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 || !dif ||
77 sk->sk_bound_dev_if == dif))
86 static inline struct sock *l2tp_ip6_bind_lookup(struct net *net,
87 struct in6_addr *laddr,
88 int dif, u32 tunnel_id)
90 struct sock *sk = __l2tp_ip6_bind_lookup(net, laddr, dif, tunnel_id);
97 /* When processing receive frames, there are two cases to
98 * consider. Data frames consist of a non-zero session-id and an
99 * optional cookie. Control frames consist of a regular L2TP header
100 * preceded by 32-bits of zeros.
102 * L2TPv3 Session Header Over IP
105 * 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
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109 * | Cookie (optional, maximum 64 bits)...
110 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114 * L2TPv3 Control Message Header Over IP
117 * 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
118 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * | (32 bits of zeros) |
120 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
122 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
123 * | Control Connection ID |
124 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
126 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
128 * All control frames are passed to userspace.
130 static int l2tp_ip6_recv(struct sk_buff *skb)
132 struct net *net = dev_net(skb->dev);
136 unsigned char *ptr, *optr;
137 struct l2tp_session *session;
138 struct l2tp_tunnel *tunnel = NULL;
141 if (!pskb_may_pull(skb, 4))
144 /* Point to L2TP header */
145 optr = ptr = skb->data;
146 session_id = ntohl(*((__be32 *) ptr));
149 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
150 * the session_id. If it is 0, the packet is a L2TP control
151 * frame and the session_id value can be discarded.
153 if (session_id == 0) {
158 /* Ok, this is a data packet. Lookup the session. */
159 session = l2tp_session_find(net, NULL, session_id);
163 tunnel = session->tunnel;
167 /* Trace packet contents, if enabled */
168 if (tunnel->debug & L2TP_MSG_DATA) {
169 length = min(32u, skb->len);
170 if (!pskb_may_pull(skb, length))
173 /* Point to L2TP header */
174 optr = ptr = skb->data;
176 pr_debug("%s: ip recv\n", tunnel->name);
177 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
180 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
181 tunnel->recv_payload_hook);
185 /* Get the tunnel_id from the L2TP header */
186 if (!pskb_may_pull(skb, 12))
189 if ((skb->data[0] & 0xc0) != 0xc0)
192 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
193 tunnel = l2tp_tunnel_find(net, tunnel_id);
197 struct ipv6hdr *iph = ipv6_hdr(skb);
199 read_lock_bh(&l2tp_ip6_lock);
200 sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, inet6_iif(skb),
203 read_unlock_bh(&l2tp_ip6_lock);
208 read_unlock_bh(&l2tp_ip6_lock);
211 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
216 return sk_receive_skb(sk, skb, 1);
226 static int l2tp_ip6_open(struct sock *sk)
228 /* Prevent autobind. We don't have ports. */
229 inet_sk(sk)->inet_num = IPPROTO_L2TP;
231 write_lock_bh(&l2tp_ip6_lock);
232 sk_add_node(sk, &l2tp_ip6_table);
233 write_unlock_bh(&l2tp_ip6_lock);
238 static void l2tp_ip6_close(struct sock *sk, long timeout)
240 write_lock_bh(&l2tp_ip6_lock);
241 hlist_del_init(&sk->sk_bind_node);
242 sk_del_node_init(sk);
243 write_unlock_bh(&l2tp_ip6_lock);
245 sk_common_release(sk);
248 static void l2tp_ip6_destroy_sock(struct sock *sk)
250 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
253 ip6_flush_pending_frames(sk);
257 l2tp_tunnel_closeall(tunnel);
261 inet6_destroy_sock(sk);
264 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
266 struct inet_sock *inet = inet_sk(sk);
267 struct ipv6_pinfo *np = inet6_sk(sk);
268 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
269 struct net *net = sock_net(sk);
275 if (addr->l2tp_family != AF_INET6)
277 if (addr_len < sizeof(*addr))
280 addr_type = ipv6_addr_type(&addr->l2tp_addr);
282 /* l2tp_ip6 sockets are IPv6 only */
283 if (addr_type == IPV6_ADDR_MAPPED)
284 return -EADDRNOTAVAIL;
286 /* L2TP is point-point, not multicast */
287 if (addr_type & IPV6_ADDR_MULTICAST)
288 return -EADDRNOTAVAIL;
293 if (!sock_flag(sk, SOCK_ZAPPED))
296 if (sk->sk_state != TCP_CLOSE)
299 bound_dev_if = sk->sk_bound_dev_if;
301 /* Check if the address belongs to the host. */
303 if (addr_type != IPV6_ADDR_ANY) {
304 struct net_device *dev = NULL;
306 if (addr_type & IPV6_ADDR_LINKLOCAL) {
307 if (addr->l2tp_scope_id)
308 bound_dev_if = addr->l2tp_scope_id;
310 /* Binding to link-local address requires an
317 dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
322 /* ipv4 addr of the socket is invalid. Only the
323 * unspecified and mapped address have a v4 equivalent.
325 v4addr = LOOPBACK4_IPV6;
326 err = -EADDRNOTAVAIL;
327 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
332 write_lock_bh(&l2tp_ip6_lock);
333 if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, bound_dev_if,
334 addr->l2tp_conn_id)) {
335 write_unlock_bh(&l2tp_ip6_lock);
340 inet->inet_saddr = v4addr;
341 inet->inet_rcv_saddr = v4addr;
342 sk->sk_bound_dev_if = bound_dev_if;
343 sk->sk_v6_rcv_saddr = addr->l2tp_addr;
344 np->saddr = addr->l2tp_addr;
346 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
348 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
349 sk_del_node_init(sk);
350 write_unlock_bh(&l2tp_ip6_lock);
352 sock_reset_flag(sk, SOCK_ZAPPED);
364 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
367 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
368 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
369 struct in6_addr *daddr;
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]))
391 /* Must bind first - autobinding does not work */
392 if (sock_flag(sk, SOCK_ZAPPED)) {
397 rc = __ip6_datagram_connect(sk, uaddr, addr_len);
401 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
403 write_lock_bh(&l2tp_ip6_lock);
404 hlist_del_init(&sk->sk_bind_node);
405 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
406 write_unlock_bh(&l2tp_ip6_lock);
414 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
416 if (sock_flag(sk, SOCK_ZAPPED))
419 return __udp_disconnect(sk, flags);
422 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
423 int *uaddr_len, int peer)
425 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
426 struct sock *sk = sock->sk;
427 struct ipv6_pinfo *np = inet6_sk(sk);
428 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
430 lsa->l2tp_family = AF_INET6;
431 lsa->l2tp_flowinfo = 0;
432 lsa->l2tp_scope_id = 0;
433 lsa->l2tp_unused = 0;
435 if (!lsk->peer_conn_id)
437 lsa->l2tp_conn_id = lsk->peer_conn_id;
438 lsa->l2tp_addr = sk->sk_v6_daddr;
440 lsa->l2tp_flowinfo = np->flow_label;
442 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
443 lsa->l2tp_addr = np->saddr;
445 lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
447 lsa->l2tp_conn_id = lsk->conn_id;
449 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
450 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
451 *uaddr_len = sizeof(*lsa);
455 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
459 /* Charge it to the socket, dropping if the queue is full. */
460 rc = sock_queue_rcv_skb(sk, skb);
467 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
472 static int l2tp_ip6_push_pending_frames(struct sock *sk)
475 __be32 *transhdr = NULL;
478 skb = skb_peek(&sk->sk_write_queue);
482 transhdr = (__be32 *)skb_transport_header(skb);
485 err = ip6_push_pending_frames(sk);
491 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
494 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
496 struct ipv6_txoptions opt_space;
497 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
498 struct in6_addr *daddr, *final_p, final;
499 struct ipv6_pinfo *np = inet6_sk(sk);
500 struct ipv6_txoptions *opt_to_free = NULL;
501 struct ipv6_txoptions *opt = NULL;
502 struct ip6_flowlabel *flowlabel = NULL;
503 struct dst_entry *dst = NULL;
505 struct sockcm_cookie sockc_unused = {0};
506 struct ipcm6_cookie ipc6;
507 int addr_len = msg->msg_namelen;
508 int transhdrlen = 4; /* zero session-id */
509 int ulen = len + transhdrlen;
512 /* Rough check on arithmetic overflow,
513 better check is made in ip6_append_data().
518 /* Mirror BSD error message compatibility */
519 if (msg->msg_flags & MSG_OOB)
523 * Get and verify the address.
525 memset(&fl6, 0, sizeof(fl6));
527 fl6.flowi6_mark = sk->sk_mark;
528 fl6.flowi6_uid = sk->sk_uid;
535 if (addr_len < SIN6_LEN_RFC2133)
538 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
539 return -EAFNOSUPPORT;
541 daddr = &lsa->l2tp_addr;
543 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
544 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
545 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
546 if (flowlabel == NULL)
552 * Otherwise it will be difficult to maintain
555 if (sk->sk_state == TCP_ESTABLISHED &&
556 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
557 daddr = &sk->sk_v6_daddr;
559 if (addr_len >= sizeof(struct sockaddr_in6) &&
560 lsa->l2tp_scope_id &&
561 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
562 fl6.flowi6_oif = lsa->l2tp_scope_id;
564 if (sk->sk_state != TCP_ESTABLISHED)
565 return -EDESTADDRREQ;
567 daddr = &sk->sk_v6_daddr;
568 fl6.flowlabel = np->flow_label;
571 if (fl6.flowi6_oif == 0)
572 fl6.flowi6_oif = sk->sk_bound_dev_if;
574 if (msg->msg_controllen) {
576 memset(opt, 0, sizeof(struct ipv6_txoptions));
577 opt->tot_len = sizeof(struct ipv6_txoptions);
580 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6,
583 fl6_sock_release(flowlabel);
586 if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
587 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
588 if (flowlabel == NULL)
591 if (!(opt->opt_nflen|opt->opt_flen))
600 opt = fl6_merge_options(&opt_space, flowlabel, opt);
601 opt = ipv6_fixup_options(&opt_space, opt);
604 fl6.flowi6_proto = sk->sk_protocol;
605 if (!ipv6_addr_any(daddr))
608 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
609 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
610 fl6.saddr = np->saddr;
612 final_p = fl6_update_dst(&fl6, opt, &final);
614 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
615 fl6.flowi6_oif = np->mcast_oif;
616 else if (!fl6.flowi6_oif)
617 fl6.flowi6_oif = np->ucast_oif;
619 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
622 ipc6.tclass = np->tclass;
624 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
626 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
633 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
635 if (ipc6.dontfrag < 0)
636 ipc6.dontfrag = np->dontfrag;
638 if (msg->msg_flags & MSG_CONFIRM)
643 err = ip6_append_data(sk, ip_generic_getfrag, msg,
644 ulen, transhdrlen, &ipc6,
645 &fl6, (struct rt6_info *)dst,
646 msg->msg_flags, &sockc_unused);
648 ip6_flush_pending_frames(sk);
649 else if (!(msg->msg_flags & MSG_MORE))
650 err = l2tp_ip6_push_pending_frames(sk);
655 fl6_sock_release(flowlabel);
656 txopt_put(opt_to_free);
658 return err < 0 ? err : len;
662 if (!(msg->msg_flags & MSG_PROBE) || len)
663 goto back_from_confirm;
668 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
669 int noblock, int flags, int *addr_len)
671 struct ipv6_pinfo *np = inet6_sk(sk);
672 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
674 int err = -EOPNOTSUPP;
681 *addr_len = sizeof(*lsa);
683 if (flags & MSG_ERRQUEUE)
684 return ipv6_recv_error(sk, msg, len, addr_len);
686 skb = skb_recv_datagram(sk, flags, noblock, &err);
692 msg->msg_flags |= MSG_TRUNC;
696 err = skb_copy_datagram_msg(skb, 0, msg, copied);
700 sock_recv_timestamp(msg, sk, skb);
702 /* Copy the address. */
704 lsa->l2tp_family = AF_INET6;
705 lsa->l2tp_unused = 0;
706 lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
707 lsa->l2tp_flowinfo = 0;
708 lsa->l2tp_scope_id = 0;
709 lsa->l2tp_conn_id = 0;
710 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
711 lsa->l2tp_scope_id = inet6_iif(skb);
715 ip6_datagram_recv_ctl(sk, msg, skb);
717 if (flags & MSG_TRUNC)
720 skb_free_datagram(sk, skb);
722 return err ? err : copied;
725 static struct proto l2tp_ip6_prot = {
727 .owner = THIS_MODULE,
728 .init = l2tp_ip6_open,
729 .close = l2tp_ip6_close,
730 .bind = l2tp_ip6_bind,
731 .connect = l2tp_ip6_connect,
732 .disconnect = l2tp_ip6_disconnect,
734 .destroy = l2tp_ip6_destroy_sock,
735 .setsockopt = ipv6_setsockopt,
736 .getsockopt = ipv6_getsockopt,
737 .sendmsg = l2tp_ip6_sendmsg,
738 .recvmsg = l2tp_ip6_recvmsg,
739 .backlog_rcv = l2tp_ip6_backlog_recv,
741 .unhash = inet_unhash,
742 .obj_size = sizeof(struct l2tp_ip6_sock),
744 .compat_setsockopt = compat_ipv6_setsockopt,
745 .compat_getsockopt = compat_ipv6_getsockopt,
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 .listen = sock_no_listen,
761 .shutdown = inet_shutdown,
762 .setsockopt = sock_common_setsockopt,
763 .getsockopt = sock_common_getsockopt,
764 .sendmsg = inet_sendmsg,
765 .recvmsg = sock_common_recvmsg,
766 .mmap = sock_no_mmap,
767 .sendpage = sock_no_sendpage,
769 .compat_setsockopt = compat_sock_common_setsockopt,
770 .compat_getsockopt = compat_sock_common_getsockopt,
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);