Merge 10.5 into 10.6

include/mysql/service_my_crypt.h

@@ -45,7 +45,7 @@ extern "C" {
 /* The max key length of all supported algorithms */
 #define MY_AES_MAX_KEY_LENGTH 32
 
-#define MY_AES_CTX_SIZE 1024
+#define MY_AES_CTX_SIZE 1040
 
 enum my_aes_mode {
     MY_AES_ECB, MY_AES_CBC

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

@@ -27,7 +27,6 @@ SET SESSION wsrep_sync_wait = 0;
 CREATE TABLE t1 (f1 INTEGER PRIMARY KEY) ENGINE=InnoDB;
 
 --connection node_2
---sleep 1
 --let $wait_condition = SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.TABLES WHERE TABLE_NAME = 't1';
 --source include/wait_condition.inc
 
@@ -47,7 +46,6 @@ INSERT INTO t1 VALUES (1),(2),(3),(4),(5);
 
 --connection node_2
 --sleep 5
-
 --let $value = query_get_value(SHOW SLAVE STATUS, Last_SQL_Error, 1)
 --connection node_1
 --disable_query_log

mysql-test/suite/rpl/r/rpl_xa_prepare_gtid_fail.result

@@ -2,14 +2,14 @@ include/master-slave.inc
 [connection master]
 connection slave;
 include/stop_slave.inc
+set @save_par_thds= @@global.slave_parallel_threads;
+set @save_strict_mode= @@global.gtid_strict_mode;
+set @save_innodb_lock_wait_timeout= @@global.innodb_lock_wait_timeout;
 change master to master_use_gtid=slave_pos;
 set @@global.slave_parallel_threads= 4;
 set @@global.slave_parallel_mode= optimistic;
 set @@global.gtid_strict_mode=ON;
-set sql_log_bin= 0;
-alter table mysql.gtid_slave_pos engine=innodb;
-call mtr.add_suppression("Deadlock found.*");
-set sql_log_bin= 1;
+set statement sql_log_bin=0 for alter table mysql.gtid_slave_pos engine=innodb;
 include/start_slave.inc
 connection master;
 create table t1 (a int primary key, b int) engine=innodb;
@@ -27,25 +27,25 @@ xa end '1';
 xa prepare '1';
 xa commit '1';
 include/save_master_gtid.inc
-connection slave;
 connection slave1;
 BEGIN;
 SELECT * FROM mysql.gtid_slave_pos WHERE seq_no=100 FOR UPDATE;
 domain_id	sub_id	server_id	seq_no
 connection slave;
 include/start_slave.inc
-include/wait_for_slave_sql_error.inc [errno=1942,1213]
+include/wait_for_slave_sql_error.inc [errno=1942]
+include/stop_slave_io.inc
 connection slave1;
 ROLLBACK;
 # Cleanup
 connection master;
 drop table t1;
 connection slave;
-include/stop_slave.inc
+# TODO: Remove after fixing MDEV-21777
 set @@global.gtid_slave_pos= "0-1-100";
-set @@global.slave_parallel_threads= 0;
-set @@global.gtid_strict_mode= 0;
-set @@global.innodb_lock_wait_timeout= 50;
+set @@global.slave_parallel_threads= @save_par_thds;
+set @@global.gtid_strict_mode= @save_strict_mode;
+set @@global.innodb_lock_wait_timeout= @save_innodb_lock_wait_timeout;
 include/start_slave.inc
 include/rpl_end.inc
 # End of rpl_xa_prepare_gtid_fail.test

mysql-test/suite/rpl/t/rpl_xa_prepare_gtid_fail.test

@@ -6,8 +6,8 @@
 # GTID slave state, then the slave should immediately quit in error, without
 # retry.
 #
-#   This tests validates the above behavior by simulating a deadlock on the
-# GTID slave state table during the second part of XA PREPARE's commit, to
+#   This tests validates the above behavior by forcing a lock-wait timeout on
+# the GTID slave state table during the second part of XA PREPARE's commit, to
 # ensure that the appropriate error is reported and the transaction was never
 # retried.
 #
@@ -23,23 +23,19 @@ source include/have_innodb.inc;
 --connection slave
 --source include/stop_slave.inc
 
---let $save_par_thds= `SELECT @@global.slave_parallel_threads`
---let $save_strict_mode= `SELECT @@global.gtid_strict_mode`
---let $save_innodb_lock_wait_timeout= `SELECT @@global.innodb_lock_wait_timeout`
+set @save_par_thds= @@global.slave_parallel_threads;
+set @save_strict_mode= @@global.gtid_strict_mode;
+set @save_innodb_lock_wait_timeout= @@global.innodb_lock_wait_timeout;
 
 change master to master_use_gtid=slave_pos;
 set @@global.slave_parallel_threads= 4;
 set @@global.slave_parallel_mode= optimistic;
 set @@global.gtid_strict_mode=ON;
 
-set sql_log_bin= 0;
-alter table mysql.gtid_slave_pos engine=innodb;
-call mtr.add_suppression("Deadlock found.*");
-set sql_log_bin= 1;
+set statement sql_log_bin=0 for alter table mysql.gtid_slave_pos engine=innodb;
 --source include/start_slave.inc
 
 --connection master
-let $datadir= `select @@datadir`;
 create table t1 (a int primary key, b int) engine=innodb;
 insert t1 values (1,1);
 --source include/save_master_gtid.inc
@@ -64,11 +60,6 @@ xa prepare '1';
 xa commit '1';
 --source include/save_master_gtid.inc
 
-
---connection slave
-
-#--eval set statement sql_log_bin=0 for insert into mysql.gtid_slave_pos values ($gtid_domain_id, 5, $gtid_server_id, $xap_seq_no)
-
 --connection slave1
 BEGIN;
 --eval SELECT * FROM mysql.gtid_slave_pos WHERE seq_no=$xap_seq_no FOR UPDATE
@@ -76,9 +67,14 @@ BEGIN;
 --connection slave
 --source include/start_slave.inc
 
---let $slave_sql_errno= 1942,1213
+--let $slave_sql_errno= 1942
 --source include/wait_for_slave_sql_error.inc
 
+# TODO: Remove after fixing MDEV-21777
+# Stop the IO thread too, so the existing relay logs are force purged on slave
+# restart, as to not re-execute the already-prepared transaction
+--source include/stop_slave_io.inc
+
 --let $retried_tx_test= query_get_value(SHOW ALL SLAVES STATUS, Retried_transactions, 1)
 if ($retried_tx_initial != $retried_tx_test)
 {
@@ -95,11 +91,11 @@ ROLLBACK;
 drop table t1;
 
 --connection slave
---source include/stop_slave.inc
+--echo # TODO: Remove after fixing MDEV-21777
 --eval set @@global.gtid_slave_pos= "$new_gtid"
---eval set @@global.slave_parallel_threads= $save_par_thds
---eval set @@global.gtid_strict_mode= $save_strict_mode
---eval set @@global.innodb_lock_wait_timeout= $save_innodb_lock_wait_timeout
+set @@global.slave_parallel_threads= @save_par_thds;
+set @@global.gtid_strict_mode= @save_strict_mode;
+set @@global.innodb_lock_wait_timeout= @save_innodb_lock_wait_timeout;
 --source include/start_slave.inc
 
 --source include/rpl_end.inc

sql/log_event_server.cc

@@ -5778,7 +5778,8 @@ int Rows_log_event::do_apply_event(rpl_group_info *rgi)
                              ignored_error_code(actual_error) : 0);
 
 #ifdef WITH_WSREP
-        if (WSREP(thd) && wsrep_ignored_error_code(this, actual_error))
+        if (WSREP(thd) && thd->wsrep_applier &&
+            wsrep_ignored_error_code(this, actual_error))
         {
           idempotent_error= true;
           thd->wsrep_has_ignored_error= true;

sql/rpl_parallel.cc

@@ -820,6 +820,10 @@ retry_event_group(rpl_group_info *rgi, rpl_parallel_thread *rpt,
   event_count= 0;
   err= 0;
   errmsg= NULL;
+#ifdef WITH_WSREP
+  thd->wsrep_cs().reset_error();
+  WSREP_DEBUG("retrying async replication event");
+#endif /* WITH_WSREP */
 
   /*
     If we already started committing before getting the deadlock (or other
@@ -920,6 +924,7 @@ retry_event_group(rpl_group_info *rgi, rpl_parallel_thread *rpt,
       err= rgi->worker_error= 1;
       my_error(ER_PRIOR_COMMIT_FAILED, MYF(0));
       mysql_mutex_unlock(&entry->LOCK_parallel_entry);
+
       goto err;
     }
     mysql_mutex_unlock(&entry->LOCK_parallel_entry);
@@ -961,7 +966,17 @@ retry_event_group(rpl_group_info *rgi, rpl_parallel_thread *rpt,
     possibility of an old deadlock kill lingering on beyond this point.
   */
   thd->reset_killed();
+#ifdef WITH_WSREP
+  if (wsrep_before_command(thd))
+  {
+    WSREP_WARN("Parallel slave worker failed at wsrep_before_command() hook");
+    err= 1;
+    goto err;
+  }
+  wsrep_start_trx_if_not_started(thd);
+  WSREP_DEBUG("parallel slave retry, after trx start");
 
+#endif /* WITH_WSREP */
   strmake_buf(log_name, ir->name);
   if ((fd= open_binlog(&rlog, log_name, &errmsg)) <0)
   {

sql/slave.cc

@@ -3908,9 +3908,19 @@ apply_event_and_update_pos_apply(Log_event* ev, THD* thd, rpl_group_info *rgi,
       default:
           WSREP_DEBUG("SQL apply failed, res %d conflict state: %s",
                       exec_res, wsrep_thd_transaction_state_str(thd));
-          rli->abort_slave= 1;
-          rli->report(ERROR_LEVEL, ER_UNKNOWN_COM_ERROR, rgi->gtid_info(),
-                      "Node has dropped from cluster");
+          /*
+            async replication thread should be stopped, if failure was
+            not due to optimistic parallel applying or if node
+            has dropped from cluster
+           */
+          if (thd->system_thread == SYSTEM_THREAD_SLAVE_SQL &&
+              ((rli->mi->using_parallel() &&
+                rli->mi->parallel_mode <= SLAVE_PARALLEL_CONSERVATIVE) ||
+               wsrep_ready == 0)) {
+            rli->abort_slave= 1;
+            rli->report(ERROR_LEVEL, ER_UNKNOWN_COM_ERROR, rgi->gtid_info(),
+                        "Node has dropped from cluster");
+          }
           break;
       }
       mysql_mutex_unlock(&thd->LOCK_thd_data);

storage/innobase/buf/buf0rea.cc

@@ -318,7 +318,8 @@ buf_read_page_low(
 	auto fio = space->io(IORequest(sync
 				       ? IORequest::READ_SYNC
 				       : IORequest::READ_ASYNC),
-			     page_id.page_no() * len, len, dst, bpage);
+			     os_offset_t{page_id.page_no()} * len, len,
+			     dst, bpage);
 
 	if (UNIV_UNLIKELY(fio.err != DB_SUCCESS)) {
 		buf_pool.corrupted_evict(bpage, buf_page_t::READ_FIX);