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
;
122 if (!pskb_may_pull(skb
, 4))
125 /* Point to L2TP header */
128 session_id
= ntohl(*((__be32
*)ptr
));
131 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
132 * the session_id. If it is 0, the packet is a L2TP control
133 * frame and the session_id value can be discarded.
135 if (session_id
== 0) {
140 /* Ok, this is a data packet. Lookup the session. */
141 session
= l2tp_session_get(net
, session_id
);
145 tunnel
= session
->tunnel
;
149 if (l2tp_v3_ensure_opt_in_linear(session
, skb
, &ptr
, &optr
))
152 l2tp_recv_common(session
, skb
, ptr
, optr
, 0, skb
->len
);
153 l2tp_session_dec_refcount(session
);
158 /* Get the tunnel_id from the L2TP header */
159 if (!pskb_may_pull(skb
, 12))
162 if ((skb
->data
[0] & 0xc0) != 0xc0)
165 tunnel_id
= ntohl(*(__be32
*)&skb
->data
[4]);
166 iph
= (struct iphdr
*)skb_network_header(skb
);
168 read_lock_bh(&l2tp_ip_lock
);
169 sk
= __l2tp_ip_bind_lookup(net
, iph
->daddr
, iph
->saddr
, inet_iif(skb
),
172 read_unlock_bh(&l2tp_ip_lock
);
176 read_unlock_bh(&l2tp_ip_lock
);
178 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
183 return sk_receive_skb(sk
, skb
, 1);
186 l2tp_session_dec_refcount(session
);
197 static int l2tp_ip_hash(struct sock
*sk
)
199 if (sk_unhashed(sk
)) {
200 write_lock_bh(&l2tp_ip_lock
);
201 sk_add_node(sk
, &l2tp_ip_table
);
202 write_unlock_bh(&l2tp_ip_lock
);
207 static void l2tp_ip_unhash(struct sock
*sk
)
211 write_lock_bh(&l2tp_ip_lock
);
212 sk_del_node_init(sk
);
213 write_unlock_bh(&l2tp_ip_lock
);
216 static int l2tp_ip_open(struct sock
*sk
)
218 /* Prevent autobind. We don't have ports. */
219 inet_sk(sk
)->inet_num
= IPPROTO_L2TP
;
225 static void l2tp_ip_close(struct sock
*sk
, long timeout
)
227 write_lock_bh(&l2tp_ip_lock
);
228 hlist_del_init(&sk
->sk_bind_node
);
229 sk_del_node_init(sk
);
230 write_unlock_bh(&l2tp_ip_lock
);
231 sk_common_release(sk
);
234 static void l2tp_ip_destroy_sock(struct sock
*sk
)
236 struct l2tp_tunnel
*tunnel
= l2tp_sk_to_tunnel(sk
);
239 while ((skb
= __skb_dequeue_tail(&sk
->sk_write_queue
)) != NULL
)
243 l2tp_tunnel_delete(tunnel
);
246 static int l2tp_ip_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
248 struct inet_sock
*inet
= inet_sk(sk
);
249 struct sockaddr_l2tpip
*addr
= (struct sockaddr_l2tpip
*)uaddr
;
250 struct net
*net
= sock_net(sk
);
254 if (addr_len
< sizeof(struct sockaddr_l2tpip
))
256 if (addr
->l2tp_family
!= AF_INET
)
262 if (!sock_flag(sk
, SOCK_ZAPPED
))
265 if (sk
->sk_state
!= TCP_CLOSE
)
268 chk_addr_ret
= inet_addr_type(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 if (addr
->l2tp_addr
.s_addr
) {
275 inet
->inet_rcv_saddr
= addr
->l2tp_addr
.s_addr
;
276 inet
->inet_saddr
= addr
->l2tp_addr
.s_addr
;
278 if (chk_addr_ret
== RTN_MULTICAST
|| chk_addr_ret
== RTN_BROADCAST
)
279 inet
->inet_saddr
= 0; /* Use device */
281 write_lock_bh(&l2tp_ip_lock
);
282 if (__l2tp_ip_bind_lookup(net
, addr
->l2tp_addr
.s_addr
, 0,
283 sk
->sk_bound_dev_if
, addr
->l2tp_conn_id
)) {
284 write_unlock_bh(&l2tp_ip_lock
);
290 l2tp_ip_sk(sk
)->conn_id
= addr
->l2tp_conn_id
;
292 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
293 sk_del_node_init(sk
);
294 write_unlock_bh(&l2tp_ip_lock
);
297 sock_reset_flag(sk
, SOCK_ZAPPED
);
305 static int l2tp_ip_connect(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
307 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*)uaddr
;
310 if (addr_len
< sizeof(*lsa
))
313 if (ipv4_is_multicast(lsa
->l2tp_addr
.s_addr
))
318 /* Must bind first - autobinding does not work */
319 if (sock_flag(sk
, SOCK_ZAPPED
)) {
324 rc
= __ip4_datagram_connect(sk
, uaddr
, addr_len
);
328 l2tp_ip_sk(sk
)->peer_conn_id
= lsa
->l2tp_conn_id
;
330 write_lock_bh(&l2tp_ip_lock
);
331 hlist_del_init(&sk
->sk_bind_node
);
332 sk_add_bind_node(sk
, &l2tp_ip_bind_table
);
333 write_unlock_bh(&l2tp_ip_lock
);
341 static int l2tp_ip_disconnect(struct sock
*sk
, int flags
)
343 if (sock_flag(sk
, SOCK_ZAPPED
))
346 return __udp_disconnect(sk
, flags
);
349 static int l2tp_ip_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
352 struct sock
*sk
= sock
->sk
;
353 struct inet_sock
*inet
= inet_sk(sk
);
354 struct l2tp_ip_sock
*lsk
= l2tp_ip_sk(sk
);
355 struct sockaddr_l2tpip
*lsa
= (struct sockaddr_l2tpip
*)uaddr
;
357 memset(lsa
, 0, sizeof(*lsa
));
358 lsa
->l2tp_family
= AF_INET
;
360 if (!inet
->inet_dport
)
362 lsa
->l2tp_conn_id
= lsk
->peer_conn_id
;
363 lsa
->l2tp_addr
.s_addr
= inet
->inet_daddr
;
365 __be32 addr
= inet
->inet_rcv_saddr
;
368 addr
= inet
->inet_saddr
;
369 lsa
->l2tp_conn_id
= lsk
->conn_id
;
370 lsa
->l2tp_addr
.s_addr
= addr
;
375 static int l2tp_ip_backlog_recv(struct sock
*sk
, struct sk_buff
*skb
)
379 /* Charge it to the socket, dropping if the queue is full. */
380 rc
= sock_queue_rcv_skb(sk
, skb
);
387 IP_INC_STATS(sock_net(sk
), IPSTATS_MIB_INDISCARDS
);
392 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
395 static int l2tp_ip_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
)
399 struct inet_sock
*inet
= inet_sk(sk
);
400 struct rtable
*rt
= NULL
;
408 if (sock_flag(sk
, SOCK_DEAD
))
411 /* Get and verify the address. */
413 DECLARE_SOCKADDR(struct sockaddr_l2tpip
*, lip
, msg
->msg_name
);
416 if (msg
->msg_namelen
< sizeof(*lip
))
419 if (lip
->l2tp_family
!= AF_INET
) {
421 if (lip
->l2tp_family
!= AF_UNSPEC
)
425 daddr
= lip
->l2tp_addr
.s_addr
;
428 if (sk
->sk_state
!= TCP_ESTABLISHED
)
431 daddr
= inet
->inet_daddr
;
435 /* Allocate a socket buffer */
437 skb
= sock_wmalloc(sk
, 2 + NET_SKB_PAD
+ sizeof(struct iphdr
) +
438 4 + len
, 0, GFP_KERNEL
);
442 /* Reserve space for headers, putting IP header on 4-byte boundary. */
443 skb_reserve(skb
, 2 + NET_SKB_PAD
);
444 skb_reset_network_header(skb
);
445 skb_reserve(skb
, sizeof(struct iphdr
));
446 skb_reset_transport_header(skb
);
448 /* Insert 0 session_id */
449 *((__be32
*)skb_put(skb
, 4)) = 0;
451 /* Copy user data into skb */
452 rc
= memcpy_from_msg(skb_put(skb
, len
), msg
, len
);
458 fl4
= &inet
->cork
.fl
.u
.ip4
;
460 rt
= (struct rtable
*)__sk_dst_check(sk
, 0);
464 const struct ip_options_rcu
*inet_opt
;
466 inet_opt
= rcu_dereference(inet
->inet_opt
);
468 /* Use correct destination address if we have options. */
469 if (inet_opt
&& inet_opt
->opt
.srr
)
470 daddr
= inet_opt
->opt
.faddr
;
472 /* If this fails, retransmit mechanism of transport layer will
473 * keep trying until route appears or the connection times
476 rt
= ip_route_output_ports(sock_net(sk
), fl4
, sk
,
477 daddr
, inet
->inet_saddr
,
478 inet
->inet_dport
, inet
->inet_sport
,
479 sk
->sk_protocol
, RT_CONN_FLAGS(sk
),
480 sk
->sk_bound_dev_if
);
484 sk_setup_caps(sk
, &rt
->dst
);
486 skb_dst_set(skb
, &rt
->dst
);
491 /* We dont need to clone dst here, it is guaranteed to not disappear.
492 * __dev_xmit_skb() might force a refcount if needed.
494 skb_dst_set_noref(skb
, &rt
->dst
);
497 /* Queue the packet to IP for output */
498 rc
= ip_queue_xmit(sk
, skb
, &inet
->cork
.fl
);
511 IP_INC_STATS(sock_net(sk
), IPSTATS_MIB_OUTNOROUTES
);
517 static int l2tp_ip_recvmsg(struct sock
*sk
, struct msghdr
*msg
,
518 size_t len
, int noblock
, int flags
, int *addr_len
)
520 struct inet_sock
*inet
= inet_sk(sk
);
522 int err
= -EOPNOTSUPP
;
523 DECLARE_SOCKADDR(struct sockaddr_in
*, sin
, msg
->msg_name
);
529 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
535 msg
->msg_flags
|= MSG_TRUNC
;
539 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
543 sock_recv_timestamp(msg
, sk
, skb
);
545 /* Copy the address. */
547 sin
->sin_family
= AF_INET
;
548 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
550 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
551 *addr_len
= sizeof(*sin
);
553 if (inet
->cmsg_flags
)
554 ip_cmsg_recv(msg
, skb
);
555 if (flags
& MSG_TRUNC
)
558 skb_free_datagram(sk
, skb
);
560 return err
? err
: copied
;
563 int l2tp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
570 amount
= sk_wmem_alloc_get(sk
);
573 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
574 skb
= skb_peek(&sk
->sk_receive_queue
);
575 amount
= skb
? skb
->len
: 0;
576 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
583 return put_user(amount
, (int __user
*)arg
);
585 EXPORT_SYMBOL_GPL(l2tp_ioctl
);
587 static struct proto l2tp_ip_prot
= {
589 .owner
= THIS_MODULE
,
590 .init
= l2tp_ip_open
,
591 .close
= l2tp_ip_close
,
592 .bind
= l2tp_ip_bind
,
593 .connect
= l2tp_ip_connect
,
594 .disconnect
= l2tp_ip_disconnect
,
596 .destroy
= l2tp_ip_destroy_sock
,
597 .setsockopt
= ip_setsockopt
,
598 .getsockopt
= ip_getsockopt
,
599 .sendmsg
= l2tp_ip_sendmsg
,
600 .recvmsg
= l2tp_ip_recvmsg
,
601 .backlog_rcv
= l2tp_ip_backlog_recv
,
602 .hash
= l2tp_ip_hash
,
603 .unhash
= l2tp_ip_unhash
,
604 .obj_size
= sizeof(struct l2tp_ip_sock
),
607 static const struct proto_ops l2tp_ip_ops
= {
609 .owner
= THIS_MODULE
,
610 .release
= inet_release
,
612 .connect
= inet_dgram_connect
,
613 .socketpair
= sock_no_socketpair
,
614 .accept
= sock_no_accept
,
615 .getname
= l2tp_ip_getname
,
616 .poll
= datagram_poll
,
618 .gettstamp
= sock_gettstamp
,
619 .listen
= sock_no_listen
,
620 .shutdown
= inet_shutdown
,
621 .setsockopt
= sock_common_setsockopt
,
622 .getsockopt
= sock_common_getsockopt
,
623 .sendmsg
= inet_sendmsg
,
624 .recvmsg
= sock_common_recvmsg
,
625 .mmap
= sock_no_mmap
,
626 .sendpage
= sock_no_sendpage
,
629 static struct inet_protosw l2tp_ip_protosw
= {
631 .protocol
= IPPROTO_L2TP
,
632 .prot
= &l2tp_ip_prot
,
636 static struct net_protocol l2tp_ip_protocol __read_mostly
= {
637 .handler
= l2tp_ip_recv
,
641 static int __init
l2tp_ip_init(void)
645 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
647 err
= proto_register(&l2tp_ip_prot
, 1);
651 err
= inet_add_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
655 inet_register_protosw(&l2tp_ip_protosw
);
659 proto_unregister(&l2tp_ip_prot
);
664 static void __exit
l2tp_ip_exit(void)
666 inet_unregister_protosw(&l2tp_ip_protosw
);
667 inet_del_protocol(&l2tp_ip_protocol
, IPPROTO_L2TP
);
668 proto_unregister(&l2tp_ip_prot
);
671 module_init(l2tp_ip_init
);
672 module_exit(l2tp_ip_exit
);
674 MODULE_LICENSE("GPL");
675 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
676 MODULE_DESCRIPTION("L2TP over IP");
677 MODULE_VERSION("1.0");
679 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
682 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET
, 2, IPPROTO_L2TP
);
683 MODULE_ALIAS_NET_PF_PROTO(PF_INET
, IPPROTO_L2TP
);