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 <asm/ioctls.h>
13 #include <linux/icmp.h>
14 #include <linux/module.h>
15 #include <linux/skbuff.h>
16 #include <linux/random.h>
17 #include <linux/socket.h>
18 #include <linux/l2tp.h>
24 #include <net/inet_common.h>
25 #include <net/inet_hashtables.h>
26 #include <net/tcp_states.h>
27 #include <net/protocol.h>
30 #include "l2tp_core.h"
33 /* inet_sock has to be the first member of l2tp_ip_sock */
34 struct inet_sock inet;
47 static DEFINE_RWLOCK(l2tp_ip_lock);
48 static struct hlist_head l2tp_ip_table;
49 static struct hlist_head l2tp_ip_bind_table;
51 static inline struct l2tp_ip_sock *l2tp_ip_sk(const struct sock *sk)
53 return (struct l2tp_ip_sock *)sk;
56 static struct sock *__l2tp_ip_bind_lookup(struct net *net, __be32 laddr, int dif, u32 tunnel_id)
58 struct hlist_node *node;
61 sk_for_each_bound(sk, node, &l2tp_ip_bind_table) {
62 struct inet_sock *inet = inet_sk(sk);
63 struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
68 if ((l2tp->conn_id == tunnel_id) &&
69 net_eq(sock_net(sk), net) &&
70 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
71 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
80 static inline struct sock *l2tp_ip_bind_lookup(struct net *net, __be32 laddr, int dif, u32 tunnel_id)
82 struct sock *sk = __l2tp_ip_bind_lookup(net, laddr, dif, tunnel_id);
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_ip_recv(struct sk_buff *skb)
127 unsigned char *ptr, *optr;
128 struct l2tp_session *session;
129 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_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 /* Point to L2TP header */
166 optr = ptr = skb->data;
168 printk(KERN_DEBUG "%s: ip recv: ", tunnel->name);
172 printk(" %02X", ptr[offset]);
173 } while (++offset < length);
178 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
183 /* Get the tunnel_id from the L2TP header */
184 if (!pskb_may_pull(skb, 12))
187 if ((skb->data[0] & 0xc0) != 0xc0)
190 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
191 tunnel = l2tp_tunnel_find(&init_net, tunnel_id);
195 struct iphdr *iph = (struct iphdr *) skb_network_header(skb);
197 read_lock_bh(&l2tp_ip_lock);
198 sk = __l2tp_ip_bind_lookup(&init_net, iph->daddr, 0, tunnel_id);
199 read_unlock_bh(&l2tp_ip_lock);
207 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
212 return sk_receive_skb(sk, skb, 1);
222 static int l2tp_ip_open(struct sock *sk)
224 /* Prevent autobind. We don't have ports. */
225 inet_sk(sk)->inet_num = IPPROTO_L2TP;
227 write_lock_bh(&l2tp_ip_lock);
228 sk_add_node(sk, &l2tp_ip_table);
229 write_unlock_bh(&l2tp_ip_lock);
234 static void l2tp_ip_close(struct sock *sk, long timeout)
236 write_lock_bh(&l2tp_ip_lock);
237 hlist_del_init(&sk->sk_bind_node);
238 hlist_del_init(&sk->sk_node);
239 write_unlock_bh(&l2tp_ip_lock);
240 sk_common_release(sk);
243 static void l2tp_ip_destroy_sock(struct sock *sk)
247 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
250 sk_refcnt_debug_dec(sk);
253 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
255 struct inet_sock *inet = inet_sk(sk);
256 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
260 if (addr_len < sizeof(struct sockaddr_l2tpip))
262 if (addr->l2tp_family != AF_INET)
266 read_lock_bh(&l2tp_ip_lock);
267 if (__l2tp_ip_bind_lookup(&init_net, addr->l2tp_addr.s_addr, sk->sk_bound_dev_if, addr->l2tp_conn_id))
270 read_unlock_bh(&l2tp_ip_lock);
273 if (!sock_flag(sk, SOCK_ZAPPED))
276 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
279 chk_addr_ret = inet_addr_type(&init_net, addr->l2tp_addr.s_addr);
280 ret = -EADDRNOTAVAIL;
281 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
282 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
285 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
286 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
287 inet->inet_saddr = 0; /* Use device */
290 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
292 write_lock_bh(&l2tp_ip_lock);
293 sk_add_bind_node(sk, &l2tp_ip_bind_table);
294 sk_del_node_init(sk);
295 write_unlock_bh(&l2tp_ip_lock);
297 sock_reset_flag(sk, SOCK_ZAPPED);
305 read_unlock_bh(&l2tp_ip_lock);
310 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
312 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
313 struct inet_sock *inet = inet_sk(sk);
319 if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
322 if (addr_len < sizeof(*lsa))
325 if (lsa->l2tp_family != AF_INET)
326 return -EAFNOSUPPORT;
332 oif = sk->sk_bound_dev_if;
333 saddr = inet->inet_saddr;
336 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
339 fl4 = &inet->cork.fl.u.ip4;
340 rt = ip_route_connect(fl4, lsa->l2tp_addr.s_addr, saddr,
341 RT_CONN_FLAGS(sk), oif,
346 if (rc == -ENETUNREACH)
347 IP_INC_STATS_BH(&init_net, IPSTATS_MIB_OUTNOROUTES);
352 if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
357 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
359 if (!inet->inet_saddr)
360 inet->inet_saddr = fl4->saddr;
361 if (!inet->inet_rcv_saddr)
362 inet->inet_rcv_saddr = fl4->saddr;
363 inet->inet_daddr = fl4->daddr;
364 sk->sk_state = TCP_ESTABLISHED;
365 inet->inet_id = jiffies;
367 sk_dst_set(sk, &rt->dst);
369 write_lock_bh(&l2tp_ip_lock);
370 hlist_del_init(&sk->sk_bind_node);
371 sk_add_bind_node(sk, &l2tp_ip_bind_table);
372 write_unlock_bh(&l2tp_ip_lock);
380 static int l2tp_ip_disconnect(struct sock *sk, int flags)
382 if (sock_flag(sk, SOCK_ZAPPED))
385 return udp_disconnect(sk, flags);
388 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
389 int *uaddr_len, int peer)
391 struct sock *sk = sock->sk;
392 struct inet_sock *inet = inet_sk(sk);
393 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
394 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
396 memset(lsa, 0, sizeof(*lsa));
397 lsa->l2tp_family = AF_INET;
399 if (!inet->inet_dport)
401 lsa->l2tp_conn_id = lsk->peer_conn_id;
402 lsa->l2tp_addr.s_addr = inet->inet_daddr;
404 __be32 addr = inet->inet_rcv_saddr;
406 addr = inet->inet_saddr;
407 lsa->l2tp_conn_id = lsk->conn_id;
408 lsa->l2tp_addr.s_addr = addr;
410 *uaddr_len = sizeof(*lsa);
414 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
418 /* Charge it to the socket, dropping if the queue is full. */
419 rc = sock_queue_rcv_skb(sk, skb);
426 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
431 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
434 static int l2tp_ip_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, size_t len)
438 struct l2tp_ip_sock *lsa = l2tp_ip_sk(sk);
439 struct inet_sock *inet = inet_sk(sk);
440 struct rtable *rt = NULL;
448 if (sock_flag(sk, SOCK_DEAD))
451 /* Get and verify the address. */
453 struct sockaddr_l2tpip *lip = (struct sockaddr_l2tpip *) msg->msg_name;
455 if (msg->msg_namelen < sizeof(*lip))
458 if (lip->l2tp_family != AF_INET) {
460 if (lip->l2tp_family != AF_UNSPEC)
464 daddr = lip->l2tp_addr.s_addr;
467 if (sk->sk_state != TCP_ESTABLISHED)
470 daddr = inet->inet_daddr;
474 /* Allocate a socket buffer */
476 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
477 4 + len, 0, GFP_KERNEL);
481 /* Reserve space for headers, putting IP header on 4-byte boundary. */
482 skb_reserve(skb, 2 + NET_SKB_PAD);
483 skb_reset_network_header(skb);
484 skb_reserve(skb, sizeof(struct iphdr));
485 skb_reset_transport_header(skb);
487 /* Insert 0 session_id */
488 *((__be32 *) skb_put(skb, 4)) = 0;
490 /* Copy user data into skb */
491 rc = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
497 fl4 = &inet->cork.fl.u.ip4;
499 rt = (struct rtable *) __sk_dst_check(sk, 0);
503 const struct ip_options_rcu *inet_opt;
505 inet_opt = rcu_dereference(inet->inet_opt);
507 /* Use correct destination address if we have options. */
508 if (inet_opt && inet_opt->opt.srr)
509 daddr = inet_opt->opt.faddr;
511 /* If this fails, retransmit mechanism of transport layer will
512 * keep trying until route appears or the connection times
515 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
516 daddr, inet->inet_saddr,
517 inet->inet_dport, inet->inet_sport,
518 sk->sk_protocol, RT_CONN_FLAGS(sk),
519 sk->sk_bound_dev_if);
523 sk_setup_caps(sk, &rt->dst);
525 skb_dst_set(skb, &rt->dst);
530 /* We dont need to clone dst here, it is guaranteed to not disappear.
531 * __dev_xmit_skb() might force a refcount if needed.
533 skb_dst_set_noref(skb, &rt->dst);
536 /* Queue the packet to IP for output */
537 rc = ip_queue_xmit(skb, &inet->cork.fl);
544 lsa->tx_bytes += len;
556 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
562 static int l2tp_ip_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
563 size_t len, int noblock, int flags, int *addr_len)
565 struct inet_sock *inet = inet_sk(sk);
566 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
568 int err = -EOPNOTSUPP;
569 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
575 skb = skb_recv_datagram(sk, flags, noblock, &err);
581 msg->msg_flags |= MSG_TRUNC;
585 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
589 sock_recv_timestamp(msg, sk, skb);
591 /* Copy the address. */
593 sin->sin_family = AF_INET;
594 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
596 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
597 *addr_len = sizeof(*sin);
599 if (inet->cmsg_flags)
600 ip_cmsg_recv(msg, skb);
601 if (flags & MSG_TRUNC)
604 skb_free_datagram(sk, skb);
612 lsk->rx_bytes += copied;
617 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
624 amount = sk_wmem_alloc_get(sk);
627 spin_lock_bh(&sk->sk_receive_queue.lock);
628 skb = skb_peek(&sk->sk_receive_queue);
629 amount = skb ? skb->len : 0;
630 spin_unlock_bh(&sk->sk_receive_queue.lock);
637 return put_user(amount, (int __user *)arg);
639 EXPORT_SYMBOL(l2tp_ioctl);
641 static struct proto l2tp_ip_prot = {
643 .owner = THIS_MODULE,
644 .init = l2tp_ip_open,
645 .close = l2tp_ip_close,
646 .bind = l2tp_ip_bind,
647 .connect = l2tp_ip_connect,
648 .disconnect = l2tp_ip_disconnect,
650 .destroy = l2tp_ip_destroy_sock,
651 .setsockopt = ip_setsockopt,
652 .getsockopt = ip_getsockopt,
653 .sendmsg = l2tp_ip_sendmsg,
654 .recvmsg = l2tp_ip_recvmsg,
655 .backlog_rcv = l2tp_ip_backlog_recv,
657 .unhash = inet_unhash,
658 .obj_size = sizeof(struct l2tp_ip_sock),
660 .compat_setsockopt = compat_ip_setsockopt,
661 .compat_getsockopt = compat_ip_getsockopt,
665 static const struct proto_ops l2tp_ip_ops = {
667 .owner = THIS_MODULE,
668 .release = inet_release,
670 .connect = inet_dgram_connect,
671 .socketpair = sock_no_socketpair,
672 .accept = sock_no_accept,
673 .getname = l2tp_ip_getname,
674 .poll = datagram_poll,
676 .listen = sock_no_listen,
677 .shutdown = inet_shutdown,
678 .setsockopt = sock_common_setsockopt,
679 .getsockopt = sock_common_getsockopt,
680 .sendmsg = inet_sendmsg,
681 .recvmsg = sock_common_recvmsg,
682 .mmap = sock_no_mmap,
683 .sendpage = sock_no_sendpage,
685 .compat_setsockopt = compat_sock_common_setsockopt,
686 .compat_getsockopt = compat_sock_common_getsockopt,
690 static struct inet_protosw l2tp_ip_protosw = {
692 .protocol = IPPROTO_L2TP,
693 .prot = &l2tp_ip_prot,
698 static struct net_protocol l2tp_ip_protocol __read_mostly = {
699 .handler = l2tp_ip_recv,
702 static int __init l2tp_ip_init(void)
706 printk(KERN_INFO "L2TP IP encapsulation support (L2TPv3)\n");
708 err = proto_register(&l2tp_ip_prot, 1);
712 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
716 inet_register_protosw(&l2tp_ip_protosw);
720 proto_unregister(&l2tp_ip_prot);
725 static void __exit l2tp_ip_exit(void)
727 inet_unregister_protosw(&l2tp_ip_protosw);
728 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
729 proto_unregister(&l2tp_ip_prot);
732 module_init(l2tp_ip_init);
733 module_exit(l2tp_ip_exit);
735 MODULE_LICENSE("GPL");
736 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
737 MODULE_DESCRIPTION("L2TP over IP");
738 MODULE_VERSION("1.0");
740 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
743 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);