ceph: correctly handle releasing an embedded cap flush
authorXiubo Li <xiubli@redhat.com>
Wed, 18 Aug 2021 13:38:42 +0000 (21:38 +0800)
committerIlya Dryomov <idryomov@gmail.com>
Wed, 25 Aug 2021 14:34:11 +0000 (16:34 +0200)
commitb2f9fa1f3bd8846f50b355fc2168236975c4d264
treee0aee74a8b50c39d3628a9d772911b740e8fbcbb
parente22ce8eb631bdc47a4a4ea7ecf4e4ba499db4f93
ceph: correctly handle releasing an embedded cap flush

The ceph_cap_flush structures are usually dynamically allocated, but
the ceph_cap_snap has an embedded one.

When force umounting, the client will try to remove all the session
caps. During this, it will free them, but that should not be done
with the ones embedded in a capsnap.

Fix this by adding a new boolean that indicates that the cap flush is
embedded in a capsnap, and skip freeing it if that's set.

At the same time, switch to using list_del_init() when detaching the
i_list and g_list heads.  It's possible for a forced umount to remove
these objects but then handle_cap_flushsnap_ack() races in and does the
list_del_init() again, corrupting memory.

Cc: stable@vger.kernel.org
URL: https://tracker.ceph.com/issues/52283
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
fs/ceph/caps.c
fs/ceph/mds_client.c
fs/ceph/snap.c
fs/ceph/super.h