2 * L2TPv3 IP encapsulation support for IPv6
4 * Copyright (c) 2012 Katalix Systems Ltd
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 #include <linux/icmp.h>
15 #include <linux/module.h>
16 #include <linux/skbuff.h>
17 #include <linux/random.h>
18 #include <linux/socket.h>
19 #include <linux/l2tp.h>
21 #include <linux/in6.h>
26 #include <net/inet_common.h>
27 #include <net/inet_hashtables.h>
28 #include <net/tcp_states.h>
29 #include <net/protocol.h>
32 #include <net/transp_v6.h>
33 #include <net/addrconf.h>
34 #include <net/ip6_route.h>
36 #include "l2tp_core.h"
38 struct l2tp_ip6_sock {
39 /* inet_sock has to be the first member of l2tp_ip6_sock */
40 struct inet_sock inet;
45 /* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
47 struct ipv6_pinfo inet6;
50 static DEFINE_RWLOCK(l2tp_ip6_lock);
51 static struct hlist_head l2tp_ip6_table;
52 static struct hlist_head l2tp_ip6_bind_table;
54 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
56 return (struct l2tp_ip6_sock *)sk;
59 static struct sock *__l2tp_ip6_bind_lookup(struct net *net,
60 struct in6_addr *laddr,
61 int dif, u32 tunnel_id)
65 sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
66 const struct in6_addr *addr = inet6_rcv_saddr(sk);
67 struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
72 if ((l2tp->conn_id == tunnel_id) &&
73 net_eq(sock_net(sk), net) &&
74 !(addr && ipv6_addr_equal(addr, laddr)) &&
75 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
84 static inline struct sock *l2tp_ip6_bind_lookup(struct net *net,
85 struct in6_addr *laddr,
86 int dif, u32 tunnel_id)
88 struct sock *sk = __l2tp_ip6_bind_lookup(net, laddr, dif, tunnel_id);
95 /* When processing receive frames, there are two cases to
96 * consider. Data frames consist of a non-zero session-id and an
97 * optional cookie. Control frames consist of a regular L2TP header
98 * preceded by 32-bits of zeros.
100 * L2TPv3 Session Header Over IP
103 * 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
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107 * | Cookie (optional, maximum 64 bits)...
108 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
110 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112 * L2TPv3 Control Message Header Over IP
115 * 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
116 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117 * | (32 bits of zeros) |
118 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
120 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121 * | Control Connection ID |
122 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
124 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
126 * All control frames are passed to userspace.
128 static int l2tp_ip6_recv(struct sk_buff *skb)
133 unsigned char *ptr, *optr;
134 struct l2tp_session *session;
135 struct l2tp_tunnel *tunnel = NULL;
138 /* Point to L2TP header */
139 optr = ptr = skb->data;
141 if (!pskb_may_pull(skb, 4))
144 session_id = ntohl(*((__be32 *) ptr));
147 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
148 * the session_id. If it is 0, the packet is a L2TP control
149 * frame and the session_id value can be discarded.
151 if (session_id == 0) {
156 /* Ok, this is a data packet. Lookup the session. */
157 session = l2tp_session_find(&init_net, NULL, session_id);
161 tunnel = session->tunnel;
165 /* Trace packet contents, if enabled */
166 if (tunnel->debug & L2TP_MSG_DATA) {
167 length = min(32u, skb->len);
168 if (!pskb_may_pull(skb, length))
171 pr_debug("%s: ip recv\n", tunnel->name);
172 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
175 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
176 tunnel->recv_payload_hook);
180 /* Get the tunnel_id from the L2TP header */
181 if (!pskb_may_pull(skb, 12))
184 if ((skb->data[0] & 0xc0) != 0xc0)
187 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
188 tunnel = l2tp_tunnel_find(&init_net, tunnel_id);
192 struct ipv6hdr *iph = ipv6_hdr(skb);
194 read_lock_bh(&l2tp_ip6_lock);
195 sk = __l2tp_ip6_bind_lookup(&init_net, &iph->daddr,
197 read_unlock_bh(&l2tp_ip6_lock);
205 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
210 return sk_receive_skb(sk, skb, 1);
220 static int l2tp_ip6_open(struct sock *sk)
222 /* Prevent autobind. We don't have ports. */
223 inet_sk(sk)->inet_num = IPPROTO_L2TP;
225 write_lock_bh(&l2tp_ip6_lock);
226 sk_add_node(sk, &l2tp_ip6_table);
227 write_unlock_bh(&l2tp_ip6_lock);
232 static void l2tp_ip6_close(struct sock *sk, long timeout)
234 write_lock_bh(&l2tp_ip6_lock);
235 hlist_del_init(&sk->sk_bind_node);
236 sk_del_node_init(sk);
237 write_unlock_bh(&l2tp_ip6_lock);
239 sk_common_release(sk);
242 static void l2tp_ip6_destroy_sock(struct sock *sk)
244 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
247 ip6_flush_pending_frames(sk);
251 l2tp_tunnel_closeall(tunnel);
255 inet6_destroy_sock(sk);
258 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
260 struct inet_sock *inet = inet_sk(sk);
261 struct ipv6_pinfo *np = inet6_sk(sk);
262 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
267 if (!sock_flag(sk, SOCK_ZAPPED))
269 if (addr->l2tp_family != AF_INET6)
271 if (addr_len < sizeof(*addr))
274 addr_type = ipv6_addr_type(&addr->l2tp_addr);
276 /* l2tp_ip6 sockets are IPv6 only */
277 if (addr_type == IPV6_ADDR_MAPPED)
278 return -EADDRNOTAVAIL;
280 /* L2TP is point-point, not multicast */
281 if (addr_type & IPV6_ADDR_MULTICAST)
282 return -EADDRNOTAVAIL;
285 read_lock_bh(&l2tp_ip6_lock);
286 if (__l2tp_ip6_bind_lookup(&init_net, &addr->l2tp_addr,
287 sk->sk_bound_dev_if, addr->l2tp_conn_id))
289 read_unlock_bh(&l2tp_ip6_lock);
294 if (sk->sk_state != TCP_CLOSE)
297 /* Check if the address belongs to the host. */
299 if (addr_type != IPV6_ADDR_ANY) {
300 struct net_device *dev = NULL;
302 if (addr_type & IPV6_ADDR_LINKLOCAL) {
303 if (addr_len >= sizeof(struct sockaddr_in6) &&
304 addr->l2tp_scope_id) {
305 /* Override any existing binding, if another
306 * one is supplied by user.
308 sk->sk_bound_dev_if = addr->l2tp_scope_id;
311 /* Binding to link-local address requires an
313 if (!sk->sk_bound_dev_if)
317 dev = dev_get_by_index_rcu(sock_net(sk),
318 sk->sk_bound_dev_if);
323 /* ipv4 addr of the socket is invalid. Only the
324 * unspecified and mapped address have a v4 equivalent.
326 v4addr = LOOPBACK4_IPV6;
327 err = -EADDRNOTAVAIL;
328 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
333 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
334 sk->sk_v6_rcv_saddr = addr->l2tp_addr;
335 np->saddr = addr->l2tp_addr;
337 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
339 write_lock_bh(&l2tp_ip6_lock);
340 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
341 sk_del_node_init(sk);
342 write_unlock_bh(&l2tp_ip6_lock);
344 sock_reset_flag(sk, SOCK_ZAPPED);
355 read_unlock_bh(&l2tp_ip6_lock);
359 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
362 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
363 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
364 struct in6_addr *daddr;
368 if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
371 if (addr_len < sizeof(*lsa))
374 if (usin->sin6_family != AF_INET6)
377 addr_type = ipv6_addr_type(&usin->sin6_addr);
378 if (addr_type & IPV6_ADDR_MULTICAST)
381 if (addr_type & IPV6_ADDR_MAPPED) {
382 daddr = &usin->sin6_addr;
383 if (ipv4_is_multicast(daddr->s6_addr32[3]))
387 rc = ip6_datagram_connect(sk, uaddr, addr_len);
391 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
393 write_lock_bh(&l2tp_ip6_lock);
394 hlist_del_init(&sk->sk_bind_node);
395 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
396 write_unlock_bh(&l2tp_ip6_lock);
403 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
405 if (sock_flag(sk, SOCK_ZAPPED))
408 return udp_disconnect(sk, flags);
411 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
412 int *uaddr_len, int peer)
414 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
415 struct sock *sk = sock->sk;
416 struct ipv6_pinfo *np = inet6_sk(sk);
417 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
419 lsa->l2tp_family = AF_INET6;
420 lsa->l2tp_flowinfo = 0;
421 lsa->l2tp_scope_id = 0;
422 lsa->l2tp_unused = 0;
424 if (!lsk->peer_conn_id)
426 lsa->l2tp_conn_id = lsk->peer_conn_id;
427 lsa->l2tp_addr = sk->sk_v6_daddr;
429 lsa->l2tp_flowinfo = np->flow_label;
431 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
432 lsa->l2tp_addr = np->saddr;
434 lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
436 lsa->l2tp_conn_id = lsk->conn_id;
438 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
439 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
440 *uaddr_len = sizeof(*lsa);
444 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
448 /* Charge it to the socket, dropping if the queue is full. */
449 rc = sock_queue_rcv_skb(sk, skb);
456 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
461 static int l2tp_ip6_push_pending_frames(struct sock *sk)
464 __be32 *transhdr = NULL;
467 skb = skb_peek(&sk->sk_write_queue);
471 transhdr = (__be32 *)skb_transport_header(skb);
474 err = ip6_push_pending_frames(sk);
480 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
483 static int l2tp_ip6_sendmsg(struct kiocb *iocb, struct sock *sk,
484 struct msghdr *msg, size_t len)
486 struct ipv6_txoptions opt_space;
487 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
488 struct in6_addr *daddr, *final_p, final;
489 struct ipv6_pinfo *np = inet6_sk(sk);
490 struct ipv6_txoptions *opt = NULL;
491 struct ip6_flowlabel *flowlabel = NULL;
492 struct dst_entry *dst = NULL;
494 int addr_len = msg->msg_namelen;
498 int transhdrlen = 4; /* zero session-id */
499 int ulen = len + transhdrlen;
502 /* Rough check on arithmetic overflow,
503 better check is made in ip6_append_data().
508 /* Mirror BSD error message compatibility */
509 if (msg->msg_flags & MSG_OOB)
513 * Get and verify the address.
515 memset(&fl6, 0, sizeof(fl6));
517 fl6.flowi6_mark = sk->sk_mark;
520 if (addr_len < SIN6_LEN_RFC2133)
523 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
524 return -EAFNOSUPPORT;
526 daddr = &lsa->l2tp_addr;
528 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
529 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
530 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
531 if (flowlabel == NULL)
537 * Otherwise it will be difficult to maintain
540 if (sk->sk_state == TCP_ESTABLISHED &&
541 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
542 daddr = &sk->sk_v6_daddr;
544 if (addr_len >= sizeof(struct sockaddr_in6) &&
545 lsa->l2tp_scope_id &&
546 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
547 fl6.flowi6_oif = lsa->l2tp_scope_id;
549 if (sk->sk_state != TCP_ESTABLISHED)
550 return -EDESTADDRREQ;
552 daddr = &sk->sk_v6_daddr;
553 fl6.flowlabel = np->flow_label;
556 if (fl6.flowi6_oif == 0)
557 fl6.flowi6_oif = sk->sk_bound_dev_if;
559 if (msg->msg_controllen) {
561 memset(opt, 0, sizeof(struct ipv6_txoptions));
562 opt->tot_len = sizeof(struct ipv6_txoptions);
564 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
565 &hlimit, &tclass, &dontfrag);
567 fl6_sock_release(flowlabel);
570 if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
571 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
572 if (flowlabel == NULL)
575 if (!(opt->opt_nflen|opt->opt_flen))
582 opt = fl6_merge_options(&opt_space, flowlabel, opt);
583 opt = ipv6_fixup_options(&opt_space, opt);
585 fl6.flowi6_proto = sk->sk_protocol;
586 if (!ipv6_addr_any(daddr))
589 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
590 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
591 fl6.saddr = np->saddr;
593 final_p = fl6_update_dst(&fl6, opt, &final);
595 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
596 fl6.flowi6_oif = np->mcast_oif;
597 else if (!fl6.flowi6_oif)
598 fl6.flowi6_oif = np->ucast_oif;
600 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
602 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
609 if (ipv6_addr_is_multicast(&fl6.daddr))
610 hlimit = np->mcast_hops;
612 hlimit = np->hop_limit;
614 hlimit = ip6_dst_hoplimit(dst);
621 dontfrag = np->dontfrag;
623 if (msg->msg_flags & MSG_CONFIRM)
628 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
629 ulen, transhdrlen, hlimit, tclass, opt,
630 &fl6, (struct rt6_info *)dst,
631 msg->msg_flags, dontfrag);
633 ip6_flush_pending_frames(sk);
634 else if (!(msg->msg_flags & MSG_MORE))
635 err = l2tp_ip6_push_pending_frames(sk);
640 fl6_sock_release(flowlabel);
642 return err < 0 ? err : len;
646 if (!(msg->msg_flags & MSG_PROBE) || len)
647 goto back_from_confirm;
652 static int l2tp_ip6_recvmsg(struct kiocb *iocb, struct sock *sk,
653 struct msghdr *msg, size_t len, int noblock,
654 int flags, int *addr_len)
656 struct ipv6_pinfo *np = inet6_sk(sk);
657 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
659 int err = -EOPNOTSUPP;
666 *addr_len = sizeof(*lsa);
668 if (flags & MSG_ERRQUEUE)
669 return ipv6_recv_error(sk, msg, len, addr_len);
671 skb = skb_recv_datagram(sk, flags, noblock, &err);
677 msg->msg_flags |= MSG_TRUNC;
681 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
685 sock_recv_timestamp(msg, sk, skb);
687 /* Copy the address. */
689 lsa->l2tp_family = AF_INET6;
690 lsa->l2tp_unused = 0;
691 lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
692 lsa->l2tp_flowinfo = 0;
693 lsa->l2tp_scope_id = 0;
694 lsa->l2tp_conn_id = 0;
695 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
696 lsa->l2tp_scope_id = IP6CB(skb)->iif;
700 ip6_datagram_recv_ctl(sk, msg, skb);
702 if (flags & MSG_TRUNC)
705 skb_free_datagram(sk, skb);
707 return err ? err : copied;
710 static struct proto l2tp_ip6_prot = {
712 .owner = THIS_MODULE,
713 .init = l2tp_ip6_open,
714 .close = l2tp_ip6_close,
715 .bind = l2tp_ip6_bind,
716 .connect = l2tp_ip6_connect,
717 .disconnect = l2tp_ip6_disconnect,
719 .destroy = l2tp_ip6_destroy_sock,
720 .setsockopt = ipv6_setsockopt,
721 .getsockopt = ipv6_getsockopt,
722 .sendmsg = l2tp_ip6_sendmsg,
723 .recvmsg = l2tp_ip6_recvmsg,
724 .backlog_rcv = l2tp_ip6_backlog_recv,
726 .unhash = inet_unhash,
727 .obj_size = sizeof(struct l2tp_ip6_sock),
729 .compat_setsockopt = compat_ipv6_setsockopt,
730 .compat_getsockopt = compat_ipv6_getsockopt,
734 static const struct proto_ops l2tp_ip6_ops = {
736 .owner = THIS_MODULE,
737 .release = inet6_release,
739 .connect = inet_dgram_connect,
740 .socketpair = sock_no_socketpair,
741 .accept = sock_no_accept,
742 .getname = l2tp_ip6_getname,
743 .poll = datagram_poll,
744 .ioctl = inet6_ioctl,
745 .listen = sock_no_listen,
746 .shutdown = inet_shutdown,
747 .setsockopt = sock_common_setsockopt,
748 .getsockopt = sock_common_getsockopt,
749 .sendmsg = inet_sendmsg,
750 .recvmsg = sock_common_recvmsg,
751 .mmap = sock_no_mmap,
752 .sendpage = sock_no_sendpage,
754 .compat_setsockopt = compat_sock_common_setsockopt,
755 .compat_getsockopt = compat_sock_common_getsockopt,
759 static struct inet_protosw l2tp_ip6_protosw = {
761 .protocol = IPPROTO_L2TP,
762 .prot = &l2tp_ip6_prot,
763 .ops = &l2tp_ip6_ops,
767 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
768 .handler = l2tp_ip6_recv,
771 static int __init l2tp_ip6_init(void)
775 pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
777 err = proto_register(&l2tp_ip6_prot, 1);
781 err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
785 inet6_register_protosw(&l2tp_ip6_protosw);
789 proto_unregister(&l2tp_ip6_prot);
794 static void __exit l2tp_ip6_exit(void)
796 inet6_unregister_protosw(&l2tp_ip6_protosw);
797 inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
798 proto_unregister(&l2tp_ip6_prot);
801 module_init(l2tp_ip6_init);
802 module_exit(l2tp_ip6_exit);
804 MODULE_LICENSE("GPL");
805 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
806 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
807 MODULE_VERSION("1.0");
809 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
812 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);