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 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 (!inet->inet_daddr || !raddr || inet->inet_daddr == raddr) &&
66 (!sk->sk_bound_dev_if || !dif ||
67 sk->sk_bound_dev_if == dif))
76 /* When processing receive frames, there are two cases to
77 * consider. Data frames consist of a non-zero session-id and an
78 * optional cookie. Control frames consist of a regular L2TP header
79 * preceded by 32-bits of zeros.
81 * L2TPv3 Session Header Over IP
84 * 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
85 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
87 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
88 * | Cookie (optional, maximum 64 bits)...
89 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
91 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
93 * L2TPv3 Control Message Header Over IP
96 * 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
97 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
98 * | (32 bits of zeros) |
99 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
101 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
102 * | Control Connection ID |
103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107 * All control frames are passed to userspace.
109 static int l2tp_ip_recv(struct sk_buff *skb)
111 struct net *net = dev_net(skb->dev);
115 unsigned char *ptr, *optr;
116 struct l2tp_session *session;
117 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_get(net, NULL, session_id, true);
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 if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
163 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
164 l2tp_session_dec_refcount(session);
169 /* Get the tunnel_id from the L2TP header */
170 if (!pskb_may_pull(skb, 12))
173 if ((skb->data[0] & 0xc0) != 0xc0)
176 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
177 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, inet_iif(skb),
183 read_unlock_bh(&l2tp_ip_lock);
187 read_unlock_bh(&l2tp_ip_lock);
189 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
194 return sk_receive_skb(sk, skb, 1);
198 session->deref(session);
199 l2tp_session_dec_refcount(session);
210 static int l2tp_ip_hash(struct sock *sk)
212 if (sk_unhashed(sk)) {
213 write_lock_bh(&l2tp_ip_lock);
214 sk_add_node(sk, &l2tp_ip_table);
215 write_unlock_bh(&l2tp_ip_lock);
220 static void l2tp_ip_unhash(struct sock *sk)
224 write_lock_bh(&l2tp_ip_lock);
225 sk_del_node_init(sk);
226 write_unlock_bh(&l2tp_ip_lock);
229 static int l2tp_ip_open(struct sock *sk)
231 /* Prevent autobind. We don't have ports. */
232 inet_sk(sk)->inet_num = IPPROTO_L2TP;
238 static void l2tp_ip_close(struct sock *sk, long timeout)
240 write_lock_bh(&l2tp_ip_lock);
241 hlist_del_init(&sk->sk_bind_node);
242 sk_del_node_init(sk);
243 write_unlock_bh(&l2tp_ip_lock);
244 sk_common_release(sk);
247 static void l2tp_ip_destroy_sock(struct sock *sk)
250 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
252 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
256 l2tp_tunnel_closeall(tunnel);
260 sk_refcnt_debug_dec(sk);
263 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
265 struct inet_sock *inet = inet_sk(sk);
266 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
267 struct net *net = sock_net(sk);
271 if (addr_len < sizeof(struct sockaddr_l2tpip))
273 if (addr->l2tp_family != AF_INET)
279 if (!sock_flag(sk, SOCK_ZAPPED))
282 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
285 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
286 ret = -EADDRNOTAVAIL;
287 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
288 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
291 if (addr->l2tp_addr.s_addr)
292 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
293 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
294 inet->inet_saddr = 0; /* Use device */
296 write_lock_bh(&l2tp_ip_lock);
297 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
298 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
299 write_unlock_bh(&l2tp_ip_lock);
305 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
307 sk_add_bind_node(sk, &l2tp_ip_bind_table);
308 sk_del_node_init(sk);
309 write_unlock_bh(&l2tp_ip_lock);
312 sock_reset_flag(sk, SOCK_ZAPPED);
320 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
322 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
325 if (addr_len < sizeof(*lsa))
328 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
333 /* Must bind first - autobinding does not work */
334 if (sock_flag(sk, SOCK_ZAPPED)) {
339 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
343 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
345 write_lock_bh(&l2tp_ip_lock);
346 hlist_del_init(&sk->sk_bind_node);
347 sk_add_bind_node(sk, &l2tp_ip_bind_table);
348 write_unlock_bh(&l2tp_ip_lock);
356 static int l2tp_ip_disconnect(struct sock *sk, int flags)
358 if (sock_flag(sk, SOCK_ZAPPED))
361 return __udp_disconnect(sk, flags);
364 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
365 int *uaddr_len, int peer)
367 struct sock *sk = sock->sk;
368 struct inet_sock *inet = inet_sk(sk);
369 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
370 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
372 memset(lsa, 0, sizeof(*lsa));
373 lsa->l2tp_family = AF_INET;
375 if (!inet->inet_dport)
377 lsa->l2tp_conn_id = lsk->peer_conn_id;
378 lsa->l2tp_addr.s_addr = inet->inet_daddr;
380 __be32 addr = inet->inet_rcv_saddr;
382 addr = inet->inet_saddr;
383 lsa->l2tp_conn_id = lsk->conn_id;
384 lsa->l2tp_addr.s_addr = addr;
386 *uaddr_len = sizeof(*lsa);
390 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
394 /* Charge it to the socket, dropping if the queue is full. */
395 rc = sock_queue_rcv_skb(sk, skb);
402 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
407 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
410 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
414 struct inet_sock *inet = inet_sk(sk);
415 struct rtable *rt = NULL;
423 if (sock_flag(sk, SOCK_DEAD))
426 /* Get and verify the address. */
428 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
430 if (msg->msg_namelen < sizeof(*lip))
433 if (lip->l2tp_family != AF_INET) {
435 if (lip->l2tp_family != AF_UNSPEC)
439 daddr = lip->l2tp_addr.s_addr;
442 if (sk->sk_state != TCP_ESTABLISHED)
445 daddr = inet->inet_daddr;
449 /* Allocate a socket buffer */
451 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
452 4 + len, 0, GFP_KERNEL);
456 /* Reserve space for headers, putting IP header on 4-byte boundary. */
457 skb_reserve(skb, 2 + NET_SKB_PAD);
458 skb_reset_network_header(skb);
459 skb_reserve(skb, sizeof(struct iphdr));
460 skb_reset_transport_header(skb);
462 /* Insert 0 session_id */
463 *((__be32 *) skb_put(skb, 4)) = 0;
465 /* Copy user data into skb */
466 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
472 fl4 = &inet->cork.fl.u.ip4;
474 rt = (struct rtable *) __sk_dst_check(sk, 0);
478 const struct ip_options_rcu *inet_opt;
480 inet_opt = rcu_dereference(inet->inet_opt);
482 /* Use correct destination address if we have options. */
483 if (inet_opt && inet_opt->opt.srr)
484 daddr = inet_opt->opt.faddr;
486 /* If this fails, retransmit mechanism of transport layer will
487 * keep trying until route appears or the connection times
490 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
491 daddr, inet->inet_saddr,
492 inet->inet_dport, inet->inet_sport,
493 sk->sk_protocol, RT_CONN_FLAGS(sk),
494 sk->sk_bound_dev_if);
498 sk_setup_caps(sk, &rt->dst);
500 skb_dst_set(skb, &rt->dst);
505 /* We dont need to clone dst here, it is guaranteed to not disappear.
506 * __dev_xmit_skb() might force a refcount if needed.
508 skb_dst_set_noref(skb, &rt->dst);
511 /* Queue the packet to IP for output */
512 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
525 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
531 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
532 size_t len, int noblock, int flags, int *addr_len)
534 struct inet_sock *inet = inet_sk(sk);
536 int err = -EOPNOTSUPP;
537 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
543 skb = skb_recv_datagram(sk, flags, noblock, &err);
549 msg->msg_flags |= MSG_TRUNC;
553 err = skb_copy_datagram_msg(skb, 0, msg, copied);
557 sock_recv_timestamp(msg, sk, skb);
559 /* Copy the address. */
561 sin->sin_family = AF_INET;
562 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
564 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
565 *addr_len = sizeof(*sin);
567 if (inet->cmsg_flags)
568 ip_cmsg_recv(msg, skb);
569 if (flags & MSG_TRUNC)
572 skb_free_datagram(sk, skb);
574 return err ? err : copied;
577 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
584 amount = sk_wmem_alloc_get(sk);
587 spin_lock_bh(&sk->sk_receive_queue.lock);
588 skb = skb_peek(&sk->sk_receive_queue);
589 amount = skb ? skb->len : 0;
590 spin_unlock_bh(&sk->sk_receive_queue.lock);
597 return put_user(amount, (int __user *)arg);
599 EXPORT_SYMBOL(l2tp_ioctl);
601 static struct proto l2tp_ip_prot = {
603 .owner = THIS_MODULE,
604 .init = l2tp_ip_open,
605 .close = l2tp_ip_close,
606 .bind = l2tp_ip_bind,
607 .connect = l2tp_ip_connect,
608 .disconnect = l2tp_ip_disconnect,
610 .destroy = l2tp_ip_destroy_sock,
611 .setsockopt = ip_setsockopt,
612 .getsockopt = ip_getsockopt,
613 .sendmsg = l2tp_ip_sendmsg,
614 .recvmsg = l2tp_ip_recvmsg,
615 .backlog_rcv = l2tp_ip_backlog_recv,
616 .hash = l2tp_ip_hash,
617 .unhash = l2tp_ip_unhash,
618 .obj_size = sizeof(struct l2tp_ip_sock),
620 .compat_setsockopt = compat_ip_setsockopt,
621 .compat_getsockopt = compat_ip_getsockopt,
625 static const struct proto_ops l2tp_ip_ops = {
627 .owner = THIS_MODULE,
628 .release = inet_release,
630 .connect = inet_dgram_connect,
631 .socketpair = sock_no_socketpair,
632 .accept = sock_no_accept,
633 .getname = l2tp_ip_getname,
634 .poll = datagram_poll,
636 .listen = sock_no_listen,
637 .shutdown = inet_shutdown,
638 .setsockopt = sock_common_setsockopt,
639 .getsockopt = sock_common_getsockopt,
640 .sendmsg = inet_sendmsg,
641 .recvmsg = sock_common_recvmsg,
642 .mmap = sock_no_mmap,
643 .sendpage = sock_no_sendpage,
645 .compat_setsockopt = compat_sock_common_setsockopt,
646 .compat_getsockopt = compat_sock_common_getsockopt,
650 static struct inet_protosw l2tp_ip_protosw = {
652 .protocol = IPPROTO_L2TP,
653 .prot = &l2tp_ip_prot,
657 static struct net_protocol l2tp_ip_protocol __read_mostly = {
658 .handler = l2tp_ip_recv,
662 static int __init l2tp_ip_init(void)
666 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
668 err = proto_register(&l2tp_ip_prot, 1);
672 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
676 inet_register_protosw(&l2tp_ip_protosw);
680 proto_unregister(&l2tp_ip_prot);
685 static void __exit l2tp_ip_exit(void)
687 inet_unregister_protosw(&l2tp_ip_protosw);
688 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
689 proto_unregister(&l2tp_ip_prot);
692 module_init(l2tp_ip_init);
693 module_exit(l2tp_ip_exit);
695 MODULE_LICENSE("GPL");
696 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
697 MODULE_DESCRIPTION("L2TP over IP");
698 MODULE_VERSION("1.0");
700 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
703 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
704 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);