Re: [RFC][PATCHSET] sanitized pathwalk machinery (v2)

[Al Viro]

On Tue, Feb 25, 2020 at 01:24:57AM +0000, Al Viro wrote:

> Incidentally, another inconsistency is LOOKUP_BENEATH treatment in case
> when we have walked out of the subtree by way of e.g. procfs symlink and
> then ran into .. in the absolute root (that's
>                 if (!follow_up(&nd->path))
>                         break;
> in follow_dotdot()).  Shouldn't that give the same reaction as ..
> in root (EXDEV on LOOKUP_BENEATH, that is)?  It doesn't...
> 
> Another one is about LOOKUP_NO_XDEV again: suppose you have process'
> root directly overmounted and cwd in the root of whatever's overmounting
> it.  Resolution of .. will stay in cwd - we have no parent within the
> chroot jail we are in, so we move to whatever's overmounting that root.
> Which is the original location.  Should we fail on LOOKUP_NO_XDEV here?
> Plain .. in the root of chroot jail (not overmounted by anything) does
> *not*...

FWIW, my preference would be the following (for non-RCU case; RCU
one is similar)

get_parent(nd)
{
	if (path_equal(&nd->path, &nd->root))
		return NULL;
	if (nd->path.dentry != nd->path.mnt->mnt_root)
		return dget_parent(nd->path.dentry);
	m = real_mount(nd->path.mnt);
        read_seqlock_excl(&mount_lock);
	while (mnt_has_parent(m)) {
		d = m->mnt_mountpoint;
		m = m->mnt_parent;
		if (&m->mnt == nd->root.mnt && d == nd->root.path) // root
			break;
		if (m->mnt_root != d) {
			if (unlikely(nd->flags & LOOKUP_NO_XDEV)) {
				read_sequnlock_excl(&mount_lock);
				return ERR_PTR(-EXDEV);
			}
			mntget(&m->mnt);
			dget(d);
			read_sequnlock_excl(&mount_lock);
			path_put(&nd->path);
			nd->path.mnt = &m->mnt;
			nd->path.dentry = d;
			nd->inode = d->d_inode;
			return dget_parent(d);
		}
	}
	read_sequnlock_excl(&mount_lock);
	return NULL;
}

with follow_dotdot() doing
	parent = get_parent(nd);
	if (unlikely(IS_ERR(parent)))
		return PTR_ERR(parent);
	if (unlikely(!parent)) {	.. in root is a rare case
		bugger off if LOOKUP_BENEATH
		parent = dget(nd->path.dentry);
	} else if (unlikely(!path_connected(nd->path.mnt, parent))) {
		dput(parent);
		return -ENOENT;
	}
	dput(nd->path.dentry);
	nd->path.dentry = parent;
	follow_mount(&nd->path);

... with the last part replaced with
	step_into(nd, WALK_NOFOLLOW, dentry, NULL, 0);
later in this series, with similar in RCU case (only there we would want
inode and seq supplied, as usual, so it would be get_parent_rcu(nd, &inode,
&seq)).