netdev - Re: [RFC PATCH 0/1] lsm: Add hook unix_path

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aV6Qh39jVg8trMlg@google.com>
Date: Wed, 7 Jan 2026 17:57:43 +0100
From: Günther Noack <gnoack@...gle.com>
To: Justin Suess <utilityemal77@...il.com>
Cc: Kuniyuki Iwashima <kuniyu@...gle.com>, Paul Moore <paul@...l-moore.com>,
	James Morris <jmorris@...ei.org>,
	"Serge E . Hallyn" <serge@...lyn.com>,
	Simon Horman <horms@...nel.org>,
	Mickaël Salaün <mic@...ikod.net>,
	linux-security-module@...r.kernel.org, Tingmao Wang <m@...wtm.org>,
	netdev@...r.kernel.org, Alexander Viro <viro@...iv.linux.org.uk>,
	Christian Brauner <brauner@...nel.org>
Subject: Re: [RFC PATCH 0/1] lsm: Add hook unix_path_connect

On Wed, Jan 07, 2026 at 07:19:02AM -0500, Justin Suess wrote:
> On 1/7/26 02:33, Kuniyuki Iwashima wrote:
> > +VFS maintainers
> >
> > On Mon, Jan 5, 2026 at 3:04 AM Günther Noack <gnoack@...gle.com> wrote:
> >> Hello!
> >>
> >> On Sun, Jan 04, 2026 at 11:46:46PM -0800, Kuniyuki Iwashima wrote:
> >>> On Wed, Dec 31, 2025 at 1:33 PM Justin Suess <utilityemal77@...il.com> wrote:
> >>>> Motivation
> >>>> ---
> >>>>
> >>>> For AF_UNIX sockets bound to a filesystem path (aka named sockets), one
> >>>> identifying object from a policy perspective is the path passed to
> >>>> connect(2). However, this operation currently restricts LSMs that rely
> >>>> on VFS-based mediation, because the pathname resolved during connect()
> >>>> is not preserved in a form visible to existing hooks before connection
> >>>> establishment.
> >>> Why can't LSM use unix_sk(other)->path in security_unix_stream_connect()
> >>> and security_unix_may_send() ?
> >> Thanks for bringing it up!
> >>
> >> That path is set by the process that acts as the listening side for
> >> the socket.  The listening and the connecting process might not live
> >> in the same mount namespace, and in that case, it would not match the
> >> path which is passed by the client in the struct sockaddr_un.
> > Thanks for the explanation !
> >
> > So basically what you need is resolving unix_sk(sk)->addr.name
> > by kern_path() and comparing its d_backing_inode(path.dentry)
> > with d_backing_inode (unix_sk(sk)->path.dendtry).
> >
> > If the new hook is only used by Landlock, I'd prefer doing that on
> > the existing connect() hooks.
> I see. Did you have a particular hook in mind to extend?
> 
> One complication I see is whatever hook this gets added to
> would also need CONFIG_SECURITY_PATH, since logically this restriction
> would fall under it:
> 
> From security/Kconfig:
> 
> config SECURITY_PATH
>     bool "Security hooks for pathname based access control"
>     depends on SECURITY
>     help
>       This enables the security hooks for pathname based access control.
>       If enabled, a security module can use these hooks to
>       implement pathname based access controls.
>       If you are unsure how to answer this question, answer N.
> 
> config SECURITY_NETWORK
>     bool "Socket and Networking Security Hooks"
>     depends on SECURITY
>     help
>       This enables the socket and networking security hooks.
>       If enabled, a security module can use these hooks to
>       implement socket and networking access controls.
>       If you are unsure how to answer this question, answer N.
> 
> Logically, this type of access control falls under both categories, so must be
> gated by both features. No existing LSM hooks are gated by both afaik, so
> there is not really an existing logical place to extend an existing hook without
> changing what features are required to be enabled for existing users.
> 
> I do see more uses for this hook that just landlock, bpf lsm hooks
> or other non-labeling LSMs like apparmor or TOMOYO could take advantage
> of this as well.

Apologies, I overlooked your reply earlier today.

The existing hooks that are called from af_unix.c are:

- security_unix_stream_connect() for SOCK_STREAM unix(7) sockets
- security_unix_may_send() for SOCK_DGRAM unix(7) sockets

Apart from that, at a higher level, there are also

- security_socket_connect()
- security_socket_sendmsg() and security_socket_recvmsg()

These are used from net/socket.c.

For the connectionless dgram Unix sockets, we would need to tell apart the cases
where sendmsg()/recvmsg() are used with and without a sockaddr.  (Dgram sockets
can be either connected with connect() and then have a fixed sockaddr, or they
can be passed a remote sockaddr with each message send and receive operation.)
This can told apart in security_socket_sendmsg() from the msg argument, but it
doesn't look like we could tell it apart from security_unix_may_send().

Landlock already depends on CONFIG_SECURITY_NETWORK and CONFIG_SECURITY_PATH,
so we would not need to have further #ifdefs to use one of these hooks.

There are other difficulties I found which worry me and which I listed in the
other mail at https://lore.kernel.org/all/aV5WTGvQB0XI8Q_N@google.com/.

—Günther