namei: only return -ECHILD from follow_dotdot_rcu()
authorAleksa Sarai <cyphar@cyphar.com>
Fri, 6 Dec 2019 14:13:26 +0000 (01:13 +1100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 11 Mar 2020 06:53:07 +0000 (07:53 +0100)
commit2739e5a9a41f75ea2e3793c1fbe328ae0c9d714e
tree67ed05c38da334e65e820042c082e8b4964d348e
parent41df68570b7ea89a537e5bd9c77cc4d7e2491cf9
namei: only return -ECHILD from follow_dotdot_rcu()

commit 2b98149c2377bff12be5dd3ce02ae0506e2dd613 upstream.

It's over-zealous to return hard errors under RCU-walk here, given that
a REF-walk will be triggered for all other cases handling ".." under
RCU.

The original purpose of this check was to ensure that if a rename occurs
such that a directory is moved outside of the bind-mount which the
resolution started in, it would be detected and blocked to avoid being
able to mess with paths outside of the bind-mount. However, triggering a
new REF-walk is just as effective a solution.

Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Fixes: 397d425dc26d ("vfs: Test for and handle paths that are unreachable from their mnt_root")
Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/namei.c