ceph: fix handling of "meta" errors
authorJeff Layton <jlayton@kernel.org>
Thu, 7 Oct 2021 18:19:49 +0000 (14:19 -0400)
committerIlya Dryomov <idryomov@gmail.com>
Tue, 19 Oct 2021 07:36:06 +0000 (09:36 +0200)
commit1bd85aa65d0e7b5e4d09240f492f37c569fdd431
tree33cc9ed572a6b30d85ad53560bc98074acd8e732
parent98d0a6fb7303a6f4a120b8b8ed05b86ff5db53e8
ceph: fix handling of "meta" errors

Currently, we check the wb_err too early for directories, before all of
the unsafe child requests have been waited on. In order to fix that we
need to check the mapping->wb_err later nearer to the end of ceph_fsync.

We also have an overly-complex method for tracking errors after
blocklisting. The errors recorded in cleanup_session_requests go to a
completely separate field in the inode, but we end up reporting them the
same way we would for any other error (in fsync).

There's no real benefit to tracking these errors in two different
places, since the only reporting mechanism for them is in fsync, and
we'd need to advance them both every time.

Given that, we can just remove i_meta_err, and convert the places that
used it to instead just use mapping->wb_err instead. That also fixes
the original problem by ensuring that we do a check_and_advance of the
wb_err at the end of the fsync op.

Cc: stable@vger.kernel.org
URL: https://tracker.ceph.com/issues/52864
Reported-by: Patrick Donnelly <pdonnell@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
fs/ceph/caps.c
fs/ceph/file.c
fs/ceph/inode.c
fs/ceph/mds_client.c
fs/ceph/super.h