Merge vajra.(none):/opt/local/work/mysql-5.0-21483

into  vajra.(none):/opt/local/work/mysql-5.1-runtime

sql/sql_insert.cc

@@ -384,6 +384,88 @@ void prepare_triggers_for_insert_stmt(TABLE *table)
 }
 
 
+/**
+  Upgrade table-level lock of INSERT statement to TL_WRITE if
+  a more concurrent lock is infeasible for some reason. This is
+  necessary for engines without internal locking support (MyISAM).
+  An engine with internal locking implementation might later
+  downgrade the lock in handler::store_lock() method.
+*/
+
+void upgrade_lock_type(THD *thd, thr_lock_type *lock_type,
+                       enum_duplicates duplic,
+                       bool is_multi_insert)
+{
+  if (duplic == DUP_UPDATE ||
+      duplic == DUP_REPLACE && *lock_type == TL_WRITE_CONCURRENT_INSERT)
+  {
+    *lock_type= TL_WRITE;
+    return;
+  }
+
+  if (*lock_type == TL_WRITE_DELAYED)
+  {
+#ifdef EMBEDDED_LIBRARY
+    /* No auxiliary threads in the embedded server. */
+    *lock_type= TL_WRITE;
+    return;
+#else
+    /*
+      We do not use delayed threads if:
+      - we're running in the safe mode or skip-new - the feature
+         is disabled in these modes
+      - we're running this query in statement level replication,
+        on a replication slave - because we must ensure serial
+        execution of queries on the slave
+      - it is INSERT .. ON DUPLICATE KEY UPDATE - in this case the
+        insert cannot be concurrent
+    */
+    if (specialflag & (SPECIAL_NO_NEW_FUNC | SPECIAL_SAFE_MODE) ||
+        thd->slave_thread ||
+        thd->variables.max_insert_delayed_threads == 0)
+    {
+      *lock_type= TL_WRITE;
+      return;
+    }
+#endif
+    bool log_on= (thd->options & OPTION_BIN_LOG ||
+                  ! (thd->security_ctx->master_access & SUPER_ACL));
+    if (global_system_variables.binlog_format == BINLOG_FORMAT_STMT &&
+        log_on && mysql_bin_log.is_open() && is_multi_insert)
+    {
+      /*
+        Statement-based binary logging does not work in this case, because:
+        a) two concurrent statements may have their rows intermixed in the
+        queue, leading to autoincrement replication problems on slave (because
+        the values generated used for one statement don't depend only on the
+        value generated for the first row of this statement, so are not
+        replicable)
+        b) if first row of the statement has an error the full statement is
+        not binlogged, while next rows of the statement may be inserted.
+        c) if first row succeeds, statement is binlogged immediately with a
+        zero error code (i.e. "no error"), if then second row fails, query
+        will fail on slave too and slave will stop (wrongly believing that the
+        master got no error).
+        So we fallback to non-delayed INSERT.
+        Note that to be fully correct, we should test the "binlog format which
+        the delayed thread is going to use for this row". But in the common case
+        where the global binlog format is not changed and the session binlog
+        format may be changed, that is equal to the global binlog format.
+        We test it without mutex for speed reasons (condition rarely true), and
+        in the common case (global not changed) it is as good as without mutex;
+        if global value is changed, anyway there is uncertainty as the delayed
+        thread may be old and use the before-the-change value.
+      */
+      *lock_type= TL_WRITE;
+    }
+  }
+}
+
+
+/**
+  INSERT statement implementation
+*/
+
 bool mysql_insert(THD *thd,TABLE_LIST *table_list,
                   List<Item> &fields,
                   List<List_item> &values_list,
@@ -419,67 +501,31 @@ bool mysql_insert(THD *thd,TABLE_LIST *table_list,
   DBUG_ENTER("mysql_insert");
 
   /*
-    in safe mode or with skip-new change delayed insert to be regular
-    if we are told to replace duplicates, the insert cannot be concurrent
-    delayed insert changed to regular in slave thread
+    Upgrade lock type if the requested lock is incompatible with
+    the current connection mode or table operation.
   */
-#ifdef EMBEDDED_LIBRARY
-  if (lock_type == TL_WRITE_DELAYED)
-    lock_type=TL_WRITE;
-#else
-  if ((lock_type == TL_WRITE_DELAYED &&
-       ((specialflag & (SPECIAL_NO_NEW_FUNC | SPECIAL_SAFE_MODE)) ||
-	thd->slave_thread || !thd->variables.max_insert_delayed_threads)) ||
-      (lock_type == TL_WRITE_CONCURRENT_INSERT && duplic == DUP_REPLACE) ||
-      (duplic == DUP_UPDATE))
-    lock_type=TL_WRITE;
-#endif
-  if ((lock_type == TL_WRITE_DELAYED) &&
-      (global_system_variables.binlog_format == BINLOG_FORMAT_STMT) &&
-      log_on && mysql_bin_log.is_open() &&
-      (values_list.elements > 1))
-  {
+  upgrade_lock_type(thd, &table_list->lock_type, duplic,
+                    values_list.elements > 1);
+  lock_type= table_list->lock_type;
+
   /*
-      Statement-based binary logging does not work in this case, because:
-      a) two concurrent statements may have their rows intermixed in the
-      queue, leading to autoincrement replication problems on slave (because
-      the values generated used for one statement don't depend only on the
-      value generated for the first row of this statement, so are not
-      replicable)
-      b) if first row of the statement has an error the full statement is
-      not binlogged, while next rows of the statement may be inserted.
-      c) if first row succeeds, statement is binlogged immediately with a
-      zero error code (i.e. "no error"), if then second row fails, query
-      will fail on slave too and slave will stop (wrongly believing that the
-      master got no error).
-      So we fallback to non-delayed INSERT.
-      Note that to be fully correct, we should test the "binlog format which
-      the delayed thread is going to use for this row". But in the common case
-      where the global binlog format is not changed and the session binlog
-      format may be changed, that is equal to the global binlog format.
-      We test it without mutex for speed reasons (condition rarely true), and
-      in the common case (global not changed) it is as good as without mutex;
-      if global value is changed, anyway there is uncertainty as the delayed
-      thread may be old and use the before-the-change value.
+    We can't write-delayed into a table locked with LOCK TABLES:
+    this will lead to a deadlock, since the delayed thread will
+    never be able to get a lock on the table. QQQ: why not
+    upgrade the lock here instead?
   */
-    lock_type= TL_WRITE;
+  if (lock_type == TL_WRITE_DELAYED && thd->locked_tables &&
+      find_locked_table(thd, table_list->db, table_list->table_name))
+  {
+    my_error(ER_DELAYED_INSERT_TABLE_LOCKED, MYF(0),
+             table_list->table_name);
+    DBUG_RETURN(TRUE);
   }
-  table_list->lock_type= lock_type;
 
 #ifndef EMBEDDED_LIBRARY
   if (lock_type == TL_WRITE_DELAYED)
   {
     res= 1;
-    if (thd->locked_tables)
-    {
-      DBUG_ASSERT(table_list->db); /* Must be set in the parser */
-      if (find_locked_table(thd, table_list->db, table_list->table_name))
-      {
-	my_error(ER_DELAYED_INSERT_TABLE_LOCKED, MYF(0),
-                 table_list->table_name);
-	DBUG_RETURN(TRUE);
-      }
-    }
     if ((table= delayed_get_table(thd,table_list)) && !thd->is_fatal_error)
     {
       /*
@@ -1502,6 +1548,12 @@ class delayed_row :public ilink {
   }
 };
 
+/**
+  delayed_insert - context of a thread responsible for delayed insert
+  into one table. When processing delayed inserts, we create an own
+  thread for every distinct table. Later on all delayed inserts directed
+  into that table are handled by a dedicated thread.
+*/
 
 class delayed_insert :public ilink {
   uint locks_in_memory;
@@ -1598,6 +1650,12 @@ class delayed_insert :public ilink {
 I_List<delayed_insert> delayed_threads;
 
 
+/**
+  Return an instance of delayed insert thread that can handle
+  inserts into a given table, if it exists. Otherwise return NULL.
+*/
+
+static
 delayed_insert *find_handler(THD *thd, TABLE_LIST *table_list)
 {
   thd->proc_info="waiting for delay_list";
@@ -1618,6 +1676,18 @@ delayed_insert *find_handler(THD *thd, TABLE_LIST *table_list)
 }
 
 
+/**
+  Attempt to find or create a delayed insert thread to handle inserts
+  into this table.
+
+  @return Return an instance of the table in the delayed thread
+  @retval  NULL  too many delayed threads OR
+                 this thread ran out of resources OR
+                 a newly created delayed insert thread ran out of resources OR
+                 the delayed insert thread failed to open the table.
+                 In the last three cases an error is set in THD.
+*/
+
 static TABLE *delayed_get_table(THD *thd,TABLE_LIST *table_list)
 {
   int error;
@@ -1726,11 +1796,17 @@ static TABLE *delayed_get_table(THD *thd,TABLE_LIST *table_list)
 }
 
 
-/*
-  As we can't let many threads modify the same TABLE structure, we create
-  an own structure for each tread.  This includes a row buffer to save the
-  column values and new fields that points to the new row buffer.
-  The memory is allocated in the client thread and is freed automaticly.
+/**
+  As we can't let many client threads modify the same TABLE
+  structure of the dedicated delayed insert thread, we create an
+  own structure for each client thread. This includes a row
+  buffer to save the column values and new fields that point to
+  the new row buffer. The memory is allocated in the client
+  thread and is freed automatically.
+
+  @pre This function is called from the client thread.  Delayed
+       insert thread mutex must be acquired before invoking this
+       function.
 */
 
 TABLE *delayed_insert::get_local_table(THD* client_thd)