dm thin: wake worker when discard is prepared
[linux-2.6.git] / net / l2tp / l2tp_ip.c
blob61d8b75d2686c0272a1618887c08c34de2b4abe5
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(struct net *net, __be32 laddr, int dif, u32 tunnel_id)
52 struct hlist_node *node;
53 struct sock *sk;
55 sk_for_each_bound(sk, node, &l2tp_ip_bind_table) {
56 struct inet_sock *inet = inet_sk(sk);
57 struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
59 if (l2tp == NULL)
60 continue;
62 if ((l2tp->conn_id == tunnel_id) &&
63 net_eq(sock_net(sk), net) &&
64 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
65 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
66 goto found;
69 sk = NULL;
70 found:
71 return sk;
74 static inline struct sock *l2tp_ip_bind_lookup(struct net *net, __be32 laddr, int dif, u32 tunnel_id)
76 struct sock *sk = __l2tp_ip_bind_lookup(net, laddr, dif, tunnel_id);
77 if (sk)
78 sock_hold(sk);
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 sock *sk;
119 u32 session_id;
120 u32 tunnel_id;
121 unsigned char *ptr, *optr;
122 struct l2tp_session *session;
123 struct l2tp_tunnel *tunnel = NULL;
124 int length;
126 /* Point to L2TP header */
127 optr = ptr = skb->data;
129 if (!pskb_may_pull(skb, 4))
130 goto discard;
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(&init_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 pr_debug("%s: ip recv\n", tunnel->name);
160 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
163 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
165 return 0;
167 pass_up:
168 /* Get the tunnel_id from the L2TP header */
169 if (!pskb_may_pull(skb, 12))
170 goto discard;
172 if ((skb->data[0] & 0xc0) != 0xc0)
173 goto discard;
175 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
176 tunnel = l2tp_tunnel_find(&init_net, tunnel_id);
177 if (tunnel != NULL)
178 sk = tunnel->sock;
179 else {
180 struct iphdr *iph = (struct iphdr *) skb_network_header(skb);
182 read_lock_bh(&l2tp_ip_lock);
183 sk = __l2tp_ip_bind_lookup(&init_net, iph->daddr, 0, tunnel_id);
184 read_unlock_bh(&l2tp_ip_lock);
187 if (sk == NULL)
188 goto discard;
190 sock_hold(sk);
192 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
193 goto discard_put;
195 nf_reset(skb);
197 return sk_receive_skb(sk, skb, 1);
199 discard_put:
200 sock_put(sk);
202 discard:
203 kfree_skb(skb);
204 return 0;
207 static int l2tp_ip_open(struct sock *sk)
209 /* Prevent autobind. We don't have ports. */
210 inet_sk(sk)->inet_num = IPPROTO_L2TP;
212 write_lock_bh(&l2tp_ip_lock);
213 sk_add_node(sk, &l2tp_ip_table);
214 write_unlock_bh(&l2tp_ip_lock);
216 return 0;
219 static void l2tp_ip_close(struct sock *sk, long timeout)
221 write_lock_bh(&l2tp_ip_lock);
222 hlist_del_init(&sk->sk_bind_node);
223 sk_del_node_init(sk);
224 write_unlock_bh(&l2tp_ip_lock);
225 sk_common_release(sk);
228 static void l2tp_ip_destroy_sock(struct sock *sk)
230 struct sk_buff *skb;
232 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
233 kfree_skb(skb);
235 sk_refcnt_debug_dec(sk);
238 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
240 struct inet_sock *inet = inet_sk(sk);
241 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
242 int ret;
243 int chk_addr_ret;
245 if (!sock_flag(sk, SOCK_ZAPPED))
246 return -EINVAL;
247 if (addr_len < sizeof(struct sockaddr_l2tpip))
248 return -EINVAL;
249 if (addr->l2tp_family != AF_INET)
250 return -EINVAL;
252 ret = -EADDRINUSE;
253 read_lock_bh(&l2tp_ip_lock);
254 if (__l2tp_ip_bind_lookup(&init_net, addr->l2tp_addr.s_addr, sk->sk_bound_dev_if, addr->l2tp_conn_id))
255 goto out_in_use;
257 read_unlock_bh(&l2tp_ip_lock);
259 lock_sock(sk);
260 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
261 goto out;
263 chk_addr_ret = inet_addr_type(&init_net, addr->l2tp_addr.s_addr);
264 ret = -EADDRNOTAVAIL;
265 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
266 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
267 goto out;
269 if (addr->l2tp_addr.s_addr)
270 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
271 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
272 inet->inet_saddr = 0; /* Use device */
273 sk_dst_reset(sk);
275 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
277 write_lock_bh(&l2tp_ip_lock);
278 sk_add_bind_node(sk, &l2tp_ip_bind_table);
279 sk_del_node_init(sk);
280 write_unlock_bh(&l2tp_ip_lock);
281 ret = 0;
282 sock_reset_flag(sk, SOCK_ZAPPED);
284 out:
285 release_sock(sk);
287 return ret;
289 out_in_use:
290 read_unlock_bh(&l2tp_ip_lock);
292 return ret;
295 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
297 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
298 int rc;
300 if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
301 return -EINVAL;
303 if (addr_len < sizeof(*lsa))
304 return -EINVAL;
306 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
307 return -EINVAL;
309 rc = ip4_datagram_connect(sk, uaddr, addr_len);
310 if (rc < 0)
311 return rc;
313 lock_sock(sk);
315 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
317 write_lock_bh(&l2tp_ip_lock);
318 hlist_del_init(&sk->sk_bind_node);
319 sk_add_bind_node(sk, &l2tp_ip_bind_table);
320 write_unlock_bh(&l2tp_ip_lock);
322 release_sock(sk);
323 return rc;
326 static int l2tp_ip_disconnect(struct sock *sk, int flags)
328 if (sock_flag(sk, SOCK_ZAPPED))
329 return 0;
331 return udp_disconnect(sk, flags);
334 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
335 int *uaddr_len, int peer)
337 struct sock *sk = sock->sk;
338 struct inet_sock *inet = inet_sk(sk);
339 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
340 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
342 memset(lsa, 0, sizeof(*lsa));
343 lsa->l2tp_family = AF_INET;
344 if (peer) {
345 if (!inet->inet_dport)
346 return -ENOTCONN;
347 lsa->l2tp_conn_id = lsk->peer_conn_id;
348 lsa->l2tp_addr.s_addr = inet->inet_daddr;
349 } else {
350 __be32 addr = inet->inet_rcv_saddr;
351 if (!addr)
352 addr = inet->inet_saddr;
353 lsa->l2tp_conn_id = lsk->conn_id;
354 lsa->l2tp_addr.s_addr = addr;
356 *uaddr_len = sizeof(*lsa);
357 return 0;
360 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
362 int rc;
364 /* Charge it to the socket, dropping if the queue is full. */
365 rc = sock_queue_rcv_skb(sk, skb);
366 if (rc < 0)
367 goto drop;
369 return 0;
371 drop:
372 IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
373 kfree_skb(skb);
374 return -1;
377 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
378 * control frames.
380 static int l2tp_ip_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, size_t len)
382 struct sk_buff *skb;
383 int rc;
384 struct inet_sock *inet = inet_sk(sk);
385 struct rtable *rt = NULL;
386 struct flowi4 *fl4;
387 int connected = 0;
388 __be32 daddr;
390 lock_sock(sk);
392 rc = -ENOTCONN;
393 if (sock_flag(sk, SOCK_DEAD))
394 goto out;
396 /* Get and verify the address. */
397 if (msg->msg_name) {
398 struct sockaddr_l2tpip *lip = (struct sockaddr_l2tpip *) msg->msg_name;
399 rc = -EINVAL;
400 if (msg->msg_namelen < sizeof(*lip))
401 goto out;
403 if (lip->l2tp_family != AF_INET) {
404 rc = -EAFNOSUPPORT;
405 if (lip->l2tp_family != AF_UNSPEC)
406 goto out;
409 daddr = lip->l2tp_addr.s_addr;
410 } else {
411 rc = -EDESTADDRREQ;
412 if (sk->sk_state != TCP_ESTABLISHED)
413 goto out;
415 daddr = inet->inet_daddr;
416 connected = 1;
419 /* Allocate a socket buffer */
420 rc = -ENOMEM;
421 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
422 4 + len, 0, GFP_KERNEL);
423 if (!skb)
424 goto error;
426 /* Reserve space for headers, putting IP header on 4-byte boundary. */
427 skb_reserve(skb, 2 + NET_SKB_PAD);
428 skb_reset_network_header(skb);
429 skb_reserve(skb, sizeof(struct iphdr));
430 skb_reset_transport_header(skb);
432 /* Insert 0 session_id */
433 *((__be32 *) skb_put(skb, 4)) = 0;
435 /* Copy user data into skb */
436 rc = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
437 if (rc < 0) {
438 kfree_skb(skb);
439 goto error;
442 fl4 = &inet->cork.fl.u.ip4;
443 if (connected)
444 rt = (struct rtable *) __sk_dst_check(sk, 0);
446 rcu_read_lock();
447 if (rt == NULL) {
448 const struct ip_options_rcu *inet_opt;
450 inet_opt = rcu_dereference(inet->inet_opt);
452 /* Use correct destination address if we have options. */
453 if (inet_opt && inet_opt->opt.srr)
454 daddr = inet_opt->opt.faddr;
456 /* If this fails, retransmit mechanism of transport layer will
457 * keep trying until route appears or the connection times
458 * itself out.
460 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
461 daddr, inet->inet_saddr,
462 inet->inet_dport, inet->inet_sport,
463 sk->sk_protocol, RT_CONN_FLAGS(sk),
464 sk->sk_bound_dev_if);
465 if (IS_ERR(rt))
466 goto no_route;
467 if (connected) {
468 sk_setup_caps(sk, &rt->dst);
469 } else {
470 skb_dst_set(skb, &rt->dst);
471 goto xmit;
475 /* We dont need to clone dst here, it is guaranteed to not disappear.
476 * __dev_xmit_skb() might force a refcount if needed.
478 skb_dst_set_noref(skb, &rt->dst);
480 xmit:
481 /* Queue the packet to IP for output */
482 rc = ip_queue_xmit(skb, &inet->cork.fl);
483 rcu_read_unlock();
485 error:
486 if (rc >= 0)
487 rc = len;
489 out:
490 release_sock(sk);
491 return rc;
493 no_route:
494 rcu_read_unlock();
495 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
496 kfree_skb(skb);
497 rc = -EHOSTUNREACH;
498 goto out;
501 static int l2tp_ip_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
502 size_t len, int noblock, int flags, int *addr_len)
504 struct inet_sock *inet = inet_sk(sk);
505 size_t copied = 0;
506 int err = -EOPNOTSUPP;
507 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
508 struct sk_buff *skb;
510 if (flags & MSG_OOB)
511 goto out;
513 if (addr_len)
514 *addr_len = sizeof(*sin);
516 skb = skb_recv_datagram(sk, flags, noblock, &err);
517 if (!skb)
518 goto out;
520 copied = skb->len;
521 if (len < copied) {
522 msg->msg_flags |= MSG_TRUNC;
523 copied = len;
526 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
527 if (err)
528 goto done;
530 sock_recv_timestamp(msg, sk, skb);
532 /* Copy the address. */
533 if (sin) {
534 sin->sin_family = AF_INET;
535 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
536 sin->sin_port = 0;
537 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
539 if (inet->cmsg_flags)
540 ip_cmsg_recv(msg, skb);
541 if (flags & MSG_TRUNC)
542 copied = skb->len;
543 done:
544 skb_free_datagram(sk, skb);
545 out:
546 return err ? err : copied;
549 static struct proto l2tp_ip_prot = {
550 .name = "L2TP/IP",
551 .owner = THIS_MODULE,
552 .init = l2tp_ip_open,
553 .close = l2tp_ip_close,
554 .bind = l2tp_ip_bind,
555 .connect = l2tp_ip_connect,
556 .disconnect = l2tp_ip_disconnect,
557 .ioctl = udp_ioctl,
558 .destroy = l2tp_ip_destroy_sock,
559 .setsockopt = ip_setsockopt,
560 .getsockopt = ip_getsockopt,
561 .sendmsg = l2tp_ip_sendmsg,
562 .recvmsg = l2tp_ip_recvmsg,
563 .backlog_rcv = l2tp_ip_backlog_recv,
564 .hash = inet_hash,
565 .unhash = inet_unhash,
566 .obj_size = sizeof(struct l2tp_ip_sock),
567 #ifdef CONFIG_COMPAT
568 .compat_setsockopt = compat_ip_setsockopt,
569 .compat_getsockopt = compat_ip_getsockopt,
570 #endif
573 static const struct proto_ops l2tp_ip_ops = {
574 .family = PF_INET,
575 .owner = THIS_MODULE,
576 .release = inet_release,
577 .bind = inet_bind,
578 .connect = inet_dgram_connect,
579 .socketpair = sock_no_socketpair,
580 .accept = sock_no_accept,
581 .getname = l2tp_ip_getname,
582 .poll = datagram_poll,
583 .ioctl = inet_ioctl,
584 .listen = sock_no_listen,
585 .shutdown = inet_shutdown,
586 .setsockopt = sock_common_setsockopt,
587 .getsockopt = sock_common_getsockopt,
588 .sendmsg = inet_sendmsg,
589 .recvmsg = sock_common_recvmsg,
590 .mmap = sock_no_mmap,
591 .sendpage = sock_no_sendpage,
592 #ifdef CONFIG_COMPAT
593 .compat_setsockopt = compat_sock_common_setsockopt,
594 .compat_getsockopt = compat_sock_common_getsockopt,
595 #endif
598 static struct inet_protosw l2tp_ip_protosw = {
599 .type = SOCK_DGRAM,
600 .protocol = IPPROTO_L2TP,
601 .prot = &l2tp_ip_prot,
602 .ops = &l2tp_ip_ops,
603 .no_check = 0,
606 static struct net_protocol l2tp_ip_protocol __read_mostly = {
607 .handler = l2tp_ip_recv,
610 static int __init l2tp_ip_init(void)
612 int err;
614 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
616 err = proto_register(&l2tp_ip_prot, 1);
617 if (err != 0)
618 goto out;
620 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
621 if (err)
622 goto out1;
624 inet_register_protosw(&l2tp_ip_protosw);
625 return 0;
627 out1:
628 proto_unregister(&l2tp_ip_prot);
629 out:
630 return err;
633 static void __exit l2tp_ip_exit(void)
635 inet_unregister_protosw(&l2tp_ip_protosw);
636 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
637 proto_unregister(&l2tp_ip_prot);
640 module_init(l2tp_ip_init);
641 module_exit(l2tp_ip_exit);
643 MODULE_LICENSE("GPL");
644 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
645 MODULE_DESCRIPTION("L2TP over IP");
646 MODULE_VERSION("1.0");
648 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
649 * enums
651 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);