autofs: revert "autofs: fix AT_NO_AUTOMOUNT not being honored"
authorIan Kent <raven@themaw.net>
Thu, 30 Nov 2017 00:11:26 +0000 (16:11 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 30 Nov 2017 02:40:43 +0000 (18:40 -0800)
commit5d38f049cee1e1c4a7ac55aa79d37d01ddcc3860
tree7ecfe3b2075538a5e882d2471f23eeb6adfc8e2b
parent43694d4bf843ddd34519e8e9de983deefeada699
autofs: revert "autofs: fix AT_NO_AUTOMOUNT not being honored"

Commit 42f461482178 ("autofs: fix AT_NO_AUTOMOUNT not being honored")
allowed the fstatat(2) system call to properly honor the AT_NO_AUTOMOUNT
flag but introduced a semantic change.

In order to honor AT_NO_AUTOMOUNT a semantic change was made to the
negative dentry case for stat family system calls in follow_automount().

This changed the unconditional triggering of an automount in this case
to no longer be done and an error returned instead.

This has caused more problems than I expected so reverting the change is
needed.

In a discussion with Neil Brown it was concluded that the automount(8)
daemon can implement this change without kernel modifications.  So that
will be done instead and the autofs module documentation updated with a
description of the problem and what needs to be done by module users for
this specific case.

Link: http://lkml.kernel.org/r/151174730120.6162.3848002191530283984.stgit@pluto.themaw.net
Fixes: 42f4614821 ("autofs: fix AT_NO_AUTOMOUNT not being honored")
Signed-off-by: Ian Kent <raven@themaw.net>
Cc: Neil Brown <neilb@suse.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: David Howells <dhowells@redhat.com>
Cc: Colin Walters <walters@redhat.com>
Cc: Ondrej Holy <oholy@redhat.com>
Cc: <stable@vger.kernel.org> [4.11+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/namei.c
include/linux/fs.h