xref: /openbmc/linux/net/l2tp/l2tp_ip6.c (revision 93f5715e)
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 	struct ipv6hdr *iph;
140 	int length;
141 
142 	if (!pskb_may_pull(skb, 4))
143 		goto discard;
144 
145 	/* Point to L2TP header */
146 	optr = ptr = skb->data;
147 	session_id = ntohl(*((__be32 *) ptr));
148 	ptr += 4;
149 
150 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
151 	 * the session_id. If it is 0, the packet is a L2TP control
152 	 * frame and the session_id value can be discarded.
153 	 */
154 	if (session_id == 0) {
155 		__skb_pull(skb, 4);
156 		goto pass_up;
157 	}
158 
159 	/* Ok, this is a data packet. Lookup the session. */
160 	session = l2tp_session_get(net, session_id);
161 	if (!session)
162 		goto discard;
163 
164 	tunnel = session->tunnel;
165 	if (!tunnel)
166 		goto discard_sess;
167 
168 	/* Trace packet contents, if enabled */
169 	if (tunnel->debug & L2TP_MSG_DATA) {
170 		length = min(32u, skb->len);
171 		if (!pskb_may_pull(skb, length))
172 			goto discard_sess;
173 
174 		/* Point to L2TP header */
175 		optr = ptr = skb->data;
176 		ptr += 4;
177 		pr_debug("%s: ip recv\n", tunnel->name);
178 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
179 	}
180 
181 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
182 	l2tp_session_dec_refcount(session);
183 
184 	return 0;
185 
186 pass_up:
187 	/* Get the tunnel_id from the L2TP header */
188 	if (!pskb_may_pull(skb, 12))
189 		goto discard;
190 
191 	if ((skb->data[0] & 0xc0) != 0xc0)
192 		goto discard;
193 
194 	tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
195 	iph = ipv6_hdr(skb);
196 
197 	read_lock_bh(&l2tp_ip6_lock);
198 	sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
199 				    inet6_iif(skb), tunnel_id);
200 	if (!sk) {
201 		read_unlock_bh(&l2tp_ip6_lock);
202 		goto discard;
203 	}
204 	sock_hold(sk);
205 	read_unlock_bh(&l2tp_ip6_lock);
206 
207 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
208 		goto discard_put;
209 
210 	nf_reset(skb);
211 
212 	return sk_receive_skb(sk, skb, 1);
213 
214 discard_sess:
215 	l2tp_session_dec_refcount(session);
216 	goto discard;
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 = sk->sk_user_data;
251 
252 	lock_sock(sk);
253 	ip6_flush_pending_frames(sk);
254 	release_sock(sk);
255 
256 	if (tunnel)
257 		l2tp_tunnel_delete(tunnel);
258 
259 	inet6_destroy_sock(sk);
260 }
261 
262 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
263 {
264 	struct inet_sock *inet = inet_sk(sk);
265 	struct ipv6_pinfo *np = inet6_sk(sk);
266 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
267 	struct net *net = sock_net(sk);
268 	__be32 v4addr = 0;
269 	int bound_dev_if;
270 	int addr_type;
271 	int err;
272 
273 	if (addr->l2tp_family != AF_INET6)
274 		return -EINVAL;
275 	if (addr_len < sizeof(*addr))
276 		return -EINVAL;
277 
278 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
279 
280 	/* l2tp_ip6 sockets are IPv6 only */
281 	if (addr_type == IPV6_ADDR_MAPPED)
282 		return -EADDRNOTAVAIL;
283 
284 	/* L2TP is point-point, not multicast */
285 	if (addr_type & IPV6_ADDR_MULTICAST)
286 		return -EADDRNOTAVAIL;
287 
288 	lock_sock(sk);
289 
290 	err = -EINVAL;
291 	if (!sock_flag(sk, SOCK_ZAPPED))
292 		goto out_unlock;
293 
294 	if (sk->sk_state != TCP_CLOSE)
295 		goto out_unlock;
296 
297 	bound_dev_if = sk->sk_bound_dev_if;
298 
299 	/* Check if the address belongs to the host. */
300 	rcu_read_lock();
301 	if (addr_type != IPV6_ADDR_ANY) {
302 		struct net_device *dev = NULL;
303 
304 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
305 			if (addr->l2tp_scope_id)
306 				bound_dev_if = addr->l2tp_scope_id;
307 
308 			/* Binding to link-local address requires an
309 			 * interface.
310 			 */
311 			if (!bound_dev_if)
312 				goto out_unlock_rcu;
313 
314 			err = -ENODEV;
315 			dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
316 			if (!dev)
317 				goto out_unlock_rcu;
318 		}
319 
320 		/* ipv4 addr of the socket is invalid.  Only the
321 		 * unspecified and mapped address have a v4 equivalent.
322 		 */
323 		v4addr = LOOPBACK4_IPV6;
324 		err = -EADDRNOTAVAIL;
325 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
326 			goto out_unlock_rcu;
327 	}
328 	rcu_read_unlock();
329 
330 	write_lock_bh(&l2tp_ip6_lock);
331 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
332 				   addr->l2tp_conn_id)) {
333 		write_unlock_bh(&l2tp_ip6_lock);
334 		err = -EADDRINUSE;
335 		goto out_unlock;
336 	}
337 
338 	inet->inet_saddr = v4addr;
339 	inet->inet_rcv_saddr = v4addr;
340 	sk->sk_bound_dev_if = bound_dev_if;
341 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
342 	np->saddr = addr->l2tp_addr;
343 
344 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
345 
346 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
347 	sk_del_node_init(sk);
348 	write_unlock_bh(&l2tp_ip6_lock);
349 
350 	sock_reset_flag(sk, SOCK_ZAPPED);
351 	release_sock(sk);
352 	return 0;
353 
354 out_unlock_rcu:
355 	rcu_read_unlock();
356 out_unlock:
357 	release_sock(sk);
358 
359 	return err;
360 }
361 
362 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
363 			    int addr_len)
364 {
365 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
366 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
367 	struct in6_addr	*daddr;
368 	int	addr_type;
369 	int rc;
370 
371 	if (addr_len < sizeof(*lsa))
372 		return -EINVAL;
373 
374 	if (usin->sin6_family != AF_INET6)
375 		return -EINVAL;
376 
377 	addr_type = ipv6_addr_type(&usin->sin6_addr);
378 	if (addr_type & IPV6_ADDR_MULTICAST)
379 		return -EINVAL;
380 
381 	if (addr_type & IPV6_ADDR_MAPPED) {
382 		daddr = &usin->sin6_addr;
383 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
384 			return -EINVAL;
385 	}
386 
387 	lock_sock(sk);
388 
389 	 /* Must bind first - autobinding does not work */
390 	if (sock_flag(sk, SOCK_ZAPPED)) {
391 		rc = -EINVAL;
392 		goto out_sk;
393 	}
394 
395 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
396 	if (rc < 0)
397 		goto out_sk;
398 
399 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
400 
401 	write_lock_bh(&l2tp_ip6_lock);
402 	hlist_del_init(&sk->sk_bind_node);
403 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
404 	write_unlock_bh(&l2tp_ip6_lock);
405 
406 out_sk:
407 	release_sock(sk);
408 
409 	return rc;
410 }
411 
412 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
413 {
414 	if (sock_flag(sk, SOCK_ZAPPED))
415 		return 0;
416 
417 	return __udp_disconnect(sk, flags);
418 }
419 
420 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
421 			    int peer)
422 {
423 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
424 	struct sock *sk = sock->sk;
425 	struct ipv6_pinfo *np = inet6_sk(sk);
426 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
427 
428 	lsa->l2tp_family = AF_INET6;
429 	lsa->l2tp_flowinfo = 0;
430 	lsa->l2tp_scope_id = 0;
431 	lsa->l2tp_unused = 0;
432 	if (peer) {
433 		if (!lsk->peer_conn_id)
434 			return -ENOTCONN;
435 		lsa->l2tp_conn_id = lsk->peer_conn_id;
436 		lsa->l2tp_addr = sk->sk_v6_daddr;
437 		if (np->sndflow)
438 			lsa->l2tp_flowinfo = np->flow_label;
439 	} else {
440 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
441 			lsa->l2tp_addr = np->saddr;
442 		else
443 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
444 
445 		lsa->l2tp_conn_id = lsk->conn_id;
446 	}
447 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
448 		lsa->l2tp_scope_id = sk->sk_bound_dev_if;
449 	return sizeof(*lsa);
450 }
451 
452 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
453 {
454 	int rc;
455 
456 	/* Charge it to the socket, dropping if the queue is full. */
457 	rc = sock_queue_rcv_skb(sk, skb);
458 	if (rc < 0)
459 		goto drop;
460 
461 	return 0;
462 
463 drop:
464 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
465 	kfree_skb(skb);
466 	return -1;
467 }
468 
469 static int l2tp_ip6_push_pending_frames(struct sock *sk)
470 {
471 	struct sk_buff *skb;
472 	__be32 *transhdr = NULL;
473 	int err = 0;
474 
475 	skb = skb_peek(&sk->sk_write_queue);
476 	if (skb == NULL)
477 		goto out;
478 
479 	transhdr = (__be32 *)skb_transport_header(skb);
480 	*transhdr = 0;
481 
482 	err = ip6_push_pending_frames(sk);
483 
484 out:
485 	return err;
486 }
487 
488 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
489  * control frames.
490  */
491 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
492 {
493 	struct ipv6_txoptions opt_space;
494 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
495 	struct in6_addr *daddr, *final_p, final;
496 	struct ipv6_pinfo *np = inet6_sk(sk);
497 	struct ipv6_txoptions *opt_to_free = NULL;
498 	struct ipv6_txoptions *opt = NULL;
499 	struct ip6_flowlabel *flowlabel = NULL;
500 	struct dst_entry *dst = NULL;
501 	struct flowi6 fl6;
502 	struct ipcm6_cookie ipc6;
503 	int addr_len = msg->msg_namelen;
504 	int transhdrlen = 4; /* zero session-id */
505 	int ulen = len + transhdrlen;
506 	int err;
507 
508 	/* Rough check on arithmetic overflow,
509 	   better check is made in ip6_append_data().
510 	 */
511 	if (len > INT_MAX)
512 		return -EMSGSIZE;
513 
514 	/* Mirror BSD error message compatibility */
515 	if (msg->msg_flags & MSG_OOB)
516 		return -EOPNOTSUPP;
517 
518 	/*
519 	 *	Get and verify the address.
520 	 */
521 	memset(&fl6, 0, sizeof(fl6));
522 
523 	fl6.flowi6_mark = sk->sk_mark;
524 	fl6.flowi6_uid = sk->sk_uid;
525 
526 	ipcm6_init(&ipc6);
527 
528 	if (lsa) {
529 		if (addr_len < SIN6_LEN_RFC2133)
530 			return -EINVAL;
531 
532 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
533 			return -EAFNOSUPPORT;
534 
535 		daddr = &lsa->l2tp_addr;
536 		if (np->sndflow) {
537 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
538 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
539 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
540 				if (flowlabel == NULL)
541 					return -EINVAL;
542 			}
543 		}
544 
545 		/*
546 		 * Otherwise it will be difficult to maintain
547 		 * sk->sk_dst_cache.
548 		 */
549 		if (sk->sk_state == TCP_ESTABLISHED &&
550 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
551 			daddr = &sk->sk_v6_daddr;
552 
553 		if (addr_len >= sizeof(struct sockaddr_in6) &&
554 		    lsa->l2tp_scope_id &&
555 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
556 			fl6.flowi6_oif = lsa->l2tp_scope_id;
557 	} else {
558 		if (sk->sk_state != TCP_ESTABLISHED)
559 			return -EDESTADDRREQ;
560 
561 		daddr = &sk->sk_v6_daddr;
562 		fl6.flowlabel = np->flow_label;
563 	}
564 
565 	if (fl6.flowi6_oif == 0)
566 		fl6.flowi6_oif = sk->sk_bound_dev_if;
567 
568 	if (msg->msg_controllen) {
569 		opt = &opt_space;
570 		memset(opt, 0, sizeof(struct ipv6_txoptions));
571 		opt->tot_len = sizeof(struct ipv6_txoptions);
572 		ipc6.opt = opt;
573 
574 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
575 		if (err < 0) {
576 			fl6_sock_release(flowlabel);
577 			return err;
578 		}
579 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
580 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
581 			if (flowlabel == NULL)
582 				return -EINVAL;
583 		}
584 		if (!(opt->opt_nflen|opt->opt_flen))
585 			opt = NULL;
586 	}
587 
588 	if (!opt) {
589 		opt = txopt_get(np);
590 		opt_to_free = opt;
591 	}
592 	if (flowlabel)
593 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
594 	opt = ipv6_fixup_options(&opt_space, opt);
595 	ipc6.opt = opt;
596 
597 	fl6.flowi6_proto = sk->sk_protocol;
598 	if (!ipv6_addr_any(daddr))
599 		fl6.daddr = *daddr;
600 	else
601 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
602 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
603 		fl6.saddr = np->saddr;
604 
605 	final_p = fl6_update_dst(&fl6, opt, &final);
606 
607 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
608 		fl6.flowi6_oif = np->mcast_oif;
609 	else if (!fl6.flowi6_oif)
610 		fl6.flowi6_oif = np->ucast_oif;
611 
612 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
613 
614 	if (ipc6.tclass < 0)
615 		ipc6.tclass = np->tclass;
616 
617 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
618 
619 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
620 	if (IS_ERR(dst)) {
621 		err = PTR_ERR(dst);
622 		goto out;
623 	}
624 
625 	if (ipc6.hlimit < 0)
626 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
627 
628 	if (ipc6.dontfrag < 0)
629 		ipc6.dontfrag = np->dontfrag;
630 
631 	if (msg->msg_flags & MSG_CONFIRM)
632 		goto do_confirm;
633 
634 back_from_confirm:
635 	lock_sock(sk);
636 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
637 			      ulen, transhdrlen, &ipc6,
638 			      &fl6, (struct rt6_info *)dst,
639 			      msg->msg_flags);
640 	if (err)
641 		ip6_flush_pending_frames(sk);
642 	else if (!(msg->msg_flags & MSG_MORE))
643 		err = l2tp_ip6_push_pending_frames(sk);
644 	release_sock(sk);
645 done:
646 	dst_release(dst);
647 out:
648 	fl6_sock_release(flowlabel);
649 	txopt_put(opt_to_free);
650 
651 	return err < 0 ? err : len;
652 
653 do_confirm:
654 	if (msg->msg_flags & MSG_PROBE)
655 		dst_confirm_neigh(dst, &fl6.daddr);
656 	if (!(msg->msg_flags & MSG_PROBE) || len)
657 		goto back_from_confirm;
658 	err = 0;
659 	goto done;
660 }
661 
662 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
663 			    int noblock, int flags, int *addr_len)
664 {
665 	struct ipv6_pinfo *np = inet6_sk(sk);
666 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
667 	size_t copied = 0;
668 	int err = -EOPNOTSUPP;
669 	struct sk_buff *skb;
670 
671 	if (flags & MSG_OOB)
672 		goto out;
673 
674 	if (addr_len)
675 		*addr_len = sizeof(*lsa);
676 
677 	if (flags & MSG_ERRQUEUE)
678 		return ipv6_recv_error(sk, msg, len, addr_len);
679 
680 	skb = skb_recv_datagram(sk, flags, noblock, &err);
681 	if (!skb)
682 		goto out;
683 
684 	copied = skb->len;
685 	if (len < copied) {
686 		msg->msg_flags |= MSG_TRUNC;
687 		copied = len;
688 	}
689 
690 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
691 	if (err)
692 		goto done;
693 
694 	sock_recv_timestamp(msg, sk, skb);
695 
696 	/* Copy the address. */
697 	if (lsa) {
698 		lsa->l2tp_family = AF_INET6;
699 		lsa->l2tp_unused = 0;
700 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
701 		lsa->l2tp_flowinfo = 0;
702 		lsa->l2tp_scope_id = 0;
703 		lsa->l2tp_conn_id = 0;
704 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
705 			lsa->l2tp_scope_id = inet6_iif(skb);
706 	}
707 
708 	if (np->rxopt.all)
709 		ip6_datagram_recv_ctl(sk, msg, skb);
710 
711 	if (flags & MSG_TRUNC)
712 		copied = skb->len;
713 done:
714 	skb_free_datagram(sk, skb);
715 out:
716 	return err ? err : copied;
717 }
718 
719 static struct proto l2tp_ip6_prot = {
720 	.name		   = "L2TP/IPv6",
721 	.owner		   = THIS_MODULE,
722 	.init		   = l2tp_ip6_open,
723 	.close		   = l2tp_ip6_close,
724 	.bind		   = l2tp_ip6_bind,
725 	.connect	   = l2tp_ip6_connect,
726 	.disconnect	   = l2tp_ip6_disconnect,
727 	.ioctl		   = l2tp_ioctl,
728 	.destroy	   = l2tp_ip6_destroy_sock,
729 	.setsockopt	   = ipv6_setsockopt,
730 	.getsockopt	   = ipv6_getsockopt,
731 	.sendmsg	   = l2tp_ip6_sendmsg,
732 	.recvmsg	   = l2tp_ip6_recvmsg,
733 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
734 	.hash		   = inet6_hash,
735 	.unhash		   = inet_unhash,
736 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
737 #ifdef CONFIG_COMPAT
738 	.compat_setsockopt = compat_ipv6_setsockopt,
739 	.compat_getsockopt = compat_ipv6_getsockopt,
740 #endif
741 };
742 
743 static const struct proto_ops l2tp_ip6_ops = {
744 	.family		   = PF_INET6,
745 	.owner		   = THIS_MODULE,
746 	.release	   = inet6_release,
747 	.bind		   = inet6_bind,
748 	.connect	   = inet_dgram_connect,
749 	.socketpair	   = sock_no_socketpair,
750 	.accept		   = sock_no_accept,
751 	.getname	   = l2tp_ip6_getname,
752 	.poll		   = datagram_poll,
753 	.ioctl		   = inet6_ioctl,
754 	.listen		   = sock_no_listen,
755 	.shutdown	   = inet_shutdown,
756 	.setsockopt	   = sock_common_setsockopt,
757 	.getsockopt	   = sock_common_getsockopt,
758 	.sendmsg	   = inet_sendmsg,
759 	.recvmsg	   = sock_common_recvmsg,
760 	.mmap		   = sock_no_mmap,
761 	.sendpage	   = sock_no_sendpage,
762 #ifdef CONFIG_COMPAT
763 	.compat_setsockopt = compat_sock_common_setsockopt,
764 	.compat_getsockopt = compat_sock_common_getsockopt,
765 #endif
766 };
767 
768 static struct inet_protosw l2tp_ip6_protosw = {
769 	.type		= SOCK_DGRAM,
770 	.protocol	= IPPROTO_L2TP,
771 	.prot		= &l2tp_ip6_prot,
772 	.ops		= &l2tp_ip6_ops,
773 };
774 
775 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
776 	.handler	= l2tp_ip6_recv,
777 };
778 
779 static int __init l2tp_ip6_init(void)
780 {
781 	int err;
782 
783 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
784 
785 	err = proto_register(&l2tp_ip6_prot, 1);
786 	if (err != 0)
787 		goto out;
788 
789 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
790 	if (err)
791 		goto out1;
792 
793 	inet6_register_protosw(&l2tp_ip6_protosw);
794 	return 0;
795 
796 out1:
797 	proto_unregister(&l2tp_ip6_prot);
798 out:
799 	return err;
800 }
801 
802 static void __exit l2tp_ip6_exit(void)
803 {
804 	inet6_unregister_protosw(&l2tp_ip6_protosw);
805 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
806 	proto_unregister(&l2tp_ip6_prot);
807 }
808 
809 module_init(l2tp_ip6_init);
810 module_exit(l2tp_ip6_exit);
811 
812 MODULE_LICENSE("GPL");
813 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
814 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
815 MODULE_VERSION("1.0");
816 
817 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
818  * enums
819  */
820 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
821 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);
822