xref: /linux/net/ipv6/datagram.c (revision 5bdd5fbb35ab0fe21bf2263106f51c5bee466a07)
1 /*
2  *	common UDP/RAW code
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *
8  *	This program is free software; you can redistribute it and/or
9  *      modify it under the terms of the GNU General Public License
10  *      as published by the Free Software Foundation; either version
11  *      2 of the License, or (at your option) any later version.
12  */
13 
14 #include <linux/capability.h>
15 #include <linux/errno.h>
16 #include <linux/types.h>
17 #include <linux/kernel.h>
18 #include <linux/interrupt.h>
19 #include <linux/socket.h>
20 #include <linux/sockios.h>
21 #include <linux/in6.h>
22 #include <linux/ipv6.h>
23 #include <linux/route.h>
24 #include <linux/slab.h>
25 #include <linux/export.h>
26 
27 #include <net/ipv6.h>
28 #include <net/ndisc.h>
29 #include <net/addrconf.h>
30 #include <net/transp_v6.h>
31 #include <net/ip6_route.h>
32 #include <net/tcp_states.h>
33 #include <net/dsfield.h>
34 
35 #include <linux/errqueue.h>
36 #include <linux/uaccess.h>
37 
38 static bool ipv6_mapped_addr_any(const struct in6_addr *a)
39 {
40 	return ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0);
41 }
42 
43 static void ip6_datagram_flow_key_init(struct flowi6 *fl6, struct sock *sk)
44 {
45 	struct inet_sock *inet = inet_sk(sk);
46 	struct ipv6_pinfo *np = inet6_sk(sk);
47 
48 	memset(fl6, 0, sizeof(*fl6));
49 	fl6->flowi6_proto = sk->sk_protocol;
50 	fl6->daddr = sk->sk_v6_daddr;
51 	fl6->saddr = np->saddr;
52 	fl6->flowi6_oif = sk->sk_bound_dev_if;
53 	fl6->flowi6_mark = sk->sk_mark;
54 	fl6->fl6_dport = inet->inet_dport;
55 	fl6->fl6_sport = inet->inet_sport;
56 	fl6->flowlabel = np->flow_label;
57 	fl6->flowi6_uid = sk->sk_uid;
58 
59 	if (!fl6->flowi6_oif)
60 		fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
61 
62 	if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr))
63 		fl6->flowi6_oif = np->mcast_oif;
64 
65 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
66 }
67 
68 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr)
69 {
70 	struct ip6_flowlabel *flowlabel = NULL;
71 	struct in6_addr *final_p, final;
72 	struct ipv6_txoptions *opt;
73 	struct dst_entry *dst;
74 	struct inet_sock *inet = inet_sk(sk);
75 	struct ipv6_pinfo *np = inet6_sk(sk);
76 	struct flowi6 fl6;
77 	int err = 0;
78 
79 	if (np->sndflow && (np->flow_label & IPV6_FLOWLABEL_MASK)) {
80 		flowlabel = fl6_sock_lookup(sk, np->flow_label);
81 		if (!flowlabel)
82 			return -EINVAL;
83 	}
84 	ip6_datagram_flow_key_init(&fl6, sk);
85 
86 	rcu_read_lock();
87 	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
88 	final_p = fl6_update_dst(&fl6, opt, &final);
89 	rcu_read_unlock();
90 
91 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
92 	if (IS_ERR(dst)) {
93 		err = PTR_ERR(dst);
94 		goto out;
95 	}
96 
97 	if (fix_sk_saddr) {
98 		if (ipv6_addr_any(&np->saddr))
99 			np->saddr = fl6.saddr;
100 
101 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
102 			sk->sk_v6_rcv_saddr = fl6.saddr;
103 			inet->inet_rcv_saddr = LOOPBACK4_IPV6;
104 			if (sk->sk_prot->rehash)
105 				sk->sk_prot->rehash(sk);
106 		}
107 	}
108 
109 	ip6_sk_dst_store_flow(sk, dst, &fl6);
110 
111 out:
112 	fl6_sock_release(flowlabel);
113 	return err;
114 }
115 
116 void ip6_datagram_release_cb(struct sock *sk)
117 {
118 	struct dst_entry *dst;
119 
120 	if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
121 		return;
122 
123 	rcu_read_lock();
124 	dst = __sk_dst_get(sk);
125 	if (!dst || !dst->obsolete ||
126 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
127 		rcu_read_unlock();
128 		return;
129 	}
130 	rcu_read_unlock();
131 
132 	ip6_datagram_dst_update(sk, false);
133 }
134 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
135 
136 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
137 			   int addr_len)
138 {
139 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
140 	struct inet_sock	*inet = inet_sk(sk);
141 	struct ipv6_pinfo	*np = inet6_sk(sk);
142 	struct in6_addr		*daddr, old_daddr;
143 	__be32			fl6_flowlabel = 0;
144 	__be32			old_fl6_flowlabel;
145 	__be16			old_dport;
146 	int			addr_type;
147 	int			err;
148 
149 	if (usin->sin6_family == AF_INET) {
150 		if (__ipv6_only_sock(sk))
151 			return -EAFNOSUPPORT;
152 		err = __ip4_datagram_connect(sk, uaddr, addr_len);
153 		goto ipv4_connected;
154 	}
155 
156 	if (addr_len < SIN6_LEN_RFC2133)
157 		return -EINVAL;
158 
159 	if (usin->sin6_family != AF_INET6)
160 		return -EAFNOSUPPORT;
161 
162 	if (np->sndflow)
163 		fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
164 
165 	if (ipv6_addr_any(&usin->sin6_addr)) {
166 		/*
167 		 *	connect to self
168 		 */
169 		if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr))
170 			ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
171 					       &usin->sin6_addr);
172 		else
173 			usin->sin6_addr = in6addr_loopback;
174 	}
175 
176 	addr_type = ipv6_addr_type(&usin->sin6_addr);
177 
178 	daddr = &usin->sin6_addr;
179 
180 	if (addr_type & IPV6_ADDR_MAPPED) {
181 		struct sockaddr_in sin;
182 
183 		if (__ipv6_only_sock(sk)) {
184 			err = -ENETUNREACH;
185 			goto out;
186 		}
187 		sin.sin_family = AF_INET;
188 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
189 		sin.sin_port = usin->sin6_port;
190 
191 		err = __ip4_datagram_connect(sk,
192 					     (struct sockaddr *) &sin,
193 					     sizeof(sin));
194 
195 ipv4_connected:
196 		if (err)
197 			goto out;
198 
199 		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
200 
201 		if (ipv6_addr_any(&np->saddr) ||
202 		    ipv6_mapped_addr_any(&np->saddr))
203 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
204 
205 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
206 		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
207 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
208 					       &sk->sk_v6_rcv_saddr);
209 			if (sk->sk_prot->rehash)
210 				sk->sk_prot->rehash(sk);
211 		}
212 
213 		goto out;
214 	}
215 
216 	if (__ipv6_addr_needs_scope_id(addr_type)) {
217 		if (addr_len >= sizeof(struct sockaddr_in6) &&
218 		    usin->sin6_scope_id) {
219 			if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) {
220 				err = -EINVAL;
221 				goto out;
222 			}
223 			sk->sk_bound_dev_if = usin->sin6_scope_id;
224 		}
225 
226 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
227 			sk->sk_bound_dev_if = np->mcast_oif;
228 
229 		/* Connect to link-local address requires an interface */
230 		if (!sk->sk_bound_dev_if) {
231 			err = -EINVAL;
232 			goto out;
233 		}
234 	}
235 
236 	/* save the current peer information before updating it */
237 	old_daddr = sk->sk_v6_daddr;
238 	old_fl6_flowlabel = np->flow_label;
239 	old_dport = inet->inet_dport;
240 
241 	sk->sk_v6_daddr = *daddr;
242 	np->flow_label = fl6_flowlabel;
243 	inet->inet_dport = usin->sin6_port;
244 
245 	/*
246 	 *	Check for a route to destination an obtain the
247 	 *	destination cache for it.
248 	 */
249 
250 	err = ip6_datagram_dst_update(sk, true);
251 	if (err) {
252 		/* Restore the socket peer info, to keep it consistent with
253 		 * the old socket state
254 		 */
255 		sk->sk_v6_daddr = old_daddr;
256 		np->flow_label = old_fl6_flowlabel;
257 		inet->inet_dport = old_dport;
258 		goto out;
259 	}
260 
261 	sk->sk_state = TCP_ESTABLISHED;
262 	sk_set_txhash(sk);
263 out:
264 	return err;
265 }
266 EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
267 
268 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
269 {
270 	int res;
271 
272 	lock_sock(sk);
273 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
274 	release_sock(sk);
275 	return res;
276 }
277 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
278 
279 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
280 				 int addr_len)
281 {
282 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
283 	if (sin6->sin6_family != AF_INET6)
284 		return -EAFNOSUPPORT;
285 	return ip6_datagram_connect(sk, uaddr, addr_len);
286 }
287 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
288 
289 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
290 		     __be16 port, u32 info, u8 *payload)
291 {
292 	struct ipv6_pinfo *np  = inet6_sk(sk);
293 	struct icmp6hdr *icmph = icmp6_hdr(skb);
294 	struct sock_exterr_skb *serr;
295 
296 	if (!np->recverr)
297 		return;
298 
299 	skb = skb_clone(skb, GFP_ATOMIC);
300 	if (!skb)
301 		return;
302 
303 	skb->protocol = htons(ETH_P_IPV6);
304 
305 	serr = SKB_EXT_ERR(skb);
306 	serr->ee.ee_errno = err;
307 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
308 	serr->ee.ee_type = icmph->icmp6_type;
309 	serr->ee.ee_code = icmph->icmp6_code;
310 	serr->ee.ee_pad = 0;
311 	serr->ee.ee_info = info;
312 	serr->ee.ee_data = 0;
313 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
314 				  skb_network_header(skb);
315 	serr->port = port;
316 
317 	__skb_pull(skb, payload - skb->data);
318 	skb_reset_transport_header(skb);
319 
320 	if (sock_queue_err_skb(sk, skb))
321 		kfree_skb(skb);
322 }
323 
324 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
325 {
326 	const struct ipv6_pinfo *np = inet6_sk(sk);
327 	struct sock_exterr_skb *serr;
328 	struct ipv6hdr *iph;
329 	struct sk_buff *skb;
330 
331 	if (!np->recverr)
332 		return;
333 
334 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
335 	if (!skb)
336 		return;
337 
338 	skb->protocol = htons(ETH_P_IPV6);
339 
340 	skb_put(skb, sizeof(struct ipv6hdr));
341 	skb_reset_network_header(skb);
342 	iph = ipv6_hdr(skb);
343 	iph->daddr = fl6->daddr;
344 
345 	serr = SKB_EXT_ERR(skb);
346 	serr->ee.ee_errno = err;
347 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
348 	serr->ee.ee_type = 0;
349 	serr->ee.ee_code = 0;
350 	serr->ee.ee_pad = 0;
351 	serr->ee.ee_info = info;
352 	serr->ee.ee_data = 0;
353 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
354 	serr->port = fl6->fl6_dport;
355 
356 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
357 	skb_reset_transport_header(skb);
358 
359 	if (sock_queue_err_skb(sk, skb))
360 		kfree_skb(skb);
361 }
362 
363 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
364 {
365 	struct ipv6_pinfo *np = inet6_sk(sk);
366 	struct ipv6hdr *iph;
367 	struct sk_buff *skb;
368 	struct ip6_mtuinfo *mtu_info;
369 
370 	if (!np->rxopt.bits.rxpmtu)
371 		return;
372 
373 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
374 	if (!skb)
375 		return;
376 
377 	skb_put(skb, sizeof(struct ipv6hdr));
378 	skb_reset_network_header(skb);
379 	iph = ipv6_hdr(skb);
380 	iph->daddr = fl6->daddr;
381 
382 	mtu_info = IP6CBMTU(skb);
383 
384 	mtu_info->ip6m_mtu = mtu;
385 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
386 	mtu_info->ip6m_addr.sin6_port = 0;
387 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
388 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
389 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
390 
391 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
392 	skb_reset_transport_header(skb);
393 
394 	skb = xchg(&np->rxpmtu, skb);
395 	kfree_skb(skb);
396 }
397 
398 /* For some errors we have valid addr_offset even with zero payload and
399  * zero port. Also, addr_offset should be supported if port is set.
400  */
401 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
402 {
403 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
404 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
405 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
406 }
407 
408 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
409  *
410  * At one point, excluding local errors was a quick test to identify icmp/icmp6
411  * errors. This is no longer true, but the test remained, so the v6 stack,
412  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
413  */
414 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
415 				      struct sock_exterr_skb *serr)
416 {
417 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
418 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
419 		return true;
420 
421 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
422 		return false;
423 
424 	if (!IP6CB(skb)->iif)
425 		return false;
426 
427 	return true;
428 }
429 
430 /*
431  *	Handle MSG_ERRQUEUE
432  */
433 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
434 {
435 	struct ipv6_pinfo *np = inet6_sk(sk);
436 	struct sock_exterr_skb *serr;
437 	struct sk_buff *skb;
438 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
439 	struct {
440 		struct sock_extended_err ee;
441 		struct sockaddr_in6	 offender;
442 	} errhdr;
443 	int err;
444 	int copied;
445 
446 	err = -EAGAIN;
447 	skb = sock_dequeue_err_skb(sk);
448 	if (!skb)
449 		goto out;
450 
451 	copied = skb->len;
452 	if (copied > len) {
453 		msg->msg_flags |= MSG_TRUNC;
454 		copied = len;
455 	}
456 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
457 	if (unlikely(err)) {
458 		kfree_skb(skb);
459 		return err;
460 	}
461 	sock_recv_timestamp(msg, sk, skb);
462 
463 	serr = SKB_EXT_ERR(skb);
464 
465 	if (sin && ipv6_datagram_support_addr(serr)) {
466 		const unsigned char *nh = skb_network_header(skb);
467 		sin->sin6_family = AF_INET6;
468 		sin->sin6_flowinfo = 0;
469 		sin->sin6_port = serr->port;
470 		if (skb->protocol == htons(ETH_P_IPV6)) {
471 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
472 								  struct ipv6hdr, daddr);
473 			sin->sin6_addr = ip6h->daddr;
474 			if (np->sndflow)
475 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
476 			sin->sin6_scope_id =
477 				ipv6_iface_scope_id(&sin->sin6_addr,
478 						    IP6CB(skb)->iif);
479 		} else {
480 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
481 					       &sin->sin6_addr);
482 			sin->sin6_scope_id = 0;
483 		}
484 		*addr_len = sizeof(*sin);
485 	}
486 
487 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
488 	sin = &errhdr.offender;
489 	memset(sin, 0, sizeof(*sin));
490 
491 	if (ip6_datagram_support_cmsg(skb, serr)) {
492 		sin->sin6_family = AF_INET6;
493 		if (np->rxopt.all)
494 			ip6_datagram_recv_common_ctl(sk, msg, skb);
495 		if (skb->protocol == htons(ETH_P_IPV6)) {
496 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
497 			if (np->rxopt.all)
498 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
499 			sin->sin6_scope_id =
500 				ipv6_iface_scope_id(&sin->sin6_addr,
501 						    IP6CB(skb)->iif);
502 		} else {
503 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
504 					       &sin->sin6_addr);
505 			if (inet_sk(sk)->cmsg_flags)
506 				ip_cmsg_recv(msg, skb);
507 		}
508 	}
509 
510 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
511 
512 	/* Now we could try to dump offended packet options */
513 
514 	msg->msg_flags |= MSG_ERRQUEUE;
515 	err = copied;
516 
517 	consume_skb(skb);
518 out:
519 	return err;
520 }
521 EXPORT_SYMBOL_GPL(ipv6_recv_error);
522 
523 /*
524  *	Handle IPV6_RECVPATHMTU
525  */
526 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
527 		     int *addr_len)
528 {
529 	struct ipv6_pinfo *np = inet6_sk(sk);
530 	struct sk_buff *skb;
531 	struct ip6_mtuinfo mtu_info;
532 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
533 	int err;
534 	int copied;
535 
536 	err = -EAGAIN;
537 	skb = xchg(&np->rxpmtu, NULL);
538 	if (!skb)
539 		goto out;
540 
541 	copied = skb->len;
542 	if (copied > len) {
543 		msg->msg_flags |= MSG_TRUNC;
544 		copied = len;
545 	}
546 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
547 	if (err)
548 		goto out_free_skb;
549 
550 	sock_recv_timestamp(msg, sk, skb);
551 
552 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
553 
554 	if (sin) {
555 		sin->sin6_family = AF_INET6;
556 		sin->sin6_flowinfo = 0;
557 		sin->sin6_port = 0;
558 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
559 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
560 		*addr_len = sizeof(*sin);
561 	}
562 
563 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
564 
565 	err = copied;
566 
567 out_free_skb:
568 	kfree_skb(skb);
569 out:
570 	return err;
571 }
572 
573 
574 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
575 				 struct sk_buff *skb)
576 {
577 	struct ipv6_pinfo *np = inet6_sk(sk);
578 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
579 
580 	if (np->rxopt.bits.rxinfo) {
581 		struct in6_pktinfo src_info;
582 
583 		if (is_ipv6) {
584 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
585 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
586 		} else {
587 			src_info.ipi6_ifindex =
588 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
589 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
590 					       &src_info.ipi6_addr);
591 		}
592 
593 		if (src_info.ipi6_ifindex >= 0)
594 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
595 				 sizeof(src_info), &src_info);
596 	}
597 }
598 
599 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
600 				    struct sk_buff *skb)
601 {
602 	struct ipv6_pinfo *np = inet6_sk(sk);
603 	struct inet6_skb_parm *opt = IP6CB(skb);
604 	unsigned char *nh = skb_network_header(skb);
605 
606 	if (np->rxopt.bits.rxhlim) {
607 		int hlim = ipv6_hdr(skb)->hop_limit;
608 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
609 	}
610 
611 	if (np->rxopt.bits.rxtclass) {
612 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
613 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
614 	}
615 
616 	if (np->rxopt.bits.rxflow) {
617 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
618 		if (flowinfo)
619 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
620 	}
621 
622 	/* HbH is allowed only once */
623 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
624 		u8 *ptr = nh + sizeof(struct ipv6hdr);
625 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
626 	}
627 
628 	if (opt->lastopt &&
629 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
630 		/*
631 		 * Silly enough, but we need to reparse in order to
632 		 * report extension headers (except for HbH)
633 		 * in order.
634 		 *
635 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
636 		 * (and WILL NOT be) defined because
637 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
638 		 */
639 		unsigned int off = sizeof(struct ipv6hdr);
640 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
641 
642 		while (off <= opt->lastopt) {
643 			unsigned int len;
644 			u8 *ptr = nh + off;
645 
646 			switch (nexthdr) {
647 			case IPPROTO_DSTOPTS:
648 				nexthdr = ptr[0];
649 				len = (ptr[1] + 1) << 3;
650 				if (np->rxopt.bits.dstopts)
651 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
652 				break;
653 			case IPPROTO_ROUTING:
654 				nexthdr = ptr[0];
655 				len = (ptr[1] + 1) << 3;
656 				if (np->rxopt.bits.srcrt)
657 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
658 				break;
659 			case IPPROTO_AH:
660 				nexthdr = ptr[0];
661 				len = (ptr[1] + 2) << 2;
662 				break;
663 			default:
664 				nexthdr = ptr[0];
665 				len = (ptr[1] + 1) << 3;
666 				break;
667 			}
668 
669 			off += len;
670 		}
671 	}
672 
673 	/* socket options in old style */
674 	if (np->rxopt.bits.rxoinfo) {
675 		struct in6_pktinfo src_info;
676 
677 		src_info.ipi6_ifindex = opt->iif;
678 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
679 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
680 	}
681 	if (np->rxopt.bits.rxohlim) {
682 		int hlim = ipv6_hdr(skb)->hop_limit;
683 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
684 	}
685 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
686 		u8 *ptr = nh + sizeof(struct ipv6hdr);
687 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
688 	}
689 	if (np->rxopt.bits.odstopts && opt->dst0) {
690 		u8 *ptr = nh + opt->dst0;
691 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
692 	}
693 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
694 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
695 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
696 	}
697 	if (np->rxopt.bits.odstopts && opt->dst1) {
698 		u8 *ptr = nh + opt->dst1;
699 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
700 	}
701 	if (np->rxopt.bits.rxorigdstaddr) {
702 		struct sockaddr_in6 sin6;
703 		__be16 *ports = (__be16 *) skb_transport_header(skb);
704 
705 		if (skb_transport_offset(skb) + 4 <= (int)skb->len) {
706 			/* All current transport protocols have the port numbers in the
707 			 * first four bytes of the transport header and this function is
708 			 * written with this assumption in mind.
709 			 */
710 
711 			sin6.sin6_family = AF_INET6;
712 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
713 			sin6.sin6_port = ports[1];
714 			sin6.sin6_flowinfo = 0;
715 			sin6.sin6_scope_id =
716 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
717 						    opt->iif);
718 
719 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
720 		}
721 	}
722 	if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
723 		int val = opt->frag_max_size;
724 
725 		put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
726 	}
727 }
728 
729 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
730 			  struct sk_buff *skb)
731 {
732 	ip6_datagram_recv_common_ctl(sk, msg, skb);
733 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
734 }
735 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
736 
737 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
738 			  struct msghdr *msg, struct flowi6 *fl6,
739 			  struct ipcm6_cookie *ipc6, struct sockcm_cookie *sockc)
740 {
741 	struct in6_pktinfo *src_info;
742 	struct cmsghdr *cmsg;
743 	struct ipv6_rt_hdr *rthdr;
744 	struct ipv6_opt_hdr *hdr;
745 	struct ipv6_txoptions *opt = ipc6->opt;
746 	int len;
747 	int err = 0;
748 
749 	for_each_cmsghdr(cmsg, msg) {
750 		int addr_type;
751 
752 		if (!CMSG_OK(msg, cmsg)) {
753 			err = -EINVAL;
754 			goto exit_f;
755 		}
756 
757 		if (cmsg->cmsg_level == SOL_SOCKET) {
758 			err = __sock_cmsg_send(sk, msg, cmsg, sockc);
759 			if (err)
760 				return err;
761 			continue;
762 		}
763 
764 		if (cmsg->cmsg_level != SOL_IPV6)
765 			continue;
766 
767 		switch (cmsg->cmsg_type) {
768 		case IPV6_PKTINFO:
769 		case IPV6_2292PKTINFO:
770 		    {
771 			struct net_device *dev = NULL;
772 
773 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
774 				err = -EINVAL;
775 				goto exit_f;
776 			}
777 
778 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
779 
780 			if (src_info->ipi6_ifindex) {
781 				if (fl6->flowi6_oif &&
782 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
783 					return -EINVAL;
784 				fl6->flowi6_oif = src_info->ipi6_ifindex;
785 			}
786 
787 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
788 
789 			rcu_read_lock();
790 			if (fl6->flowi6_oif) {
791 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
792 				if (!dev) {
793 					rcu_read_unlock();
794 					return -ENODEV;
795 				}
796 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
797 				rcu_read_unlock();
798 				return -EINVAL;
799 			}
800 
801 			if (addr_type != IPV6_ADDR_ANY) {
802 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
803 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
804 				    !ipv6_chk_addr_and_flags(net, &src_info->ipi6_addr,
805 							     dev, !strict, 0,
806 							     IFA_F_TENTATIVE) &&
807 				    !ipv6_chk_acast_addr_src(net, dev,
808 							     &src_info->ipi6_addr))
809 					err = -EINVAL;
810 				else
811 					fl6->saddr = src_info->ipi6_addr;
812 			}
813 
814 			rcu_read_unlock();
815 
816 			if (err)
817 				goto exit_f;
818 
819 			break;
820 		    }
821 
822 		case IPV6_FLOWINFO:
823 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
824 				err = -EINVAL;
825 				goto exit_f;
826 			}
827 
828 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
829 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
830 					err = -EINVAL;
831 					goto exit_f;
832 				}
833 			}
834 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
835 			break;
836 
837 		case IPV6_2292HOPOPTS:
838 		case IPV6_HOPOPTS:
839 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
840 				err = -EINVAL;
841 				goto exit_f;
842 			}
843 
844 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
845 			len = ((hdr->hdrlen + 1) << 3);
846 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
847 				err = -EINVAL;
848 				goto exit_f;
849 			}
850 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
851 				err = -EPERM;
852 				goto exit_f;
853 			}
854 			opt->opt_nflen += len;
855 			opt->hopopt = hdr;
856 			break;
857 
858 		case IPV6_2292DSTOPTS:
859 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
860 				err = -EINVAL;
861 				goto exit_f;
862 			}
863 
864 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
865 			len = ((hdr->hdrlen + 1) << 3);
866 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
867 				err = -EINVAL;
868 				goto exit_f;
869 			}
870 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
871 				err = -EPERM;
872 				goto exit_f;
873 			}
874 			if (opt->dst1opt) {
875 				err = -EINVAL;
876 				goto exit_f;
877 			}
878 			opt->opt_flen += len;
879 			opt->dst1opt = hdr;
880 			break;
881 
882 		case IPV6_DSTOPTS:
883 		case IPV6_RTHDRDSTOPTS:
884 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
885 				err = -EINVAL;
886 				goto exit_f;
887 			}
888 
889 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
890 			len = ((hdr->hdrlen + 1) << 3);
891 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
892 				err = -EINVAL;
893 				goto exit_f;
894 			}
895 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
896 				err = -EPERM;
897 				goto exit_f;
898 			}
899 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
900 				opt->opt_flen += len;
901 				opt->dst1opt = hdr;
902 			} else {
903 				opt->opt_nflen += len;
904 				opt->dst0opt = hdr;
905 			}
906 			break;
907 
908 		case IPV6_2292RTHDR:
909 		case IPV6_RTHDR:
910 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
911 				err = -EINVAL;
912 				goto exit_f;
913 			}
914 
915 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
916 
917 			switch (rthdr->type) {
918 #if IS_ENABLED(CONFIG_IPV6_MIP6)
919 			case IPV6_SRCRT_TYPE_2:
920 				if (rthdr->hdrlen != 2 ||
921 				    rthdr->segments_left != 1) {
922 					err = -EINVAL;
923 					goto exit_f;
924 				}
925 				break;
926 #endif
927 			default:
928 				err = -EINVAL;
929 				goto exit_f;
930 			}
931 
932 			len = ((rthdr->hdrlen + 1) << 3);
933 
934 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
935 				err = -EINVAL;
936 				goto exit_f;
937 			}
938 
939 			/* segments left must also match */
940 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
941 				err = -EINVAL;
942 				goto exit_f;
943 			}
944 
945 			opt->opt_nflen += len;
946 			opt->srcrt = rthdr;
947 
948 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
949 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
950 
951 				opt->opt_nflen += dsthdrlen;
952 				opt->dst0opt = opt->dst1opt;
953 				opt->dst1opt = NULL;
954 				opt->opt_flen -= dsthdrlen;
955 			}
956 
957 			break;
958 
959 		case IPV6_2292HOPLIMIT:
960 		case IPV6_HOPLIMIT:
961 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
962 				err = -EINVAL;
963 				goto exit_f;
964 			}
965 
966 			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
967 			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
968 				err = -EINVAL;
969 				goto exit_f;
970 			}
971 
972 			break;
973 
974 		case IPV6_TCLASS:
975 		    {
976 			int tc;
977 
978 			err = -EINVAL;
979 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
980 				goto exit_f;
981 
982 			tc = *(int *)CMSG_DATA(cmsg);
983 			if (tc < -1 || tc > 0xff)
984 				goto exit_f;
985 
986 			err = 0;
987 			ipc6->tclass = tc;
988 
989 			break;
990 		    }
991 
992 		case IPV6_DONTFRAG:
993 		    {
994 			int df;
995 
996 			err = -EINVAL;
997 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
998 				goto exit_f;
999 
1000 			df = *(int *)CMSG_DATA(cmsg);
1001 			if (df < 0 || df > 1)
1002 				goto exit_f;
1003 
1004 			err = 0;
1005 			ipc6->dontfrag = df;
1006 
1007 			break;
1008 		    }
1009 		default:
1010 			net_dbg_ratelimited("invalid cmsg type: %d\n",
1011 					    cmsg->cmsg_type);
1012 			err = -EINVAL;
1013 			goto exit_f;
1014 		}
1015 	}
1016 
1017 exit_f:
1018 	return err;
1019 }
1020 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1021 
1022 void __ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1023 			       __u16 srcp, __u16 destp, int rqueue, int bucket)
1024 {
1025 	const struct in6_addr *dest, *src;
1026 
1027 	dest  = &sp->sk_v6_daddr;
1028 	src   = &sp->sk_v6_rcv_saddr;
1029 	seq_printf(seq,
1030 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1031 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1032 		   bucket,
1033 		   src->s6_addr32[0], src->s6_addr32[1],
1034 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1035 		   dest->s6_addr32[0], dest->s6_addr32[1],
1036 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1037 		   sp->sk_state,
1038 		   sk_wmem_alloc_get(sp),
1039 		   rqueue,
1040 		   0, 0L, 0,
1041 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1042 		   0,
1043 		   sock_i_ino(sp),
1044 		   refcount_read(&sp->sk_refcnt), sp,
1045 		   atomic_read(&sp->sk_drops));
1046 }
1047