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 /* Point to L2TP header */
133 optr = ptr = skb->data;
135 if (!pskb_may_pull(skb, 4))
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_find(&init_net, NULL, session_id);
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 printk(KERN_DEBUG "%s: ip recv: ", tunnel->name);
169 printk(" %02X", ptr[offset]);
170 } while (++offset < length);
175 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, 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 iphdr *iph = (struct iphdr *) skb_network_header(skb);
194 read_lock_bh(&l2tp_ip_lock);
195 sk = __l2tp_ip_bind_lookup(&init_net, iph->daddr, 0, tunnel_id);
196 read_unlock_bh(&l2tp_ip_lock);
204 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
209 return sk_receive_skb(sk, skb, 1);
219 static int l2tp_ip_open(struct sock *sk)
221 /* Prevent autobind. We don't have ports. */
222 inet_sk(sk)->inet_num = IPPROTO_L2TP;
224 write_lock_bh(&l2tp_ip_lock);
225 sk_add_node(sk, &l2tp_ip_table);
226 write_unlock_bh(&l2tp_ip_lock);
231 static void l2tp_ip_close(struct sock *sk, long timeout)
233 write_lock_bh(&l2tp_ip_lock);
234 hlist_del_init(&sk->sk_bind_node);
235 hlist_del_init(&sk->sk_node);
236 write_unlock_bh(&l2tp_ip_lock);
237 sk_common_release(sk);
240 static void l2tp_ip_destroy_sock(struct sock *sk)
244 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
247 sk_refcnt_debug_dec(sk);
250 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
252 struct inet_sock *inet = inet_sk(sk);
253 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
258 read_lock_bh(&l2tp_ip_lock);
259 if (__l2tp_ip_bind_lookup(&init_net, addr->l2tp_addr.s_addr, sk->sk_bound_dev_if, addr->l2tp_conn_id))
262 read_unlock_bh(&l2tp_ip_lock);
265 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
268 chk_addr_ret = inet_addr_type(&init_net, addr->l2tp_addr.s_addr);
269 ret = -EADDRNOTAVAIL;
270 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
271 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
274 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
275 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
276 inet->inet_saddr = 0; /* Use device */
279 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
281 write_lock_bh(&l2tp_ip_lock);
282 sk_add_bind_node(sk, &l2tp_ip_bind_table);
283 sk_del_node_init(sk);
284 write_unlock_bh(&l2tp_ip_lock);
292 read_unlock_bh(&l2tp_ip_lock);
297 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
300 struct inet_sock *inet = inet_sk(sk);
301 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
307 if (addr_len < sizeof(*lsa))
311 if (lsa->l2tp_family != AF_INET)
316 oif = sk->sk_bound_dev_if;
317 saddr = inet->inet_saddr;
320 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
323 rt = ip_route_connect(lsa->l2tp_addr.s_addr, saddr,
324 RT_CONN_FLAGS(sk), oif,
329 if (rc == -ENETUNREACH)
330 IP_INC_STATS_BH(&init_net, IPSTATS_MIB_OUTNOROUTES);
335 if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
340 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
342 if (!inet->inet_saddr)
343 inet->inet_saddr = rt->rt_src;
344 if (!inet->inet_rcv_saddr)
345 inet->inet_rcv_saddr = rt->rt_src;
346 inet->inet_daddr = rt->rt_dst;
347 sk->sk_state = TCP_ESTABLISHED;
348 inet->inet_id = jiffies;
350 sk_dst_set(sk, &rt->dst);
352 write_lock_bh(&l2tp_ip_lock);
353 hlist_del_init(&sk->sk_bind_node);
354 sk_add_bind_node(sk, &l2tp_ip_bind_table);
355 write_unlock_bh(&l2tp_ip_lock);
362 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
363 int *uaddr_len, int peer)
365 struct sock *sk = sock->sk;
366 struct inet_sock *inet = inet_sk(sk);
367 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
368 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
370 memset(lsa, 0, sizeof(*lsa));
371 lsa->l2tp_family = AF_INET;
373 if (!inet->inet_dport)
375 lsa->l2tp_conn_id = lsk->peer_conn_id;
376 lsa->l2tp_addr.s_addr = inet->inet_daddr;
378 __be32 addr = inet->inet_rcv_saddr;
380 addr = inet->inet_saddr;
381 lsa->l2tp_conn_id = lsk->conn_id;
382 lsa->l2tp_addr.s_addr = addr;
384 *uaddr_len = sizeof(*lsa);
388 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
392 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
397 /* Charge it to the socket, dropping if the queue is full. */
398 rc = sock_queue_rcv_skb(sk, skb);
405 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
410 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
413 static int l2tp_ip_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, size_t len)
417 struct l2tp_ip_sock *lsa = l2tp_ip_sk(sk);
418 struct inet_sock *inet = inet_sk(sk);
419 struct ip_options *opt = inet->opt;
420 struct rtable *rt = NULL;
424 if (sock_flag(sk, SOCK_DEAD))
427 /* Get and verify the address. */
429 struct sockaddr_l2tpip *lip = (struct sockaddr_l2tpip *) msg->msg_name;
430 if (msg->msg_namelen < sizeof(*lip))
433 if (lip->l2tp_family != AF_INET) {
434 if (lip->l2tp_family != AF_UNSPEC)
435 return -EAFNOSUPPORT;
438 daddr = lip->l2tp_addr.s_addr;
440 if (sk->sk_state != TCP_ESTABLISHED)
441 return -EDESTADDRREQ;
443 daddr = inet->inet_daddr;
447 /* Allocate a socket buffer */
449 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
450 4 + len, 0, GFP_KERNEL);
454 /* Reserve space for headers, putting IP header on 4-byte boundary. */
455 skb_reserve(skb, 2 + NET_SKB_PAD);
456 skb_reset_network_header(skb);
457 skb_reserve(skb, sizeof(struct iphdr));
458 skb_reset_transport_header(skb);
460 /* Insert 0 session_id */
461 *((__be32 *) skb_put(skb, 4)) = 0;
463 /* Copy user data into skb */
464 rc = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
471 rt = (struct rtable *) __sk_dst_check(sk, 0);
474 /* Use correct destination address if we have options. */
479 struct flowi fl = { .oif = sk->sk_bound_dev_if,
481 .fl4_src = inet->inet_saddr,
482 .fl4_tos = RT_CONN_FLAGS(sk),
483 .proto = sk->sk_protocol,
484 .flags = inet_sk_flowi_flags(sk),
485 .fl_ip_sport = inet->inet_sport,
486 .fl_ip_dport = inet->inet_dport };
488 /* If this fails, retransmit mechanism of transport layer will
489 * keep trying until route appears or the connection times
492 security_sk_classify_flow(sk, &fl);
493 rt = ip_route_output_flow(sock_net(sk), &fl, sk);
497 sk_setup_caps(sk, &rt->dst);
499 skb_dst_set(skb, dst_clone(&rt->dst));
501 /* Queue the packet to IP for output */
502 rc = ip_queue_xmit(skb);
508 lsa->tx_bytes += len;
517 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
519 return -EHOSTUNREACH;
522 static int l2tp_ip_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
523 size_t len, int noblock, int flags, int *addr_len)
525 struct inet_sock *inet = inet_sk(sk);
526 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
528 int err = -EOPNOTSUPP;
529 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
536 *addr_len = sizeof(*sin);
538 skb = skb_recv_datagram(sk, flags, noblock, &err);
544 msg->msg_flags |= MSG_TRUNC;
548 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
552 sock_recv_timestamp(msg, sk, skb);
554 /* Copy the address. */
556 sin->sin_family = AF_INET;
557 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
559 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
561 if (inet->cmsg_flags)
562 ip_cmsg_recv(msg, skb);
563 if (flags & MSG_TRUNC)
566 skb_free_datagram(sk, skb);
574 lsk->rx_bytes += copied;
579 static struct proto l2tp_ip_prot = {
581 .owner = THIS_MODULE,
582 .init = l2tp_ip_open,
583 .close = l2tp_ip_close,
584 .bind = l2tp_ip_bind,
585 .connect = l2tp_ip_connect,
586 .disconnect = udp_disconnect,
588 .destroy = l2tp_ip_destroy_sock,
589 .setsockopt = ip_setsockopt,
590 .getsockopt = ip_getsockopt,
591 .sendmsg = l2tp_ip_sendmsg,
592 .recvmsg = l2tp_ip_recvmsg,
593 .backlog_rcv = l2tp_ip_backlog_recv,
595 .unhash = inet_unhash,
596 .obj_size = sizeof(struct l2tp_ip_sock),
598 .compat_setsockopt = compat_ip_setsockopt,
599 .compat_getsockopt = compat_ip_getsockopt,
603 static const struct proto_ops l2tp_ip_ops = {
605 .owner = THIS_MODULE,
606 .release = inet_release,
608 .connect = inet_dgram_connect,
609 .socketpair = sock_no_socketpair,
610 .accept = sock_no_accept,
611 .getname = l2tp_ip_getname,
612 .poll = datagram_poll,
614 .listen = sock_no_listen,
615 .shutdown = inet_shutdown,
616 .setsockopt = sock_common_setsockopt,
617 .getsockopt = sock_common_getsockopt,
618 .sendmsg = inet_sendmsg,
619 .recvmsg = sock_common_recvmsg,
620 .mmap = sock_no_mmap,
621 .sendpage = sock_no_sendpage,
623 .compat_setsockopt = compat_sock_common_setsockopt,
624 .compat_getsockopt = compat_sock_common_getsockopt,
628 static struct inet_protosw l2tp_ip_protosw = {
630 .protocol = IPPROTO_L2TP,
631 .prot = &l2tp_ip_prot,
636 static struct net_protocol l2tp_ip_protocol __read_mostly = {
637 .handler = l2tp_ip_recv,
640 static int __init l2tp_ip_init(void)
644 printk(KERN_INFO "L2TP IP encapsulation support (L2TPv3)\n");
646 err = proto_register(&l2tp_ip_prot, 1);
650 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
654 inet_register_protosw(&l2tp_ip_protosw);
658 proto_unregister(&l2tp_ip_prot);
663 static void __exit l2tp_ip_exit(void)
665 inet_unregister_protosw(&l2tp_ip_protosw);
666 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
667 proto_unregister(&l2tp_ip_prot);
670 module_init(l2tp_ip_init);
671 module_exit(l2tp_ip_exit);
673 MODULE_LICENSE("GPL");
674 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
675 MODULE_DESCRIPTION("L2TP over IP");
676 MODULE_VERSION("1.0");
678 /* Use the value of SOCK_DGRAM (2) directory, because __stringify does't like
681 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);