xref: /linux/net/l2tp/l2tp_ip6.c (revision 8f7aa3d3c7323f4ca2768a9e74ebbe359c4f8f88)
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_unsized *uaddr,
284 			 int addr_len)
285 {
286 	struct inet_sock *inet = inet_sk(sk);
287 	struct ipv6_pinfo *np = inet6_sk(sk);
288 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *)uaddr;
289 	struct net *net = sock_net(sk);
290 	struct l2tp_ip6_net *pn;
291 	__be32 v4addr = 0;
292 	int bound_dev_if;
293 	int addr_type;
294 	int err;
295 
296 	pn = l2tp_ip6_pernet(net);
297 
298 	if (addr->l2tp_family != AF_INET6)
299 		return -EINVAL;
300 	if (addr_len < sizeof(*addr))
301 		return -EINVAL;
302 
303 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
304 
305 	/* l2tp_ip6 sockets are IPv6 only */
306 	if (addr_type == IPV6_ADDR_MAPPED)
307 		return -EADDRNOTAVAIL;
308 
309 	/* L2TP is point-point, not multicast */
310 	if (addr_type & IPV6_ADDR_MULTICAST)
311 		return -EADDRNOTAVAIL;
312 
313 	lock_sock(sk);
314 
315 	err = -EINVAL;
316 	if (!sock_flag(sk, SOCK_ZAPPED))
317 		goto out_unlock;
318 
319 	if (sk->sk_state != TCP_CLOSE)
320 		goto out_unlock;
321 
322 	bound_dev_if = sk->sk_bound_dev_if;
323 
324 	/* Check if the address belongs to the host. */
325 	rcu_read_lock();
326 	if (addr_type != IPV6_ADDR_ANY) {
327 		struct net_device *dev = NULL;
328 
329 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
330 			if (addr->l2tp_scope_id)
331 				bound_dev_if = addr->l2tp_scope_id;
332 
333 			/* Binding to link-local address requires an
334 			 * interface.
335 			 */
336 			if (!bound_dev_if)
337 				goto out_unlock_rcu;
338 
339 			err = -ENODEV;
340 			dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
341 			if (!dev)
342 				goto out_unlock_rcu;
343 		}
344 
345 		/* ipv4 addr of the socket is invalid.  Only the
346 		 * unspecified and mapped address have a v4 equivalent.
347 		 */
348 		v4addr = LOOPBACK4_IPV6;
349 		err = -EADDRNOTAVAIL;
350 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
351 			goto out_unlock_rcu;
352 	}
353 	rcu_read_unlock();
354 
355 	write_lock_bh(&pn->l2tp_ip6_lock);
356 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
357 				   addr->l2tp_conn_id)) {
358 		write_unlock_bh(&pn->l2tp_ip6_lock);
359 		err = -EADDRINUSE;
360 		goto out_unlock;
361 	}
362 
363 	inet->inet_saddr = v4addr;
364 	inet->inet_rcv_saddr = v4addr;
365 	sk->sk_bound_dev_if = bound_dev_if;
366 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
367 	np->saddr = addr->l2tp_addr;
368 
369 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
370 
371 	sk_add_bind_node(sk, &pn->l2tp_ip6_bind_table);
372 	sk_del_node_init(sk);
373 	write_unlock_bh(&pn->l2tp_ip6_lock);
374 
375 	sock_reset_flag(sk, SOCK_ZAPPED);
376 	release_sock(sk);
377 	return 0;
378 
379 out_unlock_rcu:
380 	rcu_read_unlock();
381 out_unlock:
382 	release_sock(sk);
383 
384 	return err;
385 }
386 
387 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr_unsized *uaddr,
388 			    int addr_len)
389 {
390 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
391 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *)uaddr;
392 	struct in6_addr	*daddr;
393 	int	addr_type;
394 	int rc;
395 	struct l2tp_ip6_net *pn;
396 
397 	if (addr_len < sizeof(*lsa))
398 		return -EINVAL;
399 
400 	if (usin->sin6_family != AF_INET6)
401 		return -EINVAL;
402 
403 	addr_type = ipv6_addr_type(&usin->sin6_addr);
404 	if (addr_type & IPV6_ADDR_MULTICAST)
405 		return -EINVAL;
406 
407 	if (addr_type & IPV6_ADDR_MAPPED) {
408 		daddr = &usin->sin6_addr;
409 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
410 			return -EINVAL;
411 	}
412 
413 	lock_sock(sk);
414 
415 	 /* Must bind first - autobinding does not work */
416 	if (sock_flag(sk, SOCK_ZAPPED)) {
417 		rc = -EINVAL;
418 		goto out_sk;
419 	}
420 
421 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
422 	if (rc < 0)
423 		goto out_sk;
424 
425 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
426 
427 	pn = l2tp_ip6_pernet(sock_net(sk));
428 	write_lock_bh(&pn->l2tp_ip6_lock);
429 	hlist_del_init(&sk->sk_bind_node);
430 	sk_add_bind_node(sk, &pn->l2tp_ip6_bind_table);
431 	write_unlock_bh(&pn->l2tp_ip6_lock);
432 
433 out_sk:
434 	release_sock(sk);
435 
436 	return rc;
437 }
438 
439 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
440 {
441 	if (sock_flag(sk, SOCK_ZAPPED))
442 		return 0;
443 
444 	return __udp_disconnect(sk, flags);
445 }
446 
447 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
448 			    int peer)
449 {
450 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
451 	struct sock *sk = sock->sk;
452 	struct ipv6_pinfo *np = inet6_sk(sk);
453 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
454 
455 	lsa->l2tp_family = AF_INET6;
456 	lsa->l2tp_flowinfo = 0;
457 	lsa->l2tp_scope_id = 0;
458 	lsa->l2tp_unused = 0;
459 	if (peer) {
460 		if (!lsk->peer_conn_id)
461 			return -ENOTCONN;
462 		lsa->l2tp_conn_id = lsk->peer_conn_id;
463 		lsa->l2tp_addr = sk->sk_v6_daddr;
464 		if (inet6_test_bit(SNDFLOW, sk))
465 			lsa->l2tp_flowinfo = np->flow_label;
466 	} else {
467 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
468 			lsa->l2tp_addr = np->saddr;
469 		else
470 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
471 
472 		lsa->l2tp_conn_id = lsk->conn_id;
473 	}
474 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
475 		lsa->l2tp_scope_id = READ_ONCE(sk->sk_bound_dev_if);
476 	return sizeof(*lsa);
477 }
478 
479 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
480 {
481 	int rc;
482 
483 	/* Charge it to the socket, dropping if the queue is full. */
484 	rc = sock_queue_rcv_skb(sk, skb);
485 	if (rc < 0)
486 		goto drop;
487 
488 	return 0;
489 
490 drop:
491 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
492 	kfree_skb(skb);
493 	return -1;
494 }
495 
496 static int l2tp_ip6_push_pending_frames(struct sock *sk)
497 {
498 	struct sk_buff *skb;
499 	__be32 *transhdr = NULL;
500 	int err = 0;
501 
502 	skb = skb_peek(&sk->sk_write_queue);
503 	if (!skb)
504 		goto out;
505 
506 	transhdr = (__be32 *)skb_transport_header(skb);
507 	*transhdr = 0;
508 
509 	err = ip6_push_pending_frames(sk);
510 
511 out:
512 	return err;
513 }
514 
515 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
516  * control frames.
517  */
518 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
519 {
520 	struct ipv6_txoptions opt_space;
521 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
522 	struct in6_addr *daddr, *final_p, final;
523 	struct ipv6_pinfo *np = inet6_sk(sk);
524 	struct ipv6_txoptions *opt_to_free = NULL;
525 	struct ipv6_txoptions *opt = NULL;
526 	struct ip6_flowlabel *flowlabel = NULL;
527 	struct dst_entry *dst = NULL;
528 	struct flowi6 fl6;
529 	struct ipcm6_cookie ipc6;
530 	int addr_len = msg->msg_namelen;
531 	int transhdrlen = 4; /* zero session-id */
532 	int ulen;
533 	int err;
534 
535 	/* Rough check on arithmetic overflow,
536 	 * better check is made in ip6_append_data().
537 	 */
538 	if (len > INT_MAX - transhdrlen)
539 		return -EMSGSIZE;
540 
541 	/* Mirror BSD error message compatibility */
542 	if (msg->msg_flags & MSG_OOB)
543 		return -EOPNOTSUPP;
544 
545 	/* Get and verify the address */
546 	memset(&fl6, 0, sizeof(fl6));
547 
548 	fl6.flowi6_mark = READ_ONCE(sk->sk_mark);
549 	fl6.flowi6_uid = sk_uid(sk);
550 
551 	ipcm6_init_sk(&ipc6, sk);
552 
553 	if (lsa) {
554 		if (addr_len < SIN6_LEN_RFC2133)
555 			return -EINVAL;
556 
557 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
558 			return -EAFNOSUPPORT;
559 
560 		daddr = &lsa->l2tp_addr;
561 		if (inet6_test_bit(SNDFLOW, sk)) {
562 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
563 			if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
564 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
565 				if (IS_ERR(flowlabel))
566 					return -EINVAL;
567 			}
568 		}
569 
570 		/* Otherwise it will be difficult to maintain
571 		 * sk->sk_dst_cache.
572 		 */
573 		if (sk->sk_state == TCP_ESTABLISHED &&
574 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
575 			daddr = &sk->sk_v6_daddr;
576 
577 		if (addr_len >= sizeof(struct sockaddr_in6) &&
578 		    lsa->l2tp_scope_id &&
579 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
580 			fl6.flowi6_oif = lsa->l2tp_scope_id;
581 	} else {
582 		if (sk->sk_state != TCP_ESTABLISHED)
583 			return -EDESTADDRREQ;
584 
585 		daddr = &sk->sk_v6_daddr;
586 		fl6.flowlabel = np->flow_label;
587 	}
588 
589 	if (fl6.flowi6_oif == 0)
590 		fl6.flowi6_oif = READ_ONCE(sk->sk_bound_dev_if);
591 
592 	if (msg->msg_controllen) {
593 		opt = &opt_space;
594 		memset(opt, 0, sizeof(struct ipv6_txoptions));
595 		opt->tot_len = sizeof(struct ipv6_txoptions);
596 		ipc6.opt = opt;
597 
598 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
599 		if (err < 0) {
600 			fl6_sock_release(flowlabel);
601 			return err;
602 		}
603 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
604 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
605 			if (IS_ERR(flowlabel))
606 				return -EINVAL;
607 		}
608 		if (!(opt->opt_nflen | opt->opt_flen))
609 			opt = NULL;
610 	}
611 
612 	if (!opt) {
613 		opt = txopt_get(np);
614 		opt_to_free = opt;
615 	}
616 	if (flowlabel)
617 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
618 	opt = ipv6_fixup_options(&opt_space, opt);
619 	ipc6.opt = opt;
620 
621 	fl6.flowi6_proto = sk->sk_protocol;
622 	if (!ipv6_addr_any(daddr))
623 		fl6.daddr = *daddr;
624 	else
625 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
626 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
627 		fl6.saddr = np->saddr;
628 
629 	final_p = fl6_update_dst(&fl6, opt, &final);
630 
631 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
632 		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
633 	else if (!fl6.flowi6_oif)
634 		fl6.flowi6_oif = READ_ONCE(np->ucast_oif);
635 
636 	security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
637 
638 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
639 
640 	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
641 	if (IS_ERR(dst)) {
642 		err = PTR_ERR(dst);
643 		goto out;
644 	}
645 
646 	if (ipc6.hlimit < 0)
647 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
648 
649 	if (msg->msg_flags & MSG_CONFIRM)
650 		goto do_confirm;
651 
652 back_from_confirm:
653 	lock_sock(sk);
654 	ulen = len + (skb_queue_empty(&sk->sk_write_queue) ? transhdrlen : 0);
655 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
656 			      ulen, transhdrlen, &ipc6,
657 			      &fl6, dst_rt6_info(dst),
658 			      msg->msg_flags);
659 	if (err)
660 		ip6_flush_pending_frames(sk);
661 	else if (!(msg->msg_flags & MSG_MORE))
662 		err = l2tp_ip6_push_pending_frames(sk);
663 	release_sock(sk);
664 done:
665 	dst_release(dst);
666 out:
667 	fl6_sock_release(flowlabel);
668 	txopt_put(opt_to_free);
669 
670 	return err < 0 ? err : len;
671 
672 do_confirm:
673 	if (msg->msg_flags & MSG_PROBE)
674 		dst_confirm_neigh(dst, &fl6.daddr);
675 	if (!(msg->msg_flags & MSG_PROBE) || len)
676 		goto back_from_confirm;
677 	err = 0;
678 	goto done;
679 }
680 
681 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
682 			    int flags, int *addr_len)
683 {
684 	struct ipv6_pinfo *np = inet6_sk(sk);
685 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
686 	size_t copied = 0;
687 	int err = -EOPNOTSUPP;
688 	struct sk_buff *skb;
689 
690 	if (flags & MSG_OOB)
691 		goto out;
692 
693 	if (flags & MSG_ERRQUEUE)
694 		return ipv6_recv_error(sk, msg, len, addr_len);
695 
696 	skb = skb_recv_datagram(sk, flags, &err);
697 	if (!skb)
698 		goto out;
699 
700 	copied = skb->len;
701 	if (len < copied) {
702 		msg->msg_flags |= MSG_TRUNC;
703 		copied = len;
704 	}
705 
706 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
707 	if (err)
708 		goto done;
709 
710 	sock_recv_timestamp(msg, sk, skb);
711 
712 	/* Copy the address. */
713 	if (lsa) {
714 		lsa->l2tp_family = AF_INET6;
715 		lsa->l2tp_unused = 0;
716 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
717 		lsa->l2tp_flowinfo = 0;
718 		lsa->l2tp_scope_id = 0;
719 		lsa->l2tp_conn_id = 0;
720 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
721 			lsa->l2tp_scope_id = inet6_iif(skb);
722 		*addr_len = sizeof(*lsa);
723 	}
724 
725 	if (np->rxopt.all)
726 		ip6_datagram_recv_ctl(sk, msg, skb);
727 
728 	if (flags & MSG_TRUNC)
729 		copied = skb->len;
730 done:
731 	skb_free_datagram(sk, skb);
732 out:
733 	return err ? err : copied;
734 }
735 
736 static struct proto l2tp_ip6_prot = {
737 	.name		   = "L2TP/IPv6",
738 	.owner		   = THIS_MODULE,
739 	.init		   = l2tp_ip6_open,
740 	.close		   = l2tp_ip6_close,
741 	.bind		   = l2tp_ip6_bind,
742 	.connect	   = l2tp_ip6_connect,
743 	.disconnect	   = l2tp_ip6_disconnect,
744 	.ioctl		   = l2tp_ioctl,
745 	.destroy	   = l2tp_ip6_destroy_sock,
746 	.setsockopt	   = ipv6_setsockopt,
747 	.getsockopt	   = ipv6_getsockopt,
748 	.sendmsg	   = l2tp_ip6_sendmsg,
749 	.recvmsg	   = l2tp_ip6_recvmsg,
750 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
751 	.hash		   = l2tp_ip6_hash,
752 	.unhash		   = l2tp_ip6_unhash,
753 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
754 	.ipv6_pinfo_offset = offsetof(struct l2tp_ip6_sock, inet6),
755 };
756 
757 static const struct proto_ops l2tp_ip6_ops = {
758 	.family		   = PF_INET6,
759 	.owner		   = THIS_MODULE,
760 	.release	   = inet6_release,
761 	.bind		   = inet6_bind,
762 	.connect	   = inet_dgram_connect,
763 	.socketpair	   = sock_no_socketpair,
764 	.accept		   = sock_no_accept,
765 	.getname	   = l2tp_ip6_getname,
766 	.poll		   = datagram_poll,
767 	.ioctl		   = inet6_ioctl,
768 	.gettstamp	   = sock_gettstamp,
769 	.listen		   = sock_no_listen,
770 	.shutdown	   = inet_shutdown,
771 	.setsockopt	   = sock_common_setsockopt,
772 	.getsockopt	   = sock_common_getsockopt,
773 	.sendmsg	   = inet_sendmsg,
774 	.recvmsg	   = sock_common_recvmsg,
775 	.mmap		   = sock_no_mmap,
776 #ifdef CONFIG_COMPAT
777 	.compat_ioctl	   = inet6_compat_ioctl,
778 #endif
779 };
780 
781 static struct inet_protosw l2tp_ip6_protosw = {
782 	.type		= SOCK_DGRAM,
783 	.protocol	= IPPROTO_L2TP,
784 	.prot		= &l2tp_ip6_prot,
785 	.ops		= &l2tp_ip6_ops,
786 };
787 
788 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
789 	.handler	= l2tp_ip6_recv,
790 };
791 
792 static __net_init int l2tp_ip6_init_net(struct net *net)
793 {
794 	struct l2tp_ip6_net *pn = net_generic(net, l2tp_ip6_net_id);
795 
796 	rwlock_init(&pn->l2tp_ip6_lock);
797 	INIT_HLIST_HEAD(&pn->l2tp_ip6_table);
798 	INIT_HLIST_HEAD(&pn->l2tp_ip6_bind_table);
799 	return 0;
800 }
801 
802 static __net_exit void l2tp_ip6_exit_net(struct net *net)
803 {
804 	struct l2tp_ip6_net *pn = l2tp_ip6_pernet(net);
805 
806 	write_lock_bh(&pn->l2tp_ip6_lock);
807 	WARN_ON_ONCE(hlist_count_nodes(&pn->l2tp_ip6_table) != 0);
808 	WARN_ON_ONCE(hlist_count_nodes(&pn->l2tp_ip6_bind_table) != 0);
809 	write_unlock_bh(&pn->l2tp_ip6_lock);
810 }
811 
812 static struct pernet_operations l2tp_ip6_net_ops = {
813 	.init = l2tp_ip6_init_net,
814 	.exit = l2tp_ip6_exit_net,
815 	.id   = &l2tp_ip6_net_id,
816 	.size = sizeof(struct l2tp_ip6_net),
817 };
818 
819 static int __init l2tp_ip6_init(void)
820 {
821 	int err;
822 
823 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
824 
825 	err = register_pernet_device(&l2tp_ip6_net_ops);
826 	if (err)
827 		goto out;
828 
829 	err = proto_register(&l2tp_ip6_prot, 1);
830 	if (err != 0)
831 		goto out1;
832 
833 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
834 	if (err)
835 		goto out2;
836 
837 	inet6_register_protosw(&l2tp_ip6_protosw);
838 	return 0;
839 
840 out2:
841 	proto_unregister(&l2tp_ip6_prot);
842 out1:
843 	unregister_pernet_device(&l2tp_ip6_net_ops);
844 out:
845 	return err;
846 }
847 
848 static void __exit l2tp_ip6_exit(void)
849 {
850 	inet6_unregister_protosw(&l2tp_ip6_protosw);
851 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
852 	proto_unregister(&l2tp_ip6_prot);
853 	unregister_pernet_device(&l2tp_ip6_net_ops);
854 }
855 
856 module_init(l2tp_ip6_init);
857 module_exit(l2tp_ip6_exit);
858 
859 MODULE_LICENSE("GPL");
860 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
861 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
862 MODULE_VERSION("1.0");
863 
864 /* Use the values of SOCK_DGRAM (2) as type and IPPROTO_L2TP (115) as protocol,
865  * because __stringify doesn't like enums
866  */
867 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 115, 2);
868 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, 115);
869