Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 28 Mar 2025 11:22:36 +0000
From:      bugzilla-noreply@freebsd.org
To:        bugs@FreeBSD.org
Subject:   [Bug 262180] jail escaping via jail-friendly nullfs
Message-ID:  <bug-262180-227-zD2aBuTr6m@https.bugs.freebsd.org/bugzilla/>
In-Reply-To: <bug-262180-227@https.bugs.freebsd.org/bugzilla/>
References:  <bug-262180-227@https.bugs.freebsd.org/bugzilla/>

next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D262180

Mark Johnston <markj@FreeBSD.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kib@FreeBSD.org,
                   |                            |markj@FreeBSD.org

--- Comment #7 from Mark Johnston <markj@FreeBSD.org> ---
This is a consequence of the way nullfs handles dotdot lookups.  Here, a pr=
oc's
cwd is a nullfs vnode, and behind the scenes the corresponding lower vnode =
is
moved out of the exported directory.  Then, successive dotdot lookups
instantiate nullfs vnodes outside of the nullfs mountpoint.  These get retu=
rned
and effectively "cover" the jail's root dir, so the pointer equality check =
in
vfs_lookup() against ni_rootdir doesn't work because it's comparing the
returned nullfs vnode with a lower vnode.

There is a hack in null_lookup() to handle this kind of situation, and this=
 is
enough to prevent the problem if the process root is also a mountpoint.

To fix this, I think vfs_lookup() either has to compare ni_topdir/ni_rootdir
with "dp" and its aliases, using a VOP implemented by nullfs.  Or, we can s=
tart
passing the ndp to VOP_LOOKUP somehow and modify nullfs to perform this che=
ck
internally.

--=20
You are receiving this mail because:
You are the assignee for the bug.=



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-262180-227-zD2aBuTr6m>