(replayer_thd->wsrep_cs().current_error() == wsrep::e_success);
-    orig_thd->reset_kill_query();
-    my_ok(orig_thd, m_da_shadow.affected_rows, m_da_shadow.last_insert_id);
+    DBUG_ASSERT(m_thd->wsrep_cs().current_error() == wsrep::e_success);
+    m_orig_thd->reset_kill_query();
+    my_ok(m_orig_thd, m_da_shadow.affected_rows, m_da_shadow.last_insert_id);
   }
   else if (m_replay_status == wsrep::provider::error_certification_failed)
   {
-    wsrep_override_error(orig_thd, ER_LOCK_DEADLOCK);
+    wsrep_override_error(m_orig_thd, ER_LOCK_DEADLOCK);
   }
   else
   {
     DBUG_ASSERT(0);
     WSREP_ERROR("trx_replay failed for: %d, schema: %s, query: %s",
                 m_replay_status,
-                orig_thd->db.str, wsrep_thd_query(orig_thd));
+                m_orig_thd->db.str, wsrep_thd_query(m_orig_thd));
     unireg_abort(1);
   }
 }
diff --git a/sql/wsrep_thd.cc b/sql/wsrep_thd.cc
index e610d3a6c2b..f6c4d949506 100644
--- a/sql/wsrep_thd.cc
+++ b/sql/wsrep_thd.cc
@@ -1,4 +1,4 @@
-/* Copyright (C) 2013-2022 Codership Oy <info@codership.com>
+/* Copyright (C) 2013-2023 Codership Oy <info@codership.com>
 
    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
@@ -308,11 +308,11 @@ void wsrep_fire_rollbacker(THD *thd)
 }
 
 
-int wsrep_abort_thd(THD *bf_thd_ptr, THD *victim_thd_ptr, my_bool signal)
+int wsrep_abort_thd(THD *bf_thd,
+                    THD *victim_thd,
+                    my_bool signal)
 {
   DBUG_ENTER("wsrep_abort_thd");
-  THD *victim_thd= (THD *) victim_thd_ptr;
-  THD *bf_thd= (THD *) bf_thd_ptr;
 
   mysql_mutex_lock(&victim_thd->LOCK_thd_data);
 
@@ -322,18 +322,23 @@ int wsrep_abort_thd(THD *bf_thd_ptr, THD *victim_thd_ptr, my_bool signal)
   if ((WSREP(bf_thd) ||
        ((WSREP_ON || bf_thd->variables.wsrep_OSU_method == WSREP_OSU_RSU) &&
 	 wsrep_thd_is_toi(bf_thd))) &&
-       victim_thd &&
        !wsrep_thd_is_aborting(victim_thd))
   {
-      WSREP_DEBUG("wsrep_abort_thd, by: %llu, victim: %llu", (bf_thd) ?
-                  (long long)bf_thd->real_id : 0, (long long)victim_thd->real_id);
+      WSREP_DEBUG("wsrep_abort_thd, by: %llu, victim: %llu",
+                  (long long)bf_thd->real_id, (long long)victim_thd->real_id);
       mysql_mutex_unlock(&victim_thd->LOCK_thd_data);
       ha_abort_transaction(bf_thd, victim_thd, signal);
       mysql_mutex_lock(&victim_thd->LOCK_thd_data);
   }
   else
   {
-    WSREP_DEBUG("wsrep_abort_thd not effective: %p %p", bf_thd, victim_thd);
+    WSREP_DEBUG("wsrep_abort_thd not effective: bf %llu victim %llu "
+                "wsrep %d wsrep_on %d RSU %d TOI %d aborting %d",
+                (long long)bf_thd->real_id, (long long)victim_thd->real_id,
+                WSREP_NNULL(bf_thd), WSREP_ON,
+                bf_thd->variables.wsrep_OSU_method == WSREP_OSU_RSU,
+                wsrep_thd_is_toi(bf_thd),
+                wsrep_thd_is_aborting(victim_thd));
   }
 
   mysql_mutex_unlock(&victim_thd->LOCK_thd_data);
diff --git a/sql/wsrep_thd.h b/sql/wsrep_thd.h
index fd48df1494f..3d1bf3733a8 100644
--- a/sql/wsrep_thd.h
+++ b/sql/wsrep_thd.h
@@ -1,4 +1,4 @@
-/* Copyright (C) 2013-2022 Codership Oy <info@codership.com>
+/* Copyright (C) 2013-2023 Codership Oy <info@codership.com>
 
    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
@@ -88,7 +88,9 @@ bool wsrep_create_appliers(long threads, bool mutex_protected=false);
 void wsrep_create_rollbacker();
 
 bool wsrep_bf_abort(THD* bf_thd, THD* victim_thd);
-int  wsrep_abort_thd(THD *bf_thd_ptr, THD *victim_thd_ptr, my_bool signal);
+int  wsrep_abort_thd(THD *bf_thd,
+                     THD *victim_thd,
+                     my_bool signal) __attribute__((nonnull(1,2)));
 
 /*
   Helper methods to deal with thread local storage.
diff --git a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc
index 4a1117a88fe..efa27151cb0 100644
--- a/storage/innobase/handler/ha_innodb.cc
+++ b/storage/innobase/handler/ha_innodb.cc
@@ -18691,7 +18691,15 @@ void lock_wait_wsrep_kill(trx_t *bf_trx, ulong thd_id, trx_id_t trx_id)
       lock_sys.cancel_lock_wait_for_trx(vtrx);
 
       DEBUG_SYNC(bf_thd, "before_wsrep_thd_abort");
-      wsrep_thd_bf_abort(bf_thd, vthd, true);
+      if (!wsrep_thd_bf_abort(bf_thd, vthd, true))
+      {
+        wsrep_thd_LOCK(vthd);
+        wsrep_thd_set_wsrep_aborter(NULL, vthd);
+        wsrep_thd_UNLOCK(vthd);
+
+        WSREP_DEBUG("wsrep_thd_bf_abort has failed, victim %lu will survive",
+                    thd_get_thread_id(vthd));
+      }
     }
     wsrep_thd_kill_UNLOCK(vthd);
   }
diff --git a/storage/innobase/lock/lock0lock.cc b/storage/innobase/lock/lock0lock.cc
index 46cfb6e5c88..ddcb5d12853 100644
--- a/storage/innobase/lock/lock0lock.cc
+++ b/storage/innobase/lock/lock0lock.cc
@@ -693,22 +693,31 @@ lock_rec_has_to_wait(
 #endif /* HAVE_REPLICATION */
 
 #ifdef WITH_WSREP
-		/* New lock request from a transaction is using unique key
-		scan and this transaction is a wsrep high priority transaction
-		(brute force). If conflicting transaction is also wsrep high
-		priority transaction we should avoid lock conflict because
-		ordering of these transactions is already decided and
-		conflicting transaction will be later replayed. */
-		if (trx->is_wsrep_UK_scan()
-		    && wsrep_thd_is_BF(lock2->trx->mysql_thd, false)) {
-			return false;
-		}
+	/* New lock request from a transaction is using unique key
+	scan and this transaction is a wsrep high priority transaction
+	(brute force). If conflicting transaction is also wsrep high
+	priority transaction we should avoid lock conflict because
+	ordering of these transactions is already decided and
+	conflicting transaction will be later replayed. */
+	if (trx->is_wsrep_UK_scan()
+	    && wsrep_thd_is_BF(lock2->trx->mysql_thd, false)) {
+		return false;
+	}
+
+	/* If BF-BF conflict, we have to look at write set order */
+	if (trx->is_wsrep()
+	    && (type_mode & LOCK_MODE_MASK) == LOCK_X
+	    && (lock2->type_mode & LOCK_MODE_MASK) == LOCK_X
+	    && wsrep_thd_order_before(trx->mysql_thd,
+				      lock2->trx->mysql_thd)) {
+		return false;
+	}
 
-		/* We very well can let bf to wait normally as other
-		BF will be replayed in case of conflict. For debug
-		builds we will do additional sanity checks to catch
-		unsupported bf wait if any. */
-		ut_d(wsrep_assert_no_bf_bf_wait(lock2, trx));
+	/* We very well can let bf to wait normally as other
+	BF will be replayed in case of conflict. For debug
+	builds we will do additional sanity checks to catch
+	unsupported bf wait if any. */
+	ut_d(wsrep_assert_no_bf_bf_wait(lock2, trx));
 #endif /* WITH_WSREP */
 
 	return true;
@@ -1600,6 +1609,14 @@ lock_rec_has_to_wait_in_queue(const hash_cell_t &cell, const lock_t *wait_lock)
 		if (heap_no < lock_rec_get_n_bits(lock)
 		    && (p[bit_offset] & bit_mask)
 		    && lock_has_to_wait(wait_lock, lock)) {
+#ifdef WITH_WSREP
+			if (lock->trx->is_wsrep()
+			    && wsrep_thd_order_before(wait_lock->trx->mysql_thd,
+						      lock->trx->mysql_thd)) {
+				/* don't wait for another BF lock */
+				continue;
+			}
+#endif
 			return(lock);
 		}
 	}