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 struct inet_sock *inet = inet_sk(sk);
58 struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
63 if ((l2tp->conn_id == tunnel_id) &&
64 net_eq(sock_net(sk), net) &&
65 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
66 (!inet->inet_daddr || !raddr || inet->inet_daddr == raddr) &&
67 (!sk->sk_bound_dev_if || !dif ||
68 sk->sk_bound_dev_if == dif))
77 /* When processing receive frames, there are two cases to
78 * consider. Data frames consist of a non-zero session-id and an
79 * optional cookie. Control frames consist of a regular L2TP header
80 * preceded by 32-bits of zeros.
82 * L2TPv3 Session Header Over IP
85 * 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
86 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
88 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
89 * | Cookie (optional, maximum 64 bits)...
90 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
92 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
94 * L2TPv3 Control Message 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 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
99 * | (32 bits of zeros) |
100 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
101 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
102 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103 * | Control Connection ID |
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * All control frames are passed to userspace.
110 static int l2tp_ip_recv(struct sk_buff *skb)
112 struct net *net = dev_net(skb->dev);
116 unsigned char *ptr, *optr;
117 struct l2tp_session *session;
118 struct l2tp_tunnel *tunnel = NULL;
121 if (!pskb_may_pull(skb, 4))
124 /* Point to L2TP header */
125 optr = ptr = skb->data;
126 session_id = ntohl(*((__be32 *) ptr));
129 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
130 * the session_id. If it is 0, the packet is a L2TP control
131 * frame and the session_id value can be discarded.
133 if (session_id == 0) {
138 /* Ok, this is a data packet. Lookup the session. */
139 session = l2tp_session_find(net, NULL, session_id);
143 tunnel = session->tunnel;
147 /* Trace packet contents, if enabled */
148 if (tunnel->debug & L2TP_MSG_DATA) {
149 length = min(32u, skb->len);
150 if (!pskb_may_pull(skb, length))
153 /* Point to L2TP header */
154 optr = ptr = skb->data;
156 pr_debug("%s: ip recv\n", tunnel->name);
157 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
160 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
165 /* Get the tunnel_id from the L2TP header */
166 if (!pskb_may_pull(skb, 12))
169 if ((skb->data[0] & 0xc0) != 0xc0)
172 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
173 tunnel = l2tp_tunnel_find(net, tunnel_id);
177 struct iphdr *iph = (struct iphdr *) skb_network_header(skb);
179 read_lock_bh(&l2tp_ip_lock);
180 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr,
181 inet_iif(skb), tunnel_id);
183 read_unlock_bh(&l2tp_ip_lock);
188 read_unlock_bh(&l2tp_ip_lock);
191 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
196 return sk_receive_skb(sk, skb, 1);
206 static int l2tp_ip_open(struct sock *sk)
208 /* Prevent autobind. We don't have ports. */
209 inet_sk(sk)->inet_num = IPPROTO_L2TP;
211 write_lock_bh(&l2tp_ip_lock);
212 sk_add_node(sk, &l2tp_ip_table);
213 write_unlock_bh(&l2tp_ip_lock);
218 static void l2tp_ip_close(struct sock *sk, long timeout)
220 write_lock_bh(&l2tp_ip_lock);
221 hlist_del_init(&sk->sk_bind_node);
222 sk_del_node_init(sk);
223 write_unlock_bh(&l2tp_ip_lock);
224 sk_common_release(sk);
227 static void l2tp_ip_destroy_sock(struct sock *sk)
230 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
232 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
236 l2tp_tunnel_closeall(tunnel);
240 sk_refcnt_debug_dec(sk);
243 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
245 struct inet_sock *inet = inet_sk(sk);
246 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
247 struct net *net = sock_net(sk);
251 if (addr_len < sizeof(struct sockaddr_l2tpip))
253 if (addr->l2tp_family != AF_INET)
259 if (!sock_flag(sk, SOCK_ZAPPED))
262 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
265 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
266 ret = -EADDRNOTAVAIL;
267 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
268 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
271 if (addr->l2tp_addr.s_addr)
272 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
273 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
274 inet->inet_saddr = 0; /* Use device */
276 write_lock_bh(&l2tp_ip_lock);
277 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
278 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
279 write_unlock_bh(&l2tp_ip_lock);
285 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
287 sk_add_bind_node(sk, &l2tp_ip_bind_table);
288 sk_del_node_init(sk);
289 write_unlock_bh(&l2tp_ip_lock);
292 sock_reset_flag(sk, SOCK_ZAPPED);
300 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
302 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
305 if (addr_len < sizeof(*lsa))
308 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
313 /* Must bind first - autobinding does not work */
314 if (sock_flag(sk, SOCK_ZAPPED)) {
319 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
323 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
325 write_lock_bh(&l2tp_ip_lock);
326 hlist_del_init(&sk->sk_bind_node);
327 sk_add_bind_node(sk, &l2tp_ip_bind_table);
328 write_unlock_bh(&l2tp_ip_lock);
336 static int l2tp_ip_disconnect(struct sock *sk, int flags)
338 if (sock_flag(sk, SOCK_ZAPPED))
341 return __udp_disconnect(sk, flags);
344 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
345 int *uaddr_len, int peer)
347 struct sock *sk = sock->sk;
348 struct inet_sock *inet = inet_sk(sk);
349 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
350 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
352 memset(lsa, 0, sizeof(*lsa));
353 lsa->l2tp_family = AF_INET;
355 if (!inet->inet_dport)
357 lsa->l2tp_conn_id = lsk->peer_conn_id;
358 lsa->l2tp_addr.s_addr = inet->inet_daddr;
360 __be32 addr = inet->inet_rcv_saddr;
362 addr = inet->inet_saddr;
363 lsa->l2tp_conn_id = lsk->conn_id;
364 lsa->l2tp_addr.s_addr = addr;
366 *uaddr_len = sizeof(*lsa);
370 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
374 /* Charge it to the socket, dropping if the queue is full. */
375 rc = sock_queue_rcv_skb(sk, skb);
382 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
387 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
390 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
394 struct inet_sock *inet = inet_sk(sk);
395 struct rtable *rt = NULL;
403 if (sock_flag(sk, SOCK_DEAD))
406 /* Get and verify the address. */
408 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
410 if (msg->msg_namelen < sizeof(*lip))
413 if (lip->l2tp_family != AF_INET) {
415 if (lip->l2tp_family != AF_UNSPEC)
419 daddr = lip->l2tp_addr.s_addr;
422 if (sk->sk_state != TCP_ESTABLISHED)
425 daddr = inet->inet_daddr;
429 /* Allocate a socket buffer */
431 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
432 4 + len, 0, GFP_KERNEL);
436 /* Reserve space for headers, putting IP header on 4-byte boundary. */
437 skb_reserve(skb, 2 + NET_SKB_PAD);
438 skb_reset_network_header(skb);
439 skb_reserve(skb, sizeof(struct iphdr));
440 skb_reset_transport_header(skb);
442 /* Insert 0 session_id */
443 *((__be32 *) skb_put(skb, 4)) = 0;
445 /* Copy user data into skb */
446 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
452 fl4 = &inet->cork.fl.u.ip4;
454 rt = (struct rtable *) __sk_dst_check(sk, 0);
458 const struct ip_options_rcu *inet_opt;
460 inet_opt = rcu_dereference(inet->inet_opt);
462 /* Use correct destination address if we have options. */
463 if (inet_opt && inet_opt->opt.srr)
464 daddr = inet_opt->opt.faddr;
466 /* If this fails, retransmit mechanism of transport layer will
467 * keep trying until route appears or the connection times
470 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
471 daddr, inet->inet_saddr,
472 inet->inet_dport, inet->inet_sport,
473 sk->sk_protocol, RT_CONN_FLAGS(sk),
474 sk->sk_bound_dev_if);
478 sk_setup_caps(sk, &rt->dst);
480 skb_dst_set(skb, &rt->dst);
485 /* We dont need to clone dst here, it is guaranteed to not disappear.
486 * __dev_xmit_skb() might force a refcount if needed.
488 skb_dst_set_noref(skb, &rt->dst);
491 /* Queue the packet to IP for output */
492 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
505 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
511 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
512 size_t len, int noblock, int flags, int *addr_len)
514 struct inet_sock *inet = inet_sk(sk);
516 int err = -EOPNOTSUPP;
517 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
523 skb = skb_recv_datagram(sk, flags, noblock, &err);
529 msg->msg_flags |= MSG_TRUNC;
533 err = skb_copy_datagram_msg(skb, 0, msg, copied);
537 sock_recv_timestamp(msg, sk, skb);
539 /* Copy the address. */
541 sin->sin_family = AF_INET;
542 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
544 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
545 *addr_len = sizeof(*sin);
547 if (inet->cmsg_flags)
548 ip_cmsg_recv(msg, skb);
549 if (flags & MSG_TRUNC)
552 skb_free_datagram(sk, skb);
554 return err ? err : copied;
557 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
564 amount = sk_wmem_alloc_get(sk);
567 spin_lock_bh(&sk->sk_receive_queue.lock);
568 skb = skb_peek(&sk->sk_receive_queue);
569 amount = skb ? skb->len : 0;
570 spin_unlock_bh(&sk->sk_receive_queue.lock);
577 return put_user(amount, (int __user *)arg);
579 EXPORT_SYMBOL(l2tp_ioctl);
581 static struct proto l2tp_ip_prot = {
583 .owner = THIS_MODULE,
584 .init = l2tp_ip_open,
585 .close = l2tp_ip_close,
586 .bind = l2tp_ip_bind,
587 .connect = l2tp_ip_connect,
588 .disconnect = l2tp_ip_disconnect,
590 .destroy = l2tp_ip_destroy_sock,
591 .setsockopt = ip_setsockopt,
592 .getsockopt = ip_getsockopt,
593 .sendmsg = l2tp_ip_sendmsg,
594 .recvmsg = l2tp_ip_recvmsg,
595 .backlog_rcv = l2tp_ip_backlog_recv,
597 .unhash = inet_unhash,
598 .obj_size = sizeof(struct l2tp_ip_sock),
600 .compat_setsockopt = compat_ip_setsockopt,
601 .compat_getsockopt = compat_ip_getsockopt,
605 static const struct proto_ops l2tp_ip_ops = {
607 .owner = THIS_MODULE,
608 .release = inet_release,
610 .connect = inet_dgram_connect,
611 .socketpair = sock_no_socketpair,
612 .accept = sock_no_accept,
613 .getname = l2tp_ip_getname,
614 .poll = datagram_poll,
616 .listen = sock_no_listen,
617 .shutdown = inet_shutdown,
618 .setsockopt = sock_common_setsockopt,
619 .getsockopt = sock_common_getsockopt,
620 .sendmsg = inet_sendmsg,
621 .recvmsg = sock_common_recvmsg,
622 .mmap = sock_no_mmap,
623 .sendpage = sock_no_sendpage,
625 .compat_setsockopt = compat_sock_common_setsockopt,
626 .compat_getsockopt = compat_sock_common_getsockopt,
630 static struct inet_protosw l2tp_ip_protosw = {
632 .protocol = IPPROTO_L2TP,
633 .prot = &l2tp_ip_prot,
637 static struct net_protocol l2tp_ip_protocol __read_mostly = {
638 .handler = l2tp_ip_recv,
642 static int __init l2tp_ip_init(void)
646 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
648 err = proto_register(&l2tp_ip_prot, 1);
652 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
656 inet_register_protosw(&l2tp_ip_protosw);
660 proto_unregister(&l2tp_ip_prot);
665 static void __exit l2tp_ip_exit(void)
667 inet_unregister_protosw(&l2tp_ip_protosw);
668 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
669 proto_unregister(&l2tp_ip_prot);
672 module_init(l2tp_ip_init);
673 module_exit(l2tp_ip_exit);
675 MODULE_LICENSE("GPL");
676 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
677 MODULE_DESCRIPTION("L2TP over IP");
678 MODULE_VERSION("1.0");
680 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
683 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
684 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);