Merge XtraDB from Percona-Server-5.5.20-24.1 into MariaDB 5.5.

21 files changed, 304 insertions, 196 deletions

diff --git a/storage/xtradb/include/btr0cur.h b/storage/xtradb/include/btr0cur.h
index be918439f59..4f33aacc48e 100644
--- a/storage/xtradb/include/btr0cur.h
+++ b/storage/xtradb/include/btr0cur.h
@@ -743,24 +743,6 @@ struct btr_cur_struct {
 					NULL */
 	ulint		fold;		/*!< fold value used in the search if
 					flag is BTR_CUR_HASH */
-	/*----- Delete buffering -------*/
-	ulint		ibuf_cnt;	/* in searches done on insert buffer
-					trees, this contains the "counter"
-					value (the first two bytes of the
-					fourth field) extracted from the
-					page above the leaf page, from the
-					father node pointer that pointed to
-					the leaf page. in other words, it
-					contains the minimum counter value
-					for records to be inserted on the
-					chosen leaf page. If for some reason
-					this can't be read, or if the search
-					ended on the leftmost leaf page in
-					the tree (in which case the father
-					node pointer had the 'minimum
-					record' flag set), this is
-					ULINT_UNDEFINED. */
-	/*------------------------------*/
 	/* @} */
 	btr_path_t*	path_arr;	/*!< in estimating the number of
 					rows in range, we store in this array
diff --git a/storage/xtradb/include/btr0pcur.h b/storage/xtradb/include/btr0pcur.h
index f605c476844..140f94466db 100644
--- a/storage/xtradb/include/btr0pcur.h
+++ b/storage/xtradb/include/btr0pcur.h
@@ -263,14 +263,6 @@ btr_pcur_commit_specify_mtr(
 /*========================*/
 	btr_pcur_t*	pcur,	/*!< in: persistent cursor */
 	mtr_t*		mtr);	/*!< in: mtr to commit */
-/**************************************************************//**
-Tests if a cursor is detached: that is the latch mode is BTR_NO_LATCHES.
-@return	TRUE if detached */
-UNIV_INLINE
-ibool
-btr_pcur_is_detached(
-/*=================*/
-	btr_pcur_t*	pcur);	/*!< in: persistent cursor */
 /*********************************************************//**
 Moves the persistent cursor to the next record in the tree. If no records are
 left, the cursor stays 'after last in tree'.
diff --git a/storage/xtradb/include/btr0pcur.ic b/storage/xtradb/include/btr0pcur.ic
index d86601e5a32..054ce753c7d 100644
--- a/storage/xtradb/include/btr0pcur.ic
+++ b/storage/xtradb/include/btr0pcur.ic
@@ -389,38 +389,6 @@ btr_pcur_commit_specify_mtr(
 }
 
 /**************************************************************//**
-Sets the pcur latch mode to BTR_NO_LATCHES. */
-UNIV_INLINE
-void
-btr_pcur_detach(
-/*============*/
-	btr_pcur_t*	pcur)	/*!< in: persistent cursor */
-{
-	ut_a(pcur->pos_state == BTR_PCUR_IS_POSITIONED);
-
-	pcur->latch_mode = BTR_NO_LATCHES;
-
-	pcur->pos_state = BTR_PCUR_WAS_POSITIONED;
-}
-
-/**************************************************************//**
-Tests if a cursor is detached: that is the latch mode is BTR_NO_LATCHES.
-@return	TRUE if detached */
-UNIV_INLINE
-ibool
-btr_pcur_is_detached(
-/*=================*/
-	btr_pcur_t*	pcur)	/*!< in: persistent cursor */
-{
-	if (pcur->latch_mode == BTR_NO_LATCHES) {
-
-		return(TRUE);
-	}
-
-	return(FALSE);
-}
-
-/**************************************************************//**
 Sets the old_rec_buf field to NULL. */
 UNIV_INLINE
 void
diff --git a/storage/xtradb/include/btr0sea.h b/storage/xtradb/include/btr0sea.h
index 9e82602934e..d9f2cf9c81b 100644
--- a/storage/xtradb/include/btr0sea.h
+++ b/storage/xtradb/include/btr0sea.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1996, 2009, 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
@@ -137,11 +137,10 @@ UNIV_INTERN
 void
 btr_search_drop_page_hash_index(
 /*============================*/
-	buf_block_t*	block,	/*!< in: block containing index page,
+	buf_block_t*	block);	/*!< in: block containing index page,
 				s- or x-latched, or an index page
 				for which we know that
 				block->buf_fix_count == 0 */
-	dict_index_t*	index_in);
 /************************************************************************
 Drops a page hash index based on index */
 UNIV_INTERN
@@ -150,8 +149,8 @@ btr_search_drop_page_hash_index_on_index(
 /*=====================================*/
 	dict_index_t*	index);		/* in: record descriptor */
 /********************************************************************//**
-Drops a page hash index when a page is freed from a fseg to the file system.
-Drops possible hash index if the page happens to be in the buffer pool. */
+Drops a possible page hash index when a page is evicted from the buffer pool
+or freed in a file segment. */
 UNIV_INTERN
 void
 btr_search_drop_page_hash_when_freed(
@@ -235,19 +234,6 @@ void
 btr_search_s_unlock_all(void);
 /*==========================*/
 
-
-/** Flag: has the search system been enabled?
-Protected by btr_search_latch and btr_search_enabled_mutex. */
-extern char	btr_search_enabled;
-
-extern ulint	btr_search_index_num;
-
-/** Flag: whether the search system has completed its disabling process,
-It is set to TRUE right after buf_pool_drop_hash_index() in
-btr_search_disable(), indicating hash index entries are cleaned up.
-Protected by btr_search_latch and btr_search_enabled_mutex. */
-extern ibool	btr_search_fully_disabled;
-
 /** The search info struct in an index */
 struct btr_search_struct{
 	ulint	ref_count;	/*!< Number of blocks in this index tree
@@ -316,26 +302,6 @@ struct btr_search_sys_struct{
 /** The adaptive hash index */
 extern btr_search_sys_t*	btr_search_sys;
 
-/** @brief The latch protecting the adaptive search system
-
-This latch protects the
-(1) hash index;
-(2) columns of a record to which we have a pointer in the hash index;
-
-but does NOT protect:
-
-(3) next record offset field in a record;
-(4) next or previous records on the same page.
-
-Bear in mind (3) and (4) when using the hash index.
-*/
-//extern rw_lock_t*	btr_search_latch_temp;
-
-extern rw_lock_t**	btr_search_latch_part;
-
-/** The latch protecting the adaptive search system */
-//#define btr_search_latch	(*btr_search_latch_temp)
-
 #ifdef UNIV_SEARCH_PERF_STAT
 /** Number of successful adaptive hash index lookups */
 extern ulint	btr_search_n_succ;
diff --git a/storage/xtradb/include/btr0types.h b/storage/xtradb/include/btr0types.h
index 07c06fb18d7..6f515c3f58c 100644
--- a/storage/xtradb/include/btr0types.h
+++ b/storage/xtradb/include/btr0types.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1996, 2009, 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
@@ -30,6 +30,7 @@ Created 2/17/1996 Heikki Tuuri
 
 #include "rem0types.h"
 #include "page0types.h"
+#include "sync0rw.h"
 
 /** Persistent cursor */
 typedef struct btr_pcur_struct		btr_pcur_t;
@@ -38,6 +39,32 @@ typedef struct btr_cur_struct		btr_cur_t;
 /** B-tree search information for the adaptive hash index */
 typedef struct btr_search_struct	btr_search_t;
 
+/** @brief The latch protecting the adaptive search system
+
+This latch protects the
+(1) hash index;
+(2) columns of a record to which we have a pointer in the hash index;
+
+but does NOT protect:
+
+(3) next record offset field in a record;
+(4) next or previous records on the same page.
+
+Bear in mind (3) and (4) when using the hash index.
+*/
+//extern rw_lock_t*	btr_search_latch_temp;
+
+extern rw_lock_t**	btr_search_latch_part;
+
+/** The latch protecting the adaptive search system */
+//#define btr_search_latch	(*btr_search_latch_temp)
+
+/** Flag: has the search system been enabled?
+Protected by btr_search_latch. */
+extern char	btr_search_enabled;
+
+extern ulint	btr_search_index_num;
+
 #ifdef UNIV_BLOB_DEBUG
 # include "buf0types.h"
 /** An index->blobs entry for keeping track of off-page column references */
diff --git a/storage/xtradb/include/buf0buf.h b/storage/xtradb/include/buf0buf.h
index d85dac5cc91..0ea74fb9eb2 100644
--- a/storage/xtradb/include/buf0buf.h
+++ b/storage/xtradb/include/buf0buf.h
@@ -244,13 +244,11 @@ buf_pool_free(
 	ulint	n_instances);	/*!< in: numbere of instances to free */
 
 /********************************************************************//**
-Drops the adaptive hash index.  To prevent a livelock, this function
-is only to be called while holding btr_search_latch and while
-btr_search_enabled == FALSE. */
+Clears the adaptive hash index on all pages in the buffer pool. */
 UNIV_INTERN
 void
-buf_pool_drop_hash_index(void);
-/*==========================*/
+buf_pool_clear_hash_index(void);
+/*===========================*/
 
 /********************************************************************//**
 Relocate a buffer control block.  Relocates the block on the LRU list
@@ -583,17 +581,6 @@ buf_page_peek_if_too_old(
 /*=====================*/
 	const buf_page_t*	bpage);	/*!< in: block to make younger */
 /********************************************************************//**
-Returns the current state of is_hashed of a page. FALSE if the page is
-not in the pool. NOTE that this operation does not fix the page in the
-pool if it is found there.
-@return	TRUE if page hash index is built in search system */
-UNIV_INTERN
-ibool
-buf_page_peek_if_search_hashed(
-/*===========================*/
-	ulint	space,	/*!< in: space id */
-	ulint	offset);/*!< in: page number */
-/********************************************************************//**
 Gets the youngest modification log sequence number for a frame.
 Returns zero if not file page or no modification occurred yet.
 @return	newest modification to page */
@@ -952,7 +939,27 @@ buf_block_set_io_fix(
 /*=================*/
 	buf_block_t*	block,	/*!< in/out: control block */
 	enum buf_io_fix	io_fix);/*!< in: io_fix state */
-
+/*********************************************************************//**
+Makes a block sticky. A sticky block implies that even after we release
+the buf_pool->mutex and the block->mutex:
+* it cannot be removed from the flush_list
+* the block descriptor cannot be relocated
+* it cannot be removed from the LRU list
+Note that:
+* the block can still change its position in the LRU list
+* the next and previous pointers can change. */
+UNIV_INLINE
+void
+buf_page_set_sticky(
+/*================*/
+	buf_page_t*	bpage);	/*!< in/out: control block */
+/*********************************************************************//**
+Removes stickiness of a block. */
+UNIV_INLINE
+void
+buf_page_unset_sticky(
+/*==================*/
+	buf_page_t*	bpage);	/*!< in/out: control block */
 /********************************************************************//**
 Determine if a buffer block can be relocated in memory.  The block
 can be dirty, but it must not be I/O-fixed or bufferfixed. */
@@ -1569,13 +1576,16 @@ struct buf_block_struct{
 	/* @} */
 
 	/** @name Hash search fields
-	These 6 fields may only be modified when we have
+	These 5 fields may only be modified when we have
 	an x-latch on btr_search_latch AND
 	- we are holding an s-latch or x-latch on buf_block_struct::lock or
 	- we know that buf_block_struct::buf_fix_count == 0.
 
 	An exception to this is when we init or create a page
-	in the buffer pool in buf0buf.c. */
+	in the buffer pool in buf0buf.c.
+
+	Another exception is that assigning block->index = NULL
+	is allowed whenever holding an x-latch on btr_search_latch. */
 
 	/* @{ */
 
@@ -1584,21 +1594,21 @@ struct buf_block_struct{
 					pointers in the adaptive hash index
 					pointing to this frame */
 #endif /* UNIV_AHI_DEBUG || UNIV_DEBUG */
-	volatile unsigned	is_hashed:1;	/*!< TRUE if hash index has
-					already been built on this
-					page; note that it does not
-					guarantee that the index is
-					complete, though: there may
-					have been hash collisions,
-					record deletions, etc. */
 	unsigned	curr_n_fields:10;/*!< prefix length for hash indexing:
 					number of full fields */
 	unsigned	curr_n_bytes:15;/*!< number of bytes in hash
 					indexing */
 	unsigned	curr_left_side:1;/*!< TRUE or FALSE in hash indexing */
-	dict_index_t*	index;		/*!< Index for which the adaptive
-					hash index has been created. */
-	volatile rw_lock_t*	btr_search_latch;
+	dict_index_t*	index;		/*!< Index for which the
+					adaptive hash index has been
+					created, or NULL if the page
+					does not exist in the
+					index. Note that it does not
+					guarantee that the index is
+					complete, though: there may
+					have been hash collisions,
+					record deletions, etc. */
+	volatile rw_lock_t*     btr_search_latch;
 	/* @} */
 # ifdef UNIV_SYNC_DEBUG
 	/** @name Debug fields */
diff --git a/storage/xtradb/include/buf0buf.ic b/storage/xtradb/include/buf0buf.ic
index 98d575075ba..55d89b66375 100644
--- a/storage/xtradb/include/buf0buf.ic
+++ b/storage/xtradb/include/buf0buf.ic
@@ -444,6 +444,7 @@ buf_page_get_io_fix(
 	case BUF_IO_NONE:
 	case BUF_IO_READ:
 	case BUF_IO_WRITE:
+	case BUF_IO_PIN:
 		return(io_fix);
 	}
 	ut_error;
@@ -494,6 +495,49 @@ buf_block_set_io_fix(
 	buf_page_set_io_fix(&block->page, io_fix);
 }
 
+/*********************************************************************//**
+Makes a block sticky. A sticky block implies that even after we release
+the buf_pool->mutex and the block->mutex:
+* it cannot be removed from the flush_list
+* the block descriptor cannot be relocated
+* it cannot be removed from the LRU list
+Note that:
+* the block can still change its position in the LRU list
+* the next and previous pointers can change. */
+UNIV_INLINE
+void
+buf_page_set_sticky(
+/*================*/
+	buf_page_t*	bpage)	/*!< in/out: control block */
+{
+#ifdef UNIV_DEBUG
+	buf_pool_t*	buf_pool = buf_pool_from_bpage(bpage);
+	ut_ad(mutex_own(&buf_pool->LRU_list_mutex));
+#endif
+	ut_ad(mutex_own(buf_page_get_mutex(bpage)));
+	ut_ad(buf_page_get_io_fix(bpage) == BUF_IO_NONE);
+
+	bpage->io_fix = BUF_IO_PIN;
+}
+
+/*********************************************************************//**
+Removes stickiness of a block. */
+UNIV_INLINE
+void
+buf_page_unset_sticky(
+/*==================*/
+	buf_page_t*	bpage)	/*!< in/out: control block */
+{
+#ifdef UNIV_DEBUG
+	buf_pool_t*	buf_pool = buf_pool_from_bpage(bpage);
+	ut_ad(mutex_own(&buf_pool->LRU_list_mutex));
+#endif
+	ut_ad(mutex_own(buf_page_get_mutex(bpage)));
+	ut_ad(buf_page_get_io_fix(bpage) == BUF_IO_PIN);
+
+	bpage->io_fix = BUF_IO_NONE;
+}
+
 /********************************************************************//**
 Determine if a buffer block can be relocated in memory.  The block
 can be dirty, but it must not be I/O-fixed or bufferfixed. */
diff --git a/storage/xtradb/include/buf0types.h b/storage/xtradb/include/buf0types.h
index 156093aad01..d140936a886 100644
--- a/storage/xtradb/include/buf0types.h
+++ b/storage/xtradb/include/buf0types.h
@@ -57,7 +57,10 @@ enum buf_flush {
 enum buf_io_fix {
 	BUF_IO_NONE = 0,		/**< no pending I/O */
 	BUF_IO_READ,			/**< read pending */
-	BUF_IO_WRITE			/**< write pending */
+	BUF_IO_WRITE,			/**< write pending */
+	BUF_IO_PIN			/**< disallow relocation of
+					block and its removal of from
+					the flush_list */
 };
 
 /** Parameters of binary buddy system for compressed pages (buf0buddy.h) */
diff --git a/storage/xtradb/include/fil0fil.h b/storage/xtradb/include/fil0fil.h
index 840a9fbb13a..92fdca0db1b 100644
--- a/storage/xtradb/include/fil0fil.h
+++ b/storage/xtradb/include/fil0fil.h
@@ -34,6 +34,7 @@ Created 10/25/1995 Heikki Tuuri
 #include "sync0rw.h"
 #include "ibuf0types.h"
 #endif /* !UNIV_HOTBACKUP */
+#include "trx0types.h"
 
 /** When mysqld is run, the default directory "." is the mysqld datadir,
 but in the MySQL Embedded Server Library and ibbackup it is not the default
@@ -329,18 +330,23 @@ Reads the flushed lsn and arch no fields from a data file at database
 startup. */
 UNIV_INTERN
 void
-fil_read_flushed_lsn_and_arch_log_no(
-/*=================================*/
+fil_read_first_page(
+/*================*/
 	os_file_t	data_file,		/*!< in: open data file */
 	ibool		one_read_already,	/*!< in: TRUE if min and max
 						parameters below already
 						contain sensible data */
+	ulint*		flags,			/*!< out: tablespace flags */
 #ifdef UNIV_LOG_ARCHIVE
-	ulint*		min_arch_log_no,	/*!< in/out: */
-	ulint*		max_arch_log_no,	/*!< in/out: */
+	ulint*		min_arch_log_no,	/*!< out: min of archived
+						log numbers in data files */
+	ulint*		max_arch_log_no,	/*!< out: max of archived
+						log numbers in data files */
 #endif /* UNIV_LOG_ARCHIVE */
-	ib_uint64_t*	min_flushed_lsn,	/*!< in/out: */
-	ib_uint64_t*	max_flushed_lsn);	/*!< in/out: */
+	ib_uint64_t*	min_flushed_lsn,	/*!< out: min of flushed
+						lsn values in data files */
+	ib_uint64_t*	max_flushed_lsn);	/*!< out: max of flushed
+						lsn values in data files */
 /*******************************************************************//**
 Increments the count of pending insert buffer page merges, if space is not
 being deleted.
@@ -474,8 +480,11 @@ 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 */
 /********************************************************************//**
 It is possible, though very improbable, that the lsn's in the tablespace to be
 imported have risen above the current system lsn, if a lengthy purge, ibuf
diff --git a/storage/xtradb/include/fsp0fsp.h b/storage/xtradb/include/fsp0fsp.h
index 7abd3914eda..d5e235aa870 100644
--- a/storage/xtradb/include/fsp0fsp.h
+++ b/storage/xtradb/include/fsp0fsp.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1995, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 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
@@ -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
 
 *****************************************************************************/
 
@@ -34,6 +34,90 @@ Created 12/18/1995 Heikki Tuuri
 #include "page0types.h"
 #include "fsp0types.h"
 
+/* @defgroup fsp_flags InnoDB Tablespace Flag Constants @{ */
+
+/** Number of flag bits used to indicate the tablespace page size */
+#define FSP_FLAGS_WIDTH_PAGE_SSIZE	4
+/** Zero relative shift position of the PAGE_SSIZE field */
+#define FSP_FLAGS_POS_PAGE_SSIZE	6
+/** Bit mask of the PAGE_SSIZE field */
+#define FSP_FLAGS_MASK_PAGE_SSIZE				\
+		((~(~0 << FSP_FLAGS_WIDTH_PAGE_SSIZE))		\
+		<< FSP_FLAGS_POS_PAGE_SSIZE)
+/** Return the value of the PAGE_SSIZE field */
+#define FSP_FLAGS_GET_PAGE_SSIZE(flags)				\
+		((flags & FSP_FLAGS_MASK_PAGE_SSIZE)		\
+		>> FSP_FLAGS_POS_PAGE_SSIZE)
+
+/* @} */
+
+/* @defgroup Tablespace Header Constants (moved from fsp0fsp.c) @{ */
+
+/** Offset of the space header within a file page */
+#define FSP_HEADER_OFFSET	FIL_PAGE_DATA
+
+/* The data structures in files are defined just as byte strings in C */
+typedef	byte	fsp_header_t;
+typedef	byte	xdes_t;
+
+/*			SPACE HEADER
+			============
+
+File space header data structure: this data structure is contained in the
+first page of a space. The space for this header is reserved in every extent
+descriptor page, but used only in the first. */
+
+/*-------------------------------------*/
+#define FSP_SPACE_ID		0	/* space id */
+#define FSP_NOT_USED		4	/* this field contained a value up to
+					which we know that the modifications
+					in the database have been flushed to
+					the file space; not used now */
+#define	FSP_SIZE		8	/* Current size of the space in
+					pages */
+#define	FSP_FREE_LIMIT		12	/* Minimum page number for which the
+					free list has not been initialized:
+					the pages >= this limit are, by
+					definition, free; note that in a
+					single-table tablespace where size
+					< 64 pages, this number is 64, i.e.,
+					we have initialized the space
+					about the first extent, but have not
+					physically allocted those pages to the
+					file */
+#define	FSP_SPACE_FLAGS		16	/* fsp_space_t.flags, similar to
+					dict_table_t::flags */
+#define	FSP_FRAG_N_USED		20	/* number of used pages in the
+					FSP_FREE_FRAG list */
+#define	FSP_FREE		24	/* list of free extents */
+#define	FSP_FREE_FRAG		(24 + FLST_BASE_NODE_SIZE)
+					/* list of partially free extents not
+					belonging to any segment */
+#define	FSP_FULL_FRAG		(24 + 2 * FLST_BASE_NODE_SIZE)
+					/* list of full extents not belonging
+					to any segment */
+#define FSP_SEG_ID		(24 + 3 * FLST_BASE_NODE_SIZE)
+					/* 8 bytes which give the first unused
+					segment id */
+#define FSP_SEG_INODES_FULL	(32 + 3 * FLST_BASE_NODE_SIZE)
+					/* list of pages containing segment
+					headers, where all the segment inode
+					slots are reserved */
+#define FSP_SEG_INODES_FREE	(32 + 4 * FLST_BASE_NODE_SIZE)
+					/* list of pages containing segment
+					headers, where not all the segment
+					header slots are reserved */
+/*-------------------------------------*/
+/* File space header size */
+#define	FSP_HEADER_SIZE		(32 + 5 * FLST_BASE_NODE_SIZE)
+
+#define	FSP_FREE_ADD		4	/* this many free extents are added
+					to the free list from above
+					FSP_FREE_LIMIT at a time */
+/* @} */
+
+/* @} */
+
 /**********************************************************************//**
 Initializes the file space system. */
 UNIV_INTERN
@@ -352,6 +436,18 @@ fseg_print(
 	mtr_t*		mtr);	/*!< in: mtr */
 #endif /* UNIV_BTR_PRINT */
 
+/********************************************************************//**
+Extract the page size from tablespace flags.
+This feature, storing the page_ssize into the tablespace flags, is added
+to InnoDB 5.6.4.  This is here only to protect against a crash if a newer
+database is opened with this code branch.
+@return	page size of the tablespace in bytes */
+UNIV_INLINE
+ulint
+fsp_flags_get_page_size(
+/*====================*/
+	ulint	flags);	/*!< in: tablespace flags */
+
 #ifndef UNIV_NONINL
 #include "fsp0fsp.ic"
 #endif
diff --git a/storage/xtradb/include/fsp0fsp.ic b/storage/xtradb/include/fsp0fsp.ic
index 434c370b527..c92111a9d89 100644
--- a/storage/xtradb/include/fsp0fsp.ic
+++ b/storage/xtradb/include/fsp0fsp.ic
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1995, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 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
@@ -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
 
 *****************************************************************************/
 
@@ -43,3 +43,31 @@ fsp_descr_page(
 
 	return(UNIV_UNLIKELY((page_no & (zip_size - 1)) == FSP_XDES_OFFSET));
 }
+/********************************************************************//**
+Extract the page size from tablespace flags.
+This feature, storing the page_ssize into the tablespace flags, is added
+to InnoDB 5.6.4.  This is here only to protect against a crash if a newer
+database is opened with this code branch.
+@return	page size of the tablespace in bytes */
+UNIV_INLINE
+ulint
+fsp_flags_get_page_size(
+/*====================*/
+	ulint	flags)	/*!< in: tablespace flags */
+{
+	ulint	page_size = 0;
+	ulint	ssize = FSP_FLAGS_GET_PAGE_SSIZE(flags);
+
+	/* Convert from a 'log2 minus 9' to a page size in bytes. */
+	if (UNIV_UNLIKELY(ssize)) {
+		page_size = (512 << ssize);
+
+		ut_ad(page_size <= UNIV_PAGE_SIZE);
+	} else {
+		/* If the page size was not stored, then it is the
+		original 16k. */
+		page_size = UNIV_PAGE_SIZE;
+	}
+
+	return(page_size);
+}
diff --git a/storage/xtradb/include/ha0ha.h b/storage/xtradb/include/ha0ha.h
index 3299000bf3c..8bba564d153 100644
--- a/storage/xtradb/include/ha0ha.h
+++ b/storage/xtradb/include/ha0ha.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-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
@@ -31,13 +31,14 @@ Created 8/18/1994 Heikki Tuuri
 #include "hash0hash.h"
 #include "page0types.h"
 #include "buf0types.h"
+#include "rem0types.h"
 
 /*************************************************************//**
 Looks for an element in a hash table.
 @return pointer to the data of the first hash table node in chain
 having the fold number, NULL if not found */
 UNIV_INLINE
-void*
+const rec_t*
 ha_search_and_get_data(
 /*===================*/
 	hash_table_t*	table,	/*!< in: hash table */
@@ -51,11 +52,11 @@ 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 */
 
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 /** Looks for an element when we know the pointer to the data and
@@ -114,14 +115,6 @@ chosen to be a slightly bigger prime number.
 #endif /* UNIV_SYNC_DEBUG */
 
 /*************************************************************//**
-Empties a hash table and frees the memory heaps. */
-UNIV_INTERN
-void
-ha_clear(
-/*=====*/
-	hash_table_t*	table);	/*!< in, own: hash table */
-
-/*************************************************************//**
 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
 is inserted.
@@ -138,7 +131,7 @@ 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 */
 
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 /**
@@ -174,7 +167,7 @@ ha_search_and_delete_if_found(
 /*==========================*/
 	hash_table_t*	table,	/*!< in: 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 */
 #ifndef UNIV_HOTBACKUP
 /*****************************************************************//**
 Removes from the chain determined by fold all nodes whose data pointer
@@ -217,7 +210,7 @@ struct ha_node_struct {
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 	buf_block_t*	block;	/*!< buffer block containing the data, or NULL */
 #endif /* UNIV_AHI_DEBUG || UNIV_DEBUG */
-	void*		data;	/*!< pointer to the data */
+	const rec_t*	data;	/*!< pointer to the data */
 	ulint		fold;	/*!< fold value for the data */
 };
 
diff --git a/storage/xtradb/include/ha0ha.ic b/storage/xtradb/include/ha0ha.ic
index 734403c4cd9..f224776c134 100644
--- a/storage/xtradb/include/ha0ha.ic
+++ b/storage/xtradb/include/ha0ha.ic
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-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
@@ -25,6 +25,7 @@ Created 8/18/1994 Heikki Tuuri
 
 #include "ut0rnd.h"
 #include "mem0mem.h"
+#include "btr0types.h"
 
 /***********************************************************//**
 Deletes a hash node. */
@@ -39,10 +40,10 @@ ha_delete_hash_node(
 Gets a hash node data.
 @return	pointer to the data */
 UNIV_INLINE
-void*
+const rec_t*
 ha_node_get_data(
 /*=============*/
-	ha_node_t*	node)	/*!< in: hash chain node */
+	const ha_node_t*	node)	/*!< in: hash chain node */
 {
 	return(node->data);
 }
@@ -57,7 +58,7 @@ ha_node_set_data_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: pointer to the data */
+	const rec_t*	data)	/*!< in: pointer to the data */
 {
 #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
 	node->block = block;
@@ -107,39 +108,10 @@ ha_chain_get_first(
 
 /*************************************************************//**
 Looks for an element in a hash table.
-@return pointer to the first hash table node in chain having the fold
-number, NULL if not found */
-UNIV_INLINE
-ha_node_t*
-ha_search(
-/*======*/
-	hash_table_t*	table,	/*!< in: hash table */
-	ulint		fold)	/*!< in: folded value of the searched data */
-{
-	ha_node_t*	node;
-
-	ASSERT_HASH_MUTEX_OWN(table, fold);
-
-	node = ha_chain_get_first(table, fold);
-
-	while (node) {
-		if (node->fold == fold) {
-
-			return(node);
-		}
-
-		node = ha_chain_get_next(node);
-	}
-
-	return(NULL);
-}
-
-/*************************************************************//**
-Looks for an element in a hash table.
 @return pointer to the data of the first hash table node in chain
 having the fold number, NULL if not found */
 UNIV_INLINE
-void*
+const rec_t*
 ha_search_and_get_data(
 /*===================*/
 	hash_table_t*	table,	/*!< in: hash table */
@@ -148,6 +120,10 @@ ha_search_and_get_data(
 	ha_node_t*	node;
 
 	ASSERT_HASH_MUTEX_OWN(table, fold);
+#ifdef UNIV_SYNC_DEBUG
+//	ut_ad(rw_lock_own(&btr_search_latch, RW_LOCK_SHARED));
+#endif /* UNIV_SYNC_DEBUG */
+	ut_ad(btr_search_enabled);
 
 	node = ha_chain_get_first(table, fold);
 
@@ -172,12 +148,14 @@ ha_search_with_data(
 /*================*/
 	hash_table_t*	table,	/*!< in: 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 */
 {
 	ha_node_t*	node;
 
 	ASSERT_HASH_MUTEX_OWN(table, fold);
 
+	ut_ad(btr_search_enabled);
+
 	node = ha_chain_get_first(table, fold);
 
 	while (node) {
@@ -202,11 +180,15 @@ ha_search_and_delete_if_found(
 /*==========================*/
 	hash_table_t*	table,	/*!< in: 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 */
 {
 	ha_node_t*	node;
 
 	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_search_with_data(table, fold, data);
 
diff --git a/storage/xtradb/include/page0page.h b/storage/xtradb/include/page0page.h
index c1c659e2b70..1b9f9c84225 100644
--- a/storage/xtradb/include/page0page.h
+++ b/storage/xtradb/include/page0page.h
@@ -893,6 +893,7 @@ page_parse_create(
 	ulint		comp,	/*!< in: nonzero=compact page format */
 	buf_block_t*	block,	/*!< in: block or NULL */
 	mtr_t*		mtr);	/*!< in: mtr or NULL */
+#ifndef UNIV_HOTBACKUP
 /************************************************************//**
 Prints record contents including the data relevant only in
 the index page context. */
@@ -902,6 +903,7 @@ page_rec_print(
 /*===========*/
 	const rec_t*	rec,	/*!< in: physical record */
 	const ulint*	offsets);/*!< in: record descriptor */
+# ifdef UNIV_BTR_PRINT
 /***************************************************************//**
 This is used to print the contents of the directory for
 debugging purposes. */
@@ -941,6 +943,8 @@ page_print(
 				in directory */
 	ulint		rn);	/*!< in: print rn first and last records
 				in directory */
+# endif /* UNIV_BTR_PRINT */
+#endif /* !UNIV_HOTBACKUP */
 /***************************************************************//**
 The following is used to validate a record on a page. This function
 differs from rec_validate as it can also check the n_owned field and
diff --git a/storage/xtradb/include/row0upd.ic b/storage/xtradb/include/row0upd.ic
index 11db82f64da..10646241125 100644
--- a/storage/xtradb/include/row0upd.ic
+++ b/storage/xtradb/include/row0upd.ic
@@ -28,7 +28,6 @@ Created 12/27/1996 Heikki Tuuri
 # include "trx0trx.h"
 # include "trx0undo.h"
 # include "row0row.h"
-# include "btr0sea.h"
 #endif /* !UNIV_HOTBACKUP */
 #include "page0zip.h"
 
diff --git a/storage/xtradb/include/srv0srv.h b/storage/xtradb/include/srv0srv.h
index 6f985eacdf3..65b341c0f13 100644
--- a/storage/xtradb/include/srv0srv.h
+++ b/storage/xtradb/include/srv0srv.h
@@ -288,6 +288,9 @@ extern	ibool	srv_print_lock_waits;
 extern	ibool	srv_print_buf_io;
 extern	ibool	srv_print_log_io;
 extern	ibool	srv_print_latch_waits;
+
+extern	ulint	srv_flush_checkpoint_debug;
+
 #else /* UNIV_DEBUG */
 # define srv_print_thread_releases	FALSE
 # define srv_print_lock_waits		FALSE
diff --git a/storage/xtradb/include/sync0rw.h b/storage/xtradb/include/sync0rw.h
index 6420bc5d3bf..971099c91f5 100644
--- a/storage/xtradb/include/sync0rw.h
+++ b/storage/xtradb/include/sync0rw.h
@@ -543,7 +543,7 @@ mutex. */
 UNIV_INTERN
 void
 rw_lock_debug_mutex_enter(void);
-/*==========================*/
+/*===========================*/
 /******************************************************************//**
 Releases the debug mutex. */
 UNIV_INTERN
@@ -634,7 +634,8 @@ struct rw_lock_struct {
 };
 
 #ifdef UNIV_SYNC_DEBUG
-/** The structure for storing debug info of an rw-lock */
+/** The structure for storing debug info of an rw-lock.  All access to this
+structure must be protected by rw_lock_debug_mutex_enter(). */
 struct	rw_lock_debug_struct {
 
 	os_thread_id_t thread_id;  /*!< The thread id of the thread which
diff --git a/storage/xtradb/include/sync0rw.ic b/storage/xtradb/include/sync0rw.ic
index a6dfa603c59..3eaa6172631 100644
--- a/storage/xtradb/include/sync0rw.ic
+++ b/storage/xtradb/include/sync0rw.ic
@@ -406,6 +406,7 @@ rw_lock_s_lock_func(
 
 #ifdef UNIV_SYNC_DEBUG
 	ut_ad(!rw_lock_own(lock, RW_LOCK_SHARED)); /* see NOTE above */
+	ut_ad(!rw_lock_own(lock, RW_LOCK_EX));
 #endif /* UNIV_SYNC_DEBUG */
 
 	/* TODO: study performance of UNIV_LIKELY branch prediction hints. */
diff --git a/storage/xtradb/include/sync0sync.h b/storage/xtradb/include/sync0sync.h
index 6440a670f14..f9434ed54be 100644
--- a/storage/xtradb/include/sync0sync.h
+++ b/storage/xtradb/include/sync0sync.h
@@ -674,7 +674,6 @@ or row lock! */
 #define SYNC_LOG_FLUSH_ORDER	156
 #define SYNC_RECV		168
 #define	SYNC_WORK_QUEUE		162
-#define	SYNC_SEARCH_SYS_CONF	161	/* for assigning btr_search_enabled */
 #define	SYNC_SEARCH_SYS		160	/* NOTE that if we have a memory
 					heap that can be extended to the
 					buffer pool, its logical level is
diff --git a/storage/xtradb/include/trx0trx.h b/storage/xtradb/include/trx0trx.h
index a03f7aceafa..eded5c303fa 100644
--- a/storage/xtradb/include/trx0trx.h
+++ b/storage/xtradb/include/trx0trx.h
@@ -494,6 +494,7 @@ struct trx_struct{
 					this is set to 1 then registered should
 					also be set to 1. This is used in the
 					XA code */
+	unsigned	called_commit_ordered:1;/* 1 if innobase_commit_ordered has run. */
 	/*------------------------------*/
 	ulint		isolation_level;/* TRX_ISO_REPEATABLE_READ, ... */
 	ulint		check_foreigns;	/* normally TRUE, but if the user
diff --git a/storage/xtradb/include/univ.i b/storage/xtradb/include/univ.i
index 0287ea40509..1a48d68d0d7 100644
--- a/storage/xtradb/include/univ.i
+++ b/storage/xtradb/include/univ.i
@@ -54,7 +54,7 @@ Created 1/20/1994 Heikki Tuuri
 #define INNODB_VERSION_BUGFIX	8
 
 #ifndef PERCONA_INNODB_VERSION
-#define PERCONA_INNODB_VERSION 20.1
+#define PERCONA_INNODB_VERSION 24.1
 #endif
 
 /* The following is the InnoDB version as shown in