Updated with changes from Percona Server 5.1.66-rel14.1 tarball.

CMakeLists.txt

@@ -60,7 +60,7 @@ SET(INNODB_PLUGIN_SOURCES	btr/btr0btr.c btr/btr0cur.c btr/btr0pcur.c btr/btr0sea
 			ibuf/ibuf0ibuf.c
 			pars/lexyy.c pars/pars0grm.c pars/pars0opt.c pars/pars0pars.c pars/pars0sym.c
 			lock/lock0lock.c lock/lock0iter.c
-			log/log0log.c log/log0recv.c
+			log/log0log.c log/log0recv.c log/log0online.c
 			mach/mach0data.c
 			mem/mem0mem.c mem/mem0pool.c
 			mtr/mtr0log.c mtr/mtr0mtr.c

ChangeLog

+2012-08-29	The InnoDB Team
+
+	* btr/btr0btr.c, page/page0cur.c, page/page0page.c:
+	Fix Bug#14554000 CRASH IN PAGE_REC_GET_NTH_CONST(NTH=0)
+	DURING COMPRESSED PAGE SPLIT
+
+2012-08-16	The InnoDB Team
+
+	* btr/btr0cur.c:
+	Fix Bug#12595091 POSSIBLY INVALID ASSERTION IN
+	BTR_CUR_PESSIMISTIC_UPDATE()
+
+2012-08-16	The InnoDB Team
+
+	* btr/btr0btr.c, btr/btr0cur.c:
+	Fix Bug#12845774 OPTIMISTIC INSERT/UPDATE USES WRONG HEURISTICS FOR
+	COMPRESSED PAGE SIZE
+
+2012-08-16	The InnoDB Team
+
+	* btr/btr0cur.c, page/page0page.c:
+	Fix Bug#13523839 ASSERTION FAILURES ON COMPRESSED INNODB TABLES
+
+2012-08-07	The InnoDB Team
+
+	* btr/btr0pcur.c, row/row0merge.c:
+	Fix Bug#14399148 INNODB TABLES UNDER LOAD PRODUCE DUPLICATE COPIES
+	OF ROWS IN QUERIES
+
+2012-03-15	The InnoDB Team
+
+	* fil/fil0fil.c, ibuf/ibuf0ibuf.c, include/fil0fil.h,
+	lock/lock0lock.c:
+	Fix Bug#13825266 RACE IN LOCK_VALIDATE() WHEN ACCESSING PAGES
+	DIRECTLY FROM BUFFER POOL
+
+2012-03-15	The InnoDB Team
+
+	* handler/ha_innodb.cc:
+	Fix Bug#13851171STRING OVERFLOW IN INNODB CODE FOUND BY STATIC
+	ANALYSIS
+
+2012-03-15	The InnoDB Team
+
+	* include/sync0rw.ic:
+	Fix Bug#13537504 VALGRIND: COND. JUMP/MOVE DEPENDS ON
+	UNINITIALISED VALUES IN OS_THREAD_EQ
+
+2012-03-08	The InnoDB Team
+
+	* btr/btr0pcur.c:
+	Fix Bug#13807811 BTR_PCUR_RESTORE_POSITION() CAN SKIP A RECORD
+
+2012-02-28	The InnoDB Team
+
+	* btr/btr0btr.c, dict/dict0dict.c, include/btr0btr.h,
+	include/dict0dict.h, include/dict0dict.ic, include/dict0mem.h,
+	handler/handler0alter.cc, row/row0mysql.c:
+	Fix Bug#12861864 RACE CONDITION IN BTR_GET_SIZE() AND
+	DROP INDEX/TABLE/DATABASE
+
+2012-02-15	The InnoDB Team
+
+	* btr/btr0btr.c, btr/btr0cur.c, fsp/fsp0fsp.c, ibuf/ibuf0ibuf.c,
+	include/btr0btr.h, include/btr0cur.h, include/btr0cur.ic,
+	include/buf0buf.h, include/buf0buf.ic, include/fsp0fsp.h,
+	include/mtr0mtr.h, include/mtr0mtr.ic, include/page0page.h,
+	include/page0page.ic, include/trx0rec.ic, include/trx0undo.h,
+	mtr/mtr0mtr.c, page/page0cur.c, page/page0page.c, row/row0ins.c,
+	row/row0row.c, row/row0upd.c, trx/trx0rec.c, trx/trx0sys.c,
+	trx/trx0undo.c:
+	Fix Bug#13721257 RACE CONDITION IN UPDATES OR INSERTS OF WIDE RECORDS
+
+2012-02-06	The InnoDB Team
+
+	* handler/ha_innodb.cc:
+	Fix Bug#11754376 45976: INNODB LOST FILES FOR TEMPORARY TABLES ON
+	GRACEFUL SHUTDOWN
+
+2012-01-30	The InnoDB Team
+
+	* fil/fil0fil.c:
+	Fix Bug#13636122 THE ORIGINAL TABLE MISSING WHILE EXECUTE THE
+	DDL 'ALTER TABLE ADD COLUMN'
+
+2012-01-16	The InnoDB Team
+
+	* ibuf/ibuf0ibuf.c:
+	Fix Bug#13496818 ASSERTION: REC_PAGE_NO > 4 IN IBUF CONTRACTION
+
+2012-01-16	The InnoDB Team
+
+	* handler/ha_innodb.cc:
+	Fix Bug#11765438: 58406: ISSUES WITH COPYING PARTITIONED INNODB
+	TABLES FROM LINUX TO WINDOWS
+
+2012-01-04	The InnoDB Team
+
+	* row/row0mysql.c:
+	Fix Bug#12400341: INNODB CAN LEAVE ORPHAN IBD FILES AROUND
+
+2011-12-22	The InnoDB Team
+
+	* row/row0sel.c:
+	Fix Bug#63775 Server crash on handler read next after delete record.
+2011-12-21	The InnoDB Team
+
+	* include/ut0rnd.ic:
+	Fix Bug#11866367:FPE WHEN SETTING INNODB_SPIN_WAIT_DELAY
+
+2011-12-13	The InnoDB Team
+
+	* handler/ha_innodb.cc, innodb.test, innodb.result:
+	Fix Bug#13117023: InnoDB was incrementing the handler_read_key,
+	also the SSV::ha_read_key_count, at the wrong place.
+
+2011-12-10	The InnoDB Team
+
+	* include/page0page.h, page/page0page.c:
+	Fix Bug#13418887 ERROR IN DIAGNOSTIC FUNCTION PAGE_REC_PRINT()
+
+2011-11-10	The InnoDB Team
+
+	* handler/ha_innodb.cc, row/row0ins.c, innodb_replace.test:
+	Fix Bug#11759688 52020: InnoDB can still deadlock
+	on just INSERT...ON DUPLICATE KEY a.k.a. the reintroduction of
+	Bug#7975 deadlock without any locking, simple select and update
+
+2011-11-08	The InnoDB Team
+
+	* btr/btr0pcur.c, include/btr0pcur.h, include/btr0pcur.ic:
+	Fix Bug#13358468 ASSERTION FAILURE IN BTR_PCUR_GET_BLOCK
+
+2011-10-27	The InnoDB Team
+
+	* row/row0mysql.c:
+	Fix Bug #12884631 62146: TABLES ARE LOST FOR DDL
+
+2011-10-25	The InnoDB Team
+
+	* handler/ha_innodb.cc, row/row0ins.c:
+	Fix Bug#13002783 PARTIALLY UNINITIALIZED CASCADE UPDATE VECTOR
+
+2011-10-20	The InnoDB Team
+
+	* btr/btr0cur.c:
+	Fix Bug#13116045 Compilation failure using GCC 4.6.1 in btr/btr0cur.c
+
+2011-10-12	The InnoDB Team
+
+	* btr/btr0cur.c, btr/btr0sea.c, buf/buf0buf.c, buf/buf0lru.c,
+	ha/ha0ha.c, handler/ha_innodb.cc, ibuf/ibuf0ibuf.c, include/btr0sea.h,
+	include/btr0types.h, include/buf0buf.h, include/ha0ha.h,
+	include/ha0ha.ic, include/row0upd.ic, include/sync0sync.h,
+	page/page0page.c, sync/sync0sync.c:
+	Fix Bug#13006367 62487: innodb takes 3 minutes to clean up
+	the adaptive hash index at shutdown
+
+2011-10-04	The InnoDB Team
+
+	* include/sync0rw.h, sync/sync0rw.c:
+	Fix Bug#13034534 RQG TESTS FAIL ON WINDOWS WITH CRASH NEAR
+	RW_LOCK_DEBUG_PRINT
+
+2011-09-20	The InnoDB Team
+
+	* row/row0purge.c:
+	Fix Bug#12963823 CRASH IN PURGE THREAD UNDER UNUSUAL CIRCUMSTANCES
+
+2011-09-12	The InnoDB Team
+
+	* row/row0sel.c:
+	Fix Bug#12601439 CONSISTENT READ FAILURE IN COLUMN PREFIX INDEX
+
+2011-09-08	The InnoDB Team
+
+	* btr/btr0cur.c, include/page0page.h, include/row0upd.ic:
+	Fix Bug#12948130 UNNECESSARY X-LOCKING OF ADAPTIVE HASH INDEX
+
+2011-09-06	The InnoDB Team
+
+	* buf/buf0buddy.c:
+	Fix Bug#12950803 62294: BUF_BUDDY_RELOCATE CALLS GETTIMEOFDAY
+	WHILE HOLDING BUFFER POOL MUTEX
+
+2011-09-06	The InnoDB Team
+
+	* include/trx0undo.h, trx/trx0rec.c, trx/trx0undo.c:
+	Fix Bug#12547647 UPDATE LOGGING COULD EXCEED LOG PAGE SIZE
+
+2011-08-29	The InnoDB Team
+
+	* btr/btr0btr.c, btr/btr0cur.c, fsp/fsp0fsp.c,
+	include/btr0btr.h, include/btr0cur.h, include/fsp0fsp.h,
+	include/mtr0mtr.h, include/mtr0mtr.ic, mtr/mtr0mtr.c,
+	row/row0ins.c, row/row0row.c, row/row0upd.c, trx/trx0undo.c:
+	Fix Bug#12704861 Corruption after a crash during BLOB update
+	and other regressions from the fix of Bug#12612184
+
+2011-08-15	The InnoDB Team
+
+	* btr/btr0btr.c, btr/btr0cur.c, btr/btr0pcur.c, btr/btr0sea.c,
+	dict/dict0crea.c, dict/dict0dict.c, ibuf/ibuf0ibuf.c,
+	include/btr0btr.h, include/btr0btr.ic, include/sync0sync.h,
+	sync/sync0sync.c:
+	Fix Bug#11766591 59733: Possible deadlock when buffered changes
+	are to be discarded in buf_page_create()
+
 2011-08-08	The InnoDB Team
 
 	* row/row0sel.c:

Makefile.am

@@ -101,6 +101,7 @@ noinst_HEADERS=		\
 			include/lock0types.h	\
 			include/log0log.h	\
 			include/log0log.ic	\
+			include/log0online.h	\
 			include/log0recv.h	\
 			include/log0recv.ic	\
 			include/mach0data.h	\
@@ -226,7 +227,6 @@ noinst_HEADERS=		\
 			include/ut0vec.h	\
 			include/ut0vec.ic	\
 			include/ut0wqueue.h	\
-			handler/innodb_patch_info.h	\
 			mem/mem0dbg.c
 
 EXTRA_LIBRARIES=	libinnobase.a
@@ -266,6 +266,7 @@ libinnobase_a_SOURCES=	\
 			lock/lock0iter.c		\
 			lock/lock0lock.c		\
 			log/log0log.c			\
+			log/log0online.c		\
 			log/log0recv.c			\
 			mach/mach0data.c		\
 			mem/mem0mem.c			\

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
@@ -127,6 +127,8 @@ 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;
 
@@ -253,6 +255,8 @@ btr_pcur_restore_position_func(
 			cursor->rel_pos == BTR_PCUR_BEFORE_FIRST_IN_TREE,
 			index, latch_mode, btr_pcur_get_btr_cur(cursor), mtr);
 
+		cursor->latch_mode = latch_mode;
+		cursor->pos_state = BTR_PCUR_IS_POSITIONED;
 		cursor->block_when_stored = btr_pcur_get_block(cursor);
 
 		return(FALSE);
@@ -272,8 +276,10 @@ btr_pcur_restore_position_func(
 					file, line, mtr))) {
 			cursor->pos_state = BTR_PCUR_IS_POSITIONED;
 
-			buf_block_dbg_add_level(btr_pcur_get_block(cursor),
-						SYNC_TREE_NODE);
+			buf_block_dbg_add_level(
+				btr_pcur_get_block(cursor),
+				dict_index_is_ibuf(index)
+				? SYNC_IBUF_TREE_NODE : SYNC_TREE_NODE);
 
 			if (cursor->rel_pos == BTR_PCUR_ON) {
 #ifdef UNIV_DEBUG
@@ -315,13 +321,19 @@ 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,
@@ -330,19 +342,24 @@ 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 */
-
-		cursor->block_when_stored = btr_pcur_get_block(cursor);
-		cursor->modify_clock = buf_block_get_modify_clock(
+	switch (cursor->rel_pos) {
+	case BTR_PCUR_ON:
+		if (btr_pcur_is_on_user_rec(cursor)
+		    && !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;
 
@@ -350,6 +367,15 @@ btr_pcur_restore_position_func(
 
 			return(TRUE);
 		}
+#ifdef UNIV_DEBUG
+		/* fall through */
+	case BTR_PCUR_BEFORE:
+	case BTR_PCUR_AFTER:
+		break;
+	default:
+		ut_error;
+#endif /* UNIV_DEBUG */
+	}
 
 	mem_heap_free(heap);
 
@@ -396,7 +422,8 @@ btr_pcur_move_to_next_page(
 	ut_ad(next_page_no != FIL_NULL);
 
 	next_block = btr_block_get(space, zip_size, next_page_no,
-				   cursor->latch_mode, mtr);
+				   cursor->latch_mode,
+				   btr_pcur_get_btr_cur(cursor)->index, mtr);
 	next_page = buf_block_get_frame(next_block);
 
 	if (srv_pass_corrupt_table && !next_page) {

buf/buf0buddy.c

@@ -354,7 +354,6 @@ buf_buddy_relocate(
 {
 	buf_page_t*	bpage;
 	const ulint	size	= BUF_BUDDY_LOW << i;
-	ullint		usec	= ut_time_us(NULL);
 	mutex_t*	mutex;
 	ulint		space;
 	ulint		page_no;
@@ -442,6 +441,7 @@ buf_buddy_relocate(
 
 	if (mutex && buf_page_can_relocate(bpage)) {
 		/* Relocate the compressed page. */
+		ullint	usec	= ut_time_us(NULL);
 		ut_a(bpage->zip.data == src);
 		memcpy(dst, src, size);
 		bpage->zip.data = dst;

buf/buf0lru.c

@@ -48,6 +48,7 @@ Created 11/5/1995 Heikki Tuuri
 #include "page0zip.h"
 #include "log0recv.h"
 #include "srv0srv.h"
+#include "srv0start.h"
 
 /** The number of blocks from the LRU_old pointer onward, including
 the block pointed to, must be buf_LRU_old_ratio/BUF_LRU_OLD_RATIO_DIV
@@ -292,7 +293,7 @@ next_page:
 
 		//mutex_enter(&((buf_block_t*) bpage)->mutex);
 		is_fixed = bpage->buf_fix_count > 0
-			|| !((buf_block_t*) bpage)->is_hashed;
+			|| !((buf_block_t*) bpage)->index;
 		//mutex_exit(&((buf_block_t*) bpage)->mutex);
 
 		if (is_fixed) {
@@ -451,7 +452,7 @@ scan_again:
 		if (buf_page_get_state(bpage) != BUF_BLOCK_FILE_PAGE) {
 			/* This is a compressed-only block
 			descriptor. Do nothing. */
-		} else if (((buf_block_t*) bpage)->is_hashed) {
+		} else if (((buf_block_t*) bpage)->index) {
 			ulint	page_no;
 			ulint	zip_size;
 
@@ -465,7 +466,7 @@ scan_again:
 			mutex_exit(block_mutex);
 
 			/* Note that the following call will acquire
-			an S-latch on the page */
+			and release an X-latch on the page. */
 
 			btr_search_drop_page_hash_when_freed(
 				id, zip_size, page_no);
@@ -537,7 +538,7 @@ buf_LRU_mark_space_was_deleted(
 		for (j = chunk->size; j--; block++) {
 			if (buf_block_get_state(block)
 			    != BUF_BLOCK_FILE_PAGE
-			    || !block->is_hashed
+			    || !(block->index != NULL)
 			    || buf_page_get_space(&block->page) != id) {
 				continue;
 			}
@@ -1428,13 +1429,12 @@ buf_LRU_make_block_old(
 Try to free a block.  If bpage is a descriptor of a compressed-only
 page, the descriptor object will be freed as well.
 
-NOTE: If this function returns TRUE, it will temporarily
-release buf_pool_mutex.  Furthermore, the page frame will no longer be
-accessible via bpage.
+NOTE: This will temporarily release buf_pool_mutex.  Furthermore, the
+page frame will no longer be accessible via bpage.
 
-The caller must hold buf_pool_mutex and buf_page_get_mutex(bpage) and
-release these two mutexes after the call.  No other
-buf_page_get_mutex() may be held when calling this function.
+The caller must hold buf_page_get_mutex(bpage) and release this mutex
+after the call.  No other buf_page_get_mutex() may be held when
+calling this function.
 @return TRUE if freed, FALSE otherwise. */
 UNIV_INTERN
 ibool
@@ -2098,6 +2098,12 @@ func_exit:
 /********************************************************************//**
 Dump the LRU page list to the specific file. */
 #define LRU_DUMP_FILE "ib_lru_dump"
+#define LRU_DUMP_TEMP_FILE "ib_lru_dump.tmp"
+#define LRU_OS_FILE_WRITE() \
+	os_file_write(LRU_DUMP_FILE, dump_file, buffer, \
+		(buffers << UNIV_PAGE_SIZE_SHIFT) & 0xFFFFFFFFUL, \
+		(buffers >> (32 - UNIV_PAGE_SIZE_SHIFT)), \
+		UNIV_PAGE_SIZE)
 
 UNIV_INTERN
 ibool
@@ -2109,17 +2115,19 @@ buf_LRU_file_dump(void)
 	byte*		buffer_base = NULL;
 	byte*		buffer = NULL;
 	buf_page_t*	bpage;
+	buf_page_t*	first_bpage;
 	ulint		buffers;
 	ulint		offset;
-	ibool		ret = FALSE;
+	ulint		pages_written;
 	ulint		i;
+	ulint		total_pages;
 
 	for (i = 0; i < srv_n_data_files; i++) {
 		if (strstr(srv_data_file_names[i], LRU_DUMP_FILE) != NULL) {
 			fprintf(stderr,
 				" InnoDB: The name '%s' seems to be used for"
-				" innodb_data_file_path. Dumping LRU list is not"
-				" done for safeness.\n", LRU_DUMP_FILE);
+				" innodb_data_file_path. Dumping LRU list is"
+				"  not done for safeness.\n", LRU_DUMP_FILE);
 			goto end;
 		}
 	}
@@ -2132,7 +2140,7 @@ buf_LRU_file_dump(void)
 		goto end;
 	}
 
-	dump_file = os_file_create(LRU_DUMP_FILE, OS_FILE_OVERWRITE,
+	dump_file = os_file_create(LRU_DUMP_TEMP_FILE, OS_FILE_OVERWRITE,
 				OS_FILE_NORMAL, OS_DATA_FILE, &success);
 	if (!success) {
 		os_file_get_last_error(TRUE);
@@ -2142,12 +2150,21 @@ buf_LRU_file_dump(void)
 	}
 
 	mutex_enter(&LRU_list_mutex);
-	bpage = UT_LIST_GET_LAST(buf_pool->LRU);
+	bpage = first_bpage = UT_LIST_GET_FIRST(buf_pool->LRU);
+	total_pages = UT_LIST_GET_LEN(buf_pool->LRU);
 
-	buffers = offset = 0;
-	while (bpage != NULL) {
-		if (offset == 0) {
-			memset(buffer, 0, UNIV_PAGE_SIZE);
+	buffers = offset = pages_written = 0;
+	while (bpage != NULL && (pages_written++ < total_pages)) {
+
+		buf_page_t*	next_bpage = UT_LIST_GET_NEXT(LRU, bpage);
+
+		if (next_bpage == first_bpage) {
+			mutex_exit(&LRU_list_mutex);
+			success = FALSE;
+			fprintf(stderr,
+				"InnoDB: detected cycle in LRU, skipping "
+				"dump\n");
+			goto end;
 		}
 
 		mach_write_to_4(buffer + offset * 4, bpage->space);
@@ -2156,50 +2173,79 @@ buf_LRU_file_dump(void)
 		offset++;
 
 		if (offset == UNIV_PAGE_SIZE/4) {
-			success = os_file_write(LRU_DUMP_FILE, dump_file, buffer,
-					(buffers << UNIV_PAGE_SIZE_SHIFT) & 0xFFFFFFFFUL,
-					(buffers >> (32 - UNIV_PAGE_SIZE_SHIFT)),
-					UNIV_PAGE_SIZE);
+			mutex_t		*next_block_mutex = NULL;
+
+			if (srv_shutdown_state != SRV_SHUTDOWN_NONE) {
+				mutex_exit(&LRU_list_mutex);
+				success = FALSE;
+				fprintf(stderr,
+					" InnoDB: stopped dumping lru pages"
+					" because of server shutdown.\n");
+				goto end;
+			}
+
+			/* while writing file, release buffer pool mutex but
+			   keep the next page fixed so we don't worry about
+			   our list iterator becoming invalid */
+			if (next_bpage) {
+				next_block_mutex = buf_page_get_mutex(
+							next_bpage);
+
+				mutex_enter(next_block_mutex);
+				next_bpage->buf_fix_count++;
+				mutex_exit(next_block_mutex);
+			}
+			mutex_exit(&LRU_list_mutex);
+
+			success = LRU_OS_FILE_WRITE();
+
+			/* grab this again here so that next_bpage
+			   can't be purged when we drop the fix_count */
+			mutex_enter(&LRU_list_mutex);
+
+			if (next_bpage) {
+				mutex_enter(next_block_mutex);
+				next_bpage->buf_fix_count--;
+				mutex_exit(next_block_mutex);
+			}
 			if (!success) {
 				mutex_exit(&LRU_list_mutex);
 				fprintf(stderr,
-					" InnoDB: cannot write page %lu of %s\n",
+					" InnoDB: cannot write page"
+					" %lu of %s\n",
 					buffers, LRU_DUMP_FILE);
 				goto end;
 			}
 			buffers++;
 			offset = 0;
+			bpage = next_bpage;
+		} else {
+			bpage = UT_LIST_GET_NEXT(LRU, bpage);
 		}
-
-		bpage = UT_LIST_GET_PREV(LRU, bpage);
-	}
+	} /* while(bpage ...) */
 	mutex_exit(&LRU_list_mutex);
 
-	if (offset == 0) {
-		memset(buffer, 0, UNIV_PAGE_SIZE);
-	}
-
 	mach_write_to_4(buffer + offset * 4, 0xFFFFFFFFUL);
 	offset++;
 	mach_write_to_4(buffer + offset * 4, 0xFFFFFFFFUL);
 	offset++;
 
-	success = os_file_write(LRU_DUMP_FILE, dump_file, buffer,
-			(buffers << UNIV_PAGE_SIZE_SHIFT) & 0xFFFFFFFFUL,
-			(buffers >> (32 - UNIV_PAGE_SIZE_SHIFT)),
-			UNIV_PAGE_SIZE);
-	if (!success) {
-		goto end;
-	}
-
-	ret = TRUE;
+	success = LRU_OS_FILE_WRITE();
 end:
-	if (dump_file != -1)
+	if (dump_file != -1) {
+		if (success) {
+			success = os_file_flush(dump_file, TRUE);
+		}
 		os_file_close(dump_file);
+	}
+	if (success) {
+		success = os_file_rename(LRU_DUMP_TEMP_FILE,
+			LRU_DUMP_FILE);
+	}
 	if (buffer_base)
 		ut_free(buffer_base);
 
-	return(ret);
+	return(success);
 }
 
 typedef struct {
@@ -2241,6 +2287,7 @@ buf_LRU_file_restore(void)
 	dump_record_t*	records = NULL;
 	ulint		size;
 	ulint		size_high;
+	ulint 		recsize = sizeof(dump_record_t);
 	ulint		length;
 
 	dump_file = os_file_create_simple_no_error_handling(
@@ -2248,7 +2295,15 @@ buf_LRU_file_restore(void)
 	if (!success || !os_file_get_size(dump_file, &size, &size_high)) {
 		os_file_get_last_error(TRUE);
 		fprintf(stderr,
-			" InnoDB: cannot open %s\n", LRU_DUMP_FILE);
+			" InnoDB: cannot open %s,"
+			" buffer pool preload not done\n",
+			LRU_DUMP_FILE);
+		goto end;
+	}
+
+	if (size == 0 || size_high > 0 || size % recsize) {
+		fprintf(stderr, " InnoDB: broken LRU dump file,"
+			" buffer pool preload not done\n");
 		goto end;
 	}
 
@@ -2332,6 +2387,14 @@ buf_LRU_file_restore(void)
 		if (offset % 16 == 15) {
 			os_aio_simulated_wake_handler_threads();
 			buf_flush_free_margin(FALSE);
+			/* skip loading of the rest of the file if we are
+ 			   terminating anyway*/
+			if (srv_shutdown_state != SRV_SHUTDOWN_NONE) {
+				fprintf(stderr,
+					" InnoDB: stopped loading LRU pages"
+					" because of server shutdown.\n");
+				break;
+			}
 		}
 
 		zip_size = fil_space_get_zip_size(space_id);

dict/dict0boot.c

@@ -236,6 +236,166 @@ 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_dulint(*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. */
@@ -251,39 +411,23 @@ 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 */
 	dict_hdr = dict_hdr_get(&mtr);
 
-	if (ut_dulint_cmp(mtr_read_dulint(dict_hdr + DICT_HDR_XTRADB_MARK, &mtr),
-			  DICT_HDR_XTRADB_FLAG) != 0) {
-		/* not extended yet by XtraDB, need to be extended */
-		ulint	root_page_no;
+	if (ut_dulint_cmp(mtr_read_dulint(dict_hdr + DICT_HDR_XTRADB_MARK,
+					  &mtr), DICT_HDR_XTRADB_FLAG) != 0) {
 
-		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_dulint(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);
+		/* not extended yet by XtraDB, need to be extended */
+		dict_create_xtradb_sys_stats(&dict_hdr, &mtr);
 	}
 
 	/* Because we only write new row ids to disk-based data structure
@@ -464,42 +608,7 @@ 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);
 

dict/dict0crea.c

 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, 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
@@ -899,7 +899,7 @@ dict_truncate_index_tree(
 	appropriate field in the SYS_INDEXES record: this mini-transaction
 	marks the B-tree totally truncated */
 
-	btr_block_get(space, zip_size, root_page_no, RW_X_LATCH, mtr);
+	btr_block_get(space, zip_size, root_page_no, RW_X_LATCH, NULL, mtr);
 
 	btr_free_root(space, zip_size, root_page_no, mtr);
 create:

dict/dict0dict.c

 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, 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
@@ -24,6 +24,8 @@ Created 1/8/1996 Heikki Tuuri
 ***********************************************************************/
 
 #include "dict0dict.h"
+#include "m_string.h"
+#include "my_sys.h"
 
 #ifdef UNIV_NONINL
 #include "dict0dict.ic"
@@ -270,15 +272,39 @@ dict_table_stats_lock(
 	ulint			latch_mode)	/*!< in: RW_S_LATCH or
 						RW_X_LATCH */
 {
+	rw_lock_t *want, *got;
+
 	ut_ad(table != NULL);
 	ut_ad(table->magic_n == DICT_TABLE_MAGIC_N);
 
 	switch (latch_mode) {
 	case RW_S_LATCH:
-		rw_lock_s_lock(GET_TABLE_STATS_LATCH(table));
+		/* Lock one of dict_table_stats_latches in S-mode.
+		Latch is picked using table->id. table->id might be
+		changed while we are waiting for lock to be grabbed */
+		for (;;) {
+			want= GET_TABLE_STATS_LATCH(table);
+			rw_lock_s_lock(want);
+			got= GET_TABLE_STATS_LATCH(table);
+			if (want == got) {
+				break;
+			}
+			rw_lock_s_unlock(want);
+		}
 		break;
 	case RW_X_LATCH:
-		rw_lock_x_lock(GET_TABLE_STATS_LATCH(table));
+		/* Lock one of dict_table_stats_latches in X-mode.
+		Latch is picked using table->id. table->id might be
+		changed while we are waiting for lock to be grabbed */
+		for (;;) {
+			want= GET_TABLE_STATS_LATCH(table);
+			rw_lock_x_lock(want);
+			got= GET_TABLE_STATS_LATCH(table);
+			if (want == got) {
+				break;
+			}
+			rw_lock_x_unlock(want);
+		}
 		break;
 	case RW_NO_LATCH:
 		/* fall through */
@@ -1164,12 +1190,21 @@ dict_table_change_id_in_cache(
 	dict_table_t*	table,	/*!< in/out: table object already in cache */
 	dulint		new_id)	/*!< in: new id to set */
 {
+	dict_table_t table_tmp;
+
 	ut_ad(table);
 	ut_ad(mutex_own(&(dict_sys->mutex)));
 	ut_ad(table->magic_n == DICT_TABLE_MAGIC_N);
 
 	/* Remove the table from the hash table of id's */
 
+	/* Lock is needed to prevent dict_table_stats_latches from
+	being leaked. dict_table_stats_lock picks one latch using
+	table->id. We are changing table->id below. That is why
+	we also should remember the old value to unlock table */
+	dict_table_stats_lock(table, RW_X_LATCH);
+	table_tmp= *table;
+
 	HASH_DELETE(dict_table_t, id_hash, dict_sys->table_id_hash,
 		    ut_fold_dulint(table->id), table);
 	table->id = new_id;
@@ -1177,6 +1212,8 @@ dict_table_change_id_in_cache(
 	/* Add the table back to the hash table */
 	HASH_INSERT(dict_table_t, id_hash, dict_sys->table_id_hash,
 		    ut_fold_dulint(table->id), table);
+
+	dict_table_stats_unlock(&table_tmp, RW_X_LATCH);
 }
 
 /**********************************************************************//**
@@ -1724,7 +1761,9 @@ undo_size_ok:
 	new_index->stat_n_leaf_pages = 1;
 
 	new_index->page = page_no;
-	rw_lock_create(&new_index->lock, SYNC_INDEX_TREE);
+	rw_lock_create(&new_index->lock,
+		       dict_index_is_ibuf(index)
+		       ? SYNC_IBUF_INDEX_TREE : SYNC_INDEX_TREE);
 
 	if (!UNIV_UNLIKELY(new_index->type & DICT_UNIVERSAL)) {
 
@@ -2338,6 +2377,8 @@ 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);
 }
 
@@ -4306,19 +4347,26 @@ 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);
+
+		mtr_commit(&mtr);
 
 		if (size == 0) {
 			/* The root node of the tree is a leaf */
@@ -4574,12 +4622,6 @@ next_rec:
 	}
 	btr_pcur_close(&pcur);
 	mtr_commit(&mtr);
-
-	if (rests) {
-		fprintf(stderr, "InnoDB: Warning: failed to store %lu stats entries"
-				" of %s/%s to SYS_STATS system table.\n",
-				rests, index->table_name, index->name);
-	}
 }
 /*===========================================*/
 
@@ -4668,16 +4710,27 @@ 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;
-			size = btr_get_size(index, BTR_TOTAL_SIZE);
 
-			index->stat_index_size = size;
+			mtr_start(&mtr);
+			mtr_s_lock(dict_index_get_lock(index), &mtr);
 
+			size = btr_get_size(index, BTR_TOTAL_SIZE, &mtr);
+
+			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;
 			}
@@ -4694,6 +4747,7 @@ 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;
 
@@ -5334,6 +5388,28 @@ dict_table_replace_index_in_foreign_list(
 			foreign->foreign_index = new_index;
 		}
 	}
+
+
+	for (foreign = UT_LIST_GET_FIRST(table->referenced_list);
+	     foreign;
+	     foreign = UT_LIST_GET_NEXT(referenced_list, foreign)) {
+
+		dict_index_t*	new_index;
+
+		if (foreign->referenced_index == index) {
+			ut_ad(foreign->referenced_table == index->table);
+
+			new_index = dict_foreign_find_index(
+				foreign->referenced_table,
+				foreign->referenced_col_names,
+				foreign->n_fields, index,
+				/*check_charsets=*/TRUE, /*check_null=*/FALSE);
+			ut_ad(new_index || !trx->check_foreigns);
+			ut_ad(!new_index || new_index->table == index->table);
+
+			foreign->referenced_index = new_index;
+		}
+	}
 }
 
 /**********************************************************************//**

dict/dict0load.c

@@ -165,7 +165,7 @@ dict_print(void)
 	monitor printout */
 
 	mutex_enter(&kernel_mutex);
-	srv_fatal_semaphore_wait_threshold += 7200; /* 2 hours */
+	srv_fatal_semaphore_wait_threshold += SRV_SEMAPHORE_WAIT_EXTENSION;
 	mutex_exit(&kernel_mutex);
 
 	mutex_enter(&(dict_sys->mutex));
@@ -193,7 +193,7 @@ loop:
 		/* Restore the fatal semaphore wait timeout */
 
 		mutex_enter(&kernel_mutex);
-		srv_fatal_semaphore_wait_threshold -= 7200; /* 2 hours */
+		srv_fatal_semaphore_wait_threshold -= SRV_SEMAPHORE_WAIT_EXTENSION;
 		mutex_exit(&kernel_mutex);
 
 		return;
@@ -430,7 +430,7 @@ loop:
 			object and check that the .ibd file exists. */
 
 			fil_open_single_table_tablespace(FALSE, space_id,
-							 flags, name);
+							 flags, name, NULL);
 		}
 
 		mem_free(name);
@@ -1023,7 +1023,7 @@ err_exit:
 			if (!fil_open_single_table_tablespace(
 				    TRUE, space,
 				    flags == DICT_TF_COMPACT ? 0 :
-				    flags & ~(~0 << DICT_TF_BITS), name)) {
+				    flags & ~(~0 << DICT_TF_BITS), name, NULL)) {
 				/* We failed to find a sensible
 				tablespace file */
 

fil/fil0fil.c

@@ -183,7 +183,7 @@ 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;
@@ -208,12 +208,13 @@ 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
@@ -928,11 +929,6 @@ retry:
 		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) {
@@ -949,6 +945,25 @@ retry:
 
 		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++;
@@ -956,6 +971,11 @@ retry:
 		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 */
@@ -1228,7 +1248,7 @@ try_again:
 	}
 
 	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;
@@ -1237,7 +1257,7 @@ try_again:
 	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;
@@ -1840,13 +1860,12 @@ fil_read_flushed_lsn_and_arch_log_no(
 
 #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;
@@ -1857,18 +1876,18 @@ fil_inc_pending_ibuf_merges(
 
 	if (space == NULL) {
 		fprintf(stderr,
-			"InnoDB: Error: trying to do ibuf merge to a"
+			"InnoDB: Error: trying to do an operation on a"
 			" dropped tablespace %lu\n",
 			(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);
 
@@ -1876,11 +1895,11 @@ 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;
@@ -1891,13 +1910,13 @@ 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);
@@ -2187,15 +2206,15 @@ 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;
@@ -2209,9 +2228,10 @@ stop_ibuf_merges:
 				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);
 			}
 
@@ -2220,7 +2240,7 @@ stop_ibuf_merges:
 			os_thread_sleep(20000);
 			count++;
 
-			goto stop_ibuf_merges;
+			goto stop_new_ops;
 		}
 	}
 
@@ -2246,7 +2266,7 @@ try_again:
 	}
 
 	ut_a(space);
-	ut_a(space->n_pending_ibuf_merges == 0);
+	ut_a(space->n_pending_ops == 0);
 
 	space->is_being_deleted = TRUE;
 
@@ -2523,7 +2543,7 @@ 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);
@@ -3124,8 +3144,11 @@ fil_open_single_table_tablespace(
 					accessing the first page of the file */
 	ulint		id,		/*!< in: space id */
 	ulint		flags,		/*!< in: tablespace flags */
-	const char*	name)		/*!< in: table name in the
+	const char*	name,		/*!< in: table name in the
 					databasename/tablename format */
+	trx_t*		trx)		/*!< in: transaction. This is only used
+					for IMPORT TABLESPACE, must be NULL
+					otherwise */
 {
 	os_file_t	file;
 	char*		filepath;
@@ -3342,11 +3365,17 @@ skip_info:
 			/* over write space id of all pages */
 			rec_offs_init(offsets_);
 
+			/* Unlock the data dictionary to not block queries
+			accessing other tables */
+			ut_a(trx);
+			row_mysql_unlock_data_dictionary(trx);
+
 			fprintf(stderr, "InnoDB: Progress in %%:");
 
 			for (offset = 0; offset < free_limit_bytes;
 			     offset += zip_size ? zip_size : UNIV_PAGE_SIZE) {
 				ibool		page_is_corrupt;
+				ibool		is_descr_page = FALSE;
 
 				success = os_file_read(file, page,
 							(ulint)(offset & 0xFFFFFFFFUL),
@@ -3385,6 +3414,7 @@ skip_info:
 
 					/* store as descr page */
 					memcpy(descr_page, page, (zip_size ? zip_size : UNIV_PAGE_SIZE));
+					is_descr_page = TRUE;
 
 				} else if (descr_is_corrupt) {
 					/* unknown state of the page */
@@ -3461,7 +3491,8 @@ skip_info:
 						}
 					}
 
-					if (fil_page_get_type(page) == FIL_PAGE_INDEX) {
+					if (fil_page_get_type(page) ==
+					    FIL_PAGE_INDEX && !is_descr_page) {
 						dulint tmp = mach_read_from_8(page + (PAGE_HEADER + PAGE_INDEX_ID));
 
 						for (i = 0; i < n_index; i++) {
@@ -3543,6 +3574,9 @@ skip_write:
 
 			fprintf(stderr, " done.\n");
 
+			/* Reacquire the data dictionary lock */
+			row_mysql_lock_data_dictionary(trx);
+
 			/* update SYS_INDEXES set root page */
 			index = dict_table_get_first_index(table);
 			while (index) {
@@ -3678,7 +3712,6 @@ func_exit:
 		ulint		page_no;
 		ulint		zip_size;
 		ulint		height;
-		ulint		root_height = 0;
 		rec_t*		node_ptr;
 		dict_table_t*	table;
 		dict_index_t*	index;
@@ -3717,7 +3750,6 @@ func_exit:
 
 			if (height == ULINT_UNDEFINED) {
 				height = btr_page_get_level(page, &mtr);
-				root_height = height;
 			}
 
 			if (height == 0) {
@@ -3837,7 +3869,7 @@ convert_err_exit:
 
 					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);
@@ -3885,7 +3917,7 @@ convert_err_exit:
 				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,

ha/ha0ha.c

 /*****************************************************************************
 
-Copyright (c) 1994, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1994, 2011, 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
@@ -88,40 +88,6 @@ ha_create_func(
 	return(table);
 }
 
-/*************************************************************//**
-Empties a hash table and frees the memory heaps. */
-UNIV_INTERN
-void
-ha_clear(
-/*=====*/
-	hash_table_t*	table)	/*!< in, own: hash table */
-{
-	ulint	i;
-	ulint	n;
-
-	ut_ad(table);
-	ut_ad(table->magic_n == HASH_TABLE_MAGIC_N);
-#ifdef UNIV_SYNC_DEBUG
-	ut_ad(rw_lock_own(&btr_search_latch, RW_LOCK_EXCLUSIVE));
-#endif /* UNIV_SYNC_DEBUG */
-
-#ifndef UNIV_HOTBACKUP
-	/* Free the memory heaps. */
-	n = table->n_mutexes;
-
-	for (i = 0; i < n; i++) {
-		mem_heap_free(table->heaps[i]);
-	}
-#endif /* !UNIV_HOTBACKUP */
-
-	/* Clear the hash table. */
-	n = hash_get_n_cells(table);
-
-	for (i = 0; i < n; i++) {
-		hash_get_nth_cell(table, i)->node = NULL;
-	}
-}
-
 /*************************************************************//**
 Inserts an entry into a hash table. If an entry with the same fold number
 is found, its node is updated to point to the new data, and no new node
@@ -140,7 +106,7 @@ ha_insert_for_fold_func(
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 	buf_block_t*	block,	/*!< in: buffer block containing the data */
 #endif /* UNIV_AHI_DEBUG || UNIV_DEBUG */
-	void*		data)	/*!< in: data, must not be NULL */
+	const rec_t*	data)	/*!< in: data, must not be NULL */
 {
 	hash_cell_t*	cell;
 	ha_node_t*	node;
@@ -153,7 +119,11 @@ ha_insert_for_fold_func(
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 	ut_a(block->frame == page_align(data));
 #endif /* UNIV_AHI_DEBUG || UNIV_DEBUG */
+#ifdef UNIV_SYNC_DEBUG
+	ut_ad(rw_lock_own(&btr_search_latch, RW_LOCK_EX));
+#endif /* UNIV_SYNC_DEBUG */
 	ASSERT_HASH_MUTEX_OWN(table, fold);
+	ut_ad(btr_search_enabled);
 
 	hash = hash_calc_hash(fold, table);
 
@@ -173,7 +143,6 @@ ha_insert_for_fold_func(
 				prev_block->n_pointers--;
 				block->n_pointers++;
 			}
-			ut_ad(!btr_search_fully_disabled);
 # endif /* !UNIV_HOTBACKUP */
 
 			prev_node->block = block;
@@ -186,13 +155,6 @@ ha_insert_for_fold_func(
 		prev_node = prev_node->next;
 	}
 
-	/* We are in the process of disabling hash index, do not add
-	new chain node */
-	if (!btr_search_enabled) {
-		ut_ad(!btr_search_fully_disabled);
-		return(TRUE);
-	}
-
 	/* We have to allocate a new chain node */
 
 	node = mem_heap_alloc(hash_get_heap(table, fold), sizeof(ha_node_t));
@@ -250,6 +212,10 @@ ha_delete_hash_node(
 {
 	ut_ad(table);
 	ut_ad(table->magic_n == HASH_TABLE_MAGIC_N);
+#ifdef UNIV_SYNC_DEBUG
+	ut_ad(rw_lock_own(&btr_search_latch, RW_LOCK_EX));
+#endif /* UNIV_SYNC_DEBUG */
+	ut_ad(btr_search_enabled);
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 # ifndef UNIV_HOTBACKUP
 	if (table->adaptive) {
@@ -272,11 +238,11 @@ ha_search_and_update_if_found_func(
 /*===============================*/
 	hash_table_t*	table,	/*!< in/out: hash table */
 	ulint		fold,	/*!< in: folded value of the searched data */
-	void*		data,	/*!< in: pointer to the data */
+	const rec_t*	data,	/*!< in: pointer to the data */
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 	buf_block_t*	new_block,/*!< in: block containing new_data */
 #endif /* UNIV_AHI_DEBUG || UNIV_DEBUG */
-	void*		new_data)/*!< in: new pointer to the data */
+	const rec_t*	new_data)/*!< in: new pointer to the data */
 {
 	ha_node_t*	node;
 
@@ -286,6 +252,13 @@ ha_search_and_update_if_found_func(
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 	ut_a(new_block->frame == page_align(new_data));
 #endif /* UNIV_AHI_DEBUG || UNIV_DEBUG */
+#ifdef UNIV_SYNC_DEBUG
+	ut_ad(rw_lock_own(&btr_search_latch, RW_LOCK_EX));
+#endif /* UNIV_SYNC_DEBUG */
+
+	if (!btr_search_enabled) {
+		return;
+	}
 
 	node = ha_search_with_data(table, fold, data);
 
@@ -322,6 +295,10 @@ ha_remove_all_nodes_to_page(
 	ut_ad(table);
 	ut_ad(table->magic_n == HASH_TABLE_MAGIC_N);
 	ASSERT_HASH_MUTEX_OWN(table, fold);
+#ifdef UNIV_SYNC_DEBUG
+	ut_ad(rw_lock_own(&btr_search_latch, RW_LOCK_EX));
+#endif /* UNIV_SYNC_DEBUG */
+	ut_ad(btr_search_enabled);
 
 	node = ha_chain_get_first(table, fold);
 

handler/ha_innodb.h

@@ -133,9 +133,11 @@ class ha_innobase: public handler
 	const key_map* keys_to_use_for_scanning();
 
 	int open(const char *name, int mode, uint test_if_locked);
+	handler* clone(const char *name, MEM_ROOT *mem_root);
 	int close(void);
 	double scan_time();
 	double read_time(uint index, uint ranges, ha_rows rows);
+	my_bool is_fake_change_enabled(THD *thd);
 	bool is_corrupt() const;
 
 	int write_row(uchar * buf);

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
 
 *****************************************************************************/
 
@@ -668,6 +668,10 @@ ha_innobase::add_index(
 		DBUG_RETURN(HA_ERR_NO_SUCH_TABLE);
 	}
 
+	if (innodb_table->tablespace_discarded) {
+		DBUG_RETURN(-1);
+	}
+
 	/* Check that index keys are sensible */
 	error = innobase_check_index_keys(key_info, num_of_keys, innodb_table);
 
@@ -823,6 +827,8 @@ ha_innobase::add_index(
 					innodb_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. */
@@ -1024,7 +1030,9 @@ 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
@@ -1143,7 +1151,9 @@ func_exit:
 			= 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);
 	}
@@ -1209,7 +1219,9 @@ 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;

handler/i_s.h

@@ -36,7 +36,6 @@ extern struct st_mysql_plugin	i_s_innodb_cmp;
 extern struct st_mysql_plugin	i_s_innodb_cmp_reset;
 extern struct st_mysql_plugin	i_s_innodb_cmpmem;
 extern struct st_mysql_plugin	i_s_innodb_cmpmem_reset;
-extern struct st_mysql_plugin	i_s_innodb_patches;
 extern struct st_mysql_plugin	i_s_innodb_rseg;
 extern struct st_mysql_plugin	i_s_innodb_table_stats;
 extern struct st_mysql_plugin	i_s_innodb_index_stats;
@@ -44,5 +43,9 @@ extern struct st_mysql_plugin	i_s_innodb_admin_command;
 extern struct st_mysql_plugin   i_s_innodb_sys_tables;
 extern struct st_mysql_plugin   i_s_innodb_sys_indexes;
 extern struct st_mysql_plugin	i_s_innodb_sys_stats;
+extern struct st_mysql_plugin	i_s_innodb_changed_pages;
+extern struct st_mysql_plugin	i_s_innodb_buffer_page;
+extern struct st_mysql_plugin	i_s_innodb_buffer_page_lru;
+extern struct st_mysql_plugin	i_s_innodb_buffer_stats;
 
 #endif /* i_s_h */

handler/innodb_patch_info.h

-/* Copyright (C) 2002-2006 MySQL AB
-  
-   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 Foundation; version 2 of the License.
-  
-   This program is distributed in the hope that it will be useful,
-   but WITHOUT 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 */
-
-#ifdef USE_PRAGMA_INTERFACE
-#pragma interface                      /* gcc class implementation */
-#endif
-
-struct innodb_enhancement {
-       const char *file;
-       const char *name;
-       const char *comment;
-       const char *link;
-}innodb_enhancements[] = {
-{"xtradb_show_enhancements","I_S.XTRADB_ENHANCEMENTS","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_show_status","Improvements to SHOW INNODB STATUS","Memory information and lock info fixes","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_io","Improvements to InnoDB IO","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_opt_lru_count","Fix of buffer_pool mutex","Decreases contention on buffer_pool mutex on LRU operations","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_buffer_pool_pages","Information of buffer pool content","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_expand_undo_slots","expandable maximum number of undo slots","from 1024 (default) to about 4000","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_extra_rseg","allow to create extra rollback segments","When create new db, the new parameter allows to create more rollback segments","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_overwrite_relay_log_info","overwrite relay-log.info when slave recovery","Building as plugin, it is not used.","http://www.percona.com/docs/wiki/percona-xtradb:innodb_overwrite_relay_log_info"},
-{"innodb_thread_concurrency_timer_based","use InnoDB timer based concurrency throttling (backport from MySQL 5.4.0)","",""},
-{"innodb_expand_import","convert .ibd file automatically when import tablespace","the files are generated by xtrabackup export mode.","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_dict_size_limit","Limit dictionary cache size","Variable innodb_dict_size_limit in bytes","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_split_buf_pool_mutex","More fix of buffer_pool mutex","Spliting buf_pool_mutex and optimizing based on innodb_opt_lru_count","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_stats","Additional features about InnoDB statistics/optimizer","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_recovery_patches","Bugfixes and adjustments about recovery process","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_purge_thread","Enable to use purge devoted thread","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_admin_command_base","XtraDB specific command interface through i_s","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_show_lock_name","Show mutex/lock name instead of crated file/line","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_extend_slow","Extended statistics in slow.log","It is InnoDB-part only. It needs to patch also to mysqld.","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_lru_dump_restore","Dump and restore command for content of buffer pool","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_separate_doublewrite","Add option 'innodb_doublewrite_file' to separate doublewrite dedicated tablespace","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_pass_corrupt_table","Treat tables as corrupt instead of crash, when meet corrupt blocks","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_fast_checksum","Using the checksum on 32bit-unit calculation","incompatible for unpatched ver.","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_files_extend","allow >4GB transaction log files, and can vary universal page size of datafiles","incompatible for unpatched ver.","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_sys_tables_sys_indexes","Expose InnoDB SYS_TABLES and SYS_INDEXES schema tables","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{NULL, NULL, NULL, NULL}
-};

include/btr0btr.h

 /*****************************************************************************
 
-Copyright (c) 1994, 2011, Oracle and/or its affiliates. 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
 
 *****************************************************************************/
 
@@ -188,26 +188,45 @@ btr_block_get_func(
 	ulint		mode,		/*!< in: latch mode */
 	const char*	file,		/*!< in: file name */
 	ulint		line,		/*!< in: line where called */
-	mtr_t*		mtr)		/*!< in/out: mtr */
-	__attribute__((nonnull));
+# ifdef UNIV_SYNC_DEBUG
+	const dict_index_t*	index,	/*!< in: index tree, may be NULL
+					if it is not an insert buffer tree */
+# endif /* UNIV_SYNC_DEBUG */
+	mtr_t*		mtr);		/*!< in/out: mini-transaction */
+# ifdef UNIV_SYNC_DEBUG
 /** Gets a buffer page and declares its latching order level.
 @param space	tablespace identifier
 @param zip_size	compressed page size in bytes or 0 for uncompressed pages
 @param page_no	page number
 @param mode	latch mode
+@param index	index tree, may be NULL if not the insert buffer tree
 @param mtr	mini-transaction handle
 @return the block descriptor */
-# define btr_block_get(space,zip_size,page_no,mode,mtr) \
+#  define btr_block_get(space,zip_size,page_no,mode,index,mtr)	\
+	btr_block_get_func(space,zip_size,page_no,mode,		\
+			   __FILE__,__LINE__,index,mtr)
+# else /* UNIV_SYNC_DEBUG */
+/** Gets a buffer page and declares its latching order level.
+@param space	tablespace identifier
+@param zip_size	compressed page size in bytes or 0 for uncompressed pages
+@param page_no	page number
+@param mode	latch mode
+@param idx	index tree, may be NULL if not the insert buffer tree
+@param mtr	mini-transaction handle
+@return the block descriptor */
+#  define btr_block_get(space,zip_size,page_no,mode,idx,mtr)		\
 	btr_block_get_func(space,zip_size,page_no,mode,__FILE__,__LINE__,mtr)
+# endif /* UNIV_SYNC_DEBUG */
 /** Gets a buffer page and declares its latching order level.
 @param space	tablespace identifier
 @param zip_size	compressed page size in bytes or 0 for uncompressed pages
 @param page_no	page number
 @param mode	latch mode
+@param idx	index tree, may be NULL if not the insert buffer tree
 @param mtr	mini-transaction handle
 @return the uncompressed page frame */
-# define btr_page_get(space,zip_size,page_no,mode,mtr) \
-	buf_block_get_frame(btr_block_get(space,zip_size,page_no,mode,mtr))
+# define btr_page_get(space,zip_size,page_no,mode,idx,mtr)		\
+	buf_block_get_frame(btr_block_get(space,zip_size,page_no,mode,idx,mtr))
 /**************************************************************//**
 Sets the index id field of a page. */
 UNIV_INLINE
@@ -378,8 +397,7 @@ btr_free_root(
 	ulint	zip_size,	/*!< in: compressed page size in bytes
 				or 0 for uncompressed pages */
 	ulint	root_page_no,	/*!< in: root page number */
-	mtr_t*	mtr);		/*!< in: a mini-transaction which has already
-				been started */
+	mtr_t*	mtr);		/*!< in/out: mini-transaction */
 /*************************************************************//**
 Makes tree one level higher by splitting the root, and inserts
 the tuple. It is assumed that mtr contains an x-latch on the tree.
@@ -588,17 +606,23 @@ 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(
@@ -609,7 +633,12 @@ 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. */

include/db0err.h

@@ -97,6 +97,8 @@ enum db_err {
 	DB_FOREIGN_EXCEED_MAX_CASCADE,	/* Foreign key constraint related
 					cascading delete/update exceeds
 					maximum allowed depth */
+	DB_TABLE_IN_FK_CHECK,		/* table is being used in foreign
+					key check */
 
 	/* The following are partial failure codes */
 	DB_FAIL = 1000,