Many files:

Small improvements
row0mysql.c:
  Small improvements + fix the ALTER TABLE problem by introducing a lazy drop table it can use
ha_innobase.cc:
  Some fine-tuning of optimization


sql/ha_innobase.cc:
  Some fine-tuning of optimization
innobase/include/log0recv.h:
  Small improvements
innobase/include/mem0mem.h:
  Small improvements
innobase/include/mem0pool.h:
  Small improvements
innobase/include/row0mysql.h:
  Small improvements
innobase/include/srv0srv.h:
  Small improvements
innobase/include/trx0trx.h:
  Small improvements
innobase/buf/buf0lru.c:
  Small improvements
innobase/fil/fil0fil.c:
  Small improvements
innobase/log/log0recv.c:
  Small improvements
innobase/mem/mem0mem.c:
  Small improvements
innobase/mem/mem0pool.c:
  Small improvements
innobase/row/row0mysql.c:
  Small improvements + fix the ALTER TABLE problem by introducing a lazy drop table it can use
innobase/srv/srv0srv.c:
  Small improvements
innobase/srv/srv0start.c:
  Small improvements
innobase/trx/trx0purge.c:
  Small improvements
innobase/trx/trx0trx.c:
  Small improvements

innobase/buf/buf0lru.c

@@ -242,7 +242,6 @@ loop:
 	if (n_iterations > 30) {
 		ut_print_timestamp(stderr);
 		fprintf(stderr,
-		" ***********************************************\n"
 		"InnoDB: Warning: difficult to find free blocks from\n"
 		"InnoDB: the buffer pool (%lu search iterations)! Consider\n"
 		"InnoDB: increasing the buffer pool size.\n",

innobase/fil/fil0fil.c

@@ -1088,7 +1088,15 @@ loop:
 	node = UT_LIST_GET_FIRST(space->chain);
 
 	for (;;) {
-		ut_a(node);
+		if (node == NULL) {
+			fprintf(stderr,
+	"InnoDB: Error: trying to access page number %lu in space %lu\n"
+	"InnoDB: which is outside the tablespace bounds.\n"
+	"InnoDB: Byte offset %lu, len %lu, i/o type %lu\n", 
+ 			block_offset, space_id, byte_offset, len, type);
+ 			
+			ut_a(0);
+		}
 
 		if (node->size > block_offset) {
 			/* Found! */

innobase/include/log0recv.h

@@ -258,6 +258,7 @@ struct recv_sys_struct{
 extern recv_sys_t*	recv_sys;
 extern ibool		recv_recovery_on;
 extern ibool		recv_no_ibuf_operations;
+extern ibool		recv_needed_recovery;
 
 /* States of recv_addr_struct */
 #define RECV_NOT_PROCESSED	71

innobase/include/mem0mem.h

@@ -269,13 +269,24 @@ mem_realloc(
 	ulint   n,	/* in: desired number of bytes */
 	char*  	file_name,/* in: file name where called */
 	ulint 	line);  /* in: line where called */
+#ifdef MEM_PERIODIC_CHECK
+/**********************************************************************
+Goes through the list of all allocated mem blocks, checks their magic
+numbers, and reports possible corruption. */
 
+void
+mem_validate_all_blocks(void);
+/*=========================*/
+#endif
 
 /*#######################################################################*/
 	
 /* The info header of a block in a memory heap */
 
 struct mem_block_info_struct {
+	ulint   magic_n;/* magic number for debugging */
+	char	file_name[8];/* file name where the mem heap was created */
+	ulint	line;	/* line number where the mem heap was created */
 	UT_LIST_BASE_NODE_T(mem_block_t) base; /* In the first block in the
 			the list this is the base node of the list of blocks;
 			in subsequent blocks this is undefined */
@@ -299,9 +310,11 @@ struct mem_block_info_struct {
 			allocated buffer frame, which can be appended as a
 			free block to the heap, if we need more space;
 			otherwise, this is NULL */
-	ulint   magic_n;/* magic number for debugging */
+#ifdef MEM_PERIODIC_CHECK	
-	char	file_name[8];/* file name where the mem heap was created */
+	UT_LIST_NODE_T(mem_block_t) mem_block_list;
-	ulint	line;	/* line number where the mem heap was created */
+			/* List of all mem blocks allocated; protected
+			by the mem_comm_pool mutex */
+#endif
 };
 
 #define MEM_BLOCK_MAGIC_N	764741555

innobase/include/mem0pool.h

@@ -72,6 +72,18 @@ mem_pool_get_reserved(
 				/* out: reserved mmeory in bytes */
 	mem_pool_t*	pool);	/* in: memory pool */
 /************************************************************************
+Reserves the mem pool mutex. */
+
+void
+mem_pool_mutex_enter(void);
+/*======================*/
+/************************************************************************
+Releases the mem pool mutex. */
+
+void
+mem_pool_mutex_exit(void);
+/*=====================*/
+/************************************************************************
 Validates a memory pool. */
 
 ibool

innobase/include/row0mysql.h

@@ -251,6 +251,24 @@ row_table_add_foreign_constraints(
 	char*	name);		/* in: table full name in the normalized form
 				database_name/table_name */
 /*************************************************************************
+The master thread in srv0srv.c calls this regularly to drop tables which
+we must drop in background after queries to them have ended. Such lazy
+dropping of tables is needed in ALTER TABLE on Unix. */
+
+ulint
+row_drop_tables_for_mysql_in_background(void);
+/*=========================================*/
+					/* out: how many tables dropped
+					+ remaining tables in list */
+/*************************************************************************
+Get the background drop list length. NOTE: the caller must own the kernel
+mutex! */
+
+ulint
+row_get_background_drop_list_len_low(void);
+/*======================================*/
+					/* out: how many tables in list */
+/*************************************************************************
 Drops a table for MySQL. If the name of the dropped table ends to
 characters INNODB_MONITOR, then this also stops printing of monitor
 output by the master thread. */

innobase/include/srv0srv.h

@@ -249,6 +249,12 @@ mutex, for performace reasons). */
 void
 srv_active_wake_master_thread(void);
 /*===============================*/
+/***********************************************************************
+Wakes up the master thread if it is suspended or being suspended. */
+
+void
+srv_wake_master_thread(void);
+/*========================*/
 /*************************************************************************
 Puts an OS thread to wait if there are too many concurrent threads
 (>= srv_thread_concurrency) inside InnoDB. The threads wait in a FIFO queue. */

innobase/include/trx0trx.h

@@ -50,6 +50,13 @@ trx_allocate_for_mysql(void);
 /*========================*/
 				/* out, own: transaction object */
 /************************************************************************
+Creates a transaction object for background operations by the master thread. */
+
+trx_t*
+trx_allocate_for_background(void);
+/*=============================*/
+				/* out, own: transaction object */
+/************************************************************************
 Frees a transaction object. */
 
 void
@@ -63,6 +70,13 @@ void
 trx_free_for_mysql(
 /*===============*/
 	trx_t*	trx);	/* in, own: trx object */
+/************************************************************************
+Frees a transaction object of a background operation of the master thread. */
+
+void
+trx_free_for_background(
+/*====================*/
+	trx_t*	trx);	/* in, own: trx object */
 /********************************************************************
 Creates trx objects for transactions and initializes the trx list of
 trx_sys at database start. Rollback segment and undo log lists must
@@ -266,11 +280,14 @@ struct trx_sig_struct{
 					transaction is waiting a reply */
 };
 
+#define TRX_MAGIC_N	91118598
+
 /* The transaction handle; every session has a trx object which is freed only
 when the session is freed; in addition there may be session-less transactions
 rolling back after a database recovery */
 
 struct trx_struct{
+	ulint		magic_n;
 	/* All the next fields are protected by the kernel mutex, except the
 	undo logs which are protected by undo_mutex */
 	char*		op_info;	/* English text describing the

innobase/log/log0recv.c

@@ -1020,6 +1020,7 @@ loop:
 
 			if (recv_addr->state == RECV_NOT_PROCESSED) {
 				if (!has_printed) {
+	    				ut_print_timestamp(stderr);
 					fprintf(stderr, 
 "  InnoDB: Starting an apply batch of log records to the database...\n"
 "InnoDB: Progress in percents: ");

innobase/mem/mem0mem.c

@@ -75,6 +75,14 @@ After freeing, all the blocks in the heap are set to random bytes
 to help us discover errors which result from the use of
 buffers in an already freed heap. */
 
+#ifdef MEM_PERIODIC_CHECK	
+
+ibool					mem_block_list_inited;
+/* List of all mem blocks allocated; protected by the mem_comm_pool mutex */
+UT_LIST_BASE_NODE_T(mem_block_t)	mem_block_list;
+
+#endif
+			
 /*******************************************************************
 NOTE: Use the corresponding macro instead of this function.
 Allocates a single buffer of memory from the dynamic memory of
@@ -170,6 +178,18 @@ mem_heap_create_block(
 	block->file_name[7]='\0';
 	block->line = line;
 
+#ifdef MEM_PERIODIC_CHECK	
+	mem_pool_mutex_enter();
+
+	if (!mem_block_list_inited) {
+		mem_block_list_inited = TRUE;
+		UT_LIST_INIT(mem_block_list);
+	}
+	
+	UT_LIST_ADD_LAST(mem_block_list, mem_block_list, block);
+
+	mem_pool_mutex_exit();
+#endif
 	mem_block_set_len(block, len);
 	mem_block_set_type(block, type);
 	mem_block_set_free(block, MEM_BLOCK_HEADER_SIZE);
@@ -261,6 +281,13 @@ mem_heap_block_free(
 
 	UT_LIST_REMOVE(list, heap->base, block);
 		
+#ifdef MEM_PERIODIC_CHECK	
+	mem_pool_mutex_enter();
+
+	UT_LIST_REMOVE(mem_block_list, mem_block_list, block);
+
+	mem_pool_mutex_exit();
+#endif
 	type = heap->type;
 	len = block->len;
 	init_block = block->init_block;
@@ -306,3 +333,30 @@ mem_heap_free_block_free(
 		heap->free_block = NULL;
 	}
 }
+
+#ifdef MEM_PERIODIC_CHECK
+/**********************************************************************
+Goes through the list of all allocated mem blocks, checks their magic
+numbers, and reports possible corruption. */
+
+void
+mem_validate_all_blocks(void)
+/*=========================*/
+{
+	mem_block_t*	block;
+
+	mem_pool_mutex_enter();
+
+	block = UT_LIST_GET_FIRST(mem_block_list);
+
+	while (block) {
+		if (block->magic_n != MEM_BLOCK_MAGIC_N) {
+			mem_analyze_corruption((byte*)block);
+		}
+
+		block = UT_LIST_GET_NEXT(mem_block_list, block);
+	}
+
+	mem_pool_mutex_exit();
+}
+#endif

innobase/mem/mem0pool.c

@@ -78,9 +78,9 @@ pool, and after that its locks will grow into the buffer pool. */
 /* The smallest memory area total size */
 #define MEM_AREA_MIN_SIZE	(2 * MEM_AREA_EXTRA_SIZE)
 
+
 /* Data structure for a memory pool. The space is allocated using the buddy
 algorithm, where free list i contains areas of size 2 to power i. */
-
 struct mem_pool_struct{
 	byte*		buf;		/* memory pool */
 	ulint		size;		/* memory common pool size */
@@ -98,6 +98,26 @@ mem_pool_t*	mem_comm_pool	= NULL;
 
 ulint		mem_out_of_mem_err_msg_count	= 0;
 
+/************************************************************************
+Reserves the mem pool mutex. */
+
+void
+mem_pool_mutex_enter(void)
+/*======================*/
+{
+	mutex_enter(&(mem_comm_pool->mutex));
+}
+
+/************************************************************************
+Releases the mem pool mutex. */
+
+void
+mem_pool_mutex_exit(void)
+/*=====================*/
+{
+	mutex_exit(&(mem_comm_pool->mutex));
+}
+
 /************************************************************************
 Returns memory area size. */
 UNIV_INLINE
@@ -241,14 +261,14 @@ mem_pool_fill_free_list(
 		if (mem_out_of_mem_err_msg_count % 1000000000 == 0) {
 			/* We do not print the message every time: */
 
+	    		ut_print_timestamp(stderr);
+			
 			fprintf(stderr,
-	"Innobase: Warning: out of memory in additional memory pool.\n");
+	"  InnoDB: Out of memory in additional memory pool.\n"
-			fprintf(stderr,
+	"InnoDB: InnoDB will start allocating memory from the OS.\n"
-	"Innobase: Innobase will start allocating memory from the OS.\n");
+	"InnoDB: You may get better performance if you configure a bigger\n"
-			fprintf(stderr,
+        "InnoDB: value in the MySQL my.cnf file for\n"
-	"Innobase: You should restart the database with a bigger value in\n");
+	"InnoDB: innodb_additional_mem_pool_size.\n");
-			fprintf(stderr,
-     "Innobase: the MySQL .cnf file for innobase_additional_mem_pool_size.\n");
      		}
 
 		mem_out_of_mem_err_msg_count++;

innobase/row/row0mysql.c

@@ -26,6 +26,19 @@ Created 9/17/2000 Heikki Tuuri
 #include "trx0purge.h"
 #include "lock0lock.h"
 #include "rem0cmp.h"
+#include "log0log.h"
+
+/* List of tables we should drop in background. ALTER TABLE in MySQL requires
+that the table handler can drop the table in background when there are no
+queries to it any more. Protected by the kernel mutex. */
+typedef struct row_mysql_drop_struct	row_mysql_drop_t;
+struct row_mysql_drop_struct{
+	char*				table_name;
+	UT_LIST_NODE_T(row_mysql_drop_t) row_mysql_drop_list;
+};
+
+UT_LIST_BASE_NODE_T(row_mysql_drop_t)	row_mysql_drop_list;
+ibool	row_mysql_drop_list_inited 	= FALSE;
 
 /***********************************************************************
 Reads a MySQL format variable-length field (like VARCHAR) length and
@@ -172,10 +185,22 @@ handle_new_error:
 			trx_general_rollback_for_mysql(trx, TRUE, savept);
 		}
 	} else if (err == DB_TOO_BIG_RECORD) {
+           	if (savept) {
+			/* Roll back the latest, possibly incomplete
+			insertion or update */
+
+			trx_general_rollback_for_mysql(trx, TRUE, savept);
+		}
 		/* MySQL will roll back the latest SQL statement */
 	} else if (err == DB_ROW_IS_REFERENCED
 		   || err == DB_NO_REFERENCED_ROW
 		   || err == DB_CANNOT_ADD_CONSTRAINT) {
+           	if (savept) {
+			/* Roll back the latest, possibly incomplete
+			insertion or update */
+
+			trx_general_rollback_for_mysql(trx, TRUE, savept);
+		}
 		/* MySQL will roll back the latest SQL statement */
 	} else if (err == DB_LOCK_WAIT) {
 
@@ -200,6 +225,12 @@ handle_new_error:
 		trx_general_rollback_for_mysql(trx, FALSE, NULL);
 				
 	} else if (err == DB_OUT_OF_FILE_SPACE) {
+           	if (savept) {
+			/* Roll back the latest, possibly incomplete
+			insertion or update */
+
+			trx_general_rollback_for_mysql(trx, TRUE, savept);
+		}
 		/* MySQL will roll back the latest SQL statement */
 
 	} else if (err == DB_MUST_GET_MORE_FILE_SPACE) {
@@ -375,13 +406,13 @@ row_update_prebuilt_trx(
 					handle */
 	trx_t*		trx)		/* in: transaction handle */
 {	
-	if (prebuilt->magic_n != ROW_PREBUILT_ALLOCATED) {
+	if (trx->magic_n != TRX_MAGIC_N) {
 		fprintf(stderr,
-		"InnoDB: Error: trying to free a corrupt\n"
+		"InnoDB: Error: trying to use a corrupt\n"
-		"InnoDB: table handle. Magic n %lu, table name %s\n",
+		"InnoDB: trx handle. Magic n %lu\n",
-		prebuilt->magic_n, prebuilt->table->name);
+		trx->magic_n);
 
-		mem_analyze_corruption((byte*)prebuilt);
+		mem_analyze_corruption((byte*)trx);
 
 		ut_a(0);
 	}
@@ -1170,6 +1201,9 @@ row_create_table_for_mysql(
 		     	row_drop_table_for_mysql(table->name, trx, TRUE);
 		} else {
 		       	ut_a(err == DB_DUPLICATE_KEY);
+
+	    		ut_print_timestamp(stderr);
+
 			fprintf(stderr, 
      "  InnoDB: Error: table %s already exists in InnoDB internal\n"
      "InnoDB: data dictionary. Have you deleted the .frm file\n"
@@ -1349,6 +1383,164 @@ row_table_add_foreign_constraints(
 	return((int) err);
 }
 
+/*************************************************************************
+Drops a table for MySQL as a background operation. MySQL relies on Unix
+in ALTER TABLE to the fact that the table handler does not remove the
+table before all handles to it has been removed. Furhermore, the MySQL's
+call to drop table must be non-blocking. Therefore we do the drop table
+as a background operation, which is taken care of by the master thread
+in srv0srv.c. */
+static
+int
+row_drop_table_for_mysql_in_background(
+/*===================================*/
+			/* out: error code or DB_SUCCESS */
+	char*	name)	/* in: table name */
+{
+	ulint	error;
+	trx_t*	trx;
+
+	trx = trx_allocate_for_background();
+
+/*	fprintf(stderr, "InnoDB: Dropping table %s in background drop list\n",
+							name); */
+  	/* Drop the table in InnoDB */
+
+  	error = row_drop_table_for_mysql(name, trx, FALSE);
+
+	if (error != DB_SUCCESS) {
+		fprintf(stderr,
+	"InnoDB: Error: Dropping table %s in background drop list failed\n",
+								name);
+	}
+  	
+	/* Flush the log to reduce probability that the .frm files and
+	the InnoDB data dictionary get out-of-sync if the user runs
+	with innodb_flush_log_at_trx_commit = 0 */
+	
+	log_flush_up_to(ut_dulint_max, LOG_WAIT_ONE_GROUP);
+
+  	trx_commit_for_mysql(trx);
+
+  	trx_free_for_background(trx);
+
+	return(DB_SUCCESS);
+}
+
+/*************************************************************************
+The master thread in srv0srv.c calls this regularly to drop tables which
+we must drop in background after queries to them have ended. Such lazy
+dropping of tables is needed in ALTER TABLE on Unix. */
+
+ulint
+row_drop_tables_for_mysql_in_background(void)
+/*=========================================*/
+					/* out: how many tables dropped
+					+ remaining tables in list */
+{
+	row_mysql_drop_t*	drop;
+	dict_table_t*		table;
+	ulint			n_tables;
+	ulint			n_tables_dropped = 0;
+loop:	
+	mutex_enter(&kernel_mutex);
+
+	if (!row_mysql_drop_list_inited) {
+
+		UT_LIST_INIT(row_mysql_drop_list);
+		row_mysql_drop_list_inited = TRUE;
+	}
+
+	drop = UT_LIST_GET_FIRST(row_mysql_drop_list);
+	
+	n_tables = UT_LIST_GET_LEN(row_mysql_drop_list);
+
+	mutex_exit(&kernel_mutex);
+
+	if (drop == NULL) {
+
+		return(n_tables + n_tables_dropped);
+	}
+
+	mutex_enter(&(dict_sys->mutex));
+	table = dict_table_get_low(drop->table_name);
+	mutex_exit(&(dict_sys->mutex));
+
+	if (table->n_mysql_handles_opened > 0) {
+
+		return(n_tables + n_tables_dropped);
+	}
+
+	n_tables_dropped++;
+							
+	row_drop_table_for_mysql_in_background(drop->table_name);
+
+	mutex_enter(&kernel_mutex);
+
+	UT_LIST_REMOVE(row_mysql_drop_list, row_mysql_drop_list, drop);
+
+	mem_free(drop->table_name);
+
+	mem_free(drop);
+
+	mutex_exit(&kernel_mutex);
+
+	goto loop;
+}
+
+/*************************************************************************
+Get the background drop list length. NOTE: the caller must own the kernel
+mutex! */
+
+ulint
+row_get_background_drop_list_len_low(void)
+/*======================================*/
+					/* out: how many tables in list */
+{
+	ut_ad(mutex_own(&kernel_mutex));
+
+	if (!row_mysql_drop_list_inited) {
+
+		UT_LIST_INIT(row_mysql_drop_list);
+		row_mysql_drop_list_inited = TRUE;
+	}
+	
+	return(UT_LIST_GET_LEN(row_mysql_drop_list));
+}
+
+/*************************************************************************
+Adds a table to the list of tables which the master thread drops in
+background. We need this on Unix because in ALTER TABLE MySQL may call
+drop table even if the table has running queries on it. */
+static
+void
+row_add_table_to_background_drop_list(
+/*==================================*/
+	dict_table_t*	table)	/* in: table */
+{
+	row_mysql_drop_t*	drop;
+	
+	drop = mem_alloc(sizeof(row_mysql_drop_t));
+
+	drop->table_name = mem_alloc(1 + ut_strlen(table->name));
+
+	ut_memcpy(drop->table_name, table->name, 1 + ut_strlen(table->name));
+
+	mutex_enter(&kernel_mutex);
+
+	if (!row_mysql_drop_list_inited) {
+
+		UT_LIST_INIT(row_mysql_drop_list);
+		row_mysql_drop_list_inited = TRUE;
+	}
+
+	UT_LIST_ADD_LAST(row_mysql_drop_list, row_mysql_drop_list, drop);
+	
+/*	fprintf(stderr, "InnoDB: Adding table %s to background drop list\n",
+							drop->table_name); */
+	mutex_exit(&kernel_mutex);
+}
+
 /*************************************************************************
 Drops a table for MySQL. If the name of the dropped table ends to
 characters INNODB_MONITOR, then this also stops printing of monitor
@@ -1536,6 +1728,7 @@ row_drop_table_for_mysql(
 
 	if (!table) {
 		err = DB_TABLE_NOT_FOUND;
+	    	ut_print_timestamp(stderr);
 
 		fprintf(stderr, 
      	"  InnoDB: Error: table %s does not exist in the InnoDB internal\n"
@@ -1546,41 +1739,18 @@ row_drop_table_for_mysql(
 		goto funct_exit;
 	}
 
+	if (table->n_mysql_handles_opened > 0) {
+		
+		row_add_table_to_background_drop_list(table);
+
+		err = DB_SUCCESS;
+
+		goto funct_exit;
+	}
+
 	/* Remove any locks there are on the table or its records */
 	
 	lock_reset_all_on_table(table);
-loop:
-	if (table->n_mysql_handles_opened > 0) {
-		rw_lock_s_unlock(&(purge_sys->purge_is_running));
-
-		rw_lock_x_unlock(&(dict_foreign_key_check_lock));
-
-		mutex_exit(&(dict_sys->mutex));
-
-		if (rounds > 60) {
-			fprintf(stderr,
-	"InnoDB: waiting for queries to table %s to end before dropping it\n",
-								name);
-		}
-		
-		os_thread_sleep(1000000);
-
-		mutex_enter(&(dict_sys->mutex));
-		
-		rw_lock_x_lock(&(dict_foreign_key_check_lock));
-
-		rw_lock_s_lock(&(purge_sys->purge_is_running));
-
-		rounds++;
-
-		if (rounds > 120) {
-			fprintf(stderr,
-"InnoDB: Warning: queries to table %s have not ended but we continue anyway\n",
-								name);
-		} else {
-			goto loop;
-		}
-	}
 
 	trx->dict_operation = TRUE;
 	trx->table_id = table->id;
@@ -1617,6 +1787,8 @@ funct_exit:
 
 	trx->op_info = "";
 
+	srv_wake_master_thread();
+
 	return((int) err);
 }
 
@@ -1788,7 +1960,31 @@ row_rename_table_for_mysql(
 	err = trx->error_state;
 
 	if (err != DB_SUCCESS) {
-		row_mysql_handle_errors(&err, trx, thr, NULL);
+		if (err == DB_DUPLICATE_KEY) {
+	    		ut_print_timestamp(stderr);
+
+			fprintf(stderr,
+     "  InnoDB: Error: table %s exists in the InnoDB internal data\n"
+     "InnoDB: dictionary though MySQL is trying rename table %s to it.\n"
+     "InnoDB: Have you deleted the .frm file and not used DROP TABLE?\n",
+			new_name, old_name);
+
+			fprintf(stderr,
+     "InnoDB: If table %s is a temporary table #sql..., then it can be that\n"
+     "InnoDB: there are still queries running on the table, and it will be\n"
+     "InnoDB: dropped automatically when the queries end.\n", new_name);
+			
+			fprintf(stderr,
+     "InnoDB: You can drop the orphaned table inside InnoDB by\n"
+     "InnoDB: creating an InnoDB table with the same name in another\n"
+     "InnoDB: database and moving the .frm file to the current database.\n"
+     "InnoDB: Then MySQL thinks the table exists, and DROP TABLE will\n"
+     "InnoDB: succeed.\n");
+		}
+
+		trx->error_state = DB_SUCCESS;
+		trx_general_rollback_for_mysql(trx, FALSE, NULL);
+		trx->error_state = DB_SUCCESS;
 	} else {
 		ut_a(dict_table_rename_in_cache(table, new_name));
 	}

innobase/srv/srv0srv.c

@@ -20,7 +20,7 @@ Windows 2000 will have something called thread pooling
 Another possibility could be to use some very fast user space
 thread library. This might confuse NT though.
 
-(c) 1995 InnoDB Oy
+(c) 1995 Innobase Oy
 
 Created 10/8/1995 Heikki Tuuri
 *******************************************************/
@@ -49,6 +49,7 @@ Created 10/8/1995 Heikki Tuuri
 #include "btr0sea.h"
 #include "dict0load.h"
 #include "srv0start.h"
+#include "row0mysql.h"
 
 /* Buffer which can be used in printing fatal error messages */
 char	srv_fatal_errbuf[5000];
@@ -1911,17 +1912,12 @@ srv_boot(void)
 
 	srv_init();
 
-	/* Reserve the first slot for the current thread, i.e., the master
-	thread */
-
-	srv_table_reserve_slot(SRV_MASTER);
-
 	return(DB_SUCCESS);
 }
 
 /*************************************************************************
 Reserves a slot in the thread table for the current MySQL OS thread.
-NOTE! The server mutex has to be reserved by the caller! */
+NOTE! The kernel mutex has to be reserved by the caller! */
 static
 srv_slot_t*
 srv_table_reserve_slot_for_mysql(void)
@@ -1931,6 +1927,8 @@ srv_table_reserve_slot_for_mysql(void)
 	srv_slot_t*	slot;
 	ulint		i;
 
+	ut_ad(mutex_own(&kernel_mutex));
+
 	i = 0;
 	slot = srv_mysql_table + i;
 
@@ -2352,6 +2350,22 @@ srv_active_wake_master_thread(void)
 	}
 }
 
+/***********************************************************************
+Wakes up the master thread if it is suspended or being suspended. */
+
+void
+srv_wake_master_thread(void)
+/*========================*/
+{
+	srv_activity_count++;
+			
+	mutex_enter(&kernel_mutex);
+
+	srv_release_threads(SRV_MASTER, 1);
+
+	mutex_exit(&kernel_mutex);
+}
+
 /*************************************************************************
 The master thread controlling the server. */
 
@@ -2374,6 +2388,7 @@ srv_master_thread(
 	ulint		n_bytes_merged;
 	ulint		n_pages_flushed;
 	ulint		n_bytes_archived;
+	ulint		n_tables_to_drop;
 	ulint		n_ios;
 	ulint		n_ios_old;
 	ulint		n_ios_very_old;
@@ -2415,7 +2430,11 @@ loop:
 		can drop tables lazily after there no longer are SELECT
 		queries to them. */
 
-/*		row_drop_tables_for_mysql_in_background(); */
+		srv_main_thread_op_info = "doing background drop tables";
+
+		row_drop_tables_for_mysql_in_background();
+
+		srv_main_thread_op_info = "";
 
 		if (srv_force_recovery >= SRV_FORCE_NO_BACKGROUND) {
 
@@ -2465,6 +2484,11 @@ loop:
 		printf("Master thread wakes up!\n");
 	}
 
+#ifdef MEM_PERIODIC_CHECK
+	/* Check magic numbers of every allocated mem block once in 10
+	seconds */
+	mem_validate_all_blocks();
+#endif	
 	/* If there were less than 200 i/os during the 10 second period,
 	we assume that there is free disk i/o capacity available, and it
 	makes sense to do a buffer pool flush. */
@@ -2521,6 +2545,12 @@ background_loop:
 	/* In this loop we run background operations when the server
 	is quiet and we also come here about once in 10 seconds */
 
+	srv_main_thread_op_info = "doing background drop tables";
+
+	n_tables_to_drop = row_drop_tables_for_mysql_in_background();
+
+	srv_main_thread_op_info = "";
+	
 	srv_main_thread_op_info = "flushing buffer pool pages";
 
 	/* Flush a few oldest pages to make the checkpoint younger */
@@ -2604,11 +2634,13 @@ background_loop:
 	log_archive_do(FALSE, &n_bytes_archived);
 
 	if (srv_fast_shutdown && srv_shutdown_state > 0) {
-		if (n_pages_flushed + n_bytes_archived != 0) {
+		if (n_tables_to_drop + n_pages_flushed
+				+ n_bytes_archived != 0) {
 
 			goto background_loop;
 		}
-	} else if (n_pages_purged + n_bytes_merged + n_pages_flushed
+	} else if (n_tables_to_drop +
+		n_pages_purged + n_bytes_merged + n_pages_flushed
 						+ n_bytes_archived != 0) {
 		goto background_loop;
 	}
@@ -2627,6 +2659,12 @@ suspend_thread:
 
 	mutex_enter(&kernel_mutex);
 
+	if (row_get_background_drop_list_len_low() > 0) {
+		mutex_exit(&kernel_mutex);
+
+		goto loop;
+	}
+
 	event = srv_suspend_thread();
 
 	mutex_exit(&kernel_mutex);

innobase/srv/srv0start.c

@@ -270,6 +270,8 @@ open_or_create_log_file(
 	} else {
 		*log_file_created = TRUE;
 					
+	    	ut_print_timestamp(stderr);
+
 		fprintf(stderr,
 		"  InnoDB: Log file %s did not exist: new to be created\n",
 									name);
@@ -277,6 +279,9 @@ open_or_create_log_file(
 			             name, srv_log_file_size
 			>> (20 - UNIV_PAGE_SIZE_SHIFT));
 
+		fprintf(stderr,
+	    "InnoDB: Database physically writes the file full: wait...\n");
+
 		ret = os_file_set_size(name, files[i],
 					srv_calc_low32(srv_log_file_size),
 					srv_calc_high32(srv_log_file_size));
@@ -454,6 +459,7 @@ open_or_create_data_files(
 			one_created = TRUE;
 
 			if (i > 0) {
+	    			ut_print_timestamp(stderr);
 				fprintf(stderr, 
 		"  InnoDB: Data file %s did not exist: new to be created\n",
 									name);
@@ -464,6 +470,7 @@ open_or_create_data_files(
 				*create_new_db = TRUE;
 			}
 			
+	    		ut_print_timestamp(stderr);
 			fprintf(stderr, 
 				"  InnoDB: Setting file %s size to %lu MB\n",
 			       name, (srv_data_file_sizes[i]
@@ -905,6 +912,12 @@ innobase_start_or_create_for_mysql(void)
 		mtr_commit(&mtr);
 	}
 
+	if (recv_needed_recovery) {
+	    	ut_print_timestamp(stderr);
+		fprintf(stderr,
+	        "  InnoDB: Flushing modified pages from the buffer pool...\n");
+	}
+
 	log_make_checkpoint_at(ut_dulint_max, TRUE);
 
 	if (!srv_log_archive_on) {
@@ -983,7 +996,6 @@ innobase_shutdown_for_mysql(void)
 	    		ut_print_timestamp(stderr);
             		fprintf(stderr, 
 	"  InnoDB: Warning: shutting down a not properly started\n");
-	    		ut_print_timestamp(stderr);
             		fprintf(stderr, 
 	"                 InnoDB: or created database!\n");
 	  	}

innobase/trx/trx0purge.c

@@ -226,9 +226,9 @@ trx_purge_sys_create(void)
 						   value */
 	purge_sys->sess = sess_open(com_endpoint, (byte*)"purge_system", 13);
 
-	purge_sys->trx = (purge_sys->sess)->trx;
+	purge_sys->trx = purge_sys->sess->trx;
 
-	(purge_sys->trx)->type = TRX_PURGE;
+	purge_sys->trx->type = TRX_PURGE;
 
 	ut_a(trx_start_low(purge_sys->trx, ULINT_UNDEFINED));
 

innobase/trx/trx0trx.c

@@ -27,8 +27,8 @@ Created 3/26/1996 Heikki Tuuri
 
 /* Copy of the prototype for innobase_mysql_print_thd: this
 copy must be equal to the one in mysql/sql/ha_innobase.cc ! */
-void innobase_mysql_print_thd(void* thd);
 
+void innobase_mysql_print_thd(void* thd);
 
 /* Dummy session used currently in MySQL interface */
 sess_t*		trx_dummy_sess = NULL;
@@ -64,6 +64,8 @@ trx_create(
 
 	trx = mem_alloc(sizeof(trx_t));
 
+	trx->magic_n = TRX_MAGIC_N;
+
 	trx->op_info = "";
 	
 	trx->type = TRX_USER;
@@ -157,6 +159,32 @@ trx_allocate_for_mysql(void)
 	return(trx);
 }
 
+/************************************************************************
+Creates a transaction object for background operations by the master thread. */
+
+trx_t*
+trx_allocate_for_background(void)
+/*=============================*/
+				/* out, own: transaction object */
+{
+	trx_t*	trx;
+
+	mutex_enter(&kernel_mutex);
+	
+	/* Open a dummy session */
+
+	if (!trx_dummy_sess) {
+		trx_dummy_sess = sess_open(NULL, (byte*)"Dummy sess",
+						ut_strlen("Dummy sess"));
+	}
+	
+	trx = trx_create(trx_dummy_sess);
+
+	mutex_exit(&kernel_mutex);
+	
+	return(trx);
+}
+
 /************************************************************************
 Releases the search latch if trx has reserved it. */
 
@@ -181,6 +209,11 @@ trx_free(
 	trx_t*	trx)	/* in, own: trx object */
 {
 	ut_ad(mutex_own(&kernel_mutex));
+
+	ut_a(trx->magic_n == TRX_MAGIC_N);
+
+	trx->magic_n = 11112222;
+
 	ut_a(trx->conc_state == TRX_NOT_STARTED);
 	
 	mutex_free(&(trx->undo_mutex));
@@ -242,6 +275,21 @@ trx_free_for_mysql(
 	mutex_exit(&kernel_mutex);
 }
 
+/************************************************************************
+Frees a transaction object of a background operation of the master thread. */
+
+void
+trx_free_for_background(
+/*====================*/
+	trx_t*	trx)	/* in, own: trx object */
+{
+	mutex_enter(&kernel_mutex);
+	
+	trx_free(trx);
+	
+	mutex_exit(&kernel_mutex);
+}
+
 /********************************************************************
 Inserts the trx handle in the trx system trx list in the right position.
 The list is sorted on the trx id so that the biggest id is at the list

sql/ha_innobase.cc

@@ -229,7 +229,7 @@ convert_error_code_to_mysql(
 extern "C" {
 /*****************************************************************
 Prints info of a THD object (== user session thread) to the
-standatd output. NOTE that mysql/innobase/trx/trx0trx.c must contain
+standard output. NOTE that mysql/innobase/trx/trx0trx.c must contain
 the prototype for this function! */
 
 void
@@ -298,6 +298,8 @@ check_trx_exists(
 
 		thd->transaction.stmt.innobase_tid =
 		                  (void*)&innodb_dummy_stmt_trx_handle;
+	} else {
+		ut_a(trx->magic_n == TRX_MAGIC_N);
 	}
 
 	return(trx);
@@ -835,6 +837,7 @@ innobase_close_connection(
 			whose transaction should be rolled back */
 {
 	if (NULL != thd->transaction.all.innobase_tid) {
+
 	        trx_rollback_for_mysql((trx_t*)
 				(thd->transaction.all.innobase_tid));
 		trx_free_for_mysql((trx_t*)
@@ -2435,44 +2438,6 @@ ha_innobase::position(
 	ref_stored_len = len;
 }
 
-/***********************************************************************
-Tells something additional to the handler about how to do things. */
-
-int
-ha_innobase::extra(
-/*===============*/
-			   /* out: 0 or error number */
-	enum ha_extra_function operation)
-                           /* in: HA_EXTRA_DONT_USE_CURSOR_TO_UPDATE */
-{
-	row_prebuilt_t*	prebuilt = (row_prebuilt_t*) innobase_prebuilt;
-
-	switch (operation) {
- 		case HA_EXTRA_RESET:
-  		case HA_EXTRA_RESET_STATE:
-	        	prebuilt->read_just_key = 0;
-	        	break;
-		case HA_EXTRA_NO_KEYREAD:
-    			prebuilt->read_just_key = 0;
-    			break;
-	        case HA_EXTRA_DONT_USE_CURSOR_TO_UPDATE:
-			prebuilt->in_update_remember_pos = FALSE;
-			break;
-	        case HA_EXTRA_KEYREAD:
-	        	prebuilt->read_just_key = 1;
-	        	break;
-		default:/* Do nothing */
-			;
-	}
-
-	return(0);
-}
-
-int ha_innobase::reset(void)
-{
-  	return(0);
-}
-
 /*********************************************************************
 Creates a table definition to an InnoDB database. */
 static
@@ -2953,6 +2918,10 @@ ha_innobase::records_in_range(
 
    	DBUG_ENTER("records_in_range");
 
+	/* Warning: since it is not sure that MySQL calls external_lock
+	before calling this function, the trx field in prebuilt can be
+	obsolete! */
+   	
 	active_index = keynr;
 
 	key = table->key_info + active_index;
@@ -2985,6 +2954,10 @@ ha_innobase::records_in_range(
 
     	my_free((char*) key_val_buff2, MYF(0));
 
+	if (prebuilt->trx) {
+		prebuilt->trx->op_info = "";
+	}
+   	
 	DBUG_RETURN((ha_rows) n_rows);
 }
 
@@ -3005,6 +2978,10 @@ ha_innobase::estimate_number_of_rows(void)
 	ulonglong	estimate;
 	ulonglong	data_file_length;
 	
+	/* Warning: since it is not sure that MySQL calls external_lock
+	before calling this function, the trx field in prebuilt can be
+	obsolete! */
+
  	DBUG_ENTER("info");
 
 	index = dict_table_get_first_index_noninline(prebuilt->table);
@@ -3035,10 +3012,12 @@ ha_innobase::scan_time()
 {
 	row_prebuilt_t* prebuilt	= (row_prebuilt_t*) innobase_prebuilt;
 
-	/* In the following formula we assume that scanning 10 pages
+	/* Since MySQL seems to favor table scans too much over index
-	takes the same time as a disk seek: */
+	searches, we pretend that a sequential read takes the same time
+	as a random disk read, that is, we do not divide the following
+	by 10, which would be physically realistic. */
 	
-	return((double) (prebuilt->table->stat_clustered_index_size / 10));
+	return((double) (prebuilt->table->stat_clustered_index_size));
 }
 
 /*************************************************************************
@@ -3059,6 +3038,10 @@ ha_innobase::info(
 
  	DBUG_ENTER("info");
 
+	/* Warning: since it is not sure that MySQL calls external_lock
+	before calling this function, the trx field in prebuilt can be
+	obsolete! */
+   	   	
  	ib_table = prebuilt->table;
 
  	if (flag & HA_STATUS_TIME) {
@@ -3105,6 +3088,13 @@ ha_innobase::info(
    				         index->stat_n_diff_key_vals[j + 1]);
 				}
 
+				/* Since MySQL seems to favor table scans
+				too much over index searches, we pretend
+				index selectivity is 2 times better than
+				our estimate: */
+
+				rec_per_key = rec_per_key / 2;
+
 				if (rec_per_key == 0) {
 					rec_per_key = 1;
 				}
@@ -3124,6 +3114,8 @@ ha_innobase::info(
 	pointer and cause a seg fault. */
 
   	if (flag & HA_STATUS_ERRKEY) {
+		ut_a(prebuilt->trx && prebuilt->trx->magic_n == TRX_MAGIC_N);
+
 		errkey = (unsigned int) row_get_mysql_key_number_for_index(
 				       (dict_index_t*)
 				       trx_get_error_info(prebuilt->trx));
@@ -3149,6 +3141,8 @@ ha_innobase::check(
 	row_prebuilt_t* prebuilt	= (row_prebuilt_t*) innobase_prebuilt;
 	ulint		ret;
 	   	
+	ut_a(prebuilt->trx && prebuilt->trx->magic_n == TRX_MAGIC_N);
+
 	if (prebuilt->mysql_template == NULL) {
 		/* Build the template; we will use a dummy template
 		in index scans done in checking */
@@ -3182,6 +3176,10 @@ ha_innobase::update_table_comment(
   	char*		str 	= my_malloc(length + 550, MYF(0));
   	char*		pos;
 
+	/* Warning: since it is not sure that MySQL calls external_lock
+	before calling this function, the trx field in prebuilt can be
+	obsolete! */
+   	
 	if (!str) {
     		return((char*)comment);
 	}
@@ -3203,6 +3201,53 @@ ha_innobase::update_table_comment(
   	return(str);
 }
 
+/***********************************************************************
+Tells something additional to the handler about how to do things. */
+
+int
+ha_innobase::extra(
+/*===============*/
+			   /* out: 0 or error number */
+	enum ha_extra_function operation)
+                           /* in: HA_EXTRA_DONT_USE_CURSOR_TO_UPDATE */
+{
+	row_prebuilt_t*	prebuilt = (row_prebuilt_t*) innobase_prebuilt;
+
+	/* Warning: since it is not sure that MySQL calls external_lock
+	before calling this function, the trx field in prebuilt can be
+	obsolete! */
+
+	switch (operation) {
+ 		case HA_EXTRA_RESET:
+  		case HA_EXTRA_RESET_STATE:
+	        	prebuilt->read_just_key = 0;
+	        	break;
+		case HA_EXTRA_NO_KEYREAD:
+    			prebuilt->read_just_key = 0;
+    			break;
+	        case HA_EXTRA_DONT_USE_CURSOR_TO_UPDATE:
+			prebuilt->in_update_remember_pos = FALSE;
+			break;
+	        case HA_EXTRA_KEYREAD:
+	        	prebuilt->read_just_key = 1;
+	        	break;
+		default:/* Do nothing */
+			;
+	}
+
+	return(0);
+}
+
+/**********************************************************************
+????????????? */
+
+int
+ha_innobase::reset(void)
+/*====================*/
+{
+  	return(0);
+}
+
 /**********************************************************************
 As MySQL will execute an external lock for every new table it uses when it
 starts to process an SQL statement, we can use this function to store the
@@ -3427,5 +3472,4 @@ ha_innobase::get_auto_increment()
   return(nr);
 }
 
-
 #endif /* HAVE_INNOBASE_DB */