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);
54 if (!net_eq(sock_net(sk), net))
57 if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
60 if (inet->inet_rcv_saddr && laddr &&
61 inet->inet_rcv_saddr != laddr)
64 if (inet->inet_daddr && raddr && inet->inet_daddr != raddr)
67 if (l2tp->conn_id != tunnel_id)
78 /* When processing receive frames, there are two cases to
79 * consider. Data frames consist of a non-zero session-id and an
80 * optional cookie. Control frames consist of a regular L2TP header
81 * preceded by 32-bits of zeros.
83 * L2TPv3 Session Header Over IP
86 * 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
87 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
89 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
90 * | Cookie (optional, maximum 64 bits)...
91 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
93 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
95 * L2TPv3 Control Message Header Over IP
98 * 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
99 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100 * | (32 bits of zeros) |
101 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
102 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
104 * | Control Connection ID |
105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109 * All control frames are passed to userspace.
111 static int l2tp_ip_recv(struct sk_buff *skb)
113 struct net *net = dev_net(skb->dev);
117 unsigned char *ptr, *optr;
118 struct l2tp_session *session;
119 struct l2tp_tunnel *tunnel = NULL;
123 if (!pskb_may_pull(skb, 4))
126 /* Point to L2TP header */
129 session_id = ntohl(*((__be32 *)ptr));
132 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
133 * the session_id. If it is 0, the packet is a L2TP control
134 * frame and the session_id value can be discarded.
136 if (session_id == 0) {
141 /* Ok, this is a data packet. Lookup the session. */
142 session = l2tp_session_get(net, session_id);
146 tunnel = session->tunnel;
150 /* Trace packet contents, if enabled */
151 if (tunnel->debug & L2TP_MSG_DATA) {
152 length = min(32u, skb->len);
153 if (!pskb_may_pull(skb, length))
156 /* Point to L2TP header */
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_hash(struct sock *sk)
214 if (sk_unhashed(sk)) {
215 write_lock_bh(&l2tp_ip_lock);
216 sk_add_node(sk, &l2tp_ip_table);
217 write_unlock_bh(&l2tp_ip_lock);
222 static void l2tp_ip_unhash(struct sock *sk)
226 write_lock_bh(&l2tp_ip_lock);
227 sk_del_node_init(sk);
228 write_unlock_bh(&l2tp_ip_lock);
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;
240 static void l2tp_ip_close(struct sock *sk, long timeout)
242 write_lock_bh(&l2tp_ip_lock);
243 hlist_del_init(&sk->sk_bind_node);
244 sk_del_node_init(sk);
245 write_unlock_bh(&l2tp_ip_lock);
246 sk_common_release(sk);
249 static void l2tp_ip_destroy_sock(struct sock *sk)
252 struct l2tp_tunnel *tunnel = sk->sk_user_data;
254 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
258 l2tp_tunnel_delete(tunnel);
261 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
263 struct inet_sock *inet = inet_sk(sk);
264 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *)uaddr;
265 struct net *net = sock_net(sk);
269 if (addr_len < sizeof(struct sockaddr_l2tpip))
271 if (addr->l2tp_family != AF_INET)
277 if (!sock_flag(sk, SOCK_ZAPPED))
280 if (sk->sk_state != TCP_CLOSE)
283 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
284 ret = -EADDRNOTAVAIL;
285 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
286 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
289 if (addr->l2tp_addr.s_addr) {
290 inet->inet_rcv_saddr = addr->l2tp_addr.s_addr;
291 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,
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;
383 addr = inet->inet_saddr;
384 lsa->l2tp_conn_id = lsk->conn_id;
385 lsa->l2tp_addr.s_addr = addr;
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);
431 if (msg->msg_namelen < sizeof(*lip))
434 if (lip->l2tp_family != AF_INET) {
436 if (lip->l2tp_family != AF_UNSPEC)
440 daddr = lip->l2tp_addr.s_addr;
443 if (sk->sk_state != TCP_ESTABLISHED)
446 daddr = inet->inet_daddr;
450 /* Allocate a socket buffer */
452 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
453 4 + len, 0, GFP_KERNEL);
457 /* Reserve space for headers, putting IP header on 4-byte boundary. */
458 skb_reserve(skb, 2 + NET_SKB_PAD);
459 skb_reset_network_header(skb);
460 skb_reserve(skb, sizeof(struct iphdr));
461 skb_reset_transport_header(skb);
463 /* Insert 0 session_id */
464 *((__be32 *)skb_put(skb, 4)) = 0;
466 /* Copy user data into skb */
467 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
473 fl4 = &inet->cork.fl.u.ip4;
475 rt = (struct rtable *)__sk_dst_check(sk, 0);
479 const struct ip_options_rcu *inet_opt;
481 inet_opt = rcu_dereference(inet->inet_opt);
483 /* Use correct destination address if we have options. */
484 if (inet_opt && inet_opt->opt.srr)
485 daddr = inet_opt->opt.faddr;
487 /* If this fails, retransmit mechanism of transport layer will
488 * keep trying until route appears or the connection times
491 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
492 daddr, inet->inet_saddr,
493 inet->inet_dport, inet->inet_sport,
494 sk->sk_protocol, RT_CONN_FLAGS(sk),
495 sk->sk_bound_dev_if);
499 sk_setup_caps(sk, &rt->dst);
501 skb_dst_set(skb, &rt->dst);
506 /* We dont need to clone dst here, it is guaranteed to not disappear.
507 * __dev_xmit_skb() might force a refcount if needed.
509 skb_dst_set_noref(skb, &rt->dst);
512 /* Queue the packet to IP for output */
513 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
526 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
532 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
533 size_t len, int noblock, int flags, int *addr_len)
535 struct inet_sock *inet = inet_sk(sk);
537 int err = -EOPNOTSUPP;
538 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
544 skb = skb_recv_datagram(sk, flags, noblock, &err);
550 msg->msg_flags |= MSG_TRUNC;
554 err = skb_copy_datagram_msg(skb, 0, msg, copied);
558 sock_recv_timestamp(msg, sk, skb);
560 /* Copy the address. */
562 sin->sin_family = AF_INET;
563 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
565 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
566 *addr_len = sizeof(*sin);
568 if (inet->cmsg_flags)
569 ip_cmsg_recv(msg, skb);
570 if (flags & MSG_TRUNC)
573 skb_free_datagram(sk, skb);
575 return err ? err : copied;
578 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
585 amount = sk_wmem_alloc_get(sk);
588 spin_lock_bh(&sk->sk_receive_queue.lock);
589 skb = skb_peek(&sk->sk_receive_queue);
590 amount = skb ? skb->len : 0;
591 spin_unlock_bh(&sk->sk_receive_queue.lock);
598 return put_user(amount, (int __user *)arg);
600 EXPORT_SYMBOL_GPL(l2tp_ioctl);
602 static struct proto l2tp_ip_prot = {
604 .owner = THIS_MODULE,
605 .init = l2tp_ip_open,
606 .close = l2tp_ip_close,
607 .bind = l2tp_ip_bind,
608 .connect = l2tp_ip_connect,
609 .disconnect = l2tp_ip_disconnect,
611 .destroy = l2tp_ip_destroy_sock,
612 .setsockopt = ip_setsockopt,
613 .getsockopt = ip_getsockopt,
614 .sendmsg = l2tp_ip_sendmsg,
615 .recvmsg = l2tp_ip_recvmsg,
616 .backlog_rcv = l2tp_ip_backlog_recv,
617 .hash = l2tp_ip_hash,
618 .unhash = l2tp_ip_unhash,
619 .obj_size = sizeof(struct l2tp_ip_sock),
622 static const struct proto_ops l2tp_ip_ops = {
624 .owner = THIS_MODULE,
625 .release = inet_release,
627 .connect = inet_dgram_connect,
628 .socketpair = sock_no_socketpair,
629 .accept = sock_no_accept,
630 .getname = l2tp_ip_getname,
631 .poll = datagram_poll,
633 .gettstamp = sock_gettstamp,
634 .listen = sock_no_listen,
635 .shutdown = inet_shutdown,
636 .setsockopt = sock_common_setsockopt,
637 .getsockopt = sock_common_getsockopt,
638 .sendmsg = inet_sendmsg,
639 .recvmsg = sock_common_recvmsg,
640 .mmap = sock_no_mmap,
641 .sendpage = sock_no_sendpage,
644 static struct inet_protosw l2tp_ip_protosw = {
646 .protocol = IPPROTO_L2TP,
647 .prot = &l2tp_ip_prot,
651 static struct net_protocol l2tp_ip_protocol __read_mostly = {
652 .handler = l2tp_ip_recv,
656 static int __init l2tp_ip_init(void)
660 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
662 err = proto_register(&l2tp_ip_prot, 1);
666 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
670 inet_register_protosw(&l2tp_ip_protosw);
674 proto_unregister(&l2tp_ip_prot);
679 static void __exit l2tp_ip_exit(void)
681 inet_unregister_protosw(&l2tp_ip_protosw);
682 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
683 proto_unregister(&l2tp_ip_prot);
686 module_init(l2tp_ip_init);
687 module_exit(l2tp_ip_exit);
689 MODULE_LICENSE("GPL");
691 MODULE_DESCRIPTION("L2TP over IP");
692 MODULE_VERSION("1.0");
694 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
697 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
698 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);