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/inet_hashtables.h>
28 #include <net/tcp_states.h>
29 #include <net/protocol.h>
32 #include "l2tp_core.h"
35 /* inet_sock has to be the first member of l2tp_ip_sock */
36 struct inet_sock inet;
42 static DEFINE_RWLOCK(l2tp_ip_lock);
43 static struct hlist_head l2tp_ip_table;
44 static struct hlist_head l2tp_ip_bind_table;
46 static inline struct l2tp_ip_sock *l2tp_ip_sk(const struct sock *sk)
48 return (struct l2tp_ip_sock *)sk;
51 static struct sock *__l2tp_ip_bind_lookup(const struct net *net, __be32 laddr,
52 __be32 raddr, int dif, u32 tunnel_id)
56 sk_for_each_bound(sk, &l2tp_ip_bind_table) {
57 const struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
58 const struct inet_sock *inet = inet_sk(sk);
60 if (!net_eq(sock_net(sk), net))
63 if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
66 if (inet->inet_rcv_saddr && laddr &&
67 inet->inet_rcv_saddr != laddr)
70 if (inet->inet_daddr && raddr && inet->inet_daddr != raddr)
73 if (l2tp->conn_id != tunnel_id)
84 /* When processing receive frames, there are two cases to
85 * consider. Data frames consist of a non-zero session-id and an
86 * optional cookie. Control frames consist of a regular L2TP header
87 * preceded by 32-bits of zeros.
89 * L2TPv3 Session Header Over IP
92 * 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
93 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
95 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
96 * | Cookie (optional, maximum 64 bits)...
97 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
99 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
101 * L2TPv3 Control Message Header Over IP
104 * 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
105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * | (32 bits of zeros) |
107 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
109 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
110 * | Control Connection ID |
111 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115 * All control frames are passed to userspace.
117 static int l2tp_ip_recv(struct sk_buff *skb)
119 struct net *net = dev_net(skb->dev);
123 unsigned char *ptr, *optr;
124 struct l2tp_session *session;
125 struct l2tp_tunnel *tunnel = NULL;
129 if (!pskb_may_pull(skb, 4))
132 /* Point to L2TP header */
133 optr = ptr = skb->data;
134 session_id = ntohl(*((__be32 *) ptr));
137 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
138 * the session_id. If it is 0, the packet is a L2TP control
139 * frame and the session_id value can be discarded.
141 if (session_id == 0) {
146 /* Ok, this is a data packet. Lookup the session. */
147 session = l2tp_session_get(net, NULL, session_id, true);
151 tunnel = session->tunnel;
155 /* Trace packet contents, if enabled */
156 if (tunnel->debug & L2TP_MSG_DATA) {
157 length = min(32u, skb->len);
158 if (!pskb_may_pull(skb, length))
161 /* Point to L2TP header */
162 optr = ptr = skb->data;
164 pr_debug("%s: ip recv\n", tunnel->name);
165 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
168 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
169 l2tp_session_dec_refcount(session);
174 /* Get the tunnel_id from the L2TP header */
175 if (!pskb_may_pull(skb, 12))
178 if ((skb->data[0] & 0xc0) != 0xc0)
181 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
182 iph = (struct iphdr *)skb_network_header(skb);
184 read_lock_bh(&l2tp_ip_lock);
185 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr, inet_iif(skb),
188 read_unlock_bh(&l2tp_ip_lock);
192 read_unlock_bh(&l2tp_ip_lock);
194 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
199 return sk_receive_skb(sk, skb, 1);
203 session->deref(session);
204 l2tp_session_dec_refcount(session);
215 static int l2tp_ip_open(struct sock *sk)
217 /* Prevent autobind. We don't have ports. */
218 inet_sk(sk)->inet_num = IPPROTO_L2TP;
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_close(struct sock *sk, long timeout)
229 write_lock_bh(&l2tp_ip_lock);
230 hlist_del_init(&sk->sk_bind_node);
231 sk_del_node_init(sk);
232 write_unlock_bh(&l2tp_ip_lock);
233 sk_common_release(sk);
236 static void l2tp_ip_destroy_sock(struct sock *sk)
239 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
241 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
245 l2tp_tunnel_closeall(tunnel);
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;
256 struct net *net = sock_net(sk);
260 if (addr_len < sizeof(struct sockaddr_l2tpip))
262 if (addr->l2tp_family != AF_INET)
268 if (!sock_flag(sk, SOCK_ZAPPED))
271 if (sk->sk_state != TCP_CLOSE)
274 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
275 ret = -EADDRNOTAVAIL;
276 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
277 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
280 if (addr->l2tp_addr.s_addr)
281 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
282 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
283 inet->inet_saddr = 0; /* Use device */
285 write_lock_bh(&l2tp_ip_lock);
286 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
287 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
288 write_unlock_bh(&l2tp_ip_lock);
294 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
296 sk_add_bind_node(sk, &l2tp_ip_bind_table);
297 sk_del_node_init(sk);
298 write_unlock_bh(&l2tp_ip_lock);
301 sock_reset_flag(sk, SOCK_ZAPPED);
309 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
311 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
314 if (addr_len < sizeof(*lsa))
317 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
322 /* Must bind first - autobinding does not work */
323 if (sock_flag(sk, SOCK_ZAPPED)) {
328 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
332 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
334 write_lock_bh(&l2tp_ip_lock);
335 hlist_del_init(&sk->sk_bind_node);
336 sk_add_bind_node(sk, &l2tp_ip_bind_table);
337 write_unlock_bh(&l2tp_ip_lock);
345 static int l2tp_ip_disconnect(struct sock *sk, int flags)
347 if (sock_flag(sk, SOCK_ZAPPED))
350 return __udp_disconnect(sk, flags);
353 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
354 int *uaddr_len, int peer)
356 struct sock *sk = sock->sk;
357 struct inet_sock *inet = inet_sk(sk);
358 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
359 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
361 memset(lsa, 0, sizeof(*lsa));
362 lsa->l2tp_family = AF_INET;
364 if (!inet->inet_dport)
366 lsa->l2tp_conn_id = lsk->peer_conn_id;
367 lsa->l2tp_addr.s_addr = inet->inet_daddr;
369 __be32 addr = inet->inet_rcv_saddr;
371 addr = inet->inet_saddr;
372 lsa->l2tp_conn_id = lsk->conn_id;
373 lsa->l2tp_addr.s_addr = addr;
375 *uaddr_len = sizeof(*lsa);
379 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
383 /* Charge it to the socket, dropping if the queue is full. */
384 rc = sock_queue_rcv_skb(sk, skb);
391 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
396 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
399 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
403 struct inet_sock *inet = inet_sk(sk);
404 struct rtable *rt = NULL;
412 if (sock_flag(sk, SOCK_DEAD))
415 /* Get and verify the address. */
417 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
419 if (msg->msg_namelen < sizeof(*lip))
422 if (lip->l2tp_family != AF_INET) {
424 if (lip->l2tp_family != AF_UNSPEC)
428 daddr = lip->l2tp_addr.s_addr;
431 if (sk->sk_state != TCP_ESTABLISHED)
434 daddr = inet->inet_daddr;
438 /* Allocate a socket buffer */
440 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
441 4 + len, 0, GFP_KERNEL);
445 /* Reserve space for headers, putting IP header on 4-byte boundary. */
446 skb_reserve(skb, 2 + NET_SKB_PAD);
447 skb_reset_network_header(skb);
448 skb_reserve(skb, sizeof(struct iphdr));
449 skb_reset_transport_header(skb);
451 /* Insert 0 session_id */
452 *((__be32 *) skb_put(skb, 4)) = 0;
454 /* Copy user data into skb */
455 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
461 fl4 = &inet->cork.fl.u.ip4;
463 rt = (struct rtable *) __sk_dst_check(sk, 0);
467 const struct ip_options_rcu *inet_opt;
469 inet_opt = rcu_dereference(inet->inet_opt);
471 /* Use correct destination address if we have options. */
472 if (inet_opt && inet_opt->opt.srr)
473 daddr = inet_opt->opt.faddr;
475 /* If this fails, retransmit mechanism of transport layer will
476 * keep trying until route appears or the connection times
479 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
480 daddr, inet->inet_saddr,
481 inet->inet_dport, inet->inet_sport,
482 sk->sk_protocol, RT_CONN_FLAGS(sk),
483 sk->sk_bound_dev_if);
487 sk_setup_caps(sk, &rt->dst);
489 skb_dst_set(skb, &rt->dst);
494 /* We dont need to clone dst here, it is guaranteed to not disappear.
495 * __dev_xmit_skb() might force a refcount if needed.
497 skb_dst_set_noref(skb, &rt->dst);
500 /* Queue the packet to IP for output */
501 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
514 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
520 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
521 size_t len, int noblock, int flags, int *addr_len)
523 struct inet_sock *inet = inet_sk(sk);
525 int err = -EOPNOTSUPP;
526 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
532 skb = skb_recv_datagram(sk, flags, noblock, &err);
538 msg->msg_flags |= MSG_TRUNC;
542 err = skb_copy_datagram_msg(skb, 0, msg, copied);
546 sock_recv_timestamp(msg, sk, skb);
548 /* Copy the address. */
550 sin->sin_family = AF_INET;
551 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
553 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
554 *addr_len = sizeof(*sin);
556 if (inet->cmsg_flags)
557 ip_cmsg_recv(msg, skb);
558 if (flags & MSG_TRUNC)
561 skb_free_datagram(sk, skb);
563 return err ? err : copied;
566 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
573 amount = sk_wmem_alloc_get(sk);
576 spin_lock_bh(&sk->sk_receive_queue.lock);
577 skb = skb_peek(&sk->sk_receive_queue);
578 amount = skb ? skb->len : 0;
579 spin_unlock_bh(&sk->sk_receive_queue.lock);
586 return put_user(amount, (int __user *)arg);
588 EXPORT_SYMBOL(l2tp_ioctl);
590 static struct proto l2tp_ip_prot = {
592 .owner = THIS_MODULE,
593 .init = l2tp_ip_open,
594 .close = l2tp_ip_close,
595 .bind = l2tp_ip_bind,
596 .connect = l2tp_ip_connect,
597 .disconnect = l2tp_ip_disconnect,
599 .destroy = l2tp_ip_destroy_sock,
600 .setsockopt = ip_setsockopt,
601 .getsockopt = ip_getsockopt,
602 .sendmsg = l2tp_ip_sendmsg,
603 .recvmsg = l2tp_ip_recvmsg,
604 .backlog_rcv = l2tp_ip_backlog_recv,
606 .unhash = inet_unhash,
607 .obj_size = sizeof(struct l2tp_ip_sock),
609 .compat_setsockopt = compat_ip_setsockopt,
610 .compat_getsockopt = compat_ip_getsockopt,
614 static const struct proto_ops l2tp_ip_ops = {
616 .owner = THIS_MODULE,
617 .release = inet_release,
619 .connect = inet_dgram_connect,
620 .socketpair = sock_no_socketpair,
621 .accept = sock_no_accept,
622 .getname = l2tp_ip_getname,
623 .poll = datagram_poll,
625 .listen = sock_no_listen,
626 .shutdown = inet_shutdown,
627 .setsockopt = sock_common_setsockopt,
628 .getsockopt = sock_common_getsockopt,
629 .sendmsg = inet_sendmsg,
630 .recvmsg = sock_common_recvmsg,
631 .mmap = sock_no_mmap,
632 .sendpage = sock_no_sendpage,
634 .compat_setsockopt = compat_sock_common_setsockopt,
635 .compat_getsockopt = compat_sock_common_getsockopt,
639 static struct inet_protosw l2tp_ip_protosw = {
641 .protocol = IPPROTO_L2TP,
642 .prot = &l2tp_ip_prot,
646 static struct net_protocol l2tp_ip_protocol __read_mostly = {
647 .handler = l2tp_ip_recv,
651 static int __init l2tp_ip_init(void)
655 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
657 err = proto_register(&l2tp_ip_prot, 1);
661 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
665 inet_register_protosw(&l2tp_ip_protosw);
669 proto_unregister(&l2tp_ip_prot);
674 static void __exit l2tp_ip_exit(void)
676 inet_unregister_protosw(&l2tp_ip_protosw);
677 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
678 proto_unregister(&l2tp_ip_prot);
681 module_init(l2tp_ip_init);
682 module_exit(l2tp_ip_exit);
684 MODULE_LICENSE("GPL");
685 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
686 MODULE_DESCRIPTION("L2TP over IP");
687 MODULE_VERSION("1.0");
689 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
692 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
693 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);