diff --git a/block/fops.c b/block/fops.c
index 5a559fcfadf9..52ebf5bd1443 100644
--- a/block/fops.c
+++ b/block/fops.c
@@ -782,9 +782,28 @@ static ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from)
 	if (iocb->ki_flags & IOCB_DIRECT) {
 		ret = blkdev_direct_write(iocb, from);
 		if (ret >= 0 && iov_iter_count(from)) {
-			ret = direct_write_fallback(iocb, from, ret,
-					blkdev_buffered_write(iocb, from));
-			need_sync = true;
+			if (iocb->ki_flags & IOCB_NOWAIT) {
+				/*
+				 * The buffered fallback blocks on i_rwsem and
+				 * on writeback of the data it copied: return
+				 * the short direct write instead and let the
+				 * caller retry. The bytes already written
+				 * still need a flush if REQ_FUA was not set.
+				 */
+				if (!ret)
+					ret = -EAGAIN;
+				else if (!blkdev_dio_fua(iocb, bdev))
+					need_sync = true;
+			} else {
+				ssize_t ret2;
+
+				inode_lock_shared(bd_inode);
+				ret2 = blkdev_buffered_write(iocb, from);
+				inode_unlock_shared(bd_inode);
+				ret = direct_write_fallback(iocb, from, ret,
+							    ret2);
+				need_sync = true;
+			}
 		} else if (ret > 0 && !blkdev_dio_fua(iocb, bdev)) {
 			/*
 			 * The device does not support FUA, so REQ_FUA was not