xref: /linux/net/l2tp/l2tp_ip6.c (revision 84b9b44b)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* L2TPv3 IP encapsulation support for IPv6
3  *
4  * Copyright (c) 2012 Katalix Systems Ltd
5  */
6 
7 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
8 
9 #include <linux/icmp.h>
10 #include <linux/module.h>
11 #include <linux/skbuff.h>
12 #include <linux/random.h>
13 #include <linux/socket.h>
14 #include <linux/l2tp.h>
15 #include <linux/in.h>
16 #include <linux/in6.h>
17 #include <net/sock.h>
18 #include <net/ip.h>
19 #include <net/icmp.h>
20 #include <net/udp.h>
21 #include <net/inet_common.h>
22 #include <net/tcp_states.h>
23 #include <net/protocol.h>
24 #include <net/xfrm.h>
25 
26 #include <net/transp_v6.h>
27 #include <net/addrconf.h>
28 #include <net/ip6_route.h>
29 
30 #include "l2tp_core.h"
31 
32 struct l2tp_ip6_sock {
33 	/* inet_sock has to be the first member of l2tp_ip6_sock */
34 	struct inet_sock	inet;
35 
36 	u32			conn_id;
37 	u32			peer_conn_id;
38 
39 	/* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
40 	 * inet6_sk_generic
41 	 */
42 	struct ipv6_pinfo	inet6;
43 };
44 
45 static DEFINE_RWLOCK(l2tp_ip6_lock);
46 static struct hlist_head l2tp_ip6_table;
47 static struct hlist_head l2tp_ip6_bind_table;
48 
49 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
50 {
51 	return (struct l2tp_ip6_sock *)sk;
52 }
53 
54 static struct sock *__l2tp_ip6_bind_lookup(const struct net *net,
55 					   const struct in6_addr *laddr,
56 					   const struct in6_addr *raddr,
57 					   int dif, u32 tunnel_id)
58 {
59 	struct sock *sk;
60 
61 	sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
62 		const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
63 		const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
64 		const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
65 		int bound_dev_if;
66 
67 		if (!net_eq(sock_net(sk), net))
68 			continue;
69 
70 		bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
71 		if (bound_dev_if && dif && bound_dev_if != dif)
72 			continue;
73 
74 		if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
75 		    !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
76 			continue;
77 
78 		if (!ipv6_addr_any(sk_raddr) && raddr &&
79 		    !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
80 			continue;
81 
82 		if (l2tp->conn_id != tunnel_id)
83 			continue;
84 
85 		goto found;
86 	}
87 
88 	sk = NULL;
89 found:
90 	return sk;
91 }
92 
93 /* When processing receive frames, there are two cases to
94  * consider. Data frames consist of a non-zero session-id and an
95  * optional cookie. Control frames consist of a regular L2TP header
96  * preceded by 32-bits of zeros.
97  *
98  * L2TPv3 Session Header Over IP
99  *
100  *  0                   1                   2                   3
101  *  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
102  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103  * |                           Session ID                          |
104  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105  * |               Cookie (optional, maximum 64 bits)...
106  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107  *                                                                 |
108  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109  *
110  * L2TPv3 Control Message Header Over IP
111  *
112  *  0                   1                   2                   3
113  *  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
114  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115  * |                      (32 bits of zeros)                       |
116  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117  * |T|L|x|x|S|x|x|x|x|x|x|x|  Ver  |             Length            |
118  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119  * |                     Control Connection ID                     |
120  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121  * |               Ns              |               Nr              |
122  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
123  *
124  * All control frames are passed to userspace.
125  */
126 static int l2tp_ip6_recv(struct sk_buff *skb)
127 {
128 	struct net *net = dev_net(skb->dev);
129 	struct sock *sk;
130 	u32 session_id;
131 	u32 tunnel_id;
132 	unsigned char *ptr, *optr;
133 	struct l2tp_session *session;
134 	struct l2tp_tunnel *tunnel = NULL;
135 	struct ipv6hdr *iph;
136 
137 	if (!pskb_may_pull(skb, 4))
138 		goto discard;
139 
140 	/* Point to L2TP header */
141 	optr = skb->data;
142 	ptr = skb->data;
143 	session_id = ntohl(*((__be32 *)ptr));
144 	ptr += 4;
145 
146 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
147 	 * the session_id. If it is 0, the packet is a L2TP control
148 	 * frame and the session_id value can be discarded.
149 	 */
150 	if (session_id == 0) {
151 		__skb_pull(skb, 4);
152 		goto pass_up;
153 	}
154 
155 	/* Ok, this is a data packet. Lookup the session. */
156 	session = l2tp_session_get(net, session_id);
157 	if (!session)
158 		goto discard;
159 
160 	tunnel = session->tunnel;
161 	if (!tunnel)
162 		goto discard_sess;
163 
164 	if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
165 		goto discard_sess;
166 
167 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
168 	l2tp_session_dec_refcount(session);
169 
170 	return 0;
171 
172 pass_up:
173 	/* Get the tunnel_id from the L2TP header */
174 	if (!pskb_may_pull(skb, 12))
175 		goto discard;
176 
177 	if ((skb->data[0] & 0xc0) != 0xc0)
178 		goto discard;
179 
180 	tunnel_id = ntohl(*(__be32 *)&skb->data[4]);
181 	iph = ipv6_hdr(skb);
182 
183 	read_lock_bh(&l2tp_ip6_lock);
184 	sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
185 				    inet6_iif(skb), tunnel_id);
186 	if (!sk) {
187 		read_unlock_bh(&l2tp_ip6_lock);
188 		goto discard;
189 	}
190 	sock_hold(sk);
191 	read_unlock_bh(&l2tp_ip6_lock);
192 
193 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
194 		goto discard_put;
195 
196 	nf_reset_ct(skb);
197 
198 	return sk_receive_skb(sk, skb, 1);
199 
200 discard_sess:
201 	l2tp_session_dec_refcount(session);
202 	goto discard;
203 
204 discard_put:
205 	sock_put(sk);
206 
207 discard:
208 	kfree_skb(skb);
209 	return 0;
210 }
211 
212 static int l2tp_ip6_hash(struct sock *sk)
213 {
214 	if (sk_unhashed(sk)) {
215 		write_lock_bh(&l2tp_ip6_lock);
216 		sk_add_node(sk, &l2tp_ip6_table);
217 		write_unlock_bh(&l2tp_ip6_lock);
218 	}
219 	return 0;
220 }
221 
222 static void l2tp_ip6_unhash(struct sock *sk)
223 {
224 	if (sk_unhashed(sk))
225 		return;
226 	write_lock_bh(&l2tp_ip6_lock);
227 	sk_del_node_init(sk);
228 	write_unlock_bh(&l2tp_ip6_lock);
229 }
230 
231 static int l2tp_ip6_open(struct sock *sk)
232 {
233 	/* Prevent autobind. We don't have ports. */
234 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
235 
236 	l2tp_ip6_hash(sk);
237 	return 0;
238 }
239 
240 static void l2tp_ip6_close(struct sock *sk, long timeout)
241 {
242 	write_lock_bh(&l2tp_ip6_lock);
243 	hlist_del_init(&sk->sk_bind_node);
244 	sk_del_node_init(sk);
245 	write_unlock_bh(&l2tp_ip6_lock);
246 
247 	sk_common_release(sk);
248 }
249 
250 static void l2tp_ip6_destroy_sock(struct sock *sk)
251 {
252 	struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk);
253 
254 	lock_sock(sk);
255 	ip6_flush_pending_frames(sk);
256 	release_sock(sk);
257 
258 	if (tunnel)
259 		l2tp_tunnel_delete(tunnel);
260 }
261 
262 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
263 {
264 	struct inet_sock *inet = inet_sk(sk);
265 	struct ipv6_pinfo *np = inet6_sk(sk);
266 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *)uaddr;
267 	struct net *net = sock_net(sk);
268 	__be32 v4addr = 0;
269 	int bound_dev_if;
270 	int addr_type;
271 	int err;
272 
273 	if (addr->l2tp_family != AF_INET6)
274 		return -EINVAL;
275 	if (addr_len < sizeof(*addr))
276 		return -EINVAL;
277 
278 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
279 
280 	/* l2tp_ip6 sockets are IPv6 only */
281 	if (addr_type == IPV6_ADDR_MAPPED)
282 		return -EADDRNOTAVAIL;
283 
284 	/* L2TP is point-point, not multicast */
285 	if (addr_type & IPV6_ADDR_MULTICAST)
286 		return -EADDRNOTAVAIL;
287 
288 	lock_sock(sk);
289 
290 	err = -EINVAL;
291 	if (!sock_flag(sk, SOCK_ZAPPED))
292 		goto out_unlock;
293 
294 	if (sk->sk_state != TCP_CLOSE)
295 		goto out_unlock;
296 
297 	bound_dev_if = sk->sk_bound_dev_if;
298 
299 	/* Check if the address belongs to the host. */
300 	rcu_read_lock();
301 	if (addr_type != IPV6_ADDR_ANY) {
302 		struct net_device *dev = NULL;
303 
304 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
305 			if (addr->l2tp_scope_id)
306 				bound_dev_if = addr->l2tp_scope_id;
307 
308 			/* Binding to link-local address requires an
309 			 * interface.
310 			 */
311 			if (!bound_dev_if)
312 				goto out_unlock_rcu;
313 
314 			err = -ENODEV;
315 			dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
316 			if (!dev)
317 				goto out_unlock_rcu;
318 		}
319 
320 		/* ipv4 addr of the socket is invalid.  Only the
321 		 * unspecified and mapped address have a v4 equivalent.
322 		 */
323 		v4addr = LOOPBACK4_IPV6;
324 		err = -EADDRNOTAVAIL;
325 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
326 			goto out_unlock_rcu;
327 	}
328 	rcu_read_unlock();
329 
330 	write_lock_bh(&l2tp_ip6_lock);
331 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
332 				   addr->l2tp_conn_id)) {
333 		write_unlock_bh(&l2tp_ip6_lock);
334 		err = -EADDRINUSE;
335 		goto out_unlock;
336 	}
337 
338 	inet->inet_saddr = v4addr;
339 	inet->inet_rcv_saddr = v4addr;
340 	sk->sk_bound_dev_if = bound_dev_if;
341 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
342 	np->saddr = addr->l2tp_addr;
343 
344 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
345 
346 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
347 	sk_del_node_init(sk);
348 	write_unlock_bh(&l2tp_ip6_lock);
349 
350 	sock_reset_flag(sk, SOCK_ZAPPED);
351 	release_sock(sk);
352 	return 0;
353 
354 out_unlock_rcu:
355 	rcu_read_unlock();
356 out_unlock:
357 	release_sock(sk);
358 
359 	return err;
360 }
361 
362 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
363 			    int addr_len)
364 {
365 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
366 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *)uaddr;
367 	struct in6_addr	*daddr;
368 	int	addr_type;
369 	int rc;
370 
371 	if (addr_len < sizeof(*lsa))
372 		return -EINVAL;
373 
374 	if (usin->sin6_family != AF_INET6)
375 		return -EINVAL;
376 
377 	addr_type = ipv6_addr_type(&usin->sin6_addr);
378 	if (addr_type & IPV6_ADDR_MULTICAST)
379 		return -EINVAL;
380 
381 	if (addr_type & IPV6_ADDR_MAPPED) {
382 		daddr = &usin->sin6_addr;
383 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
384 			return -EINVAL;
385 	}
386 
387 	lock_sock(sk);
388 
389 	 /* Must bind first - autobinding does not work */
390 	if (sock_flag(sk, SOCK_ZAPPED)) {
391 		rc = -EINVAL;
392 		goto out_sk;
393 	}
394 
395 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
396 	if (rc < 0)
397 		goto out_sk;
398 
399 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
400 
401 	write_lock_bh(&l2tp_ip6_lock);
402 	hlist_del_init(&sk->sk_bind_node);
403 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
404 	write_unlock_bh(&l2tp_ip6_lock);
405 
406 out_sk:
407 	release_sock(sk);
408 
409 	return rc;
410 }
411 
412 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
413 {
414 	if (sock_flag(sk, SOCK_ZAPPED))
415 		return 0;
416 
417 	return __udp_disconnect(sk, flags);
418 }
419 
420 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
421 			    int peer)
422 {
423 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
424 	struct sock *sk = sock->sk;
425 	struct ipv6_pinfo *np = inet6_sk(sk);
426 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
427 
428 	lsa->l2tp_family = AF_INET6;
429 	lsa->l2tp_flowinfo = 0;
430 	lsa->l2tp_scope_id = 0;
431 	lsa->l2tp_unused = 0;
432 	if (peer) {
433 		if (!lsk->peer_conn_id)
434 			return -ENOTCONN;
435 		lsa->l2tp_conn_id = lsk->peer_conn_id;
436 		lsa->l2tp_addr = sk->sk_v6_daddr;
437 		if (np->sndflow)
438 			lsa->l2tp_flowinfo = np->flow_label;
439 	} else {
440 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
441 			lsa->l2tp_addr = np->saddr;
442 		else
443 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
444 
445 		lsa->l2tp_conn_id = lsk->conn_id;
446 	}
447 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
448 		lsa->l2tp_scope_id = READ_ONCE(sk->sk_bound_dev_if);
449 	return sizeof(*lsa);
450 }
451 
452 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
453 {
454 	int rc;
455 
456 	/* Charge it to the socket, dropping if the queue is full. */
457 	rc = sock_queue_rcv_skb(sk, skb);
458 	if (rc < 0)
459 		goto drop;
460 
461 	return 0;
462 
463 drop:
464 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
465 	kfree_skb(skb);
466 	return -1;
467 }
468 
469 static int l2tp_ip6_push_pending_frames(struct sock *sk)
470 {
471 	struct sk_buff *skb;
472 	__be32 *transhdr = NULL;
473 	int err = 0;
474 
475 	skb = skb_peek(&sk->sk_write_queue);
476 	if (!skb)
477 		goto out;
478 
479 	transhdr = (__be32 *)skb_transport_header(skb);
480 	*transhdr = 0;
481 
482 	err = ip6_push_pending_frames(sk);
483 
484 out:
485 	return err;
486 }
487 
488 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
489  * control frames.
490  */
491 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
492 {
493 	struct ipv6_txoptions opt_space;
494 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
495 	struct in6_addr *daddr, *final_p, final;
496 	struct ipv6_pinfo *np = inet6_sk(sk);
497 	struct ipv6_txoptions *opt_to_free = NULL;
498 	struct ipv6_txoptions *opt = NULL;
499 	struct ip6_flowlabel *flowlabel = NULL;
500 	struct dst_entry *dst = NULL;
501 	struct flowi6 fl6;
502 	struct ipcm6_cookie ipc6;
503 	int addr_len = msg->msg_namelen;
504 	int transhdrlen = 4; /* zero session-id */
505 	int ulen;
506 	int err;
507 
508 	/* Rough check on arithmetic overflow,
509 	 * better check is made in ip6_append_data().
510 	 */
511 	if (len > INT_MAX - transhdrlen)
512 		return -EMSGSIZE;
513 	ulen = len + transhdrlen;
514 
515 	/* Mirror BSD error message compatibility */
516 	if (msg->msg_flags & MSG_OOB)
517 		return -EOPNOTSUPP;
518 
519 	/* Get and verify the address */
520 	memset(&fl6, 0, sizeof(fl6));
521 
522 	fl6.flowi6_mark = sk->sk_mark;
523 	fl6.flowi6_uid = sk->sk_uid;
524 
525 	ipcm6_init(&ipc6);
526 
527 	if (lsa) {
528 		if (addr_len < SIN6_LEN_RFC2133)
529 			return -EINVAL;
530 
531 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
532 			return -EAFNOSUPPORT;
533 
534 		daddr = &lsa->l2tp_addr;
535 		if (np->sndflow) {
536 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
537 			if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
538 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
539 				if (IS_ERR(flowlabel))
540 					return -EINVAL;
541 			}
542 		}
543 
544 		/* Otherwise it will be difficult to maintain
545 		 * sk->sk_dst_cache.
546 		 */
547 		if (sk->sk_state == TCP_ESTABLISHED &&
548 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
549 			daddr = &sk->sk_v6_daddr;
550 
551 		if (addr_len >= sizeof(struct sockaddr_in6) &&
552 		    lsa->l2tp_scope_id &&
553 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
554 			fl6.flowi6_oif = lsa->l2tp_scope_id;
555 	} else {
556 		if (sk->sk_state != TCP_ESTABLISHED)
557 			return -EDESTADDRREQ;
558 
559 		daddr = &sk->sk_v6_daddr;
560 		fl6.flowlabel = np->flow_label;
561 	}
562 
563 	if (fl6.flowi6_oif == 0)
564 		fl6.flowi6_oif = READ_ONCE(sk->sk_bound_dev_if);
565 
566 	if (msg->msg_controllen) {
567 		opt = &opt_space;
568 		memset(opt, 0, sizeof(struct ipv6_txoptions));
569 		opt->tot_len = sizeof(struct ipv6_txoptions);
570 		ipc6.opt = opt;
571 
572 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
573 		if (err < 0) {
574 			fl6_sock_release(flowlabel);
575 			return err;
576 		}
577 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
578 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
579 			if (IS_ERR(flowlabel))
580 				return -EINVAL;
581 		}
582 		if (!(opt->opt_nflen | opt->opt_flen))
583 			opt = NULL;
584 	}
585 
586 	if (!opt) {
587 		opt = txopt_get(np);
588 		opt_to_free = opt;
589 	}
590 	if (flowlabel)
591 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
592 	opt = ipv6_fixup_options(&opt_space, opt);
593 	ipc6.opt = opt;
594 
595 	fl6.flowi6_proto = sk->sk_protocol;
596 	if (!ipv6_addr_any(daddr))
597 		fl6.daddr = *daddr;
598 	else
599 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
600 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
601 		fl6.saddr = np->saddr;
602 
603 	final_p = fl6_update_dst(&fl6, opt, &final);
604 
605 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
606 		fl6.flowi6_oif = np->mcast_oif;
607 	else if (!fl6.flowi6_oif)
608 		fl6.flowi6_oif = np->ucast_oif;
609 
610 	security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
611 
612 	if (ipc6.tclass < 0)
613 		ipc6.tclass = np->tclass;
614 
615 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
616 
617 	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
618 	if (IS_ERR(dst)) {
619 		err = PTR_ERR(dst);
620 		goto out;
621 	}
622 
623 	if (ipc6.hlimit < 0)
624 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
625 
626 	if (ipc6.dontfrag < 0)
627 		ipc6.dontfrag = np->dontfrag;
628 
629 	if (msg->msg_flags & MSG_CONFIRM)
630 		goto do_confirm;
631 
632 back_from_confirm:
633 	lock_sock(sk);
634 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
635 			      ulen, transhdrlen, &ipc6,
636 			      &fl6, (struct rt6_info *)dst,
637 			      msg->msg_flags);
638 	if (err)
639 		ip6_flush_pending_frames(sk);
640 	else if (!(msg->msg_flags & MSG_MORE))
641 		err = l2tp_ip6_push_pending_frames(sk);
642 	release_sock(sk);
643 done:
644 	dst_release(dst);
645 out:
646 	fl6_sock_release(flowlabel);
647 	txopt_put(opt_to_free);
648 
649 	return err < 0 ? err : len;
650 
651 do_confirm:
652 	if (msg->msg_flags & MSG_PROBE)
653 		dst_confirm_neigh(dst, &fl6.daddr);
654 	if (!(msg->msg_flags & MSG_PROBE) || len)
655 		goto back_from_confirm;
656 	err = 0;
657 	goto done;
658 }
659 
660 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
661 			    int flags, int *addr_len)
662 {
663 	struct ipv6_pinfo *np = inet6_sk(sk);
664 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
665 	size_t copied = 0;
666 	int err = -EOPNOTSUPP;
667 	struct sk_buff *skb;
668 
669 	if (flags & MSG_OOB)
670 		goto out;
671 
672 	if (flags & MSG_ERRQUEUE)
673 		return ipv6_recv_error(sk, msg, len, addr_len);
674 
675 	skb = skb_recv_datagram(sk, flags, &err);
676 	if (!skb)
677 		goto out;
678 
679 	copied = skb->len;
680 	if (len < copied) {
681 		msg->msg_flags |= MSG_TRUNC;
682 		copied = len;
683 	}
684 
685 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
686 	if (err)
687 		goto done;
688 
689 	sock_recv_timestamp(msg, sk, skb);
690 
691 	/* Copy the address. */
692 	if (lsa) {
693 		lsa->l2tp_family = AF_INET6;
694 		lsa->l2tp_unused = 0;
695 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
696 		lsa->l2tp_flowinfo = 0;
697 		lsa->l2tp_scope_id = 0;
698 		lsa->l2tp_conn_id = 0;
699 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
700 			lsa->l2tp_scope_id = inet6_iif(skb);
701 		*addr_len = sizeof(*lsa);
702 	}
703 
704 	if (np->rxopt.all)
705 		ip6_datagram_recv_ctl(sk, msg, skb);
706 
707 	if (flags & MSG_TRUNC)
708 		copied = skb->len;
709 done:
710 	skb_free_datagram(sk, skb);
711 out:
712 	return err ? err : copied;
713 }
714 
715 static struct proto l2tp_ip6_prot = {
716 	.name		   = "L2TP/IPv6",
717 	.owner		   = THIS_MODULE,
718 	.init		   = l2tp_ip6_open,
719 	.close		   = l2tp_ip6_close,
720 	.bind		   = l2tp_ip6_bind,
721 	.connect	   = l2tp_ip6_connect,
722 	.disconnect	   = l2tp_ip6_disconnect,
723 	.ioctl		   = l2tp_ioctl,
724 	.destroy	   = l2tp_ip6_destroy_sock,
725 	.setsockopt	   = ipv6_setsockopt,
726 	.getsockopt	   = ipv6_getsockopt,
727 	.sendmsg	   = l2tp_ip6_sendmsg,
728 	.recvmsg	   = l2tp_ip6_recvmsg,
729 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
730 	.hash		   = l2tp_ip6_hash,
731 	.unhash		   = l2tp_ip6_unhash,
732 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
733 };
734 
735 static const struct proto_ops l2tp_ip6_ops = {
736 	.family		   = PF_INET6,
737 	.owner		   = THIS_MODULE,
738 	.release	   = inet6_release,
739 	.bind		   = inet6_bind,
740 	.connect	   = inet_dgram_connect,
741 	.socketpair	   = sock_no_socketpair,
742 	.accept		   = sock_no_accept,
743 	.getname	   = l2tp_ip6_getname,
744 	.poll		   = datagram_poll,
745 	.ioctl		   = inet6_ioctl,
746 	.gettstamp	   = sock_gettstamp,
747 	.listen		   = sock_no_listen,
748 	.shutdown	   = inet_shutdown,
749 	.setsockopt	   = sock_common_setsockopt,
750 	.getsockopt	   = sock_common_getsockopt,
751 	.sendmsg	   = inet_sendmsg,
752 	.recvmsg	   = sock_common_recvmsg,
753 	.mmap		   = sock_no_mmap,
754 	.sendpage	   = sock_no_sendpage,
755 #ifdef CONFIG_COMPAT
756 	.compat_ioctl	   = inet6_compat_ioctl,
757 #endif
758 };
759 
760 static struct inet_protosw l2tp_ip6_protosw = {
761 	.type		= SOCK_DGRAM,
762 	.protocol	= IPPROTO_L2TP,
763 	.prot		= &l2tp_ip6_prot,
764 	.ops		= &l2tp_ip6_ops,
765 };
766 
767 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
768 	.handler	= l2tp_ip6_recv,
769 };
770 
771 static int __init l2tp_ip6_init(void)
772 {
773 	int err;
774 
775 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
776 
777 	err = proto_register(&l2tp_ip6_prot, 1);
778 	if (err != 0)
779 		goto out;
780 
781 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
782 	if (err)
783 		goto out1;
784 
785 	inet6_register_protosw(&l2tp_ip6_protosw);
786 	return 0;
787 
788 out1:
789 	proto_unregister(&l2tp_ip6_prot);
790 out:
791 	return err;
792 }
793 
794 static void __exit l2tp_ip6_exit(void)
795 {
796 	inet6_unregister_protosw(&l2tp_ip6_protosw);
797 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
798 	proto_unregister(&l2tp_ip6_prot);
799 }
800 
801 module_init(l2tp_ip6_init);
802 module_exit(l2tp_ip6_exit);
803 
804 MODULE_LICENSE("GPL");
805 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
806 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
807 MODULE_VERSION("1.0");
808 
809 /* Use the values of SOCK_DGRAM (2) as type and IPPROTO_L2TP (115) as protocol,
810  * because __stringify doesn't like enums
811  */
812 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 115, 2);
813 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, 115);
814