- needs gcc4 to build, be sure detect is ok

[packages/mysql.git] / mysql-innodb_extend_slow.patch

# name       : innodb_extend_slow.patch
# introduced : 11 or before
# maintainer : Yasufumi
#
#!!! notice !!!
# Any small change to this file in the main branch
# should be done or reviewed by the maintainer!
diff -ruN a/storage/innodb_plugin/buf/buf0buf.c b/storage/innodb_plugin/buf/buf0buf.c
--- a/storage/innodb_plugin/buf/buf0buf.c	2010-08-27 16:13:11.061058561 +0900
+++ b/storage/innodb_plugin/buf/buf0buf.c	2010-08-27 16:30:47.341987400 +0900
@@ -51,6 +51,40 @@
 #include "dict0dict.h"
 #include "log0recv.h"
 #include "page0zip.h"
+#include "trx0trx.h"
+
+/* prototypes for new functions added to ha_innodb.cc */
+trx_t* innobase_get_trx();
+
+inline void _increment_page_get_statistics(buf_block_t* block, trx_t* trx)
+{
+	ulint           block_hash;
+	ulint           block_hash_byte;
+	byte            block_hash_offset;
+
+	ut_ad(block);
+
+	if (!innobase_get_slow_log() || !trx || !trx->take_stats)
+		return;
+
+	if (!trx->distinct_page_access_hash) {
+		trx->distinct_page_access_hash = mem_alloc(DPAH_SIZE);
+		memset(trx->distinct_page_access_hash, 0, DPAH_SIZE);
+	}
+
+	block_hash = ut_hash_ulint((block->page.space << 20) + block->page.space +
+					block->page.offset, DPAH_SIZE << 3);
+	block_hash_byte = block_hash >> 3;
+	block_hash_offset = (byte) block_hash & 0x07;
+	if (block_hash_byte < 0 || block_hash_byte >= DPAH_SIZE)
+		fprintf(stderr, "!!! block_hash_byte = %lu  block_hash_offset = %d !!!\n", block_hash_byte, block_hash_offset);
+	if (block_hash_offset < 0 || block_hash_offset > 7)
+		fprintf(stderr, "!!! block_hash_byte = %lu  block_hash_offset = %d !!!\n", block_hash_byte, block_hash_offset);
+	if ((trx->distinct_page_access_hash[block_hash_byte] & ((byte) 0x01 << block_hash_offset)) == 0)
+		trx->distinct_page_access++;
+	trx->distinct_page_access_hash[block_hash_byte] |= (byte) 0x01 << block_hash_offset;
+	return;
+}
 
 /*
 		IMPLEMENTATION OF THE BUFFER POOL
@@ -1726,10 +1760,18 @@
 	mutex_t*	block_mutex;
 	ibool		must_read;
 	unsigned	access_time;
+	trx_t*		trx = NULL;
+	ulint		sec;
+	ulint		ms;
+	ib_uint64_t	start_time;
+	ib_uint64_t	finish_time;
 
 #ifndef UNIV_LOG_DEBUG
 	ut_ad(!ibuf_inside());
 #endif
+	if (innobase_get_slow_log()) {
+		trx = innobase_get_trx();
+	}
 	buf_pool->stat.n_page_gets++;
 
 	for (;;) {
@@ -1746,7 +1788,7 @@
 		//buf_pool_mutex_exit();
 		rw_lock_s_unlock(&page_hash_latch);
 
-		buf_read_page(space, zip_size, offset);
+		buf_read_page(space, zip_size, offset, trx);
 
 #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG
 		ut_a(++buf_dbg_counter % 37 || buf_validate());
@@ -1822,6 +1864,13 @@
 		/* Let us wait until the read operation
 		completes */
 
+		if (innobase_get_slow_log() && trx && trx->take_stats)
+		{
+			ut_usectime(&sec, &ms);
+			start_time = (ib_uint64_t)sec * 1000000 + ms;
+		} else {
+			start_time = 0;
+		}
 		for (;;) {
 			enum buf_io_fix	io_fix;
 
@@ -1836,6 +1885,12 @@
 				break;
 			}
 		}
+		if (innobase_get_slow_log() && trx && trx->take_stats && start_time)
+		{
+			ut_usectime(&sec, &ms);
+			finish_time = (ib_uint64_t)sec * 1000000 + ms;
+			trx->io_reads_wait_timer += (ulint)(finish_time - start_time);
+		}
 	}
 
 #ifdef UNIV_IBUF_COUNT_DEBUG
@@ -2092,6 +2147,11 @@
 	ibool		must_read;
 	ulint		retries = 0;
 	mutex_t*	block_mutex;
+	trx_t*          trx = NULL;
+	ulint           sec;
+	ulint           ms;
+	ib_uint64_t     start_time;
+	ib_uint64_t     finish_time;
 
 	ut_ad(mtr);
 	ut_ad(mtr->state == MTR_ACTIVE);
@@ -2106,6 +2166,9 @@
 #ifndef UNIV_LOG_DEBUG
 	ut_ad(!ibuf_inside() || ibuf_page(space, zip_size, offset, NULL));
 #endif
+	if (innobase_get_slow_log()) {
+		trx = innobase_get_trx();
+	}
 	buf_pool->stat.n_page_gets++;
 loop:
 	block = guess;
@@ -2159,7 +2222,7 @@
 			return(NULL);
 		}
 
-		if (buf_read_page(space, zip_size, offset)) {
+		if (buf_read_page(space, zip_size, offset, trx)) {
 			retries = 0;
 		} else if (retries < BUF_PAGE_READ_MAX_RETRIES) {
 			++retries;
@@ -2444,6 +2507,13 @@
 			/* Let us wait until the read operation
 			completes */
 
+			if (innobase_get_slow_log() && trx && trx->take_stats)
+			{
+				ut_usectime(&sec, &ms);
+				start_time = (ib_uint64_t)sec * 1000000 + ms;
+			} else {
+				start_time = 0;
+			}
 			for (;;) {
 				enum buf_io_fix	io_fix;
 
@@ -2458,6 +2528,12 @@
 					break;
 				}
 			}
+			if (innobase_get_slow_log() && trx && trx->take_stats && start_time)
+			{
+				ut_usectime(&sec, &ms);
+				finish_time = (ib_uint64_t)sec * 1000000 + ms;
+				trx->io_reads_wait_timer += (ulint)(finish_time - start_time);
+			}
 		}
 
 		fix_type = MTR_MEMO_BUF_FIX;
@@ -2483,13 +2559,17 @@
 		/* In the case of a first access, try to apply linear
 		read-ahead */
 
-		buf_read_ahead_linear(space, zip_size, offset);
+		buf_read_ahead_linear(space, zip_size, offset, trx);
 	}
 
 #ifdef UNIV_IBUF_COUNT_DEBUG
 	ut_a(ibuf_count_get(buf_block_get_space(block),
 			    buf_block_get_page_no(block)) == 0);
 #endif
+	if (innobase_get_slow_log()) {
+		_increment_page_get_statistics(block, trx);
+	}
+
 	return(block);
 }
 
@@ -2512,6 +2592,7 @@
 	unsigned	access_time;
 	ibool		success;
 	ulint		fix_type;
+	trx_t*		trx = NULL;
 
 	ut_ad(block);
 	ut_ad(mtr);
@@ -2589,13 +2670,17 @@
 #ifdef UNIV_DEBUG_FILE_ACCESSES
 	ut_a(block->page.file_page_was_freed == FALSE);
 #endif
+	if (innobase_get_slow_log()) {
+		trx = innobase_get_trx();
+	}
+
 	if (UNIV_UNLIKELY(!access_time)) {
 		/* In the case of a first access, try to apply linear
 		read-ahead */
 
 		buf_read_ahead_linear(buf_block_get_space(block),
 				      buf_block_get_zip_size(block),
-				      buf_block_get_page_no(block));
+				      buf_block_get_page_no(block), trx);
 	}
 
 #ifdef UNIV_IBUF_COUNT_DEBUG
@@ -2604,6 +2689,9 @@
 #endif
 	buf_pool->stat.n_page_gets++;
 
+	if (innobase_get_slow_log()) {
+		_increment_page_get_statistics(block, trx);
+	}
 	return(TRUE);
 }
 
@@ -2625,6 +2713,7 @@
 {
 	ibool		success;
 	ulint		fix_type;
+	trx_t*		trx = NULL;
 
 	ut_ad(mtr);
 	ut_ad(mtr->state == MTR_ACTIVE);
@@ -2709,6 +2798,11 @@
 #endif
 	buf_pool->stat.n_page_gets++;
 
+	if (innobase_get_slow_log()) {
+		trx = innobase_get_trx();
+		_increment_page_get_statistics(block, trx);
+	}
+
 	return(TRUE);
 }
 
diff -ruN a/storage/innodb_plugin/buf/buf0rea.c b/storage/innodb_plugin/buf/buf0rea.c
--- a/storage/innodb_plugin/buf/buf0rea.c	2010-08-27 16:23:31.014020792 +0900
+++ b/storage/innodb_plugin/buf/buf0rea.c	2010-08-27 16:30:47.342987531 +0900
@@ -75,7 +75,8 @@
 			treat the tablespace as dropped; this is a timestamp we
 			use to stop dangling page reads from a tablespace
 			which we have DISCARDed + IMPORTed back */
-	ulint	offset)	/*!< in: page number */
+	ulint	offset,	/*!< in: page number */
+	trx_t*	trx)
 {
 	buf_page_t*	bpage;
 	ulint		wake_later;
@@ -176,15 +177,15 @@
 	ut_ad(buf_page_in_file(bpage));
 
 	if (zip_size) {
-		*err = fil_io(OS_FILE_READ | wake_later,
+		*err = _fil_io(OS_FILE_READ | wake_later,
 			      sync, space, zip_size, offset, 0, zip_size,
-			      bpage->zip.data, bpage);
+			      bpage->zip.data, bpage, trx);
 	} else {
 		ut_a(buf_page_get_state(bpage) == BUF_BLOCK_FILE_PAGE);
 
-		*err = fil_io(OS_FILE_READ | wake_later,
+		*err = _fil_io(OS_FILE_READ | wake_later,
 			      sync, space, 0, offset, 0, UNIV_PAGE_SIZE,
-			      ((buf_block_t*) bpage)->frame, bpage);
+			      ((buf_block_t*) bpage)->frame, bpage, trx);
 	}
 	ut_a(*err == DB_SUCCESS);
 
@@ -209,7 +210,8 @@
 /*==========*/
 	ulint	space,	/*!< in: space id */
 	ulint	zip_size,/*!< in: compressed page size in bytes, or 0 */
-	ulint	offset)	/*!< in: page number */
+	ulint	offset,	/*!< in: page number */
+	trx_t*	trx)
 {
 	ib_int64_t	tablespace_version;
 	ulint		count;
@@ -222,7 +224,7 @@
 
 	count = buf_read_page_low(&err, TRUE, BUF_READ_ANY_PAGE, space,
 				  zip_size, FALSE,
-				  tablespace_version, offset);
+				  tablespace_version, offset, trx);
 	srv_buf_pool_reads += count;
 	if (err == DB_TABLESPACE_DELETED) {
 		ut_print_timestamp(stderr);
@@ -273,8 +275,9 @@
 /*==================*/
 	ulint	space,	/*!< in: space id */
 	ulint	zip_size,/*!< in: compressed page size in bytes, or 0 */
-	ulint	offset)	/*!< in: page number of a page; NOTE: the current thread
+	ulint	offset,	/*!< in: page number of a page; NOTE: the current thread
 			must want access to this page (see NOTE 3 above) */
+	trx_t*	trx)
 {
 	ib_int64_t	tablespace_version;
 	buf_page_t*	bpage;
@@ -497,7 +500,7 @@
 			count += buf_read_page_low(
 				&err, FALSE,
 				ibuf_mode | OS_AIO_SIMULATED_WAKE_LATER,
-				space, zip_size, FALSE, tablespace_version, i);
+				space, zip_size, FALSE, tablespace_version, i, trx);
 			if (err == DB_TABLESPACE_DELETED) {
 				ut_print_timestamp(stderr);
 				fprintf(stderr,
@@ -587,7 +590,7 @@
 		buf_read_page_low(&err, sync && (i + 1 == n_stored),
 				  BUF_READ_ANY_PAGE, space_ids[i],
 				  zip_size, TRUE, space_versions[i],
-				  page_nos[i]);
+				  page_nos[i], NULL);
 
 		if (UNIV_UNLIKELY(err == DB_TABLESPACE_DELETED)) {
 tablespace_deleted:
@@ -728,12 +731,12 @@
 		if ((i + 1 == n_stored) && sync) {
 			buf_read_page_low(&err, TRUE, BUF_READ_ANY_PAGE, space,
 					  zip_size, TRUE, tablespace_version,
-					  page_nos[i]);
+					  page_nos[i], NULL);
 		} else {
 			buf_read_page_low(&err, FALSE, BUF_READ_ANY_PAGE
 					  | OS_AIO_SIMULATED_WAKE_LATER,
 					  space, zip_size, TRUE,
-					  tablespace_version, page_nos[i]);
+					  tablespace_version, page_nos[i], NULL);
 		}
 	}
 
diff -ruN a/storage/innodb_plugin/fil/fil0fil.c b/storage/innodb_plugin/fil/fil0fil.c
--- a/storage/innodb_plugin/fil/fil0fil.c	2010-08-27 16:15:55.187400372 +0900
+++ b/storage/innodb_plugin/fil/fil0fil.c	2010-08-27 16:30:47.346992376 +0900
@@ -4325,7 +4325,7 @@
 				 node->name, node->handle, buf,
 				 offset_low, offset_high,
 				 page_size * n_pages,
-				 NULL, NULL);
+				 NULL, NULL, NULL);
 #endif
 		if (success) {
 			node->size += n_pages;
@@ -4652,7 +4652,7 @@
 i/o on a tablespace which does not exist */
 UNIV_INTERN
 ulint
-fil_io(
+_fil_io(
 /*===*/
 	ulint	type,		/*!< in: OS_FILE_READ or OS_FILE_WRITE,
 				ORed to OS_FILE_LOG, if a log i/o
@@ -4677,8 +4677,9 @@
 	void*	buf,		/*!< in/out: buffer where to store read data
 				or from where to write; in aio this must be
 				appropriately aligned */
-	void*	message)	/*!< in: message for aio handler if non-sync
+	void*	message,	/*!< in: message for aio handler if non-sync
 				aio used, else ignored */
+	trx_t*	trx)
 {
 	ulint		mode;
 	fil_space_t*	space;
@@ -4848,7 +4849,7 @@
 #else
 	/* Queue the aio request */
 	ret = os_aio(type, mode | wake_later, node->name, node->handle, buf,
-		     offset_low, offset_high, len, node, message);
+		     offset_low, offset_high, len, node, message, trx);
 #endif
 	ut_a(ret);
 
diff -ruN a/storage/innodb_plugin/handler/ha_innodb.cc b/storage/innodb_plugin/handler/ha_innodb.cc
--- a/storage/innodb_plugin/handler/ha_innodb.cc	2010-08-27 16:30:34.589021493 +0900
+++ b/storage/innodb_plugin/handler/ha_innodb.cc	2010-08-27 16:30:47.356987871 +0900
@@ -1372,6 +1372,16 @@
 	trx->check_unique_secondary = !thd_test_options(
 		thd, OPTION_RELAXED_UNIQUE_CHECKS);
 
+#ifdef EXTENDED_SLOWLOG
+	if (thd_log_slow_verbosity(thd) & SLOG_V_INNODB) {
+		trx->take_stats = TRUE;
+	} else {
+		trx->take_stats = FALSE;
+	}
+#else
+	trx->take_stats = FALSE;
+#endif
+
 	DBUG_VOID_RETURN;
 }
 
@@ -1427,6 +1437,32 @@
 }
 
 
+/*************************************************************************
+Gets current trx. */
+extern "C"
+trx_t*
+innobase_get_trx()
+{
+	THD *thd=current_thd;
+	if (likely(thd != 0)) {
+		trx_t*& trx = thd_to_trx(thd);
+		return(trx);
+	} else {
+		return(NULL);
+	}
+}
+
+extern "C"
+ibool
+innobase_get_slow_log()
+{
+#ifdef EXTENDED_SLOWLOG
+	return((ibool) thd_opt_slow_log());
+#else
+	return(FALSE);
+#endif
+}
+
 /*********************************************************************//**
 Construct ha_innobase handler. */
 UNIV_INTERN
@@ -8940,6 +8976,25 @@
 	statement has ended */
 
 	if (trx->n_mysql_tables_in_use == 0) {
+#ifdef EXTENDED_SLOWLOG
+		increment_thd_innodb_stats(thd,
+					(unsigned long long) ut_conv_dulint_to_longlong(trx->id),
+					trx->io_reads,
+					trx->io_read,
+					trx->io_reads_wait_timer,
+					trx->lock_que_wait_timer,
+					trx->innodb_que_wait_timer,
+					trx->distinct_page_access);
+
+		trx->io_reads = 0;
+		trx->io_read = 0;
+		trx->io_reads_wait_timer = 0;
+		trx->lock_que_wait_timer = 0;
+		trx->innodb_que_wait_timer = 0;
+		trx->distinct_page_access = 0;
+		if (trx->distinct_page_access_hash)
+			memset(trx->distinct_page_access_hash, 0, DPAH_SIZE);
+#endif
 
 		trx->mysql_n_tables_locked = 0;
 		prebuilt->used_in_HANDLER = FALSE;
diff -ruN a/storage/innodb_plugin/handler/innodb_patch_info.h b/storage/innodb_plugin/handler/innodb_patch_info.h
--- a/storage/innodb_plugin/handler/innodb_patch_info.h	2010-08-27 16:30:34.590004526 +0900
+++ b/storage/innodb_plugin/handler/innodb_patch_info.h	2010-08-27 16:30:47.361987777 +0900
@@ -40,5 +40,6 @@
 {"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"},
 {NULL, NULL, NULL, NULL}
 };
diff -ruN a/storage/innodb_plugin/include/buf0rea.h b/storage/innodb_plugin/include/buf0rea.h
--- a/storage/innodb_plugin/include/buf0rea.h	2010-08-27 15:54:18.078987755 +0900
+++ b/storage/innodb_plugin/include/buf0rea.h	2010-08-27 16:30:47.363031394 +0900
@@ -27,6 +27,7 @@
 #define buf0rea_h
 
 #include "univ.i"
+#include "trx0types.h"
 #include "buf0types.h"
 
 /********************************************************************//**
@@ -41,7 +42,8 @@
 /*==========*/
 	ulint	space,	/*!< in: space id */
 	ulint	zip_size,/*!< in: compressed page size in bytes, or 0 */
-	ulint	offset);/*!< in: page number */
+	ulint	offset, /*!< in: page number */
+	trx_t*	trx);
 /********************************************************************//**
 Applies linear read-ahead if in the buf_pool the page is a border page of
 a linear read-ahead area and all the pages in the area have been accessed.
@@ -72,8 +74,9 @@
 /*==================*/
 	ulint	space,	/*!< in: space id */
 	ulint	zip_size,/*!< in: compressed page size in bytes, or 0 */
-	ulint	offset);/*!< in: page number of a page; NOTE: the current thread
+	ulint	offset, /*!< in: page number of a page; NOTE: the current thread
 			must want access to this page (see NOTE 3 above) */
+	trx_t*	trx);
 /********************************************************************//**
 Issues read requests for pages which the ibuf module wants to read in, in
 order to contract the insert buffer tree. Technically, this function is like
diff -ruN a/storage/innodb_plugin/include/fil0fil.h b/storage/innodb_plugin/include/fil0fil.h
--- a/storage/innodb_plugin/include/fil0fil.h	2010-08-27 15:52:14.325059269 +0900
+++ b/storage/innodb_plugin/include/fil0fil.h	2010-08-27 16:30:47.365059512 +0900
@@ -610,9 +610,12 @@
 Reads or writes data. This operation is asynchronous (aio).
 @return DB_SUCCESS, or DB_TABLESPACE_DELETED if we are trying to do
 i/o on a tablespace which does not exist */
+#define fil_io(type, sync, space_id, zip_size, block_offset, byte_offset, len, buf, message) \
+	_fil_io(type, sync, space_id, zip_size, block_offset, byte_offset, len, buf, message, NULL)
+
 UNIV_INTERN
 ulint
-fil_io(
+_fil_io(
 /*===*/
 	ulint	type,		/*!< in: OS_FILE_READ or OS_FILE_WRITE,
 				ORed to OS_FILE_LOG, if a log i/o
@@ -637,8 +640,9 @@
 	void*	buf,		/*!< in/out: buffer where to store read data
 				or from where to write; in aio this must be
 				appropriately aligned */
-	void*	message);	/*!< in: message for aio handler if non-sync
+	void*	message,	/*!< in: message for aio handler if non-sync
 				aio used, else ignored */
+	trx_t*	trx);
 /**********************************************************************//**
 Waits for an aio operation to complete. This function is used to write the
 handler for completed requests. The aio array of pending requests is divided
diff -ruN a/storage/innodb_plugin/include/os0file.h b/storage/innodb_plugin/include/os0file.h
--- a/storage/innodb_plugin/include/os0file.h	2010-08-04 02:24:19.000000000 +0900
+++ b/storage/innodb_plugin/include/os0file.h	2010-08-27 16:30:47.366987560 +0900
@@ -36,6 +36,7 @@
 #define os0file_h
 
 #include "univ.i"
+#include "trx0types.h"
 
 #ifndef __WIN__
 #include <dirent.h>
@@ -482,9 +483,12 @@
 /*******************************************************************//**
 Requests a synchronous read operation.
 @return	TRUE if request was successful, FALSE if fail */
+#define os_file_read(file, buf, offset, offset_high, n)         \
+		_os_file_read(file, buf, offset, offset_high, n, NULL)
+
 UNIV_INTERN
 ibool
-os_file_read(
+_os_file_read(
 /*=========*/
 	os_file_t	file,	/*!< in: handle to a file */
 	void*		buf,	/*!< in: buffer where to read */
@@ -492,7 +496,8 @@
 				offset where to read */
 	ulint		offset_high,/*!< in: most significant 32 bits of
 				offset */
-	ulint		n);	/*!< in: number of bytes to read */
+	ulint		n,	/*!< in: number of bytes to read */
+	trx_t*		trx);
 /*******************************************************************//**
 Rewind file to its start, read at most size - 1 bytes from it to str, and
 NUL-terminate str. All errors are silently ignored. This function is
@@ -646,10 +651,11 @@
 				(can be used to identify a completed
 				aio operation); ignored if mode is
 				OS_AIO_SYNC */
-	void*		message2);/*!< in: message for the aio handler
+	void*		message2,/*!< in: message for the aio handler
 				(can be used to identify a completed
 				aio operation); ignored if mode is
 				OS_AIO_SYNC */
+	trx_t*		trx);
 /************************************************************************//**
 Wakes up all async i/o threads so that they know to exit themselves in
 shutdown. */
diff -ruN a/storage/innodb_plugin/include/srv0srv.h b/storage/innodb_plugin/include/srv0srv.h
--- a/storage/innodb_plugin/include/srv0srv.h	2010-08-27 16:27:30.225055856 +0900
+++ b/storage/innodb_plugin/include/srv0srv.h	2010-08-27 16:30:47.367988259 +0900
@@ -62,6 +62,9 @@
 #define SRV_AUTO_EXTEND_INCREMENT	\
 	(srv_auto_extend_increment * ((1024 * 1024) / UNIV_PAGE_SIZE))
 
+/* prototypes for new functions added to ha_innodb.cc */
+ibool	innobase_get_slow_log();
+
 /* This is set to TRUE if the MySQL user has set it in MySQL */
 extern ibool	srv_lower_case_table_names;
 
diff -ruN a/storage/innodb_plugin/include/trx0trx.h b/storage/innodb_plugin/include/trx0trx.h
--- a/storage/innodb_plugin/include/trx0trx.h	2010-08-27 16:08:45.301058614 +0900
+++ b/storage/innodb_plugin/include/trx0trx.h	2010-08-27 16:30:47.369989369 +0900
@@ -738,6 +738,17 @@
 	/*------------------------------*/
 	char detailed_error[256];	/*!< detailed error message for last
 					error, or empty. */
+	/*------------------------------*/
+	ulint		io_reads;
+	ib_uint64_t	io_read;
+	ulint		io_reads_wait_timer;
+	ib_uint64_t	lock_que_wait_ustarted;
+	ulint           lock_que_wait_timer;
+	ulint           innodb_que_wait_timer;
+	ulint           distinct_page_access;
+#define	DPAH_SIZE	8192
+	byte*		distinct_page_access_hash;
+	ibool		take_stats;
 };
 
 #define TRX_MAX_N_THREADS	32	/* maximum number of
diff -ruN a/storage/innodb_plugin/lock/lock0lock.c b/storage/innodb_plugin/lock/lock0lock.c
--- a/storage/innodb_plugin/lock/lock0lock.c	2010-08-27 15:52:14.332058513 +0900
+++ b/storage/innodb_plugin/lock/lock0lock.c	2010-08-27 16:30:47.374058285 +0900
@@ -1757,6 +1757,8 @@
 {
 	lock_t*	lock;
 	trx_t*	trx;
+	ulint   sec;
+	ulint   ms;
 
 	ut_ad(mutex_own(&kernel_mutex));
 
@@ -1815,6 +1817,10 @@
 	trx->que_state = TRX_QUE_LOCK_WAIT;
 	trx->was_chosen_as_deadlock_victim = FALSE;
 	trx->wait_started = time(NULL);
+	if (innobase_get_slow_log() && trx->take_stats) {
+		ut_usectime(&sec, &ms);
+		trx->lock_que_wait_ustarted = (ib_uint64_t)sec * 1000000 + ms;
+	}
 
 	ut_a(que_thr_stop(thr));
 
@@ -3695,6 +3701,8 @@
 {
 	lock_t*	lock;
 	trx_t*	trx;
+	ulint   sec;
+	ulint   ms;
 
 	ut_ad(mutex_own(&kernel_mutex));
 
@@ -3750,6 +3758,10 @@
 		return(DB_SUCCESS);
 	}
 
+	if (innobase_get_slow_log() && trx->take_stats) {
+		ut_usectime(&sec, &ms);
+		trx->lock_que_wait_ustarted = (ib_uint64_t)sec * 1000000 + ms;
+	}
 	trx->que_state = TRX_QUE_LOCK_WAIT;
 	trx->was_chosen_as_deadlock_victim = FALSE;
 	trx->wait_started = time(NULL);
diff -ruN a/storage/innodb_plugin/os/os0file.c b/storage/innodb_plugin/os/os0file.c
--- a/storage/innodb_plugin/os/os0file.c	2010-08-27 16:23:31.038058669 +0900
+++ b/storage/innodb_plugin/os/os0file.c	2010-08-27 16:30:47.380058815 +0900
@@ -38,6 +38,8 @@
 #include "srv0start.h"
 #include "fil0fil.h"
 #include "buf0buf.h"
+#include "trx0sys.h"
+#include "trx0trx.h"
 #include "log0recv.h"
 #ifndef UNIV_HOTBACKUP
 # include "os0sync.h"
@@ -2087,22 +2089,30 @@
 /*******************************************************************//**
 Does a synchronous read operation in Posix.
 @return	number of bytes read, -1 if error */
+#define os_file_pread(file, buf, n, offset, offset_high)        \
+		_os_file_pread(file, buf, n, offset, offset_high, NULL);
+
 static
 ssize_t
-os_file_pread(
+_os_file_pread(
 /*==========*/
 	os_file_t	file,	/*!< in: handle to a file */
 	void*		buf,	/*!< in: buffer where to read */
 	ulint		n,	/*!< in: number of bytes to read */
 	ulint		offset,	/*!< in: least significant 32 bits of file
 				offset from where to read */
-	ulint		offset_high) /*!< in: most significant 32 bits of
+	ulint		offset_high, /*!< in: most significant 32 bits of
 				offset */
+	trx_t*		trx)
 {
 	off_t	offs;
 #if defined(HAVE_PREAD) && !defined(HAVE_BROKEN_PREAD)
 	ssize_t	n_bytes;
 #endif /* HAVE_PREAD && !HAVE_BROKEN_PREAD */
+	ulint		sec;
+	ulint		ms;
+	ib_uint64_t	start_time;
+	ib_uint64_t	finish_time;
 
 	ut_a((offset & 0xFFFFFFFFUL) == offset);
 
@@ -2123,6 +2133,15 @@
 
 	os_n_file_reads++;
 
+	if (innobase_get_slow_log() && trx && trx->take_stats)
+	{
+	        trx->io_reads++;
+		trx->io_read += n;
+		ut_usectime(&sec, &ms);
+		start_time = (ib_uint64_t)sec * 1000000 + ms;
+	} else {
+		start_time = 0;
+	}
 #if defined(HAVE_PREAD) && !defined(HAVE_BROKEN_PREAD)
 	os_mutex_enter(os_file_count_mutex);
 	os_file_n_pending_preads++;
@@ -2136,6 +2155,13 @@
 	os_n_pending_reads--;
 	os_mutex_exit(os_file_count_mutex);
 
+	if (innobase_get_slow_log() && trx && trx->take_stats && start_time)
+	{
+		ut_usectime(&sec, &ms);
+		finish_time = (ib_uint64_t)sec * 1000000 + ms;
+		trx->io_reads_wait_timer += (ulint)(finish_time - start_time);
+	}
+
 	return(n_bytes);
 #else
 	{
@@ -2172,6 +2198,13 @@
 		os_n_pending_reads--;
 		os_mutex_exit(os_file_count_mutex);
 
+		if (innobase_get_slow_log() && trx && trx->take_stats && start_time)
+		{
+			ut_usectime(&sec, &ms);
+			finish_time = (ib_uint64_t)sec * 1000000 + ms;
+			trx->io_reads_wait_timer += (ulint)(finish_time - start_time);
+		}
+
 		return(ret);
 	}
 #endif
@@ -2302,7 +2335,7 @@
 @return	TRUE if request was successful, FALSE if fail */
 UNIV_INTERN
 ibool
-os_file_read(
+_os_file_read(
 /*=========*/
 	os_file_t	file,	/*!< in: handle to a file */
 	void*		buf,	/*!< in: buffer where to read */
@@ -2310,7 +2343,8 @@
 				offset where to read */
 	ulint		offset_high, /*!< in: most significant 32 bits of
 				offset */
-	ulint		n)	/*!< in: number of bytes to read */
+	ulint		n,	/*!< in: number of bytes to read */
+	trx_t*		trx)
 {
 #ifdef __WIN__
 	BOOL		ret;
@@ -2385,7 +2419,7 @@
 	os_bytes_read_since_printout += n;
 
 try_again:
-	ret = os_file_pread(file, buf, n, offset, offset_high);
+	ret = _os_file_pread(file, buf, n, offset, offset_high, trx);
 
 	if ((ulint)ret == n) {
 
@@ -3356,7 +3390,8 @@
 				offset */
 	ulint		offset_high, /*!< in: most significant 32 bits of
 				offset */
-	ulint		len)	/*!< in: length of the block to read or write */
+	ulint		len,	/*!< in: length of the block to read or write */
+	trx_t*		trx)
 {
 	os_aio_slot_t*	slot;
 	ulint		i;
@@ -3642,10 +3677,11 @@
 				(can be used to identify a completed
 				aio operation); ignored if mode is
 				OS_AIO_SYNC */
-	void*		message2)/*!< in: message for the aio handler
+	void*		message2,/*!< in: message for the aio handler
 				(can be used to identify a completed
 				aio operation); ignored if mode is
 				OS_AIO_SYNC */
+	trx_t*		trx)
 {
 	os_aio_array_t*	array;
 	os_aio_slot_t*	slot;
@@ -3687,8 +3723,8 @@
 		wait in the Windows case. */
 
 		if (type == OS_FILE_READ) {
-			return(os_file_read(file, buf, offset,
-					    offset_high, n));
+			return(_os_file_read(file, buf, offset,
+					    offset_high, n, trx));
 		}
 
 		ut_a(type == OS_FILE_WRITE);
@@ -3721,8 +3757,13 @@
 		ut_error;
 	}
 
+	if (trx && type == OS_FILE_READ)
+	{
+		trx->io_reads++;
+		trx->io_read += n;
+	}
 	slot = os_aio_array_reserve_slot(type, array, message1, message2, file,
-					 name, buf, offset, offset_high, n);
+					 name, buf, offset, offset_high, n, trx);
 	if (type == OS_FILE_READ) {
 		if (os_aio_use_native_aio) {
 #ifdef WIN_ASYNC_IO
diff -ruN a/storage/innodb_plugin/srv/srv0srv.c b/storage/innodb_plugin/srv/srv0srv.c
--- a/storage/innodb_plugin/srv/srv0srv.c	2010-08-27 16:27:30.233022109 +0900
+++ b/storage/innodb_plugin/srv/srv0srv.c	2010-08-27 16:30:47.384058509 +0900
@@ -86,6 +86,9 @@
 #include "trx0i_s.h"
 #include "os0sync.h" /* for HAVE_ATOMIC_BUILTINS */
 
+/* prototypes for new functions added to ha_innodb.cc */
+ibool	innobase_get_slow_log();
+
 /* This is set to TRUE if the MySQL user has set it in MySQL; currently
 affects only FOREIGN KEY definition parsing */
 UNIV_INTERN ibool	srv_lower_case_table_names	= FALSE;
@@ -1155,6 +1158,10 @@
 	ibool			has_slept = FALSE;
 	srv_conc_slot_t*	slot	  = NULL;
 	ulint			i;
+	ib_uint64_t             start_time = 0L;
+	ib_uint64_t             finish_time = 0L;
+	ulint                   sec;
+	ulint                   ms;
 
 	if (trx->mysql_thd != NULL
 	    && thd_is_replication_slave_thread(trx->mysql_thd)) {
@@ -1231,6 +1238,7 @@
 		switches. */
 		if (SRV_THREAD_SLEEP_DELAY > 0) {
 			os_thread_sleep(SRV_THREAD_SLEEP_DELAY);
+			trx->innodb_que_wait_timer += SRV_THREAD_SLEEP_DELAY;
 		}
 
 		trx->op_info = "";
@@ -1286,12 +1294,25 @@
 	/* Go to wait for the event; when a thread leaves InnoDB it will
 	release this thread */
 
+	if (innobase_get_slow_log() && trx->take_stats) {
+		ut_usectime(&sec, &ms);
+		start_time = (ib_uint64_t)sec * 1000000 + ms;
+	} else {
+		start_time = 0;
+	}
+
 	trx->op_info = "waiting in InnoDB queue";
 
 	os_event_wait(slot->event);
 
 	trx->op_info = "";
 
+	if (innobase_get_slow_log() && trx->take_stats && start_time) {
+		ut_usectime(&sec, &ms);
+		finish_time = (ib_uint64_t)sec * 1000000 + ms;
+		trx->innodb_que_wait_timer += (ulint)(finish_time - start_time);
+	}
+
 	os_fast_mutex_lock(&srv_conc_mutex);
 
 	srv_conc_n_waiting_threads--;
diff -ruN a/storage/innodb_plugin/trx/trx0trx.c b/storage/innodb_plugin/trx/trx0trx.c
--- a/storage/innodb_plugin/trx/trx0trx.c	2010-08-27 16:08:45.306058896 +0900
+++ b/storage/innodb_plugin/trx/trx0trx.c	2010-08-27 16:30:47.387058330 +0900
@@ -179,6 +179,15 @@
 	trx->global_read_view = NULL;
 	trx->read_view = NULL;
 
+	trx->io_reads = 0;
+	trx->io_read = 0;
+	trx->io_reads_wait_timer = 0;
+	trx->lock_que_wait_timer = 0;
+	trx->innodb_que_wait_timer = 0;
+	trx->distinct_page_access = 0;
+	trx->distinct_page_access_hash = NULL;
+	trx->take_stats = FALSE;
+
 	/* Set X/Open XA transaction identification to NULL */
 	memset(&trx->xid, 0, sizeof(trx->xid));
 	trx->xid.formatID = -1;
@@ -216,6 +225,11 @@
 
 	trx->mysql_process_no = os_proc_get_number();
 
+	if (innobase_get_slow_log() && trx->take_stats) {
+		trx->distinct_page_access_hash = mem_alloc(DPAH_SIZE);
+		memset(trx->distinct_page_access_hash, 0, DPAH_SIZE);
+	}
+
 	return(trx);
 }
 
@@ -347,6 +361,12 @@
 /*===============*/
 	trx_t*	trx)	/*!< in, own: trx object */
 {
+	if (trx->distinct_page_access_hash)
+	{
+		mem_free(trx->distinct_page_access_hash);
+		trx->distinct_page_access_hash= NULL;
+	}
+
 	mutex_enter(&kernel_mutex);
 
 	UT_LIST_REMOVE(mysql_trx_list, trx_sys->mysql_trx_list, trx);
@@ -368,6 +388,12 @@
 /*====================*/
 	trx_t*	trx)	/*!< in, own: trx object */
 {
+	if (trx->distinct_page_access_hash)
+	{
+		mem_free(trx->distinct_page_access_hash);
+		trx->distinct_page_access_hash= NULL;
+	}
+
 	mutex_enter(&kernel_mutex);
 
 	trx_free(trx);
@@ -1093,6 +1119,9 @@
 	trx_t*	trx)	/*!< in: transaction */
 {
 	que_thr_t*	thr;
+	ulint           sec;
+	ulint           ms;
+	ib_uint64_t     now;
 
 	ut_ad(mutex_own(&kernel_mutex));
 	ut_ad(trx->que_state == TRX_QUE_LOCK_WAIT);
@@ -1107,6 +1136,11 @@
 		thr = UT_LIST_GET_FIRST(trx->wait_thrs);
 	}
 
+	if (innobase_get_slow_log() && trx->take_stats) {
+		ut_usectime(&sec, &ms);
+		now = (ib_uint64_t)sec * 1000000 + ms;
+		trx->lock_que_wait_timer += (ulint)(now - trx->lock_que_wait_ustarted);
+	}
 	trx->que_state = TRX_QUE_RUNNING;
 }
 
@@ -1120,6 +1154,9 @@
 	trx_t*	trx)	/*!< in: transaction in the TRX_QUE_LOCK_WAIT state */
 {
 	que_thr_t*	thr;
+	ulint           sec;
+	ulint           ms;
+	ib_uint64_t     now;
 
 	ut_ad(mutex_own(&kernel_mutex));
 	ut_ad(trx->que_state == TRX_QUE_LOCK_WAIT);
@@ -1134,6 +1171,11 @@
 		thr = UT_LIST_GET_FIRST(trx->wait_thrs);
 	}
 
+	if (innobase_get_slow_log() && trx->take_stats) {
+		ut_usectime(&sec, &ms);
+		now = (ib_uint64_t)sec * 1000000 + ms;
+		trx->lock_que_wait_timer += (ulint)(now - trx->lock_que_wait_ustarted);
+	}
 	trx->que_state = TRX_QUE_RUNNING;
 }