MDEV-16124 fil_rename_tablespace() times out and crashes server during table-rebuilding ALTER TABLE

InnoDB insisted on closing the file handle before renaming a file. Renaming a file should never be a problem on POSIX systems. Also on Windows it should work if the file was opened in FILE_SHARE_DELETE mode. fil_space_t::stop_ios: Remove. We no longer need to stop file access during rename operations. fil_mutex_enter_and_prepare_for_io(): Remove the wait for stop_ios. fil_rename_tablespace(): Remove the retry logic; do not close the file handle. Remove the unused fault injection that was added along with the DATA DIRECTORY functionality (MySQL WL#5980). os_file_create_simple_func(), os_file_create_func(), os_file_create_simple_no_error_handling_func(): Include FILE_SHARE_DELETE in the share_mode. (We will still prevent multiple InnoDB instances from using the same files by not setting FILE_SHARE_WRITE.)

Author: dr-m

storage/innobase/fil/fil0fil.cc

@@ -845,7 +845,6 @@ fil_mutex_enter_and_prepare_for_io(
 ibool	success;
 ibool	print_info	= FALSE;
 ulint	count	= 0;
-ulint	count2	= 0;
 
 retry:
 mutex_enter(&fil_system->mutex);
@@ -862,46 +861,6 @@ fil_mutex_enter_and_prepare_for_io(
 
 space = fil_space_get_by_id(space_id);
 
-if (space != NULL && space->stop_ios) {
-/* We are going to do a rename file and want to stop new i/o's
-for a while */
-
-if (count2 > 20000) {
-fputs("InnoDB: Warning: tablespace ", stderr);
-ut_print_filename(stderr, space->name);
-fprintf(stderr,
-" has i/o ops stopped for a long time %lu\n",
-(ulong) count2);
-}
-
-mutex_exit(&fil_system->mutex);
-
-#ifndef UNIV_HOTBACKUP
-
-/* Wake the i/o-handler threads to make sure pending
-i/o's are performed */
-os_aio_simulated_wake_handler_threads();
-
-/* The sleep here is just to give IO helper threads a
-bit of time to do some work. It is not required that
-all IO related to the tablespace being renamed must
-be flushed here as we do fil_flush() in
-fil_rename_tablespace() as well. */
-os_thread_sleep(20000);
-
-#endif /* UNIV_HOTBACKUP */
-
-/* Flush tablespaces so that we can close modified
-files in the LRU list */
-fil_flush_file_spaces(FIL_TABLESPACE);
-
-os_thread_sleep(20000);
-
-count2++;
-
-goto retry;
-}
-
 if (fil_system->n_open < fil_system->max_n_open) {
 
 return;
@@ -2898,33 +2857,17 @@ fil_rename_tablespace(
 ibool	success;
 fil_space_t*	space;
 fil_node_t*	node;
-ulint	count	= 0;
 char*	new_path;
 char*	old_name;
 char*	old_path;
 const char*	not_given	= "(name not specified)";
 
 ut_a(id != 0);
 
-retry:
-count++;
-
-if (!(count % 1000)) {
-ut_print_timestamp(stderr);
-fputs(" InnoDB: Warning: problems renaming ", stderr);
-ut_print_filename(stderr,
- old_name_in ? old_name_in : not_given);
-fputs(" to ", stderr);
-ut_print_filename(stderr, new_name);
-fprintf(stderr, ", %lu iterations\n", (ulong) count);
-}
-
 mutex_enter(&fil_system->mutex);
 
 space = fil_space_get_by_id(id);
 
-DBUG_EXECUTE_IF("fil_rename_tablespace_failure_1", space = NULL; );
-
 if (space == NULL) {
 ib_logf(IB_LOG_LEVEL_ERROR,
 "Cannot find space id %lu in the tablespace "
@@ -2936,54 +2879,11 @@ fil_rename_tablespace(
 return(FALSE);
 }
 
-if (count > 25000) {
-space->stop_ios = FALSE;
-mutex_exit(&fil_system->mutex);
-
-return(FALSE);
-}
-
-/* We temporarily close the .ibd file because we do not trust that
-operating systems can rename an open file. For the closing we have to
-wait until there are no pending i/o's or flushes on the file. */
-
-space->stop_ios = TRUE;
-
 /* The following code must change when InnoDB supports
 multiple datafiles per tablespace. */
 ut_a(UT_LIST_GET_LEN(space->chain) == 1);
 node = UT_LIST_GET_FIRST(space->chain);
 
-if (node->n_pending > 0
- || node->n_pending_flushes > 0
- || node->being_extended) {
-/* There are pending i/o's or flushes or the file is
-currently being extended, sleep for a while and
-retry */
-
-mutex_exit(&fil_system->mutex);
-
-os_thread_sleep(20000);
-
-goto retry;
-
-} else if (node->modification_counter > node->flush_counter) {
-/* Flush the space */
-
-mutex_exit(&fil_system->mutex);
-
-os_thread_sleep(20000);
-
-fil_flush(id);
-
-goto retry;
-
-} else if (node->open) {
-/* Close the file */
-
-fil_node_close_file(node, fil_system);
-}
-
 /* Check that the old name in the space is right */
 
 if (old_name_in) {
@@ -3002,17 +2902,9 @@ fil_rename_tablespace(
 space, node, new_name, new_path);
 
 if (success) {
-
-DBUG_EXECUTE_IF("fil_rename_tablespace_failure_2",
-goto skip_second_rename; );
-
 success = os_file_rename(
 innodb_file_data_key, old_path, new_path);
 
-DBUG_EXECUTE_IF("fil_rename_tablespace_failure_2",
-skip_second_rename:
-success = FALSE; );
-
 if (!success) {
 /* We have to revert the changes we made
 to the tablespace memory cache */
@@ -3022,8 +2914,6 @@ fil_rename_tablespace(
 }
 }
 
-space->stop_ios = FALSE;
-
 mutex_exit(&fil_system->mutex);
 
 #ifndef UNIV_HOTBACKUP

storage/innobase/include/fil0fil.h

@@ -256,10 +256,6 @@ struct fil_space_t {
 the space corresponds to a table in the InnoDB
 data dictionary; so we can print a warning of
 orphaned tablespaces */
-ibool	stop_ios;/*!< TRUE if we want to rename the
-.ibd file of tablespace and want to
-stop temporarily posting of new i/o
-requests on the file */
 ibool	stop_new_ops;
 /*!< we set this TRUE when we start
 deleting a single-table tablespace.

storage/innobase/os/os0file.cc

@@ -2,7 +2,7 @@
 
 Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2009, Percona Inc.
-Copyright (c) 2013, 2017, MariaDB Corporation.
+Copyright (c) 2013, 2018, MariaDB Corporation.
 
 Portions of this file contain modifications contributed and copyrighted
 by Percona Inc.. Those modifications are
@@ -1186,7 +1186,8 @@ os_file_create_simple_func(
 /* Use default security attributes and no template file. */
 
 file = CreateFile(
-(LPCTSTR) name, access, FILE_SHARE_READ, NULL,
+(LPCTSTR) name, access,
+FILE_SHARE_READ | FILE_SHARE_DELETE, NULL,
 create_flag, attributes, NULL);
 
 if (file == INVALID_HANDLE_VALUE) {
@@ -1314,7 +1315,7 @@ os_file_create_simple_no_error_handling_func(
 DWORD	access;
 DWORD	create_flag;
 DWORD	attributes	= 0;
-DWORD	share_mode	= FILE_SHARE_READ;
+DWORD	share_mode	= FILE_SHARE_READ | FILE_SHARE_DELETE;
 ut_a(name);
 
 ut_a(!(create_mode & OS_FILE_ON_ERROR_SILENT));
@@ -1554,7 +1555,7 @@ os_file_create_func(
 
 #ifdef __WIN__
 DWORD	create_flag;
-DWORD	share_mode	= FILE_SHARE_READ;
+DWORD	share_mode	= FILE_SHARE_READ | FILE_SHARE_DELETE;
 
 on_error_no_exit = create_mode & OS_FILE_ON_ERROR_NO_EXIT
 ? TRUE : FALSE;

storage/xtradb/fil/fil0fil.cc

@@ -862,7 +862,6 @@ fil_mutex_enter_and_prepare_for_io(
 ibool	success;
 ibool	print_info	= FALSE;
 ulint	count	= 0;
-ulint	count2	= 0;
 
 retry:
 mutex_enter(&fil_system->mutex);
@@ -879,46 +878,6 @@ fil_mutex_enter_and_prepare_for_io(
 
 space = fil_space_get_by_id(space_id);
 
-if (space != NULL && space->stop_ios) {
-/* We are going to do a rename file and want to stop new i/o's
-for a while */
-
-if (count2 > 20000) {
-fputs("InnoDB: Warning: tablespace ", stderr);
-ut_print_filename(stderr, space->name);
-fprintf(stderr,
-" has i/o ops stopped for a long time %lu\n",
-(ulong) count2);
-}
-
-mutex_exit(&fil_system->mutex);
-
-#ifndef UNIV_HOTBACKUP
-
-/* Wake the i/o-handler threads to make sure pending
-i/o's are performed */
-os_aio_simulated_wake_handler_threads();
-
-/* The sleep here is just to give IO helper threads a
-bit of time to do some work. It is not required that
-all IO related to the tablespace being renamed must
-be flushed here as we do fil_flush() in
-fil_rename_tablespace() as well. */
-os_thread_sleep(20000);
-
-#endif /* UNIV_HOTBACKUP */
-
-/* Flush tablespaces so that we can close modified
-files in the LRU list */
-fil_flush_file_spaces(FIL_TABLESPACE);
-
-os_thread_sleep(20000);
-
-count2++;
-
-goto retry;
-}
-
 if (fil_system->n_open < fil_system->max_n_open) {
 
 return;
@@ -2950,33 +2909,17 @@ fil_rename_tablespace(
 ibool	success;
 fil_space_t*	space;
 fil_node_t*	node;
-ulint	count	= 0;
 char*	new_path;
 char*	old_name;
 char*	old_path;
 const char*	not_given	= "(name not specified)";
 
 ut_a(id != 0);
 
-retry:
-count++;
-
-if (!(count % 1000)) {
-ut_print_timestamp(stderr);
-fputs(" InnoDB: Warning: problems renaming ", stderr);
-ut_print_filename(stderr,
- old_name_in ? old_name_in : not_given);
-fputs(" to ", stderr);
-ut_print_filename(stderr, new_name);
-fprintf(stderr, ", %lu iterations\n", (ulong) count);
-}
-
 mutex_enter(&fil_system->mutex);
 
 space = fil_space_get_by_id(id);
 
-DBUG_EXECUTE_IF("fil_rename_tablespace_failure_1", space = NULL; );
-
 if (space == NULL) {
 ib_logf(IB_LOG_LEVEL_ERROR,
 "Cannot find space id %lu in the tablespace "
@@ -2988,54 +2931,11 @@ fil_rename_tablespace(
 return(FALSE);
 }
 
-if (count > 25000) {
-space->stop_ios = FALSE;
-mutex_exit(&fil_system->mutex);
-
-return(FALSE);
-}
-
-/* We temporarily close the .ibd file because we do not trust that
-operating systems can rename an open file. For the closing we have to
-wait until there are no pending i/o's or flushes on the file. */
-
-space->stop_ios = TRUE;
-
 /* The following code must change when InnoDB supports
 multiple datafiles per tablespace. */
 ut_a(UT_LIST_GET_LEN(space->chain) == 1);
 node = UT_LIST_GET_FIRST(space->chain);
 
-if (node->n_pending > 0
- || node->n_pending_flushes > 0
- || node->being_extended) {
-/* There are pending i/o's or flushes or the file is
-currently being extended, sleep for a while and
-retry */
-
-mutex_exit(&fil_system->mutex);
-
-os_thread_sleep(20000);
-
-goto retry;
-
-} else if (node->modification_counter > node->flush_counter) {
-/* Flush the space */
-
-mutex_exit(&fil_system->mutex);
-
-os_thread_sleep(20000);
-
-fil_flush(id);
-
-goto retry;
-
-} else if (node->open) {
-/* Close the file */
-
-fil_node_close_file(node, fil_system);
-}
-
 /* Check that the old name in the space is right */
 
 if (old_name_in) {
@@ -3054,17 +2954,9 @@ fil_rename_tablespace(
 space, node, new_name, new_path);
 
 if (success) {
-
-DBUG_EXECUTE_IF("fil_rename_tablespace_failure_2",
-goto skip_second_rename; );
-
 success = os_file_rename(
 innodb_file_data_key, old_path, new_path);
 
-DBUG_EXECUTE_IF("fil_rename_tablespace_failure_2",
-skip_second_rename:
-success = FALSE; );
-
 if (!success) {
 /* We have to revert the changes we made
 to the tablespace memory cache */
@@ -3074,8 +2966,6 @@ fil_rename_tablespace(
 }
 }
 
-space->stop_ios = FALSE;
-
 mutex_exit(&fil_system->mutex);
 
 #ifndef UNIV_HOTBACKUP