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 #include <linux/icmp.h>
13 #include <linux/module.h>
14 #include <linux/skbuff.h>
15 #include <linux/random.h>
16 #include <linux/socket.h>
17 #include <linux/l2tp.h>
23 #include <net/inet_common.h>
24 #include <net/inet_hashtables.h>
25 #include <net/tcp_states.h>
26 #include <net/protocol.h>
29 #include "l2tp_core.h"
32 /* inet_sock has to be the first member of l2tp_ip_sock */
33 struct inet_sock inet
;
46 static DEFINE_RWLOCK(l2tp_ip_lock
);
47 static struct hlist_head l2tp_ip_table
;
48 static struct hlist_head l2tp_ip_bind_table
;
50 static inline struct l2tp_ip_sock
*l2tp_ip_sk(const struct sock
*sk
)
52 return (struct l2tp_ip_sock
*)sk
;
55 static struct sock
*__l2tp_ip_bind_lookup(struct net
*net
, __be32 laddr
, int dif
, u32 tunnel_id
)
57 struct hlist_node
*node
;
60 sk_for_each_bound(sk
, node
, &l2tp_ip_bind_table
) {
61 struct inet_sock
*inet
= inet_sk(sk
);
62 struct l2tp_ip_sock
*l2tp
= l2tp_ip_sk(sk
);
67 if ((l2tp
->conn_id
== tunnel_id
) &&
69 (sk
->sk_net
== net
) &&
71 !(inet
->inet_rcv_saddr
&& inet
->inet_rcv_saddr
!= laddr
) &&
72 !(sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= dif
))
81 static inline struct sock
*l2tp_ip_bind_lookup(struct net
*net
, __be32 laddr
, int dif
, u32 tunnel_id
)
83 struct sock
*sk
= __l2tp_ip_bind_lookup(net
, laddr
, dif
, tunnel_id
);
90 /* When processing receive frames, there are two cases to
91 * consider. Data frames consist of a non-zero session-id and an
92 * optional cookie. Control frames consist of a regular L2TP header
93 * preceded by 32-bits of zeros.
95 * L2TPv3 Session 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 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
101 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
102 * | Cookie (optional, maximum 64 bits)...
103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107 * L2TPv3 Control Message Header Over IP
110 * 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
111 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112 * | (32 bits of zeros) |
113 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
115 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
116 * | Control Connection ID |
117 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121 * All control frames are passed to userspace.
123 static int l2tp_ip_recv(struct sk_buff
*skb
)
128 unsigned char *ptr
, *optr
;
129 struct l2tp_session
*session
;
130 struct l2tp_tunnel
*tunnel
= NULL
;
134 /* Point to L2TP header */
135 optr
= ptr
= skb
->data
;
137 if (!pskb_may_pull(skb
, 4))
140 session_id
= ntohl(*((__be32
*) ptr
));
143 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
144 * the session_id. If it is 0, the packet is a L2TP control
145 * frame and the session_id value can be discarded.
147 if (session_id
== 0) {
152 /* Ok, this is a data packet. Lookup the session. */
153 session
= l2tp_session_find(&init_net
, NULL
, session_id
);
157 tunnel
= session
->tunnel
;
161 /* Trace packet contents, if enabled */
162 if (tunnel
->debug
& L2TP_MSG_DATA
) {
163 length
= min(32u, skb
->len
);
164 if (!pskb_may_pull(skb
, length
))
167 printk(KERN_DEBUG
"%s: ip recv: ", tunnel
->name
);
171 printk(" %02X", ptr
[offset
]);
172 } while (++offset
< length
);
177 l2tp_recv_common(session
, skb
, ptr
, optr
, 0, skb
->len
, tunnel
->recv_payload_hook
);
182 /* Get the tunnel_id from the L2TP header */
183 if (!pskb_may_pull(skb
, 12))
186 if ((skb
->data
[0] & 0xc0) != 0xc0)
189 tunnel_id
= ntohl(*(__be32
*) &skb
->data
[4]);
190 tunnel
= l2tp_tunnel_find(&init_net
, tunnel_id
);
194 struct iphdr
*iph
= (struct iphdr
*) skb_network_header(skb
);
196 read_lock_bh(&l2tp_ip_lock
);
197 sk
= __l2tp_ip_bind_lookup(&init_net
, iph
->daddr
, 0, tunnel_id
);
198 read_unlock_bh(&l2tp_ip_lock
);
206 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
211 return sk_receive_skb(sk
, skb
, 1);
221 static int l2tp_ip_open(struct sock
*sk
)
223 /* Prevent autobind. We don't have ports. */
224 inet_sk(sk
)->inet_num
= IPPROTO_L2TP
;
226 write_lock_bh(&l2tp_ip_lock
);
227 sk_add_node(sk
, &l2tp_ip_table
);
228 write_unlock_bh(&l2tp_ip_lock
);
233 static void l2tp_ip_close(struct sock
*sk
, long timeout
)
235 write_lock_bh(&l2tp_ip_lock
);
236 hlist_del_init(&sk
->sk_bind_node
);
237 hlist_del_init(&sk
->sk_node
);
238 write_unlock_bh(&l2tp_ip_lock
);
239 sk_common_release(sk
);
242 static void l2tp_ip_destroy_sock(struct sock
*sk
)
246 while ((skb
= __skb_dequeue_tail(&sk
->sk_write_queue
)) != NULL
)
249 sk_refcnt_debug_dec(sk
);
252 static int l2tp_ip_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
254 struct inet_sock
*inet
= inet_sk(sk
);
255 struct sockaddr_l2tpip
*addr
= (struct sockaddr_l2tpip
*) uaddr
;
260 read_lock_bh(&l2tp_ip_lock
);
261 if (__l2tp_ip_bind_lookup(&init_net
, addr
->l2tp_addr
.s_addr
, sk
->sk_bound_dev_if
, addr
->l2tp_conn_id
))
264 read_unlock_bh(&l2tp_ip_lock
);
267 if (sk
->sk_state
!= TCP_CLOSE
|| addr_len
< sizeof(struct sockaddr_l2tpip
))
270 chk_addr_ret
= inet_addr_type(&init_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 inet
->inet_rcv_saddr
= inet
->inet_saddr
= addr
->l2tp_addr
.s_addr
;
277 if (chk_addr_ret
== RTN_MULTICAST
|| chk_addr_ret
== RTN_BROADCAST
)
278 inet
->inet_saddr
= 0; /* Use device */
281 l2tp_ip_sk(sk
)->conn_id
= addr
->l2tp_conn_id
;
283 write_lock_bh(&l2tp_ip_lock
);
284 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
285 sk_del_node_init(sk
);
286 write_unlock_bh(&l2tp_ip_lock
);
294 read_unlock_bh(&l2tp_ip_lock
);
299 static int l2tp_ip_connect(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
302 struct inet_sock
*inet
= inet_sk(sk
);
303 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*) uaddr
;
309 if (addr_len
< sizeof(*lsa
))
313 if (lsa
->l2tp_family
!= AF_INET
)
318 oif
= sk
->sk_bound_dev_if
;
319 saddr
= inet
->inet_saddr
;
322 if (ipv4_is_multicast(lsa
->l2tp_addr
.s_addr
))
325 rc
= ip_route_connect(&rt
, lsa
->l2tp_addr
.s_addr
, saddr
,
326 RT_CONN_FLAGS(sk
), oif
,
330 if (rc
== -ENETUNREACH
)
331 IP_INC_STATS_BH(&init_net
, IPSTATS_MIB_OUTNOROUTES
);
336 if (rt
->rt_flags
& (RTCF_MULTICAST
| RTCF_BROADCAST
)) {
341 l2tp_ip_sk(sk
)->peer_conn_id
= lsa
->l2tp_conn_id
;
343 if (!inet
->inet_saddr
)
344 inet
->inet_saddr
= rt
->rt_src
;
345 if (!inet
->inet_rcv_saddr
)
346 inet
->inet_rcv_saddr
= rt
->rt_src
;
347 inet
->inet_daddr
= rt
->rt_dst
;
348 sk
->sk_state
= TCP_ESTABLISHED
;
349 inet
->inet_id
= jiffies
;
351 sk_dst_set(sk
, &rt
->dst
);
353 write_lock_bh(&l2tp_ip_lock
);
354 hlist_del_init(&sk
->sk_bind_node
);
355 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
356 write_unlock_bh(&l2tp_ip_lock
);
363 static int l2tp_ip_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
364 int *uaddr_len
, int peer
)
366 struct sock
*sk
= sock
->sk
;
367 struct inet_sock
*inet
= inet_sk(sk
);
368 struct l2tp_ip_sock
*lsk
= l2tp_ip_sk(sk
);
369 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*)uaddr
;
371 memset(lsa
, 0, sizeof(*lsa
));
372 lsa
->l2tp_family
= AF_INET
;
374 if (!inet
->inet_dport
)
376 lsa
->l2tp_conn_id
= lsk
->peer_conn_id
;
377 lsa
->l2tp_addr
.s_addr
= inet
->inet_daddr
;
379 __be32 addr
= inet
->inet_rcv_saddr
;
381 addr
= inet
->inet_saddr
;
382 lsa
->l2tp_conn_id
= lsk
->conn_id
;
383 lsa
->l2tp_addr
.s_addr
= addr
;
385 *uaddr_len
= sizeof(*lsa
);
389 static int l2tp_ip_backlog_recv(struct sock
*sk
, struct sk_buff
*skb
)
393 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
398 /* Charge it to the socket, dropping if the queue is full. */
399 rc
= sock_queue_rcv_skb(sk
, skb
);
406 IP_INC_STATS(&init_net
, IPSTATS_MIB_INDISCARDS
);
411 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
414 static int l2tp_ip_sendmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
, size_t len
)
418 struct l2tp_ip_sock
*lsa
= l2tp_ip_sk(sk
);
419 struct inet_sock
*inet
= inet_sk(sk
);
420 struct ip_options
*opt
= inet
->opt
;
421 struct rtable
*rt
= NULL
;
425 if (sock_flag(sk
, SOCK_DEAD
))
428 /* Get and verify the address. */
430 struct sockaddr_l2tpip
*lip
= (struct sockaddr_l2tpip
*) msg
->msg_name
;
431 if (msg
->msg_namelen
< sizeof(*lip
))
434 if (lip
->l2tp_family
!= AF_INET
) {
435 if (lip
->l2tp_family
!= AF_UNSPEC
)
436 return -EAFNOSUPPORT
;
439 daddr
= lip
->l2tp_addr
.s_addr
;
441 if (sk
->sk_state
!= TCP_ESTABLISHED
)
442 return -EDESTADDRREQ
;
444 daddr
= inet
->inet_daddr
;
448 /* Allocate a socket buffer */
450 skb
= sock_wmalloc(sk
, 2 + NET_SKB_PAD
+ sizeof(struct iphdr
) +
451 4 + len
, 0, GFP_KERNEL
);
455 /* Reserve space for headers, putting IP header on 4-byte boundary. */
456 skb_reserve(skb
, 2 + NET_SKB_PAD
);
457 skb_reset_network_header(skb
);
458 skb_reserve(skb
, sizeof(struct iphdr
));
459 skb_reset_transport_header(skb
);
461 /* Insert 0 session_id */
462 *((__be32
*) skb_put(skb
, 4)) = 0;
464 /* Copy user data into skb */
465 rc
= memcpy_fromiovec(skb_put(skb
, len
), msg
->msg_iov
, len
);
472 rt
= (struct rtable
*) __sk_dst_check(sk
, 0);
475 /* Use correct destination address if we have options. */
480 struct flowi fl
= { .oif
= sk
->sk_bound_dev_if
,
483 .saddr
= inet
->inet_saddr
,
484 .tos
= RT_CONN_FLAGS(sk
) } },
485 .proto
= sk
->sk_protocol
,
486 .flags
= inet_sk_flowi_flags(sk
),
487 .uli_u
= { .ports
= {
488 .sport
= inet
->inet_sport
,
489 .dport
= inet
->inet_dport
} } };
491 /* If this fails, retransmit mechanism of transport layer will
492 * keep trying until route appears or the connection times
495 security_sk_classify_flow(sk
, &fl
);
496 if (ip_route_output_flow(sock_net(sk
), &rt
, &fl
, sk
, 0))
499 sk_setup_caps(sk
, &rt
->dst
);
501 skb_dst_set(skb
, dst_clone(&rt
->dst
));
503 /* Queue the packet to IP for output */
504 rc
= ip_queue_xmit(skb
);
510 lsa
->tx_bytes
+= len
;
519 IP_INC_STATS(sock_net(sk
), IPSTATS_MIB_OUTNOROUTES
);
521 return -EHOSTUNREACH
;
524 static int l2tp_ip_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
525 size_t len
, int noblock
, int flags
, int *addr_len
)
527 struct inet_sock
*inet
= inet_sk(sk
);
528 struct l2tp_ip_sock
*lsk
= l2tp_ip_sk(sk
);
530 int err
= -EOPNOTSUPP
;
531 struct sockaddr_in
*sin
= (struct sockaddr_in
*)msg
->msg_name
;
538 *addr_len
= sizeof(*sin
);
540 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
546 msg
->msg_flags
|= MSG_TRUNC
;
550 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
554 sock_recv_timestamp(msg
, sk
, skb
);
556 /* Copy the address. */
558 sin
->sin_family
= AF_INET
;
559 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
561 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
563 if (inet
->cmsg_flags
)
564 ip_cmsg_recv(msg
, skb
);
565 if (flags
& MSG_TRUNC
)
568 skb_free_datagram(sk
, skb
);
576 lsk
->rx_bytes
+= copied
;
581 struct proto l2tp_ip_prot
= {
583 .owner
= THIS_MODULE
,
584 .init
= l2tp_ip_open
,
585 .close
= l2tp_ip_close
,
586 .bind
= l2tp_ip_bind
,
587 .connect
= l2tp_ip_connect
,
588 .disconnect
= udp_disconnect
,
590 .destroy
= l2tp_ip_destroy_sock
,
591 .setsockopt
= ip_setsockopt
,
592 .getsockopt
= ip_getsockopt
,
593 .sendmsg
= l2tp_ip_sendmsg
,
594 .recvmsg
= l2tp_ip_recvmsg
,
595 .backlog_rcv
= l2tp_ip_backlog_recv
,
597 .unhash
= inet_unhash
,
598 .obj_size
= sizeof(struct l2tp_ip_sock
),
600 .compat_setsockopt
= compat_ip_setsockopt
,
601 .compat_getsockopt
= compat_ip_getsockopt
,
605 static const struct proto_ops l2tp_ip_ops
= {
607 .owner
= THIS_MODULE
,
608 .release
= inet_release
,
610 .connect
= inet_dgram_connect
,
611 .socketpair
= sock_no_socketpair
,
612 .accept
= sock_no_accept
,
613 .getname
= l2tp_ip_getname
,
614 .poll
= datagram_poll
,
616 .listen
= sock_no_listen
,
617 .shutdown
= inet_shutdown
,
618 .setsockopt
= sock_common_setsockopt
,
619 .getsockopt
= sock_common_getsockopt
,
620 .sendmsg
= inet_sendmsg
,
621 .recvmsg
= sock_common_recvmsg
,
622 .mmap
= sock_no_mmap
,
623 .sendpage
= sock_no_sendpage
,
625 .compat_setsockopt
= compat_sock_common_setsockopt
,
626 .compat_getsockopt
= compat_sock_common_getsockopt
,
630 static struct inet_protosw l2tp_ip_protosw
= {
632 .protocol
= IPPROTO_L2TP
,
633 .prot
= &l2tp_ip_prot
,
638 static struct net_protocol l2tp_ip_protocol __read_mostly
= {
639 .handler
= l2tp_ip_recv
,
642 static int __init
l2tp_ip_init(void)
646 printk(KERN_INFO
"L2TP IP encapsulation support (L2TPv3)\n");
648 err
= proto_register(&l2tp_ip_prot
, 1);
652 err
= inet_add_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
656 inet_register_protosw(&l2tp_ip_protosw
);
660 proto_unregister(&l2tp_ip_prot
);
665 static void __exit
l2tp_ip_exit(void)
667 inet_unregister_protosw(&l2tp_ip_protosw
);
668 inet_del_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
669 proto_unregister(&l2tp_ip_prot
);
672 module_init(l2tp_ip_init
);
673 module_exit(l2tp_ip_exit
);
675 MODULE_LICENSE("GPL");
676 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
677 MODULE_DESCRIPTION("L2TP over IP");
678 MODULE_VERSION("1.0");
679 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET
, SOCK_DGRAM
, IPPROTO_L2TP
);