ipv6: annotate data-races around np->mcast_oif

np->mcast_oif is read locklessly in some contexts.

Make all accesses to this field lockless, adding appropriate
annotations.

This also makes setsockopt( IPV6_MULTICAST_IF ) lockless.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/dccp/ipv6.c

@@ -669,7 +669,7 @@ static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 ipv6_pktoptions:
 	if (!((1 << sk->sk_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) {
 		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
-			np->mcast_oif = inet6_iif(opt_skb);
+			WRITE_ONCE(np->mcast_oif, inet6_iif(opt_skb));
 		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
 			WRITE_ONCE(np->mcast_hops, ipv6_hdr(opt_skb)->hop_limit);
 		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)

net/ipv6/datagram.c

@@ -60,7 +60,7 @@ static void ip6_datagram_flow_key_init(struct flowi6 *fl6,
 
 	if (!oif) {
 		if (ipv6_addr_is_multicast(&fl6->daddr))
-			oif = np->mcast_oif;
+			oif = READ_ONCE(np->mcast_oif);
 		else
 			oif = np->ucast_oif;
 	}
@@ -229,7 +229,7 @@ int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
 		}
 
 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
-			WRITE_ONCE(sk->sk_bound_dev_if, np->mcast_oif);
+			WRITE_ONCE(sk->sk_bound_dev_if, READ_ONCE(np->mcast_oif));
 
 		/* Connect to link-local address requires an interface */
 		if (!sk->sk_bound_dev_if) {

net/ipv6/icmp.c

@@ -584,7 +584,7 @@ void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info,
 	tmp_hdr.icmp6_pointer = htonl(info);
 
 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
-		fl6.flowi6_oif = np->mcast_oif;
+		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
 	else if (!fl6.flowi6_oif)
 		fl6.flowi6_oif = np->ucast_oif;
 
@@ -770,7 +770,7 @@ static enum skb_drop_reason icmpv6_echo_reply(struct sk_buff *skb)
 	np = inet6_sk(sk);
 
 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
-		fl6.flowi6_oif = np->mcast_oif;
+		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
 	else if (!fl6.flowi6_oif)
 		fl6.flowi6_oif = np->ucast_oif;
 

net/ipv6/ipv6_sockglue.c

@@ -509,6 +509,34 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		if (optlen < sizeof(int))
 			return -EINVAL;
 		return ip6_sock_set_addr_preferences(sk, val);
+	case IPV6_MULTICAST_IF:
+		if (sk->sk_type == SOCK_STREAM)
+			return -ENOPROTOOPT;
+		if (optlen < sizeof(int))
+			return -EINVAL;
+		if (val) {
+			struct net_device *dev;
+			int bound_dev_if, midx;
+
+			rcu_read_lock();
+
+			dev = dev_get_by_index_rcu(net, val);
+			if (!dev) {
+				rcu_read_unlock();
+				return -ENODEV;
+			}
+			midx = l3mdev_master_ifindex_rcu(dev);
+
+			rcu_read_unlock();
+
+			bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
+			if (bound_dev_if &&
+			    bound_dev_if != val &&
+			    (!midx || midx != bound_dev_if))
+				return -EINVAL;
+		}
+		WRITE_ONCE(np->mcast_oif, val);
+		return 0;
 	}
 	if (needs_rtnl)
 		rtnl_lock();
@@ -860,36 +888,6 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		break;
 	}
 
-	case IPV6_MULTICAST_IF:
-		if (sk->sk_type == SOCK_STREAM)
-			break;
-		if (optlen < sizeof(int))
-			goto e_inval;
-
-		if (val) {
-			struct net_device *dev;
-			int midx;
-
-			rcu_read_lock();
-
-			dev = dev_get_by_index_rcu(net, val);
-			if (!dev) {
-				rcu_read_unlock();
-				retv = -ENODEV;
-				break;
-			}
-			midx = l3mdev_master_ifindex_rcu(dev);
-
-			rcu_read_unlock();
-
-			if (sk->sk_bound_dev_if &&
-			    sk->sk_bound_dev_if != val &&
-			    (!midx || midx != sk->sk_bound_dev_if))
-				goto e_inval;
-		}
-		np->mcast_oif = val;
-		retv = 0;
-		break;
 	case IPV6_ADD_MEMBERSHIP:
 	case IPV6_DROP_MEMBERSHIP:
 	{
@@ -1161,10 +1159,12 @@ int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 		sockopt_release_sock(sk);
 		if (!skb) {
 			if (np->rxopt.bits.rxinfo) {
+				int mcast_oif = READ_ONCE(np->mcast_oif);
 				struct in6_pktinfo src_info;
-				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
+
+				src_info.ipi6_ifindex = mcast_oif ? :
 					np->sticky_pktinfo.ipi6_ifindex;
-				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
+				src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 			}
 			if (np->rxopt.bits.rxhlim) {
@@ -1178,10 +1178,12 @@ int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
 			}
 			if (np->rxopt.bits.rxoinfo) {
+				int mcast_oif = READ_ONCE(np->mcast_oif);
 				struct in6_pktinfo src_info;
-				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
+
+				src_info.ipi6_ifindex = mcast_oif ? :
 					np->sticky_pktinfo.ipi6_ifindex;
-				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
+				src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr :
 								 np->sticky_pktinfo.ipi6_addr;
 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 			}
@@ -1359,7 +1361,7 @@ int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 		break;
 
 	case IPV6_MULTICAST_IF:
-		val = np->mcast_oif;
+		val = READ_ONCE(np->mcast_oif);
 		break;
 
 	case IPV6_MULTICAST_ALL:

net/ipv6/ping.c

@@ -107,7 +107,7 @@ static int ping_v6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 		oif = np->sticky_pktinfo.ipi6_ifindex;
 
 	if (!oif && ipv6_addr_is_multicast(daddr))
-		oif = np->mcast_oif;
+		oif = READ_ONCE(np->mcast_oif);
 	else if (!oif)
 		oif = np->ucast_oif;
 
@@ -157,7 +157,7 @@ static int ping_v6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	rt = (struct rt6_info *) dst;
 
 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
-		fl6.flowi6_oif = np->mcast_oif;
+		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
 	else if (!fl6.flowi6_oif)
 		fl6.flowi6_oif = np->ucast_oif;
 

net/ipv6/raw.c

@@ -876,7 +876,7 @@ static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	final_p = fl6_update_dst(&fl6, opt, &final);
 
 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
-		fl6.flowi6_oif = np->mcast_oif;
+		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
 	else if (!fl6.flowi6_oif)
 		fl6.flowi6_oif = np->ucast_oif;
 	security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));

net/ipv6/tcp_ipv6.c

@@ -1702,7 +1702,7 @@ int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 	if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
 	    !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
 		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
-			np->mcast_oif = tcp_v6_iif(opt_skb);
+			WRITE_ONCE(np->mcast_oif, tcp_v6_iif(opt_skb));
 		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
 			WRITE_ONCE(np->mcast_hops,
 				   ipv6_hdr(opt_skb)->hop_limit);

net/ipv6/udp.c

@@ -1541,7 +1541,7 @@ int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 		connected = false;
 
 	if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr)) {
-		fl6->flowi6_oif = np->mcast_oif;
+		fl6->flowi6_oif = READ_ONCE(np->mcast_oif);
 		connected = false;
 	} else if (!fl6->flowi6_oif)
 		fl6->flowi6_oif = np->ucast_oif;

net/l2tp/l2tp_ip6.c

@@ -599,7 +599,7 @@ static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	final_p = fl6_update_dst(&fl6, opt, &final);
 
 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
-		fl6.flowi6_oif = np->mcast_oif;
+		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
 	else if (!fl6.flowi6_oif)
 		fl6.flowi6_oif = np->ucast_oif;
 

net/netfilter/ipvs/ip_vs_sync.c

@@ -1365,7 +1365,7 @@ static int set_mcast_if(struct sock *sk, struct net_device *dev)
 		struct ipv6_pinfo *np = inet6_sk(sk);
 
 		/* IPV6_MULTICAST_IF */
-		np->mcast_oif = dev->ifindex;
+		WRITE_ONCE(np->mcast_oif, dev->ifindex);
 	}
 #endif
 	release_sock(sk);

net/rds/tcp_listen.c

@@ -165,7 +165,7 @@ int rds_tcp_accept_one(struct socket *sock)
 		struct ipv6_pinfo *inet6;
 
 		inet6 = inet6_sk(new_sock->sk);
-		dev_if = inet6->mcast_oif;
+		dev_if = READ_ONCE(inet6->mcast_oif);
 	} else {
 		dev_if = new_sock->sk->sk_bound_dev_if;
 	}