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 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 struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
75 if ((l2tp->conn_id == tunnel_id) &&
76 net_eq(sock_net(sk), net) &&
77 (!sk_laddr || ipv6_addr_any(sk_laddr) || ipv6_addr_equal(sk_laddr, laddr)) &&
78 (!raddr || ipv6_addr_any(sk_raddr) || ipv6_addr_equal(sk_raddr, raddr)) &&
79 (!sk->sk_bound_dev_if || !dif ||
80 sk->sk_bound_dev_if == dif))
89 /* When processing receive frames, there are two cases to
90 * consider. Data frames consist of a non-zero session-id and an
91 * optional cookie. Control frames consist of a regular L2TP header
92 * preceded by 32-bits of zeros.
94 * L2TPv3 Session Header Over IP
97 * 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
98 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
101 * | Cookie (optional, maximum 64 bits)...
102 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * L2TPv3 Control Message Header Over IP
109 * 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
110 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111 * | (32 bits of zeros) |
112 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
114 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115 * | Control Connection ID |
116 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
118 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
120 * All control frames are passed to userspace.
122 static int l2tp_ip6_recv(struct sk_buff *skb)
124 struct net *net = dev_net(skb->dev);
128 unsigned char *ptr, *optr;
129 struct l2tp_session *session;
130 struct l2tp_tunnel *tunnel = NULL;
133 if (!pskb_may_pull(skb, 4))
136 /* Point to L2TP header */
137 optr = ptr = skb->data;
138 session_id = ntohl(*((__be32 *) ptr));
141 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
142 * the session_id. If it is 0, the packet is a L2TP control
143 * frame and the session_id value can be discarded.
145 if (session_id == 0) {
150 /* Ok, this is a data packet. Lookup the session. */
151 session = l2tp_session_get(net, NULL, session_id, true);
155 tunnel = session->tunnel;
159 /* Trace packet contents, if enabled */
160 if (tunnel->debug & L2TP_MSG_DATA) {
161 length = min(32u, skb->len);
162 if (!pskb_may_pull(skb, length))
165 /* Point to L2TP header */
166 optr = ptr = skb->data;
168 pr_debug("%s: ip recv\n", tunnel->name);
169 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
172 if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
175 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
176 tunnel->recv_payload_hook);
177 l2tp_session_dec_refcount(session);
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(net, tunnel_id);
195 struct ipv6hdr *iph = ipv6_hdr(skb);
197 read_lock_bh(&l2tp_ip6_lock);
198 sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
199 inet6_iif(skb), tunnel_id);
201 read_unlock_bh(&l2tp_ip6_lock);
206 read_unlock_bh(&l2tp_ip6_lock);
209 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
214 return sk_receive_skb(sk, skb, 1);
218 session->deref(session);
219 l2tp_session_dec_refcount(session);
230 static int l2tp_ip6_open(struct sock *sk)
232 /* Prevent autobind. We don't have ports. */
233 inet_sk(sk)->inet_num = IPPROTO_L2TP;
235 write_lock_bh(&l2tp_ip6_lock);
236 sk_add_node(sk, &l2tp_ip6_table);
237 write_unlock_bh(&l2tp_ip6_lock);
242 static void l2tp_ip6_close(struct sock *sk, long timeout)
244 write_lock_bh(&l2tp_ip6_lock);
245 hlist_del_init(&sk->sk_bind_node);
246 sk_del_node_init(sk);
247 write_unlock_bh(&l2tp_ip6_lock);
249 sk_common_release(sk);
252 static void l2tp_ip6_destroy_sock(struct sock *sk)
254 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
257 ip6_flush_pending_frames(sk);
261 l2tp_tunnel_closeall(tunnel);
265 inet6_destroy_sock(sk);
268 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
270 struct inet_sock *inet = inet_sk(sk);
271 struct ipv6_pinfo *np = inet6_sk(sk);
272 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
273 struct net *net = sock_net(sk);
279 if (addr->l2tp_family != AF_INET6)
281 if (addr_len < sizeof(*addr))
284 addr_type = ipv6_addr_type(&addr->l2tp_addr);
286 /* l2tp_ip6 sockets are IPv6 only */
287 if (addr_type == IPV6_ADDR_MAPPED)
288 return -EADDRNOTAVAIL;
290 /* L2TP is point-point, not multicast */
291 if (addr_type & IPV6_ADDR_MULTICAST)
292 return -EADDRNOTAVAIL;
297 if (!sock_flag(sk, SOCK_ZAPPED))
300 if (sk->sk_state != TCP_CLOSE)
303 bound_dev_if = sk->sk_bound_dev_if;
305 /* Check if the address belongs to the host. */
307 if (addr_type != IPV6_ADDR_ANY) {
308 struct net_device *dev = NULL;
310 if (addr_type & IPV6_ADDR_LINKLOCAL) {
311 if (addr->l2tp_scope_id)
312 bound_dev_if = addr->l2tp_scope_id;
314 /* Binding to link-local address requires an
321 dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
326 /* ipv4 addr of the socket is invalid. Only the
327 * unspecified and mapped address have a v4 equivalent.
329 v4addr = LOOPBACK4_IPV6;
330 err = -EADDRNOTAVAIL;
331 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
336 write_lock_bh(&l2tp_ip6_lock);
337 if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
338 addr->l2tp_conn_id)) {
339 write_unlock_bh(&l2tp_ip6_lock);
344 inet->inet_saddr = v4addr;
345 inet->inet_rcv_saddr = v4addr;
346 sk->sk_bound_dev_if = bound_dev_if;
347 sk->sk_v6_rcv_saddr = addr->l2tp_addr;
348 np->saddr = addr->l2tp_addr;
350 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
352 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
353 sk_del_node_init(sk);
354 write_unlock_bh(&l2tp_ip6_lock);
356 sock_reset_flag(sk, SOCK_ZAPPED);
368 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
371 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
372 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
373 struct in6_addr *daddr;
377 if (addr_len < sizeof(*lsa))
380 if (usin->sin6_family != AF_INET6)
383 addr_type = ipv6_addr_type(&usin->sin6_addr);
384 if (addr_type & IPV6_ADDR_MULTICAST)
387 if (addr_type & IPV6_ADDR_MAPPED) {
388 daddr = &usin->sin6_addr;
389 if (ipv4_is_multicast(daddr->s6_addr32[3]))
395 /* Must bind first - autobinding does not work */
396 if (sock_flag(sk, SOCK_ZAPPED)) {
401 rc = __ip6_datagram_connect(sk, uaddr, addr_len);
405 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
407 write_lock_bh(&l2tp_ip6_lock);
408 hlist_del_init(&sk->sk_bind_node);
409 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
410 write_unlock_bh(&l2tp_ip6_lock);
418 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
420 if (sock_flag(sk, SOCK_ZAPPED))
423 return __udp_disconnect(sk, flags);
426 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
427 int *uaddr_len, int peer)
429 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
430 struct sock *sk = sock->sk;
431 struct ipv6_pinfo *np = inet6_sk(sk);
432 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
434 lsa->l2tp_family = AF_INET6;
435 lsa->l2tp_flowinfo = 0;
436 lsa->l2tp_scope_id = 0;
437 lsa->l2tp_unused = 0;
439 if (!lsk->peer_conn_id)
441 lsa->l2tp_conn_id = lsk->peer_conn_id;
442 lsa->l2tp_addr = sk->sk_v6_daddr;
444 lsa->l2tp_flowinfo = np->flow_label;
446 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
447 lsa->l2tp_addr = np->saddr;
449 lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
451 lsa->l2tp_conn_id = lsk->conn_id;
453 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
454 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
455 *uaddr_len = sizeof(*lsa);
459 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
463 /* Charge it to the socket, dropping if the queue is full. */
464 rc = sock_queue_rcv_skb(sk, skb);
471 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
476 static int l2tp_ip6_push_pending_frames(struct sock *sk)
479 __be32 *transhdr = NULL;
482 skb = skb_peek(&sk->sk_write_queue);
486 transhdr = (__be32 *)skb_transport_header(skb);
489 err = ip6_push_pending_frames(sk);
495 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
498 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
500 struct ipv6_txoptions opt_space;
501 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
502 struct in6_addr *daddr, *final_p, final;
503 struct ipv6_pinfo *np = inet6_sk(sk);
504 struct ipv6_txoptions *opt_to_free = NULL;
505 struct ipv6_txoptions *opt = NULL;
506 struct ip6_flowlabel *flowlabel = NULL;
507 struct dst_entry *dst = NULL;
509 struct sockcm_cookie sockc_unused = {0};
510 struct ipcm6_cookie ipc6;
511 int addr_len = msg->msg_namelen;
512 int transhdrlen = 4; /* zero session-id */
513 int ulen = len + transhdrlen;
516 /* Rough check on arithmetic overflow,
517 better check is made in ip6_append_data().
522 /* Mirror BSD error message compatibility */
523 if (msg->msg_flags & MSG_OOB)
527 * Get and verify the address.
529 memset(&fl6, 0, sizeof(fl6));
531 fl6.flowi6_mark = sk->sk_mark;
538 if (addr_len < SIN6_LEN_RFC2133)
541 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
542 return -EAFNOSUPPORT;
544 daddr = &lsa->l2tp_addr;
546 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
547 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
548 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
549 if (flowlabel == NULL)
555 * Otherwise it will be difficult to maintain
558 if (sk->sk_state == TCP_ESTABLISHED &&
559 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
560 daddr = &sk->sk_v6_daddr;
562 if (addr_len >= sizeof(struct sockaddr_in6) &&
563 lsa->l2tp_scope_id &&
564 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
565 fl6.flowi6_oif = lsa->l2tp_scope_id;
567 if (sk->sk_state != TCP_ESTABLISHED)
568 return -EDESTADDRREQ;
570 daddr = &sk->sk_v6_daddr;
571 fl6.flowlabel = np->flow_label;
574 if (fl6.flowi6_oif == 0)
575 fl6.flowi6_oif = sk->sk_bound_dev_if;
577 if (msg->msg_controllen) {
579 memset(opt, 0, sizeof(struct ipv6_txoptions));
580 opt->tot_len = sizeof(struct ipv6_txoptions);
583 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6,
586 fl6_sock_release(flowlabel);
589 if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
590 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
591 if (flowlabel == NULL)
594 if (!(opt->opt_nflen|opt->opt_flen))
603 opt = fl6_merge_options(&opt_space, flowlabel, opt);
604 opt = ipv6_fixup_options(&opt_space, opt);
607 fl6.flowi6_proto = sk->sk_protocol;
608 if (!ipv6_addr_any(daddr))
611 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
612 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
613 fl6.saddr = np->saddr;
615 final_p = fl6_update_dst(&fl6, opt, &final);
617 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
618 fl6.flowi6_oif = np->mcast_oif;
619 else if (!fl6.flowi6_oif)
620 fl6.flowi6_oif = np->ucast_oif;
622 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
625 ipc6.tclass = np->tclass;
627 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
629 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
636 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
638 if (ipc6.dontfrag < 0)
639 ipc6.dontfrag = np->dontfrag;
641 if (msg->msg_flags & MSG_CONFIRM)
646 err = ip6_append_data(sk, ip_generic_getfrag, msg,
647 ulen, transhdrlen, &ipc6,
648 &fl6, (struct rt6_info *)dst,
649 msg->msg_flags, &sockc_unused);
651 ip6_flush_pending_frames(sk);
652 else if (!(msg->msg_flags & MSG_MORE))
653 err = l2tp_ip6_push_pending_frames(sk);
658 fl6_sock_release(flowlabel);
659 txopt_put(opt_to_free);
661 return err < 0 ? err : len;
665 if (!(msg->msg_flags & MSG_PROBE) || len)
666 goto back_from_confirm;
671 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
672 int noblock, int flags, int *addr_len)
674 struct ipv6_pinfo *np = inet6_sk(sk);
675 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
677 int err = -EOPNOTSUPP;
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);
712 *addr_len = sizeof(*lsa);
716 ip6_datagram_recv_ctl(sk, msg, skb);
718 if (flags & MSG_TRUNC)
721 skb_free_datagram(sk, skb);
723 return err ? err : copied;
726 static struct proto l2tp_ip6_prot = {
728 .owner = THIS_MODULE,
729 .init = l2tp_ip6_open,
730 .close = l2tp_ip6_close,
731 .bind = l2tp_ip6_bind,
732 .connect = l2tp_ip6_connect,
733 .disconnect = l2tp_ip6_disconnect,
735 .destroy = l2tp_ip6_destroy_sock,
736 .setsockopt = ipv6_setsockopt,
737 .getsockopt = ipv6_getsockopt,
738 .sendmsg = l2tp_ip6_sendmsg,
739 .recvmsg = l2tp_ip6_recvmsg,
740 .backlog_rcv = l2tp_ip6_backlog_recv,
742 .unhash = inet_unhash,
743 .obj_size = sizeof(struct l2tp_ip6_sock),
745 .compat_setsockopt = compat_ipv6_setsockopt,
746 .compat_getsockopt = compat_ipv6_getsockopt,
750 static const struct proto_ops l2tp_ip6_ops = {
752 .owner = THIS_MODULE,
753 .release = inet6_release,
755 .connect = inet_dgram_connect,
756 .socketpair = sock_no_socketpair,
757 .accept = sock_no_accept,
758 .getname = l2tp_ip6_getname,
759 .poll = datagram_poll,
760 .ioctl = inet6_ioctl,
761 .listen = sock_no_listen,
762 .shutdown = inet_shutdown,
763 .setsockopt = sock_common_setsockopt,
764 .getsockopt = sock_common_getsockopt,
765 .sendmsg = inet_sendmsg,
766 .recvmsg = sock_common_recvmsg,
767 .mmap = sock_no_mmap,
768 .sendpage = sock_no_sendpage,
770 .compat_setsockopt = compat_sock_common_setsockopt,
771 .compat_getsockopt = compat_sock_common_getsockopt,
775 static struct inet_protosw l2tp_ip6_protosw = {
777 .protocol = IPPROTO_L2TP,
778 .prot = &l2tp_ip6_prot,
779 .ops = &l2tp_ip6_ops,
782 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
783 .handler = l2tp_ip6_recv,
786 static int __init l2tp_ip6_init(void)
790 pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
792 err = proto_register(&l2tp_ip6_prot, 1);
796 err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
800 inet6_register_protosw(&l2tp_ip6_protosw);
804 proto_unregister(&l2tp_ip6_prot);
809 static void __exit l2tp_ip6_exit(void)
811 inet6_unregister_protosw(&l2tp_ip6_protosw);
812 inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
813 proto_unregister(&l2tp_ip6_prot);
816 module_init(l2tp_ip6_init);
817 module_exit(l2tp_ip6_exit);
819 MODULE_LICENSE("GPL");
820 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
821 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
822 MODULE_VERSION("1.0");
824 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
827 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
828 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);