xref: /openbmc/linux/net/l2tp/l2tp_ip6.c (revision 4e1a33b1)
1 /*
2  * L2TPv3 IP encapsulation support for IPv6
3  *
4  * Copyright (c) 2012 Katalix Systems Ltd
5  *
6  *	This program is free software; you can redistribute it and/or
7  *	modify it under the terms of the GNU General Public License
8  *	as published by the Free Software Foundation; either version
9  *	2 of the License, or (at your option) any later version.
10  */
11 
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 
14 #include <linux/icmp.h>
15 #include <linux/module.h>
16 #include <linux/skbuff.h>
17 #include <linux/random.h>
18 #include <linux/socket.h>
19 #include <linux/l2tp.h>
20 #include <linux/in.h>
21 #include <linux/in6.h>
22 #include <net/sock.h>
23 #include <net/ip.h>
24 #include <net/icmp.h>
25 #include <net/udp.h>
26 #include <net/inet_common.h>
27 #include <net/inet_hashtables.h>
28 #include <net/inet6_hashtables.h>
29 #include <net/tcp_states.h>
30 #include <net/protocol.h>
31 #include <net/xfrm.h>
32 
33 #include <net/transp_v6.h>
34 #include <net/addrconf.h>
35 #include <net/ip6_route.h>
36 
37 #include "l2tp_core.h"
38 
39 struct l2tp_ip6_sock {
40 	/* inet_sock has to be the first member of l2tp_ip6_sock */
41 	struct inet_sock	inet;
42 
43 	u32			conn_id;
44 	u32			peer_conn_id;
45 
46 	/* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
47 	   inet6_sk_generic */
48 	struct ipv6_pinfo	inet6;
49 };
50 
51 static DEFINE_RWLOCK(l2tp_ip6_lock);
52 static struct hlist_head l2tp_ip6_table;
53 static struct hlist_head l2tp_ip6_bind_table;
54 
55 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
56 {
57 	return (struct l2tp_ip6_sock *)sk;
58 }
59 
60 static struct sock *__l2tp_ip6_bind_lookup(const struct net *net,
61 					   const struct in6_addr *laddr,
62 					   const struct in6_addr *raddr,
63 					   int dif, u32 tunnel_id)
64 {
65 	struct sock *sk;
66 
67 	sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
68 		const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
69 		const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
70 		const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
71 
72 		if (!net_eq(sock_net(sk), net))
73 			continue;
74 
75 		if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
76 			continue;
77 
78 		if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
79 		    !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
80 			continue;
81 
82 		if (!ipv6_addr_any(sk_raddr) && raddr &&
83 		    !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
84 			continue;
85 
86 		if (l2tp->conn_id != tunnel_id)
87 			continue;
88 
89 		goto found;
90 	}
91 
92 	sk = NULL;
93 found:
94 	return sk;
95 }
96 
97 /* When processing receive frames, there are two cases to
98  * consider. Data frames consist of a non-zero session-id and an
99  * optional cookie. Control frames consist of a regular L2TP header
100  * preceded by 32-bits of zeros.
101  *
102  * L2TPv3 Session Header Over IP
103  *
104  *  0                   1                   2                   3
105  *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
106  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107  * |                           Session ID                          |
108  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109  * |               Cookie (optional, maximum 64 bits)...
110  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111  *                                                                 |
112  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113  *
114  * L2TPv3 Control Message Header Over IP
115  *
116  *  0                   1                   2                   3
117  *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
118  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119  * |                      (32 bits of zeros)                       |
120  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121  * |T|L|x|x|S|x|x|x|x|x|x|x|  Ver  |             Length            |
122  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
123  * |                     Control Connection ID                     |
124  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
125  * |               Ns              |               Nr              |
126  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
127  *
128  * All control frames are passed to userspace.
129  */
130 static int l2tp_ip6_recv(struct sk_buff *skb)
131 {
132 	struct net *net = dev_net(skb->dev);
133 	struct sock *sk;
134 	u32 session_id;
135 	u32 tunnel_id;
136 	unsigned char *ptr, *optr;
137 	struct l2tp_session *session;
138 	struct l2tp_tunnel *tunnel = NULL;
139 	int length;
140 
141 	if (!pskb_may_pull(skb, 4))
142 		goto discard;
143 
144 	/* Point to L2TP header */
145 	optr = ptr = skb->data;
146 	session_id = ntohl(*((__be32 *) ptr));
147 	ptr += 4;
148 
149 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
150 	 * the session_id. If it is 0, the packet is a L2TP control
151 	 * frame and the session_id value can be discarded.
152 	 */
153 	if (session_id == 0) {
154 		__skb_pull(skb, 4);
155 		goto pass_up;
156 	}
157 
158 	/* Ok, this is a data packet. Lookup the session. */
159 	session = l2tp_session_find(net, NULL, session_id);
160 	if (session == NULL)
161 		goto discard;
162 
163 	tunnel = session->tunnel;
164 	if (tunnel == NULL)
165 		goto discard;
166 
167 	/* Trace packet contents, if enabled */
168 	if (tunnel->debug & L2TP_MSG_DATA) {
169 		length = min(32u, skb->len);
170 		if (!pskb_may_pull(skb, length))
171 			goto discard;
172 
173 		/* Point to L2TP header */
174 		optr = ptr = skb->data;
175 		ptr += 4;
176 		pr_debug("%s: ip recv\n", tunnel->name);
177 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
178 	}
179 
180 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len,
181 			 tunnel->recv_payload_hook);
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 	tunnel = l2tp_tunnel_find(net, tunnel_id);
194 	if (tunnel != NULL)
195 		sk = tunnel->sock;
196 	else {
197 		struct ipv6hdr *iph = ipv6_hdr(skb);
198 
199 		read_lock_bh(&l2tp_ip6_lock);
200 		sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
201 					    inet6_iif(skb), tunnel_id);
202 		if (!sk) {
203 			read_unlock_bh(&l2tp_ip6_lock);
204 			goto discard;
205 		}
206 
207 		sock_hold(sk);
208 		read_unlock_bh(&l2tp_ip6_lock);
209 	}
210 
211 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
212 		goto discard_put;
213 
214 	nf_reset(skb);
215 
216 	return sk_receive_skb(sk, skb, 1);
217 
218 discard_put:
219 	sock_put(sk);
220 
221 discard:
222 	kfree_skb(skb);
223 	return 0;
224 }
225 
226 static int l2tp_ip6_open(struct sock *sk)
227 {
228 	/* Prevent autobind. We don't have ports. */
229 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
230 
231 	write_lock_bh(&l2tp_ip6_lock);
232 	sk_add_node(sk, &l2tp_ip6_table);
233 	write_unlock_bh(&l2tp_ip6_lock);
234 
235 	return 0;
236 }
237 
238 static void l2tp_ip6_close(struct sock *sk, long timeout)
239 {
240 	write_lock_bh(&l2tp_ip6_lock);
241 	hlist_del_init(&sk->sk_bind_node);
242 	sk_del_node_init(sk);
243 	write_unlock_bh(&l2tp_ip6_lock);
244 
245 	sk_common_release(sk);
246 }
247 
248 static void l2tp_ip6_destroy_sock(struct sock *sk)
249 {
250 	struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
251 
252 	lock_sock(sk);
253 	ip6_flush_pending_frames(sk);
254 	release_sock(sk);
255 
256 	if (tunnel) {
257 		l2tp_tunnel_closeall(tunnel);
258 		sock_put(sk);
259 	}
260 
261 	inet6_destroy_sock(sk);
262 }
263 
264 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
265 {
266 	struct inet_sock *inet = inet_sk(sk);
267 	struct ipv6_pinfo *np = inet6_sk(sk);
268 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
269 	struct net *net = sock_net(sk);
270 	__be32 v4addr = 0;
271 	int bound_dev_if;
272 	int addr_type;
273 	int err;
274 
275 	if (addr->l2tp_family != AF_INET6)
276 		return -EINVAL;
277 	if (addr_len < sizeof(*addr))
278 		return -EINVAL;
279 
280 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
281 
282 	/* l2tp_ip6 sockets are IPv6 only */
283 	if (addr_type == IPV6_ADDR_MAPPED)
284 		return -EADDRNOTAVAIL;
285 
286 	/* L2TP is point-point, not multicast */
287 	if (addr_type & IPV6_ADDR_MULTICAST)
288 		return -EADDRNOTAVAIL;
289 
290 	lock_sock(sk);
291 
292 	err = -EINVAL;
293 	if (!sock_flag(sk, SOCK_ZAPPED))
294 		goto out_unlock;
295 
296 	if (sk->sk_state != TCP_CLOSE)
297 		goto out_unlock;
298 
299 	bound_dev_if = sk->sk_bound_dev_if;
300 
301 	/* Check if the address belongs to the host. */
302 	rcu_read_lock();
303 	if (addr_type != IPV6_ADDR_ANY) {
304 		struct net_device *dev = NULL;
305 
306 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
307 			if (addr->l2tp_scope_id)
308 				bound_dev_if = addr->l2tp_scope_id;
309 
310 			/* Binding to link-local address requires an
311 			 * interface.
312 			 */
313 			if (!bound_dev_if)
314 				goto out_unlock_rcu;
315 
316 			err = -ENODEV;
317 			dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
318 			if (!dev)
319 				goto out_unlock_rcu;
320 		}
321 
322 		/* ipv4 addr of the socket is invalid.  Only the
323 		 * unspecified and mapped address have a v4 equivalent.
324 		 */
325 		v4addr = LOOPBACK4_IPV6;
326 		err = -EADDRNOTAVAIL;
327 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
328 			goto out_unlock_rcu;
329 	}
330 	rcu_read_unlock();
331 
332 	write_lock_bh(&l2tp_ip6_lock);
333 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
334 				   addr->l2tp_conn_id)) {
335 		write_unlock_bh(&l2tp_ip6_lock);
336 		err = -EADDRINUSE;
337 		goto out_unlock;
338 	}
339 
340 	inet->inet_saddr = v4addr;
341 	inet->inet_rcv_saddr = v4addr;
342 	sk->sk_bound_dev_if = bound_dev_if;
343 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
344 	np->saddr = addr->l2tp_addr;
345 
346 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
347 
348 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
349 	sk_del_node_init(sk);
350 	write_unlock_bh(&l2tp_ip6_lock);
351 
352 	sock_reset_flag(sk, SOCK_ZAPPED);
353 	release_sock(sk);
354 	return 0;
355 
356 out_unlock_rcu:
357 	rcu_read_unlock();
358 out_unlock:
359 	release_sock(sk);
360 
361 	return err;
362 }
363 
364 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
365 			    int addr_len)
366 {
367 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
368 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
369 	struct in6_addr	*daddr;
370 	int	addr_type;
371 	int rc;
372 
373 	if (addr_len < sizeof(*lsa))
374 		return -EINVAL;
375 
376 	if (usin->sin6_family != AF_INET6)
377 		return -EINVAL;
378 
379 	addr_type = ipv6_addr_type(&usin->sin6_addr);
380 	if (addr_type & IPV6_ADDR_MULTICAST)
381 		return -EINVAL;
382 
383 	if (addr_type & IPV6_ADDR_MAPPED) {
384 		daddr = &usin->sin6_addr;
385 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
386 			return -EINVAL;
387 	}
388 
389 	lock_sock(sk);
390 
391 	 /* Must bind first - autobinding does not work */
392 	if (sock_flag(sk, SOCK_ZAPPED)) {
393 		rc = -EINVAL;
394 		goto out_sk;
395 	}
396 
397 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
398 	if (rc < 0)
399 		goto out_sk;
400 
401 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
402 
403 	write_lock_bh(&l2tp_ip6_lock);
404 	hlist_del_init(&sk->sk_bind_node);
405 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
406 	write_unlock_bh(&l2tp_ip6_lock);
407 
408 out_sk:
409 	release_sock(sk);
410 
411 	return rc;
412 }
413 
414 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
415 {
416 	if (sock_flag(sk, SOCK_ZAPPED))
417 		return 0;
418 
419 	return __udp_disconnect(sk, flags);
420 }
421 
422 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
423 			    int *uaddr_len, int peer)
424 {
425 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
426 	struct sock *sk = sock->sk;
427 	struct ipv6_pinfo *np = inet6_sk(sk);
428 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
429 
430 	lsa->l2tp_family = AF_INET6;
431 	lsa->l2tp_flowinfo = 0;
432 	lsa->l2tp_scope_id = 0;
433 	lsa->l2tp_unused = 0;
434 	if (peer) {
435 		if (!lsk->peer_conn_id)
436 			return -ENOTCONN;
437 		lsa->l2tp_conn_id = lsk->peer_conn_id;
438 		lsa->l2tp_addr = sk->sk_v6_daddr;
439 		if (np->sndflow)
440 			lsa->l2tp_flowinfo = np->flow_label;
441 	} else {
442 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
443 			lsa->l2tp_addr = np->saddr;
444 		else
445 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
446 
447 		lsa->l2tp_conn_id = lsk->conn_id;
448 	}
449 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
450 		lsa->l2tp_scope_id = sk->sk_bound_dev_if;
451 	*uaddr_len = sizeof(*lsa);
452 	return 0;
453 }
454 
455 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
456 {
457 	int rc;
458 
459 	/* Charge it to the socket, dropping if the queue is full. */
460 	rc = sock_queue_rcv_skb(sk, skb);
461 	if (rc < 0)
462 		goto drop;
463 
464 	return 0;
465 
466 drop:
467 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
468 	kfree_skb(skb);
469 	return -1;
470 }
471 
472 static int l2tp_ip6_push_pending_frames(struct sock *sk)
473 {
474 	struct sk_buff *skb;
475 	__be32 *transhdr = NULL;
476 	int err = 0;
477 
478 	skb = skb_peek(&sk->sk_write_queue);
479 	if (skb == NULL)
480 		goto out;
481 
482 	transhdr = (__be32 *)skb_transport_header(skb);
483 	*transhdr = 0;
484 
485 	err = ip6_push_pending_frames(sk);
486 
487 out:
488 	return err;
489 }
490 
491 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
492  * control frames.
493  */
494 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
495 {
496 	struct ipv6_txoptions opt_space;
497 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
498 	struct in6_addr *daddr, *final_p, final;
499 	struct ipv6_pinfo *np = inet6_sk(sk);
500 	struct ipv6_txoptions *opt_to_free = NULL;
501 	struct ipv6_txoptions *opt = NULL;
502 	struct ip6_flowlabel *flowlabel = NULL;
503 	struct dst_entry *dst = NULL;
504 	struct flowi6 fl6;
505 	struct sockcm_cookie sockc_unused = {0};
506 	struct ipcm6_cookie ipc6;
507 	int addr_len = msg->msg_namelen;
508 	int transhdrlen = 4; /* zero session-id */
509 	int ulen = len + transhdrlen;
510 	int err;
511 
512 	/* Rough check on arithmetic overflow,
513 	   better check is made in ip6_append_data().
514 	 */
515 	if (len > INT_MAX)
516 		return -EMSGSIZE;
517 
518 	/* Mirror BSD error message compatibility */
519 	if (msg->msg_flags & MSG_OOB)
520 		return -EOPNOTSUPP;
521 
522 	/*
523 	 *	Get and verify the address.
524 	 */
525 	memset(&fl6, 0, sizeof(fl6));
526 
527 	fl6.flowi6_mark = sk->sk_mark;
528 	fl6.flowi6_uid = sk->sk_uid;
529 
530 	ipc6.hlimit = -1;
531 	ipc6.tclass = -1;
532 	ipc6.dontfrag = -1;
533 
534 	if (lsa) {
535 		if (addr_len < SIN6_LEN_RFC2133)
536 			return -EINVAL;
537 
538 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
539 			return -EAFNOSUPPORT;
540 
541 		daddr = &lsa->l2tp_addr;
542 		if (np->sndflow) {
543 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
544 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
545 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
546 				if (flowlabel == NULL)
547 					return -EINVAL;
548 			}
549 		}
550 
551 		/*
552 		 * Otherwise it will be difficult to maintain
553 		 * sk->sk_dst_cache.
554 		 */
555 		if (sk->sk_state == TCP_ESTABLISHED &&
556 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
557 			daddr = &sk->sk_v6_daddr;
558 
559 		if (addr_len >= sizeof(struct sockaddr_in6) &&
560 		    lsa->l2tp_scope_id &&
561 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
562 			fl6.flowi6_oif = lsa->l2tp_scope_id;
563 	} else {
564 		if (sk->sk_state != TCP_ESTABLISHED)
565 			return -EDESTADDRREQ;
566 
567 		daddr = &sk->sk_v6_daddr;
568 		fl6.flowlabel = np->flow_label;
569 	}
570 
571 	if (fl6.flowi6_oif == 0)
572 		fl6.flowi6_oif = sk->sk_bound_dev_if;
573 
574 	if (msg->msg_controllen) {
575 		opt = &opt_space;
576 		memset(opt, 0, sizeof(struct ipv6_txoptions));
577 		opt->tot_len = sizeof(struct ipv6_txoptions);
578 		ipc6.opt = opt;
579 
580 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6,
581 					    &sockc_unused);
582 		if (err < 0) {
583 			fl6_sock_release(flowlabel);
584 			return err;
585 		}
586 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
587 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
588 			if (flowlabel == NULL)
589 				return -EINVAL;
590 		}
591 		if (!(opt->opt_nflen|opt->opt_flen))
592 			opt = NULL;
593 	}
594 
595 	if (!opt) {
596 		opt = txopt_get(np);
597 		opt_to_free = opt;
598 	}
599 	if (flowlabel)
600 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
601 	opt = ipv6_fixup_options(&opt_space, opt);
602 	ipc6.opt = opt;
603 
604 	fl6.flowi6_proto = sk->sk_protocol;
605 	if (!ipv6_addr_any(daddr))
606 		fl6.daddr = *daddr;
607 	else
608 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
609 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
610 		fl6.saddr = np->saddr;
611 
612 	final_p = fl6_update_dst(&fl6, opt, &final);
613 
614 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
615 		fl6.flowi6_oif = np->mcast_oif;
616 	else if (!fl6.flowi6_oif)
617 		fl6.flowi6_oif = np->ucast_oif;
618 
619 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
620 
621 	if (ipc6.tclass < 0)
622 		ipc6.tclass = np->tclass;
623 
624 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
625 
626 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
627 	if (IS_ERR(dst)) {
628 		err = PTR_ERR(dst);
629 		goto out;
630 	}
631 
632 	if (ipc6.hlimit < 0)
633 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
634 
635 	if (ipc6.dontfrag < 0)
636 		ipc6.dontfrag = np->dontfrag;
637 
638 	if (msg->msg_flags & MSG_CONFIRM)
639 		goto do_confirm;
640 
641 back_from_confirm:
642 	lock_sock(sk);
643 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
644 			      ulen, transhdrlen, &ipc6,
645 			      &fl6, (struct rt6_info *)dst,
646 			      msg->msg_flags, &sockc_unused);
647 	if (err)
648 		ip6_flush_pending_frames(sk);
649 	else if (!(msg->msg_flags & MSG_MORE))
650 		err = l2tp_ip6_push_pending_frames(sk);
651 	release_sock(sk);
652 done:
653 	dst_release(dst);
654 out:
655 	fl6_sock_release(flowlabel);
656 	txopt_put(opt_to_free);
657 
658 	return err < 0 ? err : len;
659 
660 do_confirm:
661 	if (msg->msg_flags & MSG_PROBE)
662 		dst_confirm_neigh(dst, &fl6.daddr);
663 	if (!(msg->msg_flags & MSG_PROBE) || len)
664 		goto back_from_confirm;
665 	err = 0;
666 	goto done;
667 }
668 
669 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
670 			    int noblock, int flags, int *addr_len)
671 {
672 	struct ipv6_pinfo *np = inet6_sk(sk);
673 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
674 	size_t copied = 0;
675 	int err = -EOPNOTSUPP;
676 	struct sk_buff *skb;
677 
678 	if (flags & MSG_OOB)
679 		goto out;
680 
681 	if (addr_len)
682 		*addr_len = sizeof(*lsa);
683 
684 	if (flags & MSG_ERRQUEUE)
685 		return ipv6_recv_error(sk, msg, len, addr_len);
686 
687 	skb = skb_recv_datagram(sk, flags, noblock, &err);
688 	if (!skb)
689 		goto out;
690 
691 	copied = skb->len;
692 	if (len < copied) {
693 		msg->msg_flags |= MSG_TRUNC;
694 		copied = len;
695 	}
696 
697 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
698 	if (err)
699 		goto done;
700 
701 	sock_recv_timestamp(msg, sk, skb);
702 
703 	/* Copy the address. */
704 	if (lsa) {
705 		lsa->l2tp_family = AF_INET6;
706 		lsa->l2tp_unused = 0;
707 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
708 		lsa->l2tp_flowinfo = 0;
709 		lsa->l2tp_scope_id = 0;
710 		lsa->l2tp_conn_id = 0;
711 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
712 			lsa->l2tp_scope_id = inet6_iif(skb);
713 	}
714 
715 	if (np->rxopt.all)
716 		ip6_datagram_recv_ctl(sk, msg, skb);
717 
718 	if (flags & MSG_TRUNC)
719 		copied = skb->len;
720 done:
721 	skb_free_datagram(sk, skb);
722 out:
723 	return err ? err : copied;
724 }
725 
726 static struct proto l2tp_ip6_prot = {
727 	.name		   = "L2TP/IPv6",
728 	.owner		   = THIS_MODULE,
729 	.init		   = l2tp_ip6_open,
730 	.close		   = l2tp_ip6_close,
731 	.bind		   = l2tp_ip6_bind,
732 	.connect	   = l2tp_ip6_connect,
733 	.disconnect	   = l2tp_ip6_disconnect,
734 	.ioctl		   = l2tp_ioctl,
735 	.destroy	   = l2tp_ip6_destroy_sock,
736 	.setsockopt	   = ipv6_setsockopt,
737 	.getsockopt	   = ipv6_getsockopt,
738 	.sendmsg	   = l2tp_ip6_sendmsg,
739 	.recvmsg	   = l2tp_ip6_recvmsg,
740 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
741 	.hash		   = inet6_hash,
742 	.unhash		   = inet_unhash,
743 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
744 #ifdef CONFIG_COMPAT
745 	.compat_setsockopt = compat_ipv6_setsockopt,
746 	.compat_getsockopt = compat_ipv6_getsockopt,
747 #endif
748 };
749 
750 static const struct proto_ops l2tp_ip6_ops = {
751 	.family		   = PF_INET6,
752 	.owner		   = THIS_MODULE,
753 	.release	   = inet6_release,
754 	.bind		   = inet6_bind,
755 	.connect	   = inet_dgram_connect,
756 	.socketpair	   = sock_no_socketpair,
757 	.accept		   = sock_no_accept,
758 	.getname	   = l2tp_ip6_getname,
759 	.poll		   = datagram_poll,
760 	.ioctl		   = inet6_ioctl,
761 	.listen		   = sock_no_listen,
762 	.shutdown	   = inet_shutdown,
763 	.setsockopt	   = sock_common_setsockopt,
764 	.getsockopt	   = sock_common_getsockopt,
765 	.sendmsg	   = inet_sendmsg,
766 	.recvmsg	   = sock_common_recvmsg,
767 	.mmap		   = sock_no_mmap,
768 	.sendpage	   = sock_no_sendpage,
769 #ifdef CONFIG_COMPAT
770 	.compat_setsockopt = compat_sock_common_setsockopt,
771 	.compat_getsockopt = compat_sock_common_getsockopt,
772 #endif
773 };
774 
775 static struct inet_protosw l2tp_ip6_protosw = {
776 	.type		= SOCK_DGRAM,
777 	.protocol	= IPPROTO_L2TP,
778 	.prot		= &l2tp_ip6_prot,
779 	.ops		= &l2tp_ip6_ops,
780 };
781 
782 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
783 	.handler	= l2tp_ip6_recv,
784 };
785 
786 static int __init l2tp_ip6_init(void)
787 {
788 	int err;
789 
790 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
791 
792 	err = proto_register(&l2tp_ip6_prot, 1);
793 	if (err != 0)
794 		goto out;
795 
796 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
797 	if (err)
798 		goto out1;
799 
800 	inet6_register_protosw(&l2tp_ip6_protosw);
801 	return 0;
802 
803 out1:
804 	proto_unregister(&l2tp_ip6_prot);
805 out:
806 	return err;
807 }
808 
809 static void __exit l2tp_ip6_exit(void)
810 {
811 	inet6_unregister_protosw(&l2tp_ip6_protosw);
812 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
813 	proto_unregister(&l2tp_ip6_prot);
814 }
815 
816 module_init(l2tp_ip6_init);
817 module_exit(l2tp_ip6_exit);
818 
819 MODULE_LICENSE("GPL");
820 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
821 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
822 MODULE_VERSION("1.0");
823 
824 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
825  * enums
826  */
827 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
828 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);
829