xref: /openbmc/linux/net/l2tp/l2tp_ip6.c (revision 9b93eb47)
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/inet6_hashtables.h>
29 #include <net/tcp_states.h>
30 #include <net/protocol.h>
31 #include <net/xfrm.h>
32 
33 #include <net/transp_v6.h>
34 #include <net/addrconf.h>
35 #include <net/ip6_route.h>
36 
37 #include "l2tp_core.h"
38 
39 struct l2tp_ip6_sock {
40 	/* inet_sock has to be the first member of l2tp_ip6_sock */
41 	struct inet_sock	inet;
42 
43 	u32			conn_id;
44 	u32			peer_conn_id;
45 
46 	/* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
47 	   inet6_sk_generic */
48 	struct ipv6_pinfo	inet6;
49 };
50 
51 static DEFINE_RWLOCK(l2tp_ip6_lock);
52 static struct hlist_head l2tp_ip6_table;
53 static struct hlist_head l2tp_ip6_bind_table;
54 
55 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
56 {
57 	return (struct l2tp_ip6_sock *)sk;
58 }
59 
60 static struct sock *__l2tp_ip6_bind_lookup(const struct net *net,
61 					   const struct in6_addr *laddr,
62 					   const struct in6_addr *raddr,
63 					   int dif, u32 tunnel_id)
64 {
65 	struct sock *sk;
66 
67 	sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
68 		const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
69 		const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
70 		const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
71 
72 		if (!net_eq(sock_net(sk), net))
73 			continue;
74 
75 		if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
76 			continue;
77 
78 		if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
79 		    !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
80 			continue;
81 
82 		if (!ipv6_addr_any(sk_raddr) && raddr &&
83 		    !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
84 			continue;
85 
86 		if (l2tp->conn_id != tunnel_id)
87 			continue;
88 
89 		goto found;
90 	}
91 
92 	sk = NULL;
93 found:
94 	return sk;
95 }
96 
97 /* When processing receive frames, there are two cases to
98  * consider. Data frames consist of a non-zero session-id and an
99  * optional cookie. Control frames consist of a regular L2TP header
100  * preceded by 32-bits of zeros.
101  *
102  * L2TPv3 Session Header Over IP
103  *
104  *  0                   1                   2                   3
105  *  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
106  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107  * |                           Session ID                          |
108  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109  * |               Cookie (optional, maximum 64 bits)...
110  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111  *                                                                 |
112  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113  *
114  * L2TPv3 Control Message Header Over IP
115  *
116  *  0                   1                   2                   3
117  *  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
118  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119  * |                      (32 bits of zeros)                       |
120  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121  * |T|L|x|x|S|x|x|x|x|x|x|x|  Ver  |             Length            |
122  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
123  * |                     Control Connection ID                     |
124  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
125  * |               Ns              |               Nr              |
126  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
127  *
128  * All control frames are passed to userspace.
129  */
130 static int l2tp_ip6_recv(struct sk_buff *skb)
131 {
132 	struct net *net = dev_net(skb->dev);
133 	struct sock *sk;
134 	u32 session_id;
135 	u32 tunnel_id;
136 	unsigned char *ptr, *optr;
137 	struct l2tp_session *session;
138 	struct l2tp_tunnel *tunnel = NULL;
139 	struct ipv6hdr *iph;
140 	int length;
141 
142 	if (!pskb_may_pull(skb, 4))
143 		goto discard;
144 
145 	/* Point to L2TP header */
146 	optr = ptr = skb->data;
147 	session_id = ntohl(*((__be32 *) ptr));
148 	ptr += 4;
149 
150 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
151 	 * the session_id. If it is 0, the packet is a L2TP control
152 	 * frame and the session_id value can be discarded.
153 	 */
154 	if (session_id == 0) {
155 		__skb_pull(skb, 4);
156 		goto pass_up;
157 	}
158 
159 	/* Ok, this is a data packet. Lookup the session. */
160 	session = l2tp_session_get(net, session_id);
161 	if (!session)
162 		goto discard;
163 
164 	tunnel = session->tunnel;
165 	if (!tunnel)
166 		goto discard_sess;
167 
168 	/* Trace packet contents, if enabled */
169 	if (tunnel->debug & L2TP_MSG_DATA) {
170 		length = min(32u, skb->len);
171 		if (!pskb_may_pull(skb, length))
172 			goto discard_sess;
173 
174 		/* Point to L2TP header */
175 		optr = ptr = skb->data;
176 		ptr += 4;
177 		pr_debug("%s: ip recv\n", tunnel->name);
178 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
179 	}
180 
181 	if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
182 		goto discard_sess;
183 
184 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
185 	l2tp_session_dec_refcount(session);
186 
187 	return 0;
188 
189 pass_up:
190 	/* Get the tunnel_id from the L2TP header */
191 	if (!pskb_may_pull(skb, 12))
192 		goto discard;
193 
194 	if ((skb->data[0] & 0xc0) != 0xc0)
195 		goto discard;
196 
197 	tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
198 	iph = ipv6_hdr(skb);
199 
200 	read_lock_bh(&l2tp_ip6_lock);
201 	sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
202 				    inet6_iif(skb), tunnel_id);
203 	if (!sk) {
204 		read_unlock_bh(&l2tp_ip6_lock);
205 		goto discard;
206 	}
207 	sock_hold(sk);
208 	read_unlock_bh(&l2tp_ip6_lock);
209 
210 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
211 		goto discard_put;
212 
213 	nf_reset(skb);
214 
215 	return sk_receive_skb(sk, skb, 1);
216 
217 discard_sess:
218 	l2tp_session_dec_refcount(session);
219 	goto discard;
220 
221 discard_put:
222 	sock_put(sk);
223 
224 discard:
225 	kfree_skb(skb);
226 	return 0;
227 }
228 
229 static int l2tp_ip6_open(struct sock *sk)
230 {
231 	/* Prevent autobind. We don't have ports. */
232 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
233 
234 	write_lock_bh(&l2tp_ip6_lock);
235 	sk_add_node(sk, &l2tp_ip6_table);
236 	write_unlock_bh(&l2tp_ip6_lock);
237 
238 	return 0;
239 }
240 
241 static void l2tp_ip6_close(struct sock *sk, long timeout)
242 {
243 	write_lock_bh(&l2tp_ip6_lock);
244 	hlist_del_init(&sk->sk_bind_node);
245 	sk_del_node_init(sk);
246 	write_unlock_bh(&l2tp_ip6_lock);
247 
248 	sk_common_release(sk);
249 }
250 
251 static void l2tp_ip6_destroy_sock(struct sock *sk)
252 {
253 	struct l2tp_tunnel *tunnel = sk->sk_user_data;
254 
255 	lock_sock(sk);
256 	ip6_flush_pending_frames(sk);
257 	release_sock(sk);
258 
259 	if (tunnel)
260 		l2tp_tunnel_delete(tunnel);
261 
262 	inet6_destroy_sock(sk);
263 }
264 
265 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
266 {
267 	struct inet_sock *inet = inet_sk(sk);
268 	struct ipv6_pinfo *np = inet6_sk(sk);
269 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
270 	struct net *net = sock_net(sk);
271 	__be32 v4addr = 0;
272 	int bound_dev_if;
273 	int addr_type;
274 	int err;
275 
276 	if (addr->l2tp_family != AF_INET6)
277 		return -EINVAL;
278 	if (addr_len < sizeof(*addr))
279 		return -EINVAL;
280 
281 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
282 
283 	/* l2tp_ip6 sockets are IPv6 only */
284 	if (addr_type == IPV6_ADDR_MAPPED)
285 		return -EADDRNOTAVAIL;
286 
287 	/* L2TP is point-point, not multicast */
288 	if (addr_type & IPV6_ADDR_MULTICAST)
289 		return -EADDRNOTAVAIL;
290 
291 	lock_sock(sk);
292 
293 	err = -EINVAL;
294 	if (!sock_flag(sk, SOCK_ZAPPED))
295 		goto out_unlock;
296 
297 	if (sk->sk_state != TCP_CLOSE)
298 		goto out_unlock;
299 
300 	bound_dev_if = sk->sk_bound_dev_if;
301 
302 	/* Check if the address belongs to the host. */
303 	rcu_read_lock();
304 	if (addr_type != IPV6_ADDR_ANY) {
305 		struct net_device *dev = NULL;
306 
307 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
308 			if (addr->l2tp_scope_id)
309 				bound_dev_if = addr->l2tp_scope_id;
310 
311 			/* Binding to link-local address requires an
312 			 * interface.
313 			 */
314 			if (!bound_dev_if)
315 				goto out_unlock_rcu;
316 
317 			err = -ENODEV;
318 			dev = dev_get_by_index_rcu(sock_net(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 	write_lock_bh(&l2tp_ip6_lock);
334 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
335 				   addr->l2tp_conn_id)) {
336 		write_unlock_bh(&l2tp_ip6_lock);
337 		err = -EADDRINUSE;
338 		goto out_unlock;
339 	}
340 
341 	inet->inet_saddr = v4addr;
342 	inet->inet_rcv_saddr = v4addr;
343 	sk->sk_bound_dev_if = bound_dev_if;
344 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
345 	np->saddr = addr->l2tp_addr;
346 
347 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
348 
349 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
350 	sk_del_node_init(sk);
351 	write_unlock_bh(&l2tp_ip6_lock);
352 
353 	sock_reset_flag(sk, SOCK_ZAPPED);
354 	release_sock(sk);
355 	return 0;
356 
357 out_unlock_rcu:
358 	rcu_read_unlock();
359 out_unlock:
360 	release_sock(sk);
361 
362 	return err;
363 }
364 
365 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
366 			    int addr_len)
367 {
368 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
369 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
370 	struct in6_addr	*daddr;
371 	int	addr_type;
372 	int rc;
373 
374 	if (addr_len < sizeof(*lsa))
375 		return -EINVAL;
376 
377 	if (usin->sin6_family != AF_INET6)
378 		return -EINVAL;
379 
380 	addr_type = ipv6_addr_type(&usin->sin6_addr);
381 	if (addr_type & IPV6_ADDR_MULTICAST)
382 		return -EINVAL;
383 
384 	if (addr_type & IPV6_ADDR_MAPPED) {
385 		daddr = &usin->sin6_addr;
386 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
387 			return -EINVAL;
388 	}
389 
390 	lock_sock(sk);
391 
392 	 /* Must bind first - autobinding does not work */
393 	if (sock_flag(sk, SOCK_ZAPPED)) {
394 		rc = -EINVAL;
395 		goto out_sk;
396 	}
397 
398 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
399 	if (rc < 0)
400 		goto out_sk;
401 
402 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
403 
404 	write_lock_bh(&l2tp_ip6_lock);
405 	hlist_del_init(&sk->sk_bind_node);
406 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
407 	write_unlock_bh(&l2tp_ip6_lock);
408 
409 out_sk:
410 	release_sock(sk);
411 
412 	return rc;
413 }
414 
415 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
416 {
417 	if (sock_flag(sk, SOCK_ZAPPED))
418 		return 0;
419 
420 	return __udp_disconnect(sk, flags);
421 }
422 
423 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
424 			    int peer)
425 {
426 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
427 	struct sock *sk = sock->sk;
428 	struct ipv6_pinfo *np = inet6_sk(sk);
429 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
430 
431 	lsa->l2tp_family = AF_INET6;
432 	lsa->l2tp_flowinfo = 0;
433 	lsa->l2tp_scope_id = 0;
434 	lsa->l2tp_unused = 0;
435 	if (peer) {
436 		if (!lsk->peer_conn_id)
437 			return -ENOTCONN;
438 		lsa->l2tp_conn_id = lsk->peer_conn_id;
439 		lsa->l2tp_addr = sk->sk_v6_daddr;
440 		if (np->sndflow)
441 			lsa->l2tp_flowinfo = np->flow_label;
442 	} else {
443 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
444 			lsa->l2tp_addr = np->saddr;
445 		else
446 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
447 
448 		lsa->l2tp_conn_id = lsk->conn_id;
449 	}
450 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
451 		lsa->l2tp_scope_id = sk->sk_bound_dev_if;
452 	return sizeof(*lsa);
453 }
454 
455 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
456 {
457 	int rc;
458 
459 	/* Charge it to the socket, dropping if the queue is full. */
460 	rc = sock_queue_rcv_skb(sk, skb);
461 	if (rc < 0)
462 		goto drop;
463 
464 	return 0;
465 
466 drop:
467 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
468 	kfree_skb(skb);
469 	return -1;
470 }
471 
472 static int l2tp_ip6_push_pending_frames(struct sock *sk)
473 {
474 	struct sk_buff *skb;
475 	__be32 *transhdr = NULL;
476 	int err = 0;
477 
478 	skb = skb_peek(&sk->sk_write_queue);
479 	if (skb == NULL)
480 		goto out;
481 
482 	transhdr = (__be32 *)skb_transport_header(skb);
483 	*transhdr = 0;
484 
485 	err = ip6_push_pending_frames(sk);
486 
487 out:
488 	return err;
489 }
490 
491 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
492  * control frames.
493  */
494 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
495 {
496 	struct ipv6_txoptions opt_space;
497 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
498 	struct in6_addr *daddr, *final_p, final;
499 	struct ipv6_pinfo *np = inet6_sk(sk);
500 	struct ipv6_txoptions *opt_to_free = NULL;
501 	struct ipv6_txoptions *opt = NULL;
502 	struct ip6_flowlabel *flowlabel = NULL;
503 	struct dst_entry *dst = NULL;
504 	struct flowi6 fl6;
505 	struct ipcm6_cookie ipc6;
506 	int addr_len = msg->msg_namelen;
507 	int transhdrlen = 4; /* zero session-id */
508 	int ulen = len + transhdrlen;
509 	int err;
510 
511 	/* Rough check on arithmetic overflow,
512 	   better check is made in ip6_append_data().
513 	 */
514 	if (len > INT_MAX)
515 		return -EMSGSIZE;
516 
517 	/* Mirror BSD error message compatibility */
518 	if (msg->msg_flags & MSG_OOB)
519 		return -EOPNOTSUPP;
520 
521 	/*
522 	 *	Get and verify the address.
523 	 */
524 	memset(&fl6, 0, sizeof(fl6));
525 
526 	fl6.flowi6_mark = sk->sk_mark;
527 	fl6.flowi6_uid = sk->sk_uid;
528 
529 	ipcm6_init(&ipc6);
530 
531 	if (lsa) {
532 		if (addr_len < SIN6_LEN_RFC2133)
533 			return -EINVAL;
534 
535 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
536 			return -EAFNOSUPPORT;
537 
538 		daddr = &lsa->l2tp_addr;
539 		if (np->sndflow) {
540 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
541 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
542 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
543 				if (flowlabel == NULL)
544 					return -EINVAL;
545 			}
546 		}
547 
548 		/*
549 		 * Otherwise it will be difficult to maintain
550 		 * sk->sk_dst_cache.
551 		 */
552 		if (sk->sk_state == TCP_ESTABLISHED &&
553 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
554 			daddr = &sk->sk_v6_daddr;
555 
556 		if (addr_len >= sizeof(struct sockaddr_in6) &&
557 		    lsa->l2tp_scope_id &&
558 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
559 			fl6.flowi6_oif = lsa->l2tp_scope_id;
560 	} else {
561 		if (sk->sk_state != TCP_ESTABLISHED)
562 			return -EDESTADDRREQ;
563 
564 		daddr = &sk->sk_v6_daddr;
565 		fl6.flowlabel = np->flow_label;
566 	}
567 
568 	if (fl6.flowi6_oif == 0)
569 		fl6.flowi6_oif = sk->sk_bound_dev_if;
570 
571 	if (msg->msg_controllen) {
572 		opt = &opt_space;
573 		memset(opt, 0, sizeof(struct ipv6_txoptions));
574 		opt->tot_len = sizeof(struct ipv6_txoptions);
575 		ipc6.opt = opt;
576 
577 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
578 		if (err < 0) {
579 			fl6_sock_release(flowlabel);
580 			return err;
581 		}
582 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
583 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
584 			if (flowlabel == NULL)
585 				return -EINVAL;
586 		}
587 		if (!(opt->opt_nflen|opt->opt_flen))
588 			opt = NULL;
589 	}
590 
591 	if (!opt) {
592 		opt = txopt_get(np);
593 		opt_to_free = opt;
594 	}
595 	if (flowlabel)
596 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
597 	opt = ipv6_fixup_options(&opt_space, opt);
598 	ipc6.opt = opt;
599 
600 	fl6.flowi6_proto = sk->sk_protocol;
601 	if (!ipv6_addr_any(daddr))
602 		fl6.daddr = *daddr;
603 	else
604 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
605 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
606 		fl6.saddr = np->saddr;
607 
608 	final_p = fl6_update_dst(&fl6, opt, &final);
609 
610 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
611 		fl6.flowi6_oif = np->mcast_oif;
612 	else if (!fl6.flowi6_oif)
613 		fl6.flowi6_oif = np->ucast_oif;
614 
615 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
616 
617 	if (ipc6.tclass < 0)
618 		ipc6.tclass = np->tclass;
619 
620 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
621 
622 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
623 	if (IS_ERR(dst)) {
624 		err = PTR_ERR(dst);
625 		goto out;
626 	}
627 
628 	if (ipc6.hlimit < 0)
629 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
630 
631 	if (ipc6.dontfrag < 0)
632 		ipc6.dontfrag = np->dontfrag;
633 
634 	if (msg->msg_flags & MSG_CONFIRM)
635 		goto do_confirm;
636 
637 back_from_confirm:
638 	lock_sock(sk);
639 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
640 			      ulen, transhdrlen, &ipc6,
641 			      &fl6, (struct rt6_info *)dst,
642 			      msg->msg_flags);
643 	if (err)
644 		ip6_flush_pending_frames(sk);
645 	else if (!(msg->msg_flags & MSG_MORE))
646 		err = l2tp_ip6_push_pending_frames(sk);
647 	release_sock(sk);
648 done:
649 	dst_release(dst);
650 out:
651 	fl6_sock_release(flowlabel);
652 	txopt_put(opt_to_free);
653 
654 	return err < 0 ? err : len;
655 
656 do_confirm:
657 	if (msg->msg_flags & MSG_PROBE)
658 		dst_confirm_neigh(dst, &fl6.daddr);
659 	if (!(msg->msg_flags & MSG_PROBE) || len)
660 		goto back_from_confirm;
661 	err = 0;
662 	goto done;
663 }
664 
665 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
666 			    int noblock, int flags, int *addr_len)
667 {
668 	struct ipv6_pinfo *np = inet6_sk(sk);
669 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
670 	size_t copied = 0;
671 	int err = -EOPNOTSUPP;
672 	struct sk_buff *skb;
673 
674 	if (flags & MSG_OOB)
675 		goto out;
676 
677 	if (flags & MSG_ERRQUEUE)
678 		return ipv6_recv_error(sk, msg, len, addr_len);
679 
680 	skb = skb_recv_datagram(sk, flags, noblock, &err);
681 	if (!skb)
682 		goto out;
683 
684 	copied = skb->len;
685 	if (len < copied) {
686 		msg->msg_flags |= MSG_TRUNC;
687 		copied = len;
688 	}
689 
690 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
691 	if (err)
692 		goto done;
693 
694 	sock_recv_timestamp(msg, sk, skb);
695 
696 	/* Copy the address. */
697 	if (lsa) {
698 		lsa->l2tp_family = AF_INET6;
699 		lsa->l2tp_unused = 0;
700 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
701 		lsa->l2tp_flowinfo = 0;
702 		lsa->l2tp_scope_id = 0;
703 		lsa->l2tp_conn_id = 0;
704 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
705 			lsa->l2tp_scope_id = inet6_iif(skb);
706 		*addr_len = sizeof(*lsa);
707 	}
708 
709 	if (np->rxopt.all)
710 		ip6_datagram_recv_ctl(sk, msg, skb);
711 
712 	if (flags & MSG_TRUNC)
713 		copied = skb->len;
714 done:
715 	skb_free_datagram(sk, skb);
716 out:
717 	return err ? err : copied;
718 }
719 
720 static struct proto l2tp_ip6_prot = {
721 	.name		   = "L2TP/IPv6",
722 	.owner		   = THIS_MODULE,
723 	.init		   = l2tp_ip6_open,
724 	.close		   = l2tp_ip6_close,
725 	.bind		   = l2tp_ip6_bind,
726 	.connect	   = l2tp_ip6_connect,
727 	.disconnect	   = l2tp_ip6_disconnect,
728 	.ioctl		   = l2tp_ioctl,
729 	.destroy	   = l2tp_ip6_destroy_sock,
730 	.setsockopt	   = ipv6_setsockopt,
731 	.getsockopt	   = ipv6_getsockopt,
732 	.sendmsg	   = l2tp_ip6_sendmsg,
733 	.recvmsg	   = l2tp_ip6_recvmsg,
734 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
735 	.hash		   = inet6_hash,
736 	.unhash		   = inet_unhash,
737 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
738 #ifdef CONFIG_COMPAT
739 	.compat_setsockopt = compat_ipv6_setsockopt,
740 	.compat_getsockopt = compat_ipv6_getsockopt,
741 #endif
742 };
743 
744 static const struct proto_ops l2tp_ip6_ops = {
745 	.family		   = PF_INET6,
746 	.owner		   = THIS_MODULE,
747 	.release	   = inet6_release,
748 	.bind		   = inet6_bind,
749 	.connect	   = inet_dgram_connect,
750 	.socketpair	   = sock_no_socketpair,
751 	.accept		   = sock_no_accept,
752 	.getname	   = l2tp_ip6_getname,
753 	.poll		   = datagram_poll,
754 	.ioctl		   = inet6_ioctl,
755 	.gettstamp	   = sock_gettstamp,
756 	.listen		   = sock_no_listen,
757 	.shutdown	   = inet_shutdown,
758 	.setsockopt	   = sock_common_setsockopt,
759 	.getsockopt	   = sock_common_getsockopt,
760 	.sendmsg	   = inet_sendmsg,
761 	.recvmsg	   = sock_common_recvmsg,
762 	.mmap		   = sock_no_mmap,
763 	.sendpage	   = sock_no_sendpage,
764 #ifdef CONFIG_COMPAT
765 	.compat_setsockopt = compat_sock_common_setsockopt,
766 	.compat_getsockopt = compat_sock_common_getsockopt,
767 #endif
768 };
769 
770 static struct inet_protosw l2tp_ip6_protosw = {
771 	.type		= SOCK_DGRAM,
772 	.protocol	= IPPROTO_L2TP,
773 	.prot		= &l2tp_ip6_prot,
774 	.ops		= &l2tp_ip6_ops,
775 };
776 
777 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
778 	.handler	= l2tp_ip6_recv,
779 };
780 
781 static int __init l2tp_ip6_init(void)
782 {
783 	int err;
784 
785 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
786 
787 	err = proto_register(&l2tp_ip6_prot, 1);
788 	if (err != 0)
789 		goto out;
790 
791 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
792 	if (err)
793 		goto out1;
794 
795 	inet6_register_protosw(&l2tp_ip6_protosw);
796 	return 0;
797 
798 out1:
799 	proto_unregister(&l2tp_ip6_prot);
800 out:
801 	return err;
802 }
803 
804 static void __exit l2tp_ip6_exit(void)
805 {
806 	inet6_unregister_protosw(&l2tp_ip6_protosw);
807 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
808 	proto_unregister(&l2tp_ip6_prot);
809 }
810 
811 module_init(l2tp_ip6_init);
812 module_exit(l2tp_ip6_exit);
813 
814 MODULE_LICENSE("GPL");
815 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
816 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
817 MODULE_VERSION("1.0");
818 
819 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
820  * enums
821  */
822 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
823 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);
824