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_get(&init_net, NULL, session_id, true);
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);
179 l2tp_session_dec_refcount(session);
184 /* Get the tunnel_id from the L2TP header */
185 if (!pskb_may_pull(skb, 12))
188 if ((skb->data[0] & 0xc0) != 0xc0)
191 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
192 tunnel = l2tp_tunnel_find(&init_net, tunnel_id);
197 struct iphdr *iph = (struct iphdr *) skb_network_header(skb);
199 read_lock_bh(&l2tp_ip_lock);
200 sk = __l2tp_ip_bind_lookup(&init_net, iph->daddr, 0, tunnel_id);
202 read_unlock_bh(&l2tp_ip_lock);
207 read_unlock_bh(&l2tp_ip_lock);
210 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
215 return sk_receive_skb(sk, skb, 1);
219 session->deref(session);
220 l2tp_session_dec_refcount(session);
231 static int l2tp_ip_open(struct sock *sk)
233 /* Prevent autobind. We don't have ports. */
234 inet_sk(sk)->inet_num = IPPROTO_L2TP;
236 write_lock_bh(&l2tp_ip_lock);
237 sk_add_node(sk, &l2tp_ip_table);
238 write_unlock_bh(&l2tp_ip_lock);
243 static void l2tp_ip_close(struct sock *sk, long timeout)
245 write_lock_bh(&l2tp_ip_lock);
246 hlist_del_init(&sk->sk_bind_node);
247 hlist_del_init(&sk->sk_node);
248 write_unlock_bh(&l2tp_ip_lock);
249 sk_common_release(sk);
252 static void l2tp_ip_destroy_sock(struct sock *sk)
256 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
259 sk_refcnt_debug_dec(sk);
262 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
264 struct inet_sock *inet = inet_sk(sk);
265 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
269 if (addr_len < sizeof(struct sockaddr_l2tpip))
271 if (addr->l2tp_family != AF_INET)
275 read_lock_bh(&l2tp_ip_lock);
276 if (__l2tp_ip_bind_lookup(&init_net, addr->l2tp_addr.s_addr, sk->sk_bound_dev_if, addr->l2tp_conn_id))
279 read_unlock_bh(&l2tp_ip_lock);
282 if (!sock_flag(sk, SOCK_ZAPPED))
285 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
288 chk_addr_ret = inet_addr_type(&init_net, addr->l2tp_addr.s_addr);
289 ret = -EADDRNOTAVAIL;
290 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
291 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
294 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
295 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
296 inet->inet_saddr = 0; /* Use device */
299 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
301 write_lock_bh(&l2tp_ip_lock);
302 sk_add_bind_node(sk, &l2tp_ip_bind_table);
303 sk_del_node_init(sk);
304 write_unlock_bh(&l2tp_ip_lock);
306 sock_reset_flag(sk, SOCK_ZAPPED);
314 read_unlock_bh(&l2tp_ip_lock);
319 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
321 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
322 struct inet_sock *inet = inet_sk(sk);
328 if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
331 if (addr_len < sizeof(*lsa))
334 if (lsa->l2tp_family != AF_INET)
335 return -EAFNOSUPPORT;
341 oif = sk->sk_bound_dev_if;
342 saddr = inet->inet_saddr;
345 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
348 fl4 = &inet->cork.fl.u.ip4;
349 rt = ip_route_connect(fl4, lsa->l2tp_addr.s_addr, saddr,
350 RT_CONN_FLAGS(sk), oif,
355 if (rc == -ENETUNREACH)
356 IP_INC_STATS_BH(&init_net, IPSTATS_MIB_OUTNOROUTES);
361 if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
366 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
368 if (!inet->inet_saddr)
369 inet->inet_saddr = fl4->saddr;
370 if (!inet->inet_rcv_saddr)
371 inet->inet_rcv_saddr = fl4->saddr;
372 inet->inet_daddr = fl4->daddr;
373 sk->sk_state = TCP_ESTABLISHED;
374 inet->inet_id = jiffies;
376 sk_dst_set(sk, &rt->dst);
378 write_lock_bh(&l2tp_ip_lock);
379 hlist_del_init(&sk->sk_bind_node);
380 sk_add_bind_node(sk, &l2tp_ip_bind_table);
381 write_unlock_bh(&l2tp_ip_lock);
389 static int l2tp_ip_disconnect(struct sock *sk, int flags)
391 if (sock_flag(sk, SOCK_ZAPPED))
394 return udp_disconnect(sk, flags);
397 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
398 int *uaddr_len, int peer)
400 struct sock *sk = sock->sk;
401 struct inet_sock *inet = inet_sk(sk);
402 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
403 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
405 memset(lsa, 0, sizeof(*lsa));
406 lsa->l2tp_family = AF_INET;
408 if (!inet->inet_dport)
410 lsa->l2tp_conn_id = lsk->peer_conn_id;
411 lsa->l2tp_addr.s_addr = inet->inet_daddr;
413 __be32 addr = inet->inet_rcv_saddr;
415 addr = inet->inet_saddr;
416 lsa->l2tp_conn_id = lsk->conn_id;
417 lsa->l2tp_addr.s_addr = addr;
419 *uaddr_len = sizeof(*lsa);
423 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
427 /* Charge it to the socket, dropping if the queue is full. */
428 rc = sock_queue_rcv_skb(sk, skb);
435 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
440 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
443 static int l2tp_ip_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, size_t len)
447 struct l2tp_ip_sock *lsa = l2tp_ip_sk(sk);
448 struct inet_sock *inet = inet_sk(sk);
449 struct rtable *rt = NULL;
457 if (sock_flag(sk, SOCK_DEAD))
460 /* Get and verify the address. */
462 struct sockaddr_l2tpip *lip = (struct sockaddr_l2tpip *) msg->msg_name;
464 if (msg->msg_namelen < sizeof(*lip))
467 if (lip->l2tp_family != AF_INET) {
469 if (lip->l2tp_family != AF_UNSPEC)
473 daddr = lip->l2tp_addr.s_addr;
476 if (sk->sk_state != TCP_ESTABLISHED)
479 daddr = inet->inet_daddr;
483 /* Allocate a socket buffer */
485 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
486 4 + len, 0, GFP_KERNEL);
490 /* Reserve space for headers, putting IP header on 4-byte boundary. */
491 skb_reserve(skb, 2 + NET_SKB_PAD);
492 skb_reset_network_header(skb);
493 skb_reserve(skb, sizeof(struct iphdr));
494 skb_reset_transport_header(skb);
496 /* Insert 0 session_id */
497 *((__be32 *) skb_put(skb, 4)) = 0;
499 /* Copy user data into skb */
500 rc = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
506 fl4 = &inet->cork.fl.u.ip4;
508 rt = (struct rtable *) __sk_dst_check(sk, 0);
512 const struct ip_options_rcu *inet_opt;
514 inet_opt = rcu_dereference(inet->inet_opt);
516 /* Use correct destination address if we have options. */
517 if (inet_opt && inet_opt->opt.srr)
518 daddr = inet_opt->opt.faddr;
520 /* If this fails, retransmit mechanism of transport layer will
521 * keep trying until route appears or the connection times
524 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
525 daddr, inet->inet_saddr,
526 inet->inet_dport, inet->inet_sport,
527 sk->sk_protocol, RT_CONN_FLAGS(sk),
528 sk->sk_bound_dev_if);
532 sk_setup_caps(sk, &rt->dst);
534 skb_dst_set(skb, &rt->dst);
539 /* We dont need to clone dst here, it is guaranteed to not disappear.
540 * __dev_xmit_skb() might force a refcount if needed.
542 skb_dst_set_noref(skb, &rt->dst);
545 /* Queue the packet to IP for output */
546 rc = ip_queue_xmit(skb, &inet->cork.fl);
553 lsa->tx_bytes += len;
565 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
571 static int l2tp_ip_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
572 size_t len, int noblock, int flags, int *addr_len)
574 struct inet_sock *inet = inet_sk(sk);
575 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
577 int err = -EOPNOTSUPP;
578 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
584 skb = skb_recv_datagram(sk, flags, noblock, &err);
590 msg->msg_flags |= MSG_TRUNC;
594 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
598 sock_recv_timestamp(msg, sk, skb);
600 /* Copy the address. */
602 sin->sin_family = AF_INET;
603 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
605 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
606 *addr_len = sizeof(*sin);
608 if (inet->cmsg_flags)
609 ip_cmsg_recv(msg, skb);
610 if (flags & MSG_TRUNC)
613 skb_free_datagram(sk, skb);
621 lsk->rx_bytes += copied;
626 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
633 amount = sk_wmem_alloc_get(sk);
636 spin_lock_bh(&sk->sk_receive_queue.lock);
637 skb = skb_peek(&sk->sk_receive_queue);
638 amount = skb ? skb->len : 0;
639 spin_unlock_bh(&sk->sk_receive_queue.lock);
646 return put_user(amount, (int __user *)arg);
648 EXPORT_SYMBOL(l2tp_ioctl);
650 static struct proto l2tp_ip_prot = {
652 .owner = THIS_MODULE,
653 .init = l2tp_ip_open,
654 .close = l2tp_ip_close,
655 .bind = l2tp_ip_bind,
656 .connect = l2tp_ip_connect,
657 .disconnect = l2tp_ip_disconnect,
659 .destroy = l2tp_ip_destroy_sock,
660 .setsockopt = ip_setsockopt,
661 .getsockopt = ip_getsockopt,
662 .sendmsg = l2tp_ip_sendmsg,
663 .recvmsg = l2tp_ip_recvmsg,
664 .backlog_rcv = l2tp_ip_backlog_recv,
666 .unhash = inet_unhash,
667 .obj_size = sizeof(struct l2tp_ip_sock),
669 .compat_setsockopt = compat_ip_setsockopt,
670 .compat_getsockopt = compat_ip_getsockopt,
674 static const struct proto_ops l2tp_ip_ops = {
676 .owner = THIS_MODULE,
677 .release = inet_release,
679 .connect = inet_dgram_connect,
680 .socketpair = sock_no_socketpair,
681 .accept = sock_no_accept,
682 .getname = l2tp_ip_getname,
683 .poll = datagram_poll,
685 .listen = sock_no_listen,
686 .shutdown = inet_shutdown,
687 .setsockopt = sock_common_setsockopt,
688 .getsockopt = sock_common_getsockopt,
689 .sendmsg = inet_sendmsg,
690 .recvmsg = sock_common_recvmsg,
691 .mmap = sock_no_mmap,
692 .sendpage = sock_no_sendpage,
694 .compat_setsockopt = compat_sock_common_setsockopt,
695 .compat_getsockopt = compat_sock_common_getsockopt,
699 static struct inet_protosw l2tp_ip_protosw = {
701 .protocol = IPPROTO_L2TP,
702 .prot = &l2tp_ip_prot,
707 static struct net_protocol l2tp_ip_protocol __read_mostly = {
708 .handler = l2tp_ip_recv,
711 static int __init l2tp_ip_init(void)
715 printk(KERN_INFO "L2TP IP encapsulation support (L2TPv3)\n");
717 err = proto_register(&l2tp_ip_prot, 1);
721 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
725 inet_register_protosw(&l2tp_ip_protosw);
729 proto_unregister(&l2tp_ip_prot);
734 static void __exit l2tp_ip_exit(void)
736 inet_unregister_protosw(&l2tp_ip_protosw);
737 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
738 proto_unregister(&l2tp_ip_prot);
741 module_init(l2tp_ip_init);
742 module_exit(l2tp_ip_exit);
744 MODULE_LICENSE("GPL");
745 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
746 MODULE_DESCRIPTION("L2TP over IP");
747 MODULE_VERSION("1.0");
749 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
752 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);