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(struct net *net, __be32 laddr, int dif, u32 tunnel_id)
55 sk_for_each_bound(sk, &l2tp_ip_bind_table) {
56 struct inet_sock *inet = inet_sk(sk);
57 struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
62 if ((l2tp->conn_id == tunnel_id) &&
63 net_eq(sock_net(sk), net) &&
64 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
65 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
74 static inline struct sock *l2tp_ip_bind_lookup(struct net *net, __be32 laddr, int dif, u32 tunnel_id)
76 struct sock *sk = __l2tp_ip_bind_lookup(net, laddr, dif, 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;
127 if (!pskb_may_pull(skb, 4))
130 /* Point to L2TP header */
131 optr = ptr = skb->data;
132 session_id = ntohl(*((__be32 *) ptr));
135 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
136 * the session_id. If it is 0, the packet is a L2TP control
137 * frame and the session_id value can be discarded.
139 if (session_id == 0) {
144 /* Ok, this is a data packet. Lookup the session. */
145 session = l2tp_session_find(net, NULL, session_id);
149 tunnel = session->tunnel;
153 /* Trace packet contents, if enabled */
154 if (tunnel->debug & L2TP_MSG_DATA) {
155 length = min(32u, skb->len);
156 if (!pskb_may_pull(skb, length))
159 /* Point to L2TP header */
160 optr = ptr = skb->data;
162 pr_debug("%s: ip recv\n", tunnel->name);
163 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
166 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
171 /* Get the tunnel_id from the L2TP header */
172 if (!pskb_may_pull(skb, 12))
175 if ((skb->data[0] & 0xc0) != 0xc0)
178 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
179 tunnel = l2tp_tunnel_find(net, tunnel_id);
184 struct iphdr *iph = (struct iphdr *) skb_network_header(skb);
186 read_lock_bh(&l2tp_ip_lock);
187 sk = __l2tp_ip_bind_lookup(net, iph->daddr, 0, tunnel_id);
189 read_unlock_bh(&l2tp_ip_lock);
194 read_unlock_bh(&l2tp_ip_lock);
197 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
202 return sk_receive_skb(sk, skb, 1);
212 static int l2tp_ip_open(struct sock *sk)
214 /* Prevent autobind. We don't have ports. */
215 inet_sk(sk)->inet_num = IPPROTO_L2TP;
217 write_lock_bh(&l2tp_ip_lock);
218 sk_add_node(sk, &l2tp_ip_table);
219 write_unlock_bh(&l2tp_ip_lock);
224 static void l2tp_ip_close(struct sock *sk, long timeout)
226 write_lock_bh(&l2tp_ip_lock);
227 hlist_del_init(&sk->sk_bind_node);
228 sk_del_node_init(sk);
229 write_unlock_bh(&l2tp_ip_lock);
230 sk_common_release(sk);
233 static void l2tp_ip_destroy_sock(struct sock *sk)
236 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
238 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
242 l2tp_tunnel_closeall(tunnel);
246 sk_refcnt_debug_dec(sk);
249 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
251 struct inet_sock *inet = inet_sk(sk);
252 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
253 struct net *net = sock_net(sk);
257 if (addr_len < sizeof(struct sockaddr_l2tpip))
259 if (addr->l2tp_family != AF_INET)
263 read_lock_bh(&l2tp_ip_lock);
264 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr,
265 sk->sk_bound_dev_if, addr->l2tp_conn_id))
268 read_unlock_bh(&l2tp_ip_lock);
271 if (!sock_flag(sk, SOCK_ZAPPED))
274 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
277 chk_addr_ret = inet_addr_type(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 if (addr->l2tp_addr.s_addr)
284 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
285 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
286 inet->inet_saddr = 0; /* Use device */
289 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
291 write_lock_bh(&l2tp_ip_lock);
292 sk_add_bind_node(sk, &l2tp_ip_bind_table);
293 sk_del_node_init(sk);
294 write_unlock_bh(&l2tp_ip_lock);
296 sock_reset_flag(sk, SOCK_ZAPPED);
304 read_unlock_bh(&l2tp_ip_lock);
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 (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
317 if (addr_len < sizeof(*lsa))
320 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
323 rc = ip4_datagram_connect(sk, uaddr, addr_len);
329 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
331 write_lock_bh(&l2tp_ip_lock);
332 hlist_del_init(&sk->sk_bind_node);
333 sk_add_bind_node(sk, &l2tp_ip_bind_table);
334 write_unlock_bh(&l2tp_ip_lock);
340 static int l2tp_ip_disconnect(struct sock *sk, int flags)
342 if (sock_flag(sk, SOCK_ZAPPED))
345 return udp_disconnect(sk, flags);
348 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
349 int *uaddr_len, int peer)
351 struct sock *sk = sock->sk;
352 struct inet_sock *inet = inet_sk(sk);
353 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
354 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
356 memset(lsa, 0, sizeof(*lsa));
357 lsa->l2tp_family = AF_INET;
359 if (!inet->inet_dport)
361 lsa->l2tp_conn_id = lsk->peer_conn_id;
362 lsa->l2tp_addr.s_addr = inet->inet_daddr;
364 __be32 addr = inet->inet_rcv_saddr;
366 addr = inet->inet_saddr;
367 lsa->l2tp_conn_id = lsk->conn_id;
368 lsa->l2tp_addr.s_addr = addr;
370 *uaddr_len = sizeof(*lsa);
374 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
378 /* Charge it to the socket, dropping if the queue is full. */
379 rc = sock_queue_rcv_skb(sk, skb);
386 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
391 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
394 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
398 struct inet_sock *inet = inet_sk(sk);
399 struct rtable *rt = NULL;
407 if (sock_flag(sk, SOCK_DEAD))
410 /* Get and verify the address. */
412 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
414 if (msg->msg_namelen < sizeof(*lip))
417 if (lip->l2tp_family != AF_INET) {
419 if (lip->l2tp_family != AF_UNSPEC)
423 daddr = lip->l2tp_addr.s_addr;
426 if (sk->sk_state != TCP_ESTABLISHED)
429 daddr = inet->inet_daddr;
433 /* Allocate a socket buffer */
435 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
436 4 + len, 0, GFP_KERNEL);
440 /* Reserve space for headers, putting IP header on 4-byte boundary. */
441 skb_reserve(skb, 2 + NET_SKB_PAD);
442 skb_reset_network_header(skb);
443 skb_reserve(skb, sizeof(struct iphdr));
444 skb_reset_transport_header(skb);
446 /* Insert 0 session_id */
447 *((__be32 *) skb_put(skb, 4)) = 0;
449 /* Copy user data into skb */
450 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
456 fl4 = &inet->cork.fl.u.ip4;
458 rt = (struct rtable *) __sk_dst_check(sk, 0);
462 const struct ip_options_rcu *inet_opt;
464 inet_opt = rcu_dereference(inet->inet_opt);
466 /* Use correct destination address if we have options. */
467 if (inet_opt && inet_opt->opt.srr)
468 daddr = inet_opt->opt.faddr;
470 /* If this fails, retransmit mechanism of transport layer will
471 * keep trying until route appears or the connection times
474 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
475 daddr, inet->inet_saddr,
476 inet->inet_dport, inet->inet_sport,
477 sk->sk_protocol, RT_CONN_FLAGS(sk),
478 sk->sk_bound_dev_if);
482 sk_setup_caps(sk, &rt->dst);
484 skb_dst_set(skb, &rt->dst);
489 /* We dont need to clone dst here, it is guaranteed to not disappear.
490 * __dev_xmit_skb() might force a refcount if needed.
492 skb_dst_set_noref(skb, &rt->dst);
495 /* Queue the packet to IP for output */
496 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
509 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
515 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
516 size_t len, int noblock, int flags, int *addr_len)
518 struct inet_sock *inet = inet_sk(sk);
520 int err = -EOPNOTSUPP;
521 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
527 skb = skb_recv_datagram(sk, flags, noblock, &err);
533 msg->msg_flags |= MSG_TRUNC;
537 err = skb_copy_datagram_msg(skb, 0, msg, copied);
541 sock_recv_timestamp(msg, sk, skb);
543 /* Copy the address. */
545 sin->sin_family = AF_INET;
546 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
548 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
549 *addr_len = sizeof(*sin);
551 if (inet->cmsg_flags)
552 ip_cmsg_recv(msg, skb);
553 if (flags & MSG_TRUNC)
556 skb_free_datagram(sk, skb);
558 return err ? err : copied;
561 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
568 amount = sk_wmem_alloc_get(sk);
571 spin_lock_bh(&sk->sk_receive_queue.lock);
572 skb = skb_peek(&sk->sk_receive_queue);
573 amount = skb ? skb->len : 0;
574 spin_unlock_bh(&sk->sk_receive_queue.lock);
581 return put_user(amount, (int __user *)arg);
583 EXPORT_SYMBOL(l2tp_ioctl);
585 static struct proto l2tp_ip_prot = {
587 .owner = THIS_MODULE,
588 .init = l2tp_ip_open,
589 .close = l2tp_ip_close,
590 .bind = l2tp_ip_bind,
591 .connect = l2tp_ip_connect,
592 .disconnect = l2tp_ip_disconnect,
594 .destroy = l2tp_ip_destroy_sock,
595 .setsockopt = ip_setsockopt,
596 .getsockopt = ip_getsockopt,
597 .sendmsg = l2tp_ip_sendmsg,
598 .recvmsg = l2tp_ip_recvmsg,
599 .backlog_rcv = l2tp_ip_backlog_recv,
601 .unhash = inet_unhash,
602 .obj_size = sizeof(struct l2tp_ip_sock),
604 .compat_setsockopt = compat_ip_setsockopt,
605 .compat_getsockopt = compat_ip_getsockopt,
609 static const struct proto_ops l2tp_ip_ops = {
611 .owner = THIS_MODULE,
612 .release = inet_release,
614 .connect = inet_dgram_connect,
615 .socketpair = sock_no_socketpair,
616 .accept = sock_no_accept,
617 .getname = l2tp_ip_getname,
618 .poll = datagram_poll,
620 .listen = sock_no_listen,
621 .shutdown = inet_shutdown,
622 .setsockopt = sock_common_setsockopt,
623 .getsockopt = sock_common_getsockopt,
624 .sendmsg = inet_sendmsg,
625 .recvmsg = sock_common_recvmsg,
626 .mmap = sock_no_mmap,
627 .sendpage = sock_no_sendpage,
629 .compat_setsockopt = compat_sock_common_setsockopt,
630 .compat_getsockopt = compat_sock_common_getsockopt,
634 static struct inet_protosw l2tp_ip_protosw = {
636 .protocol = IPPROTO_L2TP,
637 .prot = &l2tp_ip_prot,
641 static struct net_protocol l2tp_ip_protocol __read_mostly = {
642 .handler = l2tp_ip_recv,
646 static int __init l2tp_ip_init(void)
650 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
652 err = proto_register(&l2tp_ip_prot, 1);
656 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
660 inet_register_protosw(&l2tp_ip_protosw);
664 proto_unregister(&l2tp_ip_prot);
669 static void __exit l2tp_ip_exit(void)
671 inet_unregister_protosw(&l2tp_ip_protosw);
672 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
673 proto_unregister(&l2tp_ip_prot);
676 module_init(l2tp_ip_init);
677 module_exit(l2tp_ip_exit);
679 MODULE_LICENSE("GPL");
680 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
681 MODULE_DESCRIPTION("L2TP over IP");
682 MODULE_VERSION("1.0");
684 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
687 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
688 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);