15 files changed, 169 insertions, 45 deletions

diff --git a/storage/innobase/include/btr0cur.h b/storage/innobase/include/btr0cur.h
index cb65c6bfab7..be918439f59 100644
--- a/storage/innobase/include/btr0cur.h
+++ b/storage/innobase/include/btr0cur.h
@@ -466,7 +466,10 @@ btr_estimate_n_rows_in_range(
 /*******************************************************************//**
 Estimates the number of different key values in a given index, for
 each n-column prefix of the index where n <= dict_index_get_n_unique(index).
-The estimates are stored in the array index->stat_n_diff_key_vals. */
+The estimates are stored in the array index->stat_n_diff_key_vals.
+If innodb_stats_method is nulls_ignored, we also record the number of
+non-null values for each prefix and stored the estimates in
+array index->stat_n_non_null_key_vals. */
 UNIV_INTERN
 void
 btr_estimate_number_of_different_key_vals(
@@ -497,8 +500,8 @@ file segment of the index tree.
 @return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
 UNIV_INTERN
 ulint
-btr_store_big_rec_extern_fields(
-/*============================*/
+btr_store_big_rec_extern_fields_func(
+/*=================================*/
 	dict_index_t*	index,		/*!< in: index of rec; the index tree
 					MUST be X-latched */
 	buf_block_t*	rec_block,	/*!< in/out: block containing rec */
@@ -507,10 +510,42 @@ btr_store_big_rec_extern_fields(
 					the "external storage" flags in offsets
 					will not correspond to rec when
 					this function returns */
-	big_rec_t*	big_rec_vec,	/*!< in: vector containing fields
+#ifdef UNIV_DEBUG
+	mtr_t*		local_mtr,	/*!< in: mtr containing the
+					latch to rec and to the tree */
+#endif /* UNIV_DEBUG */
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+	ibool		update_in_place,/*! in: TRUE if the record is updated
+					in place (not delete+insert) */
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
+	const big_rec_t*big_rec_vec)	/*!< in: vector containing fields
 					to be stored externally */
-	mtr_t*		local_mtr);	/*!< in: mtr containing the latch to
-					rec and to the tree */
+	__attribute__((nonnull));
+
+/** Stores the fields in big_rec_vec to the tablespace and puts pointers to
+them in rec.  The extern flags in rec will have to be set beforehand.
+The fields are stored on pages allocated from leaf node
+file segment of the index tree.
+@param index	in: clustered index; MUST be X-latched by mtr
+@param b	in/out: block containing rec; MUST be X-latched by mtr
+@param rec	in/out: clustered index record
+@param offsets	in: rec_get_offsets(rec, index);
+		the "external storage" flags in offsets will not be adjusted
+@param mtr	in: mini-transaction that holds x-latch on index and b
+@param upd	in: TRUE if the record is updated in place (not delete+insert)
+@param big	in: vector containing fields to be stored externally
+@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
+#ifdef UNIV_DEBUG
+# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
+	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,mtr,upd,big)
+#elif defined UNIV_BLOB_LIGHT_DEBUG
+# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
+	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,upd,big)
+#else
+# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
+	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,big)
+#endif
+
 /*******************************************************************//**
 Frees the space in an externally stored field to the file space
 management if the field in data is owned the externally stored field,
diff --git a/storage/innobase/include/buf0buf.h b/storage/innobase/include/buf0buf.h
index 46cae25f0f5..dbd5adee0cc 100644
--- a/storage/innobase/include/buf0buf.h
+++ b/storage/innobase/include/buf0buf.h
@@ -276,9 +276,9 @@ UNIV_INTERN
 buf_block_t*
 buf_block_alloc(
 /*============*/
-	buf_pool_t*	buf_pool,	/*!< buffer pool instance */
-	ulint		zip_size);	/*!< in: compressed page size in bytes,
-					or 0 if uncompressed tablespace */
+	buf_pool_t*	buf_pool);	/*!< in: buffer pool instance,
+					or NULL for round-robin selection
+					of the buffer pool */
 /********************************************************************//**
 Frees a buffer block which does not contain a file page. */
 UNIV_INLINE
@@ -480,7 +480,7 @@ buf_reset_check_index_page_at_flush(
 /*================================*/
 	ulint	space,	/*!< in: space id */
 	ulint	offset);/*!< in: page number */
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 /********************************************************************//**
 Sets file_page_was_freed TRUE if the page is found in the buffer pool.
 This function should be called when we free a file page and want the
@@ -505,7 +505,7 @@ buf_page_reset_file_page_was_freed(
 /*===============================*/
 	ulint	space,	/*!< in: space id */
 	ulint	offset);	/*!< in: page number */
-#endif /* UNIV_DEBUG_FILE_ACCESSES */
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 /********************************************************************//**
 Reads the freed_page_clock of a buffer block.
 @return	freed_page_clock */
@@ -1403,11 +1403,11 @@ struct buf_page_struct{
 					0 if the block was never accessed
 					in the buffer pool */
 	/* @} */
-# ifdef UNIV_DEBUG_FILE_ACCESSES
+# if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	ibool		file_page_was_freed;
 					/*!< this is set to TRUE when fsp
 					frees a page in buffer pool */
-# endif /* UNIV_DEBUG_FILE_ACCESSES */
+# endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 #endif /* !UNIV_HOTBACKUP */
 };
 
diff --git a/storage/innobase/include/buf0lru.h b/storage/innobase/include/buf0lru.h
index 7bcec633d9c..9b150188b03 100644
--- a/storage/innobase/include/buf0lru.h
+++ b/storage/innobase/include/buf0lru.h
@@ -111,12 +111,9 @@ enum buf_lru_free_block_status
 buf_LRU_free_block(
 /*===============*/
 	buf_page_t*	bpage,	/*!< in: block to be freed */
-	ibool		zip,	/*!< in: TRUE if should remove also the
+	ibool		zip)	/*!< in: TRUE if should remove also the
 				compressed page of an uncompressed page */
-	ibool*		buf_pool_mutex_released);
-				/*!< in: pointer to a variable that will
-				be assigned TRUE if buf_pool->mutex
-				was temporarily released, or NULL */
+	__attribute__((nonnull));
 /******************************************************************//**
 Try to free a replaceable block.
 @return	TRUE if found and freed */
@@ -153,9 +150,8 @@ UNIV_INTERN
 buf_block_t*
 buf_LRU_get_free_block(
 /*===================*/
-	buf_pool_t*	buf_pool,	/*!< in: preferred buffer pool */
-	ulint		zip_size);	/*!< in: compressed page size in bytes,
-					or 0 if uncompressed tablespace */
+	buf_pool_t*	buf_pool)	/*!< in/out: buffer pool instance */
+	__attribute__((nonnull,warn_unused_result));
 
 /******************************************************************//**
 Puts a block back to the free list. */
diff --git a/storage/innobase/include/dict0mem.h b/storage/innobase/include/dict0mem.h
index 7526e4dc8dc..5757b79d3ed 100644
--- a/storage/innobase/include/dict0mem.h
+++ b/storage/innobase/include/dict0mem.h
@@ -376,6 +376,12 @@ struct dict_index_struct{
 				dict_get_n_unique(index); we
 				periodically calculate new
 				estimates */
+	ib_int64_t*	stat_n_non_null_key_vals;
+				/* approximate number of non-null key values
+				for this index, for each column where
+				n < dict_get_n_unique(index); This
+				is used when innodb_stats_method is
+				"nulls_ignored". */
 	ulint		stat_index_size;
 				/*!< approximate index size in
 				database pages */
diff --git a/storage/innobase/include/dict0types.h b/storage/innobase/include/dict0types.h
index 0a9edfbfe70..687209575c9 100644
--- a/storage/innobase/include/dict0types.h
+++ b/storage/innobase/include/dict0types.h
@@ -33,11 +33,6 @@ typedef struct dict_index_struct	dict_index_t;
 typedef struct dict_table_struct	dict_table_t;
 typedef struct dict_foreign_struct	dict_foreign_t;
 
-/* A cluster object is a table object with the type field set to
-DICT_CLUSTERED */
-
-typedef dict_table_t			dict_cluster_t;
-
 typedef struct ind_node_struct		ind_node_t;
 typedef struct tab_node_struct		tab_node_t;
 
diff --git a/storage/innobase/include/rem0cmp.h b/storage/innobase/include/rem0cmp.h
index 2f751a38864..a908521c9f7 100644
--- a/storage/innobase/include/rem0cmp.h
+++ b/storage/innobase/include/rem0cmp.h
@@ -165,6 +165,10 @@ cmp_rec_rec_with_match(
 	const ulint*	offsets1,/*!< in: rec_get_offsets(rec1, index) */
 	const ulint*	offsets2,/*!< in: rec_get_offsets(rec2, index) */
 	dict_index_t*	index,	/*!< in: data dictionary index */
+	ibool		nulls_unequal,
+				/* in: TRUE if this is for index statistics
+				cardinality estimation, and innodb_stats_method
+				is "nulls_unequal" or "nulls_ignored" */
 	ulint*		matched_fields, /*!< in/out: number of already completely
 				matched fields; when the function returns,
 				contains the value the for current
diff --git a/storage/innobase/include/rem0cmp.ic b/storage/innobase/include/rem0cmp.ic
index 39ef5f4fba3..63415fe7837 100644
--- a/storage/innobase/include/rem0cmp.ic
+++ b/storage/innobase/include/rem0cmp.ic
@@ -87,5 +87,5 @@ cmp_rec_rec(
 	ulint	match_b		= 0;
 
 	return(cmp_rec_rec_with_match(rec1, rec2, offsets1, offsets2, index,
-				      &match_f, &match_b));
+				      FALSE, &match_f, &match_b));
 }
diff --git a/storage/innobase/include/row0upd.h b/storage/innobase/include/row0upd.h
index 9c2e003c609..c275c1da78e 100644
--- a/storage/innobase/include/row0upd.h
+++ b/storage/innobase/include/row0upd.h
@@ -280,19 +280,29 @@ NOTE: we compare the fields as binary strings!
 @return TRUE if update vector changes an ordering field in the index record */
 UNIV_INTERN
 ibool
-row_upd_changes_ord_field_binary(
-/*=============================*/
+row_upd_changes_ord_field_binary_func(
+/*==================================*/
+	dict_index_t*	index,	/*!< in: index of the record */
+	const upd_t*	update,	/*!< in: update vector for the row; NOTE: the
+				field numbers in this MUST be clustered index
+				positions! */
+#ifdef UNIV_DEBUG
+	const que_thr_t*thr,	/*!< in: query thread */
+#endif /* UNIV_DEBUG */
 	const dtuple_t*	row,	/*!< in: old value of row, or NULL if the
 				row and the data values in update are not
 				known when this function is called, e.g., at
 				compile time */
-	const row_ext_t*ext,	/*!< NULL, or prefixes of the externally
+	const row_ext_t*ext)	/*!< NULL, or prefixes of the externally
 				stored columns in the old row */
-	dict_index_t*	index,	/*!< in: index of the record */
-	const upd_t*	update)	/*!< in: update vector for the row; NOTE: the
-				field numbers in this MUST be clustered index
-				positions! */
-	__attribute__((nonnull(3,4), warn_unused_result));
+	__attribute__((nonnull(1,2), warn_unused_result));
+#ifdef UNIV_DEBUG
+# define row_upd_changes_ord_field_binary(index,update,thr,row,ext)	\
+	row_upd_changes_ord_field_binary_func(index,update,thr,row,ext)
+#else /* UNIV_DEBUG */
+# define row_upd_changes_ord_field_binary(index,update,thr,row,ext)	\
+	row_upd_changes_ord_field_binary_func(index,update,row,ext)
+#endif /* UNIV_DEBUG */
 /***********************************************************//**
 Checks if an update vector changes an ordering field of an index record.
 This function is fast if the update vector is short or the number of ordering
diff --git a/storage/innobase/include/srv0srv.h b/storage/innobase/include/srv0srv.h
index 612ff4c6e54..d600ad4a034 100644
--- a/storage/innobase/include/srv0srv.h
+++ b/storage/innobase/include/srv0srv.h
@@ -172,6 +172,11 @@ capacity. PCT_IO(5) -> returns the number of IO operations that
 is 5% of the max where max is srv_io_capacity.  */
 #define PCT_IO(p) ((ulong) (srv_io_capacity * ((double) p / 100.0)))
 
+/* The "innodb_stats_method" setting, decides how InnoDB is going
+to treat NULL value when collecting statistics. It is not defined
+as enum type because the configure option takes unsigned integer type. */
+extern ulong	srv_innodb_stats_method;
+
 #ifdef UNIV_LOG_ARCHIVE
 extern ibool		srv_log_archive_on;
 extern ibool		srv_archive_recovery;
@@ -418,6 +423,19 @@ enum {
 					in connection with recovery */
 };
 
+/* Alternatives for srv_innodb_stats_method, which could be changed by
+setting innodb_stats_method */
+enum srv_stats_method_name_enum {
+	SRV_STATS_NULLS_EQUAL,		/* All NULL values are treated as
+					equal. This is the default setting
+					for innodb_stats_method */
+	SRV_STATS_NULLS_UNEQUAL,	/* All NULL values are treated as
+					NOT equal. */
+	SRV_STATS_NULLS_IGNORED		/* NULL values are ignored */
+};
+
+typedef enum srv_stats_method_name_enum		srv_stats_method_name_t;
+
 #ifndef UNIV_HOTBACKUP
 /** Types of threads existing in the system. */
 enum srv_thread_type {
diff --git a/storage/innobase/include/sync0rw.h b/storage/innobase/include/sync0rw.h
index 5988fdfb382..9e5c96fa323 100644
--- a/storage/innobase/include/sync0rw.h
+++ b/storage/innobase/include/sync0rw.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
@@ -555,6 +555,7 @@ UNIV_INTERN
 void
 rw_lock_debug_print(
 /*================*/
+	FILE*			f,	/*!< in: output stream */
 	rw_lock_debug_t*	info);	/*!< in: debug struct */
 #endif /* UNIV_SYNC_DEBUG */
 
diff --git a/storage/innobase/include/trx0rseg.h b/storage/innobase/include/trx0rseg.h
index 78a7a8c4bb0..a293d9b896e 100644
--- a/storage/innobase/include/trx0rseg.h
+++ b/storage/innobase/include/trx0rseg.h
@@ -141,9 +141,7 @@ struct trx_rseg_struct{
 	ulint		id;	/*!< rollback segment id == the index of
 				its slot in the trx system file copy */
 	mutex_t		mutex;	/*!< mutex protecting the fields in this
-				struct except id; NOTE that the latching
-				order must always be kernel mutex ->
-				rseg mutex */
+				struct except id, which is constant */
 	ulint		space;	/*!< space where the rollback segment is
 				header is placed */
 	ulint		zip_size;/* compressed page size of space
diff --git a/storage/innobase/include/trx0trx.h b/storage/innobase/include/trx0trx.h
index b4a4aa0ca44..83f6182c347 100644
--- a/storage/innobase/include/trx0trx.h
+++ b/storage/innobase/include/trx0trx.h
@@ -214,12 +214,12 @@ trx_recover_for_mysql(
 /*******************************************************************//**
 This function is used to find one X/Open XA distributed transaction
 which is in the prepared state
-@return	trx or NULL */
+@return	trx or NULL; on match, the trx->xid will be invalidated */
 UNIV_INTERN
 trx_t *
 trx_get_trx_by_xid(
 /*===============*/
-	XID*	xid);	/*!< in: X/Open XA transaction identification */
+	const XID*	xid);	/*!< in: X/Open XA transaction identifier */
 /**********************************************************************//**
 If required, flushes the log to disk if we called trx_commit_for_mysql()
 with trx->flush_log_later == TRUE.
diff --git a/storage/innobase/include/univ.i b/storage/innobase/include/univ.i
index 4470d221da9..a99b8306eb8 100644
--- a/storage/innobase/include/univ.i
+++ b/storage/innobase/include/univ.i
@@ -184,14 +184,15 @@ command. Not tested on Windows. */
 						debugging without UNIV_DEBUG */
 #define UNIV_BUF_DEBUG				/* Enable buffer pool
 						debugging without UNIV_DEBUG */
+#define UNIV_BLOB_LIGHT_DEBUG			/* Enable off-page column
+						debugging without UNIV_DEBUG */
 #define UNIV_DEBUG				/* Enable ut_ad() assertions
 						and disable UNIV_INLINE */
 #define UNIV_DEBUG_LOCK_VALIDATE		/* Enable
 						ut_ad(lock_rec_validate_page())
 						assertions. */
-#define UNIV_DEBUG_FILE_ACCESSES		/* Debug .ibd file access
-						(field file_page_was_freed
-						in buf_page_t) */
+#define UNIV_DEBUG_FILE_ACCESSES		/* Enable freed block access
+						debugging without UNIV_DEBUG */
 #define UNIV_LRU_DEBUG				/* debug the buffer pool LRU */
 #define UNIV_HASH_DEBUG				/* debug HASH_ macros */
 #define UNIV_LIST_DEBUG				/* debug UT_LIST_ macros */
@@ -308,6 +309,18 @@ number does not include a terminating '\0'. InnoDB probably can handle
 longer names internally */
 #define MAX_TABLE_NAME_LEN	192
 
+/* The maximum length of a database name. Like MAX_TABLE_NAME_LEN this is
+the MySQL's NAME_LEN, see check_and_convert_db_name(). */
+#define MAX_DATABASE_NAME_LEN	MAX_TABLE_NAME_LEN
+
+/* MAX_FULL_NAME_LEN defines the full name path including the
+database name and table name. In addition, 14 bytes is added for:
+	2 for surrounding quotes around table name
+	1 for the separating dot (.)
+	9 for the #mysql50# prefix */
+#define MAX_FULL_NAME_LEN				\
+	(MAX_TABLE_NAME_LEN + MAX_DATABASE_NAME_LEN + 14)
+
 /*
 			UNIVERSAL TYPE DEFINITIONS
 			==========================
@@ -429,7 +442,7 @@ it is read or written. */
 /* Use sun_prefetch when compile with Sun Studio */
 # define UNIV_EXPECT(expr,value) (expr)
 # define UNIV_LIKELY_NULL(expr) (expr)
-# define UNIV_PREFETCH_R(addr) sun_prefetch_read_many(addr)
+# define UNIV_PREFETCH_R(addr) sun_prefetch_read_many((void*) addr)
 # define UNIV_PREFETCH_RW(addr) sun_prefetch_write_many(addr)
 #else
 /* Dummy versions of the macros */
diff --git a/storage/innobase/include/ut0vec.h b/storage/innobase/include/ut0vec.h
index a770f671cfc..0f8b955b098 100644
--- a/storage/innobase/include/ut0vec.h
+++ b/storage/innobase/include/ut0vec.h
@@ -94,6 +94,25 @@ ib_vector_get(
 	ulint		n);	/*!< in: element index to get */
 
 /****************************************************************//**
+Get last element. The vector must not be empty.
+@return	last element */
+UNIV_INLINE
+void*
+ib_vector_get_last(
+/*===============*/
+	ib_vector_t*	vec);	/*!< in: vector */
+
+/****************************************************************//**
+Set the n'th element. */
+UNIV_INLINE
+void
+ib_vector_set(
+/*==========*/
+	ib_vector_t*	vec,	/*!< in/out: vector */
+	ulint		n,	/*!< in: element index to set */
+	void*		elem);	/*!< in: data element */
+
+/****************************************************************//**
 Remove the last element from the vector. */
 UNIV_INLINE
 void*
diff --git a/storage/innobase/include/ut0vec.ic b/storage/innobase/include/ut0vec.ic
index 02e881f9bca..34c858868ce 100644
--- a/storage/innobase/include/ut0vec.ic
+++ b/storage/innobase/include/ut0vec.ic
@@ -51,6 +51,35 @@ ib_vector_get(
 }
 
 /****************************************************************//**
+Get last element. The vector must not be empty.
+@return	last element */
+UNIV_INLINE
+void*
+ib_vector_get_last(
+/*===============*/
+	ib_vector_t*	vec)	/*!< in: vector */
+{
+	ut_a(vec->used > 0);
+
+	return(vec->data[vec->used - 1]);
+}
+
+/****************************************************************//**
+Set the n'th element. */
+UNIV_INLINE
+void
+ib_vector_set(
+/*==========*/
+	ib_vector_t*	vec,	/*!< in/out: vector */
+	ulint		n,	/*!< in: element index to set */
+	void*		elem)	/*!< in: data element */
+{
+	ut_a(n < vec->used);
+
+	vec->data[n] = elem;
+}
+
+/****************************************************************//**
 Remove the last element from the vector.
 @return	last vector element */
 UNIV_INLINE