MDEV-23328 Server hang due to Galera lock conflict resolution

Mutex order violation when wsrep bf thread kills a conflicting trx,
the stack is

          wsrep_thd_LOCK()
          wsrep_kill_victim()
          lock_rec_other_has_conflicting()
          lock_clust_rec_read_check_and_lock()
          row_search_mvcc()
          ha_innobase::index_read()
          ha_innobase::rnd_pos()
          handler::ha_rnd_pos()
          handler::rnd_pos_by_record()
          handler::ha_rnd_pos_by_record()
          Rows_log_event::find_row()
          Update_rows_log_event::do_exec_row()
          Rows_log_event::do_apply_event()
          Log_event::apply_event()
          wsrep_apply_events()

and mutexes are taken in the order

          lock_sys->mutex -> victim_trx->mutex -> victim_thread->LOCK_thd_data

When a normal KILL statement is executed, the stack is

          innobase_kill_query()
          kill_handlerton()
          plugin_foreach_with_mask()
          ha_kill_query()
          THD::awake()
          kill_one_thread()

        and mutexes are

          victim_thread->LOCK_thd_data -> lock_sys->mutex -> victim_trx->mutex

This patch is the plan D variant for fixing potetial mutex locking
order exercised by BF aborting and KILL command execution.

In this approach, KILL command is replicated as TOI operation.
This guarantees total isolation for the KILL command execution
in the first node: there is no concurrent replication applying
and no concurrent DDL executing. Therefore there is no risk of
BF aborting to happen in parallel with KILL command execution
either. Potential mutex deadlocks between the different mutex
access paths with KILL command execution and BF aborting cannot
therefore happen.

TOI replication is used, in this approach,  purely as means
to provide isolated KILL command execution in the first node.
KILL command should not (and must not) be applied in secondary
nodes. In this patch, we make this sure by skipping KILL
execution in secondary nodes, in applying phase, where we
bail out if applier thread is trying to execute KILL command.
This is effective, but skipping the applying of KILL command
could happen much earlier as well.

This also fixed unprotected calls to wsrep_thd_abort
that will use wsrep_abort_transaction. This is fixed
by holding THD::LOCK_thd_data while we abort transaction.

Reviewed-by: Jan Lindström <jan.lindstrom@mariadb.com>

Author: sjaakola

mysql-test/suite/galera/r/galera_UK_conflict.result

@@ -68,9 +68,9 @@ f1	f2	f3
 10	10	0
 INSERT INTO t1 VALUES (7,7,7);
 INSERT INTO t1 VALUES (8,8,8);
-DROP TABLE t1;
-test scenario 2
-connection node_1;
+SELECT COUNT(*) FROM t1;
+COUNT(*)
+7
 CREATE TABLE t1 (f1 INTEGER PRIMARY KEY, f2 int, f3 int,  unique key keyj (f2));
 INSERT INTO t1 VALUES (1, 1, 0);
 INSERT INTO t1 VALUES (3, 3, 0);
@@ -92,9 +92,9 @@ 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;
+SELECT COUNT(*) FROM t1;
+COUNT(*)
+7
 SET SESSION wsrep_on = 1;
 SET GLOBAL wsrep_provider_options = 'dbug=';
 SET GLOBAL DEBUG_DBUG = "d,sync.wsrep_replay_cb";
@@ -125,6 +125,7 @@ f1	f2	f3
 3	3	1
 4	4	2
 5	5	2
+8	8	8
 10	10	0
 INSERT INTO t1 VALUES (7,7,7);
 INSERT INTO t1 VALUES (8,8,8);

mysql-test/suite/galera/r/galera_bf_kill_debug.result

@@ -23,22 +23,6 @@ connection node_1a;
 connection node_1b;
 connection node_2;
 connection node_2a;
-connection node_1;
-SET SESSION wsrep_sync_wait=15;
-SELECT COUNT(*) FROM parent;
-COUNT(*)
-20001
-SELECT COUNT(*) FROM child;
-COUNT(*)
-10000
-connection node_2;
-SET SESSION wsrep_sync_wait=15;
-SELECT COUNT(*) FROM parent;
-COUNT(*)
-20001
-SELECT COUNT(*) FROM child;
-COUNT(*)
-10000
 DROP TABLE child;
 DROP TABLE parent;
 DROP TABLE ten;

mysql-test/suite/galera/r/galera_toi_ddl_fk_insert.result

@@ -140,6 +140,14 @@ SELECT * FROM t1;
 # original state in node 1
 INSERT INTO t1 VALUES (7,7,7);
 INSERT INTO t1 VALUES (8,8,8);
+SELECT COUNT(*) FROM t1;
+SELECT * FROM t1;
+
+--connection node_1
+--let $wait_condition = SELECT COUNT(*) = 7 FROM t1
+--source include/wait_condition.inc
+SELECT COUNT(*) FROM t1;
+SELECT * FROM t1;
 
 DROP TABLE t1;
 
@@ -199,9 +207,9 @@ INSERT INTO t1 VALUES (5, 5, 2);
 --source include/galera_set_sync_point.inc
 
 --connection node_1
---send COMMIT
-
---connection node_1a
+--let $wait_condition = SELECT COUNT(*) = 7 FROM t1
+--source include/wait_condition.inc
+SELECT COUNT(*) FROM t1;
 # wait for the local commit to enter in commit monitor wait state
 --let $galera_sync_point = apply_monitor_slave_enter_sync commit_monitor_master_enter_sync
 --source include/galera_wait_sync_point.inc
@@ -273,4 +281,13 @@ SELECT * FROM t1;
 INSERT INTO t1 VALUES (7,7,7);
 INSERT INTO t1 VALUES (8,8,8);
 
+SELECT COUNT(*) FROM t1;
+SELECT * FROM t1;
+
+--connection node_1
+--let $wait_condition = SELECT COUNT(*) = 7 FROM t1
+--source include/wait_condition.inc
+SELECT COUNT(*) FROM t1;
+SELECT * FROM t1;
+
 DROP TABLE t1;

mysql-test/suite/galera/t/galera_UK_conflict.test

@@ -54,15 +54,11 @@ INSERT INTO parent VALUES (1, 0);
 --connection node_2a
 --reap
 
---connection node_1
-SET SESSION wsrep_sync_wait=15;
-SELECT COUNT(*) FROM parent;
-SELECT COUNT(*) FROM child;
-
---connection node_2
-SET SESSION wsrep_sync_wait=15;
-SELECT COUNT(*) FROM parent;
-SELECT COUNT(*) FROM child;
+#
+# ALTER TABLE could bf kill one or more of INSERTs to parent, so
+# the actual number of rows in PARENT depends on whether
+# the INSERT is committed before ALTER TABLE is executed
+#
 
 DROP TABLE child;
 DROP TABLE parent;

mysql-test/suite/galera/t/galera_bf_kill_debug.cnf

@@ -94,11 +94,13 @@ SELECT * FROM t1;
 --eval SET GLOBAL wsrep_auto_increment_control = $auto_increment_control_orig
 --eval SET GLOBAL auto_increment_increment = $auto_increment_increment_node1
 --eval SET GLOBAL auto_increment_offset = $auto_increment_offset_node1
+--disconnect node_1a
 
 --connection node_2
 --eval SET GLOBAL wsrep_auto_increment_control = $auto_increment_control_orig
 --eval SET GLOBAL auto_increment_increment = $auto_increment_increment_node2
 --eval SET GLOBAL auto_increment_offset = $auto_increment_offset_node2
+--disconnect node_2a
 
 --enable_query_log
 

mysql-test/suite/galera/t/galera_bf_kill_debug.test

@@ -913,7 +913,7 @@ static my_bool kill_handlerton(THD *thd, plugin_ref plugin,
 {
   handlerton *hton= plugin_hton(plugin);
 
-  mysql_mutex_assert_owner(&thd->LOCK_thd_data);
+  mysql_mutex_assert_owner(&thd->LOCK_thd_kill);
   if (hton->kill_query && thd_get_ha_data(thd, hton))
     hton->kill_query(hton, thd, *(enum thd_kill_levels *) level);
   return FALSE;