Merge with MariaDB 5.1

39 files changed, 755 insertions, 375 deletions

diff --git a/storage/xtradb/include/btr0btr.h b/storage/xtradb/include/btr0btr.h
index dde3a0bab69..1fe40965c0f 100644
--- a/storage/xtradb/include/btr0btr.h
+++ b/storage/xtradb/include/btr0btr.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1994, 2010, 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
@@ -81,6 +81,91 @@ UNIQUE definition on secondary indexes when we decide if we can use
 the insert buffer to speed up inserts */
 #define BTR_IGNORE_SEC_UNIQUE	2048
 
+#ifdef UNIV_BLOB_DEBUG
+# include "ut0rbt.h"
+/** An index->blobs entry for keeping track of off-page column references */
+struct btr_blob_dbg_struct
+{
+	unsigned	blob_page_no:32;	/*!< first BLOB page number */
+	unsigned	ref_page_no:32;		/*!< referring page number */
+	unsigned	ref_heap_no:16;		/*!< referring heap number */
+	unsigned	ref_field_no:10;	/*!< referring field number */
+	unsigned	owner:1;		/*!< TRUE if BLOB owner */
+	unsigned	always_owner:1;		/*!< TRUE if always
+						has been the BLOB owner;
+						reset to TRUE on B-tree
+						page splits and merges */
+	unsigned	del:1;			/*!< TRUE if currently
+						delete-marked */
+};
+
+/**************************************************************//**
+Add a reference to an off-page column to the index->blobs map. */
+UNIV_INTERN
+void
+btr_blob_dbg_add_blob(
+/*==================*/
+	const rec_t*	rec,		/*!< in: clustered index record */
+	ulint		field_no,	/*!< in: number of off-page column */
+	ulint		page_no,	/*!< in: start page of the column */
+	dict_index_t*	index,		/*!< in/out: index tree */
+	const char*	ctx)		/*!< in: context (for logging) */
+	__attribute__((nonnull));
+/**************************************************************//**
+Display the references to off-page columns.
+This function is to be called from a debugger,
+for example when a breakpoint on ut_dbg_assertion_failed is hit. */
+UNIV_INTERN
+void
+btr_blob_dbg_print(
+/*===============*/
+	const dict_index_t*	index)	/*!< in: index tree */
+	__attribute__((nonnull));
+/**************************************************************//**
+Check that there are no references to off-page columns from or to
+the given page. Invoked when freeing or clearing a page.
+@return TRUE when no orphan references exist */
+UNIV_INTERN
+ibool
+btr_blob_dbg_is_empty(
+/*==================*/
+	dict_index_t*	index,		/*!< in: index */
+	ulint		page_no)	/*!< in: page number */
+	__attribute__((nonnull, warn_unused_result));
+
+/**************************************************************//**
+Modify the 'deleted' flag of a record. */
+UNIV_INTERN
+void
+btr_blob_dbg_set_deleted_flag(
+/*==========================*/
+	const rec_t*		rec,	/*!< in: record */
+	dict_index_t*		index,	/*!< in/out: index */
+	const ulint*		offsets,/*!< in: rec_get_offs(rec, index) */
+	ibool			del)	/*!< in: TRUE=deleted, FALSE=exists */
+	__attribute__((nonnull));
+/**************************************************************//**
+Change the ownership of an off-page column. */
+UNIV_INTERN
+void
+btr_blob_dbg_owner(
+/*===============*/
+	const rec_t*		rec,	/*!< in: record */
+	dict_index_t*		index,	/*!< in/out: index */
+	const ulint*		offsets,/*!< in: rec_get_offs(rec, index) */
+	ulint			i,	/*!< in: ith field in rec */
+	ibool			own)	/*!< in: TRUE=owned, FALSE=disowned */
+	__attribute__((nonnull));
+/** Assert that there are no BLOB references to or from the given page. */
+# define btr_blob_dbg_assert_empty(index, page_no)	\
+	ut_a(btr_blob_dbg_is_empty(index, page_no))
+#else /* UNIV_BLOB_DEBUG */
+# define btr_blob_dbg_add_blob(rec, field_no, page, index, ctx)	((void) 0)
+# define btr_blob_dbg_set_deleted_flag(rec, index, offsets, del)((void) 0)
+# define btr_blob_dbg_owner(rec, index, offsets, i, val)	((void) 0)
+# define btr_blob_dbg_assert_empty(index, page_no)		((void) 0)
+#endif /* UNIV_BLOB_DEBUG */
+
 /**************************************************************//**
 Gets the root node of a tree and x-latches it.
 @return	root page, x-latched */
@@ -123,6 +208,17 @@ btr_block_get_func(
 @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))
+/**************************************************************//**
+Sets the index id field of a page. */
+UNIV_INLINE
+void
+btr_page_set_index_id(
+/*==================*/
+	page_t*		page,	/*!< in: page to be created */
+	page_zip_des_t*	page_zip,/*!< in: compressed page whose uncompressed
+				part will be updated, or NULL */
+	dulint		id,	/*!< in: index id */
+	mtr_t*		mtr);	/*!< in: mtr */
 #endif /* !UNIV_HOTBACKUP */
 /**************************************************************//**
 Gets the index id field of a page.
@@ -160,6 +256,17 @@ btr_page_get_next(
 	const page_t*	page,	/*!< in: index page */
 	mtr_t*		mtr);	/*!< in: mini-transaction handle */
 /********************************************************//**
+Sets the next index page field. */
+UNIV_INLINE
+void
+btr_page_set_next(
+/*==============*/
+	page_t*		page,	/*!< in: index page */
+	page_zip_des_t*	page_zip,/*!< in: compressed page whose uncompressed
+				part will be updated, or NULL */
+	ulint		next,	/*!< in: next page number */
+	mtr_t*		mtr);	/*!< in: mini-transaction handle */
+/********************************************************//**
 Gets the previous index page number.
 @return	prev page number */
 UNIV_INLINE
@@ -168,6 +275,17 @@ btr_page_get_prev(
 /*==============*/
 	const page_t*	page,	/*!< in: index page */
 	mtr_t*		mtr);	/*!< in: mini-transaction handle */
+/********************************************************//**
+Sets the previous index page field. */
+UNIV_INLINE
+void
+btr_page_set_prev(
+/*==============*/
+	page_t*		page,	/*!< in: index page */
+	page_zip_des_t*	page_zip,/*!< in: compressed page whose uncompressed
+				part will be updated, or NULL */
+	ulint		prev,	/*!< in: previous page number */
+	mtr_t*		mtr);	/*!< in: mini-transaction handle */
 /*************************************************************//**
 Gets pointer to the previous user record in the tree. It is assumed
 that the caller has appropriate latches on the page and its neighbor.
@@ -213,6 +331,18 @@ btr_node_ptr_get_child_page_no(
 /*===========================*/
 	const rec_t*	rec,	/*!< in: node pointer record */
 	const ulint*	offsets);/*!< in: array returned by rec_get_offsets() */
+/**************************************************************//**
+Creates a new index page (not the root, and also not
+used in page reorganization).  @see btr_page_empty(). */
+UNIV_INTERN
+void
+btr_page_create(
+/*============*/
+	buf_block_t*	block,	/*!< in/out: page to be created */
+	page_zip_des_t*	page_zip,/*!< in/out: compressed page, or NULL */
+	dict_index_t*	index,	/*!< in: index */
+	ulint		level,	/*!< in: the B-tree level of the page */
+	mtr_t*		mtr);	/*!< in: mtr */
 /************************************************************//**
 Creates the root node for a new index tree.
 @return	page number of the created root, FIL_NULL if did not succeed */
@@ -283,6 +413,17 @@ btr_page_reorganize(
 	dict_index_t*	index,	/*!< in: record descriptor */
 	mtr_t*		mtr);	/*!< in: mtr */
 /*************************************************************//**
+Empties an index page.  @see btr_page_create(). */
+UNIV_INTERN
+void
+btr_page_empty(
+/*===========*/
+	buf_block_t*	block,	/*!< in: page to be emptied */
+	page_zip_des_t*	page_zip,/*!< out: compressed page, or NULL */
+	dict_index_t*	index,	/*!< in: index of the page */
+	ulint		level,	/*!< in: the B-tree level of the page */
+	mtr_t*		mtr);	/*!< in: mtr */
+/*************************************************************//**
 Decides if the page should be split at the convergence point of
 inserts converging to left.
 @return	TRUE if split recommended */
@@ -341,6 +482,20 @@ btr_insert_on_non_leaf_level_func(
 # define btr_insert_on_non_leaf_level(i,l,t,m)				\
 	btr_insert_on_non_leaf_level_func(i,l,t,__FILE__,__LINE__,m)
 #endif /* !UNIV_HOTBACKUP */
+/**************************************************************//**
+Attaches the halves of an index page on the appropriate level in an
+index tree. */
+UNIV_INTERN
+void
+btr_attach_half_pages(
+/*==================*/
+	dict_index_t*	index,		/*!< in: the index tree */
+	buf_block_t*	block,		/*!< in/out: page to be split */
+	const rec_t*	split_rec,	/*!< in: first record on upper
+					half page */
+	buf_block_t*	new_block,	/*!< in/out: the new half page */
+	ulint		direction,	/*!< in: FSP_UP or FSP_DOWN */
+	mtr_t*		mtr);		/*!< in: mtr */
 /****************************************************************//**
 Sets a record as the predefined minimum record. */
 UNIV_INTERN
@@ -385,11 +540,14 @@ UNIV_INTERN
 ibool
 btr_compress(
 /*=========*/
-	btr_cur_t*	cursor,	/*!< in: cursor on the page to merge or lift;
-				the page must not be empty: in record delete
-				use btr_discard_page if the page would become
-				empty */
-	mtr_t*		mtr);	/*!< in: mtr */
+	btr_cur_t*	cursor,	/*!< in/out: cursor on the page to merge
+				or lift; the page must not be empty:
+				when deleting records, use btr_discard_page()
+				if the page would become empty */
+	ibool		adjust,	/*!< in: TRUE if should adjust the
+				cursor position even if compression occurs */
+	mtr_t*		mtr)	/*!< in/out: mini-transaction */
+	__attribute__((nonnull));
 /*************************************************************//**
 Discards a page from a B-tree. This is used to remove the last record from
 a B-tree page: the whole page must be removed at the same time. This cannot
diff --git a/storage/xtradb/include/btr0cur.h b/storage/xtradb/include/btr0cur.h
index ece3621fa97..6f4ce95d72f 100644
--- a/storage/xtradb/include/btr0cur.h
+++ b/storage/xtradb/include/btr0cur.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1994, 2010, 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
@@ -36,6 +36,9 @@ Created 10/16/1994 Heikki Tuuri
 #define BTR_NO_LOCKING_FLAG	2	/* do no record lock checking */
 #define BTR_KEEP_SYS_FLAG	4	/* sys fields will be found from the
 					update vector or inserted entry */
+#define BTR_KEEP_POS_FLAG	8	/* btr_cur_pessimistic_update()
+					must keep cursor position when
+					moving columns to big_rec */
 
 #ifndef UNIV_HOTBACKUP
 #include "que0types.h"
@@ -309,7 +312,9 @@ btr_cur_pessimistic_update(
 /*=======================*/
 	ulint		flags,	/*!< in: undo logging, locking, and rollback
 				flags */
-	btr_cur_t*	cursor,	/*!< in: cursor on the record to update */
+	btr_cur_t*	cursor,	/*!< in/out: cursor on the record to update;
+				cursor may become invalid if *big_rec == NULL
+				|| !(flags & BTR_KEEP_POS_FLAG) */
 	mem_heap_t**	heap,	/*!< in/out: pointer to memory heap, or NULL */
 	big_rec_t**	big_rec,/*!< out: big rec vector whose fields have to
 				be stored externally by the caller, or NULL */
@@ -321,6 +326,16 @@ btr_cur_pessimistic_update(
 	que_thr_t*	thr,	/*!< in: query thread */
 	mtr_t*		mtr);	/*!< in: mtr; must be committed before
 				latching any further pages */
+/*****************************************************************
+Commits and restarts a mini-transaction so that it will retain an
+x-lock on index->lock and the cursor page. */
+UNIV_INTERN
+void
+btr_cur_mtr_commit_and_start(
+/*=========================*/
+	btr_cur_t*	cursor,	/*!< in: cursor */
+	mtr_t*		mtr)	/*!< in/out: mini-transaction */
+	__attribute__((nonnull));
 /***********************************************************//**
 Marks a clustered index record deleted. Writes an undo log record to
 undo log on this delete marking. Writes in the trx id field the id
@@ -376,10 +391,13 @@ UNIV_INTERN
 ibool
 btr_cur_compress_if_useful(
 /*=======================*/
-	btr_cur_t*	cursor,	/*!< in: cursor on the page to compress;
+	btr_cur_t*	cursor,	/*!< in/out: cursor on the page to compress;
 				cursor does not stay valid if compression
 				occurs */
-	mtr_t*		mtr);	/*!< in: mtr */
+	ibool		adjust,	/*!< in: TRUE if should adjust the
+				cursor position even if compression occurs */
+	mtr_t*		mtr)	/*!< in/out: mini-transaction */
+	__attribute__((nonnull));
 /*******************************************************//**
 Removes the record on which the tree cursor is positioned. It is assumed
 that the mtr has an x-latch on the page where the cursor is positioned,
@@ -652,6 +670,11 @@ struct btr_path_struct{
 				order); value ULINT_UNDEFINED
 				denotes array end */
 	ulint	n_recs;		/*!< number of records on the page */
+	ulint	page_no;	/*!< no of the page containing the record */
+	ulint	page_level;	/*!< level of the page, if later we fetch
+				the page under page_no and it is no different
+				level then we know that the tree has been
+				reorganized */
 };
 
 #define BTR_PATH_ARRAY_N_SLOTS	250	/*!< size of path array (in slots) */
diff --git a/storage/xtradb/include/btr0cur.ic b/storage/xtradb/include/btr0cur.ic
index 280583f6ccf..c833b3e8572 100644
--- a/storage/xtradb/include/btr0cur.ic
+++ b/storage/xtradb/include/btr0cur.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
@@ -139,7 +139,7 @@ btr_cur_compress_recommendation(
 	btr_cur_t*	cursor,	/*!< in: btr cursor */
 	mtr_t*		mtr)	/*!< in: mtr */
 {
-	page_t*		page;
+	const page_t*		page;
 
 	ut_ad(mtr_memo_contains(mtr, btr_cur_get_block(cursor),
 				MTR_MEMO_PAGE_X_FIX));
diff --git a/storage/xtradb/include/btr0pcur.h b/storage/xtradb/include/btr0pcur.h
index 2334a266280..f59514d04b3 100644
--- a/storage/xtradb/include/btr0pcur.h
+++ b/storage/xtradb/include/btr0pcur.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-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
@@ -244,18 +244,6 @@ btr_pcur_restore_position_func(
 	mtr_t*		mtr);		/*!< in: mtr */
 #define btr_pcur_restore_position(l,cur,mtr)				\
 	btr_pcur_restore_position_func(l,cur,__FILE__,__LINE__,mtr)
-/**************************************************************//**
-If the latch mode of the cursor is BTR_LEAF_SEARCH or BTR_LEAF_MODIFY,
-releases the page latch and bufferfix reserved by the cursor.
-NOTE! In the case of BTR_LEAF_MODIFY, there should not exist changes
-made by the current mini-transaction to the data protected by the
-cursor latch, as then the latch must not be released until mtr_commit. */
-UNIV_INTERN
-void
-btr_pcur_release_leaf(
-/*==================*/
-	btr_pcur_t*	cursor, /*!< in: persistent cursor */
-	mtr_t*		mtr);	/*!< in: mtr */
 /*********************************************************//**
 Gets the rel_pos field for a cursor whose position has been stored.
 @return	BTR_PCUR_ON, ... */
@@ -282,10 +270,9 @@ btr_pcur_get_mtr(
 	btr_pcur_t*	cursor);	/*!< in: persistent cursor */
 /**************************************************************//**
 Commits the mtr and sets the pcur latch mode to BTR_NO_LATCHES,
-that is, the cursor becomes detached. If there have been modifications
-to the page where pcur is positioned, this can be used instead of
-btr_pcur_release_leaf. Function btr_pcur_store_position should be used
-before calling this, if restoration of cursor is wanted later. */
+that is, the cursor becomes detached.
+Function btr_pcur_store_position should be used before calling this,
+if restoration of cursor is wanted later. */
 UNIV_INLINE
 void
 btr_pcur_commit_specify_mtr(
diff --git a/storage/xtradb/include/btr0pcur.ic b/storage/xtradb/include/btr0pcur.ic
index 0c38797e6c5..0f9b969e7c5 100644
--- a/storage/xtradb/include/btr0pcur.ic
+++ b/storage/xtradb/include/btr0pcur.ic
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-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
@@ -396,10 +396,9 @@ btr_pcur_move_to_next(
 
 /**************************************************************//**
 Commits the mtr and sets the pcur latch mode to BTR_NO_LATCHES,
-that is, the cursor becomes detached. If there have been modifications
-to the page where pcur is positioned, this can be used instead of
-btr_pcur_release_leaf. Function btr_pcur_store_position should be used
-before calling this, if restoration of cursor is wanted later. */
+that is, the cursor becomes detached.
+Function btr_pcur_store_position should be used before calling this,
+if restoration of cursor is wanted later. */
 UNIV_INLINE
 void
 btr_pcur_commit_specify_mtr(
diff --git a/storage/xtradb/include/btr0types.h b/storage/xtradb/include/btr0types.h
index ef4a6b04b34..07c06fb18d7 100644
--- a/storage/xtradb/include/btr0types.h
+++ b/storage/xtradb/include/btr0types.h
@@ -38,6 +38,131 @@ 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;
 
+#ifdef UNIV_BLOB_DEBUG
+# include "buf0types.h"
+/** An index->blobs entry for keeping track of off-page column references */
+typedef struct btr_blob_dbg_struct btr_blob_dbg_t;
+
+/** Insert to index->blobs a reference to an off-page column.
+@param index	the index tree
+@param b	the reference
+@param ctx	context (for logging) */
+UNIV_INTERN
+void
+btr_blob_dbg_rbt_insert(
+/*====================*/
+	dict_index_t*		index,	/*!< in/out: index tree */
+	const btr_blob_dbg_t*	b,	/*!< in: the reference */
+	const char*		ctx)	/*!< in: context (for logging) */
+	__attribute__((nonnull));
+
+/** Remove from index->blobs a reference to an off-page column.
+@param index	the index tree
+@param b	the reference
+@param ctx	context (for logging) */
+UNIV_INTERN
+void
+btr_blob_dbg_rbt_delete(
+/*====================*/
+	dict_index_t*		index,	/*!< in/out: index tree */
+	const btr_blob_dbg_t*	b,	/*!< in: the reference */
+	const char*		ctx)	/*!< in: context (for logging) */
+	__attribute__((nonnull));
+
+/**************************************************************//**
+Add to index->blobs any references to off-page columns from a record.
+@return number of references added */
+UNIV_INTERN
+ulint
+btr_blob_dbg_add_rec(
+/*=================*/
+	const rec_t*	rec,	/*!< in: record */
+	dict_index_t*	index,	/*!< in/out: index */
+	const ulint*	offsets,/*!< in: offsets */
+	const char*	ctx)	/*!< in: context (for logging) */
+	__attribute__((nonnull));
+/**************************************************************//**
+Remove from index->blobs any references to off-page columns from a record.
+@return number of references removed */
+UNIV_INTERN
+ulint
+btr_blob_dbg_remove_rec(
+/*====================*/
+	const rec_t*	rec,	/*!< in: record */
+	dict_index_t*	index,	/*!< in/out: index */
+	const ulint*	offsets,/*!< in: offsets */
+	const char*	ctx)	/*!< in: context (for logging) */
+	__attribute__((nonnull));
+/**************************************************************//**
+Count and add to index->blobs any references to off-page columns
+from records on a page.
+@return number of references added */
+UNIV_INTERN
+ulint
+btr_blob_dbg_add(
+/*=============*/
+	const page_t*	page,	/*!< in: rewritten page */
+	dict_index_t*	index,	/*!< in/out: index */
+	const char*	ctx)	/*!< in: context (for logging) */
+	__attribute__((nonnull));
+/**************************************************************//**
+Count and remove from index->blobs any references to off-page columns
+from records on a page.
+Used when reorganizing a page, before copying the records.
+@return number of references removed */
+UNIV_INTERN
+ulint
+btr_blob_dbg_remove(
+/*================*/
+	const page_t*	page,	/*!< in: b-tree page */
+	dict_index_t*	index,	/*!< in/out: index */
+	const char*	ctx)	/*!< in: context (for logging) */
+	__attribute__((nonnull));
+/**************************************************************//**
+Restore in index->blobs any references to off-page columns
+Used when page reorganize fails due to compressed page overflow. */
+UNIV_INTERN
+void
+btr_blob_dbg_restore(
+/*=================*/
+	const page_t*	npage,	/*!< in: page that failed to compress */
+	const page_t*	page,	/*!< in: copy of original page */
+	dict_index_t*	index,	/*!< in/out: index */
+	const char*	ctx)	/*!< in: context (for logging) */
+	__attribute__((nonnull));
+
+/** Operation that processes the BLOB references of an index record
+@param[in]	rec	record on index page
+@param[in/out]	index	the index tree of the record
+@param[in]	offsets	rec_get_offsets(rec,index)
+@param[in]	ctx	context (for logging)
+@return			number of BLOB references processed */
+typedef ulint (*btr_blob_dbg_op_f)
+(const rec_t* rec,dict_index_t* index,const ulint* offsets,const char* ctx);
+
+/**************************************************************//**
+Count and process all references to off-page columns on a page.
+@return number of references processed */
+UNIV_INTERN
+ulint
+btr_blob_dbg_op(
+/*============*/
+	const page_t*		page,	/*!< in: B-tree leaf page */
+	const rec_t*		rec,	/*!< in: record to start from
+					(NULL to process the whole page) */
+	dict_index_t*		index,	/*!< in/out: index */
+	const char*		ctx,	/*!< in: context (for logging) */
+	const btr_blob_dbg_op_f	op)	/*!< in: operation on records */
+	__attribute__((nonnull(1,3,4,5)));
+#else /* UNIV_BLOB_DEBUG */
+# define btr_blob_dbg_add_rec(rec, index, offsets, ctx)		((void) 0)
+# define btr_blob_dbg_add(page, index, ctx)			((void) 0)
+# define btr_blob_dbg_remove_rec(rec, index, offsets, ctx)	((void) 0)
+# define btr_blob_dbg_remove(page, index, ctx)			((void) 0)
+# define btr_blob_dbg_restore(npage, page, index, ctx)		((void) 0)
+# define btr_blob_dbg_op(page, rec, index, ctx, op)		((void) 0)
+#endif /* UNIV_BLOB_DEBUG */
+
 /** The size of a reference to data stored on a different page.
 The reference is stored at the end of the prefix of the field
 in the index record. */
diff --git a/storage/xtradb/include/buf0buddy.h b/storage/xtradb/include/buf0buddy.h
index 3a35f8e46e9..f4c3da8692d 100644
--- a/storage/xtradb/include/buf0buddy.h
+++ b/storage/xtradb/include/buf0buddy.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 2006, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 2006, 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
@@ -37,25 +37,20 @@ Created December 2006 by Marko Makela
 /**********************************************************************//**
 Allocate a block.  The thread calling this function must hold
 buf_pool_mutex and must not hold buf_pool_zip_mutex or any
-block->mutex.  The buf_pool_mutex may only be released and reacquired
-if lru != NULL.  This function should only be used for allocating
-compressed page frames or control blocks (buf_page_t).  Allocated
-control blocks must be properly initialized immediately after
-buf_buddy_alloc() has returned the memory, before releasing
-buf_pool_mutex.
-@return	allocated block, possibly NULL if lru == NULL */
+block->mutex.  The buf_pool_mutex may be released and reacquired.
+This function should only be used for allocating compressed page frames.
+@return	allocated block, never NULL */
 UNIV_INLINE
 void*
 buf_buddy_alloc(
 /*============*/
-	ulint	size,	/*!< in: block size, up to UNIV_PAGE_SIZE */
+	ulint	size,	/*!< in: compressed page size
+			(between PAGE_ZIP_MIN_SIZE and UNIV_PAGE_SIZE) */
 	ibool*	lru,	/*!< in: pointer to a variable that will be assigned
 			TRUE if storage was allocated from the LRU list
-			and buf_pool_mutex was temporarily released,
-			or NULL if the LRU list should not be used */
+			and buf_pool_mutex was temporarily released */
 	ibool	have_page_hash_mutex)
-	__attribute__((malloc));
-
+	__attribute__((malloc, nonnull));
 /**********************************************************************//**
 Release a block. */
 UNIV_INLINE
diff --git a/storage/xtradb/include/buf0buddy.ic b/storage/xtradb/include/buf0buddy.ic
index 69659fb69d6..63241311e1f 100644
--- a/storage/xtradb/include/buf0buddy.ic
+++ b/storage/xtradb/include/buf0buddy.ic
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 2006, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 2006, 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
@@ -36,8 +36,8 @@ Created December 2006 by Marko Makela
 /**********************************************************************//**
 Allocate a block.  The thread calling this function must hold
 buf_pool_mutex and must not hold buf_pool_zip_mutex or any block->mutex.
-The buf_pool_mutex may only be released and reacquired if lru != NULL.
-@return	allocated block, possibly NULL if lru==NULL */
+The buf_pool_mutex may be released and reacquired.
+@return	allocated block, never NULL */
 UNIV_INTERN
 void*
 buf_buddy_alloc_low(
@@ -46,10 +46,9 @@ buf_buddy_alloc_low(
 			or BUF_BUDDY_SIZES */
 	ibool*	lru,	/*!< in: pointer to a variable that will be assigned
 			TRUE if storage was allocated from the LRU list
-			and buf_pool_mutex was temporarily released,
-			or NULL if the LRU list should not be used */
+			and buf_pool_mutex was temporarily released */
 	ibool	have_page_hash_mutex)
-	__attribute__((malloc));
+	__attribute__((malloc, nonnull));
 
 /**********************************************************************//**
 Deallocate a block. */
@@ -76,6 +75,8 @@ buf_buddy_get_slot(
 	ulint	i;
 	ulint	s;
 
+	ut_ad(size >= PAGE_ZIP_MIN_SIZE);
+
 	for (i = 0, s = BUF_BUDDY_LOW; s < size; i++, s <<= 1) {
 	}
 
@@ -86,27 +87,26 @@ buf_buddy_get_slot(
 /**********************************************************************//**
 Allocate a block.  The thread calling this function must hold
 buf_pool_mutex and must not hold buf_pool_zip_mutex or any
-block->mutex.  The buf_pool_mutex may only be released and reacquired
-if lru != NULL.  This function should only be used for allocating
-compressed page frames or control blocks (buf_page_t).  Allocated
-control blocks must be properly initialized immediately after
-buf_buddy_alloc() has returned the memory, before releasing
-buf_pool_mutex.
-@return	allocated block, possibly NULL if lru == NULL */
+block->mutex.  The buf_pool_mutex may be released and reacquired.
+This function should only be used for allocating compressed page frames.
+@return	allocated block, never NULL */
 UNIV_INLINE
 void*
 buf_buddy_alloc(
 /*============*/
-	ulint	size,	/*!< in: block size, up to UNIV_PAGE_SIZE */
+	ulint	size,	/*!< in: compressed page size
+			(between PAGE_ZIP_MIN_SIZE and UNIV_PAGE_SIZE) */
 	ibool*	lru,	/*!< in: pointer to a variable that will be assigned
 			TRUE if storage was allocated from the LRU list
-			and buf_pool_mutex was temporarily released,
-			or NULL if the LRU list should not be used */
+			and buf_pool_mutex was temporarily released */
 	ibool	have_page_hash_mutex)
 {
 	//ut_ad(buf_pool_mutex_own());
+	ut_ad(ut_is_2pow(size));
+	ut_ad(size >= PAGE_ZIP_MIN_SIZE);
+	ut_ad(size <= UNIV_PAGE_SIZE);
 
-	return(buf_buddy_alloc_low(buf_buddy_get_slot(size), lru, have_page_hash_mutex));
+	return((byte*) buf_buddy_alloc_low(buf_buddy_get_slot(size), lru, have_page_hash_mutex));
 }
 
 /**********************************************************************//**
@@ -121,6 +121,9 @@ buf_buddy_free(
 	ibool	have_page_hash_mutex)
 {
 	//ut_ad(buf_pool_mutex_own());
+	ut_ad(ut_is_2pow(size));
+	ut_ad(size >= PAGE_ZIP_MIN_SIZE);
+	ut_ad(size <= UNIV_PAGE_SIZE);
 
 	if (!have_page_hash_mutex) {
 		mutex_enter(&LRU_list_mutex);
diff --git a/storage/xtradb/include/buf0buf.h b/storage/xtradb/include/buf0buf.h
index bc0e9170281..838dd7f3900 100644
--- a/storage/xtradb/include/buf0buf.h
+++ b/storage/xtradb/include/buf0buf.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1995, 2010, 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
@@ -36,12 +36,13 @@ Created 11/5/1995 Heikki Tuuri
 #include "ut0rbt.h"
 #ifndef UNIV_HOTBACKUP
 #include "os0proc.h"
-#include "srv0srv.h"
 
 /** @name Modes for buf_page_get_gen */
 /* @{ */
 #define BUF_GET			10	/*!< get always */
 #define	BUF_GET_IF_IN_POOL	11	/*!< get if in pool */
+#define BUF_PEEK_IF_IN_POOL	12	/*!< get if in pool, do not make
+					the block young in the LRU list */
 #define BUF_GET_NO_LATCH	14	/*!< get and bufferfix, but
 					set no latch; we have
 					separated this case, because
@@ -140,12 +141,6 @@ buf_relocate(
 				BUF_BLOCK_ZIP_DIRTY or BUF_BLOCK_ZIP_PAGE */
 	buf_page_t*	dpage)	/*!< in/out: destination control block */
 	__attribute__((nonnull));
-/********************************************************************//**
-Resizes the buffer pool. */
-UNIV_INTERN
-void
-buf_pool_resize(void);
-/*=================*/
 /*********************************************************************//**
 Gets the current size of buffer buf_pool in bytes.
 @return	size in bytes */
@@ -162,6 +157,23 @@ ib_uint64_t
 buf_pool_get_oldest_modification(void);
 /*==================================*/
 /********************************************************************//**
+Allocates a buf_page_t descriptor. This function must succeed. In case
+of failure we assert in this function. */
+UNIV_INLINE
+buf_page_t*
+buf_page_alloc_descriptor(void)
+/*===========================*/
+	__attribute__((malloc));
+/********************************************************************//**
+Free a buf_page_t descriptor. */
+UNIV_INLINE
+void
+buf_page_free_descriptor(
+/*=====================*/
+	buf_page_t*	bpage)	/*!< in: bpage descriptor to free. */
+	__attribute__((nonnull));
+
+/********************************************************************//**
 Allocates a buffer block.
 @return	own: the allocated block, in state BUF_BLOCK_MEMORY */
 UNIV_INLINE
@@ -285,7 +297,7 @@ buf_page_get_gen(
 	ulint		rw_latch,/*!< in: RW_S_LATCH, RW_X_LATCH, RW_NO_LATCH */
 	buf_block_t*	guess,	/*!< in: guessed block or NULL */
 	ulint		mode,	/*!< in: BUF_GET, BUF_GET_IF_IN_POOL,
-				BUF_GET_NO_LATCH */
+				BUF_PEEK_IF_IN_POOL, BUF_GET_NO_LATCH */
 	const char*	file,	/*!< in: file name */
 	ulint		line,	/*!< in: line where called */
 	mtr_t*		mtr);	/*!< in: mini-transaction */
@@ -415,6 +427,18 @@ buf_block_get_freed_page_clock(
 	__attribute__((pure));
 
 /********************************************************************//**
+Tells if a block is still close enough to the MRU end of the LRU list
+meaning that it is not in danger of getting evicted and also implying
+that it has been accessed recently.
+Note that this is for heuristics only and does not reserve buffer pool
+mutex.
+@return	TRUE if block is close to MRU end of LRU */
+UNIV_INLINE
+ibool
+buf_page_peek_if_young(
+/*===================*/
+	const buf_page_t*	bpage);	/*!< in: block */
+/********************************************************************//**
 Recommends a move of a block to the start of the LRU list if there is danger
 of dropping from the buffer pool. NOTE: does not reserve the buffer pool
 mutex.
@@ -466,6 +490,31 @@ buf_block_get_modify_clock(
 #else /* !UNIV_HOTBACKUP */
 # define buf_block_modify_clock_inc(block) ((void) 0)
 #endif /* !UNIV_HOTBACKUP */
+/*******************************************************************//**
+Increments the bufferfix count. */
+UNIV_INLINE
+void
+buf_block_buf_fix_inc_func(
+/*=======================*/
+#ifdef UNIV_SYNC_DEBUG
+	const char*	file,	/*!< in: file name */
+	ulint		line,	/*!< in: line */
+#endif /* UNIV_SYNC_DEBUG */
+	buf_block_t*	block)	/*!< in/out: block to bufferfix */
+	__attribute__((nonnull));
+#ifdef UNIV_SYNC_DEBUG
+/** Increments the bufferfix count.
+@param b	in/out: block to bufferfix
+@param f	in: file name where requested
+@param l	in: line number where requested */
+# define buf_block_buf_fix_inc(b,f,l) buf_block_buf_fix_inc_func(f,l,b)
+#else /* UNIV_SYNC_DEBUG */
+/** Increments the bufferfix count.
+@param b	in/out: block to bufferfix
+@param f	in: file name where requested
+@param l	in: line number where requested */
+# define buf_block_buf_fix_inc(b,f,l) buf_block_buf_fix_inc_func(b)
+#endif /* UNIV_SYNC_DEBUG */
 /********************************************************************//**
 Calculates a page checksum which is stored to the page when it is written
 to a file. Note that we must be careful to calculate the same value
@@ -986,8 +1035,7 @@ UNIV_INTERN
 void
 buf_page_io_complete(
 /*=================*/
-	buf_page_t*	bpage,	/*!< in: pointer to the block in question */
-	trx_t*		trx);
+	buf_page_t*	bpage);	/*!< in: pointer to the block in question */
 /********************************************************************//**
 Calculates a folded value of a file page address to use in the page hash
 table.
@@ -1301,10 +1349,7 @@ struct buf_block_struct{
 /**********************************************************************//**
 Compute the hash fold value for blocks in buf_pool->zip_hash. */
 /* @{ */
-/* the fold should be relative when srv_buffer_pool_shm_key is enabled */
-#define BUF_POOL_ZIP_FOLD_PTR(ptr) (!srv_buffer_pool_shm_key\
-					?((ulint) (ptr) / UNIV_PAGE_SIZE)\
-					:((ulint) ((byte*)ptr - (byte*)(buf_pool->chunks->blocks->frame)) / UNIV_PAGE_SIZE))
+#define BUF_POOL_ZIP_FOLD_PTR(ptr) ((ulint) (ptr) / UNIV_PAGE_SIZE)
 #define BUF_POOL_ZIP_FOLD(b) BUF_POOL_ZIP_FOLD_PTR((b)->frame)
 #define BUF_POOL_ZIP_FOLD_BPAGE(b) BUF_POOL_ZIP_FOLD((buf_block_t*) (b))
 /* @} */
@@ -1330,6 +1375,8 @@ struct buf_pool_stat_struct{
 	ulint	n_pages_written;/*!< number write operations */
 	ulint	n_pages_created;/*!< number of pages created
 				in the pool with no read */
+	ulint	n_ra_pages_read_rnd;/*!< number of pages read in
+				as part of random read ahead */
 	ulint	n_ra_pages_read;/*!< number of pages read in
 				as part of read ahead */
 	ulint	n_ra_pages_evicted;/*!< number of read ahead
@@ -1453,8 +1500,10 @@ struct buf_pool_struct{
 	frames and buf_page_t descriptors of blocks that exist
 	in the buffer pool only in compressed form. */
 	/* @{ */
+#if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG
 	UT_LIST_BASE_NODE_T(buf_page_t)	zip_clean;
 					/*!< unmodified compressed pages */
+#endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */
 	UT_LIST_BASE_NODE_T(buf_page_t) zip_free[BUF_BUDDY_SIZES_MAX];
 					/*!< buddy free lists */
 //#if BUF_BUDDY_HIGH != UNIV_PAGE_SIZE
@@ -1486,8 +1535,8 @@ Use these instead of accessing buf_pool_mutex directly. */
 /** Test if buf_pool_mutex is owned. */
 #define buf_pool_mutex_own() mutex_own(&buf_pool_mutex)
 /** Acquire the buffer pool mutex. */
+/* the buf_pool_mutex is changed the latch order */
 #define buf_pool_mutex_enter() do {		\
-	ut_ad(!mutex_own(&buf_pool_zip_mutex));	\
 	mutex_enter(&buf_pool_mutex);		\
 } while (0)
 
diff --git a/storage/xtradb/include/buf0buf.ic b/storage/xtradb/include/buf0buf.ic
index 2cb0d8ef497..a081d6a34c0 100644
--- a/storage/xtradb/include/buf0buf.ic
+++ b/storage/xtradb/include/buf0buf.ic
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1995, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2011, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -62,6 +62,27 @@ buf_block_get_freed_page_clock(
 }
 
 /********************************************************************//**
+Tells if a block is still close enough to the MRU end of the LRU list
+meaning that it is not in danger of getting evicted and also implying
+that it has been accessed recently.
+Note that this is for heuristics only and does not reserve buffer pool
+mutex.
+@return	TRUE if block is close to MRU end of LRU */
+UNIV_INLINE
+ibool
+buf_page_peek_if_young(
+/*===================*/
+	const buf_page_t*	bpage)	/*!< in: block */
+{
+	/* FIXME: bpage->freed_page_clock is 31 bits */
+	return((buf_pool->freed_page_clock & ((1UL << 31) - 1))
+	       < ((ulint) bpage->freed_page_clock
+		  + (buf_pool->curr_size
+		     * (BUF_LRU_OLD_RATIO_DIV - buf_LRU_old_ratio)
+		     / (BUF_LRU_OLD_RATIO_DIV * 4))));
+}
+
+/********************************************************************//**
 Recommends a move of a block to the start of the LRU list if there is danger
 of dropping from the buffer pool. NOTE: does not reserve the buffer pool
 mutex.
@@ -89,12 +110,7 @@ buf_page_peek_if_too_old(
 		buf_pool->stat.n_pages_not_made_young++;
 		return(FALSE);
 	} else {
-		/* FIXME: bpage->freed_page_clock is 31 bits */
-		return((buf_pool->freed_page_clock & ((1UL << 31) - 1))
-		       > ((ulint) bpage->freed_page_clock
-			  + (buf_pool->curr_size
-			     * (BUF_LRU_OLD_RATIO_DIV - buf_LRU_old_ratio)
-			     / (BUF_LRU_OLD_RATIO_DIV * 4))));
+		return(!buf_page_peek_if_young(bpage));
 	}
 }
 
@@ -754,6 +770,35 @@ buf_block_get_lock_hash_val(
 }
 
 /********************************************************************//**
+Allocates a buf_page_t descriptor. This function must succeed. In case
+of failure we assert in this function.
+@return: the allocated descriptor. */
+UNIV_INLINE
+buf_page_t*
+buf_page_alloc_descriptor(void)
+/*===========================*/
+{
+	buf_page_t*	bpage;
+
+	bpage = (buf_page_t*) ut_malloc(sizeof *bpage);
+	ut_d(memset(bpage, 0, sizeof *bpage));
+	UNIV_MEM_ALLOC(bpage, sizeof *bpage);
+
+	return(bpage);
+}
+
+/********************************************************************//**
+Free a buf_page_t descriptor. */
+UNIV_INLINE
+void
+buf_page_free_descriptor(
+/*=====================*/
+	buf_page_t*	bpage)	/*!< in: bpage descriptor to free. */
+{
+	ut_free(bpage);
+}
+
+/********************************************************************//**
 Allocates a buffer block.
 @return	own: the allocated block, in state BUF_BLOCK_MEMORY */
 UNIV_INLINE
@@ -910,19 +955,6 @@ buf_block_buf_fix_inc_func(
 
 	block->page.buf_fix_count++;
 }
-#ifdef UNIV_SYNC_DEBUG
-/** Increments the bufferfix count.
-@param b	in/out: block to bufferfix
-@param f	in: file name where requested
-@param l	in: line number where requested */
-# define buf_block_buf_fix_inc(b,f,l) buf_block_buf_fix_inc_func(f,l,b)
-#else /* UNIV_SYNC_DEBUG */
-/** Increments the bufferfix count.
-@param b	in/out: block to bufferfix
-@param f	in: file name where requested
-@param l	in: line number where requested */
-# define buf_block_buf_fix_inc(b,f,l) buf_block_buf_fix_inc_func(b)
-#endif /* UNIV_SYNC_DEBUG */
 
 /*******************************************************************//**
 Decrements the bufferfix count. */
@@ -1119,7 +1151,7 @@ buf_block_dbg_add_level(
 				where we have acquired latch */
 	ulint		level)	/*!< in: latching order level */
 {
-	sync_thread_add_level(&block->lock, level);
+	sync_thread_add_level(&block->lock, level, FALSE);
 }
 #endif /* UNIV_SYNC_DEBUG */
 #endif /* !UNIV_HOTBACKUP */
diff --git a/storage/xtradb/include/buf0lru.h b/storage/xtradb/include/buf0lru.h
index fe7c067dfb7..8abebfb675c 100644
--- a/storage/xtradb/include/buf0lru.h
+++ b/storage/xtradb/include/buf0lru.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
@@ -30,18 +30,6 @@ Created 11/5/1995 Heikki Tuuri
 #include "ut0byte.h"
 #include "buf0types.h"
 
-/** The return type of buf_LRU_free_block() */
-enum buf_lru_free_block_status {
-	/** freed */
-	BUF_LRU_FREED = 0,
-	/** not freed because the caller asked to remove the
-	uncompressed frame but the control block cannot be
-	relocated */
-	BUF_LRU_CANNOT_RELOCATE,
-	/** not freed because of some other reason */
-	BUF_LRU_NOT_FREED
-};
-
 /******************************************************************//**
 Tries to remove LRU flushed blocks from the end of the LRU list and put them
 to the free list. This is beneficial for the efficiency of the insert buffer
@@ -91,6 +79,7 @@ void
 buf_LRU_mark_space_was_deleted(
 /*===========================*/
 	ulint	id);	/*!< in: space id */
+#if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG
 /********************************************************************//**
 Insert a compressed block into buf_pool->zip_clean in the LRU order. */
 UNIV_INTERN
@@ -98,22 +87,22 @@ void
 buf_LRU_insert_zip_clean(
 /*=====================*/
 	buf_page_t*	bpage);	/*!< in: pointer to the block in question */
+#endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */
 
 /******************************************************************//**
 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 BUF_LRU_FREED, it will temporarily
+NOTE: If this function returns TRUE, it 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.
-@return BUF_LRU_FREED if freed, BUF_LRU_CANNOT_RELOCATE or
-BUF_LRU_NOT_FREED otherwise. */
+@return TRUE if freed, FALSE otherwise. */
 UNIV_INTERN
-enum buf_lru_free_block_status
+ibool
 buf_LRU_free_block(
 /*===============*/
 	buf_page_t*	bpage,	/*!< in: block to be freed */
diff --git a/storage/xtradb/include/buf0types.h b/storage/xtradb/include/buf0types.h
index ce3e5ecc9c5..9dd847bdaca 100644
--- a/storage/xtradb/include/buf0types.h
+++ b/storage/xtradb/include/buf0types.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
@@ -26,6 +26,8 @@ Created 11/17/1995 Heikki Tuuri
 #ifndef buf0types_h
 #define buf0types_h
 
+#include "page0types.h"
+
 /** Buffer page (uncompressed or compressed) */
 typedef	struct buf_page_struct		buf_page_t;
 /** Buffer block for which an uncompressed page exists */
@@ -58,17 +60,10 @@ enum buf_io_fix {
 
 /** Parameters of binary buddy system for compressed pages (buf0buddy.h) */
 /* @{ */
-#if UNIV_WORD_SIZE <= 4 /* 32-bit system */
-/** Base-2 logarithm of the smallest buddy block size */
-# define BUF_BUDDY_LOW_SHIFT	6
-#else /* 64-bit system */
-/** Base-2 logarithm of the smallest buddy block size */
-# define BUF_BUDDY_LOW_SHIFT	7
-#endif
+#define BUF_BUDDY_LOW_SHIFT	PAGE_ZIP_MIN_SIZE_SHIFT
+
 #define BUF_BUDDY_LOW		(1 << BUF_BUDDY_LOW_SHIFT)
-					/*!< minimum block size in the binary
-					buddy system; must be at least
-					sizeof(buf_page_t) */
+
 #define BUF_BUDDY_SIZES		(UNIV_PAGE_SIZE_SHIFT - BUF_BUDDY_LOW_SHIFT)
 #define BUF_BUDDY_SIZES_MAX	(UNIV_PAGE_SIZE_SHIFT_MAX - BUF_BUDDY_LOW_SHIFT)
 					/*!< number of buddy sizes */
diff --git a/storage/xtradb/include/dict0mem.h b/storage/xtradb/include/dict0mem.h
index f47293bedf6..3554274847c 100644
--- a/storage/xtradb/include/dict0mem.h
+++ b/storage/xtradb/include/dict0mem.h
@@ -340,6 +340,13 @@ struct dict_index_struct{
 				index, or 0 if the index existed
 				when InnoDB was started up */
 #endif /* !UNIV_HOTBACKUP */
+#ifdef UNIV_BLOB_DEBUG
+	mutex_t		blobs_mutex;
+				/*!< mutex protecting blobs */
+	void*		blobs;	/*!< map of (page_no,heap_no,field_no)
+				to first_blob_page_no; protected by
+				blobs_mutex; @see btr_blob_dbg_t */
+#endif /* UNIV_BLOB_DEBUG */
 #ifdef UNIV_DEBUG
 	ulint		magic_n;/*!< magic number */
 /** Value of dict_index_struct::magic_n */
diff --git a/storage/xtradb/include/fil0fil.h b/storage/xtradb/include/fil0fil.h
index fbf8ca20db3..11c4cb4ba03 100644
--- a/storage/xtradb/include/fil0fil.h
+++ b/storage/xtradb/include/fil0fil.h
@@ -670,8 +670,9 @@ UNIV_INTERN
 void
 fil_flush(
 /*======*/
-	ulint	space_id);	/*!< in: file space id (this can be a group of
+	ulint	space_id,	/*!< in: file space id (this can be a group of
 				log files or a tablespace of the database) */
+	ibool	metadata);
 /**********************************************************************//**
 Flushes to disk writes in file spaces of the given type possibly cached by
 the OS. */
diff --git a/storage/xtradb/include/hash0hash.h b/storage/xtradb/include/hash0hash.h
index 492c767acc4..b17c21a45ef 100644
--- a/storage/xtradb/include/hash0hash.h
+++ b/storage/xtradb/include/hash0hash.h
@@ -49,28 +49,6 @@ hash_table_t*
 hash_create(
 /*========*/
 	ulint	n);	/*!< in: number of array cells */
-
-/*************************************************************//**
-*/
-UNIV_INTERN
-ulint
-hash_create_needed(
-/*===============*/
-	ulint	n);
-
-UNIV_INTERN
-void
-hash_create_init(
-/*=============*/
-	hash_table_t*	table,
-	ulint		n);
-
-UNIV_INTERN
-void
-hash_create_reuse(
-/*==============*/
-	hash_table_t*	table);
-
 #ifndef UNIV_HOTBACKUP
 /*************************************************************//**
 Creates a mutex array to protect a hash table. */
@@ -350,33 +328,6 @@ do {\
 	}\
 } while (0)
 
-/********************************************************************//**
-Align nodes with moving location.*/
-#define HASH_OFFSET(TABLE, NODE_TYPE, PTR_NAME, FADDR, FOFFSET, BOFFSET) \
-do {\
-	ulint		i2222;\
-	ulint		cell_count2222;\
-\
-	cell_count2222 = hash_get_n_cells(TABLE);\
-\
-	for (i2222 = 0; i2222 < cell_count2222; i2222++) {\
-		NODE_TYPE*	node2222;\
-\
-		if ((TABLE)->array[i2222].node) \
-			(TABLE)->array[i2222].node = (void*)((byte*)(TABLE)->array[i2222].node \
-			+ (((TABLE)->array[i2222].node > (void*)FADDR)?FOFFSET:BOFFSET));\
-		node2222 = HASH_GET_FIRST((TABLE), i2222);\
-\
-		while (node2222) {\
-			if (node2222->PTR_NAME) \
-				node2222->PTR_NAME = (void*)((byte*)(node2222->PTR_NAME) \
-				+ ((((void*)node2222->PTR_NAME) > (void*)FADDR)?FOFFSET:BOFFSET));\
-\
-			node2222 = node2222->PTR_NAME;\
-		}\
-	}\
-} while (0)
-
 /************************************************************//**
 Gets the mutex index for a fold value in a hash table.
 @return	mutex number */
diff --git a/storage/xtradb/include/mtr0mtr.h b/storage/xtradb/include/mtr0mtr.h
index bc3f1951be9..8a9ec8ea7f0 100644
--- a/storage/xtradb/include/mtr0mtr.h
+++ b/storage/xtradb/include/mtr0mtr.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
@@ -213,16 +213,6 @@ ulint
 mtr_set_savepoint(
 /*==============*/
 	mtr_t*	mtr);	/*!< in: mtr */
-/**********************************************************//**
-Releases the latches stored in an mtr memo down to a savepoint.
-NOTE! The mtr must not have made changes to buffer pages after the
-savepoint, as these can be handled only by mtr_commit. */
-UNIV_INTERN
-void
-mtr_rollback_to_savepoint(
-/*======================*/
-	mtr_t*	mtr,		/*!< in: mtr */
-	ulint	savepoint);	/*!< in: savepoint */
 #ifndef UNIV_HOTBACKUP
 /**********************************************************//**
 Releases the (index tree) s-latch stored in an mtr memo after a
diff --git a/storage/xtradb/include/os0file.h b/storage/xtradb/include/os0file.h
index 732e930517b..98cab5ef874 100644
--- a/storage/xtradb/include/os0file.h
+++ b/storage/xtradb/include/os0file.h
@@ -469,7 +469,8 @@ UNIV_INTERN
 ibool
 os_file_flush(
 /*==========*/
-	os_file_t	file);	/*!< in, own: handle to a file */
+	os_file_t	file,	/*!< in, own: handle to a file */
+	ibool		metadata);
 /***********************************************************************//**
 Retrieves the last error number if an error occurs in a file io function.
 The number should be retrieved before any other OS calls (because they may
diff --git a/storage/xtradb/include/os0proc.h b/storage/xtradb/include/os0proc.h
index 582cef6f803..fd46bd7db87 100644
--- a/storage/xtradb/include/os0proc.h
+++ b/storage/xtradb/include/os0proc.h
@@ -32,11 +32,6 @@ Created 9/30/1995 Heikki Tuuri
 #ifdef UNIV_LINUX
 #include <sys/ipc.h>
 #include <sys/shm.h>
-#else
-# if defined HAVE_SYS_IPC_H && HAVE_SYS_SHM_H
-#include <sys/ipc.h>
-#include <sys/shm.h>
-# endif
 #endif
 
 typedef void*			os_process_t;
@@ -75,29 +70,6 @@ os_mem_free_large(
 	ulint	size);			/*!< in: size returned by
 					os_mem_alloc_large() */
 
-
-/****************************************************************//**
-Allocates or attaches and reuses shared memory segment.
-The content is not cleared automatically.
-@return	allocated memory */
-UNIV_INTERN
-void*
-os_shm_alloc(
-/*=========*/
-	ulint*	n,			/*!< in/out: number of bytes */
-	uint	key,
-	ibool*	is_new);
-
-/****************************************************************//**
-Detach shared memory segment. */
-UNIV_INTERN
-void
-os_shm_free(
-/*========*/
-	void	*ptr,			/*!< in: pointer returned by
-					os_shm_alloc() */
-	ulint	size);			/*!< in: size returned by
-					os_shm_alloc() */
 #ifndef UNIV_NONINL
 #include "os0proc.ic"
 #endif
diff --git a/storage/xtradb/include/page0cur.ic b/storage/xtradb/include/page0cur.ic
index 3520677dfb3..81474fa35f5 100644
--- a/storage/xtradb/include/page0cur.ic
+++ b/storage/xtradb/include/page0cur.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
@@ -27,6 +27,8 @@ Created 10/4/1994 Heikki Tuuri
 #include "buf0types.h"
 
 #ifdef UNIV_DEBUG
+# include "rem0cmp.h"
+
 /*********************************************************//**
 Gets pointer to the page frame where the cursor is positioned.
 @return	page */
@@ -268,6 +270,7 @@ page_cur_tuple_insert(
 					      index, rec, offsets, mtr);
 	}
 
+	ut_ad(!rec || !cmp_dtuple_rec(tuple, rec, offsets));
 	mem_heap_free(heap);
 	return(rec);
 }
diff --git a/storage/xtradb/include/page0page.h b/storage/xtradb/include/page0page.h
index 5b2bcf7c054..aeaef030505 100644
--- a/storage/xtradb/include/page0page.h
+++ b/storage/xtradb/include/page0page.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
@@ -284,16 +284,42 @@ page_get_supremum_offset(
 	const page_t*	page);	/*!< in: page which must have record(s) */
 #define page_get_infimum_rec(page) ((page) + page_get_infimum_offset(page))
 #define page_get_supremum_rec(page) ((page) + page_get_supremum_offset(page))
+
 /************************************************************//**
-Returns the middle record of record list. If there are an even number
-of records in the list, returns the first record of upper half-list.
-@return	middle record */
+Returns the nth record of the record list.
+This is the inverse function of page_rec_get_n_recs_before().
+@return	nth record */
 UNIV_INTERN
+const rec_t*
+page_rec_get_nth_const(
+/*===================*/
+	const page_t*	page,	/*!< in: page */
+	ulint		nth)	/*!< in: nth record */
+	__attribute__((nonnull, warn_unused_result));
+/************************************************************//**
+Returns the nth record of the record list.
+This is the inverse function of page_rec_get_n_recs_before().
+@return	nth record */
+UNIV_INLINE
+rec_t*
+page_rec_get_nth(
+/*=============*/
+	page_t*	page,	/*< in: page */
+	ulint	nth)	/*!< in: nth record */
+	__attribute__((nonnull, warn_unused_result));
+
+#ifndef UNIV_HOTBACKUP
+/************************************************************//**
+Returns the middle record of the records on the page. If there is an
+even number of records in the list, returns the first record of the
+upper half-list.
+@return	middle record */
+UNIV_INLINE
 rec_t*
 page_get_middle_rec(
 /*================*/
-	page_t*	page);	/*!< in: page */
-#ifndef UNIV_HOTBACKUP
+	page_t*	page)	/*!< in: page */
+	__attribute__((nonnull, warn_unused_result));
 /*************************************************************//**
 Compares a data tuple to a physical record. Differs from the function
 cmp_dtuple_rec_with_match in the way that the record must reside on an
@@ -348,6 +374,7 @@ page_get_n_recs(
 /***************************************************************//**
 Returns the number of records before the given record in chain.
 The number includes infimum and supremum records.
+This is the inverse function of page_rec_get_nth().
 @return	number of records */
 UNIV_INTERN
 ulint
diff --git a/storage/xtradb/include/page0page.ic b/storage/xtradb/include/page0page.ic
index dab9dc742e4..b34408aed17 100644
--- a/storage/xtradb/include/page0page.ic
+++ b/storage/xtradb/include/page0page.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
@@ -423,7 +423,37 @@ page_rec_is_infimum(
 	return(page_rec_is_infimum_low(page_offset(rec)));
 }
 
+/************************************************************//**
+Returns the nth record of the record list.
+This is the inverse function of page_rec_get_n_recs_before().
+@return	nth record */
+UNIV_INLINE
+rec_t*
+page_rec_get_nth(
+/*=============*/
+	page_t*	page,	/*!< in: page */
+	ulint	nth)	/*!< in: nth record */
+{
+	return((rec_t*) page_rec_get_nth_const(page, nth));
+}
+
 #ifndef UNIV_HOTBACKUP
+/************************************************************//**
+Returns the middle record of the records on the page. If there is an
+even number of records in the list, returns the first record of the
+upper half-list.
+@return	middle record */
+UNIV_INLINE
+rec_t*
+page_get_middle_rec(
+/*================*/
+	page_t*	page)	/*!< in: page */
+{
+	ulint	middle = (page_get_n_recs(page) + PAGE_HEAP_NO_USER_LOW) / 2;
+
+	return(page_rec_get_nth(page, middle));
+}
+
 /*************************************************************//**
 Compares a data tuple to a physical record. Differs from the function
 cmp_dtuple_rec_with_match in the way that the record must reside on an
diff --git a/storage/xtradb/include/page0zip.h b/storage/xtradb/include/page0zip.h
index 4d37302ed20..fe3d2e52e0b 100644
--- a/storage/xtradb/include/page0zip.h
+++ b/storage/xtradb/include/page0zip.h
@@ -420,7 +420,7 @@ page_zip_copy_recs(
 	const page_t*		src,		/*!< in: page */
 	dict_index_t*		index,		/*!< in: index of the B-tree */
 	mtr_t*			mtr)		/*!< in: mini-transaction */
-	__attribute__((nonnull(1,2,3,4)));
+	__attribute__((nonnull));
 #endif /* !UNIV_HOTBACKUP */
 
 /**********************************************************************//**
diff --git a/storage/xtradb/include/rem0rec.h b/storage/xtradb/include/rem0rec.h
index 17d08afabb9..06de23be757 100644
--- a/storage/xtradb/include/rem0rec.h
+++ b/storage/xtradb/include/rem0rec.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
@@ -480,6 +480,18 @@ ulint
 rec_offs_any_extern(
 /*================*/
 	const ulint*	offsets);/*!< in: array returned by rec_get_offsets() */
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+/******************************************************//**
+Determine if the offsets are for a record containing null BLOB pointers.
+@return	first field containing a null BLOB pointer, or NULL if none found */
+UNIV_INLINE
+const byte*
+rec_offs_any_null_extern(
+/*=====================*/
+	const rec_t*	rec,		/*!< in: record */
+	const ulint*	offsets)	/*!< in: rec_get_offsets(rec) */
+	__attribute__((nonnull, warn_unused_result));
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
 /******************************************************//**
 Returns nonzero if the extern bit is set in nth field of rec.
 @return	nonzero if externally stored */
diff --git a/storage/xtradb/include/rem0rec.ic b/storage/xtradb/include/rem0rec.ic
index fa96c97f95e..3875a4cf814 100644
--- a/storage/xtradb/include/rem0rec.ic
+++ b/storage/xtradb/include/rem0rec.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
@@ -26,6 +26,7 @@ Created 5/30/1994 Heikki Tuuri
 #include "mach0data.h"
 #include "ut0byte.h"
 #include "dict0dict.h"
+#include "btr0types.h"
 
 /* Compact flag ORed to the extra size returned by rec_get_offsets() */
 #define REC_OFFS_COMPACT	((ulint) 1 << 31)
@@ -1087,6 +1088,44 @@ rec_offs_any_extern(
 	return(UNIV_UNLIKELY(*rec_offs_base(offsets) & REC_OFFS_EXTERNAL));
 }
 
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+/******************************************************//**
+Determine if the offsets are for a record containing null BLOB pointers.
+@return	first field containing a null BLOB pointer, or NULL if none found */
+UNIV_INLINE
+const byte*
+rec_offs_any_null_extern(
+/*=====================*/
+	const rec_t*	rec,		/*!< in: record */
+	const ulint*	offsets)	/*!< in: rec_get_offsets(rec) */
+{
+	ulint	i;
+	ut_ad(rec_offs_validate(rec, NULL, offsets));
+
+	if (!rec_offs_any_extern(offsets)) {
+		return(NULL);
+	}
+
+	for (i = 0; i < rec_offs_n_fields(offsets); i++) {
+		if (rec_offs_nth_extern(offsets, i)) {
+			ulint		len;
+			const byte*	field
+				= rec_get_nth_field(rec, offsets, i, &len);
+
+			ut_a(len >= BTR_EXTERN_FIELD_REF_SIZE);
+			if (!memcmp(field + len
+				    - BTR_EXTERN_FIELD_REF_SIZE,
+				    field_ref_zero,
+				    BTR_EXTERN_FIELD_REF_SIZE)) {
+				return(field);
+			}
+		}
+	}
+
+	return(NULL);
+}
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
+
 /******************************************************//**
 Returns nonzero if the extern bit is set in nth field of rec.
 @return	nonzero if externally stored */
diff --git a/storage/xtradb/include/row0row.h b/storage/xtradb/include/row0row.h
index 723b7b53395..36fb26482ce 100644
--- a/storage/xtradb/include/row0row.h
+++ b/storage/xtradb/include/row0row.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
@@ -38,16 +38,16 @@ Created 4/20/1996 Heikki Tuuri
 #include "btr0types.h"
 
 /*********************************************************************//**
-Gets the offset of the trx id field, in bytes relative to the origin of
+Gets the offset of the DB_TRX_ID field, in bytes relative to the origin of
 a clustered index record.
 @return	offset of DATA_TRX_ID */
-UNIV_INTERN
+UNIV_INLINE
 ulint
 row_get_trx_id_offset(
 /*==================*/
-	const rec_t*	rec,	/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets);/*!< in: rec_get_offsets(rec, index) */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: record offsets */
+	__attribute__((nonnull, warn_unused_result));
 /*********************************************************************//**
 Reads the trx id field from a clustered index record.
 @return	value of the field */
@@ -55,9 +55,10 @@ UNIV_INLINE
 trx_id_t
 row_get_rec_trx_id(
 /*===============*/
-	const rec_t*	rec,	/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets);/*!< in: rec_get_offsets(rec, index) */
+	const rec_t*		rec,	/*!< in: record */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
+	__attribute__((nonnull, warn_unused_result));
 /*********************************************************************//**
 Reads the roll pointer field from a clustered index record.
 @return	value of the field */
@@ -65,9 +66,10 @@ UNIV_INLINE
 roll_ptr_t
 row_get_rec_roll_ptr(
 /*=================*/
-	const rec_t*	rec,	/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets);/*!< in: rec_get_offsets(rec, index) */
+	const rec_t*		rec,	/*!< in: record */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
+	__attribute__((nonnull, warn_unused_result));
 /*****************************************************************//**
 When an insert or purge to a table is performed, this function builds
 the entry to be inserted into or purged from an index on the table.
diff --git a/storage/xtradb/include/row0row.ic b/storage/xtradb/include/row0row.ic
index 05c007641af..0b9ca982af8 100644
--- a/storage/xtradb/include/row0row.ic
+++ b/storage/xtradb/include/row0row.ic
@@ -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
@@ -28,15 +28,42 @@ Created 4/20/1996 Heikki Tuuri
 #include "trx0undo.h"
 
 /*********************************************************************//**
+Gets the offset of the DB_TRX_ID field, in bytes relative to the origin of
+a clustered index record.
+@return	offset of DATA_TRX_ID */
+UNIV_INLINE
+ulint
+row_get_trx_id_offset(
+/*==================*/
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: record offsets */
+{
+	ulint	pos;
+	ulint	offset;
+	ulint	len;
+
+	ut_ad(dict_index_is_clust(index));
+	ut_ad(rec_offs_validate(NULL, index, offsets));
+
+	pos = dict_index_get_sys_col_pos(index, DATA_TRX_ID);
+
+	offset = rec_get_nth_field_offs(offsets, pos, &len);
+
+	ut_ad(len == DATA_TRX_ID_LEN);
+
+	return(offset);
+}
+
+/*********************************************************************//**
 Reads the trx id field from a clustered index record.
 @return	value of the field */
 UNIV_INLINE
 trx_id_t
 row_get_rec_trx_id(
 /*===============*/
-	const rec_t*	rec,	/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets)/*!< in: rec_get_offsets(rec, index) */
+	const rec_t*		rec,	/*!< in: record */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
 {
 	ulint	offset;
 
@@ -46,7 +73,7 @@ row_get_rec_trx_id(
 	offset = index->trx_id_offset;
 
 	if (!offset) {
-		offset = row_get_trx_id_offset(rec, index, offsets);
+		offset = row_get_trx_id_offset(index, offsets);
 	}
 
 	return(trx_read_trx_id(rec + offset));
@@ -59,9 +86,9 @@ UNIV_INLINE
 roll_ptr_t
 row_get_rec_roll_ptr(
 /*=================*/
-	const rec_t*	rec,	/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets)/*!< in: rec_get_offsets(rec, index) */
+	const rec_t*		rec,	/*!< in: record */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
 {
 	ulint	offset;
 
@@ -71,7 +98,7 @@ row_get_rec_roll_ptr(
 	offset = index->trx_id_offset;
 
 	if (!offset) {
-		offset = row_get_trx_id_offset(rec, index, offsets);
+		offset = row_get_trx_id_offset(index, offsets);
 	}
 
 	return(trx_read_roll_ptr(rec + offset + DATA_TRX_ID_LEN));
diff --git a/storage/xtradb/include/row0upd.ic b/storage/xtradb/include/row0upd.ic
index 18e22f1eca9..0894ed373b0 100644
--- a/storage/xtradb/include/row0upd.ic
+++ b/storage/xtradb/include/row0upd.ic
@@ -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
@@ -171,7 +171,7 @@ row_upd_rec_sys_fields(
 		ulint	offset = index->trx_id_offset;
 
 		if (!offset) {
-			offset = row_get_trx_id_offset(rec, index, offsets);
+			offset = row_get_trx_id_offset(index, offsets);
 		}
 
 #if DATA_TRX_ID + 1 != DATA_ROLL_PTR
diff --git a/storage/xtradb/include/srv0srv.h b/storage/xtradb/include/srv0srv.h
index d4329d16a62..c120db5cbcc 100644
--- a/storage/xtradb/include/srv0srv.h
+++ b/storage/xtradb/include/srv0srv.h
@@ -159,13 +159,10 @@ extern ulint	srv_buf_pool_curr_size;	/*!< current size in bytes */
 extern ulint	srv_mem_pool_size;
 extern ulint	srv_lock_table_size;
 
-extern uint	srv_buffer_pool_shm_key;
-extern ibool	srv_buffer_pool_shm_is_reused;
-extern ibool	srv_buffer_pool_shm_checksum;
-
 extern ibool	srv_thread_concurrency_timer_based;
 
 extern ulint	srv_n_file_io_threads;
+extern my_bool	srv_random_read_ahead;
 extern ulong	srv_read_ahead_threshold;
 extern ulint	srv_n_read_io_threads;
 extern ulint	srv_n_write_io_threads;
@@ -288,6 +285,7 @@ extern	ibool	srv_print_latch_waits;
 extern ulint	srv_activity_count;
 extern ulint	srv_fatal_semaphore_wait_threshold;
 extern ulint	srv_dml_needed_delay;
+extern lint	srv_kill_idle_transaction;
 
 extern mutex_t*	kernel_mutex_temp;/* mutex protecting the server, trx structs,
 				query threads, and lock table: we allocate
@@ -351,6 +349,9 @@ extern ulint srv_buf_pool_reads;
 /** Time in seconds between automatic buffer pool dumps */
 extern uint srv_auto_lru_dump;
 
+/** Whether startup should be blocked until buffer pool is fully restored */
+extern ibool srv_blocking_lru_restore;
+
 /** Status variables to be passed to MySQL */
 typedef struct export_var_struct export_struc;
 
@@ -694,6 +695,7 @@ struct export_var_struct{
 	ulint innodb_buffer_pool_wait_free;	/*!< srv_buf_pool_wait_free */
 	ulint innodb_buffer_pool_pages_flushed;	/*!< srv_buf_pool_flushed */
 	ulint innodb_buffer_pool_write_requests;/*!< srv_buf_pool_write_requests */
+	ulint innodb_buffer_pool_read_ahead_rnd;/*!< srv_read_ahead_rnd */
 	ulint innodb_buffer_pool_read_ahead;	/*!< srv_read_ahead */
 	ulint innodb_buffer_pool_read_ahead_evicted;/*!< srv_read_ahead evicted*/
         ulint innodb_deadlocks;                 /* ??? */
diff --git a/storage/xtradb/include/sync0arr.h b/storage/xtradb/include/sync0arr.h
index 5f1280f5e28..6e931346238 100644
--- a/storage/xtradb/include/sync0arr.h
+++ b/storage/xtradb/include/sync0arr.h
@@ -115,8 +115,11 @@ Prints warnings of long semaphore waits to stderr.
 @return	TRUE if fatal semaphore wait threshold was exceeded */
 UNIV_INTERN
 ibool
-sync_array_print_long_waits(void);
-/*=============================*/
+sync_array_print_long_waits(
+/*========================*/
+	os_thread_id_t*	waiter,	/*!< out: longest waiting thread */
+	const void**	sema)	/*!< out: longest-waited-for semaphore */
+	__attribute__((nonnull));
 /********************************************************************//**
 Validates the integrity of the wait array. Checks
 that the number of reserved cells equals the count variable. */
diff --git a/storage/xtradb/include/sync0rw.ic b/storage/xtradb/include/sync0rw.ic
index 7116f1b7c9b..485a63a1b18 100644
--- a/storage/xtradb/include/sync0rw.ic
+++ b/storage/xtradb/include/sync0rw.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.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -603,16 +603,16 @@ rw_lock_x_unlock_direct(
 
 	ut_ad((lock->lock_word % X_LOCK_DECR) == 0);
 
-#ifdef UNIV_SYNC_DEBUG
-	rw_lock_remove_debug_info(lock, 0, RW_LOCK_EX);
-#endif
-
 	if (lock->lock_word == 0) {
 		lock->recursive = FALSE;
 		UNIV_MEM_INVALID(&lock->writer_thread,
 				 sizeof lock->writer_thread);
 	}
 
+#ifdef UNIV_SYNC_DEBUG
+	rw_lock_remove_debug_info(lock, 0, RW_LOCK_EX);
+#endif
+
 	lock->lock_word += X_LOCK_DECR;
 
 	ut_ad(!lock->waiters);
diff --git a/storage/xtradb/include/sync0sync.h b/storage/xtradb/include/sync0sync.h
index f2ff83101ab..8b9a075e875 100644
--- a/storage/xtradb/include/sync0sync.h
+++ b/storage/xtradb/include/sync0sync.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1995, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2011, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -219,8 +219,10 @@ void
 sync_thread_add_level(
 /*==================*/
 	void*	latch,	/*!< in: pointer to a mutex or an rw-lock */
-	ulint	level);	/*!< in: level in the latching order; if
+	ulint	level,	/*!< in: level in the latching order; if
 			SYNC_LEVEL_VARYING, nothing is done */
+	ibool	relock)	/*!< in: TRUE if re-entering an x-lock */
+	__attribute__((nonnull));
 /******************************************************************//**
 Removes a latch from the thread level array if it is found there.
 @return TRUE if found in the array; it is no error if the latch is
diff --git a/storage/xtradb/include/trx0sys.h b/storage/xtradb/include/trx0sys.h
index 2637189f37e..eafa1ab6409 100644
--- a/storage/xtradb/include/trx0sys.h
+++ b/storage/xtradb/include/trx0sys.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
@@ -317,6 +317,17 @@ ibool
 trx_in_trx_list(
 /*============*/
 	trx_t*	in_trx);/*!< in: trx */
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+/***********************************************************//**
+Assert that a transaction has been recovered.
+@return TRUE */
+UNIV_INLINE
+ibool
+trx_assert_recovered(
+/*=================*/
+	trx_id_t	trx_id)		/*!< in: transaction identifier */
+	__attribute__((warn_unused_result));
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
 /*****************************************************************//**
 Updates the offset information about the end of the MySQL binlog entry
 which corresponds to the transaction just being committed. In a MySQL
diff --git a/storage/xtradb/include/trx0sys.ic b/storage/xtradb/include/trx0sys.ic
index 5e0f07c8b9d..234fc0b92e9 100644
--- a/storage/xtradb/include/trx0sys.ic
+++ b/storage/xtradb/include/trx0sys.ic
@@ -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
@@ -311,6 +311,28 @@ trx_get_on_id(
 	return(NULL);
 }
 
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+/***********************************************************//**
+Assert that a transaction has been recovered.
+@return TRUE */
+UNIV_INLINE
+ibool
+trx_assert_recovered(
+/*=================*/
+	trx_id_t	trx_id)		/*!< in: transaction identifier */
+{
+	trx_t*		trx;
+
+	mutex_enter(&kernel_mutex);
+	trx = trx_get_on_id(trx_id);
+	ut_a(trx);
+	ut_a(trx->is_recovered);
+	mutex_exit(&kernel_mutex);
+
+	return(TRUE);
+}
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
+
 /****************************************************************//**
 Returns the minumum trx id in trx list. This is the smallest id for which
 the trx can possibly be active. (But, you must look at the trx->conc_state to
diff --git a/storage/xtradb/include/trx0trx.h b/storage/xtradb/include/trx0trx.h
index 173c63918d3..700476bcd85 100644
--- a/storage/xtradb/include/trx0trx.h
+++ b/storage/xtradb/include/trx0trx.h
@@ -44,6 +44,9 @@ extern sess_t*	trx_dummy_sess;
 /** Number of transactions currently allocated for MySQL: protected by
 the kernel mutex */
 extern ulint	trx_n_mysql_transactions;
+/** Number of transactions currently in the XA PREPARED state: protected by
+the kernel mutex */
+extern ulint	trx_n_prepared;
 
 /********************************************************************//**
 Releases the search latch if trx has reserved it. */
@@ -108,6 +111,14 @@ trx_free(
 /*=====*/
 	trx_t*	trx);	/*!< in, own: trx object */
 /********************************************************************//**
+At shutdown, frees a transaction object that is in the PREPARED state. */
+UNIV_INTERN
+void
+trx_free_prepared(
+/*==============*/
+	trx_t*	trx)	/*!< in, own: trx object */
+	__attribute__((nonnull));
+/********************************************************************//**
 Frees a transaction object for MySQL. */
 UNIV_INTERN
 void
@@ -498,6 +509,7 @@ struct trx_struct{
 					150 bytes in the undo log size as then
 					we skip XA steps */
 	ulint		flush_log_at_trx_commit_session;
+	ulint		fake_changes;
 	ulint		flush_log_later;/* In 2PC, we hold the
 					prepare_commit mutex across
 					both phases. In that case, we
@@ -589,6 +601,8 @@ struct trx_struct{
 	ulint		mysql_process_no;/* since in Linux, 'top' reports
 					process id's and not thread id's, we
 					store the process number too */
+	time_t		idle_start;
+	ib_int64_t	last_stmt_start;
 	/*------------------------------*/
 	ulint		n_mysql_tables_in_use; /* number of Innobase tables
 					used in the processing of the current
diff --git a/storage/xtradb/include/trx0undo.h b/storage/xtradb/include/trx0undo.h
index a084f2394b5..4f15cd85833 100644
--- a/storage/xtradb/include/trx0undo.h
+++ b/storage/xtradb/include/trx0undo.h
@@ -298,6 +298,15 @@ void
 trx_undo_insert_cleanup(
 /*====================*/
 	trx_t*	trx);	/*!< in: transaction handle */
+
+/********************************************************************//**
+At shutdown, frees the undo logs of a PREPARED transaction. */
+UNIV_INTERN
+void
+trx_undo_free_prepared(
+/*===================*/
+	trx_t*	trx)	/*!< in/out: PREPARED transaction */
+	__attribute__((nonnull));
 #endif /* !UNIV_HOTBACKUP */
 /***********************************************************//**
 Parses the redo log entry of an undo log page initialization.
diff --git a/storage/xtradb/include/univ.i b/storage/xtradb/include/univ.i
index 7b11a16dae9..902d0c94ddd 100644
--- a/storage/xtradb/include/univ.i
+++ b/storage/xtradb/include/univ.i
@@ -1,8 +1,7 @@
 /*****************************************************************************
 
-Copyright (c) 1994, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1994, 2011, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
-Copyright (c) 2009, Sun Microsystems, Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
 Google, Inc. Those modifications are gratefully acknowledged and are described
@@ -46,8 +45,8 @@ Created 1/20/1994 Heikki Tuuri
 
 #define INNODB_VERSION_MAJOR	1
 #define INNODB_VERSION_MINOR	0
-#define INNODB_VERSION_BUGFIX	15
-#define PERCONA_INNODB_VERSION 12.7
+#define INNODB_VERSION_BUGFIX	17
+#define PERCONA_INNODB_VERSION 13.0
 
 /* The following is the InnoDB version as shown in
 SELECT plugin_version FROM information_schema.plugins;
@@ -197,6 +196,8 @@ this will break redo log file compatibility, but it may be useful when
 debugging redo log application problems. */
 #define UNIV_MEM_DEBUG				/* detect memory leaks etc */
 #define UNIV_IBUF_DEBUG				/* debug the insert buffer */
+#define UNIV_BLOB_DEBUG				/* track BLOB ownership;
+assumes that no BLOBs survive server restart */
 #define UNIV_IBUF_COUNT_DEBUG			/* debug the insert buffer;
 this limits the database to IBUF_COUNT_N_SPACES and IBUF_COUNT_N_PAGES,
 and the insert buffer must be empty when the database is started */
diff --git a/storage/xtradb/include/ut0lst.h b/storage/xtradb/include/ut0lst.h
index 245dfc226c3..7b15c052978 100644
--- a/storage/xtradb/include/ut0lst.h
+++ b/storage/xtradb/include/ut0lst.h
@@ -257,48 +257,5 @@ do {									\
 	ut_a(ut_list_node_313 == NULL);					\
 } while (0)
 
-/********************************************************************//**
-Align nodes with moving location.
-@param NAME		the name of the list
-@param TYPE		node type
-@param BASE		base node (not a pointer to it)
-@param OFFSET		offset moved */
-#define UT_LIST_OFFSET(NAME, TYPE, BASE, FADDR, FOFFSET, BOFFSET)	\
-do {									\
-	ulint	ut_list_i_313;						\
-	TYPE*	ut_list_node_313;					\
-									\
-	if ((BASE).start)						\
-		(BASE).start = (void*)((byte*)((BASE).start)			\
-			+ (((void*)((BASE).start) > (void*)FADDR)?FOFFSET:BOFFSET));\
-	if ((BASE).end)							\
-		(BASE).end   = (void*)((byte*)((BASE).end)			\
-			+ (((void*)((BASE).end) > (void*)FADDR)?FOFFSET:BOFFSET));\
-									\
-	ut_list_node_313 = (BASE).start;				\
-									\
-	for (ut_list_i_313 = (BASE).count; ut_list_i_313--; ) {		\
-		ut_a(ut_list_node_313);					\
-		if ((ut_list_node_313->NAME).prev)			\
-			(ut_list_node_313->NAME).prev = (void*)((byte*)((ut_list_node_313->NAME).prev)\
-				+ (((void*)((ut_list_node_313->NAME).prev) > (void*)FADDR)?FOFFSET:BOFFSET));\
-		if ((ut_list_node_313->NAME).next)			\
-			(ut_list_node_313->NAME).next =	(void*)((byte*)((ut_list_node_313->NAME).next)\
-				+ (((void*)((ut_list_node_313->NAME).next)> (void*)FADDR)?FOFFSET:BOFFSET));\
-		ut_list_node_313 = (ut_list_node_313->NAME).next;	\
-	}								\
-									\
-	ut_a(ut_list_node_313 == NULL);					\
-									\
-	ut_list_node_313 = (BASE).end;					\
-									\
-	for (ut_list_i_313 = (BASE).count; ut_list_i_313--; ) {		\
-		ut_a(ut_list_node_313);					\
-		ut_list_node_313 = (ut_list_node_313->NAME).prev;	\
-	}								\
-									\
-	ut_a(ut_list_node_313 == NULL);					\
-} while (0)
-
 #endif
 
diff --git a/storage/xtradb/include/ut0mem.h b/storage/xtradb/include/ut0mem.h
index f14606be966..9c6ee9049ec 100644
--- a/storage/xtradb/include/ut0mem.h
+++ b/storage/xtradb/include/ut0mem.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
@@ -210,43 +210,6 @@ ut_strlcpy_rev(
 	ulint		size);	/*!< in: size of destination buffer */
 
 /**********************************************************************//**
-Compute strlen(ut_strcpyq(str, q)).
-@return	length of the string when quoted */
-UNIV_INLINE
-ulint
-ut_strlenq(
-/*=======*/
-	const char*	str,	/*!< in: null-terminated string */
-	char		q);	/*!< in: the quote character */
-
-/**********************************************************************//**
-Make a quoted copy of a NUL-terminated string.	Leading and trailing
-quotes will not be included; only embedded quotes will be escaped.
-See also ut_strlenq() and ut_memcpyq().
-@return	pointer to end of dest */
-UNIV_INTERN
-char*
-ut_strcpyq(
-/*=======*/
-	char*		dest,	/*!< in: output buffer */
-	char		q,	/*!< in: the quote character */
-	const char*	src);	/*!< in: null-terminated string */
-
-/**********************************************************************//**
-Make a quoted copy of a fixed-length string.  Leading and trailing
-quotes will not be included; only embedded quotes will be escaped.
-See also ut_strlenq() and ut_strcpyq().
-@return	pointer to end of dest */
-UNIV_INTERN
-char*
-ut_memcpyq(
-/*=======*/
-	char*		dest,	/*!< in: output buffer */
-	char		q,	/*!< in: the quote character */
-	const char*	src,	/*!< in: string to be quoted */
-	ulint		len);	/*!< in: length of src */
-
-/**********************************************************************//**
 Return the number of times s2 occurs in s1. Overlapping instances of s2
 are only counted once.
 @return	the number of times s2 occurs in s1 */
diff --git a/storage/xtradb/include/ut0mem.ic b/storage/xtradb/include/ut0mem.ic
index f36c28f1989..c06e2b3ae81 100644
--- a/storage/xtradb/include/ut0mem.ic
+++ b/storage/xtradb/include/ut0mem.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
@@ -99,27 +99,6 @@ ut_strcmp(const char* str1, const char* str2)
 }
 
 /**********************************************************************//**
-Compute strlen(ut_strcpyq(str, q)).
-@return	length of the string when quoted */
-UNIV_INLINE
-ulint
-ut_strlenq(
-/*=======*/
-	const char*	str,	/*!< in: null-terminated string */
-	char		q)	/*!< in: the quote character */
-{
-	ulint len;
-
-	for (len = 0; *str; len++, str++) {
-		if (*str == q) {
-			len++;
-		}
-	}
-
-	return(len);
-}
-
-/**********************************************************************//**
 Converts a raw binary data to a NUL-terminated hex string. The output is
 truncated if there is not enough space in "hex", make sure "hex_size" is at
 least (2 * raw_size + 1) if you do not want this to happen. Returns the