__generic_file_write_iter(): fix handling of sync error after DIO
authorAl Viro <viro@zeniv.linux.org.uk>
Fri, 8 Aug 2014 16:39:16 +0000 (12:39 -0400)
committerAl Viro <viro@zeniv.linux.org.uk>
Mon, 11 Aug 2014 16:27:40 +0000 (12:27 -0400)
If DIO results in short write and sync write fails, we want to bugger off
whether the DIO part has written anything or not; the logics on the return
will take care of the right return value.

Cc: stable@vger.kernel.org [3.16]
Reported-by: Anton Altaparmakov <aia21@cam.ac.uk>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
mm/filemap.c

index 900edfa..8163e04 100644 (file)
@@ -2584,7 +2584,7 @@ ssize_t __generic_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
                 * that this differs from normal direct-io semantics, which
                 * will return -EFOO even if some bytes were written.
                 */
-               if (unlikely(status < 0) && !written) {
+               if (unlikely(status < 0)) {
                        err = status;
                        goto out;
                }