writeback: merely wakeup flusher thread if work allocation fails for WB_SYNC_NONE
Christoph Hellwig [Fri, 11 Sep 2009 07:47:56 +0000 (09:47 +0200)]
Since it's an opportunistic writeback and not a data integrity action,
don't punt to blocking writeback. Just wakeup the thread and it will
flush old data.

Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

fs/fs-writeback.c

index 628235c..783ed44 100644 (file)
@@ -75,13 +75,6 @@ static inline void bdi_work_init(struct bdi_work *work,
        work->state = WS_USED;
 }
 
-static inline void bdi_work_init_on_stack(struct bdi_work *work,
-                                         struct writeback_control *wbc)
-{
-       bdi_work_init(work, wbc);
-       work->state |= WS_ONSTACK;
-}
-
 /**
  * writeback_in_progress - determine whether there is writeback in progress
  * @bdi: the device's backing_dev_info structure.
@@ -207,34 +200,23 @@ static struct bdi_work *bdi_alloc_work(struct writeback_control *wbc)
 
 void bdi_start_writeback(struct writeback_control *wbc)
 {
-       const bool must_wait = wbc->sync_mode == WB_SYNC_ALL;
-       struct bdi_work work_stack, *work = NULL;
-
-       if (!must_wait)
-               work = bdi_alloc_work(wbc);
+       /*
+        * WB_SYNC_NONE is opportunistic writeback. If this allocation fails,
+        * bdi_queue_work() will wake up the thread and flush old data. This
+        * should ensure some amount of progress in freeing memory.
+        */
+       if (wbc->sync_mode != WB_SYNC_ALL) {
+               struct bdi_work *w = bdi_alloc_work(wbc);
 
-       if (!work) {
-               work = &work_stack;
-               bdi_work_init_on_stack(work, wbc);
-       }
+               bdi_queue_work(wbc->bdi, w);
+       } else {
+               struct bdi_work work;
 
-       bdi_queue_work(wbc->bdi, work);
+               bdi_work_init(&work, wbc);
+               work.state |= WS_ONSTACK;
 
-       /*
-        * If the sync mode is WB_SYNC_ALL, block waiting for the work to
-        * complete. If not, we only need to wait for the work to be started,
-        * if we allocated it on-stack. We use the same mechanism, if the
-        * wait bit is set in the bdi_work struct, then threads will not
-        * clear pending until after they are done.
-        *
-        * Note that work == &work_stack if must_wait is true, so we don't
-        * need to do call_rcu() here ever, since the completion path will
-        * have done that for us.
-        */
-       if (must_wait || work == &work_stack) {
-               bdi_wait_on_work_clear(work);
-               if (work != &work_stack)
-                       call_rcu(&work->rcu_head, bdi_work_free);
+               bdi_queue_work(wbc->bdi, &work);
+               bdi_wait_on_work_clear(&work);
        }
 }