sync with rev73 extensions

Makefile.in

@@ -713,7 +713,8 @@ check_have_atomic_pthread_t:
 		echo '#define HAVE_ATOMIC_PTHREAD_T' > include/ut0auxconf.h ; \
 	fi
 
-all: check_have_atomic_pthread_t all-am
+# This is temprary fix for http://bugs.mysql.com/43740
+all: all-am
 
 .SUFFIXES:
 .SUFFIXES: .c .cc .lo .o .obj

btr/btr0cur.c

@@ -3202,7 +3202,9 @@ btr_estimate_number_of_different_key_vals(
 	ulint		n_cols;
 	ulint		matched_fields;
 	ulint		matched_bytes;
+	ib_int64_t	n_recs	= 0;
 	ib_int64_t*	n_diff;
+	ib_int64_t*	n_not_nulls;
 	ullint		n_sample_pages; /* number of pages to sample */
 	ulint		not_empty_flag	= 0;
 	ulint		total_external_size = 0;
@@ -3215,6 +3217,7 @@ btr_estimate_number_of_different_key_vals(
 	ulint		offsets_next_rec_[REC_OFFS_NORMAL_SIZE];
 	ulint*		offsets_rec	= offsets_rec_;
 	ulint*		offsets_next_rec= offsets_next_rec_;
+	ulint		stats_method	= srv_stats_method;
 	rec_offs_init(offsets_rec_);
 	rec_offs_init(offsets_next_rec_);
 
@@ -3222,6 +3225,10 @@ btr_estimate_number_of_different_key_vals(
 
 	n_diff = mem_zalloc((n_cols + 1) * sizeof(ib_int64_t));
 
+	if (stats_method == SRV_STATS_METHOD_IGNORE_NULLS) {
+		n_not_nulls = mem_zalloc((n_cols + 1) * sizeof(ib_int64_t));
+	}
+
 	/* It makes no sense to test more pages than are contained
 	in the index, thus we lower the number if it is too high */
 	if (srv_stats_sample_pages > index->stat_index_size) {
@@ -3260,6 +3267,20 @@ btr_estimate_number_of_different_key_vals(
 		}
 
 		while (rec != supremum) {
+			/* count recs */
+			if (stats_method == SRV_STATS_METHOD_IGNORE_NULLS) {
+				n_recs++;
+				for (j = 0; j <= n_cols; j++) {
+					ulint	f_len;
+					rec_get_nth_field(rec, offsets_rec,
+							  j, &f_len);
+					if (f_len == UNIV_SQL_NULL)
+						break;
+
+					n_not_nulls[j]++;
+				}
+			}
+
 			rec_t*	next_rec = page_rec_get_next(rec);
 			if (next_rec == supremum) {
 				break;
@@ -3274,7 +3295,7 @@ btr_estimate_number_of_different_key_vals(
 			cmp_rec_rec_with_match(rec, next_rec,
 					       offsets_rec, offsets_next_rec,
 					       index, &matched_fields,
-					       &matched_bytes);
+					       &matched_bytes, srv_stats_method);
 
 			for (j = matched_fields + 1; j <= n_cols; j++) {
 				/* We add one if this index record has
@@ -3359,9 +3380,21 @@ btr_estimate_number_of_different_key_vals(
 		}
 
 		index->stat_n_diff_key_vals[j] += add_on;
+
+		/* revision for 'nulls_ignored' */
+		if (stats_method == SRV_STATS_METHOD_IGNORE_NULLS) {
+			if (!n_not_nulls[j])
+				n_not_nulls[j] = 1;
+			index->stat_n_diff_key_vals[j] =
+				index->stat_n_diff_key_vals[j] * n_recs
+					/ n_not_nulls[j];
+		}
 	}
 
 	mem_free(n_diff);
+	if (stats_method == SRV_STATS_METHOD_IGNORE_NULLS) {
+		mem_free(n_not_nulls);
+	}
 	if (UNIV_LIKELY_NULL(heap)) {
 		mem_heap_free(heap);
 	}
@@ -3733,7 +3766,8 @@ btr_blob_free(
 
 	mtr_commit(mtr);
 
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
+	mutex_enter(&LRU_list_mutex);
 	mutex_enter(&block->mutex);
 
 	/* Only free the block if it is still allocated to
@@ -3744,17 +3778,22 @@ btr_blob_free(
 	    && buf_block_get_space(block) == space
 	    && buf_block_get_page_no(block) == page_no) {
 
-		if (buf_LRU_free_block(&block->page, all, NULL)
+		if (buf_LRU_free_block(&block->page, all, NULL, TRUE)
 		    != BUF_LRU_FREED
-		    && all && block->page.zip.data) {
+		    && all && block->page.zip.data
+		    /* Now, buf_LRU_free_block() may release mutex temporarily */
+		    && buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE
+		    && buf_block_get_space(block) == space
+		    && buf_block_get_page_no(block) == page_no) {
 			/* Attempt to deallocate the uncompressed page
 			if the whole block cannot be deallocted. */
 
-			buf_LRU_free_block(&block->page, FALSE, NULL);
+			buf_LRU_free_block(&block->page, FALSE, NULL, TRUE);
 		}
 	}
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	mutex_exit(&LRU_list_mutex);
 	mutex_exit(&block->mutex);
 }
 

btr/btr0sea.c

@@ -1731,7 +1731,8 @@ btr_search_validate(void)
 	rec_offs_init(offsets_);
 
 	rw_lock_x_lock(&btr_search_latch);
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
+	rw_lock_x_lock(&page_hash_latch);
 
 	cell_count = hash_get_n_cells(btr_search_sys->hash_index);
 
@@ -1739,11 +1740,13 @@ btr_search_validate(void)
 		/* We release btr_search_latch every once in a while to
 		give other queries a chance to run. */
 		if ((i != 0) && ((i % chunk_size) == 0)) {
-			buf_pool_mutex_exit();
+			//buf_pool_mutex_exit();
+			rw_lock_x_unlock(&page_hash_latch);
 			rw_lock_x_unlock(&btr_search_latch);
 			os_thread_yield();
 			rw_lock_x_lock(&btr_search_latch);
-			buf_pool_mutex_enter();
+			//buf_pool_mutex_enter();
+			rw_lock_x_lock(&page_hash_latch);
 		}
 
 		node = hash_get_nth_cell(btr_search_sys->hash_index, i)->node;
@@ -1850,11 +1853,13 @@ btr_search_validate(void)
 		/* We release btr_search_latch every once in a while to
 		give other queries a chance to run. */
 		if (i != 0) {
-			buf_pool_mutex_exit();
+			//buf_pool_mutex_exit();
+			rw_lock_x_unlock(&page_hash_latch);
 			rw_lock_x_unlock(&btr_search_latch);
 			os_thread_yield();
 			rw_lock_x_lock(&btr_search_latch);
-			buf_pool_mutex_enter();
+			//buf_pool_mutex_enter();
+			rw_lock_x_lock(&page_hash_latch);
 		}
 
 		if (!ha_validate(btr_search_sys->hash_index, i, end_index)) {
@@ -1862,7 +1867,8 @@ btr_search_validate(void)
 		}
 	}
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	rw_lock_x_unlock(&page_hash_latch);
 	rw_lock_x_unlock(&btr_search_latch);
 	if (UNIV_LIKELY_NULL(heap)) {
 		mem_heap_free(heap);

buf/buf0rea.c

@@ -246,18 +246,22 @@ buf_read_ahead_random(
 
 	LRU_recent_limit = buf_LRU_get_recent_limit();
 
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
+	mutex_enter(&buf_pool_mutex);
 
 	if (buf_pool->n_pend_reads
 	    > buf_pool->curr_size / BUF_READ_AHEAD_PEND_LIMIT) {
-		buf_pool_mutex_exit();
+		//buf_pool_mutex_exit();
+		mutex_exit(&buf_pool_mutex);
 
 		return(0);
 	}
+	mutex_exit(&buf_pool_mutex);
 
 	/* Count how many blocks in the area have been recently accessed,
 	that is, reside near the start of the LRU list. */
 
+	rw_lock_s_lock(&page_hash_latch);
 	for (i = low; i < high; i++) {
 		const buf_page_t*	bpage = buf_page_hash_get(space, i);
 
@@ -269,13 +273,15 @@ buf_read_ahead_random(
 
 			if (recent_blocks >= BUF_READ_AHEAD_RANDOM_THRESHOLD) {
 
-				buf_pool_mutex_exit();
+				//buf_pool_mutex_exit();
+				rw_lock_s_unlock(&page_hash_latch);
 				goto read_ahead;
 			}
 		}
 	}
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	rw_lock_s_unlock(&page_hash_latch);
 	/* Do nothing */
 	return(0);
 
@@ -469,10 +475,12 @@ buf_read_ahead_linear(
 
 	tablespace_version = fil_space_get_version(space);
 
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
+	mutex_enter(&buf_pool_mutex);
 
 	if (high > fil_space_get_size(space)) {
-		buf_pool_mutex_exit();
+		//buf_pool_mutex_exit();
+		mutex_exit(&buf_pool_mutex);
 		/* The area is not whole, return */
 
 		return(0);
@@ -480,10 +488,12 @@ buf_read_ahead_linear(
 
 	if (buf_pool->n_pend_reads
 	    > buf_pool->curr_size / BUF_READ_AHEAD_PEND_LIMIT) {
-		buf_pool_mutex_exit();
+		//buf_pool_mutex_exit();
+		mutex_exit(&buf_pool_mutex);
 
 		return(0);
 	}
+	mutex_exit(&buf_pool_mutex);
 
 	/* Check that almost all pages in the area have been accessed; if
 	offset == low, the accesses must be in a descending order, otherwise,
@@ -497,6 +507,7 @@ buf_read_ahead_linear(
 
 	fail_count = 0;
 
+	rw_lock_s_lock(&page_hash_latch);
 	for (i = low; i < high; i++) {
 		bpage = buf_page_hash_get(space, i);
 
@@ -520,7 +531,8 @@ buf_read_ahead_linear(
 	    * LINEAR_AREA_THRESHOLD_COEF) {
 		/* Too many failures: return */
 
-		buf_pool_mutex_exit();
+		//buf_pool_mutex_exit();
+		rw_lock_s_unlock(&page_hash_latch);
 
 		return(0);
 	}
@@ -531,7 +543,8 @@ buf_read_ahead_linear(
 	bpage = buf_page_hash_get(space, offset);
 
 	if (bpage == NULL) {
-		buf_pool_mutex_exit();
+		//buf_pool_mutex_exit();
+		rw_lock_s_unlock(&page_hash_latch);
 
 		return(0);
 	}
@@ -557,7 +570,8 @@ buf_read_ahead_linear(
 	pred_offset = fil_page_get_prev(frame);
 	succ_offset = fil_page_get_next(frame);
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	rw_lock_s_unlock(&page_hash_latch);
 
 	if ((offset == low) && (succ_offset == offset + 1)) {
 

dict/dict0boot.c

@@ -265,6 +265,7 @@ dict_boot(void)
 	system tables */
 	/*-------------------------*/
 	table = dict_mem_table_create("SYS_TABLES", DICT_HDR_SPACE, 8, 0);
+	table->n_mysql_handles_opened = 1; /* for pin */
 
 	dict_mem_table_add_col(table, heap, "NAME", DATA_BINARY, 0, 0);
 	dict_mem_table_add_col(table, heap, "ID", DATA_BINARY, 0, 0);
@@ -314,6 +315,7 @@ dict_boot(void)
 
 	/*-------------------------*/
 	table = dict_mem_table_create("SYS_COLUMNS", DICT_HDR_SPACE, 7, 0);
+	table->n_mysql_handles_opened = 1; /* for pin */
 
 	dict_mem_table_add_col(table, heap, "TABLE_ID", DATA_BINARY, 0, 0);
 	dict_mem_table_add_col(table, heap, "POS", DATA_INT, 0, 4);
@@ -346,6 +348,7 @@ dict_boot(void)
 
 	/*-------------------------*/
 	table = dict_mem_table_create("SYS_INDEXES", DICT_HDR_SPACE, 7, 0);
+	table->n_mysql_handles_opened = 1; /* for pin */
 
 	dict_mem_table_add_col(table, heap, "TABLE_ID", DATA_BINARY, 0, 0);
 	dict_mem_table_add_col(table, heap, "ID", DATA_BINARY, 0, 0);
@@ -388,6 +391,7 @@ dict_boot(void)
 
 	/*-------------------------*/
 	table = dict_mem_table_create("SYS_FIELDS", DICT_HDR_SPACE, 3, 0);
+	table->n_mysql_handles_opened = 1; /* for pin */
 
 	dict_mem_table_add_col(table, heap, "INDEX_ID", DATA_BINARY, 0, 0);
 	dict_mem_table_add_col(table, heap, "POS", DATA_INT, 0, 4);

dict/dict0crea.c

@@ -1184,6 +1184,9 @@ dict_create_or_check_foreign_constraint_tables(void)
 		/* Foreign constraint system tables have already been
 		created, and they are ok */
 
+		table1->n_mysql_handles_opened = 1; /* for pin */
+		table2->n_mysql_handles_opened = 1; /* for pin */
+
 		mutex_exit(&(dict_sys->mutex));
 
 		return(DB_SUCCESS);
@@ -1265,6 +1268,11 @@ dict_create_or_check_foreign_constraint_tables(void)
 
 	trx_commit_for_mysql(trx);
 
+	table1 = dict_table_get_low("SYS_FOREIGN");
+	table2 = dict_table_get_low("SYS_FOREIGN_COLS");
+	table1->n_mysql_handles_opened = 1; /* for pin */
+	table2->n_mysql_handles_opened = 1; /* for pin */
+
 	row_mysql_unlock_data_dictionary(trx);
 
 	trx_free_for_mysql(trx);

dict/dict0dict.c

@@ -545,6 +545,8 @@ dict_table_get_on_id(
 
 	table = dict_table_get_on_id_low(table_id);
 
+	dict_table_LRU_trim(table);
+
 	mutex_exit(&(dict_sys->mutex));
 
 	return(table);
@@ -659,6 +661,8 @@ dict_table_get(
 		table->n_mysql_handles_opened++;
 	}
 
+	dict_table_LRU_trim(table);
+
 	mutex_exit(&(dict_sys->mutex));
 
 	if (table != NULL) {
@@ -1153,6 +1157,64 @@ dict_table_remove_from_cache(
 	dict_mem_table_free(table);
 }
 
+/**************************************************************************
+Frees tables from the end of table_LRU if the dictionary cache occupies
+too much space. */
+UNIV_INTERN
+void
+dict_table_LRU_trim(
+/*================*/
+	dict_table_t*	self)
+{
+	dict_table_t*	table;
+	dict_table_t*	prev_table;
+	dict_foreign_t*	foreign;
+	ulint		n_removed;
+	ulint		n_have_parent;
+	ulint		cached_foreign_tables;
+
+#ifdef UNIV_SYNC_DEBUG
+	ut_ad(mutex_own(&(dict_sys->mutex)));
+#endif /* UNIV_SYNC_DEBUG */
+
+retry:
+	n_removed = n_have_parent = 0;
+	table = UT_LIST_GET_LAST(dict_sys->table_LRU);
+
+	while ( srv_dict_size_limit && table
+		&& ((dict_sys->table_hash->n_cells
+		     + dict_sys->table_id_hash->n_cells) * sizeof(hash_cell_t)
+		    + dict_sys->size) > srv_dict_size_limit ) {
+		prev_table = UT_LIST_GET_PREV(table_LRU, table);
+
+		if (table == self || table->n_mysql_handles_opened)
+			goto next_loop;
+
+		cached_foreign_tables = 0;
+		foreign = UT_LIST_GET_FIRST(table->foreign_list);
+		while (foreign != NULL) {
+			if (foreign->referenced_table)
+				cached_foreign_tables++;
+			foreign = UT_LIST_GET_NEXT(foreign_list, foreign);
+		}
+
+		if (cached_foreign_tables == 0) {
+			dict_table_remove_from_cache(table);
+			n_removed++;
+		} else {
+			n_have_parent++;
+		}
+next_loop:
+		table = prev_table;
+	}
+
+	if ( srv_dict_size_limit && n_have_parent && n_removed
+		&& ((dict_sys->table_hash->n_cells
+		     + dict_sys->table_id_hash->n_cells) * sizeof(hash_cell_t)
+		    + dict_sys->size) > srv_dict_size_limit )
+		goto retry;
+}
+
 /********************************************************************
 If the given column name is reserved for InnoDB system columns, return
 TRUE. */
@@ -4276,7 +4338,8 @@ dict_table_print_low(
 
 	ut_ad(mutex_own(&(dict_sys->mutex)));
 
-	dict_update_statistics_low(table, TRUE);
+	if (srv_stats_auto_update)
+		dict_update_statistics_low(table, TRUE);
 
 	fprintf(stderr,
 		"--------------------------------------\n"

dict/dict0load.c

@@ -223,7 +223,7 @@ dict_print(void)
 			/* The table definition was corrupt if there
 			is no index */
 
-			if (dict_table_get_first_index(table)) {
+			if (srv_stats_auto_update && dict_table_get_first_index(table)) {
 				dict_update_statistics_low(table, TRUE);
 			}
 

fil/fil0fil.c

@@ -42,6 +42,10 @@ Created 10/25/1995 Heikki Tuuri
 #include "mtr0log.h"
 #include "dict0dict.h"
 #include "page0zip.h"
+#include "trx0trx.h"
+#include "trx0sys.h"
+#include "pars0pars.h"
+#include "row0mysql.h"
 
 
 /*
@@ -2977,7 +2981,7 @@ fil_open_single_table_tablespace(
 	ut_a(flags != DICT_TF_COMPACT);
 
 	file = os_file_create_simple_no_error_handling(
-		filepath, OS_FILE_OPEN, OS_FILE_READ_ONLY, &success);
+		filepath, OS_FILE_OPEN, OS_FILE_READ_WRITE, &success);
 	if (!success) {
 		/* The following call prints an error message */
 		os_file_get_last_error(TRUE);
@@ -3025,6 +3029,275 @@ fil_open_single_table_tablespace(
 	space_id = fsp_header_get_space_id(page);
 	space_flags = fsp_header_get_flags(page);
 
+	if (srv_expand_import && (space_id != id || space_flags != flags)) {
+		dulint		old_id[31];
+		dulint		new_id[31];
+		ulint		root_page[31];
+		ulint		n_index;
+		os_file_t	info_file = -1;
+		char*		info_file_path;
+		ulint	i;
+		int		len;
+		ib_uint64_t	current_lsn;
+
+		current_lsn = log_get_lsn();
+
+		/* overwrite fsp header */
+		fsp_header_init_fields(page, id, flags);
+		mach_write_to_4(page + FIL_PAGE_ARCH_LOG_NO_OR_SPACE_ID, id);
+		space_id = id;
+		space_flags = flags;
+		if (mach_read_ull(page + FIL_PAGE_FILE_FLUSH_LSN) > current_lsn)
+			mach_write_ull(page + FIL_PAGE_FILE_FLUSH_LSN, current_lsn);
+		mach_write_to_4(page + FIL_PAGE_SPACE_OR_CHKSUM,
+				srv_use_checksums
+				? buf_calc_page_new_checksum(page)
+						: BUF_NO_CHECKSUM_MAGIC);
+		mach_write_to_4(page + UNIV_PAGE_SIZE - FIL_PAGE_END_LSN_OLD_CHKSUM,
+				srv_use_checksums
+				? buf_calc_page_old_checksum(page)
+						: BUF_NO_CHECKSUM_MAGIC);
+		success = os_file_write(filepath, file, page, 0, 0, UNIV_PAGE_SIZE);
+
+		/* get file size */
+		ulint		size_low, size_high, size;
+		ib_int64_t	size_bytes;
+		os_file_get_size(file, &size_low, &size_high);
+		size_bytes = (((ib_int64_t)size_high) << 32)
+				+ (ib_int64_t)size_low;
+
+		/* get cruster index information */
+		dict_table_t*   table;
+		dict_index_t*   index;
+		table = dict_table_get_low(name);
+		index = dict_table_get_first_index(table);
+		ut_a(index->page==3);
+
+
+		/* read metadata from .exp file */
+		n_index = 0;
+		bzero(old_id, sizeof(old_id));
+		bzero(new_id, sizeof(new_id));
+		bzero(root_page, sizeof(root_page));
+
+		info_file_path = fil_make_ibd_name(name, FALSE);
+		len = strlen(info_file_path);
+		info_file_path[len - 3] = 'e';
+		info_file_path[len - 2] = 'x';
+		info_file_path[len - 1] = 'p';
+
+		info_file = os_file_create_simple_no_error_handling(
+				info_file_path, OS_FILE_OPEN, OS_FILE_READ_ONLY, &success);
+		if (!success) {
+			fprintf(stderr, "InnoDB: cannot open %s\n", info_file_path);
+			goto skip_info;
+		}
+		success = os_file_read(info_file, page, 0, 0, UNIV_PAGE_SIZE);
+		if (!success) {
+			fprintf(stderr, "InnoDB: cannot read %s\n", info_file_path);
+			goto skip_info;
+		}
+		if (mach_read_from_4(page) != 0x78706f72UL
+		    || mach_read_from_4(page + 4) != 0x74696e66UL) {
+			fprintf(stderr, "InnoDB: %s seems not to be a correct .exp file\n", info_file_path);
+			goto skip_info;
+		}
+
+		fprintf(stderr, "InnoDB: import: extended import of %s is started.\n", name);
+
+		n_index = mach_read_from_4(page + 8);
+		fprintf(stderr, "InnoDB: import: %lu indexes are detected.\n", (ulong)n_index);
+		for (i = 0; i < n_index; i++) {
+			new_id[i] =
+				dict_table_get_index_on_name(table,
+						(page + (i + 1) * 512 + 12))->id;
+			old_id[i] = mach_read_from_8(page + (i + 1) * 512);
+			root_page[i] = mach_read_from_4(page + (i + 1) * 512 + 8);
+		}
+
+skip_info:
+		if (info_file != -1)
+			os_file_close(info_file);
+
+		/*
+		if (size_bytes >= 1024 * 1024) {
+			size_bytes = ut_2pow_round(size_bytes, 1024 * 1024);
+		}
+		*/
+		if (!(flags & DICT_TF_ZSSIZE_MASK)) {
+			mem_heap_t*	heap = NULL;
+			ulint		offsets_[REC_OFFS_NORMAL_SIZE];
+			ulint*		offsets = offsets_;
+			size = (ulint) (size_bytes / UNIV_PAGE_SIZE);
+			/* over write space id of all pages */
+			ib_int64_t     offset;
+
+			rec_offs_init(offsets_);
+
+			fprintf(stderr, "InnoDB: Progress in %:");
+
+			for (offset = 0; offset < size_bytes; offset += UNIV_PAGE_SIZE) {
+				success = os_file_read(file, page,
+							(ulint)(offset & 0xFFFFFFFFUL),
+							(ulint)(offset >> 32), UNIV_PAGE_SIZE);
+				if (mach_read_from_4(page + FIL_PAGE_OFFSET) || !offset) {
+					mach_write_to_4(page + FIL_PAGE_ARCH_LOG_NO_OR_SPACE_ID, id);
+
+					for (i = 0; i < n_index; i++) {
+						if (offset / UNIV_PAGE_SIZE == root_page[i]) {
+							/* this is index root page */
+							mach_write_to_4(page + FIL_PAGE_DATA + PAGE_BTR_SEG_LEAF
+											+ FSEG_HDR_SPACE, id);
+							mach_write_to_4(page + FIL_PAGE_DATA + PAGE_BTR_SEG_TOP
+											+ FSEG_HDR_SPACE, id);
+							break;
+						}
+					}
+
+					if (fil_page_get_type(page) == FIL_PAGE_INDEX) {
+						dulint tmp = mach_read_from_8(page + (PAGE_HEADER + PAGE_INDEX_ID));
+
+						if (mach_read_from_2(page + PAGE_HEADER + PAGE_LEVEL) == 0
+						    && ut_dulint_cmp(old_id[0], tmp) == 0) {
+							/* leaf page of cluster index, reset trx_id of records */
+							rec_t*	rec;
+							rec_t*	supremum;
+							ulint	n_recs;
+
+							supremum = page_get_supremum_rec(page);
+							rec = page_rec_get_next(page_get_infimum_rec(page));
+							n_recs = page_get_n_recs(page);
+
+							while (rec && rec != supremum && n_recs > 0) {
+								ulint	offset = index->trx_id_offset;
+								if (!offset) {
+									offsets = rec_get_offsets(rec, index, offsets,
+											ULINT_UNDEFINED, &heap);
+									offset = row_get_trx_id_offset(rec, index, offsets);
+								}
+								trx_write_trx_id(rec + offset, ut_dulint_create(0, 1));
+								rec = page_rec_get_next(rec);
+								n_recs--;
+							}
+						}
+
+						for (i = 0; i < n_index; i++) {
+							if (ut_dulint_cmp(old_id[i], tmp) == 0) {
+								mach_write_to_8(page + (PAGE_HEADER + PAGE_INDEX_ID), new_id[i]);
+								break;
+							}
+						}
+					}
+
+					if (mach_read_ull(page + FIL_PAGE_LSN) > current_lsn) {
+						mach_write_ull(page + FIL_PAGE_LSN, current_lsn);
+						mach_write_ull(page + UNIV_PAGE_SIZE - FIL_PAGE_END_LSN_OLD_CHKSUM,
+										current_lsn);
+					}
+
+					mach_write_to_4(page + FIL_PAGE_SPACE_OR_CHKSUM,
+							srv_use_checksums
+							? buf_calc_page_new_checksum(page)
+									: BUF_NO_CHECKSUM_MAGIC);
+					mach_write_to_4(page + UNIV_PAGE_SIZE - FIL_PAGE_END_LSN_OLD_CHKSUM,
+							srv_use_checksums
+							? buf_calc_page_old_checksum(page)
+									: BUF_NO_CHECKSUM_MAGIC);
+
+					success = os_file_write(filepath, file, page,
+								(ulint)(offset & 0xFFFFFFFFUL),
+								(ulint)(offset >> 32), UNIV_PAGE_SIZE);
+				}
+
+				if (size_bytes
+				    && ((ib_int64_t)((offset + UNIV_PAGE_SIZE) * 100) / size_bytes)
+					!= ((offset * 100) / size_bytes)) {
+					fprintf(stderr, " %lu",
+						(ulong)((ib_int64_t)((offset + UNIV_PAGE_SIZE) * 100) / size_bytes));
+				}
+			}
+
+			fprintf(stderr, " done.\n");
+
+			/* update SYS_INDEXES set root page */
+			index = dict_table_get_first_index(table);
+			while (index) {
+				for (i = 0; i < n_index; i++) {
+					if (ut_dulint_cmp(new_id[i], index->id) == 0) {
+						break;
+					}
+				}
+
+				if (i != n_index
+				    && root_page[i] != index->page) {
+					/* must update */
+					ulint	error;
+					trx_t*	trx;
+					pars_info_t*	info = NULL;
+
+					trx = trx_allocate_for_mysql();
+					trx->op_info = "extended import";
+
+					info = pars_info_create();
+
+					pars_info_add_dulint_literal(info, "indexid", new_id[i]);
+					pars_info_add_int4_literal(info, "new_page", (lint) root_page[i]);
+
+					error = que_eval_sql(info,
+						"PROCEDURE UPDATE_INDEX_PAGE () IS\n"
+						"BEGIN\n"
+						"UPDATE SYS_INDEXES"
+						" SET PAGE_NO = :new_page"
+						" WHERE ID = :indexid;\n"
+						"COMMIT WORK;\n"
+						"END;\n",
+						FALSE, trx);
+
+					if (error != DB_SUCCESS) {
+						fprintf(stderr, "InnoDB: failed to update SYS_INDEXES\n");
+					}
+
+					trx_commit_for_mysql(trx);
+
+					trx_free_for_mysql(trx);
+
+					index->page = root_page[i];
+				}
+
+				index = dict_table_get_next_index(index);
+			}
+			if (UNIV_LIKELY_NULL(heap)) {
+				mem_heap_free(heap);
+			}
+		} else {
+			/* zip page? */
+			size = (ulint)
+			(size_bytes
+					/ dict_table_flags_to_zip_size(flags));
+			fprintf(stderr, "InnoDB: import: table %s seems to be in newer format."
+					" It may not be able to treated for now.\n", name);
+		}
+		/* .exp file should be removed */
+		success = os_file_delete(info_file_path);
+		if (!success) {
+			success = os_file_delete_if_exists(info_file_path);
+		}
+		mem_free(info_file_path);
+
+		fil_system_t*	system	= fil_system;
+		mutex_enter(&(system->mutex));
+		fil_node_t*	node = NULL;
+		fil_space_t*	space;
+		space = fil_space_get_by_id(id);
+		if (space)
+			node = UT_LIST_GET_FIRST(space->chain);
+		if (node && node->size < size) {
+			space->size += (size - node->size);
+			node->size = size;
+		}
+		mutex_exit(&(system->mutex));
+	}
+
 	ut_free(buf2);
 
 	if (UNIV_UNLIKELY(space_id != id || space_flags != flags)) {

handler/i_s.cc

@@ -45,6 +45,7 @@ extern "C" {
 #include "dict0dict.h" /* for dict_index_get_if_in_cache */
 #include "trx0rseg.h" /* for trx_rseg_struct */
 #include "trx0sys.h" /* for trx_sys */
+#include "dict0dict.h" /* for dict_sys */
 /* from buf0buf.c */
 struct buf_chunk_struct{
 	ulint		mem_size;	/* allocated size of the chunk */
@@ -2282,7 +2283,8 @@ i_s_cmpmem_fill_low(
 
 	RETURN_IF_INNODB_NOT_STARTED(tables->schema_table_name);
 
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
+	mutex_enter(&zip_free_mutex);
 
 	for (uint x = 0; x <= BUF_BUDDY_SIZES; x++) {
 		buf_buddy_stat_t*	buddy_stat = &buf_buddy_stat[x];
@@ -2308,7 +2310,8 @@ i_s_cmpmem_fill_low(
 		}
 	}
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	mutex_exit(&zip_free_mutex);
 	DBUG_RETURN(status);
 }
 
@@ -2653,3 +2656,299 @@ UNIV_INTERN struct st_mysql_plugin	i_s_innodb_rseg =
 	/* void* */
 	STRUCT_FLD(__reserved1, NULL)
 };
+
+/***********************************************************************
+*/
+static ST_FIELD_INFO	i_s_innodb_table_stats_info[] =
+{
+	{STRUCT_FLD(field_name,		"table_name"),
+	 STRUCT_FLD(field_length,	NAME_LEN),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_STRING),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	0),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"rows"),
+	 STRUCT_FLD(field_length,	MY_INT64_NUM_DECIMAL_DIGITS),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_LONGLONG),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	MY_I_S_UNSIGNED),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"clust_size"),
+	 STRUCT_FLD(field_length,	MY_INT64_NUM_DECIMAL_DIGITS),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_LONGLONG),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	MY_I_S_UNSIGNED),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"other_size"),
+	 STRUCT_FLD(field_length,	MY_INT64_NUM_DECIMAL_DIGITS),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_LONGLONG),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	MY_I_S_UNSIGNED),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"modified"),
+	 STRUCT_FLD(field_length,	MY_INT64_NUM_DECIMAL_DIGITS),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_LONGLONG),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	MY_I_S_UNSIGNED),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	END_OF_ST_FIELD_INFO
+};
+
+static ST_FIELD_INFO	i_s_innodb_index_stats_info[] =
+{
+	{STRUCT_FLD(field_name,		"table_name"),
+	 STRUCT_FLD(field_length,	NAME_LEN),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_STRING),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	0),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"index_name"),
+	 STRUCT_FLD(field_length,	NAME_LEN),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_STRING),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	0),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"fields"),
+	 STRUCT_FLD(field_length,	MY_INT64_NUM_DECIMAL_DIGITS),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_LONGLONG),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	MY_I_S_UNSIGNED),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"row_per_keys"),
+	 STRUCT_FLD(field_length,	256),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_STRING),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	0),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"index_size"),
+	 STRUCT_FLD(field_length,	MY_INT64_NUM_DECIMAL_DIGITS),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_LONGLONG),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	MY_I_S_UNSIGNED),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	{STRUCT_FLD(field_name,		"leaf_pages"),
+	 STRUCT_FLD(field_length,	MY_INT64_NUM_DECIMAL_DIGITS),
+	 STRUCT_FLD(field_type,		MYSQL_TYPE_LONGLONG),
+	 STRUCT_FLD(value,		0),
+	 STRUCT_FLD(field_flags,	MY_I_S_UNSIGNED),
+	 STRUCT_FLD(old_name,		""),
+	 STRUCT_FLD(open_method,	SKIP_OPEN_TABLE)},
+
+	END_OF_ST_FIELD_INFO
+};
+
+static
+int
+i_s_innodb_table_stats_fill(
+/*========================*/
+	THD*		thd,
+	TABLE_LIST*	tables,
+	COND*		cond)
+{
+	TABLE*	i_s_table	= (TABLE *) tables->table;
+	int	status	= 0;
+	dict_table_t*	table;
+
+	DBUG_ENTER("i_s_innodb_table_stats_fill");
+
+	/* deny access to non-superusers */
+	if (check_global_access(thd, PROCESS_ACL)) {
+		DBUG_RETURN(0);
+	}
+
+	mutex_enter(&(dict_sys->mutex));
+
+	table = UT_LIST_GET_FIRST(dict_sys->table_LRU);
+
+	while (table) {
+		if (table->stat_clustered_index_size == 0) {
+			table = UT_LIST_GET_NEXT(table_LRU, table);
+			continue;
+		}
+
+		field_store_string(i_s_table->field[0], table->name);
+		i_s_table->field[1]->store(table->stat_n_rows);
+		i_s_table->field[2]->store(table->stat_clustered_index_size);
+		i_s_table->field[3]->store(table->stat_sum_of_other_index_sizes);
+		i_s_table->field[4]->store(table->stat_modified_counter);
+
+		if (schema_table_store_record(thd, i_s_table)) {
+			status = 1;
+			break;
+		}
+
+		table = UT_LIST_GET_NEXT(table_LRU, table);
+	}
+
+	mutex_exit(&(dict_sys->mutex));
+
+	DBUG_RETURN(status);
+}
+
+static
+int
+i_s_innodb_index_stats_fill(
+/*========================*/
+	THD*		thd,
+	TABLE_LIST*	tables,
+	COND*		cond)
+{
+	TABLE*	i_s_table	= (TABLE *) tables->table;
+	int	status	= 0;
+	dict_table_t*	table;
+	dict_index_t*	index;
+
+	DBUG_ENTER("i_s_innodb_index_stats_fill");
+
+	/* deny access to non-superusers */
+	if (check_global_access(thd, PROCESS_ACL)) {
+		DBUG_RETURN(0);
+	}
+
+	mutex_enter(&(dict_sys->mutex));
+
+	table = UT_LIST_GET_FIRST(dict_sys->table_LRU);
+
+	while (table) {
+		if (table->stat_clustered_index_size == 0) {
+			table = UT_LIST_GET_NEXT(table_LRU, table);
+			continue;
+		}
+
+		ib_int64_t	n_rows = table->stat_n_rows;
+
+		if (n_rows < 0) {
+			n_rows = 0;
+		}
+
+		index = dict_table_get_first_index(table);
+
+		while (index) {
+			char	buff[256+1];
+			char	row_per_keys[256+1];
+			ulint	i;
+
+			field_store_string(i_s_table->field[0], table->name);
+			field_store_string(i_s_table->field[1], index->name);
+			i_s_table->field[2]->store(index->n_uniq);
+
+			row_per_keys[0] = '\0';
+			if (index->stat_n_diff_key_vals) {
+				for (i = 1; i <= index->n_uniq; i++) {
+					ib_int64_t	rec_per_key;
+					if (index->stat_n_diff_key_vals[i]) {
+						rec_per_key = n_rows / index->stat_n_diff_key_vals[i];
+					} else {
+						rec_per_key = n_rows;
+					}
+					snprintf(buff, 256, (i == index->n_uniq)?"%llu":"%llu, ",
+						 rec_per_key);
+					strncat(row_per_keys, buff, 256 - strlen(row_per_keys));
+				}
+			}
+			field_store_string(i_s_table->field[3], row_per_keys);
+
+			i_s_table->field[4]->store(index->stat_index_size);
+			i_s_table->field[5]->store(index->stat_n_leaf_pages);
+
+			if (schema_table_store_record(thd, i_s_table)) {
+				status = 1;
+				break;
+			}
+
+			index = dict_table_get_next_index(index);
+		}
+
+		if (status == 1) {
+			break;
+		}
+
+		table = UT_LIST_GET_NEXT(table_LRU, table);
+	}
+
+	mutex_exit(&(dict_sys->mutex));
+
+	DBUG_RETURN(status);
+}
+
+static
+int
+i_s_innodb_table_stats_init(
+/*========================*/
+	void*   p)
+{
+	DBUG_ENTER("i_s_innodb_table_stats_init");
+	ST_SCHEMA_TABLE* schema = (ST_SCHEMA_TABLE*) p;
+
+	schema->fields_info = i_s_innodb_table_stats_info;
+	schema->fill_table = i_s_innodb_table_stats_fill;
+
+	DBUG_RETURN(0);
+}
+
+static
+int
+i_s_innodb_index_stats_init(
+/*========================*/
+	void*	p)
+{
+	DBUG_ENTER("i_s_innodb_index_stats_init");
+	ST_SCHEMA_TABLE* schema = (ST_SCHEMA_TABLE*) p;
+
+	schema->fields_info = i_s_innodb_index_stats_info;
+	schema->fill_table = i_s_innodb_index_stats_fill;
+
+	DBUG_RETURN(0);
+}
+
+UNIV_INTERN struct st_mysql_plugin	i_s_innodb_table_stats =
+{
+	STRUCT_FLD(type, MYSQL_INFORMATION_SCHEMA_PLUGIN),
+	STRUCT_FLD(info, &i_s_info),
+	STRUCT_FLD(name, "INNODB_TABLE_STATS"),
+	STRUCT_FLD(author, plugin_author),
+	STRUCT_FLD(descr, "InnoDB table statistics in memory"),
+	STRUCT_FLD(license, PLUGIN_LICENSE_GPL),
+	STRUCT_FLD(init, i_s_innodb_table_stats_init),
+	STRUCT_FLD(deinit, i_s_common_deinit),
+	STRUCT_FLD(version, 0x0100 /* 1.0 */),
+	STRUCT_FLD(status_vars, NULL),
+	STRUCT_FLD(system_vars, NULL),
+	STRUCT_FLD(__reserved1, NULL)
+};
+
+UNIV_INTERN struct st_mysql_plugin	i_s_innodb_index_stats =
+{
+	STRUCT_FLD(type, MYSQL_INFORMATION_SCHEMA_PLUGIN),
+	STRUCT_FLD(info, &i_s_info),
+	STRUCT_FLD(name, "INNODB_INDEX_STATS"),
+	STRUCT_FLD(author, plugin_author),
+	STRUCT_FLD(descr, "InnoDB index statistics in memory"),
+	STRUCT_FLD(license, PLUGIN_LICENSE_GPL),
+	STRUCT_FLD(init, i_s_innodb_index_stats_init),
+	STRUCT_FLD(deinit, i_s_common_deinit),
+	STRUCT_FLD(version, 0x0100 /* 1.0 */),
+	STRUCT_FLD(status_vars, NULL),
+	STRUCT_FLD(system_vars, NULL),
+	STRUCT_FLD(__reserved1, NULL)
+};

handler/i_s.h

@@ -37,5 +37,7 @@ extern struct st_mysql_plugin	i_s_innodb_cmpmem;
 extern struct st_mysql_plugin	i_s_innodb_cmpmem_reset;
 extern struct st_mysql_plugin	i_s_innodb_patches;
 extern struct st_mysql_plugin	i_s_innodb_rseg;
+extern struct st_mysql_plugin	i_s_innodb_table_stats;
+extern struct st_mysql_plugin	i_s_innodb_index_stats;
 
 #endif /* i_s_h */

handler/innodb_patch_info.h

@@ -31,5 +31,11 @@ struct innodb_enhancement {
 {"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_extra_rseg","allow to create extra rollback segments","When create new db, the new parameter allows to create more rollback segments","http://www.percona.com/docs/wiki/percona-xtradb"},
 {"innodb_overwrite_relay_log_info","overwrite relay-log.info when slave recovery","Building as plugin, it is not used.","http://www.percona.com/docs/wiki/percona-xtradb:innodb_overwrite_relay_log_info"},
+{"innodb_pause_in_spin","use 'pause' instruction during spin loop for x86 (gcc)","","http://www.percona.com/docs/wiki/percona-xtradb"},
+{"innodb_thread_concurrency_timer_based","use InnoDB timer based concurrency throttling (backport from MySQL 5.4.0)","",""},
+{"innodb_expand_import","convert .ibd file automatically when import tablespace","the files are generated by xtrabackup export mode.","http://www.percona.com/docs/wiki/percona-xtradb"},
+{"innodb_dict_size_limit","Limit dictionary cache size","Variable innodb_dict_size_limit in bytes","http://www.percona.com/docs/wiki/percona-xtradb"},
+{"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_stats","Additional features about InnoDB statistics/optimizer","","http://www.percona.com/docs/wiki/percona-xtradb"},
 {NULL, NULL, NULL, NULL}
 };

ibuf/ibuf0ibuf.c

@@ -472,6 +472,7 @@ ibuf_init_at_db_start(void)
 
 	/* Use old-style record format for the insert buffer. */
 	table = dict_mem_table_create(IBUF_TABLE_NAME, IBUF_SPACE_ID, 1, 0);
+	table->n_mysql_handles_opened = 1; /* for pin */
 
 	dict_mem_table_add_col(table, heap, "DUMMY_COLUMN", DATA_BINARY, 0, 0);
 

include/buf0buddy.h

@@ -49,10 +49,11 @@ 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));
 
 /**************************************************************************
@@ -63,7 +64,8 @@ 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

@@ -44,10 +44,11 @@ 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));
 
 /**************************************************************************
@@ -58,8 +59,9 @@ 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));
 
 /**************************************************************************
@@ -98,14 +100,15 @@ 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());
+	//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));
 }
 
 /**************************************************************************
@@ -116,11 +119,24 @@ 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());
+	//ut_ad(buf_pool_mutex_own());
+
+	if (!have_page_hash_mutex) {
+		mutex_enter(&LRU_list_mutex);
+		rw_lock_x_lock(&page_hash_latch);
+	}
+
+	mutex_enter(&zip_free_mutex);
+	buf_buddy_free_low(buf, buf_buddy_get_slot(size), TRUE);
+	mutex_exit(&zip_free_mutex);
 
-	buf_buddy_free_low(buf, buf_buddy_get_slot(size));
+	if (!have_page_hash_mutex) {
+		mutex_exit(&LRU_list_mutex);
+		rw_lock_x_unlock(&page_hash_latch);
+	}
 }
 
 #ifdef UNIV_MATERIALIZE

include/buf0buf.h

@@ -1024,7 +1024,7 @@ struct buf_page_struct{
 
 	/* 2. Page flushing fields; protected by buf_pool_mutex */
 
-	UT_LIST_NODE_T(buf_page_t) list;
+	/* UT_LIST_NODE_T(buf_page_t) list; */
 					/* based on state, this is a list
 					node in one of the following lists
 					in buf_pool:
@@ -1034,6 +1034,10 @@ struct buf_page_struct{
 					BUF_BLOCK_ZIP_DIRTY:	flush_list
 					BUF_BLOCK_ZIP_PAGE:	zip_clean
 					BUF_BLOCK_ZIP_FREE:	zip_free[] */
+	/* resplit for optimistic use */
+	UT_LIST_NODE_T(buf_page_t) free;
+	UT_LIST_NODE_T(buf_page_t) flush_list;
+	UT_LIST_NODE_T(buf_page_t) zip_list; /* zip_clean or zip_free[] */
 #ifdef UNIV_DEBUG
 	ibool		in_flush_list;	/* TRUE if in buf_pool->flush_list;
 					when buf_pool_mutex is free, the
@@ -1104,11 +1108,11 @@ struct buf_block_struct{
 					a block is in the unzip_LRU list
 					if page.state == BUF_BLOCK_FILE_PAGE
 					and page.zip.data != NULL */
-#ifdef UNIV_DEBUG
+//#ifdef UNIV_DEBUG
 	ibool		in_unzip_LRU_list;/* TRUE if the page is in the
 					decompressed LRU list;
 					used in debugging */
-#endif /* UNIV_DEBUG */
+//#endif /* UNIV_DEBUG */
 	byte*		frame;		/* pointer to buffer frame which
 					is of size UNIV_PAGE_SIZE, and
 					aligned to an address divisible by
@@ -1316,6 +1320,12 @@ struct buf_pool_struct{
 /* mutex protecting the buffer pool struct and control blocks, except the
 read-write lock in them */
 extern mutex_t	buf_pool_mutex;
+extern mutex_t	LRU_list_mutex;
+extern mutex_t	flush_list_mutex;
+extern rw_lock_t	page_hash_latch;
+extern mutex_t	free_list_mutex;
+extern mutex_t	zip_free_mutex;
+extern mutex_t	zip_hash_mutex;
 /* mutex protecting the control blocks of compressed-only pages
 (of type buf_page_t, not buf_block_t) */
 extern mutex_t	buf_pool_zip_mutex;

include/buf0buf.ic

@@ -100,7 +100,9 @@ buf_pool_get_oldest_modification(void)
 	buf_page_t*	bpage;
 	ib_uint64_t	lsn;
 
-	buf_pool_mutex_enter();
+try_again:
+	//buf_pool_mutex_enter();
+	mutex_enter(&flush_list_mutex);
 
 	bpage = UT_LIST_GET_LAST(buf_pool->flush_list);
 
@@ -109,9 +111,14 @@ buf_pool_get_oldest_modification(void)
 	} else {
 		ut_ad(bpage->in_flush_list);
 		lsn = bpage->oldest_modification;
+		if (lsn == 0) {
+			mutex_exit(&flush_list_mutex);
+			goto try_again;
+		}
 	}
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	mutex_exit(&flush_list_mutex);
 
 	/* The returned answer may be out of date: the flush_list can
 	change after the mutex has been released. */
@@ -128,7 +135,8 @@ buf_pool_clock_tic(void)
 /*====================*/
 			/* out: new clock value */
 {
-	ut_ad(buf_pool_mutex_own());
+	//ut_ad(buf_pool_mutex_own());
+	ut_ad(mutex_own(&LRU_list_mutex));
 
 	buf_pool->ulint_clock++;
 
@@ -246,7 +254,7 @@ buf_page_in_file(
 	case BUF_BLOCK_ZIP_FREE:
 		/* This is a free page in buf_pool->zip_free[].
 		Such pages should only be accessed by the buddy allocator. */
-		ut_error;
+		/* ut_error; */ /* optimistic */
 		break;
 	case BUF_BLOCK_ZIP_PAGE:
 	case BUF_BLOCK_ZIP_DIRTY:
@@ -288,7 +296,7 @@ buf_page_get_LRU_position(
 	const buf_page_t*	bpage)	/* in: control block */
 {
 	ut_ad(buf_page_in_file(bpage));
-	ut_ad(buf_pool_mutex_own());
+	//ut_ad(buf_pool_mutex_own()); /* This is used in optimistic */
 
 	return(bpage->LRU_position);
 }
@@ -305,7 +313,7 @@ buf_page_get_mutex(
 {
 	switch (buf_page_get_state(bpage)) {
 	case BUF_BLOCK_ZIP_FREE:
-		ut_error;
+		/* ut_error; */ /* optimistic */
 		return(NULL);
 	case BUF_BLOCK_ZIP_PAGE:
 	case BUF_BLOCK_ZIP_DIRTY:
@@ -410,7 +418,7 @@ buf_page_set_io_fix(
 	buf_page_t*	bpage,	/* in/out: control block */
 	enum buf_io_fix	io_fix)	/* in: io_fix state */
 {
-	ut_ad(buf_pool_mutex_own());
+	//ut_ad(buf_pool_mutex_own());
 	ut_ad(mutex_own(buf_page_get_mutex(bpage)));
 
 	bpage->io_fix = io_fix;
@@ -438,12 +446,13 @@ buf_page_can_relocate(
 /*==================*/
 	const buf_page_t*	bpage)	/* control block being relocated */
 {
-	ut_ad(buf_pool_mutex_own());
+	//ut_ad(buf_pool_mutex_own());
 	ut_ad(mutex_own(buf_page_get_mutex(bpage)));
 	ut_ad(buf_page_in_file(bpage));
-	ut_ad(bpage->in_LRU_list);
+	/* optimistic */
+	//ut_ad(bpage->in_LRU_list);
 
-	return(buf_page_get_io_fix(bpage) == BUF_IO_NONE
+	return(bpage->in_LRU_list && bpage->io_fix == BUF_IO_NONE
 	       && bpage->buf_fix_count == 0);
 }
 
@@ -457,7 +466,7 @@ buf_page_is_old(
 	const buf_page_t*	bpage)	/* in: control block */
 {
 	ut_ad(buf_page_in_file(bpage));
-	ut_ad(buf_pool_mutex_own());
+	//ut_ad(buf_pool_mutex_own()); /* This is used in optimistic */
 
 	return(bpage->old);
 }
@@ -472,7 +481,8 @@ buf_page_set_old(
 	ibool		old)	/* in: old */
 {
 	ut_a(buf_page_in_file(bpage));
-	ut_ad(buf_pool_mutex_own());
+	//ut_ad(buf_pool_mutex_own());
+	ut_ad(mutex_own(&LRU_list_mutex));
 	ut_ad(bpage->in_LRU_list);
 
 #ifdef UNIV_LRU_DEBUG
@@ -728,17 +738,17 @@ buf_block_free(
 /*===========*/
 	buf_block_t*	block)	/* in, own: block to be freed */
 {
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
 
 	mutex_enter(&block->mutex);
 
 	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);
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
 }
 
 /*************************************************************************
@@ -783,14 +793,23 @@ buf_page_io_query(
 	buf_page_t*	bpage)	/* in: buf_pool block, must be bufferfixed */
 {
 	ibool	io_fixed;
+	mutex_t* block_mutex = buf_page_get_mutex(bpage);
 
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
+retry_lock:
+	mutex_enter(block_mutex);
+	if (block_mutex != buf_page_get_mutex(bpage)) {
+		mutex_exit(block_mutex);
+		block_mutex = buf_page_get_mutex(bpage);
+		goto retry_lock;
+	}
 
 	ut_ad(buf_page_in_file(bpage));
 	ut_ad(bpage->buf_fix_count > 0);
 
 	io_fixed = buf_page_get_io_fix(bpage) != BUF_IO_NONE;
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	mutex_exit(block_mutex);
 
 	return(io_fixed);
 }
@@ -809,7 +828,13 @@ buf_page_get_newest_modification(
 	ib_uint64_t	lsn;
 	mutex_t*	block_mutex = buf_page_get_mutex(bpage);
 
+retry_lock:
 	mutex_enter(block_mutex);
+	if (block_mutex != buf_page_get_mutex(bpage)) {
+		mutex_exit(block_mutex);
+		block_mutex = buf_page_get_mutex(bpage);
+		goto retry_lock;
+	}
 
 	if (buf_page_in_file(bpage)) {
 		lsn = bpage->newest_modification;
@@ -833,7 +858,7 @@ buf_block_modify_clock_inc(
 	buf_block_t*	block)	/* in: block */
 {
 #ifdef UNIV_SYNC_DEBUG
-	ut_ad((buf_pool_mutex_own()
+	ut_ad((mutex_own(&LRU_list_mutex)
 	       && (block->page.buf_fix_count == 0))
 	      || rw_lock_own(&(block->lock), RW_LOCK_EXCLUSIVE));
 #endif /* UNIV_SYNC_DEBUG */
@@ -917,7 +942,11 @@ buf_page_hash_get(
 	ulint		fold;
 
 	ut_ad(buf_pool);
-	ut_ad(buf_pool_mutex_own());
+	//ut_ad(buf_pool_mutex_own());
+#ifdef UNIV_SYNC_DEBUG
+	ut_ad(rw_lock_own(&page_hash_latch, RW_LOCK_EX)
+	      || rw_lock_own(&page_hash_latch, RW_LOCK_SHARED));
+#endif
 
 	/* Look for the page in the hash table */
 
@@ -966,11 +995,13 @@ buf_page_peek(
 {
 	const buf_page_t*	bpage;
 
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
+	rw_lock_s_lock(&page_hash_latch);
 
 	bpage = buf_page_hash_get(space, offset);
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	rw_lock_s_unlock(&page_hash_latch);
 
 	return(bpage != NULL);
 }
@@ -1032,11 +1063,14 @@ buf_page_release(
 	ut_a(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
 	ut_a(block->page.buf_fix_count > 0);
 
+	/* buf_flush_note_modification() should be called before this function. */
+/*
 	if (rw_latch == RW_X_LATCH && mtr->modifications) {
 		buf_pool_mutex_enter();
 		buf_flush_note_modification(block, mtr);
 		buf_pool_mutex_exit();
 	}
+*/
 
 	mutex_enter(&block->mutex);
 

include/buf0flu.ic

@@ -53,13 +53,23 @@ buf_flush_note_modification(
 	buf_block_t*	block,	/* in: block which is modified */
 	mtr_t*		mtr)	/* in: mtr */
 {
+	ibool	use_LRU_mutex = FALSE;
+
+	if (UT_LIST_GET_LEN(buf_pool->unzip_LRU))
+		use_LRU_mutex = TRUE;
+
+	if (use_LRU_mutex)
+		mutex_enter(&LRU_list_mutex);
+
+	mutex_enter(&block->mutex);
+
 	ut_ad(block);
 	ut_ad(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
 	ut_ad(block->page.buf_fix_count > 0);
 #ifdef UNIV_SYNC_DEBUG
 	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_EX));
 #endif /* UNIV_SYNC_DEBUG */
-	ut_ad(buf_pool_mutex_own());
+	//ut_ad(buf_pool_mutex_own());
 
 	ut_ad(mtr->start_lsn != 0);
 	ut_ad(mtr->modifications);
@@ -68,16 +78,23 @@ buf_flush_note_modification(
 	block->page.newest_modification = mtr->end_lsn;
 
 	if (!block->page.oldest_modification) {
+		mutex_enter(&flush_list_mutex);
 
 		block->page.oldest_modification = mtr->start_lsn;
 		ut_ad(block->page.oldest_modification != 0);
 
 		buf_flush_insert_into_flush_list(block);
+		mutex_exit(&flush_list_mutex);
 	} else {
 		ut_ad(block->page.oldest_modification <= mtr->start_lsn);
 	}
 
+	mutex_exit(&block->mutex);
+
 	++srv_buf_pool_write_requests;
+
+	if (use_LRU_mutex)
+		mutex_exit(&LRU_list_mutex);
 }
 
 /************************************************************************
@@ -92,6 +109,16 @@ buf_flush_recv_note_modification(
 	ib_uint64_t	end_lsn)	/* in: end lsn of the last mtr in the
 					set of mtr's */
 {
+	ibool	use_LRU_mutex = FALSE;
+
+	if(UT_LIST_GET_LEN(buf_pool->unzip_LRU))
+		use_LRU_mutex = TRUE;
+
+	if (use_LRU_mutex)
+		mutex_enter(&LRU_list_mutex);
+
+	mutex_enter(&(block->mutex));
+
 	ut_ad(block);
 	ut_ad(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
 	ut_ad(block->page.buf_fix_count > 0);
@@ -99,22 +126,27 @@ buf_flush_recv_note_modification(
 	ut_ad(rw_lock_own(&(block->lock), RW_LOCK_EX));
 #endif /* UNIV_SYNC_DEBUG */
 
-	buf_pool_mutex_enter();
+	//buf_pool_mutex_enter();
 
 	ut_ad(block->page.newest_modification <= end_lsn);
 
 	block->page.newest_modification = end_lsn;
 
 	if (!block->page.oldest_modification) {
+		mutex_enter(&flush_list_mutex);
 
 		block->page.oldest_modification = start_lsn;
 
 		ut_ad(block->page.oldest_modification != 0);
 
 		buf_flush_insert_sorted_into_flush_list(block);
+		mutex_exit(&flush_list_mutex);
 	} else {
 		ut_ad(block->page.oldest_modification <= start_lsn);
 	}
 
-	buf_pool_mutex_exit();
+	//buf_pool_mutex_exit();
+	if (use_LRU_mutex)
+		mutex_exit(&LRU_list_mutex);
+	mutex_exit(&(block->mutex));
 }

include/buf0lru.h

@@ -122,10 +122,11 @@ buf_LRU_free_block(
 	buf_page_t*	bpage,	/* in: block to be freed */
 	ibool		zip,	/* in: TRUE if should remove also the
 				compressed page of an uncompressed page */
-	ibool*		buf_pool_mutex_released);
+	ibool*		buf_pool_mutex_released,
 				/* in: pointer to a variable that will
 				be assigned TRUE if buf_pool_mutex
 				was temporarily released, or NULL */
+	ibool		have_LRU_mutex);
 /**********************************************************************
 Try to free a replaceable block. */
 UNIV_INTERN
@@ -169,7 +170,8 @@ 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/dict0dict.h

@@ -1102,6 +1102,12 @@ dict_table_get_index_on_name_and_min_id(
 				/* out: index, NULL if does not exist */
 	dict_table_t*	table,	/* in: table */
 	const char*	name);	/* in: name of the index to find */
+
+UNIV_INTERN
+void
+dict_table_LRU_trim(
+/*================*/
+	dict_table_t*	self);
 /* Buffers for storing detailed information about the latest foreign key
 and unique key errors */
 extern FILE*	dict_foreign_err_file;

include/dict0dict.ic

@@ -723,6 +723,13 @@ dict_table_check_if_in_cache_low(
 	HASH_SEARCH(name_hash, dict_sys->table_hash, table_fold,
 		    dict_table_t*, table, ut_ad(table->cached),
 		    !strcmp(table->name, table_name));
+
+	/* make young in table_LRU */
+	if (table) {
+		UT_LIST_REMOVE(table_LRU, dict_sys->table_LRU, table);
+		UT_LIST_ADD_FIRST(table_LRU, dict_sys->table_LRU, table);
+	}
+
 	return(table);
 }
 
@@ -776,6 +783,12 @@ dict_table_get_on_id_low(
 		table = dict_load_table_on_id(table_id);
 	}
 
+	/* make young in table_LRU */
+	if (table) {
+		UT_LIST_REMOVE(table_LRU, dict_sys->table_LRU, table);
+		UT_LIST_ADD_FIRST(table_LRU, dict_sys->table_LRU, table);
+	}
+
 	ut_ad(!table || table->cached);
 
 	/* TODO: should get the type information from MySQL */

include/log0log.h

@@ -186,6 +186,13 @@ void
 log_buffer_flush_to_disk(void);
 /*==========================*/
 /********************************************************************
+Flushes the log buffer. Forces it to disk depending on the value of
+the configuration parameter innodb_flush_log_at_trx_commit. */
+UNIV_INTERN
+void
+log_buffer_flush_maybe_sync(void);
+/*=============================*/
+/********************************************************************
 Advances the smallest lsn for which there are unflushed dirty blocks in the
 buffer pool and also may make a new checkpoint. NOTE: this function may only
 be called if the calling thread owns no synchronization objects! */

include/rem0cmp.h

@@ -177,10 +177,11 @@ cmp_rec_rec_with_match(
 				matched fields; when the function returns,
 				contains the value the for current
 				comparison */
-	ulint*		matched_bytes);/* in/out: number of already matched
+	ulint*		matched_bytes, /* in/out: number of already matched
 				bytes within the first field not completely
 				matched; when the function returns, contains
 				the value for the current comparison */
+	ulint		stats_method);
 /*****************************************************************
 This function is used to compare two physical records. Only the common
 first fields are compared. */

include/rem0cmp.ic

@@ -88,5 +88,5 @@ cmp_rec_rec(
 	ulint	match_b		= 0;
 
 	return(cmp_rec_rec_with_match(rec1, rec2, offsets1, offsets2, index,
-				      &match_f, &match_b));
+				      &match_f, &match_b, 0));
 }

include/srv0srv.h

@@ -127,6 +127,8 @@ extern ulint	srv_buf_pool_curr_size;	/* current size in bytes */
 extern ulint	srv_mem_pool_size;
 extern ulint	srv_lock_table_size;
 
+extern ibool	srv_thread_concurrency_timer_based;
+
 extern ulint	srv_n_file_io_threads;
 extern ulint	srv_n_read_io_threads;
 extern ulint	srv_n_write_io_threads;
@@ -163,6 +165,11 @@ extern ulint	srv_fast_shutdown;	 /* If this is 1, do not do a
 extern ibool	srv_innodb_status;
 
 extern unsigned long long	srv_stats_sample_pages;
+extern ulint	srv_stats_method;
+#define SRV_STATS_METHOD_NULLS_EQUAL     0
+#define SRV_STATS_METHOD_NULLS_NOT_EQUAL 1
+#define SRV_STATS_METHOD_IGNORE_NULLS    2
+extern ulint	srv_stats_auto_update;
 
 extern ibool	srv_use_doublewrite_buf;
 extern ibool	srv_use_checksums;
@@ -184,8 +191,10 @@ extern ulint	srv_enable_unsafe_group_commit;
 extern ulint	srv_read_ahead;
 extern ulint	srv_adaptive_checkpoint;
 
-extern ulint	srv_extra_rsegments;
+extern ulint	srv_expand_import;
 
+extern ulint	srv_extra_rsegments;
+extern ulint	srv_dict_size_limit;
 /*-------------------------------------------*/
 
 extern ulint	srv_n_rows_inserted;
@@ -552,6 +561,7 @@ struct export_var_struct{
 	ulint innodb_data_writes;
 	ulint innodb_data_written;
 	ulint innodb_data_reads;
+	ulint innodb_dict_tables;
 	ulint innodb_buffer_pool_pages_total;
 	ulint innodb_buffer_pool_pages_data;
 	ulint innodb_buffer_pool_pages_dirty;

include/sync0sync.h

@@ -464,8 +464,14 @@ or row lock! */
 					SYNC_SEARCH_SYS, as memory allocation
 					can call routines there! Otherwise
 					the level is SYNC_MEM_HASH. */
+#define SYNC_BUF_LRU_LIST	157
+#define SYNC_BUF_PAGE_HASH	156
+#define	SYNC_BUF_BLOCK		155
+#define SYNC_BUF_FREE_LIST	153
+#define SYNC_BUF_ZIP_FREE	152
+#define SYNC_BUF_ZIP_HASH	151
 #define	SYNC_BUF_POOL		150
-#define	SYNC_BUF_BLOCK		149
+#define SYNC_BUF_FLUSH_LIST	149
 #define SYNC_DOUBLEWRITE	140
 #define	SYNC_ANY_LATCH		135
 #define SYNC_THR_LOCAL		133

include/ut0auxconf.h

@@ -12,3 +12,8 @@ If by any chance Makefile.in and ./configure are regenerated and thus
 the hack from Makefile.in wiped away then the "real" check from plug.in
 will take over.
 */
+/* This is temprary fix for http://bugs.mysql.com/43740 */
+/* force to enable */
+#ifdef HAVE_GCC_ATOMIC_BUILTINS
+#define HAVE_ATOMIC_PTHREAD_T
+#endif

log/log0log.c

@@ -1525,6 +1525,26 @@ log_buffer_flush_to_disk(void)
 	log_write_up_to(lsn, LOG_WAIT_ALL_GROUPS, TRUE);
 }
 
+/********************************************************************
+Flush the log buffer. Force it to disk depending on the value of
+innodb_flush_log_at_trx_commit. */
+UNIV_INTERN
+void
+log_buffer_flush_maybe_sync(void)
+/*=============================*/
+{
+	ib_uint64_t	lsn;
+
+	mutex_enter(&(log_sys->mutex));
+
+	lsn = log_sys->lsn;
+
+	mutex_exit(&(log_sys->mutex));
+
+	/* Force log buffer to disk when innodb_flush_log_at_trx_commit = 1. */
+	log_write_up_to(lsn, LOG_WAIT_ALL_GROUPS,
+			srv_flush_log_at_trx_commit == 1 ? TRUE : FALSE);
+}
 /********************************************************************
 Tries to establish a big enough margin of free space in the log buffer, such
 that a new log entry can be catenated without an immediate need for a flush. */

mtr/mtr0mtr.c

@@ -102,6 +102,38 @@ mtr_memo_pop_all(
 	}
 }
 
+UNIV_INLINE
+void
+mtr_memo_note_modification_all(
+/*===========================*/
+	mtr_t*	mtr)	/* in: mtr */
+{
+	mtr_memo_slot_t* slot;
+	dyn_array_t*	memo;
+	ulint		offset;
+
+	ut_ad(mtr);
+	ut_ad(mtr->magic_n == MTR_MAGIC_N);
+	ut_ad(mtr->state == MTR_COMMITTING); /* Currently only used in
+					     commit */
+	ut_ad(mtr->modifications);
+
+	memo = &(mtr->memo);
+
+	offset = dyn_array_get_data_size(memo);
+
+	while (offset > 0) {
+		offset -= sizeof(mtr_memo_slot_t);
+		slot = dyn_array_get_element(memo, offset);
+
+		if (UNIV_LIKELY(slot->object != NULL) &&
+		    slot->type == MTR_MEMO_PAGE_X_FIX) {
+			buf_flush_note_modification(
+				(buf_block_t*)slot->object, mtr);
+		}
+	}
+}
+
 /****************************************************************
 Writes the contents of a mini-transaction log, if any, to the database log. */
 static
@@ -180,6 +212,8 @@ mtr_commit(
 
 	if (write_log) {
 		mtr_log_reserve_and_write(mtr);
+
+		mtr_memo_note_modification_all(mtr);
 	}
 
 	/* We first update the modification info to buffer pages, and only
@@ -190,12 +224,13 @@ mtr_commit(
 	required when we insert modified buffer pages in to the flush list
 	which must be sorted on oldest_modification. */
 
-	mtr_memo_pop_all(mtr);
-
 	if (write_log) {
 		log_release();
 	}
 
+	/* All unlocking has been moved here, after log_sys mutex release. */
+	mtr_memo_pop_all(mtr);
+
 	ut_d(mtr->state = MTR_COMMITTED);
 	dyn_array_free(&(mtr->memo));
 	dyn_array_free(&(mtr->log));
@@ -263,6 +298,12 @@ mtr_memo_release(
 		slot = dyn_array_get_element(memo, offset);
 
 		if ((object == slot->object) && (type == slot->type)) {
+			if (mtr->modifications &&
+			    UNIV_LIKELY(slot->object != NULL) &&
+			    slot->type == MTR_MEMO_PAGE_X_FIX) {
+				buf_flush_note_modification(
+					(buf_block_t*)slot->object, mtr);
+			}
 
 			mtr_memo_slot_release(mtr, slot);
 

mysql-test/innodb-index.result

@@ -46,13 +46,6 @@ t1	CREATE TABLE `t1` (
   KEY `d2` (`d`),
   KEY `b` (`b`)
 ) ENGINE=InnoDB DEFAULT CHARSET=latin1
-CREATE TABLE `t1#1`(a INT PRIMARY KEY) ENGINE=InnoDB;
-alter table t1 add unique index (c), add index (d);
-ERROR HY000: Table 'test.t1#1' already exists
-rename table `t1#1` to `t1#2`;
-alter table t1 add unique index (c), add index (d);
-ERROR HY000: Table 'test.t1#2' already exists
-drop table `t1#2`;
 alter table t1 add unique index (c), add index (d);
 show create table t1;
 Table	Create Table

mysql-test/innodb-index.test

@@ -17,16 +17,6 @@ show create table t1;
 alter table t1 add index (b);
 show create table t1;
 
-# Check how existing tables interfere with temporary tables.
-CREATE TABLE `t1#1`(a INT PRIMARY KEY) ENGINE=InnoDB;
-
---error 156
-alter table t1 add unique index (c), add index (d);
-rename table `t1#1` to `t1#2`;
---error 156
-alter table t1 add unique index (c), add index (d);
-drop table `t1#2`;
-
 alter table t1 add unique index (c), add index (d);
 show create table t1;
 explain select * from t1 force index(c) order by c;

mysql-test/innodb_bug34300.test

@@ -30,3 +30,4 @@ ALTER TABLE bug34300 ADD COLUMN (f10 INT);
 SELECT f4, f8 FROM bug34300;
 
 DROP TABLE bug34300;
+SET @@global.max_allowed_packet=1048576;

mysql-test/innodb_bug36169.result

+set @old_innodb_file_per_table=@@innodb_file_per_table;
+set @old_innodb_file_format=@@innodb_file_format;
+set @old_innodb_file_format_check=@@innodb_file_format_check;
 SET GLOBAL innodb_file_format='Barracuda';
 SET GLOBAL innodb_file_per_table=ON;

mysql-test/innodb_bug36169.test

@@ -4,6 +4,9 @@
 #
 
 -- source include/have_innodb.inc
+set @old_innodb_file_per_table=@@innodb_file_per_table;
+set @old_innodb_file_format=@@innodb_file_format;
+set @old_innodb_file_format_check=@@innodb_file_format_check;
 
 SET GLOBAL innodb_file_format='Barracuda';
 SET GLOBAL innodb_file_per_table=ON;
@@ -1153,3 +1156,7 @@ DROP TABLE IF EXISTS table4;
 DROP TABLE IF EXISTS table5;
 DROP TABLE IF EXISTS table6;
 
+set global innodb_file_per_table=@old_innodb_file_per_table;
+set global innodb_file_format=@old_innodb_file_format;
+set global innodb_file_format_check=@old_innodb_file_format_check;
+

mysql-test/innodb_bug36172.test

@@ -13,6 +13,9 @@ SET storage_engine=InnoDB;
 
 -- disable_query_log
 -- disable_result_log
+set @old_innodb_file_per_table=@@innodb_file_per_table;
+set @old_innodb_file_format=@@innodb_file_format;
+set @old_innodb_file_format_check=@@innodb_file_format_check;
 
 SET GLOBAL innodb_file_format='Barracuda';
 SET GLOBAL innodb_file_per_table=on;
@@ -24,3 +27,7 @@ CHECK TABLE table0 EXTENDED;
 INSERT IGNORE INTO `table0` SET `col19` = '19940127002709', `col20` = 2383927.9055146948, `col21` = 4293243420.5621204000, `col22` = '20511211123705', `col23` = 4289899778.6573381000, `col24` = 4293449279.0540481000, `col25` = 'emphysemic', `col26` = 'dentally', `col27` = '2347406', `col28` = 'eruct', `col30` = 1222, `col31` = 4294372994.9941406000, `col32` = 4291385574.1173744000, `col33` = 'borrowing\'s', `col34` = 'septics', `col35` = 'ratter\'s', `col36` = 'Kaye', `col37` = 'Florentia', `col38` = 'allium', `col39` = 'barkeep', `col40` = '19510407003441', `col41` = 4293559200.4215522000, `col42` = 22482, `col43` = 'decussate', `col44` = 'Brom\'s', `col45` = 'violated', `col46` = 4925506.4635456400, `col47` = 930549, `col48` = '51296066', `col49` = 'voluminously', `col50` = '29306676', `col51` = -88, `col52` = -2153690, `col53` = 4290250202.1464887000, `col54` = 'expropriation', `col55` = 'Aberdeen\'s', `col56` = 20343, `col58` = '19640415171532', `col59` = 'extern', `col60` = 'Ubana', `col61` = 4290487961.8539081000, `col62` = '2147', `col63` = -24271, `col64` = '20750801194548', `col65` = 'Cunaxa\'s', `col66` = 'pasticcio', `col67` = 2795817, `col68` = 'Indore\'s', `col70` = 6864127, `col71` = '1817832', `col72` = '20540506114211', `col73` = '20040101012300', `col74` = 'rationalized', `col75` = '45522', `col76` = 'indene', `col77` = -6964559, `col78` = 4247535.5266884370, `col79` = '20720416124357', `col80` = '2143', `col81` = 4292060102.4466386000, `col82` = 'striving', `col83` = 'boneblack\'s', `col84` = 'redolent', `col85` = 6489697.9009369183, `col86` = 4287473465.9731131000, `col87` = 7726015, `col88` = 'perplexed', `col89` = '17153791', `col90` = 5478587.1108127078, `col91` = 4287091404.7004304000, `col92` = 'Boulez\'s', `col93` = '2931278';
 CHECK TABLE table0 EXTENDED;
 DROP TABLE table0;
+set global innodb_file_per_table=@old_innodb_file_per_table;
+set global innodb_file_format=@old_innodb_file_format;
+set global innodb_file_format_check=@old_innodb_file_format_check;
+

rem/rem0cmp.c

@@ -892,10 +892,11 @@ cmp_rec_rec_with_match(
 				matched fields; when the function returns,
 				contains the value the for current
 				comparison */
-	ulint*		matched_bytes) /* in/out: number of already matched
+	ulint*		matched_bytes, /* in/out: number of already matched
 				bytes within the first field not completely
 				matched; when the function returns, contains
 				the value for the current comparison */
+	ulint		stats_method)
 {
 #ifndef UNIV_HOTBACKUP
 	ulint		rec1_n_fields;	/* the number of fields in rec */
@@ -989,7 +990,11 @@ cmp_rec_rec_with_match(
 
 				if (rec1_f_len == rec2_f_len) {
 
-					goto next_field;
+					if (stats_method == SRV_STATS_METHOD_NULLS_EQUAL) {
+						goto next_field;
+					} else {
+						ret = -1;
+					}
 
 				} else if (rec2_f_len == UNIV_SQL_NULL) {
 

row/row0mysql.c

@@ -854,6 +854,9 @@ row_update_statistics_if_needed(
 
 	table->stat_modified_counter = counter + 1;
 
+	if (!srv_stats_auto_update)
+		return;
+
 	/* Calculate new statistics if 1 / 16 of table has been modified
 	since the last time a statistics batch was run, or if
 	stat_modified_counter > 2 000 000 000 (to avoid wrap-around).

scripts/install_innodb_plugins.sql

@@ -12,3 +12,5 @@ INSTALL PLUGIN INNODB_BUFFER_POOL_PAGES SONAME 'ha_innodb.so';
 INSTALL PLUGIN INNODB_BUFFER_POOL_PAGES_BLOB SONAME 'ha_innodb.so';
 INSTALL PLUGIN INNODB_BUFFER_POOL_PAGES_INDEX SONAME 'ha_innodb.so';
 INSTALL PLUGIN innodb_rseg SONAME 'ha_innodb.so';
+INSTALL PLUGIN innodb_table_stats SONAME 'ha_innodb.so';
+INSTALL PLUGIN innodb_index_stats SONAME 'ha_innodb.so';