xref: /linux/net/l2tp/l2tp_ip6.c (revision f5c31bcf604db54470868f3118a60dc4a9ba8813)
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 	struct ipv6_pinfo	inet6;
40 };
41 
42 static DEFINE_RWLOCK(l2tp_ip6_lock);
43 static struct hlist_head l2tp_ip6_table;
44 static struct hlist_head l2tp_ip6_bind_table;
45 
46 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
47 {
48 	return (struct l2tp_ip6_sock *)sk;
49 }
50 
51 static struct sock *__l2tp_ip6_bind_lookup(const struct net *net,
52 					   const struct in6_addr *laddr,
53 					   const struct in6_addr *raddr,
54 					   int dif, u32 tunnel_id)
55 {
56 	struct sock *sk;
57 
58 	sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
59 		const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
60 		const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
61 		const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
62 		int bound_dev_if;
63 
64 		if (!net_eq(sock_net(sk), net))
65 			continue;
66 
67 		bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
68 		if (bound_dev_if && dif && bound_dev_if != dif)
69 			continue;
70 
71 		if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
72 		    !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
73 			continue;
74 
75 		if (!ipv6_addr_any(sk_raddr) && raddr &&
76 		    !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
77 			continue;
78 
79 		if (l2tp->conn_id != tunnel_id)
80 			continue;
81 
82 		goto found;
83 	}
84 
85 	sk = NULL;
86 found:
87 	return sk;
88 }
89 
90 /* When processing receive frames, there are two cases to
91  * consider. Data frames consist of a non-zero session-id and an
92  * optional cookie. Control frames consist of a regular L2TP header
93  * preceded by 32-bits of zeros.
94  *
95  * L2TPv3 Session Header Over IP
96  *
97  *  0                   1                   2                   3
98  *  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
99  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100  * |                           Session ID                          |
101  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
102  * |               Cookie (optional, maximum 64 bits)...
103  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
104  *                                                                 |
105  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106  *
107  * L2TPv3 Control Message Header Over IP
108  *
109  *  0                   1                   2                   3
110  *  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
111  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112  * |                      (32 bits of zeros)                       |
113  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114  * |T|L|x|x|S|x|x|x|x|x|x|x|  Ver  |             Length            |
115  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
116  * |                     Control Connection ID                     |
117  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
118  * |               Ns              |               Nr              |
119  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
120  *
121  * All control frames are passed to userspace.
122  */
123 static int l2tp_ip6_recv(struct sk_buff *skb)
124 {
125 	struct net *net = dev_net(skb->dev);
126 	struct sock *sk;
127 	u32 session_id;
128 	u32 tunnel_id;
129 	unsigned char *ptr, *optr;
130 	struct l2tp_session *session;
131 	struct l2tp_tunnel *tunnel = NULL;
132 	struct ipv6hdr *iph;
133 
134 	if (!pskb_may_pull(skb, 4))
135 		goto discard;
136 
137 	/* Point to L2TP header */
138 	optr = skb->data;
139 	ptr = skb->data;
140 	session_id = ntohl(*((__be32 *)ptr));
141 	ptr += 4;
142 
143 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
144 	 * the session_id. If it is 0, the packet is a L2TP control
145 	 * frame and the session_id value can be discarded.
146 	 */
147 	if (session_id == 0) {
148 		__skb_pull(skb, 4);
149 		goto pass_up;
150 	}
151 
152 	/* Ok, this is a data packet. Lookup the session. */
153 	session = l2tp_session_get(net, session_id);
154 	if (!session)
155 		goto discard;
156 
157 	tunnel = session->tunnel;
158 	if (!tunnel)
159 		goto discard_sess;
160 
161 	if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
162 		goto discard_sess;
163 
164 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
165 	l2tp_session_dec_refcount(session);
166 
167 	return 0;
168 
169 pass_up:
170 	/* Get the tunnel_id from the L2TP header */
171 	if (!pskb_may_pull(skb, 12))
172 		goto discard;
173 
174 	if ((skb->data[0] & 0xc0) != 0xc0)
175 		goto discard;
176 
177 	tunnel_id = ntohl(*(__be32 *)&skb->data[4]);
178 	iph = ipv6_hdr(skb);
179 
180 	read_lock_bh(&l2tp_ip6_lock);
181 	sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
182 				    inet6_iif(skb), tunnel_id);
183 	if (!sk) {
184 		read_unlock_bh(&l2tp_ip6_lock);
185 		goto discard;
186 	}
187 	sock_hold(sk);
188 	read_unlock_bh(&l2tp_ip6_lock);
189 
190 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
191 		goto discard_put;
192 
193 	nf_reset_ct(skb);
194 
195 	return sk_receive_skb(sk, skb, 1);
196 
197 discard_sess:
198 	l2tp_session_dec_refcount(session);
199 	goto discard;
200 
201 discard_put:
202 	sock_put(sk);
203 
204 discard:
205 	kfree_skb(skb);
206 	return 0;
207 }
208 
209 static int l2tp_ip6_hash(struct sock *sk)
210 {
211 	if (sk_unhashed(sk)) {
212 		write_lock_bh(&l2tp_ip6_lock);
213 		sk_add_node(sk, &l2tp_ip6_table);
214 		write_unlock_bh(&l2tp_ip6_lock);
215 	}
216 	return 0;
217 }
218 
219 static void l2tp_ip6_unhash(struct sock *sk)
220 {
221 	if (sk_unhashed(sk))
222 		return;
223 	write_lock_bh(&l2tp_ip6_lock);
224 	sk_del_node_init(sk);
225 	write_unlock_bh(&l2tp_ip6_lock);
226 }
227 
228 static int l2tp_ip6_open(struct sock *sk)
229 {
230 	/* Prevent autobind. We don't have ports. */
231 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
232 
233 	l2tp_ip6_hash(sk);
234 	return 0;
235 }
236 
237 static void l2tp_ip6_close(struct sock *sk, long timeout)
238 {
239 	write_lock_bh(&l2tp_ip6_lock);
240 	hlist_del_init(&sk->sk_bind_node);
241 	sk_del_node_init(sk);
242 	write_unlock_bh(&l2tp_ip6_lock);
243 
244 	sk_common_release(sk);
245 }
246 
247 static void l2tp_ip6_destroy_sock(struct sock *sk)
248 {
249 	struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk);
250 
251 	lock_sock(sk);
252 	ip6_flush_pending_frames(sk);
253 	release_sock(sk);
254 
255 	if (tunnel)
256 		l2tp_tunnel_delete(tunnel);
257 }
258 
259 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
260 {
261 	struct inet_sock *inet = inet_sk(sk);
262 	struct ipv6_pinfo *np = inet6_sk(sk);
263 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *)uaddr;
264 	struct net *net = sock_net(sk);
265 	__be32 v4addr = 0;
266 	int bound_dev_if;
267 	int addr_type;
268 	int err;
269 
270 	if (addr->l2tp_family != AF_INET6)
271 		return -EINVAL;
272 	if (addr_len < sizeof(*addr))
273 		return -EINVAL;
274 
275 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
276 
277 	/* l2tp_ip6 sockets are IPv6 only */
278 	if (addr_type == IPV6_ADDR_MAPPED)
279 		return -EADDRNOTAVAIL;
280 
281 	/* L2TP is point-point, not multicast */
282 	if (addr_type & IPV6_ADDR_MULTICAST)
283 		return -EADDRNOTAVAIL;
284 
285 	lock_sock(sk);
286 
287 	err = -EINVAL;
288 	if (!sock_flag(sk, SOCK_ZAPPED))
289 		goto out_unlock;
290 
291 	if (sk->sk_state != TCP_CLOSE)
292 		goto out_unlock;
293 
294 	bound_dev_if = sk->sk_bound_dev_if;
295 
296 	/* Check if the address belongs to the host. */
297 	rcu_read_lock();
298 	if (addr_type != IPV6_ADDR_ANY) {
299 		struct net_device *dev = NULL;
300 
301 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
302 			if (addr->l2tp_scope_id)
303 				bound_dev_if = addr->l2tp_scope_id;
304 
305 			/* Binding to link-local address requires an
306 			 * interface.
307 			 */
308 			if (!bound_dev_if)
309 				goto out_unlock_rcu;
310 
311 			err = -ENODEV;
312 			dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
313 			if (!dev)
314 				goto out_unlock_rcu;
315 		}
316 
317 		/* ipv4 addr of the socket is invalid.  Only the
318 		 * unspecified and mapped address have a v4 equivalent.
319 		 */
320 		v4addr = LOOPBACK4_IPV6;
321 		err = -EADDRNOTAVAIL;
322 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
323 			goto out_unlock_rcu;
324 	}
325 	rcu_read_unlock();
326 
327 	write_lock_bh(&l2tp_ip6_lock);
328 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
329 				   addr->l2tp_conn_id)) {
330 		write_unlock_bh(&l2tp_ip6_lock);
331 		err = -EADDRINUSE;
332 		goto out_unlock;
333 	}
334 
335 	inet->inet_saddr = v4addr;
336 	inet->inet_rcv_saddr = v4addr;
337 	sk->sk_bound_dev_if = bound_dev_if;
338 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
339 	np->saddr = addr->l2tp_addr;
340 
341 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
342 
343 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
344 	sk_del_node_init(sk);
345 	write_unlock_bh(&l2tp_ip6_lock);
346 
347 	sock_reset_flag(sk, SOCK_ZAPPED);
348 	release_sock(sk);
349 	return 0;
350 
351 out_unlock_rcu:
352 	rcu_read_unlock();
353 out_unlock:
354 	release_sock(sk);
355 
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 (addr_len < sizeof(*lsa))
369 		return -EINVAL;
370 
371 	if (usin->sin6_family != AF_INET6)
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 	lock_sock(sk);
385 
386 	 /* Must bind first - autobinding does not work */
387 	if (sock_flag(sk, SOCK_ZAPPED)) {
388 		rc = -EINVAL;
389 		goto out_sk;
390 	}
391 
392 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
393 	if (rc < 0)
394 		goto out_sk;
395 
396 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
397 
398 	write_lock_bh(&l2tp_ip6_lock);
399 	hlist_del_init(&sk->sk_bind_node);
400 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
401 	write_unlock_bh(&l2tp_ip6_lock);
402 
403 out_sk:
404 	release_sock(sk);
405 
406 	return rc;
407 }
408 
409 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
410 {
411 	if (sock_flag(sk, SOCK_ZAPPED))
412 		return 0;
413 
414 	return __udp_disconnect(sk, flags);
415 }
416 
417 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
418 			    int peer)
419 {
420 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
421 	struct sock *sk = sock->sk;
422 	struct ipv6_pinfo *np = inet6_sk(sk);
423 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
424 
425 	lsa->l2tp_family = AF_INET6;
426 	lsa->l2tp_flowinfo = 0;
427 	lsa->l2tp_scope_id = 0;
428 	lsa->l2tp_unused = 0;
429 	if (peer) {
430 		if (!lsk->peer_conn_id)
431 			return -ENOTCONN;
432 		lsa->l2tp_conn_id = lsk->peer_conn_id;
433 		lsa->l2tp_addr = sk->sk_v6_daddr;
434 		if (inet6_test_bit(SNDFLOW, sk))
435 			lsa->l2tp_flowinfo = np->flow_label;
436 	} else {
437 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
438 			lsa->l2tp_addr = np->saddr;
439 		else
440 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
441 
442 		lsa->l2tp_conn_id = lsk->conn_id;
443 	}
444 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
445 		lsa->l2tp_scope_id = READ_ONCE(sk->sk_bound_dev_if);
446 	return sizeof(*lsa);
447 }
448 
449 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
450 {
451 	int rc;
452 
453 	/* Charge it to the socket, dropping if the queue is full. */
454 	rc = sock_queue_rcv_skb(sk, skb);
455 	if (rc < 0)
456 		goto drop;
457 
458 	return 0;
459 
460 drop:
461 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
462 	kfree_skb(skb);
463 	return -1;
464 }
465 
466 static int l2tp_ip6_push_pending_frames(struct sock *sk)
467 {
468 	struct sk_buff *skb;
469 	__be32 *transhdr = NULL;
470 	int err = 0;
471 
472 	skb = skb_peek(&sk->sk_write_queue);
473 	if (!skb)
474 		goto out;
475 
476 	transhdr = (__be32 *)skb_transport_header(skb);
477 	*transhdr = 0;
478 
479 	err = ip6_push_pending_frames(sk);
480 
481 out:
482 	return err;
483 }
484 
485 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
486  * control frames.
487  */
488 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
489 {
490 	struct ipv6_txoptions opt_space;
491 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
492 	struct in6_addr *daddr, *final_p, final;
493 	struct ipv6_pinfo *np = inet6_sk(sk);
494 	struct ipv6_txoptions *opt_to_free = NULL;
495 	struct ipv6_txoptions *opt = NULL;
496 	struct ip6_flowlabel *flowlabel = NULL;
497 	struct dst_entry *dst = NULL;
498 	struct flowi6 fl6;
499 	struct ipcm6_cookie ipc6;
500 	int addr_len = msg->msg_namelen;
501 	int transhdrlen = 4; /* zero session-id */
502 	int ulen;
503 	int err;
504 
505 	/* Rough check on arithmetic overflow,
506 	 * better check is made in ip6_append_data().
507 	 */
508 	if (len > INT_MAX - transhdrlen)
509 		return -EMSGSIZE;
510 
511 	/* Mirror BSD error message compatibility */
512 	if (msg->msg_flags & MSG_OOB)
513 		return -EOPNOTSUPP;
514 
515 	/* Get and verify the address */
516 	memset(&fl6, 0, sizeof(fl6));
517 
518 	fl6.flowi6_mark = READ_ONCE(sk->sk_mark);
519 	fl6.flowi6_uid = sk->sk_uid;
520 
521 	ipcm6_init(&ipc6);
522 
523 	if (lsa) {
524 		if (addr_len < SIN6_LEN_RFC2133)
525 			return -EINVAL;
526 
527 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
528 			return -EAFNOSUPPORT;
529 
530 		daddr = &lsa->l2tp_addr;
531 		if (inet6_test_bit(SNDFLOW, sk)) {
532 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
533 			if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
534 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
535 				if (IS_ERR(flowlabel))
536 					return -EINVAL;
537 			}
538 		}
539 
540 		/* Otherwise it will be difficult to maintain
541 		 * sk->sk_dst_cache.
542 		 */
543 		if (sk->sk_state == TCP_ESTABLISHED &&
544 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
545 			daddr = &sk->sk_v6_daddr;
546 
547 		if (addr_len >= sizeof(struct sockaddr_in6) &&
548 		    lsa->l2tp_scope_id &&
549 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
550 			fl6.flowi6_oif = lsa->l2tp_scope_id;
551 	} else {
552 		if (sk->sk_state != TCP_ESTABLISHED)
553 			return -EDESTADDRREQ;
554 
555 		daddr = &sk->sk_v6_daddr;
556 		fl6.flowlabel = np->flow_label;
557 	}
558 
559 	if (fl6.flowi6_oif == 0)
560 		fl6.flowi6_oif = READ_ONCE(sk->sk_bound_dev_if);
561 
562 	if (msg->msg_controllen) {
563 		opt = &opt_space;
564 		memset(opt, 0, sizeof(struct ipv6_txoptions));
565 		opt->tot_len = sizeof(struct ipv6_txoptions);
566 		ipc6.opt = opt;
567 
568 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
569 		if (err < 0) {
570 			fl6_sock_release(flowlabel);
571 			return err;
572 		}
573 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
574 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
575 			if (IS_ERR(flowlabel))
576 				return -EINVAL;
577 		}
578 		if (!(opt->opt_nflen | opt->opt_flen))
579 			opt = NULL;
580 	}
581 
582 	if (!opt) {
583 		opt = txopt_get(np);
584 		opt_to_free = opt;
585 	}
586 	if (flowlabel)
587 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
588 	opt = ipv6_fixup_options(&opt_space, opt);
589 	ipc6.opt = opt;
590 
591 	fl6.flowi6_proto = sk->sk_protocol;
592 	if (!ipv6_addr_any(daddr))
593 		fl6.daddr = *daddr;
594 	else
595 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
596 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
597 		fl6.saddr = np->saddr;
598 
599 	final_p = fl6_update_dst(&fl6, opt, &final);
600 
601 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
602 		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
603 	else if (!fl6.flowi6_oif)
604 		fl6.flowi6_oif = READ_ONCE(np->ucast_oif);
605 
606 	security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
607 
608 	if (ipc6.tclass < 0)
609 		ipc6.tclass = np->tclass;
610 
611 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
612 
613 	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
614 	if (IS_ERR(dst)) {
615 		err = PTR_ERR(dst);
616 		goto out;
617 	}
618 
619 	if (ipc6.hlimit < 0)
620 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
621 
622 	if (ipc6.dontfrag < 0)
623 		ipc6.dontfrag = inet6_test_bit(DONTFRAG, sk);
624 
625 	if (msg->msg_flags & MSG_CONFIRM)
626 		goto do_confirm;
627 
628 back_from_confirm:
629 	lock_sock(sk);
630 	ulen = len + (skb_queue_empty(&sk->sk_write_queue) ? transhdrlen : 0);
631 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
632 			      ulen, transhdrlen, &ipc6,
633 			      &fl6, (struct rt6_info *)dst,
634 			      msg->msg_flags);
635 	if (err)
636 		ip6_flush_pending_frames(sk);
637 	else if (!(msg->msg_flags & MSG_MORE))
638 		err = l2tp_ip6_push_pending_frames(sk);
639 	release_sock(sk);
640 done:
641 	dst_release(dst);
642 out:
643 	fl6_sock_release(flowlabel);
644 	txopt_put(opt_to_free);
645 
646 	return err < 0 ? err : len;
647 
648 do_confirm:
649 	if (msg->msg_flags & MSG_PROBE)
650 		dst_confirm_neigh(dst, &fl6.daddr);
651 	if (!(msg->msg_flags & MSG_PROBE) || len)
652 		goto back_from_confirm;
653 	err = 0;
654 	goto done;
655 }
656 
657 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
658 			    int flags, int *addr_len)
659 {
660 	struct ipv6_pinfo *np = inet6_sk(sk);
661 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
662 	size_t copied = 0;
663 	int err = -EOPNOTSUPP;
664 	struct sk_buff *skb;
665 
666 	if (flags & MSG_OOB)
667 		goto out;
668 
669 	if (flags & MSG_ERRQUEUE)
670 		return ipv6_recv_error(sk, msg, len, addr_len);
671 
672 	skb = skb_recv_datagram(sk, flags, &err);
673 	if (!skb)
674 		goto out;
675 
676 	copied = skb->len;
677 	if (len < copied) {
678 		msg->msg_flags |= MSG_TRUNC;
679 		copied = len;
680 	}
681 
682 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
683 	if (err)
684 		goto done;
685 
686 	sock_recv_timestamp(msg, sk, skb);
687 
688 	/* Copy the address. */
689 	if (lsa) {
690 		lsa->l2tp_family = AF_INET6;
691 		lsa->l2tp_unused = 0;
692 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
693 		lsa->l2tp_flowinfo = 0;
694 		lsa->l2tp_scope_id = 0;
695 		lsa->l2tp_conn_id = 0;
696 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
697 			lsa->l2tp_scope_id = inet6_iif(skb);
698 		*addr_len = sizeof(*lsa);
699 	}
700 
701 	if (np->rxopt.all)
702 		ip6_datagram_recv_ctl(sk, msg, skb);
703 
704 	if (flags & MSG_TRUNC)
705 		copied = skb->len;
706 done:
707 	skb_free_datagram(sk, skb);
708 out:
709 	return err ? err : copied;
710 }
711 
712 static struct proto l2tp_ip6_prot = {
713 	.name		   = "L2TP/IPv6",
714 	.owner		   = THIS_MODULE,
715 	.init		   = l2tp_ip6_open,
716 	.close		   = l2tp_ip6_close,
717 	.bind		   = l2tp_ip6_bind,
718 	.connect	   = l2tp_ip6_connect,
719 	.disconnect	   = l2tp_ip6_disconnect,
720 	.ioctl		   = l2tp_ioctl,
721 	.destroy	   = l2tp_ip6_destroy_sock,
722 	.setsockopt	   = ipv6_setsockopt,
723 	.getsockopt	   = ipv6_getsockopt,
724 	.sendmsg	   = l2tp_ip6_sendmsg,
725 	.recvmsg	   = l2tp_ip6_recvmsg,
726 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
727 	.hash		   = l2tp_ip6_hash,
728 	.unhash		   = l2tp_ip6_unhash,
729 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
730 	.ipv6_pinfo_offset = offsetof(struct l2tp_ip6_sock, inet6),
731 };
732 
733 static const struct proto_ops l2tp_ip6_ops = {
734 	.family		   = PF_INET6,
735 	.owner		   = THIS_MODULE,
736 	.release	   = inet6_release,
737 	.bind		   = inet6_bind,
738 	.connect	   = inet_dgram_connect,
739 	.socketpair	   = sock_no_socketpair,
740 	.accept		   = sock_no_accept,
741 	.getname	   = l2tp_ip6_getname,
742 	.poll		   = datagram_poll,
743 	.ioctl		   = inet6_ioctl,
744 	.gettstamp	   = sock_gettstamp,
745 	.listen		   = sock_no_listen,
746 	.shutdown	   = inet_shutdown,
747 	.setsockopt	   = sock_common_setsockopt,
748 	.getsockopt	   = sock_common_getsockopt,
749 	.sendmsg	   = inet_sendmsg,
750 	.recvmsg	   = sock_common_recvmsg,
751 	.mmap		   = sock_no_mmap,
752 #ifdef CONFIG_COMPAT
753 	.compat_ioctl	   = inet6_compat_ioctl,
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 };
763 
764 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
765 	.handler	= l2tp_ip6_recv,
766 };
767 
768 static int __init l2tp_ip6_init(void)
769 {
770 	int err;
771 
772 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
773 
774 	err = proto_register(&l2tp_ip6_prot, 1);
775 	if (err != 0)
776 		goto out;
777 
778 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
779 	if (err)
780 		goto out1;
781 
782 	inet6_register_protosw(&l2tp_ip6_protosw);
783 	return 0;
784 
785 out1:
786 	proto_unregister(&l2tp_ip6_prot);
787 out:
788 	return err;
789 }
790 
791 static void __exit l2tp_ip6_exit(void)
792 {
793 	inet6_unregister_protosw(&l2tp_ip6_protosw);
794 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
795 	proto_unregister(&l2tp_ip6_prot);
796 }
797 
798 module_init(l2tp_ip6_init);
799 module_exit(l2tp_ip6_exit);
800 
801 MODULE_LICENSE("GPL");
802 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
803 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
804 MODULE_VERSION("1.0");
805 
806 /* Use the values of SOCK_DGRAM (2) as type and IPPROTO_L2TP (115) as protocol,
807  * because __stringify doesn't like enums
808  */
809 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 115, 2);
810 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, 115);
811