mysql-5.5 merge

42 files changed, 822 insertions, 674 deletions

diff --git a/storage/archive/ha_archive.cc b/storage/archive/ha_archive.cc
index 4008165865a..b5ec94ba503 100644
--- a/storage/archive/ha_archive.cc
+++ b/storage/archive/ha_archive.cc
@@ -830,6 +830,7 @@ uint32 ha_archive::max_row_length(const uchar *buf)
        ptr != end ;
        ptr++)
   {
+    if (!table->field[*ptr]->is_null())
       length += 2 + ((Field_blob*)table->field[*ptr])->get_length();
   }
 
@@ -1186,6 +1187,17 @@ int ha_archive::unpack_row(azio_stream *file_to_read, uchar *record)
 
   /* Copy null bits */
   const uchar *ptr= record_buffer->buffer;
+  /*
+    Field::unpack() is not called when field is NULL. For VARCHAR
+    Field::unpack() only unpacks as much bytes as occupied by field
+    value. In these cases respective memory area on record buffer is
+    not initialized.
+
+    These uninitialized areas may be accessed by CHECKSUM TABLE or
+    by optimizer using temporary table (BUG#12997905). We may remove
+    this memset() when they're fixed.
+  */
+  memset(record, 0, table->s->reclength);
   memcpy(record, ptr, table->s->null_bytes);
   ptr+= table->s->null_bytes;
   for (Field **field=table->field ; *field ; field++)
@@ -1691,13 +1703,15 @@ int ha_archive::check(THD* thd, HA_CHECK_OPT* check_opt)
 {
   int rc= 0;
   const char *old_proc_info;
-  ha_rows count= share->rows_recorded;
+  ha_rows count;
   DBUG_ENTER("ha_archive::check");
 
   old_proc_info= thd_proc_info(thd, "Checking table");
-  /* Flush any waiting data */
   mysql_mutex_lock(&share->mutex);
-  azflush(&(share->archive_write), Z_SYNC_FLUSH);
+  count= share->rows_recorded;
+  /* Flush any waiting data */
+  if (share->archive_write_open)
+    azflush(&(share->archive_write), Z_SYNC_FLUSH);
   mysql_mutex_unlock(&share->mutex);
 
   if (init_archive_reader())
@@ -1707,18 +1721,34 @@ int ha_archive::check(THD* thd, HA_CHECK_OPT* check_opt)
     start of the file.
   */
   read_data_header(&archive);
+  for (ha_rows cur_count= count; cur_count; cur_count--)
+  {
+    if ((rc= get_row(&archive, table->record[0])))
+      goto error;
+  }
+  /*
+    Now read records that may have been inserted concurrently.
+    Acquire share->mutex so tail of the table is not modified by
+    concurrent writers.
+  */
+  mysql_mutex_lock(&share->mutex);
+  count= share->rows_recorded - count;
+  if (share->archive_write_open)
+    azflush(&(share->archive_write), Z_SYNC_FLUSH);
   while (!(rc= get_row(&archive, table->record[0])))
     count--;
-
-  thd_proc_info(thd, old_proc_info);
+  mysql_mutex_unlock(&share->mutex);
 
   if ((rc && rc != HA_ERR_END_OF_FILE) || count)  
-  {
-    share->crashed= FALSE;
-    DBUG_RETURN(HA_ADMIN_CORRUPT);
-  }
+    goto error;
 
+  thd_proc_info(thd, old_proc_info);
   DBUG_RETURN(HA_ADMIN_OK);
+
+error:
+  thd_proc_info(thd, old_proc_info);
+  share->crashed= FALSE;
+  DBUG_RETURN(HA_ADMIN_CORRUPT);
 }
 
 /*
diff --git a/storage/federated/ha_federated.cc b/storage/federated/ha_federated.cc
index 1091d88ffc8..4294f4b094f 100644
--- a/storage/federated/ha_federated.cc
+++ b/storage/federated/ha_federated.cc
@@ -1686,6 +1686,16 @@ int ha_federated::close(void)
   mysql_close(mysql);
   mysql= NULL;
 
+  /*
+    mysql_close() might return an error if a remote server's gone
+    for some reason. If that happens while removing a table from
+    the table cache, the error will be propagated to a client even
+    if the original query was not issued against the FEDERATED table.
+    So, don't propagate errors from mysql_close().
+  */
+  if (table->in_use)
+    table->in_use->clear_error();
+
   DBUG_RETURN(free_share(share));
 }
 
diff --git a/storage/innobase/btr/btr0cur.c b/storage/innobase/btr/btr0cur.c
index 625721133fd..03346337d3f 100644
--- a/storage/innobase/btr/btr0cur.c
+++ b/storage/innobase/btr/btr0cur.c
@@ -458,8 +458,6 @@ btr_cur_search_to_nth_level(
 	cursor->flag = BTR_CUR_BINARY;
 	cursor->index = index;
 
-	cursor->ibuf_cnt = ULINT_UNDEFINED;
-
 #ifndef BTR_CUR_ADAPT
 	guess = NULL;
 #else
@@ -747,21 +745,8 @@ retry_page_get:
 			/* We're doing a search on an ibuf tree and we're one
 			level above the leaf page. */
 
-			ulint	is_min_rec;
-
 			ut_ad(level == 0);
 
-			is_min_rec = rec_get_info_bits(node_ptr, 0)
-				& REC_INFO_MIN_REC_FLAG;
-
-			if (!is_min_rec) {
-				cursor->ibuf_cnt
-					= ibuf_rec_get_counter(node_ptr);
-
-				ut_a(cursor->ibuf_cnt <= 0xFFFF
-				     || cursor->ibuf_cnt == ULINT_UNDEFINED);
-			}
-
 			buf_mode = BUF_GET;
 			rw_latch = RW_NO_LATCH;
 			goto retry_page_get;
diff --git a/storage/innobase/btr/btr0pcur.c b/storage/innobase/btr/btr0pcur.c
index 57d9752649f..fb153556b2f 100644
--- a/storage/innobase/btr/btr0pcur.c
+++ b/storage/innobase/btr/btr0pcur.c
@@ -52,12 +52,13 @@ btr_pcur_create_for_mysql(void)
 }
 
 /**************************************************************//**
-Frees the memory for a persistent cursor object. */
+Resets a persistent cursor object, freeing ::old_rec_buf if it is
+allocated and resetting the other members to their initial values. */
 UNIV_INTERN
 void
-btr_pcur_free_for_mysql(
-/*====================*/
-	btr_pcur_t*	cursor)	/*!< in, own: persistent cursor */
+btr_pcur_reset(
+/*===========*/
+	btr_pcur_t*	cursor)	/*!< in, out: persistent cursor */
 {
 	if (cursor->old_rec_buf != NULL) {
 
@@ -66,6 +67,7 @@ btr_pcur_free_for_mysql(
 		cursor->old_rec_buf = NULL;
 	}
 
+	cursor->btr_cur.index = NULL;
 	cursor->btr_cur.page_cur.rec = NULL;
 	cursor->old_rec = NULL;
 	cursor->old_n_fields = 0;
@@ -73,7 +75,17 @@ btr_pcur_free_for_mysql(
 
 	cursor->latch_mode = BTR_NO_LATCHES;
 	cursor->pos_state = BTR_PCUR_NOT_POSITIONED;
+}
 
+/**************************************************************//**
+Frees the memory for a persistent cursor object. */
+UNIV_INTERN
+void
+btr_pcur_free_for_mysql(
+/*====================*/
+	btr_pcur_t*	cursor)	/*!< in, own: persistent cursor */
+{
+	btr_pcur_reset(cursor);
 	mem_free(cursor);
 }
 
diff --git a/storage/innobase/buf/buf0buf.c b/storage/innobase/buf/buf0buf.c
index b5004a0834c..890d0282286 100644
--- a/storage/innobase/buf/buf0buf.c
+++ b/storage/innobase/buf/buf0buf.c
@@ -3885,6 +3885,9 @@ buf_pool_validate_instance(
 					ut_a(rw_lock_is_locked(&block->lock,
 							       RW_LOCK_EX));
 					break;
+
+				case BUF_IO_PIN:
+					break;
 				}
 
 				n_lru++;
@@ -3914,6 +3917,7 @@ buf_pool_validate_instance(
 		ut_a(buf_page_get_state(b) == BUF_BLOCK_ZIP_PAGE);
 		switch (buf_page_get_io_fix(b)) {
 		case BUF_IO_NONE:
+		case BUF_IO_PIN:
 			/* All clean blocks should be I/O-unfixed. */
 			break;
 		case BUF_IO_READ:
@@ -3953,6 +3957,7 @@ buf_pool_validate_instance(
 			switch (buf_page_get_io_fix(b)) {
 			case BUF_IO_NONE:
 			case BUF_IO_READ:
+			case BUF_IO_PIN:
 				break;
 			case BUF_IO_WRITE:
 				switch (buf_page_get_flush_type(b)) {
diff --git a/storage/innobase/buf/buf0lru.c b/storage/innobase/buf/buf0lru.c
index 510f6eefba5..15b0ad40aaa 100644
--- a/storage/innobase/buf/buf0lru.c
+++ b/storage/innobase/buf/buf0lru.c
@@ -68,8 +68,12 @@ allowed to point to either end of the LRU list. */
 
 /** When dropping the search hash index entries before deleting an ibd
 file, we build a local array of pages belonging to that tablespace
-in the buffer pool. Following is the size of that array. */
-#define BUF_LRU_DROP_SEARCH_HASH_SIZE	1024
+in the buffer pool. Following is the size of that array.
+We also release buf_pool->mutex after scanning this many pages of the
+flush_list when dropping a table. This is to ensure that other threads
+are not blocked for extended period of time when using very large
+buffer pools. */
+#define BUF_LRU_DROP_SEARCH_SIZE	1024
 
 /** If we switch on the InnoDB monitor because there are too few available
 frames in the buffer pool, we set this to TRUE */
@@ -210,7 +214,7 @@ buf_LRU_drop_page_hash_batch(
 	ulint	i;
 
 	ut_ad(arr != NULL);
-	ut_ad(count <= BUF_LRU_DROP_SEARCH_HASH_SIZE);
+	ut_ad(count <= BUF_LRU_DROP_SEARCH_SIZE);
 
 	for (i = 0; i < count; ++i) {
 		btr_search_drop_page_hash_when_freed(space_id, zip_size,
@@ -244,7 +248,7 @@ buf_LRU_drop_page_hash_for_tablespace(
 	}
 
 	page_arr = ut_malloc(
-		sizeof(ulint) * BUF_LRU_DROP_SEARCH_HASH_SIZE);
+		sizeof(ulint) * BUF_LRU_DROP_SEARCH_SIZE);
 
 	buf_pool_mutex_enter(buf_pool);
 	num_entries = 0;
@@ -283,10 +287,10 @@ next_page:
 		/* Store the page number so that we can drop the hash
 		index in a batch later. */
 		page_arr[num_entries] = bpage->offset;
-		ut_a(num_entries < BUF_LRU_DROP_SEARCH_HASH_SIZE);
+		ut_a(num_entries < BUF_LRU_DROP_SEARCH_SIZE);
 		++num_entries;
 
-		if (num_entries < BUF_LRU_DROP_SEARCH_HASH_SIZE) {
+		if (num_entries < BUF_LRU_DROP_SEARCH_SIZE) {
 			goto next_page;
 		}
 
@@ -331,37 +335,40 @@ next_page:
 }
 
 /******************************************************************//**
-Invalidates all pages belonging to a given tablespace inside a specific
+Remove all dirty pages belonging to a given tablespace inside a specific
 buffer pool instance when we are deleting the data file(s) of that
-tablespace. */
+tablespace. The pages still remain a part of LRU and are evicted from
+the list as they age towards the tail of the LRU. */
 static
 void
-buf_LRU_invalidate_tablespace_buf_pool_instance(
-/*============================================*/
+buf_LRU_remove_dirty_pages_for_tablespace(
+/*======================================*/
 	buf_pool_t*	buf_pool,	/*!< buffer pool instance */
 	ulint		id)		/*!< in: space id */
 {
 	buf_page_t*	bpage;
 	ibool		all_freed;
+	ulint		i;
 
 scan_again:
 	buf_pool_mutex_enter(buf_pool);
+	buf_flush_list_mutex_enter(buf_pool);
 
 	all_freed = TRUE;
 
-	bpage = UT_LIST_GET_LAST(buf_pool->LRU);
+	for (bpage = UT_LIST_GET_LAST(buf_pool->flush_list), i = 0;
+	     bpage != NULL; ++i) {
 
-	while (bpage != NULL) {
 		buf_page_t*	prev_bpage;
 		mutex_t*	block_mutex = NULL;
 
 		ut_a(buf_page_in_file(bpage));
 
-		prev_bpage = UT_LIST_GET_PREV(LRU, bpage);
+		prev_bpage = UT_LIST_GET_PREV(list, bpage);
 
 		/* bpage->space and bpage->io_fix are protected by
-		buf_pool_mutex and block_mutex.  It is safe to check
-		them while holding buf_pool_mutex only. */
+		buf_pool->mutex and block_mutex. It is safe to check
+		them while holding buf_pool->mutex only. */
 
 		if (buf_page_get_space(bpage) != id) {
 			/* Skip this block, as it does not belong to
@@ -374,79 +381,83 @@ scan_again:
 
 			all_freed = FALSE;
 			goto next_page;
-		} else {
-			block_mutex = buf_page_get_mutex(bpage);
-			mutex_enter(block_mutex);
+		}
 
-			if (bpage->buf_fix_count > 0) {
+		/* We have to release the flush_list_mutex to obey the
+		latching order. We are however guaranteed that the page
+		will stay in the flush_list because buf_flush_remove()
+		needs buf_pool->mutex as well. */
+		buf_flush_list_mutex_exit(buf_pool);
+		block_mutex = buf_page_get_mutex(bpage);
+		mutex_enter(block_mutex);
 
-				mutex_exit(block_mutex);
-				/* We cannot remove this page during
-				this scan yet; maybe the system is
-				currently reading it in, or flushing
-				the modifications to the file */
+		if (bpage->buf_fix_count > 0) {
+			mutex_exit(block_mutex);
+			buf_flush_list_mutex_enter(buf_pool);
 
-				all_freed = FALSE;
+			/* We cannot remove this page during
+			this scan yet; maybe the system is
+			currently reading it in, or flushing
+			the modifications to the file */
 
-				goto next_page;
-			}
+			all_freed = FALSE;
+			goto next_page;
 		}
 
-		ut_ad(mutex_own(block_mutex));
+		ut_ad(bpage->oldest_modification != 0);
 
-#ifdef UNIV_DEBUG
-		if (buf_debug_prints) {
-			fprintf(stderr,
-				"Dropping space %lu page %lu\n",
-				(ulong) buf_page_get_space(bpage),
-				(ulong) buf_page_get_page_no(bpage));
-		}
-#endif
-		if (buf_page_get_state(bpage) != BUF_BLOCK_FILE_PAGE) {
-			/* This is a compressed-only block
-			descriptor. Do nothing. */
-		} else if (((buf_block_t*) bpage)->index) {
-			ulint	page_no;
-			ulint	zip_size;
+		buf_flush_remove(bpage);
 
-			buf_pool_mutex_exit(buf_pool);
-
-			zip_size = buf_page_get_zip_size(bpage);
-			page_no = buf_page_get_page_no(bpage);
+		mutex_exit(block_mutex);
+		buf_flush_list_mutex_enter(buf_pool);
+next_page:
+		bpage = prev_bpage;
 
-			mutex_exit(block_mutex);
+		if (!bpage) {
+			break;
+		}
 
-			/* Note that the following call will acquire
-			and release an X-latch on the page. */
+		/* Every BUF_LRU_DROP_SEARCH_SIZE iterations in the
+		loop we release buf_pool->mutex to let other threads
+		do their job. */
+		if (i < BUF_LRU_DROP_SEARCH_SIZE) {
+			continue;
+		}
 
-			btr_search_drop_page_hash_when_freed(
-				id, zip_size, page_no);
-			goto scan_again;
+		/* We IO-fix the block to make sure that the block
+		stays in its position in the flush_list. */
+		if (buf_page_get_io_fix(bpage) != BUF_IO_NONE) {
+			/* Block is already IO-fixed. We don't
+			want to change the value. Lets leave
+			this block alone. */
+			continue;
 		}
 
-		if (bpage->oldest_modification != 0) {
+		buf_flush_list_mutex_exit(buf_pool);
+		block_mutex = buf_page_get_mutex(bpage);
+		mutex_enter(block_mutex);
+		buf_page_set_sticky(bpage);
+		mutex_exit(block_mutex);
 
-			buf_flush_remove(bpage);
-		}
+		/* Now it is safe to release the buf_pool->mutex. */
+		buf_pool_mutex_exit(buf_pool);
+		os_thread_yield();
+		buf_pool_mutex_enter(buf_pool);
 
-		/* Remove from the LRU list. */
+		mutex_enter(block_mutex);
+		buf_page_unset_sticky(bpage);
+		mutex_exit(block_mutex);
 
-		if (buf_LRU_block_remove_hashed_page(bpage, TRUE)
-		    != BUF_BLOCK_ZIP_FREE) {
-			buf_LRU_block_free_hashed_page((buf_block_t*) bpage);
-			mutex_exit(block_mutex);
-		} else {
-			/* The block_mutex should have been released
-			by buf_LRU_block_remove_hashed_page() when it
-			returns BUF_BLOCK_ZIP_FREE. */
-			ut_ad(block_mutex == &buf_pool->zip_mutex);
-			ut_ad(!mutex_own(block_mutex));
-		}
-next_page:
-		bpage = prev_bpage;
+		buf_flush_list_mutex_enter(buf_pool);
+		ut_ad(bpage->in_flush_list);
+
+		i = 0;
 	}
 
 	buf_pool_mutex_exit(buf_pool);
+	buf_flush_list_mutex_exit(buf_pool);
+
+	ut_ad(buf_flush_validate(buf_pool));
 
 	if (!all_freed) {
 		os_thread_sleep(20000);
@@ -477,7 +488,7 @@ buf_LRU_invalidate_tablespace(
 
 		buf_pool = buf_pool_from_array(i);
 		buf_LRU_drop_page_hash_for_tablespace(buf_pool, id);
-		buf_LRU_invalidate_tablespace_buf_pool_instance(buf_pool, id);
+		buf_LRU_remove_dirty_pages_for_tablespace(buf_pool, id);
 	}
 }
 
@@ -1532,8 +1543,9 @@ alloc:
 			/* Prevent buf_page_get_gen() from
 			decompressing the block while we release
 			buf_pool->mutex and block_mutex. */
-			b->buf_fix_count++;
-			b->io_fix = BUF_IO_READ;
+			mutex_enter(&buf_pool->zip_mutex);
+			buf_page_set_sticky(b);
+			mutex_exit(&buf_pool->zip_mutex);
 		}
 
 		buf_pool_mutex_exit(buf_pool);
@@ -1573,8 +1585,7 @@ alloc:
 
 		if (b) {
 			mutex_enter(&buf_pool->zip_mutex);
-			b->buf_fix_count--;
-			buf_page_set_io_fix(b, BUF_IO_NONE);
+			buf_page_unset_sticky(b);
 			mutex_exit(&buf_pool->zip_mutex);
 		}
 
diff --git a/storage/innobase/dict/dict0dict.c b/storage/innobase/dict/dict0dict.c
index 2a2c7652817..9dc3cef229e 100644
--- a/storage/innobase/dict/dict0dict.c
+++ b/storage/innobase/dict/dict0dict.c
@@ -911,6 +911,11 @@ dict_index_find_on_id_low(
 	dict_table_t*	table;
 	dict_index_t*	index;
 
+	/* This can happen if the system tablespace is the wrong page size */
+	if (dict_sys == NULL) {
+		return(NULL);
+	}
+
 	table = UT_LIST_GET_FIRST(dict_sys->table_LRU);
 
 	while (table) {
@@ -3069,10 +3074,15 @@ dict_scan_table_name(
 		memcpy(ref, database_name, database_name_len);
 		ref[database_name_len] = '/';
 		memcpy(ref + database_name_len + 1, table_name, table_name_len + 1);
+
 	} else {
+#ifndef __WIN__
 		if (innobase_get_lower_case_table_names() == 1) {
 			innobase_casedn_str(ref);
 		}
+#else
+		innobase_casedn_str(ref);
+#endif /* !__WIN__ */
 		*table = dict_table_get_low(ref);
 	}
 
diff --git a/storage/innobase/dict/dict0load.c b/storage/innobase/dict/dict0load.c
index 0a5cfb87140..44a0ec5b338 100644
--- a/storage/innobase/dict/dict0load.c
+++ b/storage/innobase/dict/dict0load.c
@@ -2066,8 +2066,9 @@ static
 void
 dict_load_foreign_cols(
 /*===================*/
-	const char*	id,	/*!< in: foreign constraint id as a
-				null-terminated string */
+	const char*	id,	/*!< in: foreign constraint id, not
+				necessary '\0'-terminated */
+	ulint		id_len,	/*!< in: id length */
 	dict_foreign_t*	foreign)/*!< in: foreign constraint object */
 {
 	dict_table_t*	sys_foreign_cols;
@@ -2097,7 +2098,7 @@ dict_load_foreign_cols(
 	tuple = dtuple_create(foreign->heap, 1);
 	dfield = dtuple_get_nth_field(tuple, 0);
 
-	dfield_set_data(dfield, id, ut_strlen(id));
+	dfield_set_data(dfield, id, id_len);
 	dict_index_copy_types(tuple, sys_index, 1);
 
 	btr_pcur_open_on_user_rec(sys_index, tuple, PAGE_CUR_GE,
@@ -2110,7 +2111,7 @@ dict_load_foreign_cols(
 		ut_a(!rec_get_deleted_flag(rec, 0));
 
 		field = rec_get_nth_field_old(rec, 0, &len);
-		ut_a(len == ut_strlen(id));
+		ut_a(len == id_len);
 		ut_a(ut_memcmp(id, field, len) == 0);
 
 		field = rec_get_nth_field_old(rec, 1, &len);
@@ -2139,8 +2140,9 @@ static
 ulint
 dict_load_foreign(
 /*==============*/
-	const char*	id,	/*!< in: foreign constraint id as a
-				null-terminated string */
+	const char*	id,	/*!< in: foreign constraint id, not
+				necessary '\0'-terminated */
+	ulint		id_len,	/*!< in: id length */
 	ibool		check_charsets,
 				/*!< in: TRUE=check charset compatibility */
 	ibool		check_recursive)
@@ -2176,7 +2178,7 @@ dict_load_foreign(
 	tuple = dtuple_create(heap2, 1);
 	dfield = dtuple_get_nth_field(tuple, 0);
 
-	dfield_set_data(dfield, id, ut_strlen(id));
+	dfield_set_data(dfield, id, id_len);
 	dict_index_copy_types(tuple, sys_index, 1);
 
 	btr_pcur_open_on_user_rec(sys_index, tuple, PAGE_CUR_GE,
@@ -2188,8 +2190,8 @@ dict_load_foreign(
 		/* Not found */
 
 		fprintf(stderr,
-			"InnoDB: Error A: cannot load foreign constraint %s\n",
-			id);
+			"InnoDB: Error A: cannot load foreign constraint "
+			"%.*s\n", (int) id_len, id);
 
 		btr_pcur_close(&pcur);
 		mtr_commit(&mtr);
@@ -2201,11 +2203,11 @@ dict_load_foreign(
 	field = rec_get_nth_field_old(rec, 0, &len);
 
 	/* Check if the id in record is the searched one */
-	if (len != ut_strlen(id) || ut_memcmp(id, field, len) != 0) {
+	if (len != id_len || ut_memcmp(id, field, len) != 0) {
 
 		fprintf(stderr,
-			"InnoDB: Error B: cannot load foreign constraint %s\n",
-			id);
+			"InnoDB: Error B: cannot load foreign constraint "
+			"%.*s\n", (int) id_len, id);
 
 		btr_pcur_close(&pcur);
 		mtr_commit(&mtr);
@@ -2231,7 +2233,7 @@ dict_load_foreign(
 	foreign->type = (unsigned int) (n_fields_and_type >> 24);
 	foreign->n_fields = (unsigned int) (n_fields_and_type & 0x3FFUL);
 
-	foreign->id = mem_heap_strdup(foreign->heap, id);
+	foreign->id = mem_heap_strdupl(foreign->heap, id, id_len);
 
 	field = rec_get_nth_field_old(rec, 3, &len);
 
@@ -2247,7 +2249,7 @@ dict_load_foreign(
 	btr_pcur_close(&pcur);
 	mtr_commit(&mtr);
 
-	dict_load_foreign_cols(id, foreign);
+	dict_load_foreign_cols(id, id_len, foreign);
 
 	ref_table = dict_table_check_if_in_cache_low(
 			foreign->referenced_table_name_lookup);
@@ -2326,8 +2328,8 @@ dict_load_foreigns(
 	ibool		check_charsets)	/*!< in: TRUE=check charset
 					compatibility */
 {
+	char		tuple_buf[DTUPLE_EST_ALLOC(1)];
 	btr_pcur_t	pcur;
-	mem_heap_t*	heap;
 	dtuple_t*	tuple;
 	dfield_t*	dfield;
 	dict_index_t*	sec_index;
@@ -2335,7 +2337,6 @@ dict_load_foreigns(
 	const rec_t*	rec;
 	const byte*	field;
 	ulint		len;
-	char*		id ;
 	ulint		err;
 	mtr_t		mtr;
 
@@ -2362,9 +2363,8 @@ dict_load_foreigns(
 	sec_index = dict_table_get_next_index(
 		dict_table_get_first_index(sys_foreign));
 start_load:
-	heap = mem_heap_create(256);
 
-	tuple  = dtuple_create(heap, 1);
+	tuple = dtuple_create_from_mem(tuple_buf, sizeof(tuple_buf), 1);
 	dfield = dtuple_get_nth_field(tuple, 0);
 
 	dfield_set_data(dfield, table_name, ut_strlen(table_name));
@@ -2418,7 +2418,6 @@ loop:
 
 	/* Now we get a foreign key constraint id */
 	field = rec_get_nth_field_old(rec, 1, &len);
-	id = mem_heap_strdupl(heap, (char*) field, len);
 
 	btr_pcur_store_position(&pcur, &mtr);
 
@@ -2426,11 +2425,11 @@ loop:
 
 	/* Load the foreign constraint definition to the dictionary cache */
 
-	err = dict_load_foreign(id, check_charsets, check_recursive);
+	err = dict_load_foreign((char*) field, len, check_charsets,
+				check_recursive);
 
 	if (err != DB_SUCCESS) {
 		btr_pcur_close(&pcur);
-		mem_heap_free(heap);
 
 		return(err);
 	}
@@ -2446,7 +2445,6 @@ next_rec:
 load_next_index:
 	btr_pcur_close(&pcur);
 	mtr_commit(&mtr);
-	mem_heap_free(heap);
 
 	sec_index = dict_table_get_next_index(sec_index);
 
diff --git a/storage/innobase/fil/fil0fil.c b/storage/innobase/fil/fil0fil.c
index 196f4bd3f42..2e4c6aeeb60 100644
--- a/storage/innobase/fil/fil0fil.c
+++ b/storage/innobase/fil/fil0fil.c
@@ -1803,36 +1803,44 @@ fil_write_flushed_lsn_to_data_files(
 }
 
 /*******************************************************************//**
-Reads the flushed lsn and arch no fields from a data file at database
-startup. */
+Reads the flushed lsn, arch no, and tablespace flag fields from a data
+file at database startup. */
 UNIV_INTERN
 void
-fil_read_flushed_lsn_and_arch_log_no(
-/*=================================*/
+fil_read_first_page(
+/*================*/
 	os_file_t	data_file,		/*!< in: open data file */
 	ibool		one_read_already,	/*!< in: TRUE if min and max
 						parameters below already
 						contain sensible data */
+	ulint*		flags,			/*!< out: tablespace flags */
 #ifdef UNIV_LOG_ARCHIVE
-	ulint*		min_arch_log_no,	/*!< in/out: */
-	ulint*		max_arch_log_no,	/*!< in/out: */
+	ulint*		min_arch_log_no,	/*!< out: min of archived
+						log numbers in data files */
+	ulint*		max_arch_log_no,	/*!< out: max of archived
+						log numbers in data files */
 #endif /* UNIV_LOG_ARCHIVE */
-	ib_uint64_t*	min_flushed_lsn,	/*!< in/out: */
-	ib_uint64_t*	max_flushed_lsn)	/*!< in/out: */
+	ib_uint64_t*	min_flushed_lsn,	/*!< out: min of flushed
+						lsn values in data files */
+	ib_uint64_t*	max_flushed_lsn)	/*!< out: max of flushed
+						lsn values in data files */
 {
 	byte*		buf;
-	byte*		buf2;
+	page_t*		page;
 	ib_uint64_t	flushed_lsn;
 
-	buf2 = ut_malloc(2 * UNIV_PAGE_SIZE);
+	buf = ut_malloc(2 * UNIV_PAGE_SIZE);
 	/* Align the memory for a possible read from a raw device */
-	buf = ut_align(buf2, UNIV_PAGE_SIZE);
+	page = ut_align(buf, UNIV_PAGE_SIZE);
 
-	os_file_read(data_file, buf, 0, 0, UNIV_PAGE_SIZE);
+	os_file_read(data_file, page, 0, 0, UNIV_PAGE_SIZE);
 
-	flushed_lsn = mach_read_from_8(buf + FIL_PAGE_FILE_FLUSH_LSN);
+	*flags = mach_read_from_4(page +
+		FSP_HEADER_OFFSET + FSP_SPACE_FLAGS);
 
-	ut_free(buf2);
+	flushed_lsn = mach_read_from_8(page + FIL_PAGE_FILE_FLUSH_LSN);
+
+	ut_free(buf);
 
 	if (!one_read_already) {
 		*min_flushed_lsn = flushed_lsn;
diff --git a/storage/innobase/fsp/fsp0fsp.c b/storage/innobase/fsp/fsp0fsp.c
index f31e6c03ba1..2d626405c5c 100644
--- a/storage/innobase/fsp/fsp0fsp.c
+++ b/storage/innobase/fsp/fsp0fsp.c
@@ -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
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -50,67 +50,6 @@ Created 11/29/1995 Heikki Tuuri
 #include "dict0mem.h"
 
 
-#define FSP_HEADER_OFFSET	FIL_PAGE_DATA	/* Offset of the space header
-						within a file page */
-
-/* The data structures in files are defined just as byte strings in C */
-typedef	byte	fsp_header_t;
-typedef	byte	xdes_t;
-
-/*			SPACE HEADER
-			============
-
-File space header data structure: this data structure is contained in the
-first page of a space. The space for this header is reserved in every extent
-descriptor page, but used only in the first. */
-
-/*-------------------------------------*/
-#define FSP_SPACE_ID		0	/* space id */
-#define FSP_NOT_USED		4	/* this field contained a value up to
-					which we know that the modifications
-					in the database have been flushed to
-					the file space; not used now */
-#define	FSP_SIZE		8	/* Current size of the space in
-					pages */
-#define	FSP_FREE_LIMIT		12	/* Minimum page number for which the
-					free list has not been initialized:
-					the pages >= this limit are, by
-					definition, free; note that in a
-					single-table tablespace where size
-					< 64 pages, this number is 64, i.e.,
-					we have initialized the space
-					about the first extent, but have not
-					physically allocted those pages to the
-					file */
-#define	FSP_SPACE_FLAGS		16	/* table->flags & ~DICT_TF_COMPACT */
-#define	FSP_FRAG_N_USED		20	/* number of used pages in the
-					FSP_FREE_FRAG list */
-#define	FSP_FREE		24	/* list of free extents */
-#define	FSP_FREE_FRAG		(24 + FLST_BASE_NODE_SIZE)
-					/* list of partially free extents not
-					belonging to any segment */
-#define	FSP_FULL_FRAG		(24 + 2 * FLST_BASE_NODE_SIZE)
-					/* list of full extents not belonging
-					to any segment */
-#define FSP_SEG_ID		(24 + 3 * FLST_BASE_NODE_SIZE)
-					/* 8 bytes which give the first unused
-					segment id */
-#define FSP_SEG_INODES_FULL	(32 + 3 * FLST_BASE_NODE_SIZE)
-					/* list of pages containing segment
-					headers, where all the segment inode
-					slots are reserved */
-#define FSP_SEG_INODES_FREE	(32 + 4 * FLST_BASE_NODE_SIZE)
-					/* list of pages containing segment
-					headers, where not all the segment
-					header slots are reserved */
-/*-------------------------------------*/
-/* File space header size */
-#define	FSP_HEADER_SIZE		(32 + 5 * FLST_BASE_NODE_SIZE)
-
-#define	FSP_FREE_ADD		4	/* this many free extents are added
-					to the free list from above
-					FSP_FREE_LIMIT at a time */
-
 /*			FILE SEGMENT INODE
 			==================
 
@@ -332,7 +271,7 @@ fseg_alloc_free_page_low(
 				inserted there in order, into which
 				direction they go alphabetically: FSP_DOWN,
 				FSP_UP, FSP_NO_DIR */
-	mtr_t*		mtr);	/*!< in/out: mini-transaction */
+	mtr_t*		mtr);	/*!< in: mtr handle */
 #endif /* !UNIV_HOTBACKUP */
 
 /**********************************************************************//**
@@ -1547,7 +1486,7 @@ fsp_alloc_free_page(
 	ulint	zip_size,/*!< in: compressed page size in bytes
 			or 0 for uncompressed pages */
 	ulint	hint,	/*!< in: hint of which page would be desirable */
-	mtr_t*	mtr)	/*!< in/out: mini-transaction */
+	mtr_t*	mtr)	/*!< in: mtr handle */
 {
 	fsp_header_t*	header;
 	fil_addr_t	first;
@@ -2576,7 +2515,7 @@ fseg_alloc_free_page_low(
 				inserted there in order, into which
 				direction they go alphabetically: FSP_DOWN,
 				FSP_UP, FSP_NO_DIR */
-	mtr_t*		mtr)	/*!< in/out: mini-transaction */
+	mtr_t*		mtr)	/*!< in: mtr handle */
 {
 	fsp_header_t*	space_header;
 	ulint		space_size;
@@ -2824,7 +2763,7 @@ fseg_alloc_free_page_general(
 				with fsp_reserve_free_extents, then there
 				is no need to do the check for this individual
 				page */
-	mtr_t*		mtr)	/*!< in/out: mini-transaction */
+	mtr_t*		mtr)	/*!< in: mtr handle */
 {
 	fseg_inode_t*	inode;
 	ulint		space;
diff --git a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc
index 9a8bd67e8a4..27ad19456cc 100644
--- a/storage/innobase/handler/ha_innodb.cc
+++ b/storage/innobase/handler/ha_innodb.cc
@@ -3790,22 +3790,9 @@ ha_innobase::open(
 		DBUG_RETURN(1);
 	}
 
-	/* Create buffers for packing the fields of a record. Why
-	table->reclength did not work here? Obviously, because char
-	fields when packed actually became 1 byte longer, when we also
-	stored the string length as the first byte. */
-
-	upd_and_key_val_buff_len =
-				table->s->reclength + table->s->max_key_length
-							+ MAX_REF_PARTS * 3;
-	if (!(uchar*) my_multi_malloc(MYF(MY_WME),
-			&upd_buff, upd_and_key_val_buff_len,
-			&key_val_buff, upd_and_key_val_buff_len,
-			NullS)) {
-		free_share(share);
-
-		DBUG_RETURN(1);
-	}
+	/* Will be allocated if it is needed in ::update_row() */
+	upd_buf = NULL;
+	upd_buf_size = 0;
 
 	/* We look for pattern #P# to see if the table is partitioned
 	MySQL table. The retry logic for partitioned tables is a
@@ -3846,7 +3833,6 @@ retry:
 				"how you can resolve the problem.\n",
 				norm_name);
 		free_share(share);
-		my_free(upd_buff);
 		my_errno = ENOENT;
 
 		DBUG_RETURN(HA_ERR_NO_SUCH_TABLE);
@@ -3862,16 +3848,14 @@ retry:
 				"how you can resolve the problem.\n",
 				norm_name);
 		free_share(share);
-		my_free(upd_buff);
 		my_errno = ENOENT;
 
 		dict_table_decrement_handle_count(ib_table, FALSE);
 		DBUG_RETURN(HA_ERR_NO_SUCH_TABLE);
 	}
 
-	prebuilt = row_create_prebuilt(ib_table);
+	prebuilt = row_create_prebuilt(ib_table, table->s->reclength);
 
-	prebuilt->mysql_row_len = table->s->reclength;
 	prebuilt->default_rec = table->s->default_values;
 	ut_ad(prebuilt->default_rec);
 
@@ -4060,7 +4044,13 @@ ha_innobase::close(void)
 
 	row_prebuilt_free(prebuilt, FALSE);
 
-	my_free(upd_buff);
+	if (upd_buf != NULL) {
+		ut_ad(upd_buf_size != 0);
+		my_free(upd_buf);
+		upd_buf = NULL;
+		upd_buf_size = 0;
+	}
+
 	free_share(share);
 
 	/* Tell InnoDB server that there might be work for
@@ -5327,6 +5317,23 @@ ha_innobase::update_row(
 
 	ut_a(prebuilt->trx == trx);
 
+	if (upd_buf == NULL) {
+		ut_ad(upd_buf_size == 0);
+
+		/* Create a buffer for packing the fields of a record. Why
+		table->reclength did not work here? Obviously, because char
+		fields when packed actually became 1 byte longer, when we also
+		stored the string length as the first byte. */
+
+		upd_buf_size = table->s->reclength + table->s->max_key_length
+			+ MAX_REF_PARTS * 3;
+		upd_buf = (uchar*) my_malloc(upd_buf_size, MYF(MY_WME));
+		if (upd_buf == NULL) {
+			upd_buf_size = 0;
+			DBUG_RETURN(HA_ERR_OUT_OF_MEM);
+		}
+	}
+
 	ha_statistic_increment(&SSV::ha_update_count);
 
 	if (table->timestamp_field_type & TIMESTAMP_AUTO_SET_ON_UPDATE)
@@ -5339,11 +5346,10 @@ ha_innobase::update_row(
 	}
 
 	/* Build an update vector from the modified fields in the rows
-	(uses upd_buff of the handle) */
+	(uses upd_buf of the handle) */
 
 	calc_row_difference(uvect, (uchar*) old_row, new_row, table,
-			upd_buff, (ulint)upd_and_key_val_buff_len,
-			prebuilt, user_thd);
+			    upd_buf, upd_buf_size, prebuilt, user_thd);
 
 	/* This is not a delete */
 	prebuilt->upd_node->is_delete = FALSE;
@@ -5720,8 +5726,7 @@ ha_innobase::index_read(
 
 		row_sel_convert_mysql_key_to_innobase(
 			prebuilt->search_tuple,
-			(byte*) key_val_buff,
-			(ulint)upd_and_key_val_buff_len,
+			srch_key_val1, sizeof(srch_key_val1),
 			index,
 			(byte*) key_ptr,
 			(ulint) key_len,
@@ -5819,7 +5824,6 @@ ha_innobase::innobase_get_index(
 	dict_index_t*	index = 0;
 
 	DBUG_ENTER("innobase_get_index");
-	ha_statistic_increment(&SSV::ha_read_key_count);
 
 	if (keynr != MAX_KEY && table->s->keys > 0) {
 		key = table->key_info + keynr;
@@ -5833,13 +5837,13 @@ ha_innobase::innobase_get_index(
 			table. Only print message if the index translation
 			table exists */
 			if (share->idx_trans_tbl.index_mapping) {
-				sql_print_error("InnoDB could not find "
-						"index %s key no %u for "
-						"table %s through its "
-						"index translation table",
-						key ? key->name : "NULL",
-						keynr,
-						prebuilt->table->name);
+				sql_print_warning("InnoDB could not find "
+						  "index %s key no %u for "
+						  "table %s through its "
+						  "index translation table",
+						  key ? key->name : "NULL",
+						  keynr,
+						  prebuilt->table->name);
 			}
 
 			index = dict_table_get_index_on_name(prebuilt->table,
@@ -7541,12 +7545,6 @@ ha_innobase::records_in_range(
 {
 	KEY*		key;
 	dict_index_t*	index;
-	uchar*		key_val_buff2	= (uchar*) my_malloc(
-						  table->s->reclength
-					+ table->s->max_key_length + 100,
-								MYF(MY_FAE));
-	ulint		buff2_len = table->s->reclength
-					+ table->s->max_key_length + 100;
 	dtuple_t*	range_start;
 	dtuple_t*	range_end;
 	ib_int64_t	n_rows;
@@ -7598,8 +7596,8 @@ ha_innobase::records_in_range(
 	dict_index_copy_types(range_end, index, key->key_parts);
 
 	row_sel_convert_mysql_key_to_innobase(
-				range_start, (byte*) key_val_buff,
-				(ulint)upd_and_key_val_buff_len,
+				range_start,
+				srch_key_val1, sizeof(srch_key_val1),
 				index,
 				(byte*) (min_key ? min_key->key :
 					 (const uchar*) 0),
@@ -7610,8 +7608,9 @@ ha_innobase::records_in_range(
 		    : range_start->n_fields == 0);
 
 	row_sel_convert_mysql_key_to_innobase(
-				range_end, (byte*) key_val_buff2,
-				buff2_len, index,
+				range_end,
+				srch_key_val2, sizeof(srch_key_val2),
+				index,
 				(byte*) (max_key ? max_key->key :
 					 (const uchar*) 0),
 				(ulint) (max_key ? max_key->length : 0),
@@ -7638,7 +7637,6 @@ ha_innobase::records_in_range(
 	mem_heap_free(heap);
 
 func_exit:
-	my_free(key_val_buff2);
 
 	prebuilt->trx->op_info = (char*)"";
 
@@ -8322,7 +8320,10 @@ ha_innobase::check(
 		putc('\n', stderr);
 #endif
 
-		if (!btr_validate_index(index, prebuilt->trx)) {
+		/* If this is an index being created, break */
+		if (*index->name == TEMP_INDEX_PREFIX) {
+			break;
+		}  else if (!btr_validate_index(index, prebuilt->trx)) {
 			is_ok = FALSE;
 
 			innobase_format_name(
diff --git a/storage/innobase/handler/ha_innodb.h b/storage/innobase/handler/ha_innodb.h
index e6c9e955827..7cce0c4a16c 100644
--- a/storage/innobase/handler/ha_innodb.h
+++ b/storage/innobase/handler/ha_innodb.h
@@ -78,13 +78,14 @@ class ha_innobase: public handler
 	INNOBASE_SHARE*	share;		/*!< information for MySQL
 					table locking */
 
-	uchar*		upd_buff;	/*!< buffer used in updates */
-	uchar*		key_val_buff;	/*!< buffer used in converting
+	uchar*		upd_buf;	/*!< buffer used in updates */
+	ulint		upd_buf_size;	/*!< the size of upd_buf in bytes */
+	uchar		srch_key_val1[REC_VERSION_56_MAX_INDEX_COL_LEN + 2];
+	uchar		srch_key_val2[REC_VERSION_56_MAX_INDEX_COL_LEN + 2];
+					/*!< buffers used in converting
 					search key values from MySQL format
-					to Innodb format */
-	ulong		upd_and_key_val_buff_len;
-					/* the length of each of the previous
-					two buffers */
+					to InnoDB format. "+ 2" for the two
+					bytes where the length is stored */
 	Table_flags	int_table_flags;
 	uint		primary_key;
 	ulong		start_of_scan;	/*!< this is set to 1 when we are
diff --git a/storage/innobase/handler/handler0alter.cc b/storage/innobase/handler/handler0alter.cc
index 6d5b7b4668f..c6754660b84 100644
--- a/storage/innobase/handler/handler0alter.cc
+++ b/storage/innobase/handler/handler0alter.cc
@@ -1008,7 +1008,12 @@ ha_innobase::final_add_index(
 			row_prebuilt_free(prebuilt, TRUE);
 			error = row_merge_drop_table(trx, old_table);
 			add->indexed_table->n_mysql_handles_opened++;
-			prebuilt = row_create_prebuilt(add->indexed_table);
+			prebuilt = row_create_prebuilt(add->indexed_table,
+				0 /* XXX Do we know the mysql_row_len here?
+				Before the addition of this parameter to
+				row_create_prebuilt() the mysql_row_len
+				member was left 0 (from zalloc) in the
+				prebuilt object. */);
 		}
 
 		err = convert_error_code_to_mysql(
diff --git a/storage/innobase/ibuf/ibuf0ibuf.c b/storage/innobase/ibuf/ibuf0ibuf.c
index 0676a7be0f7..47ec1365cb8 100644
--- a/storage/innobase/ibuf/ibuf0ibuf.c
+++ b/storage/innobase/ibuf/ibuf0ibuf.c
@@ -254,11 +254,20 @@ ibuf_count_check(
 					list of the ibuf */
 /* @} */
 
+#define IBUF_REC_FIELD_SPACE	0	/*!< in the pre-4.1 format,
+					the page number. later, the space_id */
+#define IBUF_REC_FIELD_MARKER	1	/*!< starting with 4.1, a marker
+					consisting of 1 byte that is 0 */
+#define IBUF_REC_FIELD_PAGE	2	/*!< starting with 4.1, the
+					page number */
+#define IBUF_REC_FIELD_METADATA	3	/* the metadata field */
+#define IBUF_REC_FIELD_USER	4	/* first user field */
+
 /* Various constants for checking the type of an ibuf record and extracting
 data from it. For details, see the description of the record format at the
 top of this file. */
 
-/** @name Format of the fourth column of an insert buffer record
+/** @name Format of the IBUF_REC_FIELD_METADATA of an insert buffer record
 The fourth column in the MySQL 5.5 format contains an operation
 type, counter, and some flags. */
 /* @{ */
@@ -1233,13 +1242,13 @@ ibuf_rec_get_page_no_func(
 	ut_ad(ibuf_inside(mtr));
 	ut_ad(rec_get_n_fields_old(rec) > 2);
 
-	field = rec_get_nth_field_old(rec, 1, &len);
+	field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_MARKER, &len);
 
 	if (len == 1) {
 		/* This is of the >= 4.1.x record format */
 		ut_a(trx_sys_multiple_tablespace_format);
 
-		field = rec_get_nth_field_old(rec, 2, &len);
+		field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_PAGE, &len);
 	} else {
 		ut_a(trx_doublewrite_must_reset_space_ids);
 		ut_a(!trx_sys_multiple_tablespace_format);
@@ -1279,13 +1288,13 @@ ibuf_rec_get_space_func(
 	ut_ad(ibuf_inside(mtr));
 	ut_ad(rec_get_n_fields_old(rec) > 2);
 
-	field = rec_get_nth_field_old(rec, 1, &len);
+	field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_MARKER, &len);
 
 	if (len == 1) {
 		/* This is of the >= 4.1.x record format */
 
 		ut_a(trx_sys_multiple_tablespace_format);
-		field = rec_get_nth_field_old(rec, 0, &len);
+		field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_SPACE, &len);
 		ut_a(len == 4);
 
 		return(mach_read_from_4(field));
@@ -1335,9 +1344,9 @@ ibuf_rec_get_info_func(
 	      || mtr_memo_contains_page(mtr, rec, MTR_MEMO_PAGE_S_FIX));
 	ut_ad(ibuf_inside(mtr));
 	fields = rec_get_n_fields_old(rec);
-	ut_a(fields > 4);
+	ut_a(fields > IBUF_REC_FIELD_USER);
 
-	types = rec_get_nth_field_old(rec, 3, &len);
+	types = rec_get_nth_field_old(rec, IBUF_REC_FIELD_METADATA, &len);
 
 	info_len_local = len % DATA_NEW_ORDER_NULL_TYPE_BUF_SIZE;
 
@@ -1363,7 +1372,8 @@ ibuf_rec_get_info_func(
 
 	ut_a(op_local < IBUF_OP_COUNT);
 	ut_a((len - info_len_local) ==
-	     (fields - 4) * DATA_NEW_ORDER_NULL_TYPE_BUF_SIZE);
+	     (fields - IBUF_REC_FIELD_USER)
+	     * DATA_NEW_ORDER_NULL_TYPE_BUF_SIZE);
 
 	if (op) {
 		*op = op_local;
@@ -1407,7 +1417,7 @@ ibuf_rec_get_op_type_func(
 	ut_ad(ibuf_inside(mtr));
 	ut_ad(rec_get_n_fields_old(rec) > 2);
 
-	(void) rec_get_nth_field_old(rec, 1, &len);
+	(void) rec_get_nth_field_old(rec, IBUF_REC_FIELD_MARKER, &len);
 
 	if (len > 1) {
 		/* This is a < 4.1.x format record */
@@ -1436,12 +1446,12 @@ ibuf_rec_get_counter(
 	const byte*	ptr;
 	ulint		len;
 
-	if (rec_get_n_fields_old(rec) < 4) {
+	if (rec_get_n_fields_old(rec) <= IBUF_REC_FIELD_METADATA) {
 
 		return(ULINT_UNDEFINED);
 	}
 
-	ptr = rec_get_nth_field_old(rec, 3, &len);
+	ptr = rec_get_nth_field_old(rec, IBUF_REC_FIELD_METADATA, &len);
 
 	if (len >= 2) {
 
@@ -1666,7 +1676,7 @@ ibuf_build_entry_from_ibuf_rec_func(
 	      || mtr_memo_contains_page(mtr, ibuf_rec, MTR_MEMO_PAGE_S_FIX));
 	ut_ad(ibuf_inside(mtr));
 
-	data = rec_get_nth_field_old(ibuf_rec, 1, &len);
+	data = rec_get_nth_field_old(ibuf_rec, IBUF_REC_FIELD_MARKER, &len);
 
 	if (len > 1) {
 		/* This a < 4.1.x format record */
@@ -1678,13 +1688,13 @@ ibuf_build_entry_from_ibuf_rec_func(
 
 	ut_a(trx_sys_multiple_tablespace_format);
 	ut_a(*data == 0);
-	ut_a(rec_get_n_fields_old(ibuf_rec) > 4);
+	ut_a(rec_get_n_fields_old(ibuf_rec) > IBUF_REC_FIELD_USER);
 
-	n_fields = rec_get_n_fields_old(ibuf_rec) - 4;
+	n_fields = rec_get_n_fields_old(ibuf_rec) - IBUF_REC_FIELD_USER;
 
 	tuple = dtuple_create(heap, n_fields);
 
-	types = rec_get_nth_field_old(ibuf_rec, 3, &len);
+	types = rec_get_nth_field_old(ibuf_rec, IBUF_REC_FIELD_METADATA, &len);
 
 	ibuf_rec_get_info(mtr, ibuf_rec, NULL, &comp, &info_len, NULL);
 
@@ -1698,7 +1708,8 @@ ibuf_build_entry_from_ibuf_rec_func(
 	for (i = 0; i < n_fields; i++) {
 		field = dtuple_get_nth_field(tuple, i);
 
-		data = rec_get_nth_field_old(ibuf_rec, i + 4, &len);
+		data = rec_get_nth_field_old(
+			ibuf_rec, i + IBUF_REC_FIELD_USER, &len);
 
 		dfield_set_data(field, data, len);
 
@@ -1745,7 +1756,7 @@ ibuf_rec_get_size(
 		field_offset = 2;
 		types_offset = DATA_ORDER_NULL_TYPE_BUF_SIZE;
 	} else {
-		field_offset = 4;
+		field_offset = IBUF_REC_FIELD_USER;
 		types_offset = DATA_NEW_ORDER_NULL_TYPE_BUF_SIZE;
 	}
 
@@ -1806,7 +1817,7 @@ ibuf_rec_get_volume_func(
 	ut_ad(ibuf_inside(mtr));
 	ut_ad(rec_get_n_fields_old(ibuf_rec) > 2);
 
-	data = rec_get_nth_field_old(ibuf_rec, 1, &len);
+	data = rec_get_nth_field_old(ibuf_rec, IBUF_REC_FIELD_MARKER, &len);
 	pre_4_1 = (len > 1);
 
 	if (pre_4_1) {
@@ -1829,7 +1840,8 @@ ibuf_rec_get_volume_func(
 		ut_a(trx_sys_multiple_tablespace_format);
 		ut_a(*data == 0);
 
-		types = rec_get_nth_field_old(ibuf_rec, 3, &len);
+		types = rec_get_nth_field_old(
+			ibuf_rec, IBUF_REC_FIELD_METADATA, &len);
 
 		ibuf_rec_get_info(mtr, ibuf_rec, &op, &comp, &info_len, NULL);
 
@@ -1859,7 +1871,8 @@ ibuf_rec_get_volume_func(
 		}
 
 		types += info_len;
-		n_fields = rec_get_n_fields_old(ibuf_rec) - 4;
+		n_fields = rec_get_n_fields_old(ibuf_rec)
+			- IBUF_REC_FIELD_USER;
 	}
 
 	data_size = ibuf_rec_get_size(ibuf_rec, types, n_fields, pre_4_1, comp);
@@ -1914,11 +1927,11 @@ ibuf_entry_build(
 
 	n_fields = dtuple_get_n_fields(entry);
 
-	tuple = dtuple_create(heap, n_fields + 4);
+	tuple = dtuple_create(heap, n_fields + IBUF_REC_FIELD_USER);
 
 	/* 1) Space Id */
 
-	field = dtuple_get_nth_field(tuple, 0);
+	field = dtuple_get_nth_field(tuple, IBUF_REC_FIELD_SPACE);
 
 	buf = mem_heap_alloc(heap, 4);
 
@@ -1928,7 +1941,7 @@ ibuf_entry_build(
 
 	/* 2) Marker byte */
 
-	field = dtuple_get_nth_field(tuple, 1);
+	field = dtuple_get_nth_field(tuple, IBUF_REC_FIELD_MARKER);
 
 	buf = mem_heap_alloc(heap, 1);
 
@@ -1940,7 +1953,7 @@ ibuf_entry_build(
 
 	/* 3) Page number */
 
-	field = dtuple_get_nth_field(tuple, 2);
+	field = dtuple_get_nth_field(tuple, IBUF_REC_FIELD_PAGE);
 
 	buf = mem_heap_alloc(heap, 4);
 
@@ -1988,10 +2001,7 @@ ibuf_entry_build(
 		ulint			fixed_len;
 		const dict_field_t*	ifield;
 
-		/* We add 4 below because we have the 4 extra fields at the
-		start of an ibuf record */
-
-		field = dtuple_get_nth_field(tuple, i + 4);
+		field = dtuple_get_nth_field(tuple, i + IBUF_REC_FIELD_USER);
 		entry_field = dtuple_get_nth_field(entry, i);
 		dfield_copy(field, entry_field);
 
@@ -2024,13 +2034,13 @@ ibuf_entry_build(
 
 	/* 4) Type info, part #2 */
 
-	field = dtuple_get_nth_field(tuple, 3);
+	field = dtuple_get_nth_field(tuple, IBUF_REC_FIELD_METADATA);
 
 	dfield_set_data(field, type_info, ti - type_info);
 
 	/* Set all the types in the new tuple binary */
 
-	dtuple_set_types_binary(tuple, n_fields + 4);
+	dtuple_set_types_binary(tuple, n_fields + IBUF_REC_FIELD_USER);
 
 	return(tuple);
 }
@@ -2090,11 +2100,11 @@ ibuf_new_search_tuple_build(
 
 	ut_a(trx_sys_multiple_tablespace_format);
 
-	tuple = dtuple_create(heap, 3);
+	tuple = dtuple_create(heap, IBUF_REC_FIELD_METADATA);
 
 	/* Store the space id in tuple */
 
-	field = dtuple_get_nth_field(tuple, 0);
+	field = dtuple_get_nth_field(tuple, IBUF_REC_FIELD_SPACE);
 
 	buf = mem_heap_alloc(heap, 4);
 
@@ -2104,7 +2114,7 @@ ibuf_new_search_tuple_build(
 
 	/* Store the new format record marker byte */
 
-	field = dtuple_get_nth_field(tuple, 1);
+	field = dtuple_get_nth_field(tuple, IBUF_REC_FIELD_MARKER);
 
 	buf = mem_heap_alloc(heap, 1);
 
@@ -2114,7 +2124,7 @@ ibuf_new_search_tuple_build(
 
 	/* Store the page number in tuple */
 
-	field = dtuple_get_nth_field(tuple, 2);
+	field = dtuple_get_nth_field(tuple, IBUF_REC_FIELD_PAGE);
 
 	buf = mem_heap_alloc(heap, 4);
 
@@ -2122,7 +2132,7 @@ ibuf_new_search_tuple_build(
 
 	dfield_set_data(field, buf, 4);
 
-	dtuple_set_types_binary(tuple, 3);
+	dtuple_set_types_binary(tuple, IBUF_REC_FIELD_METADATA);
 
 	return(tuple);
 }
@@ -2789,8 +2799,10 @@ ibuf_get_volume_buffered_hash(
 	ulint		fold;
 	ulint		bitmask;
 
-	len = ibuf_rec_get_size(rec, types, rec_get_n_fields_old(rec) - 4,
-				FALSE, comp);
+	len = ibuf_rec_get_size(
+		rec, types,
+		rec_get_n_fields_old(rec) - IBUF_REC_FIELD_USER,
+		FALSE, comp);
 	fold = ut_fold_binary(data, len);
 
 	hash += (fold / (CHAR_BIT * sizeof *hash)) % size;
@@ -2842,8 +2854,8 @@ ibuf_get_volume_buffered_count_func(
 	ut_ad(ibuf_inside(mtr));
 
 	n_fields = rec_get_n_fields_old(rec);
-	ut_ad(n_fields > 4);
-	n_fields -= 4;
+	ut_ad(n_fields > IBUF_REC_FIELD_USER);
+	n_fields -= IBUF_REC_FIELD_USER;
 
 	rec_get_nth_field_offs_old(rec, 1, &len);
 	/* This function is only invoked when buffering new
@@ -2852,7 +2864,7 @@ ibuf_get_volume_buffered_count_func(
 	ut_a(len == 1);
 	ut_ad(trx_sys_multiple_tablespace_format);
 
-	types = rec_get_nth_field_old(rec, 3, &len);
+	types = rec_get_nth_field_old(rec, IBUF_REC_FIELD_METADATA, &len);
 
 	switch (UNIV_EXPECT(len % DATA_NEW_ORDER_NULL_TYPE_BUF_SIZE,
 			    IBUF_REC_INFO_SIZE)) {
@@ -3164,7 +3176,7 @@ ibuf_update_max_tablespace_id(void)
 	} else {
 		rec = btr_pcur_get_rec(&pcur);
 
-		field = rec_get_nth_field_old(rec, 0, &len);
+		field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_SPACE, &len);
 
 		ut_a(len == 4);
 
@@ -3186,10 +3198,12 @@ ibuf_update_max_tablespace_id(void)
 	ibuf_get_entry_counter_low_func(rec,space,page_no)
 #endif
 /****************************************************************//**
-Helper function for ibuf_set_entry_counter. Checks if rec is for (space,
-page_no), and if so, reads counter value from it and returns that + 1.
-Otherwise, returns 0.
-@return	new counter value, or 0 */
+Helper function for ibuf_get_entry_counter_func. Checks if rec is for
+(space, page_no), and if so, reads counter value from it and returns
+that + 1.
+@retval ULINT_UNDEFINED if the record does not contain any counter
+@retval 0 if the record is not for (space, page_no)
+@retval 1 + previous counter value, otherwise */
 static
 ulint
 ibuf_get_entry_counter_low_func(
@@ -3210,7 +3224,7 @@ ibuf_get_entry_counter_low_func(
 	      || mtr_memo_contains_page(mtr, rec, MTR_MEMO_PAGE_S_FIX));
 	ut_ad(rec_get_n_fields_old(rec) > 2);
 
-	field = rec_get_nth_field_old(rec, 1, &len);
+	field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_MARKER, &len);
 
 	if (UNIV_UNLIKELY(len != 1)) {
 		/* pre-4.1 format */
@@ -3223,7 +3237,7 @@ ibuf_get_entry_counter_low_func(
 	ut_a(trx_sys_multiple_tablespace_format);
 
 	/* Check the tablespace identifier. */
-	field = rec_get_nth_field_old(rec, 0, &len);
+	field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_SPACE, &len);
 	ut_a(len == 4);
 
 	if (mach_read_from_4(field) != space) {
@@ -3232,7 +3246,7 @@ ibuf_get_entry_counter_low_func(
 	}
 
 	/* Check the page offset. */
-	field = rec_get_nth_field_old(rec, 2, &len);
+	field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_PAGE, &len);
 	ut_a(len == 4);
 
 	if (mach_read_from_4(field) != page_no) {
@@ -3241,7 +3255,7 @@ ibuf_get_entry_counter_low_func(
 	}
 
 	/* Check if the record contains a counter field. */
-	field = rec_get_nth_field_old(rec, 3, &len);
+	field = rec_get_nth_field_old(rec, IBUF_REC_FIELD_METADATA, &len);
 
 	switch (len % DATA_NEW_ORDER_NULL_TYPE_BUF_SIZE) {
 	default:
@@ -3257,147 +3271,61 @@ ibuf_get_entry_counter_low_func(
 	}
 }
 
+#ifdef UNIV_DEBUG
+# define ibuf_get_entry_counter(space,page_no,rec,mtr,exact_leaf) \
+	ibuf_get_entry_counter_func(space,page_no,rec,mtr,exact_leaf)
+#else /* UNIV_DEBUG */
+# define ibuf_get_entry_counter(space,page_no,rec,mtr,exact_leaf) \
+	ibuf_get_entry_counter_func(space,page_no,rec,exact_leaf)
+#endif
+
 /****************************************************************//**
-Set the counter field in entry to the correct value based on the current
+Calculate the counter field for an entry based on the current
 last record in ibuf for (space, page_no).
-@return	FALSE if we should abort this insertion to ibuf */
+@return	the counter field, or ULINT_UNDEFINED
+if we should abort this insertion to ibuf */
 static
-ibool
-ibuf_set_entry_counter(
-/*===================*/
-	dtuple_t*	entry,		/*!< in/out: entry to patch */
+ulint
+ibuf_get_entry_counter_func(
+/*========================*/
 	ulint		space,		/*!< in: space id of entry */
 	ulint		page_no,	/*!< in: page number of entry */
-	btr_pcur_t*	pcur,		/*!< in: pcur positioned on the record
-					found by btr_pcur_open(.., entry,
-					PAGE_CUR_LE, ..., pcur, ...) */
-	ibool		is_optimistic,	/*!< in: is this an optimistic insert */
-	mtr_t*		mtr)		/*!< in: mtr */
+	const rec_t*	rec,		/*!< in: the record preceding the
+					insertion point */
+#ifdef UNIV_DEBUG
+	mtr_t*		mtr,		/*!< in: mini-transaction */
+#endif /* UNIV_DEBUG */
+	ibool		only_leaf)	/*!< in: TRUE if this is the only
+					leaf page that can contain entries
+					for (space,page_no), that is, there
+					was no exact match for (space,page_no)
+					in the node pointer */
 {
-	dfield_t*	field;
-	byte*		data;
-	ulint		counter = 0;
-
-	/* pcur points to either a user rec or to a page's infimum record. */
 	ut_ad(ibuf_inside(mtr));
-	ut_ad(mtr_memo_contains(mtr, btr_pcur_get_block(pcur),
-				MTR_MEMO_PAGE_X_FIX));
-	ut_ad(page_validate(btr_pcur_get_page(pcur), ibuf->index));
-
-	if (btr_pcur_is_on_user_rec(pcur)) {
-
-		counter = ibuf_get_entry_counter_low(
-			mtr, btr_pcur_get_rec(pcur), space, page_no);
-
-		if (UNIV_UNLIKELY(counter == ULINT_UNDEFINED)) {
-			/* The record lacks a counter field.
-			Such old records must be merged before
-			new records can be buffered. */
-
-			return(FALSE);
-		}
-	} else if (btr_pcur_is_before_first_in_tree(pcur, mtr)) {
-		/* Ibuf tree is either completely empty, or the insert
-		position is at the very first record of a non-empty tree. In
-		either case we have no previous records for (space,
-		page_no). */
-
-		counter = 0;
-	} else if (btr_pcur_is_before_first_on_page(pcur)) {
-		btr_cur_t*	cursor = btr_pcur_get_btr_cur(pcur);
-
-		if (cursor->low_match < 3) {
-			/* If low_match < 3, we know that the father node
-			pointer did not contain the searched for (space,
-			page_no), which means that the search ended on the
-			right page regardless of the counter value, and
-			since we're at the infimum record, there are no
-			existing records. */
-
-			counter = 0;
-		} else {
-			rec_t*		rec;
-			const page_t*	page;
-			buf_block_t*	block;
-			page_t*		prev_page;
-			ulint		prev_page_no;
-
-			ut_a(cursor->ibuf_cnt != ULINT_UNDEFINED);
-
-			page = btr_pcur_get_page(pcur);
-			prev_page_no = btr_page_get_prev(page, mtr);
-
-			ut_a(prev_page_no != FIL_NULL);
-
-			block = buf_page_get(
-				IBUF_SPACE_ID, 0, prev_page_no,
-				RW_X_LATCH, mtr);
+	ut_ad(mtr_memo_contains_page(mtr, rec, MTR_MEMO_PAGE_X_FIX));
+	ut_ad(page_validate(page_align(rec), ibuf->index));
 
-			buf_block_dbg_add_level(block, SYNC_IBUF_TREE_NODE);
-
-			prev_page = buf_block_get_frame(block);
-
-			rec = page_rec_get_prev(
-				page_get_supremum_rec(prev_page));
-
-			ut_ad(page_rec_is_user_rec(rec));
-
-			counter = ibuf_get_entry_counter_low(
-				mtr, rec, space, page_no);
-
-			if (UNIV_UNLIKELY(counter == ULINT_UNDEFINED)) {
-				/* The record lacks a counter field.
-				Such old records must be merged before
-				new records can be buffered. */
-
-				return(FALSE);
-			}
-
-			if (counter < cursor->ibuf_cnt) {
-				/* Search ended on the wrong page. */
-
-				if (is_optimistic) {
-					/* In an optimistic insert, we can
-					shift the insert position to the left
-					page, since it only needs an X-latch
-					on the page itself, which the
-					original search acquired for us. */
-
-					btr_cur_position(
-						ibuf->index, rec, block,
-						btr_pcur_get_btr_cur(pcur));
-				} else {
-					/* We can't shift the insert
-					position to the left page in a
-					pessimistic insert since it would
-					require an X-latch on the left
-					page's left page, so we have to
-					abort. */
-
-					return(FALSE);
-				}
-			} else {
-				/* The counter field in the father node is
-				the same as we would insert; we don't know
-				whether the insert should go to this page or
-				the left page (the later fields can differ),
-				so refuse the insert. */
-
-				return(FALSE);
-			}
-		}
+	if (page_rec_is_supremum(rec)) {
+		/* This is just for safety. The record should be a
+		page infimum or a user record. */
+		ut_ad(0);
+		return(ULINT_UNDEFINED);
+	} else if (!page_rec_is_infimum(rec)) {
+		return(ibuf_get_entry_counter_low(mtr, rec, space, page_no));
+	} else if (only_leaf
+		   || fil_page_get_prev(page_align(rec)) == FIL_NULL) {
+		/* The parent node pointer did not contain the
+		searched for (space, page_no), which means that the
+		search ended on the correct page regardless of the
+		counter value, and since we're at the infimum record,
+		there are no existing records. */
+		return(0);
 	} else {
-		/* The cursor is not positioned at or before a user record. */
-		return(FALSE);
+		/* We used to read the previous page here. It would
+		break the latching order, because the caller has
+		buffer-fixed an insert buffer bitmap page. */
+		return(ULINT_UNDEFINED);
 	}
-
-	/* Patch counter value in already built entry. */
-	field = dtuple_get_nth_field(entry, 3);
-	data = dfield_get_data(field);
-
-	mach_write_to_2(data + IBUF_REC_OFFSET_COUNTER, counter);
-
-	return(TRUE);
 }
 
 /*********************************************************************//**
@@ -3604,16 +3532,27 @@ fail_exit:
 		}
 	}
 
-	/* Patch correct counter value to the entry to insert. This can
-	change the insert position, which can result in the need to abort in
-	some cases. */
-	if (!no_counter
-	    && !ibuf_set_entry_counter(ibuf_entry, space, page_no, &pcur,
-				       mode == BTR_MODIFY_PREV, &mtr)) {
+	if (!no_counter) {
+		/* Patch correct counter value to the entry to
+		insert. This can change the insert position, which can
+		result in the need to abort in some cases. */
+		ulint		counter = ibuf_get_entry_counter(
+			space, page_no, btr_pcur_get_rec(&pcur), &mtr,
+			btr_pcur_get_btr_cur(&pcur)->low_match
+			< IBUF_REC_FIELD_METADATA);
+		dfield_t*	field;
+
+		if (counter == ULINT_UNDEFINED) {
 bitmap_fail:
-		ibuf_mtr_commit(&bitmap_mtr);
+			ibuf_mtr_commit(&bitmap_mtr);
+			goto fail_exit;
+		}
 
-		goto fail_exit;
+		field = dtuple_get_nth_field(
+			ibuf_entry, IBUF_REC_FIELD_METADATA);
+		mach_write_to_2(
+			(byte*) dfield_get_data(field)
+			+ IBUF_REC_OFFSET_COUNTER, counter);
 	}
 
 	/* Set the bitmap bit denoting that the insert buffer contains
diff --git a/storage/innobase/include/btr0cur.h b/storage/innobase/include/btr0cur.h
index be918439f59..4f33aacc48e 100644
--- a/storage/innobase/include/btr0cur.h
+++ b/storage/innobase/include/btr0cur.h
@@ -743,24 +743,6 @@ struct btr_cur_struct {
 					NULL */
 	ulint		fold;		/*!< fold value used in the search if
 					flag is BTR_CUR_HASH */
-	/*----- Delete buffering -------*/
-	ulint		ibuf_cnt;	/* in searches done on insert buffer
-					trees, this contains the "counter"
-					value (the first two bytes of the
-					fourth field) extracted from the
-					page above the leaf page, from the
-					father node pointer that pointed to
-					the leaf page. in other words, it
-					contains the minimum counter value
-					for records to be inserted on the
-					chosen leaf page. If for some reason
-					this can't be read, or if the search
-					ended on the leftmost leaf page in
-					the tree (in which case the father
-					node pointer had the 'minimum
-					record' flag set), this is
-					ULINT_UNDEFINED. */
-	/*------------------------------*/
 	/* @} */
 	btr_path_t*	path_arr;	/*!< in estimating the number of
 					rows in range, we store in this array
diff --git a/storage/innobase/include/btr0pcur.h b/storage/innobase/include/btr0pcur.h
index 140f94466db..2ebd70a6f23 100644
--- a/storage/innobase/include/btr0pcur.h
+++ b/storage/innobase/include/btr0pcur.h
@@ -53,6 +53,16 @@ UNIV_INTERN
 btr_pcur_t*
 btr_pcur_create_for_mysql(void);
 /*============================*/
+
+/**************************************************************//**
+Resets a persistent cursor object, freeing ::old_rec_buf if it is
+allocated and resetting the other members to their initial values. */
+UNIV_INTERN
+void
+btr_pcur_reset(
+/*===========*/
+	btr_pcur_t*	cursor);/*!< in, out: persistent cursor */
+
 /**************************************************************//**
 Frees the memory for a persistent cursor object. */
 UNIV_INTERN
diff --git a/storage/innobase/include/buf0buf.h b/storage/innobase/include/buf0buf.h
index c0ff7b1766b..456f077a13d 100644
--- a/storage/innobase/include/buf0buf.h
+++ b/storage/innobase/include/buf0buf.h
@@ -910,7 +910,27 @@ buf_block_set_io_fix(
 /*=================*/
 	buf_block_t*	block,	/*!< in/out: control block */
 	enum buf_io_fix	io_fix);/*!< in: io_fix state */
-
+/*********************************************************************//**
+Makes a block sticky. A sticky block implies that even after we release
+the buf_pool->mutex and the block->mutex:
+* it cannot be removed from the flush_list
+* the block descriptor cannot be relocated
+* it cannot be removed from the LRU list
+Note that:
+* the block can still change its position in the LRU list
+* the next and previous pointers can change. */
+UNIV_INLINE
+void
+buf_page_set_sticky(
+/*================*/
+	buf_page_t*	bpage);	/*!< in/out: control block */
+/*********************************************************************//**
+Removes stickiness of a block. */
+UNIV_INLINE
+void
+buf_page_unset_sticky(
+/*==================*/
+	buf_page_t*	bpage);	/*!< in/out: control block */
 /********************************************************************//**
 Determine if a buffer block can be relocated in memory.  The block
 can be dirty, but it must not be I/O-fixed or bufferfixed. */
diff --git a/storage/innobase/include/buf0buf.ic b/storage/innobase/include/buf0buf.ic
index b65b5133c15..99e55df3312 100644
--- a/storage/innobase/include/buf0buf.ic
+++ b/storage/innobase/include/buf0buf.ic
@@ -414,6 +414,7 @@ buf_page_get_io_fix(
 	case BUF_IO_NONE:
 	case BUF_IO_READ:
 	case BUF_IO_WRITE:
+	case BUF_IO_PIN:
 		return(io_fix);
 	}
 	ut_error;
@@ -464,6 +465,49 @@ buf_block_set_io_fix(
 	buf_page_set_io_fix(&block->page, io_fix);
 }
 
+/*********************************************************************//**
+Makes a block sticky. A sticky block implies that even after we release
+the buf_pool->mutex and the block->mutex:
+* it cannot be removed from the flush_list
+* the block descriptor cannot be relocated
+* it cannot be removed from the LRU list
+Note that:
+* the block can still change its position in the LRU list
+* the next and previous pointers can change. */
+UNIV_INLINE
+void
+buf_page_set_sticky(
+/*================*/
+	buf_page_t*	bpage)	/*!< in/out: control block */
+{
+#ifdef UNIV_DEBUG
+	buf_pool_t*	buf_pool = buf_pool_from_bpage(bpage);
+	ut_ad(buf_pool_mutex_own(buf_pool));
+#endif
+	ut_ad(mutex_own(buf_page_get_mutex(bpage)));
+	ut_ad(buf_page_get_io_fix(bpage) == BUF_IO_NONE);
+
+	bpage->io_fix = BUF_IO_PIN;
+}
+
+/*********************************************************************//**
+Removes stickiness of a block. */
+UNIV_INLINE
+void
+buf_page_unset_sticky(
+/*==================*/
+	buf_page_t*	bpage)	/*!< in/out: control block */
+{
+#ifdef UNIV_DEBUG
+	buf_pool_t*	buf_pool = buf_pool_from_bpage(bpage);
+	ut_ad(buf_pool_mutex_own(buf_pool));
+#endif
+	ut_ad(mutex_own(buf_page_get_mutex(bpage)));
+	ut_ad(buf_page_get_io_fix(bpage) == BUF_IO_PIN);
+
+	bpage->io_fix = BUF_IO_NONE;
+}
+
 /********************************************************************//**
 Determine if a buffer block can be relocated in memory.  The block
 can be dirty, but it must not be I/O-fixed or bufferfixed. */
diff --git a/storage/innobase/include/buf0types.h b/storage/innobase/include/buf0types.h
index 0cc2defb3ff..12b9e22f673 100644
--- a/storage/innobase/include/buf0types.h
+++ b/storage/innobase/include/buf0types.h
@@ -57,7 +57,10 @@ enum buf_flush {
 enum buf_io_fix {
 	BUF_IO_NONE = 0,		/**< no pending I/O */
 	BUF_IO_READ,			/**< read pending */
-	BUF_IO_WRITE			/**< write pending */
+	BUF_IO_WRITE,			/**< write pending */
+	BUF_IO_PIN			/**< disallow relocation of
+					block and its removal of from
+					the flush_list */
 };
 
 /** Parameters of binary buddy system for compressed pages (buf0buddy.h) */
diff --git a/storage/innobase/include/data0data.h b/storage/innobase/include/data0data.h
index f7bdd29ed90..6d3c2988fdc 100644
--- a/storage/innobase/include/data0data.h
+++ b/storage/innobase/include/data0data.h
@@ -231,6 +231,26 @@ dtuple_set_n_fields_cmp(
 	dtuple_t*	tuple,		/*!< in: tuple */
 	ulint		n_fields_cmp);	/*!< in: number of fields used in
 					comparisons in rem0cmp.* */
+
+/* Estimate the number of bytes that are going to be allocated when
+creating a new dtuple_t object */
+#define DTUPLE_EST_ALLOC(n_fields)	\
+	(sizeof(dtuple_t) + (n_fields) * sizeof(dfield_t))
+
+/**********************************************************//**
+Creates a data tuple from an already allocated chunk of memory.
+The size of the chunk must be at least DTUPLE_EST_ALLOC(n_fields).
+The default value for number of fields used in record comparisons
+for this tuple is n_fields.
+@return	created tuple (inside buf) */
+UNIV_INLINE
+dtuple_t*
+dtuple_create_from_mem(
+/*===================*/
+	void*	buf,		/*!< in, out: buffer to use */
+	ulint	buf_size,	/*!< in: buffer size */
+	ulint	n_fields);	/*!< in: number of fields */
+
 /**********************************************************//**
 Creates a data tuple to a memory heap. The default value for number
 of fields used in record comparisons for this tuple is n_fields.
@@ -240,7 +260,8 @@ dtuple_t*
 dtuple_create(
 /*==========*/
 	mem_heap_t*	heap,	/*!< in: memory heap where the tuple
-				is created */
+				is created, DTUPLE_EST_ALLOC(n_fields)
+				bytes will be allocated from this heap */
 	ulint		n_fields); /*!< in: number of fields */
 
 /**********************************************************//**
diff --git a/storage/innobase/include/data0data.ic b/storage/innobase/include/data0data.ic
index 5c0f8039c80..205fa397987 100644
--- a/storage/innobase/include/data0data.ic
+++ b/storage/innobase/include/data0data.ic
@@ -348,23 +348,25 @@ dtuple_get_nth_field(
 #endif /* UNIV_DEBUG */
 
 /**********************************************************//**
-Creates a data tuple to a memory heap. The default value for number
-of fields used in record comparisons for this tuple is n_fields.
-@return	own: created tuple */
+Creates a data tuple from an already allocated chunk of memory.
+The size of the chunk must be at least DTUPLE_EST_ALLOC(n_fields).
+The default value for number of fields used in record comparisons
+for this tuple is n_fields.
+@return	created tuple (inside buf) */
 UNIV_INLINE
 dtuple_t*
-dtuple_create(
-/*==========*/
-	mem_heap_t*	heap,	/*!< in: memory heap where the tuple
-				is created */
-	ulint		n_fields) /*!< in: number of fields */
+dtuple_create_from_mem(
+/*===================*/
+	void*	buf,		/*!< in, out: buffer to use */
+	ulint	buf_size,	/*!< in: buffer size */
+	ulint	n_fields)	/*!< in: number of fields */
 {
 	dtuple_t*	tuple;
 
-	ut_ad(heap);
+	ut_ad(buf != NULL);
+	ut_a(buf_size >= DTUPLE_EST_ALLOC(n_fields));
 
-	tuple = (dtuple_t*) mem_heap_alloc(heap, sizeof(dtuple_t)
-					   + n_fields * sizeof(dfield_t));
+	tuple = (dtuple_t*) buf;
 	tuple->info_bits = 0;
 	tuple->n_fields = n_fields;
 	tuple->n_fields_cmp = n_fields;
@@ -386,9 +388,38 @@ dtuple_create(
 			dfield_get_type(field)->mtype = DATA_ERROR;
 		}
 	}
+#endif
+	return(tuple);
+}
+
+/**********************************************************//**
+Creates a data tuple to a memory heap. The default value for number
+of fields used in record comparisons for this tuple is n_fields.
+@return	own: created tuple */
+UNIV_INLINE
+dtuple_t*
+dtuple_create(
+/*==========*/
+	mem_heap_t*	heap,	/*!< in: memory heap where the tuple
+				is created, DTUPLE_EST_ALLOC(n_fields)
+				bytes will be allocated from this heap */
+	ulint		n_fields) /*!< in: number of fields */
+{
+	void*		buf;
+	ulint		buf_size;
+	dtuple_t*	tuple;
+
+	ut_ad(heap);
+
+	buf_size = DTUPLE_EST_ALLOC(n_fields);
+	buf = mem_heap_alloc(heap, buf_size);
 
+	tuple = dtuple_create_from_mem(buf, buf_size, n_fields);
+
+#ifdef UNIV_DEBUG
 	UNIV_MEM_INVALID(tuple->fields, n_fields * sizeof *tuple->fields);
 #endif
+
 	return(tuple);
 }
 
diff --git a/storage/innobase/include/fil0fil.h b/storage/innobase/include/fil0fil.h
index 3a6336f1a01..d50b0cb4162 100644
--- a/storage/innobase/include/fil0fil.h
+++ b/storage/innobase/include/fil0fil.h
@@ -328,18 +328,23 @@ Reads the flushed lsn and arch no fields from a data file at database
 startup. */
 UNIV_INTERN
 void
-fil_read_flushed_lsn_and_arch_log_no(
-/*=================================*/
+fil_read_first_page(
+/*================*/
 	os_file_t	data_file,		/*!< in: open data file */
 	ibool		one_read_already,	/*!< in: TRUE if min and max
 						parameters below already
 						contain sensible data */
+	ulint*		flags,			/*!< out: tablespace flags */
 #ifdef UNIV_LOG_ARCHIVE
-	ulint*		min_arch_log_no,	/*!< in/out: */
-	ulint*		max_arch_log_no,	/*!< in/out: */
+	ulint*		min_arch_log_no,	/*!< out: min of archived
+						log numbers in data files */
+	ulint*		max_arch_log_no,	/*!< out: max of archived
+						log numbers in data files */
 #endif /* UNIV_LOG_ARCHIVE */
-	ib_uint64_t*	min_flushed_lsn,	/*!< in/out: */
-	ib_uint64_t*	max_flushed_lsn);	/*!< in/out: */
+	ib_uint64_t*	min_flushed_lsn,	/*!< out: min of flushed
+						lsn values in data files */
+	ib_uint64_t*	max_flushed_lsn);	/*!< out: max of flushed
+						lsn values in data files */
 /*******************************************************************//**
 Increments the count of pending insert buffer page merges, if space is not
 being deleted.
diff --git a/storage/innobase/include/fsp0fsp.h b/storage/innobase/include/fsp0fsp.h
index 6e70fbf6f66..4efabacb2cb 100644
--- a/storage/innobase/include/fsp0fsp.h
+++ b/storage/innobase/include/fsp0fsp.h
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1995, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2011, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -34,6 +34,90 @@ Created 12/18/1995 Heikki Tuuri
 #include "page0types.h"
 #include "fsp0types.h"
 
+/* @defgroup fsp_flags InnoDB Tablespace Flag Constants @{ */
+
+/** Number of flag bits used to indicate the tablespace page size */
+#define FSP_FLAGS_WIDTH_PAGE_SSIZE	4
+/** Zero relative shift position of the PAGE_SSIZE field */
+#define FSP_FLAGS_POS_PAGE_SSIZE	6
+/** Bit mask of the PAGE_SSIZE field */
+#define FSP_FLAGS_MASK_PAGE_SSIZE				\
+		((~(~0 << FSP_FLAGS_WIDTH_PAGE_SSIZE))		\
+		<< FSP_FLAGS_POS_PAGE_SSIZE)
+/** Return the value of the PAGE_SSIZE field */
+#define FSP_FLAGS_GET_PAGE_SSIZE(flags)				\
+		((flags & FSP_FLAGS_MASK_PAGE_SSIZE)		\
+		>> FSP_FLAGS_POS_PAGE_SSIZE)
+
+/* @} */
+
+/* @defgroup Tablespace Header Constants (moved from fsp0fsp.c) @{ */
+
+/** Offset of the space header within a file page */
+#define FSP_HEADER_OFFSET	FIL_PAGE_DATA
+
+/* The data structures in files are defined just as byte strings in C */
+typedef	byte	fsp_header_t;
+typedef	byte	xdes_t;
+
+/*			SPACE HEADER
+			============
+
+File space header data structure: this data structure is contained in the
+first page of a space. The space for this header is reserved in every extent
+descriptor page, but used only in the first. */
+
+/*-------------------------------------*/
+#define FSP_SPACE_ID		0	/* space id */
+#define FSP_NOT_USED		4	/* this field contained a value up to
+					which we know that the modifications
+					in the database have been flushed to
+					the file space; not used now */
+#define	FSP_SIZE		8	/* Current size of the space in
+					pages */
+#define	FSP_FREE_LIMIT		12	/* Minimum page number for which the
+					free list has not been initialized:
+					the pages >= this limit are, by
+					definition, free; note that in a
+					single-table tablespace where size
+					< 64 pages, this number is 64, i.e.,
+					we have initialized the space
+					about the first extent, but have not
+					physically allocted those pages to the
+					file */
+#define	FSP_SPACE_FLAGS		16	/* fsp_space_t.flags, similar to
+					dict_table_t::flags */
+#define	FSP_FRAG_N_USED		20	/* number of used pages in the
+					FSP_FREE_FRAG list */
+#define	FSP_FREE		24	/* list of free extents */
+#define	FSP_FREE_FRAG		(24 + FLST_BASE_NODE_SIZE)
+					/* list of partially free extents not
+					belonging to any segment */
+#define	FSP_FULL_FRAG		(24 + 2 * FLST_BASE_NODE_SIZE)
+					/* list of full extents not belonging
+					to any segment */
+#define FSP_SEG_ID		(24 + 3 * FLST_BASE_NODE_SIZE)
+					/* 8 bytes which give the first unused
+					segment id */
+#define FSP_SEG_INODES_FULL	(32 + 3 * FLST_BASE_NODE_SIZE)
+					/* list of pages containing segment
+					headers, where all the segment inode
+					slots are reserved */
+#define FSP_SEG_INODES_FREE	(32 + 4 * FLST_BASE_NODE_SIZE)
+					/* list of pages containing segment
+					headers, where not all the segment
+					header slots are reserved */
+/*-------------------------------------*/
+/* File space header size */
+#define	FSP_HEADER_SIZE		(32 + 5 * FLST_BASE_NODE_SIZE)
+
+#define	FSP_FREE_ADD		4	/* this many free extents are added
+					to the free list from above
+					FSP_FREE_LIMIT at a time */
+/* @} */
+
+/* @} */
+
 /**********************************************************************//**
 Initializes the file space system. */
 UNIV_INTERN
@@ -353,6 +437,18 @@ fseg_print(
 	mtr_t*		mtr);	/*!< in: mtr */
 #endif /* UNIV_BTR_PRINT */
 
+/********************************************************************//**
+Extract the page size from tablespace flags.
+This feature, storing the page_ssize into the tablespace flags, is added
+to InnoDB 5.6.4.  This is here only to protect against a crash if a newer
+database is opened with this code branch.
+@return	page size of the tablespace in bytes */
+UNIV_INLINE
+ulint
+fsp_flags_get_page_size(
+/*====================*/
+	ulint	flags);	/*!< in: tablespace flags */
+
 #ifndef UNIV_NONINL
 #include "fsp0fsp.ic"
 #endif
diff --git a/storage/innobase/include/fsp0fsp.ic b/storage/innobase/include/fsp0fsp.ic
index 434c370b527..c92111a9d89 100644
--- a/storage/innobase/include/fsp0fsp.ic
+++ b/storage/innobase/include/fsp0fsp.ic
@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1995, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1995, 2011, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc., 59 Temple
-Place, Suite 330, Boston, MA 02111-1307 USA
+this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
 
 *****************************************************************************/
 
@@ -43,3 +43,31 @@ fsp_descr_page(
 
 	return(UNIV_UNLIKELY((page_no & (zip_size - 1)) == FSP_XDES_OFFSET));
 }
+/********************************************************************//**
+Extract the page size from tablespace flags.
+This feature, storing the page_ssize into the tablespace flags, is added
+to InnoDB 5.6.4.  This is here only to protect against a crash if a newer
+database is opened with this code branch.
+@return	page size of the tablespace in bytes */
+UNIV_INLINE
+ulint
+fsp_flags_get_page_size(
+/*====================*/
+	ulint	flags)	/*!< in: tablespace flags */
+{
+	ulint	page_size = 0;
+	ulint	ssize = FSP_FLAGS_GET_PAGE_SSIZE(flags);
+
+	/* Convert from a 'log2 minus 9' to a page size in bytes. */
+	if (UNIV_UNLIKELY(ssize)) {
+		page_size = (512 << ssize);
+
+		ut_ad(page_size <= UNIV_PAGE_SIZE);
+	} else {
+		/* If the page size was not stored, then it is the
+		original 16k. */
+		page_size = UNIV_PAGE_SIZE;
+	}
+
+	return(page_size);
+}
diff --git a/storage/innobase/include/mem0mem.ic b/storage/innobase/include/mem0mem.ic
index d214c3fe6c9..c70615e1ca9 100644
--- a/storage/innobase/include/mem0mem.ic
+++ b/storage/innobase/include/mem0mem.ic
@@ -209,10 +209,6 @@ mem_heap_alloc(
 	buf = (byte*)buf + MEM_FIELD_HEADER_SIZE;
 
 #endif
-#ifdef UNIV_SET_MEM_TO_ZERO
-	UNIV_MEM_ALLOC(buf, n);
-	memset(buf, '\0', n);
-#endif
 	UNIV_MEM_ALLOC(buf, n);
 	return(buf);
 }
diff --git a/storage/innobase/include/mtr0mtr.ic b/storage/innobase/include/mtr0mtr.ic
index 960c25d6051..1db4a4bd735 100644
--- a/storage/innobase/include/mtr0mtr.ic
+++ b/storage/innobase/include/mtr0mtr.ic
@@ -44,6 +44,7 @@ mtr_start(
 
 	mtr->log_mode = MTR_LOG_ALL;
 	mtr->modifications = FALSE;
+	mtr->inside_ibuf = FALSE;
 	mtr->n_log_recs = 0;
 
 	ut_d(mtr->state = MTR_ACTIVE);
diff --git a/storage/innobase/include/page0page.h b/storage/innobase/include/page0page.h
index 6a82e820312..540ec7855bc 100644
--- a/storage/innobase/include/page0page.h
+++ b/storage/innobase/include/page0page.h
@@ -893,6 +893,7 @@ page_parse_create(
 	ulint		comp,	/*!< in: nonzero=compact page format */
 	buf_block_t*	block,	/*!< in: block or NULL */
 	mtr_t*		mtr);	/*!< in: mtr or NULL */
+#ifndef UNIV_HOTBACKUP
 /************************************************************//**
 Prints record contents including the data relevant only in
 the index page context. */
@@ -902,6 +903,7 @@ page_rec_print(
 /*===========*/
 	const rec_t*	rec,	/*!< in: physical record */
 	const ulint*	offsets);/*!< in: record descriptor */
+# ifdef UNIV_BTR_PRINT
 /***************************************************************//**
 This is used to print the contents of the directory for
 debugging purposes. */
@@ -941,6 +943,8 @@ page_print(
 				in directory */
 	ulint		rn);	/*!< in: print rn first and last records
 				in directory */
+# endif /* UNIV_BTR_PRINT */
+#endif /* !UNIV_HOTBACKUP */
 /***************************************************************//**
 The following is used to validate a record on a page. This function
 differs from rec_validate as it can also check the n_owned field and
diff --git a/storage/innobase/include/row0mysql.h b/storage/innobase/include/row0mysql.h
index dd619406ab9..e17fd584110 100644
--- a/storage/innobase/include/row0mysql.h
+++ b/storage/innobase/include/row0mysql.h
@@ -168,7 +168,9 @@ UNIV_INTERN
 row_prebuilt_t*
 row_create_prebuilt(
 /*================*/
-	dict_table_t*	table);	/*!< in: Innobase table handle */
+	dict_table_t*	table,		/*!< in: Innobase table handle */
+	ulint		mysql_row_len);	/*!< in: length in bytes of a row in
+					the MySQL format */
 /********************************************************************//**
 Free a prebuilt struct for a MySQL table handle. */
 UNIV_INTERN
@@ -672,9 +674,9 @@ struct row_prebuilt_struct {
 					in inserts */
 	que_fork_t*	upd_graph;	/*!< Innobase SQL query graph used
 					in updates or deletes */
-	btr_pcur_t*	pcur;		/*!< persistent cursor used in selects
+	btr_pcur_t	pcur;		/*!< persistent cursor used in selects
 					and updates */
-	btr_pcur_t*	clust_pcur;	/*!< persistent cursor used in
+	btr_pcur_t	clust_pcur;	/*!< persistent cursor used in
 					some selects and updates */
 	que_fork_t*	sel_graph;	/*!< dummy query graph used in
 					selects */
diff --git a/storage/innobase/include/row0sel.h b/storage/innobase/include/row0sel.h
index 8544b9d08ba..1c4ea6f7244 100644
--- a/storage/innobase/include/row0sel.h
+++ b/storage/innobase/include/row0sel.h
@@ -128,7 +128,12 @@ row_sel_convert_mysql_key_to_innobase(
 					in the tuple is already according
 					to index! */
 	byte*		buf,		/*!< in: buffer to use in field
-					conversions */
+					conversions; NOTE that dtuple->data
+					may end up pointing inside buf so
+					do not discard that buffer while
+					the tuple is being used. See
+					row_mysql_store_col_in_innobase_format()
+					in the case of DATA_INT */
 	ulint		buf_len,	/*!< in: buffer length */
 	dict_index_t*	index,		/*!< in: index of the key value */
 	const byte*	key_ptr,	/*!< in: MySQL key value */
diff --git a/storage/innobase/include/univ.i b/storage/innobase/include/univ.i
index 21d1eb03b99..50837d262c3 100644
--- a/storage/innobase/include/univ.i
+++ b/storage/innobase/include/univ.i
@@ -154,14 +154,6 @@ resolved */
 /*			DEBUG VERSION CONTROL
 			===================== */
 
-/* The following flag will make InnoDB to initialize
-all memory it allocates to zero. It hides Purify
-warnings about reading unallocated memory unless
-memory is read outside the allocated blocks. */
-/*
-#define UNIV_INIT_MEM_TO_ZERO
-*/
-
 /* When this macro is defined then additional test functions will be
 compiled. These functions live at the end of each relevant source file
 and have "test_" prefix. These functions are not called from anywhere in
@@ -231,15 +223,6 @@ operations (very slow); also UNIV_DEBUG must be defined */
 #define UNIV_BTR_DEBUG				/* check B-tree links */
 #define UNIV_LIGHT_MEM_DEBUG			/* light memory debugging */
 
-#ifdef HAVE_purify
-/* The following sets all new allocated memory to zero before use:
-this can be used to eliminate unnecessary Purify warnings, but note that
-it also masks many bugs Purify could detect. For detailed Purify analysis it
-is best to remove the define below and look through the warnings one
-by one. */
-#define UNIV_SET_MEM_TO_ZERO
-#endif
-
 /*
 #define UNIV_SQL_DEBUG
 #define UNIV_LOG_DEBUG
diff --git a/storage/innobase/include/ut0mem.h b/storage/innobase/include/ut0mem.h
index faf6f242883..39f5f20dc6d 100644
--- a/storage/innobase/include/ut0mem.h
+++ b/storage/innobase/include/ut0mem.h
@@ -78,40 +78,19 @@ ut_mem_init(void);
 /*=============*/
 
 /**********************************************************************//**
-Allocates memory. Sets it also to zero if UNIV_SET_MEM_TO_ZERO is
-defined and set_to_zero is TRUE.
+Allocates memory.
 @return	own: allocated memory */
 UNIV_INTERN
 void*
 ut_malloc_low(
 /*==========*/
 	ulint	n,			/*!< in: number of bytes to allocate */
-	ibool	set_to_zero,		/*!< in: TRUE if allocated memory
-					should be set to zero if
-					UNIV_SET_MEM_TO_ZERO is defined */
-	ibool	assert_on_error);	/*!< in: if TRUE, we crash mysqld if
+	ibool	assert_on_error)	/*!< in: if TRUE, we crash mysqld if
 					the memory cannot be allocated */
+	__attribute__((malloc));
 /**********************************************************************//**
-Allocates memory. Sets it also to zero if UNIV_SET_MEM_TO_ZERO is
-defined.
-@return	own: allocated memory */
-UNIV_INTERN
-void*
-ut_malloc(
-/*======*/
-	ulint	n);	/*!< in: number of bytes to allocate */
-#ifndef UNIV_HOTBACKUP
-/**********************************************************************//**
-Tests if malloc of n bytes would succeed. ut_malloc() asserts if memory runs
-out. It cannot be used if we want to return an error message. Prints to
-stderr a message if fails.
-@return	TRUE if succeeded */
-UNIV_INTERN
-ibool
-ut_test_malloc(
-/*===========*/
-	ulint	n);	/*!< in: try to allocate this many bytes */
-#endif /* !UNIV_HOTBACKUP */
+Allocates memory. */
+#define ut_malloc(n) ut_malloc_low(n, TRUE)
 /**********************************************************************//**
 Frees a memory block allocated with ut_malloc. Freeing a NULL pointer is
 a nop. */
diff --git a/storage/innobase/include/ut0rnd.ic b/storage/innobase/include/ut0rnd.ic
index df99c384bbe..795b8ab7a85 100644
--- a/storage/innobase/include/ut0rnd.ic
+++ b/storage/innobase/include/ut0rnd.ic
@@ -114,7 +114,7 @@ ut_rnd_interval(
 
 	rnd = ut_rnd_gen_ulint();
 
-	return(low + (rnd % (high - low + 1)));
+	return(low + (rnd % (high - low)));
 }
 
 /*********************************************************//**
diff --git a/storage/innobase/mem/mem0pool.c b/storage/innobase/mem/mem0pool.c
index dc68cf8eb24..50dbe526d64 100644
--- a/storage/innobase/mem/mem0pool.c
+++ b/storage/innobase/mem/mem0pool.c
@@ -228,11 +228,7 @@ mem_pool_create(
 
 	pool = ut_malloc(sizeof(mem_pool_t));
 
-	/* We do not set the memory to zero (FALSE) in the pool,
-	but only when allocated at a higher level in mem0mem.c.
-	This is to avoid masking useful Purify warnings. */
-
-	pool->buf = ut_malloc_low(size, FALSE, TRUE);
+	pool->buf = ut_malloc_low(size, TRUE);
 	pool->size = size;
 
 	mutex_create(mem_pool_mutex_key, &pool->mutex, SYNC_MEM_POOL);
diff --git a/storage/innobase/os/os0proc.c b/storage/innobase/os/os0proc.c
index 0f56a608f38..68321e1aaf9 100644
--- a/storage/innobase/os/os0proc.c
+++ b/storage/innobase/os/os0proc.c
@@ -111,9 +111,6 @@ os_mem_alloc_large(
 		os_fast_mutex_lock(&ut_list_mutex);
 		ut_total_allocated_memory += size;
 		os_fast_mutex_unlock(&ut_list_mutex);
-# ifdef UNIV_SET_MEM_TO_ZERO
-		memset(ptr, '\0', size);
-# endif
 		UNIV_MEM_ALLOC(ptr, size);
 		return(ptr);
 	}
diff --git a/storage/innobase/page/page0page.c b/storage/innobase/page/page0page.c
index 102274d66f3..4858929082a 100644
--- a/storage/innobase/page/page0page.c
+++ b/storage/innobase/page/page0page.c
@@ -1591,13 +1591,14 @@ page_rec_print(
 			" n_owned: %lu; heap_no: %lu; next rec: %lu\n",
 			(ulong) rec_get_n_owned_old(rec),
 			(ulong) rec_get_heap_no_old(rec),
-			(ulong) rec_get_next_offs(rec, TRUE));
+			(ulong) rec_get_next_offs(rec, FALSE));
 	}
 
 	page_rec_check(rec);
 	rec_validate(rec, offsets);
 }
 
+# ifdef UNIV_BTR_PRINT
 /***************************************************************//**
 This is used to print the contents of the directory for
 debugging purposes. */
@@ -1758,6 +1759,7 @@ page_print(
 	page_dir_print(page, dn);
 	page_print_list(block, index, rn);
 }
+# endif /* UNIV_BTR_PRINT */
 #endif /* !UNIV_HOTBACKUP */
 
 /***************************************************************//**
diff --git a/storage/innobase/pars/pars0pars.c b/storage/innobase/pars/pars0pars.c
index ef107f2896f..86f54195682 100644
--- a/storage/innobase/pars/pars0pars.c
+++ b/storage/innobase/pars/pars0pars.c
@@ -1857,7 +1857,7 @@ pars_sql(
 
 	ut_ad(str);
 
-	heap = mem_heap_create(256);
+	heap = mem_heap_create(16000);
 
 	/* Currently, the parser is not reentrant: */
 	ut_ad(mutex_own(&(dict_sys->mutex)));
diff --git a/storage/innobase/row/row0mysql.c b/storage/innobase/row/row0mysql.c
index d06411e09f0..996a49f76e8 100644
--- a/storage/innobase/row/row0mysql.c
+++ b/storage/innobase/row/row0mysql.c
@@ -289,21 +289,21 @@ row_mysql_pad_col(
 		/* space=0x0020 */
 		pad_end = pad + len;
 		ut_a(!(len % 2));
-		do {
+		while (pad < pad_end) {
 			*pad++ = 0x00;
 			*pad++ = 0x20;
-		} while (pad < pad_end);
+		};
 		break;
 	case 4:
 		/* space=0x00000020 */
 		pad_end = pad + len;
 		ut_a(!(len % 4));
-		do {
+		while (pad < pad_end) {
 			*pad++ = 0x00;
 			*pad++ = 0x00;
 			*pad++ = 0x00;
 			*pad++ = 0x20;
-		} while (pad < pad_end);
+		}
 		break;
 	}
 }
@@ -667,17 +667,60 @@ UNIV_INTERN
 row_prebuilt_t*
 row_create_prebuilt(
 /*================*/
-	dict_table_t*	table)	/*!< in: Innobase table handle */
+	dict_table_t*	table,		/*!< in: Innobase table handle */
+	ulint		mysql_row_len)	/*!< in: length in bytes of a row in
+					the MySQL format */
 {
 	row_prebuilt_t*	prebuilt;
 	mem_heap_t*	heap;
 	dict_index_t*	clust_index;
 	dtuple_t*	ref;
 	ulint		ref_len;
+	ulint		search_tuple_n_fields;
+
+	search_tuple_n_fields = 2 * dict_table_get_n_cols(table);
+
+	clust_index = dict_table_get_first_index(table);
 
-	heap = mem_heap_create(sizeof *prebuilt + 128);
+	/* Make sure that search_tuple is long enough for clustered index */
+	ut_a(2 * dict_table_get_n_cols(table) >= clust_index->n_fields);
+
+	ref_len = dict_index_get_n_unique(clust_index);
 
-	prebuilt = mem_heap_zalloc(heap, sizeof *prebuilt);
+#define PREBUILT_HEAP_INITIAL_SIZE	\
+	( \
+	sizeof(*prebuilt) \
+	/* allocd in this function */ \
+	+ DTUPLE_EST_ALLOC(search_tuple_n_fields) \
+	+ DTUPLE_EST_ALLOC(ref_len) \
+	/* allocd in row_prebuild_sel_graph() */ \
+	+ sizeof(sel_node_t) \
+	+ sizeof(que_fork_t) \
+	+ sizeof(que_thr_t) \
+	/* allocd in row_get_prebuilt_update_vector() */ \
+	+ sizeof(upd_node_t) \
+	+ sizeof(upd_t) \
+	+ sizeof(upd_field_t) \
+	  * dict_table_get_n_cols(table) \
+	+ sizeof(que_fork_t) \
+	+ sizeof(que_thr_t) \
+	/* allocd in row_get_prebuilt_insert_row() */ \
+	+ sizeof(ins_node_t) \
+	/* mysql_row_len could be huge and we are not \
+	sure if this prebuilt instance is going to be \
+	used in inserts */ \
+	+ (mysql_row_len < 256 ? mysql_row_len : 0) \
+	+ DTUPLE_EST_ALLOC(dict_table_get_n_cols(table)) \
+	+ sizeof(que_fork_t) \
+	+ sizeof(que_thr_t) \
+	)
+
+	/* We allocate enough space for the objects that are likely to
+	be created later in order to minimize the number of malloc()
+	calls */
+	heap = mem_heap_create(PREBUILT_HEAP_INITIAL_SIZE);
+
+	prebuilt = mem_heap_zalloc(heap, sizeof(*prebuilt));
 
 	prebuilt->magic_n = ROW_PREBUILT_ALLOCATED;
 	prebuilt->magic_n2 = ROW_PREBUILT_ALLOCATED;
@@ -687,23 +730,15 @@ row_create_prebuilt(
 	prebuilt->sql_stat_start = TRUE;
 	prebuilt->heap = heap;
 
-	prebuilt->pcur = btr_pcur_create_for_mysql();
-	prebuilt->clust_pcur = btr_pcur_create_for_mysql();
+	btr_pcur_reset(&prebuilt->pcur);
+	btr_pcur_reset(&prebuilt->clust_pcur);
 
 	prebuilt->select_lock_type = LOCK_NONE;
 	prebuilt->stored_select_lock_type = 99999999;
 	UNIV_MEM_INVALID(&prebuilt->stored_select_lock_type,
 			 sizeof prebuilt->stored_select_lock_type);
 
-	prebuilt->search_tuple = dtuple_create(
-		heap, 2 * dict_table_get_n_cols(table));
-
-	clust_index = dict_table_get_first_index(table);
-
-	/* Make sure that search_tuple is long enough for clustered index */
-	ut_a(2 * dict_table_get_n_cols(table) >= clust_index->n_fields);
-
-	ref_len = dict_index_get_n_unique(clust_index);
+	prebuilt->search_tuple = dtuple_create(heap, search_tuple_n_fields);
 
 	ref = dtuple_create(heap, ref_len);
 
@@ -720,6 +755,8 @@ row_create_prebuilt(
 
 	prebuilt->autoinc_last_value = 0;
 
+	prebuilt->mysql_row_len = mysql_row_len;
+
 	return(prebuilt);
 }
 
@@ -755,8 +792,8 @@ row_prebuilt_free(
 	prebuilt->magic_n = ROW_PREBUILT_FREED;
 	prebuilt->magic_n2 = ROW_PREBUILT_FREED;
 
-	btr_pcur_free_for_mysql(prebuilt->pcur);
-	btr_pcur_free_for_mysql(prebuilt->clust_pcur);
+	btr_pcur_reset(&prebuilt->pcur);
+	btr_pcur_reset(&prebuilt->clust_pcur);
 
 	if (prebuilt->mysql_template) {
 		mem_free(prebuilt->mysql_template);
@@ -1416,11 +1453,11 @@ row_update_for_mysql(
 
 	clust_index = dict_table_get_first_index(table);
 
-	if (prebuilt->pcur->btr_cur.index == clust_index) {
-		btr_pcur_copy_stored_position(node->pcur, prebuilt->pcur);
+	if (prebuilt->pcur.btr_cur.index == clust_index) {
+		btr_pcur_copy_stored_position(node->pcur, &prebuilt->pcur);
 	} else {
 		btr_pcur_copy_stored_position(node->pcur,
-					      prebuilt->clust_pcur);
+					      &prebuilt->clust_pcur);
 	}
 
 	ut_a(node->pcur->rel_pos == BTR_PCUR_ON);
@@ -1524,8 +1561,8 @@ row_unlock_for_mysql(
 					clust_pcur, and we do not need
 					to reposition the cursors. */
 {
-	btr_pcur_t*	pcur		= prebuilt->pcur;
-	btr_pcur_t*	clust_pcur	= prebuilt->clust_pcur;
+	btr_pcur_t*	pcur		= &prebuilt->pcur;
+	btr_pcur_t*	clust_pcur	= &prebuilt->clust_pcur;
 	trx_t*		trx		= prebuilt->trx;
 
 	ut_ad(prebuilt && trx);
diff --git a/storage/innobase/row/row0sel.c b/storage/innobase/row/row0sel.c
index 5d8f53f68da..20d45c1884d 100644
--- a/storage/innobase/row/row0sel.c
+++ b/storage/innobase/row/row0sel.c
@@ -2301,7 +2301,12 @@ row_sel_convert_mysql_key_to_innobase(
 					in the tuple is already according
 					to index! */
 	byte*		buf,		/*!< in: buffer to use in field
-					conversions */
+					conversions; NOTE that dtuple->data
+					may end up pointing inside buf so
+					do not discard that buffer while
+					the tuple is being used. See
+					row_mysql_store_col_in_innobase_format()
+					in the case of DATA_INT */
 	ulint		buf_len,	/*!< in: buffer length */
 	dict_index_t*	index,		/*!< in: index of the key value */
 	const byte*	key_ptr,	/*!< in: MySQL key value */
@@ -2433,6 +2438,7 @@ row_sel_convert_mysql_key_to_innobase(
 		/* Storing may use at most data_len bytes of buf */
 
 		if (UNIV_LIKELY(!is_null)) {
+			ut_a(buf + data_len <= original_buf + buf_len);
 			row_mysql_store_col_in_innobase_format(
 				dfield, buf,
 				FALSE, /* MySQL key value format col */
@@ -2915,17 +2921,17 @@ row_sel_get_clust_rec_for_mysql(
 
 	btr_pcur_open_with_no_init(clust_index, prebuilt->clust_ref,
 				   PAGE_CUR_LE, BTR_SEARCH_LEAF,
-				   prebuilt->clust_pcur, 0, mtr);
+				   &prebuilt->clust_pcur, 0, mtr);
 
-	clust_rec = btr_pcur_get_rec(prebuilt->clust_pcur);
+	clust_rec = btr_pcur_get_rec(&prebuilt->clust_pcur);
 
-	prebuilt->clust_pcur->trx_if_known = trx;
+	prebuilt->clust_pcur.trx_if_known = trx;
 
 	/* Note: only if the search ends up on a non-infimum record is the
 	low_match value the real match to the search tuple */
 
 	if (!page_rec_is_user_rec(clust_rec)
-	    || btr_pcur_get_low_match(prebuilt->clust_pcur)
+	    || btr_pcur_get_low_match(&prebuilt->clust_pcur)
 	    < dict_index_get_n_unique(clust_index)) {
 
 		/* In a rare case it is possible that no clust rec is found
@@ -2974,7 +2980,7 @@ row_sel_get_clust_rec_for_mysql(
 		we set a LOCK_REC_NOT_GAP type lock */
 
 		err = lock_clust_rec_read_check_and_lock(
-			0, btr_pcur_get_block(prebuilt->clust_pcur),
+			0, btr_pcur_get_block(&prebuilt->clust_pcur),
 			clust_rec, clust_index, *offsets,
 			prebuilt->select_lock_type, LOCK_REC_NOT_GAP, thr);
 		switch (err) {
@@ -3052,7 +3058,7 @@ func_exit:
 		/* We may use the cursor in update or in unlock_row():
 		store its position */
 
-		btr_pcur_store_position(prebuilt->clust_pcur, mtr);
+		btr_pcur_store_position(&prebuilt->clust_pcur, mtr);
 	}
 
 err_exit:
@@ -3300,7 +3306,7 @@ row_sel_try_search_shortcut_for_mysql(
 {
 	dict_index_t*	index		= prebuilt->index;
 	const dtuple_t*	search_tuple	= prebuilt->search_tuple;
-	btr_pcur_t*	pcur		= prebuilt->pcur;
+	btr_pcur_t*	pcur		= &prebuilt->pcur;
 	trx_t*		trx		= prebuilt->trx;
 	const rec_t*	rec;
 
@@ -3389,7 +3395,7 @@ row_search_for_mysql(
 	dict_index_t*	index		= prebuilt->index;
 	ibool		comp		= dict_table_is_comp(index->table);
 	const dtuple_t*	search_tuple	= prebuilt->search_tuple;
-	btr_pcur_t*	pcur		= prebuilt->pcur;
+	btr_pcur_t*	pcur		= &prebuilt->pcur;
 	trx_t*		trx		= prebuilt->trx;
 	dict_index_t*	clust_index;
 	que_thr_t*	thr;
diff --git a/storage/innobase/srv/srv0start.c b/storage/innobase/srv/srv0start.c
index cf11e75b8e0..edc655e93b9 100644
--- a/storage/innobase/srv/srv0start.c
+++ b/storage/innobase/srv/srv0start.c
@@ -733,6 +733,7 @@ open_or_create_data_files(
 	ibool	one_created	= FALSE;
 	ulint	size;
 	ulint	size_high;
+	ulint	flags;
 	ulint	rounded_size_pages;
 	char	name[10000];
 
@@ -914,12 +915,31 @@ open_or_create_data_files(
 				return(DB_ERROR);
 			}
 skip_size_check:
-			fil_read_flushed_lsn_and_arch_log_no(
-				files[i], one_opened,
+			fil_read_first_page(
+				files[i], one_opened, &flags,
 #ifdef UNIV_LOG_ARCHIVE
 				min_arch_log_no, max_arch_log_no,
 #endif /* UNIV_LOG_ARCHIVE */
 				min_flushed_lsn, max_flushed_lsn);
+
+			if (UNIV_PAGE_SIZE
+			    != fsp_flags_get_page_size(flags)) {
+
+				ut_print_timestamp(stderr);
+				fprintf(stderr,
+					" InnoDB: Error: data file %s"
+					" uses page size %lu,\n",
+					name,
+					fsp_flags_get_page_size(flags));
+				ut_print_timestamp(stderr);
+				fprintf(stderr,
+					" InnoDB: but the only supported"
+					" page size in this release is=%lu\n",
+					(ulong) UNIV_PAGE_SIZE);
+
+				return(DB_ERROR);
+			}
+
 			one_opened = TRUE;
 		} else {
 			/* We created the data file and now write it full of
diff --git a/storage/innobase/ut/ut0mem.c b/storage/innobase/ut/ut0mem.c
index 303fdd6dd44..cb6b050beca 100644
--- a/storage/innobase/ut/ut0mem.c
+++ b/storage/innobase/ut/ut0mem.c
@@ -84,17 +84,13 @@ ut_mem_init(void)
 #endif /* !UNIV_HOTBACKUP */
 
 /**********************************************************************//**
-Allocates memory. Sets it also to zero if UNIV_SET_MEM_TO_ZERO is
-defined and set_to_zero is TRUE.
+Allocates memory.
 @return	own: allocated memory */
 UNIV_INTERN
 void*
 ut_malloc_low(
 /*==========*/
 	ulint	n,		/*!< in: number of bytes to allocate */
-	ibool	set_to_zero,	/*!< in: TRUE if allocated memory should be
-				set to zero if UNIV_SET_MEM_TO_ZERO is
-				defined */
 	ibool	assert_on_error)/*!< in: if TRUE, we crash mysqld if the
 				memory cannot be allocated */
 {
@@ -106,12 +102,6 @@ ut_malloc_low(
 		ret = malloc(n);
 		ut_a(ret || !assert_on_error);
 
-#ifdef UNIV_SET_MEM_TO_ZERO
-		if (set_to_zero) {
-			memset(ret, '\0', n);
-			UNIV_MEM_ALLOC(ret, n);
-		}
-#endif
 		return(ret);
 	}
 
@@ -193,12 +183,6 @@ retry:
 		}
 	}
 
-	if (set_to_zero) {
-#ifdef UNIV_SET_MEM_TO_ZERO
-		memset(ret, '\0', n + sizeof(ut_mem_block_t));
-#endif
-	}
-
 	UNIV_MEM_ALLOC(ret, n + sizeof(ut_mem_block_t));
 
 	((ut_mem_block_t*)ret)->size = n + sizeof(ut_mem_block_t);
@@ -215,75 +199,11 @@ retry:
 	void*	ret = malloc(n);
 	ut_a(ret || !assert_on_error);
 
-# ifdef UNIV_SET_MEM_TO_ZERO
-	if (set_to_zero) {
-		memset(ret, '\0', n);
-	}
-# endif
 	return(ret);
 #endif /* !UNIV_HOTBACKUP */
 }
 
 /**********************************************************************//**
-Allocates memory. Sets it also to zero if UNIV_SET_MEM_TO_ZERO is
-defined.
-@return	own: allocated memory */
-UNIV_INTERN
-void*
-ut_malloc(
-/*======*/
-	ulint	n)	/*!< in: number of bytes to allocate */
-{
-#ifndef UNIV_HOTBACKUP
-	return(ut_malloc_low(n, TRUE, TRUE));
-#else /* !UNIV_HOTBACKUP */
-	return(malloc(n));
-#endif /* !UNIV_HOTBACKUP */
-}
-
-#ifndef UNIV_HOTBACKUP
-/**********************************************************************//**
-Tests if malloc of n bytes would succeed. ut_malloc() asserts if memory runs
-out. It cannot be used if we want to return an error message. Prints to
-stderr a message if fails.
-@return	TRUE if succeeded */
-UNIV_INTERN
-ibool
-ut_test_malloc(
-/*===========*/
-	ulint	n)	/*!< in: try to allocate this many bytes */
-{
-	void*	ret;
-
-	ret = malloc(n);
-
-	if (ret == NULL) {
-		ut_print_timestamp(stderr);
-		fprintf(stderr,
-			"  InnoDB: Error: cannot allocate"
-			" %lu bytes of memory for\n"
-			"InnoDB: a BLOB with malloc! Total allocated memory\n"
-			"InnoDB: by InnoDB %lu bytes."
-			" Operating system errno: %d\n"
-			"InnoDB: Check if you should increase"
-			" the swap file or\n"
-			"InnoDB: ulimits of your operating system.\n"
-			"InnoDB: On FreeBSD check you have"
-			" compiled the OS with\n"
-			"InnoDB: a big enough maximum process size.\n",
-			(ulong) n,
-			(ulong) ut_total_allocated_memory,
-			(int) errno);
-		return(FALSE);
-	}
-
-	free(ret);
-
-	return(TRUE);
-}
-#endif /* !UNIV_HOTBACKUP */
-
-/**********************************************************************//**
 Frees a memory block allocated with ut_malloc. Freeing a NULL pointer is
 a nop. */
 UNIV_INTERN
diff --git a/storage/myisam/mi_close.c b/storage/myisam/mi_close.c
index 6e9ee7747ae..e58c2e0f189 100644
--- a/storage/myisam/mi_close.c
+++ b/storage/myisam/mi_close.c
@@ -87,7 +87,12 @@ int mi_close(register MI_INFO *info)
     }
 #ifdef HAVE_MMAP
     if (share->file_map)
-      _mi_unmap_file(info);
+    {
+      if (share->options & HA_OPTION_COMPRESS_RECORD)
+        _mi_unmap_file(info);
+      else
+        mi_munmap_file(info);
+    }
 #endif
     if (share->decode_trees)
     {
diff --git a/storage/myisam/mi_packrec.c b/storage/myisam/mi_packrec.c
index 917bc9f4409..c154ec3b817 100644
--- a/storage/myisam/mi_packrec.c
+++ b/storage/myisam/mi_packrec.c
@@ -1550,13 +1550,14 @@ my_bool _mi_memmap_file(MI_INFO *info)
 
 void _mi_unmap_file(MI_INFO *info)
 {
-  (void) my_munmap((char*) info->s->file_map,
-                   (size_t) info->s->mmaped_length + MEMMAP_EXTRA_MARGIN);
+  DBUG_ASSERT(info->s->options & HA_OPTION_COMPRESS_RECORD);
+
+  (void) my_munmap((char*) info->s->file_map, (size_t) info->s->mmaped_length);
 
   if (myisam_mmap_size != SIZE_T_MAX)
   {
     mysql_mutex_lock(&THR_LOCK_myisam_mmap);
-    myisam_mmap_used-= info->s->mmaped_length + MEMMAP_EXTRA_MARGIN;
+    myisam_mmap_used-= info->s->mmaped_length;
     mysql_mutex_unlock(&THR_LOCK_myisam_mmap);
   }
 }