1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* L2TPv3 IP encapsulation support for IPv6
4 * Copyright (c) 2012 Katalix Systems Ltd
7 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 #include <linux/icmp.h>
10 #include <linux/module.h>
11 #include <linux/skbuff.h>
12 #include <linux/random.h>
13 #include <linux/socket.h>
14 #include <linux/l2tp.h>
16 #include <linux/in6.h>
21 #include <net/inet_common.h>
22 #include <net/tcp_states.h>
23 #include <net/protocol.h>
26 #include <net/transp_v6.h>
27 #include <net/addrconf.h>
28 #include <net/ip6_route.h>
30 #include "l2tp_core.h"
32 struct l2tp_ip6_sock {
33 /* inet_sock has to be the first member of l2tp_ip6_sock */
34 struct inet_sock inet;
39 /* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
42 struct ipv6_pinfo inet6;
45 static DEFINE_RWLOCK(l2tp_ip6_lock);
46 static struct hlist_head l2tp_ip6_table;
47 static struct hlist_head l2tp_ip6_bind_table;
49 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
51 return (struct l2tp_ip6_sock *)sk;
54 static struct sock *__l2tp_ip6_bind_lookup(const struct net *net,
55 const struct in6_addr *laddr,
56 const struct in6_addr *raddr,
57 int dif, u32 tunnel_id)
61 sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
62 const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
63 const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
64 const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
66 if (!net_eq(sock_net(sk), net))
69 if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
72 if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
73 !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
76 if (!ipv6_addr_any(sk_raddr) && raddr &&
77 !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
80 if (l2tp->conn_id != tunnel_id)
91 /* When processing receive frames, there are two cases to
92 * consider. Data frames consist of a non-zero session-id and an
93 * optional cookie. Control frames consist of a regular L2TP header
94 * preceded by 32-bits of zeros.
96 * L2TPv3 Session Header Over IP
99 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
100 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
102 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103 * | Cookie (optional, maximum 64 bits)...
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * L2TPv3 Control Message Header Over IP
111 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
112 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113 * | (32 bits of zeros) |
114 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
116 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117 * | Control Connection ID |
118 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
120 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
122 * All control frames are passed to userspace.
124 static int l2tp_ip6_recv(struct sk_buff *skb)
126 struct net *net = dev_net(skb->dev);
130 unsigned char *ptr, *optr;
131 struct l2tp_session *session;
132 struct l2tp_tunnel *tunnel = NULL;
136 if (!pskb_may_pull(skb, 4))
139 /* Point to L2TP header */
142 session_id = ntohl(*((__be32 *)ptr));
145 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
146 * the session_id. If it is 0, the packet is a L2TP control
147 * frame and the session_id value can be discarded.
149 if (session_id == 0) {
154 /* Ok, this is a data packet. Lookup the session. */
155 session = l2tp_session_get(net, session_id);
159 tunnel = session->tunnel;
163 /* Trace packet contents, if enabled */
164 if (tunnel->debug & L2TP_MSG_DATA) {
165 length = min(32u, skb->len);
166 if (!pskb_may_pull(skb, length))
169 /* Point to L2TP header */
173 pr_debug("%s: ip recv\n", tunnel->name);
174 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
177 if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
180 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
181 l2tp_session_dec_refcount(session);
186 /* Get the tunnel_id from the L2TP header */
187 if (!pskb_may_pull(skb, 12))
190 if ((skb->data[0] & 0xc0) != 0xc0)
193 tunnel_id = ntohl(*(__be32 *)&skb->data[4]);
196 read_lock_bh(&l2tp_ip6_lock);
197 sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
198 inet6_iif(skb), tunnel_id);
200 read_unlock_bh(&l2tp_ip6_lock);
204 read_unlock_bh(&l2tp_ip6_lock);
206 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
211 return sk_receive_skb(sk, skb, 1);
214 l2tp_session_dec_refcount(session);
225 static int l2tp_ip6_hash(struct sock *sk)
227 if (sk_unhashed(sk)) {
228 write_lock_bh(&l2tp_ip6_lock);
229 sk_add_node(sk, &l2tp_ip6_table);
230 write_unlock_bh(&l2tp_ip6_lock);
235 static void l2tp_ip6_unhash(struct sock *sk)
239 write_lock_bh(&l2tp_ip6_lock);
240 sk_del_node_init(sk);
241 write_unlock_bh(&l2tp_ip6_lock);
244 static int l2tp_ip6_open(struct sock *sk)
246 /* Prevent autobind. We don't have ports. */
247 inet_sk(sk)->inet_num = IPPROTO_L2TP;
253 static void l2tp_ip6_close(struct sock *sk, long timeout)
255 write_lock_bh(&l2tp_ip6_lock);
256 hlist_del_init(&sk->sk_bind_node);
257 sk_del_node_init(sk);
258 write_unlock_bh(&l2tp_ip6_lock);
260 sk_common_release(sk);
263 static void l2tp_ip6_destroy_sock(struct sock *sk)
265 struct l2tp_tunnel *tunnel = sk->sk_user_data;
268 ip6_flush_pending_frames(sk);
272 l2tp_tunnel_delete(tunnel);
274 inet6_destroy_sock(sk);
277 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
279 struct inet_sock *inet = inet_sk(sk);
280 struct ipv6_pinfo *np = inet6_sk(sk);
281 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *)uaddr;
282 struct net *net = sock_net(sk);
288 if (addr->l2tp_family != AF_INET6)
290 if (addr_len < sizeof(*addr))
293 addr_type = ipv6_addr_type(&addr->l2tp_addr);
295 /* l2tp_ip6 sockets are IPv6 only */
296 if (addr_type == IPV6_ADDR_MAPPED)
297 return -EADDRNOTAVAIL;
299 /* L2TP is point-point, not multicast */
300 if (addr_type & IPV6_ADDR_MULTICAST)
301 return -EADDRNOTAVAIL;
306 if (!sock_flag(sk, SOCK_ZAPPED))
309 if (sk->sk_state != TCP_CLOSE)
312 bound_dev_if = sk->sk_bound_dev_if;
314 /* Check if the address belongs to the host. */
316 if (addr_type != IPV6_ADDR_ANY) {
317 struct net_device *dev = NULL;
319 if (addr_type & IPV6_ADDR_LINKLOCAL) {
320 if (addr->l2tp_scope_id)
321 bound_dev_if = addr->l2tp_scope_id;
323 /* Binding to link-local address requires an
330 dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
335 /* ipv4 addr of the socket is invalid. Only the
336 * unspecified and mapped address have a v4 equivalent.
338 v4addr = LOOPBACK4_IPV6;
339 err = -EADDRNOTAVAIL;
340 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
345 write_lock_bh(&l2tp_ip6_lock);
346 if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
347 addr->l2tp_conn_id)) {
348 write_unlock_bh(&l2tp_ip6_lock);
353 inet->inet_saddr = v4addr;
354 inet->inet_rcv_saddr = v4addr;
355 sk->sk_bound_dev_if = bound_dev_if;
356 sk->sk_v6_rcv_saddr = addr->l2tp_addr;
357 np->saddr = addr->l2tp_addr;
359 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
361 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
362 sk_del_node_init(sk);
363 write_unlock_bh(&l2tp_ip6_lock);
365 sock_reset_flag(sk, SOCK_ZAPPED);
377 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
380 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
381 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
382 struct in6_addr *daddr;
386 if (addr_len < sizeof(*lsa))
389 if (usin->sin6_family != AF_INET6)
392 addr_type = ipv6_addr_type(&usin->sin6_addr);
393 if (addr_type & IPV6_ADDR_MULTICAST)
396 if (addr_type & IPV6_ADDR_MAPPED) {
397 daddr = &usin->sin6_addr;
398 if (ipv4_is_multicast(daddr->s6_addr32[3]))
404 /* Must bind first - autobinding does not work */
405 if (sock_flag(sk, SOCK_ZAPPED)) {
410 rc = __ip6_datagram_connect(sk, uaddr, addr_len);
414 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
416 write_lock_bh(&l2tp_ip6_lock);
417 hlist_del_init(&sk->sk_bind_node);
418 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
419 write_unlock_bh(&l2tp_ip6_lock);
427 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
429 if (sock_flag(sk, SOCK_ZAPPED))
432 return __udp_disconnect(sk, flags);
435 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
438 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
439 struct sock *sk = sock->sk;
440 struct ipv6_pinfo *np = inet6_sk(sk);
441 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
443 lsa->l2tp_family = AF_INET6;
444 lsa->l2tp_flowinfo = 0;
445 lsa->l2tp_scope_id = 0;
446 lsa->l2tp_unused = 0;
448 if (!lsk->peer_conn_id)
450 lsa->l2tp_conn_id = lsk->peer_conn_id;
451 lsa->l2tp_addr = sk->sk_v6_daddr;
453 lsa->l2tp_flowinfo = np->flow_label;
455 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
456 lsa->l2tp_addr = np->saddr;
458 lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
460 lsa->l2tp_conn_id = lsk->conn_id;
462 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
463 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
467 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
471 /* Charge it to the socket, dropping if the queue is full. */
472 rc = sock_queue_rcv_skb(sk, skb);
479 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
484 static int l2tp_ip6_push_pending_frames(struct sock *sk)
487 __be32 *transhdr = NULL;
490 skb = skb_peek(&sk->sk_write_queue);
494 transhdr = (__be32 *)skb_transport_header(skb);
497 err = ip6_push_pending_frames(sk);
503 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
506 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
508 struct ipv6_txoptions opt_space;
509 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
510 struct in6_addr *daddr, *final_p, final;
511 struct ipv6_pinfo *np = inet6_sk(sk);
512 struct ipv6_txoptions *opt_to_free = NULL;
513 struct ipv6_txoptions *opt = NULL;
514 struct ip6_flowlabel *flowlabel = NULL;
515 struct dst_entry *dst = NULL;
517 struct ipcm6_cookie ipc6;
518 int addr_len = msg->msg_namelen;
519 int transhdrlen = 4; /* zero session-id */
520 int ulen = len + transhdrlen;
523 /* Rough check on arithmetic overflow,
524 * better check is made in ip6_append_data().
529 /* Mirror BSD error message compatibility */
530 if (msg->msg_flags & MSG_OOB)
533 /* Get and verify the address */
534 memset(&fl6, 0, sizeof(fl6));
536 fl6.flowi6_mark = sk->sk_mark;
537 fl6.flowi6_uid = sk->sk_uid;
542 if (addr_len < SIN6_LEN_RFC2133)
545 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
546 return -EAFNOSUPPORT;
548 daddr = &lsa->l2tp_addr;
550 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
551 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
552 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
553 if (IS_ERR(flowlabel))
558 /* Otherwise it will be difficult to maintain
561 if (sk->sk_state == TCP_ESTABLISHED &&
562 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
563 daddr = &sk->sk_v6_daddr;
565 if (addr_len >= sizeof(struct sockaddr_in6) &&
566 lsa->l2tp_scope_id &&
567 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
568 fl6.flowi6_oif = lsa->l2tp_scope_id;
570 if (sk->sk_state != TCP_ESTABLISHED)
571 return -EDESTADDRREQ;
573 daddr = &sk->sk_v6_daddr;
574 fl6.flowlabel = np->flow_label;
577 if (fl6.flowi6_oif == 0)
578 fl6.flowi6_oif = sk->sk_bound_dev_if;
580 if (msg->msg_controllen) {
582 memset(opt, 0, sizeof(struct ipv6_txoptions));
583 opt->tot_len = sizeof(struct ipv6_txoptions);
586 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
588 fl6_sock_release(flowlabel);
591 if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
592 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
593 if (IS_ERR(flowlabel))
596 if (!(opt->opt_nflen | opt->opt_flen))
605 opt = fl6_merge_options(&opt_space, flowlabel, opt);
606 opt = ipv6_fixup_options(&opt_space, opt);
609 fl6.flowi6_proto = sk->sk_protocol;
610 if (!ipv6_addr_any(daddr))
613 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
614 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
615 fl6.saddr = np->saddr;
617 final_p = fl6_update_dst(&fl6, opt, &final);
619 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
620 fl6.flowi6_oif = np->mcast_oif;
621 else if (!fl6.flowi6_oif)
622 fl6.flowi6_oif = np->ucast_oif;
624 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
627 ipc6.tclass = np->tclass;
629 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
631 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
638 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
640 if (ipc6.dontfrag < 0)
641 ipc6.dontfrag = np->dontfrag;
643 if (msg->msg_flags & MSG_CONFIRM)
648 err = ip6_append_data(sk, ip_generic_getfrag, msg,
649 ulen, transhdrlen, &ipc6,
650 &fl6, (struct rt6_info *)dst,
653 ip6_flush_pending_frames(sk);
654 else if (!(msg->msg_flags & MSG_MORE))
655 err = l2tp_ip6_push_pending_frames(sk);
660 fl6_sock_release(flowlabel);
661 txopt_put(opt_to_free);
663 return err < 0 ? err : len;
666 if (msg->msg_flags & MSG_PROBE)
667 dst_confirm_neigh(dst, &fl6.daddr);
668 if (!(msg->msg_flags & MSG_PROBE) || len)
669 goto back_from_confirm;
674 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
675 int noblock, int flags, int *addr_len)
677 struct ipv6_pinfo *np = inet6_sk(sk);
678 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
680 int err = -EOPNOTSUPP;
686 if (flags & MSG_ERRQUEUE)
687 return ipv6_recv_error(sk, msg, len, addr_len);
689 skb = skb_recv_datagram(sk, flags, noblock, &err);
695 msg->msg_flags |= MSG_TRUNC;
699 err = skb_copy_datagram_msg(skb, 0, msg, copied);
703 sock_recv_timestamp(msg, sk, skb);
705 /* Copy the address. */
707 lsa->l2tp_family = AF_INET6;
708 lsa->l2tp_unused = 0;
709 lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
710 lsa->l2tp_flowinfo = 0;
711 lsa->l2tp_scope_id = 0;
712 lsa->l2tp_conn_id = 0;
713 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
714 lsa->l2tp_scope_id = inet6_iif(skb);
715 *addr_len = sizeof(*lsa);
719 ip6_datagram_recv_ctl(sk, msg, skb);
721 if (flags & MSG_TRUNC)
724 skb_free_datagram(sk, skb);
726 return err ? err : copied;
729 static struct proto l2tp_ip6_prot = {
731 .owner = THIS_MODULE,
732 .init = l2tp_ip6_open,
733 .close = l2tp_ip6_close,
734 .bind = l2tp_ip6_bind,
735 .connect = l2tp_ip6_connect,
736 .disconnect = l2tp_ip6_disconnect,
738 .destroy = l2tp_ip6_destroy_sock,
739 .setsockopt = ipv6_setsockopt,
740 .getsockopt = ipv6_getsockopt,
741 .sendmsg = l2tp_ip6_sendmsg,
742 .recvmsg = l2tp_ip6_recvmsg,
743 .backlog_rcv = l2tp_ip6_backlog_recv,
744 .hash = l2tp_ip6_hash,
745 .unhash = l2tp_ip6_unhash,
746 .obj_size = sizeof(struct l2tp_ip6_sock),
749 static const struct proto_ops l2tp_ip6_ops = {
751 .owner = THIS_MODULE,
752 .release = inet6_release,
754 .connect = inet_dgram_connect,
755 .socketpair = sock_no_socketpair,
756 .accept = sock_no_accept,
757 .getname = l2tp_ip6_getname,
758 .poll = datagram_poll,
759 .ioctl = inet6_ioctl,
760 .gettstamp = sock_gettstamp,
761 .listen = sock_no_listen,
762 .shutdown = inet_shutdown,
763 .setsockopt = sock_common_setsockopt,
764 .getsockopt = sock_common_getsockopt,
765 .sendmsg = inet_sendmsg,
766 .recvmsg = sock_common_recvmsg,
767 .mmap = sock_no_mmap,
768 .sendpage = sock_no_sendpage,
770 .compat_ioctl = inet6_compat_ioctl,
774 static struct inet_protosw l2tp_ip6_protosw = {
776 .protocol = IPPROTO_L2TP,
777 .prot = &l2tp_ip6_prot,
778 .ops = &l2tp_ip6_ops,
781 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
782 .handler = l2tp_ip6_recv,
785 static int __init l2tp_ip6_init(void)
789 pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
791 err = proto_register(&l2tp_ip6_prot, 1);
795 err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
799 inet6_register_protosw(&l2tp_ip6_protosw);
803 proto_unregister(&l2tp_ip6_prot);
808 static void __exit l2tp_ip6_exit(void)
810 inet6_unregister_protosw(&l2tp_ip6_protosw);
811 inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
812 proto_unregister(&l2tp_ip6_prot);
815 module_init(l2tp_ip6_init);
816 module_exit(l2tp_ip6_exit);
818 MODULE_LICENSE("GPL");
819 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
820 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
821 MODULE_VERSION("1.0");
823 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
826 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
827 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);