Many files:

  Merge InnoDB-3.23.50

innobase/btr/btr0btr.c

@@ -570,6 +570,19 @@ btr_page_get_father_for_rec(
 
 	node_ptr = btr_cur_get_rec(&cursor);
 
+	if (btr_node_ptr_get_child_page_no(node_ptr) !=
+                                                buf_frame_get_page_no(page)) {
+      		fprintf(stderr,
+"InnoDB: Corruption of an index tree: table %s, index %s,\n"
+"InnoDB: father ptr page no %lu, child page no %lu\n",
+                    (UT_LIST_GET_FIRST(tree->tree_indexes))->table_name,
+                    (UT_LIST_GET_FIRST(tree->tree_indexes))->name,
+                    btr_node_ptr_get_child_page_no(node_ptr),
+                    buf_frame_get_page_no(page));
+     		page_rec_print(page_rec_get_next(page_get_infimum_rec(page)));
+     		page_rec_print(node_ptr);
+	}
+
 	ut_a(btr_node_ptr_get_child_page_no(node_ptr) ==
 						buf_frame_get_page_no(page));
 	mem_heap_free(heap);

innobase/btr/btr0cur.c

@@ -204,7 +204,7 @@ btr_cur_search_to_nth_level(
 				the caller uses his search latch
 				to protect the record! */
 	btr_cur_t*	cursor, /* in/out: tree cursor; the cursor page is
-				   s- or x-latched, but see also above! */
+				s- or x-latched, but see also above! */
 	ulint		has_search_latch,/* in: info on the latch mode the
 				caller currently has on btr_search_latch:
 				RW_S_LATCH, or 0 */

innobase/btr/btr0sea.c

@@ -743,7 +743,7 @@ btr_search_guess_on_hash(
 	
 #ifdef notdefined
 	/* These lines of code can be used in a debug version to check
-	correctness of the searched cursor position: */
+	the correctness of the searched cursor position: */
 	
 	info->last_hash_succ = FALSE;
 

innobase/buf/buf0buf.c

@@ -220,6 +220,10 @@ buf_calc_page_checksum(
 {
   	ulint checksum;
 
+	/* Since the fields FIL_PAGE_FILE_FLUSH_LSN and ..._ARCH_LOG_NO
+	are written outside the buffer pool to the first pages of data
+	files, we have to skip them in page checksum calculation */
+  	
   	checksum = ut_fold_binary(page, FIL_PAGE_FILE_FLUSH_LSN);
   		+ ut_fold_binary(page + FIL_PAGE_DATA,
 				UNIV_PAGE_SIZE - FIL_PAGE_DATA
@@ -279,8 +283,9 @@ buf_page_print(
 
 	ut_sprintf_buf(buf, read_buf, UNIV_PAGE_SIZE);
 
+	ut_print_timestamp(stderr);
 	fprintf(stderr,
-	"InnoDB: Page dump in ascii and hex (%lu bytes):\n%s",
+	"  InnoDB: Page dump in ascii and hex (%lu bytes):\n%s",
 					UNIV_PAGE_SIZE, buf);
 	fprintf(stderr, "InnoDB: End of page dump\n");
 
@@ -288,7 +293,8 @@ buf_page_print(
 
 	checksum = buf_calc_page_checksum(read_buf);
 
-	fprintf(stderr, "InnoDB: Page checksum %lu stored checksum %lu\n",
+	ut_print_timestamp(stderr);
+	fprintf(stderr, "  InnoDB: Page checksum %lu stored checksum %lu\n",
 			checksum, mach_read_from_4(read_buf
                                         + UNIV_PAGE_SIZE
 					- FIL_PAGE_END_LSN)); 
@@ -1358,47 +1364,87 @@ buf_page_io_complete(
 /*=================*/
 	buf_block_t*	block)	/* in: pointer to the block in question */
 {
-	dulint		id;
 	dict_index_t*	index;
+	dulint		id;
 	ulint		io_type;
-
+	ulint		read_page_no;
+	
 	ut_ad(block);
 
 	io_type = block->io_fix;
 
 	if (io_type == BUF_IO_READ) {
+		/* If this page is not uninitialized and not in the
+		doublewrite buffer, then the page number should be the
+		same as in block */
+
+		read_page_no = mach_read_from_4((block->frame)
+						+ FIL_PAGE_OFFSET);
+		if (read_page_no != 0
+			&& !trx_doublewrite_page_inside(read_page_no)
+	    		&& read_page_no != block->offset) {
+
+			fprintf(stderr,
+"InnoDB: Error: page n:o stored in the page read in is %lu, should be %lu!\n",
+				read_page_no, block->offset);
+		}
+#ifdef notdefined
+		if (block->offset != 0 && read_page_no == 0) {
+			/* Check that the page is really uninited */
+
+			for (i = 0; i < UNIV_PAGE_SIZE; i++) {
+
+				if (*((block->frame) + i) != '\0') {
+					fprintf(stderr,
+"InnoDB: Error: page n:o in the page read in is 0, but page %lu is inited!\n",
+						block->offset);
+					break;
+				}
+			}
+		}
+#endif
 		/* From version 3.23.38 up we store the page checksum
-		   to the 4 upper bytes of the page end lsn field */
+		   to the 4 first bytes of the page end lsn field */
 
 		if (buf_page_is_corrupted(block->frame)) {
 		  	fprintf(stderr,
-			  "InnoDB: Database page corruption or a failed\n"
-			  "InnoDB: file read of page %lu.\n", block->offset);
+		"InnoDB: Database page corruption on disk or a failed\n"
+		"InnoDB: file read of page %lu.\n", block->offset);
 			  
 		  	fprintf(stderr,
-			  "InnoDB: You may have to recover from a backup.\n");
+		"InnoDB: You may have to recover from a backup.\n");
 
 			buf_page_print(block->frame);
 
 		  	fprintf(stderr,
-			  "InnoDB: Database page corruption or a failed\n"
-			  "InnoDB: file read of page %lu.\n", block->offset);
+		"InnoDB: Database page corruption on disk or a failed\n"
+		"InnoDB: file read of page %lu.\n", block->offset);
 		  	fprintf(stderr,
-			  "InnoDB: You may have to recover from a backup.\n");
+		"InnoDB: You may have to recover from a backup.\n");
 			fprintf(stderr,
-			  "InnoDB: It is also possible that your operating\n"
-			  "InnoDB: system has corrupted its own file cache\n"
-			  "InnoDB: and rebooting your computer removes the\n"
-			  "InnoDB: error.\n");
+		"InnoDB: It is also possible that your operating\n"
+		"InnoDB: system has corrupted its own file cache\n"
+		"InnoDB: and rebooting your computer removes the\n"
+		"InnoDB: error.\n"
+		"InnoDB: If the corrupt page is an index page\n"
+		"InnoDB: you can also try to fix the corruption\n"
+		"InnoDB: by dumping, dropping, and reimporting\n"
+		"InnoDB: the corrupt table. You can use CHECK\n"
+		"InnoDB: TABLE to scan your table for corruption.\n"
+		"InnoDB: Look also at section 6.1 of\n"
+		"InnoDB: http://www.innodb.com/ibman.html about\n"
+		"InnoDB: forcing recovery.\n");
 			  
-			if (srv_force_recovery < SRV_FORCE_IGNORE_CORRUPT) { 
+			if (srv_force_recovery < SRV_FORCE_IGNORE_CORRUPT) {
+				fprintf(stderr,
+	"InnoDB: Ending processing because of a corrupt database page.\n");
 		  		exit(1);
 		  	}
 		}
 
 		if (recv_recovery_is_on()) {
-			recv_recover_page(TRUE, block->frame, block->space,
-								block->offset);
+			recv_recover_page(FALSE, TRUE, block->frame,
+						block->space, block->offset);
 		}
 
 		if (!recv_no_ibuf_operations) {

innobase/buf/buf0flu.c

@@ -327,6 +327,34 @@ buf_flush_post_to_doublewrite_buf(
 	mutex_exit(&(trx_doublewrite->mutex));
 }
 
+/************************************************************************
+Initializes a page for writing to the tablespace. */
+
+void
+buf_flush_init_for_writing(
+/*=======================*/
+	byte*	page,		/* in: page */
+	dulint	newest_lsn,	/* in: newest modification lsn to the page */
+	ulint	space,		/* in: space id */
+	ulint	page_no)	/* in: page number */
+{	
+	/* Write the newest modification lsn to the page */
+	mach_write_to_8(page + FIL_PAGE_LSN, newest_lsn);
+
+	mach_write_to_8(page + UNIV_PAGE_SIZE - FIL_PAGE_END_LSN, newest_lsn);
+
+	/* Write to the page the space id and page number */
+
+	mach_write_to_4(page + FIL_PAGE_SPACE, space);
+	mach_write_to_4(page + FIL_PAGE_OFFSET, page_no);
+
+	/* We overwrite the first 4 bytes of the end lsn field to store
+	a page checksum */
+
+	mach_write_to_4(page + UNIV_PAGE_SIZE - FIL_PAGE_END_LSN,
+					buf_calc_page_checksum(page));
+}
+
 /************************************************************************
 Does an asynchronous write of a buffer page. NOTE: in simulated aio and
 also when the doublewrite buffer is used, we must call
@@ -349,23 +377,8 @@ buf_flush_write_block_low(
 	/* Force the log to the disk before writing the modified block */
 	log_flush_up_to(block->newest_modification, LOG_WAIT_ALL_GROUPS);
 #endif	
-	/* Write the newest modification lsn to the page */
-	mach_write_to_8(block->frame + FIL_PAGE_LSN,
-						block->newest_modification);
-	mach_write_to_8(block->frame + UNIV_PAGE_SIZE - FIL_PAGE_END_LSN,
-						block->newest_modification);
-
-	/* Write to the page the space id and page number */
-
-	mach_write_to_4(block->frame + FIL_PAGE_SPACE, block->space);
-	mach_write_to_4(block->frame + FIL_PAGE_OFFSET, block->offset);
-
-	/* We overwrite the first 4 bytes of the end lsn field to store
-	a page checksum */
-
-	mach_write_to_4(block->frame + UNIV_PAGE_SIZE - FIL_PAGE_END_LSN,
-			buf_calc_page_checksum(block->frame));
-
+	buf_flush_init_for_writing(block->frame, block->newest_modification,
+						block->space, block->offset);
 	if (!trx_doublewrite) {
 		fil_io(OS_FILE_WRITE | OS_AIO_SIMULATED_WAKE_LATER,
 			FALSE, block->space, block->offset, 0, UNIV_PAGE_SIZE,

innobase/dict/dict0dict.c

@@ -281,7 +281,8 @@ dict_table_autoinc_initialize(
 }
 
 /************************************************************************
-Gets the next autoinc value, 0 if not yet initialized. */
+Gets the next autoinc value, 0 if not yet initialized. If initialized,
+increments the counter by 1. */
 
 ib_longlong
 dict_table_autoinc_get(
@@ -306,6 +307,32 @@ dict_table_autoinc_get(
 	return(value);
 }
 
+/************************************************************************
+Reads the autoinc counter value, 0 if not yet initialized. Does not
+increment the counter. */
+
+ib_longlong
+dict_table_autoinc_read(
+/*====================*/
+				/* out: value of the counter */
+	dict_table_t*	table)	/* in: table */
+{
+	ib_longlong	value;
+
+	mutex_enter(&(table->autoinc_mutex));
+
+	if (!table->autoinc_inited) {
+
+		value = 0;
+	} else {
+		value = table->autoinc;
+	}
+	
+	mutex_exit(&(table->autoinc_mutex));
+
+	return(value);
+}
+
 /************************************************************************
 Updates the autoinc counter if the value supplied is bigger than the
 current value. If not inited, does nothing. */
@@ -644,7 +671,10 @@ dict_table_rename_in_cache(
 /*=======================*/
 					/* out: TRUE if success */
 	dict_table_t*	table,		/* in: table */
-	char*		new_name)	/* in: new name */
+	char*		new_name,	/* in: new name */
+	ibool		rename_also_foreigns)/* in: in ALTER TABLE we want
+					to preserve the original table name
+					in constraints which reference it */
 {
 	dict_foreign_t*	foreign;
 	dict_index_t*	index;
@@ -702,6 +732,41 @@ dict_table_rename_in_cache(
 		index = dict_table_get_next_index(index);
 	}
 
+	if (!rename_also_foreigns) {
+		/* In ALTER TABLE we think of the rename table operation
+		in the direction table -> temporary table (#sql...)
+		as dropping the table with the old name and creating
+		a new with the new name. Thus we kind of drop the
+		constraints from the dictionary cache here. The foreign key
+		constraints will be inherited to the new table from the
+		system tables through a call of dict_load_foreigns. */
+	
+		/* Remove the foreign constraints from the cache */
+		foreign = UT_LIST_GET_LAST(table->foreign_list);
+
+		while (foreign != NULL) {
+			dict_foreign_remove_from_cache(foreign);
+			foreign = UT_LIST_GET_LAST(table->foreign_list);
+		}
+
+		/* Reset table field in referencing constraints */
+
+		foreign = UT_LIST_GET_FIRST(table->referenced_list);
+
+		while (foreign != NULL) {
+			foreign->referenced_table = NULL;
+			foreign->referenced_index = NULL;
+		
+			foreign = UT_LIST_GET_NEXT(referenced_list, foreign);
+		}
+
+		/* Make the list of referencing constraints empty */
+
+		UT_LIST_INIT(table->referenced_list);
+		
+		return(TRUE);
+	}
+
 	/* Update the table name fields in foreign constraints */
 
 	foreign = UT_LIST_GET_FIRST(table->foreign_list);
@@ -768,8 +833,6 @@ dict_table_remove_from_cache(
 	foreign = UT_LIST_GET_LAST(table->foreign_list);
 
 	while (foreign != NULL) {
-		ut_a(0 == ut_strcmp(foreign->foreign_table_name, table->name));
-
 		dict_foreign_remove_from_cache(foreign);
 		foreign = UT_LIST_GET_LAST(table->foreign_list);
 	}
@@ -779,8 +842,6 @@ dict_table_remove_from_cache(
 	foreign = UT_LIST_GET_FIRST(table->referenced_list);
 
 	while (foreign != NULL) {
-		ut_a(0 == ut_strcmp(foreign->referenced_table_name,
-								table->name));
 		foreign->referenced_table = NULL;
 		foreign->referenced_index = NULL;
 		
@@ -1628,8 +1689,9 @@ dict_foreign_add_to_cache(
 {
 	dict_table_t*	for_table;
 	dict_table_t*	ref_table;
-	dict_foreign_t*	for_in_cache	= NULL;
+	dict_foreign_t*	for_in_cache			= NULL;
 	dict_index_t*	index;
+	ibool		added_to_referenced_list	= FALSE;
 
 	ut_ad(mutex_own(&(dict_sys->mutex)));
 
@@ -1673,6 +1735,7 @@ dict_foreign_add_to_cache(
 		UT_LIST_ADD_LAST(referenced_list,
 					ref_table->referenced_list,
 					for_in_cache);
+		added_to_referenced_list = TRUE;
 	}
 
 	if (for_in_cache->foreign_table == NULL && for_table) {
@@ -1683,6 +1746,12 @@ dict_foreign_add_to_cache(
 
 		if (index == NULL) {
 			if (for_in_cache == foreign) {
+				if (added_to_referenced_list) {
+					UT_LIST_REMOVE(referenced_list,
+						ref_table->referenced_list,
+						for_in_cache);
+				}
+			
 				mem_heap_free(foreign->heap);
 			}
 
@@ -1802,9 +1871,14 @@ dict_scan_col(
 		return(ptr);
 	}
 
+	if (*ptr == '`') {
+		ptr++;
+	}
+
 	old_ptr = ptr;
 	
-	while (!isspace(*ptr) && *ptr != ',' && *ptr != ')') {
+	while (!isspace(*ptr) && *ptr != ',' && *ptr != ')' && 	*ptr != '`') {
+
 		ptr++;
 	}
 
@@ -1825,6 +1899,10 @@ dict_scan_col(
 		}
 	}
 	
+	if (*ptr == '`') {
+		ptr++;
+	}
+
 	return(ptr);
 }
 
@@ -1855,9 +1933,13 @@ dict_scan_table_name(
 		return(ptr);
 	}
 
+	if (*ptr == '`') {
+		ptr++;
+	}
+
 	old_ptr = ptr;
 	
-	while (!isspace(*ptr) && *ptr != '(') {
+	while (!isspace(*ptr) && *ptr != '(' && *ptr != '`') {
 		if (*ptr == '.') {
 			dot_ptr = ptr;
 		}
@@ -1898,6 +1980,10 @@ dict_scan_table_name(
 
 	*table = dict_table_get_low(second_table_name);
 
+	if (*ptr == '`') {
+		ptr++;
+	}
+
 	return(ptr);
 }
 
@@ -1940,8 +2026,8 @@ dict_create_foreign_constraints(
 /*============================*/
 				/* out: error code or DB_SUCCESS */
 	trx_t*	trx,		/* in: transaction */
-	char*	sql_string,	/* in: table create statement where
-				foreign keys are declared like:
+	char*	sql_string,	/* in: table create or ALTER TABLE
+				statement where foreign keys are declared like:
 				FOREIGN KEY (a, b) REFERENCES table2(c, d),
 				table2 can be written also with the database
 				name before it: test.table2; the default
@@ -1967,10 +2053,11 @@ dict_create_foreign_constraints(
 	if (table == NULL) {
 		return(DB_ERROR);
 	}
+
 loop:
 	ptr = dict_scan_to(ptr, "FOREIGN");
 
-	if (*ptr == '\0' || dict_bracket_count(sql_string, ptr) != 1) {
+	if (*ptr == '\0') {
 
 		/* The following call adds the foreign key constraints
 		to the data dictionary system tables on disk */
@@ -2883,12 +2970,89 @@ dict_field_print_low(
 	printf(" %s", field->name);
 }
 
+/**************************************************************************
+Sprintfs to a string info on foreign keys of a table in a format suitable
+for CREATE TABLE. */
+static
+void
+dict_print_info_on_foreign_keys_in_create_format(
+/*=============================================*/
+	char*		buf,	/* in: auxiliary buffer of 10000 chars */
+	char*		str,	/* in/out: pointer to a string */
+	ulint		len,	/* in: space in str available for info */
+	dict_table_t*	table)	/* in: table */
+{
+
+	dict_foreign_t*	foreign;
+	ulint		i;
+	char*		buf2;
+
+	buf2 = buf;
+
+	mutex_enter(&(dict_sys->mutex));
+
+	foreign = UT_LIST_GET_FIRST(table->foreign_list);
+
+	if (foreign == NULL) {
+		mutex_exit(&(dict_sys->mutex));
+
+		return;
+	}
+
+	while (foreign != NULL) {
+		buf2 += sprintf(buf2, ",\n  FOREIGN KEY (");
+
+		for (i = 0; i < foreign->n_fields; i++) {
+			buf2 += sprintf(buf2, "`%s`",
+					foreign->foreign_col_names[i]);
+			
+			if (i + 1 < foreign->n_fields) {
+				buf2 += sprintf(buf2, ", ");
+			}
+		}
+
+		buf2 += sprintf(buf2, ") REFERENCES `%s` (",
+					foreign->referenced_table_name);
+		/* Change the '/' in the table name to '.' */
+
+		for (i = ut_strlen(buf); i > 0; i--) {
+			if (buf[i] == '/') {
+
+				buf[i] = '.';
+
+				break;
+			}
+		}
+	
+		for (i = 0; i < foreign->n_fields; i++) {
+			buf2 += sprintf(buf2, "`%s`",
+					foreign->referenced_col_names[i]);
+			if (i + 1 < foreign->n_fields) {
+				buf2 += sprintf(buf2, ", ");
+			}
+		}
+
+		buf2 += sprintf(buf2, ")");
+
+		foreign = UT_LIST_GET_NEXT(foreign_list, foreign);
+	}
+
+	mutex_exit(&(dict_sys->mutex));
+
+	buf[len - 1] = '\0';
+	ut_memcpy(str, buf, len);
+}
+
 /**************************************************************************
 Sprintfs to a string info on foreign keys of a table. */
 
 void
 dict_print_info_on_foreign_keys(
 /*============================*/
+	ibool		create_table_format, /* in: if TRUE then print in
+				a format suitable to be inserted into
+				a CREATE TABLE, otherwise in the format
+				of SHOW TABLE STATUS */
 	char*		str,	/* in/out: pointer to a string */
 	ulint		len,	/* in: space in str available for info */
 	dict_table_t*	table)	/* in: table */
@@ -2898,6 +3062,12 @@ dict_print_info_on_foreign_keys(
 	char*		buf2;
 	char		buf[10000];
 
+	if (create_table_format) {
+		dict_print_info_on_foreign_keys_in_create_format(
+						buf, str, len, table);
+		return;
+	}
+
 	buf2 = buf;
 
 	mutex_enter(&(dict_sys->mutex));
@@ -2911,11 +3081,12 @@ dict_print_info_on_foreign_keys(
 	}
 
 	while (foreign != NULL) {
-		buf2 += sprintf(buf2, "; (");			
-		
+		buf2 += sprintf(buf2, "; (");
+
 		for (i = 0; i < foreign->n_fields; i++) {
 			buf2 += sprintf(buf2, "%s",
 					foreign->foreign_col_names[i]);
+			
 			if (i + 1 < foreign->n_fields) {
 				buf2 += sprintf(buf2, " ");
 			}

innobase/dict/dict0load.c

@@ -688,7 +688,16 @@ dict_load_indexes(
 		
 			dict_load_fields(table, index, heap);
 
-			dict_index_add_to_cache(table, index);
+			if (index->type & DICT_CLUSTERED == 0
+			    && NULL == dict_table_get_first_index(table)) {
+
+				fprintf(stderr,
+	"InnoDB: Error: trying to load index %s for table %s\n"
+	"InnoDB: but the first index was not clustered\n",
+				index->name, table->name);
+			} else {
+				dict_index_add_to_cache(table, index);
+			}
 		}
 
 		btr_pcur_move_to_next_user_rec(&pcur, &mtr);

innobase/fil/fil0fil.c

@@ -89,8 +89,8 @@ struct fil_node_struct {
 	char*		name;	/* the file name or path */
 	ibool		open;	/* TRUE if file open */
 	os_file_t	handle;	/* OS handle to the file, if file open */
-	ulint		size;	/* size of the file in database blocks
-				(where the possible last incomplete block
+	ulint		size;	/* size of the file in database pages
+				(where the possible last incomplete megabyte
 				is ignored) */
 	ulint		n_pending;
 				/* count of pending i/o-ops on this file */
@@ -945,6 +945,76 @@ fil_node_complete_io(
 	}
 }
 		
+/**************************************************************************
+Tries to extend a data file by the number of pages given. Any fractions of a
+megabyte are ignored. */
+
+ibool
+fil_extend_last_data_file(
+/*======================*/
+				/* out: TRUE if success, also if we run
+				out of disk space we may return TRUE */
+	ulint*	actual_increase,/* out: number of pages we were able to
+				extend, here the orginal size of the file and
+				the resulting size of the file are rounded
+				downwards to a full megabyte, and the
+				difference expressed in pages is returned */
+	ulint	size_increase)	/* in: try to extend this many pages */
+{
+	fil_node_t*	node;
+	fil_space_t*	space;
+	fil_system_t*	system		= fil_system;
+	byte*		buf;
+	ibool		success;
+	ulint		i;
+
+	mutex_enter(&(system->mutex));
+
+	HASH_SEARCH(hash, system->spaces, 0, space, space->id == 0);
+
+	ut_a(space);
+	
+	node = UT_LIST_GET_LAST(space->chain);
+
+	fil_node_prepare_for_io(node, system, space);
+
+	buf = mem_alloc(1024 * 1024);
+
+	memset(buf, '\0', 1024 * 1024);
+
+	for (i = 0; i < size_increase / ((1024 * 1024) / UNIV_PAGE_SIZE); i++) {
+
+		success = os_file_write(node->name, node->handle, buf,
+			(node->size << UNIV_PAGE_SIZE_SHIFT) & 0xFFFFFFFF,
+			node->size >> (32 - UNIV_PAGE_SIZE_SHIFT),
+			1024 * 1024);
+
+		if (!success) {
+
+			break;
+		}
+
+		node->size += ((1024 * 1024) / UNIV_PAGE_SIZE);
+		space->size += ((1024 * 1024) / UNIV_PAGE_SIZE);
+
+		os_has_said_disk_full = FALSE;
+	}
+
+	mem_free(buf);
+
+	fil_node_complete_io(node, system, OS_FILE_WRITE);
+
+	mutex_exit(&(system->mutex));	
+
+	*actual_increase = i * ((1024 * 1024) / UNIV_PAGE_SIZE);
+
+	fil_flush(0);
+
+	srv_data_file_sizes[srv_n_data_files - 1] += *actual_increase;
+
+	return(TRUE);
+}
+
 /************************************************************************
 Reads or writes data. This operation is asynchronous (aio). */
 
@@ -966,9 +1036,9 @@ fil_io(
 	ulint	byte_offset,	/* in: remainder of offset in bytes; in
 				aio this must be divisible by the OS block
 				size */
-	ulint	len,		/* in: how many bytes to read; this must
-				not cross a file boundary; in aio this must
-				be a block size multiple */
+	ulint	len,		/* in: how many bytes to read or write; this
+				must not cross a file boundary; in aio this
+				must be a block size multiple */
 	void*	buf,		/* in/out: buffer where to store read data
 				or from where to write; in aio this must be
 				appropriately aligned */

innobase/fsp/fsp0fsp.c

@@ -50,7 +50,7 @@ descriptor page, but used only in the first. */
 #define	FSP_FREE_LIMIT		12	/* Minimum page number for which the
 					free list has not been initialized:
 					the pages >= this limit are, by
-					definition free */
+					definition, free */
 #define	FSP_LOWEST_NO_WRITE	16	/* The lowest page offset for which
 					the page has not been written to disk
 					(if it has been written, we know that
@@ -898,6 +898,106 @@ fsp_header_inc_size(
 	mlog_write_ulint(header + FSP_SIZE, size + size_inc, MLOG_4BYTES, mtr); 
 }
 
+/**************************************************************************
+Gets the current free limit of a tablespace. The free limit means the
+place of the first page which has never been put to the the free list
+for allocation. The space above that address is initialized to zero.
+Sets also the global variable log_fsp_current_free_limit. */
+
+ulint
+fsp_header_get_free_limit(
+/*======================*/
+			/* out: free limit in megabytes */
+	ulint	space)	/* in: space id */
+{
+	fsp_header_t*	header;
+	ulint		limit;
+	mtr_t		mtr;
+
+	ut_a(space == 0); /* We have only one log_fsp_current_... variable */
+	
+	mtr_start(&mtr);
+
+	mtr_x_lock(fil_space_get_latch(space), &mtr);	
+
+	header = fsp_get_space_header(space, &mtr);
+
+	limit = mtr_read_ulint(header + FSP_FREE_LIMIT, MLOG_4BYTES, &mtr);
+
+	limit = limit / ((1024 * 1024) / UNIV_PAGE_SIZE);
+	
+	log_fsp_current_free_limit_set_and_checkpoint(limit);
+
+	mtr_commit(&mtr);
+
+	return(limit);
+}
+
+/***************************************************************************
+Tries to extend the last data file file if it is defined as auto-extending. */
+static
+ibool
+fsp_try_extend_last_file(
+/*=====================*/
+					/* out: FALSE if not auto-extending */
+	ulint*		actual_increase,/* out: actual increase in pages */
+	ulint		space,		/* in: space */
+	fsp_header_t*	header,		/* in: space header */
+	mtr_t*		mtr)		/* in: mtr */
+{
+	ulint	size;
+	ulint	size_increase;
+	ibool	success;
+
+	ut_a(space == 0);
+
+	*actual_increase = 0;
+
+	if (!srv_auto_extend_last_data_file) {
+
+		return(FALSE);
+	}
+
+	size = mtr_read_ulint(header + FSP_SIZE, MLOG_4BYTES, mtr);
+
+	if (srv_last_file_size_max != 0) {
+		if (srv_last_file_size_max
+			 < srv_data_file_sizes[srv_n_data_files - 1]) {
+
+			fprintf(stderr,
+"InnoDB: Error: Last data file size is %lu, max size allowed %lu\n",
+				srv_data_file_sizes[srv_n_data_files - 1],
+				srv_last_file_size_max);
+		}
+
+		size_increase = srv_last_file_size_max
+				 - srv_data_file_sizes[srv_n_data_files - 1];
+		if (size_increase > SRV_AUTO_EXTEND_INCREMENT) {
+			size_increase = SRV_AUTO_EXTEND_INCREMENT;
+		}
+	} else {
+		size_increase = SRV_AUTO_EXTEND_INCREMENT;
+	}
+				
+	if (size_increase == 0) {
+		return(TRUE);
+	}
+	
+	/* Extend the data file. If we are not able to extend
+	the full requested length, the function tells us
+	the number of full megabytes (but the unit is pages!)
+	we were able to extend. */
+				
+	success = fil_extend_last_data_file(actual_increase, size_increase);
+
+	if (success) {
+		mlog_write_ulint(header + FSP_SIZE, size + *actual_increase,
+							MLOG_4BYTES, mtr);
+	}
+
+	return(TRUE);
+}
+
 /**************************************************************************
 Puts new extents to the free list if there are free extents above the free
 limit. If an extent happens to contain an extent descriptor page, the extent
@@ -917,8 +1017,9 @@ fsp_fill_free_list(
 	ulint	frag_n_used;
 	page_t*	descr_page;
 	page_t*	ibuf_page;
-	mtr_t	ibuf_mtr;
+	ulint	actual_increase;
 	ulint	i;
+	mtr_t	ibuf_mtr;
 
 	ut_ad(header && mtr);
 	
@@ -926,12 +1027,28 @@ fsp_fill_free_list(
 	size = mtr_read_ulint(header + FSP_SIZE, MLOG_4BYTES, mtr);
 	limit = mtr_read_ulint(header + FSP_FREE_LIMIT, MLOG_4BYTES, mtr);
 
+	if (srv_auto_extend_last_data_file
+			&& size < limit + FSP_EXTENT_SIZE * FSP_FREE_ADD) {
+
+		/* Try to increase the last data file size */
+		fsp_try_extend_last_file(&actual_increase, space, header,
+									mtr);
+		size = mtr_read_ulint(header + FSP_SIZE, MLOG_4BYTES, mtr);
+	}
+
 	i = limit;
 		
 	while ((i + FSP_EXTENT_SIZE <= size) && (count < FSP_FREE_ADD)) {
 
 		mlog_write_ulint(header + FSP_FREE_LIMIT, i + FSP_EXTENT_SIZE,
 							MLOG_4BYTES, mtr); 
+
+		/* Update the free limit info in the log system and make
+		a checkpoint */
+		log_fsp_current_free_limit_set_and_checkpoint(
+				(i + FSP_EXTENT_SIZE)
+				/ ((1024 * 1024) / UNIV_PAGE_SIZE));
+
 		if (0 == i % XDES_DESCRIBED_PER_PAGE) {
 
 			/* We are going to initialize a new descriptor page
@@ -1172,6 +1289,7 @@ fsp_free_page(
 	xdes_t*		descr;
 	ulint		state;
 	ulint		frag_n_used;
+	char		buf[1000];
 	
 	ut_ad(mtr);
 
@@ -1183,10 +1301,38 @@ fsp_free_page(
 
 	state = xdes_get_state(descr, mtr);
 	
-	ut_a((state == XDES_FREE_FRAG) || (state == XDES_FULL_FRAG));
+	if (state != XDES_FREE_FRAG && state != XDES_FULL_FRAG) {
+		fprintf(stderr,
+"InnoDB: Error: File space extent descriptor of page %lu has state %lu\n",
+								page, state);
+		ut_sprintf_buf(buf, ((byte*)descr) - 50, 200);
 
-	ut_a(xdes_get_bit(descr, XDES_FREE_BIT, page % FSP_EXTENT_SIZE, mtr)
-								== FALSE);
+		fprintf(stderr, "InnoDB: Dump of descriptor: %s\n", buf);
+		
+		if (state == XDES_FREE) {
+			/* We put here some fault tolerance: if the page
+			is already free, return without doing anything! */
+
+			return;
+		}
+
+		ut_a(0);
+	}
+
+	if (xdes_get_bit(descr, XDES_FREE_BIT, page % FSP_EXTENT_SIZE, mtr)
+								== TRUE) {
+		fprintf(stderr,
+"InnoDB: Error: File space extent descriptor of page %lu says it is free\n",
+									page);
+		ut_sprintf_buf(buf, ((byte*)descr) - 50, 200);
+
+		fprintf(stderr, "InnoDB: Dump of descriptor: %s\n", buf);
+
+		/* We put here some fault tolerance: if the page
+		is already free, return without doing anything! */
+
+		return;
+	}
 
 	xdes_set_bit(descr, XDES_FREE_BIT, page % FSP_EXTENT_SIZE, TRUE, mtr);
 	xdes_set_bit(descr, XDES_CLEAN_BIT, page % FSP_EXTENT_SIZE, TRUE, mtr);
@@ -2243,13 +2389,15 @@ fsp_reserve_free_extents(
 	mtr_t*	mtr)	/* in: mtr */
 {
 	fsp_header_t*	space_header;
+	rw_lock_t*	latch;
 	ulint		n_free_list_ext;
 	ulint		free_limit;
 	ulint		size;
 	ulint		n_free;
 	ulint		n_free_up;
 	ulint		reserve;
-	rw_lock_t*	latch;
+	ibool		success;
+	ulint		n_pages_added;
 
 	ut_ad(mtr);	
 	ut_ad(!mutex_own(&kernel_mutex)
@@ -2260,7 +2408,7 @@ fsp_reserve_free_extents(
 	mtr_x_lock(latch, mtr);
 
 	space_header = fsp_get_space_header(space, mtr);
-
+try_again:
 	size = mtr_read_ulint(space_header + FSP_SIZE, MLOG_4BYTES, mtr);
 	
 	n_free_list_ext = flst_get_len(space_header + FSP_FREE, mtr);
@@ -2291,7 +2439,7 @@ fsp_reserve_free_extents(
 
 		if (n_free <= reserve + n_ext) {
 
-			return(FALSE);
+			goto try_to_extend;
 		}
 	} else if (alloc_type == FSP_UNDO) {
 		/* We reserve 1 % of the space size to cleaning operations */
@@ -2300,13 +2448,26 @@ fsp_reserve_free_extents(
 
 		if (n_free <= reserve + n_ext) {
 
-			return(FALSE);
+			goto try_to_extend;
 		}
 	} else {
 		ut_a(alloc_type == FSP_CLEANING);
 	}
 
-	return(fil_space_reserve_free_extents(space, n_free, n_ext));
+	success = fil_space_reserve_free_extents(space, n_free, n_ext);
+
+	if (success) {
+		return(TRUE);
+	}
+try_to_extend:
+	success = fsp_try_extend_last_file(&n_pages_added, space,
+							space_header, mtr);
+	if (success && n_pages_added > 0) {
+
+		goto try_again;
+	}
+
+	return(FALSE);
 }
 
 /**************************************************************************

innobase/include/buf0flu.h

@@ -28,6 +28,16 @@ a margin of replaceable pages there. */
 void
 buf_flush_free_margin(void);
 /*=======================*/
+/************************************************************************
+Initializes a page for writing to the tablespace. */
+
+void
+buf_flush_init_for_writing(
+/*=======================*/
+	byte*	page,		/* in: page */
+	dulint	newest_lsn,	/* in: newest modification lsn to the page */
+	ulint	space,		/* in: space id */
+	ulint	page_no);	/* in: page number */
 /***********************************************************************
 This utility flushes dirty blocks from the end of the LRU list or flush_list.
 NOTE 1: in the case of an LRU flush the calling thread may own latches to

innobase/include/dict0dict.h

@@ -105,7 +105,8 @@ dict_table_autoinc_initialize(
 	dict_table_t*	table,	/* in: table */
 	ib_longlong	value);	/* in: value which was assigned to a row */
 /************************************************************************
-Gets the next autoinc value, 0 if not yet initialized. */
+Gets the next autoinc value, 0 if not yet initialized. If initialized,
+increments the counter by 1. */
 
 ib_longlong
 dict_table_autoinc_get(
@@ -113,6 +114,15 @@ dict_table_autoinc_get(
 				/* out: value for a new row, or 0 */
 	dict_table_t*	table);	/* in: table */
 /************************************************************************
+Reads the autoinc counter value, 0 if not yet initialized. Does not
+increment the counter. */
+
+ib_longlong
+dict_table_autoinc_read(
+/*====================*/
+				/* out: value of the counter */
+	dict_table_t*	table);	/* in: table */
+/************************************************************************
 Updates the autoinc counter if the value supplied is bigger than the
 current value. If not inited, does nothing. */
 
@@ -143,7 +153,10 @@ dict_table_rename_in_cache(
 /*=======================*/
 					/* out: TRUE if success */
 	dict_table_t*	table,		/* in: table */
-	char*		new_name);	/* in: new name */
+	char*		new_name,	/* in: new name */
+	ibool		rename_also_foreigns);/* in: in ALTER TABLE we want
+					to preserve the original table name
+					in constraints which reference it */
 /**************************************************************************
 Adds a foreign key constraint object to the dictionary cache. May free
 the object if there already is an object with the same identifier in.
@@ -284,6 +297,10 @@ Sprintfs to a string info on foreign keys of a table. */
 void
 dict_print_info_on_foreign_keys(
 /*============================*/
+	ibool		create_table_format, /* in: if TRUE then print in
+				a format suitable to be inserted into
+				a CREATE TABLE, otherwise in the format
+				of SHOW TABLE STATUS */
 	char*		str,	/* in/out: pointer to a string */
 	ulint		len,	/* in: space in str available for info */
 	dict_table_t*	table);	/* in: table */

innobase/include/fil0fil.h

@@ -64,8 +64,10 @@ extern fil_addr_t	fil_addr_null;
 #define FIL_PAGE_DATA		38	/* start of the data on the page */
 
 /* File page trailer */
-#define FIL_PAGE_END_LSN	8	/* this should be same as
-					FIL_PAGE_LSN */
+#define FIL_PAGE_END_LSN	8	/* the low 4 bytes of this are used
+					to store the page checksum, the
+					last 4 bytes should be identical
+					to the last 4 bytes of FIL_PAGE_LSN */
 #define FIL_PAGE_DATA_END	8
 
 /* File page types */
@@ -134,6 +136,21 @@ fil_space_truncate_start(
 	ulint	trunc_len);	/* in: truncate by this much; it is an error
 				if this does not equal to the combined size of
 				some initial files in the space */
+/**************************************************************************
+Tries to extend a data file by the number of pages given. Any fractions of a
+megabyte are ignored. */
+
+ibool
+fil_extend_last_data_file(
+/*======================*/
+				/* out: TRUE if success, also if we run
+				out of disk space we may return TRUE */
+	ulint*	actual_increase,/* out: number of pages we were able to
+				extend, here the orginal size of the file and
+				the resulting size of the file are rounded
+				downwards to a full megabyte, and the
+				difference expressed in pages is returned */
+	ulint	size_increase);	/* in: try to extend this many pages */
 /***********************************************************************
 Frees a space object from a file system. Closes the files in the chain
 but does not delete them. */

innobase/include/fsp0fsp.h

@@ -46,6 +46,17 @@ void
 fsp_init(void);
 /*==========*/
 /**************************************************************************
+Gets the current free limit of a tablespace. The free limit means the
+place of the first page which has never been put to the the free list
+for allocation. The space above that address is initialized to zero.
+Sets also the global variable log_fsp_current_free_limit. */
+
+ulint
+fsp_header_get_free_limit(
+/*======================*/
+			/* out: free limit in megabytes */
+	ulint	space);	/* in: space id */
+/**************************************************************************
 Initializes the space header of a new created space. */
 
 void

innobase/include/log0log.h

@@ -26,6 +26,32 @@ extern 	ibool	log_debug_writes;
 #define	LOG_WAIT_ALL_GROUPS	93
 #define LOG_MAX_N_GROUPS	32
 
+/********************************************************************
+Sets the global variable log_fsp_current_free_limit. Also makes a checkpoint,
+so that we know that the limit has been written to a log checkpoint field
+on disk. */
+
+void
+log_fsp_current_free_limit_set_and_checkpoint(
+/*==========================================*/
+	ulint	limit);	/* in: limit to set */
+/***********************************************************************
+Calculates where in log files we find a specified lsn. */
+
+ulint
+log_calc_where_lsn_is(
+/*==================*/
+						/* out: log file number */
+	ib_longlong*	log_file_offset,	/* out: offset in that file
+						(including the header) */
+	dulint		first_header_lsn,	/* in: first log file start
+						lsn */
+	dulint		lsn,			/* in: lsn whose position to
+						determine */
+	ulint		n_log_files,		/* in: total number of log
+						files */
+	ib_longlong	log_file_size);		/* in: log file size
+						(including the header) */
 /****************************************************************
 Writes to the log the string given. The log must be released with
 log_release. */
@@ -225,6 +251,16 @@ Writes checkpoint info to groups. */
 void
 log_groups_write_checkpoint_info(void);
 /*==================================*/
+/**********************************************************
+Writes info to a buffer of a log group when log files are created in
+backup restoration. */
+
+void
+log_reset_first_header_and_checkpoint(
+/*==================================*/
+	byte*	hdr_buf,/* in: buffer which will be written to the start
+			of the first log file */
+	dulint	lsn);	/* in: lsn of the start of the first log file */
 /************************************************************************
 Starts an archiving operation. */
 
@@ -507,7 +543,16 @@ extern log_t*	log_sys;
 							+ LOG_MAX_N_GROUPS * 8)
 #define LOG_CHECKPOINT_CHECKSUM_1 	LOG_CHECKPOINT_ARRAY_END
 #define LOG_CHECKPOINT_CHECKSUM_2 	(4 + LOG_CHECKPOINT_ARRAY_END)
-#define LOG_CHECKPOINT_SIZE		(8 + LOG_CHECKPOINT_ARRAY_END)
+#define LOG_CHECKPOINT_FSP_FREE_LIMIT	(8 + LOG_CHECKPOINT_ARRAY_END)
+					/* current fsp free limit in the
+					tablespace, in units of one megabyte */
+#define LOG_CHECKPOINT_FSP_MAGIC_N	(12 + LOG_CHECKPOINT_ARRAY_END)
+					/* this magic number tells if the
+					checkpoint contains the above field:
+					the field was added to InnoDB-3.23.50 */
+#define LOG_CHECKPOINT_SIZE		(16 + LOG_CHECKPOINT_ARRAY_END)
+
+#define LOG_CHECKPOINT_FSP_MAGIC_N_VAL	1441231243
 
 /* Offsets of a log file header */
 #define LOG_GROUP_ID		0	/* log group number */

innobase/include/log0recv.h

@@ -15,6 +15,39 @@ Created 9/20/1997 Heikki Tuuri
 #include "hash0hash.h"
 #include "log0log.h"
 
+/***********************************************************************
+Reads the checkpoint info needed in hot backup. */
+
+ibool
+recv_read_cp_info_for_backup(
+/*=========================*/
+			/* out: TRUE if success */
+	byte*	hdr,	/* in: buffer containing the log group header */
+	dulint*	lsn,	/* out: checkpoint lsn */
+	ulint*	offset,	/* out: checkpoint offset in the log group */
+	ulint*	fsp_limit,/* out: fsp limit, 1000000000 if the database
+			is running with < version 3.23.50 of InnoDB */
+	dulint*	cp_no,	/* out: checkpoint number */
+	dulint*	first_header_lsn);
+			/* out: lsn of of the start of the first log file */
+/***********************************************************************
+Scans the log segment and n_bytes_scanned is set to the length of valid
+log scanned. */
+
+void
+recv_scan_log_seg_for_backup(
+/*=========================*/
+	byte*		buf,		/* in: buffer containing log data */
+	ulint		buf_len,	/* in: data length in that buffer */
+	dulint*		scanned_lsn,	/* in/out: lsn of buffer start,
+					we return scanned lsn */
+	ulint*		scanned_checkpoint_no,
+					/* in/out: 4 lowest bytes of the
+					highest scanned checkpoint number so
+					far */
+	ulint*		n_bytes_scanned);/* out: how much we were able to
+					scan, smaller than buf_len if log
+					data ended here */
 /***********************************************************************
 Returns TRUE if recovery is currently running. */
 UNIV_INLINE
@@ -35,6 +68,10 @@ read in, or also for a page already in the buffer pool. */
 void
 recv_recover_page(
 /*==============*/
+	ibool	recover_backup,	/* in: TRUE if we are recovering a backup
+				page: then we do not acquire any latches
+				since the page was read in outside the
+				buffer pool */
 	ibool	just_read_in,	/* in: TRUE if the i/o-handler calls this for
 				a freshly read page */
 	page_t*	page,		/* in: buffer page */
@@ -69,8 +106,15 @@ recv_scan_log_recs(
 /*===============*/
 				/* out: TRUE if limit_lsn has been reached, or
 				not able to scan any more in this log group */
+	ibool	apply_automatically,/* in: TRUE if we want this function to
+				apply log records automatically when the
+				hash table becomes full; in the hot backup tool
+				the tool does the applying, not this
+				function */
+	ulint	available_memory,/* in: we let the hash table of recs to grow
+				to this size, at the maximum */
 	ibool	store_to_hash,	/* in: TRUE if the records should be stored
-				to the hash table; this is set FALSE if just
+				to the hash table; this is set to FALSE if just
 				debug checking is needed */
 	byte*	buf,		/* in: buffer containing a log segment or
 				garbage */
@@ -92,6 +136,16 @@ recv_reset_logs(
 	ibool	new_logs_created);/* in: TRUE if resetting logs is done
 				at the log creation; FALSE if it is done
 				after archive recovery */
+/**********************************************************
+Creates new log files after a backup has been restored. */
+
+void
+recv_reset_log_files_for_backup(
+/*============================*/
+	char*	log_dir,	/* in: log file directory path */
+	ulint	n_log_files,	/* in: number of log files */
+	ulint	log_file_size,	/* in: log file size */
+	dulint	lsn);		/* in: new start lsn */
 /************************************************************
 Creates the recovery system. */
 
@@ -102,8 +156,11 @@ recv_sys_create(void);
 Inits the recovery system for a recovery operation. */
 
 void
-recv_sys_init(void);
-/*===============*/
+recv_sys_init(
+/*==========*/
+	ibool	recover_from_backup,	/* in: TRUE if this is called
+					to recover from a hot backup */
+	ulint	available_memory);	/* in: available memory in bytes */
 /***********************************************************************
 Empties the hash table of stored log records, applying them to appropriate
 pages. */
@@ -118,6 +175,17 @@ recv_apply_hashed_log_recs(
 				disk and invalidated in buffer pool: this
 				alternative means that no new log records
 				can be generated during the application */
+/***********************************************************************
+Applies log records in the hash table to a backup. */
+
+void
+recv_apply_log_recs_for_backup(
+/*===========================*/
+	ulint	n_data_files,	/* in: number of data files */
+	char**	data_files,	/* in: array containing the paths to the
+				data files */
+	ulint*	file_sizes);	/* in: sizes of the data files in database
+				pages */
 /************************************************************
 Recovers from archived log files, and also from log files, if they exist. */
 
@@ -260,6 +328,14 @@ extern ibool		recv_recovery_on;
 extern ibool		recv_no_ibuf_operations;
 extern ibool		recv_needed_recovery;
 
+/* Size of the parsing buffer; it must accommodate RECV_SCAN_SIZE many
+times! */ 
+#define RECV_PARSING_BUF_SIZE	(2 * 1024 * 1024)
+
+/* Size of block reads when the log groups are scanned forward to do a
+roll-forward */
+#define RECV_SCAN_SIZE		(4 * UNIV_PAGE_SIZE)
+
 /* States of recv_addr_struct */
 #define RECV_NOT_PROCESSED	71
 #define RECV_BEING_READ		72

innobase/include/mem0mem.h

@@ -41,11 +41,11 @@ page buffer pool; the latter method is used for very big heaps */
 
 /* The following start size is used for the first block in the memory heap if
 the size is not specified, i.e., 0 is given as the parameter in the call of
-create. The standard size is the maximum size of the blocks used for
+create. The standard size is the maximum (payload) size of the blocks used for
 allocations of small buffers. */
 
 #define MEM_BLOCK_START_SIZE            64
-#define MEM_BLOCK_STANDARD_SIZE         8192
+#define MEM_BLOCK_STANDARD_SIZE         8000
 
 /* If a memory heap is allowed to grow into the buffer pool, the following
 is the maximum size for a single allocated buffer: */

innobase/include/os0file.h

@@ -11,6 +11,12 @@ Created 10/21/1995 Heikki Tuuri
 
 #include "univ.i"
 
+
+/* If the following is set to TRUE, we do not call os_file_flush in every
+os_file_write */
+extern ibool	os_do_not_call_flush_at_each_write;
+extern ibool	os_has_said_disk_full;
+
 #ifdef __WIN__
 
 /* We define always WIN_ASYNC_IO, and check at run-time whether
@@ -55,6 +61,9 @@ log. */
 #define	OS_FILE_CREATE			52
 #define OS_FILE_OVERWRITE		53
 
+#define OS_FILE_READ_ONLY 		333
+#define	OS_FILE_READ_WRITE		444
+
 /* Options for file_create */
 #define	OS_FILE_AIO			61
 #define	OS_FILE_NORMAL			62
@@ -118,6 +127,27 @@ os_get_os_version(void);
 /*===================*/
                   /* out: OS_WIN95, OS_WIN31, OS_WINNT (2000 == NT) */
 /********************************************************************
+Creates the seek mutexes used in positioned reads and writes. */
+
+void
+os_io_init_simple(void);
+/*===================*/
+/********************************************************************
+A simple function to open or create a file. */
+
+os_file_t
+os_file_create_simple(
+/*==================*/
+			/* out, own: handle to the file, not defined if error,
+			error number can be retrieved with os_get_last_error */
+	char*	name,	/* in: name of the file or path as a null-terminated
+			string */
+	ulint	create_mode,/* in: OS_FILE_OPEN if an existing file is opened
+			(if does not exist, error), or OS_FILE_CREATE if a new
+			file is created (if exists, error) */
+	ulint	access_type,/* in: OS_FILE_READ_ONLY or OS_FILE_READ_WRITE */
+	ibool*	success);/* out: TRUE if succeed, FALSE if error */
+/********************************************************************
 Opens an existing file or creates a new. */
 
 os_file_t

innobase/include/row0mysql.h

@@ -402,13 +402,13 @@ struct row_prebuilt_struct {
 	byte*		ins_upd_rec_buff;/* buffer for storing data converted
 					to the Innobase format from the MySQL
 					format */
-	ibool		in_update_remember_pos;
-					/* if an update is processed, then if
-					this flag is set to TRUE, it means
-					that the stored cursor position in
-					SELECT is the right position also
-					for the update: we can just restore
-					the cursor and save CPU time */
+	ibool		hint_no_need_to_fetch_extra_cols;
+					/* normally this is TRUE, but
+					MySQL will set this to FALSE
+					if we might be required to fetch also
+					other columns than mentioned in the
+					query: the clustered index column(s),
+					or an auto-increment column*/
 	upd_node_t*	upd_node;	/* Innobase SQL update node used
 					to perform updates and deletes */
 	que_fork_t*	ins_graph;	/* Innobase SQL query graph used

innobase/include/srv0srv.h

@@ -24,10 +24,13 @@ extern char	srv_fatal_errbuf[];
 thread starts running */
 extern os_event_t	srv_lock_timeout_thread_event;
 
+/* If the last data file is auto-extended, we add this many pages to it
+at a time */
+#define SRV_AUTO_EXTEND_INCREMENT   (8 * ((1024 * 1024) / UNIV_PAGE_SIZE))
+
 /* Server parameters which are read from the initfile */
 
 extern char*	srv_data_home;
-extern char*	srv_logs_home;
 extern char*	srv_arch_dir;
 
 extern ulint	srv_n_data_files;
@@ -35,6 +38,9 @@ extern char**	srv_data_file_names;
 extern ulint*	srv_data_file_sizes;
 extern ulint*   srv_data_file_is_raw_partition;
 
+extern ibool	srv_auto_extend_last_data_file;
+extern ulint	srv_last_file_size_max;
+
 extern ibool	srv_created_new_raw;
 
 #define SRV_NEW_RAW    1
@@ -185,6 +191,19 @@ srv_boot(void);
 /*==========*/
 			/* out: DB_SUCCESS or error code */
 /*************************************************************************
+Initializes the server. */
+
+void
+srv_init(void);
+/*==========*/
+/*************************************************************************
+Initializes the synchronization primitives, memory system, and the thread
+local storage. */
+
+void
+srv_general_init(void);
+/*==================*/
+/*************************************************************************
 Gets the number of threads in the system. */
 
 ulint

innobase/include/srv0start.h

@@ -12,6 +12,56 @@ Created 10/10/1995 Heikki Tuuri
 
 #include "univ.i"
 
+/*************************************************************************
+Normalizes a directory path for Windows: converts slashes to backslashes. */
+
+void
+srv_normalize_path_for_win(
+/*=======================*/
+	char*	str);	/* in/out: null-terminated character string */
+/*************************************************************************
+Adds a slash or a backslash to the end of a string if it is missing
+and the string is not empty. */
+
+char*
+srv_add_path_separator_if_needed(
+/*=============================*/
+			/* out, own: string which has the separator if the
+			string is not empty */
+	char*	str);	/* in: null-terminated character string */
+/*************************************************************************
+Reads the data files and their sizes from a character string given in
+the .cnf file. */
+
+ibool
+srv_parse_data_file_paths_and_sizes(
+/*================================*/
+					/* out: TRUE if ok, FALSE if parsing
+					error */
+	char*	str,			/* in: the data file path string */
+	char***	data_file_names,	/* out, own: array of data file
+					names */
+	ulint**	data_file_sizes,	/* out, own: array of data file sizes
+					in megabytes */
+	ulint**	data_file_is_raw_partition,/* out, own: array of flags
+					showing which data files are raw
+					partitions */
+	ulint*	n_data_files,		/* out: number of data files */
+	ibool*	is_auto_extending,	/* out: TRUE if the last data file is
+					auto-extending */
+	ulint*	max_auto_extend_size);	/* out: max auto extend size for the
+					last file if specified, 0 if not */
+/*************************************************************************
+Reads log group home directories from a character string given in
+the .cnf file. */
+
+ibool
+srv_parse_log_group_home_dirs(
+/*==========================*/
+					/* out: TRUE if ok, FALSE if parsing
+					error */
+	char*	str,			/* in: character string */
+	char***	log_group_home_dirs);	/* out, own: log group home dirs */
 /********************************************************************
 Starts Innobase and creates a new database if database files
 are not found and the user wants. Server parameters are

innobase/include/trx0sys.h

@@ -44,6 +44,15 @@ half-written pages in the data files. */
 void
 trx_sys_doublewrite_restore_corrupt_pages(void);
 /*===========================================*/
+/********************************************************************
+Determines if a page number is located inside the doublewrite buffer. */
+
+ibool
+trx_doublewrite_page_inside(
+/*========================*/
+				/* out: TRUE if the location is inside
+				the two blocks of the doublewrite buffer */
+	ulint	page_no);	/* in: page number */
 /*******************************************************************
 Checks if a page address is the trx sys header page. */
 UNIV_INLINE

innobase/include/ut0byte.h

@@ -55,6 +55,15 @@ ut_dulint_get_low(
 			/* out: 32 bits in ulint */
 	dulint	d);	/* in: dulint */
 /***********************************************************
+Converts a dulint (a struct of 2 ulints) to ib_longlong, which is a 64-bit
+integer type. */
+UNIV_INLINE
+ib_longlong
+ut_conv_dulint_to_longlong(
+/*=======================*/
+			/* out: value in ib_longlong type */
+	dulint	d);	/* in: dulint */
+/***********************************************************
 Tests if a dulint is zero. */
 UNIV_INLINE
 ibool

innobase/include/ut0rnd.h

@@ -35,7 +35,7 @@ ut_rnd_gen_next_ulint(
 /*************************************************************
 The following function generates 'random' ulint integers which
 enumerate the value space (let there be N of them) of ulint integers
-in a pseudo random fashion. Note that the same integer is repeated
+in a pseudo-random fashion. Note that the same integer is repeated
 always after N calls to the generator. */
 UNIV_INLINE
 ulint

innobase/include/ut0ut.h

@@ -17,6 +17,16 @@ Created 1/20/1994 Heikki Tuuri
 
 typedef time_t	ib_time_t;
 
+/************************************************************
+Gets the high 32 bits in a ulint. That is makes a shift >> 32,
+but since there seem to be compiler bugs in both gcc and Visual C++,
+we do this by a special conversion. */
+
+ulint
+ut_get_high32(
+/*==========*/
+			/* out: a >> 32 */
+	ulint	a);	/* in: ulint */
 /**********************************************************
 Calculates the minimum of two ulints. */
 UNIV_INLINE
@@ -144,6 +154,15 @@ void
 ut_print_timestamp(
 /*===============*/
 	FILE*  file); /* in: file where to print */
+/**************************************************************
+Returns current year, month, day. */
+
+void
+ut_get_year_month_day(
+/*==================*/
+	ulint*	year,	/* out: current year */
+	ulint*	month,	/* out: month */
+	ulint*	day);	/* out: day */
 /*****************************************************************
 Runs an idle loop on CPU. The argument gives the desired delay
 in microseconds on 100 MHz Pentium + Visual C++. */

innobase/log/log0log.c

 /******************************************************
 Database log
 
-(c) 1995-1997 InnoDB Oy
+(c) 1995-1997 Innobase Oy
 
 Created 12/9/1995 Heikki Tuuri
 *******************************************************/
@@ -24,6 +24,9 @@ Created 12/9/1995 Heikki Tuuri
 #include "trx0sys.h"
 #include "trx0trx.h"
 
+/* Current free limit; protected by the log sys mutex; 0 means uninitialized */
+ulint	log_fsp_current_free_limit		= 0;
+
 /* Global log system variable */
 log_t*	log_sys	= NULL;
 
@@ -95,6 +98,32 @@ void
 log_archive_margin(void);
 /*====================*/
 
+/********************************************************************
+Sets the global variable log_fsp_current_free_limit. Also makes a checkpoint,
+so that we know that the limit has been written to a log checkpoint field
+on disk. */
+
+void
+log_fsp_current_free_limit_set_and_checkpoint(
+/*==========================================*/
+	ulint	limit)	/* in: limit to set */
+{
+	ibool	success;
+
+	mutex_enter(&(log_sys->mutex));
+
+	log_fsp_current_free_limit = limit;
+
+	mutex_exit(&(log_sys->mutex));
+
+	/* Try to make a synchronous checkpoint */
+	
+	success = FALSE;
+
+	while (!success) {
+		success = log_checkpoint(TRUE, TRUE);
+	}
+}
 
 /********************************************************************
 Returns the oldest modified block lsn in the pool, or log_sys->lsn if none
@@ -436,6 +465,51 @@ log_group_calc_lsn_offset(
 	return(log_group_calc_real_offset(offset, group));
 }
 
+/***********************************************************************
+Calculates where in log files we find a specified lsn. */
+
+ulint
+log_calc_where_lsn_is(
+/*==================*/
+						/* out: log file number */
+	ib_longlong*	log_file_offset,	/* out: offset in that file
+						(including the header) */
+	dulint		first_header_lsn,	/* in: first log file start
+						lsn */
+	dulint		lsn,			/* in: lsn whose position to
+						determine */
+	ulint		n_log_files,		/* in: total number of log
+						files */
+	ib_longlong	log_file_size)		/* in: log file size
+						(including the header) */
+{
+	ib_longlong	ib_lsn;
+	ib_longlong	ib_first_header_lsn;
+	ib_longlong	capacity	= log_file_size - LOG_FILE_HDR_SIZE;
+	ulint		file_no;
+	ib_longlong	add_this_many;
+	
+	ib_lsn = ut_conv_dulint_to_longlong(lsn);
+	ib_first_header_lsn = ut_conv_dulint_to_longlong(first_header_lsn);
+
+	if (ib_lsn < ib_first_header_lsn) {
+		add_this_many = 1 + (ib_first_header_lsn - ib_lsn)
+				/ (capacity * (ib_longlong)n_log_files);
+		ib_lsn += add_this_many
+		          * capacity * (ib_longlong)n_log_files;
+	}
+
+	ut_a(ib_lsn >= ib_first_header_lsn);
+	
+	file_no = ((ulint)((ib_lsn - ib_first_header_lsn) / capacity))
+			  % n_log_files;
+	*log_file_offset = (ib_lsn - ib_first_header_lsn) % capacity;
+
+	*log_file_offset = *log_file_offset + LOG_FILE_HDR_SIZE;
+
+	return(file_no);
+}
+
 /************************************************************
 Sets the field values in group to correspond to a given lsn. For this function
 to work, the values must already be correctly initialized to correspond to
@@ -653,7 +727,7 @@ log_init(void)
 
 #ifdef UNIV_LOG_DEBUG
 	recv_sys_create();
-	recv_sys_init();
+	recv_sys_init(FALSE, buf_pool_get_curr_size());
 
 	recv_sys->parse_start_lsn = log_sys->lsn;
 	recv_sys->scanned_lsn = log_sys->lsn;
@@ -961,7 +1035,7 @@ log_group_write_buf(
 	ibool	sync;
 	ibool	write_header;
 	ulint	next_offset;
-
+	
 	ut_ad(mutex_own(&(log_sys->mutex)));
 	ut_ad(len % OS_FILE_LOG_BLOCK_SIZE == 0);
 	ut_ad(ut_dulint_get_low(start_lsn) % OS_FILE_LOG_BLOCK_SIZE == 0);
@@ -1002,9 +1076,28 @@ log_group_write_buf(
 	}
 	
 	if (log_debug_writes) {
+		ulint	i;
+
+		printf(
+		"Writing log file segment to group %lu offset %lu len %lu\n"
+		"start lsn %lu %lu\n",
+			group->id, next_offset, write_len,
+			ut_dulint_get_high(start_lsn),
+			ut_dulint_get_low(start_lsn));
 		printf(
-		"Writing log file segment to group %lu offset %lu len %lu\n",
-					group->id, next_offset, write_len);
+		"First block n:o %lu last block n:o %lu\n",
+			log_block_get_hdr_no(buf),
+			log_block_get_hdr_no(
+				buf + write_len - OS_FILE_LOG_BLOCK_SIZE));
+		ut_a(log_block_get_hdr_no(buf)
+			== log_block_convert_lsn_to_no(start_lsn));
+		
+		for (i = 0; i < write_len / OS_FILE_LOG_BLOCK_SIZE; i++) {
+
+			ut_a(log_block_get_hdr_no(buf) + i
+				== log_block_get_hdr_no(buf
+					+ i * OS_FILE_LOG_BLOCK_SIZE));
+		}
 	}
 
 	if (log_do_write) {
@@ -1346,7 +1439,7 @@ log_group_checkpoint(
 	ulint	i;
 
 	ut_ad(mutex_own(&(log_sys->mutex)));
-	ut_ad(LOG_CHECKPOINT_SIZE <= OS_FILE_LOG_BLOCK_SIZE);
+	ut_a(LOG_CHECKPOINT_SIZE <= OS_FILE_LOG_BLOCK_SIZE);
 	
 	buf = group->checkpoint_buf;
 	
@@ -1394,6 +1487,15 @@ log_group_checkpoint(
 			LOG_CHECKPOINT_CHECKSUM_2 - LOG_CHECKPOINT_LSN);
 	mach_write_to_4(buf + LOG_CHECKPOINT_CHECKSUM_2, fold);
 
+	/* Starting from InnoDB-3.23.50, we also write info on allocated
+	size in the tablespace */
+
+	mach_write_to_4(buf + LOG_CHECKPOINT_FSP_FREE_LIMIT,
+						log_fsp_current_free_limit);
+
+	mach_write_to_4(buf + LOG_CHECKPOINT_FSP_MAGIC_N,
+					LOG_CHECKPOINT_FSP_MAGIC_N_VAL);
+
 	/* We alternate the physical place of the checkpoint info in the first
 	log file */
 	
@@ -1428,6 +1530,48 @@ log_group_checkpoint(
 	}
 }
 
+/**********************************************************
+Writes info to a buffer of a log group when log files are created in
+backup restoration. */
+
+void
+log_reset_first_header_and_checkpoint(
+/*==================================*/
+	byte*	hdr_buf,/* in: buffer which will be written to the start
+			of the first log file */
+	dulint	lsn)	/* in: lsn of the start of the first log file
+			+ LOG_BLOCK_HDR_SIZE */
+{
+	ulint	fold;
+	byte*	buf;
+
+	mach_write_to_4(hdr_buf + LOG_GROUP_ID, 0);
+	mach_write_to_8(hdr_buf + LOG_FILE_START_LSN, lsn);
+
+	buf = hdr_buf + LOG_CHECKPOINT_1;
+	
+	mach_write_to_8(buf + LOG_CHECKPOINT_NO, ut_dulint_zero);
+	mach_write_to_8(buf + LOG_CHECKPOINT_LSN, lsn);
+
+	mach_write_to_4(buf + LOG_CHECKPOINT_OFFSET,
+				LOG_FILE_HDR_SIZE + LOG_BLOCK_HDR_SIZE);
+								
+	mach_write_to_4(buf + LOG_CHECKPOINT_LOG_BUF_SIZE, 2 * 1024 * 1024);
+
+	mach_write_to_8(buf + LOG_CHECKPOINT_ARCHIVED_LSN, ut_dulint_max);
+
+	fold = ut_fold_binary(buf, LOG_CHECKPOINT_CHECKSUM_1);
+	mach_write_to_4(buf + LOG_CHECKPOINT_CHECKSUM_1, fold);
+
+	fold = ut_fold_binary(buf + LOG_CHECKPOINT_LSN,
+			LOG_CHECKPOINT_CHECKSUM_2 - LOG_CHECKPOINT_LSN);
+	mach_write_to_4(buf + LOG_CHECKPOINT_CHECKSUM_2, fold);
+
+	/* Starting from InnoDB-3.23.50, we should also write info on
+	allocated size in the tablespace, but unfortunately we do not
+	know it here */
+}
+
 /**********************************************************
 Reads a checkpoint info from a log group header to log_sys->checkpoint_buf. */
 
@@ -2795,7 +2939,10 @@ log_check_log_recs(
 
 	ut_memcpy(scan_buf, start, end - start);
 	
-	recv_scan_log_recs(FALSE, scan_buf, end - start,
+	recv_scan_log_recs(TRUE,
+				buf_pool_get_curr_size() -
+				RECV_POOL_N_FREE_BLOCKS * UNIV_PAGE_SIZE,	
+				FALSE, scan_buf, end - start,
 				ut_dulint_align_down(buf_start_lsn,
 						OS_FILE_LOG_BLOCK_SIZE),
 			&contiguous_lsn, &scanned_lsn);

innobase/mem/mem0mem.c

@@ -234,7 +234,8 @@ mem_heap_add_block(
 	new_size = 2 * mem_block_get_len(block);
 
 	if (heap->type != MEM_HEAP_DYNAMIC) {
-		ut_ad(n <= MEM_MAX_ALLOC_IN_BUF);
+		/* From the buffer pool we allocate buffer frames */
+		ut_a(n <= MEM_MAX_ALLOC_IN_BUF);
 
 		if (new_size > MEM_MAX_ALLOC_IN_BUF) {
 			new_size = MEM_MAX_ALLOC_IN_BUF;
@@ -249,7 +250,7 @@ mem_heap_add_block(
 	}
 	
 	new_block = mem_heap_create_block(heap, new_size, NULL, heap->type,
-					heap->file_name, heap->line);
+						heap->file_name, heap->line);
 	if (new_block == NULL) {
 
 		return(NULL);

innobase/rem/rem0cmp.c

@@ -55,7 +55,8 @@ cmp_debug_dtuple_rec_with_match(
 				contains the value for current comparison */
 /*****************************************************************
 This function is used to compare two data fields for which the data type
-is such that we must use MySQL code to compare them. */
+is such that we must use MySQL code to compare them. The prototype here
+must be a copy of the the one in ha_innobase.cc! */
 
 int
 innobase_mysql_cmp(

innobase/row/row0ins.c

@@ -391,7 +391,7 @@ row_ins_check_foreign_constraint(
 				/* out: DB_SUCCESS, DB_LOCK_WAIT,
 				DB_NO_REFERENCED_ROW,
 				or DB_ROW_IS_REFERENCED */
-	ibool		check_ref,/* in: TRUE If we want to check that
+	ibool		check_ref,/* in: TRUE if we want to check that
 				the referenced table is ok, FALSE if we
 				want to to check the foreign key table */
 	dict_foreign_t*	foreign,/* in: foreign constraint; NOTE that the
@@ -411,10 +411,23 @@ row_ins_check_foreign_constraint(
 	ibool		moved;
 	int		cmp;
 	ulint		err;
+	ulint		i;
 	mtr_t		mtr;
 
 	ut_ad(rw_lock_own(&dict_foreign_key_check_lock, RW_LOCK_SHARED));
 
+	/* If any of the foreign key fields in entry is SQL NULL, we
+	suppress the foreign key check: this is compatible with Oracle,
+	for example */
+
+	for (i = 0; i < foreign->n_fields; i++) {
+		if (UNIV_SQL_NULL == dfield_get_len(
+                                         dtuple_get_nth_field(entry, i))) {
+
+			return(DB_SUCCESS);
+		}
+	}
+
 	if (check_ref) {
 		check_table = foreign->referenced_table;
 		check_index = foreign->referenced_index;
@@ -591,6 +604,8 @@ row_ins_scan_sec_index_for_duplicate(
 	dtuple_t*	entry,	/* in: index entry */
 	que_thr_t*	thr)	/* in: query thread */
 {
+	ulint		n_unique;
+	ulint		i;
 	int		cmp;
 	ulint		n_fields_cmp;
 	rec_t*		rec;
@@ -599,6 +614,20 @@ row_ins_scan_sec_index_for_duplicate(
 	ibool		moved;
 	mtr_t		mtr;
 
+	n_unique = dict_index_get_n_unique(index);
+
+	/* If the secondary index is unique, but one of the fields in the
+	n_unique first fields is NULL, a unique key violation cannot occur,
+	since we define NULL != NULL in this case */
+
+	for (i = 0; i < n_unique; i++) {
+		if (UNIV_SQL_NULL == dfield_get_len(
+                                         dtuple_get_nth_field(entry, i))) {
+
+			return(DB_SUCCESS);
+		}
+	}
+
 	mtr_start(&mtr);
 
 	/* Store old value on n_fields_cmp */

innobase/row/row0mysql.c

@@ -1881,6 +1881,28 @@ row_drop_database_for_mysql(
 	return(err);
 }
 
+/*************************************************************************
+Checks if a table name contains the string "/#sql" which denotes temporary
+tables in MySQL. */
+static
+ibool
+row_is_mysql_tmp_table_name(
+/*========================*/
+			/* out: TRUE if temporary table */
+	char*	name)	/* in: table name in the form 'database/tablename' */
+{
+	ulint	i;
+
+	for (i = 0; i <= ut_strlen(name) - 5; i++) {
+		if (ut_memcmp(name + i, "/#sql", 5) == 0) {
+
+			return(TRUE);
+		}
+	}
+
+	return(FALSE);
+}
+
 /*************************************************************************
 Renames a table for MySQL. */
 
@@ -1944,16 +1966,27 @@ row_rename_table_for_mysql(
 	str2 = 
 	"';\nold_table_name := '";
 
-	str3 =
-	"';\n"
-	"UPDATE SYS_TABLES SET NAME = new_table_name\n"
-	"WHERE NAME = old_table_name;\n"
-	"UPDATE SYS_FOREIGN SET FOR_NAME = new_table_name\n"
-	"WHERE FOR_NAME = old_table_name;\n"
-	"UPDATE SYS_FOREIGN SET REF_NAME = new_table_name\n"
-	"WHERE REF_NAME = old_table_name;\n"
-	"COMMIT WORK;\n"
-	"END;\n";
+	if (row_is_mysql_tmp_table_name(new_name)) {
+
+		/* We want to preserve the original foreign key
+		constraint definitions despite the name change */
+
+		str3 =
+		"';\n"
+		"UPDATE SYS_TABLES SET NAME = new_table_name\n"
+		"WHERE NAME = old_table_name;\n"
+		"END;\n";
+	} else {
+		str3 =
+		"';\n"
+		"UPDATE SYS_TABLES SET NAME = new_table_name\n"
+		"WHERE NAME = old_table_name;\n"
+		"UPDATE SYS_FOREIGN SET FOR_NAME = new_table_name\n"
+		"WHERE FOR_NAME = old_table_name;\n"
+		"UPDATE SYS_FOREIGN SET REF_NAME = new_table_name\n"
+		"WHERE REF_NAME = old_table_name;\n"
+		"END;\n";
+	}
 
 	len = ut_strlen(str1);
 
@@ -2028,7 +2061,32 @@ row_rename_table_for_mysql(
 		trx_general_rollback_for_mysql(trx, FALSE, NULL);
 		trx->error_state = DB_SUCCESS;
 	} else {
-		ut_a(dict_table_rename_in_cache(table, new_name));
+		ut_a(dict_table_rename_in_cache(table, new_name,
+				!row_is_mysql_tmp_table_name(new_name)));
+
+		if (row_is_mysql_tmp_table_name(old_name)) {
+
+			err = dict_load_foreigns(new_name);
+
+			if (err != DB_SUCCESS) {
+
+	    			ut_print_timestamp(stderr);
+
+				fprintf(stderr,
+     "  InnoDB: Error: in ALTER TABLE table %s\n"
+     "InnoDB: has or is referenced in foreign key constraints\n"
+     "InnoDB: which are not compatible with the new table definition.\n",
+     new_name);
+     
+				ut_a(dict_table_rename_in_cache(table,
+							old_name, FALSE));
+						
+				trx->error_state = DB_SUCCESS;
+				trx_general_rollback_for_mysql(trx, FALSE,
+									NULL);
+				trx->error_state = DB_SUCCESS;
+			}
+		}
 	}
 funct_exit:	
 	mutex_exit(&(dict_sys->mutex));

innobase/row/row0sel.c

@@ -2233,7 +2233,7 @@ row_sel_get_clust_rec_for_mysql(
 		(or old_vers) is not rec; in that case we must ignore
 		such row because in our snapshot rec would not have existed.
 		Remember that from rec we cannot see directly which transaction
-		id corrsponds to it: we have to go to the clustered index
+		id corresponds to it: we have to go to the clustered index
 		record. A query where we want to fetch all rows where
 		the secondary index value is in some interval would return
 		a wrong result if we would not drop rows which we come to
@@ -2244,6 +2244,12 @@ row_sel_get_clust_rec_for_mysql(
 		    && !row_sel_sec_rec_is_for_clust_rec(rec, sec_index,
 						clust_rec, clust_index)) {
 			clust_rec = NULL;
+		} else {
+#ifdef UNIV_SEARCH_DEBUG
+			ut_a(clust_rec == NULL ||
+			    row_sel_sec_rec_is_for_clust_rec(rec, sec_index,
+						clust_rec, clust_index));
+#endif		
 		}
 	}
 
@@ -2399,7 +2405,12 @@ row_sel_try_search_shortcut_for_mysql(
 
 	btr_pcur_open_with_no_init(index, search_tuple, PAGE_CUR_GE,
 					BTR_SEARCH_LEAF, pcur,
-					RW_S_LATCH, mtr);
+#ifndef UNIV_SEARCH_DEBUG
+					RW_S_LATCH,
+#else
+					0,
+#endif
+					mtr);
 	rec = btr_pcur_get_rec(pcur);
 	
 	if (!page_rec_is_user_rec(rec)) {
@@ -2623,15 +2634,18 @@ row_search_for_mysql(
 				
 				goto no_shortcut;
 			}
-			
+#ifndef UNIV_SEARCH_DEBUG			
 			if (!trx->has_search_latch) {
 				rw_lock_s_lock(&btr_search_latch);
 				trx->has_search_latch = TRUE;
 			}
-
+#endif
 			shortcut = row_sel_try_search_shortcut_for_mysql(&rec,
 							       prebuilt, &mtr);
 			if (shortcut == SEL_FOUND) {
+#ifdef UNIV_SEARCH_DEBUG
+				ut_a(0 == cmp_dtuple_rec(search_tuple, rec));
+#endif 
 				row_sel_store_mysql_rec(buf, prebuilt, rec);
 	
  				mtr_commit(&mtr);
@@ -2793,7 +2807,9 @@ row_search_for_mysql(
 		/* The record matches enough */
 
 		ut_ad(mode == PAGE_CUR_GE);
-	
+#ifdef UNIV_SEARCH_DEBUG
+		ut_a(0 == cmp_dtuple_rec(search_tuple, rec));
+#endif	
 	} else if (match_mode == ROW_SEL_EXACT) {
 		/* Test if the index record matches completely to search_tuple
 		in prebuilt: if not, then we return with DB_RECORD_NOT_FOUND */

innobase/row/row0upd.c

@@ -142,7 +142,7 @@ row_upd_index_is_referenced(
 
 /*************************************************************************
 Checks if possible foreign key constraints hold after a delete of the record
-under pcur. NOTE that this function will temporarily commit mtr and lose
+under pcur. NOTE that this function will temporarily commit mtr and lose the
 pcur position! */
 static
 ulint

innobase/srv/srv0srv.c

@@ -69,13 +69,19 @@ char*	srv_main_thread_op_info = "";
 names, where the file name itself may also contain a path */
 
 char*	srv_data_home 	= NULL;
-char*	srv_logs_home 	= NULL;
 char*	srv_arch_dir 	= NULL;
 
 ulint	srv_n_data_files = 0;
 char**	srv_data_file_names = NULL;
 ulint*	srv_data_file_sizes = NULL;	/* size in database pages */ 
 
+ibool	srv_auto_extend_last_data_file	= FALSE; /* if TRUE, then we
+						 auto-extend the last data
+						 file */
+ulint	srv_last_file_size_max	= 0;		 /* if != 0, this tells
+						 the max size auto-extending
+						 may increase the last data
+						 file size */
 ulint*  srv_data_file_is_raw_partition = NULL;
 
 /* If the following is TRUE we do not allow inserts etc. This protects
@@ -1596,7 +1602,7 @@ srv_read_initfile(
 
 /*************************************************************************
 Initializes the server. */
-static
+
 void
 srv_init(void)
 /*==========*/
@@ -1664,7 +1670,7 @@ srv_init(void)
 /*************************************************************************
 Initializes the synchronization primitives, memory system, and the thread
 local storage. */
-static
+
 void
 srv_general_init(void)
 /*==================*/
@@ -1686,6 +1692,7 @@ srv_conc_enter_innodb(
 	trx_t*	trx)	/* in: transaction object associated with the
 			thread */
 {
+	ibool			has_slept	= FALSE;
 	srv_conc_slot_t*	slot;
 	ulint			i;
 
@@ -1703,7 +1710,7 @@ srv_conc_enter_innodb(
 
 		return;
 	}
-
+retry:
 	os_fast_mutex_lock(&srv_conc_mutex);
 
 	if (srv_conc_n_threads < (lint)srv_thread_concurrency) {
@@ -1716,7 +1723,23 @@ srv_conc_enter_innodb(
 
 		return;
 	}
+
+	/* If the transaction is not holding resources, let it sleep
+	for 100 milliseconds, and try again then */
 	
+	if (!has_slept && !trx->has_search_latch
+	    && NULL == UT_LIST_GET_FIRST(trx->trx_locks)) {
+
+	    	has_slept = TRUE; /* We let is sleep only once to avoid
+	    			  starvation */
+
+	    	os_fast_mutex_unlock(&srv_conc_mutex);
+
+	    	os_thread_sleep(100000);
+
+		goto retry;
+	}	    	
+
 	/* Too many threads inside: put the current thread to a queue */
 
 	for (i = 0; i < OS_THREAD_MAX_N; i++) {
@@ -1908,6 +1931,9 @@ srv_normalize_init_values(void)
 					* ((1024 * 1024) / UNIV_PAGE_SIZE);
 	}		
 
+	srv_last_file_size_max = srv_last_file_size_max
+					* ((1024 * 1024) / UNIV_PAGE_SIZE);
+		
 	srv_log_file_size = srv_log_file_size / UNIV_PAGE_SIZE;
 
 	srv_log_buffer_size = srv_log_buffer_size / UNIV_PAGE_SIZE;

innobase/trx/trx0sys.c

@@ -20,11 +20,42 @@ Created 3/26/1996 Heikki Tuuri
 #include "srv0srv.h"
 #include "trx0purge.h"
 #include "log0log.h"
+#include "os0file.h"
 
 /* The transaction system */
 trx_sys_t*		trx_sys 	= NULL;
 trx_doublewrite_t*	trx_doublewrite = NULL;
 
+/********************************************************************
+Determines if a page number is located inside the doublewrite buffer. */
+
+ibool
+trx_doublewrite_page_inside(
+/*========================*/
+				/* out: TRUE if the location is inside
+				the two blocks of the doublewrite buffer */
+	ulint	page_no)	/* in: page number */
+{
+	if (trx_doublewrite == NULL) {
+
+		return(FALSE);
+	}
+
+	if (page_no >= trx_doublewrite->block1
+	    && page_no < trx_doublewrite->block1
+					+ TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
+		return(TRUE);
+	}
+
+	if (page_no >= trx_doublewrite->block2
+	    && page_no < trx_doublewrite->block2
+					+ TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
+		return(TRUE);
+	}
+
+	return(FALSE);
+}
+
 /********************************************************************
 Creates or initialializes the doublewrite buffer at a database start. */
 static
@@ -36,6 +67,11 @@ trx_doublewrite_init(
 {
 	trx_doublewrite = mem_alloc(sizeof(trx_doublewrite_t));
 
+	/* When we have the doublewrite buffer in use, we do not need to
+	call os_file_flush (Unix fsync) after every write. */
+	
+	os_do_not_call_flush_at_each_write = TRUE;
+	
 	mutex_create(&(trx_doublewrite->mutex));
 	mutex_set_level(&(trx_doublewrite->mutex), SYNC_DOUBLEWRITE);
 

innobase/ut/ut0mem.c

@@ -121,6 +121,7 @@ ut_malloc(
 {
         return(ut_malloc_low(n, TRUE));
 }
+
 /**************************************************************************
 Frees a memory block allocated with ut_malloc. */
 

innobase/ut/ut0ut.c

@@ -16,6 +16,24 @@ Created 5/11/1994 Heikki Tuuri
 
 ibool	ut_always_false	= FALSE;
 
+/************************************************************
+Gets the high 32 bits in a ulint. That is makes a shift >> 32,
+but since there seem to be compiler bugs in both gcc and Visual C++,
+we do this by a special conversion. */
+
+ulint
+ut_get_high32(
+/*==========*/
+			/* out: a >> 32 */
+	ulint	a)	/* in: ulint */
+{
+	if (sizeof(ulint) == 4) {
+		return(0);
+	}
+
+	return(a >> 32);
+}
+
 /************************************************************
 The following function returns a clock time in milliseconds. */
 
@@ -58,11 +76,11 @@ ut_print_timestamp(
 	FILE*  file) /* in: file where to print */
 {
 #ifdef __WIN__
-  SYSTEMTIME cal_tm;
+  	SYSTEMTIME cal_tm;
 
-  GetLocalTime(&cal_tm);
+  	GetLocalTime(&cal_tm);
 
-  fprintf(file,"%02d%02d%02d %2d:%02d:%02d",
+  	fprintf(file,"%02d%02d%02d %2d:%02d:%02d",
 	  (int)cal_tm.wYear % 100,
 	  (int)cal_tm.wMonth,
 	  (int)cal_tm.wDay,
@@ -70,23 +88,21 @@ ut_print_timestamp(
 	  (int)cal_tm.wMinute,
 	  (int)cal_tm.wSecond);
 #else
+	struct tm  cal_tm;
+  	struct tm* cal_tm_ptr;
+  	time_t     tm;
 
-  struct tm  cal_tm;
-  struct tm* cal_tm_ptr;
-  time_t     tm;
-
-  time(&tm);
+  	time(&tm);
 
 #ifdef HAVE_LOCALTIME_R
-  localtime_r(&tm, &cal_tm);
-  cal_tm_ptr = &cal_tm;
+  	localtime_r(&tm, &cal_tm);
+  	cal_tm_ptr = &cal_tm;
 #else
-  cal_tm_ptr = localtime(&tm);
+  	cal_tm_ptr = localtime(&tm);
 #endif
-
-  fprintf(file,"%02d%02d%02d %2d:%02d:%02d",
+  	fprintf(file,"%02d%02d%02d %2d:%02d:%02d",
 	  cal_tm_ptr->tm_year % 100,
-	  cal_tm_ptr->tm_mon+1,
+	  cal_tm_ptr->tm_mon + 1,
 	  cal_tm_ptr->tm_mday,
 	  cal_tm_ptr->tm_hour,
 	  cal_tm_ptr->tm_min,
@@ -94,6 +110,39 @@ ut_print_timestamp(
 #endif
 }
 
+/**************************************************************
+Returns current year, month, day. */
+
+void
+ut_get_year_month_day(
+/*==================*/
+	ulint*	year,	/* out: current year */
+	ulint*	month,	/* out: month */
+	ulint*	day)	/* out: day */
+{
+#ifdef __WIN__
+  	SYSTEMTIME cal_tm;
+
+  	GetLocalTime(&cal_tm);
+
+  	*year = (ulint)cal_tm.wYear;
+  	*month = (ulint)cal_tm.wMonth;
+  	*day = (ulint)cal_tm.wDay;
+#else
+  	struct tm  cal_tm;
+  	struct tm* cal_tm_ptr;
+  	time_t     tm;
+
+  	time(&tm);
+
+  	cal_tm_ptr = localtime(&tm);
+
+  	*year = (ulint)cal_tm_ptr->tm_year;
+  	*month = (ulint)cal_tm_ptr->tm_mon + 1;
+  	*day = (ulint)cal_tm_ptr->tm_mday;
+#endif
+}
+
 /*****************************************************************
 Runs an idle loop on CPU. The argument gives the desired delay
 in microseconds on 100 MHz Pentium + Visual C++. */

sql/ha_innobase.h

@@ -154,7 +154,8 @@ class ha_innobase: public handler
 	int rename_table(const char* from, const char* to);
 	int check(THD* thd, HA_CHECK_OPT* check_opt);
         char* update_table_comment(const char* comment);
-
+	char* get_foreign_key_create_info();
+	void free_foreign_key_create_info(char* str);	
   	THR_LOCK_DATA **store_lock(THD *thd, THR_LOCK_DATA **to,
 			     		enum thr_lock_type lock_type);
 	longlong get_auto_increment();