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
) &&
68 net_eq(sock_net(sk
), net
) &&
69 !(inet
->inet_rcv_saddr
&& inet
->inet_rcv_saddr
!= laddr
) &&
70 !(sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= dif
))
79 static inline struct sock
*l2tp_ip_bind_lookup(struct net
*net
, __be32 laddr
, int dif
, u32 tunnel_id
)
81 struct sock
*sk
= __l2tp_ip_bind_lookup(net
, laddr
, dif
, tunnel_id
);
88 /* When processing receive frames, there are two cases to
89 * consider. Data frames consist of a non-zero session-id and an
90 * optional cookie. Control frames consist of a regular L2TP header
91 * preceded by 32-bits of zeros.
93 * L2TPv3 Session Header Over IP
96 * 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
97 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
99 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100 * | Cookie (optional, maximum 64 bits)...
101 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105 * L2TPv3 Control Message Header Over IP
108 * 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
109 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
110 * | (32 bits of zeros) |
111 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
113 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114 * | Control Connection ID |
115 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * All control frames are passed to userspace.
121 static int l2tp_ip_recv(struct sk_buff
*skb
)
126 unsigned char *ptr
, *optr
;
127 struct l2tp_session
*session
;
128 struct l2tp_tunnel
*tunnel
= NULL
;
132 /* Point to L2TP header */
133 optr
= ptr
= skb
->data
;
135 if (!pskb_may_pull(skb
, 4))
138 session_id
= ntohl(*((__be32
*) ptr
));
141 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
142 * the session_id. If it is 0, the packet is a L2TP control
143 * frame and the session_id value can be discarded.
145 if (session_id
== 0) {
150 /* Ok, this is a data packet. Lookup the session. */
151 session
= l2tp_session_find(&init_net
, NULL
, session_id
);
155 tunnel
= session
->tunnel
;
159 /* Trace packet contents, if enabled */
160 if (tunnel
->debug
& L2TP_MSG_DATA
) {
161 length
= min(32u, skb
->len
);
162 if (!pskb_may_pull(skb
, length
))
165 printk(KERN_DEBUG
"%s: ip recv: ", tunnel
->name
);
169 printk(" %02X", ptr
[offset
]);
170 } while (++offset
< length
);
175 l2tp_recv_common(session
, skb
, ptr
, optr
, 0, skb
->len
, tunnel
->recv_payload_hook
);
180 /* Get the tunnel_id from the L2TP header */
181 if (!pskb_may_pull(skb
, 12))
184 if ((skb
->data
[0] & 0xc0) != 0xc0)
187 tunnel_id
= ntohl(*(__be32
*) &skb
->data
[4]);
188 tunnel
= l2tp_tunnel_find(&init_net
, tunnel_id
);
192 struct iphdr
*iph
= (struct iphdr
*) skb_network_header(skb
);
194 read_lock_bh(&l2tp_ip_lock
);
195 sk
= __l2tp_ip_bind_lookup(&init_net
, iph
->daddr
, 0, tunnel_id
);
196 read_unlock_bh(&l2tp_ip_lock
);
204 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
209 return sk_receive_skb(sk
, skb
, 1);
219 static int l2tp_ip_open(struct sock
*sk
)
221 /* Prevent autobind. We don't have ports. */
222 inet_sk(sk
)->inet_num
= IPPROTO_L2TP
;
224 write_lock_bh(&l2tp_ip_lock
);
225 sk_add_node(sk
, &l2tp_ip_table
);
226 write_unlock_bh(&l2tp_ip_lock
);
231 static void l2tp_ip_close(struct sock
*sk
, long timeout
)
233 write_lock_bh(&l2tp_ip_lock
);
234 hlist_del_init(&sk
->sk_bind_node
);
235 hlist_del_init(&sk
->sk_node
);
236 write_unlock_bh(&l2tp_ip_lock
);
237 sk_common_release(sk
);
240 static void l2tp_ip_destroy_sock(struct sock
*sk
)
244 while ((skb
= __skb_dequeue_tail(&sk
->sk_write_queue
)) != NULL
)
247 sk_refcnt_debug_dec(sk
);
250 static int l2tp_ip_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
252 struct inet_sock
*inet
= inet_sk(sk
);
253 struct sockaddr_l2tpip
*addr
= (struct sockaddr_l2tpip
*) uaddr
;
258 read_lock_bh(&l2tp_ip_lock
);
259 if (__l2tp_ip_bind_lookup(&init_net
, addr
->l2tp_addr
.s_addr
, sk
->sk_bound_dev_if
, addr
->l2tp_conn_id
))
262 read_unlock_bh(&l2tp_ip_lock
);
265 if (sk
->sk_state
!= TCP_CLOSE
|| addr_len
< sizeof(struct sockaddr_l2tpip
))
268 chk_addr_ret
= inet_addr_type(&init_net
, addr
->l2tp_addr
.s_addr
);
269 ret
= -EADDRNOTAVAIL
;
270 if (addr
->l2tp_addr
.s_addr
&& chk_addr_ret
!= RTN_LOCAL
&&
271 chk_addr_ret
!= RTN_MULTICAST
&& chk_addr_ret
!= RTN_BROADCAST
)
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 */
279 l2tp_ip_sk(sk
)->conn_id
= addr
->l2tp_conn_id
;
281 write_lock_bh(&l2tp_ip_lock
);
282 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
283 sk_del_node_init(sk
);
284 write_unlock_bh(&l2tp_ip_lock
);
292 read_unlock_bh(&l2tp_ip_lock
);
297 static int l2tp_ip_connect(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
299 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*) uaddr
;
300 struct inet_sock
*inet
= inet_sk(sk
);
307 if (addr_len
< sizeof(*lsa
))
311 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 fl4
= &inet
->cork
.fl
.u
.ip4
;
326 rt
= ip_route_connect(fl4
, lsa
->l2tp_addr
.s_addr
, saddr
,
327 RT_CONN_FLAGS(sk
), oif
,
332 if (rc
== -ENETUNREACH
)
333 IP_INC_STATS_BH(&init_net
, IPSTATS_MIB_OUTNOROUTES
);
338 if (rt
->rt_flags
& (RTCF_MULTICAST
| RTCF_BROADCAST
)) {
343 l2tp_ip_sk(sk
)->peer_conn_id
= lsa
->l2tp_conn_id
;
345 if (!inet
->inet_saddr
)
346 inet
->inet_saddr
= fl4
->saddr
;
347 if (!inet
->inet_rcv_saddr
)
348 inet
->inet_rcv_saddr
= fl4
->saddr
;
349 inet
->inet_daddr
= fl4
->daddr
;
350 sk
->sk_state
= TCP_ESTABLISHED
;
351 inet
->inet_id
= jiffies
;
353 sk_dst_set(sk
, &rt
->dst
);
355 write_lock_bh(&l2tp_ip_lock
);
356 hlist_del_init(&sk
->sk_bind_node
);
357 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
358 write_unlock_bh(&l2tp_ip_lock
);
366 static int l2tp_ip_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
367 int *uaddr_len
, int peer
)
369 struct sock
*sk
= sock
->sk
;
370 struct inet_sock
*inet
= inet_sk(sk
);
371 struct l2tp_ip_sock
*lsk
= l2tp_ip_sk(sk
);
372 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*)uaddr
;
374 memset(lsa
, 0, sizeof(*lsa
));
375 lsa
->l2tp_family
= AF_INET
;
377 if (!inet
->inet_dport
)
379 lsa
->l2tp_conn_id
= lsk
->peer_conn_id
;
380 lsa
->l2tp_addr
.s_addr
= inet
->inet_daddr
;
382 __be32 addr
= inet
->inet_rcv_saddr
;
384 addr
= inet
->inet_saddr
;
385 lsa
->l2tp_conn_id
= lsk
->conn_id
;
386 lsa
->l2tp_addr
.s_addr
= addr
;
388 *uaddr_len
= sizeof(*lsa
);
392 static int l2tp_ip_backlog_recv(struct sock
*sk
, struct sk_buff
*skb
)
396 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
401 /* Charge it to the socket, dropping if the queue is full. */
402 rc
= sock_queue_rcv_skb(sk
, skb
);
409 IP_INC_STATS(&init_net
, IPSTATS_MIB_INDISCARDS
);
414 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
417 static int l2tp_ip_sendmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
, size_t len
)
421 struct l2tp_ip_sock
*lsa
= l2tp_ip_sk(sk
);
422 struct inet_sock
*inet
= inet_sk(sk
);
423 struct rtable
*rt
= NULL
;
431 if (sock_flag(sk
, SOCK_DEAD
))
434 /* Get and verify the address. */
436 struct sockaddr_l2tpip
*lip
= (struct sockaddr_l2tpip
*) msg
->msg_name
;
438 if (msg
->msg_namelen
< sizeof(*lip
))
441 if (lip
->l2tp_family
!= AF_INET
) {
443 if (lip
->l2tp_family
!= AF_UNSPEC
)
447 daddr
= lip
->l2tp_addr
.s_addr
;
449 if (sk
->sk_state
!= TCP_ESTABLISHED
)
450 return -EDESTADDRREQ
;
452 daddr
= inet
->inet_daddr
;
456 /* Allocate a socket buffer */
458 skb
= sock_wmalloc(sk
, 2 + NET_SKB_PAD
+ sizeof(struct iphdr
) +
459 4 + len
, 0, GFP_KERNEL
);
463 /* Reserve space for headers, putting IP header on 4-byte boundary. */
464 skb_reserve(skb
, 2 + NET_SKB_PAD
);
465 skb_reset_network_header(skb
);
466 skb_reserve(skb
, sizeof(struct iphdr
));
467 skb_reset_transport_header(skb
);
469 /* Insert 0 session_id */
470 *((__be32
*) skb_put(skb
, 4)) = 0;
472 /* Copy user data into skb */
473 rc
= memcpy_fromiovec(skb_put(skb
, len
), msg
->msg_iov
, len
);
479 fl4
= &inet
->cork
.fl
.u
.ip4
;
481 rt
= (struct rtable
*) __sk_dst_check(sk
, 0);
485 const struct ip_options_rcu
*inet_opt
;
487 inet_opt
= rcu_dereference(inet
->inet_opt
);
489 /* Use correct destination address if we have options. */
490 if (inet_opt
&& inet_opt
->opt
.srr
)
491 daddr
= inet_opt
->opt
.faddr
;
493 /* If this fails, retransmit mechanism of transport layer will
494 * keep trying until route appears or the connection times
497 rt
= ip_route_output_ports(sock_net(sk
), fl4
, sk
,
498 daddr
, inet
->inet_saddr
,
499 inet
->inet_dport
, inet
->inet_sport
,
500 sk
->sk_protocol
, RT_CONN_FLAGS(sk
),
501 sk
->sk_bound_dev_if
);
505 sk_setup_caps(sk
, &rt
->dst
);
507 dst_release(&rt
->dst
); /* safe since we hold rcu_read_lock */
510 /* We dont need to clone dst here, it is guaranteed to not disappear.
511 * __dev_xmit_skb() might force a refcount if needed.
513 skb_dst_set_noref(skb
, &rt
->dst
);
515 /* Queue the packet to IP for output */
516 rc
= ip_queue_xmit(skb
, &inet
->cork
.fl
);
523 lsa
->tx_bytes
+= len
;
535 IP_INC_STATS(sock_net(sk
), IPSTATS_MIB_OUTNOROUTES
);
541 static int l2tp_ip_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
542 size_t len
, int noblock
, int flags
, int *addr_len
)
544 struct inet_sock
*inet
= inet_sk(sk
);
545 struct l2tp_ip_sock
*lsk
= l2tp_ip_sk(sk
);
547 int err
= -EOPNOTSUPP
;
548 struct sockaddr_in
*sin
= (struct sockaddr_in
*)msg
->msg_name
;
555 *addr_len
= sizeof(*sin
);
557 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
563 msg
->msg_flags
|= MSG_TRUNC
;
567 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
571 sock_recv_timestamp(msg
, sk
, skb
);
573 /* Copy the address. */
575 sin
->sin_family
= AF_INET
;
576 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
578 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
580 if (inet
->cmsg_flags
)
581 ip_cmsg_recv(msg
, skb
);
582 if (flags
& MSG_TRUNC
)
585 skb_free_datagram(sk
, skb
);
593 lsk
->rx_bytes
+= copied
;
598 static struct proto l2tp_ip_prot
= {
600 .owner
= THIS_MODULE
,
601 .init
= l2tp_ip_open
,
602 .close
= l2tp_ip_close
,
603 .bind
= l2tp_ip_bind
,
604 .connect
= l2tp_ip_connect
,
605 .disconnect
= udp_disconnect
,
607 .destroy
= l2tp_ip_destroy_sock
,
608 .setsockopt
= ip_setsockopt
,
609 .getsockopt
= ip_getsockopt
,
610 .sendmsg
= l2tp_ip_sendmsg
,
611 .recvmsg
= l2tp_ip_recvmsg
,
612 .backlog_rcv
= l2tp_ip_backlog_recv
,
614 .unhash
= inet_unhash
,
615 .obj_size
= sizeof(struct l2tp_ip_sock
),
617 .compat_setsockopt
= compat_ip_setsockopt
,
618 .compat_getsockopt
= compat_ip_getsockopt
,
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 .listen
= sock_no_listen
,
634 .shutdown
= inet_shutdown
,
635 .setsockopt
= sock_common_setsockopt
,
636 .getsockopt
= sock_common_getsockopt
,
637 .sendmsg
= inet_sendmsg
,
638 .recvmsg
= sock_common_recvmsg
,
639 .mmap
= sock_no_mmap
,
640 .sendpage
= sock_no_sendpage
,
642 .compat_setsockopt
= compat_sock_common_setsockopt
,
643 .compat_getsockopt
= compat_sock_common_getsockopt
,
647 static struct inet_protosw l2tp_ip_protosw
= {
649 .protocol
= IPPROTO_L2TP
,
650 .prot
= &l2tp_ip_prot
,
655 static struct net_protocol l2tp_ip_protocol __read_mostly
= {
656 .handler
= l2tp_ip_recv
,
659 static int __init
l2tp_ip_init(void)
663 printk(KERN_INFO
"L2TP IP encapsulation support (L2TPv3)\n");
665 err
= proto_register(&l2tp_ip_prot
, 1);
669 err
= inet_add_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
673 inet_register_protosw(&l2tp_ip_protosw
);
677 proto_unregister(&l2tp_ip_prot
);
682 static void __exit
l2tp_ip_exit(void)
684 inet_unregister_protosw(&l2tp_ip_protosw
);
685 inet_del_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
686 proto_unregister(&l2tp_ip_prot
);
689 module_init(l2tp_ip_init
);
690 module_exit(l2tp_ip_exit
);
692 MODULE_LICENSE("GPL");
693 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
694 MODULE_DESCRIPTION("L2TP over IP");
695 MODULE_VERSION("1.0");
697 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
700 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET
, 2, IPPROTO_L2TP
);