Apply InnoDB snapshot innodb-5.1-ss1751.

Bug #16979: AUTO_INC lock in InnoDB works a table level lock
  Add a table level counter that tracks the number of AUTOINC locks that are
  pending and/or granted on a table. We peek at this value to determine whether
  a transaction doing a simple INSERT in innodb_autoinc_lock_mode = 1, needs to
  acquire the AUTOINC lock or not. This change is related to Bug# 16979.
Bug #27950: Duplicate entry error in auto-inc after mysqld restart
  We check whether the AUTOINC sub-system has been initialized (first) by
  holding the AUTOINC mutex and if initialization is required then we
  initialize using our normal procedure.


storage/innobase/dict/dict0boot.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
storage/innobase/dict/dict0crea.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
storage/innobase/dict/dict0load.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
storage/innobase/dict/dict0mem.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1749:
  Add a table level counter that tracks the number of AUTOINC locks that are
  pending and/or granted on a table. We peek at this value to determine whether
  a transaction doing a simple INSERT in innodb_autoinc_lock_mode = 1, needs to
  acquire the AUTOINC lock or not. This change is related to Bug# 16979.
storage/innobase/handler/ha_innodb.cc:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1735:
  Add variable "innodb_autoinc_lock_mode"  to control the behavior of the
  AUTOINC locking. There are three modes, 0 for backward compatibility, 1 for the
  new style locking (default, safe for statement-based replication) and
  2 for no AUTOINC locking (unsafe for statement-based replication).
  
  
  Revision r1750:
  We check whether the AUTOINC sub-system has been initialized (first) by
  holding the AUTOINC mutex and if initialization is required then we
  initialize using our normal procedure. This change is related to Bug#27950.
  
  
  Revision r1746:
  Remove some code from ha_innodb.cc that was already enclosed in #if 0.
  The code was related to replication and group commit, and it was
  unreachable already as of MySQL 5.0.
  
  Approved by Heikki
  
  
  Revision r1749:
  Add a table level counter that tracks the number of AUTOINC locks that are
  pending and/or granted on a table. We peek at this value to determine whether
  a transaction doing a simple INSERT in innodb_autoinc_lock_mode = 1, needs to
  acquire the AUTOINC lock or not. This change is related to Bug# 16979.
storage/innobase/include/dict0mem.h:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1749:
  Add a table level counter that tracks the number of AUTOINC locks that are
  pending and/or granted on a table. We peek at this value to determine whether
  a transaction doing a simple INSERT in innodb_autoinc_lock_mode = 1, needs to
  acquire the AUTOINC lock or not. This change is related to Bug# 16979.
  
  
  
  Revision r1747:
  Correct an outdated comment about dict_table_t:col_names that should have
  been adjusted in r1719, or r1264 in branches/zip.
storage/innobase/include/mem0mem.ic:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1744:
  Merge r1739 from branches/zip:
  
  mem_heap_free_heap_top(): Declare the memory freed with UNIV_MEM_FREE().
  Before this change, the memory freed by mem_heap_empty() was not completely
  flagged free in UNIV_DEBUG_VALGRIND builds.  After this change, Valgrind
  will hopefully catch all errors caught by UNIV_MEM_DEBUG.
storage/innobase/include/sync0rw.ic:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1751:
  rw_lock_s_lock_func(): Correct a typo in a comment.
storage/innobase/include/trx0trx.h:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1742:
  Remove some unused fields of trx_t: mysql_master_log_file_name,
  mysql_master_log_pos, repl_wait_binlog_name, repl_wait_binlog_pos.
  
  Approved by Heikki.
storage/innobase/lock/lock0lock.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1749:
  Add a table level counter that tracks the number of AUTOINC locks that are
  pending and/or granted on a table. We peek at this value to determine whether
  a transaction doing a simple INSERT in innodb_autoinc_lock_mode = 1, needs to
  acquire the AUTOINC lock or not. This change is related to Bug# 16979.
storage/innobase/mtr/mtr0log.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
storage/innobase/pars/pars0pars.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
storage/innobase/srv/srv0srv.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
storage/innobase/trx/trx0sys.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1743:
  trx_sys_print_mysql_binlog_offset_from_page(): Enclose the definition
  in #ifdef UNIV_HOTBACKUP, to match the function declaration.
storage/innobase/trx/trx0trx.c:
  Apply InnoDB snapshot innodb-5.1-ss1751.
  
  Revision r1742:
  Remove some unused fields of trx_t: mysql_master_log_file_name,
  mysql_master_log_pos, repl_wait_binlog_name, repl_wait_binlog_pos.
  
  Approved by Heikki.

storage/innobase/dict/dict0boot.c

@@ -211,6 +211,7 @@ dict_boot(void)
 	dict_table_t*	table;
 	dict_index_t*	index;
 	dict_hdr_t*	dict_hdr;
+	mem_heap_t*	heap;
 	mtr_t		mtr;
 
 	mtr_start(&mtr);
@@ -218,6 +219,8 @@ dict_boot(void)
 	/* Create the hash tables etc. */
 	dict_init();
 
+	heap = mem_heap_create(450);
+
 	mutex_enter(&(dict_sys->mutex));
 
 	/* Get the dictionary header */
@@ -244,19 +247,20 @@ dict_boot(void)
 	/*-------------------------*/
 	table = dict_mem_table_create("SYS_TABLES", DICT_HDR_SPACE, 8, 0);
 
-	dict_mem_table_add_col(table, "NAME", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "ID", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "N_COLS", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "TYPE", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "MIX_ID", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "MIX_LEN", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "CLUSTER_NAME", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "SPACE", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "NAME", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "ID", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "N_COLS", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "TYPE", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "MIX_ID", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "MIX_LEN", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "CLUSTER_NAME", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "SPACE", DATA_INT, 0, 4);
 
 	table->id = DICT_TABLES_ID;
 
-	dict_table_add_to_cache(table);
+	dict_table_add_to_cache(table, heap);
 	dict_sys->sys_tables = table;
+	mem_heap_empty(heap);
 
 	index = dict_mem_index_create("SYS_TABLES", "CLUST_IND",
 				      DICT_HDR_SPACE,
@@ -283,18 +287,19 @@ dict_boot(void)
 	/*-------------------------*/
 	table = dict_mem_table_create("SYS_COLUMNS", DICT_HDR_SPACE, 7, 0);
 
-	dict_mem_table_add_col(table, "TABLE_ID", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "POS", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "NAME", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "MTYPE", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "PRTYPE", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "LEN", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "PREC", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "TABLE_ID", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "POS", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "NAME", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "MTYPE", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "PRTYPE", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "LEN", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "PREC", DATA_INT, 0, 4);
 
 	table->id = DICT_COLUMNS_ID;
 
-	dict_table_add_to_cache(table);
+	dict_table_add_to_cache(table, heap);
 	dict_sys->sys_columns = table;
+	mem_heap_empty(heap);
 
 	index = dict_mem_index_create("SYS_COLUMNS", "CLUST_IND",
 				      DICT_HDR_SPACE,
@@ -311,13 +316,13 @@ dict_boot(void)
 	/*-------------------------*/
 	table = dict_mem_table_create("SYS_INDEXES", DICT_HDR_SPACE, 7, 0);
 
-	dict_mem_table_add_col(table, "TABLE_ID", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "ID", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "NAME", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "N_FIELDS", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "TYPE", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "SPACE", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "PAGE_NO", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "TABLE_ID", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "ID", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "NAME", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "N_FIELDS", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "TYPE", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "SPACE", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "PAGE_NO", DATA_INT, 0, 4);
 
 	/* The '+ 2' below comes from the 2 system fields */
 #if DICT_SYS_INDEXES_PAGE_NO_FIELD != 6 + 2
@@ -331,8 +336,9 @@ dict_boot(void)
 #endif
 
 	table->id = DICT_INDEXES_ID;
-	dict_table_add_to_cache(table);
+	dict_table_add_to_cache(table, heap);
 	dict_sys->sys_indexes = table;
+	mem_heap_empty(heap);
 
 	index = dict_mem_index_create("SYS_INDEXES", "CLUST_IND",
 				      DICT_HDR_SPACE,
@@ -349,13 +355,14 @@ dict_boot(void)
 	/*-------------------------*/
 	table = dict_mem_table_create("SYS_FIELDS", DICT_HDR_SPACE, 3, 0);
 
-	dict_mem_table_add_col(table, "INDEX_ID", DATA_BINARY, 0, 0);
-	dict_mem_table_add_col(table, "POS", DATA_INT, 0, 4);
-	dict_mem_table_add_col(table, "COL_NAME", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "INDEX_ID", DATA_BINARY, 0, 0);
+	dict_mem_table_add_col(table, heap, "POS", DATA_INT, 0, 4);
+	dict_mem_table_add_col(table, heap, "COL_NAME", DATA_BINARY, 0, 0);
 
 	table->id = DICT_FIELDS_ID;
-	dict_table_add_to_cache(table);
+	dict_table_add_to_cache(table, heap);
 	dict_sys->sys_fields = table;
+	mem_heap_free(heap);
 
 	index = dict_mem_index_create("SYS_FIELDS", "CLUST_IND",
 				      DICT_HDR_SPACE,

storage/innobase/dict/dict0crea.c

@@ -960,7 +960,7 @@ dict_create_table_step(
 
 	if (node->state == TABLE_ADD_TO_CACHE) {
 
-		dict_table_add_to_cache(node->table);
+		dict_table_add_to_cache(node->table, node->heap);
 
 		err = DB_SUCCESS;
 	}

storage/innobase/dict/dict0load.c

@@ -423,7 +423,8 @@ dict_load_columns(
 
 		ut_a(name_of_col_is(sys_columns, sys_index, 8, "PREC"));
 
-		dict_mem_table_add_col(table, name, mtype, prtype, col_len);
+		dict_mem_table_add_col(table, heap, name,
+				       mtype, prtype, col_len);
 		btr_pcur_move_to_next_user_rec(&pcur, &mtr);
 	}
 
@@ -746,7 +747,7 @@ dict_load_table(
 
 	ut_ad(mutex_own(&(dict_sys->mutex)));
 
-	heap = mem_heap_create(1000);
+	heap = mem_heap_create(32000);
 
 	mtr_start(&mtr);
 
@@ -852,7 +853,9 @@ err_exit:
 
 	dict_load_columns(table, heap);
 
-	dict_table_add_to_cache(table);
+	dict_table_add_to_cache(table, heap);
+
+	mem_heap_empty(heap);
 
 	dict_load_indexes(table, heap);
 

storage/innobase/dict/dict0mem.c

@@ -95,6 +95,10 @@ dict_mem_table_create(
 	default to 1 here.*/
 	table->autoinc_increment = 1;
 
+	/* The number of transactions that are either waiting on the
+	AUTOINC lock or have been granted the lock. */
+	table->n_waiting_or_granted_auto_inc_locks = 0;
+
 #ifdef UNIV_DEBUG
 	table->magic_n = DICT_TABLE_MAGIC_N;
 #endif /* UNIV_DEBUG */

storage/innobase/handler/ha_innodb.cc

@@ -92,11 +92,15 @@ extern "C" {
 #include "../storage/innobase/include/ha_prototypes.h"
 }
 
+static const long AUTOINC_OLD_STYLE_LOCKING = 0;
+static const long AUTOINC_NEW_STYLE_LOCKING = 1;
+static const long AUTOINC_NO_LOCKING = 2;
+
 static long innobase_mirrored_log_groups, innobase_log_files_in_group,
 	innobase_log_buffer_size, innobase_buffer_pool_awe_mem_mb,
 	innobase_additional_mem_pool_size, innobase_file_io_threads,
 	innobase_lock_wait_timeout, innobase_force_recovery,
-	innobase_open_files;
+	innobase_open_files, innobase_autoinc_lock_mode;
 
 static long long innobase_buffer_pool_size, innobase_log_file_size;
 
@@ -1928,110 +1932,6 @@ retry:
 	DBUG_RETURN(0);
 }
 
-#if 0
-/* TODO: put the
-MySQL-4.1 functionality back to 5.0. This is needed to get InnoDB Hot Backup
-to work. */
-
-/*********************************************************************
-This is called when MySQL writes the binlog entry for the current
-transaction. Writes to the InnoDB tablespace info which tells where the
-MySQL binlog entry for the current transaction ended. Also commits the
-transaction inside InnoDB but does NOT flush InnoDB log files to disk.
-To flush you have to call innobase_commit_complete(). We have separated
-flushing to eliminate the bottleneck of LOCK_log in log.cc which disabled
-InnoDB's group commit capability. */
-static
-int
-innobase_report_binlog_offset_and_commit(
-/*=====================================*/
-				/* out: 0 */
-        handlerton *hton, /* in: Innodb handlerton */ 
-	THD*	thd,		/* in: user thread */
-	void*	trx_handle,	/* in: InnoDB trx handle */
-	char*	log_file_name,	/* in: latest binlog file name */
-	my_off_t end_offset)	/* in: the offset in the binlog file
-				   up to which we wrote */
-{
-	trx_t*	trx;
-
-	trx = (trx_t*)trx_handle;
-
-	ut_a(trx != NULL);
-
-	trx->mysql_log_file_name = log_file_name;
-	trx->mysql_log_offset = (ib_longlong)end_offset;
-
-	trx->flush_log_later = TRUE;
-
-	innobase_commit(hton, thd, TRUE);
-
-	trx->flush_log_later = FALSE;
-
-	return(0);
-}
-
-/***********************************************************************
-This function stores the binlog offset and flushes logs. */
-static
-void
-innobase_store_binlog_offset_and_flush_log(
-/*=======================================*/
-	char*		binlog_name,	/* in: binlog name */
-	longlong	offset)		/* in: binlog offset */
-{
-	mtr_t mtr;
-
-	assert(binlog_name != NULL);
-
-	/* Start a mini-transaction */
-	mtr_start_noninline(&mtr);
-
-	/* Update the latest MySQL binlog name and offset info
-	in trx sys header */
-
-	trx_sys_update_mysql_binlog_offset(
-		binlog_name,
-		offset,
-		TRX_SYS_MYSQL_LOG_INFO, &mtr);
-
-	/* Commits the mini-transaction */
-	mtr_commit(&mtr);
-
-	/* Synchronous flush of the log buffer to disk */
-	log_buffer_flush_to_disk();
-}
-
-/*********************************************************************
-This is called after MySQL has written the binlog entry for the current
-transaction. Flushes the InnoDB log files to disk if required. */
-static
-int
-innobase_commit_complete(
-/*=====================*/
-				/* out: 0 */
-	THD*	thd)		/* in: user thread */
-{
-	trx_t*	trx;
-
-	trx = thd_to_trx(thd);
-
-	if (trx && trx->active_trans) {
-
-		trx->active_trans = 0;
-
-		if (UNIV_UNLIKELY(srv_flush_log_at_trx_commit == 0)) {
-
-			return(0);
-		}
-
-		trx_commit_complete_for_mysql(trx);
-	}
-
-	return(0);
-}
-#endif
-
 /*********************************************************************
 Rolls back a transaction or the latest SQL statement. */
 static
@@ -3353,24 +3253,46 @@ ha_innobase::innobase_autoinc_lock(void)
 {
 	ulint		error = DB_SUCCESS;
 
-	if (thd_sql_command(user_thd) == SQLCOM_INSERT) {
+	switch (innobase_autoinc_lock_mode) {
+	case AUTOINC_NO_LOCKING:
+		/* Acquire only the AUTOINC mutex. */
 		dict_table_autoinc_lock(prebuilt->table);
+		break;
 
-		/* We peek at the dict_table_t::auto_inc_lock to check if
-		another statement has locked it */
-		if (prebuilt->trx->auto_inc_lock != NULL) {
-			/* Release the mutex to avoid deadlocks */
-			dict_table_autoinc_unlock(prebuilt->table);
-
-			goto acquire_auto_inc_lock;
+	case AUTOINC_NEW_STYLE_LOCKING:
+		/* For simple (single/multi) row INSERTs, we fallback to the
+		old style only if another transaction has already acquired
+		the AUTOINC lock on behalf of a LOAD FILE or INSERT ... SELECT
+		etc. type of statement. */
+		if (thd_sql_command(user_thd) == SQLCOM_INSERT) {
+			dict_table_t*	table = prebuilt->table;
+
+			/* Acquire the AUTOINC mutex. */
+			dict_table_autoinc_lock(table);
+
+			/* We need to check that another transaction isn't
+			already holding the AUTOINC lock on the table. */
+			if (table->n_waiting_or_granted_auto_inc_locks) {
+				/* Release the mutex to avoid deadlocks. */
+				dict_table_autoinc_unlock(table);
+			} else {
+				break;
+			}
 		}
-	} else {
-acquire_auto_inc_lock:
+		/* Fall through to old style locking. */
+
+	case AUTOINC_OLD_STYLE_LOCKING:
 		error = row_lock_table_autoinc_for_mysql(prebuilt);
 
 		if (error == DB_SUCCESS) {
+
+			/* Acquire the AUTOINC mutex. */
 			dict_table_autoinc_lock(prebuilt->table);
 		}
+		break;
+
+	default:
+		ut_error;
 	}
 
 	return(ulong(error));
@@ -3610,6 +3532,7 @@ no_commit:
 
 			if (auto_inc > prebuilt->last_value) {
 set_max_autoinc:
+				ut_a(prebuilt->table->autoinc_increment > 0);
 				auto_inc += prebuilt->table->autoinc_increment;
 
 				innobase_set_max_autoinc(auto_inc);
@@ -3884,12 +3807,23 @@ ha_innobase::delete_row(
 	if (table->found_next_number_field && record == table->record[0]) {
 		ulonglong	dummy = 0;
 
-		error = innobase_get_auto_increment(&dummy);
+		/* First check whether the AUTOINC sub-system has been
+		initialized using the AUTOINC mutex. If not then we
+		do it the "proper" way, by acquiring the heavier locks. */
+		dict_table_autoinc_lock(prebuilt->table);
 
-		if (error == DB_SUCCESS) {
+		if (!prebuilt->table->autoinc_inited) {
+			dict_table_autoinc_unlock(prebuilt->table);
+
+			error = innobase_get_auto_increment(&dummy);
+
+			if (error == DB_SUCCESS) {
+				dict_table_autoinc_unlock(prebuilt->table);
+			} else {
+				goto error_exit;
+			}
+		} else  {
 			dict_table_autoinc_unlock(prebuilt->table);
-		} else {
-			goto error_exit;
 		}
 	}
 
@@ -7411,13 +7345,24 @@ ha_innobase::get_auto_increment(
 
 	*nb_reserved_values = prebuilt->trx->n_autoinc_rows;
 
-	/* Compute the last value in the interval */
-	prebuilt->last_value = *first_value + (*nb_reserved_values * increment);
+	/* With old style AUTOINC locking we only update the table's
+	AUTOINC counter after attempting to insert the row. */
+	if (innobase_autoinc_lock_mode != AUTOINC_OLD_STYLE_LOCKING) {
 
-	ut_a(prebuilt->last_value >= *first_value);
+		/* Compute the last value in the interval */
+		prebuilt->last_value = *first_value +
+		    (*nb_reserved_values * increment);
 
-	/* Update the table autoinc variable */
-	dict_table_autoinc_update(prebuilt->table, prebuilt->last_value);
+		ut_a(prebuilt->last_value >= *first_value);
+
+		/* Update the table autoinc variable */
+		dict_table_autoinc_update(
+			prebuilt->table, prebuilt->last_value);
+	} else {
+		/* This will force write_row() into attempting an update
+		of the table's AUTOINC counter. */
+		prebuilt->last_value = 0;
+	}
 
 	/* The increment to be used to increase the AUTOINC value, we use
 	this in write_row() and update_row() to increase the autoinc counter
@@ -8080,6 +8025,17 @@ static MYSQL_SYSVAR_STR(data_file_path, innobase_data_file_path,
   "Path to individual files and their sizes.",
   NULL, NULL, NULL);
 
+static MYSQL_SYSVAR_LONG(autoinc_lock_mode, innobase_autoinc_lock_mode,
+  PLUGIN_VAR_RQCMDARG,
+  "The AUTOINC lock modes supported by InnoDB:\n"
+  "  0 => Old style AUTOINC locking (for backward compatibility)\n"
+  "  1 => New style AUTOINC locking\n"
+  "  2 => No AUTOINC locking (unsafe for SBR)",
+  NULL, NULL,
+  AUTOINC_NEW_STYLE_LOCKING,	/* Default setting */
+  AUTOINC_OLD_STYLE_LOCKING,	/* Minimum value */
+  AUTOINC_NO_LOCKING, 0);	/* Maximum value */
+
 static struct st_mysql_sys_var* innobase_system_variables[]= {
   MYSQL_SYSVAR(additional_mem_pool_size),
   MYSQL_SYSVAR(autoextend_increment),
@@ -8118,6 +8074,7 @@ static struct st_mysql_sys_var* innobase_system_variables[]= {
   MYSQL_SYSVAR(table_locks),
   MYSQL_SYSVAR(thread_concurrency),
   MYSQL_SYSVAR(thread_sleep_delay),
+  MYSQL_SYSVAR(autoinc_lock_mode),
   NULL
 };
 

storage/innobase/include/dict0mem.h

@@ -315,11 +315,11 @@ struct dict_table_struct{
 	unsigned	n_cols:10;/* number of columns */
 	dict_col_t*	cols;	/* array of column descriptions */
 	const char*	col_names;
-				/* n_def column names packed in an
-				"name1\0name2\0...nameN\0" array. until
-				n_def reaches n_cols, this is allocated with
-				ut_malloc, and the final size array is
-				allocated through the table's heap. */
+				/* Column names packed in a character string
+				"name1\0name2\0...nameN\0".  Until
+				the string contains n_cols, it will be
+				allocated from a temporary heap.  The final
+				string will be allocated from table->heap. */
 	hash_node_t	name_hash; /* hash chain node */
 	hash_node_t	id_hash; /* hash chain node */
 	UT_LIST_BASE_NODE_T(dict_index_t)
@@ -416,6 +416,16 @@ struct dict_table_struct{
 				/* The increment step of the auto increment
 				column. Value must be greater than or equal
 				to 1 */
+	ulong		n_waiting_or_granted_auto_inc_locks;
+				/* This counter is used to track the number
+				of granted and pending autoinc locks on this
+				table. This value is set after acquiring the
+				kernel mutex but we peek the contents to
+				determine whether other transactions have
+				acquired the AUTOINC lock or not. Of course
+				only one transaction can be granted the
+				lock but there can be multiple waiters. */
+
 #ifdef UNIV_DEBUG
 	ulint		magic_n;/* magic number */
 # define DICT_TABLE_MAGIC_N	76333786

storage/innobase/include/mem0mem.ic

@@ -271,15 +271,19 @@ mem_heap_free_heap_top(
 	ut_ad(mem_block_get_start(block) <= mem_block_get_free(block));
 
 	/* In the debug version erase block from top up */
-
-	mem_erase_buf(old_top, (byte*)block + block->len - old_top);
+	{
+		ulint	len = (byte*)block + block->len - old_top;
+		mem_erase_buf(old_top, len);
+		UNIV_MEM_FREE(old_top, len);
+	}
 
 	/* Update allocated memory count */
 	mutex_enter(&mem_hash_mutex);
 	mem_current_allocated_memory -= (total_size - size);
 	mutex_exit(&mem_hash_mutex);
-
-#endif
+#else /* UNIV_MEM_DEBUG */
+	UNIV_MEM_FREE(old_top, (byte*)block + block->len - old_top);
+#endif /* UNIV_MEM_DEBUG */
 
 	/* If free == start, we may free the block if it is not the first
 	one */

storage/innobase/include/sync0rw.ic

@@ -231,7 +231,7 @@ rw_lock_s_lock_func(
 	owns an s-lock here, it may end up in a deadlock with another thread
 	which requests an x-lock here. Therefore, we will forbid recursive
 	s-locking of a latch: the following assert will warn the programmer
-	of the possibility of a tjis kind of deadlock. If we want to implement
+	of the possibility of this kind of a deadlock. If we want to implement
 	safe recursive s-locking, we should keep in a list the thread ids of
 	the threads which have s-locked a latch. This would use some CPU
 	time. */

storage/innobase/include/trx0trx.h

@@ -482,31 +482,6 @@ struct trx_struct{
 	ib_longlong	mysql_log_offset;/* if MySQL binlog is used, this field
 					contains the end offset of the binlog
 					entry */
-	const char*	mysql_master_log_file_name;
-					/* if the database server is a MySQL
-					replication slave, we have here the
-					master binlog name up to which
-					replication has processed; otherwise
-					this is a pointer to a null
-					character */
-	ib_longlong	mysql_master_log_pos;
-					/* if the database server is a MySQL
-					replication slave, this is the
-					position in the log file up to which
-					replication has processed */
-	/* A MySQL variable mysql_thd->synchronous_repl tells if we have
-	to use synchronous replication. See ha_innodb.cc. */
-	char*		repl_wait_binlog_name;/* NULL, or if synchronous MySQL
-					replication is used, the binlog name
-					up to which we must communicate the
-					binlog to the slave, before returning
-					from a commit; this is the same as
-					mysql_log_file_name, but we allocate
-					and copy the name to a separate buffer
-					here */
-	ib_longlong	repl_wait_binlog_pos;/* see above at
-					repl_wait_binlog_name */
-
 	os_thread_id_t	mysql_thread_id;/* id of the MySQL thread associated
 					with this transaction object */
 	ulint		mysql_process_no;/* since in Linux, 'top' reports

storage/innobase/lock/lock0lock.c

@@ -3386,6 +3386,10 @@ lock_table_create(
 	ut_ad(table && trx);
 	ut_ad(mutex_own(&kernel_mutex));
 
+	if ((type_mode & LOCK_MODE_MASK) == LOCK_AUTO_INC) {
+		++table->n_waiting_or_granted_auto_inc_locks;
+	}
+
 	if (type_mode == LOCK_AUTO_INC) {
 		/* Only one trx can have the lock on the table
 		at a time: we may use the memory preallocated
@@ -3436,6 +3440,9 @@ lock_table_remove_low(
 
 	if (lock == trx->auto_inc_lock) {
 		trx->auto_inc_lock = NULL;
+
+		ut_a(table->n_waiting_or_granted_auto_inc_locks > 0);
+		--table->n_waiting_or_granted_auto_inc_locks;
 	}
 
 	UT_LIST_REMOVE(trx_locks, trx->trx_locks, lock);

storage/innobase/mtr/mtr0log.c

@@ -517,8 +517,9 @@ mlog_parse_index(
 		n = mach_read_from_2(ptr);
 		ptr += 2;
 		n_uniq = mach_read_from_2(ptr);
+		ptr += 2;
 		ut_ad(n_uniq <= n);
-		if (end_ptr < ptr + (n + 1) * 2) {
+		if (end_ptr < ptr + n * 2) {
 			return(NULL);
 		}
 	} else {
@@ -531,18 +532,18 @@ mlog_parse_index(
 	ind->table = table;
 	ind->n_uniq = (unsigned int) n_uniq;
 	if (n_uniq != n) {
+		ut_a(n_uniq + DATA_ROLL_PTR <= n);
 		ind->type = DICT_CLUSTERED;
 	}
-	/* avoid ut_ad(index->cached) in dict_index_get_n_unique_in_tree */
-	ind->cached = TRUE;
 	if (comp) {
 		for (i = 0; i < n; i++) {
-			ulint	len = mach_read_from_2(ptr += 2);
+			ulint	len = mach_read_from_2(ptr);
+			ptr += 2;
 			/* The high-order bit of len is the NOT NULL flag;
 			the rest is 0 or 0x7fff for variable-length fields,
 			and 1..0x7ffe for fixed-length fields. */
 			dict_mem_table_add_col(
-				table, "DUMMY",
+				table, NULL, NULL,
 				((len + 1) & 0x7fff) <= 1
 				? DATA_BINARY : DATA_FIXBINARY,
 				len & 0x8000 ? DATA_NOT_NULL : 0,
@@ -552,8 +553,23 @@ mlog_parse_index(
 					   dict_table_get_nth_col(table, i),
 					   0);
 		}
-		ptr += 2;
+		dict_table_add_system_columns(table, table->heap);
+		if (n_uniq != n) {
+			/* Identify DB_TRX_ID and DB_ROLL_PTR in the index. */
+			ut_a(DATA_TRX_ID_LEN
+			     == dict_index_get_nth_col(ind, DATA_TRX_ID - 1
+						       + n_uniq)->len);
+			ut_a(DATA_ROLL_PTR_LEN
+			     == dict_index_get_nth_col(ind, DATA_ROLL_PTR - 1
+						       + n_uniq)->len);
+			ind->fields[DATA_TRX_ID - 1 + n_uniq].col
+				= &table->cols[n + DATA_TRX_ID];
+			ind->fields[DATA_ROLL_PTR - 1 + n_uniq].col
+				= &table->cols[n + DATA_ROLL_PTR];
+		}
 	}
+	/* avoid ut_ad(index->cached) in dict_index_get_n_unique_in_tree */
+	ind->cached = TRUE;
 	*index = ind;
 	return(ptr);
 }

storage/innobase/pars/pars0pars.c

@@ -1640,7 +1640,8 @@ pars_create_table(
 	while (column) {
 		dtype = dfield_get_type(que_node_get_val(column));
 
-		dict_mem_table_add_col(table, column->name, dtype->mtype,
+		dict_mem_table_add_col(table, table->heap,
+				       column->name, dtype->mtype,
 				       dtype->prtype, dtype->len);
 		column->resolved = TRUE;
 		column->token_type = SYM_COLUMN;

storage/innobase/srv/srv0srv.c

@@ -907,7 +907,7 @@ srv_init(void)
 	/* create dummy table and index for old-style infimum and supremum */
 	table = dict_mem_table_create("SYS_DUMMY1",
 				      DICT_HDR_SPACE, 1, 0);
-	dict_mem_table_add_col(table, "DUMMY", DATA_CHAR,
+	dict_mem_table_add_col(table, NULL, NULL, DATA_CHAR,
 			       DATA_ENGLISH | DATA_NOT_NULL, 8);
 
 	srv_sys->dummy_ind1 = dict_mem_index_create(
@@ -918,7 +918,7 @@ srv_init(void)
 	/* create dummy table and index for new-style infimum and supremum */
 	table = dict_mem_table_create("SYS_DUMMY2",
 				      DICT_HDR_SPACE, 1, DICT_TF_COMPACT);
-	dict_mem_table_add_col(table, "DUMMY", DATA_CHAR,
+	dict_mem_table_add_col(table, NULL, NULL, DATA_CHAR,
 			       DATA_ENGLISH | DATA_NOT_NULL, 8);
 	srv_sys->dummy_ind2 = dict_mem_index_create(
 		"SYS_DUMMY2", "SYS_DUMMY2", DICT_HDR_SPACE, 0, 1);

storage/innobase/trx/trx0sys.c

@@ -646,6 +646,7 @@ trx_sys_update_mysql_binlog_offset(
 			 MLOG_4BYTES, mtr);
 }
 
+#ifdef UNIV_HOTBACKUP
 /*********************************************************************
 Prints to stderr the MySQL binlog info in the system header if the
 magic number shows it valid. */
@@ -677,6 +678,7 @@ trx_sys_print_mysql_binlog_offset_from_page(
 			+ TRX_SYS_MYSQL_LOG_NAME);
 	}
 }
+#endif /* UNIV_HOTBACKUP */
 
 /*********************************************************************
 Stores the MySQL binlog offset info in the trx system header if

storage/innobase/trx/trx0trx.c

@@ -139,11 +139,6 @@ trx_create(
 
 	trx->mysql_log_file_name = NULL;
 	trx->mysql_log_offset = 0;
-	trx->mysql_master_log_file_name = "";
-	trx->mysql_master_log_pos = 0;
-
-	trx->repl_wait_binlog_name = NULL;
-	trx->repl_wait_binlog_pos = 0;
 
 	mutex_create(&trx->undo_mutex, SYNC_TRX_UNDO);
 
@@ -325,11 +320,6 @@ trx_free(
 		trx_undo_arr_free(trx->undo_no_arr);
 	}
 
-	if (trx->repl_wait_binlog_name != NULL) {
-
-		mem_free(trx->repl_wait_binlog_name);
-	}
-
 	ut_a(UT_LIST_GET_LEN(trx->signals) == 0);
 	ut_a(UT_LIST_GET_LEN(trx->reply_signals) == 0);
 
@@ -809,14 +799,6 @@ trx_commit_off_kernel(
 			trx->mysql_log_file_name = NULL;
 		}
 
-		if (trx->mysql_master_log_file_name[0] != '\0') {
-			/* This database server is a MySQL replication slave */
-			trx_sys_update_mysql_binlog_offset(
-				trx->mysql_master_log_file_name,
-				trx->mysql_master_log_pos,
-				TRX_SYS_MYSQL_MASTER_LOG_INFO, &mtr);
-		}
-
 		/* The following call commits the mini-transaction, making the
 		whole transaction committed in the file-based world, at this
 		log sequence number. The transaction becomes 'durable' when