vmxnet3: Fix inconsistent LRO state after initialization
[linux-2.6/linux-mips.git] / net / l2tp / l2tp_ip.c
blobfce9bd3bd3feb61c3b6a2fbc078d9a5b8f7462c5
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 #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>
18 #include <linux/in.h>
19 #include <net/sock.h>
20 #include <net/ip.h>
21 #include <net/icmp.h>
22 #include <net/udp.h>
23 #include <net/inet_common.h>
24 #include <net/inet_hashtables.h>
25 #include <net/tcp_states.h>
26 #include <net/protocol.h>
27 #include <net/xfrm.h>
29 #include "l2tp_core.h"
31 struct l2tp_ip_sock {
32 /* inet_sock has to be the first member of l2tp_ip_sock */
33 struct inet_sock inet;
35 __u32 conn_id;
36 __u32 peer_conn_id;
38 __u64 tx_packets;
39 __u64 tx_bytes;
40 __u64 tx_errors;
41 __u64 rx_packets;
42 __u64 rx_bytes;
43 __u64 rx_errors;
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;
58 struct sock *sk;
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);
64 if (l2tp == NULL)
65 continue;
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))
71 goto found;
74 sk = NULL;
75 found:
76 return sk;
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);
82 if (sk)
83 sock_hold(sk);
85 return sk;
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
95 * 0 1 2 3
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 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
98 * | Session ID |
99 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100 * | Cookie (optional, maximum 64 bits)...
101 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105 * L2TPv3 Control Message Header Over IP
107 * 0 1 2 3
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 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
116 * | Ns | Nr |
117 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * All control frames are passed to userspace.
121 static int l2tp_ip_recv(struct sk_buff *skb)
123 struct sock *sk;
124 u32 session_id;
125 u32 tunnel_id;
126 unsigned char *ptr, *optr;
127 struct l2tp_session *session;
128 struct l2tp_tunnel *tunnel = NULL;
129 int length;
130 int offset;
132 /* Point to L2TP header */
133 optr = ptr = skb->data;
135 if (!pskb_may_pull(skb, 4))
136 goto discard;
138 session_id = ntohl(*((__be32 *) ptr));
139 ptr += 4;
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) {
146 __skb_pull(skb, 4);
147 goto pass_up;
150 /* Ok, this is a data packet. Lookup the session. */
151 session = l2tp_session_find(&init_net, NULL, session_id);
152 if (session == NULL)
153 goto discard;
155 tunnel = session->tunnel;
156 if (tunnel == NULL)
157 goto discard;
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))
163 goto discard;
165 printk(KERN_DEBUG "%s: ip recv: ", tunnel->name);
167 offset = 0;
168 do {
169 printk(" %02X", ptr[offset]);
170 } while (++offset < length);
172 printk("\n");
175 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
177 return 0;
179 pass_up:
180 /* Get the tunnel_id from the L2TP header */
181 if (!pskb_may_pull(skb, 12))
182 goto discard;
184 if ((skb->data[0] & 0xc0) != 0xc0)
185 goto discard;
187 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
188 tunnel = l2tp_tunnel_find(&init_net, tunnel_id);
189 if (tunnel != NULL)
190 sk = tunnel->sock;
191 else {
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);
199 if (sk == NULL)
200 goto discard;
202 sock_hold(sk);
204 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
205 goto discard_put;
207 nf_reset(skb);
209 return sk_receive_skb(sk, skb, 1);
211 discard_put:
212 sock_put(sk);
214 discard:
215 kfree_skb(skb);
216 return 0;
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);
228 return 0;
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)
242 struct sk_buff *skb;
244 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
245 kfree_skb(skb);
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;
254 int ret = -EINVAL;
255 int chk_addr_ret;
257 ret = -EADDRINUSE;
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))
260 goto out_in_use;
262 read_unlock_bh(&l2tp_ip_lock);
264 lock_sock(sk);
265 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
266 goto out;
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)
272 goto out;
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 */
277 sk_dst_reset(sk);
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);
285 ret = 0;
286 out:
287 release_sock(sk);
289 return ret;
291 out_in_use:
292 read_unlock_bh(&l2tp_ip_lock);
294 return ret;
297 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
299 int rc;
300 struct inet_sock *inet = inet_sk(sk);
301 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
302 struct rtable *rt;
303 __be32 saddr;
304 int oif;
306 rc = -EINVAL;
307 if (addr_len < sizeof(*lsa))
308 goto out;
310 rc = -EAFNOSUPPORT;
311 if (lsa->l2tp_family != AF_INET)
312 goto out;
314 sk_dst_reset(sk);
316 oif = sk->sk_bound_dev_if;
317 saddr = inet->inet_saddr;
319 rc = -EINVAL;
320 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
321 goto out;
323 rt = ip_route_connect(lsa->l2tp_addr.s_addr, saddr,
324 RT_CONN_FLAGS(sk), oif,
325 IPPROTO_L2TP,
326 0, 0, sk, true);
327 if (IS_ERR(rt)) {
328 rc = PTR_ERR(rt);
329 if (rc == -ENETUNREACH)
330 IP_INC_STATS_BH(&init_net, IPSTATS_MIB_OUTNOROUTES);
331 goto out;
334 rc = -ENETUNREACH;
335 if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
336 ip_rt_put(rt);
337 goto out;
340 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
342 if (!inet->inet_saddr)
343 inet->inet_saddr = rt->rt_src;
344 if (!inet->inet_rcv_saddr)
345 inet->inet_rcv_saddr = rt->rt_src;
346 inet->inet_daddr = rt->rt_dst;
347 sk->sk_state = TCP_ESTABLISHED;
348 inet->inet_id = jiffies;
350 sk_dst_set(sk, &rt->dst);
352 write_lock_bh(&l2tp_ip_lock);
353 hlist_del_init(&sk->sk_bind_node);
354 sk_add_bind_node(sk, &l2tp_ip_bind_table);
355 write_unlock_bh(&l2tp_ip_lock);
357 rc = 0;
358 out:
359 return rc;
362 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
363 int *uaddr_len, int peer)
365 struct sock *sk = sock->sk;
366 struct inet_sock *inet = inet_sk(sk);
367 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
368 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
370 memset(lsa, 0, sizeof(*lsa));
371 lsa->l2tp_family = AF_INET;
372 if (peer) {
373 if (!inet->inet_dport)
374 return -ENOTCONN;
375 lsa->l2tp_conn_id = lsk->peer_conn_id;
376 lsa->l2tp_addr.s_addr = inet->inet_daddr;
377 } else {
378 __be32 addr = inet->inet_rcv_saddr;
379 if (!addr)
380 addr = inet->inet_saddr;
381 lsa->l2tp_conn_id = lsk->conn_id;
382 lsa->l2tp_addr.s_addr = addr;
384 *uaddr_len = sizeof(*lsa);
385 return 0;
388 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
390 int rc;
392 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
393 goto drop;
395 nf_reset(skb);
397 /* Charge it to the socket, dropping if the queue is full. */
398 rc = sock_queue_rcv_skb(sk, skb);
399 if (rc < 0)
400 goto drop;
402 return 0;
404 drop:
405 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
406 kfree_skb(skb);
407 return -1;
410 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
411 * control frames.
413 static int l2tp_ip_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, size_t len)
415 struct sk_buff *skb;
416 int rc;
417 struct l2tp_ip_sock *lsa = l2tp_ip_sk(sk);
418 struct inet_sock *inet = inet_sk(sk);
419 struct ip_options *opt = inet->opt;
420 struct rtable *rt = NULL;
421 int connected = 0;
422 __be32 daddr;
424 if (sock_flag(sk, SOCK_DEAD))
425 return -ENOTCONN;
427 /* Get and verify the address. */
428 if (msg->msg_name) {
429 struct sockaddr_l2tpip *lip = (struct sockaddr_l2tpip *) msg->msg_name;
430 if (msg->msg_namelen < sizeof(*lip))
431 return -EINVAL;
433 if (lip->l2tp_family != AF_INET) {
434 if (lip->l2tp_family != AF_UNSPEC)
435 return -EAFNOSUPPORT;
438 daddr = lip->l2tp_addr.s_addr;
439 } else {
440 if (sk->sk_state != TCP_ESTABLISHED)
441 return -EDESTADDRREQ;
443 daddr = inet->inet_daddr;
444 connected = 1;
447 /* Allocate a socket buffer */
448 rc = -ENOMEM;
449 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
450 4 + len, 0, GFP_KERNEL);
451 if (!skb)
452 goto error;
454 /* Reserve space for headers, putting IP header on 4-byte boundary. */
455 skb_reserve(skb, 2 + NET_SKB_PAD);
456 skb_reset_network_header(skb);
457 skb_reserve(skb, sizeof(struct iphdr));
458 skb_reset_transport_header(skb);
460 /* Insert 0 session_id */
461 *((__be32 *) skb_put(skb, 4)) = 0;
463 /* Copy user data into skb */
464 rc = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
465 if (rc < 0) {
466 kfree_skb(skb);
467 goto error;
470 if (connected)
471 rt = (struct rtable *) __sk_dst_check(sk, 0);
473 if (rt == NULL) {
474 /* Use correct destination address if we have options. */
475 if (opt && opt->srr)
476 daddr = opt->faddr;
478 /* If this fails, retransmit mechanism of transport layer will
479 * keep trying until route appears or the connection times
480 * itself out.
482 rt = ip_route_output_ports(sock_net(sk), sk,
483 daddr, inet->inet_saddr,
484 inet->inet_dport, inet->inet_sport,
485 sk->sk_protocol, RT_CONN_FLAGS(sk),
486 sk->sk_bound_dev_if);
487 if (IS_ERR(rt))
488 goto no_route;
489 sk_setup_caps(sk, &rt->dst);
491 skb_dst_set(skb, dst_clone(&rt->dst));
493 /* Queue the packet to IP for output */
494 rc = ip_queue_xmit(skb);
496 error:
497 /* Update stats */
498 if (rc >= 0) {
499 lsa->tx_packets++;
500 lsa->tx_bytes += len;
501 rc = len;
502 } else {
503 lsa->tx_errors++;
506 return rc;
508 no_route:
509 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
510 kfree_skb(skb);
511 return -EHOSTUNREACH;
514 static int l2tp_ip_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
515 size_t len, int noblock, int flags, int *addr_len)
517 struct inet_sock *inet = inet_sk(sk);
518 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
519 size_t copied = 0;
520 int err = -EOPNOTSUPP;
521 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
522 struct sk_buff *skb;
524 if (flags & MSG_OOB)
525 goto out;
527 if (addr_len)
528 *addr_len = sizeof(*sin);
530 skb = skb_recv_datagram(sk, flags, noblock, &err);
531 if (!skb)
532 goto out;
534 copied = skb->len;
535 if (len < copied) {
536 msg->msg_flags |= MSG_TRUNC;
537 copied = len;
540 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
541 if (err)
542 goto done;
544 sock_recv_timestamp(msg, sk, skb);
546 /* Copy the address. */
547 if (sin) {
548 sin->sin_family = AF_INET;
549 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
550 sin->sin_port = 0;
551 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
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 if (err) {
561 lsk->rx_errors++;
562 return err;
565 lsk->rx_packets++;
566 lsk->rx_bytes += copied;
568 return copied;
571 static struct proto l2tp_ip_prot = {
572 .name = "L2TP/IP",
573 .owner = THIS_MODULE,
574 .init = l2tp_ip_open,
575 .close = l2tp_ip_close,
576 .bind = l2tp_ip_bind,
577 .connect = l2tp_ip_connect,
578 .disconnect = udp_disconnect,
579 .ioctl = udp_ioctl,
580 .destroy = l2tp_ip_destroy_sock,
581 .setsockopt = ip_setsockopt,
582 .getsockopt = ip_getsockopt,
583 .sendmsg = l2tp_ip_sendmsg,
584 .recvmsg = l2tp_ip_recvmsg,
585 .backlog_rcv = l2tp_ip_backlog_recv,
586 .hash = inet_hash,
587 .unhash = inet_unhash,
588 .obj_size = sizeof(struct l2tp_ip_sock),
589 #ifdef CONFIG_COMPAT
590 .compat_setsockopt = compat_ip_setsockopt,
591 .compat_getsockopt = compat_ip_getsockopt,
592 #endif
595 static const struct proto_ops l2tp_ip_ops = {
596 .family = PF_INET,
597 .owner = THIS_MODULE,
598 .release = inet_release,
599 .bind = inet_bind,
600 .connect = inet_dgram_connect,
601 .socketpair = sock_no_socketpair,
602 .accept = sock_no_accept,
603 .getname = l2tp_ip_getname,
604 .poll = datagram_poll,
605 .ioctl = inet_ioctl,
606 .listen = sock_no_listen,
607 .shutdown = inet_shutdown,
608 .setsockopt = sock_common_setsockopt,
609 .getsockopt = sock_common_getsockopt,
610 .sendmsg = inet_sendmsg,
611 .recvmsg = sock_common_recvmsg,
612 .mmap = sock_no_mmap,
613 .sendpage = sock_no_sendpage,
614 #ifdef CONFIG_COMPAT
615 .compat_setsockopt = compat_sock_common_setsockopt,
616 .compat_getsockopt = compat_sock_common_getsockopt,
617 #endif
620 static struct inet_protosw l2tp_ip_protosw = {
621 .type = SOCK_DGRAM,
622 .protocol = IPPROTO_L2TP,
623 .prot = &l2tp_ip_prot,
624 .ops = &l2tp_ip_ops,
625 .no_check = 0,
628 static struct net_protocol l2tp_ip_protocol __read_mostly = {
629 .handler = l2tp_ip_recv,
632 static int __init l2tp_ip_init(void)
634 int err;
636 printk(KERN_INFO "L2TP IP encapsulation support (L2TPv3)\n");
638 err = proto_register(&l2tp_ip_prot, 1);
639 if (err != 0)
640 goto out;
642 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
643 if (err)
644 goto out1;
646 inet_register_protosw(&l2tp_ip_protosw);
647 return 0;
649 out1:
650 proto_unregister(&l2tp_ip_prot);
651 out:
652 return err;
655 static void __exit l2tp_ip_exit(void)
657 inet_unregister_protosw(&l2tp_ip_protosw);
658 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
659 proto_unregister(&l2tp_ip_prot);
662 module_init(l2tp_ip_init);
663 module_exit(l2tp_ip_exit);
665 MODULE_LICENSE("GPL");
666 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
667 MODULE_DESCRIPTION("L2TP over IP");
668 MODULE_VERSION("1.0");
670 /* Use the value of SOCK_DGRAM (2) directory, because __stringify does't like
671 * enums
673 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);