MDEV-37354 : SIGSEGV in Wsrep_server_service::release_high_priority_service | discard_streaming_applier

This issue seems to be already fixed. However, to avoid future problems:

Wsrep_server_service::release_storage_service
  Add assertion that storage service is not nullptr and
  contains thd. In production binaries add guard to
  not use nullptr.

Wsrep_server_service::release_high_priority_service
  Add assertion that high_priority service is not nullptr and
  contains thd. In production binaries add guard to
  not use nullptr.

wsrep_is_BF_lock_timeout
  Remove printing of record lock because its page might
  not be latched leading to assertion in multi-master
  testing.

Author: janlindstrom

mysql-test/suite/galera/r/galera_mdev_31517.result

@@ -1,5 +1,9 @@
 connection node_2;
 connection node_1;
+# Make sure that the test is operating on the right version of galera library.
+# Correct Galera library found
+connection node_1;
+connection node_2;
 connection node_2;
 CALL mtr.add_suppression("unknown variable 'non_existing_variable=ON'");
 CALL mtr.add_suppression("Aborting");

mysql-test/suite/galera/t/galera_mdev_31517.test

@@ -24,6 +24,11 @@
 --let $galera_version=26.4.25
 source ../wsrep/include/check_galera_version.inc;
 
+# Save original auto_increment_offset values.
+--let $node_1=node_1
+--let $node_2=node_2
+--source include/auto_increment_offset_save.inc
+
 # Suppress expected errors and warnings:
 --connection node_2
 CALL mtr.add_suppression("unknown variable 'non_existing_variable=ON'");
@@ -129,4 +134,5 @@ DROP TABLE t;
 EOF
 --remove_file $TEST_LOG.copy
 
+--source include/auto_increment_offset_restore.inc
 --source include/galera_end.inc

sql/wsrep_server_service.cc

@@ -80,11 +80,22 @@ void Wsrep_server_service::release_storage_service(
 {
   Wsrep_storage_service* ss=
     static_cast<Wsrep_storage_service*>(storage_service);
-  THD* thd= ss->m_thd;
-  wsrep_reset_threadvars(thd);
-  server_threads.erase(thd);
-  delete ss;
-  delete thd;
+  DBUG_ASSERT(ss && ss->m_thd);
+
+  // Do not crash server on production
+  if (ss)
+  {
+    THD* thd= ss->m_thd;
+    if (thd)
+    {
+      wsrep_reset_threadvars(thd);
+      server_threads.erase(thd);
+      delete ss;
+      delete thd;
+    }
+    else
+      delete ss;
+  }
 }
 
 Wsrep_applier_service*
@@ -141,12 +152,23 @@ void Wsrep_server_service::release_high_priority_service(wsrep::high_priority_se
 {
   Wsrep_high_priority_service* hps=
     static_cast<Wsrep_high_priority_service*>(high_priority_service);
-  THD* thd= hps->m_thd;
-  delete hps;
-  wsrep_store_threadvars(thd);
-  server_threads.erase(thd);
-  delete thd;
-  wsrep_delete_threadvars();
+  DBUG_ASSERT(hps && hps->m_thd);
+
+  // Do not crash server on production
+  if (hps)
+  {
+    THD* thd= hps->m_thd;
+    if (thd)
+    {
+      delete hps;
+      wsrep_store_threadvars(thd);
+      server_threads.erase(thd);
+      delete thd;
+      wsrep_delete_threadvars();
+    }
+    else
+      delete hps;
+  }
 }
 
 void Wsrep_server_service::background_rollback(

storage/innobase/lock/lock0lock.cc

@@ -670,25 +670,9 @@ bool wsrep_is_BF_lock_timeout(const trx_t &trx)
              << " error: " << trx.error_state
              << " query: " << wsrep_thd_query(trx.mysql_thd);
 
-  if (const lock_t*wait_lock = trx.lock.wait_lock)
-  {
-    const my_hrtime_t now= my_hrtime_coarse();
-    const my_hrtime_t suspend_time= trx.lock.suspend_time;
-    fprintf(stderr,
-            "------- TRX HAS BEEN WAITING %llu us"
-            " FOR THIS LOCK TO BE GRANTED:\n",
-            now.val - suspend_time.val);
-
-    if (!wait_lock->is_table()) {
-      mtr_t mtr;
-      lock_rec_print(stderr, wait_lock, mtr);
-    } else {
-      lock_table_print(stderr, wait_lock);
-    }
-
-    fprintf(stderr, "------------------\n");
-  }
-
+  // TODO: Can't use lock_rec_print from here because
+  // record lock page might not be latched and we are
+  // actually only interested lock information.
   return true;
 }