sync with extensions rev 23

buf/buf0buddy.c

@@ -162,7 +162,8 @@ static
 void
 buf_buddy_block_free(
 /*=================*/
-	void*	buf)	/* in: buffer frame to deallocate */
+	void*	buf,	/* in: buffer frame to deallocate */
+	ibool	have_page_hash_mutex)
 {
 	const ulint	fold	= BUF_POOL_ZIP_FOLD_PTR(buf);
 	buf_page_t*	bpage;
@@ -190,7 +191,7 @@ buf_buddy_block_free(
 
 	block = (buf_block_t*) bpage;
 	mutex_enter(&block->mutex);
-	buf_LRU_block_free_non_file_page(block);
+	buf_LRU_block_free_non_file_page(block, have_page_hash_mutex);
 	mutex_exit(&block->mutex);
 
 	ut_ad(buf_buddy_n_frames > 0);
@@ -274,10 +275,11 @@ buf_buddy_alloc_low(
 			possibly NULL if lru==NULL */
 	ulint	i,	/* in: index of buf_pool->zip_free[],
 			or BUF_BUDDY_SIZES */
-	ibool*	lru)	/* in: pointer to a variable that will be assigned
+	ibool*	lru,	/* in: pointer to a variable that will be assigned
 			TRUE if storage was allocated from the LRU list
 			and buf_pool_mutex was temporarily released,
 			or NULL if the LRU list should not be used */
+	ibool	have_page_hash_mutex)
 {
 	buf_block_t*	block;
 
@@ -312,9 +314,15 @@ buf_buddy_alloc_low(
 
 	/* Try replacing an uncompressed page in the buffer pool. */
 	//buf_pool_mutex_exit();
+	mutex_exit(&LRU_list_mutex);
+	if (have_page_hash_mutex)
+		mutex_exit(&page_hash_mutex);
 	block = buf_LRU_get_free_block(0);
 	*lru = TRUE;
 	//buf_pool_mutex_enter();
+	mutex_enter(&LRU_list_mutex);
+	if (have_page_hash_mutex)
+		mutex_enter(&page_hash_mutex);
 
 alloc_big:
 	buf_buddy_block_register(block);
@@ -407,7 +415,8 @@ buf_buddy_relocate(
 			/* out: TRUE if relocated */
 	void*	src,	/* in: block to relocate */
 	void*	dst,	/* in: free block to relocate to */
-	ulint	i)	/* in: index of buf_pool->zip_free[] */
+	ulint	i,	/* in: index of buf_pool->zip_free[] */
+	ibool	have_page_hash_mutex)
 {
 	buf_page_t*	bpage;
 	const ulint	size	= BUF_BUDDY_LOW << i;
@@ -433,12 +442,14 @@ buf_buddy_relocate(
 	actually is a properly initialized buf_page_t object. */
 
 	if (size >= PAGE_ZIP_MIN_SIZE) {
-		mutex_exit(&zip_free_mutex);
+		if (!have_page_hash_mutex)
+			mutex_exit(&zip_free_mutex);
 
 		/* This is a compressed page. */
 		mutex_t*	mutex;
 
-		mutex_enter(&page_hash_mutex);
+		if (!have_page_hash_mutex)
+			mutex_enter(&page_hash_mutex);
 		/* The src block may be split into smaller blocks,
 		some of which may be free.  Thus, the
 		mach_read_from_4() calls below may attempt to read
@@ -459,8 +470,10 @@ buf_buddy_relocate(
 			added to buf_pool->page_hash yet.  Obviously,
 			it cannot be relocated. */
 
-			mutex_exit(&page_hash_mutex);
-			mutex_enter(&zip_free_mutex);
+			if (!have_page_hash_mutex) {
+				mutex_exit(&page_hash_mutex);
+				mutex_enter(&zip_free_mutex);
+			}
 			return(FALSE);
 		}
 
@@ -470,11 +483,17 @@ buf_buddy_relocate(
 			For the sake of simplicity, give up. */
 			ut_ad(page_zip_get_size(&bpage->zip) < size);
 
-			mutex_exit(&page_hash_mutex);
-			mutex_enter(&zip_free_mutex);
+			if (!have_page_hash_mutex) {
+				mutex_exit(&page_hash_mutex);
+				mutex_enter(&zip_free_mutex);
+			}
 			return(FALSE);
 		}
 
+		/* To keep latch order */
+		if (have_page_hash_mutex)
+			mutex_exit(&zip_free_mutex);
+
 		/* The block must have been allocated, but it may
 		contain uninitialized data. */
 		UNIV_MEM_ASSERT_W(src, size);
@@ -482,7 +501,8 @@ buf_buddy_relocate(
 		mutex = buf_page_get_mutex(bpage);
 
 		mutex_enter(mutex);
-		mutex_exit(&page_hash_mutex);
+		if (!have_page_hash_mutex)
+			mutex_exit(&page_hash_mutex);
 
 		if (buf_page_can_relocate(bpage)) {
 			/* Relocate the compressed page. */
@@ -529,7 +549,8 @@ buf_buddy_free_low(
 /*===============*/
 	void*	buf,	/* in: block to be freed, must not be
 			pointed to by the buffer pool */
-	ulint	i)	/* in: index of buf_pool->zip_free[] */
+	ulint	i,	/* in: index of buf_pool->zip_free[] */
+	ibool	have_page_hash_mutex)
 {
 	buf_page_t*	bpage;
 	buf_page_t*	buddy;
@@ -546,7 +567,7 @@ buf_buddy_free_low(
 	ut_d(((buf_page_t*) buf)->state = BUF_BLOCK_ZIP_FREE);
 
 	if (i == BUF_BUDDY_SIZES) {
-		buf_buddy_block_free(buf);
+		buf_buddy_block_free(buf, have_page_hash_mutex);
 		return;
 	}
 
@@ -614,7 +635,7 @@ buf_buddy_free_low(
 		buf_buddy_remove_from_free(bpage, i);
 
 		/* Try to relocate the buddy of buf to the free block. */
-		if (buf_buddy_relocate(buddy, bpage, i)) {
+		if (buf_buddy_relocate(buddy, bpage, i, have_page_hash_mutex)) {
 
 			ut_d(buddy->state = BUF_BLOCK_ZIP_FREE);
 			goto buddy_free2;
@@ -642,7 +663,7 @@ buf_buddy_free_low(
 		}
 #endif /* UNIV_DEBUG && !UNIV_DEBUG_VALGRIND */
 
-		if (buf_buddy_relocate(buddy, buf, i)) {
+		if (buf_buddy_relocate(buddy, buf, i, have_page_hash_mutex)) {
 
 			buf = bpage;
 			UNIV_MEM_VALID(bpage, BUF_BUDDY_LOW << i);

buf/buf0buf.c

@@ -1863,6 +1863,7 @@ buf_page_get_gen(
 	ibool		accessed;
 	ulint		fix_type;
 	ibool		must_read;
+	mutex_t*	block_mutex;
 
 	ut_ad(mtr);
 	ut_ad((rw_latch == RW_S_LATCH)
@@ -1881,7 +1882,8 @@ buf_page_get_gen(
 	//buf_pool_mutex_enter();
 
 	if (block) {
-		mutex_enter(&block->mutex);
+		block_mutex = buf_page_get_mutex(block);
+		mutex_enter(block_mutex);
 
 		/* If the guess is a compressed page descriptor that
 		has been allocated by buf_buddy_alloc(), it may have
@@ -1896,7 +1898,7 @@ buf_page_get_gen(
 		    || space != block->page.space
 		    || buf_block_get_state(block) != BUF_BLOCK_FILE_PAGE) {
 
-			mutex_exit(&block->mutex);
+			mutex_exit(block_mutex);
 
 			block = guess = NULL;
 		} else {
@@ -1909,7 +1911,8 @@ buf_page_get_gen(
 		mutex_enter(&page_hash_mutex);
 		block = (buf_block_t*) buf_page_hash_get(space, offset);
 		if (block) {
-			mutex_enter(&block->mutex);
+			block_mutex = buf_page_get_mutex(block);
+			mutex_enter(block_mutex);
 		}
 		mutex_exit(&page_hash_mutex);
 	}
@@ -1940,7 +1943,7 @@ buf_page_get_gen(
 	if (must_read && mode == BUF_GET_IF_IN_POOL) {
 		/* The page is only being read to buffer */
 		//buf_pool_mutex_exit();
-		mutex_exit(&block->mutex);
+		mutex_exit(block_mutex);
 
 		return(NULL);
 	}
@@ -1965,7 +1968,7 @@ buf_page_get_gen(
 			/* The block is buffer-fixed or I/O-fixed.
 			Try again later. */
 			//buf_pool_mutex_exit();
-			mutex_exit(&block->mutex);
+			mutex_exit(block_mutex);
 			os_thread_sleep(WAIT_FOR_READ);
 
 			goto loop;
@@ -1973,16 +1976,17 @@ buf_page_get_gen(
 
 		/* Allocate an uncompressed page. */
 		//buf_pool_mutex_exit();
-		mutex_exit(&block->mutex);
+		mutex_exit(block_mutex);
 
 		block = buf_LRU_get_free_block(0);
 		ut_a(block);
+		block_mutex = buf_page_get_mutex(block);
 
 		//buf_pool_mutex_enter();
 		mutex_enter(&flush_list_mutex);
 		mutex_enter(&LRU_list_mutex);
 		mutex_enter(&page_hash_mutex);
-		mutex_enter(&block->mutex);
+		mutex_enter(block_mutex);
 
 		{
 			buf_page_t*	hash_bpage
@@ -1993,12 +1997,13 @@ buf_page_get_gen(
 				while buf_pool_mutex was released.
 				Free the block that was allocated. */
 
-				buf_LRU_block_free_non_file_page(block);
-				mutex_exit(&block->mutex);
+				buf_LRU_block_free_non_file_page(block, TRUE);
+				mutex_exit(block_mutex);
 
 				block = (buf_block_t*) hash_bpage;
+				block_mutex = buf_page_get_mutex(block);
 
-				mutex_enter(&block->mutex);
+				mutex_enter(block_mutex);
 				mutex_exit(&page_hash_mutex);
 				mutex_exit(&LRU_list_mutex);
 				mutex_exit(&flush_list_mutex);
@@ -2015,7 +2020,7 @@ buf_page_get_gen(
 			Free the block that was allocated and try again.
 			This should be extremely unlikely. */
 
-			buf_LRU_block_free_non_file_page(block);
+			buf_LRU_block_free_non_file_page(block, TRUE);
 			//mutex_exit(&block->mutex);
 
 			mutex_exit(&page_hash_mutex);
@@ -2084,10 +2089,10 @@ buf_page_get_gen(
 		mutex_exit(&buf_pool_mutex);
 
 		rw_lock_x_lock(&block->lock);
-		mutex_exit(&block->mutex);
+		mutex_exit(block_mutex);
 		mutex_exit(&buf_pool_zip_mutex);
 
-		buf_buddy_free(bpage, sizeof *bpage);
+		buf_buddy_free(bpage, sizeof *bpage, FALSE);
 
 		//buf_pool_mutex_exit();
 
@@ -2102,7 +2107,7 @@ buf_page_get_gen(
 
 		/* Unfix and unlatch the block. */
 		//buf_pool_mutex_enter();
-		mutex_enter(&block->mutex);
+		mutex_enter(block_mutex);
 		mutex_enter(&buf_pool_mutex);
 		buf_pool->n_pend_unzip--;
 		mutex_exit(&buf_pool_mutex);
@@ -2143,7 +2148,7 @@ buf_page_get_gen(
 
 	buf_page_set_accessed(&block->page, TRUE);
 
-	mutex_exit(&block->mutex);
+	mutex_exit(block_mutex);
 
 	buf_block_make_young(&block->page);
 
@@ -2694,7 +2699,7 @@ buf_page_init_for_read(
 err_exit:
 		if (block) {
 			mutex_enter(&block->mutex);
-			buf_LRU_block_free_non_file_page(block);
+			buf_LRU_block_free_non_file_page(block, TRUE);
 			mutex_exit(&block->mutex);
 		}
 
@@ -2753,9 +2758,9 @@ buf_page_init_for_read(
 			operation until after the block descriptor has
 			been added to buf_pool->LRU and
 			buf_pool->page_hash. */
-			//mutex_exit(&block->mutex); /*##may be able to removed?##*/
-			data = buf_buddy_alloc(zip_size, &lru);
-			//mutex_enter(&block->mutex); /*##may be able to removed?##*/
+			mutex_exit(&block->mutex);
+			data = buf_buddy_alloc(zip_size, &lru, FALSE);
+			mutex_enter(&block->mutex);
 			block->page.zip.data = data;
 
 			/* To maintain the invariant
@@ -2779,8 +2784,8 @@ buf_page_init_for_read(
 		control block (bpage), in order to avoid the
 		invocation of buf_buddy_relocate_block() on
 		uninitialized data. */
-		data = buf_buddy_alloc(zip_size, &lru);
-		bpage = buf_buddy_alloc(sizeof *bpage, &lru);
+		data = buf_buddy_alloc(zip_size, &lru, TRUE);
+		bpage = buf_buddy_alloc(sizeof *bpage, &lru, TRUE);
 
 		/* If buf_buddy_alloc() allocated storage from the LRU list,
 		it released and reacquired buf_pool_mutex.  Thus, we must
@@ -2931,15 +2936,15 @@ buf_page_create(
 		rw_lock_x_lock(&block->lock);
 
 		page_zip_set_size(&block->page.zip, zip_size);
-		//mutex_exit(&block->mutex); /*##may be able to removed?##*/
+		mutex_exit(&block->mutex);
 		/* buf_pool_mutex may be released and reacquired by
 		buf_buddy_alloc().  Thus, we must release block->mutex
 		in order not to break the latching order in
 		the reacquisition of buf_pool_mutex.  We also must
 		defer this operation until after the block descriptor
 		has been added to buf_pool->LRU and buf_pool->page_hash. */
-		data = buf_buddy_alloc(zip_size, &lru);
-		//mutex_enter(&block->mutex); /*##may be able to removed?##*/
+		data = buf_buddy_alloc(zip_size, &lru, FALSE);
+		mutex_enter(&block->mutex);
 		block->page.zip.data = data;
 
 		/* To maintain the invariant

buf/buf0lru.c

@@ -112,8 +112,9 @@ static
 void
 buf_LRU_block_free_hashed_page(
 /*===========================*/
-	buf_block_t*	block);	/* in: block, must contain a file page and
+	buf_block_t*	block,	/* in: block, must contain a file page and
 				be in a state where it can be freed */
+	ibool		have_page_hash_mutex);
 
 /**********************************************************************
 Determines if the unzip_LRU list should be used for evicting a victim
@@ -389,7 +390,7 @@ buf_LRU_invalidate_tablespace(
 			if (buf_LRU_block_remove_hashed_page(bpage, TRUE)
 			    != BUF_BLOCK_ZIP_FREE) {
 				buf_LRU_block_free_hashed_page((buf_block_t*)
-							       bpage);
+							       bpage, TRUE);
 			} else {
 				/* The compressed block descriptor
 				(bpage) has been deallocated and
@@ -860,7 +861,7 @@ buf_LRU_get_free_block(
 		if (UNIV_UNLIKELY(zip_size)) {
 			ibool	lru;
 			page_zip_set_size(&block->page.zip, zip_size);
-			block->page.zip.data = buf_buddy_alloc(zip_size, &lru);
+			block->page.zip.data = buf_buddy_alloc(zip_size, &lru, FALSE);
 			UNIV_MEM_DESC(block->page.zip.data, zip_size, block);
 		} else {
 			page_zip_set_size(&block->page.zip, 0);
@@ -1392,7 +1393,7 @@ buf_LRU_free_block(
 		from the LRU list), refuse to free bpage. */
 alloc:
 		buf_pool_mutex_exit_forbid();
-		b = buf_buddy_alloc(sizeof *b, NULL);
+		b = buf_buddy_alloc(sizeof *b, NULL, FALSE);
 		buf_pool_mutex_exit_allow();
 
 		if (UNIV_UNLIKELY(!b)) {
@@ -1424,7 +1425,7 @@ buf_LRU_free_block(
 	    || ( bpage->oldest_modification
 		 && buf_page_get_state(bpage) != BUF_BLOCK_FILE_PAGE ) ) {
 		if (b) {
-			buf_buddy_free(b, sizeof *b);
+			buf_buddy_free(b, sizeof *b, TRUE);
 		}
 		mutex_exit(&LRU_list_mutex);
 		mutex_exit(&page_hash_mutex);
@@ -1599,7 +1600,7 @@ buf_LRU_free_block(
 			mutex_exit(&buf_pool_zip_mutex);
 		}
 
-		buf_LRU_block_free_hashed_page((buf_block_t*) bpage);
+		buf_LRU_block_free_hashed_page((buf_block_t*) bpage, FALSE);
 	} else {
 		/* It may be bug of 1.0.2 */
 		//mutex_enter(block_mutex);
@@ -1614,7 +1615,8 @@ UNIV_INTERN
 void
 buf_LRU_block_free_non_file_page(
 /*=============================*/
-	buf_block_t*	block)	/* in: block, must not contain a file page */
+	buf_block_t*	block,	/* in: block, must not contain a file page */
+	ibool		have_page_hash_mutex)
 {
 	void*	data;
 
@@ -1652,7 +1654,7 @@ buf_LRU_block_free_non_file_page(
 		block->page.zip.data = NULL;
 		mutex_exit(&block->mutex);
 		buf_pool_mutex_exit_forbid();
-		buf_buddy_free(data, page_zip_get_size(&block->page.zip));
+		buf_buddy_free(data, page_zip_get_size(&block->page.zip), have_page_hash_mutex);
 		buf_pool_mutex_exit_allow();
 		mutex_enter(&block->mutex);
 		page_zip_set_size(&block->page.zip, 0);
@@ -1823,8 +1825,8 @@ buf_LRU_block_remove_hashed_page(
 		mutex_exit(&buf_pool_zip_mutex);
 		buf_pool_mutex_exit_forbid();
 		buf_buddy_free(bpage->zip.data,
-			       page_zip_get_size(&bpage->zip));
-		buf_buddy_free(bpage, sizeof(*bpage));
+			       page_zip_get_size(&bpage->zip), TRUE);
+		buf_buddy_free(bpage, sizeof(*bpage), TRUE);
 		buf_pool_mutex_exit_allow();
 		UNIV_MEM_UNDESC(bpage);
 		return(BUF_BLOCK_ZIP_FREE);
@@ -1844,7 +1846,7 @@ buf_LRU_block_remove_hashed_page(
 			bpage->zip.data = NULL;
 
 			mutex_exit(&((buf_block_t*) bpage)->mutex);
-			buf_buddy_free(data, page_zip_get_size(&bpage->zip));
+			buf_buddy_free(data, page_zip_get_size(&bpage->zip), TRUE);
 			mutex_enter(&((buf_block_t*) bpage)->mutex);
 			page_zip_set_size(&bpage->zip, 0);
 		}
@@ -1870,15 +1872,16 @@ static
 void
 buf_LRU_block_free_hashed_page(
 /*===========================*/
-	buf_block_t*	block)	/* in: block, must contain a file page and
+	buf_block_t*	block,	/* in: block, must contain a file page and
 				be in a state where it can be freed */
+	ibool		have_page_hash_mutex)
 {
 	//ut_ad(buf_pool_mutex_own());
 	ut_ad(mutex_own(&block->mutex));
 
 	buf_block_set_state(block, BUF_BLOCK_MEMORY);
 
-	buf_LRU_block_free_non_file_page(block);
+	buf_LRU_block_free_non_file_page(block, have_page_hash_mutex);
 }
 
 /************************************************************************

handler/ha_innodb.cc

@@ -167,6 +167,7 @@ static char*	innobase_log_arch_dir			= NULL;
 #endif /* UNIV_LOG_ARCHIVE */
 static my_bool	innobase_use_doublewrite		= TRUE;
 static my_bool	innobase_use_checksums			= TRUE;
+static my_bool	innobase_extra_undoslots		= FALSE;
 static my_bool	innobase_locks_unsafe_for_binlog	= FALSE;
 static my_bool	innobase_rollback_on_timeout		= FALSE;
 static my_bool	innobase_create_status_file		= FALSE;
@@ -1960,6 +1961,8 @@ innobase_init(
 		goto error;
 	}
 
+	srv_extra_undoslots = (ibool) innobase_extra_undoslots;
+
 	/* -------------- Log files ---------------------------*/
 
 	/* The default dir for log files is the datadir of MySQL */
@@ -9357,6 +9360,13 @@ static MYSQL_SYSVAR_STR(data_home_dir, innobase_data_home_dir,
   "The common part for InnoDB table spaces.",
   NULL, NULL, NULL);
 
+static MYSQL_SYSVAR_BOOL(extra_undoslots, innobase_extra_undoslots,
+  PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY,
+  "Enable to use about 4000 undo slots instead of default 1024. "
+  "#### Attention: Once you enable this parameter, "
+  "don't use the datafile for normal mysqld or ibbackup! ####",
+  NULL, NULL, FALSE);
+
 static MYSQL_SYSVAR_BOOL(doublewrite, innobase_use_doublewrite,
   PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY,
   "Enable InnoDB doublewrite buffer (enabled by default). "
@@ -9606,6 +9616,7 @@ static struct st_mysql_sys_var* innobase_system_variables[]= {
   MYSQL_SYSVAR(data_file_path),
   MYSQL_SYSVAR(data_home_dir),
   MYSQL_SYSVAR(doublewrite),
+  MYSQL_SYSVAR(extra_undoslots),
   MYSQL_SYSVAR(fast_shutdown),
   MYSQL_SYSVAR(file_io_threads),
   MYSQL_SYSVAR(file_per_table),

handler/innodb_patch_info.h

@@ -30,5 +30,6 @@ struct innodb_enhancement {
 {"innodb_opt_lru_count","Fix of buffer_pool mutex","Decreases contention on buffer_pool mutex on LRU operations","http://www.percona.com/docs/wiki/percona-xtradb:innodb_opt_lru_count"},
 {"innodb_buffer_pool_pages","Information of buffer pool content","","http://www.percona.com/docs/wiki/percona-xtradb:innodb_buffer_pool_pages"},
 {"innodb_split_buf_pool_mutex","More fix of buffer_pool mutex","Spliting buf_pool_mutex and optimizing based on innodb_opt_lru_count","http://www.percona.com/docs/wiki/percona-xtradb:innodb_split_buf_pool_mutex"},
+{"innodb_expand_undo_slots","expandable maximum number of undo slots","from 1024 (default) to about 4000","http://www.percona.com/docs/wiki/percona-xtradb:innodb_expand_undo_slots"},
 {NULL, NULL, NULL, NULL}
 };

include/buf0buddy.h

@@ -33,10 +33,11 @@ buf_buddy_alloc(
 			/* out: allocated block,
 			possibly NULL if lru == NULL */
 	ulint	size,	/* in: block size, up to UNIV_PAGE_SIZE */
-	ibool*	lru)	/* in: pointer to a variable that will be assigned
+	ibool*	lru,	/* in: pointer to a variable that will be assigned
 			TRUE if storage was allocated from the LRU list
 			and buf_pool_mutex was temporarily released,
 			or NULL if the LRU list should not be used */
+	ibool	have_page_hash_mutex)
 	__attribute__((malloc));
 
 /**************************************************************************
@@ -47,7 +48,8 @@ buf_buddy_free(
 /*===========*/
 	void*	buf,	/* in: block to be freed, must not be
 			pointed to by the buffer pool */
-	ulint	size)	/* in: block size, up to UNIV_PAGE_SIZE */
+	ulint	size,	/* in: block size, up to UNIV_PAGE_SIZE */
+	ibool	have_page_hash_mutex)
 	__attribute__((nonnull));
 
 /** Statistics of buddy blocks of a given size. */

include/buf0buddy.ic

@@ -28,10 +28,11 @@ buf_buddy_alloc_low(
 			possibly NULL if lru==NULL */
 	ulint	i,	/* in: index of buf_pool->zip_free[],
 			or BUF_BUDDY_SIZES */
-	ibool*	lru)	/* in: pointer to a variable that will be assigned
+	ibool*	lru,	/* in: pointer to a variable that will be assigned
 			TRUE if storage was allocated from the LRU list
 			and buf_pool_mutex was temporarily released,
 			or NULL if the LRU list should not be used */
+	ibool	have_page_hash_mutex)
 	__attribute__((malloc));
 
 /**************************************************************************
@@ -42,8 +43,9 @@ buf_buddy_free_low(
 /*===============*/
 	void*	buf,	/* in: block to be freed, must not be
 			pointed to by the buffer pool */
-	ulint	i)	/* in: index of buf_pool->zip_free[],
+	ulint	i,	/* in: index of buf_pool->zip_free[],
 			or BUF_BUDDY_SIZES */
+	ibool	have_page_hash_mutex)
 	__attribute__((nonnull));
 
 /**************************************************************************
@@ -81,14 +83,15 @@ buf_buddy_alloc(
 			/* out: allocated block,
 			possibly NULL if lru == NULL */
 	ulint	size,	/* in: block size, up to UNIV_PAGE_SIZE */
-	ibool*	lru)	/* in: pointer to a variable that will be assigned
+	ibool*	lru,	/* in: pointer to a variable that will be assigned
 			TRUE if storage was allocated from the LRU list
 			and buf_pool_mutex was temporarily released,
 			or NULL if the LRU list should not be used */
+	ibool	have_page_hash_mutex)
 {
 	//ut_ad(buf_pool_mutex_own());
 
-	return(buf_buddy_alloc_low(buf_buddy_get_slot(size), lru));
+	return(buf_buddy_alloc_low(buf_buddy_get_slot(size), lru, have_page_hash_mutex));
 }
 
 /**************************************************************************
@@ -99,12 +102,13 @@ buf_buddy_free(
 /*===========*/
 	void*	buf,	/* in: block to be freed, must not be
 			pointed to by the buffer pool */
-	ulint	size)	/* in: block size, up to UNIV_PAGE_SIZE */
+	ulint	size,	/* in: block size, up to UNIV_PAGE_SIZE */
+	ibool	have_page_hash_mutex)
 {
 	//ut_ad(buf_pool_mutex_own());
 
 	mutex_enter(&zip_free_mutex);
-	buf_buddy_free_low(buf, buf_buddy_get_slot(size));
+	buf_buddy_free_low(buf, buf_buddy_get_slot(size), have_page_hash_mutex);
 	mutex_exit(&zip_free_mutex);
 }
 

include/buf0buf.ic

@@ -736,7 +736,7 @@ buf_block_free(
 
 	ut_a(buf_block_get_state(block) != BUF_BLOCK_FILE_PAGE);
 
-	buf_LRU_block_free_non_file_page(block);
+	buf_LRU_block_free_non_file_page(block, FALSE);
 
 	mutex_exit(&block->mutex);
 

include/buf0lru.h

@@ -147,7 +147,8 @@ UNIV_INTERN
 void
 buf_LRU_block_free_non_file_page(
 /*=============================*/
-	buf_block_t*	block);	/* in: block, must not contain a file page */
+	buf_block_t*	block,	/* in: block, must not contain a file page */
+	ibool		have_page_hash_mutex);
 /**********************************************************************
 Adds a block to the LRU list. */
 UNIV_INTERN

include/srv0srv.h

@@ -73,6 +73,8 @@ extern char**	srv_data_file_names;
 extern ulint*	srv_data_file_sizes;
 extern ulint*	srv_data_file_is_raw_partition;
 
+extern ibool	srv_extra_undoslots;
+
 extern ibool	srv_auto_extend_last_data_file;
 extern ulint	srv_last_file_size_max;
 extern ulong	srv_auto_extend_increment;

include/trx0rseg.h

@@ -115,8 +115,11 @@ trx_rseg_create(
 	mtr_t*	mtr);		/* in: mtr */
 
 
+/* Real max value may be 4076 in usual. But reserve 4 slot for safety or etc... */
+#define TRX_RSEG_N_EXTRA_SLOTS	(((UNIV_PAGE_SIZE - (FIL_PAGE_DATA + FIL_PAGE_DATA_END + TRX_RSEG_UNDO_SLOTS)) / TRX_RSEG_SLOT_SIZE) - 4)
+
 /* Number of undo log slots in a rollback segment file copy */
-#define TRX_RSEG_N_SLOTS	(UNIV_PAGE_SIZE / 16)
+#define TRX_RSEG_N_SLOTS	(srv_extra_undoslots ? TRX_RSEG_N_EXTRA_SLOTS : (UNIV_PAGE_SIZE / 16))
 
 /* Maximum number of transactions supported by a single rollback segment */
 #define TRX_RSEG_MAX_N_TRXS	(TRX_RSEG_N_SLOTS / 2)

include/univ.i

@@ -12,7 +12,7 @@ Created 1/20/1994 Heikki Tuuri
 #define INNODB_VERSION_MAJOR	1
 #define INNODB_VERSION_MINOR	0
 #define INNODB_VERSION_BUGFIX	2
-#define PERCONA_INNODB_VERSION	2
+#define PERCONA_INNODB_VERSION	1
 
 /* The following is the InnoDB version as shown in
 SELECT plugin_version FROM information_schema.plugins;

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 @@
+@@ -64,6 +64,17 @@
  TRIGGERS
  USER_PRIVILEGES
  VIEWS
-+INNODB_CMP_RESET
++INNODB_CMP
++INNODB_BUFFER_POOL_PAGES
++INNODB_CMPMEM
++INNODB_BUFFER_POOL_PAGES_INDEX
++XTRADB_ENHANCEMENTS
 +INNODB_TRX
-+INNODB_CMPMEM_RESET
++INNODB_BUFFER_POOL_PAGES_BLOB
 +INNODB_LOCK_WAITS
-+INNODB_CMPMEM
-+INNODB_CMP
++INNODB_CMP_RESET
 +INNODB_LOCKS
++INNODB_CMPMEM_RESET
  columns_priv
  db
  event
-@@ -795,6 +802,8 @@
+@@ -795,6 +806,8 @@
  TABLES	UPDATE_TIME	datetime
  TABLES	CHECK_TIME	datetime
  TRIGGERS	CREATED	datetime
@@ -24,47 +28,74 @@ 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 @@
+@@ -843,12 +856,13 @@
+ TABLE_CONSTRAINTS	TABLE_NAME	select
+ TABLE_PRIVILEGES	TABLE_NAME	select
+ VIEWS	TABLE_NAME	select
++INNODB_BUFFER_POOL_PAGES_INDEX	table_name	select
+ delete from mysql.user where user='mysqltest_4';
+ delete from mysql.db where user='mysqltest_4';
  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;
  table_schema	count(*)
 -information_schema	28
-+information_schema	35
++information_schema	39
  mysql	22
  create table t1 (i int, j int);
  create trigger trg1 before insert on t1 for each row
-@@ -1263,6 +1272,13 @@
+@@ -1263,6 +1277,17 @@
  TRIGGERS	TRIGGER_SCHEMA
  USER_PRIVILEGES	GRANTEE
  VIEWS	TABLE_SCHEMA
-+INNODB_CMP_RESET	page_size
++INNODB_CMP	page_size
++INNODB_BUFFER_POOL_PAGES	page_type
++INNODB_CMPMEM	page_size
++INNODB_BUFFER_POOL_PAGES_INDEX	schema_name
++XTRADB_ENHANCEMENTS	name
 +INNODB_TRX	trx_id
-+INNODB_CMPMEM_RESET	page_size
++INNODB_BUFFER_POOL_PAGES_BLOB	space_id
 +INNODB_LOCK_WAITS	requesting_trx_id
-+INNODB_CMPMEM	page_size
-+INNODB_CMP	page_size
++INNODB_CMP_RESET	page_size
 +INNODB_LOCKS	lock_id
++INNODB_CMPMEM_RESET	page_size
  SELECT t.table_name, c1.column_name
  FROM information_schema.tables t
  INNER JOIN
-@@ -1306,6 +1322,13 @@
+@@ -1306,14 +1331,25 @@
  TRIGGERS	TRIGGER_SCHEMA
  USER_PRIVILEGES	GRANTEE
  VIEWS	TABLE_SCHEMA
-+INNODB_CMP_RESET	page_size
++INNODB_CMP	page_size
++INNODB_BUFFER_POOL_PAGES	page_type
++INNODB_CMPMEM	page_size
++INNODB_BUFFER_POOL_PAGES_INDEX	schema_name
++XTRADB_ENHANCEMENTS	name
 +INNODB_TRX	trx_id
-+INNODB_CMPMEM_RESET	page_size
++INNODB_BUFFER_POOL_PAGES_BLOB	space_id
 +INNODB_LOCK_WAITS	requesting_trx_id
-+INNODB_CMPMEM	page_size
-+INNODB_CMP	page_size
++INNODB_CMP_RESET	page_size
 +INNODB_LOCKS	lock_id
++INNODB_CMPMEM_RESET	page_size
  SELECT MAX(table_name) FROM information_schema.tables;
  MAX(table_name)
- VIEWS
-@@ -1382,6 +1405,13 @@
+-VIEWS
++XTRADB_ENHANCEMENTS
+ SELECT table_name from information_schema.tables
+ WHERE table_name=(SELECT MAX(table_name)
+ FROM information_schema.tables);
+ table_name
+-VIEWS
++XTRADB_ENHANCEMENTS
+ DROP TABLE IF EXISTS bug23037;
+ DROP FUNCTION IF EXISTS get_value;
+ SELECT COLUMN_NAME, MD5(COLUMN_DEFAULT), LENGTH(COLUMN_DEFAULT) FROM INFORMATION_SCHEMA.COLUMNS WHERE TABLE_NAME='bug23037';
+@@ -1382,6 +1418,16 @@
  FILES	information_schema.FILES	1
  GLOBAL_STATUS	information_schema.GLOBAL_STATUS	1
  GLOBAL_VARIABLES	information_schema.GLOBAL_VARIABLES	1
++INNODB_BUFFER_POOL_PAGES	information_schema.INNODB_BUFFER_POOL_PAGES	1
++INNODB_BUFFER_POOL_PAGES_BLOB	information_schema.INNODB_BUFFER_POOL_PAGES_BLOB	1
++INNODB_BUFFER_POOL_PAGES_INDEX	information_schema.INNODB_BUFFER_POOL_PAGES_INDEX	1
 +INNODB_CMP	information_schema.INNODB_CMP	1
 +INNODB_CMPMEM	information_schema.INNODB_CMPMEM	1
 +INNODB_CMPMEM_RESET	information_schema.INNODB_CMPMEM_RESET	1
@@ -75,51 +106,71 @@ diff mysql-test/r/information_schema.result.orig mysql-test/r/information_schema
  KEY_COLUMN_USAGE	information_schema.KEY_COLUMN_USAGE	1
  PARTITIONS	information_schema.PARTITIONS	1
  PLUGINS	information_schema.PLUGINS	1
+@@ -1400,6 +1446,7 @@
+ TRIGGERS	information_schema.TRIGGERS	1
+ USER_PRIVILEGES	information_schema.USER_PRIVILEGES	1
+ VIEWS	information_schema.VIEWS	1
++XTRADB_ENHANCEMENTS	information_schema.XTRADB_ENHANCEMENTS	1
+ create table t1(f1 int);
+ create view v1 as select f1+1 as a from t1;
+ create table t2 (f1 int, f2 int);
 diff mysql-test/r/information_schema_db.result.orig mysql-test/r/information_schema_db.result
 --- mysql-test/r/information_schema_db.result.orig	2008-08-04 09:27:49.000000000 +0300
 +++ mysql-test/r/information_schema_db.result	2008-10-07 12:26:31.000000000 +0300
-@@ -33,6 +33,13 @@
+@@ -33,6 +33,17 @@
  TRIGGERS
  USER_PRIVILEGES
  VIEWS
-+INNODB_CMP_RESET
++INNODB_CMP
++INNODB_BUFFER_POOL_PAGES
++INNODB_CMPMEM
++INNODB_BUFFER_POOL_PAGES_INDEX
++XTRADB_ENHANCEMENTS
 +INNODB_TRX
-+INNODB_CMPMEM_RESET
++INNODB_BUFFER_POOL_PAGES_BLOB
 +INNODB_LOCK_WAITS
-+INNODB_CMPMEM
-+INNODB_CMP
++INNODB_CMP_RESET
 +INNODB_LOCKS
++INNODB_CMPMEM_RESET
  show tables from INFORMATION_SCHEMA like 'T%';
  Tables_in_information_schema (T%)
  TABLES
 diff mysql-test/r/mysqlshow.result.orig mysql-test/r/mysqlshow.result
 --- mysql-test/r/mysqlshow.result.orig	2008-08-04 09:27:51.000000000 +0300
 +++ mysql-test/r/mysqlshow.result	2008-10-07 12:35:39.000000000 +0300
-@@ -107,6 +107,13 @@
+@@ -107,6 +107,17 @@
  | TRIGGERS                              |
  | USER_PRIVILEGES                       |
  | VIEWS                                 |
-+| INNODB_CMP_RESET                      |
++| INNODB_CMP                            |
++| INNODB_BUFFER_POOL_PAGES              |
++| INNODB_CMPMEM                         |
++| INNODB_BUFFER_POOL_PAGES_INDEX        |
++| XTRADB_ENHANCEMENTS                   |
 +| INNODB_TRX                            |
-+| INNODB_CMPMEM_RESET                   |
++| INNODB_BUFFER_POOL_PAGES_BLOB         |
 +| INNODB_LOCK_WAITS                     |
-+| INNODB_CMPMEM                         |
-+| INNODB_CMP                            |
++| INNODB_CMP_RESET                      |
 +| INNODB_LOCKS                          |
++| INNODB_CMPMEM_RESET                   |
  +---------------------------------------+
  Database: INFORMATION_SCHEMA
  +---------------------------------------+
-@@ -140,6 +147,13 @@
+@@ -140,6 +151,17 @@
  | TRIGGERS                              |
  | USER_PRIVILEGES                       |
  | VIEWS                                 |
-+| INNODB_CMP_RESET                      |
++| INNODB_CMP                            |
++| INNODB_BUFFER_POOL_PAGES              |
++| INNODB_CMPMEM                         |
++| INNODB_BUFFER_POOL_PAGES_INDEX        |
++| XTRADB_ENHANCEMENTS                   |
 +| INNODB_TRX                            |
-+| INNODB_CMPMEM_RESET                   |
++| INNODB_BUFFER_POOL_PAGES_BLOB         |
 +| INNODB_LOCK_WAITS                     |
-+| INNODB_CMPMEM                         |
-+| INNODB_CMP                            |
++| INNODB_CMP_RESET                      |
 +| INNODB_LOCKS                          |
++| INNODB_CMPMEM_RESET                   |
  +---------------------------------------+
  Wildcard: inf_rmation_schema
  +--------------------+

srv/srv0srv.c

@@ -107,6 +107,8 @@ UNIV_INTERN char**	srv_data_file_names = NULL;
 /* size in database pages */
 UNIV_INTERN ulint*	srv_data_file_sizes = NULL;
 
+UNIV_INTERN ibool	srv_extra_undoslots = FALSE;
+
 /* if TRUE, then we auto-extend the last data file */
 UNIV_INTERN ibool	srv_auto_extend_last_data_file	= FALSE;
 /* if != 0, this tells the max size auto-extending may increase the

trx/trx0undo.c

@@ -1366,9 +1366,47 @@ trx_undo_lists_init(
 	rseg_header = trx_rsegf_get_new(rseg->space, rseg->zip_size,
 					rseg->page_no, &mtr);
 
+	if (!srv_extra_undoslots) {
+		/* uses direct call for avoid "Assertion failure" */
+		//page_no = trx_rsegf_get_nth_undo(rseg_header, TRX_RSEG_N_EXTRA_SLOTS - 1, &mtr);
+		page_no = mtr_read_ulint(rseg_header + TRX_RSEG_UNDO_SLOTS
+					 + (TRX_RSEG_N_EXTRA_SLOTS - 1) * TRX_RSEG_SLOT_SIZE,
+					 MLOG_4BYTES, &mtr);
+		if (page_no != 0) {
+			/* check extended slots are not used */
+			for (i = TRX_RSEG_N_SLOTS; i < TRX_RSEG_N_EXTRA_SLOTS; i++) {
+				/* uses direct call for avoid "Assertion failure" */
+				page_no = mtr_read_ulint(rseg_header + TRX_RSEG_UNDO_SLOTS
+							 + i * TRX_RSEG_SLOT_SIZE,
+							 MLOG_4BYTES, &mtr);
+				if (page_no != FIL_NULL) {
+					srv_extra_undoslots = TRUE;
+					fprintf(stderr,
+"InnoDB: Error: innodb_extra_undoslots option is disabled, but it was enabled before.\n"
+"InnoDB: The datafile is not normal for mysqld and disabled innodb_extra_undoslots.\n"
+"InnoDB: Enable innodb_extra_undoslots if it was enabled before, and\n"
+"InnoDB: ### don't use this datafile with other mysqld or ibbackup! ###\n"
+"InnoDB: Cannot continue operation for the safety. Calling exit(1).\n");
+					exit(1);
+				}
+			}
+			fprintf(stderr,
+"InnoDB: Warning: innodb_extra_undoslots option is disabled, but it was  enabled before.\n"
+"InnoDB: But extended undo slots seem not used, so continue operation.\n");
+		}
+	}
+
 	for (i = 0; i < TRX_RSEG_N_SLOTS; i++) {
 		page_no = trx_rsegf_get_nth_undo(rseg_header, i, &mtr);
 
+		/* If it was not initialized when the datafile created,
+		page_no will be 0 for the extended slots after that */
+
+		if (page_no == 0) {
+			page_no = FIL_NULL;
+			trx_rsegf_set_nth_undo(rseg_header, i, page_no, &mtr);
+		}
+
 		/* In forced recovery: try to avoid operations which look
 		at database pages; undo logs are rapidly changing data, and
 		the probability that they are in an inconsistent state is