xref: /openbmc/linux/net/l2tp/l2tp_ip6.c (revision a0ae2562c6c4b2721d9fddba63b7286c13517d9f)
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, NULL, 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 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
182 			 tunnel->recv_payload_hook);
183 	l2tp_session_dec_refcount(session);
184 
185 	return 0;
186 
187 pass_up:
188 	/* Get the tunnel_id from the L2TP header */
189 	if (!pskb_may_pull(skb, 12))
190 		goto discard;
191 
192 	if ((skb->data[0] & 0xc0) != 0xc0)
193 		goto discard;
194 
195 	tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
196 	iph = ipv6_hdr(skb);
197 
198 	read_lock_bh(&l2tp_ip6_lock);
199 	sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
200 				    inet6_iif(skb), tunnel_id);
201 	if (!sk) {
202 		read_unlock_bh(&l2tp_ip6_lock);
203 		goto discard;
204 	}
205 	sock_hold(sk);
206 	read_unlock_bh(&l2tp_ip6_lock);
207 
208 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
209 		goto discard_put;
210 
211 	nf_reset(skb);
212 
213 	return sk_receive_skb(sk, skb, 1);
214 
215 discard_sess:
216 	l2tp_session_dec_refcount(session);
217 	goto discard;
218 
219 discard_put:
220 	sock_put(sk);
221 
222 discard:
223 	kfree_skb(skb);
224 	return 0;
225 }
226 
227 static int l2tp_ip6_open(struct sock *sk)
228 {
229 	/* Prevent autobind. We don't have ports. */
230 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
231 
232 	write_lock_bh(&l2tp_ip6_lock);
233 	sk_add_node(sk, &l2tp_ip6_table);
234 	write_unlock_bh(&l2tp_ip6_lock);
235 
236 	return 0;
237 }
238 
239 static void l2tp_ip6_close(struct sock *sk, long timeout)
240 {
241 	write_lock_bh(&l2tp_ip6_lock);
242 	hlist_del_init(&sk->sk_bind_node);
243 	sk_del_node_init(sk);
244 	write_unlock_bh(&l2tp_ip6_lock);
245 
246 	sk_common_release(sk);
247 }
248 
249 static void l2tp_ip6_destroy_sock(struct sock *sk)
250 {
251 	struct l2tp_tunnel *tunnel = sk->sk_user_data;
252 
253 	lock_sock(sk);
254 	ip6_flush_pending_frames(sk);
255 	release_sock(sk);
256 
257 	if (tunnel)
258 		l2tp_tunnel_delete(tunnel);
259 
260 	inet6_destroy_sock(sk);
261 }
262 
263 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
264 {
265 	struct inet_sock *inet = inet_sk(sk);
266 	struct ipv6_pinfo *np = inet6_sk(sk);
267 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
268 	struct net *net = sock_net(sk);
269 	__be32 v4addr = 0;
270 	int bound_dev_if;
271 	int addr_type;
272 	int err;
273 
274 	if (addr->l2tp_family != AF_INET6)
275 		return -EINVAL;
276 	if (addr_len < sizeof(*addr))
277 		return -EINVAL;
278 
279 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
280 
281 	/* l2tp_ip6 sockets are IPv6 only */
282 	if (addr_type == IPV6_ADDR_MAPPED)
283 		return -EADDRNOTAVAIL;
284 
285 	/* L2TP is point-point, not multicast */
286 	if (addr_type & IPV6_ADDR_MULTICAST)
287 		return -EADDRNOTAVAIL;
288 
289 	lock_sock(sk);
290 
291 	err = -EINVAL;
292 	if (!sock_flag(sk, SOCK_ZAPPED))
293 		goto out_unlock;
294 
295 	if (sk->sk_state != TCP_CLOSE)
296 		goto out_unlock;
297 
298 	bound_dev_if = sk->sk_bound_dev_if;
299 
300 	/* Check if the address belongs to the host. */
301 	rcu_read_lock();
302 	if (addr_type != IPV6_ADDR_ANY) {
303 		struct net_device *dev = NULL;
304 
305 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
306 			if (addr->l2tp_scope_id)
307 				bound_dev_if = addr->l2tp_scope_id;
308 
309 			/* Binding to link-local address requires an
310 			 * interface.
311 			 */
312 			if (!bound_dev_if)
313 				goto out_unlock_rcu;
314 
315 			err = -ENODEV;
316 			dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
317 			if (!dev)
318 				goto out_unlock_rcu;
319 		}
320 
321 		/* ipv4 addr of the socket is invalid.  Only the
322 		 * unspecified and mapped address have a v4 equivalent.
323 		 */
324 		v4addr = LOOPBACK4_IPV6;
325 		err = -EADDRNOTAVAIL;
326 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
327 			goto out_unlock_rcu;
328 	}
329 	rcu_read_unlock();
330 
331 	write_lock_bh(&l2tp_ip6_lock);
332 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
333 				   addr->l2tp_conn_id)) {
334 		write_unlock_bh(&l2tp_ip6_lock);
335 		err = -EADDRINUSE;
336 		goto out_unlock;
337 	}
338 
339 	inet->inet_saddr = v4addr;
340 	inet->inet_rcv_saddr = v4addr;
341 	sk->sk_bound_dev_if = bound_dev_if;
342 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
343 	np->saddr = addr->l2tp_addr;
344 
345 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
346 
347 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
348 	sk_del_node_init(sk);
349 	write_unlock_bh(&l2tp_ip6_lock);
350 
351 	sock_reset_flag(sk, SOCK_ZAPPED);
352 	release_sock(sk);
353 	return 0;
354 
355 out_unlock_rcu:
356 	rcu_read_unlock();
357 out_unlock:
358 	release_sock(sk);
359 
360 	return err;
361 }
362 
363 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
364 			    int addr_len)
365 {
366 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
367 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
368 	struct in6_addr	*daddr;
369 	int	addr_type;
370 	int rc;
371 
372 	if (addr_len < sizeof(*lsa))
373 		return -EINVAL;
374 
375 	if (usin->sin6_family != AF_INET6)
376 		return -EINVAL;
377 
378 	addr_type = ipv6_addr_type(&usin->sin6_addr);
379 	if (addr_type & IPV6_ADDR_MULTICAST)
380 		return -EINVAL;
381 
382 	if (addr_type & IPV6_ADDR_MAPPED) {
383 		daddr = &usin->sin6_addr;
384 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
385 			return -EINVAL;
386 	}
387 
388 	lock_sock(sk);
389 
390 	 /* Must bind first - autobinding does not work */
391 	if (sock_flag(sk, SOCK_ZAPPED)) {
392 		rc = -EINVAL;
393 		goto out_sk;
394 	}
395 
396 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
397 	if (rc < 0)
398 		goto out_sk;
399 
400 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
401 
402 	write_lock_bh(&l2tp_ip6_lock);
403 	hlist_del_init(&sk->sk_bind_node);
404 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
405 	write_unlock_bh(&l2tp_ip6_lock);
406 
407 out_sk:
408 	release_sock(sk);
409 
410 	return rc;
411 }
412 
413 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
414 {
415 	if (sock_flag(sk, SOCK_ZAPPED))
416 		return 0;
417 
418 	return __udp_disconnect(sk, flags);
419 }
420 
421 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
422 			    int peer)
423 {
424 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
425 	struct sock *sk = sock->sk;
426 	struct ipv6_pinfo *np = inet6_sk(sk);
427 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
428 
429 	lsa->l2tp_family = AF_INET6;
430 	lsa->l2tp_flowinfo = 0;
431 	lsa->l2tp_scope_id = 0;
432 	lsa->l2tp_unused = 0;
433 	if (peer) {
434 		if (!lsk->peer_conn_id)
435 			return -ENOTCONN;
436 		lsa->l2tp_conn_id = lsk->peer_conn_id;
437 		lsa->l2tp_addr = sk->sk_v6_daddr;
438 		if (np->sndflow)
439 			lsa->l2tp_flowinfo = np->flow_label;
440 	} else {
441 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
442 			lsa->l2tp_addr = np->saddr;
443 		else
444 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
445 
446 		lsa->l2tp_conn_id = lsk->conn_id;
447 	}
448 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
449 		lsa->l2tp_scope_id = sk->sk_bound_dev_if;
450 	return sizeof(*lsa);
451 }
452 
453 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
454 {
455 	int rc;
456 
457 	/* Charge it to the socket, dropping if the queue is full. */
458 	rc = sock_queue_rcv_skb(sk, skb);
459 	if (rc < 0)
460 		goto drop;
461 
462 	return 0;
463 
464 drop:
465 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
466 	kfree_skb(skb);
467 	return -1;
468 }
469 
470 static int l2tp_ip6_push_pending_frames(struct sock *sk)
471 {
472 	struct sk_buff *skb;
473 	__be32 *transhdr = NULL;
474 	int err = 0;
475 
476 	skb = skb_peek(&sk->sk_write_queue);
477 	if (skb == NULL)
478 		goto out;
479 
480 	transhdr = (__be32 *)skb_transport_header(skb);
481 	*transhdr = 0;
482 
483 	err = ip6_push_pending_frames(sk);
484 
485 out:
486 	return err;
487 }
488 
489 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
490  * control frames.
491  */
492 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
493 {
494 	struct ipv6_txoptions opt_space;
495 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
496 	struct in6_addr *daddr, *final_p, final;
497 	struct ipv6_pinfo *np = inet6_sk(sk);
498 	struct ipv6_txoptions *opt_to_free = NULL;
499 	struct ipv6_txoptions *opt = NULL;
500 	struct ip6_flowlabel *flowlabel = NULL;
501 	struct dst_entry *dst = NULL;
502 	struct flowi6 fl6;
503 	struct ipcm6_cookie ipc6;
504 	int addr_len = msg->msg_namelen;
505 	int transhdrlen = 4; /* zero session-id */
506 	int ulen = len + transhdrlen;
507 	int err;
508 
509 	/* Rough check on arithmetic overflow,
510 	   better check is made in ip6_append_data().
511 	 */
512 	if (len > INT_MAX)
513 		return -EMSGSIZE;
514 
515 	/* Mirror BSD error message compatibility */
516 	if (msg->msg_flags & MSG_OOB)
517 		return -EOPNOTSUPP;
518 
519 	/*
520 	 *	Get and verify the address.
521 	 */
522 	memset(&fl6, 0, sizeof(fl6));
523 
524 	fl6.flowi6_mark = sk->sk_mark;
525 	fl6.flowi6_uid = sk->sk_uid;
526 
527 	ipcm6_init(&ipc6);
528 
529 	if (lsa) {
530 		if (addr_len < SIN6_LEN_RFC2133)
531 			return -EINVAL;
532 
533 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
534 			return -EAFNOSUPPORT;
535 
536 		daddr = &lsa->l2tp_addr;
537 		if (np->sndflow) {
538 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
539 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
540 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
541 				if (flowlabel == NULL)
542 					return -EINVAL;
543 			}
544 		}
545 
546 		/*
547 		 * Otherwise it will be difficult to maintain
548 		 * sk->sk_dst_cache.
549 		 */
550 		if (sk->sk_state == TCP_ESTABLISHED &&
551 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
552 			daddr = &sk->sk_v6_daddr;
553 
554 		if (addr_len >= sizeof(struct sockaddr_in6) &&
555 		    lsa->l2tp_scope_id &&
556 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
557 			fl6.flowi6_oif = lsa->l2tp_scope_id;
558 	} else {
559 		if (sk->sk_state != TCP_ESTABLISHED)
560 			return -EDESTADDRREQ;
561 
562 		daddr = &sk->sk_v6_daddr;
563 		fl6.flowlabel = np->flow_label;
564 	}
565 
566 	if (fl6.flowi6_oif == 0)
567 		fl6.flowi6_oif = sk->sk_bound_dev_if;
568 
569 	if (msg->msg_controllen) {
570 		opt = &opt_space;
571 		memset(opt, 0, sizeof(struct ipv6_txoptions));
572 		opt->tot_len = sizeof(struct ipv6_txoptions);
573 		ipc6.opt = opt;
574 
575 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
576 		if (err < 0) {
577 			fl6_sock_release(flowlabel);
578 			return err;
579 		}
580 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
581 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
582 			if (flowlabel == NULL)
583 				return -EINVAL;
584 		}
585 		if (!(opt->opt_nflen|opt->opt_flen))
586 			opt = NULL;
587 	}
588 
589 	if (!opt) {
590 		opt = txopt_get(np);
591 		opt_to_free = opt;
592 	}
593 	if (flowlabel)
594 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
595 	opt = ipv6_fixup_options(&opt_space, opt);
596 	ipc6.opt = opt;
597 
598 	fl6.flowi6_proto = sk->sk_protocol;
599 	if (!ipv6_addr_any(daddr))
600 		fl6.daddr = *daddr;
601 	else
602 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
603 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
604 		fl6.saddr = np->saddr;
605 
606 	final_p = fl6_update_dst(&fl6, opt, &final);
607 
608 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
609 		fl6.flowi6_oif = np->mcast_oif;
610 	else if (!fl6.flowi6_oif)
611 		fl6.flowi6_oif = np->ucast_oif;
612 
613 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
614 
615 	if (ipc6.tclass < 0)
616 		ipc6.tclass = np->tclass;
617 
618 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
619 
620 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
621 	if (IS_ERR(dst)) {
622 		err = PTR_ERR(dst);
623 		goto out;
624 	}
625 
626 	if (ipc6.hlimit < 0)
627 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
628 
629 	if (ipc6.dontfrag < 0)
630 		ipc6.dontfrag = np->dontfrag;
631 
632 	if (msg->msg_flags & MSG_CONFIRM)
633 		goto do_confirm;
634 
635 back_from_confirm:
636 	lock_sock(sk);
637 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
638 			      ulen, transhdrlen, &ipc6,
639 			      &fl6, (struct rt6_info *)dst,
640 			      msg->msg_flags);
641 	if (err)
642 		ip6_flush_pending_frames(sk);
643 	else if (!(msg->msg_flags & MSG_MORE))
644 		err = l2tp_ip6_push_pending_frames(sk);
645 	release_sock(sk);
646 done:
647 	dst_release(dst);
648 out:
649 	fl6_sock_release(flowlabel);
650 	txopt_put(opt_to_free);
651 
652 	return err < 0 ? err : len;
653 
654 do_confirm:
655 	if (msg->msg_flags & MSG_PROBE)
656 		dst_confirm_neigh(dst, &fl6.daddr);
657 	if (!(msg->msg_flags & MSG_PROBE) || len)
658 		goto back_from_confirm;
659 	err = 0;
660 	goto done;
661 }
662 
663 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
664 			    int noblock, int flags, int *addr_len)
665 {
666 	struct ipv6_pinfo *np = inet6_sk(sk);
667 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
668 	size_t copied = 0;
669 	int err = -EOPNOTSUPP;
670 	struct sk_buff *skb;
671 
672 	if (flags & MSG_OOB)
673 		goto out;
674 
675 	if (addr_len)
676 		*addr_len = sizeof(*lsa);
677 
678 	if (flags & MSG_ERRQUEUE)
679 		return ipv6_recv_error(sk, msg, len, addr_len);
680 
681 	skb = skb_recv_datagram(sk, flags, noblock, &err);
682 	if (!skb)
683 		goto out;
684 
685 	copied = skb->len;
686 	if (len < copied) {
687 		msg->msg_flags |= MSG_TRUNC;
688 		copied = len;
689 	}
690 
691 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
692 	if (err)
693 		goto done;
694 
695 	sock_recv_timestamp(msg, sk, skb);
696 
697 	/* Copy the address. */
698 	if (lsa) {
699 		lsa->l2tp_family = AF_INET6;
700 		lsa->l2tp_unused = 0;
701 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
702 		lsa->l2tp_flowinfo = 0;
703 		lsa->l2tp_scope_id = 0;
704 		lsa->l2tp_conn_id = 0;
705 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
706 			lsa->l2tp_scope_id = inet6_iif(skb);
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 	.listen		   = sock_no_listen,
756 	.shutdown	   = inet_shutdown,
757 	.setsockopt	   = sock_common_setsockopt,
758 	.getsockopt	   = sock_common_getsockopt,
759 	.sendmsg	   = inet_sendmsg,
760 	.recvmsg	   = sock_common_recvmsg,
761 	.mmap		   = sock_no_mmap,
762 	.sendpage	   = sock_no_sendpage,
763 #ifdef CONFIG_COMPAT
764 	.compat_setsockopt = compat_sock_common_setsockopt,
765 	.compat_getsockopt = compat_sock_common_getsockopt,
766 #endif
767 };
768 
769 static struct inet_protosw l2tp_ip6_protosw = {
770 	.type		= SOCK_DGRAM,
771 	.protocol	= IPPROTO_L2TP,
772 	.prot		= &l2tp_ip6_prot,
773 	.ops		= &l2tp_ip6_ops,
774 };
775 
776 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
777 	.handler	= l2tp_ip6_recv,
778 };
779 
780 static int __init l2tp_ip6_init(void)
781 {
782 	int err;
783 
784 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
785 
786 	err = proto_register(&l2tp_ip6_prot, 1);
787 	if (err != 0)
788 		goto out;
789 
790 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
791 	if (err)
792 		goto out1;
793 
794 	inet6_register_protosw(&l2tp_ip6_protosw);
795 	return 0;
796 
797 out1:
798 	proto_unregister(&l2tp_ip6_prot);
799 out:
800 	return err;
801 }
802 
803 static void __exit l2tp_ip6_exit(void)
804 {
805 	inet6_unregister_protosw(&l2tp_ip6_protosw);
806 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
807 	proto_unregister(&l2tp_ip6_prot);
808 }
809 
810 module_init(l2tp_ip6_init);
811 module_exit(l2tp_ip6_exit);
812 
813 MODULE_LICENSE("GPL");
814 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
815 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
816 MODULE_VERSION("1.0");
817 
818 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
819  * enums
820  */
821 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
822 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);
823