MDEV-25389 : Assertion `!is_thread_specific || (mysqld_server_initialized && thd)' failed in void my_malloc_size_cb_func(long long int, my_bool)

If wsrep slave thread creation fails for some reason we need to handle
this error correctly and set actual running slave threads accordingly.

6 files changed, 56 insertions, 6 deletions

diff --git a/mysql-test/suite/galera/r/MDEV-25389.result b/mysql-test/suite/galera/r/MDEV-25389.result
new file mode 100644
index 00000000000..9a142676d92
--- /dev/null
+++ b/mysql-test/suite/galera/r/MDEV-25389.result
@@ -0,0 +1,15 @@
+connection node_2;
+connection node_1;
+connection node_2;
+call mtr.add_suppression("WSREP: Failed to create/initialize system thread");
+SET GLOBAL debug_dbug='+d,simulate_failed_connection_1';
+SET GLOBAL wsrep_slave_threads=2;
+ERROR HY000: Incorrect arguments to SET
+SELECT @@wsrep_slave_threads;
+@@wsrep_slave_threads
+1
+SET GLOBAL debug_dbug='';
+SET GLOBAL wsrep_slave_threads=1;
+SELECT @@wsrep_slave_threads;
+@@wsrep_slave_threads
+1
diff --git a/mysql-test/suite/galera/t/MDEV-25389.test b/mysql-test/suite/galera/t/MDEV-25389.test
new file mode 100644
index 00000000000..dcb21ee6136
--- /dev/null
+++ b/mysql-test/suite/galera/t/MDEV-25389.test
@@ -0,0 +1,13 @@
+--source include/galera_cluster.inc
+--source include/have_debug.inc
+--source include/have_debug_sync.inc
+
+--connection node_2
+call mtr.add_suppression("WSREP: Failed to create/initialize system thread");
+SET GLOBAL debug_dbug='+d,simulate_failed_connection_1';
+--error ER_WRONG_ARGUMENTS
+SET GLOBAL wsrep_slave_threads=2;
+SELECT @@wsrep_slave_threads;
+SET GLOBAL debug_dbug='';
+SET GLOBAL wsrep_slave_threads=1;
+SELECT @@wsrep_slave_threads;
diff --git a/sql/wsrep_mysqld.cc b/sql/wsrep_mysqld.cc
index d7cbc545161..f4efaeee42e 100644
--- a/sql/wsrep_mysqld.cc
+++ b/sql/wsrep_mysqld.cc
@@ -126,6 +126,7 @@ ulong wsrep_trx_fragment_unit= WSREP_FRAG_BYTES;
 ulong wsrep_SR_store_type= WSREP_SR_STORE_TABLE;
 uint  wsrep_ignore_apply_errors= 0;
 
+std::atomic <bool> wsrep_thread_create_failed;
 
 /*
  * End configuration options
@@ -2940,7 +2941,9 @@ void* start_wsrep_THD(void *arg)
   {
     close_connection(thd, ER_OUT_OF_RESOURCES);
     statistic_increment(aborted_connects,&LOCK_status);
-    MYSQL_CALLBACK(thread_scheduler, end_thread, (thd, 0));
+    // This will signal error to wsrep_slave_threads_update
+    wsrep_thread_create_failed.store(true, std::memory_order_relaxed);
+    WSREP_DEBUG("start_wsrep_THD: init_new_connection_thread failed");
     goto error;
   }
 
diff --git a/sql/wsrep_mysqld.h b/sql/wsrep_mysqld.h
index 3fc0e397514..201e3e1a44a 100644
--- a/sql/wsrep_mysqld.h
+++ b/sql/wsrep_mysqld.h
@@ -1,4 +1,4 @@
-/* Copyright 2008-2021 Codership Oy <http://www.codership.com>
+/* Copyright 2008-2022 Codership Oy <http://www.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
@@ -88,6 +88,7 @@ extern ulong       wsrep_running_rollbacker_threads;
 extern bool        wsrep_new_cluster;
 extern bool        wsrep_gtid_mode;
 extern uint32      wsrep_gtid_domain_id;
+extern std::atomic <bool > wsrep_thread_create_failed;
 
 enum enum_wsrep_reject_types {
   WSREP_REJECT_NONE,    /* nothing rejected */
diff --git a/sql/wsrep_thd.cc b/sql/wsrep_thd.cc
index 8853ea6a499..5922b39e718 100644
--- a/sql/wsrep_thd.cc
+++ b/sql/wsrep_thd.cc
@@ -1,4 +1,4 @@
-/* Copyright (C) 2013-2021 Codership Oy <info@codership.com>
+/* Copyright (C) 2013-2022 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
diff --git a/sql/wsrep_var.cc b/sql/wsrep_var.cc
index 8919715b188..dcca0ed08c2 100644
--- a/sql/wsrep_var.cc
+++ b/sql/wsrep_var.cc
@@ -1,4 +1,4 @@
-/* Copyright 2008-2021 Codership Oy <http://www.codership.com>
+/* Copyright 2008-2022 Codership Oy <http://www.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
@@ -753,12 +753,30 @@ bool wsrep_slave_threads_update (sys_var *self, THD* thd, enum_var_type type)
   if (wsrep_slave_count_change > 0)
   {
     WSREP_DEBUG("Creating %d applier threads, total %ld", wsrep_slave_count_change, wsrep_slave_threads);
+    wsrep_thread_create_failed.store(false, std::memory_order_relaxed);
     res= wsrep_create_appliers(wsrep_slave_count_change, true);
+    mysql_mutex_unlock(&LOCK_global_system_variables);
+    mysql_mutex_unlock(&LOCK_wsrep_slave_threads);
+    // Thread creation and execution is asyncronous, therefore we need
+    // wait them to be started or error produced
+    while (wsrep_running_applier_threads != (ulong)wsrep_slave_threads &&
+	   !wsrep_thread_create_failed.load(std::memory_order_relaxed))
+    {
+	    my_sleep(1000);
+    }
+
+    mysql_mutex_lock(&LOCK_global_system_variables);
+
+    if (wsrep_thread_create_failed.load(std::memory_order_relaxed)) {
+      wsrep_slave_threads= wsrep_running_applier_threads;
+      return true;
+    }
+
     WSREP_DEBUG("Running %lu applier threads", wsrep_running_applier_threads);
     wsrep_slave_count_change = 0;
   }
-
-  mysql_mutex_unlock(&LOCK_wsrep_slave_threads);
+  else
+     mysql_mutex_unlock(&LOCK_wsrep_slave_threads);
 
   return res;
 }