diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
index ae813e609932168ec2e74056fb598aca0ad65907..a5e516a40e7a359cdae8b2bf289175e971b6e6c9 100644
--- a/fs/fuse/dev.c
+++ b/fs/fuse/dev.c
@@ -165,9 +165,13 @@ static bool fuse_block_alloc(struct fuse_conn *fc, bool for_background)
 
 static void fuse_drop_waiting(struct fuse_conn *fc)
 {
-	if (fc->connected) {
-		atomic_dec(&fc->num_waiting);
-	} else if (atomic_dec_and_test(&fc->num_waiting)) {
+	/*
+	 * lockess check of fc->connected is okay, because atomic_dec_and_test()
+	 * provides a memory barrier mached with the one in fuse_wait_aborted()
+	 * to ensure no wake-up is missed.
+	 */
+	if (atomic_dec_and_test(&fc->num_waiting) &&
+	    !READ_ONCE(fc->connected)) {
 		/* wake up aborters */
 		wake_up_all(&fc->blocked_waitq);
 	}
@@ -1768,8 +1772,10 @@ static int fuse_retrieve(struct fuse_conn *fc, struct inode *inode,
 	req->in.args[1].size = total_len;
 
 	err = fuse_request_send_notify_reply(fc, req, outarg->notify_unique);
-	if (err)
+	if (err) {
 		fuse_retrieve_end(fc, req);
+		fuse_put_request(fc, req);
+	}
 
 	return err;
 }
@@ -2219,6 +2225,8 @@ EXPORT_SYMBOL_GPL(fuse_abort_conn);
 
 void fuse_wait_aborted(struct fuse_conn *fc)
 {
+	/* matches implicit memory barrier in fuse_drop_waiting() */
+	smp_mb();
 	wait_event(fc->blocked_waitq, atomic_read(&fc->num_waiting) == 0);
 }
 
diff --git a/fs/fuse/file.c b/fs/fuse/file.c
index cc2121b37bf5f7d3bb0a57398a2abfc6948ad321..b52f9baaa3e7b9c98478a8c115748ae71fb7b0e1 100644
--- a/fs/fuse/file.c
+++ b/fs/fuse/file.c
@@ -2924,10 +2924,12 @@ fuse_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
 	}
 
 	if (io->async) {
+		bool blocking = io->blocking;
+
 		fuse_aio_complete(io, ret < 0 ? ret : 0, -1);
 
 		/* we have a non-extending, async request, so return */
-		if (!io->blocking)
+		if (!blocking)
 			return -EIOCBQUEUED;
 
 		wait_for_completion(&wait);