Re: [PATCH v5 11/14] Add start_renaming_two_dentries()

On Wed, Nov 5, 2025 at 7:56 PM NeilBrown <[email protected]> wrote:
>
> From: NeilBrown <[email protected]>
>
> A few callers want to lock for a rename and already have both dentries.
> Also debugfs does want to perform a lookup but doesn't want permission
> checking, so start_renaming_dentry() cannot be used.
>
> This patch introduces start_renaming_two_dentries() which is given both
> dentries.  debugfs performs one lookup itself.  As it will only continue
> with a negative dentry and as those cannot be renamed or unlinked, it is
> safe to do the lookup before getting the rename locks.
>
> overlayfs uses start_renaming_two_dentries() in three places and  selinux
> uses it twice in sel_make_policy_nodes().
>
> In sel_make_policy_nodes() we now lock for rename twice instead of just
> once so the combined operation is no longer atomic w.r.t the parent
> directory locks.  As selinux_state.policy_mutex is held across the whole
> operation this does open up any interesting races.
>
> Reviewed-by: Amir Goldstein <[email protected]>
> Signed-off-by: NeilBrown <[email protected]>
>
> ---
> changes since v3:
>  added missing assignment to rd.mnt_idmap in ovl_cleanup_and_whiteout
> ---

> diff --git a/security/selinux/selinuxfs.c b/security/selinux/selinuxfs.c
> index 232e087bce3e..a224ef9bb831 100644
> --- a/security/selinux/selinuxfs.c
> +++ b/security/selinux/selinuxfs.c
> @@ -539,22 +540,30 @@ static int sel_make_policy_nodes(struct selinux_fs_info *fsi,
>         if (ret)
>                 goto out;
>
> -       lock_rename(tmp_parent, fsi->sb->s_root);
> +       rd.old_parent = tmp_parent;
> +       rd.new_parent = fsi->sb->s_root;
>
>         /* booleans */
> -       d_exchange(tmp_bool_dir, fsi->bool_dir);
> +       ret = start_renaming_two_dentries(&rd, tmp_bool_dir, fsi->bool_dir);
> +       if (!ret) {
> +               d_exchange(tmp_bool_dir, fsi->bool_dir);

I would recommend an immediate goto out if ret != 0; we don't want to
silently fall through and possibly reset ret on the next
start_renaming_two_dentries() call, thereby ultimately returning 0 to
the caller and acting as if nothing bad happened.

>
> -       swap(fsi->bool_num, bool_num);
> -       swap(fsi->bool_pending_names, bool_names);
> -       swap(fsi->bool_pending_values, bool_values);
> +               swap(fsi->bool_num, bool_num);
> +               swap(fsi->bool_pending_names, bool_names);
> +               swap(fsi->bool_pending_values, bool_values);
>
> -       fsi->bool_dir = tmp_bool_dir;
> +               fsi->bool_dir = tmp_bool_dir;
> +               end_renaming(&rd);
> +       }
>
>         /* classes */
> -       d_exchange(tmp_class_dir, fsi->class_dir);
> -       fsi->class_dir = tmp_class_dir;
> +       ret = start_renaming_two_dentries(&rd, tmp_class_dir, fsi->class_dir);
> +       if (ret == 0) {
> +               d_exchange(tmp_class_dir, fsi->class_dir);
> +               fsi->class_dir = tmp_class_dir;
>
> -       unlock_rename(tmp_parent, fsi->sb->s_root);
> +               end_renaming(&rd);
> +       }
>
>  out:
>         sel_remove_old_bool_data(bool_num, bool_names, bool_values);
> --
> 2.50.0.107.gf914562f5916.dirty
>