Merge XtraDB from Percona-Server 5.1.66-rel14.1 into MariaDB 5.1.

mysql-test/r/information_schema.result

@@ -54,9 +54,13 @@ EVENTS
 FILES
 GLOBAL_STATUS
 GLOBAL_VARIABLES
+INNODB_BUFFER_PAGE
+INNODB_BUFFER_PAGE_LRU
 INNODB_BUFFER_POOL_PAGES
 INNODB_BUFFER_POOL_PAGES_BLOB
 INNODB_BUFFER_POOL_PAGES_INDEX
+INNODB_BUFFER_POOL_STATS
+INNODB_CHANGED_PAGES
 INNODB_CMP
 INNODB_CMPMEM
 INNODB_CMPMEM_RESET
@@ -89,7 +93,6 @@ TRIGGERS
 USER_PRIVILEGES
 VIEWS
 XTRADB_ADMIN_COMMAND
-XTRADB_ENHANCEMENTS
 columns_priv
 db
 event
@@ -860,6 +863,8 @@ TABLE_NAME	COLUMN_NAME	PRIVILEGES
 COLUMNS	TABLE_NAME	select
 COLUMN_PRIVILEGES	TABLE_NAME	select
 FILES	TABLE_NAME	select
+INNODB_BUFFER_PAGE	TABLE_NAME	select
+INNODB_BUFFER_PAGE_LRU	TABLE_NAME	select
 INNODB_INDEX_STATS	table_name	select
 INNODB_TABLE_STATS	table_name	select
 KEY_COLUMN_USAGE	TABLE_NAME	select
@@ -1245,12 +1250,12 @@ DROP PROCEDURE p1;
 DROP USER mysql_bug20230@localhost;
 SELECT MAX(table_name) FROM information_schema.tables WHERE table_schema IN ('mysql', 'INFORMATION_SCHEMA', 'test');
 MAX(table_name)
-XTRADB_ENHANCEMENTS
+XTRADB_ADMIN_COMMAND
 SELECT table_name from information_schema.tables
 WHERE table_name=(SELECT MAX(table_name)
 FROM information_schema.tables WHERE table_schema IN ('mysql', 'INFORMATION_SCHEMA', 'test'));
 table_name
-XTRADB_ENHANCEMENTS
+XTRADB_ADMIN_COMMAND
 DROP TABLE IF EXISTS bug23037;
 DROP FUNCTION IF EXISTS get_value;
 SELECT COLUMN_NAME, MD5(COLUMN_DEFAULT), LENGTH(COLUMN_DEFAULT) FROM INFORMATION_SCHEMA.COLUMNS WHERE TABLE_NAME='bug23037';

mysql-test/r/information_schema_all_engines.result

@@ -33,13 +33,13 @@ INNODB_BUFFER_POOL_PAGES
 PBXT_STATISTICS
 INNODB_CMP
 INNODB_RSEG
-XTRADB_ENHANCEMENTS
+INNODB_INDEX_STATS
 INNODB_BUFFER_POOL_PAGES_INDEX
 XTRADB_ADMIN_COMMAND
 INNODB_TRX
 INNODB_SYS_TABLES
 INNODB_LOCK_WAITS
-INNODB_SYS_STATS
+INNODB_BUFFER_POOL_STATS
 INNODB_LOCKS
 INNODB_CMPMEM
 INNODB_TABLE_STATS
@@ -47,7 +47,10 @@ INNODB_SYS_INDEXES
 INNODB_CMP_RESET
 INNODB_BUFFER_POOL_PAGES_BLOB
 INNODB_CMPMEM_RESET
-INNODB_INDEX_STATS
+INNODB_BUFFER_PAGE
+INNODB_CHANGED_PAGES
+INNODB_SYS_STATS
+INNODB_BUFFER_PAGE_LRU
 SELECT t.table_name, c1.column_name
 FROM information_schema.tables t
 INNER JOIN
@@ -95,13 +98,13 @@ INNODB_BUFFER_POOL_PAGES	page_type
 PBXT_STATISTICS	ID
 INNODB_CMP	page_size
 INNODB_RSEG	rseg_id
-XTRADB_ENHANCEMENTS	name
+INNODB_INDEX_STATS	table_schema
 INNODB_BUFFER_POOL_PAGES_INDEX	index_id
 XTRADB_ADMIN_COMMAND	result_message
 INNODB_TRX	trx_id
 INNODB_SYS_TABLES	SCHEMA
 INNODB_LOCK_WAITS	requesting_trx_id
-INNODB_SYS_STATS	INDEX_ID
+INNODB_BUFFER_POOL_STATS	POOL_SIZE
 INNODB_LOCKS	lock_id
 INNODB_CMPMEM	page_size
 INNODB_TABLE_STATS	table_schema
@@ -109,7 +112,10 @@ INNODB_SYS_INDEXES	TABLE_ID
 INNODB_CMP_RESET	page_size
 INNODB_BUFFER_POOL_PAGES_BLOB	space_id
 INNODB_CMPMEM_RESET	page_size
-INNODB_INDEX_STATS	table_schema
+INNODB_BUFFER_PAGE	BLOCK_ID
+INNODB_CHANGED_PAGES	space_id
+INNODB_SYS_STATS	INDEX_ID
+INNODB_BUFFER_PAGE_LRU	LRU_POSITION
 SELECT t.table_name, c1.column_name
 FROM information_schema.tables t
 INNER JOIN
@@ -157,13 +163,13 @@ INNODB_BUFFER_POOL_PAGES	page_type
 PBXT_STATISTICS	ID
 INNODB_CMP	page_size
 INNODB_RSEG	rseg_id
-XTRADB_ENHANCEMENTS	name
+INNODB_INDEX_STATS	table_schema
 INNODB_BUFFER_POOL_PAGES_INDEX	index_id
 XTRADB_ADMIN_COMMAND	result_message
 INNODB_TRX	trx_id
 INNODB_SYS_TABLES	SCHEMA
 INNODB_LOCK_WAITS	requesting_trx_id
-INNODB_SYS_STATS	INDEX_ID
+INNODB_BUFFER_POOL_STATS	POOL_SIZE
 INNODB_LOCKS	lock_id
 INNODB_CMPMEM	page_size
 INNODB_TABLE_STATS	table_schema
@@ -171,7 +177,10 @@ INNODB_SYS_INDEXES	TABLE_ID
 INNODB_CMP_RESET	page_size
 INNODB_BUFFER_POOL_PAGES_BLOB	space_id
 INNODB_CMPMEM_RESET	page_size
-INNODB_INDEX_STATS	table_schema
+INNODB_BUFFER_PAGE	BLOCK_ID
+INNODB_CHANGED_PAGES	space_id
+INNODB_SYS_STATS	INDEX_ID
+INNODB_BUFFER_PAGE_LRU	LRU_POSITION
 select 1 as f1 from information_schema.tables  where "CHARACTER_SETS"=
 (select cast(table_name as char)  from information_schema.tables
 order by table_name limit 1) limit 1;
@@ -203,9 +212,13 @@ EVENTS	information_schema.EVENTS	1
 FILES	information_schema.FILES	1
 GLOBAL_STATUS	information_schema.GLOBAL_STATUS	1
 GLOBAL_VARIABLES	information_schema.GLOBAL_VARIABLES	1
+INNODB_BUFFER_PAGE	information_schema.INNODB_BUFFER_PAGE	1
+INNODB_BUFFER_PAGE_LRU	information_schema.INNODB_BUFFER_PAGE_LRU	1
 INNODB_BUFFER_POOL_PAGES	information_schema.INNODB_BUFFER_POOL_PAGES	1
 INNODB_BUFFER_POOL_PAGES_BLOB	information_schema.INNODB_BUFFER_POOL_PAGES_BLOB	1
 INNODB_BUFFER_POOL_PAGES_INDEX	information_schema.INNODB_BUFFER_POOL_PAGES_INDEX	1
+INNODB_BUFFER_POOL_STATS	information_schema.INNODB_BUFFER_POOL_STATS	1
+INNODB_CHANGED_PAGES	information_schema.INNODB_CHANGED_PAGES	1
 INNODB_CMP	information_schema.INNODB_CMP	1
 INNODB_CMPMEM	information_schema.INNODB_CMPMEM	1
 INNODB_CMPMEM_RESET	information_schema.INNODB_CMPMEM_RESET	1
@@ -238,7 +251,6 @@ TABLE_PRIVILEGES	information_schema.TABLE_PRIVILEGES	1
 TRIGGERS	information_schema.TRIGGERS	1
 USER_PRIVILEGES	information_schema.USER_PRIVILEGES	1
 VIEWS	information_schema.VIEWS	1
-XTRADB_ENHANCEMENTS	information_schema.XTRADB_ENHANCEMENTS	1
 Database: information_schema
 +---------------------------------------+
 |                Tables                 |
@@ -275,13 +287,13 @@ Database: information_schema
 | PBXT_STATISTICS                       |
 | INNODB_CMP                            |
 | INNODB_RSEG                           |
-| XTRADB_ENHANCEMENTS                   |
+| INNODB_INDEX_STATS                    |
 | INNODB_BUFFER_POOL_PAGES_INDEX        |
 | XTRADB_ADMIN_COMMAND                  |
 | INNODB_TRX                            |
 | INNODB_SYS_TABLES                     |
 | INNODB_LOCK_WAITS                     |
-| INNODB_SYS_STATS                      |
+| INNODB_BUFFER_POOL_STATS              |
 | INNODB_LOCKS                          |
 | INNODB_CMPMEM                         |
 | INNODB_TABLE_STATS                    |
@@ -289,7 +301,10 @@ Database: information_schema
 | INNODB_CMP_RESET                      |
 | INNODB_BUFFER_POOL_PAGES_BLOB         |
 | INNODB_CMPMEM_RESET                   |
-| INNODB_INDEX_STATS                    |
+| INNODB_BUFFER_PAGE                    |
+| INNODB_CHANGED_PAGES                  |
+| INNODB_SYS_STATS                      |
+| INNODB_BUFFER_PAGE_LRU                |
 +---------------------------------------+
 Database: INFORMATION_SCHEMA
 +---------------------------------------+
@@ -327,13 +342,13 @@ Database: INFORMATION_SCHEMA
 | PBXT_STATISTICS                       |
 | INNODB_CMP                            |
 | INNODB_RSEG                           |
-| XTRADB_ENHANCEMENTS                   |
+| INNODB_INDEX_STATS                    |
 | INNODB_BUFFER_POOL_PAGES_INDEX        |
 | XTRADB_ADMIN_COMMAND                  |
 | INNODB_TRX                            |
 | INNODB_SYS_TABLES                     |
 | INNODB_LOCK_WAITS                     |
-| INNODB_SYS_STATS                      |
+| INNODB_BUFFER_POOL_STATS              |
 | INNODB_LOCKS                          |
 | INNODB_CMPMEM                         |
 | INNODB_TABLE_STATS                    |
@@ -341,7 +356,10 @@ Database: INFORMATION_SCHEMA
 | INNODB_CMP_RESET                      |
 | INNODB_BUFFER_POOL_PAGES_BLOB         |
 | INNODB_CMPMEM_RESET                   |
-| INNODB_INDEX_STATS                    |
+| INNODB_BUFFER_PAGE                    |
+| INNODB_CHANGED_PAGES                  |
+| INNODB_SYS_STATS                      |
+| INNODB_BUFFER_PAGE_LRU                |
 +---------------------------------------+
 Wildcard: inf_rmation_schema
 +--------------------+
@@ -351,5 +369,5 @@ Wildcard: inf_rmation_schema
 +--------------------+
 SELECT table_schema, count(*) FROM information_schema.TABLES WHERE table_schema IN ('mysql', 'INFORMATION_SCHEMA', 'test', 'mysqltest') AND table_name<>'ndb_binlog_index' AND table_name<>'ndb_apply_status' GROUP BY TABLE_SCHEMA;
 table_schema	count(*)
-information_schema	47
+information_schema	50
 mysql	22

mysql-test/r/innodb.result

@@ -3198,7 +3198,7 @@ c21 CHAR(255), c22 CHAR(255), c23 CHAR(255), c24 CHAR(255),
 c25 CHAR(255), c26 CHAR(255), c27 CHAR(255), c28 CHAR(255),
 c29 CHAR(255), c30 CHAR(255), c31 CHAR(255), c32 CHAR(255)
 ) ENGINE = InnoDB;
-ERROR 42000: Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. You have to change some columns to TEXT or BLOBs
+ERROR 42000: Row size too large (> 8126). Changing some columns to TEXT or BLOB or using ROW_FORMAT=DYNAMIC or ROW_FORMAT=COMPRESSED may help. In current row format, BLOB prefix of 768 bytes is stored inline.
 DROP TABLE IF EXISTS t1;
 Warnings:
 Note	1051	Unknown table 't1'

mysql-test/suite/innodb_plugin/r/innodb-index.result

@@ -1096,7 +1096,7 @@ PRIMARY KEY (b(10), a), INDEX (c(10))
 INSERT INTO bug12547647 VALUES (5,repeat('khdfo5AlOq',1900),repeat('g',7731));
 COMMIT;
 UPDATE bug12547647 SET c = REPEAT('b',16928);
-ERROR 42000: Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. You have to change some columns to TEXT or BLOBs
+ERROR 42000: Row size too large (> 8126). Changing some columns to TEXT or BLOB may help. In current row format, BLOB prefix of 0 bytes is stored inline.
 DROP TABLE bug12547647;
 SET @r=REPEAT('a',500);
 CREATE TABLE t1(a INT,

mysql-test/suite/innodb_plugin/r/innodb-zip.result

@@ -125,12 +125,12 @@ CREATE TABLE t1(
 c TEXT NOT NULL, d TEXT NOT NULL,
 PRIMARY KEY (c(767),d(767)))
 ENGINE=InnoDB ROW_FORMAT=COMPRESSED KEY_BLOCK_SIZE=1 CHARSET=ASCII;
-ERROR 42000: Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. You have to change some columns to TEXT or BLOBs
+ERROR 42000: Row size too large (> 8126). Changing some columns to TEXT or BLOB may help. In current row format, BLOB prefix of 0 bytes is stored inline.
 CREATE TABLE t1(
 c TEXT NOT NULL, d TEXT NOT NULL,
 PRIMARY KEY (c(767),d(767)))
 ENGINE=InnoDB ROW_FORMAT=COMPRESSED KEY_BLOCK_SIZE=2 CHARSET=ASCII;
-ERROR 42000: Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. You have to change some columns to TEXT or BLOBs
+ERROR 42000: Row size too large (> 8126). Changing some columns to TEXT or BLOB may help. In current row format, BLOB prefix of 0 bytes is stored inline.
 CREATE TABLE t1(
 c TEXT NOT NULL, d TEXT NOT NULL,
 PRIMARY KEY (c(767),d(767)))
@@ -138,7 +138,7 @@ ENGINE=InnoDB ROW_FORMAT=COMPRESSED KEY_BLOCK_SIZE=4 CHARSET=ASCII;
 drop table t1;
 CREATE TABLE t1(c TEXT, PRIMARY KEY (c(440)))
 ENGINE=InnoDB ROW_FORMAT=COMPRESSED KEY_BLOCK_SIZE=1 CHARSET=ASCII;
-ERROR 42000: Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. You have to change some columns to TEXT or BLOBs
+ERROR 42000: Row size too large (> 8126). Changing some columns to TEXT or BLOB may help. In current row format, BLOB prefix of 0 bytes is stored inline.
 CREATE TABLE t1(c TEXT, PRIMARY KEY (c(438)))
 ENGINE=InnoDB ROW_FORMAT=COMPRESSED KEY_BLOCK_SIZE=1 CHARSET=ASCII;
 INSERT INTO t1 VALUES(REPEAT('A',512)),(REPEAT('B',512));

mysql-test/suite/innodb_plugin/r/innodb.result

@@ -3151,7 +3151,7 @@ c21 CHAR(255), c22 CHAR(255), c23 CHAR(255), c24 CHAR(255),
 c25 CHAR(255), c26 CHAR(255), c27 CHAR(255), c28 CHAR(255),
 c29 CHAR(255), c30 CHAR(255), c31 CHAR(255), c32 CHAR(255)
 ) ENGINE = InnoDB;
-ERROR 42000: Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. You have to change some columns to TEXT or BLOBs
+ERROR 42000: Row size too large (> 8126). Changing some columns to TEXT or BLOB or using ROW_FORMAT=DYNAMIC or ROW_FORMAT=COMPRESSED may help. In current row format, BLOB prefix of 768 bytes is stored inline.
 DROP TABLE IF EXISTS t1;
 Warnings:
 Note	1051	Unknown table 't1'

mysql-test/suite/innodb_plugin/r/innodb_bug53591.result

@@ -8,7 +8,7 @@ ERROR HY000: Too big row
 SHOW WARNINGS;
 Level	Code	Message
 Error	139	Too big row
-Error	1118	Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. You have to change some columns to TEXT or BLOBs
+Error	1118	Row size too large (> 8126). Changing some columns to TEXT or BLOB may help. In current row format, BLOB prefix of 0 bytes is stored inline.
 DROP TABLE bug53591;
 SET GLOBAL innodb_file_format=Antelope;
 SET GLOBAL innodb_file_per_table=0;

mysql-test/suite/innodb_plugin/r/innodb_misc1.result

@@ -774,7 +774,7 @@ c21 CHAR(255), c22 CHAR(255), c23 CHAR(255), c24 CHAR(255),
 c25 CHAR(255), c26 CHAR(255), c27 CHAR(255), c28 CHAR(255),
 c29 CHAR(255), c30 CHAR(255), c31 CHAR(255), c32 CHAR(255)
 ) ENGINE = InnoDB;
-ERROR 42000: Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. You have to change some columns to TEXT or BLOBs
+ERROR 42000: Row size too large (> 8126). Changing some columns to TEXT or BLOB or using ROW_FORMAT=DYNAMIC or ROW_FORMAT=COMPRESSED may help. In current row format, BLOB prefix of 768 bytes is stored inline.
 SET innodb_strict_mode=OFF;
 DROP TABLE IF EXISTS t1;
 Warnings:

storage/xtradb/CMakeLists.txt

@@ -62,7 +62,7 @@ SET(XTRADB_SOURCES	btr/btr0btr.c btr/btr0cur.c btr/btr0pcur.c btr/btr0sea.c
 			ibuf/ibuf0ibuf.c
 			pars/lexyy.c pars/pars0grm.c pars/pars0opt.c pars/pars0pars.c pars/pars0sym.c
 			lock/lock0lock.c lock/lock0iter.c
-			log/log0log.c log/log0recv.c
+			log/log0log.c log/log0recv.c log/log0online.c
 			mach/mach0data.c
 			mem/mem0mem.c mem/mem0pool.c
 			mtr/mtr0log.c mtr/mtr0mtr.c

storage/xtradb/ChangeLog

@@ -1,3 +1,32 @@
+2012-08-29	The InnoDB Team
+
+	* btr/btr0btr.c, page/page0cur.c, page/page0page.c:
+	Fix Bug#14554000 CRASH IN PAGE_REC_GET_NTH_CONST(NTH=0)
+	DURING COMPRESSED PAGE SPLIT
+
+2012-08-16	The InnoDB Team
+
+	* btr/btr0cur.c:
+	Fix Bug#12595091 POSSIBLY INVALID ASSERTION IN
+	BTR_CUR_PESSIMISTIC_UPDATE()
+
+2012-08-16	The InnoDB Team
+
+	* btr/btr0btr.c, btr/btr0cur.c:
+	Fix Bug#12845774 OPTIMISTIC INSERT/UPDATE USES WRONG HEURISTICS FOR
+	COMPRESSED PAGE SIZE
+
+2012-08-16	The InnoDB Team
+
+	* btr/btr0cur.c, page/page0page.c:
+	Fix Bug#13523839 ASSERTION FAILURES ON COMPRESSED INNODB TABLES
+
+2012-08-07	The InnoDB Team
+
+	* btr/btr0pcur.c, row/row0merge.c:
+	Fix Bug#14399148 INNODB TABLES UNDER LOAD PRODUCE DUPLICATE COPIES
+	OF ROWS IN QUERIES
+
 2012-03-15	The InnoDB Team
 
 	* fil/fil0fil.c, ibuf/ibuf0ibuf.c, include/fil0fil.h,

storage/xtradb/Makefile.am

@@ -101,6 +101,7 @@ noinst_HEADERS=		\
 			include/lock0types.h	\
 			include/log0log.h	\
 			include/log0log.ic	\
+			include/log0online.h	\
 			include/log0recv.h	\
 			include/log0recv.ic	\
 			include/mach0data.h	\
@@ -226,7 +227,6 @@ noinst_HEADERS=		\
 			include/ut0vec.h	\
 			include/ut0vec.ic	\
 			include/ut0wqueue.h	\
-			handler/innodb_patch_info.h	\
 			mem/mem0dbg.c
 
 noinst_LTLIBRARIES=	@plugin_xtradb_static_target@
@@ -265,6 +265,7 @@ libxtradb_la_SOURCES=	\
 			lock/lock0iter.c		\
 			lock/lock0lock.c		\
 			log/log0log.c			\
+			log/log0online.c		\
 			log/log0recv.c			\
 			mach/mach0data.c		\
 			mem/mem0mem.c			\

storage/xtradb/btr/btr0btr.c

@@ -664,6 +664,12 @@ btr_root_fseg_validate(
 {
 	ulint	offset = mach_read_from_2(seg_header + FSEG_HDR_OFFSET);
 
+	if (UNIV_UNLIKELY(srv_pass_corrupt_table)) {
+		return (mach_read_from_4(seg_header + FSEG_HDR_SPACE) == space)
+			&& (offset >= FIL_PAGE_DATA)
+			&& (offset <= UNIV_PAGE_SIZE - FIL_PAGE_DATA_END);
+	}
+
 	ut_a(mach_read_from_4(seg_header + FSEG_HDR_SPACE) == space);
 	ut_a(offset >= FIL_PAGE_DATA);
 	ut_a(offset <= UNIV_PAGE_SIZE - FIL_PAGE_DATA_END);
@@ -704,6 +710,17 @@ btr_root_block_get(
 	if (!dict_index_is_ibuf(index)) {
 		const page_t*	root = buf_block_get_frame(block);
 
+		if (UNIV_UNLIKELY(srv_pass_corrupt_table)) {
+			if (!btr_root_fseg_validate(FIL_PAGE_DATA
+						    + PAGE_BTR_SEG_LEAF
+						    + root, space))
+				return(NULL);
+			if (!btr_root_fseg_validate(FIL_PAGE_DATA
+						    + PAGE_BTR_SEG_TOP
+						    + root, space))
+				return(NULL);
+			return(block);
+		}
 		ut_a(btr_root_fseg_validate(FIL_PAGE_DATA + PAGE_BTR_SEG_LEAF
 					    + root, space));
 		ut_a(btr_root_fseg_validate(FIL_PAGE_DATA + PAGE_BTR_SEG_TOP
@@ -1852,6 +1869,7 @@ btr_root_raise_and_insert(
 	root = btr_cur_get_page(cursor);
 	root_block = btr_cur_get_block(cursor);
 	root_page_zip = buf_block_get_page_zip(root_block);
+	ut_ad(page_get_n_recs(root) > 0);
 #ifdef UNIV_ZIP_DEBUG
 	ut_a(!root_page_zip || page_zip_validate(root_page_zip, root));
 #endif /* UNIV_ZIP_DEBUG */
@@ -2332,12 +2350,20 @@ btr_insert_on_non_leaf_level_func(
 				    BTR_CONT_MODIFY_TREE,
 				    &cursor, 0, file, line, mtr);
 
-	err = btr_cur_pessimistic_insert(BTR_NO_LOCKING_FLAG
-					 | BTR_KEEP_SYS_FLAG
-					 | BTR_NO_UNDO_LOG_FLAG,
-					 &cursor, tuple, &rec,
-					 &dummy_big_rec, 0, NULL, mtr);
-	ut_a(err == DB_SUCCESS);
+	ut_ad(cursor.flag == BTR_CUR_BINARY);
+
+	err = btr_cur_optimistic_insert(
+		BTR_NO_LOCKING_FLAG | BTR_KEEP_SYS_FLAG
+		| BTR_NO_UNDO_LOG_FLAG, &cursor, tuple, &rec,
+		&dummy_big_rec, 0, NULL, mtr);
+
+	if (err == DB_FAIL) {
+		err = btr_cur_pessimistic_insert(
+			BTR_NO_LOCKING_FLAG | BTR_KEEP_SYS_FLAG
+			| BTR_NO_UNDO_LOG_FLAG,
+			&cursor, tuple, &rec, &dummy_big_rec, 0, NULL, mtr);
+		ut_a(err == DB_SUCCESS);
+	}
 }
 
 /**************************************************************//**
@@ -3262,6 +3288,7 @@ btr_compress(
 
 	if (adjust) {
 		nth_rec = page_rec_get_n_recs_before(btr_cur_get_rec(cursor));
+		ut_ad(nth_rec > 0);
 	}
 
 	/* Decide the page to which we try to merge and which will inherit
@@ -3497,6 +3524,7 @@ func_exit:
 	mem_heap_free(heap);
 
 	if (adjust) {
+		ut_ad(nth_rec > 0);
 		btr_cur_position(
 			index,
 			page_rec_get_nth(merge_block->frame, nth_rec),
@@ -4009,8 +4037,22 @@ btr_index_page_validate(
 {
 	page_cur_t	cur;
 	ibool		ret	= TRUE;
+#ifndef DBUG_OFF
+	ulint		nth	= 1;
+#endif /* !DBUG_OFF */
 
 	page_cur_set_before_first(block, &cur);
+
+	/* Directory slot 0 should only contain the infimum record. */
+	DBUG_EXECUTE_IF("check_table_rec_next",
+			ut_a(page_rec_get_nth_const(
+				     page_cur_get_page(&cur), 0)
+			     == cur.rec);
+			ut_a(page_dir_slot_get_n_owned(
+				     page_dir_get_nth_slot(
+					     page_cur_get_page(&cur), 0))
+			     == 1););
+
 	page_cur_move_to_next(&cur);
 
 	for (;;) {
@@ -4024,6 +4066,16 @@ btr_index_page_validate(
 			return(FALSE);
 		}
 
+		/* Verify that page_rec_get_nth_const() is correctly
+		retrieving each record. */
+		DBUG_EXECUTE_IF("check_table_rec_next",
+				ut_a(cur.rec == page_rec_get_nth_const(
+					     page_cur_get_page(&cur),
+					     page_rec_get_n_recs_before(
+						     cur.rec)));
+				ut_a(nth++ == page_rec_get_n_recs_before(
+					     cur.rec)););
+
 		page_cur_move_to_next(&cur);
 	}
 
@@ -4435,6 +4487,12 @@ btr_validate_index(
 	mtr_x_lock(dict_index_get_lock(index), &mtr);
 
 	root = btr_root_get(index, &mtr);
+
+	if (UNIV_UNLIKELY(srv_pass_corrupt_table && !root)) {
+		mtr_commit(&mtr);
+		return(FALSE);
+	}
+
 	n = btr_page_get_level(root, &mtr);
 
 	for (i = 0; i <= n && !trx_is_interrupted(trx); i++) {

storage/xtradb/btr/btr0cur.c

@@ -1307,7 +1307,12 @@ fail_err:
 
 		if (UNIV_UNLIKELY(reorg)) {
 			ut_a(zip_size);
-			ut_a(*rec);
+			/* It's possible for rec to be NULL if the
+			page is compressed.  This is because a
+			reorganized page may become incompressible. */
+			if (!*rec) {
+				goto fail;
+			}
 		}
 	}
 
@@ -1443,20 +1448,9 @@ btr_cur_pessimistic_insert(
 	ut_ad((thr && thr_get_trx(thr)->fake_changes) || mtr_memo_contains(mtr, btr_cur_get_block(cursor),
 				MTR_MEMO_PAGE_X_FIX));
 
-	/* Try first an optimistic insert; reset the cursor flag: we do not
-	assume anything of how it was positioned */
-
 	cursor->flag = BTR_CUR_BINARY;
 
-	err = btr_cur_optimistic_insert(flags, cursor, entry, rec,
-					big_rec, n_ext, thr, mtr);
-	if (err != DB_FAIL) {
-
-		return(err);
-	}
-
-	/* Retry with a pessimistic insert. Check locks and write to undo log,
-	if specified */
+	/* Check locks and write to undo log, if specified */
 
 	err = btr_cur_ins_lock_and_undo(flags, cursor, entry,
 					thr, mtr, &dummy_inh);
@@ -2083,8 +2077,12 @@ any_extern:
 		goto err_exit;
 	}
 
-	max_size = old_rec_size
-		+ page_get_max_insert_size_after_reorganize(page, 1);
+	/* We do not attempt to reorganize if the page is compressed.
+	This is because the page may fail to compress after reorganization. */
+	max_size = page_zip
+		? page_get_max_insert_size(page, 1)
+		: (old_rec_size
+		   + page_get_max_insert_size_after_reorganize(page, 1));
 
 	if (!(((max_size >= BTR_CUR_PAGE_REORGANIZE_LIMIT)
 	       && (max_size >= new_rec_size))
@@ -2452,7 +2450,12 @@ make_external:
 		err = DB_SUCCESS;
 		goto return_after_reservations;
 	} else {
-		ut_a(optim_err != DB_UNDERFLOW);
+		/* If the page is compressed and it initially
+		compresses very well, and there is a subsequent insert
+		of a badly-compressing record, it is possible for
+		btr_cur_optimistic_update() to return DB_UNDERFLOW and
+		btr_cur_insert_if_possible() to return FALSE. */
+		ut_a(page_zip || optim_err != DB_UNDERFLOW);
 
 		/* Out of space: reset the free bits. */
 		if (!dict_index_is_clust(index)
@@ -2480,8 +2483,10 @@ make_external:
 	record on its page? */
 	was_first = page_cur_is_before_first(page_cursor);
 
-	/* The first parameter means that no lock checking and undo logging
-	is made in the insert */
+	/* Lock checks and undo logging were already performed by
+	btr_cur_upd_lock_and_undo(). We do not try
+	btr_cur_optimistic_insert() because
+	btr_cur_insert_if_possible() already failed above. */
 
 	err = btr_cur_pessimistic_insert(BTR_NO_UNDO_LOG_FLAG
 					 | BTR_NO_LOCKING_FLAG
@@ -3483,6 +3488,8 @@ btr_estimate_n_rows_in_range(
 				n_rows = n_rows * 2;
 			}
 
+			DBUG_EXECUTE_IF("bug14007649", return(n_rows););
+
 			/* Do not estimate the number of rows in the range
 			to over 1 / 2 of the estimated rows in the whole
 			table */
@@ -3582,9 +3589,9 @@ btr_record_not_null_field_in_rec(
 
 	for (i = 0; i < n_unique; i++) {
 		if (rec_offs_nth_sql_null(offsets, i)) {
-			/* Break if we hit the first NULL value */
 			break;
 		}
+
 		n_not_null[i]++;
 	}
 }

storage/xtradb/btr/btr0pcur.c

@@ -342,44 +342,39 @@ btr_pcur_restore_position_func(
 	/* Restore the old search mode */
 	cursor->search_mode = old_mode;
 
-	if (btr_pcur_is_on_user_rec(cursor)) {
-		switch (cursor->rel_pos) {
-		case BTR_PCUR_ON:
-			if (!cmp_dtuple_rec(
-				    tuple, btr_pcur_get_rec(cursor),
-				    rec_get_offsets(btr_pcur_get_rec(cursor),
-						    index, NULL,
-						    ULINT_UNDEFINED, &heap))) {
+	switch (cursor->rel_pos) {
+	case BTR_PCUR_ON:
+		if (btr_pcur_is_on_user_rec(cursor)
+		    && !cmp_dtuple_rec(
+			    tuple, btr_pcur_get_rec(cursor),
+			    rec_get_offsets(btr_pcur_get_rec(cursor),
+					    index, NULL,
+					    ULINT_UNDEFINED, &heap))) {
 
-				/* We have to store the NEW value for
-				the modify clock, since the cursor can
-				now be on a different page! But we can
-				retain the value of old_rec */
+			/* We have to store the NEW value for
+			the modify clock, since the cursor can
+			now be on a different page! But we can
+			retain the value of old_rec */
 
-				cursor->block_when_stored =
-					btr_pcur_get_block(cursor);
-				cursor->modify_clock =
-					buf_block_get_modify_clock(
-						cursor->block_when_stored);
-				cursor->old_stored = BTR_PCUR_OLD_STORED;
+			cursor->block_when_stored =
+				btr_pcur_get_block(cursor);
+			cursor->modify_clock =
+				buf_block_get_modify_clock(
+					cursor->block_when_stored);
+			cursor->old_stored = BTR_PCUR_OLD_STORED;
 
-				mem_heap_free(heap);
+			mem_heap_free(heap);
 
-				return(TRUE);
-			}
-
-			break;
-		case BTR_PCUR_BEFORE:
-			page_cur_move_to_next(btr_pcur_get_page_cur(cursor));
-			break;
-		case BTR_PCUR_AFTER:
-			page_cur_move_to_prev(btr_pcur_get_page_cur(cursor));
-			break;
-#ifdef UNIV_DEBUG
-		default:
-			ut_error;
-#endif /* UNIV_DEBUG */
+			return(TRUE);
 		}
+#ifdef UNIV_DEBUG
+		/* fall through */
+	case BTR_PCUR_BEFORE:
+	case BTR_PCUR_AFTER:
+		break;
+	default:
+		ut_error;
+#endif /* UNIV_DEBUG */
 	}
 
 	mem_heap_free(heap);

storage/xtradb/buf/buf0buf.c

@@ -2123,6 +2123,7 @@ wait_until_unfixed:
 	if (mode == BUF_GET_IF_IN_POOL && ibuf_debug) {
 		/* Try to evict the block from the buffer pool, to use the
 		insert buffer as much as possible. */
+		ulint	page_no	= buf_block_get_page_no(block);
 
 		if (buf_LRU_free_block(&block->page, TRUE, FALSE)) {
 			//buf_pool_mutex_exit();
@@ -2131,6 +2132,18 @@ wait_until_unfixed:
 				"innodb_change_buffering_debug evict %u %u\n",
 				(unsigned) space, (unsigned) offset);
 			return(NULL);
+		} else if (UNIV_UNLIKELY(buf_block_get_state(block)
+					 != BUF_BLOCK_FILE_PAGE
+				|| (buf_block_get_page_no(block) != page_no)
+				|| (buf_block_get_space(block) != space))) {
+
+				/* buf_LRU_free_block temporarily releases the
+				block mutex, and now block points to something
+				else. */
+				mutex_exit(block_mutex);
+				block = NULL;
+				goto loop2;
+
 		} else if (buf_flush_page_try(block)) {
 			fprintf(stderr,
 				"innodb_change_buffering_debug flush %u %u\n",
@@ -4078,6 +4091,133 @@ buf_get_free_list_len(void)
 
 	return(len);
 }
+
+/*******************************************************************//**
+Collect buffer pool stats information for a buffer pool. Also
+record aggregated stats if there are more than one buffer pool
+in the server */
+UNIV_INTERN
+void
+buf_stats_get_pool_info(
+/*====================*/
+	buf_pool_info_t*	pool_info)	/*!< in/out: buffer pool info
+						to fill */
+{
+	time_t			current_time;
+	double			time_elapsed;
+
+	buf_pool_mutex_enter();
+
+	pool_info->pool_size = buf_pool->curr_size;
+
+	pool_info->lru_len = UT_LIST_GET_LEN(buf_pool->LRU);
+
+	pool_info->old_lru_len = buf_pool->LRU_old_len;
+
+	pool_info->free_list_len = UT_LIST_GET_LEN(buf_pool->free);
+
+	pool_info->flush_list_len = UT_LIST_GET_LEN(buf_pool->flush_list);
+
+	pool_info->n_pend_unzip = UT_LIST_GET_LEN(buf_pool->unzip_LRU);
+
+	pool_info->n_pend_reads = buf_pool->n_pend_reads;
+
+	pool_info->n_pending_flush_lru =
+		 (buf_pool->n_flush[BUF_FLUSH_LRU]
+		  + buf_pool->init_flush[BUF_FLUSH_LRU]);
+
+	pool_info->n_pending_flush_list =
+		 (buf_pool->n_flush[BUF_FLUSH_LIST]
+		  + buf_pool->init_flush[BUF_FLUSH_LIST]);
+
+	pool_info->n_pending_flush_single_page =
+		 (buf_pool->n_flush[BUF_FLUSH_SINGLE_PAGE]
+		  + buf_pool->init_flush[BUF_FLUSH_SINGLE_PAGE]);
+
+	current_time = time(NULL);
+	time_elapsed = 0.001 + difftime(current_time,
+					buf_pool->last_printout_time);
+
+	pool_info->n_pages_made_young = buf_pool->stat.n_pages_made_young;
+
+	pool_info->n_pages_not_made_young =
+		buf_pool->stat.n_pages_not_made_young;
+
+	pool_info->n_pages_read = buf_pool->stat.n_pages_read;
+
+	pool_info->n_pages_created = buf_pool->stat.n_pages_created;
+
+	pool_info->n_pages_written = buf_pool->stat.n_pages_written;
+
+	pool_info->n_page_gets = buf_pool->stat.n_page_gets;
+
+	pool_info->n_ra_pages_read_rnd = buf_pool->stat.n_ra_pages_read_rnd;
+	pool_info->n_ra_pages_read = buf_pool->stat.n_ra_pages_read;
+
+	pool_info->n_ra_pages_evicted = buf_pool->stat.n_ra_pages_evicted;
+
+	pool_info->page_made_young_rate =
+		 (buf_pool->stat.n_pages_made_young
+		  - buf_pool->old_stat.n_pages_made_young) / time_elapsed;
+
+	pool_info->page_not_made_young_rate =
+		 (buf_pool->stat.n_pages_not_made_young
+		  - buf_pool->old_stat.n_pages_not_made_young) / time_elapsed;
+
+	pool_info->pages_read_rate =
+		(buf_pool->stat.n_pages_read
+		  - buf_pool->old_stat.n_pages_read) / time_elapsed;
+
+	pool_info->pages_created_rate =
+		(buf_pool->stat.n_pages_created
+		 - buf_pool->old_stat.n_pages_created) / time_elapsed;
+
+	pool_info->pages_written_rate =
+		(buf_pool->stat.n_pages_written
+		 - buf_pool->old_stat.n_pages_written) / time_elapsed;
+
+	pool_info->n_page_get_delta = buf_pool->stat.n_page_gets
+				      - buf_pool->old_stat.n_page_gets;
+
+	if (pool_info->n_page_get_delta) {
+		pool_info->page_read_delta = buf_pool->stat.n_pages_read
+					     - buf_pool->old_stat.n_pages_read;
+
+		pool_info->young_making_delta =
+			buf_pool->stat.n_pages_made_young
+			- buf_pool->old_stat.n_pages_made_young;
+
+		pool_info->not_young_making_delta =
+			buf_pool->stat.n_pages_not_made_young
+			- buf_pool->old_stat.n_pages_not_made_young;
+	}
+	pool_info->pages_readahead_rnd_rate =
+		 (buf_pool->stat.n_ra_pages_read_rnd
+		  - buf_pool->old_stat.n_ra_pages_read_rnd) / time_elapsed;
+
+
+	pool_info->pages_readahead_rate =
+		 (buf_pool->stat.n_ra_pages_read
+		  - buf_pool->old_stat.n_ra_pages_read) / time_elapsed;
+
+	pool_info->pages_evicted_rate =
+		(buf_pool->stat.n_ra_pages_evicted
+		 - buf_pool->old_stat.n_ra_pages_evicted) / time_elapsed;
+
+	pool_info->unzip_lru_len = UT_LIST_GET_LEN(buf_pool->unzip_LRU);
+
+	pool_info->io_sum = buf_LRU_stat_sum.io;
+
+	pool_info->io_cur = buf_LRU_stat_cur.io;
+
+	pool_info->unzip_sum = buf_LRU_stat_sum.unzip;
+
+	pool_info->unzip_cur = buf_LRU_stat_cur.unzip;
+
+	buf_refresh_io_stats();
+	buf_pool_mutex_exit();
+}
+
 #else /* !UNIV_HOTBACKUP */
 /********************************************************************//**
 Inits a page to the buffer buf_pool, for use in ibbackup --restore. */
@@ -4108,3 +4248,5 @@ buf_page_init_for_backup_restore(
 	}
 }
 #endif /* !UNIV_HOTBACKUP */
+
+

storage/xtradb/buf/buf0lru.c

@@ -48,6 +48,7 @@ Created 11/5/1995 Heikki Tuuri
 #include "page0zip.h"
 #include "log0recv.h"
 #include "srv0srv.h"
+#include "srv0start.h"
 
 /** The number of blocks from the LRU_old pointer onward, including
 the block pointed to, must be buf_LRU_old_ratio/BUF_LRU_OLD_RATIO_DIV
@@ -1428,13 +1429,12 @@ buf_LRU_make_block_old(
 Try to free a block.  If bpage is a descriptor of a compressed-only
 page, the descriptor object will be freed as well.
 
-NOTE: If this function returns TRUE, it will temporarily
-release buf_pool_mutex.  Furthermore, the page frame will no longer be
-accessible via bpage.
+NOTE: This will temporarily release buf_pool_mutex.  Furthermore, the
+page frame will no longer be accessible via bpage.
 
-The caller must hold buf_pool_mutex and buf_page_get_mutex(bpage) and
-release these two mutexes after the call.  No other
-buf_page_get_mutex() may be held when calling this function.
+The caller must hold buf_page_get_mutex(bpage) and release this mutex
+after the call.  No other buf_page_get_mutex() may be held when
+calling this function.
 @return TRUE if freed, FALSE otherwise. */
 UNIV_INTERN
 ibool
@@ -2098,6 +2098,12 @@ func_exit:
 /********************************************************************//**
 Dump the LRU page list to the specific file. */
 #define LRU_DUMP_FILE "ib_lru_dump"
+#define LRU_DUMP_TEMP_FILE "ib_lru_dump.tmp"
+#define LRU_OS_FILE_WRITE() \
+	os_file_write(LRU_DUMP_FILE, dump_file, buffer, \
+		(buffers << UNIV_PAGE_SIZE_SHIFT) & 0xFFFFFFFFUL, \
+		(buffers >> (32 - UNIV_PAGE_SIZE_SHIFT)), \
+		UNIV_PAGE_SIZE)
 
 UNIV_INTERN
 ibool
@@ -2109,17 +2115,19 @@ buf_LRU_file_dump(void)
 	byte*		buffer_base = NULL;
 	byte*		buffer = NULL;
 	buf_page_t*	bpage;
+	buf_page_t*	first_bpage;
 	ulint		buffers;
 	ulint		offset;
-	ibool		ret = FALSE;
+	ulint		pages_written;
 	ulint		i;
+	ulint		total_pages;
 
 	for (i = 0; i < srv_n_data_files; i++) {
 		if (strstr(srv_data_file_names[i], LRU_DUMP_FILE) != NULL) {
 			fprintf(stderr,
 				" InnoDB: The name '%s' seems to be used for"
-				" innodb_data_file_path. Dumping LRU list is not"
-				" done for safeness.\n", LRU_DUMP_FILE);
+				" innodb_data_file_path. Dumping LRU list is"
+				"  not done for safeness.\n", LRU_DUMP_FILE);
 			goto end;
 		}
 	}
@@ -2132,7 +2140,7 @@ buf_LRU_file_dump(void)
 		goto end;
 	}
 
-	dump_file = os_file_create(LRU_DUMP_FILE, OS_FILE_OVERWRITE,
+	dump_file = os_file_create(LRU_DUMP_TEMP_FILE, OS_FILE_OVERWRITE,
 				OS_FILE_NORMAL, OS_DATA_FILE, &success);
 	if (!success) {
 		os_file_get_last_error(TRUE);
@@ -2142,12 +2150,21 @@ buf_LRU_file_dump(void)
 	}
 
 	mutex_enter(&LRU_list_mutex);
-	bpage = UT_LIST_GET_LAST(buf_pool->LRU);
+	bpage = first_bpage = UT_LIST_GET_FIRST(buf_pool->LRU);
+	total_pages = UT_LIST_GET_LEN(buf_pool->LRU);
 
-	buffers = offset = 0;
-	while (bpage != NULL) {
-		if (offset == 0) {
-			memset(buffer, 0, UNIV_PAGE_SIZE);
+	buffers = offset = pages_written = 0;
+	while (bpage != NULL && (pages_written++ < total_pages)) {
+
+		buf_page_t*	next_bpage = UT_LIST_GET_NEXT(LRU, bpage);
+
+		if (next_bpage == first_bpage) {
+			mutex_exit(&LRU_list_mutex);
+			success = FALSE;
+			fprintf(stderr,
+				"InnoDB: detected cycle in LRU, skipping "
+				"dump\n");
+			goto end;
 		}
 
 		mach_write_to_4(buffer + offset * 4, bpage->space);
@@ -2156,50 +2173,79 @@ buf_LRU_file_dump(void)
 		offset++;
 
 		if (offset == UNIV_PAGE_SIZE/4) {
-			success = os_file_write(LRU_DUMP_FILE, dump_file, buffer,
-					(buffers << UNIV_PAGE_SIZE_SHIFT) & 0xFFFFFFFFUL,
-					(buffers >> (32 - UNIV_PAGE_SIZE_SHIFT)),
-					UNIV_PAGE_SIZE);
+			mutex_t		*next_block_mutex = NULL;
+
+			if (srv_shutdown_state != SRV_SHUTDOWN_NONE) {
+				mutex_exit(&LRU_list_mutex);
+				success = FALSE;
+				fprintf(stderr,
+					" InnoDB: stopped dumping lru pages"
+					" because of server shutdown.\n");
+				goto end;
+			}
+
+			/* while writing file, release buffer pool mutex but
+			   keep the next page fixed so we don't worry about
+			   our list iterator becoming invalid */
+			if (next_bpage) {
+				next_block_mutex = buf_page_get_mutex(
+							next_bpage);
+
+				mutex_enter(next_block_mutex);
+				next_bpage->buf_fix_count++;
+				mutex_exit(next_block_mutex);
+			}
+			mutex_exit(&LRU_list_mutex);
+
+			success = LRU_OS_FILE_WRITE();
+
+			/* grab this again here so that next_bpage
+			   can't be purged when we drop the fix_count */
+			mutex_enter(&LRU_list_mutex);
+
+			if (next_bpage) {
+				mutex_enter(next_block_mutex);
+				next_bpage->buf_fix_count--;
+				mutex_exit(next_block_mutex);
+			}
 			if (!success) {
 				mutex_exit(&LRU_list_mutex);
 				fprintf(stderr,
-					" InnoDB: cannot write page %lu of %s\n",
+					" InnoDB: cannot write page"
+					" %lu of %s\n",
 					buffers, LRU_DUMP_FILE);
 				goto end;
 			}
 			buffers++;
 			offset = 0;
+			bpage = next_bpage;
+		} else {
+			bpage = UT_LIST_GET_NEXT(LRU, bpage);
 		}
-
-		bpage = UT_LIST_GET_PREV(LRU, bpage);
-	}
+	} /* while(bpage ...) */
 	mutex_exit(&LRU_list_mutex);
 
-	if (offset == 0) {
-		memset(buffer, 0, UNIV_PAGE_SIZE);
-	}
-
 	mach_write_to_4(buffer + offset * 4, 0xFFFFFFFFUL);
 	offset++;
 	mach_write_to_4(buffer + offset * 4, 0xFFFFFFFFUL);
 	offset++;
 
-	success = os_file_write(LRU_DUMP_FILE, dump_file, buffer,
-			(buffers << UNIV_PAGE_SIZE_SHIFT) & 0xFFFFFFFFUL,
-			(buffers >> (32 - UNIV_PAGE_SIZE_SHIFT)),
-			UNIV_PAGE_SIZE);
-	if (!success) {
-		goto end;
-	}
-
-	ret = TRUE;
+	success = LRU_OS_FILE_WRITE();
 end:
-	if (dump_file != (os_file_t) -1)
+	if (dump_file != (os_file_t) -1) {
+		if (success) {
+			success = os_file_flush(dump_file, TRUE);
+		}
 		os_file_close(dump_file);
+	}
+	if (success) {
+		success = os_file_rename(LRU_DUMP_TEMP_FILE,
+			LRU_DUMP_FILE);
+	}
 	if (buffer_base)
 		ut_free(buffer_base);
 
-	return(ret);
+	return(success);
 }
 
 typedef struct {
@@ -2241,6 +2287,7 @@ buf_LRU_file_restore(void)
 	dump_record_t*	records = NULL;
 	ulint		size;
 	ulint		size_high;
+	ulint 		recsize = sizeof(dump_record_t);
 	ulint		length;
 
 	dump_file = os_file_create_simple_no_error_handling(
@@ -2248,7 +2295,15 @@ buf_LRU_file_restore(void)
 	if (!success || !os_file_get_size(dump_file, &size, &size_high)) {
 		os_file_get_last_error(TRUE);
 		fprintf(stderr,
-			" InnoDB: cannot open %s\n", LRU_DUMP_FILE);
+			" InnoDB: cannot open %s,"
+			" buffer pool preload not done\n",
+			LRU_DUMP_FILE);
+		goto end;
+	}
+
+	if (size == 0 || size_high > 0 || size % recsize) {
+		fprintf(stderr, " InnoDB: broken LRU dump file,"
+			" buffer pool preload not done\n");
 		goto end;
 	}
 
@@ -2332,6 +2387,14 @@ buf_LRU_file_restore(void)
 		if (offset % 16 == 15) {
 			os_aio_simulated_wake_handler_threads();
 			buf_flush_free_margin(FALSE);
+			/* skip loading of the rest of the file if we are
+ 			   terminating anyway*/
+			if (srv_shutdown_state != SRV_SHUTDOWN_NONE) {
+				fprintf(stderr,
+					" InnoDB: stopped loading LRU pages"
+					" because of server shutdown.\n");
+				break;
+			}
 		}
 
 		zip_size = fil_space_get_zip_size(space_id);

storage/xtradb/dict/dict0boot.c

@@ -236,6 +236,166 @@ dict_hdr_create(
 	return(TRUE);
 }
 
+/*****************************************************************//**
+Verifies the SYS_STATS table by scanning its clustered index.  This
+function may only be called at InnoDB startup time.
+
+@return	TRUE if SYS_STATS was verified successfully */
+UNIV_INTERN
+ibool
+dict_verify_xtradb_sys_stats(void)
+/*==============================*/
+{
+	dict_index_t* sys_stats_index;
+	ulint	      saved_srv_pass_corrupt_table = srv_pass_corrupt_table;
+	ibool	      result;
+
+	sys_stats_index = dict_table_get_first_index(dict_sys->sys_stats);
+
+	/* Since this may be called only during server startup, avoid hitting
+	   various asserts by using XtraDB pass_corrupt_table option. */
+	srv_pass_corrupt_table = 1;
+	result = btr_validate_index(sys_stats_index, NULL);
+	srv_pass_corrupt_table = saved_srv_pass_corrupt_table;
+
+	return result;
+}
+
+/*****************************************************************//**
+Creates the B-tree for the SYS_STATS clustered index, adds the XtraDB
+mark and the id of the index to the dictionary header page.  Rewrites
+both passed args. */
+static
+void
+dict_create_xtradb_sys_stats(
+/*=========================*/
+	dict_hdr_t**	dict_hdr,	/*!< in/out: dictionary header */
+	mtr_t*		mtr)		/*!< in/out: mtr */
+{
+	ulint	root_page_no;
+
+	root_page_no = btr_create(DICT_CLUSTERED | DICT_UNIQUE,
+				  DICT_HDR_SPACE, 0, DICT_STATS_ID,
+				  dict_ind_redundant, mtr);
+	if (root_page_no == FIL_NULL) {
+		fprintf(stderr, "InnoDB: Warning: failed to create SYS_STATS btr.\n");
+		srv_use_sys_stats_table = FALSE;
+	} else {
+		mlog_write_ulint(*dict_hdr + DICT_HDR_STATS, root_page_no,
+				 MLOG_4BYTES, mtr);
+		mlog_write_dulint(*dict_hdr + DICT_HDR_XTRADB_MARK,
+				  DICT_HDR_XTRADB_FLAG, mtr);
+	}
+	mtr_commit(mtr);
+	/* restart mtr */
+	mtr_start(mtr);
+	*dict_hdr = dict_hdr_get(mtr);
+}
+
+/*****************************************************************//**
+Create the table and index structure of SYS_STATS for the dictionary
+cache and add it there.  If called for the first time, also support
+wrong root page id injection for testing purposes. */
+static
+void
+dict_add_to_cache_xtradb_sys_stats(
+/*===============================*/
+	ibool		first_time __attribute__((unused)),
+					/*!< in: first invocation flag. If
+					TRUE, optionally inject wrong root page
+					id */
+	mem_heap_t*	heap,		/*!< in: memory heap for table/index
+					allocation */
+	dict_hdr_t*	dict_hdr,	/*!< in: dictionary header */
+	mtr_t*		mtr)		/*!< in: mtr */
+{
+	dict_table_t*	table;
+	dict_index_t*	index;
+	ulint		root_page_id;
+	ulint		error;
+
+	table = dict_mem_table_create("SYS_STATS", DICT_HDR_SPACE, 4, 0);
+	table->n_mysql_handles_opened = 1; /* for pin */
+
+	dict_mem_table_add_col(table, heap, "INDEX_ID", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "KEY_COLS", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "DIFF_VALS", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "NON_NULL_VALS", DATA_BINARY, 0, 0);
+
+	/* The '+ 2' below comes from the fields DB_TRX_ID, DB_ROLL_PTR */
+#if DICT_SYS_STATS_DIFF_VALS_FIELD != 2 + 2
+#error "DICT_SYS_STATS_DIFF_VALS_FIELD != 2 + 2"
+#endif
+#if DICT_SYS_STATS_NON_NULL_VALS_FIELD != 3 + 2
+#error "DICT_SYS_STATS_NON_NULL_VALS_FIELD != 3 + 2"
+#endif
+
+	table->id = DICT_STATS_ID;
+	dict_table_add_to_cache(table, heap);
+	dict_sys->sys_stats = table;
+	mem_heap_empty(heap);
+
+	index = dict_mem_index_create("SYS_STATS", "CLUST_IND",
+				      DICT_HDR_SPACE,
+				      DICT_UNIQUE | DICT_CLUSTERED, 2);
+
+	dict_mem_index_add_field(index, "INDEX_ID", 0);
+	dict_mem_index_add_field(index, "KEY_COLS", 0);
+
+	index->id = DICT_STATS_ID;
+
+	root_page_id = mtr_read_ulint(dict_hdr + DICT_HDR_STATS, MLOG_4BYTES,
+				      mtr);
+#ifdef UNIV_DEBUG
+	if ((srv_sys_stats_root_page != 0) && first_time)
+		root_page_id = srv_sys_stats_root_page;
+#endif
+	error = dict_index_add_to_cache(table, index, root_page_id, FALSE);
+	ut_a(error == DB_SUCCESS);
+
+	mem_heap_empty(heap);
+}
+
+/*****************************************************************//**
+Discard the existing dictionary cache SYS_STATS information, create and
+add it there anew.  Does not touch the old SYS_STATS tablespace page
+under the assumption that they are corrupted or overwritten for other
+purposes. */
+UNIV_INTERN
+void
+dict_recreate_xtradb_sys_stats(void)
+/*================================*/
+{
+	mtr_t		mtr;
+	dict_hdr_t*	dict_hdr;
+	dict_index_t*	sys_stats_clust_idx;
+	mem_heap_t*	heap;
+
+	heap = mem_heap_create(450);
+
+	mutex_enter(&(dict_sys->mutex));
+
+	sys_stats_clust_idx = dict_table_get_first_index(dict_sys->sys_stats);
+	dict_index_remove_from_cache(dict_sys->sys_stats, sys_stats_clust_idx);
+
+	dict_table_remove_from_cache(dict_sys->sys_stats);
+
+	dict_sys->sys_stats = NULL;
+
+	mtr_start(&mtr);
+
+	dict_hdr = dict_hdr_get(&mtr);
+
+	dict_create_xtradb_sys_stats(&dict_hdr, &mtr);
+	dict_add_to_cache_xtradb_sys_stats(FALSE, heap, dict_hdr, &mtr);
+
+	mem_heap_free(heap);
+
+	mtr_commit(&mtr);
+
+	mutex_exit(&(dict_sys->mutex));
+}
+
 /*****************************************************************//**
 Initializes the data dictionary memory structures when the database is
 started. This function is also called when the data dictionary is created. */
@@ -251,39 +411,23 @@ dict_boot(void)
 	mtr_t		mtr;
 	ulint		error;
 
+	heap = mem_heap_create(450);
+
 	mtr_start(&mtr);
 
 	/* Create the hash tables etc. */
 	dict_init();
 
-	heap = mem_heap_create(450);
-
 	mutex_enter(&(dict_sys->mutex));
 
 	/* Get the dictionary header */
 	dict_hdr = dict_hdr_get(&mtr);
 
-	if (ut_dulint_cmp(mtr_read_dulint(dict_hdr + DICT_HDR_XTRADB_MARK, &mtr),
-			  DICT_HDR_XTRADB_FLAG) != 0) {
-		/* not extended yet by XtraDB, need to be extended */
-		ulint	root_page_no;
+	if (ut_dulint_cmp(mtr_read_dulint(dict_hdr + DICT_HDR_XTRADB_MARK,
+					  &mtr), DICT_HDR_XTRADB_FLAG) != 0) {
 
-		root_page_no = btr_create(DICT_CLUSTERED | DICT_UNIQUE,
-					  DICT_HDR_SPACE, 0, DICT_STATS_ID,
-					  dict_ind_redundant, &mtr);
-		if (root_page_no == FIL_NULL) {
-			fprintf(stderr, "InnoDB: Warning: failed to create SYS_STATS btr.\n");
-			srv_use_sys_stats_table = FALSE;
-		} else {
-			mlog_write_ulint(dict_hdr + DICT_HDR_STATS, root_page_no,
-					 MLOG_4BYTES, &mtr);
-			mlog_write_dulint(dict_hdr + DICT_HDR_XTRADB_MARK,
-					  DICT_HDR_XTRADB_FLAG, &mtr);
-		}
-		mtr_commit(&mtr);
-		/* restart mtr */
-		mtr_start(&mtr);
-		dict_hdr = dict_hdr_get(&mtr);
+		/* not extended yet by XtraDB, need to be extended */
+		dict_create_xtradb_sys_stats(&dict_hdr, &mtr);
 	}
 
 	/* Because we only write new row ids to disk-based data structure
@@ -464,42 +608,7 @@ dict_boot(void)
 					FALSE);
 	ut_a(error == DB_SUCCESS);
 
-	/*-------------------------*/
-	table = dict_mem_table_create("SYS_STATS", DICT_HDR_SPACE, 4, 0);
-	table->n_mysql_handles_opened = 1; /* for pin */
-
-	dict_mem_table_add_col(table, heap, "INDEX_ID", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, heap, "KEY_COLS", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, heap, "DIFF_VALS", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, heap, "NON_NULL_VALS", DATA_BINARY, 0, 0);
-
-	/* The '+ 2' below comes from the fields DB_TRX_ID, DB_ROLL_PTR */
-#if DICT_SYS_STATS_DIFF_VALS_FIELD != 2 + 2
-#error "DICT_SYS_STATS_DIFF_VALS_FIELD != 2 + 2"
-#endif
-#if DICT_SYS_STATS_NON_NULL_VALS_FIELD != 3 + 2
-#error "DICT_SYS_STATS_NON_NULL_VALS_FIELD != 3 + 2"
-#endif
-
-	table->id = DICT_STATS_ID;
-	dict_table_add_to_cache(table, heap);
-	dict_sys->sys_stats = table;
-	mem_heap_empty(heap);
-
-	index = dict_mem_index_create("SYS_STATS", "CLUST_IND",
-				      DICT_HDR_SPACE,
-				      DICT_UNIQUE | DICT_CLUSTERED, 2);
-
-	dict_mem_index_add_field(index, "INDEX_ID", 0);
-	dict_mem_index_add_field(index, "KEY_COLS", 0);
-
-	index->id = DICT_STATS_ID;
-	error = dict_index_add_to_cache(table, index,
-					mtr_read_ulint(dict_hdr
-						       + DICT_HDR_STATS,
-						       MLOG_4BYTES, &mtr),
-					FALSE);
-	ut_a(error == DB_SUCCESS);
+	dict_add_to_cache_xtradb_sys_stats(TRUE, heap, dict_hdr, &mtr);
 
 	mem_heap_free(heap);
 

storage/xtradb/dict/dict0dict.c

@@ -4622,12 +4622,6 @@ next_rec:
 	}
 	btr_pcur_close(&pcur);
 	mtr_commit(&mtr);
-
-	if (rests) {
-		fprintf(stderr, "InnoDB: Warning: failed to store %lu stats entries"
-				" of %s/%s to SYS_STATS system table.\n",
-				rests, index->table_name, index->name);
-	}
 }
 /*===========================================*/
 
@@ -5394,6 +5388,28 @@ dict_table_replace_index_in_foreign_list(
 			foreign->foreign_index = new_index;
 		}
 	}
+
+
+	for (foreign = UT_LIST_GET_FIRST(table->referenced_list);
+	     foreign;
+	     foreign = UT_LIST_GET_NEXT(referenced_list, foreign)) {
+
+		dict_index_t*	new_index;
+
+		if (foreign->referenced_index == index) {
+			ut_ad(foreign->referenced_table == index->table);
+
+			new_index = dict_foreign_find_index(
+				foreign->referenced_table,
+				foreign->referenced_col_names,
+				foreign->n_fields, index,
+				/*check_charsets=*/TRUE, /*check_null=*/FALSE);
+			ut_ad(new_index || !trx->check_foreigns);
+			ut_ad(!new_index || new_index->table == index->table);
+
+			foreign->referenced_index = new_index;
+		}
+	}
 }
 
 /**********************************************************************//**

storage/xtradb/dict/dict0load.c

@@ -165,7 +165,7 @@ dict_print(void)
 	monitor printout */
 
 	mutex_enter(&kernel_mutex);
-	srv_fatal_semaphore_wait_threshold += 7200; /* 2 hours */
+	srv_fatal_semaphore_wait_threshold += SRV_SEMAPHORE_WAIT_EXTENSION;
 	mutex_exit(&kernel_mutex);
 
 	mutex_enter(&(dict_sys->mutex));
@@ -193,7 +193,7 @@ loop:
 		/* Restore the fatal semaphore wait timeout */
 
 		mutex_enter(&kernel_mutex);
-		srv_fatal_semaphore_wait_threshold -= 7200; /* 2 hours */
+		srv_fatal_semaphore_wait_threshold -= SRV_SEMAPHORE_WAIT_EXTENSION;
 		mutex_exit(&kernel_mutex);
 
 		return;

storage/xtradb/fil/fil0fil.c

@@ -1876,7 +1876,7 @@ fil_inc_pending_ops(
 
 	if (space == NULL) {
 		fprintf(stderr,
-			"InnoDB: Error: trying to do ibuf merge to a"
+			"InnoDB: Error: trying to do an operation on a"
 			" dropped tablespace %lu\n",
 			(ulong) id);
 	}
@@ -3375,6 +3375,7 @@ skip_info:
 			for (offset = 0; offset < free_limit_bytes;
 			     offset += zip_size ? zip_size : UNIV_PAGE_SIZE) {
 				ibool		page_is_corrupt;
+				ibool		is_descr_page = FALSE;
 
 				success = os_file_read(file, page,
 							(ulint)(offset & 0xFFFFFFFFUL),
@@ -3413,6 +3414,7 @@ skip_info:
 
 					/* store as descr page */
 					memcpy(descr_page, page, (zip_size ? zip_size : UNIV_PAGE_SIZE));
+					is_descr_page = TRUE;
 
 				} else if (descr_is_corrupt) {
 					/* unknown state of the page */
@@ -3489,7 +3491,8 @@ skip_info:
 						}
 					}
 
-					if (fil_page_get_type(page) == FIL_PAGE_INDEX) {
+					if (fil_page_get_type(page) ==
+					    FIL_PAGE_INDEX && !is_descr_page) {
 						dulint tmp = mach_read_from_8(page + (PAGE_HEADER + PAGE_INDEX_ID));
 
 						for (i = 0; i < n_index; i++) {

storage/xtradb/handler/ha_innodb.cc

@@ -195,6 +195,9 @@ static my_bool	innobase_rollback_on_timeout		= FALSE;
 static my_bool	innobase_create_status_file		= FALSE;
 static my_bool	innobase_stats_on_metadata		= TRUE;
 static my_bool	innobase_use_sys_stats_table		= FALSE;
+#ifdef UNIV_DEBUG
+static ulong    innobase_sys_stats_root_page		= 0;
+#endif
 static my_bool	innobase_buffer_pool_shm_checksum	= TRUE;
 static uint	innobase_buffer_pool_shm_key		= 0;
 
@@ -939,11 +942,23 @@ convert_error_code_to_mysql(
 	case DB_TABLE_NOT_FOUND:
 		return(HA_ERR_NO_SUCH_TABLE);
 
-	case DB_TOO_BIG_RECORD:
-		my_error(ER_TOO_BIG_ROWSIZE, MYF(0),
-			 page_get_free_space_of_empty(flags
-						      & DICT_TF_COMPACT) / 2);
+	case DB_TOO_BIG_RECORD: {
+		/* If prefix is true then a 768-byte prefix is stored
+		locally for BLOB fields. Refer to dict_table_get_format() */
+		bool prefix = ((flags & DICT_TF_FORMAT_MASK)
+		 	       >> DICT_TF_FORMAT_SHIFT) < UNIV_FORMAT_B;
+		my_printf_error(ER_TOO_BIG_ROWSIZE,
+			"Row size too large (> %lu). Changing some columns "
+			"to TEXT or BLOB %smay help. In current row "
+			"format, BLOB prefix of %d bytes is stored inline.",
+			MYF(0),
+			page_get_free_space_of_empty(flags &
+				DICT_TF_COMPACT) / 2,
+			prefix ? "or using ROW_FORMAT=DYNAMIC "
+			"or ROW_FORMAT=COMPRESSED ": "",
+			prefix ? DICT_MAX_INDEX_COL_LEN : 0);
 		return(HA_ERR_TO_BIG_ROW);
+	}
 
 	case DB_NO_SAVEPOINT:
 		return(HA_ERR_NO_SAVEPOINT);
@@ -2369,6 +2384,10 @@ mem_free_and_error:
 
 	srv_use_sys_stats_table = (ibool) innobase_use_sys_stats_table;
 
+#ifdef UNIV_DEBUG
+	srv_sys_stats_root_page = innobase_sys_stats_root_page;
+#endif
+
 	/* -------------- Log files ---------------------------*/
 
 	/* The default dir for log files is the datadir of MySQL */
@@ -4267,6 +4286,27 @@ table_opened:
 	DBUG_RETURN(0);
 }
 
+UNIV_INTERN
+handler*
+ha_innobase::clone(
+/*===============*/
+	const char*	name,		/*!< in: table name */
+	MEM_ROOT*	mem_root)	/*!< in: memory context */
+{
+	ha_innobase* new_handler;
+
+	DBUG_ENTER("ha_innobase::clone");
+
+	new_handler = static_cast<ha_innobase*>(handler::clone(name,
+							       mem_root));
+	if (new_handler) {
+		new_handler->prebuilt->select_lock_type
+			= prebuilt->select_lock_type;
+	}
+
+	DBUG_RETURN(new_handler);
+}
+
 UNIV_INTERN
 uint
 ha_innobase::max_supported_key_part_length() const
@@ -8684,7 +8724,7 @@ ha_innobase::check(
 
 	/* Enlarge the fatal lock wait timeout during CHECK TABLE. */
 	mutex_enter(&kernel_mutex);
-	srv_fatal_semaphore_wait_threshold += 7200; /* 2 hours */
+	srv_fatal_semaphore_wait_threshold += SRV_SEMAPHORE_WAIT_EXTENSION;
 	mutex_exit(&kernel_mutex);
 
 	for (index = dict_table_get_first_index(prebuilt->table);
@@ -8780,7 +8820,7 @@ ha_innobase::check(
 
 	/* Restore the fatal lock wait timeout after CHECK TABLE. */
 	mutex_enter(&kernel_mutex);
-	srv_fatal_semaphore_wait_threshold -= 7200; /* 2 hours */
+	srv_fatal_semaphore_wait_threshold -= SRV_SEMAPHORE_WAIT_EXTENSION;
 	mutex_exit(&kernel_mutex);
 
 	prebuilt->trx->op_info = "";
@@ -11751,6 +11791,13 @@ static MYSQL_SYSVAR_BOOL(use_sys_stats_table, innobase_use_sys_stats_table,
   "So you should use ANALYZE TABLE command intentionally.",
   NULL, NULL, FALSE);
 
+#ifdef UNIV_DEBUG
+static MYSQL_SYSVAR_ULONG(persistent_stats_root_page,
+  innobase_sys_stats_root_page, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY,
+  "Override the SYS_STATS root page id, 0 = no override (for testing only)",
+  NULL, NULL, 0, 0, ULONG_MAX, 0);
+#endif
+
 static MYSQL_SYSVAR_BOOL(adaptive_hash_index, btr_search_enabled,
   PLUGIN_VAR_OPCMDARG,
   "Enable InnoDB adaptive hash index (enabled by default).  "
@@ -11934,6 +11981,18 @@ static MYSQL_SYSVAR_ENUM(stats_method, srv_innodb_stats_method,
   "NULLS_UNEQUAL and NULLS_IGNORED",
    NULL, NULL, SRV_STATS_NULLS_EQUAL, &innodb_stats_method_typelib);
 
+static MYSQL_SYSVAR_BOOL(track_changed_pages, srv_track_changed_pages,
+    PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY,
+    "Track the redo log for changed pages and output a changed page bitmap",
+    NULL, NULL, FALSE);
+
+static MYSQL_SYSVAR_ULONGLONG(changed_pages_limit, srv_changed_pages_limit,
+  PLUGIN_VAR_RQCMDARG,
+  "The maximum number of rows for "
+  "INFORMATION_SCHEMA.INNODB_CHANGED_PAGES table, "
+  "0 - unlimited",
+  NULL, NULL, 1000000, 0, ~0ULL, 0);
+
 #if defined UNIV_DEBUG || defined UNIV_IBUF_DEBUG
 static MYSQL_SYSVAR_UINT(change_buffering_debug, ibuf_debug,
   PLUGIN_VAR_RQCMDARG,
@@ -12070,7 +12129,7 @@ static MYSQL_SYSVAR_UINT(auto_lru_dump, srv_auto_lru_dump,
   NULL, NULL, 0, 0, UINT_MAX32, 0);
 
 static MYSQL_SYSVAR_BOOL(blocking_lru_restore, innobase_blocking_lru_restore,
-  PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY,
+  PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_READONLY,
   "Block XtraDB startup process until buffer pool is full restored from a "
   "dump file (if present). Disabled by default.",
   NULL, NULL, FALSE);
@@ -12149,6 +12208,9 @@ static struct st_mysql_sys_var* innobase_system_variables[]= {
   MYSQL_SYSVAR(stats_auto_update),
   MYSQL_SYSVAR(stats_update_need_lock),
   MYSQL_SYSVAR(use_sys_stats_table),
+#ifdef UNIV_DEBUG
+  MYSQL_SYSVAR(persistent_stats_root_page),
+#endif
   MYSQL_SYSVAR(stats_sample_pages),
   MYSQL_SYSVAR(adaptive_hash_index),
   MYSQL_SYSVAR(stats_method),
@@ -12180,6 +12242,8 @@ static struct st_mysql_sys_var* innobase_system_variables[]= {
   MYSQL_SYSVAR(dict_size_limit),
   MYSQL_SYSVAR(use_sys_malloc),
   MYSQL_SYSVAR(change_buffering),
+  MYSQL_SYSVAR(track_changed_pages),
+  MYSQL_SYSVAR(changed_pages_limit),
 #if defined UNIV_DEBUG || defined UNIV_IBUF_DEBUG
   MYSQL_SYSVAR(change_buffering_debug),
 #endif /* UNIV_DEBUG || UNIV_IBUF_DEBUG */
@@ -12230,7 +12294,10 @@ i_s_innodb_admin_command,
 i_s_innodb_sys_tables,
 i_s_innodb_sys_indexes,
 i_s_innodb_sys_stats,
-i_s_innodb_patches
+i_s_innodb_changed_pages,
+i_s_innodb_buffer_page,
+i_s_innodb_buffer_page_lru,
+i_s_innodb_buffer_stats
 mysql_declare_plugin_end;
 
 /** @brief Initialize the default value of innodb_commit_concurrency.

storage/xtradb/handler/ha_innodb.h

@@ -133,6 +133,7 @@ class ha_innobase: public handler
 	const key_map* keys_to_use_for_scanning();
 
 	int open(const char *name, int mode, uint test_if_locked);
+	handler* clone(const char *name, MEM_ROOT *mem_root);
 	int close(void);
 	double scan_time();
 	double read_time(uint index, uint ranges, ha_rows rows);

storage/xtradb/handler/handler0alter.cc

@@ -668,6 +668,10 @@ ha_innobase::add_index(
 		DBUG_RETURN(HA_ERR_NO_SUCH_TABLE);
 	}
 
+	if (innodb_table->tablespace_discarded) {
+		DBUG_RETURN(-1);
+	}
+
 	/* Check that index keys are sensible */
 	error = innobase_check_index_keys(key_info, num_of_keys, innodb_table);
 
@@ -823,6 +827,8 @@ ha_innobase::add_index(
 					innodb_table, indexed_table,
 					index, num_of_idx, table);
 
+	DBUG_EXECUTE_IF("crash_innodb_add_index_after", DBUG_SUICIDE(););
+
 error_handling:
 	/* After an error, remove all those index definitions from the
 	dictionary which were defined. */

storage/xtradb/handler/i_s.h

@@ -36,7 +36,6 @@ extern struct st_mysql_plugin	i_s_innodb_cmp;
 extern struct st_mysql_plugin	i_s_innodb_cmp_reset;
 extern struct st_mysql_plugin	i_s_innodb_cmpmem;
 extern struct st_mysql_plugin	i_s_innodb_cmpmem_reset;
-extern struct st_mysql_plugin	i_s_innodb_patches;
 extern struct st_mysql_plugin	i_s_innodb_rseg;
 extern struct st_mysql_plugin	i_s_innodb_table_stats;
 extern struct st_mysql_plugin	i_s_innodb_index_stats;
@@ -44,5 +43,9 @@ extern struct st_mysql_plugin	i_s_innodb_admin_command;
 extern struct st_mysql_plugin   i_s_innodb_sys_tables;
 extern struct st_mysql_plugin   i_s_innodb_sys_indexes;
 extern struct st_mysql_plugin	i_s_innodb_sys_stats;
+extern struct st_mysql_plugin	i_s_innodb_changed_pages;
+extern struct st_mysql_plugin	i_s_innodb_buffer_page;
+extern struct st_mysql_plugin	i_s_innodb_buffer_page_lru;
+extern struct st_mysql_plugin	i_s_innodb_buffer_stats;
 
 #endif /* i_s_h */

storage/xtradb/handler/innodb_patch_info.h

@@ -1,51 +0,0 @@
-/* Copyright (C) 2002-2006 MySQL AB
-  
-   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 Foundation; version 2 of the License.
-  
-   This program is distributed in the hope that it will be useful,
-   but WITHOUT 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 */
-
-#ifdef USE_PRAGMA_INTERFACE
-#pragma interface                      /* gcc class implementation */
-#endif
-
-struct innodb_enhancement {
-       const char *file;
-       const char *name;
-       const char *comment;
-       const char *link;
-}innodb_enhancements[] = {
-{"xtradb_show_enhancements","I_S.XTRADB_ENHANCEMENTS","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_show_status","Improvements to SHOW INNODB STATUS","Memory information and lock info fixes","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_io","Improvements to InnoDB IO","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_opt_lru_count","Fix of buffer_pool mutex","Decreases contention on buffer_pool mutex on LRU operations","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_buffer_pool_pages","Information of buffer pool content","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_expand_undo_slots","expandable maximum number of undo slots","from 1024 (default) to about 4000","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_extra_rseg","allow to create extra rollback segments","When create new db, the new parameter allows to create more rollback segments","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_overwrite_relay_log_info","overwrite relay-log.info when slave recovery","Building as plugin, it is not used.","http://www.percona.com/docs/wiki/percona-xtradb:innodb_overwrite_relay_log_info"},
-{"innodb_thread_concurrency_timer_based","use InnoDB timer based concurrency throttling (backport from MySQL 5.4.0)","",""},
-{"innodb_expand_import","convert .ibd file automatically when import tablespace","the files are generated by xtrabackup export mode.","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_dict_size_limit","Limit dictionary cache size","Variable innodb_dict_size_limit in bytes","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_split_buf_pool_mutex","More fix of buffer_pool mutex","Spliting buf_pool_mutex and optimizing based on innodb_opt_lru_count","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_stats","Additional features about InnoDB statistics/optimizer","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_recovery_patches","Bugfixes and adjustments about recovery process","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_purge_thread","Enable to use purge devoted thread","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_admin_command_base","XtraDB specific command interface through i_s","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_show_lock_name","Show mutex/lock name instead of crated file/line","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_extend_slow","Extended statistics in slow.log","It is InnoDB-part only. It needs to patch also to mysqld.","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_lru_dump_restore","Dump and restore command for content of buffer pool","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_separate_doublewrite","Add option 'innodb_doublewrite_file' to separate doublewrite dedicated tablespace","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_pass_corrupt_table","Treat tables as corrupt instead of crash, when meet corrupt blocks","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_fast_checksum","Using the checksum on 32bit-unit calculation","incompatible for unpatched ver.","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_files_extend","allow >4GB transaction log files, and can vary universal page size of datafiles","incompatible for unpatched ver.","http://www.percona.com/docs/wiki/percona-xtradb"},
-{"innodb_sys_tables_sys_indexes","Expose InnoDB SYS_TABLES and SYS_INDEXES schema tables","","http://www.percona.com/docs/wiki/percona-xtradb"},
-{NULL, NULL, NULL, NULL}
-};

storage/xtradb/ibuf/ibuf0ibuf.c

@@ -2759,11 +2759,19 @@ ibuf_insert_low(
 
 		root = ibuf_tree_root_get(&mtr);
 
-		err = btr_cur_pessimistic_insert(BTR_NO_LOCKING_FLAG
-						 | BTR_NO_UNDO_LOG_FLAG,
-						 cursor,
-						 ibuf_entry, &ins_rec,
-						 &dummy_big_rec, 0, thr, &mtr);
+		err = btr_cur_optimistic_insert(
+			BTR_NO_LOCKING_FLAG | BTR_NO_UNDO_LOG_FLAG,
+			cursor, ibuf_entry, &ins_rec,
+			&dummy_big_rec, 0, thr, &mtr);
+
+		if (err == DB_FAIL) {
+			err = btr_cur_pessimistic_insert(
+				BTR_NO_LOCKING_FLAG
+				| BTR_NO_UNDO_LOG_FLAG,
+				cursor, ibuf_entry, &ins_rec,
+				&dummy_big_rec, 0, thr, &mtr);
+		}
+
 		if (err == DB_SUCCESS) {
 			/* Update the page max trx id field */
 			page_update_max_trx_id(btr_cur_get_block(cursor), NULL,

storage/xtradb/include/buf0buf.h

@@ -103,6 +103,81 @@ enum buf_page_state {
 					before putting to the free list */
 };
 
+/** This structure defines information we will fetch from each buffer pool. It
+will be used to print table IO stats */
+struct buf_pool_info_struct{
+	/* General buffer pool info */
+	ulint	pool_size;		/*!< Buffer Pool size in pages */
+	ulint	lru_len;		/*!< Length of buf_pool->LRU */
+	ulint	old_lru_len;		/*!< buf_pool->LRU_old_len */
+	ulint	free_list_len;		/*!< Length of buf_pool->free list */
+	ulint	flush_list_len;		/*!< Length of buf_pool->flush_list */
+	ulint	n_pend_unzip;		/*!< buf_pool->n_pend_unzip, pages
+					pending decompress */
+	ulint	n_pend_reads;		/*!< buf_pool->n_pend_reads, pages
+					pending read */
+	ulint	n_pending_flush_lru;	/*!< Pages pending flush in LRU */
+	ulint	n_pending_flush_single_page;/*!< Pages pending to be
+					flushed as part of single page
+					flushes issued by various user
+					threads */
+	ulint	n_pending_flush_list;	/*!< Pages pending flush in FLUSH
+					LIST */
+	ulint	n_pages_made_young;	/*!< number of pages made young */
+	ulint	n_pages_not_made_young;	/*!< number of pages not made young */
+	ulint	n_pages_read;		/*!< buf_pool->n_pages_read */
+	ulint	n_pages_created;	/*!< buf_pool->n_pages_created */
+	ulint	n_pages_written;	/*!< buf_pool->n_pages_written */
+	ulint	n_page_gets;		/*!< buf_pool->n_page_gets */
+	ulint	n_ra_pages_read_rnd;	/*!< buf_pool->n_ra_pages_read_rnd,
+					number of pages readahead */
+	ulint	n_ra_pages_read;	/*!< buf_pool->n_ra_pages_read, number
+					of pages readahead */
+	ulint	n_ra_pages_evicted;	/*!< buf_pool->n_ra_pages_evicted,
+					number of readahead pages evicted
+					without access */
+	ulint	n_page_get_delta;	/*!< num of buffer pool page gets since
+					last printout */
+
+	/* Buffer pool access stats */
+	double	page_made_young_rate;	/*!< page made young rate in pages
+					per second */
+	double	page_not_made_young_rate;/*!< page not made young rate
+					in pages per second */
+	double	pages_read_rate;	/*!< num of pages read per second */
+	double	pages_created_rate;	/*!< num of pages create per second */
+	double	pages_written_rate;	/*!< num of  pages written per second */
+	ulint	page_read_delta;	/*!< num of pages read since last
+					printout */
+	ulint	young_making_delta;	/*!< num of pages made young since
+					last printout */
+	ulint	not_young_making_delta;	/*!< num of pages not make young since
+					last printout */
+
+	/* Statistics about read ahead algorithm.  */
+	double	pages_readahead_rnd_rate;/*!< random readahead rate in pages per
+					second */
+	double	pages_readahead_rate;	/*!< readahead rate in pages per
+					second */
+	double	pages_evicted_rate;	/*!< rate of readahead page evicted
+					without access, in pages per second */
+
+	/* Stats about LRU eviction */
+	ulint	unzip_lru_len;		/*!< length of buf_pool->unzip_LRU
+					list */
+	/* Counters for LRU policy */
+	ulint	io_sum;			/*!< buf_LRU_stat_sum.io */
+	ulint	io_cur;			/*!< buf_LRU_stat_cur.io, num of IO
+					for current interval */
+	ulint	unzip_sum;		/*!< buf_LRU_stat_sum.unzip */
+	ulint	unzip_cur;		/*!< buf_LRU_stat_cur.unzip, num
+					pages decompressed in current
+					interval */
+};
+
+typedef struct buf_pool_info_struct	buf_pool_info_t;
+
+
 #ifndef UNIV_HOTBACKUP
 /********************************************************************//**
 Creates the buffer pool.
@@ -623,6 +698,16 @@ void
 buf_print_io(
 /*=========*/
 	FILE*	file);	/*!< in: file where to print */
+/*******************************************************************//**
+Collect buffer pool stats information for a buffer pool. Also
+record aggregated stats if there are more than one buffer pool
+in the server */
+UNIV_INTERN
+void
+buf_stats_get_pool_info(
+/*====================*/
+	buf_pool_info_t*        pool_info); /*!< in/out: buffer pool info
+                                                to fill */
 /*********************************************************************//**
 Returns the ratio in percents of modified pages in the buffer pool /
 database pages in the buffer pool.
@@ -1051,12 +1136,27 @@ UNIV_INTERN
 ulint
 buf_get_free_list_len(void);
 /*=======================*/
+
+/*********************************************************************//**
+Get the nth chunk's buffer block in the specified buffer pool.
+@return the nth chunk's buffer block. */
+UNIV_INLINE
+buf_block_t*
+buf_get_nth_chunk_block(
+/*====================*/
+	const buf_pool_t* buf_pool,     /*!< in: buffer pool instance */
+	ulint           n,              /*!< in: nth chunk in the buffer pool */
+	ulint*          chunk_size);    /*!< in: chunk size */
+
 #endif /* !UNIV_HOTBACKUP */
 
 
 /** The common buffer control block structure
 for compressed and uncompressed frames */
 
+/** Number of bits used for buffer page states. */
+#define BUF_PAGE_STATE_BITS     3
+
 struct buf_page_struct{
 	/** @name General fields
 	None of these bit-fields must be modified without holding
@@ -1071,7 +1171,8 @@ struct buf_page_struct{
 	unsigned	offset:32;	/*!< page number; also protected
 					by buf_pool_mutex. */
 
-	unsigned	state:3;	/*!< state of the control block; also
+	unsigned	state:BUF_PAGE_STATE_BITS;
+					/*!< state of the control block; also
 					protected by buf_pool_mutex.
 					State transitions from
 					BUF_BLOCK_READY_FOR_USE to

storage/xtradb/include/buf0buf.ic

@@ -36,6 +36,7 @@ Created 11/5/1995 Heikki Tuuri
 #include "buf0lru.h"
 #include "buf0rea.h"
 #include "srv0srv.h"
+
 /********************************************************************//**
 Reads the freed_page_clock of a buffer block.
 @return	freed_page_clock */
@@ -1154,4 +1155,23 @@ buf_block_dbg_add_level(
 	sync_thread_add_level(&block->lock, level, FALSE);
 }
 #endif /* UNIV_SYNC_DEBUG */
+
+/*********************************************************************//**
+Get the nth chunk's buffer block in the specified buffer pool.
+@return the nth chunk's buffer block. */
+UNIV_INLINE
+buf_block_t*
+buf_get_nth_chunk_block(
+/*====================*/
+        const buf_pool_t* buf_pool,     /*!< in: buffer pool instance */
+        ulint           n,              /*!< in: nth chunk in the buffer pool */
+        ulint*          chunk_size)     /*!< in: chunk size */
+{
+        const buf_chunk_t*      chunk;
+
+        chunk = buf_pool->chunks + n;
+        *chunk_size = chunk->size;
+        return(chunk->blocks);
+}
 #endif /* !UNIV_HOTBACKUP */
+

storage/xtradb/include/buf0lru.h

@@ -93,13 +93,12 @@ buf_LRU_insert_zip_clean(
 Try to free a block.  If bpage is a descriptor of a compressed-only
 page, the descriptor object will be freed as well.
 
-NOTE: If this function returns TRUE, it will temporarily
-release buf_pool_mutex.  Furthermore, the page frame will no longer be
-accessible via bpage.
+NOTE: This will temporarily release buf_pool_mutex.  Furthermore, the
+page frame will no longer be accessible via bpage.
 
-The caller must hold buf_pool_mutex and buf_page_get_mutex(bpage) and
-release these two mutexes after the call.  No other
-buf_page_get_mutex() may be held when calling this function.
+The caller must hold buf_page_get_mutex(bpage) and release this mutex
+after the call.  No other buf_page_get_mutex() may be held when
+calling this function.
 @return TRUE if freed, FALSE otherwise. */
 UNIV_INTERN
 ibool

storage/xtradb/include/dict0boot.h

@@ -91,6 +91,26 @@ void
 dict_create(void);
 /*=============*/
 
+/*****************************************************************//**
+Verifies the SYS_STATS table by scanning its clustered index.  This
+function may only be called at InnoDB startup time.
+
+@return	TRUE if SYS_STATS was verified successfully */
+UNIV_INTERN
+ibool
+dict_verify_xtradb_sys_stats(void);
+/*==============================*/
+
+/*****************************************************************//**
+Discard the existing dictionary cache SYS_STATS information, create and
+add it there anew.  Does not touch the old SYS_STATS tablespace page
+under the assumption that they are corrupted or overwritten for other
+purposes. */
+UNIV_INTERN
+void
+dict_recreate_xtradb_sys_stats(void);
+/*================================*/
+
 
 /* Space id and page no where the dictionary header resides */
 #define	DICT_HDR_SPACE		0	/* the SYSTEM tablespace */

storage/xtradb/include/fil0fil.h

@@ -142,6 +142,8 @@ extern fil_addr_t	fil_addr_null;
 #define FIL_PAGE_TYPE_BLOB	10	/*!< Uncompressed BLOB page */
 #define FIL_PAGE_TYPE_ZBLOB	11	/*!< First compressed BLOB page */
 #define FIL_PAGE_TYPE_ZBLOB2	12	/*!< Subsequent compressed BLOB page */
+#define FIL_PAGE_TYPE_LAST	FIL_PAGE_TYPE_ZBLOB2
+					/*!< Last page type */
 /* @} */
 
 /** Space types @{ */

storage/xtradb/include/log0log.h

@@ -41,6 +41,9 @@ Created 12/9/1995 Heikki Tuuri
 #include "sync0rw.h"
 #endif /* !UNIV_HOTBACKUP */
 
+/* Type used for all log sequence number storage and arithmetics */
+typedef ib_uint64_t		lsn_t;
+
 /** Redo log buffer */
 typedef struct log_struct	log_t;
 /** Redo log group */
@@ -953,6 +956,11 @@ struct log_struct{
 					become signaled */
 	/* @} */
 #endif /* UNIV_LOG_ARCHIVE */
+	ib_uint64_t	tracked_lsn;	/*!< log tracking has advanced to this
+					lsn.  Field accessed atomically where
+					64-bit atomic ops are supported,
+					protected by the log sys mutex
+					otherwise. */
 };
 
 #ifdef UNIV_LOG_ARCHIVE

storage/xtradb/include/log0online.h

@@ -0,0 +1,111 @@
+/*****************************************************************************
+
+Copyright (c) 2011-2012, Percona Inc. 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
+Foundation; version 2 of the License.
+
+This program is distributed in the hope that it will be useful, but WITHOUT
+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
+
+*****************************************************************************/
+
+/**************************************************//**
+@file include/log0online.h
+Online database log parsing for changed page tracking
+*******************************************************/
+
+#ifndef log0online_h
+#define log0online_h
+
+#include "univ.i"
+#include "os0file.h"
+
+/*********************************************************************//**
+Initializes the online log following subsytem. */
+UNIV_INTERN
+void
+log_online_read_init();
+/*===================*/
+
+/*********************************************************************//**
+Shuts down the online log following subsystem. */
+UNIV_INTERN
+void
+log_online_read_shutdown();
+/*=======================*/
+
+/*********************************************************************//**
+Reads and parses the redo log up to last checkpoint LSN to build the changed
+page bitmap which is then written to disk.  */
+UNIV_INTERN
+void
+log_online_follow_redo_log();
+/*=========================*/
+
+/** The iterator through all bits of changed pages bitmap blocks */
+struct log_bitmap_iterator_struct
+{
+	char		in_name[FN_REFLEN]; /*!< the file name for bitmap
+					    input */
+	os_file_t	in;                 /*!< the bitmap input file */
+	ib_uint64_t	in_offset;          /*!< the next write position in the
+					    bitmap output file */
+	ib_uint32_t	bit_offset;         /*!< bit offset inside of bitmap
+					    block*/
+	ib_uint64_t	start_lsn;          /*!< Start lsn of the block */
+	ib_uint64_t	end_lsn;            /*!< End lsn of the block */
+	ib_uint32_t	space_id;           /*!< Block space id */
+	ib_uint32_t	first_page_id;      /*!< First block page id */
+	ibool		changed;            /*!< true if current page was changed */
+	byte*		page;               /*!< Bitmap block */
+};
+
+typedef struct log_bitmap_iterator_struct log_bitmap_iterator_t;
+
+#define LOG_BITMAP_ITERATOR_START_LSN(i) \
+        ((i).start_lsn)
+#define LOG_BITMAP_ITERATOR_END_LSN(i) \
+        ((i).end_lsn)
+#define LOG_BITMAP_ITERATOR_SPACE_ID(i) \
+        ((i).space_id)
+#define LOG_BITMAP_ITERATOR_PAGE_NUM(i) \
+        ((i).first_page_id + (i).bit_offset)
+#define LOG_BITMAP_ITERATOR_PAGE_CHANGED(i) \
+        ((i).changed)
+
+/*********************************************************************//**
+Initializes log bitmap iterator.
+@return TRUE if the iterator is initialized OK, FALSE otherwise. */
+UNIV_INTERN
+ibool
+log_online_bitmap_iterator_init(
+/*============================*/
+	log_bitmap_iterator_t *i); /*!<in/out:  iterator */
+
+/*********************************************************************//**
+Releases log bitmap iterator. */
+UNIV_INTERN
+void
+log_online_bitmap_iterator_release(
+/*===============================*/
+	log_bitmap_iterator_t *i); /*!<in/out:  iterator */
+
+/*********************************************************************//**
+Iterates through bits of saved bitmap blocks.
+Sequentially reads blocks from bitmap file(s) and interates through
+their bits. Ignores blocks with wrong checksum.
+@return TRUE if iteration is successful, FALSE if all bits are iterated. */
+UNIV_INTERN
+ibool
+log_online_bitmap_iterator_next(
+/*============================*/
+	log_bitmap_iterator_t *i); /*!<in/out: iterator */
+
+#endif

storage/xtradb/include/log0recv.h

@@ -32,6 +32,28 @@ Created 9/20/1997 Heikki Tuuri
 #include "hash0hash.h"
 #include "log0log.h"
 
+/******************************************************//**
+Checks the 4-byte checksum to the trailer checksum field of a log
+block.  We also accept a log block in the old format before
+InnoDB-3.23.52 where the checksum field contains the log block number.
+@return TRUE if ok, or if the log block may be in the format of InnoDB
+version predating 3.23.52 */
+UNIV_INTERN
+ibool
+log_block_checksum_is_ok_or_old_format(
+/*===================================*/
+	const byte*	block);	/*!< in: pointer to a log block */
+
+/*******************************************************//**
+Calculates the new value for lsn when more data is added to the log. */
+UNIV_INTERN
+ib_uint64_t
+recv_calc_lsn_on_data_add(
+/*======================*/
+	ib_uint64_t	lsn,	/*!< in: old lsn */
+	ib_uint64_t	len);	/*!< in: this many bytes of data is
+				added, log block headers not included */
+
 #ifdef UNIV_HOTBACKUP
 extern ibool	recv_replay_file_ops;
 
@@ -182,6 +204,21 @@ UNIV_INTERN
 void
 recv_recovery_rollback_active(void);
 /*===============================*/
+
+/*******************************************************************//**
+Tries to parse a single log record and returns its length.
+@return	length of the record, or 0 if the record was not complete */
+UNIV_INTERN
+ulint
+recv_parse_log_rec(
+/*===============*/
+	byte*	ptr,	/*!< in: pointer to a buffer */
+	byte*	end_ptr,/*!< in: pointer to the buffer end */
+	byte*	type,	/*!< out: type */
+	ulint*	space,	/*!< out: space id */
+	ulint*	page_no,/*!< out: page number */
+	byte**	body);	/*!< out: log record body start */
+
 /*******************************************************//**
 Scans log from a buffer and stores new log data to the parsing buffer.
 Parses and hashes the log records if new data found.  Unless

storage/xtradb/include/os0file.h

@@ -463,6 +463,14 @@ os_file_set_eof(
 /*============*/
 	FILE*		file);	/*!< in: file to be truncated */
 /***********************************************************************//**
+Truncates a file at the specified position.
+@return TRUE if success */
+UNIV_INTERN
+ibool
+os_file_set_eof_at(
+	os_file_t	file, /*!< in: handle to a file */
+	ib_uint64_t	new_len);/*!< in: new file length */
+/***********************************************************************//**
 Flushes the write buffers of a given file to the disk.
 @return	TRUE if success */
 UNIV_INTERN

storage/xtradb/include/os0sync.h

@@ -287,7 +287,11 @@ Atomic compare-and-swap and increment for InnoDB. */
 
 #if defined(HAVE_IB_GCC_ATOMIC_BUILTINS)
 
-#define HAVE_ATOMIC_BUILTINS
+# define HAVE_ATOMIC_BUILTINS
+
+# ifdef HAVE_IB_GCC_ATOMIC_BUILTINS_64
+#  define HAVE_ATOMIC_BUILTINS_64
+# endif
 
 /**********************************************************//**
 Returns true if swapped, ptr is pointer to target, old_val is value to
@@ -326,6 +330,9 @@ amount of increment. */
 # define os_atomic_increment_ulint(ptr, amount) \
 	os_atomic_increment(ptr, amount)
 
+# define os_atomic_increment_uint64(ptr, amount) \
+	os_atomic_increment(ptr, amount)
+
 /**********************************************************//**
 Returns the old value of *ptr, atomically sets *ptr to new_val */
 
@@ -334,12 +341,13 @@ Returns the old value of *ptr, atomically sets *ptr to new_val */
 
 #elif defined(HAVE_IB_SOLARIS_ATOMICS)
 
-#define HAVE_ATOMIC_BUILTINS
+# define HAVE_ATOMIC_BUILTINS
+# define HAVE_ATOMIC_BUILTINS_64
 
 /* If not compiling with GCC or GCC doesn't support the atomic
 intrinsics and running on Solaris >= 10 use Solaris atomics */
 
-#include <atomic.h>
+# include <atomic.h>
 
 /**********************************************************//**
 Returns true if swapped, ptr is pointer to target, old_val is value to
@@ -379,6 +387,9 @@ amount of increment. */
 # define os_atomic_increment_ulint(ptr, amount) \
 	atomic_add_long_nv(ptr, amount)
 
+# define os_atomic_increment_uint64(ptr, amount) \
+	atomic_add_64_nv(ptr, amount)
+
 /**********************************************************//**
 Returns the old value of *ptr, atomically sets *ptr to new_val */
 
@@ -387,7 +398,11 @@ Returns the old value of *ptr, atomically sets *ptr to new_val */
 
 #elif defined(HAVE_WINDOWS_ATOMICS)
 
-#define HAVE_ATOMIC_BUILTINS
+# define HAVE_ATOMIC_BUILTINS
+
+# ifndef _WIN32
+#  define HAVE_ATOMIC_BUILTINS_64
+# endif
 
 /* On Windows, use Windows atomics / interlocked */
 # ifdef _WIN64
@@ -425,6 +440,11 @@ amount of increment. */
 # define os_atomic_increment_ulint(ptr, amount) \
 	((ulint) (win_xchg_and_add(ptr, amount) + amount))
 
+# define os_atomic_increment_uint64(ptr, amount)		\
+	((ib_uint64_t) (InterlockedExchangeAdd64(		\
+				 (ib_int64_t*) ptr,		\
+				 (ib_int64_t) amount) + amount))
+
 /**********************************************************//**
 Returns the old value of *ptr, atomically sets *ptr to new_val.
 InterlockedExchange() operates on LONG, and the LONG will be

storage/xtradb/include/srv0srv.h

@@ -60,6 +60,14 @@ extern os_event_t	srv_lock_timeout_thread_event;
 /* This event is set at shutdown to wakeup threads from sleep */
 extern os_event_t	srv_shutdown_event;
 
+/* This event is set on checkpoint completion to wake the redo log parser
+thread */
+extern os_event_t	srv_checkpoint_completed_event;
+
+/* This event is set on the online redo log following thread exit to signal
+that the (slow) shutdown may proceed */
+extern os_event_t	srv_redo_log_thread_finished_event;
+
 /* If the last data file is auto-extended, we add this many pages to it
 at a time */
 #define SRV_AUTO_EXTEND_INCREMENT	\
@@ -126,6 +134,11 @@ extern ibool	srv_recovery_stats;
 
 extern ulint	srv_use_purge_thread;
 
+extern my_bool	srv_track_changed_pages;
+
+extern
+ulonglong       srv_changed_pages_limit;
+
 extern ibool	srv_auto_extend_last_data_file;
 extern ulint	srv_last_file_size_max;
 extern char**	srv_log_group_home_dirs;
@@ -213,6 +226,9 @@ extern unsigned long long	srv_stats_sample_pages;
 extern ulong	srv_stats_auto_update;
 extern ulint	srv_stats_update_need_lock;
 extern ibool	srv_use_sys_stats_table;
+#ifdef UNIV_DEBUG
+extern ulong	srv_sys_stats_root_page;
+#endif
 
 extern ibool	srv_use_doublewrite_buf;
 extern ibool	srv_use_checksums;
@@ -284,6 +300,7 @@ extern	ibool	srv_print_latch_waits;
 
 extern ulint	srv_activity_count;
 extern ulint	srv_fatal_semaphore_wait_threshold;
+#define SRV_SEMAPHORE_WAIT_EXTENSION	7200
 extern ulint	srv_dml_needed_delay;
 extern long long	srv_kill_idle_transaction;
 
@@ -644,6 +661,15 @@ srv_LRU_dump_restore_thread(
 	void*	arg);	/*!< in: a dummy parameter required by
 			os_thread_create */
 /******************************************************************//**
+A thread which follows the redo log and outputs the changed page bitmap.
+@return a dummy value */
+UNIV_INTERN
+os_thread_ret_t
+srv_redo_log_follow_thread(
+/*=======================*/
+	void*	arg);	/*!< in: a dummy parameter required by
+			os_thread_create */
+/******************************************************************//**
 Outputs to a file the output of the InnoDB Monitor.
 @return FALSE if not all information printed
 due to failure to obtain necessary mutex */

storage/xtradb/include/univ.i

@@ -46,7 +46,7 @@ Created 1/20/1994 Heikki Tuuri
 #define INNODB_VERSION_MAJOR	1
 #define INNODB_VERSION_MINOR	0
 #define INNODB_VERSION_BUGFIX	17
-#define PERCONA_INNODB_VERSION 13.0
+#define PERCONA_INNODB_VERSION 14.1
 
 /* The following is the InnoDB version as shown in
 SELECT plugin_version FROM information_schema.plugins;
@@ -270,6 +270,24 @@ management to ensure correct alignment for doubles etc. */
 			========================
 */
 
+/** There are currently two InnoDB file formats which are used to group
+features with similar restrictions and dependencies. Using an enum allows
+switch statements to give a compiler warning when a new one is introduced. */
+enum innodb_file_formats_enum {
+	/** Antelope File Format: InnoDB/MySQL up to 5.1.
+	This format includes REDUNDANT and COMPACT row formats */
+	UNIV_FORMAT_A		= 0,
+
+	/** Barracuda File Format: Introduced in InnoDB plugin for 5.1:
+	This format includes COMPRESSED and DYNAMIC row formats.  It
+	includes the ability to create secondary indexes from data that
+	is not on the clustered index page and the ability to store more
+	data off the clustered index page. */
+	UNIV_FORMAT_B		= 1
+};
+
+typedef enum innodb_file_formats_enum innodb_file_formats_t;
+
 /* The 2-logarithm of UNIV_PAGE_SIZE: */
 /* #define UNIV_PAGE_SIZE_SHIFT	14 */
 #define UNIV_PAGE_SIZE_SHIFT_MAX	14

storage/xtradb/include/ut0rbt.h

@@ -110,6 +110,10 @@ struct ib_rbt_bound_struct {
 /* Compare a key with the node value (t is tree, k is key, n is node)*/
 #define rbt_compare(t, k, n) (t->compare(k, n->value))
 
+/* Node size. FIXME: name might clash, but currently it does not, so for easier
+maintenance do not rename it for now. */
+#define	SIZEOF_NODE(t)	((sizeof(ib_rbt_node_t) + t->sizeof_value) - 1)
+
 /****************************************************************//**
 Free an instance of  a red black tree */
 UNIV_INTERN
@@ -181,6 +185,18 @@ rbt_add_node(
 	const void*	value);		/*!< in: this value is copied
 					to the node */
 /****************************************************************//**
+Add a new caller-provided node to tree at the specified position.
+The node must have its key fields initialized correctly.
+@return added node */
+UNIV_INTERN
+const ib_rbt_node_t*
+rbt_add_preallocated_node(
+/*======================*/
+	ib_rbt_t*	tree,		/*!< in: rb tree */
+	ib_rbt_bound_t*	parent,		/*!< in: parent */
+	ib_rbt_node_t*	node);		/*!< in: node */
+
+/****************************************************************//**
 Return the left most data node in the tree
 @return	left most node */
 UNIV_INTERN
@@ -267,6 +283,13 @@ rbt_clear(
 /*======*/
 	ib_rbt_t*	tree);		/*!< in: rb tree */
 /****************************************************************//**
+Clear the tree without deleting and freeing its nodes. */
+UNIV_INTERN
+void
+rbt_reset(
+/*======*/
+	 ib_rbt_t*	tree);		/*!< in: rb tree */
+/****************************************************************//**
 Merge the node from dst into src. Return the number of nodes merged.
 @return	no. of recs merged */
 UNIV_INTERN

storage/xtradb/log/log0log.c

@@ -201,6 +201,54 @@ log_buf_pool_get_oldest_modification(void)
 	return(lsn);
 }
 
+/****************************************************************//**
+Safely reads the log_sys->tracked_lsn value.  Uses atomic operations
+if available, otherwise this field is protected with the log system
+mutex.  The writer counterpart function is log_set_tracked_lsn() in
+log0online.c.
+
+@return log_sys->tracked_lsn value. */
+UNIV_INLINE
+ib_uint64_t
+log_get_tracked_lsn()
+{
+#ifdef HAVE_ATOMIC_BUILTINS_64
+	return os_atomic_increment_uint64(&log_sys->tracked_lsn, 0);
+#else
+	ut_ad(mutex_own(&(log_sys->mutex)));
+	return log_sys->tracked_lsn;
+#endif
+}
+
+/****************************************************************//**
+Checks if the log groups have a big enough margin of free space in
+so that a new log entry can be written without overwriting log data
+that is not read by the changed page bitmap thread.
+@return TRUE if there is not enough free space. */
+static
+ibool
+log_check_tracking_margin(
+	ulint	lsn_advance)	/*!< in: an upper limit on how much log data we
+				plan to write.  If zero, the margin will be
+				checked for the already-written log. */
+{
+	ib_uint64_t	tracked_lsn;
+	ulint		tracked_lsn_age;
+
+	if (!srv_track_changed_pages) {
+		return FALSE;
+	}
+
+	ut_ad(mutex_own(&(log_sys->mutex)));
+
+	tracked_lsn = log_get_tracked_lsn();
+	tracked_lsn_age = log_sys->lsn - tracked_lsn;
+
+	/* The overwrite would happen when log_sys->log_group_capacity is
+	exceeded, but we use max_checkpoint_age for an extra safety margin. */
+	return tracked_lsn_age + lsn_advance > log_sys->max_checkpoint_age;
+}
+
 /************************************************************//**
 Opens the log for log_write_low. The log must be closed with log_close and
 released with log_release.
@@ -217,9 +265,7 @@ log_reserve_and_open(
 	ulint	archived_lsn_age;
 	ulint	dummy;
 #endif /* UNIV_LOG_ARCHIVE */
-#ifdef UNIV_DEBUG
 	ulint	count			= 0;
-#endif /* UNIV_DEBUG */
 
 	ut_a(len < log->buf_size / 2);
 loop:
@@ -247,6 +293,19 @@ loop:
 		goto loop;
 	}
 
+	if (log_check_tracking_margin(len_upper_limit) && (++count < 50)) {
+
+		/* This log write would violate the untracked LSN free space
+		margin.  Limit this to 50 retries as there might be situations
+		where we have no choice but to proceed anyway, i.e. if the log
+		is about to be overflown, log tracking or not. */
+		mutex_exit(&(log->mutex));
+
+		os_thread_sleep(10000);
+
+		goto loop;
+	}
+
 #ifdef UNIV_LOG_ARCHIVE
 	if (log->archiving_state != LOG_ARCH_OFF) {
 
@@ -385,6 +444,8 @@ log_close(void)
 	ulint		first_rec_group;
 	ib_uint64_t	oldest_lsn;
 	ib_uint64_t	lsn;
+	ib_uint64_t	tracked_lsn;
+	ulint		tracked_lsn_age;
 	log_t*		log	= log_sys;
 	ib_uint64_t	checkpoint_age;
 
@@ -411,6 +472,19 @@ log_close(void)
 		log->check_flush_or_checkpoint = TRUE;
 	}
 
+	if (srv_track_changed_pages) {
+
+		tracked_lsn = log_get_tracked_lsn();
+		tracked_lsn_age = lsn - tracked_lsn;
+
+		if (tracked_lsn_age >= log->log_group_capacity) {
+
+			fprintf(stderr, " InnoDB: Error: the age of the "
+				"oldest untracked record exceeds the log "
+				"group capacity!\n");
+		}
+	}
+
 	checkpoint_age = lsn - log->last_checkpoint_lsn;
 
 	if (checkpoint_age >= log->log_group_capacity) {
@@ -872,6 +946,8 @@ log_init(void)
 	log_sys->archiving_on = os_event_create(NULL);
 #endif /* UNIV_LOG_ARCHIVE */
 
+	log_sys->tracked_lsn = 0;
+
 	/*----------------------------*/
 
 	log_block_init(log_sys->buf, log_sys->lsn);
@@ -1721,6 +1797,12 @@ log_io_complete_checkpoint(void)
 	}
 
 	mutex_exit(&(log_sys->mutex));
+
+	/* Wake the redo log watching thread to parse the log up to this
+	checkpoint. */
+	if (srv_track_changed_pages) {
+		os_event_set(srv_checkpoint_completed_event);
+	}
 }
 
 /*******************************************************************//**
@@ -3065,6 +3147,15 @@ loop:
 
 	log_checkpoint_margin();
 
+	mutex_enter(&(log_sys->mutex));
+	if (log_check_tracking_margin(0)) {
+
+		mutex_exit(&(log_sys->mutex));
+		os_thread_sleep(10000);
+		goto loop;
+	}
+	mutex_exit(&(log_sys->mutex));
+
 #ifdef UNIV_LOG_ARCHIVE
 	log_archive_margin();
 #endif /* UNIV_LOG_ARCHIVE */
@@ -3093,6 +3184,7 @@ logs_empty_and_mark_files_at_shutdown(void)
 /*=======================================*/
 {
 	ib_uint64_t	lsn;
+	ib_uint64_t	tracked_lsn;
 	ulint		arch_log_no;
 
 	if (srv_print_verbose_log) {
@@ -3198,9 +3290,12 @@ loop:
 
 	mutex_enter(&(log_sys->mutex));
 
+	tracked_lsn = log_get_tracked_lsn();
+
 	lsn = log_sys->lsn;
 
 	if (lsn != log_sys->last_checkpoint_lsn
+	    || (srv_track_changed_pages	&& (tracked_lsn != log_sys->last_checkpoint_lsn))
 #ifdef UNIV_LOG_ARCHIVE
 	    || (srv_log_archive_on
 		&& lsn != log_sys->archived_lsn + LOG_BLOCK_HDR_SIZE)
@@ -3255,6 +3350,11 @@ loop:
 
 	srv_shutdown_state = SRV_SHUTDOWN_LAST_PHASE;
 
+	/* Signal the log following thread to quit */
+	if (srv_track_changed_pages) {
+		os_event_set(srv_checkpoint_completed_event);
+	}
+
 	/* Make some checks that the server really is quiet */
 	ut_a(srv_n_threads_active[SRV_MASTER] == 0);
 	ut_a(buf_all_freed());
@@ -3274,6 +3374,10 @@ loop:
 
 	fil_flush_file_spaces(FIL_TABLESPACE);
 
+	if (srv_track_changed_pages) {
+		os_event_wait(srv_redo_log_thread_finished_event);
+	}
+
 	fil_close_all_files();
 
 	/* Make some checks that the server really is quiet */
@@ -3399,6 +3503,18 @@ log_print(
 		((log_sys->n_log_ios - log_sys->n_log_ios_old)
 		 / time_elapsed));
 
+	if (srv_track_changed_pages) {
+
+		/* The maximum tracked LSN age is equal to the maximum
+		checkpoint age */
+		fprintf(file,
+			"Log tracking enabled\n"
+			"Log tracked up to   %llu\n"
+			"Max tracked LSN age %lu\n",
+			log_get_tracked_lsn(),
+			log_sys->max_checkpoint_age);
+	}
+
 	log_sys->n_log_ios_old = log_sys->n_log_ios;
 	log_sys->last_printout_time = current_time;
 

storage/xtradb/log/log0recv.c

@@ -840,7 +840,7 @@ block.  We also accept a log block in the old format before
 InnoDB-3.23.52 where the checksum field contains the log block number.
 @return TRUE if ok, or if the log block may be in the format of InnoDB
 version predating 3.23.52 */
-static
+UNIV_INTERN
 ibool
 log_block_checksum_is_ok_or_old_format(
 /*===================================*/
@@ -2084,7 +2084,7 @@ skip_this_recv_addr:
 /*******************************************************************//**
 Tries to parse a single log record and returns its length.
 @return	length of the record, or 0 if the record was not complete */
-static
+UNIV_INTERN
 ulint
 recv_parse_log_rec(
 /*===============*/
@@ -2155,7 +2155,7 @@ recv_parse_log_rec(
 
 /*******************************************************//**
 Calculates the new value for lsn when more data is added to the log. */
-static
+UNIV_INTERN
 ib_uint64_t
 recv_calc_lsn_on_data_add(
 /*======================*/
@@ -3542,6 +3542,8 @@ recv_reset_logs(
 	log_sys->archived_lsn = log_sys->lsn;
 #endif /* UNIV_LOG_ARCHIVE */
 
+	log_sys->tracked_lsn = log_sys->lsn;
+
 	log_block_init(log_sys->buf, log_sys->lsn);
 	log_block_set_first_rec_group(log_sys->buf, LOG_BLOCK_HDR_SIZE);
 

storage/xtradb/os/os0file.c

@@ -1939,6 +1939,25 @@ os_file_set_eof(
 #endif /* __WIN__ */
 }
 
+/***********************************************************************//**
+Truncates a file at the specified position.
+@return TRUE if success */
+UNIV_INTERN
+ibool
+os_file_set_eof_at(
+	os_file_t	file, /*!< in: handle to a file */
+	ib_uint64_t	new_len)/*!< in: new file length */
+{
+#ifdef __WIN__
+	/* TODO: untested! */
+	return(!_chsize_s(file, new_len));
+#else
+	/* TODO: works only with -D_FILE_OFFSET_BITS=64 ? */
+	return(!ftruncate(file, new_len));
+#endif
+}
+
+
 #ifndef __WIN__
 /***********************************************************************//**
 Wrapper to fsync(2) that retries the call on some errors.

storage/xtradb/page/page0cur.c

@@ -1902,6 +1902,7 @@ page_cur_delete_rec(
 
 	/* Save to local variables some data associated with current_rec */
 	cur_slot_no = page_dir_find_owner_slot(current_rec);
+	ut_ad(cur_slot_no > 0);
 	cur_dir_slot = page_dir_get_nth_slot(page, cur_slot_no);
 	cur_n_owned = page_dir_slot_get_n_owned(cur_dir_slot);
 

storage/xtradb/page/page0page.c

@@ -780,17 +780,23 @@ page_copy_rec_list_start(
 	if (UNIV_LIKELY_NULL(new_page_zip)) {
 		mtr_set_log_mode(mtr, log_mode);
 
+		DBUG_EXECUTE_IF("page_copy_rec_list_start_compress_fail",
+				goto zip_reorganize;);
+
 		if (UNIV_UNLIKELY
 		    (!page_zip_compress(new_page_zip, new_page, index, mtr))) {
+			ulint	ret_pos;
+#ifndef DBUG_OFF
+zip_reorganize:
+#endif /* DBUG_OFF */
 			/* Before trying to reorganize the page,
 			store the number of preceding records on the page. */
-			ulint	ret_pos
-				= page_rec_get_n_recs_before(ret);
+			ret_pos = page_rec_get_n_recs_before(ret);
 			/* Before copying, "ret" was the predecessor
 			of the predefined supremum record.  If it was
 			the predefined infimum record, then it would
-			still be the infimum.  Thus, the assertion
-			ut_a(ret_pos > 0) would fail here. */
+			still be the infimum, and we would have
+			ret_pos == 0. */
 
 			if (UNIV_UNLIKELY
 			    (!page_zip_reorganize(new_block, index, mtr))) {
@@ -806,15 +812,10 @@ page_copy_rec_list_start(
 				btr_blob_dbg_add(new_page, index,
 						 "copy_start_reorg_fail");
 				return(NULL);
-			} else {
-				/* The page was reorganized:
-				Seek to ret_pos. */
-				ret = new_page + PAGE_NEW_INFIMUM;
-
-				do {
-					ret = rec_get_next_ptr(ret, TRUE);
-				} while (--ret_pos);
 			}
+
+			/* The page was reorganized: Seek to ret_pos. */
+			ret = page_rec_get_nth(new_page, ret_pos);
 		}
 	}
 
@@ -1050,6 +1051,7 @@ page_delete_rec_list_end(
 
 		n_owned = rec_get_n_owned_new(rec2) - count;
 		slot_index = page_dir_find_owner_slot(rec2);
+		ut_ad(slot_index > 0);
 		slot = page_dir_get_nth_slot(page, slot_index);
 	} else {
 		rec_t*	rec2	= rec;
@@ -1065,6 +1067,7 @@ page_delete_rec_list_end(
 
 		n_owned = rec_get_n_owned_old(rec2) - count;
 		slot_index = page_dir_find_owner_slot(rec2);
+		ut_ad(slot_index > 0);
 		slot = page_dir_get_nth_slot(page, slot_index);
 	}
 
@@ -1491,6 +1494,10 @@ page_rec_get_nth_const(
 	ulint			n_owned;
 	const rec_t*		rec;
 
+	if (nth == 0) {
+		return(page_get_infimum_rec(page));
+	}
+
 	ut_ad(nth < UNIV_PAGE_SIZE / (REC_N_NEW_EXTRA_BYTES + 1));
 
 	for (i = 0;; i++) {
@@ -1584,7 +1591,7 @@ page_rec_get_n_recs_before(
 	n--;
 
 	ut_ad(n >= 0);
-	ut_ad(n < UNIV_PAGE_SIZE / (REC_N_NEW_EXTRA_BYTES + 1));
+	ut_ad((ulint) n < UNIV_PAGE_SIZE / (REC_N_NEW_EXTRA_BYTES + 1));
 
 	return((ulint) n);
 }

storage/xtradb/plug.in

@@ -93,7 +93,6 @@ MYSQL_PLUGIN_ACTIONS(xtradb,  [
 	if (res != 10 || c != 123) {
 	  return(1);
 	}
-
 	return(0);
       }
     ],
@@ -107,6 +106,42 @@ MYSQL_PLUGIN_ACTIONS(xtradb,  [
     ]
   )
 
+  AC_MSG_CHECKING(whether GCC 64-bit atomic builtins are available)
+  # either define HAVE_IB_GCC_ATOMIC_BUILTINS_64 or not
+  AC_TRY_RUN(
+    [
+      #include <stdint.h>
+      int main()
+      {
+        int64_t x, y, res;
+
+	x = 10;
+	y = 123;
+	res = __sync_bool_compare_and_swap(&x, x, y);
+	if (!res || x != y) {
+	  return(1);
+	}
+
+	x = 10;
+	y = 123;
+	res = __sync_add_and_fetch(&x, y);
+	if (res != 123 + 10 || x != 123 + 10) {
+	  return(1);
+	}
+
+	return(0);
+      }
+    ],
+    [
+      AC_DEFINE([HAVE_IB_GCC_ATOMIC_BUILTINS_64], [1],
+                [GCC 64-bit atomic builtins are available])
+      AC_MSG_RESULT(yes)
+    ],
+    [
+      AC_MSG_RESULT(no)
+    ]
+  )
+
   AC_MSG_CHECKING(whether pthread_t can be used by GCC atomic builtins)
   # either define HAVE_IB_ATOMIC_PTHREAD_T_GCC or not
   AC_TRY_RUN(

storage/xtradb/row/row0ins.c

@@ -1296,7 +1296,8 @@ run_again:
 		check_index = foreign->foreign_index;
 	}
 
-	if (check_table == NULL || check_table->ibd_file_missing) {
+	if (check_table == NULL || check_table->ibd_file_missing
+	    || check_index == NULL) {
 		if (check_ref) {
 			FILE*	ef = dict_foreign_err_file;
 
@@ -1331,9 +1332,6 @@ run_again:
 		goto exit_func;
 	}
 
-	ut_a(check_table);
-	ut_a(check_index);
-
 	if (check_table != table) {
 		/* We already have a LOCK_IX on table, but not necessarily
 		on check_table */
@@ -2194,9 +2192,16 @@ row_ins_index_entry_low(
 
 				goto function_exit;
 			}
-			err = btr_cur_pessimistic_insert(
+
+			err = btr_cur_optimistic_insert(
 				0, &cursor, entry, &insert_rec, &big_rec,
 				n_ext, thr, &mtr);
+
+			if (err == DB_FAIL) {
+				err = btr_cur_pessimistic_insert(
+					0, &cursor, entry, &insert_rec,
+					&big_rec, n_ext, thr, &mtr);
+			}
 		}
 	}
 

storage/xtradb/row/row0merge.c

@@ -1214,11 +1214,25 @@ row_merge_read_clustered_index(
 				goto err_exit;
 			}
 
+			/* Store the cursor position on the last user
+			record on the page. */
+			btr_pcur_move_to_prev_on_page(&pcur);
+			/* Leaf pages must never be empty, unless
+			this is the only page in the index tree. */
+			ut_ad(btr_pcur_is_on_user_rec(&pcur)
+			      || buf_block_get_page_no(
+				      btr_pcur_get_block(&pcur))
+			      == clust_index->page);
+
 			btr_pcur_store_position(&pcur, &mtr);
 			mtr_commit(&mtr);
 			mtr_start(&mtr);
+			/* Restore position on the record, or its
+			predecessor if the record was purged
+			meanwhile. */
 			btr_pcur_restore_position(BTR_SEARCH_LEAF,
 						  &pcur, &mtr);
+			/* Move to the successor of the original record. */
 			has_next = btr_pcur_move_to_next_user_rec(&pcur, &mtr);
 		}
 

storage/xtradb/row/row0mysql.c

@@ -3636,7 +3636,7 @@ row_mysql_drop_temp_tables(void)
 		btr_pcur_store_position(&pcur, &mtr);
 		btr_pcur_commit_specify_mtr(&pcur, &mtr);
 
-		table = dict_load_table(table_name);
+		table = dict_table_get_low(table_name);
 
 		if (table) {
 			row_drop_table_for_mysql(table_name, trx, FALSE);

storage/xtradb/row/row0vers.c

@@ -208,18 +208,6 @@ row_vers_impl_x_locked_off_kernel(
 		vers_del = rec_get_deleted_flag(prev_version, comp);
 		prev_trx_id = row_get_rec_trx_id(prev_version, clust_index,
 						 clust_offsets);
-
-		/* If the trx_id and prev_trx_id are different and if
-		the prev_version is marked deleted then the
-		prev_trx_id must have already committed for the trx_id
-		to be able to modify the row. Therefore, prev_trx_id
-		cannot hold any implicit lock. */
-		if (vers_del && 0 != ut_dulint_cmp(trx_id, prev_trx_id)) {
-
-			mutex_enter(&kernel_mutex);
-			break;
-		}
-
 		/* The stack of versions is locked by mtr.  Thus, it
 		is safe to fetch the prefixes for externally stored
 		columns. */

storage/xtradb/scripts/install_innodb_plugins.sql

@@ -7,3 +7,6 @@ INSTALL PLUGIN innodb_cmp SONAME 'ha_innodb.so';
 INSTALL PLUGIN innodb_cmp_reset SONAME 'ha_innodb.so';
 INSTALL PLUGIN innodb_cmpmem SONAME 'ha_innodb.so';
 INSTALL PLUGIN innodb_cmpmem_reset SONAME 'ha_innodb.so';
+INSTALL PLUGIN innodb_buffer_pool_stats SONAME 'ha_innodb.so';
+INSTALL PLUGIN innodb_buffer_page SONAME 'ha_innodb.so';
+INSTALL PLUGIN innodb_buffer_page_lru SONAME 'ha_innodb.so';

storage/xtradb/scripts/install_innodb_plugins_win.sql

@@ -7,3 +7,6 @@ INSTALL PLUGIN innodb_cmp SONAME 'ha_innodb.dll';
 INSTALL PLUGIN innodb_cmp_reset SONAME 'ha_innodb.dll';
 INSTALL PLUGIN innodb_cmpmem SONAME 'ha_innodb.dll';
 INSTALL PLUGIN innodb_cmpmem_reset SONAME 'ha_innodb.dll';
+INSTALL PLUGIN innodb_buffer_pool_stats SONAME 'ha_innodb.dll';
+INSTALL PLUGIN innodb_buffer_page SONAME 'ha_innodb.dll';
+INSTALL PLUGIN innodb_buffer_page_lru SONAME 'ha_innodb.dll';

storage/xtradb/srv/srv0srv.c

@@ -69,6 +69,7 @@ Created 10/8/1995 Heikki Tuuri
 #include "thr0loc.h"
 #include "que0que.h"
 #include "srv0que.h"
+#include "log0online.h"
 #include "log0recv.h"
 #include "pars0pars.h"
 #include "usr0sess.h"
@@ -161,6 +162,10 @@ UNIV_INTERN ibool	srv_recovery_stats = FALSE;
 
 UNIV_INTERN ulint	srv_use_purge_thread = 0;
 
+UNIV_INTERN my_bool	srv_track_changed_pages = TRUE;
+
+UNIV_INTERN ulonglong	srv_changed_pages_limit = 0;
+
 /* if TRUE, then we auto-extend the last data file */
 UNIV_INTERN ibool	srv_auto_extend_last_data_file	= FALSE;
 /* if != 0, this tells the max size auto-extending may increase the
@@ -405,6 +410,9 @@ UNIV_INTERN unsigned long long	srv_stats_sample_pages = 8;
 UNIV_INTERN ulong	srv_stats_auto_update = 1;
 UNIV_INTERN ulint	srv_stats_update_need_lock = 1;
 UNIV_INTERN ibool	srv_use_sys_stats_table = FALSE;
+#ifdef UNIV_DEBUG
+UNIV_INTERN ulong	srv_sys_stats_root_page = 0;
+#endif
 
 UNIV_INTERN ibool	srv_use_doublewrite_buf	= TRUE;
 UNIV_INTERN ibool	srv_use_checksums = TRUE;
@@ -724,6 +732,10 @@ UNIV_INTERN os_event_t	srv_lock_timeout_thread_event;
 
 UNIV_INTERN os_event_t	srv_shutdown_event;
 
+UNIV_INTERN os_event_t	srv_checkpoint_completed_event;
+
+UNIV_INTERN os_event_t	srv_redo_log_thread_finished_event;
+
 UNIV_INTERN srv_sys_t*	srv_sys	= NULL;
 
 /* padding to prevent other memory update hotspots from residing on
@@ -1031,6 +1043,9 @@ srv_init(void)
 	srv_lock_timeout_thread_event = os_event_create(NULL);
 	srv_shutdown_event = os_event_create(NULL);
 
+	srv_checkpoint_completed_event = os_event_create(NULL);
+	srv_redo_log_thread_finished_event = os_event_create(NULL);
+
 	for (i = 0; i < SRV_MASTER + 1; i++) {
 		srv_n_threads_active[i] = 0;
 		srv_n_threads[i] = 0;
@@ -1178,7 +1193,7 @@ retry:
 static void
 srv_conc_exit_innodb_timer_based(trx_t* trx)
 {
-        (void) os_atomic_increment_lint(&srv_conc_n_threads, -1);
+	(void) os_atomic_increment_lint(&srv_conc_n_threads, -1);
 	trx->declared_to_be_inside_innodb = FALSE;
 	trx->n_tickets_to_enter_innodb = 0;
 	return;
@@ -1385,7 +1400,7 @@ srv_conc_force_enter_innodb(
 	ut_ad(srv_conc_n_threads >= 0);
 #ifdef HAVE_ATOMIC_BUILTINS
 	if (srv_thread_concurrency_timer_based) {
-                (void) os_atomic_increment_lint(&srv_conc_n_threads, 1);
+		(void) os_atomic_increment_lint(&srv_conc_n_threads, 1);
 		trx->declared_to_be_inside_innodb = TRUE;
 		trx->n_tickets_to_enter_innodb = 1;
 		return;
@@ -2674,6 +2689,41 @@ exit_func:
 	OS_THREAD_DUMMY_RETURN;
 }
 
+/******************************************************************//**
+A thread which follows the redo log and outputs the changed page bitmap.
+@return a dummy value */
+UNIV_INTERN
+os_thread_ret_t
+srv_redo_log_follow_thread(
+/*=======================*/
+	void*	arg __attribute__((unused)))	/*!< in: a dummy parameter
+						     required by
+						     os_thread_create */
+{
+#ifdef UNIV_DEBUG_THREAD_CREATION
+	fprintf(stderr, "Redo log follower thread starts, id %lu\n",
+		os_thread_pf(os_thread_get_curr_id()));
+#endif
+	my_thread_init();
+
+	do {
+		os_event_wait(srv_checkpoint_completed_event);
+		os_event_reset(srv_checkpoint_completed_event);
+
+		log_online_follow_redo_log();
+
+	} while (srv_shutdown_state < SRV_SHUTDOWN_LAST_PHASE);
+
+	log_online_read_shutdown();
+	os_event_set(srv_redo_log_thread_finished_event);
+
+	my_thread_end();
+	os_thread_exit(NULL);
+
+	OS_THREAD_DUMMY_RETURN;
+}
+
+
 /*******************************************************************//**
 Tells the InnoDB server that there has been activity in the database
 and wakes up the master thread if it is suspended (not sleeping). Used

storage/xtradb/srv/srv0start.c

@@ -51,6 +51,7 @@ Created 2/16/1996 Heikki Tuuri
 #include "rem0rec.h"
 #include "mtr0mtr.h"
 #include "log0log.h"
+#include "log0online.h"
 #include "log0recv.h"
 #include "page0page.h"
 #include "page0cur.h"
@@ -127,9 +128,9 @@ static mutex_t		ios_mutex;
 static ulint		ios;
 
 /** io_handler_thread parameters for thread identification */
-static ulint		n[SRV_MAX_N_IO_THREADS + 7 + UNIV_MAX_PARALLELISM];
+static ulint		n[SRV_MAX_N_IO_THREADS + 8 + UNIV_MAX_PARALLELISM];
 /** io_handler_thread identifiers */
-static os_thread_id_t	thread_ids[SRV_MAX_N_IO_THREADS + 7 + UNIV_MAX_PARALLELISM];
+static os_thread_id_t	thread_ids[SRV_MAX_N_IO_THREADS + 8 + UNIV_MAX_PARALLELISM];
 
 /** We use this mutex to test the return value of pthread_mutex_trylock
    on successful locking. HP-UX does NOT return 0, though Linux et al do. */
@@ -1823,6 +1824,12 @@ innobase_start_or_create_for_mysql(void)
 
 			trx_sys_dummy_create(TRX_DOUBLEWRITE_SPACE);
 		}
+
+		if (UNIV_UNLIKELY(!dict_verify_xtradb_sys_stats())) {
+			fprintf(stderr, "InnoDB: Warning: "
+				"SYS_STATS table corrupted, recreating\n");
+			dict_recreate_xtradb_sys_stats();
+		}
 	}
 
 	if (!create_new_db && sum_of_new_sizes > 0) {
@@ -1885,6 +1892,19 @@ innobase_start_or_create_for_mysql(void)
 	if (srv_auto_lru_dump && srv_blocking_lru_restore)
 		buf_LRU_file_restore();
 
+	if (srv_track_changed_pages) {
+
+		/* Initialize the log tracking subsystem here to block
+		server startup until it's completed due to the potential
+		need to re-read previous server run's log. */
+		log_online_read_init();
+
+		/* Create the thread that follows the redo log to output the
+		changed page bitmap */
+		os_thread_create(&srv_redo_log_follow_thread, NULL,
+				 thread_ids + 5 + SRV_MAX_N_IO_THREADS);
+	}
+
 	srv_is_being_started = FALSE;
 
 	if (trx_doublewrite == NULL) {

storage/xtradb/sync/sync0arr.c

@@ -926,6 +926,11 @@ sync_array_print_long_waits(
 	ibool		fatal = FALSE;
 	double		longest_diff = 0;
 
+	/* For huge tables, skip the check during CHECK TABLE etc... */
+	if (fatal_timeout > SRV_SEMAPHORE_WAIT_EXTENSION) {
+		return(FALSE);
+	}
+
 	for (i = 0; i < sync_primary_wait_array->n_cells; i++) {
 
 		double	diff;

storage/xtradb/ut/ut0rbt.c

@@ -48,7 +48,6 @@ red-black properties:
 #endif
 
 #define ROOT(t)		(t->root->left)
-#define	SIZEOF_NODE(t)	((sizeof(ib_rbt_node_t) + t->sizeof_value) - 1)
 
 /****************************************************************//**
 Print out the sub-tree recursively. */
@@ -829,6 +828,21 @@ rbt_add_node(
 	node = (ib_rbt_node_t*) ut_malloc(SIZEOF_NODE(tree));
 
 	memcpy(node->value, value, tree->sizeof_value);
+	return(rbt_add_preallocated_node(tree, parent, node));
+}
+
+/****************************************************************//**
+Add a new caller-provided node to tree at the specified position.
+The node must have its key fields initialized correctly.
+@return added node */
+UNIV_INTERN
+const ib_rbt_node_t*
+rbt_add_preallocated_node(
+/*======================*/
+	ib_rbt_t*	tree,		/*!< in: rb tree */
+	ib_rbt_bound_t*	parent,		/*!< in: parent */
+	ib_rbt_node_t*	node)		/*!< in: node */
+{
 	node->parent = node->left = node->right = tree->nil;
 
 	/* If tree is empty */
@@ -1137,7 +1151,17 @@ rbt_clear(
 	ib_rbt_t*	tree)		/*!< in: rb tree */
 {
 	rbt_free_node(ROOT(tree), tree->nil);
+	rbt_reset(tree);
+}
 
+/****************************************************************//**
+Clear the tree without deleting and freeing its nodes. */
+UNIV_INTERN
+void
+rbt_reset(
+/*======*/
+	ib_rbt_t*	tree)		/*!< in: rb tree */
+{
 	tree->n_nodes = 0;
 	tree->root->left = tree->root->right = tree->nil;
 }