xref: /openbmc/linux/net/l2tp/l2tp_ip6.c (revision 930beb5a)
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 sock *sk;
64 
65 	sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
66 		const struct in6_addr *addr = inet6_rcv_saddr(sk);
67 		struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
68 
69 		if (l2tp == NULL)
70 			continue;
71 
72 		if ((l2tp->conn_id == tunnel_id) &&
73 		    net_eq(sock_net(sk), net) &&
74 		    !(addr && ipv6_addr_equal(addr, laddr)) &&
75 		    !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
76 			goto found;
77 	}
78 
79 	sk = NULL;
80 found:
81 	return sk;
82 }
83 
84 static inline struct sock *l2tp_ip6_bind_lookup(struct net *net,
85 						struct in6_addr *laddr,
86 						int dif, u32 tunnel_id)
87 {
88 	struct sock *sk = __l2tp_ip6_bind_lookup(net, laddr, dif, tunnel_id);
89 	if (sk)
90 		sock_hold(sk);
91 
92 	return sk;
93 }
94 
95 /* When processing receive frames, there are two cases to
96  * consider. Data frames consist of a non-zero session-id and an
97  * optional cookie. Control frames consist of a regular L2TP header
98  * preceded by 32-bits of zeros.
99  *
100  * L2TPv3 Session Header Over IP
101  *
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  * |                           Session ID                          |
106  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107  * |               Cookie (optional, maximum 64 bits)...
108  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109  *                                                                 |
110  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111  *
112  * L2TPv3 Control Message Header Over IP
113  *
114  *  0                   1                   2                   3
115  *  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
116  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117  * |                      (32 bits of zeros)                       |
118  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119  * |T|L|x|x|S|x|x|x|x|x|x|x|  Ver  |             Length            |
120  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121  * |                     Control Connection ID                     |
122  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
123  * |               Ns              |               Nr              |
124  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
125  *
126  * All control frames are passed to userspace.
127  */
128 static int l2tp_ip6_recv(struct sk_buff *skb)
129 {
130 	struct sock *sk;
131 	u32 session_id;
132 	u32 tunnel_id;
133 	unsigned char *ptr, *optr;
134 	struct l2tp_session *session;
135 	struct l2tp_tunnel *tunnel = NULL;
136 	int length;
137 
138 	/* Point to L2TP header */
139 	optr = ptr = skb->data;
140 
141 	if (!pskb_may_pull(skb, 4))
142 		goto discard;
143 
144 	session_id = ntohl(*((__be32 *) ptr));
145 	ptr += 4;
146 
147 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
148 	 * the session_id. If it is 0, the packet is a L2TP control
149 	 * frame and the session_id value can be discarded.
150 	 */
151 	if (session_id == 0) {
152 		__skb_pull(skb, 4);
153 		goto pass_up;
154 	}
155 
156 	/* Ok, this is a data packet. Lookup the session. */
157 	session = l2tp_session_find(&init_net, NULL, session_id);
158 	if (session == NULL)
159 		goto discard;
160 
161 	tunnel = session->tunnel;
162 	if (tunnel == NULL)
163 		goto discard;
164 
165 	/* Trace packet contents, if enabled */
166 	if (tunnel->debug & L2TP_MSG_DATA) {
167 		length = min(32u, skb->len);
168 		if (!pskb_may_pull(skb, length))
169 			goto discard;
170 
171 		pr_debug("%s: ip recv\n", tunnel->name);
172 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
173 	}
174 
175 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
176 			 tunnel->recv_payload_hook);
177 	return 0;
178 
179 pass_up:
180 	/* Get the tunnel_id from the L2TP header */
181 	if (!pskb_may_pull(skb, 12))
182 		goto discard;
183 
184 	if ((skb->data[0] & 0xc0) != 0xc0)
185 		goto discard;
186 
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 ipv6hdr *iph = ipv6_hdr(skb);
193 
194 		read_lock_bh(&l2tp_ip6_lock);
195 		sk = __l2tp_ip6_bind_lookup(&init_net, &iph->daddr,
196 					    0, tunnel_id);
197 		read_unlock_bh(&l2tp_ip6_lock);
198 	}
199 
200 	if (sk == NULL)
201 		goto discard;
202 
203 	sock_hold(sk);
204 
205 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
206 		goto discard_put;
207 
208 	nf_reset(skb);
209 
210 	return sk_receive_skb(sk, skb, 1);
211 
212 discard_put:
213 	sock_put(sk);
214 
215 discard:
216 	kfree_skb(skb);
217 	return 0;
218 }
219 
220 static int l2tp_ip6_open(struct sock *sk)
221 {
222 	/* Prevent autobind. We don't have ports. */
223 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
224 
225 	write_lock_bh(&l2tp_ip6_lock);
226 	sk_add_node(sk, &l2tp_ip6_table);
227 	write_unlock_bh(&l2tp_ip6_lock);
228 
229 	return 0;
230 }
231 
232 static void l2tp_ip6_close(struct sock *sk, long timeout)
233 {
234 	write_lock_bh(&l2tp_ip6_lock);
235 	hlist_del_init(&sk->sk_bind_node);
236 	sk_del_node_init(sk);
237 	write_unlock_bh(&l2tp_ip6_lock);
238 
239 	sk_common_release(sk);
240 }
241 
242 static void l2tp_ip6_destroy_sock(struct sock *sk)
243 {
244 	struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
245 
246 	lock_sock(sk);
247 	ip6_flush_pending_frames(sk);
248 	release_sock(sk);
249 
250 	if (tunnel) {
251 		l2tp_tunnel_closeall(tunnel);
252 		sock_put(sk);
253 	}
254 
255 	inet6_destroy_sock(sk);
256 }
257 
258 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
259 {
260 	struct inet_sock *inet = inet_sk(sk);
261 	struct ipv6_pinfo *np = inet6_sk(sk);
262 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
263 	__be32 v4addr = 0;
264 	int addr_type;
265 	int err;
266 
267 	if (!sock_flag(sk, SOCK_ZAPPED))
268 		return -EINVAL;
269 	if (addr->l2tp_family != AF_INET6)
270 		return -EINVAL;
271 	if (addr_len < sizeof(*addr))
272 		return -EINVAL;
273 
274 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
275 
276 	/* l2tp_ip6 sockets are IPv6 only */
277 	if (addr_type == IPV6_ADDR_MAPPED)
278 		return -EADDRNOTAVAIL;
279 
280 	/* L2TP is point-point, not multicast */
281 	if (addr_type & IPV6_ADDR_MULTICAST)
282 		return -EADDRNOTAVAIL;
283 
284 	err = -EADDRINUSE;
285 	read_lock_bh(&l2tp_ip6_lock);
286 	if (__l2tp_ip6_bind_lookup(&init_net, &addr->l2tp_addr,
287 				   sk->sk_bound_dev_if, addr->l2tp_conn_id))
288 		goto out_in_use;
289 	read_unlock_bh(&l2tp_ip6_lock);
290 
291 	lock_sock(sk);
292 
293 	err = -EINVAL;
294 	if (sk->sk_state != TCP_CLOSE)
295 		goto out_unlock;
296 
297 	/* Check if the address belongs to the host. */
298 	rcu_read_lock();
299 	if (addr_type != IPV6_ADDR_ANY) {
300 		struct net_device *dev = NULL;
301 
302 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
303 			if (addr_len >= sizeof(struct sockaddr_in6) &&
304 			    addr->l2tp_scope_id) {
305 				/* Override any existing binding, if another
306 				 * one is supplied by user.
307 				 */
308 				sk->sk_bound_dev_if = addr->l2tp_scope_id;
309 			}
310 
311 			/* Binding to link-local address requires an
312 			   interface */
313 			if (!sk->sk_bound_dev_if)
314 				goto out_unlock_rcu;
315 
316 			err = -ENODEV;
317 			dev = dev_get_by_index_rcu(sock_net(sk),
318 						   sk->sk_bound_dev_if);
319 			if (!dev)
320 				goto out_unlock_rcu;
321 		}
322 
323 		/* ipv4 addr of the socket is invalid.  Only the
324 		 * unspecified and mapped address have a v4 equivalent.
325 		 */
326 		v4addr = LOOPBACK4_IPV6;
327 		err = -EADDRNOTAVAIL;
328 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
329 			goto out_unlock_rcu;
330 	}
331 	rcu_read_unlock();
332 
333 	inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
334 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
335 	np->saddr = addr->l2tp_addr;
336 
337 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
338 
339 	write_lock_bh(&l2tp_ip6_lock);
340 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
341 	sk_del_node_init(sk);
342 	write_unlock_bh(&l2tp_ip6_lock);
343 
344 	sock_reset_flag(sk, SOCK_ZAPPED);
345 	release_sock(sk);
346 	return 0;
347 
348 out_unlock_rcu:
349 	rcu_read_unlock();
350 out_unlock:
351 	release_sock(sk);
352 	return err;
353 
354 out_in_use:
355 	read_unlock_bh(&l2tp_ip6_lock);
356 	return err;
357 }
358 
359 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
360 			    int addr_len)
361 {
362 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
363 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
364 	struct in6_addr	*daddr;
365 	int	addr_type;
366 	int rc;
367 
368 	if (sock_flag(sk, SOCK_ZAPPED)) /* Must bind first - autobinding does not work */
369 		return -EINVAL;
370 
371 	if (addr_len < sizeof(*lsa))
372 		return -EINVAL;
373 
374 	addr_type = ipv6_addr_type(&usin->sin6_addr);
375 	if (addr_type & IPV6_ADDR_MULTICAST)
376 		return -EINVAL;
377 
378 	if (addr_type & IPV6_ADDR_MAPPED) {
379 		daddr = &usin->sin6_addr;
380 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
381 			return -EINVAL;
382 	}
383 
384 	rc = ip6_datagram_connect(sk, uaddr, addr_len);
385 
386 	lock_sock(sk);
387 
388 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
389 
390 	write_lock_bh(&l2tp_ip6_lock);
391 	hlist_del_init(&sk->sk_bind_node);
392 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
393 	write_unlock_bh(&l2tp_ip6_lock);
394 
395 	release_sock(sk);
396 
397 	return rc;
398 }
399 
400 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
401 {
402 	if (sock_flag(sk, SOCK_ZAPPED))
403 		return 0;
404 
405 	return udp_disconnect(sk, flags);
406 }
407 
408 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
409 			    int *uaddr_len, int peer)
410 {
411 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
412 	struct sock *sk = sock->sk;
413 	struct ipv6_pinfo *np = inet6_sk(sk);
414 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
415 
416 	lsa->l2tp_family = AF_INET6;
417 	lsa->l2tp_flowinfo = 0;
418 	lsa->l2tp_scope_id = 0;
419 	lsa->l2tp_unused = 0;
420 	if (peer) {
421 		if (!lsk->peer_conn_id)
422 			return -ENOTCONN;
423 		lsa->l2tp_conn_id = lsk->peer_conn_id;
424 		lsa->l2tp_addr = sk->sk_v6_daddr;
425 		if (np->sndflow)
426 			lsa->l2tp_flowinfo = np->flow_label;
427 	} else {
428 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
429 			lsa->l2tp_addr = np->saddr;
430 		else
431 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
432 
433 		lsa->l2tp_conn_id = lsk->conn_id;
434 	}
435 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
436 		lsa->l2tp_scope_id = sk->sk_bound_dev_if;
437 	*uaddr_len = sizeof(*lsa);
438 	return 0;
439 }
440 
441 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
442 {
443 	int rc;
444 
445 	/* Charge it to the socket, dropping if the queue is full. */
446 	rc = sock_queue_rcv_skb(sk, skb);
447 	if (rc < 0)
448 		goto drop;
449 
450 	return 0;
451 
452 drop:
453 	IP_INC_STATS(&init_net, IPSTATS_MIB_INDISCARDS);
454 	kfree_skb(skb);
455 	return -1;
456 }
457 
458 static int l2tp_ip6_push_pending_frames(struct sock *sk)
459 {
460 	struct sk_buff *skb;
461 	__be32 *transhdr = NULL;
462 	int err = 0;
463 
464 	skb = skb_peek(&sk->sk_write_queue);
465 	if (skb == NULL)
466 		goto out;
467 
468 	transhdr = (__be32 *)skb_transport_header(skb);
469 	*transhdr = 0;
470 
471 	err = ip6_push_pending_frames(sk);
472 
473 out:
474 	return err;
475 }
476 
477 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
478  * control frames.
479  */
480 static int l2tp_ip6_sendmsg(struct kiocb *iocb, struct sock *sk,
481 			    struct msghdr *msg, size_t len)
482 {
483 	struct ipv6_txoptions opt_space;
484 	struct sockaddr_l2tpip6 *lsa =
485 		(struct sockaddr_l2tpip6 *) msg->msg_name;
486 	struct in6_addr *daddr, *final_p, final;
487 	struct ipv6_pinfo *np = inet6_sk(sk);
488 	struct ipv6_txoptions *opt = NULL;
489 	struct ip6_flowlabel *flowlabel = NULL;
490 	struct dst_entry *dst = NULL;
491 	struct flowi6 fl6;
492 	int addr_len = msg->msg_namelen;
493 	int hlimit = -1;
494 	int tclass = -1;
495 	int dontfrag = -1;
496 	int transhdrlen = 4; /* zero session-id */
497 	int ulen = len + transhdrlen;
498 	int err;
499 
500 	/* Rough check on arithmetic overflow,
501 	   better check is made in ip6_append_data().
502 	 */
503 	if (len > INT_MAX)
504 		return -EMSGSIZE;
505 
506 	/* Mirror BSD error message compatibility */
507 	if (msg->msg_flags & MSG_OOB)
508 		return -EOPNOTSUPP;
509 
510 	/*
511 	 *	Get and verify the address.
512 	 */
513 	memset(&fl6, 0, sizeof(fl6));
514 
515 	fl6.flowi6_mark = sk->sk_mark;
516 
517 	if (lsa) {
518 		if (addr_len < SIN6_LEN_RFC2133)
519 			return -EINVAL;
520 
521 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
522 			return -EAFNOSUPPORT;
523 
524 		daddr = &lsa->l2tp_addr;
525 		if (np->sndflow) {
526 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
527 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
528 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
529 				if (flowlabel == NULL)
530 					return -EINVAL;
531 			}
532 		}
533 
534 		/*
535 		 * Otherwise it will be difficult to maintain
536 		 * sk->sk_dst_cache.
537 		 */
538 		if (sk->sk_state == TCP_ESTABLISHED &&
539 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
540 			daddr = &sk->sk_v6_daddr;
541 
542 		if (addr_len >= sizeof(struct sockaddr_in6) &&
543 		    lsa->l2tp_scope_id &&
544 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
545 			fl6.flowi6_oif = lsa->l2tp_scope_id;
546 	} else {
547 		if (sk->sk_state != TCP_ESTABLISHED)
548 			return -EDESTADDRREQ;
549 
550 		daddr = &sk->sk_v6_daddr;
551 		fl6.flowlabel = np->flow_label;
552 	}
553 
554 	if (fl6.flowi6_oif == 0)
555 		fl6.flowi6_oif = sk->sk_bound_dev_if;
556 
557 	if (msg->msg_controllen) {
558 		opt = &opt_space;
559 		memset(opt, 0, sizeof(struct ipv6_txoptions));
560 		opt->tot_len = sizeof(struct ipv6_txoptions);
561 
562 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
563 					    &hlimit, &tclass, &dontfrag);
564 		if (err < 0) {
565 			fl6_sock_release(flowlabel);
566 			return err;
567 		}
568 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
569 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
570 			if (flowlabel == NULL)
571 				return -EINVAL;
572 		}
573 		if (!(opt->opt_nflen|opt->opt_flen))
574 			opt = NULL;
575 	}
576 
577 	if (opt == NULL)
578 		opt = np->opt;
579 	if (flowlabel)
580 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
581 	opt = ipv6_fixup_options(&opt_space, opt);
582 
583 	fl6.flowi6_proto = sk->sk_protocol;
584 	if (!ipv6_addr_any(daddr))
585 		fl6.daddr = *daddr;
586 	else
587 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
588 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
589 		fl6.saddr = np->saddr;
590 
591 	final_p = fl6_update_dst(&fl6, opt, &final);
592 
593 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
594 		fl6.flowi6_oif = np->mcast_oif;
595 	else if (!fl6.flowi6_oif)
596 		fl6.flowi6_oif = np->ucast_oif;
597 
598 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
599 
600 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
601 	if (IS_ERR(dst)) {
602 		err = PTR_ERR(dst);
603 		goto out;
604 	}
605 
606 	if (hlimit < 0) {
607 		if (ipv6_addr_is_multicast(&fl6.daddr))
608 			hlimit = np->mcast_hops;
609 		else
610 			hlimit = np->hop_limit;
611 		if (hlimit < 0)
612 			hlimit = ip6_dst_hoplimit(dst);
613 	}
614 
615 	if (tclass < 0)
616 		tclass = np->tclass;
617 
618 	if (dontfrag < 0)
619 		dontfrag = np->dontfrag;
620 
621 	if (msg->msg_flags & MSG_CONFIRM)
622 		goto do_confirm;
623 
624 back_from_confirm:
625 	lock_sock(sk);
626 	err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
627 			      ulen, transhdrlen, hlimit, tclass, opt,
628 			      &fl6, (struct rt6_info *)dst,
629 			      msg->msg_flags, dontfrag);
630 	if (err)
631 		ip6_flush_pending_frames(sk);
632 	else if (!(msg->msg_flags & MSG_MORE))
633 		err = l2tp_ip6_push_pending_frames(sk);
634 	release_sock(sk);
635 done:
636 	dst_release(dst);
637 out:
638 	fl6_sock_release(flowlabel);
639 
640 	return err < 0 ? err : len;
641 
642 do_confirm:
643 	dst_confirm(dst);
644 	if (!(msg->msg_flags & MSG_PROBE) || len)
645 		goto back_from_confirm;
646 	err = 0;
647 	goto done;
648 }
649 
650 static int l2tp_ip6_recvmsg(struct kiocb *iocb, struct sock *sk,
651 			    struct msghdr *msg, size_t len, int noblock,
652 			    int flags, int *addr_len)
653 {
654 	struct ipv6_pinfo *np = inet6_sk(sk);
655 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)msg->msg_name;
656 	size_t copied = 0;
657 	int err = -EOPNOTSUPP;
658 	struct sk_buff *skb;
659 
660 	if (flags & MSG_OOB)
661 		goto out;
662 
663 	if (addr_len)
664 		*addr_len = sizeof(*lsa);
665 
666 	if (flags & MSG_ERRQUEUE)
667 		return ipv6_recv_error(sk, msg, len, addr_len);
668 
669 	skb = skb_recv_datagram(sk, flags, noblock, &err);
670 	if (!skb)
671 		goto out;
672 
673 	copied = skb->len;
674 	if (len < copied) {
675 		msg->msg_flags |= MSG_TRUNC;
676 		copied = len;
677 	}
678 
679 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
680 	if (err)
681 		goto done;
682 
683 	sock_recv_timestamp(msg, sk, skb);
684 
685 	/* Copy the address. */
686 	if (lsa) {
687 		lsa->l2tp_family = AF_INET6;
688 		lsa->l2tp_unused = 0;
689 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
690 		lsa->l2tp_flowinfo = 0;
691 		lsa->l2tp_scope_id = 0;
692 		lsa->l2tp_conn_id = 0;
693 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
694 			lsa->l2tp_scope_id = IP6CB(skb)->iif;
695 	}
696 
697 	if (np->rxopt.all)
698 		ip6_datagram_recv_ctl(sk, msg, skb);
699 
700 	if (flags & MSG_TRUNC)
701 		copied = skb->len;
702 done:
703 	skb_free_datagram(sk, skb);
704 out:
705 	return err ? err : copied;
706 }
707 
708 static struct proto l2tp_ip6_prot = {
709 	.name		   = "L2TP/IPv6",
710 	.owner		   = THIS_MODULE,
711 	.init		   = l2tp_ip6_open,
712 	.close		   = l2tp_ip6_close,
713 	.bind		   = l2tp_ip6_bind,
714 	.connect	   = l2tp_ip6_connect,
715 	.disconnect	   = l2tp_ip6_disconnect,
716 	.ioctl		   = udp_ioctl,
717 	.destroy	   = l2tp_ip6_destroy_sock,
718 	.setsockopt	   = ipv6_setsockopt,
719 	.getsockopt	   = ipv6_getsockopt,
720 	.sendmsg	   = l2tp_ip6_sendmsg,
721 	.recvmsg	   = l2tp_ip6_recvmsg,
722 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
723 	.hash		   = inet_hash,
724 	.unhash		   = inet_unhash,
725 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
726 #ifdef CONFIG_COMPAT
727 	.compat_setsockopt = compat_ipv6_setsockopt,
728 	.compat_getsockopt = compat_ipv6_getsockopt,
729 #endif
730 };
731 
732 static const struct proto_ops l2tp_ip6_ops = {
733 	.family		   = PF_INET6,
734 	.owner		   = THIS_MODULE,
735 	.release	   = inet6_release,
736 	.bind		   = inet6_bind,
737 	.connect	   = inet_dgram_connect,
738 	.socketpair	   = sock_no_socketpair,
739 	.accept		   = sock_no_accept,
740 	.getname	   = l2tp_ip6_getname,
741 	.poll		   = datagram_poll,
742 	.ioctl		   = inet6_ioctl,
743 	.listen		   = sock_no_listen,
744 	.shutdown	   = inet_shutdown,
745 	.setsockopt	   = sock_common_setsockopt,
746 	.getsockopt	   = sock_common_getsockopt,
747 	.sendmsg	   = inet_sendmsg,
748 	.recvmsg	   = sock_common_recvmsg,
749 	.mmap		   = sock_no_mmap,
750 	.sendpage	   = sock_no_sendpage,
751 #ifdef CONFIG_COMPAT
752 	.compat_setsockopt = compat_sock_common_setsockopt,
753 	.compat_getsockopt = compat_sock_common_getsockopt,
754 #endif
755 };
756 
757 static struct inet_protosw l2tp_ip6_protosw = {
758 	.type		= SOCK_DGRAM,
759 	.protocol	= IPPROTO_L2TP,
760 	.prot		= &l2tp_ip6_prot,
761 	.ops		= &l2tp_ip6_ops,
762 	.no_check	= 0,
763 };
764 
765 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
766 	.handler	= l2tp_ip6_recv,
767 };
768 
769 static int __init l2tp_ip6_init(void)
770 {
771 	int err;
772 
773 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
774 
775 	err = proto_register(&l2tp_ip6_prot, 1);
776 	if (err != 0)
777 		goto out;
778 
779 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
780 	if (err)
781 		goto out1;
782 
783 	inet6_register_protosw(&l2tp_ip6_protosw);
784 	return 0;
785 
786 out1:
787 	proto_unregister(&l2tp_ip6_prot);
788 out:
789 	return err;
790 }
791 
792 static void __exit l2tp_ip6_exit(void)
793 {
794 	inet6_unregister_protosw(&l2tp_ip6_protosw);
795 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
796 	proto_unregister(&l2tp_ip6_prot);
797 }
798 
799 module_init(l2tp_ip6_init);
800 module_exit(l2tp_ip6_exit);
801 
802 MODULE_LICENSE("GPL");
803 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
804 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
805 MODULE_VERSION("1.0");
806 
807 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
808  * enums
809  */
810 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
811