xref: /linux/net/ipv6/datagram.c (revision 50f0033d1a0f3a8e9eed09ab68067fbb57b0669d)
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 <asm/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 
58 	if (!fl6->flowi6_oif)
59 		fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
60 
61 	if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr))
62 		fl6->flowi6_oif = np->mcast_oif;
63 
64 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
65 }
66 
67 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr)
68 {
69 	struct ip6_flowlabel *flowlabel = NULL;
70 	struct in6_addr *final_p, final;
71 	struct ipv6_txoptions *opt;
72 	struct dst_entry *dst;
73 	struct inet_sock *inet = inet_sk(sk);
74 	struct ipv6_pinfo *np = inet6_sk(sk);
75 	struct flowi6 fl6;
76 	int err = 0;
77 
78 	if (np->sndflow && (np->flow_label & IPV6_FLOWLABEL_MASK)) {
79 		flowlabel = fl6_sock_lookup(sk, np->flow_label);
80 		if (!flowlabel)
81 			return -EINVAL;
82 	}
83 	ip6_datagram_flow_key_init(&fl6, sk);
84 
85 	rcu_read_lock();
86 	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
87 	final_p = fl6_update_dst(&fl6, opt, &final);
88 	rcu_read_unlock();
89 
90 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
91 	if (IS_ERR(dst)) {
92 		err = PTR_ERR(dst);
93 		goto out;
94 	}
95 
96 	if (fix_sk_saddr) {
97 		if (ipv6_addr_any(&np->saddr))
98 			np->saddr = fl6.saddr;
99 
100 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
101 			sk->sk_v6_rcv_saddr = fl6.saddr;
102 			inet->inet_rcv_saddr = LOOPBACK4_IPV6;
103 			if (sk->sk_prot->rehash)
104 				sk->sk_prot->rehash(sk);
105 		}
106 	}
107 
108 	ip6_dst_store(sk, dst,
109 		      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
110 		      &sk->sk_v6_daddr : NULL,
111 #ifdef CONFIG_IPV6_SUBTREES
112 		      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
113 		      &np->saddr :
114 #endif
115 		      NULL);
116 
117 out:
118 	fl6_sock_release(flowlabel);
119 	return err;
120 }
121 
122 void ip6_datagram_release_cb(struct sock *sk)
123 {
124 	struct dst_entry *dst;
125 
126 	if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
127 		return;
128 
129 	rcu_read_lock();
130 	dst = __sk_dst_get(sk);
131 	if (!dst || !dst->obsolete ||
132 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
133 		rcu_read_unlock();
134 		return;
135 	}
136 	rcu_read_unlock();
137 
138 	ip6_datagram_dst_update(sk, false);
139 }
140 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
141 
142 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
143 			   int addr_len)
144 {
145 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
146 	struct inet_sock	*inet = inet_sk(sk);
147 	struct ipv6_pinfo	*np = inet6_sk(sk);
148 	struct in6_addr		*daddr;
149 	int			addr_type;
150 	int			err;
151 	__be32			fl6_flowlabel = 0;
152 
153 	if (usin->sin6_family == AF_INET) {
154 		if (__ipv6_only_sock(sk))
155 			return -EAFNOSUPPORT;
156 		err = __ip4_datagram_connect(sk, uaddr, addr_len);
157 		goto ipv4_connected;
158 	}
159 
160 	if (addr_len < SIN6_LEN_RFC2133)
161 		return -EINVAL;
162 
163 	if (usin->sin6_family != AF_INET6)
164 		return -EAFNOSUPPORT;
165 
166 	if (np->sndflow)
167 		fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
168 
169 	addr_type = ipv6_addr_type(&usin->sin6_addr);
170 
171 	if (addr_type == IPV6_ADDR_ANY) {
172 		/*
173 		 *	connect to self
174 		 */
175 		usin->sin6_addr.s6_addr[15] = 0x01;
176 	}
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->sk_bound_dev_if &&
220 			    sk->sk_bound_dev_if != usin->sin6_scope_id) {
221 				err = -EINVAL;
222 				goto out;
223 			}
224 			sk->sk_bound_dev_if = usin->sin6_scope_id;
225 		}
226 
227 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
228 			sk->sk_bound_dev_if = np->mcast_oif;
229 
230 		/* Connect to link-local address requires an interface */
231 		if (!sk->sk_bound_dev_if) {
232 			err = -EINVAL;
233 			goto out;
234 		}
235 	}
236 
237 	sk->sk_v6_daddr = *daddr;
238 	np->flow_label = fl6_flowlabel;
239 
240 	inet->inet_dport = usin->sin6_port;
241 
242 	/*
243 	 *	Check for a route to destination an obtain the
244 	 *	destination cache for it.
245 	 */
246 
247 	err = ip6_datagram_dst_update(sk, true);
248 	if (err)
249 		goto out;
250 
251 	sk->sk_state = TCP_ESTABLISHED;
252 	sk_set_txhash(sk);
253 out:
254 	return err;
255 }
256 EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
257 
258 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
259 {
260 	int res;
261 
262 	lock_sock(sk);
263 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
264 	release_sock(sk);
265 	return res;
266 }
267 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
268 
269 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
270 				 int addr_len)
271 {
272 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
273 	if (sin6->sin6_family != AF_INET6)
274 		return -EAFNOSUPPORT;
275 	return ip6_datagram_connect(sk, uaddr, addr_len);
276 }
277 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
278 
279 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
280 		     __be16 port, u32 info, u8 *payload)
281 {
282 	struct ipv6_pinfo *np  = inet6_sk(sk);
283 	struct icmp6hdr *icmph = icmp6_hdr(skb);
284 	struct sock_exterr_skb *serr;
285 
286 	if (!np->recverr)
287 		return;
288 
289 	skb = skb_clone(skb, GFP_ATOMIC);
290 	if (!skb)
291 		return;
292 
293 	skb->protocol = htons(ETH_P_IPV6);
294 
295 	serr = SKB_EXT_ERR(skb);
296 	serr->ee.ee_errno = err;
297 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
298 	serr->ee.ee_type = icmph->icmp6_type;
299 	serr->ee.ee_code = icmph->icmp6_code;
300 	serr->ee.ee_pad = 0;
301 	serr->ee.ee_info = info;
302 	serr->ee.ee_data = 0;
303 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
304 				  skb_network_header(skb);
305 	serr->port = port;
306 
307 	__skb_pull(skb, payload - skb->data);
308 	skb_reset_transport_header(skb);
309 
310 	if (sock_queue_err_skb(sk, skb))
311 		kfree_skb(skb);
312 }
313 
314 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
315 {
316 	const struct ipv6_pinfo *np = inet6_sk(sk);
317 	struct sock_exterr_skb *serr;
318 	struct ipv6hdr *iph;
319 	struct sk_buff *skb;
320 
321 	if (!np->recverr)
322 		return;
323 
324 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
325 	if (!skb)
326 		return;
327 
328 	skb->protocol = htons(ETH_P_IPV6);
329 
330 	skb_put(skb, sizeof(struct ipv6hdr));
331 	skb_reset_network_header(skb);
332 	iph = ipv6_hdr(skb);
333 	iph->daddr = fl6->daddr;
334 
335 	serr = SKB_EXT_ERR(skb);
336 	serr->ee.ee_errno = err;
337 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
338 	serr->ee.ee_type = 0;
339 	serr->ee.ee_code = 0;
340 	serr->ee.ee_pad = 0;
341 	serr->ee.ee_info = info;
342 	serr->ee.ee_data = 0;
343 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
344 	serr->port = fl6->fl6_dport;
345 
346 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
347 	skb_reset_transport_header(skb);
348 
349 	if (sock_queue_err_skb(sk, skb))
350 		kfree_skb(skb);
351 }
352 
353 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
354 {
355 	struct ipv6_pinfo *np = inet6_sk(sk);
356 	struct ipv6hdr *iph;
357 	struct sk_buff *skb;
358 	struct ip6_mtuinfo *mtu_info;
359 
360 	if (!np->rxopt.bits.rxpmtu)
361 		return;
362 
363 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
364 	if (!skb)
365 		return;
366 
367 	skb_put(skb, sizeof(struct ipv6hdr));
368 	skb_reset_network_header(skb);
369 	iph = ipv6_hdr(skb);
370 	iph->daddr = fl6->daddr;
371 
372 	mtu_info = IP6CBMTU(skb);
373 
374 	mtu_info->ip6m_mtu = mtu;
375 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
376 	mtu_info->ip6m_addr.sin6_port = 0;
377 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
378 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
379 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
380 
381 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
382 	skb_reset_transport_header(skb);
383 
384 	skb = xchg(&np->rxpmtu, skb);
385 	kfree_skb(skb);
386 }
387 
388 /* For some errors we have valid addr_offset even with zero payload and
389  * zero port. Also, addr_offset should be supported if port is set.
390  */
391 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
392 {
393 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
394 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
395 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
396 }
397 
398 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
399  *
400  * At one point, excluding local errors was a quick test to identify icmp/icmp6
401  * errors. This is no longer true, but the test remained, so the v6 stack,
402  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
403  *
404  * Timestamp code paths do not initialize the fields expected by cmsg:
405  * the PKTINFO fields in skb->cb[]. Fill those in here.
406  */
407 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
408 				      struct sock_exterr_skb *serr)
409 {
410 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
411 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
412 		return true;
413 
414 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
415 		return false;
416 
417 	if (!skb->dev)
418 		return false;
419 
420 	if (skb->protocol == htons(ETH_P_IPV6))
421 		IP6CB(skb)->iif = skb->dev->ifindex;
422 	else
423 		PKTINFO_SKB_CB(skb)->ipi_ifindex = skb->dev->ifindex;
424 
425 	return true;
426 }
427 
428 /*
429  *	Handle MSG_ERRQUEUE
430  */
431 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
432 {
433 	struct ipv6_pinfo *np = inet6_sk(sk);
434 	struct sock_exterr_skb *serr;
435 	struct sk_buff *skb;
436 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
437 	struct {
438 		struct sock_extended_err ee;
439 		struct sockaddr_in6	 offender;
440 	} errhdr;
441 	int err;
442 	int copied;
443 
444 	err = -EAGAIN;
445 	skb = sock_dequeue_err_skb(sk);
446 	if (!skb)
447 		goto out;
448 
449 	copied = skb->len;
450 	if (copied > len) {
451 		msg->msg_flags |= MSG_TRUNC;
452 		copied = len;
453 	}
454 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
455 	if (unlikely(err)) {
456 		kfree_skb(skb);
457 		return err;
458 	}
459 	sock_recv_timestamp(msg, sk, skb);
460 
461 	serr = SKB_EXT_ERR(skb);
462 
463 	if (sin && ipv6_datagram_support_addr(serr)) {
464 		const unsigned char *nh = skb_network_header(skb);
465 		sin->sin6_family = AF_INET6;
466 		sin->sin6_flowinfo = 0;
467 		sin->sin6_port = serr->port;
468 		if (skb->protocol == htons(ETH_P_IPV6)) {
469 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
470 								  struct ipv6hdr, daddr);
471 			sin->sin6_addr = ip6h->daddr;
472 			if (np->sndflow)
473 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
474 			sin->sin6_scope_id =
475 				ipv6_iface_scope_id(&sin->sin6_addr,
476 						    IP6CB(skb)->iif);
477 		} else {
478 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
479 					       &sin->sin6_addr);
480 			sin->sin6_scope_id = 0;
481 		}
482 		*addr_len = sizeof(*sin);
483 	}
484 
485 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
486 	sin = &errhdr.offender;
487 	memset(sin, 0, sizeof(*sin));
488 
489 	if (ip6_datagram_support_cmsg(skb, serr)) {
490 		sin->sin6_family = AF_INET6;
491 		if (np->rxopt.all)
492 			ip6_datagram_recv_common_ctl(sk, msg, skb);
493 		if (skb->protocol == htons(ETH_P_IPV6)) {
494 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
495 			if (np->rxopt.all)
496 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
497 			sin->sin6_scope_id =
498 				ipv6_iface_scope_id(&sin->sin6_addr,
499 						    IP6CB(skb)->iif);
500 		} else {
501 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
502 					       &sin->sin6_addr);
503 			if (inet_sk(sk)->cmsg_flags)
504 				ip_cmsg_recv(msg, skb);
505 		}
506 	}
507 
508 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
509 
510 	/* Now we could try to dump offended packet options */
511 
512 	msg->msg_flags |= MSG_ERRQUEUE;
513 	err = copied;
514 
515 	consume_skb(skb);
516 out:
517 	return err;
518 }
519 EXPORT_SYMBOL_GPL(ipv6_recv_error);
520 
521 /*
522  *	Handle IPV6_RECVPATHMTU
523  */
524 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
525 		     int *addr_len)
526 {
527 	struct ipv6_pinfo *np = inet6_sk(sk);
528 	struct sk_buff *skb;
529 	struct ip6_mtuinfo mtu_info;
530 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
531 	int err;
532 	int copied;
533 
534 	err = -EAGAIN;
535 	skb = xchg(&np->rxpmtu, NULL);
536 	if (!skb)
537 		goto out;
538 
539 	copied = skb->len;
540 	if (copied > len) {
541 		msg->msg_flags |= MSG_TRUNC;
542 		copied = len;
543 	}
544 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
545 	if (err)
546 		goto out_free_skb;
547 
548 	sock_recv_timestamp(msg, sk, skb);
549 
550 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
551 
552 	if (sin) {
553 		sin->sin6_family = AF_INET6;
554 		sin->sin6_flowinfo = 0;
555 		sin->sin6_port = 0;
556 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
557 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
558 		*addr_len = sizeof(*sin);
559 	}
560 
561 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
562 
563 	err = copied;
564 
565 out_free_skb:
566 	kfree_skb(skb);
567 out:
568 	return err;
569 }
570 
571 
572 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
573 				 struct sk_buff *skb)
574 {
575 	struct ipv6_pinfo *np = inet6_sk(sk);
576 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
577 
578 	if (np->rxopt.bits.rxinfo) {
579 		struct in6_pktinfo src_info;
580 
581 		if (is_ipv6) {
582 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
583 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
584 		} else {
585 			src_info.ipi6_ifindex =
586 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
587 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
588 					       &src_info.ipi6_addr);
589 		}
590 
591 		if (src_info.ipi6_ifindex >= 0)
592 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
593 				 sizeof(src_info), &src_info);
594 	}
595 }
596 
597 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
598 				    struct sk_buff *skb)
599 {
600 	struct ipv6_pinfo *np = inet6_sk(sk);
601 	struct inet6_skb_parm *opt = IP6CB(skb);
602 	unsigned char *nh = skb_network_header(skb);
603 
604 	if (np->rxopt.bits.rxhlim) {
605 		int hlim = ipv6_hdr(skb)->hop_limit;
606 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
607 	}
608 
609 	if (np->rxopt.bits.rxtclass) {
610 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
611 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
612 	}
613 
614 	if (np->rxopt.bits.rxflow) {
615 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
616 		if (flowinfo)
617 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
618 	}
619 
620 	/* HbH is allowed only once */
621 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
622 		u8 *ptr = nh + sizeof(struct ipv6hdr);
623 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
624 	}
625 
626 	if (opt->lastopt &&
627 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
628 		/*
629 		 * Silly enough, but we need to reparse in order to
630 		 * report extension headers (except for HbH)
631 		 * in order.
632 		 *
633 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
634 		 * (and WILL NOT be) defined because
635 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
636 		 */
637 		unsigned int off = sizeof(struct ipv6hdr);
638 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
639 
640 		while (off <= opt->lastopt) {
641 			unsigned int len;
642 			u8 *ptr = nh + off;
643 
644 			switch (nexthdr) {
645 			case IPPROTO_DSTOPTS:
646 				nexthdr = ptr[0];
647 				len = (ptr[1] + 1) << 3;
648 				if (np->rxopt.bits.dstopts)
649 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
650 				break;
651 			case IPPROTO_ROUTING:
652 				nexthdr = ptr[0];
653 				len = (ptr[1] + 1) << 3;
654 				if (np->rxopt.bits.srcrt)
655 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
656 				break;
657 			case IPPROTO_AH:
658 				nexthdr = ptr[0];
659 				len = (ptr[1] + 2) << 2;
660 				break;
661 			default:
662 				nexthdr = ptr[0];
663 				len = (ptr[1] + 1) << 3;
664 				break;
665 			}
666 
667 			off += len;
668 		}
669 	}
670 
671 	/* socket options in old style */
672 	if (np->rxopt.bits.rxoinfo) {
673 		struct in6_pktinfo src_info;
674 
675 		src_info.ipi6_ifindex = opt->iif;
676 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
677 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
678 	}
679 	if (np->rxopt.bits.rxohlim) {
680 		int hlim = ipv6_hdr(skb)->hop_limit;
681 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
682 	}
683 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
684 		u8 *ptr = nh + sizeof(struct ipv6hdr);
685 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
686 	}
687 	if (np->rxopt.bits.odstopts && opt->dst0) {
688 		u8 *ptr = nh + opt->dst0;
689 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
690 	}
691 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
692 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
693 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
694 	}
695 	if (np->rxopt.bits.odstopts && opt->dst1) {
696 		u8 *ptr = nh + opt->dst1;
697 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
698 	}
699 	if (np->rxopt.bits.rxorigdstaddr) {
700 		struct sockaddr_in6 sin6;
701 		__be16 *ports = (__be16 *) skb_transport_header(skb);
702 
703 		if (skb_transport_offset(skb) + 4 <= skb->len) {
704 			/* All current transport protocols have the port numbers in the
705 			 * first four bytes of the transport header and this function is
706 			 * written with this assumption in mind.
707 			 */
708 
709 			sin6.sin6_family = AF_INET6;
710 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
711 			sin6.sin6_port = ports[1];
712 			sin6.sin6_flowinfo = 0;
713 			sin6.sin6_scope_id =
714 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
715 						    opt->iif);
716 
717 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
718 		}
719 	}
720 }
721 
722 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
723 			  struct sk_buff *skb)
724 {
725 	ip6_datagram_recv_common_ctl(sk, msg, skb);
726 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
727 }
728 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
729 
730 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
731 			  struct msghdr *msg, struct flowi6 *fl6,
732 			  struct ipcm6_cookie *ipc6, struct sockcm_cookie *sockc)
733 {
734 	struct in6_pktinfo *src_info;
735 	struct cmsghdr *cmsg;
736 	struct ipv6_rt_hdr *rthdr;
737 	struct ipv6_opt_hdr *hdr;
738 	struct ipv6_txoptions *opt = ipc6->opt;
739 	int len;
740 	int err = 0;
741 
742 	for_each_cmsghdr(cmsg, msg) {
743 		int addr_type;
744 
745 		if (!CMSG_OK(msg, cmsg)) {
746 			err = -EINVAL;
747 			goto exit_f;
748 		}
749 
750 		if (cmsg->cmsg_level == SOL_SOCKET) {
751 			err = __sock_cmsg_send(sk, msg, cmsg, sockc);
752 			if (err)
753 				return err;
754 			continue;
755 		}
756 
757 		if (cmsg->cmsg_level != SOL_IPV6)
758 			continue;
759 
760 		switch (cmsg->cmsg_type) {
761 		case IPV6_PKTINFO:
762 		case IPV6_2292PKTINFO:
763 		    {
764 			struct net_device *dev = NULL;
765 
766 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
767 				err = -EINVAL;
768 				goto exit_f;
769 			}
770 
771 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
772 
773 			if (src_info->ipi6_ifindex) {
774 				if (fl6->flowi6_oif &&
775 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
776 					return -EINVAL;
777 				fl6->flowi6_oif = src_info->ipi6_ifindex;
778 			}
779 
780 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
781 
782 			rcu_read_lock();
783 			if (fl6->flowi6_oif) {
784 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
785 				if (!dev) {
786 					rcu_read_unlock();
787 					return -ENODEV;
788 				}
789 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
790 				rcu_read_unlock();
791 				return -EINVAL;
792 			}
793 
794 			if (addr_type != IPV6_ADDR_ANY) {
795 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
796 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
797 				    !ipv6_chk_addr(net, &src_info->ipi6_addr,
798 						   strict ? dev : NULL, 0) &&
799 				    !ipv6_chk_acast_addr_src(net, dev,
800 							     &src_info->ipi6_addr))
801 					err = -EINVAL;
802 				else
803 					fl6->saddr = src_info->ipi6_addr;
804 			}
805 
806 			rcu_read_unlock();
807 
808 			if (err)
809 				goto exit_f;
810 
811 			break;
812 		    }
813 
814 		case IPV6_FLOWINFO:
815 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
816 				err = -EINVAL;
817 				goto exit_f;
818 			}
819 
820 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
821 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
822 					err = -EINVAL;
823 					goto exit_f;
824 				}
825 			}
826 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
827 			break;
828 
829 		case IPV6_2292HOPOPTS:
830 		case IPV6_HOPOPTS:
831 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
832 				err = -EINVAL;
833 				goto exit_f;
834 			}
835 
836 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
837 			len = ((hdr->hdrlen + 1) << 3);
838 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
839 				err = -EINVAL;
840 				goto exit_f;
841 			}
842 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
843 				err = -EPERM;
844 				goto exit_f;
845 			}
846 			opt->opt_nflen += len;
847 			opt->hopopt = hdr;
848 			break;
849 
850 		case IPV6_2292DSTOPTS:
851 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
852 				err = -EINVAL;
853 				goto exit_f;
854 			}
855 
856 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
857 			len = ((hdr->hdrlen + 1) << 3);
858 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
859 				err = -EINVAL;
860 				goto exit_f;
861 			}
862 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
863 				err = -EPERM;
864 				goto exit_f;
865 			}
866 			if (opt->dst1opt) {
867 				err = -EINVAL;
868 				goto exit_f;
869 			}
870 			opt->opt_flen += len;
871 			opt->dst1opt = hdr;
872 			break;
873 
874 		case IPV6_DSTOPTS:
875 		case IPV6_RTHDRDSTOPTS:
876 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
877 				err = -EINVAL;
878 				goto exit_f;
879 			}
880 
881 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
882 			len = ((hdr->hdrlen + 1) << 3);
883 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
884 				err = -EINVAL;
885 				goto exit_f;
886 			}
887 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
888 				err = -EPERM;
889 				goto exit_f;
890 			}
891 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
892 				opt->opt_flen += len;
893 				opt->dst1opt = hdr;
894 			} else {
895 				opt->opt_nflen += len;
896 				opt->dst0opt = hdr;
897 			}
898 			break;
899 
900 		case IPV6_2292RTHDR:
901 		case IPV6_RTHDR:
902 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
903 				err = -EINVAL;
904 				goto exit_f;
905 			}
906 
907 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
908 
909 			switch (rthdr->type) {
910 #if IS_ENABLED(CONFIG_IPV6_MIP6)
911 			case IPV6_SRCRT_TYPE_2:
912 				if (rthdr->hdrlen != 2 ||
913 				    rthdr->segments_left != 1) {
914 					err = -EINVAL;
915 					goto exit_f;
916 				}
917 				break;
918 #endif
919 			default:
920 				err = -EINVAL;
921 				goto exit_f;
922 			}
923 
924 			len = ((rthdr->hdrlen + 1) << 3);
925 
926 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
927 				err = -EINVAL;
928 				goto exit_f;
929 			}
930 
931 			/* segments left must also match */
932 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
933 				err = -EINVAL;
934 				goto exit_f;
935 			}
936 
937 			opt->opt_nflen += len;
938 			opt->srcrt = rthdr;
939 
940 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
941 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
942 
943 				opt->opt_nflen += dsthdrlen;
944 				opt->dst0opt = opt->dst1opt;
945 				opt->dst1opt = NULL;
946 				opt->opt_flen -= dsthdrlen;
947 			}
948 
949 			break;
950 
951 		case IPV6_2292HOPLIMIT:
952 		case IPV6_HOPLIMIT:
953 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
954 				err = -EINVAL;
955 				goto exit_f;
956 			}
957 
958 			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
959 			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
960 				err = -EINVAL;
961 				goto exit_f;
962 			}
963 
964 			break;
965 
966 		case IPV6_TCLASS:
967 		    {
968 			int tc;
969 
970 			err = -EINVAL;
971 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
972 				goto exit_f;
973 
974 			tc = *(int *)CMSG_DATA(cmsg);
975 			if (tc < -1 || tc > 0xff)
976 				goto exit_f;
977 
978 			err = 0;
979 			ipc6->tclass = tc;
980 
981 			break;
982 		    }
983 
984 		case IPV6_DONTFRAG:
985 		    {
986 			int df;
987 
988 			err = -EINVAL;
989 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
990 				goto exit_f;
991 
992 			df = *(int *)CMSG_DATA(cmsg);
993 			if (df < 0 || df > 1)
994 				goto exit_f;
995 
996 			err = 0;
997 			ipc6->dontfrag = df;
998 
999 			break;
1000 		    }
1001 		default:
1002 			net_dbg_ratelimited("invalid cmsg type: %d\n",
1003 					    cmsg->cmsg_type);
1004 			err = -EINVAL;
1005 			goto exit_f;
1006 		}
1007 	}
1008 
1009 exit_f:
1010 	return err;
1011 }
1012 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1013 
1014 void ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1015 			     __u16 srcp, __u16 destp, int bucket)
1016 {
1017 	const struct in6_addr *dest, *src;
1018 
1019 	dest  = &sp->sk_v6_daddr;
1020 	src   = &sp->sk_v6_rcv_saddr;
1021 	seq_printf(seq,
1022 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1023 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1024 		   bucket,
1025 		   src->s6_addr32[0], src->s6_addr32[1],
1026 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1027 		   dest->s6_addr32[0], dest->s6_addr32[1],
1028 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1029 		   sp->sk_state,
1030 		   sk_wmem_alloc_get(sp),
1031 		   sk_rmem_alloc_get(sp),
1032 		   0, 0L, 0,
1033 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1034 		   0,
1035 		   sock_i_ino(sp),
1036 		   atomic_read(&sp->sk_refcnt), sp,
1037 		   atomic_read(&sp->sk_drops));
1038 }
1039