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(const struct net *net,
61 const struct in6_addr *laddr,
62 const struct in6_addr *raddr,
63 int dif, u32 tunnel_id)
67 sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
68 const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
69 const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
70 const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
72 if (!net_eq(sock_net(sk), net))
75 if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
78 if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
79 !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
82 if (!ipv6_addr_any(sk_raddr) && raddr &&
83 !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
86 if (l2tp->conn_id != 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;
142 if (!pskb_may_pull(skb, 4))
145 /* Point to L2TP header */
146 optr = ptr = skb->data;
147 session_id = ntohl(*((__be32 *) ptr));
150 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
151 * the session_id. If it is 0, the packet is a L2TP control
152 * frame and the session_id value can be discarded.
154 if (session_id == 0) {
159 /* Ok, this is a data packet. Lookup the session. */
160 session = l2tp_session_get(net, NULL, session_id);
164 tunnel = session->tunnel;
168 /* Trace packet contents, if enabled */
169 if (tunnel->debug & L2TP_MSG_DATA) {
170 length = min(32u, skb->len);
171 if (!pskb_may_pull(skb, length))
174 /* Point to L2TP header */
175 optr = ptr = skb->data;
177 pr_debug("%s: ip recv\n", tunnel->name);
178 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
181 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
182 tunnel->recv_payload_hook);
183 l2tp_session_dec_refcount(session);
188 /* Get the tunnel_id from the L2TP header */
189 if (!pskb_may_pull(skb, 12))
192 if ((skb->data[0] & 0xc0) != 0xc0)
195 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
198 read_lock_bh(&l2tp_ip6_lock);
199 sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
200 inet6_iif(skb), tunnel_id);
202 read_unlock_bh(&l2tp_ip6_lock);
206 read_unlock_bh(&l2tp_ip6_lock);
208 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
213 return sk_receive_skb(sk, skb, 1);
216 l2tp_session_dec_refcount(session);
227 static int l2tp_ip6_open(struct sock *sk)
229 /* Prevent autobind. We don't have ports. */
230 inet_sk(sk)->inet_num = IPPROTO_L2TP;
232 write_lock_bh(&l2tp_ip6_lock);
233 sk_add_node(sk, &l2tp_ip6_table);
234 write_unlock_bh(&l2tp_ip6_lock);
239 static void l2tp_ip6_close(struct sock *sk, long timeout)
241 write_lock_bh(&l2tp_ip6_lock);
242 hlist_del_init(&sk->sk_bind_node);
243 sk_del_node_init(sk);
244 write_unlock_bh(&l2tp_ip6_lock);
246 sk_common_release(sk);
249 static void l2tp_ip6_destroy_sock(struct sock *sk)
251 struct l2tp_tunnel *tunnel = sk->sk_user_data;
254 ip6_flush_pending_frames(sk);
258 l2tp_tunnel_delete(tunnel);
260 inet6_destroy_sock(sk);
263 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
265 struct inet_sock *inet = inet_sk(sk);
266 struct ipv6_pinfo *np = inet6_sk(sk);
267 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
268 struct net *net = sock_net(sk);
274 if (addr->l2tp_family != AF_INET6)
276 if (addr_len < sizeof(*addr))
279 addr_type = ipv6_addr_type(&addr->l2tp_addr);
281 /* l2tp_ip6 sockets are IPv6 only */
282 if (addr_type == IPV6_ADDR_MAPPED)
283 return -EADDRNOTAVAIL;
285 /* L2TP is point-point, not multicast */
286 if (addr_type & IPV6_ADDR_MULTICAST)
287 return -EADDRNOTAVAIL;
292 if (!sock_flag(sk, SOCK_ZAPPED))
295 if (sk->sk_state != TCP_CLOSE)
298 bound_dev_if = sk->sk_bound_dev_if;
300 /* Check if the address belongs to the host. */
302 if (addr_type != IPV6_ADDR_ANY) {
303 struct net_device *dev = NULL;
305 if (addr_type & IPV6_ADDR_LINKLOCAL) {
306 if (addr->l2tp_scope_id)
307 bound_dev_if = addr->l2tp_scope_id;
309 /* Binding to link-local address requires an
316 dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
321 /* ipv4 addr of the socket is invalid. Only the
322 * unspecified and mapped address have a v4 equivalent.
324 v4addr = LOOPBACK4_IPV6;
325 err = -EADDRNOTAVAIL;
326 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
331 write_lock_bh(&l2tp_ip6_lock);
332 if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
333 addr->l2tp_conn_id)) {
334 write_unlock_bh(&l2tp_ip6_lock);
339 inet->inet_saddr = v4addr;
340 inet->inet_rcv_saddr = v4addr;
341 sk->sk_bound_dev_if = bound_dev_if;
342 sk->sk_v6_rcv_saddr = addr->l2tp_addr;
343 np->saddr = addr->l2tp_addr;
345 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
347 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
348 sk_del_node_init(sk);
349 write_unlock_bh(&l2tp_ip6_lock);
351 sock_reset_flag(sk, SOCK_ZAPPED);
363 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
366 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
367 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
368 struct in6_addr *daddr;
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]))
390 /* Must bind first - autobinding does not work */
391 if (sock_flag(sk, SOCK_ZAPPED)) {
396 rc = __ip6_datagram_connect(sk, uaddr, addr_len);
400 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
402 write_lock_bh(&l2tp_ip6_lock);
403 hlist_del_init(&sk->sk_bind_node);
404 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
405 write_unlock_bh(&l2tp_ip6_lock);
413 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
415 if (sock_flag(sk, SOCK_ZAPPED))
418 return __udp_disconnect(sk, flags);
421 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
424 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
425 struct sock *sk = sock->sk;
426 struct ipv6_pinfo *np = inet6_sk(sk);
427 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
429 lsa->l2tp_family = AF_INET6;
430 lsa->l2tp_flowinfo = 0;
431 lsa->l2tp_scope_id = 0;
432 lsa->l2tp_unused = 0;
434 if (!lsk->peer_conn_id)
436 lsa->l2tp_conn_id = lsk->peer_conn_id;
437 lsa->l2tp_addr = sk->sk_v6_daddr;
439 lsa->l2tp_flowinfo = np->flow_label;
441 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
442 lsa->l2tp_addr = np->saddr;
444 lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
446 lsa->l2tp_conn_id = lsk->conn_id;
448 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
449 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
453 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
457 /* Charge it to the socket, dropping if the queue is full. */
458 rc = sock_queue_rcv_skb(sk, skb);
465 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
470 static int l2tp_ip6_push_pending_frames(struct sock *sk)
473 __be32 *transhdr = NULL;
476 skb = skb_peek(&sk->sk_write_queue);
480 transhdr = (__be32 *)skb_transport_header(skb);
483 err = ip6_push_pending_frames(sk);
489 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
492 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
494 struct ipv6_txoptions opt_space;
495 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
496 struct in6_addr *daddr, *final_p, final;
497 struct ipv6_pinfo *np = inet6_sk(sk);
498 struct ipv6_txoptions *opt_to_free = NULL;
499 struct ipv6_txoptions *opt = NULL;
500 struct ip6_flowlabel *flowlabel = NULL;
501 struct dst_entry *dst = NULL;
503 struct sockcm_cookie sockc_unused = {0};
504 struct ipcm6_cookie ipc6;
505 int addr_len = msg->msg_namelen;
506 int transhdrlen = 4; /* zero session-id */
507 int ulen = len + transhdrlen;
510 /* Rough check on arithmetic overflow,
511 better check is made in ip6_append_data().
516 /* Mirror BSD error message compatibility */
517 if (msg->msg_flags & MSG_OOB)
521 * Get and verify the address.
523 memset(&fl6, 0, sizeof(fl6));
525 fl6.flowi6_mark = sk->sk_mark;
526 fl6.flowi6_uid = sk->sk_uid;
533 if (addr_len < SIN6_LEN_RFC2133)
536 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
537 return -EAFNOSUPPORT;
539 daddr = &lsa->l2tp_addr;
541 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
542 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
543 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
544 if (flowlabel == NULL)
550 * Otherwise it will be difficult to maintain
553 if (sk->sk_state == TCP_ESTABLISHED &&
554 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
555 daddr = &sk->sk_v6_daddr;
557 if (addr_len >= sizeof(struct sockaddr_in6) &&
558 lsa->l2tp_scope_id &&
559 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
560 fl6.flowi6_oif = lsa->l2tp_scope_id;
562 if (sk->sk_state != TCP_ESTABLISHED)
563 return -EDESTADDRREQ;
565 daddr = &sk->sk_v6_daddr;
566 fl6.flowlabel = np->flow_label;
569 if (fl6.flowi6_oif == 0)
570 fl6.flowi6_oif = sk->sk_bound_dev_if;
572 if (msg->msg_controllen) {
574 memset(opt, 0, sizeof(struct ipv6_txoptions));
575 opt->tot_len = sizeof(struct ipv6_txoptions);
578 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6,
581 fl6_sock_release(flowlabel);
584 if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
585 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
586 if (flowlabel == NULL)
589 if (!(opt->opt_nflen|opt->opt_flen))
598 opt = fl6_merge_options(&opt_space, flowlabel, opt);
599 opt = ipv6_fixup_options(&opt_space, opt);
602 fl6.flowi6_proto = sk->sk_protocol;
603 if (!ipv6_addr_any(daddr))
606 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
607 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
608 fl6.saddr = np->saddr;
610 final_p = fl6_update_dst(&fl6, opt, &final);
612 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
613 fl6.flowi6_oif = np->mcast_oif;
614 else if (!fl6.flowi6_oif)
615 fl6.flowi6_oif = np->ucast_oif;
617 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
620 ipc6.tclass = np->tclass;
622 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
624 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
631 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
633 if (ipc6.dontfrag < 0)
634 ipc6.dontfrag = np->dontfrag;
636 if (msg->msg_flags & MSG_CONFIRM)
641 err = ip6_append_data(sk, ip_generic_getfrag, msg,
642 ulen, transhdrlen, &ipc6,
643 &fl6, (struct rt6_info *)dst,
644 msg->msg_flags, &sockc_unused);
646 ip6_flush_pending_frames(sk);
647 else if (!(msg->msg_flags & MSG_MORE))
648 err = l2tp_ip6_push_pending_frames(sk);
653 fl6_sock_release(flowlabel);
654 txopt_put(opt_to_free);
656 return err < 0 ? err : len;
659 if (msg->msg_flags & MSG_PROBE)
660 dst_confirm_neigh(dst, &fl6.daddr);
661 if (!(msg->msg_flags & MSG_PROBE) || len)
662 goto back_from_confirm;
667 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
668 int noblock, int flags, int *addr_len)
670 struct ipv6_pinfo *np = inet6_sk(sk);
671 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
673 int err = -EOPNOTSUPP;
680 *addr_len = sizeof(*lsa);
682 if (flags & MSG_ERRQUEUE)
683 return ipv6_recv_error(sk, msg, len, addr_len);
685 skb = skb_recv_datagram(sk, flags, noblock, &err);
691 msg->msg_flags |= MSG_TRUNC;
695 err = skb_copy_datagram_msg(skb, 0, msg, copied);
699 sock_recv_timestamp(msg, sk, skb);
701 /* Copy the address. */
703 lsa->l2tp_family = AF_INET6;
704 lsa->l2tp_unused = 0;
705 lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
706 lsa->l2tp_flowinfo = 0;
707 lsa->l2tp_scope_id = 0;
708 lsa->l2tp_conn_id = 0;
709 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
710 lsa->l2tp_scope_id = inet6_iif(skb);
714 ip6_datagram_recv_ctl(sk, msg, skb);
716 if (flags & MSG_TRUNC)
719 skb_free_datagram(sk, skb);
721 return err ? err : copied;
724 static struct proto l2tp_ip6_prot = {
726 .owner = THIS_MODULE,
727 .init = l2tp_ip6_open,
728 .close = l2tp_ip6_close,
729 .bind = l2tp_ip6_bind,
730 .connect = l2tp_ip6_connect,
731 .disconnect = l2tp_ip6_disconnect,
733 .destroy = l2tp_ip6_destroy_sock,
734 .setsockopt = ipv6_setsockopt,
735 .getsockopt = ipv6_getsockopt,
736 .sendmsg = l2tp_ip6_sendmsg,
737 .recvmsg = l2tp_ip6_recvmsg,
738 .backlog_rcv = l2tp_ip6_backlog_recv,
740 .unhash = inet_unhash,
741 .obj_size = sizeof(struct l2tp_ip6_sock),
743 .compat_setsockopt = compat_ipv6_setsockopt,
744 .compat_getsockopt = compat_ipv6_getsockopt,
748 static const struct proto_ops l2tp_ip6_ops = {
750 .owner = THIS_MODULE,
751 .release = inet6_release,
753 .connect = inet_dgram_connect,
754 .socketpair = sock_no_socketpair,
755 .accept = sock_no_accept,
756 .getname = l2tp_ip6_getname,
757 .poll = datagram_poll,
758 .ioctl = inet6_ioctl,
759 .listen = sock_no_listen,
760 .shutdown = inet_shutdown,
761 .setsockopt = sock_common_setsockopt,
762 .getsockopt = sock_common_getsockopt,
763 .sendmsg = inet_sendmsg,
764 .recvmsg = sock_common_recvmsg,
765 .mmap = sock_no_mmap,
766 .sendpage = sock_no_sendpage,
768 .compat_setsockopt = compat_sock_common_setsockopt,
769 .compat_getsockopt = compat_sock_common_getsockopt,
773 static struct inet_protosw l2tp_ip6_protosw = {
775 .protocol = IPPROTO_L2TP,
776 .prot = &l2tp_ip6_prot,
777 .ops = &l2tp_ip6_ops,
780 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
781 .handler = l2tp_ip6_recv,
784 static int __init l2tp_ip6_init(void)
788 pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
790 err = proto_register(&l2tp_ip6_prot, 1);
794 err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
798 inet6_register_protosw(&l2tp_ip6_protosw);
802 proto_unregister(&l2tp_ip6_prot);
807 static void __exit l2tp_ip6_exit(void)
809 inet6_unregister_protosw(&l2tp_ip6_protosw);
810 inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
811 proto_unregister(&l2tp_ip6_prot);
814 module_init(l2tp_ip6_init);
815 module_exit(l2tp_ip6_exit);
817 MODULE_LICENSE("GPL");
818 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
819 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
820 MODULE_VERSION("1.0");
822 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
825 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
826 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);