2 * L2TPv3 IP encapsulation support
4 * Copyright (c) 2008,2009,2010 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 #include <linux/icmp.h>
13 #include <linux/module.h>
14 #include <linux/skbuff.h>
15 #include <linux/random.h>
16 #include <linux/socket.h>
17 #include <linux/l2tp.h>
23 #include <net/inet_common.h>
24 #include <net/inet_hashtables.h>
25 #include <net/tcp_states.h>
26 #include <net/protocol.h>
29 #include "l2tp_core.h"
32 /* inet_sock has to be the first member of l2tp_ip_sock */
33 struct inet_sock inet;
46 static DEFINE_RWLOCK(l2tp_ip_lock);
47 static struct hlist_head l2tp_ip_table;
48 static struct hlist_head l2tp_ip_bind_table;
50 static inline struct l2tp_ip_sock *l2tp_ip_sk(const struct sock *sk)
52 return (struct l2tp_ip_sock *)sk;
55 static struct sock *__l2tp_ip_bind_lookup(struct net *net, __be32 laddr, int dif, u32 tunnel_id)
57 struct hlist_node *node;
60 sk_for_each_bound(sk, node, &l2tp_ip_bind_table) {
61 struct inet_sock *inet = inet_sk(sk);
62 struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
67 if ((l2tp->conn_id == tunnel_id) &&
68 net_eq(sock_net(sk), net) &&
69 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
70 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
79 static inline struct sock *l2tp_ip_bind_lookup(struct net *net, __be32 laddr, int dif, u32 tunnel_id)
81 struct sock *sk = __l2tp_ip_bind_lookup(net, laddr, dif, tunnel_id);
88 /* When processing receive frames, there are two cases to
89 * consider. Data frames consist of a non-zero session-id and an
90 * optional cookie. Control frames consist of a regular L2TP header
91 * preceded by 32-bits of zeros.
93 * L2TPv3 Session Header Over IP
96 * 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
97 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
99 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100 * | Cookie (optional, maximum 64 bits)...
101 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105 * L2TPv3 Control Message Header Over IP
108 * 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
109 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
110 * | (32 bits of zeros) |
111 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
113 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114 * | Control Connection ID |
115 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * All control frames are passed to userspace.
121 static int l2tp_ip_recv(struct sk_buff *skb)
126 unsigned char *ptr, *optr;
127 struct l2tp_session *session;
128 struct l2tp_tunnel *tunnel = NULL;
132 if (!pskb_may_pull(skb, 4))
135 /* Point to L2TP header */
136 optr = ptr = skb->data;
137 session_id = ntohl(*((__be32 *) ptr));
140 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
141 * the session_id. If it is 0, the packet is a L2TP control
142 * frame and the session_id value can be discarded.
144 if (session_id == 0) {
149 /* Ok, this is a data packet. Lookup the session. */
150 session = l2tp_session_find(&init_net, NULL, session_id);
154 tunnel = session->tunnel;
158 /* Trace packet contents, if enabled */
159 if (tunnel->debug & L2TP_MSG_DATA) {
160 length = min(32u, skb->len);
161 if (!pskb_may_pull(skb, length))
164 /* Point to L2TP header */
165 optr = ptr = skb->data;
167 printk(KERN_DEBUG "%s: ip recv: ", tunnel->name);
171 printk(" %02X", ptr[offset]);
172 } while (++offset < length);
177 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
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(&init_net, tunnel_id);
194 struct iphdr *iph = (struct iphdr *) skb_network_header(skb);
196 read_lock_bh(&l2tp_ip_lock);
197 sk = __l2tp_ip_bind_lookup(&init_net, iph->daddr, 0, tunnel_id);
198 read_unlock_bh(&l2tp_ip_lock);
206 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
211 return sk_receive_skb(sk, skb, 1);
221 static int l2tp_ip_open(struct sock *sk)
223 /* Prevent autobind. We don't have ports. */
224 inet_sk(sk)->inet_num = IPPROTO_L2TP;
226 write_lock_bh(&l2tp_ip_lock);
227 sk_add_node(sk, &l2tp_ip_table);
228 write_unlock_bh(&l2tp_ip_lock);
233 static void l2tp_ip_close(struct sock *sk, long timeout)
235 write_lock_bh(&l2tp_ip_lock);
236 hlist_del_init(&sk->sk_bind_node);
237 hlist_del_init(&sk->sk_node);
238 write_unlock_bh(&l2tp_ip_lock);
239 sk_common_release(sk);
242 static void l2tp_ip_destroy_sock(struct sock *sk)
246 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
249 sk_refcnt_debug_dec(sk);
252 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
254 struct inet_sock *inet = inet_sk(sk);
255 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
259 if (!sock_flag(sk, SOCK_ZAPPED))
261 if (addr_len < sizeof(struct sockaddr_l2tpip))
263 if (addr->l2tp_family != AF_INET)
267 read_lock_bh(&l2tp_ip_lock);
268 if (__l2tp_ip_bind_lookup(&init_net, addr->l2tp_addr.s_addr, sk->sk_bound_dev_if, addr->l2tp_conn_id))
271 read_unlock_bh(&l2tp_ip_lock);
274 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
277 chk_addr_ret = inet_addr_type(&init_net, addr->l2tp_addr.s_addr);
278 ret = -EADDRNOTAVAIL;
279 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
280 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
283 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
284 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
285 inet->inet_saddr = 0; /* Use device */
288 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
290 write_lock_bh(&l2tp_ip_lock);
291 sk_add_bind_node(sk, &l2tp_ip_bind_table);
292 sk_del_node_init(sk);
293 write_unlock_bh(&l2tp_ip_lock);
295 sock_reset_flag(sk, SOCK_ZAPPED);
303 read_unlock_bh(&l2tp_ip_lock);
308 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
310 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
311 struct inet_sock *inet = inet_sk(sk);
317 if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
320 if (addr_len < sizeof(*lsa))
323 if (lsa->l2tp_family != AF_INET)
324 return -EAFNOSUPPORT;
330 oif = sk->sk_bound_dev_if;
331 saddr = inet->inet_saddr;
334 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
337 fl4 = &inet->cork.fl.u.ip4;
338 rt = ip_route_connect(fl4, lsa->l2tp_addr.s_addr, saddr,
339 RT_CONN_FLAGS(sk), oif,
344 if (rc == -ENETUNREACH)
345 IP_INC_STATS_BH(&init_net, IPSTATS_MIB_OUTNOROUTES);
350 if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
355 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
357 if (!inet->inet_saddr)
358 inet->inet_saddr = fl4->saddr;
359 if (!inet->inet_rcv_saddr)
360 inet->inet_rcv_saddr = fl4->saddr;
361 inet->inet_daddr = fl4->daddr;
362 sk->sk_state = TCP_ESTABLISHED;
363 inet->inet_id = jiffies;
365 sk_dst_set(sk, &rt->dst);
367 write_lock_bh(&l2tp_ip_lock);
368 hlist_del_init(&sk->sk_bind_node);
369 sk_add_bind_node(sk, &l2tp_ip_bind_table);
370 write_unlock_bh(&l2tp_ip_lock);
378 static int l2tp_ip_disconnect(struct sock *sk, int flags)
380 if (sock_flag(sk, SOCK_ZAPPED))
383 return udp_disconnect(sk, flags);
386 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
387 int *uaddr_len, int peer)
389 struct sock *sk = sock->sk;
390 struct inet_sock *inet = inet_sk(sk);
391 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
392 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
394 memset(lsa, 0, sizeof(*lsa));
395 lsa->l2tp_family = AF_INET;
397 if (!inet->inet_dport)
399 lsa->l2tp_conn_id = lsk->peer_conn_id;
400 lsa->l2tp_addr.s_addr = inet->inet_daddr;
402 __be32 addr = inet->inet_rcv_saddr;
404 addr = inet->inet_saddr;
405 lsa->l2tp_conn_id = lsk->conn_id;
406 lsa->l2tp_addr.s_addr = addr;
408 *uaddr_len = sizeof(*lsa);
412 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
416 /* Charge it to the socket, dropping if the queue is full. */
417 rc = sock_queue_rcv_skb(sk, skb);
424 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
429 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
432 static int l2tp_ip_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, size_t len)
436 struct l2tp_ip_sock *lsa = l2tp_ip_sk(sk);
437 struct inet_sock *inet = inet_sk(sk);
438 struct rtable *rt = NULL;
446 if (sock_flag(sk, SOCK_DEAD))
449 /* Get and verify the address. */
451 struct sockaddr_l2tpip *lip = (struct sockaddr_l2tpip *) msg->msg_name;
453 if (msg->msg_namelen < sizeof(*lip))
456 if (lip->l2tp_family != AF_INET) {
458 if (lip->l2tp_family != AF_UNSPEC)
462 daddr = lip->l2tp_addr.s_addr;
465 if (sk->sk_state != TCP_ESTABLISHED)
468 daddr = inet->inet_daddr;
472 /* Allocate a socket buffer */
474 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
475 4 + len, 0, GFP_KERNEL);
479 /* Reserve space for headers, putting IP header on 4-byte boundary. */
480 skb_reserve(skb, 2 + NET_SKB_PAD);
481 skb_reset_network_header(skb);
482 skb_reserve(skb, sizeof(struct iphdr));
483 skb_reset_transport_header(skb);
485 /* Insert 0 session_id */
486 *((__be32 *) skb_put(skb, 4)) = 0;
488 /* Copy user data into skb */
489 rc = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
495 fl4 = &inet->cork.fl.u.ip4;
497 rt = (struct rtable *) __sk_dst_check(sk, 0);
501 const struct ip_options_rcu *inet_opt;
503 inet_opt = rcu_dereference(inet->inet_opt);
505 /* Use correct destination address if we have options. */
506 if (inet_opt && inet_opt->opt.srr)
507 daddr = inet_opt->opt.faddr;
509 /* If this fails, retransmit mechanism of transport layer will
510 * keep trying until route appears or the connection times
513 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
514 daddr, inet->inet_saddr,
515 inet->inet_dport, inet->inet_sport,
516 sk->sk_protocol, RT_CONN_FLAGS(sk),
517 sk->sk_bound_dev_if);
521 sk_setup_caps(sk, &rt->dst);
523 skb_dst_set(skb, &rt->dst);
528 /* We dont need to clone dst here, it is guaranteed to not disappear.
529 * __dev_xmit_skb() might force a refcount if needed.
531 skb_dst_set_noref(skb, &rt->dst);
534 /* Queue the packet to IP for output */
535 rc = ip_queue_xmit(skb, &inet->cork.fl);
542 lsa->tx_bytes += len;
554 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
560 static int l2tp_ip_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
561 size_t len, int noblock, int flags, int *addr_len)
563 struct inet_sock *inet = inet_sk(sk);
564 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
566 int err = -EOPNOTSUPP;
567 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
573 skb = skb_recv_datagram(sk, flags, noblock, &err);
579 msg->msg_flags |= MSG_TRUNC;
583 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
587 sock_recv_timestamp(msg, sk, skb);
589 /* Copy the address. */
591 sin->sin_family = AF_INET;
592 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
594 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
595 *addr_len = sizeof(*sin);
597 if (inet->cmsg_flags)
598 ip_cmsg_recv(msg, skb);
599 if (flags & MSG_TRUNC)
602 skb_free_datagram(sk, skb);
610 lsk->rx_bytes += copied;
615 static struct proto l2tp_ip_prot = {
617 .owner = THIS_MODULE,
618 .init = l2tp_ip_open,
619 .close = l2tp_ip_close,
620 .bind = l2tp_ip_bind,
621 .connect = l2tp_ip_connect,
622 .disconnect = l2tp_ip_disconnect,
624 .destroy = l2tp_ip_destroy_sock,
625 .setsockopt = ip_setsockopt,
626 .getsockopt = ip_getsockopt,
627 .sendmsg = l2tp_ip_sendmsg,
628 .recvmsg = l2tp_ip_recvmsg,
629 .backlog_rcv = l2tp_ip_backlog_recv,
631 .unhash = inet_unhash,
632 .obj_size = sizeof(struct l2tp_ip_sock),
634 .compat_setsockopt = compat_ip_setsockopt,
635 .compat_getsockopt = compat_ip_getsockopt,
639 static const struct proto_ops l2tp_ip_ops = {
641 .owner = THIS_MODULE,
642 .release = inet_release,
644 .connect = inet_dgram_connect,
645 .socketpair = sock_no_socketpair,
646 .accept = sock_no_accept,
647 .getname = l2tp_ip_getname,
648 .poll = datagram_poll,
650 .listen = sock_no_listen,
651 .shutdown = inet_shutdown,
652 .setsockopt = sock_common_setsockopt,
653 .getsockopt = sock_common_getsockopt,
654 .sendmsg = inet_sendmsg,
655 .recvmsg = sock_common_recvmsg,
656 .mmap = sock_no_mmap,
657 .sendpage = sock_no_sendpage,
659 .compat_setsockopt = compat_sock_common_setsockopt,
660 .compat_getsockopt = compat_sock_common_getsockopt,
664 static struct inet_protosw l2tp_ip_protosw = {
666 .protocol = IPPROTO_L2TP,
667 .prot = &l2tp_ip_prot,
672 static struct net_protocol l2tp_ip_protocol __read_mostly = {
673 .handler = l2tp_ip_recv,
676 static int __init l2tp_ip_init(void)
680 printk(KERN_INFO "L2TP IP encapsulation support (L2TPv3)\n");
682 err = proto_register(&l2tp_ip_prot, 1);
686 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
690 inet_register_protosw(&l2tp_ip_protosw);
694 proto_unregister(&l2tp_ip_prot);
699 static void __exit l2tp_ip_exit(void)
701 inet_unregister_protosw(&l2tp_ip_protosw);
702 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
703 proto_unregister(&l2tp_ip_prot);
706 module_init(l2tp_ip_init);
707 module_exit(l2tp_ip_exit);
709 MODULE_LICENSE("GPL");
710 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
711 MODULE_DESCRIPTION("L2TP over IP");
712 MODULE_VERSION("1.0");
714 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
717 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);