selinux: fix handling of empty opts in selinux_fs_context_submount()

selinux_set_mnt_opts() relies on the fact that the mount options pointer
is always NULL when all options are unset (specifically in its
!selinux_initialized() branch. However, the new
selinux_fs_context_submount() hook breaks this rule by allocating a new
structure even if no options are set. That causes any submount created
before a SELinux policy is loaded to be rejected in
selinux_set_mnt_opts().

Fix this by making selinux_fs_context_submount() leave fc->security
set to NULL when there are no options to be copied from the reference
superblock.

Cc: <stable@vger.kernel.org>
Reported-by: Adam Williamson <awilliam@redhat.com>
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2236345
Fixes: d80a8f1 ("vfs, security: Fix automount superblock LSM init problem, preventing NFS sb sharing")
Signed-off-by: Ondrej Mosnacek <omosnace@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Paul Moore <paul@paul-moore.com>

Author: WOnder93

security/selinux/hooks.c

@@ -2775,14 +2775,20 @@ static int selinux_umount(struct vfsmount *mnt, int flags)
 static int selinux_fs_context_submount(struct fs_context *fc,
 				   struct super_block *reference)
 {
-	const struct superblock_security_struct *sbsec;
+	const struct superblock_security_struct *sbsec = selinux_superblock(reference);
 	struct selinux_mnt_opts *opts;
 
+	/*
+	 * Ensure that fc->security remains NULL when no options are set
+	 * as expected by selinux_set_mnt_opts().
+	 */
+	if (!(sbsec->flags & (FSCONTEXT_MNT|CONTEXT_MNT|DEFCONTEXT_MNT)))
+		return 0;
+
 	opts = kzalloc(sizeof(*opts), GFP_KERNEL);
 	if (!opts)
 		return -ENOMEM;
 
-	sbsec = selinux_superblock(reference);
 	if (sbsec->flags & FSCONTEXT_MNT)
 		opts->fscontext_sid = sbsec->sid;
 	if (sbsec->flags & CONTEXT_MNT)