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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 #include <asm/ioctls.h>
15 #include <linux/icmp.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/random.h>
19 #include <linux/socket.h>
20 #include <linux/l2tp.h>
26 #include <net/inet_common.h>
27 #include <net/tcp_states.h>
28 #include <net/protocol.h>
31 #include "l2tp_core.h"
34 /* inet_sock has to be the first member of l2tp_ip_sock */
35 struct inet_sock inet;
41 static DEFINE_RWLOCK(l2tp_ip_lock);
42 static struct hlist_head l2tp_ip_table;
43 static struct hlist_head l2tp_ip_bind_table;
45 static inline struct l2tp_ip_sock *l2tp_ip_sk(const struct sock *sk)
47 return (struct l2tp_ip_sock *)sk;
50 static struct sock *__l2tp_ip_bind_lookup(const struct net *net, __be32 laddr,
51 __be32 raddr, int dif, u32 tunnel_id)
55 sk_for_each_bound(sk, &l2tp_ip_bind_table) {
56 const struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
57 const struct inet_sock *inet = inet_sk(sk);
59 if (!net_eq(sock_net(sk), net))
62 if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
65 if (inet->inet_rcv_saddr && laddr &&
66 inet->inet_rcv_saddr != laddr)
69 if (inet->inet_daddr && raddr && inet->inet_daddr != raddr)
72 if (l2tp->conn_id != tunnel_id)
83 /* When processing receive frames, there are two cases to
84 * consider. Data frames consist of a non-zero session-id and an
85 * optional cookie. Control frames consist of a regular L2TP header
86 * preceded by 32-bits of zeros.
88 * L2TPv3 Session Header Over IP
91 * 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
92 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
94 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
95 * | Cookie (optional, maximum 64 bits)...
96 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
98 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100 * L2TPv3 Control Message Header Over IP
103 * 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
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105 * | (32 bits of zeros) |
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
108 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109 * | Control Connection ID |
110 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114 * All control frames are passed to userspace.
116 static int l2tp_ip_recv(struct sk_buff *skb)
118 struct net *net = dev_net(skb->dev);
122 unsigned char *ptr, *optr;
123 struct l2tp_session *session;
124 struct l2tp_tunnel *tunnel = NULL;
128 if (!pskb_may_pull(skb, 4))
131 /* Point to L2TP header */
132 optr = ptr = skb->data;
133 session_id = ntohl(*((__be32 *) ptr));
136 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
137 * the session_id. If it is 0, the packet is a L2TP control
138 * frame and the session_id value can be discarded.
140 if (session_id == 0) {
145 /* Ok, this is a data packet. Lookup the session. */
146 session = l2tp_session_get(net, NULL, session_id, true);
150 tunnel = session->tunnel;
154 /* Trace packet contents, if enabled */
155 if (tunnel->debug & L2TP_MSG_DATA) {
156 length = min(32u, skb->len);
157 if (!pskb_may_pull(skb, length))
160 /* Point to L2TP header */
161 optr = ptr = skb->data;
163 pr_debug("%s: ip recv\n", tunnel->name);
164 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
167 if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
170 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
171 l2tp_session_dec_refcount(session);
176 /* Get the tunnel_id from the L2TP header */
177 if (!pskb_may_pull(skb, 12))
180 if ((skb->data[0] & 0xc0) != 0xc0)
183 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
184 iph = (struct iphdr *)skb_network_header(skb);
186 read_lock_bh(&l2tp_ip_lock);
187 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr, inet_iif(skb),
190 read_unlock_bh(&l2tp_ip_lock);
194 read_unlock_bh(&l2tp_ip_lock);
196 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
201 return sk_receive_skb(sk, skb, 1);
205 session->deref(session);
206 l2tp_session_dec_refcount(session);
217 static int l2tp_ip_hash(struct sock *sk)
219 if (sk_unhashed(sk)) {
220 write_lock_bh(&l2tp_ip_lock);
221 sk_add_node(sk, &l2tp_ip_table);
222 write_unlock_bh(&l2tp_ip_lock);
227 static void l2tp_ip_unhash(struct sock *sk)
231 write_lock_bh(&l2tp_ip_lock);
232 sk_del_node_init(sk);
233 write_unlock_bh(&l2tp_ip_lock);
236 static int l2tp_ip_open(struct sock *sk)
238 /* Prevent autobind. We don't have ports. */
239 inet_sk(sk)->inet_num = IPPROTO_L2TP;
245 static void l2tp_ip_close(struct sock *sk, long timeout)
247 write_lock_bh(&l2tp_ip_lock);
248 hlist_del_init(&sk->sk_bind_node);
249 sk_del_node_init(sk);
250 write_unlock_bh(&l2tp_ip_lock);
251 sk_common_release(sk);
254 static void l2tp_ip_destroy_sock(struct sock *sk)
257 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
259 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
263 l2tp_tunnel_closeall(tunnel);
267 sk_refcnt_debug_dec(sk);
270 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
272 struct inet_sock *inet = inet_sk(sk);
273 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
274 struct net *net = sock_net(sk);
278 if (addr_len < sizeof(struct sockaddr_l2tpip))
280 if (addr->l2tp_family != AF_INET)
286 if (!sock_flag(sk, SOCK_ZAPPED))
289 if (sk->sk_state != TCP_CLOSE)
292 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
293 ret = -EADDRNOTAVAIL;
294 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
295 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
298 if (addr->l2tp_addr.s_addr)
299 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
300 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
301 inet->inet_saddr = 0; /* Use device */
303 write_lock_bh(&l2tp_ip_lock);
304 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
305 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
306 write_unlock_bh(&l2tp_ip_lock);
312 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
314 sk_add_bind_node(sk, &l2tp_ip_bind_table);
315 sk_del_node_init(sk);
316 write_unlock_bh(&l2tp_ip_lock);
319 sock_reset_flag(sk, SOCK_ZAPPED);
327 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
329 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
332 if (addr_len < sizeof(*lsa))
335 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
340 /* Must bind first - autobinding does not work */
341 if (sock_flag(sk, SOCK_ZAPPED)) {
346 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
350 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
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);
363 static int l2tp_ip_disconnect(struct sock *sk, int flags)
365 if (sock_flag(sk, SOCK_ZAPPED))
368 return __udp_disconnect(sk, flags);
371 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
372 int *uaddr_len, int peer)
374 struct sock *sk = sock->sk;
375 struct inet_sock *inet = inet_sk(sk);
376 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
377 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
379 memset(lsa, 0, sizeof(*lsa));
380 lsa->l2tp_family = AF_INET;
382 if (!inet->inet_dport)
384 lsa->l2tp_conn_id = lsk->peer_conn_id;
385 lsa->l2tp_addr.s_addr = inet->inet_daddr;
387 __be32 addr = inet->inet_rcv_saddr;
389 addr = inet->inet_saddr;
390 lsa->l2tp_conn_id = lsk->conn_id;
391 lsa->l2tp_addr.s_addr = addr;
393 *uaddr_len = sizeof(*lsa);
397 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
401 /* Charge it to the socket, dropping if the queue is full. */
402 rc = sock_queue_rcv_skb(sk, skb);
409 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
414 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
417 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
421 struct inet_sock *inet = inet_sk(sk);
422 struct rtable *rt = NULL;
430 if (sock_flag(sk, SOCK_DEAD))
433 /* Get and verify the address. */
435 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
437 if (msg->msg_namelen < sizeof(*lip))
440 if (lip->l2tp_family != AF_INET) {
442 if (lip->l2tp_family != AF_UNSPEC)
446 daddr = lip->l2tp_addr.s_addr;
449 if (sk->sk_state != TCP_ESTABLISHED)
452 daddr = inet->inet_daddr;
456 /* Allocate a socket buffer */
458 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
459 4 + len, 0, GFP_KERNEL);
463 /* Reserve space for headers, putting IP header on 4-byte boundary. */
464 skb_reserve(skb, 2 + NET_SKB_PAD);
465 skb_reset_network_header(skb);
466 skb_reserve(skb, sizeof(struct iphdr));
467 skb_reset_transport_header(skb);
469 /* Insert 0 session_id */
470 *((__be32 *) skb_put(skb, 4)) = 0;
472 /* Copy user data into skb */
473 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
479 fl4 = &inet->cork.fl.u.ip4;
481 rt = (struct rtable *) __sk_dst_check(sk, 0);
485 const struct ip_options_rcu *inet_opt;
487 inet_opt = rcu_dereference(inet->inet_opt);
489 /* Use correct destination address if we have options. */
490 if (inet_opt && inet_opt->opt.srr)
491 daddr = inet_opt->opt.faddr;
493 /* If this fails, retransmit mechanism of transport layer will
494 * keep trying until route appears or the connection times
497 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
498 daddr, inet->inet_saddr,
499 inet->inet_dport, inet->inet_sport,
500 sk->sk_protocol, RT_CONN_FLAGS(sk),
501 sk->sk_bound_dev_if);
505 sk_setup_caps(sk, &rt->dst);
507 skb_dst_set(skb, &rt->dst);
512 /* We dont need to clone dst here, it is guaranteed to not disappear.
513 * __dev_xmit_skb() might force a refcount if needed.
515 skb_dst_set_noref(skb, &rt->dst);
518 /* Queue the packet to IP for output */
519 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
532 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
538 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
539 size_t len, int noblock, int flags, int *addr_len)
541 struct inet_sock *inet = inet_sk(sk);
543 int err = -EOPNOTSUPP;
544 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
550 skb = skb_recv_datagram(sk, flags, noblock, &err);
556 msg->msg_flags |= MSG_TRUNC;
560 err = skb_copy_datagram_msg(skb, 0, msg, copied);
564 sock_recv_timestamp(msg, sk, skb);
566 /* Copy the address. */
568 sin->sin_family = AF_INET;
569 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
571 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
572 *addr_len = sizeof(*sin);
574 if (inet->cmsg_flags)
575 ip_cmsg_recv(msg, skb);
576 if (flags & MSG_TRUNC)
579 skb_free_datagram(sk, skb);
581 return err ? err : copied;
584 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
591 amount = sk_wmem_alloc_get(sk);
594 spin_lock_bh(&sk->sk_receive_queue.lock);
595 skb = skb_peek(&sk->sk_receive_queue);
596 amount = skb ? skb->len : 0;
597 spin_unlock_bh(&sk->sk_receive_queue.lock);
604 return put_user(amount, (int __user *)arg);
606 EXPORT_SYMBOL(l2tp_ioctl);
608 static struct proto l2tp_ip_prot = {
610 .owner = THIS_MODULE,
611 .init = l2tp_ip_open,
612 .close = l2tp_ip_close,
613 .bind = l2tp_ip_bind,
614 .connect = l2tp_ip_connect,
615 .disconnect = l2tp_ip_disconnect,
617 .destroy = l2tp_ip_destroy_sock,
618 .setsockopt = ip_setsockopt,
619 .getsockopt = ip_getsockopt,
620 .sendmsg = l2tp_ip_sendmsg,
621 .recvmsg = l2tp_ip_recvmsg,
622 .backlog_rcv = l2tp_ip_backlog_recv,
623 .hash = l2tp_ip_hash,
624 .unhash = l2tp_ip_unhash,
625 .obj_size = sizeof(struct l2tp_ip_sock),
627 .compat_setsockopt = compat_ip_setsockopt,
628 .compat_getsockopt = compat_ip_getsockopt,
632 static const struct proto_ops l2tp_ip_ops = {
634 .owner = THIS_MODULE,
635 .release = inet_release,
637 .connect = inet_dgram_connect,
638 .socketpair = sock_no_socketpair,
639 .accept = sock_no_accept,
640 .getname = l2tp_ip_getname,
641 .poll = datagram_poll,
643 .listen = sock_no_listen,
644 .shutdown = inet_shutdown,
645 .setsockopt = sock_common_setsockopt,
646 .getsockopt = sock_common_getsockopt,
647 .sendmsg = inet_sendmsg,
648 .recvmsg = sock_common_recvmsg,
649 .mmap = sock_no_mmap,
650 .sendpage = sock_no_sendpage,
652 .compat_setsockopt = compat_sock_common_setsockopt,
653 .compat_getsockopt = compat_sock_common_getsockopt,
657 static struct inet_protosw l2tp_ip_protosw = {
659 .protocol = IPPROTO_L2TP,
660 .prot = &l2tp_ip_prot,
664 static struct net_protocol l2tp_ip_protocol __read_mostly = {
665 .handler = l2tp_ip_recv,
669 static int __init l2tp_ip_init(void)
673 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
675 err = proto_register(&l2tp_ip_prot, 1);
679 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
683 inet_register_protosw(&l2tp_ip_protosw);
687 proto_unregister(&l2tp_ip_prot);
692 static void __exit l2tp_ip_exit(void)
694 inet_unregister_protosw(&l2tp_ip_protosw);
695 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
696 proto_unregister(&l2tp_ip_prot);
699 module_init(l2tp_ip_init);
700 module_exit(l2tp_ip_exit);
702 MODULE_LICENSE("GPL");
703 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
704 MODULE_DESCRIPTION("L2TP over IP");
705 MODULE_VERSION("1.0");
707 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
710 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
711 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);