mlxsw: spectrum: Implement TC flower offload
[linux-2.6/btrfs-unstable.git] / net / l2tp / l2tp_ip.c
blobb07a859f21bd3f96999aa9c58f09c0f57667e90b
1 /*
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>
20 #include <linux/in.h>
21 #include <net/sock.h>
22 #include <net/ip.h>
23 #include <net/icmp.h>
24 #include <net/udp.h>
25 #include <net/inet_common.h>
26 #include <net/inet_hashtables.h>
27 #include <net/tcp_states.h>
28 #include <net/protocol.h>
29 #include <net/xfrm.h>
31 #include "l2tp_core.h"
33 struct l2tp_ip_sock {
34 /* inet_sock has to be the first member of l2tp_ip_sock */
35 struct inet_sock inet;
37 u32 conn_id;
38 u32 peer_conn_id;
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(const struct net *net, __be32 laddr,
51 __be32 raddr, int dif, u32 tunnel_id)
53 struct sock *sk;
55 sk_for_each_bound(sk, &l2tp_ip_bind_table) {
56 const struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
57 const struct inet_sock *inet = inet_sk(sk);
59 if (!net_eq(sock_net(sk), net))
60 continue;
62 if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
63 continue;
65 if (inet->inet_rcv_saddr && laddr &&
66 inet->inet_rcv_saddr != laddr)
67 continue;
69 if (inet->inet_daddr && raddr && inet->inet_daddr != raddr)
70 continue;
72 if (l2tp->conn_id != tunnel_id)
73 continue;
75 goto found;
78 sk = NULL;
79 found:
80 return sk;
83 /* When processing receive frames, there are two cases to
84 * consider. Data frames consist of a non-zero session-id and an
85 * optional cookie. Control frames consist of a regular L2TP header
86 * preceded by 32-bits of zeros.
88 * L2TPv3 Session Header Over IP
90 * 0 1 2 3
91 * 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
92 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
93 * | Session ID |
94 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
95 * | Cookie (optional, maximum 64 bits)...
96 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
97 * |
98 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100 * L2TPv3 Control Message Header Over IP
102 * 0 1 2 3
103 * 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
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105 * | (32 bits of zeros) |
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
108 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109 * | Control Connection ID |
110 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111 * | Ns | Nr |
112 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114 * All control frames are passed to userspace.
116 static int l2tp_ip_recv(struct sk_buff *skb)
118 struct net *net = dev_net(skb->dev);
119 struct sock *sk;
120 u32 session_id;
121 u32 tunnel_id;
122 unsigned char *ptr, *optr;
123 struct l2tp_session *session;
124 struct l2tp_tunnel *tunnel = NULL;
125 int length;
127 if (!pskb_may_pull(skb, 4))
128 goto discard;
130 /* Point to L2TP header */
131 optr = ptr = skb->data;
132 session_id = ntohl(*((__be32 *) ptr));
133 ptr += 4;
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) {
140 __skb_pull(skb, 4);
141 goto pass_up;
144 /* Ok, this is a data packet. Lookup the session. */
145 session = l2tp_session_find(net, NULL, session_id);
146 if (session == NULL)
147 goto discard;
149 tunnel = session->tunnel;
150 if (tunnel == NULL)
151 goto discard;
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))
157 goto discard;
159 /* Point to L2TP header */
160 optr = ptr = skb->data;
161 ptr += 4;
162 pr_debug("%s: ip recv\n", tunnel->name);
163 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
166 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
168 return 0;
170 pass_up:
171 /* Get the tunnel_id from the L2TP header */
172 if (!pskb_may_pull(skb, 12))
173 goto discard;
175 if ((skb->data[0] & 0xc0) != 0xc0)
176 goto discard;
178 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
179 tunnel = l2tp_tunnel_find(net, tunnel_id);
180 if (tunnel != NULL)
181 sk = tunnel->sock;
182 else {
183 struct iphdr *iph = (struct iphdr *) skb_network_header(skb);
185 read_lock_bh(&l2tp_ip_lock);
186 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr,
187 inet_iif(skb), tunnel_id);
188 if (!sk) {
189 read_unlock_bh(&l2tp_ip_lock);
190 goto discard;
193 sock_hold(sk);
194 read_unlock_bh(&l2tp_ip_lock);
197 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
198 goto discard_put;
200 nf_reset(skb);
202 return sk_receive_skb(sk, skb, 1);
204 discard_put:
205 sock_put(sk);
207 discard:
208 kfree_skb(skb);
209 return 0;
212 static int l2tp_ip_open(struct sock *sk)
214 /* Prevent autobind. We don't have ports. */
215 inet_sk(sk)->inet_num = IPPROTO_L2TP;
217 write_lock_bh(&l2tp_ip_lock);
218 sk_add_node(sk, &l2tp_ip_table);
219 write_unlock_bh(&l2tp_ip_lock);
221 return 0;
224 static void l2tp_ip_close(struct sock *sk, long timeout)
226 write_lock_bh(&l2tp_ip_lock);
227 hlist_del_init(&sk->sk_bind_node);
228 sk_del_node_init(sk);
229 write_unlock_bh(&l2tp_ip_lock);
230 sk_common_release(sk);
233 static void l2tp_ip_destroy_sock(struct sock *sk)
235 struct sk_buff *skb;
236 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
238 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
239 kfree_skb(skb);
241 if (tunnel) {
242 l2tp_tunnel_closeall(tunnel);
243 sock_put(sk);
246 sk_refcnt_debug_dec(sk);
249 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
251 struct inet_sock *inet = inet_sk(sk);
252 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
253 struct net *net = sock_net(sk);
254 int ret;
255 int chk_addr_ret;
257 if (addr_len < sizeof(struct sockaddr_l2tpip))
258 return -EINVAL;
259 if (addr->l2tp_family != AF_INET)
260 return -EINVAL;
262 lock_sock(sk);
264 ret = -EINVAL;
265 if (!sock_flag(sk, SOCK_ZAPPED))
266 goto out;
268 if (sk->sk_state != TCP_CLOSE)
269 goto out;
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)
275 goto out;
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 */
282 write_lock_bh(&l2tp_ip_lock);
283 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
284 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
285 write_unlock_bh(&l2tp_ip_lock);
286 ret = -EADDRINUSE;
287 goto out;
290 sk_dst_reset(sk);
291 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
293 sk_add_bind_node(sk, &l2tp_ip_bind_table);
294 sk_del_node_init(sk);
295 write_unlock_bh(&l2tp_ip_lock);
297 ret = 0;
298 sock_reset_flag(sk, SOCK_ZAPPED);
300 out:
301 release_sock(sk);
303 return ret;
306 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
308 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
309 int rc;
311 if (addr_len < sizeof(*lsa))
312 return -EINVAL;
314 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
315 return -EINVAL;
317 lock_sock(sk);
319 /* Must bind first - autobinding does not work */
320 if (sock_flag(sk, SOCK_ZAPPED)) {
321 rc = -EINVAL;
322 goto out_sk;
325 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
326 if (rc < 0)
327 goto out_sk;
329 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
331 write_lock_bh(&l2tp_ip_lock);
332 hlist_del_init(&sk->sk_bind_node);
333 sk_add_bind_node(sk, &l2tp_ip_bind_table);
334 write_unlock_bh(&l2tp_ip_lock);
336 out_sk:
337 release_sock(sk);
339 return rc;
342 static int l2tp_ip_disconnect(struct sock *sk, int flags)
344 if (sock_flag(sk, SOCK_ZAPPED))
345 return 0;
347 return __udp_disconnect(sk, flags);
350 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
351 int *uaddr_len, int peer)
353 struct sock *sk = sock->sk;
354 struct inet_sock *inet = inet_sk(sk);
355 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
356 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
358 memset(lsa, 0, sizeof(*lsa));
359 lsa->l2tp_family = AF_INET;
360 if (peer) {
361 if (!inet->inet_dport)
362 return -ENOTCONN;
363 lsa->l2tp_conn_id = lsk->peer_conn_id;
364 lsa->l2tp_addr.s_addr = inet->inet_daddr;
365 } else {
366 __be32 addr = inet->inet_rcv_saddr;
367 if (!addr)
368 addr = inet->inet_saddr;
369 lsa->l2tp_conn_id = lsk->conn_id;
370 lsa->l2tp_addr.s_addr = addr;
372 *uaddr_len = sizeof(*lsa);
373 return 0;
376 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
378 int rc;
380 /* Charge it to the socket, dropping if the queue is full. */
381 rc = sock_queue_rcv_skb(sk, skb);
382 if (rc < 0)
383 goto drop;
385 return 0;
387 drop:
388 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
389 kfree_skb(skb);
390 return -1;
393 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
394 * control frames.
396 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
398 struct sk_buff *skb;
399 int rc;
400 struct inet_sock *inet = inet_sk(sk);
401 struct rtable *rt = NULL;
402 struct flowi4 *fl4;
403 int connected = 0;
404 __be32 daddr;
406 lock_sock(sk);
408 rc = -ENOTCONN;
409 if (sock_flag(sk, SOCK_DEAD))
410 goto out;
412 /* Get and verify the address. */
413 if (msg->msg_name) {
414 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
415 rc = -EINVAL;
416 if (msg->msg_namelen < sizeof(*lip))
417 goto out;
419 if (lip->l2tp_family != AF_INET) {
420 rc = -EAFNOSUPPORT;
421 if (lip->l2tp_family != AF_UNSPEC)
422 goto out;
425 daddr = lip->l2tp_addr.s_addr;
426 } else {
427 rc = -EDESTADDRREQ;
428 if (sk->sk_state != TCP_ESTABLISHED)
429 goto out;
431 daddr = inet->inet_daddr;
432 connected = 1;
435 /* Allocate a socket buffer */
436 rc = -ENOMEM;
437 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
438 4 + len, 0, GFP_KERNEL);
439 if (!skb)
440 goto error;
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);
453 if (rc < 0) {
454 kfree_skb(skb);
455 goto error;
458 fl4 = &inet->cork.fl.u.ip4;
459 if (connected)
460 rt = (struct rtable *) __sk_dst_check(sk, 0);
462 rcu_read_lock();
463 if (rt == NULL) {
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
474 * itself out.
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);
481 if (IS_ERR(rt))
482 goto no_route;
483 if (connected) {
484 sk_setup_caps(sk, &rt->dst);
485 } else {
486 skb_dst_set(skb, &rt->dst);
487 goto xmit;
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);
496 xmit:
497 /* Queue the packet to IP for output */
498 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
499 rcu_read_unlock();
501 error:
502 if (rc >= 0)
503 rc = len;
505 out:
506 release_sock(sk);
507 return rc;
509 no_route:
510 rcu_read_unlock();
511 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
512 kfree_skb(skb);
513 rc = -EHOSTUNREACH;
514 goto out;
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);
521 size_t copied = 0;
522 int err = -EOPNOTSUPP;
523 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
524 struct sk_buff *skb;
526 if (flags & MSG_OOB)
527 goto out;
529 skb = skb_recv_datagram(sk, flags, noblock, &err);
530 if (!skb)
531 goto out;
533 copied = skb->len;
534 if (len < copied) {
535 msg->msg_flags |= MSG_TRUNC;
536 copied = len;
539 err = skb_copy_datagram_msg(skb, 0, msg, copied);
540 if (err)
541 goto done;
543 sock_recv_timestamp(msg, sk, skb);
545 /* Copy the address. */
546 if (sin) {
547 sin->sin_family = AF_INET;
548 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
549 sin->sin_port = 0;
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)
556 copied = skb->len;
557 done:
558 skb_free_datagram(sk, skb);
559 out:
560 return err ? err : copied;
563 static struct proto l2tp_ip_prot = {
564 .name = "L2TP/IP",
565 .owner = THIS_MODULE,
566 .init = l2tp_ip_open,
567 .close = l2tp_ip_close,
568 .bind = l2tp_ip_bind,
569 .connect = l2tp_ip_connect,
570 .disconnect = l2tp_ip_disconnect,
571 .ioctl = udp_ioctl,
572 .destroy = l2tp_ip_destroy_sock,
573 .setsockopt = ip_setsockopt,
574 .getsockopt = ip_getsockopt,
575 .sendmsg = l2tp_ip_sendmsg,
576 .recvmsg = l2tp_ip_recvmsg,
577 .backlog_rcv = l2tp_ip_backlog_recv,
578 .hash = inet_hash,
579 .unhash = inet_unhash,
580 .obj_size = sizeof(struct l2tp_ip_sock),
581 #ifdef CONFIG_COMPAT
582 .compat_setsockopt = compat_ip_setsockopt,
583 .compat_getsockopt = compat_ip_getsockopt,
584 #endif
587 static const struct proto_ops l2tp_ip_ops = {
588 .family = PF_INET,
589 .owner = THIS_MODULE,
590 .release = inet_release,
591 .bind = inet_bind,
592 .connect = inet_dgram_connect,
593 .socketpair = sock_no_socketpair,
594 .accept = sock_no_accept,
595 .getname = l2tp_ip_getname,
596 .poll = datagram_poll,
597 .ioctl = inet_ioctl,
598 .listen = sock_no_listen,
599 .shutdown = inet_shutdown,
600 .setsockopt = sock_common_setsockopt,
601 .getsockopt = sock_common_getsockopt,
602 .sendmsg = inet_sendmsg,
603 .recvmsg = sock_common_recvmsg,
604 .mmap = sock_no_mmap,
605 .sendpage = sock_no_sendpage,
606 #ifdef CONFIG_COMPAT
607 .compat_setsockopt = compat_sock_common_setsockopt,
608 .compat_getsockopt = compat_sock_common_getsockopt,
609 #endif
612 static struct inet_protosw l2tp_ip_protosw = {
613 .type = SOCK_DGRAM,
614 .protocol = IPPROTO_L2TP,
615 .prot = &l2tp_ip_prot,
616 .ops = &l2tp_ip_ops,
619 static struct net_protocol l2tp_ip_protocol __read_mostly = {
620 .handler = l2tp_ip_recv,
621 .netns_ok = 1,
624 static int __init l2tp_ip_init(void)
626 int err;
628 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
630 err = proto_register(&l2tp_ip_prot, 1);
631 if (err != 0)
632 goto out;
634 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
635 if (err)
636 goto out1;
638 inet_register_protosw(&l2tp_ip_protosw);
639 return 0;
641 out1:
642 proto_unregister(&l2tp_ip_prot);
643 out:
644 return err;
647 static void __exit l2tp_ip_exit(void)
649 inet_unregister_protosw(&l2tp_ip_protosw);
650 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
651 proto_unregister(&l2tp_ip_prot);
654 module_init(l2tp_ip_init);
655 module_exit(l2tp_ip_exit);
657 MODULE_LICENSE("GPL");
658 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
659 MODULE_DESCRIPTION("L2TP over IP");
660 MODULE_VERSION("1.0");
662 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
663 * enums
665 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
666 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);