Many files:

  Multiple tablespaces for InnoDB
sql_table.cc:
  Tell explicitly that InnoDB should retrieve all columns in CHECKSUM TABLE
sql_update.cc, sql_select.cc, my_base.h:
  More descriptive flag name HA_EXTRA_RETRIEVE_ALL_COLS

include/my_base.h

@@ -121,7 +121,10 @@ enum ha_extra_function {
   HA_EXTRA_RESET_STATE,			/* Reset positions */
   HA_EXTRA_IGNORE_DUP_KEY,		/* Dup keys don't rollback everything*/
   HA_EXTRA_NO_IGNORE_DUP_KEY,
-  HA_EXTRA_DONT_USE_CURSOR_TO_UPDATE,	/* Cursor will not be used for update */
+  HA_EXTRA_RETRIEVE_ALL_COLS,		/* Instructs InnoDB to retrieve all
+					columns, not just those where
+					field->query_id is the same as the
+					current query id */
   HA_EXTRA_PREPARE_FOR_DELETE,
   HA_EXTRA_PREPARE_FOR_UPDATE,		/* Remove read cache if problems */
   HA_EXTRA_PRELOAD_BUFFER_SIZE          /* Set buffer size for preloading */

innobase/btr/btr0cur.c

@@ -957,7 +957,7 @@ btr_cur_optimistic_insert(
 	/* Now, try the insert */
 
 	*rec = page_cur_insert_rec_low(page_cursor, entry, data_size,
-								NULL, mtr);	
+								NULL, mtr);
 	if (!(*rec)) {
 		/* If the record did not fit, reorganize */
 		btr_page_reorganize(page, mtr);
@@ -1048,6 +1048,7 @@ btr_cur_pessimistic_insert(
 	ibool		dummy_inh;
 	ibool		success;
 	ulint		n_extents	= 0;
+	ulint		n_reserved;
 	
 	ut_ad(dtuple_check_typed(entry));
 
@@ -1067,7 +1068,7 @@ btr_cur_pessimistic_insert(
 	cursor->flag = BTR_CUR_BINARY;
 
 	err = btr_cur_optimistic_insert(flags, cursor, entry, rec, big_rec,
-								thr, mtr);	
+								thr, mtr);
 	if (err != DB_FAIL) {
 
 		return(err);
@@ -1090,7 +1091,7 @@ btr_cur_pessimistic_insert(
 
 		n_extents = cursor->tree_height / 16 + 3;
 
-		success = fsp_reserve_free_extents(index->space,
+		success = fsp_reserve_free_extents(&n_reserved, index->space,
 						n_extents, FSP_NORMAL, mtr);
 		if (!success) {
 			err = DB_OUT_OF_FILE_SPACE;
@@ -1112,7 +1113,7 @@ btr_cur_pessimistic_insert(
 		
 			if (n_extents > 0) {
 			        fil_space_release_free_extents(index->space,
-								n_extents);
+								n_reserved);
 			}
 			return(DB_TOO_BIG_RECORD);
 		}
@@ -1140,7 +1141,7 @@ btr_cur_pessimistic_insert(
 	err = DB_SUCCESS;
 
 	if (n_extents > 0) {
-		fil_space_release_free_extents(index->space, n_extents);
+		fil_space_release_free_extents(index->space, n_reserved);
 	}
 
 	*big_rec = big_rec_vec;
@@ -1721,6 +1722,7 @@ btr_cur_pessimistic_update(
 	ibool		was_first;
 	ibool		success;
 	ulint		n_extents	= 0;
+	ulint		n_reserved;
 	ulint*		ext_vect;
 	ulint		n_ext_vect;
 	ulint		reserve_flag;
@@ -1767,7 +1769,8 @@ btr_cur_pessimistic_update(
 			reserve_flag = FSP_NORMAL;
 		}
 		
-		success = fsp_reserve_free_extents(cursor->index->space,
+		success = fsp_reserve_free_extents(&n_reserved,
+						cursor->index->space,
 						n_extents, reserve_flag, mtr);
 		if (!success) {
 			err = DB_OUT_OF_FILE_SPACE;
@@ -1916,7 +1919,7 @@ btr_cur_pessimistic_update(
 
 	if (n_extents > 0) {
 		fil_space_release_free_extents(cursor->index->space,
-							n_extents);
+							n_reserved);
 	}
 
 	*big_rec = big_rec_vec;
@@ -2387,6 +2390,7 @@ btr_cur_pessimistic_delete(
 	rec_t*		rec;
 	dtuple_t*	node_ptr;
 	ulint		n_extents	= 0;
+	ulint		n_reserved;
 	ibool		success;
 	ibool		ret		= FALSE;
 	mem_heap_t*	heap;
@@ -2405,7 +2409,8 @@ btr_cur_pessimistic_delete(
 
 		n_extents = cursor->tree_height / 32 + 1;
 
-		success = fsp_reserve_free_extents(cursor->index->space,
+		success = fsp_reserve_free_extents(&n_reserved,
+						cursor->index->space,
 						n_extents, FSP_CLEANING, mtr);
 		if (!success) {
 			*err = DB_OUT_OF_FILE_SPACE;
@@ -2484,7 +2489,8 @@ btr_cur_pessimistic_delete(
 	}
 
 	if (n_extents > 0) {
-		fil_space_release_free_extents(cursor->index->space, n_extents);
+		fil_space_release_free_extents(cursor->index->space,
+								n_reserved);
 	}
 
 	return(ret);
@@ -3156,7 +3162,7 @@ btr_store_big_rec_extern_fields(
 	ut_ad(mtr_memo_contains(local_mtr, dict_tree_get_lock(index->tree),
 							MTR_MEMO_X_LOCK));
 	ut_ad(mtr_memo_contains(local_mtr, buf_block_align(rec),
-							MTR_MEMO_PAGE_X_FIX));	
+							MTR_MEMO_PAGE_X_FIX));
 	ut_a(index->type & DICT_CLUSTERED);
 							
 	space_id = buf_frame_get_space_id(rec);
@@ -3322,7 +3328,7 @@ btr_free_externally_stored_field(
 	ut_ad(mtr_memo_contains(local_mtr, dict_tree_get_lock(index->tree),
 							MTR_MEMO_X_LOCK));
 	ut_ad(mtr_memo_contains(local_mtr, buf_block_align(data),
-							MTR_MEMO_PAGE_X_FIX));	
+							MTR_MEMO_PAGE_X_FIX));
 	ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);
 	local_len -= BTR_EXTERN_FIELD_REF_SIZE;
 	

innobase/btr/btr0sea.c

@@ -1022,12 +1022,14 @@ btr_search_drop_page_hash_when_freed(
 	
 	mtr_start(&mtr);
 
-	/* We assume that if the caller has a latch on the page,
-	then the caller has already dropped the hash index for the page,
-	and we never get here. Therefore we can acquire the s-latch to
-	the page without fearing a deadlock. */
+	/* We assume that if the caller has a latch on the page, then the
+	caller has already dropped the hash index for the page, and we never
+	get here. Therefore we can acquire the s-latch to the page without
+	having to fear a deadlock. */
 	
-	page = buf_page_get(space, page_no, RW_S_LATCH, &mtr);
+	page = buf_page_get_gen(space, page_no, RW_S_LATCH, NULL,
+				BUF_GET_IF_IN_POOL, IB__FILE__, __LINE__,
+				&mtr);
 
 	buf_page_dbg_add_level(page, SYNC_TREE_NODE_FROM_HASH);
 	

innobase/buf/buf0buf.c

@@ -243,9 +243,10 @@ buf_calc_page_new_checksum(
 {
   	ulint checksum;
 
-	/* Since the fields FIL_PAGE_FILE_FLUSH_LSN and ..._ARCH_LOG_NO
-	are written outside the buffer pool to the first pages of data
-	files, we have to skip them in the page checksum calculation.
+        /* Since the field FIL_PAGE_FILE_FLUSH_LSN, and in versions <= 4.1.x
+        ..._ARCH_LOG_NO, are written outside the buffer pool to the first
+        pages of data files, we have to skip them in the page checksum
+        calculation.
 	We must also skip the field FIL_PAGE_SPACE_OR_CHKSUM where the
 	checksum is stored, and also the last 8 bytes of page because
 	there we store the old formula checksum. */
@@ -255,7 +256,7 @@ buf_calc_page_new_checksum(
   		   + ut_fold_binary(page + FIL_PAGE_DATA, 
 				           UNIV_PAGE_SIZE - FIL_PAGE_DATA
 				           - FIL_PAGE_END_LSN_OLD_CHKSUM);
-  	checksum = checksum & 0xFFFFFFFF;
+  	checksum = checksum & 0xFFFFFFFFUL;
 
   	return(checksum);
 }
@@ -278,7 +279,7 @@ buf_calc_page_old_checksum(
   	
   	checksum = ut_fold_binary(page, FIL_PAGE_FILE_FLUSH_LSN);
 
-  	checksum = checksum & 0xFFFFFFFF;
+  	checksum = checksum & 0xFFFFFFFFUL;
 
   	return(checksum);
 }
@@ -378,7 +379,7 @@ buf_page_print(
 
 	ut_print_timestamp(stderr);
 	fprintf(stderr,
-	"  InnoDB: Page dump in ascii and hex (%lu bytes):\n%s",
+"  InnoDB: Page dump in ascii and hex (%lu bytes):\n%s",
 					(ulint)UNIV_PAGE_SIZE, buf);
 	fprintf(stderr, "InnoDB: End of page dump\n");
 
@@ -396,11 +397,16 @@ buf_page_print(
 			mach_read_from_4(read_buf + UNIV_PAGE_SIZE
 					- FIL_PAGE_END_LSN_OLD_CHKSUM));
 	fprintf(stderr,
-	"InnoDB: Page lsn %lu %lu, low 4 bytes of lsn at page end %lu\n",
+"InnoDB: Page lsn %lu %lu, low 4 bytes of lsn at page end %lu\n"
+"InnoDB: Page number (if stored to page already) %lu,\n"
+"InnoDB: space id (if created with >= MySQL-4.1.1 and stored already) %lu\n",
 		mach_read_from_4(read_buf + FIL_PAGE_LSN),
 		mach_read_from_4(read_buf + FIL_PAGE_LSN + 4),
 		mach_read_from_4(read_buf + UNIV_PAGE_SIZE
-					- FIL_PAGE_END_LSN_OLD_CHKSUM + 4));
+					- FIL_PAGE_END_LSN_OLD_CHKSUM + 4),
+		mach_read_from_4(read_buf + FIL_PAGE_OFFSET),
+		mach_read_from_4(read_buf + FIL_PAGE_ARCH_LOG_NO_OR_SPACE_ID));
+
 	if (mach_read_from_2(read_buf + TRX_UNDO_PAGE_HDR + TRX_UNDO_PAGE_TYPE)
 	    == TRX_UNDO_INSERT) {
 	    	fprintf(stderr,
@@ -414,10 +420,7 @@ buf_page_print(
 
 	if (fil_page_get_type(read_buf) == FIL_PAGE_INDEX) {
 	    	fprintf(stderr,
-			"InnoDB: Page may be an index page ");
-
-		fprintf(stderr,
-			"where index id is %lu %lu\n",
+"InnoDB: Page may be an index page where index id is %lu %lu\n",
 			ut_dulint_get_high(btr_page_get_index_id(read_buf)),
 			ut_dulint_get_low(btr_page_get_index_id(read_buf)));
 
@@ -435,7 +438,6 @@ buf_page_print(
 						index->name);
 			}
 		}
-	  
 	} else if (fil_page_get_type(read_buf) == FIL_PAGE_INODE) {
 		fprintf(stderr, "InnoDB: Page may be an 'inode' page\n");
 	} else if (fil_page_get_type(read_buf) == FIL_PAGE_IBUF_FREE_LIST) {
@@ -581,8 +583,8 @@ buf_pool_init(
 		the window */
 
 		os_awe_map_physical_mem_to_window(buf_pool->frame_zero,
-					n_frames *
-					(UNIV_PAGE_SIZE / OS_AWE_X86_PAGE_SIZE),
+				n_frames *
+				(UNIV_PAGE_SIZE / OS_AWE_X86_PAGE_SIZE),
 					buf_pool->awe_info);
 		/*----------------------------------------*/
 	}
@@ -1554,25 +1556,35 @@ buf_page_init(
 
 /************************************************************************
 Function which inits a page for read to the buffer buf_pool. If the page is
-already in buf_pool, does nothing. Sets the io_fix flag to BUF_IO_READ and
-sets a non-recursive exclusive lock on the buffer frame. The io-handler must
-take care that the flag is cleared and the lock released later. This is one
-of the functions which perform the state transition NOT_USED => FILE_PAGE to
-a block (the other is buf_page_create). */ 
+(1) already in buf_pool, or
+(2) if we specify to read only ibuf pages and the page is not an ibuf page, or
+(3) if the space is deleted or being deleted,
+then this function does nothing.
+Sets the io_fix flag to BUF_IO_READ and sets a non-recursive exclusive lock
+on the buffer frame. The io-handler must take care that the flag is cleared
+and the lock released later. This is one of the functions which perform the
+state transition NOT_USED => FILE_PAGE to a block (the other is
+buf_page_create). */ 
 
 buf_block_t*
 buf_page_init_for_read(
 /*===================*/
-			/* out: pointer to the block or NULL */
-	ulint	mode,	/* in: BUF_READ_IBUF_PAGES_ONLY, ... */
-	ulint	space,	/* in: space id */
-	ulint	offset)	/* in: page number */
+				/* out: pointer to the block or NULL */
+	ulint*		err,	/* out: DB_SUCCESS or DB_TABLESPACE_DELETED */
+	ulint		mode,	/* in: BUF_READ_IBUF_PAGES_ONLY, ... */
+	ulint		space,	/* in: space id */
+	ib_longlong	tablespace_version,/* in: prevents reading from a wrong
+				version of the tablespace in case we have done
+				DISCARD + IMPORT */
+	ulint		offset)	/* in: page number */
 {
 	buf_block_t*	block;
 	mtr_t		mtr;
-	
+
 	ut_ad(buf_pool);
 
+	*err = DB_SUCCESS;
+
 	if (mode == BUF_READ_IBUF_PAGES_ONLY) {
 		/* It is a read-ahead within an ibuf routine */
 
@@ -1596,10 +1608,17 @@ buf_page_init_for_read(
 	ut_ad(block);
 
 	mutex_enter(&(buf_pool->mutex));
-	
-	if (NULL != buf_page_hash_get(space, offset)) {
 
-		/* The page is already in buf_pool, return */
+	if (fil_tablespace_deleted_or_being_deleted_in_mem(space,
+							tablespace_version)) {
+		*err = DB_TABLESPACE_DELETED;
+	}
+
+	if (*err == DB_TABLESPACE_DELETED
+	    || NULL != buf_page_hash_get(space, offset)) {
+
+		/* The page belongs to a space which has been deleted or is
+		being deleted, or the page is already in buf_pool, return */
 
 		mutex_exit(&(buf_pool->mutex));
 		buf_block_free(block);
@@ -1715,7 +1734,7 @@ buf_page_create(
 	/* Delete possible entries for the page from the insert buffer:
 	such can exist if the page belonged to an index which was dropped */
 
-	ibuf_merge_or_delete_for_page(NULL, space, offset);	
+	ibuf_merge_or_delete_for_page(NULL, space, offset, TRUE);
 
 	/* Flush pages from the end of the LRU list if necessary */
 	buf_flush_free_margin();
@@ -1828,7 +1847,7 @@ buf_page_io_complete(
 
 		if (!recv_no_ibuf_operations) {
 			ibuf_merge_or_delete_for_page(block->frame,
-						block->space, block->offset);
+					block->space, block->offset, TRUE);
 		}
 	}
 	
@@ -2294,7 +2313,7 @@ buf_all_freed(void)
 
 			if (!buf_flush_ready_for_replace(block)) {
 
-			    	/* printf("Page %lu %lu still fixed or dirty\n",
+			     /* printf("Page %lu %lu still fixed or dirty\n",
 			    		block->space, block->offset); */
 			    	ut_error;
 			}

innobase/buf/buf0flu.c

@@ -361,16 +361,15 @@ buf_flush_init_for_writing(
 	ulint	space,		/* in: space id */
 	ulint	page_no)	/* in: page number */
 {	
-	UT_NOT_USED(space);
-
 	/* Write the newest modification lsn to the page header and trailer */
 	mach_write_to_8(page + FIL_PAGE_LSN, newest_lsn);
 
 	mach_write_to_8(page + UNIV_PAGE_SIZE - FIL_PAGE_END_LSN_OLD_CHKSUM,
 								newest_lsn);
-	/* Write the page number */
+	/* Write the page number and the space id */
 
 	mach_write_to_4(page + FIL_PAGE_OFFSET, page_no);
+        mach_write_to_4(page + FIL_PAGE_ARCH_LOG_NO_OR_SPACE_ID, space);
 
 	/* Store the new formula checksum */
 

innobase/buf/buf0lru.c

@@ -61,6 +61,87 @@ buf_LRU_block_free_hashed_page(
 	buf_block_t*	block);	/* in: block, must contain a file page and
 				be in a state where it can be freed */
 
+/**********************************************************************
+Invalidates all pages belonging to a given tablespace when we are deleting
+the data file(s) of that tablespace. */
+
+void
+buf_LRU_invalidate_tablespace(
+/*==========================*/
+	ulint	id)	/* in: space id */
+{
+	buf_block_t*	block;
+	ulint		page_no;
+	ibool		all_freed;
+
+scan_again:
+	mutex_enter(&(buf_pool->mutex));
+	
+	all_freed = TRUE;
+	
+	block = UT_LIST_GET_LAST(buf_pool->LRU);
+
+	while (block != NULL) {
+		if (block->space == id
+		    && (block->buf_fix_count > 0 || block->io_fix != 0)) {
+
+			/* We cannot remove this page during this scan yet;
+			maybe the system is currently reading it in, or
+			flushing the modifications to the file */
+			
+			all_freed = FALSE;
+
+			goto next_page;
+		}
+
+		if (block->space == id) {
+			if (buf_debug_prints) {
+				printf(
+				"Dropping space %lu page %lu\n",
+					block->space, block->offset);
+			}
+
+			if (block->is_hashed) {
+				page_no = block->offset;
+			
+				mutex_exit(&(buf_pool->mutex));
+
+				/* Note that the following call will acquire
+				an S-latch on the page */
+
+				btr_search_drop_page_hash_when_freed(id,
+								page_no);
+				goto scan_again;
+			}
+
+			if (0 != ut_dulint_cmp(block->oldest_modification,
+							ut_dulint_zero)) {
+
+				/* Remove from the flush list of modified
+				blocks */
+				block->oldest_modification = ut_dulint_zero;
+
+				UT_LIST_REMOVE(flush_list, 
+						buf_pool->flush_list, block);
+			}
+
+			/* Remove from the LRU list */
+			buf_LRU_block_remove_hashed_page(block);
+			buf_LRU_block_free_hashed_page(block);
+		}
+next_page:
+		block = UT_LIST_GET_PREV(LRU, block);
+	}
+
+	mutex_exit(&(buf_pool->mutex));
+	
+	if (!all_freed) {
+		os_thread_sleep(20000);
+
+	        goto scan_again;
+	}
+}
+
 /**********************************************************************
 Gets the minimum LRU_position field for the blocks in an initial segment
 (determined by BUF_LRU_INITIAL_RATIO) of the LRU list. The limit is not

innobase/data/data0type.c

@@ -12,7 +12,7 @@ Created 1/16/1996 Heikki Tuuri
 #include "data0type.ic"
 #endif
 
-dtype_t		dtype_binary_val = {DATA_BINARY, 0, 0, 0};
+dtype_t		dtype_binary_val = {DATA_BINARY, 0, 0, 0, 0};
 dtype_t* 	dtype_binary 	= &dtype_binary_val;
 
 /*************************************************************************

innobase/dict/dict0boot.c

@@ -419,6 +419,4 @@ dict_create(void)
 	dict_boot();
 
 	dict_insert_initial_data();
-
-	sync_order_checks_on = TRUE;
 }

innobase/dict/dict0crea.c

@@ -264,6 +264,8 @@ dict_build_table_def_step(
 	dict_table_t*	table;
 	dict_table_t*	cluster_table;
 	dtuple_t*	row;
+	ulint		error;
+	mtr_t		mtr;
 
 	UT_NOT_USED(thr);
 	ut_ad(mutex_own(&(dict_sys->mutex)));
@@ -291,6 +293,29 @@ dict_build_table_def_step(
 		table->mix_id = dict_hdr_get_new_id(DICT_HDR_MIX_ID);
 	}
 
+	if (srv_file_per_table) {
+		/* We create a new single-table tablespace for the table.
+		We initially let it be 4 pages:
+		- page 0 is the fsp header and an extent descriptor page,
+		- page 1 is an ibuf bitmap page,
+		- page 2 is the first inode page,
+		- page 3 will contain the root of the clustered index of the
+		  table we create here. */
+
+		error = fil_create_new_single_table_tablespace(
+					&(table->space), table->name, 4);
+		if (error != DB_SUCCESS) {
+
+			return(error);
+		}
+
+		mtr_start(&mtr);
+
+		fsp_header_init(table->space, 4, &mtr);
+		
+		mtr_commit(&mtr);
+	}
+
 	row = dict_create_sys_tables_tuple(table, node->heap);
 
 	ins_node_set_new_row(node->tab_def, row);
@@ -317,7 +342,6 @@ dict_build_col_def_step(
 }
 
 #ifdef notdefined
-
 /*************************************************************************
 Creates the single index for a cluster: it contains all the columns of
 the cluster definition in the order they were defined. */
@@ -508,8 +532,8 @@ dict_create_sys_fields_tuple(
 }	
 
 /*********************************************************************
-Creates the tuple with which the index entry is searched for
-writing the index tree root page number, if such a tree is created. */
+Creates the tuple with which the index entry is searched for writing the index
+tree root page number, if such a tree is created. */
 static
 dtuple_t*
 dict_create_search_tuple(
@@ -577,10 +601,10 @@ dict_build_index_def_step(
 	
 	index->id = dict_hdr_get_new_id(DICT_HDR_INDEX_ID);
 
-	if (index->type & DICT_CLUSTERED) {
-		/* Inherit the space from the table */
-		index->space = table->space;
-	}
+	/* Inherit the space id from the table; we store all indexes of a
+	table in the same tablespace */
+
+	index->space = table->space;
 
 	index->page_no = FIL_NULL;
 	
@@ -664,6 +688,9 @@ dict_create_index_tree_step(
 
 	index->page_no = btr_create(index->type, index->space, index->id,
 									&mtr);
+	/* printf("Created a new index tree in space %lu root page %lu\n",
+					index->space, index->page_no); */
+
 	page_rec_write_index_page_no(btr_pcur_get_rec(&pcur),
 					DICT_SYS_INDEXES_PAGE_NO_FIELD,
 					index->page_no, &mtr);
@@ -712,7 +739,14 @@ dict_drop_index_tree(
 	ut_ad(len == 4);
 
 	space = mtr_read_ulint(ptr, MLOG_4BYTES, mtr);
-	
+
+	if (!fil_tablespace_exists_in_mem(space)) {
+		/* It is a single table tablespace and the .ibd file is
+		missing: do nothing */
+
+		return;
+	}
+
 	/* We free all the pages but the root page first; this operation
 	may span several mini-transactions */
 
@@ -722,6 +756,8 @@ dict_drop_index_tree(
 	we write FIL_NULL to the appropriate field in the SYS_INDEXES
 	record: this mini-transaction marks the B-tree totally freed */
 	
+	/* printf("Dropping index tree in space %lu root page %lu\n", space,
+							 root_page_no); */
 	btr_free_root(space, root_page_no, mtr);
 
 	page_rec_write_index_page_no(rec, DICT_SYS_INDEXES_PAGE_NO_FIELD,
@@ -746,7 +782,6 @@ dict_create_default_index(
 
 	dict_create_index(index, trx); 
 }
-
 #endif
 
 /*************************************************************************

innobase/dict/dict0dict.c

@@ -139,7 +139,8 @@ dict_tree_find_index_low(
 /*=====================*/
 				/* out: index */
 	dict_tree_t*	tree,	/* in: index tree */
-	rec_t*		rec);	/* in: record for which to find correct index */
+	rec_t*		rec);	/* in: record for which to find correct
+				index */
 /**************************************************************************
 Removes a foreign constraint struct from the dictionet cache. */
 static
@@ -717,7 +718,7 @@ dict_table_get_and_increment_handle_count(
 	mutex_exit(&(dict_sys->mutex));
 
 	if (table != NULL) {
-	        if (!table->stat_initialized) {
+	        if (!table->stat_initialized && !table->ibd_file_missing) {
 			dict_update_statistics(table);
 		}
 	}
@@ -869,6 +870,7 @@ dict_table_rename_in_cache(
 	ulint		fold;
 	ulint		old_size;
 	char*		name_buf;
+	ibool		success;
 	ulint		i;
 	
 	ut_ad(table);
@@ -884,6 +886,21 @@ dict_table_rename_in_cache(
 		HASH_SEARCH(name_hash, dict_sys->table_hash, fold, table2,
 				(ut_strcmp(table2->name, new_name) == 0));
 		if (table2) {
+			fprintf(stderr,
+"InnoDB: Error: dictionary cache already contains a table of name %s\n",
+	 							     new_name);
+			return(FALSE);
+		}
+	}
+
+	/* If the table is stored in a single-table tablespace, rename the
+	.ibd file */
+
+	if (table->space != 0) {
+		success = fil_rename_tablespace(table->name, table->space,
+								new_name);
+		if (!success) {
+
 			return(FALSE);
 		}
 	}
@@ -909,7 +926,6 @@ dict_table_rename_in_cache(
 	/* Add table to hash table of tables */
 	HASH_INSERT(dict_table_t, name_hash, dict_sys->table_hash, fold,
 								   table);
-	
 	dict_sys->size += (mem_heap_get_size(table->heap) - old_size);
 
 	/* Update the table_name field in indexes */
@@ -999,6 +1015,31 @@ dict_table_rename_in_cache(
 	return(TRUE);
 }
 
+/**************************************************************************
+Change the id of a table object in the dictionary cache. This is used in
+DISCARD TABLESPACE. */
+
+void
+dict_table_change_id_in_cache(
+/*==========================*/
+	dict_table_t*	table,	/* in: table object already in cache */
+	dulint		new_id)	/* in: new id to set */
+{
+	ut_ad(table);
+	ut_ad(mutex_own(&(dict_sys->mutex)));
+	ut_ad(table->magic_n == DICT_TABLE_MAGIC_N);
+
+	/* Remove the table from the hash table of id's */
+
+	HASH_DELETE(dict_table_t, id_hash, dict_sys->table_id_hash,
+					ut_fold_dulint(table->id), table);
+	table->id = new_id;
+
+	/* Add the table back to the hash table */
+	HASH_INSERT(dict_table_t, id_hash, dict_sys->table_id_hash,
+					ut_fold_dulint(table->id), table);
+}
+
 /**************************************************************************
 Removes a table object from the dictionary cache. */
 
@@ -3295,8 +3336,8 @@ dict_tree_free(
 /*===========*/
 	dict_tree_t*	tree)	/* in, own: index tree */
 {
-	ut_ad(tree);
-	ut_ad(tree->magic_n == DICT_TREE_MAGIC_N);
+	ut_a(tree);
+	ut_a(tree->magic_n == DICT_TREE_MAGIC_N);
 
 	rw_lock_free(&(tree->lock));
 	mem_free(tree);
@@ -3310,7 +3351,8 @@ dict_tree_find_index_low(
 /*=====================*/
 				/* out: index */
 	dict_tree_t*	tree,	/* in: index tree */
-	rec_t*		rec)	/* in: record for which to find correct index */
+	rec_t*		rec)	/* in: record for which to find correct
+				index */
 {
 	dict_index_t*	index;
 	dict_table_t*	table;
@@ -3348,7 +3390,8 @@ dict_tree_find_index(
 /*=================*/
 				/* out: index */
 	dict_tree_t*	tree,	/* in: index tree */
-	rec_t*		rec)	/* in: record for which to find correct index */
+	rec_t*		rec)	/* in: record for which to find correct
+				index */
 {
 	dict_index_t*	index;
 	
@@ -3438,7 +3481,8 @@ dict_tree_build_node_ptr(
 /*=====================*/
 				/* out, own: node pointer */
 	dict_tree_t*	tree,	/* in: index tree */
-	rec_t*		rec,	/* in: record for which to build node pointer */
+	rec_t*		rec,	/* in: record for which to build node
+				pointer */
 	ulint		page_no,/* in: page number to put in node pointer */
 	mem_heap_t*	heap,	/* in: memory heap where pointer created */
 	ulint           level)  /* in: level of rec in tree: 0 means leaf
@@ -3600,6 +3644,16 @@ dict_update_statistics_low(
 	ulint		size;
 	ulint		sum_of_index_sizes	= 0;
 
+	if (table->ibd_file_missing) {
+		ut_print_timestamp(stderr);
+		fprintf(stderr,
+"  InnoDB: cannot calculate statistics for table %s\n"
+"InnoDB: because the .ibd file is missing. See section 15.1 of\n"
+"InnoDB: http:/www.innodb.com/ibman.html for help\n", table->name);
+
+		return;
+	}
+
 	/* If we have set a high innodb_force_recovery level, do not calculate
 	statistics, as a badly corrupted index can cause a crash in it. */
 

innobase/dict/dict0load.c

@@ -19,6 +19,7 @@ Created 4/24/1996 Heikki Tuuri
 #include "mach0data.h"
 #include "dict0dict.h"
 #include "dict0boot.h"
+#include "srv0start.h"
 
 /************************************************************************
 Finds the first table name in the given database. */
@@ -120,8 +121,8 @@ dict_print(void)
 	rec_t*		rec;
 	byte*		field;
 	ulint		len;
-	char		table_name[10000];
 	mtr_t		mtr;
+	char		table_name[10000];
 	
 	mutex_enter(&(dict_sys->mutex));
 
@@ -185,6 +186,100 @@ dict_print(void)
 	goto loop;
 }
 
+/************************************************************************
+In a crash recovery we already have all the tablespace objects created.
+This function compares the space id information in the InnoDB data dictionary
+to what we already read with fil_load_single_table_tablespaces().
+In a normal startup we just scan the biggest space id, and store it to
+fil_system. */
+
+void
+dict_check_tablespaces_or_store_max_id(
+/*===================================*/
+	ibool	in_crash_recovery)	/* in: are we doing a crash recovery */
+{
+	dict_table_t*	sys_tables;
+	dict_index_t*	sys_index;
+	btr_pcur_t	pcur;
+	rec_t*		rec;
+	byte*		field;
+	ulint		len;
+	ulint		space_id;
+	ulint		max_space_id	= 0;
+	mtr_t		mtr;
+	char		name[OS_FILE_MAX_PATH];
+	
+	mutex_enter(&(dict_sys->mutex));
+
+	mtr_start(&mtr);
+
+	sys_tables = dict_table_get_low((char *) "SYS_TABLES");
+	sys_index = UT_LIST_GET_FIRST(sys_tables->indexes);
+
+	btr_pcur_open_at_index_side(TRUE, sys_index, BTR_SEARCH_LEAF, &pcur,
+								TRUE, &mtr);
+loop:
+	btr_pcur_move_to_next_user_rec(&pcur, &mtr);
+
+	rec = btr_pcur_get_rec(&pcur);
+
+	if (!btr_pcur_is_on_user_rec(&pcur, &mtr)) {
+		/* end of index */
+
+		btr_pcur_close(&pcur);
+		mtr_commit(&mtr);
+		
+		/* We must make the tablespace cache aware of the biggest
+		known space id */
+
+		/* printf("Biggest space id in data dictionary %lu\n",
+							    max_space_id); */
+		fil_set_max_space_id_if_bigger(max_space_id);
+
+		mutex_exit(&(dict_sys->mutex));
+
+		return;
+	}	
+
+	field = rec_get_nth_field(rec, 0, &len);
+
+	if (!rec_get_deleted_flag(rec)) {
+
+		/* We found one */
+
+		ut_a(len < OS_FILE_MAX_PATH - 10);
+		ut_memcpy(name, field, len);
+		name[len] = '\0';
+
+		field = rec_get_nth_field(rec, 9, &len);
+		ut_a(len == 4);
+			
+		space_id = mach_read_from_4(field);
+
+		btr_pcur_store_position(&pcur, &mtr);
+
+		mtr_commit(&mtr);
+		
+		if (space_id != 0 && in_crash_recovery) {
+			/* Check that the tablespace (the .ibd file) really
+			exists; print a warning to the .err log if not */
+			
+			fil_space_for_table_exists_in_mem(space_id, name,
+								TRUE, TRUE);
+		}
+		
+		if (space_id > max_space_id) {
+			max_space_id = space_id;
+		}
+
+		mtr_start(&mtr);
+
+		btr_pcur_restore_position(BTR_SEARCH_LEAF, &pcur, &mtr);
+	}
+
+	goto loop;
+}
+
 /************************************************************************
 Loads definitions for table columns. */
 static
@@ -359,13 +454,13 @@ dict_load_fields(
 
 		pos_and_prefix_len = mach_read_from_4(field);
 
-		ut_a((pos_and_prefix_len & 0xFFFF) == i
-		     || (pos_and_prefix_len & 0xFFFF0000) == (i << 16));
+		ut_a((pos_and_prefix_len & 0xFFFFUL) == i
+		     || (pos_and_prefix_len & 0xFFFF0000UL) == (i << 16));
 
 		if ((i == 0 && pos_and_prefix_len > 0)
-		    || (pos_and_prefix_len & 0xFFFF0000) > 0) {
+		    || (pos_and_prefix_len & 0xFFFF0000UL) > 0) {
 
-		        prefix_len = pos_and_prefix_len & 0xFFFF;
+		        prefix_len = pos_and_prefix_len & 0xFFFFUL;
 		} else {
 		        prefix_len = 0;
 		}
@@ -540,8 +635,8 @@ dict_load_indexes(
 			    && (0 == ut_memcmp(name_buf, (char*) "ID_IND",
 							name_len))))) {
 
-			/* The index was created in memory already in
-			booting */
+			/* The index was created in memory already at booting
+			of the database server */
 		} else {
  			index = dict_mem_index_create(table->name, name_buf,
 						space, type, n_fields);
@@ -572,9 +667,14 @@ dictionary cache. */
 dict_table_t*
 dict_load_table(
 /*============*/
-			/* out: table, NULL if does not exist */
-	char*	name)	/* in: table name */
+			/* out: table, NULL if does not exist; if the table is
+			stored in an .ibd file, but the file does not exist,
+			then we set the ibd_file_missing flag TRUE in the table
+			object we return */
+	char*	name)	/* in: table name in the databasename/tablename
+			format */
 {
+	ibool		ibd_file_missing	= FALSE;
 	dict_table_t*	table;
 	dict_table_t*	sys_tables;
 	btr_pcur_t	pcur;
@@ -641,6 +741,23 @@ dict_load_table(
 	field = rec_get_nth_field(rec, 9, &len);
 	space = mach_read_from_4(field);
 
+	/* Check if the tablespace exists and has the right name */
+	if (space != 0) {
+		if (fil_space_for_table_exists_in_mem(space, name, FALSE,
+								   FALSE)) {
+			/* Ok; (if we did a crash recovery then the tablespace
+			can already be in the memory cache) */
+		} else {
+			/* Try to open the tablespace */
+			if (!fil_open_single_table_tablespace(space, name)) {
+				/* We failed to find a sensible tablespace
+				file */
+
+				ibd_file_missing = TRUE;
+			}
+		}
+	}
+
 	ut_a(0 == ut_strcmp((char *) "N_COLS",
 		dict_field_get_col(
 		dict_index_get_nth_field(
@@ -651,6 +768,8 @@ dict_load_table(
 
 	table = dict_mem_table_create(name, space, n_cols);
 
+	table->ibd_file_missing = ibd_file_missing;
+
 	ut_a(0 == ut_strcmp((char *) "ID",
 		dict_field_get_col(
 		dict_index_get_nth_field(
@@ -1003,7 +1122,7 @@ dict_load_foreign(
 	/* We store the type to the bits 24-31 of n_fields */
 	
 	foreign->type = foreign->n_fields >> 24;
-	foreign->n_fields = foreign->n_fields & 0xFFFFFF;
+	foreign->n_fields = foreign->n_fields & 0xFFFFFFUL;
 	
 	foreign->id = mem_heap_alloc(foreign->heap, ut_strlen(id) + 1);
 				

innobase/dict/dict0mem.c

@@ -56,6 +56,8 @@ dict_mem_table_create(
 	table->type = DICT_TABLE_ORDINARY;
 	table->name = str;
 	table->space = space;
+	table->ibd_file_missing = FALSE;
+	table->tablespace_discarded = FALSE;
 	table->n_def = 0;
 	table->n_cols = n_cols + DATA_N_SYS_COLS;
 	table->mem_fix = 0;

innobase/ha/ha0ha.c

@@ -294,10 +294,10 @@ ha_print_info(
 {
 	hash_cell_t*	cell;
 /*
-	ha_node_t*	node;
-	ulint		len	= 0;
-	ulint		max_len	= 0;
-	ulint		nodes	= 0;
+        ha_node_t*      node;
+        ulint           len     = 0;
+        ulint           max_len = 0;
+        ulint           nodes   = 0;
 */
 	ulint		cells	= 0;
 	ulint		n_bufs;

innobase/include/btr0btr.ic

@@ -189,6 +189,7 @@ btr_node_ptr_get_child_page_no(
 	ulint	n_fields;
 	byte*	field;
 	ulint	len;
+	ulint	page_no;
 
 	n_fields = rec_get_n_fields(rec);
 
@@ -197,7 +198,16 @@ btr_node_ptr_get_child_page_no(
 
 	ut_ad(len == 4);
 	
-	return(mach_read_from_4(field));
+	page_no = mach_read_from_4(field);
+
+	if (page_no == 0) {
+		fprintf(stderr,
+"InnoDB: a nonsensical page number 0 in a node ptr record at offset %lu\n",
+		       (ulint)(rec - buf_frame_align(rec)));
+		buf_page_print(buf_frame_align(rec));
+	}
+
+	return(page_no);
 }
 
 /******************************************************************

innobase/include/btr0pcur.ic

@@ -564,7 +564,7 @@ btr_pcur_open_at_index_side(
 	}
 
 	btr_cur_open_at_index_side(from_left, index, latch_mode,
-					btr_pcur_get_btr_cur(pcur), mtr);	
+					btr_pcur_get_btr_cur(pcur), mtr);
 	pcur->pos_state = BTR_PCUR_IS_POSITIONED;
 
 	pcur->old_stored = BTR_PCUR_OLD_NOT_STORED;

innobase/include/buf0buf.h

@@ -626,19 +626,27 @@ buf_pool_get_nth_block(
 	ulint		i);	/* in: index of the block */
 /************************************************************************
 Function which inits a page for read to the buffer buf_pool. If the page is
-already in buf_pool, does nothing. Sets the io_fix flag to BUF_IO_READ and
-sets a non-recursive exclusive lock on the buffer frame. The io-handler must
-take care that the flag is cleared and the lock released later. This is one
-of the functions which perform the state transition NOT_USED => FILE_PAGE to
-a block (the other is buf_page_create). */ 
+(1) already in buf_pool, or
+(2) if we specify to read only ibuf pages and the page is not an ibuf page, or
+(3) if the space is deleted or being deleted,
+then this function does nothing.
+Sets the io_fix flag to BUF_IO_READ and sets a non-recursive exclusive lock
+on the buffer frame. The io-handler must take care that the flag is cleared
+and the lock released later. This is one of the functions which perform the
+state transition NOT_USED => FILE_PAGE to a block (the other is
+buf_page_create). */ 
 
 buf_block_t*
 buf_page_init_for_read(
 /*===================*/
-			/* out: pointer to the block */
-	ulint	mode,	/* in: BUF_READ_IBUF_PAGES_ONLY, ... */
-	ulint	space,	/* in: space id */
-	ulint	offset);/* in: page number */
+				/* out: pointer to the block or NULL */
+	ulint*		err,	/* out: DB_SUCCESS or DB_TABLESPACE_DELETED */
+	ulint		mode,	/* in: BUF_READ_IBUF_PAGES_ONLY, ... */
+	ulint		space,	/* in: space id */
+	ib_longlong	tablespace_version,/* in: prevents reading from a wrong
+				version of the tablespace in case we have done
+				DISCARD + IMPORT */
+	ulint		offset);/* in: page number */
 /************************************************************************
 Completes an asynchronous read or write request of a file page to or from
 the buffer pool. */

innobase/include/buf0lru.h

@@ -36,6 +36,16 @@ These are low-level functions
 
 #define BUF_LRU_FREE_SEARCH_LEN		(5 + 2 * BUF_READ_AHEAD_AREA)
 
+/**********************************************************************
+Invalidates all pages belonging to a given tablespace when we are deleting
+the data file(s) of that tablespace. A PROBLEM: if readahead is being started,
+what guarantees that it will not try to read in pages after this operation has
+completed? */
+
+void
+buf_LRU_invalidate_tablespace(
+/*==========================*/
+	ulint	id);	/* in: space id */
 /**********************************************************************
 Gets the minimum LRU_position field for the blocks in an initial segment
 (determined by BUF_LRU_INITIAL_RATIO) of the LRU list. The limit is not

innobase/include/buf0rea.h

@@ -59,7 +59,7 @@ buf_read_ahead_linear(
 			must want access to this page (see NOTE 3 above) */
 /************************************************************************
 Issues read requests for pages which the ibuf module wants to read in, in
-order to contract insert buffer trees. Technically, this function is like
+order to contract the insert buffer tree. Technically, this function is like
 a read-ahead function. */
 
 void
@@ -68,9 +68,14 @@ buf_read_ibuf_merge_pages(
 	ibool	sync,		/* in: TRUE if the caller wants this function
 				to wait for the highest address page to get
 				read in, before this function returns */
-	ulint	space,		/* in: space id */
-	ulint*	page_nos,	/* in: array of page numbers to read, with
-				the highest page number last in the array */
+	ulint*	space_ids,	/* in: array of space ids */
+	ib_longlong* space_versions,/* in: the spaces must have this version
+				number (timestamp), otherwise we discard the
+				read; we use this to cancel reads if
+				DISCARD + IMPORT may have changed the
+				tablespace size */
+	ulint*	page_nos,	/* in: array of page numbers to read, with the
+				highest page number the last in the array */
 	ulint	n_stored);	/* in: number of page numbers in the array */
 /************************************************************************
 Issues read requests for pages which recovery wants to read in. */

innobase/include/db0err.h

@@ -48,6 +48,11 @@ Created 5/24/1996 Heikki Tuuri
 					from a table failed */
 #define DB_NO_SAVEPOINT		42	/* no savepoint exists with the given
 					name */
+#define	DB_TABLESPACE_ALREADY_EXISTS 43 /* we cannot create a new single-table
+				        tablespace because a file of the same
+					name already exists */
+#define DB_TABLESPACE_DELETED	44	/* tablespace does not exist or is
+					being dropped right now */
 					
 /* The following are partial failure codes */
 #define DB_FAIL 		1000

innobase/include/dict0boot.h

@@ -93,7 +93,7 @@ dict_create(void);
 					indexes; ibuf tables and indexes are
 					assigned as the id the number
 					DICT_IBUF_ID_MIN plus the space id */
-#define DICT_IBUF_ID_MIN	ut_dulint_create(0xFFFFFFFF, 0)
+#define DICT_IBUF_ID_MIN	ut_dulint_create(0xFFFFFFFFUL, 0)
 					
 /* The offset of the dictionary header on the page */
 #define	DICT_HDR		FSEG_PAGE_DATA

innobase/include/dict0mem.h

@@ -309,6 +309,13 @@ struct dict_table_struct{
 	char*		name;	/* table name */
 	ulint		space;	/* space where the clustered index of the
 				table is placed */
+	ibool		ibd_file_missing;/* TRUE if this is in a single-table
+				tablespace and the .ibd file is missing; then
+				we must return in ha_innodb.cc an error if the
+				user tries to query such an orphaned table */
+	ibool		tablespace_discarded;/* this flag is set TRUE when the
+				user calls DISCARD TABLESPACE on this table,
+				and reset to FALSE in IMPORT TABLESPACE */
 	hash_node_t	name_hash; /* hash chain node */
 	hash_node_t	id_hash; /* hash chain node */
 	ulint		n_def;	/* number of columns defined so far */

innobase/include/dyn0dyn.ic

@@ -7,7 +7,7 @@ Created 2/5/1996 Heikki Tuuri
 *******************************************************/
 
 #define DYN_BLOCK_MAGIC_N	375767
-#define DYN_BLOCK_FULL_FLAG	0x1000000
+#define DYN_BLOCK_FULL_FLAG	0x1000000UL
 
 /****************************************************************
 Adds a new block to a dyn array. */