Merge with 3.23.51

Fixed wrong usage of sprintf() in ha_innodb.cc


BitKeeper/etc/logging_ok:
  auto-union
BitKeeper/deleted/.del-global.h~e80d28157acfdcb5:
  Auto merged
Docs/manual.ja.texi:
  Auto merged
include/config-os2.h:
  Auto merged
innobase/btr/btr0btr.c:
  Auto merged
innobase/btr/btr0cur.c:
  Auto merged
innobase/btr/btr0sea.c:
  Auto merged
innobase/buf/buf0buf.c:
  Auto merged
innobase/data/data0data.c:
  Auto merged
innobase/dict/dict0crea.c:
  Auto merged
innobase/dict/dict0dict.c:
  Auto merged
innobase/fil/fil0fil.c:
  Auto merged
innobase/fsp/fsp0fsp.c:
  Auto merged
innobase/ibuf/ibuf0ibuf.c:
  Auto merged
innobase/include/buf0buf.ic:
  Auto merged
innobase/include/srv0srv.h:
  Auto merged
innobase/lock/lock0lock.c:
  Auto merged
innobase/log/log0log.c:
  Auto merged
innobase/mtr/mtr0mtr.c:
  Auto merged
innobase/os/os0file.c:
  Auto merged
innobase/page/page0cur.c:
  Auto merged
innobase/page/page0page.c:
  Auto merged
innobase/rem/rem0cmp.c:
  Auto merged
innobase/row/row0ins.c:
  Auto merged
innobase/row/row0purge.c:
  Auto merged
innobase/row/row0umod.c:
  Auto merged
innobase/row/row0upd.c:
  Auto merged
innobase/sync/sync0arr.c:
  Auto merged
innobase/sync/sync0sync.c:
  Auto merged
innobase/trx/trx0roll.c:
  Auto merged
innobase/trx/trx0trx.c:
  Auto merged
innobase/trx/trx0undo.c:
  Auto merged
innobase/ut/ut0ut.c:
  Auto merged
isam/pack_isam.c:
  Auto merged
mysys/Makefile.am:
  Auto merged
sql/ha_innodb.h:
  Auto merged
sql/key.cc:
  Auto merged
sql/mini_client.cc:
  Auto merged
sql/sql_delete.cc:
  Auto merged
sql/sql_lex.h:
  Auto merged
sql/sql_table.cc:
  Auto merged
sql/sql_update.cc:
  Auto merged
sql/table.cc:
  Auto merged
support-files/my-huge.cnf.sh:
  Auto merged
support-files/my-large.cnf.sh:
  Auto merged
support-files/my-medium.cnf.sh:
  Auto merged
support-files/my-small.cnf.sh:
  Auto merged
Docs/manual.texi:
  Merge with 3.23.51
configure.in:
  Merge with 3.23.51
include/Makefile.am:
  Merge with 3.23.51
innobase/dict/dict0load.c:
  Merge with 3.23.51
innobase/include/univ.i:
  Merge with 3.23.51
innobase/row/row0mysql.c:
  Merge with 3.23.51
innobase/srv/srv0srv.c:
  Merge with 3.23.51
innobase/srv/srv0start.c:
  Merge with 3.23.51
myisam/mi_cache.c:
  Merge with 3.23.51
myisammrg/myrg_rnext.c:
  Merge with 3.23.51
myisammrg/myrg_rprev.c:
  Merge with 3.23.51
sql/ha_innodb.cc:
  Merge with 3.23.51
  Changed used of sprintf() to make code portable.
sql/handler.cc:
  Indentation change
sql/lex.h:
  Comment cleanup
sql/log.cc:
  Merge with 3.23.51
sql/log_event.h:
  Merge with 3.23.51
sql/mysql_priv.h:
  Merge with 3.23.51
sql/mysqld.cc:
  Merge with 3.23.51
  Cleaned up handling of innodb_flush_log_at_trx_commit
sql/share/portuguese/errmsg.txt:
  Merge with 3.23.51
sql/slave.cc:
  Not merged as this affects many files; Will be fixed in next changeset
sql/slave.h:
  Merge with 3.23.51
sql/sql_acl.cc:
  Merge with 3.23.51 (no changes)
sql/sql_db.cc:
  Merge with 3.23.51
sql/sql_parse.cc:
  Merge with 3.23.51
sql/sql_yacc.yy:
  Merge with 3.23.51
  (Renamed NO_FOREIGN_KEY_CHECKS and RELAXED_UNIQUE_CHECKS)
support-files/mysql.spec.sh:
  Merge with 3.23.51

2 files changed, 153 insertions, 40 deletions

diff --git a/innobase/os/os0file.c b/innobase/os/os0file.c
index ee4045febde..ae3c8a45f62 100644
--- a/innobase/os/os0file.c
+++ b/innobase/os/os0file.c
@@ -22,6 +22,16 @@ Created 10/21/1995 Heikki Tuuri
 
 #endif
 
+/* This specifies the file permissions InnoDB uses when it craetes files in
+Unix; the value of os_innodb_umask is initialized in ha_innodb.cc to
+my_umask */
+
+#ifndef __WIN__
+ulint	os_innodb_umask		= S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP;
+#else
+ulint	os_innodb_umask		= 0;
+#endif
+
 /* If the following is set to TRUE, we do not call os_file_flush in every
 os_file_write. We can set this TRUE if the doublewrite buffer is used. */
 ibool	os_do_not_call_flush_at_each_write	= FALSE;
@@ -32,7 +42,7 @@ OS does not provide an atomic pread or pwrite, or similar */
 os_mutex_t	os_file_seek_mutexes[OS_FILE_N_SEEK_MUTEXES];
 
 /* In simulated aio, merge at most this many consecutive i/os */
-#define OS_AIO_MERGE_N_CONSECUTIVE	32
+#define OS_AIO_MERGE_N_CONSECUTIVE	64
 
 /* If this flag is TRUE, then we will use the native aio of the
 OS (provided we compiled Innobase with it in), otherwise we will
@@ -40,6 +50,8 @@ use simulated aio we build below with threads */
 
 ibool	os_aio_use_native_aio	= FALSE;
 
+ibool	os_aio_print_debug	= FALSE;
+
 /* The aio array slot structure */
 typedef struct os_aio_slot_struct	os_aio_slot_t;
 
@@ -115,7 +127,12 @@ os_aio_array_t*	os_aio_sync_array	= NULL;
 
 ulint	os_aio_n_segments	= ULINT_UNDEFINED;
 
+/* If the following is TRUE, read i/o handler threads try to
+wait until a batch of new read requests have been posted */
+ibool	os_aio_recommend_sleep_for_read_threads	= FALSE;
+
 ulint	os_n_file_reads		= 0;
+ulint	os_bytes_read_since_printout = 0;
 ulint	os_n_file_writes	= 0;
 ulint	os_n_fsyncs		= 0;
 ulint	os_n_file_reads_old	= 0;
@@ -412,8 +429,8 @@ try_again:
 	}
 
 	if (create_mode == OS_FILE_CREATE) {
-	        file = open(name, create_flag, S_IRUSR | S_IWUSR | S_IRGRP
-			                     | S_IWGRP | S_IROTH | S_IWOTH);
+	        file = open(name, create_flag, S_IRUSR | S_IWUSR
+						| S_IRGRP | S_IWGRP);
         } else {
                 file = open(name, create_flag);
         }
@@ -548,8 +565,7 @@ try_again:
 	}
 #endif
 	if (create_mode == OS_FILE_CREATE) {
-	        file = open(name, create_flag, S_IRUSR | S_IWUSR | S_IRGRP
-			                     | S_IWGRP | S_IROTH | S_IWOTH);
+	        file = open(name, create_flag, os_innodb_umask);
         } else {
                 file = open(name, create_flag);
         }
@@ -673,6 +689,7 @@ os_file_set_size(
 	ulint   	n_bytes;
 	ibool		ret;
 	byte*   	buf;
+	byte*   	buf2;
 	ulint   	i;
 
 	ut_a(size == (size & 0xFFFFFFFF));
@@ -680,7 +697,10 @@ os_file_set_size(
 	/* We use a very big 8 MB buffer in writing because Linux may be
 	extremely slow in fsync on 1 MB writes */
 
-	buf = ut_malloc(UNIV_PAGE_SIZE * 512);
+	buf2 = ut_malloc(UNIV_PAGE_SIZE * 513);
+
+	/* Align the buffer for possible raw i/o */
+	buf = ut_align(buf2, UNIV_PAGE_SIZE);
 
 	/* Write buffer full of zeros */
 	for (i = 0; i < UNIV_PAGE_SIZE * 512; i++) {
@@ -702,13 +722,13 @@ os_file_set_size(
 				    (ulint)(offset >> 32),
 				n_bytes);
 	        if (!ret) {
-			ut_free(buf);
+			ut_free(buf2);
 	         	goto error_handling;
 	        }
 	        offset += n_bytes;
 	}
 
-	ut_free(buf);
+	ut_free(buf2);
 
 	ret = os_file_flush(file);
 
@@ -734,6 +754,8 @@ os_file_flush(
 
 	ut_a(file);
 
+	os_n_fsyncs++;
+
 	ret = FlushFileBuffers(file);
 
 	if (ret) {
@@ -742,6 +764,10 @@ os_file_flush(
 
 	os_file_handle_error(file, NULL);
 
+	/* It is a fatal error if a file flush does not succeed, because then
+	the database can get corrupt on disk */
+	ut_a(0);
+
 	return(FALSE);
 #else
 	int	ret;
@@ -764,11 +790,17 @@ os_file_flush(
 	        return(TRUE);
 	}
 
+	ut_print_timestamp(stderr);
+	
 	fprintf(stderr,
-		"InnoDB: Error: the OS said file flush did not succeed\n");
+		"  InnoDB: Error: the OS said file flush did not succeed\n");
 
 	os_file_handle_error(file, NULL);
 
+	/* It is a fatal error if a file flush does not succeed, because then
+	the database can get corrupt on disk */
+	ut_a(0);
+
 	return(FALSE);
 #endif
 }
@@ -954,6 +986,7 @@ os_file_read(
 	ut_a((offset & 0xFFFFFFFF) == offset);
 
 	os_n_file_reads++;
+	os_bytes_read_since_printout += n;
 
 try_again:	
 	ut_ad(file);
@@ -1062,7 +1095,9 @@ os_file_write(
 
 		fprintf(stderr,
 "  InnoDB: Error: File pointer positioning to file %s failed at\n"
-"InnoDB: offset %lu %lu. Operating system error number %lu.\n",
+"InnoDB: offset %lu %lu. Operating system error number %lu.\n"
+"InnoDB: Look from section 13.2 at http://www.innodb.com/ibman.html\n"
+"InnoDB: what the error number means.\n",
 			name, offset_high, offset,
 			(ulint)GetLastError());
 
@@ -1093,8 +1128,10 @@ os_file_write(
 "  InnoDB: Error: Write to file %s failed at offset %lu %lu.\n"
 "InnoDB: %lu bytes should have been written, only %lu were written.\n"
 "InnoDB: Operating system error number %lu.\n"
+"InnoDB: Look from section 13.2 at http://www.innodb.com/ibman.html\n"
+"InnoDB: what the error number means.\n"
 "InnoDB: Check that your OS and file system support files of this size.\n"
-"InnoDB: Check also the disk is not full or a disk quota exceeded.\n",
+"InnoDB: Check also that the disk is not full or a disk quota exceeded.\n",
 			name, offset_high, offset, n, len,
 			(ulint)GetLastError());
 
@@ -1120,10 +1157,12 @@ os_file_write(
 "  InnoDB: Error: Write to file %s failed at offset %lu %lu.\n"
 "InnoDB: %lu bytes should have been written, only %lu were written.\n"
 "InnoDB: Operating system error number %lu.\n"
+"InnoDB: Look from section 13.2 at http://www.innodb.com/ibman.html\n"
+"InnoDB: what the error number means or use the perror program of MySQL.\n"
 "InnoDB: Check that your OS and file system support files of this size.\n"
-"InnoDB: Check also the disk is not full or a disk quota exceeded.\n",
-			name, offset_high, offset, n, ret, (ulint)errno);
-
+"InnoDB: Check also that the disk is not full or a disk quota exceeded.\n",
+			name, offset_high, offset, n, (ulint)ret,
+							(ulint)errno);
 		os_has_said_disk_full = TRUE;
 	}
 
@@ -1623,13 +1662,40 @@ os_aio_simulated_wake_handler_threads(void)
 		/* We do not use simulated aio: do nothing */
 
 		return;
-	}
+	}	
+
+	os_aio_recommend_sleep_for_read_threads	= FALSE;
 
 	for (i = 0; i < os_aio_n_segments; i++) {
 		os_aio_simulated_wake_handler_thread(i);
 	}
 }
 
+/**************************************************************************
+This function can be called if one wants to post a batch of reads and
+prefers an i/o-handler thread to handle them all at once later. You must
+call os_aio_simulated_wake_handler_threads later to ensure the threads
+are not left sleeping! */
+
+void
+os_aio_simulated_put_read_threads_to_sleep(void)
+/*============================================*/
+{
+	os_aio_array_t*	array;
+	ulint		g;
+
+	os_aio_recommend_sleep_for_read_threads	= TRUE;
+
+	for (g = 0; g < os_aio_n_segments; g++) {
+		os_aio_get_array_and_local_segment(&array, g);
+
+		if (array == os_aio_read_array) {
+		
+			os_event_reset(os_aio_segment_wait_events[g]);
+		}
+	}
+}
+
 /***********************************************************************
 Requests an asynchronous i/o operation. */
 
@@ -1685,7 +1751,6 @@ os_aio(
 	ut_ad(buf);
 	ut_ad(n > 0);
 	ut_ad(n % OS_FILE_LOG_BLOCK_SIZE == 0);
-	ut_ad((ulint)buf % OS_FILE_LOG_BLOCK_SIZE == 0)
 	ut_ad(offset % OS_FILE_LOG_BLOCK_SIZE == 0);
 	ut_ad(os_aio_validate());
 
@@ -2036,18 +2101,14 @@ os_aio_simulated_handle(
 	ulint		offs;
 	ulint		lowest_offset;
 	byte*		combined_buf;
+	byte*		combined_buf2;
 	ibool		ret;
 	ulint		n;
 	ulint		i;
-
+	
 	segment = os_aio_get_array_and_local_segment(&array, global_segment);
 	
 restart:
-	/* Give other threads chance to add several i/os to the array
-	at once */
-	
-	os_thread_yield();
-
 	/* NOTE! We only access constant fields in os_aio_array. Therefore
 	we do not have to acquire the protecting mutex yet */
 
@@ -2058,6 +2119,15 @@ restart:
 
 	/* Look through n slots after the segment * n'th slot */
 
+	if (array == os_aio_read_array
+	    && os_aio_recommend_sleep_for_read_threads) {
+
+		/* Give other threads chance to add several i/os to the array
+		at once. */
+
+		goto recommended_sleep;
+	}
+	
 	os_mutex_enter(array->mutex);
 
 	/* Check if there is a slot for which the i/o has already been
@@ -2068,6 +2138,11 @@ restart:
 
 		if (slot->reserved && slot->io_already_done) {
 
+			if (os_aio_print_debug) {
+				fprintf(stderr,
+"InnoDB: i/o for slot %lu already done, returning\n", i);
+			}
+
 			ret = TRUE;
 			
 			goto slot_io_done;
@@ -2149,9 +2224,11 @@ consecutive_loop:
 		/* We can use the buffer of the i/o request */
 		combined_buf = slot->buf;
 	} else {
-		combined_buf = ut_malloc(total_len);
+		combined_buf2 = ut_malloc(total_len + UNIV_PAGE_SIZE);
+
+		ut_a(combined_buf2);
 
-		ut_a(combined_buf);
+		combined_buf = ut_align(combined_buf2, UNIV_PAGE_SIZE);
 	}
 	
 	/* We release the array mutex for the time of the i/o: NOTE that
@@ -2174,6 +2251,13 @@ consecutive_loop:
 
 	srv_io_thread_op_info[global_segment] = (char*) "doing file i/o";
 
+	if (os_aio_print_debug) {
+		fprintf(stderr,
+"InnoDB: doing i/o of type %lu at offset %lu %lu, length %lu\n",
+			slot->type, slot->offset_high, slot->offset,
+			total_len);
+	}
+
 	/* Do the i/o with ordinary, synchronous i/o functions: */
 	if (slot->type == OS_FILE_WRITE) {
 		ret = os_file_write(slot->name, slot->file, combined_buf,
@@ -2203,7 +2287,7 @@ consecutive_loop:
 	}
 
 	if (n_consecutive > 1) {
-		ut_free(combined_buf);
+		ut_free(combined_buf2);
 	}
 
 	os_mutex_enter(array->mutex);
@@ -2241,10 +2325,18 @@ wait_for_io:
 
 	os_mutex_exit(array->mutex);
 
-	srv_io_thread_op_info[global_segment] = (char*) "waiting for i/o request";
+recommended_sleep:
+	srv_io_thread_op_info[global_segment] =
+				(char*)"waiting for i/o request";
 
 	os_event_wait(os_aio_segment_wait_events[global_segment]);
 
+	if (os_aio_print_debug) {
+		fprintf(stderr,
+"InnoDB: i/o handler thread for i/o segment %lu wakes up\n",
+			global_segment);
+	}
+	
 	goto restart;
 }
 
@@ -2305,22 +2397,30 @@ os_aio_validate(void)
 Prints info of the aio arrays. */
 
 void
-os_aio_print(void)
-/*==============*/
+os_aio_print(
+/*=========*/
+	char*	buf,	/* in/out: buffer where to print */
+	char*	buf_end)/* in: buffer end */
 {
 	os_aio_array_t*	array;
 	os_aio_slot_t*	slot;
 	ulint		n_reserved;
 	time_t		current_time;
 	double		time_elapsed;
+	double		avg_bytes_read;
 	ulint		i;
 
+	if (buf_end - buf < 1000) {
+
+		return;
+	}
+
 	for (i = 0; i < srv_n_file_io_threads; i++) {
-		printf("I/O thread %lu state: %s\n", i,
+		buf += sprintf(buf, "I/O thread %lu state: %s\n", i,
 					srv_io_thread_op_info[i]);
 	}
 
-	printf("Pending normal aio reads:");
+	buf += sprintf(buf, "Pending normal aio reads:");
 
 	array = os_aio_read_array;
 loop:
@@ -2347,12 +2447,12 @@ loop:
 
 	ut_a(array->n_reserved == n_reserved);
 
-	printf(" %lu", n_reserved);
+	buf += sprintf(buf, " %lu", n_reserved);
 	
 	os_mutex_exit(array->mutex);
 
 	if (array == os_aio_read_array) {
-		printf(", aio writes:");
+		buf += sprintf(buf, ", aio writes:");
 	
 		array = os_aio_write_array;
 
@@ -2360,38 +2460,50 @@ loop:
 	}
 
 	if (array == os_aio_write_array) {
-		printf(",\n ibuf aio reads:");
+		buf += sprintf(buf, ",\n ibuf aio reads:");
 		array = os_aio_ibuf_array;
 
 		goto loop;
 	}
 
 	if (array == os_aio_ibuf_array) {
-		printf(", log i/o's:");
+		buf += sprintf(buf, ", log i/o's:");
 		array = os_aio_log_array;
 
 		goto loop;
 	}
 
 	if (array == os_aio_log_array) {
-		printf(", sync i/o's:");		
+		buf += sprintf(buf, ", sync i/o's:");		
 		array = os_aio_sync_array;
 
 		goto loop;
 	}
 
-	printf("\n");
+	buf += sprintf(buf, "\n");
 	
 	current_time = time(NULL);
 	time_elapsed = difftime(current_time, os_last_printout);
 
-	printf("Pending flushes (fsync) log: %lu; buffer pool: %lu\n",
+	buf += sprintf(buf,
+		"Pending flushes (fsync) log: %lu; buffer pool: %lu\n",
 	       fil_n_pending_log_flushes, fil_n_pending_tablespace_flushes);
-	printf("%lu OS file reads, %lu OS file writes, %lu OS fsyncs\n",
+	buf += sprintf(buf,
+		"%lu OS file reads, %lu OS file writes, %lu OS fsyncs\n",
 		os_n_file_reads, os_n_file_writes, os_n_fsyncs);
-	printf("%.2f reads/s, %.2f writes/s, %.2f fsyncs/s\n",
+
+	if (os_n_file_reads == os_n_file_reads_old) {
+		avg_bytes_read = 0.0;
+	} else {
+		avg_bytes_read = os_bytes_read_since_printout /
+				(os_n_file_reads - os_n_file_reads_old);
+	}
+
+	buf += sprintf(buf,
+"%.2f reads/s, %lu avg bytes/read, %.2f writes/s, %.2f fsyncs/s\n",
 		(os_n_file_reads - os_n_file_reads_old)
 		/ time_elapsed,
+		(ulint)avg_bytes_read,
 		(os_n_file_writes - os_n_file_writes_old)
 		/ time_elapsed,
 		(os_n_fsyncs - os_n_fsyncs_old)
@@ -2400,6 +2512,7 @@ loop:
 	os_n_file_reads_old = os_n_file_reads;
 	os_n_file_writes_old = os_n_file_writes;
 	os_n_fsyncs_old = os_n_fsyncs;
+	os_bytes_read_since_printout = 0;
 	
 	os_last_printout = current_time;
 }
diff --git a/innobase/os/os0sync.c b/innobase/os/os0sync.c
index 2cf5160d055..c1345de0d55 100644
--- a/innobase/os/os0sync.c
+++ b/innobase/os/os0sync.c
@@ -435,7 +435,7 @@ os_fast_mutex_init(
 	
 	InitializeCriticalSection((LPCRITICAL_SECTION) fast_mutex);
 #else
-	pthread_mutex_init(fast_mutex, NULL);
+	pthread_mutex_init(fast_mutex, MY_MUTEX_INIT_FAST);
 #endif
 }