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/inet6_hashtables.h>
29 #include <net/tcp_states.h>
30 #include <net/protocol.h>
33 #include <net/transp_v6.h>
34 #include <net/addrconf.h>
35 #include <net/ip6_route.h>
37 #include "l2tp_core.h"
39 struct l2tp_ip6_sock {
40 /* inet_sock has to be the first member of l2tp_ip6_sock */
41 struct inet_sock inet;
46 /* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
48 struct ipv6_pinfo inet6;
51 static DEFINE_RWLOCK(l2tp_ip6_lock);
52 static struct hlist_head l2tp_ip6_table;
53 static struct hlist_head l2tp_ip6_bind_table;
55 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
57 return (struct l2tp_ip6_sock *)sk;
60 static struct sock *__l2tp_ip6_bind_lookup(struct net *net,
61 struct in6_addr *laddr,
62 int dif, u32 tunnel_id)
66 sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
67 const struct in6_addr *addr = inet6_rcv_saddr(sk);
68 struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
73 if ((l2tp->conn_id == tunnel_id) &&
74 net_eq(sock_net(sk), net) &&
75 !(addr && ipv6_addr_equal(addr, laddr)) &&
76 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
85 static inline struct sock *l2tp_ip6_bind_lookup(struct net *net,
86 struct in6_addr *laddr,
87 int dif, u32 tunnel_id)
89 struct sock *sk = __l2tp_ip6_bind_lookup(net, laddr, dif, tunnel_id);
96 /* When processing receive frames, there are two cases to
97 * consider. Data frames consist of a non-zero session-id and an
98 * optional cookie. Control frames consist of a regular L2TP header
99 * preceded by 32-bits of zeros.
101 * L2TPv3 Session Header Over IP
104 * 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
105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * | Cookie (optional, maximum 64 bits)...
109 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113 * L2TPv3 Control Message Header Over IP
116 * 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
117 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
118 * | (32 bits of zeros) |
119 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
120 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
121 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
122 * | Control Connection ID |
123 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
125 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
127 * All control frames are passed to userspace.
129 static int l2tp_ip6_recv(struct sk_buff *skb)
134 unsigned char *ptr, *optr;
135 struct l2tp_session *session;
136 struct l2tp_tunnel *tunnel = NULL;
139 /* Point to L2TP header */
140 optr = ptr = skb->data;
142 if (!pskb_may_pull(skb, 4))
145 session_id = ntohl(*((__be32 *) ptr));
148 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
149 * the session_id. If it is 0, the packet is a L2TP control
150 * frame and the session_id value can be discarded.
152 if (session_id == 0) {
157 /* Ok, this is a data packet. Lookup the session. */
158 session = l2tp_session_find(&init_net, NULL, session_id);
162 tunnel = session->tunnel;
166 /* Trace packet contents, if enabled */
167 if (tunnel->debug & L2TP_MSG_DATA) {
168 length = min(32u, skb->len);
169 if (!pskb_may_pull(skb, length))
172 pr_debug("%s: ip recv\n", tunnel->name);
173 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
176 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
177 tunnel->recv_payload_hook);
181 /* Get the tunnel_id from the L2TP header */
182 if (!pskb_may_pull(skb, 12))
185 if ((skb->data[0] & 0xc0) != 0xc0)
188 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
189 tunnel = l2tp_tunnel_find(&init_net, tunnel_id);
193 struct ipv6hdr *iph = ipv6_hdr(skb);
195 read_lock_bh(&l2tp_ip6_lock);
196 sk = __l2tp_ip6_bind_lookup(&init_net, &iph->daddr,
198 read_unlock_bh(&l2tp_ip6_lock);
206 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
211 return sk_receive_skb(sk, skb, 1);
221 static int l2tp_ip6_open(struct sock *sk)
223 /* Prevent autobind. We don't have ports. */
224 inet_sk(sk)->inet_num = IPPROTO_L2TP;
226 write_lock_bh(&l2tp_ip6_lock);
227 sk_add_node(sk, &l2tp_ip6_table);
228 write_unlock_bh(&l2tp_ip6_lock);
233 static void l2tp_ip6_close(struct sock *sk, long timeout)
235 write_lock_bh(&l2tp_ip6_lock);
236 hlist_del_init(&sk->sk_bind_node);
237 sk_del_node_init(sk);
238 write_unlock_bh(&l2tp_ip6_lock);
240 sk_common_release(sk);
243 static void l2tp_ip6_destroy_sock(struct sock *sk)
245 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
248 ip6_flush_pending_frames(sk);
252 l2tp_tunnel_closeall(tunnel);
256 inet6_destroy_sock(sk);
259 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
261 struct inet_sock *inet = inet_sk(sk);
262 struct ipv6_pinfo *np = inet6_sk(sk);
263 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
268 if (!sock_flag(sk, SOCK_ZAPPED))
270 if (addr->l2tp_family != AF_INET6)
272 if (addr_len < sizeof(*addr))
275 addr_type = ipv6_addr_type(&addr->l2tp_addr);
277 /* l2tp_ip6 sockets are IPv6 only */
278 if (addr_type == IPV6_ADDR_MAPPED)
279 return -EADDRNOTAVAIL;
281 /* L2TP is point-point, not multicast */
282 if (addr_type & IPV6_ADDR_MULTICAST)
283 return -EADDRNOTAVAIL;
286 read_lock_bh(&l2tp_ip6_lock);
287 if (__l2tp_ip6_bind_lookup(&init_net, &addr->l2tp_addr,
288 sk->sk_bound_dev_if, addr->l2tp_conn_id))
290 read_unlock_bh(&l2tp_ip6_lock);
295 if (sk->sk_state != TCP_CLOSE)
298 /* Check if the address belongs to the host. */
300 if (addr_type != IPV6_ADDR_ANY) {
301 struct net_device *dev = NULL;
303 if (addr_type & IPV6_ADDR_LINKLOCAL) {
304 if (addr_len >= sizeof(struct sockaddr_in6) &&
305 addr->l2tp_scope_id) {
306 /* Override any existing binding, if another
307 * one is supplied by user.
309 sk->sk_bound_dev_if = addr->l2tp_scope_id;
312 /* Binding to link-local address requires an
314 if (!sk->sk_bound_dev_if)
318 dev = dev_get_by_index_rcu(sock_net(sk),
319 sk->sk_bound_dev_if);
324 /* ipv4 addr of the socket is invalid. Only the
325 * unspecified and mapped address have a v4 equivalent.
327 v4addr = LOOPBACK4_IPV6;
328 err = -EADDRNOTAVAIL;
329 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
334 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
335 sk->sk_v6_rcv_saddr = addr->l2tp_addr;
336 np->saddr = addr->l2tp_addr;
338 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
340 write_lock_bh(&l2tp_ip6_lock);
341 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
342 sk_del_node_init(sk);
343 write_unlock_bh(&l2tp_ip6_lock);
345 sock_reset_flag(sk, SOCK_ZAPPED);
356 read_unlock_bh(&l2tp_ip6_lock);
360 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
363 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
364 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
365 struct in6_addr *daddr;
369 if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
372 if (addr_len < sizeof(*lsa))
375 if (usin->sin6_family != AF_INET6)
378 addr_type = ipv6_addr_type(&usin->sin6_addr);
379 if (addr_type & IPV6_ADDR_MULTICAST)
382 if (addr_type & IPV6_ADDR_MAPPED) {
383 daddr = &usin->sin6_addr;
384 if (ipv4_is_multicast(daddr->s6_addr32[3]))
388 rc = ip6_datagram_connect(sk, uaddr, addr_len);
392 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
394 write_lock_bh(&l2tp_ip6_lock);
395 hlist_del_init(&sk->sk_bind_node);
396 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
397 write_unlock_bh(&l2tp_ip6_lock);
404 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
406 if (sock_flag(sk, SOCK_ZAPPED))
409 return udp_disconnect(sk, flags);
412 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
413 int *uaddr_len, int peer)
415 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
416 struct sock *sk = sock->sk;
417 struct ipv6_pinfo *np = inet6_sk(sk);
418 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
420 lsa->l2tp_family = AF_INET6;
421 lsa->l2tp_flowinfo = 0;
422 lsa->l2tp_scope_id = 0;
423 lsa->l2tp_unused = 0;
425 if (!lsk->peer_conn_id)
427 lsa->l2tp_conn_id = lsk->peer_conn_id;
428 lsa->l2tp_addr = sk->sk_v6_daddr;
430 lsa->l2tp_flowinfo = np->flow_label;
432 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
433 lsa->l2tp_addr = np->saddr;
435 lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
437 lsa->l2tp_conn_id = lsk->conn_id;
439 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
440 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
441 *uaddr_len = sizeof(*lsa);
445 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
449 /* Charge it to the socket, dropping if the queue is full. */
450 rc = sock_queue_rcv_skb(sk, skb);
457 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
462 static int l2tp_ip6_push_pending_frames(struct sock *sk)
465 __be32 *transhdr = NULL;
468 skb = skb_peek(&sk->sk_write_queue);
472 transhdr = (__be32 *)skb_transport_header(skb);
475 err = ip6_push_pending_frames(sk);
481 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
484 static int l2tp_ip6_sendmsg(struct sock *sk, 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_to_free = NULL;
491 struct ipv6_txoptions *opt = NULL;
492 struct ip6_flowlabel *flowlabel = NULL;
493 struct dst_entry *dst = NULL;
495 int addr_len = msg->msg_namelen;
499 int transhdrlen = 4; /* zero session-id */
500 int ulen = len + transhdrlen;
503 /* Rough check on arithmetic overflow,
504 better check is made in ip6_append_data().
509 /* Mirror BSD error message compatibility */
510 if (msg->msg_flags & MSG_OOB)
514 * Get and verify the address.
516 memset(&fl6, 0, sizeof(fl6));
518 fl6.flowi6_mark = sk->sk_mark;
521 if (addr_len < SIN6_LEN_RFC2133)
524 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
525 return -EAFNOSUPPORT;
527 daddr = &lsa->l2tp_addr;
529 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
530 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
531 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
532 if (flowlabel == NULL)
538 * Otherwise it will be difficult to maintain
541 if (sk->sk_state == TCP_ESTABLISHED &&
542 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
543 daddr = &sk->sk_v6_daddr;
545 if (addr_len >= sizeof(struct sockaddr_in6) &&
546 lsa->l2tp_scope_id &&
547 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
548 fl6.flowi6_oif = lsa->l2tp_scope_id;
550 if (sk->sk_state != TCP_ESTABLISHED)
551 return -EDESTADDRREQ;
553 daddr = &sk->sk_v6_daddr;
554 fl6.flowlabel = np->flow_label;
557 if (fl6.flowi6_oif == 0)
558 fl6.flowi6_oif = sk->sk_bound_dev_if;
560 if (msg->msg_controllen) {
562 memset(opt, 0, sizeof(struct ipv6_txoptions));
563 opt->tot_len = sizeof(struct ipv6_txoptions);
565 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
566 &hlimit, &tclass, &dontfrag);
568 fl6_sock_release(flowlabel);
571 if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
572 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
573 if (flowlabel == NULL)
576 if (!(opt->opt_nflen|opt->opt_flen))
585 opt = fl6_merge_options(&opt_space, flowlabel, opt);
586 opt = ipv6_fixup_options(&opt_space, opt);
588 fl6.flowi6_proto = sk->sk_protocol;
589 if (!ipv6_addr_any(daddr))
592 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
593 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
594 fl6.saddr = np->saddr;
596 final_p = fl6_update_dst(&fl6, opt, &final);
598 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
599 fl6.flowi6_oif = np->mcast_oif;
600 else if (!fl6.flowi6_oif)
601 fl6.flowi6_oif = np->ucast_oif;
603 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
605 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
612 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
618 dontfrag = np->dontfrag;
620 if (msg->msg_flags & MSG_CONFIRM)
625 err = ip6_append_data(sk, ip_generic_getfrag, msg,
626 ulen, transhdrlen, hlimit, tclass, opt,
627 &fl6, (struct rt6_info *)dst,
628 msg->msg_flags, dontfrag);
630 ip6_flush_pending_frames(sk);
631 else if (!(msg->msg_flags & MSG_MORE))
632 err = l2tp_ip6_push_pending_frames(sk);
637 fl6_sock_release(flowlabel);
638 txopt_put(opt_to_free);
640 return err < 0 ? err : len;
644 if (!(msg->msg_flags & MSG_PROBE) || len)
645 goto back_from_confirm;
650 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
651 int noblock, int flags, int *addr_len)
653 struct ipv6_pinfo *np = inet6_sk(sk);
654 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
656 int err = -EOPNOTSUPP;
663 *addr_len = sizeof(*lsa);
665 if (flags & MSG_ERRQUEUE)
666 return ipv6_recv_error(sk, msg, len, addr_len);
668 skb = skb_recv_datagram(sk, flags, noblock, &err);
674 msg->msg_flags |= MSG_TRUNC;
678 err = skb_copy_datagram_msg(skb, 0, msg, copied);
682 sock_recv_timestamp(msg, sk, skb);
684 /* Copy the address. */
686 lsa->l2tp_family = AF_INET6;
687 lsa->l2tp_unused = 0;
688 lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
689 lsa->l2tp_flowinfo = 0;
690 lsa->l2tp_scope_id = 0;
691 lsa->l2tp_conn_id = 0;
692 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
693 lsa->l2tp_scope_id = inet6_iif(skb);
697 ip6_datagram_recv_ctl(sk, msg, skb);
699 if (flags & MSG_TRUNC)
702 skb_free_datagram(sk, skb);
704 return err ? err : copied;
707 static struct proto l2tp_ip6_prot = {
709 .owner = THIS_MODULE,
710 .init = l2tp_ip6_open,
711 .close = l2tp_ip6_close,
712 .bind = l2tp_ip6_bind,
713 .connect = l2tp_ip6_connect,
714 .disconnect = l2tp_ip6_disconnect,
716 .destroy = l2tp_ip6_destroy_sock,
717 .setsockopt = ipv6_setsockopt,
718 .getsockopt = ipv6_getsockopt,
719 .sendmsg = l2tp_ip6_sendmsg,
720 .recvmsg = l2tp_ip6_recvmsg,
721 .backlog_rcv = l2tp_ip6_backlog_recv,
723 .unhash = inet_unhash,
724 .obj_size = sizeof(struct l2tp_ip6_sock),
726 .compat_setsockopt = compat_ipv6_setsockopt,
727 .compat_getsockopt = compat_ipv6_getsockopt,
731 static const struct proto_ops l2tp_ip6_ops = {
733 .owner = THIS_MODULE,
734 .release = inet6_release,
736 .connect = inet_dgram_connect,
737 .socketpair = sock_no_socketpair,
738 .accept = sock_no_accept,
739 .getname = l2tp_ip6_getname,
740 .poll = datagram_poll,
741 .ioctl = inet6_ioctl,
742 .listen = sock_no_listen,
743 .shutdown = inet_shutdown,
744 .setsockopt = sock_common_setsockopt,
745 .getsockopt = sock_common_getsockopt,
746 .sendmsg = inet_sendmsg,
747 .recvmsg = sock_common_recvmsg,
748 .mmap = sock_no_mmap,
749 .sendpage = sock_no_sendpage,
751 .compat_setsockopt = compat_sock_common_setsockopt,
752 .compat_getsockopt = compat_sock_common_getsockopt,
756 static struct inet_protosw l2tp_ip6_protosw = {
758 .protocol = IPPROTO_L2TP,
759 .prot = &l2tp_ip6_prot,
760 .ops = &l2tp_ip6_ops,
763 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
764 .handler = l2tp_ip6_recv,
767 static int __init l2tp_ip6_init(void)
771 pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
773 err = proto_register(&l2tp_ip6_prot, 1);
777 err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
781 inet6_register_protosw(&l2tp_ip6_protosw);
785 proto_unregister(&l2tp_ip6_prot);
790 static void __exit l2tp_ip6_exit(void)
792 inet6_unregister_protosw(&l2tp_ip6_protosw);
793 inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
794 proto_unregister(&l2tp_ip6_prot);
797 module_init(l2tp_ip6_init);
798 module_exit(l2tp_ip6_exit);
800 MODULE_LICENSE("GPL");
801 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
802 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
803 MODULE_VERSION("1.0");
805 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
808 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
809 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);