xref: /openbmc/linux/net/l2tp/l2tp_ip6.c (revision 04dc82e1)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * L2TPv3 IP encapsulation support for IPv6
4  *
5  * Copyright (c) 2012 Katalix Systems Ltd
6  */
7 
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 
10 #include <linux/icmp.h>
11 #include <linux/module.h>
12 #include <linux/skbuff.h>
13 #include <linux/random.h>
14 #include <linux/socket.h>
15 #include <linux/l2tp.h>
16 #include <linux/in.h>
17 #include <linux/in6.h>
18 #include <net/sock.h>
19 #include <net/ip.h>
20 #include <net/icmp.h>
21 #include <net/udp.h>
22 #include <net/inet_common.h>
23 #include <net/inet_hashtables.h>
24 #include <net/inet6_hashtables.h>
25 #include <net/tcp_states.h>
26 #include <net/protocol.h>
27 #include <net/xfrm.h>
28 
29 #include <net/transp_v6.h>
30 #include <net/addrconf.h>
31 #include <net/ip6_route.h>
32 
33 #include "l2tp_core.h"
34 
35 struct l2tp_ip6_sock {
36 	/* inet_sock has to be the first member of l2tp_ip6_sock */
37 	struct inet_sock	inet;
38 
39 	u32			conn_id;
40 	u32			peer_conn_id;
41 
42 	/* ipv6_pinfo has to be the last member of l2tp_ip6_sock, see
43 	   inet6_sk_generic */
44 	struct ipv6_pinfo	inet6;
45 };
46 
47 static DEFINE_RWLOCK(l2tp_ip6_lock);
48 static struct hlist_head l2tp_ip6_table;
49 static struct hlist_head l2tp_ip6_bind_table;
50 
51 static inline struct l2tp_ip6_sock *l2tp_ip6_sk(const struct sock *sk)
52 {
53 	return (struct l2tp_ip6_sock *)sk;
54 }
55 
56 static struct sock *__l2tp_ip6_bind_lookup(const struct net *net,
57 					   const struct in6_addr *laddr,
58 					   const struct in6_addr *raddr,
59 					   int dif, u32 tunnel_id)
60 {
61 	struct sock *sk;
62 
63 	sk_for_each_bound(sk, &l2tp_ip6_bind_table) {
64 		const struct in6_addr *sk_laddr = inet6_rcv_saddr(sk);
65 		const struct in6_addr *sk_raddr = &sk->sk_v6_daddr;
66 		const struct l2tp_ip6_sock *l2tp = l2tp_ip6_sk(sk);
67 
68 		if (!net_eq(sock_net(sk), net))
69 			continue;
70 
71 		if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
72 			continue;
73 
74 		if (sk_laddr && !ipv6_addr_any(sk_laddr) &&
75 		    !ipv6_addr_any(laddr) && !ipv6_addr_equal(sk_laddr, laddr))
76 			continue;
77 
78 		if (!ipv6_addr_any(sk_raddr) && raddr &&
79 		    !ipv6_addr_any(raddr) && !ipv6_addr_equal(sk_raddr, raddr))
80 			continue;
81 
82 		if (l2tp->conn_id != tunnel_id)
83 			continue;
84 
85 		goto found;
86 	}
87 
88 	sk = NULL;
89 found:
90 	return sk;
91 }
92 
93 /* When processing receive frames, there are two cases to
94  * consider. Data frames consist of a non-zero session-id and an
95  * optional cookie. Control frames consist of a regular L2TP header
96  * preceded by 32-bits of zeros.
97  *
98  * L2TPv3 Session Header Over IP
99  *
100  *  0                   1                   2                   3
101  *  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
102  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103  * |                           Session ID                          |
104  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105  * |               Cookie (optional, maximum 64 bits)...
106  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
107  *                                                                 |
108  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
109  *
110  * L2TPv3 Control Message Header Over IP
111  *
112  *  0                   1                   2                   3
113  *  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
114  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115  * |                      (32 bits of zeros)                       |
116  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117  * |T|L|x|x|S|x|x|x|x|x|x|x|  Ver  |             Length            |
118  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119  * |                     Control Connection ID                     |
120  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121  * |               Ns              |               Nr              |
122  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
123  *
124  * All control frames are passed to userspace.
125  */
126 static int l2tp_ip6_recv(struct sk_buff *skb)
127 {
128 	struct net *net = dev_net(skb->dev);
129 	struct sock *sk;
130 	u32 session_id;
131 	u32 tunnel_id;
132 	unsigned char *ptr, *optr;
133 	struct l2tp_session *session;
134 	struct l2tp_tunnel *tunnel = NULL;
135 	struct ipv6hdr *iph;
136 	int length;
137 
138 	if (!pskb_may_pull(skb, 4))
139 		goto discard;
140 
141 	/* Point to L2TP header */
142 	optr = ptr = skb->data;
143 	session_id = ntohl(*((__be32 *) ptr));
144 	ptr += 4;
145 
146 	/* RFC3931: L2TP/IP packets have the first 4 bytes containing
147 	 * the session_id. If it is 0, the packet is a L2TP control
148 	 * frame and the session_id value can be discarded.
149 	 */
150 	if (session_id == 0) {
151 		__skb_pull(skb, 4);
152 		goto pass_up;
153 	}
154 
155 	/* Ok, this is a data packet. Lookup the session. */
156 	session = l2tp_session_get(net, session_id);
157 	if (!session)
158 		goto discard;
159 
160 	tunnel = session->tunnel;
161 	if (!tunnel)
162 		goto discard_sess;
163 
164 	/* Trace packet contents, if enabled */
165 	if (tunnel->debug & L2TP_MSG_DATA) {
166 		length = min(32u, skb->len);
167 		if (!pskb_may_pull(skb, length))
168 			goto discard_sess;
169 
170 		/* Point to L2TP header */
171 		optr = ptr = skb->data;
172 		ptr += 4;
173 		pr_debug("%s: ip recv\n", tunnel->name);
174 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
175 	}
176 
177 	if (l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
178 		goto discard_sess;
179 
180 	l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
181 	l2tp_session_dec_refcount(session);
182 
183 	return 0;
184 
185 pass_up:
186 	/* Get the tunnel_id from the L2TP header */
187 	if (!pskb_may_pull(skb, 12))
188 		goto discard;
189 
190 	if ((skb->data[0] & 0xc0) != 0xc0)
191 		goto discard;
192 
193 	tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
194 	iph = ipv6_hdr(skb);
195 
196 	read_lock_bh(&l2tp_ip6_lock);
197 	sk = __l2tp_ip6_bind_lookup(net, &iph->daddr, &iph->saddr,
198 				    inet6_iif(skb), tunnel_id);
199 	if (!sk) {
200 		read_unlock_bh(&l2tp_ip6_lock);
201 		goto discard;
202 	}
203 	sock_hold(sk);
204 	read_unlock_bh(&l2tp_ip6_lock);
205 
206 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
207 		goto discard_put;
208 
209 	nf_reset(skb);
210 
211 	return sk_receive_skb(sk, skb, 1);
212 
213 discard_sess:
214 	l2tp_session_dec_refcount(session);
215 	goto discard;
216 
217 discard_put:
218 	sock_put(sk);
219 
220 discard:
221 	kfree_skb(skb);
222 	return 0;
223 }
224 
225 static int l2tp_ip6_open(struct sock *sk)
226 {
227 	/* Prevent autobind. We don't have ports. */
228 	inet_sk(sk)->inet_num = IPPROTO_L2TP;
229 
230 	write_lock_bh(&l2tp_ip6_lock);
231 	sk_add_node(sk, &l2tp_ip6_table);
232 	write_unlock_bh(&l2tp_ip6_lock);
233 
234 	return 0;
235 }
236 
237 static void l2tp_ip6_close(struct sock *sk, long timeout)
238 {
239 	write_lock_bh(&l2tp_ip6_lock);
240 	hlist_del_init(&sk->sk_bind_node);
241 	sk_del_node_init(sk);
242 	write_unlock_bh(&l2tp_ip6_lock);
243 
244 	sk_common_release(sk);
245 }
246 
247 static void l2tp_ip6_destroy_sock(struct sock *sk)
248 {
249 	struct l2tp_tunnel *tunnel = sk->sk_user_data;
250 
251 	lock_sock(sk);
252 	ip6_flush_pending_frames(sk);
253 	release_sock(sk);
254 
255 	if (tunnel)
256 		l2tp_tunnel_delete(tunnel);
257 
258 	inet6_destroy_sock(sk);
259 }
260 
261 static int l2tp_ip6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
262 {
263 	struct inet_sock *inet = inet_sk(sk);
264 	struct ipv6_pinfo *np = inet6_sk(sk);
265 	struct sockaddr_l2tpip6 *addr = (struct sockaddr_l2tpip6 *) uaddr;
266 	struct net *net = sock_net(sk);
267 	__be32 v4addr = 0;
268 	int bound_dev_if;
269 	int addr_type;
270 	int err;
271 
272 	if (addr->l2tp_family != AF_INET6)
273 		return -EINVAL;
274 	if (addr_len < sizeof(*addr))
275 		return -EINVAL;
276 
277 	addr_type = ipv6_addr_type(&addr->l2tp_addr);
278 
279 	/* l2tp_ip6 sockets are IPv6 only */
280 	if (addr_type == IPV6_ADDR_MAPPED)
281 		return -EADDRNOTAVAIL;
282 
283 	/* L2TP is point-point, not multicast */
284 	if (addr_type & IPV6_ADDR_MULTICAST)
285 		return -EADDRNOTAVAIL;
286 
287 	lock_sock(sk);
288 
289 	err = -EINVAL;
290 	if (!sock_flag(sk, SOCK_ZAPPED))
291 		goto out_unlock;
292 
293 	if (sk->sk_state != TCP_CLOSE)
294 		goto out_unlock;
295 
296 	bound_dev_if = sk->sk_bound_dev_if;
297 
298 	/* Check if the address belongs to the host. */
299 	rcu_read_lock();
300 	if (addr_type != IPV6_ADDR_ANY) {
301 		struct net_device *dev = NULL;
302 
303 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
304 			if (addr->l2tp_scope_id)
305 				bound_dev_if = addr->l2tp_scope_id;
306 
307 			/* Binding to link-local address requires an
308 			 * interface.
309 			 */
310 			if (!bound_dev_if)
311 				goto out_unlock_rcu;
312 
313 			err = -ENODEV;
314 			dev = dev_get_by_index_rcu(sock_net(sk), bound_dev_if);
315 			if (!dev)
316 				goto out_unlock_rcu;
317 		}
318 
319 		/* ipv4 addr of the socket is invalid.  Only the
320 		 * unspecified and mapped address have a v4 equivalent.
321 		 */
322 		v4addr = LOOPBACK4_IPV6;
323 		err = -EADDRNOTAVAIL;
324 		if (!ipv6_chk_addr(sock_net(sk), &addr->l2tp_addr, dev, 0))
325 			goto out_unlock_rcu;
326 	}
327 	rcu_read_unlock();
328 
329 	write_lock_bh(&l2tp_ip6_lock);
330 	if (__l2tp_ip6_bind_lookup(net, &addr->l2tp_addr, NULL, bound_dev_if,
331 				   addr->l2tp_conn_id)) {
332 		write_unlock_bh(&l2tp_ip6_lock);
333 		err = -EADDRINUSE;
334 		goto out_unlock;
335 	}
336 
337 	inet->inet_saddr = v4addr;
338 	inet->inet_rcv_saddr = v4addr;
339 	sk->sk_bound_dev_if = bound_dev_if;
340 	sk->sk_v6_rcv_saddr = addr->l2tp_addr;
341 	np->saddr = addr->l2tp_addr;
342 
343 	l2tp_ip6_sk(sk)->conn_id = addr->l2tp_conn_id;
344 
345 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
346 	sk_del_node_init(sk);
347 	write_unlock_bh(&l2tp_ip6_lock);
348 
349 	sock_reset_flag(sk, SOCK_ZAPPED);
350 	release_sock(sk);
351 	return 0;
352 
353 out_unlock_rcu:
354 	rcu_read_unlock();
355 out_unlock:
356 	release_sock(sk);
357 
358 	return err;
359 }
360 
361 static int l2tp_ip6_connect(struct sock *sk, struct sockaddr *uaddr,
362 			    int addr_len)
363 {
364 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *) uaddr;
365 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
366 	struct in6_addr	*daddr;
367 	int	addr_type;
368 	int rc;
369 
370 	if (addr_len < sizeof(*lsa))
371 		return -EINVAL;
372 
373 	if (usin->sin6_family != AF_INET6)
374 		return -EINVAL;
375 
376 	addr_type = ipv6_addr_type(&usin->sin6_addr);
377 	if (addr_type & IPV6_ADDR_MULTICAST)
378 		return -EINVAL;
379 
380 	if (addr_type & IPV6_ADDR_MAPPED) {
381 		daddr = &usin->sin6_addr;
382 		if (ipv4_is_multicast(daddr->s6_addr32[3]))
383 			return -EINVAL;
384 	}
385 
386 	lock_sock(sk);
387 
388 	 /* Must bind first - autobinding does not work */
389 	if (sock_flag(sk, SOCK_ZAPPED)) {
390 		rc = -EINVAL;
391 		goto out_sk;
392 	}
393 
394 	rc = __ip6_datagram_connect(sk, uaddr, addr_len);
395 	if (rc < 0)
396 		goto out_sk;
397 
398 	l2tp_ip6_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
399 
400 	write_lock_bh(&l2tp_ip6_lock);
401 	hlist_del_init(&sk->sk_bind_node);
402 	sk_add_bind_node(sk, &l2tp_ip6_bind_table);
403 	write_unlock_bh(&l2tp_ip6_lock);
404 
405 out_sk:
406 	release_sock(sk);
407 
408 	return rc;
409 }
410 
411 static int l2tp_ip6_disconnect(struct sock *sk, int flags)
412 {
413 	if (sock_flag(sk, SOCK_ZAPPED))
414 		return 0;
415 
416 	return __udp_disconnect(sk, flags);
417 }
418 
419 static int l2tp_ip6_getname(struct socket *sock, struct sockaddr *uaddr,
420 			    int peer)
421 {
422 	struct sockaddr_l2tpip6 *lsa = (struct sockaddr_l2tpip6 *)uaddr;
423 	struct sock *sk = sock->sk;
424 	struct ipv6_pinfo *np = inet6_sk(sk);
425 	struct l2tp_ip6_sock *lsk = l2tp_ip6_sk(sk);
426 
427 	lsa->l2tp_family = AF_INET6;
428 	lsa->l2tp_flowinfo = 0;
429 	lsa->l2tp_scope_id = 0;
430 	lsa->l2tp_unused = 0;
431 	if (peer) {
432 		if (!lsk->peer_conn_id)
433 			return -ENOTCONN;
434 		lsa->l2tp_conn_id = lsk->peer_conn_id;
435 		lsa->l2tp_addr = sk->sk_v6_daddr;
436 		if (np->sndflow)
437 			lsa->l2tp_flowinfo = np->flow_label;
438 	} else {
439 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
440 			lsa->l2tp_addr = np->saddr;
441 		else
442 			lsa->l2tp_addr = sk->sk_v6_rcv_saddr;
443 
444 		lsa->l2tp_conn_id = lsk->conn_id;
445 	}
446 	if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
447 		lsa->l2tp_scope_id = sk->sk_bound_dev_if;
448 	return sizeof(*lsa);
449 }
450 
451 static int l2tp_ip6_backlog_recv(struct sock *sk, struct sk_buff *skb)
452 {
453 	int rc;
454 
455 	/* Charge it to the socket, dropping if the queue is full. */
456 	rc = sock_queue_rcv_skb(sk, skb);
457 	if (rc < 0)
458 		goto drop;
459 
460 	return 0;
461 
462 drop:
463 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
464 	kfree_skb(skb);
465 	return -1;
466 }
467 
468 static int l2tp_ip6_push_pending_frames(struct sock *sk)
469 {
470 	struct sk_buff *skb;
471 	__be32 *transhdr = NULL;
472 	int err = 0;
473 
474 	skb = skb_peek(&sk->sk_write_queue);
475 	if (skb == NULL)
476 		goto out;
477 
478 	transhdr = (__be32 *)skb_transport_header(skb);
479 	*transhdr = 0;
480 
481 	err = ip6_push_pending_frames(sk);
482 
483 out:
484 	return err;
485 }
486 
487 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
488  * control frames.
489  */
490 static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
491 {
492 	struct ipv6_txoptions opt_space;
493 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
494 	struct in6_addr *daddr, *final_p, final;
495 	struct ipv6_pinfo *np = inet6_sk(sk);
496 	struct ipv6_txoptions *opt_to_free = NULL;
497 	struct ipv6_txoptions *opt = NULL;
498 	struct ip6_flowlabel *flowlabel = NULL;
499 	struct dst_entry *dst = NULL;
500 	struct flowi6 fl6;
501 	struct ipcm6_cookie ipc6;
502 	int addr_len = msg->msg_namelen;
503 	int transhdrlen = 4; /* zero session-id */
504 	int ulen = len + transhdrlen;
505 	int err;
506 
507 	/* Rough check on arithmetic overflow,
508 	   better check is made in ip6_append_data().
509 	 */
510 	if (len > INT_MAX)
511 		return -EMSGSIZE;
512 
513 	/* Mirror BSD error message compatibility */
514 	if (msg->msg_flags & MSG_OOB)
515 		return -EOPNOTSUPP;
516 
517 	/*
518 	 *	Get and verify the address.
519 	 */
520 	memset(&fl6, 0, sizeof(fl6));
521 
522 	fl6.flowi6_mark = sk->sk_mark;
523 	fl6.flowi6_uid = sk->sk_uid;
524 
525 	ipcm6_init(&ipc6);
526 
527 	if (lsa) {
528 		if (addr_len < SIN6_LEN_RFC2133)
529 			return -EINVAL;
530 
531 		if (lsa->l2tp_family && lsa->l2tp_family != AF_INET6)
532 			return -EAFNOSUPPORT;
533 
534 		daddr = &lsa->l2tp_addr;
535 		if (np->sndflow) {
536 			fl6.flowlabel = lsa->l2tp_flowinfo & IPV6_FLOWINFO_MASK;
537 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
538 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
539 				if (flowlabel == NULL)
540 					return -EINVAL;
541 			}
542 		}
543 
544 		/*
545 		 * Otherwise it will be difficult to maintain
546 		 * sk->sk_dst_cache.
547 		 */
548 		if (sk->sk_state == TCP_ESTABLISHED &&
549 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
550 			daddr = &sk->sk_v6_daddr;
551 
552 		if (addr_len >= sizeof(struct sockaddr_in6) &&
553 		    lsa->l2tp_scope_id &&
554 		    ipv6_addr_type(daddr) & IPV6_ADDR_LINKLOCAL)
555 			fl6.flowi6_oif = lsa->l2tp_scope_id;
556 	} else {
557 		if (sk->sk_state != TCP_ESTABLISHED)
558 			return -EDESTADDRREQ;
559 
560 		daddr = &sk->sk_v6_daddr;
561 		fl6.flowlabel = np->flow_label;
562 	}
563 
564 	if (fl6.flowi6_oif == 0)
565 		fl6.flowi6_oif = sk->sk_bound_dev_if;
566 
567 	if (msg->msg_controllen) {
568 		opt = &opt_space;
569 		memset(opt, 0, sizeof(struct ipv6_txoptions));
570 		opt->tot_len = sizeof(struct ipv6_txoptions);
571 		ipc6.opt = opt;
572 
573 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
574 		if (err < 0) {
575 			fl6_sock_release(flowlabel);
576 			return err;
577 		}
578 		if ((fl6.flowlabel & IPV6_FLOWLABEL_MASK) && !flowlabel) {
579 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
580 			if (flowlabel == NULL)
581 				return -EINVAL;
582 		}
583 		if (!(opt->opt_nflen|opt->opt_flen))
584 			opt = NULL;
585 	}
586 
587 	if (!opt) {
588 		opt = txopt_get(np);
589 		opt_to_free = opt;
590 	}
591 	if (flowlabel)
592 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
593 	opt = ipv6_fixup_options(&opt_space, opt);
594 	ipc6.opt = opt;
595 
596 	fl6.flowi6_proto = sk->sk_protocol;
597 	if (!ipv6_addr_any(daddr))
598 		fl6.daddr = *daddr;
599 	else
600 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
601 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
602 		fl6.saddr = np->saddr;
603 
604 	final_p = fl6_update_dst(&fl6, opt, &final);
605 
606 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
607 		fl6.flowi6_oif = np->mcast_oif;
608 	else if (!fl6.flowi6_oif)
609 		fl6.flowi6_oif = np->ucast_oif;
610 
611 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
612 
613 	if (ipc6.tclass < 0)
614 		ipc6.tclass = np->tclass;
615 
616 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
617 
618 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
619 	if (IS_ERR(dst)) {
620 		err = PTR_ERR(dst);
621 		goto out;
622 	}
623 
624 	if (ipc6.hlimit < 0)
625 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
626 
627 	if (ipc6.dontfrag < 0)
628 		ipc6.dontfrag = np->dontfrag;
629 
630 	if (msg->msg_flags & MSG_CONFIRM)
631 		goto do_confirm;
632 
633 back_from_confirm:
634 	lock_sock(sk);
635 	err = ip6_append_data(sk, ip_generic_getfrag, msg,
636 			      ulen, transhdrlen, &ipc6,
637 			      &fl6, (struct rt6_info *)dst,
638 			      msg->msg_flags);
639 	if (err)
640 		ip6_flush_pending_frames(sk);
641 	else if (!(msg->msg_flags & MSG_MORE))
642 		err = l2tp_ip6_push_pending_frames(sk);
643 	release_sock(sk);
644 done:
645 	dst_release(dst);
646 out:
647 	fl6_sock_release(flowlabel);
648 	txopt_put(opt_to_free);
649 
650 	return err < 0 ? err : len;
651 
652 do_confirm:
653 	if (msg->msg_flags & MSG_PROBE)
654 		dst_confirm_neigh(dst, &fl6.daddr);
655 	if (!(msg->msg_flags & MSG_PROBE) || len)
656 		goto back_from_confirm;
657 	err = 0;
658 	goto done;
659 }
660 
661 static int l2tp_ip6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
662 			    int noblock, int flags, int *addr_len)
663 {
664 	struct ipv6_pinfo *np = inet6_sk(sk);
665 	DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name);
666 	size_t copied = 0;
667 	int err = -EOPNOTSUPP;
668 	struct sk_buff *skb;
669 
670 	if (flags & MSG_OOB)
671 		goto out;
672 
673 	if (flags & MSG_ERRQUEUE)
674 		return ipv6_recv_error(sk, msg, len, addr_len);
675 
676 	skb = skb_recv_datagram(sk, flags, noblock, &err);
677 	if (!skb)
678 		goto out;
679 
680 	copied = skb->len;
681 	if (len < copied) {
682 		msg->msg_flags |= MSG_TRUNC;
683 		copied = len;
684 	}
685 
686 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
687 	if (err)
688 		goto done;
689 
690 	sock_recv_timestamp(msg, sk, skb);
691 
692 	/* Copy the address. */
693 	if (lsa) {
694 		lsa->l2tp_family = AF_INET6;
695 		lsa->l2tp_unused = 0;
696 		lsa->l2tp_addr = ipv6_hdr(skb)->saddr;
697 		lsa->l2tp_flowinfo = 0;
698 		lsa->l2tp_scope_id = 0;
699 		lsa->l2tp_conn_id = 0;
700 		if (ipv6_addr_type(&lsa->l2tp_addr) & IPV6_ADDR_LINKLOCAL)
701 			lsa->l2tp_scope_id = inet6_iif(skb);
702 		*addr_len = sizeof(*lsa);
703 	}
704 
705 	if (np->rxopt.all)
706 		ip6_datagram_recv_ctl(sk, msg, skb);
707 
708 	if (flags & MSG_TRUNC)
709 		copied = skb->len;
710 done:
711 	skb_free_datagram(sk, skb);
712 out:
713 	return err ? err : copied;
714 }
715 
716 static struct proto l2tp_ip6_prot = {
717 	.name		   = "L2TP/IPv6",
718 	.owner		   = THIS_MODULE,
719 	.init		   = l2tp_ip6_open,
720 	.close		   = l2tp_ip6_close,
721 	.bind		   = l2tp_ip6_bind,
722 	.connect	   = l2tp_ip6_connect,
723 	.disconnect	   = l2tp_ip6_disconnect,
724 	.ioctl		   = l2tp_ioctl,
725 	.destroy	   = l2tp_ip6_destroy_sock,
726 	.setsockopt	   = ipv6_setsockopt,
727 	.getsockopt	   = ipv6_getsockopt,
728 	.sendmsg	   = l2tp_ip6_sendmsg,
729 	.recvmsg	   = l2tp_ip6_recvmsg,
730 	.backlog_rcv	   = l2tp_ip6_backlog_recv,
731 	.hash		   = inet6_hash,
732 	.unhash		   = inet_unhash,
733 	.obj_size	   = sizeof(struct l2tp_ip6_sock),
734 #ifdef CONFIG_COMPAT
735 	.compat_setsockopt = compat_ipv6_setsockopt,
736 	.compat_getsockopt = compat_ipv6_getsockopt,
737 #endif
738 };
739 
740 static const struct proto_ops l2tp_ip6_ops = {
741 	.family		   = PF_INET6,
742 	.owner		   = THIS_MODULE,
743 	.release	   = inet6_release,
744 	.bind		   = inet6_bind,
745 	.connect	   = inet_dgram_connect,
746 	.socketpair	   = sock_no_socketpair,
747 	.accept		   = sock_no_accept,
748 	.getname	   = l2tp_ip6_getname,
749 	.poll		   = datagram_poll,
750 	.ioctl		   = inet6_ioctl,
751 	.gettstamp	   = sock_gettstamp,
752 	.listen		   = sock_no_listen,
753 	.shutdown	   = inet_shutdown,
754 	.setsockopt	   = sock_common_setsockopt,
755 	.getsockopt	   = sock_common_getsockopt,
756 	.sendmsg	   = inet_sendmsg,
757 	.recvmsg	   = sock_common_recvmsg,
758 	.mmap		   = sock_no_mmap,
759 	.sendpage	   = sock_no_sendpage,
760 #ifdef CONFIG_COMPAT
761 	.compat_setsockopt = compat_sock_common_setsockopt,
762 	.compat_getsockopt = compat_sock_common_getsockopt,
763 #endif
764 };
765 
766 static struct inet_protosw l2tp_ip6_protosw = {
767 	.type		= SOCK_DGRAM,
768 	.protocol	= IPPROTO_L2TP,
769 	.prot		= &l2tp_ip6_prot,
770 	.ops		= &l2tp_ip6_ops,
771 };
772 
773 static struct inet6_protocol l2tp_ip6_protocol __read_mostly = {
774 	.handler	= l2tp_ip6_recv,
775 };
776 
777 static int __init l2tp_ip6_init(void)
778 {
779 	int err;
780 
781 	pr_info("L2TP IP encapsulation support for IPv6 (L2TPv3)\n");
782 
783 	err = proto_register(&l2tp_ip6_prot, 1);
784 	if (err != 0)
785 		goto out;
786 
787 	err = inet6_add_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
788 	if (err)
789 		goto out1;
790 
791 	inet6_register_protosw(&l2tp_ip6_protosw);
792 	return 0;
793 
794 out1:
795 	proto_unregister(&l2tp_ip6_prot);
796 out:
797 	return err;
798 }
799 
800 static void __exit l2tp_ip6_exit(void)
801 {
802 	inet6_unregister_protosw(&l2tp_ip6_protosw);
803 	inet6_del_protocol(&l2tp_ip6_protocol, IPPROTO_L2TP);
804 	proto_unregister(&l2tp_ip6_prot);
805 }
806 
807 module_init(l2tp_ip6_init);
808 module_exit(l2tp_ip6_exit);
809 
810 MODULE_LICENSE("GPL");
811 MODULE_AUTHOR("Chris Elston <celston@katalix.com>");
812 MODULE_DESCRIPTION("L2TP IP encapsulation for IPv6");
813 MODULE_VERSION("1.0");
814 
815 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
816  * enums
817  */
818 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 2, IPPROTO_L2TP);
819 MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_L2TP);
820