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 <linux/icmp.h>
15 #include <linux/module.h>
16 #include <linux/skbuff.h>
17 #include <linux/random.h>
18 #include <linux/socket.h>
19 #include <linux/l2tp.h>
25 #include <net/inet_common.h>
26 #include <net/inet_hashtables.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(struct net
*net
, __be32 laddr
, int dif
, u32 tunnel_id
)
54 sk_for_each_bound(sk
, &l2tp_ip_bind_table
) {
55 struct inet_sock
*inet
= inet_sk(sk
);
56 struct l2tp_ip_sock
*l2tp
= l2tp_ip_sk(sk
);
61 if ((l2tp
->conn_id
== tunnel_id
) &&
62 net_eq(sock_net(sk
), net
) &&
63 !(inet
->inet_rcv_saddr
&& inet
->inet_rcv_saddr
!= laddr
) &&
64 !(sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= dif
))
73 static inline struct sock
*l2tp_ip_bind_lookup(struct net
*net
, __be32 laddr
, int dif
, u32 tunnel_id
)
75 struct sock
*sk
= __l2tp_ip_bind_lookup(net
, laddr
, dif
, tunnel_id
);
82 /* When processing receive frames, there are two cases to
83 * consider. Data frames consist of a non-zero session-id and an
84 * optional cookie. Control frames consist of a regular L2TP header
85 * preceded by 32-bits of zeros.
87 * L2TPv3 Session Header Over IP
90 * 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
91 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
93 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
94 * | Cookie (optional, maximum 64 bits)...
95 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
97 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
99 * L2TPv3 Control Message Header Over IP
102 * 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
103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
104 * | (32 bits of zeros) |
105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
107 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * | Control Connection ID |
109 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113 * All control frames are passed to userspace.
115 static int l2tp_ip_recv(struct sk_buff
*skb
)
117 struct net
*net
= dev_net(skb
->dev
);
121 unsigned char *ptr
, *optr
;
122 struct l2tp_session
*session
;
123 struct l2tp_tunnel
*tunnel
= NULL
;
126 /* Point to L2TP header */
127 optr
= ptr
= skb
->data
;
129 if (!pskb_may_pull(skb
, 4))
132 session_id
= ntohl(*((__be32
*) ptr
));
135 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
136 * the session_id. If it is 0, the packet is a L2TP control
137 * frame and the session_id value can be discarded.
139 if (session_id
== 0) {
144 /* Ok, this is a data packet. Lookup the session. */
145 session
= l2tp_session_find(net
, NULL
, session_id
);
149 tunnel
= session
->tunnel
;
153 /* Trace packet contents, if enabled */
154 if (tunnel
->debug
& L2TP_MSG_DATA
) {
155 length
= min(32u, skb
->len
);
156 if (!pskb_may_pull(skb
, length
))
159 pr_debug("%s: ip recv\n", tunnel
->name
);
160 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET
, ptr
, length
);
163 l2tp_recv_common(session
, skb
, ptr
, optr
, 0, skb
->len
, tunnel
->recv_payload_hook
);
168 /* Get the tunnel_id from the L2TP header */
169 if (!pskb_may_pull(skb
, 12))
172 if ((skb
->data
[0] & 0xc0) != 0xc0)
175 tunnel_id
= ntohl(*(__be32
*) &skb
->data
[4]);
176 tunnel
= l2tp_tunnel_find(net
, tunnel_id
);
180 struct iphdr
*iph
= (struct iphdr
*) skb_network_header(skb
);
182 read_lock_bh(&l2tp_ip_lock
);
183 sk
= __l2tp_ip_bind_lookup(net
, iph
->daddr
, 0, tunnel_id
);
184 read_unlock_bh(&l2tp_ip_lock
);
192 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
197 return sk_receive_skb(sk
, skb
, 1);
207 static int l2tp_ip_open(struct sock
*sk
)
209 /* Prevent autobind. We don't have ports. */
210 inet_sk(sk
)->inet_num
= IPPROTO_L2TP
;
212 write_lock_bh(&l2tp_ip_lock
);
213 sk_add_node(sk
, &l2tp_ip_table
);
214 write_unlock_bh(&l2tp_ip_lock
);
219 static void l2tp_ip_close(struct sock
*sk
, long timeout
)
221 write_lock_bh(&l2tp_ip_lock
);
222 hlist_del_init(&sk
->sk_bind_node
);
223 sk_del_node_init(sk
);
224 write_unlock_bh(&l2tp_ip_lock
);
225 sk_common_release(sk
);
228 static void l2tp_ip_destroy_sock(struct sock
*sk
)
231 struct l2tp_tunnel
*tunnel
= l2tp_sock_to_tunnel(sk
);
233 while ((skb
= __skb_dequeue_tail(&sk
->sk_write_queue
)) != NULL
)
237 l2tp_tunnel_closeall(tunnel
);
241 sk_refcnt_debug_dec(sk
);
244 static int l2tp_ip_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
246 struct inet_sock
*inet
= inet_sk(sk
);
247 struct sockaddr_l2tpip
*addr
= (struct sockaddr_l2tpip
*) uaddr
;
248 struct net
*net
= sock_net(sk
);
252 if (!sock_flag(sk
, SOCK_ZAPPED
))
254 if (addr_len
< sizeof(struct sockaddr_l2tpip
))
256 if (addr
->l2tp_family
!= AF_INET
)
260 read_lock_bh(&l2tp_ip_lock
);
261 if (__l2tp_ip_bind_lookup(net
, addr
->l2tp_addr
.s_addr
,
262 sk
->sk_bound_dev_if
, addr
->l2tp_conn_id
))
265 read_unlock_bh(&l2tp_ip_lock
);
268 if (sk
->sk_state
!= TCP_CLOSE
|| addr_len
< sizeof(struct sockaddr_l2tpip
))
271 chk_addr_ret
= inet_addr_type(net
, addr
->l2tp_addr
.s_addr
);
272 ret
= -EADDRNOTAVAIL
;
273 if (addr
->l2tp_addr
.s_addr
&& chk_addr_ret
!= RTN_LOCAL
&&
274 chk_addr_ret
!= RTN_MULTICAST
&& chk_addr_ret
!= RTN_BROADCAST
)
277 if (addr
->l2tp_addr
.s_addr
)
278 inet
->inet_rcv_saddr
= inet
->inet_saddr
= addr
->l2tp_addr
.s_addr
;
279 if (chk_addr_ret
== RTN_MULTICAST
|| chk_addr_ret
== RTN_BROADCAST
)
280 inet
->inet_saddr
= 0; /* Use device */
283 l2tp_ip_sk(sk
)->conn_id
= addr
->l2tp_conn_id
;
285 write_lock_bh(&l2tp_ip_lock
);
286 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
287 sk_del_node_init(sk
);
288 write_unlock_bh(&l2tp_ip_lock
);
290 sock_reset_flag(sk
, SOCK_ZAPPED
);
298 read_unlock_bh(&l2tp_ip_lock
);
303 static int l2tp_ip_connect(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
305 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*) uaddr
;
308 if (sock_flag(sk
, SOCK_ZAPPED
)) /* Must bind first - autobinding does not work */
311 if (addr_len
< sizeof(*lsa
))
314 if (ipv4_is_multicast(lsa
->l2tp_addr
.s_addr
))
317 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
);
334 static int l2tp_ip_disconnect(struct sock
*sk
, int flags
)
336 if (sock_flag(sk
, SOCK_ZAPPED
))
339 return udp_disconnect(sk
, flags
);
342 static int l2tp_ip_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
343 int *uaddr_len
, int peer
)
345 struct sock
*sk
= sock
->sk
;
346 struct inet_sock
*inet
= inet_sk(sk
);
347 struct l2tp_ip_sock
*lsk
= l2tp_ip_sk(sk
);
348 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*)uaddr
;
350 memset(lsa
, 0, sizeof(*lsa
));
351 lsa
->l2tp_family
= AF_INET
;
353 if (!inet
->inet_dport
)
355 lsa
->l2tp_conn_id
= lsk
->peer_conn_id
;
356 lsa
->l2tp_addr
.s_addr
= inet
->inet_daddr
;
358 __be32 addr
= inet
->inet_rcv_saddr
;
360 addr
= inet
->inet_saddr
;
361 lsa
->l2tp_conn_id
= lsk
->conn_id
;
362 lsa
->l2tp_addr
.s_addr
= addr
;
364 *uaddr_len
= sizeof(*lsa
);
368 static int l2tp_ip_backlog_recv(struct sock
*sk
, struct sk_buff
*skb
)
372 /* Charge it to the socket, dropping if the queue is full. */
373 rc
= sock_queue_rcv_skb(sk
, skb
);
380 IP_INC_STATS(sock_net(sk
), IPSTATS_MIB_INDISCARDS
);
385 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
388 static int l2tp_ip_sendmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
, size_t len
)
392 struct inet_sock
*inet
= inet_sk(sk
);
393 struct rtable
*rt
= NULL
;
401 if (sock_flag(sk
, SOCK_DEAD
))
404 /* Get and verify the address. */
406 DECLARE_SOCKADDR(struct sockaddr_l2tpip
*, lip
, msg
->msg_name
);
408 if (msg
->msg_namelen
< sizeof(*lip
))
411 if (lip
->l2tp_family
!= AF_INET
) {
413 if (lip
->l2tp_family
!= AF_UNSPEC
)
417 daddr
= lip
->l2tp_addr
.s_addr
;
420 if (sk
->sk_state
!= TCP_ESTABLISHED
)
423 daddr
= inet
->inet_daddr
;
427 /* Allocate a socket buffer */
429 skb
= sock_wmalloc(sk
, 2 + NET_SKB_PAD
+ sizeof(struct iphdr
) +
430 4 + len
, 0, GFP_KERNEL
);
434 /* Reserve space for headers, putting IP header on 4-byte boundary. */
435 skb_reserve(skb
, 2 + NET_SKB_PAD
);
436 skb_reset_network_header(skb
);
437 skb_reserve(skb
, sizeof(struct iphdr
));
438 skb_reset_transport_header(skb
);
440 /* Insert 0 session_id */
441 *((__be32
*) skb_put(skb
, 4)) = 0;
443 /* Copy user data into skb */
444 rc
= memcpy_fromiovec(skb_put(skb
, len
), msg
->msg_iov
, len
);
450 fl4
= &inet
->cork
.fl
.u
.ip4
;
452 rt
= (struct rtable
*) __sk_dst_check(sk
, 0);
456 const struct ip_options_rcu
*inet_opt
;
458 inet_opt
= rcu_dereference(inet
->inet_opt
);
460 /* Use correct destination address if we have options. */
461 if (inet_opt
&& inet_opt
->opt
.srr
)
462 daddr
= inet_opt
->opt
.faddr
;
464 /* If this fails, retransmit mechanism of transport layer will
465 * keep trying until route appears or the connection times
468 rt
= ip_route_output_ports(sock_net(sk
), fl4
, sk
,
469 daddr
, inet
->inet_saddr
,
470 inet
->inet_dport
, inet
->inet_sport
,
471 sk
->sk_protocol
, RT_CONN_FLAGS(sk
),
472 sk
->sk_bound_dev_if
);
476 sk_setup_caps(sk
, &rt
->dst
);
478 skb_dst_set(skb
, &rt
->dst
);
483 /* We dont need to clone dst here, it is guaranteed to not disappear.
484 * __dev_xmit_skb() might force a refcount if needed.
486 skb_dst_set_noref(skb
, &rt
->dst
);
489 /* Queue the packet to IP for output */
490 rc
= ip_queue_xmit(skb
, &inet
->cork
.fl
);
503 IP_INC_STATS(sock_net(sk
), IPSTATS_MIB_OUTNOROUTES
);
509 static int l2tp_ip_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
510 size_t len
, int noblock
, int flags
, int *addr_len
)
512 struct inet_sock
*inet
= inet_sk(sk
);
514 int err
= -EOPNOTSUPP
;
515 DECLARE_SOCKADDR(struct sockaddr_in
*, sin
, msg
->msg_name
);
521 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
527 msg
->msg_flags
|= MSG_TRUNC
;
531 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
535 sock_recv_timestamp(msg
, sk
, skb
);
537 /* Copy the address. */
539 sin
->sin_family
= AF_INET
;
540 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
542 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
543 *addr_len
= sizeof(*sin
);
545 if (inet
->cmsg_flags
)
546 ip_cmsg_recv(msg
, skb
);
547 if (flags
& MSG_TRUNC
)
550 skb_free_datagram(sk
, skb
);
552 return err
? err
: copied
;
555 static struct proto l2tp_ip_prot
= {
557 .owner
= THIS_MODULE
,
558 .init
= l2tp_ip_open
,
559 .close
= l2tp_ip_close
,
560 .bind
= l2tp_ip_bind
,
561 .connect
= l2tp_ip_connect
,
562 .disconnect
= l2tp_ip_disconnect
,
564 .destroy
= l2tp_ip_destroy_sock
,
565 .setsockopt
= ip_setsockopt
,
566 .getsockopt
= ip_getsockopt
,
567 .sendmsg
= l2tp_ip_sendmsg
,
568 .recvmsg
= l2tp_ip_recvmsg
,
569 .backlog_rcv
= l2tp_ip_backlog_recv
,
571 .unhash
= inet_unhash
,
572 .obj_size
= sizeof(struct l2tp_ip_sock
),
574 .compat_setsockopt
= compat_ip_setsockopt
,
575 .compat_getsockopt
= compat_ip_getsockopt
,
579 static const struct proto_ops l2tp_ip_ops
= {
581 .owner
= THIS_MODULE
,
582 .release
= inet_release
,
584 .connect
= inet_dgram_connect
,
585 .socketpair
= sock_no_socketpair
,
586 .accept
= sock_no_accept
,
587 .getname
= l2tp_ip_getname
,
588 .poll
= datagram_poll
,
590 .listen
= sock_no_listen
,
591 .shutdown
= inet_shutdown
,
592 .setsockopt
= sock_common_setsockopt
,
593 .getsockopt
= sock_common_getsockopt
,
594 .sendmsg
= inet_sendmsg
,
595 .recvmsg
= sock_common_recvmsg
,
596 .mmap
= sock_no_mmap
,
597 .sendpage
= sock_no_sendpage
,
599 .compat_setsockopt
= compat_sock_common_setsockopt
,
600 .compat_getsockopt
= compat_sock_common_getsockopt
,
604 static struct inet_protosw l2tp_ip_protosw
= {
606 .protocol
= IPPROTO_L2TP
,
607 .prot
= &l2tp_ip_prot
,
612 static struct net_protocol l2tp_ip_protocol __read_mostly
= {
613 .handler
= l2tp_ip_recv
,
617 static int __init
l2tp_ip_init(void)
621 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
623 err
= proto_register(&l2tp_ip_prot
, 1);
627 err
= inet_add_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
631 inet_register_protosw(&l2tp_ip_protosw
);
635 proto_unregister(&l2tp_ip_prot
);
640 static void __exit
l2tp_ip_exit(void)
642 inet_unregister_protosw(&l2tp_ip_protosw
);
643 inet_del_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
644 proto_unregister(&l2tp_ip_prot
);
647 module_init(l2tp_ip_init
);
648 module_exit(l2tp_ip_exit
);
650 MODULE_LICENSE("GPL");
651 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
652 MODULE_DESCRIPTION("L2TP over IP");
653 MODULE_VERSION("1.0");
655 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
658 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET
, 2, IPPROTO_L2TP
);