1 # name : innodb_extend_slow.patch
2 # introduced : 11 or before
3 # maintainer : Yasufumi
6 # Any small change to this file in the main branch
7 # should be done or reviewed by the maintainer!
8 diff -ruN a/storage/innobase/buf/buf0buf.c b/storage/innobase/buf/buf0buf.c
9 --- a/storage/innobase/buf/buf0buf.c 2010-12-03 15:49:59.175955882 +0900
10 +++ b/storage/innobase/buf/buf0buf.c 2010-12-03 17:42:42.074307123 +0900
12 #include "dict0dict.h"
17 +/* prototypes for new functions added to ha_innodb.cc */
18 +trx_t* innobase_get_trx();
20 +inline void _increment_page_get_statistics(buf_block_t* block, trx_t* trx)
23 + ulint block_hash_byte;
24 + byte block_hash_offset;
28 + if (!innobase_get_slow_log() || !trx || !trx->take_stats)
31 + if (!trx->distinct_page_access_hash) {
32 + trx->distinct_page_access_hash = mem_alloc(DPAH_SIZE);
33 + memset(trx->distinct_page_access_hash, 0, DPAH_SIZE);
36 + block_hash = ut_hash_ulint((block->page.space << 20) + block->page.space +
37 + block->page.offset, DPAH_SIZE << 3);
38 + block_hash_byte = block_hash >> 3;
39 + block_hash_offset = (byte) block_hash & 0x07;
40 + if (block_hash_byte < 0 || block_hash_byte >= DPAH_SIZE)
41 + fprintf(stderr, "!!! block_hash_byte = %lu block_hash_offset = %d !!!\n", block_hash_byte, block_hash_offset);
42 + if (block_hash_offset < 0 || block_hash_offset > 7)
43 + fprintf(stderr, "!!! block_hash_byte = %lu block_hash_offset = %d !!!\n", block_hash_byte, block_hash_offset);
44 + if ((trx->distinct_page_access_hash[block_hash_byte] & ((byte) 0x01 << block_hash_offset)) == 0)
45 + trx->distinct_page_access++;
46 + trx->distinct_page_access_hash[block_hash_byte] |= (byte) 0x01 << block_hash_offset;
51 IMPLEMENTATION OF THE BUFFER POOL
52 @@ -2399,11 +2433,19 @@
59 + ib_uint64_t start_time;
60 + ib_uint64_t finish_time;
61 buf_pool_t* buf_pool = buf_pool_get(space, offset);
63 #ifndef UNIV_LOG_DEBUG
64 ut_ad(!ibuf_inside());
66 + if (innobase_get_slow_log()) {
67 + trx = innobase_get_trx();
69 buf_pool->stat.n_page_gets++;
73 //buf_pool_mutex_exit(buf_pool);
74 rw_lock_s_unlock(&buf_pool->page_hash_latch);
76 - buf_read_page(space, zip_size, offset);
77 + buf_read_page(space, zip_size, offset, trx);
79 #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG
80 ut_a(++buf_dbg_counter % 37 || buf_validate());
81 @@ -2499,6 +2541,13 @@
82 /* Let us wait until the read operation
85 + if (innobase_get_slow_log() && trx && trx->take_stats)
87 + ut_usectime(&sec, &ms);
88 + start_time = (ib_uint64_t)sec * 1000000 + ms;
93 enum buf_io_fix io_fix;
95 @@ -2513,6 +2562,12 @@
99 + if (innobase_get_slow_log() && trx && trx->take_stats && start_time)
101 + ut_usectime(&sec, &ms);
102 + finish_time = (ib_uint64_t)sec * 1000000 + ms;
103 + trx->io_reads_wait_timer += (ulint)(finish_time - start_time);
107 #ifdef UNIV_IBUF_COUNT_DEBUG
108 @@ -2825,6 +2880,11 @@
111 mutex_t* block_mutex = NULL;
115 + ib_uint64_t start_time;
116 + ib_uint64_t finish_time;
117 buf_pool_t* buf_pool = buf_pool_get(space, offset);
120 @@ -2842,6 +2902,9 @@
121 #ifndef UNIV_LOG_DEBUG
122 ut_ad(!ibuf_inside() || ibuf_page(space, zip_size, offset, NULL));
124 + if (innobase_get_slow_log()) {
125 + trx = innobase_get_trx();
127 buf_pool->stat.n_page_gets++;
128 fold = buf_page_address_fold(space, offset);
130 @@ -2915,7 +2978,7 @@
134 - if (buf_read_page(space, zip_size, offset)) {
135 + if (buf_read_page(space, zip_size, offset, trx)) {
137 } else if (retries < BUF_PAGE_READ_MAX_RETRIES) {
139 @@ -3178,6 +3241,13 @@
140 /* Let us wait until the read operation
143 + if (innobase_get_slow_log() && trx && trx->take_stats)
145 + ut_usectime(&sec, &ms);
146 + start_time = (ib_uint64_t)sec * 1000000 + ms;
151 enum buf_io_fix io_fix;
153 @@ -3192,6 +3262,12 @@
157 + if (innobase_get_slow_log() && trx && trx->take_stats && start_time)
159 + ut_usectime(&sec, &ms);
160 + finish_time = (ib_uint64_t)sec * 1000000 + ms;
161 + trx->io_reads_wait_timer += (ulint)(finish_time - start_time);
165 fix_type = MTR_MEMO_BUF_FIX;
166 @@ -3217,13 +3293,17 @@
167 /* In the case of a first access, try to apply linear
170 - buf_read_ahead_linear(space, zip_size, offset);
171 + buf_read_ahead_linear(space, zip_size, offset, trx);
174 #ifdef UNIV_IBUF_COUNT_DEBUG
175 ut_a(ibuf_count_get(buf_block_get_space(block),
176 buf_block_get_page_no(block)) == 0);
178 + if (innobase_get_slow_log()) {
179 + _increment_page_get_statistics(block, trx);
185 @@ -3247,6 +3327,7 @@
186 unsigned access_time;
193 @@ -3324,13 +3405,17 @@
194 #ifdef UNIV_DEBUG_FILE_ACCESSES
195 ut_a(block->page.file_page_was_freed == FALSE);
197 + if (innobase_get_slow_log()) {
198 + trx = innobase_get_trx();
201 if (UNIV_UNLIKELY(!access_time)) {
202 /* In the case of a first access, try to apply linear
205 buf_read_ahead_linear(buf_block_get_space(block),
206 buf_block_get_zip_size(block),
207 - buf_block_get_page_no(block));
208 + buf_block_get_page_no(block), trx);
211 #ifdef UNIV_IBUF_COUNT_DEBUG
212 @@ -3340,6 +3425,9 @@
213 buf_pool = buf_pool_from_block(block);
214 buf_pool->stat.n_page_gets++;
216 + if (innobase_get_slow_log()) {
217 + _increment_page_get_statistics(block, trx);
222 @@ -3362,6 +3450,7 @@
223 buf_pool_t* buf_pool;
229 ut_ad(mtr->state == MTR_ACTIVE);
230 @@ -3448,6 +3537,11 @@
232 buf_pool->stat.n_page_gets++;
234 + if (innobase_get_slow_log()) {
235 + trx = innobase_get_trx();
236 + _increment_page_get_statistics(block, trx);
242 diff -ruN a/storage/innobase/buf/buf0rea.c b/storage/innobase/buf/buf0rea.c
243 --- a/storage/innobase/buf/buf0rea.c 2010-12-03 17:32:15.617037263 +0900
244 +++ b/storage/innobase/buf/buf0rea.c 2010-12-03 17:42:42.075297193 +0900
246 treat the tablespace as dropped; this is a timestamp we
247 use to stop dangling page reads from a tablespace
248 which we have DISCARDed + IMPORTed back */
249 - ulint offset) /*!< in: page number */
250 + ulint offset, /*!< in: page number */
255 @@ -179,15 +180,15 @@
257 thd_wait_begin(NULL, THD_WAIT_DISKIO);
259 - *err = fil_io(OS_FILE_READ | wake_later,
260 + *err = _fil_io(OS_FILE_READ | wake_later,
261 sync, space, zip_size, offset, 0, zip_size,
262 - bpage->zip.data, bpage);
263 + bpage->zip.data, bpage, trx);
265 ut_a(buf_page_get_state(bpage) == BUF_BLOCK_FILE_PAGE);
267 - *err = fil_io(OS_FILE_READ | wake_later,
268 + *err = _fil_io(OS_FILE_READ | wake_later,
269 sync, space, 0, offset, 0, UNIV_PAGE_SIZE,
270 - ((buf_block_t*) bpage)->frame, bpage);
271 + ((buf_block_t*) bpage)->frame, bpage, trx);
274 ut_a(*err == DB_SUCCESS);
277 ulint space, /*!< in: space id */
278 ulint zip_size,/*!< in: compressed page size in bytes, or 0 */
279 - ulint offset) /*!< in: page number */
280 + ulint offset, /*!< in: page number */
283 buf_pool_t* buf_pool = buf_pool_get(space, offset);
284 ib_int64_t tablespace_version;
287 count = buf_read_page_low(&err, TRUE, BUF_READ_ANY_PAGE, space,
289 - tablespace_version, offset);
290 + tablespace_version, offset, trx);
291 srv_buf_pool_reads += count;
292 if (err == DB_TABLESPACE_DELETED) {
293 ut_print_timestamp(stderr);
295 /*==================*/
296 ulint space, /*!< in: space id */
297 ulint zip_size,/*!< in: compressed page size in bytes, or 0 */
298 - ulint offset) /*!< in: page number of a page; NOTE: the current thread
299 + ulint offset, /*!< in: page number of a page; NOTE: the current thread
300 must want access to this page (see NOTE 3 above) */
303 buf_pool_t* buf_pool = buf_pool_get(space, offset);
304 ib_int64_t tablespace_version;
306 count += buf_read_page_low(
308 ibuf_mode | OS_AIO_SIMULATED_WAKE_LATER,
309 - space, zip_size, FALSE, tablespace_version, i);
310 + space, zip_size, FALSE, tablespace_version, i, trx);
311 if (err == DB_TABLESPACE_DELETED) {
312 ut_print_timestamp(stderr);
315 buf_read_page_low(&err, sync && (i + 1 == n_stored),
316 BUF_READ_ANY_PAGE, space_ids[i],
317 zip_size, TRUE, space_versions[i],
319 + page_nos[i], NULL);
321 if (UNIV_UNLIKELY(err == DB_TABLESPACE_DELETED)) {
323 @@ -736,12 +739,12 @@
324 if ((i + 1 == n_stored) && sync) {
325 buf_read_page_low(&err, TRUE, BUF_READ_ANY_PAGE, space,
326 zip_size, TRUE, tablespace_version,
328 + page_nos[i], NULL);
330 buf_read_page_low(&err, FALSE, BUF_READ_ANY_PAGE
331 | OS_AIO_SIMULATED_WAKE_LATER,
332 space, zip_size, TRUE,
333 - tablespace_version, page_nos[i]);
334 + tablespace_version, page_nos[i], NULL);
338 diff -ruN a/storage/innobase/fil/fil0fil.c b/storage/innobase/fil/fil0fil.c
339 --- a/storage/innobase/fil/fil0fil.c 2010-12-03 15:53:54.610037199 +0900
340 +++ b/storage/innobase/fil/fil0fil.c 2010-12-03 17:42:42.079064198 +0900
341 @@ -4349,7 +4349,7 @@
342 node->name, node->handle, buf,
343 offset_low, offset_high,
349 node->size += n_pages;
350 @@ -4676,7 +4676,7 @@
351 i/o on a tablespace which does not exist */
357 ulint type, /*!< in: OS_FILE_READ or OS_FILE_WRITE,
358 ORed to OS_FILE_LOG, if a log i/o
359 @@ -4701,8 +4701,9 @@
360 void* buf, /*!< in/out: buffer where to store read data
361 or from where to write; in aio this must be
362 appropriately aligned */
363 - void* message) /*!< in: message for aio handler if non-sync
364 + void* message, /*!< in: message for aio handler if non-sync
365 aio used, else ignored */
370 @@ -4872,7 +4873,7 @@
372 /* Queue the aio request */
373 ret = os_aio(type, mode | wake_later, node->name, node->handle, buf,
374 - offset_low, offset_high, len, node, message);
375 + offset_low, offset_high, len, node, message, trx);
379 diff -ruN a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc
380 --- a/storage/innobase/handler/ha_innodb.cc 2010-12-03 17:36:44.293955189 +0900
381 +++ b/storage/innobase/handler/ha_innodb.cc 2010-12-03 17:42:42.090024586 +0900
382 @@ -1528,6 +1528,16 @@
383 trx->check_unique_secondary = !thd_test_options(
384 thd, OPTION_RELAXED_UNIQUE_CHECKS);
386 +#ifdef EXTENDED_SLOWLOG
387 + if (thd_log_slow_verbosity(thd) & SLOG_V_INNODB) {
388 + trx->take_stats = TRUE;
390 + trx->take_stats = FALSE;
393 + trx->take_stats = FALSE;
399 @@ -1583,6 +1593,32 @@
403 +/*************************************************************************
404 +Gets current trx. */
409 + THD *thd=current_thd;
410 + if (likely(thd != 0)) {
411 + trx_t*& trx = thd_to_trx(thd);
420 +innobase_get_slow_log()
422 +#ifdef EXTENDED_SLOWLOG
423 + return((ibool) thd_opt_slow_log());
429 /*********************************************************************//**
430 Construct ha_innobase handler. */
432 @@ -9179,6 +9215,25 @@
433 statement has ended */
435 if (trx->n_mysql_tables_in_use == 0) {
436 +#ifdef EXTENDED_SLOWLOG
437 + increment_thd_innodb_stats(thd,
438 + (unsigned long long) trx->id,
441 + trx->io_reads_wait_timer,
442 + trx->lock_que_wait_timer,
443 + trx->innodb_que_wait_timer,
444 + trx->distinct_page_access);
448 + trx->io_reads_wait_timer = 0;
449 + trx->lock_que_wait_timer = 0;
450 + trx->innodb_que_wait_timer = 0;
451 + trx->distinct_page_access = 0;
452 + if (trx->distinct_page_access_hash)
453 + memset(trx->distinct_page_access_hash, 0, DPAH_SIZE);
456 trx->mysql_n_tables_locked = 0;
457 prebuilt->used_in_HANDLER = FALSE;
458 diff -ruN a/storage/innobase/handler/innodb_patch_info.h b/storage/innobase/handler/innodb_patch_info.h
459 --- a/storage/innobase/handler/innodb_patch_info.h 2010-12-03 17:36:44.293955189 +0900
460 +++ b/storage/innobase/handler/innodb_patch_info.h 2010-12-03 17:42:42.094955866 +0900
462 {"innodb_recovery_patches","Bugfixes and adjustments about recovery process","","http://www.percona.com/docs/wiki/percona-xtradb"},
463 {"innodb_admin_command_base","XtraDB specific command interface through i_s","","http://www.percona.com/docs/wiki/percona-xtradb"},
464 {"innodb_show_lock_name","Show mutex/lock name instead of crated file/line","","http://www.percona.com/docs/wiki/percona-xtradb"},
465 +{"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"},
466 {NULL, NULL, NULL, NULL}
468 diff -ruN a/storage/innobase/include/buf0rea.h b/storage/innobase/include/buf0rea.h
469 --- a/storage/innobase/include/buf0rea.h 2010-12-03 15:18:48.891024406 +0900
470 +++ b/storage/innobase/include/buf0rea.h 2010-12-03 17:42:42.096026873 +0900
475 +#include "trx0types.h"
476 #include "buf0types.h"
478 /********************************************************************//**
481 ulint space, /*!< in: space id */
482 ulint zip_size,/*!< in: compressed page size in bytes, or 0 */
483 - ulint offset);/*!< in: page number */
484 + ulint offset, /*!< in: page number */
486 /********************************************************************//**
487 Applies linear read-ahead if in the buf_pool the page is a border page of
488 a linear read-ahead area and all the pages in the area have been accessed.
490 /*==================*/
491 ulint space, /*!< in: space id */
492 ulint zip_size,/*!< in: compressed page size in bytes, or 0 */
493 - ulint offset);/*!< in: page number of a page; NOTE: the current thread
494 + ulint offset, /*!< in: page number of a page; NOTE: the current thread
495 must want access to this page (see NOTE 3 above) */
497 /********************************************************************//**
498 Issues read requests for pages which the ibuf module wants to read in, in
499 order to contract the insert buffer tree. Technically, this function is like
500 diff -ruN a/storage/innobase/include/fil0fil.h b/storage/innobase/include/fil0fil.h
501 --- a/storage/innobase/include/fil0fil.h 2010-12-03 15:09:51.290958543 +0900
502 +++ b/storage/innobase/include/fil0fil.h 2010-12-03 17:42:42.097027548 +0900
504 Reads or writes data. This operation is asynchronous (aio).
505 @return DB_SUCCESS, or DB_TABLESPACE_DELETED if we are trying to do
506 i/o on a tablespace which does not exist */
507 +#define fil_io(type, sync, space_id, zip_size, block_offset, byte_offset, len, buf, message) \
508 + _fil_io(type, sync, space_id, zip_size, block_offset, byte_offset, len, buf, message, NULL)
515 ulint type, /*!< in: OS_FILE_READ or OS_FILE_WRITE,
516 ORed to OS_FILE_LOG, if a log i/o
518 void* buf, /*!< in/out: buffer where to store read data
519 or from where to write; in aio this must be
520 appropriately aligned */
521 - void* message); /*!< in: message for aio handler if non-sync
522 + void* message, /*!< in: message for aio handler if non-sync
523 aio used, else ignored */
525 /**********************************************************************//**
526 Waits for an aio operation to complete. This function is used to write the
527 handler for completed requests. The aio array of pending requests is divided
528 diff -ruN a/storage/innobase/include/os0file.h b/storage/innobase/include/os0file.h
529 --- a/storage/innobase/include/os0file.h 2010-11-03 07:01:13.000000000 +0900
530 +++ b/storage/innobase/include/os0file.h 2010-12-03 17:42:42.100023783 +0900
535 +#include "trx0types.h"
539 @@ -277,13 +278,17 @@
540 pfs_os_file_close_func(file, __FILE__, __LINE__)
542 # define os_aio(type, mode, name, file, buf, offset, offset_high, \
543 - n, message1, message2) \
544 + n, message1, message2, trx) \
545 pfs_os_aio_func(type, mode, name, file, buf, offset, \
546 - offset_high, n, message1, message2, \
547 + offset_high, n, message1, message2, trx, \
550 # define os_file_read(file, buf, offset, offset_high, n) \
551 - pfs_os_file_read_func(file, buf, offset, offset_high, n, \
552 + pfs_os_file_read_func(file, buf, offset, offset_high, n, NULL, \
553 + __FILE__, __LINE__)
555 +# define os_file_read_trx(file, buf, offset, offset_high, n, trx) \
556 + pfs_os_file_read_func(file, buf, offset, offset_high, n, trx, \
559 # define os_file_read_no_error_handling(file, buf, offset, \
560 @@ -319,12 +324,15 @@
561 # define os_file_close(file) os_file_close_func(file)
563 # define os_aio(type, mode, name, file, buf, offset, offset_high, \
564 - n, message1, message2) \
565 + n, message1, message2, trx) \
566 os_aio_func(type, mode, name, file, buf, offset, offset_high, n,\
567 - message1, message2)
568 + message1, message2, trx)
570 # define os_file_read(file, buf, offset, offset_high, n) \
571 - os_file_read_func(file, buf, offset, offset_high, n)
572 + os_file_read_func(file, buf, offset, offset_high, n, NULL)
574 +# define os_file_read_trx(file, buf, offset, offset_high, n, trx) \
575 + os_file_read_func(file, buf, offset, offset_high, n, trx)
577 # define os_file_read_no_error_handling(file, buf, offset, \
580 ulint offset_high,/*!< in: most significant 32 bits of
582 ulint n, /*!< in: number of bytes to read */
584 const char* src_file,/*!< in: file name where func invoked */
585 ulint src_line);/*!< in: line where the func invoked */
588 (can be used to identify a completed
589 aio operation); ignored if mode is
592 const char* src_file,/*!< in: file name where func invoked */
593 ulint src_line);/*!< in: line where the func invoked */
594 /*******************************************************************//**
596 offset where to read */
597 ulint offset_high,/*!< in: most significant 32 bits of
599 - ulint n); /*!< in: number of bytes to read */
600 + ulint n, /*!< in: number of bytes to read */
602 /*******************************************************************//**
603 Rewind file to its start, read at most size - 1 bytes from it to str, and
604 NUL-terminate str. All errors are silently ignored. This function is
605 @@ -1044,10 +1055,11 @@
606 (can be used to identify a completed
607 aio operation); ignored if mode is
609 - void* message2);/*!< in: message for the aio handler
610 + void* message2,/*!< in: message for the aio handler
611 (can be used to identify a completed
612 aio operation); ignored if mode is
615 /************************************************************************//**
616 Wakes up all async i/o threads so that they know to exit themselves in
618 diff -ruN a/storage/innobase/include/os0file.ic b/storage/innobase/include/os0file.ic
619 --- a/storage/innobase/include/os0file.ic 2010-11-03 07:01:13.000000000 +0900
620 +++ b/storage/innobase/include/os0file.ic 2010-12-03 17:42:42.102024458 +0900
622 (can be used to identify a completed
623 aio operation); ignored if mode is
626 const char* src_file,/*!< in: file name where func invoked */
627 ulint src_line)/*!< in: line where the func invoked */
632 result = os_aio_func(type, mode, name, file, buf, offset, offset_high,
633 - n, message1, message2);
634 + n, message1, message2, trx);
636 register_pfs_file_io_end(locker, n);
639 ulint offset_high,/*!< in: most significant 32 bits of
641 ulint n, /*!< in: number of bytes to read */
643 const char* src_file,/*!< in: file name where func invoked */
644 ulint src_line)/*!< in: line where the func invoked */
647 register_pfs_file_io_begin(&state, locker, file, n, PSI_FILE_READ,
650 - result = os_file_read_func(file, buf, offset, offset_high, n);
651 + result = os_file_read_func(file, buf, offset, offset_high, n, trx);
653 register_pfs_file_io_end(locker, n);
655 diff -ruN a/storage/innobase/include/srv0srv.h b/storage/innobase/include/srv0srv.h
656 --- a/storage/innobase/include/srv0srv.h 2010-12-03 17:32:15.634987408 +0900
657 +++ b/storage/innobase/include/srv0srv.h 2010-12-03 17:42:42.104028644 +0900
659 #define SRV_AUTO_EXTEND_INCREMENT \
660 (srv_auto_extend_increment * ((1024 * 1024) / UNIV_PAGE_SIZE))
662 +/* prototypes for new functions added to ha_innodb.cc */
663 +ibool innobase_get_slow_log();
665 /* This is set to TRUE if the MySQL user has set it in MySQL */
666 extern ibool srv_lower_case_table_names;
668 diff -ruN a/storage/innobase/include/trx0trx.h b/storage/innobase/include/trx0trx.h
669 --- a/storage/innobase/include/trx0trx.h 2010-12-03 15:41:52.049372966 +0900
670 +++ b/storage/innobase/include/trx0trx.h 2010-12-03 17:42:42.107024532 +0900
672 /*------------------------------*/
673 char detailed_error[256]; /*!< detailed error message for last
675 + /*------------------------------*/
677 + ib_uint64_t io_read;
678 + ulint io_reads_wait_timer;
679 + ib_uint64_t lock_que_wait_ustarted;
680 + ulint lock_que_wait_timer;
681 + ulint innodb_que_wait_timer;
682 + ulint distinct_page_access;
683 +#define DPAH_SIZE 8192
684 + byte* distinct_page_access_hash;
688 #define TRX_MAX_N_THREADS 32 /* maximum number of
689 diff -ruN a/storage/innobase/lock/lock0lock.c b/storage/innobase/lock/lock0lock.c
690 --- a/storage/innobase/lock/lock0lock.c 2010-12-03 15:09:51.297986437 +0900
691 +++ b/storage/innobase/lock/lock0lock.c 2010-12-03 17:42:42.111024587 +0900
692 @@ -1755,6 +1755,8 @@
699 ut_ad(mutex_own(&kernel_mutex));
701 @@ -1813,6 +1815,10 @@
702 trx->que_state = TRX_QUE_LOCK_WAIT;
703 trx->was_chosen_as_deadlock_victim = FALSE;
704 trx->wait_started = time(NULL);
705 + if (innobase_get_slow_log() && trx->take_stats) {
706 + ut_usectime(&sec, &ms);
707 + trx->lock_que_wait_ustarted = (ib_uint64_t)sec * 1000000 + ms;
710 ut_a(que_thr_stop(thr));
712 @@ -3692,6 +3698,8 @@
719 ut_ad(mutex_own(&kernel_mutex));
721 @@ -3747,6 +3755,10 @@
725 + if (innobase_get_slow_log() && trx->take_stats) {
726 + ut_usectime(&sec, &ms);
727 + trx->lock_que_wait_ustarted = (ib_uint64_t)sec * 1000000 + ms;
729 trx->que_state = TRX_QUE_LOCK_WAIT;
730 trx->was_chosen_as_deadlock_victim = FALSE;
731 trx->wait_started = time(NULL);
732 diff -ruN a/storage/innobase/os/os0file.c b/storage/innobase/os/os0file.c
733 --- a/storage/innobase/os/os0file.c 2010-12-03 17:32:15.644024974 +0900
734 +++ b/storage/innobase/os/os0file.c 2010-12-03 17:42:42.117023467 +0900
736 #include "srv0start.h"
739 +#include "trx0sys.h"
740 +#include "trx0trx.h"
741 #include "log0recv.h"
742 #ifndef UNIV_HOTBACKUP
743 # include "os0sync.h"
744 @@ -2175,13 +2177,18 @@
745 ulint n, /*!< in: number of bytes to read */
746 ulint offset, /*!< in: least significant 32 bits of file
747 offset from where to read */
748 - ulint offset_high) /*!< in: most significant 32 bits of
749 + ulint offset_high, /*!< in: most significant 32 bits of
754 #if defined(HAVE_PREAD) && !defined(HAVE_BROKEN_PREAD)
756 #endif /* HAVE_PREAD && !HAVE_BROKEN_PREAD */
759 + ib_uint64_t start_time;
760 + ib_uint64_t finish_time;
762 ut_a((offset & 0xFFFFFFFFUL) == offset);
764 @@ -2202,6 +2209,15 @@
768 + if (innobase_get_slow_log() && trx && trx->take_stats)
772 + ut_usectime(&sec, &ms);
773 + start_time = (ib_uint64_t)sec * 1000000 + ms;
777 #if defined(HAVE_PREAD) && !defined(HAVE_BROKEN_PREAD)
778 os_mutex_enter(os_file_count_mutex);
779 os_file_n_pending_preads++;
780 @@ -2215,6 +2231,13 @@
781 os_n_pending_reads--;
782 os_mutex_exit(os_file_count_mutex);
784 + if (innobase_get_slow_log() && trx && trx->take_stats && start_time)
786 + ut_usectime(&sec, &ms);
787 + finish_time = (ib_uint64_t)sec * 1000000 + ms;
788 + trx->io_reads_wait_timer += (ulint)(finish_time - start_time);
794 @@ -2251,6 +2274,13 @@
795 os_n_pending_reads--;
796 os_mutex_exit(os_file_count_mutex);
798 + if (innobase_get_slow_log() && trx && trx->take_stats && start_time)
800 + ut_usectime(&sec, &ms);
801 + finish_time = (ib_uint64_t)sec * 1000000 + ms;
802 + trx->io_reads_wait_timer += (ulint)(finish_time - start_time);
808 @@ -2391,7 +2421,8 @@
809 offset where to read */
810 ulint offset_high, /*!< in: most significant 32 bits of
812 - ulint n) /*!< in: number of bytes to read */
813 + ulint n, /*!< in: number of bytes to read */
818 @@ -2463,7 +2494,7 @@
819 os_bytes_read_since_printout += n;
822 - ret = os_file_pread(file, buf, n, offset, offset_high);
823 + ret = os_file_pread(file, buf, n, offset, offset_high, trx);
825 if ((ulint)ret == n) {
827 @@ -2589,7 +2620,7 @@
828 os_bytes_read_since_printout += n;
831 - ret = os_file_pread(file, buf, n, offset, offset_high);
832 + ret = os_file_pread(file, buf, n, offset, offset_high, NULL);
834 if ((ulint)ret == n) {
836 @@ -3608,7 +3639,8 @@
838 ulint offset_high, /*!< in: most significant 32 bits of
840 - ulint len) /*!< in: length of the block to read or write */
841 + ulint len, /*!< in: length of the block to read or write */
844 os_aio_slot_t* slot = NULL;
846 @@ -3976,10 +4008,11 @@
847 (can be used to identify a completed
848 aio operation); ignored if mode is
850 - void* message2)/*!< in: message for the aio handler
851 + void* message2,/*!< in: message for the aio handler
852 (can be used to identify a completed
853 aio operation); ignored if mode is
857 os_aio_array_t* array;
859 @@ -4017,8 +4050,8 @@
860 wait in the Windows case. */
862 if (type == OS_FILE_READ) {
863 - return(os_file_read(file, buf, offset,
865 + return(os_file_read_trx(file, buf, offset,
866 + offset_high, n, trx));
869 ut_a(type == OS_FILE_WRITE);
870 @@ -4056,8 +4089,13 @@
874 + if (trx && type == OS_FILE_READ)
879 slot = os_aio_array_reserve_slot(type, array, message1, message2, file,
880 - name, buf, offset, offset_high, n);
881 + name, buf, offset, offset_high, n, trx);
882 if (type == OS_FILE_READ) {
883 if (srv_use_native_aio) {
885 diff -ruN a/storage/innobase/srv/srv0srv.c b/storage/innobase/srv/srv0srv.c
886 --- a/storage/innobase/srv/srv0srv.c 2010-12-03 17:32:15.648024399 +0900
887 +++ b/storage/innobase/srv/srv0srv.c 2010-12-03 17:45:05.067023254 +0900
889 #include "mysql/plugin.h"
890 #include "mysql/service_thd_wait.h"
892 +/* prototypes for new functions added to ha_innodb.cc */
893 +ibool innobase_get_slow_log();
895 /* This is set to TRUE if the MySQL user has set it in MySQL; currently
896 affects only FOREIGN KEY definition parsing */
897 UNIV_INTERN ibool srv_lower_case_table_names = FALSE;
898 @@ -1204,6 +1207,10 @@
899 ibool has_slept = FALSE;
900 srv_conc_slot_t* slot = NULL;
902 + ib_uint64_t start_time = 0L;
903 + ib_uint64_t finish_time = 0L;
907 if (trx->mysql_thd != NULL
908 && thd_is_replication_slave_thread(trx->mysql_thd)) {
909 @@ -1280,6 +1287,7 @@
911 if (SRV_THREAD_SLEEP_DELAY > 0) {
912 os_thread_sleep(SRV_THREAD_SLEEP_DELAY);
913 + trx->innodb_que_wait_timer += SRV_THREAD_SLEEP_DELAY;
917 @@ -1335,6 +1343,13 @@
918 /* Go to wait for the event; when a thread leaves InnoDB it will
919 release this thread */
921 + if (innobase_get_slow_log() && trx->take_stats) {
922 + ut_usectime(&sec, &ms);
923 + start_time = (ib_uint64_t)sec * 1000000 + ms;
928 trx->op_info = "waiting in InnoDB queue";
930 thd_wait_begin(trx->mysql_thd, THD_WAIT_ROW_TABLE_LOCK);
931 @@ -1343,6 +1358,12 @@
935 + if (innobase_get_slow_log() && trx->take_stats && start_time) {
936 + ut_usectime(&sec, &ms);
937 + finish_time = (ib_uint64_t)sec * 1000000 + ms;
938 + trx->innodb_que_wait_timer += (ulint)(finish_time - start_time);
941 os_fast_mutex_lock(&srv_conc_mutex);
943 srv_conc_n_waiting_threads--;
944 diff -ruN a/storage/innobase/trx/trx0trx.c b/storage/innobase/trx/trx0trx.c
945 --- a/storage/innobase/trx/trx0trx.c 2010-12-03 15:41:52.053955669 +0900
946 +++ b/storage/innobase/trx/trx0trx.c 2010-12-03 17:42:42.127023410 +0900
948 trx->global_read_view = NULL;
949 trx->read_view = NULL;
953 + trx->io_reads_wait_timer = 0;
954 + trx->lock_que_wait_timer = 0;
955 + trx->innodb_que_wait_timer = 0;
956 + trx->distinct_page_access = 0;
957 + trx->distinct_page_access_hash = NULL;
958 + trx->take_stats = FALSE;
960 /* Set X/Open XA transaction identification to NULL */
961 memset(&trx->xid, 0, sizeof(trx->xid));
962 trx->xid.formatID = -1;
965 trx->mysql_process_no = os_proc_get_number();
967 + if (innobase_get_slow_log() && trx->take_stats) {
968 + trx->distinct_page_access_hash = mem_alloc(DPAH_SIZE);
969 + memset(trx->distinct_page_access_hash, 0, DPAH_SIZE);
977 trx_t* trx) /*!< in, own: trx object */
979 + if (trx->distinct_page_access_hash)
981 + mem_free(trx->distinct_page_access_hash);
982 + trx->distinct_page_access_hash= NULL;
985 mutex_enter(&kernel_mutex);
987 UT_LIST_REMOVE(mysql_trx_list, trx_sys->mysql_trx_list, trx);
989 /*====================*/
990 trx_t* trx) /*!< in, own: trx object */
992 + if (trx->distinct_page_access_hash)
994 + mem_free(trx->distinct_page_access_hash);
995 + trx->distinct_page_access_hash= NULL;
998 mutex_enter(&kernel_mutex);
1001 @@ -1094,6 +1120,9 @@
1002 trx_t* trx) /*!< in: transaction */
1009 ut_ad(mutex_own(&kernel_mutex));
1010 ut_ad(trx->que_state == TRX_QUE_LOCK_WAIT);
1011 @@ -1108,6 +1137,11 @@
1012 thr = UT_LIST_GET_FIRST(trx->wait_thrs);
1015 + if (innobase_get_slow_log() && trx->take_stats) {
1016 + ut_usectime(&sec, &ms);
1017 + now = (ib_uint64_t)sec * 1000000 + ms;
1018 + trx->lock_que_wait_timer += (ulint)(now - trx->lock_que_wait_ustarted);
1020 trx->que_state = TRX_QUE_RUNNING;
1023 @@ -1121,6 +1155,9 @@
1024 trx_t* trx) /*!< in: transaction in the TRX_QUE_LOCK_WAIT state */
1031 ut_ad(mutex_own(&kernel_mutex));
1032 ut_ad(trx->que_state == TRX_QUE_LOCK_WAIT);
1033 @@ -1135,6 +1172,11 @@
1034 thr = UT_LIST_GET_FIRST(trx->wait_thrs);
1037 + if (innobase_get_slow_log() && trx->take_stats) {
1038 + ut_usectime(&sec, &ms);
1039 + now = (ib_uint64_t)sec * 1000000 + ms;
1040 + trx->lock_que_wait_timer += (ulint)(now - trx->lock_que_wait_ustarted);
1042 trx->que_state = TRX_QUE_RUNNING;