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 if (!pskb_may_pull(skb, 4))
141 /* Point to L2TP header */
142 optr = ptr = skb->data;
143 session_id = ntohl(*((__be32 *) ptr));
146 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
147 * the session_id. If it is 0, the packet is a L2TP control
148 * frame and the session_id value can be discarded.
150 if (session_id == 0) {
155 /* Ok, this is a data packet. Lookup the session. */
156 session = l2tp_session_find(&init_net, NULL, session_id);
160 tunnel = session->tunnel;
164 /* Trace packet contents, if enabled */
165 if (tunnel->debug & L2TP_MSG_DATA) {
166 length = min(32u, skb->len);
167 if (!pskb_may_pull(skb, length))
170 /* Point to L2TP header */
171 optr = ptr = skb->data;
173 pr_debug("%s: ip recv\n", tunnel->name);
174 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
177 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
178 tunnel->recv_payload_hook);
182 /* Get the tunnel_id from the L2TP header */
183 if (!pskb_may_pull(skb, 12))
186 if ((skb->data[0] & 0xc0) != 0xc0)
189 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
190 tunnel = l2tp_tunnel_find(&init_net, tunnel_id);
194 struct ipv6hdr *iph = ipv6_hdr(skb);
196 read_lock_bh(&l2tp_ip6_lock);
197 sk = __l2tp_ip6_bind_lookup(&init_net, &iph->daddr,
199 read_unlock_bh(&l2tp_ip6_lock);
207 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
212 return sk_receive_skb(sk, skb, 1);
222 static int l2tp_ip6_open(struct sock *sk)
224 /* Prevent autobind. We don't have ports. */
225 inet_sk(sk)->inet_num = IPPROTO_L2TP;
227 write_lock_bh(&l2tp_ip6_lock);
228 sk_add_node(sk, &l2tp_ip6_table);
229 write_unlock_bh(&l2tp_ip6_lock);
234 static void l2tp_ip6_close(struct sock *sk, long timeout)
236 write_lock_bh(&l2tp_ip6_lock);
237 hlist_del_init(&sk->sk_bind_node);
238 sk_del_node_init(sk);
239 write_unlock_bh(&l2tp_ip6_lock);
241 sk_common_release(sk);
244 static void l2tp_ip6_destroy_sock(struct sock *sk)
246 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
249 ip6_flush_pending_frames(sk);
253 l2tp_tunnel_closeall(tunnel);
257 inet6_destroy_sock(sk);
260 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
262 struct inet_sock *inet = inet_sk(sk);
263 struct ipv6_pinfo *np = inet6_sk(sk);
264 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
269 if (!sock_flag(sk, SOCK_ZAPPED))
271 if (addr->l2tp_family != AF_INET6)
273 if (addr_len < sizeof(*addr))
276 addr_type = ipv6_addr_type(&addr->l2tp_addr);
278 /* l2tp_ip6 sockets are IPv6 only */
279 if (addr_type == IPV6_ADDR_MAPPED)
280 return -EADDRNOTAVAIL;
282 /* L2TP is point-point, not multicast */
283 if (addr_type & IPV6_ADDR_MULTICAST)
284 return -EADDRNOTAVAIL;
287 read_lock_bh(&l2tp_ip6_lock);
288 if (__l2tp_ip6_bind_lookup(&init_net, &addr->l2tp_addr,
289 sk->sk_bound_dev_if, addr->l2tp_conn_id))
291 read_unlock_bh(&l2tp_ip6_lock);
296 if (sk->sk_state != TCP_CLOSE)
299 /* Check if the address belongs to the host. */
301 if (addr_type != IPV6_ADDR_ANY) {
302 struct net_device *dev = NULL;
304 if (addr_type & IPV6_ADDR_LINKLOCAL) {
305 if (addr_len >= sizeof(struct sockaddr_in6) &&
306 addr->l2tp_scope_id) {
307 /* Override any existing binding, if another
308 * one is supplied by user.
310 sk->sk_bound_dev_if = addr->l2tp_scope_id;
313 /* Binding to link-local address requires an
315 if (!sk->sk_bound_dev_if)
319 dev = dev_get_by_index_rcu(sock_net(sk),
320 sk->sk_bound_dev_if);
325 /* ipv4 addr of the socket is invalid. Only the
326 * unspecified and mapped address have a v4 equivalent.
328 v4addr = LOOPBACK4_IPV6;
329 err = -EADDRNOTAVAIL;
330 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
335 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
336 sk->sk_v6_rcv_saddr = addr->l2tp_addr;
337 np->saddr = addr->l2tp_addr;
339 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
341 write_lock_bh(&l2tp_ip6_lock);
342 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
343 sk_del_node_init(sk);
344 write_unlock_bh(&l2tp_ip6_lock);
346 sock_reset_flag(sk, SOCK_ZAPPED);
357 read_unlock_bh(&l2tp_ip6_lock);
361 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
364 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
365 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
366 struct in6_addr *daddr;
370 if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
373 if (addr_len < sizeof(*lsa))
376 if (usin->sin6_family != AF_INET6)
379 addr_type = ipv6_addr_type(&usin->sin6_addr);
380 if (addr_type & IPV6_ADDR_MULTICAST)
383 if (addr_type & IPV6_ADDR_MAPPED) {
384 daddr = &usin->sin6_addr;
385 if (ipv4_is_multicast(daddr->s6_addr32[3]))
389 rc = ip6_datagram_connect(sk, uaddr, addr_len);
393 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
395 write_lock_bh(&l2tp_ip6_lock);
396 hlist_del_init(&sk->sk_bind_node);
397 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
398 write_unlock_bh(&l2tp_ip6_lock);
405 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
407 if (sock_flag(sk, SOCK_ZAPPED))
410 return udp_disconnect(sk, flags);
413 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
414 int *uaddr_len, int peer)
416 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
417 struct sock *sk = sock->sk;
418 struct ipv6_pinfo *np = inet6_sk(sk);
419 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
421 lsa->l2tp_family = AF_INET6;
422 lsa->l2tp_flowinfo = 0;
423 lsa->l2tp_scope_id = 0;
424 lsa->l2tp_unused = 0;
426 if (!lsk->peer_conn_id)
428 lsa->l2tp_conn_id = lsk->peer_conn_id;
429 lsa->l2tp_addr = sk->sk_v6_daddr;
431 lsa->l2tp_flowinfo = np->flow_label;
433 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
434 lsa->l2tp_addr = np->saddr;
436 lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
438 lsa->l2tp_conn_id = lsk->conn_id;
440 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
441 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
442 *uaddr_len = sizeof(*lsa);
446 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
450 /* Charge it to the socket, dropping if the queue is full. */
451 rc = sock_queue_rcv_skb(sk, skb);
458 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
463 static int l2tp_ip6_push_pending_frames(struct sock *sk)
466 __be32 *transhdr = NULL;
469 skb = skb_peek(&sk->sk_write_queue);
473 transhdr = (__be32 *)skb_transport_header(skb);
476 err = ip6_push_pending_frames(sk);
482 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
485 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
487 struct ipv6_txoptions opt_space;
488 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
489 struct in6_addr *daddr, *final_p, final;
490 struct ipv6_pinfo *np = inet6_sk(sk);
491 struct ipv6_txoptions *opt_to_free = NULL;
492 struct ipv6_txoptions *opt = NULL;
493 struct ip6_flowlabel *flowlabel = NULL;
494 struct dst_entry *dst = NULL;
496 int addr_len = msg->msg_namelen;
500 int transhdrlen = 4; /* zero session-id */
501 int ulen = len + transhdrlen;
504 /* Rough check on arithmetic overflow,
505 better check is made in ip6_append_data().
510 /* Mirror BSD error message compatibility */
511 if (msg->msg_flags & MSG_OOB)
515 * Get and verify the address.
517 memset(&fl6, 0, sizeof(fl6));
519 fl6.flowi6_mark = sk->sk_mark;
522 if (addr_len < SIN6_LEN_RFC2133)
525 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
526 return -EAFNOSUPPORT;
528 daddr = &lsa->l2tp_addr;
530 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
531 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
532 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
533 if (flowlabel == NULL)
539 * Otherwise it will be difficult to maintain
542 if (sk->sk_state == TCP_ESTABLISHED &&
543 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
544 daddr = &sk->sk_v6_daddr;
546 if (addr_len >= sizeof(struct sockaddr_in6) &&
547 lsa->l2tp_scope_id &&
548 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
549 fl6.flowi6_oif = lsa->l2tp_scope_id;
551 if (sk->sk_state != TCP_ESTABLISHED)
552 return -EDESTADDRREQ;
554 daddr = &sk->sk_v6_daddr;
555 fl6.flowlabel = np->flow_label;
558 if (fl6.flowi6_oif == 0)
559 fl6.flowi6_oif = sk->sk_bound_dev_if;
561 if (msg->msg_controllen) {
563 memset(opt, 0, sizeof(struct ipv6_txoptions));
564 opt->tot_len = sizeof(struct ipv6_txoptions);
566 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
567 &hlimit, &tclass, &dontfrag);
569 fl6_sock_release(flowlabel);
572 if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
573 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
574 if (flowlabel == NULL)
577 if (!(opt->opt_nflen|opt->opt_flen))
586 opt = fl6_merge_options(&opt_space, flowlabel, opt);
587 opt = ipv6_fixup_options(&opt_space, opt);
589 fl6.flowi6_proto = sk->sk_protocol;
590 if (!ipv6_addr_any(daddr))
593 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
594 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
595 fl6.saddr = np->saddr;
597 final_p = fl6_update_dst(&fl6, opt, &final);
599 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
600 fl6.flowi6_oif = np->mcast_oif;
601 else if (!fl6.flowi6_oif)
602 fl6.flowi6_oif = np->ucast_oif;
604 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
606 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
613 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
619 dontfrag = np->dontfrag;
621 if (msg->msg_flags & MSG_CONFIRM)
626 err = ip6_append_data(sk, ip_generic_getfrag, msg,
627 ulen, transhdrlen, hlimit, tclass, opt,
628 &fl6, (struct rt6_info *)dst,
629 msg->msg_flags, dontfrag);
631 ip6_flush_pending_frames(sk);
632 else if (!(msg->msg_flags & MSG_MORE))
633 err = l2tp_ip6_push_pending_frames(sk);
638 fl6_sock_release(flowlabel);
639 txopt_put(opt_to_free);
641 return err < 0 ? err : len;
645 if (!(msg->msg_flags & MSG_PROBE) || len)
646 goto back_from_confirm;
651 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
652 int noblock, int flags, int *addr_len)
654 struct ipv6_pinfo *np = inet6_sk(sk);
655 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
657 int err = -EOPNOTSUPP;
664 *addr_len = sizeof(*lsa);
666 if (flags & MSG_ERRQUEUE)
667 return ipv6_recv_error(sk, msg, len, addr_len);
669 skb = skb_recv_datagram(sk, flags, noblock, &err);
675 msg->msg_flags |= MSG_TRUNC;
679 err = skb_copy_datagram_msg(skb, 0, msg, copied);
683 sock_recv_timestamp(msg, sk, skb);
685 /* Copy the address. */
687 lsa->l2tp_family = AF_INET6;
688 lsa->l2tp_unused = 0;
689 lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
690 lsa->l2tp_flowinfo = 0;
691 lsa->l2tp_scope_id = 0;
692 lsa->l2tp_conn_id = 0;
693 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
694 lsa->l2tp_scope_id = inet6_iif(skb);
698 ip6_datagram_recv_ctl(sk, msg, skb);
700 if (flags & MSG_TRUNC)
703 skb_free_datagram(sk, skb);
705 return err ? err : copied;
708 static struct proto l2tp_ip6_prot = {
710 .owner = THIS_MODULE,
711 .init = l2tp_ip6_open,
712 .close = l2tp_ip6_close,
713 .bind = l2tp_ip6_bind,
714 .connect = l2tp_ip6_connect,
715 .disconnect = l2tp_ip6_disconnect,
717 .destroy = l2tp_ip6_destroy_sock,
718 .setsockopt = ipv6_setsockopt,
719 .getsockopt = ipv6_getsockopt,
720 .sendmsg = l2tp_ip6_sendmsg,
721 .recvmsg = l2tp_ip6_recvmsg,
722 .backlog_rcv = l2tp_ip6_backlog_recv,
724 .unhash = inet_unhash,
725 .obj_size = sizeof(struct l2tp_ip6_sock),
727 .compat_setsockopt = compat_ipv6_setsockopt,
728 .compat_getsockopt = compat_ipv6_getsockopt,
732 static const struct proto_ops l2tp_ip6_ops = {
734 .owner = THIS_MODULE,
735 .release = inet6_release,
737 .connect = inet_dgram_connect,
738 .socketpair = sock_no_socketpair,
739 .accept = sock_no_accept,
740 .getname = l2tp_ip6_getname,
741 .poll = datagram_poll,
742 .ioctl = inet6_ioctl,
743 .listen = sock_no_listen,
744 .shutdown = inet_shutdown,
745 .setsockopt = sock_common_setsockopt,
746 .getsockopt = sock_common_getsockopt,
747 .sendmsg = inet_sendmsg,
748 .recvmsg = sock_common_recvmsg,
749 .mmap = sock_no_mmap,
750 .sendpage = sock_no_sendpage,
752 .compat_setsockopt = compat_sock_common_setsockopt,
753 .compat_getsockopt = compat_sock_common_getsockopt,
757 static struct inet_protosw l2tp_ip6_protosw = {
759 .protocol = IPPROTO_L2TP,
760 .prot = &l2tp_ip6_prot,
761 .ops = &l2tp_ip6_ops,
764 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
765 .handler = l2tp_ip6_recv,
768 static int __init l2tp_ip6_init(void)
772 pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
774 err = proto_register(&l2tp_ip6_prot, 1);
778 err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
782 inet6_register_protosw(&l2tp_ip6_protosw);
786 proto_unregister(&l2tp_ip6_prot);
791 static void __exit l2tp_ip6_exit(void)
793 inet6_unregister_protosw(&l2tp_ip6_protosw);
794 inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
795 proto_unregister(&l2tp_ip6_prot);
798 module_init(l2tp_ip6_init);
799 module_exit(l2tp_ip6_exit);
801 MODULE_LICENSE("GPL");
802 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
803 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
804 MODULE_VERSION("1.0");
806 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
809 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
810 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);