This ChangeSet must be null-merged to 5.1. Applied innodb-5.0-ss982, -ss998, -ss1003

Fixes:
- Bug #15815: Very poor performance with multiple queries running concurrently
- Bug #22868: 'Thread thrashing' with > 50 concurrent conns under an upd-intensive workloadw
- Bug #23769: Debug assertion failure with innodb_locks_unsafe_for_binlog
- Bug #24089: Race condition in fil_flush_file_spaces()


innobase/buf/buf0buf.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1000:
  branches/5.0: Merge r999 from trunk:
  
  Reduce buffer pool mutex contention under >= 4 big concurrent
  CPU-bound SELECT queries.  (Bug #22868)
  
  Fix: replace the mutex by one mutex protecting the 'flush list'
  (and the free list) and several mutexes protecting portions of the
  buffer pool, where we keep several indivudual LRU lists of pages.
  
  This patch is from Sunny Bains and Heikki Tuuri.
innobase/buf/buf0flu.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1000:
  branches/5.0: Merge r999 from trunk:
  
  Reduce buffer pool mutex contention under >= 4 big concurrent
  CPU-bound SELECT queries.  (Bug #22868)
  
  Fix: replace the mutex by one mutex protecting the 'flush list'
  (and the free list) and several mutexes protecting portions of the
  buffer pool, where we keep several indivudual LRU lists of pages.
  
  This patch is from Sunny Bains and Heikki Tuuri.
innobase/buf/buf0lru.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1000:
  branches/5.0: Merge r999 from trunk:
  
  Reduce buffer pool mutex contention under >= 4 big concurrent
  CPU-bound SELECT queries.  (Bug #22868)
  
  Fix: replace the mutex by one mutex protecting the 'flush list'
  (and the free list) and several mutexes protecting portions of the
  buffer pool, where we keep several indivudual LRU lists of pages.
  
  This patch is from Sunny Bains and Heikki Tuuri.
innobase/dict/dict0crea.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r974:
  branches/5.0: Port r973 from trunk.
  
  Do not break the latching order in TRUNCATE TABLE.
  
  dict_truncate_index_tree(): Replace parameter rec_t* rec with
  btr_pcur_t* pcur.  Reposition pcur before calling btr_create().
  
  sync_thread_add_level(): Remove the relaxation of the assertion added in r968.
innobase/fil/fil0fil.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1003:
  branches/5.0: Merge r1002 from trunk:
  
  fil_flush_file_spaces(): Copy the system->unflushed_spaces list to an
  array while holding the mutex.  This removes the crash-triggering
  race condition that was introduced when fixing Bug 15653.  (Bug #24089)
innobase/include/buf0buf.h:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1000:
  branches/5.0: Merge r999 from trunk:
  
  Reduce buffer pool mutex contention under >= 4 big concurrent
  CPU-bound SELECT queries.  (Bug #22868)
  
  Fix: replace the mutex by one mutex protecting the 'flush list'
  (and the free list) and several mutexes protecting portions of the
  buffer pool, where we keep several indivudual LRU lists of pages.
  
  This patch is from Sunny Bains and Heikki Tuuri.
innobase/include/buf0buf.ic:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1000:
  branches/5.0: Merge r999 from trunk:
  
  Reduce buffer pool mutex contention under >= 4 big concurrent
  CPU-bound SELECT queries.  (Bug #22868)
  
  Fix: replace the mutex by one mutex protecting the 'flush list'
  (and the free list) and several mutexes protecting portions of the
  buffer pool, where we keep several indivudual LRU lists of pages.
  
  This patch is from Sunny Bains and Heikki Tuuri.
innobase/include/dict0crea.h:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r974:
  branches/5.0: Port r973 from trunk.
  
  Do not break the latching order in TRUNCATE TABLE.
  
  dict_truncate_index_tree(): Replace parameter rec_t* rec with
  btr_pcur_t* pcur.  Reposition pcur before calling btr_create().
  
  sync_thread_add_level(): Remove the relaxation of the assertion added in r968.
innobase/include/sync0arr.h:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1001:
  branches/5.0: Reduce locking contention:
  
  Bug #15815: 'Thread thrashing' with > 50 concurrent connections under
  an update-intensive workload.
  Fix: Introduce one event per InnoDB semaphore.
  
  This patch is from Sunny Bains and Heikki Tuuri.
  This patch will not be merged to trunk (MySQL/InnoDB 5.1) yet,
  because it tries to address the problem in a different way.
innobase/include/sync0rw.h:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1001:
  branches/5.0: Reduce locking contention:
  
  Bug #15815: 'Thread thrashing' with > 50 concurrent connections under
  an update-intensive workload.
  Fix: Introduce one event per InnoDB semaphore.
  
  This patch is from Sunny Bains and Heikki Tuuri.
  This patch will not be merged to trunk (MySQL/InnoDB 5.1) yet,
  because it tries to address the problem in a different way.
innobase/include/sync0rw.ic:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1001:
  branches/5.0: Reduce locking contention:
  
  Bug #15815: 'Thread thrashing' with > 50 concurrent connections under
  an update-intensive workload.
  Fix: Introduce one event per InnoDB semaphore.
  
  This patch is from Sunny Bains and Heikki Tuuri.
  This patch will not be merged to trunk (MySQL/InnoDB 5.1) yet,
  because it tries to address the problem in a different way.
innobase/include/sync0sync.h:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1001:
  branches/5.0: Reduce locking contention:
  
  Bug #15815: 'Thread thrashing' with > 50 concurrent connections under
  an update-intensive workload.
  Fix: Introduce one event per InnoDB semaphore.
  
  This patch is from Sunny Bains and Heikki Tuuri.
  This patch will not be merged to trunk (MySQL/InnoDB 5.1) yet,
  because it tries to address the problem in a different way.
innobase/os/os0sync.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1001:
  branches/5.0: Reduce locking contention:
  
  Bug #15815: 'Thread thrashing' with > 50 concurrent connections under
  an update-intensive workload.
  Fix: Introduce one event per InnoDB semaphore.
  
  This patch is from Sunny Bains and Heikki Tuuri.
  This patch will not be merged to trunk (MySQL/InnoDB 5.1) yet,
  because it tries to address the problem in a different way.
innobase/row/row0mysql.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r974:
  branches/5.0: Port r973 from trunk.
  
  Do not break the latching order in TRUNCATE TABLE.
  
  dict_truncate_index_tree(): Replace parameter rec_t* rec with
  btr_pcur_t* pcur.  Reposition pcur before calling btr_create().
  
  sync_thread_add_level(): Remove the relaxation of the assertion added in r968.
innobase/row/row0sel.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r982:
  branches/5.0: row_sel(): Do not try to acquire a LOCK_REC_NOT_GAP lock
  on the supremum record.  Instead, skip to the next record.  (Bug #23769)
  This fix was backported from r623 in the 5.1 tree.
innobase/srv/srv0start.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r926:
  Refer to bug: 22268. Since no one tries to run 5.0 on Windows 95/ME it was
  decided to raise the limit of srv_max_n_threads to 10000 on Windows.
innobase/sync/sync0arr.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1001:
  branches/5.0: Reduce locking contention:
  
  Bug #15815: 'Thread thrashing' with > 50 concurrent connections under
  an update-intensive workload.
  Fix: Introduce one event per InnoDB semaphore.
  
  This patch is from Sunny Bains and Heikki Tuuri.
  This patch will not be merged to trunk (MySQL/InnoDB 5.1) yet,
  because it tries to address the problem in a different way.
innobase/sync/sync0rw.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r1001:
  branches/5.0: Reduce locking contention:
  
  Bug #15815: 'Thread thrashing' with > 50 concurrent connections under
  an update-intensive workload.
  Fix: Introduce one event per InnoDB semaphore.
  
  This patch is from Sunny Bains and Heikki Tuuri.
  This patch will not be merged to trunk (MySQL/InnoDB 5.1) yet,
  because it tries to address the problem in a different way.
innobase/sync/sync0sync.c:
  Applied innodb-5.0-ss982, -ss998, -ss1003
  
  Revision r969:
  branches/5.0: Port r968 from trunk:
  
  sync_thread_add_level(): When level == SYNC_TREE_NODE, allow the latching
  order to be violated if the thread holds dict_operation_lock, whose level is
  SYNC_DICT_OPERATION.  This removes the assertion failure of TRUNCATE TABLE
  #ifdef UNIV_SYNC_DEBUG.
  
  
  Revision r974:
  branches/5.0: Port r973 from trunk.
  
  Do not break the latching order in TRUNCATE TABLE.
  
  dict_truncate_index_tree(): Replace parameter rec_t* rec with
  btr_pcur_t* pcur.  Reposition pcur before calling btr_create().
  
  sync_thread_add_level(): Remove the relaxation of the assertion added in r968.
  
  
  Revision r1001:
  branches/5.0: Reduce locking contention:
  
  Bug #15815: 'Thread thrashing' with > 50 concurrent connections under
  an update-intensive workload.
  Fix: Introduce one event per InnoDB semaphore.
  
  This patch is from Sunny Bains and Heikki Tuuri.
  This patch will not be merged to trunk (MySQL/InnoDB 5.1) yet,
  because it tries to address the problem in a different way.

3 files changed, 49 insertions, 80 deletions

diff --git a/innobase/sync/sync0arr.c b/innobase/sync/sync0arr.c
index 198ef49ca9f..64f9310bad3 100644
--- a/innobase/sync/sync0arr.c
+++ b/innobase/sync/sync0arr.c
@@ -62,9 +62,6 @@ struct sync_cell_struct {
 	ibool		waiting;	/* TRUE if the thread has already
 					called sync_array_event_wait
 					on this cell */
-	ibool		event_set;	/* TRUE if the event is set */
-        os_event_t 	event;   	/* operating system event
-                                        semaphore handle */
 	time_t		reservation_time;/* time when the thread reserved
 					the wait cell */
 };
@@ -218,10 +215,7 @@ sync_array_create(
         for (i = 0; i < n_cells; i++) {
 		cell = sync_array_get_nth_cell(arr, i);        	
                 cell->wait_object = NULL;
-
-                /* Create an operating system event semaphore with no name */
-                cell->event = os_event_create(NULL);
-		cell->event_set = FALSE; /* it is created in reset state */
+		cell->waiting = FALSE;
 	}
 
 	return(arr);
@@ -235,19 +229,12 @@ sync_array_free(
 /*============*/
 	sync_array_t*	arr)	/* in, own: sync wait array */
 {
-        ulint           i;
-        sync_cell_t*   	cell;
 	ulint		protection;
 
         ut_a(arr->n_reserved == 0);
         
 	sync_array_validate(arr);
         
-        for (i = 0; i < arr->n_cells; i++) {
-		cell = sync_array_get_nth_cell(arr, i);        	
-		os_event_free(cell->event);
-        }
-
 	protection = arr->protection;
 
         /* Release the mutex protecting the wait array complex */
@@ -293,27 +280,19 @@ sync_array_validate(
 }
 
 /***********************************************************************
-Puts the cell event in set state. */
-static
-void
-sync_cell_event_set(
-/*================*/
-	sync_cell_t*	cell)	/* in: array cell */
-{
-	os_event_set(cell->event);
-	cell->event_set = TRUE;
-}		
-
-/***********************************************************************
 Puts the cell event in reset state. */
 static
 void
 sync_cell_event_reset(
 /*==================*/
-	sync_cell_t*	cell)	/* in: array cell */
+	ulint		type,	/* in: lock type mutex/rw_lock */
+	void*		object) /* in: the rw_lock/mutex object */
 {
-	os_event_reset(cell->event);
-	cell->event_set = FALSE;
+	if (type == SYNC_MUTEX) {
+		os_event_reset(((mutex_t *) object)->event);
+	} else {
+		os_event_reset(((rw_lock_t *) object)->event);
+	}
 }		
 
 /**********************************************************************
@@ -346,14 +325,7 @@ sync_array_reserve_cell(
 
                 if (cell->wait_object == NULL) {
 
-                        /* Make sure the event is reset */
-			if (cell->event_set) {	
-                        	sync_cell_event_reset(cell);
-			}
-
-			cell->reservation_time = time(NULL);
-			cell->thread = os_thread_get_curr_id();
-
+			cell->waiting = FALSE;
 			cell->wait_object = object;
 
 			if (type == SYNC_MUTEX) {
@@ -363,7 +335,6 @@ sync_array_reserve_cell(
 			}
 				
 			cell->request_type = type;
-			cell->waiting = FALSE;
 			
 			cell->file = file;
 			cell->line = line;
@@ -373,6 +344,13 @@ sync_array_reserve_cell(
 			*index = i;
 
 			sync_array_exit(arr);
+
+			/* Make sure the event is reset */
+			sync_cell_event_reset(type, object);
+
+			cell->reservation_time = time(NULL);
+
+			cell->thread = os_thread_get_curr_id();
                         
                         return;
                 }
@@ -408,7 +386,12 @@ sync_array_wait_event(
 	ut_a(!cell->waiting);
 	ut_ad(os_thread_get_curr_id() == cell->thread);
 
-       	event = cell->event;
+	if (cell->request_type == SYNC_MUTEX) {
+		event = ((mutex_t*) cell->wait_object)->event;
+	} else {
+		event = ((rw_lock_t*) cell->wait_object)->event;
+	}
+
        	cell->waiting = TRUE;
 
 #ifdef UNIV_SYNC_DEBUG
@@ -510,10 +493,6 @@ sync_array_cell_print(
         if (!cell->waiting) {
 		fputs("wait has ended\n", file);
 	}
-
-        if (cell->event_set) {
-		fputs("wait is ending\n", file);
-	}
 }
 
 #ifdef UNIV_SYNC_DEBUG
@@ -623,7 +602,7 @@ sync_array_detect_deadlock(
 	
 	depth++;
 	
-	if (cell->event_set || !cell->waiting) {
+	if (!cell->waiting) {
 
 		return(FALSE); /* No deadlock here */
 	}
@@ -802,6 +781,7 @@ sync_array_free_cell(
 
         ut_a(cell->wait_object != NULL);
 
+	cell->waiting = FALSE;
 	cell->wait_object =  NULL;
 
 	ut_a(arr->n_reserved > 0);
@@ -811,44 +791,17 @@ sync_array_free_cell(
 }
 
 /**************************************************************************
-Looks for the cells in the wait array which refer to the wait object
-specified, and sets their corresponding events to the signaled state. In this
-way releases the threads waiting for the object to contend for the object.
-It is possible that no such cell is found, in which case does nothing. */
+Increments the signalled count. */
 
 void
-sync_array_signal_object(
-/*=====================*/
-	sync_array_t*	arr,	/* in: wait array */
-	void*		object)	/* in: wait object */
+sync_array_object_signalled(
+/*========================*/
+	sync_array_t*	arr)	/* in: wait array */
 {
-        sync_cell_t*   	cell;
-        ulint           count;
-        ulint           i;
-
         sync_array_enter(arr);
 
 	arr->sg_count++;
 
-	i = 0;
-	count = 0;
-
-        while (count < arr->n_reserved) {
-
-        	cell = sync_array_get_nth_cell(arr, i);
-
-                if (cell->wait_object != NULL) {
-
-                        count++;
-                        if (cell->wait_object == object) {
-
-                        	sync_cell_event_set(cell);
-                        }
-                }
-
-                i++;
-        }
-
         sync_array_exit(arr);
 }
 
@@ -881,7 +834,17 @@ sync_arr_wake_threads_if_sema_free(void)
 
                         if (sync_arr_cell_can_wake_up(cell)) {
 
-                        	sync_cell_event_set(cell);
+				if (cell->request_type == SYNC_MUTEX) {
+					mutex_t*	mutex;
+
+					mutex = cell->wait_object;
+					os_event_set(mutex->event);
+				} else {
+					rw_lock_t*	lock;
+
+					lock = cell->wait_object;
+					os_event_set(lock->event);
+				}
                         }
                 }
 
@@ -911,7 +874,7 @@ sync_array_print_long_waits(void)
 
         	cell = sync_array_get_nth_cell(sync_primary_wait_array, i);
 
-                if (cell->wait_object != NULL
+		if (cell->wait_object != NULL && cell->waiting
 		    && difftime(time(NULL), cell->reservation_time) > 240) {
 			fputs("InnoDB: Warning: a long semaphore wait:\n",
 				stderr);
@@ -919,7 +882,7 @@ sync_array_print_long_waits(void)
 			noticed = TRUE;
                 }
 
-                if (cell->wait_object != NULL
+		if (cell->wait_object != NULL && cell->waiting
 		    && difftime(time(NULL), cell->reservation_time)
 		    > fatal_timeout) {
 			fatal = TRUE;
diff --git a/innobase/sync/sync0rw.c b/innobase/sync/sync0rw.c
index 973b46fdd50..050de73db9e 100644
--- a/innobase/sync/sync0rw.c
+++ b/innobase/sync/sync0rw.c
@@ -128,6 +128,7 @@ rw_lock_create_func(
 	lock->last_x_file_name = "not yet reserved";
 	lock->last_s_line = 0;
 	lock->last_x_line = 0;
+	lock->event = os_event_create(NULL);
 
 	mutex_enter(&rw_lock_list_mutex);
 	
@@ -163,6 +164,7 @@ rw_lock_free(
 	mutex_free(rw_lock_get_mutex(lock));
 
 	mutex_enter(&rw_lock_list_mutex);
+	os_event_free(lock->event);
 
 	if (UT_LIST_GET_PREV(list, lock)) {
 		ut_a(UT_LIST_GET_PREV(list, lock)->magic_n == RW_LOCK_MAGIC_N);
diff --git a/innobase/sync/sync0sync.c b/innobase/sync/sync0sync.c
index 43249f4b96f..95bf83dce79 100644
--- a/innobase/sync/sync0sync.c
+++ b/innobase/sync/sync0sync.c
@@ -212,6 +212,7 @@ mutex_create_func(
 	os_fast_mutex_init(&(mutex->os_fast_mutex));
 	mutex->lock_word = 0;
 #endif
+	mutex->event = os_event_create(NULL);
 	mutex_set_waiters(mutex, 0);
 	mutex->magic_n = MUTEX_MAGIC_N;
 #ifdef UNIV_SYNC_DEBUG
@@ -288,6 +289,8 @@ mutex_free(
 		mutex_exit(&mutex_list_mutex);
 	}
 
+	os_event_free(mutex->event);
+
 #if !defined(_WIN32) || !defined(UNIV_CAN_USE_X86_ASSEMBLER) 
 	os_fast_mutex_free(&(mutex->os_fast_mutex));
 #endif
@@ -564,8 +567,8 @@ mutex_signal_object(
 
 	/* The memory order of resetting the waiters field and
 	signaling the object is important. See LEMMA 1 above. */
-
-	sync_array_signal_object(sync_primary_wait_array, mutex);
+	os_event_set(mutex->event);
+	sync_array_object_signalled(sync_primary_wait_array);
 }
 
 #ifdef UNIV_SYNC_DEBUG
@@ -1114,6 +1117,7 @@ sync_thread_add_level(
 		ut_a(sync_thread_levels_g(array, SYNC_PURGE_SYS));
 	} else if (level == SYNC_TREE_NODE) {
 		ut_a(sync_thread_levels_contain(array, SYNC_INDEX_TREE)
+		     || sync_thread_levels_contain(array, SYNC_DICT_OPERATION)
 		     || sync_thread_levels_g(array, SYNC_TREE_NODE - 1));
 	} else if (level == SYNC_TREE_NODE_FROM_HASH) {
 		ut_a(1);