ipv6: add complete rcu protection around np->opt

This patch addresses multiple problems :

UDP/RAW sendmsg() need to get a stable struct ipv6_txoptions
while socket is not locked : Other threads can change np->opt
concurrently. Dmitry posted a syzkaller
(http://github.com/google/syzkaller) program desmonstrating
use-after-free.

Starting with TCP/DCCP lockless listeners, tcp_v6_syn_recv_sock()
and dccp_v6_request_recv_sock() also need to use RCU protection
to dereference np->opt once (before calling ipv6_dup_options())

This patch adds full RCU protection to np->opt

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/ipv6.h

@@ -227,7 +227,7 @@ struct ipv6_pinfo {
 	struct ipv6_ac_socklist	*ipv6_ac_list;
 	struct ipv6_fl_socklist __rcu *ipv6_fl_list;
 
-	struct ipv6_txoptions	*opt;
+	struct ipv6_txoptions __rcu	*opt;
 	struct sk_buff		*pktoptions;
 	struct sk_buff		*rxpmtu;
 	struct inet6_cork	cork;

include/net/ipv6.h

@@ -205,6 +205,7 @@ extern rwlock_t ip6_ra_lock;
  */
 
 struct ipv6_txoptions {
+	atomic_t		refcnt;
 	/* Length of this structure */
 	int			tot_len;
 
@@ -217,7 +218,7 @@ struct ipv6_txoptions {
 	struct ipv6_opt_hdr	*dst0opt;
 	struct ipv6_rt_hdr	*srcrt;	/* Routing Header */
 	struct ipv6_opt_hdr	*dst1opt;
-
+	struct rcu_head		rcu;
 	/* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */
 };
 
@@ -252,6 +253,24 @@ struct ipv6_fl_socklist {
 	struct rcu_head			rcu;
 };
 
+static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np)
+{
+	struct ipv6_txoptions *opt;
+
+	rcu_read_lock();
+	opt = rcu_dereference(np->opt);
+	if (opt && !atomic_inc_not_zero(&opt->refcnt))
+		opt = NULL;
+	rcu_read_unlock();
+	return opt;
+}
+
+static inline void txopt_put(struct ipv6_txoptions *opt)
+{
+	if (opt && atomic_dec_and_test(&opt->refcnt))
+		kfree_rcu(opt, rcu);
+}
+
 struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk, __be32 label);
 struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space,
 					 struct ip6_flowlabel *fl,

net/dccp/ipv6.c

@@ -202,7 +202,9 @@ static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req
 	security_req_classify_flow(req, flowi6_to_flowi(&fl6));
 
 
-	final_p = fl6_update_dst(&fl6, np->opt, &final);
+	rcu_read_lock();
+	final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
+	rcu_read_unlock();
 
 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 	if (IS_ERR(dst)) {
@@ -219,7 +221,10 @@ static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req
 							 &ireq->ir_v6_loc_addr,
 							 &ireq->ir_v6_rmt_addr);
 		fl6.daddr = ireq->ir_v6_rmt_addr;
-		err = ip6_xmit(sk, skb, &fl6, np->opt, np->tclass);
+		rcu_read_lock();
+		err = ip6_xmit(sk, skb, &fl6, rcu_dereference(np->opt),
+			       np->tclass);
+		rcu_read_unlock();
 		err = net_xmit_eval(err);
 	}
 
@@ -387,6 +392,7 @@ static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
 	struct inet_request_sock *ireq = inet_rsk(req);
 	struct ipv6_pinfo *newnp;
 	const struct ipv6_pinfo *np = inet6_sk(sk);
+	struct ipv6_txoptions *opt;
 	struct inet_sock *newinet;
 	struct dccp6_sock *newdp6;
 	struct sock *newsk;
@@ -488,13 +494,15 @@ static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
 	 * Yes, keeping reference count would be much more clever, but we make
 	 * one more one thing there: reattach optmem to newsk.
 	 */
-	if (np->opt != NULL)
+	opt = rcu_dereference(np->opt);
-		newnp->opt = ipv6_dup_options(newsk, np->opt);
+	if (opt) {
-
+		opt = ipv6_dup_options(newsk, opt);
+		RCU_INIT_POINTER(newnp->opt, opt);
+	}
 	inet_csk(newsk)->icsk_ext_hdr_len = 0;
-	if (newnp->opt != NULL)
+	if (opt)
-		inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
+		inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
-						     newnp->opt->opt_flen);
+						    opt->opt_flen;
 
 	dccp_sync_mss(newsk, dst_mtu(dst));
 
@@ -757,6 +765,7 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 	struct ipv6_pinfo *np = inet6_sk(sk);
 	struct dccp_sock *dp = dccp_sk(sk);
 	struct in6_addr *saddr = NULL, *final_p, final;
+	struct ipv6_txoptions *opt;
 	struct flowi6 fl6;
 	struct dst_entry *dst;
 	int addr_type;
@@ -856,7 +865,8 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 	fl6.fl6_sport = inet->inet_sport;
 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 
-	final_p = fl6_update_dst(&fl6, np->opt, &final);
+	opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk));
+	final_p = fl6_update_dst(&fl6, opt, &final);
 
 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 	if (IS_ERR(dst)) {
@@ -876,9 +886,8 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 	__ip6_dst_store(sk, dst, NULL, NULL);
 
 	icsk->icsk_ext_hdr_len = 0;
-	if (np->opt != NULL)
+	if (opt)
-		icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
+		icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
-					  np->opt->opt_nflen);
 
 	inet->inet_dport = usin->sin6_port;
 

net/ipv6/af_inet6.c

@@ -428,9 +428,11 @@ void inet6_destroy_sock(struct sock *sk)
 
 	/* Free tx options */
 
-	opt = xchg(&np->opt, NULL);
+	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
-	if (opt)
+	if (opt) {
-		sock_kfree_s(sk, opt, opt->tot_len);
+		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
+		txopt_put(opt);
+	}
 }
 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 
@@ -659,7 +661,10 @@ int inet6_sk_rebuild_header(struct sock *sk)
 		fl6.fl6_sport = inet->inet_sport;
 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 
-		final_p = fl6_update_dst(&fl6, np->opt, &final);
+		rcu_read_lock();
+		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
+					 &final);
+		rcu_read_unlock();
 
 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 		if (IS_ERR(dst)) {

net/ipv6/datagram.c

@@ -167,8 +167,10 @@ static int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int a
 
 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 
-	opt = flowlabel ? flowlabel->opt : np->opt;
+	rcu_read_lock();
+	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
 	final_p = fl6_update_dst(&fl6, opt, &final);
+	rcu_read_unlock();
 
 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 	err = 0;

net/ipv6/exthdrs.c

@@ -727,6 +727,7 @@ ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
 			*((char **)&opt2->dst1opt) += dif;
 		if (opt2->srcrt)
 			*((char **)&opt2->srcrt) += dif;
+		atomic_set(&opt2->refcnt, 1);
 	}
 	return opt2;
 }
@@ -790,7 +791,7 @@ ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
 		return ERR_PTR(-ENOBUFS);
 
 	memset(opt2, 0, tot_len);
-
+	atomic_set(&opt2->refcnt, 1);
 	opt2->tot_len = tot_len;
 	p = (char *)(opt2 + 1);
 

net/ipv6/inet6_connection_sock.c

@@ -78,7 +78,9 @@ struct dst_entry *inet6_csk_route_req(const struct sock *sk,
 	memset(fl6, 0, sizeof(*fl6));
 	fl6->flowi6_proto = proto;
 	fl6->daddr = ireq->ir_v6_rmt_addr;
-	final_p = fl6_update_dst(fl6, np->opt, &final);
+	rcu_read_lock();
+	final_p = fl6_update_dst(fl6, rcu_dereference(np->opt), &final);
+	rcu_read_unlock();
 	fl6->saddr = ireq->ir_v6_loc_addr;
 	fl6->flowi6_oif = ireq->ir_iif;
 	fl6->flowi6_mark = ireq->ir_mark;
@@ -142,7 +144,9 @@ static struct dst_entry *inet6_csk_route_socket(struct sock *sk,
 	fl6->fl6_dport = inet->inet_dport;
 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
 
-	final_p = fl6_update_dst(fl6, np->opt, &final);
+	rcu_read_lock();
+	final_p = fl6_update_dst(fl6, rcu_dereference(np->opt), &final);
+	rcu_read_unlock();
 
 	dst = __inet6_csk_dst_check(sk, np->dst_cookie);
 	if (!dst) {
@@ -175,7 +179,8 @@ int inet6_csk_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl_unused
 	/* Restore final destination back after routing done */
 	fl6.daddr = sk->sk_v6_daddr;
 
-	res = ip6_xmit(sk, skb, &fl6, np->opt, np->tclass);
+	res = ip6_xmit(sk, skb, &fl6, rcu_dereference(np->opt),
+		       np->tclass);
 	rcu_read_unlock();
 	return res;
 }

net/ipv6/ipv6_sockglue.c

@@ -111,7 +111,8 @@ struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
 			icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 		}
 	}
-	opt = xchg(&inet6_sk(sk)->opt, opt);
+	opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
+		   opt);
 	sk_dst_reset(sk);
 
 	return opt;
@@ -231,9 +232,12 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 				sk->sk_socket->ops = &inet_dgram_ops;
 				sk->sk_family = PF_INET;
 			}
-			opt = xchg(&np->opt, NULL);
+			opt = xchg((__force struct ipv6_txoptions **)&np->opt,
-			if (opt)
+				   NULL);
-				sock_kfree_s(sk, opt, opt->tot_len);
+			if (opt) {
+				atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
+				txopt_put(opt);
+			}
 			pktopt = xchg(&np->pktoptions, NULL);
 			kfree_skb(pktopt);
 
@@ -403,7 +407,8 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		if (optname != IPV6_RTHDR && !ns_capable(net->user_ns, CAP_NET_RAW))
 			break;
 
-		opt = ipv6_renew_options(sk, np->opt, optname,
+		opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk));
+		opt = ipv6_renew_options(sk, opt, optname,
 					 (struct ipv6_opt_hdr __user *)optval,
 					 optlen);
 		if (IS_ERR(opt)) {
@@ -432,8 +437,10 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		retv = 0;
 		opt = ipv6_update_options(sk, opt);
 sticky_done:
-		if (opt)
+		if (opt) {
-			sock_kfree_s(sk, opt, opt->tot_len);
+			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
+			txopt_put(opt);
+		}
 		break;
 	}
 
@@ -486,6 +493,7 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 			break;
 
 		memset(opt, 0, sizeof(*opt));
+		atomic_set(&opt->refcnt, 1);
 		opt->tot_len = sizeof(*opt) + optlen;
 		retv = -EFAULT;
 		if (copy_from_user(opt+1, optval, optlen))
@@ -502,8 +510,10 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		retv = 0;
 		opt = ipv6_update_options(sk, opt);
 done:
-		if (opt)
+		if (opt) {
-			sock_kfree_s(sk, opt, opt->tot_len);
+			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
+			txopt_put(opt);
+		}
 		break;
 	}
 	case IPV6_UNICAST_HOPS:
@@ -1110,10 +1120,11 @@ static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 	case IPV6_RTHDR:
 	case IPV6_DSTOPTS:
 	{
+		struct ipv6_txoptions *opt;
 
 		lock_sock(sk);
-		len = ipv6_getsockopt_sticky(sk, np->opt,
+		opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk));
-					     optname, optval, len);
+		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
 		release_sock(sk);
 		/* check if ipv6_getsockopt_sticky() returns err code */
 		if (len < 0)

net/ipv6/raw.c

@@ -733,6 +733,7 @@ static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
 
 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 {
+	struct ipv6_txoptions *opt_to_free = NULL;
 	struct ipv6_txoptions opt_space;
 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 	struct in6_addr *daddr, *final_p, final;
@@ -839,8 +840,10 @@ static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 		if (!(opt->opt_nflen|opt->opt_flen))
 			opt = NULL;
 	}
-	if (!opt)
+	if (!opt) {
-		opt = np->opt;
+		opt = txopt_get(np);
+		opt_to_free = opt;
+		}
 	if (flowlabel)
 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
 	opt = ipv6_fixup_options(&opt_space, opt);
@@ -906,6 +909,7 @@ static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	dst_release(dst);
 out:
 	fl6_sock_release(flowlabel);
+	txopt_put(opt_to_free);
 	return err < 0 ? err : len;
 do_confirm:
 	dst_confirm(dst);

net/ipv6/syncookies.c

@@ -222,7 +222,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
 		memset(&fl6, 0, sizeof(fl6));
 		fl6.flowi6_proto = IPPROTO_TCP;
 		fl6.daddr = ireq->ir_v6_rmt_addr;
-		final_p = fl6_update_dst(&fl6, np->opt, &final);
+		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
 		fl6.saddr = ireq->ir_v6_loc_addr;
 		fl6.flowi6_oif = sk->sk_bound_dev_if;
 		fl6.flowi6_mark = ireq->ir_mark;

net/ipv6/tcp_ipv6.c

@@ -120,6 +120,7 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 	struct ipv6_pinfo *np = inet6_sk(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct in6_addr *saddr = NULL, *final_p, final;
+	struct ipv6_txoptions *opt;
 	struct flowi6 fl6;
 	struct dst_entry *dst;
 	int addr_type;
@@ -235,7 +236,8 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 	fl6.fl6_dport = usin->sin6_port;
 	fl6.fl6_sport = inet->inet_sport;
 
-	final_p = fl6_update_dst(&fl6, np->opt, &final);
+	opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk));
+	final_p = fl6_update_dst(&fl6, opt, &final);
 
 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 
@@ -263,9 +265,9 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 		tcp_fetch_timewait_stamp(sk, dst);
 
 	icsk->icsk_ext_hdr_len = 0;
-	if (np->opt)
+	if (opt)
-		icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
+		icsk->icsk_ext_hdr_len = opt->opt_flen +
-					  np->opt->opt_nflen);
+					 opt->opt_nflen;
 
 	tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
 
@@ -461,7 +463,8 @@ static int tcp_v6_send_synack(const struct sock *sk, struct dst_entry *dst,
 		if (np->repflow && ireq->pktopts)
 			fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));
 
-		err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
+		err = ip6_xmit(sk, skb, fl6, rcu_dereference(np->opt),
+			       np->tclass);
 		err = net_xmit_eval(err);
 	}
 
@@ -972,6 +975,7 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 	struct inet_request_sock *ireq;
 	struct ipv6_pinfo *newnp;
 	const struct ipv6_pinfo *np = inet6_sk(sk);
+	struct ipv6_txoptions *opt;
 	struct tcp6_sock *newtcp6sk;
 	struct inet_sock *newinet;
 	struct tcp_sock *newtp;
@@ -1098,13 +1102,15 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 	   but we make one more one thing there: reattach optmem
 	   to newsk.
 	 */
-	if (np->opt)
+	opt = rcu_dereference(np->opt);
-		newnp->opt = ipv6_dup_options(newsk, np->opt);
+	if (opt) {
-
+		opt = ipv6_dup_options(newsk, opt);
+		RCU_INIT_POINTER(newnp->opt, opt);
+	}
 	inet_csk(newsk)->icsk_ext_hdr_len = 0;
-	if (newnp->opt)
+	if (opt)
-		inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
+		inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
-						     newnp->opt->opt_flen);
+						    opt->opt_flen;
 
 	tcp_ca_openreq_child(newsk, dst);
 

net/ipv6/udp.c

@@ -1110,6 +1110,7 @@ int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 	struct in6_addr *daddr, *final_p, final;
 	struct ipv6_txoptions *opt = NULL;
+	struct ipv6_txoptions *opt_to_free = NULL;
 	struct ip6_flowlabel *flowlabel = NULL;
 	struct flowi6 fl6;
 	struct dst_entry *dst;
@@ -1263,8 +1264,10 @@ int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 			opt = NULL;
 		connected = 0;
 	}
-	if (!opt)
+	if (!opt) {
-		opt = np->opt;
+		opt = txopt_get(np);
+		opt_to_free = opt;
+	}
 	if (flowlabel)
 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
 	opt = ipv6_fixup_options(&opt_space, opt);
@@ -1373,6 +1376,7 @@ int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 out:
 	dst_release(dst);
 	fl6_sock_release(flowlabel);
+	txopt_put(opt_to_free);
 	if (!err)
 		return len;
 	/*

net/l2tp/l2tp_ip6.c

@@ -486,6 +486,7 @@ static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
 	struct in6_addr *daddr, *final_p, final;
 	struct ipv6_pinfo *np = inet6_sk(sk);
+	struct ipv6_txoptions *opt_to_free = NULL;
 	struct ipv6_txoptions *opt = NULL;
 	struct ip6_flowlabel *flowlabel = NULL;
 	struct dst_entry *dst = NULL;
@@ -575,8 +576,10 @@ static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 			opt = NULL;
 	}
 
-	if (opt == NULL)
+	if (!opt) {
-		opt = np->opt;
+		opt = txopt_get(np);
+		opt_to_free = opt;
+	}
 	if (flowlabel)
 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
 	opt = ipv6_fixup_options(&opt_space, opt);
@@ -631,6 +634,7 @@ static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	dst_release(dst);
 out:
 	fl6_sock_release(flowlabel);
+	txopt_put(opt_to_free);
 
 	return err < 0 ? err : len;