1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* L2TPv3 IP encapsulation support
4 * Copyright (c) 2008,2009,2010 Katalix Systems Ltd
7 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 #include <asm/ioctls.h>
10 #include <linux/icmp.h>
11 #include <linux/module.h>
12 #include <linux/skbuff.h>
13 #include <linux/random.h>
14 #include <linux/socket.h>
15 #include <linux/l2tp.h>
21 #include <net/inet_common.h>
22 #include <net/tcp_states.h>
23 #include <net/protocol.h>
26 #include "l2tp_core.h"
29 /* inet_sock has to be the first member of l2tp_ip_sock */
30 struct inet_sock inet;
36 static DEFINE_RWLOCK(l2tp_ip_lock);
37 static struct hlist_head l2tp_ip_table;
38 static struct hlist_head l2tp_ip_bind_table;
40 static inline struct l2tp_ip_sock *l2tp_ip_sk(const struct sock *sk)
42 return (struct l2tp_ip_sock *)sk;
45 static struct sock *__l2tp_ip_bind_lookup(const struct net *net, __be32 laddr,
46 __be32 raddr, int dif, u32 tunnel_id)
50 sk_for_each_bound(sk, &l2tp_ip_bind_table) {
51 const struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
52 const struct inet_sock *inet = inet_sk(sk);
55 if (!net_eq(sock_net(sk), net))
58 bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
59 if (bound_dev_if && dif && bound_dev_if != dif)
62 if (inet->inet_rcv_saddr && laddr &&
63 inet->inet_rcv_saddr != laddr)
66 if (inet->inet_daddr && raddr && inet->inet_daddr != raddr)
69 if (l2tp->conn_id != tunnel_id)
80 /* When processing receive frames, there are two cases to
81 * consider. Data frames consist of a non-zero session-id and an
82 * optional cookie. Control frames consist of a regular L2TP header
83 * preceded by 32-bits of zeros.
85 * L2TPv3 Session Header Over IP
88 * 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
89 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
91 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
92 * | Cookie (optional, maximum 64 bits)...
93 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
95 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
97 * L2TPv3 Control Message Header Over IP
100 * 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
101 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
102 * | (32 bits of zeros) |
103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
104 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * | Control Connection ID |
107 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111 * All control frames are passed to userspace.
113 static int l2tp_ip_recv(struct sk_buff *skb)
115 struct net *net = dev_net(skb->dev);
119 unsigned char *ptr, *optr;
120 struct l2tp_session *session;
121 struct l2tp_tunnel *tunnel = NULL;
124 if (!pskb_may_pull(skb, 4))
127 /* Point to L2TP header */
130 session_id = ntohl(*((__be32 *)ptr));
133 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
134 * the session_id. If it is 0, the packet is a L2TP control
135 * frame and the session_id value can be discarded.
137 if (session_id == 0) {
142 /* Ok, this is a data packet. Lookup the session. */
143 session = l2tp_session_get(net, session_id);
147 tunnel = session->tunnel;
151 if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
154 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
155 l2tp_session_dec_refcount(session);
160 /* Get the tunnel_id from the L2TP header */
161 if (!pskb_may_pull(skb, 12))
164 if ((skb->data[0] & 0xc0) != 0xc0)
167 tunnel_id = ntohl(*(__be32 *)&skb->data[4]);
168 iph = (struct iphdr *)skb_network_header(skb);
170 read_lock_bh(&l2tp_ip_lock);
171 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr, inet_iif(skb),
174 read_unlock_bh(&l2tp_ip_lock);
178 read_unlock_bh(&l2tp_ip_lock);
180 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
185 return sk_receive_skb(sk, skb, 1);
188 l2tp_session_dec_refcount(session);
199 static int l2tp_ip_hash(struct sock *sk)
201 if (sk_unhashed(sk)) {
202 write_lock_bh(&l2tp_ip_lock);
203 sk_add_node(sk, &l2tp_ip_table);
204 write_unlock_bh(&l2tp_ip_lock);
209 static void l2tp_ip_unhash(struct sock *sk)
213 write_lock_bh(&l2tp_ip_lock);
214 sk_del_node_init(sk);
215 write_unlock_bh(&l2tp_ip_lock);
218 static int l2tp_ip_open(struct sock *sk)
220 /* Prevent autobind. We don't have ports. */
221 inet_sk(sk)->inet_num = IPPROTO_L2TP;
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)
238 struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk);
241 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
245 l2tp_tunnel_delete(tunnel);
248 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
250 struct inet_sock *inet = inet_sk(sk);
251 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *)uaddr;
252 struct net *net = sock_net(sk);
256 if (addr_len < sizeof(struct sockaddr_l2tpip))
258 if (addr->l2tp_family != AF_INET)
264 if (!sock_flag(sk, SOCK_ZAPPED))
267 if (sk->sk_state != TCP_CLOSE)
270 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
271 ret = -EADDRNOTAVAIL;
272 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
273 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
276 if (addr->l2tp_addr.s_addr) {
277 inet->inet_rcv_saddr = addr->l2tp_addr.s_addr;
278 inet->inet_saddr = addr->l2tp_addr.s_addr;
280 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
281 inet->inet_saddr = 0; /* Use device */
283 write_lock_bh(&l2tp_ip_lock);
284 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
285 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
286 write_unlock_bh(&l2tp_ip_lock);
292 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
294 sk_add_bind_node(sk, &l2tp_ip_bind_table);
295 sk_del_node_init(sk);
296 write_unlock_bh(&l2tp_ip_lock);
299 sock_reset_flag(sk, SOCK_ZAPPED);
307 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
309 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
312 if (addr_len < sizeof(*lsa))
315 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
320 /* Must bind first - autobinding does not work */
321 if (sock_flag(sk, SOCK_ZAPPED)) {
326 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
330 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
332 write_lock_bh(&l2tp_ip_lock);
333 hlist_del_init(&sk->sk_bind_node);
334 sk_add_bind_node(sk, &l2tp_ip_bind_table);
335 write_unlock_bh(&l2tp_ip_lock);
343 static int l2tp_ip_disconnect(struct sock *sk, int flags)
345 if (sock_flag(sk, SOCK_ZAPPED))
348 return __udp_disconnect(sk, flags);
351 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
354 struct sock *sk = sock->sk;
355 struct inet_sock *inet = inet_sk(sk);
356 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
357 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
359 memset(lsa, 0, sizeof(*lsa));
360 lsa->l2tp_family = AF_INET;
362 if (!inet->inet_dport)
364 lsa->l2tp_conn_id = lsk->peer_conn_id;
365 lsa->l2tp_addr.s_addr = inet->inet_daddr;
367 __be32 addr = inet->inet_rcv_saddr;
370 addr = inet->inet_saddr;
371 lsa->l2tp_conn_id = lsk->conn_id;
372 lsa->l2tp_addr.s_addr = addr;
377 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
381 /* Charge it to the socket, dropping if the queue is full. */
382 rc = sock_queue_rcv_skb(sk, skb);
389 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
394 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
397 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
401 struct inet_sock *inet = inet_sk(sk);
402 struct rtable *rt = NULL;
410 if (sock_flag(sk, SOCK_DEAD))
413 /* Get and verify the address. */
415 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
418 if (msg->msg_namelen < sizeof(*lip))
421 if (lip->l2tp_family != AF_INET) {
423 if (lip->l2tp_family != AF_UNSPEC)
427 daddr = lip->l2tp_addr.s_addr;
430 if (sk->sk_state != TCP_ESTABLISHED)
433 daddr = inet->inet_daddr;
437 /* Allocate a socket buffer */
439 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
440 4 + len, 0, GFP_KERNEL);
444 /* Reserve space for headers, putting IP header on 4-byte boundary. */
445 skb_reserve(skb, 2 + NET_SKB_PAD);
446 skb_reset_network_header(skb);
447 skb_reserve(skb, sizeof(struct iphdr));
448 skb_reset_transport_header(skb);
450 /* Insert 0 session_id */
451 *((__be32 *)skb_put(skb, 4)) = 0;
453 /* Copy user data into skb */
454 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
460 fl4 = &inet->cork.fl.u.ip4;
462 rt = (struct rtable *)__sk_dst_check(sk, 0);
466 const struct ip_options_rcu *inet_opt;
468 inet_opt = rcu_dereference(inet->inet_opt);
470 /* Use correct destination address if we have options. */
471 if (inet_opt && inet_opt->opt.srr)
472 daddr = inet_opt->opt.faddr;
474 /* If this fails, retransmit mechanism of transport layer will
475 * keep trying until route appears or the connection times
478 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
479 daddr, inet->inet_saddr,
480 inet->inet_dport, inet->inet_sport,
481 sk->sk_protocol, RT_CONN_FLAGS(sk),
482 sk->sk_bound_dev_if);
486 sk_setup_caps(sk, &rt->dst);
488 skb_dst_set(skb, &rt->dst);
493 /* We don't need to clone dst here, it is guaranteed to not disappear.
494 * __dev_xmit_skb() might force a refcount if needed.
496 skb_dst_set_noref(skb, &rt->dst);
499 /* Queue the packet to IP for output */
500 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
513 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
519 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
520 size_t len, int flags, int *addr_len)
522 struct inet_sock *inet = inet_sk(sk);
524 int err = -EOPNOTSUPP;
525 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
531 skb = skb_recv_datagram(sk, flags, &err);
537 msg->msg_flags |= MSG_TRUNC;
541 err = skb_copy_datagram_msg(skb, 0, msg, copied);
545 sock_recv_timestamp(msg, sk, skb);
547 /* Copy the address. */
549 sin->sin_family = AF_INET;
550 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
552 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
553 *addr_len = sizeof(*sin);
555 if (inet_cmsg_flags(inet))
556 ip_cmsg_recv(msg, skb);
557 if (flags & MSG_TRUNC)
560 skb_free_datagram(sk, skb);
562 return err ? err : copied;
565 int l2tp_ioctl(struct sock *sk, int cmd, int *karg)
571 *karg = sk_wmem_alloc_get(sk);
574 spin_lock_bh(&sk->sk_receive_queue.lock);
575 skb = skb_peek(&sk->sk_receive_queue);
576 *karg = skb ? skb->len : 0;
577 spin_unlock_bh(&sk->sk_receive_queue.lock);
586 EXPORT_SYMBOL_GPL(l2tp_ioctl);
588 static struct proto l2tp_ip_prot = {
590 .owner = THIS_MODULE,
591 .init = l2tp_ip_open,
592 .close = l2tp_ip_close,
593 .bind = l2tp_ip_bind,
594 .connect = l2tp_ip_connect,
595 .disconnect = l2tp_ip_disconnect,
597 .destroy = l2tp_ip_destroy_sock,
598 .setsockopt = ip_setsockopt,
599 .getsockopt = ip_getsockopt,
600 .sendmsg = l2tp_ip_sendmsg,
601 .recvmsg = l2tp_ip_recvmsg,
602 .backlog_rcv = l2tp_ip_backlog_recv,
603 .hash = l2tp_ip_hash,
604 .unhash = l2tp_ip_unhash,
605 .obj_size = sizeof(struct l2tp_ip_sock),
608 static const struct proto_ops l2tp_ip_ops = {
610 .owner = THIS_MODULE,
611 .release = inet_release,
613 .connect = inet_dgram_connect,
614 .socketpair = sock_no_socketpair,
615 .accept = sock_no_accept,
616 .getname = l2tp_ip_getname,
617 .poll = datagram_poll,
619 .gettstamp = sock_gettstamp,
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,
629 static struct inet_protosw l2tp_ip_protosw = {
631 .protocol = IPPROTO_L2TP,
632 .prot = &l2tp_ip_prot,
636 static struct net_protocol l2tp_ip_protocol __read_mostly = {
637 .handler = l2tp_ip_recv,
640 static int __init l2tp_ip_init(void)
644 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
646 err = proto_register(&l2tp_ip_prot, 1);
650 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
654 inet_register_protosw(&l2tp_ip_protosw);
658 proto_unregister(&l2tp_ip_prot);
663 static void __exit l2tp_ip_exit(void)
665 inet_unregister_protosw(&l2tp_ip_protosw);
666 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
667 proto_unregister(&l2tp_ip_prot);
670 module_init(l2tp_ip_init);
671 module_exit(l2tp_ip_exit);
673 MODULE_LICENSE("GPL");
674 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
675 MODULE_DESCRIPTION("L2TP over IP");
676 MODULE_VERSION("1.0");
678 /* Use the values of SOCK_DGRAM (2) as type and IPPROTO_L2TP (115) as protocol,
679 * because __stringify doesn't like enums
681 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 115, 2);
682 MODULE_ALIAS_NET_PF_PROTO(PF_INET, 115);