MDEV-7424: InnoDB: Assertion failure in thread 139901753345792

  in file buf0mtflu.cc line 439

  Analysis: At shutdown multi-threaded flush sends a exit work
  items to all mtflush threads. We wait until the work queue is
  empty. However, as we did not hold the mutex, some other thread
  could also put work-items to work queue.

  Fix: Take mutex before adding exit work items to work queue and
  wait until all work-items are really processed. Release
  mutex after we have marked that multi-threaded flush is not
  anymore active.

  Fix test failure on innodb_bug12902967 caused by unnecessary
  info output on xtradb/buf/buf0mtflush.cc.

   Do not try to enable atomic writes if the file type
   is not OS_DATA_FILE. Atomic writes are unnecessary
   for log files. If we try to enable atomic writes
   to log writes that are stored to media supporting
   atomic writes we will end up problems later.

3 files changed, 22 insertions, 15 deletions

diff --git a/storage/innobase/buf/buf0mtflu.cc b/storage/innobase/buf/buf0mtflu.cc
index c14f9048ae5..e990ba785e7 100644
--- a/storage/innobase/buf/buf0mtflu.cc
+++ b/storage/innobase/buf/buf0mtflu.cc
@@ -1,7 +1,7 @@
 /*****************************************************************************
 
 Copyright (C) 2013, 2014, Fusion-io. All Rights Reserved.
-Copyright (C) 2013, 2014, SkySQL Ab. All Rights Reserved.
+Copyright (C) 2013, 2015, MariaDB Corporation. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -202,6 +202,7 @@ buf_mtflu_flush_pool_instance(
 		return 0;
 	}
 
+	memset(&n, 0, sizeof(flush_counters_t));
 
     	if (work_item->wr.flush_type == BUF_FLUSH_LRU) {
         	/* srv_LRU_scan_depth can be arbitrarily large value.
@@ -379,9 +380,6 @@ buf_mtflu_io_thread_exit(void)
 
 	mtflush_io->gwt_status = WTHR_KILL_IT;
 
-	fprintf(stderr, "InnoDB: [Note]: Signal mtflush_io_threads to exit [%lu]\n",
-		srv_mtflush_threads);
-
 	/* This lock is to safequard against timing bug: flush request take
 	this mutex before sending work items to be processed by flush
 	threads. Inside flush thread we assume that work queue contains only
@@ -409,6 +407,9 @@ buf_mtflu_io_thread_exit(void)
 			mtflush_io->wheap);
 	}
 
+	/* Requests sent */
+	os_fast_mutex_unlock(&mtflush_mtx);
+
 	/* Wait until all work items on a work queue are processed */
 	while(!ib_wqueue_is_empty(mtflush_io->wq)) {
 		/* Wait */
@@ -433,11 +434,14 @@ buf_mtflu_io_thread_exit(void)
 	/* Wait about 1/2 sec to allow threads really exit */
 	os_thread_sleep(MT_WAIT_IN_USECS);
 
+	/* Make sure that work queue is empty */
 	while(!ib_wqueue_is_empty(mtflush_io->wq))
 	{
 		ib_wqueue_nowait(mtflush_io->wq);
 	}
 
+	os_fast_mutex_lock(&mtflush_mtx);
+
 	ut_a(ib_wqueue_is_empty(mtflush_io->wq));
 	ut_a(ib_wqueue_is_empty(mtflush_io->wr_cq));
 	ut_a(ib_wqueue_is_empty(mtflush_io->rd_cq));
@@ -447,14 +451,18 @@ buf_mtflu_io_thread_exit(void)
 	ib_wqueue_free(mtflush_io->wr_cq);
 	ib_wqueue_free(mtflush_io->rd_cq);
 
-	/* Requests sent */
-	os_fast_mutex_unlock(&mtflush_mtx);
-	os_fast_mutex_free(&mtflush_mtx);
-	os_fast_mutex_free(&mtflush_io->thread_global_mtx);
+	mtflush_io->wq = NULL;
+	mtflush_io->wr_cq  = NULL;
+	mtflush_io->rd_cq = NULL;
+	mtflush_work_initialized = 0;
 
 	/* Free heap */
 	mem_heap_free(mtflush_io->wheap);
 	mem_heap_free(mtflush_io->rheap);
+
+	os_fast_mutex_unlock(&mtflush_mtx);
+	os_fast_mutex_free(&mtflush_mtx);
+	os_fast_mutex_free(&mtflush_io->thread_global_mtx);
 }
 
 /******************************************************************//**
@@ -541,6 +549,10 @@ buf_mtflu_flush_work_items(
 	mem_heap_t* reply_heap;
 	wrk_t work_item[MTFLUSH_MAX_WORKER];
 
+	if (mtflush_ctx->gwt_status == WTHR_KILL_IT) {
+		return 0;
+	}
+
 	/* Allocate heap where all work items used and queue
 	node items areallocated */
 	work_heap = mem_heap_create(0);
diff --git a/storage/innobase/os/os0file.cc b/storage/innobase/os/os0file.cc
index d0025b829ec..1f013a622ff 100644
--- a/storage/innobase/os/os0file.cc
+++ b/storage/innobase/os/os0file.cc
@@ -1919,7 +1919,7 @@ os_file_create_func(
 	try to set atomic writes and if that fails when creating a new
 	table, produce a error. If atomic writes are used on existing
 	file, ignore error and use traditional writes for that file */
-	if (file != INVALID_HANDLE_VALUE
+	if (file != INVALID_HANDLE_VALUE && type == OS_DATA_FILE
 	    && (awrites == ATOMIC_WRITES_ON ||
 		(srv_use_atomic_writes && awrites == ATOMIC_WRITES_DEFAULT))
 	    && !os_file_set_atomic_writes(name, file)) {
@@ -2065,7 +2065,7 @@ os_file_create_func(
 	try to set atomic writes and if that fails when creating a new
 	table, produce a error. If atomic writes are used on existing
 	file, ignore error and use traditional writes for that file */
-	if (file != -1
+	if (file != -1 && type == OS_DATA_FILE
 	    && (awrites == ATOMIC_WRITES_ON ||
 		(srv_use_atomic_writes && awrites == ATOMIC_WRITES_DEFAULT))
 	    && !os_file_set_atomic_writes(name, file)) {
diff --git a/storage/innobase/srv/srv0start.cc b/storage/innobase/srv/srv0start.cc
index 2692636dcb5..fe16f40a9d6 100644
--- a/storage/innobase/srv/srv0start.cc
+++ b/storage/innobase/srv/srv0start.cc
@@ -2741,11 +2741,6 @@ files_checked:
 				srv_mtflush_threads,
 				mtflush_ctx,
 				(thread_ids + 6 + 32));
-
-#if UNIV_DEBUG
-			fprintf(stderr, "InnoDB: Note: %s:%d buf-pool-instances:%lu mtflush_threads %lu\n",
-				__FILE__, __LINE__, srv_buf_pool_instances, srv_mtflush_threads);
-#endif
 		}
 
 		os_thread_create(buf_flush_page_cleaner_thread, NULL, NULL);