packet: fix bitfield update race

Updates to the bitfields in struct packet_sock are not atomic.
Serialize these read-modify-write cycles.

Move po->running into a separate variable. Its writes are protected by
po->bind_lock (except for one startup case at packet_create). Also
replace a textual precondition warning with lockdep annotation.

All others are set only in packet_setsockopt. Serialize these
updates by holding the socket lock. Analogous to other field updates,
also hold the lock when testing whether a ring is active (pg_vec).

Fixes: 8dc419447415 ("[PACKET]: Add optional checksum computation for recvmsg")
Reported-by: DaeRyong Jeong <threeearcat@gmail.com>
Reported-by: Byoungyoung Lee <byoungyoung@purdue.edu>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/packet/af_packet.c

@@ -329,11 +329,11 @@ static void packet_pick_tx_queue(struct net_device *dev, struct sk_buff *skb)
 	skb_set_queue_mapping(skb, queue_index);
 }
 
-/* register_prot_hook must be invoked with the po->bind_lock held,
+/* __register_prot_hook must be invoked through register_prot_hook
  * or from a context in which asynchronous accesses to the packet
  * socket is not possible (packet_create()).
  */
-static void register_prot_hook(struct sock *sk)
+static void __register_prot_hook(struct sock *sk)
 {
 	struct packet_sock *po = pkt_sk(sk);
 
@@ -348,8 +348,13 @@ static void register_prot_hook(struct sock *sk)
 	}
 }
 
-/* {,__}unregister_prot_hook() must be invoked with the po->bind_lock
- * held.   If the sync parameter is true, we will temporarily drop
+static void register_prot_hook(struct sock *sk)
+{
+	lockdep_assert_held_once(&pkt_sk(sk)->bind_lock);
+	__register_prot_hook(sk);
+}
+
+/* If the sync parameter is true, we will temporarily drop
  * the po->bind_lock and do a synchronize_net to make sure no
  * asynchronous packet processing paths still refer to the elements
  * of po->prot_hook.  If the sync parameter is false, it is the
@@ -359,6 +364,8 @@ static void __unregister_prot_hook(struct sock *sk, bool sync)
 {
 	struct packet_sock *po = pkt_sk(sk);
 
+	lockdep_assert_held_once(&po->bind_lock);
+
 	po->running = 0;
 
 	if (po->fanout)
@@ -3252,7 +3259,7 @@ static int packet_create(struct net *net, struct socket *sock, int protocol,
 
 	if (proto) {
 		po->prot_hook.type = proto;
-		register_prot_hook(sk);
+		__register_prot_hook(sk);
 	}
 
 	mutex_lock(&net->packet.sklist_lock);
@@ -3732,12 +3739,18 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 
 		if (optlen != sizeof(val))
 			return -EINVAL;
-		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
-			return -EBUSY;
 		if (copy_from_user(&val, optval, sizeof(val)))
 			return -EFAULT;
-		po->tp_loss = !!val;
-		return 0;
+
+		lock_sock(sk);
+		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) {
+			ret = -EBUSY;
+		} else {
+			po->tp_loss = !!val;
+			ret = 0;
+		}
+		release_sock(sk);
+		return ret;
 	}
 	case PACKET_AUXDATA:
 	{
@@ -3748,7 +3761,9 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 		if (copy_from_user(&val, optval, sizeof(val)))
 			return -EFAULT;
 
+		lock_sock(sk);
 		po->auxdata = !!val;
+		release_sock(sk);
 		return 0;
 	}
 	case PACKET_ORIGDEV:
@@ -3760,7 +3775,9 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 		if (copy_from_user(&val, optval, sizeof(val)))
 			return -EFAULT;
 
+		lock_sock(sk);
 		po->origdev = !!val;
+		release_sock(sk);
 		return 0;
 	}
 	case PACKET_VNET_HDR:
@@ -3769,15 +3786,20 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 
 		if (sock->type != SOCK_RAW)
 			return -EINVAL;
-		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
-			return -EBUSY;
 		if (optlen < sizeof(val))
 			return -EINVAL;
 		if (copy_from_user(&val, optval, sizeof(val)))
 			return -EFAULT;
 
-		po->has_vnet_hdr = !!val;
-		return 0;
+		lock_sock(sk);
+		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) {
+			ret = -EBUSY;
+		} else {
+			po->has_vnet_hdr = !!val;
+			ret = 0;
+		}
+		release_sock(sk);
+		return ret;
 	}
 	case PACKET_TIMESTAMP:
 	{
@@ -3815,11 +3837,17 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 
 		if (optlen != sizeof(val))
 			return -EINVAL;
-		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
-			return -EBUSY;
 		if (copy_from_user(&val, optval, sizeof(val)))
 			return -EFAULT;
-		po->tp_tx_has_off = !!val;
+
+		lock_sock(sk);
+		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) {
+			ret = -EBUSY;
+		} else {
+			po->tp_tx_has_off = !!val;
+			ret = 0;
+		}
+		release_sock(sk);
 		return 0;
 	}
 	case PACKET_QDISC_BYPASS:

net/packet/internal.h

@@ -112,10 +112,12 @@ struct packet_sock {
 	int			copy_thresh;
 	spinlock_t		bind_lock;
 	struct mutex		pg_vec_lock;
-	unsigned int		running:1,	/* prot_hook is attached*/
-				auxdata:1,
+	unsigned int		running;	/* bind_lock must be held */
+	unsigned int		auxdata:1,	/* writer must hold sock lock */
 				origdev:1,
-				has_vnet_hdr:1;
+				has_vnet_hdr:1,
+				tp_loss:1,
+				tp_tx_has_off:1;
 	int			pressure;
 	int			ifindex;	/* bound device		*/
 	__be16			num;
@@ -125,8 +127,6 @@ struct packet_sock {
 	enum tpacket_versions	tp_version;
 	unsigned int		tp_hdrlen;
 	unsigned int		tp_reserve;
-	unsigned int		tp_loss:1;
-	unsigned int		tp_tx_has_off:1;
 	unsigned int		tp_tstamp;
 	struct net_device __rcu	*cached_dev;
 	int			(*xmit)(struct sk_buff *skb);