xref: /linux/net/l2tp/l2tp_ip6.c (revision 25aee3debe0464f6c680173041fa3de30ec9ff54)
1 /*
2  * L2TPv3 IP encapsulation support for IPv6
3  *
4  * Copyright (c) 2012 Katalix Systems Ltd
5  *
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.
10  */
11 
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 
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 <linux/in6.h>
22 #include <net/sock.h>
23 #include <net/ip.h>
24 #include <net/icmp.h>
25 #include <net/udp.h>
26 #include <net/inet_common.h>
27 #include <net/inet_hashtables.h>
28 #include <net/tcp_states.h>
29 #include <net/protocol.h>
30 #include <net/xfrm.h>
31 
32 #include <net/transp_v6.h>
33 #include <net/addrconf.h>
34 #include <net/ip6_route.h>
35 
36 #include "l2tp_core.h"
37 
38 struct l2tp_ip6_sock {
39 	/* inet_sock has to be the first member of l2tp_ip6_sock */
40 	struct inet_sock	inet;
41 
42 	u32			conn_id;
43 	u32			peer_conn_id;
44 
45 	/* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
46 	   inet6_sk_generic */
47 	struct ipv6_pinfo	inet6;
48 };
49 
50 static DEFINE_RWLOCK(l2tp_ip6_lock);
51 static struct hlist_head l2tp_ip6_table;
52 static struct hlist_head l2tp_ip6_bind_table;
53 
54 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
55 {
56 	return (struct l2tp_ip6_sock *)sk;
57 }
58 
59 static struct sock *__l2tp_ip6_bind_lookup(struct net *net,
60 					   struct in6_addr *laddr,
61 					   int dif, u32 tunnel_id)
62 {
63 	struct hlist_node *node;
64 	struct sock *sk;
65 
66 	sk_for_each_bound(sk, node, &l2tp_ip6_bind_table) {
67 		struct in6_addr *addr = inet6_rcv_saddr(sk);
68 		struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
69 
70 		if (l2tp == NULL)
71 			continue;
72 
73 		if ((l2tp->conn_id == tunnel_id) &&
74 		    net_eq(sock_net(sk), net) &&
75 		    !(addr && ipv6_addr_equal(addr, laddr)) &&
76 		    !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
77 			goto found;
78 	}
79 
80 	sk = NULL;
81 found:
82 	return sk;
83 }
84 
85 static inline struct sock *l2tp_ip6_bind_lookup(struct net *net,
86 						struct in6_addr *laddr,
87 						int dif, u32 tunnel_id)
88 {
89 	struct sock *sk = __l2tp_ip6_bind_lookup(net, laddr, dif, tunnel_id);
90 	if (sk)
91 		sock_hold(sk);
92 
93 	return sk;
94 }
95 
96 /* When processing receive frames, there are two cases to
97  * consider. Data frames consist of a non-zero session-id and an
98  * optional cookie. Control frames consist of a regular L2TP header
99  * preceded by 32-bits of zeros.
100  *
101  * L2TPv3 Session Header Over IP
102  *
103  *  0                   1                   2                   3
104  *  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
105  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106  * |                           Session ID                          |
107  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108  * |               Cookie (optional, maximum 64 bits)...
109  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
110  *                                                                 |
111  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112  *
113  * L2TPv3 Control Message Header Over IP
114  *
115  *  0                   1                   2                   3
116  *  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
117  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
118  * |                      (32 bits of zeros)                       |
119  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
120  * |T|L|x|x|S|x|x|x|x|x|x|x|  Ver  |             Length            |
121  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
122  * |                     Control Connection ID                     |
123  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
124  * |               Ns              |               Nr              |
125  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
126  *
127  * All control frames are passed to userspace.
128  */
129 static int l2tp_ip6_recv(struct sk_buff *skb)
130 {
131 	struct sock *sk;
132 	u32 session_id;
133 	u32 tunnel_id;
134 	unsigned char *ptr, *optr;
135 	struct l2tp_session *session;
136 	struct l2tp_tunnel *tunnel = NULL;
137 	int length;
138 
139 	/* Point to L2TP header */
140 	optr = ptr = skb->data;
141 
142 	if (!pskb_may_pull(skb, 4))
143 		goto discard;
144 
145 	session_id = ntohl(*((__be32 *) ptr));
146 	ptr += 4;
147 
148 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
149 	 * the session_id. If it is 0, the packet is a L2TP control
150 	 * frame and the session_id value can be discarded.
151 	 */
152 	if (session_id == 0) {
153 		__skb_pull(skb, 4);
154 		goto pass_up;
155 	}
156 
157 	/* Ok, this is a data packet. Lookup the session. */
158 	session = l2tp_session_find(&init_net, NULL, session_id);
159 	if (session == NULL)
160 		goto discard;
161 
162 	tunnel = session->tunnel;
163 	if (tunnel == NULL)
164 		goto discard;
165 
166 	/* Trace packet contents, if enabled */
167 	if (tunnel->debug & L2TP_MSG_DATA) {
168 		length = min(32u, skb->len);
169 		if (!pskb_may_pull(skb, length))
170 			goto discard;
171 
172 		pr_debug("%s: ip recv\n", tunnel->name);
173 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
174 	}
175 
176 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
177 			 tunnel->recv_payload_hook);
178 	return 0;
179 
180 pass_up:
181 	/* Get the tunnel_id from the L2TP header */
182 	if (!pskb_may_pull(skb, 12))
183 		goto discard;
184 
185 	if ((skb->data[0] & 0xc0) != 0xc0)
186 		goto discard;
187 
188 	tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
189 	tunnel = l2tp_tunnel_find(&init_net, tunnel_id);
190 	if (tunnel != NULL)
191 		sk = tunnel->sock;
192 	else {
193 		struct ipv6hdr *iph = ipv6_hdr(skb);
194 
195 		read_lock_bh(&l2tp_ip6_lock);
196 		sk = __l2tp_ip6_bind_lookup(&init_net, &iph->daddr,
197 					    0, tunnel_id);
198 		read_unlock_bh(&l2tp_ip6_lock);
199 	}
200 
201 	if (sk == NULL)
202 		goto discard;
203 
204 	sock_hold(sk);
205 
206 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
207 		goto discard_put;
208 
209 	nf_reset(skb);
210 
211 	return sk_receive_skb(sk, skb, 1);
212 
213 discard_put:
214 	sock_put(sk);
215 
216 discard:
217 	kfree_skb(skb);
218 	return 0;
219 }
220 
221 static int l2tp_ip6_open(struct sock *sk)
222 {
223 	/* Prevent autobind. We don't have ports. */
224 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
225 
226 	write_lock_bh(&l2tp_ip6_lock);
227 	sk_add_node(sk, &l2tp_ip6_table);
228 	write_unlock_bh(&l2tp_ip6_lock);
229 
230 	return 0;
231 }
232 
233 static void l2tp_ip6_close(struct sock *sk, long timeout)
234 {
235 	write_lock_bh(&l2tp_ip6_lock);
236 	hlist_del_init(&sk->sk_bind_node);
237 	sk_del_node_init(sk);
238 	write_unlock_bh(&l2tp_ip6_lock);
239 
240 	sk_common_release(sk);
241 }
242 
243 static void l2tp_ip6_destroy_sock(struct sock *sk)
244 {
245 	lock_sock(sk);
246 	ip6_flush_pending_frames(sk);
247 	release_sock(sk);
248 
249 	inet6_destroy_sock(sk);
250 }
251 
252 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
253 {
254 	struct inet_sock *inet = inet_sk(sk);
255 	struct ipv6_pinfo *np = inet6_sk(sk);
256 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
257 	__be32 v4addr = 0;
258 	int addr_type;
259 	int err;
260 
261 	if (!sock_flag(sk, SOCK_ZAPPED))
262 		return -EINVAL;
263 	if (addr->l2tp_family != AF_INET6)
264 		return -EINVAL;
265 	if (addr_len < sizeof(*addr))
266 		return -EINVAL;
267 
268 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
269 
270 	/* l2tp_ip6 sockets are IPv6 only */
271 	if (addr_type == IPV6_ADDR_MAPPED)
272 		return -EADDRNOTAVAIL;
273 
274 	/* L2TP is point-point, not multicast */
275 	if (addr_type & IPV6_ADDR_MULTICAST)
276 		return -EADDRNOTAVAIL;
277 
278 	err = -EADDRINUSE;
279 	read_lock_bh(&l2tp_ip6_lock);
280 	if (__l2tp_ip6_bind_lookup(&init_net, &addr->l2tp_addr,
281 				   sk->sk_bound_dev_if, addr->l2tp_conn_id))
282 		goto out_in_use;
283 	read_unlock_bh(&l2tp_ip6_lock);
284 
285 	lock_sock(sk);
286 
287 	err = -EINVAL;
288 	if (sk->sk_state != TCP_CLOSE)
289 		goto out_unlock;
290 
291 	/* Check if the address belongs to the host. */
292 	rcu_read_lock();
293 	if (addr_type != IPV6_ADDR_ANY) {
294 		struct net_device *dev = NULL;
295 
296 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
297 			if (addr_len >= sizeof(struct sockaddr_in6) &&
298 			    addr->l2tp_scope_id) {
299 				/* Override any existing binding, if another
300 				 * one is supplied by user.
301 				 */
302 				sk->sk_bound_dev_if = addr->l2tp_scope_id;
303 			}
304 
305 			/* Binding to link-local address requires an
306 			   interface */
307 			if (!sk->sk_bound_dev_if)
308 				goto out_unlock_rcu;
309 
310 			err = -ENODEV;
311 			dev = dev_get_by_index_rcu(sock_net(sk),
312 						   sk->sk_bound_dev_if);
313 			if (!dev)
314 				goto out_unlock_rcu;
315 		}
316 
317 		/* ipv4 addr of the socket is invalid.  Only the
318 		 * unspecified and mapped address have a v4 equivalent.
319 		 */
320 		v4addr = LOOPBACK4_IPV6;
321 		err = -EADDRNOTAVAIL;
322 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
323 			goto out_unlock_rcu;
324 	}
325 	rcu_read_unlock();
326 
327 	inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
328 	np->rcv_saddr = addr->l2tp_addr;
329 	np->saddr = addr->l2tp_addr;
330 
331 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
332 
333 	write_lock_bh(&l2tp_ip6_lock);
334 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
335 	sk_del_node_init(sk);
336 	write_unlock_bh(&l2tp_ip6_lock);
337 
338 	sock_reset_flag(sk, SOCK_ZAPPED);
339 	release_sock(sk);
340 	return 0;
341 
342 out_unlock_rcu:
343 	rcu_read_unlock();
344 out_unlock:
345 	release_sock(sk);
346 	return err;
347 
348 out_in_use:
349 	read_unlock_bh(&l2tp_ip6_lock);
350 	return err;
351 }
352 
353 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
354 			    int addr_len)
355 {
356 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
357 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
358 	struct in6_addr	*daddr;
359 	int	addr_type;
360 	int rc;
361 
362 	if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
363 		return -EINVAL;
364 
365 	if (addr_len < sizeof(*lsa))
366 		return -EINVAL;
367 
368 	addr_type = ipv6_addr_type(&usin->sin6_addr);
369 	if (addr_type & IPV6_ADDR_MULTICAST)
370 		return -EINVAL;
371 
372 	if (addr_type & IPV6_ADDR_MAPPED) {
373 		daddr = &usin->sin6_addr;
374 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
375 			return -EINVAL;
376 	}
377 
378 	rc = ip6_datagram_connect(sk, uaddr, addr_len);
379 
380 	lock_sock(sk);
381 
382 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
383 
384 	write_lock_bh(&l2tp_ip6_lock);
385 	hlist_del_init(&sk->sk_bind_node);
386 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
387 	write_unlock_bh(&l2tp_ip6_lock);
388 
389 	release_sock(sk);
390 
391 	return rc;
392 }
393 
394 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
395 {
396 	if (sock_flag(sk, SOCK_ZAPPED))
397 		return 0;
398 
399 	return udp_disconnect(sk, flags);
400 }
401 
402 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
403 			    int *uaddr_len, int peer)
404 {
405 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
406 	struct sock *sk = sock->sk;
407 	struct ipv6_pinfo *np = inet6_sk(sk);
408 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
409 
410 	lsa->l2tp_family = AF_INET6;
411 	lsa->l2tp_flowinfo = 0;
412 	lsa->l2tp_scope_id = 0;
413 	if (peer) {
414 		if (!lsk->peer_conn_id)
415 			return -ENOTCONN;
416 		lsa->l2tp_conn_id = lsk->peer_conn_id;
417 		lsa->l2tp_addr = np->daddr;
418 		if (np->sndflow)
419 			lsa->l2tp_flowinfo = np->flow_label;
420 	} else {
421 		if (ipv6_addr_any(&np->rcv_saddr))
422 			lsa->l2tp_addr = np->saddr;
423 		else
424 			lsa->l2tp_addr = np->rcv_saddr;
425 
426 		lsa->l2tp_conn_id = lsk->conn_id;
427 	}
428 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
429 		lsa->l2tp_scope_id = sk->sk_bound_dev_if;
430 	*uaddr_len = sizeof(*lsa);
431 	return 0;
432 }
433 
434 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
435 {
436 	int rc;
437 
438 	/* Charge it to the socket, dropping if the queue is full. */
439 	rc = sock_queue_rcv_skb(sk, skb);
440 	if (rc < 0)
441 		goto drop;
442 
443 	return 0;
444 
445 drop:
446 	IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
447 	kfree_skb(skb);
448 	return -1;
449 }
450 
451 static int l2tp_ip6_push_pending_frames(struct sock *sk)
452 {
453 	struct sk_buff *skb;
454 	__be32 *transhdr = NULL;
455 	int err = 0;
456 
457 	skb = skb_peek(&sk->sk_write_queue);
458 	if (skb == NULL)
459 		goto out;
460 
461 	transhdr = (__be32 *)skb_transport_header(skb);
462 	*transhdr = 0;
463 
464 	err = ip6_push_pending_frames(sk);
465 
466 out:
467 	return err;
468 }
469 
470 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
471  * control frames.
472  */
473 static int l2tp_ip6_sendmsg(struct kiocb *iocb, struct sock *sk,
474 			    struct msghdr *msg, size_t len)
475 {
476 	struct ipv6_txoptions opt_space;
477 	struct sockaddr_l2tpip6 *lsa =
478 		(struct sockaddr_l2tpip6 *) msg->msg_name;
479 	struct in6_addr *daddr, *final_p, final;
480 	struct ipv6_pinfo *np = inet6_sk(sk);
481 	struct ipv6_txoptions *opt = NULL;
482 	struct ip6_flowlabel *flowlabel = NULL;
483 	struct dst_entry *dst = NULL;
484 	struct flowi6 fl6;
485 	int addr_len = msg->msg_namelen;
486 	int hlimit = -1;
487 	int tclass = -1;
488 	int dontfrag = -1;
489 	int transhdrlen = 4; /* zero session-id */
490 	int ulen = len + transhdrlen;
491 	int err;
492 
493 	/* Rough check on arithmetic overflow,
494 	   better check is made in ip6_append_data().
495 	 */
496 	if (len > INT_MAX)
497 		return -EMSGSIZE;
498 
499 	/* Mirror BSD error message compatibility */
500 	if (msg->msg_flags & MSG_OOB)
501 		return -EOPNOTSUPP;
502 
503 	/*
504 	 *	Get and verify the address.
505 	 */
506 	memset(&fl6, 0, sizeof(fl6));
507 
508 	fl6.flowi6_mark = sk->sk_mark;
509 
510 	if (lsa) {
511 		if (addr_len < SIN6_LEN_RFC2133)
512 			return -EINVAL;
513 
514 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
515 			return -EAFNOSUPPORT;
516 
517 		daddr = &lsa->l2tp_addr;
518 		if (np->sndflow) {
519 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
520 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
521 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
522 				if (flowlabel == NULL)
523 					return -EINVAL;
524 				daddr = &flowlabel->dst;
525 			}
526 		}
527 
528 		/*
529 		 * Otherwise it will be difficult to maintain
530 		 * sk->sk_dst_cache.
531 		 */
532 		if (sk->sk_state == TCP_ESTABLISHED &&
533 		    ipv6_addr_equal(daddr, &np->daddr))
534 			daddr = &np->daddr;
535 
536 		if (addr_len >= sizeof(struct sockaddr_in6) &&
537 		    lsa->l2tp_scope_id &&
538 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
539 			fl6.flowi6_oif = lsa->l2tp_scope_id;
540 	} else {
541 		if (sk->sk_state != TCP_ESTABLISHED)
542 			return -EDESTADDRREQ;
543 
544 		daddr = &np->daddr;
545 		fl6.flowlabel = np->flow_label;
546 	}
547 
548 	if (fl6.flowi6_oif == 0)
549 		fl6.flowi6_oif = sk->sk_bound_dev_if;
550 
551 	if (msg->msg_controllen) {
552 		opt = &opt_space;
553 		memset(opt, 0, sizeof(struct ipv6_txoptions));
554 		opt->tot_len = sizeof(struct ipv6_txoptions);
555 
556 		err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
557 					&hlimit, &tclass, &dontfrag);
558 		if (err < 0) {
559 			fl6_sock_release(flowlabel);
560 			return err;
561 		}
562 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
563 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
564 			if (flowlabel == NULL)
565 				return -EINVAL;
566 		}
567 		if (!(opt->opt_nflen|opt->opt_flen))
568 			opt = NULL;
569 	}
570 
571 	if (opt == NULL)
572 		opt = np->opt;
573 	if (flowlabel)
574 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
575 	opt = ipv6_fixup_options(&opt_space, opt);
576 
577 	fl6.flowi6_proto = sk->sk_protocol;
578 	if (!ipv6_addr_any(daddr))
579 		fl6.daddr = *daddr;
580 	else
581 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
582 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
583 		fl6.saddr = np->saddr;
584 
585 	final_p = fl6_update_dst(&fl6, opt, &final);
586 
587 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
588 		fl6.flowi6_oif = np->mcast_oif;
589 	else if (!fl6.flowi6_oif)
590 		fl6.flowi6_oif = np->ucast_oif;
591 
592 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
593 
594 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
595 	if (IS_ERR(dst)) {
596 		err = PTR_ERR(dst);
597 		goto out;
598 	}
599 
600 	if (hlimit < 0) {
601 		if (ipv6_addr_is_multicast(&fl6.daddr))
602 			hlimit = np->mcast_hops;
603 		else
604 			hlimit = np->hop_limit;
605 		if (hlimit < 0)
606 			hlimit = ip6_dst_hoplimit(dst);
607 	}
608 
609 	if (tclass < 0)
610 		tclass = np->tclass;
611 
612 	if (dontfrag < 0)
613 		dontfrag = np->dontfrag;
614 
615 	if (msg->msg_flags & MSG_CONFIRM)
616 		goto do_confirm;
617 
618 back_from_confirm:
619 	lock_sock(sk);
620 	err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
621 			      ulen, transhdrlen, hlimit, tclass, opt,
622 			      &fl6, (struct rt6_info *)dst,
623 			      msg->msg_flags, dontfrag);
624 	if (err)
625 		ip6_flush_pending_frames(sk);
626 	else if (!(msg->msg_flags & MSG_MORE))
627 		err = l2tp_ip6_push_pending_frames(sk);
628 	release_sock(sk);
629 done:
630 	dst_release(dst);
631 out:
632 	fl6_sock_release(flowlabel);
633 
634 	return err < 0 ? err : len;
635 
636 do_confirm:
637 	dst_confirm(dst);
638 	if (!(msg->msg_flags & MSG_PROBE) || len)
639 		goto back_from_confirm;
640 	err = 0;
641 	goto done;
642 }
643 
644 static int l2tp_ip6_recvmsg(struct kiocb *iocb, struct sock *sk,
645 			    struct msghdr *msg, size_t len, int noblock,
646 			    int flags, int *addr_len)
647 {
648 	struct inet_sock *inet = inet_sk(sk);
649 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)msg->msg_name;
650 	size_t copied = 0;
651 	int err = -EOPNOTSUPP;
652 	struct sk_buff *skb;
653 
654 	if (flags & MSG_OOB)
655 		goto out;
656 
657 	if (addr_len)
658 		*addr_len = sizeof(*lsa);
659 
660 	if (flags & MSG_ERRQUEUE)
661 		return ipv6_recv_error(sk, msg, len);
662 
663 	skb = skb_recv_datagram(sk, flags, noblock, &err);
664 	if (!skb)
665 		goto out;
666 
667 	copied = skb->len;
668 	if (len < copied) {
669 		msg->msg_flags |= MSG_TRUNC;
670 		copied = len;
671 	}
672 
673 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
674 	if (err)
675 		goto done;
676 
677 	sock_recv_timestamp(msg, sk, skb);
678 
679 	/* Copy the address. */
680 	if (lsa) {
681 		lsa->l2tp_family = AF_INET6;
682 		lsa->l2tp_unused = 0;
683 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
684 		lsa->l2tp_flowinfo = 0;
685 		lsa->l2tp_scope_id = 0;
686 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
687 			lsa->l2tp_scope_id = IP6CB(skb)->iif;
688 	}
689 
690 	if (inet->cmsg_flags)
691 		ip_cmsg_recv(msg, skb);
692 
693 	if (flags & MSG_TRUNC)
694 		copied = skb->len;
695 done:
696 	skb_free_datagram(sk, skb);
697 out:
698 	return err ? err : copied;
699 }
700 
701 static struct proto l2tp_ip6_prot = {
702 	.name		   = "L2TP/IPv6",
703 	.owner		   = THIS_MODULE,
704 	.init		   = l2tp_ip6_open,
705 	.close		   = l2tp_ip6_close,
706 	.bind		   = l2tp_ip6_bind,
707 	.connect	   = l2tp_ip6_connect,
708 	.disconnect	   = l2tp_ip6_disconnect,
709 	.ioctl		   = udp_ioctl,
710 	.destroy	   = l2tp_ip6_destroy_sock,
711 	.setsockopt	   = ipv6_setsockopt,
712 	.getsockopt	   = ipv6_getsockopt,
713 	.sendmsg	   = l2tp_ip6_sendmsg,
714 	.recvmsg	   = l2tp_ip6_recvmsg,
715 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
716 	.hash		   = inet_hash,
717 	.unhash		   = inet_unhash,
718 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
719 #ifdef CONFIG_COMPAT
720 	.compat_setsockopt = compat_ipv6_setsockopt,
721 	.compat_getsockopt = compat_ipv6_getsockopt,
722 #endif
723 };
724 
725 static const struct proto_ops l2tp_ip6_ops = {
726 	.family		   = PF_INET6,
727 	.owner		   = THIS_MODULE,
728 	.release	   = inet6_release,
729 	.bind		   = inet6_bind,
730 	.connect	   = inet_dgram_connect,
731 	.socketpair	   = sock_no_socketpair,
732 	.accept		   = sock_no_accept,
733 	.getname	   = l2tp_ip6_getname,
734 	.poll		   = datagram_poll,
735 	.ioctl		   = inet6_ioctl,
736 	.listen		   = sock_no_listen,
737 	.shutdown	   = inet_shutdown,
738 	.setsockopt	   = sock_common_setsockopt,
739 	.getsockopt	   = sock_common_getsockopt,
740 	.sendmsg	   = inet_sendmsg,
741 	.recvmsg	   = sock_common_recvmsg,
742 	.mmap		   = sock_no_mmap,
743 	.sendpage	   = sock_no_sendpage,
744 #ifdef CONFIG_COMPAT
745 	.compat_setsockopt = compat_sock_common_setsockopt,
746 	.compat_getsockopt = compat_sock_common_getsockopt,
747 #endif
748 };
749 
750 static struct inet_protosw l2tp_ip6_protosw = {
751 	.type		= SOCK_DGRAM,
752 	.protocol	= IPPROTO_L2TP,
753 	.prot		= &l2tp_ip6_prot,
754 	.ops		= &l2tp_ip6_ops,
755 	.no_check	= 0,
756 };
757 
758 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
759 	.handler	= l2tp_ip6_recv,
760 };
761 
762 static int __init l2tp_ip6_init(void)
763 {
764 	int err;
765 
766 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
767 
768 	err = proto_register(&l2tp_ip6_prot, 1);
769 	if (err != 0)
770 		goto out;
771 
772 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
773 	if (err)
774 		goto out1;
775 
776 	inet6_register_protosw(&l2tp_ip6_protosw);
777 	return 0;
778 
779 out1:
780 	proto_unregister(&l2tp_ip6_prot);
781 out:
782 	return err;
783 }
784 
785 static void __exit l2tp_ip6_exit(void)
786 {
787 	inet6_unregister_protosw(&l2tp_ip6_protosw);
788 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
789 	proto_unregister(&l2tp_ip6_prot);
790 }
791 
792 module_init(l2tp_ip6_init);
793 module_exit(l2tp_ip6_exit);
794 
795 MODULE_LICENSE("GPL");
796 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
797 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
798 MODULE_VERSION("1.0");
799 
800 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
801  * enums
802  */
803 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
804