merged with XtraDB 1.1.8-26.0

mysql-test/suite/innodb/t/innodb-blob.test

@@ -4,11 +4,6 @@
 
 --source include/have_innodb.inc
 
-if (`select plugin_auth_version <= "1.1.8-24.1" from information_schema.plugins where plugin_name='innodb'`)
-{
-  --skip Not fixed in XtraDB 1.1.8-24.1 or earlier
-}
-
 # DEBUG_SYNC must be compiled in.
 --source include/have_debug_sync.inc
 

mysql-test/suite/innodb/t/innodb-index.test

 -- source include/have_innodb.inc
 
-if (`select plugin_auth_version <= "1.0.17-13.0" from information_schema.plugins where plugin_name='innodb'`)
-{
-  --skip Not fixed in XtraDB 1.0.17-13.0 or earlier
-}
-
 let $MYSQLD_DATADIR= `select @@datadir`;
 
 let $per_table=`select @@innodb_file_per_table`;

mysql-test/suite/innodb/t/innodb-lock.test

 --source include/have_innodb.inc
 
-if (`select plugin_auth_version <= "1.0.17-13.0" from information_schema.plugins where plugin_name='innodb'`)
-{
-  --skip Not fixed in XtraDB 1.0.17-13.0 or earlier
-}
-
 #
 # Check and select innodb lock type
 #

mysql-test/suite/innodb/t/innodb_bug13635833.test

@@ -2,11 +2,6 @@
 --source include/have_debug_sync.inc
 --source include/not_embedded.inc
 
-if (`select plugin_auth_version <= "1.1.8-24.1" from information_schema.plugins where plugin_name='innodb'`)
-{
-  --skip Not fixed in XtraDB 1.1.8-24.1 or earlier
-}
-
 SET DEBUG_SYNC='reset';
 
 # Save the initial number of concurrent sessions

mysql-test/suite/innodb/t/innodb_bug14007649.test

 --source include/have_innodb.inc
 --source include/have_debug.inc
 
-if (`select plugin_auth_version <= "1.1.8-24.1" from information_schema.plugins where plugin_name='innodb'`)
+if (`select plugin_auth_version <= "1.1.8-26.0" from information_schema.plugins where plugin_name='innodb'`)
 {
-  --skip Not fixed in XtraDB 1.1.8-24.1 or earlier
+  --skip Not fixed in XtraDB 1.1.8-26.0 or earlier
 }
 
 create table t1 (

mysql-test/suite/percona/disabled.def

+percona_sync_flush: removed from XtraDB-26.0
+
 percona_suppress_log_warning_1592:    Feature not merged into MariaDB
 percona_log_slow_slave_statements:    Feature not merged into MariaDB
 percona_log_slow_slave_statements-and-use_global_long_query_time:    Feature not merged into MariaDB

mysql-test/suite/sys_vars/disabled.def

@@ -11,4 +11,5 @@
 ##############################################################################
 
 innodb_trx_rseg_n_slots_debug_basic: variable innodb_trx_rseg_n_slots_debug is removed in MariaDB
+innodb_flush_checkpoint_debug_basic: removed from XtraDB-26.0
 

mysql-test/suite/sys_vars/t/innodb_trx_rseg_n_slots_debug_basic.test

 --source include/have_innodb.inc
 --source include/have_debug.inc
 
-if (`select plugin_auth_version <= "1.1.8-24.1" from information_schema.plugins where plugin_name='innodb'`)
-{
-  --skip Not supported by XtraDB 1.1.8-24.1 or earlier
-}
-
 SET @start_global_value = @@global.innodb_trx_rseg_n_slots_debug;
 SELECT @start_global_value;
 

storage/xtradb/btr/btr0pcur.c

 /*****************************************************************************
 
-Copyright (c) 1996, 2011, Oracle and/or its affiliates. All Rights Reserved.
+Copyright (c) 1996, 2012, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -52,12 +52,13 @@ btr_pcur_create_for_mysql(void)
 }
 
 /**************************************************************//**
-Frees the memory for a persistent cursor object. */
+Resets a persistent cursor object, freeing ::old_rec_buf if it is
+allocated and resetting the other members to their initial values. */
 UNIV_INTERN
 void
-btr_pcur_free_for_mysql(
-/*====================*/
-	btr_pcur_t*	cursor)	/*!< in, own: persistent cursor */
+btr_pcur_reset(
+/*===========*/
+	btr_pcur_t*	cursor)	/*!< in, out: persistent cursor */
 {
 	if (cursor->old_rec_buf != NULL) {
 
@@ -66,6 +67,7 @@ btr_pcur_free_for_mysql(
 		cursor->old_rec_buf = NULL;
 	}
 
+	cursor->btr_cur.index = NULL;
 	cursor->btr_cur.page_cur.rec = NULL;
 	cursor->old_rec = NULL;
 	cursor->old_n_fields = 0;
@@ -73,7 +75,17 @@ btr_pcur_free_for_mysql(
 
 	cursor->latch_mode = BTR_NO_LATCHES;
 	cursor->pos_state = BTR_PCUR_NOT_POSITIONED;
+}
 
+/**************************************************************//**
+Frees the memory for a persistent cursor object. */
+UNIV_INTERN
+void
+btr_pcur_free_for_mysql(
+/*====================*/
+	btr_pcur_t*	cursor)	/*!< in, own: persistent cursor */
+{
+	btr_pcur_reset(cursor);
 	mem_free(cursor);
 }
 
@@ -127,6 +139,8 @@ btr_pcur_store_position(
 
 		ut_a(btr_page_get_next(page, mtr) == FIL_NULL);
 		ut_a(btr_page_get_prev(page, mtr) == FIL_NULL);
+		ut_ad(page_is_leaf(page));
+		ut_ad(page_get_page_no(page) == index->page);
 
 		cursor->old_stored = BTR_PCUR_OLD_STORED;
 
@@ -319,13 +333,19 @@ btr_pcur_restore_position_func(
 	/* Save the old search mode of the cursor */
 	old_mode = cursor->search_mode;
 
-	if (UNIV_LIKELY(cursor->rel_pos == BTR_PCUR_ON)) {
+	switch (cursor->rel_pos) {
+	case BTR_PCUR_ON:
 		mode = PAGE_CUR_LE;
-	} else if (cursor->rel_pos == BTR_PCUR_AFTER) {
+		break;
+	case BTR_PCUR_AFTER:
 		mode = PAGE_CUR_G;
-	} else {
-		ut_ad(cursor->rel_pos == BTR_PCUR_BEFORE);
+		break;
+	case BTR_PCUR_BEFORE:
 		mode = PAGE_CUR_L;
+		break;
+	default:
+		ut_error;
+		mode = 0;
 	}
 
 	btr_pcur_open_with_no_init_func(index, tuple, mode, latch_mode,
@@ -334,25 +354,44 @@ btr_pcur_restore_position_func(
 	/* Restore the old search mode */
 	cursor->search_mode = old_mode;
 
-	if (cursor->rel_pos == BTR_PCUR_ON
-	    && btr_pcur_is_on_user_rec(cursor)
-	    && 0 == cmp_dtuple_rec(tuple, btr_pcur_get_rec(cursor),
-				   rec_get_offsets(
-					   btr_pcur_get_rec(cursor), index,
-					   NULL, ULINT_UNDEFINED, &heap))) {
-
-		/* We have to store the NEW value for the modify clock, since
-		the cursor can now be on a different page! But we can retain
-		the value of old_rec */
+	if (btr_pcur_is_on_user_rec(cursor)) {
+		switch (cursor->rel_pos) {
+		case BTR_PCUR_ON:
+			if (!cmp_dtuple_rec(
+				    tuple, btr_pcur_get_rec(cursor),
+				    rec_get_offsets(btr_pcur_get_rec(cursor),
+						    index, NULL,
+						    ULINT_UNDEFINED, &heap))) {
+
+				/* We have to store the NEW value for
+				the modify clock, since the cursor can
+				now be on a different page! But we can
+				retain the value of old_rec */
+
+				cursor->block_when_stored =
+					btr_pcur_get_block(cursor);
+				cursor->modify_clock =
+					buf_block_get_modify_clock(
+						cursor->block_when_stored);
+				cursor->old_stored = BTR_PCUR_OLD_STORED;
 
-		cursor->block_when_stored = btr_pcur_get_block(cursor);
-		cursor->modify_clock = buf_block_get_modify_clock(
-			cursor->block_when_stored);
-		cursor->old_stored = BTR_PCUR_OLD_STORED;
+				mem_heap_free(heap);
 
-		mem_heap_free(heap);
+				return(TRUE);
+			}
 
-		return(TRUE);
+			break;
+		case BTR_PCUR_BEFORE:
+			page_cur_move_to_next(btr_pcur_get_page_cur(cursor));
+			break;
+		case BTR_PCUR_AFTER:
+			page_cur_move_to_prev(btr_pcur_get_page_cur(cursor));
+			break;
+#ifdef UNIV_DEBUG
+		default:
+			ut_error;
+#endif /* UNIV_DEBUG */
+		}
 	}
 
 	mem_heap_free(heap);

storage/xtradb/btr/btr0sea.c

 /*****************************************************************************
 
-Copyright (c) 1996, 2011, Oracle and/or its affiliates. All Rights Reserved.
+Copyright (c) 1996, 2012, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -1240,7 +1240,7 @@ btr_search_drop_page_hash_index(
 			index->name, (ulong) block->n_pointers);
 		rw_lock_x_unlock(btr_search_get_latch(index->id));
 
-		btr_search_validate();
+		ut_ad(btr_search_validate());
 	} else {
 		rw_lock_x_unlock(btr_search_get_latch(index->id));
 	}
@@ -2143,7 +2143,9 @@ btr_search_validate(void)
 					(ulong) block->curr_left_side);
 
 				if (n_page_dumps < 20) {
-					buf_page_print(page, 0);
+					buf_page_print(
+						page, 0,
+						BUF_PAGE_PRINT_NO_CRASH);
 					n_page_dumps++;
 				}
 			}

storage/xtradb/buf/buf0buf.c

 /*****************************************************************************
 
-Copyright (c) 1995, 2011, Oracle and/or its affiliates. All Rights Reserved.
+Copyright (c) 1995, 2012, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -57,7 +57,9 @@ Created 11/5/1995 Heikki Tuuri
 /* prototypes for new functions added to ha_innodb.cc */
 trx_t* innobase_get_trx();
 
-static inline void _increment_page_get_statistics(buf_block_t* block, trx_t* trx)
+static inline
+void
+_increment_page_get_statistics(buf_block_t* block, trx_t* trx)
 {
 	ulint           block_hash;
 	ulint           block_hash_byte;
@@ -680,8 +682,12 @@ void
 buf_page_print(
 /*===========*/
 	const byte*	read_buf,	/*!< in: a database page */
-	ulint		zip_size)	/*!< in: compressed page size, or
-				0 for uncompressed pages */
+	ulint		zip_size,	/*!< in: compressed page size, or
+					0 for uncompressed pages */
+	ulint		flags)		/*!< in: 0 or
+					BUF_PAGE_PRINT_NO_CRASH or
+					BUF_PAGE_PRINT_NO_FULL */
+
 {
 #ifndef UNIV_HOTBACKUP
 	dict_index_t*	index;
@@ -695,11 +701,14 @@ buf_page_print(
 		size = UNIV_PAGE_SIZE;
 	}
 
-	ut_print_timestamp(stderr);
-	fprintf(stderr, "  InnoDB: Page dump in ascii and hex (%lu bytes):\n",
-		(ulong) size);
-	ut_print_buf(stderr, read_buf, size);
-	fputs("\nInnoDB: End of page dump\n", stderr);
+	if (!(flags & BUF_PAGE_PRINT_NO_FULL)) {
+		ut_print_timestamp(stderr);
+		fprintf(stderr,
+			"  InnoDB: Page dump in ascii and hex (%lu bytes):\n",
+			(ulong) size);
+		ut_print_buf(stderr, read_buf, size);
+		fputs("\nInnoDB: End of page dump\n", stderr);
+	}
 
 	if (zip_size) {
 		/* Print compressed page. */
@@ -873,6 +882,8 @@ buf_page_print(
 		      stderr);
 		break;
 	}
+
+	ut_ad(flags & BUF_PAGE_PRINT_NO_CRASH);
 }
 
 #ifndef UNIV_HOTBACKUP
@@ -1078,11 +1089,8 @@ buf_chunk_init(
 	for (i = chunk->size; i--; ) {
 
 		buf_block_init(buf_pool, block, frame);
+		UNIV_MEM_INVALID(block->frame, UNIV_PAGE_SIZE);
 
-#ifdef HAVE_valgrind
-		/* Wipe contents of frame to eliminate a Purify warning */
-		memset(block->frame, '\0', UNIV_PAGE_SIZE);
-#endif
 		/* Add the block to the free list */
 		mutex_enter(&buf_pool->free_list_mutex);
 		UT_LIST_ADD_LAST(free, buf_pool->free, (&block->page));
@@ -2768,7 +2776,7 @@ buf_page_get_gen(
 
 		block->page.buf_fix_count = 1;
 		buf_block_set_io_fix(block, BUF_IO_READ);
-		rw_lock_x_lock_func(&block->lock, 0, file, line);
+		rw_lock_x_lock_inline(&block->lock, 0, file, line);
 
 		UNIV_MEM_INVALID(bpage, sizeof *bpage);
 
@@ -2931,14 +2939,14 @@ buf_page_get_gen(
 		break;
 
 	case RW_S_LATCH:
-		rw_lock_s_lock_func(&(block->lock), 0, file, line);
+		rw_lock_s_lock_inline(&(block->lock), 0, file, line);
 
 		fix_type = MTR_MEMO_PAGE_S_FIX;
 		break;
 
 	default:
 		ut_ad(rw_latch == RW_X_LATCH);
-		rw_lock_x_lock_func(&(block->lock), 0, file, line);
+		rw_lock_x_lock_inline(&(block->lock), 0, file, line);
 
 		fix_type = MTR_MEMO_PAGE_X_FIX;
 		break;
@@ -3023,8 +3031,8 @@ buf_page_optimistic_get(
 						file, line);
 		fix_type = MTR_MEMO_PAGE_S_FIX;
 	} else {
-		success = rw_lock_x_lock_func_nowait(&(block->lock),
-						     file, line);
+		success = rw_lock_x_lock_func_nowait_inline(&(block->lock),
+							    file, line);
 		fix_type = MTR_MEMO_PAGE_X_FIX;
 	}
 
@@ -3165,8 +3173,8 @@ buf_page_get_known_nowait(
 						file, line);
 		fix_type = MTR_MEMO_PAGE_S_FIX;
 	} else {
-		success = rw_lock_x_lock_func_nowait(&(block->lock),
-						     file, line);
+		success = rw_lock_x_lock_func_nowait_inline(&(block->lock),
+							    file, line);
 		fix_type = MTR_MEMO_PAGE_X_FIX;
 	}
 
@@ -3186,7 +3194,7 @@ buf_page_get_known_nowait(
 	ut_a(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
 #endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */
 #if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
-	ut_a(block->page.file_page_was_freed == FALSE);
+	ut_a(mode == BUF_KEEP_OLD || !block->page.file_page_was_freed);
 #endif
 
 #ifdef UNIV_IBUF_COUNT_DEBUG
@@ -3261,8 +3269,8 @@ buf_page_try_get_func(
 		S-latch. */
 
 		fix_type = MTR_MEMO_PAGE_X_FIX;
-		success = rw_lock_x_lock_func_nowait(&block->lock,
-						     file, line);
+		success = rw_lock_x_lock_func_nowait_inline(&block->lock,
+							    file, line);
 	}
 
 	if (!success) {
@@ -3997,7 +4005,8 @@ buf_page_io_complete(
 				"InnoDB: You may have to recover"
 				" from a backup.\n",
 				(ulong) bpage->offset);
-			buf_page_print(frame, buf_page_get_zip_size(bpage));
+			buf_page_print(frame, buf_page_get_zip_size(bpage),
+				       BUF_PAGE_PRINT_NO_CRASH);
 			fprintf(stderr,
 				"InnoDB: Database page corruption on disk"
 				" or a failed\n"
@@ -5285,34 +5294,32 @@ buf_all_freed(void)
 /*********************************************************************//**
 Checks that there currently are no pending i/o-operations for the buffer
 pool.
-@return	TRUE if there is no pending i/o */
+@return	number of pending i/o */
 UNIV_INTERN
-ibool
-buf_pool_check_no_pending_io(void)
-/*==============================*/
+ulint
+buf_pool_check_num_pending_io(void)
+/*===============================*/
 {
 	ulint		i;
-	ibool		ret = TRUE;
+	ulint		pending_io = 0;
 
 	buf_pool_mutex_enter_all();
 
-	for (i = 0; i < srv_buf_pool_instances && ret; i++) {
+	for (i = 0; i < srv_buf_pool_instances; i++) {
 		const buf_pool_t*	buf_pool;
 
 		buf_pool = buf_pool_from_array(i);
 
-		if (buf_pool->n_pend_reads
-		    + buf_pool->n_flush[BUF_FLUSH_LRU]
-		    + buf_pool->n_flush[BUF_FLUSH_LIST]
-		    + buf_pool->n_flush[BUF_FLUSH_SINGLE_PAGE]) {
+		pending_io += buf_pool->n_pend_reads
+			      + buf_pool->n_flush[BUF_FLUSH_LRU]
+			      + buf_pool->n_flush[BUF_FLUSH_LIST]
+			      + buf_pool->n_flush[BUF_FLUSH_SINGLE_PAGE];
 
-			ret = FALSE;
-		}
 	}
 
 	buf_pool_mutex_exit_all();
 
-	return(ret);
+	return(pending_io);
 }
 
 #if 0

storage/xtradb/buf/buf0flu.c

@@ -767,7 +767,8 @@ buf_flush_buffered_writes(void)
 			if (UNIV_UNLIKELY
 			    (!page_simple_validate_new(block->frame))) {
 corrupted_page:
-				buf_page_print(block->frame, 0);
+				buf_page_print(block->frame, 0,
+					       BUF_PAGE_PRINT_NO_CRASH);
 
 				ut_print_timestamp(stderr);
 				fprintf(stderr,
@@ -2036,22 +2037,6 @@ buf_flush_list(
 
 		buf_pool = buf_pool_from_array(i);
 
-		if (lsn_limit != IB_ULONGLONG_MAX) {
-			buf_page_t*	bpage;
-
-			buf_flush_list_mutex_enter(buf_pool);
-			bpage = UT_LIST_GET_LAST(buf_pool->flush_list);
-			if (!bpage
-			    || bpage->oldest_modification >= lsn_limit) {
-
-				buf_flush_list_mutex_exit(buf_pool);
-				continue;
-			} else {
-
-				buf_flush_list_mutex_exit(buf_pool);
-			}
-		}
-
 		if (!buf_flush_start(buf_pool, BUF_FLUSH_LIST)) {
 			/* We have two choices here. If lsn_limit was
 			specified then skipping an instance of buffer

storage/xtradb/dict/dict0boot.c

@@ -239,6 +239,166 @@ dict_hdr_create(
 	return(TRUE);
 }
 
+/*****************************************************************//**
+Verifies the SYS_STATS table by scanning its clustered index.  This
+function may only be called at InnoDB startup time.
+
+@return	TRUE if SYS_STATS was verified successfully */
+UNIV_INTERN
+ibool
+dict_verify_xtradb_sys_stats(void)
+/*==============================*/
+{
+	dict_index_t* sys_stats_index;
+	ulint	      saved_srv_pass_corrupt_table = srv_pass_corrupt_table;
+	ibool	      result;
+
+	sys_stats_index = dict_table_get_first_index(dict_sys->sys_stats);
+
+	/* Since this may be called only during server startup, avoid hitting
+	   various asserts by using XtraDB pass_corrupt_table option. */
+	srv_pass_corrupt_table = 1;
+	result = btr_validate_index(sys_stats_index, NULL);
+	srv_pass_corrupt_table = saved_srv_pass_corrupt_table;
+
+	return result;
+}
+
+/*****************************************************************//**
+Creates the B-tree for the SYS_STATS clustered index, adds the XtraDB
+mark and the id of the index to the dictionary header page.  Rewrites
+both passed args. */
+static
+void
+dict_create_xtradb_sys_stats(
+/*=========================*/
+	dict_hdr_t**	dict_hdr,	/*!< in/out: dictionary header */
+	mtr_t*		mtr)		/*!< in/out: mtr */
+{
+	ulint	root_page_no;
+
+	root_page_no = btr_create(DICT_CLUSTERED | DICT_UNIQUE,
+				  DICT_HDR_SPACE, 0, DICT_STATS_ID,
+				  dict_ind_redundant, mtr);
+	if (root_page_no == FIL_NULL) {
+		fprintf(stderr, "InnoDB: Warning: failed to create SYS_STATS btr.\n");
+		srv_use_sys_stats_table = FALSE;
+	} else {
+		mlog_write_ulint(*dict_hdr + DICT_HDR_STATS, root_page_no,
+				 MLOG_4BYTES, mtr);
+		mlog_write_ull(*dict_hdr + DICT_HDR_XTRADB_MARK,
+			       DICT_HDR_XTRADB_FLAG, mtr);
+	}
+	mtr_commit(mtr);
+	/* restart mtr */
+	mtr_start(mtr);
+	*dict_hdr = dict_hdr_get(mtr);
+}
+
+/*****************************************************************//**
+Create the table and index structure of SYS_STATS for the dictionary
+cache and add it there.  If called for the first time, also support
+wrong root page id injection for testing purposes. */
+static
+void
+dict_add_to_cache_xtradb_sys_stats(
+/*===============================*/
+	ibool		first_time __attribute__((unused)),
+					/*!< in: first invocation flag. If
+					TRUE, optionally inject wrong root page
+					id */
+	mem_heap_t*	heap,		/*!< in: memory heap for table/index
+					allocation */
+	dict_hdr_t*	dict_hdr,	/*!< in: dictionary header */
+	mtr_t*		mtr)		/*!< in: mtr */
+{
+	dict_table_t*	table;
+	dict_index_t*	index;
+	ulint		root_page_id;
+	ulint		error;
+
+	table = dict_mem_table_create("SYS_STATS", DICT_HDR_SPACE, 4, 0);
+	table->n_mysql_handles_opened = 1; /* for pin */
+
+	dict_mem_table_add_col(table, heap, "INDEX_ID", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "KEY_COLS", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "DIFF_VALS", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "NON_NULL_VALS", DATA_BINARY, 0, 0);
+
+	/* The '+ 2' below comes from the fields DB_TRX_ID, DB_ROLL_PTR */
+#if DICT_SYS_STATS_DIFF_VALS_FIELD != 2 + 2
+#error "DICT_SYS_STATS_DIFF_VALS_FIELD != 2 + 2"
+#endif
+#if DICT_SYS_STATS_NON_NULL_VALS_FIELD != 3 + 2
+#error "DICT_SYS_STATS_NON_NULL_VALS_FIELD != 3 + 2"
+#endif
+
+	table->id = DICT_STATS_ID;
+	dict_table_add_to_cache(table, heap);
+	dict_sys->sys_stats = table;
+	mem_heap_empty(heap);
+
+	index = dict_mem_index_create("SYS_STATS", "CLUST_IND",
+				      DICT_HDR_SPACE,
+				      DICT_UNIQUE | DICT_CLUSTERED, 2);
+
+	dict_mem_index_add_field(index, "INDEX_ID", 0);
+	dict_mem_index_add_field(index, "KEY_COLS", 0);
+
+	index->id = DICT_STATS_ID;
+
+	root_page_id = mtr_read_ulint(dict_hdr + DICT_HDR_STATS, MLOG_4BYTES,
+				      mtr);
+#ifdef UNIV_DEBUG
+	if ((srv_sys_stats_root_page != 0) && first_time)
+		root_page_id = srv_sys_stats_root_page;
+#endif
+	error = dict_index_add_to_cache(table, index, root_page_id, FALSE);
+	ut_a(error == DB_SUCCESS);
+
+	mem_heap_empty(heap);
+}
+
+/*****************************************************************//**
+Discard the existing dictionary cache SYS_STATS information, create and
+add it there anew.  Does not touch the old SYS_STATS tablespace page
+under the assumption that they are corrupted or overwritten for other
+purposes. */
+UNIV_INTERN
+void
+dict_recreate_xtradb_sys_stats(void)
+/*================================*/
+{
+	mtr_t		mtr;
+	dict_hdr_t*	dict_hdr;
+	dict_index_t*	sys_stats_clust_idx;
+	mem_heap_t*	heap;
+
+	heap = mem_heap_create(450);
+
+	mutex_enter(&(dict_sys->mutex));
+
+	sys_stats_clust_idx = dict_table_get_first_index(dict_sys->sys_stats);
+	dict_index_remove_from_cache(dict_sys->sys_stats, sys_stats_clust_idx);
+
+	dict_table_remove_from_cache(dict_sys->sys_stats);
+
+	dict_sys->sys_stats = NULL;
+
+	mtr_start(&mtr);
+
+	dict_hdr = dict_hdr_get(&mtr);
+
+	dict_create_xtradb_sys_stats(&dict_hdr, &mtr);
+	dict_add_to_cache_xtradb_sys_stats(FALSE, heap, dict_hdr, &mtr);
+
+	mem_heap_free(heap);
+
+	mtr_commit(&mtr);
+
+	mutex_exit(&(dict_sys->mutex));
+}
+
 /*****************************************************************//**
 Initializes the data dictionary memory structures when the database is
 started. This function is also called when the data dictionary is created. */
@@ -254,13 +414,13 @@ dict_boot(void)
 	mtr_t		mtr;
 	ulint		error;
 
+	heap = mem_heap_create(450);
+
 	mtr_start(&mtr);
 
 	/* Create the hash tables etc. */
 	dict_init();
 
-	heap = mem_heap_create(450);
-
 	mutex_enter(&(dict_sys->mutex));
 
 	/* Get the dictionary header */
@@ -268,25 +428,9 @@ dict_boot(void)
 
 	if (mach_read_from_8(dict_hdr + DICT_HDR_XTRADB_MARK)
 	    != DICT_HDR_XTRADB_FLAG) {
+
 		/* not extended yet by XtraDB, need to be extended */
-		ulint	root_page_no;
-
-		root_page_no = btr_create(DICT_CLUSTERED | DICT_UNIQUE,
-					  DICT_HDR_SPACE, 0, DICT_STATS_ID,
-					  dict_ind_redundant, &mtr);
-		if (root_page_no == FIL_NULL) {
-			fprintf(stderr, "InnoDB: Warning: failed to create SYS_STATS btr.\n");
-			srv_use_sys_stats_table = FALSE;
-		} else {
-			mlog_write_ulint(dict_hdr + DICT_HDR_STATS, root_page_no,
-					 MLOG_4BYTES, &mtr);
-			mlog_write_ull(dict_hdr + DICT_HDR_XTRADB_MARK,
-					  DICT_HDR_XTRADB_FLAG, &mtr);
-		}
-		mtr_commit(&mtr);
-		/* restart mtr */
-		mtr_start(&mtr);
-		dict_hdr = dict_hdr_get(&mtr);
+		dict_create_xtradb_sys_stats(&dict_hdr, &mtr);
 	}
 
 	/* Because we only write new row ids to disk-based data structure
@@ -465,42 +609,7 @@ dict_boot(void)
 					FALSE);
 	ut_a(error == DB_SUCCESS);
 
-	/*-------------------------*/
-	table = dict_mem_table_create("SYS_STATS", DICT_HDR_SPACE, 4, 0);
-	table->n_mysql_handles_opened = 1; /* for pin */
-
-	dict_mem_table_add_col(table, heap, "INDEX_ID", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, heap, "KEY_COLS", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, heap, "DIFF_VALS", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, heap, "NON_NULL_VALS", DATA_BINARY, 0, 0);
-
-	/* The '+ 2' below comes from the fields DB_TRX_ID, DB_ROLL_PTR */
-#if DICT_SYS_STATS_DIFF_VALS_FIELD != 2 + 2
-#error "DICT_SYS_STATS_DIFF_VALS_FIELD != 2 + 2"
-#endif
-#if DICT_SYS_STATS_NON_NULL_VALS_FIELD != 3 + 2
-#error "DICT_SYS_STATS_NON_NULL_VALS_FIELD != 3 + 2"
-#endif
-
-	table->id = DICT_STATS_ID;
-	dict_table_add_to_cache(table, heap);
-	dict_sys->sys_stats = table;
-	mem_heap_empty(heap);
-
-	index = dict_mem_index_create("SYS_STATS", "CLUST_IND",
-				      DICT_HDR_SPACE,
-				      DICT_UNIQUE | DICT_CLUSTERED, 2);
-
-	dict_mem_index_add_field(index, "INDEX_ID", 0);
-	dict_mem_index_add_field(index, "KEY_COLS", 0);
-
-	index->id = DICT_STATS_ID;
-	error = dict_index_add_to_cache(table, index,
-					mtr_read_ulint(dict_hdr
-						       + DICT_HDR_STATS,
-						       MLOG_4BYTES, &mtr),
-					FALSE);
-	ut_a(error == DB_SUCCESS);
+	dict_add_to_cache_xtradb_sys_stats(TRUE, heap, dict_hdr, &mtr);
 
 	mem_heap_free(heap);
 

storage/xtradb/dict/dict0dict.c

 /*****************************************************************************
 
-Copyright (c) 1996, 2011, Oracle and/or its affiliates. All Rights Reserved.
+Copyright (c) 1996, 2012, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -56,6 +56,8 @@ UNIV_INTERN dict_index_t*	dict_ind_compact;
 #include "ha_prototypes.h" /* innobase_strcasecmp(), innobase_casedn_str()*/
 #include "row0upd.h"
 #include "srv0start.h" /* SRV_LOG_SPACE_FIRST_ID */
+#include "m_string.h"
+#include "my_sys.h"
 
 #include <ctype.h>
 
@@ -2400,6 +2402,8 @@ dict_foreign_free(
 /*==============*/
 	dict_foreign_t*	foreign)	/*!< in, own: foreign key struct */
 {
+	ut_a(foreign->foreign_table->n_foreign_key_checks_running == 0);
+
 	mem_heap_free(foreign->heap);
 }
 
@@ -4381,25 +4385,32 @@ dict_reload_statistics(
 	heap = mem_heap_create(1000);
 
 	while (index) {
+		mtr_t mtr;
+
 		if (table->is_corrupt) {
 			ut_a(srv_pass_corrupt_table);
 			mem_heap_free(heap);
 			return(FALSE);
 		}
 
-		size = btr_get_size(index, BTR_TOTAL_SIZE);
+		mtr_start(&mtr);
+		mtr_s_lock(dict_index_get_lock(index), &mtr);
+
+		size = btr_get_size(index, BTR_TOTAL_SIZE, &mtr);
 
 		index->stat_index_size = size;
 
 		*sum_of_index_sizes += size;
 
-		size = btr_get_size(index, BTR_N_LEAF_PAGES);
+		size = btr_get_size(index, BTR_N_LEAF_PAGES, &mtr);
 
 		if (size == 0) {
 			/* The root node of the tree is a leaf */
 			size = 1;
 		}
 
+		mtr_commit(&mtr);
+
 		index->stat_n_leaf_pages = size;
 
 /*===========================================*/
@@ -4731,6 +4742,7 @@ dict_update_statistics(
 		    (srv_force_recovery < SRV_FORCE_NO_IBUF_MERGE
 		     || (srv_force_recovery < SRV_FORCE_NO_LOG_REDO
 			 && dict_index_is_clust(index)))) {
+			mtr_t	mtr;
 			ulint	size;
 
 			if (table->is_corrupt) {
@@ -4739,15 +4751,24 @@ dict_update_statistics(
 				return;
 			}
 
-			size = btr_get_size(index, BTR_TOTAL_SIZE);
+			mtr_start(&mtr);
+			mtr_s_lock(dict_index_get_lock(index), &mtr);
 
-			index->stat_index_size = size;
+			size = btr_get_size(index, BTR_TOTAL_SIZE, &mtr);
 
-			sum_of_index_sizes += size;
+			if (size != ULINT_UNDEFINED) {
+				sum_of_index_sizes += size;
+				index->stat_index_size = size;
+				size = btr_get_size(
+					index, BTR_N_LEAF_PAGES, &mtr);
+			}
 
-			size = btr_get_size(index, BTR_N_LEAF_PAGES);
+			mtr_commit(&mtr);
 
-			if (size == 0) {
+			switch (size) {
+			case ULINT_UNDEFINED:
+				goto fake_statistics;
+			case 0:
 				/* The root node of the tree is a leaf */
 				size = 1;
 			}
@@ -4764,6 +4785,7 @@ dict_update_statistics(
 			various means, also via secondary indexes. */
 			ulint	i;
 
+fake_statistics:
 			sum_of_index_sizes++;
 			index->stat_index_size = index->stat_n_leaf_pages = 1;
 

storage/xtradb/dict/dict0load.c

@@ -2137,8 +2137,9 @@ static
 void
 dict_load_foreign_cols(
 /*===================*/
-	const char*	id,	/*!< in: foreign constraint id as a
-				null-terminated string */
+	const char*	id,	/*!< in: foreign constraint id, not
+				necessary '\0'-terminated */
+	ulint		id_len,	/*!< in: id length */
 	dict_foreign_t*	foreign)/*!< in: foreign constraint object */
 {
 	dict_table_t*	sys_foreign_cols;
@@ -2168,7 +2169,7 @@ dict_load_foreign_cols(
 	tuple = dtuple_create(foreign->heap, 1);
 	dfield = dtuple_get_nth_field(tuple, 0);
 
-	dfield_set_data(dfield, id, ut_strlen(id));
+	dfield_set_data(dfield, id, id_len);
 	dict_index_copy_types(tuple, sys_index, 1);
 
 	btr_pcur_open_on_user_rec(sys_index, tuple, PAGE_CUR_GE,
@@ -2181,7 +2182,7 @@ dict_load_foreign_cols(
 		ut_a(!rec_get_deleted_flag(rec, 0));
 
 		field = rec_get_nth_field_old(rec, 0, &len);
-		ut_a(len == ut_strlen(id));
+		ut_a(len == id_len);
 		ut_a(ut_memcmp(id, field, len) == 0);
 
 		field = rec_get_nth_field_old(rec, 1, &len);
@@ -2210,8 +2211,9 @@ static
 ulint
 dict_load_foreign(
 /*==============*/
-	const char*	id,	/*!< in: foreign constraint id as a
-				null-terminated string */
+	const char*	id,	/*!< in: foreign constraint id, not
+				necessary '\0'-terminated */
+	ulint		id_len,	/*!< in: id length */
 	ibool		check_charsets,
 				/*!< in: TRUE=check charset compatibility */
 	ibool		check_recursive)
@@ -2247,7 +2249,7 @@ dict_load_foreign(
 	tuple = dtuple_create(heap2, 1);
 	dfield = dtuple_get_nth_field(tuple, 0);
 
-	dfield_set_data(dfield, id, ut_strlen(id));
+	dfield_set_data(dfield, id, id_len);
 	dict_index_copy_types(tuple, sys_index, 1);
 
 	btr_pcur_open_on_user_rec(sys_index, tuple, PAGE_CUR_GE,
@@ -2259,8 +2261,8 @@ dict_load_foreign(
 		/* Not found */
 
 		fprintf(stderr,
-			"InnoDB: Error A: cannot load foreign constraint %s\n",
-			id);
+			"InnoDB: Error A: cannot load foreign constraint "
+			"%.*s\n", (int) id_len, id);
 
 		btr_pcur_close(&pcur);
 		mtr_commit(&mtr);
@@ -2272,11 +2274,11 @@ dict_load_foreign(
 	field = rec_get_nth_field_old(rec, 0, &len);
 
 	/* Check if the id in record is the searched one */
-	if (len != ut_strlen(id) || ut_memcmp(id, field, len) != 0) {
+	if (len != id_len || ut_memcmp(id, field, len) != 0) {
 
 		fprintf(stderr,
-			"InnoDB: Error B: cannot load foreign constraint %s\n",
-			id);
+			"InnoDB: Error B: cannot load foreign constraint "
+			"%.*s\n", (int) id_len, id);
 
 		btr_pcur_close(&pcur);
 		mtr_commit(&mtr);
@@ -2302,7 +2304,7 @@ dict_load_foreign(
 	foreign->type = (unsigned int) (n_fields_and_type >> 24);
 	foreign->n_fields = (unsigned int) (n_fields_and_type & 0x3FFUL);
 
-	foreign->id = mem_heap_strdup(foreign->heap, id);
+	foreign->id = mem_heap_strdupl(foreign->heap, id, id_len);
 
 	field = rec_get_nth_field_old(rec, 3, &len);
 
@@ -2318,7 +2320,7 @@ dict_load_foreign(
 	btr_pcur_close(&pcur);
 	mtr_commit(&mtr);
 
-	dict_load_foreign_cols(id, foreign);
+	dict_load_foreign_cols(id, id_len, foreign);
 
 	ref_table = dict_table_check_if_in_cache_low(
 			foreign->referenced_table_name_lookup);
@@ -2397,8 +2399,8 @@ dict_load_foreigns(
 	ibool		check_charsets)	/*!< in: TRUE=check charset
 					compatibility */
 {
+	char		tuple_buf[DTUPLE_EST_ALLOC(1)];
 	btr_pcur_t	pcur;
-	mem_heap_t*	heap;
 	dtuple_t*	tuple;
 	dfield_t*	dfield;
 	dict_index_t*	sec_index;
@@ -2406,7 +2408,6 @@ dict_load_foreigns(
 	const rec_t*	rec;
 	const byte*	field;
 	ulint		len;
-	char*		id ;
 	ulint		err;
 	mtr_t		mtr;
 
@@ -2433,9 +2434,8 @@ dict_load_foreigns(
 	sec_index = dict_table_get_next_index(
 		dict_table_get_first_index(sys_foreign));
 start_load:
-	heap = mem_heap_create(256);
 
-	tuple  = dtuple_create(heap, 1);
+	tuple = dtuple_create_from_mem(tuple_buf, sizeof(tuple_buf), 1);
 	dfield = dtuple_get_nth_field(tuple, 0);
 
 	dfield_set_data(dfield, table_name, ut_strlen(table_name));
@@ -2489,7 +2489,6 @@ dict_load_foreigns(
 
 	/* Now we get a foreign key constraint id */
 	field = rec_get_nth_field_old(rec, 1, &len);
-	id = mem_heap_strdupl(heap, (char*) field, len);
 
 	btr_pcur_store_position(&pcur, &mtr);
 
@@ -2497,11 +2496,11 @@ dict_load_foreigns(
 
 	/* Load the foreign constraint definition to the dictionary cache */
 
-	err = dict_load_foreign(id, check_charsets, check_recursive);
+	err = dict_load_foreign((char*) field, len, check_charsets,
+				check_recursive);
 
 	if (err != DB_SUCCESS) {
 		btr_pcur_close(&pcur);
-		mem_heap_free(heap);
 
 		return(err);
 	}
@@ -2517,7 +2516,6 @@ dict_load_foreigns(
 load_next_index:
 	btr_pcur_close(&pcur);
 	mtr_commit(&mtr);
-	mem_heap_free(heap);
 
 	sec_index = dict_table_get_next_index(sec_index);
 

storage/xtradb/fil/fil0fil.c

@@ -193,7 +193,7 @@ struct fil_space_struct {
 				.ibd file of tablespace and want to
 				stop temporarily posting of new i/o
 				requests on the file */
-	ibool		stop_ibuf_merges;
+	ibool		stop_new_ops;
 				/*!< we set this TRUE when we start
 				deleting a single-table tablespace */
 	ibool		is_being_deleted;
@@ -218,12 +218,13 @@ struct fil_space_struct {
 	ulint		n_pending_flushes; /*!< this is positive when flushing
 				the tablespace to disk; dropping of the
 				tablespace is forbidden if this is positive */
-	ulint		n_pending_ibuf_merges;/*!< this is positive
-				when merging insert buffer entries to
-				a page so that we may need to access
-				the ibuf bitmap page in the
-				tablespade: dropping of the tablespace
-				is forbidden if this is positive */
+	ulint		n_pending_ops;/*!< this is positive when we
+				have pending operations against this
+				tablespace. The pending operations can
+				be ibuf merges or lock validation code
+				trying to read a block.
+				Dropping of the tablespace is forbidden
+				if this is positive */
 	hash_node_t	hash;	/*!< hash chain node */
 	hash_node_t	name_hash;/*!< hash chain the name_hash table */
 #ifndef UNIV_HOTBACKUP
@@ -974,11 +975,6 @@ fil_mutex_enter_and_prepare_for_io(
 		return;
 	}
 
-	if (fil_system->n_open < fil_system->max_n_open) {
-
-		return;
-	}
-
 	space = fil_space_get_by_id(space_id);
 
 	if (space != NULL && space->stop_ios) {
@@ -995,6 +991,25 @@ fil_mutex_enter_and_prepare_for_io(
 
 		mutex_exit(&fil_system->mutex);
 
+#ifndef UNIV_HOTBACKUP
+
+		/* Wake the i/o-handler threads to make sure pending
+		i/o's are performed */
+		os_aio_simulated_wake_handler_threads();
+
+		/* The sleep here is just to give IO helper threads a
+		bit of time to do some work. It is not required that
+		all IO related to the tablespace being renamed must
+		be flushed here as we do fil_flush() in
+		fil_rename_tablespace() as well. */
+		os_thread_sleep(20000);
+
+#endif /* UNIV_HOTBACKUP */
+
+		/* Flush tablespaces so that we can close modified
+		files in the LRU list */
+		fil_flush_file_spaces(FIL_TABLESPACE);
+
 		os_thread_sleep(20000);
 
 		count2++;
@@ -1002,6 +1017,11 @@ fil_mutex_enter_and_prepare_for_io(
 		goto retry;
 	}
 
+	if (fil_system->n_open < fil_system->max_n_open) {
+
+		return;
+	}
+
 	/* If the file is already open, no need to do anything; if the space
 	does not exist, we handle the situation in the function which called
 	this function */
@@ -1274,7 +1294,7 @@ fil_space_create(
 	}
 
 	space->stop_ios = FALSE;
-	space->stop_ibuf_merges = FALSE;
+	space->stop_new_ops = FALSE;
 	space->is_being_deleted = FALSE;
 	space->purpose = purpose;
 	space->size = 0;
@@ -1283,7 +1303,7 @@ fil_space_create(
 	space->n_reserved_extents = 0;
 
 	space->n_pending_flushes = 0;
-	space->n_pending_ibuf_merges = 0;
+	space->n_pending_ops = 0;
 
 	UT_LIST_INIT(space->chain);
 	space->magic_n = FIL_SPACE_MAGIC_N;
@@ -1896,13 +1916,12 @@ fil_read_first_page(
 
 #ifndef UNIV_HOTBACKUP
 /*******************************************************************//**
-Increments the count of pending insert buffer page merges, if space is not
-being deleted.
-@return	TRUE if being deleted, and ibuf merges should be skipped */
+Increments the count of pending operation, if space is not being deleted.
+@return	TRUE if being deleted, and operation should be skipped */
 UNIV_INTERN
 ibool
-fil_inc_pending_ibuf_merges(
-/*========================*/
+fil_inc_pending_ops(
+/*================*/
 	ulint	id)	/*!< in: space id */
 {
 	fil_space_t*	space;
@@ -1918,13 +1937,13 @@ fil_inc_pending_ibuf_merges(
 			(ulong) id);
 	}
 
-	if (space == NULL || space->stop_ibuf_merges) {
+	if (space == NULL || space->stop_new_ops) {
 		mutex_exit(&fil_system->mutex);
 
 		return(TRUE);
 	}
 
-	space->n_pending_ibuf_merges++;
+	space->n_pending_ops++;
 
 	mutex_exit(&fil_system->mutex);
 
@@ -1932,11 +1951,11 @@ fil_inc_pending_ibuf_merges(
 }
 
 /*******************************************************************//**
-Decrements the count of pending insert buffer page merges. */
+Decrements the count of pending operations. */
 UNIV_INTERN
 void
-fil_decr_pending_ibuf_merges(
-/*=========================*/
+fil_decr_pending_ops(
+/*=================*/
 	ulint	id)	/*!< in: space id */
 {
 	fil_space_t*	space;
@@ -1947,13 +1966,13 @@ fil_decr_pending_ibuf_merges(
 
 	if (space == NULL) {
 		fprintf(stderr,
-			"InnoDB: Error: decrementing ibuf merge of a"
-			" dropped tablespace %lu\n",
+			"InnoDB: Error: decrementing pending operation"
+			" of a dropped tablespace %lu\n",
 			(ulong) id);
 	}
 
 	if (space != NULL) {
-		space->n_pending_ibuf_merges--;
+		space->n_pending_ops--;
 	}
 
 	mutex_exit(&fil_system->mutex);
@@ -2164,7 +2183,7 @@ fil_op_log_parse_or_replay(
 	switch (type) {
 	case MLOG_FILE_DELETE:
 		if (fil_tablespace_exists_in_mem(space_id)) {
-			ut_a(fil_delete_tablespace(space_id));
+			ut_a(fil_delete_tablespace(space_id, TRUE));
 		}
 
 		break;
@@ -2234,7 +2253,9 @@ UNIV_INTERN
 ibool
 fil_delete_tablespace(
 /*==================*/
-	ulint	id)	/*!< in: space id */
+	ulint	id,		/*!< in: space id */
+	ibool	evict_all)	/*!< in: TRUE if we want all pages
+				evicted from LRU. */
 {
 	ibool		success;
 	fil_space_t*	space;
@@ -2243,15 +2264,15 @@ fil_delete_tablespace(
 	char*		path;
 
 	ut_a(id != 0);
-stop_ibuf_merges:
+stop_new_ops:
 	mutex_enter(&fil_system->mutex);
 
 	space = fil_space_get_by_id(id);
 
 	if (space != NULL) {
-		space->stop_ibuf_merges = TRUE;
+		space->stop_new_ops = TRUE;
 
-		if (space->n_pending_ibuf_merges == 0) {
+		if (space->n_pending_ops == 0) {
 			mutex_exit(&fil_system->mutex);
 
 			count = 0;
@@ -2265,9 +2286,10 @@ fil_delete_tablespace(
 				ut_print_filename(stderr, space->name);
 				fprintf(stderr, ",\n"
 					"InnoDB: but there are %lu pending"
-					" ibuf merges on it.\n"
+					" operations (most likely ibuf merges)"
+					" on it.\n"
 					"InnoDB: Loop %lu.\n",
-					(ulong) space->n_pending_ibuf_merges,
+					(ulong) space->n_pending_ops,
 					(ulong) count);
 			}
 
@@ -2276,7 +2298,7 @@ fil_delete_tablespace(
 			os_thread_sleep(20000);
 			count++;
 
-			goto stop_ibuf_merges;
+			goto stop_new_ops;
 		}
 	}
 
@@ -2302,7 +2324,7 @@ fil_delete_tablespace(
 	}
 
 	ut_a(space);
-	ut_a(space->n_pending_ibuf_merges == 0);
+	ut_a(space->n_pending_ops == 0);
 
 	space->is_being_deleted = TRUE;
 
@@ -2358,7 +2380,11 @@ fil_delete_tablespace(
 	if (srv_lazy_drop_table) {
 		buf_LRU_mark_space_was_deleted(id);
 	} else {
-	buf_LRU_invalidate_tablespace(id);
+	buf_LRU_flush_or_remove_pages(
+		id, evict_all
+		? BUF_REMOVE_ALL_NO_WRITE
+		: BUF_REMOVE_FLUSH_NO_WRITE);
+
 	}
 #endif
 	/* printf("Deleting tablespace %s id %lu\n", space->name, id); */
@@ -2447,7 +2473,7 @@ fil_discard_tablespace(
 {
 	ibool	success;
 
-	success = fil_delete_tablespace(id);
+	success = fil_delete_tablespace(id, TRUE);
 
 	if (!success) {
 		fprintf(stderr,
@@ -2579,7 +2605,7 @@ fil_rename_tablespace(
 retry:
 	count++;
 
-	if (count > 1000) {
+	if (!(count % 1000)) {
 		ut_print_timestamp(stderr);
 		fputs("  InnoDB: Warning: problems renaming ", stderr);
 		ut_print_filename(stderr, old_name);
@@ -3905,7 +3931,7 @@ fil_open_single_table_tablespace(
 
 					level = btr_page_get_level(page, &mtr);
 
-					new_block = btr_page_alloc(index, 0, FSP_NO_DIR, level, &mtr);
+					new_block = btr_page_alloc(index, 0, FSP_NO_DIR, level, &mtr, &mtr);
 					new_page = buf_block_get_frame(new_block);
 					new_page_zip = buf_block_get_page_zip(new_block);
 					btr_page_create(new_block, new_page_zip, index, level, &mtr);
@@ -3953,7 +3979,7 @@ fil_open_single_table_tablespace(
 				split_rec = page_get_middle_rec(page);
 
 				new_block = btr_page_alloc(index, page_no + 1, FSP_UP,
-							   btr_page_get_level(page, &mtr), &mtr);
+							   btr_page_get_level(page, &mtr), &mtr, &mtr);
 				new_page = buf_block_get_frame(new_block);
 				new_page_zip = buf_block_get_page_zip(new_block);
 				btr_page_create(new_block, new_page_zip, index,

storage/xtradb/handler/ha_innodb.h

@@ -79,13 +79,14 @@ class ha_innobase: public handler
 	INNOBASE_SHARE*	share;		/*!< information for MySQL
 					table locking */
 
-	uchar*		upd_buff;	/*!< buffer used in updates */
-	uchar*		key_val_buff;	/*!< buffer used in converting
+	uchar*		upd_buf;	/*!< buffer used in updates */
+	ulint		upd_buf_size;	/*!< the size of upd_buf in bytes */
+	uchar		srch_key_val1[REC_VERSION_56_MAX_INDEX_COL_LEN + 2];
+	uchar		srch_key_val2[REC_VERSION_56_MAX_INDEX_COL_LEN + 2];
+					/*!< buffers used in converting
 					search key values from MySQL format
-					to Innodb format */
-	ulong		upd_and_key_val_buff_len;
-					/* the length of each of the previous
-					two buffers */
+					to InnoDB format. "+ 2" for the two
+					bytes where the length is stored */
 	Table_flags	int_table_flags;
 	uint		primary_key;
 	ulong		start_of_scan;	/*!< this is set to 1 when we are

storage/xtradb/handler/handler0alter.cc

 /*****************************************************************************
 
-Copyright (c) 2005, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 2005, 2012, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -878,6 +878,8 @@ ha_innobase::add_index(
 					prebuilt->table, indexed_table,
 					index, num_of_idx, table);
 
+	DBUG_EXECUTE_IF("crash_innodb_add_index_after", DBUG_SUICIDE(););
+
 error_handling:
 	/* After an error, remove all those index definitions from the
 	dictionary which were defined. */
@@ -1020,7 +1022,12 @@ ha_innobase::final_add_index(
 			row_prebuilt_free(prebuilt, TRUE);
 			error = row_merge_drop_table(trx, old_table);
 			add->indexed_table->n_mysql_handles_opened++;
-			prebuilt = row_create_prebuilt(add->indexed_table);
+			prebuilt = row_create_prebuilt(add->indexed_table,
+				0 /* XXX Do we know the mysql_row_len here?
+				Before the addition of this parameter to
+				row_create_prebuilt() the mysql_row_len
+				member was left 0 (from zalloc) in the
+				prebuilt object. */);
 		}
 
 		err = convert_error_code_to_mysql(
@@ -1154,7 +1161,9 @@ ha_innobase::prepare_drop_index(
 			goto func_exit;
 		}
 
+		rw_lock_x_lock(dict_index_get_lock(index));
 		index->to_be_dropped = TRUE;
+		rw_lock_x_unlock(dict_index_get_lock(index));
 	}
 
 	/* If FOREIGN_KEY_CHECKS = 1 you may not drop an index defined
@@ -1273,7 +1282,9 @@ ha_innobase::prepare_drop_index(
 			= dict_table_get_first_index(prebuilt->table);
 
 		do {
+			rw_lock_x_lock(dict_index_get_lock(index));
 			index->to_be_dropped = FALSE;
+			rw_lock_x_unlock(dict_index_get_lock(index));
 			index = dict_table_get_next_index(index);
 		} while (index);
 	}
@@ -1339,7 +1350,9 @@ ha_innobase::final_drop_index(
 		for (index = dict_table_get_first_index(prebuilt->table);
 		     index; index = dict_table_get_next_index(index)) {
 
+			rw_lock_x_lock(dict_index_get_lock(index));
 			index->to_be_dropped = FALSE;
+			rw_lock_x_unlock(dict_index_get_lock(index));
 		}
 
 		goto func_exit;

storage/xtradb/ibuf/ibuf0ibuf.c

 /*****************************************************************************
 
-Copyright (c) 1997, 2011, Oracle and/or its affiliates. All Rights Reserved.
+Copyright (c) 1997, 2012, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -1523,7 +1523,7 @@ ibuf_add_ops(
 
 	for (i = 0; i < IBUF_OP_COUNT; i++) {
 #ifdef HAVE_ATOMIC_BUILTINS
-		os_atomic_increment_ulint(&arr[i], ops[i]);
+		(void) os_atomic_increment_ulint(&arr[i], ops[i]);
 #else /* HAVE_ATOMIC_BUILTINS */
 		arr[i] += ops[i];
 #endif /* HAVE_ATOMIC_BUILTINS */
@@ -2222,14 +2222,14 @@ ibool
 ibuf_add_free_page(void)
 /*====================*/
 {
-	mtr_t	mtr;
-	page_t*	header_page;
-	ulint	flags;
-	ulint	zip_size;
-	ulint	page_no;
-	page_t*	page;
-	page_t*	root;
-	page_t*	bitmap_page;
+	mtr_t		mtr;
+	page_t*		header_page;
+	ulint		flags;
+	ulint		zip_size;
+	buf_block_t*	block;
+	page_t*		page;
+	page_t*		root;
+	page_t*		bitmap_page;
 
 	mtr_start(&mtr);
 
@@ -2250,28 +2250,23 @@ ibuf_add_free_page(void)
 	of a deadlock. This is the reason why we created a special ibuf
 	header page apart from the ibuf tree. */
 
-	page_no = fseg_alloc_free_page(
+	block = fseg_alloc_free_page(
 		header_page + IBUF_HEADER + IBUF_TREE_SEG_HEADER, 0, FSP_UP,
 		&mtr);
 
-	if (UNIV_UNLIKELY(page_no == FIL_NULL)) {
+	if (block == NULL) {
 		mtr_commit(&mtr);
 
 		return(FALSE);
-	} else {
-		buf_block_t*	block = buf_page_get(
-			IBUF_SPACE_ID, 0, page_no, RW_X_LATCH, &mtr);
-
-		ibuf_enter(&mtr);
-
-		mutex_enter(&ibuf_mutex);
-
-		root = ibuf_tree_root_get(&mtr);
+	}
 
-		buf_block_dbg_add_level(block, SYNC_IBUF_TREE_NODE_NEW);
+	ut_ad(rw_lock_get_x_lock_count(&block->lock) == 1);
+	ibuf_enter(&mtr);
+	mutex_enter(&ibuf_mutex);
+	root = ibuf_tree_root_get(&mtr);
 
-		page = buf_block_get_frame(block);
-	}
+	buf_block_dbg_add_level(block, SYNC_IBUF_TREE_NODE_NEW);
+	page = buf_block_get_frame(block);
 
 	/* Add the page to the free list and update the ibuf size data */
 
@@ -2288,12 +2283,13 @@ ibuf_add_free_page(void)
 	(level 2 page) */
 
 	bitmap_page = ibuf_bitmap_get_map_page(
-		IBUF_SPACE_ID, page_no, zip_size, &mtr);
+		IBUF_SPACE_ID, buf_block_get_page_no(block), zip_size, &mtr);
 
 	mutex_exit(&ibuf_mutex);
 
 	ibuf_bitmap_page_set_bits(
-		bitmap_page, page_no, zip_size, IBUF_BITMAP_IBUF, TRUE, &mtr);
+		bitmap_page, buf_block_get_page_no(block), zip_size,
+		IBUF_BITMAP_IBUF, TRUE, &mtr);
 
 	ibuf_mtr_commit(&mtr);
 
@@ -2588,7 +2584,15 @@ ibuf_get_merge_page_nos_func(
 		} else {
 			rec_page_no = ibuf_rec_get_page_no(mtr, rec);
 			rec_space_id = ibuf_rec_get_space(mtr, rec);
-			ut_ad(rec_page_no > IBUF_TREE_ROOT_PAGE_NO);
+			/* In the system tablespace, the smallest
+			possible secondary index leaf page number is
+			bigger than IBUF_TREE_ROOT_PAGE_NO (4). In
+			other tablespaces, the clustered index tree is
+			created at page 3, which makes page 4 the
+			smallest possible secondary index leaf page
+			(and that only after DROP INDEX). */
+			ut_ad(rec_page_no
+			      > IBUF_TREE_ROOT_PAGE_NO - (rec_space_id != 0));
 		}
 
 #ifdef UNIV_IBUF_DEBUG
@@ -3920,6 +3924,7 @@ ibuf_insert_to_index_page_low(
 
 	fputs("InnoDB: Submit a detailed bug report"
 	      " to http://bugs.mysql.com\n", stderr);
+	ut_ad(0);
 }
 
 /************************************************************************
@@ -3970,9 +3975,10 @@ ibuf_insert_to_index_page(
 		      "InnoDB: but the number of fields does not match!\n",
 		      stderr);
 dump:
-		buf_page_print(page, 0);
+		buf_page_print(page, 0, BUF_PAGE_PRINT_NO_CRASH);
 
 		dtuple_print(stderr, entry);
+		ut_ad(0);
 
 		fputs("InnoDB: The table where where"
 		      " this index record belongs\n"
@@ -4113,6 +4119,11 @@ ibuf_set_del_mark(
 							  TRUE, mtr);
 		}
 	} else {
+		const page_t*		page
+			= page_cur_get_page(&page_cur);
+		const buf_block_t*	block
+			= page_cur_get_block(&page_cur);
+
 		ut_print_timestamp(stderr);
 		fputs("  InnoDB: unable to find a record to delete-mark\n",
 		      stderr);
@@ -4121,10 +4132,14 @@ ibuf_set_del_mark(
 		fputs("\n"
 		      "InnoDB: record ", stderr);
 		rec_print(stderr, page_cur_get_rec(&page_cur), index);
-		putc('\n', stderr);
-		fputs("\n"
-		      "InnoDB: Submit a detailed bug report"
-		      " to http://bugs.mysql.com\n", stderr);
+		fprintf(stderr, "\nspace %u offset %u"
+			" (%u records, index id %llu)\n"
+			"InnoDB: Submit a detailed bug report"
+			" to http://bugs.mysql.com\n",
+			(unsigned) buf_block_get_space(block),
+			(unsigned) buf_block_get_page_no(block),
+			(unsigned) page_get_n_recs(page),
+			(ulonglong) btr_page_get_index_id(page));
 		ut_ad(0);
 	}
 }
@@ -4168,12 +4183,31 @@ ibuf_delete(
 		offsets = rec_get_offsets(
 			rec, index, offsets, ULINT_UNDEFINED, &heap);
 
-		/* Refuse to delete the last record. */
-		ut_a(page_get_n_recs(page) > 1);
-
-		/* The record should have been marked for deletion. */
-		ut_ad(REC_INFO_DELETED_FLAG
-		      & rec_get_info_bits(rec, page_is_comp(page)));
+		if (page_get_n_recs(page) <= 1
+		    || !(REC_INFO_DELETED_FLAG
+			 & rec_get_info_bits(rec, page_is_comp(page)))) {
+			/* Refuse to purge the last record or a
+			record that has not been marked for deletion. */
+			ut_print_timestamp(stderr);
+			fputs("  InnoDB: unable to purge a record\n",
+			      stderr);
+			fputs("InnoDB: tuple ", stderr);
+			dtuple_print(stderr, entry);
+			fputs("\n"
+			      "InnoDB: record ", stderr);
+			rec_print_new(stderr, rec, offsets);
+			fprintf(stderr, "\nspace %u offset %u"
+				" (%u records, index id %llu)\n"
+				"InnoDB: Submit a detailed bug report"
+				" to http://bugs.mysql.com\n",
+				(unsigned) buf_block_get_space(block),
+				(unsigned) buf_block_get_page_no(block),
+				(unsigned) page_get_n_recs(page),
+				(ulonglong) btr_page_get_index_id(page));
+
+			ut_ad(0);
+			return;
+		}
 
 		lock_update_delete(block, rec);
 
@@ -4259,6 +4293,7 @@ ibuf_restore_pos(
 
 		fprintf(stderr, "InnoDB: ibuf tree ok\n");
 		fflush(stderr);
+		ut_ad(0);
 	}
 
 	return(FALSE);
@@ -4441,7 +4476,7 @@ ibuf_merge_or_delete_for_page(
 		function. When the counter is > 0, that prevents tablespace
 		from being dropped. */
 
-		tablespace_being_deleted = fil_inc_pending_ibuf_merges(space);
+		tablespace_being_deleted = fil_inc_pending_ops(space);
 
 		if (UNIV_UNLIKELY(tablespace_being_deleted)) {
 			/* Do not try to read the bitmap page from space;
@@ -4467,7 +4502,7 @@ ibuf_merge_or_delete_for_page(
 				/* No inserts buffered for this page */
 
 				if (!tablespace_being_deleted) {
-					fil_decr_pending_ibuf_merges(space);
+					fil_decr_pending_ops(space);
 				}
 
 				return;
@@ -4516,12 +4551,14 @@ ibuf_merge_or_delete_for_page(
 
 			bitmap_page = ibuf_bitmap_get_map_page(space, page_no,
 							       zip_size, &mtr);
-			buf_page_print(bitmap_page, 0);
+			buf_page_print(bitmap_page, 0,
+				       BUF_PAGE_PRINT_NO_CRASH);
 			ibuf_mtr_commit(&mtr);
 
 			fputs("\nInnoDB: Dump of the page:\n", stderr);
 
-			buf_page_print(block->frame, 0);
+			buf_page_print(block->frame, 0,
+				       BUF_PAGE_PRINT_NO_CRASH);
 
 			fprintf(stderr,
 				"InnoDB: Error: corruption in the tablespace."
@@ -4541,6 +4578,7 @@ ibuf_merge_or_delete_for_page(
 				(ulong) page_no,
 				(ulong)
 				fil_page_get_type(block->frame));
+			ut_ad(0);
 		}
 	}
 
@@ -4747,7 +4785,7 @@ ibuf_merge_or_delete_for_page(
 	mem_heap_free(heap);
 
 #ifdef HAVE_ATOMIC_BUILTINS
-	os_atomic_increment_ulint(&ibuf->n_merges, 1);
+	(void) os_atomic_increment_ulint(&ibuf->n_merges, 1);
 	ibuf_add_ops(ibuf->n_merged_ops, mops);
 	ibuf_add_ops(ibuf->n_discarded_ops, dops);
 #else /* HAVE_ATOMIC_BUILTINS */
@@ -4763,7 +4801,7 @@ ibuf_merge_or_delete_for_page(
 
 	if (update_ibuf_bitmap && !tablespace_being_deleted) {
 
-		fil_decr_pending_ibuf_merges(space);
+		fil_decr_pending_ops(space);
 	}
 
 #ifdef UNIV_IBUF_COUNT_DEBUG

storage/xtradb/include/btr0btr.h

 /*****************************************************************************
 
-Copyright (c) 1994, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1994, 2012, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -92,6 +92,26 @@ insert/delete buffer when the record is not in the buffer pool. */
 buffer when the record is not in the buffer pool. */
 #define BTR_DELETE		8192
 
+/**************************************************************//**
+Report that an index page is corrupted. */
+UNIV_INTERN
+void
+btr_corruption_report(
+/*==================*/
+	const buf_block_t*	block,	/*!< in: corrupted block */
+	const dict_index_t*	index)	/*!< in: index tree */
+	UNIV_COLD __attribute__((nonnull));
+
+/** Assert that a B-tree page is not corrupted.
+@param block buffer block containing a B-tree page
+@param index the B-tree index */
+#define btr_assert_not_corrupted(block, index)			\
+	if ((ibool) !!page_is_comp(buf_block_get_frame(block))	\
+	    != dict_table_is_comp((index)->table)) {		\
+		btr_corruption_report(block, index);		\
+		ut_error;					\
+	}
+
 #ifdef UNIV_BLOB_DEBUG
 # include "ut0rbt.h"
 /** An index->blobs entry for keeping track of off-page column references */
@@ -569,11 +589,14 @@ UNIV_INTERN
 ibool
 btr_compress(
 /*=========*/
-	btr_cur_t*	cursor,	/*!< in: cursor on the page to merge or lift;
-				the page must not be empty: in record delete
-				use btr_discard_page if the page would become
-				empty */
-	mtr_t*		mtr);	/*!< in: mtr */
+	btr_cur_t*	cursor,	/*!< in/out: cursor on the page to merge
+				or lift; the page must not be empty:
+				when deleting records, use btr_discard_page()
+				if the page would become empty */
+	ibool		adjust,	/*!< in: TRUE if should adjust the
+				cursor position even if compression occurs */
+	mtr_t*		mtr)	/*!< in/out: mini-transaction */
+	__attribute__((nonnull));
 /*************************************************************//**
 Discards a page from a B-tree. This is used to remove the last record from
 a B-tree page: the whole page must be removed at the same time. This cannot
@@ -614,17 +637,23 @@ btr_parse_page_reorganize(
 #ifndef UNIV_HOTBACKUP
 /**************************************************************//**
 Gets the number of pages in a B-tree.
-@return	number of pages */
+@return	number of pages, or ULINT_UNDEFINED if the index is unavailable */
 UNIV_INTERN
 ulint
 btr_get_size(
 /*=========*/
 	dict_index_t*	index,	/*!< in: index */
-	ulint		flag);	/*!< in: BTR_N_LEAF_PAGES or BTR_TOTAL_SIZE */
+	ulint		flag,	/*!< in: BTR_N_LEAF_PAGES or BTR_TOTAL_SIZE */
+	mtr_t*		mtr)	/*!< in/out: mini-transaction where index
+				is s-latched */
+	__attribute__((nonnull, warn_unused_result));
 /**************************************************************//**
 Allocates a new file page to be used in an index tree. NOTE: we assume
 that the caller has made the reservation for free extents!
-@return	new allocated block, x-latched; NULL if out of space */
+@retval NULL if no page could be allocated
+@retval block, rw_lock_x_lock_count(&block->lock) == 1 if allocation succeeded
+(init_mtr == mtr, or the page was not previously freed in mtr)
+@retval block (not allocated or initialized) otherwise */
 UNIV_INTERN
 buf_block_t*
 btr_page_alloc(
@@ -635,7 +664,12 @@ btr_page_alloc(
 					page split is made */
 	ulint		level,		/*!< in: level where the page is placed
 					in the tree */
-	mtr_t*		mtr);		/*!< in: mtr */
+	mtr_t*		mtr,		/*!< in/out: mini-transaction
+					for the allocation */
+	mtr_t*		init_mtr)	/*!< in/out: mini-transaction
+					for x-latching and initializing
+					the page */
+	__attribute__((nonnull, warn_unused_result));
 /**************************************************************//**
 Frees a file page used in an index tree. NOTE: cannot free field external
 storage pages because the page must contain info on its level. */

storage/xtradb/include/btr0btr.ic

@@ -279,7 +279,7 @@ btr_node_ptr_get_child_page_no(
 			"InnoDB: a nonsensical page number 0"
 			" in a node ptr record at offset %lu\n",
 			(ulong) page_offset(rec));
-		buf_page_print(page_align(rec), 0);
+		buf_page_print(page_align(rec), 0, 0);
 	}
 
 	return(page_no);

storage/xtradb/include/btr0cur.h

 /*****************************************************************************
 
-Copyright (c) 1994, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1994, 2012, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -36,6 +36,9 @@ Created 10/16/1994 Heikki Tuuri
 #define BTR_NO_LOCKING_FLAG	2	/* do no record lock checking */
 #define BTR_KEEP_SYS_FLAG	4	/* sys fields will be found from the
 					update vector or inserted entry */
+#define BTR_KEEP_POS_FLAG	8	/* btr_cur_pessimistic_update()
+					must keep cursor position when
+					moving columns to big_rec */
 
 #ifndef UNIV_HOTBACKUP
 #include "que0types.h"
@@ -310,7 +313,9 @@ btr_cur_pessimistic_update(
 /*=======================*/
 	ulint		flags,	/*!< in: undo logging, locking, and rollback
 				flags */
-	btr_cur_t*	cursor,	/*!< in: cursor on the record to update */
+	btr_cur_t*	cursor,	/*!< in/out: cursor on the record to update;
+				cursor may become invalid if *big_rec == NULL
+				|| !(flags & BTR_KEEP_POS_FLAG) */
 	mem_heap_t**	heap,	/*!< in/out: pointer to memory heap, or NULL */
 	big_rec_t**	big_rec,/*!< out: big rec vector whose fields have to
 				be stored externally by the caller, or NULL */
@@ -364,10 +369,13 @@ UNIV_INTERN
 ibool
 btr_cur_compress_if_useful(
 /*=======================*/
-	btr_cur_t*	cursor,	/*!< in: cursor on the page to compress;
+	btr_cur_t*	cursor,	/*!< in/out: cursor on the page to compress;
 				cursor does not stay valid if compression
 				occurs */
-	mtr_t*		mtr);	/*!< in: mtr */
+	ibool		adjust,	/*!< in: TRUE if should adjust the
+				cursor position even if compression occurs */
+	mtr_t*		mtr)	/*!< in/out: mini-transaction */
+	__attribute__((nonnull));
 /*******************************************************//**
 Removes the record on which the tree cursor is positioned. It is assumed
 that the mtr has an x-latch on the page where the cursor is positioned,
@@ -492,6 +500,27 @@ btr_cur_disown_inherited_fields(
 	const upd_t*	update,	/*!< in: update vector */
 	mtr_t*		mtr)	/*!< in/out: mini-transaction */
 	__attribute__((nonnull(2,3,4,5,6)));
+
+/** Operation code for btr_store_big_rec_extern_fields(). */
+enum blob_op {
+	/** Store off-page columns for a freshly inserted record */
+	BTR_STORE_INSERT = 0,
+	/** Store off-page columns for an insert by update */
+	BTR_STORE_INSERT_UPDATE,
+	/** Store off-page columns for an update */
+	BTR_STORE_UPDATE
+};
+
+/*******************************************************************//**
+Determine if an operation on off-page columns is an update.
+@return TRUE if op != BTR_STORE_INSERT */
+UNIV_INLINE
+ibool
+btr_blob_op_is_update(
+/*==================*/
+	enum blob_op	op)	/*!< in: operation */
+	__attribute__((warn_unused_result));
+
 /*******************************************************************//**
 Stores the fields in big_rec_vec to the tablespace and puts pointers to
 them in rec.  The extern flags in rec will have to be set beforehand.
@@ -499,52 +528,23 @@ The fields are stored on pages allocated from leaf node
 file segment of the index tree.
 @return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
 UNIV_INTERN
-ulint
-btr_store_big_rec_extern_fields_func(
-/*=================================*/
+enum db_err
+btr_store_big_rec_extern_fields(
+/*============================*/
 	dict_index_t*	index,		/*!< in: index of rec; the index tree
 					MUST be X-latched */
 	buf_block_t*	rec_block,	/*!< in/out: block containing rec */
-	rec_t*		rec,		/*!< in: record */
+	rec_t*		rec,		/*!< in/out: record */
 	const ulint*	offsets,	/*!< in: rec_get_offsets(rec, index);
 					the "external storage" flags in offsets
 					will not correspond to rec when
 					this function returns */
-#ifdef UNIV_DEBUG
-	mtr_t*		local_mtr,	/*!< in: mtr containing the
-					latch to rec and to the tree */
-#endif /* UNIV_DEBUG */
-#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
-	ibool		update_in_place,/*! in: TRUE if the record is updated
-					in place (not delete+insert) */
-#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
-	const big_rec_t*big_rec_vec)	/*!< in: vector containing fields
+	const big_rec_t*big_rec_vec,	/*!< in: vector containing fields
 					to be stored externally */
-	__attribute__((nonnull));
-
-/** Stores the fields in big_rec_vec to the tablespace and puts pointers to
-them in rec.  The extern flags in rec will have to be set beforehand.
-The fields are stored on pages allocated from leaf node
-file segment of the index tree.
-@param index	in: clustered index; MUST be X-latched by mtr
-@param b	in/out: block containing rec; MUST be X-latched by mtr
-@param rec	in/out: clustered index record
-@param offsets	in: rec_get_offsets(rec, index);
-		the "external storage" flags in offsets will not be adjusted
-@param mtr	in: mini-transaction that holds x-latch on index and b
-@param upd	in: TRUE if the record is updated in place (not delete+insert)
-@param big	in: vector containing fields to be stored externally
-@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
-#ifdef UNIV_DEBUG
-# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
-	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,mtr,upd,big)
-#elif defined UNIV_BLOB_LIGHT_DEBUG
-# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
-	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,upd,big)
-#else
-# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
-	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,big)
-#endif
+	mtr_t*		btr_mtr,	/*!< in: mtr containing the
+					latches to the clustered index */
+	enum blob_op	op)		/*! in: operation code */
+	__attribute__((nonnull, warn_unused_result));
 
 /*******************************************************************//**
 Frees the space in an externally stored field to the file space

storage/xtradb/include/btr0cur.ic

 /*****************************************************************************
 
-Copyright (c) 1994, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1994, 2012, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -139,7 +139,7 @@ btr_cur_compress_recommendation(
 	btr_cur_t*	cursor,	/*!< in: btr cursor */
 	mtr_t*		mtr)	/*!< in: mtr */
 {
-	page_t*		page;
+	const page_t*	page;
 
 	ut_ad(mtr_memo_contains(mtr, btr_cur_get_block(cursor),
 				MTR_MEMO_PAGE_X_FIX));
@@ -197,4 +197,25 @@ btr_cur_can_delete_without_compress(
 
 	return(TRUE);
 }
+
+/*******************************************************************//**
+Determine if an operation on off-page columns is an update.
+@return TRUE if op != BTR_STORE_INSERT */
+UNIV_INLINE
+ibool
+btr_blob_op_is_update(
+/*==================*/
+	enum blob_op	op)	/*!< in: operation */
+{
+	switch (op) {
+	case BTR_STORE_INSERT:
+		return(FALSE);
+	case BTR_STORE_INSERT_UPDATE:
+	case BTR_STORE_UPDATE:
+		return(TRUE);
+	}
+
+	ut_ad(0);
+	return(FALSE);
+}
 #endif /* !UNIV_HOTBACKUP */

storage/xtradb/include/btr0pcur.h

@@ -53,6 +53,16 @@ UNIV_INTERN
 btr_pcur_t*
 btr_pcur_create_for_mysql(void);
 /*============================*/
+
+/**************************************************************//**
+Resets a persistent cursor object, freeing ::old_rec_buf if it is
+allocated and resetting the other members to their initial values. */
+UNIV_INTERN
+void
+btr_pcur_reset(
+/*===========*/
+	btr_pcur_t*	cursor);/*!< in, out: persistent cursor */
+
 /**************************************************************//**
 Frees the memory for a persistent cursor object. */
 UNIV_INTERN

storage/xtradb/include/buf0buf.h

 /*****************************************************************************
 
-Copyright (c) 1995, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2012, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -611,6 +611,31 @@ buf_block_get_modify_clock(
 #else /* !UNIV_HOTBACKUP */
 # define buf_block_modify_clock_inc(block) ((void) 0)
 #endif /* !UNIV_HOTBACKUP */
+/*******************************************************************//**
+Increments the bufferfix count. */
+UNIV_INLINE
+void
+buf_block_buf_fix_inc_func(
+/*=======================*/
+#ifdef UNIV_SYNC_DEBUG
+	const char*	file,	/*!< in: file name */
+	ulint		line,	/*!< in: line */
+#endif /* UNIV_SYNC_DEBUG */
+	buf_block_t*	block)	/*!< in/out: block to bufferfix */
+	__attribute__((nonnull));
+#ifdef UNIV_SYNC_DEBUG
+/** Increments the bufferfix count.
+@param b	in/out: block to bufferfix
+@param f	in: file name where requested
+@param l	in: line number where requested */
+# define buf_block_buf_fix_inc(b,f,l) buf_block_buf_fix_inc_func(f,l,b)
+#else /* UNIV_SYNC_DEBUG */
+/** Increments the bufferfix count.
+@param b	in/out: block to bufferfix
+@param f	in: file name where requested
+@param l	in: line number where requested */
+# define buf_block_buf_fix_inc(b,f,l) buf_block_buf_fix_inc_func(b)
+#endif /* UNIV_SYNC_DEBUG */
 /********************************************************************//**
 Calculates a page checksum which is stored to the page when it is written
 to a file. Note that we must be careful to calculate the same value
@@ -700,6 +725,13 @@ buf_print(void);
 /*============*/
 #endif /* UNIV_DEBUG_PRINT || UNIV_DEBUG || UNIV_BUF_DEBUG */
 #endif /* !UNIV_HOTBACKUP */
+enum buf_page_print_flags {
+	/** Do not crash at the end of buf_page_print(). */
+	BUF_PAGE_PRINT_NO_CRASH	= 1,
+	/** Do not print the full page dump. */
+	BUF_PAGE_PRINT_NO_FULL = 2
+};
+
 /********************************************************************//**
 Prints a page to stderr. */
 UNIV_INTERN
@@ -707,8 +739,12 @@ void
 buf_page_print(
 /*===========*/
 	const byte*	read_buf,	/*!< in: a database page */
-	ulint		zip_size);	/*!< in: compressed page size, or
+	ulint		zip_size,	/*!< in: compressed page size, or
 					0 for uncompressed pages */
+	ulint		flags)		/*!< in: 0 or
+					BUF_PAGE_PRINT_NO_CRASH or
+					BUF_PAGE_PRINT_NO_FULL */
+	__attribute__((nonnull));
 /********************************************************************//**
 Decompress a block.
 @return	TRUE if successful */
@@ -773,11 +809,11 @@ buf_all_freed(void);
 /*********************************************************************//**
 Checks that there currently are no pending i/o-operations for the buffer
 pool.
-@return	TRUE if there is no pending i/o */
+@return	number of pending i/o operations */
 UNIV_INTERN
-ibool
-buf_pool_check_no_pending_io(void);
-/*==============================*/
+ulint
+buf_pool_check_num_pending_io(void);
+/*===============================*/
 /*********************************************************************//**
 Invalidates the file pages in the buffer pool when an archive recovery is
 completed. All the file pages buffered must be in a replaceable state when

storage/xtradb/include/buf0buf.ic

 /*****************************************************************************
 
-Copyright (c) 1995, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2012, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -18,8 +18,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -1026,19 +1026,6 @@ buf_block_buf_fix_inc_func(
 
 	block->page.buf_fix_count++;
 }
-#ifdef UNIV_SYNC_DEBUG
-/** Increments the bufferfix count.
-@param b	in/out: block to bufferfix
-@param f	in: file name where requested
-@param l	in: line number where requested */
-# define buf_block_buf_fix_inc(b,f,l) buf_block_buf_fix_inc_func(f,l,b)
-#else /* UNIV_SYNC_DEBUG */
-/** Increments the bufferfix count.
-@param b	in/out: block to bufferfix
-@param f	in: file name where requested
-@param l	in: line number where requested */
-# define buf_block_buf_fix_inc(b,f,l) buf_block_buf_fix_inc_func(b)
-#endif /* UNIV_SYNC_DEBUG */
 
 /*******************************************************************//**
 Decrements the bufferfix count. */
@@ -1295,7 +1282,7 @@ buf_block_dbg_add_level(
 				where we have acquired latch */
 	ulint		level)	/*!< in: latching order level */
 {
-	sync_thread_add_level(&block->lock, level);
+	sync_thread_add_level(&block->lock, level, FALSE);
 }
 #endif /* UNIV_SYNC_DEBUG */
 /********************************************************************//**

storage/xtradb/include/buf0lru.h

@@ -64,15 +64,14 @@ These are low-level functions
 #define BUF_LRU_FREE_SEARCH_LEN(b)	(5 + 2 * BUF_READ_AHEAD_AREA(b))
 
 /******************************************************************//**
-Invalidates all pages belonging to a given tablespace when we are deleting
-the data file(s) of that tablespace. A PROBLEM: if readahead is being started,
-what guarantees that it will not try to read in pages after this operation has
-completed? */
+Removes all pages belonging to a given tablespace. */
 UNIV_INTERN
 void
-buf_LRU_invalidate_tablespace(
+buf_LRU_flush_or_remove_pages(
 /*==========================*/
-	ulint	id);	/*!< in: space id */
+	ulint			id,	/*!< in: space id */
+	enum buf_remove_t	buf_remove);/*!< in: remove or flush
+					strategy */
 
 /******************************************************************//**
 */