1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* L2TPv3 IP encapsulation support for IPv6
4 * Copyright (c) 2012 Katalix Systems Ltd
7 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 #include <linux/icmp.h>
10 #include <linux/module.h>
11 #include <linux/skbuff.h>
12 #include <linux/random.h>
13 #include <linux/socket.h>
14 #include <linux/l2tp.h>
16 #include <linux/in6.h>
21 #include <net/inet_common.h>
22 #include <net/tcp_states.h>
23 #include <net/protocol.h>
26 #include <net/transp_v6.h>
27 #include <net/addrconf.h>
28 #include <net/ip6_route.h>
30 #include "l2tp_core.h"
32 struct l2tp_ip6_sock {
33 /* inet_sock has to be the first member of l2tp_ip6_sock */
34 struct inet_sock inet;
39 /* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
42 struct ipv6_pinfo inet6;
45 static DEFINE_RWLOCK(l2tp_ip6_lock);
46 static struct hlist_head l2tp_ip6_table;
47 static struct hlist_head l2tp_ip6_bind_table;
49 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
51 return (struct l2tp_ip6_sock *)sk;
54 static struct sock *__l2tp_ip6_bind_lookup(const struct net *net,
55 const struct in6_addr *laddr,
56 const struct in6_addr *raddr,
57 int dif, u32 tunnel_id)
61 sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
62 const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
63 const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
64 const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
66 if (!net_eq(sock_net(sk), net))
69 if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
72 if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
73 !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
76 if (!ipv6_addr_any(sk_raddr) && raddr &&
77 !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
80 if (l2tp->conn_id != tunnel_id)
91 /* When processing receive frames, there are two cases to
92 * consider. Data frames consist of a non-zero session-id and an
93 * optional cookie. Control frames consist of a regular L2TP header
94 * preceded by 32-bits of zeros.
96 * L2TPv3 Session Header Over IP
99 * 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
100 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
102 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103 * | Cookie (optional, maximum 64 bits)...
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * L2TPv3 Control Message Header Over IP
111 * 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
112 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113 * | (32 bits of zeros) |
114 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
116 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117 * | Control Connection ID |
118 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
120 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
122 * All control frames are passed to userspace.
124 static int l2tp_ip6_recv(struct sk_buff *skb)
126 struct net *net = dev_net(skb->dev);
130 unsigned char *ptr, *optr;
131 struct l2tp_session *session;
132 struct l2tp_tunnel *tunnel = NULL;
136 if (!pskb_may_pull(skb, 4))
139 /* Point to L2TP header */
140 optr = ptr = skb->data;
141 session_id = ntohl(*((__be32 *)ptr));
144 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
145 * the session_id. If it is 0, the packet is a L2TP control
146 * frame and the session_id value can be discarded.
148 if (session_id == 0) {
153 /* Ok, this is a data packet. Lookup the session. */
154 session = l2tp_session_get(net, session_id);
158 tunnel = session->tunnel;
162 /* Trace packet contents, if enabled */
163 if (tunnel->debug & L2TP_MSG_DATA) {
164 length = min(32u, skb->len);
165 if (!pskb_may_pull(skb, length))
168 /* Point to L2TP header */
169 optr = ptr = skb->data;
171 pr_debug("%s: ip recv\n", tunnel->name);
172 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
175 if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
178 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
179 l2tp_session_dec_refcount(session);
184 /* Get the tunnel_id from the L2TP header */
185 if (!pskb_may_pull(skb, 12))
188 if ((skb->data[0] & 0xc0) != 0xc0)
191 tunnel_id = ntohl(*(__be32 *)&skb->data[4]);
194 read_lock_bh(&l2tp_ip6_lock);
195 sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
196 inet6_iif(skb), tunnel_id);
198 read_unlock_bh(&l2tp_ip6_lock);
202 read_unlock_bh(&l2tp_ip6_lock);
204 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
209 return sk_receive_skb(sk, skb, 1);
212 l2tp_session_dec_refcount(session);
223 static int l2tp_ip6_hash(struct sock *sk)
225 if (sk_unhashed(sk)) {
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_unhash(struct sock *sk)
237 write_lock_bh(&l2tp_ip6_lock);
238 sk_del_node_init(sk);
239 write_unlock_bh(&l2tp_ip6_lock);
242 static int l2tp_ip6_open(struct sock *sk)
244 /* Prevent autobind. We don't have ports. */
245 inet_sk(sk)->inet_num = IPPROTO_L2TP;
251 static void l2tp_ip6_close(struct sock *sk, long timeout)
253 write_lock_bh(&l2tp_ip6_lock);
254 hlist_del_init(&sk->sk_bind_node);
255 sk_del_node_init(sk);
256 write_unlock_bh(&l2tp_ip6_lock);
258 sk_common_release(sk);
261 static void l2tp_ip6_destroy_sock(struct sock *sk)
263 struct l2tp_tunnel *tunnel = sk->sk_user_data;
266 ip6_flush_pending_frames(sk);
270 l2tp_tunnel_delete(tunnel);
272 inet6_destroy_sock(sk);
275 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
277 struct inet_sock *inet = inet_sk(sk);
278 struct ipv6_pinfo *np = inet6_sk(sk);
279 struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *)uaddr;
280 struct net *net = sock_net(sk);
286 if (addr->l2tp_family != AF_INET6)
288 if (addr_len < sizeof(*addr))
291 addr_type = ipv6_addr_type(&addr->l2tp_addr);
293 /* l2tp_ip6 sockets are IPv6 only */
294 if (addr_type == IPV6_ADDR_MAPPED)
295 return -EADDRNOTAVAIL;
297 /* L2TP is point-point, not multicast */
298 if (addr_type & IPV6_ADDR_MULTICAST)
299 return -EADDRNOTAVAIL;
304 if (!sock_flag(sk, SOCK_ZAPPED))
307 if (sk->sk_state != TCP_CLOSE)
310 bound_dev_if = sk->sk_bound_dev_if;
312 /* Check if the address belongs to the host. */
314 if (addr_type != IPV6_ADDR_ANY) {
315 struct net_device *dev = NULL;
317 if (addr_type & IPV6_ADDR_LINKLOCAL) {
318 if (addr->l2tp_scope_id)
319 bound_dev_if = addr->l2tp_scope_id;
321 /* Binding to link-local address requires an
328 dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
333 /* ipv4 addr of the socket is invalid. Only the
334 * unspecified and mapped address have a v4 equivalent.
336 v4addr = LOOPBACK4_IPV6;
337 err = -EADDRNOTAVAIL;
338 if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
343 write_lock_bh(&l2tp_ip6_lock);
344 if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
345 addr->l2tp_conn_id)) {
346 write_unlock_bh(&l2tp_ip6_lock);
351 inet->inet_saddr = v4addr;
352 inet->inet_rcv_saddr = v4addr;
353 sk->sk_bound_dev_if = bound_dev_if;
354 sk->sk_v6_rcv_saddr = addr->l2tp_addr;
355 np->saddr = addr->l2tp_addr;
357 l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
359 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
360 sk_del_node_init(sk);
361 write_unlock_bh(&l2tp_ip6_lock);
363 sock_reset_flag(sk, SOCK_ZAPPED);
375 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
378 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
379 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
380 struct in6_addr *daddr;
384 if (addr_len < sizeof(*lsa))
387 if (usin->sin6_family != AF_INET6)
390 addr_type = ipv6_addr_type(&usin->sin6_addr);
391 if (addr_type & IPV6_ADDR_MULTICAST)
394 if (addr_type & IPV6_ADDR_MAPPED) {
395 daddr = &usin->sin6_addr;
396 if (ipv4_is_multicast(daddr->s6_addr32[3]))
402 /* Must bind first - autobinding does not work */
403 if (sock_flag(sk, SOCK_ZAPPED)) {
408 rc = __ip6_datagram_connect(sk, uaddr, addr_len);
412 l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
414 write_lock_bh(&l2tp_ip6_lock);
415 hlist_del_init(&sk->sk_bind_node);
416 sk_add_bind_node(sk, &l2tp_ip6_bind_table);
417 write_unlock_bh(&l2tp_ip6_lock);
425 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
427 if (sock_flag(sk, SOCK_ZAPPED))
430 return __udp_disconnect(sk, flags);
433 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
436 struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
437 struct sock *sk = sock->sk;
438 struct ipv6_pinfo *np = inet6_sk(sk);
439 struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
441 lsa->l2tp_family = AF_INET6;
442 lsa->l2tp_flowinfo = 0;
443 lsa->l2tp_scope_id = 0;
444 lsa->l2tp_unused = 0;
446 if (!lsk->peer_conn_id)
448 lsa->l2tp_conn_id = lsk->peer_conn_id;
449 lsa->l2tp_addr = sk->sk_v6_daddr;
451 lsa->l2tp_flowinfo = np->flow_label;
453 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
454 lsa->l2tp_addr = np->saddr;
456 lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
458 lsa->l2tp_conn_id = lsk->conn_id;
460 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
461 lsa->l2tp_scope_id = sk->sk_bound_dev_if;
465 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
469 /* Charge it to the socket, dropping if the queue is full. */
470 rc = sock_queue_rcv_skb(sk, skb);
477 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
482 static int l2tp_ip6_push_pending_frames(struct sock *sk)
485 __be32 *transhdr = NULL;
488 skb = skb_peek(&sk->sk_write_queue);
492 transhdr = (__be32 *)skb_transport_header(skb);
495 err = ip6_push_pending_frames(sk);
501 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
504 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
506 struct ipv6_txoptions opt_space;
507 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
508 struct in6_addr *daddr, *final_p, final;
509 struct ipv6_pinfo *np = inet6_sk(sk);
510 struct ipv6_txoptions *opt_to_free = NULL;
511 struct ipv6_txoptions *opt = NULL;
512 struct ip6_flowlabel *flowlabel = NULL;
513 struct dst_entry *dst = NULL;
515 struct ipcm6_cookie ipc6;
516 int addr_len = msg->msg_namelen;
517 int transhdrlen = 4; /* zero session-id */
518 int ulen = len + transhdrlen;
521 /* Rough check on arithmetic overflow,
522 * better check is made in ip6_append_data().
527 /* Mirror BSD error message compatibility */
528 if (msg->msg_flags & MSG_OOB)
531 /* Get and verify the address */
532 memset(&fl6, 0, sizeof(fl6));
534 fl6.flowi6_mark = sk->sk_mark;
535 fl6.flowi6_uid = sk->sk_uid;
540 if (addr_len < SIN6_LEN_RFC2133)
543 if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
544 return -EAFNOSUPPORT;
546 daddr = &lsa->l2tp_addr;
548 fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
549 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
550 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
551 if (IS_ERR(flowlabel))
556 /* Otherwise it will be difficult to maintain
559 if (sk->sk_state == TCP_ESTABLISHED &&
560 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
561 daddr = &sk->sk_v6_daddr;
563 if (addr_len >= sizeof(struct sockaddr_in6) &&
564 lsa->l2tp_scope_id &&
565 ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
566 fl6.flowi6_oif = lsa->l2tp_scope_id;
568 if (sk->sk_state != TCP_ESTABLISHED)
569 return -EDESTADDRREQ;
571 daddr = &sk->sk_v6_daddr;
572 fl6.flowlabel = np->flow_label;
575 if (fl6.flowi6_oif == 0)
576 fl6.flowi6_oif = sk->sk_bound_dev_if;
578 if (msg->msg_controllen) {
580 memset(opt, 0, sizeof(struct ipv6_txoptions));
581 opt->tot_len = sizeof(struct ipv6_txoptions);
584 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 (IS_ERR(flowlabel))
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(sock_net(sk), 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,
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;
664 if (msg->msg_flags & MSG_PROBE)
665 dst_confirm_neigh(dst, &fl6.daddr);
666 if (!(msg->msg_flags & MSG_PROBE) || len)
667 goto back_from_confirm;
672 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
673 int noblock, int flags, int *addr_len)
675 struct ipv6_pinfo *np = inet6_sk(sk);
676 DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
678 int err = -EOPNOTSUPP;
684 if (flags & MSG_ERRQUEUE)
685 return ipv6_recv_error(sk, msg, len, addr_len);
687 skb = skb_recv_datagram(sk, flags, noblock, &err);
693 msg->msg_flags |= MSG_TRUNC;
697 err = skb_copy_datagram_msg(skb, 0, msg, copied);
701 sock_recv_timestamp(msg, sk, skb);
703 /* Copy the address. */
705 lsa->l2tp_family = AF_INET6;
706 lsa->l2tp_unused = 0;
707 lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
708 lsa->l2tp_flowinfo = 0;
709 lsa->l2tp_scope_id = 0;
710 lsa->l2tp_conn_id = 0;
711 if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
712 lsa->l2tp_scope_id = inet6_iif(skb);
713 *addr_len = sizeof(*lsa);
717 ip6_datagram_recv_ctl(sk, msg, skb);
719 if (flags & MSG_TRUNC)
722 skb_free_datagram(sk, skb);
724 return err ? err : copied;
727 static struct proto l2tp_ip6_prot = {
729 .owner = THIS_MODULE,
730 .init = l2tp_ip6_open,
731 .close = l2tp_ip6_close,
732 .bind = l2tp_ip6_bind,
733 .connect = l2tp_ip6_connect,
734 .disconnect = l2tp_ip6_disconnect,
736 .destroy = l2tp_ip6_destroy_sock,
737 .setsockopt = ipv6_setsockopt,
738 .getsockopt = ipv6_getsockopt,
739 .sendmsg = l2tp_ip6_sendmsg,
740 .recvmsg = l2tp_ip6_recvmsg,
741 .backlog_rcv = l2tp_ip6_backlog_recv,
742 .hash = l2tp_ip6_hash,
743 .unhash = l2tp_ip6_unhash,
744 .obj_size = sizeof(struct l2tp_ip6_sock),
747 static const struct proto_ops l2tp_ip6_ops = {
749 .owner = THIS_MODULE,
750 .release = inet6_release,
752 .connect = inet_dgram_connect,
753 .socketpair = sock_no_socketpair,
754 .accept = sock_no_accept,
755 .getname = l2tp_ip6_getname,
756 .poll = datagram_poll,
757 .ioctl = inet6_ioctl,
758 .gettstamp = sock_gettstamp,
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_ioctl = inet6_compat_ioctl,
772 static struct inet_protosw l2tp_ip6_protosw = {
774 .protocol = IPPROTO_L2TP,
775 .prot = &l2tp_ip6_prot,
776 .ops = &l2tp_ip6_ops,
779 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
780 .handler = l2tp_ip6_recv,
783 static int __init l2tp_ip6_init(void)
787 pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
789 err = proto_register(&l2tp_ip6_prot, 1);
793 err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
797 inet6_register_protosw(&l2tp_ip6_protosw);
801 proto_unregister(&l2tp_ip6_prot);
806 static void __exit l2tp_ip6_exit(void)
808 inet6_unregister_protosw(&l2tp_ip6_protosw);
809 inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
810 proto_unregister(&l2tp_ip6_prot);
813 module_init(l2tp_ip6_init);
814 module_exit(l2tp_ip6_exit);
816 MODULE_LICENSE("GPL");
817 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
818 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
819 MODULE_VERSION("1.0");
821 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
824 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
825 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);