xref: /linux/net/l2tp/l2tp_ip6.c (revision 10fbe8c082fdde74b1f0814bc30e194cf330cdf7)
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 #include <net/net_namespace.h>
26 #include <net/netns/generic.h>
27 
28 #include <net/transp_v6.h>
29 #include <net/addrconf.h>
30 #include <net/ip6_route.h>
31 
32 #include "l2tp_core.h"
33 
34 /* per-net private data for this module */
35 static unsigned int l2tp_ip6_net_id;
36 struct l2tp_ip6_net {
37 	rwlock_t l2tp_ip6_lock;
38 	struct hlist_head l2tp_ip6_table;
39 	struct hlist_head l2tp_ip6_bind_table;
40 };
41 
42 struct l2tp_ip6_sock {
43 	/* inet_sock has to be the first member of l2tp_ip6_sock */
44 	struct inet_sock	inet;
45 
46 	u32			conn_id;
47 	u32			peer_conn_id;
48 
49 	struct ipv6_pinfo	inet6;
50 };
51 
52 static struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
53 {
54 	return (struct l2tp_ip6_sock *)sk;
55 }
56 
57 static struct l2tp_ip6_net *l2tp_ip6_pernet(const struct net *net)
58 {
59 	return net_generic(net, l2tp_ip6_net_id);
60 }
61 
62 static struct sock *__l2tp_ip6_bind_lookup(const struct net *net,
63 					   const struct in6_addr *laddr,
64 					   const struct in6_addr *raddr,
65 					   int dif, u32 tunnel_id)
66 {
67 	struct l2tp_ip6_net *pn = l2tp_ip6_pernet(net);
68 	struct sock *sk;
69 
70 	sk_for_each_bound(sk, &pn->l2tp_ip6_bind_table) {
71 		const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
72 		const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
73 		const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
74 		int bound_dev_if;
75 
76 		if (!net_eq(sock_net(sk), net))
77 			continue;
78 
79 		bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
80 		if (bound_dev_if && dif && bound_dev_if != dif)
81 			continue;
82 
83 		if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
84 		    !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
85 			continue;
86 
87 		if (!ipv6_addr_any(sk_raddr) && raddr &&
88 		    !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
89 			continue;
90 
91 		if (l2tp->conn_id != tunnel_id)
92 			continue;
93 
94 		goto found;
95 	}
96 
97 	sk = NULL;
98 found:
99 	return sk;
100 }
101 
102 /* When processing receive frames, there are two cases to
103  * consider. Data frames consist of a non-zero session-id and an
104  * optional cookie. Control frames consist of a regular L2TP header
105  * preceded by 32-bits of zeros.
106  *
107  * L2TPv3 Session 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  * |                           Session ID                          |
113  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114  * |               Cookie (optional, maximum 64 bits)...
115  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
116  *                                                                 |
117  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
118  *
119  * L2TPv3 Control Message Header Over IP
120  *
121  *  0                   1                   2                   3
122  *  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
123  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
124  * |                      (32 bits of zeros)                       |
125  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
126  * |T|L|x|x|S|x|x|x|x|x|x|x|  Ver  |             Length            |
127  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
128  * |                     Control Connection ID                     |
129  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
130  * |               Ns              |               Nr              |
131  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
132  *
133  * All control frames are passed to userspace.
134  */
135 static int l2tp_ip6_recv(struct sk_buff *skb)
136 {
137 	struct net *net = dev_net(skb->dev);
138 	struct l2tp_ip6_net *pn;
139 	struct sock *sk;
140 	u32 session_id;
141 	u32 tunnel_id;
142 	unsigned char *ptr, *optr;
143 	struct l2tp_session *session;
144 	struct l2tp_tunnel *tunnel = NULL;
145 	struct ipv6hdr *iph;
146 
147 	pn = l2tp_ip6_pernet(net);
148 
149 	if (!pskb_may_pull(skb, 4))
150 		goto discard;
151 
152 	/* Point to L2TP header */
153 	optr = skb->data;
154 	ptr = skb->data;
155 	session_id = ntohl(*((__be32 *)ptr));
156 	ptr += 4;
157 
158 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
159 	 * the session_id. If it is 0, the packet is a L2TP control
160 	 * frame and the session_id value can be discarded.
161 	 */
162 	if (session_id == 0) {
163 		__skb_pull(skb, 4);
164 		goto pass_up;
165 	}
166 
167 	/* Ok, this is a data packet. Lookup the session. */
168 	session = l2tp_v3_session_get(net, NULL, session_id);
169 	if (!session)
170 		goto discard;
171 
172 	tunnel = session->tunnel;
173 	if (!tunnel)
174 		goto discard_sess;
175 
176 	if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
177 		goto discard_sess;
178 
179 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
180 	l2tp_session_put(session);
181 
182 	return 0;
183 
184 pass_up:
185 	/* Get the tunnel_id from the L2TP header */
186 	if (!pskb_may_pull(skb, 12))
187 		goto discard;
188 
189 	if ((skb->data[0] & 0xc0) != 0xc0)
190 		goto discard;
191 
192 	tunnel_id = ntohl(*(__be32 *)&skb->data[4]);
193 	iph = ipv6_hdr(skb);
194 
195 	read_lock_bh(&pn->l2tp_ip6_lock);
196 	sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
197 				    inet6_iif(skb), tunnel_id);
198 	if (!sk) {
199 		read_unlock_bh(&pn->l2tp_ip6_lock);
200 		goto discard;
201 	}
202 	sock_hold(sk);
203 	read_unlock_bh(&pn->l2tp_ip6_lock);
204 
205 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
206 		goto discard_put;
207 
208 	nf_reset_ct(skb);
209 
210 	return sk_receive_skb(sk, skb, 1);
211 
212 discard_sess:
213 	l2tp_session_put(session);
214 	goto discard;
215 
216 discard_put:
217 	sock_put(sk);
218 
219 discard:
220 	kfree_skb(skb);
221 	return 0;
222 }
223 
224 static int l2tp_ip6_hash(struct sock *sk)
225 {
226 	struct l2tp_ip6_net *pn = l2tp_ip6_pernet(sock_net(sk));
227 
228 	if (sk_unhashed(sk)) {
229 		write_lock_bh(&pn->l2tp_ip6_lock);
230 		sk_add_node(sk, &pn->l2tp_ip6_table);
231 		write_unlock_bh(&pn->l2tp_ip6_lock);
232 	}
233 	return 0;
234 }
235 
236 static void l2tp_ip6_unhash(struct sock *sk)
237 {
238 	struct l2tp_ip6_net *pn = l2tp_ip6_pernet(sock_net(sk));
239 
240 	if (sk_unhashed(sk))
241 		return;
242 	write_lock_bh(&pn->l2tp_ip6_lock);
243 	sk_del_node_init(sk);
244 	write_unlock_bh(&pn->l2tp_ip6_lock);
245 }
246 
247 static int l2tp_ip6_open(struct sock *sk)
248 {
249 	/* Prevent autobind. We don't have ports. */
250 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
251 
252 	l2tp_ip6_hash(sk);
253 	return 0;
254 }
255 
256 static void l2tp_ip6_close(struct sock *sk, long timeout)
257 {
258 	struct l2tp_ip6_net *pn = l2tp_ip6_pernet(sock_net(sk));
259 
260 	write_lock_bh(&pn->l2tp_ip6_lock);
261 	hlist_del_init(&sk->sk_bind_node);
262 	sk_del_node_init(sk);
263 	write_unlock_bh(&pn->l2tp_ip6_lock);
264 
265 	sk_common_release(sk);
266 }
267 
268 static void l2tp_ip6_destroy_sock(struct sock *sk)
269 {
270 	struct l2tp_tunnel *tunnel;
271 
272 	lock_sock(sk);
273 	ip6_flush_pending_frames(sk);
274 	release_sock(sk);
275 
276 	tunnel = l2tp_sk_to_tunnel(sk);
277 	if (tunnel) {
278 		l2tp_tunnel_delete(tunnel);
279 		l2tp_tunnel_put(tunnel);
280 	}
281 }
282 
283 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
284 {
285 	struct inet_sock *inet = inet_sk(sk);
286 	struct ipv6_pinfo *np = inet6_sk(sk);
287 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *)uaddr;
288 	struct net *net = sock_net(sk);
289 	struct l2tp_ip6_net *pn;
290 	__be32 v4addr = 0;
291 	int bound_dev_if;
292 	int addr_type;
293 	int err;
294 
295 	pn = l2tp_ip6_pernet(net);
296 
297 	if (addr->l2tp_family != AF_INET6)
298 		return -EINVAL;
299 	if (addr_len < sizeof(*addr))
300 		return -EINVAL;
301 
302 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
303 
304 	/* l2tp_ip6 sockets are IPv6 only */
305 	if (addr_type == IPV6_ADDR_MAPPED)
306 		return -EADDRNOTAVAIL;
307 
308 	/* L2TP is point-point, not multicast */
309 	if (addr_type & IPV6_ADDR_MULTICAST)
310 		return -EADDRNOTAVAIL;
311 
312 	lock_sock(sk);
313 
314 	err = -EINVAL;
315 	if (!sock_flag(sk, SOCK_ZAPPED))
316 		goto out_unlock;
317 
318 	if (sk->sk_state != TCP_CLOSE)
319 		goto out_unlock;
320 
321 	bound_dev_if = sk->sk_bound_dev_if;
322 
323 	/* Check if the address belongs to the host. */
324 	rcu_read_lock();
325 	if (addr_type != IPV6_ADDR_ANY) {
326 		struct net_device *dev = NULL;
327 
328 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
329 			if (addr->l2tp_scope_id)
330 				bound_dev_if = addr->l2tp_scope_id;
331 
332 			/* Binding to link-local address requires an
333 			 * interface.
334 			 */
335 			if (!bound_dev_if)
336 				goto out_unlock_rcu;
337 
338 			err = -ENODEV;
339 			dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
340 			if (!dev)
341 				goto out_unlock_rcu;
342 		}
343 
344 		/* ipv4 addr of the socket is invalid.  Only the
345 		 * unspecified and mapped address have a v4 equivalent.
346 		 */
347 		v4addr = LOOPBACK4_IPV6;
348 		err = -EADDRNOTAVAIL;
349 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
350 			goto out_unlock_rcu;
351 	}
352 	rcu_read_unlock();
353 
354 	write_lock_bh(&pn->l2tp_ip6_lock);
355 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
356 				   addr->l2tp_conn_id)) {
357 		write_unlock_bh(&pn->l2tp_ip6_lock);
358 		err = -EADDRINUSE;
359 		goto out_unlock;
360 	}
361 
362 	inet->inet_saddr = v4addr;
363 	inet->inet_rcv_saddr = v4addr;
364 	sk->sk_bound_dev_if = bound_dev_if;
365 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
366 	np->saddr = addr->l2tp_addr;
367 
368 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
369 
370 	sk_add_bind_node(sk, &pn->l2tp_ip6_bind_table);
371 	sk_del_node_init(sk);
372 	write_unlock_bh(&pn->l2tp_ip6_lock);
373 
374 	sock_reset_flag(sk, SOCK_ZAPPED);
375 	release_sock(sk);
376 	return 0;
377 
378 out_unlock_rcu:
379 	rcu_read_unlock();
380 out_unlock:
381 	release_sock(sk);
382 
383 	return err;
384 }
385 
386 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
387 			    int addr_len)
388 {
389 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
390 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *)uaddr;
391 	struct in6_addr	*daddr;
392 	int	addr_type;
393 	int rc;
394 	struct l2tp_ip6_net *pn;
395 
396 	if (addr_len < sizeof(*lsa))
397 		return -EINVAL;
398 
399 	if (usin->sin6_family != AF_INET6)
400 		return -EINVAL;
401 
402 	addr_type = ipv6_addr_type(&usin->sin6_addr);
403 	if (addr_type & IPV6_ADDR_MULTICAST)
404 		return -EINVAL;
405 
406 	if (addr_type & IPV6_ADDR_MAPPED) {
407 		daddr = &usin->sin6_addr;
408 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
409 			return -EINVAL;
410 	}
411 
412 	lock_sock(sk);
413 
414 	 /* Must bind first - autobinding does not work */
415 	if (sock_flag(sk, SOCK_ZAPPED)) {
416 		rc = -EINVAL;
417 		goto out_sk;
418 	}
419 
420 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
421 	if (rc < 0)
422 		goto out_sk;
423 
424 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
425 
426 	pn = l2tp_ip6_pernet(sock_net(sk));
427 	write_lock_bh(&pn->l2tp_ip6_lock);
428 	hlist_del_init(&sk->sk_bind_node);
429 	sk_add_bind_node(sk, &pn->l2tp_ip6_bind_table);
430 	write_unlock_bh(&pn->l2tp_ip6_lock);
431 
432 out_sk:
433 	release_sock(sk);
434 
435 	return rc;
436 }
437 
438 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
439 {
440 	if (sock_flag(sk, SOCK_ZAPPED))
441 		return 0;
442 
443 	return __udp_disconnect(sk, flags);
444 }
445 
446 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
447 			    int peer)
448 {
449 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
450 	struct sock *sk = sock->sk;
451 	struct ipv6_pinfo *np = inet6_sk(sk);
452 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
453 
454 	lsa->l2tp_family = AF_INET6;
455 	lsa->l2tp_flowinfo = 0;
456 	lsa->l2tp_scope_id = 0;
457 	lsa->l2tp_unused = 0;
458 	if (peer) {
459 		if (!lsk->peer_conn_id)
460 			return -ENOTCONN;
461 		lsa->l2tp_conn_id = lsk->peer_conn_id;
462 		lsa->l2tp_addr = sk->sk_v6_daddr;
463 		if (inet6_test_bit(SNDFLOW, sk))
464 			lsa->l2tp_flowinfo = np->flow_label;
465 	} else {
466 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
467 			lsa->l2tp_addr = np->saddr;
468 		else
469 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
470 
471 		lsa->l2tp_conn_id = lsk->conn_id;
472 	}
473 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
474 		lsa->l2tp_scope_id = READ_ONCE(sk->sk_bound_dev_if);
475 	return sizeof(*lsa);
476 }
477 
478 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
479 {
480 	int rc;
481 
482 	/* Charge it to the socket, dropping if the queue is full. */
483 	rc = sock_queue_rcv_skb(sk, skb);
484 	if (rc < 0)
485 		goto drop;
486 
487 	return 0;
488 
489 drop:
490 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
491 	kfree_skb(skb);
492 	return -1;
493 }
494 
495 static int l2tp_ip6_push_pending_frames(struct sock *sk)
496 {
497 	struct sk_buff *skb;
498 	__be32 *transhdr = NULL;
499 	int err = 0;
500 
501 	skb = skb_peek(&sk->sk_write_queue);
502 	if (!skb)
503 		goto out;
504 
505 	transhdr = (__be32 *)skb_transport_header(skb);
506 	*transhdr = 0;
507 
508 	err = ip6_push_pending_frames(sk);
509 
510 out:
511 	return err;
512 }
513 
514 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
515  * control frames.
516  */
517 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
518 {
519 	struct ipv6_txoptions opt_space;
520 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
521 	struct in6_addr *daddr, *final_p, final;
522 	struct ipv6_pinfo *np = inet6_sk(sk);
523 	struct ipv6_txoptions *opt_to_free = NULL;
524 	struct ipv6_txoptions *opt = NULL;
525 	struct ip6_flowlabel *flowlabel = NULL;
526 	struct dst_entry *dst = NULL;
527 	struct flowi6 fl6;
528 	struct ipcm6_cookie ipc6;
529 	int addr_len = msg->msg_namelen;
530 	int transhdrlen = 4; /* zero session-id */
531 	int ulen;
532 	int err;
533 
534 	/* Rough check on arithmetic overflow,
535 	 * better check is made in ip6_append_data().
536 	 */
537 	if (len > INT_MAX - transhdrlen)
538 		return -EMSGSIZE;
539 
540 	/* Mirror BSD error message compatibility */
541 	if (msg->msg_flags & MSG_OOB)
542 		return -EOPNOTSUPP;
543 
544 	/* Get and verify the address */
545 	memset(&fl6, 0, sizeof(fl6));
546 
547 	fl6.flowi6_mark = READ_ONCE(sk->sk_mark);
548 	fl6.flowi6_uid = sk->sk_uid;
549 
550 	ipcm6_init(&ipc6);
551 
552 	if (lsa) {
553 		if (addr_len < SIN6_LEN_RFC2133)
554 			return -EINVAL;
555 
556 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
557 			return -EAFNOSUPPORT;
558 
559 		daddr = &lsa->l2tp_addr;
560 		if (inet6_test_bit(SNDFLOW, sk)) {
561 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
562 			if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
563 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
564 				if (IS_ERR(flowlabel))
565 					return -EINVAL;
566 			}
567 		}
568 
569 		/* Otherwise it will be difficult to maintain
570 		 * sk->sk_dst_cache.
571 		 */
572 		if (sk->sk_state == TCP_ESTABLISHED &&
573 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
574 			daddr = &sk->sk_v6_daddr;
575 
576 		if (addr_len >= sizeof(struct sockaddr_in6) &&
577 		    lsa->l2tp_scope_id &&
578 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
579 			fl6.flowi6_oif = lsa->l2tp_scope_id;
580 	} else {
581 		if (sk->sk_state != TCP_ESTABLISHED)
582 			return -EDESTADDRREQ;
583 
584 		daddr = &sk->sk_v6_daddr;
585 		fl6.flowlabel = np->flow_label;
586 	}
587 
588 	if (fl6.flowi6_oif == 0)
589 		fl6.flowi6_oif = READ_ONCE(sk->sk_bound_dev_if);
590 
591 	if (msg->msg_controllen) {
592 		opt = &opt_space;
593 		memset(opt, 0, sizeof(struct ipv6_txoptions));
594 		opt->tot_len = sizeof(struct ipv6_txoptions);
595 		ipc6.opt = opt;
596 
597 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
598 		if (err < 0) {
599 			fl6_sock_release(flowlabel);
600 			return err;
601 		}
602 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
603 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
604 			if (IS_ERR(flowlabel))
605 				return -EINVAL;
606 		}
607 		if (!(opt->opt_nflen | opt->opt_flen))
608 			opt = NULL;
609 	}
610 
611 	if (!opt) {
612 		opt = txopt_get(np);
613 		opt_to_free = opt;
614 	}
615 	if (flowlabel)
616 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
617 	opt = ipv6_fixup_options(&opt_space, opt);
618 	ipc6.opt = opt;
619 
620 	fl6.flowi6_proto = sk->sk_protocol;
621 	if (!ipv6_addr_any(daddr))
622 		fl6.daddr = *daddr;
623 	else
624 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
625 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
626 		fl6.saddr = np->saddr;
627 
628 	final_p = fl6_update_dst(&fl6, opt, &final);
629 
630 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
631 		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
632 	else if (!fl6.flowi6_oif)
633 		fl6.flowi6_oif = READ_ONCE(np->ucast_oif);
634 
635 	security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
636 
637 	if (ipc6.tclass < 0)
638 		ipc6.tclass = np->tclass;
639 
640 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
641 
642 	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
643 	if (IS_ERR(dst)) {
644 		err = PTR_ERR(dst);
645 		goto out;
646 	}
647 
648 	if (ipc6.hlimit < 0)
649 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
650 
651 	if (ipc6.dontfrag < 0)
652 		ipc6.dontfrag = inet6_test_bit(DONTFRAG, sk);
653 
654 	if (msg->msg_flags & MSG_CONFIRM)
655 		goto do_confirm;
656 
657 back_from_confirm:
658 	lock_sock(sk);
659 	ulen = len + (skb_queue_empty(&sk->sk_write_queue) ? transhdrlen : 0);
660 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
661 			      ulen, transhdrlen, &ipc6,
662 			      &fl6, dst_rt6_info(dst),
663 			      msg->msg_flags);
664 	if (err)
665 		ip6_flush_pending_frames(sk);
666 	else if (!(msg->msg_flags & MSG_MORE))
667 		err = l2tp_ip6_push_pending_frames(sk);
668 	release_sock(sk);
669 done:
670 	dst_release(dst);
671 out:
672 	fl6_sock_release(flowlabel);
673 	txopt_put(opt_to_free);
674 
675 	return err < 0 ? err : len;
676 
677 do_confirm:
678 	if (msg->msg_flags & MSG_PROBE)
679 		dst_confirm_neigh(dst, &fl6.daddr);
680 	if (!(msg->msg_flags & MSG_PROBE) || len)
681 		goto back_from_confirm;
682 	err = 0;
683 	goto done;
684 }
685 
686 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
687 			    int flags, int *addr_len)
688 {
689 	struct ipv6_pinfo *np = inet6_sk(sk);
690 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
691 	size_t copied = 0;
692 	int err = -EOPNOTSUPP;
693 	struct sk_buff *skb;
694 
695 	if (flags & MSG_OOB)
696 		goto out;
697 
698 	if (flags & MSG_ERRQUEUE)
699 		return ipv6_recv_error(sk, msg, len, addr_len);
700 
701 	skb = skb_recv_datagram(sk, flags, &err);
702 	if (!skb)
703 		goto out;
704 
705 	copied = skb->len;
706 	if (len < copied) {
707 		msg->msg_flags |= MSG_TRUNC;
708 		copied = len;
709 	}
710 
711 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
712 	if (err)
713 		goto done;
714 
715 	sock_recv_timestamp(msg, sk, skb);
716 
717 	/* Copy the address. */
718 	if (lsa) {
719 		lsa->l2tp_family = AF_INET6;
720 		lsa->l2tp_unused = 0;
721 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
722 		lsa->l2tp_flowinfo = 0;
723 		lsa->l2tp_scope_id = 0;
724 		lsa->l2tp_conn_id = 0;
725 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
726 			lsa->l2tp_scope_id = inet6_iif(skb);
727 		*addr_len = sizeof(*lsa);
728 	}
729 
730 	if (np->rxopt.all)
731 		ip6_datagram_recv_ctl(sk, msg, skb);
732 
733 	if (flags & MSG_TRUNC)
734 		copied = skb->len;
735 done:
736 	skb_free_datagram(sk, skb);
737 out:
738 	return err ? err : copied;
739 }
740 
741 static struct proto l2tp_ip6_prot = {
742 	.name		   = "L2TP/IPv6",
743 	.owner		   = THIS_MODULE,
744 	.init		   = l2tp_ip6_open,
745 	.close		   = l2tp_ip6_close,
746 	.bind		   = l2tp_ip6_bind,
747 	.connect	   = l2tp_ip6_connect,
748 	.disconnect	   = l2tp_ip6_disconnect,
749 	.ioctl		   = l2tp_ioctl,
750 	.destroy	   = l2tp_ip6_destroy_sock,
751 	.setsockopt	   = ipv6_setsockopt,
752 	.getsockopt	   = ipv6_getsockopt,
753 	.sendmsg	   = l2tp_ip6_sendmsg,
754 	.recvmsg	   = l2tp_ip6_recvmsg,
755 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
756 	.hash		   = l2tp_ip6_hash,
757 	.unhash		   = l2tp_ip6_unhash,
758 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
759 	.ipv6_pinfo_offset = offsetof(struct l2tp_ip6_sock, inet6),
760 };
761 
762 static const struct proto_ops l2tp_ip6_ops = {
763 	.family		   = PF_INET6,
764 	.owner		   = THIS_MODULE,
765 	.release	   = inet6_release,
766 	.bind		   = inet6_bind,
767 	.connect	   = inet_dgram_connect,
768 	.socketpair	   = sock_no_socketpair,
769 	.accept		   = sock_no_accept,
770 	.getname	   = l2tp_ip6_getname,
771 	.poll		   = datagram_poll,
772 	.ioctl		   = inet6_ioctl,
773 	.gettstamp	   = sock_gettstamp,
774 	.listen		   = sock_no_listen,
775 	.shutdown	   = inet_shutdown,
776 	.setsockopt	   = sock_common_setsockopt,
777 	.getsockopt	   = sock_common_getsockopt,
778 	.sendmsg	   = inet_sendmsg,
779 	.recvmsg	   = sock_common_recvmsg,
780 	.mmap		   = sock_no_mmap,
781 #ifdef CONFIG_COMPAT
782 	.compat_ioctl	   = inet6_compat_ioctl,
783 #endif
784 };
785 
786 static struct inet_protosw l2tp_ip6_protosw = {
787 	.type		= SOCK_DGRAM,
788 	.protocol	= IPPROTO_L2TP,
789 	.prot		= &l2tp_ip6_prot,
790 	.ops		= &l2tp_ip6_ops,
791 };
792 
793 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
794 	.handler	= l2tp_ip6_recv,
795 };
796 
797 static __net_init int l2tp_ip6_init_net(struct net *net)
798 {
799 	struct l2tp_ip6_net *pn = net_generic(net, l2tp_ip6_net_id);
800 
801 	rwlock_init(&pn->l2tp_ip6_lock);
802 	INIT_HLIST_HEAD(&pn->l2tp_ip6_table);
803 	INIT_HLIST_HEAD(&pn->l2tp_ip6_bind_table);
804 	return 0;
805 }
806 
807 static __net_exit void l2tp_ip6_exit_net(struct net *net)
808 {
809 	struct l2tp_ip6_net *pn = l2tp_ip6_pernet(net);
810 
811 	write_lock_bh(&pn->l2tp_ip6_lock);
812 	WARN_ON_ONCE(hlist_count_nodes(&pn->l2tp_ip6_table) != 0);
813 	WARN_ON_ONCE(hlist_count_nodes(&pn->l2tp_ip6_bind_table) != 0);
814 	write_unlock_bh(&pn->l2tp_ip6_lock);
815 }
816 
817 static struct pernet_operations l2tp_ip6_net_ops = {
818 	.init = l2tp_ip6_init_net,
819 	.exit = l2tp_ip6_exit_net,
820 	.id   = &l2tp_ip6_net_id,
821 	.size = sizeof(struct l2tp_ip6_net),
822 };
823 
824 static int __init l2tp_ip6_init(void)
825 {
826 	int err;
827 
828 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
829 
830 	err = register_pernet_device(&l2tp_ip6_net_ops);
831 	if (err)
832 		goto out;
833 
834 	err = proto_register(&l2tp_ip6_prot, 1);
835 	if (err != 0)
836 		goto out1;
837 
838 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
839 	if (err)
840 		goto out2;
841 
842 	inet6_register_protosw(&l2tp_ip6_protosw);
843 	return 0;
844 
845 out2:
846 	proto_unregister(&l2tp_ip6_prot);
847 out1:
848 	unregister_pernet_device(&l2tp_ip6_net_ops);
849 out:
850 	return err;
851 }
852 
853 static void __exit l2tp_ip6_exit(void)
854 {
855 	inet6_unregister_protosw(&l2tp_ip6_protosw);
856 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
857 	proto_unregister(&l2tp_ip6_prot);
858 	unregister_pernet_device(&l2tp_ip6_net_ops);
859 }
860 
861 module_init(l2tp_ip6_init);
862 module_exit(l2tp_ip6_exit);
863 
864 MODULE_LICENSE("GPL");
865 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
866 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
867 MODULE_VERSION("1.0");
868 
869 /* Use the values of SOCK_DGRAM (2) as type and IPPROTO_L2TP (115) as protocol,
870  * because __stringify doesn't like enums
871  */
872 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 115, 2);
873 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, 115);
874