1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * L2TPv3 IP encapsulation support
5 * Copyright (c) 2008,2009,2010 Katalix Systems Ltd
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <asm/ioctls.h>
11 #include <linux/icmp.h>
12 #include <linux/module.h>
13 #include <linux/skbuff.h>
14 #include <linux/random.h>
15 #include <linux/socket.h>
16 #include <linux/l2tp.h>
22 #include <net/inet_common.h>
23 #include <net/inet_hashtables.h>
24 #include <net/tcp_states.h>
25 #include <net/protocol.h>
28 #include "l2tp_core.h"
31 /* inet_sock has to be the first member of l2tp_ip_sock */
32 struct inet_sock inet;
38 static DEFINE_RWLOCK(l2tp_ip_lock);
39 static struct hlist_head l2tp_ip_table;
40 static struct hlist_head l2tp_ip_bind_table;
42 static inline struct l2tp_ip_sock *l2tp_ip_sk(const struct sock *sk)
44 return (struct l2tp_ip_sock *)sk;
47 static struct sock *__l2tp_ip_bind_lookup(const struct net *net, __be32 laddr,
48 __be32 raddr, int dif, u32 tunnel_id)
52 sk_for_each_bound(sk, &l2tp_ip_bind_table) {
53 const struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
54 const struct inet_sock *inet = inet_sk(sk);
56 if (!net_eq(sock_net(sk), net))
59 if (sk->sk_bound_dev_if && dif && sk->sk_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;
125 if (!pskb_may_pull(skb, 4))
128 /* Point to L2TP header */
129 optr = ptr = skb->data;
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 /* Trace packet contents, if enabled */
152 if (tunnel->debug & L2TP_MSG_DATA) {
153 length = min(32u, skb->len);
154 if (!pskb_may_pull(skb, length))
157 /* Point to L2TP header */
158 optr = ptr = skb->data;
160 pr_debug("%s: ip recv\n", tunnel->name);
161 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
164 if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
167 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
168 l2tp_session_dec_refcount(session);
173 /* Get the tunnel_id from the L2TP header */
174 if (!pskb_may_pull(skb, 12))
177 if ((skb->data[0] & 0xc0) != 0xc0)
180 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
181 iph = (struct iphdr *)skb_network_header(skb);
183 read_lock_bh(&l2tp_ip_lock);
184 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr, inet_iif(skb),
187 read_unlock_bh(&l2tp_ip_lock);
191 read_unlock_bh(&l2tp_ip_lock);
193 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
198 return sk_receive_skb(sk, skb, 1);
201 l2tp_session_dec_refcount(session);
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 = sk->sk_user_data;
238 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
242 l2tp_tunnel_delete(tunnel);
245 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
247 struct inet_sock *inet = inet_sk(sk);
248 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
249 struct net *net = sock_net(sk);
253 if (addr_len < sizeof(struct sockaddr_l2tpip))
255 if (addr->l2tp_family != AF_INET)
261 if (!sock_flag(sk, SOCK_ZAPPED))
264 if (sk->sk_state != TCP_CLOSE)
267 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
268 ret = -EADDRNOTAVAIL;
269 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
270 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
273 if (addr->l2tp_addr.s_addr)
274 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
275 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
276 inet->inet_saddr = 0; /* Use device */
278 write_lock_bh(&l2tp_ip_lock);
279 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
280 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
281 write_unlock_bh(&l2tp_ip_lock);
287 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
289 sk_add_bind_node(sk, &l2tp_ip_bind_table);
290 sk_del_node_init(sk);
291 write_unlock_bh(&l2tp_ip_lock);
294 sock_reset_flag(sk, SOCK_ZAPPED);
302 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
304 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
307 if (addr_len < sizeof(*lsa))
310 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
315 /* Must bind first - autobinding does not work */
316 if (sock_flag(sk, SOCK_ZAPPED)) {
321 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
325 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
327 write_lock_bh(&l2tp_ip_lock);
328 hlist_del_init(&sk->sk_bind_node);
329 sk_add_bind_node(sk, &l2tp_ip_bind_table);
330 write_unlock_bh(&l2tp_ip_lock);
338 static int l2tp_ip_disconnect(struct sock *sk, int flags)
340 if (sock_flag(sk, SOCK_ZAPPED))
343 return __udp_disconnect(sk, flags);
346 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
349 struct sock *sk = sock->sk;
350 struct inet_sock *inet = inet_sk(sk);
351 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
352 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
354 memset(lsa, 0, sizeof(*lsa));
355 lsa->l2tp_family = AF_INET;
357 if (!inet->inet_dport)
359 lsa->l2tp_conn_id = lsk->peer_conn_id;
360 lsa->l2tp_addr.s_addr = inet->inet_daddr;
362 __be32 addr = inet->inet_rcv_saddr;
364 addr = inet->inet_saddr;
365 lsa->l2tp_conn_id = lsk->conn_id;
366 lsa->l2tp_addr.s_addr = addr;
371 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
375 /* Charge it to the socket, dropping if the queue is full. */
376 rc = sock_queue_rcv_skb(sk, skb);
383 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
388 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
391 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
395 struct inet_sock *inet = inet_sk(sk);
396 struct rtable *rt = NULL;
404 if (sock_flag(sk, SOCK_DEAD))
407 /* Get and verify the address. */
409 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
411 if (msg->msg_namelen < sizeof(*lip))
414 if (lip->l2tp_family != AF_INET) {
416 if (lip->l2tp_family != AF_UNSPEC)
420 daddr = lip->l2tp_addr.s_addr;
423 if (sk->sk_state != TCP_ESTABLISHED)
426 daddr = inet->inet_daddr;
430 /* Allocate a socket buffer */
432 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
433 4 + len, 0, GFP_KERNEL);
437 /* Reserve space for headers, putting IP header on 4-byte boundary. */
438 skb_reserve(skb, 2 + NET_SKB_PAD);
439 skb_reset_network_header(skb);
440 skb_reserve(skb, sizeof(struct iphdr));
441 skb_reset_transport_header(skb);
443 /* Insert 0 session_id */
444 *((__be32 *) skb_put(skb, 4)) = 0;
446 /* Copy user data into skb */
447 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
453 fl4 = &inet->cork.fl.u.ip4;
455 rt = (struct rtable *) __sk_dst_check(sk, 0);
459 const struct ip_options_rcu *inet_opt;
461 inet_opt = rcu_dereference(inet->inet_opt);
463 /* Use correct destination address if we have options. */
464 if (inet_opt && inet_opt->opt.srr)
465 daddr = inet_opt->opt.faddr;
467 /* If this fails, retransmit mechanism of transport layer will
468 * keep trying until route appears or the connection times
471 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
472 daddr, inet->inet_saddr,
473 inet->inet_dport, inet->inet_sport,
474 sk->sk_protocol, RT_CONN_FLAGS(sk),
475 sk->sk_bound_dev_if);
479 sk_setup_caps(sk, &rt->dst);
481 skb_dst_set(skb, &rt->dst);
486 /* We dont need to clone dst here, it is guaranteed to not disappear.
487 * __dev_xmit_skb() might force a refcount if needed.
489 skb_dst_set_noref(skb, &rt->dst);
492 /* Queue the packet to IP for output */
493 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
506 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
512 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
513 size_t len, int noblock, int flags, int *addr_len)
515 struct inet_sock *inet = inet_sk(sk);
517 int err = -EOPNOTSUPP;
518 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
524 skb = skb_recv_datagram(sk, flags, noblock, &err);
530 msg->msg_flags |= MSG_TRUNC;
534 err = skb_copy_datagram_msg(skb, 0, msg, copied);
538 sock_recv_timestamp(msg, sk, skb);
540 /* Copy the address. */
542 sin->sin_family = AF_INET;
543 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
545 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
546 *addr_len = sizeof(*sin);
548 if (inet->cmsg_flags)
549 ip_cmsg_recv(msg, skb);
550 if (flags & MSG_TRUNC)
553 skb_free_datagram(sk, skb);
555 return err ? err : copied;
558 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
565 amount = sk_wmem_alloc_get(sk);
568 spin_lock_bh(&sk->sk_receive_queue.lock);
569 skb = skb_peek(&sk->sk_receive_queue);
570 amount = skb ? skb->len : 0;
571 spin_unlock_bh(&sk->sk_receive_queue.lock);
578 return put_user(amount, (int __user *)arg);
580 EXPORT_SYMBOL(l2tp_ioctl);
582 static struct proto l2tp_ip_prot = {
584 .owner = THIS_MODULE,
585 .init = l2tp_ip_open,
586 .close = l2tp_ip_close,
587 .bind = l2tp_ip_bind,
588 .connect = l2tp_ip_connect,
589 .disconnect = l2tp_ip_disconnect,
591 .destroy = l2tp_ip_destroy_sock,
592 .setsockopt = ip_setsockopt,
593 .getsockopt = ip_getsockopt,
594 .sendmsg = l2tp_ip_sendmsg,
595 .recvmsg = l2tp_ip_recvmsg,
596 .backlog_rcv = l2tp_ip_backlog_recv,
598 .unhash = inet_unhash,
599 .obj_size = sizeof(struct l2tp_ip_sock),
601 .compat_setsockopt = compat_ip_setsockopt,
602 .compat_getsockopt = compat_ip_getsockopt,
606 static const struct proto_ops l2tp_ip_ops = {
608 .owner = THIS_MODULE,
609 .release = inet_release,
611 .connect = inet_dgram_connect,
612 .socketpair = sock_no_socketpair,
613 .accept = sock_no_accept,
614 .getname = l2tp_ip_getname,
615 .poll = datagram_poll,
617 .gettstamp = sock_gettstamp,
618 .listen = sock_no_listen,
619 .shutdown = inet_shutdown,
620 .setsockopt = sock_common_setsockopt,
621 .getsockopt = sock_common_getsockopt,
622 .sendmsg = inet_sendmsg,
623 .recvmsg = sock_common_recvmsg,
624 .mmap = sock_no_mmap,
625 .sendpage = sock_no_sendpage,
627 .compat_setsockopt = compat_sock_common_setsockopt,
628 .compat_getsockopt = compat_sock_common_getsockopt,
632 static struct inet_protosw l2tp_ip_protosw = {
634 .protocol = IPPROTO_L2TP,
635 .prot = &l2tp_ip_prot,
639 static struct net_protocol l2tp_ip_protocol __read_mostly = {
640 .handler = l2tp_ip_recv,
644 static int __init l2tp_ip_init(void)
648 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
650 err = proto_register(&l2tp_ip_prot, 1);
654 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
658 inet_register_protosw(&l2tp_ip_protosw);
662 proto_unregister(&l2tp_ip_prot);
667 static void __exit l2tp_ip_exit(void)
669 inet_unregister_protosw(&l2tp_ip_protosw);
670 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
671 proto_unregister(&l2tp_ip_prot);
674 module_init(l2tp_ip_init);
675 module_exit(l2tp_ip_exit);
677 MODULE_LICENSE("GPL");
679 MODULE_DESCRIPTION("L2TP over IP");
680 MODULE_VERSION("1.0");
682 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
685 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
686 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);