[PATCH net 0/4] l2tp: register sessions atomically

[PATCH net 0/4] l2tp: register sessions atomically

[Guillaume Nault]

Currently l2tp_session_create() allocates a session, partially
initialises it and finally registers it. It therefore exposes sessions
that aren't fully initialised to the rest of the system, because
pseudo-wire specific initialisation can only happen after
l2tp_session_create() returns.
This leads to several crashes when these sessions are used or deleted.

This series starts by splitting session registration out of
l2tp_session_create() (patch #1). Thus allowing pseudo-wires code to
terminate the initialisation phase before registration.

Then patch #2 fixes the eth pseudo-wire code. This requires protecting
the session's netdevice pointer with RCU, because it still needs to be
updated concurrently after the session got registered.

Remaining patches take care of ppp pseudo-wires. RCU protection is
needed there too, for the same reasons. This time it's the pppol2tp
socket pointer that gets protected. For clarity, and since the
conversion requires more modifications, introducing RCU is done in
its own patch (#3). Then patch #4 only has to take care of fixing
sessions initialisation and registration (and adapting part of the
deletion process).


Guillaume Nault (4):
  l2tp: don't register sessions in l2tp_session_create()
  l2tp: initialise l2tp_eth sessions before registering them
  l2tp: protect sock pointer of struct pppol2tp_session with RCU
  l2tp: initialise PPP sessions before registering them

 net/l2tp/l2tp_core.c |  21 ++---
 net/l2tp/l2tp_core.h |   3 +
 net/l2tp/l2tp_eth.c  |  99 ++++++++++++++++-----
 net/l2tp/l2tp_ppp.c  | 238 ++++++++++++++++++++++++++++++++-------------------
 4 files changed, 238 insertions(+), 123 deletions(-)

-- 
2.15.0.rc2

[PATCH net 1/4] l2tp: don't register sessions in l2tp_session_create()

[Guillaume Nault]

Sessions created by l2tp_session_create() aren't fully initialised:
some pseudo-wire specific operations need to be done before making the
session usable. Therefore the PPP and Ethernet pseudo-wires continue
working on the returned l2tp session while it's already been exposed to
the rest of the system.
This can lead to various issues. In particular, the session may enter
the deletion process before having been fully initialised, which will
confuse the session removal code.

This patch moves session registration out of l2tp_session_create(), so
that callers can control when the session is exposed to the rest of the
system. This is done by the new l2tp_session_register() function.

Only pppol2tp_session_create() can be easily converted to avoid
modifying its session after registration (the debug message is dropped
in order to avoid the need for holding a reference on the session).

For pppol2tp_connect() and l2tp_eth_create()), more work is needed.
That'll be done in followup patches. For now, let's just register the
session right after its creation, like it was done before. The only
difference is that we can easily take a reference on the session before
registering it, so, at least, we're sure it's not going to be freed
while we're working on it.

Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
---
 net/l2tp/l2tp_core.c | 21 +++++++--------------
 net/l2tp/l2tp_core.h |  3 +++
 net/l2tp/l2tp_eth.c  |  9 +++++++++
 net/l2tp/l2tp_ppp.c  | 23 +++++++++++++++++------
 4 files changed, 36 insertions(+), 20 deletions(-)

diff --git a/net/l2tp/l2tp_core.c b/net/l2tp/l2tp_core.c
index 02d61101b108..afeecab955be 100644
--- a/net/l2tp/l2tp_core.c
+++ b/net/l2tp/l2tp_core.c
@@ -322,8 +322,8 @@ struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net,
 }
 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname);
 
-static int l2tp_session_add_to_tunnel(struct l2tp_tunnel *tunnel,
-				      struct l2tp_session *session)
+int l2tp_session_register(struct l2tp_session *session,
+			  struct l2tp_tunnel *tunnel)
 {
 	struct l2tp_session *session_walk;
 	struct hlist_head *g_head;
@@ -371,6 +371,10 @@ static int l2tp_session_add_to_tunnel(struct l2tp_tunnel *tunnel,
 	hlist_add_head(&session->hlist, head);
 	write_unlock_bh(&tunnel->hlist_lock);
 
+	/* Ignore management session in session count value */
+	if (session->session_id != 0)
+		atomic_inc(&l2tp_session_count);
+
 	return 0;
 
 err_tlock_pnlock:
@@ -380,6 +384,7 @@ static int l2tp_session_add_to_tunnel(struct l2tp_tunnel *tunnel,
 
 	return err;
 }
+EXPORT_SYMBOL_GPL(l2tp_session_register);
 
 /* Lookup a tunnel by id
  */
@@ -1788,7 +1793,6 @@ EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id, u32 peer_session_id, struct l2tp_session_cfg *cfg)
 {
 	struct l2tp_session *session;
-	int err;
 
 	session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
 	if (session != NULL) {
@@ -1846,17 +1850,6 @@ struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunn
 
 		refcount_set(&session->ref_count, 1);
 
-		err = l2tp_session_add_to_tunnel(tunnel, session);
-		if (err) {
-			kfree(session);
-
-			return ERR_PTR(err);
-		}
-
-		/* Ignore management session in session count value */
-		if (session->session_id != 0)
-			atomic_inc(&l2tp_session_count);
-
 		return session;
 	}
 
diff --git a/net/l2tp/l2tp_core.h b/net/l2tp/l2tp_core.h
index 67c79d9b5c6c..77caa5966736 100644
--- a/net/l2tp/l2tp_core.h
+++ b/net/l2tp/l2tp_core.h
@@ -263,6 +263,9 @@ struct l2tp_session *l2tp_session_create(int priv_size,
 					 struct l2tp_tunnel *tunnel,
 					 u32 session_id, u32 peer_session_id,
 					 struct l2tp_session_cfg *cfg);
+int l2tp_session_register(struct l2tp_session *session,
+			  struct l2tp_tunnel *tunnel);
+
 void __l2tp_session_unhash(struct l2tp_session *session);
 int l2tp_session_delete(struct l2tp_session *session);
 void l2tp_session_free(struct l2tp_session *session);
diff --git a/net/l2tp/l2tp_eth.c b/net/l2tp/l2tp_eth.c
index 014a7bc2a872..a7d76f5f31ff 100644
--- a/net/l2tp/l2tp_eth.c
+++ b/net/l2tp/l2tp_eth.c
@@ -271,6 +271,13 @@ static int l2tp_eth_create(struct net *net, struct l2tp_tunnel *tunnel,
 		goto out;
 	}
 
+	l2tp_session_inc_refcount(session);
+	rc = l2tp_session_register(session, tunnel);
+	if (rc < 0) {
+		kfree(session);
+		goto out;
+	}
+
 	dev = alloc_netdev(sizeof(*priv), name, name_assign_type,
 			   l2tp_eth_dev_setup);
 	if (!dev) {
@@ -304,6 +311,7 @@ static int l2tp_eth_create(struct net *net, struct l2tp_tunnel *tunnel,
 	__module_get(THIS_MODULE);
 	/* Must be done after register_netdev() */
 	strlcpy(session->ifname, dev->name, IFNAMSIZ);
+	l2tp_session_dec_refcount(session);
 
 	dev_hold(dev);
 
@@ -314,6 +322,7 @@ static int l2tp_eth_create(struct net *net, struct l2tp_tunnel *tunnel,
 	spriv->dev = NULL;
 out_del_session:
 	l2tp_session_delete(session);
+	l2tp_session_dec_refcount(session);
 out:
 	return rc;
 }
diff --git a/net/l2tp/l2tp_ppp.c b/net/l2tp/l2tp_ppp.c
index f50452b919d5..40cf7a78e331 100644
--- a/net/l2tp/l2tp_ppp.c
+++ b/net/l2tp/l2tp_ppp.c
@@ -715,6 +715,14 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 			error = PTR_ERR(session);
 			goto end;
 		}
+
+		l2tp_session_inc_refcount(session);
+		error = l2tp_session_register(session, tunnel);
+		if (error < 0) {
+			kfree(session);
+			goto end;
+		}
+		drop_refcnt = true;
 	}
 
 	/* Associate session with its PPPoL2TP socket */
@@ -800,7 +808,7 @@ static int pppol2tp_session_create(struct net *net, struct l2tp_tunnel *tunnel,
 	/* Error if tunnel socket is not prepped */
 	if (!tunnel->sock) {
 		error = -ENOENT;
-		goto out;
+		goto err;
 	}
 
 	/* Default MTU values. */
@@ -815,18 +823,21 @@ static int pppol2tp_session_create(struct net *net, struct l2tp_tunnel *tunnel,
 				      peer_session_id, cfg);
 	if (IS_ERR(session)) {
 		error = PTR_ERR(session);
-		goto out;
+		goto err;
 	}
 
 	ps = l2tp_session_priv(session);
 	ps->tunnel_sock = tunnel->sock;
 
-	l2tp_info(session, L2TP_MSG_CONTROL, "%s: created\n",
-		  session->name);
+	error = l2tp_session_register(session, tunnel);
+	if (error < 0)
+		goto err_sess;
 
-	error = 0;
+	return 0;
 
-out:
+err_sess:
+	kfree(session);
+err:
 	return error;
 }
 
-- 
2.15.0.rc2

[PATCH net 2/4] l2tp: initialise l2tp_eth sessions before registering them

[Guillaume Nault]

Sessions must be initialised before being made externally visible by
l2tp_session_register(). Otherwise the session may be concurrently
deleted before being initialised, which can confuse the deletion path
and eventually lead to kernel oops.

Therefore, we need to move l2tp_session_register() down in
l2tp_eth_create(), but also handle the intermediate step where only the
session or the netdevice has been registered.

We can't just call l2tp_session_register() in ->ndo_init() because
we'd have no way to properly undo this operation in ->ndo_uninit().
Instead, let's register the session and the netdevice in two different
steps and protect the session's device pointer with RCU.

And now that we allow the session's .dev field to be NULL, we don't
need to prevent the netdevice from being removed anymore. So we can
drop the dev_hold() and dev_put() calls in l2tp_eth_create() and
l2tp_eth_dev_uninit().

Fixes: d9e31d17ceba ("l2tp: Add L2TP ethernet pseudowire support")
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
---
 net/l2tp/l2tp_eth.c | 106 +++++++++++++++++++++++++++++++++++++---------------
 1 file changed, 75 insertions(+), 31 deletions(-)

diff --git a/net/l2tp/l2tp_eth.c b/net/l2tp/l2tp_eth.c
index a7d76f5f31ff..d29bfee291cb 100644
--- a/net/l2tp/l2tp_eth.c
+++ b/net/l2tp/l2tp_eth.c
@@ -54,7 +54,7 @@ struct l2tp_eth {
 
 /* via l2tp_session_priv() */
 struct l2tp_eth_sess {
-	struct net_device	*dev;
+	struct net_device __rcu *dev;
 };
 
 
@@ -72,7 +72,14 @@ static int l2tp_eth_dev_init(struct net_device *dev)
 
 static void l2tp_eth_dev_uninit(struct net_device *dev)
 {
-	dev_put(dev);
+	struct l2tp_eth *priv = netdev_priv(dev);
+	struct l2tp_eth_sess *spriv;
+
+	spriv = l2tp_session_priv(priv->session);
+	RCU_INIT_POINTER(spriv->dev, NULL);
+	/* No need for synchronize_net() here. We're called by
+	 * unregister_netdev*(), which does the synchronisation for us.
+	 */
 }
 
 static int l2tp_eth_dev_xmit(struct sk_buff *skb, struct net_device *dev)
@@ -130,8 +137,8 @@ static void l2tp_eth_dev_setup(struct net_device *dev)
 static void l2tp_eth_dev_recv(struct l2tp_session *session, struct sk_buff *skb, int data_len)
 {
 	struct l2tp_eth_sess *spriv = l2tp_session_priv(session);
-	struct net_device *dev = spriv->dev;
-	struct l2tp_eth *priv = netdev_priv(dev);
+	struct net_device *dev;
+	struct l2tp_eth *priv;
 
 	if (session->debug & L2TP_MSG_DATA) {
 		unsigned int length;
@@ -155,16 +162,25 @@ static void l2tp_eth_dev_recv(struct l2tp_session *session, struct sk_buff *skb,
 	skb_dst_drop(skb);
 	nf_reset(skb);
 
+	rcu_read_lock();
+	dev = rcu_dereference(spriv->dev);
+	if (!dev)
+		goto error_rcu;
+
+	priv = netdev_priv(dev);
 	if (dev_forward_skb(dev, skb) == NET_RX_SUCCESS) {
 		atomic_long_inc(&priv->rx_packets);
 		atomic_long_add(data_len, &priv->rx_bytes);
 	} else {
 		atomic_long_inc(&priv->rx_errors);
 	}
+	rcu_read_unlock();
+
 	return;
 
+error_rcu:
+	rcu_read_unlock();
 error:
-	atomic_long_inc(&priv->rx_errors);
 	kfree_skb(skb);
 }
 
@@ -175,11 +191,15 @@ static void l2tp_eth_delete(struct l2tp_session *session)
 
 	if (session) {
 		spriv = l2tp_session_priv(session);
-		dev = spriv->dev;
+
+		rtnl_lock();
+		dev = rtnl_dereference(spriv->dev);
 		if (dev) {
-			unregister_netdev(dev);
-			spriv->dev = NULL;
+			unregister_netdevice(dev);
+			rtnl_unlock();
 			module_put(THIS_MODULE);
+		} else {
+			rtnl_unlock();
 		}
 	}
 }
@@ -189,9 +209,20 @@ static void l2tp_eth_show(struct seq_file *m, void *arg)
 {
 	struct l2tp_session *session = arg;
 	struct l2tp_eth_sess *spriv = l2tp_session_priv(session);
-	struct net_device *dev = spriv->dev;
+	struct net_device *dev;
+
+	rcu_read_lock();
+	dev = rcu_dereference(spriv->dev);
+	if (!dev) {
+		rcu_read_unlock();
+		return;
+	}
+	dev_hold(dev);
+	rcu_read_unlock();
 
 	seq_printf(m, "   interface %s\n", dev->name);
+
+	dev_put(dev);
 }
 #endif
 
@@ -268,21 +299,14 @@ static int l2tp_eth_create(struct net *net, struct l2tp_tunnel *tunnel,
 				      peer_session_id, cfg);
 	if (IS_ERR(session)) {
 		rc = PTR_ERR(session);
-		goto out;
-	}
-
-	l2tp_session_inc_refcount(session);
-	rc = l2tp_session_register(session, tunnel);
-	if (rc < 0) {
-		kfree(session);
-		goto out;
+		goto err;
 	}
 
 	dev = alloc_netdev(sizeof(*priv), name, name_assign_type,
 			   l2tp_eth_dev_setup);
 	if (!dev) {
 		rc = -ENOMEM;
-		goto out_del_session;
+		goto err_sess;
 	}
 
 	dev_net_set(dev, net);
@@ -302,28 +326,48 @@ static int l2tp_eth_create(struct net *net, struct l2tp_tunnel *tunnel,
 #endif
 
 	spriv = l2tp_session_priv(session);
-	spriv->dev = dev;
 
-	rc = register_netdev(dev);
-	if (rc < 0)
-		goto out_del_dev;
+	l2tp_session_inc_refcount(session);
+
+	rtnl_lock();
+
+	/* Register both device and session while holding the rtnl lock. This
+	 * ensures that l2tp_eth_delete() will see that there's a device to
+	 * unregister, even if it happened to run before we assign spriv->dev.
+	 */
+	rc = l2tp_session_register(session, tunnel);
+	if (rc < 0) {
+		rtnl_unlock();
+		goto err_sess_dev;
+	}
+
+	rc = register_netdevice(dev);
+	if (rc < 0) {
+		rtnl_unlock();
+		l2tp_session_delete(session);
+		l2tp_session_dec_refcount(session);
+		free_netdev(dev);
+
+		return rc;
+	}
 
-	__module_get(THIS_MODULE);
-	/* Must be done after register_netdev() */
 	strlcpy(session->ifname, dev->name, IFNAMSIZ);
+	rcu_assign_pointer(spriv->dev, dev);
+
+	rtnl_unlock();
+
 	l2tp_session_dec_refcount(session);
 
-	dev_hold(dev);
+	__module_get(THIS_MODULE);
 
 	return 0;
 
-out_del_dev:
-	free_netdev(dev);
-	spriv->dev = NULL;
-out_del_session:
-	l2tp_session_delete(session);
+err_sess_dev:
 	l2tp_session_dec_refcount(session);
-out:
+	free_netdev(dev);
+err_sess:
+	kfree(session);
+err:
 	return rc;
 }
 
-- 
2.15.0.rc2

[PATCH net 3/4] l2tp: protect sock pointer of struct pppol2tp_session with RCU

[Guillaume Nault]

pppol2tp_session_create() registers sessions that can't have their
corresponding socket initialised. This socket has to be created by
userspace, then connected to the session by pppol2tp_connect().
Therefore, we need to protect the pppol2tp socket pointer of L2TP
sessions, so that it can safely be updated when userspace is connecting
or closing the socket. This will eventually allow pppol2tp_connect()
to avoid generating transient states while initialising its parts of the
session.

To this end, this patch protects the pppol2tp socket pointer using RCU.

The pppol2tp socket pointer is still set in pppol2tp_connect(), but
only once we know the function isn't going to fail. It's eventually
reset by pppol2tp_release(), which now has to wait for a grace period
to elapse before it can drop the last reference on the socket. This
ensures that pppol2tp_session_get_sock() can safely grab a reference
on the socket, even after ps->sk is reset to NULL but before this
operation actually gets visible from pppol2tp_session_get_sock().

The rest is standard RCU conversion: pppol2tp_recv(), which already
runs in atomic context, is simply enclosed by rcu_read_lock() and
rcu_read_unlock(), while other functions are converted to use
pppol2tp_session_get_sock() followed by sock_put().
pppol2tp_session_setsockopt() is a special case. It used to retrieve
the pppol2tp socket from the L2TP session, which itself was retrieved
from the pppol2tp socket. Therefore we can just avoid dereferencing
ps->sk and directly use the original socket pointer instead.

With all users of ps->sk now handling NULL and concurrent updates, the
L2TP ->ref() and ->deref() callbacks aren't needed anymore. Therefore,
rather than converting pppol2tp_session_sock_hold() and
pppol2tp_session_sock_put(), we can just drop them.

Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
---
 net/l2tp/l2tp_ppp.c | 154 ++++++++++++++++++++++++++++++++++------------------
 1 file changed, 101 insertions(+), 53 deletions(-)

diff --git a/net/l2tp/l2tp_ppp.c b/net/l2tp/l2tp_ppp.c
index 40cf7a78e331..d40d5492c148 100644
--- a/net/l2tp/l2tp_ppp.c
+++ b/net/l2tp/l2tp_ppp.c
@@ -122,8 +122,11 @@
 struct pppol2tp_session {
 	int			owner;		/* pid that opened the socket */
 
-	struct sock		*sock;		/* Pointer to the session
+	struct mutex		sk_lock;	/* Protects .sk */
+	struct sock __rcu	*sk;		/* Pointer to the session
 						 * PPPoX socket */
+	struct sock		*__sk;		/* Copy of .sk, for cleanup */
+	struct rcu_head		rcu;		/* For asynchronous release */
 	struct sock		*tunnel_sock;	/* Pointer to the tunnel UDP
 						 * socket */
 	int			flags;		/* accessed by PPPIOCGFLAGS.
@@ -138,6 +141,24 @@ static const struct ppp_channel_ops pppol2tp_chan_ops = {
 
 static const struct proto_ops pppol2tp_ops;
 
+/* Retrieves the pppol2tp socket associated to a session.
+ * A reference is held on the returned socket, so this function must be paired
+ * with sock_put().
+ */
+static struct sock *pppol2tp_session_get_sock(struct l2tp_session *session)
+{
+	struct pppol2tp_session *ps = l2tp_session_priv(session);
+	struct sock *sk;
+
+	rcu_read_lock();
+	sk = rcu_dereference(ps->sk);
+	if (sk)
+		sock_hold(sk);
+	rcu_read_unlock();
+
+	return sk;
+}
+
 /* Helpers to obtain tunnel/session contexts from sockets.
  */
 static inline struct l2tp_session *pppol2tp_sock_to_session(struct sock *sk)
@@ -224,7 +245,8 @@ static void pppol2tp_recv(struct l2tp_session *session, struct sk_buff *skb, int
 	/* If the socket is bound, send it in to PPP's input queue. Otherwise
 	 * queue it on the session socket.
 	 */
-	sk = ps->sock;
+	rcu_read_lock();
+	sk = rcu_dereference(ps->sk);
 	if (sk == NULL)
 		goto no_sock;
 
@@ -247,30 +269,16 @@ static void pppol2tp_recv(struct l2tp_session *session, struct sk_buff *skb, int
 			kfree_skb(skb);
 		}
 	}
+	rcu_read_unlock();
 
 	return;
 
 no_sock:
+	rcu_read_unlock();
 	l2tp_info(session, L2TP_MSG_DATA, "%s: no socket\n", session->name);
 	kfree_skb(skb);
 }
 
-static void pppol2tp_session_sock_hold(struct l2tp_session *session)
-{
-	struct pppol2tp_session *ps = l2tp_session_priv(session);
-
-	if (ps->sock)
-		sock_hold(ps->sock);
-}
-
-static void pppol2tp_session_sock_put(struct l2tp_session *session)
-{
-	struct pppol2tp_session *ps = l2tp_session_priv(session);
-
-	if (ps->sock)
-		sock_put(ps->sock);
-}
-
 /************************************************************************
  * Transmit handling
  ***********************************************************************/
@@ -431,14 +439,16 @@ static int pppol2tp_xmit(struct ppp_channel *chan, struct sk_buff *skb)
  */
 static void pppol2tp_session_close(struct l2tp_session *session)
 {
-	struct pppol2tp_session *ps = l2tp_session_priv(session);
-	struct sock *sk = ps->sock;
-	struct socket *sock = sk->sk_socket;
+	struct sock *sk;
 
 	BUG_ON(session->magic != L2TP_SESSION_MAGIC);
 
-	if (sock)
-		inet_shutdown(sock, SEND_SHUTDOWN);
+	sk = pppol2tp_session_get_sock(session);
+	if (sk) {
+		if (sk->sk_socket)
+			inet_shutdown(sk->sk_socket, SEND_SHUTDOWN);
+		sock_put(sk);
+	}
 
 	/* Don't let the session go away before our socket does */
 	l2tp_session_inc_refcount(session);
@@ -461,6 +471,14 @@ static void pppol2tp_session_destruct(struct sock *sk)
 	}
 }
 
+static void pppol2tp_put_sk(struct rcu_head *head)
+{
+	struct pppol2tp_session *ps;
+
+	ps = container_of(head, typeof(*ps), rcu);
+	sock_put(ps->__sk);
+}
+
 /* Called when the PPPoX socket (session) is closed.
  */
 static int pppol2tp_release(struct socket *sock)
@@ -486,11 +504,24 @@ static int pppol2tp_release(struct socket *sock)
 
 	session = pppol2tp_sock_to_session(sk);
 
-	/* Purge any queued data */
 	if (session != NULL) {
+		struct pppol2tp_session *ps;
+
 		__l2tp_session_unhash(session);
 		l2tp_session_queue_purge(session);
-		sock_put(sk);
+
+		ps = l2tp_session_priv(session);
+		mutex_lock(&ps->sk_lock);
+		ps->__sk = rcu_dereference_protected(ps->sk,
+						     lockdep_is_held(&ps->sk_lock));
+		RCU_INIT_POINTER(ps->sk, NULL);
+		mutex_unlock(&ps->sk_lock);
+		call_rcu(&ps->rcu, pppol2tp_put_sk);
+
+		/* Rely on the sock_put() call at the end of the function for
+		 * dropping the reference held by pppol2tp_sock_to_session().
+		 * The last reference will be dropped by pppol2tp_put_sk().
+		 */
 	}
 	release_sock(sk);
 
@@ -557,12 +588,14 @@ static int pppol2tp_create(struct net *net, struct socket *sock, int kern)
 static void pppol2tp_show(struct seq_file *m, void *arg)
 {
 	struct l2tp_session *session = arg;
-	struct pppol2tp_session *ps = l2tp_session_priv(session);
+	struct sock *sk;
+
+	sk = pppol2tp_session_get_sock(session);
+	if (sk) {
+		struct pppox_sock *po = pppox_sk(sk);
 
-	if (ps) {
-		struct pppox_sock *po = pppox_sk(ps->sock);
-		if (po)
-			seq_printf(m, "   interface %s\n", ppp_dev_name(&po->chan));
+		seq_printf(m, "   interface %s\n", ppp_dev_name(&po->chan));
+		sock_put(sk);
 	}
 }
 #endif
@@ -693,13 +726,17 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 		/* Using a pre-existing session is fine as long as it hasn't
 		 * been connected yet.
 		 */
-		if (ps->sock) {
+		mutex_lock(&ps->sk_lock);
+		if (rcu_dereference_protected(ps->sk,
+					      lockdep_is_held(&ps->sk_lock))) {
+			mutex_unlock(&ps->sk_lock);
 			error = -EEXIST;
 			goto end;
 		}
 
 		/* consistency checks */
 		if (ps->tunnel_sock != tunnel->sock) {
+			mutex_unlock(&ps->sk_lock);
 			error = -EEXIST;
 			goto end;
 		}
@@ -716,19 +753,21 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 			goto end;
 		}
 
+		ps = l2tp_session_priv(session);
+		mutex_init(&ps->sk_lock);
 		l2tp_session_inc_refcount(session);
+
+		mutex_lock(&ps->sk_lock);
 		error = l2tp_session_register(session, tunnel);
 		if (error < 0) {
+			mutex_unlock(&ps->sk_lock);
 			kfree(session);
 			goto end;
 		}
 		drop_refcnt = true;
 	}
 
-	/* Associate session with its PPPoL2TP socket */
-	ps = l2tp_session_priv(session);
 	ps->owner	     = current->pid;
-	ps->sock	     = sk;
 	ps->tunnel_sock = tunnel->sock;
 
 	session->recv_skb	= pppol2tp_recv;
@@ -737,12 +776,6 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 	session->show		= pppol2tp_show;
 #endif
 
-	/* We need to know each time a skb is dropped from the reorder
-	 * queue.
-	 */
-	session->ref = pppol2tp_session_sock_hold;
-	session->deref = pppol2tp_session_sock_put;
-
 	/* If PMTU discovery was enabled, use the MTU that was discovered */
 	dst = sk_dst_get(tunnel->sock);
 	if (dst != NULL) {
@@ -776,12 +809,17 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 	po->chan.mtu	 = session->mtu;
 
 	error = ppp_register_net_channel(sock_net(sk), &po->chan);
-	if (error)
+	if (error) {
+		mutex_unlock(&ps->sk_lock);
 		goto end;
+	}
 
 out_no_ppp:
 	/* This is how we get the session context from the socket. */
 	sk->sk_user_data = session;
+	rcu_assign_pointer(ps->sk, sk);
+	mutex_unlock(&ps->sk_lock);
+
 	sk->sk_state = PPPOX_CONNECTED;
 	l2tp_info(session, L2TP_MSG_CONTROL, "%s: created\n",
 		  session->name);
@@ -827,6 +865,7 @@ static int pppol2tp_session_create(struct net *net, struct l2tp_tunnel *tunnel,
 	}
 
 	ps = l2tp_session_priv(session);
+	mutex_init(&ps->sk_lock);
 	ps->tunnel_sock = tunnel->sock;
 
 	error = l2tp_session_register(session, tunnel);
@@ -998,12 +1037,10 @@ static int pppol2tp_session_ioctl(struct l2tp_session *session,
 		 "%s: pppol2tp_session_ioctl(cmd=%#x, arg=%#lx)\n",
 		 session->name, cmd, arg);
 
-	sk = ps->sock;
+	sk = pppol2tp_session_get_sock(session);
 	if (!sk)
 		return -EBADR;
 
-	sock_hold(sk);
-
 	switch (cmd) {
 	case SIOCGIFMTU:
 		err = -ENXIO;
@@ -1279,7 +1316,6 @@ static int pppol2tp_session_setsockopt(struct sock *sk,
 				       int optname, int val)
 {
 	int err = 0;
-	struct pppol2tp_session *ps = l2tp_session_priv(session);
 
 	switch (optname) {
 	case PPPOL2TP_SO_RECVSEQ:
@@ -1300,8 +1336,8 @@ static int pppol2tp_session_setsockopt(struct sock *sk,
 		}
 		session->send_seq = !!val;
 		{
-			struct sock *ssk      = ps->sock;
-			struct pppox_sock *po = pppox_sk(ssk);
+			struct pppox_sock *po = pppox_sk(sk);
+
 			po->chan.hdrlen = val ? PPPOL2TP_L2TP_HDR_SIZE_SEQ :
 				PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
 		}
@@ -1640,8 +1676,9 @@ static void pppol2tp_seq_session_show(struct seq_file *m, void *v)
 {
 	struct l2tp_session *session = v;
 	struct l2tp_tunnel *tunnel = session->tunnel;
-	struct pppol2tp_session *ps = l2tp_session_priv(session);
-	struct pppox_sock *po = pppox_sk(ps->sock);
+	unsigned char state;
+	char user_data_ok;
+	struct sock *sk;
 	u32 ip = 0;
 	u16 port = 0;
 
@@ -1651,6 +1688,15 @@ static void pppol2tp_seq_session_show(struct seq_file *m, void *v)
 		port = ntohs(inet->inet_sport);
 	}
 
+	sk = pppol2tp_session_get_sock(session);
+	if (sk) {
+		state = sk->sk_state;
+		user_data_ok = (session == sk->sk_user_data) ? 'Y' : 'N';
+	} else {
+		state = 0;
+		user_data_ok = 'N';
+	}
+
 	seq_printf(m, "  SESSION '%s' %08X/%d %04X/%04X -> "
 		   "%04X/%04X %d %c\n",
 		   session->name, ip, port,
@@ -1658,9 +1704,7 @@ static void pppol2tp_seq_session_show(struct seq_file *m, void *v)
 		   session->session_id,
 		   tunnel->peer_tunnel_id,
 		   session->peer_session_id,
-		   ps->sock->sk_state,
-		   (session == ps->sock->sk_user_data) ?
-		   'Y' : 'N');
+		   state, user_data_ok);
 	seq_printf(m, "   %d/%d/%c/%c/%s %08x %u\n",
 		   session->mtu, session->mru,
 		   session->recv_seq ? 'R' : '-',
@@ -1677,8 +1721,12 @@ static void pppol2tp_seq_session_show(struct seq_file *m, void *v)
 		   atomic_long_read(&session->stats.rx_bytes),
 		   atomic_long_read(&session->stats.rx_errors));
 
-	if (po)
+	if (sk) {
+		struct pppox_sock *po = pppox_sk(sk);
+
 		seq_printf(m, "   interface %s\n", ppp_dev_name(&po->chan));
+		sock_put(sk);
+	}
 }
 
 static int pppol2tp_seq_show(struct seq_file *m, void *v)
-- 
2.15.0.rc2

[PATCH net 4/4] l2tp: initialise PPP sessions before registering them

[Guillaume Nault]

pppol2tp_connect() initialises L2TP sessions after they've been exposed
to the rest of the system by l2tp_session_register(). This puts
sessions into transient states that are the source of several races, in
particular with session's deletion path.

This patch centralises the initialisation code into
pppol2tp_session_init(), which is called before the registration phase.
The only field that can't be set before session registration is the
pppol2tp socket pointer, which has already been converted to RCU. So
pppol2tp_connect() should now be race-free.

The session's .session_close() callback is now set before registration.
Therefore, it's always called when l2tp_core deletes the session, even
if it was created by pppol2tp_session_create() and hasn't been plugged
to a pppol2tp socket yet. That'd prevent session free because the extra
reference taken by pppol2tp_session_close() wouldn't be dropped by the
socket's ->sk_destruct() callback (pppol2tp_session_destruct()).
We could set .session_close() only while connecting a session to its
pppol2tp socket, or teach pppol2tp_session_close() to avoid grabbing a
reference when the session isn't connected, but that'd require adding
some form of synchronisation to be race free.

Instead of that, we can just let the pppol2tp socket hold a reference
on the session as soon as it starts depending on it (that is, in
pppol2tp_connect()). Then we don't need to utilise
pppol2tp_session_close() to hold a reference at the last moment to
prevent l2tp_core from dropping it.

When releasing the socket, pppol2tp_release() now deletes the session
using the standard l2tp_session_delete() function, instead of merely
removing it from hash tables. l2tp_session_delete() drops the reference
the sessions holds on itself, but also makes sure it doesn't remove a
session twice. So it can safely be called, even if l2tp_core already
tried, or is concurrently trying, to remove the session.
Finally, pppol2tp_session_destruct() drops the reference held by the
socket.

Fixes: fd558d186df2 ("l2tp: Split pppol2tp patch into separate l2tp and ppp parts")
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
---
 net/l2tp/l2tp_ppp.c | 69 +++++++++++++++++++++++++++++------------------------
 1 file changed, 38 insertions(+), 31 deletions(-)

diff --git a/net/l2tp/l2tp_ppp.c b/net/l2tp/l2tp_ppp.c
index d40d5492c148..845aba543dce 100644
--- a/net/l2tp/l2tp_ppp.c
+++ b/net/l2tp/l2tp_ppp.c
@@ -449,9 +449,6 @@ static void pppol2tp_session_close(struct l2tp_session *session)
 			inet_shutdown(sk->sk_socket, SEND_SHUTDOWN);
 		sock_put(sk);
 	}
-
-	/* Don't let the session go away before our socket does */
-	l2tp_session_inc_refcount(session);
 }
 
 /* Really kill the session socket. (Called from sock_put() if
@@ -507,8 +504,7 @@ static int pppol2tp_release(struct socket *sock)
 	if (session != NULL) {
 		struct pppol2tp_session *ps;
 
-		__l2tp_session_unhash(session);
-		l2tp_session_queue_purge(session);
+		l2tp_session_delete(session);
 
 		ps = l2tp_session_priv(session);
 		mutex_lock(&ps->sk_lock);
@@ -600,6 +596,35 @@ static void pppol2tp_show(struct seq_file *m, void *arg)
 }
 #endif
 
+static void pppol2tp_session_init(struct l2tp_session *session)
+{
+	struct pppol2tp_session *ps;
+	struct dst_entry *dst;
+
+	session->recv_skb = pppol2tp_recv;
+	session->session_close = pppol2tp_session_close;
+#if IS_ENABLED(CONFIG_L2TP_DEBUGFS)
+	session->show = pppol2tp_show;
+#endif
+
+	ps = l2tp_session_priv(session);
+	mutex_init(&ps->sk_lock);
+	ps->tunnel_sock = session->tunnel->sock;
+	ps->owner = current->pid;
+
+	/* If PMTU discovery was enabled, use the MTU that was discovered */
+	dst = sk_dst_get(session->tunnel->sock);
+	if (dst) {
+		u32 pmtu = dst_mtu(dst);
+
+		if (pmtu) {
+			session->mtu = pmtu - PPPOL2TP_HEADER_OVERHEAD;
+			session->mru = pmtu - PPPOL2TP_HEADER_OVERHEAD;
+		}
+		dst_release(dst);
+	}
+}
+
 /* connect() handler. Attach a PPPoX socket to a tunnel UDP socket
  */
 static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
@@ -611,7 +636,6 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 	struct l2tp_session *session = NULL;
 	struct l2tp_tunnel *tunnel;
 	struct pppol2tp_session *ps;
-	struct dst_entry *dst;
 	struct l2tp_session_cfg cfg = { 0, };
 	int error = 0;
 	u32 tunnel_id, peer_tunnel_id;
@@ -753,8 +777,8 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 			goto end;
 		}
 
+		pppol2tp_session_init(session);
 		ps = l2tp_session_priv(session);
-		mutex_init(&ps->sk_lock);
 		l2tp_session_inc_refcount(session);
 
 		mutex_lock(&ps->sk_lock);
@@ -767,26 +791,6 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 		drop_refcnt = true;
 	}
 
-	ps->owner	     = current->pid;
-	ps->tunnel_sock = tunnel->sock;
-
-	session->recv_skb	= pppol2tp_recv;
-	session->session_close	= pppol2tp_session_close;
-#if IS_ENABLED(CONFIG_L2TP_DEBUGFS)
-	session->show		= pppol2tp_show;
-#endif
-
-	/* If PMTU discovery was enabled, use the MTU that was discovered */
-	dst = sk_dst_get(tunnel->sock);
-	if (dst != NULL) {
-		u32 pmtu = dst_mtu(dst);
-
-		if (pmtu != 0)
-			session->mtu = session->mru = pmtu -
-				PPPOL2TP_HEADER_OVERHEAD;
-		dst_release(dst);
-	}
-
 	/* Special case: if source & dest session_id == 0x0000, this
 	 * socket is being created to manage the tunnel. Just set up
 	 * the internal context for use by ioctl() and sockopt()
@@ -820,6 +824,12 @@ static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
 	rcu_assign_pointer(ps->sk, sk);
 	mutex_unlock(&ps->sk_lock);
 
+	/* Keep the reference we've grabbed on the session: sk doesn't expect
+	 * the session to disappear. pppol2tp_session_destruct() is responsible
+	 * for dropping it.
+	 */
+	drop_refcnt = false;
+
 	sk->sk_state = PPPOX_CONNECTED;
 	l2tp_info(session, L2TP_MSG_CONTROL, "%s: created\n",
 		  session->name);
@@ -841,7 +851,6 @@ static int pppol2tp_session_create(struct net *net, struct l2tp_tunnel *tunnel,
 {
 	int error;
 	struct l2tp_session *session;
-	struct pppol2tp_session *ps;
 
 	/* Error if tunnel socket is not prepped */
 	if (!tunnel->sock) {
@@ -864,9 +873,7 @@ static int pppol2tp_session_create(struct net *net, struct l2tp_tunnel *tunnel,
 		goto err;
 	}
 
-	ps = l2tp_session_priv(session);
-	mutex_init(&ps->sk_lock);
-	ps->tunnel_sock = tunnel->sock;
+	pppol2tp_session_init(session);
 
 	error = l2tp_session_register(session, tunnel);
 	if (error < 0)
-- 
2.15.0.rc2

Re: [PATCH net 0/4] l2tp: register sessions atomically

[David Miller]

From: Guillaume Nault <g.nault@alphalink.fr>
Date: Fri, 27 Oct 2017 16:51:48 +0200

> Currently l2tp_session_create() allocates a session, partially
> initialises it and finally registers it. It therefore exposes sessions
> that aren't fully initialised to the rest of the system, because
> pseudo-wire specific initialisation can only happen after
> l2tp_session_create() returns.
> This leads to several crashes when these sessions are used or deleted.
> 
> This series starts by splitting session registration out of
> l2tp_session_create() (patch #1). Thus allowing pseudo-wires code to
> terminate the initialisation phase before registration.
> 
> Then patch #2 fixes the eth pseudo-wire code. This requires protecting
> the session's netdevice pointer with RCU, because it still needs to be
> updated concurrently after the session got registered.
> 
> Remaining patches take care of ppp pseudo-wires. RCU protection is
> needed there too, for the same reasons. This time it's the pppol2tp
> socket pointer that gets protected. For clarity, and since the
> conversion requires more modifications, introducing RCU is done in
> its own patch (#3). Then patch #4 only has to take care of fixing
> sessions initialisation and registration (and adapting part of the
> deletion process).

Series applied, thank you.