branches/innodb+: Merge revisions 4072:4150 from branches/zip:

  ------------------------------------------------------------------------
  r4074 | vasil | 2009-01-31 08:05:24 +0200 (Sat, 31 Jan 2009) | 4 lines

  branches/zip:

  Adjust the failing patch patches/information_schema.diff.
  ------------------------------------------------------------------------
  r4076 | vasil | 2009-02-02 09:32:04 +0200 (Mon, 02 Feb 2009) | 4 lines

  branches/zip:

  Add ChangeLog entry for the change in r4072.
  ------------------------------------------------------------------------
  r4077 | marko | 2009-02-02 10:48:05 +0200 (Mon, 02 Feb 2009) | 2 lines

  branches/zip: innobase_start_or_create_for_mysql(): Remove a factual error
  in the function comment.  Parameters are not read from a file "srv_init".
  ------------------------------------------------------------------------
  r4081 | marko | 2009-02-02 14:28:17 +0200 (Mon, 02 Feb 2009) | 4 lines

  branches/zip: Enclose some backup functions in #ifdef UNIV_HOTBACKUP.

  recv_read_cp_info_for_backup(), recv_scan_log_seg_for_backup():
  These functions are only called by InnoDB Hot Backup.
  ------------------------------------------------------------------------
  r4082 | vasil | 2009-02-02 18:24:08 +0200 (Mon, 02 Feb 2009) | 10 lines

  branches/zip:

  Fix a mysql-test failure in innodb-zip:

  main.innodb-zip                          [ fail ]
          Test ended at 2009-02-02 18:13:25

  CURRENT_TEST: main.innodb-zip
  mysqltest: At line 160: Found line beginning with --  that didn't contain a valid mysqltest command, check your syntax or use # if you intended to write a comment
  ------------------------------------------------------------------------
  r4083 | vasil | 2009-02-02 18:33:20 +0200 (Mon, 02 Feb 2009) | 6 lines

  branches/zip:

  Fix the failing innodb-zip test to restore the environment as it was before
  the test execution because a newly added feature in the mysql-test framework
  does check for this.
  ------------------------------------------------------------------------
  r4088 | calvin | 2009-02-03 02:35:56 +0200 (Tue, 03 Feb 2009) | 8 lines

  branches/zip: fix a compiler error and a warning

  Both are minor changes:
  1) Compiler error introduced in r4072: double ';' at the end.
  2) Warning introduced in r3613: \mem\mem0pool.c(481) :
  warning C4098: 'mem_area_free' : 'void' function returning a value

  Approved by: Sunny (IM)
  ------------------------------------------------------------------------
  r4098 | marko | 2009-02-03 09:52:45 +0200 (Tue, 03 Feb 2009) | 4 lines

  branches/zip: mem_area_free(): Correct a bug that was introduced in r4088.
  free() is not the same as ut_free().  ut_free() pairs with ut_malloc(),
  not malloc().  free() pairs with malloc() and some other functions.
  ------------------------------------------------------------------------
  r4114 | marko | 2009-02-04 16:09:24 +0200 (Wed, 04 Feb 2009) | 2 lines

  branches/zip: buf_block_align(): Fix a bogus debug assertion
  that was introduced in r4036, to address Issue #161.
  ------------------------------------------------------------------------
  r4139 | vasil | 2009-02-09 13:47:16 +0200 (Mon, 09 Feb 2009) | 5 lines

  branches/zip:

  Remove mysql-test/patches/bug35261.diff because that bug has been fixed
  in the MySQL repository.
  ------------------------------------------------------------------------
  r4141 | marko | 2009-02-09 15:35:50 +0200 (Mon, 09 Feb 2009) | 1 line

  branches/zip: fil_write_lsn_and_arch_no_to_file(): Plug a memory leak.
  ------------------------------------------------------------------------
  r4144 | inaam | 2009-02-10 01:36:25 +0200 (Tue, 10 Feb 2009) | 9 lines

  branches/zip            rb://30

  This patch changes the innodb mutexes and rw_locks implementation.
  On supported platforms it uses GCC builtin atomics. These changes
  are based on the patch sent by Mark Callaghan of Google under BSD
  license. More technical discussion can be found at rb://30

  Approved by: Heikki
  ------------------------------------------------------------------------
  r4145 | vasil | 2009-02-10 07:34:43 +0200 (Tue, 10 Feb 2009) | 9 lines

  branches/zip:

  Non-functional change: Fix a compilation warning introduced in r4144:

  gcc -DHAVE_CONFIG_H -I. -I../../include -I../../include -I../../include -I../../regex -I../../storage/innobase/include -I../../sql -I.   -Werror         -Wall -g   -MT libinnobase_a-sync0arr.o -MD -MP -MF .deps/libinnobase_a-sync0arr.Tpo -c -o libinnobase_a-sync0arr.o `test -f 'sync/sync0arr.c' || echo './'`sync/sync0arr.c
  cc1: warnings being treated as errors
  sync/sync0arr.c: In function 'sync_array_object_signalled':
  sync/sync0arr.c:869: warning: pointer targets in passing argument 1 of 'os_atomic_increment' differ in signedness
  ------------------------------------------------------------------------
  r4148 | marko | 2009-02-10 10:38:41 +0200 (Tue, 10 Feb 2009) | 12 lines

  branches/zip: Map ut_malloc(), ut_realloc(), ut_free() to
  malloc(), realloc(), free() when innodb_use_sys_malloc is set.

  ut_free_all_mem(): If innodb_use_sys_malloc is set, do nothing,
  because then ut_mem_block_list_inited will never be set.

  log_init(): Use mem_alloc() instead of ut_malloc(), so that the
  memory will be freed.  (Tested with Valgrind, although it is not
  clear why the memory would be freed.)

  rb://86 approved by Heikki Tuuri and Ken Jacobs.  This addresses Issue #168.
  ------------------------------------------------------------------------
  r4149 | marko | 2009-02-10 11:09:15 +0200 (Tue, 10 Feb 2009) | 1 line

  branches/zip: ChangeLog: Document recent changes.
  ------------------------------------------------------------------------
  r4150 | marko | 2009-02-10 11:51:43 +0200 (Tue, 10 Feb 2009) | 6 lines

  branches/zip: get_share(), free_share(): Make table locking case sensitive.
  If lower_case_table_names=1, MySQL will pass the table names in lower case.
  Thus, we can use a binary comparison (strcmp) in the hash table.

  rb://87 approved by Heikki Tuuri, to address Bug #41676 and Issue #167.
  ------------------------------------------------------------------------

ChangeLog

+2009-02-10	The InnoDB Team
+
+	* handler/ha_innodb.h, handler/ha_innodb.cc:
+	Fix Bug#41676 Table names are case insensitive in locking
+
+2009-02-10	The InnoDB Team
+
+	* ut/ut0mem.c:
+	Map ut_malloc_low(), ut_realloc(), and ut_free() directly to
+	malloc(), realloc(), and free() when innodb_use_sys_malloc is set.
+	As a side effect, ut_total_allocated_memory ("Total memory allocated"
+	in the "BUFFER POOL AND MEMORY" section of SHOW ENGINE INNODB STATUS)
+	will exclude any memory allocated by these functions when
+	innodb_use_sys_malloc is set.
+
+2009-02-10	The InnoDB Team
+
+	* btr/btr0cur.c, btr/btr0sea.c, buf/buf0buf.c, handler/ha_innodb.cc,
+	include/buf0buf.ic, include/os0sync.h, include/os0sync.ic,
+	include/srv0srv.h, include/sync0rw.h, include/sync0rw.ic,
+	include/sync0sync.h, include/sync0sync.ic, include/univ.i,
+	row/row0sel.c, srv/srv0srv.c, srv/srv0start.c,
+	sync/sync0arr.c, sync/sync0rw.c, sync/sync0sync.c:
+	On those platforms that support it, implement the synchronization
+	primitives of InnoDB mutexes and read/write locks with GCC atomic
+	builtins instead of Pthreads mutexes and InnoDB mutexes.  These
+	changes are based on a patch supplied by Mark Callaghan of Google
+	under a BSD license.
+
+2009-01-30	The InnoDB Team
+
+	* btr/btr0cur.c, btr/btr0sea.c, buf/buf0buf.c, handler/ha_innodb.cc,
+	include/btr0sea.h, include/buf0buf.h, include/sync0sync.h,
+	sync/sync0sync.c:
+	Make the configuration parameter innodb_adaptive_hash_index dynamic,
+	so that it can be changed at runtime.
+
 2009-01-29	The InnoDB Team
 
 	* handler/ha_innodb.cc, include/ibuf0ibuf.h, include/ibuf0ibuf.ic,
@@ -107,7 +144,10 @@
 	mysql-test/innodb-use-sys-malloc.test:
 	Implement the configuration parameter innodb_use_sys_malloc
 	(false by default), for disabling InnoDB's internal memory allocator
-	and using system malloc/free instead.
+	and using system malloc/free instead.  The "BUFFER POOL AND MEMORY"
+	section of SHOW ENGINE INNODB STATUS will report
+	"in additional pool allocated allocated 0" when
+	innodb_use_sys_malloc is set.
 
 2008-12-30	The InnoDB Team
 

btr/btr0cur.c

@@ -16,6 +16,38 @@ by crashing the database and doing a roll-forward.
 
 Created 10/16/1994 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #include "btr0cur.h"
 
@@ -416,7 +448,7 @@ btr_cur_search_to_nth_level(
 
 	/* Ibuf does not use adaptive hash; this is prevented by the
 	latch_mode check below. */
-	if (btr_search_latch.writer == RW_LOCK_NOT_LOCKED
+	if (rw_lock_get_writer(&btr_search_latch) == RW_LOCK_NOT_LOCKED
 	    && latch_mode <= BTR_MODIFY_LEAF
 	    && info->last_hash_succ
 	    && !estimate

btr/btr0sea.c

@@ -6,6 +6,39 @@ The index tree adaptive search
 Created 2/17/1996 Heikki Tuuri
 *************************************************************************/
 
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
+
 #include "btr0sea.h"
 #ifdef UNIV_NONINL
 #include "btr0sea.ic"
@@ -820,8 +853,8 @@ btr_search_guess_on_hash(
 		}
 	}
 
-	ut_ad(btr_search_latch.writer != RW_LOCK_EX);
-	ut_ad(btr_search_latch.reader_count > 0);
+	ut_ad(rw_lock_get_writer(&btr_search_latch) != RW_LOCK_EX);
+	ut_ad(rw_lock_get_reader_count(&btr_search_latch) > 0);
 
 	rec = ha_search_and_get_data(btr_search_sys->hash_index, fold);
 

buf/buf0buf.c

@@ -19,6 +19,38 @@ The database buffer buf_pool
 
 Created 11/5/1995 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #include "buf0buf.h"
 
@@ -1016,7 +1048,7 @@ buf_pool_drop_hash_index(void)
 	ut_ad(!btr_search_enabled);
 
 	do {
-		buf_chunk_t*	chunks	= buf_pool->chunks;;
+		buf_chunk_t*	chunks	= buf_pool->chunks;
 		buf_chunk_t*	chunk	= chunks + buf_pool->n_chunks;
 
 		released_search_latch = FALSE;
@@ -1924,12 +1956,52 @@ buf_block_align(
 			buf_block_init() so that block[n].frame ==
 			block->frame + n * UNIV_PAGE_SIZE.  Check it. */
 			ut_ad(block->frame == page_align(ptr));
-			/* The space id and page number should be
-			stamped on the page. */
-			ut_ad(block->page.space
-			      == page_get_space_id(page_align(ptr)));
-			ut_ad(block->page.offset
-			      == page_get_page_no(page_align(ptr)));
+#ifdef UNIV_DEBUG
+			/* A thread that updates these fields must
+			hold buf_pool_mutex and block->mutex.  Acquire
+			only the latter. */
+			mutex_enter(&block->mutex);
+
+			switch (buf_block_get_state(block)) {
+			case BUF_BLOCK_ZIP_FREE:
+			case BUF_BLOCK_ZIP_PAGE:
+			case BUF_BLOCK_ZIP_DIRTY:
+				/* These types should only be used in
+				the compressed buffer pool, whose
+				memory is allocated from
+				buf_pool->chunks, in UNIV_PAGE_SIZE
+				blocks flagged as BUF_BLOCK_MEMORY. */
+				ut_error;
+				break;
+			case BUF_BLOCK_NOT_USED:
+			case BUF_BLOCK_READY_FOR_USE:
+			case BUF_BLOCK_MEMORY:
+				/* Some data structures contain
+				"guess" pointers to file pages.  The
+				file pages may have been freed and
+				reused.  Do not complain. */
+				break;
+			case BUF_BLOCK_REMOVE_HASH:
+				/* buf_LRU_block_remove_hashed_page()
+				will overwrite the FIL_PAGE_OFFSET and
+				FIL_PAGE_ARCH_LOG_NO_OR_SPACE_ID with
+				0xff and set the state to
+				BUF_BLOCK_REMOVE_HASH. */
+				ut_ad(page_get_space_id(page_align(ptr))
+				      == 0xffffffff);
+				ut_ad(page_get_page_no(page_align(ptr))
+				      == 0xffffffff);
+				break;
+			case BUF_BLOCK_FILE_PAGE:
+				ut_ad(block->page.space
+				      == page_get_space_id(page_align(ptr)));
+				ut_ad(block->page.offset
+				      == page_get_page_no(page_align(ptr)));
+				break;
+			}
+
+			mutex_exit(&block->mutex);
+#endif /* UNIV_DEBUG */
 
 			return(block);
 		}
@@ -2366,8 +2438,8 @@ buf_page_optimistic_get_func(
 			   buf_block_get_page_no(block), NULL));
 
 	if (rw_latch == RW_S_LATCH) {
-		success = rw_lock_s_lock_func_nowait(&(block->lock),
-						     file, line);
+		success = rw_lock_s_lock_nowait(&(block->lock),
+						file, line);
 		fix_type = MTR_MEMO_PAGE_S_FIX;
 	} else {
 		success = rw_lock_x_lock_func_nowait(&(block->lock),
@@ -2478,8 +2550,8 @@ buf_page_get_known_nowait(
 	ut_ad(!ibuf_inside() || (mode == BUF_KEEP_OLD));
 
 	if (rw_latch == RW_S_LATCH) {
-		success = rw_lock_s_lock_func_nowait(&(block->lock),
-						     file, line);
+		success = rw_lock_s_lock_nowait(&(block->lock),
+						file, line);
 		fix_type = MTR_MEMO_PAGE_S_FIX;
 	} else {
 		success = rw_lock_x_lock_func_nowait(&(block->lock),
@@ -2556,7 +2628,7 @@ buf_page_try_get_func(
 	mutex_exit(&block->mutex);
 
 	fix_type = MTR_MEMO_PAGE_S_FIX;
-	success = rw_lock_s_lock_func_nowait(&block->lock, file, line);
+	success = rw_lock_s_lock_nowait(&block->lock, file, line);
 
 	if (!success) {
 		/* Let us try to get an X-latch. If the current thread

fil/fil0fil.c

@@ -1678,6 +1678,8 @@ fil_write_lsn_and_arch_no_to_file(
 
 	fil_write(TRUE, 0, 0, sum_of_sizes, 0, UNIV_PAGE_SIZE, buf, NULL);
 
+	mem_free(buf1);
+
 	return(DB_SUCCESS);
 }
 

handler/ha_innodb.cc

@@ -13,6 +13,38 @@
    along with this program; if not, write to the Free Software
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307	 USA */
 
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 /* TODO list for the InnoDB handler in 5.0:
   - Remove the flag trx->active_trans and look at trx->conc_state
   - fix savepoint functions to use savepoint storage area
@@ -466,6 +498,8 @@ static SHOW_VAR innodb_status_variables[]= {
   (char*) &export_vars.innodb_dblwr_pages_written,	  SHOW_LONG},
   {"dblwr_writes",
   (char*) &export_vars.innodb_dblwr_writes,		  SHOW_LONG},
+  {"have_atomic_builtins",
+  (char*) &export_vars.innodb_have_atomic_builtins,	  SHOW_BOOL},
   {"log_waits",
   (char*) &export_vars.innodb_log_waits,		  SHOW_LONG},
   {"log_write_requests",
@@ -8045,7 +8079,8 @@ innodb_mutex_show_status(
 	stat_print_fn*	stat_print)
 {
 	char buf1[IO_SIZE], buf2[IO_SIZE];
-	mutex_t*  mutex;
+	mutex_t*	mutex;
+	rw_lock_t*	lock;
 #ifdef UNIV_DEBUG
 	ulint	  rw_lock_count= 0;
 	ulint	  rw_lock_count_spin_loop= 0;
@@ -8116,6 +8151,29 @@ innodb_mutex_show_status(
 
 	mutex_exit(&mutex_list_mutex);
 
+	mutex_enter(&rw_lock_list_mutex);
+
+	lock = UT_LIST_GET_FIRST(rw_lock_list);
+
+	while (lock != NULL) {
+		if (lock->count_os_wait) {
+			buf1len= my_snprintf(buf1, sizeof(buf1), "%s:%lu",
+                                    lock->cfile_name, (ulong) lock->cline);
+			buf2len= my_snprintf(buf2, sizeof(buf2),
+                                    "os_waits=%lu", lock->count_os_wait);
+
+			if (stat_print(thd, innobase_hton_name,
+				       hton_name_len, buf1, buf1len,
+				       buf2, buf2len)) {
+				mutex_exit(&rw_lock_list_mutex);
+				DBUG_RETURN(1);
+			}
+		}
+		lock = UT_LIST_GET_NEXT(list, lock);
+	}
+
+	mutex_exit(&rw_lock_list_mutex);
+
 #ifdef UNIV_DEBUG
 	buf2len= my_snprintf(buf2, sizeof(buf2),
 		"count=%lu, spin_waits=%lu, spin_rounds=%lu, "
@@ -8151,56 +8209,35 @@ bool innobase_show_status(handlerton *hton, THD* thd,
 	}
 }
 
-
 /****************************************************************************
  Handling the shared INNOBASE_SHARE structure that is needed to provide table
  locking.
 ****************************************************************************/
 
-/****************************************************************************
-Folds a string in system_charset_info. */
-static
-ulint
-innobase_fold_name(
-/*===============*/
-				/* out: fold value of the name */
-	const uchar*	name,	/* in: string to be folded */
-	size_t		length)	/* in: length of the name in bytes */
-{
-	ulong n1 = 1, n2 = 4;
-
-	system_charset_info->coll->hash_sort(system_charset_info,
-					     name, length, &n1, &n2);
-	return((ulint) n1);
-}
-
 static INNOBASE_SHARE* get_share(const char* table_name)
 {
 	INNOBASE_SHARE *share;
 	pthread_mutex_lock(&innobase_share_mutex);
-	uint length=(uint) strlen(table_name);
 
-	ulint	fold = innobase_fold_name((const uchar*) table_name, length);
+	ulint	fold = ut_fold_string(table_name);
 
 	HASH_SEARCH(table_name_hash, innobase_open_tables, fold,
 		    INNOBASE_SHARE*, share,
 		    ut_ad(share->use_count > 0),
-		    !my_strnncoll(system_charset_info,
-				  share->table_name,
-				  share->table_name_length,
-				  (const uchar*) table_name, length));
+		    !strcmp(share->table_name, table_name));
 
 	if (!share) {
 
+		uint length = (uint) strlen(table_name);
+
 		/* TODO: invoke HASH_MIGRATE if innobase_open_tables
 		grows too big */
 
 		share = (INNOBASE_SHARE *) my_malloc(sizeof(*share)+length+1,
 			MYF(MY_FAE | MY_ZEROFILL));
 
-		share->table_name_length = length;
-		share->table_name = (uchar*) memcpy(share + 1,
-						    table_name, length + 1);
+		share->table_name = (char*) memcpy(share + 1,
+						   table_name, length + 1);
 
 		HASH_INSERT(INNOBASE_SHARE, table_name_hash,
 			    innobase_open_tables, fold, share);
@@ -8221,24 +8258,18 @@ static void free_share(INNOBASE_SHARE* share)
 
 #ifdef UNIV_DEBUG
 	INNOBASE_SHARE* share2;
-	ulint fold = innobase_fold_name(share->table_name,
-					share->table_name_length);
+	ulint	fold = ut_fold_string(share->table_name);
 
 	HASH_SEARCH(table_name_hash, innobase_open_tables, fold,
 		    INNOBASE_SHARE*, share2,
 		    ut_ad(share->use_count > 0),
-		    !my_strnncoll(system_charset_info,
-				  share->table_name,
-				  share->table_name_length,
-				  share2->table_name,
-				  share2->table_name_length));
+		    !strcmp(share->table_name, share2->table_name));
 
 	ut_a(share2 == share);
 #endif /* UNIV_DEBUG */
 
 	if (!--share->use_count) {
-		ulint	fold = innobase_fold_name(share->table_name,
-						  share->table_name_length);
+		ulint	fold = ut_fold_string(share->table_name);
 
 		HASH_DELETE(INNOBASE_SHARE, table_name_hash,
 			    innobase_open_tables, fold, share);

handler/ha_innodb.h

@@ -27,8 +27,8 @@
 typedef struct st_innobase_share {
   THR_LOCK lock;
   pthread_mutex_t mutex;
-  const uchar *table_name;
-  uint table_name_length,use_count;
+  const char* table_name;
+  uint use_count;
   void* table_name_hash;
 } INNOBASE_SHARE;
 

include/buf0buf.ic

@@ -5,6 +5,38 @@ The database buffer buf_pool
 
 Created 11/5/1995 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #include "buf0flu.h"
 #include "buf0lru.h"
@@ -851,7 +883,7 @@ buf_block_buf_fix_inc_func(
 #ifdef UNIV_SYNC_DEBUG
 	ibool	ret;
 
-	ret = rw_lock_s_lock_func_nowait(&(block->debug_latch), file, line);
+	ret = rw_lock_s_lock_nowait(&(block->debug_latch), file, line);
 	ut_a(ret);
 #endif /* UNIV_SYNC_DEBUG */
 	ut_ad(mutex_own(&block->mutex));

include/log0recv.h

@@ -17,7 +17,6 @@ Created 9/20/1997 Heikki Tuuri
 
 #ifdef UNIV_HOTBACKUP
 extern ibool	recv_replay_file_ops;
-#endif /* UNIV_HOTBACKUP */
 
 /***********************************************************************
 Reads the checkpoint info needed in hot backup. */
@@ -55,6 +54,7 @@ recv_scan_log_seg_for_backup(
 	ulint*		n_bytes_scanned);/* out: how much we were able to
 					scan, smaller than buf_len if log
 					data ended here */
+#endif /* UNIV_HOTBACKUP */
 /***********************************************************************
 Returns TRUE if recovery is currently running. */
 UNIV_INLINE

include/os0sync.h

@@ -6,6 +6,38 @@ synchronization primitives.
 
 Created 9/6/1995 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 #ifndef os0sync_h
 #define os0sync_h
 
@@ -261,6 +293,29 @@ os_fast_mutex_free(
 /*===============*/
 	os_fast_mutex_t*	fast_mutex);	/* in: mutex to free */
 
+#ifdef HAVE_GCC_ATOMIC_BUILTINS
+/**************************************************************
+Atomic compare-and-swap for InnoDB. Currently requires GCC atomic builtins. */
+UNIV_INLINE
+ibool
+os_compare_and_swap(
+/*================*/
+						/* out: true if swapped */
+	volatile lint*		ptr,		/* in: pointer to target */
+	lint			oldVal,		/* in: value to compare to */
+	lint			newVal);	/* in: value to swap in */
+/**************************************************************
+Atomic increment for InnoDB. Currently requires GCC atomic builtins. */
+UNIV_INLINE
+lint
+os_atomic_increment(
+/*================*/
+						/* out: resulting value */
+	volatile lint*		ptr,		/* in: pointer to target */
+	lint			amount);	/* in: amount of increment */
+
+#endif /* HAVE_GCC_ATOMIC_BUILTINS */
+
 #ifndef UNIV_NONINL
 #include "os0sync.ic"
 #endif

include/os0sync.ic

@@ -5,6 +5,38 @@ The interface to the operating system synchronization primitives.
 
 Created 9/6/1995 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #ifdef __WIN__
 #include <winbase.h>
@@ -44,3 +76,38 @@ os_fast_mutex_trylock(
 #endif
 #endif
 }
+
+#ifdef HAVE_GCC_ATOMIC_BUILTINS
+/**************************************************************
+Atomic compare-and-swap for InnoDB. Currently requires GCC atomic builtins. */
+UNIV_INLINE
+ibool
+os_compare_and_swap(
+/*================*/
+					/* out: true if swapped */
+	volatile lint*	ptr,		/* in: pointer to target */
+	lint		oldVal,		/* in: value to compare to */
+	lint		newVal)		/* in: value to swap in */
+{
+	if(__sync_bool_compare_and_swap(ptr, oldVal, newVal)) {
+		return(TRUE);
+	}
+
+	return(FALSE);
+}
+
+/**************************************************************
+Atomic increment for InnoDB. Currently requires GCC atomic builtins. */
+UNIV_INLINE
+lint
+os_atomic_increment(
+/*================*/
+					/* out: resulting value */
+	volatile lint*	ptr,		/* in: pointer to target */
+	lint		amount)		/* in: amount of increment */
+{
+	lint newVal = __sync_add_and_fetch(ptr, amount);
+	return newVal;
+}
+
+#endif /* HAVE_GCC_ATOMIC_BUILTINS */

include/srv0srv.h

@@ -5,7 +5,38 @@ The server main program
 
 Created 10/10/1995 Heikki Tuuri
 *******************************************************/
-
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #ifndef srv0srv_h
 #define srv0srv_h
@@ -533,6 +564,7 @@ struct export_var_struct{
 	ulint innodb_buffer_pool_read_ahead_rnd;
 	ulint innodb_dblwr_pages_written;
 	ulint innodb_dblwr_writes;
+	ibool innodb_have_atomic_builtins;
 	ulint innodb_log_waits;
 	ulint innodb_log_write_requests;
 	ulint innodb_log_writes;
@@ -569,4 +601,3 @@ struct srv_sys_struct{
 extern ulint	srv_n_threads_active[];
 
 #endif
-

include/srv0start.h

@@ -65,8 +65,7 @@ srv_add_path_separator_if_needed(
 	char*	str);	/* in: null-terminated character string */
 /********************************************************************
 Starts Innobase and creates a new database if database files
-are not found and the user wants. Server parameters are
-read from a file of name "srv_init" in the ib_home directory. */
+are not found and the user wants. */
 UNIV_INTERN
 int
 innobase_start_or_create_for_mysql(void);

include/sync0sync.h

@@ -5,6 +5,38 @@ Mutex, the basic synchronization primitive
 
 Created 9/5/1995 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #ifndef sync0sync_h
 #define sync0sync_h
@@ -237,7 +269,7 @@ mutex_n_reserved(void);
 NOT to be used outside this module except in debugging! Gets the value
 of the lock word. */
 UNIV_INLINE
-ulint
+byte
 mutex_get_lock_word(
 /*================*/
 	const mutex_t*	mutex);	/* in: mutex */
@@ -463,9 +495,11 @@ implementation of a mutual exclusion semaphore. */
 
 struct mutex_struct {
 	os_event_t	event;	/* Used by sync0arr.c for the wait queue */
-	ulint	lock_word;	/* This ulint is the target of the atomic
-				test-and-set instruction in Win32 */
-#if defined WIN32 && defined UNIV_CAN_USE_X86_ASSEMBLER
+	byte	lock_word;	/* This byte is the target of the atomic
+				test-and-set instruction in Win32 and
+				x86 32/64 with GCC 4.1.0 or later version */
+#if defined(_WIN32) && defined(UNIV_CAN_USE_X86_ASSEMBLER)
+#elif defined(HAVE_GCC_ATOMIC_BUILTINS)
 #else
 	os_fast_mutex_t
 		os_fast_mutex;	/* In other systems we use this OS mutex
@@ -519,8 +553,7 @@ to 20 microseconds. */
 /* The number of system calls made in this module. Intended for performance
 monitoring. */
 
-extern	ulint	mutex_system_call_count;
-extern	ulint	mutex_exit_count;
+extern	ib_int64_t	mutex_exit_count;
 
 #ifdef UNIV_SYNC_DEBUG
 /* Latching order checks start when this is set TRUE */

include/sync0sync.ic

@@ -5,16 +5,38 @@ Mutex, the basic synchronization primitive
 
 Created 9/5/1995 Heikki Tuuri
 *******************************************************/
-
-#if defined(not_defined) && defined(__GNUC__) && defined(UNIV_INTEL_X86)
-/* %z0: Use the size of operand %0 which in our case is *m to determine
-instruction size, it should end up as xchgl. "1" in the input constraint,
-says that "in" has to go in the same place as "out".*/
-#define TAS(m, in, out) \
-	asm volatile ("xchg%z0 %2, %0" \
-	: "=g" (*(m)), "=r" (out) \
-	: "1" (in))	/* Note: "1" here refers to "=r" (out) */
-#endif
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 /**********************************************************************
 Sets the waiters field in a mutex. */
@@ -59,7 +81,7 @@ mutex_signal_object(
 Performs an atomic test-and-set instruction to the lock_word field of a
 mutex. */
 UNIV_INLINE
-ulint
+byte
 mutex_test_and_set(
 /*===============*/
 				/* out: the previous value of lock_word: 0 or
@@ -67,18 +89,18 @@ mutex_test_and_set(
 	mutex_t*	mutex)	/* in: mutex */
 {
 #if defined(_WIN32) && defined(UNIV_CAN_USE_X86_ASSEMBLER)
-	ulint	res;
-	ulint*	lw;		/* assembler code is used to ensure that
+	byte	res;
+	byte*	lw;		/* assembler code is used to ensure that
 				lock_word is loaded from memory */
 	ut_ad(mutex);
-	ut_ad(sizeof(ulint) == 4);
+	ut_ad(sizeof(byte) == 1);
 
 	lw = &(mutex->lock_word);
 
 	__asm	MOV	ECX, lw
 		__asm	MOV	EDX, 1
-		__asm	XCHG	EDX, DWORD PTR [ECX]
-		__asm	MOV	res, EDX
+		__asm	XCHG	DL, BYTE PTR [ECX]
+		__asm	MOV	res, DL
 
 		/* The fence below would prevent this thread from
 		reading the data structure protected by the mutex
@@ -98,12 +120,8 @@ mutex_test_and_set(
 		/* mutex_fence(); */
 
 		return(res);
-#elif defined(not_defined) && defined(__GNUC__) && defined(UNIV_INTEL_X86)
-	ulint	res;
-
-	TAS(&mutex->lock_word, 1, res);
-
-	return(res);
+#elif defined(HAVE_GCC_ATOMIC_BUILTINS)
+	return __sync_lock_test_and_set(&(mutex->lock_word), 1);
 #else
 	ibool	ret;
 
@@ -117,7 +135,7 @@ mutex_test_and_set(
 		mutex->lock_word = 1;
 	}
 
-	return(ret);
+	return((byte)ret);
 #endif
 }
 
@@ -131,7 +149,7 @@ mutex_reset_lock_word(
 	mutex_t*	mutex)	/* in: mutex */
 {
 #if defined(_WIN32) && defined(UNIV_CAN_USE_X86_ASSEMBLER)
-	ulint*	lw;		/* assembler code is used to ensure that
+	byte*	lw;		/* assembler code is used to ensure that
 				lock_word is loaded from memory */
 	ut_ad(mutex);
 
@@ -139,11 +157,12 @@ mutex_reset_lock_word(
 
 	__asm	MOV	EDX, 0
 		__asm	MOV	ECX, lw
-		__asm	XCHG	EDX, DWORD PTR [ECX]
-#elif defined(not_defined) && defined(__GNUC__) && defined(UNIV_INTEL_X86)
-	ulint	res;
-
-	TAS(&mutex->lock_word, 0, res);
+		__asm	XCHG	DL, BYTE PTR [ECX]
+#elif defined(HAVE_GCC_ATOMIC_BUILTINS)
+	/* In theory __sync_lock_release should be used to release the lock.
+	Unfortunately, it does not work properly alone. The workaround is
+	that more conservative __sync_lock_test_and_set is used instead. */
+	__sync_lock_test_and_set(&(mutex->lock_word), 0);
 #else
 	mutex->lock_word = 0;
 
@@ -154,12 +173,12 @@ mutex_reset_lock_word(
 /**********************************************************************
 Gets the value of the lock word. */
 UNIV_INLINE
-ulint
+byte
 mutex_get_lock_word(
 /*================*/
 	const mutex_t*	mutex)	/* in: mutex */
 {
-	const volatile ulint*	ptr;	/* declared volatile to ensure that
+	const volatile byte*	ptr;	/* declared volatile to ensure that
 					lock_word is loaded from memory */
 	ut_ad(mutex);
 

include/univ.i

@@ -5,6 +5,38 @@ Version control for database, common definitions, and include files
 
 Created 1/20/1994 Heikki Tuuri
 ****************************************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #ifndef univ_i
 #define univ_i
@@ -90,6 +122,13 @@ of the 32-bit x86 assembler in mutex operations. */
 #  define UNIV_CAN_USE_X86_ASSEMBLER
 # endif
 
+/* For InnoDB rw_locks to work with atomics we need the thread_id
+to be no more than machine word wide. The following enables using
+atomics for InnoDB rw_locks where these conditions are met. */
+# if defined(HAVE_GCC_ATOMIC_BUILTINS) && defined(UNIV_LINUX)
+#  define INNODB_RW_LOCKS_USE_ATOMICS
+# endif
+
 /* We only try to do explicit inlining of functions with gcc and
 Microsoft Visual C++ */
 

log/log0log.c

@@ -739,7 +739,7 @@ log_init(void)
 	ut_a(LOG_BUFFER_SIZE >= 16 * OS_FILE_LOG_BLOCK_SIZE);
 	ut_a(LOG_BUFFER_SIZE >= 4 * UNIV_PAGE_SIZE);
 
-	buf = ut_malloc(LOG_BUFFER_SIZE + OS_FILE_LOG_BLOCK_SIZE);
+	buf = mem_alloc(LOG_BUFFER_SIZE + OS_FILE_LOG_BLOCK_SIZE);
 	log_sys->buf = ut_align(buf, OS_FILE_LOG_BLOCK_SIZE);
 
 	log_sys->buf_size = LOG_BUFFER_SIZE;

log/log0recv.c

@@ -591,6 +591,7 @@ recv_find_max_checkpoint(
 	return(DB_SUCCESS);
 }
 
+#ifdef UNIV_HOTBACKUP
 /***********************************************************************
 Reads the checkpoint info needed in hot backup. */
 UNIV_INTERN
@@ -661,6 +662,7 @@ recv_read_cp_info_for_backup(
 
 	return(TRUE);
 }
+#endif /* UNIV_HOTBACKUP */
 
 /**********************************************************
 Checks the 4-byte checksum to the trailer checksum field of a log block.
@@ -698,6 +700,7 @@ log_block_checksum_is_ok_or_old_format(
 	return(FALSE);
 }
 
+#ifdef UNIV_HOTBACKUP
 /***********************************************************************
 Scans the log segment and n_bytes_scanned is set to the length of valid
 log scanned. */
@@ -787,6 +790,7 @@ recv_scan_log_seg_for_backup(
 		}
 	}
 }
+#endif /* UNIV_HOTBACKUP */
 
 /***********************************************************************
 Tries to parse a single log record body and also applies it to a page if

mem/mem0pool.c

@@ -478,7 +478,9 @@ mem_area_free(
 	ulint		n;
 
 	if (srv_use_sys_malloc) {
-		return(free(ptr));
+		free(ptr);
+
+		return;
 	}
 
 	/* It may be that the area was really allocated from the OS with

mysql-test/innodb-zip.result

@@ -419,3 +419,6 @@ select @@innodb_file_format_check;
 @@innodb_file_format_check
 Barracuda
 drop table normal_table, zip_table;
+set global innodb_file_format=Antelope;
+set global innodb_file_per_table=0;
+set global innodb_file_format_check=Antelope;

mysql-test/innodb-zip.test

@@ -2,6 +2,7 @@
 
 let $per_table=`select @@innodb_file_per_table`;
 let $format=`select @@innodb_file_format`;
+let $innodb_file_format_check_orig=`select @@innodb_file_format_check`;
 set global innodb_file_per_table=off;
 set global innodb_file_format=`0`;
 
@@ -157,7 +158,8 @@ set global innodb_file_format=`1a`;
 set global innodb_file_format=``;
 
 #test strict mode.
---enable_errors
+# this does not work anymore, has been removed from mysqltest
+# -- enable_errors
 set global innodb_file_per_table = on;
 set global innodb_file_format = `1`;
 
@@ -330,3 +332,11 @@ show table status;
 select @@innodb_file_format_check;
 drop table normal_table, zip_table;
 -- disable_result_log
+
+#
+# restore environment to the state it was before this test execution
+#
+
+eval set global innodb_file_format=$format;
+eval set global innodb_file_per_table=$per_table;
+eval set global innodb_file_format_check=$innodb_file_format_check_orig;

mysql-test/patches/bug35261.diff

---- mysql-test/t/date_formats.test.orig	2007-06-15 02:53:07.000000000 +0300
-+++ mysql-test/t/date_formats.test	2008-03-19 17:25:10.000000000 +0200
-@@ -7,9 +7,15 @@
- --enable_warnings
- 
- --replace_result ROW <format> STATEMENT <format> MIXED <format>
--SHOW GLOBAL VARIABLES LIKE "%e_format";
-+SELECT variable_name, variable_value
-+FROM information_schema.global_variables
-+WHERE variable_name IN ('date_format', 'datetime_format', 'time_format')
-+ORDER BY variable_name;
- --replace_result ROW <format> STATEMENT <format> MIXED <format>
--SHOW SESSION VARIABLES LIKE "%e_format";
-+SELECT variable_name, variable_value
-+FROM information_schema.session_variables
-+WHERE variable_name IN ('date_format', 'datetime_format', 'time_format')
-+ORDER BY variable_name;
- 
- #
- # Test setting a lot of different formats to see which formats are accepted and
-@@ -37,7 +43,10 @@
- set datetime_format= '%h:%i:%s.%f %p %Y-%m-%d';
- 
- --replace_result ROW <format> STATEMENT <format> MIXED <format>
--SHOW SESSION VARIABLES LIKE "%e_format";
-+SELECT variable_name, variable_value
-+FROM information_schema.session_variables
-+WHERE variable_name IN ('date_format', 'datetime_format', 'time_format')
-+ORDER BY variable_name;
- 
- --error 1231
- SET time_format='%h:%i:%s';
---- mysql-test/r/date_formats.result.orig	2008-02-12 21:09:14.000000000 +0200
-+++ mysql-test/r/date_formats.result	2008-03-19 17:26:33.000000000 +0200
-@@ -1,14 +1,20 @@
- drop table if exists t1;
--SHOW GLOBAL VARIABLES LIKE "%e_format";
--Variable_name	Value
--date_format	%d.%m.%Y
--datetime_format	%Y-%m-%d %H:%i:%s
--time_format	%H.%i.%s
--SHOW SESSION VARIABLES LIKE "%e_format";
--Variable_name	Value
--date_format	%d.%m.%Y
--datetime_format	%Y-%m-%d %H:%i:%s
--time_format	%H.%i.%s
-+SELECT variable_name, variable_value
-+FROM information_schema.global_variables
-+WHERE variable_name IN ('date_format', 'datetime_format', 'time_format')
-+ORDER BY variable_name;
-+variable_name	variable_value
-+DATETIME_FORMAT	%Y-%m-%d %H:%i:%s
-+DATE_FORMAT	%d.%m.%Y
-+TIME_FORMAT	%H.%i.%s
-+SELECT variable_name, variable_value
-+FROM information_schema.session_variables
-+WHERE variable_name IN ('date_format', 'datetime_format', 'time_format')
-+ORDER BY variable_name;
-+variable_name	variable_value
-+DATETIME_FORMAT	%Y-%m-%d %H:%i:%s
-+DATE_FORMAT	%d.%m.%Y
-+TIME_FORMAT	%H.%i.%s
- SET time_format='%H%i%s';
- SET time_format='%H:%i:%s.%f';
- SET time_format='%h-%i-%s.%f%p';
-@@ -26,11 +32,14 @@
- set datetime_format= '%H:%i:%s.%f %m-%d-%Y';
- set datetime_format= '%h:%i:%s %p %Y-%m-%d';
- set datetime_format= '%h:%i:%s.%f %p %Y-%m-%d';
--SHOW SESSION VARIABLES LIKE "%e_format";
--Variable_name	Value
--date_format	%m-%d-%Y
--datetime_format	%h:%i:%s.%f %p %Y-%m-%d
--time_format	%h:%i:%s%p
-+SELECT variable_name, variable_value
-+FROM information_schema.session_variables
-+WHERE variable_name IN ('date_format', 'datetime_format', 'time_format')
-+ORDER BY variable_name;
-+variable_name	variable_value
-+DATETIME_FORMAT	%h:%i:%s.%f %p %Y-%m-%d
-+DATE_FORMAT	%m-%d-%Y
-+TIME_FORMAT	%h:%i:%s%p
- SET time_format='%h:%i:%s';
- ERROR 42000: Variable 'time_format' can't be set to the value of '%h:%i:%s'
- SET time_format='%H %i:%s';

mysql-test/patches/information_schema.diff

-diff mysql-test/r/information_schema.result.orig mysql-test/r/information_schema.result
---- mysql-test/r/information_schema.result.orig	2008-08-04 09:27:49.000000000 +0300
-+++ mysql-test/r/information_schema.result	2008-10-07 11:21:51.000000000 +0300
-@@ -64,6 +64,13 @@
+--- mysql-test/r/information_schema.result.orig	2009-01-31 03:38:50.000000000 +0200
++++ mysql-test/r/information_schema.result	2009-01-31 07:51:58.000000000 +0200
+@@ -71,6 +71,13 @@
  TRIGGERS
  USER_PRIVILEGES
  VIEWS
@@ -15,7 +14,7 @@ diff mysql-test/r/information_schema.result.orig mysql-test/r/information_schema
  columns_priv
  db
  event
-@@ -795,6 +802,8 @@
+@@ -799,6 +806,8 @@
  TABLES	UPDATE_TIME	datetime
  TABLES	CHECK_TIME	datetime
  TRIGGERS	CREATED	datetime
@@ -24,16 +23,16 @@ diff mysql-test/r/information_schema.result.orig mysql-test/r/information_schema
  event	execute_at	datetime
  event	last_executed	datetime
  event	starts	datetime
-@@ -848,7 +857,7 @@
+@@ -852,7 +861,7 @@
  flush privileges;
- SELECT table_schema, count(*) FROM information_schema.TABLES where table_name<>'ndb_binlog_index' AND table_name<>'ndb_apply_status' GROUP BY TABLE_SCHEMA;
+ SELECT table_schema, count(*) FROM information_schema.TABLES WHERE table_schema IN ('mysql', 'INFORMATION_SCHEMA', 'test', 'mysqltest') AND table_name<>'ndb_binlog_index' AND table_name<>'ndb_apply_status' GROUP BY TABLE_SCHEMA;
  table_schema	count(*)
 -information_schema	28
 +information_schema	35
  mysql	22
  create table t1 (i int, j int);
  create trigger trg1 before insert on t1 for each row
-@@ -1263,6 +1272,13 @@
+@@ -1267,6 +1276,13 @@
  TRIGGERS	TRIGGER_SCHEMA
  USER_PRIVILEGES	GRANTEE
  VIEWS	TABLE_SCHEMA
@@ -47,7 +46,7 @@ diff mysql-test/r/information_schema.result.orig mysql-test/r/information_schema
  SELECT t.table_name, c1.column_name
  FROM information_schema.tables t
  INNER JOIN
-@@ -1306,6 +1322,13 @@
+@@ -1310,6 +1326,13 @@
  TRIGGERS	TRIGGER_SCHEMA
  USER_PRIVILEGES	GRANTEE
  VIEWS	TABLE_SCHEMA
@@ -58,10 +57,10 @@ diff mysql-test/r/information_schema.result.orig mysql-test/r/information_schema
 +INNODB_CMPMEM	page_size
 +INNODB_CMP	page_size
 +INNODB_LOCKS	lock_id
- SELECT MAX(table_name) FROM information_schema.tables;
+ SELECT MAX(table_name) FROM information_schema.tables WHERE table_schema IN ('mysql', 'INFORMATION_SCHEMA', 'test');
  MAX(table_name)
  VIEWS
-@@ -1382,6 +1405,13 @@
+@@ -1386,6 +1409,13 @@
  FILES	information_schema.FILES	1
  GLOBAL_STATUS	information_schema.GLOBAL_STATUS	1
  GLOBAL_VARIABLES	information_schema.GLOBAL_VARIABLES	1

row/row0sel.c

@@ -5,6 +5,38 @@ Select
 
 Created 12/19/1997 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #include "row0sel.h"
 
@@ -1339,7 +1371,7 @@ row_sel(
 			rw_lock_s_lock(&btr_search_latch);
 
 			search_latch_locked = TRUE;
-		} else if (btr_search_latch.writer_is_wait_ex) {
+		} else if (rw_lock_get_writer(&btr_search_latch) == RW_LOCK_WAIT_EX) {
 
 			/* There is an x-latch request waiting: release the
 			s-latch for a moment; as an s-latch here is often
@@ -3364,7 +3396,7 @@ row_search_for_mysql(
 	/* PHASE 0: Release a possible s-latch we are holding on the
 	adaptive hash index latch if there is someone waiting behind */
 
-	if (UNIV_UNLIKELY(btr_search_latch.writer != RW_LOCK_NOT_LOCKED)
+	if (UNIV_UNLIKELY(rw_lock_get_writer(&btr_search_latch) != RW_LOCK_NOT_LOCKED)
 	    && trx->has_search_latch) {
 
 		/* There is an x-latch request on the adaptive hash index:

srv/srv0srv.c

@@ -24,6 +24,38 @@ thread library. This might confuse NT though.
 
 Created 10/8/1995 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 /* Dummy comment */
 #include "srv0srv.h"
 
@@ -1828,6 +1860,11 @@ srv_export_innodb_status(void)
 	export_vars.innodb_buffer_pool_pages_misc = buf_pool->curr_size
 		- UT_LIST_GET_LEN(buf_pool->LRU)
 		- UT_LIST_GET_LEN(buf_pool->free);
+#ifdef HAVE_GCC_ATOMIC_BUILTINS
+	export_vars.innodb_have_atomic_builtins = 1;
+#else
+	export_vars.innodb_have_atomic_builtins = 0;
+#endif
 	export_vars.innodb_page_size = UNIV_PAGE_SIZE;
 	export_vars.innodb_log_waits = srv_log_waits;
 	export_vars.innodb_os_log_written = srv_os_log_written;

srv/srv0start.c

@@ -5,6 +5,38 @@ Starts the InnoDB database server
 
 Created 2/16/1996 Heikki Tuuri
 *************************************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #include "os0proc.h"
 #include "sync0sync.h"
@@ -956,8 +988,7 @@ open_or_create_data_files(
 
 /********************************************************************
 Starts InnoDB and creates a new database if database files
-are not found and the user wants. Server parameters are
-read from a file of name "srv_init" in the ib_home directory. */
+are not found and the user wants. */
 UNIV_INTERN
 int
 innobase_start_or_create_for_mysql(void)
@@ -1053,6 +1084,17 @@ innobase_start_or_create_for_mysql(void)
 			"InnoDB: The InnoDB memory heap is disabled\n");
 	}
 
+#ifdef HAVE_GCC_ATOMIC_BUILTINS
+#ifdef INNODB_RW_LOCKS_USE_ATOMICS
+	fprintf(stderr,
+		"InnoDB: Mutex and rw_lock use GCC atomic builtins.\n");
+#else
+	fprintf(stderr,
+		"InnoDB: Mutex use GCC atomic builtins.\n");
+#endif
+
+#endif
+
 	/* Since InnoDB does not currently clean up all its internal data
 	structures in MySQL Embedded Server Library server_end(), we
 	print an error message if someone tries to start up InnoDB a

sync/sync0arr.c

@@ -5,6 +5,38 @@ The wait array used in synchronization primitives
 
 Created 9/5/1995 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #include "sync0arr.h"
 #ifdef UNIV_NONINL
@@ -295,25 +327,21 @@ sync_array_validate(
 }
 
 /***********************************************************************
-Puts the cell event in reset state. */
+Returns the event that the thread owning the cell waits for. */
 static
-ib_int64_t
-sync_cell_event_reset(
-/*==================*/
-				/* out: value of signal_count
-				at the time of reset. */
-	ulint		type,	/* in: lock type mutex/rw_lock */
-	void*		object) /* in: the rw_lock/mutex object */
+os_event_t
+sync_cell_get_event(
+/*================*/
+	sync_cell_t*	cell) /* in: non-empty sync array cell */
 {
+	ulint type = cell->request_type;
+
 	if (type == SYNC_MUTEX) {
-		return(os_event_reset(((mutex_t *) object)->event));
-#ifdef __WIN__
+		return(((mutex_t *) cell->wait_object)->event);
 	} else if (type == RW_LOCK_WAIT_EX) {
-		return(os_event_reset(
-		       ((rw_lock_t *) object)->wait_ex_event));
-#endif
-	} else {
-		return(os_event_reset(((rw_lock_t *) object)->event));
+		return(((rw_lock_t *) cell->wait_object)->wait_ex_event);
+	} else { /* RW_LOCK_SHARED and RW_LOCK_EX wait on the same event */
+		return(((rw_lock_t *) cell->wait_object)->event);
 	}
 }
 
@@ -332,6 +360,7 @@ sync_array_reserve_cell(
 	ulint*		index)	/* out: index of the reserved cell */
 {
 	sync_cell_t*	cell;
+	os_event_t      event;
 	ulint		i;
 
 	ut_a(object);
@@ -370,8 +399,8 @@ sync_array_reserve_cell(
 			/* Make sure the event is reset and also store
 			the value of signal_count at which the event
 			was reset. */
-			cell->signal_count = sync_cell_event_reset(type,
-								object);
+                        event = sync_cell_get_event(cell);
+			cell->signal_count = os_event_reset(event);
 
 			cell->reservation_time = time(NULL);
 
@@ -411,19 +440,7 @@ sync_array_wait_event(
 	ut_a(!cell->waiting);
 	ut_ad(os_thread_get_curr_id() == cell->thread);
 
-	if (cell->request_type == SYNC_MUTEX) {
-		event = ((mutex_t*) cell->wait_object)->event;
-#ifdef __WIN__
-	/* On windows if the thread about to wait is the one which
-	has set the state of the rw_lock to RW_LOCK_WAIT_EX, then
-	it waits on a special event i.e.: wait_ex_event. */
-	} else if (cell->request_type == RW_LOCK_WAIT_EX) {
-		event = ((rw_lock_t*) cell->wait_object)->wait_ex_event;
-#endif
-	} else {
-		event = ((rw_lock_t*) cell->wait_object)->event;
-	}
-
+	event = sync_cell_get_event(cell);
 		cell->waiting = TRUE;
 
 #ifdef UNIV_SYNC_DEBUG
@@ -462,6 +479,7 @@ sync_array_cell_print(
 	mutex_t*	mutex;
 	rw_lock_t*	rwlock;
 	ulint		type;
+	ulint		writer;
 
 	type = cell->request_type;
 
@@ -491,9 +509,7 @@ sync_array_cell_print(
 			(ulong) mutex->waiters);
 
 	} else if (type == RW_LOCK_EX
-#ifdef __WIN__
 		   || type == RW_LOCK_WAIT_EX
-#endif
 		   || type == RW_LOCK_SHARED) {
 
 		fputs(type == RW_LOCK_EX ? "X-lock on" : "S-lock on", file);
@@ -504,22 +520,25 @@ sync_array_cell_print(
 			" RW-latch at %p created in file %s line %lu\n",
 			(void*) rwlock, rwlock->cfile_name,
 			(ulong) rwlock->cline);
-		if (rwlock->writer != RW_LOCK_NOT_LOCKED) {
+		writer = rw_lock_get_writer(rwlock);
+		if (writer != RW_LOCK_NOT_LOCKED) {
 			fprintf(file,
 				"a writer (thread id %lu) has"
 				" reserved it in mode %s",
 				(ulong) os_thread_pf(rwlock->writer_thread),
-				rwlock->writer == RW_LOCK_EX
+				writer == RW_LOCK_EX
 				? " exclusive\n"
 				: " wait exclusive\n");
 		}
 
 		fprintf(file,
-			"number of readers %lu, waiters flag %lu\n"
+			"number of readers %lu, waiters flag %lu, "
+                        "lock_word: %lx\n"
 			"Last time read locked in file %s line %lu\n"
 			"Last time write locked in file %s line %lu\n",
-			(ulong) rwlock->reader_count,
+			(ulong) rw_lock_get_reader_count(rwlock),
 			(ulong) rwlock->waiters,
+			rwlock->lock_word,
 			rwlock->last_s_file_name,
 			(ulong) rwlock->last_s_line,
 			rwlock->last_x_file_name,
@@ -778,28 +797,30 @@ sync_arr_cell_can_wake_up(
 			return(TRUE);
 		}
 
-	} else if (cell->request_type == RW_LOCK_EX
-		   || cell->request_type == RW_LOCK_WAIT_EX) {
+	} else if (cell->request_type == RW_LOCK_EX) {
 
 		lock = cell->wait_object;
 
-		if (rw_lock_get_reader_count(lock) == 0
-		    && rw_lock_get_writer(lock) == RW_LOCK_NOT_LOCKED) {
+		if (lock->lock_word > 0) {
+		/* Either unlocked or only read locked. */
 
 			return(TRUE);
 		}
 
-		if (rw_lock_get_reader_count(lock) == 0
-		    && rw_lock_get_writer(lock) == RW_LOCK_WAIT_EX
-		    && os_thread_eq(lock->writer_thread, cell->thread)) {
+        } else if (cell->request_type == RW_LOCK_WAIT_EX) {
+
+		lock = cell->wait_object;
+
+                /* lock_word == 0 means all readers have left */
+		if (lock->lock_word == 0) {
 
 			return(TRUE);
 		}
-
 	} else if (cell->request_type == RW_LOCK_SHARED) {
 		lock = cell->wait_object;
 
-		if (rw_lock_get_writer(lock) == RW_LOCK_NOT_LOCKED) {
+                /* lock_word > 0 means no writer or reserved writer */
+		if (lock->lock_word > 0) {
 
 			return(TRUE);
 		}
@@ -844,11 +865,15 @@ sync_array_object_signalled(
 /*========================*/
 	sync_array_t*	arr)	/* in: wait array */
 {
+#ifdef HAVE_GCC_ATOMIC_BUILTINS
+	os_atomic_increment((lint*) &arr->sg_count, 1);
+#else
 	sync_array_enter(arr);
 
 	arr->sg_count++;
 
 	sync_array_exit(arr);
+#endif
 }
 
 /**************************************************************************
@@ -868,6 +893,7 @@ sync_arr_wake_threads_if_sema_free(void)
 	sync_cell_t*	cell;
 	ulint		count;
 	ulint		i;
+	os_event_t      event;
 
 	sync_array_enter(arr);
 
@@ -877,36 +903,20 @@ sync_arr_wake_threads_if_sema_free(void)
 	while (count < arr->n_reserved) {
 
 		cell = sync_array_get_nth_cell(arr, i);
+		i++;
 
-		if (cell->wait_object != NULL) {
-
+		if (cell->wait_object == NULL) {
+			continue;
+		}
 			count++;
 
 			if (sync_arr_cell_can_wake_up(cell)) {
 
-				if (cell->request_type == SYNC_MUTEX) {
-					mutex_t*	mutex;
+			event = sync_cell_get_event(cell);
 
-					mutex = cell->wait_object;
-					os_event_set(mutex->event);
-#ifdef __WIN__
-				} else if (cell->request_type
-					   == RW_LOCK_WAIT_EX) {
-					rw_lock_t*	lock;
-
-					lock = cell->wait_object;
-					os_event_set(lock->wait_ex_event);
-#endif
-				} else {
-					rw_lock_t*	lock;
-
-					lock = cell->wait_object;
-					os_event_set(lock->event);
-				}
-			}
+			os_event_set(event);
 		}
 
-		i++;
 	}
 
 	sync_array_exit(arr);
@@ -1026,4 +1036,3 @@ sync_array_print_info(
 
 	sync_array_exit(arr);
 }
-

sync/sync0sync.c

@@ -5,6 +5,38 @@ Mutex, the basic synchronization primitive
 
 Created 9/5/1995 Heikki Tuuri
 *******************************************************/
+/***********************************************************************
+# Copyright (c) 2008, Google Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#	* Redistributions of source code must retain the above copyright
+#	  notice, this list of conditions and the following disclaimer.
+#	* Redistributions in binary form must reproduce the above
+#	  copyright notice, this list of conditions and the following
+#	  disclaimer in the documentation and/or other materials
+#	  provided with the distribution.
+#	* Neither the name of the Google Inc. nor the names of its
+#	  contributors may be used to endorse or promote products
+#	  derived from this software without specific prior written
+#	  permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+# Note, the BSD license applies to the new code. The old code is GPL.
+***********************************************************************/
 
 #include "sync0sync.h"
 #ifdef UNIV_NONINL
@@ -138,18 +170,13 @@ Therefore, this thread is guaranteed to catch the os_set_event()
 signalled unconditionally at the release of the lock.
 Q.E.D. */
 
-/* The number of system calls made in this module. Intended for performance
-monitoring. */
-
-UNIV_INTERN ulint	mutex_system_call_count		= 0;
-
 /* Number of spin waits on mutexes: for performance monitoring */
 
 /* round=one iteration of a spin loop */
-UNIV_INTERN ulint	mutex_spin_round_count		= 0;
-UNIV_INTERN ulint	mutex_spin_wait_count		= 0;
-UNIV_INTERN ulint	mutex_os_wait_count		= 0;
-UNIV_INTERN ulint	mutex_exit_count		= 0;
+UNIV_INTERN ib_int64_t	mutex_spin_round_count		= 0;
+UNIV_INTERN ib_int64_t	mutex_spin_wait_count		= 0;
+UNIV_INTERN ib_int64_t	mutex_os_wait_count		= 0;
+UNIV_INTERN ib_int64_t	mutex_exit_count		= 0;
 
 /* The global array of wait cells for implementation of the database's own
 mutexes and read-write locks */
@@ -219,6 +246,8 @@ mutex_create_func(
 {
 #if defined(_WIN32) && defined(UNIV_CAN_USE_X86_ASSEMBLER)
 	mutex_reset_lock_word(mutex);
+#elif defined(HAVE_GCC_ATOMIC_BUILTINS)
+	mutex_reset_lock_word(mutex);
 #else
 	os_fast_mutex_init(&(mutex->os_fast_mutex));
 	mutex->lock_word = 0;
@@ -309,7 +338,9 @@ mutex_free(
 
 	os_event_free(mutex->event);
 
-#if !defined(_WIN32) || !defined(UNIV_CAN_USE_X86_ASSEMBLER)
+#if defined(_WIN32) && defined(UNIV_CAN_USE_X86_ASSEMBLER)
+#elif defined(HAVE_GCC_ATOMIC_BUILTINS)
+#else
 	os_fast_mutex_free(&(mutex->os_fast_mutex));
 #endif
 	/* If we free the mutex protecting the mutex list (freeing is
@@ -426,6 +457,12 @@ mutex_spin_wait(
 #endif /* UNIV_DEBUG && !UNIV_HOTBACKUP */
 	ut_ad(mutex);
 
+	/* This update is not thread safe, but we don't mind if the count
+	isn't exact. Moved out of ifdef that follows because we are willing
+	to sacrifice the cost of counting this as the data is valuable.
+	Count the number of calls to mutex_spin_wait. */
+	mutex_spin_wait_count++;
+
 mutex_loop:
 
 	i = 0;
@@ -438,7 +475,6 @@ mutex_spin_wait(
 
 spin_loop:
 #if defined UNIV_DEBUG && !defined UNIV_HOTBACKUP
-	mutex_spin_wait_count++;
 	mutex->count_spin_loop++;
 #endif /* UNIV_DEBUG && !UNIV_HOTBACKUP */
 
@@ -503,8 +539,6 @@ mutex_spin_wait(
 	sync_array_reserve_cell(sync_primary_wait_array, mutex,
 				SYNC_MUTEX, file_name, line, &index);
 
-	mutex_system_call_count++;
-
 	/* The memory order of the array reservation and the change in the
 	waiters field is important: when we suspend a thread, we first
 	reserve the cell and then set waiters field to 1. When threads are
@@ -551,7 +585,6 @@ mutex_spin_wait(
 		mutex->cfile_name, (ulong) mutex->cline, (ulong) i);
 #endif
 
-	mutex_system_call_count++;
 	mutex_os_wait_count++;
 
 #ifndef UNIV_HOTBACKUP
@@ -1340,21 +1373,31 @@ sync_print_wait_info(
 	FILE*	file)		/* in: file where to print */
 {
 #ifdef UNIV_SYNC_DEBUG
-	fprintf(file, "Mutex exits %lu, rws exits %lu, rwx exits %lu\n",
+	fprintf(file, "Mutex exits %llu, rws exits %llu, rwx exits %llu\n",
 		mutex_exit_count, rw_s_exit_count, rw_x_exit_count);
 #endif
 
 	fprintf(file,
-		"Mutex spin waits %lu, rounds %lu, OS waits %lu\n"
-		"RW-shared spins %lu, OS waits %lu;"
-		" RW-excl spins %lu, OS waits %lu\n",
-		(ulong) mutex_spin_wait_count,
-		(ulong) mutex_spin_round_count,
-		(ulong) mutex_os_wait_count,
-		(ulong) rw_s_spin_wait_count,
-		(ulong) rw_s_os_wait_count,
-		(ulong) rw_x_spin_wait_count,
-		(ulong) rw_x_os_wait_count);
+		"Mutex spin waits %llu, rounds %llu, OS waits %llu\n"
+		"RW-shared spins %llu, OS waits %llu;"
+		" RW-excl spins %llu, OS waits %llu\n",
+		mutex_spin_wait_count,
+		mutex_spin_round_count,
+		mutex_os_wait_count,
+		rw_s_spin_wait_count,
+		rw_s_os_wait_count,
+		rw_x_spin_wait_count,
+		rw_x_os_wait_count);
+
+	fprintf(file,
+		"Spin rounds per wait: %.2f mutex, %.2f RW-shared, "
+		"%.2f RW-excl\n",
+		(double) mutex_spin_round_count /
+		(mutex_spin_wait_count ? mutex_spin_wait_count : 1),
+		(double) rw_s_spin_round_count /
+		(rw_s_spin_wait_count ? rw_s_spin_wait_count : 1),
+		(double) rw_x_spin_round_count /
+		(rw_x_spin_wait_count ? rw_x_spin_wait_count : 1));
 }
 
 /***********************************************************************

ut/ut0mem.c

@@ -15,6 +15,7 @@ Created 5/11/1994 Heikki Tuuri
 #include "mem0mem.h"
 #include "os0sync.h"
 #include "os0thread.h"
+#include "srv0srv.h"
 
 /* This struct is placed first in every allocated memory block */
 typedef struct ut_mem_block_struct ut_mem_block_t;
@@ -68,14 +69,29 @@ ut_malloc_low(
 	ibool	assert_on_error)/* in: if TRUE, we crash mysqld if the
 				memory cannot be allocated */
 {
-	ulint	retry_count	= 0;
+	ulint	retry_count;
 	void*	ret;
 
+	if (srv_use_sys_malloc) {
+		ret = malloc(n);
+		ut_a(ret || !assert_on_error);
+
+#ifdef UNIV_SET_MEM_TO_ZERO
+		if (set_to_zero) {
+			memset(ret, '\0', n);
+			UNIV_MEM_ALLOC(ret, n);
+		}
+#endif
+		return(ret);
+	}
+
 	ut_ad((sizeof(ut_mem_block_t) % 8) == 0); /* check alignment ok */
 
-	if (!ut_mem_block_list_inited) {
+	if (UNIV_UNLIKELY(!ut_mem_block_list_inited)) {
 		ut_mem_block_list_init();
 	}
+
+	retry_count = 0;
 retry:
 	os_fast_mutex_lock(&ut_list_mutex);
 
@@ -239,6 +255,11 @@ ut_free(
 {
 	ut_mem_block_t* block;
 
+	if (srv_use_sys_malloc) {
+		free(ptr);
+		return;
+	}
+
 	block = (ut_mem_block_t*)((byte*)ptr - sizeof(ut_mem_block_t));
 
 	os_fast_mutex_lock(&ut_list_mutex);
@@ -291,6 +312,10 @@ ut_realloc(
 	ulint		min_size;
 	void*		new_ptr;
 
+	if (srv_use_sys_malloc) {
+		return(realloc(ptr, size));
+	}
+
 	if (ptr == NULL) {
 
 		return(ut_malloc(size));
@@ -338,6 +363,11 @@ ut_free_all_mem(void)
 {
 	ut_mem_block_t* block;
 
+	if (!ut_mem_block_list_inited) {
+		return;
+	}
+
+	ut_mem_block_list_inited = FALSE;
 	os_fast_mutex_free(&ut_list_mutex);
 
 	while ((block = UT_LIST_GET_FIRST(ut_mem_block_list))) {