Merge mysql-5.5-innodb -> mysql-5.5

mysql-test/valgrind.supp

@@ -835,3 +835,45 @@
    fun:lf_hash_search
 }
 
+{
+   Bug 59874 Valgrind warning in InnoDB compression code
+   Memcheck:Cond
+   fun:*
+   fun:*
+   fun:deflate
+   fun:btr_store_big_rec_extern_fields_func
+   fun:row_ins_index_entry_low
+   fun:row_ins_index_entry
+   fun:row_ins_index_entry_step
+   fun:row_ins
+   fun:row_ins_step
+   fun:row_insert_for_mysql
+}
+
+{
+   In page0zip.c we have already checked that the memory is initialized before calling deflate()
+   Memcheck:Cond
+   fun:*
+   fun:*
+   fun:deflate
+   fun:page_zip_compress
+}
+
+{
+   In page0zip.c we have already checked that the memory is initialized before calling deflate()
+   Memcheck:Cond
+   fun:*
+   fun:*
+   fun:deflate
+   fun:page_zip_compress_deflate
+}
+
+{
+   Bug 59875 Valgrind warning in buf0buddy.c
+   Memcheck:Addr1
+   fun:mach_read_from_4
+   fun:buf_buddy_relocate
+   fun:buf_buddy_free_low
+   fun:buf_buddy_free
+   fun:buf_LRU_block_remove_hashed_page
+}

storage/innobase/btr/btr0cur.c

@@ -197,7 +197,7 @@ static
 ulint
 btr_rec_get_externally_stored_len(
 /*==============================*/
-	rec_t*		rec,	/*!< in: record */
+	const rec_t*	rec,	/*!< in: record */
 	const ulint*	offsets);/*!< in: array returned by rec_get_offsets() */
 #endif /* !UNIV_HOTBACKUP */
 
@@ -1886,8 +1886,8 @@ btr_cur_update_in_place(
 		NOT call it if index is secondary */
 
 		if (!dict_index_is_clust(index)
-		    || row_upd_changes_ord_field_binary(NULL, NULL,
-							index, update)) {
+		    || row_upd_changes_ord_field_binary(index, update, thr,
+							NULL, NULL)) {
 
 			/* Remove possible hash index pointer to this record */
 			btr_search_update_hash_on_delete(cursor);
@@ -3761,6 +3761,35 @@ btr_estimate_number_of_different_key_vals(
 
 /*================== EXTERNAL STORAGE OF BIG FIELDS ===================*/
 
+/***********************************************************//**
+Gets the offset of the pointer to the externally stored part of a field.
+@return	offset of the pointer to the externally stored part */
+static
+ulint
+btr_rec_get_field_ref_offs(
+/*=======================*/
+	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
+	ulint		n)	/*!< in: index of the external field */
+{
+	ulint	field_ref_offs;
+	ulint	local_len;
+
+	ut_a(rec_offs_nth_extern(offsets, n));
+	field_ref_offs = rec_get_nth_field_offs(offsets, n, &local_len);
+	ut_a(local_len != UNIV_SQL_NULL);
+	ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);
+
+	return(field_ref_offs + local_len - BTR_EXTERN_FIELD_REF_SIZE);
+}
+
+/** Gets a pointer to the externally stored part of a field.
+@param rec	record
+@param offsets	rec_get_offsets(rec)
+@param n	index of the externally stored field
+@return pointer to the externally stored part */
+#define btr_rec_get_field_ref(rec, offsets, n)			\
+	((rec) + btr_rec_get_field_ref_offs(offsets, n))
+
 /***********************************************************//**
 Gets the externally stored size of a record, in units of a database page.
 @return	externally stored part, in units of a database page */
@@ -3768,27 +3797,26 @@ static
 ulint
 btr_rec_get_externally_stored_len(
 /*==============================*/
-	rec_t*		rec,	/*!< in: record */
+	const rec_t*	rec,	/*!< in: record */
 	const ulint*	offsets)/*!< in: array returned by rec_get_offsets() */
 {
 	ulint	n_fields;
-	byte*	data;
-	ulint	local_len;
-	ulint	extern_len;
 	ulint	total_extern_len = 0;
 	ulint	i;
 
 	ut_ad(!rec_offs_comp(offsets) || !rec_get_node_ptr_flag(rec));
+
+	if (!rec_offs_any_extern(offsets)) {
+		return(0);
+	}
+
 	n_fields = rec_offs_n_fields(offsets);
 
 	for (i = 0; i < n_fields; i++) {
 		if (rec_offs_nth_extern(offsets, i)) {
 
-			data = rec_get_nth_field(rec, offsets, i, &local_len);
-
-			local_len -= BTR_EXTERN_FIELD_REF_SIZE;
-
-			extern_len = mach_read_from_4(data + local_len
+			ulint	extern_len = mach_read_from_4(
+				btr_rec_get_field_ref(rec, offsets, i)
 				+ BTR_EXTERN_LEN + 4);
 
 			total_extern_len += ut_calc_align(extern_len,
@@ -3819,7 +3847,7 @@ btr_cur_set_ownership_of_extern_field(
 	ulint	byte_val;
 
 	data = rec_get_nth_field(rec, offsets, i, &local_len);
-
+	ut_ad(rec_offs_nth_extern(offsets, i));
 	ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);
 
 	local_len -= BTR_EXTERN_FIELD_REF_SIZE;
@@ -3829,6 +3857,9 @@ btr_cur_set_ownership_of_extern_field(
 	if (val) {
 		byte_val = byte_val & (~BTR_EXTERN_OWNER_FLAG);
 	} else {
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+		ut_a(!(byte_val & BTR_EXTERN_OWNER_FLAG));
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
 		byte_val = byte_val | BTR_EXTERN_OWNER_FLAG;
 	}
 
@@ -4067,8 +4098,8 @@ file segment of the index tree.
 @return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
 UNIV_INTERN
 ulint
-btr_store_big_rec_extern_fields(
-/*============================*/
+btr_store_big_rec_extern_fields_func(
+/*=================================*/
 	dict_index_t*	index,		/*!< in: index of rec; the index tree
 					MUST be X-latched */
 	buf_block_t*	rec_block,	/*!< in/out: block containing rec */
@@ -4077,11 +4108,17 @@ btr_store_big_rec_extern_fields(
 					the "external storage" flags in offsets
 					will not correspond to rec when
 					this function returns */
-	big_rec_t*	big_rec_vec,	/*!< in: vector containing fields
+#ifdef UNIV_DEBUG
+	mtr_t*		local_mtr,	/*!< in: mtr containing the
+					latch to rec and to the tree */
+#endif /* UNIV_DEBUG */
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+	ibool		update_in_place,/*! in: TRUE if the record is updated
+					in place (not delete+insert) */
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
+	const big_rec_t*big_rec_vec)	/*!< in: vector containing fields
 					to be stored externally */
-	mtr_t*		local_mtr __attribute__((unused))) /*!< in: mtr
-					containing the latch to rec and to the
-					tree */
+
 {
 	ulint	rec_page_no;
 	byte*	field_ref;
@@ -4099,6 +4136,7 @@ btr_store_big_rec_extern_fields(
 	z_stream c_stream;
 
 	ut_ad(rec_offs_validate(rec, index, offsets));
+	ut_ad(rec_offs_any_extern(offsets));
 	ut_ad(mtr_memo_contains(local_mtr, dict_index_get_lock(index),
 				MTR_MEMO_X_LOCK));
 	ut_ad(mtr_memo_contains(local_mtr, rec_block, MTR_MEMO_PAGE_X_FIX));
@@ -4130,21 +4168,37 @@ btr_store_big_rec_extern_fields(
 		ut_a(err == Z_OK);
 	}
 
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+	/* All pointers to externally stored columns in the record
+	must either be zero or they must be pointers to inherited
+	columns, owned by this record or an earlier record version. */
+	for (i = 0; i < rec_offs_n_fields(offsets); i++) {
+		if (!rec_offs_nth_extern(offsets, i)) {
+			continue;
+		}
+		field_ref = btr_rec_get_field_ref(rec, offsets, i);
+
+		ut_a(!(field_ref[BTR_EXTERN_LEN] & BTR_EXTERN_OWNER_FLAG));
+		/* Either this must be an update in place,
+		or the BLOB must be inherited, or the BLOB pointer
+		must be zero (will be written in this function). */
+		ut_a(update_in_place
+		     || (field_ref[BTR_EXTERN_LEN] & BTR_EXTERN_INHERITED_FLAG)
+		     || !memcmp(field_ref, field_ref_zero,
+				BTR_EXTERN_FIELD_REF_SIZE));
+	}
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
 	/* We have to create a file segment to the tablespace
 	for each field and put the pointer to the field in rec */
 
 	for (i = 0; i < big_rec_vec->n_fields; i++) {
-		ut_ad(rec_offs_nth_extern(offsets,
-					  big_rec_vec->fields[i].field_no));
-		{
-			ulint	local_len;
-			field_ref = rec_get_nth_field(
-				rec, offsets, big_rec_vec->fields[i].field_no,
-				&local_len);
-			ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);
-			local_len -= BTR_EXTERN_FIELD_REF_SIZE;
-			field_ref += local_len;
-		}
+		field_ref = btr_rec_get_field_ref(
+			rec, offsets, big_rec_vec->fields[i].field_no);
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+		/* A zero BLOB pointer should have been initially inserted. */
+		ut_a(!memcmp(field_ref, field_ref_zero,
+			     BTR_EXTERN_FIELD_REF_SIZE));
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
 		extern_len = big_rec_vec->fields[i].len;
 		UNIV_MEM_ASSERT_RW(big_rec_vec->fields[i].data,
 				   extern_len);
@@ -4426,6 +4480,23 @@ btr_store_big_rec_extern_fields(
 		mem_heap_free(heap);
 	}
 
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+	/* All pointers to externally stored columns in the record
+	must be valid. */
+	for (i = 0; i < rec_offs_n_fields(offsets); i++) {
+		if (!rec_offs_nth_extern(offsets, i)) {
+			continue;
+		}
+
+		field_ref = btr_rec_get_field_ref(rec, offsets, i);
+
+		/* The pointer must not be zero. */
+		ut_a(0 != memcmp(field_ref, field_ref_zero,
+				 BTR_EXTERN_FIELD_REF_SIZE));
+		/* The column must not be disowned by this record. */
+		ut_a(!(field_ref[BTR_EXTERN_LEN] & BTR_EXTERN_OWNER_FLAG));
+	}
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
 	return(DB_SUCCESS);
 }
 
@@ -4448,6 +4519,7 @@ btr_check_blob_fil_page_type(
 	if (UNIV_UNLIKELY(type != FIL_PAGE_TYPE_BLOB)) {
 		ulint	flags = fil_space_get_flags(space_id);
 
+#ifndef UNIV_DEBUG /* Improve debug test coverage */
 		if (UNIV_LIKELY
 		    ((flags & DICT_TF_FORMAT_MASK) == DICT_TF_FORMAT_51)) {
 			/* Old versions of InnoDB did not initialize
@@ -4456,6 +4528,7 @@ btr_check_blob_fil_page_type(
 			a BLOB page that is in Antelope format.*/
 			return;
 		}
+#endif /* !UNIV_DEBUG */
 
 		ut_print_timestamp(stderr);
 		fprintf(stderr,
@@ -4505,23 +4578,13 @@ btr_free_externally_stored_field(
 	ulint		page_no;
 	ulint		next_page_no;
 	mtr_t		mtr;
-#ifdef UNIV_DEBUG
+
 	ut_ad(mtr_memo_contains(local_mtr, dict_index_get_lock(index),
 				MTR_MEMO_X_LOCK));
 	ut_ad(mtr_memo_contains_page(local_mtr, field_ref,
 				     MTR_MEMO_PAGE_X_FIX));
 	ut_ad(!rec || rec_offs_validate(rec, index, offsets));
-
-	if (rec) {
-		ulint	local_len;
-		const byte*	f = rec_get_nth_field(rec, offsets,
-						      i, &local_len);
-		ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);
-		local_len -= BTR_EXTERN_FIELD_REF_SIZE;
-		f += local_len;
-		ut_ad(f == field_ref);
-	}
-#endif /* UNIV_DEBUG */
+	ut_ad(!rec || field_ref == btr_rec_get_field_ref(rec, offsets, i));
 
 	if (UNIV_UNLIKELY(!memcmp(field_ref, field_ref_zero,
 				  BTR_EXTERN_FIELD_REF_SIZE))) {
@@ -4684,13 +4747,8 @@ btr_rec_free_externally_stored_fields(
 
 	for (i = 0; i < n_fields; i++) {
 		if (rec_offs_nth_extern(offsets, i)) {
-			ulint	len;
-			byte*	data
-				= rec_get_nth_field(rec, offsets, i, &len);
-			ut_a(len >= BTR_EXTERN_FIELD_REF_SIZE);
-
 			btr_free_externally_stored_field(
-				index, data + len - BTR_EXTERN_FIELD_REF_SIZE,
+				index, btr_rec_get_field_ref(rec, offsets, i),
 				rec, offsets, page_zip, i, rb_ctx, mtr);
 		}
 	}

storage/innobase/buf/buf0buf.c

@@ -865,9 +865,9 @@ buf_block_init(
 
 	block->modify_clock = 0;
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	block->page.file_page_was_freed = FALSE;
-#endif /* UNIV_DEBUG_FILE_ACCESSES */
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 
 	block->check_index_page_at_flush = FALSE;
 	block->index = NULL;
@@ -2229,7 +2229,7 @@ buf_page_peek_if_search_hashed(
 	return(is_hashed);
 }
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 /********************************************************************//**
 Sets file_page_was_freed TRUE if the page is found in the buffer pool.
 This function should be called when we free a file page and want the
@@ -2252,6 +2252,8 @@ buf_page_set_file_page_was_freed(
 
 	if (bpage) {
 		ut_ad(!buf_pool_watch_is_sentinel(buf_pool, bpage));
+		/* bpage->file_page_was_freed can already hold
+		when this code is invoked from dict_drop_index_tree() */
 		bpage->file_page_was_freed = TRUE;
 	}
 
@@ -2289,7 +2291,7 @@ buf_page_reset_file_page_was_freed(
 
 	return(bpage);
 }
-#endif /* UNIV_DEBUG_FILE_ACCESSES */
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 
 /********************************************************************//**
 Get read access to a compressed page (usually of type
@@ -2390,7 +2392,7 @@ buf_page_get_zip(
 
 	buf_page_set_accessed_make_young(bpage, access_time);
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	ut_a(!bpage->file_page_was_freed);
 #endif
 
@@ -3058,7 +3060,7 @@ buf_page_get_gen(
 
 	buf_page_set_accessed_make_young(&block->page, access_time);
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	ut_a(!block->page.file_page_was_freed);
 #endif
 
@@ -3217,7 +3219,7 @@ buf_page_optimistic_get(
 	ut_a(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
 #endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	ut_a(block->page.file_page_was_freed == FALSE);
 #endif
 	if (UNIV_UNLIKELY(!access_time)) {
@@ -3329,7 +3331,7 @@ buf_page_get_known_nowait(
 	ut_a(block->page.buf_fix_count > 0);
 	ut_a(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
 #endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	ut_a(block->page.file_page_was_freed == FALSE);
 #endif
 
@@ -3415,9 +3417,9 @@ buf_page_try_get_func(
 	ut_a(block->page.buf_fix_count > 0);
 	ut_a(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
 #endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	ut_a(block->page.file_page_was_freed == FALSE);
-#endif /* UNIV_DEBUG_FILE_ACCESSES */
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 	buf_block_dbg_add_level(block, SYNC_NO_ORDER_CHECK);
 
 	buf_pool->stat.n_page_gets++;
@@ -3446,9 +3448,9 @@ buf_page_init_low(
 	bpage->newest_modification = 0;
 	bpage->oldest_modification = 0;
 	HASH_INVALIDATE(bpage, hash);
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	bpage->file_page_was_freed = FALSE;
-#endif /* UNIV_DEBUG_FILE_ACCESSES */
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 }
 
 /********************************************************************//**
@@ -3807,9 +3809,9 @@ buf_page_create(
 #ifdef UNIV_IBUF_COUNT_DEBUG
 		ut_a(ibuf_count_get(space, offset) == 0);
 #endif
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 		block->page.file_page_was_freed = FALSE;
-#endif /* UNIV_DEBUG_FILE_ACCESSES */
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 
 		/* Page can be found in buf_pool */
 		buf_pool_mutex_exit(buf_pool);

storage/innobase/dict/dict0load.c

@@ -1552,7 +1552,7 @@ dict_load_table_low(
 				"InnoDB: in InnoDB data dictionary"
 				" has unknown type %lx.\n",
 				(ulong) flags);
-			return(NULL);
+			return("incorrect flags in SYS_TABLES");
 		}
 	} else {
 		flags = 0;

storage/innobase/fsp/fsp0fsp.c

@@ -3435,9 +3435,9 @@ fseg_free_page(
 
 	fseg_free_page_low(seg_inode, space, zip_size, page, mtr);
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	buf_page_set_file_page_was_freed(space, page);
-#endif
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 }
 
 /**********************************************************************//**
@@ -3503,13 +3503,13 @@ fseg_free_extent(
 
 	fsp_free_extent(space, zip_size, page, mtr);
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	for (i = 0; i < FSP_EXTENT_SIZE; i++) {
 
 		buf_page_set_file_page_was_freed(space,
 						 first_page_in_extent + i);
 	}
-#endif
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 }
 
 /**********************************************************************//**

storage/innobase/ibuf/ibuf0ibuf.c

@@ -2276,9 +2276,9 @@ ibuf_remove_free_page(void)
 	fseg_free_page(header_page + IBUF_HEADER + IBUF_TREE_SEG_HEADER,
 		       IBUF_SPACE_ID, page_no, &mtr);
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	buf_page_reset_file_page_was_freed(IBUF_SPACE_ID, page_no);
-#endif
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 
 	ibuf_enter();
 
@@ -2322,9 +2322,9 @@ ibuf_remove_free_page(void)
 	ibuf_bitmap_page_set_bits(
 		bitmap_page, page_no, zip_size, IBUF_BITMAP_IBUF, FALSE, &mtr);
 
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	buf_page_set_file_page_was_freed(IBUF_SPACE_ID, page_no);
-#endif
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 	mtr_commit(&mtr);
 
 	ibuf_exit();

storage/innobase/include/btr0cur.h

@@ -500,8 +500,8 @@ file segment of the index tree.
 @return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
 UNIV_INTERN
 ulint
-btr_store_big_rec_extern_fields(
-/*============================*/
+btr_store_big_rec_extern_fields_func(
+/*=================================*/
 	dict_index_t*	index,		/*!< in: index of rec; the index tree
 					MUST be X-latched */
 	buf_block_t*	rec_block,	/*!< in/out: block containing rec */
@@ -510,10 +510,42 @@ btr_store_big_rec_extern_fields(
 					the "external storage" flags in offsets
 					will not correspond to rec when
 					this function returns */
-	big_rec_t*	big_rec_vec,	/*!< in: vector containing fields
+#ifdef UNIV_DEBUG
+	mtr_t*		local_mtr,	/*!< in: mtr containing the
+					latch to rec and to the tree */
+#endif /* UNIV_DEBUG */
+#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
+	ibool		update_in_place,/*! in: TRUE if the record is updated
+					in place (not delete+insert) */
+#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
+	const big_rec_t*big_rec_vec)	/*!< in: vector containing fields
 					to be stored externally */
-	mtr_t*		local_mtr);	/*!< in: mtr containing the latch to
-					rec and to the tree */
+	__attribute__((nonnull));
+
+/** Stores the fields in big_rec_vec to the tablespace and puts pointers to
+them in rec.  The extern flags in rec will have to be set beforehand.
+The fields are stored on pages allocated from leaf node
+file segment of the index tree.
+@param index	in: clustered index; MUST be X-latched by mtr
+@param b	in/out: block containing rec; MUST be X-latched by mtr
+@param rec	in/out: clustered index record
+@param offsets	in: rec_get_offsets(rec, index);
+		the "external storage" flags in offsets will not be adjusted
+@param mtr	in: mini-transaction that holds x-latch on index and b
+@param upd	in: TRUE if the record is updated in place (not delete+insert)
+@param big	in: vector containing fields to be stored externally
+@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
+#ifdef UNIV_DEBUG
+# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
+	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,mtr,upd,big)
+#elif defined UNIV_BLOB_LIGHT_DEBUG
+# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
+	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,upd,big)
+#else
+# define btr_store_big_rec_extern_fields(index,b,rec,offsets,mtr,upd,big) \
+	btr_store_big_rec_extern_fields_func(index,b,rec,offsets,big)
+#endif
+
 /*******************************************************************//**
 Frees the space in an externally stored field to the file space
 management if the field in data is owned the externally stored field,

storage/innobase/include/buf0buf.h

@@ -480,7 +480,7 @@ buf_reset_check_index_page_at_flush(
 /*================================*/
 	ulint	space,	/*!< in: space id */
 	ulint	offset);/*!< in: page number */
-#ifdef UNIV_DEBUG_FILE_ACCESSES
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 /********************************************************************//**
 Sets file_page_was_freed TRUE if the page is found in the buffer pool.
 This function should be called when we free a file page and want the
@@ -505,7 +505,7 @@ buf_page_reset_file_page_was_freed(
 /*===============================*/
 	ulint	space,	/*!< in: space id */
 	ulint	offset);	/*!< in: page number */
-#endif /* UNIV_DEBUG_FILE_ACCESSES */
+#endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 /********************************************************************//**
 Reads the freed_page_clock of a buffer block.
 @return	freed_page_clock */
@@ -1403,11 +1403,11 @@ struct buf_page_struct{
 					0 if the block was never accessed
 					in the buffer pool */
 	/* @} */
-# ifdef UNIV_DEBUG_FILE_ACCESSES
+# if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
 	ibool		file_page_was_freed;
 					/*!< this is set to TRUE when fsp
 					frees a page in buffer pool */
-# endif /* UNIV_DEBUG_FILE_ACCESSES */
+# endif /* UNIV_DEBUG_FILE_ACCESSES || UNIV_DEBUG */
 #endif /* !UNIV_HOTBACKUP */
 };
 

storage/innobase/include/row0upd.h

@@ -280,19 +280,29 @@ NOTE: we compare the fields as binary strings!
 @return TRUE if update vector changes an ordering field in the index record */
 UNIV_INTERN
 ibool
-row_upd_changes_ord_field_binary(
-/*=============================*/
+row_upd_changes_ord_field_binary_func(
+/*==================================*/
+	dict_index_t*	index,	/*!< in: index of the record */
+	const upd_t*	update,	/*!< in: update vector for the row; NOTE: the
+				field numbers in this MUST be clustered index
+				positions! */
+#ifdef UNIV_DEBUG
+	const que_thr_t*thr,	/*!< in: query thread */
+#endif /* UNIV_DEBUG */
 	const dtuple_t*	row,	/*!< in: old value of row, or NULL if the
 				row and the data values in update are not
 				known when this function is called, e.g., at
 				compile time */
-	const row_ext_t*ext,	/*!< NULL, or prefixes of the externally
+	const row_ext_t*ext)	/*!< NULL, or prefixes of the externally
 				stored columns in the old row */
-	dict_index_t*	index,	/*!< in: index of the record */
-	const upd_t*	update)	/*!< in: update vector for the row; NOTE: the
-				field numbers in this MUST be clustered index
-				positions! */
-	__attribute__((nonnull(3,4), warn_unused_result));
+	__attribute__((nonnull(1,2), warn_unused_result));
+#ifdef UNIV_DEBUG
+# define row_upd_changes_ord_field_binary(index,update,thr,row,ext)	\
+	row_upd_changes_ord_field_binary_func(index,update,thr,row,ext)
+#else /* UNIV_DEBUG */
+# define row_upd_changes_ord_field_binary(index,update,thr,row,ext)	\
+	row_upd_changes_ord_field_binary_func(index,update,row,ext)
+#endif /* UNIV_DEBUG */
 /***********************************************************//**
 Checks if an update vector changes an ordering field of an index record.
 This function is fast if the update vector is short or the number of ordering

storage/innobase/include/univ.i

@@ -184,14 +184,15 @@ command. Not tested on Windows. */
 						debugging without UNIV_DEBUG */
 #define UNIV_BUF_DEBUG				/* Enable buffer pool
 						debugging without UNIV_DEBUG */
+#define UNIV_BLOB_LIGHT_DEBUG			/* Enable off-page column
+						debugging without UNIV_DEBUG */
 #define UNIV_DEBUG				/* Enable ut_ad() assertions
 						and disable UNIV_INLINE */
 #define UNIV_DEBUG_LOCK_VALIDATE		/* Enable
 						ut_ad(lock_rec_validate_page())
 						assertions. */
-#define UNIV_DEBUG_FILE_ACCESSES		/* Debug .ibd file access
-						(field file_page_was_freed
-						in buf_page_t) */
+#define UNIV_DEBUG_FILE_ACCESSES		/* Enable freed block access
+						debugging without UNIV_DEBUG */
 #define UNIV_LRU_DEBUG				/* debug the buffer pool LRU */
 #define UNIV_HASH_DEBUG				/* debug HASH_ macros */
 #define UNIV_LIST_DEBUG				/* debug UT_LIST_ macros */

storage/innobase/row/row0ins.c

@@ -2122,7 +2122,7 @@ row_ins_index_entry_low(
 
 		err = btr_store_big_rec_extern_fields(
 			index, btr_cur_get_block(&cursor),
-			rec, offsets, big_rec, &mtr);
+			rec, offsets, &mtr, FALSE, big_rec);
 
 		if (modify) {
 			dtuple_big_rec_free(big_rec);

storage/innobase/row/row0mysql.c

@@ -1939,15 +1939,13 @@ row_create_table_for_mysql(
 
 	err = trx->error_state;
 
-	if (UNIV_UNLIKELY(err != DB_SUCCESS)) {
+	switch (err) {
+	case DB_SUCCESS:
+		break;
+	case DB_OUT_OF_FILE_SPACE:
 		trx->error_state = DB_SUCCESS;
 		trx_general_rollback_for_mysql(trx, NULL);
-		/* TO DO: free table?  The code below will dereference
-		table->name, though. */
-	}
 
-	switch (err) {
-	case DB_OUT_OF_FILE_SPACE:
 		ut_print_timestamp(stderr);
 		fputs("  InnoDB: Warning: cannot create table ",
 		      stderr);
@@ -1962,9 +1960,13 @@ row_create_table_for_mysql(
 		break;
 
 	case DB_DUPLICATE_KEY:
+	default:
 		/* We may also get err == DB_ERROR if the .ibd file for the
 		table already exists */
 
+		trx->error_state = DB_SUCCESS;
+		trx_general_rollback_for_mysql(trx, NULL);
+		dict_mem_table_free(table);
 		break;
 	}
 

storage/innobase/row/row0purge.c

@@ -489,8 +489,11 @@ Purges an update of an existing record. Also purges an update of a delete
 marked record if that record contained an externally stored field. */
 static
 void
-row_purge_upd_exist_or_extern(
-/*==========================*/
+row_purge_upd_exist_or_extern_func(
+/*===============================*/
+#ifdef UNIV_DEBUG
+	const que_thr_t*thr,	/*!< in: query thread */
+#endif /* UNIV_DEBUG */
 	purge_node_t*	node)	/*!< in: row purge node */
 {
 	mem_heap_t*	heap;
@@ -515,8 +518,8 @@ row_purge_upd_exist_or_extern(
 	while (node->index != NULL) {
 		index = node->index;
 
-		if (row_upd_changes_ord_field_binary(NULL, NULL, node->index,
-						     node->update)) {
+		if (row_upd_changes_ord_field_binary(node->index, node->update,
+						     thr, NULL, NULL)) {
 			/* Build the older version of the index entry */
 			entry = row_build_index_entry(node->row, NULL,
 						      index, heap);
@@ -598,6 +601,14 @@ row_purge_upd_exist_or_extern(
 	}
 }
 
+#ifdef UNIV_DEBUG
+# define row_purge_upd_exist_or_extern(thr,node)	\
+	row_purge_upd_exist_or_extern_func(thr,node)
+#else /* UNIV_DEBUG */
+# define row_purge_upd_exist_or_extern(thr,node)	\
+	row_purge_upd_exist_or_extern_func(node)
+#endif /* UNIV_DEBUG */
+
 /***********************************************************//**
 Parses the row reference and other info in a modify undo log record.
 @return TRUE if purge operation required: NOTE that then the CALLER
@@ -704,47 +715,32 @@ row_purge_parse_undo_rec(
 /***********************************************************//**
 Fetches an undo log record and does the purge for the recorded operation.
 If none left, or the current purge completed, returns the control to the
-parent node, which is always a query thread node.
-@return	DB_SUCCESS if operation successfully completed, else error code */
-static
-ulint
+parent node, which is always a query thread node. */
+static __attribute__((nonnull))
+void
 row_purge(
 /*======*/
 	purge_node_t*	node,	/*!< in: row purge node */
 	que_thr_t*	thr)	/*!< in: query thread */
 {
-	roll_ptr_t	roll_ptr;
-	ibool		purge_needed;
 	ibool		updated_extern;
-	trx_t*		trx;
 
-	ut_ad(node && thr);
-
-	trx = thr_get_trx(thr);
+	ut_ad(node);
+	ut_ad(thr);
 
-	node->undo_rec = trx_purge_fetch_next_rec(&roll_ptr,
-						  &(node->reservation),
+	node->undo_rec = trx_purge_fetch_next_rec(&node->roll_ptr,
+						  &node->reservation,
 						  node->heap);
 	if (!node->undo_rec) {
 		/* Purge completed for this query thread */
 
 		thr->run_node = que_node_get_parent(node);
 
-		return(DB_SUCCESS);
-	}
-
-	node->roll_ptr = roll_ptr;
-
-	if (node->undo_rec == &trx_purge_dummy_rec) {
-		purge_needed = FALSE;
-	} else {
-		purge_needed = row_purge_parse_undo_rec(node, &updated_extern,
-							thr);
-		/* If purge_needed == TRUE, we must also remember to unfreeze
-		data dictionary! */
+		return;
 	}
 
-	if (purge_needed) {
+	if (node->undo_rec != &trx_purge_dummy_rec
+	    && row_purge_parse_undo_rec(node, &updated_extern, thr)) {
 		node->found_clust = FALSE;
 
 		node->index = dict_table_get_next_index(
@@ -756,14 +752,14 @@ row_purge(
 		} else if (updated_extern
 			   || node->rec_type == TRX_UNDO_UPD_EXIST_REC) {
 
-			row_purge_upd_exist_or_extern(node);
+			row_purge_upd_exist_or_extern(thr, node);
 		}
 
 		if (node->found_clust) {
 			btr_pcur_close(&(node->pcur));
 		}
 
-		row_mysql_unfreeze_data_dictionary(trx);
+		row_mysql_unfreeze_data_dictionary(thr_get_trx(thr));
 	}
 
 	/* Do some cleanup */
@@ -771,8 +767,6 @@ row_purge(
 	mem_heap_empty(node->heap);
 
 	thr->run_node = node;
-
-	return(DB_SUCCESS);
 }
 
 /***********************************************************//**
@@ -786,9 +780,6 @@ row_purge_step(
 	que_thr_t*	thr)	/*!< in: query thread */
 {
 	purge_node_t*	node;
-#ifdef UNIV_DEBUG
-	ulint		err;
-#endif /* UNIV_DEBUG */
 
 	ut_ad(thr);
 
@@ -796,12 +787,7 @@ row_purge_step(
 
 	ut_ad(que_node_get_type(node) == QUE_NODE_PURGE);
 
-#ifdef UNIV_DEBUG
-	err =
-#endif /* UNIV_DEBUG */
 	row_purge(node, thr);
 
-	ut_ad(err == DB_SUCCESS);
-
 	return(thr);
 }

storage/innobase/row/row0umod.c

@@ -173,40 +173,26 @@ row_undo_mod_remove_clust_low(
 	mtr_t*		mtr,	/*!< in: mtr */
 	ulint		mode)	/*!< in: BTR_MODIFY_LEAF or BTR_MODIFY_TREE */
 {
-	btr_pcur_t*	pcur;
 	btr_cur_t*	btr_cur;
 	ulint		err;
-	ibool		success;
 
 	ut_ad(node->rec_type == TRX_UNDO_UPD_DEL_REC);
-	pcur = &(node->pcur);
-	btr_cur = btr_pcur_get_btr_cur(pcur);
 
-	success = btr_pcur_restore_position(mode, pcur, mtr);
+	/* Find out if the record has been purged already
+	or if we can remove it. */
 
-	if (!success) {
+	if (!btr_pcur_restore_position(mode, &node->pcur, mtr)
+	    || row_vers_must_preserve_del_marked(node->new_trx_id, mtr)) {
 
 		return(DB_SUCCESS);
 	}
 
-	/* Find out if we can remove the whole clustered index record */
-
-	if (node->rec_type == TRX_UNDO_UPD_DEL_REC
-	    && !row_vers_must_preserve_del_marked(node->new_trx_id, mtr)) {
-
-		/* Ok, we can remove */
-	} else {
-		return(DB_SUCCESS);
-	}
+	btr_cur = btr_pcur_get_btr_cur(&node->pcur);
 
 	if (mode == BTR_MODIFY_LEAF) {
-		success = btr_cur_optimistic_delete(btr_cur, mtr);
-
-		if (success) {
-			err = DB_SUCCESS;
-		} else {
-			err = DB_FAIL;
-		}
+		err = btr_cur_optimistic_delete(btr_cur, mtr)
+			? DB_SUCCESS
+			: DB_FAIL;
 	} else {
 		ut_ad(mode == BTR_MODIFY_TREE);
 
@@ -693,8 +679,9 @@ row_undo_mod_upd_exist_sec(
 	while (node->index != NULL) {
 		index = node->index;
 
-		if (row_upd_changes_ord_field_binary(
-			    node->row, node->ext, node->index, node->update)) {
+		if (row_upd_changes_ord_field_binary(node->index, node->update,
+						     thr,
+						     node->row, node->ext)) {
 
 			/* Build the newest version of the index entry */
 			entry = row_build_index_entry(node->row, node->ext,

storage/innobase/row/row0upd.c

@@ -1192,25 +1192,31 @@ NOTE: we compare the fields as binary strings!
 @return TRUE if update vector changes an ordering field in the index record */
 UNIV_INTERN
 ibool
-row_upd_changes_ord_field_binary(
-/*=============================*/
+row_upd_changes_ord_field_binary_func(
+/*==================================*/
+	dict_index_t*	index,	/*!< in: index of the record */
+	const upd_t*	update,	/*!< in: update vector for the row; NOTE: the
+				field numbers in this MUST be clustered index
+				positions! */
+#ifdef UNIV_DEBUG
+	const que_thr_t*thr,	/*!< in: query thread */
+#endif /* UNIV_DEBUG */
 	const dtuple_t*	row,	/*!< in: old value of row, or NULL if the
 				row and the data values in update are not
 				known when this function is called, e.g., at
 				compile time */
-	const row_ext_t*ext,	/*!< NULL, or prefixes of the externally
+	const row_ext_t*ext)	/*!< NULL, or prefixes of the externally
 				stored columns in the old row */
-	dict_index_t*	index,	/*!< in: index of the record */
-	const upd_t*	update)	/*!< in: update vector for the row; NOTE: the
-				field numbers in this MUST be clustered index
-				positions! */
 {
 	ulint			n_unique;
 	ulint			i;
 	const dict_index_t*	clust_index;
 
-	ut_ad(update);
 	ut_ad(index);
+	ut_ad(update);
+	ut_ad(thr);
+	ut_ad(thr->graph);
+	ut_ad(thr->graph->trx);
 
 	n_unique = dict_index_get_n_unique(index);
 
@@ -1263,9 +1269,14 @@ row_upd_changes_ord_field_binary(
 
 			if (UNIV_LIKELY_NULL(buf)) {
 				if (UNIV_UNLIKELY(buf == field_ref_zero)) {
-					/* This should never happen, but
-					we try to fail safe here. */
-					ut_ad(0);
+					/* The externally stored field
+					was not written yet. This
+					record should only be seen by
+					recv_recovery_rollback_active(),
+					when the server had crashed before
+					storing the field. */
+					ut_ad(thr->graph->trx->is_recovered);
+					ut_ad(trx_is_recv(thr->graph->trx));
 					return(TRUE);
 				}
 
@@ -1640,8 +1651,8 @@ row_upd_sec_step(
 	ut_ad(!dict_index_is_clust(node->index));
 
 	if (node->state == UPD_NODE_UPDATE_ALL_SEC
-	    || row_upd_changes_ord_field_binary(node->row, node->ext,
-						node->index, node->update)) {
+	    || row_upd_changes_ord_field_binary(node->index, node->update,
+						thr, node->row, node->ext)) {
 		return(row_upd_sec_index_entry(node, thr));
 	}
 
@@ -1971,7 +1982,7 @@ row_upd_clust_rec(
 			index, btr_cur_get_block(btr_cur), rec,
 			rec_get_offsets(rec, index, offsets_,
 					ULINT_UNDEFINED, &heap),
-			big_rec, mtr);
+			mtr, TRUE, big_rec);
 		mtr_commit(mtr);
 	}
 
@@ -2169,8 +2180,8 @@ row_upd_clust_step(
 
 	row_upd_store_row(node);
 
-	if (row_upd_changes_ord_field_binary(node->row, node->ext, index,
-					     node->update)) {
+	if (row_upd_changes_ord_field_binary(index, node->update, thr,
+					     node->row, node->ext)) {
 
 		/* Update causes an ordering field (ordering fields within
 		the B-tree) of the clustered index record to change: perform

storage/innobase/srv/srv0start.c

@@ -1327,13 +1327,16 @@ innobase_start_or_create_for_mysql(void)
 
 	ut_a(srv_n_file_io_threads <= SRV_MAX_N_IO_THREADS);
 
-	/* TODO: Investigate if SRV_N_PENDING_IOS_PER_THREAD (32) limit
-	still applies to windows. */
-	if (!srv_use_native_aio) {
 	io_limit = 8 * SRV_N_PENDING_IOS_PER_THREAD;
-	} else {
+
+	/* On Windows when using native aio the number of aio requests
+	that a thread can handle at a given time is limited to 32
+	i.e.: SRV_N_PENDING_IOS_PER_THREAD */
+# ifdef __WIN__
+	if (srv_use_native_aio) {
 		io_limit = SRV_N_PENDING_IOS_PER_THREAD;
 	}
+# endif /* __WIN__ */
 
 	os_aio_init(io_limit,
 		    srv_n_read_io_threads,

storage/innobase/trx/trx0roll.c

@@ -47,8 +47,8 @@ Created 3/26/1996 Heikki Tuuri
 rollback */
 #define TRX_ROLL_TRUNC_THRESHOLD	1
 
-/** In crash recovery, the current trx to be rolled back */
-static trx_t*		trx_roll_crash_recv_trx	= NULL;
+/** In crash recovery, the current trx to be rolled back; NULL otherwise */
+static const trx_t*	trx_roll_crash_recv_trx	= NULL;
 
 /** In crash recovery we set this to the undo n:o of the current trx to be
 rolled back. Then we can print how many % the rollback has progressed. */