~tsarev/percona-server/5.5-processlist_rows_stats-sporadic_fails-fix : contents of innodb_dict_size

~tsarev/percona-server/5.5-processlist_rows_stats-sporadic_fails-fix : (revision 130)
# name       : innodb_dict_size_limit.patch
# introduced : 11 or before
# maintainer : Yasufumi
#
#!!! notice !!!
# Any small change to this file in the main branch
# should be done or reviewed by the maintainer!
diff -ruN a/storage/innobase/btr/btr0sea.c b/storage/innobase/btr/btr0sea.c
--- a/storage/innobase/btr/btr0sea.c	2010-11-03 07:01:13.000000000 +0900
+++ b/storage/innobase/btr/btr0sea.c	2010-12-03 15:45:47.503988924 +0900
@@ -1185,6 +1185,179 @@
 	mem_free(folds);
 }
 
+/************************************************************************
+Drops a page hash index based on index */
+UNIV_INTERN
+void
+btr_search_drop_page_hash_index_on_index(
+/*=====================================*/
+	dict_index_t*	index)		/* in: record descriptor */
+{
+
+	hash_table_t*	table;
+	buf_block_t*	block;
+	ulint		n_fields;
+	ulint		n_bytes;
+	const page_t*		page;
+	const rec_t*		rec;
+	ulint		fold;
+	ulint		prev_fold;
+	index_id_t	index_id;
+	ulint		n_cached;
+	ulint		n_recs;
+	ulint*		folds;
+	ulint		i, j;
+	mem_heap_t*	heap	= NULL;
+	ulint*		offsets;
+	ibool		released_search_latch;
+
+	rw_lock_s_lock(&btr_search_latch);
+
+	table = btr_search_sys->hash_index;
+
+	for (j = 0; j < srv_buf_pool_instances; j++) {
+		buf_pool_t*	buf_pool;
+
+		buf_pool = buf_pool_from_array(j);
+
+		do {
+			buf_chunk_t*	chunks	= buf_pool->chunks;
+			buf_chunk_t*	chunk	= chunks + buf_pool->n_chunks;
+
+			released_search_latch = FALSE;
+
+			while (--chunk >= chunks) {
+				block	= chunk->blocks;
+				i	= chunk->size;
+
+retry:
+				for (; i--; block++) {
+					if (buf_block_get_state(block)
+					    != BUF_BLOCK_FILE_PAGE
+					    || block->index != index
+					    || !block->is_hashed) {
+						continue;
+					}
+
+					page = block->frame;
+
+					/* from btr_search_drop_page_hash_index() */
+					n_fields = block->curr_n_fields;
+					n_bytes = block->curr_n_bytes;
+
+
+					/* keeping latch order */
+					rw_lock_s_unlock(&btr_search_latch);
+					released_search_latch = TRUE;
+					rw_lock_x_lock(&block->lock);
+
+
+					ut_a(n_fields + n_bytes > 0);
+
+					n_recs = page_get_n_recs(page);
+
+					/* Calculate and cache fold values into an array for fast deletion
+					from the hash index */
+
+					folds = mem_alloc(n_recs * sizeof(ulint));
+
+					n_cached = 0;
+
+					rec = page_get_infimum_rec(page);
+					rec = page_rec_get_next_low(rec, page_is_comp(page));
+
+					index_id = btr_page_get_index_id(page);
+	
+					ut_a(index_id == index->id);
+
+					prev_fold = 0;
+
+					offsets = NULL;
+
+					while (!page_rec_is_supremum(rec)) {
+						offsets = rec_get_offsets(rec, index, offsets,
+									n_fields + (n_bytes > 0), &heap);
+						ut_a(rec_offs_n_fields(offsets) == n_fields + (n_bytes > 0));
+						fold = rec_fold(rec, offsets, n_fields, n_bytes, index_id);
+
+						if (fold == prev_fold && prev_fold != 0) {
+
+							goto next_rec;
+						}
+
+						/* Remove all hash nodes pointing to this page from the
+						hash chain */
+
+						folds[n_cached] = fold;
+						n_cached++;
+next_rec:
+						rec = page_rec_get_next_low(rec, page_rec_is_comp(rec));
+						prev_fold = fold;
+					}
+
+					if (UNIV_LIKELY_NULL(heap)) {
+						mem_heap_empty(heap);
+					}
+
+					rw_lock_x_lock(&btr_search_latch);
+
+					if (UNIV_UNLIKELY(!block->is_hashed)) {
+						goto cleanup;
+					}
+
+					ut_a(block->index == index);
+
+					if (UNIV_UNLIKELY(block->curr_n_fields != n_fields)
+					    || UNIV_UNLIKELY(block->curr_n_bytes != n_bytes)) {
+						rw_lock_x_unlock(&btr_search_latch);
+						rw_lock_x_unlock(&block->lock);
+
+						mem_free(folds);
+
+						rw_lock_s_lock(&btr_search_latch);
+						goto retry;
+					}
+
+					for (i = 0; i < n_cached; i++) {
+
+						ha_remove_all_nodes_to_page(table, folds[i], page);
+					}
+
+					ut_a(index->search_info->ref_count > 0);
+					index->search_info->ref_count--;
+
+					block->is_hashed = FALSE;
+					block->index = NULL;
+
+cleanup:
+#if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG
+					if (UNIV_UNLIKELY(block->n_pointers)) {
+						/* Corruption */
+						ut_print_timestamp(stderr);
+						fprintf(stderr,
+"InnoDB: The adaptive hash index is corrupted. After dropping\n"
+"InnoDB: the hash index to a page of %s, %lu hash nodes still remain.\n",
+							index->name, (ulong) block->n_pointers);
+					}
+#endif /* UNIV_AHI_DEBUG || UNIV_DEBUG */
+					rw_lock_x_unlock(&btr_search_latch);
+					rw_lock_x_unlock(&block->lock);
+
+					mem_free(folds);
+
+					rw_lock_s_lock(&btr_search_latch);
+				}
+			}
+		} while (released_search_latch);
+	}
+
+	rw_lock_s_unlock(&btr_search_latch);
+
+	if (UNIV_LIKELY_NULL(heap)) {
+		mem_heap_free(heap);
+	}
+}
+
 /********************************************************************//**
 Drops a page hash index when a page is freed from a fseg to the file system.
 Drops possible hash index if the page happens to be in the buffer pool. */
diff -ruN a/storage/innobase/buf/buf0buf.c b/storage/innobase/buf/buf0buf.c
--- a/storage/innobase/buf/buf0buf.c	2011-02-01 18:00:03.000000000 +0900
+++ b/storage/innobase/buf/buf0buf.c	2011-02-01 18:01:59.000000000 +0900
@@ -294,14 +294,14 @@
 # endif /* !PFS_SKIP_BUFFER_MUTEX_RWLOCK */
 #endif /* UNIV_PFS_MUTEX || UNIV_PFS_RWLOCK */
 
-/** A chunk of buffers.  The buffer pool is allocated in chunks. */
-struct buf_chunk_struct{
-	ulint		mem_size;	/*!< allocated size of the chunk */
-	ulint		size;		/*!< size of frames[] and blocks[] */
-	void*		mem;		/*!< pointer to the memory area which
-					was allocated for the frames */
-	buf_block_t*	blocks;		/*!< array of buffer control blocks */
-};
+/** A chunk of buffers.  The buffer pool is allocated in chunks. (moved to buf0buf.h)*/
+//struct buf_chunk_struct{
+//	ulint		mem_size;	/*!< allocated size of the chunk */
+//	ulint		size;		/*!< size of frames[] and blocks[] */
+//	void*		mem;		/*!< pointer to the memory area which
+//					was allocated for the frames */
+//	buf_block_t*	blocks;		/*!< array of buffer control blocks */
+//};
 #endif /* !UNIV_HOTBACKUP */
 
 /********************************************************************//**
diff -ruN a/storage/innobase/dict/dict0boot.c b/storage/innobase/dict/dict0boot.c
--- a/storage/innobase/dict/dict0boot.c	2010-11-03 07:01:13.000000000 +0900
+++ b/storage/innobase/dict/dict0boot.c	2010-12-03 15:45:47.503988924 +0900
@@ -284,6 +284,7 @@
 	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);
@@ -336,6 +337,7 @@
 
 	/*-------------------------*/
 	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);
@@ -368,6 +370,7 @@
 
 	/*-------------------------*/
 	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);
@@ -413,6 +416,7 @@
 
 	/*-------------------------*/
 	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);
diff -ruN a/storage/innobase/dict/dict0crea.c b/storage/innobase/dict/dict0crea.c
--- a/storage/innobase/dict/dict0crea.c	2010-11-03 07:01:13.000000000 +0900
+++ b/storage/innobase/dict/dict0crea.c	2010-12-03 15:45:47.521955810 +0900
@@ -1210,6 +1210,9 @@
 		/* 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);
@@ -1291,6 +1294,11 @@
 
 	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);
diff -ruN a/storage/innobase/dict/dict0dict.c b/storage/innobase/dict/dict0dict.c
--- a/storage/innobase/dict/dict0dict.c	2010-11-03 07:01:13.000000000 +0900
+++ b/storage/innobase/dict/dict0dict.c	2010-12-03 15:45:47.525953769 +0900
@@ -626,6 +626,8 @@
 
 	table = dict_table_get_on_id_low(table_id);
 
+	dict_table_LRU_trim(table);
+
 	mutex_exit(&(dict_sys->mutex));
 
 	return(table);
@@ -744,6 +746,8 @@
 		table->n_mysql_handles_opened++;
 	}
 
+	dict_table_LRU_trim(table);
+
 	mutex_exit(&(dict_sys->mutex));
 
 	if (table != NULL) {
@@ -1259,6 +1263,64 @@
 	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.
@@ -1728,6 +1790,11 @@
 	ut_ad(index->magic_n == DICT_INDEX_MAGIC_N);
 	ut_ad(mutex_own(&(dict_sys->mutex)));
 
+	/* remove all entry of the index from adaptive hash index,
+	because removing from adaptive hash index needs dict_index */
+	if (btr_search_enabled && srv_dict_size_limit)
+		btr_search_drop_page_hash_index_on_index(index);
+
 	/* We always create search info whether or not adaptive
 	hash index is enabled or not. */
 	info = index->search_info;
diff -ruN a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc
--- a/storage/innobase/handler/ha_innodb.cc	2010-12-03 15:43:57.294986852 +0900
+++ b/storage/innobase/handler/ha_innodb.cc	2010-12-03 15:45:47.534959966 +0900
@@ -674,6 +674,8 @@
   (char*) &export_vars.innodb_dblwr_pages_written,	  SHOW_LONG},
   {"dblwr_writes",
   (char*) &export_vars.innodb_dblwr_writes,		  SHOW_LONG},
+  {"dict_tables",
+  (char*) &export_vars.innodb_dict_tables,		  SHOW_LONG},
   {"have_atomic_builtins",
   (char*) &export_vars.innodb_have_atomic_builtins,	  SHOW_BOOL},
   {"log_waits",
@@ -11642,6 +11644,11 @@
   "Choose method of innodb_adaptive_flushing. (native, [estimate], keep_average)",
   NULL, innodb_adaptive_flushing_method_update, 1, &adaptive_flushing_method_typelib);
 
+static MYSQL_SYSVAR_ULONG(dict_size_limit, srv_dict_size_limit,
+  PLUGIN_VAR_RQCMDARG,
+  "Limit the allocated memory for dictionary cache. (0: unlimited)",
+  NULL, NULL, 0, 0, LONG_MAX, 0);
+
 static struct st_mysql_sys_var* innobase_system_variables[]= {
   MYSQL_SYSVAR(additional_mem_pool_size),
   MYSQL_SYSVAR(autoextend_increment),
@@ -11709,6 +11716,7 @@
   MYSQL_SYSVAR(flush_neighbor_pages),
   MYSQL_SYSVAR(read_ahead),
   MYSQL_SYSVAR(adaptive_flushing_method),
+  MYSQL_SYSVAR(dict_size_limit),
   MYSQL_SYSVAR(use_sys_malloc),
   MYSQL_SYSVAR(use_native_aio),
   MYSQL_SYSVAR(change_buffering),
diff -ruN a/storage/innobase/ibuf/ibuf0ibuf.c b/storage/innobase/ibuf/ibuf0ibuf.c
--- a/storage/innobase/ibuf/ibuf0ibuf.c	2010-12-03 15:18:48.889024455 +0900
+++ b/storage/innobase/ibuf/ibuf0ibuf.c	2010-12-03 15:45:47.553025057 +0900
@@ -566,6 +566,7 @@
 
 	/* 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);
 
diff -ruN a/storage/innobase/include/btr0sea.h b/storage/innobase/include/btr0sea.h
--- a/storage/innobase/include/btr0sea.h	2010-11-03 07:01:13.000000000 +0900
+++ b/storage/innobase/include/btr0sea.h	2010-12-03 15:45:47.555024229 +0900
@@ -140,6 +140,13 @@
 				s- or x-latched, or an index page
 				for which we know that
 				block->buf_fix_count == 0 */
+/************************************************************************
+Drops a page hash index based on index */
+UNIV_INTERN
+void
+btr_search_drop_page_hash_index_on_index(
+/*=====================================*/
+	dict_index_t*	index);		/* in: record descriptor */
 /********************************************************************//**
 Drops a page hash index when a page is freed from a fseg to the file system.
 Drops possible hash index if the page happens to be in the buffer pool. */
diff -ruN a/storage/innobase/include/buf0buf.h b/storage/innobase/include/buf0buf.h
--- a/storage/innobase/include/buf0buf.h	2011-02-01 18:00:03.000000000 +0900
+++ b/storage/innobase/include/buf0buf.h	2011-02-01 18:03:29.000000000 +0900
@@ -1555,6 +1555,15 @@
 #define BUF_POOL_ZIP_FOLD_BPAGE(b) BUF_POOL_ZIP_FOLD((buf_block_t*) (b))
 /* @} */
 
+/** A chunk of buffers.  The buffer pool is allocated in chunks. */
+struct buf_chunk_struct{
+	ulint		mem_size;	/*!< allocated size of the chunk */
+	ulint		size;		/*!< size of frames[] and blocks[] */
+	void*		mem;		/*!< pointer to the memory area which
+					was allocated for the frames */
+	buf_block_t*	blocks;		/*!< array of buffer control blocks */
+};
+
 /** @brief The buffer pool statistics structure. */
 struct buf_pool_stat_struct{
 	ulint	n_page_gets;	/*!< number of page gets performed;
diff -ruN a/storage/innobase/include/dict0dict.h b/storage/innobase/include/dict0dict.h
--- a/storage/innobase/include/dict0dict.h	2010-11-03 07:01:13.000000000 +0900
+++ b/storage/innobase/include/dict0dict.h	2010-12-03 15:45:47.558024515 +0900
@@ -1170,6 +1170,12 @@
 /*====================================*/
 	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;
diff -ruN a/storage/innobase/include/dict0dict.ic b/storage/innobase/include/dict0dict.ic
--- a/storage/innobase/include/dict0dict.ic	2010-11-03 07:01:13.000000000 +0900
+++ b/storage/innobase/include/dict0dict.ic	2010-12-03 15:45:47.560024398 +0900
@@ -824,6 +824,13 @@
 	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);
 }
 
@@ -905,6 +912,12 @@
 		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 */
diff -ruN a/storage/innobase/include/srv0srv.h b/storage/innobase/include/srv0srv.h
--- a/storage/innobase/include/srv0srv.h	2010-12-03 15:43:57.297067100 +0900
+++ b/storage/innobase/include/srv0srv.h	2010-12-03 15:45:47.562024404 +0900
@@ -229,6 +229,7 @@
 extern ulint	srv_read_ahead;
 extern ulint	srv_adaptive_flushing_method;
 
+extern ulint	srv_dict_size_limit;
 /*-------------------------------------------*/
 
 extern ulint	srv_n_rows_inserted;
@@ -709,6 +710,7 @@
 	ulint innodb_data_writes;		/*!< I/O write requests */
 	ulint innodb_data_written;		/*!< Data bytes written */
 	ulint innodb_data_reads;		/*!< I/O read requests */
+	ulint innodb_dict_tables;
 	ulint innodb_buffer_pool_pages_total;	/*!< Buffer pool size */
 	ulint innodb_buffer_pool_pages_data;	/*!< Data pages */
 	ulint innodb_buffer_pool_pages_dirty;	/*!< Dirty data pages */
diff -ruN a/storage/innobase/srv/srv0srv.c b/storage/innobase/srv/srv0srv.c
--- a/storage/innobase/srv/srv0srv.c	2010-12-03 15:43:57.301024390 +0900
+++ b/storage/innobase/srv/srv0srv.c	2010-12-03 15:45:47.565023830 +0900
@@ -415,6 +415,8 @@
 UNIV_INTERN ulint	srv_enable_unsafe_group_commit = 0; /* 0:disable 1:enable */
 UNIV_INTERN ulint	srv_read_ahead = 3; /* 1: random  2: linear  3: Both */
 UNIV_INTERN ulint	srv_adaptive_flushing_method = 0; /* 0: native  1: estimate  2: keep_average */
+
+UNIV_INTERN ulint	srv_dict_size_limit = 0;
 /*-------------------------------------------*/
 UNIV_INTERN ulong	srv_n_spin_wait_rounds	= 30;
 UNIV_INTERN ulong	srv_n_free_tickets_to_enter = 500;
@@ -2220,6 +2222,7 @@
 	export_vars.innodb_data_reads = os_n_file_reads;
 	export_vars.innodb_data_writes = os_n_file_writes;
 	export_vars.innodb_data_written = srv_data_written;
+	export_vars.innodb_dict_tables= (dict_sys ? UT_LIST_GET_LEN(dict_sys->table_LRU) : 0);
 	export_vars.innodb_buffer_pool_read_requests = stat.n_page_gets;
 	export_vars.innodb_buffer_pool_write_requests
 		= srv_buf_pool_write_requests;