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 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)
245 ip6_flush_pending_frames(sk);
248 inet6_destroy_sock(sk);
251 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
253 struct inet_sock *inet = inet_sk(sk);
254 struct ipv6_pinfo *np = inet6_sk(sk);
255 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
260 if (!sock_flag(sk, SOCK_ZAPPED))
262 if (addr->l2tp_family != AF_INET6)
264 if (addr_len < sizeof(*addr))
267 addr_type = ipv6_addr_type(&addr->l2tp_addr);
269 /* l2tp_ip6 sockets are IPv6 only */
270 if (addr_type == IPV6_ADDR_MAPPED)
271 return -EADDRNOTAVAIL;
273 /* L2TP is point-point, not multicast */
274 if (addr_type & IPV6_ADDR_MULTICAST)
275 return -EADDRNOTAVAIL;
278 read_lock_bh(&l2tp_ip6_lock);
279 if (__l2tp_ip6_bind_lookup(&init_net, &addr->l2tp_addr,
280 sk->sk_bound_dev_if, addr->l2tp_conn_id))
282 read_unlock_bh(&l2tp_ip6_lock);
287 if (sk->sk_state != TCP_CLOSE)
290 /* Check if the address belongs to the host. */
292 if (addr_type != IPV6_ADDR_ANY) {
293 struct net_device *dev = NULL;
295 if (addr_type & IPV6_ADDR_LINKLOCAL) {
296 if (addr_len >= sizeof(struct sockaddr_in6) &&
297 addr->l2tp_scope_id) {
298 /* Override any existing binding, if another
299 * one is supplied by user.
301 sk->sk_bound_dev_if = addr->l2tp_scope_id;
304 /* Binding to link-local address requires an
306 if (!sk->sk_bound_dev_if)
310 dev = dev_get_by_index_rcu(sock_net(sk),
311 sk->sk_bound_dev_if);
316 /* ipv4 addr of the socket is invalid. Only the
317 * unspecified and mapped address have a v4 equivalent.
319 v4addr = LOOPBACK4_IPV6;
320 err = -EADDRNOTAVAIL;
321 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
326 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
327 np->rcv_saddr = addr->l2tp_addr;
328 np->saddr = addr->l2tp_addr;
330 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
332 write_lock_bh(&l2tp_ip6_lock);
333 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
334 sk_del_node_init(sk);
335 write_unlock_bh(&l2tp_ip6_lock);
337 sock_reset_flag(sk, SOCK_ZAPPED);
348 read_unlock_bh(&l2tp_ip6_lock);
352 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
355 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
356 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
357 struct in6_addr *daddr;
361 if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
364 if (addr_len < sizeof(*lsa))
367 addr_type = ipv6_addr_type(&usin->sin6_addr);
368 if (addr_type & IPV6_ADDR_MULTICAST)
371 if (addr_type & IPV6_ADDR_MAPPED) {
372 daddr = &usin->sin6_addr;
373 if (ipv4_is_multicast(daddr->s6_addr32[3]))
377 rc = ip6_datagram_connect(sk, uaddr, addr_len);
381 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
383 write_lock_bh(&l2tp_ip6_lock);
384 hlist_del_init(&sk->sk_bind_node);
385 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
386 write_unlock_bh(&l2tp_ip6_lock);
393 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
395 if (sock_flag(sk, SOCK_ZAPPED))
398 return udp_disconnect(sk, flags);
401 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
402 int *uaddr_len, int peer)
404 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
405 struct sock *sk = sock->sk;
406 struct ipv6_pinfo *np = inet6_sk(sk);
407 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
409 lsa->l2tp_family = AF_INET6;
410 lsa->l2tp_flowinfo = 0;
411 lsa->l2tp_scope_id = 0;
412 lsa->l2tp_unused = 0;
414 if (!lsk->peer_conn_id)
416 lsa->l2tp_conn_id = lsk->peer_conn_id;
417 lsa->l2tp_addr = np->daddr;
419 lsa->l2tp_flowinfo = np->flow_label;
421 if (ipv6_addr_any(&np->rcv_saddr))
422 lsa->l2tp_addr = np->saddr;
424 lsa->l2tp_addr = np->rcv_saddr;
426 lsa->l2tp_conn_id = lsk->conn_id;
428 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
429 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
430 *uaddr_len = sizeof(*lsa);
434 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
438 /* Charge it to the socket, dropping if the queue is full. */
439 rc = sock_queue_rcv_skb(sk, skb);
446 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
451 static int l2tp_ip6_push_pending_frames(struct sock *sk)
454 __be32 *transhdr = NULL;
457 skb = skb_peek(&sk->sk_write_queue);
461 transhdr = (__be32 *)skb_transport_header(skb);
464 err = ip6_push_pending_frames(sk);
470 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
473 static int l2tp_ip6_sendmsg(struct kiocb *iocb, struct sock *sk,
474 struct msghdr *msg, size_t len)
476 struct ipv6_txoptions opt_space;
477 struct sockaddr_l2tpip6 *lsa =
478 (struct sockaddr_l2tpip6 *) msg->msg_name;
479 struct in6_addr *daddr, *final_p, final;
480 struct ipv6_pinfo *np = inet6_sk(sk);
481 struct ipv6_txoptions *opt = NULL;
482 struct ip6_flowlabel *flowlabel = NULL;
483 struct dst_entry *dst = NULL;
485 int addr_len = msg->msg_namelen;
489 int transhdrlen = 4; /* zero session-id */
490 int ulen = len + transhdrlen;
493 /* Rough check on arithmetic overflow,
494 better check is made in ip6_append_data().
499 /* Mirror BSD error message compatibility */
500 if (msg->msg_flags & MSG_OOB)
504 * Get and verify the address.
506 memset(&fl6, 0, sizeof(fl6));
508 fl6.flowi6_mark = sk->sk_mark;
511 if (addr_len < SIN6_LEN_RFC2133)
514 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
515 return -EAFNOSUPPORT;
517 daddr = &lsa->l2tp_addr;
519 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
520 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
521 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
522 if (flowlabel == NULL)
524 daddr = &flowlabel->dst;
529 * Otherwise it will be difficult to maintain
532 if (sk->sk_state == TCP_ESTABLISHED &&
533 ipv6_addr_equal(daddr, &np->daddr))
536 if (addr_len >= sizeof(struct sockaddr_in6) &&
537 lsa->l2tp_scope_id &&
538 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
539 fl6.flowi6_oif = lsa->l2tp_scope_id;
541 if (sk->sk_state != TCP_ESTABLISHED)
542 return -EDESTADDRREQ;
545 fl6.flowlabel = np->flow_label;
548 if (fl6.flowi6_oif == 0)
549 fl6.flowi6_oif = sk->sk_bound_dev_if;
551 if (msg->msg_controllen) {
553 memset(opt, 0, sizeof(struct ipv6_txoptions));
554 opt->tot_len = sizeof(struct ipv6_txoptions);
556 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
557 &hlimit, &tclass, &dontfrag);
559 fl6_sock_release(flowlabel);
562 if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
563 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
564 if (flowlabel == NULL)
567 if (!(opt->opt_nflen|opt->opt_flen))
574 opt = fl6_merge_options(&opt_space, flowlabel, opt);
575 opt = ipv6_fixup_options(&opt_space, opt);
577 fl6.flowi6_proto = sk->sk_protocol;
578 if (!ipv6_addr_any(daddr))
581 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
582 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
583 fl6.saddr = np->saddr;
585 final_p = fl6_update_dst(&fl6, opt, &final);
587 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
588 fl6.flowi6_oif = np->mcast_oif;
589 else if (!fl6.flowi6_oif)
590 fl6.flowi6_oif = np->ucast_oif;
592 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
594 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
601 if (ipv6_addr_is_multicast(&fl6.daddr))
602 hlimit = np->mcast_hops;
604 hlimit = np->hop_limit;
606 hlimit = ip6_dst_hoplimit(dst);
613 dontfrag = np->dontfrag;
615 if (msg->msg_flags & MSG_CONFIRM)
620 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
621 ulen, transhdrlen, hlimit, tclass, opt,
622 &fl6, (struct rt6_info *)dst,
623 msg->msg_flags, dontfrag);
625 ip6_flush_pending_frames(sk);
626 else if (!(msg->msg_flags & MSG_MORE))
627 err = l2tp_ip6_push_pending_frames(sk);
632 fl6_sock_release(flowlabel);
634 return err < 0 ? err : len;
638 if (!(msg->msg_flags & MSG_PROBE) || len)
639 goto back_from_confirm;
644 static int l2tp_ip6_recvmsg(struct kiocb *iocb, struct sock *sk,
645 struct msghdr *msg, size_t len, int noblock,
646 int flags, int *addr_len)
648 struct ipv6_pinfo *np = inet6_sk(sk);
649 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)msg->msg_name;
651 int err = -EOPNOTSUPP;
658 *addr_len = sizeof(*lsa);
660 if (flags & MSG_ERRQUEUE)
661 return ipv6_recv_error(sk, msg, len);
663 skb = skb_recv_datagram(sk, flags, noblock, &err);
669 msg->msg_flags |= MSG_TRUNC;
673 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
677 sock_recv_timestamp(msg, sk, skb);
679 /* Copy the address. */
681 lsa->l2tp_family = AF_INET6;
682 lsa->l2tp_unused = 0;
683 lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
684 lsa->l2tp_flowinfo = 0;
685 lsa->l2tp_scope_id = 0;
686 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
687 lsa->l2tp_scope_id = IP6CB(skb)->iif;
691 ip6_datagram_recv_ctl(sk, msg, skb);
693 if (flags & MSG_TRUNC)
696 skb_free_datagram(sk, skb);
698 return err ? err : copied;
701 static struct proto l2tp_ip6_prot = {
703 .owner = THIS_MODULE,
704 .init = l2tp_ip6_open,
705 .close = l2tp_ip6_close,
706 .bind = l2tp_ip6_bind,
707 .connect = l2tp_ip6_connect,
708 .disconnect = l2tp_ip6_disconnect,
710 .destroy = l2tp_ip6_destroy_sock,
711 .setsockopt = ipv6_setsockopt,
712 .getsockopt = ipv6_getsockopt,
713 .sendmsg = l2tp_ip6_sendmsg,
714 .recvmsg = l2tp_ip6_recvmsg,
715 .backlog_rcv = l2tp_ip6_backlog_recv,
717 .unhash = inet_unhash,
718 .obj_size = sizeof(struct l2tp_ip6_sock),
720 .compat_setsockopt = compat_ipv6_setsockopt,
721 .compat_getsockopt = compat_ipv6_getsockopt,
725 static const struct proto_ops l2tp_ip6_ops = {
727 .owner = THIS_MODULE,
728 .release = inet6_release,
730 .connect = inet_dgram_connect,
731 .socketpair = sock_no_socketpair,
732 .accept = sock_no_accept,
733 .getname = l2tp_ip6_getname,
734 .poll = datagram_poll,
735 .ioctl = inet6_ioctl,
736 .listen = sock_no_listen,
737 .shutdown = inet_shutdown,
738 .setsockopt = sock_common_setsockopt,
739 .getsockopt = sock_common_getsockopt,
740 .sendmsg = inet_sendmsg,
741 .recvmsg = sock_common_recvmsg,
742 .mmap = sock_no_mmap,
743 .sendpage = sock_no_sendpage,
745 .compat_setsockopt = compat_sock_common_setsockopt,
746 .compat_getsockopt = compat_sock_common_getsockopt,
750 static struct inet_protosw l2tp_ip6_protosw = {
752 .protocol = IPPROTO_L2TP,
753 .prot = &l2tp_ip6_prot,
754 .ops = &l2tp_ip6_ops,
758 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
759 .handler = l2tp_ip6_recv,
762 static int __init l2tp_ip6_init(void)
766 pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
768 err = proto_register(&l2tp_ip6_prot, 1);
772 err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
776 inet6_register_protosw(&l2tp_ip6_protosw);
780 proto_unregister(&l2tp_ip6_prot);
785 static void __exit l2tp_ip6_exit(void)
787 inet6_unregister_protosw(&l2tp_ip6_protosw);
788 inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
789 proto_unregister(&l2tp_ip6_prot);
792 module_init(l2tp_ip6_init);
793 module_exit(l2tp_ip6_exit);
795 MODULE_LICENSE("GPL");
796 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
797 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
798 MODULE_VERSION("1.0");
800 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
803 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);