merge

innobase/btr/btr0btr.c

@@ -2347,6 +2347,8 @@ btr_validate_level(
 	
 	mtr_start(&mtr);
 
+	mtr_x_lock(dict_tree_get_lock(tree), &mtr);
+	
 	page = btr_root_get(tree, &mtr);
 
 	space = buf_frame_get_space_id(page);

innobase/btr/btr0cur.c

@@ -256,7 +256,8 @@ btr_cur_search_to_nth_level(
 #ifdef UNIV_SEARCH_PERF_STAT
 	info->n_searches++;
 #endif	
-	if (latch_mode <= BTR_MODIFY_LEAF && info->last_hash_succ
+	if (btr_search_latch.writer != RW_LOCK_NOT_LOCKED
+		&& latch_mode <= BTR_MODIFY_LEAF && info->last_hash_succ
 		&& !estimate
 	        && btr_search_guess_on_hash(index, info, tuple, mode,
 						latch_mode, cursor,
@@ -344,9 +345,7 @@ btr_cur_search_to_nth_level(
 retry_page_get:		
 		page = buf_page_get_gen(space, page_no, rw_latch, guess,
 					buf_mode,
-#ifdef UNIV_SYNC_DEBUG
 					IB__FILE__, __LINE__,
-#endif
 					mtr);
 
 		if (page == NULL) {
@@ -515,9 +514,7 @@ btr_cur_open_at_index_side(
 	for (;;) {
 		page = buf_page_get_gen(space, page_no, RW_NO_LATCH, NULL,
 					BUF_GET,
-#ifdef UNIV_SYNC_DEBUG
 					IB__FILE__, __LINE__,
-#endif
 					mtr);
 		ut_ad(0 == ut_dulint_cmp(tree->id,
 						btr_page_get_index_id(page)));
@@ -604,9 +601,7 @@ btr_cur_open_at_rnd_pos(
 	for (;;) {
 		page = buf_page_get_gen(space, page_no, RW_NO_LATCH, NULL,
 					BUF_GET,
-#ifdef UNIV_SYNC_DEBUG
 					IB__FILE__, __LINE__,
-#endif
 					mtr);
 		ut_ad(0 == ut_dulint_cmp(tree->id,
 						btr_page_get_index_id(page)));
@@ -1222,6 +1217,57 @@ btr_cur_parse_update_in_place(
 	return(ptr);
 }
 
+/*****************************************************************
+Updates a secondary index record when the update causes no size
+changes in its fields. The only case when this function is currently
+called is that in a char field characters change to others which
+are identified in the collation order. */
+
+ulint
+btr_cur_update_sec_rec_in_place(
+/*============================*/
+				/* out: DB_SUCCESS or error number */
+	btr_cur_t*	cursor,	/* in: cursor on the record to update;
+				cursor stays valid and positioned on the
+				same record */
+	upd_t*		update,	/* in: update vector */
+	que_thr_t*	thr,	/* in: query thread */
+	mtr_t*		mtr)	/* in: mtr */
+{
+	dict_index_t*	index 		= cursor->index;
+	dict_index_t*	clust_index;
+	ulint		err;
+	rec_t*		rec;
+	dulint		roll_ptr	= ut_dulint_zero;
+	trx_t*		trx		= thr_get_trx(thr);
+
+	/* Only secondary index records are updated using this function */
+	ut_ad(0 == (index->type & DICT_CLUSTERED));
+
+	rec = btr_cur_get_rec(cursor);
+	
+	err = lock_sec_rec_modify_check_and_lock(0, rec, index, thr);
+
+	if (err != DB_SUCCESS) {
+
+		return(err);
+	}
+
+	/* Remove possible hash index pointer to this record */
+	btr_search_update_hash_on_delete(cursor);
+
+	row_upd_rec_in_place(rec, update);
+
+	clust_index = dict_table_get_first_index(index->table);
+
+	/* Note that roll_ptr is really just a dummy value since
+	a secondary index record does not contain any sys columns */
+
+	btr_cur_update_in_place_log(BTR_KEEP_SYS_FLAG, rec, clust_index,
+						update, trx, roll_ptr, mtr);
+	return(DB_SUCCESS);
+}
+
 /*****************************************************************
 Updates a record when the update causes no size changes in its fields. */
 
@@ -1248,7 +1294,7 @@ btr_cur_update_in_place(
 	ibool		was_delete_marked;
 
 	/* Only clustered index records are updated using this function */
-	ut_ad((cursor->index)->type & DICT_CLUSTERED);
+	ut_ad(cursor->index->type & DICT_CLUSTERED);
 
 	rec = btr_cur_get_rec(cursor);
 	index = cursor->index;
@@ -2477,27 +2523,33 @@ btr_estimate_n_rows_in_range(
 }
 
 /***********************************************************************
-Estimates the number of different key values in a given index. */
+Estimates the number of different key values in a given index, for
+each n-column prefix of the index where n <= dict_index_get_n_unique(index).
+The estimates are stored in the array index->stat_n_diff_key_vals. */
 
-ulint
+void
 btr_estimate_number_of_different_key_vals(
 /*======================================*/
-				/* out: estimated number of key values */
 	dict_index_t*	index)	/* in: index */
 {
 	btr_cur_t	cursor;
 	page_t*		page;
 	rec_t*		rec;
-	ulint		total_n_recs		= 0;
-	ulint		n_diff_in_page;
-	ulint		n_diff			= 0;
+	ulint		n_cols;
 	ulint		matched_fields;
 	ulint		matched_bytes;
+	ulint*		n_diff;
+	ulint		not_empty_flag	= 0;
 	ulint		i;
+	ulint		j;
 	mtr_t		mtr;
 
-	if (index->type & DICT_UNIQUE) {
-		return(index->table->stat_n_rows);
+	n_cols = dict_index_get_n_unique(index);
+
+	n_diff = mem_alloc((n_cols + 1) * sizeof(ib_longlong));
+
+	for (j = 0; j <= n_cols; j++) {
+		n_diff[j] = 0;
 	}
 
 	/* We sample some pages in the index to get an estimate */
@@ -2507,17 +2559,19 @@ btr_estimate_number_of_different_key_vals(
 
 		btr_cur_open_at_rnd_pos(index, BTR_SEARCH_LEAF, &cursor, &mtr);
 		
-		/* Count the number of different key values minus one on this
-		index page: we subtract one because otherwise our algorithm
-		would give a wrong estimate for an index where there is
-		just one key value */
+		/* Count the number of different key values minus one
+		for each prefix of the key on this index page: we subtract
+		one because otherwise our algorithm would give a wrong
+		estimate for an index where there is just one key value */
 
 		page = btr_cur_get_page(&cursor);
 
 		rec = page_get_infimum_rec(page);
 		rec = page_rec_get_next(rec);
 
-		n_diff_in_page = 0;
+		if (rec != page_get_supremum_rec(page)) {
+			not_empty_flag = 1;
+		}
 		
 		while (rec != page_get_supremum_rec(page)
 		       && page_rec_get_next(rec)
@@ -2528,30 +2582,30 @@ btr_estimate_number_of_different_key_vals(
 			cmp_rec_rec_with_match(rec, page_rec_get_next(rec),
 						index, &matched_fields,
 						&matched_bytes);
-			if (matched_fields <
-				dict_index_get_n_ordering_defined_by_user(
-								index)) {
-				n_diff_in_page++;
+
+			for (j = matched_fields + 1; j <= n_cols; j++) {
+				n_diff[j]++;
 			}
 					
 			rec = page_rec_get_next(rec);
 		}
 		
-		n_diff += n_diff_in_page;
-		
-		total_n_recs += page_get_n_recs(page);
-		
 		mtr_commit(&mtr);
 	}
 
-	if (n_diff == 0) {
-		/* We play safe and assume that there are just two different
-		key values in the index */
+	/* If we saw k borders between different key values on
+	BTR_KEY_VAL_ESTIMATE_N_PAGES leaf pages, we can estimate how many
+	there will be in index->stat_n_leaf_pages */
 	
-		return(2);
+	for (j = 0; j <= n_cols; j++) {
+		index->stat_n_diff_key_vals[j] =
+				(n_diff[j] * index->stat_n_leaf_pages
+				 + BTR_KEY_VAL_ESTIMATE_N_PAGES - 1
+				 + not_empty_flag)
+		                	/ BTR_KEY_VAL_ESTIMATE_N_PAGES;
 	}
 	
-	return(index->table->stat_n_rows / (total_n_recs / n_diff));
+	mem_free(n_diff);
 }
 
 /*================== EXTERNAL STORAGE OF BIG FIELDS ===================*/

innobase/btr/btr0pcur.c

@@ -62,8 +62,10 @@ btr_pcur_free_for_mysql(
 /******************************************************************
 The position of the cursor is stored by taking an initial segment of the
 record the cursor is positioned on, before, or after, and copying it to the
-cursor data structure. NOTE that the page where the cursor is positioned
-must not be empty! */
+cursor data structure, or just setting a flag if the cursor id before the
+first in an EMPTY tree, or after the last in an EMPTY tree. NOTE that the
+page where the cursor is positioned must not be empty if the index tree is
+not totally empty! */
 
 void
 btr_pcur_store_position(
@@ -93,9 +95,21 @@ btr_pcur_store_position(
 	ut_a(cursor->latch_mode != BTR_NO_LATCHES);
 
 	if (page_get_n_recs(page) == 0) {
+		/* It must be an empty index tree */
 
-		/* Cannot store position! */
-		btr_pcur_close(cursor);
+		ut_a(btr_page_get_next(page, mtr) == FIL_NULL
+		     && btr_page_get_prev(page, mtr) == FIL_NULL);
+
+		if (rec == page_get_supremum_rec(page)) {
+
+			cursor->rel_pos = BTR_PCUR_AFTER_LAST_IN_TREE;
+			cursor->old_stored = BTR_PCUR_OLD_STORED;
+
+			return;
+		}
+
+		cursor->rel_pos = BTR_PCUR_BEFORE_FIRST_IN_TREE;
+		cursor->old_stored = BTR_PCUR_OLD_STORED;
 
 		return;
 	} 
@@ -140,13 +154,15 @@ btr_pcur_copy_stored_position(
 
 	ut_memcpy((byte*)pcur_receive, (byte*)pcur_donate, sizeof(btr_pcur_t));
 
+	if (pcur_donate->old_rec_buf) {
+
 		pcur_receive->old_rec_buf = mem_alloc(pcur_donate->buf_size);
 	
 		ut_memcpy(pcur_receive->old_rec_buf, pcur_donate->old_rec_buf,
 						pcur_donate->buf_size);
 		pcur_receive->old_rec = pcur_receive->old_rec_buf
 			+ (pcur_donate->old_rec - pcur_donate->old_rec_buf);
-	
+	}	
 }
 
 /******************************************************************
@@ -158,7 +174,9 @@ to the last record LESS OR EQUAL to the stored record;
 the last record LESS than the user record which was the successor of the page
 infimum;
 (3) cursor was positioned on the page supremum: restores to the first record
-GREATER than the user record which was the predecessor of the supremum. */
+GREATER than the user record which was the predecessor of the supremum.
+(4) cursor was positioned before the first or after the last in an empty tree:
+restores to before first or after the last in the tree. */
 
 ibool
 btr_pcur_restore_position(
@@ -177,17 +195,33 @@ btr_pcur_restore_position(
 	dtuple_t*	tuple;
 	ulint		mode;
 	ulint		old_mode;
+	ibool		from_left;
 	mem_heap_t*	heap;
 
-	ut_a((cursor->pos_state == BTR_PCUR_WAS_POSITIONED)
-			|| (cursor->pos_state == BTR_PCUR_IS_POSITIONED));
+	ut_a(cursor->pos_state == BTR_PCUR_WAS_POSITIONED
+			|| cursor->pos_state == BTR_PCUR_IS_POSITIONED);
 	ut_a(cursor->old_stored == BTR_PCUR_OLD_STORED);
+
+	if (cursor->rel_pos == BTR_PCUR_AFTER_LAST_IN_TREE
+	    || cursor->rel_pos == BTR_PCUR_BEFORE_FIRST_IN_TREE) {
+
+	    	if (cursor->rel_pos == BTR_PCUR_BEFORE_FIRST_IN_TREE) {
+	    		from_left = TRUE;
+	    	} else {
+	    		from_left = FALSE;
+	    	}
+
+		btr_cur_open_at_index_side(from_left,
+			btr_pcur_get_btr_cur(cursor)->index, latch_mode,
+					btr_pcur_get_btr_cur(cursor), mtr);
+		return(FALSE);
+	}
+	
 	ut_a(cursor->old_rec);
 
 	page = btr_cur_get_page(btr_pcur_get_btr_cur(cursor));
 
-	if ((latch_mode == BTR_SEARCH_LEAF)
-					|| (latch_mode == BTR_MODIFY_LEAF)) {
+	if (latch_mode == BTR_SEARCH_LEAF || latch_mode == BTR_MODIFY_LEAF) {
 		/* Try optimistic restoration */
 	    
 		if (buf_page_optimistic_get(latch_mode, page,
@@ -242,16 +276,15 @@ btr_pcur_restore_position(
 	/* Restore the old search mode */
 	cursor->search_mode = old_mode;
 
-	if ((cursor->rel_pos == BTR_PCUR_ON)
+	if (cursor->rel_pos == BTR_PCUR_ON
 	    && btr_pcur_is_on_user_rec(cursor, mtr)
-		&& (0 == cmp_dtuple_rec(tuple, btr_pcur_get_rec(cursor)))) {
+	    && 0 == cmp_dtuple_rec(tuple, btr_pcur_get_rec(cursor))) {
 
 	        /* We have to store the NEW value for the modify clock, since
 	        the cursor can now be on a different page! */
 
 	        cursor->modify_clock = buf_frame_get_modify_clock(
-				    buf_frame_align(
-						    btr_pcur_get_rec(cursor)));
+				    buf_frame_align(btr_pcur_get_rec(cursor)));
 		mem_heap_free(heap);
 
 		return(TRUE);
@@ -366,6 +399,7 @@ btr_pcur_move_backward_from_page(
 
 		latch_mode2 = BTR_MODIFY_PREV;
 	} else {
+		latch_mode2 = 0; /* To eliminate compiler warning */
 		ut_error;
 	}
 

innobase/btr/btr0sea.c

@@ -680,9 +680,7 @@ btr_search_guess_on_hash(
 
 		success = buf_page_get_known_nowait(latch_mode, page,
 						BUF_MAKE_YOUNG,
-#ifdef UNIV_SYNC_DEBUG
 						IB__FILE__, __LINE__,
-#endif
 						mtr);
 
 		rw_lock_s_unlock(&btr_search_latch);

innobase/buf/buf0lru.c

@@ -551,6 +551,10 @@ buf_LRU_block_free_non_file_page(
 
 	block->state = BUF_BLOCK_NOT_USED;
 
+#ifdef UNIV_DEBUG	
+	/* Wipe contents of page to reveal possible stale pointers to it */
+	memset(block->frame, '\0', UNIV_PAGE_SIZE);
+#endif	
 	UT_LIST_ADD_FIRST(free, buf_pool->free, block);
 }
 

innobase/data/data0data.c

@@ -14,6 +14,7 @@ Created 5/30/1994 Heikki Tuuri
 
 #include "ut0rnd.h"
 #include "rem0rec.h"
+#include "rem0cmp.h"
 #include "page0page.h"
 #include "dict0dict.h"
 #include "btr0cur.h"
@@ -63,6 +64,53 @@ dtuple_get_nth_field_noninline(
 	return(dtuple_get_nth_field(tuple, n));
 }
 
+/****************************************************************
+Returns TRUE if lengths of two dtuples are equal and respective data fields
+in them are equal when compared with collation in char fields (not as binary
+strings). */
+
+ibool
+dtuple_datas_are_ordering_equal(
+/*============================*/
+				/* out: TRUE if length and fieds are equal
+				when compared with cmp_data_data:
+				NOTE: in character type fields some letters
+				are identified with others! (collation) */
+	dtuple_t*	tuple1,	/* in: tuple 1 */
+	dtuple_t*	tuple2)	/* in: tuple 2 */
+{
+	dfield_t*	field1;
+	dfield_t*	field2;
+	ulint		n_fields;
+	ulint		i;
+
+	ut_ad(tuple1 && tuple2);
+	ut_ad(tuple1->magic_n = DATA_TUPLE_MAGIC_N);
+	ut_ad(tuple2->magic_n = DATA_TUPLE_MAGIC_N);
+	ut_ad(dtuple_check_typed(tuple1));
+	ut_ad(dtuple_check_typed(tuple2));
+
+	n_fields = dtuple_get_n_fields(tuple1);
+
+	if (n_fields != dtuple_get_n_fields(tuple2)) {
+
+		return(FALSE);
+	}
+	
+	for (i = 0; i < n_fields; i++) {
+
+		field1 = dtuple_get_nth_field(tuple1, i);
+		field2 = dtuple_get_nth_field(tuple2, i);
+
+		if (0 != cmp_dfield_dfield(field1, field2)) {
+		
+			return(FALSE);
+		}			
+	}
+	
+	return(TRUE);
+}
+
 /*************************************************************************
 Creates a dtuple for use in MySQL. */
 
@@ -408,7 +456,7 @@ dtuple_convert_big_rec(
 	ulint		size;
 	ulint		n_fields;
 	ulint		longest;
-	ulint		longest_i;
+	ulint		longest_i		= ULINT_MAX;
 	ibool		is_externally_stored;
 	ulint		i;
 	ulint		j;

innobase/data/data0type.c

@@ -28,7 +28,6 @@ dtype_validate(
 	ut_a((type->mtype >= DATA_VARCHAR) && (type->mtype <= DATA_MYSQL));
 	
 	if (type->mtype == DATA_SYS) {
-		ut_a(type->prtype >= DATA_ROW_ID);
 		ut_a(type->prtype <= DATA_MIX_ID);
 	}
 
@@ -45,11 +44,10 @@ dtype_print(
 {
 	ulint	mtype;
 	ulint	prtype;
+	ulint	len;
 	
 	ut_a(type);
 
-	printf("DATA TYPE: ");
-	
 	mtype = type->mtype;
 	prtype = type->prtype;
 	if (mtype == DATA_VARCHAR) {
@@ -65,17 +63,24 @@ dtype_print(
 	} else if (mtype == DATA_SYS) {
 		printf("DATA_SYS");
 	} else {
-		printf("unknown type %lu", mtype);
+		printf("type %lu", mtype);
 	}
 
+	len = type->len;
+	
 	if ((type->mtype == DATA_SYS)
 	   || (type->mtype == DATA_VARCHAR)
 	   || (type->mtype == DATA_CHAR)) {
 		printf(" ");
 		if (prtype == DATA_ROW_ID) {
 			printf("DATA_ROW_ID");
+			len = DATA_ROW_ID_LEN;
 		} else if (prtype == DATA_ROLL_PTR) {
 			printf("DATA_ROLL_PTR");
+			len = DATA_ROLL_PTR_LEN;
+		} else if (prtype == DATA_TRX_ID) {
+			printf("DATA_TRX_ID");
+			len = DATA_TRX_ID_LEN;
 		} else if (prtype == DATA_MIX_ID) {
 			printf("DATA_MIX_ID");
 		} else if (prtype == DATA_ENGLISH) {
@@ -83,9 +88,9 @@ dtype_print(
 		} else if (prtype == DATA_FINNISH) {
 			printf("DATA_FINNISH");
 		} else {
-			printf("unknown prtype %lu", mtype);
+			printf("prtype %lu", mtype);
 		}
 	}
 
-	printf("; len %lu prec %lu\n", type->len, type->prec);
+	printf(" len %lu prec %lu", len, type->prec);
 }

innobase/dict/dict0crea.c

@@ -17,9 +17,13 @@ Created 1/8/1996 Heikki Tuuri
 #include "page0page.h"
 #include "mach0data.h"
 #include "dict0boot.h"
+#include "dict0dict.h"
 #include "que0que.h"
 #include "row0ins.h"
+#include "row0mysql.h"
 #include "pars0pars.h"
+#include "trx0roll.h"
+#include "usr0sess.h"
 
 /*********************************************************************
 Based on a table object, this function builds the entry to be inserted
@@ -1019,3 +1023,228 @@ function_exit:
 
 	return(thr);
 } 
+
+/********************************************************************
+Creates the foreign key constraints system tables inside InnoDB
+at database creation or database start if they are not found or are
+not of the right form. */
+
+ulint
+dict_create_or_check_foreign_constraint_tables(void)
+/*================================================*/
+				/* out: DB_SUCCESS or error code */
+{
+	dict_table_t*	table1;
+	dict_table_t*	table2;
+	que_thr_t*	thr;
+	que_t*		graph;
+	ulint		error;
+	trx_t*		trx;
+	char*		str;
+
+	mutex_enter(&(dict_sys->mutex));
+
+	table1 = dict_table_get_low("SYS_FOREIGN");
+	table2 = dict_table_get_low("SYS_FOREIGN_COLS");
+	
+	if (table1 && table2
+            && UT_LIST_GET_LEN(table1->indexes) == 3
+            && UT_LIST_GET_LEN(table2->indexes) == 1) {
+
+            	/* Foreign constraint system tables have already been
+            	created, and they are ok */
+
+		mutex_exit(&(dict_sys->mutex));
+
+            	return(DB_SUCCESS);
+        }
+
+	trx = trx_allocate_for_mysql();
+	
+	trx->op_info = "creating foreign key sys tables";
+
+	if (table1) {
+		fprintf(stderr,
+		"InnoDB: dropping incompletely created SYS_FOREIGN table\n");
+		row_drop_table_for_mysql("SYS_FOREIGN", trx, TRUE);
+	}
+
+	if (table2) {
+		fprintf(stderr,
+	"InnoDB: dropping incompletely created SYS_FOREIGN_COLS table\n");
+		row_drop_table_for_mysql("SYS_FOREIGN_COLS", trx, TRUE);
+	}
+
+	fprintf(stderr,
+		"InnoDB: creating foreign key constraint system tables\n");
+
+	/* NOTE: in dict_load_foreigns we use the fact that
+	there are 2 secondary indexes on SYS_FOREIGN, and they
+	are defined just like below */
+	
+	str =
+	"PROCEDURE CREATE_FOREIGN_SYS_TABLES_PROC () IS\n"
+	"BEGIN\n"
+	"CREATE TABLE\n"
+	"SYS_FOREIGN(ID CHAR, FOR_NAME CHAR, REF_NAME CHAR, N_COLS INT);\n"
+	"CREATE UNIQUE CLUSTERED INDEX ID_IND ON SYS_FOREIGN (ID);\n"
+	"CREATE INDEX FOR_IND ON SYS_FOREIGN (FOR_NAME);\n"
+	"CREATE INDEX REF_IND ON SYS_FOREIGN (REF_NAME);\n"
+	"CREATE TABLE\n"
+  "SYS_FOREIGN_COLS(ID CHAR, POS INT, FOR_COL_NAME CHAR, REF_COL_NAME CHAR);\n"
+	"CREATE UNIQUE CLUSTERED INDEX ID_IND ON SYS_FOREIGN_COLS (ID, POS);\n"
+	"COMMIT WORK;\n"
+	"END;\n";
+
+	graph = pars_sql(str);
+
+	ut_a(graph);
+
+	graph->trx = trx;
+	trx->graph = NULL;
+
+	graph->fork_type = QUE_FORK_MYSQL_INTERFACE;
+
+	ut_a(thr = que_fork_start_command(graph, SESS_COMM_EXECUTE, 0));
+
+	que_run_threads(thr);
+
+	error = trx->error_state;
+
+	if (error != DB_SUCCESS) {
+		ut_a(error == DB_OUT_OF_FILE_SPACE);
+
+		fprintf(stderr, "InnoDB: creation failed\n");
+		fprintf(stderr, "InnoDB: tablespace is full\n");
+		fprintf(stderr,
+		"InnoDB: dropping incompletely created SYS_FOREIGN tables\n");
+
+		row_drop_table_for_mysql("SYS_FOREIGN", trx, TRUE);
+		row_drop_table_for_mysql("SYS_FOREIGN_COLS", trx, TRUE);
+
+		error = DB_MUST_GET_MORE_FILE_SPACE;
+	}
+
+	que_graph_free(graph);
+	
+	trx->op_info = "";
+
+  	trx_free_for_mysql(trx);
+
+  	if (error == DB_SUCCESS) {
+		fprintf(stderr,
+		"InnoDB: foreign key constraint system tables created\n");
+	}
+
+	mutex_exit(&(dict_sys->mutex));
+
+	return(error);
+}
+
+/************************************************************************
+Adds foreign key definitions to data dictionary tables in the database. */
+
+ulint
+dict_create_add_foreigns_to_dictionary(
+/*===================================*/
+				/* out: error code or DB_SUCCESS */
+	dict_table_t*	table,	/* in: table */
+	trx_t*		trx)	/* in: transaction */
+{
+	dict_foreign_t*	foreign;
+	que_thr_t*	thr;
+	que_t*		graph;
+	dulint		id;	
+	ulint		len;
+	ulint		error;
+	ulint		i;
+	char		buf2[50];
+	char		buf[10000];
+
+	ut_ad(mutex_own(&(dict_sys->mutex)));	
+
+	if (NULL == dict_table_get_low("SYS_FOREIGN")) {
+		fprintf(stderr,
+     "InnoDB: table SYS_FOREIGN not found from internal data dictionary\n");
+		return(DB_ERROR);
+	}
+
+	foreign = UT_LIST_GET_FIRST(table->foreign_list);
+loop:
+	if (foreign == NULL) {
+
+		return(DB_SUCCESS);
+	}
+
+	/* Build an InnoDB stored procedure which will insert the necessary
+	rows to SYS_FOREIGN and SYS_FOREIGN_COLS */
+
+	len = 0;
+	
+	len += sprintf(buf,
+	"PROCEDURE ADD_FOREIGN_DEFS_PROC () IS\n"
+	"BEGIN\n");
+
+	/* We allocate the new id from the sequence of table id's */
+	id = dict_hdr_get_new_id(DICT_HDR_TABLE_ID);
+
+	sprintf(buf2, "%lu_%lu", ut_dulint_get_high(id),
+						ut_dulint_get_low(id));
+	foreign->id = mem_heap_alloc(foreign->heap, ut_strlen(buf2) + 1);
+	ut_memcpy(foreign->id, buf2, ut_strlen(buf2) + 1);
+	
+	len += sprintf(buf + len,
+	"INSERT INTO SYS_FOREIGN VALUES('%lu_%lu', '%s', '%s', %lu);\n",
+					ut_dulint_get_high(id),
+					ut_dulint_get_low(id),
+					table->name,
+					foreign->referenced_table_name,
+					foreign->n_fields);
+
+	for (i = 0; i < foreign->n_fields; i++) {
+
+		len += sprintf(buf + len,
+	"INSERT INTO SYS_FOREIGN_COLS VALUES('%lu_%lu', %lu, '%s', '%s');\n",
+					ut_dulint_get_high(id),
+					ut_dulint_get_low(id),
+					i,
+					foreign->foreign_col_names[i],
+					foreign->referenced_col_names[i]);
+	}
+
+	len += sprintf(buf + len,"COMMIT WORK;\nEND;\n");
+
+	graph = pars_sql(buf);
+
+	ut_a(graph);
+
+	graph->trx = trx;
+	trx->graph = NULL;
+
+	graph->fork_type = QUE_FORK_MYSQL_INTERFACE;
+
+	ut_a(thr = que_fork_start_command(graph, SESS_COMM_EXECUTE, 0));
+
+	que_run_threads(thr);
+
+	error = trx->error_state;
+
+	que_graph_free(graph);
+
+	if (error != DB_SUCCESS) {
+		ut_a(error == DB_OUT_OF_FILE_SPACE);
+
+		fprintf(stderr, "InnoDB: foreign constraint creation failed\n");
+		fprintf(stderr, "InnoDB: tablespace is full\n");
+
+		trx_general_rollback_for_mysql(trx, FALSE, NULL);
+
+		error = DB_MUST_GET_MORE_FILE_SPACE;
+
+		return(error);
+	}
+	
+	foreign = UT_LIST_GET_NEXT(foreign_list, foreign);
+
+	goto loop;
+}

innobase/dict/dict0mem.c

@@ -18,6 +18,7 @@ Created 1/8/1996 Heikki Tuuri
 #include "dict0dict.h"
 #include "que0que.h"
 #include "pars0pars.h"
+#include "lock0lock.h"
 
 #define	DICT_HEAP_SIZE		100	/* initial memory heap size when
 					creating a table or index object */
@@ -63,7 +64,12 @@ dict_mem_table_create(
 	table->cols = mem_heap_alloc(heap, (n_cols + DATA_N_SYS_COLS)
 							* sizeof(dict_col_t));
 	UT_LIST_INIT(table->indexes);
+
+	table->auto_inc_lock = mem_heap_alloc(heap, lock_get_size());
+
 	UT_LIST_INIT(table->locks);
+	UT_LIST_INIT(table->foreign_list);
+	UT_LIST_INIT(table->referenced_list);
 
 	table->does_not_fit_in_memory = FALSE;
 
@@ -199,12 +205,49 @@ dict_mem_index_create(
 						* sizeof(dict_field_t));
 					/* The '1 +' above prevents allocation
 					of an empty mem block */
+	index->stat_n_diff_key_vals = NULL;
+
 	index->cached = FALSE;
 	index->magic_n = DICT_INDEX_MAGIC_N;
 
 	return(index);
 }
 
+/**************************************************************************
+Creates and initializes a foreign constraint memory object. */
+
+dict_foreign_t*
+dict_mem_foreign_create(void)
+/*=========================*/
+				/* out, own: foreign constraint struct */
+{
+	dict_foreign_t*	foreign;
+	mem_heap_t*	heap;
+
+	heap = mem_heap_create(100);
+
+	foreign = mem_heap_alloc(heap, sizeof(dict_foreign_t));
+
+	foreign->heap = heap;
+
+	foreign->id = NULL;
+
+	foreign->foreign_table_name = NULL;
+	foreign->foreign_table = NULL;
+	foreign->foreign_col_names = NULL;
+
+	foreign->referenced_table_name = NULL;
+	foreign->referenced_table = NULL;
+	foreign->referenced_col_names = NULL;
+
+	foreign->n_fields = 0;
+
+	foreign->foreign_index = NULL;
+	foreign->referenced_index = NULL;
+
+	return(foreign);
+}
+
 /**************************************************************************
 Adds a field definition to an index. NOTE: does not take a copy
 of the column name if the field is a column. The memory occupied

innobase/fil/fil0fil.c

@@ -77,6 +77,9 @@ out of the LRU-list and keep a count of pending operations. When an operation
 completes, we decrement the count and return the file node to the LRU-list if
 the count drops to zero. */
 
+ulint	fil_n_pending_log_flushes		= 0;
+ulint	fil_n_pending_tablespace_flushes	= 0;
+
 /* Null file address */
 fil_addr_t	fil_addr_null = {FIL_NULL, 0};
 
@@ -856,6 +859,15 @@ fil_node_prepare_for_io(
 
 			last_node = UT_LIST_GET_LAST(system->LRU);
 
+			if (last_node == NULL) {
+				fprintf(stderr,
+	"InnoDB: Error: cannot close any file to open another for i/o\n"
+	"InnoDB: Pending i/o's on %lu files exist\n",
+					system->n_open_pending);
+
+				ut_a(0);
+			}
+
 			fil_node_close(last_node, system);
 		}
 
@@ -973,6 +985,7 @@ fil_io(
 	ibool		ret;
 	ulint		is_log;
 	ulint		wake_later;
+	ulint		count;
 	
 	is_log = type & OS_FILE_LOG;
 	type = type & ~OS_FILE_LOG;
@@ -996,7 +1009,7 @@ fil_io(
 #endif
 	if (sync) {
 		mode = OS_AIO_SYNC;
-	} else if ((type == OS_FILE_READ) && !is_log
+	} else if (type == OS_FILE_READ && !is_log
 				&& ibuf_page(space_id, block_offset)) {
 		mode = OS_AIO_IBUF;
 	} else if (is_log) {
@@ -1006,9 +1019,44 @@ fil_io(
 	}
 
 	system = fil_system;
+
+	count = 0;
 loop:
+	count++;
+	
+	/* NOTE that there is a possibility of a hang here:
+	if the read i/o-handler thread needs to complete
+	a read by reading from the insert buffer, it may need to
+	post another read. But if the maximum number of files
+	are already open, it cannot proceed from here! */
+	
 	mutex_enter(&(system->mutex));
 	
+	if (count < 500 && !is_log && !ibuf_inside()
+	    && system->n_open_pending >= (3 * system->max_n_open) / 4) {
+
+	    	/* We are not doing an ibuf operation: leave a
+	    	safety margin of openable files for possible ibuf
+	    	merges needed in page read completion */
+
+		mutex_exit(&(system->mutex));
+
+		/* Wake the i/o-handler threads to make sure pending
+		i/o's are handled and eventually we can open the file */
+		
+		os_aio_simulated_wake_handler_threads();
+
+		os_thread_sleep(100000);
+
+		if (count > 50) {
+			fprintf(stderr,
+		"InnoDB: Warning: waiting for file closes to proceed\n"
+		"InnoDB: round %lu\n", count);
+		}
+
+		goto loop;
+	}
+
 	if (system->n_open_pending == system->max_n_open) {
 
 		/* It is not sure we can open the file if it is closed: wait */
@@ -1018,6 +1066,14 @@ loop:
 
 		mutex_exit(&(system->mutex));
 
+		/* Wake the i/o-handler threads to make sure pending
+		i/o's are handled and eventually we can open the file */
+		
+		os_aio_simulated_wake_handler_threads();
+
+		fprintf(stderr,
+		"InnoDB: Warning: max allowed number of files is open\n");
+
 		os_event_wait(event);
 
 		goto loop;
@@ -1160,6 +1216,7 @@ fil_aio_wait(
 #elif defined(POSIX_ASYNC_IO)
 		ret = os_aio_posix_handle(segment, &fil_node, &message);
 #else
+		ret = 0; /* Eliminate compiler warning */
 		ut_a(0);
 #endif
 	} else {
@@ -1220,6 +1277,12 @@ fil_flush(
 
 			node->is_modified = FALSE;
 			
+			if (space->purpose == FIL_TABLESPACE) {
+				fil_n_pending_tablespace_flushes++;
+			} else {
+				fil_n_pending_log_flushes++;
+			}
+
 			mutex_exit(&(system->mutex));
 
 			/* Note that it is not certain, when we have
@@ -1233,6 +1296,12 @@ fil_flush(
 			os_file_flush(file);
 			
 			mutex_enter(&(system->mutex));
+
+			if (space->purpose == FIL_TABLESPACE) {
+				fil_n_pending_tablespace_flushes--;
+			} else {
+				fil_n_pending_log_flushes--;
+			}
 		}
 
 		node = UT_LIST_GET_NEXT(chain, node);
@@ -1377,7 +1446,7 @@ fil_page_set_type(
 	ulint	type)	/* in: type */
 {
 	ut_ad(page);
-	ut_ad((type == FIL_PAGE_INDEX) || (type == FIL_PAGE_INDEX));
+	ut_ad((type == FIL_PAGE_INDEX) || (type == FIL_PAGE_UNDO_LOG));
 
 	mach_write_to_2(page + FIL_PAGE_TYPE, type);
 }	

innobase/ibuf/ibuf0ibuf.c

@@ -1013,7 +1013,7 @@ ibuf_rec_get_volume(
 	ulint	i;
 
 	ut_ad(ibuf_inside());
-	ut_ad(rec_get_n_fields(rec) > 2);
+	ut_ad(rec_get_n_fields(ibuf_rec) > 2);
 	
 	n_fields = rec_get_n_fields(ibuf_rec) - 2;
 
@@ -1624,13 +1624,14 @@ ibuf_get_merge_page_nos(
 
 /*************************************************************************
 Contracts insert buffer trees by reading pages to the buffer pool. */
-
+static
 ulint
-ibuf_contract(
-/*==========*/
+ibuf_contract_ext(
+/*==============*/
 			/* out: a lower limit for the combined size in bytes
 			of entries which will be merged from ibuf trees to the
 			pages read, 0 if ibuf is empty */
+	ulint*	n_pages,/* out: number of pages to which merged */
 	ibool	sync)	/* in: TRUE if the caller wants to wait for the
 			issued read with the highest tablespace address
 			to complete */
@@ -1644,6 +1645,8 @@ ibuf_contract(
 	ulint		n_stored;
 	ulint		sum_sizes;
 	mtr_t		mtr;
+
+	*n_pages = 0;
 loop:
 	ut_ad(!ibuf_inside());
 
@@ -1730,9 +1733,64 @@ loop:
 
 	buf_read_ibuf_merge_pages(sync, space, page_nos, n_stored);
 
+	*n_pages = n_stored;
+	
 	return(sum_sizes + 1);
 }
 
+/*************************************************************************
+Contracts insert buffer trees by reading pages to the buffer pool. */
+
+ulint
+ibuf_contract(
+/*==========*/
+			/* out: a lower limit for the combined size in bytes
+			of entries which will be merged from ibuf trees to the
+			pages read, 0 if ibuf is empty */
+	ibool	sync)	/* in: TRUE if the caller wants to wait for the
+			issued read with the highest tablespace address
+			to complete */
+{
+	ulint	n_pages;
+
+	return(ibuf_contract_ext(&n_pages, sync));
+}
+
+/*************************************************************************
+Contracts insert buffer trees by reading pages to the buffer pool. */
+
+ulint
+ibuf_contract_for_n_pages(
+/*======================*/
+			/* out: a lower limit for the combined size in bytes
+			of entries which will be merged from ibuf trees to the
+			pages read, 0 if ibuf is empty */
+	ibool	sync,	/* in: TRUE if the caller wants to wait for the
+			issued read with the highest tablespace address
+			to complete */
+	ulint	n_pages)/* in: try to read at least this many pages to
+			the buffer pool and merge the ibuf contents to
+			them */
+{
+	ulint	sum_bytes	= 0;
+	ulint	sum_pages 	= 0;
+	ulint	n_bytes;
+	ulint	n_pag2;
+	
+	while (sum_pages < n_pages) {
+		n_bytes = ibuf_contract_ext(&n_pag2, sync);
+		
+		if (n_bytes == 0) {
+			return(sum_bytes);
+		}
+
+		sum_bytes += n_bytes;
+		sum_pages += n_pag2;
+	}
+
+	return(sum_bytes);
+}
+
 /*************************************************************************
 Contract insert buffer trees after insert if they are too big. */
 UNIV_INLINE
@@ -2253,8 +2311,6 @@ ibuf_insert_to_index_page(
 	if (low_match == dtuple_get_n_fields(entry)) {
 		rec = page_cur_get_rec(&page_cur);
 		
-		ut_ad(rec_get_deleted_flag(rec));
-		
 		btr_cur_del_unmark_for_ibuf(rec, mtr);
 	} else {
 		rec = page_cur_tuple_insert(&page_cur, entry, mtr);
@@ -2306,6 +2362,8 @@ ibuf_delete_rec(
 				should belong */
 	btr_pcur_t*	pcur,	/* in: pcur positioned on the record to
 				delete, having latch mode BTR_MODIFY_LEAF */
+	dtuple_t*	search_tuple,
+				/* in: search tuple for entries of page_no */
 	mtr_t*		mtr)	/* in: mtr */
 {
 	ibool		success;
@@ -2336,7 +2394,28 @@ ibuf_delete_rec(
 
 	mtr_start(mtr);
 	
-	ut_a(btr_pcur_restore_position(BTR_MODIFY_TREE, pcur, mtr));
+	success = btr_pcur_restore_position(BTR_MODIFY_TREE, pcur, mtr);
+
+	if (!success) {
+		fprintf(stderr,
+		"InnoDB: ERROR: Send the output to heikki.tuuri@innodb.com\n");
+		fprintf(stderr, "InnoDB: ibuf cursor restoration fails!\n");
+		fprintf(stderr, "InnoDB: ibuf record inserted to page %lu\n",
+								page_no);
+		rec_print(btr_pcur_get_rec(pcur));
+		rec_print(pcur->old_rec);
+		dtuple_print(search_tuple);
+
+		rec_print(page_rec_get_next(btr_pcur_get_rec(pcur)));
+
+		mtr_commit(mtr);
+
+		fprintf(stderr, "InnoDB: Validating insert buffer tree:\n");
+		ut_a(btr_validate_tree(ibuf_data->index->tree));		
+		fprintf(stderr, "InnoDB: Ibuf tree ok\n");
+	}
+	
+	ut_a(success);
 
 	root = ibuf_tree_root_get(ibuf_data, space, mtr);
 
@@ -2393,7 +2472,10 @@ ibuf_merge_or_delete_for_page(
 	dulint		max_trx_id;
 	mtr_t		mtr;
 
-	/* TODO: get MySQL type info to use in ibuf_insert_to_index_page */
+	if (srv_force_recovery >= SRV_FORCE_NO_IBUF_MERGE) {
+
+		return;
+	}
 	
 #ifdef UNIV_LOG_DEBUG
 	if (space % 2 != 0) {
@@ -2451,11 +2533,8 @@ loop:
 	if (page) {
 		success = buf_page_get_known_nowait(RW_X_LATCH, page,
 					BUF_KEEP_OLD,
-#ifdef UNIV_SYNC_DEBUG
 					IB__FILE__, __LINE__,
-#endif
 					&mtr);
-
 		ut_a(success);
 
 		buf_page_dbg_add_level(page, SYNC_TREE_NODE);
@@ -2508,13 +2587,13 @@ loop:
 					/ IBUF_PAGE_SIZE_PER_FREE_SPACE);
 #endif
 			ibuf_insert_to_index_page(entry, page, &mtr);
+		}
 
 		n_inserts++;
-		}
 		
 		/* Delete the record from ibuf */
-		closed = ibuf_delete_rec(space, page_no, &pcur, &mtr);
-
+		closed = ibuf_delete_rec(space, page_no, &pcur, search_tuple,
+									&mtr);
 		if (closed) {
 			/* Deletion was pessimistic and mtr was committed:
 			we start from the beginning again */
@@ -2524,6 +2603,7 @@ loop:
 
 		if (btr_pcur_is_after_last_on_page(&pcur, &mtr)) {
 			mtr_commit(&mtr);
+ 			btr_pcur_close(&pcur);
 
 			goto loop;
 		}
@@ -2619,8 +2699,6 @@ ibuf_print(void)
 #endif
 	mutex_enter(&ibuf_mutex);
 
-	printf("Ibuf size %lu max size %lu\n", ibuf->size, ibuf->max_size);
-
 	data = UT_LIST_GET_FIRST(ibuf->data_list);
 
 	while (data) {

innobase/include/btr0cur.h

@@ -188,6 +188,22 @@ btr_cur_pessimistic_insert(
 	que_thr_t*	thr,	/* in: query thread or NULL */
 	mtr_t*		mtr);	/* in: mtr */
 /*****************************************************************
+Updates a secondary index record when the update causes no size
+changes in its fields. The only case when this function is currently
+called is that in a char field characters change to others which
+are identified in the collation order. */
+
+ulint
+btr_cur_update_sec_rec_in_place(
+/*============================*/
+				/* out: DB_SUCCESS or error number */
+	btr_cur_t*	cursor,	/* in: cursor on the record to update;
+				cursor stays valid and positioned on the
+				same record */
+	upd_t*		update,	/* in: update vector */
+	que_thr_t*	thr,	/* in: query thread */
+	mtr_t*		mtr);	/* in: mtr */
+/*****************************************************************
 Updates a record when the update causes no size changes in its fields. */
 
 ulint
@@ -411,12 +427,13 @@ btr_estimate_n_rows_in_range(
 	dtuple_t*	tuple2,	/* in: range end, may also be empty tuple */
 	ulint		mode2);	/* in: search mode for range end */
 /***********************************************************************
-Estimates the number of different key values in a given index. */
+Estimates the number of different key values in a given index, for
+each n-column prefix of the index where n <= dict_index_get_n_unique(index).
+The estimates are stored in the array index->stat_n_diff_key_vals. */
 
-ulint
+void
 btr_estimate_number_of_different_key_vals(
 /*======================================*/
-				/* out: estimated number of key values */
 	dict_index_t*	index);	/* in: index */
 /***********************************************************************
 Marks not updated extern fields as not-owned by this record. The ownership

innobase/include/btr0pcur.h

@@ -22,6 +22,12 @@ Created 2/23/1996 Heikki Tuuri
 #define BTR_PCUR_ON			1
 #define BTR_PCUR_BEFORE			2
 #define BTR_PCUR_AFTER			3
+/* Note that if the tree is not empty, btr_pcur_store_position does not
+use the following, but only uses the above three alternatives, where the
+position is stored relative to a specific record: this makes implementation
+of a scroll cursor easier */
+#define BTR_PCUR_BEFORE_FIRST_IN_TREE	4	/* in an empty tree */
+#define BTR_PCUR_AFTER_LAST_IN_TREE	5	/* in an empty tree */
 
 /******************************************************************
 Allocates memory for a persistent cursor object and initializes the cursor. */
@@ -170,8 +176,10 @@ btr_pcur_close(
 /******************************************************************
 The position of the cursor is stored by taking an initial segment of the
 record the cursor is positioned on, before, or after, and copying it to the
-cursor data structure. NOTE that the page where the cursor is positioned
-must not be empty! */
+cursor data structure, or just setting a flag if the cursor id before the
+first in an EMPTY tree, or after the last in an EMPTY tree. NOTE that the
+page where the cursor is positioned must not be empty if the index tree is
+not totally empty! */
 
 void
 btr_pcur_store_position(
@@ -179,26 +187,6 @@ btr_pcur_store_position(
 	btr_pcur_t*	cursor, /* in: persistent cursor */
 	mtr_t*		mtr);	/* in: mtr */
 /******************************************************************
-If the latch mode of the cursor is BTR_LEAF_SEARCH or BTR_LEAF_MODIFY,
-releases the page latch and bufferfix reserved by the cursor.
-NOTE! In the case of BTR_LEAF_MODIFY, there should not exist changes
-made by the current mini-transaction to the data protected by the
-cursor latch, as then the latch must not be released until mtr_commit. */
-
-void
-btr_pcur_release_leaf(
-/*==================*/
-	btr_pcur_t*	cursor, /* in: persistent cursor */
-	mtr_t*		mtr);	/* in: mtr */
-/*************************************************************
-Gets the rel_pos field for a cursor whose position has been stored. */
-UNIV_INLINE
-ulint
-btr_pcur_get_rel_pos(
-/*=================*/
-				/* out: BTR_PCUR_ON, ... */
-	btr_pcur_t*	cursor);/* in: persistent cursor */
-/******************************************************************
 Restores the stored position of a persistent cursor bufferfixing the page and
 obtaining the specified latches. If the cursor position was saved when the
 (1) cursor was positioned on a user record: this function restores the position
@@ -207,7 +195,9 @@ to the last record LESS OR EQUAL to the stored record;
 the last record LESS than the user record which was the successor of the page
 infimum;
 (3) cursor was positioned on the page supremum: restores to the first record
-GREATER than the user record which was the predecessor of the supremum. */
+GREATER than the user record which was the predecessor of the supremum.
+(4) cursor was positioned before the first or after the last in an empty tree:
+restores to before first or after the last in the tree. */
 
 ibool
 btr_pcur_restore_position(
@@ -220,6 +210,26 @@ btr_pcur_restore_position(
 	ulint		latch_mode,	/* in: BTR_SEARCH_LEAF, ... */
 	btr_pcur_t*	cursor, 	/* in: detached persistent cursor */
 	mtr_t*		mtr);		/* in: mtr */
+/******************************************************************
+If the latch mode of the cursor is BTR_LEAF_SEARCH or BTR_LEAF_MODIFY,
+releases the page latch and bufferfix reserved by the cursor.
+NOTE! In the case of BTR_LEAF_MODIFY, there should not exist changes
+made by the current mini-transaction to the data protected by the
+cursor latch, as then the latch must not be released until mtr_commit. */
+
+void
+btr_pcur_release_leaf(
+/*==================*/
+	btr_pcur_t*	cursor, /* in: persistent cursor */
+	mtr_t*		mtr);	/* in: mtr */
+/*************************************************************
+Gets the rel_pos field for a cursor whose position has been stored. */
+UNIV_INLINE
+ulint
+btr_pcur_get_rel_pos(
+/*=================*/
+				/* out: BTR_PCUR_ON, ... */
+	btr_pcur_t*	cursor);/* in: persistent cursor */
 /*************************************************************
 Sets the mtr field for a pcur. */
 UNIV_INLINE
@@ -458,7 +468,7 @@ struct btr_pcur_struct{
 	ulint		search_mode;	/* PAGE_CUR_G, ... */
 	/*-----------------------------*/
 	/* NOTE that the following fields may possess dynamically allocated
-	memory, which should be freed if not needed anymore! */
+	memory which should be freed if not needed anymore! */
 
 	mtr_t*		mtr;		/* NULL, or this field may contain
 					a mini-transaction which holds the

innobase/include/btr0pcur.ic

@@ -19,8 +19,8 @@ btr_pcur_get_rel_pos(
 	ut_ad(cursor);
 	ut_ad(cursor->old_rec);
 	ut_ad(cursor->old_stored == BTR_PCUR_OLD_STORED);
-	ut_ad((cursor->pos_state == BTR_PCUR_WAS_POSITIONED)
-			|| (cursor->pos_state == BTR_PCUR_IS_POSITIONED));
+	ut_ad(cursor->pos_state == BTR_PCUR_WAS_POSITIONED
+			|| cursor->pos_state == BTR_PCUR_IS_POSITIONED);
 
 	return(cursor->rel_pos);
 }

innobase/include/btr0sea.h

@@ -262,6 +262,12 @@ index */
 
 #define BTR_SEARCH_ON_HASH_LIMIT	3
 
+/* We do this many searches before trying to keep the search latch over calls
+from MySQL. If we notice someone waiting for the latch, we again set this
+much timeout. This is to reduce contention. */
+
+#define BTR_SEA_TIMEOUT			10000
+
 #ifndef UNIV_NONINL
 #include "btr0sea.ic"
 #endif

innobase/include/buf0buf.h

@@ -116,53 +116,30 @@ buf_frame_copy(
 NOTE! The following macros should be used instead of buf_page_get_gen,
 to improve debugging. Only values RW_S_LATCH and RW_X_LATCH are allowed
 in LA! */
-#ifdef UNIV_SYNC_DEBUG
 #define buf_page_get(SP, OF, LA, MTR)    buf_page_get_gen(\
 				SP, OF, LA, NULL,\
 				BUF_GET, IB__FILE__, __LINE__, MTR)
-#else
-#define buf_page_get(SP, OF, LA, MTR)    buf_page_get_gen(\
-				SP, OF, LA, NULL,\
-				BUF_GET, MTR)
-#endif
 /******************************************************************
 Use these macros to bufferfix a page with no latching. Remember not to
 read the contents of the page unless you know it is safe. Do not modify
 the contents of the page! We have separated this case, because it is
 error-prone programming not to set a latch, and it should be used
 with care. */
-#ifdef UNIV_SYNC_DEBUG
 #define buf_page_get_with_no_latch(SP, OF, MTR)    buf_page_get_gen(\
 				SP, OF, RW_NO_LATCH, NULL,\
 				BUF_GET_NO_LATCH, IB__FILE__, __LINE__, MTR)
-#else
-#define buf_page_get_with_no_latch(SP, OF, MTR)    buf_page_get_gen(\
-				SP, OF, RW_NO_LATCH, NULL,\
-				BUF_GET_NO_LATCH, MTR)
-#endif
 /******************************************************************
 NOTE! The following macros should be used instead of buf_page_get_gen, to
 improve debugging. Only values RW_S_LATCH and RW_X_LATCH are allowed as LA! */
-#ifdef UNIV_SYNC_DEBUG
 #define buf_page_get_nowait(SP, OF, LA, MTR)    buf_page_get_gen(\
 				SP, OF, LA, NULL,\
 				BUF_GET_NOWAIT, IB__FILE__, __LINE__, MTR)
-#else
-#define buf_page_get_nowait(SP, OF, LA, MTR)    buf_page_get_gen(\
-				SP, OF, LA, NULL,\
-				BUF_GET_NOWAIT, MTR)
-#endif
 /******************************************************************
 NOTE! The following macros should be used instead of
 buf_page_optimistic_get_func, to improve debugging. Only values RW_S_LATCH and
 RW_X_LATCH are allowed as LA! */
-#ifdef UNIV_SYNC_DEBUG
 #define buf_page_optimistic_get(LA, G, MC, MTR) buf_page_optimistic_get_func(\
 				LA, G, MC, IB__FILE__, __LINE__, MTR)
-#else
-#define buf_page_optimistic_get(LA, G, MC, MTR) buf_page_optimistic_get_func(\
-				LA, G, MC, MTR)
-#endif
 /************************************************************************
 This is the general function used to get optimistic access to a database
 page. */
@@ -175,10 +152,8 @@ buf_page_optimistic_get_func(
 	buf_frame_t*	guess,	/* in: guessed frame */
 	dulint		modify_clock,/* in: modify clock value if mode is
 				..._GUESS_ON_CLOCK */
-#ifdef UNIV_SYNC_DEBUG
 	char*		file,	/* in: file name */
 	ulint		line,	/* in: line where called */
-#endif
 	mtr_t*		mtr);	/* in: mini-transaction */
 /************************************************************************
 Tries to get the page, but if file io is required, releases all latches
@@ -210,10 +185,8 @@ buf_page_get_known_nowait(
 	ulint		rw_latch,/* in: RW_S_LATCH, RW_X_LATCH */
 	buf_frame_t*	guess,	/* in: the known page frame */
 	ulint		mode,	/* in: BUF_MAKE_YOUNG or BUF_KEEP_OLD */
-#ifdef UNIV_SYNC_DEBUG
 	char*		file,	/* in: file name */
 	ulint		line,	/* in: line where called */
-#endif
 	mtr_t*		mtr);	/* in: mini-transaction */
 /************************************************************************
 This is the general function used to get access to a database page. */
@@ -228,10 +201,8 @@ buf_page_get_gen(
 	buf_frame_t*	guess,	/* in: guessed frame or NULL */
 	ulint		mode,	/* in: BUF_GET, BUF_GET_IF_IN_POOL,
 				BUF_GET_NO_LATCH */
-#ifdef UNIV_SYNC_DEBUG
 	char*		file,	/* in: file name */
 	ulint		line,	/* in: line where called */
-#endif
 	mtr_t*		mtr);	/* in: mini-transaction */
 /************************************************************************
 Initializes a page to the buffer buf_pool. The page is usually not read
@@ -455,6 +426,13 @@ Validates the buffer pool data structure. */
 ibool
 buf_validate(void);
 /*==============*/
+/************************************************************************
+Prints a page to stderr. */
+
+void
+buf_page_print(
+/*===========*/
+	byte*	read_buf);	/* in: a database page */
 /*************************************************************************
 Prints info of the buffer pool data structure. */
 
@@ -462,6 +440,12 @@ void
 buf_print(void);
 /*===========*/
 /*************************************************************************
+Returns the number of pending buf pool ios. */
+
+ulint
+buf_get_n_pending_ios(void);
+/*=======================*/
+/*************************************************************************
 Prints info of the buffer i/o. */
 
 void
@@ -760,6 +744,8 @@ struct buf_pool_struct{
 	byte*		frame_zero;	/* pointer to the first buffer frame:
 					this may differ from frame_mem, because
 					this is aligned by the frame size */
+	byte*		high_end;	/* pointer to the end of the
+					buffer pool */
 	buf_block_t*	blocks;		/* array of buffer control blocks */
 	ulint		max_size;	/* number of control blocks ==
 					maximum pool size in pages */
@@ -767,6 +753,9 @@ struct buf_pool_struct{
 	hash_table_t*	page_hash;	/* hash table of the file pages */
 
 	ulint		n_pend_reads;	/* number of pending read operations */
+
+	time_t		last_printout_time; /* when buf_print was last time
+					called */
 	ulint		n_pages_read;	/* number read operations */
 	ulint		n_pages_written;/* number write operations */
 	ulint		n_pages_created;/* number of pages created in the pool
@@ -782,6 +771,9 @@ struct buf_pool_struct{
 					hit rate */
 	ulint		n_pages_read_old;/* n_pages_read when buf_print was
 					last time called */
+	ulint		n_pages_written_old;/* number write operations */
+	ulint		n_pages_created_old;/* number of pages created in
+					the pool with no read */
 	/* 2. Page flushing algorithm fields */
 
 	UT_LIST_BASE_NODE_T(buf_block_t) flush_list;

innobase/include/buf0buf.ic

@@ -486,11 +486,7 @@ buf_block_buf_fix_inc_debug(
 {
 	ibool	ret;
 	
-	ret = rw_lock_s_lock_func_nowait(&(block->debug_latch)
-#ifdef UNIV_SYNC_DEBUG	
-				,file, line
-#endif
-			    );
+	ret = rw_lock_s_lock_func_nowait(&(block->debug_latch), file, line);
 
 	ut_ad(ret == TRUE);
 
@@ -557,9 +553,7 @@ buf_page_get_release_on_io(
 
 	frame = buf_page_get_gen(space, offset, rw_latch, guess,
 				BUF_GET_IF_IN_POOL,
-#ifdef UNIV_SYNC_DEBUG
 				IB__FILE__, __LINE__,
-#endif
 				mtr);
 	if (frame != NULL) {
 

innobase/include/data0data.h

@@ -116,8 +116,8 @@ dfield_copy(
 Tests if data length and content is equal for two dfields. */
 UNIV_INLINE
 ibool
-dfield_datas_are_equal(
-/*===================*/
+dfield_datas_are_binary_equal(
+/*==========================*/
 				/* out: TRUE if equal */
 	dfield_t*	field1,	/* in: field */
 	dfield_t*	field2);/* in: field */
@@ -125,8 +125,8 @@ dfield_datas_are_equal(
 Tests if dfield data length and content is equal to the given. */
 UNIV_INLINE
 ibool
-dfield_data_is_equal(
-/*=================*/
+dfield_data_is_binary_equal(
+/*========================*/
 				/* out: TRUE if equal */
 	dfield_t*	field,	/* in: field */
 	ulint		len,	/* in: data length or UNIV_SQL_NULL */
@@ -230,14 +230,18 @@ dtuple_get_data_size(
 	dtuple_t*	tuple);	/* in: typed data tuple */
 /****************************************************************
 Returns TRUE if lengths of two dtuples are equal and respective data fields
-in them are equal. */
-UNIV_INLINE
+in them are equal when compared with collation in char fields (not as binary
+strings). */
+
 ibool
-dtuple_datas_are_equal(
-/*===================*/
-				/* out: TRUE if length and datas are equal */
+dtuple_datas_are_ordering_equal(
+/*============================*/
+				/* out: TRUE if length and fieds are equal
+				when compared with cmp_data_data:
+				NOTE: in character type fields some letters
+				are identified with others! (collation) */
 	dtuple_t*	tuple1,	/* in: tuple 1 */
-	dtuple_t*	tuple2);	/* in: tuple 2 */
+	dtuple_t*	tuple2);/* in: tuple 2 */
 /****************************************************************
 Folds a prefix given as the number of fields of a tuple. */
 UNIV_INLINE
@@ -447,7 +451,7 @@ struct dfield_struct{
 
 struct dtuple_struct {
 	ulint		info_bits;	/* info bits of an index record:
-					default is 0; this field is used
+					the default is 0; this field is used
 					if an index record is built from
 					a data tuple */
 	ulint		n_fields;	/* number of fields in dtuple */

innobase/include/data0data.ic

@@ -133,8 +133,8 @@ dfield_copy(
 Tests if data length and content is equal for two dfields. */
 UNIV_INLINE
 ibool
-dfield_datas_are_equal(
-/*===================*/
+dfield_datas_are_binary_equal(
+/*==========================*/
 				/* out: TRUE if equal */
 	dfield_t*	field1,	/* in: field */
 	dfield_t*	field2)	/* in: field */
@@ -157,8 +157,8 @@ dfield_datas_are_equal(
 Tests if dfield data length and content is equal to the given. */
 UNIV_INLINE
 ibool
-dfield_data_is_equal(
-/*=================*/
+dfield_data_is_binary_equal(
+/*========================*/
 				/* out: TRUE if equal */
 	dfield_t*	field,	/* in: field */
 	ulint		len,	/* in: data length or UNIV_SQL_NULL */
@@ -169,8 +169,7 @@ dfield_data_is_equal(
 		return(FALSE);
 	}
 
-	if ((len != UNIV_SQL_NULL)
-	    && (0 != ut_memcmp(field->data, data, len))) {
+	if (len != UNIV_SQL_NULL && 0 != ut_memcmp(field->data, data, len)) {
 	    	
 		return(FALSE);
 	}
@@ -342,65 +341,6 @@ dtuple_get_data_size(
 	return(sum);
 }
 
-/****************************************************************
-Returns TRUE if lengths of two dtuples are equal and respective data fields
-in them are equal. */
-UNIV_INLINE
-ibool
-dtuple_datas_are_equal(
-/*===================*/
-				/* out: TRUE if length and datas are equal */
-	dtuple_t*	tuple1,	/* in: tuple 1 */
-	dtuple_t*	tuple2)	/* in: tuple 2 */
-{
-	dfield_t*	field1;
-	dfield_t*	field2;
-	ulint		n_fields;
-	byte*		data1;
-	byte*		data2;
-	ulint		len1;
-	ulint		len2;
-	ulint		i;
-
-	ut_ad(tuple1 && tuple2);
-	ut_ad(tuple1->magic_n = DATA_TUPLE_MAGIC_N);
-	ut_ad(tuple2->magic_n = DATA_TUPLE_MAGIC_N);
-	ut_ad(dtuple_check_typed(tuple1));
-	ut_ad(dtuple_check_typed(tuple2));
-
-	n_fields = dtuple_get_n_fields(tuple1);
-
-	if (n_fields != dtuple_get_n_fields(tuple2)) {
-
-		return(FALSE);
-	}
-	
-	for (i = 0; i < n_fields; i++) {
-
-		field1 = dtuple_get_nth_field(tuple1, i);
-		data1 = (byte*) dfield_get_data(field1);
-		len1 = dfield_get_len(field1);	
-
-		field2 = dtuple_get_nth_field(tuple2, i);
-		data2 = (byte*) dfield_get_data(field2);
-		len2 = dfield_get_len(field2);
-
-		if (len1 != len2) {
-
-			return(FALSE);
-		}
-
-		if (len1 != UNIV_SQL_NULL) {
-			if (ut_memcmp(data1, data2, len1) != 0) {
-
-				return(FALSE);
-			}
-		}
-	}
-	
-	return(TRUE);
-}
-
 /***********************************************************************
 Sets types of fields binary in a tuple. */
 UNIV_INLINE

innobase/include/data0type.h

@@ -124,17 +124,6 @@ dtype_get_pad_char(
 				/* out: padding character code, or
 				ULINT_UNDEFINED if no padding specified */
 	dtype_t*	type);	/* in: typeumn */
-/*************************************************************************
-Transforms the character code so that it is ordered appropriately
-for the language. */
-UNIV_INLINE
-ulint
-dtype_collate(
-/*==========*/
-				/* out: padding character */
-	dtype_t*	type,	/* in: type */
-	ulint		code);	/* in: character code stored in database
-				record */
 /***************************************************************************
 Returns the size of a fixed size data type, 0 if not a fixed size type. */
 UNIV_INLINE

innobase/include/data0type.ic

@@ -120,23 +120,6 @@ dtype_get_pad_char(
 	return(ULINT_UNDEFINED);
 }
 
-/*************************************************************************
-Transforms the character code so that it is ordered appropriately for the
-language. */
-UNIV_INLINE
-ulint
-dtype_collate(
-/*==========*/
-				/* out: collation order position */
-	dtype_t*	type,	/* in: type */
-	ulint		code)	/* in: character code stored in database
-				record */
-{
-	ut_ad((type->mtype == DATA_CHAR) || (type->mtype == DATA_VARCHAR));
-
-	return(toupper(code));
-}
-
 /**************************************************************************
 Stores to a type the information which determines its alphabetical
 ordering. */
@@ -198,6 +181,10 @@ dtype_get_fixed_size(
 
 	case DATA_SYS: 	if (type->prtype == DATA_ROW_ID) {
 				return(DATA_ROW_ID_LEN);
+			} else if (type->prtype == DATA_TRX_ID) {
+				return(DATA_TRX_ID_LEN);
+			} else if (type->prtype == DATA_ROLL_PTR) {
+				return(DATA_ROLL_PTR_LEN);
 			} else {
 				return(0);
 			}

innobase/include/db0err.h

@@ -27,11 +27,20 @@ Created 5/24/1996 Heikki Tuuri
 #define DB_CLUSTER_NOT_FOUND	30
 #define DB_TABLE_NOT_FOUND	31
 #define DB_MUST_GET_MORE_FILE_SPACE 32	/* the database has to be stopped
-					and restrated with more file space */
+					and restarted with more file space */
 #define DB_TABLE_IS_BEING_USED	33
 #define DB_TOO_BIG_RECORD	34	/* a record in an index would become
 					bigger than 1/2 free space in a page
 					frame */
+#define DB_LOCK_WAIT_TIMEOUT	35	/* lock wait lasted too long */
+#define DB_NO_REFERENCED_ROW	36	/* referenced key value not found
+					for a foreign key in an insert or
+					update of a row */
+#define DB_ROW_IS_REFERENCED	37	/* cannot delete or update a row
+					because it contains a key value
+					which is referenced */
+#define DB_CANNOT_ADD_CONSTRAINT 38	/* adding a foreign key constraint
+					to a table failed */
 
 /* The following are partial failure codes */
 #define DB_FAIL 		1000

innobase/include/dict0crea.h

@@ -71,6 +71,24 @@ dict_drop_index_tree(
 	rec_t*	rec,	/* in: record in the clustered index of SYS_INDEXES
 			table */
 	mtr_t*	mtr);	/* in: mtr having the latch on the record page */
+/********************************************************************
+Creates the foreign key constraints system tables inside InnoDB
+at database creation or database start if they are not found or are
+not of the right form. */
+
+ulint
+dict_create_or_check_foreign_constraint_tables(void);
+/*================================================*/
+				/* out: DB_SUCCESS or error code */
+/************************************************************************
+Adds foreign key definitions to data dictionary tables in the database. */
+
+ulint
+dict_create_add_foreigns_to_dictionary(
+/*===================================*/
+				/* out: error code or DB_SUCCESS */
+	dict_table_t*	table,	/* in: table */
+	trx_t*		trx);	/* in: transaction */
 
 
 /* Table create node structure */

innobase/include/dict0dict.h

@@ -138,6 +138,38 @@ dict_table_rename_in_cache(
 	dict_table_t*	table,		/* in: table */
 	char*		new_name);	/* in: new name */
 /**************************************************************************
+Adds a foreign key constraint object to the dictionary cache. May free
+the object if there already is an object with the same identifier in.
+At least one of foreign table or referenced table must already be in
+the dictionary cache! */
+
+ulint
+dict_foreign_add_to_cache(
+/*======================*/
+					/* out: DB_SUCCESS or error code */
+	dict_foreign_t*	foreign);	/* in, own: foreign key constraint */
+/*************************************************************************
+Scans a table create SQL string and adds to the data dictionary
+the foreign key constraints declared in the string. This function
+should be called after the indexes for a table have been created.
+Each foreign key constraint must be accompanied with indexes in
+bot participating tables. The indexes are allowed to contain more
+fields than mentioned in the constraint. */
+
+ulint
+dict_create_foreign_constraints(
+/*============================*/
+				/* out: error code or DB_SUCCESS */
+	trx_t*	trx,		/* in: transaction */
+	char*	sql_string,	/* in: table create statement where
+				foreign keys are declared like:
+				FOREIGN KEY (a, b) REFERENCES table2(c, d),
+				table2 can be written also with the database
+				name before it: test.table2; the default
+				database id the database of parameter name */
+	char*	name);		/* in: table full name in the normalized form
+				database_name/table_name */
+/**************************************************************************
 Returns a table object and memoryfixes it. NOTE! This is a high-level
 function to be used mainly from outside the 'dict' directory. Inside this
 directory dict_table_get_low is usually the appropriate function. */
@@ -174,6 +206,14 @@ dict_table_release(
 /*===============*/
 	dict_table_t*	table);	/* in: table to be released */
 /**************************************************************************
+Checks if a table is in the dictionary cache. */
+UNIV_INLINE
+dict_table_t*
+dict_table_check_if_in_cache_low(
+/*==============================*/
+				/* out: table, NULL if not found */
+	char*	table_name);	/* in: table name */
+/**************************************************************************
 Gets a table; loads it to the dictionary cache if necessary. A low-level
 function. */
 UNIV_INLINE
@@ -208,6 +248,13 @@ dict_table_print(
 /*=============*/
 	dict_table_t*	table);	/* in: table */
 /**************************************************************************
+Prints a table data. */
+
+void
+dict_table_print_low(
+/*=================*/
+	dict_table_t*	table);	/* in: table */
+/**************************************************************************
 Prints a table data when we know the table name. */
 
 void
@@ -319,6 +366,16 @@ dict_table_copy_types(
 	dtuple_t*	tuple,	/* in: data tuple */
 	dict_table_t*	table);	/* in: index */
 /**************************************************************************
+Looks for an index with the given id. NOTE that we do not reserve
+the dictionary mutex: this function is for emergency purposes like
+printing info of a corrupt database page! */
+
+dict_index_t*
+dict_index_find_on_id_low(
+/*======================*/
+			/* out: index or NULL if not found from cache */
+	dulint	id);	/* in: index id */
+/**************************************************************************
 Adds an index to dictionary cache. */
 
 ibool
@@ -640,6 +697,23 @@ dict_tree_get_space_reserve(
 				reserved for updates */
 	dict_tree_t*	tree);	/* in: a tree */
 /*************************************************************************
+Calculates the minimum record length in an index. */
+
+ulint
+dict_index_calc_min_rec_len(
+/*========================*/
+	dict_index_t*	index);	/* in: index */
+/*************************************************************************
+Calculates new estimates for table and index statistics. The statistics
+are used in query optimization. */
+
+void
+dict_update_statistics_low(
+/*=======================*/
+	dict_table_t*	table,		/* in: table */
+	ibool		has_dict_mutex);/* in: TRUE if the caller has the
+					dictionary mutex */	
+/*************************************************************************
 Calculates new estimates for table and index statistics. The statistics
 are used in query optimization. */
 
@@ -662,6 +736,7 @@ dict_mutex_exit_for_mysql(void);
 
 
 extern dict_sys_t*	dict_sys;	/* the dictionary system */
+extern rw_lock_t	dict_foreign_key_check_lock;
 
 /* Dictionary system struct */
 struct dict_sys_struct{

innobase/include/dict0dict.ic

@@ -532,12 +532,11 @@ dict_tree_get_space_reserve(
 }
 
 /**************************************************************************
-Gets a table; loads it to the dictionary cache if necessary. A low-level
-function. */
+Checks if a table is in the dictionary cache. */
 UNIV_INLINE
 dict_table_t*
-dict_table_get_low(
-/*===============*/
+dict_table_check_if_in_cache_low(
+/*==============================*/
 				/* out: table, NULL if not found */
 	char*	table_name)	/* in: table name */
 {
@@ -552,6 +551,26 @@ dict_table_get_low(
 
 	HASH_SEARCH(name_hash, dict_sys->table_hash, table_fold, table,
 				ut_strcmp(table->name, table_name) == 0);
+	return(table);
+}
+
+/**************************************************************************
+Gets a table; loads it to the dictionary cache if necessary. A low-level
+function. */
+UNIV_INLINE
+dict_table_t*
+dict_table_get_low(
+/*===============*/
+				/* out: table, NULL if not found */
+	char*	table_name)	/* in: table name */
+{
+	dict_table_t*	table;
+	
+	ut_ad(table_name);
+	ut_ad(mutex_own(&(dict_sys->mutex)));
+	
+	table = dict_table_check_if_in_cache_low(table_name);
+	
 	if (table == NULL) {
 		table = dict_load_table(table_name);
 	}
@@ -603,6 +622,7 @@ dict_table_get_on_id_low(
 	dict_table_t*	table;
 	ulint		fold;
 
+	ut_ad(mutex_own(&(dict_sys->mutex)));
 	UT_NOT_USED(trx);
 	
 	/* Look for the table name in the hash table */

innobase/include/dict0load.h

@@ -14,9 +14,20 @@ Created 4/24/1996 Heikki Tuuri
 #include "dict0types.h"
 #include "ut0byte.h"
 
+/************************************************************************
+Finds the first table name in the given database. */
+
+char*
+dict_get_first_table_name_in_db(
+/*============================*/
+			/* out, own: table name, NULL if does not exist;
+			the caller must free the memory in the string! */
+	char*	name);	/* in: database name which ends to '/' */
 /************************************************************************
 Loads a table definition and also all its index definitions, and also
-the cluster definition, if the table is a member in a cluster. */
+the cluster definition if the table is a member in a cluster. Also loads
+all foreign key constraints where the foreign key is in the table or where
+a foreign key references columns in this table. */
 
 dict_table_t*
 dict_load_table(
@@ -40,6 +51,25 @@ void
 dict_load_sys_table(
 /*================*/
 	dict_table_t*	table);	/* in: system table */
+/***************************************************************************
+Loads foreign key constraints where the table is either the foreign key
+holder or where the table is referenced by a foreign key. Adds these
+constraints to the data dictionary. Note that we know that the dictionary
+cache already contains all constraints where the other relevant table is
+already in the dictionary cache. */
+
+ulint
+dict_load_foreigns(
+/*===============*/
+				/* out: DB_SUCCESS or error code */
+	char*	table_name);	/* in: table name */
+/************************************************************************
+Prints to the standard output information on all tables found in the data
+dictionary system table. */
+
+void
+dict_print(void);
+/*============*/
 
 
 #ifndef UNIV_NONINL

innobase/include/dict0mem.h

@@ -123,6 +123,13 @@ dict_mem_index_free(
 /*================*/
 	dict_index_t*	index);	/* in: index */
 /**************************************************************************
+Creates and initializes a foreign constraint memory object. */
+
+dict_foreign_t*
+dict_mem_foreign_create(void);
+/*=========================*/
+				/* out, own: foreign constraint struct */
+/**************************************************************************
 Creates a procedure memory object. */
 
 dict_proc_t*
@@ -221,15 +228,56 @@ struct dict_index_struct{
 				dictionary cache */
 	btr_search_t*	search_info; /* info used in optimistic searches */
 	/*----------------------*/
-	ulint		stat_n_diff_key_vals;
+	ib_longlong*	stat_n_diff_key_vals;
 				/* approximate number of different key values
-				for this index; we periodically calculate
-				new estimates */
+				for this index, for each n-column prefix
+				where n <= dict_get_n_unique(index); we
+				periodically calculate new estimates */
 	ulint		stat_index_size;
 				/* approximate index size in database pages */
+	ulint		stat_n_leaf_pages;
+				/* approximate number of leaf pages in the
+				index tree */
 	ulint		magic_n;/* magic number */
 };
 
+/* Data structure for a foreign key constraint; an example:
+FOREIGN KEY (A, B) REFERENCES TABLE2 (C, D) */
+
+struct dict_foreign_struct{
+	mem_heap_t*	heap;		/* this object is allocated from
+					this memory heap */
+	char*		id;		/* id of the constraint as a
+					null-terminated string */
+	char*		foreign_table_name;/* foreign table name */
+	dict_table_t*	foreign_table;	/* table where the foreign key is */
+	char**		foreign_col_names;/* names of the columns in the
+					foreign key */
+	char*		referenced_table_name;/* referenced table name */
+	dict_table_t*	referenced_table;/* table where the referenced key
+					is */
+	char**		referenced_col_names;/* names of the referenced
+					columns in the referenced table */
+	ulint		n_fields;	/* number of indexes' first fields
+					for which the the foreign key
+					constraint is defined: we allow the
+					indexes to contain more fields than
+					mentioned in the constraint, as long
+					as the first fields are as mentioned */ 
+	dict_index_t*	foreign_index;	/* foreign index; we require that
+					both tables contain explicitly defined
+					indexes for the constraint: InnoDB
+					does not generate new indexes
+					implicitly */
+	dict_index_t*	referenced_index;/* referenced index */
+	UT_LIST_NODE_T(dict_foreign_t)
+			foreign_list;	/* list node for foreign keys of the
+					table */
+	UT_LIST_NODE_T(dict_foreign_t)
+			referenced_list;/* list node for referenced keys of the
+					table */
+};
+
 #define	DICT_INDEX_MAGIC_N	76789786
 
 /* Data structure for a database table */
@@ -247,6 +295,13 @@ struct dict_table_struct{
 	dict_col_t*	cols;	/* array of column descriptions */
 	UT_LIST_BASE_NODE_T(dict_index_t)
 			indexes; /* list of indexes of the table */
+	UT_LIST_BASE_NODE_T(dict_foreign_t)
+			foreign_list;/* list of foreign key constraints
+				in the table; these refer to columns
+				in other tables */
+	UT_LIST_BASE_NODE_T(dict_foreign_t)
+			referenced_list;/* list of foreign key constraints
+				which refer to this table */
 	UT_LIST_NODE_T(dict_table_t)
 			table_LRU; /* node of the LRU list of tables */
 	ulint		mem_fix;/* count of how many times the table 
@@ -254,6 +309,13 @@ struct dict_table_struct{
 				currently NOT used */
 	ibool		cached;	/* TRUE if the table object has been added
 				to the dictionary cache */
+	lock_t*		auto_inc_lock;/* a buffer for an auto-inc lock
+				for this table: we allocate the memory here
+				so that individual transactions can get it
+				and release it without a need to allocate
+				space from the lock heap of the trx:
+				otherwise the lock heap would grow rapidly
+				if we do a large insert from a select */
 	UT_LIST_BASE_NODE_T(lock_t)
 			locks; /* list of locks on the table */
 	/*----------------------*/
@@ -278,7 +340,7 @@ struct dict_table_struct{
 				forget about value TRUE if it has to reload
 				the table definition from disk */
 	/*----------------------*/
-	ulint		stat_n_rows;
+	ib_longlong	stat_n_rows;
 				/* approximate number of rows in the table;
 				we periodically calculate new estimates */
 	ulint		stat_clustered_index_size;

innobase/include/dict0types.h

@@ -16,6 +16,7 @@ typedef struct dict_index_struct	dict_index_t;
 typedef struct dict_tree_struct		dict_tree_t;
 typedef struct dict_table_struct	dict_table_t;
 typedef struct dict_proc_struct		dict_proc_t;
+typedef struct dict_foreign_struct	dict_foreign_t;
 
 /* A cluster object is a table object with the type field set to
 DICT_CLUSTERED */

innobase/include/fil0fil.h

@@ -76,6 +76,9 @@ extern fil_addr_t	fil_addr_null;
 #define FIL_TABLESPACE 		501
 #define FIL_LOG			502
 
+extern ulint	fil_n_pending_log_flushes;
+extern ulint	fil_n_pending_tablespace_flushes;
+
 /***********************************************************************
 Reserves a right to open a single file. The right must be released with
 fil_release_right_to_open. */

innobase/include/ibuf0ibuf.h

@@ -226,6 +226,21 @@ ibuf_contract(
 			issued read with the highest tablespace address
 			to complete */
 /*************************************************************************
+Contracts insert buffer trees by reading pages to the buffer pool. */
+
+ulint
+ibuf_contract_for_n_pages(
+/*======================*/
+			/* out: a lower limit for the combined size in bytes
+			of entries which will be merged from ibuf trees to the
+			pages read, 0 if ibuf is empty */
+	ibool	sync,	/* in: TRUE if the caller wants to wait for the
+			issued read with the highest tablespace address
+			to complete */
+	ulint	n_pages);/* in: try to read at least this many pages to
+			the buffer pool and merge the ibuf contents to
+			them */
+/*************************************************************************
 Parses a redo log record of an ibuf bitmap page init. */
 
 byte*

innobase/include/lock0lock.h

@@ -21,15 +21,13 @@ Created 5/7/1996 Heikki Tuuri
 
 extern ibool	lock_print_waits;
 
-/*****************************************************************
-Cancels a waiting record lock request and releases the waiting transaction
-that requested it. NOTE: does NOT check if waiting lock requests behind this
-one can now be granted! */
+/*************************************************************************
+Gets the size of a lock struct. */
 
-void
-lock_rec_cancel(
-/*============*/
-	lock_t*	lock);	/* in: waiting record lock request */
+ulint
+lock_get_size(void);
+/*===============*/
+			/* out: size in bytes */
 /*************************************************************************
 Creates the lock system at database start. */
 
@@ -388,6 +386,14 @@ lock_is_on_table(
 				/* out: TRUE if there are lock(s) */
 	dict_table_t*	table);	/* in: database table in dictionary cache */
 /*************************************************************************
+Releases an auto-inc lock a transaction possibly has on a table.
+Releases possible other transactions waiting for this lock. */
+
+void
+lock_table_unlock_auto_inc(
+/*=======================*/
+	trx_t*	trx);	/* in: transaction */
+/*************************************************************************
 Releases transaction locks, and releases possible other transactions waiting
 because of these locks. */
 
@@ -396,6 +402,14 @@ lock_release_off_kernel(
 /*====================*/
 	trx_t*	trx);	/* in: transaction */
 /*************************************************************************
+Cancels a waiting lock request and releases possible other transactions
+waiting behind it. */
+
+void
+lock_cancel_waiting_and_release(
+/*============================*/
+	lock_t*	lock);	/* in: waiting lock request */
+/*************************************************************************
 Resets all locks, both table and record locks, on a table to be dropped.
 No lock is allowed to be a wait lock. */
 
@@ -495,6 +509,8 @@ extern lock_sys_t*	lock_sys;
 #define	LOCK_IX		3	/* intention exclusive */
 #define	LOCK_S		4	/* shared */
 #define	LOCK_X		5	/* exclusive */
+#define	LOCK_AUTO_INC	6	/* locks the auto-inc counter of a table
+				in an exclusive mode */
 #define LOCK_MODE_MASK	0xF	/* mask used to extract mode from the
 				type_mode field in a lock */
 #define LOCK_TABLE	16	/* these type values should be so high that */

innobase/include/log0log.h

@@ -659,6 +659,11 @@ struct log_struct{
 					mutex! */
 	ulint		n_log_ios;	/* number of log i/os initiated thus
 					far */
+	ulint		n_log_ios_old;	/* number of log i/o's at the
+					previous printout */
+	time_t		last_printout_time;/* when log_print was last time
+					called */
+
 	/* Fields involved in checkpoints */
 	ulint		max_modified_age_async;
 					/* when this recommended value for lsn

innobase/include/mtr0mtr.h

@@ -203,20 +203,12 @@ mtr_read_dulint(
 	mtr_t*	mtr);	/* in: mini-transaction handle */
 /*************************************************************************
 This macro locks an rw-lock in s-mode. */
-#ifdef UNIV_SYNC_DEBUG
 #define mtr_s_lock(B, MTR)	mtr_s_lock_func((B), IB__FILE__, __LINE__,\
 						(MTR))
-#else
-#define mtr_s_lock(B, MTR)	mtr_s_lock_func((B), (MTR))
-#endif
 /*************************************************************************
 This macro locks an rw-lock in x-mode. */
-#ifdef UNIV_SYNC_DEBUG
 #define mtr_x_lock(B, MTR)	mtr_x_lock_func((B), IB__FILE__, __LINE__,\
 						(MTR))
-#else
-#define mtr_x_lock(B, MTR)	mtr_x_lock_func((B), (MTR))
-#endif
 /*************************************************************************
 NOTE! Use the macro above!
 Locks a lock in s-mode. */
@@ -225,10 +217,8 @@ void
 mtr_s_lock_func(
 /*============*/
 	rw_lock_t*	lock,	/* in: rw-lock */
-#ifdef UNIV_SYNC_DEBUG
 	char*		file,	/* in: file name */
 	ulint		line,	/* in: line number */
-#endif
 	mtr_t*		mtr);	/* in: mtr */
 /*************************************************************************
 NOTE! Use the macro above!
@@ -238,10 +228,8 @@ void
 mtr_x_lock_func(
 /*============*/
 	rw_lock_t*	lock,	/* in: rw-lock */
-#ifdef UNIV_SYNC_DEBUG
 	char*		file,	/* in: file name */
 	ulint		line,	/* in: line number */
-#endif
 	mtr_t*		mtr);	/* in: mtr */
 
 /*******************************************************

innobase/include/mtr0mtr.ic

@@ -217,20 +217,14 @@ void
 mtr_s_lock_func(
 /*============*/
 	rw_lock_t*	lock,	/* in: rw-lock */
-#ifdef UNIV_SYNC_DEBUG
 	char*		file,	/* in: file name */
 	ulint		line,	/* in: line number */
-#endif
 	mtr_t*		mtr)	/* in: mtr */
 {
 	ut_ad(mtr);
 	ut_ad(lock);
 
-	rw_lock_s_lock_func(lock
-		#ifdef UNIV_SYNC_DEBUG
-				,0, file, line
-		#endif
-			    );
+	rw_lock_s_lock_func(lock, 0, file, line);
 
 	mtr_memo_push(mtr, lock, MTR_MEMO_S_LOCK);
 }
@@ -242,20 +236,14 @@ void
 mtr_x_lock_func(
 /*============*/
 	rw_lock_t*	lock,	/* in: rw-lock */
-#ifdef UNIV_SYNC_DEBUG
 	char*		file,	/* in: file name */
 	ulint		line,	/* in: line number */
-#endif
 	mtr_t*		mtr)	/* in: mtr */
 {
 	ut_ad(mtr);
 	ut_ad(lock);
 
-	rw_lock_x_lock_func(lock, 0
-		#ifdef UNIV_SYNC_DEBUG
-				, file, line
-		#endif
-			    );
+	rw_lock_x_lock_func(lock, 0, file, line);
 
 	mtr_memo_push(mtr, lock, MTR_MEMO_X_LOCK);
 }

innobase/include/rem0cmp.h

@@ -15,6 +15,32 @@ Created 7/1/1994 Heikki Tuuri
 #include "dict0dict.h"
 #include "rem0rec.h"
 
+/*****************************************************************
+Returns TRUE if two types are equal for comparison purposes. */
+
+ibool
+cmp_types_are_equal(
+/*================*/
+				/* out: TRUE if the types are considered
+				equal in comparisons */
+	dtype_t*	type1,	/* in: type 1 */
+	dtype_t*	type2);	/* in: type 2 */
+/*****************************************************************
+This function is used to compare two data fields for which we know the
+data type. */
+UNIV_INLINE
+int
+cmp_data_data(
+/*==========*/	
+				/* out: 1, 0, -1, if data1 is greater, equal, 
+				less than data2, respectively */
+	dtype_t*	cur_type,/* in: data type of the fields */
+	byte*		data1,	/* in: data field (== a pointer to a memory
+				buffer) */
+	ulint		len1,	/* in: data field length or UNIV_SQL_NULL */
+	byte*		data2,	/* in: data field (== a pointer to a memory
+				buffer) */
+	ulint		len2);	/* in: data field length or UNIV_SQL_NULL */
 /*****************************************************************
 This function is used to compare two dfields where at least the first
 has its data type field set. */

innobase/include/row0ins.h

@@ -16,6 +16,28 @@ Created 4/20/1996 Heikki Tuuri
 #include "trx0types.h"
 #include "row0types.h"
 	
+/*******************************************************************
+Checks if foreign key constraint fails for an index entry. Sets shared locks
+which lock either the success or the failure of the constraint. NOTE that
+the caller must have a shared latch on dict_foreign_key_check_lock. */
+
+ulint
+row_ins_check_foreign_constraint(
+/*=============================*/
+				/* out: DB_SUCCESS, DB_LOCK_WAIT,
+				DB_NO_REFERENCED_ROW,
+				or DB_ROW_IS_REFERENCED */
+	ibool		check_ref,/* in: TRUE If we want to check that
+				the referenced table is ok, FALSE if we
+				want to to check the foreign key table */
+	dict_foreign_t*	foreign,/* in: foreign constraint; NOTE that the
+				tables mentioned in it must be in the
+				dictionary cache if they exist at all */
+	dict_table_t*	table,	/* in: if check_ref is TRUE, then the foreign
+				table, else the referenced table */
+	dict_index_t*	index,	/* in: index in table */
+	dtuple_t*	entry,	/* in: index entry for index */
+	que_thr_t*	thr);	/* in: query thread */
 /*************************************************************************
 Creates an insert node struct. */
 

innobase/include/row0mysql.h

@@ -133,6 +133,26 @@ row_update_prebuilt_trx(
 					handle */
 	trx_t*		trx);		/* in: transaction handle */
 /*************************************************************************
+Unlocks an AUTO_INC type lock possibly reserved by trx. */
+
+void		  	
+row_unlock_table_autoinc_for_mysql(
+/*===============================*/
+	trx_t*	trx);	/* in: transaction */
+/*************************************************************************
+Sets an AUTO_INC type lock on the table mentioned in prebuilt. The
+AUTO_INC lock gives exclusive access to the auto-inc counter of the
+table. The lock is reserved only for the duration of an SQL statement.
+It is not compatible with another AUTO_INC or exclusive lock on the
+table. */
+
+int
+row_lock_table_autoinc_for_mysql(
+/*=============================*/
+					/* out: error code or DB_SUCCESS */
+	row_prebuilt_t*	prebuilt);	/* in: prebuilt struct in the MySQL
+					table handle */
+/*************************************************************************
 Does an insert for MySQL. */
 
 int
@@ -211,6 +231,26 @@ row_create_index_for_mysql(
 	dict_index_t*	index,		/* in: index defintion */
 	trx_t*		trx);		/* in: transaction handle */
 /*************************************************************************
+Scans a table create SQL string and adds to the data dictionary
+the foreign key constraints declared in the string. This function
+should be called after the indexes for a table have been created.
+Each foreign key constraint must be accompanied with indexes in
+bot participating tables. The indexes are allowed to contain more
+fields than mentioned in the constraint. */
+
+int
+row_table_add_foreign_constraints(
+/*==============================*/
+				/* out: error code or DB_SUCCESS */
+	trx_t*	trx,		/* in: transaction */
+	char*	sql_string,	/* in: table create statement where
+				foreign keys are declared like:
+				FOREIGN KEY (a, b) REFERENCES table2(c, d),
+				table2 can be written also with the database
+				name before it: test.table2 */
+	char*	name);		/* in: table full name in the normalized form
+				database_name/table_name */
+/*************************************************************************
 Drops a table for MySQL. If the name of the dropped table ends to
 characters INNODB_MONITOR, then this also stops printing of monitor
 output by the master thread. */
@@ -224,6 +264,15 @@ row_drop_table_for_mysql(
 	ibool	has_dict_mutex);/* in: TRUE if the caller already owns the
 				dictionary system mutex */
 /*************************************************************************
+Drops a database for MySQL. */
+
+int
+row_drop_database_for_mysql(
+/*========================*/
+			/* out: error code or DB_SUCCESS */
+	char*	name,	/* in: database name which ends to '/' */
+	trx_t*	trx);	/* in: transaction handle */
+/*************************************************************************
 Renames a table for MySQL. */
 
 int

innobase/include/row0upd.h

@@ -47,8 +47,7 @@ upd_get_nth_field(
 	upd_t*	update,	/* in: update vector */
 	ulint	n);	/* in: field position in update vector */
 /*************************************************************************
-Sets the clustered index field number to be updated by an update vector
-field. */
+Sets an index field number to be updated by an update vector field. */
 UNIV_INLINE
 void
 upd_field_set_field_no(
@@ -56,7 +55,7 @@ upd_field_set_field_no(
 	upd_field_t*	upd_field,	/* in: update vector field */
 	ulint		field_no,	/* in: field number in a clustered
 					index */
-	dict_index_t*	index);		/* in: clustered index */
+	dict_index_t*	index);		/* in: index */
 /*************************************************************************
 Writes into the redo log the values of trx id and roll ptr and enough info
 to determine their positions within a clustered index record. */
@@ -136,13 +135,27 @@ row_upd_rec_in_place(
 	rec_t*	rec,	/* in/out: record where replaced */
 	upd_t*	update);/* in: update vector */
 /*******************************************************************
+Builds an update vector from those fields which in a secondary index entry
+differ from a record that has the equal ordering fields. NOTE: we compare
+the fields as binary strings! */
+
+upd_t*
+row_upd_build_sec_rec_difference_binary(
+/*====================================*/
+				/* out, own: update vector of differing
+				fields */
+	dict_index_t*	index,	/* in: index */
+	dtuple_t*	entry,	/* in: entry to insert */
+	rec_t*		rec,	/* in: secondary index record */
+	mem_heap_t*	heap);	/* in: memory heap from which allocated */
+/*******************************************************************
 Builds an update vector from those fields, excluding the roll ptr and
 trx id fields, which in an index entry differ from a record that has
-the equal ordering fields. */
+the equal ordering fields. NOTE: we compare the fields as binary strings! */
 
 upd_t*
-row_upd_build_difference(
-/*=====================*/
+row_upd_build_difference_binary(
+/*============================*/
 				/* out, own: update vector of differing
 				fields, excluding roll ptr and trx id */
 	dict_index_t*	index,	/* in: clustered index */
@@ -175,13 +188,16 @@ row_upd_clust_index_replace_new_col_vals(
 /***************************************************************
 Checks if an update vector changes an ordering field of an index record.
 This function is fast if the update vector is short or the number of ordering
-fields in the index is small. Otherwise, this can be quadratic. */
+fields in the index is small. Otherwise, this can be quadratic.
+NOTE: we compare the fields as binary strings! */
  
 ibool
-row_upd_changes_ord_field(
-/*======================*/
+row_upd_changes_ord_field_binary(
+/*=============================*/
 				/* out: TRUE if update vector changes
-				an ordering field in the index record */
+				an ordering field in the index record;
+				NOTE: the fields are compared as binary
+				strings */
 	dtuple_t*	row,	/* in: old value of row, or NULL if the
 				row and the data values in update are not
 				known when this function is called, e.g., at
@@ -191,11 +207,12 @@ row_upd_changes_ord_field(
 /***************************************************************
 Checks if an update vector changes an ordering field of an index record.
 This function is fast if the update vector is short or the number of ordering
-fields in the index is small. Otherwise, this can be quadratic. */
+fields in the index is small. Otherwise, this can be quadratic.
+NOTE: we compare the fields as binary strings! */
 
 ibool
-row_upd_changes_some_index_ord_field(
-/*=================================*/
+row_upd_changes_some_index_ord_field_binary(
+/*========================================*/
 				/* out: TRUE if update vector may change
 				an ordering field in an index record */
 	dict_table_t*	table,	/* in: table */

innobase/include/row0upd.ic

@@ -70,8 +70,7 @@ upd_get_nth_field(
 }
 
 /*************************************************************************
-Sets the clustered index field number to be updated by an update vector
-field. */
+Sets an index field number to be updated by an update vector field. */
 UNIV_INLINE
 void
 upd_field_set_field_no(
@@ -79,12 +78,18 @@ upd_field_set_field_no(
 	upd_field_t*	upd_field,	/* in: update vector field */
 	ulint		field_no,	/* in: field number in a clustered
 					index */
-	dict_index_t*	index)		/* in: clustered index */
+	dict_index_t*	index)		/* in: index */
 {	
-	ut_ad(index->type & DICT_CLUSTERED);
-	
 	upd_field->field_no = field_no;
 
+	if (field_no >= dict_index_get_n_fields(index)) {
+		fprintf(stderr,
+		"InnoDB: Error: trying to access field %lu in table %s\n"
+		"InnoDB: index %s, but index has only %lu fields\n",
+		field_no, index->table_name, index->name,
+		dict_index_get_n_fields(index));
+	}
+
 	dtype_copy(dfield_get_type(&(upd_field->new_val)),
 				dict_index_get_nth_type(index, field_no));
 }

innobase/include/srv0srv.h

@@ -16,6 +16,11 @@ Created 10/10/1995 Heikki Tuuri
 #include "com0com.h"
 #include "que0types.h"
 
+
+/* When this event is set the lock timeout and InnoDB monitor
+thread starts running */
+extern os_event_t	srv_lock_timeout_thread_event;
+
 /* Server parameters which are read from the initfile */
 
 extern char*	srv_data_home;
@@ -27,6 +32,8 @@ extern char**	srv_data_file_names;
 extern ulint*	srv_data_file_sizes;
 extern ulint*   srv_data_file_is_raw_partition;
 
+extern ibool	srv_created_new_raw;
+
 #define SRV_NEW_RAW    1
 #define SRV_OLD_RAW    2
 
@@ -39,6 +46,8 @@ extern ibool	srv_log_archive_on;
 extern ulint	srv_log_buffer_size;
 extern ibool	srv_flush_log_at_trx_commit;
 
+extern byte	srv_latin1_ordering[256];/* The sort order table of the latin1
+					character set */
 extern ibool	srv_use_native_aio;		
 
 extern ulint	srv_pool_size;
@@ -54,6 +63,7 @@ extern ulint	srv_lock_wait_timeout;
 
 extern char*    srv_unix_file_flush_method_str;
 extern ulint    srv_unix_file_flush_method;
+extern ulint	srv_force_recovery;
 
 extern ibool	srv_use_doublewrite_buf;
 
@@ -71,6 +81,7 @@ extern ibool	srv_print_innodb_monitor;
 extern ibool    srv_print_innodb_lock_monitor;
 extern ibool    srv_print_innodb_tablespace_monitor;
 extern ibool    srv_print_verbose_log;
+extern ibool    srv_print_innodb_table_monitor;
 
 extern ulint	srv_n_spin_wait_rounds;
 extern ulint	srv_spin_wait_delay;
@@ -133,6 +144,25 @@ what these mean */
 #define SRV_UNIX_LITTLESYNC  3
 #define SRV_UNIX_NOSYNC      4
 
+/* Alternatives for srv_force_recovery. Non-zero values are intended
+to help the user get a damaged database up so that he can dump intact
+tables and rows with SELECT INTO OUTFILE. The database must not otherwise
+be used with these options! A bigger number below means that all precautions
+of lower numbers are included. */
+
+#define SRV_FORCE_IGNORE_CORRUPT 1	/* let the server run even if it
+					detects a corrupt page */
+#define SRV_FORCE_NO_BACKGROUND	2 	/* prevent the main thread from
+					running: if a crash would occur
+					in purge, this prevents it */
+#define SRV_FORCE_NO_TRX_UNDO	3	/* do not run trx rollback after
+					recovery */
+#define SRV_FORCE_NO_IBUF_MERGE	4	/* prevent also ibuf operations:
+					if they would cause a crash, better
+					not do them */
+#define SRV_FORCE_NO_LOG_REDO	5	/* do not do the log roll-forward
+					in connection with recovery */
+					
 /*************************************************************************
 Boots Innobase server. */
 
@@ -225,15 +255,30 @@ srv_release_mysql_thread_if_suspended(
 	que_thr_t*	thr);	/* in: query thread associated with the
 				MySQL OS thread  */
 /*************************************************************************
-A thread which wakes up threads whose lock wait may have lasted too long. */
+A thread which wakes up threads whose lock wait may have lasted too long.
+This also prints the info output by various InnoDB monitors. */
 
 #ifndef __WIN__
 void*
 #else
 ulint
 #endif
-srv_lock_timeout_monitor_thread(
-/*============================*/
+srv_lock_timeout_and_monitor_thread(
+/*================================*/
+			/* out: a dummy parameter */
+	void*	arg);	/* in: a dummy parameter required by
+			os_thread_create */
+/*************************************************************************
+A thread which prints warnings about semaphore waits which have lasted
+too long. These can be used to track bugs which cause hangs. */
+
+#ifndef __WIN__
+void*
+#else
+ulint
+#endif
+srv_error_monitor_thread(
+/*=====================*/
 			/* out: a dummy parameter */
 	void*	arg);	/* in: a dummy parameter required by
 			os_thread_create */

innobase/include/sync0arr.h

@@ -51,12 +51,8 @@ sync_array_reserve_cell(
         sync_array_t*	arr,	/* in: wait array */
         void*   	object, /* in: pointer to the object to wait for */
         ulint		type,	/* in: lock request type */
-	#ifdef UNIV_SYNC_DEBUG
-        char*		file,	/* in: in debug version file where
-        			requested */
-        ulint		line,	/* in: in the debug version line where
-        			requested */
-	#endif
+        char*		file,	/* in: file where requested */
+        ulint		line,	/* in: line where requested */
         ulint*   	index); /* out: index of the reserved cell */
 /**********************************************************************
 This function should be called when a thread starts to wait on
@@ -90,6 +86,20 @@ sync_array_signal_object(
 /*=====================*/
 	sync_array_t*	arr,	/* in: wait array */
 	void*		object);/* in: wait object */
+/**************************************************************************
+If the wakeup algorithm does not work perfectly at semaphore relases,
+this function will do the waking (see the comment in mutex_exit). This
+function should be called about every 1 second in the server. */
+
+void
+sync_arr_wake_threads_if_sema_free(void);
+/*====================================*/
+/**************************************************************************
+Prints warnings of long semaphore waits to stderr. Currently > 120 sec. */
+
+void
+sync_array_print_long_waits(void);
+/*=============================*/
 /************************************************************************
 Validates the integrity of the wait array. Checks
 that the number of reserved cells equals the count variable. */

innobase/include/sync0ipm.ic

@@ -92,7 +92,7 @@ loop:
 	loop_count++;
 	ut_ad(loop_count < 15);
 
-	if (mutex_enter_nowait(mutex) == 0) {
+	if (mutex_enter_nowait(mutex, IB__FILE__, __LINE__) == 0) {
 		/* Succeeded! */
 
 		return(0);
@@ -105,7 +105,7 @@ loop:
 	/* Order is important here: FIRST reset event, then set waiters */
 	ip_mutex_set_waiters(ip_mutex, 1);
 
-	if (mutex_enter_nowait(mutex) == 0) {
+	if (mutex_enter_nowait(mutex, IB__FILE__, __LINE__) == 0) {
 		/* Succeeded! */
 
 		return(0);

innobase/include/sync0rw.h

@@ -46,9 +46,10 @@ extern ibool		rw_lock_debug_waiters;	/* This is set to TRUE, if
 extern	ulint	rw_s_system_call_count;
 extern	ulint	rw_s_spin_wait_count;
 extern	ulint	rw_s_exit_count;
-
+extern	ulint	rw_s_os_wait_count;
 extern	ulint	rw_x_system_call_count;
 extern	ulint	rw_x_spin_wait_count;
+extern	ulint	rw_x_os_wait_count;
 extern	ulint	rw_x_exit_count;
 
 /**********************************************************************
@@ -92,32 +93,20 @@ rw_lock_validate(
 NOTE! The following macros should be used in rw s-locking, not the
 corresponding function. */
 
-#ifdef UNIV_SYNC_DEBUG
 #define rw_lock_s_lock(M)    rw_lock_s_lock_func(\
 					  (M), 0, IB__FILE__, __LINE__)
-#else
-#define rw_lock_s_lock(M)    rw_lock_s_lock_func(M)
-#endif
 /******************************************************************
 NOTE! The following macros should be used in rw s-locking, not the
 corresponding function. */
 
-#ifdef UNIV_SYNC_DEBUG
 #define rw_lock_s_lock_gen(M, P)    rw_lock_s_lock_func(\
 					  (M), (P), IB__FILE__, __LINE__)
-#else
-#define rw_lock_s_lock_gen(M, P)    rw_lock_s_lock_func(M)
-#endif
 /******************************************************************
 NOTE! The following macros should be used in rw s-locking, not the
 corresponding function. */
 
-#ifdef UNIV_SYNC_DEBUG
 #define rw_lock_s_lock_nowait(M)    rw_lock_s_lock_func_nowait(\
 					     (M), IB__FILE__, __LINE__)
-#else
-#define rw_lock_s_lock_nowait(M)    rw_lock_s_lock_func_nowait(M)
-#endif
 /**********************************************************************
 NOTE! Use the corresponding macro, not directly this function, except if
 you supply the file name and line number. Lock an rw-lock in shared mode
@@ -129,14 +118,11 @@ UNIV_INLINE
 void
 rw_lock_s_lock_func(
 /*================*/
-        rw_lock_t*   	lock  	/* in: pointer to rw-lock */
-	#ifdef UNIV_SYNC_DEBUG
-	,ulint		pass,	/* in: pass value; != 0, if the lock will
+        rw_lock_t*   	lock,  	/* in: pointer to rw-lock */
+	ulint		pass,	/* in: pass value; != 0, if the lock will
 				be passed to another thread to unlock */
-	char*		file_name, /* in: file name where lock requested */
-	ulint		line	/* in: line where requested */
-	#endif
-);
+	char*		file_name,/* in: file name where lock requested */
+	ulint		line);	/* in: line where requested */
 /**********************************************************************
 NOTE! Use the corresponding macro, not directly this function, except if
 you supply the file name and line number. Lock an rw-lock in shared mode
@@ -146,12 +132,9 @@ ibool
 rw_lock_s_lock_func_nowait(
 /*=======================*/
 				/* out: TRUE if success */
-        rw_lock_t*   	lock  	/* in: pointer to rw-lock */
-	#ifdef UNIV_SYNC_DEBUG
-	,char*		file_name, /* in: file name where lock requested */
-	ulint		line	/* in: line where requested */
-	#endif
-);
+        rw_lock_t*   	lock,  	/* in: pointer to rw-lock */
+	char*		file_name,/* in: file name where lock requested */
+	ulint		line);	/* in: line where requested */
 /**********************************************************************
 NOTE! Use the corresponding macro, not directly this function! Lock an
 rw-lock in exclusive mode for the current thread if the lock can be
@@ -161,12 +144,9 @@ ibool
 rw_lock_x_lock_func_nowait(
 /*=======================*/
 				/* out: TRUE if success */
-        rw_lock_t*   	lock  	/* in: pointer to rw-lock */
-	#ifdef UNIV_SYNC_DEBUG
-	,char*		file_name, /* in: file name where lock requested */
-	ulint		line	/* in: line where requested */
-	#endif
-);
+        rw_lock_t*   	lock,  	/* in: pointer to rw-lock */
+	char*		file_name,/* in: file name where lock requested */
+	ulint		line);	/* in: line where requested */
 /**********************************************************************
 Releases a shared mode lock. */
 UNIV_INLINE
@@ -199,32 +179,20 @@ Releases a shared mode lock. */
 NOTE! The following macro should be used in rw x-locking, not the
 corresponding function. */
 
-#ifdef UNIV_SYNC_DEBUG
 #define rw_lock_x_lock(M)    rw_lock_x_lock_func(\
 					  (M), 0, IB__FILE__, __LINE__)
-#else
-#define rw_lock_x_lock(M)    rw_lock_x_lock_func(M, 0)
-#endif
 /******************************************************************
 NOTE! The following macro should be used in rw x-locking, not the
 corresponding function. */
 
-#ifdef UNIV_SYNC_DEBUG
 #define rw_lock_x_lock_gen(M, P)    rw_lock_x_lock_func(\
 					  (M), (P), IB__FILE__, __LINE__)
-#else
-#define rw_lock_x_lock_gen(M, P)    rw_lock_x_lock_func(M, P)
-#endif
 /******************************************************************
 NOTE! The following macros should be used in rw x-locking, not the
 corresponding function. */
 
-#ifdef UNIV_SYNC_DEBUG
 #define rw_lock_x_lock_nowait(M)    rw_lock_x_lock_func_nowait(\
 					     (M), IB__FILE__, __LINE__)
-#else
-#define rw_lock_x_lock_nowait(M)    rw_lock_x_lock_func_nowait(M)
-#endif
 /**********************************************************************
 NOTE! Use the corresponding macro, not directly this function! Lock an
 rw-lock in exclusive mode for the current thread. If the rw-lock is locked
@@ -239,13 +207,10 @@ void
 rw_lock_x_lock_func(
 /*================*/
         rw_lock_t*   	lock,  	/* in: pointer to rw-lock */
-	ulint		pass	/* in: pass value; != 0, if the lock will
+	ulint		pass,	/* in: pass value; != 0, if the lock will
 				be passed to another thread to unlock */
-	#ifdef UNIV_SYNC_DEBUG
-	,char*		file_name, /* in: file name where lock requested */
-	ulint		line	/* in: line where requested */
-	#endif
-);
+	char*		file_name,/* in: file name where lock requested */
+	ulint		line);	/* in: line where requested */
 /**********************************************************************
 Releases an exclusive mode lock. */
 UNIV_INLINE
@@ -283,10 +248,8 @@ void
 rw_lock_s_lock_direct(
 /*==================*/
         rw_lock_t*   	lock  	/* in: pointer to rw-lock */
-	#ifdef UNIV_SYNC_DEBUG
 	,char*		file_name, /* in: file name where lock requested */
 	ulint		line	/* in: line where requested */
-	#endif
 );
 /**********************************************************************
 Low-level function which locks an rw-lock in x-mode when we know that it
@@ -297,10 +260,8 @@ void
 rw_lock_x_lock_direct(
 /*==================*/
         rw_lock_t*   	lock  	/* in: pointer to rw-lock */
-	#ifdef UNIV_SYNC_DEBUG
 	,char*		file_name, /* in: file name where lock requested */
 	ulint		line	/* in: line where requested */
-	#endif
 );
 /**********************************************************************
 This function is used in the insert buffer to move the ownership of an
@@ -349,6 +310,23 @@ rw_lock_get_x_lock_count(
 /*=====================*/
 				/* out: value of writer_count */
 	rw_lock_t*	lock);	/* in: rw-lock */
+/************************************************************************
+Accessor functions for rw lock. */
+UNIV_INLINE
+ulint
+rw_lock_get_waiters(
+/*================*/
+	rw_lock_t*	lock);
+UNIV_INLINE
+ulint
+rw_lock_get_writer(
+/*===============*/
+	rw_lock_t*	lock);
+UNIV_INLINE
+ulint
+rw_lock_get_reader_count(
+/*=====================*/
+	rw_lock_t*	lock);
 /**********************************************************************
 Checks if the thread has locked the rw-lock in the specified mode, with
 the pass value == 0. */
@@ -414,9 +392,6 @@ rw_lock_debug_print(
 /*================*/
 	rw_lock_debug_t*	info);	/* in: debug struct */
 
-
-#define	RW_CNAME_LEN	8
-
 /* NOTE! The structure appears here only for the compiler to know its size.
 Do not use its fields directly! The structure used in the spin lock
 implementation of a read-write lock. Several threads may have a shared lock
@@ -447,7 +422,7 @@ struct rw_lock_struct {
 	ulint	waiters;	/* This ulint is set to 1 if there are
 				waiters (readers or writers) in the global
 				wait array, waiting for this rw_lock.
-				Otherwise, = 0. */
+				Otherwise, == 0. */
 	ibool	writer_is_wait_ex;
 				/* This is TRUE if the writer field is
 				RW_LOCK_WAIT_EX; this field is located far
@@ -463,9 +438,12 @@ struct rw_lock_struct {
 				info list of the lock */
 	ulint	level;		/* Debug version: level in the global latching
 				order; default SYNC_LEVEL_NONE */
-	char	cfile_name[RW_CNAME_LEN];
-				/* File name where lock created */
+	char*	cfile_name;	/* File name where lock created */
 	ulint	cline;		/* Line where created */
+	char*	last_s_file_name;/* File name where last time s-locked */
+	char*	last_x_file_name;/* File name where last time x-locked */
+	ulint	last_s_line;	/* Line number where last time s-locked */
+	ulint	last_x_line;	/* Line number where last time x-locked */
 	ulint	magic_n;
 };
 

innobase/thr/thr0loc.c

@@ -224,5 +224,5 @@ thr_local_init(void)
 	thr_local_hash = hash_create(OS_THREAD_MAX_N + 100);
 
 	mutex_create(&thr_local_mutex);
-	mutex_set_level(&thr_local_mutex, SYNC_ANY_LATCH);
+	mutex_set_level(&thr_local_mutex, SYNC_THR_LOCAL);
 }