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 <asm/ioctls.h>
15 #include <linux/icmp.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/random.h>
19 #include <linux/socket.h>
20 #include <linux/l2tp.h>
26 #include <net/inet_common.h>
27 #include <net/inet_hashtables.h>
28 #include <net/tcp_states.h>
29 #include <net/protocol.h>
32 #include "l2tp_core.h"
35 /* inet_sock has to be the first member of l2tp_ip_sock */
36 struct inet_sock inet
;
42 static DEFINE_RWLOCK(l2tp_ip_lock
);
43 static struct hlist_head l2tp_ip_table
;
44 static struct hlist_head l2tp_ip_bind_table
;
46 static inline struct l2tp_ip_sock
*l2tp_ip_sk(const struct sock
*sk
)
48 return (struct l2tp_ip_sock
*)sk
;
51 static struct sock
*__l2tp_ip_bind_lookup(const struct net
*net
, __be32 laddr
,
52 __be32 raddr
, int dif
, u32 tunnel_id
)
56 sk_for_each_bound(sk
, &l2tp_ip_bind_table
) {
57 struct inet_sock
*inet
= inet_sk(sk
);
58 struct l2tp_ip_sock
*l2tp
= l2tp_ip_sk(sk
);
63 if ((l2tp
->conn_id
== tunnel_id
) &&
64 net_eq(sock_net(sk
), net
) &&
65 !(inet
->inet_rcv_saddr
&& inet
->inet_rcv_saddr
!= laddr
) &&
66 (!inet
->inet_daddr
|| !raddr
|| inet
->inet_daddr
== raddr
) &&
67 (!sk
->sk_bound_dev_if
|| !dif
||
68 sk
->sk_bound_dev_if
== dif
))
77 /* When processing receive frames, there are two cases to
78 * consider. Data frames consist of a non-zero session-id and an
79 * optional cookie. Control frames consist of a regular L2TP header
80 * preceded by 32-bits of zeros.
82 * L2TPv3 Session Header Over IP
85 * 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
86 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
88 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
89 * | Cookie (optional, maximum 64 bits)...
90 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
92 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
94 * L2TPv3 Control Message Header Over IP
97 * 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
98 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
99 * | (32 bits of zeros) |
100 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
101 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
102 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103 * | Control Connection ID |
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * All control frames are passed to userspace.
110 static int l2tp_ip_recv(struct sk_buff
*skb
)
112 struct net
*net
= dev_net(skb
->dev
);
116 unsigned char *ptr
, *optr
;
117 struct l2tp_session
*session
;
118 struct l2tp_tunnel
*tunnel
= NULL
;
121 if (!pskb_may_pull(skb
, 4))
124 /* Point to L2TP header */
125 optr
= ptr
= skb
->data
;
126 session_id
= ntohl(*((__be32
*) ptr
));
129 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
130 * the session_id. If it is 0, the packet is a L2TP control
131 * frame and the session_id value can be discarded.
133 if (session_id
== 0) {
138 /* Ok, this is a data packet. Lookup the session. */
139 session
= l2tp_session_get(net
, NULL
, session_id
, true);
143 tunnel
= session
->tunnel
;
147 /* Trace packet contents, if enabled */
148 if (tunnel
->debug
& L2TP_MSG_DATA
) {
149 length
= min(32u, skb
->len
);
150 if (!pskb_may_pull(skb
, length
))
153 /* Point to L2TP header */
154 optr
= ptr
= skb
->data
;
156 pr_debug("%s: ip recv\n", tunnel
->name
);
157 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET
, ptr
, length
);
160 if (l2tp_v3_ensure_opt_in_linear(session
, skb
, &ptr
, &optr
))
163 l2tp_recv_common(session
, skb
, ptr
, optr
, 0, skb
->len
, tunnel
->recv_payload_hook
);
164 l2tp_session_dec_refcount(session
);
169 /* Get the tunnel_id from the L2TP header */
170 if (!pskb_may_pull(skb
, 12))
173 if ((skb
->data
[0] & 0xc0) != 0xc0)
176 tunnel_id
= ntohl(*(__be32
*) &skb
->data
[4]);
177 tunnel
= l2tp_tunnel_find(net
, tunnel_id
);
182 struct iphdr
*iph
= (struct iphdr
*) skb_network_header(skb
);
184 read_lock_bh(&l2tp_ip_lock
);
185 sk
= __l2tp_ip_bind_lookup(net
, iph
->daddr
, iph
->saddr
,
186 inet_iif(skb
), tunnel_id
);
188 read_unlock_bh(&l2tp_ip_lock
);
193 read_unlock_bh(&l2tp_ip_lock
);
196 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
201 return sk_receive_skb(sk
, skb
, 1);
205 session
->deref(session
);
206 l2tp_session_dec_refcount(session
);
217 static int l2tp_ip_open(struct sock
*sk
)
219 /* Prevent autobind. We don't have ports. */
220 inet_sk(sk
)->inet_num
= IPPROTO_L2TP
;
222 write_lock_bh(&l2tp_ip_lock
);
223 sk_add_node(sk
, &l2tp_ip_table
);
224 write_unlock_bh(&l2tp_ip_lock
);
229 static void l2tp_ip_close(struct sock
*sk
, long timeout
)
231 write_lock_bh(&l2tp_ip_lock
);
232 hlist_del_init(&sk
->sk_bind_node
);
233 sk_del_node_init(sk
);
234 write_unlock_bh(&l2tp_ip_lock
);
235 sk_common_release(sk
);
238 static void l2tp_ip_destroy_sock(struct sock
*sk
)
241 struct l2tp_tunnel
*tunnel
= l2tp_sock_to_tunnel(sk
);
243 while ((skb
= __skb_dequeue_tail(&sk
->sk_write_queue
)) != NULL
)
247 l2tp_tunnel_closeall(tunnel
);
251 sk_refcnt_debug_dec(sk
);
254 static int l2tp_ip_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
256 struct inet_sock
*inet
= inet_sk(sk
);
257 struct sockaddr_l2tpip
*addr
= (struct sockaddr_l2tpip
*) uaddr
;
258 struct net
*net
= sock_net(sk
);
262 if (addr_len
< sizeof(struct sockaddr_l2tpip
))
264 if (addr
->l2tp_family
!= AF_INET
)
270 if (!sock_flag(sk
, SOCK_ZAPPED
))
273 if (sk
->sk_state
!= TCP_CLOSE
|| addr_len
< sizeof(struct sockaddr_l2tpip
))
276 chk_addr_ret
= inet_addr_type(net
, addr
->l2tp_addr
.s_addr
);
277 ret
= -EADDRNOTAVAIL
;
278 if (addr
->l2tp_addr
.s_addr
&& chk_addr_ret
!= RTN_LOCAL
&&
279 chk_addr_ret
!= RTN_MULTICAST
&& chk_addr_ret
!= RTN_BROADCAST
)
282 if (addr
->l2tp_addr
.s_addr
)
283 inet
->inet_rcv_saddr
= inet
->inet_saddr
= addr
->l2tp_addr
.s_addr
;
284 if (chk_addr_ret
== RTN_MULTICAST
|| chk_addr_ret
== RTN_BROADCAST
)
285 inet
->inet_saddr
= 0; /* Use device */
287 write_lock_bh(&l2tp_ip_lock
);
288 if (__l2tp_ip_bind_lookup(net
, addr
->l2tp_addr
.s_addr
, 0,
289 sk
->sk_bound_dev_if
, addr
->l2tp_conn_id
)) {
290 write_unlock_bh(&l2tp_ip_lock
);
296 l2tp_ip_sk(sk
)->conn_id
= addr
->l2tp_conn_id
;
298 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
299 sk_del_node_init(sk
);
300 write_unlock_bh(&l2tp_ip_lock
);
303 sock_reset_flag(sk
, SOCK_ZAPPED
);
311 static int l2tp_ip_connect(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
313 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*) uaddr
;
316 if (addr_len
< sizeof(*lsa
))
319 if (ipv4_is_multicast(lsa
->l2tp_addr
.s_addr
))
324 /* Must bind first - autobinding does not work */
325 if (sock_flag(sk
, SOCK_ZAPPED
)) {
330 rc
= __ip4_datagram_connect(sk
, uaddr
, addr_len
);
334 l2tp_ip_sk(sk
)->peer_conn_id
= lsa
->l2tp_conn_id
;
336 write_lock_bh(&l2tp_ip_lock
);
337 hlist_del_init(&sk
->sk_bind_node
);
338 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
339 write_unlock_bh(&l2tp_ip_lock
);
347 static int l2tp_ip_disconnect(struct sock
*sk
, int flags
)
349 if (sock_flag(sk
, SOCK_ZAPPED
))
352 return __udp_disconnect(sk
, flags
);
355 static int l2tp_ip_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
356 int *uaddr_len
, int peer
)
358 struct sock
*sk
= sock
->sk
;
359 struct inet_sock
*inet
= inet_sk(sk
);
360 struct l2tp_ip_sock
*lsk
= l2tp_ip_sk(sk
);
361 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*)uaddr
;
363 memset(lsa
, 0, sizeof(*lsa
));
364 lsa
->l2tp_family
= AF_INET
;
366 if (!inet
->inet_dport
)
368 lsa
->l2tp_conn_id
= lsk
->peer_conn_id
;
369 lsa
->l2tp_addr
.s_addr
= inet
->inet_daddr
;
371 __be32 addr
= inet
->inet_rcv_saddr
;
373 addr
= inet
->inet_saddr
;
374 lsa
->l2tp_conn_id
= lsk
->conn_id
;
375 lsa
->l2tp_addr
.s_addr
= addr
;
377 *uaddr_len
= sizeof(*lsa
);
381 static int l2tp_ip_backlog_recv(struct sock
*sk
, struct sk_buff
*skb
)
385 /* Charge it to the socket, dropping if the queue is full. */
386 rc
= sock_queue_rcv_skb(sk
, skb
);
393 IP_INC_STATS(sock_net(sk
), IPSTATS_MIB_INDISCARDS
);
398 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
401 static int l2tp_ip_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
)
405 struct inet_sock
*inet
= inet_sk(sk
);
406 struct rtable
*rt
= NULL
;
414 if (sock_flag(sk
, SOCK_DEAD
))
417 /* Get and verify the address. */
419 DECLARE_SOCKADDR(struct sockaddr_l2tpip
*, lip
, msg
->msg_name
);
421 if (msg
->msg_namelen
< sizeof(*lip
))
424 if (lip
->l2tp_family
!= AF_INET
) {
426 if (lip
->l2tp_family
!= AF_UNSPEC
)
430 daddr
= lip
->l2tp_addr
.s_addr
;
433 if (sk
->sk_state
!= TCP_ESTABLISHED
)
436 daddr
= inet
->inet_daddr
;
440 /* Allocate a socket buffer */
442 skb
= sock_wmalloc(sk
, 2 + NET_SKB_PAD
+ sizeof(struct iphdr
) +
443 4 + len
, 0, GFP_KERNEL
);
447 /* Reserve space for headers, putting IP header on 4-byte boundary. */
448 skb_reserve(skb
, 2 + NET_SKB_PAD
);
449 skb_reset_network_header(skb
);
450 skb_reserve(skb
, sizeof(struct iphdr
));
451 skb_reset_transport_header(skb
);
453 /* Insert 0 session_id */
454 *((__be32
*) skb_put(skb
, 4)) = 0;
456 /* Copy user data into skb */
457 rc
= memcpy_from_msg(skb_put(skb
, len
), msg
, len
);
463 fl4
= &inet
->cork
.fl
.u
.ip4
;
465 rt
= (struct rtable
*) __sk_dst_check(sk
, 0);
469 const struct ip_options_rcu
*inet_opt
;
471 inet_opt
= rcu_dereference(inet
->inet_opt
);
473 /* Use correct destination address if we have options. */
474 if (inet_opt
&& inet_opt
->opt
.srr
)
475 daddr
= inet_opt
->opt
.faddr
;
477 /* If this fails, retransmit mechanism of transport layer will
478 * keep trying until route appears or the connection times
481 rt
= ip_route_output_ports(sock_net(sk
), fl4
, sk
,
482 daddr
, inet
->inet_saddr
,
483 inet
->inet_dport
, inet
->inet_sport
,
484 sk
->sk_protocol
, RT_CONN_FLAGS(sk
),
485 sk
->sk_bound_dev_if
);
489 sk_setup_caps(sk
, &rt
->dst
);
491 skb_dst_set(skb
, &rt
->dst
);
496 /* We dont need to clone dst here, it is guaranteed to not disappear.
497 * __dev_xmit_skb() might force a refcount if needed.
499 skb_dst_set_noref(skb
, &rt
->dst
);
502 /* Queue the packet to IP for output */
503 rc
= ip_queue_xmit(sk
, skb
, &inet
->cork
.fl
);
516 IP_INC_STATS(sock_net(sk
), IPSTATS_MIB_OUTNOROUTES
);
522 static int l2tp_ip_recvmsg(struct sock
*sk
, struct msghdr
*msg
,
523 size_t len
, int noblock
, int flags
, int *addr_len
)
525 struct inet_sock
*inet
= inet_sk(sk
);
527 int err
= -EOPNOTSUPP
;
528 DECLARE_SOCKADDR(struct sockaddr_in
*, sin
, msg
->msg_name
);
534 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
540 msg
->msg_flags
|= MSG_TRUNC
;
544 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
548 sock_recv_timestamp(msg
, sk
, skb
);
550 /* Copy the address. */
552 sin
->sin_family
= AF_INET
;
553 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
555 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
556 *addr_len
= sizeof(*sin
);
558 if (inet
->cmsg_flags
)
559 ip_cmsg_recv(msg
, skb
);
560 if (flags
& MSG_TRUNC
)
563 skb_free_datagram(sk
, skb
);
565 return err
? err
: copied
;
568 int l2tp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
575 amount
= sk_wmem_alloc_get(sk
);
578 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
579 skb
= skb_peek(&sk
->sk_receive_queue
);
580 amount
= skb
? skb
->len
: 0;
581 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
588 return put_user(amount
, (int __user
*)arg
);
590 EXPORT_SYMBOL(l2tp_ioctl
);
592 static struct proto l2tp_ip_prot
= {
594 .owner
= THIS_MODULE
,
595 .init
= l2tp_ip_open
,
596 .close
= l2tp_ip_close
,
597 .bind
= l2tp_ip_bind
,
598 .connect
= l2tp_ip_connect
,
599 .disconnect
= l2tp_ip_disconnect
,
601 .destroy
= l2tp_ip_destroy_sock
,
602 .setsockopt
= ip_setsockopt
,
603 .getsockopt
= ip_getsockopt
,
604 .sendmsg
= l2tp_ip_sendmsg
,
605 .recvmsg
= l2tp_ip_recvmsg
,
606 .backlog_rcv
= l2tp_ip_backlog_recv
,
608 .unhash
= inet_unhash
,
609 .obj_size
= sizeof(struct l2tp_ip_sock
),
611 .compat_setsockopt
= compat_ip_setsockopt
,
612 .compat_getsockopt
= compat_ip_getsockopt
,
616 static const struct proto_ops l2tp_ip_ops
= {
618 .owner
= THIS_MODULE
,
619 .release
= inet_release
,
621 .connect
= inet_dgram_connect
,
622 .socketpair
= sock_no_socketpair
,
623 .accept
= sock_no_accept
,
624 .getname
= l2tp_ip_getname
,
625 .poll
= datagram_poll
,
627 .listen
= sock_no_listen
,
628 .shutdown
= inet_shutdown
,
629 .setsockopt
= sock_common_setsockopt
,
630 .getsockopt
= sock_common_getsockopt
,
631 .sendmsg
= inet_sendmsg
,
632 .recvmsg
= sock_common_recvmsg
,
633 .mmap
= sock_no_mmap
,
634 .sendpage
= sock_no_sendpage
,
636 .compat_setsockopt
= compat_sock_common_setsockopt
,
637 .compat_getsockopt
= compat_sock_common_getsockopt
,
641 static struct inet_protosw l2tp_ip_protosw
= {
643 .protocol
= IPPROTO_L2TP
,
644 .prot
= &l2tp_ip_prot
,
648 static struct net_protocol l2tp_ip_protocol __read_mostly
= {
649 .handler
= l2tp_ip_recv
,
653 static int __init
l2tp_ip_init(void)
657 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
659 err
= proto_register(&l2tp_ip_prot
, 1);
663 err
= inet_add_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
667 inet_register_protosw(&l2tp_ip_protosw
);
671 proto_unregister(&l2tp_ip_prot
);
676 static void __exit
l2tp_ip_exit(void)
678 inet_unregister_protosw(&l2tp_ip_protosw
);
679 inet_del_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
680 proto_unregister(&l2tp_ip_prot
);
683 module_init(l2tp_ip_init
);
684 module_exit(l2tp_ip_exit
);
686 MODULE_LICENSE("GPL");
687 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
688 MODULE_DESCRIPTION("L2TP over IP");
689 MODULE_VERSION("1.0");
691 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
694 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET
, 2, IPPROTO_L2TP
);
695 MODULE_ALIAS_NET_PF_PROTO(PF_INET
, IPPROTO_L2TP
);