Speed up connection time:

-Change my_rnd() slightly to make it safer if two threads use it at the same time.
-Avoid some sprintf and strmov in vio.
-Changed thread_count to be automaticly incremented (instead of under LOCK_thread_count).
-Thread cache now uses LOCK_thread_cache instead of LOCK_thread_count.
-Moved delete thd out from LOCK_thread_count.
-Save some mysql_cond_broadcast(&COND_thread_count) calls.
-Removed call to getsockname() during connect.
-Initialize random generator without locks.

Other things:
-Fixed test cases that depends on changes for LOCK_grant
-Added thread_safe_decrement32() and thread_safe_increment32()
-Removed sql_rnd_with_mutex() and get_thread_running()
-In check_table_access() don't lock LOCK_grant if we can resolve the grant with user or db level grants (the normal case).
-Don't use a lock for setting THD->query_id.
-Fixed bug where thd->set_query_id() could be set to same value by multiple threads.

Thanks to Yoshinori Matsunobu for the benchmark of connection speed and to
Domas Mituzas for the inspiration for many of the fixes.


include/violite.h:
  Change desc to a string pointer
mysql-test/suite/perfschema/r/all_instances.result:
  Added new mutex
mysql-test/suite/perfschema/t/func_mutex.test:
  Test for LOCK_system_variables_hash instead of LOCK_grant, as LOCK_grant is not anymore always taken for SELECT's.
mysys/my_gethwaddr.c:
  More DBUG
mysys/my_rnd.c:
  Change my_rnd() slightly to make it safer if two threads use it at the same time.
sql/event_scheduler.cc:
  Changed thread_count to be automically incremented
  Moved some safe things out from LOCK_thread_count.
  Simplify deleting of THD for running thread.
sql/mysqld.cc:
  Changed thread_count to be automically incremented
  Thread cache now uses LOCK_thread_cache instead of LOCK_thread_count
  Added delete_running_thd()
  Moved delete thd out from LOCK_thread_count
  More DBUG
  Only call  mysql_cond_broadcast(&COND_thread_count) if thread_count is 0
  Removed call to getsockname() (old not anymore needed check)
sql/mysqld.h:
  Removed sql_rnd_with_mutex() (not needed anymore)
  Removed not used function get_thread_running()
  Added thread_safe_decrement32() and thread_safe_increment32()
  Simplified dec_thread_running() and inc_thread_running()
  next_query_id() should return the original value for global_query_id, not the next one.
  (Bug introduced with MySQL 5.5 merge).
sql/sql_acl.cc:
  In check_table_access() don't lock LOCK_grant if we can resolve the grant with user or db level grants (the normal case).
sql/sql_class.cc:
  Removed thd_lock_thread_count() and thd_unlock_thread_count()
  Initialize random generator without locks
  Don't use a lock for setting THD->query_id.
  (This is only accessed by thread owning the THD)
sql/sql_class.h:
  Don't use a lock for setting THD->query_id.
sql/sql_insert.cc:
  Changed thread_count to be automically incremented
sql/sql_parse.cc:
  Changed thread_count to be automically incremented
  Fixed bug where thd->set_query_id() could be set to same value by multiple threads.
vio/vio.c:
  Don't generate 'desc' with sprintf/strmov. Assign a pointer instead.
  (Good enough as this is just for debugging)

8 files changed, 207 insertions, 193 deletions

diff --git a/sql/event_scheduler.cc b/sql/event_scheduler.cc
index 8c3e57778ab..4a4ae5a732f 100644
--- a/sql/event_scheduler.cc
+++ b/sql/event_scheduler.cc
@@ -132,11 +132,11 @@ post_init_event_thread(THD *thd)
     return TRUE;
   }
 
+  thread_safe_increment32(&thread_count, &thread_count_lock);
   mysql_mutex_lock(&LOCK_thread_count);
   threads.append(thd);
-  thread_count++;
-  inc_thread_running();
   mysql_mutex_unlock(&LOCK_thread_count);
+  inc_thread_running();
   return FALSE;
 }
 
@@ -154,12 +154,8 @@ deinit_event_thread(THD *thd)
 {
   thd->proc_info= "Clearing";
   DBUG_PRINT("exit", ("Event thread finishing"));
-  mysql_mutex_lock(&LOCK_thread_count);
-  thread_count--;
-  dec_thread_running();
-  delete thd;
-  mysql_cond_broadcast(&COND_thread_count);
-  mysql_mutex_unlock(&LOCK_thread_count);
+
+  delete_running_thd(thd);
 }
 
 
@@ -436,12 +432,7 @@ Event_scheduler::start()
     ret= TRUE;
 
     new_thd->proc_info= "Clearing";
-    mysql_mutex_lock(&LOCK_thread_count);
-    thread_count--;
-    dec_thread_running();
-    delete new_thd;
-    mysql_cond_broadcast(&COND_thread_count);
-    mysql_mutex_unlock(&LOCK_thread_count);
+    delete_running_thd(new_thd);
   }
 end:
   UNLOCK_DATA();
@@ -570,12 +561,7 @@ error:
   if (new_thd)
   {
     new_thd->proc_info= "Clearing";
-    mysql_mutex_lock(&LOCK_thread_count);
-    thread_count--;
-    dec_thread_running();
-    delete new_thd;
-    mysql_cond_broadcast(&COND_thread_count);
-    mysql_mutex_unlock(&LOCK_thread_count);
+    delete_running_thd(new_thd);
   }
   delete event_name;
   DBUG_RETURN(TRUE);
diff --git a/sql/mysqld.cc b/sql/mysqld.cc
index c662b3279ef..fb954908ac9 100644
--- a/sql/mysqld.cc
+++ b/sql/mysqld.cc
@@ -465,7 +465,7 @@ ulong delay_key_write_options;
 uint protocol_version;
 uint lower_case_table_names;
 ulong tc_heuristic_recover= 0;
-uint volatile thread_count;
+int32 thread_count;
 int32 thread_running;
 ulong thread_created;
 ulong back_log, connect_timeout, concurrency, server_id;
@@ -489,6 +489,7 @@ ulong executed_events=0;
 query_id_t global_query_id;
 my_atomic_rwlock_t global_query_id_lock;
 my_atomic_rwlock_t thread_running_lock;
+my_atomic_rwlock_t thread_count_lock;
 my_atomic_rwlock_t statistics_lock;
 ulong aborted_threads, aborted_connects;
 ulong delayed_insert_timeout, delayed_insert_limit, delayed_queue_size;
@@ -663,7 +664,7 @@ SHOW_COMP_OPTION have_openssl;
 
 pthread_key(MEM_ROOT**,THR_MALLOC);
 pthread_key(THD*, THR_THD);
-mysql_mutex_t LOCK_thread_count;
+mysql_mutex_t LOCK_thread_count, LOCK_thread_cache;
 mysql_mutex_t
   LOCK_status, LOCK_error_log, LOCK_short_uuid_generator,
   LOCK_delayed_insert, LOCK_delayed_status, LOCK_delayed_create,
@@ -761,7 +762,8 @@ PSI_mutex_key key_BINLOG_LOCK_index, key_BINLOG_LOCK_xid_list,
   key_relay_log_info_log_space_lock, key_relay_log_info_run_lock,
   key_relay_log_info_sleep_lock,
   key_structure_guard_mutex, key_TABLE_SHARE_LOCK_ha_data,
-  key_LOCK_error_messages, key_LOG_INFO_lock, key_LOCK_thread_count,
+  key_LOCK_error_messages, key_LOG_INFO_lock,
+  key_LOCK_thread_count, key_LOCK_thread_cache,
   key_PARTITION_LOCK_auto_inc;
 PSI_mutex_key key_RELAYLOG_LOCK_index;
 
@@ -832,6 +834,7 @@ static PSI_mutex_info all_server_mutexes[]=
   { &key_LOCK_commit_ordered, "LOCK_commit_ordered", PSI_FLAG_GLOBAL},
   { &key_LOG_INFO_lock, "LOG_INFO::lock", 0},
   { &key_LOCK_thread_count, "LOCK_thread_count", PSI_FLAG_GLOBAL},
+  { &key_LOCK_thread_cache, "LOCK_thread_cache", PSI_FLAG_GLOBAL},
   { &key_PARTITION_LOCK_auto_inc, "HA_DATA_PARTITION::LOCK_auto_inc", 0}
 };
 
@@ -1472,7 +1475,7 @@ static void close_connections(void)
   end_slave();
 
   /* Give threads time to die. */
-  for (int i= 0; thread_count && i < 100; i++)
+  for (int i= 0; *(volatile int32*) &thread_count && i < 100; i++)
     my_sleep(20000);
 
   /*
@@ -1886,6 +1889,7 @@ void clean_up(bool print_message)
   sys_var_end();
   my_atomic_rwlock_destroy(&global_query_id_lock);
   my_atomic_rwlock_destroy(&thread_running_lock);
+  my_atomic_rwlock_destroy(&thread_count_lock);
   my_atomic_rwlock_destroy(&statistics_lock); 
   mysql_mutex_lock(&LOCK_thread_count);
   DBUG_PRINT("quit", ("got thread count lock"));
@@ -1929,6 +1933,7 @@ static void clean_up_mutexes()
   DBUG_ENTER("clean_up_mutexes");
   mysql_rwlock_destroy(&LOCK_grant);
   mysql_mutex_destroy(&LOCK_thread_count);
+  mysql_mutex_destroy(&LOCK_thread_cache);
   mysql_mutex_destroy(&LOCK_status);
   mysql_mutex_destroy(&LOCK_delayed_insert);
   mysql_mutex_destroy(&LOCK_delayed_status);
@@ -2456,6 +2461,28 @@ void dec_connection_count(THD *thd)
 
 
 /*
+  Delete THD and decrement thread counters, including thread_running
+*/
+
+void delete_running_thd(THD *thd)
+{
+  mysql_mutex_lock(&LOCK_thread_count);
+  thd->unlink();
+  mysql_mutex_unlock(&LOCK_thread_count);
+
+  delete thd;
+  dec_thread_running();
+  thread_safe_decrement32(&thread_count, &thread_count_lock);
+  if (!thread_count)
+  {
+    mysql_mutex_lock(&LOCK_thread_count);
+    mysql_cond_broadcast(&COND_thread_count);
+    mysql_mutex_unlock(&LOCK_thread_count);
+  }
+}
+
+
+/*
   Unlink thd from global list of available connections and free thd
 
   SYNOPSIS
@@ -2479,7 +2506,6 @@ void unlink_thd(THD *thd)
   mysql_mutex_unlock(&LOCK_status);
 
   mysql_mutex_lock(&LOCK_thread_count);
-  thread_count--;
   thd->unlink();
   /*
     Used by binlog_reset_master.  It would be cleaner to use
@@ -2487,13 +2513,11 @@ void unlink_thd(THD *thd)
     sync feature has been shut down at this point.
   */
   DBUG_EXECUTE_IF("sleep_after_lock_thread_count_before_delete_thd", sleep(5););
-  /*
-    We must delete thd inside the lock to ensure that we don't start cleanup
-    before THD is deleted
-  */
-  delete thd;
   mysql_mutex_unlock(&LOCK_thread_count);
 
+  delete thd;
+  thread_safe_decrement32(&thread_count, &thread_count_lock);
+
   DBUG_VOID_RETURN;
 }
 
@@ -2505,7 +2529,7 @@ void unlink_thd(THD *thd)
     cache_thread()
 
   NOTES
-    LOCK_thread_count has to be locked
+    LOCK_thread_cache is used to protect the cache variables
 
   RETURN
     0  Thread was not put in cache
@@ -2516,7 +2540,9 @@ void unlink_thd(THD *thd)
 
 static bool cache_thread()
 {
-  mysql_mutex_assert_owner(&LOCK_thread_count);
+  DBUG_ENTER("cache_thread");
+
+  mysql_mutex_lock(&LOCK_thread_cache);
   if (cached_thread_count < thread_cache_size &&
       ! abort_loop && !kill_cached_threads)
   {
@@ -2534,7 +2560,7 @@ static bool cache_thread()
 #endif
 
     while (!abort_loop && ! wake_thread && ! kill_cached_threads)
-      mysql_cond_wait(&COND_thread_cache, &LOCK_thread_count);
+      mysql_cond_wait(&COND_thread_cache, &LOCK_thread_cache);
     cached_thread_count--;
     if (kill_cached_threads)
       mysql_cond_signal(&COND_flush_thread_cache);
@@ -2543,6 +2569,8 @@ static bool cache_thread()
       THD *thd;
       wake_thread--;
       thd= thread_cache.get();
+      mysql_mutex_unlock(&LOCK_thread_cache);
+
       thd->thread_stack= (char*) &thd;          // For store_globals
       (void) thd->store_globals();
 
@@ -2568,11 +2596,16 @@ static bool cache_thread()
       thd->mysys_var->abort= 0;
       thd->thr_create_utime= microsecond_interval_timer();
       thd->start_utime= thd->thr_create_utime;
+
+      /* Link thd into list of all active threads (THD's) */
+      mysql_mutex_lock(&LOCK_thread_count);
       threads.append(thd);
-      return(1);
+      mysql_mutex_unlock(&LOCK_thread_count);
+      DBUG_RETURN(1);
     }
   }
-  return(0);
+  mysql_mutex_unlock(&LOCK_thread_cache);
+  DBUG_RETURN(0);
 }
 
 
@@ -2603,17 +2636,23 @@ bool one_thread_per_connection_end(THD *thd, bool put_in_cache)
   set_current_thd(0);
   if (put_in_cache)
   {
-    mysql_mutex_lock(&LOCK_thread_count);
-    put_in_cache= cache_thread();
-    mysql_mutex_unlock(&LOCK_thread_count);
-    if (put_in_cache)
+    if (cache_thread())
       DBUG_RETURN(0);                             // Thread is reused
   }
 
-  /* It's safe to broadcast outside a lock (COND... is not deleted here) */
-  DBUG_PRINT("signal", ("Broadcasting COND_thread_count"));
-  mysql_cond_broadcast(&COND_thread_count);
-
+  /*
+    It's safe to check for thread_count outside of the mutex
+    as we are only interested to see if it was counted to 0 by the
+    above unlink_thd() call. We should only signal COND_thread_count if
+    thread_count is likely to be 0. (false positives are ok)
+  */
+  if (!thread_count)
+  {
+    mysql_mutex_lock(&LOCK_thread_count);
+    DBUG_PRINT("signal", ("Broadcasting COND_thread_count"));
+    mysql_cond_broadcast(&COND_thread_count);
+    mysql_mutex_unlock(&LOCK_thread_count);
+  }
   DBUG_LEAVE;                                   // Must match DBUG_ENTER()
   my_thread_end();
 
@@ -2624,15 +2663,17 @@ bool one_thread_per_connection_end(THD *thd, bool put_in_cache)
 
 void flush_thread_cache()
 {
-  mysql_mutex_lock(&LOCK_thread_count);
+  DBUG_ENTER("flush_thread_cache");
+  mysql_mutex_lock(&LOCK_thread_cache);
   kill_cached_threads++;
   while (cached_thread_count)
   {
     mysql_cond_broadcast(&COND_thread_cache);
-    mysql_cond_wait(&COND_flush_thread_cache, &LOCK_thread_count);
+    mysql_cond_wait(&COND_flush_thread_cache, &LOCK_thread_cache);
   }
   kill_cached_threads--;
-  mysql_mutex_unlock(&LOCK_thread_count);
+  mysql_mutex_unlock(&LOCK_thread_cache);
+  DBUG_VOID_RETURN;
 }
 
 
@@ -3973,6 +4014,7 @@ static int init_thread_environment()
 {
   DBUG_ENTER("init_thread_environment");
   mysql_mutex_init(key_LOCK_thread_count, &LOCK_thread_count, MY_MUTEX_INIT_FAST);
+  mysql_mutex_init(key_LOCK_thread_cache, &LOCK_thread_cache, MY_MUTEX_INIT_FAST);
   mysql_mutex_init(key_LOCK_status, &LOCK_status, MY_MUTEX_INIT_FAST);
   mysql_mutex_init(key_LOCK_delayed_insert,
                    &LOCK_delayed_insert, MY_MUTEX_INIT_FAST);
@@ -5363,7 +5405,7 @@ static void bootstrap(MYSQL_FILE *file)
   thd->max_client_packet_length= thd->net.max_packet;
   thd->security_ctx->master_access= ~(ulong)0;
   thd->thread_id= thd->variables.pseudo_thread_id= thread_id++;
-  thread_count++;
+  thread_count++;                        // Safe as only one thread running
   in_bootstrap= TRUE;
 
   bootstrap_file=file;
@@ -5444,54 +5486,69 @@ void handle_connection_in_main_thread(THD *thd)
 
 void create_thread_to_handle_connection(THD *thd)
 {
+  DBUG_ENTER("create_thread_to_handle_connection");
+  mysql_mutex_assert_owner(&LOCK_thread_count);
+
+  /* Check if we can get thread from the cache */
   if (cached_thread_count > wake_thread)
   {
-    /* Get thread from cache */
-    thread_cache.push_back(thd);
-    wake_thread++;
-    mysql_cond_signal(&COND_thread_cache);
-  }
-  else
-  {
-    char error_message_buff[MYSQL_ERRMSG_SIZE];
-    /* Create new thread to handle connection */
-    int error;
-    thread_created++;
-    threads.append(thd);
-    DBUG_PRINT("info",(("creating thread %lu"), thd->thread_id));
-    thd->prior_thr_create_utime= microsecond_interval_timer();
-    if ((error= mysql_thread_create(key_thread_one_connection,
-                                    &thd->real_id, &connection_attrib,
-                                    handle_one_connection,
-                                    (void*) thd)))
+    mysql_mutex_lock(&LOCK_thread_cache);
+    /* Recheck condition when we have the lock */
+    if (cached_thread_count > wake_thread)
     {
-      /* purecov: begin inspected */
-      DBUG_PRINT("error",
-                 ("Can't create thread to handle request (error %d)",
-                  error));
-      thread_count--;
-      thd->killed= KILL_CONNECTION;             // Safety
       mysql_mutex_unlock(&LOCK_thread_count);
+      /* Get thread from cache */
+      thread_cache.push_back(thd);
+      wake_thread++;
+      mysql_cond_signal(&COND_thread_cache);
+      mysql_mutex_unlock(&LOCK_thread_cache);
+      DBUG_PRINT("info",("Thread created"));
+      DBUG_VOID_RETURN;
+    }
+    mysql_mutex_unlock(&LOCK_thread_cache);
+  }
 
-      mysql_mutex_lock(&LOCK_connection_count);
-      (*thd->scheduler->connection_count)--;
-      mysql_mutex_unlock(&LOCK_connection_count);
+  char error_message_buff[MYSQL_ERRMSG_SIZE];
+  /* Create new thread to handle connection */
+  int error;
+  thread_created++;
+  threads.append(thd);
+  DBUG_PRINT("info",(("creating thread %lu"), thd->thread_id));
+  thd->prior_thr_create_utime= microsecond_interval_timer();
+  if ((error= mysql_thread_create(key_thread_one_connection,
+                                  &thd->real_id, &connection_attrib,
+                                  handle_one_connection,
+                                  (void*) thd)))
+  {
+    /* purecov: begin inspected */
+    DBUG_PRINT("error",
+               ("Can't create thread to handle request (error %d)",
+                error));
+    thd->killed= KILL_CONNECTION;             // Safety
+    mysql_mutex_unlock(&LOCK_thread_count);
 
-      statistic_increment(aborted_connects,&LOCK_status);
-      /* Can't use my_error() since store_globals has not been called. */
-      my_snprintf(error_message_buff, sizeof(error_message_buff),
-                  ER_THD(thd, ER_CANT_CREATE_THREAD), error);
-      net_send_error(thd, ER_CANT_CREATE_THREAD, error_message_buff, NULL);
-      close_connection(thd, ER_OUT_OF_RESOURCES);
-      mysql_mutex_lock(&LOCK_thread_count);
-      delete thd;
-      mysql_mutex_unlock(&LOCK_thread_count);
-      return;
-      /* purecov: end */
-    }
+    mysql_mutex_lock(&LOCK_connection_count);
+    (*thd->scheduler->connection_count)--;
+    mysql_mutex_unlock(&LOCK_connection_count);
+
+    statistic_increment(aborted_connects,&LOCK_status);
+    /* Can't use my_error() since store_globals has not been called. */
+    my_snprintf(error_message_buff, sizeof(error_message_buff),
+                ER_THD(thd, ER_CANT_CREATE_THREAD), error);
+    net_send_error(thd, ER_CANT_CREATE_THREAD, error_message_buff, NULL);
+    close_connection(thd, ER_OUT_OF_RESOURCES);
+
+    mysql_mutex_lock(&LOCK_thread_count);
+    thd->unlink();
+    mysql_mutex_unlock(&LOCK_thread_count);
+    delete thd;
+    thread_safe_decrement32(&thread_count, &thread_count_lock);
+    return;
+    /* purecov: end */
   }
   mysql_mutex_unlock(&LOCK_thread_count);
   DBUG_PRINT("info",("Thread created"));
+  DBUG_VOID_RETURN;
 }
 
 
@@ -5538,10 +5595,10 @@ static void create_new_thread(THD *thd)
 
   mysql_mutex_unlock(&LOCK_connection_count);
 
-  /* Start a new thread to handle connection. */
+  thread_safe_increment32(&thread_count, &thread_count_lock);
 
+  /* Start a new thread to handle connection. */
   mysql_mutex_lock(&LOCK_thread_count);
-
   /*
     The initialization of thread_id is done in create_embedded_thd() for
     the embedded library.
@@ -5549,8 +5606,6 @@ static void create_new_thread(THD *thd)
   */
   thd->thread_id= thd->variables.pseudo_thread_id= thread_id++;
 
-  thread_count++;
-
   MYSQL_CALLBACK(thd->scheduler, add_connection, (thd));
 
   DBUG_VOID_RETURN;
@@ -5758,20 +5813,6 @@ void handle_connections_sockets()
     }
 #endif /* HAVE_LIBWRAP */
 
-    {
-      size_socket dummyLen;
-      struct sockaddr_storage dummy;
-      dummyLen = sizeof(dummy);
-      if (  getsockname(new_sock,(struct sockaddr *)&dummy, 
-                  (SOCKET_SIZE_TYPE *)&dummyLen) < 0  )
-      {
-	sql_perror("Error on new connection socket");
-	(void) mysql_socket_shutdown(new_sock, SHUT_RDWR);
-	(void) closesocket(new_sock);
-	continue;
-      }
-    }
-
     /*
     ** Don't allow too many connections
     */
@@ -7402,6 +7443,7 @@ static int mysql_init_variables(void)
   global_query_id= thread_id= 1L;
   my_atomic_rwlock_init(&global_query_id_lock);
   my_atomic_rwlock_init(&thread_running_lock);
+  my_atomic_rwlock_init(&thread_count_lock);
   my_atomic_rwlock_init(&statistics_lock);
   strmov(server_version, MYSQL_SERVER_VERSION);
   threads.empty();
@@ -8461,12 +8503,8 @@ void refresh_status(THD *thd)
 
   /*
     Set max_used_connections to the number of currently open
-    connections.  Lock LOCK_thread_count out of LOCK_status to avoid
-    deadlocks.  Status reset becomes not atomic, but status data is
-    not exact anyway.
+    connections.  This is not perfect, but status data is not exact anyway.
   */
-  mysql_mutex_lock(&LOCK_thread_count);
   max_used_connections= thread_count-delayed_insert_threads;
-  mysql_mutex_unlock(&LOCK_thread_count);
 }
 
diff --git a/sql/mysqld.h b/sql/mysqld.h
index 6bde25f08fb..587673de37c 100644
--- a/sql/mysqld.h
+++ b/sql/mysqld.h
@@ -56,6 +56,7 @@ void kill_mysql(void);
 void close_connection(THD *thd, uint sql_errno= 0);
 void handle_connection_in_main_thread(THD *thd);
 void create_thread_to_handle_connection(THD *thd);
+void delete_running_thd(THD *thd);
 void unlink_thd(THD *thd);
 bool one_thread_per_connection_end(THD *thd, bool put_in_cache);
 void flush_thread_cache();
@@ -89,7 +90,6 @@ extern bool opt_ignore_builtin_innodb;
 extern my_bool opt_character_set_client_handshake;
 extern bool volatile abort_loop;
 extern bool in_bootstrap;
-extern uint volatile thread_count;
 extern uint connection_count;
 extern my_bool opt_safe_user_create;
 extern my_bool opt_safe_show_db, opt_local_infile, opt_myisam_use_mmap;
@@ -352,7 +352,8 @@ extern mysql_rwlock_t LOCK_system_variables_hash;
 extern mysql_cond_t COND_thread_count;
 extern mysql_cond_t COND_manager;
 extern int32 thread_running;
-extern my_atomic_rwlock_t thread_running_lock;
+extern int32 thread_count;
+extern my_atomic_rwlock_t thread_running_lock, thread_count_lock;
 
 extern char *opt_ssl_ca, *opt_ssl_capath, *opt_ssl_cert, *opt_ssl_cipher,
             *opt_ssl_key;
@@ -449,7 +450,7 @@ inline query_id_t next_query_id()
   my_atomic_rwlock_wrlock(&global_query_id_lock);
   id= my_atomic_add64(&global_query_id, 1);
   my_atomic_rwlock_wrunlock(&global_query_id_lock);
-  return (id+1);
+  return (id);
 }
 
 inline query_id_t get_query_id()
@@ -479,42 +480,30 @@ inline void table_case_convert(char * name, uint length)
                                      name, length, name, length);
 }
 
-inline ulong sql_rnd_with_mutex()
+inline void thread_safe_increment32(int32 *value, my_atomic_rwlock_t *lock)
 {
-  mysql_mutex_lock(&LOCK_thread_count);
-  ulong tmp=(ulong) (my_rnd(&sql_rand) * 0xffffffff); /* make all bits random */
-  mysql_mutex_unlock(&LOCK_thread_count);
-  return tmp;
+  my_atomic_rwlock_wrlock(lock);
+  (void) my_atomic_add32(value, 1);
+  my_atomic_rwlock_wrunlock(lock);
 }
 
-inline int32
-inc_thread_running()
+inline void thread_safe_decrement32(int32 *value, my_atomic_rwlock_t *lock)
 {
-  int32 num_thread_running;
-  my_atomic_rwlock_wrlock(&thread_running_lock);
-  num_thread_running= my_atomic_add32(&thread_running, 1);
-  my_atomic_rwlock_wrunlock(&thread_running_lock);
-  return (num_thread_running+1);
+  my_atomic_rwlock_wrlock(lock);
+  (void) my_atomic_add32(value, -1);
+  my_atomic_rwlock_wrunlock(lock);
 }
 
-inline int32
-dec_thread_running()
+inline void
+inc_thread_running()
 {
-  int32 num_thread_running;
-  my_atomic_rwlock_wrlock(&thread_running_lock);
-  num_thread_running= my_atomic_add32(&thread_running, -1);
-  my_atomic_rwlock_wrunlock(&thread_running_lock);
-  return (num_thread_running-1);
+  thread_safe_increment32(&thread_running, &thread_running_lock);
 }
 
-inline int32
-get_thread_running()
+inline void
+dec_thread_running()
 {
-  int32 num_thread_running;
-  my_atomic_rwlock_wrlock(&thread_running_lock);
-  num_thread_running= my_atomic_load32(&thread_running);
-  my_atomic_rwlock_wrunlock(&thread_running_lock);
-  return num_thread_running;
+  thread_safe_decrement32(&thread_running, &thread_running_lock);
 }
 
 void set_server_version(void);
diff --git a/sql/sql_acl.cc b/sql/sql_acl.cc
index 2f854251ce5..25b54abf983 100644
--- a/sql/sql_acl.cc
+++ b/sql/sql_acl.cc
@@ -4530,12 +4530,16 @@ end:
   @see check_access
   @see check_table_access
 
-  @note This functions assumes that either number of tables to be inspected
+  @note
+     This functions assumes that either number of tables to be inspected
      by it is limited explicitly (i.e. is is not UINT_MAX) or table list
      used and thd->lex->query_tables_own_last value correspond to each
      other (the latter should be either 0 or point to next_global member
      of one of elements of this table list).
 
+     We delay locking of LOCK_grant until we really need it as we assume that
+     most privileges be resolved with user or db level accesses.
+
    @return Access status
      @retval FALSE Access granted; But column privileges might need to be
       checked.
@@ -4552,6 +4556,8 @@ bool check_grant(THD *thd, ulong want_access, TABLE_LIST *tables,
   Security_context *sctx= thd->security_ctx;
   uint i;
   ulong orig_want_access= want_access;
+  my_bool locked= 0;
+  GRANT_TABLE *grant_table;
   DBUG_ENTER("check_grant");
   DBUG_ASSERT(number > 0);
 
@@ -4575,11 +4581,9 @@ bool check_grant(THD *thd, ulong want_access, TABLE_LIST *tables,
     */
     tl->grant.orig_want_privilege= (want_access & ~SHOW_VIEW_ACL);
   }
+  number= i;
 
-  mysql_rwlock_rdlock(&LOCK_grant);
-  for (tl= tables;
-       tl && number-- && tl != first_not_own_table;
-       tl= tl->next_global)
+  for (tl= tables; number-- ; tl= tl->next_global)
   {
     sctx = test(tl->security_ctx) ? tl->security_ctx : thd->security_ctx;
 
@@ -4632,13 +4636,18 @@ bool check_grant(THD *thd, ulong want_access, TABLE_LIST *tables,
       }
       continue;
     }
-    GRANT_TABLE *grant_table= table_hash_search(sctx->host, sctx->ip,
-                                                tl->get_db_name(),
-                                                sctx->priv_user,
-                                                tl->get_table_name(),
-                                                FALSE);
 
-    if (!grant_table)
+    if (!locked)
+    {
+      locked= 1;
+      mysql_rwlock_rdlock(&LOCK_grant);
+    }
+
+    if (!(grant_table= table_hash_search(sctx->host, sctx->ip,
+                                         tl->get_db_name(),
+                                         sctx->priv_user,
+                                         tl->get_table_name(),
+                                         FALSE)))
     {
       want_access &= ~tl->grant.privilege;
       goto err;					// No grants
@@ -4665,11 +4674,13 @@ bool check_grant(THD *thd, ulong want_access, TABLE_LIST *tables,
       goto err;					// impossible
     }
   }
-  mysql_rwlock_unlock(&LOCK_grant);
+  if (locked)
+    mysql_rwlock_unlock(&LOCK_grant);
   DBUG_RETURN(FALSE);
 
 err:
-  mysql_rwlock_unlock(&LOCK_grant);
+  if (locked)
+    mysql_rwlock_unlock(&LOCK_grant);
   if (!no_errors)				// Not a silent skip of table
   {
     char command[128];
diff --git a/sql/sql_class.cc b/sql/sql_class.cc
index 155c6365e76..9fd3827ab39 100644
--- a/sql/sql_class.cc
+++ b/sql/sql_class.cc
@@ -350,29 +350,6 @@ void thd_set_thread_stack(THD *thd, char *stack_start)
 }
 
 /**
-  Lock connection data for the set of connections this connection
-  belongs to
-
-  @param thd                       THD object
-*/
-void thd_lock_thread_count(THD *)
-{
-  mysql_mutex_lock(&LOCK_thread_count);
-}
-
-/**
-  Lock connection data for the set of connections this connection
-  belongs to
-
-  @param thd                       THD object
-*/
-void thd_unlock_thread_count(THD *)
-{
-  mysql_cond_broadcast(&COND_thread_count);
-  mysql_mutex_unlock(&LOCK_thread_count);
-}
-
-/**
   Close the socket used by this connection
 
   @param thd                THD object
@@ -947,7 +924,14 @@ THD::THD()
   protocol_binary.init(this);
 
   tablespace_op=FALSE;
-  tmp= sql_rnd_with_mutex();
+
+  /*
+    Initialize the random generator. We call my_rnd() without a lock as
+    it's not really critical if two threads modifies the structure at the
+    same time.  We ensure that we have an unique number foreach thread
+    by adding the address of the stack.
+  */
+  tmp= (ulong) (my_rnd(&sql_rand) * 0xffffffff);
   my_rnd_init(&rand, tmp + (ulong) &rand, tmp + (ulong) ::global_query_id);
   substitute_null_with_insert_id = FALSE;
   thr_lock_info_init(&lock_info); /* safety: will be reset after start */
@@ -4342,17 +4326,8 @@ void THD::set_query_and_id(char *query_arg, uint32 query_length_arg,
 {
   mysql_mutex_lock(&LOCK_thd_data);
   set_query_inner(query_arg, query_length_arg, cs);
-  query_id= new_query_id;
   mysql_mutex_unlock(&LOCK_thd_data);
-}
-
-/** Assign a new value to thd->query_id.  */
-
-void THD::set_query_id(query_id_t new_query_id)
-{
-  mysql_mutex_lock(&LOCK_thd_data);
   query_id= new_query_id;
-  mysql_mutex_unlock(&LOCK_thd_data);
 }
 
 /** Assign a new value to thd->mysys_var.  */
diff --git a/sql/sql_class.h b/sql/sql_class.h
index 87b9783c1ed..01558603f68 100644
--- a/sql/sql_class.h
+++ b/sql/sql_class.h
@@ -3076,7 +3076,10 @@ public:
   { set_query(CSET_STRING()); }
   void set_query_and_id(char *query_arg, uint32 query_length_arg,
                         CHARSET_INFO *cs, query_id_t new_query_id);
-  void set_query_id(query_id_t new_query_id);
+  void set_query_id(query_id_t new_query_id)
+  {
+    query_id= new_query_id;
+  }
   void set_open_tables(TABLE *open_tables_arg)
   {
     mysql_mutex_lock(&LOCK_thd_data);
diff --git a/sql/sql_insert.cc b/sql/sql_insert.cc
index 833e134bff9..efbf35fe575 100644
--- a/sql/sql_insert.cc
+++ b/sql/sql_insert.cc
@@ -2038,9 +2038,9 @@ public:
     thd.unlink();				// Must be unlinked under lock
     my_free(thd.query());
     thd.security_ctx->user= thd.security_ctx->host=0;
-    thread_count--;
     delayed_insert_threads--;
     mysql_mutex_unlock(&LOCK_thread_count);
+    thread_safe_decrement32(&thread_count, &thread_count_lock);
     mysql_cond_broadcast(&COND_thread_count); /* Tell main we are ready */
   }
 
@@ -2175,9 +2175,9 @@ bool delayed_get_table(THD *thd, MDL_request *grl_protection_request,
     {
       if (!(di= new Delayed_insert()))
         goto end_create;
-      mysql_mutex_lock(&LOCK_thread_count);
-      thread_count++;
-      mysql_mutex_unlock(&LOCK_thread_count);
+
+      thread_safe_increment32(&thread_count, &thread_count_lock);
+
       /*
         Annotating delayed inserts is not supported.
       */
diff --git a/sql/sql_parse.cc b/sql/sql_parse.cc
index 0abb249d97b..bdf2bd17589 100644
--- a/sql/sql_parse.cc
+++ b/sql/sql_parse.cc
@@ -645,9 +645,10 @@ end:
   delete thd;
 
 #ifndef EMBEDDED_LIBRARY
-  mysql_mutex_lock(&LOCK_thread_count);
-  thread_count--;
+  thread_safe_decrement32(&thread_count, &thread_count_lock);
   in_bootstrap= FALSE;
+
+  mysql_mutex_lock(&LOCK_thread_count);
   mysql_cond_broadcast(&COND_thread_count);
   mysql_mutex_unlock(&LOCK_thread_count);
   my_thread_end();
@@ -930,9 +931,16 @@ bool dispatch_command(enum enum_server_command command, THD *thd,
   thd->query_plan_flags= QPLAN_INIT;
   thd->lex->sql_command= SQLCOM_END; /* to avoid confusing VIEW detectors */
   thd->set_time();
-  thd->set_query_id(get_query_id());
   if (!(server_command_flags[command] & CF_SKIP_QUERY_ID))
-    next_query_id();
+    thd->set_query_id(next_query_id());
+  else
+  {
+    /*
+      ping, get statistics or similar stateless command.
+      No reason to increase query id here.
+    */
+    thd->set_query_id(get_query_id());
+  }
   inc_thread_running();
 
   if (!(server_command_flags[command] & CF_SKIP_QUESTIONS))
@@ -5021,6 +5029,10 @@ check_access(THD *thd, ulong want_access, const char *db, ulong *save_priv,
 
   if ((db != NULL) && (db != any_db))
   {
+    /*
+      Check if this is reserved database, like information schema or
+      performance schema
+    */
     const ACL_internal_schema_access *access;
     access= get_cached_schema_access(grant_internal_info, db);
     if (access)